The AKT isoforms 1 and 2 drive B cell fate decisions during the germinal center response

B cell–intrinsic AKT1/2 are essential for germinal center formation, antibody production and affinity maturation, and plasma cell differentiation.

The PI3K/Akt/Foxo1 pathway is a prominent signaling hub in B cells and is particularly relevant during immune responses. It is also altered in B cell lymphomas by different types of genetic alterations, and thus of great interest. In this manuscript, Zhu and colleagues dissect the role of different isoforms of the Akt kinase in mature B cells and particularly germinal center responses, combining different conditional alleles of Akt 1, 3 and Foxo1. They also include some partial data on the impact of FOXO1 T24A mutations in these responses, which occurs in a subset of germinal center-derived B cell non-Hodgkin lymphomas. The authors find that signaling through Akt1/2 is essential to mount successful germinal center responses. They show, in ex vivo and in vivo studies, that the activity of these enzymes is necessary for germinal center formation, immunoglobulin class switch recombination and affinity maturation of antibodies. They also find that these enzymes are important in the signaling cascade determining plasma cell differentiation. Most of these effects are due to defects in B cell proliferation, survival and metabolic fitness upon B cell receptor stimulation. Interestingly, many of these defects can be rescued (in the context of germinal center B cells) by co-stimulation through the CD40 receptor, a signal also essential for germinal center formation and homeostasis. The authors provide some results explaining how CD40 is able to correct these defects at the molecular level.
There is significant interest in better understanding the roles of the PI3K/Akt/FOXO1 pathway in germinal center biology, given its essentiality to the biology of this compartment and its suspected relevance during lymphomagenesis. Thus, the results reported in this manuscript can be of interest to a wide range of readers. The added description of a couple of new, original mouse models can also be of interest to many. However, a few points would need to be addressed experimentally to help better interpret some of the key findings in the manuscript. Addressing these points would also add conceptual clarity to some of the points discussed in the manuscript.
>Main Points: 1-Mentions to "control" mice in all figures and throughout the manuscript is a bit confusing. For some experiments, Cgamma1-Cre only mice are used as controls, while in others the authors use Cg1-Cre x Akt2-/-mice. This seems important since, based on Fig. EV1-A, Cg1-Cre Akt2-/-mice show statistically significant defects in, for example, their ability to generate high affinity antibodies. A more homogeneous set of controls or, at least, clear specification of the exact genotype of the animals used as controls in each experiment both in the figures and figure legends should be provided (it is not the case, now). Figure 4: Some observations in the behavior of Foxo1 wt and T24A proteins are unexpected and may require some argumentation. Concurrent phosphorylation of Foxo1 T24 and S256 is expected to recruit 14-3-3 proteins and result in Foxo1 nuclear export. The authors show that this seems to be observed as p-T24 + p-S256 Foxo1 is depleted from nuclear fractions at 30 minutes post-stimulation, although no changes in the cytosolic fraction or total Foxo1 are observed (the authors imply degradation of cytosolic foxo1). However, similar oscillations in Foxo1 levels in the nuclear fraction occur with the T24A/T24A mutant. Note that the levels at 30 minutes are clearly reduced when compared to those at 10 minutes post-stimulation, and the dynamics of total Foxo1 protein are not different from those in wildtype B cells. These results are confusing, at least as how they relate to the authors' interpretation.

2-
3-One important finding in this study is that CD40 activation can rescue key deficiencies of the response of Akt1/2 ko B cells to BCR stimulation. CD40 activation also rescues in significant degree the defects in GC formation, including Tfh numbers. In GC B cells, mTOR activity is dependent on mTOR activation (Ersching et al, Immmunity 2017;46(6)). This activity is essential to support the accumulation of biomass in affinity selected B cells. Notably, the authors show in Fig.6 that an important defect in Akt1/2 ko B cells is a severe reduction in the levels of pS6 in response to BCR stimulation. Thus one could reason that the ability of CD40 to rescue the defects of Akt1/2 ko B cells would be due to a "restoration" of mTOR activity. However, pS6 levels have not been analyzed in the experiments shown in Fig.8, despite the severe reduction in pS6 levels in Akt1/2 ko B cells (Fig.6). This appears to be an important piece of missing evidence. 4-Cyclin D2 is a MYC transcriptional target, and its induction appears to be dependent on the activation of PI3K/Akt signaling and FOXO inactivation (Bouchard et al, EMBO J 2004;23(14)). Thus, the inability of Akt1/2 ko B cells to activate Cyclin D2 expression, regardless of stimulation and despite the increase in MYC levels, doesn't seem surprising. CD40 stimulation does not seem to rescue these defects, despite upon addition of CD40, Akt1/2 ko B cells are able to divide. The authors suggest that rescue is dependent on Cyclin D3, but no evidence is shown. Are the levels of Cyclin D3 protein or RNA in Akt1/2 B cells changing in response to CD40 stimulation? 5-What occurs to FOXO in these experiments (Fig.8)?
6-In the discussion, the authors mention unpublished results describing enhanced proliferation of Foxo1 T24A B cells in response to CD40 and BCR stimulation. Although this may be consistent with previous studies on the potential role of these mutations in supporting B cell survival and proliferation (Trihn et al, 2013;Kabrani et al, 2018), do not seem to fit with the phenotypes of Akt1/2 ko B cells (which are expected to have constitutively active, non-phosphorylated nuclear Foxo1 but instead, fail to proliferate in response to BCR stimulation). These observations are difficult to interpret in light of the findings reported in this manuscript.
>Additional points: 7-The immunoblot results in Fig. EV2-A suggest that loss of Akt2 (maybe also Akt3) may impact the overall levels of Akt1. Only one sample per genotype is included in this analysis, but this may be important, since Akt2-/-mice seem to actually have a significant defect in antibody affinity responses ( Fig. EV1) 8- Figure 4B: Two distinct populations of IgG1+ cells are observed in the CSR ex vivo experiments. Does this correspond to differences in the balance between surface and intracellular IgG1 protein? 9- Fig. 4D: The correlation that the authors imply between AID levels and the fraction of cells undergoing CSR does not seem to correspond to the experimental results. Specifically, the levels of AID protein in Foxo1 KO B cells are higher than those in WT or WT/T24A cells (yet, Foxo1 KO cells do not switch); and the levels in WT/T24A cells are not different from those in WT/WT cells, yet they show clear differences in the percentage of IgG1 positive cells. How is this explained? 10- Fig. 5B: The histogram depicting the number of cell divisions in LPS-stimulated B cells would suggest that all B cells in this experiment divided synchronously. Commonly, entry of B cells into cycle upon any cytokine stimulation is asynchronous. Are these results reproducible? 11- Fig. 5D: How many biological replicates for this experiment? 12-In Fig.6B the authors also show that Akt1/2 ko B cells also fail to induce expression of Mcl-1 in response to BCR activation. Loss of mTORC1 signaling and loss of Mcl1-1 expression have been previously connected and suggest that Mcl-1 expression is sensitive to metabolic states (Coloff et al, Cancer Res 2011;71(15) and Mills et al, PNAS 2008;105(31)). Does Mcl-1 expression change upon addition of CD40? (As shown in Fig.8, CD40 corrects some of the metabolic changes in Akt1/2 ko cells in response to BCR stimulation).
13-Previous studies indicate that an important role for IRF4 during plasma cell differentiation is the regulation of BLIMP1 expression. IRF4 also controls expression of AID (Sciammas et al, Immunity 2006;25(5)). In Fig.5D, the authors show that while IRF4 levels are increased in Foxo1 ko B cells after LPS stimulation, this does not correlate with an increase in BLIMP1, despite a higher number of B cells appear to adopt a plasma cell phenotype. Can this be reconciled with those previous studies? How do cells become plasma cells in greater numbers without observing an increase in BLIMP1?
14-Why did the authors use CD19-Cre only B cells as controls in the mitochondrial stress studies (seahorse) depicted in Fig. 8F? 1 Referee #2: The manuscript entitled "The AKT isoforms 1 and 2 drive B cell fate decisions during the germinal center response" shows that mice lacking AKT1/2 isoforms in B cells failed to maintain GCs, consequently mice lack antibody titers (probably because they are not producing antibody-forming cells) and reduced affinity maturation. Data show AKT1/2 isoforms are needed to induce proliferation and survival of B cells after BCR stimulation. Data suggest that the mechanisms behind these findings rely on FOXO1 inactivation promoting IRF4-driven PC differentiation. Lastly, CD40 stimulation rescue B cells by restoring proliferative expansion and energy production.

Main Points:
Comment 1: Although introduction nicely summarized supporting literature, this reviewer thinks introduction can be shortened.

Answer:
According to your suggestion，we shortened the introduction from 1168 to 872 words.

Comment 2:
Authors discussed GC reduction is mainly on LZ phase of differentiation. An experiment to support this statement is to look for DZ/LZ GC B cells. If authors' assumption is true, DZ GC B cells will be normal while LZ GC B cells will be reduced.

Answer:
Thank you very much for your comments. According to your suggestion，we performed experiments to examine DZ and LZ B cells in SRBC-immunized WT Control (C1 Cre ) and C1 Cre x Akt1 f/f x Akt2 -/mice. We found that LZ GC B cells were reduced to a greater extent than DZ GC B cells in AKT1/2 KO mice which leaded to a relatively higher DZ/LZ ratio in AKT1/2 KO mice than that in WT Control mice.  ). Two-tailed t tests were used to test statistical significance for (B). Symbols represent individual mice studied. Error bars represent mean ± SEM. ***, P < 0.001.

Comment 3:
Another possibility is that GC B cells are dying as a consequence of AKT1/2 deletion. Then, they are not able to maintain their proliferation or even mature the affinity. Because there are not cells surviving after AKT1/2 deletion, it would be expected affinity maturation will be dampened. Can authors show the in vivo viability/proliferation of GC B cells perhaps by means of active caspase-3 staining/EdU incorporation or intracellular staining with Ki67? That would show if cells are more prone to die after AKT1/2 deletion.

Answer:
Thank you very much for your comments. In order to rule out the possibility that cell viability influences GC B cell proliferation or even the affinity in Akt1/2 KO mice, we crossed the CD19 Cre x Akt1 f/f x Akt2 -/mice with Bcl2 transgenic mice. Our results showed that enforced Bcl2 expression failed to rescue the loss of AKT1/2-deficient GC B cells in vivo, suggesting the loss of AKT1/2-deficient GC B cells is not due to the inability of these cells to transmit pro-survival signals. The data can be found in Representative FACS plots for GC B cell proliferation, as indicated by BrdU incorporation (left panel) and the frequencies of B220 + BrdU + cells in total GC B cells (right panel) were shown. Two-tailed t tests were used to test statistical significance for (A and B). Symbols represent individual mice studied. Error bars represent mean ± SEM. ***, P < 0.001.

Comment 4:
It has been proposed memory B cells appear early after GC onset, e.g. day 7 after immunization with NP-CGG. Have you evaluated earlier time points to see if these differences exist or if something that is happening late during the GC reaction?

Answer:
Thank you very much for your comments. According to your suggestion，we performed experiments using the approach we previously developed to examine SRBC-specific memory B cells as early as 7 days postimmunization. and the frequencies of SRBC-specific memory B cells (B220 + eFluor670 + GL7 -CD80 + ) as a percentage of total SRBC-specific B cells in the spleen (right panel) were shown. Two-tailed t tests were used to test statistical 4 significance for (A and B). Symbols represent individual mice studied. Error bars represent mean ± SEM. **, P < 0.01; ***, P < 0.001.

Comment 5:
Affinity maturation increases with time, did you evaluate later time points e.g. d21? Affinity reduction correlates with the absence of GC B cells, the actual antibody-producing cells won't be produced if GC B cells are absent. If titers are quite low, it is rather difficult to calculate the ratio NP4/NP23. Can you please show the curves for titers calculation (possibly on supplementary figures)?

Answer:
Thank you very much for your comments. According to your suggestion, we performed experiments to examine the NP23-specific IgG and IgM tires in the serum of NP25-CGG-immunized WT Control (C1 Cre ) and C1 Cre x Akt1 f/f x Akt2 -/mice on D7, D14, and D21 postimmunization. We found that the titers of serum NP23-specific IgG and IgM were low on D7 postimmunization. The tires increased with time and peaked on D21.
Whereas the titers of NP23-specific IgM were comparable between WT Control (C1 Cre ) and C1 Cre x Akt1 f/f x Akt2 -/mice, the titers of NP23-specific IgG were less in C1 Cre x Akt1 f/f x Akt2 -/mice than those in WT Control (C1 Cre ) mice on D14 and D21 postimmunization. We did examine the titers of serum NP-specific IgG and IgM from the mice in Fig 2 on D21 postimmunization. Since the titers of NP-specific IgG slightly increased on D21 compared to those on D14 in C1 Cre x Akt1 f/f x Akt2 -/mice and the results on D21 were consistent to those on D14, so we just showed the results on D14 as a representative in Fig 2. We are happy to provide the data on D21 if requested. Loss of AKT1/2 inhibits the production of NP23-specific IgG. Mice of each genotype (n=9) were immunized i.p. with NP25-CGG and the serum were collected on D7, D14, and D21 postimmunization. The titers of NP23-specific IgG (left panel) and IgM (right panel) were measured by ELISA. Statistical analysis was done with two-way ANOVA. Error bars represent mean ± SEM. ***, P < 0.001.

Comment 6:
Authors conclude "these findings demonstrate that AKT1/2 are required for antigen-specific IgG1 affinity maturation". Although affinity maturation seems reduced, the direct link between AKT1/2 deletion and affinity maturation mechanisms is still missing. How AKT1/2 deficiency is interfering with the SHM? Could authors speculate about this? 5 Answer: Thank you very much for your comments. SHM has been considered to be an event at DNA level. Cell proliferation can provide a large amount DNA templates for SHM. Given that a critical role of AKT1/2 in GC B cell proliferation, we speculate that loss of AKT1/2 inhibits SHM-mediated antibody affinity maturation. We mentioned this in the discussion.

Comment 7:
If AKT1/2 absence is inducing GC B cells death, then reduced affinity maturation is not a direct effect of AKT deficiency. Have the authors assessed the production of NP-specific plasma cells?

Answer:
Thank you very much for your comments. We agree that AKT1/2 are important for the survival of GC B cells. Because enforced Bcl2 expression didn't rescue the loss of AKT1/2-deficient GC B cells in vivo, we think that reduced affinity maturation is largely caused by impaired cell proliferation in AKT1/2-deficient GC B cells. In the 2nd and 3rd paragraphs of discussion, we explained the possibilities how cell proliferation affects the antibody affinity maturation and GC positive selection. According to your suggestion, we performed experiments to assess the production of NP-specific plasma cells. We found that the frequencies of NP-specific plasma cells were significantly reduced in AKT1/2 KO mice compared to that in WT Ctrl mice 21 days after NP-CGG immunization. The frequencies of NP-specific plasma cells (NP + B220 lo CD138 + ) as a percentage of total NP-specific cells in the spleen were shown. Two-tailed t tests were used to test statistical significance for (B). Symbols represent individual mice studied. Error bars represent mean ± SEM. ***, P < 0.001.

Comment 8:
6 Can author show the summary of data on figure 3B?

Answer:
According to your suggestion, we summarized the data in Fig 3A and B.

Comment 9:
Actually, all B cells were rescued with in vivo anti-CD40 administration (Fig 8A.). Why is this happening? If this is an effect on all B cells, then it is not exclusive for GC B cells rescue. An experiment that might show the effect on rescuing GC B cells exclusively could be in vivo anti-CD40 administration on the Cg1CrexAkt1f/fxAkt2-/-, where only GC B cells are affected.

Answer:
Thank you very much for your comments. We think that the rescue of B cells by in vivo anti-CD40 administration could partly be caused by upregulation of Mcl-1 (an antiapoptotic protein of the Bcl2 family) expression by anti-CD40 stimulation. The data can be found in Fig

Answer:
Thank you very much for your suggestions. We specified "control" mice in manuscripts.
For example, if C1 Cre only mice are used as controls, we specify them as "WT Ctrl (C1 Cre )" in both figures and figure legends. We also added and mentioned that "Heterozygous Cre recombinase mice were used in experiments" in the Materials and Methods.

Comment 11:
1D. Could you please mention in the figure legend, how many events are in the displayed dot plot? It seems the frequency of cells is really few on the C1cre x Aktf/f x Akt2-/-.

Answer:
According to your suggestion, we added the numbers of events to be displayed in the Dot Plot graph in the legends of Fig 1D. Comment 12:

Fig 3C. Have you analyzed frequencies of IgG1-producing cells by FACS?
Answer: Yes, we did surface IgG1 staining on the GC B cells in the SRBC-immunized mice by FACS. The results showed that AKT1-deficient mice had more IgG1 expression on GC B cells than AKT2 or AKT3-deficient mice. The results were consistent to the serum SRBC-specific IgG1 production in those mice in Fig 3C. Because of the limited space, we did not show these data. We are happy to provide them if requested.

Referee #3
The PI3K/Akt/Foxo1 pathway is a prominent signaling hub in B cells and is particularly relevant during immune responses. It is also altered in B cell lymphomas by different types of genetic alterations, and thus of great interest. In this manuscript, Zhu and colleagues dissect the role of different isoforms of the Akt kinase in mature B cells and particularly germinal center responses, combining different conditional alleles of Akt 1, 3 and Foxo1. They also include some partial data on the impact of FOXO1 T24A mutations in these responses, which occurs in a subset of germinal center-derived B cell non-Hodgkin lymphomas. The authors find that signaling through Akt1/2 is essential to mount successful germinal center responses. They show, in ex vivo and in vivo studies, that the activity of these enzymes is necessary for germinal center formation, immunoglobulin class switch recombination and affinity maturation of antibodies. They also find that these enzymes are important in the signaling cascade determining plasma 8 cell differentiation. Most of these effects are due to defects in B cell proliferation, survival and metabolic fitness upon B cell receptor stimulation. Interestingly, many of these defects can be rescued (in the context of germinal center B cells) by co-stimulation through the CD40 receptor, a signal also essential for germinal center formation and homeostasis. The authors provide some results explaining how CD40 is able to correct these defects at the molecular level.
There is significant interest in better understanding the roles of the PI3K/Akt/FOXO1 pathway in germinal center biology, given its essentiality to the biology of this compartment and its suspected relevance during lymphomagenesis. Thus, the results reported in this manuscript can be of interest to a wide range of readers. The added description of a couple of new, original mouse models can also be of interest to many. However, a few points would need to be addressed experimentally to help better interpret some of the key findings in the manuscript. Addressing these points would also add conceptual clarity to some of the points discussed in the manuscript.

Main Points:
Comment 1: Mentions to "control" mice in all figures and throughout the manuscript is a bit confusing.
For some experiments, Cgamma1-Cre only mice are used as controls, while in others the authors use Cg1-Cre x Akt2-/-mice. This seems important since, based on Fig EV1-A, Cg1-Cre Akt2-/-mice show statistically significant defects in, for example, their ability to generate high affinity antibodies. A more homogeneous set of controls or, at least, clear specification of the exact genotype of the animals used as controls in each experiment both in the figures and figure legends should be provided (it is not the case, now).

Answer:
Thank you very much for your suggestions. We specified "control" mice in manuscripts.
For example, if C1 Cre only mice are used as controls, we specify them as "WT Ctrl (C1 Cre )" in both figures and figure legends. We also added and mentioned that "Heterozygous Cre recombinase mice were used in experiments" in the Materials and Methods.
Comment 2: Figure 4: Some observations in the behavior of Foxo1 wt and T24A proteins are unexpected and may require some argumentation. Concurrent phosphorylation of Foxo1 T24 and S256 is expected to recruit 14-3-3 proteins and result in Foxo1 nuclear export. The authors show that this seems to be observed as p-T24 + p-S256 Foxo1 is depleted from nuclear fractions at 30 minutes post-stimulation, although no changes in the cytosolic fraction or total Foxo1 are observed (the authors imply degradation of cytosolic foxo1). However, similar oscillations in Foxo1 levels in the nuclear fraction occur with the T24A/T24A mutant. Note that the levels at 30 minutes are clearly reduced when compared to those at 10 minutes post-stimulation, and the dynamics of total Foxo1 protein are not 9 different from those in wildtype B cells. These results are confusing, at least as how they relate to the authors' interpretation.

Answer:
Thank you very much for your comments. We think that unchanged total FOXO1 protein levels in Foxo1 T24A/T24A mutant B cells could be due to two possible reasons. First, the depletion of p-FOXO1 may not be reflected in total FOXO1 protein because only a small subset of total FOXO1 underwent phosphorylation and degradation upon anti-BCR and anti-CD40 stimulation. This is the case we found in Fig S8. We observed that the total FOXO1 levels were not changed very much although p-FOXO1 (T24) and p-FOXO1 (S256) were largely depleted after anti-BCR and anti-CD40 stimulation for 24 h. Second, the duration of stimulation matters. Short time anti-BCR or anti-CD40 stimulation (30 min) may not be long enough to observe changes in total FOXO1 levels. By contrast, we found that long time anti-CD40 stimulation (96 h) and LPS stimulation (72 h) can significantly decrease total FOXO1 levels ( Fig 4D and Fig 5E).

Comment 3:
One important finding in this study is that CD40 activation can rescue key deficiencies of the response of Akt1/2 ko B cells to BCR stimulation. CD40 activation also rescues in significant degree the defects in GC formation, including Tfh numbers. In GC B cells, mTOR activity is dependent on mTOR activation (Ersching et al, Immmunity 2017; 46 (6)). This activity is essential to support the accumulation of biomass in affinity selected B cells.
Notably, the authors show in Fig6 that an important defect in Akt1/2 ko B cells is a severe reduction in the levels of pS6 in response to BCR stimulation. Thus one could reason that the ability of CD40 to rescue the defects of Akt1/2 ko B cells would be due to a "restoration" of mTOR activity. However, pS6 levels have not been analyzed in the experiments shown in Fig8, despite the severe reduction in pS6 levels in Akt1/2 ko B cells (Fig 6). This appears to be an important piece of missing evidence.

Answer:
Thank you very much for your comments. According to your suggestion, we performed experiments to analyze p-S6 (S240/244) levels in the experiments shown in Fig 8. We found that anti-CD40 or anti-BCR stimulation alone poorly induced p-S6 (S240/244) expression in AKT1/2 KO B cells compared to that in WT B cells. Combined anti-CD40 and anti-BCR stimulation strongly and synergistically induced p-S6 (S240/244) in AKT1/2 KO B cells. The new data are shown in Fig S7.

Comment 4:
Cyclin D2 is a MYC transcriptional target, and its induction appears to be dependent on the activation of PI3K/Akt signaling and FOXO inactivation (Bouchard et al, EMBO J 2004;23(14)

Answer:
According to your suggestion, we performed experiments to analyze the levels of total FOXO1, and p-FOXO1 (T24), and p-FOXO1 (S256) in the experiments shown in Fig 8. We found that baseline levels of both p-FOXO1 (T24) and p-FOXO1 (S256) were less in AKT1/2 KO B cells than those in WT B cells. After combined anti-CD40 and anti-BCR stimulation for 24 h, the levels of both p-FOXO1 (T24) and p-FOXO1 (S256) were significantly reduced in both WT and AKT1/2 KO B cells. While the levels of total FOXO1 in AKT1/2 KO B cells were not clearly affected, interestingly, we observed that the levels of total FOXO1 in WT B cells were slightly reduced probably due to protease-mediated degradation indicated by smeared bands. The new data are shown in Fig S8.

Comment 6:
In the discussion, the authors mention unpublished results describing enhanced proliferation of Foxo1 T24A B cells in response to CD40 and BCR stimulation. Although this may be consistent with previous studies on the potential role of these mutations in supporting B cell survival and proliferation (Trihn et al, 2013;Kabrani et al, 2018), do not seem to fit with the phenotypes of Akt1/2 ko B cells (which are expected to have constitutively active, non-phosphorylated nuclear Foxo1 but instead, fail to proliferate in response to BCR stimulation). These observations are difficult to interpret in light of the findings reported in this manuscript.

Answer:
Thank you very much for your comments. We agree that AKT1/2 KO B cells have constitutively active, non-phosphorylated nuclear FOXO1. However, FOXO1 is just one of many downstream targets of the PI3K/AKT signaling pathway. So, it is possible that the phenotypes of AKT1/2 KO B cells are different from those of Foxo1 T24A B cells.
Additional points:

Comment 7:
The immunoblot results in Fig EV2-A suggest that loss of Akt2 (maybe also Akt3) may impact the overall levels of Akt1. Only one sample per genotype is included in this analysis, but this may be important, since Akt2-/-mice seem to actually have a significant defect in antibody affinity responses ( Fig EV1) Answer: Thank you very much for your comments. Fig EV2 is now Fig S2 in revised manuscript. We agree that one AKT isoform is possible to compensate for the loss of the other(s) which may reveal the functional interchangeability of AKT isoforms.
Comment 8: Figure 4B: Two distinct populations of IgG1+ cells are observed in the CSR ex vivo experiments. Does this correspond to differences in the balance between surface and intracellular IgG1 protein?

Answer:
Thank you very much for your comments. Please note that the cells were not fixed and it was surface staining. Two distinct populations of IgG1+ cells could partly be due to different proliferation rates of B cells stimulated by cytokines because Ig class switching requires cell proliferation. We repeated the experiments under the same conditions, and gated the round high IgG1 population as positive.  Fig 4D: The correlation that the authors imply between AID levels and the fraction of cells undergoing CSR does not seem to correspond to the experimental results. Specifically, the levels of AID protein in Foxo1 KO B cells are higher than those in WT or WT/T24A cells (yet, Foxo1 KO cells do not switch); and the levels in WT/T24A cells are not different from those in WT/WT cells, yet they show clear differences in the percentage of IgG1 positive cells. How is this explained?

Answer:
Thank you very much for your comments. We appreciate the reviewer pointed out that the levels of AID protein in FOXO1 KO B cells are higher than those in WT or WT/T24A. We think that it could be due to an incomplete deletion of FOXO1 by tamoxifen administration. In order to obtain a complete deletion of FOXO1, we increased frequency of tamoxifen administration (50mg/kg) from 3 to 5 consecutive days. We also appreciate the reviewer pointed out that the levels of AID in WT/T24A cells are not different from those in WT/WT cells. Actually, there are barely detectable AID proteins in WT/T24A and WT/WT cells probably due to insufficient protein loading. We increased the amount of protein in Western blot and found that the levels of AID in WT/T24A cells were higher than those in WT/WT cells. The new results are shown in Fig 4D instead of the old one.

Answer:
We did the experiments 3 times (original Fig 5D in manuscripts plus 2 replicates). Please find the data from the replicates in our answer to comment 13.

Comment 13:
Previous studies indicate that an important role for IRF4 during plasma cell differentiation is the regulation of BLIMP1 expression. IRF4 also controls expression of AID (Sciammas et al, Immunity 2006;25(5)). In Fig5D, the authors show that while IRF4 levels are 13 increased in Foxo1 KO B cells after LPS stimulation, this does not correlate with an increase in BLIMP1, despite a higher number of B cells appear to adopt a plasma cell phenotype. Can this be reconciled with those previous studies? How do cells become plasma cells in greater numbers without observing an increase in BLIMP1?

Answer:
Thank you very much for your comments. In order to confirm our data in Fig 5D, we independently repeated the experiments twice. Consistent to the results in Fig 5D, we found that IRF4 levels were significantly increased in FOXO1 KO B cells after LPS stimulation (panel A), this did not correlate with a clear increase in BLIMP1 in the two repeated experiments (panel B). Interestingly, previous findings showed that IRF4 deficiency inhibited LPS-induced plasma cell differentiation without affecting Blimp1 expression. (Ulf Klein et al, Transcription factor IRF4 controls plasma cell differentiation and class-switch recombination. Nat Immunol. 2006 Jul; 7(7): 773-82, Fig 4a and 4b).
Based on these observations, we think that increased IRF4 expression in LPS-stimulated FOXO1 KO B cells may compensate BLIMP1 to promote plasma cell differentiation. We noticed that Klein's data showed that IRF4 was not essential for BLIMP1 expression, whereas Sciammas' data showed that IRF4 was. So, whether or not IRF4 regulates BLIMP1 expression is controversial. Loss of FOXO1 promotes IRF4-driven plasma cell differentiation. (A) Expressions of IRF4 and PAX5 as determined using flow cytometry of purified splenic B cells from tamoxifen-treated mice of each genotype cultured in LPS for 3 days. (B) Corresponding cultures in (A) were analyzed via Western blot for FOXO1, IRF4, BLIMP1, and PAX5 protein.

Comment 14:
Why did the authors use CD19-Cre only B cells as controls in the mitochondrial stress