On 2016 Jan 04, Avital Rodal commented:

First, we would like to clarify that our FCHSD1 and FCHSD2 constructs, which generate protrusions in cultured cells, are not >500 aa as Dr. Gould suggested in her comment on 12/31. As described in Becalska AN, 2013, we quantitatively determined robust cellular activities for FCHSD1[1-417] (41 amino acids longer than the construct in McDonald NA, 2015 and for FCHSD2[1-414] (only 18 amino acids longer than the construct in McDonald NA, 2015). While we have not yet further characterized these two mammalian proteins in vitro, the same purified fragment of their Drosophila homolog, Nwk, generates ridges and scallops on liposomes and flattens, pinches, or crumples GUVs of the appropriate lipid composition Becalska AN, 2013, Kelley CF, 2015. These in vitro activities occur in the absence of the cytoskeleton or additional cellular factors. Further, when actin polymerization is inhibited in cells, the Nwk F-BAR still generates small buds, analogous to its scalloping activity in vitro Becalska AN, 2013. The cellular activity of Nwk requires both the concave surface and the tips of the canonical F-BAR, suggesting that the short additional C-terminal alpha-helical segment in our constructs is critical for F-BAR-dependent membrane bending activity, similar to SrGAPs Guerrier S, 2009. With this information, readers can make their own assessment about how the lack of activity reported by McDonald NA, 2015 from slightly shorter fragments of FCHSD1 and FCHSD2 could be related to the robust activity we have previously reported.

Second, our interpretation that the mutants generated in McDonald NA, 2015 do not uncouple membrane binding and oligomerization arises from their data showing no biochemical difference in membrane binding affinity between mutants in the basic oligomerization interface (K163E) compared to the acidic oligomerization interface (E30K, E152K) (Fig. 5D,E). Their assumption was that the acidic patch mutants would not affect electrostatic membrane binding. However, these mutants impair binding to charged membranes to the same extent as the basic patch mutants, instead of the expected intermediate membrane binding affinity if only oligomerization (and thus avidity) was affected. Since the mutants behave identically, it suggests either that membrane binding affinity and oligomerization are intrinsically coupled, or at least that these specific mutations do not uncouple them.

On 2015 Dec 31, Kathleen L Gould commented:

We appreciate the productive comments by Dr. Rodal regarding the important biological function of F-BAR proteins. We completely agree that deciphering the nature of the “non-canonical” function of various F-BAR domain proteins is an important area of future research.

Dr. Rodal takes issue with two aspects of our study: first with a lack of discussion of srGAP and Nwk literature, and second with the experimental methods used to detect membrane bending.

First, we agree that additional F-BAR proteins have been identified with “non-canonical” membrane remodeling abilities in vitro including srGAPs and Drosophila Nwk proteins. Due to space constraints we could not extensively discuss this previous work, but cited and summarized it in Table S1. We utilized the same assays which originally discovered the activities of srGAP and Nwk to conclude Cdc15 and 6 other human F-BAR domains do not bend membranes. We tested a wide range of additional binding conditions (Figure S1), including testing diverse lipid compositions (unpublished data), which resulted in identical results. Indeed, we believe these examples of non-canonical F-BAR membrane assemblies further support the idea that F-BAR domains utilize diverse modes of oligomerization for functions upon the membrane.

Second, Dr. Rodal indicates her group “published in 2013 that the F-BAR domains of FCHSD1 and FCHSD2 generate extensive membrane protrusions (to which the protein localizes) in both S2 cells and HEK cells, similar to Drosophila Nwk”. Indeed, these experiments and those with Gas7 were performed in vivo, with a full complement of cellular machinery. The results of such an in vivo experiment cannot substantiate the conclusion that these F-BAR domains physically deform the membrane. To directly test for membrane deformation, an in vitro experiment with the isolated domain is required, as we have performed (Figure 3 and S1). Dr. Rodal correctly points out the protrusions observed for FCHSD1/2 and Gas7 in vivo were actin dependent, indicating these proteins are organizing the actin cytoskeleton to generate protrusions, not necessarily directly remodeling of the membrane on their own. Furthermore, additional portions of FCHSD1/2 and Gas7 were present in the constructs used in these experiments. F-BAR domains comprise ~300-350 aa that fold into banana-shaped molecules as evidenced by >10 crystal structures. Large constructs were used in Dr. Rodal's experiments and these additional elements may confer unknown interactions and/or activities to the proteins in vivo.

In sum, though our specific interpretations may differ, we appreciate the interest in our work and encourage all researchers in this area as we together pursue this exciting area.

On 2015 Dec 27, Avital Rodal commented:

F-BAR domains: diversity in oligomerization and membrane bending activities

Avital Rodal, Brandeis University

McDonald et al. report that several oligomerizing yeast and mammalian F-BAR proteins do not form membrane tubules in vitro or in heterologous cells. They go on to show that surfaces required for oligomerization in vitro are required for the in vivo functions of several of these proteins. They conclude that oligomerization, but not membrane bending, underlies the in vivo roles of these F-BAR proteins, and that their main function is to recruit and organize other proteins at the membrane. However, their conclusion that these proteins do not bend membranes is only supported by negative results (i.e. the authors did not observe deformation in vitro or in heterologous cells), and in some cases contradict published results. Further, the authors do not discuss a previous body of literature that shows that several F-BAR proteins, including SRGAPs and FCHSD proteins, generate non-canonical (i.e. non-tubular) membrane deformations that would not have been detected in their assays.

Several groups have shown that the SRGAP family of F-BAR proteins generate negative membrane curvature (i.e. away from the protein-decorated face of the membrane) in multiple contexts: in purified systems in vitro, in heterologous cells upon overexpression, and in vivo during neuronal protrusion formation (Guerrier S, 2009; Carlson BR, 2011; Coutinho-Budd J, 2012). This activity is not consistent with tubular arrays observed for canonical F-BAR proteins (Frost A, 2008). Further, our group found that the FCHSD family of F-BAR proteins, which includes Drosophila Nervous Wreck (Nwk), exhibit an unusual higher order assembly that leads to non-tubule membrane remodeling. Using single particle EM, we showed that Nwk assembles into zig-zags on membranes instead of linear filaments typical of canonical F-BARs, and that this resulted in membrane ridges (Becalska AN, 2013). These deformations led to actin-dependent protrusions in heterologous cells, similar to a previous model for formation of cellular microspikes by the F-BAR protein syndapin (Becalska AN, 2013; Kelley CF, 2015; Shimada A, 2010). This activity does not require a novel membrane binding surface for any of these F-BAR domains. Instead these proteins use a conventional concave membrane binding surface, and appear to oligomerize into non-canonical arrays to deform membranes. This is not likely to be a special case for these specific F-BAR proteins, but rather suggests that different members of this protein family oligomerize into variable types of higher order arrays to generate and/or sense different types of membrane curvatures, which are neither tubules (the "dogma" for F-BAR domains (Traub LM, 2015)) or flat membranes (as proposed by McDonald et al.).

McDonald et al. report that the Nwk homologues FCHSD1 and FCHSD2 do not bend membranes in vitro or in cells, and state that theirs is the first study to report their activities. In fact, we published in 2013 that the F-BAR domains of FCHSD1 and FCHSD2 generate extensive membrane protrusions (to which the protein localizes) in both S2 cells and HEK cells, similar to Drosophila Nwk (Becalska AN, 2013). They may have failed to detect membrane remodeling activity for FCHSD1 and FCHSD2 in cells because their constructs omitted part of a C terminal alpha-helical extension to the F-BAR domain that is essential for function in SRGAPs (Guerrier S, 2009). Indeed, another of their non-membrane bending mammalian proteins, Gas7, has been reported to generate cellular protrusions upon full-length protein overexpression (She BR, 2002). It remains to be tested if Cdc15 or the other apparently non-membrane remodeling mammalian F-BAR proteins in their study also show activity in vivo or in vitro when a more extended region of the protein is studied.

In addition to the issue of potentially using inactive protein fragments, the specific in vitro and in vivo assays used by McDonald et al. could easily have missed non-canonical membrane bending activities. Several types of deformations are subtle on giant unilamellar vesicles (e.g. flattening, ridging, or any deformation that occurs on a ~100-200 nm scale rather than the micron scale of tubules), or are not detectable by negative stain (e.g. ridged, negatively curved, or flattened liposomes appear very similar to dried undecorated liposomes), or are unresolvable by light microscopy in cells. Cryo-EM of liposomes or thin sectioning and EM of cells is necessary to detect smaller scale deformation. Indeed, only large scale deformations like tubulation would have been detectable in the assays they used. Further, BAR domain membrane remodeling depends on a large set of parameters (Simunovic M, 2015), many of which were not tested by McDonald et al. For example, we recently showed that membrane binding and membrane deformation are not correlated, and that Nwk only deforms membranes within a limited "sweet spot" of membrane charge. This is likely dependent on F-BAR domain assembly and orientation on the membrane, which favors concave side-down under stringent binding conditions (Kelley CF, 2015). The activities of F-BAR proteins like Nwk/FCHSD1/FCHSD2 are not likely to have been detected by McDonald et al at 5% PI(4)P, the only lipid composition they tested for GUV and liposome deformation assays. Indeed, two more members of their set of six “non-deforming” F-BAR proteins, Fer and Fes, were previously shown to generate membrane tubules in vitro at 10% PI(4,5)P2 (Tsujita K, 2006; McPherson VA, 2009).

Thus, though McDonald et al. may be able to make a case against tubulation for a few of these six human F-BAR proteins (as has previously been demonstrated for both SRGAPs and Nwks), they do not test other types of membrane bending or enough parameters to conclude that these proteins do not have membrane remodeling activities. Instead, the most compelling conclusion from our work, the SRGAP work, and McDonald et al. is that F-BAR domains oligomerize on membranes into diverse higher order assemblies, and that tubular scaffolds (for which there is indeed little in vivo evidence) are just one potential way to deploy F-BAR oligomers. A non-membrane-deforming assembly, as they suggest for Cdc15, is a plausible variation on this theme for some subset of F-BAR proteins, but the limited negative data they provide are not convincing enough at this point to rule out other models, nor do they indicate that this is the rule for non-tubulating F-BAR proteins. In addition, we note that since the mutants generated in McDonald et al. do not fully uncouple membrane binding affinity from oligomerization (because the tips are part of the membrane-binding surface), an alternative model that remains consistent with all of their data is that some F-BAR domains, including Cdc15, may function as individual, non-oligomerized dimers on the membrane.

On 2015 Dec 27, Avital Rodal commented:

F-BAR domains: diversity in oligomerization and membrane bending activities

Avital Rodal, Brandeis University

McDonald et al. report that several oligomerizing yeast and mammalian F-BAR proteins do not form membrane tubules in vitro or in heterologous cells. They go on to show that surfaces required for oligomerization in vitro are required for the in vivo functions of several of these proteins. They conclude that oligomerization, but not membrane bending, underlies the in vivo roles of these F-BAR proteins, and that their main function is to recruit and organize other proteins at the membrane. However, their conclusion that these proteins do not bend membranes is only supported by negative results (i.e. the authors did not observe deformation in vitro or in heterologous cells), and in some cases contradict published results. Further, the authors do not discuss a previous body of literature that shows that several F-BAR proteins, including SRGAPs and FCHSD proteins, generate non-canonical (i.e. non-tubular) membrane deformations that would not have been detected in their assays.

Several groups have shown that the SRGAP family of F-BAR proteins generate negative membrane curvature (i.e. away from the protein-decorated face of the membrane) in multiple contexts: in purified systems in vitro, in heterologous cells upon overexpression, and in vivo during neuronal protrusion formation (Guerrier S, 2009; Carlson BR, 2011; Coutinho-Budd J, 2012). This activity is not consistent with tubular arrays observed for canonical F-BAR proteins (Frost A, 2008). Further, our group found that the FCHSD family of F-BAR proteins, which includes Drosophila Nervous Wreck (Nwk), exhibit an unusual higher order assembly that leads to non-tubule membrane remodeling. Using single particle EM, we showed that Nwk assembles into zig-zags on membranes instead of linear filaments typical of canonical F-BARs, and that this resulted in membrane ridges (Becalska AN, 2013). These deformations led to actin-dependent protrusions in heterologous cells, similar to a previous model for formation of cellular microspikes by the F-BAR protein syndapin (Becalska AN, 2013; Kelley CF, 2015; Shimada A, 2010). This activity does not require a novel membrane binding surface for any of these F-BAR domains. Instead these proteins use a conventional concave membrane binding surface, and appear to oligomerize into non-canonical arrays to deform membranes. This is not likely to be a special case for these specific F-BAR proteins, but rather suggests that different members of this protein family oligomerize into variable types of higher order arrays to generate and/or sense different types of membrane curvatures, which are neither tubules (the "dogma" for F-BAR domains (Traub LM, 2015)) or flat membranes (as proposed by McDonald et al.).

McDonald et al. report that the Nwk homologues FCHSD1 and FCHSD2 do not bend membranes in vitro or in cells, and state that theirs is the first study to report their activities. In fact, we published in 2013 that the F-BAR domains of FCHSD1 and FCHSD2 generate extensive membrane protrusions (to which the protein localizes) in both S2 cells and HEK cells, similar to Drosophila Nwk (Becalska AN, 2013). They may have failed to detect membrane remodeling activity for FCHSD1 and FCHSD2 in cells because their constructs omitted part of a C terminal alpha-helical extension to the F-BAR domain that is essential for function in SRGAPs (Guerrier S, 2009). Indeed, another of their non-membrane bending mammalian proteins, Gas7, has been reported to generate cellular protrusions upon full-length protein overexpression (She BR, 2002). It remains to be tested if Cdc15 or the other apparently non-membrane remodeling mammalian F-BAR proteins in their study also show activity in vivo or in vitro when a more extended region of the protein is studied.

In addition to the issue of potentially using inactive protein fragments, the specific in vitro and in vivo assays used by McDonald et al. could easily have missed non-canonical membrane bending activities. Several types of deformations are subtle on giant unilamellar vesicles (e.g. flattening, ridging, or any deformation that occurs on a ~100-200 nm scale rather than the micron scale of tubules), or are not detectable by negative stain (e.g. ridged, negatively curved, or flattened liposomes appear very similar to dried undecorated liposomes), or are unresolvable by light microscopy in cells. Cryo-EM of liposomes or thin sectioning and EM of cells is necessary to detect smaller scale deformation. Indeed, only large scale deformations like tubulation would have been detectable in the assays they used. Further, BAR domain membrane remodeling depends on a large set of parameters (Simunovic M, 2015), many of which were not tested by McDonald et al. For example, we recently showed that membrane binding and membrane deformation are not correlated, and that Nwk only deforms membranes within a limited "sweet spot" of membrane charge. This is likely dependent on F-BAR domain assembly and orientation on the membrane, which favors concave side-down under stringent binding conditions (Kelley CF, 2015). The activities of F-BAR proteins like Nwk/FCHSD1/FCHSD2 are not likely to have been detected by McDonald et al at 5% PI(4)P, the only lipid composition they tested for GUV and liposome deformation assays. Indeed, two more members of their set of six “non-deforming” F-BAR proteins, Fer and Fes, were previously shown to generate membrane tubules in vitro at 10% PI(4,5)P2 (Tsujita K, 2006; McPherson VA, 2009).

Thus, though McDonald et al. may be able to make a case against tubulation for a few of these six human F-BAR proteins (as has previously been demonstrated for both SRGAPs and Nwks), they do not test other types of membrane bending or enough parameters to conclude that these proteins do not have membrane remodeling activities. Instead, the most compelling conclusion from our work, the SRGAP work, and McDonald et al. is that F-BAR domains oligomerize on membranes into diverse higher order assemblies, and that tubular scaffolds (for which there is indeed little in vivo evidence) are just one potential way to deploy F-BAR oligomers. A non-membrane-deforming assembly, as they suggest for Cdc15, is a plausible variation on this theme for some subset of F-BAR proteins, but the limited negative data they provide are not convincing enough at this point to rule out other models, nor do they indicate that this is the rule for non-tubulating F-BAR proteins. In addition, we note that since the mutants generated in McDonald et al. do not fully uncouple membrane binding affinity from oligomerization (because the tips are part of the membrane-binding surface), an alternative model that remains consistent with all of their data is that some F-BAR domains, including Cdc15, may function as individual, non-oligomerized dimers on the membrane.

On 2016 Jan 04, Avital Rodal commented:

First, we would like to clarify that our FCHSD1 and FCHSD2 constructs, which generate protrusions in cultured cells, are not >500 aa as Dr. Gould suggested in her comment on 12/31. As described in Becalska AN, 2013, we quantitatively determined robust cellular activities for FCHSD1[1-417] (41 amino acids longer than the construct in McDonald NA, 2015 and for FCHSD2[1-414] (only 18 amino acids longer than the construct in McDonald NA, 2015). While we have not yet further characterized these two mammalian proteins in vitro, the same purified fragment of their Drosophila homolog, Nwk, generates ridges and scallops on liposomes and flattens, pinches, or crumples GUVs of the appropriate lipid composition Becalska AN, 2013, Kelley CF, 2015. These in vitro activities occur in the absence of the cytoskeleton or additional cellular factors. Further, when actin polymerization is inhibited in cells, the Nwk F-BAR still generates small buds, analogous to its scalloping activity in vitro Becalska AN, 2013. The cellular activity of Nwk requires both the concave surface and the tips of the canonical F-BAR, suggesting that the short additional C-terminal alpha-helical segment in our constructs is critical for F-BAR-dependent membrane bending activity, similar to SrGAPs Guerrier S, 2009. With this information, readers can make their own assessment about how the lack of activity reported by McDonald NA, 2015 from slightly shorter fragments of FCHSD1 and FCHSD2 could be related to the robust activity we have previously reported.

Second, our interpretation that the mutants generated in McDonald NA, 2015 do not uncouple membrane binding and oligomerization arises from their data showing no biochemical difference in membrane binding affinity between mutants in the basic oligomerization interface (K163E) compared to the acidic oligomerization interface (E30K, E152K) (Fig. 5D,E). Their assumption was that the acidic patch mutants would not affect electrostatic membrane binding. However, these mutants impair binding to charged membranes to the same extent as the basic patch mutants, instead of the expected intermediate membrane binding affinity if only oligomerization (and thus avidity) was affected. Since the mutants behave identically, it suggests either that membrane binding affinity and oligomerization are intrinsically coupled, or at least that these specific mutations do not uncouple them.