Sulfations of sugars, such as heparan sulfates (HS), or tyrosines require the universal sulfate donor 3′-phospho-adenosine-5′-phosphosulfate (PAPS) to be transported from the cytosol into the Golgi. Metazoan genomes encode two putative PAPS transporters (PAPST1 and PAPST2), which have been shown in vitro to preferentially transport PAPS across membranes. We have identified the C. elegans orthologs of PAPST1 and PAPST2 and named them pst-1 and pst-2, respectively. We show that pst-1 is essential for viability in C. elegans, functions non-redundantly with pst-2, and can act non-autonomously to mediate essential functions. Additionally, pst-1 is required for specific aspects of nervous system development rather than for formation of the major neuronal ganglia or fascicles. Neuronal defects correlate with reduced complexity of HS modification patterns, as measured by direct biochemical analysis. Our results suggest that pst-1 functions in metazoans to establish the complex HS modification patterns that are required for the development of neuronal connectivity.
- Heparan sulfate
- PAPS transport
- Neuronal development
- Neuronal connectivity
- Embryonic development
Sulfations are ubiquitous modifications of biological macromolecules found in all cells from bacteria to mammals. They occur as post-translational modifications of specific tyrosine residues in proteins (Moore, 2003), of polysaccharides such as heparan sulfates (HS) (Esko and Lindahl, 2001) and of small molecules such as sterols (Strott, 2002).
Cells actively transport inorganic sulfate into the cytosol, and mutations in an inorganic sulfate transporter have been shown to be responsible for diastrophic dysplasia in humans (Hastbäcka et al., 1994). In the cytosol, a PAPS synthase converts inorganic sulfate to 3′-phospho-adenosine-5′-phosphosulfate (PAPS) (Strott, 2002), the universal sulfate donor for all enzymatic reactions in vivo. PAPS synthase is a bifunctional enzyme that has both sulfurylase and kinase activity, and mutations in a gene coding for a PAPS synthase result in spondyloepimetaphyseal dysplasia in humans (ul Haque et al., 1998). The Caenorhabditis elegans genome harbors a single orthologous gene for the PAPS synthase, named pps-1, which is essential for viability (Dejima et al., 2006) (Table 1).
Sulfotransferases can be classified on the basis of their substrate specificities and subcellular localization. For instance, Golgi resident type II transmembrane sulfotransferases carry out sulfations of sugars and tyrosines (Esko and Lindahl, 2001; Moore, 2003). Because all sulfotransferases require PAPS as sulfate donor and PAPS is synthesized in the cytosol, reactions in the Golgi require transport of PAPS across at least one membrane. This transport is mediated by specific multi-transmembrane-spanning solute carrier proteins that are called PAPS transporters 1 and 2 (PAPST1 and PAPST2) in vertebrates and Slalom in Drosophila (Clement et al., 2008; Kamiyama et al., 2006; Kamiyama et al., 2003; Lüders et al., 2003). Human PAPST1 and PAPST2 and Slalom are similar to nucleotide sugar transporters and, in vitro, preferentially transport PAPS across membranes (Kamiyama et al., 2006; Kamiyama et al., 2003; Lüders et al., 2003). Drosophila slalom is required for wingless and hedgehog signaling as well as for the establishment of the dorsal-ventral axis of the Drosophila oocyte (Lüders et al., 2003), consistent with the role of sulfated molecules in these processes (Sen et al., 1998).
Tyrosine sulfation is a post-translational protein modification in the Golgi. Sulfation of extracellular tyrosines of G-protein-coupled chemokine receptors has been shown to be required for specific ligand-receptor interactions (Kehoe and Bertozzi, 2000; Moore, 2003). For instance, tyrosine sulfation of a specific extracellular residue in CCR5 (C-C chemokine receptor type 5) is crucial for its function as a co-receptor for the gp120-CD4 complex required for HIV entry into cells (Farzan et al., 1999).
Heparan sulfates are unbranched polysaccharides of alternating hexuronic acid and glucosamine residues, and are attached to proteins to form HS proteoglycans. The molecular complexity of the rather simple heparan polymer is greatly increased by non-uniform epimerizations and sulfations of the sugar moieties in the Golgi (Esko and Lindahl, 2001). Individual modifications are crucial for cell-cell signaling pathways during various aspects of development (Bülow and Hobert, 2006; Häcker et al., 2005; Lee and Chien, 2004; Van Vactor et al., 2006). Recent systematic analyses of loss-of-function mutations in several HS sulfotransferases have documented a role for particular HS modifications in neural development (Bülow and Hobert, 2004; Kinnunen et al., 2005; Pratt et al., 2006). However, the function of PAPS transport and overall sulfation in development of the nervous system has not been investigated.
We report here the identification and characterization of pst-1, the nematode ortholog of PAPST1/slalom. Our experiments show that pst-1 is required for viability, acts non-redundantly with pst-2 and can act non-autonomously in its essential function. Animals with a null mutation in pst-1 display defects in subsets of neuronal classes, including highly specific defects in the formation of collateral axonal branches and motor synapses. The neuronal defects correlate with heparan sulfate of reduced molecular complexity, as determined by direct biochemical analyses. Together, our studies show that sulfation of heparan is crucial for the establishment of neuronal connectivity.
Identification of genes involved in sulfation in C. elegans
From sequence analysis of the C. elegans genome we identified two genes coding for putative PAPS transporters, which are solute carrier proteins of the SLC35B2 and SLC35B3 class, respectively. We named the orthologous genes of PAPST1 and PAPST2 in C. elegans pst-1 and pst-2, respectively (Table 1; supplementary material Figs S1, S2). Further, there are two gene coding for predicted protein-tyrosine sulfotransferases, named tpst-1 and tpst-2 (tyrosine protein sulfotransferase 1 and 2, respectively) (Table 1, supplementary material Fig. S3). Additionally, the C. elegans genome harbors one gene (ssu-1) coding for a sulfotransferase with high homology to mammalian cytosolic alcohol sulfotransferases (Table 1, supplementary material Fig. S3). This gene might be required in C. elegans for the biosynthesis of a secreted compound(s) involved in dauer formation and locomotion (Carroll et al., 2006). We also identified at least five genes with similarity to genes encoding heparan sulfate (HS) sulfotransferases in the C. elegans genome. In addition to the genes encoding HS 2-O- and 6-O-sulfotransferases, hst-2 and hst-6 (Bülow and Hobert, 2004; Kinnunen et al., 2005), we found a single ortholog for the N-deactetylase/N-sulfotransferase (hst-1) and two genes (hst-3.1 and hst-3.2) coding for the predicted HS 3-O-sulfotransferases (Table 1). We did not find genes with homology to chondroitin sulfotransferases within the C. elegans genome, consistent with the absence of sulfation of chondroitin in C. elegans (Toyoda et al., 2000) (R.A.T. and H.E.B., unpublished results).
pst-1 is required for kal-1-induced axonal branching and misrouting
Kallmann Syndrome protein 1 (KAL1; also known as anosmin-1) is an extracellular cell adhesion molecule. It comprises a cysteine-rich region, a protease-inhibitor domain and four fibronectin III repeats, and is mutant in the X-linked form of Kallmann Syndrome, a neural targeting and migration defect (Cadman et al., 2006). To gain insight into KAL1 function and to identify genetically interacting loci, we had previously conducted a modifier screen of an axon-branching defect induced by overexpression of the C. elegans ortholog kal-1 (Bülow et al., 2002). Of four strong suppressor mutations of the axon-branching phenotype, two mutations are in the gene encoding HS 6-O-sulfotransferase (hst-6) and one in the HS glucuronyl C5 epimerase (hse-5) (Bülow et al., 2002; Bülow and Hobert, 2004). These genes code for enzymes involved in the modifications of heparan sulfate. We describe here the characterization of the fourth strong suppressor mutation, the recessive allele ot20 (Fig. 1B, supplementary material Fig. S4). We mapped ot20 to the proximal left arm of chromosome V (Fig. 1A) and identified a single point mutation in M03F8.2, the gene coding for PST-1 (see Materials and Methods for details). The point mutation in ot20 results in a change from serine to proline at position 193 of PST-1, which affects all splice variants. Serine 193 lies in a predicted luminal loop of 66 amino acids between the third and fourth transmembrane domains (Fig. 1E) and is perfectly conserved among all members of the SLC35B2 subfamily, of which the PAPST1 proteins form a part (supplementary material Fig. S2A). The genomic region covering pst-1 rescues the mutant phenotypes, that is, rescues the lethality (Fig. 1A) and reverts the suppression of branching (Fig. 1C). Furthermore, RNA-mediated interference against pst-1, but not against adjacent genes in the genomic region, phenocopies the ot20 mutant phenotype, that is, suppresses the kal-1-induced axon-branching phenotype (Fig. 1D). We conclude that ot20 is a loss-of-function mutation in the predicted PAPS transporter encoded by pst-1.
The ot20 allele displays a temperature-sensitive lethality, with decreased temperature (15°C) resulting in embryonic and early larval arrest. We reasoned that ot20 might represent a hypomorphic allele and that a complete loss-of-function allele could be inviable. Numerous essential genes have been identified on the left arm of chromosome V, and grouped into complementation groups by deficiency mapping and complementation tests (Johnsen and Baillie, 1991; Stewart et al., 1991). Using the cold-sensitive lethality, we conducted complementation tests between ot20 and the same deficiencies. We found that complementation patterns of ot20 are consistent with four essential genes: let-429, let-439, let-462 and let-479 (supplementary material Table S1). Of those, only the four alleles of let-462 fail to complement ot20 and we found single nucleotide changes in three alleles of let-462 in the coding region of pst-1. Both s1481 and s1590 harbor missense mutations that change perfectly conserved glycine residues to charged amino acids in the predicted second and seventh transmembrane domain, respectively (Fig. 1E, supplementary material Fig. S2A). Comparable mutations in conserved glycine residues of transmembrane domains have been associated with loss of transport activity in other nucleotide-sugar transporters (Eckhardt et al., 1998). The allele s1956 harbors a nonsense mutation that affects all splice variants and leads to a protein that is truncated in the second transmembrane domain (Fig. 1E, supplementary material Fig. S2A), thus probably representing a molecular null allele. We conclude that let-462 and pst-1 are the same gene, and refer to this gene in this paper as pst-1.
pst-1 function is essential for late embryonic and larval development
Animals homozygous for the null allele pst-1(s1956) die during embryonic and larval stages. We first asked whether pst-1 is maternally contributed. To this end, we compared the lethality of homozygous null mutant animals coming from a heterozygous mother (lacking zygotic pst-1) with animals lacking both maternal and zygotic pst-1. We found 63% (n=332) embryonic lethality in zygotic null animals compared with 88% (n=389) embryonic lethality in animals lacking both maternal and zygotic expression. This demonstrates that pst-1 is maternally contributed (Fig. 2A). The lethality of the null allele is not cold-sensitive (Fig. 2B), nor is it the result of apparent changes in cell fate because markers for hypodermal fate, neuronal fate and body wall muscle are normally expressed (Fig. 2D,F; Fig. 3; and data not shown).
To determine why pst-1(s1956) mutant animals die, we followed the development of mutant embryos. We found that s1956 null mutant embryos develop slower than wild type (data not shown) and that the majority of embryos (72.1%, n=1019) arrest shortly before or at the threefold stage, the last embryonic stage (Fig. 2C). Mutant embryos at the comma stage, which marks the beginning of embryonic elongation, display normal organization of the hypodermis. By contrast, the hypodermis is often severely disorganized in embryos arresting at the threefold stage (Fig. 2D). We conclude that pst-1 function is required for late embryonic development and that earlier embryonic stages (including gastrulation, morphogenesis and initial elongation) are independent of pst-1 function.
Of the embryos without maternal and zygotic pst-1, 12% do hatch, but die within four days after arresting at the L1 or L2 larval stage (data not shown). Of these animals, 62.2% display defects at the L1 and L2 larval molts compared with 3.8% (n=212) of control animals (Fig. 2E). In addition, 41.5% (n=270) of the animals display a pharynx detachment phenotype, compared with 1.6% (n=126) of control animals (Fig. 2F).
We next assayed the lethality of the pst-1(ot20) temperature (cold)-sensitive allele. When pst-1(ot20) animals are shifted to the non-permissive temperature at the L3 stage, the lethality of their progeny is similar to the null allele lacking zygotic pst-1 (Fig. 2G). However, pst-1(ot20) animals grown at the permissive temperature show only a slightly higher lethality than wild-type animals (P=0.10) (Fig. 2G). This indicates that, with regard to the essential functions, the temperature-sensitive allele pst-1(ot20) is a very strong loss-of-function allele at the non-permissive temperature and close to normal at the permissive temperature (compare Fig. 2A and 2G).
We used the pst-1(ot20) allele to determine the point at which pst-1 function is required during development. Eggs were grown at the permissive temperature and shifted to the non-permissive temperature at various time-points during development (Fig. 2H). We found that the lethality decreases when the worms are shifted to the non-permissive temperature at increasingly later developmental stages (Fig. 2H). Because 20% of pst-1(ot20) animals die even when shifted to the non-permissive temperature as adults, it is possible that sulfation is required throughout life. Nonetheless, our experiments show that pst-1 function, and by inference sulfation, is crucial during late embryonic and larval stages.
pst-1 is required for specific aspects of neural development
We initially identified pst-1(ot20) as a modifier of neurite-branching morphology (Bülow et al., 2002). Therefore, we sought to determine the effect of reduced sulfation on development of the nervous system. To maximally reduce gene function, we conducted our analyses wherever possible with the null allele pst-1(s1956) lacking both maternal and zygotic pst-1. We found that overall organization of the nervous system is intact in animals lacking all pst-1 function. Neurons are organized into the major ganglia and fascicles (Fig. 3A), indicating no gross disruption of neuronal cell fate or development. Using a panel of green fluorescent protein (GFP) reporter strains that allow visualization of individual nerve cells in high resolution, we found that several classes of neurons appear to develop largely independent of pst-1 function (Table 2). For instance, the AIY interneurons, which are located in the retrovesicular ganglion and send their axons into the nerve ring, display essentially wild-type axonal morphologies (Fig. 3B; Table 2).
We observed specific neuronal migration defects in a subset of neurons. The hermaphrodite-specific neurons (HSNs) display migration defects in pst-1 null mutants whereas the canal-associated neurons (CANs) are largely unaffected (Fig. 3C). Both CAN neurons and HSN neurons are pairs of neurons in the C. elegans hermaphrodite that migrate during embryonic stages from the head and tail region, respectively, towards the midbody of the worm (White et al., 1986). Our findings indicate that HSN and CAN neurons require distinct cues for correct migration, one of which is pst-1-dependent and the other largely pst-1-independent.
We next tested the requirement of sulfations for axonal midline guidance. We found that PVQ and PVP neurons, a set of posterior ventral nerve cord (VNC) interneurons that send axons anteriorly in the VNC, display 43.3% and 34.7% midline crossover defects, respectively (Fig. 3D, Table 2). Similarly, we detected defects in axonal midline guidance of HSN as well as guidance and fasciculation defects in D-type motor neurons (Table 2; supplementary material Fig. S5). By contrast, AVK neurons, a pair of anterior VNC interneurons that execute conceptually similar guidance choices to PVQ and PVP but send their axons posteriorly from the head, display only minor defects (Table 2). Likewise, the AVG and PVT pioneer neurons of the VNC appear unaffected by genetic removal of pst-1 (data not shown). AVG and PVT are pioneer neurons that migrate in the ventral nerve cord in anterior-posterior and posterior-anterior directions, respectively. Our studies show that not all axonal guidance choices at the midline of the VNC require pst-1 activity and that there is no correlation between pst-1 function and anterior-posterior or posterior-anterior guidance (Table 2).
We also identified defects in specific aspects of development of PLM touch neurons. The anterior processes of the PLM neurons, which are located in the tail of the animal, fail to stop at defined positions along the length of the animal and instead overextend (Fig. 3E). Additionally, we observed defects in the formation of synaptic branches of PLM neurons in pst-1 mutants (Fig. 3F). These phenotypes are reminiscent of phenotypes in synaptogenesis mutants (Jin, 2005), prompting us to analyze synaptic structure in more detail. We focused our analyzes on a well-characterized motor synapse formed between GABA-ergic D-type motor neurons and body-wall muscles (Jin, 2005). We found reduced synaptic labeling with a synaptobrevin GFP presynaptic reporter in D-type motor neurons (Fig. 3G), which indicates defects in presynaptic development. Specifically, the fluorescence intensity of synapses labeled in the ventral cord was reduced by 35% in pst-1(s1956) null mutants compared with isogenic controls (Fig. 3H). Taken together, our experiments show that pst-1 is required for highly specific aspects of neural development but is dispensable for overall development of the nervous system. Specifically, pst-1 is important for the developmental guidance choices of a select number of neurons and for the establishment of neuronal connections in motor neurons and, possibly, in touch neurons (Fig. 3, Table 2).
Cellular focus of pst-1 functions
To determine where pst-1 function is required, we conducted cell-specific rescue experiments with the two most abundant pst-1 transcripts, pst-1a and pst-1b. We first assayed for rescue of the cold-sensitive lethality of the pst-1(ot20) allele. Expression of either pst-1 isoform in body-wall muscles, the hypodermis or neurons rescued the cold-sensitive lethality of pst-1(ot20) (Fig. 4A). Surprisingly, expression in the pair of AIY interneurons alone was able to rescue the lethality of pst-1(ot20) (Fig. 4A) and the pst-1(s1956) null mutant (data not shown). By contrast, a piece of genomic DNA adjacent to pst-1 (PCR C, Fig. 4A) was not able to rescue transgenically the lethality. These findings suggest that pst-1 acts non-autonomously in its essential functions during development and that both pst-1a and pst-1b isoforms can supply this essential function.
To investigate whether the neuronal branching phenotype and the lethality phenotype are mediated by the same focus of action of pst-1, we determined where pst-1 function is required for the kal-1-dependent branching phenotype in AIY interneurons, which is suppressed by the pst-1(ot20) loss-of-function mutation. We assayed the AIY-branching phenotype (Bülow et al., 2002) using the same transgenic lines that were used in the rescue assay of the essential functions (Fig. 4A). We found that expression of neither pst-1a nor pst-1b in hypodermal tissues can rescue the branching phenotype. Expression of pst-1a or pst-1b in neurons, muscle or the AIY interneurons alone can partially rescue the phenotype (Fig. 4B). Similar results were obtained with rescue of the PLM-overextension phenotype (data not shown). One possible explanation for these findings is that sulfation is required in more than one tissue and that expression of pst-1 in one tissue alone might not be sufficient for full rescue of the neuronal phenotypes. An alternate explanation is that our constructs do not provide pst-1 function at the correct time or place. Further experiments are required to resolve this issue.
The two putative PAPS transporters PST-1 and PST-2 act non-redundantly in vivo
The lethality observed in pst-1 mutants indicates that the two genes coding for the predicted PAPS transporters PST-1 and PST-2 serve non-redundant functions. Such lack of redundancy could result from non-overlapping cellular expression, non-overlapping subcellular localization or functional non-redundancy. To address this question, we first determined in which tissues pst-1 and pst-2 are expressed. To this end, we constructed transgenic worms carrying reporter constructs that monitor transcription of both genes by expression of yellow (YFP) or cyan (CFP) fluorescent protein (Fig. 5A). Visible expression of the pst-1 reporter is first seen at around the twofold stage. Expression of the pst-1 reporter is detected in the lateral blast cells that will later either fuse to the hypodermal syncytium or give rise to seam cells, a group of specialized hypodermal cells (Fig. 5B). Expression continues to be detected throughout life in the seam cells and in pharyngeal gland cells. In addition, weaker expression is seen in a number of unidentified cells in the head, possibly in amphid support cells (Fig. 5B) and the duct and/or pore cell, as well as, possibly, in the intestine (data not shown). Expression of the pst-2 reporter appears generally weaker than the pst-1 reporter. It is first seen at around the 350-cell stage and can be detected in ectodermal cells (Fig. 5C). During larval and adult stages, pst-2 reporter expression appears restricted to the intestine and, in the adult stage, also to the toroidal epithelial cells that constitute the vulva (Fig. 5C). In addition, the pst-2 reporter appears to be very weakly expressed in hypodermal tissues. In summary, our expression analyses indicate that whereas pst-1 and pst-2 expression appears to be non-overlapping during postembryonic stages (except for possible intestinal coexpression), coexpression of both genes might occur during embryonic stages.
To test whether PST-1 and PST-2 are present in the same subcellular compartment in vivo we conducted colocalization studies. We constructed transgenic animals that expressed fluorescently tagged versions of PST-1 (YFP) and PST-2 (CFP) together with a Golgi marker fused to a red fluorescent protein (mCherry) (Fig. 6A). We found that pst-1 and pst-2 reporter fusions only partially colocalize with both each other (Fig. 6B) and a Golgi marker (Fig. 6C,D). These results indicate that PST-1 and PST-2, despite their similar activities in vitro, might have distinct functions in the Golgi apparatus.
Because the two putative PAPS transporters PST-1 and PST-2 could be coexpressed, both on the cellular and subcellular level, we asked whether pst-2 can complement pst-1 function in a pst-1 mutant background. We created transgenic animals in which either pst-1 or pst-2 cDNAs were driven by a short pst-1 promoter fragment that drives expression in seam cells, amphids, amphid support cells and intestinal cells (supplementary material Fig. S7). As expected, the pst-1 cDNA could rescue both the lethality and the kal-1-dependent neuronal phenotype of the pst-1(ot20) mutant (Fig. 4C,D). By contrast, the pst-2 cDNA driven under the identical pst-1 promoter fragment was not nearly as efficient at rescuing either the lethality or the neuronal phenotype (Fig. 4C,D). The fact that pst-2 can partially complement loss of pst-1 under certain experimental conditions is in accord with the interpretation that pst-2 encodes a bona fide PAPS transporter, but that the two genes serve distinct and non-redundant functions under physiological conditions.
pst-1 is required for heparan sulfation
The function of pst-1 could be required for either heparan sulfation or tyrosine sulfation, or both. To distinguish between these possibilities, we conducted three sets of experiments. First, we determined whether pst-1 mutants phenocopy loss-of-function phenotypes of a tyrosine protein sulfotransferase, a result we would predict if the PAPS transported by PST-1 is being utilized for tyrosine sulfation. RNAi-mediated knockdown of the gene encoding tyrosine protein sulfotransferase (tpst-1) results in suppression of the rolling phenotype caused by mutations in rol-6, a gene coding for a cuticle collagen (Kim et al., 2005). We find that RNAi against tpst-1 but not pst-1, pst-2 or pst-1 in combination with pst-2 efficiently suppresses the rolling phenotype caused by a mutation in rol-6 (Fig. 7A). To further corroborate this result, we constructed a double-mutant between rol-6 and the temperature-sensitive allele pst-1(ot20), which is a strong loss-of-function allele at the non-permissive temperature (Fig. 2G). We found that pst-1(ot20) at the non-permissive temperature barely suppresses the rolling phenotype of rol-6 mutants (Fig. 7B). Taken together, we conclude that pst-1 has, at best, minor functions in tyrosine sulfation.
We next used genetic double- and triple-mutant analyses to test whether pst-1 and genes in the HS sulfation pathway act in the same genetic pathway. We made use of a double mutant in two HS sulfotransferases (hst-2 and hst-6) that is severely defective in HS sulfation (R.A.T. and H.E.B., unpublished results). If a pst-1; hst-6 hst-2 triple-null mutant were more severe than either the pst-1 single or the hst-6 hst-2 double mutant, we would conclude that pst-1 acts in a pathway that is in parallel to the heparan sulfation pathways compromised in the hst-6 hst-2 double mutant. By contrast, no enhancement in the triple mutant would indicate that pst-1 acts genetically within the heparan sulfation pathways, at least in the cellular contexts analyzed. We found no enhanced defects in a pst-1; hst-2 hst-6 triple-null mutant compared with the pst-1 single-null mutant for two different aspects of touch neuron development (Fig. 7C,D). This demonstrates that pst-1 acts genetically in the same pathways as hst-2 and hst-6 and indicates that tyrosine sulfation might play no major role in neuronal development, at least in the cellular context (PLM) tested here.
Third, we asked whether pst-1 is required for the synthesis and modification of heparan sulfate in vivo. To this end, we directly analyzed HS from pst-1 mutant animals and determined the relative amount of each HS disaccharide that results from complete enzymatic depolymerization with bacterial heparinases. If pst-1 function were required for heparan sulfation, we would expect an alteration in the composition of sulfated disaccharides purified from mutant animals. We found that the composition of sulfated disaccharides is altered in pst-1(ot20) mutants (Fig. 7E,F). The fraction of monosulfated disaccharide (D0S0; ΔUA-GlcNS) is increased whereas the fraction of di-sulfated (D2S0; ΔUA2S-GlcNS) and tri-sulfated (D2S6; ΔUA2S-GlcNS6S) disaccharides is reduced. Additionally, the 6-O, N-sulfated disaccharide D0S6 (ΔUA-GlcNS6S) becomes apparent in the pst-1 mutant but is virtually undetected in HS from wild-type animals (Fig. 7E,F).
The extent of sulfation of HS can be expressed as the mole of O-sulfates divided by the mole of sulfated disaccharides. We use this ratio, which we term the `sulfation extent' as a measure of operational complexity. The sulfation extent is reduced by 37% (from 0.458 to 0.288) in pst-1 mutant animals compared with wild type (Fig. 7G). This indicates that HS from pst-1(ot20) mutant animals contains fewer disaccharides with two or more sulfates, resulting in HS sulfation patterns of reduced complexity in pst-1 mutant animals. Taken together, our analyses demonstrate that pst-1 is required in vivo for sulfation of HS and that the reduced complexity of HS modification patterns in pst-1(ot20) mutant animals correlates with defects in neuronal development (Fig. 3, Table 2).
Sulfation and development
Our analyses show that sulfation plays an essential role during embryonic and larval development and organogenesis. Early embryonic development in mutant embryos appears to proceed normally during gastrulation and morphogenesis and we observe no apparent defects in tissue organization at these early embryonic stages. Instead, we found that the majority of embryos arrest after the twofold stage, often with severely disorganized hypodermal tissues (Fig. 2), that is after the initial stages of embryonic elongation. Similar defects with late arrest after the twofold embryonic stage have been described for mutants of extracellular matrix components such as α(IV)collagen. These mutants display defects in muscle-hypodermal interactions and attachment (Gupta et al., 1997). Similarly, the pharynx phenotype that we observe in larvae is reminiscent of loss-of-function phenotypes in mutants of laminin and genes required for the synthesis of HS (Franks et al., 2006; Huang et al., 2003). Our findings are consistent with the notion that the extracellular matrix plays crucial roles in the attachment between tissues, and that sulfation, possibly of HS, is a pivotal part of extracellular matrix function in these processes. Whether HS is involved in mechanically attaching tissues or functions by modulating extracellular signaling remains to be determined.
Intriguingly, we can rescue the cold-sensitive embryonic lethality of pst-1 mutants by expression of pst-1 from any tissue that we tested, and in as few as two cells (AIY interneurons) (Fig. 4, supplementary material Fig. S6). Given the predicted localization of pst-1 to the Golgi, this finding indicates non-autonomous rescue of the lethal phenotype. In flies, defects in formation of the embryonic dorsal-ventral axis as a result of a mutation in the gene encoding PAPS synthase can also be rescued non-autonomously (Zhu et al., 2007). These findings imply a secreted and sulfated compound that is indispensable for viability and requires pst-1. This compound might be a sulfated protein, sugar or PAPS itself. The C. elegans PAPS synthase is expressed predominantly in seam cells and a number of gland cells (Dejima et al., 2006), a pattern that is strikingly similar to the expression patterns we observed for the reporter of the PAPS transporter PST-1 (Fig. 5B). The orthologous Drosophila PAPS transporter slalom also displays very restricted expression in the salivary gland placode (Lüders et al., 2003). Together, these results suggest that not all cells produce PAPS but that rather there might be PAPS `factories'. From these cells, PAPS could diffuse either through GAP junctions or be secreted through the action of pst-1 and subsequently taken up (by unknown mechanisms) from the extracellular environment by cells that utilize PAPS to provide essential functions. Alternatively, PAPS itself could play an essential role in the extracellular space as a sulfate donor of an as-yet-unknown sulfotransferase reaction, or function through unknown mechanisms.
pst-1 and sulfation
Our analyses of the development of the nervous system in animals that lack pst-1 show a restricted set of defects (Table 2, Fig. 3). These studies document that sulfation is a crucial aspect of neuronal development. But is pst-1 required for sulfation of glycosaminoglycans (GAGs) such as HS or tyrosine residues in proteins or, possibly, both? We provide evidence that mutations in pst-1 primarily affect HS sulfation in C. elegans (Fig. 7). First, loss of pst-1 function does not phenocopy loss of tyrosine sulfation. Second, the null allele of pst-1 cannot be enhanced by null alleles in genes coding for HS sulfotransferases, arguing that pst-1 acts genetically in the same pathway and is not required for tyrosine sulfation, at least in the cellular contexts that we tested. Third, our biochemical analyses show that the complexity of HS sulfation patterns is reduced in pst-1 mutants. We observed a significant shift from di- and tri-sulfated disaccharides to mono-sulfated disaccharides. Thus, our analysis of the pst-1 hypomorphic allele provides in vivo evidence for the observation that, in vitro, the concentration of PAPS can influence the structure of HS modification patterns, possibly by modulating HS sulfotransferases such as the HS deacetylase/N-sulfotransferase (encoded by NDST/hst-1) (Carlsson et al., 2008).
All metazoan genomes contain two genes (PAPST1 and PAPST2) coding for the closely related presumptive PAPS transporters, PAPST1 and PAPST2. Both proteins belong to distinct subfamilies of solute carrier proteins and both have been shown to preferentially transport PAPS across membranes in vitro (Kamiyama et al., 2006; Kamiyama et al., 2003). Our finding that, in C. elegans, pst-2 cannot efficiently complement for loss of pst-1 in at least two cellular contexts suggests that pst-1 and pst-2 might not be functionally redundant under physiological conditions. Together with the observation that pst-1 appears to be predominantly involved in HS sulfation and less so in tyrosine sulfation (Fig. 7), this raises the intriguing possibility that in vivo pst-1 might primarily function in the sulfation of GAGs such as HS, whereas pst-2 might be involved in other sulfation processes such as tyrosine sulfation. For example, GAG and tyrosine sulfation might be compartmentalized within the Golgi apparatus such that both PST-1 and PST-2, in conjunction with specific sulfotransferases, form complexes that modify the respective macromolecules. Yet, the activities of pst-1 and pst-2 might not be completely specific because RNAi-mediated knockdown of PAPST1 in cell culture shows a small effect on tyrosine sulfation (Kamiyama et al., 2006) and pst-1 mutants display a 3% suppression of the rolling phenotype (Fig. 7B). This interpretation is also consistent with the partial colocalization that we observe between a PST-1 and a PST-2 fusion (Fig. 6). Because RNAi-mediated knockdown of pst-2 results in no visible phenotypes [(Gönczy et al., 2000; Kamath and Ahringer, 2003; Rual et al., 2004; Sonnichsen et al., 2005), Fig. 7 and data not shown], mutant analysis of pst-2 will be required to determine the in vivo functions of pst-2.
HS sulfation and the nervous system
We have previously shown that HS modifications play specific and instructive roles during development of the nervous system in C. elegans (Bülow and Hobert, 2004; Bülow et al., 2008). This has lent support to the `HS code' hypothesis, which postulates that diverse HS sugar modifications might provide information that is crucial for the determination of cell-specific interactions (Bülow and Hobert, 2006; Holt and Dickson, 2005). Our findings here show that by manipulating PAPS transport we can selectively simplify HS sulfation patterns, and thus at least partially erase the postulated HS code. Interestingly, the overall organization of the nervous system is intact in pst-1 null mutants and we detect only very specific defects in neural development. Taken together, our results suggest that the molecular complexity of the HS code is required in vivo to ensure the formation of appropriate neuronal connectivity with high fidelity, rather than being required for establishment of the nervous system `grid' by the major ganglia and fascicles.
Materials and Methods
OH142: otIs76 mgIs18, pst-1(ot20); BC3700: dpy-18(e364)/eT1 III; unc-46(e177) let-462(s1594)/eT1V; BC2282: dpy-18(e364)/eT1III; unc-46(e177) let-462(s1481)/eT1V; BC2391: dpy-18(e364)/eT1III; unc-46(e177) let-462(s1590)/eT1V (Johnsen and Baillie, 1991); BC2883: dpy-18(e364)/eT1III; unc-46(e177) let-462(s1956)/eT1V (Stewart et al., 1991) (BC2883 and BC2391 were kind gifts of David Baillie and Robert Johnsen, both of Simon Fraser University, Burnaby, B.C., Canada). The kal-1 misexpressing trangenic strains are: OH2374: otIs76 mgIs18IV (AIY expression of kal-1); OH2375 otIs81 (pan-neuronal expression of kal-1). Transgenic strains used for phenotypic analyses are listed in supplementary material Table S2.
Molecular cloning and transgenic rescue
The cDNAs of pst-1a, pst-1b and pst-2 were PCR-amplified with gene-specific primers and cloned under control of a hypodermis-specific dpy-7prom promoter (Gilleard et al., 1997). The transcription units were verified by sequencing and subcloned under the control of the F25B3.3prom, ttx-3prom, pst-1aprom and myo-3prom (gift of Andrew Fire, Stanford University, Stanford, California). The F25B3.3prom, ttx-3prom and myo-3prom promoters drive expression of the cDNAs pan-neuronally, in AIY interneurons and in muscle, respectively. The pst-1aprom promoter expresses in seam cells, intestine, amphids and amphid sheath cells (supplementary material Fig. S7). Recombinant DNA constructs and genomic DNA pieces were injected at 5 ng/μl or 1 ng/μl, respectively, together with pRF4 [a plasmid carrying the dominant rol-6(sy1006) allele] as injection marker at 100 ng/μl or with ceh-22::gfp (pharyngeal GFP) and pBluescript at 50 ng/μl each. Translational reporters of pst-1, pst-2 and mannosidase II were made by a PCR fusion technique (Hobert, 2002) and injected at 10 ng/μl together with ceh-22::gfp (pharyngeal GFP) and pBluescript at 50 ng/μl each into BC2883: dpy-18(e364)/eT1III; unc-46(e177) let-462(s1956)/eT1V. The PST-1A::YFP translational fusion does not rescue the lethality associated with the pst-1/let-462(s1956) null allele.
Transcriptional expression constructs for pst-1 and pst-2 were engineered using a recombineering protocol (Tursun et al., 2009). Briefly, the intercistronic region of a C. elegans operon was inserted immediately after the stop codon of the gene of interest, followed by the coding region of a fluorescent protein. This construct represents an artificial operon and should result in a multicistronic transcript of the gene of interest (pst-1 or pst-2) and the fluorescent protein. The `downstream' gene (i.e. the fluorescent protein) is transpliced to a SL2 leader sequence, which is a hallmark of operonic transcripts in C. elegans. Constructs were verified by sequencing and injected into N2 wild-type animals (20 ng/μl together with 100 ng/μl pRF4).
Positional cloning of ot20
Using suppression of kal-1-dependent axonal branching (Bülow et al., 2002), ot20 was placed by STS mapping into a 10.3 MU interval between stP3 and bP1 on LGV. Using the co-segregating cold-sensitive lethality at 15°C we further mapped ot20 into a 100 kb interval (Fig. 1A) using the polymorphic strain CB4856. The region contains 20 predicted genes, two of which (acy-2 and pcm-1) could be excluded by complementation tests. Transformation rescue experiments narrowed the region to a single gene (M03F8.2a/b) coding for the C. elegans PAPS transporter 1, which we name pst-1. ot20 is a missense mutation changing a thymidine residue to a cytidine at position 14929 in cosmid M03F8. In addition, let-462 alleles harbor mutations in the coding region of pst-1: s1480, and s1590 are missense mutations that change guanosine to adenosine at position 16931, and 14666 in cosmid M03F8, respectively. The predicted null allele s1956 changes glycine to threonine at position 14639 in cosmid M03F8. No molecular lesion could be identified in the s1594 allele.
Molecular characterization of pst-1
Sequencing of 14 EST clones (yk715c1, yk1139c05, yk1407h04, yk1471c09, yk1607g07, yk1612a06, yk1413f03, yk1592d02, yk1408e08, yk1471g07, yk1754c08 and yk1724h04 kindly provided by Yuji Kohara, National Institute of Genetics, Japan) revealed that pst-1 exists in at least three alternatively spliced variants pst-1a, pst-1b and pst-1c. The pst-1a isoform is SL2 trans-spliced. The pst-1a and pst-1b forms differ in their usage of an alternative first exon and encode two proteins of 424 and 439 amino acids, respectively. pst-1c contains the 1B exon but is alternatively spliced to exon 8, resulting in a shorter protein of 408 amino acids.
Heparan sulfate analysis
Preparation of GAGs for compositional analysis
For GAG production, worms were grown in liquid culture (Stiernagle, 1999). After lyophilization, worms were resuspended in 10 ml of acetone and disrupted with a Fisher brand Polytron for 3-5 minutes. The homogenate was de-lipidated in acetone, clarified by centrifugation and lyophilized. Worm powder (150-300 mg) was solubilized and crude GAGs were purified as described (Toyoda et al., 2000). Crude GAGs were incubated with 800 μl of diethylaminoethyl cellulose for 1 hour at 4°C. The resin was washed three times with 5 ml of 150 mM NaCl, 50 mM Tris-HCl pH 8.0. Purified GAGs were eluted with 800 μl of 1 M NaCl, 50 mM Tris-HCl pH 8.0 for 10 minutes (repeated three times). Eluates were pooled, desalted on a PD-10 column (GE Healthcare), lyophilized and resuspended in 30 μl of distilled H2O.
A 1 μl aliquot of the sample was digested with 25 mIU of Chondroitinase AC (Ibex) in 30 mM Tris-acetate pH 8.0 for 6 hours at 37°C. The remaining sample was digested with a mixture containing 1 mIU each of Heparinase I, Heparinase II and Heparinase III (Ibex) in 20 mM Tris-acetate pH 7.0, 2 mM calcium acetate for 6 hours at 37°C. After digestion, the chondroitin and HS disaccharides were isolated by filtration using an YM-3 microcon filtration device (Fisher).
The protocol for determining the unsaturated disaccharide composition of heparan sulfate was modified from the method previously described (Toyoda et al., 2000). The chromatographic equipment included a U3000 HPLC (Dionex), two single piston pumps (Eldex Laboratories), a RF2000 fluorescence detector (Dionex) with a 12 μl flow cell volume, a dry reaction bath (FH-40) and a thermocontroller (TC-55) (Brinkman Instruments). Samples were analyzed using a Senshu Pak C22 3 μm Docosil column (2×100 mm) or a Dionex C18 3 μm Acclaim 120 column (2.1×150 mm). The flow rates were 0.300 ml/minute for the U3000 gradient pump and 0.175 ml/minute for the reaction pumps. Buffer A: H2O. Buffer B: 200 mM NaCl. Buffer C: 10 mM tetrabutylammonium hydrogensulfate. Buffer D: 50% acetonitrile. Injection volume was 20 μl. Initial conditions were 1.0% B, 12.0% C, 17.0% D. Gradient was as follows: 1-55% B over 9.0 minutes, 55-70% B over 1.5 minutes, 70% B for 4 minutes, 1% B for 10 minutes. C and D were constant at 12% and 17%, respectively. Post-column derivatization was achieved by adding 1.0% NaOH and 0.5% 2-cyanoacetamide to the effluent from the HPLC and passing the mixture through a reaction coil (0.25 mm inner diameter × 50 feet) at 125°C. Peaks were identified fluorometrically (excitation wavelength, 346 nm; emission wavelength, 410 nm) and quantified using known unsaturated chondroitin and HS disaccharide standards (V-labs, Covington, LA).
Phenotypic analyses and microscopy
All mutant alleles were backcrossed at least three times before further analyses. Unless indicated otherwise, experiments were performed at 20°C in young adults (ot20) or the L1 stage (s1956) and images were acquired using an Axioimager Z1 microscope Apotome (Zeiss) using Zeiss Axioimaging software. All confocal images were acquired on a LEICA SP 2 confocal microscope equipped for sequential scan of the emission lines. For colocalization analysis, CFP was excited with a 405 nm laser, YFP with the 514 nm line, and RFP at 561 nm. Images were exported as TIF files; channels were merged where applicable using ImageJ (NIH, Bethesda, MD) and assembled in Adobe Illustrator.
Synaptobrevin::GFP expression was quantified as described (Schuske et al., 2007) with minor modifications. Images of mutant and control worms at the L1-L2 stage were taken using identical settings on the same day after the laser had stabilized. A maximum intensity projection of the whole ventral nerve cord consisting of three to four optical sections was usually analyzed for each worm. The ventral nerve cord was outlined for each worm and the relative fluorescence was calculated as the mean fluorescent value per pixel of the cord (three to eight observations per animal, excluding the cell bodies) using the Image J software.
Temperature-shift experiments were performed with animals grown at 20°C. After egg preparation, animals were either directly transferred to the non-permissive temperature at 15°C or synchronized by allowing them to hatch without food overnight. The next day, animals were moved to food at 20°C and subsequently transferred to the non-permissive temperature at 15°C at different developmental stages. Four days later, the percentage of animals still alive was recorded. Statistical significance was calculated using the Z-test and is indicated in all panels as: ns, not significant; *P<0.05; **P<0.005; ***P<0.0005. The Bonferroni correction was applied where necessary.
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/122/24/4492/DC1
We are grateful to Oliver Hobert at Columbia University (New York) in whose laboratory this study was initiated. We thank Qi Chen, Matthew Attreed, Jason Brandler and Peter Weinberg for expert technical assistance. We thank Anna Kalinowski, and Wendy Yip for help during SNP mapping of ot20, Dominic Didiano for RNAi experiments, Yuji Kohara for yk clones, Hanna Fares and Baris Tursun for DNA constructs and several colleagues, in particular David Baillie, Robert Johnsen and the C. elegans Genetics Center for strains. Some of the strains are part of the Genetic Toolkit, which is funded by the NIH National Center for Research Resources (NCRR). We thank Oliver Hobert, Alicia Meléndez, Ji Y. Sze and members of the Bülow laboratory for comments on the manuscript. This work was supported in part by grants from the National Institutes of Health (5R01HD055380 to H.E.B.; 5T32NS07098 to R.A.T.). H.E.B. is an Alfred P. Sloan scholar. Deposited in PMC for release after 12 months.
- Accepted September 23, 2009.
- © The Company of Biologists Limited 2009