An in vitro fatty acylation assay reveals a mechanism for Wnt recognition by the acyltransferase Porcupine

Wnt proteins are a family of secreted signaling proteins that play key roles in regulating cell proliferation in both embryonic and adult tissues. Production of active Wnt depends on attachment of palmitoleate, a monounsaturated fatty acid, to a conserved serine by the acyltransferase Porcupine (PORCN). Studies of PORCN activity relied on cell-based fatty acylation and signaling assays as no direct enzyme assay had yet been developed. Here, we present the first in vitro assay that accurately recapitulates PORCN-mediated fatty acylation of a Wnt substrate. The critical feature is the use of a double disulfide-bonded Wnt peptide that mimics the two-dimensional structure surrounding the Wnt acylation site. PORCN-mediated Wnt acylation was abolished when the Wnt peptide was treated with DTT, and did not occur with a linear (non-disulfide-bonded) peptide, or when the double disulfide-bonded Wnt peptide contained Ala substituted for the Ser acylation site. We exploited this in vitro Wnt acylation assay to provide direct evidence that the small molecule LGK974, which is in clinical trials for managing Wnt-driven tumors, is a bona fide PORCN inhibitor whose IC50 for inhibition of Wnt fatty acylation in vitro closely matches that for inhibition of Wnt signaling. Side-by-side comparison of PORCN and Hedgehog acyltransferase (HHAT), two enzymes that attach 16-carbon fatty acids to secreted proteins, revealed that neither enzyme will accept the other's fatty acyl-CoA or peptide substrates. These findings illustrate the unique enzyme–substrate selectivity exhibited by members of the membrane-bound O-acyl transferase family.

Wnt proteins comprise a family of secreted signaling molecules that regulate embryonic development as well as tissue homeostasis and tumorigenesis in adults (1). Signaling by Wnt proteins is dependent on covalent attachment of the monounsaturated fatty acid, palmitoleic acid, to a conserved serine residue found in nearly all Wnt family members (Ser 209 in Wnt3a) (2,3). Palmitoleoylation exerts pleiotropic effects on Wnt biology: it is required for Wnt proteins to bind to the chaperone Wntless, for Wnt protein secretion, and for Wnt binding to the Wnt co-receptor Frizzled (4 -6). Studies of Wnt fatty acylation in cell-based labeling assays revealed that lipid modification of Wnt proteins occurs in two steps. Stearoyl-CoA desaturase first converts palmitoyl-CoA into palmitoleoyl-CoA, which is then utilized as a substrate by the Wnt acyltransferase Porcupine (PORCN) 2 (7). To date, structural information on PORCN is lacking and Wnt acylation has not been reconstituted in an enzymatic assay in vitro.
PORCN is a member of the membrane-bound O-acyl transferase (MBOAT) family of multipass membrane proteins that catalyze fatty acylation of lipid or protein substrates (8). Three members of the MBOAT family transfer fatty acids to secreted proteins: HHAT, the palmitoyl acyltransferase for Hedgehog proteins; GOAT, which transfers octanoate to ghrelin; and PORCN (9). Each of these fatty acylated proteins contributes to disease states in humans. Because fatty acylation is essential for disease-causing activity, HHAT, GOAT, and PORCN have gained attention as drug targets. Peptide-based fatty acylation assays have led to the development of enzyme inhibitors for HHAT and GOAT (10,11). Using cell-based Wnt signaling assays, small molecule inhibitors of PORCN have been identified, one of which (LGK974) has progressed to a clinical trial for treatment of Wnt-driven solid tumors (12,13). However, in the absence of an in vitro assay, the ability of LGK974 to directly inhibit PORCN-mediated Wnt acylation has never been demonstrated. Development of an enzymatic assay for PORCN would shed light on how this enzyme recognizes its substrates and therefore would considerably enhance our knowledge of PORCN-mediated catalysis.

Development of an in vitro assay that recapitulates PORCNmediated Wnt acylation
We first sought to develop an in vitro assay to directly monitor fatty acylation of Wnt by PORCN. The principle of the assay was based on conditions successfully established for Shh palmitoylation by HHAT (14), with a reaction mix consisting of a biotinylated peptide containing residues surrounding the fatty acylation site, microsomal membranes from cells express-  (Fig. 1A). We hypothesized that recognition of the serine acylation site, located internally within Wnt, might be different from Shh, where acylation occurs at the N-terminal cysteine. In the three-dimensional crystal structure of Xenopus Wnt8, the acylated serine is located at the base of a loop whose structure is constrained by two sets of disulfide-bonded cysteines (4). We therefore designed a peptide containing amino acids 199 -219 of Wnt3a with disulfide bonds between Cys 203 -Cys 217 and Cys 205 -Cys 212 , and appended with a C-terminal PEG-biotin (WT Wnt (S-S)) ( To verify that the disulfide bonds were present in the correct positions, the peptides were digested with trypsin and analyzed by LC-MS on an Orbitrap high-resolution mass spectrometer. The measured monoisotopic mass of the intact peptide was within 10 ppm of the theoretical mass, and the measured monoisotopic mass of the tryptic peptides matches within 10 ppm of the theoretical monoisotopic mass (Fig. 1C). Thus, the correct disposition of the two sets of disulfide-bonded cysteines was experimentally verified. Incubation of the disulfide-bonded peptide WT Wnt (S-S) with [ 125 I]IC15:1 CoA and membranes from cells expressing wild-type PORCN resulted in a 5-fold increase of radiolabel incorporation when compared with the signal obtained with membranes from cells transfected with empty vector (Fig. 1A). No radiolabel incorporation above background was obtained when the disulfide-bonded peptide lacking the acylation site (S209A WNT) was used as a substrate. Moreover, the acylation signal was vastly reduced with membranes containing H341A PORCN, a putative active site PORCN mutant. The WT Wnt (S-S) acylation reaction was time-dependent ( Fig. 1D) and saltsensitive (Fig. 1E). No radiolabel incorporation into WT WNT was observed when free fatty acid, [ 125 I]IC15:1, was used instead of [ 125 I]IC15:1 CoA (Fig. 1F). We next tested the importance of disulfide-bonded cysteines in the peptide substrate.
Pretreatment of WT WNT (S-S) peptide with DTT, prior to the addition of the other reaction components, reduced the signal to background levels (Fig. 1G). This loss of activity was not due to an effect of DTT on PORCN, because pretreatment of PORCN-containing membranes with DTT, followed by membrane re-isolation, had no effect on Wnt peptide acylation (Fig.  1G). Taken together, these data strongly suggest that PORCNmediated Wnt fatty acylation can be accurately recapitulated in vitro and that the reaction is dependent on the presence of disulfide-bonded cysteines surrounding the Wnt acylation site.

FDH mutations in Porcn exhibit impaired Wnt acylation activity in vitro
Mutations in the PORCN gene cause focal dermal hypoplasia, an X-linked disorder characterized by multiple congenital abnormalities (16). Several of the reported FDH-associated missense mutations in human PORCN have been analyzed using a cell-based Wnt acylation assay, and have been shown to exhibit decreased acylation activity as well as decreased protein stability (15). To test whether enzymatic activity of PORCN is directly affected by these FDH mutations, we analyzed Wnt peptide acylation in vitro using membranes from cells expressing wild-type or mutant PORCN. The effects on PORCN-mediated Wnt acylation activity in vitro mirrored those obtained with the cell-based Wnt acylation assay: R228C PORCN exhibited activity nearly identical to WT PORCN, whereas the other mutants were compromised ( Fig. 2A).

Direct inhibition of PORCN activity by LGK974
Wnt signaling plays a key role in multiple tumor types, especially breast and colon cancer (17). Given that Wnt acylation is essential for signaling activity, inhibitors of PORCN have the potential to serve as chemotherapeutics in Wnt-driven cancers. High-throughput screening was used to identify GNF-1331, a small molecule that inhibits Wnt secretion in cell-based assays of Wnt signaling and also binds to PORCN-containing membranes (13). Pharmacologic optimization of GNF-1331 resulted in the generation of LGK974, which displaces radiolabeled GNF-1331 from PORCN-containing membranes (IC 50 ϭ 1 nM) and inhibits Wnt signaling in a co-culture assay (IC 50 ϭ 0.4 nM) (13). LGK974 has therefore been designated a PORCN inhibitor and is currently in Phase I clinical trials. To date, however, LGK974 has never been tested for its ability to directly inhibit the enzymatic activity of PORCN. Using our in vitro assay, dose-dependent inhibition of Wnt peptide acylation by LGK974 occurred with an IC 50 ϭ 12.9 nM, which closely mirrors the IC 50 ϭ 7.5 nM obtained for LGK974 in a luciferase-based reporter assay of Wnt signaling (Fig. 2, B and C).

In vitro analyses demonstrate substrate specificity of PORCN and HHAT
PORCN and HHAT both catalyze attachment of 16-carbon fatty acids (palmitoleate or palmitate, respectively) to secreted signaling proteins, and both enzymes are members of the MBOAT family. We previously showed that the small molecule

Discussion
One of the molecular hallmarks of Wnt proteins is the presence of 22-24 conserved cysteine residues within all Wnt family members. The four cysteines located immediately upstream and downstream of Wnt3a Ser 209 (Cys 203 , Cys 205 , Cys 212 , and Cys 217 ) had been previously been demonstrated to be required for fatty acylation and signaling by Wnt3a (15,18). Moreover, GFP fusions containing 21-and 31-amino acid sequences from Wnt1 that include the four cysteines incorporate alkyne-labeled palmitate in cell-based acylation assays, whereas an 11-amino acid sequence that lacks the outer cysteine pair does not (19). Using synthetic peptides that encompass the sequence surrounding the Ser acylation site, we show that the mere presence of Cys residues in a Wnt substrate is not sufficient, and that disulfide bonding of Cys residues is critical for PORCNmediated acylation. It is likely that this unique two-dimensional structure places constraints on the orientation of Ser 209 so that it can be correctly presented for binding to and acylation by PORCN. The double disulfide-bonded Wnt peptides were extremely technically challenging to synthesize, and as a result, production was time-and labor-intensive and costly. Nonetheless, the 21-residue WT Wnt (S-S) peptide represents a minimal length PORCN substrate for in vitro assays. Further refinement of the geometric requirements within the Wnt peptide substrate could be directed toward replacement of native disulfides with synthetic bridges or disulfide bond mimetics or the use of cyclized peptides.
LGK974 is currently in clinical trials for safety and efficacy in treating Wnt-driven tumors. Although it is marketed as a PORCN inhibitor, the ability of LGK974 to directly inhibit PORCN-mediated Wnt acylation has never been tested. Thus, it remained formally possible that inhibition of Wnt secretion and signaling that occurred with LGK974 treatment in cells could be due to an effect on one or more of the multiple steps required to produce active Wnt ligand, e.g. inhibition of stearoyl-CoA desaturase or binding to Wntless. Here, we exploit the in vitro Wnt acylation assay to provide the first direct evidence that LGK974 is a bona fide PORCN inhibitor whose IC 50 for inhibition of Wnt fatty acylation closely matches that for inhibition of Wnt signaling. Future mechanistic studies can now be carried out to determine how LGK974 inhibits PORCN-mediated catalysis.
Understanding the molecular basis for enzyme-substrate recognition and selectivity remains one of the major challenges for MBOAT proteins in general, and PORCN and HHAT in particular (20). The development of an in vitro Wnt acylation assay now makes it possible to directly compare, side by side, two MBOAT enzymes that both catalyze attachment of 16-carbon fatty acids to secreted proteins. Neither enzyme will accept the other's fatty acyl-CoA or protein (peptide) substrates. These biochemical features explain why targeted disruption of PORCN is not genetically rescued by HHAT, and vice versa. A two-dimensional transmembrane topology map for HHAT reveals the presence of 10 transmembrane domains and 2 reentrant loops (21,22). Although a similar topological analysis of PORCN is not yet available, it is likely to be equally complex. The next step toward understanding how these multipass membrane proteins function will likely require three-dimensional structural information.

Plasmids, cell culture, transfection, and membrane preparation
N-terminal FLAG-tagged Porcn cDNA was generated from a murine Porcupine clone in pcDNA3.1 (a gift from Dr. Joseph Goldstein (UT Southwestern)). Porcn point mutants were generated by site-directed mutagenesis using the QuikChange mutagenesis kit (Agilent Technologies). Super TOP/FOP and pRL-TK plasmids were a kind gift from Dr. Anthony M. C. Brown (Weill Cornell Medical College, New York, NY). HEK 293FT cells (Invitrogen) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 50 units/ml penicillin, 50 g/ml streptomycin, 500 g/ml Geneticin, 1 mM GlutaMAX (Invitrogen), 1 mM sodium pyruvate, and 0.1 mM nonessential amino acids. Cells were transfected with 6 g of Porcn cDNA using Lipofectamine (Invitrogen), split 1:2 the following day, and then grown for 24 h. P100 membranes were prepared from hypotonically lysed cells, following centrifugation at 100,000 ϫ g (P100) as described (14).

Substrates for in vitro fatty acylation assays
IC15:1 was synthesized by the Organic Synthesis Core Lab of Memorial Sloan Kettering Cancer Center, and converted into [ 125 I]IC15:1 using 125 I-NaI (PerkinElmer). Synthesis of [ 125 I]IC15:1 CoA using acyl-coenzyme A synthetase (Sigma catalog number A3352) was carried out as described previously (14). IC16, [ 125 I]IC16, and [ 125 I]IC16 CoA were generated as described (14). C-terminal biotinylated Shh peptide (CGP-GRGFGKR) was prepared as described (14).  (Fremont, CA), and then analyzed by HPLC to assess purity to be greater than 95% pure. To confirm the identity and location of the disulfide bonds, a 1-l aliquot was diluted in 25 l of 100 mM ammonium bicarbonate and subsequently digested with trypsin at a 1:100 enzyme-to-substrate ratio at 37°C for 4 h. The trypsin-digested peptides as well as undigested peptides were further diluted 1:50 with 0.1% formic acid. 2-l aliquots were characterized by high-resolution LC-MS on an Orbitrap mass spectrometer (Thermo Scientific LTQ Orbitrap XL) coupled with a Waters nano-ACQUITY LC via a Proxeon2 nano-electrospray ionization source. Peptides were eluted from a nano-C18 reverse-phase column with a 1-50% binary gradient, 0.1% formic acid (Buffer A) and acetonitrile with 0.1% formic acid (Buffer B), over 90 min at a flow rate of 300 nl/min. MS1 data from m/z 380 -2000 were acquired in profile mode at 60,000 resolution.
The C-terminal biotinylated (Biotin-PEG NovaTag TM ), disulfide-bonded Wnt3a WT peptide mass spectra showed signals for the ϩ3, ϩ4, and ϩ5 charge state with the ϩ4 charge state and observed monoisotopic mass shown (Fig. 1C). As seen in Fig. 1B, there are three possible arrangements of the two disulfide bonds: Cys 203 -Cys 217 and Cys 205 -Cys 212 ; Cys 203 -Cys 205 and Cys 212 -Cys 217 ; and Cys 203 -Cys 212 and Cys 205 -Cys 217 . We took advantage of the lysine residues within the peptide sequence and the location of the disulfide bonds to experimentally confirm the proper orientation of the disulfide bonds. After trypsin digestion, two distinct products will be generated for the desired Cys 203 -Cys 217 (calculated monoisotopic mass (MϩH) ϩ1 ϭ 1117.4703) and Cys 205 -Cys 212 (calculated monoisotopic mass (MϩH) ϩ1 ϭ 1270.5697) linked peptide.

In vitro Wnt and Shh peptide fatty acylation assays
10 g of P100 membranes harvested from Porcupine-expressing HEK 293FT cells were incubated with 100 M Wnt peptide and 30 l of reaction buffer (167 mM MES, pH 6.5, 0.083% Triton X-100, 167 M [ 125 I]IC15:1 CoA) for 1 h at 37°C. For Shh assays, 10 g of P100 membranes harvested from HHAT-expressing HEK 293FT cells were incubated with 100 M Shh peptide, 167 M [ 125 I]IC16 CoA, and 30 l of reaction buffer for 1 h at room temperature (14). After incubation, 400 l of radioimmunoprecipitation assay buffer and 50 l of streptavidin-agarose beads were added, and the mixture was incubated for 1 h at 4°C with continuous mixing. Biotinylated peptides were pelleted by centrifugation at 1000 ϫ g for 5 min. Pellets were washed three times with 500 l of radioimmunoprecipitation assay buffer. [ 125 I]IC15:1 or [ 125 I]IC16 incorporation was measured in a PerkinElmer ␥ counter.

Wnt signaling activity assays
To measure paracrine Wnt signaling activity, HEK293FT cells were transfected with 3 g of Super TopFlash or Super FOP and 0.3 g of pRL-TK, and 24 h after transfection, cells were co-cultured in 12-well plates with L-Wnt3a cells transfected with 6 g of Porcn at a 3:1 ratio (L-Wnt3a:293FT) in the presence of the indicated drugs for 24 h. Wnt3a pathway activity was detected using a Dual-Luciferase reporter assay (Promega) and recorded as relative luciferase units using a BioTek Synergy TM H1 microplate reader.