R-spondins can potentiate WNT signaling without LGR receptors

The WNT signaling pathway regulates patterning and morphogenesis during embryonic development and promotes tissue renewal and regeneration in adults. Some WNT responses in vertebrates depend on a second signal provided by the R-spondin family of four secreted proteins (RSPO1-4) that drive the renewal of stem cells in many tissues. RSPOs markedly amplify target cell sensitivity to WNT ligands by neutralizing two transmembrane E3 ligases, ZNRF3 and RNF43, which reduce cell-surface levels of WNT receptors. Chromosomal translocations that increase RSPO expression or that inactivate ZNRF3/RNF43 can drive human cancers. RSPOs contain tandem furin-like repeats (FU1 and FU2), a thrombospondin type I (TSP) domain, and a basic region (BR). RSPOs simultaneously engage ZNRF3/RNF43 through their FU1 domain and one of three leucine-rich repeat-containing G-protein coupled receptors (LGR4-6) through their FU2 domain, triggering the clearance of ZNRF3/RNF43 and the consequent rise in WNT receptor levels. LGRs are selectively expressed in various tissue stem cells and are considered the primary high-affinity receptors for RSPOs. Using purified mutant and chimeric RSPOs and cell lines lacking various receptors, we show that RSPO2 and RSPO3, but not RSPO1 and RSPO4, can potentiate WNT/β-catenin signaling in the absence of all three LGRs. The ZNRF3/RNF43-interacting FU1 domain was necessary for LGR-independent signaling, while the LGR-interacting FU2 domain was dispensable. The FU1 domain of RSPO3 was also sufficient to confer LGR-independence when transplanted to RSPO1, demonstrating that its interaction with ZNRF3/RNF43 dictates LGR-independent signaling. The enigmatic TSP/BR domains of RSPOs and their interaction with heparan sulfate proteoglycans (HSPGs), previously considered dispensable for WNT/β-catenin signaling, became essential in the absence of LGRs. These results define two alternative modes of RSPO-mediated signaling that share a common dependence on ZNRF3/RNF43, but differ in their use of either LGRs or HSPGs, with implications for understanding their mechanism of action, biological functions and evolutionary origins.

LGRs. The ZNRF3/RNF43-interacting FU1 domain was necessary for LGR-independent 23 signaling, while the LGR-interacting FU2 domain was dispensable. The FU1 domain of RSPO3 24 was also sufficient to confer LGR-independence when transplanted to RSPO1, demonstrating 25 that its interaction with ZNRF3/RNF43 dictates LGR-independent signaling. The enigmatic 26 TSP/BR domains of RSPOs and their interaction with heparan sulfate proteoglycans (HSPGs), 27 previously considered dispensable for WNT/b-catenin signaling 16,17 , became essential in the 28 absence of LGRs. These results define two alternative modes of RSPO-mediated signaling that 29 3 share a common dependence on ZNRF3/RNF43, but differ in their use of either LGRs or 30 HSPGs, with implications for understanding their mechanism of action, biological functions and 31 evolutionary origins. 32 In previous work 18 , we generated and thoroughly characterized a haploid human cell line 33 (HAP1-7TGP) that harbors a fluorescent transcriptional reporter for WNT/b-catenin signaling. 34 Both the fluorescence of this synthetic reporter and the transcription of endogenous WNT target 35 genes in HAP1-7TGP cells can be activated by WNT ligands, and these WNT responses can be 36 strongly potentiated by RSPOs 18 . HAP1-7TGP cells do not secrete WNT ligands and thus their 37 response to RSPOs requires the co-administration of a low concentration of WNT. A 38 comprehensive set of unbiased genetic screens in HAP1-7TGP identified most of the known 39 components required for a signaling response to RSPO and WNT ligands, establishing this cell 40 line as a valid and genetically tractable system for the study of this pathway 18 . 41 We made the serendipitous and unexpected observation that RSPO3 could potently 42 enhance WNT reporter fluorescence driven by a low concentration of WNT3A in two 43 independently derived HAP1-7TGP clonal cell lines carrying loss of function mutations in LGR4 44 (LGR4 KO cells; see Methods and Supplementary Data File 1) (Fig. 1a). In contrast, RSPO1 was 45 inactive in LGR4 KO cells. RSPO1 and RSPO3 had equivalent activity in wild-type (WT) HAP1-46 7TGP cells, demonstrating that both ligands were active, and responses in both WT and LGR4 KO 47 7 RSPO1 to potentiate WNT signaling in LGR4/5/6 KO cells (Fig. 3b, d). Conversely, replacing the 122 FU1 domain of RSPO3 with that of RSPO1 drastically reduced the signaling capacity of RSPO3 123 in LGR4/5/6 KO cells (Fig. 3b,d). In important control experiments, all chimeric ligands showed 124 equivalent activity in WT cells, establishing ligand integrity (Fig. 3b, c). Thus, a difference in the 125 interaction between ZNRF3/RNF43 and the FU1 domains of RSPO1 and RSPO3 is the crucial 126 determinant of LGR-independent signaling. Of note, the affinities of the FU1-FU2 fragment of 127 RSPO2 (25 nM) and RSPO3 (60 nM) for ZNRF3 have been reported to be much higher than 128 those of RSPO1 (6.8 µM) and RSPO4 (300 µM) 12 . Indeed, these affinities correlate with the 129 capacity of RSPO2 and RSPO3, but not RSPO1 or RSPO4, to promote LGR-independent 130 signaling ( Fig. 1b). While the TSP/BR domains of RSPO3 were required for LGR-independent 131 signaling, they were not sufficient because replacement of the TSP/BR domains of RSPO1 with 132 those of RSPO3 did not confer the capacity to signal in LGR4/5/6 KO cells (Fig. 3e). In fact, the 133 TSP/BR domains of RSPO1 and RSPO3 seemed interchangeable for signaling activity in both 134 WT and LGR4/5/6 KO cells (Fig. 3e). 135 These results suggested that the WNT-potentiating activity of RSPO3 in the absence of 136 the LGRs depends on its interaction with ZNRF3/RNF43 through the FU1 domain and an 137 additional interaction with an alternative co-receptor through the TSP/BR domains. We 138 considered the previous observation that the TSP/BR domains of RSPOs can bind to heparin 21 . 139 Addition of heparin to the culture medium completely blocked signaling by RSPO3 in 140 LGR4/5/6 KO cells, but had only a partial inhibitory effect on WT cells, in which RSPO3 can also 141 signal through LGRs (Fig. 4a). 142 The TSP/BR domains of RSPOs can mediate interactions with the two major families of 143 cell-surface heparan sulfate proteoglycans (HSPGs), the transmembrane syndecans and the 144 8 glycophosphatidylinositol (GPI)-linked glypicans 17 . In humans, both protein families are 145 encoded by multiple, partially redundant genes: four syndecan genes (SDC1-4) and six glypican 146 genes (GPC1-6) 22 , all of which are expressed in HAP1 cells (Extended Data Table 1), making 147 their genetic analysis challenging. Since all syndecans and glypicans must be post-translationally 148 modified with heparan sulfate chains for receptor function, we disrupted EXTL3, a gene 149 encoding a glycosyltransferase that is specifically required for HSPG biosynthesis, but 150 dispensable for the synthesis of other glycosaminoglycans and proteoglycans 23 . The loss of 151 EXTL3 in LGR4/5/6 KO cells led to an 81% reduction in RSPO3-mediated potentiation of WNT 152 signaling (Fig. 4b). In contrast, the loss of EXTL3 in WT cells only reduced signaling by 34%, 153 likely because RSPO3 can also signal through LGR receptors in WT cells. In an important 154 control, the loss of EXTL3 did not affect signaling induced by addition of WNT3A alone or by 155 inhibition of the b-catenin destruction complex kinase GSK3 in either WT or LGR4/5/6 KO cells 156 (Fig. 4c). 157 To distinguish between syndecans and glypicans, we took advantage of the fact that only 158 glypicans are anchored to the cell surface by a GPI linkage. Disrupting PIGL, a gene required for 159 GPI-anchor biosynthesis, or disrupting both GPC4 and GPC6 (the two glypican genes identified 160 in our previous haploid genetic screens 18 ) in LGR4/5/6 KO cells did not impair LGR-independent 161 potentiation of WNT signaling by RSPO3 (Fig. 4d). 162 These results suggest that the interaction of the TSP/BR domains of RSPO3 with cell 163 surface HSPGs, possibly syndecans, provides an alternative mechanism that neutralizes 164 ZNRF3/RNF43 in the absence of LGR receptors (Fig. 4e). Cell-surface HSPGs are known to 165 mediate the efficient endocytosis of multiple cargoes 24 . Hence, we speculate that the 166 simultaneous interaction of RSPO3 with ZNRF3/RNF43 through its FU1 domain and cell 9 surface HSPGs through its TSP/BR domains provides an LGR-independent route for the 168 endocytosis and clearance of ZNRF3/RNF43 from the cell surface (Fig. 4e), and the consequent 169 rise in WNT receptor levels. 170 Our work shows that RSPOs can potentiate WNT signals in the absence of LGR 171 receptors, expression of which has been hitherto considered the hallmark of RSPO-responsive 172 cells. Future work will define the developmental, regenerative, and oncogenic contexts in which 173 this LGR-independent mode of signaling is used to amplify target cell responses to WNT 174 ligands. The mutant and chimeric RSPO ligands we described should allow the selective 175 modulation of these alternate modes of signaling to dissect their biological roles. 176 10

Methods 177
The following materials and methods relevant to this manuscript have been described 178 previously 18 : cell lines and growth conditions, preparation of WNT3A conditioned medium 179 (CM) and construction of the HAP1-7TGP WNT reporter haploid cell line. File 2) were generated synthetically as gBlocks Gene Fragments (IDT), flanked at the 5' and 3' 219 ends, respectively, by 24 bp overhangs overlapping the sequence upstream of the unique AgeI 220 site and downstream of the unique KpnI site in the pHLsec-HA-Tev-Fc-Avi-1D4 vector. The 12 gBlocks were subcloned into pHLsec-HA-Tev-Fc-Avi-1D4, digested with AgeI and KpnI, using 222 the NEBuilder HiFi DNA Assembly Master Mix (NEB Cat. # E2621L). 223 To remove the dimerizing Fc tag from RSPO3 DTSP/BR, a fragment lacking the TSP and 224 BR domains of RSPO3 was amplified by PCR using forward primer pHL-SEC-RSPO3-F-gibson 225 respectively. The PCR product and pHLsec-HA-Avi-1D4 vector were both digested with AgeI 230 and KpnI, and ligated to produce pHLsec-HA-RSPO3DTSP/BR-Avi-1D4. 231 All constructs were sequenced fully and will be deposited in Addgene. 232 233

Analysis of WNT reporter fluorescence 234
To measure WNT reporter activity in HAP1-7TGP cells or derivatives thereof, ~24 hrs 235 before treatment cells were seeded in 96-well plates at a density of 1.5x10 4 per well and grown in 236 (BD Biosciences) using a 488 laser and 505LP, 530/30BP filters, or on a BD Accuri RUO 244

Special Order System (BD Biosciences). 245
For the experiments shown in Figs. 1c, 2b, 2e, 3b, 3e and 4b-d, cells were treated in 246 duplicate or triplicate wells, fluorescence data for 5,000-10,000 singlet-gated cells was collected, 247 and the median EGFP fluorescence for each well and/or the average +/-standard deviation ( CNBr-activated sepharose 4B was dissolved in 50 ml of 1 mM HCl and allowed to swell. The 303 resin was transferred to an Econo-Pac chromatography column (Biorad Cat. # 7321010) and 304 washed by gravity flow with 50 ml of 1 mM HCl, followed by 50 ml of 0.1 M NaHCO 3 , 0.5 M 305 NaCl, pH 8.5. 14 mg of Rho 1D4 antibody were dissolved in 0.1 M NaHCO 3 , 0.5 M NaCl, pH 306 8.5, and incubated with the resin overnight, rotating at 4°C. The resin was washed with 50 ml of 307 0.2 M glycine, pH 8.0, and incubated for 2 hrs in the same buffer, rotating at RT. The resin was 308 washed sequentially with 50 ml each of: 0.1 M NaHCO 3 , 0.5 M NaCl, pH 8.5; 0.1 M NaOAc, 309 0.5 M NaCl, pH 4.0; 0.1 M NaHCO 3 , 0.5 M NaCl, pH 8.5; PBS, 10 mM NaN 3 . The packed resin 310 16 was resuspended in an equal volume of PBS, 10 mM NaN 3 to make a ~50% slurry, aliquoted and 311 stored at 4°C. 312 300 µl of the ~50% slurry of Rho 1D4 resin was added to a 50 ml conical tube containing 313 the CM, and the suspension was incubated 10 hrs rocking at 4°C. Following binding and during 314 all subsequent washes, the resin was collected by centrifugation for 5 min at 400 x g in a 315 swinging bucket rotor. The beads were wash three times at RT with 25 ml PBS by resuspending 316 the beads in buffer and mixing by inverting for ~1 min. Following the third wash the resin was 317 transferred to a 1.5 ml Eppendorf tube and washed three times with 1.4 ml of PBS, 10% glycerol. 318 Following the last wash, the buffer was aspirated and the resin was resuspended in 150 µl 319 of PBS, 10% glycerol to obtain a ~50% slurry. Tagged RSPO protein was eluted by adding 3 µl 320

of a 25 mM stock of 1D4 peptide ((NH 3 )-T-E-T-S-Q-V-A-P-A-(COOH)) for a final 321
concentration of 250 µM. Elution was carried out by rotating the tube sideways overnight at 4°C. 322 Following centrifugation of the resin, the eluate was recovered and reserved on ice. The resin 323 was resuspended in 150 µl of PBS, 10% glycerol, and 250 µM 1D4 peptide was added. A second 324 round of elution was carried out for 1 hr at RT. Following centrifugation of the resin, the second 325 eluate was recovered and pooled with the first. The final eluate was centrifuged once again to 326 remove residual resin, and the supernatant was aliquoted, frozen in liquid nitrogen and stored at -

Preparation of figures and statistical analysis 359
Illustrations were prepared using PowerPoint (Microsoft) and Illustrator CS6 (Adobe). 360 Tables and supplementary files were prepared using Excel and Word (Microsoft). Bar graphs, 361 dose-response graphs and circle graphs were prepared using Prism 7 (GraphPad Software) and 362 statistical analysis was performed using the same software. For comparisons between two 363 datasets, significance was determined by unpaired t test; for comparisons between more than two 364 datasets, significance was determined by one-way ANOVA. Significance is indicated as **** (p 365 < 0.0001), ** (p < 0.01), * (p < 0.05) or ns (not significant). Pictures of gels and immunoblots 366 were only adjusted for contrast and brightness when necessary for clarity using Photoshop CS6 367 (Adobe), and were arranged in Illustrator CS6. 368 369

Data availability 370
All data generated or analyzed during this study are included in this published article (and 371 its supplementary information files).    LGR4/5/6 KO 29  LGR-independent signaling by RSPO3 requires heparan sulfate proteoglycans. 513 a. WNT reporter induction (calculated from the median WNT reporter fluorescence from 5,000 514 cells) in WT HAP1-7TGP and LGR4/5/6 KO cells stimulated with 1.43% WNT3A CM, 2 nM 515 untagged RSPO3 and the indicated concentrations of heparin. The fold-induction over 1.43% 516 WNT3A CM alone in the absence of heparin was normalized to 100%. 517 b. WNT reporter induction (calculated from the average WNT reporter fluorescence of triplicate 518 wells) in the indicated cell lines following treatment with 2.78% WNT3A CM and 20 ng/ml 519 untagged RSPO3. The fold-induction was normalized to the average fold-induction for WT (left 520 two genotypes) or for LGR4/5/6 KO (right two genotypes) cells. Each circle represents a unique 521 clonal cell line (determined by genotyping, Supplementary Data File 1) and the average of data 522 from two independent clonal cell lines for each genotype is indicated by a horizontal line. 523 Significance was determined as described in Methods. 524 c. WNT reporter fluorescence (average +/-SD from triplicate wells) for the same clonal cell lines 525 depicted in b. Where indicated, cells were treated with a sub-saturating concentration (11.1%) of 526 WNT3A CM or with 10 µM of the GSK3 inhibitor CHIR-99021. 527 d. WNT reporter fluorescence (average +/-SD from duplicate wells) following treatment with a 528 low concentration of WNT3A CM (2.78% for LGR4/5/6 KO cells, or 11.1% for LGR4/5/6 KO ; 529 PIGL KO and LGR4/5/6 KO ; GPC4/6 KO cells) alone or in combination with 20 ng/ml RSPO3, or 530 with 10 µM of the GSK3 inhibitor CHIR-99021. Since depletion of PIGL or of GPC4 and GPC6 531 reduces signaling at low WNT concentrations 18 , different WNT3A CM concentrations were used 532 to achieve comparable signaling responses to WNT3A alone in all cell lines so that potentiation 533 by the further addition of RSPO3 could be directly compared. Each circle represents a unique 534 32 clonal cell line, and the average of data from two independent clonal cell lines for each genotype 535 is indicated by a horizontal line. 536 e. Proposed models for LGR-dependent and LGR-independent signaling by RSPOs. See text for 537 details.