Directional migration of neurons and neuronal precursor cells is a central process in nervous system development. In the nematode Caenorhabditis elegans, the two Q neuroblasts polarize and migrate in opposite directions along the anteroposterior body axis. Several key regulators of Q cell polarization have been identified, including MIG-21, DPY-19/DPY19L1, the netrin receptor UNC-40/DCC, the Fat-like cadherin CDH-4 and CDH-3/Fat, which we describe in this study. How these different transmembrane proteins act together to direct Q neuroblast polarization and migration is still largely unknown. Here, we demonstrate that MIG-21 and DPY-19, CDH-3 and CDH-4, and UNC-40 define three distinct pathways that have partially redundant roles in protrusion formation, but also separate functions in regulating protrusion direction. Moreover, we show that the MIG-21, DPY-19 and Fat-like cadherin pathways control the localization and clustering of UNC-40 at the leading edge of the polarizing Q neuroblast, and that this is independent of the UNC-40 ligands UNC-6/netrin and MADD-4. Our results provide insight into a novel mechanism for ligand-independent localization of UNC-40 that directs the activity of UNC-40 along the anteroposterior axis.
Cell migration plays a central role in the development of the nervous system. For example, migration of neurons and neuronal precursors is required to form the different cortical layers of the vertebrate brain, and neural crest cells migrate throughout the embryo to generate the peripheral nervous system. The Q neuroblasts of the nematode Caenorhabditis elegans provide a powerful model system for studying neuronal cell migration at single cell resolution in vivo (Middelkoop and Korswagen, 2014). The two Q neuroblasts are generated at the end of embryogenesis as sister cells of the hypodermal seam cell V5 (Sulston and Horvitz, 1977). They are initially located at similar positions (between seam cells V4 and V5) on the left (QL) and right (QR) lateral sides of the animal, but during the first few hours of larval development, the two neuroblasts polarize and migrate in opposite directions along the anteroposterior body axis (Fig. S1A). QL extends a protrusion towards the posterior and migrates to a position dorsal to seam cell V5, whereas QR polarizes in the opposite direction and migrates over seam cell V4. Interestingly, this asymmetric polarization and migration is linked to the subsequent response of the Q neuroblast to Wnt ligands (Honigberg and Kenyon, 2000; Middelkoop et al., 2012; Whangbo and Kenyon, 1999). Thus, the polarization and migration of QL towards the posterior initiates canonical, β-catenin-dependent Wnt signaling and expression of the target gene mab-5. Expression of this Hox gene in turn induces the QL descendants to either arrest their migration (QL.p) or to continue towards a position in the tail (QL.a) (Salser and Kenyon, 1992) (Fig. S1B). The anterior polarization and migration of QR, on the other hand, leads to a non-canonical Wnt signaling response that mediates the long-range migration of the QR descendants to positions in the anterior body region (Mentink et al., 2014; Zinovyeva et al., 2008).
Several genes have been identified that disrupt the asymmetric polarization and migration of the Q neuroblasts. Among these are five genes – mig-21, dpy-19, ptp-3, cdh-4 and unc-40 – that encode transmembrane proteins (Honigberg and Kenyon, 2000; Middelkoop et al., 2012; Schmitz et al., 2008; Sundararajan and Lundquist, 2012; Sundararajan et al., 2014). MIG-21 has no close orthologs in other species, but shares similarity with the netrin co-receptor UNC-5 and semaphorin 5A through its two extracellular thrombospondin repeats. DPY-19 is an evolutionarily conserved multi-pass transmembrane protein, and one of its mammalian orthologs, DPY19L1, is required for the radial migration of glutaminergic neurons in the developing cerebral cortex (Watanabe et al., 2011). Interestingly, it was shown that DPY-19 functions as a c-mannosyltransferase that glycosylates the thrombospondin repeats of UNC-5 and MIG-21 (Buettner et al., 2013). ptp-3 encodes a LAR-type receptor protein tyrosine phosphatase (RPTP) (Ackley et al., 2005; Harrington et al., 2002). LAR-RPTPs contain extracellular fibronectin type III repeats and immunoglobulin domains, and two intracellular phosphatase domains. They interact with components of the extracellular matrix and play an important role in axon guidance, synapse development and cell migration (Ackley et al., 2005; Harrington et al., 2002; Um and Ko, 2013). cdh-4 encodes one of the two Fat-like cadherin orthologs of C. elegans (Pettitt, 2005). Fat-like cadherins function in the planar polarization of epithelial tissues (Matis and Axelrod, 2013), in tissue size regulation (Yin and Pan, 2007) and in cell migration (Horne-Badovinac, 2017; Schmitz et al., 2008; Sundararajan et al., 2014). In mammalian cells, Fat1 localizes at the leading edge of migrating cells, and is required for directional polarization and protrusive activity at the lamellipodium (Moeller et al., 2004; Tanoue and Takeichi, 2004). unc-40 encodes an ortholog of the netrin receptor DCC (Chan et al., 1996). UNC-40 and its ligand UNC-6/netrin control the migration of axonal growth cones along the dorsoventral body axis (Wadsworth, 2002). Moreover, UNC-40 is required for the ventral polarization and invasion of the anchor cell during vulva development (Hagedorn et al., 2013; Wang et al., 2014). In addition to netrin-dependent signaling, also ligand-independent functions of UNC-40 have been described. Thus, UNC-40 can interact with the Robo ortholog SAX-3 to potentiate the response of axonal growth cones to the SLT-1/Slt guidance cue (Yu et al., 2002) and signaling of the Wnt ligand EGL-20 through the Frizzled MIG-1 and the Van Gogh ortholog VANG-1 has been shown to influence the localization of UNC-40 in the developing HSN neuron (Tang and Wadsworth, 2014). Interestingly, UNC-40 may also function independently of netrin in the Q neuroblast lineage, as polarization and migration are not affected in unc-6/netrin mutants (Honigberg and Kenyon, 2000).
Genetic epistasis experiments have shown that mig-21, dpy-19, ptp-3/Lar and cdh-4/Fat have overlapping functions in the polarization and migration of the Q neuroblasts, while unc-40/DCC functions in a separate pathway (Middelkoop et al., 2012; Sundararajan and Lundquist, 2012; Sundararajan et al., 2014). Moreover, it has been found that there are distinct left-right asymmetric differences in the functional interactions between these regulators. Thus, unc-40 acts in parallel to mig-21 and ptp-3 to mediate posterior migration of QL, while these two pathways may mutually inhibit each other to direct the protrusion of QR towards the anterior (Sundararajan and Lundquist, 2012). Furthermore, cdh-4 has been found to function together with mig-21 and ptp-3 in QL, but in both the unc-40 and mig-21 - ptp-3 pathways in QR (Sundararajan et al., 2014). Apart from these genetic interactions, the underlying mechanism that mediates the robust left-right asymmetric polarization of the Q neuroblasts is still largely unknown.
Here, we show that a second Fat-like cadherin, CDH-3, is also required for the correct polarization of the Q neuroblasts. To further define the functional interaction between cdh-3, cdh-4, mig-21, dpy-19 and unc-40, we performed a detailed quantitative analysis of protrusion formation and directional outgrowth in null or strong loss-of-function mutants. Our results show that mig-21, dpy-19, cdh-3, cdh-4 and unc-40 define three distinct pathways. We found that the unc-40 pathway plays a prominent role in the formation of a major, lamellipodium-like protrusion, while the mig-21-dpy-19 pathway and the Fat-like cadherins have additional functions in specifying the direction of the growing protrusion. Moreover, we show that cdh-4 has cell autonomous as well as non-autonomous functions in protrusion formation and directional outgrowth. Mechanistically, we show that the mig-21-dpy-19 and cdh-3-cdh-4 pathways are required for the ligand-independent localization of UNC-40/DCC at the leading edge of the Q neuroblast protrusion. We conclude that a complex interplay between the three pathways, which acts in part through the localization of UNC-40 at the protrusive front, mediates the left-right asymmetric polarization of the Q neuroblasts.
Quantitative analysis of Q neuroblast polarization and migration reveals distinct functions of UNC-40/DCC, Fat-like cadherins and the transmembrane proteins MIG-21 and DPY-19 in protrusion formation and direction
The netrin receptor UNC-40/DCC, the transmembrane proteins MIG-21 and DPY-19, and the Fat-like cadherin CDH-4 are required for the left-right asymmetric polarization and migration of the QL and QR neuroblasts (Middelkoop et al., 2012; Schmitz et al., 2008; Sundararajan and Lundquist, 2012; Sundararajan et al., 2014). In addition, we found that a second Fat-like cadherin, CDH-3, is involved in this process, as demonstrated by the effect of cdh-3 mutation on Q neuroblast polarization and the subsequent migration of the Q cell descendants (Fig. 1B, Fig. S1F).
To further examine the function of these transmembrane proteins in polarization and migration, we quantified the effects of strong loss-of-function or null mutations. The first step in Q neuroblast migration is the formation of a single lamellipodium-like protrusion that extends towards the anterior in QR and the posterior in QL. To visualize the Q neuroblasts, we used a transgenic marker that labels the plasma membrane (GFP-PH) and the nucleus (H2B-GFP) of the Q neuroblasts and the lateral hypodermal seam cells (Wildwater et al., 2011) (Fig. 1A, Fig. S1C). We measured the ability of the Q cells to form a major lamellipodium-like protrusion, and the direction of this protrusion with respect to the anteroposterior body axis (as defined by a line between the centers of the nuclei of the Q neuroblast and the seam cell V5) (Fig. 1A) (Middelkoop et al., 2012). In wild-type animals, QL extends a protrusion towards the dorsal side of V5 (with a predominant angle of 20-40°), while QR sends a protrusion over V4 (with an angle between 140 and 160°) (Fig. 1B). During this phase, filopodia-like membrane extensions are formed at the leading edge of the protrusion (Movies 1 and 2). In mig-21(u787) and dpy-19(e1314) mutants, the direction of the QL and QR protrusions was altered. In agreement with previous studies (Middelkoop et al., 2012), we found that 40-50% of QL and QR cells formed a short broadly shaped protrusion towards the dorsal side (60-120°) (Fig. 1B, Fig. S1C,D, see Table S1 for statistical analysis), as well as smaller spike-like filopodia in other directions. In 11-16% of animals, the QL and QR neuroblasts failed to form a major protrusion. As discussed below, this phenotype was enhanced in the double mutant, indicating that mig-21 and dpy-19 have an additional partially redundant role in protrusion formation. Loss of cdh-4 also affected protrusion formation [20-23% of QL and QR neuroblasts did not form a lamellipodium-like protrusion in cdh-4(hd40) mutants] (Fig. 1B, Fig. S1C). However, the effect on protrusion direction was different from mig-21 and dpy-19. Although there was a relatively modest effect on protrusion direction in QL, there was a striking reversal of direction in QR, with 36% of the cells polarizing towards the posterior instead of the anterior. Time-lapse imaging showed increased formation of small ectopic protrusions (Movie 3) and we observed that the major protrusion can collapse and reform in a different direction (Movie 4). In cdh-3(pk87) mutants, there was a mild defect in polarization and protrusion direction, but not the reversal in QR polarity that we observed in the cdh-4 mutant (Fig. 1B, Fig. S1C). However, cdh-3 mutants did show ectopic protrusive activity and instability of the major protrusion (Movie 5). unc-40(e271) mutants showed the highest percentage (35-37%) of unpolarized cells (Fig. 1B). We found that these unpolarized cells do show protrusive activity, but that these protrusions are unstable and unable to grow into a lamellipodium-like protrusion (Movie 6). In cells that did form a major protrusion, the average length of the protrusion was reduced (Fig. S2B), while the direction of the protrusion was mostly correct (Fig. 1B, Fig. S1C,D).
Once the protrusion has fully extended, the cell body follows; QL divides at a position dorsal to V5, and QR at a position dorsal to V4 (Fig. 2A). In agreement with previous studies (Middelkoop et al., 2012; Sundararajan and Lundquist, 2012; Sundararajan et al., 2014), we found that all the mutants showed defects in the migration of the QL and QR cell bodies (Fig. 2B,C). However, the reduction in migration distance was more severe in mig-21, dpy-19 and cdh-4 mutants than in the unc-40 and cdh-3 mutants, which is consistent with the effect of these mutations on protrusion formation and directional outgrowth. Finally, we examined the position of the Q.p descendants (Q.paa and Q.pap, abbreviated as Q.pax) (Fig. S1B), which is linked to initial polarization and migration through the activation of the Wnt-dependent anterior or posterior migration pathways (Honigberg and Kenyon, 2000; Middelkoop et al., 2012). Again, there was a strong correlation between the severity of defects in initial polarization and the subsequent left-right asymmetric migration of the Q.pax (Fig. S1F).
Based on these results, we conclude that unc-40 is predominantly required for the formation of a major lamellipodium-like protrusion, while mig-21, dpy-19 and the Fat-like cadherin cdh-4 – and to a lesser extent cdh-3 – have additional functions in controlling the directional outgrowth of the protrusion. As part of this guidance mechanism, mig-21 and dpy-19 are required for the asymmetric polarization of QL and QR along the anteroposterior body axis, while cdh-4 is particularly important for the anterior polarization of QR. Finally, as all four mutants showed ectopic protrusive activity, the coordinated function of these different transmembrane proteins is required for the formation of a single major protrusion.
UNC-40/DCC, MIG-21, DPY-19 and CDH-3 act cell-autonomously, while the Fat-like cadherin CDH-4 has both cell-autonomous and non-autonomous roles in Q neuroblast polarization
We and others have previously shown that mig-21 is expressed in the Q neuroblasts and that it functions cell-autonomously in Q cell polarization and migration (Middelkoop et al., 2012; Sundararajan and Lundquist, 2012). To examine the expression of dpy-19, we used CRISPR/Cas9-mediated genome editing to endogenously tag the dpy-19 gene with gfp (Dickinson et al., 2013). We found that the resulting DPY-19::GFP fusion protein is functional (data not shown) and that it is expressed in the QL and QR neuroblasts, the seam cells, and the hypodermal syncytium hyp7 (Fig. 3A). The endogenous DPY-19 protein localizes to the nuclear periphery, which is in agreement with earlier studies using DPY-19 overexpression (Honigberg and Kenyon, 2000). To investigate whether DPY-19 is required in the Q cells or the surrounding hypodermal cells, we used the promotor of egl-17 (Branda and Stern, 2000; Middelkoop et al., 2012) to specifically express dpy-19 in the Q cell lineage of dpy-19(e1314) mutants. Using the final position of the Q.pax cells as a proxy, we found that Q cell-specific expression of DPY-19 was sufficient for correct Q neuroblast polarization and migration (Fig. S2D), which was further confirmed using transgenic mosaics (Table S4). We conclude that MIG-21 and DPY-19 both function cell autonomously in the Q neuroblasts, which is consistent with the shared molecular function of the two proteins (Buettner et al., 2013).
Studies with transgenic reporters have shown that unc-40 is expressed in the Q neuroblast lineage (Honigberg and Kenyon, 2000). To examine the endogenous expression of unc-40 during Q cell polarization and migration, we used single-molecule mRNA fluorescent in situ hybridization (smFISH) (Raj et al., 2008) to quantitatively measure unc-40 mRNA abundance. At the early L1 larval stage, unc-40 expression was detected in the nerve ring and the Q neuroblasts (Fig. 3B, Fig. S2A). We found that the number of unc-40 mRNA spots was similar in QL and QR, with a positive correlation between spot abundance and migration distance (QL, r=0.4081, P<0.0022; QR, r=−0.2838, P<0.0324) (Fig. 3E), indicating that unc-40 is symmetrically expressed in the two lineages and that expression increases during the polarization and migration process. Earlier studies using Q cell-specific rescue and tissue-specific RNAi-mediated knock-down have shown that unc-40 functions cell autonomously in the Q cell lineage (Sundararajan and Lundquist, 2012). To corroborate these findings and to generate a tool for robust Q cell-specific depletion of UNC-40, we used CRISPR/Cas9-mediated genome editing to insert a ZF1 tag at the C terminus of UNC-40. This sequence is recognized by ZIF-1, the substrate-binding subunit of a ubiquitin-ligase complex, which targets the tagged protein for proteasomal degradation (Armenti et al., 2014). We found that ubiquitous expression of ZIF-1 induced an uncoordinated phenotype that was similar to that observed in the unc-40 null mutant (data not shown), indicating that the tagged UNC-40 protein is efficiently degraded. Next, we used the egl-17 promotor to specifically express ZIF-1 in the Q neuroblast lineage. We found that the effect of UNC-40 depletion on protrusion formation, polarization and the migration of the Q cell and its descendants was similar to the unc-40 null mutant (Fig. 3H, Fig. S3A-C), confirming the cell-autonomous function of UNC-40 in this process.
We also used smFISH to examine the expression of cdh-4 and cdh-3. Consistent with transgenic reporter studies (Sundararajan et al., 2014), we observed cdh-4 mRNA spots in the Q neuroblasts, as well as a wide range of other cells, including the neighboring seam and P cells, ventral nerve cord neurons, and cells in the head region (Fig. 3C, Fig. S2C). In contrast, we found that during the initial polarization and migration phase, cdh-3 was specifically expressed in the Q neuroblasts, with additional expression in only a few unidentified cells in the head (Fig. 3D, Fig. S2B). Similar to unc-40, there was no significant difference in expression between QL and QR. Moreover, for both cdh-3 and cdh-4, the expression level correlated with migration distance (cdh-3 QL, Spearman r=0.5765, P<0.0001, QR, Spearman r=−0.4838, P<0.0004; cdh-4 QL, Spearman r=0.3176, P<0.03, QR, Spearman r=−0.5070, P<0.0004), indicating that expression increases during polarization and migration (Fig. 3F,G). To examine the site of action of cdh-3 and cdh-4, we used CRISPR/Cas9-mediated genome editing to insert a ZF1 tag at the C terminus and depleted the endogenously tagged proteins using Q lineage-specific expression of ZIF-1. Depletion of CDH-3 closely phenocopied the mutant phenotype (Fig. 3H, Fig. S3A-C). A similar effect on protrusion formation was also observed when CDH-4 was depleted from the Q neuroblast lineage [an average of 22 versus 24% of unpolarized cells in cdh-4(hd40) and CDH-4 depletion, respectively], but the reversal of QR polarity observed in the cdh-4 null mutant was strikingly absent (Fig. 3H).
We conclude that MIG-21, DPY-19, UNC-40/DCC and the Fat-like cadherin CDH-3 function cell autonomously in protrusion formation and guidance. CDH-4/Fat, on the other hand, has a dual role, with a cell-autonomous function in protrusion formation and a cell non-autonomous function in directing the protrusion of QR towards the anterior.
MIG-21, DPY-19 and the Fat-like cadherins function in parallel to UNC-40/DCC
To examine genetic interactions between the different polarity regulators, we generated compound mutants. For mig-21 and dpy-19, we found that the phenotype of the double mutant was similar to that of the single mutants (Fig. 1C), with a clear dorsal bias in protrusion direction. However, the percentage of unpolarized cells was increased in the double mutant. These results indicate that mig-21 and dpy-19 function together in regulating protrusion direction, which is in agreement with previous genetic studies (Middelkoop et al., 2012) and the observation that MIG-21 is a substrate of DPY-19 c-mannosyltransferase activity (Buettner et al., 2013). However, our results suggest that there is also a minor parallel activity in protrusion formation.
Double mutants between cdh-3 and cdh-4 were not viable (data not shown), but we found that homozygous offspring of cdh-4(hd40) cdh-3(pk87) heterozygotes develop to the L1 stage. Using the final position of the Q.pax as a proxy for Q polarization and migration, we found that the phenotype of the maternally rescued double mutant was indistinguishable from the cdh-4 single mutant (Fig. S4C). These results indicate that cdh-3 and cdh-4 function in a common pathway, in which cdh-4 has a predominant role.
Next, we combined mutations in the cdh-3-cdh-4/Fat and mig-21-dpy-19 pathways with unc-40/DCC. As cdh-4 and unc-40 double mutants are not viable (Sundararajan et al., 2014), we used Q lineage-specific expression of ZIF-1 to deplete endogenously tagged UNC-40::ZF1 in a cdh-4(hd40) mutant background (Fig. 1C). In agreement with a previous study using tissue-specific RNAi (Sundararajan et al., 2014), the combined loss of cdh-4 and UNC-40 resulted in a strongly enhanced phenotype. Quantification showed that 63% of QL and 71% of QR neuroblasts remained unpolarized. Moreover, the cells that did form a protrusion no longer showed the posterior bias observed in the cdh-4 single mutant. Of note, 6% of QL and QR cells polarized at an angle ventral to the anteroposterior body axis, a phenotype that was not observed in the single mutants. As the triple mutant of mig-21, dpy-19 and unc-40 was also not viable (data not shown), we used a similar approach to deplete UNC-40 from mig-21(u787) dpy-19(e1314) double mutants. Again confirming earlier RNAi experiments (Sundararajan and Lundquist, 2012), this resulted in a strongly enhanced phenotype. We found that there was a strong defect in protrusion formation: 79-80% of Q cells remained unpolarized (Fig. 1C), leading to a strong reduction in initial migration (Fig. S4A), misorientation of the position Q.a relative to Q.p (Fig. S4D) and almost complete anterior migration of the QR.pax (Fig. S4B). Furthermore, of the ∼20% of cells that did form a major protrusion, the predominant polarization direction was shifted from dorsal in the mig-21 dpy-19 double mutant to anterior in the triple mutant (Fig. 1C). Consistent with these static measurements, time-lapse imaging revealed that the Q neuroblasts either show limited protrusive activity (Movie 7) or form a larger protrusion towards the anterior (Movie 8). However, even when a major lamellipodium was formed, we did not observe migration of the cell body.
Taken together, these results show that mig-21 and dpy-19, and the Fat-like cadherins cdh-3 and cdh-4, are part of linear genetic pathways that act in parallel to unc-40/DCC. Loss of either of these pathways strongly enhances the polarization defect of unc-40. However, in Q cells that still form a major protrusion, loss of cdh-4 and unc-40 leads to random polarization, whereas in the absence of mig-21, dpy-19 and unc-40, the Q cells predominantly polarize towards the anterior.
The asymmetric polarization of QL and QR is independent of centrosome position
The position of the centrosome, and the axis that it forms with the nucleus, is a key polarity determinant in migrating cells (Luxton and Gundersen, 2011). In many migrating cell types, the centrosome is positioned anterior to the nucleus. Here, it forms a microtubule-organizing center that mediates polarized vesicular transport to the leading edge and is important for nuclear movement. Using the centrosomal marker CMD-1::GFP (Chai et al., 2012), we have found that in both QL and QR, the centrosome is localized at the anterior side of the nucleus (Fig. 4A,B). A consequence of this symmetric localization is that the centrosome is positioned on the same side as the protrusion in QR, while it is at the opposite side of the nucleus in QL. Measurements of centrosome position – quantified as the angle of the centrosome and the Q cell nucleus with respect to the anteroposterior body axis – revealed that while the centrosome is anterior to the nucleus in both QL and QR, the angle of centrosome position is more variable in QL (Fig. 4C-E). We speculate that this variability is related to the posterior polarization of QL, which may induce microtubule pulling forces that influence the position of the centrosome at the opposite side of the nucleus. The average position of the centrosome was not altered in mig-21, dpy-19 or unc-40 mutants, but variability was increased in QR. Moreover, there was no correlation between centrosome position and protrusion direction in the mutants (Fig. 4F,G). We conclude that the left-right asymmetric polarization of the Q neuroblasts is independent of centrosome position. However, the anterior localization of the centrosome does correlate with the anterior bias in protrusion direction in the absence of the MIG-21 - DPY-19 and UNC-40/DCC pathways, indicating that it may have a function in specifying an anterior ground polarity.
Anterior polarity in the absence of the UNC-40/DCC and MIG-21-DPY-19 pathways is dependent on CDH-4/Fat
Given the prominent role of CDH-4 in the anterior polarization of QR, we investigated whether it is required for the anterior polarization observed in the absence of the MIG-21-DPY-19 and UNC-40/DCC pathways. Depletion of mig-21 by RNAi and UNC-40 by ZIF-1-mediated protein depletion in a cdh-4(hd40) mutant background resulted in a further increase in the percentage of Q cells that remain unpolarized (Fig. 4H) and failed to migrate (Fig. S4A), as confirmed by the almost fully penetrant effect on Q.pax migration (Fig. S4B). Importantly, among the 10% of Q cells that did form a protrusion, the anterior bias in polarization direction was lost (Fig. 4H), with cells polarizing in multiple directions, including angles below the anteroposterior body axis. Based on these results, we propose that the cell non-autonomous CDH-4/Fat pathway provides an anteriorly directed ground polarity, which is modified in QL by the MIG-21-DPY-19 and UNC-40/DCC pathways to mediate posterior polarization.
The MIG-21-DPY-19 pathway and CDH-4/Fat mediate ligand-independent localization of UNC-40/DCC
UNC-40/DCC is a key driver of actin-dependent protrusion formation (Quinn and Wadsworth, 2008). Studies of migrating neuronal growth cones (Kulkarni et al., 2013) and the polarization and invasion of the anchor cell during vulva formation (Wang et al., 2014) have shown that UNC-40 randomly clusters along the plasma membrane through an interlocked positive- and negative-feedback mechanism. Directional binding of the ligand UNC-6/netrin stabilizes these clusters and thereby induces protrusive activity at the appropriate side of the cell. Our observation that the Q neuroblasts frequently remain unpolarized in unc-40 mutants demonstrates that UNC-40 is also an important mediator of protrusion formation during Q cell polarization (Fig. 1B). Interestingly, however, it has previously been shown that the ligand UNC-6/netrin is not required (Honigberg and Kenyon, 2000). To corroborate these findings and to investigate the role of the alternative UNC-40 ligand MADD-4 (Chan et al., 1996; Tu et al., 2015), we examined protrusion formation and direction in unc-6(ev400) and madd-4(kr270) single and double mutants. Loss of unc-6 and madd-4 did not significantly affect protrusion formation, direction and migration (Fig. 5A, Fig. S5A,B), supporting the notion that UNC-40 functions ligand independently in Q neuroblast polarization.
To examine the subcellular localization of UNC-40, we endogenously tagged UNC-40 with GFP using CRISPR/Cas9-mediated genome editing. The C-terminal fusion of UNC-40 with GFP was functional (Fig. S6A) and, consistent with unc-40 mRNA expression (Fig. 3A,E), UNC-40::GFP expression increased during Q cell polarization and migration (Fig. S6B). The endogenous UNC-40::GFP protein localized diffusely along the plasma membrane, but also formed prominent punctate clusters at the cell periphery (Fig. 5B). Consistent with the role of UNC-40 in protrusion formation, quantification showed that these UNC-40 clusters predominantly localize at the leading edge of the polarizing Q cell (Fig. 5B,E). Of note, we found that these clusters dissolve when the Q neuroblast divides, but reappear at the leading edge of the migrating daughter cells QR.a and QR.p (Fig. S6B).
In the anchor cell, random clustering of UNC-40 in the absence of the UNC-6/netrin ligand leads to the formation of multiple, unstable and randomly directed protrusions (Wang et al., 2014). As mig-21, dpy-19 and cdh-4 mutants also display such ectopic spike-like protrusions, we investigated whether these genes are required for the stabilization and polarized localization of UNC-40. We found that in both mig-21(u787) dpy-19(e1414) double mutants and cdh-4(hd40) single mutants, the specific localization of UNC-40 at the anterior or posterior side of the Q neuroblast was lost (Fig. 5C,D). Quantification showed dispersed localization of UNC-40 clusters along the Q cell membrane and a clear reduction in the ratio of clusters at the anterior versus posterior side of the cell (Fig. 5E). In addition, we found that membrane localization and clustering of UNC-40 was significantly increased in mig-21 dpy-19 double mutants, while it was decreased in cdh-4 mutants (Fig. 5E,F), indicating that the two pathways also have distinct functions in controlling the localization or stability of UNC-40 at the cell membrane. Taken together, these results support the notion that the MIG-21-DPY-19 and the CDH-4/Fat pathways are part of a novel ligand-independent mechanism that controls the polarized activity of UNC-40.
The MIG-21-DPY-19 pathway prevents the Q neuroblasts from responding to an UNC-6/netrin-dependent dorsalizing signal
The UNC-40 ligand UNC-6/netrin plays a central role in polarization and migration along the dorsoventral axis (Wadsworth, 2002). Interestingly, we observed a prominent dorsal bias in protrusion direction in the mig-21 dpy-19 single and double mutants (Fig. 1B,C). As this dorsal polarization is dependent on UNC-40 (Fig. 1C), we investigated whether loss of the MIG-21-DPY-19 pathway triggers an ectopic response to UNC-6. We generated a mig-21 dpy-19; unc-6(ev400) triple mutant and found that loss of unc-6 strongly suppressed dorsal polarization (from 36-44% to 16%, P<0.01), with a clear shift towards posterior polarization in QL and anterior polarization in QR (Fig. 5A), an increase in migration distance (Fig. S5A) (with a concomitant increase in Q.a and Q.p mispositioning, Fig. S5C), and a partial rescue of QR.pax migration (Fig. S5B). These results show that the Q neuroblasts can respond to UNC-6, but that this response is normally prevented by the MIG-21-DPY-19 pathway. This is consistent with a role of the MIG-21-DPY-19 pathway in localizing UNC-40 at the leading edge, and supports the model that the MIG-21 - DPY-19 and CDH-4/Fat pathways are part of a ligand independent mechanism that redirects the activity of UNC-40 from the dorsoventral to the anteroposterior body axis.
The MIG-21-DPY-19, CDH-3-CDH-4/Fat and UNC-40/DCC pathways separately control directional outgrowth of the Q neuroblast protrusion
The transmembrane proteins MIG-21, DPY-19, CDH-4/Fat and UNC-40/DCC play a central role in the left-right asymmetric polarization and migration of the Q neuroblasts (Middelkoop et al., 2012; Schmitz et al., 2008; Sundararajan and Lundquist, 2012; Sundararajan et al., 2014). Previous studies have suggested that CDH-4 acts together with MIG-21 and UNC-40 in the anterior polarization of QR, while it functions in parallel to UNC-40 in the posterior polarization of QL (Sundararajan et al., 2014). Our detailed quantitative analysis shows that these interactions are more complex. We found that unc-40/DCC, mig-21 and dpy-19, and cdh-4/Fat and cdh-3 – the Fat-like cadherin gene identified in this study – represent three separate genetic pathways that control distinct aspects of the polarization process. Thus, in unc-40/DCC mutants, the direction of polarization is more variable than in wild type. In mig-21 and dpy-19 single or double mutants, there is a clear dorsal bias in polarization. Finally, in cdh-4/Fat mutants, there is an increased variability in polarization direction and a striking reversal in the polarity of QR. This polarity reversal is not observed in cdh-3 mutants, but double mutant analysis indicates that both Fat-like cadherins are part of the same pathway. Interestingly, we have found that Q lineage-specific depletion of CDH-4 did not reproduce the reversal in QR polarity observed in the null mutant, indicating that this is a cell non-autonomous function of CDH-4. This is consistent with previous mosaic rescue experiments (Sundararajan et al., 2014) and is in agreement with the expression of cdh-4 in neighboring seam cells and juvenile ventral cord neurons. Cadherin family members can form homo- or heterotypic interactions through their extracellular domains (Gooding et al., 2004) and similar interactions between CDH-4, CDH-3 and other cadherin-related proteins, such as CDH-1, may underlie the cell non-autonomous function of CDH-4 in QR polarization. Moreover, as has been described for Fat2 in collective cell migration in Drosophila, where Fat2 stabilizes Lar at the leading edge of neighboring cells (Barlan et al., 2017), CDH-4 may interact with PTP-3/Lar to promote anterior polarization of the Q neuroblasts.
The parallel function of the three pathways is also demonstrated in mutant combinations. Depletion of UNC-40 in a mig-21; dpy-19 double-mutant background resulted in a strong defect in protrusion formation, which is discussed below. Interestingly, in cells that did form a major lamellipodium, the polarization direction was predominantly anterior. This anterior bias is correlated with the position of the centrosome, which is located anterior to the nucleus in both QL and QR. The symmetric localization of the centrosome derives from the division of the hypodermal QLV5L and QRV5R precursor cells that generate the two Q neuroblasts (as anterior daughters that both inherit an anterior centrosome) and their sister cells: the seam cells V5L and V5R (the posterior daughters with a posterior centrosome, data not shown) (Sulston et al., 1983). In migrating cells, the centrosome is often localized at the front of the nucleus and it is thought that this contributes to the formation of a polarized microtubule network that transports components of the cell polarity and migration machinery to the leading edge of the cell (Etienne-Manneville, 2013; Luxton and Gundersen, 2011). Although centrosome position did not correlate with protrusion direction in QL and QR, it may contribute to the anterior bias that emerges in the absence of the MIG-21-DPY-19 and UNC-40 pathways. Such an anterior ground polarity may also be linked to the non-cell autonomous activity of CDH-4. Thus, loss of cdh-4 predominantly affects the anterior polarization of QR and the anterior bias observed in the absence of the MIG-21-DPY-19 and UNC-40 pathways is lost when CDH-4 is depleted. An intriguing possibility is therefore that the CDH-4 pathway links centrosome position with anterior polarization in QR, while this activity is modulated or overruled to allow posterior polarization in QL.
The UNC-40/DCC, CDH-3-CDH-4/Fat and MIG-21-DPY-19 pathways are required for the formation of a major lamellipodium-like protrusion
We found that UNC-40/DCC has a prominent role in protrusion formation. In unc-40 mutants, the Q neuroblasts frequently fail to form a major protrusion. This is also observed when UNC-40 is specifically depleted from the Q cell lineage, indicating that UNC-40 acts cell autonomously in this process. UNC-40 is an established regulator of protrusion formation in axonal growth cones and the directional outgrowth of invasive protrusions of the anchor cell during vulva development (Quinn and Wadsworth, 2008; Wang et al., 2014). Consistent with a direct function in protrusion formation, we found that endogenous UNC-40 specifically localizes at the leading edge of the Q cell protrusion. Although the intracellular effectors of UNC-40 vary in different cellular contexts, UNC-40 has been shown to act upstream of various regulators of actin dynamics, including the actin-binding protein UNC-115/AbLIM, the Rho family GTPase CED-10/Rac1 and its regulators TIAM-1 and UNC-73/Trio (Demarco et al., 2012; Yu et al., 2002), and UNC-34, an ortholog of the enabled/vasodilator-stimulated phosphoprotein (Ena/VASP), a key mediator of actin filament formation in migrating cells and growth cones (Gitai et al., 2003). Of these, CED-10/Rac1, UNC-73/Trio and UNC-34/Ena are required for Q cell migration (Dyer et al., 2010; Shakir et al., 2006), indicating that UNC-40 may signal through a CED-10/Rac1 and UNC-34/Ena dependent mechanism to form a major protrusion.
The Fat-like cadherins CDH-4 and, to a lesser extent, CDH-3, are also required for protrusion formation. Depletion of CDH-4 from the Q cell lineage revealed that this function is cell autonomous, demonstrating that – in addition to its cell non-autonomous function in specifying protrusion direction – it is also required in the Q neuroblasts for protrusion formation. To date, no downstream effectors of CDH-3 and CDH-4 have been identified. Studies in mammalian cells have shown that Fat1 binds Ena/VASP through a proline-rich class I EVH1 domain (Moeller et al., 2004; Tanoue and Takeichi, 2004). Interestingly, a consensus PPxxF class II type EVH1 binding motif is present in the intracellular domain of CDH-4 and this domain is directly adjacent to another proline-rich motif (PVVPPPL) (Schmitz et al., 2008), indicating that CDH-4 may act through an UNC-34/Ena-dependent mechanism as well.
mig-21 and dpy-19 single mutants show only a slight defect in protrusion formation, but this phenotype is enhanced in the double mutant, indicating that the two genes have a partial, non-overlapping function in regulating protrusive activity. Loss of mig-21 and dpy-19 also strongly enhances the defect in protrusion formation when UNC-40 is depleted or when depletion of UNC-40 is combined with loss of cdh-4. MIG-21 is a thrombospondin repeat-containing protein that is modified by the c-mannosyltransferase activity of DPY-19 (Buettner et al., 2013). How MIG-21 and DPY-19 are involved in protrusion formation is not known, but based on genetic epistasis analysis, mig-21 acts in the same pathway as ptp-3/Lar (Sundararajan and Lundquist, 2012). LAR-type RPTPs functionally interact with some of the same actin regulators (such as Trio and ENA/VASP) (Um and Ko, 2013) that act downstream of UNC-40/DCC and possibly CDH-4. Consistent with such potential overlap in downstream effectors, the three pathways act partially redundantly, with strong synergistic interactions when both the MIG-21-DPY-19 and UNC-40 or CDH-4 and UNC-40 pathways are depleted. We speculate that crosstalk at the level of shared downstream effectors may integrate the activity of the three pathways in protrusion formation and link it to the directional outgrowth mechanisms discussed above.
A ligand-independent function of UNC-40/DCC in Q neuroblast polarization
Previous studies on axon outgrowth and anchor cell polarization have shown that UNC-40 randomly clusters along the plasma membrane through an interlocked positive- and negative-feedback mechanism (Kulkarni et al., 2013; Wang et al., 2014). These spontaneously formed clusters are sufficient to form small protrusions, but for the formation of a larger lamellipodium-like protrusion, these clusters need to be stabilized. Such stabilization can be mediated through binding of the ligand UNC-6/netrin, which forms a ventral-to-dorsal gradient and polarizes protrusive activity along this axis (Wadsworth, 2002). UNC-6 can act in a concentration-dependent manner to promote outgrowth of the protrusion towards (or away from) its source of expression. Recent studies have shown that directionality can also be achieved through a stochastic mechanism in which extracellular cues act in a concentration-independent manner to influence the probability of UNC-40 localization and protrusion formation at a specific side of the cell (Limerick et al., 2018; Wadsworth, 2018). Interestingly, it has previously been shown that UNC-40 acts independently of UNC-6/netrin in Q neuroblast polarization (Honigberg and Kenyon, 2000). We could confirm these observations, and we show that protrusion formation is also unaffected by loss of the alternative UNC-40 ligand MADD-4 or the combined loss of both ligands. Although we cannot rule out the possibility that another, as yet unidentified, netrin-like ligand may be involved, these results strongly suggest that UNC-40 functions in a ligand-independent manner.
Endogenous UNC-40 forms distinct punctate clusters at the leading edge of the Q cell protrusion. We found that this highly polarized localization is dependent on the MIG-21-DPY-19 and CDH-4/Fat pathways. Thus, in mig-21 dpy-19 double mutants and cdh-4 single mutants, the UNC-40 clusters are still formed, but their localization is more dispersed along the Q neuroblast membrane. Consistently, we found that the Q neuroblasts form multiple ectopic, filipodia-like protrusions in these mutant backgrounds. Based on these results, we propose that the MIG-21-DPY-19 and CDH-4/Fat pathways have a dual function in protrusion formation. As discussed above, both pathways can directly promote protrusion formation, explaining why double mutants between these pathways and unc-40 display an enhanced defect in outgrowth activity. However, our results show that the two pathways also provide a ligand-independent mechanism for localized UNC-40 stabilization. Such a ligand-independent mechanism may be necessary to focus the activity of UNC-40 along the anteroposterior axis, instead of the dorsoventral axis that is specified by the UNC-6/netrin gradient. In support of such a model is our observation that the Q neuroblasts polarize towards the dorsal side in mig-21 dpy-19 double mutants and that this dorsal bias is lost in the absence of unc-6.
An important issue is how the MIG-21-DPY-19 and CDH-4/Fat pathways control the asymmetric localization of UNC-40 and how this is linked to their own function in polarity establishment. Several mechanisms have been proposed that can localize and polarize the activity of guidance receptors like UNC-40. In canonical chemoattraction, the concentration gradient of an extracellular cue results in polarization and activation of its receptor along this gradient (Tessier-Lavigne and Goodman, 1996). Alternatively, receptor activation may be aligned with cell intrinsic polarity axes without the need of direct extracellular guidance cues (Gujar et al., 2018; Mahekta et al., 2019; Limerick et al., 2018 and Wadsworth, 2018). Although we cannot discriminate between these possibilities, our observation that the polarization of the Q neuroblasts is mainly a cell-autonomous process – which is modified by extracellular CDH-4/Fat to achieve anterior polarization of QR – may favor the latter possibility. Given the evolutionary conservation of DPY-19, PTP-3/LAR and the Fat-like cadherins, ligand-independent UNC-40/DCC stabilization through these transmembrane proteins may provide a general mechanism for uncoupling UNC-40/DCC signaling from netrin guidance cues.
MATERIALS AND METHODS
C. elegans strains and culture
Unless noted otherwise, C. elegans strains were cultured at 20°C using standard conditions (Lewis and Fleming, 1995). The Bristol N2 strain was used as wild type. The alleles and transgenes used in this study are: LGI, unc-40(e271) (Brenner, 1974); unc-40(hu242[unc-40::gfp^flag3x]), unc-40(hu226[unc-40::zf1]); LGII, muIs32[Pmec-7::gfp; lin-15(+)] (Ch'ng et al., 2003); madd-4(kr270) (Seetharaman et al., 2011); LGIII, cdh-4(hd40) (Schmitz et al., 2008), cdh-4(hu240[cdh-4::zf1]), mig-21(u787) (Du and Chalfie, 2001), cdh-3(pk87) (Pettitt et al., 1996), cdh-3(hu238[cdh-3::zf1]), dpy-19(e1314) (Honigberg and Kenyon, 2000), dpy-19(hu257[dpy-19::gfp^SEC^flag3x]); LGIV, dpy-20(e1362); LGV: heIs63[Pwrt-2::gfp-ph; Pwrt-2::h2b::gfp; Plin-48::tomato] (Wildwater et al., 2011); LGX, unc-6(ev400), huIs166[Pwrt-2::h2b::mcherry; Pwrt::ph::mcherry; dpy-20(+)]; and linkage group unknown, casIs22[Pegl-17::gfp-TEV-S-cmd-1; Pegl-17::myri-mcherry; Pegl-17::mcherry-TEV-S-his-23; rol-6(dn)] (Chai et al., 2012), huIs181[Pegl-17::zif-1-SL2-mcherry; Pmyo-2::mcherry]. The following extrachromosomal arrays were generated: huEx439[Pegl-17::dpy-19; Pmyo-2::gfp] and huEx440[Pegl-17::dpy-19; Pmyo-2::gfp]. We used hT2 (I:III) [bli-4(e937) let-?(q782) qIs48] to generate a balanced cdh-4 cdh-3/hT2 strain. As recombination is not fully suppressed in this strain, all animals examined were checked by PCR for presence of the cdh-4 allele. Strains containing dpy-19(e1314) were grown at 15°C, but were shifted to 20°C a day prior to analysis. Synchronized populations of animals were obtained by collecting L1 larvae 0-1 h after hatching and growing them for 1 h for protrusion formation and directionality assays, 2 h for the analysis of centrosome position, 4-6 h for the position of the first Q neuroblast division and 12 h for determining Q.pax position.
Microscopy and fluorescence quantification
For confocal, epifluorescence and DIC microscopy, animals were mounted on 2% agarose pads containing 10 mM sodium azide. Static imaging of Q neuroblast polarization, migration and the position of the first division were performed using a Zeiss Axioscop with Axiocam camera and images were processed using ImageJ and Adobe Photoshop (Middelkoop et al., 2012). For imaging endogenous DPY-19::GFP, a PerkinElmer Ultraview Vox spinning disk confocal microscope (100× objective, 77.5% 488 nm laser power, exposure 800 ms, 9% 561 nm laser power, exposure 600 ms) was used. For UNC-40::GFP, the settings were 100× objective, 7% 488 nm laser power, exposure 800 ms, 2% 561 nm laser power, exposure 500 ms, 0.33 µm z-stack. To quantify membrane-associated UNC-40::GFP, line scans (counter clockwise or clockwise, starting from the point of contact with the neighboring anterior or posterior seam cell for QR and QL, respectively) were made using ImageJ. After background subtraction, average GFP intensity plus 1× standard deviation in wild type was used to define UNC-40::GFP clusters in images of wild-type and mutant Q neuroblasts. GFP clusters were divided proportionally to the line scan to determine clusters on the anterior versus the posterior side of the cell. To determine membrane GFP localization, average fluorescence at the membrane was divided by the average fluorescence of a line scan through the cytoplasm.
Samples for live imaging were prepared as previously described (Middelkoop et al., 2012). Animals were imaged using a PerkinElmer Ultraview Vox spinning disk confocal microscope (100× objective, 2-6% 488 nm laser power (gradually increasing from 2% to 6% during the course of the experiment), exposure 500 ms, 0.5 µm z-stack). Z-stacks were made every minute, for ∼300 min, using Volocity software. For the unc-40::zf1, mig-21(u787) dpy-19(e1314), Pegl-17::zif-1 strain, z-stacks were made every 10 min. Image sequences were made by selecting z-slices at every time point in which the Q neuroblast was in focus. Movies were then processed from image sequences using ImageJ software. To compensate for movement, individual time frames were first aligned with respect to the Q cell nucleus using a normalized cross-correlation in a custom-written Matlab script. Subsequently, stacks were further aligned in Fiji by a rigid body transformation using the StackReg plug-in (Thevenaz et al., 1998).
Single molecule fluorescence in situ hybridization
smFISH was performed as described (Raj et al., 2008). In short, wild-type animals carrying the heIs63 transgene were synchronized 0-5 h after hatching and fixed using 4% formaldehyde and 70% ethanol. Hybridization was carried out for >12 h at 37°C. unc-40, cdh-4 and cdh-3 oligonucleotide probes were designed using the web-based algorithm at singlemoleculefish.com and chemically coupled to Cy5. Z-stacks of 0.5 µm step-size were collected using a Leica DM6000 microscope equipped with a Leica DFC360FX camera, 100× objective and a Y5 filter cube (Cy5). Images were further analyzed and processed with ImageJ software. Quantification was performed as described previously (Middelkoop et al., 2012) by manually counting mRNA spots in Q neuroblasts, defined by the heIs63 transgene. Only fluorescent spots visible in at least two neighboring z-slices (0.5 µm step size) were counted.
Quantification of Q neuroblast polarization, migration and centrosome position
Protrusion formation, direction and the position of the first division were quantified as described (Middelkoop et al., 2012). The final position of the Q.pax was determined relative to the seam cells using DIC microscopy (Coudreuse et al., 2006) or by epifluorescence using the muIs32 (Ch'ng et al., 2003) or huIs181 transgenes. In some experiments, the final positions of the Q descendants AVM (QR.paa) and PVM (QL.paa) were scored relative to the developing vulva in L4 animals using the muIs32 transgene. The centrosome was visualized using the casIs22 transgene (Chai et al., 2012). The position of the centrosome was quantified relative to the anteroposterior body axis, and the position of the centrosome with respect to the direction of the protrusive front was determined by measuring the angle of the protrusion with the anteroposterior axis and subtracting the angle made by the centrosome.
Q-lineage-specific expression of dpy-19
The dpy-19-coding sequence was amplified from cDNA using primers dpy-19 FW and dpy-19 RV (Table S5). The PCR product was cloned into pDONR221 using Gateway technology. Pegl-17::dpy-19::unc-54-3′UTR was generated by three-way Gateway cloning of the eg-17 promotor, dpy-19-coding sequence and the unc-54 3′ UTR (Middelkoop et al., 2012) entry clones into the destination vector pCFJ150. The Pegl-17::dpy-19::unc-54 3′ UTR plasmid was injected at 10 ng/µl together with Pmyo-2::GFP (5 ng/µl) and pBluescriptII (135 ng/µl) into dpy-19(e1314) mutants. Two independent transgenic lines (huEx439 and huEx440) were isolated and used in this study.
CRISPR/Cas9-mediated genome editing
A zf1 tag was inserted at the C terminus of unc-40, cdh-4 and cdh-3 using single-stranded oligo DNA (ssODN) repair templates as described previously (Arribere et al., 2014; Paix et al., 2015). Sequences of the guide RNAs and repair templates are provided in Table S5. For inserting GFP at the C terminus of unc-40 and dpy-19, the previously described self-excising cassette (SEC) approach was used (Dickinson et al., 2015). sgRNAs were generated by in vitro transcription from PCR fragments containing the T7 promotor and the sgRNA sequence of interest. Primers to generate the PCR fragments for sgRNA synthesis and the homology arms are in Table S5. The homology arms were cloned into the GFP-containing vector pDD282 as described previously (Dickinson et al., 2015). Guide RNAs and repair templates were co-injected with recombinant SpCas9 (D'Astolfo et al., 2015) as described previously (Arribere et al., 2014; Paix et al., 2015).
Protrusion formation and directionality, position of division and Q.pax localization were compared using Fisher's exact test (Tables S1-S3) or Student's t-test as indicated. The correlation between smFISH spot abundance and migration distance was calculated using Spearman's rank correlation coefficient. For centrosome position, Levene's test was performed to address homogeneity of variance. To compare the number of UNC-40::GFP clusters on the anterior versus the posterior side of wild-type and mutant Q neuroblasts, Tukey's multiple comparison tests were performed. The membrane UNC-40::GFP localization was compared using the Holm-Sidak's multiple comparisons test.
We thank members of the Korswagen group for critically reading the manuscript, the Hubrecht Imaging Center (HIC) for assistance with microscopy and the Utrecht University Gene Editing Facility for recombinant SpCas9 protein. Some strains were provided by the CGC, which is funded by the NIH Office of Research Infrastructure Programs (P40 OD010440).
Conceptualization: H.C.K., A.E., T.C.M.; Methodology: H.C.K., A.E., T.C.M.; Investigation: A.E., T.C.M., M.C.B., E.B.; Writing - original draft: H.C.K., A.E., T.C.M.; Writing - review & editing: H.C.K., A.E., T.C.M.; Supervision: H.C.K.; Project administration: H.C.K.; Funding acquisition: H.C.K.
This work was funded by the research program 14NOISE01 of the Foundation for Fundamental Research on Matter (FOM), which is financially supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (NWO).
The authors declare no competing or financial interests.