The POU transcription factor UNC-86 controls the timing and ventral guidance of C. elegans axon growth

Abstract

The in vivo mechanisms that coordinate the timing of axon growth and guidance are not well understood. In the C. elegans hermaphrodite specific neurons, the lin-4 microRNA controls the stage of axon initiation independent of the UNC-40 and SAX-3 ventral guidance receptors. lin-4 loss-of-function mutants exhibit marked delays in axon outgrowth, while lin-4 overexpression, leads to precocious growth in the L3. Here we show that loss of the POU transcription factor UNC-86 not only results in penetrant ventral axon growth defects in the HSNs, but also causes processes to extend in the L1, three stages earlier than wild-type. This temporal shift is not dependent on UNC-40 or SAX-3, and does not require the presence of lin-4. We propose that unc-86(lf) HSN axons are misguided due to the temporal decoupling of axon initiation and ventral guidance responses.

Introduction

To generate functional neural circuits, axonal growth cones must migrate appropriately to their targets. Several conserved families of extracellular cues—including the netrins, slits, semaphorins, and ephrins—as well as their corresponding receptors specify the direction of these migrations (Guan and Rao, 2003). While a great deal is known about the expression and downstream effectors of these guidance families (Butler and Tear, 2007; Polleux et al., 2007; O’Donnell et al., 2009), the ways in which these signaling pathways are coordinated spatially and temporally with axon growth are less well understood. In vitro experiments have revealed the growth-promoting features of guidance cues like netrin (Kennedy, 2000), revealing that axons can extend in response to extrinsic spatial patterning cues. However, observations from the C. elegans hermaphrodite specific motor neurons (HSNs) have shown that growth can also be initiated independently of the unc-6/netrin and slt-1/slit ventral signals (Desai et al., 1988; Wadsworth et al., 1996; Zallen et al., 1998; Hao et al., 2001; Adler et al., 2006; Olsson-Carter and Slack, 2010), suggesting that independent mechanisms ensure the correct timing of axon initiation relative to the activation of these guidance pathways.

The C. elegans heterochronic pathway was first defined for its role in directing stage-specific cell-division patterns in the hypodermal seam cells (Moss, 2007), although subsets of heterochronic genes were later shown to control developmental timing in the vulva and neurons as well (Hallam and Jin, 1998; Aurelio et al., 2003; Johnson et al., 2007). In the HSNs, three members of the heterochronic pathway, the lin-4 microRNA (miRNA) and its targets lin-14 and lin-28, direct when HSN axons extend (Olsson-Carter and Slack, 2010). Axon outgrowth is delayed in lin-4 mutants, while it is initiated one stage too early upon lin-4 overexpression or in lin-14 or lin-28 loss-of-function mutants (Olsson-Carter and Slack, 2010)

Here we identify an additional gene that controls the timing of HSN axon outgrowth, the neuronally expressed POU homeodomain transcription factor UNC-86. POU proteins are required for a range of developmental events including the maintenance of vertebrate stem cell identity and the temporal specification of Drosophila neuroblasts (Herr et al., 1988; Pearson and Doe, 2004; Yu et al., 2007), and one family, UNC-86/Acj6/BRN3, controls axon growth and pathfinding in C. elegans, Drosophila, and vertebrates (Certel et al., 2000; Erkman et al., 2000; Sze et al., 2002; Komiyama et al., 2003; Badea et al., 2009). In the HSNs, unc-86 is constitutively expressed after cell cycle exit (Finney and Ruvkun, 1990) and was previously predicted to act with the zinc-finger transcription factor SEM-4 and the unidentified EGL-45 protein to promote late HSN maturation events, including axon outgrowth ((Desai et al., 1988; Basson and Horvitz, 1996)).

Here we show that unc-86 null mutants exhibit penetrant defects in both growth and ventral guidance, with HSN axons extending approximately three stages prematurely along the anterior/posterior axis. This precocious axon initiation is independent of the two well-conserved guidance pathways—unc-6/netrin and slt-1/slit—that normally orient HSN ventral growth (Desai et al., 1988; Wadsworth et al., 1996; Zallen et al., 1998; Hao et al., 2001; Adler et al., 2006). Based on these findings, we predict that the pathfinding errors in unc-86 mutants are not due to changes in the expression of unc-6 or slt-1 pathway members, but instead stem directly from a shift in the relative timing of axon initiation and activation of ventral guidance responses.

Results

unc-86 is necessary for ventral axon guidance in the HSNs

The HSNs are born in the tail and migrate anteriorly to the mid-body during embryonic development (Desai et al., 1988). In the L1 larval stage, the neurons extend short neurites that are restricted ventrally after approximately 4 hours (Adler et al., 2006), and by the end of L2, they have formed a dynamic leading edge. The HSNs complete ventral migration toward the ventral nerve cord (VNC) by early L3, and at the L3/L4 transition, select a single ventral axon that later turns anteriorly (Fig. 1)(Desai et al., 1988; Adler et al., 2006; Olsson-Carter and Slack, 2010).

Fig. 1. Wild-type HSNs initiate axon growth in the ventral direction.

Top: Cartoon representation of an L4-stage animal, with the CAN neuron and excretory canal dorsal to the HSN. Image not drawn exactly to scale. Bottom: Images of a representative L4-stage wild-type HSN expressing lin-4∷GFP(zaIs1) that extends its axon ventrally and then anteriorly. Left: DIC merged with GFP. Right: GFP only. The arrow points to the excretory canal, arrowheads designate the HSN axon paths, the scale bars represent 5 μm, and anterior is to the left and ventral is down. The CAN neuron is in a distinct focal plane and is thus undetectable. The GFP images were pseudocolored green.

To investigate a potential function for unc-86 during HSN axon development, left HSNs (HSNLs) were visualized at the L4 larval stage in two presumptive unc-86 null mutants using the lin-4∷GFP(zaIs1) reporter strain. HSNL axons were visible in mutant animals at this stage (Fig. 2A), but they exhibited pronounced guidance defects (Fig. 2B-C). While wild-type HSNs first extend axons ventrally (Fig. 1)(Desai et al., 1988; Adler et al., 2006; Olsson-Carter and Slack, 2010), nearly all mutant HSNs initiated outgrowth in the anterior or posterior directions (Fig. 2B-C), reminiscent of the guidance defects observed in UNC-6/Netrin and SLT-1/Slit pathway mutants (Desai et al., 1988; Wadsworth et al., 1996; Zallen et al., 1998; Hao et al., 2001; Adler et al., 2006). For unc-86(e1416), there was a statistically significant preference for extending in the anterior versus posterior direction (p=0.008, two-sample z-test; Fig. 2B), a preference that was not observed in unc-86(n846) (p=0.098). The majority of mutant HSNLs extended only a single axon, although a small percentage grew a second short process of less than or equal to three times the HSN longitudinal diameter [9.8% unc-86(n846) (n=51; Standard Error of Proportion (S.E.P.)=4.5%) and 5.3% unc-86(e1416) (n=57; S.E.P.=3.5%)]. In both mutants, the HSN cell bodies were also dorsally displaced relative to the wild-type position in the L4 (Figs. 1, 2C), reflecting a failure in ventral cell migration.

Fig. 2. UNC-86 is required for the ventral growth of HSN axons.

A: The proportion of left HSNs with detectable axon outgrowth, visualized with the lin-4∷GFP(zaIs1) reporter strain. n>50 for each genotype. B: The percentage of HSNL axons scored in (A) extending in the direction indicated, and representing the first direction in which axons extended a distance of 1.5 anterior/posterior cell diameters. ***: p<0.0001 when comparing proportions between wild-type and the indicated mutants, calculated using the two-sample z-test. For (A) and (B), error bars represent standard error of proportion. C: Representative images of a single unc-86(n846) (upper two panels) or unc-86(e1416) (bottom panel) L4-stage HSN with ventral guidance defects. Cell somas were often directly adjacent to the excretory canal, a more dorsal position than in wild-type (top panels; data not shown). For unc-86(n846), images of the same animal were acquired using a GFP bandpass filter (upper image) or DIC (merged with GFP, lower image). Arrows point to the excretory canal, just ventral to the CAN neuron. Arrowheads designate the HSN axon paths, the scale bars represent 5 μm, and anterior is to the left and ventral is down. GFP images for unc-86(n846) were pseudocolored green.

The left HSN was initially chosen for these experiments because it has previously served as a model for postembryonic HSN differentiation events, including axon guidance and synapse formation (Wightman et al., 1997; Shen and Bargmann, 2003; Shen et al., 2004). Moreover, HSNL axons grow anteriorly as part of a much smaller axon bundle than seen for the right HSN (HSNR) (White et al., 1986; Garriga et al., 1993), potentially simplifying studies of HSN growth cone migrations. However, when we investigated HSNR axon guidance in unc-86 mutants, we observed the same penetrant ventral growth defects as those seen for HSNLs, with 95.7% (S.E.P.=6.05) of unc-86(n846) and 95.2% (S.E.P.=4.15) of unc-86(e1416) HSNRs initiating outgrowth laterally. For this reason, we opted to collect and pool HSNR and HSNL data for experiments described in Figures 3-6.

Fig. 3. unc-86 controls the timing and guidance of HSN axon outgrowth.

A-B: Representative wild-type HSN (A) or unc-86(n846) (B) expressing the unc-86∷myr-GFP reporter at the L1 stage. No axon has extended in wild-type, while unc-86(n846) exhibits precocious outgrowth. For (A) and (B), scale bars represent 5 μm, and anterior is to the left and ventral is down. C-D: The proportion of wild-type and unc-86 mutant animals with axon outgrowth in mixed L1 populations using unc-86∷myr-GFP (myr-GFP) (C) or lin-4∷GFP (GFP) (D) reporter strains. n>50 for all genotypes. E: The proportion of HSN axons scored in (C-D) with extension along the anterior/posterior (A/P) or dorsal/ventral (D/V) axes for the indicated genetic backgrounds. Assigned axes represent the first direction in which axons extended a distance of 1.5 anterior/posterior cell diameters. For C-E, error bars represent standard error of proportion, and ***: p<0.0001 for the difference between wild-type and unc-86(n846) (C-D) or A/P and D/V axon outgrowth (E), calculated using the two-sample z-test.

Fig. 6. unc-86(n846) suppresses the delay in HSN axon outgrowth observed in loss-of-function mutants for the lin-4 microRNA.

A: The percentage of unc-86(n846); lin-4(e912) animals that extended axons during the L1 stage. Axons were visualized with the unc-86∷myr-GFP reporter (designated myr-GFP) and lin-4∷GFP (designated GFP). Error bars represent standard error of proportion. n>50 B: GFP expression in the HSNs normalized to L1-stage values for two unc-86 transcriptional fusion strains: unc-86∷myr-GFP containing the unc-54 3’UTR and unc-86∷GFP Line 2 containing the endogenous unc-86 3’UTR. No statistically significant differences in relative mean pixel intensity were observed between the two strains at the L2, L3, and L4 larval stages. C: Expression of the lin-14∷GFP(zaIs2) reporter strain in a wild-type or unc-86(n846) mutant background at the L1 and L2 larval stages. The mean pixel intensity values were not statistically different for wild-type and unc-86(n846) at both stages, but significant down-regulation between L1 and L2 was observed for each strain (**: p=0.002). D: Expression of lin-28∷GFP Line 10-2 in a wild-type or unc-86(n846) mutant background. While no difference was observed between wild-type and unc-86(n846) values at L1 or L3, each background exhibited significant down-regulation by the L3 (***: p<0.0001 for wild-type and p=0.0002 for unc-86(n846)). For A, n>50 for each genetic background. For B, C, and D, n≥10 for each time point and error bars represent standard error of the mean. p-values were calculated using the two-sample t-test.

unc-86 controls the timing of HSN axon outgrowth

While HSN axons initially failed to extend ventrally in both unc-86(n846) and unc-86(e1416) mutant strains, approximately one-third did display detectable ventral outgrowth at some point during their trajectories (unc-86(n846): 34.2% (n=73; S.E.P. 5.4%); unc-86(e1416): 29.2% (n=96; S.E.P. 4.6%)). This finding introduces the possibility that the pathfinding defects seen in unc-86 mutants do not stem from complete failures in ventral guidance pathway(s) per se, but may arise from a change in the location or timing of ventral signaling events relative to HSN axon initiation. In fact, when tracking HSN development in mixed-stage unc-86(n846) animals, we observed HSN axons extend prematurely in the L1, prior to the stage when the unc-6/netrin guidance response is normally activated in the HSNs (Adler et al., 2006).

To further characterize this timing defect, we scored the effects of the two unc-86 null mutations on HSN morphology in animals expressing GFP under the lin-4 or unc-86 promoters (Olsson-Carter and Slack, 2010);(Adler et al., 2006). Since unc-86 is constitutively expressed in the HSNs and lin-4 is up-regulated during the L1 (Finney and Ruvkun, 1990; Baumeister et al., 1996; Adler et al., 2006; Olsson-Carter and Slack, 2010), the unc-86∷myr-GFP strain enabled us to score cell morphology at earlier developmental time-points. For the purposes of this study, axons were defined as processes extending greater than or equal to the length of three HSN cell diameters (measured along the anterior/posterior axis) to distinguish them from immature neurites normally observed at this stage, and animals were scored as positive for outgrowth if at least one of the two HSNs extended an axon.

In mixed L1-stage populations, unc-86(lf) animals expressing the unc-86∷myr-GFP (Fig. 3A-C) or lin-4∷GFP (Fig. 3D) reporters exhibited precocious axon outgrowth relative to wild-type, and as was observed for the L4-stage mutants, these axons failed to extend ventrally (Fig. 3E). Since progressively longer HSN processes were also seen in subsequent larval stages (Fig. 4D; data not shown), it is unlikely that these L1-stage axons were retracted and later re-initiated at the L3/L4 transition.

Fig. 4. The UNC-40/DCC receptor is expressed and membrane-targeted in unc-86(n846).

A-B: The proportion of wild-type or unc-86(n846) animals with detectable unc-40∷SL2∷YFP expression in mixed populations at each larval stage (A: Line 1c; B: Line 6). Error bars represent standard error of proportion, and for each condition, n≥50 (A) or ≥25 (B). ***: p<0.0001. **: p=0.007. *: p<0.05 (0.043 for L1, 0.035 for L2, and 0.033 for L4). p-values were calculated using the two-sample z-test. C: For Line 6, representative L4-stage HSNs from wild-type (upper row) and unc-86(n846) (lower row) animals. Images of the same animal were acquired using the Yellow GFP (YFP) filter or DIC, and YFP-expressing cells were pseudocolored green. Left: YFP; Right: YFP/DIC merge. Arrowheads indicate the CAN neuron, while arrows refer to the HSN. Note that in unc-86(n846), the HSN cell body is dorsally displaced relative to the developing vulva, reflecting a failure in ventral migration. D: A strain expressing the UNC-40∷GFP translational fusion construct was crossed into unc-86(n846) and scored for membrane localization. Wild-type and unc-86(n846) L1-stage animals display UNC-40 membrane targeting with no preferential ventral localization (arrowhead). In the L2, UNC-40 has accumulated primarily at the ventral surface (arrowhead) in wild-type animals. In L2-stage unc-86(n846) mutants, UNC-40 is distributed throughout the precociously extending axon (arrowheads), with preferential accumulation in the ventral/posterior cell soma and a posterior membrane protrusion (arrow). For C and D, scale bars represent 5 μm, and anterior is to the left and ventral is down.

UNC-86 is not required for the expression and membrane localization of UNC-40

The UNC-40/DCC Netrin receptor is necessary for HSN ventral polarization, migration, and axon initiation (Desai et al., 1988; Wadsworth et al., 1996; Adler et al., 2006). In wild-type animals, unc-40 is transcriptionally up-regulated in the HSNs during the L1, the same stage that UNC-6/Netrin is detected in cells of the ventral nerve cord (VNC) (Desai et al., 1988; Chan et al., 1996; Wadsworth et al., 1996; Adler et al., 2006). Initially UNC-40/DCC is distributed throughout the HSN plasma membrane, but later in the L2 it accumulates primarily at the ventral surface, a shift in localization that is unc-6-dependent (Adler et al., 2006).

Like UNC-86, netrin signaling is necessary for ventral guidance but not growth of HSN axons (Desai et al., 1988; Wadsworth et al., 1996; Adler et al., 2006), and in mutants for either UNC-6/Netrin or its receptor UNC-40/DCC, axons often extend only along the anterior/posterior axis (Desai et al., 1988; Wadsworth et al., 1996; Adler et al., 2006). This suggested that UNC-86 may function by positively regulating the expression of unc-6 and/or unc-40. UNC-86 cannot serve as a direct transcriptional regulator of UNC-6/Netrin since unc-86 and unc-6 are not co-transcribed in the same cells during embryonic or larval development (Baumeister et al., 1996; Wadsworth et al., 1996). However, UNC-86 and UNC-40 are both present in the HSNs, and thus unc-86 could control ventral guidance by directly promoting unc-40 expression.

To address this possibility, we generated two independent lines that each expressed an unintegrated construct containing the endogenous unc-40 promoter fused to the unc-40 minigene, the SL2 trans-splice site, YFP and the unc-54 3’UTR (construct was a gift from D. Colón-Ramos (Colon-Ramos et al., 2007)). The unc-40 minigene was derived from both genomic and cDNA, and is able to rescue the gross unc and dpy phenotypes observed in unc-40(e271) mutants (D. Colón-Ramos, personal communication). unc-54 codes for a C. elegans muscle myosin class II heavy chain, and its 3’UTR is commonly used in transcriptional fusion constructs (Okkema et al., 1993). Since YFP was downstream of the SL2 site and under the control of the unc-54 3’UTR, its expression could only be used to mark unc-40 transcription initiation and not later post-transcriptional regulatory events.

In the HSNs, unc-40 transcription was not observed in wild-type or unc-86(n846) animals during embryogenesis, but was up-regulated in both genetic backgrounds in mixed-stage L1 animals (Fig. 4A-C). It was then maintained for the remainder of larval development (Fig. 4A-B), confirming that the unc-86(n846) ventral guidance defects cannot be explained by the loss of unc-40 transcription. Since the YFP construct was not stably integrated into the genome, differences in the proportion of HSNs with detectable YFP could reflect changes in the timing and/or level of transgene expression. We did not observe precocious YFP in unc-86(n846) animals, however, suggesting that the increase in YFP-expressing HSNs could be due to a release of unc-86-mediated inhibition.

Pathfinding errors can be caused by misregulation of guidance family members at both the transcriptional and post-transcriptional levels (Butler and Tear, 2007; Polleux et al., 2007; O’Donnell et al., 2009). Thus, the ventral guidance phenotypes in unc-86(n846) could also stem from errors in UNC-40 translation or cellular localization, leading to the inability of cells to detect the UNC-6 ventral signal. We tested this possibility in a strain expressing an UNC-40∷GFP translational fusion construct under the control of the unc-86 promoter (gift from C. Bargmann;(Adler et al., 2006)), which has previously been shown to rescue the unc-40(e271) axon guidance defects in the HSNs (Adler et al., 2006). In L1- and L2-stage unc-86(n846) animals, UNC-40∷GFP was expressed and targeted to the plasma membrane (Fig. 4D) and its membrane distribution in the L1 was comparable to that observed for wild-type (Fig. 4D). While it was difficult to directly compare L2-stage patterns of UNC-40∷GFP localization due to the markedly different morphologies of mutant and wild-type HSNs (Fig. 4D), we could conclude that unc-86(lf) outgrowth defects were unlikely to be due to general deficiencies in UNC-40 synthesis or membrane targeting.

Precocious axon outgrowth in unc-86(lf) is not dependent on signaling through the UNC-40/DCC and SAX-3/Robo guidance receptors

Netrin family members can guide as well as promote axon extension (Ishii et al., 1992; Serafini et al., 1994; Wadsworth et al., 1996; Adler et al., 2006), and thus the pathfinding and premature growth phenotypes observed in unc-86 mutants could result from ectopic activation of the Netrin pathway. These defects could also stem from the possible up-regulation of unc-40 in unc-86(n846) HSNs (Fig. 4A-B), which may effectively enhance sensitivity to weaker anterior/posterior netrin sources present during early larval development (Wadsworth et al., 1996; Levy-Strumpf and Culotti, 2007). However, when we scored double mutants lacking unc-86 and either unc-6/netrin or unc-40/DCC, we found that HSNs continued to initiate outgrowth along the anterior/posterior axis at the L1 stage (Fig. 5A-B), confirming that the unc-86(lf) premature growth phenotype is not dependent on netrin signaling.

Fig. 5. Precocious HSN axon initiation in unc-86(lf) is not dependent on signaling through the UNC-40/DCC or SAX-3/Robo guidance receptors.

A: The proportion of L1-stage HSN axon outgrowth in unc-86(n846) animals that also contained a mutation in either unc-6/netrin or its receptor unc-40/DCC. n≥50 for all three genotypes. B: The proportion of axons in (A) with initial outgrowth along the A/P or D/V axes for the genotypes indicated. C: The proportion of L1-stage HSN axon outgrowth in unc-86(n846); sax-3(ky123) double mutant animals (n=30). D: The proportion of HSN axons in (C) with initial extension along the A/P versus D/V axes. For (A)-(D), axons were visualized using the unc-86∷myr-GFP reporter strain, and error bars represent standard error of proportion. For (B) and (D), assigned axes reflect the first direction in which axons extended a distance of 1.5 anterior/posterior cell diameters. ***: p<0.0001 for the difference in the percentage of HSNs extending axons along each axis, calculated using the two-sample z-test.

A second guidance pathway containing the SLT-1/Slit ligand and SAX-3/Robo receptor has also been shown to promote ventral growth of HSN axons, with approximately one-third of HSNs extending only laterally in the putative null mutant, sax-3(ky123) (Zallen et al., 1998; Hao et al., 2001). This pathway is also important embryonically in guiding the posterior cellular migrations of the two CAN neurons to their positions just anterior to the HSNs (Hao et al., 2001), introducing the possibility that signaling through the SAX-3/Robo receptor could also promote and guide early larval HSN axonal migrations along the anterior/posterior axis in unc-86(n846) animals. Yet in unc-86(n846); sax-3(ky123) double mutants, we continued to observe both the premature L1-stage outgrowth (Fig. 5C) and failed ventral pathfinding (Fig. 5D) defects seen in the HSNs of unc-86(lf) single mutants. We therefore concluded that the unc-86(lf) phenotypes are not caused by ectopic signaling through SAX-3/Robo.

Delayed axon growth in loss-of-function mutants for the lin-4 microRNA is suppressed by unc-86(lf)

Three members of the C. elegans heterochronic pathway, including the lin-4 miRNA and its targets, the LIN-14 transcription factor and LIN-28 RNA-binding protein, also control the timing of HSN axon outgrowth independent of the Netrin and Slit guidance pathways (Olsson-Carter and Slack, 2010). lin-4 is up-regulated in the HSNs during the L1, leading to the down-regulation of LIN-14 and LIN-28 during the L2 and L3 larval stages, respectively. When lin-14 and lin-28 expression persists in lin-4 loss-of-function mutants, axon extension is delayed in L4 and adult animals (Olsson-Carter and Slack, 2010). Conversely in lin-14(lf) or lin-28(lf), outgrowth occurs approximately one stage prematurely in the L3. To assess the interaction between unc-86 and the heterochronic pathway, we generated unc-86(n846); lin-4(e912) double mutants and scored for axon growth. Animals still extended axons precociously in the L1 (Fig. 6A), demonstrating that unc-86(n846) can suppress the lin-4(e912) retarded axon growth phenotype and that unc-86 functions either in parallel to or downstream of lin-4.

However, unc-86 is unlikely to be a direct lin-4 target. Not only did we fail to find candidate lin-4 binding sites in the unc-86 3’UTR using currently accepted parameters (Bartel, 2009), but expression of GFP reporter constructs containing either the unc-54 or unc-86 3’UTRs also displayed no significant differences in average pixel intensities at the L2, L3, and L4 stages relative to L1 values (Fig. 6B). This suggests that post-transcriptional regulation of unc-86 expression—at least that mediated through the 3’UTR—is minimal, consistent with the findings from antibody staining that UNC-86 protein is not down-regulated during development but is instead constitutively present after cell cycle exit (Finney and Ruvkun, 1990; Baumeister et al., 1996).

We next investigated whether unc-86 might affect expression patterns of the lin-4 targets lin-14 and lin-28. The lin-14∷GFP and lin-28∷GFP reporter strains described previously (Olsson-Carter and Slack, 2010) still displayed down-regulation at the L2 and L3 stages, respectively, in unc-86(n846) (Fig. 6C-D). In addition, mean pixel intensity values at these time points were not significantly different for wild-type and unc-86(n846), showing that there was no detectable change in lin-14 and lin-28 transcription in the absence of unc-86.

Discussion

We have shown that loss of the conserved POU domain gene unc-86 results in penetrant defects in timing of outgrowth and ventral guidance of HSN axons, and propose that the axon misroutings stem directly from a shift in the relative timing of growth versus guidance responses during early larval development. Ultimately, these pathfinding errors in combination with cell migration defects lead to devastating effects on HSN function, since processes are no longer positioned to synapse appropriately onto the vulval muscle cells (Figs. 2C, 4C) and stimulate egg-laying (Schafer, 2006).

unc-86 is a novel postmitotic neuronal heterochronic gene that functions independently of UNC-40/DCC and SAX-3/Robo

In Drosophila mitotic neuroblast (NB) lineages, the POU family members Pdm1/2 function as part of a transcriptional cascade that specifies the temporal identities of neuronal progeny (Pearson and Doe, 2004). In the work presented here, we have shown that the POU protein UNC-86 also controls developmental timing. In contrast to Drosophila Pdm1/2, however, unc-86 is not expressed in HSN precursor cells, is only up-regulated in the HSNs after cell cycle exit, and does not appear to affect the timing of early HSN differentiation events such as cell birth or anterior migration (Desai et al., 1988; Finney and Ruvkun, 1990; Baumeister et al., 1996).

unc-86 is also distinct from members of the C. elegans heterochronic pathway, the set of genes that controls the relative timing of stage-specific developmental events in the hypodermis and other tissues (Hallam and Jin, 1998; Johnson et al., 2005; Moss, 2007) Not only is unc-86 expressed specifically in neuronal lineages (Finney and Ruvkun, 1990; Baumeister et al., 1996), but its expression does not appear to be developmentally regulated based on immunostaining and analysis of transcriptional reporters (Fig. 6B;(Finney and Ruvkun, 1990; Baumeister et al., 1996)). Ultimately, it may function as part of a separate timing mechanism. The lin-4 miRNA and its targets lin-14 and lin-28 act as a developmental switch that controls HSN axon initiation (Olsson-Carter and Slack, 2010), and epistasis analysis reveals that unc-86 acts either down-stream of or in parallel to lin-4. However, reporter assays suggest that unc-86 is unlikely to be targeted by lin-4 directly, and the intensity and pattern of down-regulation of lin-14 and lin-28 are the same for unc-86(n846) as for wild-type.

Constitutive overexpression of lin-4 with the unc-86 promoter or loss-of-function mutations in lin-14 or lin-28 only shift outgrowth one larval stage earlier (Olsson-Carter and Slack, 2010), suggesting that other factor(s) must prevent premature axon growth in the L1 and L2. Here we have shown that unc-86 functions as one of these inhibitors. Yet because it continues to be expressed at the L3/L4 transition—the time when HSN axon initiation normally occurs in wild-type animals—additional factors must be required to release its inhibition of axon growth at this later stage of development. One gene that could play such a role is the zinc-finger transcription factor sem-4. sem-4 is also required for larval-stage HSN maturation (Desai et al., 1988; Basson and Horvitz, 1996), and has previously been shown to act with unc-86 and other transcriptional regulators to control development of the C. elegans touch cells (Mitani et al., 1993). However, precocious axon outgrowth was not observed in sem-4(n1378) animals (0% L1-stage outgrowth (n=32; S.E.P. 3.8%)), and axon initiation in the L4 occurred in the ventral direction, as seen in wild-type (data not shown).

Transcription factors like UNC-86 have traditionally been thought to direct the timing of axon outgrowth by modulating the expression of guidance cues and/or their receptors (Polleux et al., 2007). While we cannot exclude the possibility that unc-86 null mutants extend axons prematurely due to loss and/or ectopic expression of spatial patterning genes like the wnts (Zinovyeva et al., 2008), the observation that precocious L1-stage axon initiation does not require signaling through UNC-40/DCC or SAX-3/Robo, the two receptors known to be required for HSN ventral growth, is consistent with the independent regulation of axon growth and guidance in these neurons (Adler et al., 2006; Olsson-Carter and Slack, 2010). If this is indeed the case, the misroutings of unc-86(lf) HSN axons could simply be due to functional changes in these or other unidentified guidance systems. Alternatively, they could stem directly from the shift in timing of axon initiation relative to the establishment of netrin and/or slit signaling.

unc-86 is required for late, but not early, differentiation events in the HSNs

unc-86 is not detected in HSN precursor cells (Finney and Ruvkun, 1990) and appears to be only partially necessary for HSN specification (Desai et al., 1988). In addition to exhibiting normal anterior migration relative to the CAN neurons by the L1 stage (Desai et al., 1988), unc-86(lf) HSNs express a number of genes known to be up-regulated during HSN development, including the lin-4 miRNA (Figs. 1-2, 6)(Olsson-Carter and Slack, 2010) and the unc-40/DCC netrin receptor (Fig. 4C). These observations coupled with the previous finding that unc-86 is required for serotonin synthesis in the HSNs (Desai et al., 1988; Sze et al., 2002) are consistent with a role for unc-86 in late postmitotic maturation (Desai et al., 1988). It also likely functions cell-autonomously, since it is not expressed in cells neighboring the HSNs or in neurons of the ventral cord (Finney and Ruvkun, 1990; Baumeister et al., 1996). In fact, previous studies have employed the unc-86 promoter to test cell autonomy and/or drive gene expression in the HSNs (Wu et al., 2001; Gitai et al., 2003; Shen and Bargmann, 2003; Adler et al., 2006; Olsson-Carter and Slack, 2010).

However, since unc-86 is known to be broadly important in C. elegans neuronal lineage and identity specification (Finney et al., 1988; Finney and Ruvkun, 1990; Baumeister et al., 1996; Duggan et al., 1998), the possibility cannot be excluded that unc-86(lf) mutant phenotypes reflect a transformation in cell fate. HSN identities could be shifted to those of the two chemosensory PHB phasmid neurons, since they are descendents of the same progenitor cell as the HSNs but do not express unc-86 (Finney and Ruvkun, 1990; Hall and Altun, 2008). Yet while unc-86(lf) mutant HSNs migrate anteriorly from their place of birth in the tail, extend a single long process, and fail to produce the serotonin neurotransmitter in the adult (Desai et al., 1988; Sze et al., 2002), the PHBs remain in the posterior of the animal, extend axons during embryogenesis, and have been reported to synthesize serotonin (Sawin et al., 2000; Hall and Altun, 2008). Other potential candidates are the two PVQ interneurons, which are also present in the HSN lineage and fail to express unc-86 (Finney and Ruvkun, 1990; Hall and Altun, 2008). However, these cells are also positioned in the tail and complete axon outgrowth by the beginning of L1 (Hall and Altun, 2008), earlier than observed in unc-86(lf) HSNs.

An alternative possibility is that failures in the execution of later events in HSN development—such as serotonin synthesis—stem simply from defects in earlier maturation steps, including axon guidance. However, previous evidence suggests that the HSN axon misroutings in unc-86(lf) animals are not necessarily caused by the absence of ventral cell migration, which normally occurs just prior to axon initiation at the L3 stage (Desai et al., 1988; Adler et al., 2006). In loss-of-function mutants for the ENA/VASP protein UNC-34 as well as several other genes, the HSN cell bodies fail to migrate properly but still exhibit ventral outgrowth (Desai et al., 1988; Adler et al., 2006). On the other hand, axon extension prior to ventral cell migration could alter the competence of the HSNs to migrate normally in the L3 (Desai et al., 1988; Adler et al., 2006). Currently, the factors affecting the decision to migrate versus initiate axon growth are not well understood in the HSNs or other neurons (Yu and Bargmann, 2001; da Silva and Dotti, 2002).

MODEL: unc-86 guides axons by controlling the relative timing of axon growth and activation of ventral guidance pathways

During the L1, wild-type HSNs break symmetry in response to the ventral UNC-6/Netrin source, but they have not yet exhibited later features of netrin signaling required for axon guidance, including cytoplasmic thickening and redistribution of UNC-40 at the ventral surface (Adler et al., 2006). unc-86(lf) mutants do not exhibit evidence of precocious unc-40 expression or membrane localization (Fig. 4A-B, 4D), suggesting that they initiate HSN axon outgrowth before the netrin response is fully activated in the L2. The same may be true for signaling through SAX-3/Robo, since the receptor is normally not up-regulated in the HSNs until the L2 larval stage (Zallen et al., 1998).

Similar to the findings presented here, loss of the murine UNC-86 homolog Brn-3.2 leads to defects in retinal ganglion cell (RGC) axon guidance and optic nerve morphology that are reminiscent of netrin signaling mutants without markedly affecting the expression of Netrin-1 or its receptor, DCC (Erkman et al., 2000). An actin-binding protein, AbLim, was identified through differential expression analysis as a potential Brn-3.2 target, providing a direct link between Brn-3.2 and genes involved in the cytoskeletal rearrangements required for axonal elongation (Erkman et al., 2000). Interestingly, the C. elegans AbLim homolog, UNC-115, not only promotes axon outgrowth but also acts downstream of UNC-40/DCC (Lundquist et al., 1998; Gitai et al., 2003), suggesting that UNC-86 may control HSN axon initiation at least in part by inhibiting unc-115 expression until the appropriate developmental stage. Other candidates for downstream targets of unc-86 include the Enabled homolog unc-34 (Gitai et al., 2003; Krause et al., 2003; Lebrand et al., 2004), the Rac GTPase ced-10 (Gitai et al., 2003), and the actin regulator mig-10/lamellipodin (Manser et al., 1997; Krause et al., 2004; Adler et al., 2006). unc-34, ced-10, and mig-10 have all been shown to function as effectors in the netrin and slit guidance pathways (Yu et al., 2002; Gitai et al., 2003; Adler et al., 2006; Chang et al., 2006; Quinn et al., 2006).

Taken together, these data support a model in which unc-86 cell-autonomously inhibits cytoskeletal factors required for axon growth, and loss of inhibition in unc-86(lf) leads to premature axon initiation. Moreover, unc-86(lf) HSN axons are likely misrouted along the longitudinal axis because processes extend at a stage when the Netrin and Slit signaling pathways are not fully active. In summary, we propose that the unc-86(lf) HSN pathfinding defects stem directly from a shift in the timing of outgrowth relative to ventral guidance responses. Since the unc-86 family is not only important for C. elegans axon development but also plays key roles in axon pathfinding in Drosophila and higher animals, this model could represent a novel and conserved mechanism for generating complex neural networks using a limited set of guidance molecules.

Experimental Procedures

Nematode Strains

Animals were cultured at 20 °C using standard methods (Brenner, 1974). The wild-type strain was N2 Bristol, mutant alleles included unc-40(e271), unc-40(e1430) (marked with dpy-5(e61)), unc-86(e1416), unc-86(n846), unc-6(ev400), sax-3(ky123), and lin-4(e912). Previously described reporters were lin-4∷GFP(zaIs1) (Boehm and Slack, 2005), lin-14∷GFP(zaIs2) (Olsson-Carter and Slack, 2010), lin-28∷GFP Line 10-2 (Olsson-Carter and Slack, 2010), and unc-86∷myr-GFP(kyIs262) and unc-86∷unc-40∷GFP(kyEx1212) (Adler et al., 2006, gifts from C. Bargmann). The independent unc-40∷SL1∷YFP Lines 1c and 6 were generated by injecting 10 ng/μl unc-40 minigene∷SL2∷YFP (construct was a gift from D. Colón-Ramos; (Colon-Ramos et al., 2007)) and 5 ng/μl of the myo-2∷dsRed2 injection marker into young adult N2 animals as previously described (Mello and Fire, 1995). For Line 6, 85 ng/μl EcoR1-cut pBluescript SK was also added to minimize toxicity associated with the co-injected constructs. unc-86∷GFP Line 2 containing the unc-86 3’UTR was produced using an injection mix containing 1 ng/μl of the unc-86∷GFP∷unc-86 transcriptional fusion construct, 5 ng/ul myo-2∷dsRed2 marker (Olsson-Carter and Slack, 2010), and 44 ng/μl of EcoR1-cut pBluescript SK. For unc-40∷SL1∷YFP Line 6 and unc-86∷GFP∷unc-86 Line 2, EcoR1-cut pBluescript was column-purified using the Qiagen QIAquick PCR purification kit.

Generating the unc-86∷GFP∷unc-86 construct

The unc-86 3’UTR was PCR-amplified from N2 genomic DNA using the Novagen Hot Start DNA polymerase, a 50°C annealing temperature, and U86UTRFOR (5’-GAATTCTTTCGTTTTCGTGAACACATTTTC) and U86UTRREV (ACTAGTGAATTATATTTATTTTACGAGGCAT) primers flanked by EcoR1 and Spe1 restriction sites, respectively. The PCR product was then inserted into pCR2.1-TOPO using the Invitrogen TOPO TA Cloning Kit. A pCR2.1-TOPO clone containing the sequence-verified unc-86 3’UTR was digested with EcoR1 and Spe1, and the resulting UTR fragment was ligated to the EcoR1/Spe1-cut pPD95.67 GFP expression vector backbone. To generate unc-86∷GFP∷unc-86, the pSM-86 construct (gift from C. Bargmann, Rockefeller University, New York, NY (Gitai et al., 2003);(Adler et al., 2006)) was digested with Sph1 and Xma1, and the 5.1 kilobase promoter fragment (Baumeister et al., 1996) was inserted into the Sph1/Xma-cut pPD95.67 construct containing the unc-86 3’UTR. All digested DNA was gel-purified using the Qiagen QIAquick Gel Extraction Kit, and ligations were performed with NEB T4 DNA ligase.

Microscopy

Animals were immobilized with 5 mM levamisol on 2% agarose pads (Figs. 2-6) or directly on a slide (Fig. 1), and scored on the Zeiss Axioplan 2 using the Endow (Chroma #41017; Figs. 1-3, 4D, 5-6) and/or Yellow (Chroma #41028; Fig. 4A-C) GFP bandpass filter sets. Images were acquired with the Zeiss Plan-Apochromat 100x/1.4 N.A. objective, AxioCam MRm camera and the automated AxioVision (v. 4.6) multidimensional acquisition module, and they were processed using AxioVision (v. 4.6) or Adobe Photoshop CS3 software.

For Fig. 6B-D, fluorescence data were acquired after HSNs were first identified by DIC. Average pixel intensity values for these cells were determined using the Zeiss AxioVision AutoMeasure module, and image acquisition and processing parameters were identical for all animals expressing the same reporter construct. The HSNs were selected primarily based on pixel intensity, but separation lines were drawn when necessary. Pixel intensity values of axonal processes were not included in our analyses. For unc-86∷GFP Line 2, HSNs which failed to express detectable transgene due to genetic mosaicism were also excluded. The two-sample t-test was used to determine the significance of the difference between mean values.

Staging and scoring the HSNs

Staging was performed as described previously (Ambros and Horvitz, 1987; Olsson-Carter and Slack, 2010). HSNs were identified based on overall morphology, position relative to other cells and anatomical structures, and expression of cell-specific markers (White et al., 1986), and if cells could not be conclusively identified using these criteria, they were not scored. To distinguish axons from more immature neurites, an HSN process was scored as an axon if it had extended a distance of at least three times the anterior/posterior cell diameter. Unless indicated otherwise, animals were considered positive for axon outgrowth or reporter gene expression if the phenotype was observed in at least one of the two HSNs. In Figs. 2B, 3E, 5B, and 5D, initial axon guidance was assigned by determining the direction in which axons first extended at least 1.5 anterior/posterior cell diameters. When neurons extended more than one process or a branched axon, the direction of extension of the longest process or branch was scored. In the rare case where a neuron extended multiple processes or branches of the same length, it was excluded from analysis.

For lin-4∷GFP experiments, HSNs were more clearly visualized by preferentially knocking down GFP in non-neuronal cells (Simmer et al., 2002) using the GFP L4440 RNAi feeding vector (gift from A. Fire) and previously established methods (Timmons and Fire, 1998). The standard error of proportion was calculated for all proportional data, and the significance of the difference between two proportions was determined using the two-sample z-test.