Structure

Short Article Structural Insights into the Inhibition of Wnt Signaling by Cancer Antigen 5T4/Wnt-Activated Inhibitory Factor 1 Yuguang Zhao,1,3 Tomas Malinauskas,1,2,3 Karl Harlos,1 and E. Yvonne Jones1,* 1Division

of Structural Biology, Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, OX3 7BN, UK address: Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724, USA 3These authors contributed equally to this work *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2014.01.009 This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/). 2Present

SUMMARY

The tumor antigen 5T4/WAIF1 (Wnt-activated inhibitory factor 1; also known as Trophoblast glycoprotein TPBG) is a cell surface protein targeted in multiple cancer immunotherapy clinical trials. Recently, it has been shown that 5T4/WAIF1 inhibits Wnt/ b-catenin signaling, a signaling system central to many developmental and pathological processes. Wnt/b-catenin signaling is controlled by multiple inhibitors and activators. Here, we report crystal structures for the extracellular domain of 5T4/WAIF1 at 1.8 A˚ resolution. They reveal a highly glycosylated, rigid core, comprising eight leucine-rich repeats (LRRs), which serves as a platform to present evolutionarily conserved surface residues in the N-terminal LRR1. Structural and cell-based analyses, coupled with previously reported in vivo data, suggest that Tyr325 plus the LRR1 surface centered on a second exposed aromatic residue, Phe97, are essential for inhibition of Wnt/b-catenin signaling. These results provide a structural basis for the development of 5T4/WAIF1-targeted therapies that preserve or block 5T4/WAIF1-mediated inhibition of Wnt/b-catenin signaling.

INTRODUCTION 5T4/WAIF1 (also known as trophoblast glycoprotein, TPBG; 5T4 oncofetal trophoblast glycoprotein; and Wnt-activated inhibitory factor 1, WAIF1) is a vertebrate-specific, single-pass transmembrane protein first identified in human placental tissues (Hole and Stern, 1988). 5T4/WAIF1 is rarely expressed in normal adult tissues, but is present at high levels in placenta and in most common tumors, typically more than 80% of carcinomas of the kidney, breast, colon, prostate, and ovary (Hole and Stern, 1988; Southall et al., 1990; Starzynska et al., 1994). Thus, 5T4/ WAIF1 has the characteristics of an oncofetal antigen, highlighting it as a possible candidate for use as a diagnostic marker or target for cancer treatment. A modified vaccinia virus Ankara (MVA) encoding human 5T4/WAIF1 (designated TroVax or MVA612 Structure 22, 612–620, April 8, 2014 ª2014 The Authors

5T4) induced an antitumor immune response in mouse cancer models (Woods et al., 2002; Mulryan et al., 2002) and has been evaluated in clinical trials targeting colorectal cancer, renal cell carcinoma, and refractory prostate cancer (Kim et al., 2010). For the most comprehensive phase 3 clinical study of MVA5T4 carried out to date (in 733 patients with metastatic renal cancer), no difference was observed in survival for the overall study population, but analyses suggested that subgroups of patients could benefit (Amato et al., 2010). 5T4/WAIF1-targeted antibody-based therapies are also under active development (Boghaert et al., 2008; Sapra et al., 2013). Despite the potential therapeutic value of 5T4/WAIF1 as a target in oncology, knowledge of the function and molecular mechanism of action of this cell surface molecule remains sparse. 5T4/WAIF1 is predicted to contain multiple leucine rich-repeats (LRRs) in the extracellular domain (Myers et al., 1994), a transmembrane helix and a cytoplasmic region (Figure 1A; Figure S1 available online). The cytoplasmic PDZ-binding motif Ser-Asp-Val of 5T4/WAIF1 has been reported to interact with the PDZ domain of TIP-2/GIPC (Awan et al., 2002), a cytoplasmic protein that associates with vesicles located near the cell membrane (De Vries et al., 1998). Further downstream mechanisms of signal transduction remain unknown. Recently, 5T4/WAIF1 has been found to inhibit the Wnt/b-catenin signaling pathway (Kagermeier-Schenk et al., 2011), a key pathway in embryonic development and a major target for anticancer therapeutics (MacDonald et al., 2009; Polakis, 2012). KagermeierSchenk and colleagues (Kagermeier-Schenk et al., 2011) reported that one of three 5T4/WAIF1 paralogs in zebrafish, 5T4/Waif1a, co-immunoprecipitated with the Wnt coreceptor LRP6. However, the extracellular domains of 5T4/Waif1a and LRP6 did not interact directly; this indirect interaction appears to require colocalization of both partners in the membrane and/or bridging by an unknown molecule. 5T4/WAIF1 is heavily glycosylated (Hole and Stern, 1988) posing challenges for production in standard Escherichia colibased protein overexpression systems. To date, biophysical studies of 5T4/WAIF1 are lacking. As a first step in elucidating the molecular mechanisms governing 5T4/WAIF1 function and its role in Wnt/b-catenin signaling, we established protocols for the production of the soluble extracellular domain of 5T4/ WAIF1 (5T4/WAIF1Ecto) using a human embryonic kidney 293 (HEK293) cell-based stable expression system. We determined three structures of the extracellular domain of 5T4/WAIF1Ecto

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

from two crystal forms at 1.75–1.77 A˚ resolution. Our structural, evolutionary, and cell-based analyses, coupled to previously reported in vivo studies of two zebrafish 5T4/Waif1 paralogs, 5T4/Waif1a and 5T4/Waif1c, reveal key residues of 5T4/WAIF1 that are essential for inhibition of the Wnt/b-catenin signaling pathway.

0.17 A˚ for 218 equivalent Ca atoms from crystal 1 and crystal 2, chain A; 0.25 A˚ (231 atoms) for crystal 1 and crystal 2, chain B (Figure S3). LRRs 7–8 provide a platform onto which the CTcap is integrated and contain additional elements (an a helix C300–E311 at the C terminus of LRR 8; Figure 1C). Finally, both N- and C-terminal caps contain two disulfide bonds each that further stabilize the seamless LRR core of 5T4/WAIF1Ecto.

RESULTS AND DISCUSSION Production of the Extracellular Domain of 5T4/WAIF1 Human 5T4/WAIF1 contains 420 amino acid residues and is a single-pass transmembrane protein (Figures 1A and S1). Sequence analysis predicts a domain structure comprising a secretion signal (residues 1–31), serine-rich motif (32–53), LRRs (60–344), a transmembrane region (356–376), and a short cytoplasmic tail (377–420; Figure S1). We expressed a full-length LRR region of 5T4/WAIF1 (D60–D345, 5T4/WAIF1Ecto) with a C-terminal rhodopsin 1D4 tag (Hodges et al., 1988) as a secreted construct in HEK293 GnTI( ) cells (Reeves et al., 2002; Aricescu et al., 2006; Zhao et al., 2011). A stable cell line was generated by using a PhiC31 integrase-assisted vector with puromycin selection, pURD (see Experimental Procedures). We purified 5T4/ WAIF1Ecto using antibody-affinity chromatography followed by size exclusion chromatography. Multi-angle light scattering (MALS) analysis indicated that the purified 5T4/WAIF1Ecto was monomeric in solution (Figure S2A). Crystal Structures of 5T4/WAIF1Ecto Reveal the LRR Core Decorated with Loops and Glycans Enzymatically deglycosylated 5T4/WAIF1Ecto crystallized in two crystal forms, 1 and 2, with one and two molecules in the asymmetric unit, respectively (Table 1). Initial crystallographic phases were calculated by molecular replacement using LRRs of Netrin G ligand 1 (NGL1; Protein Data Bank [PDB] ID 3ZYJ; Seiradake et al., 2011) as a search model. Final models from crystal forms 1 and 2 were refined to an Rwork/Rfree of 18.2%/22.2% and 17.7%/21.7% at 1.75 A˚ and 1.77 A˚ resolution, respectively. All three structures of 5T4/WAIF1Ecto reveal a compact core of eight LRRs. The LRRs build-up a right-handed slightly twisted solenoid in which the short b strands are arranged like a ladder in the concave face of 5T4/WAIF1Ecto (Figures 1B and 1C). The eight LRRs are flanked by two, N- and C-terminal, cysteinerich capping modules (NT-cap and CT-cap, respectively) that complete the overall structural framework of 5T4/WAIF1Ecto. The eight LRRs could be classified into three distinct groups (Figure 1C). LRRs 1–3 and 6 contain a sequence motif Lxx(x)V/ LxxLxLxxxxLxxL/VxxxxF (where x is any amino acid). In LRRs 1–3, the motif provides a scaffold for a rigid hydrophobic core where three consecutive phenylalanine residues (F111, F138, and F162) rest on the continuous bed of leucines (Figure 1C). LRRs 4 and 5 form a second distinct group because they contain a buried phenylalanine at different positions compared to LRRs 1–3, and 6. A flexible loop (N166–P172) that links LRR 3 to LRR 4 is ordered only in one out of the three 5T4/WAIF1Ecto structures, crystal 2, chain B (Figure S3). Similarly, a flexible insertion (P185–S194) with a short 310 helix is ordered only in crystal 1 (Figures 1B and S3). Other than these two loops, the three 5T4/WAIF1Ecto molecules are essentially identical in the two crystal forms with root mean square deviation (rmsd) of

A Potential 5T4/WAIF1-Protein Interaction Hotspot We investigated the surface characteristics of 5T4/WAIF1Ecto to delineate the areas that had properties indicative of interaction sites. Specific protein-protein interactions are rarely mediated by flexible, heterogeneous glycans. 5T4/WAIF1Ecto contains seven predicted asparagine-linked glycosylation sites at N81, N124, N166, N192, N243, N256, and N275 (Figures 1A and S1). Electron density corresponding to five of these seven potential glycans is visible in crystal 1 (Figure 1B). The glycan on N166 is not visible because it is located in a disordered loop (N166– P172); however, fragmentary electron density proximal to N166 is visible in crystal 2, chain B, where loop N166–P172 is stabilized by a neighboring molecule in the crystal lattice. The glycan on N192 is not visible, possibly because of flexibility as the putative site is exposed in solvent channels for both crystal forms. Thus, our crystal structures suggest that at least five of the putative N-linked glycosylation sites of 5T4/WAIF1Ecto are glycosylated. These five glycosylated asparagines therefore flag positions on the ectodomain structure that are less likely to mediate 5T4/ WAIF1-protein interactions. The concave face of 5T4/WAIF1Ecto contains an extended, negatively charged region that spans five LRRs: E177 in LRR4; E216, LRR5; D240 LRR6; E266 and D267, LRR7; and D292, LRR8 (Figure 1D). This region is bordered by the glycosylated N124 on LRR2 and an array of positively charged residues: R213, R214, R237, H238, H264, and R288. These residues bind sulfate ions in crystal form 2 (Figure 1E). Notably, similar positively charged surfaces mediate interactions among Wnts, their inhibitors, and heparan sulfate proteoglycans (HSPG; Malinauskas et al., 2011), suggesting that 5T4/WAIF1 interactions with its binding partners could be modulated by HSPGs. 5T4/ WAIF1Ecto did not bind to a heparin column (Figure S2B), consistent with the calculated isoelectric point of 5.36, arguing against 5T4-heparan sulfate interactions. However, this finding does not exclude the possibility that charged regions of 5T4/WAIF1 could interact with HSPGs and/or other proteins. Protein-protein interactions are frequently mediated by rigid, larger than 600 A˚2 surfaces enriched in aromatic, evolutionarily conserved residues (Moreira et al., 2007). 5T4/WAIF1Ecto has five aromatic side chains (F97, Y224, Y251, F290, and Y325) exposed on its surface (Figure 1F). We further investigated which surface exposed residues of 5T4/WAIF1 are conserved across 5T4/WAIF1 family members in vertebrates. Notably, a cluster of evolutionarily conserved residues (E67, S69, and K76) is centered on F97 of LRR1 (Figure 1F). Thus F97 represents a potential ‘‘hot spot’’ that might mediate 5T4/WAIF1-protein interactions and/or might be important for the inhibition of Wnt/b-catenin signaling by 5T4/WAIF1. Similarly, the aromatic character of Y325 is conserved, and mammalian 5T4/WAIF1 sequences and zebrafish Waif1a contain either tyrosine or phenylalanine at this position. Structure 22, 612–620, April 8, 2014 ª2014 The Authors 613

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

(legend on next page)

614 Structure 22, 612–620, April 8, 2014 ª2014 The Authors

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

Comparison of 5T4/WAIF1Ecto to Structurally and Functionally Related Proteins In a second approach to identify where the potential ligandbinding regions of 5T4/WAIF1 are located, we compared 5T4/ WAIF1Ecto to other structures of LRR-containing proteins and their complexes. We used the secondary structure matching program PDBeFold (Krissinel and Henrick, 2004) for protein structure alignment to identify the closest structural homologs of 5T4/WAIF1Ecto. The most similar protein is Netrin G ligand 1 (NGL1; PDB ID 3ZYJ, chain A), which binds to Netrin G1 via its concave surface (Figure 1G; Seiradake et al., 2011). 5T4/ WAIF1Ecto and NGL1 could be superimposed based on 222 equivalent Ca atoms with an rmsd of 2.0 A˚, highlighting the overall conservation of the capped LRR fold. Despite this structural similarity, the sequence identity between 5T4/WAIF1 and NGL1 is low (10%) and provides no evidence for any similarity in protein-binding mode. A second structurally similar protein, glycoprotein 1ba (PDB ID 1P9A, chain G; rmsd 2.3 A˚ for 225 equivalent Ca atoms), recognizes its ligand, thrombin, using a combination of flexible C-terminal tail and two LRRs (Celikel et al., 2003). Glycoprotein 1ba wraps around one side of another ligand, von Willebrand factor A1 domain, using N- and C-terminal b-fingers that are absent in 5T4/WAIF1Ecto (Huizinga et al., 2002). 5T4/WAIF1Ecto lacks surfaces or sequence motifs that are present in other extracellular inhibitors or receptors of the Wnt signaling pathway. It does not have the surface-exposed NxI motif that mediates interactions between the Wnt co-receptor LRP6 and two inhibitors, Dickkopf and Sclerostin (Bourhis et al., 2011). There is an absence of hydrophobic grooves (present in Wnt8 receptor Frizzled8, Janda et al., 2012) or pockets (present in the Wnt inhibitory factor 1; Malinauskas, 2008; Malinauskas et al., 2011) that could mediate 5T4/WAIF1Ecto inhibitory function by sequestering Wnt-linked palmitoleyl moieties. The only other inhibitor of Wnt signaling that has a LRR fold is Tsukushi, which functions as a Wnt signaling inhibitor by competing with Wnt protein for binding to the receptor Frizzled4 (Ohta et al., 2011). However, the sequence identity between 5T4/WAIF1 and Tsukushi is relatively low, 20%, suggesting that the mode of Wnt inhibition may be different. Taken together, our comparison of 5T4/WAIF1 to other LRR-containing protein recognition modules and inhibitors of Wnt signaling suggests that 5T4/WAIF1 is a

unique antagonist of Wnt signaling. Our analysis of evolutionarily conserved surface residues (see previous section) had identified a putative protein-binding hotspot; therefore, we focused further analyses onto this region, which encompasses the NT-cap and LRR1. It contains the surface exposed side chains of S69, E67, T74, K76, N95, and, most notably, F97 (Figures 2A and 2B). Lys76 and Phe97 of 5T4/WAIF1 Are Essential for the Inhibition of Wnt Signaling We used a cell-based Wnt/b-catenin signaling assay (DasGupta et al., 2005) to investigate the role of the conserved surface residues clustered around F97 of human 5T4/WAIF1 (Figure 2). Based on our crystal structures, we selected residues for sitedirected mutagenesis that have their side chains exposed on the exterior of 5T4/WAIF1 and are therefore less likely to disturb the overall structural integrity (Figure 2A). Most full-length, wildtype and mutant, 5T4/WAIF1 constructs were trafficked to the cell membrane as assessed by fluorescence microscopy (Figure S4A), except two variants, N124Q and R214E, which were not secreted as 5T4/WAIF1Ecto constructs. Wild-type and mutant extracellular domains were expressed and secreted at similar levels (except N124Q and R214E) as assessed by western blot and FACS analyses (Figures S4B and S4C), consistent with them all being folded, glycosylated, and passing cellular quality control systems (Trombetta and Parodi, 2003). Consistent with previous studies (Kagermeier-Schenk et al., 2011), conditioned media containing secreted mouse Wnt3a activated the Wnt/b-catenin signaling pathway in HEK293T cells, and this signaling was inhibited both by the Wnt antagonist Dickkopf (a positive control) and wild-type 5T4/WAIF1 (Figures 2C and S5). Combined mutations of two residues, N338A and A340G, exposed at the CT-cap did not have an effect on the inhibitory function of 5T4/WAIF1 (Figures 2A and 2C). Similarly, combined mutations E189R, R190A, and Q191A in an evolutionarily divergent region, the flexible insertion in LRR4 (Figures 1B, 1C, 2A, and 2C), did not affect the 5T4/WAIF1-mediated inhibition. We investigated two sulfate ion-contacting residues, R214 and E261, using charge reversal mutations, R214E and E261R, respectively (Figure 1E). E261R did not have an effect on the Wnt-inhibitory function of 5T4/WAIF1 (Figures 2C and S5), whereas R214E abolished secretion of 5T4/WAIF1Ecto R214E

Figure 1. Crystal Structure of the Human 5T4/WAIF1Ecto (A) Domain organization of human 5T4/WAIF1. Seven glycosylation sites are marked with hexagons. The 5T4/WAIF1Ecto is colored in blue-to-red transition; serine-rich and transmembrane regions are gray. (B) Ribbon diagram of 5T4/WAIF1Ecto in two views that differ by a 90 rotation around a vertical axis. 5T4/WAIF1Ecto is colored as in (A). Asn-linked N-acetylglucosamines are shown as magenta sticks. Disulfide bonds are shown as gray connected spheres. (C) The structure-based alignment of LRRs of 5T4/WAIF1Ecto reveals repetitive patterns of leucines (or similar hydrophobic residues, valines, and isoleucine) that build up the framework of 5T4/WAIF1Ecto. LRRs 1–3 and LRR6 form a distinctive group; each of these LRRs contains a buried phenylalanine (pink in the bottom panel), which contributes to the tightly packed hydrophobic core of 5T4/WAIF1Ecto. On top of the LRR-based core, the architecture of 5T4/WAIF1Ecto is further stabilized by hydrogen bonding patterns between multiple, three residue-long b strands (highlighted in gray background). (D) Electrostatic properties of 5T4/WAIF1Ecto. The protein is shown as solvent-accessible surface colored by electrostatic potential at ± 5 kT/e (red, acidic; blue, basic). The orientation of 5T4/WAIF1Ecto on the left side is the same as in (B, left side). Glycan moieties are shown as yellow sticks. Charged residues discussed in the text are indicated. (E) A highly charged, sulfate-binding region on LRRs 5–8, colored as in (B). Side chains of sulfate-binding residues are labeled and shown as sticks (carbons, gray to orange; nitrogen, blue; sulfur, yellow; oxygen, red). Distances between atoms are shown in angstroms. (F) The surface of 5T4/WAIF1Ecto is colored by residue conservation (conserved, magenta; variable, cyan). Sequences of 45 members of the 5T4/WAIF1 family were included in the sequence conservation analysis. Five surface-exposed aromatic residues are indicated; only one, F97, is evolutionarily conserved. (G) Superposition of 5T4/WAIF1Ecto (green) onto Netrin G ligand 1 (NGL1; blue) in complex with Netrin G1 (gray surface) illustrates the potential of the concave face of LRRs to recognize LRR-binding proteins.

Structure 22, 612–620, April 8, 2014 ª2014 The Authors 615

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

Table 1. X-Ray Crystallography Data Collection and Refinement Statistics 5T4/WAIF1Ecto (Crystal Form 1)

5T4/WAIF1Ecto (Crystal Form 2)

Data Collection X-ray source

Diamond Light Diamond Light Source, beamline I02 Source, beamline I02

Wavelength (A˚)

1.06

1.06

Resolution range (A˚)

49.31–1.75 (1.80–1.75)

54.33–1.77 (1.82–1.77)

Space group

P 21 21 21

P 1 21 1

Unit cell a, b, c (A˚)

49.31, 67.31, 96.38

49.6, 95.82, 65.97

a, b, g ( )

90, 90, 90

90, 91.14, 90

Total reflections

317,855

300,698 (21,206)

Unique reflections

33,062

59,830 (4,362)

Multiplicity

9.6 (9.3)

5.0 (4.9)

Completeness (%)

99.8 (99.2)

99.7 (99.5)

Mean I/sigma (I)

17.9 (3.1)

13.8 (2.5)

Rmerge (all I+ and I ) (%)a 6.6 (81.3)

6.3 (76.4)

Refinement Resolution range (A˚)

43.94–1.75 (1.80–1.75)

54.33–1.77 (1.82–1.77)

Number of reflections, work/test set

31,330/1,674

56,784/3,022

Number of atoms (protein/ligands, glycans/waters)

2,493 (2,239/ 130/124)

4,712 (4,276/ 257/179)

Protein residues

283

552

Mean B-factors (protein/ligands, glycans/waters) (A˚2)

36 (35/55/40)

32 (32/52/37)

Rwork (%)

18.2 (27.9)

17.7 (24.5)

Rfree (%)

22.2 (29.4)

21.7 (29.8)

Rmsd from ideal values (bonds) (A˚)

0.016

0.016

Rmsd from ideal values (angles,  )

1.8

1.9

Ramachandran plotb Favored regions (%)

95.7

95.4

Outliers (%)

0.0

0.2

Values for the highest resolution shell are shown in parentheses. a As defined in Aimless (Evans and Murshudov, 2013). b As defined in MolProbity (Chen et al., 2010).

(Figure S4B), suggesting aberrant trafficking of this construct within the cell. Consistent with these observations, 5T4/WAIF1 R214E did not inhibit Wnt signaling (Figures 2C and S5). We note that positive charge at the position corresponding to R214 in human 5T4/WAIF1 is relatively conserved across species (Arg/Lys/His in 11 of 15 species; Figure S1), suggesting that the charge characteristics of this position play a role in the folding and trafficking of 5T4/WAIF1. Similarly to R214E, mutation of glycosylated N124 to glutamine impaired trafficking to the cell surface and Wntinhibitory function of full-length 5T4/WAIF1 N124Q (Figures 616 Structure 22, 612–620, April 8, 2014 ª2014 The Authors

S4A and 2C, respectively). Consistent with the behavior of the full-length protein, 5T4/WAIF1Ecto N124Q was not secreted (Figure S4B). Mutation of the highly conserved phenylalanine residue, F97, to threonine (F97T) significantly (p = 0.0002) reduced the Wntinhibitory function of human 5T4/WAIF1 (Figure 2C). The side chain of F97 is sandwiched between the glycosylated N124 and evolutionarily conserved K76 (Figure 2B). The juxtaposition of F97 and K76 could provide further stabilization to this region because of cation-p interactions between the -NH3+ group of K76 and the adjacent delocalized electrons of F97, respectively (Gallivan and Dougherty, 1999). The average distance between the nitrogen of the K76 side chain and the carbon atoms of the F97 ring is 5.4 A˚ (SD, 0.3 A˚) in the two crystal forms. Indeed, 5T4/WAIF1 mutant K76A inhibited Wnt signaling significantly (p < 0.0001) less efficiently compared to wild-type 5T4/WAIF1 in our cell-based assay (Figure 2C). Mutation of the less conserved N95 to alanine did not influence the Wnt-inhibitory function. Interestingly, a similar lack of effect on function was observed for a mutation, T74V, of the conserved T74 (Figure 2C). Thus, while sensitive to mutation of F97 and K76, the inhibitory function of 5T4/WAIF1 appears relatively robust to mutation of neighboring residues (Figures 2B and 2C). Mutants E67A and S69A also showed less marked effects on Wnt-inhibitory function than F97T and K76A (Figure 2C). In addition, we tested the effect of mutation Y325A on the Wnt-inhibitory function of 5T4/WAIF1. Multiple vertebrates contain phenylalanine at the corresponding position, including zebrafish Wnt-inhibitory Waif1a (F274; zebrafish noninhibitory Waif1c contains S283 at the equivalent position; Figure S1). 5T4/WAIF1Ecto Y325A was expressed and secreted at a similar level compared to wild-type 5T4/WAIF1Ecto (Figure S4). The fulllength 5T4/WAIF1 Y325A did not inhibit Wnt3a signaling in the cell-based assay (Figures 2C and S5), suggesting that, in addition to the region around F97 and K76, other surfaces of 5T4/ WAIF1 may be important for the inhibition of Wnt signaling. Notably, the surface proximal to Y325 is not obstructed by Asn-linked glycosylation and thus may be accessible for the interactions between 5T4/WAIF1 and currently unknown binding partners. Our results suggest that residues K76 and F97 are essential for the inhibition of the Wnt/b-catenin signaling pathway by human 5T4/WAIF1. The previously reported findings for zebrafish paralogs 5T4/Waif1a and 5T4/Waif1c (Kagermeier-Schenk et al., 2011) shed new light when revisited in the context of our data. Whereas 5T4/Waif1a inhibited Wnt/b-catenin signaling, 5T4/Waif1c did not. A chimera comprising the 5T4/ Waif1a ectodomain plus 5T4/Waif1c cytoplasmic domain inhibited Wnt signaling, whereas the 5T4/Waif1c ectodomain plus 5T4/Waif1a transmembrane and cytoplasmic regions did not inhibit. These results indicated that the 5T4/Waif1 cytoplasmic region is dispensable for Wnt signaling inhibition; however, they also raise the question as to which differences between the ectodomains of 5T4/Waif1a and 5T4/Waif1c are responsible for the difference in inhibitory function. We therefore analyzed the ectodomain sequences (Figure 2D). We found that the Wnt-inhibitory 5T4/Waif1a in zebrafish has K40, F61, and F274 at the positions corresponding to K76, F97, and Y325 in human 5T4/WAIF1. In contrast, the noninhibitory 5T4/Waif1c

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

Figure 2. An Evolutionarily Conserved Region of 5T4/WAIF1Ecto Is Essential for the Inhibition of the Wnt Signaling Pathway (A) The surface of the 5T4/WAIF1Ecto is colored by residue conservation as in Figure 1F. Residues investigated in the Wnt-responsive cell-based assay are indicated. (B) A surface-exposed region that mediates the Wnt-inhibitory function of 5T4/WAIF1. b Strands of 5T4/WAIF1Ecto are colored as in Figure 1B, numbering corresponds to Figure S1. Nitrogen and oxygen atoms of selected side chains and glycans are shown in blue and red, respectively. Disulfide bonds are shown as gray connected spheres. (C) The inhibition of Wnt3a signaling in a cellular assay by wild-type and mutant constructs of 5T4/ WAIF1 and the Wnt inhibitor Dickkopf. The signaling was induced using conditioned media containing secreted mouse Wnt3a. Wnt3a signaling was inhibited by wild-type 5T4/WAIF1 and Dickkopf, but it was significantly less inhibited by 5T4/WAIF1 mutant constructs K76A and F97N. The experiment was repeated three times (results from experiment 1 are shown here, from experiments 2 and 3; Figure S5), each time in quadruplicate, and error bars show SD. The p values were calculated using paired t test are shown for the wild-type K76A, F97T, and Y325A pairs. The other mutant constructs of 5T4/WAIF1 did not show significant inhibition of Wnt signaling as suggested by p values (p > 0.05). Columns corresponding to two mutant constructs of 5T4/WAIF1 that exhibited impaired trafficking to the cell surface (N124Q and R214E; Figure S4) are shown in gray. (D) Amino acid sequence alignment of the N-terminal regions of the 5T4/WAIF1 family members. K76 and F97, which are essential for the Wntinhibitory function of human 5T4/WAIF1 (Figure 2C), are marked with red stars. Corresponding residues N52 and N73 in the noninhibitory 5T4/ Waif1c from zebrafish are framed in black. Boundaries between the NT-cap, LRR1, and LRR2 are shown below the alignment. An alignment of the full-length 5T4/WAIF1 proteins is presented in Figure S1.

has N52, N73, and S283 at the corresponding positions, further supporting the notion that these surface-exposed residues play a key role in the inhibition of Wnt/b-catenin signaling by 5T4/ WAIF1 family members. Future Directions and Implications for Cancer Therapies Our structural, evolutionary, and cell-based analyses, combined with previously reported in vivo studies in zebrafish (KagermeierSchenk et al., 2011), suggest that the conserved surface centered on K76 and F97, plus a second area involving Y325, is essential for the inhibition of the Wnt/b-catenin signaling pathway by human 5T4/WAIF1. However, the proteins that bind to these surfaces remain to be identified. The protocols we report here for the production of wild-type and mutant 5T4/ WAIF1 proteins could provide useful reagents for the identification and validation of 5T4/WAIF1 binding partners (cf. the strategy of So¨llner and Wright, 2009), to add further intriguing insights into the varied mechanisms by which Wnt signaling can be modulated. Our studies detail the specific atomic-level features

that link 5T4/WAIF1, a tumor antigen that is the focus of multiple late phase clinical trials, with the Wnt signaling pathway, a wellestablished but challenging target in oncology (Polakis, 2012). These data provide a structural basis for the assessment of current therapeutic anti-5T4/WAIF1 antibodies (Boghaert et al., 2008; Sapra et al., 2013) in terms of their ability to preserve or block 5T4/WAIF1-mediated inhibition of Wnt/b-catenin signaling with consequent benefits and risks. EXPERIMENTAL PROCEDURES Protein Production We cloned human 5T4/WAIF1 (UniProtKB/Swiss-Prot Q13641) extracellular domain (residues D60–D345) construct 5T4/WAIF1Ecto from the IMAGE clone 438906 (Source BioScience) into a stable cell line generation vector pURD that we had recently engineered in-house (Figure S6). HEK293S GnTI( ) cells (Reeves et al., 2002) were cotransfected with pURD-5T4/WAIF1Ecto and a PhiC31 integrase expression vector (pgk-phiC31). The polyclonal cell population resulting from puromycin (2 mg/ml) selection was used for the protein production. The secreted 5T4/WAIF1Ecto protein has additional amino acids

Structure 22, 612–620, April 8, 2014 ª2014 The Authors 617

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

from the cloning vector: N-terminal ETG and C-terminal GTETSQVAPA sequences (last nine residues are Rhodopsin 1D4 tag; Molday and MacKenzie, 1983). For protein purification, the conditioned media were passed through the anti-1D4 tag antibody (University of British Columbia) covalently linked to Sepharose beads at 4 C (CNBr-activated Sepharose 4 Fast Flow, GE Healthcare) and eluted with 1D4 peptide. For deglycosylation, 5T4/WAIF1Ecto was treated with endo-b-N-acetylglucosaminidase F1 (37 C, 1 hr) and further purified by size-exclusion chromatography (Superdex 200 16/60 column, GE Healthcare) in 20 mM Tris-HCl, pH 7.5, and 150 mM NaCl. Crystallization and Structure Determination 5T4/WAIF1Ecto was concentrated to 5.5 mg/ml and crystallized using the sitting drop, vapor diffusion method at 21 C (Walter et al., 2005). The crystallization drop contained 100 nl of concentrated 5T4/WAIF1Ecto, 100 nl of 25% w/v polyethylene glycol (PEG) 3350, 0.1 M citrate pH 3.5, and 0.5 nl of 0.1 M NaOH (crystal form 1). For crystal form 2, the crystallization drop contained 100 nl of concentrated 5T4/WAIF1Ecto plus 100 nl of 0.2 M (NH4)2SO4, 30% w/v PEG 4000. Crystals were flash-frozen by immersion into a reservoir solution supplemented with 25% v/v glycerol followed by transfer to liquid nitrogen. The crystals were kept at 173 C during X-ray diffraction data collection. Diffraction data were indexed and integrated using XDS (Kabsch, 2010), and scaled and merged using Aimless (Evans and Murshudov, 2013). Initial X-ray phases were determined by molecular replacement using the NGL1 structure (PDB ID 3ZYJ; Seiradake et al., 2011) as a search model in Phaser (McCoy et al., 2007). Structures were built using Buccaneer (Cowtan, 2008), PHENIX AutoBuild (Terwilliger et al., 2008), further refined in REFMAC (Murshudov et al., 1997), validated using MolProbity (Chen et al., 2010), and analyzed using ConSurf (Glaser et al., 2003) and the PyMOL Molecular Graphics System (Schro¨dinger). Data collection and refinement statistics are shown in Table 1. Multi-Angle Light Scattering Analysis MALS analysis of 5T4/WAIF1Ecto was performed during analytical size exclusion chromatography using Superdex 75 10/300 GL column (GE Healthcare Life Sciences) connected to static light-scattering (DAWN HELEOS II, Wyatt Technology), differential refractive index (Optilab rEX, Wyatt Technology), and Agilent 1200 UV (Agilent Technologies) detectors. Purified 5T4/WAIF1Ecto (2.24 mg/ml, 0.1 ml) was loaded onto a column equilibrated in 150 mM NaCl, 10 mM HEPES, at pH 7.5. Data were analyzed using the ASTRA software package (Wyatt Technology). Heparin Affinity Chromatography Purified 5T4/WAIF1Ecto (0.358 mg/ml, 0.69 ml) was loaded onto a 5 ml HiTrap heparin HP column (GE Healthcare Life Sciences) equilibrated in 40 mM NaCl, 10 mM HEPES, and pH 7.5. Elution of 5T4/WAIF1Ecto (flow rate 2 ml/min) was followed by absorption at 280 nm. A linear NaCl gradient, from 40 mM NaCl, 10 mM HEPES, pH 7.5 to 1 M NaCl, 10 mM HEPES, pH 7.5 over ten column volumes was applied but 5T4/WAIF1Ecto did not bind to the column and was eluted with 40 mM NaCl, 10 mM HEPES, at pH 7.5. Cellular Assays for Wnt Signaling Wnt3a-responsive cell-based assays were carried out using HEK293T cells in quadruplicate (in 96-well plates, 105 cells/well). Cells were cotransfected with a SuperTopFlash firefly luciferase plasmid (300 ng/well; DasGupta et al., 2005), a constitutive Renilla luciferase plasmid (pRL-TK from Promega 100 ng/well), and a 5T4/WAIF1 wild-type (residues 1–420, cloned into the pneo-sec vector via EcoR1 and Xho1 sites; pneo-sec is similar to pHLsec [Aricescu et al., 2006], but it contains a neomycin/kanamycin resistance selection marker) or a mutation variant or Dkk plasmid (Chen et al., 2011; 300 ng/well). Lipofectamine 2000 (Invitrogen) was used according to the manufacturer’s recommendations. Sixteen hours after transfection, the media were replaced with the conditioned media from the mouse L-Wnt3a cell line for Wnt signaling induction. The firefly and Renilla luciferase activities were measured 24 hr later with the Dual-Glo luciferase reporter assay system (Promega) using an Ascent Lunimoskan luminometer (Labsystems). The enzymatic activity of firefly luciferase was normalized compared to constitutive Renilla luciferase activity.

618 Structure 22, 612–620, April 8, 2014 ª2014 The Authors

Expression Analysis of 5T4/WAIF1 Constructs For western blot analyses of secreted 5T4/WAIF1 constructs, residues 32–355 were cloned into a modified pHLsec vector with a C-terminal 1D4 tag and expressed in HEK293T cells for 2 days. 5T4/WAIF1 samples (cell lysates and conditioned media) were reduced with dithiothreitol and heated for 5 min at 100 C before loading on the SDS-PAGE gels. Western blots were probed with mouse anti-1D4 tag (University of British Columbia), anti-His tag (Penta His, QIAGEN), anti-a-tubulin (Sigma), and anti-b-actin (Abcam) antibodies, followed by goat anti-mouse IgG-horseradish peroxidise (HRP) conjugate (Sigma). HRP was detected using the Amersham ECL western blotting detection kit (GE Healthcare Life Sciences). For the expression analysis of human 5T4/WAIF1 on the cell surface, wild-type and mutant constructs (residues 32–420) were cloned into a modified pHLsec vector (pHLsec-mVenus), which has a C-terminal monomeric fluorescent protein mVenus (Nagai et al., 2002) tag (a gift from A. Clayton and A. R. Aricescu, Oxford) and expressed in COS-7 cells for 2 days. Images of the cells were taken using the Leica SP8-X SMD inverted microscope. Flow cytometry analysis of 5T4/WAIF1 tagged with a fluorescent monoVenus protein was performed in HEK293T cells. The expression of monoVenus was monitored via flow cytometry on a CyAn ADP Analyzer (Beckman Coulter). The argon-ion laser was tuned to 488 nm with 100 mW of power, and mono-Venus fluorescence detected in FL1 through a 530/ 40-nm bandpass filter. Data analysis was performed using FlowJo software (Tree Star).

ACCESSION NUMBERS The PDB accession numbers for the structure factors and coordinates corresponding to the crystal forms 1 and 2 of 5T4/WAIF1Ecto are 4CNM and 4CNC, respectively.

SUPPLEMENTAL INFORMATION Supplemental Information includes six figures and can be found with this article online at http://dx.doi.org/10.1016/j.str.2014.01.009. ACKNOWLEDGMENTS We thank Diamond Light Source for beamtime (proposal mx8423) and the staff of beamline I02 for assistance with crystal testing and data collection. We are grateful to Thomas Walter for assistance with crystallization, Benjamin Davies for the PhiC31 integrase expression vector, Keith Morris for confocal imaging, Veronica T. Chang for flow cytometry analysis, and members of the Division of Structural Biology for helpful discussion and assistance. This work was funded by Cancer Research UK, the UK Medical Research Council (A10976 and G9900061 to E.Y.J.) and the Wellcome Trust (grant 090532/Z/09/Z supporting the Wellcome Trust Centre for Human Genetics). Y.Z. was supported by the BioStruct-X project funded by the seventh Framework Programme (FP7) of the European Commission. T.M. was supported by the UK Medical Research Council. Received: October 15, 2013 Revised: December 30, 2013 Accepted: January 13, 2014 Published: February 27, 2014 REFERENCES Amato, R.J., Hawkins, R.E., Kaufman, H.L., Thompson, J.A., Tomczak, P., Szczylik, C., McDonald, M., Eastty, S., Shingler, W.H., de Belin, J., et al. (2010). Vaccination of metastatic renal cancer patients with MVA-5T4: a randomized, double-blind, placebo-controlled phase III study. Clin. Cancer Res. 16, 5539–5547. Aricescu, A.R., Lu, W., and Jones, E.Y. (2006). A time- and cost-efficient system for high-level protein production in mammalian cells. Acta Crystallogr. D Biol. Crystallogr. 62, 1243–1250.

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

Awan, A., Lucic, M.R., Shaw, D.M., Sheppard, F., Westwater, C., Lyons, S.A., and Stern, P.L. (2002). 5T4 interacts with TIP-2/GIPC, a PDZ protein, with implications for metastasis. Biochem. Biophys. Res. Commun. 290, 1030– 1036. Boghaert, E.R., Sridharan, L., Khandke, K.M., Armellino, D., Ryan, M.G., Myers, K., Harrop, R., Kunz, A., Hamann, P.R., Marquette, K., et al. (2008). The oncofetal protein, 5T4, is a suitable target for antibody-guided anti-cancer chemotherapy with calicheamicin. Int. J. Oncol. 32, 221–234. Bourhis, E., Wang, W., Tam, C., Hwang, J., Zhang, Y., Spittler, D., Huang, O.W., Gong, Y., Estevez, A., Zilberleyb, I., et al. (2011). Wnt antagonists bind through a short peptide to the first b-propeller domain of LRP5/6. Structure 19, 1433–1442. Celikel, R., McClintock, R.A., Roberts, J.R., Mendolicchio, G.L., Ware, J., Varughese, K.I., and Ruggeri, Z.M. (2003). Modulation of alpha-thrombin function by distinct interactions with platelet glycoprotein Ibalpha. Science 301, 218–221. Chen, V.B., Arendall, W.B., 3rd, Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010). MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21. Chen, S., Bubeck, D., MacDonald, B.T., Liang, W.X., Mao, J.H., Malinauskas, T., Llorca, O., Aricescu, A.R., Siebold, C., He, X., and Jones, E.Y. (2011). Structural and functional studies of LRP6 ectodomain reveal a platform for Wnt signaling. Dev. Cell 21, 848–861. Cowtan, K. (2008). Fitting molecular fragments into electron density. Acta Crystallogr. D Biol. Crystallogr. 64, 83–89. DasGupta, R., Kaykas, A., Moon, R.T., and Perrimon, N. (2005). Functional genomic analysis of the Wnt-wingless signaling pathway. Science 308, 826–833. De Vries, L., Lou, X., Zhao, G., Zheng, B., and Farquhar, M.G. (1998). GIPC, a PDZ domain containing protein, interacts specifically with the C terminus of RGS-GAIP. Proc. Natl. Acad. Sci. USA 95, 12340–12345. Evans, P.R., and Murshudov, G.N. (2013). How good are my data and what is the resolution? Acta Crystallogr. D Biol. Crystallogr. 69, 1204–1214. Gallivan, J.P., and Dougherty, D.A. (1999). Cation-pi interactions in structural biology. Proc. Natl. Acad. Sci. USA 96, 9459–9464. Glaser, F., Pupko, T., Paz, I., Bell, R.E., Bechor-Shental, D., Martz, E., and Ben-Tal, N. (2003). ConSurf: identification of functional regions in proteins by surface-mapping of phylogenetic information. Bioinformatics 19, 163–164. Hodges, R.S., Heaton, R.J., Parker, J.M., Molday, L., and Molday, R.S. (1988). Antigen-antibody interaction. Synthetic peptides define linear antigenic determinants recognized by monoclonal antibodies directed to the cytoplasmic carboxyl terminus of rhodopsin. J. Biol. Chem. 263, 11768– 11775. Hole, N., and Stern, P.L. (1988). A 72 kD trophoblast glycoprotein defined by a monoclonal antibody. Br. J. Cancer 57, 239–246. Huizinga, E.G., Tsuji, S., Romijn, R.A., Schiphorst, M.E., de Groot, P.G., Sixma, J.J., and Gros, P. (2002). Structures of glycoprotein Ibalpha and its complex with von Willebrand factor A1 domain. Science 297, 1176– 1179. Janda, C.Y., Waghray, D., Levin, A.M., Thomas, C., and Garcia, K.C. (2012). Structural basis of Wnt recognition by Frizzled. Science 337, 59–64. Kabsch, W. (2010). XDS. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132. Kagermeier-Schenk, B., Wehner, D., Ozhan-Kizil, G., Yamamoto, H., Li, J., Kirchner, K., Hoffmann, C., Stern, P., Kikuchi, A., Schambony, A., and Weidinger, G. (2011). Waif1/5T4 inhibits Wnt/b-catenin signaling and activates noncanonical Wnt pathways by modifying LRP6 subcellular localization. Dev. Cell 21, 1129–1143. Kim, D.W., Krishnamurthy, V., Bines, S.D., and Kaufman, H.L. (2010). TroVax, a recombinant modified vaccinia Ankara virus encoding 5T4: lessons learned and future development. Hum. Vaccin. 6, 784–791.

Krissinel, E., and Henrick, K. (2004). Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions. Acta Crystallogr. D Biol. Crystallogr. 60, 2256–2268. MacDonald, B.T., Tamai, K., and He, X. (2009). Wnt/b-catenin signaling: components, mechanisms, and diseases. Dev. Cell 17, 9–26. Malinauskas, T. (2008). Docking of fatty acids into the WIF domain of the human Wnt inhibitory factor-1. Lipids 43, 227–230. Malinauskas, T., Aricescu, A.R., Lu, W., Siebold, C., and Jones, E.Y. (2011). Modular mechanism of Wnt signaling inhibition by Wnt inhibitory factor 1. Nat. Struct. Mol. Biol. 18, 886–893. McCoy, A.J., Grosse-Kunstleve, R.W., Adams, P.D., Winn, M.D., Storoni, L.C., and Read, R.J. (2007). Phaser crystallographic software. J. Appl. Cryst. 40, 658–674. Molday, R.S., and MacKenzie, D. (1983). Monoclonal antibodies to rhodopsin: characterization, cross-reactivity, and application as structural probes. Biochemistry 22, 653–660. Moreira, I.S., Fernandes, P.A., and Ramos, M.J. (2007). Hot spots—a review of the protein-protein interface determinant amino-acid residues. Proteins 68, 803–812. Mulryan, K., Ryan, M.G., Myers, K.A., Shaw, D., Wang, W., Kingsman, S.M., Stern, P.L., and Carroll, M.W. (2002). Attenuated recombinant vaccinia virus expressing oncofetal antigen (tumor-associated antigen) 5T4 induces active therapy of established tumors. Mol. Cancer Ther. 1, 1129–1137. Murshudov, G.N., Vagin, A.A., and Dodson, E.J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255. Myers, K.A., Rahi-Saund, V., Davison, M.D., Young, J.A., Cheater, A.J., and Stern, P.L. (1994). Isolation of a cDNA encoding 5T4 oncofetal trophoblast glycoprotein. An antigen associated with metastasis contains leucine-rich repeats. J. Biol. Chem. 269, 9319–9324. Nagai, T., Ibata, K., Park, E.S., Kubota, M., Mikoshiba, K., and Miyawaki, A. (2002). A variant of yellow fluorescent protein with fast and efficient maturation for cell-biological applications. Nat. Biotechnol. 20, 87–90. Ohta, K., Ito, A., Kuriyama, S., Lupo, G., Kosaka, M., Ohnuma, S., Nakagawa, S., and Tanaka, H. (2011). Tsukushi functions as a Wnt signaling inhibitor by competing with Wnt2b for binding to transmembrane protein Frizzled4. Proc. Natl. Acad. Sci. USA 108, 14962–14967. Polakis, P. (2012). Drugging Wnt signalling in cancer. EMBO J. 31, 2737–2746. Reeves, P.J., Callewaert, N., Contreras, R., and Khorana, H.G. (2002). Structure and function in rhodopsin: high-level expression of rhodopsin with restricted and homogeneous N-glycosylation by a tetracycline-inducible N-acetylglucosaminyltransferase I-negative HEK293S stable mammalian cell line. Proc. Natl. Acad. Sci. USA 99, 13419–13424. Sapra, P., Damelin, M., Dijoseph, J., Marquette, K., Geles, K.G., Golas, J., Dougher, M., Narayanan, B., Giannakou, A., Khandke, K., et al. (2013). Long-term tumor regression induced by an antibody-drug conjugate that targets 5T4, an oncofetal antigen expressed on tumor-initiating cells. Mol. Cancer Ther. 12, 38–47. Seiradake, E., Coles, C.H., Perestenko, P.V., Harlos, K., McIlhinney, R.A., Aricescu, A.R., and Jones, E.Y. (2011). Structural basis for cell surface patterning through NetrinG-NGL interactions. EMBO J. 30, 4479–4488. So¨llner, C., and Wright, G.J. (2009). A cell surface interaction network of neural leucine-rich repeat receptors. Genome Biol. 10, R99. Southall, P.J., Boxer, G.M., Bagshawe, K.D., Hole, N., Bromley, M., and Stern, P.L. (1990). Immunohistological distribution of 5T4 antigen in normal and malignant tissues. Br. J. Cancer 61, 89–95. Starzynska, T., Marsh, P.J., Schofield, P.F., Roberts, S.A., Myers, K.A., and Stern, P.L. (1994). Prognostic significance of 5T4 oncofetal antigen expression in colorectal carcinoma. Br. J. Cancer 69, 899–902. Terwilliger, T.C., Grosse-Kunstleve, R.W., Afonine, P.V., Moriarty, N.W., Zwart, P.H., Hung, L.W., Read, R.J., and Adams, P.D. (2008). Iterative model building, structure refinement and density modification with the PHENIX AutoBuild wizard. Acta Crystallogr. D Biol. Crystallogr. 64, 61–69.

Structure 22, 612–620, April 8, 2014 ª2014 The Authors 619

Structure Inhibition of Wnt Signaling by Cancer Antigen 5T4

Trombetta, E.S., and Parodi, A.J. (2003). Quality control and protein folding in the secretory pathway. Annu. Rev. Cell Dev. Biol. 19, 649–676. Walter, T.S., Diprose, J.M., Mayo, C.J., Siebold, C., Pickford, M.G., Carter, L., Sutton, G.C., Berrow, N.S., Brown, J., Berry, I.M., et al. (2005). A procedure for setting up high-throughput nanolitre crystallization experiments. Crystallization workflow for initial screening, automated storage, imaging and optimization. Acta Crystallogr. D Biol. Crystallogr. 61, 651–657.

620 Structure 22, 612–620, April 8, 2014 ª2014 The Authors

Woods, A.M., Wang, W.W., Shaw, D.M., Ward, C.M., Carroll, M.W., Rees, B.R., and Stern, P.L. (2002). Characterization of the murine 5T4 oncofoetal antigen: a target for immunotherapy in cancer. Biochem. J. 366, 353–365. Zhao, Y., Bishop, B., Clay, J.E., Lu, W., Jones, M., Daenke, S., Siebold, C., Stuart, D.I., Jones, E.Y., and Aricescu, A.R. (2011). Automation of large scale transient protein expression in mammalian cells. J. Struct. Biol. 175, 209–215.