R-spondins are BMP receptor antagonists in Xenopus early embryonic development

BMP signaling plays key roles in development, stem cells, adult tissue homeostasis, and disease. How BMP receptors are extracellularly modulated and in which physiological context, is therefore of prime importance. R-spondins (RSPOs) are a small family of secreted proteins that co-activate WNT signaling and function as potent stem cell effectors and oncogenes. Evidence is mounting that RSPOs act WNT-independently but how and in which physiological processes remains enigmatic. Here we show that RSPO2 and RSPO3 also act as BMP antagonists. RSPO2 is a high affinity ligand for the type I BMP receptor BMPR1A/ALK3, and it engages ZNRF3 to trigger internalization and degradation of BMPR1A. In early Xenopus embryos, Rspo2 is a negative feedback inhibitor in the BMP4 synexpression group and regulates dorsoventral axis formation. We conclude that R-spondins are bifunctional ligands, which activate WNT- and inhibit BMP signaling via ZNRF3, with implications for development and cancer.

B one Morphogenetic Proteins (BMPs) are a subfamily of TGFβ growth factors that exert a plethora of crucial functions in embryonic development, adult tissue homeostasis, as well as regeneration, and they underlie human pathology such as skeletal disorders, cancer, and fibrosis in multiple organs [1][2][3][4][5] . Due to their accessibility, extracellular components of the BMP pathway are of particular interest as therapeutic targets 6 and mechanistic understanding of receptor modulation should improve the ability to manipulate BMP-dependent processes.
BMPs signal through a tetrameric receptor kinase complex composed of type I (BMPR1A/ALK3, BMPR1B/ALK6, ACVR1/ ALK2, or ACVRL1/ALK1) and type II receptors (BMPR2, ACVR2A, ACVR2B) 7 . Ligands and receptors combine in a combinatorial fashion 8 and phosphorylate SMAD1, 5, and 8, which enter the nucleus with SMAD4 to regulate target gene expression 9,10 . There exists a multitude of extracellular modulators of TGFβ signaling, either soluble or membrane-associated proteins that control ligand availability, processing, ligand-receptor interaction, and receptor activation 11 . However, only two BMP receptor antagonists are known, which directly bind and inhibit receptor function, the TGFβ-family proteins BMP3 and Inhibin 12,13 .
Unexpectedly, recent studies showed that RSPO2 and RSPO3 can potentiate WNT signaling in the absence of all three LGRs in vitro and in vivo 27,30 . Moreover, WNT and RSPO ligands are functionally non-equivalent since e.g., WNT ligand overexpression cannot induce crypt expansion in contrast to RSPO2 or RSPO3 31 and RSPO2 and WNT1 have distinct effects on mammary epithelial cell growth 32 and cochlea development 33 . Hence, these inconsistencies in our current understanding raise the questions: do RSPOs possess WNT-independent functions? Do they engage other receptors? If so, in which physiological processes is this relevant?
Here we show that RSPO2 and RSPO3 are high affinity ligands for the BMP receptor BMPR1A/ALK3. RSPO2 forms a ternary complex between BMPR1A and the E3 ligase ZNRF3, which triggers endocytosis and degradation of the BMP receptor. We show that Rspo2 antagonizes BMP signaling during embryonic axis formation in Xenopus. By gain-of-function and loss-of-function experiments rspo2 cooperates with Spemann organizer effectors to regulate the BMP morphogen gradient, which controls dorsoventral axis formation. Our study reveals R-spondins as BMP receptor antagonists in development, inviting re-interpretation of the mode of action of R-spondins and ZNRF3 in stem cell and cancer biology.
To delineate the domains required for BMP inhibition, we analyzed deletion mutants of RSPO2 and found both the TSP1domains and FU-domains to be important for signaling inhibition (Fig. 1i, j) 24 . We next investigated RSPO2 deficiency in H1581 cells, a human large cell lung carcinoma cell line that expresses high levels of RSPO2 ( Supplementary Fig. 1a). Knockdown of RSPO2 but not LRP5/6 sensitized H1581 cells to BMP stimulation (Fig. 1k, l, Supplementary Fig. 1m, n). We conclude that RSPO2 and RSPO3 antagonize BMP signaling independently of WNT signaling.
Rspo2 antagonizes BMP signaling during Xenopus embryonic axis development. To analyze if Rspo2 inhibits BMP signaling in vivo, we turned to early Xenopus development. In the early amphibian embryo, the Spemann organizer is a small evolutionary conserved signaling center, which plays an eminent role in regulating embryonic axis formation and neural induction. One essential molecular mechanism underlying Spemann organizer function resides in its secretion of BMP antagonists, which create a BMP morphogen gradient that patterns the embryo [35][36][37] . Since rspo2 is expressed and functions in WNT-mediated myogenesis of early Xenopus embryos 20 , we analyzed if it may have an additional role as BMP antagonist in axial patterning.
bmp4 overexpression ventralizes Xenopus embryos, resulting in small heads and enlarged ventral structures 38 . Injection of wildtype rspo2 mRNA, but neither its ΔFU1/2 nor ΔTSP1 deletion mutants rescued these bmp4-induced malformations (Fig. 2a, b). This domain requirement is different from that for WNT signaling activation, where only FU1 and FU2 but not the TSP1 domain are essential 20 . Conversely, injection of a previously characterized rspo2 antisense Morpholino (Mo) 20 increased endogenous BMP signaling, and this was unaffected by lrp6 Mo (Fig. 2c) 39 . Strikingly, coinjection of bmp4 Mo and rspo2 Mo neutralized each other in BMP signaling reporter assay (Fig. 2d), BMP target gene expression (vent1, sizzled) (Fig. 2e, f), as well as defects in dorsoventral axis development ( Supplementary Fig. 2a,  b). Typically, overexpression of common BMP antagonists such as noggin or chordin that sequester BMP ligands, leads to strongly dosalized Xenopus embryos, with enlarged heads and cement glands [35][36][37] . In contrast, overexpression of rspo2 failed to induce enlarged heads but instead induced spina bifida with reduced head structures, yielding the first indication that rspo2 does not act by the common mode of sequestering BMP ligands ( Supplementary Fig. 2c).
Rspo2 is a negative feedback regulator in the Xenopus BMP4 synexpression group. In early vertebrate embryos, genes belonging to certain signaling networks form characteristic synexpression groups, i.e., genetic modules composed of genes   that show tight spatio-temporal RNA coexpression and that function in the respective signaling pathway 40 . A wellcharacterized example is the BMP4 synexpression group, members of which are expressed like this growth factor-dorsally in the eye, heart and proctodeum of tailbud stage Xenopus embryos (Fig. 3a). This group consists of at least eight members, which all encode positive or negative feedback components of the BMP signaling cascade as studied in early development, including ligands, receptors and downstream components of the pathway 41 . Interestingly, we found that rspo2 is part of the BMP4 synexpression group, being coexpressed with bmp4 from gastrula to tadpole stages (Fig. 3a), suggesting that its expression depends on BMP signaling as for other synexpressed genes. To test this idea, we employed Xenopus animal cap explants, which express low levels of rspo2 and bmp4 to monitor rspo2 induction upon bmp4 overexpression (Fig. 3b). Indeed, bmp4 induced rspo2 expression by qRT-PCR (Fig. 3c) and in situ hybridization (Fig. 3d, e), similar to bmp4 direct targets sizzled (Fig. 3c-e) and vent1 (Fig. 3c). To test whether rspo2 is an immediate early target of BMP4, we blocked protein synthesis with cycloheximide (CHX) 41 . Interestingly, while induction of the direct BMP4 targets sizzled and vent1 by bmp4 was unaffected by CHX, rspo2 induction was inhibited ( Fig. 3b-e). We conclude that rspo2 is a negative feedback inhibitor within the BMP4 synexpression group and that it is an indirect BMP target gene, whose expression may depend on transcription factors of the e.g., Vent or Msx families 41,42 (Fig. 3f).
RSPO2 and RSPO3 bind BMPR1A via the TSP1 domain to antagonize BMP signaling. Given that RSPOs act by promoting receptor endocytosis 14,17 , we postulated that RSPO2 might regulate BMP signaling through its receptors: ACVR1, BMPR1A and BMPR1B. To test this hypothesis, we analyzed the effect of RSPO1-4 treatment on BMP signaling induced by constitutively active ACVR1/BMPR1A/BMPR1B (ACVR1/BMPR1A/BMPR1B QD ). Interestingly, RSPO2 and RSPO3 treatment specifically inhibited BMPR1A QD but not ACVR1 QD or BMPR1B QD , while RSPO1 and RSPO4 had no effect to any of the constitutively active receptors (Fig. 4a-c).
Indeed, cell surface binding assay and in vitro binding assay revealed that RSPO2 and RSPO3, but not RSPO1 and RSPO4, bound the extracellular domain (ECD) of BMPR1A (Fig. 4d, e, Supplementary Fig. 5a). RSPO2 showed high affinity with BMPR1A ECD (K d ≈ 4.8 nM) (Fig. 4f), comparable to the RSPO-LGR interaction 24 . To further delineate the domains required for BMPR1A binding, we analyzed deletion mutants of RSPO2 in cell surface binding assays with BMPR1A ECD, and found BMPR1A binding required the TSP1 but not the FU domains of RSPO2, while, conversely, LGR binding required the FU domains but not TSP1 ( Supplementary Fig. 5b, c). The importance of the TSP1 domain was confirmed by in vitro binding assay showing that the isolated TSP1 domain of RSPO2, but not RSPO1, was sufficient to interact directly with BMPR1A ECD (Fig. 4g, h). Similarly, BMPR1A binding required the TSP1 domain also in RSPO3, suggesting that an analogous mode of binding applies to RSPO2 and RSPO3 ( Supplementary Fig. 5d, e).
The importance of the TSP1 domain in BMP inhibition was further corroborated in Xenopus, where we took advantage of the fact that the TSP1-domain is encoded by a distinct exon in the 3′end of the rspo2 gene. We generated a rspo2 Mo (rspo2 ΔTSP Mo), which specifically abolished TSP1-domain splicing, yielding 3' truncated rspo2 mRNA lacking the TSP1 domain but retaining the FU domains (Fig. 5a). Microinjection of rspo2 ΔTSP Mo resulted in ventralized tadpoles with shorter axis and reduced heads compared to control tadpoles, which was partially rescued by introducing a non-targeted rspo2 mRNA ( Supplementary  Fig. 6a, b). rspo2 ΔTSP Morphants had no effect on WNT signaling (Fig. 5b), confirming that it does not interfere with Rspo2 FU domains that are essential for WNT activation. However, rspo2 ΔTSP Mo increased BMP signaling (Fig. 5c). Similar to chordin and rspo2 Morphants, rspo2 ΔTSP Morphants showed expanded expression of the BMP target genes vent1 and sizzled in gastrulae (Fig. 5d, e, Supplementary Fig. 6c, d) 38 , and corresponding tadpoles were ventralized, displaying decreased bf1 and myoD and increased sizzled expression (Fig. 5f, g) 38 . Coexpression of dominant negative bmpr1a (bmpr1a DN ) rescued these defects ( Fig. 5d, g, Supplementary Fig. 6c, d). Taken together, these results emphasize that the TSP1 domain is a key element in providing target specificity to RSPOs, both in vitro and in vivo, and that it dictates their BMP-inhibitory function. b Quantification of embryonic phenotypes shown in a. 'Ventralized' represents embryos with both small head and enlarged ventral structure, reminiscent of BMP hyperactivation. 'CE defect' refers to embryos with convergent extension (gastrulation) defects, unrelated to BMP signaling. Note that rspo2 mRNA dosage used in a was below those that cause gastrulation defects. n = number of embryos. c, d BMP-(vent2) reporter assays with Xenopus laevis neurulae (St.15) injected with reporter plasmids and the indicated Mo at 4-cell stage. n = biologically independent samples and data are displayed as means ± SD. ns, not significant. *P < 0.1, **P < 0.01, ****P < 0.0001 from two-tailed unpaired t-test. e In situ hybridization of vent1 and sizzled in Xenopus laevis gastrulae (St.11, dorsal to the top, vegetal view) injected as indicated. D, dorsal, V, ventral. Asterisk, abolishment of the expression. Dashed line, dorsal blastopore lip (dbl). Scale bar, 0.5 mm. f Quantification of embryonic phenotypes shown in (e). 'Expressed', normal, increased or reappearance of vent1/sizzled expression. 'Abolished', complete absence of vent1/sizzled expression. Data are pooled from two independent experiments. n = number of embryos. g Microinjection strategy and representative phenotypes of Xenopus tropicalis tadpole (St.30) Crispants and tadpoles (St.30) injected with bmp4 mRNA or lrp6 Mo. At 1-cell stage, Cas9 protein with guide RNA (gRNA) targeting rspo2 or chd, or both gRNAs were injected animally. Dashed lines, head size. Arrowheads, enlarged ventral structure. h Quantification embryonic phenotypes shown in g. 'Severe' showed small head, enlarged ventral tissues and short body axis. 'Mild' showed one or two of the defects described above. 'Normal' showed no visible differences to the uninjected control. n = number of embryos. ns, not significant. **P < 0.01, ***P < 0.001, ****P < 0.0001 from two-tailed χ2 test comparing normal versus ventralized phenotypes (b), twotailed χ2 test comparing expressed versus abolished (f), or two-tailed χ2 test comparing normal versus severe and mild defects h. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19373-w ARTICLE NATURE COMMUNICATIONS | (2020) 11:5570 | https://doi.org/10.1038/s41467-020-19373-w | www.nature.com/naturecommunications RSPO2 destabilizes the BMP receptor BMPR1A. To investigate the consequence of RSPO-BMPR1A binding, we monitored BMPR1A protein levels upon RSPO2 knockdown in H1581 cells and found that siRSPO2 treatment increased BMPR1A protein levels ( Fig. 6a). Similarly in Xenopus whole embryos, microinjection of mRNA encoding rspo2 but not rspo2 ΔFU1/2 or rspo2 ΔTSP decreased protein levels from coinjected bmpr1a-EYFP mRNA (Fig. 6b). Immunofluorescence microscopy (IF) of Xenopus animal cap explants showed that Bmpr1a-EYFP localizes to the plasma membrane, where it was once again reduced by rspo2 but not by rspo2 ΔFU1/2 or rspo2 ΔTSP mRNA ( Fig. 6c-e). Focusing on Xenopus ventrolateral marginal zone (VLMZ) explants, where endogenous rspo2, bmpr1a and bmp4 are coexpressed, showed that ablation of rspo2 by Mo injection results in significant increase of Bmpr1a-EYFP plasma membrane levels ( Fig. 6f-h). Moreover, in VLMZ from rspo2 ΔTSP Morphants, Bmpr1a levels were also increased ( Fig. 6f-h), which was confirmed by western blot analysis (Fig. 6i). Altogether, our results suggest that RSPO2 destabilizes BMPR1A.
RSPO2 requires ZNRF3 to antagonize BMP receptor signaling. We next turned to the role of the FU domains in RSPO2, which are also required for inhibition of BMP signaling (Figs. 1j, 2a, b and 6b-e). FU1 and FU2 domains confer RSPO binding to ZNRF3/RNF43 and LGRs, respectively 27 . Since our results demonstrated an LGR-independent mode of action (Fig. 1f), and Dashed lines, dorsal blastopore lip (dbl); anp, anterior neural plate; di, diencephalon; mhb, mid-hindbrain boundary; ov, otic vesicle; a, atria; nt, neural tube; de, dorsal eye; h, heart; pdm, proctodeum. Scale bar, 0.5 mm. b Microinjection and experimental scheme for c-e. 2 or 4 cell stage Xenopus laevis embryos were animally injected with control (ppl) or bmp4 mRNA. The animal cap (AC) explants were dissected from injected embryos at stage 8, and either treated or untreated with cycloheximide (CHX) until control embryos reached stage 10 for qRT-PCR (c) or in situ hybridization (d, e). c qRT-PCR of rspo2, sizzled, and vent1 expression in the AC explants injected and treated as indicated. Data are pooled from three independent experiments with similar results and displayed as means ± SD. ns, not significant, *P < 0.05, **P < 0.01, ****P < 0.0001 from two-tailed unpaired t-test. d In situ hybridization of rspo2 and sizzled in the AC explants injected and treated as indicated. e Quantification of (d). n = number of the AC explants. f Model for Rspo2 function as a negative feedback inhibitor of BMP4 in Xenopus dorsoventral patterning.
Interestingly, znrf3 Morphants at neurula showed increased BMP signaling by BMP-reporter assay and rspo2 mRNA coinjection could not reduce it (Fig. 7c). Moreover, IF in Xenopus animal cap explants showed that rspo2-induced destabilization of Bmpr1a protein levels was prevented by ZNRF3 ΔR (Fig. 7d, e). Altogether, these results support that to function as BMP antagonist, RSPO2 requires ZNRF3.
RSPO2 bridges BMPR1A and ZNRF3 and triggers BMP receptor clearance from the cell surface. The interaction of RSPO2 and RSPO3 with BMPR1A, as well as ZNRF3, suggested that R-spondins bridge both transmembrane proteins. In vitro binding assays (Fig. 8a, b) and colocalization by IF (Fig. 8c, d, Supplementary Fig. 9a, b), confirmed that ZNRF3 interacted with BMPR1A in the presence of RSPO2 or RSPO3 but not of RSPO1. Emphasizing once again the importance of the FU1 and TSP1 domains for this interaction, in vitro ZNRF3-BMPR1A-RSPO2 ternary complex formation was prevented by TSP1, FU1/2, or FU1 deletion ( Supplementary Fig. 9c-g), whereas it remained intact upon FU2 deletion ( Supplementary Fig. 9h).
Since ZNRF3/RNF43 eliminate WNT receptors from the cell surface by co-internalization and lysosomal degradation 25,26 , we considered an analogous function in BMPR1A turnover. We monitored BMPR1A localization by IF in H1581 cells and found that it was absent from the plasma membrane but abundantly colocalized with ZNRF3 in cytoplasmic vesicles (Fig. 8e, i), suggesting that it may be internalized by endogenous RSPO2. Indeed, upon knockdown of RSPO2, but not LRP6 or LGR4/5, BMPR1A accumulated at the plasma membrane (Fig. 8f-i). Importantly, IF (Fig. 8j-m) and cell surface biotinylation assays (Fig. 8n) showed that upon ZNRF3/RNF43 siRNA treatment, BMPR1A also accumulated at the plasma membrane.
To test if RSPO2/ZNRF3 target BMPR1A for endocytosis and lysosomal degradation, we treated cells with the clathrin inhibitor monodansylcadaverin (MDC), which eliminated inhibition of BMP signaling by RSPO2 (Fig. 8o). In addition, siRSPO2 abolished the colocalization of BMPR1A with the early endosome marker EEA1 (Fig. 8p-q) and lysosomal marker Lamp1 ( Fig. 8r-s), suggesting that RSPO2 binding promotes BMPR1A internalization and degradation via ZNRF3 ternary complex formation. Consistently, 20 min exposure to RSPO2 increased internalized BMPR1A in cell surface biotinylation assays in H1581 cells (Supplementary Fig. 10a) and induced vesicular Bmpr1a-EYFP in Xenopus animal caps ( Supplementary Fig. 10b, c). Taken together, our results support a model ( Supplementary  Fig. 10d) wherein RSPO2 bridges ZNRF3 and BMPR1A and routes the ternary complex towards clathrin-mediated endocytosis for lysosomal degradation, thereby antagonizing BMP signaling. We suggest that a similar mechanism applies to RSPO3 but not RSPO1 and RSPO4.

Discussion
The three main findings of our study are (i) the discovery Rspondins as BMP receptor antagonists, (ii) that RSPO2 depletes BMPR1A/ALK3 by engaging ZNRF3 for internalization and lysosomal degradation, and (iii) that in Xenopus, rspo2 is a negative feedback inhibitor of the BMP4 synexpression group, which cooperates with Spemann organizer effectors to inhibit BMP signaling during axis formation. Given the importance of RSPOs and BMPs as developmental regulators, as well as growth factors of normal and malignant stem cells, these conclusions have implications for development and cancer.
With regard to stem cells, R-spondins are a key ingredient of the culture media, which have made the organoid revolution possible 21,22 and their rational use requires an understanding of their mechanism of action. For example, the fact that R-spondins inhibit BMP signaling may explain the reported non-equivalence of WNT and RSPO ligands in stem cells and development [31][32][33] . It may also explain their potency as stem cell growth factors, as e.g., intestinal stem cells requires both, WNT activation and BMP inhibition 21,22 .
TGFβ growth factors play an eminent role in biology and medicine, and their receptor signaling is exquisitely regulated extracellularly with over 20 TGFβ antagonists, most of which antagonize signaling by ligand sequestration (e.g., Cerberus, Chordin, Follistatin, Gremlin, Noggin, and Sost) 1,11 . Two extracellular BMP receptor antagonists are known, BMP3 and Inhibin 12,13 . Both are TGFβ family members, whose unproductive binding to type II receptors prevents signal transmission. Relatedly, the BMP antagonist BAMBI is a BMP pseudoreceptor lacking kinase activity, which also leads to formation of a deadend complex with BMP receptors 44 . In contrast, RSPO2 and RSPO3 share no sequence homology with TGFβ family members, they inhibit type I instead of type II BMP receptors, and they do so by engaging the ZNRF3 E3 transmembrane ubiquitin ligase to internalize BMPR1A. RSPO2 thereby routes BMPR1A to j TOPFlash reporter assay in HEPG2 cells upon WNT3A with or without (i) as indicated. k BRE reporter assay in HEPG2 cells upon BMP4 with or without (i) as indicated. For reporter assays (a-c, j, k), n = 3 biologically independent samples; In vitro binding assays (e, h), n = 3 experimentally independent samples. All data are displayed as mean ± SD. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 from two-tailed unpaired t-test (a, b, e, h, j, k) or one-way ANOVA test (c).
clathrin-mediated endocytosis for lysosomal degradation. This mode of action resembles the function of the Spastic Paraplegia related gene NIPA1, a transmembrane antagonist, which promotes BMP receptor type II endocytosis and lysosomal degradation 45 .
BMPR1A engages not only various BMPs but also GDFs 1 , and hence RSPO-mediated inhibition may potentially affect signaling in multiple contexts. On the other hand, the specificity of RSPO2 for BMPR1A may provide therapeutic opportunities on the background of pleiotropic BMP ligands effects.
RSPO2 engages ZNRF3 to antagonize BMP signaling, implying that ZNRF3 is also a negative regulator not only of WNT, but also BMP signaling. Consistently, our results indicate that loss of ZNRF3 increases BMP signaling. Moreover, ZNRF3 overexpression induces expression of Spemann organizer genes in Xenopus embryos, which is characteristic not only for WNT but also BMP inhibition 43 . In WNT signaling, the role of RSPO2 is to protect WNT receptors from ubiquitination and internalization by ZNRF3, by forming a ternary complex with LGR4-6 and triggering endocytosis. In contrast, during BMP signaling, RSPO2 directly forms a ternary complex with ZNRF3-BMPR1A to internalize and degrade the type I receptor. Our data also imply a possible function of RNF43 in antagonizing BMP signaling, inviting a closer inspection of its loss-of-function phenotypes 25,26 .
A number of studies emphasized the importance of the Furin domains in RSPOs, which are necessary and sufficient for activation of WNT signaling 17,20,26,28 , however, the role of the TSP1 domain has received less attention. We found that the specificity of RSPOs for BMP signaling is dictated by the TSP1 domain, which binds directly to BMPR1A. Unlike RSPO2 and RSPO3, RSPO1 and RSPO4 do not inhibit BMP signaling, the key difference residing in the TSP1 domain, as domain swapping of the TSP1 domain is sufficient to confer BMP inhibition upon RSPO1. i Quantification of h. n = areas analyzed and data are displayed as mean ± SD. ns, not significant, ****P < 0.0001 from one-way ANOVA analysis. All data are displayed as mean ± SD. The P values for reporter assays (a-c, f, g) were determined from two-tailed unpaired t-test. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. The physiological role in vivo is highlighted by rspo2 Morphants specifically lacking the TSP1 domain, which displayed phenotypic defects due to BMP hyperactivation (Fig. 5). The TSP1 domain also binds to heparin sulfate proteoglycans (HSPG) e.g., syndecans (SDC) 27,29,30 , which raises the possibility of cooperation between RSPOs and SDC in BMP receptor regulation. Indeed, SDC1 and SDC3 have been implicated as negative regulators in BMP signaling, but the underlying mechanisms remained unclear 47,48 . Hence, it will be interesting to investigate the role of SDCs in BMPR1A-RSPO interactions. HSPGs are also coreceptors in FGF signaling, which may explain why misexpressed rspo2 can inhibit FGF signaling in Xenopus animal cap explants 49 . We established that Xenopus Rspo2 cooperates with Noggin and Chordin released by the Spemann organizer in repressing BMP signaling to modulate the BMP morphogen gradient, which controls axial patterning. Yet, overexpression of rspo2 unlike of noggin and chordin, does not strongly dorsalize early embryos. The reason is that instead of sequestering BMP ligands, RSPO2 specifically targets the BMP receptor BMPR1A in early Xenopus embryos and that BMPR1A and BMPR1B play overlapping roles in dorsoventral patterning 50,51 . Also unlike noggin and chordin, rspo2 is not expressed in the organizer but is a negative feedback inhibitor of the BMP4 synexpression group, similar to the BMP pseudoreceptor bambi 44,52 . Like bambi, rspo2 is an indirect BMP4 target gene, which may require Vent or Msx transcription factors for expression. Negative feedback in BMP signaling expands the dynamic BMP signaling range essential for proper embryonic patterning and reduce inter-individual phenotypic and molecular variability in Xenopus embryos 42 . Indeed, rspo2 deficiency by itself has only mild effects on axis formation and dorsoventral marker gene expression, while defects manifest upon misbalance of BMP signaling (bmp4-overexpression, noggin/chordin knockdown). Functional redundancy between BMP antagonists is a characteristic feature observed in fish, frog, and mouse embryos [53][54][55][56] . We note that the mouse Rspo2 expression pattern at E9.5 mimics that of mouse Bmp4, including forebrain, midbrain/hindbrain junction, branchial arches and limb apical ectodermal ridge 57,58 . Thus, although Rspo2 deficient mouse embryos gastrulate normally 59 , it may be fruitful to analyze compound mutants between Rspo2 and BMP antagonists for axial defects. The fact that R-spondins are bifunctional ligands, which activate WNT signaling and inhibit BMP signaling has implications for development, stem cell biology, and cancer. Mechanistically, the general picture emerging is that R-spondins function as adapters, which escort client extracellular proteins for ZNRF3/RNF43-mediated degradation, e.g., LGR4-6 and BMPR1A. Our results assign a key role to the largely ignored TSP1 domain of R-spondins in providing target specificity. The substantial sequence variability between TSP1 domains of RSPO1-4 invites screening for additional RSPO receptor targets beyond BMPR1A.
Xenopus laevis and Xenopus tropicalis. Xenopus laevis frogs were obtained from Nasco. Xenopus tropicalis frogs were obtained from Nasco, National Xenopus Resource (NXR) and European Xenopus Resource Centre (EXRC). All X.laevis and X.tropicalis experiments were approved by the state review board of Baden-Württemberg, Germany (permit number 35-9185.81/G-141/18 (Regierungspräsidium Karlsruhe)) and executed according to federal and institutional guidelines and regulations. Developmental stages of the embryos were determined according to Nieuwkoop and Faber (Xenbase). No statistical analysis was done to adjust sample size before the experiments. No randomization of injection order was used during the experiments.
Generation of conditioned medium. HEK293T cells were seeded in 15 cm culture dishes and transiently transfected with RSPOs-AP, RSPOs-flag, BMPR1A ECD -AP, DKK1 or WNT surrogate plasmids. After 24 h, media were changed with fresh DMEM, 10% FBS, 1% L-glutamine and 1% penicillin-streptomycin and cultured 6 days at 32°C. Conditioned media were harvested three times every two days, centrifuged and validated by TOPFlash assay or western blot analyses. Mouse WNT3A conditioned medium was produced from mouse L-cells stably transfected with WNT3A (ATCC CRL-2647) following the manufacturer's instruction. For human RSPO2 ΔFU1 , RSPO2 ΔFU2 , Xenopus Rspo2 ΔFU1 , and Rspo2 F107E mutants conditioned media, HEK293T cells were seeded in 12 well culture plates and transfected with 500 μg of each plasmid, and harvested three times every two days. Production of the media was validated with western blot analyses and AP activity analyses. Fig. 8 RSPO2 bridges BMPR1A and ZNRF3 and triggers BMP receptor clearance from the cell surface. a, b In vitro binding assay between ZNRF3 and BMPR1A ECD mediated by RSPO1-3. ZNRF3-Fc protein was used as a bait, with sequential RSPO1-3 protein and BMPR1A ECD -AP treatment. Bound BMPR1A ECD to ZNRF3 was detected by chromogenic AP assay. n = 3 experimentally independent samples. c, d IF in H1581 cells transfected with BMPR1A-HA and ZNRF3-flag DNA upon RSPO2 and RSPO3-HRP treatment for 3 h. RSPOs (red) were visualized with tyramid signal amplification. BMPR1A (blue) and ZNRF3 (green) were stained against HA and flag antibody. White arrowheads, colocalized BMPR1A/RSPO2; white arrows, colocalized BMPR1A/ RSPO2-3/ZNRF3; yellow arrow, colocalized BMPR1A/RSPO2-3/ZNRF3 in magnified inset; Dashed lines, nucleus. Scale bar, 20 μm. e-h IF of colocalized BMPR1A (green)/ZNRF3 (red) in H1581 cells treated with siRNA as indicated. Nuclei were stained with Hoechst. Scale bar, 20 μm. i Quantification of BMPR1A colocalizing with ZNRF3 from e-h. j-l IF of BMPR1A (green) in H1581 cells treated with siRNA as indicated. m Quantification of cells harboring membrane localized BMPR1A from j-l. Scale bar, 20 μm. n Cell surface biotinylation assay in H1581 cells treated with BMPR1A-HA and siRNA as indicated. Co, control; R2, RSPO2; ZR, ZNRF3/RNF43 siRNA. After labeling surface proteins with Biotin, lysates were pulled down with streptavidin beads and subjected to Western blot analysis. Transferrin receptor (TfR), a loading control. TCL, Total cell lysate. Data shows results representative for 2 from 3 independent experiments. o BRE reporter assay in HEPG2 cells treated as indicated. MDC, monodansylcadaverin. n = 3 biological replicates. p IF of colocalized BMPR1A (green)/EEA1 (magenta) in H1581 cells treated with siRNA. White arrowheads, colocalized BMPR1A/EEA1 in magnified inset. q Quantification of p. Scale bar, 20 μm. r IF of colocalized BMPR1A (green)/Lamp1 (red) in H1581 cells treated with siRNA as indicated. White arrowheads, colocalized BMPR1A/Lamp1 in magnified inset. Scale bar, 20 μm. s Quantification of r. For i, m, q, s, n = the number of cells pooled from 2 independent experiments. Data for (a, b, i, m, o, q, s) are displayed as mean ± SD. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 from two-tailed unpaired t-test. Luciferase reporter assays. BRE luciferase assays were executed using 300,000 ml −1 of HEPG2 cells in 24-well plates. PGL3-BRE-Luciferase (500 ng ml −1 ) and pRL-TK-Renilla plasmids (50 ng ml −1 ) were transfected using Lipofectamine 3000. After 24 h, cells were serum starved 2 h and stimulated 14-16 h with 80 ng ml −1 recombinant human BMP4 protein (R&D systems 314-BP) along with RSPO1-4 conditioned medium. TOPFlash luciferase assays were executed in HEK293T cells using 10,000 cells per well in 96-well plates. TOPFlash-Luciferase (5 ng) and pRL-TK-Renilla plasmids (1 ng) per well were transfected using X-tremeGENE9 transfection reagent (Roche 06365787001). After 48 h, cells were stimulated 24 h with WNT3A conditioned medium along with RSPO conditioned medium. Luciferase activity was measured with the Dual luciferase reporter assay system (Promega E1960). Firefly luminescence (BRE or TOPFlash) was normalized to Renilla. Statistical analyses were made with Graphpad PRISM7 software.
Cell surface biotinylation assay. H1581 cells were seeded in 6 cm culture dishes and transfected with 50 nM of indicated siRNAs for 3 days and 2 μg of BMPR1A-HA DNA for 2 days. Surface proteins were biotinylated with 0.25 mg ml −1 sulfo-NHS-LC-LC-Biotin (ThermoFisher 21338) at 4°C for 30 min. The reaction was quenched by 10 mM Monoethanolamine and cells were harvested and lysed with Triton X-100 lysis buffer. 200-300 μg of lysate was incubated with 20 μl streptavidin agarose (ThermoFisher 20359) to pull-down biotinylated surface proteins and subjected to Western blot.
Xenopus laevis whole-mount in situ hybridization. Whole-mount in situ hybridizations of Xenopus embryos were performed using digoxigenin (DIG)-labeled probes according to the standard protocol (https://www.xenbase.org) 62 . Antisense RNA probes against rspo2 and bmp4 were generated by in vitro transcription with T7 RNA polymerase (Promega P2075) 20 . Probes against bmpr1a and znrf3 were prepared using full-size Xenopus bmpr1a ORF or znrf3 ORF as a template, linearized with Xho I (NEB R0146S) and transcribed with T7 RNA polymerase. Mo and mRNA injected embryos were collected at stage 11 (gastrula) or 32 (tadpole) for in situ hybridization. Images were obtained using AxioCam MRc 5 microscope (Zeiss). Embryos in each image were selected using Magnetic Lasso tool or Magic Wand tool of Adobe Photoshop CS6 software, and pasted into the uniform background color for presentation.
Xenopus microinjection and phenotype analysis. In vitro fertilization, microinjection and culture of Xenopus embryos were performed according to the standard protocol (https://www.xenbase.org). X.laevis embryos were microinjected with reporter DNAs, in vitro transcribed mRNAs or antisense morpholino oligonucleotide (Mo) using Harvard Apparatus microinjection system. Mos for rspo2 20 , lrp6 39 , chordin, bmp4 38 , znrf3 43 and standard control were purchased from Gen-eTools. rspo2 ΔTSP Mo was designed based on rspo2 sequence (Supplementary Table 3). X.laevis 4-cell stage embryos were microinjected 5 nl per each blastomere equatorially and cultured until indicated stages. Equal amount of total mRNA or Mo were injected by adjustment with ppl or standard control Mo. Scoring of phenotypes was executed blind from two individuals, and data are representative images from at least two independent experiments. Embryos in each image were selected using Magnetic Lasso tool or Magic Wand tool of Adobe Photoshop CS6 software, and pasted into the uniform background color for presentation.
Xenopus tropicalis CRISPR/Cas9-mediated mutagenesis. The 5′ region of genomic sequences from X.tropicalis chordin (NM_001142657.1) and noggin (NM_001171898.1) were searched for guide RNA (gRNA) targeting sites using an online prediction tool (https://crispr.cos.uni-heidelberg.de). Primers for rspo2, noggin and chordin were designed or chosen 27 for PCR-based gRNA template assembly (Supplementary Table 4) 63 . A primer lacking any target sequences was used as control gRNA. PCR reactions were performed with Phusion Hot Start Flex DNA Polymerase (NEB M0535), followed by in vitro transcription using MEGAscript T7 Transcription Kit (Invitrogen AM1334). Embryos were microinjected at one to two-cell stages with a mixture of 50 pg of gRNA and 1 ng of recombinant Cas9 protein (Toolgen) per embryo. Injected embryos were cultured until stage 30, fixed with MEMFA for phenotypical analysis. Scoring of phenotypes was executed at stage 30 with blinding from two individuals, and data are representative images from three independent experiments. Defects were categorized by the severity of ventralization. 'Severe' showed small head, enlarged ventral tissues and short body axis. 'Mild' showed one or two of the defects described above. 'Normal' showed no visible differences to the uninjected control. Xenopus tropicalis T7 Endonuclease I assay. To validate CRISPR/Cas9-mediated genome editing, three embryos of each injection set were lysed at stage 30 for genotyping PCR reactions as described 63 . For primer sequences, see Supplementary  Table 4. All target sequences were amplified with Roti-Pol Hot-TaqS Mix (Roth 9248). After denaturation for 3 min at 94°C and reannealing (ramp 0.1°C per sec), the PCR products were incubated with 3 U of T7 Endonuclease I for 45 min at 37°C . Cleavage results were visualized on a 2% agarose gel. Cycloheximide treatment on Xenopus laevis animal cap explants. Xenopus laevis animal caps were dissected at stage 8 and treated with 30 μg ml −1 of cycloheximide (CHX) (Sigma C7698) until control embryos reached stage 10. CHX treatment was validated since cell division was retarded compared to untreated control. In situ hybridization and qRT-PCR were performed with same methods used in whole embryos.
Immunofluorescence. 150,000 H1581 cells were grown on coverslips in 12-well plates, followed by siRNA and DNA transfection. After 48 h cells were fixed in 4% PFA for 10 min. Cells were treated with primary antibodies (1:250) overnight at 4°C, and secondary antibodies (1:500) and Hoechst dye (1:500) were applied for 2 h at room temperature. For Tyramide Signal Amplification to detect RSPOs, H1581 cells were treated with RSPO1-3 ΔC -HRP conditioned media 3 h at 37°C. Cells were washed with Hanks' balanced salt solution (HBSS) and fixed for 30 min with 0.5 mM dithiobis (succinimidyl) propionate (DSP) (Thermo 22585) crosslinker in HBSS supplemented with 10 mM HEPES, followed by permeabilization with 0.1% saponin in TSA buffer (100 mM Tris, pH 8.8, 10 mM imidazole). The TSA reaction was executed for 30 min in dark with the 30 μM tyramide-Rhodamine in TSA buffer supplemented with 0.003% H 2 O 2 . Cells were washed and further stained with anti-HA and anti-Flag antibodies (1:250) 24,39 . Quantification was done using Image J v1.51k software. For X. laevis embryos, bmpr1a-EYFP and membrane-RFP mRNAs were coinjected with the indicated mRNAs or Mos. Embryos were dissected for animal or ventrolateral explants at stage 9 or stage 11.5, respectively. Explants were immediately fixed with 4% PFA for 2 h and mounted with Fluoromount-G (ThermoFisher 00495802). Images were obtained using LSM 700 (Zeiss). Data are representative images from two independent experiments. For quantification, Pearson's correlation coefficient for EYFP and RFP was analyzed using 16-30 random areas harboring 10 cells chosen from 6-10 embryos per each set.
Cell surface binding assay. Two hundred and fifty nanogram of human BMPR1A-HA and Xenopus tropicalis LGR4 DNA were transfected in HEK293T cells and incubated with 1.5 U ml −1 conditioned media for 3 h on ice. After several washes with PBS and crosslinking with DSP, cells were treated with 2 mM Levamisole for 20 min to inactivate endogenous AP activities and developed with BM-Purple (Sigma 11442074001). Cells were mounted with Fluoromount G. Images were obtained using LEICA DMIL microscope/Canon DS126311 camera.
Quantitative real-time PCR. Cultured cells were lysed in Macherey-Nagel RA1 buffer containing 1% β-mercaptoethanol and total RNAs were isolated using NucleoSpin RNA isolation kit (Macherey-Nagel 740955). Reverse transcription and PCR amplification were performed as described before 64 . For Xenopus laevis, animal cap explants were harvested at stage 10 and lysed in 1 ml of TRizol (Ambion 18914101). Extraction and precipitation of RNA were performed following the manufacturer's instruction. Reverse transcription was executed with 1 μg RNA using SuperScript II reverse transcriptase and random primers (Invitrogen P/N58875). PCR amplification with the obtained cDNA was performed using UPL (Universal Probe Library, Roche, #15 (rspo2), #140 (sizzled), #145 (vent1)) probes and corresponding primers, further analyzed by LightCycle 480. Primers used in this study are listed in Supplementary Table 1. Graphs show relative gene expressions to GAPDH. Data are displayed as mean with SD from multiple experimental replicates. Statistical analyses were performed using the GraphPad PRISM7 software.