A Farnesyltransferase Acts to Inhibit Ectopic Neurite Formation in C. elegans

Genetic pathways that regulate nascent neurite formation play a critical role in neuronal morphogenesis. The core planar cell polarity components VANG-1/Van Gogh and PRKL-1/Prickle are involved in blocking inappropriate neurite formation in a subset of motor neurons in C. elegans. A genetic screen for mutants that display supernumerary neurites was performed to identify additional factors involved in this process. This screen identified mutations in fntb-1, the β subunit of farnesyltransferase. We show that fntb-1 is expressed in neurons and acts cell-autonomously to regulate neurite formation. Prickle proteins are known to be post-translationally modified by farnesylation at their C-terminal CAAX motifs. We show that PRKL-1 can be recruited to the plasma membrane in both a CAAX-dependent and CAAX-independent manner but that PRKL-1 can only inhibit neurite formation in a CAAX-dependent manner.


Introduction
Initial neurite formation and extension is a key stage in the transition from neuroblast to mature neuron.Cell intrinsic factors that orchestrate the localized regulation of actin and microtubules determine the site of nascent neurite outgrowth [1][2][3][4].In the complex environments existing in vivo, these factors are polarized at sites of neurite outgrowth in response to extracellular cues [5,6].However, in addition to neurite promoting factors, mechanisms that inhibit neurite formation help to sculpt neuronal connections by blocking inappropriate neurites during development and by maintaining neuronal morphology as neurons age [7,8].The genetic pathways involved in inhibiting neurite formation remain poorly understood.
The Frizzled/Planar Cell Polarity (Fz/PCP) pathway, a non-canonical Wnt pathway, is involved in mediating polarized cytoskeletal rearrangements that underlie several developmental processes including ordered cell movements during gastrulation, neuronal migration and axon guidance [9,10].In C. elegans, a PCP-like pathway that includes the core PCP proteins PRKL-1/Prickle and VANG-1/Van Gogh acts in a subset of motor neurons to block neurite formation along the anterior-posterior (AP) axis in order to restrict neurite outgrowth to an orthologous axis [11].PRKL-1-mediated neurite inhibition, at least in part, appears to involve localization to the plasma membrane of neuronal somas.However, other components of this pathway, in particular those which act in membrane targeting remain to be discovered.
Herein, we present the findings of a genetic screen for mutations that, like those in prkl-1, fail to block inappropriate neurite formation from neuronal somas.In addition to new prkl-1 and vang-1 alleles, this screen identified mutations in fntb-1, the β subunit of farnesyltransferase (FTase).FTases are cytoplasmic proteins that transfer a prenyl (farnesyl) group to C-terminal CAAX-motif containing proteins [12].Prenylation is an important post-translational modification that mediates protein insertion into cellular membranes.We show that FNTB-1 acts cell autonomously in motor neurons and that the fntb-1 mutations recovered in our screen are predicted to disrupt FTase activity.CAAX-containing Prickle proteins are well established as evolutionary conserved targets of FTases [13,14].We show that the PRKL-1 CAAX motif is required to block ectopic neurite formation and that PRKL-1 is targeted to the plasma membrane of neuronal somas in both a CAAX-dependent and CAAX-independent VANG-1-dependent manner.

Genetic screen for VC4 and VC5 neurite defects
Worms carrying the VC4 and VC5 reporter transgene cyIs4[Pcat-1::GFP pRF4(rol-6)] [15] were mutagenized with 50mM ethylmethanesulfonate (EMS) as described by Brenner [16].Young adult F1 progeny of these worms were then transferred to freshly seeded plates (5 F1 worms/plate) and allowed to self-propagate.30-40 young adult roller progeny from each F1 plate were then transferred in a drop of M9 to slides prepared with 2% agarose pads and covered with glass cover slips.These worms were visually screened for AP-directed ectopic VC4 and VC5 neurites under 20x magnification on an AxioplanII fluorescence microscope.The roller background facilitated the visualization of VC4 and VC5 neurites on the ventral side and ensured that worms, embedded in the agarose pad, were immobilized.Worms displaying a VC neurite defect were recovered from the slide by carefully sliding off the cover slip and transferring individual worms to fresh plates to self-propagate.Potential mutants were rescreened to authenticate the outgrowth defect and then outcrossed at least twice before further characterization.

VC4 and VC5 morphology and PRKL-1 localization
VC4 and VC5 morphology was visualized in young adult hermaphrodites using the cyIs4 [Pcat-1::GFP] reporter.Worms were immobilized with 10 mM levamisole (Sigma) and imaged using an AxioplanII fluorescence microscope.An ectopic VC neurite was defined as a protrusion from the cell body that was greater than the length of one VC cell body (~5 μm).The relevance of the CAAX motif for plasma membrane localization in VC4 and VC5 neurons was assessed by quantifying the localization of full length and CAAX-deleted GFP::PRKL-1 (PRKL-1::ΔCTVS) in wt and a vang-1(tm1422) null backgrounds.The unc-4 promoter was used to express GFP::PRKL-1 in VC neurons.VC4 and VC5 were scored during early L4 and identified by unc-4p::GFP::PRKL-1 fluorescence and their stereotypical positions flanking the vulva.Plasma membrane localization was quantified by binning observations into two primary categories: (1) 10 or more membrane punctae (many puncta) and (2) fewer than 10 membrane punctae (few puncta).

Results and Discussion
A forward genetic screen identifies five genes required to block VC neurite outgrowth along the AP axis Egg-laying in C. elegans is mediated by a circuit consisting of the VC and HSN motor neurons and the vulval sex muscles [19].The VC neurons are a set of six motor neurons positioned along the AP axis with VC4 and VC5 flanking the vulva (Fig 1A).These neurons display stereotypical differences in axon extension along either an AP or left-right (LR) trajectory during formation of the 'egg-laying' circuit [20] and thus provide a morphologically simple and genetically accessible model to investigate genetic programs involved in neurite emergence.We have previously shown that a PCP-like pathway that includes vang-1, prkl-1 and dsh-1 is required in VC4 and VC5 to maintain neuronal morphology [11].Normally, VC4 and VC5 extend two neurites along a circular path around the vulva.In PCP mutants, VC4 and VC5 will often extend an additional neurite along the AP axis directed away from the vulva resulting in distinct tripolar-like morphologies (Fig 1B and 1C).To identify new components involved in suppressing VC neurite outgrowth, we performed a forward genetic screen for neurite outgrowth defective (nde) mutants (Fig 1D ).
A visual screen of the F2 progeny of approximately 6000 EMS mutagenized cyIs4[cat-1p:: gfp] worms identified 14 mutants in which VC4 and VC5 extended neurites inappropriately along the AP axis.Since our previous study showed that vang-1, prkl-1 and dsh-1 act to block ectopic neurite formation [11], we performed standard genetic complementation tests to determine if any of these genes were identified in our screen (data not shown).This analysis followed by sequencing revealed three new alleles each of vang-1 and prkl-1 (nde-1 and nde-2 respectively) and a single new allele of dsh-1 (nde-3) (Tables 1 and 2).Sequencing did not identify a polymorphism in the prkl-1 coding region of nde-2(zy17) despite zy17 failing to complement prkl-1(ok3182) and mapping to a small interval on LGIV that included the prkl-1 locus (data not shown).It is possible that zy17 represents a mutation in prkl-1 regulatory sequence or a complex rearrangement of the prkl-1 locus.All vang-1, prkl-1 and dsh-1 mutants displayed highly penetrant morphology defects consistent with previous findings (Fig 1E).
The seven remaining mutants were assigned to two complementation groups, nde-4 (5 alleles) and nde-5 (2 alleles) (Table 1).These mutants were mapped using single nucleotide polymorphisms (SNPs) between Hawaiian and Bristol (N2) strains [21] to chromosomes V and I respectively (S2 Table ).The penetrance of ectopic VC neurites in nde-4 mutants were similar in severity to those of prkl-1 mutants (Fig 1E).In prkl-1(ok3182) null mutants, 94% of VC4 and VC5 neurons displayed ectopic neurites compared to 90% in nde-4(zy23).nde-5 mutants displayed less severe defects (approximately 30% ectopic neurites) compared to the others (Fig 1E).Most mutants showed qualitatively similar VC4 and VC5 morphologies (Fig 1C and 1F) except for nde-5 mutants which often displayed multiple neurites extending from cell somas (Fig 1G).nde-4 mutants also displayed low brood sizes, egg retention and small body size.The recovery of multiple alleles for most complementation groups suggests that the screen may have approached saturation for viable mutants.

A farnesyltransferase is involved in VC neurite formation
Because nde-4 mutants displayed highly penetrant morphology defects and represented the largest complementation group with 5 alleles, we sought to identify the gene product to gain further insight into how VC neurite extension is regulated.While outcrossing we found that nde-4 mutations could not be separated from our cyIs4 reporter indicating a linkage to chromosome V.We subsequently mapped nde-4 to a roughly 2 Mb region on chromosome V between SNPs pkP5070 (AH10) and pkP5129 (F57G8) (S2 Table ).A systematic search of genes within this region using the genome browser at Wormbase.org(version WS249) revealed fntb-1(F23B12.6), the β subunit of farnesyltransferase (FTase), as a possible candidate (Fig 2A ).A genome-wide analysis of trans-spliced mRNAs identified fntb-1 as the upstream gene in an operon with F23B12.7 [22].fntb-1 was flagged as a good candidate for nde-4 as PRKL-1 contains a carboxy-terminal prenylation signal (CAAX motif) and Prickle proteins are known to be post-translationally modified by FTases [13,14].We confirmed fntb-1 as the nde-4 locus by showing that a 2.5 kb fntb-1 genomic fragment containing 485 bp upstream of the ATG start and 332 bp downstream of the stop codon rescued nde-4(zy7) VC defects (Fig 2B ) and by identifying single nucleotide changes in all five alleles (Table 2).nde-4 will henceforth be referred to as fntb-1.

FNTB-1 mutants are predicted to disrupt FTase catalytic activity
FTases transfer a 15-carbon prenyl (farnesyl) group from farnesyl diphosphate (FPP) to CAAX-motif containing proteins.They function as heterodimers consisting of a regulatory α subunit (FNTA) and catalytic β subunit (FNTB) that contains the substrate and FPP-binding pockets [12].FNTA-1 and FNTB-1 are the sole worm orthologues of the alpha and beta-subunits respectively.In this study, we will focus primarily on the beta-subunit.Compared to human FNTB, C. elegans FNTB-1 shows strong sequence conservation (46% amino acid identity, 60% amino acid similarity) across its entire length (S1 Fig).
Of the five fntb-1 mutants identified in our screen, four are missense mutations (zy7, zy20, zy23, zy24) that alter conserved residues and one is a nonsense mutant (zy21) that introduces a stop codon at arginine 305 (S1 Fig) .Interestingly, zy21 animals are viable despite the existence of a deletion allele, fntb-1(ok590), generated by the C. elegans Gene Knockout Consortium that is not viable.To understand how fntb-1 mutations might disrupt protein function we used the crystal structure of human FTase (PDB ID: 1S63) [18] to map the location of mutations in the highly homologous worm FTase (Fig 3A).
This analysis suggests that the FNTB-1 truncated protein generated in zy21 would lack residues and secondary structure that form the FPP-binding catalytic domain and thus should completely lack enzymatic activity.How to explain the non-viability of ok589?A likely explanation is that because the ok590 deletion removes approximately 100bp upstream of the fntb-1/ F23B12.7 operon start in addition to fntb-1 coding region, it might also disrupt the expression of the co-transcribed F23B12.7 gene.The non-viability of ok590 may thus result from loss of F23B12.7 singly or in combination with fntb-1.In contrast to fntb-1 mutants, ok269, a deletion in fnta-1, is non-viable.FNTA is a shared subunit in both FTase and another prenyltransferase, geranylgeranyltransferase (GGTase), and may therefore participate in the post-translational modification of a greater complement of substrates, some of which may be essential for viability [12].
How might the other conserved residues disrupted in our fntb-1 mutants affect FTase function?The G238E substitution in zy23 maps to G250 in human FNTB.This residue is associated with the active site.Any substitution at position G238 would result in an elongation of the side chain and therefore interfere with the productive binding of the FPP lipid substrate (Fig 3B).zy24 (C92Y) maps to human FNTB residue C104.A tyrosine substitution at this position is predicted to compromise the packing and stability of the enzyme as any tyrosine rotomer would result in a steric clash with either the alpha-helices of FNTB or a loop of FNTA (Fig 3C).VC morphology defects in zy23 and zy24 are quantitatively similar to the zy21 null, suggesting that they may also completely disrupt protein function.zy7 (E321K) and zy20 (S289N) substitutions map to residues Q336 and S301 in human FNTB.The consequences for substitutions at these residues are less clear but they may also compromise protein stability by disrupting alpha helices.It is important to note however that in the absence of empirical data, this analysis is speculative.

FNTB-1 acts in VC4 and VC5
To determine where fntb-1 is expressed during the period of VC neurite outgrowth we made GFP promoter and protein fusions.We showed that a genomic fragment containing 485 bp of fntb-1 promoter sequence was able to rescue VC morphology defects in fntb-1(-/-) mutants.This promoter was also able to rescue VC defects when used to drive fntb-1 cDNA expression (Fig 2A and 2B) suggesting that the regulatory elements contained therein are sufficient to properly express fntb-1 and therefore suitable for GFP reporters.In L4 animals when VC neurite outgrowth is taking place, fntb-1 promoter activity was found in VC neurons and vulval cells (Fig 4A and 4D).This expression pattern is similar to that of prkl-1 at the same stage [11].We previously showed that prkl-1 acts cell autonomously to block ectopic VC neurite outgrowth [11].We therefore asked if fntb-1 was required in VC neurons as would be expected if PRKL-1 was a target of FTase activity.The unc-4 promoter was used to drive fntb-1 expression in a subset of neurons including the VC neurons.We found that unc-4p::FNTB-1 expression was able to restore proper VC morphology in fntb-1(-/-) mutants indicating that fntb-1 acts cell autonomously to promote proper VC morphology (Fig 2B).In addition to VC neurons, fntb-1 was expressed in motor neurons in the ventral nerve cord, mid-body region neurons (HSNs and CANs), and various neurons in the head and tail (Fig 4A -4C).In non-neural tissue, fntb-1 expression was observed in body wall and pharyngeal muscle and rectal epithelial cells (Fig 4A -4C).To determine where FNTB-1 accumulates in VC neurons, we generated transgenic animals carrying an fntb-1 cDNA fused in frame to GFP and driven from the fntb-1 promoter.Expression from the Pfntb-1::FNTB-1::GFP transgene was able to rescue fntb-1 VC defects (Fig 2B ) suggesting that the C-terminal GFP tag does not interfere with FNTB-1 function.In VC neurons, FNTB-1::GFP protein accumulated in the cell soma and along axons but was excluded from the nucleus as expected for a cytoplasmic protein (Fig 4E).

Farnesylation is important for PRKL-1-mediated inhibition of ectopic neurites in VC neurons
A farnesyl group is covalently attached to C-terminal CAAX motif containing substrates (Fig 5A).Prickle orthologues with mutated or deleted CAAX motifs are not farnesylated [13,14].To determine the functional importance of the PRKL-1 CAAX motif, CTVS, we compared the ability of unc-4 promoter-driven full length PRKL-1 and PRKL-1::ΔCTVS to restore proper VC morphology in prkl-1 mutants.We found that PRKL-1::ΔCTVS expression was far less effective at rescuing ectopic VC neurites in prkl-1 mutants compared to full length PRKL-1 (Fig 5B ) indicating that the CAAX signal is important.The mild but significant decrease in ectopic neurites when PRKL-1::ΔCTVS was expressed in prkl-1(-/-) mutants may be the result of overexpression from the multi-copy transgenic array.Alternatively, functional non-farnesylated PRKL-1 may normally play some role in blocking ectopic neurite formation.
We next sought to determine the effect of replacing the PRKL-1 CAAX motif with a heterologous prenylation signal.We reasoned that if the PRKL-1 CTVS motif acts as a prenylation signal, then substitution with a CAAX motif from a heterologous protein should not interfere with PRKL-1 function.The Ras and Rho family of small GTPases are well characterized prenylation substrates for FTases [23].We therefore substituted the C-terminal 15 amino acids of PRKL (VRMAKKKKSSRCTVS) with the corresponding region (LHPKPQKKKKSCNIM) from MIG-2, a C. elegans Rho GTPase [24].We swapped an extended C-terminal region as CAAX motifs are usually preceded by a poly-basic region that is important for prenylation [25].We found that expression of PRKL-1::CNIM is largely indistinguishable from full length PRKL-1 in rescuing prkl-1 VC morphology defects (Fig 5B).Combined, CAAX deletion and substitution support the notion that the CAAX motif is important for PRKL-1 to maintain normal VC morphology.
Previous findings showed that PRKL-1 overexpression is able to partially rescue the ectopic VC neurite defects in vang-1 mutants suggesting that PRKL-1 acts downstream of VANG-1 [11].We therefore asked if PRKL-1 overexpression could also rescue VC defects in fntb-1 mutants.A positive outcome would suggest that prenylation is not a necessary requirement for PRKL-1 to block ectopic neurite outgrowth.We found that PRKL-1 overexpression does not rescue fntb-1 mutants (Fig 5C) consistent with the notion that PRKL-1 prenylation is important for maintaining VC morphology.However, we cannot exclude the possibility that FTase modifies another component, in addition to PRKL-1, that is also important for blocking ectopic VC neurite formation.

Farnesylation is important for PRKL-1 membrane localization
Asymmetric Prickle accumulation on the plasma membrane of Drosophila epithelial cells [26] and vertebrate cells during convergent extension [27,28] is thought to play a key role in PCP signaling.In Drosophila, the requirement for Prickle prenylation in mediating membrane localization appears to be both prickle isoform and cell context-dependent [14,29,30].In zebrafish, Prickle membrane localization in cells undergoing gastrulation is partially dependent on its CAAX motif [31].Prenylation is also important for activity that does not reside at the plasma  [32] and polarity establishment in early embryos [33].
VC4 and VC5 extend neurites around the vulval epithelium beginning at the early L4 stage.During this period, PRKL-1 was found in the cytoplasm and in a punctate, but not an obvious asymmetric, distribution on the plasma membrane of VC neurons [11].Since PRKL-1:: ΔCAAX expression is unable to fully restore proper VC morphology in prkl-1 mutants, we asked if the CAAX domain is required for PRKL-1 localization to the plasma membrane.These results suggest that the CAAX domain, and therefore farnesylation, is important for PRKL-1 localization to the membrane of VC neurons.
However, if membrane localization is sufficient for proper neurite formation, then PRKL-1:: ΔCTVS localization to the membrane of 48% of neurons (n = 54) in wild type animals appears at odds with the almost complete failure of PRKL-1::ΔCTVS to rescue prkl-1(-/-) neurite defects (Fig 5B ) and the high penetrance of defects in fntb-1(-/-) mutants.These observations suggest that a C-terminal farnesyl group may be involved in more than simply mediating membrane insertion.One possibility is that PRKL-1 farnesylation acts to stabilize or promote an interaction between VANG-1 or another pathway component required for neurite inhibition.Such a notion has been proposed in Drosophila, where Prickle farnesylation is believed to promote physical association with Van Gogh to organize the polarity of wing cells [14].

Conclusions
We show for the first time that an FTase acts to establish proper neuronal morphology by blocking supernumerary neurite formation in C. elegans.The results of our genetic screen are consistent with previous work showing that a PCP-like pathway consisting of prkl-1, vang-1 and dsh-1 acts to establish and maintain neuronal soma morphology.We show that localization of PRKL-1, an FTase substrate, to the membrane of neuronal somas is partially dependent on both its CAAX motif and recruitment by VANG-1.During establishment of epithelial planar polarity in Drosophila, Prickle farnesylation has been shown to be important in some cellular contexts but not others.Prickle farnesylation has also been linked to both membrane and nuclear targeting in PCP-mediated processes as varied as epithelial cell polarity and neuronal and mesenchymal migration.These findings combined with our own highlight the diversity of roles played by farnesylation in the context of PCP-like signaling.

Fig 1 .
Fig 1.A genetic screen for VC4 and VC5 neurite outgrowth defective (Nde) mutants.(A) Worm schematic showing the position of VC1-6 (ventral view).VC4 and VC5 neurons extend neurites along a LR axis and VC1-3 and VC6 extend neurites along the AP body axis.(A) Wild-type VC4 and VC5.(B) In prkl-1 and vang-1 (not shown) mutants, VC4 and VC5 display ectopic neurites (arrows) directed along the AP axis.(D) Schematic outline of a forward genetic screen for VC4 and VC5 neurite outgrowth defective (nde) mutants.(E) Quantification of ectopic neurite defects in nde mutants.Graph shows the percentage of VC4 and VC5 neurons (pooled) displaying at least one ectopic neurite.Error bars represent the 95% confidence interval of the proportion.For each genotype, n>160.(F and G) Representative images of VC4 and VC5 in nde-4 (F) and nde-5 (G) mutants.Ectopic neurites are marked with arrows.All lines contain the cyIs4[Pcat-1:: GFP] reporter background.All images are ventral views.Scale bars, 10μm.doi:10.1371/journal.pone.0157537.g001

Fig 3 .
Fig 3. FNTB-1 missense mutations mapped to human FTase.(A) Crystal structure of the human protein FTase (PDB ID: 1S63) heterodimer (α subunit; red and β subunit; blue).Yellow circles mark the location of residues C104, G250, S301 and Q336 of the worm FTase.Lipid substrate farnesyl diphosphate, (FPP: isoprenoid moiety, green; diphosphate, red) and zinc ion (grey sphere) mark the site of catalysis.(B) Enlarged view of the FTase active site.G250 (yellow surface representation) makes van der Waals contact with FPP (grey surface).Any mutation at position 250 would result in a larger side chain that would affect FPP binding.(C) Enlarged view around residue C92 in worm FTase (C104 human FTase; green stick).The four major rotamers for the C104Y mutation and their corresponding van der Waals surfaces are shown in yellow.A van der Waals surface calculated around the alpha subunit (red) and beta subunit (blue) is shown.Any tyrosine rotamer results in a steric class with the protein.This mutation is predicted to compromise the packing and the stability of the enzyme.doi:10.1371/journal.pone.0157537.g003

Table 2 .
Sequence changes in nde mutants.