Membrane remodeling by FAM92A1 during brain development regulates neuronal morphology, synaptic function, and cognition

The Bin/Amphiphysin/Rvs (BAR) domain protein FAM92A1 is a multifunctional protein engaged in regulating mitochondrial ultrastructure and ciliogenesis, but its physiological role in the brain remains unclear. Here, we show that FAM92A1 is expressed in neurons starting from embryonic development. FAM92A1 knockout in mice results in altered brain morphology and age-associated cognitive deficits, potentially due to neuronal degeneration and disrupted synaptic plasticity. Specifically, FAM92A1 deficiency impairs diverse neuronal membrane morphology, including the mitochondrial inner membrane, myelin sheath, and synapses, indicating its roles in membrane remodeling and maintenance. By determining the crystal structure of the FAM92A1 BAR domain, combined with atomistic molecular dynamics simulations, we uncover that FAM92A1 interacts with phosphoinositide- and cardiolipin-containing membranes to induce lipid-clustering and membrane curvature. Altogether, these findings reveal the physiological role of FAM92A1 in the brain, highlighting its impact on synaptic plasticity and neural function through the regulation of membrane remodeling and endocytic processes.

Fig. 6A: Cristae morphology needs to be quantified from several independent micrographs.Fig. 6M/N: Synaptic area appears to be increased in heterozygous, but not so much in homozygous ko mice.Any explanation for this?While it is an interesting observation worth to report, it seems difficult to conclude from these results that FAM92A1 has a role in regulating synaptic vesicle number/synaptic activity.Please discuss.Fig. 7A and Fig. 1B: Please provide a specificity control for the FAM92A1 antibody in immunostainings by showing that the respective ko cells and ko brain slices are not stained by the antibody (obviously using the same protocols as for wt).Is only a single FAM92A1 band recognized in the Western blots or are there also non-specific bands?Has the antibody been raised against the BAR domain or some region outside the BAR domain?Please mention the latter in the methods.
Fig. 7A top: The endogenous FAM92A1 staining looks mostly punctate and cytoplasmic with some minor mitochondrial co-localization.I do not see much plasma membrane staining, beyond some random dots.The authors should repeat these experiments using TIRF microscopy and staining of the endogenous FAM92A1 protein, together with a marker for clathrin coated pits (CCPs) and/or markers for clathrin-independent uptake routes.Without a clear co-localization of FAM92A1 to CCPs, it is difficult to judge whether the observed difference in CCP morphology (Fig. 7L) and endocytic uptake (Figure 7G-H) is a direct effect on clathrin-mediated endocytosis or indirectly caused, for example by defective mitochondria/lipid metabolism.
Fig. 7A bottom: The staining pattern of GFP-tagged FAM92A1 looks different from the endogenous staining, e.g. more cellular structures are stained which could be an over-expression artefact.I would only show the endogenous stain (if it is specific, see point above).Same worries for Fig. 7B.Is this endogenous staining (as stated in the main text) or staining of FAM92A1-GFP (as stated in the figure)?Please show the endogenous FAM92A1 staining for these experiments.
Fig. 8: The structural data for the FAM91A1 BAR domain appear convincing, based on the provided data table and in the absence of a validation report.However, the follow-up evaluation of the structural data is not conclusive: While the significance of the BAR domain dimer may be inferred by comparison with other BAR domains, the suggested oligomer interface cannot -BAR domain proteins use different oligomerization interfaces for their assembly on membranes.Does FAM91A1 form a dimer in solution and can it be disrupted by mutations in the dimer interface?Are the observed dimer and oligomer interfaces required for the interaction with liposomes and membrane tubulation?Fig. 8: The authors should discuss or show a scheme how the positive curvature of the BAR domain dimer relates to its localization in the mitochondrial matrix/cristae.Minor Fig. 8: Please add the domain architecture of FAM92A1 with the amino acid numbers, also showing the crystallized construct, e.g. even if it is only a BAR domain.The absence of an SH3 domain may hint for a localization other than CCPs, as most BAR domain proteins at CCPs contain an SH3 domain required for their recruitment.
At some sites, PMIDs instead of references are provided.
Reviewer #3 (Remarks to the Author): Wang et al. investigate the neuronal functions of FAM92A1, a BAR domain protein previously shown to be important for membrane remodeling and proper mitochondrial structure and function.Using a newly generated murine KO model, the authors here report that global deletion of FAM92A1 results in pleiotropic effects in the CNS, ranging from deficits in synapse structure to neuronal network dysfunction and impaired memory.In the second part of the ms, the authors solve the crystal structure of FAMP92A1 and perform MD simulations to elucidate its membrane interactions.

General points -----(1)
The authors conclude that their characterization of CNS effects in FAM92A1 KO mice revealed the "molecular basis of the pathophysiology of FAM92A1-linked neurological disorders in humans" (l.108).However, which disorder(s) do the authors refer to here?In l. 63, they cite Debost-Legrand et al. (2013) who reported a case of autism associated with 8q21.3 deletion, which may or may not be due to FAMP92A1 loss.Even if this were the case, how does the microencephaly found in these patients relate to brain swelling observed in FAM92A1 KO mice?Mutations in FAM92A1 have further been linked to polydactyly in human and mice (Schrauwen et al. 2019 PMID: 30395363), but this phenotype was not explored in the present ms.A related question is whether male infertility in mice as observed in the present ms has been found in patients with mutations in FAM92A1?In sum, the above conclusion is not fully supported by the data provided.
(2) Most investigations were performed at an age when a substantial percentage of KO mice had already died (Fig. 1H).This suggests that that the surviving mice were severely sick at the time of investigation, implying that the observed alterations may well be due to systemic factors rather than more direct effects of FAM92A1 in the CNS.Indeed, FAM92A1 seems to be ubiquitously expressed from embryonic stages onward (Figs. 1 and S1).This issue may be particularly relevant for "high-level functions" such as synaptic transmission , neuronal network dynamics (Fig. 5) or behavior (Fig. 3).As the authors analyzed a conventional KO, FAM92A1 was absent from the onset of embryogenesis, which further makes it impossible to discriminate between acute effects (at the time of investigation) vs. developmental impairments.One way of addressing this question is by inducing a more spatially and temporally restricted KO or KD of FAM92A1.An advanced mechanistic understanding would considerably increase the relevance of the ms.
(3) The ms suffers from several shortcomings regarding the reporting of data analysis and statistics, which challenge several of the authors' main conclusions (see below).
Additional points & suggestions ------Statistical details can only be found in the figure legends and are frequently incomplete.To increase the transparency of reporting, I strongly suggest adding a supplementary table containing detailed statistical information for all datasets included.-Incomplete statistics: (i) ANOVA in Fig. 1G: Details of post-hoc tests need to be added.(ii) Fig. 1B, C, D, E: The number of biological replicates (n) is lacking.(iii) Fig. 1G: "n = 3 replicates/group": Are these biological or technical replicates?(iv) l. 152 "departure from the expected Mendelian inheritance pattern": Please add test details.(v) l. 189 "Though the statistical differences were less prominent, a consistent reduction […]": This conclusion is invalid, as Fig. 2E shows that there is no significant difference in any of the analyzed hippocampal areas!(vi) Fig. 3G, H: n is lacking.(vii) Fig. 6A: n is lacking (viii) Fig. 6G: Please add statistical details. (ix) Fig. 6H: The use of a standard ANOVA is questionable, as g-ratio depends strongly on axon diameter (Fig. 6G).ANCOVA or similar may be more appropriate.What is n here (mice/slices/axons)? (x) Fig. 6N: I could not find any statistical details, i.e., the conclusion is not adequately supported by data.
-The relevance and mechanism of brain "swelling" (Fig. 1) remains unclear, as neither changes in CSF nor volume changes of white or grey matter can explain it.This discrepancy should be resolved.
-Fig.2C: The unit should be added.-Fig.2G-H: The reported reduction is not evident from the sample images.Also, how does this relate to the volume seen with MRI? -l.215 "reduced regional gray matter volume, which may be attributed to neuronal degeneration": This conclusion is not convincing, as no reduction of gray matter volume (MRI) was seen in DG, where Fluoro-Jade staining was performed.It remains unclear to me why these histological analyses have focused on DG rather than those areas in which gray matter volume changes were found by MRI.-Fig.3G-J: These analyses are of peripheral interest to the present ms.-Fig.4: It remains unclear which neurons were traced in Sholl analyses (l.278: "hippocampal neurons").Please add all relevant details to Methods.Again, the brain region analyzed is obviously not motivated by the MRI findings in Figs. 1 and 2. Please explain.
-Fig.5A-C: Methods related to Ca2+ imaging are completely lacking.It is unclear to me how the experiments were conducted and what exactly was measured/quantified.For example, the authors mention spontaneous activity, but Fig. 5B shows trials.Please add all relevant details to Methods.
-Fig.5D-E and l. 330 ("impact of FAM92A1 on excitatory synaptic transmission is more pronounced at individual synaptic sites"): The data provided do not support such a conclusion!Also note that mEPSC and sEPSC amplitudes are basically the same.-l.333 "further supported by the decreased number of glutamate release sites per neuron": A reduction in mEPSC frequency does not necessarily imply a lower number of release sites, as it is also strongly dependent on release probability.Please correct.-Fig.5E: Why are the lower traces interrupted?What does the green line in the left trace indicate?-MEA recordings: The results are very difficult to read, mainly because the definitions of the reported quantities are not properly introduced.For example, "synchrony index" is not defined.Please add all analytical details to Methods and rephrase Results for clarity.What do the gray bars in Fig. 5J indicate?-l.424-425 "area of synaptic vesicles (SVs) … above 5000 nm³": Did the authors analyze areas or volumes?-l.89 "structural characteristics have not been elucidated yet": Why is Breugel et al. 2022 (PMID:

Point-by-Point Response to Reviewers' Comments
We deeply appreciate the insightful comments and suggestions provided by the reviewers, which have significantly improved the quality of our work.In response to these comments and suggestions, we have incorporated new data, reanalyzed existing data, and restructured specific sections of the manuscript accordingly.We believe that these additional works significantly improve the quality and comprehensiveness of our manuscript.Below, we present a detailed, point-by-point response to the reviewers' comments, with reviewer remarks shown in black and our responses in blue.Please be aware that the figure citations in our response pertain to the new (post-revision) figures.We have highlighted the corresponding changes within the manuscript in red.We look forward to your feedback and the opportunity to have our work published in Nature Communications.
Reviewer #1 (Remarks to the Author): As requested by the editor my comments will be mainly on evaluation of mouse MRI and voxel-based morphometry experiments.For this part of the work Wang et al. examined the brain morphology of FAM92A1 knockout and Wilt type mice as a control using volumetric MRI.In summary, authors described a brain swelling and a reduction in gray matter in FAM92A1 KO mice.Specifically, authors used 7T MRI (Bruker) and did the images acquisition on anesthetized animals.Authors found an increased brain volume in FAM92A1 KO mice.Then, to investigate more in detail this increased of brain volume, authors performed a whole brain voxel-based morphometry (VBM) of the gray matter.The authors find that in general gray matter volume is decreased in FAM92A1 KO mice.These changes seem very localized and subtle.
Pre and post processing of the data are appropriate and well conducted.All together the MRI experiments presented by Wang et al. are well designed with correct statistical analyses.

Response:
We thank the reviewer for the appreciation of our study and the constructive comments.Following the suggestions, we have conducted additional experiments as the reviewer's suggested method and addressed this reviewer's concerns as described below.We hope that our response sufficiently addresses the concern raised by the reviewer.

Minor comments
-As these changes are subtle why author didn't do MRI on fixed brain as classically done for VBM (see: PMID: 25199916IF: 11.0 Q1 PMID: 22557981IF: 4.7 Q2 ) to increase the resolution of the images?
Response: We appreciate the reviewer's valuable comments.As suggested, we have acquired new MRI scans with enhanced contrast by utilizing fixed mouse brains perfused with 2 mM ProHance.Due to the super time-consuming nature of scanning the entire brain and the limited available homozygous mice suitable for perfusion, we were only able to scan three mice from each wild-type and FAM92A1 +/-group for analysis.Despite the small number of mice used, the deformation-based morphometry (DBM) analysis of the volumetric morphology of the mouse brain and the voxel-based morphometry (VBM) analysis of gray matter revealed a noticeable decrease in volume outside of the lateral ventricles (LVs), accompanied by a reduced gray matter density within the diminished cortical area (Supplementary Fig. 2e).These data suggest that the suggested method can enhance sensitivity to detect subtle differences in hippocampal subregions.We thank the reviewer for suggested methods, we have now used and cited this work.
We have supplemented test details and new data in our revised manuscript.Changes in corresponding Materials (Lines 997-1001 and 1029-1041), Results descriptions (Lines 203-226), and Figure  This is an interesting and noteworthy study uncovering a role of FAM92A1 in neuronal development and membrane remodeling.However, the wide breadth of different disciplines used to characterize the function of on FAM92A1 comes at the expense that some of the experiments are not conducted in the required detail to be fully conclusive (at least in the second part).My expertise is in structural biology and endocytic trafficking so I will mostly comment on experiments starting from Fig. 6.

Response:
We thank the reviewer for the appreciation of our study and the constructive comments.Following the suggestions, we have conducted additional experiments and addressed the reviewer's major concerns as described below.We appreciate the invaluable comments provided during the review process.This included the incorporation of new MRI data and a reanalysis of previous MRI data to comprehensively understand the cause of brain swelling.
Based on the latest findings, we have meticulously revised the description and succinctly summarized the results in the manuscript.We believe that the quality of our study has been significantly improved by addressing these concerns.
We hope that our response adequately addresses the concerns raised by the reviewer.

Major:
Fig. 6A: Cristae morphology needs to be quantified from several independent micrographs.

Response:
We appreciate the reviewer's comments.The impact of FAM92A1 depletion on mitochondrial morphology, including mitochondrial diameter, perimeter, and the number of cristae per mitochondrial perimeter, has been assessed using electron micrographs as we previously reported.In line with our earlier findings, FAM92A1 depletion results in swollen mitochondria with reduced or even absent cristae (Fig. 6b).
We have supplemented these new data in our revised manuscript.Fig. 6M/N: Synaptic area appears to be increased in heterozygous, but not so much in homozygous ko mice.Any explanation for this?While it is an interesting observation worth to report, it seems difficult to conclude from these results that FAM92A1 has a role in regulating synaptic vesicle number/synaptic activity.Please discuss.

Response:
We appreciate the reviewer's critical comments.As pointed out by the reviewer, the changes in the average area of SVs in the FAM92A1 +/-group were more pronounced than those in the FAM92A1 -/-group when compared to the SV area in the FAM92A1 +/+ group.In the quantification of the synaptic area in previous Fig.6M, we analyzed data from a similar number of synaptic vesicles (SVs), rather than SVs from the same number of synapses.In fact, through quantification of SVs number (Fig. 6o) by a stereological approach, the number of SVs was remarkably increased in the FAM92A1 -/-synapses in contrast to FAM92A1 +/+ synapses.Hence, to get a similar number of SVs (FAM92A1 +/+ , n = 361; FAM92A1 +/-, n = 351; FAM92A1 -/-, n = 346) for statistical analysis, a smaller number of synapses were indeed incorporated in FAM92A1 -/-group compared to the FAM92A1 +/+ group in the previous Fig.6M.Due to most of the SVs among all groups with a diameter < 60 nm (Fig. 6p), the average area of SVs was not evident in the FAM92A1 -/-group.
To address this concern, SVs from the same number of synapses (n = 15 synapses per group) were used for quantification in the revised manuscript (Supplementary Fig. 5e, f).Compared to the SV area in the FAM92A1 +/+ group, changes in the average area of SVs in the FAM92A1 +/-group were still more evident than that in the FAM92A1 - /-group.Loss of FAM92A1 not only resulted in decreased endocytosis but also impaired mitochondrial inner membrane architecture and energy production (Fig. 6a-f).
Neuronal activity, including neurotransmitter release and synaptic vesicle membrane retrieval at synapses, relies heavily on energy.Depletion of FAM92A1 not only impaired the FAM92A1-involved step of synaptic vesicle membrane retrieval but also inhibited all energy-dependent endocytic processes of synaptic vesicle membrane (Li et al. 2021. PMID: 34782781;Devine et al. 2018. PMID: 29348666).Compared to FAM92A1 homozygous mice, both mitochondrial architecture and function were less affected in the FAM92A1 heterozygous mice (Fig. 6a-f), indicating less suffering from the energy shortage.As a result, in contrast to the FAM92A1 homozygous, the endocytic process of the synaptic vesicle membrane might occur more frequently in the FAM92A1 heterozygous.Consequently, more SVs might exhibit enlarged areas due to defects in membrane curvature during endocytosis of the synaptic vesicle membrane.Therefore, we speculated that the aberrant structure of SVs and decreased synaptic activity result from the combined effects of impaired mitochondrial energy provision and the endocytic process of the synaptic vesicle membrane.
Moreover, to observe the changes in SVs, both changes in synaptic number and area have been quantified using the FIB-SEM electron micrographs in the revised manuscript.Synapses displaying distinguishable pre-and postsynaptic areas were measured on a micrograph with dimensions of 2209 × 205 pixels.Quantification was conducted on every 50 th slice, with a total of 18 slices used for analysis (Fig. 6n).Quantification results revealed that FAM92A1 depletion caused an increased ratio of SVs with a diameter larger than 80 nm.The average area of SVs was found to be increased in the FAM92A1-depleted synapses (Fig. 6o-q).Notably, the heterozygous mouse data were not included in this analysis due to the large dataset.This 3D EM data will be deposited in EMPIAR and made publicly visible.
We have supplemented these new data in our revised manuscript.Changes in corresponding Materials (Lines  Response: We appreciate the reviewer's valuable and critical comments.Both Fig. 1b and 7a depict the localization of FAM92A1 in the brain and HT22 cells, respectively.We tested two commercial FAM92A1 antibodies purchased from Sigma (CAT#HPA034760) and Proteintech (CAT#24803-1-AP) in different experiments.Our test results revealed that both antibodies have their limitations, as is common with technical challenges associated with antibodies.
The antigen sequences used for raising these antibodies have been indicated in Supplementary Fig. 1e.The FAM92A1 antibody from Proteintech recognized multiple non-specific bands across all samples, including primary fibroblast and brain tissue (Response Fig. 1b).The antibody from Sigma displayed fewer non-specific bands.
The two guide RNAs were separately targeted to the upstream of Fam92a1 exon 1 and downstream of Fam92a1 exon 6, resulting in the microdeletion of the first six exons (Supplementary Fig. 1c).According to the resolved crystal structure of FAM92A1 BAR dimer, the N-terminus residues 2-211 encoded by genes spanning exon 1 to exon 7 is the FAM92A1 BAR domain.Both FAM92A1 antibodies can recognize the part of the sequence of the BAR domain.The FAM92A1 antibody from Sigma recognizes amino acids encoded by the Fam92a1 gene spanning exon 7 to exon 8, and the FAM92A1 antibody from Proteintech recognizes amino acids encoded by the Fam92a1 gene spanning exon 4 to exon 8 (Supplementary Fig. 1e).Due to the deletion of the first six exons in the FAM92A1 knockout mouse, we utilized the antibody from Proteintech for the knockout efficiency identification through western blot, except for immunostaining.This decision was made because the Proteintech antibody did not perform well in immunostaining experiments, exhibiting non-specific diffused localization in cells due to its recognition of multiple non-specific bands (Response Fig. 1a, c).The antibody from Sigma was used to detect the localization of endogenous FAM92A1.
However, although the FAM92A1 antibody from Sigma displayed fewer non-specific bands, it exhibited a minor weak non-specific background signal in KO cells and slices in immunostaining experiments (Response Fig. 1a).We chose not to include these data in the manuscript due to the potential for misleading information caused by antibody limitations.
We have supplemented these new data in our revised manuscript.Changes in corresponding Results (Lines 146-157) descriptions and Figure legends (Supplementary file, Lines 37-46) are marked in red.
Fig. 7A top: The endogenous FAM92A1 staining looks mostly punctate and cytoplasmic with some minor mitochondrial co-localization.I do not see much plasma membrane staining, beyond some random dots.The authors should repeat these experiments using TIRF microscopy and staining of the endogenous FAM92A1 protein, together with a marker for clathrin coated pits (CCPs) and/or markers for clathrin-independent uptake routes.Without a clear co-localization of FAM92A1 to CCPs, it is difficult to judge whether the observed difference in CCP morphology (Fig. 7L) and endocytic uptake (Figure 7G-H) is a direct effect on clathrin-mediated endocytosis or indirectly caused, for example by defective mitochondria/lipid metabolism.

Response:
We appreciate the reviewer's valuable comments.As suggested, we utilized TIRF microscopy to determine the localization of endogenous FAM92A1 on the plasma membrane and its co-localization with clathrin and caveolin.
The results revealed that FAM92A1 resides in the plasma membrane, although its abundance there was relatively lower compared to its enriched abundance in mitochondria.Moreover, it was found to co-localize with both clathrin and caveolin (Fig. 7p).These findings suggest the involvement of FAM92A1 in both endocytic pathways.This parallels another BAR domain protein, endophilinA2, which is also implicated in membrane remodeling during both clathrin-mediated endocytosis (CME) and clathrin-independent endocytosis (CIE) (Boucrot et al. 2014. PMID: 29425511 Bertot et al. 2018. PMID: 25517094).Consistently, our new immuno-electron microscopy data confirmed the localization of endogenous FAM92A1 on the plasma membrane (The detailed explanation and results were shown in the next response).Collectively, these data indicate that FAM92A1 localizes to the plasma membrane and is required for endocytosis.
We have supplemented these new data in our revised manuscript.Changes in corresponding Materials (Lines  Response: We appreciate the reviewer's critical comments.As suggested, we have used the FAM92A1 antibody to visualize the subcellular localization of endogenous FAM92A1 both in the HT22 cells and primary hippocampal neurons through immunofluorescence and Immuno-electron microscopy (Immuno-EM).The immunofluorescent images presented that FAM92A1 localized to mitochondria and the plasma membrane in HT22 cells (Fig. 7a).The Immuno-EM also revealed that FAM92A1 was located inside mitochondria and at or near the periphery of the plasma membrane (Fig. 7b).Similar results were visualized in the primary hippocampal neurons, FAM92A1 was highly expressed in neurons, with partial localization to both the mitochondrial protein VDAC and synaptic proteins SV2, vGlut1, and PSD95 (Fig. 7c, d and Supplementary Fig. 6d).We completely understand the concern of the reviewer regarding the potential artifact of overexpressed FAM92A1-GFP.Therefore, we moved all FAM92A1-GFP overexpressed results into the supplementary figures (Supplementary Fig. 6c-d).
We have supplemented these new data in our revised manuscript.Changes in corresponding Materials (Lines  Response: We appreciate the reviewer's valuable comments.As suggested, we simultaneously mutated three hydrophobic residues at the dimer interface observed in the structure to negatively charged residues, generating a mutant construct (FAM92A1 Leu70Glu-Phe73Glu-Phe184Glu, abbreviated as mutant) (Supplementary Fig. 7e).
These mutations aimed to disrupt the dimer interface packing.Compared to the dimerized FAM92A1 wild-type BAR domain protein in equilibrium (Wang et al. 2019. PMID: 30404948), the purified mutant protein existed as a monomer in equilibrium, with minor aggregates (Supplementary Fig. 7f).As shown in the SEC-MALS profile, the major peak in the chromatogram corresponds to a monomeric species, not seen for the WT (Wang et al 2019. PMID: 30404948).
This result indicates the importance of these residues for formation of the FAM92A1 BAR dimer Although the interface mutant lost its capacity to form dimers compared to WT, co-sedimentation assay revealed that the mutant protein preserved its membrane binding ability with a minor decrease compared to the WT protein (Supplementary Fig. 7g), while mutation of the membrane binding residues of FAM92A1 BAR (as we previously detected Mut3 and Mut5) have been demonstrated to impair its membrane affinity (Wang et al. 2019. PMID: 30404948).Despite retaining the affinity of liposomes, the mutant protein abolished its ability to sculpt the spherical liposomes into narrow tubules, suggesting the absence of membrane remodeling activity (Supplementary Fig. 7h).
These data demonstrate that the dimerization of the FAM92A1 BAR domain is necessary for membrane sculpting, but not dispensable for binding to the membrane interaction.
The oligomerization interface involves a head-to-head interaction in the crystal lattice (Supplementary Fig. 8b, c).
However, we noticed non-specific aggregates exist in all the samples, making it difficult to define the oligomerization mutants.The role of protein oligomerization in membrane-interaction and -remodeling is thus challenging to study.
We have discussed this in the revised manuscript.
We have supplemented these new data in our revised manuscript.

Response:
We appreciate the reviewer's insightful comments.The crystal structure of the FAM92A1 BAR domain demonstrates that its curvature closely resembles that of classical BAR domains, which are banana-shaped.Similar to other BAR domain proteins, the concave surface of the FAM92A1 BAR domain generates positive membrane curvature, as we have shown using molecular simulations in this study (Fig. 7h, j) and also our previous giant vesicle model (Wang et al., 2019).Additionally, our previous Immuno-EM data showed that endogenous FAM92A1 resides in the mitochondrial matrix in close proximity to the inner membrane and clusters mostly along the cristae (Figure 1E in Wang et al., 2019).Some localization at the base and tip of the cristae was also observed.Combining these observations with the molecular simulation data, it suggest that FAM92A1 interacts with the mitochondrial inner membrane to generate positive curvature, initiating membrane invaginations to form cristae, and maintains the cristae structure by binding to the inner membrane.We have discussed this in the revised manuscript, and changes in the corresponding Discussion (Lines 819-934) descriptions are marked in red.

Minor
Fig. 8: Please add the domain architecture of FAM92A1 with the amino acid numbers, also showing the crystallized construct, e.g. even if it is only a BAR domain.The absence of an SH3 domain may hint for a localization other than CCPs, as most BAR domain proteins at CCPs contain an SH3 domain required for their recruitment.

Response:
We appreciate the reviewer's valuable comments.As suggested, we have added the aligned amino acid sequences of human and mouse FAM92A1, the domain architecture of FAM92A1 with amino acid numbering, and the details of the construct used for crystallization of the FAM92A1 BAR domain in the revised manuscript (Supplementary Fig. 1e and 7a).At the beginning of crystallizing the FAM92A1 BAR domain, according to the secondary structure of human FAM92A1 and the implied BAR domain from Uniprot, we predicted the N-terminal residues 1-219 constituted the BAR domain (Wang et al. 2019. PMID: 30404948).Therefore, we expressed and purified the N-terminal residues 1-219 (Supplementary Fig. 7a).In fact, based on the resolved crystal structure of the human FAM92A1 BAR domain, the N-terminal residues 2-211 is the BAR domain.
FAM92A1 predominantly localizes to mitochondria, with some localization on the plasma membrane.Our new data revealed that FAM92A1 co-localized with the endocytic markers clathrin and caveolin (Fig. 7p), suggesting its involvement in both endocytic pathways.FAM92A1 likely plays a role in generating positive curvature of the plasma membrane during the endocytosis process, rather than interacting with other proteins such as N-WASP through an SH3 domain, to coordinate with the actin cytoskeleton during the endocytic process (Defne et al. 2007. PMID: 17609109;Ringstad et al. 1999. PMID: 10677033).However, the detailed molecular mechanism requires further studies in the future.We have discussed this in the revised manuscript.
We have supplemented these new data in our revised manuscript.Response: We appreciate the reviewer for this helpful suggestion.As suggested, we have thoroughly reviewed the entire article, and these kinds of problems in the numbers and errors have been revised in the revised manuscript.
At some sites, PMIDs instead of references are provided.

Response:
We thank the reviewer's kind suggestion.We have thoroughly reviewed the entire article and revised these, and the PMIDs have been replaced with references in some sites of the main text.
Reviewer #3 (Remarks to the Author): Wang et al. investigate the neuronal functions of FAM92A1, a BAR domain protein previously shown to be important for membrane remodeling and proper mitochondrial structure and function.Using a newly generated murine KO model, the authors here report that global deletion of FAM92A1 results in pleiotropic effects in the CNS, ranging from deficits in synapse structure to neuronal network dysfunction and impaired memory.In the second part of the ms, the authors solve the crystal structure of FAMP92A1 and perform MD simulations to elucidate its membrane interactions.

Response:
We thank the reviewer for the appreciation of our study and the constructive comments.Following the suggestions, we have conducted additional experiments and addressed this reviewer's major concerns as described below.We believe that the quality of our study is significantly improved by addressing these concerns and we hope that our response sufficiently addresses the concerns raised by the reviewer.

General points
----- (1) The authors conclude that their characterization of CNS effects in FAM92A1 KO mice revealed the "molecular basis of the pathophysiology of FAM92A1-linked neurological disorders in humans" (l.108).However, which disorder(s) do the authors refer to here?In l. 63, they cite Debost-Legrand et al. (2013) who reported a case of autism associated with 8q21.3 deletion, which may or may not be due to FAMP92A1 loss.Even if this were the case, how does the microencephaly found in these patients relate to brain swelling observed in FAM92A1 KO mice?Mutations in FAM92A1 have further been linked to polydactyly in human and mice (Schrauwen et al. 2019 PMID: 30395363IF: 6.2 Q1 ), but this phenotype was not explored in the present ms.A related question is whether male infertility in mice as observed in the present ms has been found in patients with mutations in FAM92A1?In sum, the above conclusion is not fully supported by the data provided.(2) Most investigations were performed at an age when a substantial percentage of KO mice had already died (Fig. 1H).This suggests that that the surviving mice were severely sick at the time of investigation, implying that the observed alterations may well be due to systemic factors rather than more direct effects of FAM92A1 in the CNS.Indeed, FAM92A1 seems to be ubiquitously expressed from embryonic stages onward (Figs. 1 and S1).This issue may be particularly relevant for "high-level functions" such as synaptic transmission , neuronal network dynamics (Fig. 5) or behavior (Fig. 3).As the authors analyzed a conventional KO, FAM92A1 was absent from the onset of embryogenesis, which further makes it impossible to discriminate between acute effects (at the time of investigation) vs. developmental impairments.One way of addressing this question is by inducing a more spatially and temporally restricted KO or KD of FAM92A1.An advanced mechanistic understanding would considerably increase the relevance of the ms.
(3) The ms suffers from several shortcomings regarding the reporting of data analysis and statistics, which challenge several of the authors' main conclusions (see below).

Response:
We appreciate the reviewer's comments.
(1) We agree with the reviewer that the confirmed neurological disorders associated with FAM92A1, including male infertility in humans as well as experimental data, are currently unknown.The confirmed disease related to the FAM92A1 mutation is polydactyly (Schrauwen et al. 2019. PMID: 30395363).Consistently, during mouse breeding in this study, four of 26 FAM92A1 knockout mice were found to have visible polysyndactyly.For example, mouse with polysyndactyly was observed on the right hind paw (Response Fig. 2).Because this manuscript aimed to investigate the potential role of FAM92A1 in the central nervous system.The reported abnormal digit morphology in the Fam92a -/-mice, like osteoma (Schrauwen et al. 2019. PMID: 30395363), which is not visible through the eye but is detectable using X-ray.Anyway, the visible polysyndactyly in the FAM92A1 -/-mice further confirmed the role of FAM92A1 in maintaining the normal skeleton morphology.
As for the reported case of an autism patient with an 8q21.3 microdeletion, it remains uncertain whether it is merely attributable to FAM92A1 loss.Although the three patients with polydactyly did not exhibit evident neurological problems, neurological disorders may become apparent with age progression, according to our behavioral result of age-associated memory decline and cognitive deficits (Fig. 4 and Supplementary Fig. 4).Hence, the direct link of FAM92A1 mutation with the neurological disorder in humans is hardly known at present.We aimed to elucidate potential associations through IPA analysis of differentially expressed genes following FAM92A1 loss.This bioinformatic investigation revealed a clustering of neurological disorders, including dementia and Alzheimer's disease (AD) (Fig. 3r).These analyses collectively imply that the loss of FAM92A1 expression heightens the susceptibility to neurological disorders.However, the exact neurological disease associated with FAM92A1 requires further investigation.
According to our previous study and current study, we revealed that membrane-remodeling mediated by FAM92A1 is crucial from embryonic development onward.Consequently, the absence of FAM92A1-caused abnormal inner membrane architecture leads to mitochondrial dysfunction, including energy shortage, oxidative stress, and probably metabolic dysfunction.These factors collectively contribute to neurodevelopmental anomalies, a hallmark pathology observed in microcephaly.Notably, an increasing number of patients or infants with mutations in mitochondrial proteins exhibit microcephaly (Nathalie el al. 2020. PMID: 32294449;Ruth et al. 2016. PMID: 26992161;José-Mario et al. 2015. PMID: 25650066).For instance, a patient harboring a homozygous variant in NDUFA8, affecting mitochondrial complex I, also manifests developmental delay and microcephaly (Yukiko et al. 2020. PMID: PMID: 32385911).Additionally, the malformation of brain morphology caused by FAM92A1 deficiency may worsen with decreased neuronal plasticity (Fig. 4 and 5).Moreover, brain swelling in the FAM92A1-deficient mice may reversely lead to increased pressure inside the skull, potentially causing further damage to brain tissue and impeding brain growth.Taken together, according to our experimental results, the observed global developmental delay, autism, and microcephaly in the patient with microdeletions around the FAM92A1 loci may represent a combined consequence of systemic neurodevelopmental defects.
For male infertility, the reported study did not report whether three men with the homozygous Fam92a mutation have problems with male infertility.And the pedigree did not show linkage analysis for the offspring of these three men (Schrauwen et al. 2019. PMID: 30395363).Hence, we are unable to determine whether the male infertility of FAM92A1-depleted mice also occurs in patients with mutations in FAM92A1.Because the critical role of mitochondria in sperm function is widely studied, extending beyond energy production to encompass various functions throughout gamete production and reproduction (Singh et al. 2020. PMID: 32534048).Therefore, it is not surprising that FAM92A1 knockout mice exhibit male infertility.Importantly, this phenotype further implies the pivotal role of FAM92A1 in mitochondria.Because the founded male infertility is beyond the scope of this study.We included this phenomenon, which has not been reported in other studies to date, in the manuscript to provide a comprehensive overview of the symptoms observed in FAM92A1 knockout mice.We hope that future studies or case reports will explore this aspect in more detail.
(2) We acknowledge the Reviewer's insightful advice.According to the reviewer's comments, we investigated the role of FAM92A1 in the central nervous system (CNS) using AAV-mediated spatially and temporally knockdown neuronal FAM92A1 in the mouse hippocampus.A miR-30-based short hairpin RNA (shRNA) technique was applied to specifically knockdown the FAM92A1 transcript within the hippocampus.Two siRNA sequences targeting FAM92A1 (FAM92A1 siRNA #1 and #4) (Supplementary Fig. 3a, b) were individually cloned into a neurotropic AAV vector (pAAV-hSyn-EGFP-3xFlag), generating pAAV-hSyn-EGFP-3xFlag-miR30shRNA (Fam92a1) (abbreviated as shFam92a1) (Fig. 3a).Three weeks after AAV injection into the hippocampus (Fig. 3b-d), a series of learning and memory-involved behavioral tests were conducted to assess the acute effects of FAM92A1 depletion on mouse behavior.
Consistent with the memory defects of FAM92A1 knockout mice, mice with FAM92A1 knockdown in the hippocampal neuron presented an attenuated performance in memory tasks, particularly in tasks involving spatial and contextual memory retrieval (Fig. 3g, i, k, n, and Supplementary Fig. 4b, e, f, i).In contrast to the control group, mice subjected to FAM92A1 knockdown exhibited decreased accuracy and spent less time in the target quadrant (SW) during the Morris water maze task (Fig. 3g, i, k).Additionally, they demonstrated impaired recall of fear memory when reintroduced to the same contextual environment in the fear conditioning test (Fig. 3n), and decreased ratio of correct choice in T-maze (Supplementary Fig. 4i).Similar to FAM92A1 knockout mice, FAM92A1 knockdown has less affection on the recognition memory of mice (Supplementary Fig. 4l).These data collectively demonstrate the role of FAM92A1 in preserving spatial and contextual memory.
In addition to the behavioral tests, our previous study has investigated changes in neuronal morphology following the knockdown of FAM92A1 in cultured hippocampal neurons.In contrast to the enriched presence of mushroomlike spines, FAM92A1 knockdown led to a noticeable reduction in spine density and the absence of mushroom-like morphology (Fig. 4j).For the affection of FAM92A1 knockdown on neuronal function, we previously recorded the spontaneous electrophysiological activity of cultured hippocampal neural networks using the microelectrode arrays (MEA) technology.Due to the low probability of visible FAM92A1 -/-embryos, we failed to identify the visible FAM92A1 -/-embryos even after four attempts.Consequently, we were unable to compare the activity difference between FAM92A1 knockout and knockdown neurons.
legends (Supplementary file, Lines 62-69) are marked in red.Reviewer #2 (Remarks to the Author): Wang and colleagues report a comprehensive functional analysis of the BAR-domain containing protein FAM92A1, including the generation and morphological characterization of gene knockout mice, behavioral studies, neuronal activity measurements, experiments probing morphology and function of mitochondria, cell-based assays, such as colocalization and endocytic uptake experiments, EM analyses of synapses, structural biology and molecular dynamics simulations.Based on their analyses, they propose a function of FAM92A1 in the regulation of synaptic plasticity and neural function by controlling endocytosis and mitochondrial membrane remodeling.
Changes in corresponding Results (Lines 438-444) descriptions and Figure legends (Lines 2134-2135) are marked in red.

Fig
Fig. 7A and Fig. 1B: Please provide a specificity control for the FAM92A1 antibody in immuno-stainings by showing that the respective ko cells and ko brain slices are not stained by the antibody (obviously using the same protocols as for wt).Is only a single FAM92A1 band recognized in the Western blots or are there also non-specific bands?Has the antibody been raised against the BAR domain or some region outside the BAR domain?Please mention the latter in 1109-1122), Results (Lines 573-580) descriptions, and Figure legends (Lines 2126-2128) are marked in red.

Fig
Fig. 7A bottom: The staining pattern of GFP-tagged FAM92A1 looks different from the endogenous staining, e.g. more cellular structures are stained which could be an over-expression artefact.I would only show the endogenous stain (if it is specific, see point above).Same worries for Fig. 7B.Is this endogenous staining (as stated in the main text) or staining of FAM92A1-GFP (as stated in the figure)?Please show the endogenous FAM92A1 staining for these experiments.

Fig. 8 :
Fig.8: The structural data for the FAM91A1 BAR domain appear convincing, based on the provided data table and in the absence of a validation report.However, the follow-up evaluation of the structural data is not conclusive:While the significance of the BAR domain dimer may be inferred by comparison with other BAR domains, the suggested oligomer interface cannot -BAR domain proteins use different oligomerization interfaces for their assembly on membranes.Does FAM91A1 form a dimer in solution and can it be disrupted by mutations in the dimer interface?Are the observed dimer and oligomer interfaces required for the interaction with liposomes and membrane tubulation?
Fig. 8: The authors should discuss or show a scheme how the positive curvature of the BAR domain dimer relates to its localization in the mitochondrial matrix/cristae.