Liposomal Packaging Generates Wnt Protein with In Vivo Biological Activity

Wnt signals exercise strong cell-biological and regenerative effects of considerable therapeutic value. There are, however, no specific Wnt agonists and no method for in vivo delivery of purified Wnt proteins. Wnts contain lipid adducts that are required for activity and we exploited this lipophilicity by packaging purified Wnt3a protein into lipid vesicles. Rather than being encapsulated, Wnts are tethered to the liposomal surface, where they enhance and sustain Wnt signaling in vitro. Molecules that effectively antagonize soluble Wnt3a protein but are ineffective against the Wnt3a signal presented by a cell in a paracrine or autocrine manner are also unable to block liposomal Wnt3a activity, suggesting that liposomal packaging mimics the biological state of active Wnts. When delivered subcutaneously, Wnt3a liposomes induce hair follicle neogenesis, demonstrating their robust biological activity in a regenerative context.


Introduction
Wnt signals are implicated in the self-renewal and proliferation of stem cells from a variety of adult tissues [1][2][3] but despite numerous large-scale screenings [4][5][6][7], no specific small molecule Wnt agonists have been identified. As a consequence, the therapeutic application of Wnt proteins -of which there are potentially many [8][9][10] -has been severely hampered. We exploited a unique and essential feature of Wnt proteins, their covalent modification by several lipids, to package the purified proteins in a manner that both stabilizes and enhances their in vivo biological activity. Palmitoylation is essential for Wnt activity [1,11]; therefore we reasoned that liposomes might serve as an ideal delivery vehicle for such a hydrophobic molecule.
Liposomes are spherical nanovesicles consisting of an aqueous core enclosed in one or more phospholipid layers (reviewed in [12]). Initially, liposomes were developed in an attempt to improve the pharmacokinetics and tissue distribution of chemotherapeutic agents [13][14][15]. Consequently, bioengineers and chemical engineers have invested considerable time and effort into manufacturing liposomes that retain the drug or molecule of interest and prevent its degradation. Such preparations would ideally effectively evade detection by the reticuloendothelial system; could be targeted to the tissue of interest; and could be induced to release the drug/molecule when required. For example, the addition of polyethylene glycol (PEG) can prolong the circulatory half-life of liposomes, perhaps acting through steric hindrance ( [14][15][16]; reviewed in [17]).
At present, the primary application for liposome technology is for the treatment of cancers (reviewed in [14,15,18]). In this clinical scenario, the objective is to deliver cytotoxic drugs to a tumor whilst simultaneously preserving cell viability in the rest of the body. Our objective differed slightly: we wanted to develop a method to deliver Wnts to a tissue that simultaneously preserved biological activity and restricted diffusion of the protein.

Purification of Wnt3a liposomes
Mouse Wnt3a protein was purified as described [1], without the heparin purification step. The product, containing approximately equal amounts of Wnt3a and bovine serum albumin, was concentrated further to 250 ng/ml in PBS with 1% CHAPS.
Unless otherwise indicated, 14 mmol of DMPC in chloroform were dried to a thin film in a 10 ml round bottom flask using nitrogen gas and were further evaporated in a vacuum overnight. Purified Wnt3a in 1% CHAPS in 16 PBS was then diluted in 16 PBS to a total concentration of 1-1.3 mg/ml. This solution was then added to the 10 ml flask and vortexed vigorously until the solution was cloudy and there was no lipid visible on the bottom of the flask. The lipid solution was then extruded 40 times through a 100-200 nm polycarbonate membrane in a thermo-barrel extruder held at 30-32uC (Avanti Polar Lipids, Inc).
To separate liposome-associated Wnt3a from free Wnt3a, the extruded solution was spun in an Optima TLX Ultracentrifuge (Beckman Coulter, Fullerton, CA) at 28,000 rpm for a minimum of 30 minutes at 4uC. The supernatant was removed and the lipid pellet re-suspended in 16 DMEM (Mediatech, Inc., Herndon, VA). The liposomes, if not used immediately, were stored at 4uC.

Cell culture and activity assays for Wnt3a liposomes
Mouse LSL cells were grown at 37uC and 5% CO 2 in 16 DMEM, 10% FBS, and 1% Penicillin/Streptomycin (Mediatech, Inc., Herndon, VA). The cells were plated in 96-well plates with an initial density of 25,000 cells/well and allowed to recover overnight. The cells were then treated as described and incubated for an additional 17 hours. As the LSL cells constitutively express bgalactosidase and express luciferase in response to Tcf/Lef binding, activity was assessed via the Dual-LightH Combined Reporter Gene Assay System (Applied Biosystems). Relative luciferase units were measured and normalized against b-galactosidase activity. Error bars indicate standard deviation. All assays were done in triplicate. Student's T test was employed to determine statistical significance.

Determination of effective Wnt3a concentration in Wnt liposomes
Increasing amounts of purified Wnt3a protein were added to LSL cells, grown in 96-well plates, in order to generate an activity gradient. In a parallel set of experiments, LSL cells were exposed to different volumes of Wnt3a liposomes. All experiments were performed in triplicate. Activity of the purified protein was then determined by luciferase activity (as described above) and liposomal Wnt3a activity was plotted on the same graph. From these data we interpolated the concentration of active Wnt3a in a given volume of liposomes.

Trypsin digestion of Wnt3a liposomes
Exo-liposomal Wnt3a protein was removed by subjecting liposomes to trypsin digestion. Briefly, 75 ml of Wnt3a liposomes were added to 405 ml of 16DMEM and 20 ml 16Trypsin solution (Mediatech, Inc., Herndon, VA) for a final mixture containing 4.0% trypsin. Liposomes were then incubated at 37uC for 0, 5, 10, 15, or 20 minutes. The proteolytic reaction was quenched by the addition of 1 ml of 16 DMEM containing 10% FBS. The samples were ultra-centrifuged at 28,000 rpm for 30 min at 4uC; the supernatant was then removed and the lipid pellet re-suspended in 16DMEM. The samples were then added to LSL cells in triplicate, or prepared for standard Western blot analysis.

Isolation and characterization of Wnt antagonists
We undertook a small molecule Wnt antagonist high-throughput screen using a mammalian cell-based reporter assay. The reporter plasmid (Super8xTOPflash) was stably introduced into the mouse C3H10T1/2 cell line, and the best 10T1/2 Super8-xTOPflash clone gave ,800 fold up-regulation of luciferase activity after 24 hour incubation with Wnt3a conditioned media.
Under these assay conditions we screened ,205,000 small molecule compounds at a concentration of 10 mM and identified several potential Wnt antagonists (hereafter referred to as Ant). These antagonists were further evaluated using a luciferase reporter driven by the SV40 large T antigen promoter (for detecting general non-specificity) and a minimal promoter with multimerized Gli binding sites (for detecting activity on the Hedgehog pathway). The primary screen hit for the compound series described in this paper was effective at inhibiting Wnt signaling (IC50,1-2 mM) but did not function as a Hedgehog inhibitor (IC50.20 mM) and showed no inhibitory effect on SV40 promoter driven luciferase reporter at concentrations up to 30 mM. In the derivative Ant 1.4, either a bromide or a chlorine atom was substituted in the ortho position to enhance activity over the unsubstituted molecule.
Evaluation of Wnt antagonist activity in autocrine and paracrine cultures C3H10T1/2 cells were transfected with a Super8xTOPflash luciferase reporter plasmid (Upstate Cell Signaling Solutions). PA-1 cells (ATCC) were transfected with a TOPflash (Upstate Cell Signaling Solutions) luciferase reporter plasmid. Both cell lines were maintained using standard tissue culture protocols.
PA-1 cells exhibit autocrine Wnt signaling [19]; to monitor the ability of antagonists to block autocrine Wnt activity, cells were plated in 96-well plates at 20,000 cells/well in growth medium; 24 hrs later, the cells were changed to fresh growth medium and the antagonists were added in the presence of 20% Wnt3a condition medium, or 0.1 mg/ml purified Wnt3a protein. After 24 hrs, plates were assayed for luciferase activity with the LucLite kit (Packard).
To test the ability of the antagonists to block an exogenous Wnt signal and autocrine Wnt signal simultaneously, PA-1 cells transfected with Super8xTOPflash reporter were plated as described above; 24 hrs later, the cells were changed to fresh growth medium and the antagonists were added in the presence of 20% Wnt3a condition medium, or 0.1 mg/ml purified Wnt3a protein. After 24 hrs, plates were assayed for luciferase activity with the LucLite kit (Packard).
To test the ability of the antagonists to block paracrine Wnt signaling, we co-cultured 10T1/2 Super8xTOPflash reporter cells with L cells stably expressing Wnt 3a protein. In all assays, hFzd8 CRD, an antagonist of Wnt signaling [19], served as a positive control for Wnt inhibition.

Wnt antagonist activity: liposomal preparation
Wnt antagonist (Ant) 1.4Cl was diluted from 20 mM in DMSO to final concentrations of 2 mM or 20 mM in 16 DMEM. LSL cells were treated with 15 mL of Wnt3a liposomes or 15 mL of purified Wnt3a protein and varying concentrations of Ant 1.4Cl ranging from 0.01 mM to 10 mM and allowed to incubate overnight. The activity assay was conducted as previously mentioned and was conducted in triplicates. Percent inhibition values are relative to non-antagonist treated Wnt3a liposome or free Wnt3a protein activity as appropriate.

Animal studies
All experiments were performed in accordance with Stanford University Animal Care and Use Committee guidelines. Six threeweek old, male CD-1 mice were purchased from Charles River Laboratories, Inc. (Wilmington, MA). Animals were housed in a light-and temperature-controlled environment and given food and water ad libitum.
Before subcutaneous injection of Wnt3a/PBS liposomes, the back of the mice was shaved and cleansed with antiseptic Betadine. Two regions, each 1 cm in diameter, were marked approximately 5 mm lateral from the midline. Mice received subcutaneous injections of either 20 ml Wnt3a liposomes or PBS liposomes every other day. Tissues were harvested after 14 days (7 injections) and fixed in 4% Paraformaldehyde overnight. Samples were processed for paraffin embedding, sectioned at 8 mm thickness and stained with H&E for histological evaluation.

Results
Our first goal was to determine if liposomal packaging affected the activity of Wnt3a protein. We generated Wnt3a liposomes and compared their activity to purified soluble Wnt3a protein using the LSL dual-reporter cell assay [20]. These cells show robust, dosedependent expression of luciferase in response to Wnt3a treatment (Fig. S1).
In our initial formulation of Wnt liposomes we first tested the activity of Wnt protein associated with a liposome. The liposomal preparation of Wnt3a was made by combining 1.0 mg/ml of Wnt3a protein with lipids to generate a solution with an estimated concentration of 0.5 mg/ml. Instead of being distributed equivalently in the supernatant and the pellet, however, Western blot analyses indicated that the lipid pellet contained the majority of the Wnt3a protein (Fig. 1A). This lipid-associated Wnt3a exhibited activity using the same luciferase assay (Fig. 1A).
Because of its lipid modifications, we speculated that Wnt3a protein would associate with the lipid bilayer in such a way that Wnt positioned on the exo-liposomal surface would be available for receptor binding but Wnt positioned in the endo-liposomal surface would not. To test this hypothesis we subjected the Wnt3a liposomes to trypsin digestion to remove protein on the exo-liposomal surface. Trypsin digestion completely inactivated the Wnt3a liposomes (Fig. 1B). The ,20% of the Wnt3a protein remained associated with the liposomal fraction where it did not elicit any activity (Fig. 1C). We therefore conclude that the majority (80%) of Wnt3a is positioned on the exo-liposomal surface where it exhibits biological activity, while a small percentage remains localized to the endo-liposomal environment where it is unavailable for signaling (Fig. 1C).

Liposomal packaging enhances Wnt3a activity
During fabrication of Wnt3a liposomes, we calculated that ,55% of the input protein is incorporated in a manner that exhibits biological activity, while the remaining 45% is either lost in the supernatant, or is sequestered in the endo-liposomal space. Does the liposomal presentation of Wnt3a affect its biological activity? We compared the activity of Wnt3a protein against Wnt3a liposomes containing the same concentration of active Wnt3a on the exo-liposomal surface ( Fig. 2A). Based on the use of equivalent concentrations of active Wnt3a, we calculate that the liposomal preparation of Wnt3a exhibits a 5-fold increase in biological activity compared to the isolated protein ( Fig. 2A). Liposomal packaging preserves Wnt3a activity We also tested whether liposomal packaging preserved the biological activity of Wnt3a over time. TOPflash cells were plated at different cell densities in order to insure that cells were in the exponential growth phase throughout the experiment. Cells were treated with various concentrations of Wnt3a protein, or Wnt3a liposomes that were normalized for having similar activity. Reporter activity was determined at each time point and normalized against cell number. We found that Wnt3a and Wnt3a liposomes exhibited similar biological activity after 24 and 48 h in culture but Wnt3a liposomes showed significantly higher activity after 72 h in comparison to Wnt3a protein (Fig. 2B). Therefore, Wnt3a positioned on the exo-liposomal surface is active, and this packaging scheme potentiates and stabilizes the activity of Wnt3a in vitro.
We examined whether DMPC could be replaced by other lipids with varying carbon side chain lengths (i.e. DPPC has a fourteen carbon backbone whereas DMPC has a twelve carbon backbone). We found that lipid formulations considerably influenced the activity of the liposomes, with DMPC being optimal (Fig. 3), despite the fact that Western analyses indicated that a similar amount of total protein was incorporated into each liposomal preparation (Fig. 3).

Identification of Wnt antagonists by high through-put screening
In parallel series of experiments, we sought to identify molecules that were effective at inhibiting Wnt signaling. In our initial screening we used 10T1/2 cells transfected with the Super8-xTOPflash Wnt reporter construct, and treated the cells with soluble Wnt3a. We focused on a molecule that blocked Wntinduced reporter activity (Fig. 4) and also repressed Axin2 expression in 10T1/2 cells and SAX1 and GAD1 expression in teratocarcinoma cells induced with exogenous Wnt3a (data not shown).
In an effort to improve potency of the antagonists we synthesized over 200 derivatives and tested these derivatives in the 10T1/2 Super8xTOPflash reporter assay. The most potent antagonists had IC50s less than 1 nM, indicating over 1000-fold enhancement in potency by chemical modification (data not shown). Epistasis analysis indicates that this class of small-molecule inhibitors acts upstream of b-catenin at the level of Wnt ligands or receptors (data not shown). Of these, Ant 1.4Br and one of its derivatives, Ant 1.4Cl, were tested further for their ability to inhibit autocrine Wnt signaling (Fig. 4A).

Antagonists discriminate between an autocrine and an exogenous Wnt signal
The ability of Ant 1.4Br and Ant 1.4Cl to block autocrine Wnt signaling was tested in three separate cell lines (NCCIT, NTera2, and PA-1) all of which were transfected with a luciferase-based Wnt reporter construct. We compared the inhibitory effects of Ant 1.4Br and Ant 1.4Cl against hFzd8CRD. In an autocrine signaling assay, hFzd8CRD repressed luciferase activity by 50% (Fig. 4B). In contrast, Ant 1.4Br and 1.4Cl inhibited luciferase activity by only 10% (Fig. 4B).
Ant1.4Br and 1.4Cl were also ineffective against paracrine Wnt signaling. When cultured alone, 10T1/2 Super8xTOPflash reporter cells showed no luciferase activity but when co-cultured with L cells stably expressing Wnt3a protein, then luciferase activity was increased 25-fold, which was largely inhibited by hFzd8-CRD (Fig. 4C). Ant 1.4Br and its derivatives, however, were largely ineffective and only repressed luciferase activity to a maximum of 30% (Fig. 4C).

Wnt antagonists are ineffective against liposomal Wnt
Ant1.4Br and Ant 1.4Cl compounds discriminate between purified Wnt3a, and Wnt3a secreted from cells in a paracrine/ autocrine assay. This difference was not attributable to alterations in the protein itself either, since the paracrine Wnt signal was from L cells and these were the same cells used to produce the purified Wnt protein. Instead, these data imply that paracrine/autocrine

cells transfected with a TOPflash
Wnt reporter construct were co-cultured with LSL cells secreting Wnt3a. No luciferase activity was detectable in 10T1/2 cells grown alone (control, brown line); luciferase activity was at a maximum when cells were co-cultured with LSL cells secreting Wnt3a (blue line). Fzd8CRD (yellow) inhibited paracrine-induced Wnt reporter activity in a dose-dependent manner, to a maximum of 90%. Ant 1.4Br (red) reduced luciferase activity by less than 20%. (D) LSL cells transfected with a TOPflash Wnt reporter construct were treated with either exogenous Wnt3a or liposomal Wnt3a in the presence of Ant 1.4Cl or hFzd8CRD. Ant 1.4Cl inhibited exogenous Wnt3a-induced reporter activity by 80% but inhibited liposomal Wnt-induced reporter activity by only 30%. hFzd8CRD (yellow) reduced Wnt3a-induced reporter activity by 80% in Wnt3 treated cells and by 70% in liposome treated cells. The symbols * and # reflect statistical significance (Student's T-test, p,0.05). doi:10.1371/journal.pone.0002930.g004 signaling is mediated by a conformation of Wnt that differs from the purified protein. One possibility is that Wnt secreted by cells during autocrine/paracrine signaling is associated with a lipid raft or vesicle that facilitates its transport [21] and that liposomal packaging mimics this biological state.
We hypothesized that association of Wnt with a lipid vesicle, either during autocrine/paracrine signaling or when packaged in a liposome, might impede the activity of the antagonists. We directly tested this hypothesis by treating LSL SuperTOPflash cells with purified Wnt3a or liposomal Wnt3a in the presence of Ant 1.4Cl. Ant1.4Cl blocked ,80% of exogenous Wnt3a-induced luciferase activity, which was comparable to the inhibitory effect of hFzd8-CRD (Fig. 4D). When we treated SuperTOPflash cells with liposomal Wnt3a and Ant1.4Cl, we observed a significant reduction in inhibitory potential of the antagonist. Again, hFzd8-CRD was effective in blocking liposomal Wnt3a-induced luciferase activity (Fig. 4D). Thus, Ant 1.4Br and Ant 1.4Cl were potent antagonists of purified Wnt3a under a variety of assay conditions but were ineffective against autocrine, paracrine, and liposomal Wnt3a.

Liposomal packaging of Wnt3a potentiates its biological activity in vivo
Secreted Wnt is different than purified Wnt, perhaps because the former is associated with a lipid raft. Some data suggest that Wnts are associated with a lipid membrane when they are shuttled between cells [21][22][23]. If this hypothesis is true then we reasoned that the in vivo efficacy of the purified protein would be enhanced by packaging Wnts in liposomes.
We tested the efficacy of the Wnt3a liposomes in an in vivo context, where genetic experiments have demonstrated a function for b-catenin dependent Wnt signaling in hair follicle neogenesis [24]. Wnt3a liposomes or PBS (control) liposomes were injected subcutaneously into the dorsal surface of 6 week old Axin2LacZ/+ mice, in which the LacZ gene is under control of the endogenous Wnt target Axin2 [25,26]. Twenty-four hours later, Xgal staining was considerably stronger in the epidermal/dermal junction and around hair follicles in animals treated with Wnt3a liposomes compared to those treated with PBS liposomes (n = 6 for each condition; Fig. 5A,B). The cellular distribution of the staining, however, was unchanged (Fig. 5A,B).
We assessed the tissue response to Wnt3a liposomes. Sites treated with PBS liposomes showed no discernable change from un-injected controls on post-injection d14 (n = 6; Fig. 5C). In sharp contrast, sites treated with Wnt3a liposomes exhibited robust hair follicle neogenesis (n = 6; Fig. 5D). The dermal thickness was increased by 2-fold and there were significantly more hair follicles within that layer; in addition, the region normally containing subcutaneous fat was filled with hair follicles (Fig. 5D).

Liposomal packaging potentiates the effects of Wnt proteins in vivo
Because of their potential therapeutic value, multiple large scale screens have been conducted in an attempt to identify small molecule agonists of the Wnt pathway [4]. Although molecules that synergize with Wnt proteins have been uncovered, none act as agonists in the absence of Wnts. This characteristic appears to be unique to Wnts. Unlike other pathways regulated by lipidmodified morphogens including Hedgehogs, Wnt signaling requires the assembly of two receptors plus a large and complex cytoplasmic group of molecules including Axin, Dsh, and GSK-3b. This may indeed only by accomplished by Wnt protein in the proper configuration and not by other molecules. Furthermore, most assays and experiments rely on conditioned media or viral over-expression approaches to study the effects of Wnts on cell function, which oftentimes confounds the interpretation of experimental results because of the inherent inability to control the amount, duration, and activity of Wnts.
With the advent of methods to purify Wnt proteins [1,2] some of these difficulties have been circumvented but Wnt proteins are not stable for extended periods of time (Fig. 2). We found that liposomal packaging enhanced and sustained the biological activity of Wnt proteins both in vitro and in vivo (Fig. 2). Furthermore, liposomal Wnt3a exhibited the same biological specificity as purified Wnt3a, as demonstrated by the patterns of Xgal staining following delivery in multiple strains of Wnt reporter mice (Fig. 5 and data not shown).

Liposomal packaging mimics biological transport
Wnt proteins contain a lipid adduct that is required for activity ( [1]; reviewed in [27,28]) but precisely how this lipid modification affects the intercellular transport of Wnt proteins is unclear. Some evidence suggests that palmitoylation is required for Wnt secretion [29,30]. The lipid adduct may be the method by which Wnt is tethered to a cell membrane, which in turn would restrict diffusion and thus allow local concentration of Wnt proteins to reach a threshold level required for biological activity [31].

A biomimetic approach for Wnt delivery in vivo
In Drosophila there is some evidence suggesting that the Wg protein is transported over many cell diameters in small vesicular structures [22]. These data hint at an appealing hypothesis: liposomal packaging mimics the method by which Wnts are normally secreted from cells. Four lines of experimental evidence support this conclusion. First, liposomal packaging potentiates the activity of purified Wnt3a. Second, Wnt antagonists that are effective against purified protein and are ineffective against Wnt secreted from cells are also ineffective against liposomal Wnt. This selective antagonism is likely due to interactions between the antagonists and a region of the Wnt protein that is available in its purified state, but hidden when Wnt is secreted in a paracrine manner or if Wnt is associated with a liposome. Third, liposomal Wnt3a increases reporter activity in vivo but does not alter the distribution of Wnt responsive cells. These data suggest that liposomal Wnt3a acts similar to endogenous Wnts. Fourth, Wnt3a liposomes exhibit robust activity in a biologically relevant model of hair follicle neogenesis.

Therapeutic strategies to exploit liposomal Wnts
Given the role of Wnt signaling in many regenerative processes (planaria; fish tails) the delivery of Wnt protein as a biological reagent has obvious clinical applications. In addition to its well described role in inducing hair development and growth [24,32], Wnts are also implicated in the self-renewal and proliferation of hematopoietic stem cells [2]; mesenchymal stem cells; and neural stem cells. Wnts may also be an effective means to stimulate bone formation after injury or in disease states such as osteoporosis [33]. Any strategy that attempts to target the Wnt pathway to augment tissue regeneration will have to take into consideration the need to selectively and locally activate signaling in the tissue or area of interest, whilst simultaneously restricting Wnt signaling in other parts of the body. Future experiments will focus on the potential to stimulate tissue regeneration using Wnt liposome based approaches.