Disrupting the ciliary gradient of active Arl3 affects rod photoreceptor nuclear migration

Article activity feed

1. 

   Version published to 10.7554/elife.80533 on eLife

   Jan 4, 2023
2. 

   eLife

   Dec 2, 2022

   Author Response*

   Reviewer #1 (Public Review):

   ARL3 is a small GTPase that localizes to the primary cilium and plays a role in regulating the localization of some specific ciliary membrane proteins, including PDEδ and NPHP3. Mutations in this gene cause Joubert syndrome, a type of ciliopathy characterized by cerebellar malformation, and retinal degeneration. While the majority of the diseases occur in an autosomal recessive manner, two mutations in ARL3 (D67V and Y90C) have been reported to cause autosomal dominant retinal diseases. In the current paper, Travis et al. sought to understand the pathogenesis of the diseases caused by the two autosomal dominant mutations. They found that D67V acts as a constitutive active mutation, whereas Y90C is a fast-cycling mutant, which can be activated in a guanine nucleotide exchange factor (GEF) …

   Author Response*

   Reviewer #1 (Public Review):

   ARL3 is a small GTPase that localizes to the primary cilium and plays a role in regulating the localization of some specific ciliary membrane proteins, including PDEδ and NPHP3. Mutations in this gene cause Joubert syndrome, a type of ciliopathy characterized by cerebellar malformation, and retinal degeneration. While the majority of the diseases occur in an autosomal recessive manner, two mutations in ARL3 (D67V and Y90C) have been reported to cause autosomal dominant retinal diseases. In the current paper, Travis et al. sought to understand the pathogenesis of the diseases caused by the two autosomal dominant mutations. They found that D67V acts as a constitutive active mutation, whereas Y90C is a fast-cycling mutant, which can be activated in a guanine nucleotide exchange factor (GEF) independent manner. Since the fast-cycle mutant did not bind to the effector proteins in vitro (likely because the guanine nucleotide falls off from the mutant ARL3, which has a lower affinity to GDP/GTP), they developed a method to snapshot the interaction between ARL3 and its effector. Using this method, they showed that the Y90C mutant indeed has increased interaction with the effectors, suggesting that Y90C is an overactive form of ARL3. They then addressed how photoreceptor cells are affected by these two mutations using a mouse model and found that the mutations disrupt the proper migration of the photoreceptor cells.

   Strengths:

   • The paper is well written, and it was easy to understand what the authors did from the figure legends and the methods section.

   • It was easy to find out what is known or unknown, as the paper has accurate references.

   • The authors developed a method to analyze a snapshot of the interaction between ARL3 and its interactors.

   • The paper has an in vivo model and connects the biochemical characteristics of ARL3 to in vivo cellular phenotypes.

   Weaknesses:

   (1) I understand that authors focused on nuclear migration defect as the phenotype was first described in ARL3-Q71L transgenic mice. The similar phenotype observed in RP2 knockout mice further supports the idea that the defect is caused by the hyperactivation of ARL3. Indeed, the defect is not reported in the ARL3 knockout mice, however, I feel that it does not necessarily mean that the defect is not caused by loss of function. Although it has not been assessed, ARL3 knockout mice might have the same defect. Therefore, I think analyzing both the migration defect and trafficking defect would be more informative, rather than focusing on the migration defect. The fact that the relationship between nuclear migration defect and the retinal degeneration phenotype is not entirely clear further enhances the importance of analyzing the trafficking defect.

   Does the expression of ARL3-Y90C also cause the trafficking defect? If it is the case, you can separate the nuclear migration phenotype from the one caused by the trafficking defect. Would the expression of lipidated cargo(s) rescue the trafficking defect as well?

   I think many questions can be addressed by analyzing the localization of the lipidated cargos, such as PDEδ and GRK1.

   The effect of Arl3-Y90C expression on trafficking of lipidated cargos is an interesting question. Previous papers showed mislocalization of lipidated outer segment proteins in Arl3-KO rods and down-regulation or subtle mislocalization in Arl3-Q71L overexpressing rods. So, this was one of the first things we investigated; however, we never observed mislocalization of ciliary or outer segment lipidated cargos (i.e. GRK1, transducin, Rab28, and PDE) in wild type mature rods that were overexpressing Arl3 mutants, and many were tested. It was through these experiments that we first identified the pronounced nuclear migration defect. Rod photoreceptor nuclear migration is a developmental process that is completed by P10, so Arl3-Y90C overexpression is causing a developmental defect. When rods are positioning their nuclei in the ONL, they are still “immature” as their primary cilium has not begun to elaborate disc membranes for light capture. All our analysis was performed in mature rods, so it is not surprising that we did not observe any lipidated trafficking defects at this timepoint. Since the developmental timing of the nuclear migration defect is important for our manuscript, we have added this to our introduction. Additionally, we use “immature” photoreceptors for the cartoon diagrams showing how Arl3 activity is altered by different mutation and rescue experiments, since formation of the mature outer segment occurs post-migration.

   (2) I am not quite sure if the nuclear migration was assessed properly. Based on the pictures in Fig.1, some of the FLAG-negative cells also seem to be migrating to INL (please see Fig.1C and Fig.1D). Is this biologically normal during development? Could this analysis be affected by the thickness of OPL, the layer between ONL and INL? Also, the picture is cut out in the middle of INL. Could authors include more layers, such as IPL, of the retina in the picture, so that we can evaluate INL and OPL better? Taking this into account, I think it is worth measuring the nuclear position of FLAG-negative cells as a negative control in all the experiments.

   Our electroporation technique results in a small population of rods that express our constructs of interest (~5-15% with a patch). All the experiments were performed in wild type retina which develop normal retinal layers, so analysis of the nuclear position of FLAG-positive cells with the retina is cell autonomous. Migration defects are assessed by differences in the skew of FLAG-labeled rods relative to the boundaries of the wild type ONL, which is marked by Hoechst nuclear stain (also a measure of the FLAG-negative rods). Wild type photoreceptors nuclei are not found within the INL, the nuclei in that layer belong to either horizontal cells or bipolar cells both of which are not targeted by our electroporation approach. As a control, we show that when wild type Arl3-FLAG was expressed FLAG-labeled rods were never observed within the INL. We have now included the % of displaced nuclei in Table 1.

   (3) The way that the authors showed the Y90C mutant of ARL3 is a fast-cycling mutant is not very compelling. In Figure 2C, the authors showed that ARL3 Y90C can bind to PDEδ, its effector, once it is pre-loaded with GTP. The authors also showed that the mutant can bind to its effector even without EDTA as long as an excess amount of GTP is added. The authors used endogenous ARL3 as a control to compare the effects between wild-type and mutants. I see that this experiment has multiple pitfalls. First, ideally, this type of experiment needs to be done with a purified protein using fluorescent guanine nucleotide/radioactive guanine nucleotide (e.g. nucleotide loading assay or nucleotide exchange assay) to directly access the kinetics of nucleotide exchange. However, I do understand that this is out of the authors' expertise. In the authors' experimental setting, I am not sure loading the protein with GTP in the presence of the EDTA means anything more than confirming that the protein is intact. Theoretically, wild-type and a fast-cycling mutant can load GTP with similar efficiency in the presence of EDTA. Then during immuno-precipitation, GTP falls off from the Y90C mutant faster than wild-type (because a fast-cycling mutant theoretically has a lower affinity to guanine nucleotides), assuming that GTP was not added during immuno-precipitation (GTP addition was not mentioned in the method, but could authors confirm this?). But in this case, the kinetic of GTP dissociation can be affected by many factors, including the presence of GAP in the reaction, the dissociation constant of Y90C, the volume of the buffer used, and the number of washing steps. Thus, it is not very easy to estimate the difference between wild-type and Y90C. Besides, using endogenous ARL3 rather than ARL3-wild type FLAG as a control can be dangerous. I have experienced that a tagged protein is cleaved to a protein that has a similar size to endogenous protein. (I expressed GFP-protein X in knockout cells lacking protein X, and saw the band at the position where the endogenous protein is observed in wild-type cells). So, the endogenous band that the authors showed could come from the cleaved FLAG-Arl3. (Authors can easily confirm this by having wild-type not expressing FLAG-tagged ARL3, though).

   An alternative experiment that I would suggest is doing immuno-precipitation in the buffer containing: 1) no guanine nucleotide, 2) 10mM GDP, or 3) 10mM GTP in the cells expressing the following protein: 1) ARL3 wild-type FLAG, 2) ARL3 Y90C FLAG, or 3) ARL3 D129N FLAG. 10mM guanine nucleotide should be added throughout the process including washing. This experiment might also be affected by many factors, but variability should be lower than the experiment presented in Fig 2C. ARL3-wild type FLAG is also a better control here than endogenous protein.

   Variability due to the factors you mention is a concern, but we were able to repeatedly obtain the same results using our method—admittedly our method is testing whether the mutated Arl3 can exchange under a certain condition more than exactly how. We know that we are not providing precise kinetics or elucidating the underlying mechanism for how these mutations lead to what we are calling fast cycling. While that information is important, it is outside the scope of this paper.

   As you mention, an important conclusion from the PDEδ binding experiments is that we confirm the Arl3-Y90C protein is intact by showing it can indeed bind nucleotide as long as there is an excess of GTP (Fig 2B. The interesting finding from these experiments is that Arl3-Y90C binds GTP even in the presence of magnesium, a behavior not observed for wild type Arl3. We feel that showing that endogenous Arl3 is not activated in the presence magnesium in each of our preparations is a lovely internal control. However, we agree that showing wild type Arl3-FLAG in these assays is an important negative control and have now included this blot as Fig 2-Sup Fig 1.

   (4) In Fig.3, the authors attempted to take a snapshot of the interaction between ARL3 and multiple effector proteins. The three bands that were enriched in the Q71L cells were found as RP2, UNC119, and BART by mass spec (Fig.3B). These bands were used as a readout for the subsequent experiments. I am not quite sure why the authors used this approach rather than using the cell line that expresses both FLAG-ARL3 and GFP tagged protein of interest, just like what the authors did in Fig3G. The reasons why I prefer the latter approach are the following: FLAG bands that correspond to the three proteins (RP2, UNC119, and BART) in wild-type cells are very close to the detection limit, 2) authors failed to confirm that the lowest band actually comes from BART, 3) authors cannot access some important effector proteins, such as PDEδ because 293 cells might not express them. All of the problems can be solved by using the approach that was taken in Figur 3G.

   If the authors chose the former approach because of some specific reason, I would appreciate it if the authors could explain that in the main text of the paper.

   In vitro crosslinking experiments were performed to test whether overexpression of Arl3 mutants resulted in an active cellular Arl3 without artificially changing any components of the GTPase cycle. We feel these experiments are highly elegant as they allow us to take a snapshot of native Arl3 activities without compromising the analysis by artificially altering GAP/GEF/effector interactions through overexpression or during lysis (as we show that the concentration of GTP/Mg could alter interactions in Fig 2). While AD293T cells are not rod photoreceptors, we are able to use this system to better understand how the Arl3 mutants alter the level of activity within the cell. Yes, this experimental assay is novel, but we confirmed the identity of the effectors by Western and mass spec, used positive and negative controls in each experiment, and show that the method is highly reproducible. We agree with Reviewers 2 and 3 that using this method to study the cellular activity of fast cycling Arl3 mutants is a strength of our paper.

   (5) ALR3 Y90C causes nuclear migration defect. Given that Y90C is a fast-cycling mutant (hyperactive) and has a high affinity to ARL13B, the nuclear migration defect might come from either the increased activity of ARL3 or sequestration of ARL13B, which can act as a GEF for ARL3 but potentially have other functions. If my understanding is correct, the authors concluded that the defect caused by ARL3-Y90C is likely due to hyper-activation of the protein, as Y90C/T31N mutant, which cannot bind to effectors but still retains the ability to capture ARL13B, did not cause migration defect. But I am a little confused by the fact that Y90C/R149H, which is unable to bind to ARL13B (Fig.2C) but still retains the ability to interact with the effectors (Fig.3F), did not have migration defect (Fig.7B). Wouldn't this mean that the sequestration of ARL13B could contribute to the phenotype?

   If my understanding is correct, the authors are trying to say that both hyper-activation of cytosolic ARL3 and the defect in endogenous ARL3 activation in cilium is necessary to cause migration defect. I am not very convinced by this hypothesis, and still think that the defect could be caused by sequestration of ARL13B to the cytoplasm.

   Then why Y90C/T31N did not cause the defect even though they can sequester ARL13B? This might be explained by the localization of the ARL13B mutants. If Y90C can localize to the cilium while the double mutant, Y90C/T31N, does not, then only Y90C might be able to inhibit the ARL13B function in the cilium. This could explain the lack of the defect in the cells expressing Y90C/T31N.

   It would be helpful to understand how exactly the fast-cycling mutant causes the defect if the authors can provide more information, including localization of ARL3 (wild-type and mutants) as well as key proteins, such as ARL13B and the effector proteins. Assessing ARL13B defect seems to be particularly important to me because ARL13B deficiency has been connected to neuronal migration defect (Higginbotham et al., 2012)

   What I am trying to say here is that how the defect is caused is likely very complex. So, providing more information without sticking to one specific hypothesis might be important for readers/authors to accurately interpret the data.

   Our data shows that for the fast cycling Arl3-Y90C mutation both features: blocking endogenous Arl3 activation in the cilium (through Arl13B binding) and increasing activity of Arl3-Y90C in the cell body are required to produce a nuclear migration defect. We find that we can rescue migration defects by either restoring activation in the cilium or reducing GTP activity outside the cilium. As long as there is more Arl3-GTP activity in the cilium, then the rod can handle aberrant Arl3-GTP activity in the cell body. The Y90C/R149H was a critical result that led to our hypothesis that there is a gradient between the two compartments that is used for proper migration. One interesting point is that absence of any activity does not produce the migration phenotype, further suggesting that an imbalance in the gradient is important.

   We performed new experiments to investigate whether Arl3-Y90C is sequestering Arl13B away from the cilium but found that localization of Arl13B (both endogenous and overexpressed) is not altered by expression of Arl3-Y90C – see Fig 3-SupFig 1-2.

   It is an interesting question as to how different Arl3-FLAG constructs are localized within the photoreceptor. Sadly, we did not analyze the data in a way that would allow us to draw any conclusion about the localization of different Arl3-FLAG constructs. In general, we observed FLAG localization throughout the photoreceptor cell and focused our imaging on the FLAG staining around the nucleus so we could further analyze ONL position. Looking back through our images, some of mutants might have a more prominent localization within a specific subcellular compartment (e.g. Arl3-D67V is more prominent in the inner segment than outer segment and Arl3-Y90C appears to have dominant outer segment localization). Likely, these differences represent each mutant binding a particular effector: D67V to RP2 and Y90C to Arl13B, which we show biochemically. Ideally, Arl3 mutant localization would be analyzed during development to provide a more direct link to the nuclear migration defect, a future direction for our lab. We have updated our manuscript to be more transparent about the potential differences in rod localization of Arl3 mutants.

   (6) The rescue experiments that the authors presented in Fig.5-6 are striking and would build a base for future therapy of the diseases caused by ARL3 defects. However, I believe more examinations are needed to accurately interpret the data. The authors did this rescue experiment by co-injecting ARL3-FLAG and chaperons/cargos if I understand the method section correctly. But I feel we can interpret this data correctly only when ARL3-FLAG and chaperons/cargos are co-expressed in the same cells. I think a better way to analyze the data might be by comparing the nuclear migration phenotype between ARL3-FLAG only and ARL3-FLAG;chaperons/cargos double-positive cells.

   Our lab has found that the initial estimates by the Cepko Lab that co-injection of two plasmids results in above 90% of rods expressing both proteins is accurate (see reference Matsuda and Cepko PNAS 2004). Since we only assess nuclear position of FLAG-labeled rods, it is true that a small percentage of cells in this analysis express the Arl3-FLAG mutant and not the chaperone/cargo; however, inclusion of these cells really only bolsters our findings as complete rescue would likely be even more robust than measured.

   Reviewer #2 (Public Review):

   The small GTPase Arl3 (Arf-like 3) is a well-characterized component of primary cilia, including the outer segment of photoreceptors, which contain specialized cilia. Arl3 is critical for the import of multiple lipid-modified proteins into cilia that are vital to ciliary function. Human mutations in Arl3 are reported to cause both autosomal recessive and dominant inherited retinal dystrophies, but the mechanisms through which these mutations disrupt photoreceptor development are not known. Here the authors show that two dominant Arl3 mutants, Arl3-D67V and Arl3-Y90C exhibit increased activity, but for different reasons. Arl3-D67V is constitutively active (unable to hydrolyze GTP), whereas Arl3-Y90C is a classic rapid-cycling mutant, able to bind GTP spontaneously (independent of its guanine nucleotide exchange factor Arl13) but still able to complete the GTPase cycle by hydrolyzing GTP. Expression of either mutant in developing murine retinas results in a nuclear migration defect, specifically aberrant localization of rod nuclei to the inner rather than outer nuclear layer. In this sense, they phenocopy another well-characterized constitutively active mutant, Arl3-Q71L. Normal nuclear distribution could be restored by overexpression of Arl3 effectors, suggesting that the active mutants disrupt nuclear migration, at least in part, by sequestering Arl3 effectors.

   While the data are reasonably clear and convincing, there are several instances where the conclusions drawn are either confusing or problematic. Specifically:

   1. Although retinal rod cells are ciliated in their outer segment, the authors never actually examine ciliation here. Their only morphological readout is nuclear migration. How does nuclear migration failure impact ciliogenesis in the outer segment?

   Imaging was performed in mature retinas at P21 after outer segment formation is completed. Electroporation only targets a small population of cells for which we observed normal outer segments structures in all conditions tested — therefore we conclude that ciliogenesis is unaffected. Previous literature has also showed that defects in rod nuclear migration do not affect ciliation of the outer segment.

   1. The Arl3-Y90C mutant seems to act physiologically more like a dominant-negative than an activated mutant. A second mutation in Y90C (R149H) that blocks binding to the GEF Arl13 abrogates the nuclear migration defect, suggesting that Y90C is preventing activation of endogenous Arl3 by sequestering the GEF. Yet overexpression of effectors or cargos still rescues nuclear migration in the presence of Y90C, suggesting that it also sequesters effectors. How do the authors explain this?

   We agree with this interpretation. We have now included language about Arl3-Y90C’s role as a dominant negative in that it blocks Arl13B activity. The interesting caveat to this black and white usage is that blocking Arl13B would suggest a reduction in endogenous Arl3 activity in rods (which we find to be true, see Fig 5A). However, the migration defect phenotype mimics overly active Arl3 (Arl3-Q71L) and not a loss of function in Arl3 (Arl3-T31N). Using in vivo crosslinking experiments, we show that the fast cycling nature of Arl3-Y90C also causes GEF-independent activation of Arl3 (Fig 4D-E) that leads to the migration defect. Our rescue data shows that only the combination of both effects – reduced Arl3 activity in the cilium and GEF-independent Arl3 activation outside the cilium - is enough to disrupt the ciliary gradient and produce the migration defect.

   1. Fig. 1 suggests that an Arl3-T31N mutant has no phenotype. This is a canonical mutation in small GTPases that typically renders them dominant negative. The lack of phenotype is surprising since most dominant-negative mutants act by sequestering their GEFs, thereby preventing activation of the endogenous GTPase. Fig. 2C suggests that this may not be the case for Arl3-T31N, which binds Arl13 only weakly. Some of this confusion may arise from the fact that Arl13 is not a typical GEF. It is very unusual for one GTPase to directly promote nucleotide exchange on another. Does Arl3-T31N affect ciliation in the rod outer segment, or in other ciliated cells? Some discussion of this point is warranted here.

   Our paper finds that Arl3 mutants must produce an aberrant activity outside the cilium, whether through constitutive activity (seen for D67V and Q71L) or fast cycling (seen for Y90C and D129N) to cause the migration defect. Since T31N does not cause excess Arl3 activity in cells (see Fig 4) even if it does have some dominant negative activity toward Arl13B, then it is still not enough to cause the migration phenotype. This was directly tested in Fig 5, where we increase T31N binding to Arl13B by introducing Y90C/T31N and still do not see migration defect. Our results are also in line with a previous study showing that despite rapid photoreceptor degeneration in a retina-specific conditional Arl3 knockout mouse the outer segments were initially formed, in contrast the retina-specific conditional Arl13B knockout mouse did disrupt photoreceptor ciliogenesis leading to a more rapid degeneration (Hanke-Gogokhia, JBC 2017). Since complete loss of Arl3 activity did not disrupt ciliogenesis, it is unlikely that expression of Arl3-T31N in wild type retinas would alter outer segment formation, and we observed that outer segments formed in all Arl3 mutants.

   1. Oddly, Arl3-Y90C does robustly bind Arl13 (Fig. 2C), while at the same time binding to effectors (Fig. 3D/E), although less strongly than the canonical Q71L constitutively active mutant (Fig. 2A). As noted in point #2, the Y90C/R149H double mutant, which fails to bind Arl13, abrogates the nuclear migration defect observed with Y90C alone. Although the authors refer to Y90C as "rapid cycling" its phenotype is more similar to a dominant-negative than an activated mutant.

   We agree with this interpretation. We have now included language about Arl3-Y90C’s role as a dominant negative in that it blocks Arl13B activity. However, the rapid cycling behavior is important to cause the phenotype.

   1. The authors also mention that loss of Arl3 has no phenotype in their assay, however, Arl3 knockout mice exhibit severe retinal degeneration. How do they explain this?

   Our study finds that not all human Arl3 mutations will target the same cellular process even though they all result in degeneration. Arl3 knockout mice show drastic alterations in lipidated protein trafficking to the rod outer segment in mature retinas, a phenotype that we did not observe by expressing the dominant Arl3 mutants in wild type rods. Since our tools are not designed to study degeneration of rods, the precise mechanisms of degeneration caused by loss of function or dominant mutations remains to be determined. We outline some ideas in the discussion, but more work needs to be done before making any big statements regarding this. We hope that our manuscript will inspire clinicians to take a closer look at human patients to determine if there are subtle differences between disease presentation for dominant and recessive forms Arl3 inherited mutations. This is beyond the scope of our expertise.

   Reviewer #3 (Public Review):

   This work provides mechanistic insights into two recently described dominant variants of Arl3, a small GTPase, namely mutations D67V and Y90C. The authors identified a phenotype of these dominant variants during the development of rod photoreceptors by in vivo experiments in mice. They specifically observed a defect in rod nuclear migration to their final outer nuclear layer. This phenotype has been previously observed in another constitutively active variant of Arl3, Q71L. The authors performed a series of extensive and thorough biochemical assays to clarify the mode of action of these variants, mostly the Y90C variant, comparing the behavior of these variants to previously described mutants and combining multiple variants by mutagenesis. They also developed a new in vivo crosslinking strategy to be able to identify transient states of protein-protein interactions. They finally performed phenotypic rescue experiments by co-expression of various relevant proteins interacting/involved with Arl3. They finally propose a model based on differential subcellular compartmentalization of Arl3 activation which when disrupted leads to rod nuclei misplacement. These data add to the current understanding of contribution of different Arl3 variants causing human retinal degeneration, which has strong potential translational implications.

   Strengths:

   Relevance of Arl3 dominant variants to human retinal degeneration. Identification of Y90C variant as a "fast cycling" GTPase, and not as a predicted destabilizer of the protein structure.

   New method of crosslinking to enable snapshots of endogenous protein-protein interactions.

   Weaknesses:

   • The relevance of this study is justified by the fact that newly identified dominant variants of Arl3 have been associated to retinal degeneration. However, the authors never assess a degeneration phenotype.

   Electroporation technique allows for rapid expression of constructs, but the sparse expression makes it a poor means to study retinal degeneration. This is important to examine in the future using robust genetic mouse models.

   • The authors show new dominant variants of Arl3, namely Y90C and D67V, cause rod nuclear mislocalization. This phenotype is interesting but this was previously observed with other constitutively active mutation of Arl3, Q71L, and therefore is not novel.

   Yes, the Q71L paper is well cited in our manuscript and set the basis for many of our experiments.

   • The main claim of this paper is that subcellular compartmentalization of Alr3 activation to the cilium (the so called gradient by the authors) is required for proper rod nuclear migration to their final outer nuclear layer destination. The authors provide multiple experiments to support this model, but this is never directly demonstrated.

   We are not aware of any methods that could be done to directly show the subcellular localization of active Arl3-GTP within rod photoreceptors. We agree that we have provided many experiments that support our hypothesis that altering the Arl3-GTP gradient between cilium and cell body produces a nuclear migration defect. Some of our favorites include Fig 6, where we find that the migration phenotype is only rescued with expression of ciliary cargos and not rescued by non-ciliary cargos. Also, the new data requested by reviewers showing Arl13B expression in the cilium can restore the Y90C defect further supports that the Arl3 ciliary gradient is necessary for proper nuclear migration.

   Read the original source
3. 

   eLife

   Aug 16, 2022

   Evaluation Summary:

   The current paper is of interest to cell biologists studying ciliogenesis and specifically vertebrate photoreceptors, which are specialized cilia. The study shows that mutations in the small GTP binding protein ARL3 known to cause dominant inherited retinal dystrophies in humans result in ARL3 hyperactivity, disrupting the normal ciliary gradient of ARL3 activity and leading to altered retinal development. The authors demonstrate restored normal nuclear distribution by overexpression of ARL3 effectors, suggesting that the active mutants disrupt nuclear migration, at least in part, by sequestering ARL3 effectors. Overall, the experiments are properly controlled, executed, and analyzed and involve a series of extensive biochemical analyses complemented with in vivo phenotypic assessment. The development of a method to …

   Evaluation Summary:

   The current paper is of interest to cell biologists studying ciliogenesis and specifically vertebrate photoreceptors, which are specialized cilia. The study shows that mutations in the small GTP binding protein ARL3 known to cause dominant inherited retinal dystrophies in humans result in ARL3 hyperactivity, disrupting the normal ciliary gradient of ARL3 activity and leading to altered retinal development. The authors demonstrate restored normal nuclear distribution by overexpression of ARL3 effectors, suggesting that the active mutants disrupt nuclear migration, at least in part, by sequestering ARL3 effectors. Overall, the experiments are properly controlled, executed, and analyzed and involve a series of extensive biochemical analyses complemented with in vivo phenotypic assessment. The development of a method to analyze snapshots of the interaction between ARL3 and its interactors is also a strength of the paper, however, significant concerns remain regarding links between nuclear migration failure and ciliogenesis in the outer segment, and alternative possibilities that could explain the phenotype of the ARL3 Y90C mutant with respect to its sequestration of the GEF ARL13B. Addressing these major concerns would improve the manuscript and could have considerable impact on the field.

   (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

   Read the original source
4. 

   eLife

   Aug 16, 2022

   Reviewer #1 (Public Review):

   ARL3 is a small GTPase that localizes to the primary cilium and plays a role in regulating the localization of some specific ciliary membrane proteins, including PDEδ and NPHP3. Mutations in this gene cause Joubert syndrome, a type of ciliopathy characterized by cerebellar malformation, and retinal degeneration. While the majority of the diseases occur in an autosomal recessive manner, two mutations in ARL3 (D67V and Y90C) have been reported to cause autosomal dominant retinal diseases. In the current paper, Travis et al. sought to understand the pathogenesis of the diseases caused by the two autosomal dominant mutations. They found that D67V acts as a constitutive active mutation, whereas Y90C is a fast-cycling mutant, which can be activated in a guanine nucleotide exchange factor (GEF) independent manner. …

   Reviewer #1 (Public Review):

   ARL3 is a small GTPase that localizes to the primary cilium and plays a role in regulating the localization of some specific ciliary membrane proteins, including PDEδ and NPHP3. Mutations in this gene cause Joubert syndrome, a type of ciliopathy characterized by cerebellar malformation, and retinal degeneration. While the majority of the diseases occur in an autosomal recessive manner, two mutations in ARL3 (D67V and Y90C) have been reported to cause autosomal dominant retinal diseases. In the current paper, Travis et al. sought to understand the pathogenesis of the diseases caused by the two autosomal dominant mutations. They found that D67V acts as a constitutive active mutation, whereas Y90C is a fast-cycling mutant, which can be activated in a guanine nucleotide exchange factor (GEF) independent manner. Since the fast-cycle mutant did not bind to the effector proteins in vitro (likely because the guanine nucleotide falls off from the mutant ARL3, which has a lower affinity to GDP/GTP), they developed a method to snapshot the interaction between ARL3 and its effector. Using this method, they showed that the Y90C mutant indeed has increased interaction with the effectors, suggesting that Y90C is an overactive form of ARL3. They then addressed how photoreceptor cells are affected by these two mutations using a mouse model and found that the mutations disrupt the proper migration of the photoreceptor cells.

   Strengths:
   • The paper is well written, and it was easy to understand what the authors did from the figure legends and the methods section.
   • It was easy to find out what is known or unknown, as the paper has accurate references.
   • The authors developed a method to analyze a snapshot of the interaction between ARL3 and its interactors.
   • The paper has an in vivo model and connects the biochemical characteristics of ARL3 to in vivo cellular phenotypes.

   Weaknesses:
   (1) I understand that authors focused on nuclear migration defect as the phenotype was first described in ARL3-Q71L transgenic mice. The similar phenotype observed in RP2 knockout mice further supports the idea that the defect is caused by the hyperactivation of ARL3. Indeed, the defect is not reported in the ARL3 knockout mice, however, I feel that it does not necessarily mean that the defect is not caused by loss of function. Although it has not been assessed, ARL3 knockout mice might have the same defect. Therefore, I think analyzing both the migration defect and trafficking defect would be more informative, rather than focusing on the migration defect. The fact that the relationship between nuclear migration defect and the retinal degeneration phenotype is not entirely clear further enhances the importance of analyzing the trafficking defect.
   Does the expression of ARL3-Y90C also cause the trafficking defect? If it is the case, you can separate the nuclear migration phenotype from the one caused by the trafficking defect. Would the expression of lipidated cargo(s) rescue the trafficking defect as well?
   I think many questions can be addressed by analyzing the localization of the lipidated cargos, such as PDE and GRK1.

   (2) I am not quite sure if the nuclear migration was assessed properly. Based on the pictures in Fig.1, some of the FLAG-negative cells also seem to be migrating to INL (please see Fig.1C and Fig.1D). Is this biologically normal during development? Could this analysis be affected by the thickness of OPL, the layer between ONL and INL? Also, the picture is cut out in the middle of INL. Could authors include more layers, such as IPL, of the retina in the picture, so that we can evaluate INL and OPL better? Taking this into account, I think it is worth measuring the nuclear position of FLAG-negative cells as a negative control in all the experiments.

   (3) The way that the authors showed the Y90C mutant of ARL3 is a fast-cycling mutant is not very compelling. In Figure 2C, the authors showed that ARL3 Y90C can bind to PDEδ, its effector, once it is pre-loaded with GTP. The authors also showed that the mutant can bind to its effector even without EDTA as long as an excess amount of GTP is added. The authors used endogenous ARL3 as a control to compare the effects between wild-type and mutants. I see that this experiment has multiple pitfalls. First, ideally, this type of experiment needs to be done with a purified protein using fluorescent guanine nucleotide/radioactive guanine nucleotide (e.g. nucleotide loading assay or nucleotide exchange assay) to directly access the kinetics of nucleotide exchange. However, I do understand that this is out of the authors' expertise. In the authors' experimental setting, I am not sure loading the protein with GTP in the presence of the EDTA means anything more than confirming that the protein is intact. Theoretically, wild-type and a fast-cycling mutant can load GTP with similar efficiency in the presence of EDTA. Then during immuno-precipitation, GTP falls off from the Y90C mutant faster than wild-type (because a fast-cycling mutant theoretically has a lower affinity to guanine nucleotides), assuming that GTP was not added during immuno-precipitation (GTP addition was not mentioned in the method, but could authors confirm this?). But in this case, the kinetic of GTP dissociation can be affected by many factors, including the presence of GAP in the reaction, the dissociation constant of Y90C, the volume of the buffer used, and the number of washing steps. Thus, it is not very easy to estimate the difference between wild-type and Y90C. Besides, using endogenous ARL3 rather than ARL3-wild type FLAG as a control can be dangerous. I have experienced that a tagged protein is cleaved to a protein that has a similar size to endogenous protein. (I expressed GFP-protein X in knockout cells lacking protein X, and saw the band at the position where the endogenous protein is observed in wild-type cells). So, the endogenous band that the authors showed could come from the cleaved FLAG-Arl3. (Authors can easily confirm this by having wild-type not expressing FLAG-tagged ARL3, though).

   An alternative experiment that I would suggest is doing immuno-precipitation in the buffer containing: 1) no guanine nucleotide, 2) 10mM GDP, or 3) 10mM GTP in the cells expressing the following protein: 1) ARL3 wild-type FLAG, 2) ARL3 Y90C FLAG, or 3) ARL3 D129N FLAG. 10mM guanine nucleotide should be added throughout the process including washing. This experiment might also be affected by many factors, but variability should be lower than the experiment presented in Fig 2C. ARL3-wild type FLAG is also a better control here than endogenous protein.

   (4) In Fig.3, the authors attempted to take a snapshot of the interaction between ARL3 and multiple effector proteins. The three bands that were enriched in the Q71L cells were found as RP2, UNC119, and BART by mass spec (Fig.3B). These bands were used as a readout for the subsequent experiments. I am not quite sure why the authors used this approach rather than using the cell line that expresses both FLAG-ARL3 and GFP tagged protein of interest, just like what the authors did in Fig3G. The reasons why I prefer the latter approach are the following: FLAG bands that correspond to the three proteins (RP2, UNC119, and BART) in wild-type cells are very close to the detection limit, 2) authors failed to confirm that the lowest band actually comes from BART, 3) authors cannot access some important effector proteins, such as PDE because 293 cells might not express them. All of the problems can be solved by using the approach that was taken in Figur 3G.
   If the authors chose the former approach because of some specific reason, I would appreciate it if the authors could explain that in the main text of the paper.

   (5) ALR3 Y90C causes nuclear migration defect. Given that Y90C is a fast-cycling mutant (hyperactive) and has a high affinity to ARL13B, the nuclear migration defect might come from either the increased activity of ARL3 or sequestration of ARL13B, which can act as a GEF for ARL3 but potentially have other functions. If my understanding is correct, the authors concluded that the defect caused by ARL3-Y90C is likely due to hyper-activation of the protein, as Y90C/T31N mutant, which cannot bind to effectors but still retains the ability to capture ARL13B, did not cause migration defect. But I am a little confused by the fact that Y90C/R149H, which is unable to bind to ARL13B (Fig.2C) but still retains the ability to interact with the effectors (Fig.3F), did not have migration defect (Fig.7B). Wouldn't this mean that the sequestration of ARL13B could contribute to the phenotype?
   If my understanding is correct, the authors are trying to say that both hyper-activation of cytosolic ARL3 and the defect in endogenous ARL3 activation in cilium is necessary to cause migration defect. I am not very convinced by this hypothesis, and still think that the defect could be caused by sequestration of ARL13B to the cytoplasm.
   Then why Y90C/T31N did not cause the defect even though they can sequester ARL13B? This might be explained by the localization of the ARL13B mutants. If Y90C can localize to the cilium while the double mutant, Y90C/T31N, does not, then only Y90C might be able to inhibit the ARL13B function in the cilium. This could explain the lack of the defect in the cells expressing Y90C/T31N.
   It would be helpful to understand how exactly the fast-cycling mutant causes the defect if the authors can provide more information, including localization of ARL3 (wild-type and mutants) as well as key proteins, such as ARL13B and the effector proteins. Assessing ARL13B defect seems to be particularly important to me because ARL13B deficiency has been connected to neuronal migration defect (Higginbotham et al., 2012)
   What I am trying to say here is that how the defect is caused is likely very complex. So, providing more information without sticking to one specific hypothesis might be important for readers/authors to accurately interpret the data.

   (6) The rescue experiments that the authors presented in Fig.5-6 are striking and would build a base for future therapy of the diseases caused by ARL3 defects. However, I believe more examinations are needed to accurately interpret the data. The authors did this rescue experiment by co-injecting ARL3-FLAG and chaperons/cargos if I understand the method section correctly. But I feel we can interpret this data correctly only when ARL3-FLAG and chaperons/cargos are co-expressed in the same cells. I think a better way to analyze the data might be by comparing the nuclear migration phenotype between ARL3-FLAG only and ARL3-FLAG;chaperons/cargos double-positive cells.

   Read the original source
5. 

   eLife

   Aug 16, 2022

   Reviewer #2 (Public Review):

   The small GTPase Arl3 (Arf-like 3) is a well-characterized component of primary cilia, including the outer segment of photoreceptors, which contain specialized cilia. Arl3 is critical for the import of multiple lipid-modified proteins into cilia that are vital to ciliary function. Human mutations in Arl3 are reported to cause both autosomal recessive and dominant inherited retinal dystrophies, but the mechanisms through which these mutations disrupt photoreceptor development are not known. Here the authors show that two dominant Arl3 mutants, Arl3-D67V and Arl3-Y90C exhibit increased activity, but for different reasons. Arl3-D67V is constitutively active (unable to hydrolyze GTP), whereas Arl3-Y90C is a classic rapid-cycling mutant, able to bind GTP spontaneously (independent of its guanine nucleotide …

   Reviewer #2 (Public Review):

   The small GTPase Arl3 (Arf-like 3) is a well-characterized component of primary cilia, including the outer segment of photoreceptors, which contain specialized cilia. Arl3 is critical for the import of multiple lipid-modified proteins into cilia that are vital to ciliary function. Human mutations in Arl3 are reported to cause both autosomal recessive and dominant inherited retinal dystrophies, but the mechanisms through which these mutations disrupt photoreceptor development are not known. Here the authors show that two dominant Arl3 mutants, Arl3-D67V and Arl3-Y90C exhibit increased activity, but for different reasons. Arl3-D67V is constitutively active (unable to hydrolyze GTP), whereas Arl3-Y90C is a classic rapid-cycling mutant, able to bind GTP spontaneously (independent of its guanine nucleotide exchange factor Arl13) but still able to complete the GTPase cycle by hydrolyzing GTP. Expression of either mutant in developing murine retinas results in a nuclear migration defect, specifically aberrant localization of rod nuclei to the inner rather than outer nuclear layer. In this sense, they phenocopy another well-characterized constitutively active mutant, Arl3-Q71L. Normal nuclear distribution could be restored by overexpression of Arl3 effectors, suggesting that the active mutants disrupt nuclear migration, at least in part, by sequestering Arl3 effectors.

   While the data are reasonably clear and convincing, there are several instances where the conclusions drawn are either confusing or problematic. Specifically:
   1. Although retinal rod cells are ciliated in their outer segment, the authors never actually examine ciliation here. Their only morphological readout is nuclear migration. How does nuclear migration failure impact ciliogenesis in the outer segment?
   2. The Arl3-Y90C mutant seems to act physiologically more like a dominant-negative than an activated mutant. A second mutation in Y90C (R149H) that blocks binding to the GEF Arl13 abrogates the nuclear migration defect, suggesting that Y90C is preventing activation of endogenous Arl3 by sequestering the GEF. Yet overexpression of effectors or cargos still rescues nuclear migration in the presence of Y90C, suggesting that it also sequesters effectors. How do the authors explain this?
   3. Fig. 1 suggests that an Arl3-T31N mutant has no phenotype. This is a canonical mutation in small GTPases that typically renders them dominant negative. The lack of phenotype is surprising since most dominant-negative mutants act by sequestering their GEFs, thereby preventing activation of the endogenous GTPase. Fig. 2C suggests that this may not be the case for Arl3-T31N, which binds Arl13 only weakly. Some of this confusion may arise from the fact that Arl13 is not a typical GEF. It is very unusual for one GTPase to directly promote nucleotide exchange on another. Does Arl3-T31N affect ciliation in the rod outer segment, or in other ciliated cells? Some discussion of this point is warranted here.
   4. Oddly, Arl3-Y90C does robustly bind Arl13 (Fig. 2C), while at the same time binding to effectors (Fig. 3D/E), although less strongly than the canonical Q71L constitutively active mutant (Fig. 2A). As noted in point #2, the Y90C/R149H double mutant, which fails to bind Arl13, abrogates the nuclear migration defect observed with Y90C alone. Although the authors refer to Y90C as "rapid cycling" its phenotype is more similar to a dominant-negative than an activated mutant.
   5. The authors also mention that loss of Arl3 has no phenotype in their assay, however, Arl3 knockout mice exhibit severe retinal degeneration. How do they explain this?

   Read the original source
6. 

   eLife

   Aug 16, 2022

   Reviewer #3 (Public Review):

   This work provides mechanistic insights into two recently described dominant variants of Arl3, a small GTPase, namely mutations D67V and Y90C. The authors identified a phenotype of these dominant variants during the development of rod photoreceptors by in vivo experiments in mice. They specifically observed a defect in rod nuclear migration to their final outer nuclear layer. This phenotype has been previously observed in another constitutively active variant of Arl3, Q71L. The authors performed a series of extensive and thorough biochemical assays to clarify the mode of action of these variants, mostly the Y90C variant, comparing the behavior of these variants to previously described mutants and combining multiple variants by mutagenesis. They also developed a new in vivo crosslinking strategy to be able to …

   Reviewer #3 (Public Review):

   This work provides mechanistic insights into two recently described dominant variants of Arl3, a small GTPase, namely mutations D67V and Y90C. The authors identified a phenotype of these dominant variants during the development of rod photoreceptors by in vivo experiments in mice. They specifically observed a defect in rod nuclear migration to their final outer nuclear layer. This phenotype has been previously observed in another constitutively active variant of Arl3, Q71L. The authors performed a series of extensive and thorough biochemical assays to clarify the mode of action of these variants, mostly the Y90C variant, comparing the behavior of these variants to previously described mutants and combining multiple variants by mutagenesis. They also developed a new in vivo crosslinking strategy to be able to identify transient states of protein-protein interactions. They finally performed phenotypic rescue experiments by co-expression of various relevant proteins interacting/involved with Arl3. They finally propose a model based on differential subcellular compartmentalization of Arl3 activation which when disrupted leads to rod nuclei misplacement. These data add to the current understanding of contribution of different Arl3 variants causing human retinal degeneration, which has strong potential translational implications.

   Strengths:
   Relevance of Arl3 dominant variants to human retinal degeneration.
   Identification of Y90C variant as a "fast cycling" GTPase, and not as a predicted destabilizer of the protein structure.
   New method of crosslinking to enable snapshots of endogenous protein-protein interactions.

   Weaknesses:
   - The relevance of this study is justified by the fact that newly identified dominant variants of Arl3 have been associated to retinal degeneration. However, the authors never assess a degeneration phenotype.
   - The authors show new dominant variants of Arl3, namely Y90C and D67V, cause rod nuclear mislocalization. This phenotype is interesting but this was previously observed with other constitutively active mutation of Arl3, Q71L, and therefore is not novel.
   - The main claim of this paper is that subcellular compartmentalization of Alr3 activation to the cilium (the so called gradient by the authors) is required for proper rod nuclear migration to their final outer nuclear layer destination. The authors provide multiple experiments to support this model, but this is never directly demonstrated.

   Read the original source
7. 

   Version published to 10.1101/2022.05.02.490367v2 on bioRxiv

   May 23, 2022
8. 

   Version published to 10.1101/2022.05.02.490367v1 on bioRxiv

   May 3, 2022