Wnt family member 4 (WNT4) and WNT3A activate cell-autonomous Wnt signaling independent of porcupine O-acyltransferase or Wnt secretion

Porcupine O-acyltransferase (PORCN) is considered essential for Wnt secretion and signaling. However, we observed that PORCN inhibition does not phenocopy the effects of WNT4 knockdown in WNT4-dependent breast cancer cells. This suggests a unique relationship between PORCN and WNT4 signaling. To examine the role of PORCN in WNT4 signaling, here we overexpressed WNT4 or WNT3A in breast cancer, ovarian cancer, and fibrosarcoma cell lines. Conditioned media from these lines and co-culture systems were used to assess the dependence of Wnt secretion and activity on the critical Wnt secretion proteins PORCN and Wnt ligand secretion (WLS) mediator. We observed that WLS is universally required for Wnt secretion and paracrine signaling. In contrast, the dependence of WNT3A secretion and activity on PORCN varied across the cell lines, and WNT4 secretion was PORCN-independent in all models. Surprisingly, WNT4 did not exhibit paracrine activity in any tested context. Absent the expected paracrine activity of secreted WNT4, we identified cell-autonomous Wnt signaling activation by WNT4 and WNT3A, independent of PORCN or Wnt secretion. The PORCN-independent, cell-autonomous Wnt signaling demonstrated here may be critical in WNT4-driven cellular contexts or in those that are considered to have dysfunctional Wnt signaling.

Wnt 2 signaling is an ancestrally-conserved pathway that plays fundamental roles in embryonic development and adult tissue homeostasis. Dysregulation of Wnt signaling is a causative factor for a range of human pathologies, including several forms of cancer (1). As a result, inhibition of Wnt signaling has become an attractive therapeutic target in ongoing clinical trials, with some strategies targeting the upstream activation of signaling by Wnt proteins (1)(2)(3). Wnt proteins compose a family of secreted glycoproteins that act as intercellular ligands, which stimulate a myriad of signal transduction cascades regulating cellular proliferation, stem cell renewal, cell motility, angiogenesis, and apoptosis (1,(4)(5)(6). Wnt proteins are posttranslationally modified by the O-acyltransferase Porcupine (PORCN), which palmitoylates Wnt proteins at single serine residues (2,(7)(8)(9). This lipidation forms a binding motif for interaction with Wntless (WLS), which chaperones Wnt proteins to the plasma membrane for secretion (8,10,11). Once secreted, Wnt proteins signal in a paracrine manner, binding nearby receptor complexes. Wnts typically bind a Frizzled protein (FZD) in conjunction with the LRP5 or LRP6 co-receptor, resulting in activation of the Disheveled second messenger proteins (DVL1/2/3 in humans) and initiation of Wnt signaling, through either canonical (␤-catenindependent) or noncanonical (␤-catenin-independent) pathways (1,4). The essential initiating step in Wnt processing is palmitoylation by PORCN, which has prompted the development of PORCN inhibitors, including IWP compounds (11), WNT974 (also known as LGK974) (3), and others (2,12). PORCN inhibitors have been shown to block Wnt secretion, inhibit downstream Wnt signaling, and suppress Wnt-driven tumor growth in animal models (3,13,14), with WNT974 currently in phase I/II clinical trials for cancer treatment (NCT01351103 and NCT02278133). Based on these observations, PORCN inhibitors are an attractive strategy to target Wnt-driven pathologies.
The Wnt protein WNT4 is critical in organogenesis of endocrine organs and regulation of bone mass, and it underlies steroid hormone-related phenotypes in humans (15)(16)(17)(18)(19)(20)(21)(22). WNT4 dysregulation via loss-of-function mutation results in developmental female to male sex reversal (23)(24)(25)(26). Similarly, WNT4 polymorphisms are associated with endocrine dysfunction, gynecological malignancies, reduced bone density with prema-ture skeletal aging, and related phenotypes (27)(28)(29)(30)(31)(32)(33). WNT4 is also critical in mammary gland development, as Wnt4 knockout in mouse mammary gland prevents progesterone-driven ductal elongation and branching during pregnancy (34,35). In this context, activated progesterone receptor drives expression of Wnt4 in mammary gland luminal cells resulting in paracrine signaling that supports maintenance of the mammary stem cell niche (6,(36)(37)(38). Despite these observed critical roles of WNT4 in both normal and malignant tissues, WNT4 signaling is crudely understood due to varied context-dependent functions. In a cell type-and tissue-specific manner, WNT4 (human or murine) has been shown to regulate either ␤-catenindependent or -independent Wnt signaling and has been shown to either activate or suppress signaling (described in references herein). Furthermore, conflicting reports exist whether murine Wnt4 can or cannot activate Wnt signaling via ␤-catenin in the murine mammary gland (36,39). As such, WNT4 has been described as a "problem child" among Wnt proteins. It is also unclear which FZD complexes are utilized as receptors by WNT4, as WNT4 is often required for distinct, nonredundant functions versus other Wnt proteins (35). Because WNT4 has myriad downstream signaling effects, inhibition of WNT4 upstream of Wnt effector pathways (e.g. with PORCN inhibitors) is an attractive approach to block WNT4 signaling in a "pathway-indifferent" manner to treat WNT4related pathologies.
We reported that regulation of WNT4 expression is co-opted by the estrogen receptor in a subtype of breast cancer, invasive lobular carcinoma (ILC) (40,41). Estrogen-driven WNT4 is required in ILC cells for estrogen-induced proliferation and survival, as well as anti-estrogen resistance (41). Although WNT4-driven signaling in ILC has yet to be fully elucidated, ILC cells lack the capacity to engage canonical Wnt signaling, as the characteristic genetic loss of E-cadherin in ILC leads to loss of ␤-catenin protein (41,42). This suggests WNT4 drives ␤-catenin-independent Wnt signaling in ILC cells. Although the specific pathways activated by WNT4 are unknown, PORCN inhibition should be an effective strategy to block WNT4 upstream and treat ILC. However, treatment of ILC cells with PORCN inhibitors did not suppress growth or survival. These unexpected results initiated further studies into the mechanisms underlying WNT4 secretion and signaling. In this report, we show WNT4 secretion is mediated by atypical mechanisms. Our observations challenge the paradigm that PORCN-mediated secretion is required for Wnt signaling and suggest a novel process by which Wnt proteins, including WNT4, can initiate ␤-catenin-independent Wnt signaling.

PORCN inhibition does not mimic WNT4 siRNA in lobular carcinoma cells
We hypothesized that because ILC cells are dependent on WNT4 for proliferation and survival (41), inhibition of PORCN would phenocopy WNT4 siRNA by blocking WNT4 secretion and downstream signaling. Proliferation and cell death were monitored by live-cell imaging of MM134 (ILC) cells either transfected with siRNA targeting PORCN (siPORCN) or treated with PORCN inhibitor (PORCNi) LGK974. Proliferation and cell death were compared with untreated cells and cells treated with the anti-estrogen fulvestrant (Fulv) or transfected with siRNA-targeting WNT4 (siWNT4). As we reported previously (41), siRNA-mediated WNT4 knockdown or Fulv halts proliferation, and WNT4 knockdown induces cell death (Fig. 1A). However, neither genetic nor chemical PORCN inhibition had any effect on cell proliferation or survival of MM134 cells (Fig. 1, A and B). Similar results were obtained in the ILC cell line SUM44PE, as PORCN inhibitor at concentrations up to 1 M did not affect proliferation (Fig. S1). These data suggest PORCN inhibition is not sufficient to inhibit WNT4 function, and WNT4 signaling likely occurs via PORCN-independent mechanisms.

PORCN-independent Wnt signaling WNT4 secretion is WLS-dependent but PORCN-independent
Because targeting PORCN did not phenocopy WNT4 knockdown, we further examined the role of PORCN in WNT4 secretion. To facilitate Wnt secretion studies, we overexpressed WNT3A or WNT4 in MM134 (MM134:W3 and MM134:W4; Fig. 2A and Table 1), and we measured secreted Wnt proteins in conditioned medium. Of note, because epitope tags may alter Wnt secretion and activity (e.g. Ref. 8), we performed all studies with nontagged Wnt constructs. A general workflow for experiments assessing Wnt secretion and function, for each figure, is shown in Fig. S2.
PORCN-mediated palmitoylation of Wnt proteins is commonly described as required for Wnt binding to WLS and transport to the cell surface for secretion (see Introduction), so we examined the requisite of PORCN (using PORCNi and siPORCN) or WLS (using siWLS) for Wnt secretion. Secreted WNT3A and WNT4 were detected in conditioned medium from MM134:W3 and MM134:W4, respectively (Fig. 2B). Consistent with the lack of effect of cell proliferation, PORCNi treatment had no effect on WNT4 secretion, and WNT3A secretion was also unaffected by PORCNi (Fig. 2B, top). Similarly, siPORCN had no effect on secretion of either WNT4 or WNT3A (Fig. 2B, bottom; efficacy of PORCNi and siPORCN were confirmed below). However, WLS was required for Wnt secretion, as siWLS suppressed secretion of both WNT3A and WNT4 from MM134 (Fig. 2B, bottom). qPCR confirmation of PORCN and WLS knockdown is shown in Fig. S3A. These data suggest that Wnt processing and secretion may be atypical in ILC cells, but the PORCN-independent secretion of WNT4 is a potential mechanism of PORCNi resistance (Fig. 1).
Endogenously-expressed WNT4 was also detected in conditioned medium from HT1080 and HT1080-PKO cell lines (Fig.  2D). We observed that secreted WNT4 can be resolved by electrophoresis as a doublet. The larger species was PORCN-dependent and not detected in HT1080-PKO, suggesting that these species represent palmitoylated versus nonpalmitoylated proteins. Notably, in HT1080:W3 endogenously-secreted WNT4 shifted to the larger species (Fig. 2D, column 2), potentially due to positive feedback activation of PORCN activity (44). We observed loss of both species in conditioned media after WNT4 knockdown by siRNA, confirming both secreted species as WNT4 (Fig. S3B). These data indicate that although WNT4 is modified by PORCN, PORCN is not required for WNT4 secretion. WLS knockdown suppressed secretion of both WNT3A and WNT4 from HT1080 (Fig. 2E), confirming that WNT3A secretion depends on both WLS and PORCN, whereas WNT4 secretion is WLS-dependent but PORCN-independent.
Although WNT4 secretion was PORCN-independent in both MM134 and HT1080, WNT4 appeared to be post-translationally modified during secretion (Fig. 2D). Also, in both MM134 and HT1080, secreted WNT4 migrated as a highermolecular-weight species than WNT4 from cell lysate (Fig.  S3C). The increased molecular weight is due, at least in part, to glycosylation (45), as treatment with tunicamycin (N-linked glycosylation inhibitor) decreased the apparent molecular weight of secreted WNT4 in either cell line (Fig. S3C). Notably, MM134 secreted two WNT4 species when treated with tunicamycin, suggesting WNT4 modification can be variable and cell context-specific. To determine whether the PORCN-independent secretion of WNT4 was also independent of palmitoylation or other acylation, we performed Triton X-114 phase separation of conditioned medium from MM134:W4 (see under "Experimental procedures" and Fig. S3D) (46). WNT4 was detected only in the aqueous, hydrophilic phase, suggesting that secreted WNT4 is not modified by palmitoylation or other acylation.
Together, these data demonstrate that WNT4 secretion is PORCN-independent and that Wnt protein processing and signaling are more broadly atypical in ILC. PORCN-independent WNT4 secretion may explain the disparate effects of siWNT4 versus PORCN inhibition. However, it is unclear whether Wnt proteins secreted independently of PORCN are competent to activate paracrine Wnt signaling.

WNT4 overexpression does not activate paracrine Wnt signaling
Because palmitoylation of Wnt proteins is typically necessary for activation of downstream signaling, we assessed whether WNT4 secreted independent of PORCN could activate canonical or noncanonical Wnt signaling pathways in a paracrine manner. Paracrine signaling was tested by treating HT1080-PKO cells ("receiver" cells, PKO reduces endogenous paracrine Wnt signaling; see Fig. S2) with conditioned medium from MM134 cells, with or without Wnt overexpression. After 24 h of treatment with conditioned medium, we assessed phosphorylation of DVL2, DVL3, and LRP6 (induced in both canonical and noncanonical Wnt signaling), and expression of ␤-catenin target gene AXIN2 (specific marker of canonical, ␤-catenindependent Wnt signaling). Conditioned medium from MM134: W4 or MM134:W3 was unable to activate Wnt signaling in receiver cells by these measures (i.e. phospho-DVL/LRP6 or AXIN2 induction; Fig. 3, A and B) despite the presence of Wnt protein in conditioned medium from these cells (Fig. 2B)   LGK974, and medium was allowed to condition for 7 days. Bottom, MM134 was transfected with siPORCN or siWLS, and after 24 h, medium was changed and conditioned for 7 days. Total protein was extracted from medium as above for the immunoblot. C, HT1080 constitutive Wnt overexpression cell lines were created as detailed under "Experimental procedures." Left, qPCR for WNT3A and WNT4. Points represent three biological replicates. Right, immunoblot for cellular expression of WNT3A and WNT4. Overexpression shows endogenous WNT4 expression. D, HT1080 medium was conditioned for 5 days as described under "Experimental procedures," prior to total protein extraction for immunoblot. E, HT1080 cells were transfected with siPORCN or siWLS, and after 24 h, medium was changed and conditioned for 5 days. Total protein was extracted from medium as above for the immunoblot. siRNA experiments are representative of three independent experiments.  2D). Because the Wnt-signaling phosphorylation events that we assayed may peak and return to baseline well within 24 h, we also examined DVL/LRP6 phosphorylation after 2 h of treatment with conditioned medium. These results corroborated our 24-h data; neither WNT3A nor WNT4 from MM134 models activated DVL/LRP6, and WNT4 from neither HT1080 nor HT1080-PKO activated DVL/LRP6 (Fig. S4A). We also examined whether paracrine WNT4 specifically activated GPCRmediated Wnt signaling (e.g. JNK or PKC phosphorylation, independent of DVL activation) (47). However, these Wnt signaling pathways were not activated by WNT4 (or WNT3A) in HT1080-PKO (Fig. S4B). Taken together, secreted WNT4 from MM134, HT1080, or HT1080-PKO was unable to activate paracrine Wnt signaling. Paracrine signaling by Wnt proteins can be mediated by Wnt secretion as well as by Wnt presentation on the cell surface and subsequent cell-cell contacts (48,49). To determine whether paracrine WNT4 signaling is initiated specifically by cell-cell contact, we assessed whether Wnt signaling could be activated in "receiver" cells via co-culture with Wnt-overexpressing cells. HT1080-PKO cells transfected with the TOP-FLASH reporter ("receiver" cells) were co-cultured with cells expressing WNT3A or WNT4 (Fig. 3E). WNT3A-expressing cells activated TOP-FLASH in co-cultured cells in a PORCN-dependent manner (i.e. blocked by LGK974 and absent from HT1080-PKO:W3 cells). However, under no condition was WNT4 able to activate TOP-FLASH in co-cultured cells. Although WNT3A is able to mediate paracrine Wnt signaling in both secreted and co-culture models, neither secreted nor cell surface WNT4 mediates paracrine signaling.

PORCN-independent
Because WNT4 from neither HT1080 nor MM134 cells was able to activate paracrine Wnt signaling, we treated HT1080-PKO cells with recombinant human Wnt proteins (rWNT3A and rWNT4; 10 -500 ng/ml). Of note, because we did not observe activation of GPCR-mediated Wnt signaling in this model (Fig. S4B), as above we utilized DVL/LRP6 phosphorylation and AXIN2 induction as indicators of activated Wnt signaling. Wnt signaling was activated by all measures by rWNT3A, whereas rWNT4 failed to activate Wnt signaling at any concentration (Fig. 3, F and G). These data suggest that HT1080 may be nonresponsive to paracrine WNT4, raising the need for orthogonal systems to validate the function of both secreted and recombinant WNT4. To address this, we used MC3T3-E1 as an additional "receiver" cell line (Fig. S2). MC3T3-E1 cells are another bone-like model highly responsive to exogenous Wnt protein and induce alkaline phosphatase production in response to paracrine Wnt ligand activity (measured by colorimetric assay, see "Experimental procedures" (50)). Conditioned medium as above was used to treat "receiver" 3T3-E1 cells, and alkaline phosphatase (AP) activity was used as the readout for activation of paracrine Wnt signaling. Both rWNT3A and rWNT4 increased AP activity (Fig. 4A), and both rWNT3A and rWNT4 induced DVL and LRP6 phosphorylation in 3T3-E1 cells (Fig. S5A). This confirmed that 3T3-E1 cells responded to paracrine WNT4 and could be used to assess the activity of secreted Wnts in conditioned medium. Conditioned medium from HT1080:W3, but not HT1080-PKO:W3, induced AP activity in 3T3-E1 (Fig. 4B), consistent with a requirement of PORCN for paracrine WNT3A signaling. However, although rWNT4 induced AP activity, conditioned medium from neither HT1080:W4 nor HT1080-PKO:W4 induced AP activity in 3T3-E1 (Fig. 4B). Parallel results were obtained using MM134, as conditioned medium from MM134:W3 induced AP activity in 3T3-E1, which was blocked by PORCNi treatment, but no AP activity was induced with MM134:W4 conditioned medium (Fig. 4C). We considered that the concentration of secreted WNT4 protein may be insufficient to activate 3T3-E1, but secreted WNT3A or WNT4 is within the range of active rWNT concentrations in this assay (Fig. S5, B-D). Notably, commercially available rWNT3A and rWNT4 are produced lacking the N-terminal signal peptide, and both rWNT3A and rWNT4 migrated as a smaller peptide than the corresponding Wnt protein secreted from either HT1080 or MM134 (Fig. S5, B and C). Taken together, these results indicate secreted Wnt proteins are not equivalent to recombinant proteins and have distinct capacities for activating paracrine Wnt signaling. However, although secreted WNT3A activated paracrine Wnt signaling in a context-dependent manner (based on both source and receiver cells), secreted WNT4

PORCN-independent Wnt signaling
was unable to act as a functional paracrine-signaling ligand in any tested context.

PORCN-independent WNT4 secretion and lack of paracrine activity are observed in varied model systems
We further examined Wnt secretion and function via WNT3A or WNT4 overexpression in a second ILC cell line (SUM44PE), an additional breast cancer cell line (HCC1428), and an ovarian cancer cell line (PEO1) ( Table 1). Like bone, ovarian cancer is a relevant context as WNT4 mediates Müllerian tissue and ovary development (23,25). Wnt expression and secretion were assessed as above, and paracrine activity of secreted Wnt proteins was tested using the 3T3-E1 receiver model.
Secreted WNT3A and WNT4 were detected after overexpression in SUM44PE (ILC, Fig. 5A), and WNT4 secretion in SUM44PE was PORCN-independent but WLS-dependent (Fig.  5B). Neither WNT3A nor WNT4 induced AP activity versus parental SUM44PE cells (Fig. 5C). These data with SUM44PE are consistent with our observations of PORCN-independent Wnt secretion and inactive paracrine activity of secreted Wnt proteins in ILC. Secreted WNT4 also lacked any paracrine activity from both HCC1428 (Fig. 5, D and E) and PEO1 (Fig. 5, F and G). In total, we were unable to detect paracrine WNT4 activity in five cell lines derived from three different WNT4responsive tissues of origin (mammary gland, bone, and ovary), further supporting that secreted WNT4 does not mediate paracrine signaling in WNT4-expressing cells.

Secreted or paracrine WNT4 is not required for ILC cell proliferation and viability
Because secreted WNT4 did not activate paracrine Wnt signaling in MM134 or SUM44PE cells, we hypothesized that WNT4 secretion is dispensable for WNT4 function in ILC cells. To confirm this, we examined whether blocking WNT4 secretion by WLS knockdown would phenocopy WNT4 knockdown in MM134 cells. MM134 cells were transfected with siRNAs targeting WNT4, PORCN, or WLS, followed by cell proliferation and death assays as above. Despite the loss of WNT4 secretion, siWLS had no detrimental effect on MM134 proliferation or viability and did not phenocopy siWNT4 (Fig. S6, A and B). Similarly, we attempted to rescue WNT4 knockdown with conditioned medium from MM134 (with or without Wnt overexpression) or rWNT protein. Exogenous Wnt protein had no effect on siWNT4-mediated growth inhibition or cell death (Fig. S6, C and D). These data support that secretion may not mediate the activation of WNT4-driven signaling.

WNT4 and WNT3A activate cell-autonomous signaling
Our observations suggest WNT4 may activate signaling not via paracrine or autocrine mechanisms (i.e. mediated by Wnt secretion or trafficking to the cell membrane), but rather by an intracellular cell-autonomous mechanism. To examine cell-autonomous Wnt-induced signaling, we assessed DVL and LRP6 phosphorylation and AXIN2 expression directly in HT1080 and HT1080-PKO Wnt-overexpressing cells ( Fig. S2 and Fig. 6, A  and B). Consistent with conditioned medium/receiver cell experiments, both HT1080:W3 and HT1080:W4 had increased DVL and LRP6 phosphorylation (Fig. 6A), but only HT1080: W3 had increased AXIN2 expression (Fig. 6B). In HT1080-PKO, overexpression of either WNT3A or WNT4 induced cellautonomous DVL2 and DVL3 phosphorylation (Fig. 6A). DVL activation was independent of paracrine signaling, as Wnt overexpression in HT1080-PKO did not activate LRP6 or induce AXIN2 expression. Interestingly, this cell-autonomous DVL activation by Wnt overexpression occurred despite the inability of HT1080 to respond to paracrine WNT4 (Fig. 3) and the lack of WNT3A secretion in HT1080-PKO (Fig. 2). Similarly, Wnt overexpression did not activate LRP6 in MM134 (Fig. S7) further supporting that secreted WNT4 does not activate paracrine signaling. These data support that WNT4 and WNT3A can mediate intracellular signaling independent of secretion or activation of extracellular signaling (i.e. LRP6).
As WNT4 is required for bone regeneration and cell proliferation (22), we examined whether WNT4 is similarly essential for proliferation and/or viability of HT1080 or HT1080-PKO. We hypothesized that WNT4 might be dispensable in HT1080, due to redundant Wnt signaling. However, without functional PORCN for secretion and paracrine signaling of Wnt family members, HT1080-PKO may become reliant on cell-autonomous PORCN-independent WNT4 signaling. WNT4 knockdown induced ϳ21% cell death at 48 h post-knockdown in HT1080 (Fig. 6C), leading to a modest suppression of proliferation (Fig. 6D). In contrast, WNT4 knockdown in HT1080-PKO strongly suppressed growth, and cell death was accelerated (ϳ70% cell death at 48h post-knockdown). Importantly, minimal cell death was induced by WLS knockdown in HT1080 or HT1080-PKO, despite global suppression of Wnt secretion in the latter with ablation of both PORCN and WLS. The sensitivity of HT1080-PKO to knockdown of WNT4, but not WLS, supports a critical role for secretion-independent functions of WNT4, in particular in the absence of paracrine signaling by other Wnt proteins.  D). A and C, immunoblots of whole-cell lysates from the treated HT1080-PKO cells were run and probed for DVL2, DVL3, total LRP6, and phosphorylated LRP6. B and D, mRNA from the treated HT1080-PKO cells was extracted for qPCR for AXIN2 mRNA expression levels versus RPLP0. Cells treated with 62.5 ng/ml rWNT3A as a positive control. Points represent two biological replicates; line ϭ mean Ϯ S.D.; data represent three independent experiments. Statistics were obtained using ANOVA with Dunnett's multiple correction; ns ϭ not significant. E, HT1080-PKO cells were transfected with a TOP-FLASH reporter plasmid and then co-cultured with either HT1080 or HT1080-PKO WNT overexpressing (OE) cells, with or without 10 nM LGK974. Ctrl represents TOP-FLASH-transfected HT1080-PKO without co-culture. WNT signaling activity, as measured by luminescence, was performed using a dual-luciferase assay. Statistics obtained using Student's unpaired t test compared with the no co-culture control (Ctrl). * represents p Ͻ 0.05. Points represent three biological replicates; line ϭ mean Ϯ S.D. Results are representative of two independent experiments. F and G, HT1080-PKO control cells were treated for 24 h with recombinant WNT protein (at concentrations of 10, 50, 100, 250, or 500 ng/ml). Immunoblots of whole-cell lysates were run (F), and mRNA extracted for qPCR was performed as above (G). Points represent two biological replicates; line ϭ mean Ϯ S.D.; data represent two independent experiments. Statistics obtained using ANOVA with Dunnett's multiple correction.

PORCN-independent Wnt signaling
To confirm that cell-autonomous PORCN-independent WNT signaling was similarly independent of WLS, we knocked down WLS or PORCN in HT1080-PKO Wnt-overexpressing cells and assessed DVL3 activation (Fig. 6E). siPORCN did not affect DVL3 activation, consistent with our above observations. DVL3 activation in HT1080-PKO:W3 was suppressed, suggesting disparate roles of PORCN and WLS in activating paracrine versus cell-autonomous signaling. However, siWLS did not impact cell-autonomous DVL3 activation by WNT4, further supporting a secretion-independent cell-autonomous signaling mechanism for WNT4.
Our data show that PORCN and acylation are not required for WNT4 signaling. Typically, PORCN-mediated acylation embeds Wnt proteins into the endoplasmic reticulum membrane and leads to trafficking through the Golgi for secretion (45,51); this process renders Wnt proteins hydrophobic. However, WNT4 that mediates cell-autonomous signaling independent of PORCN may remain hydrophilic. To test this, we applied Triton X-114 phase separation to cell lysates from Wntoverexpressing cells (Fig. 6, F and G). In HT1080:W3, WNT3A was detected in both aqueous and detergent phases. However, WNT3A was absent in detergent phase from HT1080-PKO: W3, consistent with a loss of PORCN-mediated acylation. Both endogenous and overexpressed WNT4 were detected in the detergent phase in HT1080, but WNT4 shifted to the aqueous phase in HT1080-PKO. The presence of endogenous WNT4 largely in the aqueous phase in HT1080-PKO, taken together with the sensitivity of HT1080-PKO to WNT4 knockdown (Fig.  6D), supports that nonacylated WNT4 is contributing to cell signaling. Similarly, in MM134 cells (Fig. 6G), endogenous WNT4 was detectable primarily in the aqueous phase, supporting that cell-autonomous WNT4 signaling in MM134 is mediated by nonacylated WNT4.

Discussion
The Wnt-modifying enzyme PORCN is commonly described as a gatekeeper for the secretion of Wnt proteins, and thus PORCN inhibition is an approach to broadly block Wnt signaling without targeting cell type-or tissue-specific downstream Wnt pathways. WNT4 signaling is required for survival and proliferation of ILC cells, but we show that WNT4 activity is PORCN-independent. PORCN was not required for WNT4 secretion from a panel of cell lines, as genetic or chemical PORCN blockade had no effect on WNT4 secretion. However, WNT4 was not capable of activating paracrine Wnt signaling in any model tested, despite the ability of recombinant human WNT4 to do so in a context-dependent manner. These data together suggest that secreted WNT4 is unique and may not be responsible for driving signaling in WNT4-expressing cells. Instead, we determined that WNT4 and WNT3A can activate cell-autonomous, intracellular signaling independent of secretion ( Fig. 7 and Fig. S8). This unique mode of Wnt signaling is likely essential for the survival and proliferation of WNT4-dependent cells.
Our observations of PORCN-independent WNT4 signaling are supported by other studies showing a disconnect between PORCN activity and Wnt secretion/signaling, and they support that the requirement for PORCN is context-dependent and not absolute. Nusse and co-workers (52) reported PORCN-independent secretion and activity of Drosophila WntD. WntD is secreted at high levels in fly tissues and cell culture models LGK974:   LGK974, siPORCN, or siWLS (as in Fig. 2). C, E, and G, alkaline phosphatase production induced by conditioned media from WNT3A-or WNT4-overexpressing cell lines: C, SUM44PE; E, HCC1428; or G, PEO1. Points represent four biological replicates; line ϭ mean Ϯ S.D. Statistics were obtained using ANOVA with Dunnett's multiple correction. Above data are representative of two independent experiments. ns ϭ not significant.    (24), discussed below) blocks WNT4 palmitoylation but not secretion, yet ultimately it is associated with WNT4 loss-of-function, consistent with a disconnect between PORCN, WNT4 secretion, and WNT4 signaling. These studies together show Wnt proteins secreted in a PORCN-independent manner can be active in paracrine-signaling models, but to our knowledge, this is the first report of PORCN-independent Wnt activity that is also independent of secretion. Of note, Kurita et al. (56) recently demonstrated that Wnt4 siRNA in mouse pancreatic ␤-cells suppressed glucoseinduced insulin secretion, but recombinant Wnt4 had no effect on insulin secretion. These data parallel our findings and sup-port that WNT4 signals independent of PORCN, secretion, and paracrine activity. Our data, together with the above reports on PORCN-independent Wnt signaling, highlight that the roles of PORCN and WLS in Wnt modification, secretion, and signaling are Wntspecific, as well as cell type-and context-dependent. This observed context-dependence includes the specific "receiver" cells in question; this is perhaps best understood in the context of differentially-expressed FZD proteins, which may still mediate cell-autonomous Wnt signaling (57). However, the contextdependence of Wnt signaling also includes the source of the Wnt protein ( Fig. 7 and Fig. S8). For example, WNT3A secreted from HT1080 robustly activated Wnt signaling in 3T3-E1 or HT1080-PKO cells, and PORCN was required for both secretion and paracrine activity. In contrast, WNT3A secreted from MM134 cells activated Wnt signaling in 3T3-E1 but not HT1080-PKO cells, and PORCN was required for paracrine activity but not secretion. Cell type-specific regulation of Wnt protein post-translational modification (e.g. glycosylation patterns) may drive these differences (54,58), which parallel our observation that recombinant Wnt proteins were more promiscuously able to activate signaling than secreted Wnt proteins (e.g. rWNT4, but not secreted WNT4, activated 3T3-E1 Wnt signaling). Future studies should incorporate both secreted and rWnt proteins, because rWnt proteins may thus represent a specific functional context versus secreted Wnt protein and have distinct paracrine activity.

PORCN-independent Wnt signaling
WNT4 and WNT3A likely signal via at least three distinct mechanisms: 1) as a secreted protein with PORCN modification; 2) as a secreted protein without PORCN modification; and 3) by a cell-autonomous mechanism independent of secretion. This may explain the myriad context-dependent signaling controlled by WNT4, including activating canonical ␤-catenin activity (15,36,59), repressing ␤-catenindriven transcription (60,61), or activating noncanonical Wnt-signaling pathways (22,62). The signaling differences we observed between recombinant and secreted WNT4 indicate differential protein processing may guide Wnt proteins to activate distinct signaling pathways. The commercial recombinant Wnt proteins used in our study lack the N-terminal signal peptide (residues 1-22 for WNT4); however, Wnt signal peptides may have important roles in the regulation of signaling activity (45). The WNT4 signal peptide has a uniquely high content of arginine (14%) and serine (18%) compared with other human Wnts (average of 5 and 9%, respectively), whereas WNT3A has no charged or polar residues in its signal peptide. Mutation in the signal peptide (L12P) of WNT4 has also been linked to Mayer-Rokitansky-Küster-Hauser syndrome (24). The L12P mutant functions as a dominant-negative inhibitor and suppresses the activity of WT  Figure 7. WNT4 activates signaling in an atypical manner independent of secretion. WNT4 can be modified by PORCN during the process of secretion, but WNT4 can instead be secreted (or presented on the cell membrane surface) in a PORCN-independent, WLS-dependent manner. In either case, secreted WNT4 (or WNT4 embedded in the cell membrane) does not have paracrine activity in any context tested herein. Cell-autonomous WNT4 signaling can activate DVL proteins intracellularly and likely signals via other pathways in ILC cells.

PORCN-independent Wnt signaling
WNT4 when co-expressed. Although the L12P mutant protein is not palmitoylated, it is secreted and does not prevent secretion of WT protein (24). These observations are consistent with our findings of PORCN-independent WNT4 secretion. The mechanism of dominant-negative activity has not been described but suggests distinct forms of WNT4 may drive cellautonomous Wnt signaling. Furthermore, understanding the cellular localization of WNT4, and the effects of acylation or other processing, will be important toward understanding function. Our phase separation data suggest that because WNT4 is aqueous in ILC cells (Fig. 6G), it likely is not embedded in a cellular membrane as typical of other Wnt proteins (45,51) and is present in a distinct cellular compartment. Characterizing distinct WNT4 species is an important future direction.
ILC is a unique context for paracrine Wnt signaling, as we observed that both WNT3A and WNT4 were secreted from ILC cells in a PORCN-independent manner. Although WNT3A activity remained dependent on PORCN (similar to Wnt5a in 293T cells (63)), the role of WNT4 processing in paracrine activity is unclear. However, this atypical Wnt processing in ILC cells (also observed in relation to glycosylation) may be related to broader Wnt signaling dysfunction. The genetic hallmark of ILC is the loss of E-cadherin (CDH1) (42), which leads to dysfunction of catenin proteins, including activation of p120 catenin (64) and inactivation of ␤-catenin. E-cadherin loss in ILC leads to a loss of ␤-catenin protein in both patient tumors and cell lines (41,42), and as a result, a ␤-catenindriven TOP-Flash reporter cannot be activated in ILC cells (41). Catenin dysfunction was previously postulated as being linked to PORCNi sensitivity, and ILC patients were specifically included in a trial of WNT974 (NCT01351103). This trial opened in 2011, but by 2015 ILC patients were removed from the inclusion criteria. It is unclear whether this is due to accrual problems or a lack of efficacy, as neither have been reported for ILC patients on this trial, although our data suggest PORCNi are unlikely to have clinical efficacy for ILC. This highlights the importance of defining the unique context for Wnt signaling in ILC, in particular for WNT4, based on our prior findings (41). Our laboratory has begun to characterize WNT4-driven signaling in ILC cells (65), which are likely mediated by PORCN-independent, cell-autonomous WNT4 signaling.
Our study uses diverse cell-line models to investigate Wnt secretion and paracrine activity of Wnt proteins. We used nontagged Wnt expression constructs, eliminating previously described complications caused by epitope tags on Wnt proteins. Wnt signaling activity was measured by complementary pathway readouts that facilitated our study of context-dependent Wnt signaling activities. The cell-line models used for Wnt secretion were all cancer-derived, and thus further study is needed to translate our findings to normal tissue, developmental, or in vivo contexts. Notably, a recent study by Speer et al. (66) demonstrated that mutation of the critical palmitoylation residue did not suppress the secretion and/or activity of Wnt proteins in a zebrafish development model, in a Wnt-specific context, supporting a role for PORCN-independent Wnt signaling in development. Regarding our cell models, we observed identical processing and secretion for endogenous WNT4 as for overexpressed WNT4 (models herein express low or no endogenous WNT3A), but overexpression was required to facilitate signaling experiments. Future studies will determine the contribution of endogenous Wnt protein levels to activating the signaling pathways discussed herein.
The secretion and paracrine activity of Wnt proteins are heavily context-dependent, as WNT3A and WNT4 present with differing dependence on PORCN for secretion and paracrine activity in distinct model systems. Secretion of WNT4 is PORCN-independent, but WNT4 did not present paracrine activity in cells that were dependent on WNT4, indicative of secretion-independent, cell-autonomous activity. Both WNT4 and WNT3A presented cell-autonomous activity independent of secretion. Our studies identify a PORCN-independent mode of Wnt signaling that may be critical to understanding WNT4driven cellular contexts or those that are otherwise considered to have dysfunctional Wnt signaling.
For treatments with conditioned medium, conditioned medium was collected as below (see under "Protein extraction from conditioned medium"). Conditioned medium was filtered to remove cellular debris (0.45 m). Receiver cells (HT1080-PKO) were treated with a mixture of fresh base medium and conditioned medium at a 1:1 ratio, unless indicated otherwise, for 2-24 h as indicated.

Reagents and plasmids
LGK974 was obtained from Cayman Chemical (

Protein extraction from conditioned medium
HT1080, MM134, HCC1428, or PEO1 cells were plated in full medium (as above, 10% FBS). Decreased FBS is required to reduce competition of serum proteins with secreted Wnt proteins during extraction. 24 h post-plating, medium was changed to reduced serum medium: DMEM/F12 ϩ 2% FBS (HT1080, PEO1); DMEM ϩ 2% FBS (HCC1428); and DMEM/L15 ϩ 5% FBS (MM134). Conditioned medium was harvested 4 -5 days later for HT1080 cells and 6 -7 days later for MM134, HCC1428, and PEO1 cells, typically once medium acidification was apparent by phenol red color change. SUM44PE cells are normally cultured in low serum (DMEM/F12 ϩ 2% charcoalstripped FBS), so standard medium was allowed to condition for 6 -7 days before harvesting. Medium was centrifuged at 300 ϫ g for 4 min to pellet any cells or debris. The supernatant was then syringe-filtered using a 0.2-m filter. The protein concentration of the conditioned medium was measured using a Pierce BCA protein assay kit (Thermo Fisher Scientific, catalog no. 23225), and medium volumes, normalized to total protein, were adjusted with sterile-filtered PBS. StrataClean Resin (Agilent Technologies, Santa Clara, CA, catalog no. 400714) was added to the conditioned medium at a ratio of 10 l of resin to 100 g of medium protein, and vortexed to resuspend the resin. The medium ϩ resin mixture was incubated, rotating at 4°C for 30 min, and then centrifuged at 425 ϫ g for 1 min at 4°C. The supernatant was then aspirated, and the resin was washed using sterile-filtered PBS. The resin was centrifuged again at 425 ϫ g for 1 min at 4°C, and the supernatant was aspirated. An equal volume of 2ϫ Laemmli sample buffer (Bio-Rad, catalog no. 1610737) was added to the resin to release bound protein, and the slurry was heated at 95°C for 5 min. Slurry equivalent to 100 g of conditioned medium protein was run on SDS-polyacrylamide gels to detect secreted WNT3A or WNT4 via immunoblotting (below).

Alkaline phosphatase assay
MC3T3-E1 cells were seeded into a 96-well plate at 10,000 cells/well. 24 h later, the cells were treated with either conditioned media (CM) or recombinant WNTs (rWNTs). 72 h later, alkaline phosphatase activity was assessed, using para-nitrophenyl phosphate as a substrate and measuring absorbance at 405 nm. Assay protocol is based on Nakamura et al. (69) with buffer modifications provided by the recombinant Wnt manufacturer (R&D Systems). A more detailed protocol is available upon request. para-Nitrophenyl phosphate tablets were obtained from Cayman Chemical (catalog no. 400090).

Co-culture and dual-luciferase assay
HT1080-PKO cells were transfected with both a TOP-FLASH and a Renilla reporter plasmid. 24 h later, these cells were co-cultured with either HT1080 or HT1080-PKO WNToverexpressing cells, with or without 10 nM LGK974 treatment. 48 h later, luciferase and Renilla activity was assessed using Promega's dual-luciferase assay (Promega, Madison, WI, catalog no. E1910) and a BioTek Synergy 2 microplate reader (BioTek, Winooski, VT) with a dual injector system according to the manufacturer's instructions.

Proliferation and viability assays
Cells were seeded in a 96-well plate, and 24 h later cells were treated with 100 nM SYTOX TM Green (Thermo Fisher Scientific, catalog no. S7020). The plate was then placed into an Incucyte Zoom (Essen Bioscience, Ann Arbor, MI) for 4 -5 days where they were imaged every 4 h at ϫ10 magnification. Cell confluence and cell death (SYTOX TM Green-positive counts) were assessed using Incucyte S3 software (version 2018A). After time-course completion, total dsDNA was measured by hypotonic lysis of cells in ultrapure H 2 O, followed by addition of Hoechst 33258 (Thermo Fisher Scientific, catalog no. 62249) at 1 g/ml in Tris-NaCl buffer (10 mM Tris, 2 M NaCl, pH 7.4) at an equivalent volume to lysate. Fluorescence (360 nm excitation/460 nm emission) was measured on a Bio-Tek Synergy 2 microplate reader.

Triton X-114 phase separation
Separation of aqueous and detergent phases of conditioned medium or cell lysates was performed based on previously published protocols (46,70,71), with minor modifications.
Conditioned medium-Ice-cold Triton X-114 Buffer (10 mM Tris base, 150 mM NaCl, 2% Triton X-114, pH 7.6) was added to medium at a 1:1 ratio and incubated at 37°C for 45 min or until the top aqueous phase was clearly distinguishable from the bottom Triton X-114 pellet (detergent). Samples were then centrifuged at 2,000 ϫ g for 10 min at 37°C. The aqueous and detergent phases were carefully removed and stored separately on ice. StrataClean Resin was added to the aqueous phase as above, with the exception of TBS used in place of PBS. 2ϫ Laemmli sample buffer was added to both aqueous and detergent phase samples and heated to 70°C for 10 min. 25-50% of total material was loaded for SDS-PAGE and immunoblotting (typically equivalent to 25-50 g of total protein from conditioned medium).
Cell lysate-Whole-cell lysates were harvested as above using Triton X-114 Buffer (10 mM Tris base, 150 mM NaCl, 2% Triton X-114, pH 7.6) supplemented with protease/phosphatase inhibitors. Fresh Triton X-114 Buffer was added at a 1:1 ratio, and processing was continued as for conditioned medium. SDS-PAGE was performed as with conditioned medium (no StrataClean-based protein extraction from the aqueous phase); aqueous and detergent phases were loaded at a 1:2 ratio to improve detection of protein from the detergent phase. Note that maintaining the detergent phase sample (in Laemmli buffer) at 70°C reduced sample viscosity to improve gel loading.

Statistical considerations
Graphpad Prism version 7.02 was used for all graphical representations and statistical analyses. All Western blotting figures are representative of at least two to three independent experiments.

Software
Prism (GraphPad Software, San Diego, CA, version 7.02) and Image Studio (LI-COR Biosciences, Lincoln, NE, version 5.2) were used to obtain and analyze that data presented in this manuscript.
Data associated with experiments herein will be available at an Open Science Framework repository (72).