EGL-20/Wnt and MAB-5/Hox Act Sequentially to Inhibit Anterior Migration of Neuroblasts in C. elegans

Directed neuroblast and neuronal migration is important in the proper development of nervous systems. In C. elegans the bilateral Q neuroblasts QR (on the right) and QL (on the left) undergo an identical pattern of cell division and differentiation but migrate in opposite directions (QR and descendants anteriorly and QL and descendants posteriorly). EGL-20/Wnt, via canonical Wnt signaling, drives the expression of MAB-5/Hox in QL but not QR. MAB-5 acts as a determinant of posterior migration, and mab-5 and egl-20 mutants display anterior QL descendant migrations. Here we analyze the behaviors of QR and QL descendants as they begin their anterior and posterior migrations, and the effects of EGL-20 and MAB-5 on these behaviors. The anterior and posterior daughters of QR (QR.a/p) after the first division immediately polarize and begin anterior migration, whereas QL.a/p remain rounded and non-migratory. After ~1 hour, QL.a migrates posteriorly over QL.p. We find that in egl-20/Wnt, bar-1/β-catenin, and mab-5/Hox mutants, QL.a/p polarize and migrate anteriorly, indicating that these molecules normally inhibit anterior migration of QL.a/p. In egl-20/Wnt mutants, QL.a/p immediately polarize and begin migration, whereas in bar-1/β-catenin and mab-5/Hox, the cells transiently retain a rounded, non-migratory morphology before anterior migration. Thus, EGL-20/Wnt mediates an acute inhibition of anterior migration independently of BAR-1/β-catenin and MAB-5/Hox, and a later, possible transcriptional response mediated by BAR-1/β-catenin and MAB-5/Hox. In addition to inhibiting anterior migration, MAB-5/Hox also cell-autonomously promotes posterior migration of QL.a (and QR.a in a mab-5 gain-of-function).


Introduction
The directed migration of neurons and neuroblasts is important in nervous system development to establish proper connectivity and circuits. Wnt signaling has been broadly implicated in mammalian cortical and hippocampal neurogenesis [1][2][3]. The Q neuroblasts in C. elegans have been a useful system in which to dissect the molecular mechanisms of directed neuroblast AQR migration (Fig 1). Genes affected by the mutations were mapped by whole genome single nucleotide polymorphism resequencing using the Cloudmap protocol [20] (data not shown). Both mapped to a region in the center of LG IV, and each harbored a mutation in the egl-20 gene, which encodes a C. elegans Wnt molecule. egl-20 has been shown previously to affect PQR but not AQR migration [7], consistent with the phenotypes of lq42 and lq74. lq42 introduced a premature stop codon at arginine 296 (C to T), and lq74 was a missense mutation changing cysteine 248 to tyrosine (G to A) (see Methods). Both failed to complement egl-20 (n585) for PQR migration (data not shown). The egl-20(gk453010) allele was generated by the Million Mutation Project (present in strain VC40076) [21]. gk453010 is a G to A transition resulting in an arginine 245 to opal stop codon mutation. The egl-20(hu105) allele also introduces a premature stop codon [22,23]. lq42, lq74, hu105, and gk453010 each displayed nearly completely penetrant anterior PQR migration (Fig 1), suggesting that they are all putative null mutations, with the exception of the lq74 missense mutation which may retain some function (1% PQR did not migrate completely to head).
AQR and PQR migration was analyzed in mutants of the four other C. elegans Wnt genes cwn-1, cwn-2, lin-44, and mom-2 (Fig 1). Putative null alleles of cwn-2 and lin-44 displayed no defects. A null cwn-1 allele displayed 29% failure of AQR to complete anterior migration. mom-2, which has not been previously reported to affect Q descendant migrations, displayed 3% AQR and PQR migration defects in the partial loss of function or77M+ allele (M+ is used to designate homozygous animals from a heterozygous mother and therefore with a wild-type maternal contribution of gene function). Thus, egl-20, cwn-1, and, mom-2 each affect AQR and PQR migration in distinct ways.
Redundancy among Wnts in Q descendant migrations has been reported previously [18]. The Wnt triple mutant egl-20(n585) cwn-2(ok895); cwn-1(ok546) displayed highly penetrant PQR anterior migration and a failure of both AQR and PQR migration along the anterior route (Fig 1). No AQR or PQR migrated posteriorly in this genotype. Mutation of lin-44 in the egl-20 (n585M+) cwn-2(ok895M+); cwn-1(ok546); lin-44(n1792) quadruple mutant resulted in posterior migration of both AQR and PQR (28% and 37%) (Fig 1). This indicates that lin-44 activity inhibits posterior AQR and PQR migration in the absence of egl-20, cwn-1, and cwn-2, and suggests that lin-44 can act as a posterior repellant. Previous studies indicated that lin-44 caused posterior displacement of QL descendants PVM and SDQL [18], and HSN [24], in certain multiple Wnt mutant combinations, consistent with a posterior repellant activity of lin-44. The triple and quadruple Wnt mutants were scored with maternal contribution of both egl-20 and cwn-2, so we cannot be certain that maternal Wnt function remains in these mutants. MIG-14/ Wntless is required for Wnt processing and function and affects QL descendant migrations [25]. Two mig-14 mutants had less severe defects than the Wnt triple and quadruple mutants (Fig 1). The likely hypomorph ga62 had weak effects on both AQR and PQR, and the lethal inframe deletion mutation or78, with wild-type maternal contribution, resembled egl-20 and only affected PQR migration. These data suggest either that these mutations did not completely eliminate mig-14 function, or that mig-14 is not involved in all Wnt-related signaling events.

Wnt Signaling Might Not Affect Early Q Neuroblast Migration
The above results, along with previous studies [18,26], indicate that the five C. elegans Wnt genes and four Frizzled genes mig-1, lin-17, cfz-2, and mom-5 act redundantly to control Q descendant migration. However, egl-20 alone and the egl-20(n585) cwn-2; cwn-1 triple mutant displayed no defects in early Q anterior-posterior protrusion and migration, despite severe AQR and PQR descendant migration defects [7,26] (Fig 1). To further investigate the role of Wnts in early Q protrusion and migration, we analyzed early QL and QR protrusion and migration with the ayIs9[Pegl-17::gfp] transgene, expressed in the early Q cells [27][28][29]. Two additional mutant combinations were tested: the Wnt triple mutant egl-20(lq42M+); mom-2 (or77M+); lin-44(n1792) and the Frizzled double mutant mom-5(ne12M+) lin-17(n761M+). Neither displayed defects in direction of initial QL and QR migration (QL divided atop V5 and QR atop V4 (n 10)) (data not shown). We have not scored early Q migrations in the absence of all Wnt signaling due to maternal perdurance and/or incomplete knockdown of all Wnt genes at once. However, we have not observed early Q directional migration defects in any Wnt signaling mutant combination tested, despite strong AQR and PQR migration defects. This suggests Wnt signaling might not be involved in early Q anterior-posterior protrusion and migration, with function restricted to later migrations of the Q descendants.

QR Daughters Begin Migration Immediately after QR Division
In order to understand how EGL-20/Wnt signaling drives posterior QL descendant migration, the behaviors of QR.a/p and QL.a/p just after division were analyzed. We monitored the position of the Q cell daughters using a Pegl-17::gfp transgene ayIs9 expressed in the Q cells (Fig 2). Newly-hatched L1 larvae were synchronized within the same hour-long age range at half-hour timepoints after hatching (e.g. 3.5-4.5h, 4.0-5.0h, 4.5-5.5h, 5.0-6.0h, 5.5-6.5h, 6.0-7.0h) (see Methods for synchronization of L1 larvae). The positions of QR, QL, and their daughters were determined in 20 animals at each timepoint. At 3.5-4.5h, 19/20 of the QR examined had divided, whereas only 13/20 of the QL had divided. This trend held true in all genotypes analyzed, suggesting that QL divides slightly later than QR. Indeed, at 4.0-5.0h, 4/20 QL had not yet divided, but 20/20 of the QRs had divided.
After division, the QR and QL daughters had a rounded morphology with little or no protrusion (Figs 2A, 2B and 3). Shortly after division in the 4.0-5.0h timepoint, the QR daughters polarized and begun migrating to the anterior (Fig 2C and 2D). Migrating QR.a/p cells first elongated in the anterior-posterior axis and then extended F-actin-rich lamellipodia-like protrusions in the direction of migration (to the anterior), similar to what has been previously reported ( Fig 3A and S1 Movie) [30][31][32]. The QR.a/p cells maintained this polarized, migratory morphology as they migrated anteriorly until their second round of cell division at approximately 7h post-hatching (Figs 2D-2I and 3A). At 6.0-7.0, QR.a/p cells began to lose the polarized morphology and assumed a rounded morphology in preparation for the second round of cell division (Figs 2I and 3A). Time-lapse imaging with the casIs330 transgene (S1 Movie, S2 Movie, and Fig 3) [30] resulted in an approximately 2-fold delay in migration and division relative to the ayIs9 transgene.
To account for variability in the timing of migration, QR.a/p migration was quantified in 20 animals at each time point using ayIs9 (Fig 4). Cells were scored as migratory if they extended protrusions to the anterior, often accompanied by migration from their birthplace and also separation from one another (Figs 2E, 2F and 3A). At 4.0-5.0h, 6/20 of the QR daughters had polarized and started migrating to the anterior (Fig 4). At 4.5-5.5h, 16/20 were migrating, and at 5.0-6.0, 18/20. By 5.5-6.5h, all (20/20) of the QR.a/p were polarized and migrating to the anterior (Fig 4).

QL Daughters Remain Rounded and Non-Migratory
In contrast to QR.a/p, the QL.a/p daughters did not polarize and migrate anteriorly. From 3.5-5.5h, at a time when QR.a/p had polarized and begun migration, the QL.a/p cells retained a rounded, non-migratory morphology and did not migrate (Figs 2A'-2F' and 3B). Occasionally, protrusions were observed on non-migratory QL.a/p (e.g. Fig 2F'), but they did not assume the elongated morphology and robust anterior protrusion observed in QR.a/p. At 5.0-6.0h, the anterior QL.a daughters began migrating posteriorly over the anterior QL.p daughters, which retained a rounded morphology (Figs 2G'-2H' and 3B) (S1 Movie). QL.p showed no significant migration over this time period, consistent with previous observations [33]. Often, QL.a apparently migrated over the top of QL.p (Figs 2H' and 3B). QL.a was scored as migrating posteriorly if it was over QL.p, or if it was indistinguishable from QL.p suggesting it was migrating in front of or behind QL.p (data not shown).
Quantification of migration in 20 animals showed that only 1/20 QL.a began posterior migration at the 5.0-6.0h timepoint (Fig 4). At 5.5-6.5h, 13/20 of the anterior QL.a cells had migrated. At 6.0-7.0h, all QL.a had migrated posterior to QL.p and began to assume a rounded morphology preceding the second round of cell division (Fig 2I'). Posterior QL.p showed little . Individual images were taken from S1 and S2 Movies at the timepoints in minutes indicated in the lower right of each micrograph. Red indicates MYR::mCherry at membranes and in chromatin (HIS-24::mCherry), and green represents Factin (MOEabd::GFP). (A) QR and descendants are indicated. Arrows point to the F-actin-rich lamellipodial protrusions that accompany polarization and anterior migration. F-actin also accumulated at the furrows between dividing cells (e.g. between QR.a and QR.p at 40min, and between QR.aa and QR.ap at 200min). (B) QL and descendant migration. Arrows point to the extensions from QL.a posteriorly over QL.p. In these time-lapse imaging experiments with the casIs330 transgene, migrations and cell divisions were delayed bỹ 2-fold compared to ayIs9 still images, but the same pattern of QL.a/p and QR.a/p migration as observed with ayIs9 still images was conserved. or no polarization or migration in the entire time before the second round of cell division (7.0-8.0h after hatching).
In sum, QR.a/p polarized and began to migrate anteriorly shortly after division. The division of QL was delayed compared to QR, and the QL.a/p daughters remained rounded and nonmigratory in the 3.5-5.0h time window while the QR.a/p daughters migrated to the anterior. At 5.0-6.0h, QL.a began migrating posteriorly over QL.p and continued posterior migration. QL.p did not migrate in this time period. A schematic summary of QR.a/p and QL.a/p migratory behavior is shown in Fig 5.
As expected, direction and extent of initial QR and QL migration prior to the first Q cell division was unaffected by egl-20 and mab-5 (Figs 4, 6A and 7A). Furthermore, the QR.a/p anterior and posterior daughters polarized and migrated to the anterior in all cases (Figs 4, 6 and 7). Previous work showed that in egl-20 mutants, the posterior QR.p daughter displayed variable posterior polarization during migration [35]. We did not follow this phenotype in these studies.
In egl-20(n585), QL.a/p polarized and migrated anteriorly shortly after division at the 4.0-5.0h timepoint, similar to QR.a/p (Fig 4). In some cases, QL.a/p polarized and began to migrate before QR.a/p (Fig 6B). Upon quantification, 3/20 QL.a/p at the 4.0-5.0 timepoint and 14/20 at the 4.5-5.5h timepoint had polarized and begun anterior migration, similar to QR.a/p (Fig 4). QL.a/p resembled QR.a/p in their anterior migrations through the remaining timepoints ( Fig  6C). Thus, in the absence of EGL-20/Wnt, QL.a/p polarized and migrated anteriorly, indicating that EGL-20 inhibits polarization and anterior migration of QL.a/p.
In mab-5 mutants, QL.a/p also polarized and migrated to the anterior similar to egl-20 ( Fig  7). Upon quantification of two mab-5 mutants, 2/20 QL.a/p polarized and began migrating to the anterior at the 4.0-5.0h timepoint in both mutants, and 7/20 and 4/20 at the 4.5-5.5h timepoint (Fig 4). By 5.5-6.5h, all QL.a/p had migrated anteriorly in both mab-5 mutants (Fig 4). These results suggest that MAB-5/Hox also inhibits QL.a/p anterior polarization and migration. mab-5 Gain-Of-Function Causes Posterior QR.a/p Migration mab-5(e1751) is a gain-of-function variant that results in ectopic expression of mab-5 in cells, including in QRs which normally do not express it [36]. mab-5(e1751) mutants display posterior migration of both QL and QR descendant neurons AQR and PQR, the opposite of mab-5 loss-of-function [19] (Fig 1). We analyzed Q descendant behaviors in mab-5(e1751). Initial Q migrations were normal, as QR migrated anteriorly and QL posteriorly, indicating that mab-5 (e1751) does not affect initial Q migration (Fig 8A). QL.a/p behavior generally resembled what is seen in wild type (persistent rounded and non-migratory morphology until 5.0-6.0h, with QL.a beginning to migrate posteriorly over QL.p at this time) (Fig 8).
Strikingly, QR.a/p behaved similarly to QL.a/p in mab-5(e1751) (Fig 8). Some QR.a/p slightly separated from one another, but did not show the polarized and anterior migratory behavior seen in wild-type QR.a/p from 4h onward. Eventually, QR.a migrated posteriorly over QR.p just as QL.a did, beginning at the 5.5-6.5h timepoint (Figs 4 and 8). mab-5(e1751) resulted in a delay in QR.a/p and QL.a/p posterior migration relative to wild-type QL.a/p (Fig  4). At the 6.0-7.0h timepoint, all of QL.a had migrated posteriorly in wild-type where only 12/ 20 QR.a and 10/20 QL.a had migrated posteriorly in mab-5(e1751). Together with mab-5 lossof-function, these data suggest that MAB-5 normally inhibits anterior migration of QL.a/p and can also inhibit QR.a/p anterior migration in the gain-of-function mutant (Fig 5). In addition to inhibiting anterior migration, mab-5 activity can also induce posterior migration of the QR. a cell, similar to wild-type QL.a. mab-5 is expressed in QL as well as in other cells surrounding QL in the posterior of the animal. We constructed a transgene driving mab-5 expression from the egl-17 promoter expressed in the early Q cells. Q descendant neurons AQR and PQR both migrated posteriorly in animals harboring this transgene, similar to mab-5(e1751) (Fig 1). Initial QL and QR migration were normal in Pegl-17::mab-5 animals, but both QL.a/p and QR.a/p resembled mab-5(e1751) (they remained rounded and non-migratory) (Fig 9). These data indicate that MAB-5 expression in the Q cells themselves is sufficient to prevent anterior migration of QL.a/p and QR.a/p and drive their posterior migration, consistent with previous results showing autonomy of mab-5 function in neuronal descendant migrations [15].
To explore this difference between mab-5 and egl-20 further, we determined, at the 4.5-5.5h timepoint, the proportion of QL.a/p that had polarized and begun to migrate anteriorly in animals with migrating QR.a/p (Fig 10). In wild type, 4% of QL.a had begun migration (to the posterior), while in egl-20(n585) and egl-20(hu105), 78% and 80% of QL.a/p had begun anterior migration (p < 0.05). Furthermore, we noted that 8% of QL.a/p polarized and migrated before QR.a/p in egl-20(n585) (Fig 6B).

MAB-5/Hox and EGL-20/Wnt Inhibit Anterior Migration
Here we describe the distinct behaviors of QL and QR descendants in response to Wnt signaling. After the first Q cell division, QR.a/p and QL.a/p had a rounded morphology with little or no protrusion. Shortly after division, QR.a/p began to elongate in the anterior-posterior axis by extending F-actin-rich lamellipodial protrusions to the anterior. They then began anterior migration with polarized, migratory morphology for approximately 2h, at which time they stopped migrating, become rounded, and underwent their second round of cell division. In contrast, QL.a/p remain rounded with little or no protrusive morphology and migration. Our results show that EGL-20/Wnt and MAB-5/Hox are required to prevent QL.a/p from polarizing and migrating to the anterior. When MAB-5/Hox was ectopically expressed in QR.a/p via the mab-5(e1751) mutation or transgenic expression of mab-5 in both Q cells, QR.a/p polarization and migration was similarly inhibited.

EGL-20/Wnt has MAB-5-Independent and MAB-5-Dependent Roles
Both egl-20 and mab-5 loss-of-function mutants displayed anterior QL.a/p migration. However, we noted a delay in anterior migration in mab-5 not observed in egl-20. QL.a/p in mab-5 remain rounded and non-migratory for a short time (~0.5-1h) before they polarized and migrated anteriorly. In contrast, QL.a/p in egl-20 mutants resembled QR.a/p: they immediately polarized and began anterior migration. mab-5 expression is induced in QL by canonical Wnt signaling via EGL-20/Wnt and BAR-1/ β-catenin. The delay in QL.a/p anterior migration was also observed in bar-1/β-catenin mutants, similar to mab-5. This suggests that EGL-20/Wnt has at least two distinct roles in inhibiting migration (Fig 11). EGL-20/Wnt might acutely inhibit migration in a MAB-5 and BAR-1-independent pathway, and later, via canonical Wnt signaling and MAB-5 expression, consolidate this inhibition. Consistent with this idea of stepwise regulation of Q descendant migration by EGL-20, previous studies showed that QR descendants display cell-intrinsic temporal responses to Wnt signals that determines their final positions in the anterior-posterior axis [35].
How might EGL-20 and MAB-5 inhibit QL.a/p anterior migration? EGL-20 could activate an acute response by acting directly on the ability of the cell to polarize and migrate (e.g. e.g. inhibiting cytoskeletal rearrangements and protrusive events necessary for migration). This response does not require BAR-1 or MAB-5, suggesting that it might be a non-canonical Wnt signaling pathway that acts directly on protrusive ability (e.g. the cytoskeleton). There is precedence for BAR-1-independent Wnt signaling in later QR descendant migrations, which are shortened in Wnt mutants independent of BAR-1/β-catenin [18]. In addition to the acute role, which is transient, EGL-20 can also activate canonical Wnt signaling via BAR-1/β-catenin resulting in MAB-5 expression. The acute EGL-20 signal might inhibit anterior migration until MAB-5 is expressed, which begins in QL during its initial migration and before division [17]. In this scenario, the delay in anterior migration in mab-5 mutants might indicate the lag time in mab-5 expression and regulation of target gene expression to inhibit anterior migration. Wnt signaling and mab-5 expression remain tied in a feedback loop that ensures consistent and continuous levels of mab-5 expression [17].
MAB-5 might consolidate inhibition of anterior migration by regulating genes involved in polarization and migration and/or genes involved in responding to an A-P guidance cue. Indeed, expression of the Hox factor LIN-39 is inhibited in QL.a/p by MAB-5 [30]. LIN-39 normally promotes anterior migration in QR.a/p by driving the expression of the transmembrane MIG-13 molecule [30]. In parallel to SDN-1/Syndecan [37], MIG-13 responds to an A-P guidance cue by mediating cytoskeletal rearrangements underlying anterior protrusion and migration [30]. Thus, MAB-5 might inhibit anterior QL.a/p migration in part by inhibiting LIN-39 expression in QL.a/p.

MAB-5 Promotes Posterior Migration
While QL.a/p normally remain rounded and non-migratory while QR.a/p migrate anteriorly, QL.a eventually migrates posteriorly over QL.p. QL.a posterior migration is dependent on MAB-5, as we found that ectopic expression of MAB-5 in both QL and QR resulted in posterior migration of both anterior daughters. Thus, MAB-5 is not simply inhibiting migration. One possibility is that by inhibiting anterior migration, MAB-5 allows QL.a/p to respond to a later de novo guidance signal that directs posterior migration. By not migrating anteriorly, QL.a/p are in a position to respond to this new signal. Alternatively, MAB-5 might regulate gene expression in QL.a/p that alters response to a posterior guidance cue, possibly the same A-P cue used by QR.a/p to migrate anteriorly. As Wnts regulate Q descendant migrations, MAB-5 might alter the response of QL.a/p to Wnt signals directing anterior-posterior migrations.
The Wnt quadruple mutant cwn-1; egl-20 cwn-2; lin-44 showed some posterior migration of both AQR and PQR. EGL-20/Wnt is required to activate mab-5 expression, so it is possible that in the Wnt quadruple, posterior migration occurs in the absence of mab-5. This indicates that Wnts constitute the anterior-posterior guidance system to which mab-5 modifies responses. In the absence of Wnts, some cells migrate posteriorly despite not having mab-5 expression due to a disrupted anterior-posterior guidance system. It is also possible that mab-5 is activated in an egl-20-independent manner in both AQR and PQR in this quadruple mutant and is responsible for posterior migration.
In sum, our results indicate that both QR.a/p and QL.a/p can both respond to an anterior migration signal. Normally, EGL-20 prevents QL.a/p from responding by first acutely inhibiting migration, likely due to the inherent sensitivity of QL to the EGL-20 signal [16]. EGL-20 also activates mab-5 expression in QL.a/p via BAR-1/ β-catenin and canonical Wnt signaling, which likely results in changes in gene expression that maintains inhibition of anterior migration. MAB-5-induced gene expression might also then promote posterior migration. However, it is also possible that by inhibiting anterior migration, mab-5 allows the QL.a/p cells to respond to a later posterior migration signal that is not present early in QL.a/p migration. These results indicate that egl-20 and mab-5 mutations do not transform QL into a QR-like fate, but rather discretely modify how these cells differentially respond to anterior-posterior guidance information.
These studies and others are beginning to paint a picture of the early migration events of the Q neuroblasts and descendants. Initial anterior QR and posterior QL migration are regulated by inherent differences in the functions of the transmembrane molecules UNC-40/DCC and PTP-3/LAR in QR versus QL [29], and does not appear to involve Wnt signaling. While the signal specifying initial anterior versus posterior migration is unknown, the Fat-like Cadherin CDH-4 non-autonomously controls UNC-40/DCC and PTP-3/LAR function in QR versus QL [26]. Due to posterior migration of QL and inherent differences in sensitivity, an EGL-20/Wnt signal acutely inhibits QL.a/p anterior migrations, possibly via a non-canonical mechanism. Via canonical Wnt signaling and BAR-1/β-catenin, EGL-20/Wnt also induces MAB-5 expression in QL.a/p, which inhibits LIN-39/Hox and thus MIG-13 expression and consolidates inhibition of anterior migration. Additionally, MAB-5 might regulate other genes that inhibit anterior migration and direct posterior migration. It will be important to identify the potential non-canonical mechanism of EGL-20/Wnt inhibition of anterior migration, and to define other genes regulated by MAB-5 to autonomously inhibit anterior migration and promote posterior migration.
The Pegl-17::mab-5::gfp transgene was produced by placing a full-length mab-5 cDNA downstream of the egl-17 promoter and fused in frame to the gfp coding region at the 3' end (C-terminal tag). The sequence of this plasmid, pEL862, is available upon request. Six independent extrachromosomal arrays and two independent integrants (lqIs220 and lqIs221) had a similar effect of causing posterior AQR migration (Fig 1).

egl-20 and mab-5 Allele Sequencing
The lesions associated with egl-20(lq42 and lq74) and mab-5(mu114) were determined by whole genome sequencing using Cloudmap [20] and confirmed with Sanger sequencing of the region. egl-20(lq42) was a C to T transition at chromosome IV position 9,813,864 (Wormbase release WS249) resulting in an arginine to stop; egl-20(lq74) was a G to A transition at 9,814,127 resulting in a cysteine to tyrosine missense; egl-20(gk453010) was a C to T transition at 9,814,137 resulting in an arginine to stop; and mab-5(mu114) was a G to A transition at chromosome III position 7,783,349 in the first exon of mab-5 resulting in a tryptophan to stop. mab-5(e2088) was a complex rearrangement involving the second exon of mab-5 at position 7,782,932. The exact molecular nature of e2088 could not be determined.

L1 Synchronization and Q Cell Imaging
L1 larvae were synchronized by hatching as previously described. Gravid adults and larvae were washed from plates on which many eggs had been laid. Eggs were allowed to hatch for one hour, and newly hatched larvae, all within an hour's age of one another, were washed off and allowed to develop for specific times: 3.5h for the 3.5-4.5h timepoint; 4h for the 4-5h timepoint; 4.5h for the 4.5-5.5h timepoint; 5h for the 5-6h timepoint; 5.5h for the 5.5-6.5h timepoint; 6h for the 6-7h timepoint; 6.5h for the 6.5-7.5h timepoint; and 7h for the 7-8h timepoint. At the specified timepoint, L1 larvae were mounted for microscopic inspection and imaging of Q cell position and morphology using the ayIs9[Pegl-17::gfp] transgene.

Scoring Q Cell Position
From micrographs of L1 larvae expressing ayIs9[Pegl-17::gfp], the morphology and position of QL.a/p and QR.a/p were determined. QR.a/p cells were scored as migrating if they had assumed an elongated migratory morphology with anterior protrusions and/or if QL.a had separated from QL.p (Figs 2 and 3). Cells that remained rounded and adjacent to one another were scored as non-migratory. Posterior migration of QL.a in wild type and QL.a and QR.a in mab-5 gainof-function was defined by QX.a extending protrusions on top of QX.p, or if QX.a could not be distinguished from QX.p, indicating that QX.a was migrating behind or in front of QX.p (Figs 2  and 3). Significance of differences in Fig 4 were determined using Fisher's exact test.
For data in Fig 10, L1 larvae were synchronized and imaged at the 4.5-5.5h timepoint, when most QR.a/p had migrated anteriorly and QL.a/p remained rounded and non-migratory. The migrations of QL.a/p were scored in those animals with migratory QR.a/p (n = 50 for each genotype). Significances of difference were determined by Fisher's exact test.

Time-Lapse Imaging of QL and QR
For time-lapse imaging, the casIs330 transgene was used to visualize QL, QR and descendants. casIs330 contains the Q-cell promoter egl-17 driving the expression of myr::mCherry (to mark membranes), his-24::mCherry (to mark chromatin), and MOEabd::gfp, the actin-binding domain of human moesin fused to GFP, to mark F-actin [31]. Images were acquired every 2 minutes using a spinning disk confocal microscope (Zeiss Axio Observer Z1, with Yokogawa CSU-X1 Spinning Disk Unit) using previously-described methods to anesthetize and immobilize animals. Images were analyzed and assembled using ImageJ and Adobe Photoshop. Timelapse imaging with the casIs330 transgene delayed migrations and cell divisions by~2-fold compared to ayIs9 still images, but the same pattern of QL.a/p and QR.a/p migration as observed with ayIs9 still images was conserved.

Scoring AQR and PQR Migration
AQR and PQR are neuronal descendants of QR and QL, respectively, and migrate the longest distances of Q descendants into the head and tail of the animal. AQR normally migrates anteriorly to a region near the anterior deirid and the posterior pharyngeal bulb, and PQR normally migrates posteriorly near the phasmid ganglia posterior to the anus in the tail. AQR and PQR position was scored as described previously. The animal is divided into five regions in the anterior-posterior axis: Position 1 is the normal position of AQR; position 2 is posterior to position 1 but anterior to the vulva; position 3 is adjacent to the vulva; position 4 is the position of Q cell birth; and position 5 is the normal position of PQR posterior to the anus. AQR and PQR position was scored in 100 animals of each genotype indicated in Fig 1. Supporting Information S1 Movie. Division and migration of QR.a/p in wild-type. casIs330[Pegl-17::myr::mCherry, Pegl-17::HIS-24::mCherry, Pegl-17:: MOEabd::GFP] was used to image QR division and QR.a/p migrations. mCherry labeled cell membranes and chromatin (red), and GFP labeled F-actin (green). Images were acquired every 2 minutes using a spinning disk confocal microscope (Zeiss Axio Observer Z1, with Yokogawa CSU-X1 Spinning Disk Unit). Total imaging time was 200 minutes. Time-lapse imaging with casIs330 resulted in an approximately 2-fold delay in events relative to ayIs9 in timepoint analysis. Anterior is left, and dorsal is up. (AVI) S2 Movie. Division and migration of QL.a/p in wild-type. casIs330[Pegl-17::myr::mCherry, Pegl-17::HIS-24::mCherry, Pegl-17:: MOEabd::GFP] was used to image QL division and QL.a/p migrations. mCherry labels cell membranes and chromatin (red) and GFP labels F-actin (green). Images were acquired every 2 minutes using a spinning disk confocal microscope (Zeiss Axio Observer Z1, with Yokogawa CSU-X1 Spinning Disk Unit). Total imaging time was 250 minutes. Time-lapse imaging with casIs330 resulted in an approximately 2-fold delay in events relative to ayIs9 in timepoint analysis. Anterior is left, and dorsal is up. (AVI)