Structure-based design of a protein immunogen that displays an HIV-1 gp41 neutralizing epitope

Abstract

Antibody Z13e1 is a relatively broadly neutralizing anti-HIV-1 antibody that recognizes the membrane proximal external region (MPER) of the HIV-1 envelope (Env) glycoprotein gp41. Based on the crystal structure of an MPER epitope peptide in complex with Z13e1 Fab, we identified an unrelated protein, IL-22, with a surface-exposed region that is structurally homologous in its backbone to the gp41 Z13e1 epitope. By grafting the gp41 Z13e1 epitope sequence onto the structurally homologous region in IL-22, we engineered a novel protein (Z13-IL22-2) that contains the MPER epitope sequence for use as a potential immunogen and as a reagent for detection of Z13e1-like antibodies. The Z13-IL22-2 protein binds Fab Z13e1 with a K_d of 73nM. The crystal structure of Z13-IL22-2 in complex with Fab Z13e1 shows that the epitope region is faithfully replicated in the Fab-bound scaffold protein; however isothermal calorimetry studies indicate that Fab binding to Z13-IL22-2 is not a lock-and-key event, leaving open the question of whether conformational changes upon binding occur in the Fab, or Z13-IL-22, or in both.

Introduction

Since its emergence as a major health threat over 25 years ago, HIV-1 has continued to spread throughout the global population. Currently, an estimated 33 million people are infected with the virus (www.unaids.org), with many living in third world areas where access to treatment is limited. Patient viral loads can be lowered to undetectable levels with highly active antiretroviral therapy (HAART), which utilizes combinations of antiviral drugs to inhibit a variety of viral proteins; however, the viral load rebounds quickly when such drugs are withdrawn. Thus, an effective vaccine to halt the spread of the HIV-1 virus is critically needed.

Due to radical sequence diversity within the assortment of viral clades of HIV-1, an effective vaccine based on an antibody response must be able to block infection from a plethora of viral isolates. Until lately, only a handful of monoclonal antibodies that mediate such broad neutralization had been discovered, but several new broadly neutralizing antibodies have recently been reported ^{1; 2; 3; 4; 5}. Studies of HIV-1 infected individuals show that 10–30% do, in fact, develop broadly neutralizing sera over time ⁶. Importantly, passive immunization with individual or cocktail mixtures of the known potent, broadly neutralizing antibodies (bNAbs) can protect nonhuman primates against SHIV challenge ^{7; 8; 9; 10; 11}, even at low serum neutralizing titers ^{12; 13}. Thus, an immunogen capable of eliciting bNAbs prior to viral exposure will likely prove a valuable component of a successful vaccine.

The bNAbs that have been structurally characterized by x-ray crystallography have unique properties that enable recognition of target epitopes on either the Env receptor binding subunit gp120 or the transmembrane fusion glycoprotein gp41 ^{14; 15}. Antibody b12 is able to access the highly conserved, but deeply buried, CD4 binding site on gp120 ^{16; 17}, using only its heavy chain, and an even more potent set of antibodies against the same epitope (VRC01, VRC02, VRC03) can neutralize over 90% of a representative panel of HIV-1 strains and subtypes ^{3; 18}. Antibody 2G12 recognizes an unusually dense cluster of high-mannose carbohydrate on the gp120 silent face by adopting a novel, domain-swapped Fab dimer that provides a high avidity binding site for the glycan cluster ^{19; 20}. Recently, two new trimer-specific antibodies were discovered that have extremely potent and broad neutralization properties (PG9 and PG16)¹. Currently, no crystal structures for these PG antibodies in complex with their gp120 epitopes are published, but the structure of the unliganded PG16 Fab^{21; 22} reveals an extremely long, but highly structured, complementarity determining region (CDR) H3 that plays a critical role in the neutralization²¹. While the aforementioned antibodies all target the gp120 glycoprotein, several bNAbs have also been found that target gp41^{23; 24; 25; 26; 27; 28}, specifically the gp41 membrane proximal external region (MPER). Their epitopes reside within gp41 residues 659–683, which are located after the C-heptad repeat and just prior to the transmembrane region. Three of these bNAbs have been structurally analyzed thus far: 2F5, 4E10, and Z13e1^{29; 30; 31; 32; 33}. These antibodies are thought to recognize the pre-fusion intermediate conformation of gp41 ³⁴, and their epitopes encompass residues 662–668 (ELDKWAS; 2F5), 672–680 (WFDITNWLW; 4E10) and 670–677 (WNWFDITN; Z13e1). Interestingly, although 4E10 and Z13e1 bind to overlapping epitopes, crystal structures of the two antibodies in complex with epitope peptides reveal that they recognize two discretely different MPER conformations ^{21; 29; 30}. Since gp41 is expected to undergo large conformational rearrangements during entry of the virus into the target cell, it is thought that the two conformations recognized by 4E10 and Z13e1 may exist at different stages in the viral entry process ³³. Although 4E10 and Z13e1 recognize approximately the same epitope region, they have different levels of neutralization potency and breadth of neutralization, with 4E10 being the more potent neutralizer by about a factor of 10. Z13e1 is limited in its neutralization breadth to isolates with Asp or Glu at position 674, corresponding to about 59% of known viral sequences²⁵. Shared features that are found in all three MPER antibodies include a hydrophobic binding site with long CDR H3 loops, which are thought to enable close contact of the antibodies with the viral membrane during neutralization ^{25; 29; 30; 31; 32; 33}. Significantly, mutation of some of the hydrophobic CDR H3 residues of 2F5 and 4E10 abolishes neutralization without affecting binding to the core peptide antigen, signifying the presence of other hydrophobic interactions, including with the viral membrane ^{35; 36; 37; 38; 39; 40}.

While the crystal structures of these bNAbs in complex with their antigens have elegantly illustrated the mechanisms used to effect their broad neutralization and provided structural templates for the critical epitope regions on the virus, utilizing this information in immunogen design has proven a challenging task ^{41; 42}. It is generally agreed that the ‘native’, cleaved, trimeric gp120/gp41 displayed on cell surfaces or viral membrane is the most structurally relevant immunogen, but translating this construct to a stable, soluble entity has yet to be achieved. Other immunogens that have been tested include monomeric gp120 or gp160 ^{43; 44; 45; 46}, as well as a large number of different peptides that represent linear epitopes scattered throughout the Env proteins. A recently developed ‘antigenically resurfaced’ gp120 antigen³ has been successfully utilized to isolate new and potent broadly neutralizing antibodies, but it is not yet known how that construct will perform in immunization trials. Other approaches to immunogen design include constraining peptide fragments of the parent protein to adopt conformations similar to those in their gp120 or gp41 parental proteins ^{28; 29; 47; 48; 49; 50; 51; 52; 53; 54; 55}. Such approaches utilize the introduction of disulfide bonds or other linkers or the inclusion of unusual amino acids that stabilize secondary structure. Unfortunately, none of these immunogens have yet been successfully translated into a useful vaccine.

Here, we present the design, synthesis and analysis of an immunogen displaying the gp41 MPER residues that are recognized by bNAb Z13e1. Novel immunogen constructs are especially needed for the hydrophobic MPER, as this region is thought to be recognized by anti-MPER neutralizing antibodies during the prehairpin-intermediate stage of membrane fusion ^{34; 35; 56; 57; 58}. Previously described MPER immunogens²⁸ include a large number of synthetic β-turn and helical peptides containing the 2F5 or 4E10/Z13 epitope or both ^{29; 48; 54; 59; 60; 61; 62}. The MPER sequence has also been grafted into surface loops of proteins on intact virus particles ^{63; 64; 65; 66; 67; 68}, as well as monomeric proteins ^{39; 69; 70}. The work described here utilizes a method whereby a protein ‘scaffold’ is used to stabilize the linear epitope of interest in a biologically relevant conformation. Such approaches were pioneered in the mid 1990’s where elegant work using the scorpion toxin charybdotoxin as a scaffold for a structurally homologous loop from a snake toxin resulted in a construct that could be used to elicit neutralizing antibodies against the parent snake toxin ⁷¹. More recently, extensive and detailed work to develop and analyze scaffold proteins to present the HIV-1 gp41 epitopes for 2F5 and 4E10 have been described ^{72; 73}. We identified a scaffold protein by searching for proteins with regions of structural homology to the Z13e1 epitope as defined by the crystal structure of an MPER peptide bound to Z13e1, and analyzing those proteins by computer-aided, visual inspection for potential binding to Fab Z13e1 via their epitope-homologous region. The candidate template protein sequence was mutated in the epitope-homologous region to introduce the Z13e1 MPER epitope sequence, and tested for binding to Z13e1. We produced two variants of the human protein IL-22 that display the Z13e1 MPER epitope, and describe their design and production, and the crystal structure for one of these proteins in complex with Fab Z13e1.

Results

Database search

The gp41 MPER peptide coordinates (peptide 178-1, W⁶⁷⁰NWFDITN⁶⁷⁷, chain P from PDB ID 3fn0 ³³) were used as a target to search the Protein Data Bank (PDB) ⁷⁴ for proteins with structurally homologous regions to the Z13e1 epitope on gp41. Backbone φ/ψ torsion angles were calculated for the target epitope peptide and for each protein in the PDB, and the torsion angles were compared by computationally ‘sliding’ the epitope peptide along the length of the template protein being tested. For example, torsion angles from a 7-residue peptide are compared to those for residues 1–7, 2–8, 3–9, etc. of the template protein. If any one of the phi/psi torsion angles differs beyond a threshold user input value (usually between 30° to 60°), the match is considered poor, and the comparison to the next short segment within that protein is initiated. The method does not require superposition of the peptide and protein and, thus, is not dependent on the quality of that superposition. The method also results in good structural agreement between peptide and protein over all of the backbone atoms. The chosen epitope-display protein coordinates are then superimposed over the target MPER peptide (bound to Fab Z13e1) by superimposition of the Cα atoms from the homologous segment. The superposition is then visually examined for clashes between the epitope-display protein and Z13e1 Fab. For Z13e1, using a 40° cutoff on torsion angle differences and φ/ψ angles from MPER residues P671-P677, we tested 41,485 PDB coordinate files, found 177 potential hits, and, after visual inspection, eliminated 95 of these coordinate sets due to severe clashes between the epitope-display protein and Fab Z13e1. The remaining 82 coordinate sets were approximately ranked based on their Cα root mean square deviations from the MPER peptide, the likelihood of possible minor clashes with Fab Z13e1, and for more subjective qualities including availability of the target gene, size of the protein, potential for expression in E. coli, and high resolution of the crystal structure for the wild-type (WT) protein. The epitope-display protein reported here is based on human IL-22 (chain B from PDB file 1m4r)⁷⁵ For the seven residue positions with structural similarity in IL-22 and the MPER search peptide, the overall root mean square deviation on Cα atoms was 0.56Å (Table 1, Figure 1; P671-P677 from MPER peptide (NWFDITN) and 65–71 from WT IL-22 (LADNNTD); n.b. throughout the manuscript, P will be used as chain identifier for residues from the MPER peptide, L and H are the chain identifiers for the Fab light and heavy chains and the IL-22 proteins have no chain identifiers). The largest deviation in backbone torsion angle between the two, seven-residue stretches was 31.7° for the ψ angle of residues P675/69 (Table 1). The superposition of the MPER peptide and corresponding segment from WT IL-22 is shown in Figure 1.

Table 1.

Torsion angle differences for MPER peptide bound to FabZ13e1 and corresponding residues used as a template in IL-22.

MPER residue (3fn0)	IL-22 Residue (1m4r)	Δphi (°)	Δpsi (°)	RMSD (Cα) per residue (Å)
NP671	L65	14.6	9.5	0.78
WP672	A66	21.2	29.0	0.42
FP673	D67	18.6	5.7	0.58
DP674	N68	8.1	2.9	0.35
IP675	N69	3.9	31.7	0.58
TP676	T70	26.2	3.6	0.40
NP677	D71	2.9	3.6	0.65

Fig. 1.

Comparison of template Fab-peptide crystal structure, modeled IL-22-Fab complex structure, and final Z13-IL-22-2-Fab crystal structure. (a) Crystal structure showing the variable domain of Fab Z13e1 in complex with gp41 MPER peptide (from PDB file 3fn0). The light and heavy chains are shown in light and dark blue, and the peptide is shown in red. The peptide coordinates were used to search the PDB for proteins with regions of structural homology to the peptide. (b) Model of human IL-22 (1m4r, chain B) superimposed onto the Z13e1-peptide crystal structure by superposition of IL-22 (gray) residues 65–71 with peptide (red) residues P671-P677. (c) Crystal structure of Z13-IL22-2 (yellow) epitope display protein in complex with Fab Z13e1. The epitope region is highlighted in red. This figure and other figures depicting the structure were made using MolScript¹⁰⁶, BobScript¹⁰⁷ and Raster3D ¹⁰⁸.

Production of epitope-display IL-22 mutants

Two variants of the Z13-IL22 epitope-display protein that differ by the number of gp41 residues introduced into the IL-22 framework were produced. The Z13e1 epitope includes gp41 residues 670–677(WNWFDITN); however, as Phe^P673 and Asn^P677 do not contact Fab, they were not included in the epitope-display protein design. The two mutant proteins that we describe here (Z13-IL22-1 and Z13-IL22-2) include human IL-22 residues 39–179, with residues LADNNT (65–70) replaced by residues NWDDIT (Z13-IL22-1), and with residues SLADNNT (64–70) replaced by residues WNWDDIT (Z13-IL22-2). These two mutants differ only at position 64, where we considered that replacing WT IL-22 Ser⁶⁴ with Trp (P670 in the MPER) might disrupt the IL-22 structure. However, Trp^P670 makes extensive contacts with Fab Z13e1³³, and is required for tight binding to linear MPER peptides ²⁶. Thus, both options at this position were tested. The two mutants were cloned, expressed in E. coli as inclusion bodies, and refolded by rapid dilution.

Z13e1 binds both variant proteins as shown by both Isothermal Titration Calorimetry (ITC) and ELISA experiments. In direct immobilization ELISA with the full length MPER peptide 179.4 (LLELDKWASLWNWFDITNWLWYIKKKK)²⁶ as a positive control, both Z13-IL22-1 and Z13-IL22-2 bound to mAb Z13e1, while no binding was detected for WT IL-22 (Supplemental Figure 1). Direct immobilization ELISA was also used to show that mAb 4E10 does not bind to WT IL-22, Z13-IL22-1, Z13-IL22-2, but does bind the positive control peptide 132 (SLWNWFDITNWLWYIKKKK)²⁶(Supplemental Figure 1). The IL-22 variants were also tested against Z13e1 using a solution competition approach (Supplemental Figure 1). In this assay, the IL-22 variants compete with a biotinylated high affinity peptide (SLWNWFDITNWLWRRK(biotin)-NH₂)²⁶ for mAb Z13e1 binding. As shown in Supplemental Figure 1, only mutant Z13-IL22-2 and positive control peptide 132 were able to interact with mAb Z13e1 in the presence of the bio-peptide with apparent affinities of approximately 150 nM and 110 nM, respectively. The negative peptide control 94.1 (NWFDITNWLWYIKKKK)²⁶ , which does not have the essential Trp residue prior to the first Asn, did not bind at the highest concentration tested (>5 µM). Finally, Z13-IL22-1, Z13-IL22-2 and positive control peptide 132 were tested for binding to Fab Z13e1 using ITC. The K_d values determined were 15±6 µM for Z13-IL22-1, 73±8 nM for Z13-IL22-2, and 29±2 nM for peptide 132 (Figure 2, Table 2). Binding of Fab Z13e1 to Z13-IL22-2 and the peptide was also examined at 3 different temperatures (293K, 298K, and 303K). These experiments show a constant change in Gibbs energy of binding that is characterized by a favorable enthalpy change coupled to an unfavorable change in entropy at higher temperatures. Furthermore, the calculated ΔC_p values of −0.65kcal K⁻¹ mol⁻¹ for both peptide and Z13-IL22-2 binding events are within the range of typical protein-protein interactions ⁷⁶.

Fig. 2.

Isothermal titration calorimetry of FabZ13e1 with peptide 132, Z13-IL22-1, and Z13-IL22-2. Calorimetry data for the titration of (a) peptide 132, (b) Z13-IL22-2, and (c) Z13-IL22-1, all with Fab Z13e1 in 20mM Tris, 150mM NaCl, pH 8.0. Top panels show raw data while lower panels show normalized integrated areas for each spike. The temperature dependence of the thermodynamic parameters is shown for Fab binding to (d) peptide 132 and to (e) Z13-IL22-2. The heat capacity changes (ΔC_p) were calculated from the slope of the ΔH plot.

Table 2.

ITC measurements for Fab13e1 with peptide 132, Z13-IL22-1, and Z13-IL22-2. ΔG was calculated from the basic thermodynamic principle: ΔG = ΔH − TΔS. ΔC_p was obtained from the slope of ΔH in the temperature dependence plot.

Construct	ΔG (kcal mol−1)			ΔH (kcal mol−1)			−TΔS (kcal mol−1)			Kd (nM)†	ΔCp (kcal K−1 mol−1)*
	293 K	298 K	303 K	293 K	298 K	303 K	293 K	298 K	303 K
Peptide 132	−10.17	−10.21	−10.44	−9.29	−12.67	−15.74	−0.88	2.46	5.30	29 ± 2	−0.64 ± 0.23
Z13-IL22-2	−9.65	−9.85	−9.78	−7.46	−10.78	−13.90	−2.19	0.94	4.12	73 ± 8	−0.64 ± 0.15
Z13-IL22-1	---	---	−7.33	---	---	−7.47	---	---	0.14	15,000 ± 6,000	---

^†Values for the binding constant are an average of three or more independent ITC measurements given with the standard error of the mean (SEM).

^*The associated ΔC_p error represents the calculated 95% confidence interval.

Crystal structure of Z13-IL22-2 in complex with Fab Z13e1

In good agreement with the affinity measurements, the Fab Z13e1 complex with Z13-IL22-1 was not stable during purification over a size exclusion column and was, therefore, not tested in crystallization trials. However, the higher affinity Fab Z13e1/Z13-IL22-2 complex was purified by size exclusion chromatography, concentrated to 7 mg/ml, crystallized and its structure determined to 2.15 Å resolution (Figure 3). The crystals contain one Fab-protein complex in the asymmetric unit. The Z13-IL22-2 construct also contained a 21-residue N-terminal region containing a 6XHis affinity tag and thrombin cleavage site. This tag was not removed prior to crystallization and was not visible in the electron density maps. The final model has R_cryst and R_free values of 20.7 and 26.0%, respectively.

Fig. 3.

Fab Z13e1/Z13-IL22-2 structure. (a) Z13-IL22-2 is shown in yellow, Fab light and heavy chains in light and dark gray, and Fab CDR loops L1, L2, L3, H1, H2, and H3 in blue, magenta, light green, cyan, pink, and green, respectively. The approximate location of disordered regions in the Z13-IL22-2 and the Fab are shown with dotted lines. (b) Sequence comparison of human IL-22 (1m4r, chain B) and Z13-IL22-2. The two sequences are aligned with helices highlighted by rectangles. Residues highlighted in red make up the Z13e1 epitope, and residues shown in italics are disordered in the Fab/Z13-IL22-2 structure. Secondary structure evaluated with DSSP ¹⁰⁹. (c) Stereoview of the superposition of WT human IL-22 (gray) and Z13-IL22-2 (yellow). The Z13e1 epitope region is highlighted in red, and disordered regions are shown with dotted lines.

The Z13-IL22-2 protein bound to Fab differs from WT human IL-22 in 5 amino-acid positions, and maintains the overall IL-22 fold (Figure 3) of a 6-helix, up-down bundle. Residues 39, 47–57 (the A1-A2 connecting region and first half of the A2 helix) and 103–108 (the CD loop and beginning of helix D) are disordered in the electron density maps, preventing their inclusion in the crystal structure coordinates. The region of the protein that contains the MPER residues (58–71, A2-B loop) has moved away from the body of the protein by about 8Å (Figure 3, Supplemental Figure 2) compared to its position in WT IL-22. When the Z13-IL22-2 and WT IL-22 structures are superimposed on their structurally similar core residues (72–101 and 119–179), the overall RMSD on the Cα atoms for those residues is 1.03Å. After superposition, residues 58–71 (that were not included in the superposition) differ by an RMSD of 7.96Å. If residues 58–71 are used for the superposition, the RMSD for those 14 atoms is 2.09Å. Thus, the Z13-IL22-2 core region is very similar to that of the WT IL-22 protein, and the 58–71 loop regions are similar between the two proteins, but the two structural elements have moved apart from one another, possibly due to the Ser⁶⁴Trp mutation. The two disordered regions in Z13-IL22-2 may result from loss of contact to the 58–71 loop (Supplemental Figure 2). The disorder and loop movement observed here are not seen in any of the 5 different crystal structures (3dlq, 3g9v, 1m4r, 1ykb, 3dgc) ^{75; 77; 78; 79; 80} of human WT IL-22 in both liganded and unliganded states, which show little variation with respect to each other (Supplemental Figure 3).

The liganded Fab Z13e1 has the expected structure of four immunoglobulin-fold domains, arranged into a variable and constant region related by an elbow angle of 166°, almost identical to that seen for Z13e1 in the MPER peptide-bound structure (168°; 3fn0) (Figure 3). Two regions of disorder were found at H126-H132 and H193-H204, and are frequently observed as flexible regions at the distal end of the Fab constant region. These regions were not included in the final model. The first two residues of the light chain are also not visible, and are also omitted from the model. All six CDR loops are well ordered, with canonical structures 2 for CDR L1, 1 for CDR L2, and 1 for CDR H2. H1 is similar in overall conformation to canonical structure 1; however, the H30 carbonyl oxygen has flipped, so that residues H30 and H31, that are the i+1 and i+2 residues in a β-turn, form a type II rather than the type I turn normally seen for Fabs at this position. CDR L3 has an unusual sequence, derived during in vitro affinity maturation of Z13e1, and as a result adopts a novel structure ³³ that helps to stabilize the interaction between the light and heavy chains. Significantly, the extreme tip of the CDR H3 is more ordered in the Fab/Z13-IL22-2 structure due to contacts with Z13-IL22-2 residues 75–77, 80, 169, and 172–173 (Figure 3, Table 3), and is in a slightly different position than in the Fab-peptide structure (Supplemental Figure 4). These additional contacts may be of value in thinking about the design of the next generation of immunogens. Furthermore, CDR H3 in the peptide-bound structure lacks ordered electron density for the side chains of residues H99-H100c, but these side chains are clearly resolved in the Z13-IL22-2-bound Fab complex (Figure 4, Supplemental Figure 4).

Table 3.

Comparison of molecular surface area buried in Fab Z13e1/Z13-IL22-2 and Fab Z13e1/ MPER peptide structures. The buried surfaces on these two complexes are compared and aligned to indicate how the corresponding residues from Fab Z13e1 interact with the MPER peptide or when its equivalent when transplanted into the Z13-IL22-2 protein. Clearly, other residues for Z13-IL-22-2 are contacted in addition to the MPER residues.

Z13-IL22-2/Z13e1 structure (PDB 3q1s)				gp41 MPER peptide/Z13e1 structure (PDB 3fn0)
Z13- IL22-2 Residue	Buried MSA (Å2)	Z13e1 Residue	Buried MSA (Å2)	Peptide Residue	Buried MSA (Å2)	Z13e1 Residue	Buried MSA (Å2)
Lys61	13.5	LeuL92	16.7			LeuL92	15.1
Gly62	16.9	IleH30	17.9			IleH30	12.8
Trp64	61.7	AsnH31	27.2	TrpP670	79.3	AsnH31	37.0
Asn65	70.2	TyrH32	6.9	AsnP671	68.1	TyrH32	6.7
Trp66	47.0	TyrH33	49.0	TrpP672	54.4	TyrH33	50.2
Asp67	4.7	HisH50	18.6	PheP673	9.0	HisH50	16.3
Asp68	55.3	IleH52	44.4	AspP674	51.8	IleH52	41.3
Ile69	75.0	TyrH53	38.9	IleP675	69.4	TyrH53	41.6
Thr70	67.3	GlyH54	26.1	ThrP676	67.5	GlyH54	31.2
Asp71	20.0	ThrH56	26.3	AspP677	26.4	ThrH56	27.1
Arg73	5.9	ThrH57	1.3	LysP678	3.2	ThrH57	2.9
Ile75	18.9	LysH58	38.6			LysH58	38.2
Gly76	14.3	ArgH71	1.9			ArgH71	2.5
Glu77	38.9	IleH97	35.1			IleH97	37.8
Phe80	31.5	GlyH98	13.7			GlyH98	21.0
Arg88	16.3	ValH99	35.2			ValH99	17.2
Leu169	22.3	PheH100b	65.7
Met172	23.6	LeuH100c	71.7
Ser173	16.4	Asn H100d	10.9			AsnH100d	7.3
		TyrH100e	53.2			TyrH100e	55.4
						TyrH100f	1.1
		TyrH100g	12.9			TyrH100g	10.3
Total	619.9		612.3		429.1		473.0

Fig. 4.

Detailed binding site interactions in the Fab-protein and Fab-peptide complexes. (a) Stereoview of the Fab Z13e1 and Z13-IL22-2 epitope contact residues, with hydrogen bonds shown as dotted lines. The Z13-IL22-2 is shown as a yellow ball-and-stick model, and the Fab interacting region is depicted as tubes, with contact residues in ball-and-stick. CDR and framework regions are colored as in Fig. 3. (b) Stereoview of the Fab Z13e1 in complex with MPER peptide (PDB 3fn0) shows very similar contacts for the epitope region, although the position of Trp^P670 (corresponding to Trp⁶⁴ in panel a) changes in the Z13-IL22-2 structure.

Comparison of epitope region in gp41 peptide and Z13-IL22-2

The contact region between Z13-IL22-2 and Fab Z13e1 is very well ordered (Figure 5). Around 620 and 612 Ų of molecular surface area are buried on the Z13-IL22-2 and Fab, respectively (Table 3). The Fab light chain (via CDR L3) contributes only 2.7% of the buried surface area, while 16.5%, 31.7%, and 48.7% of the buried surface is contributed by CDRs H1, H2, and H3, respectively. Framework residue Arg^H71 also makes a minor contact (<1 % of the buried surface). A total of 99 van der Waals contacts, 7 hydrogen bonds and 1 salt bridge make up the interaction between Fab and Z13-IL22-2 (Tables 4 and 5). Six of the hydrogen bonds and the salt bridge are conserved between the Z13-IL22-2 and peptide-bound (PDB 3fn0) structures. The Z13-IL22-2 structure has an additional hydrogen bond between Thr⁷⁰ and Lys^H58, which is too long to be classified as a hydrogen bond in the peptide-bound structure. On the other hand, the peptide-bound structure has a hydrogen bond between Trp^P670Nε1 and Asn^H31Oδ1 that does not form in the Z13-IL22-2 structure due to movement of Trp⁶⁴ The Fab/Z13-IL22-2 buried surface area is about 25% larger than that in the Fab-peptide structure where 429 and 473 Ų of molecular surface are buried on the peptide and Fab. The increase in buried surface area is due to additional interactions between Fab and non-epitope residues in Z13-IL22-2. A superposition of equivalent peptide residues P670-P678 from the Fab Z13e1 complex structure (PDB 3fn0) with Z13-IL22-2 residues 64–72 gives an RMSD of 0.38Å for Cα atoms. The side chain of Trp⁶⁴ has flipped in Z13-IL22-2 to form hydrophobic interactions with Ile⁶⁹ (Figure 5). This Trp indole in Z13-IL22-2 cannot adopt the same position as in the MPER peptide structure due to clashes with the protein core of Z13-IL22-2.

Fig. 5.

The gp41 Z13e1 epitope region of Z13-IL22-2. (a) Stereoview of 2Fo-Fc electron density for Z13-IL22-2 residues 63–72. The residue numbers in parentheses are the corresponding residue number in gp41. If there is no number in parenthesis, that residue was not mutated to the gp41 sequence. (b) Superposition of Z13e1 gp41 peptide epitope (red, from PDB 3fn0) and corresponding region in Z13-IL22-2 (yellow). The one major change in the structure is the location of the Trp⁶⁴/Trp^P670 side chain.

Table 4.

Comparison between hydrogen bonds and salt links between Fab Z13e1 residues (left) and Z13-IL22-2 (middle) and MPER peptide (right) ³³.

Fab atom	Z13-IL22-2 atom	Distance (Å)	MPER peptide atom	Distance (Å)
TyrH33 OH	Trp64 O	2.8	TrpP670 O	2.6
AsnH31 N	Asn65 Oδ1	2.8	AsnP671 Oδ1	2.8
AsnH31 Oδ1	Asn65 Nδ2	2.8	AsnP671 Nδ2	2.7
GlyH54 N	Asn65 O	2.7	AsnP671 O	2.8
HisH50 Nε2	Asp68 Oδ2	2.8	AspP674 Oδ2	2.8
TyrH33 OH	Ile69 N	2.9	IleP675 N	2.9
LysH58 Nζ	Thr70 Oδ1	2.7	ThP676 Oγ1	(4.0)
LysH58 Nζ	Asp71 Oδ1	2.9	AsnP677 Oδ1	2.7
AsnH31 Oδ1		-	TrpP670 Nε1	3.3

Table 5.

van der Waals contacts between Z13-IL22-2 and Fab Z13e1. Equivalent positions in the MPER are listed in parentheses for Z13-IL22-2.

Z13-IL22-2 residue	Fab Z13e1 residue
Trp64(P670)	TyrH33, IleH97, ValH99
Asn65(P671)	IleH30, AsnH31, IleH52, TyrH53, GlyH54
Trp66(P672)	IleH52, TyrH53, GlyH54
Phe67(P673)	TyrH33, IleH52
Asp68(P674)	TyrH33, HisH50, IleH52, LysH58
Ile69(P675)	TyrH33, IleH97, GlyH98, ValH99, TyrH100e
Thr70(P676)	LeuL92, HisH50, LysH58, TyrH100e, TyrH100g
Asp71(P677)	LysH58
Ile75	LeuH100c
Gly76	LeuH100c
Glu77	PheH100b
Phe80	PheH100b
Ser173	LeuH100c

Discussion

A novel protein Z13-IL22-2 that binds the HIV-1 neutralizing antibody Z13e1 with a K_d of 73±8nM has been designed, cloned, expressed, and purified and its crystal structure determined in complex with Fab Z13e1. While the structure of the MPER-containing region in Z13-IL22-2 when bound to the Fab is almost identical to that of the equivalent gp41 MPER peptide bound to the same Fab, the region containing the epitope in Z13-IL-22-2 (residues 64–72), is shifted from its native position in the WT IL-22 protein. Thus, it is not clear whether the 64–72 region also adopts the Z13e1-bound conformation in the unliganded state. As attempts to crystallize the unliganded Z13-IL22-2 have been unsuccessful thus far, direct structural evidence for the epitope conformation in the absence of antibody is not available. While the Ser⁶⁴Trp mutation increases affinity for Fab Z13e1 (Table 2), it might also contribute to disorder in the Z13-IL22-2 protein (Supplemental Figure 2). Structural analysis of the buried surface area at the interface in relation to the ΔC_p determined in ITC measurements ⁸¹ suggest that the binding of scaffold to Fab is probably not a lock-and-key type event, because the experimental ΔC_p exceeds the theoretical ΔC_p value calculated for the burial of the interface alone. However, the lack of unliganded structures for both Z13-IL22-2 and the Fab make it impossible to speculate on the source of these additional contributions. It is possible that the epitope region in unliganded Z13-Il22-2 is flexible, and becomes more ordered upon binding Fab; however, if this is indeed the case, it might actually be beneficial in immunization, as epitope flexibility has been shown to positively correlate with immunogenicity in other scaffold design studies ⁷³ and a flexible MPER region might have more ways to bind and activate different germline precursor BCRs. Nonetheless, future variants will include substitution of Trp⁶⁴ with a smaller, hydrophobic residue (Phe, Leu, Val) to try to maintain hydrophobic contacts with the Fab without disrupting the IL-22 structure or, alternatively, presentation of a Trp side chain from a different location on IL-22, such as residue Lys⁶¹, which is located about one α-helical turn from Trp⁶⁴. Currently, Fab Z13e1 binds Z13-IL22-2 (73nM) with an affinity about 2 times lower than that of Fab for peptide 132 (29nM). This difference may be due to the extra flanking residues present in the peptide, or relate to the need to unfold the region surrounding the Z13-IL22-2 epitope before its binds to Fab. Whatever the reasons for the difference in affinities, future variants will be optimized for tighter binding, as well as for stabilization of the IL-22 scaffold protein.

IL-22 is a member of the IL-10 cytokine family, and is an important component of the immune response (reviewed in^{82; 83}). IL-22 mediates its inflammatory response by binding to a two-component receptor (IL-22R1/IL-10R2) that then activates the Stat3 signaling cascade. IL-22 is important in preventing and aiding in the repair of tissue damage, but can be up-regulated in patients with chronic inflammatory diseases, such as rheumatoid arthritis or psoriasis. It is expected that human IL-22 would not be immunogenic and that immune responses against Z13-IL22-2 and its variants would, therefore, be focused onto the introduced, non-self, MPER epitope region. However, it is possible that anti-IL-22 autoantibodies will eventually be generated, and it is also possible that potential binding of IL-22 variants to their natural receptor will lead to problems. A solution to the latter issue is to mutate one or more residues within the receptor binding site of IL-22 to prevent interaction with its receptor. With regard to potential autoimmunity concerns, glycosylation sites might be introduced to mask ^{84; 85; 86} non-epitope regions of the protein if any undesired immunogenicity problems arise. Alternatively, our approach taken with IL-22 can be applied to other structurally compatible, epitope-display molecules, any of which can be optimized through further engineering.

The 2F5 minimal MPER epitope has been displayed in surface loops on human rhinovirus constructs, by insertion via random length linkers ⁶⁷. These constructs have been reported to elicit ‘broad, but modest’, neutralizing sera in guinea pigs, giving some encouragement for epitope-display for vaccine design strategies in general. Additionally, 4E10 and 2F5 scaffold proteins have been shown to elicit rabbit antibodies that bind their gp41 targets with high affinity, although with no associated neutralization activity ^{72; 73; 87; 88}. Thus, several potential limitations still need to be overcome for these and other structure-assisted vaccine designs, including the difficulty in relating antigenicity and immunogenicity for a given protein region, and the challenge of eliciting an antibody that binds the desired epitope with the appropriate direction and angle of approach such that it can also access its cognate epitope on cell surface HIV-1 Env trimers. The MPER antibodies in particular must be able to fit into highly sequestered regions that may only be briefly exposed, so that the angle of attack on their epitopes is likely a critical component. Whether a long CDR H3 is the only solution to access these recessed regions in Env is unclear. Significantly, while most of the known HIV-1 antibody (both neutralizing and non-neutralizing) H3 regions are longer than the average 12–13 residues seen in human antibody sequences^{89; 90}, the design of immunogens or vaccination strategies that can preferentially elicit longer CDR H3’s is a major challenge. Although Z13e1 has a relatively long CDR H3 (17 residues), it is shorter than those of 2F5 (22) or 4E10 (18) and, thus, Z13e1-like bNAbs may be slightly easier to elicit. Furthermore, the Z13e1 CDR H3 makes interactions with Z13-IL22-2 that are not made with the free gp41 MPER peptide. As no crystal structure of Z13e1 is available with the intact gp120/gp41 trimer, it is not known whether CDR H3 contacts additional residues in the gp41/gp120 subunits, or perhaps the viral membrane during neutralization^{35; 91; 92; 93}. Thus, it is unclear whether or not the Z13-IL22-2 contacts to CDR H3 are desirable from a vaccine design perspective. However, modulation of the Z13-IL22-2 H3 contact residues may be a useful variable to exploit in attempts to influence the length and composition of CDR H3.

In addition to their intended use as anti-HIV-1 immunogen constructs, epitope-display proteins that correctly stabilize a conformational epitope, may be useful for serum specificity profiling, for example to distinguish between Z13e1-like and 4E10-like antibodies that each recognize different conformations of the same linear epitope. Epitope-display proteins could also be used as indirect probes of non-MPER interacting regions of the MPER antibodies to aid in elucidating requirements for neutralization of these regions. Finally, while Z13e1 is less potent and broadly neutralizing than 2F5 and 4E10, it has one property that may make its MPER epitope more attractive as a template for vaccine design. 2F5 and 4E10 have been proposed to contact membrane and extract membrane-embedded MPER residues during viral neutralization^{35; 91; 92; 93}; however, binding of Z13e1 to the MPER does not appear to require such extraction or involve changes in the membrane structure ⁹². If accessing the Z13e1 epitope does not require such a mechanism, it may prove easier to elicit a Z13e1-like antibody. Finally, the MPER conformation recognized by Z13e1 may have the potential to mediate more potent neutralizing activity, perhaps through more extensive interaction with the antibody light chain, and future variants may be utilized to explore this possibility. Work is currently in progress to express Z13-IL22-2 in mammalian cells and to work towards stabilizing the epitope region for testing as immunogens.

Materials and methods

Database search

The gp41 MPER peptide coordinates (peptide 178-1;W⁶⁷⁰NWFDITN⁶⁷⁷; chain P from PDB ID 3fn0 ³³) were used as a target to search the Protein Data Bank (PDB) ⁷⁴ for proteins with structurally homologous regions. Backbone φ/ψ torsion angles were calculated with the program DIHDRL (author E.E. Abola) and the angles of the target peptide and test proteins were compared using an in-house Fortran program. A 40° cutoff for torsion angle differences was used. Superpositions of potential epitope-display proteins and Z13e1-peptide were performed using the McLachlan algorithm ⁹⁴ as implemented in the program ProFit (Martin, A.C.R. and Porter, C.T., http://www.bioinf.org.uk/software/profit/) and visually checked for clashes between the epitope-display protein and the Z13e1 Fab using Coot ⁹⁵. The 41,485 PDB coordinate files available at the time were tested, and 177 potential files were identified that met the torsion angle criterion. After visual examination, 95 of these coordinate sets were eliminated because of severe clashes between the epitope-display protein and Fab Z13e1. All non-human proteins were then eliminated, leaving us with 21 PDB files, representing 14 different human proteins (Supplemental Table 1). We focused on human targets to minimize potential immunogenicity problems in humans and because many of these human genes are readily and inexpensively available from the I.M.A.G.E. Consortium⁹⁶ IL-22 was chosen for its good fit to the target gp41 epitope (RMSD = 0.56 Å for 7 residues), its small size (151 residues), its ability to be expressed in E. coli for the initial tests, its native aggregation state (monomeric), and the small number of bad (i.e. too close) van del Waals contacts in the interface of the modeled Fab-scaffold complex.

Cloning and mutagenesis

The cDNA clone for human IL-22 was obtained from Thermo Scientific Open Biosystems (Image Reference: number 30416027) in the pDNR-LIB donor vector. IL-22 residues 39–179 were subcloned into a pET15b+ vector (Novagen) with an N-terminal 6XHis tag separated from the protein by a thrombin cleavage site (21 residues). Using QuikChange mutagenesis (Stratagene), mutations were made to include different residues from the Z13e1 MPER epitope for two different mutant proteins (Figure 3b).

Expression and purification of Z13-IL22 and complex with Fab Z13e1

The vectors containing the WT IL-22 and the Z13-IL22 mutants were used to transform Rosetta 2(DE3) competent cells (Novagen). Sixty ml overnight (o/n) cultures were grown in Terrific Broth (Research Products International) supplemented with 0.5% glucose and 50 µg/ml ampicillin. Five hundred ml cultures of Overnight Express TB media (Novagen) supplemented with 50 µg/ml ampicillin were inoculated with 25ml each of the o/n culture for each construct and grown at 37°C for 2 hrs, at which time the temperature was lowered to 30°C. Cultures were shaken at 300rpm and grown for approximately 24 hr from the initial inoculation. Cells were harvested by centrifugation at 4000g for 30 min. The pelleted cells were resuspended in 50mM Tris (Tris(hydroxymethyl)aminomethane), 100mM NaCl, pH 8, with Complete EDTA free protease inhibitor (Roche) and lysed in a cell disrupter (Avestin) at 15kpsi, twice. The lysed cells were pelleted at 4000g for 30 min. The cell pellet including inclusion bodies was dissolved in 6M guanidine hydrochloride, 20mM Tris, 2.5mM TCEP (Tris (2-carboxyethyl)phosphine), pH 8, and stirred at room temperature (RT) for 2–3 hours (hrs). The denatured protein in guanidine hydrochloride was then shaken o/n with Ni NTA Superflow nickel chelating resin (Qiagen). Resin was poured into a glass Econo-column (Bio-Rad), washed with 5 column volumes of 6M guanidine hydrochloride, 20mM Tris, 2.5mM TCEP, 20mM imidazole, pH 8, and the bound protein was eluted from the resin with 6M guanidine hydrochloride, 20mM Tris, 2.5mM TCEP, 500mM imidazole, pH 8, collecting 1ml fractions. Denatured protein was refolded by rapid dilution into 500ml 50mM Tris, 500mM L-arginine, 1mM reduced glutathione, and 0.1mM oxidized glutathione, pH 8, and stirred in the cold for 48 hrs. Protein was concentrated with a stirred ultrafiltration cell (Millipore) with 3000 MW cutoff regenerated cellulose filter. At a concentration of approximately 0.5mg/ml, the protein was primarily monomeric as determined by size exclusion chromatography. The N-terminal tag was not removed prior to crystallization or for any ELISA experiments. Fab Z13e1 was prepared by proteolysis of IgG Z13e1 as described previously ³³. Z13-IL22 mutants (0.5 mg/ml) were mixed with Fab Z13e1 (10 mg/ml) in an approximate 1:1 molar ratio with a slight excess of the Z13-IL22 constructs. The complex was allowed to incubate o/n at 4°C, concentrated to 2mg/ml, and purified by size-exclusion chromatography on a Superdex 75, 10/30 column with 20mM Tris, 150mM NaCl, pH 7.5. The complex of Z13e1 with Z13-IL22-1 was not stable during the size exclusion process; however, the complex with Z13-IL22-2 remained intact during the size exclusion step. Fractions containing pure Fab-protein complex were pooled and concentrated to 7mg/ml.

Isothermal Titration Calorimetry

ITC measurements were performed using a MicroCal iTC200 instrument (GE). Prior to the experiments, all proteins and peptide were dialyzed against 20mM Tris, 150mM NaCl, pH 8.0 buffer. Protein and peptide concentrations were determined by absorbance at 280nm using calculated extinction coefficients (ProtParam⁹⁷). Experiments for Z13-IL22-2 and peptide 132 were carried out in duplicate at 3 different temperatures (293K, 298K, 303K), while low affinity Z13-IL22-1 was analyzed in duplicate at 303K. Fab Z13e1 was present in the syringe at concentrations ranging from 60–100 µM, depending on the binding partner, while Z13-IL22-1, Z13-IL22-2, and peptide 132 were in the cell at concentrations ranging from 6–10 µM. One experiment consisted of 16 injections of 2.5µl each, with injection duration of 5 s, injection interval of 180 s and reference power of 5µcals. Association constants (K_a) and molar reaction enthalpy (ΔH) were calculated by fitting the integrated titration peaks with Origin 7.0 software using a single-site binding model. The entropic change ΔS and the change in Gibbs free energy ΔG were then calculated as ΔS=ΔH+RlnK_A and ΔG=ΔH-TΔS, respectively. The heat capacity change, ΔC_p was calculated from the slope of the ΔH temperature dependence graph using the GraphPad Prism software.

ELISAs

Direct immobilization ELISA

Costar 96-well half area plates were coated o/n at 4°C with 100 ng/well in PBS of peptide or IL-22 variants. Plates were washed with TPBS (PBS containing 0.05% Tween-20) and blocked with 4% non-fat dry milk (NFDM) in TPBS for 1 hr at RT. Meanwhile, antibody dilution plates were prepared by serially diluting human bNAb Z13e1 (or 4E10) in the presence of 1% NFDM in TPBS. In the next step, the antibody dilutions were transferred onto the blocked antigen-coated plate and incubated for 1 hr at RT. After washing with TPBS, the plates were subsequently incubated for 1 hr at RT with a peroxidase-labeled goat anti-human IgG, F(ab’)₂-fragment specific conjugate (Jackson ImmunoResearch Laboratories, for Z13e1 ELISA) or a peroxidase-labeled goat anti-human (H+L) specific conjugate (Pierce, for 4E10 ELISA). Plates were finally washed again with TPBS and developed with tetramethylbenzidine (TMB) substrate according to the manufacturer’s instructions (Pierce). The optical density at a wavelength of 450 nm was read on a microplate reader (Molecular Devices).

Solution competition ELISA

The solution competition ELISA was adapted from Nelson et al. ²⁶. In brief, 96-well plates (Corning) were coated o/n with 200 ng/well neutravidin (Pierce) in PBS at 4°C. Plates were washed and blocked with 4% NFDM in TPBS for 1 hr at RT. In the meantime, peptides and IL-22 variants (3-fold dilution series in 1% NFDM/TPBS at a starting concentration of 20 µM) were mixed with a constant concentration of biotinylated peptide SLWNWFDITNWLWRRK(biotin)-NH2 and mAb Z13e1. The antibody/peptide and IL-22 variants mixture was incubated for 2 hrs at 37°C, subsequently transferred to the washed neutravidin plates and incubated at 37°C for 30 min. Finally, the plates were incubated with a 1:1,000 dilution of a peroxidase-labeled goat anti-human IgG F(ab’)₂-specific conjugate (Jackson ImmunoResearch Laboratories). After 1 hr at RT, plates were washed, developed, stopped, and read as described above. The concentration of competitor peptide and IL-22 variants corresponding to a half-maximal signal was determined by interpolation of the resulting binding curve. Each competing peptide and IL-22 variants was tested in duplicate in at least two separate experiments.

Crystallization and data collection

The Fab/Z13-IL22-2 complex was screened for crystallization using the IAVI-JCSG-TSRI CrystalMation robot (Rigaku). Different crystallization conditions (n=384) were tested at 2 different temperatures (4° and 20°C) with drops of 100nl protein mixed with 100nl well solution. Crystals were obtained in several similar conditions. Fine screens of the crystallization conditions were carried out manually using 24-well Cryschem sitting drop plates (Hampton Research) with drops of 0.3µl protein mixed with 0.3µl well solution to yield small, rod-shaped crystals from 19% PEG 4000, 10% isopropanol, 0.1 M Hepes, pH 7.5. Crystals were cryoprotected by a brief immersion in 70% well buffer, 30% glycerol, followed by immediate flash-cooling in liquid nitrogen. Data were collected at APS beamline 23ID-B to a resolution of 2.15 Å resolution, with an overall R_sym of 11.7% and 99.9% completeness (Table 6). Data were processed and scaled with HKL-2000 ⁹⁸. The space group is P2₁2₁2₁ with unit cell dimensions a=56.6Å, b=99.2Å, c=105.5Å.

Table 6.

X-ray data collection and refinement statistics.

Beamline	APS 23-ID-B
Wavelength (Å)	1.058
Resolution (Å)	2.15 (2.19-2.15)
Space group, a,b,c (Å)	P21212; 56.62, 99.22, 105.53
# observations	164,323 (8240)
# unique reflections	33,242 (1656)
Completeness (%)	99.9 (100.0)
Rsym (%)	11.7 (51.0)
Average I/σ	16.0 (2.9)
Refinement statistics (all refl. > 0.0 F/σF)
Resolution (Å)	49.63-2.15 (2.21-2.15)
# reflections (working set)	31,381 (2273)
# reflections (test set)	1639 (122)
Rcryst (%)	20.7 (26.4)
Rfree (%)	26.0 (30.6)
Stereochemical parameters
Average isotropic B (Å2) Fab variable domain	31.6
Average isotropic B (Å2) Fab constant domain	46.2
Average isotropic B (Å2) Z13-IL22-2	77.9
Average isotropic B (Å2) Waters	20.6
rmsd bond lengths (Å)	0.018
rmsd bond angles (°)	1.7
Ramachandran plot distribution from Procheck110	90.6% core
	8.6% allowed
	0.7% generously allowed
	0.2% disallowed (residue L51 which is in a conserved γ turn)

R_sym = Σ|I_i − <I_i>|/Σ|I_i|, where I_i is the scaled intensity of the ith measurement and <I_i> is the mean intensity for that reflection.

R_cryst = Σ‖F_obs| − |F_calc‖/Σ|F_obs|, where F_calc and F_obs are the calculated and observed structure factor amplitudes, respectively.

R_free = as for R_cryst, but for 5.0% of the total reflections chosen at random and omitted from refinement.

Structure determination and refinement

The structure was determined by the molecular replacement method using Phaser⁹⁹, with data to 3 Å resolution and models for the Z13e1 Fab (PDB ID: 3fn0, ³³) and human IL-22 (chain I, PDB ID: 3dlq⁷⁷), with residues 40–74 removed from the IL-22 model. The Fab was located by Phaser with a log-likelihood gain of 1601, and the IL-22 was then placed for a final log-likelihood gain of 2710. Model building was carried out using Coot-0.2⁹⁵ and refinement was implemented with the Refmac5 program ¹⁰⁰. Final R_cryst and R_free values are 20.7 and 26.0% for all data to 2.15Å. Electron density for the epitope region of the Z13-IL22-2 is very clear and well ordered (Figure 5). Several regions of the structure had no visible electron density, and were omitted from the final model (Fab residues L1-2, H126-132, H193-204 and Z13-IL22-2 residues 47–57, and 103–118; the 21-residue N-terminal expression and purification tag on Z13-IL22-2 is also not visible). All structural comparisons are with human IL-22 from PDB file 1m4r, chain B (WT human IL-22 expressed in E. coli; 2.0 Å resolution⁷⁵). Buried molecular surface areas were analyzed with MS ¹⁰¹using a 1.7 Å probe radius and standard van der Waals radii¹⁰², and van der Waals contacts and hydrogen bonds were evaluated with Contacsym^{103; 104} and HBPLUS¹⁰⁵.

Accession Number

Crystal structure coordinates and data have been deposited with the Protein Data Bank with accession number 3q1s.

Supplementary Material

01

Supplemental Fig. 1. ELISA binding curves. (a) Direct immobilization ELISA. IgG Z13e1 binds to IL-22 variants, Z13-IL22-1 and Z13-IL22-2, and positive control peptide 179.4, but not to WT IL-22 and blank. (b) In-solution competition ELISA. IgG Z13e1 binds tightly to Z13-IL22-2 (~150nM) and to the positive control, biotinylated, high affinity peptide 132 (~110nM), but not to the negative peptide 94.1 control at the highest concentration tested. (c) Direct immobilization ELISA for binding of IgG 4E10 to peptide 132 and IL-22 variants. 4E10 at the highest concentration tested (10 µg/ml) does not bind WT IL-22, Z13-IL22-1 or Z13-IL22-2, but binds positive control peptide 132.

02

Supplemental Fig. 2. Movement of epitope-display loop. (a) In WT human IL-22 (1m4r, chain B), the AB loop interacts with the F1 helix through hydrogen bonds between residues Asp⁶⁷ to Lys¹⁶² and Asn⁶⁹ to Ser¹⁵⁸. (b) In Z13-IL22-2, the AB loop has moved away from the F1 helix, likely due to the Asn⁶⁹Ile mutation, which eliminates one possible hydrogen bond, and the presence of the bulky Trp⁶⁴, which is too large to fit into the WT structure at this position, and may also lead to unwinding of the A2 helix.

03

Supplemental Fig.3. Stereoview of IL-22 structures from different crystal forms and complexes. PDB 1m4r chain a, b (yellow), PDB 1ykb chain a (green), PDB 3dgc chain l, m (blue), PDB 3dlq (cyan), PDB 3g9v chain b, d (magenta). The structures are all very similar.

04

Supplemental Fig. 4. Comparison of Z13e1 structure bound to peptide, and bound to Z13e1-IL22. (a) Fab Z13e1 from the peptide-bound structure (gray; from PDB 3fn0) is compared with Z13e1 Fab from the Z13-IL22-2 bound structure (light and heavy chains are colored light and dark blue, with CDR loops colored as in Figure 3). The only major change in the Fab structure is at the tip of CDR H3. (b) Superposition of CDR H3 from Z13e1 bound to Z13-IL22-2 (green) and bound to gp41 peptide (gray; from PDB 3fn0). The tip of CDR H3 makes no contact to peptide in the Z13e1 peptide-bound structure, but does contact the Z13-IL22-2 protein. Residues with no ordered density in the Z13e1 peptide bound structure (H99-H100c) have clear side-chain density in the Z13-IL22-2 bound structure.

Abbreviations

HIV-1

human immunodeficiency virus type 1

Env

envelope

SHIV

simian-human immunodeficiency virus

CDR

complementarity determining region

MPER

membrane proximal external region

WT

wild-type