|Home | About | Journals | Submit | Contact Us | Français|
The Caenorhabditis elegans sax-1 gene regulates several aspects of neuronal cell shape. sax-1 mutants have expanded cell bodies and ectopic neurites in many classes of neurons, suggesting that SAX-1 functions to restrict cell and neurite growth. The ectopic neurites in sensory neurons of sax-1 mutants resemble the defects caused by decreased sensory activity. However, the activity-dependent pathway, mediated in part by the UNC-43 calcium/calmodulin-dependent kinase II, functions in parallel with SAX-1 to suppress neurite initiation. sax-1 encodes a serine/threonine kinase in the Ndr family that is related to the Orb6 (Schizosaccharomyces pombe), Warts/Lats (Drosophila), and COT-1 (Neurospora) kinases that function in cell shape regulation. These kinases have similarity to Rho kinases but lack consensus Rho-binding domains. Dominant negative mutations in the C. elegans RhoA GTPase cause neuronal cell shape defects similar to those of sax-1 mutants, and genetic interactions between rhoA and sax-1 suggest shared functions. These results suggest that SAX-1/Ndr kinases are endogenous inhibitors of neurite initiation and cell spreading.
Neuronal cells have complex morphologies, with distinctive cell bodies, axons, and dendrites. Although extrinsic cues that regulate axon guidance and branching have been defined (Tessier-Lavigne and Goodman, 1996 ; Mueller, 1999 ; Wang et al., 1999 ), less is known about the intrinsic determinants of cell shape. The mechanisms that determine a neuron's competence to form neurites, the number of neurites per cell, and the subcellular site of neurite initiation are unknown.
Rho family GTPases affect the actin cytoskeleton and morphology of many cell types (Hall, 1994 ). These GTPases exert their activity by binding and regulating multiple targets, including actin-binding proteins and several classes of kinases (Tapon and Hall, 1997 ). The Rho family GTPases Rac and Cdc42 are implicated in cell motility and axon outgrowth (Ridley et al., 1992 ; Luo et al., 1994 ; Kozma et al., 1995 ; Nobes and Hall, 1995 ; Luo et al., 1997 ), and Rho is implicated in contractile events such as stress fiber formation and neurite retraction (Paterson et al., 1990 ; Ridley and Hall, 1992 ; Jalink et al., 1994 ; Gebbink et al., 1997 ). Rho acts in part through the Rho kinase family, which includes Rho kinase (Leung et al., 1995 ; Ishizaki et al., 1996 ; Matsui et al., 1996 ), LET-502 (Wissmann et al., 1997 ), Genghis Khan (Luo et al., 1997 ), Citron (Di Cunto et al., 1998 ; Madaule et al., 1998 ), and MRCK (Leung et al., 1998 ). These proteins share a related kinase catalytic domain as well as domains that allow GTPase association. Mutations in the Caenorhabditis elegans LET-502 kinase prevent the epidermal cell shape changes that drive embryonic elongation, whereas mutations in Genghis Khan disrupt actin structures in the Drosophila egg chamber (Luo et al., 1997 ; Wissmann et al., 1997 ). These phenotypes, together with functional studies in cultured cells (Leung et al., 1996 ; Amano et al., 1997 ; Ishizaki et al., 1997 ), implicate Rho kinases in the regulation of cell morphology.
Members of a distinct family of serine/threonine kinases, including Orb6, COT-1, and Warts/Lats, are required for the regulation of cell morphology and cell division (Yarden et al., 1992 ; Justice et al., 1995 ; Xu et al., 1995 ; Verde et al., 1998 ). Kinases in the Orb6/COT-1/Warts family are closely related to Rho kinases in the kinase catalytic domain but lack consensus Rho-binding motifs. The pathways that regulate these kinases are not fully defined. Orb6 functions downstream of Pak1/Shk1 kinase (Verde et al., 1998 ), which acts together with the Cdc42 GTPase in cell shape regulation (Marcus et al., 1995 ; Ottilie et al., 1995 ). An expressed sequence tag (EST) in C. elegans led to the identification of Ndr, a new member of the Orb6/COT-1/Warts kinase family that is conserved in C. elegans, Drosophila, and humans (Millward et al., 1995 ). The function of the Ndr kinases is not known.
A screen for mutants with altered neuronal morphology in C. elegans identified mutations in the sax-1 and sax-2 genes (Zallen et al., 1999 ). Here we show that sax-1 encodes the C. elegans Ndr kinase, a member of the Orb6/COT-1/Warts family. sax-1 and sax-2 mutants exhibit defects in neuronal cell shape and polarity: cells appear expanded and irregular instead of compact and spherical, and they initiate ectopic neurites in addition to the normal axon and dendrite. Similar cell shape defects are caused by dominant negative mutations in the C. elegans RhoA GTPase. In sensory neurons, ectopic neurites are also caused by mutations that disrupt neuronal activity (Coburn and Bargmann, 1996 ; Coburn et al., 1998 ; Peckol et al., 1999 ). We find that the activity-dependent pathway for neurite initiation is modulated by the UNC-43 calcium/calmodulin-dependent protein kinase II (CaMKII) and functions in parallel with the SAX-1 (GenBank accession number AF275634) kinase to regulate neurite initiation.
Wild-type animals were C. elegans variety Bristol, strain N2. Strains were maintained at 20 or 25°C by standard methods (Brenner, 1974 ). Some strains were provided by the Caenorhabditis Genetics Center (St. Paul, MN).
sax-1 and sax-2 mutant alleles (Zallen et al., 1999 ) were outcrossed twice by kyIs4 X; him-5(e1490) V, once by kyIs4 X, and once by N2. Sensory axon defects detected by the ceh-23::gfp kyIs4 transgene were followed to map sax-1(ky211) to LGX. Five of five lon-2 non–lin-18 unc-78 recombinants and two of four lon-2 lin-18 non–unc-78 recombinants were mutant for sax-1. Two of seven lon-2 non–fax-1 recombinants were mutant for sax-1.
The stP40 restriction fragment length polymorphism that differs between the Bristol strains RW7000 and N2 was used to further map sax-1. Recombinants were isolated from unc-20(e112ts) sax-1(ky211) lon-2(e648)/RW7000 heterozygotes. None of three unc-20 sax-1 non–lon-2 recombinants and two of two unc-20 non–sax-1 lon-2 recombinants segregated the RW7000 stP40 polymorphism. None of two lon-2 sax-1 non–unc-20 recombinants and three of three lon-2 non–sax-1 unc-20 recombinants segregated the stP40 polymorphism. These mapping data placed sax-1 to the left of lin-18 and to the right or close on the left of stP40 on LGX.
Transgenic strains were created as described (Mello et al., 1991 ). Multiple lines from each injection were characterized for rescue of the sensory axon defects in the sax-1(ky211) kyIs4 X strain. Cosmids spanning the region between stP40 and lin-18 were injected at 10 ng/μl each in pools of four to five cosmids with the use of the dominant pRF4 rol-6(su1006) plasmid at 100 ng/μl as a coinjection marker. Cosmids from the rescuing pool were injected individually at 30 ng/μl with pRF4. Rescue activity of the R11G1 cosmid was retained in a 17-kilobase (kb) SacII–NarI subclone of the R11G1 cosmid cloned into the SacII–ClaI sites of pBSKII (pJAZ19, injected at 30 ng/μl) and a 7.7-kb SacII–XhoI subclone of the R11G1 cosmid cloned into the SacII–XhoI sites of pBSKII (pJAZ29, injected at 50 ng/μl). A deletion in the pJAZ19 sax-1 rescuing construct was made by digestion with BalI and religation, removing 1 kb of genomic sequence that deletes conserved kinase domains V (part), VIa, VIb, and VII (part) and breaks within an intron, which is predicted to disrupt sax-1 splicing. This deletion construct (pJAZ30) was injected at 30 ng/μl with pRF4.
A high-stringency screen of 5 × 105 plaques of a mixed-stage C. elegans cDNA library (Barstead and Waterston, 1989 ) that used the cm11b8 cDNA as a probe (a generous gift from R. Waterston, Washington University, St. Louis, MO) identified 15 full-length sax-1 cDNAs and 1 partial cDNA. Six cDNAs were fully sequenced and the rest were partially sequenced, encoding a predicted protein of 467 or 469 amino acids. The sax-1 sequence was confirmed, and its genomic organization was determined by aligning the cDNA with reported genomic sequences from the C. elegans Sequencing Consortium (1998) ; however, it differs from the gene predicted in that region (see below). The cDNAs were flanked by a 38-base pair 5′ untranslated region (UTR), a 335-base pair 3′ UTR, and a poly(A) tail.
The sax-1(ky211) mutation was identified by sequencing the sax-1 ORF and splice junctions from genomic DNA amplified with the use of the Expand PCR kit (Boehringer Mannheim, Indianapolis, IN). PCR fragments were sequenced on one strand with the use of the fmol sequencing kit (Promega, Madison, WI). The mutation was confirmed by sequencing a separately amplified PCR fragment.
The C. elegans Sequencing Consortium (1998) predicted a longer 1356-amino acid protein in the sax-1 region, including sax-1 exons and additional downstream exons. To determine whether these downstream exons could encode alternative exons of sax-1, we sequenced five ESTs kindly provided by Y. Kohara (National Institute of Genetics, Mishima, Japan); the longest contained 712 amino acids in addition to the 3′ UTR. The 5′ end of this transcript was identified by reverse transcription (RT)-PCR from wild-type N2 RNA prepared by Trizol extraction (GIBCO-BRL, Gaithersburg, MD) with the use of 3′ primers from the predicted ESTs in combination with a 5′ primer to the C. elegans SL2 spliced leader sequence. The SL2 spliced leader sequence is present at the 5′ end of C. elegans genes that are located downstream in an operon (Zorio et al., 1994 ). The full-length transcript encodes a predicted 811-amino acid protein. We named this transcript stg-1, for sax-1 three-prime gene. The stg-1 coding region was not required for sax-1 rescue. No transcripts were identified that extended from the stg-1 coding region into the sax-1 coding region or the SL1 spliced leader sequence. An additional in-frame stg-1 exon upstream of the SL2 splice junction was present in one of the five cDNAs; however, the 5′ end of this transcript was not detected by RT-PCR. Therefore, sax-1 and stg-1 appear to encode distinct protein products that may belong to a common operon, although it is possible that they are also included in a single transcript that was not detected by ESTs or RT-PCR.
Axons were scored in living adult animals, except for larval animals scored in Figure Figure2.2. The AWC neurons were visualized with an integrated str-2::gfp transgene (kyIs140 I; Dwyer et al., 1998 ). The ASER neuron was visualized with an integrated gcy-5::gfp transgene (kyIs164 II; Yu et al., 1997 ). The ASJ neurons shown in Figures Figures11 and and22 and described in Table Table11 were visualized with an integrated tax-2Δ::gfp transgene (kyIs150 IV; Coburn and Bargmann, 1996 ; Peckol et al., 1999 ). tax-2Δ::gfp labels ASJ along with the AWB, AWC, ASG, ASI, and ASK amphid chemosensory neuron pairs. Transgenes were integrated by trimethylpsoralen and UV radiation and outcrossed three to seven times. In Figures Figures33 and and5,5, ASJ neurons were visualized in the absence of a gfp transgene by exposing adult animals to 15 μg/ml 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) in M9 buffer (Brenner, 1974 ) for 1.5 h to label ASJ along with the ADL, ASH, ASI, ASK, and AWB amphid chemosensory neuron pairs. The ASJ defects scored by the tax-2Δ::gfp transgene or DiI filling differed in penetrance by up to 20%, but in all cases qualitatively similar results were obtained with the use of both tax-2Δ::gfp and DiI filling of strains lacking a gfp transgene.
Animals were scored as having ectopic neurite defects if at least one neuron had an ectopic thin or thick process that was longer than the diameter of the cell body. Lateral ectopic neurites emerged directly from the cell body in sax-1 and sax-2 mutants and grew posteriorly along the lateral body wall. tax-4 mutants exhibit both lateral and ventral neurite defects in ASJ (Peckol et al., 1999 ), whereas sax-1 and sax-2 mutants exhibit only lateral neurites. Only lateral neurites were included in this analysis. In Table Table1,1, Figure Figure2,2, and Figure Figure5,5, ASE and AWC neurons were scored as defective in cell shape if the mutant cell body appeared at least twice the size of a wild-type cell body. Because rhoA-expressing cells had milder defects than sax-1 mutants, in Figure Figure66 ASE neurons were scored as defective in cell shape if the cell body was expanded 1.5-fold compared with wild-type neurons.
Statistical analysis was conducted with the use of Primer of Biostatistics software (Stanton A. Glantz, McGraw-Hill, New York) and Kaleidagraph (Synergy Software, Reading, PA). Confocal images were acquired with the use of LaserSharp Acquisition version 2.1A software (Bio-Rad, Richmond, CA), an Optiphot-2 microscope (Nikon, Garden City, NY), and the MRC-1024 Laser Scanning Confocal Imaging System (Bio-Rad). Epifluorescence images were acquired with the use of an Axioplan 2 microscope (Zeiss, Thornwood, NY) and assembled with the use of Adobe Photoshop (Adobe, Mountain View, CA). The image in Figure Figure5D5D was acquired with a scientific-grade cooled charge-coupled device camera on a multiwavelength wide-field three-dimensional microscopy system (Hiraoka et al., 1990 ).
A full-length translational SAX-1::GFP fusion was constructed by cloning Green Fluorescent Protein (GFP) flanked by splice acceptor and donor sites (pPD103.75) into the blunted AflII site in the last intron of the SacII–XbaI sax-1 genomic fragment in a pBSKII+ backbone. This SAX-1::GFP transgene (pJAZ31) was injected at 50 ng/μl into a lin-15(n765ts) strain with the use of 30 ng/μl pJM23 lin-15 plasmid as a coinjection marker (Huang et al., 1994 ). To assess whether the SAX-1::GFP tag was functional, SAX-1::GFP was injected at 100 ng/μl into sax-1(ky211) X and sax-1(ky491) X mutant animals with pRF4, and rescue of the ectopic neurite phenotype was scored by DiI filling. The sax-1 expression pattern was characterized in animals that contained the SAX-1::GFP plasmid as an unstable extrachromosomal array. Chemosensory neurons were scored for rescue of ectopic neurite defects by DiI filling of transgenic animals.
The srh-1::gfp construct contained a 3-kb fragment of the srh-1 promoter (cosmid R09F10 nucleotides 24,398–27,397) engineered by PCR to contain a 5′ SphI site and a 3′ BamHI site and cloned into the corresponding sites of the pPD95.77 GFP vector by Yongmei Zhang (Howard Hughes Medical Institute, University of California, San Francisco, CA). The gcy-5 constructs contained a 2.7-kb fragment upstream of the gcy-5 start codon engineered by PCR to contain a 5′ SphI site and a 3′ BamHI site and cloned into the corresponding sites of the pPD95.75 GFP vector. GFP vectors were kindly provided by A. Fire, S. Xu, J. Ahnn, and G. Seydoux (Carnegie Institution of Washington, Baltimore, MD).
A KpnI site was engineered immediately upstream of the sax-1 cDNA start codon by PCR. The complete sax-1 cDNA was cut at the 5′ KpnI site and the 3′ NaeI site in the sax-1 3′ UTR. The sax-1 cDNA was joined to the downstream unc-54 3′ UTR (cut with EcoRI and blunted) and the upstream srh-1 or gcy-5 promoters (cut with KpnI). gcy-5::sax-1 was injected at 50 ng/μl with pRF4 into the kyIs164 II; sax-1(ky491) X strain. srh-1::sax-1 was injected at 50 ng/μl with pRF4 into the sax-1(ky491) X strain and the kyIs164 II; sax-1(ky491) X strain.
To express the SAX-1::GFP tag in a single cell type, the full-length SAX-1::GFP clone with GFP in the final intron was cloned into the srh-1::sax-1 fusion gene (pJAZ22) in two steps. The srh-1::sax-1::gfp fusion gene (pJAZ35) was injected at 100 ng/μl into wild-type N2 animals with pRF4, and ASJ morphology was scored by DiI filling.
A sax-1 deletion mutation was isolated by PCR as described (Dernburg et al., 1998 ). A deletion library of 106 genomes was constructed with the use of the ethyl methanesulfate (EMS) mutagen for half and UV/trimethylpsoralen for half of the library. Nested PCR primers were used to amplify a 3-kb fragment containing the sax-1 kinase domain. A 1.7-kb deletion band was identified in a pool from the EMS part of the library and sequenced. A single animal homozygous for the deletion was isolated after three rounds of sib selection. The sax-1(ky491) X deletion mutant was outcrossed once by N2 and once by lon-2(e678) X.
A full-length cDNA of the rhoA gene (Chen and Lim, 1994 ) was amplified by PCR from hexamer-primed C. elegans cDNA (a gift of Mario de Bono, Howard Hughes Medical Institute, University of California, San Francisco, CA) with the use of the primers 5′GAAGTCGACACGAGTACGGGTGTTC and 3′CCATGGAATTAGAGAGAAGAAGAGCAGAC, which contained artificial SalI and NcoI sites, respectively. The cDNA was subcloned into SalI–NcoI sites of the vector pPD49.26, sequenced, and determined not to contain any PCR-induced mutations. The gcy-5 promoter (Yu et al., 1997 ) was subcloned into the SphI–BamHI sites of the same vector with the use of artificial primers to create the restriction sites. Point mutations Q63L, T19N, and D13T in the rhoA gene were generated with the use of the Quikchange kit (Stratagene, La Jolla, CA). After mutagenesis, the complete rhoA coding region was sequenced to confirm the presence of the desired mutation and the absence of other mutations. rhoA plasmids were injected at 50 ng/μl with rol-6 as a cotransformation marker. gcy-5::sax-1 was injected at 50 ng/μl with the D13T mutant, and the gcy-5 promoter plasmid was injected at 50 ng/μl with the D13T mutant as a control.
ASER cell shape was scored in adult animals raised at 25°C with the use of the kyIs164 transgene.
sax-1 and sax-2 mutations cause ectopic neurite initiation from the ASI sensory neurons, the CAN neuroendocrine neurons, and glr-1–expressing interneurons (Zallen et al., 1999 ). Characterization of sax-1 and sax-2 mutants with cell-specific GFP markers for the AWC, ASE, and ASJ sensory neurons revealed similar defects in neurite initiation as well as defects in neuronal cell shape (Figure (Figure1,1, Table Table1).1). In sax-1 and sax-2 mutant animals, axon guidance of chemosensory neurons in the embryo was normal, producing a bipolar neuron with one dendrite that connects to the tip of the nose and one axon that grows circumferentially in the nerve ring neuropil. However, some sensory neurons had an ectopic neurite in addition to the normal dendrite and axon (Figure (Figure1,1, H and I). These ectopic neurites extended posteriorly from the cell body for up to 25 μm and were up to ~1 μm in diameter; a normal axon is 75 μm long and 0.1 μm in diameter. In addition, the sensory neuron cell bodies appeared expanded and irregular in shape (Figure (Figure1,1, B, C, E, and F). The expanded regions of the cell body are flattened and extended, like lamellipodia, but their structures have not been characterized in detail. The three sensory neuron types exhibited a range of phenotypes, with primarily cell shape defects in the AWC and ASE neurons and ectopic neurites in the ASJ neurons (Table (Table1). 1).
Despite the aberrant AWC morphology, sax-1 and sax-2 mutant animals generated nearly wild-type chemotaxis responses to attractive odorants detected by AWC (Figure (Figure2A),2A), suggesting that these cell shape defects do not eliminate AWC function. sax-1 and sax-2 mutant animals appeared morphologically and behaviorally normal, with no obvious defects in locomotion or egg laying.
The ectopic neurite phenotypes of sax-1 and sax-2 mutants are reminiscent of defects caused by mutations that disrupt the electrical activity of sensory neurons, which result in late-onset ectopic neurite initiation during larval and adult stages (Peckol et al., 1999 ). To determine whether sax-1 and sax-2 also affect a late developmental process, we examined ectopic neurites in the ASJ sensory neurons at different larval stages. Although ASJ axon guidance is completed in the embryo, the ectopic neurites in sax-1 and sax-2 mutants increased in severity throughout the four larval stages and continued to appear in adults (Figure (Figure2B).2B). Activity mutants are also temperature-sensitive for ectopic neurite formation (Peckol et al., 1999 ), and more ectopic neurites are observed in sax-1 and sax-2 mutants at higher temperatures (Table (Table1)1) (Zallen et al., 1999 ).
Unlike ectopic neurites, the cell shape defects in sax-1 and sax-2 mutants are not observed in mutants with defects in sensory activity and have not been described previously in C. elegans. The AWC cell shape defects were apparent even at the L1 stage, although they increased slightly in severity between the first and second larval stages (Figure (Figure2C).2C). The ASE cell shape defects were temperature-sensitive in sax-1 but not in sax-2 mutants, whereas AWC cell shape defects were temperature-sensitive in only one sax-1 allele (Table (Table11).
The ectopic neurite defects, late onset, and temperature-sensitivity of sax-1 mutants were all similar to the defects in mutants with altered sensory transduction or electrical activity. To determine whether SAX-1 participates in an activity-dependent pathway, we generated double mutants defective in sax-1 and the cGMP-gated sensory transduction channel tax-4 (Komatsu et al., 1996 ). ASJ neurons in the sax-1; tax-4 double mutant displayed significantly more ectopic neurites than a null mutant in either gene (Figure (Figure3A),3A), indicating that these genes function in two different pathways to suppress neurite initiation.
The SAX-1–related Ndr kinase can be regulated by calcium (Millward et al., 1998 ; see below), and calcium-dependent kinases such as CaMKII have been implicated in the regulation of neuronal morphology in response to neural activity (Wang et al., 1994 ; Wu and Cline, 1998 ). C. elegans has a single predicted CaMKII homologue encoded by the unc-43 gene (Reiner et al., 1999 ). To determine whether CaMKII is involved in the regulation of cell morphology in C. elegans, we examined chemosensory neurons in unc-43 mutants. Both the null unc-43(n1186lf) mutation and the gain-of-function unc-43(n498gf) mutation caused occasional ectopic neurites in the ASJ (Figure (Figure3B)3B) and ASE neurons [13% defective in unc-43(lf), n = 194; 3% defective in unc-43(gf), n = 159], but no cell shape defects.
Genetic evidence indicates that the UNC-43 CaMKII functions in the activity-dependent pathway to regulate neurite initiation. The ASJ defects in the tax-4; unc-43(lf) double mutant were not enhanced compared with the tax-4 single mutant (Figure (Figure3B), 3B), suggesting that UNC-43 functions in the same pathway as the TAX-4 sensory transduction channel. The tax-4; unc-43(gf) double mutant was significantly suppressed for the tax-4 ASJ defects, suggesting that UNC-43 acts downstream of TAX-4. However, tax-4 null mutants were significantly more defective than unc-43 null mutants at 25°C (Figure (Figure3B),3B), indicating that TAX-4 carries out additional UNC-43–independent functions.
UNC-43, like TAX-4, functions in parallel with the SAX-1 kinase. The ASJ neurons in the sax-1; unc-43(lf) double mutant were significantly more defective than in a null mutant in either gene, indicating that SAX-1 and UNC-43 operate in parallel pathways (Figure (Figure3C).3C). Interestingly, the sax-1; unc-43(gf) double mutant was partially suppressed compared with the sax-1 single mutant at 25°C, suggesting that increased activity of the UNC-43 kinase can partially compensate for decreased SAX-1 kinase activity.
sax-1 and sax-2 mutants have similar defects in neurite initiation and cell shape. The ASJ defects in sax-1; sax-2 double mutants were no more severe than those in either single mutant (Figure (Figure3D),3D), suggesting that these genes act in the same genetic pathway. We conclude that SAX-1 and SAX-2 function together, at least partly in parallel with an activity-dependent pathway that regulates neurite initiation through the action of the TAX-4 sensory transduction channel and the UNC-43 CaMKII.
sax-1 was mapped to the X chromosome between the stP40 polymorphism and the lin-18 gene, and cosmid pools covering this region were injected into the sax-1(ky211) strain. The R11G1 cosmid and its subclones rescued sax-1 mutant defects (Figure (Figure4A).4A). Sequence information from the C. elegans Sequencing Consortium predicted that the smallest rescuing subclone contained a serine/threonine kinase. A 1-kb deletion within the predicted kinase domain completely abolished rescuing activity (Figure (Figure4A).4A). To confirm that this gene was disrupted in sax-1 mutant animals, we sequenced sax-1(ky211) and identified a G-to-A transition mutation in a predicted splice acceptor site. Fifteen full-length sax-1 cDNAs from a mixed-stage C. elegans cDNA library (Barstead and Waterston, 1989 ) define a 1.8-kb primary transcript that encodes a predicted protein of 467 amino acids. An additional 2 amino acids were present at the 11th splice acceptor site in 1 of 15 clones. The 467-amino acid protein overlaps with the first third of the putative protein R11G1.4, which was predicted to encode a 1356-amino acid protein based on genomic sequence. The sax-1 transcript was contained within the 7.7-kb rescuing subclone, which included 1.9 kb of upstream sequence and 2.4 kb of downstream sequence (Figure (Figure4A). 4A).
sax-1 encodes a protein with homology to serine/threonine protein kinases, including the 11 highly conserved motifs that constitute the catalytic domain (Hanks et al., 1988 ). The SAX-1 kinase belongs to a family of serine/threonine kinases that is conserved from yeast to humans (Figure (Figure4,4, B and C). SAX-1 is most closely related to the Ndr kinases (Millward et al., 1995 ), with 62% amino acid identity to human Ndr and 60% identity to Drosophila Ndr. Certain features distinguish Ndr kinases from other serine/threonine kinases, including a conserved N-terminal region and a 34- to 45-amino acid spacer between kinase motifs VII and VIII. Several close relatives of Ndr possess these features, including Schizosaccharomyces pombe Orb6, Neurospora COT-1, and Drosophila Warts/Lats (Yarden et al., 1992 ; Justice et al., 1995 ; Xu et al., 1995 ; Verde et al., 1998 ). However, the Ndr kinases form a distinct class: C. elegans has both SAX-1/Ndr and a Warts/Lats homologue (T20F10.1), and Drosophila has an Ndr kinase as well as Warts/Lats.
SAX-1 is related to Rho kinases (Figure (Figure4C),4C), with 35% amino acid identity and 52% amino acid similarity to human p160ROCK/ROKβ. However, Rho kinases lack the SAX-1 spacer region between kinase subdomains VII and VIII, possess divergent N termini, and contain an additional C-terminal region that mediates Rho association (Leung et al., 1995 ; Fujisawa et al., 1996 ).
Immediately 3′ of the predicted sax-1 gene lies a second predicted ORF with a C1 motif (phorbol ester/diacylglycerol binding) and a C2 motif (calcium binding) that was predicted to be part of the same R11G1.4 protein as SAX-1 by the C. elegans Sequencing Consortium. We found that this gene encodes a separate 811-amino acid protein that does not overlap with SAX-1 (see MATERIALS AND METHODS) and is not required for sax-1 rescue. Its mRNA begins with an SL2 splice leader, suggesting that it may be in a common operon with sax-1.
The sax-1(ky211) mutation is predicted to disrupt sax-1 splicing and may cause a partial or complete loss of gene function. To obtain a null allele of sax-1, we used a PCR-based strategy to isolate a deletion mutant from a library of 106 EMS- or UV-trimethylpsoralen–mutagenized animals (see MATERIALS AND METHODS). The sax-1(ky491) mutation represents a 1.3-kb deletion of genomic sequence that leads to a frame shift in the sax-1 ORF early in the kinase domain. sax-1(ky491) is predicted to encode a truncated gene product containing only 3 of the 11 conserved kinase domains. Both break points of the deletion occur in exons, and the deleted sequences encompass conserved domains IV, V, and VIA of the sax-1 kinase region. Based on the early frame shift and the complete elimination of conserved kinase sequences, sax-1(ky491) is a candidate null allele. The sax-1(ky491) deletion mutation was fully recessive and caused defects similar to those caused by the sax-1(ky211) allele (Table (Table1,1, Figures Figures1E1E and and2A),2A), indicating that the partially penetrant, temperature-sensitive defects in the ASJ and ASE neurons represent the sax-1 null phenotype.
To determine the pattern of sax-1 expression, we inserted a gfp reporter flanked by splice sites into the final sax-1 intron, 15 amino acids before the stop codon. The SAX-1::GFP fusion contained all exons, introns, and 5′ regulatory regions present in the rescuing sax-1 subclone. The SAX-1::GFP tag rescued the ASJ defects of sax-1(ky211) and sax-1(ky491) mutant animals (Figure (Figure4A).4A). SAX-1::GFP was widely expressed in embryos and continued to be expressed in larvae in neurons that contribute axons to the nerve ring (Figure (Figure5A),5A), hypodermal cells, including lateral seam cells (Figure (Figure5B),5B), and muscle, where the SAX-1::GFP tag displayed a punctate localization (Figure (Figure5C).5C). The rescuing SAX-1::GFP fusion was present in the nucleus and cytoplasm of cells (Figure (Figure5,5, A–D).
Although they express SAX-1::GFP, hypodermal seam cells do not have obvious defects in sax-1 mutants. They secrete normal alae, indicating a correct apical/basal polarity, and have an apparently normal cell shape as assessed by expression of the cell junction marker JAM-1::GFP (Mohler et al., 1998 ).
Because SAX-1::GFP is broadly expressed, we wanted to determine whether SAX-1 functions cell autonomously in chemosensory neurons to regulate cell shape and neurite initiation. We used the gcy-5 promoter (Yu et al., 1997 ) to drive sax-1 expression in the right ASE neuron. gcy-5::sax-1 expressed in ASER was able to fully rescue the ASER defects of sax-1 null mutant animals (Figure (Figure5E),5E), suggesting that SAX-1 can function cell autonomously to regulate ASER cell shape. However, the ASJ ectopic neurite defects were also partially rescued in these animals (Figure (Figure5E).5E). These results are consistent with two possibilities: SAX-1 could function cell autonomously in ASER and nonautonomously in ASJ, or the gcy-5 promoter may be expressed at a low level in ASJ, allowing partial ASJ rescue.
To determine whether SAX-1 can function cell autonomously in ASJ to repress neurite initiation, we used an independent promoter to express SAX-1 in ASJ but not in ASE. The promoter for the predicted seven-transmembrane-domain receptor srh-1 drives expression in the two ASJ sensory neurons and pharyngeal muscle (Y. Zhang and C.I.B., unpublished results). Expression of the sax-1 cDNA from the srh-1 promoter fully rescued the ectopic neurite defects in the ASJ neurons of sax-1(ky491) null mutant animals but did not rescue the ASER defects (Figure (Figure5E).5E). To determine whether rescuing activity was due to sax-1 expression in ASJ or in pharyngeal muscle, we coinjected the srh-1::sax-1 transgene and an srh-1::gfp fusion to form an unstable extrachromosomal array whose site of expression can be monitored by gfp fluorescence. Mosaic animals with gfp expression in ASJ but not in the pharyngeal muscle were rescued for the ASJ defects (14 of 15 mosaic animals rescued). This result suggests that sax-1 expression in ASJ is sufficient for ASJ rescue, indicating that SAX-1 is likely to function cell autonomously in ASJ to regulate neurite initiation.
An srh-1::sax-1::gfp tag also rescued the ASJ ectopic neurite defects of sax-1 mutants [three independent transgenic lines: 10–12% defective, n = 46–80, each line significantly rescued at p < 0.001; control sax-1(ky491) animals: 61% defective, n = 218]. In larval and adult animals, SAX-1::GFP fluorescence was detected throughout the ASJ cell body, including the nucleus and the cytoplasm, and occasionally in the proximal axon and dendrite (Figure (Figure55D).
SAX-1 belongs to a kinase family that includes the Rho kinases, but its targets are unknown. To determine whether SAX-1 might have functions similar to those of the Rho kinases, we explored the activity of the C. elegans RhoA GTPase in sensory neurons. C. elegans RhoA shares 87% identity with vertebrate RhoA GTPases and is expressed in nerve ring neurons during larval stages (Chen and Lim, 1994 ). Use of RNA interference to disrupt rhoA function caused embryonic lethality (E.A. Lundquist and C.I.B., unpublished observations), precluding a straightforward loss-of-function analysis. Therefore, we generated mutations in conserved residues in the rhoA ORF that are predicted to confer dominant negative (rhoAT19N) or gain-of-function (rhoAQ63L) properties (Feig and Cooper, 1988 ; Bourne et al., 1991 ). Wild-type rhoA, rhoAT19N, and rhoAQ63L were expressed from the gcy-5 promoter (Yu et al., 1997 ) to drive expression in the ASER chemosensory neuron in late embryonic-, larval-, and adult-stage animals. The rhoAT19N allele, which is predicted to reduce endogenous rhoA function, caused ASER cell shape defects that resembled those of sax-1 mutant animals (Figure (Figure6D).6D). A second mutation in the GTP-binding domain, rhoAD13T, also caused cell shape defects (Figure (Figure6,6, B and D), suggesting that this allele also functions in a dominant negative manner. Minimal defects were caused by expression of rhoA or rhoAQ63L (Figure (Figure6D).6D). These results suggest that RhoA-dependent pathways can regulate cell shape in chemosensory neurons. However, the defects observed in rhoA-expressing animals were less marked than those in sax-1 mutants. Whereas sax-1 mutant cell bodies were usually more than twice as large as wild type, the rhoAT19N or rhoAD13T cell bodies were rarely more than 1.5-fold larger than wild type.
A sax-1; rhoAT19N strain had defects that were no more severe than the sax-1 single mutant, suggesting that sax-1 and rhoA affect a similar process (Figure (Figure6E).6E). In agreement with this model, overexpression of sax-1 partly suppressed rhoAD13T defects (Figure (Figure6,6, C and D), and overexpression of rhoA or rhoAQ63L partly suppressed sax-1 defects (Figure (Figure66E).
During normal development, neurite outgrowth is limited to discrete axons and dendrites, and the compact neuronal cell body is refractory to neurite initiation. In sax-1 mutants, chemosensory neurons and other neuron types have an extended, irregularly shaped cell body and ectopic neurite-like processes. sax-1 activity may stabilize neuronal morphology, maintain cell polarity, or localize factors that make the rigid cell body distinct from the exploratory axons. sax-1 encodes a serine/threonine kinase related to the Drosophila and human Ndr kinases (Millward et al., 1995 ) in a kinase family that includes S. pombe Orb6, Neurospora COT-1, and Drosophila Warts/Lats. The function of the Ndr kinase subfamily has not been analyzed previously, but Orb6, COT-1, and Warts/Lats all influence cell shape (Yarden et al., 1992 ; Justice et al., 1995 ; Xu et al., 1995 ; Verde et al., 1998 ), suggesting that these kinases participate in an evolutionarily conserved mechanism that regulates cell morphology. SAX-1 is the first member of this family found to affect neurons.
SAX-1 and related kinases may preferentially affect particular subcellular compartments or cell types. sax-1 mutants have abnormal cell bodies, but axon and dendrite guidance in the same neurons appears to be normal. Drosophila warts/lats mutants exhibit localized epithelial cell shape defects, with abnormal apical surfaces but normal basolateral morphologies (Justice et al., 1995 ). C. elegans, Drosophila, and humans all have both an ndr/sax-1–like gene and a warts/lats-like gene, suggesting that these two kinases have distinct functions. sax-1 defects become more severe as animals mature, suggesting an ongoing requirement for SAX-1 kinase activity in the maintenance of cell morphology. The fission yeast Orb6 kinase is similarly required for cell shape maintenance; arrested elongate cells become spherical when orb-6 function is disrupted (Verde et al., 1995 ). Warts/Lats and Orb6 play an additional role in cell division (Justice et al., 1995 ; Xu et al., 1995 ; Verde et al., 1998 ; St. John et al., 1999 ), but we have not observed cell proliferation defects in sax-1 mutants.
The functions of SAX-1 in C. elegans are reminiscent of the activities of RhoA GTPase in cultured mammalian neurons: inactivation of RhoA promotes cell flattening and neurite outgrowth, whereas activation of RhoA causes cell rounding and decreased neurite outgrowth (Jalink et al., 1994 ; Gebbink et al., 1997 ). The effect of RhoA on neurite outgrowth has been shown to be mediated by Rho kinase (Amano et al., 1998 ; Hirose et al., 1998 ; Katoh et al., 1998 ). We found that dominant negative rhoA alleles or mutations in the SAX-1 kinase caused similar cell shape changes in C. elegans sensory neurons. Our genetic results are consistent with a model in which SAX-1 and RhoA act in a common process affecting cell shape. These experiments should be interpreted with caution, because they involve overexpression of wild-type and mutant proteins. However, standard genetic analysis of this pathway is likely to be difficult, because RNA interference of rhoA resulted in early embryonic lethality (E. Lundquist and C.I.B., unpublished data). Mutations in other GTPase pathways, including the C. elegans MIG-2 GTPase or the UNC-73 Dbl-homology exchange factor (Zipkin et al., 1997 ; Steven et al., 1998 ), do not cause cell shape defects in ASER, indicating that the effects of RhoA expression on cell shape are selective.
We propose that SAX-1, like Rho GTPase and Rho kinase, regulates the cytoskeleton in neuronal cells to inhibit cell spreading and neurite initiation. Human Ndr kinase localizes primarily to the nucleus in cultured cells, but a subpopulation of human Ndr protein is present at the cell periphery, where it could interact with the cytoskeleton (Millward et al., 1995 ). Orb6/COT-1/Warts family kinases are related to Rho kinases but do not contain a Rho-binding domain and other regulatory domains. It is possible that SAX-1 and Rho kinases can phosphorylate similar targets but are regulated differently.
The SAX-1 kinase functions in parallel with an activity-dependent pathway to inhibit neurite initiation in sensory neurons. Ectopic chemosensory neurites appear late in development in mutants with abnormal sensory transduction, ion channel function, or development of the dendritic sensory endings (Coburn and Bargmann, 1996 ; Coburn et al., 1998 ; Peckol et al., 1999 ). The C. elegans unc-43 gene encodes CaMKII (Reiner et al., 1999 ), and we observed mild neurite defects in unc-43 mutants. Double-mutant analysis suggests that the calcium-permeable sensory channel TAX-4 functions upstream of the UNC-43 CaMKII in the activity-dependent pathway. CaMKII has also been implicated in the activity-dependent regulation of axonal and dendritic branching in flies and vertebrates (Budnik et al., 1990 ; Wang et al., 1994 ; Wu and Cline, 1998 ).
Although the ectopic neurite defects in sax-1 and activity mutants are similar, double-mutant analysis argues that the SAX-1 kinase functions at least partly in parallel with TAX-4/UNC-43, not in an identical pathway. Moreover, the activity-dependent pathway affects only neurite initiation, whereas SAX-1 also regulates cell shape. The human Ndr kinase is regulated by calcium through a domain that is conserved in SAX-1 (Millward et al., 1998 ), suggesting that SAX-1, like UNC-43, might respond to neuronal activity and calcium. Alternatively, these two pathways may represent activity-dependent (TAX-4 and UNC-43) and activity-independent (SAX-1) mechanisms that maintain sensory neuron morphology.
Interestingly, increased activity of the UNC-43 CaMKII can partially compensate for decreased activity of SAX-1, perhaps by regulating a common downstream target. SAX-1/Ndr targets are unknown, but the related Rho kinase and CaMKII both phosphorylate the same residue on myosin light chain (Edelman et al., 1990 ; Amano et al., 1996 ). Phosphorylated myosin can regulate actin assembly and inhibit neurite extension (Tan et al., 1992 ; Wang et al., 1996 ), so this is one candidate among many that could be regulated by SAX-1 and UNC-43. The characterization of SAX-2 and other components of the SAX-1 pathway should provide further insight into the mechanisms that regulate neuronal morphology.
We are grateful to Orion Weiner and John Sedat for deconvolution microscopy of the SAX-1::GFP fusion, Jeff Simske, Yongmei Zhang, Andy Fire, and Emily Troemel for GFP markers, and Rachel Kindt, Orion Weiner, and Maria Gallegos for comments on the manuscript and useful advice during the course of this work. Some of the strains used in this study were obtained from the Caenorhabditis Genetics Center. This work was supported by the Howard Hughes Medical Institute. J.A.Z. and D.M.T. were predoctoral fellows of the National Science Foundation. E.L.P. was supported by a research fellowship from the American Heart Association, California Affiliate. C.I.B. is an Investigator of the Howard Hughes Medical Institute.