P-type ATPases are a large family of transporters that are conserved from unicellular eukaryotes and plants to metazoans. Structurally and functionally, they fall into five subfamilies, P1 to P5, of which the latter is further divided into P5A and P5B-type ATPases. Unlike for other P-type ATPases, no mutant phenotypes for the P5A-type ATPases have been described in metazoans. Here, we show that the catp-8/P5A-ATPase in the nematode Caenorhabditis elegans is involved in multiple aspects of neuronal patterning, including neuronal migrations as well as axon guidance and dendrite patterning. A functional fluorescent reporter fusion shows the CATP-8/P5A-ATPase is expressed in most, if not all, tissues in the endoplasmic reticulum and that catp-8 can function both in neurons and surrounding tissues from where it orchestrates neuronal development. Genetically, catp-8 acts in multiple pathways during these processes, including the Wnt signaling and the Menorin pathway. Imaging studies suggest that the catp-8/P5A-ATPase is necessary for proper localization of cell-surface transmembrane proteins to dendrites of sensory neurons, but likely not for their trafficking. In summary, we propose that CATP-8/P5A-ATPase serves a function in the ER during development of select neurons, by localizing certain transmembrane, and possibly, secreted proteins.
Development of a nervous system is a crucial step in metazoan development that requires the coordinated interactions of many genetic pathways and tissues. This process can be broken down into at least three components. First, in metazoans, a majority of neurons are actually born distant from their final location in adult animals. Therefore, neurons have to migrate from the place where they are born to their final location in the nervous system, a process that is guided by both intrinsic and extrinsic factors [1,2]. Second, axonal processes, the cellular structures that mediate output of nerve cells, need to grow and navigate, guided by extracellular guidance factors, towards their cellular targets with whom they form specific synaptic connections [3,4]. Third, the often elaborately sculpted dendrites, a neuron’s receiving cellular structures, are patterned by dedicated pathways that originate from different tissues and act both cell-autonomously and non-autonomously [5–7].
The nematode is an excellent system to study basic aspects of conserved processes involved in neural patterning. Several cellular paradigms in worms can be studied at single cell resolution, and often, ample genetic information already exists for these processes. For example, the somatosensory PVD neurons are a powerful system to study conserved mechanisms of dendrite patterning [8,9]. These mechanisms involve the interactions of different tissues that coordinately establish the stereotyped dendritic trees of PVD neurons. A complex of two transmembrane cell adhesion molecules, SAX-7/L1CAM and MNR-1/Menorin, function from the skin in concert with a secreted chemokine LECT-2/Chondromodulin II from muscle to instruct the growing PVD dendrites via a leucine rich transmembrane receptor DMA-1/LRR-TM expressed in dendrites [8,9]. Other well-studied paradigms include neuronal migrations, such as the anteriorly directed embryonic migration of hermaphrodite specific (HSN) neurons from the tail region towards the midbody region  or the pair of laterally located Q neuroblasts, which undergo stereotypical divisions and migrations during larval stages . All cell migrations appear to be controlled, at least in part, by the conserved Wnt signaling system, among others [12,13].
The diverse P-type ATPase family of transporters is an evolutionarily highly conserved family of transporters in eukaryotes that, respectively, move multiple molecular species, including ions and lipids across membranes through the hydrolysis of ATP. These P-type ATPases are multi-transmembrane proteins that can be divided into five subfamilies based on structural and functional characteristics, P1 through P5 [14,15], of which the latter can be further divided into the P5A and P5B family . The P1 through P3 subfamilies serve as ion transporters, whereas the P4 subfamily has been shown to function as flippases that move lipids across membranes . The P5B family of transporters was recently reported to transport polyamines across membranes . Mutations in the P5A family of transporters in both plants (A. thaliana) and unicellular eukaryotes (S. cerevisiae ) were shown to display pleiotropic phenotypes, including defects in phosphate homeostasis and male gametogenesis [18–20], phospholipid and sterol homeostasis , and targeting of mitochondrial outer membrane proteins . However, it remained unclear what functions P5A-ATPases serve in metazoans.
We show here that the sole C. elegans ortholog catp-8/P5A-ATPase is required for specific aspects of nervous system development, including dendrite patterning, axonal guidance decisions and neuronal migrations. A CATP-8/P5A-ATPase reporter shows expression in the endoplasmic reticulum in most if not all tissues albeit at different levels. Moreover, the catp-8/P5A-ATPase can function both cell-autonomously and non-autonomously in multiple genetic pathways in to regulate neuronal development. Additionally, we find that in catp-8/ATPase mutants, the amounts of functional reporters for select transmembrane proteins (the leucine rich transmembrane receptor DMA-1/LRR-TM and the HPO-30/Claudin-like molecule) are strongly reduced on higher order dendritic branches of PVD dendrites. Taken together, these results suggest that the CATP-8/P5A-ATPase functions in multiple pathways during neural patterning and is important for correct localization of transmembrane cell surface proteins and possibly secreted proteins. While this study was in progress, two studies reporting a function for catp-8/P5A-ATPase in PVD patterning in C. elegans were published [23,24]. Feng et al. suggested a role for catp-8/P5A-ATPase in dendrite patterning and targeting of DMA-1/LRR-TM in dendrites  and Qin et al. suggested that catp-8 was important for ER homeostasis by preventing mistargeting of mitochondrial proteins independently of ER-associated degradation (ERAD) as well as dendrite patterning . Similarly, a crystal structure of the yeast homolog Spf1 in conjunction with biochemical studies suggested a role for P5A-ATPases in ER and mitochondrial quality control .
In a genetic screen for animals with defects in dendrite arborization of PVD somatosensory neurons, we isolated the allele dz224 with characteristic defects in dendrite patterning of PVD. Using a combination of whole genome sequencing and mapping (S1 Fig, Materials and Methods)), we found that the molecular lesion in dz224 results in a premature stop codon after 55 amino acids in catp-8, which encodes the sole P5A-type ATPase in the C. elegans genome (Fig 1A–1C). Based on the molecular nature of dz224, we predict this allele to result in strong, if not complete loss of function of the catp-8/P5A-ATPase.
To more precisely determine the function of catp-8 in dendrite patterning, we conducted morphometric analyses of PVD somatosensory dendrites. During larval stages two primary dendrites emanate from the PVD cell body on either side of the animal in an anterior and posterior direction, respectively, and through sequential perpendicular branching of secondary, tertiary, and quaternary branches establish highly stereotyped dendritic trees with semblance to Menorah-like candelabras (Fig 1D)[27–29]. We traced dendrites in a region 100μm anterior to the cell body of young adult animals and determined the number and length of secondary, tertiary and quaternary dendritic branches. Compared to wild type animals, we found the total number of secondary branches unchanged, whereas the number of tertiary and quaternary branches were significantly reduced in catp-8(dz224) mutants (Fig 1E and 1G–1I). However, the number of secondary branches that reached the line of the tertiary branches was significantly reduced, resulting in a reduced aggregate length of secondary branches (S1B and S1C Fig). Overall, many dendritic trees displayed a conspicuous defect, where secondary branches appeared to lack proper tertiary branches, and merely bifurcated into two quaternary dendrites rather than forming perpendicular tertiary branches. This observation was also borne out by a reduced number of quaternary branches per secondary branches, approaching an average of two quaternary branches per secondary branch (i.e. per menorah) as opposed to approximately four in control animals, and a decrease in aggregate length of secondary, tertiary and quaternary branches (Figs 1J and S1B–S1E). In addition, we observed increased numbers of ectopic secondary and ectopic tertiary branches, possibly reflective of a lack of a functioning feed-back loop that has been postulated to restrict secondary branching once stable tertiary branches have been formed .
To corroborate these findings, we obtained three additional alleles of catp-8/P5A ATPase, including gk860114 from the million mutation project , which results in a premature stop codon after 593 amino acids (Fig 1A). We used CRISPR/Cas9 based genome editing to create two additional alleles (Fig 1A and 1C): dz290 , which deletes two codons within a perfectly conserved stretch of the actuator domain , and a missense mutation (E289K), which changes a perfectly conserved glutamic acid to lysine (E289K) in the actuator domain (Fig 1A and 1C), and is analogous to a mutation described in a pair of sisters with intellectual disability . Finally, we introduced the E289K mutation on its own to create dz291. Of these, all but the dz291 allele displayed strong phenotypes in PVD patterning that were indistinguishable from the dz224 allele (Figs 1G–1J and S1B–S1E). Consistent with the nonsense mutations in dz224 and gk860114, both alleles behaved recessively and failed to complement each other, suggesting loss of catp-8 function (S1F Fig). To test, whether the gk860114 allele behaved genetically as a null allele, we analyzed animals where catp-8(gk860114) was placed in trans to the mDf7 deficiency, which removes the genomic region encoding catp-8 and fails to complement catp-8 mutant alleles (S1A Fig). We found that these animals did not display a more severe phenotype in PVD patterning than catp-8(gk860114) homozygous animals, demonstrating that gk860114 behaves as a genetic null allele (S1G Fig). Animals carrying the catp-8(gk860114) null allele appear superficially normal, have a normal body length at the L1 larval stage (S1H Fig), but seem to grow slower. Taken together these findings suggest that catp-8 function is required for PVD patterning, especially for the extension of tertiary and quaternary dendritic branches. Moreover, the E289K mutation, which is analogous to a mutation shared by the pair of sisters with intellectual disability , does not compromise catp-8 sufficiently to result in morphological phenotypes, at least in the context of PVD patterning or neuronal migration (see below). Lastly, the defects due to mutations in the actuator domain are consistent with the interpretation that enzymatic activity of CATP-8/P5A type ATPase is required for PVD patterning.
We next asked whether catp-8 is only necessary for development of the highly branched dendritic trees of PVD dendrites or more generally, for other aspects of neural patterning. To this end, we crossed catp-8 mutants with a panel of GFP reporters that label individual neurons or classes of neurons, including motor neurons, interneurons, sensory neurons and touch receptor neurons. We found that mutant alleles of catp-8 displayed defects in neurite patterning of some but not all neurons. For example, in catp-8 mutant animals the D-type motor neurons displayed neurite extension defects in the ventral and dorsal nerve cords, whereas the DA/DB-motor neurons appeared indistinguishable from control animals (Table 1, S2A–S2D Fig). This is significant insofar as the processes of both types of motor neurons make similar navigational choices and travel, at least in the nerve cords, in the same fascicles. However, neither patterning of the pair of AIY interneurons, nor the amphid sensory neurons or phasmid sensory neurons was obviously dependent on catp-8 function (Table 1, S2E–S2J Fig). Lastly, we investigated axon guidance choices of HSN and PVQ neurons at the ventral midline, where both neurons send axons in the left and right ventral nerve cords from the midbody or tail to the head region, respectively. We observed significant defects in axon midline guidance of HSN neurons with cross overs, but no defects for PVQ neurons (Table 1, S2K–S2N Fig). Taken together, our findings show that axonal patterning of some but not all neurons is dependent on catp-8/P5A-ATPase function. Importantly, even neurons that make similar guidance choices and travel in the same nerve fascicles such as the D-type and DA/DB motor neurons or the HSN and PVQ neurons, can develop either dependent on catp-8 function or, independently.
|Neurons Examined (Marker Used)||control [%]||N||catp-8 [%]||N||P Value 1|
|Motor Neurons 2|
|Gaps in ventral nerve cord||10.8||93||41.3||109||****|
|Gaps in dorsal nerve cord||12.9||93||26.6||109||*|
|0 L/R choice commissure||60.0||40||23.6||55||***|
|1 L/R choice commissure||25.0||40||41.8||55||**|
|≥ 2 L/R choice commissure||15.0||40||32.7||55|
|defasciculation of DNC||0.0||31||11.4||44||0.07|
|defasciculation of VNC||19.4||31||36.4||44||0.13|
|defasciculation in both||3.2||31||2.3||44||ns|
|Gaps in ventral nerve cord||0||102||0||108||ns|
|Gaps in dorsal nerve cord||0||102||0||108||ns|
|L/R choice of commissures||4.8||102||5.7||108||ns|
|Ventral nerve cord crossover2||0.0||85||12.5||72||***|
|Ventral cord crossover||11.8||101||7||100||ns|
|Ectopic axonal branching||2||120||2||124||ns|
|Amphid Neurons 4|
|Anterior nerve ring||0||50||0||50||ns|
|Posterior nerve ring||0||50||0||50||ns|
|Phasmid Neurons 3|
We next investigated a set of stereotyped neuronal and non-neuronal cellular migrations, including both embryonic and postembryonic migrations, as well as anteriorly-posteriorly and posteriorly-anteriorly directed migrations (Fig 2A). We first investigated two embryonic migrations: the pair of ALM mechanosensory neurons, which migrate from the head region towards the midbody region (i.e. anteriorly-posteriorly directed), and the hermaphrodite specific neurons (HSNs), a set of motor neurons that migrate from the tail towards the mid body region (i.e. posteriorly-anteriorly directed). Loss of catp-8 function resulted in severe migration defects in anteriorly directed HSN migrations, which often failed to reach their final destination around the midbody and instead stopped migration prematurely (Fig 2B and 2E). In contrast, loss of catp-8 had no effect on posteriorly directed ALM neuronal migrations (Fig 2C and 2F). We next investigated a set of postembryonic migrations, specifically of the Q cell descendants . The bilateral pair of Q cells are neuroblast cells, which are born during the mid L1 larval stage. On the right side of the animal, the QR cell migrates anteriorly while undergoing a series of stereotyped cell divisions as follows. The first division results in two more blast cells, of which one divides again and gives rise to the AQR cell and the other undergoes programmed cell death . The other blast cell also gives rise to one cell that undergoes programmed cell death and another that after a further division forms the AVM and SDQR neurons (Fig 2A). The AQR and AVM/SDQR neurons therefore originate from two subbranches of QR cell descendants, and are positioned in the head or a lateral anterior section of the worms, respectively. The Q cell descendants on the left side of the animal undergo a posteriorly directed migration while also giving rise, in a similar pattern of divisions, to PQR and PVM/SDQL neurons (Fig 2A). PQR travels to the tail whereas PVM and SDQL remain in a lateral posterior section of the animals. We found that the anteriorly directed migrations of both AQR and AVM, which represent both sub branches of QR descendants, are not obviously affected in catp-8 mutants, whereas the posteriorly directed migrations of PQR and PVM neurons display defects (Fig 2C, 2D, 2G, and 2H). Specifically, PQR and PVM erroneously migrate anteriorly towards the head instead of posteriorly. This observation suggests that the choice of directionality in migration of QL descendants on the left side of the animals is dependent on catp-8/P5A ATPase function, rather than migratory capacity itself. Moreover, we found that gk860114 homozygous mutants originating from homozygous mothers displayed Q cell migration defects, whereas gk860114 homozygous mutants originating from gk860114/+ heterozygous animals did not display this phenotype, suggesting a maternal contribution of catp-8 during this postembryonic migration (S1I Fig). In contrast, we observed a similar range of defects in HSN migration and PVD dendrite patterning in gk860114 homozygous mutants coming from heterozygous mothers compared to those originating from homozygous mothers (S1F and S1J Fig, cf also Fig 2E), leading us to conclude that Q cell migration requires less catp-8 function compared to HSN and PVD patterning. Lastly, we investigated two pairs of non-neuronal cells, the coelomocytes, which migrate embryonically in a posterior direction . We found that coelomocyte migration proceeds normally in catp-8 mutants (Fig 2I). Collectively, our experiments show that both embryonic and postembryonic cell migrations require catp-8/P5A-ATPase function, including anteriorly and posteriorly directed migrations.
To gain insight into the cellular focus of action of the catp-8/P5A-ATPase , we used CRISPR/Cas9 engineering to insert a tagRFP fluorescent protein at the N-terminus of CATP-8 (Fig 1A). We found these animals to show widespread, if not ubiquitous expression of CATP-8, including in the epidermis, muscle, pharynx, intestine, and the major neuronal ganglia (Fig 3A–3D). The expression appears intracellular in a perinuclear pattern and different tissues appear to express different levels of CATP-8. Of note, we detected expression in both neurons that were phenotypically affected in catp-8 mutants as well as neurons that were not visibly affected. For example, clear expression of the endogenous tagRFP::CATP-8 reporter was seen in HSN, PVM, and PVD neurons, but also in ALM neurons and others (Fig 3E–3H). Importantly, PVD morphology and PQR cell position appeared completely normal in animals carrying the endogenous fluorescent tag, suggesting that the N-terminal tagRFP::CATP-8 fusion is fully functional. We found similar expression patterns in transgenic animals expressing GFP under control of the catp-8 5’ region immediately upstream of the ATG start codon, where we also saw widespread expression from embryonic stages onwards, including in muscle, intestine, pharynx and other tissues (S3 Fig). Collectively, our studies show that catp-8 is widely, if not ubiquitously, expressed albeit at different levels.
Previous work showed that the yeast Spf1 (and Arabidopsis MALE GAMETOGENESIS IMPAIRED ANTHERS (MIA) or PDR2) homologs are localized to the endoplasmic reticulum (ER)[19,20,35]. We therefore aimed to determine the subcellular localization of CATP-8 in C. elegans. To this end, we used the functional N-terminal fusion of tagRFP to CATP-8 and drove expression from a single copy insertion transgene under control of the heterologous dpy-7p promoter, which is specific for the epidermis . We chose to use this single copy, functional transgene over the endogenously tagged catp-8 locus to minimize interference with signal from other tissues (cf. Fig 3). This epidermally-expressed transgene (dzSi3 ), which also displayed perinuclear staining (and rescued PVD patterning defects) was crossed with a panel of organellar reporters—for (1) the ER, (2) the cis/medial Golgi, (3) early endosomes, (4) late endosomes and lysosomes, (5) autophagosomes, and (6) lysosomes. We found the strongest colocalization with a marker for the ER and no correlation with markers for Golgi, late endosomes or lysosomes (Fig 4). These findings suggest that in metazoans CATP-8/P5A-ATPase is localized to the ER similarly to plants and unicellular eukaryotes. We next tested whether the gross morphology of these cellular compartments was defective in catp-8 mutant animals. We found no apparent defects in the perinuclear localization, patterns or intensity of any of the organelle reporters examined (S4 Fig). Taken together, our findings suggest that the CATP-8/P5A-ATPase serves a function in the ER, which may be conserved from unicellular eukaryotes and plants to animals.
To determine where catp-8 may function during different developmental processes, we performed transgenic rescue experiments, where we expressed a catp-8 cDNA under control of heterologous promoters in catp-8 mutant animals. We found that patterning defects of PVD dendrites were rescued by pan-neuronal transgenic expression of catp-8 , but not by expression in muscle (Figs 5A–5G and S5A–S5C). Interestingly, a single copy transgene expressing the functional N-terminal tagRFP::CATP-8 fusion under control of the heterologous epidermis-specific dpy-7p promoter also rescued PVD patterning defects in catp-8 mutants (Figs 5A–5G and S5A–S5C). These observations suggest that catp-8 can serve redundant functions in different tissues, i.e. both cell-autonomously and non-autonomously, but may be sufficient in either neurons or the epidermis. Surprisingly, embryonic and postembryonic cell migration defects of HSN, PQR and, PVM neurons were completely rescued by expression of catp-8 in muscle, but not in the epidermis, the touch receptor neurons (i.e. PVM), or the precursors that give rise to the Q cells (Figs 5H–5J and S5D and S5E). One out of three transgenic lines expressing the catp-8 cDNA under control of a pan-neuronal reporter showed partial rescue of the neuronal migration defects of PVM and PQR neurons (Figs 5I,5J and S5D and S5E), suggesting that overexpression in neurons can partially compensate for loss of catp-8 in other tissues, e.g. muscle. We conclude that at least for these cellular migrations, catp-8 appears to function non-autonomously in muscle. Therefore catp-8 can serve both autonomous and non-autonomous functions during neural patterning, including in dendrite morphogenesis and different neuronal migrations.
The conserved Menorin pathway shapes somatosensory patterning of PVD dendrites [8,9]. Briefly, the two putative cell adhesion molecules SAX-7/L1CAM and MNR-1/Menorin form a complex in the epidermis, which, together with the muscle derived chemokine LECT-2/Chondromodulin II forms a high affinity substrate for the leucine rich repeat transmembrane receptor DMA-1/LRR-TM, which functions in PVD neurons [8,9]. Downstream of DMA-1/LRR-TM in PVD, a conserved set of intracellular molecules such as the TIAM-1/GEF guanine nucleotide exchange factor and HPO-30/Claudin act together to regulate the actin cytoskeleton [8,9,37]. Previous genetic experiments established that SAX-7/L1CAM, LECT-2/Chondromodulin II and MNR-1/Menorin function in a linear genetic pathway with the DMA-1/LRR-TM receptor, but also suggested that DMA-1 can serve functions in PVD patterning that are independent of SAX-7/L1CAM, MNR-1/Menorin and LECT-2/Chondromodulin II [30,38]. To determine the genetic relationship between catp-8 and the Menorin pathway, we analyzed double null mutants of catp-8 and various components of the Menorin pathway using morphometric analyses. We traced and quantified the number and length of secondary, tertiary and quaternary dendrites 100μm anterior to the PVD cell bodies as previously described . With regard to the number of secondary dendrites, we found that the catp-8 mutant was epistatic in catp-8; mnr-1 and lect-2; catp-8 double mutants, i.e. the double mutant resembled the catp-8 single mutant rather than displaying the increased number of secondary branches observed in the mnr-1/Menorin and lect-2/Chondromodulin II single mutants (Fig 6A). In double mutants with dma-1, hpo-30 or tiam-1 the phenotypes were not further enhanced by concomitant genetic removal of catp-8. With regard to the number of tertiary branches, we made similar observations, although in some cases the double mutants between catp-8 and for instance mnr-1/Menorin, or hpo-30/Claudin appeared phenotypically stronger than the single mutants (Fig 6B). For the number of quaternary branches, double mutants between catp-8 and genes in the Menorin pathway invariably displayed the stronger phenotype observed in mutants of the Menorin pathway (Fig 6C). We observed similar genetic interactions when investigating the aggregate lengths of secondary, tertiary and quaternary branches (S6 Fig). We conclude that catp-8 functions in the Menorin pathway and likely within both branches, i.e. the mnr-1/Menorin-dependent as well as -independent branches, both of which feed into the DMA-1/LRR-TM receptor in PVD neurons.
The functions of catp-8 in PVD neurons and the localization to the ER prompted us to test whether the localization or trafficking of two transmembrane molecules, which are expected to transit the ER, is affected in catp-8 mutants. Specifically, we analyzed the localization of reporters for the single transmembrane receptor DMA-1/LRR-TM (DMA-1::GFP)  and the four transmembrane, claudin-like molecule HPO-30/Claudin (HPO-30::tagBFP) . We found that overall DMA-1::GFP expression was noticeably dimmer in catp-8 mutants (Fig 6D and 6E). When we quantified the fluorescence in the cell body and the primary dendrite we found it strongly reduced in catp-8 versus control animals (Fig 6F and 6G). Interestingly, the ratio of fluorescence in the cell body versus the primary dendrite was significantly increased (Fig 6H). These observations could be a reflection of reduced trafficking towards the periphery or reduced amounts of DMA-1::GFP/vesicle. To distinguish between these possibilities, we quantified the number of puncta, previously suggested to represent the vesicular fraction of DMA-1::GFP [41,42], in primary and tertiary dendrites of catp-8 and control animals. We found the number of puncta of DMA-1::GFP in neither primary nor tertiary dendritic branches significantly different in catp-8 mutant versus control animals (S7B and S7C Fig). We therefore propose that catp-8 does not serve a function in vesicle trafficking. Instead, catp-8 may function to facilitate correct membrane localization of DMA-1/LRR-TM. Alternatively, but not mutually exclusive, catp-8 could serve a role in loading DMA-1::GFP into vesicles.
As previously reported, HPO-30::tagBFP localizes primarily to primary and quaternary dendritic branches . We found that HPO-30::tagBFP fluorescence in distal dendrites appeared reduced in catp-8 mutant animals (Fig 6I and 6J). Although HPO-30::tagBFP expression in the cell body did not seem visibly different, HPO-30::tagRFP fluorescence in quaternary dendrite branches was strongly reduced in catp-8 mutant versus control animals (Fig 6K). Yet, the number of HPO-30::tagRFP puncta in the PVD primary branch remained unaffected, implying again that trafficking of HPO-30::tagBFP does not require catp-8 function (S7D Fig).
Because epidermal expression of catp-8/P5A-ATPase could rescue PVD defects in catp-8 mutant animals, we tested, whether another conserved transmembrane cell adhesion molecule SAX-7/L1CAM, which functions in the epidermis to shape PVD dendrites, is affected in catp-8 mutant animals. We found that SAX-7::GFP localization appears qualitatively and quantitatively unchanged in catp-8 mutants compared to control animals, or in animals that overexpressed tagRFP::CATP-8 transgenically in the epidermis (S7E and S7F Fig). Similarly, localization of the secreted LECT-2/Chondromodulin II was not affected in catp-8 mutants (S7G Fig). We conclude, catp-8/P5A-ATPase function is required for the regulation and localization of some but not all transmembrane proteins.
Our studies show that catp-8 is required for both embryonic and postembryonic neuronal migrations, including the anteriorly directed embryonic migrations of HSN neurons and the posteriorly directed postembryonic migrations of the Q cell descendants PVM and PQR. Both of these cell migrations are genetically well-described and depend on, among other pathways, Wnt signaling [10,11,13]. Briefly, the egl-20/Wnt ligand is expressed in posterior tissues, including body wall muscle [43,44], is believed to form a gradient  and to function through the G-protein coupled receptor mig-1/Frizzled in HSN neurons to mediate HSN migration . While other Wnts and frizzled receptors play smaller, largely redundant roles in HSN migration , most functions appear to be mediated by the MIG-1/Frizzled receptor in HSN neurons, because a mig-1; egl-20 double mutant does not display more severe defects in HSN migration than the single mutants . To investigate the genetic relationship between catp-8/P5A-ATPase and Wnt signaling, we therefore created a double null mutant between catp-8 and mig-1. We found that the HSN migration phenotype in mig-1; catp-8 double mutant animals was not enhanced compared to the more severe of the single mutants (mig-1/Fz ) (Fig 7A). This suggests that catp-8 functions genetically in a pathway with mig-1/Frizzled and, by inference, Wnt signaling.
The migration of Q cell descendants occurs in two phases . In an initial phase, the QL and QR cells, both of which are located laterally on either side of the animals, are polarized and migrate about the distance of one cell diameter over the posteriorly and anteriorly located epidermal blast cells, respectively (Figs 7B and S7A). This is followed by a second phase, in which the QL and QR cells migrate further posteriorly and anteriorly, respectively, while undergoing the stereotyped cell divisions (Fig 2A). The first phase is controlled by at least three partially redundant genetic pathways, which are defined by genes encoding: (1) the transmembrane cell adhesion molecule UNC-40/DCC (Deleted in Colorectal Cancer), (2) the transmembrane protein MIG-21 and the conserved, putative C-mannosyl transferase DPY-19/DPY19L [48–50], and (3) the cadherin proteins CDH-3, CDH-4, and the LAR-like receptor tyrosine phosphatase PTP-3/LAR [51–53]. Mutations in any of these genes result in defects in the initial polarization and, consequently, migration. In contrast, we find that initial polarization and migration of QL and QR cells proceeds normally in catp-8 null mutant animals, suggesting that catp-8 functions primarily during the second phase of migration (Figs 7B and S7A). Consistent with this interpretation, we found that unc-40; catp-8 double mutants exhibited a more severe phenotype than either single mutant alone (Fig 7C–7E). Similar results were found in mig-21; catp-8 and ptp-3; catp-8 double mutant animals, which also showed a more severe phenotype compared to the single mutants (Fig 7C–7E). We conclude that catp-8 serves no individual role during the first phase of Q cell migration, but rather functions during the second phase of Q cell migration.
We next focused our analysis on the second phase of migration and in particular on AVM and PVM migrations. The mab-5/HOX-C antennapedia-like homeobox transcription factor is necessary and sufficient in QL cells after the initial polarization to mediate the second phase of their posterior migration  and is activated in response to the egl-20/Wnt morphogen that is expressed in posterior tissues . Similar to loss of catp-8, loss of mab-5 results in QL cell migration defects, where QL cell descendants migrate in an anterior rather than posterior direction . Because the mab-5 mutant phenotype is essentially fully penetrant, we could not test for possible enhancement in a mab-5; catp-8 double mutant. However, in a mab-5(gf) gain of function allele mab-5 is activated in both QL and QR cells. Migration of the QL descendants is not affected in the mab-5(gf) gain of function mutants, whereas QR descendants fail to migrate in an anterior direction . We therefore tested the epistatic relationship between the mab-5(gf) allele and the catp-8 loss of function mutant. We found that the aberrant anterior migration of the QL descendant PVM in catp-8 single mutants was completely suppressed in a mab-5(gf); catp-8 double mutant (Fig 7D and 7F). In contrast, the migration defects of QR descendants in a mab-5(gf) remained unaffected by concomitant loss of catp-8 (Fig 7D and 7G). Thus, catp-8/P5A-ATPase functions upstream of or in parallel to the mab-5/HOX-C homeobox transcription factor.
In this study we describe multiple functions of the catp-8/P5A-ATPase in metazoan development. Specifically, we show that catp-8 is required for different aspects of nervous system development, including axonal guidance, dendrite patterning and neuronal migrations. A reporter for CATP-8/P5A-ATPase is expressed in the ER of most if not all tissues and can function both cell-autonomously and non-autonomously during neuronal development. Genetically, catp-8/P5A functions in multiple pathways, including the conserved Menorin pathway during dendrite development and Wnt signaling during neuronal migrations. We find a substantial degree of genetic redundancy with different developmental pathways, suggesting that catp-8/P5A-ATPase may serve diverse, yet specific functions in different cellular contexts Lastly, catp-8/P5A-ATPase is necessary to correctly localize reporters for some but not all transmembrane proteins.
Not all neurons are affected by loss of catp-8 function. For example, the PVD somatosensory neurons require catp-8 for the elaboration of dendritic trees, but the dendrites of amphid sensory neurons seem unaffected. Similarly, axonal guidance or extension of HSN and D-type motor neurons requires catp-8 function, whereas guidance of PVQ axons or DA/DB motor neurons appears independent of catp-8 , respectively. It is interesting to note that PVQ and HSN (but also D-type and DA/DB motor neurons to some extent) make similar navigational choices, are part of the same fascicles and depend on similar axon guidance pathways, such as Netrin signaling or Slit/Robo signaling [46,55,56], although we cannot exclude that the extracellular or signaling environment is distinct at different times during development. It should be noted that Wnt signaling is also required for correct axon guidance of both HSN and PVQ axons at the midline [46,57] and, at least for the latter cellular context, has been shown to function in parallel with Netrin signaling . A possible explanation is that different genetic redundancies for catp-8 exist in different cells; in other words, in some cells, catp-8 may serve unique roles for which no functional alternative to catp-8 exists whereas in other cells catp-8 has functions that can be partially served by other genes. Such genes could be the related catp-6 or catp- 7 ATPases, although at least in PVD patterning they play no redundant roles . Alternatively, the molecules that are required for development of a given cell may be distinct in different neurons, such that some neurons depend on more or less proteins requiring catp-8 function.
Transgenic rescue experiments suggest both cell-autonomous and non-autonomous functions for catp-8. For example, PVD can develop normally with catp-8/P5A-ATPase function present in either neurons or the epidermis. This finding is in contrast to findings by Feng et al. who could rescue catp-8 defects by transgenic expression of catp-8 in neurons, but not in muscle or the epidermis . One possible explanation for this discrepancy is different levels in transgene expression (we use a single copy transgenes for epidermal expression whereas Feng et al. utilized multi copy transgenes ). This hypothesis is supported by the fact that the Menorin pathway is known to be dosage sensitive and can be both positively and negatively regulated [58,59].
In contrast to this tissue redundancy, the migration defects of Q cell descendants are rescued only by non-autonomous expression of catp-8 in muscle, but not by expression in neurons or other tissues. These results strongly argue for a non-autonomous function of catp-8/P5A-ATPase in muscle during migration. However, no transmembrane proteins are known to function from muscle in Q cell migration, although muscle is clearly important, as both the RNA-Binding Protein ETR-1/CELF1 and the extracellular matrix protein SPON-1/F-spondin can function in muscle cells to mediate Q cell migration [60,61]. It is therefore tempting to speculate that secretion of the EGL-20/Wnt ligand could be regulated by catp-8/P5A-ATPase (Fig 8). First, we found, that during HSN migration (and possibly also Q cell migration), Wnt signaling functions in concert with the catp-8/P5A-ATPase . Second, EGL-20/Wnt has been shown to be expressed from posterior body wall muscle among other posterior tissues [44,45]. Third, the cellular migrations that are affected in catp-8/P5A-ATPase mutant animals (PVM, PQR, HSN) are all dependent on egl-20/Wnt whereas cell migrations that are independent of catp-8/P5A-ATPase (ALM, coelomocytes) are also independent of egl-20/Wnt . Lastly, a mab-5(gf) mutant rescues the catp-8/P5A-ATPase mutant and is therefore likely downstream of catp-8 –just as mab-5 is downstream of egl-20/Wnt . Further experiments are needed to definitively establish whether and how EGL-20/Wnt secretion or localization is dependent on catp-8/P5A-ATPase function.
Previous studies in yeast and plants suggested various roles for P5A-ATPases, such as functions in phosphate homeostasis and male gametogenesis [18–20], phospholipid and sterol homeostasis  and targeting of mitochondrial outer membrane proteins . Therefore, which functions P5A ATPases may serve in metazoans remains a major question. Here, we provide genetic evidence that catp-8/P5A-ATPase is required for the localization of sufficient amounts of the DMA-1/LRR-TM receptor in the membrane of higher order PVD dendrites during morphogenesis. Similarly, the four transmembrane protein HPO-30/Claudin, which also functions in PVD neurons during dendrite patterning  requires catp-8 to be correctly localized to distal, higher order dendrites. Importantly, the number of vesicles containing the respective reporters and, which are presumed to represent endocytic/transport vesicles [41,42] along the dendritic branches, remain unaffected. Intriguingly, localization of the conserved SAX-7/L1CAM cell adhesion molecule in either neurons or the epidermis is not obviously affected. Therefore, catp-8 may be involved in the insertion/localization of certain transmembrane proteins but not others to the plasma membrane (Fig 8). These findings are consistent with another recent study showing similar effects on DMA-1::GFP [23,24], although it was also reported by Qin et al. that the levels of a HPO-30 reporter remained unchanged . A possible explanation for this discrepancy is that we measured fluorescence in distal dendrites rather than the cell body . The findings of a role for the catp-8/P5A ATPase in the localization of some cell surface proteins is seemingly at odds with two other studies in yeast and C. elegans suggesting that P5A-ATPases primarily maintain ER homeostasis by functioning as dislocases to remove mislocalized mitochondrial proteins [24,25]. However, several arguments can be made that P5A-ATPases serve additional functions beyond removing mislocalized mitochondrial proteins. First, genetic manipulations in C. elegans that suppress ER defects in PVD due to mislocalized DRP-1/Drp1 mitochondrial fission protein to the ER in catp-8 mutants, fail to restore the correct dendritic morphology  suggesting that catp-8/P5A-ATPase serves additional functions beyond maintaining ER integrity to mediate PVD morphogenesis. Second, proteomic studies in HeLa cells found not only the levels of mitochondrial, but also other transmembrane and secreted proteins changed in ATP13A1/P5A-ATPAse mutant cells. For example, a number of secreted growth factors, including the platelet derived growth factor D (PDGFD) or transforming growth factor beta (TGFB) as well as axon guidance molecules such as ephrin A5 (EFNA5), to name but a few, were significantly reduced in abundance upon loss of ATP13A1 catalytic activity . In the context of these studies, our findings argue for a more diverse, yet likely specific role for catp-8 in the localization of certain transmembrane proteins and regulation of secreted proteins (Fig 8). Further experiments will be required to delineate the full spectrum of genetic and biochemical functions of P5A-type ATPases.
All C. elegans strains were grown on nematode growth medium plates with E. coli (OP50) as a food source as described , usually at 20°C unless otherwise specified. Strains used in this work can be found in S1 Table. The catp-8 alleles dz224 and gk860114 were backcrossed four and three times, respectively, prior to further analysis.
Using a combination of Hawaiian SNP mapping and sequencing  the allele dz224 was localized to a region between 8Mb and 13Mb on LG IV, which was further refined by deficiency mapping to between 10.2Mb and 12Mb (S1 Fig). This 1.8Mb interval comprised 7 non-synonymous missense mutations and one non-sense mutation in catp-8, which was confirmed by Sanger sequencing: an A to a T transversion at position 11,479,010 on LG IV (WB270), resulting in an ochre premature stop codon after Y55 in C10C6.6 (catp-8). We found that catp-8(gk860411) failed to complement dz224, further suggesting that both alleles are affecting the same gene.
Cloning of all constructs were carried out using standard molecular biology methods. A list of primers used can be found in S2 Table. For a list of plasmids used and detailed descriptions of constructs see S3 and S4 Tables. All extrachromosomal arrays were generated by injecting a mixture of the desired plasmids with pBluescript to a final concentration of 100 ng/μl of DNA. A detailed list of all the extrachromosomal transgenic lines generated can be found in S5 Table.
For CRISPR/Cas9 mediated genome editing, appropriate guide RNA sites were selected using IDT Cas9 crRNA design tool and, oligos in the general form of TAATACGACTCACTATA(gRNA)GTTTTAGAGCTAGAAATAGCAAG were ordered, where (gRNA) is the 20nt of the guide RNA sequence before the PAM motif, optimized for T7 promoter transcription. These oligos were used in PCR reactions as a forward primer in conjunction with the universal reverse primer AAAAGCACCGACTCGGTG (oLT337) to generate sgRNA transcription templates from pDD162 (containing the sequence of tracrRNA). sgRNA was then transcribed from this template using the HiScribe T7 in-vitro transcription kit (NEB) and purified with the Monarch RNA cleanup kit (NEB). The resulting sgRNA was used in an injection mix at 20 ng/μl per sgRNA, together with 250 ng/μl of Alt-R Cas9 endonuclease (IDT) and repair templates (100 ng/μl for single-strand oligos, 200 ng/μl for PCR products or plasmids). Injection and CRISPR efficiency were monitored through a Co-CRISPR strategy with dpy-10(cn64) or unc-58(e665) conversions, or through rol-6(su1006) co-injection markers . A list of all genome-edited strains and detailed methodology can be found in S6 Table.
All worms were immobilized using 1 mM levamisole and mounted on 4% agarose pads for phenotypic analysis.
AQR/PQR migrations were scored in wdIs52 (Is[F49H12.4p::GFP]) in L2 to young adults with a Plan-Apochromat 10x/0.25 objective on a Zeiss Axioimager Z1. Cells were considered in the head if anterior to the intestine, in the tail if posterior to the intestine, and in the mid body otherwise. AVM/PVM migrations were scored in zdIs5 (Is[mec-4p::GFP]) in L4 to young adults with a Plan-Apochromat 10x/0.25 objective on a Zeiss Axioimager Z1. Cells were scored based on their relative position (anterior or posterior) to the vulva. HSNs migrations were scored in dzIs75 (Is[kal-1p9::GFP]) in adult worms with a Plan-Apochromat 16x/0.5 objective on a Zeiss Axioimager Z1. Cells were considered to be at the vulva if the cell bodies were positioned less than or equal to an egg’s length posterior to the vulva. ALM was scored in zdIs5 (Is[mec-4p::GFP]) in synchronized L1 animals. Animals are imaged using a Plan-Apochromat 10x/0.25 objective on a Zeiss Axioimager Z1 Apotome. Using ImageJ, the distance of the more anterior cell body from the nose was measured and divided by the total length of the animal. Coelomocytes were scored in dzEx2140 at the L4 to young adult stage with a Plan-Apochromat 16x/0.5 objective on a Zeiss Axioimager Z1. Coelomocytes were considered to be in the head region if in the anterior quarter of the worm.
HSN axon guidance at the ventral midline was scored in dzIs75 or mgIs71 at the L4 to adult worms stage with a Plan-Apochromat 40x/1.3 objective on a Zeiss Axioimager Z1. Crossover defects of axons were scored if any portion of the axons of HSNL and HSNR were touching. DD/VD neurons were scored in juIs76 at the L4 to adult worm stage with a Plan-Apochromat 40x/1.3 objective on a Zeiss Axioimager Z1. Axon guidance was categorized into two measurements, L/R choice and gaps in the dorsal or ventral cord. For L/R choice, any worm with 1 or more axons found on the opposite side of the stereotyped location was counted as having an L/R choice defect as indicated. Additionally, any worm with a visible discontinuation of the dorsal and/or ventral cord between any two neurons (gaps) was considered to have a dorsal and/or ventral cord defect respectively. DA/DB motor neurons were scored in evIs82b at the L4 to adult stage with a Plan-Apochromat 40x/1.3 objective on a Zeiss Axioimager. Axon guidance was categorized into two measurements, L/R choice and gaps in the dorsal or ventral cord. For L/R choice, any worm with 1 or more axons found on the opposite side of the stereotyped location was counted as having an L/R choice defect. Any worm with a visible discontinuation of the dorsal and/or ventral cord between any two neurons (gaps) was considered to have a dorsal or ventral cord defect, respectively. AIY axonal morphology was scored with mgIs32 at the L4 to adult stage with a Plan-Apochromat 16x/0.5 objective on a Zeiss Axioimager Z1. Cells were considered to be overbranching if an additional protrusion off of the cell body at least the length of the cell body was present. PVQ axonal guidance was scored in oyIs14 at the L4 to adult stage with a Plan-Apochromat 16x/0.5 objective on a Zeiss Axioimager Z1. Crossover defects were counted if any portion of the axons of PVQL and PVQR anterior to the vulva were not visibly separable. For DiI staining, well-fed adult worms were washed from plates with 1 mL M9, pelleted, resuspendend in 1 mL M9 before 5 μL of DiI stock solution (2 mg/mL DiI (Molecular Probes, #D282) in dimethyl formamide) was added. Animals were incubated with light shaking for 3 hours, spun and washed with M9 two times, followed by transfer to fresh plates. After 30 minutes recovery, animals were immobilized in 1 mM levamisole and neuronal processes were assessed for defects and imaged with a Plan-Apochromat 40x/1.3 or 63x/1.4 objective on a Zeiss Axioimager Z1 Apotome.
Fluorescent images were captured in live C. elegans at the L4 larval stage using a Plan-Fluor Nikon 40x/1.3x on a Nikon CSU-W1 Spinning Disk Confocal. At least 15 L4 larval animals were scored per genotype. Optical sections were collected and maximum intensity projections adjusted for optimal contrast to resolve detailed morphology were used for further analysis. For tracing, the 100 μm section of the primary branch anterior to the cell body was used for morphometric analyses using the NeuronJ plugin of the FIJI software. Branches were defined as follows: 2° dendritic branches as any neurite branching out of the primary dendrite; 3° branches as neurites branching at the end of a 2° branch in proximity to the tertiary line along the border of outer body wall muscles and the lateral epidermis; 4° dendritic branches as those originating from 3° dendritic branches and extending towards the dorsal or ventral nerve cords, respectively.
All DMA-1::GFP fluorescent images were captured in strains containing qyIs369 at the L4 larval stage using a Plan-Fluor Nikon 60x/1.4x objective and GFP filter setting on a Nikon CSU-W1 Spinning Disk Confocal with identical imaging parameters and appropriate z-stack. Fluorescence measurements of the z-projected image were subsequently performed in ImageJ as follows. For Intensity of GFP in the PVD cell body an outline of the cell body was drawn as Region of Interest (ROI) and the average signal in the ROI was quantified. An ROI of the same shape was placed outside of the cell body to measure the background signal, which was subtracted from the fluorescent signal. For the intensity of GFP in the PVD primary dendrite, the primary dendrite was traced for 100 μm anterior to the cell body, and an ROI was defined as 10 pixels (= 1.6 μm) thick along the traced line, where the average signal in the ROI was quantified. A ROI of the same size was placed next to the dendrite to measure the background signal, which was subtracted from the fluorescent signal. The number of puncta within 100 μm of the primary dendrite anterior to the cell body were counted manually.
All HPO-30::tagBFP fluorescent images were captured in strains containing dzEx2140 and wdIs52 at the L4 larval stage using a 40x/1.3 objective on a Zeiss Axioimager Z1 Apotome with identical imaging parameters for tagBFP and GFP filter setting, and appropriate z-stacks. Quantification of the z-projected images were subsequently performed in ImageJ as follows. For the percentage of quaternary dendrites containing HPO-30::tagBFP, the number of quaternary branches with visible tagBFP signal within 100 μm anterior to the cell body were counted in the tagBFP channel, then divided by the number of quaternary dendrites as imaged in the GFP channel. For the number of HPO-30::tagBFP puncta, quantification was performed by manually counting the puncta within 100 μm anterior to the cell body.
Fluorescent images for the strains containing the ddIs290 fosmid based SAX-7::GFP reporter or dz249[lect-2::GFP] endogenous knock-in were captured at the late L4 larval stage using a Plan-Apochromat 40x/1.3 objective on a Zeiss Axioimager Z1 Apotome at an exposure time of 151 ms, followed by appropriate z-stacks. Neuronal and epidermal patterning were assessed, and the PVD phenotype was observed in a dzIs53 background.
L4 animals carrying dzSi3 and single-copy insertions of GFP subcellular compartment markers were imaged using a Plan-Fluor Nikon 60x/1.4x objective on a Nikon CSU-W1 Spinning Disk. Single planes from a confocal image stack in which both the catp-8 and organellar reporters were clearly visible were manually thresholded, then analyzed using the Correlation Plot application in Metamorph 7.7 to establish a Pearson’s Correlation Coefficient for each image pair.
Animals carrying the heIs63 transgene were synchronized and imaged 4.5 hours after hatching using a Plan-Fluor Nikon 100x/1.4x objective on a Nikon CSU-W1 Spinning Disk Confocal, before the first cell division of the Q neuroblast. The normalized position of the Q neuroblast was calculated by first drawing a straight line between the nuclei of V4 and V5, then the perpendicular location of nuclei of Q neuroblasts on the line was noted. The normalized position was calculated by subtracting the distance between the nuclei of V5 and Q neuroblasts (negative if posterior to V5) from that between the nuclei of V4 and Q neuroblasts (negative if anterior to V4) and divided by that between the nuclei of V4 and V5.
All statistical analyses were performed using the Prism 9 Statistical Software suite (GraphPad). All pairwise migration phenotypes are analyzed with Fisher exact test when applicable, or Chi-squared test as indicated. All comparisons of averages/proportions are analyzed with one-way ANOVA with Tukey correction, the Kruskal-Wallis test, the Mann-Whitney test, or the Z-test as applicable.
We thank S. Emmons, A. Meléndez, Y. Salzberg, and members of the Bülow laboratory for comments on the manuscript and for helpful discussions during the course of this project. We thank S. Cook for the initial observation of PQR cell migration defects, R. Poole for help with the analysis of whole genome sequencing data and, Yuji Kohara for cDNA clones and M. Harterink for a plasmid; some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs.