A Synthetic Lethal Screen Identifies a Role for Lin-44/Wnt in C. elegans Embryogenesis

Background The C. elegans proteins PTP-3/LAR-RPTP and SDN-1/Syndecan are conserved cell adhesion molecules. Loss-of-function (LOF) mutations in either ptp-3 or sdn-1 result in low penetrance embryonic developmental defects. Work from other systems has shown that syndecans can function as ligands for LAR receptors in vivo. We used double mutant analysis to test whether ptp-3 and sdn-1 function in a linear genetic pathway during C. elegans embryogenesis. Results We found animals with LOF in both sdn-1 and ptp-3 exhibited a highly penetrant synthetic lethality (SynLet), with only a small percentage of animals surviving to adulthood. Analysis of the survivors demonstrated that these animals had a synergistic increase in the penetrance of embryonic developmental defects. Together, these data strongly suggested PTP-3 and SDN-1 function in parallel during embryogenesis. We subsequently used RNAi to knockdown ~3,600 genes predicted to encode secreted and/or transmembrane molecules to identify genes that interacted with ptp-3 or sdn-1. We found that the Wnt ligand, lin-44, was SynLet with sdn-1, but not ptp-3. We used 4-dimensional time-lapse analysis to characterize the interaction between lin-44 and sdn-1. We found evidence that loss of lin-44 caused defects in the polarization and migration of endodermal precursors during gastrulation, a previously undescribed role for lin-44 that is strongly enhanced by the loss of sdn-1. Conclusions PTP-3 and SDN-1 function in compensatory pathways during C. elegans embryonic and larval development, as simultaneous loss of both genes has dire consequences for organismal survival. The Wnt ligand lin-44 contributes to the early stages of gastrulation in parallel to sdn-1, but in a genetic pathway with ptp-3. Overall, the SynLet phenotype provides a robust platform to identify ptp-3 and sdn-1 interacting genes, as well as other genes that function in development, yet might be missed in traditional forward genetic screens.


Introduction
Cell adhesion molecules (CAMs) provide multiple functions during the development and homeostasis of an organism. In C. elegans, multiple CAMs contribute to early embryonic development and loss-of-function (LOF) mutations in these can result in cellular, tissue and/or organismal abnormalities [1][2][3][4]. Interestingly, these molecules often appear to act in semi-redundant ways, where input from multiple CAMs are required for the fidelity of a specific developmental event [3,[5][6][7]. This can be best observed when LOF in a single gene has a modest effect on viability, but LOF in two genes in combination can have severe effects leading to highly penetrant lethality or arrest. This synergistic effect, known as Synthetic Lethality (SynLet), can be harnessed to uncover genetic interactions between functional pathways [8][9][10][11][12][13][14].
Syndecans are cell-surface associated HSPGs that have been implicated in a broad range of developmental events, and have been linked to the modulation of several secreted morphogens including Wnts and Fibroblast Growth Factors (FGFs) [36][37][38][39][40][41][42][43]. The extracellular domain of syndecans can be post-translationally modified with HS-or CS-side chains and the intracellular domain can interact with cytoplasmic signaling effectors via a PDZ binding motif. The C. elegans genome encodes a single syndecan, SDN-1, which, by sequence, is most similar to human syndecan-2 (SDC2) [44][45][46].
Here we provide genetic evidence that ptp-3B and sdn-1 function in parallel signaling pathways during C. elegans embryonic development. ptp-3; sdn-1 double mutants exhibit a highly penetrant SynLet phenotype, with development arresting during embryogenesis or in the first larval stage (L1). A small percentage of animals do progress to adulthood, but exhibit sterility or low fecundity with all offspring arresting during development.
Using an RNAi library comprised primarily of predicted secreted proteins, we screened for genes that exhibited a SynLet phenotype with worms homozygous for either ptp-3 or sdn-1 LOF mutations. From a screen of 3,652 clones, we isolated 25 candidate SynLet genes, and several additional candidate genes displayed an increase in lethality or slow growth phenotype in either the ptp-3 or sdn-1 background, but did not meet our threshold criteria for synthetic lethality. Among these, we found that the Wnt ligand, lin-44, was strongly SynLet with sdn-1, but not ptp-3. Using a time-lapse microscopy approach we found defects in the ingression of the endodermal precursor cells Ea and Ep in lin-44 mutants. This phenotype was significantly enhanced when sdn-1 was also removed. This is the first data indicating that LIN-44 contributes to gastrulation events in C. elegans and demonstrates the power of using double mutant analyses to uncover novel roles for well-characterized genes.

RNAi feeding control strains used
RNAi clones were compiled from the Ahringer library [48]. Three RNAi controls were used during each round of screening, an empty vector (L4440), and clones targeting ptp-3 (II-7J03; Overlapping CDS: C09D8.1) or sdn-1 (pEVL202). The sdn-1 fragment designated for RNAi was obtained by polymerase chain reaction (PCR) from N2 genomic DNA. The fragment was then cloned into the pCR8 TOPO cloning vector (Life Technologies) and recombined via an LR reaction into a dual T7-feeding vector derived from the L4440 vector (a kind gift of A. Fire). The resulting plasmid was transformed into the HT115(DE3) RNase III-deficient E. coli strain [49]. The following primer pairs were used for PCR amplification of sdn-1: forward 5'-TTTTCTTTTAGAACCCTTTTGC-3' reverse: 3'-CATCAATTTATCATCTCGCAAC-5'.
RNAi feeding assay (6-well format) HT115 bacterial strains, containing the RNAi clones of interest, were grown overnight at 37°C in 1.5 mL LB plus ampicillin, tetracycline and nystatin. 0.1 ml of the overnight cultures were aliquoted on single 6-well plates of NGM containing carbenicillin, tetracycline, and 1mM IPTG and grown overnight at 37°C. Approximately 3-4 L4 worms were dispensed in M9 into the top wells (1, 2 and 3) on the 6-well plates. The plates were left at approximately 20°C for five days before the worms were scored for phenotypes and 3-4 L4 worms were transferred to the bottom wells (4, 5 and 6). The bottom wells were then scored after six more days. The experimenter was blind to the RNAi clone being tested in all assays. All clones that exhibited any level of SynLet were rescreened as described above to verify the interaction. Clones were then rescreened if they displayed SynLet or slow growth with any of the query strains. We categorized SynLet as embryonic lethal (Emb), larval lethal (Lvl) or adult lethal (Adl). The following additional phenotypes were also scored: Lethal (Let), body morphology defects (Bmd), uncoordinated (Unc), sickness (Sck), sterility (Ste), slow growing (Gro), protruding vulva (Pvl), dumpy (Dpy) and/or egg laying defective (Egl). All phenotypes were compared between the strains on that 6-well plate alone.

RNAi clone sequencing
Clones isolated from the screen were grown as single clones in liquid LB cultures containing ampicillin, tetracycline and nystatin. Cultures were purified using a Qiagen spin mini-prep kit then sequenced using a modified forward T7 primer (5'-ACTCACTATAGGGAGACCGG-3').
Time-lapse microscopy of embryonic development 4-dimensional video microscopy was carried out using an Olympus BX61 microscope equipped with a 63x magnification oil immersion objective and motorized z-axis stage. Z-stacks (27-35 slices) were collected every 2 minutes at 1 micron spacing using a Retiga camera (Q-Imaging). Data sets were analyzed via the Bioformats-Importer and 4D Browser plugins in ImageJ [50,51]. The following time points were collected; Ea/Ep ingression, gastrulation cleft opening, cleft closure, and comma stage. Analysis of developmental time points was normalized relative to Ea/Ep ingression and comma stage. Embryonic lethal phenotypes were categorized according to [52]. Ea/Ep ingression behavior was scored as follows; I, normal (Ea/Ep ingress together); II, no ingression (Ea/Ep remain on the outer surface of the embryo); III, skewed (either Ea or Ep ingresses before its partner); IV, unclassified (Ea/Ep migration obscured or unclear).

Lethality analysis
A single L4 hermaphrodite was placed on a single NGM plate to initiate the assay. Every 24 hours the hermaphrodite was transferred to a new plate. 24 hours after the removal of the hermaphrodite the plate was analyzed for embryos (Emb) or dead L1 larvae (Lva). The plate was rescreened 24 hours later for later larval lethality (Lva). Adults were transferred until they died, or until they stopped giving rise to offspring. All animals were maintained at 20°C, except during scoring. Experimenter was blind to the genotype during scoring.

Epithelial morphology assays
Epithelial tight junctions were visualized with jcIs1 [P-ajm-1-AJM-1::GFP], which localizes to epithelial junctions, essentially outlining all epithelial cells from about the lima bean stage onwards. Multiple embryos were harvested by bleaching and > 100 assayed for epithelial morphology. Embryos with cells that were obviously mispositioned, (e.g. dorsal cells on the ventral side), or misshapen in a way that was strikingly different from wild-type, (e.g. rounded cells along the lateral aspect where cuboidal cells are normally present), were scored as a mutant.

Results and Discussion
Dual loss of ptp-3 and sdn-1 results in synthetic lethality Both ptp-3 and sdn-1 have previously described roles in C. elegans embryonic development [5,7,44,53]. The ptp-3 locus encodes three distinct transcripts, each with independent promoters [15] (Fig 1). Mutations in ptp-3B exhibit a low level of embryonic (Emb) and larval lethality (Lva) as well as Variable-abnormal body morphology defects (Vab) (Fig 2 and Table 1). However, most ptp-3 LOF mutants are superficially normal in appearance. Similarly, sdn-1 mutants exhibit low levels of Emb and Lva offspring (Table 1), but most grow to adulthood, where they exhibit uncoordinated movement (Unc) and egg-laying defects (Egl).
Based on the low penetrance viability defects in ptp-3 mutants, and previous observations that the Drosophila LAR receptor, DLAR, can bind syndecan [33,34], we asked whether ptp-3 and sdn-1 could be interacting genetically during C. elegans development. If PTP-3 and SDN-1 were acting as a ligand-receptor pair, we would have expected that ptp-3; sdn-1 double mutants would exhibit embryonic lethality at a rate similar to the single mutant genetic backgrounds. In contrast we found that animals lacking both sdn-1 and ptp-3 were essentially inviable (Table 1). To score the development of these animals we generated sdn-1(zh20) homozygotes where a strong ptp-3 LOF mutation, mu245, was balanced with a chromosomal inversion, mIn1, which is marked with a recessive mutation (dpy-10), and a dominant pharyngeal GFP insertion, mIs14 (phGfp) (see Materials and methods).
We found an increase in the embryonic and larval lethality in offspring from sdn-1(zh20); ptp-3(mu245)juIs76/mIn1mIs14 mothers compared to sdn-1(zh20); juIs76 or ptp-3(mu245) juIs76 alone. Compared to the expected 25% of the brood, only 12.6% of live hatchlings were ptp-3; sdn-1 (126/1000 offspring from a total of 5 mothers). Further, when these broods were analyzed on the first day of adulthood, only 2% (2/114 offspring) of the adults were sdn-1 (zh20); ptp-3(mu245)juIs76). Surviving sdn-1(zh20); ptp-3(mu245)juIs76 adults appeared sickly, were largely paralyzed and had no viable offspring (78 Emb and 5 Lva L1s from 4 mothers). SynLet phenotypes were also observed when we tested other strong ptp-3 LOF mutations  ptp-3(mu245) mutant animals can hatch as normal looking L1 larvae (arrow), but can also die during embryogenesis (asterisk shows an embryo that ruptured at elongation). B: ptp-3 mutants can exhibit defects in ventral neuroblast migration (B1 outlined area), which result in a persistent gastrulation cleft on the ventral surface. When these fail to close, internal cells extrude through the opening (arrowheads) during elongation, including ptp-3(mu256) and ptp-3(op147), and a second deletion allele in sdn-1, ok449. None of the ptp-3; sdn-1 double mutant combinations were viable as homozygous strains, and each had to be maintained as balanced ptp-3/mIn1mIs14 heterozygotes for propagation. In addition, RNAi targeting sdn-1 was lethal in ptp-3(mu245) animals, and RNAi targeting ptp-3 was lethal in sdn-1(zh20) animals. These data confirm that the SynLet phenotypes observed are due to the alleles in question and not caused by closely linked background mutations.
We more closely analyzed the sdn-1(zh20); ptp-3(mu245) homozygous animals to determine the point at which they were arresting. We found instances of early embryonic arrest (pre-morphogenesis), embryonic rupture during epidermal enclosure, and arrest at the L1 stage as misshapen larvae. These are similar arrest points to those observed in embryos from sdn-1(zh20); ptp-3(mu245)/mIn1mIs14 mothers. However, the higher rate of embryonic arrest in the offspring of sdn-1(zh20); ptp-3(mu245) animals than from sdn-1(zh20); ptp-3(mu245)/mIn1mIs14 mothers strongly suggests a maternal contribution of ptp-3 to embryonic development. Since the offspring of sdn-1(zh20); ptp-3(mu245) animals, which should lack any maternal contribution, demonstrated a variable arrest point, we concluded there are likely several stages of development to which both SDN-1 and PTP-3 contribute, in partially compensatory ways. However, we cannot completely discount that defects that occur early in development may present as variable arrest phenotypes. Overall these results indicate that ptp-3 and sdn-1 have overlapping function during embryogenesis and that loss of both genes results in dire consequences for organismal survival.
We previously demonstrated that two of the isoforms produced by the ptp-3 locus have differential localization [5,15]. PTP-3A localization is restricted to synapses, while PTP-3B is leading to embryonic lethality. C: ptp-3 mutants can also exhibit variably abnormal herniations in body morphology (empty arrowhead). D: sdn-1(zh20) mutants also exhibit low penetrance embryonic lethality (asterisks show ruptured embryos). E: Gastrulation cleft closure defects are observed in some sdn-1(zh20) mutant embryos (outlined area). Again, this can lead to embryonic rupture at ventral enclosure (arrowhead shows cells oozing from within). associated with cell-cell junctions during embryogenesis but also localizes to axons during axon outgrowth. Consistent with their different localizations, PTP-3A and PTP-3B appear to function as distinct genetic units as loss of ptp-3A has no embryonic patterning or axon guidance defects, yet exhibits a fully penetrant synaptic morphology defect that is equivalent to a complete loss of function for the ptp-3 locus [15]. Similarly, PTP-3B is capable of rescuing the embryonic development, cell migration, and axon outgrowth defects associated with ptp-3 LOF mutations [1,5,17,54]. The ptp-3(mu245) lesion specifically affects the PTP-3A and PTP-3B isoforms, and we did not observe a SynLet phenotype when we used a LOF mutation in ptp-3 that specifically affects the PTP-3A isoform, ptp-3A(ok244) (Fig 1). This is consistent with our previous results suggesting that PTP-3A does not obviously contribute to epidermal development, while PTP-3B does.
To better understand what might be contributing to ptp-3; sdn-1 lethality during embryogenesis, we used 4D time-lapse microscopy to observe cell division and migration during the first 10 hours of embryonic development. The first cell migration event that occurs during C. elegans development is the onset of gastrulation, where the gut precursor cells, Ea and Ep, ingress into the center of the embryo [55]. In wild-type embryos, Ea and Ep ingress in concert, side-by-side, with the space vacated by their ingression filled by movements from six surrounding cells [56]. The next major developmental event occurs when cells of the endodermal and mesodermal lineages begin to ingress at the posterior end, leaving a transient gastrulation cleft on the ventral surface (Fig 3) [57]. Onset of cleft opening was phenotypically normal in the embryos observed (n = 10) although one embryo showed gastrulation cleft opening at the anterior end. However, the relative timing of gastrulation cleft opening was significantly delayed in ptp-3 embryos when compared to Ea/Ep ingression and comma stage. The gastrulation cleft is flanked by neuroblasts; in wild-type embryos, these normally migrate towards the midline of the ventral surface, closing the cleft in about 55 minutes [53]. These subsequently form a substrate for epithelial cell migration, which intercalate and extend from the dorsal and lateral surfaces to enclose the embryo during epiboly.
In ptp-3; sdn-1 mutants, 8/14 embryos showed gastrulation clefts that persisted until epithelial extension and ventral enclosure (Fig 3). All of these embryos ruptured prior to comma stage, at the onset of embryonic elongation (class I phenotype, [52]). Two large cells were frequently seen in the center of these enlarged clefts that showed no adhesion to the cells surrounding them. Based on previous cell lineaging experiments, we tentatively identified these as the germ line precursor cells Z2 and Z3 (M. L. Hudson, unpublished observations). Additional embryos showed gross disorganization during development, with lateral loss of cell-cell contacts on the embryo surface prior to cleft opening. In the embryos that survived ventral enclosure, defects were also observed in tail morphology. While gastrulation cleft opening appears to be significantly delayed in ptp-3; sdn-1 mutants compared to ptp-3 alone, only 3/10 ptp-3; sdn-1 mutants could be scored for this phenotype as most embryos rupture prior to reaching comma stage, and hence cannot be scored for the final developmental timeline marker. As such, this apparent suppression of ptp-3 developmental timing defects may be misleading. Overall, the most common cause of embryonic lethality was failure to close the gastrulation cleft prior to ventral enclosure. These data confirm previously identified roles for SDN-1 and PTP-3 in embryonic morphogenesis [5,53], and suggest that SDN-1 and PTP-3 function in parallel, part-redundant pathways to control either cell adhesion, neuroblast migration or both.

An RNAi screen for SynLet interactions with ptp-3 and sdn-1
The observation that simultaneous LOF in ptp-3 and sdn-1 resulted in a highly penetrant Syn-Let phenotype suggested that we could use these backgrounds to identify additional genes that contribute to embryonic development [8]. We employed RNAi knockdown to systematically screen for genes that lead to a synergistic SynLet phenotype in either ptp-3(mu256) and/or sdn-1(zh20) mutant backgrounds. RNAi clones that were identified from the first round of screening were retested at least four times to confirm the results. The threshold for declaring an interaction as synthetic lethal was >75% Emb. We also identified genes that were SynLet with both ptp-3 and sdn-1, suggesting these may function in yet another independent parallel pathway, or may contribute in overlapping fashion to both the ptp-3 and sdn-1 pathways.
We found 11 genes that showed a SynLet phenotype in ptp-3(mu256) animals, but only limited or no lethality in a wild type background (Table 2, S1 Table). Two of these genes, vab-1 and unc-40, have previously been found to be SynLet with ptp-3, indicating that our screen was capable of identifying relevant genetic interactions [5,7,17,46]. We also found genes that are known to be individually lethal via complete loss-of-function mutations, including bli-3 and mek-2. However, in our assays, RNAi knockdown in the wild type background was insufficient to cause highly penetrant lethality. We conclude that this approach enabled us to identify genetic interactions that might be missed using traditional loss-of-function alleles. Seven of the genes found to be SynLet in the ptp-3 background had an attenuated effect when knocked down in the sdn-1 background (Table 2). These genes are candidates to function in sdn-1 mediated development. Of these several have been associated with the formation or function of the nervous system (including unc-40, vab-1, mek-2 and C27C7.5). Because syndecans are associated with neural development it suggests either failures in neural development can interfere with normal embryogenesis, or that these molecules perform non-neural developmental functions [2,58]. A second theme that emerged when analyzing the list is that several of the genes have an association with gametogenesis or the germline. For example, perm-4 is expressed in oocytes, and regulates an interaction with sperm, while the C04F12.7 gene is coexpressed with several sperm-specific genes. VAB-1 has also been linked to the function of germline maintenance [59]. Finally, the ZC190.5 has been previously identified as a suppressor of the egl-9 locus [60]. EGL-9 encodes a proline hydroxylase that negatively regulates HIF-1 signaling [61], but also participates in neural development [62]. Thus, it will be interesting to determine whether the identification of these genes implicates syndecan in oxygen sensing or germline development/function.
We isolated 15 clones that generated a SynLet phenotype in sdn-1 mutants, but had no effect, or an attenuated one, when knocked down in wild type animals. Of these, 11 of these were also less penetrant in ptp-3 animals, suggesting they are candidates to function in ptp-3-dependent development. One of the most interesting candidates to emerge was the Wnt ligand, lin-44 (see below). Wnt ligands contribute to multiple facets of organismal development throughout the animal kingdom. Interestingly, a recent paper describes an interaction between sdn-1 and another C. elegans Wnt ligand, mom-2, where SDN-1 concentrates the MIG-5/Dishevelled protein in early embryogenesis [62].
We also identified several genes that may contribute to post-translational modifications of proteins, possibly including Wnts, such as an O-acyltransferase (oac-9), a mannosidase (mans-1), a pterin-4-α-carbinolamine dehydratase (pcbd-1), a putative sugar transporter (B0041.5), a subtilisin-like endoproteases (bli-4) and a dual-oxidase (bli-3). It should be possible to use the lin-44-sdn-1 interaction we describe below to tease out potential contributions of these molecules to Wnt-dependent functions. Two of the genes are likely involved in mitochondrial function (stl-1, and F43G9.3), although it is unclear whether this function is contributing to the embryonic lethality when knocked down in sdn-1.
An unanticipated outcome of the RNAi SynLet screen was the discovery of three genes where RNAi was lethal to our control strain yet showed incomplete penetrance in ptp-3 or sdn-1 mutant backgrounds. Two of those genes, C18B12.4 and mnp-1 are reported to cause lethality when mutated, or when knocked down in wild-type animals via RNAi [63][64][65][66]. The third gene, F17B5.6, is predicted to code for a glycosyl transferase and has not previously been reported to have a role in embryonic development. mnp-1, which encodes a 781 amino acid protein related to the M1 family of metalloproteinases, is required during embryonic development to facilitate muscle cell migrations from lateral to dorsal and ventral positions [66]. Previous work has demonstrated that mnp-1 genetically interacts with the Eph Receptor vab-1, which is SynLet with ptp-3, suggesting that the interactions between these genes maybe more complicated than previously suggested. Interestingly, mnp-1 is predicted to be catalytically inactive due to the lack of three of four essential zinc-binding amino acids, and thus may function in more of a structural role, perhaps by occluding peptidase sites to promote structural integrity.
In the course of our screen we found other genes that interacted genetically with ptp-3 and/ or sdn-1, but these interactions were either too variable, or did not repeat in multiple assays, to formally conclude their role in embryonic and larval development (S1 Table). Some of these caused slowed growth (Gro) or an apparent sickness (Sck) that lead to decreased viability over the experimental window, while others may have caused increased changes in the body morphology defects (Bmd or Vab) (S1 Table). Although these were not included in the list of Syn-Let candidates, the genes may merit analysis in the future when attempting to further understand how PTP-3 or SDN-1 contribute to morphogenesis.

Epidermal junction defects do not correlate with SynLet phenotypes
PTP-3B is associated with cell junctions during the cellular migrations and tissue rearrangements that occur during embryogenesis, and ptp-3B mutants exhibit low-penetrance epidermal morphology (Vab) defects [5]. SDN-1 also localizes to the plasma membrane, being concentrated at cell-cell junctions in early embryos [62]. We hypothesized that lethality might arise from disruption of epidermal junction formation or patterning. To assay this we examined a marker for epidermal junctions, AJM-1::GFP, in sdn-1 or ptp-3 mutants alone and when grown on enhancer gene RNAi expressing bacteria.
In wild type animals, AJM-1::GFP can be seen accumulating at cell junctions outlining the epidermal cells, starting around the lima bean stage of embryogenesis, and persisting throughout development (Fig 4). Epidermal cell junctions in wild-type animals are well organized, and only rarely display gaps or misshapen cells. In contrast, we found that both ptp-3 and sdn-1 mutants had apparent cell-shape changes consistent with defects in either cell positioning or cell polarity (Fig 4, Table 3). Interestingly, the presence of disorganized cells as visualized by AJM-1::GFP did not correlate with the embryonic or larval lethality in the various mutant backgrounds assayed. For example, only 4% (4/100) ptp-3 mutants had obvious defects in the AJM-1::GFP pattern (Table 3). In contrast, 87% (87/100) of sdn-1 mutants exhibited abnormalities in AJM-1::GFP expression. Despite this, the rate of embryonic lethality in these two backgrounds is similar. Also, when we used RNAi to knock down gene expression in our SynLet screen, we found no correlation between the effect of RNAi on embryonic development and changes in the expression pattern of the epidermal junction marker. Overall this suggests that while these genes may contribute to the positioning of epidermal cells or the organization of epidermal junctions, the proper localization of AJM-1::GFP to these sites is insufficient to explain the lethality observed in the different genetic backgrounds. This is similar to previous reports for cell adhesion molecules in C. elegans, e.g. loss of the E-cadherin-like HMR-1, which results in lethality, yet the animals can form and maintain intact epidermal junctions, even when some junctions are failing [67].

The Wnt ligand LIN-44 functions in gastrulation
One of the strongest SynLet interactions uncovered in our screen was between the Wnt ligand lin-44 and sdn-1 (Fig 3, Table 1). Wnts and syndecan have been shown to function together in multiple developmental contexts in both C. elegans and other systems [43,46,67,68]. The C. elegans genome encodes five Wnt ligands; cwn-1, cwn-2, egl-20, lin-44 and mom-2. Of these only mom-2 has been shown to function in embryonic development, as loss of function in mom-2 results in maternal-effect embryonic lethality [56], although loss of multiple Wnts can result in a more penetrant lethality [69]. In our screen, knockdown of lin-44 caused a robust increase in the embryonic and larval lethality of sdn-1 mutants, but not in ptp-3 or wild-type animals. We did not observe lethality in sdn-1 animals treated with egl-20 or cwn-2 RNAi, while mom-2 knockdown caused embryonic lethality in all backgrounds tested. cwn-1 was not present in our RNAi library hence was not assayed.
To better understand the morphogenetic defects behind the lin-44 and sdn-1 interaction, we built and analyzed a double LOF line, lin-44(n1792); sdn-1(zh20). lin-44(n1792) is predicted to be a complete loss of function mutation at the lin-44 locus, hence this strain was genetically null for both genes in question. We found that sdn-1; lin-44 double mutants showed a significant increase in the penetrance of embryonic lethality compared to either single mutant (Table 1).
Using time-lapse video microscopy, we found highly penetrant defects in the migration of endodermal precursor cells Ea and Ep, in lin-44; sdn-1 double mutants at the 24-cell stage of development (Fig 5). In wild type animals, gastrulation begins when Ea and Ep rotate and then ingress from the surface of the embryo to the center [51]. In lin-44; sdn-1 double mutants, 48%  A: In C. elegans, gastrulation is initiated by the inward migration of the endodermal precursor cells Ea and Ep (black asterisks). In sdn-1 and ptp-3 mutant animals, the cells ingress, become completely surrounded by neighboring cells, then divide laterally (white asterisks). Note that Ea and Ep are completely internalized prior to the lateral cell division. In lin-44 and sdn-1 mutants, Ea/Ep ingression is often asynchronous. In addition, some lin-44 embryos show a more severe phenotype, where the Ea/Ep cells completely fail to ingress. The subsequent lateral cell division positions two of the daughter cells onto the surface of the embryo, generating a "Gut on the exterior", or Gex phenotype [77]. Similar but of embryos (11/24) showed defective Ea/Ep ingression, compared to 21% of lin-44 (5/23) and 20% of sdn-1 (3/15) single mutant embryos. As a comparison, we also examined ptp-3; sdn-1 embryos for Ea/Ep ingression failure. 29% (2/7) embryos showed defects in this process. In one embryo, the Ea/Ep cells completely failed to ingress, while in another, Ep ingressed before Ea. These data are not significantly different from sdn-1 mutants alone, suggesting that ptp-3 has no obvious role in early gastrulation. These defects in Ea/Ep ingression often resulted in endodermal cells appearing on the embryo's surface later in development (Gut on the exterior, or Gex phenotype), with catastrophic consequences for subsequent epithelial cell migrations ( Fig  5). Further analysis of our time-lapse data revealed that 15% (3/20) of lin-44 embryos showed defects in neuroblast migration as manifested by an increase in gastrulation cleft duration and failure of epithelial cells to enclose the embryo (Fig 3A and Fig 5). This is likely due to mis-positioned gut cells inhibiting or blocking epithelial cell migrations, or causing defects in overall embryonic organization. While the lin-44; sdn-1 Ea/Ep ingression phenotypes appear to be additive when compared to each single mutant, the increase in embryonic lethality is clearly synergistic ( Table 1). As such, it appears that small defects in Ea/Ep ingression early in development lead to severe consequences in later developmental events. Together these results indicate that both lin-44 and sdn-1 contribute to the normal migration of Ea and Ep cells at the onset of gastrulation, and that ptp-3 has no obvious role in this process. In addition, both lin-44 and sdn-1 may also be involved in controlling neuroblast migration and gastrulation cleft closure later in embryogenesis, although in the case of lin-44, we cannot rule out that defects at later time points are a consequence of earlier defects in Ea/Ep migration. In other contexts LIN-44 has been shown to signal through the LIN-17 Frizzled-like receptor. We made double mutants of lin-17 with sdn-1, but found no significant increase in the Emb phenotype compared to sdn-1 animals alone (Table 1). lin-17; sdn-1 double mutant adults were strongly Unc, and often died early, around day 2-3 of adulthood, compared to sdn-1 adults, which live for~7-10 days (B.D. Ackley unpublished observations). This suggests that LIN-44 affects gastrulation via a different receptor, and does not specifically function via LIN-17.
Finally, we examined whether ptp-3 loss of function also synergized with lin-44. We found that double mutants of ptp-3(mu245); lin-44(n1792) exhibited Emb lethality at a rate similar to ptp-3(mu245) single mutants (Table 1), although there was a slight increase in larval lethality in the double mutants compared to each single mutant background. This suggests that lin-44 and ptp-3 may be functioning in the same genetic pathway during embryonic development. Overall, our results are consistent with ptp-3 contributing to lin-44-dependent embryogenesis, and that at least one of the potential reasons that ptp-3; sdn-1 double mutants die is because of a disruption in the lin-44 / ptp-3 signaling pathway.

Conclusions
Embryonic development requires an orchestrated set of cell migrations and rearrangements, and the proper modulation of cell adhesion is critical to this process. Here we have demonstrated that the type IIa RPTP, ptp-3 and the syndecan ortholog, sdn-1, contribute to parallel genetic pathways during C. elegans embryogenesis. The dual loss of these molecules results in developmental failure due to defects in two major cellular rearrangements, gastrulation and epiboly. Both sdn-1 and ptp-3 have roles at the onset of gastrulation, which is the first major cellular rearrangement seen in C. elegans development. In addition, both are clearly required for cell migration events later in development at gastrulation cleft closure. Failure to close the gastrulation cleft leads to subsequent defects in hypodermal enclosure and developmental failure via ventral rupture at the onset of embryonic elongation.
At first glance our results seem at odds with data from other organisms, where LAR-RPTPs and syndecans have been found to function as a ligand-receptor pair [33,34]. In the Drosophila neuromuscular junction, cis-interactions on the neural membrane between DLar and syndecan interfere with trans-interactions between neural DLar and muscle-derived glypican. The binding of DLar to syndecan reduces the adhesion at the NMJ and serves to permit expansion of the structure. However, the C. elegans isoform most similar to DLar, shown to bind syndecan, is PTP-3A, which is genetically and functionally distinct from the isoform we found to cause embryonic defects, PTP-3B. In our assays ptp-3A had no effect on embryonic viability, either alone or in combination with sdn-1, suggesting that the differences between our observations and others are due to the isoform being analyzed.
The PTP-3B isoform that our data implicate as functioning in parallel to SDN-1 appears to be conserved evolutionarily in vertebrates, but is not obviously found in Drosophila. In vertebrates a short isoform of the LAR/PTP-3F receptor has been identified that has the same domain architecture as PTP-3B, being comprised of 5 Fibronectin type III domains in the extracellular portion and two tandem phosphatase domains intracellularly (Uniprot-H0Y4H1). Thus, it is possible that the interaction we have observed between LAR and syndecan is conserved in vertebrates as well.
Our results suggest that in C. elegans embryonic development, the syndecan and LAR proteins have overlapping roles, and can partially compensate for the absence of each other. Further, we found that animals lacking both ptp-3 and sdn-1 exhibited variable points of developmental arrest. This was most apparent in the animals that lacked any maternal ptp-3 contribution, which arrested embryogenesis during epiboly or at hatching, after completing embryogenesis. The variable arrest points of the sdn-1; ptp-3 double mutants suggests that there are likely multiple phases of development in which SDN-1 and PTP-3 function in parallel to provide essential functions. However, an alternative hypothesis is that an early defect at the onset of gastrulation can lead to a variable arrest point later in development. Based on our time-lapse analysis, our data favor the former, but it is possible that subtle defects during early development are occurring. To address this point we will need to identify the mechanism(s) that underlie the lethality of the ptp-3; sdn-1 double mutant animals.
Additional factors likely provide some compensatory function in the absence of both PTP-3 and SDN-1, permitting animals to get through critical developmental periods. For example, some sdn-1; ptp-3 embryos fail to complete epiboly; if this were the earliest function of PTP-3 and SDN-1, and they could not be compensated for during this process, we would expect 100% of the embryos to arrest at that point in development. However, we find that some animals complete epiboly, but arrest at a later time point. The variable arrest points seen in the double mutants suggest that there are potentially multiple proteins that could partially compensate for the loss of ptp-3 and/or sdn-1 during development, including the Eph-ephrin, and slit-robo pathways, which have previously described roles in this process [6,7,52].
The SynLet phenotype observed in ptp-3; sdn-1 double mutants provided us with a powerful platform to identify novel genetic interactions required for embryonic development. We found that we can induce lethality in either the ptp-3 or sdn-1 mutant backgrounds by knockdown of orthogonal genes using RNA interference (RNAi). Like most screens using RNAi we observed some variability in the efficacy, compared to known loss-of-function mutations. While this may complicate the interpretation of epistatic relationships, we found RNAi had a robust effect for the purposes of discovery.
Based on the results of our SynLet screen, we propose that there are at least three, and likely more, different cell-adhesion pathways functioning semi-redundantly during C. elegans development, at least two of which utilize ptp-3 and sdn-1. This is not surprising, as development requires an integrated symphony of cell movements, wherein adhesion must be transiently changed in an orchestrated fashion. However, it is interesting to note that some cell adhesion proteins, e.g. laminin, perlecan, collagen IV, and integrins are categorically essential to embryonic viability, whereas others, like ptp-3 or sdn-1, have a more flexible requirement. As laminin, perlecan, and collagen IV are all components of the basal lamina, and integrins receptors for some of these proteins, it suggests that the basal lamina is a crucial reference point for cell migrations. Cell adhesion, via proteins like PTP-3B or SDN-1, appears to be a more redundant process, with multiple proteins capable of contributing to this role.
Our SynLet screen has identified multiple genes with previously undiscovered roles in embryonic development. The identification of mnp-1 lethality being reduced by mutations in ptp-3 and sdn-1 lethality suggests a complex interplay between cell adhesion and matrix remodeling proteins during embryogenesis. The Eph receptor vab-1 has previously been shown to function in parallel with mnp-1 during C. elegans embryonic muscle cell migration [66]. vab-1 also functions in parallel with ptp-3 to regulate epidermal migration during morphogenesis [5]. The genetic interactions we observe suggest that loss of ptp-3 was protective to animals where mnp-1 had been knocked down. Thus, it will be interesting to determine if this reveals a previously unknown role for ptp-3 in muscle cell migration, or a role for mnp-1 in epidermal cell migration.
One of the interactions we uncovered using our SynLet RNAi approach was a genetic interaction between sdn-1 and lin-44, one of the five Wnt ligands encoded by the C. elegans genome [70]. Prior to this, mom-2 was the only Wnt ligand that had been demonstrated to affect gastrulation. However, it has been shown that other Wnt ligands can influence this process as cwn-1 and cwn-2 mutations enhance the mom-2 lethal phenotype. The fact that the cwn-1;cwn-2; mom-2 triple mutants exhibit a fully penetrant lethality indicates at least some functional redundancy of Wnt ligands in embryogenesis [69]. mom-2 contributes both to endodermal specification and gastrulation through partially overlapping functions [56]. Ultimately, MOM-2, functioning through the frizzled-like receptor MOM-5, results in the phosphorylation of myosin light chain to induce constriction of the apical surfaces of the Ea and Ep cells to induce their internalization. Recent work has also uncovered a novel role for SDN-1 in embryogenesis, where it functions in a MOM-2-dependent pathway to control the orientation of the mitotic spindle earlier in embryogenesis (6 to 8 cell stage of development) [71].
Here we find that lin-44 also affects the internalization of the Ea and Ep cells, albeit to a lesser extent than mom-2. The loss of sdn-1 significantly enhances the penetrance of Ea and Ep ingression defects, and synergistically causes a highly penetrant embryonic lethality. Further, genetic evidence suggests that the LIN-17 frizzled-like receptor does not function in this event, although previous reports have generally found that LIN-44 signals through LIN-17 [72][73][74][75][76].
Previous work suggested that LIN-44 appears to prime cells for other Wnt-signals. For example, in the PLM mechanosensory neurons, LIN-44 activity was required to induce asymmetric localization of LIN-17 which then acted as a receptor for the EGL-20 Wnt ligand [73,76]. One possibility is that LIN-44 also primes the Ea and Ep cells to respond to the MOM-2 ligand, although this remains to be determined.
The use of parallel genetic backgrounds to identify SynLet interactions allowed us to immediately assign a candidate gene into a genetic pathway, based on the outcome of the screen. For instance, lin-44 RNAi knockdown strongly enhanced sdn-1 Lof phenotypes, yet showed no synergistic phenotypes in a ptp-3 mutant background. This suggests that ptp-3 somehow functions in the Wnt pathway during C. elegans embryonic development. The lin-44; ptp-3 double mutants had an increase in larval lethality over the single mutant backgrounds, it did not result in complete synthetic lethality as nearly 70% of the animals survived to adulthood. This indicates that during larval development, ptp-3 and lin-44 may actually function in separate pathways, which contrasts with their linear genetic relationship during embryogenesis. Our future studies will harness sdn-1 synergistic effects to allow us to further explore the mechanisms by which lin-44 and ptp-3 affect gastrulation in C. elegans.
Supporting Information S1 Table. Total Results from RNAi screen. S1 Table provides a complete list of the genes targeted by RNAi in our screen, along with the phenotypes observed in the different mutant backgrounds. Because all clones were screened a minimum of two times the phenotypes observed could be from either assay. (XLSX)