Report

Wnt Gradient Formation Requires Retromer Function in Wnt-Producing Cells

See allHide authors and affiliations

Science  12 May 2006:
Vol. 312, Issue 5775, pp. 921-924
DOI: 10.1126/science.1124856

Abstract

Wnt proteins function as morphogens that can form long-range concentration gradients to pattern developing tissues. Here, we show that the retromer, a multiprotein complex involved in intracellular protein trafficking, is required for long-range signaling of the Caenorhabditis elegans Wnt ortholog EGL-20. The retromer functions in EGL-20–producing cells to allow the formation of an EGL-20 gradient along the anteroposterior axis. This function is evolutionarily conserved, because Wnt target gene expression is also impaired in the absence of the retromer complex in vertebrates. These results demonstrate that the ability of Wnt to regulate long-range patterning events is dependent on a critical and conserved function of the retromer complex within Wnt-producing cells.

During C. elegans early larval development, EGL-20 is expressed by a group of cells located at the posterior end of the animal (1). EGL-20 controls the anterior migration of the HSN neurons (2) and the polarity of the division of the epidermal seam cell V5 (3). In addition, EGL-20 regulates the left/right asymmetric migration of the more distantly located Q neuroblasts (2) (fig. S1A). EGL-20 generates this asymmetry by specifically activating the expression of the Hox gene mab-5 in the left Q cell (QL) (1). MAB-5 in turn directs the migration of the QL daughter cells (QL.d) toward the posterior (4). The Q daughter cells on the right side (QR.d) migrate in the default anterior direction. We used this asymmetric migration as an assay to identify novel components of the EGL-20 pathway. In a genome-wide RNA-mediated interference (RNAi)–based screen (5), we found that an ortholog of the yeast retromer complex subunit Vps35p (table S1) is required for posterior localization of the QL.d. In yeast, the retromer directs endosome-to-Golgi retrieval of proteins such as the carboxypeptidase Y receptor Vps10p (6). In vertebrate epithelial cells, it physically interacts with the immunoglobulin receptor (IgR) and mediates basal-to-apical transcytosis of the IgR-IgA complex (7).

We isolated a deletion allele that likely represents the vps-35 null phenotype (fig. S1B). vps-35(hu68) mutants showed loss of mab-5 expression in the QL lineage and anterior migration of the QL.d (Fig. 1, A and B). Further analysis of vps-35(hu68) showed that the vps-35 and egl-20 mutant phenotypes are strikingly similar. Thus, vps-35 is also required for the EGL-20–dependent anterior migration of the QR.d and HSN neurons and for establishing the correct polarity of the V5 division (fig. S2A and table S5). Apart from these phenotypes, vps-35 mutants showed no obvious additional defects, indicating that other signaling pathways are not substantially affected. Mutation of the two additional core retromer components vps-26 and vps-29 resulted in similar phenotypes (fig. S2A and table S2). Loss of vps-5, which encodes an accessory subunit of the retromer (6, 8), did not affect Q cell migration. However, it did enhance the QL.d migration defect of vps-29(tm1320) and vps-26(RNAi) (fig. S2A and table S2), indicating that VPS-5 has a minor role in this process.

Fig. 1.

The retromer complex is required for Q cell migration. (A) In vps-35(hu68), a mab-5::lacZ reporter is not expressed in the QL.d. Whole-mount larvae were fixed 3 to 5 hours after hatching and stained for β-galactosidase activity. The DNA stain 4′,6′-diamidino-2-phenylindole (DAPI) was used for cell identification. (B) Anterior localization of the QL.d in vps-35(hu68). The final positions of the QL.d (n = 50) were determined relative to the invariant positions of the seam cells (V1 to V6) after their first division. Bars indicate the percentage of cells at each position. Dashed lines indicate the wild-type position.

EGL-20 signaling is mediated by canonical as well as noncanonical Wnt pathways (9), and the requirement of the retromer complex for both signaling functions indicates that it acts before the divergence point of these different pathways. Epistatic analysis showed that VPS-35 functions upstream of the BAR-1/β-catenin destruction complex (fig. S2A and table S2), suggesting that it acts at the level of Frizzled or Dishevelled, or upstream of EGL-20 itself. To distinguish between these possibilities, we investigated whether the retromer complex is required in EGL-20–responding or –producing cells. Because vps-35 is widely expressed, we used different tissue-specific promoters to express vps-35 in a vps-35(hu68) mutant background (table S3 and fig. S2B). We found that expression of vps-35 in cells that respond to EGL-20, such as the Q neuroblasts and the HSN neurons, failed to rescue the mutant phenotype. Rescue was, however, obtained when vps-35 was expressed in all or part of the EGL-20–producing cells (see supporting online material). Taken together, these results demonstrate that the presence of the retromer complex in EGL-20–producing cells is necessary and sufficient for its function in EGL-20 signaling.

Mutation of the retromer complex strongly disrupts EGL-20 function, but other Wnts such as MOM-2 and LIN-44 are only mildly affected (table S4). The known functions of MOM-2 and LIN-44 depend on short-range signaling (10, 11), whereas EGL-20 is the only Wnt that has been demonstrated to signal over a distance (1). We therefore investigated whether loss of retromer function differently affects long- and short-range functions of EGL-20. To test this, we compared the effect of different mutations in egl-20, vps-35, and vps-26 on the Q neuroblasts, which are located relatively far from the group of EGL-20–producing cells, and the V5 cell, which is positioned closer. All mutations similarly disrupted QL.d migration (fig. S2), but V5 polarity was more strongly affected in egl-20 null mutants than in vps-35(hu68) and vps-26(tm1523) (table S5). These results suggest that long-range EGL-20 signaling is more dependent on retromer function than is short-range signaling.

In the Drosophila wing imaginal disc, Wingless forms a concentration gradient that enables signaling over a distance (12). In C. elegans, egl-20 is expressed in the posterior of the animal and triggers responses in more anteriorly located cells (fig. S1A). Together with the different effects on long- versus short-range EGL-20 signaling, this suggests that the retromer may be required for the formation of an EGL-20 gradient. We therefore investigated whether loss of retromer function affects EGL-20 localization. Immunostaining of animals expressing a functional EGL-20::proteinA fusion under the control of the egl-20 promoter revealed that EGL-20 forms an anteroposterior gradient (Fig. 2, A to C, and E). The gradient is mainly visible at early stages, when EGL-20–dependent processes such as mab-5 regulation in the Q neuroblasts take place. As observed for Wingless in Drosophila (12), EGL-20 shows a punctate staining. The EGL-20 gradient was absent in vps-35(hu68) mutants (Fig. 2D). Taken together, these results show that the retromer is involved in a critical step within EGL-20–expressing cells that allows EGL-20 to spread anteriorly.

Fig. 2.

The retromer complex is required for EGL-20 gradient formation. Whole-mount larvae carrying the integrated Pegl-20::egl-20::proteinA transgene were fixed 1 to 5 hours after hatching and stained with a fluorescein isothiocyanate–coupled antibody. (A and B) In wild-type animals, EGL-20 forms a punctate anteroposterior gradient. (C) This gradient is not affected in bar-1(ga80) mutants. (D) In vps-35(hu68) mutants, the EGL-20 gradient is absent. Arrowheads indicate P11/12 (23). White bars delimit the group of egl-20–expressing cells. Dashed lines represent the range of the gradient. A similar gradient was observed in live animals using an EGL-20::Venus fusion (fig. S4). (E) Schematic representation of the EGL-20 gradient. The punctate EGL-20 staining is indicated by green dots; egl-20–expressing cells are in blue. (F) Loss of vps-35 does not affect global levels of EGL-20. Western blots of total first and second larval stage lysates were stained for EGL-20::proteinA. Neg, nontransgenic animals. Equal loading was confirmed using an antibody to α-tubulin.

The retromer complex is conserved from yeast to vertebrates (table S1). Therefore, we investigated its potential role in vertebrate Wnt signaling. RNAi-mediated knockdown of Vps35 in human embryonic kidney (HEK) cells inhibited the activation of a TCF reporter gene upon Wnt3A transfection (Fig. 3, A and B). As expected from the C. elegans data, Vps35(RNAi) had no effect when the reporter was activated downstream of Wnt by overexpression of a constitutively active TCF–β-catenin fusion (13). This shows that Vps35 has a conserved function that is also required for effective signaling in tissue culture. Furthermore, to address the in vivo effect of Vps35 knockdown in vertebrates, we used a morpholino (MO) approach in Xenopus tropicalis (fig. S3). Ectopic ventral activation of the Wnt pathway induces a secondary body axis (14, 15). When a Vps35 MO was co-injected ventrally with Wnt1 RNA, axis duplication was strongly inhibited (Fig. 4A). Again, Vps35 knockdown had no effect on axis duplication when Wnt target gene expression was ventrally induced by a constitutively active TCF3-Armadillo fusion (16). The absence of a functional retromer complex also impaired endogenous Wnt signaling. Asymmetric injection of Vps35 MO specifically inhibited the expression of the Wnt target genes XmyoD (17) and Xslug (18) on the injected side (Fig. 4B) but did not affect the mesodermal marker Xbra. Co-injection of β-catenin or TCF3-Armadillo expression constructs rescued XmyoD and Xslug expression, showing that the vertebrate retromer specifically acts in the Wnt pathway, upstream of β-catenin (Fig. 4B and table S6). By analogy with our results in C. elegans, we propose that Wnt spreading and long-range signaling in vertebrates are dependent on a functional retromer.

Fig. 3.

Vps35 RNAi in mammalian cells impairs Wnt signaling but not Wnt secretion. (A) A stable HEK TR clone inducible for Vps35 RNAi (HEKVps35) was obtained using a set of three different targeting constructs. Down-regulation of Vps35 mRNA was checked 96 hours after doxycycline induction by Northern blot. Equal loading was confirmed by ethidium bromide staining (EtBr). (B) Activation of a TCF reporter (black bars) by Wnt3A transfection is inhibited by doxycycline-induced Vps35 RNAi. Activation by TCF–β-catenin is not affected. White bars represent negative controls. Data are presented as means ± SD. (C) Wnt3A levels are identical in the medium of cells induced or noninduced for Vps35 RNAi, demonstrating that Wnt secretion was not impaired despite a slightly lower amount of Wnt3A::proteinA in the corresponding total cell lysates. Neg., Nontransfected control; Dox, doxycycline.

Fig. 4.

The retromer complex is required in vivo for vertebrate Wnt signaling. (A) Vps35 MO inhibits Wnt1-induced double axis formation but has no effect on a constitutively active TCF3-Armadillo fusion (n ≥ 50 in each assay). Data are presented as means ± SD. (B) Vps35 knockdown inhibits endogenous Wnt target gene expression on the MO-injected side. In situ hybridizations for the Wnt target genes XmyoD and Xslug were performed at the appropriate stages. Co-injection of β-catenin or TCF3-Armadillo expression constructs rescued Wnt target gene expression (representative embryos are shown). Arrowheads indicate the side of injection. N.I., noninjected control.

Retromer function may be required for the formation of an active Wnt protein, for secretion, or for other processes within the expressing cells that enable long-range Wnt signaling. Several lines of evidence suggest that retromer function is not required for the formation and secretion of an active form of Wnt. First, the specificity of the retromer mutant phenotype argues against a general role in active Wnt formation. Thus, whereas mutation of the Porcupine ortholog MOM-1 (which is required for active Wnt formation) (19, 20) leads to embryonic lethality (10, 21), mutation of the retromer complex primarily disrupts EGL-20 function. Second, our observation that loss of retromer function mainly affects long-range EGL-20 signaling indicates that short-range–acting forms of EGL-20 and other Wnts are still produced and secreted. Third, mutation of vps-35 did not affect EGL-20 levels within the EGL-20–producing cells (as observed by immunostaining and with a functional EGL-20::GFP fusion) (Fig. 2, D and F), indicating that EGL-20 is normally produced and secreted. Finally, knockdown of Vps35 in transfected HEK cells did not affect Wnt3A secretion (Fig. 3C).

It has recently been shown that binding of Wnt to lipoprotein particles is important for long-range gradient formation (22). The association of Wnt with these particles may take place within Wnt-producing cells, in an endosomal compartment that contains both Wnt and internalized lipoprotein particles. This mechanism would require Wnt to be sorted out of the default secretory pathway by a specific cargo receptor and to be transported to endosomes before secretion. We speculate that, similar to its function in Vps10p trafficking, the retromer complex may recycle the Wnt cargo receptor from the endosome to the Golgi network. In the absence of retromer, a lack of Wnt cargo receptors in the Golgi would lead to Wnt secretion via the default pathway. This would impair Wnt association with lipoprotein particles and limit its range of signaling.

Wnt gradient formation is a complex and tightly regulated process. Factors expressed at the surface of cells along the gradient domain control its spatial extension and determine the range of its effects. However, our results show that the establishment of a Wnt concentration gradient is not dependent only on the capacity of neighboring cells to attract and spread Wnt proteins. An early permissive mechanism within the Wnt-producing cells, which requires retromer function, is an essential step that allows secreted Wnt molecules to respond to the different cues that will shape the gradient.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1124856/DC1

Materials and Methods

Figs. S1 to S4

Tables S1 to S6

References

References and Notes

View Abstract

Navigate This Article