Genetic specification of left–right asymmetry in the diaphragm muscles and their motor innervation

Essential revisions:

1) Addition of supplemental imaging controls to demonstrate the reproducibility and quantification schemes used to assess L-R asymmetry of the phrenic nerves across animals:a) Some of the left-right traces do not incorporate all of the projection data, which may inflate the matches/discrepancies in the overlay of L vs R nerves. Showing sample variability would strengthen the rigor of the analyses.

b) The methods used to score the branching patterns of phrenic axons are not consistent. The red-green trace projections lack branches and details that are visible in the primary data images (e.g., the control nerves in with Figure 4G). It would be helpful to see more examples of the raw and "traced" data that were used to generate the data plots, as supplemental data. One suggestion for appreciating the stereotypy of projections of left and right nerves from embryo to embryo is to overlay the traces from multiple specimens. This would help readers assess the background level of deviation from nerve to nerve.

We fully understand these concerns. It is effectively a major point to make it clear to the referees and the readers that all our quantifications were done on immunolabeled diaphragms, and not based on the traces that we showed in the pictures. Indeed, the traces were thought to help the reader identifying the branches of interest and to better appreciate the general phrenic nerve pattern of the primary and secondary order branches that innervate the lateral muscles. We thought that simplified traces also nicely highlight the different angles by which left and right primary nerves split into dorsal and ventral branches (T and V shapes).

In the revised version, we rearranged the Figure 1 and its supplements to provide additional illustrations to clarify these points.

To better explain which branches were considered, we included an image of the color-coded traces of lateral and crural branches drawn from Neurofilament staining of wholemount diaphragm (Figure 1—figure supplement 1A). We also added an image describing the method used for quantification in Figure 1—figure supplement 1B.

To demonstrate the stereotypy of the left-right asymmetry that we measured, we incorporated in Figure 1D-E the histograms showing the average value measured for L and R phrenic nerves together with the SEM, which illustrates the variation observed between individual embryos for the 2 parameters measured. In addition, we added in Figure 1—figure supplement 3B a ladder graph showing the paired left and right values of the defasciculation distance for eight E14.5 embryos. As suggested, we now provide several examples of wholemount diaphragms labeled with neurofilament with their respective branch traces (Figure 1—figure supplement 3C). We hope this will better document the stereotypy of the asymmetry.

2) Mechanisms of how enrichment of the short isoform of Robo 1 specifies a different phrenic nerve pattern on the right side:

a) An important further inquiry, which could be done with the mice in hand, is to determine whether cleavage of Robo1 on the left vs. right populations of motor neurons altered in MMP2 mutants. This would bolster the conclusion that differential levels of MMP2 are crucial for setting the L-R differences in Robo1 processing. As they currently stand, the data connecting MMP2 to Robo1 is based solely on the MMP2 expression pattern and gain of function experiments, which are suggestive yet not definitive.

We fully understand the relevance of this concern and agree on the fact that we did not provide a direct link between Robo protein forms and MMP2. We worked in two directions to strengthen our conclusions. We first thought that a quite feasible way to document this link would be to show that manipulating endogenous MMP activity in the Left and Right motoneuron fraction impacts on Robo processing. To do so, we dissected cervical ventral spinal cord tissue enriched in phrenic motoneurons from HB9+ embryos. First Left and Right tissues were let for two hours in zymography buffer which preserves MMP activity and then processed for immunoblot to compare Robo protein forms. First, we found that the Right/Left ratio of Robo cleaved form was maintained during the incubation, further supporting that higher MMP activity in the Right population results in higher rate of Robo processing. Second, we inhibited endogenous MMP activity in the motoneuron tissues using GM6001. Under this condition, short forms of Robo1 are not anymore Right-enriched. We hope that these experiments further show that differential MMP activity can generate a different pattern of Robo processing in the left and in the right cervical motoneuron population.

The second direction, suggested by the referees, was to examine whether the L/R ratio of Robo processing is modified in MMP2 mutants. Although very pertinent, we anticipated that it would be highly challenging to provide significant conclusion, using western blot approach, which is the only way for distinguishing long and short Robo forms. First, the amplitude of the L/R difference is narrow. Second MMP mutants are reported to show high levels of compensation (Holmbeck et al., 2004; Kukreja et al., 2015) and we have observed during the analysis that levels of phrenic nerve symmetrization in MMP2 mutants is variable. Nevertheless, we conducted these experiments and crossed MMP2 and HB9+ lines to highlight the motoneuron population for our dissection procedure. Two independent littermates of 5 embryos each were analyzed and L/R ratio was reduced, compared to wild- types. This analysis thus provides additional support for linking MMP activity and Robo processing. We would like to stress that given the nature of the samples, it is necessary to pool a substantial number of embryos. Reaching statistical significance would mean replicating these experiments at least 10 times, which was not feasible during the revision timing and indeed incompatible with the fact that these experiments are performed on knockout tissue.

Nevertheless all together, we are convinced that these additional results strengthen our interpretation and hope they now provide convincing support for contribution of MMP2 to the differential processing of Robo1 in left and right motoneurons.

b) The authors should more directly state how they envision the signaling to work. Is asymmetry, as presented in the quantification, a reflection of left-right differences in branching and outgrowth? Robo1 has previously been implicating in branching/outgrowth. They might comment on how their results compare to data in "Biochemical purification of a mammalian slit protein as a positive regulator of sensory axon elongation and branching", Cell 96, 771-784 (1999). This study is in line with the descriptive observations in the present manuscript.

We agree that several properties have been assigned to Slits, which need to be considered in our interpretation of the Slit/Robo function in the setting of Left and Right phrenic innervation and are now discussed in the revised manuscript (Discussion section). Indeed, when reaching the diaphragm, the phrenic nerves undergo successive defasciculations, leading to the individualization of axonal bundles, referred to as “branches”. This term gives confusion because these bundles are not “branches per see” meaning produced by sprouting of novel axon from pre-existing one. In the terminal step of the innervation, individual axons undergo branching, but this is a stage that we did not investigate in our study, concentrating on the initial defasciculation processes. Therefore, we believe that the asymmetry of phrenic patterns reported here result from differential regulation of axon defasciculation and growth, rather than branching per se. This interpretation is coherent with the study by Jaworski and Tessier-Lavigne (Jaworski and Tessier-Lavigne, 2012), reporting that defasciculation is affected by loss of Slit-Robo function, both receptors and ligands being produced by the motoneurons themselves. The simplest mechanistic model would be that Slit-Robo interactions between phrenic axons promote fasciculation. This signaling is higher in the Left than in the Right population, due to higher Abl activity and lower Robo processing in the Left than in the Right. We do not exclude that the Slit-Robo signaling could also regulate other aspects, such as axon speed, orientation of growth in the target, and terminal branching, but studying these contributions were the main focus of the present study.

c). Comment on the fact that there seem to be no L-R Robo1 ligand differences.

We now introduced a section of discussion in our manuscript, in which we quote that an asymmetric signaling can be generated with equivalent initial levels of Slits.

3) One of the interesting aspects is analysis of differences in fasciculation and growth behavior of phrenic motor neurons; unfortunately, these data are relegated to Figure 3—figure supplement 1. The methods are given only in the Figure legend and not at all in the Methods for the Supplementary Figures but are quite unintelligible. The differences seem to be great, and although it is unclear whether Robo-Slit signaling would be involved in fasciculation as proposed by Jaworski and Tessier-Lavigne in their earlier study, these data are of interest given the L/R asymmetry issue and are therefore important data for the present study.

Additional information has been added in the method section, in a separate paragraph, so the reader can easily have access to our method of quantification.

Even if the data stay in Supplementary data, issues include: why do axons extend only from one aspect of the explant? In f. why was only one aspect of the explant measured for bundle width, and what is the genotype of the explant? Also statistical analysis is not well stated for this aspect of the study.

Our figure was unclear since it induced confusions. Indeed, quantifications have been made on the whole explants, without any bias of choices. The magnified panels showing only part of the explants were presented to more clearly illustrate the different fasciculation aspect, which was not visible in pictures of whole explants. We now clarified the figure legend and stated in the experimental procedure that measures were done all around the explants. The statistics have been inserted in the corresponding figure legend (and the transparent form).

4) There is a noticeable gap in connecting the very apt initial observations of changes in phrenic nerve asymmetry in Pitx2 and Rfx3 mutants to the later findings of changes in motor neuron expression of MMP2, Robo1 processing, etc. Assessment of changes in the L-R differences in MMP2 and Robo1 processing in Pitx2 and Rfx3 mutants would improve the cohesion of the studies. If MMP2/Robo1 processing are involved, there should be an equilibration of MMP2/Robo1 levels in left vs right phrenic motor neurons in Pitx2 and Rfx3 mutants.

Unfortunately, we have no access anymore to the Pitx2 colony. Fixed embryos were provided by our collaborator, Dr D. Martin, who did not maintain the colony in her lab. Nevertheless, to address this point, we concentrated on RFX3 mutant embryos. We had to consider several constraints. First, due to the partial penetrance of the phenotype, embryos must be considered individually. Second, we need to focus on MMP activity in motoneurons, excluding MMP activity in other populations of the ventral spinal cord, such as the progenitors that express it at high levels. This excludes the approaches of western blot and RTPCR because we cannot separate progenitors and motoneurons in the dissection procedure.

We therefore selected the technique of in situ zymography, that is very sensitive, in dissociated motoneuron cultures, that allow assessing MMP activity of motoneurons isolated from embryos that we previously identified as affected, based on their lung isomerism. The analysis demonstrated that L/R difference of MMP activity is gone in these embryos. These data have been inserted in Figure 5E.

These new data thus establish a link between early L/R specification and the molecular effectors of L/R asymmetry in phrenic motoneurons, which we hope meets the expectation of the referees.

[Editors' note: further revisions were requested, as described below.]

[…] The reviewers, Reviewing Editor, and Senior Editors agree that valuable observations are presented in the first part in Figures 1–4 – on patterns of branching, that asymmetry is controlled by Nodal signaling and seen in "intrinsic" differences in phrenic motoneurons, with the short isoform of Robo enriched on the right side to presumably control fasciculation differently than on the left. These are data well worth publishing and have merit even if the mechanistic aspect is unclear. Ordinarily, as with many of the other prominent journals, for the manuscript is to be acceptable, the authors would need to additionally provide mechanism. We understand that not all studies can end with a definitive mechanism. In repeated consultation with the reviewers, the Reviewing Editor, and the Senior Editor, we initially considered recommending that you either omit the experiments that are not substantiated by the statistical analyses, or describe the results as evidence that your hypothesis was not supported by the data. In either case, this would diminish the offering of the study for eLife, and we imagine would not be agreeable to you. Thus, the end, we have decided that the paper with its current components and incomplete outcome of analyses is not acceptable for publication in eLife. Instead, we encourage that you rethink the study with sufficient statistical power to ensure that you will have adequate sample size for a strong conclusion on L-R differences in Robo1 processing through MMP2, the proposed route through which L-R axon asymmetry could be achieved, either way.

Two major issues were raised by the referees: i) lack of statistical significance for the L/R difference in Robo forms observed in immunoblots of motoneuron-containing tissue, and ii) lack of statistical significance for the analysis of Robo processing by MMP2. We have addressed these concerns as follows.

i) Lack of statistical significance for the L/R difference in Robo forms observed in immunoblots of motoneuron-containing tissue.

The data presented in our prior manuscript were not statistically significant and therefore did not support our conclusion of a L/R difference in Robo forms in spinal cord motor neurons. We have now recognized that this was due to an inadequate method of analysis. To address this issue, and as recommended in the literature (Degasperi et al., 2014), we have now normalized the L/R data for comparison of L/R data between different western-blots and obtained statistical significance. The source data file provided with the new version of the manuscript contains the raw and normalized data. To consider that data obtained by immunoblotting may not follow the normal distribution assumption required for the t-test that is classically used (Kreutz et al., 2007), we also performed a Mann-Whitney test. Both tests yielded similar and significant p values (p= 0.016200 t-test; p=0.015873 Mann-Whithey). Thus, we believe that the statistical analysis of our data now supports the conclusion that Robo forms differ in left and right motoneuron-containing spinal cord tissue.

ii) Lack of statistical significance for the analysis of Robo processing by MMP2.

Even though MMP can process Robo (Seki et al., 2010), our analysis of Robo processing by MMP2 in spinal cord tissue did not reach statistical significance. This may have several reasons, including the likely possibility that MMP2 is not the only protease involved in generating left-right asymmetries of Robo forms in spinal cord tissue. Accordingly, our microarray analysis showed that the genes for 7 MMPs and 13 ADAMs metalloproteinases are expressed in left and right motoneuron-containing spinal cord tissue. This includes Adam10, whose homolog in Drosophila, Kusbanian, has been implicated in Robo processing (Coleman et al., 2010). Moreover, different MMPs can cleave the same protein (Kukreja et al., 2015; Prudova et al., 2010) and some of the short Robo form is preserved in MMP2 mutant tissue.

Finally, we had not previously explained the consideration that MMP2 has several known targets in addition to Robos (Dean and Overall, 2007), and several of these are also expressed by phrenic motoneurons and involved in their patterning, such as NCAM (Allan and Greer, 1998; Dean and Overall, 2007). Accordingly, the defect observed after MMP2 genetic deletion might affect alternative, or additional signaling pathways involved in axon guidance.

For this reason, we have revised our manuscript to present the findings on the Robo and MMP2 pathways in phrenic nerve left-right asymmetries independently from each other and only discussed a possible link. Accordingly, we have presented two examples of signaling pathways that contribute to the generation of left-right asymmetries in motor neurons. This revised format should better reflect the complexity of the signaling mechanisms underlying the establishment of left and right phrenic nerve patterns and whilst opening up the field to further investigation into the mechanisms underlying Robo processing (Blockus and Chédotal, 2016).

In summary, we have revised our manuscript substantially to underscore how a left-right genetic program initiated by Nodal sets into motion a complex signaling network that controls a range of events required for the stereotypic and asymmetric pattern diaphragm innervation.

List of major changes in the revised manuscript:

1) Improved statistical analysis of left and right differences in Robo forms detected by immunoblotting, including corresponding revision of the results and experimental procedures as well as data representation (Figure 5B).

2) Differences in the MMP and Robo pathways are now presented independently, and the wording that bridges the Robo and MMP2 studies has accordingly been modified.

3) The Discussion has been modified in several ways. Firstly, the paragraph on Slit/Robo signaling was changed to better discuss potential roles of Robo1 asymmetric processing. Secondly, a new section has been added to discuss the role of MMP2, including its several substrates, and to explain why MMP2 genetic loss is expected to lead to only a partial symmetrisation.

4) The analysis of ABL1 levels and Abl1 mutant phenotype has been removed, as ABL1 is not a mediator only of the Slit-Robo pathway, but operates downstream of several signaling receptors.

[Editors' note: further revisions were requested, as described below.]

The reviewers, both new, feel that your presentation in this third submission is much improved over the previous two versions, which they have viewed. The reviewers thought your findings on left-right asymmetry in phrenic nerve innervation of the mouse diaphragm are important, particularly the nodal-mediated signaling of the asymmetry, the anatomical differences in the fasciculation differences in vitro, and the molecular screen pinpointing MMP processing of the Robo receptor to effect the l-r differences. However, there are several issues that need to be addressed before acceptance, as outlined below and in the appended reviews:

In your last rebuttal and in the text, you explain two major amendments: First, you state that you addressed the lack of statistical significance in the L/R difference in Robo forms in immunoblots, and we acknowledge your efforts. However, one of the present reviewers criticized the western blot analysis with the new statistical test used (based on the Degaspari method) and considered the normalization strategy invalid. Upon normalizing the levels of the short band to the full length band, one would predict that if the short form derives from the long form, the levels of the long form would decrease and the short form would increase. Please comment.

We understand the concern of the referee. A normalization is obligatory to eliminate the experimental variability. Normalization to the full-length form is classically used for assessing protein processing (Gatto et al., 2014; Lin et al., 2008). Nevertheless, another option is to normalize to the tubulin. We obtained the same ratio when the normalization is done using the tubulin. Therefore, the variability of the long form between lines is mainly due to differences of loading. We thus kept in the principal figure the normalization on the protein of interest, and present the normalization to tubulin in the associated supplemental figure (Figure 5—figure supplement 1D).

More generally we agree with the referee that the short Robo forms might derive from the long Robo form. Nevertheless, we would like to highlight that such linear relationship between long and short forms cannot be deduced from western blot analysis. Indeed, it is not known whether the short and long forms have the same stability and half-life. Moreover, it is not known whether the Robo1 antibody recognizes with the same efficiency the long and the short forms.

Here are some additional elements on the western blot normalization. We normalized the lines within for each blot to take into account the variability of samples and deposition. This normalization enable comparison between left and right samples, but the normalized values vary between different Western blots for several reasons, which explains why a normalization between different blots is needed. First each western blot represents about 7 different embryos, individually dissected. This might generate some heterogeneity in the biological samples. Second, the re-use of antibodies and reagents to reveal the proteins might induce differences of band intensity between different western blots. Normalizing the western blots do not change the left-right ratio but allowed taking into account in the determination of the variability the heterogeneity between samples and reagents.

Second, your previous analysis of Robo processing by MMP2 in spinal cord tissue did not reach statistical significance. As with the previous reviewers, the present reviewers stressed that analyzing the mutants is a key experiment, and that it is crucial to your argument as such data would demonstrate a biochemical link between MMP2 activity and Robo1 receptor cleavage. Increasing the "n" would strengthen your case, but in a potentially "dangerous' direction, one of seeking the desired result and this route would be a lengthy as well.

You observed a partial symmetrization of the phrenic branches, with a right pattern that resembles the left pattern in control littermates in E14.5 Mmp2-/- embryos. But now you take the stance that the defect observed with MMP2 genetic deletion might affect alternative, or additional signaling pathways in axon guidance and that MMP2 may be one of several proteases, and thus chose to present the findings on the Robo and MMP2 pathways in phrenic nerve left-right asymmetries "independently" from each other, suggesting but not concluding that this relationship may be linked. While this does not solidify "mechanism", given the rounds of work you have done to improve the manuscript, we now feel that your revision is acceptable. However, we request that you acknowledge overtly that your attempts to demonstrate Robo1 processing in MMP2 mutants lacked significance, and shorten the text in the Discussion where you discuss multiple MMP2 targets and proteases.

We understand that the referees wish that we mention in the manuscript our attempt to link MMP2 and Robo1 and the lack of statistical significance of the data. We have now added in the Discussion a paragraph stating:

– First that incubation of cervical spinal cord tissue with active MMP2 significantly increased the short Robo1 forms (using the same normalization method as for assessing the left-right differences) (Supplementary File 1)

– And second that however, our analysis of Robo1 in cervical motoneuron tissue from Mmp2 mutants did not allow to detect a significant difference. (Supplementary File 2)

We also shortened our Discussion on the contribution of other protease pathways.

In addition to these two major points of contention by the various the reviewers, as often happens with re-review by a new set of reviewers, additional experiments were suggested. One suggestion is to determine whether the cleaved Robo1 receptors have any guidance or fasciculation function. Your explant system would be ideal to investigate how MMP2 mutant motor neurons respond to Slits compared to wild type neurons after bath application or by co-culture with Slit-expressing cells, an analysis that would strengthen your hypothesis and data on the asymmetric axon innervation patterns in vivo. However, this would also take time and hopefully could be the basis of "next-step" experiments.

We fully agree that investigating the functional properties of the balance of short and long Robo forms is an important step towards understanding the Slit-Robo signaling. Effectively this is tricky because MMP2 loss of function does not produce an optimal outcome on the balance. We have no entry point, neither from our own data nor from the literature, into how the Robo forms are produced and regulated, that would allow a straightforward investigation. This is also an additional reason for why this investigation certainly requires a fully dedicated investigation program.