1. Neuroscience
Download icon

Endoglycan plays a role in axon guidance by modulating cell adhesion

  1. Thomas Baeriswyl
  2. Alexandre Dumoulin
  3. Martina Schaettin
  4. Georgia Tsapara
  5. Vera Niederkofler
  6. Denise Helbling
  7. Evelyn Avilés
  8. Jeannine A Frei
  9. Nicole H Wilson
  10. Matthias Gesemann
  11. Beat Kunz
  12. Esther T Stoeckli  Is a corresponding author
  1. Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Switzerland
Research Article
  • Cited 0
  • Views 618
  • Annotations
Cite this article as: eLife 2021;10:e64767 doi: 10.7554/eLife.64767

Abstract

Axon navigation depends on the interactions between guidance molecules along the trajectory and specific receptors on the growth cone. However, our in vitro and in vivo studies on the role of Endoglycan demonstrate that in addition to specific guidance cue – receptor interactions, axon guidance depends on fine-tuning of cell-cell adhesion. Endoglycan, a sialomucin, plays a role in axon guidance in the central nervous system of chicken embryos, but it is neither an axon guidance cue nor a receptor. Rather, Endoglycan acts as a negative regulator of molecular interactions based on evidence from in vitro experiments demonstrating reduced adhesion of growth cones. In the absence of Endoglycan, commissural axons fail to properly navigate the midline of the spinal cord. Taken together, our in vivo and in vitro results support the hypothesis that Endoglycan acts as a negative regulator of cell-cell adhesion in commissural axon guidance.

Introduction

Cell migration and axonal pathfinding are crucial aspects of neural development. Neurons are born in proliferative zones from where they migrate to their final destination. After arrival, they send out axons that have to navigate through the tissue to find the target cells with which they establish synaptic contacts. Intuitively, it is clear that the same cues provided by the environment can guide both cells and axons to their target. Although we know relatively little about guidance cues for cells compared to guidance cues for axons, both processes are dependent on proper cell-cell contacts (Gomez and Letourneau, 2014; Short et al., 2016).

One of the best-studied model systems for axon guidance are the commissural neurons located in the dorsolateral spinal cord (de Ramon Francàs et al., 2017; Stoeckli, 2017 and Stoeckli, 2018). These neurons send out their axons toward the ventral midline under the influence of the roof plate-derived repellents BMP7 (Augsburger et al., 1999) and Draxin (Islam et al., 2009). At the same time, axons are attracted to the floor plate, their intermediate target, by Netrin (for a review on Netrin function, see Boyer and Gupton, 2018), VEGF (Ruiz de Almodovar et al., 2011), and Shh (Yam et al., 2009 and Yam et al., 2012). At the floor-plate border, commissural axons require the short-range guidance cues Contactin2 (aka Axonin-1) and NrCAM to enter the midline area (Stoeckli and Landmesser, 1995; Stoeckli et al., 1997; Fitzli et al., 2000; Pekarik et al., 2003). Slits and their receptors, the Robos, were shown to be required as negative signals involved in pushing axons out of the midline area (Long et al., 2004; Blockus and Chédotal, 2016; Pignata et al., 2019). Members of the Semaphorin family are also involved in midline crossing, either as negative signals mediated by Neuropilin-2 (Zou et al., 2000; Parra and Zou, 2010; Nawabi et al., 2010; Charoy et al., 2012), or as receptors for floor-plate derived PlexinA2 (Andermatt et al., 2014a). Once commissural axons exit the floor-plate area, they turn rostrally along the longitudinal axis of the spinal cord. Morphogens of the Wnt family (Lyuksyutova et al., 2003; Domanitskaya et al., 2010; Avilés and Stoeckli, 2016) and Shh (Bourikas et al., 2005; Wilson and Stoeckli, 2013) were identified as guidance cues directing post-crossing commissural axons rostrally. In the same screen that resulted in the discovery of Shh as a repellent for post-crossing commissural axons (Bourikas et al., 2005), we found another candidate that interfered with the rostral turn of post-crossing commissural axons. This candidate gene was identified as Endoglycan.

Endoglycan is a member of the CD34 family of sialomucins (Nielsen and McNagny, 2008; Sassetti et al., 2000; Furness and McNagny, 2006). The family includes CD34, Podocalyxin (also known as Thrombomucin, PCLP-1, MEP21, or gp135), and Endoglycan (also known as Podocalyxin-like 2). They are single-pass transmembrane proteins with highly conserved transmembrane and cytoplasmic domains. A C-terminal PDZ recognition site is found in all three family members (Furness and McNagny, 2006; Nielsen and McNagny, 2008). The hallmark of sialomucins is their bulky extracellular domain that is negatively charged due to extensive N- and O-glycosylation. Despite the fact that CD34 was identified more than 20 years ago, very little is known about its function. It has been widely used as a marker for hematopoietic stem cells and precursors. Similarly, Podocalyxin is expressed on hematopoietic stem and precursor cells. In contrast to CD34, Podocalyxin was found in podocytes of the developing kidney (Kerjaschki et al., 1984; Doyonnas et al., 2005; Furness and McNagny, 2006). In the absence of Podocalyxin, podocytes do not differentiate, resulting in kidney failure and thus perinatal lethality in mice (Doyonnas et al., 2001). Podocalyxin, but not CD34, is expressed widely in the developing and mature mouse brain (Vitureira et al., 2005; Vitureira et al., 2010). Podocalyxin was shown to induce microvilli and regulate cell-cell adhesion via its binding to the NHERF (Na+/H+ exchanger regulatory factor) family of adaptor proteins that link Podocalyxin to the actin cytoskeleton (Nielsen et al., 2007; Nielsen and McNagny, 2008; Nielsen and McNagny, 2009). Like Podocalyxin, Endoglycan is expressed in the brain and in the kidney. Only low levels were found in hematopoietic tissues (Sassetti et al., 2000). To date, nothing is known about the function of Endoglycan.

Based on its temporal and spatial expression pattern, we first analyzed the function of Endoglycan in the embryonic chicken spinal cord. In the absence of Endoglycan, commissural axons failed to turn rostrally upon floor-plate exit. Occasionally, they were observed to turn already inside the floor-plate area. Furthermore, the trajectory of commissural axons in the midline area was tortuous in embryos lacking Endoglycan, but straight in control embryos. Live imaging data of dI1 axons crossing the floor plate confirmed changes in axon – floor plate interaction. In addition, axons were growing more slowly after silencing Endoglycan and faster after overexpression of Endoglycan. In vitro assays confirmed a lower adhesive strength of growth cones to a layer of HEK cells expressing Endoglycan. Similarly, adhesion between growth cones overexpressing Endoglycan and control HEK cells was reduced, in agreement with a model suggesting that Endoglycan acts as a negative regulator of cell-cell adhesion during axon guidance.

Results

Endoglycan was identified as a candidate guidance cue for commissural axons

In a subtractive hybridization screen, we identified differentially expressed floor-plate genes as candidate guidance cues directing axons from dorsolateral commissural neurons (dI1 neurons) along the longitudinal axis after midline crossing (Bourikas et al., 2005; see Materials and methods). Candidates with an expression pattern that was compatible with a role in commissural axon navigation at the midline were selected for functional analysis using in ovo RNAi (Pekarik et al., 2003; Wilson and Stoeckli, 2011). One of these candidates that interfered with the correct rostral turning of commissural axons after midline crossing turned out to be Endoglycan, a member of the CD34 family of sialomucins.

CD34 family members share a common domain organization that consists of a mucin-like domain followed by a cysteine-containing globular domain, a membrane associated stalk region, a transmembrane spanning domain and the cytoplasmic domain (Figure 1; Sassetti et al., 2000; Furness and McNagny, 2006; Nielsen and McNagny, 2008). With the exception of the mucin-like domain at the N-terminus, the conservation between species orthologues of CD34, Endoglycan and Podocalyxin is in the range of 80%, but drops below 40% within the mucin domain. Homologies of these paralogous proteins within the same species are generally only in the range of 40% (Figure 1), demonstrating that, while they might share a similar overall structure, the structure can be built by quite diverse primary amino acid sequences.

Domain organization, exon alignment, and conservation of the CD34 family of sialomucins.

(A) Schematic drawing of the domain organization of CD34 family members. They all contain an N-terminal, highly glycosylated, mucin domain (green), a cysteine-containing globular domain (CGD, yellow), a juxtamembrane stalk region (SR, brown), as well as a transmembrane alpha-helix (TM, red) and a cytoplasmic domain (CD, blue). O-linked glycosylation sites within the mucin domain are depicted in light blue, whereas further sialylated residues are symbolized in purple. N-linked glycosylation sites are not shown. Note, that the indicated glycosylation sites in this scheme are only symbolizing the extensive amount of glycosylation in sialomucins and are not representing the actual position of glycosylation. (B) Exon organization and domain conservation of sialomucins. Transcripts of CD34 family members are encoded by eight separate exons (colored boxes). While the length of exons coding for the cysteine containing globular domain (yellow), the juxtamembrane stalk region (brown), and the cytoplasmic domain (blue) are more or less conserved (exon sizes are given within the boxes), exons coding for the mucin domain (green) vary markedly in their length and organization. The translational start sites are highlighted by the ATG and the end of the coding sequences are indicated by the given Stop codon. Protein homology between the different chicken sialomucins is depicted by the large numbers between the exon pictograms. All domains were compared separately and the colors used indicate the corresponding domains. The first number indicates identical amino acids between the compared proteins, the second number represents conserved residues and the number in brackets designates the amount of gap positions within the alignment of the domains (e.g. 28/44 (30)). The alignment was done using MUSCLE version 3.7 configured to the highest accuracy (Edgar, 2004). Single gap positions were scored with high penalties, whereas extensions of calculated gaps were less stringent. Using such parameters homologous regions of only distantly related sequences can be identified. Note, that within the mucin domain only some blocks, interspaced by sometimes large gap regions, are conserved between the different proteins.

Endoglycan was expressed mainly in the nervous system during development, as levels in non-neuronal tissues were much lower (Figure 2 and Figure 2—figure supplement 1). In the neural tube, Endoglycan was expressed ubiquitously, including floor-plate cells at HH21 (Hamburger and Hamilton stage 21; Hamburger and Hamilton, 1951; Figure 2). By HH25, expression was still found throughout the neural tube with higher levels detected in dorsal neurons (including dI1 neurons) and motoneurons. Endoglycan expression was also maintained in the floor plate (Figure 2B). For functional analysis, dsRNA was produced from the Endoglycan cDNA fragment obtained from a screen and used for in ovo electroporation of the spinal cord at HH18 (Figure 2D). The analysis of commissural axons’ trajectories at HH26 by DiI tracing in 'open-book' preparations (Figure 2E; quantified in Figure 2L) revealed either failure to turn or erroneous caudal turns along the contralateral floor-plate border in embryos lacking Endoglycan in the floor plate (Figure 2G), in only the dorsal spinal cord (Figure 2J), or in one half of the spinal cord, including the floor plate (Figure 2H,I). Furthermore, axons were occasionally found to turn prematurely either before midline crossing or within the floor-plate area. Detailed analysis of the axonal morphology in the floor-plate area revealed a tortuous, ‘corkscrew’-like trajectory in embryos lacking Endoglycan in dI1 neurons and the floor plate (Figure 2I), whereas axons crossed the midline in a straight trajectory in untreated control embryos (Figure 2F) and in embryos injected and electroporated with dsRNA derived from either CD34 (not shown) or Podocalyxin (Figure 2K).

Figure 2 with 3 supplements see all
Endoglycan is required for correct turning of post-crossing commissural axons.

(A,B) Endoglycan is expressed in the developing neural tube during commissural axon guidance. Endoglycan is expressed throughout the neural tube at HH21, including the floor plate (white arrowhead) (A). (B) At HH25, Endoglycan is still found in most cells of the spinal cord. High levels are found in motoneurons and interneurons, including the dorsal dI1 neurons (black arrowhead), and in the floor plate (white arrowhead). No staining was found when hybridization was carried out with a sense probe (C). Commissural axon pathfinding was analyzed in ‘open-book’ preparations (D,E; see Materials and methods for details). The positions of the electrodes for dorsal and ventral electroporation are indicated (D). In control embryos at HH26, commissural axons have crossed the floor plate and turned rostrally along the contralateral floor-plate border (F). In contrast, after downregulation of Endoglycan (G–J) commissural axons failed to turn along the contralateral floor-plate border or they turned randomly either rostrally or caudally (arrowheads in G-J). Occasionally, axons were turning already inside the floor plate (open arrowhead in G). A closer look at the morphology of the axons in the floor plate revealed their tortuous, ‘corkscrew'-like trajectory across the midline at many DiI injection sites (open arrowheads in I). To knockdown Endoglycan either in the floor plate or in commissural neurons, the ventral or dorsal spinal cord was targeted as indicated in (D) (see inserts in G and J, respectively). Phenotypes were the same as those observed after targeting one half of the spinal cord including the floor plate (H,I). Pathfinding was normal in embryos electroporated with dsRNA derived from Podocalyxin (K). The quantification of injection sites with pathfinding errors after targeting the floor plate or one half of the spinal cord is shown in (L). Pathfinding errors were seen only at 6.7±3.4% of the injection sites in untreated control embryos (n=10 embryos, 45 injection sites). In control embryos injected and electroporated with the EGFP plasmid alone, pathfinding errors were found at 16.2±6% of the injection sites (n=17 embryos, 92 injection sites). Injection and electroporation of dsRNA derived from CD34 (24.6±5.8%, n=8 embryos, 80 injection sites) and Podocalyxin (23.3±3.9%, n=17 embryos, 147 injection sites) did not affect midline crossing and turning behavior of commissural axons. By contrast, 82.3±5.6% (n=11 embryos, 65 sites) and 61.7±6.4% (n=18, 161 sites) of the injection sites in embryos injected with dsRNA derived from the 3’-UTR or the ORF of Endoglycan, respectively, showed aberrant pathfinding of commissural axons. One-way ANOVA with Tukey’s multiple comparisons test. P values ****<0.0001, compared to EGFP-injected control groups. The two groups electroporated with dsRNA derived from Endoglycan were not different from each other. Values represent average percentage of DiI injection sites per embryo with aberrant axonal navigation ± standard error of the mean. Source data and statistics are available in the Figure 2—source data 1 spreadsheet. Bar: 50 μm.

To demonstrate specificity of Endoglycan downregulation and to verify that the phenotype was not due to an off-target effect, we used three non-overlapping cDNA fragments to produce dsRNA. All fragments resulted in the same phenotypes. Downregulation of Endoglycan with dsRNA derived from the ORF resulted in 61.7 ± 6.4% injection sites with aberrant axon guidance. The effect on axon guidance was also seen with dsRNA derived from the 3’UTR, with 82.3 ± 5.6% of the injection sites with aberrant axon guidance (Figure 2L). In contrast, aberrant axonal pathfinding was seen only at 6.7 ± 3.4% of the injection sites in untreated control embryos. Values were 16.2 ± 6.0 for EGFP-expressing control embryos, 24.6 ± 5.8% for embryos transfected with dsRNA derived from CD34, and 23.3 ± 3.9% for embryos transfected with dsRNA derived from Podocalyxin. Thus, silencing either CD34 or Podocalyxin did not interfere with correct navigation of axons at the midline. Because both of them were expressed in the developing spinal cord (Figure 2—figure supplement 2), these results further support the specificity of the observed effects of Endoglycan silencing.

Lack of Endoglycan affects the morphology of the floor plate only after dI1 axons have crossed the midline

Because the hallmark of sialomucins is their bulky, negatively charged extracellular domain with extensive glycosylation, a role as regulators of cell-cell adhesion has been postulated (Vitureira et al., 2010; Takeda et al., 2000; Nielsen and McNagny, 2008 and Nielsen and McNagny, 2009). This together with our observation that commissural axons have a 'corkscrew'-like phenotype in the midline area in Endoglycan-deficient embryos prompted us to analyze the morphology of the floor plate. Sections were taken from the lumbar level of the spinal cord at HH25 from control-treated and experimental embryos and stained for HNF3β/FoxA2 to label floor-plate cells, and for Contactin2 (aka Axonin-1) to label commissural axons (Figure 3). In untreated (Figure 3A–C) and control-treated embryos (Figure 3D–F), HNF3β/FoxA2-positive cells were aligned to form the characteristic triangular shape of the floor plate. In particular, the ventral border of the floor plate, where commissural axons traverse the midline was smooth, because all floor-plate cells were precisely aligned (Figure 3A,D). In contrast, floor-plate cells were no longer aligned to form a smooth ventral border in embryos lacking Endoglycan after electroporation of dsEndo into one half of the spinal cord (Figure 3G,J). On the one hand, floor-plate cells were found dislocated into the commissure formed by the Contactin2-positive axons (arrowheads in Figure 3I,L). On the other hand, the floor plate appeared to have gaps in embryos lacking Endoglycan. In addition, the floor-plate width was significantly narrower in embryos lacking Endoglycan in comparison to age-matched controls (Figure 3S,T). These changes in floor-plate morphology were not due to differences in cell differentiation or patterning (Figure 3—figure supplement 1). Furthermore, we can exclude cell death as a contributor to the changes in the floor plate, as we found no Cleaved Caspase-3-positive floor-plate cells in any of the conditions (Figure 3—figure supplement 2).

Figure 3 with 4 supplements see all
After downregulation of Endoglycan in one half of the spinal cord, the floor-plate morphology is compromised only after axonal midline crossing.

In untreated (A-C) and control-treated embryos (D-F), the floor plate is of triangular shape with floor-plate cells precisely aligned at the ventral border. There is no overlap between the floor plate (visualized by HNF3β staining; red) and the commissure (visualized by anti-Axonin1 staining; green). The shape of the floor plate is no longer triangular in embryos lacking Endoglycan (G-L). The floor-plate cells are not aligned ventrally (arrowheads in G, I, J, and L) and the floor plate appears to have gaps. This change in morphology is only seen at HH25, when midline crossing is completed. When the floor-plate morphology was analyzed at HH21, there was no difference between control (M-O) and experimental embryos electroporated with dsRNA derived from Endoglycan (P-R). Note that some more ventral commissural axon populations have crossed the floor plate at this stage. But overall, the number of axons that form the commissure at HH21 is still very small. The width of the floor plate (indicated by asterisks) was measured (S,T). There was no significant difference in spinal cord width at HH25 (400.2 ± 54.5 µm in untreated controls, 438.2 ± 30.3 µm in EGFP-expressing controls, and 394 ± 12.0 µm in dsEndo embryos), but floor plates were significantly narrower in embryos lacking Endoglycan (T; 61.6 ± 2.9 µm; n = 7 embryos; p<0.001) compared to untreated (82.4 ± 1.8 µm; n = 6 embryos) and EGFP-injected control embryos (83.6 ± 2.6 µm; n = 6 embryos). One-way ANOVA with Tukey’s multiple comparisons test. The commissure had a tendency to be wider in experimental compared to control embryos but the effect was not statistically significant (U). The width of the floor plate was not different between groups when measured at HH21 (V) with 70.1 ± 4.2 µm (n = 3) for untreated and 70.6 ± 2.5 µm for EGFP controls (n = 4), compared to 71.8 ± 2.0 µm for experimental embryos (n = 3). No difference was seen in the width of the commissure. Two-tailed T-test. ****p<0.0001. Mean ± SEM are given. Electroporation was targeted to one side of the spinal cord as shown in the insert in Figure 2H. Bar: 50 µm. Source data and statistics are available in Figure 3—source data 1 spreadsheet.

When embryos lacking Endoglycan were analyzed at HH21 (Figure 3P–R), that is at a time point when dI1 axons have reached but not yet crossed the floor plate, the morphology and the width of the floor plate were not different from controls (Figure 3M–O). In contrast to measurements of floor-plate width at HH25 (Figure 3T), the values for experimental and control embryos were not different at HH21 (Figure 3V). Taken together, these results suggested that the absence of Endoglycan did not affect primarily cell-cell adhesion between floor-plate cells. Rather the altered floor-plate morphology appeared to be an indirect effect of changes in axon to floor-plate adhesion.

To provide more evidence for this idea, we first ruled out an effect of the perturbation of Endoglycan levels on the expression of known guidance cues of dI1 axons, such as Contactin2 (Axonin-1) or NrCAM (Figure 3—figure supplement 3). Similarly, we did not find changes in the expression of Shh and Wnt5a, morphogens that are known to direct post-crossing dI1 axons rostrally (Figure 3—figure supplement 4; Bourikas et al., 2005; Lyuksyutova et al., 2003).

An alternative way of demonstrating the requirement for Endoglycan in both floor plate and commissural axons were rescue experiments (Figure 4). We used dsRNA derived from the 3’-UTR and expressed the ORF of Endoglycan either under control of the Math1 enhancer (expression only in dI1 neurons) or the Hoxa1 enhancer for floor-plate-specific expression. Because expression of these plasmids in control embryos (overexpression) resulted in aberrant behavior of axons at the floor plate, we used three different concentrations of plasmid for our rescue experiments and obtained a dose-dependent effect on axon guidance. Expression of high doses of Endoglycan was never able to rescue the axon guidance phenotype. However, axon guidance was not different from control embryos after transfection of dI1 neurons with a low concentration, or after transfection of floor-plate cells with a medium concentration of the Endoglycan ORF (Figure 4B; Table 1). Interestingly, the source of Endoglycan did not matter, but the amount of Endoglycan did. These findings were consistent with the idea that Endoglycan could be regulating adhesion, as both too much, but also too little adhesion would be a problem for axonal navigation. Furthermore, these results suggested that Endoglycan did not act as a receptor.

Too much or too little Endoglycan causes aberrant axon guidance.

Because silencing Endoglycan either in commissural neurons or in the floor plate caused the same type of axon guidance defects, we wanted to test the idea that the presence of an adequate amount, but not the source of Endoglycan was important. We therefore downregulated Endoglycan by transfection of dsRNA derived from the 3’UTR of Endoglycan into one half of the spinal cord. We then tried to rescue the aberrant axon guidance phenotype by co-electroporation of the Endoglycan ORF specifically in dI1 neurons (using the Math1 enhancer, red) or in the floor plate (using the Hoxa1 enhancer; blue, A and B). The rescue constructs were used at a concentration of 150 (L = low), 300 (M = medium), and 750 (H = high) ng/µl, respectively. In both cases, rescue was only possible with one concentration: the medium concentration of the Endoglycan plasmid driven by the Hoxa1 promoter and the low concentration of the plasmid driven by the Math1 promoter. The lowest concentration of the Hoxa1-driven construct and the two higher concentrations of the Math1-driven constructs were not able to rescue the aberrant phenotype. Note that the amounts of Endoglycan cannot be compared between the Math1- and the Hoxa1 enhancers, as they differ in their potency to drive expression. However, we can conclude a response in a dose-dependent manner in both cases. Statistical analysis by one-way ANOVA: *p<0.05, **p<0.01, ***p<0.001. Values are shown ± standard deviation. See Table 1 for values. Source data and statistics are available in Figure 4—source data 1 spreadsheet.

Figure 4—source data 1

Raw data and statistics for rescue experiments.

https://cdn.elifesciences.org/articles/64767/elife-64767-fig4-data1-v2.xlsx
Table 1
The amount, but not the source of Endoglycan matters.
TreatmentNo of embryosNo of inj. sites% inj sites normal PTp-Value% inj sites stalling% inj sites no turn
Untreated1411185.3 ± 2.60.9990.8 ± 0.814.6 ± 2.7
EGFP148580.5 ± 6.715.1 ± 2.418.6 ± 7.4
dsEndo2116130.2 ± 4.4<0.000131.6 ± 550.1 ± 3.9
β-actin::EndoOE169149.1 ± 5.40.017720 ± 5.130 ± 5
Hoxa1::Endo-L2523438.5 ± 5.7<0.000124.1 ± 4.844.5 ± 5
Hoxa1::Endo-M1813660.3 ± 6.70.368610.9 ± 4.232.1 ± 5
Hoxa1::Endo-H77224.6 ± 10.1<0.000140.0 ± 11.344.1 ± 7
Math1::Endo-L1515759.6 ± 6.80.373812.3 ± 3.530.9 ± 5.4
Math1::Endo-M78635 ± 10.70.003131.7 ± 7.242.4 ± 9.3
Math1::Endo-H88626.3 ± 6.5<0.000150.8 ± 6.440.5 ± 6.1
  1. Concomitant expression of Endoglycan could rescue the aberrant axon guidance phenotype induced by the downregulation of Endoglycan throughout the spinal cord. It did not matter whether Endoglycan was expressed under the Hoxa1 enhancer for specific expression in floor-plate cells, or under the Math1 enhancer for specific expression in dI1 neurons. However, the rescue effect was dose-dependent. Too little, or too much Endoglycan was inducing axon guidance defects. For rescue, Endoglycan cDNA under the control of the Hoxa1 enhancer (Hoxa1::Endo) or the Math1 enhancer (Math1::Endo) were injected at 150 ng/µl (low, L), 300 ng/µl (medium, M), or 750 ng/µl (high, H). The same criteria for quantification were applied as for the results shown in Figure 2. There was no fundamental difference between too much or too little stickiness. In both conditions, we found stalling in the floor plate and failure to turn into the longitudinal axis. The only difference was that we did not find any ‘corkscrew’ phenotypes after overexpression of Endoglycan. The number of embryos and the number of DiI injection sites analyzed per group are indicated. The average % of injection sites with normal axon guidance phenotypes (± standard error of the mean) and the p-value for the comparison between the respective group and the control-treated (EGFP-expressing) group are given. The last two columns list the average values for injection sites with the majority of axons stalling in the floor plate and the majority of axons not turning into the longitudinal axis. Because some injection sites can have both phenotypes, the values do not add up to 100%.

Endoglycan is a negative regulator of cell adhesion

The observation that downregulation of Endoglycan seemed to increase the adhesion between commissural axons and floor-plate cells, together with the knowledge about its molecular features, led us to hypothesize that Endoglycan might act as a negative regulator of cell-cell adhesion. As a first test, we counted the number of commissural neurons that adhered to a layer of HEK cells stably expressing Endoglycan compared to control HEK cells (Figure 5). Expression of Endoglycan reduced the number of commissural neurons on Endoglycan-expressing HEK cells compared to control HEK cells. These suggested ‘anti-adhesive’ properties of Endoglycan depended on its post-translational modification (Figure 5B,D). Enzymatic removal of sialic acid by Neuraminidase or of O-linked glycans by O-glycosidase abolished the ‘anti-adhesive’ properties of Endoglycan expressed in HEK cells. Endoglycan had a similar effect on the adhesion of motoneurons to a layer of HEK cells expressing Endoglycan that also depended on the post-translational modification of Endoglycan with O-glycosylation and sialylation (Figure 5—figure supplements 1 and 2).

Figure 5 with 2 supplements see all
Endoglycan expression reduces adhesion of commissural neurons in vitro.

(A-C) Commissural neurons dissected from HH25/26 chicken embryos were cultured on a layer of control HEK cells or HEK cells stably expressing human Endoglycan. Neurons were allowed to attach for 16 hr. Staining for Axonin-1 revealed a pronounced decrease in the number of commissural neurons on HEK cells expressing Endoglycan compared to control HEK cells (A and D). For each replicate, the number of neurons attached to HEK cells expressing Endoglycan was normalized to the number of cells attached to control HEK cells (D). The number of commissural neurons attached to control HEK cells was more than twice the number on Endoglycan-expressing HEK cells (only 0.47 ± 0.34 compared to 1 ± 0.35). Treatment with Neuraminidase (+N) or O-glycosidase (+O) abolished the difference. See Table 2 for values. N(replicates)=4. ns (not significant), ***p<0.001, ****p<0.0001, one-way ANOVA with Tukey’s multiple comparisons test. Error bars represent standard deviations. O, O-glycosidase; N, Neuraminidase; vehic, vehicle; ctrl, control. Scale bar: 50 µm. Source data and statistics are available in Figure 5—source data 1 spreadsheet.

Figure 5—source data 1

Commissural neuron counts on HEK cells expressing Endoglycan.

https://cdn.elifesciences.org/articles/64767/elife-64767-fig5-data1-v2.xlsx
Table 2
Quantification of the number of commissural neurons adhering to a layer of HEK cells and midline crossing of dI1 axons using live imaging.
Figure 5
PartNameValueStdevn(pictures)N(replicates)
DHEKctrl vehic.1.000.35344
DHEKEndo-myc vehic.0.470.34344
DHEKctrl vehic.+N0.880.36284
DHEKEndo-myc vehic.+N1.030.49284
DHEKctrl vehic.+O1.190.70294
DHEKEndo-myc vehic.+O1.170.50294
Figure 10
PartNameValueStdevn(axons)N(embryos)
BEndo Ctrl5.381.251613
BdsEndo5.501.201683
BEndo OE4.351.382343
CEndo Ctrl Ent-Mid2.740.941613
CEndo Ctrl Mid-Ex2.641.011613
CdsEndo Ent-Mid3.060.881683
CdsEndo Mid-Ex2.451.021683
CEndo OE Ent-Mid2.100.712343
CEndo OE Mid-Ex2.260.832343
DEndo Ctrl0.510.131613
DdsEndo0.560.121683
DEndo OE0.480.112343
EEndo Ctrl0.490.131613
EdsEndo0.440.121683
EEndo OE0.520.112343
FEndo Ctrl49.4912.84703
FdsEndo48.6713.47683
FEndo OE43.5014.84833
GEndo Ctrl47.4511.58703
GdsEndo40.8710.44683
GEndo OE41.0911.18833
HEndo Ctrl106.0528.72703
HdsEndo108.0830.59683
HEndo OE87.5228.01833
Figure 11
A1 st half49.4912.84703
A2nd half47.4511.58703
AFP exit106.0528.72703
B1 st half48.6713.47683
B2nd half40.8710.44683
BFP exit108.0830.59683
C1 st half43.5014.84833
C2nd half41.0911.18833
CFP exit87.5228.01833
  1. The table contains the detailed values from results presented in Figure 5, Figure 10, and Figure 11. Ent, entry; Mid, midline; Ex, exit; FP, floor plate; stdev, standard deviation.

As an additional experiment to assess differences in adhesive strength modulated by Endoglycan, we adapted the growth cone blasting assay developed by Lemmon and colleagues (Lemmon et al., 1992). mRFP-expressing commissural neurons were cultured on a layer of HEK cells and growth cones were blasted off with a constant flow of buffer delivered from a glass micropipette (Figure 6A, Video 1). Thus, the time it took to detach the growth cone from the HEK cell layer was taken as an indirect measurement of adhesive strength. On average, it took only about half the time to detach commissural growth cones from HEK cells expressing Endoglycan compared to control HEK cells (Figure 6B). In line with this difference, overexpression of Endoglycan in neurons growing on control HEK cells significantly decreased the time it took to blast off growth cones compared to control neurons (Figure 6C). The difference in adhesive strength was not explained or markedly affected by differences in adhesive area between growth cones and HEK cells, as the growth cone areas did not differ between control and experimental conditions (Figure 7). Taken together, our in vitro experiments demonstrated that Endoglycan negatively regulated the adhesive strength between commissural growth cones and the HEK cells no matter whether Endoglycan was expressed in the growth cone or in the HEK cells.

Endoglycan reduces the adhesive strength of commissural neuron growth cones.

(A) Example of snapshots taken before and after growth cone blasting of mRFP-positive commissural neurons cultured on a HEK cell layer (shown in the brightfield channel). White arrowheads and asterisks show the location of the growth cone and the approximate location of the micropipette tip, respectively. (B) Control mRFP-transfected commissural neurons were plated either on a layer of control HEK cells or HEK cells expressing Endoglycan. The time it took to detach a growth cone from the HEK cells was taken as a measure for the adhesive strength. The adhesion of growth cones to HEK cells expressing Endoglycan was significantly reduced (0.59 ± 0.26 compared to 1 ± 0.61 on control HEK cells). N(replicates)=3, n(growth cones)=40 (HEKCtrl) and 50 (HEKEndo-myc). (C) Similar observations were made when commissural neurons were transfected with Endoglycan instead of the HEK cells. Detachment was faster for growth cones of mRFP-transfected neurons co-transfected with Endoglycan and plated on control HEK cells (0.83 ± 0.33) compared to control neurons transfected only with mRFP (1 ± 0.41). N(replicates)=6, n(growth cones)=52 (CNCtrl) and 51 (CNEndo-myc). *p<0.05, ****p<0.0001, Two tailed Mann-Whitney test. Error bars represent standard deviation. CN, commissural neuron; Ctrl, control. Scale bar: 50 μm. Source data and statistics are available in Figure 6—source data 1 spreadsheet.

Reduced attachment of growth cones in the presence of higher levels of Endoglycan cannot be explained by changes in growth cone area.

To exclude that the observed differences in adhesive strength between growth cones and HEK cells were influenced by growth cone size, we measured growth cone areas by tracing the edge of the mRFP-positive growth cones from images taken before blasting with a x20 objective. Note that the numbers of growth cones in the growth cone blasting experiment (Figure 6) and the area measurements shown here differ because in two replicates in ~50% of the measured growth cones only videos were taken but no still images. Therefore, these growth cones are not included in the size measurement. The area of each growth cone was measured in imageJ using the tracing tool and the value was normalized to the average growth cone area of the control condition (either HEKCtrl or CNCtrl) for each replicate. A.U., arbitrary unit. Values were 1 ± 0.69 for HEKCtrl (A) versus 0.99 ± 0.61 for HEKEndo-myc (B), and 1 ± 0.54 for CNCtrl (C) versus 0.93 ± 0.62 for CNEndo-myc (D; mean ± standard deviation). Statistical analysis: E, p=0.9694 (ns, unpaired T-test), N (replicates) = 3; n (growth cones)=27 (HEKCtrl) and 25 (HEKEndo-myc); F, p=0.3261 (ns, Mann Whitney test), N (replicates) = 6; n (growth cones)=50 (CNCtrl) and 49 (CNEndo-myc). Bar: 10 µm. Source data and statistics are available in Figure 7—source data 1 spreadsheet.

Video 1
Example of growth cone blasting assay.

Snapshots taken from a video showing the detachment of an mRFP-positive commissural neuron growth cone (shown in black, indicated by white arrowhead) on a layer of HEK cells (bright-field view) before and after detachment. The position of the tip of the glass micropipette is indicated by asterisk.

Next, we tested our hypothesis that Endoglycan was a negative regulator of adhesion by manipulating the balance of adhesion between commissural axons and the floor plate in vivo. We had previously used a similar approach to demonstrate a role of RabGDI in Robo trafficking (Philipp et al., 2012). Commissural axons cross the midline because of the positive signals provided by the interaction of floor-plate NrCAM with growth cone Contactin2 (Stoeckli and Landmesser, 1995; Stoeckli et al., 1997; Fitzli et al., 2000). In the absence of NrCAM or Contactin2, commissural axons fail to enter the floor plate and turn into the longitudinal axis prematurely along the ipsilateral floor-plate border. The positive signal derived from the Contactin2/NrCAM interaction depends on sufficient contact between growth cone and floor-plate cells. Thus, we hypothesized that the failure to detect the positive signal due to lower NrCAM levels on the floor-plate cells could be counteracted by a forced increase in growth cone-floor plate contact. We reasoned that the concomitant downregulation of NrCAM and Endoglycan would rescue the NrCAM phenotype, because the decrease in adhesion due to lower NrCAM, resulting in the failure of commissural axons to enter the floor plate, would be counteracted by an increase in adhesion in the absence of Endoglycan. This is indeed what we observed (Figure 8). As found previously (Stoeckli and Landmesser, 1995; Pekarik et al., 2003), axons were frequently turning prematurely along the ipsilateral floor-plate border in the absence of NrCAM (Figure 8A). In accordance with our hypothesis, ipsilateral turns were reduced to control levels when NrCAM and Endoglycan were downregulated concomitantly (Figure 8B,G). The rescue of the NrCAM phenotype was only seen for Endoglycan, as concomitant downregulation of NrCAM with Podocalyxin or CD34 had no effect on ipsilateral turns (Figure 8).

Downregulation of Endoglycan, but not its family members, rescues the axon guidance phenotype induced by downregulation of NrCAM.

The perturbation of axon/floor-plate contact by downregulation of NrCAM resulted in the failure of commissural axons to enter the floor-plate area and caused their premature turns along the ipsilateral floor-plate border (arrowhead in A) at 63.6 ± 8.8% of the injection sites (n = 7 embryos, 90 injection sites; G). These results are in line with previous reports (Stoeckli and Landmesser, 1995; Philipp et al., 2012). When both NrCAM and Endoglycan were downregulated, the number of ipsilateral turns was reduced to control levels (B,G; 7.3 ± 5.8%, n = 10 embryos; 78 injection sites), consistent with the idea that a decrease in adhesion due to a lack of NrCAM can be balanced by an increase in adhesion between floor plate and growth cones due to a lack of Endoglycan. Downregulation of either Podocalyxin (C) or CD34 (E) did not impair axon guidance (see also Figure 2; 12.3 ± 5.2% (n = 9) and 7.25 ± 4.3% (n = 8), respectively). In contrast to Endoglycan, neither concomitant downregulation of Podocalyxin (D) nor CD34 (F) could rescue the NrCAM-induced ipsilateral turns, as aberrant axon behavior was still observed at 56.7 ± 9.2% (n = 8 embryos, 77 injection sites) and 49.6 ± 8.1% (n = 10 embryos, 100 injection sites), respectively. The floor plate is indicated by dashed lines. For statistical analysis, one-way ANOVA followed by Tukey’s multiple comparisons test was used, **p<0.01 or lower, see source data and statistics in Figure 8—source data 1 spreadsheet; (ns) p≥0.05. Bar: 50 µm.

Figure 8—source data 1

Quantification of ipsilateral turns - raw data.

https://cdn.elifesciences.org/articles/64767/elife-64767-fig8-data1-v2.xlsx

Endoglycan levels modulate growth cone movement in the floor plate

To get more insight into the role of Endoglycan in the regulation of contacts between axons and floor-plate cells, we established ex vivo live imaging of commissural axons during midline crossing. Intact spinal cords of HH22 chicken embryos, which were co-injected and unilaterally electroporated with constructs expressing farnesylated td-Tomato (td-Tomato-f) under the control of the dI1 neuron-specific Math1 enhancer together with farnesylated EGFP (EGFP-f) under the control of the β-actin promoter, were cultured and imaged for 24 hr (Figure 9). This method allowed us to follow the behavior and trajectories of the very first wave of single dI1 axons entering, crossing, and exiting the floor plate in a conserved environment (arrowheads, Figure 9A1–A3). The expression of EGFP-f in all transfected cells and brightfield images helped us to define the floor-plate boundaries (white dashed lines) and midline (yellow dashed lines in Figure 9B,C).

Live imaging of cultured intact spinal cords revealed major impacts of different Endoglycan levels on midline crossing.

(A-C) Live imaging allowed tracing and quantitative analysis of dI1 axons’ trajectories in cultured intact chicken spinal cords. (A1-3) The behavior and trajectory of single tdTomato-positive dI1 axons could be tracked over time when they crossed the floor plate and turned rostrally (yellow and white arrowheads). (B-C) EGFP-F expression under the β-actin promoter and brightfield images helped to visualize the floor-plate boundaries (white dashed line) and the midline (yellow dashed line). (D1, E1, F1) Temporally color-coded projections of 24 hr time-lapse movies (Video 2). Kymograph analysis of the regions of interest selected in the floor plate of each condition shown in (D1,E1,F1) was used to calculate growth cone speed during floor-plate crossing (D2, E2, F2) and after turning into the longitudinal axis (D3, E3, F3). Yellow dashed lines outline a representative example of the slope (velocity) of a single axon crossing the floor plate in each condition. TdTomato-positive axons in control-injected spinal cords (D1-3) crossed the floor plate at a steady speed of 26 µm/h (D2) and turned rostrally in a highly organized manner (D3). In contrast, growth cone speed in the first half of the floor plate that was electroporated with dsRNA derived from Endoglycan (dsEndo) was markedly slowed down to only 13 µm/hr. In the second, non-electroporated half of the floor plate, axons electroporated with dsEndo were faster than control axons (44 µm/hr; E2). Axons overexpressing Endoglycan were faster in both halves of the floor plate (50 µm/hr; F2). Downregulation or overexpression of Endoglycan clearly impacted the rostral turning behavior visualized by less organized patterns (D3-F3). Asterisks mark axons stalling and thus causing a ‘smeared’ pattern in the kymographs. Arrowheads indicate caudally turning axons. R, rostral; C, caudal; ipsi, ipsilateral; contra, contralateral; FP, floor plate. Scale bars: 50 µm.

Video 2
24 hr time-lapse recordings of tdTomato-positive dI1 axons (shown in black) in control (Endo control), Endoglycan knockdown (dsEndo), and Endoglycan overexpression (Endo OE) conditions.

Dashed lines represent floor plate boundaries. R, rostral; C, caudal.

Spinal cords of embryos co-injected with the Math1::tdTomato-f plasmid and dsRNA derived from Endoglycan (dsEndo) or a plasmid encoding chicken Endoglycan under the β-actin promoter (Endo OE) were imaged for 24 hr and compared to spinal cords dissected from control-injected embryos (Endo ctrl, Video 2, temporally color-coded projections in Figure 9D1,E1and F1). In contrast to control-injected spinal cords, the post-crossing segment of dI1 axons was disorganized in dsEndo and Endo OE conditions. In both these conditions, caudal turns were seen (Video 2). As our in vivo data suggested a difference in adhesion between floor-plate cells and dI1 axons, we analyzed axonal midline crossing with kymographs in two different regions of interest (ROI; shown in Figure 9D1, E1, F1). This allowed us to follow growth cone movement across the floor plate and along the floor-plate border. Interestingly, our analyses indicated that the transfection of dsRNA derived from Endoglycan in dI1 neurons and floor-plate cells led to a decrease in the growth cones’ speed in the first half of the floor plate (13 µm/hr) and in an increase in the second half (44 µm/hr; Figure 9E2) compared to control-injected spinal cords (Figure 9D2 and Video 3), where speed in the first and second halves did not differ (26 µm/hr). In spinal cords overexpressing Endoglycan, growth cone speed was accelerated in the entire floor plate (50 µm/h; Figure 9F2; Video 3). The analysis of axon growth in a second ROI confirmed the disorganization seen in both mutants in Video 2. Although the axonal trajectories in control-injected embryos (Figure 9D3) were well organized and mostly parallel, axonal behavior in mutants caused ‘smeared’ patterns (asterisks Figure 9E3) due to stalling and pixels moving obviously in caudal direction indicating caudal axonal turns (arrowheads in Figure 9F3). These phenotypes confirmed our analyses of open-book preparations of spinal cords lacking Endoglycan (Figure 2). Axonal stalling (arrowheads) and caudal turns (arrows) could also be observed at the floor-plate exit site of spinal cord overexpressing Endoglycan (Video 4).

Video 3
Representative examples of the floor plate crossing of tdTomato-positive dI1 axons (shown in black) taken from 24 hr time-lapse recordings from control (Endo control), Endoglycan knockdown (dsEndo), and Endoglycan overexpression (Endo OE) conditions.

An arrowhead in each condition points at the migrating growth cone. Dashed lines represent floor-plate boundaries.

Video 4
Example of a 24 hr time-lapse recordings of tdTomato-positive dI1 axons (shown in black) showing guidance defects at the exit site of the floor plate in Endoglycan overexpression condition.

Stalling growth cones are shown by arrowheads and caudally turning growth cones by arrows. Rostral is up. Dashed line represents the floor-plate exit site.

The obvious differences in axonal behavior in experimental compared to control spinal cords was corroborated by quantitative analyses of specific aspects. Firstly, we quantified how much time the growth cones spent migrating from the floor-plate entry site to the exit site (Figure 10A1). Confirming the observations made in our videos and kymographic analysis, growth cones overexpressing Endoglycan crossed the floor plate faster, in only 4.4 ± 1.4 hr (mean ± SD), compared to controls (5.4 ± 1.3 hr) and the dsEndo condition (5.5 ± 1.2; Figure 10B). Furthermore, we compared the average time for crossing each half of the floor plate for each condition (Figure 10A1, C). Growth cones migrated equally fast through both halves in controls (2.7 ± 0.9 hr versus 2.6 ± 1.0 hr, Figure 10C). After overexpression of Endoglycan, there was a slight but significant shortening of the time growth cones spent crossing the first half compared to the second half (2.1 ± 0.7 hr versus 2.3 ± 0.8 hr, Figure 10C). In contrast, the unilateral silencing of Endoglycan induced a highly significant difference in the migration speed of growth cones in the first (electroporated) versus the second half of the floor plate. It took 3.1 ± 0.9 hr to cross the first half but only 2.5 ± 1.0 hr to cross the second half (Figure 10C). An alternative way of demonstrating the differences in migration speed is shown in Figure 10D,E. We calculated the ratios of the time spent in the first (Figure 10D) or the second half of the floor plate (Figure 10E) compared to the total time used for floor-plate crossing for the different conditions (Figure 10B). Indeed, knockdown of Endoglycan induced a significant increase in the ratio spent in the first half of the floor plate (0.56 ± 0.1) compared to both control (0.51 ± 0.1) and overexpression of Endoglycan (0.48 ± 0.1; Figure 10D). In the second half of the floor plate, there was a significant decrease in spinal cords electroporated with dsEndo (0.44 ± 0.1) compared to control (0.49 ± 0.1) and Endoglycan-overexpressing spinal cords (0.52 ± 0.1; Figure 10E).

Too much or too little of Endoglycan impaired the timing and morphology of single dI1 growth cones migrating in the floor plate.

Data at the single axon level extracted from 24 hr time-lapse recordings of tdTomato-positive dI1 axons crossing the floor plate. (A1) The time of floor-plate crossing was measured for the entire floor plate, for the first and for the second half for each condition. (A2) The average growth cone area was measured in the first half, the second half and at the exit site of the floor plate for each condition. (B) Overexpression of Endoglycan significantly decreased the time axons needed to cross the entire floor plate compared to control and dsEndo conditions (Kruskal-Wallis test with Dunn’s multiple-comparisons test). (C) The average time of crossing the first half and the second half of the floor plate was compared. Interestingly, there was a highly significant difference in the spinal cords unilaterally electroporated with dsEndo, as axons spent much longer in the first compared to the second half of the floor plate. There was no difference between the two halves of the floor plate in the control condition, but there was a significant decrease in growth cone speed between the first (electroporated) half of the floor plate and the second half, where only axons were overexpressing Endoglycan (Wilcoxon test). (D) and (E) The ratios of the time axons spent in the first half (D) or the second half (E) of the floor plate divided by the time they needed to cross it entirely were compared between conditions. Unilateral knockdown of Endoglycan resulted in a significant increase of the ratio in the first half and a decrease in the second half compared to both control and overexpression conditions (one-way ANOVA with Sidak’s multiple-comparisons test). (F-H) The average dI1 growth cone area at each position of the floor plate (as depicted in A2) was compared across all conditions. (F) Overexpression of Endoglycan induced a significant reduction in the average growth cone area compared to the control condition (Kruskal-Wallis test with Dunn’s multiple-comparisons test) but not compared to Endoglycan knockdown (p value = 0.08). (G) In the second half of the floor plate, the average growth cone area was reduced in both Endoglycan knockdown and overexpression condition compared to control (one-way ANOVA with Sidak’s multiple-comparisons test). (H) At the floor-plate exit site, overexpression of Endoglycan induced a significant decrease in the average growth cone area compared to both control and knockdown conditions (one-way ANOVA with Sidak’s multiple-comparisons test). Error bars represent standard deviation. p<0.0001 (****), p<0.001 (***), p<0.01 (**), p<0.05 (*), and p≥0.05 (ns) for all tests. N(embryos)=3 for each condition; n(axons, panel C-E)=161 (Endo Ctrl), 168 (dsEndo), 234 (Endo OE); n(axons, panel F-H)=70 (Endo Ctrl), 78 (dsEndo), 83 (Endo OE). See Table 2 for detailed results. Ent, entry; Mid, midline; Ex, exit; ipsi, ipsilateral; contra, contralateral; FP, floor plate; GC, growth cone. Source data and statistics are available in Figure 10—source data 1 spreadsheet.

Secondly, we also analyzed growth cone morphologies by comparing the average area in the first and the second halves, as well as at the exit site of the floor plate (Figure 10A2). The difference in growth speed was reflected in growth cone morphology and size (Figure 11 and Video 2). Growth cones tended to be small and have a simple morphology at fast speed. At choice points, like the floor-plate exit site, growth cone size and complexity increased. The average area of growth cones in the floor plate in control spinal cords was 50 ± 13 µm2 in the first, and 47 ± 12 µm2 in the second half of the floor plate (Figure 11A). The growth cone area significantly increased at the floor-plate exit site, where growth cones need to choose to grow rostrally rather than caudally (Figure 11A). At the exit site, growth cone area was 106 ± 29 µm2 in control embryos. In agreement with their faster speed in the floor plate overexpressing Endoglycan, growth cones were significantly smaller in the first (43.5 ± 15 µm2) and the second half (41 ± 11 µm2) of the floor plate, as well as at the exit site (88 ± 28 µm2) compared to controls (Figure 10F–H and Figure 11C). Reduction of Endoglycan expression in the axons and in the first half of the floor plate resulted in reduced migration speed, but the average size of the growth cones was not significantly different from controls. (Figure 10F). The fact that growth cone were significantly faster in the second, non-transfected half of the floor plate was reflected by a significant reduction in growth cone area (41 ± 10 µm2) compared to control (49 ± 13 µm2) and compared to the first, transfected half (49 ± 13 µm2; Figure 10G, Figure 11B, Table 2).

Growth cone size is enlarged at the floor-plate exit site.

Data at the single axon level extracted from 24 hr time-lapse recordings of dI1 axons crossing the floor plate. The average growth cone area was measured in the first half, the second half and at the exit site of the floor plate for each condition. (A) No difference in the area was detected between growth cones in the first half and the second half of the floor plate in controls. However, growth cones were found to be much enlarged at the exit site compared to when they were in the floor plate (Friedman test with Dunn’s multiple-comparisons test). (B) Unilateral down regulation of Endoglycan induced a significant decrease in the average growth cone area in the second part of the floor plate compared to the first part. However, dI1 growth cones still got much larger when exiting the floor plate compared to when they were located in the floor plate (one-way ANOVA with Sidak’s multiple-comparisons test). (C) After unilateral overexpression of Endoglycan, no difference in growth cone area was detected between the first and the second half of the floor plate. Like in all other conditions, they were found to be much enlarged at the exit site (Friedman test with Dunn’s multiple-comparisons test). Error bars represent standard deviation. p<0.0001 (****) for all tests. N(embryos)=3 for each condition; n(axons)=70 (Endo Ctrl), 78 (ds Endo), 83 (Endo OE). See Table 2 for detailed results. Source data and statistics are available in Figure 11—source data 1 spreadsheet.

Finally, live imaging of growth cones crossing the floor plate provided support for our hypothesis that axon-floor plate contact was causing the displacement of floor-plate cells observed after knockdown of Endoglycan, the ‘corkscrew’ phenotype. The tortuous, 'corkscrew'-like phenotype of axons was seen exclusively in spinal cords after knockdown of Endoglycan (Figure 12 and Videos 5 and 6). We could observe roundish EGFP-f-positive cells (arrows) that obstructed the smooth trajectory of dI1 axons in the commissure (arrowheads in Figure 12A1-5 and Video 5). Although we could not use markers to identify these cells as floor-plate cells, their position indicated that they had to be mislocalized floor-plate cells. Moreover, dI1 axons were found to form loops in the layer where floor-plate cell somata were localized (arrowhead in Figure 12B1-5 and Video 6). In the first half of the floor plate electroporated with dsEndo, clusters of roundish EGFP-f-positive cells in the commissure (arrows) were apparently causing axons to deviate from their trajectory by strongly adhering to them (arrowheads in Figure 12C1-5 and Video 6). We never found such aberrant behavior in control embryos or in embryos overexpressing Endoglycan. Moreover, we only observed these events after many dI1 axons had already crossed the floor plate (after at least 10 hr), supporting the hypothesis that the phenotype was due to excessive growth cone-floor plate adhesion resulting in floor-plate cell displacement. Furthermore, these observations suggest that Endoglycan regulates migratory speed of growth cones by modulating their adhesion to floor-plate cells.

Live imaging of dI1 axons after perturbation of Endoglycan expression explains ‘corkscew-like’ phenotypes by aberrant interactions between axons and floor-plate cells.

(A-C) 'Corkscrew'-like phenotypes of dI1 axons expressing farnesylated tdTomato were observed by live imaging in the first half of the floor plate (electroporated half) only after Endoglycan was silenced (see also Videos 5 and 6). (A) Dislocated EGFP-positive cells (arrows) caused dI1 axons to deviate from a smooth trajectory inducing a 'corkscrew'-like morphology (arrowheads, A1-4). These axons and cells were located in the commissure as shown in a coronal view (yellow arrowhead, A5 and Video 5). (B) Some axons were found to form a loop (arrowheads, B1-4) invading the layer of the floor plate where somata of floor-plate cells are located as shown in the transverse view (yellow arrowheads, B5 and Video 6). (C) Clusters of dislocated roundish EGFP-positive cells (arrows) seemed to retain growth cones in the first half of the floor plate (arrowheads). These axons and cells were located in the commissure as shown in a coronal view (yellow arrowhead, C5 and Video 6). D, dorsal; V, ventral. Scale bars: 10 µm.

Video 5
The trajectory of single dI1 axons in the first half of the floor plate (electroporated half) in the absence of Endoglycan was aberrant and showed a 'corkscrew'-like phenotype.

Example of a tdTomato-positive dI1 axon taken from a 24 hr time-lapse recording of an ‘Endoglycan-knockdown’ spinal cord. Arrowheads show how the growth cone is migrating within the first half (electroporated half) of the floor plate and enters in contact with mislocated EGFP-positive cells (arrows) that induced a 'corkscrew' like morphology of the shaft. The 3D coronal rotation clearly shows that the axon and cells are located within the commissure (arrowhead).

Video 6
The trajectory of single dI1 axons in the first half of the floor plate (electroporated half) in the absence of Endoglycan was aberrant and showed a 'corkscrew'-like phenotype.

Example of tdTomato-farnesylated-positive dI1 axons taken from 24 hr time-lapse recordings of an ‘Endoglycan-knockdown’ spinal cord. Arrowheads show how two growth cones are migrating within the first half (electroporated half) of the floor plate. The first one made a loop within the floor-plate cell layer (arrowheads), as shown by the 3D coronal rotation. The second one was attracted toward mislocalized EGFP-positive cells (arrow), made contact with them and then carried out a U-turn toward them before continuing its migration in the direction of the midline. The 3D coronal view confirmed that the second growth cone during its U-turn and the mislocalized cells were located within the commissure (arrowhead).

Taken together, our live imaging studies support results from in vitro adhesion experiments indicating that the level of Endoglycan expression modulates adhesive strength between dI1 commissural growth cones and floor-plate cells. In contrast, axon-axon interactions did not seem to be different in the presence and absence of Endoglycan, as we did not find any effect on pre-crossing axons after perturbation of Endoglycan levels (Figure 13).

Endoglycan does not affect pre-crossing commissural axons.

We found no differences in timing or trajectories of pre-crossing commissural axons labeled by the co-electroporation of Math1::tdTomato-F, when we compared control embryos electroporated with an EGFP plasmid (A,B) with embryos electroporated with dsEndo (C,D) or embryos overexpressing Endoglycan (E,F).

In summary, our results demonstrate a vital role for Endoglycan in commissural axon guidance at the ventral midline. The observed phenotype is consistent with the hypothesis that Endoglycan is an essential regulator of cell-cell contacts by modulating the strength of adhesion between commissural axon growth cones and floor-plate cells. This model is supported by observations in vitro and in vivo. Neuronal attachment was negatively affected by the presence of an excess of Endoglycan in a glycosylation-dependent manner, indicating that Endoglycan decreases adhesive strength during neural circuit assembly.

Discussion

We identified Endoglycan, a member of the CD34 family of sialomucins, in a screen for axon guidance cues involved in commissural axon pathfinding at the midline of the spinal cord. The phenotypes obtained in our in vivo experiments and in live imaging observations in intact spinal cords are consistent with an anti-adhesive role of Endoglycan. This is further supported by the reduced adhesion of growth cones in in vitro assays. Thus, Endoglycan may act like a ‘lubricant’ modulating the motility of growth cones in the floor plate by lowering stickiness. Such a function is supported by the structural features of sialomucins. The function of CD34 family members has not been characterized in detail, but all the results obtained so far are compatible with an anti-adhesive role (Nielsen and McNagny, 2008). One exception are reports from lymph node cells, the so-called high endothelial venules (HEVs), where a very specific glycosylation patterns was implicated in the interaction of CD34 and Endoglycan with L-selectin (Furness and McNagny, 2006). However, in agreement with most published studies on the role of CD34 and Podocalyxin (for reviews see Furness and McNagny, 2006; Nielsen and McNagny, 2008 and Nielsen and McNagny, 2009), our observations suggest that Endoglycan acts as an anti-adhesive rather than as adhesive factor. This model is supported by results from in vivo and in vitro experiments that confirm a negative effect of Endoglycan on cell-cell adhesion (Figure 14).

Endoglycan modulates cell-cell contact by interference with adhesive strength.

Based on our in vivo and in vitro studies, we postulate a model for Endoglycan function in neural circuit formation that suggests an ‘anti-adhesive’ role by modulation of many specific molecular interactions due to decreasing cell-cell contact. This model is consistent with our rescue experiments demonstrating that the source of Endoglycan did not matter, but the expression level did, as aberrant phenotypes were prevented, when Endoglycan was expressed either in the axon or in the floor plate.

The adhesion-modulating effect of Endoglycan is mediated by the negatively charged mucin domain. Similar to the role suggested for the polysialic acid modification of NCAM (Rutishauser, 2008; Brusés and Rutishauser, 2001; Burgess et al., 2008), Endoglycan could lower cell-cell adhesion by increasing the distance between adjacent cell membranes due to repulsion caused by the bulky, negatively charged posttranslational modifications of its extracellular domains. A similar effect was found for PSA-NCAM in hindlimb innervation (Tang et al., 1994; Landmesser et al., 1990) and in the visual system, where retinal ganglion cell axons innervating the tectum were found to regulate axon-axon adhesion versus axon-target cell adhesion (Rutishauser et al., 1988). The same mechanism was found in motoneurons, where axon-axon versus axon-muscle fiber adhesion was a determining factor for the appropriate innervation pattern. In contrast to PSA-NCAM that continues to play a role in synaptic plasticity in the adult nervous system, the function of Endoglycan appears to be restricted to development. Our findings together with structural features of Endoglycan suggest that the mechanisms by which Endoglycan reduces adhesion of growth cones and floor-plate cells might be similar. Additional studies will be required to verify whether the effect of Endoglycan as modulator of adhesive strength uses the exact same mechanism as PSA-NCAM.

At first sight, the effect of Endoglycan on floor-plate morphology appears to suggest a positive regulation of cell-cell adhesion. Floor-plate cells are precisely aligned in control embryos but are protruding into the commissure in the absence of Endoglycan. Therefore, one might conclude that in the absence of Endoglycan cell-cell adhesion between floor-plate cells is compromised, resulting in the observed structural changes. However, this scenario can be ruled out based on the analysis of younger embryos. At HH21, the floor plate was intact in the absence of Endoglycan, indicating that Endoglycan is not required for adhesion between floor-plate cells (Figure 3). The morphology of the floor plate is only compromised once many axons have crossed the midline (Figure 3G–L). These findings are supported by our live imaging data of growth cones crossing the floor plate (Figure 12, Videos 5 and 6). Contacts between commissural axons and floor-plate cells have to be broken when later crossing commissural axons arrive and cross (Yaginuma et al., 1991). Commissural axons crossing the floor plate are suggested to do so by close interaction with short filopodial processes of floor-plate cells (Dumoulin et al., 2021). Thus, the aberrant morphology of the floor plate at HH25 could be explained by the inability of axons to break contacts with floor-plate cells in the absence of Endoglycan, consistent with our hypothesis that Endoglycan is a negative regulator of adhesion. Further support for this hypothesis was contributed by in vitro findings that the adhesive strength between growth cones of commissural neurons and HEK cells was reduced when Endoglycan was expressed either in the HEK cells or in the neurons (Figure 6). Live imaging data demonstrated that the perturbation of the balance in growth cone-floor plate adhesion led to impaired timing of midline crossing, which in turn might also interfere with the correct sensing and reading of guidance cues by dI1 growth cones, and prevented them from making the correct decision at the floor-plate exit site (Figures 9 and 10).

Thus, we concluded that the function of Endoglycan in commissural axon guidance is to lower cell adhesion: the absence of Endoglycan results in too much stickiness. At the midline of the spinal cord, excessive adhesion causes axons to adhere too much to floor-plate cells and prevents their displacement by follower axons. Rather than acting as a guidance cue or guidance receptor, we suggest that Endoglycan affects neural circuit formation by modulating the interaction of many different guidance cues and their surface receptors.

In summary, we propose an ‘anti-adhesive’ role for Endoglycan in axon guidance that is fine-tuning the balance between adhesion and de-adhesion (Figure 14). Precise regulation of cell-cell contacts is required in both processes and is fundamental for developmental processes that depend on a high degree of plasticity and a plethora of specific molecular interactions.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
AntibodyAnti-digoxigenin-AP antibody (rabbit polyclonal)RocheRRID:AB_514497ISH (1:10000)
AntibodyAnti-myc (rabbit polyclonal)AbcamRRID: AB_307014IF (1:500)
AntibodyAnti-Hnf3β (mouse monoclonal)DSHBRRID: AB_2278498IF (supernatant)
AntibodyAnti-Axonin-1 (goat polyclonal)Stoeckli and Landmesser, 1995N/AIF (1:500)
AntibodyAnti-Axonin-1 (rabbit polyclonal)Stoeckli and Landmesser, 1995N/AIF (1:1000)
AntibodyAnti-neurofilament-M (RMO-270, mouse monoclonal)ThermoFisher ScientificRRID: AB_2532998IF (1:200)
AntibodyAnti-GFP-FITC (goat polyclonal)RocklandRRID: AB_218187IF (1:400)
AntibodyAnti-NrCAM(mouse monoclonal)Fitzli et al., 2000N/AIF (1:1000)
AntibodyAnti-Islet1 (mouse monoclonal)DSHBRRID: AB_528315IF (supernatant)
AntibodyAnti-Nkx2.2 (mouse monoclonal)DSHBRRID: AB_531794IF (supernatant)
AntibodyAnti-Pax3 (mouse monoclonal)DSHBRRID: AB_528426IF (supernatant)
AntibodyAnti-Pax6 (mouse monoclonal)DSHBRRID: AB_528427IF (supernatant)
AntibodyCleaved Caspase-3 (rabbit polyclonal)Cell SignalingRRID: AB_2341188IF (1:200)
Cell line (H. sapiens)HEK293TAmerican Type Culture CollectionRRID: CVCL_0063
Cell line (H. sapiens)HEK293T-PODXL2-myc(Endo-myc)This paperN/A
Recombinant DNA reagentMath1::chEndoglycan (plasmid)This paperN/A
Recombinant DNA reagentHoxa1::chEndoglycan (plasmid)This paperN/A
Recombinant DNA reagentβ-actin::chEndoglycan (plasmid)This paperN/A
Recombinant DNA reagentβ-actin::EGFP-F (plasmid)This paperN/A
Recombinant DNA reagentβ-actin::mRFP (plasmid)Wilson and Stoeckli, 2011N/A
Recombinant DNA reagentMath1::tdTomato-F (plasmid)Wilson and Stoeckli, 2011N/A
Chemical compound, drugO-glycosidase and NeuraminidaseNEBCat# E0540S
Software, algorithmFiji/ImageJSchindelin et al., 2012RRID:SCR_002285
Biological sample
(Gallus gallus)
Hubbard JA57 strainBrüterei Stöckli, OhmstalN/A

Identification and cloning of Endoglycan

Request a detailed protocol

We had used a PCR-based subtractive hybridization screen to search for guidance cues for post-crossing commissural axons (for details, see Bourikas et al., 2005). To this end, we isolated floor-plate cells from HH26 and HH20 embryos (Hamburger and Hamilton, 1951). Among the differentially expressed genes, we found a sequence from the 3’-UTR of Endoglycan. Subsequently, a cDNA fragment from the coding sequence of Endoglycan (PODXL2; 1028–1546 bp) was obtained by RT-PCR using total RNA isolated from HH40 cerebellum. For reverse transcription, 1 µg total RNA was mixed with 0.3 µl RNasin (Promega), 1 µl dNTPs (5 mM), 1 µl random nonamers, 1 µl DTT (Promega), in 20 µl Superscript II buffer (Invitrogen). Reverse transcription was carried out for 1 hr at 42°C. Two µl of this mixture were used for PCR with 2.5 µl forward primer (10 µM; 5’-CAGACACGCAGACTCTTTC-3’) and 2.5 µl reverse primer (10 µM; 5’-CTAAAGATGTGTGTCTTCCTCA-3’) using the Expand Long Template PCR System (Roche). The PCR conditions were 35 cycles at 95°C for 30 s, 57°C for 30 s and 68°C for 3 min. The PCR product was cut with BamHI/BclI and cloned into pBluescript II KS. For cloning of full-length chicken Endoglycan, we used 5’-ATGGTGAGAGGAGCTGCG-3’ and 5’-GTGTTTGAGGAAGACACACATCTTTAG-3’ as forward and reverse primers, respectively. A plasmid containing the full-length ORF of human Endoglycan was obtained from SourceBioScience.

Preparation of DIG-labeled RNA probes and in situ hybridization

Request a detailed protocol

For in vitro transcription, 1 µg of the linearized and purified plasmids encoding Endoglycan (EndoORF: 1028-1546pb, Endo3’UTR: 3150–3743 bp and 5070–5754 bp; numbers are derived from the human sequence), Podocalyxin (ChEST190L9), and CD34 (ChEST91D7) were used to prepare DIG-labeled in situ probes as described earlier (Mauti et al., 2006). The same fragments were used to prepare dsRNA (Pekarik et al., 2003; Baeriswyl et al., 2008; Andermatt and Stoeckli, 2014b).

Northern blot

Request a detailed protocol

Total RNA was extracted from cerebrum, cerebellum, spinal cord, muscle, heart, lung, and kidney from HH38 embryos using the RNeasy Mini Kit (Qiagen) and loaded on a denaturing formaldehyde gel (4.5 µg of total RNA per lane). The RNA was blotted onto a positively charged nylon membrane (Roche) overnight, using 10x SSC as a transfer medium. The membranes were hybridized with 1.5 µg preheated DIG-labeled RNA probes for Endoglycan and GAPDH at 68°C overnight. The membrane was then washed twice with 2xSSC/0.1%SDS for 5 min at room temperature and twice with 0.1xSSC/0.5% SDS for 20 min at 68°C. For detection, buffer 2 (2% blocking reagent dissolved in 0.1 M maleic acid, 0.15 M NaCl, pH 7.5) was added for 2–3 hr at room temperature. After incubation with anti-digoxigenin-AP antibody dissolved in buffer 2 (1:10,000; Roche, RRID:AB_514497) for 30 min at room temperature the membrane was washed twice in washing buffer (0.3% Tween 20 dissolved in 0.1 M maleic acid, 0.15 M NaCl, pH 7.5) for 20 min. Subsequently, detection buffer (0.1 M Tris-HCl, 0.1 M NaCl, pH 9.5) was applied for 2 min before adding CDP-star (25 mM, cat# C0712, Roche) for 5 min in the dark. For detection of the chemiluminescence a Kodak BioMAX XAR film was used.

In ovo RNAi

Request a detailed protocol

For functional studies in the spinal cord, we silenced Endoglycan with three different long dsRNAs. They were produced from bp 1028–1546 of the ORF, as well as bp 3150–3743 and bp 5070–5754 from the 3’UTR. The fact that we obtained the same phenotype with three different, non-overlapping dsRNAs derived from Endoglycan confirms the specificity of the approach and the absence of off-target effects. dsRNA was produced as detailed in Pekarik et al., 2003 and Wilson and Stoeckli, 2011. Because no antibodies recognizing chicken Endoglycan are available, we used in situ hybridization to assess the successful downregulation of the target mRNA (Figure 2—figure supplement 3). Downregulation efficiency was about 40%. Because we transfect only around 50% of the cells in the electroporated area, transfected cells express only very low levels of Endoglycan.

For rescue experiments, the dsRNA was co-injected with 150 (low), 300 (middle), or 750 ng/µl (high) plasmid encoding the ORF of chicken Endoglycan. The ORF was either expressed under the control of the Math1 promoter for dI1 neuron-specific expression, or the Hoxa1 promoter for floor-plate specific expression of Endoglycan.

Commissural neuron and motoneuron adhesion assay

Request a detailed protocol

Dissociated commissural or motoneurons of HH25/26 chicken embryos were cultured as described previously (Avilés and Stoeckli, 2016; Mauti et al., 2006) either on HEK293T cells stably expressing human Endoglycan-myc under the control of the CMV promoter or on untransfected HEK293T (ATTC, Cat# CRL-3216, RRID:CVCL_063) cells as control. The plasmid encoding human Endoglycan was obtained from SourceBioScience (Nottingham, UK). Cultures of commissural neurons were fixed with 4% paraformaldehyde for 15 min at 37°C and stained with goat anti-Axonin1(Contactin2) and rabbit anti-myc antibodies (Abcam, RRID:AB_307014). The number of Axonin-1-positive neurons was counted from seven to nine images per replicate taken with a ×40 water objective in random regions of a well where the confluence of HEK cells was at least 60%. The number of neurons per image was normalized to the average number on control HEK cells for each replicate. Cultures of motoneurons were fixed for 1 hr at room temperature in 4% paraformaldehyde and stained with mouse anti-neurofilament (RMO 270; Thermofisher Scientific, RRID:AB_2532998) and rabbit anti-myc antibodies (Abcam, RRID:AB_307014). The number of neurofilament-positive cells was counted in 16 randomly selected frames (0.4 mm2). Similar results were obtained in three independent experiments. One representative example is shown in Figure 5—figure supplement 1. Both commissural and motoneurons were tested for adhesive strength after HEK cells expressing Endoglycan were treated with O-glycosidase (8’000 U/ml) or α2–3,6,8 Neuraminidase (5 U/ml; NEB Cat# E0540S, kit with both enzymes) for 2 hr before commissural neurons or motoneurons were added (Figure 5; Figure 5—figure supplement 2).

Growth cone blasting assay

Request a detailed protocol

Commissural neurons were dissected from the most dorsal region of spinal cords of HH25/26 embryos that were unilaterally electroporated in ovo at HH17-18 with a plasmid encoding mRFP under the β-actin promotor (30 ng/μl) or co-electroporated with a plasmid encoding the open-reading frame of Endoglycan under the β-actin promotor (300 ng/μl). Dissociated neurons were plated (400 neurons per mm2) on a layer of HEK cells (control or expressing Endoglycan, ~60% confluent) plated the day before on poly-L-lysine-coated (20 μg/ml) 35-mm dishes (Sarstedt). Neurons on HEK cells were cultured overnight at 37°C with 5% CO2 in medium consisting of MEM/Glutamax (Gibco), 4 mg/ml Albumax (Gibco), N3 (100 μg/ml transferrin, 10 μg/ml insulin, 20 ng/ml triiodothyronine, 40 nM progesterone, 200 ng/ml corticosterone, 200 μM putrescine, 60 nM sodium selenite; all from Sigma) and 1 mM pyruvate. The growth cone blasting assay was adapted from Lemmon et al., 1992. Before blasting growth cones, a final concentration of 20 mM HEPES was added to the cells to maintain the pH of the medium. Cells were maintained at 37°C in a temperature-controlled chamber (Life Imaging Services) and were visualized with an inverted IX81 microscope (Olympus) equipped with a DP80 camera (Olympus) and a ×20 air objective (2CPLFL PM20x/0.40, Olympus). Growth cones were blasted off the HEK cells with a glass micropipette with an opening of 8–13 μm positioned 75 μm away from the leading edge of the growth cone and 50 μm above the growth cone. The height of the micropipette tip above the cells was set by adding 50 μm in the z-axis to the focus plane on the HEK cell surface where the growth cone was localized using the Olympus CellSens Dimension 2.2 software. Then, the micropipette tip was brought into the focus of this z-localization at a distance of 75 μm away from the leading edge of the growth cone that was previously set with the tracing tool of the same software. The glass micropipette was connected to a pump controlled by a liquid chromatography controller (LCC-500, Pharmacia) and a constant flux of 20 μl/min of sterile PBS containing phenol red (Gibco) was directed at the growth cone until it detached. For reproducible results, it was important to use degassed PBS, as the formation of tiny bubbles in the tubing otherwise caused variability in the force exerted on the growth cones. The time growth cones took to detach from the HEK cells was measured manually by a person blind to the experimental condition and for each growth cone a video was taken. A second person (not blind to the experimental condition) was switching between controls and experimental cells to make sure that small instabilities in the pump, the flow of PBS or the temperature of the set-up would not introduce artefacts. We only counted the time a growth cone took to detach, if the HEK cell layer was intact after blasting (see Video 1). Note that only results generated with the same glass micropipette were normalized to the control and compared to each other. The size of growth cones within each group was heterogeneous, but there was no difference in average growth cone size between the different conditions (Figure 7).

Immunohistochemistry

Request a detailed protocol

Cryostat sections were rinsed in PBS at 37°C for 3 min followed by 3 min in cold water. Subsequently, the sections were incubated in 20 mM lysine in 0.1 M sodium phosphate (pH 7.4) for 30 min at room temperature before being rinsed in PBS three times for 10 min. The tissue was permeabilized with 0.1% Triton in PBS for 30 min at room temperature and then washed again three times with PBS for 10 min. To prevent unspecific binding of the antibody, the tissue was blocked with 10% fetal calf serum (FCS) in PBS for one hour. Goat anti-GFP (1:400; Rockland, RRID:AB_218187), anti-axonin-1 (rabbit 1:1000 or goat 1:500), anti-NrCAM (goat 1:1000), mouse anti-HNF3β (supernatant; 4C7, DSHB), mouse anti-Islet1 (supernatant; 40.2D6, DSHB, RRID:AB_528315), mouse anti-Nkx2.2 (supernatant; 74.5A5, DSHB, RRID:AB_531794), mouse anti-Pax3 and Pax6 (supernatants, DSHB, RRID:AB_528426 and AB_528427, respectively) were dissolved in 10% FCS/PBS and incubated overnight at 4°C. After three washes in PBS, 10% FCS in PBS was applied again for one hour, followed by the incubation with goat anti-rabbit IgG-Alexa488 (1:250; Molecular Probes, RRID:AB_2576217), donkey anti-rabbit IgG-Cy3 (1:200; Jackson ImmunoResearch, RRID:AB_2307443) or goat anti-mouse IgG-Cy3 (1:250; Jackson ImmunoResearch, RRID:AB_2338680) diluted in 10% FCS in PBS for 90 min at room temperature. The tissue was rinsed five times in PBS for 12 min and then mounted in Celvol (Celanese) or Mowiol. The staining of cryostat sections was analyzed with an upright microscope equipped with fluorescence optics (Olympus BX51). Apoptosis was analyzed as described previously (Baeriswyl and Stoeckli, 2008). For analysis of cell death in the floor plate, we used cleaved caspase-3 staining of sections taken from HH25 embryos using a cleaved caspase-3 polyclonal antibody (1:200, Cat# 9661S, Cell Signaling, RRID:AB_2341188).

Quantification of commissural axon guidance errors

Request a detailed protocol

To analyze commissural axon growth and guidance, the embryos were sacrificed between HH25 and 26. The spinal cord was removed, opened at the roof plate (‘open-book’ preparation) and fixed in 4% paraformaldehyde (PFA) for 40 min to 1 hr at room temperature. To visualize the trajectories of commissural axons, Fast-Dil (5 mg/ml, dissolved in ethanol, Molecular Probes) was injected into the dorsal part of the spinal cord as described previously (Wilson and Stoeckli, 2012). Dil injections sites with pathfinding errors were analyzed by a person blind to the experimental condition, using an upright microscope equipped with fluorescence optics (Olympus BX51). All measurements including floor-plate width, thickness of the commissure, and spinal cord width were performed with the analySIS Five software from Soft Imaging System. For all measurements, embryos injected with dsRNA derived from Endoglycan were compared with embryos injected with the EGFP plasmid only, and untreated controls. For statistical analyses, ANOVA with Tukey's or Sidak's multiple comparisons test was used. Details are given in the Figure legends or in the source data.

Live imaging

Request a detailed protocol

Plasmids encoding farnesylated td-Tomato under the Math1 enhancer, upstream of the β-globin minimal promoter, for dI1 neuron-specific expression (Math1::tdTomato-f), and farnesylated EGFP under the β-actin promoter (β-actin::EGFP-f) were co-injected into the central canal of the chicken neural tube in ovo at HH17/18 and unilaterally electroporated, using a BTX ECM830 square-wave electroporator (five pulses at 25 V with 50 ms duration each), as previously described (Wilson and Stoeckli, 2012). For the perturbation of Endoglycan levels, either 300 ng/µl dsRNA derived from the 3’-UTR of Endoglycan or a plasmid encoding the open-reading frame of Endoglycan under the β-actin promoter were co-injected with the Math1::tdTomato-f plasmid. After electroporation, embryos were covered with sterile PBS and eggs were sealed with tape and incubated at 39°C for 26–30 hr until embryos reached stage HH22.

For live imaging, embryos were sacrificed at HH22. Dissection, mounting and imaging of intact spinal cord were carried out as described in detail in Dumoulin et al., 2021. Intact spinal cords were dissected and embedded with the ventral side down in a drop (100 µl) of 0.5% low-melting agarose (FMC; Pignata et al., 2019) containing a 6:7 ratio of spinal cord medium (MEM with Glutamax (Gibco) supplemented with 4 mg/ml Albumax (Gibco), 1 mM pyruvate (Sigma), 100 Units/ml Penicillin, and 100 µg/ml Streptomycin in a 35-mm Ibidi µ-Dish with glass bottom (Ibidi, #81158)). Once the agarose polymerized, 200 µl of spinal cord medium were added to the drop and live imaging was started.

Live imaging recordings were carried out with an Olympus IX83 inverted microscope equipped with a spinning disk unit (CSU-X1 10’000 rpm, Yokogawa). Cultured spinal cords were kept at 37°C with 5% CO2 and 95% air in a PeCon cell vivo chamber. Temperature and CO2-levels were controlled by the cell vivo temperature controller and the CO2 controller units (PeCon). Spinal cords were incubated for at least 30 min before imaging was started. We acquired 18–35 planes (1.5 µm spacing) of 2 × 2 binned z-stack images every 15 min for 24 hr with a 20x air objective (UPLSAPO 20x/0.75, Olympus) and an Orca-Flash 4.0 camera (Hamamatsu) with the help of Olympus CellSens Dimension 2.2 software. Z-stacks and maximum projections of Z-stack videos were evaluated and processed using Fiji/ImageJ (Schindelin et al., 2012). Temporally color-coded projections were generated using Fiji/ImageJ. Kymograph analysis of axons crossing the floor plate or exiting it was performed as previously described (Medioni et al., 2015), using a region of interest (ROI) selection, the re-slice function, and the z-projection of the re-sliced results in Fiji/ImageJ, which allowed following pixel movements within the horizontal axis. The ROI in the floor plate was selected as a 120 × 20 µm2 rectangle and the one in the post-crossing segment was a rectangle of 175 × 104 µm2. Note that the post-crossing segment ROI was rotated by 90° before running the kymograph analysis. The MtrackJ plugin (Meijering et al., 2012) was used to virtually trace single tdTomato-positive dI1 axons crossing the floor plate. Only axons that enter, cross and exit the floor plate during the 24 hr imaging period were traced and quantified. Overlays of traced axons with GFP and brightfield channels were used to assess the time the axons needed to cross the floor plate. Videos of axons with the ‘corkscrew’ phenotype were generated and assembled from z-stacks that were 2D deconvolved (nearest neighbor) using the Olympus CellSens Dimension 2.2 and Fiji/Image J software, respectively.

Data availability

All data generated and analyzed during this study are included in the manuscript and supporting files.

References

Decision letter

  1. Carol A Mason
    Reviewing Editor; Columbia University, United States
  2. Didier YR Stainier
    Senior Editor; Max Planck Institute for Heart and Lung Research, Germany
  3. Kristian Franze
    Reviewer; University of Cambridge, United Kingdom

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Your study reports the role of Endoglycan in the navigation of commissural axons in the spinal cord. Using a range of in vitro and in vivo approaches and time-lapse imaging, you have highlighted the interactions between the axons and their local environment during the floor plate crossing. In particular, in a novel in vitro experiment, neuronal attachment was tested when exposed to forces exerted by a flow of a buffer solution in the presence of Endoglycan. Such experiments strengthen the view that Endoglycan regulates adhesion between commissural axons and their first intermediate target, the floor plate, thereby facilitating axon growth through the ventral spinal cord midline. Your study highlights the importance of balancing adhesive and anti-adhesive forces between growth cones and guidepost cells for proper axon guidance.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Endoglycan plays a role in axon guidance and neuronal migration by negatively regulating cell-cell adhesion" for consideration by eLife. Your article has been reviewed by a Senior Editor, a Reviewing Editor, and three reviewers. The following individual involved in review of your submission has agreed to reveal their identity: Valerie Castellani (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife at this time.

The reviewers found your study proposing the sialomucin Endoglycan as a new player in axon guidance in commissural axon growth and Purkinje cell migration of interest, especially in its role in regulating cell adhesion in these processes. Indeed, the mechanisms that regulate and adjust cell-cell adhesive contacts and how these contacts are coupled to guidance decisions are poorly understood, and your study, which was carefully executed, addresses an important and underestimated aspect of axon guidance. However, the reviewers share the opinion that the experimental evidence supporting the mode of action of Endoglycan as an anti-adhesive factor was lacking, in both the in vivo and in vitro settings.

in vivo, you demonstrate through RNA interference that Endoglycan is necessary for proper navigation and migration, but the basic mechanisms underlying the contribution of Endoglycan and how it functions at specific steps of the navigation rather than all along, were not considered to be addressed in depth. While your knockdown experiments cause misrouting and increased tortuosity, and your rescue experiments implicate a role for Endoglycan in guidance, it is unclear how increased adhesion when Endoglycan is disrupted is responsible for these outcomes. The "loosening" of the floor plate cells rather than their enhanced clumping is a puzzle; might the loosening of the floor plate cells cause the misrouting? And a further puzzling aspect is that axon fasciculation near the floor plate when Endoglycan is perturbed appears unchanged.

A more difficult criticism to address, is direct evidence for involvement of Endoglycan in regulating cell-cell adhesion: to demonstrate changes in fasciculation rather than neurite length (which could vary due to many factors), to present both high and low levels of Endoglycan to axons in the non-neuronal cells, and to directly measure cell-cell adhesion strength in the presence of Endoglycan. These experiments are challenging yet would be important and very welcome to the field, should you be able to execute them.

A final note is that the Purkinje cell experiments were little criticized but detracted a bit from the main story.

We hope that these comments, in full below, can aid you in revising your manuscript.Reviewer #1:

In this manuscript, Baeriswyl et al., convincingly show that the sialomucin Endoglycan plays an important role in commissural axon growth and Purkinje cell migration, and suggest that the main function of Endoglycan is to unspecifically regulate cell-cell adhesion.

The manuscript thus identifies a new player in axon guidance (confirming data from a previous screen by the group). However, it is currently not exactly clear how Endoglycan affects axon growth. While the authors make some strong claims about adhesion, there is currently no direct evidence for an involvement of Endoglycan in regulating cell-cell adhesion.

If Endoglycan indeed opposes cell adhesion, why don't the floor plate cells stick together better in the knockdowns than in controls (rather than the other way around as it is shown in Figure 2)? The authors address this issue in the Discussion, however, their arguments are not very convincing. The fact that at HH21 the floor plate was intact in the absence of Endoglycan doesn't necessarily mean that cell adhesion is not affected or that Endoglycan is not required for adhesion between floorplate cells. An alternative explanation might be that there are only weak mechanical forces acting on these cells at early stages, which might increase during development and eventually be strong enough to tear apart cells that have reduced adhesion.

Furthermore, if cell-cell adhesion is changed, I would also expect a change in fasciculation of axon bundles on their way to the floor plate. Figure 1 seems to suggest that this might not be the case. To address this issue, fasciculation should be quantified.

I also don't find the in vitro experiments very convincing. Figure 5 shows that neurite length is decreased in Endoglycan overexpressing motoneurons but not when it is overexpressed in COS cells. First, if the main effect of Endoglycan is the unspecific regulation of cell-cell adhesion, I would have expected a stronger effect of the overexpression in COS cells even if the transfection efficiency was only ~50%. Second, it would be good to see a similar set of experiments with decreased expression of Endoglycan. What would the authors expect? And lastly, changes in neurite length can be caused in many different ways, certainly not allowing to conclude “that Endoglycan acts as a negative regulator of cell-cell adhesion”.

Similarly, Figure 5—figure supplement 1 only shows a few images, no quantification of data is shown. For how long were neurons allowed to grow? And even if there are reproducibly less neurons growing on transfected HEK cells, this could be a consequence of many factors (one of them of cause being a change in adhesion).

In summary, the manuscript presents new and exciting data on axon guidance, which might be very relevant for many other systems in which cell-cell adhesion is important. However, while some of the data suggest that Endoglycan might mainly act through the regulation of cell adhesion, this remains an hypothesis until it is directly shown. In order to convincingly demonstrate how Endoglycan regulates axon growth, the authors should directly measure cell-cell adhesion strength and quantify its dependence on the presence of Endoglycan.

Reviewer #2:

In this manuscript by Baeriswyl et al., the authors describe defects in spinal commissural axon guidance and cerebellar Purkinje cell migration that occur after interfering with the function of the sialomucin Endoglycan. Baeriswyl et al., suggest that these phenotypes reflect a function of Endoglycan in reducing cell-cell adhesion.

The authors show that Endoglycan knockdown in the chick spinal cord floor plate or floor plate plus commissural neurons disrupts clustering of floor plate cells and causes increased axon tortuosity during floor plate crossing and abnormal caudal turning of commissural axons after midline crossing. Expression of Endoglycan using commissural neuron- or floor plate-specific promoters can rescue these axon guidance defects, depending on expression levels. The authors further show that knockdown of Endoglycan in the floor plate can suppress the premature commissural axon turning phenotype that results from NrCAM knockdown, and in vitro experiments support the idea that Endoglycan can reduce adhesion of motor neurons to heterologous cells. Lastly, knockdown of Endoglycan in the cerebellum disrupts Purkinje cell positioning and overall cerebellar morphology.

This work presents a series of novel observations, and the idea that Endoglycan modulates cell-cell adhesion to control axon guidance and neuronal migration is of potential interest to the field, even though such an anti-adhesive mechanism for axon guidance has been extensively explored in the context of other molecules, e.g. NCAM. The main conclusions about the mechanism of Endoglycan function in axon guidance and cell migration, however, are not sufficiently supported by the data, and some important control experiments are missing. These problems, together with the limited conceptual novelty of the findings, do not make this manuscript a strong candidate for publication in eLife.

1) The experiments involving Endoglycan knockdown in the spinal cord are difficult to interpret for multiple reasons. It is not clear that increased axon-floor plate adhesion is responsible for the observed effects.

a) Knockdown is targeted to the floor plate or "one half of the spinal cord including the floor plate". Data that validate successful knockdown in the targeted cell populations should be included. Also, why is knockdown in commissural neurons alone not attempted?

b) The data showing normal floor plate differentiation (Figure S3) are not sufficient to exclude abnormal tissue patterning as a cause for the axon guidance defects. Shh expression appears reduced (panel C), which could explain abnormal turning of axons after midline crossing. Patterning along the anterior-posterior axis and expression of rostro-caudal guidance cue gradients should be examined quantitatively.

c) Are the axon guidance defects simply a result of reduced floor plate cell clustering? Can guidance errors be observed before changes in floor plate morphology?

d) The authors claim that changes in floor plate morphology coincide with commissural axon crossing of the midline and interpret this as support for the idea that axon-floor plate contact causes the change in floor plate cell clustering. However, axons have clearly already crossed the floor plate at HH21 (Figure 2N,O) before the floor plate phenotype appears. Moreover, successful knockdown of Endoglycan by HH21 would have to be demonstrated to allow interpretation of this experiment. Lastly, even if the timing was consistent, it should be directly tested whether reduced floor plate cell clustering after Endoglycan knockdown depends on commissural axon crossing, e.g. by ablating commissural neurons.

e) It is important to show that a control dsRNA will not "rescue" the NrCAM knockdown defect. Moreover, rescue of the defect by Endoglycan knockdown could result from reduced floor plate repulsion and other mechanisms besides an effect on axon-floor plate adhesion, so the explanatory power of this experiment is limited. Lastly, does loss of NrCAM affect the Endoglycan knockdown phenotypes (turning after crossing, tortuous trajectory during crossing)?

2) Direct support for an anti-adhesive role of Endoglycan in commissural neurons and Purkinje cells is lacking.

a) How would increased floor plate adhesion explain the rostro-caudal axon guidance defects?

b) The in vitro results using spinal motor neurons do not connect to the relevant cell types. Could commissural neuron and Purkinje cell adhesion/growth be studied in similar assays?

c) It would be important to model adhesive commissural axon-floor plate interactions in vitro and study the effect of Endoglycan gain or loss of function in this system. The same applies for Purkinje neurons and the substrate for their migration.

d) The cerebellar phenotype after Endoglycan knockdown could be explained by numerous mechanisms other than increased Purkinje neuron adhesion. Could removal of cell adhesion molecules rescue the migration defect?

Reviewer #3:

The authors took advantage of a substractive hybridization screen that they made several years ago to investigate the functions of endoglycan in the navigation of spinal commissural projections and in the development to the cerebellum. Endoglycan belongs to a family of cyalomucins also comprising CD34 and Podocalyxin, known to regulate cell-cell interactions through anti-adhesive properties.

The authors report for the first time a requirement for endoglycan in the two contexts that they have examined, which is thus an interesting finding. They also propose that endoglycan functions are mediated via anti-adhesive properties.

The mechanisms that regulate and adjust cell- cell adhesive contacts and how these contacts are coupled to guidance decisions are yet poorly known. Therefore, the study addresses an important and underestimated aspect of axon guidance. Nevertheless, I have the feeling that, beyond the demonstration through RNA interference that endoglycan is necessary for proper navigation and migration, the basic mechanisms underlying the contribution of endoglycan and what makes it that it plays a role at some specific steps of the navigation rather than all along, have been superficially addressed. We are indeed let with a very unclear picture on how it contributes to enable proper navigation of the floor plate, and proper purkinje cell migration. The substractive screen was designed to pick up genes important for rostral turning but from the described phenotypes, it looks like what is primarily affected is the FP crossing. Indeed, the axons still appear able to turn, even though the turning is abnormally disconnected to the crossing. Overall and for these reasons, I have a number of issues which I think are necessary to address. They are listed below.

1) Subtractive hybridization screen: the authors should describe the general strategy and method, even though both have been reported in their previous studies. The rationale is needed to correlate the approach with the expected and observed phenotypes of endoglycan knock-down. In particular, the screen was designed for identifying cues instructing rostral turning guidance. How is regulation of cell adhesion during floor plate relevant for the turning? should be expect a rostral-caudal gradient of endoglycan?

2) In Figure 2 at HH21: the axonin labeling seen in the picture is indeed very strong, not as if there were only a few axons that already had crossed, as indicated in the result section of the manuscript.

3) The HH21 untreated control is not shown. This is needed to compare the shape of the FP at this stage or at least would it be the case, the authors should make it clear that the shape is expected to be similar to that at later stage.

The FP appears strongly disorganized, based on the staining of FP cell soma and nuclei. It would strongly add to illustrate how radial fibers of FP cells look like, because commissural axons navigate within this fiber network. A better morphological characterization of the FP structure is also needed. Are cells losing their bipolar morphology? Are the radial fiber still present and attached to the basal side? In the cerebellum, endoglycan depletion alters cell proliferation, could it be the case for FP cells?

The authors consider that since at HH21 the FP is unaffected when only a few axons hare reached the FP, the FP disorganization at HH26 is therefore an indirect consequence of the presence of the axons. This conclusion is weak and lacks experimental support.

4) The disorganization of FP cells could well alter the presentation pattern of local guidance cues or local cell adhesion molecules. Expression profiles of the principal players should be investigated at protein levels, when antibodies are available.

5) How knock-down of endoglycan impacts on levels of endoglycan is not shown. This is also true for dose-dependent rescue: the authors report that axon guidance was rescued by addition of endoglycan in dI1 neurons at a low concentration or in floor-plate cells at a medium concentration. One would like to see whether these different conditions really result in differences of endoglycan levels. It might be possible to get insights into endoglycan levels be done, maybe with western blot of pieces of spinal cords. On the least, knock-down efficiency and rescue could be assessed on cells transfected with tagged endoglycan?

6) The anti-adhesive role of endoglycan during FP crossing is rather deduced from the literature than supported by the data. The authors need to investigate first whether manipulating the adhesive/anti-adhesive balance results in alterations of FP crossing and post-crossing. According to their model, abrogating endoglycan results in an unbalanced weight of NrCAM over contactin. The authors could for example test whether over expressing contactin strictly mimics NrCAM/endoglycan knock-down.

7) Why were the in vitro experiments achieved with motoneurons rather than commissural neurons? Moreover, the performed analysis rather assesses whether endoglycan regulates axon outgrowth, not really cell-cell contacts and adhesion. it is somehow surprising that this outgrowth effect, if it applies also to commissural axons, does not result in alteration of the pre-crossing navigation, for example by delaying the growth towards the FP.

8) "Cell-autonomous" versus "non-cell autonomous" endoglycan contributions are unclear for FP navigation. On the one hand endoglycan is reported enriched in the FP at HH20, a stage when there are not impact of endoglycan KD on FP cell organization. Later on, higher expression is found in the dorsal spinal cord where commissural neurons are, and at this stage they have crossed the FP. Thus, this increase might rather be important for post-crossing navigation.

9) Is endoglycan relevant for axon fasciculation?

10) The impact of endoglycan deletion on cerebellar development is rather impressive. The interpretations are simplest because Purkinje cells are the only source of endoglycan. This makes it possible to reconstitute the sequence of direct and indirect events leading to the different abnormalities the authors found. Nevertheless, very few attempts are done for characterizing the nature of the interactions and adhesion required during Purkinje cell migration. One would like to know more about the adhesion molecules that are needed for this cell-type, or the nature of the migration process that is impaired. Is it that the leading process sticks to the substrate, or is the nuclei translocation prevented?

11) Also, the Discussion should be extended on the question of folding. Is folding defect a direct consequence of the decrease amount of produced granule cells? or the lack of purkinje cells at their final position?

12) The authors mention that “Pathfinding was normal in embryos electroporated with dsRNA derived from Podocalyxin". They also quote in their introduction that this cyalomucin is also expressed in the developing nervous system. Therefore, why this experiment was done is unclear, as well as a conclusion lacks. Was it done to document a specific contribution of endoglycan over the other cyalomucins? or a specific functional property of endoglycan?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your work entitled "Endoglycan plays a role in axon guidance and neuronal migration by negatively regulating cell-cell adhesion" for consideration by eLife. Your article has been reviewed by a Reviewing Editor and a Senior Editor, a Reviewing Editor, and three reviewers. The following individual involved in review of your submission has agreed to reveal their identity: Valerie Castellani (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

The three reviewers believe that your study offers some new, potentially important insights into the regulation of axon pathfinding. The findings are certainly new and exciting. think that the work nicely shows that Endoglycan contributes to commissural axon navigation. In this revision, the phenotypes have been characterized in more detail, and important control experiments were added.

Nonetheless, the reviewers do not find supporting evidence for your model, that Endoglycan facilitates floor plate navigation by controlling the strength of "adhesion and anti-adhesion" contacts resulting from cell-axon or axon-axon contacts during navigation.

In the reviewer consultation after the reviews below were submitted, the reviewers were in agreement in believing that the in vivo analyses alone do not truly inform on the mechanism of Eendoglycan-mediated cell adhesion. They would like to see in vitro experiments of the type your lab should be able to execute, but using commissural, not motor, neurons, to complement your in vivo analysis. Examining adhesion of commissural neurons to heterologous cells and/or floor plate cells in vitro, combined with manipulations of Endoglycan expression, could directly support (or not) the idea that Endoglycan modulates adhesive interactions between commissural axons and floor plate cells. Without such experiments, the statements about adhesion need to be toned down; adhesion should be critically discussed as one possible mechanism.

Since the requested experiments would take longer than two months, especially in this time of lab ramping-up, your manuscript will be rejected at this time.

Reviewer #1:

In the revised version of their manuscript, Baeriswyl et al., have addressed several of my concerns. Additional control experiments have helped solidify some of the authors' conclusions, and the live imaging experiments provide new information about the effect of Endoglycan knockdown and overexpression on commissural axon behavior during floor plate crossing. However, direct support for the anti-adhesive or "lubricating" role of Endoglycan in the context of commissural axon-floor plate or Purkinje neuron-substrate interactions is still lacking. I am surprised that the authors decided not to include some very doable experiments in their extensive revision work to strengthen this hypothesis. I therefore believe that the authors either need to tone down their conclusions about molecular mechanism or perform the necessary experiments to test their interpretation more directly.

1) The experiments involving motor neurons cultured on Endoglycan-expressing HEK cells should be carried out with commissural neurons instead. These neurons are easy to grow in culture, making this a very feasible experiment. Examining adhesion of commissural neurons to heterologous cells and/or floor plate cells in vitro, combined with manipulations of Endoglycan expression, could directly support (or not) the idea that Endoglycan modulates adhesive interactions between commissural axons and floor plate cells. Without direct evidence of this kind, the language centered around anti-adhesive or lubricating properties of Endoglycan in an axon guidance context does not seem justified.

2) The data on cerebellum development after Endoglycan knockdown are difficult to interpret due to the dramatic effects on overall tissue morphology. Without extensive experiments focused on molecular mechanisms underlying the phenotype, these findings do not integrate well with the rest of the paper and should be removed.

Reviewer #2:

The authors have conducted several new experiments to support their hypothesis that Endoglycan is regulating cell-cell adhesion. While I still think that the manuscript presents exciting new data on axon guidance, the authors still do not provide any direct evidence for an involvement of Endoglycan in regulating cell-cell adhesion.

I appreciate that the authors tried to measure adhesive strength in an in vitro assay as designed by Vance Lemmon and colleagues. However, I am a bit concerned about their statement "we did not succeed in measuring the differences reproducibly". Perhaps there are just no differences?

Unfortunately, also the ablation of commissural neurons suggested by reviewer 2 was not attempted. This experiment would have tested whether the reduced floor plate cell clustering after Endoglycan knockdown indeed depends on direct interactions between axons and floor place cells, which is currently not clear.

The authors present new figures (Figure 6, Figure 7 and Figure 8) and movies "demonstrating that the observed differences in axonal midline crossing are indeed explained by aberrant interactions between growth cones and floor-plate cells". However, the nature of these interactions remains unclear. The 'corkscrew' phenotype shown in Figure 8, Video 4 and Video 5 could also be explained by a change in microtubule dynamics (Krieg et al., 2017). Endoglycan could very well interfere with other signals present during floor plate crossing, providing an alternative explanation for the observed axonal growth patterns.

The authors also claim that "we now only show adhesion of cells (Figure 4)." Figure 4 shows a reduction in cell number, which could not only be a consequence of reduced adhesion but also because of other effects, such as apoptosis.

If I understand correctly, the authors suggest that axons pull on FP cells: "Most importantly, evidence from our in vivo and live cell imaging experiments suggests that axons cause the disruption of the floor plate (see Figure 8 and Video 1 and Video 3, subsection “Endoglycan acts as “lubricant” for growth cone movement in the floor plate”)." I cannot see any such evidence in the presented data. Time lapse movies showing how axons pull on FP cells are currently missing. The ablation of commissural neurons mentioned above would also provide such evidence.

Overall, I'm not sure if I understand the concept of a lubricant in this context. The authors state: "Endoglycan is not affecting adhesion between floor-plate cells, as it is not a cell adhesion molecule, or an “anti-adhesion” molecule to be precise. Rather Endoglycan acts as a “lubricant” that is required between axons and floor-plate cells, or between cells that move with respect to each other, because it allows for the necessary dynamics in the adhesion between growth cones and floor-plate cells mediated by adhesion molecules." What is the difference between anti-adhesion and a lubricant? Both reduce friction, it should be the same? And do the authors suggest that whole axons are moving over the floor plate: "In contrast, at the floor plate axons need to move relative to the floor plate."? As far as I'm aware, axons per se do not move much. Growth is mostly achieved at the distal end of the axon. And if a lubricant would allow sliding there, growth cones could not generate forces required for their motility. Furthermore, if adhesion would be reduced, growth cones should also be smaller. Yet, Figure 11 seems to suggest that growth cone sizes are similar in all conditions.

Thus, while the authors clearly demonstrate, for the first time, the importance of Endoglycan in axon guidance and neuronal migration, I am still not convinced that it acts mostly through regulation of cell-cell adhesion as suggested by the authors.

Reviewer #3:

The study has been significantly extended, in particular with a series of novel experiments performed with live imaging. The resulting data clarify the role of endoglycan and its mode of action. For example, the authors observed that manipulations of endoglycan levels have impact on the velocity of commissural axons and growth cone size. Differences appear tiny but statistically significant. They propose a model whereby endoglycan provided by FP cells and axons acts as a lubricant to facilitate the FP navigation and the axon-FP cell contacts. This is an interesting idea that beyond endoglycan, shed light on yet poorly understood roles of modulators.

In addition, the authors addressed several concerns that were raised by their previous version of the work. They checked whether manipulations of endoglycan alter the expression of molecules known to guide commissural axons, which as far as we can tell from general expression patterns is not the case.

I also think that the part on the cerebellum model is now more integrated to the rest of the work, and I agree that it enable to make a broader message on the functional properties of endoglycan.

I still consider that, for better consistency, the experiments to demonstrate the regulation of adhesion by endoglycan should have been carried out on the population of neurons that is investigated here (commissural neurons) and not on motoneurons. That said, I believe that what the authors report with motoneurons might apply to commissural neurons and the findings are consistent with their model on commissural axon navigation. The authors modified their initial analysis, now concentrating on adhesion rather than outgrowth, which is to me very pertinent. They also carried out some biochemical analysis to document endoglycan mode of action, showing contribution of glycosylation.

Overall, I found that the study is improved by the novel data.

Here are some specific comments:

For all figures related to the time-lapse imaging.: the number of experiments, embryos etc… are not indicated.

It seems to me that there are some redundancy between Figure 6 and Figure 7 in the questions that the experiments address.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Endoglycan plays a role in axon guidance and neuronal migration by negatively regulating cell-cell adhesion" for consideration by eLife. Your article has been reviewed by Didier Stainier as the Senior Editor, a Reviewing Editor, and three reviewers. The following individual involved in review of your submission has agreed to reveal their identity: Kristian Franze (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

This study presents a role for the sialomucin protein Endoglycan in axon guidance and neuronal migration. The authors describe defects in spinal commissural axon guidance and cerebellar Purkinje cell migration that occur after interfering with Endoglycan function. With an array of in vivo and in vitro approaches, especially a novel in vitro experiment, in which neurons detached faster from other cells in the presence of Endoglycan when exposed to forces exerted by a flow of buffer solution, the authors strengthen the view that Endoglycan reduces adhesion between commissural axons and their first intermediate target, the floor plate, thereby facilitating axon growth across the spinal cord midline. This work establishes a novel function for Endoglycan in axon guidance and highlights the importance of balancing adhesive and anti-adhesive forces between growth cones and guidepost cells for proper axon navigation, adding to our understanding of how proper neuronal networks are formed during development.

Essential revisions:

Your study highlights Endoglycan in regulating neuronal growth and migration, pointing to a mechanism involving anti-adhesion. The reviewers felt that you have addressed all the major criticisms, and with the addition of your apt in vitro experiments addressing cell adhesion (new Figure 5), that the manuscript is generally improved. They comment that this will be "a very nice paper that significantly contributes to the field".

There are still revisions called for, as listed below, primarily altering the wording used to describe the action of endoglycan. Moreover, the reviewers encourage you to consider removing the data on the cerebellum.

I) Cerebellar data: Two of the three reviewers urge you to consider removing the cerebellar data on Purkinje cells, as this analysis distracts from the now well fleshed-out message relating to axon guidance. The effects of Endoglycan knockdown on Purkinje cell migration are compelling per se, but the interpretation of these defects is complicated by the overall disrupted morphology of the cerebellum and lack of data that directly demonstrate whether and how Endoglycan modulates Purkinje cell adhesion to other relevant cell types. In your rebuttal, you indicate that defects in granule cell proliferation are consistent with impaired Purkinje cell migration. This is certainly true, but numerous alternative interpretations exist, and no attempt is made to probe the authors' model further. Without extensive experiments focused on molecular mechanisms underlying the phenotype, these findings do not integrate well with the rest of the paper. You might consider publishing the cerebellar work as a smaller study. One question is whether cell migration/translocation and axon outgrowth rely on adhesion/anti-adhesion in the cerebellum as they do in the floor plate.

II) Endoglycan at the floor plate:

1) “Lubricant"? vs negatively regulating cell adhesion:

a) While you now provide an in vitro experiment suggesting a role of endoglycans in regulating cell adhesion, there is no experimental evidence for a role of endoglycans as a lubricant in vivo (which would certainly be very difficult to achieve, if at all possible). This should be kept in mind when interpreting the data. You are thus encouraged to tone down the wording referring to the proposed mechanism throughout the manuscript, for example in statements such as "Endoglycan acts as “lubricant” for growth cone movement in the floor plate".

b) The second part of the title is not backed up by the data ("by negatively regulating cell-cell adhesion"); the abstract should explicitly state that adhesion is affected in vitro.

c) Statements such as "Thus, the aberrant morphology of the floor plate at HH25 is explained by the inability of axons to break contacts with floor-plate cells in the absence of Endoglycan" should be phrased more carefully. Replacing "is explained" by "could be explained" would not make the study weaker but rather protect you in the future should it turn out that there is a different mechanism at play. Finally, because of the potential problems mentioned above associated with the blast experiment, you might omit the word "strong" in the sentence "Strong support for this hypothesis was contributed by in vitro findings that the adhesive strength… ".

d) It is admittedly difficult to distinguish a lubricant effect from a system that would control the strength of an adhesion or anti-adhesion effect (for example by controlling the number of binding partners or their availability). If the idea is that endoglycan does not target a specific adhesion mechanism but rather has a general effect, would you consider the role of endoglycan to be similar to the one attributed to PSA of NCAM adhesion molecules? This should be discussed.

2) Blasting assay: The growth cone blasting assay introduced in the revised version of the manuscript is a very difficult experiment, and your efforts should be praised. You have carefully considered some of the confounding factors; however, some open questions remain.

a) Factors such as the distance of the pipette from the substrate (i.e., from the top surface of the HEK cells) and the area and height of the growth cones will also be crucial determinants of the time required to detach growth cones. How was the height of the pipette tip above the cells controlled, and how accurately could it be determined? Was the growth cone area the same between the groups? Where the experiments done blindly?

b) Furthermore, why was PBS used, which usually does not contain any calcium? Calcium has a key role in cell adhesion.

Details of these aspects should be provided (including images of growth cones at higher magnification and, if possible, a quantification of growth cone areas), and the experiment should be critically discussed.

3) A more detailed description of what constitutes an abnormal injection site in Figure 3 would greatly help with interpretation of those data. Quantification of "normal injection sites" in Figure 3 somewhat unsatisfying, especially for the comparison of gain-of-function and loss-of-function effects. After all, more adhesion and less adhesion should not produce the exact same phenotypes. Could the abnormal injection sites be further categorized, e.g. as stalled, 'corkscrew', premature turning, aberrant postcrossing A-P guidance etc.?

4) You have now used a carpet of HEK cells rather than just a cell adhesion molecule coated on plastic, which you argue "probably also helped to avoid artefacts". It is not clear how a homogeneous, controlled coating of a plastic surface with purified proteins should cause more artifacts than cells expressing a multitude of different proteins at their surface which may vary in space and time. While it is very difficult to see any details in Figure 4, HEK cells in the two groups (Figure 4A) seem to have different morphologies. If this is true, changes in HEK cells could lead to changes in cell-cell contacts to neurons that are independent of Endoglycans. This should be discussed.

5) In the Results section, the interpretation of defects as a problem with adhesion is introduced very early. It would be better to draw that conclusion later, after the experiments in Figure 4 and Figure 5. In general, some of the conclusions in the Results section that combine outcomes from multiple experiments might be better suited for inclusion in the Discussion.

https://doi.org/10.7554/eLife.64767.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

The reviewers found your study proposing the sialomucin Endoglycan as a new player in axon guidance in commissural axon growth and Purkinje cell migration of interest, especially in its role in regulating cell adhesion in these processes. Indeed, the mechanisms that regulate and adjust cell-cell adhesive contacts and how these contacts are coupled to guidance decisions are poorly understood, and your study, which was carefully executed, addresses an important and underestimated aspect of axon guidance. However, the reviewers share the opinion that the experimental evidence supporting the mode of action of Endoglycan as an anti-adhesive factor was lacking, in both the in vivo and in vitro settings.

in vivo, you demonstrate through RNA interference that Endoglycan is necessary for proper navigation and migration, but the basic mechanisms underlying the contribution of Endoglycan and how it functions at specific steps of the navigation rather than all along, were not considered to be addressed in depth. While your knockdown experiments cause misrouting and increased tortuosity, and your rescue experiments implicate a role for Endoglycan in guidance, it is unclear how increased adhesion when Endoglycan is disrupted is responsible for these outcomes. The "loosening" of the floor plate cells rather than their enhanced clumping is a puzzle; might the loosening of the floor plate cells cause the misrouting? And a further puzzling aspect is that axon fasciculation near the floor plate when Endoglycan is perturbed appears unchanged.

We thank the editor and reviewers for the efforts and comments made during the evaluation of our manuscript. We are happy that our work was considered to be carefully executed and that is addresses an important and underestimated aspect of axon guidance. Because the consensus was that the mode of Endoglycan action was not sufficiently explained in our in vivo and in vitro studies, we added many new experiments supporting our conclusion that Endoglycan affects neural circuit formation by modulating cell-cell adhesion during cell migration and axon guidance. In particular, and as detailed below in our one-to-one response to the reviewers’ comments, we have added ex vivo live cell imaging in intact spinal cords to provide additional results supporting our model. Furthermore, we demonstrate that glycosylation is required for Endoglycan function, in line with our hypothesis that Endoglycan acts by lowering cell-cell contacts through its high glycosylation content in the mucin domain. We added control experiments that exclude an indirect effect of Endoglycan on axon guidance at the spinal cord midline through changes in the expression of known axon guidance cues.

Initially, we were also surprised by our findings that the floor plate integrity suffers in the absence of Endoglycan. As indicated, one might think that downregulation of Endoglycan would strengthen the adhesion between floor-plate cells rather than “loosening” them. However, it looks like Endoglycan is not affecting adhesion between floor-plate cells, as it is not a cell adhesion molecule, or an “anti-adhesion” molecule to be precise. Rather Endoglycan acts as a “lubricant” that is required between axons and floorplate cells, or between cells that move with respect to each other, because it allows for the necessary dynamics in the adhesion between growth cones and floor-plate cells mediated by adhesion molecules. This was our conclusion based on all our results presented in the original version of our manuscript. We have made considerable efforts to add further support for this conclusion in our revised manuscript by demonstrating that axon – floor-plate cell interactions change with the amount of Endoglycan (Figure 6, Figure 7, Figure 8 of the revised manuscript). The reviewers also found it puzzling that axon fasciculation near the floor plate was not perturbed. In agreement with our view of Endoglycan function, this makes sense. Endoglycan would not be required between axons that move in the same direction, as they do not need to move against each other. Rather they move in sync, in loose contact but without fasciculation in the strict sense. In contrast, at the floor plate axons need to move relative to the floor plate. Later crossing axons are dislocating previously crossing axons by breaking their contact with the floor-plate cells. Thus, here, a “lubricant” is required to allow axons to move across the midline by balancing adhesion in a dynamic way. We realize that the mechanism of Endoglycan action was maybe not explained clearly enough. We thus tried to put more emphasis on the idea of Endoglycan as a lubricant. With this in mind, it makes sense that the floor-plate integrity is not affected in the absence of Endoglycan unless axons are exerting forces. Please see below for a point-to-point response to issues raised by reviewers.

A more difficult criticism to address, is direct evidence for involvement of Endoglycan in regulating cell-cell adhesion: to demonstrate changes in fasciculation rather than neurite length (which could vary due to many factors), to present both high and low levels of Endoglycan to axons in the non-neuronal cells, and to directly measure cell-cell adhesion strength in the presence of Endoglycan. These experiments are challenging yet would be important and very welcome to the field, should you be able to execute them.

Indeed, it has taken us much more than two months to carry out all the additional experiments. We have tried measurements of adhesion in vitro but this was not very convincing. Therefore, we have changed the strategy and added new live imaging studies and additional in vivo experiments. The live imaging experiments support the fact that Endoglycan affects axon – floor-plate adhesion rather than axon-axon contacts before or after midline crossing. Taken together our previous and our new results strongly support our conclusions that Endoglycan acts as a “lubricant” allowing axons to navigate their intermediate target, the floor plate. See below for details.

A final note is that the Purkinje cell experiments were little criticized but detracted a bit from the main story.

We respectfully disagree. These experiments clearly take away the focus from axon-axon interactions, which is never what we suggested, but unfortunately was not sufficiently clear. In fact, we think that these experiments add to the conclusion that Endoglycan acts as a “lubricant” by regulating cell-cell adhesion in a more general way, not restricted to midline crossing. We have rephrased the text to better explain this point.

Reviewer #1:

In this manuscript, Baeriswyl et al., convincingly show that the sialomucin Endoglycan plays an important role in commissural axon growth and Purkinje cell migration, and suggest that the main function of Endoglycan is to unspecifically regulate cell-cell adhesion.

The manuscript thus identifies a new player in axon guidance (confirming data from a previous screen by the group). However, it is currently not exactly clear how Endoglycan affects axon growth. While the authors make some strong claims about adhesion, there is currently no direct evidence for an involvement of Endoglycan in regulating cell-cell adhesion.

Endoglycan does not affect axon growth per se. For instance, we did not see any difference in the arrival of commissural axons at the floor plate (see also Figure 13). Axons were not found to arrive at the floor-plate entry site later in embryos treated with dsEndo or earlier in embryos overexpressing Endoglycan. The difference was however significant during floor-plate crossing, as shown by our live imaging data added as Figure 6, Figure 7, Table 2, Figure 11, Video 1, Video 2, Video 3, Video 4, Video 5, Results.

If Endoglycan indeed opposes cell adhesion, why don't the floor plate cells stick together better in the knockdowns than in controls (rather than the other way around as it is shown in Figure 2)? The authors address this issue in the Discussion, however, their arguments are not very convincing. The fact that at HH21 the floor plate was intact in the absence of Endoglycan doesn't necessarily mean that cell adhesion is not affected or that Endoglycan is not required for adhesion between floorplate cells. An alternative explanation might be that there are only weak mechanical forces acting on these cells at early stages, which might increase during development and eventually be strong enough to tear apart cells that have reduced adhesion.

Furthermore, if cell-cell adhesion is changed, I would also expect a change in fasciculation of axon bundles on their way to the floor plate. Figure 1 seems to suggest that this might not be the case. To address this issue, fasciculation should be quantified.

Initially, we were also surprised to find the changes in floor-plate integrity. However, at second glance, our data clearly support a role of Endoglycan as a “lubricant”. It is not the opposite of an adhesion molecule but a dynamic regulator of adhesion that allows the movement of growth cones across the midline. Axons that cross the floor plate need to dislocate the axons that have crossed before. This process generates stress/force between the axons and the floor-plate cells. In the absence of the lubricant, the mechanical forces are too big and the floor plate is disrupted. Therefore, we find single cells dislocated into the axonal commissure, Purkinje cells stuck on their way to the periphery, but never a general disassembly of the tissue. This is clearly illustrated by the integrity of the spinal cord tissue after electroporation of dsRNA or after overexpression of Endoglycan (new Figure 3—figure supplement 1, Figure 3—figure supplement 3 and Figure 13, Results).

Reviewer 1 suggests an alternative explanation: only weak mechanical forces would act at HH21 and therefore might not be sufficient to explain, why the floor plate does not lose its integrity at that stage, but would do at HH25, when forces get stronger. This is actually absolutely in line with our explanation: Axons (the producers of mechanical forces) are compromising floor-plate integrity because they stick too strongly to the floor-plate cells in the absence of the lubricant Endoglycan (See Video 4 and Video 5, subsection “Endoglycan acts as ‘lubricant’ for growth cone movement in the floor plate”). As pointed out, we cannot completely rule out that the adhesion between floor plate cells is also affected. However, the “gaps” in the floor plate can only be explained by a higher adhesion between axons and floor-plate cells, if cells would adhere less strongly to each other and therefore the floor plate would “fall apart” only later, why would it do so in the absence of Endoglycan that enhances the adhesiveness?

Similarly, our findings that the Endoglycan loss-of-function phenotype can also be rescued when Endoglycan is provided by the axons further supports our “lubricant” hypothesis. Strongest support comes from our live imaging data, where we clearly find that the interaction between axons and floorplate cells is compromised but only after early axons have crossed the floor plate, in line with our observations and interpretation of the findings shown in Figure 2.

I also don't find the in vitro experiments very convincing. Figure 5 shows that neurite length is decreased in Endoglycan overexpressing motoneurons but not when it is overexpressed in COS cells. First, if the main effect of Endoglycan is the unspecific regulation of cell-cell adhesion, I would have expected a stronger effect of the overexpression in COS cells even if the transfection efficiency was only ~50%. Second, it would be good to see a similar set of experiments with decreased expression of Endoglycan. What would the authors expect? And lastly, changes in neurite length can be caused in many different ways, certainly not allowing to conclude “that Endoglycan acts as a negative regulator of cell-cell adhesion”.

We have removed Figure 5 and replaced it by new Figure 4 and Figure 5—figure supplement 2 (subsection “Endoglycan is a negative regulator of cell adhesion”). We realize that these results were confusing in the way they were presented. Figure 4 shows that fewer neurons attach to HEK cells stably expressing Endoglycan, whereas Figure 5—figure supplement 2shows that the “anti-adhesive” effect of Endoglycan is dependent on its glycosylation in the mucin domain, as removal of poly-sialic acid or O-linked glycosylation abolishes the anti-adhesive effect.

Similarly, Figure 5—figure supplement 1 only shows a few images, no quantification of data is shown. For how long were neurons allowed to grow? And even if there are reproducibly less neurons growing on transfected HEK cells, this could be a consequence of many factors (one of them of cause being a change in adhesion).

See above. We have removed (old) Figure 4 looking at neurite length. We have quantified the decrease in adhesiveness by counting cells per area in new Figure 4 (Results; see also Figure 5—figure supplement 2) for a role in glycosylation in the adhesion-modulating effect.

In summary, the manuscript presents new and exciting data on axon guidance, which might be very relevant for many other systems in which cell-cell adhesion is important. However, while some of the data suggest that Endoglycan might mainly act through the regulation of cell adhesion, this remains an hypothesis until it is directly shown. In order to convincingly demonstrate how Endoglycan regulates axon growth, the authors should directly measure cell-cell adhesion strength and quantify its dependence on the presence of Endoglycan.

We have included live imaging analyses as additional evidence and strong support for our conclusion that Endoglycan is a regulator of cell-cell adhesion by acting as a lubricant (Figure 6, Figure 6, Figure 8; Video 1, Video 2, Video 3, Video 4, Video 5, Results). Please note that it is NOT our intention to say that Endoglycan regulates axon growth. This does not appear to be the case, as axons get to the floor plate at the expected time (observations made by the analyses of open-book preparations, see Figure 2, and observed in live imaging preparations). There is no change in the pre-crossing axon tract (Figure 13). Axons are only affected when they are interacting with their intermediate target, this is during midline crossing, where indeed axons are slower in the absence of Endoglycan (thus stick more) and faster upon overexpression of Endoglycan (when they stick less). We do not think that they are affected in their intrinsic growth potential. This is also reflected by the fact that axons are faster in the second half of the floor plate under the conditions shown in Figure 6E2. If axons would be affected in intrinsic growth potential, they would be expected to maintain their speed.

We accept the criticism that we do not provide a direct measurement for the adhesive strength with and without Endoglycan. We tried to measure adhesive strength in an in vitro assay as designed by Vance Lemmon and colleagues (Lemmon et al., (1992)). However, we did not succeed in measuring the differences reproducibly. We therefore opted to carry out the live imaging studies, which support a change in the interaction/adhesion between floor-plate cells and growth cones (see Figure 8 and Video4, Video 5).

Reviewer #2:

In this manuscript by Baeriswyl et al., the authors describe defects in spinal commissural axon guidance and cerebellar Purkinje cell migration that occur after interfering with the function of the sialomucin Endoglycan. Baeriswyl et al., suggest that these phenotypes reflect a function of Endoglycan in reducing cell-cell adhesion.

The authors show that Endoglycan knockdown in the chick spinal cord floor plate or floor plate plus commissural neurons disrupts clustering of floor plate cells and causes increased axon tortuosity during floor plate crossing and abnormal caudal turning of commissural axons after midline crossing. Expression of Endoglycan using commissural neuron- or floor plate-specific promoters can rescue these axon guidance defects, depending on expression levels. The authors further show that knockdown of Endoglycan in the floor plate can suppress the premature commissural axon turning phenotype that results from NrCAM knockdown, and in vitro experiments support the idea that Endoglycan can reduce adhesion of motor neurons to heterologous cells. Lastly, knockdown of Endoglycan in the cerebellum disrupts Purkinje cell positioning and overall cerebellar morphology.

This work presents a series of novel observations, and the idea that Endoglycan modulates cell-cell adhesion to control axon guidance and neuronal migration is of potential interest to the field, even though such an anti-adhesive mechanism for axon guidance has been extensively explored in the context of other molecules, e.g. NCAM. The main conclusions about the mechanism of Endoglycan function in axon guidance and cell migration, however, are not sufficiently supported by the data, and some important control experiments are missing. These problems, together with the limited conceptual novelty of the findings, do not make this manuscript a strong candidate for publication in eLife.

1) The experiments involving Endoglycan knockdown in the spinal cord are difficult to interpret for multiple reasons. It is not clear that increased axon-floor plate adhesion is responsible for the observed effects.

We have added live imaging studies (Figure 6, Figure 7, Figure 8) demonstrating that the observed differences in axonal midline crossing are indeed explained by aberrant interactions between growth cones and floor-plate cells (see Figure 8 and Video 4, Video 5; subsection “Endoglycan acts as ‘lubricant’ for growth cone movement in the floor plate”).

a) Knockdown is targeted to the floor plate or "one half of the spinal cord including the floor plate". Data that validate successful knockdown in the targeted cell populations should be included. Also, why is knockdown in commissural neurons alone not attempted?

We did knockdown Endoglycan only dorsally (see graphic in Figure 1D and Figure 1J, Results). Moreover, the requirement for Endoglycan in dI1 neurons can also be concluded from our rescue experiments, where the expression of Endoglycan only in dI1 neurons, using the Math1::Endoglycan construct, was sufficient to rescue the phenotype observed after downregulation of Endoglycan in one side of the spinal cord including the floor plate (Results).

b) The data showing normal floor plate differentiation (Figure S3) are not sufficient to exclude abnormal tissue patterning as a cause for the axon guidance defects. Shh expression appears reduced (panel C), which could explain abnormal turning of axons after midline crossing. Patterning along the anterior-posterior axis and expression of rostro-caudal guidance cue gradients should be examined quantitatively.

We have added more detailed analyses of spinal cord patterning after downregulation of Endoglycan in one half of the spinal cord including the floor plate (new Figure 3—figure supplement 1, Results). Furthermore, we extended the analysis of Shh expression to include also Wnt5a along the anteroposterior axis and found no changes in any of the experimental groups compared to controls (Figure 3—figure supplement 4, Results). Along the same lines, we also did not find any differences in the expression of the known guidance molecules Axonin-1/Contactin2 and NrCAM (Figure 3—figure supplement 3, Results).

c) Are the axon guidance defects simply a result of reduced floor plate cell clustering? Can guidance errors be observed before changes in floor plate morphology?

The gaps in the floor plate are unlikely to explain the axon guidance phenotypes alone, as the floor plate is still largely intact as a structure. Most importantly, evidence from our in vivo and live cell imaging experiments suggests that axons cause the disruption of the floor plate (see Figure 8 and Video 4, Video 5; Results). See also response to reviewer 1.

d) The authors claim that changes in floor plate morphology coincide with commissural axon crossing of the midline and interpret this as support for the idea that axon-floor plate contact causes the change in floor plate cell clustering. However, axons have clearly already crossed the floor plate at HH21 (Figure 2N,O) before the floor plate phenotype appears. Moreover, successful knockdown of Endoglycan by HH21 would have to be demonstrated to allow interpretation of this experiment. Lastly, even if the timing was consistent, it should be directly tested whether reduced floor plate cell clustering after Endoglycan knockdown depends on commissural axon crossing, e.g. by ablating commissural neurons.

As the reviewer correctly states, some axons have already crossed the midline by HH21. These are mostly more ventral populations of commissural neurons. The changes in floor plate integrity require axons to dislocate previously crossing axon from contact with floor-plate cells. This is why we chose to show HH21, a stage, where we did not yet see the gaps in the floor plate but where some early axons have already crossed (Figure 2). This is to show that the floor plate does not disintegrate due to downregulation of Endoglycan but is affected only at later stages when axon-floor-plate cell contacts have to be dissolved by later following axons. This conclusion is also confirmed by our live-imaging results where we saw aberrant axon-floor-plate cell interactions only after 10 hours, indicating that it is the excessive adhesion of axons with the floor-plate cells that results in the dislocation of floor-plate cells (Figure 8, Video 4 and Video 5).

e) It is important to show that a control dsRNA will not "rescue" the NrCAM knockdown defect. Moreover, rescue of the defect by Endoglycan knockdown could result from reduced floor plate repulsion and other mechanisms besides an effect on axon-floor plate adhesion, so the explanatory power of this experiment is limited. Lastly, does loss of NrCAM affect the Endoglycan knockdown phenotypes (turning after crossing, tortuous trajectory during crossing)?

As suggested by the reviewer, we have extended our analysis of the concomitant downregulation of NrCAM and Endoglycan with dsCD34 and dsPodxl (Figure 5, Results). Both dsRNAs were not able to counteract the loss of NrCAM function, as axons were still found to turn aberrantly along the ipsilateral floor-plate border. Both genes are expressed in the spinal cord during the time window of commissural axon navigation across the floor plate (Figure 2—figure supplement 2), although mRNA for Podxl1 was only found in precursors of dI1 neurons, not mature neurons. However, the protein may of course persist in neurons.

And indeed, the “corkscrew-phenotype” was no longer observed after concomitant downregulation of NrCAM and Endoglycan.

2) Direct support for an anti-adhesive role of Endoglycan in commissural neurons and Purkinje cells is lacking.

a) How would increased floor plate adhesion explain the rostro-caudal axon guidance defects?

This is still an open question. We can rule out a change in morphogen gradients, as both the Shh gradient and Wnt5a expression were unchanged in embryos electroporated with dsEndoglycan (Figure 3—figure supplement 4).

We can only speculate that the changes in axon-floor-plate cell interactions will prevent growth cones from reading these cues correctly (Discussion).

b) The in vitro results using spinal motor neurons do not connect to the relevant cell types. Could commissural neuron and Purkinje cell adhesion/growth be studied in similar assays?

We do not have a method to culture Purkinje cells and because this in vitro assay is used to support our results from in vivo experiments, we opted for live cell imaging rather than an in vitro assay with commissural neurons. These live cell imaging experiments clearly demonstrate that aberrant interactions between growth cones and floor plate cells explain the observed in vivo phenotypes (see Figure 8 and Video 4 and Video 5).

c) It would be important to model adhesive commissural axon-floor plate interactions in vitro and study the effect of Endoglycan gain or loss of function in this system. The same applies for Purkinje neurons and the substrate for their migration.

An in vitro model of floor plate-growth cone interactions would be very difficult to model, as on both sides a plethora of molecules is involved in a dynamically changing pattern. Therefore, as mentioned above, we opted for an ex vivo live imaging approach. This allows for visualization of individual growth cones and their interaction with the floor plate, thus providing the advantages of an in vitro system, but without the disadvantages of the in vitro system, as the in vivo complexity is maintained. The results of these experiments are completely in line with our model and confirm that Endoglycan fine-tunes the interactions between growth cones and floor-plate cells. (See Figure 6, Figure 7, and Figure 8, along with the Video 1, Video 2, Video 3, Video 4, Video 5, Results).

d) The cerebellar phenotype after Endoglycan knockdown could be explained by numerous mechanisms other than increased Purkinje neuron adhesion. Could removal of cell adhesion molecules rescue the migration defect?

In contrast to the situation at the floor plate, the molecules involved in Purkinje cell migration have not been identified. Therefore, it is not possible to do the experiment suggested by the reviewer.

Reviewer #3:

The authors took advantage of a substractive hybridization screen that they made several years ago to investigate the functions of endoglycan in the navigation of spinal commissural projections and in the development to the cerebellum. Endoglycan belongs to a family of cyalomucins also comprising CD34 and Podocalyxin, known to regulate cell-cell interactions through anti-adhesive properties.

The authors report for the first time a requirement for endoglycan in the two contexts that they have examined, which is thus an interesting finding. They also propose that endoglycan functions are mediated via anti-adhesive properties.

The mechanisms that regulate and adjust cell- cell adhesive contacts and how these contacts are coupled to guidance decisions are yet poorly known. Therefore, the study addresses an important and underestimated aspect of axon guidance. Nevertheless, I have the feeling that, beyond the demonstration through RNA interference that endoglycan is necessary for proper navigation and migration, the basic mechanisms underlying the contribution of endoglycan and what makes it that it plays a role at some specific steps of the navigation rather than all along, have been superficially addressed. We are indeed let with a very unclear picture on how it contributes to enable proper navigation of the floor plate, and proper purkinje cell migration. The substractive screen was designed to pick up genes important for rostral turning but from the described phenotypes, it looks like what is primarily affected is the FP crossing. Indeed, the axons still appear able to turn, even though the turning is abnormally disconnected to the crossing. Overall and for these reasons, I have a number of issues which I think are necessary to address. They are listed below.

1) Subtractive hybridization screen: the authors should describe the general strategy and method, even though both have been reported in their previous studies. The rationale is needed to correlate the approach with the expected and observed phenotypes of endoglycan knock-down. In particular, the screen was designed for identifying cues instructing rostral turning guidance. How is regulation of cell adhesion during floor plate relevant for the turning? should be expect a rostral-caudal gradient of endoglycan?

The screen was designed to identify genes with differential expression in the floor plate at stage HH26, when axons have crossed, compared to HH19/20, before dI1 axons have entered the floor plate. From the identified genes, we randomly selected some for functional testing. Endoglycan was one of them. In fact, it is not even clear why it showed up in our screen, as it is expressed in the floor plate at both stages. Therefore, we feel that it is not indicated to put too much emphasis on this screen. The reason why we looked at this gene in more detail really was its functional contribution to proper floor plate navigation. We did add more information about the screen to the Materials and methods part.

2) In Figure 2 at HH21: the axonin labeling seen in the picture is indeed very strong, not as if there were only a few axons that already had crossed, as indicated in the result section of the manuscript.

The reason to choose HH21 was to use the oldest stage, where we did not see gaps in the floor plate, because only more ventral populations of commissural axons have already crossed the midline. No or only few dI1 axons are in the floor plate at this stage. This strongly supports our conclusion that the floor plate is not directly affected by the loss of Endoglycan but that it is the dynamic contact between axons and the floor plate that causes the changes in floor-plate integrity. Later crossing axons are competing with previously crossing axons for contact with the floor-plate cells. These findings are further supported by the newly added live imaging experiments showing changes in growth cone-floor plate interaction after downregulation of Endoglycan or overexpression of Endoglycan that are consistent with the idea that adhesion between growth cones and floor-plate cells is fine-tuned by the amount of Endoglycan. (Please see Figure 6, Figure 7, and Figure 8, along with the Video 1, Video 2, Video 3, Video 4, Video 5, Results).

3) The HH21 untreated control is not shown. This is needed to compare the shape of the FP at this stage or at least would it be the case, the authors should make it clear that the shape is expected to be similar to that at later stage.

We have added untreated controls to new Figure 2, panels M-O, quantification in V and W.

The FP appears strongly disorganized, based on the staining of FP cell soma and nuclei. It would strongly add to illustrate how radial fibers of FP cells look like, because commissural axons navigate within this fiber network. A better morphological characterization of the FP structure is also needed. Are cells losing their bipolar morphology? Are the radial fiber still present and attached to the basal side? In the cerebellum, endoglycan depletion alters cell proliferation, could it be the case for FP cells?

The authors consider that since at HH21 the FP is unaffected when only a few axons hare reached the FP, the FP disorganization at HH26 is therefore an indirect consequence of the presence of the axons. This conclusion is weak and lacks experimental support.

We have looked at apoptosis of floor-plate cells, but did not see any Cleaved-caspase3-positive cells in any of the conditions. Furthermore, based on the analysis of patterning (Figure 3—figure supplement 1) or the expression of Shh and Wnt5a in the floor plate, we do not have any indication that the floor plate may have any major problems in structure or function. This is also based on our analysis of live axons crossing the floor plate. There are single mislocalized cells in the absence of Endoglycan, as also shown in Figure 2, but no major disruption of floor plate morphology or function (Video 1, Video 2, Video 3, Video 4, Video 5, Figure 8).

4) The disorganization of FP cells could well alter the presentation pattern of local guidance cues or local cell adhesion molecules. Expression profiles of the principal players should be investigated at protein levels, when antibodies are available.

We have carried out these control experiments and compared the expression of Contactin-2/Axonin-1, NrCAM and the expression of Shh and Wnt5a between controls and the experimental groups with less or more Endoglycan (electroporation with dsEndo or overexpression of Endoglycan, respectively). We did not find any differences in expression of any of these well-known guidance cues, suggesting that Endoglycan does not act on axonal midline crossing by interfering with the expression of other guidance cues but rather by modulating the interaction between growth cone and floor-plate molecules. Results are shown in Figure 3—figure supplement 3 and Figure 3—figure supplement 4, respectively; Results.

5) How knock-down of endoglycan impacts on levels of endoglycan is not shown. This is also true for dose-dependent rescue: the authors report that axon guidance was rescued by addition of endoglycan in dI1 neurons at a low concentration or in floor-plate cells at a medium concentration. One would like to see whether these different conditions really result in differences of endoglycan levels. It might be possible to get insights into endoglycan levels be done, maybe with western blot of pieces of spinal cords. On the least, knock-down efficiency and rescue could be assessed on cells transfected with tagged endoglycan?

We have added the quantification of Endoglycan knockdown as Figure 2—figure supplement 3. Because we do not have antibodies that recognize Endoglycan, quantification of Endoglycan with the precision level required to measure these differences is not possible. We would like to point out that the electroporation parameters used in these experiments successfully transfect about 50% of the cells. Therefore, knockdown of Endoglycan will result in only 50% of Endoglycan decrease as the theoretical maximum. However, at the cellular level, our results indicate that the loss of Endoglycan expression is close to or complete.

We did not attempt to see differences between the different concentrations used in our rescue experiments (shown as Figure 3). The two enhancers have different efficiencies in driving expression. Again, we would need specific antibodies to be able to see differences.

6) The anti-adhesive role of endoglycan during FP crossing is rather deduced from the literature than supported by the data. The authors need to investigate first whether manipulating the adhesive/anti-adhesive balance results in alterations of FP crossing and post-crossing. According to their model, abrogating endoglycan results in an unbalanced weight of NrCAM over contactin. The authors could for example test whether over expressing contactin strictly mimics NrCAM/endoglycan knock-down.

We have added new evidence for the change in adhesive properties in our Videos and in Figure 8, where the aberrant contact between axons and floor-plate cells is shown in detail (Results). Furthermore, we show that the anti-adhesive function is due to the glycosylation of Endoglycan (Figure 5—figure supplement 2; Results). We extended the analysis of the counterbalance between Contactin2/NrCAM interactions (the adhesion-promoting interaction) and Endoglycan level (the adhesion-impeding interaction) in new Figure 5 (Results). We do not quite understand the statement of the reviewer: …”unbalanced weight of NrCAM over contactin”. The experiments here and those done in a previous study (Philipp et al., 2012) are in line with the notion that Contactin2 and NrCAM interact to make axons enter the floor plate. This interaction is weakened when NrCAM is downregulated, resulting in aberrant ipsilateral turns. However, the concomitant strengthening of axonfloor-plate interactions by downregulation of Endoglycan rescues the phenotype specifically. Other family members of Endoglycan, Podocalyxin1 or CD34, have no effect.

7) Why were the in vitro experiments achieved with motoneurons rather than commissural neurons? Moreover, the performed analysis rather assesses whether endoglycan regulates axon outgrowth, not really cell-cell contacts and adhesion. it is somehow surprising that this outgrowth effect, if it applies also to commissural axons, does not result in alteration of the pre-crossing navigation, for example by delaying the growth towards the FP.

We have changed this experiment, as we do not think that Endoglycan has any effect on axon growth. Therefore, we now only show adhesion of cells (Figure 4). However, we have expanded the analysis and show that the anti-adhesive effect is due to the glycosylation of Endoglycan (Figure 5—figure supplement 2; Results). We did not find any difference in axonal arrival at the floor plate, nor did we see any differences in the pre-crossing trajectory between the experimental groups or in comparison to the controls (Figure 13; Results).

8) "Cell-autonomous" versus "non-cell autonomous" endoglycan contributions are unclear for FP navigation. On the one hand endoglycan is reported enriched in the FP at HH20, a stage when there are not impact of endoglycan KD on FP cell organization. Later on, higher expression is found in the dorsal spinal cord where commissural neurons are, and at this stage they have crossed the FP. Thus, this increase might rather be important for post-crossing navigation.

We have carried out studies to address the location of Endoglycan function in our system. Based on these (Figure 3), it does not matter where Endoglycan comes from, as the rescue works when we replace Endoglycan either in dI1 neurons (with the Math1 enhancer-driven construct) or in the floor plate (with the Hoxa1-driven construct). These findings are in line and support our conclusion that Endoglycan acts as a lubricant: The amount matters but not where it comes from. (Results).

9) Is endoglycan relevant for axon fasciculation?

No, we did not see any changes in axons extending towards or across the midline (Figure 13).

10) The impact of endoglycan deletion on cerebellar development is rather impressive. The interpretations are simplest because Purkinje cells are the only source of endoglycan. This makes it possible to reconstitute the sequence of direct and indirect events leading to the different abnormalities the authors found. Nevertheless, very few attempts are done for characterizing the nature of the interactions and adhesion required during Purkinje cell migration. One would like to know more about the adhesion molecules that are needed for this cell-type, or the nature of the migration process that is impaired. Is it that the leading process sticks to the substrate, or is the nuclei translocation prevented?

We agree that it would be great to know more about the adhesive interactions driving Purkinje cell migration. However, this is clearly beyond the scope of this study.

11) Also, the Discussion should be extended on the question of folding. Is folding defect a direct consequence of the decrease amount of produced granule cells? or the lack of purkinje cells at their final position?

It has been shown that Purkinje cells are the source of Shh and that Shh signaling promotes granule cell proliferation. The increasing number of granule cells is responsible for the folding of the cerebellar anlage. These studies are very concisely summarized in a review by De Luca et al., (2016). We have added this to the text (Discussion).

12) The authors mention that “Pathfinding was normal in embryos electroporated with dsRNA derived from Podocalyxin". They also quote in their introduction that this cyalomucin is also expressed in the developing nervous system. Therefore, why this experiment was done is unclear, as well as a conclusion lacks. Was it done to document a specific contribution of endoglycan over the other cyalomucins? or a specific functional property of endoglycan?

The analysis of the other two sialomucin family members supports the specificity of our approach. We used them as controls in our initial loss-of-function studies (Figure 1). Furthermore, we extended the in vivo experiments demonstrating the balance of adhesion between axons and floor plate by adding the effect of dsPodocalyxin and dsCD34 (Figure 5).

[Editors’ note: what follows is the authors’ response to the second round of review.]

The three reviewers believe that your study offers some new, potentially important insights into the regulation of axon pathfinding. The findings are certainly new and exciting. think that the work nicely shows that Endoglycan contributes to commissural axon navigation. In this revision, the phenotypes have been characterized in more detail, and important control experiments were added.

Nonetheless, the reviewers do not find supporting evidence for your model, that Endoglycan facilitates floor plate navigation by controlling the strength of "adhesion and anti-adhesion" contacts resulting from cell-axon or axon-axon contacts during navigation.

We are glad to hear that the reviewers consider our study demonstrating a regulatory role of Endoglycan in axon guidance new and exciting. Of course, we are also disappointed that the reviewers are still not convinced of the mechanisms of Endoglycan activity that we propose. We kept trying to find a solution for the direct measurement of adhesive strength in vitro. Despite the fact that this seems to be a simple experiment, it is not. However, in the meantime, we did succeed in finding a way that demonstrates directly that Endoglycan changes the adhesive strength of growth cones of commissural neurons and carpet cells. For this purpose, we adapted the “growth cone blasting” experiment used by Vance Lemmon and colleagues in the 90s (see below).

In the reviewer consultation after the reviews below were submitted, the reviewers were in agreement in believing that the in vivo analyses alone do not truly inform on the mechanism of Eendoglycan-mediated cell adhesion. They would like to see in vitro experiments of the type your lab should be able to execute, but using commissural, not motor, neurons, to complement your in vivo analysis. Examining adhesion of commissural neurons to heterologous cells and/or floor plate cells in vitro, combined with manipulations of Endoglycan expression, could directly support (or not) the idea that Endoglycan modulates adhesive interactions between commissural axons and floor plate cells. Without such experiments, the statements about adhesion need to be toned down; adhesion should be critically discussed as one possible mechanism.

As suggested by the reviewers, we have carried out in vitro experiments that demonstrate directly that the adhesion between growth cones of commissural neurons and HEK cells expressing Endoglycan is strongly reduced compared to control HEK cells (new Figure 5). As also shown in our in vivo experiments, and the in vitro experiments presented in our previous versions of the manuscript, Endoglycan lowers adhesion no matter whether it is provided by the neurons or by the floor plate or HEK cell carpet. This is what you would expect from a “lubricant”. In the current, revised version, of our manuscript, we have added adhesion assays with commissural neurons and also demonstrated that the effect of Endoglycan depends on its post-translational modification (new Figure 4). Thus, as suggested by the reviewers, we have repeated our findings from the original manuscript, where we did these experiments with motoneurons, with commissural neurons. One of the reasons to use motoneurons for the in vitro adhesion assays was to demonstrate that the effect of Endoglycan is not restricted to commissural neurons but would generally be valid in areas where it is expressed. For this reason, we kept the results obtained with motoneurons in the supplementary material (Supplementary Figures 7 and 8).

Reviewer #1:

In the revised version of their manuscript, Baeriswyl et al., have addressed several of my concerns. Additional control experiments have helped solidify some of the authors' conclusions, and the live imaging experiments provide new information about the effect of Endoglycan knockdown and overexpression on commissural axon behavior during floor plate crossing. However, direct support for the anti-adhesive or "lubricating" role of Endoglycan in the context of commissural axon-floor plate or Purkinje neuron-substrate interactions is still lacking. I am surprised that the authors decided not to include some very doable experiments in their extensive revision work to strengthen this hypothesis. I therefore believe that the authors either need to tone down their conclusions about molecular mechanism or perform the necessary experiments to test their interpretation more directly.

We hope that we could relieve any doubts about the validity of our claims about the “lubricant” or adhesion modulating function of Endoglycan with our new in vitro experiments involving commissural neurons (Figure 4 and Figure 5).

1) The experiments involving motor neurons cultured on Endoglycan-expressing HEK cells should be carried out with commissural neurons instead. These neurons are easy to grow in culture, making this a very feasible experiment. Examining adhesion of commissural neurons to heterologous cells and/or floor plate cells in vitro, combined with manipulations of Endoglycan expression, could directly support (or not) the idea that Endoglycan modulates adhesive interactions between commissural axons and floor plate cells. Without direct evidence of this kind, the language centered around anti-adhesive or lubricating properties of Endoglycan in an axon guidance context does not seem justified.

As mentioned above, these experiments have been done now (Figure 4). We have gone one step further and adapted a “growth cone blasting” experiment (Lemmon et al., 1992) that was originally used for cell adhesion molecules coated on tissue culture plastic to our needs.

In brief, we have cultured commissural neurons on a carpet of HEK cells (either control cells or cells expressing Endoglycan) and then we measured the time it took to detach growth cones from the HEK cell carpet depending on the presence of Endoglycan. To this end, we have placed a micropipette from which we directed a constant flow of PBS at the growth cone from a defined angle. We only compared times between control and Endoglycan conditions that were obtained with the same pipette to make sure that the forces affecting the growth cones were constant and reproducible. Using this assay, we could demonstrate that growth cones were detached (or blasted off) much faster, when they were growing on HEK cells expressing Endoglycan compared to control HEK cells.

As we expected, we found the same result when we used HEK cells without Endoglycan and overexpressed Endoglycan in the neurons instead. This again demonstrated that the source of Endoglycan is not important but its presence, in line with what you would expect from a lubricant.

2) The data on cerebellum development after Endoglycan knockdown are difficult to interpret due to the dramatic effects on overall tissue morphology. Without extensive experiments focused on molecular mechanisms underlying the phenotype, these findings do not integrate well with the rest of the paper and should be removed.

Yes, we agree that the changes in cerebellar morphology are dramatic. However, this is what has been described as a consequence of reduced granule cell proliferation. Because granule cell proliferation is dependent on Shh released by Purkinje cells, the observed phenotype is absolutely in line with these descriptions and in line with the developmental trajectory of the cerebellum. The review by De Luca et al., (2016) summarizes the literature and the studies describing the effect of Purkinje cell-derived Shh on granule cell proliferation and cerebellar morphology.

Reviewer #2:

The authors have conducted several new experiments to support their hypothesis that Endoglycan is regulating cell-cell adhesion. While I still think that the manuscript presents exciting new data on axon guidance, the authors still do not provide any direct evidence for an involvement of Endoglycan in regulating cell-cell adhesion.

I appreciate that the authors tried to measure adhesive strength in an in vitro assay as designed by Vance Lemmon and colleagues. However, I am a bit concerned about their statement "we did not succeed in measuring the differences reproducibly". Perhaps there are just no differences?

No, our difficulties were not due to the fact that there were no differences. But the experiment as we tried it initially was extremely artificial and subject to variations. We kept trying and reduced the pulsatile nature of the medium stream from the micropipette. Using a cell carpet rather than just a cell adhesion molecule coated on plastic probably also helped to avoid artefacts. With the version of the “growth cone blasting” experiments as presented in the revised manuscript, we are confident to have directly shown that the adhesive forces between growth cones and HEK cells are reduced by the presence of Endoglycan.

Unfortunately, also the ablation of commissural neurons suggested by reviewer 2 was not attempted. This experiment would have tested whether the reduced floor plate cell clustering after Endoglycan knockdown indeed depends on direct interactions between axons and floor place cells, which is currently not clear.

We have opted for the above-mentioned version of demonstrating the anti-adhesive role of Endoglycan. Ablating commissural neurons in very young spinal cord preparations would have caused tremendous morphological changes and artifacts that would prevent the clear interpretation of the results.

The authors present new figures (Figure 6, Figure 7 and Figure 8) and movies "demonstrating that the observed differences in axonal midline crossing are indeed explained by aberrant interactions between growth cones and floor-plate cells". However, the nature of these interactions remains unclear. The 'corkscrew' phenotype shown in Figure 8, Video 4 and Video 5 could also be explained by a change in microtubule dynamics (Krieg et al., 2017). Endoglycan could very well interfere with other signals present during floor plate crossing, providing an alternative explanation for the observed axonal growth patterns.

Because our experiments clearly show that Endoglycan has its effect independent of its source, we exclude an effect on microtubule dynamics. It would also be unlikely that a change in microtubule dynamics would manifest itself only in one half of the floor plate or not at all in pre-crossing axon growth. For these reasons, we belief that our explanation that is supported by live imaging results and now corroborated by the new in vitro adhesion assays better explains the activity of Endoglycan.

The authors also claim that "we now only show adhesion of cells (Figure 4)." Figure 4 shows a reduction in cell number, which could not only be a consequence of reduced adhesion but also because of other effects, such as apoptosis.

We have no evidence for any effect on apoptosis, as in none of our experiments Endoglycan had an effect on cell death, neither enhancing nor preventing cell death.

If I understand correctly, the authors suggest that axons pull on FP cells: "Most importantly, evidence from our in vivo and live cell imaging experiments suggests that axons cause the disruption of the floor plate (see Figure 8 and Video 1 and Video 3, Results)." I cannot see any such evidence in the presented data. Time lapse movies showing how axons pull on FP cells are currently missing. The ablation of commissural neurons mentioned above would also provide such evidence.

Maybe our wording was not clear enough. What we mean is that growth cones and axons crossing the floor plate stick to floor-plate cells mediated by cell adhesion molecules, but obviously, this interaction is transient. Axons crossing the axons after the first ones have done so need to displace the previously crossing axons from the floor-plate cells. This requires a detachment between axons and floor-plate cells. However, if this adhesion is too strong (as seen in the absence of Endoglycan) then the “weakest bond” that is broken is not the axon-floorplate cell contact but it may be the contact between floor-plate cells that results in the gaps in floor plate morphology (Figure 2). Another consequence of this increased adhesion between growth cones and floor-plate cells is the “curling back” of axons during floor plate crossing (now Figure 9 and Video 5 and Video 6). The enhanced bond between growth cone and floor-plate cell distorts the straight trajectory of axons resulting in the 'corkscrew' phenotype.

Overall, I'm not sure if I understand the concept of a lubricant in this context. The authors state: "Endoglycan is not affecting adhesion between floor-plate cells, as it is not a cell adhesion molecule, or an “anti-adhesion” molecule to be precise. Rather Endoglycan acts as a “lubricant” that is required between axons and floor-plate cells, or between cells that move with respect to each other, because it allows for the necessary dynamics in the adhesion between growth cones and floor-plate cells mediated by adhesion molecules." What is the difference between anti-adhesion and a lubricant? Both reduce friction, it should be the same? And do the authors suggest that whole axons are moving over the floor plate: "In contrast, at the floor plate axons need to move relative to the floor plate."? As far as I'm aware, axons per se do not move much. Growth is mostly achieved at the distal end of the axon. And if a lubricant would allow sliding there, growth cones could not generate forces required for their motility. Furthermore, if adhesion would be reduced, growth cones should also be smaller. Yet, Figure 11 seems to suggest that growth cone sizes are similar in all conditions.

The reviewer refers to the rebuttal letter, where I tried to explain how we envisage the role of Endoglycan. But it looks like the reviewer misunderstood our concept of how Endoglycan acts. What we wanted to point out is that Endoglycan in not an adhesion molecule but also not a repulsive molecule. Rather, our in vivo and in vitro results support a model that Endoglycan is modulating the interaction of a variety of cell adhesion molecules without specifically binding to them. The activity is best compared to a lubricant that allows movement of two parts against each other due to a decrease in friction. In biological terms, this would be a decrease in cell-cell adhesion by preventing tight interaction of cell adhesion molecules. Endoglycan with its extensive O-glycosylation and sialylation is perfectly suited to interfere with molecular interactions between growth cones and floor plate cells or between migrating Purkinje cells and other precursors in the developing cerebellum. We showed that the post-translational modifications, O-glycosylation and sialylation, are both required for the function of Endoglycan.

Thus, while the authors clearly demonstrate, for the first time, the importance of Endoglycan in axon guidance and neuronal migration, I am still not convinced that it acts mostly through regulation of cell-cell adhesion as suggested by the authors.

We hope that the additional in vitro experiments demonstrating clearly a decrease in cell-growth cone adhesion due to the presence of Endoglycan, no matter whether it is provided by the cell carpet or the growth cone, convinces the reviewer of our model.

Reviewer #3:

The study has been significantly extended, in particular with a series of novel experiments performed with live imaging. The resulting data clarify the role of endoglycan and its mode of action. For example, the authors observed that manipulations of endoglycan levels have impact on the velocity of commissural axons and growth cone size. Differences appear tiny but statistically significant. They propose a model whereby endoglycan provided by FP cells and axons acts as a lubricant to facilitate the FP navigation and the axon-FP cell contacts. This is an interesting idea that beyond endoglycan, shed light on yet poorly understood roles of modulators.

In addition, the authors addressed several concerns that were raised by their previous version of the work. They checked whether manipulations of endoglycan alter the expression of molecules known to guide commissural axons, which as far as we can tell from general expression patterns is not the case.

I also think that the part on the cerebellum model is now more integrated to the rest of the work, and I agree that it enable to make a broader message on the functional properties of endoglycan.

We thank the reviewer for this positive assessment of our revised version of the study.

I still consider that, for better consistency, the experiments to demonstrate the regulation of adhesion by endoglycan should have been carried out on the population of neurons that is investigated here (commissural neurons) and not on motoneurons. That said, I believe that what the authors report with motoneurons might apply to commissural neurons and the findings are consistent with their model on commissural axon navigation. The authors modified their initial analysis, now concentrating on adhesion rather than outgrowth, which is to me very pertinent. They also carried out some biochemical analysis to document endoglycan mode of action, showing contribution of glycosylation.

As suggested, we have repeated the adhesion assay originally done with motoneurons with commissural neurons (now Figure 4). In addition, we have added a series of “growth cone blasting” experiments where we either cultures commissural neurons on HEK cells expressing Endoglycan or overexpressed Endoglycan in commissural neurons. These experiments demonstrate that Endoglycan reduces growth cone adhesion no matter where it comes from. Growth cones were detached faster when they were targeted by a stream of PBS delivered from a micropipette (new Figure 5).

Overall, I found that the study is improved by the novel data.

Here are some specific comments:

For all figures related to the time-lapse imaging.: the number of experiments, embryos etc… are not indicated.

The number of replicates and/or embryos are given in Table 2 and we have added the values to the figure legend (Figure 8).

It seems to me that there are some redundancy between Figure 6 and Figure 7 in the questions that the experiments address.

Yes, Figure 6 and Figure 7 (now Figure 7 and Figure 8) are in fact the same series of experiments. Figure 8 is the quantification of the observations explained in Figure 7. We chose to divide the imaging examples and the quantification into two figures due to size.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

Your study highlights Endoglycan in regulating neuronal growth and migration, pointing to a mechanism involving anti-adhesion. The reviewers felt that you have addressed all the major criticisms, and with the addition of your apt in vitro experiments addressing cell adhesion (new Figure 5), that the manuscript is generally improved. They comment that this will be "a very nice paper that significantly contributes to the field".

There are still revisions called for, as listed below, primarily altering the wording used to describe the action of endoglycan. Moreover, the reviewers encourage you to consider removing the data on the cerebellum.

I) Cerebellar data: Two of the three reviewers urge you to consider removing the cerebellar data on Purkinje cells, as this analysis distracts from the now well fleshed-out message relating to axon guidance. The effects of Endoglycan knockdown on Purkinje cell migration are compelling per se, but the interpretation of these defects is complicated by the overall disrupted morphology of the cerebellum and lack of data that directly demonstrate whether and how Endoglycan modulates Purkinje cell adhesion to other relevant cell types. In your rebuttal, you indicate that defects in granule cell proliferation are consistent with impaired Purkinje cell migration. This is certainly true, but numerous alternative interpretations exist, and no attempt is made to probe the authors' model further. Without extensive experiments focused on molecular mechanisms underlying the phenotype, these findings do not integrate well with the rest of the paper. You might consider publishing the cerebellar work as a smaller study. One question is whether cell migration/translocation and axon outgrowth rely on adhesion/anti-adhesion in the cerebellum as they do in the floor plate.

We have decided to remove the part on the cerebellum from our paper. Our idea was to demonstrate that Endoglycan plays a general role in adhesion regulation in the CNS. But this clearly made the paper very long and added a second focus. Therefore, we removed Figure 10, Figure 11, Figure 12 of the original manuscript and the paragraphs from the Result section reporting our observations in the cerebellum. We rephrased parts of the Discussion accordingly.

II) Endoglycan at the floor plate:

1) “Lubricant"? vs negatively regulating cell adhesion:

a) While you now provide an in vitro experiment suggesting a role of endoglycans in regulating cell adhesion, there is no experimental evidence for a role of endoglycans as a lubricant in vivo (which would certainly be very difficult to achieve, if at all possible). This should be kept in mind when interpreting the data. You are thus encouraged to tone down the wording referring to the proposed mechanism throughout the manuscript, for example in statements such as "Endoglycan acts as “lubricant” for growth cone movement in the floor plate".

We have followed the suggestion and removed the work lubricant from the result section. We only use the word “lubricant” once in the Discussion to provide a metaphor for the anti-adhesive role of Endoglycan that facilitates motility of growth cones in the floor plate. We also toned down the wording throughout the text to be more cautious with our conclusions.

The reviewers caution that our evidence that Endoglycan lowers adhesive strength comes from in vitro experiments and that it would be very difficult or impossible to get similar data in vivo. This is certainly true. However, we would like to point out that supporting evidence that lower adhesive forces contribute to the observed phenotypes also comes from our ex vivo observations with live imaging. The Video 5 and Video 6 (like all other videos) show growth cones and axons in intact spinal cords. The videos and the frames shown in Figure 12 present axons stuck to floor-plate cells for some time as a possible explanation for the tortuous path taken by axons. We are aware that this ex vivo preparation is still not in vivo, but the axons are growing in their intact environment. We cannot measure adhesion in this situation but the observation of axons crossing floor plates in control spinal cords compared to spinal cords lacking Endoglycan suggests that cell-cell contacts are clearly different. Nonetheless, we have toned down our conclusions about the proposed mechanism of Endoglycan function, as suggested.

b) The second part of the title is not backed up by the data ("by negatively regulating cell-cell adhesion"); the abstract should explicitly state that adhesion is affected in vitro.

We have changed the title and rephrased the Abstract according to the suggestions by the reviewers.

c) Statements such as "Thus, the aberrant morphology of the floor plate at HH25 is explained by the inability of axons to break contacts with floor-plate cells in the absence of Endoglycan" should be phrased more carefully. Replacing "is explained" by "could be explained" would not make the study weaker but rather protect you in the future should it turn out that there is a different mechanism at play. Finally, because of the potential problems mentioned above associated with the blast experiment, you might omit the word "strong" in the sentence "Strong support for this hypothesis was contributed by in vitro findings that the adhesive strength… ".

We have rephrased the respective passages as suggested (Materials and methods).

d) It is admittedly difficult to distinguish a lubricant effect from a system that would control the strength of an adhesion or anti-adhesion effect (for example by controlling the number of binding partners or their availability). If the idea is that endoglycan does not target a specific adhesion mechanism but rather has a general effect, would you consider the role of endoglycan to be similar to the one attributed to PSA of NCAM adhesion molecules? This should be discussed.

We have rephrased our paragraph comparing the roles of PSA-NCAM and Endoglycan to strengthen the similarity between the mechanisms (Materials and methods).

2) Blasting assay: The growth cone blasting assay introduced in the revised version of the manuscript is a very difficult experiment, and your efforts should be praised. You have carefully considered some of the confounding factors; however, some open questions remain.

a) Factors such as the distance of the pipette from the substrate (i.e., from the top surface of the HEK cells) and the area and height of the growth cones will also be crucial determinants of the time required to detach growth cones. How was the height of the pipette tip above the cells controlled, and how accurately could it be determined? Was the growth cone area the same between the groups? Where the experiments done blindly?

We have added a detailed description of the technical details that were crucial for the growth cone blasting experiments. The experiments were done by two people (A.D. and B.K.). The position of the pipette tip was carefully controlled by the motorized microscope stage taking the focal plane of the growth cone as basis. The horizontal distance was measured between the tip of the pipette and the edge of the targeted growth cone before the pump was started. The time it took to detach the growth cone from the HEK cell carpet was taken by a person blind to the experimental condition. To avoid influences on the time measurements from drifts in temperature or flow rate due to changes in the oxygen content in the PBS, one person was not blind to the experimental condition and switched between experimental conditions on purpose. We added a detailed description to the Materials and method section. We also added measurements of growth cone areas between the groups as new Figure 7, including source data. The analysis clearly demonstrated that despite some variability in growth cone size within each group. There was no significant difference between the different groups.

b) Furthermore, why was PBS used, which usually does not contain any calcium? Calcium has a key role in cell adhesion.

Details of these aspects should be provided (including images of growth cones at higher magnification and, if possible, a quantification of growth cone areas), and the experiment should be critically discussed.

To get reproducible results, it was extremely important to degas the solution used to blast the growth cones. Culture medium could not have been degassed without changes in pH. Because the same solution was used for both control and experimental conditions, and because the absolute value was not of importance, we belief that the local decrease in Ca2+ does not matter. Importantly, the volumes added to the culture dish were very small compared to the volume of the medium. We have added a quantification of growth cone areas (Results and Figure 7, including source data showing growth cones at higher magnification along with measurements of growth cone areas).

3) A more detailed description of what constitutes an abnormal injection site in Figure 3 would greatly help with interpretation of those data. Quantification of "normal injection sites" in Figure 3 somewhat unsatisfying, especially for the comparison of gain-of-function and loss-of-function effects. After all, more adhesion and less adhesion should not produce the exact same phenotypes. Could the abnormal injection sites be further categorized, e.g. as stalled, 'corkscrew', premature turning, aberrant postcrossing A-P guidance etc.?

We have added an explanation about the phenotypes to the quantification in Table 1. Because the “failure to turn into the longitudinal axis” and the “stalling in the floor plate” are not independent of each other (strong stalling in the floor plate of close to all fibers will prevent the analysis of the “no turn” phenotype) we can only quantitatively compare the DiI injection sites with normal trajectories. Nonetheless, we have exchanged Table 1. The new version includes the average percentage of DiI injection sites with the “FP stalling” and the “no turn” phenotypes.

4) You have now used a carpet of HEK cells rather than just a cell adhesion molecule coated on plastic, which you argue "probably also helped to avoid artefacts". It is not clear how a homogeneous, controlled coating of a plastic surface with purified proteins should cause more artifacts than cells expressing a multitude of different proteins at their surface which may vary in space and time. While it is very difficult to see any details in Figure 4, HEK cells in the two groups (Figure 4A) seem to have different morphologies. If this is true, changes in HEK cells could lead to changes in cell-cell contacts to neurons that are independent of Endoglycans. This should be discussed.

We respectfully disagree. Axons are never growing on hard surfaces in vivo. They are adhering to cells in a 3D context. To have just one molecule offered together with tissue culture plastic is a very unphysiological environment. The HEK cells are inherently diverse in morphology and morphology might be different when one compares very confluent versus less confluent areas. However, there was no difference in morphology between the different conditions. The reviewers probably are maybe referring to the few round green cells in Figure 5A? These might be either dying cells or cells in the process of division. They have accumulated more background fluorescence. To show that there is a cell carpet in Figure 5A, the image was taken with suboptimal conditions.

5) In the Results section, the interpretation of defects as a problem with adhesion is introduced very early. It would be better to draw that conclusion later, after the experiments in Figure 4 and Figure 5. In general, some of the conclusions in the Results section that combine outcomes from multiple experiments might be better suited for inclusion in the Discussion.

We have rephrased this paragraph to make it sound like a model to be tested with further experiments rather than a conclusion.

https://doi.org/10.7554/eLife.64767.sa2

Article and author information

Author details

  1. Thomas Baeriswyl

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Formal analysis, Investigation, Methodology, Writing - original draft
    Contributed equally with
    Alexandre Dumoulin, Martina Schaettin and Georgia Tsapara
    Competing interests
    No competing interests declared
  2. Alexandre Dumoulin

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Thomas Baeriswyl, Martina Schaettin and Georgia Tsapara
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2420-6877
  3. Martina Schaettin

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Formal analysis, Investigation, Methodology, Writing - review and editing
    Contributed equally with
    Thomas Baeriswyl, Alexandre Dumoulin and Georgia Tsapara
    Competing interests
    No competing interests declared
  4. Georgia Tsapara

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Formal analysis, Investigation, Writing - original draft
    Contributed equally with
    Thomas Baeriswyl, Alexandre Dumoulin and Martina Schaettin
    Competing interests
    No competing interests declared
  5. Vera Niederkofler

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Formal analysis, Investigation, Writing - original draft
    Competing interests
    No competing interests declared
  6. Denise Helbling

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  7. Evelyn Avilés

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Resources
    Competing interests
    No competing interests declared
  8. Jeannine A Frei

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Resources, Investigation
    Competing interests
    No competing interests declared
  9. Nicole H Wilson

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Resources, Supervision
    Competing interests
    No competing interests declared
  10. Matthias Gesemann

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Data curation, Investigation, Visualization
    Competing interests
    No competing interests declared
  11. Beat Kunz

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  12. Esther T Stoeckli

    Department of Molecular Life Sciences and Neuroscience Center Zurich, University of Zurich, Zurich, Switzerland
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Investigation, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    esther.stoeckli@mls.uzh.ch
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8485-0648

Funding

Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung

  • Esther T Stoeckli

Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (Brain Plasticy and Repair)

  • Esther T Stoeckli

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Tiziana Flego for excellent technical assistance, Alexandra Moniz and John Darby Cole for help with glycosylation dependence experiment, and members of the lab for discussions and critical reading of the manuscript. This project was supported by the Swiss National Science Foundation and the NCCR Brain Plasticity and Repair (Center of Transgenesis Expertise).

Senior Editor

  1. Didier YR Stainier, Max Planck Institute for Heart and Lung Research, Germany

Reviewing Editor

  1. Carol A Mason, Columbia University, United States

Reviewer

  1. Kristian Franze, University of Cambridge, United Kingdom

Publication history

  1. Received: November 10, 2020
  2. Accepted: February 25, 2021
  3. Accepted Manuscript published: March 2, 2021 (version 1)
  4. Version of Record published: March 10, 2021 (version 2)

Copyright

© 2021, Baeriswyl et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 618
    Page views
  • 101
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    Piotr Michaluk et al.
    Research Article

    Glutamate uptake by astroglial transporters confines excitatory transmission to the synaptic cleft. The efficiency of this mechanism depends on the transporter dynamics in the astrocyte membrane, which remains poorly understood. Here, we visualise the main glial glutamate transporter GLT1 by generating its pH-sensitive fluorescent analogue, GLT1-SEP. FRAP-based imaging shows that 70-75% of GLT1-SEP dwell on the surface of rat brain astroglia, recycling with a lifetime of ~22 s. Genetic deletion of the C-terminus accelerates GLT1-SEP membrane turnover while disrupting its surface pattern, as revealed by single-molecule localisation microscopy. Excitatory activity boosts surface mobility of GLT1-SEP, involving its C-terminus, metabotropic glutamate receptors, intracellular Ca2+ and calcineurin-phosphatase activity, but not the broad-range kinase activity. The results suggest that membrane turnover, rather than lateral diffusion, is the main 'redeployment' route for the immobile fraction (20-30%) of surface-expressed GLT1. This finding reveals an important mechanism helping to control extrasynaptic escape of glutamate.

    1. Neuroscience
    Eric Torsten Reifenstein et al.
    Research Article

    Remembering the temporal order of a sequence of events is a task easily performed by humans in everyday life, but the underlying neuronal mechanisms are unclear. This problem is particularly intriguing as human behavior often proceeds on a time scale of seconds, which is in stark contrast to the much faster millisecond time-scale of neuronal processing in our brains. One long-held hypothesis in sequence learning suggests that a particular temporal fine-structure of neuronal activity - termed 'phase precession' - enables the compression of slow behavioral sequences down to the fast time scale of the induction of synaptic plasticity. Using mathematical analysis and computer simulations, we find that - for short enough synaptic learning windows - phase precession can improve temporal-order learning tremendously and that the asymmetric part of the synaptic learning window is essential for temporal-order learning. To test these predictions, we suggest experiments that selectively alter phase precession or the learning window and evaluate memory of temporal order.