Stability of isolated antibody-antigen complexes as a predictive tool for selecting toxin neutralizing antibodies.

Ricin is an A-B ribosome inactivating protein (RIP) toxin composed of an A-chain subunit (RTA) that contains a catalytic N-glycosidase and a B-chain (RTB) lectin domain that binds cell surface glycans. Ricin exploits retrograde transport to enter into the Golgi and the endoplasmic reticulum, and then dislocates into the cytoplasm where it can reach its substrate, the rRNA. A subset of isolated antibodies (Abs) raised against the RTA subunit protect against ricin intoxication, and RTA-based vaccine immunogens have been shown to provide long-lasting protective immunity against the holotoxin. Anti-RTA Abs are unlikely to cross a membrane and reach the cytoplasm to inhibit the enzymatic activity of the A-chain. Moreover, there is not a strict correlation between the apparent binding affinity (Ka) of anti-RTA Abs and their ability to successfully neutralize ricin toxicity. Some anti-RTA antibodies are toxin-neutralizing, whereas others are not. We hypothesize that neutralizing anti-RTA Abs may interfere selectively with conformational change(s) or partial unfolding required for toxin internalization. To test this hypothesis, we measured the melting temperatures (Tm) of neutralizing single-domain Ab (sdAb)-antigen (Ag) complexes relative to the Tm of the free antigen (Tm-shift = Tmcomplex - TmAg), and observed increases in the Tmcomplex of 9-20 degrees. In contrast, non-neutralizing sdAb-Ag complexes shifted the TmComplex by only 6-7 degrees. A strong linear correlation (r2 = 0.992) was observed between the magnitude of the Tm-shift and the viability of living cells treated with the sdAb and ricin holotoxin. The Tm-shift of the sdAb-Ag complex provided a quantitative biophysical parameter that could be used to predict and rank-order the toxin-neutralizing activities of Abs. We determined the first structure of an sdAb-RTA1-33/44-198 complex, and examined other sdAb-RTA complexes. We found that neutralizing sdAb bound to regions involved in the early stages of unfolding. These Abs likely interfere with steps preceding or following endocytosis that require conformational changes. This method may have utility for the characterization or rapid screening of other Ab that act to prevent conformational changes or unfolding as part of their mechanism of action.

Introduction

For antibodies developed to treat or prevent infectious diseases or poisoning by protein toxins, the location of the antibody's (Ab) binding site (epitope) can, in some cases, explain its protective effects. For example, an Ab may bind a toxin or pathogen to block or compete with a cell surface receptor and, thereby, reduce or prevent internalization. There are many cases, however, in which the mechanism of action (MOA) of an Ab is not obvious. For toxins, as well as viruses, Abs recognizing the same antigen can be segregated into 2 categories, neutralizing and non-neutralizing, suggesting that some Abs exert additional effects beyond simple binding. Understanding how Ab-mediated toxin neutralization occurs is significant because it can lead to improved toxin-neutralizing Ab-based diagnostics, engineered vaccines, and new small molecule therapeutic strategies.

The assessment of the protective effects of Abs traditionally has relied upon dissociation constants (Kd), cell-based assays, and animal studies. Epitopes are categorized as protective or non-protective and the assignment of the epitope to either category is empirically determined. In some cases, non-protective “decoy” epitopes can be found adjacent to protective epitopes, and so the location of an epitope alone is not used to determine if an Ab is protective or non-protective. Even Abs that competitively bind to overlapping epitopes can still differ in their toxin-neutralizing activities. A simple biophysical measurement that can functionally distinguish between a highly protective neutralizing Ab versus a moderate or weakly neutralizing Ab would thus be useful for identifying the different types of Abs elicited by vaccination. Without a biophysical parameter, the selection of a therapeutic Ab or the combinatorial formulation of an Ab cocktail relies on trial and error, requiring a large number of cell-based assays and in some cases animal studies to determine the most effective combinations. For highly toxic Select Agents, a biophysical method that uses a toxoid also may reduce the need and associated costs for experiments conducted under biosafety level-2/3 (BSL2/3) containment.

Research and development of improved therapeutics for ricin intoxication is ongoing. Ricin is a ribosome inactivating protein (RIP) toxin. Its A-chain (RTA) catalyzes the depurination of A4324 in the α-sarcin/ricin GAGA tetra loop of the 28S rRNA. The B-chain (RTB) is a lectin that binds galactose residues on the cell surface. Since RTA is a highly efficient catalyst, only a few internalized RTA molecules are required to inactivate all of the ribosomes within a target cell. The depurination of the rRNA leads to inhibition of protein synthesis, apoptosis and ultimately cell death.

RTA enters the cytosol by retrograde transport. After endosomal uptake RIPs can be destroyed in the lysosome or transported in early endosomes to the trans-Golgi network and then to the endoplasmic reticulum (ER). To reach the cytosol, RTA exploits the ER-associated protein degradation (ERAD) pathway. Unfolded proteins in the ER are normally transported into the cytosol for ubiquitination followed by degradation by the proteasome. A small fraction of RTA in the ER dislocates into the cytoplasm and refolds to a catalytically active enzyme in the cytosol.

Several anti-RTA Abs elicited by RTA subunit vaccine immunogens are known to protect against intoxication in animals. These immunogens do not contain the RTB lectin domain. Two RTA subunit vaccines, RiVax and RTA1-33/44-198 (also called RVEC), have been tested as vaccine immunogens in Phase 1 clinical studies (Fig. 1). RTA1-33/44-198 was shown to protect mice against 10 LD50 of ricin delivered intranasally. Additionally, the neutralizing activities of single-domain antibodies (sdAb) and antigen (Ag)-binding fragments (Fab) in cell-based assays are independent of the Fc domain. It has been suggested that anti-RTA Abs may work by direct inhibition of RTA enzymatic activity; however, an anti-RTA Ab able to inhibit the enzymatic activity was found to be non-neutralizing. Moreover, Abs that enter into the cell during endocytosis are compartmentally separated from ribosomes in the cytosol by a membrane, and likely do not prevent or interfere with RTA substrate binding within the cytoplasm to exert their protective effects. So how do these anti-RTA Abs protect against ricin intoxication?

Figure 1.

Ricin is an A-B toxin: the catalytic A-chain (RTA) is shown in ribbon, and the RTB lectin domain in worm (PDB 2AAI). Glycans are shown in stick. The RiVax vaccine is an inactive RTA variant containing 2 mutations, V76M and Y80A (PDB 3SRP). RTA1-33/44-198 R48C/T77C is a thermostabilized RTA variant created by truncating RTA at residue 198, removing a loop between residues 34–43, and incorporating a disulfide (R48C/T77C) (PDB 3LC9). The structure of RTA1-33/44-198 in complex with the toxin-neutralizing sdAb A3C8 (PDB 5SV4) is shown on the right. The epitope (N97-F108) bound by the most potent neutralizing anti-ricin antibody (UNIVAX 70/138, also known as R70) is shown in hot pink. The epitope of sdAb A3C8 overlaps with the immunodominant UNIVAX epitope. The immunodominant human B-cell epitope identified by Castelletti et al. (L161-I175) is shown in green, and a T-cell epitope is shown in blue.

We hypothesize that some toxin-neutralizing Abs may block the unfolding of RTA or prevent a conformational change within RTA and thereby inhibit intoxication. RTA's ability to unfold and refold is central to its currently accepted mechanism of target cell entry and intoxication. Like Ab binding, the nucleation of protein unfolding is thought to occur at specific locations in the 3 dimensional structure. Identification and conformational restriction of these regions has been exploited for the stabilization of enzymes and proteins. The melting temperature (Tm) of a protein reflects a protein's intrinsic propensity to unfold. Thermal denaturation of a protein or protein-protein complex can be monitored by circular dichroism (CD), the mid-point of the transition yields the apparent Tm. Thermodynamic analysis of folding-unfolding requires that the 2 states coexist at the Tm with equal populations (ΔG = 0). Since the unfolding of RTA yields a non-equilibrium condition by way of aggregation, the ΔG at the Tm is not zero. Both thermal and chemical denaturation of RTA is an irreversible process. Thus, the measured Tm values of RTA and its complexes are denoted as apparent Tm values, and should not be interpreted as a thermodynamic measure of stability. Nevertheless, the apparent Tm's can still yield useful information regarding the relative stability of these proteins and protein-protein complexes. Both CD and differential scanning calorimetry are label-free methods that have been used to measure the Tm of proteins, protein-protein complexes, and protein-small molecule complexes in solution. The Tm can be used to examine the stabilizing effects of conformationally restrictive mutations or, in this case, conformationally restrictive Ab binding.

We previously incorporated disulfide bonds within specific regions of RTA predicted to be involved in the early stages of unfolding based partly upon in silico predictions. The elevation of the Tm of a protein can be practically useful during manufacturing and for the development of long shelf-life vaccine immunogens, or for the development of field-expedient protein-based assay reagents that do not require refrigeration. Incorporation of a disulfide bond at 2 sites within RTA led to 2 engineered immunogens with enhanced stability as judged by a higher apparent Tm. Disulfides incorporated at 7 other sites had little or no effect on the Tm. Our findings with engineered disulfides suggest that specific structural regions within RTA may play a preferential role in controlling overall protein stability. Likewise, earlier work showed that the introduction of an intrachain disulfide bond within RTA (S215C/M255C) could interfere with ricin toxicity. We hypothesized that, if partial unfolding of RTA is necessary for toxin internalization, toxin-neutralizing Abs may act like our engineered disulfides by locally stabilizing the folded structure and, thereby, interfering with intoxication.

We propose here that the magnitude of the Tm-shift induced by an Ab may be useful in predicting the toxin-neutralizing activities of anti-RTA Abs that prevent conformational change or unfolding. To test this hypothesis, we measured the Tm of RTA and compared it with the Tm of RTA in complex with a set of previously identified neutralizing and non-neutralizing sdAb at low pH (to mimic the endosomal environment, pH 4.5) or at pH 7.4 (to mimic the extracellular environment). We found that the magnitude of the Tm shift induced by the Ab was highly correlated with the level of protection in cell-based assays. These results suggest a potential MOA for anti-RTA neutralizing Abs, and identify a simple biophysical parameter that may be useful in conjunction with cell-based assays for quantitatively assessing and comparing other protective Ab or small molecules as experimental therapeutics.

Results

Structure of the A3C8 single-domain antibody

The A3C8 sdAb was first crystallized. Data collection and refinement statistics are shown in Table 1. A3C8 was found to be monomeric using a calibrated gel-filtration column (Table 2). The structure of the A3C8 sdAb (PDB 5SV4) was used to solve the structure of the sdAb-RTA1-33/44-198 complex by molecular replacement.

Table 1.

X-ray crystallography data collection and refinement statistics.

	sdAb A3C8	A3C8:RTA1-33/44-198 Complex
PDB	5SV4	5SV3
Space group	P2(1)2(1)2(1)	P2(1)2(1)2
Unit Cell Dimensions (Å)	35.88, 44.40, 143.77	35.88, 117.76, 130.43
Wavelength (Å)	1.54	1.54
Resolution Range (Å)a	47.92–2.70 (2.80–2.70)	65.22–2.73 (2.83–2.73)
Unique Reflections	6,823 (689)	15,475 (1,536)
Rsymb	0.166 (0.445)	0.187 (0.454)
I/σI	12.2 (3.5)	11.9 (4.5)
Completeness	100.0 (100.0)	99.9 (100.0)
Redundancy	11.5 (8.2)	14.0 (14.2)
Refinement Statistics:
Resolution (Å)	47.92–2.70	65.22–2.73
No. of reflections	6,455	14,603
Rfactorc	0.201	0.228
Rfreed	0.246 (5%)b	0.267 (5%)b
Number of Atoms:
Protein	1922	4538
Solvent	60	78
Other	10	10
Average B-factors (Å2)
Protein	18.1 (A), 36.7 (B)	29.2(A), 22.9(B), 22.0(C), 32.5(D)
Solvent	18.1	32.1
R.m.s.d. from ideal geometry:
Bond lengths (Å)	0.010	0.007
Bond angles (degrees)	1.482	1.17
Ramachandran plot
Most favored regions (%)	93.6%	93.0%
Additional allowed regions (%)	6.4%	7.0%
Generously allowed regions (%)	0.0%	0.0%
Disallowed regions (%)	0.0%	0.0%

Values in parentheses are for the outer most data shell

Rfree for test set and size of test set as % total reflections in parentheses.

Table 2.

Apparent Tm values of SdAb, RTA, RTA1-33/44-198, and their complexes in 1x PBS pH 4.5 or 7.4. Gel-filtration columns were calibrated, and elution volumes are shown for the free proteins and their complexes. The MW has been calculated based upon the protein sequence.

Protein	Type of Antibody	Kd (M)	Apparent Tm (°C)	ΔTm (°C)	pH	Wavelength (nm)		Elution Volume (mL)	Calculated MW (kDa)
Ricin A-Chain Constructs
RTA	—	—	52.2 ± 0.1	—	4.5	222	G75	11.0	30.112
RTA	—	—	51.6 ± 0.7	—	7.4	222	G200	16.1	30.112
RTA1-33/44-198	—	—	57.6 ± 0.5	—	4.5	214	G75	13.0	22.595
RTA1-33/44-198	—	—	55.5 ± 0.1	—	7.4	222	G200	16.3	22.595
Anti-RTA Antibodies
A3C8	Neutralizing	—	66.2 ± 0.4	—	4.5	207, 222	G75	12.3	15.063
A3C8	Neutralizing	—	70.7 ± 0.1	—	7.4	222	G200	16.6	15.063
C8	Neutralizing	—	58.1 ± 0.5	—	4.5	207, 222	G75	12.5	13.576
C8	Neutralizing	—	60.5 ± 0.3	—	7.4	207, 222	G200	16.8	13.576
D10	Neutralizing	—	66.8 ± 0.4	—	7.4	207, 222	G200	17.0	13.999
D12f	Neutralizing	—	78.8 ± 0.4	—	7.4	222	G200	16.6	14.224
H1W12V	Non-Neutralizing	—	67.0 ± 0.4	—	7.4	207	G200	16.6	14.766
C10neg	Non-Neutralizing	—	77.9 ± 1.2	—	7.4	207	G75	13.2	13.941
F6	Non-Neutralizing	—	76.3 ± 0.4	—	7.4	207	G75	12.5	14.942
E1	(Non-Neutralizing)a	—	66.4 ± 0.5	—	7.4	207, 222	G75	12.7	14.723
RTA1-33/44-198-Antibody Complexes
RTA1-33/44-198:A3C8	Neutralizing	1.48 E-08	66.6 ± 0.5	9.0	4.5	214	G75	11.1	37.658
RTA1-33/44-198:A3C8	Neutralizing	7.22 E-10	68.3 ± 0.2	12.8	7.4	222	G200	14.8	37.658
RTA1-33/44-198:A3C8	Neutralizing	7.22 E-10	69.3 ± 0.5	13.8	7.4	222	G75	10.8	37.658
RTA1-33/44-198:C8	Neutralizing	1.55 E-08	64.2 ± 0.1	6.6	4.5	214	G75	11.2	36.171
RTA1-33/44-198:C8	Neutralizing	1.40 E-10	69.87 ± 0.03	14.4	7.4	222	G75	10.9	36.171
RTA1-33/44-198:D10	Neutralizing	1.75 E-10	71.1 ± 0.2	15.6	7.4	222	G75	10.9	36.594
RTA1-33/44-198:D12f	Neutralizing	1.16 E-10	75.0 ± 0.3	19.5	7.4	222	G75	10.9	36.819
RTA1-33/44-198:H1W12V	Non-Neutralizing	—	No Complex	—	7.4	—	G75	12.1	37.361
RTA1-33/44-198:C10neg	Non-Neutralizing	—	No Complex	—	7.4	—	G75	12.2,13.4	36.536
RTA1-33/44-198:F6	Non-Neutralizing	3.13E-08	59.9 ± 0.2	4.4	7.4	222	G75	11.1	37.537
RTA1-33/44-198:E1	(Non-Neutralizing)a	3.58 E-09	69.6 ± 0.3	14.1	7.4	222	G75	10.8	37.318
RTA-Antibody Complexes
RTA:A3C8	Neutralizing	4.32 E-09	58.3 ± 0.3	6.1	4.5	222	G75	9.7	45.175
RTA:A3C8	Neutralizing	4.6 E-10	60.3 ± 0.7	8.7	7.4	222	G75	10.5	45.175
RTA:A3C8							Neutralizing	4.6 E-10	61.0 ± 1.1	9.4	7.4	222	G200	14.5	45.175
RTA:C8	Neutralizing	5.98 E-09	58.9 ± 0.2	6.7	4.5	214	G75	9.8	43.688
RTA:C8	Neutralizing	1.63 E-10	71.9 ± 0.3	20.3	7.4	222	G200	14.8	43.688
RTA:D10	Neutralizing	1.44 E-10	66.4 ± 0.4	14.8	7.4	222	G200	14.8	44.094
RTA:D12f	Neutralizing	2.27 E-11	67.7 ± 0.6	16.1	7.4	222	G200	14.8	44.336
RTA:H1W12V	Non-Neutralizing	5.18 E-10	57.2 ± 0.6	5.6	7.4	222	G200	14.6	44.878
RTA:C10neg	Non-Neutralizing	—	No Complex	—	7.4	222	G75	11.1, 13.1	43.851
RTA:F6	Non-Neutralizing	6.04 E-09	57.1 ± 0.4	5.5	7.4	222	G75	10.4	45.054
RTA:E1	(Non-Neutralizing)a	1.25 E-09	58.4 ± 0.3	6.8	7.4	222	G75	10.5	44.835

aThis antibody is reported to be a non-neutralizing, but has also been reported to compete with a neutralizing antibody

Structure of the A3C8:RTA1-33/44-198 complex

Previously, we crystallized 2 thermostabilized variants of RTA1-33/44-198 containing disulfide bonds (Fig. 1); however, we were not able to crystallize the parent protein, RTA1-33/44-198, alone. RTA1-33/44-198 and its variants contain a deletion of loop residues 34–43; these hydrophobic loop residues pack with the C-terminal 199–267 residues (Fig. S1). The removal of residues 34–43 and 199–267 was previously shown to increase the Tm by 8 degrees compared with RTA.

Co-crystallization of RTA1-33/44-198 with sdAb A3C∼8 enabled us to solve the first structure of the original RTA1-33/44-198 variant and examine the conformation of the sdAb-bound antigen. When sdAb-bound RTA1-33/44-198 was compared with the disulfide bonded variants (PDB 3MK9 and 3LC9), conformational differences were observed in 3 regions: residues 33–52; residues 131–136 and residues 182–188 (Fig. S2). The differences in residues 33–52 are likely due to the disulfide bonds that were incorporated at R48C/T77C and V49C/E99C. In the sdAb A3C8:RTA1-33/44-198 complex, the conformation of the loop residues 131–136 was more similar to that of RTA in the ricin holotoxin (PDB 2AAI), than to the conformation found in the disulfide-bonded variants (PDB 3LC9, 4IMV, and 3MK9). In the structures of the disulfide variants, the conformation of loop residues 131–136 found in RTA (PDB 1J1M) would clash with an adjacent molecule at the crystal packing interface. We had previously shown that the B-factors of residues 131–136 were higher in the disulfide-bonded variants compared with RTA, and that the density for the loop was poor, suggesting static or dynamic disorder of these residues. In the A3C8:RTA1-33/44-198 complex (PDB 5SV3), these residues also were involved in a crystal packing interface, but had well-defined density while residues 188–198 (near the 33–44 loop) appeared disordered. The C-terminal domain packs with the 131–136 loop residues and residues 182–188. The C-terminal domain and residues 182–188 contact RTB (Fig. 1, Fig. S1, S2). The conformation of these loop residues may be stabilized by the C-terminal domain residues 199–267 or by sdAb A3C8 binding. However, a structure of the free RTA1-33/44-198 is still needed to confirm this effect.

Effects of the Q98A, Q112A and R114A mutations on sdAb A3C8 binding

The epitope recognized by A3C8 was initially identified using the crystal structure of the complex. The epitope then was confirmed by site-directed mutagenesis, Western blotting, and surface plasmon resonance (SPR; Fig. 2).

Figure 2.

SdAb A3C8 epitope identification. (A) Western analysis to identify residues important for Ab binding using mutants of RTA. His-tagged sdAb A3C8 was used as the primary Ab. A mouse anti-His-tag Ab was used as the secondary Ab, and an anti-mouse alkaline phosphatase linked Ab was used for detection. (B) SdAb A3C8 Ab-Ag interface (PDB 5SV3) showing the location of 3 mutations (pink spheres): Q98A, Q112A, and R114A. (C) Apparent Ab on and off rates measured by SPR for RTA1-33/44-198 and the Q98A, Q112A, and R114A variants. Tight binding by sdAb A3C8 was not observed to the R114A variant by SPR or Western analysis.

A C-terminal His-tagged sdAb and anti-His-tag IgG were used to detect RTA and its variants in Western blot analysis. Antigens were separated by SDS-PAGE. Based upon the crystal structure (PDB 5SV3), 3 residues were selected for mutation: Q98A, Q112A, and R114A. Other previously characterized disulfide-bonded RTA1-33/44-198 variants were included: R48C/T77C; V49C/E99C; A90C/V111C; R125C/A165C; E138C/P143C; N141C/I192C; and R48C/T77C/D75N. A total of 10 RTA1-33/44-198 variants were examined in addition to recombinant RTA and RTA purified from Ricinus communis. Staining was observed for all of the RTA1-33/44-198 variants (0.5 µg per lane) with the exception of R114A, indicating that R114A mutation disrupted sdAb A3C8 binding. The Q112A variant showed slightly reduced staining.

SPR was used to examine changes in affinity to the epitope since multiple contacts were observed between the antibody and antigen (Table 3). For sdAb A3C8, the R114A mutation led to the most significant increase in the dissociation constant (49-fold higher Kd), and the Q112A mutation lead to a ˜2-fold higher Kd compared with RTA1-33/44-198.

Table 3.

Dissociation constants measured by SPR for sdAb and RTA and its variants. Binding of the A3C8 antibody was disrupted by the R114A mutation in both Western blots and SPR experiments. DB denotes ‘disrupted binding’: the response units (y-axis) for these mutants were <10% of the parent protein and the on and off rates could not be accurately determined for the mutant. Dashes indicate that the change in Kd was < 2-fold from the parent protein. The disruption mutations were used to identify the general locations of the epitopes recognized by these Abs.

	sdAb	kon (1/M*sec)	koff (1/sec)	Kd (M)	Fold Difference in Kd (RTA vs. Mutants)	Fold Difference in Kd (RTA1-33/44-198 vs. Mutants)
RTA	A3C8	700,000	0.000263	4.6 × 10−10	—
RiVax	A3C8	974,000	0.000205	2.11 × 10−10	—
RTA 1-33/44-198	A3C8	552,000	0.000346	6.27 × 10−10	—	—
RTA 1-33/44-198, Q98A	A3C8	437,000	0.000306	7.01 × 10−10		—
RTA 1-33/44-198, Q112A	A3C8	483,000	0.000484	1.00 × 10−9		—
RTA 1-33/44-198, R114A	A3C8	531,000	0.0164	3.09 × 10−8		49-fold
RTA	C8	1,320,000	0.000215	1.63 × 10−10	—
RiVax	C8	1,800,000	0.000205	1.14 × 10−10	—
RTA 1-33/44-198	C8	1,090,000	0.000152	1.40 × 10−10	—	—
RTA 1-33/44-198, Q98A	C8	1,210,000	0.000228	1.89 × 10−10		—
RTA 1-33/44-198, Q112A	C8	836,000	0.000224	2.68 × 10−10		—
RTA 1-33/44-198, R114A	C8	DB	DB			≥100-fold
RTA	D10	396,000	0.0000570	1.44 × 10−10	—
RiVax	D10	539,000	0.0000979	1.82 × 10−10	—
RTA 1-33/44-198	D10	416,000	0.0000729	1.75 × 10−10	—	—
RTA 1-33/44-198, Q98A	D10	439,000	0.0000688	1.57 × 10−10		—
RTA 1-33/44-198, Q112A	D10	500,000	0.0000915	1.83 × 10−10		—
RTA 1-33/44-198, R114A	D10	732,000	0.000183	2.50 × 10−10		—
RTA	D12f	418,000	0.00000948	2.27 × 10−11	—
RiVax	D12f	476,000	0.0000149	3.12 × 10−11	—
RTA 1-33/44-198	D12f	350,000	0.0000407	1.16 × 10−10	5-fold	—
RTA 1-33/44-198, Q98A	D12f	388,000	0.0000286	7.38 × 10−11		—
RTA 1-33/44-198, Q112A	D12f	447,000	0.0000307	6.88 × 10−11		—
RTA 1-33/44-198, R114A	D12f	680,000	0.00324	4.77 × 10−9		41-fold
RTA	F6	197,000	0.00119	6.04 × 10−9	—
RiVax	F6	167,000	0.00128	7.65 × 10−9	—
RTA 1-33/44-198	F6	358,000	0.0112	3.13 × 10−8	5-fold	—
RTA 1-33/44-198, Q98A	F6	313,000	0.0116	3.70 × 10−8		—
RTA 1-33/44-198, Q112A	F6	371,000	0.0128	3.45 × 10−8		—
RTA 1-33/44-198, R114A	F6	DB	DB			≥100-fold
RTA	E1	188,000	0.000235	1.25 × 10−9	—
RiVax	E1	194,000	0.000194	1.00 × 10−9	—
RTA 1-33/44-198	E1	296,000	0.00106	3.58 × 10−9	3-fold	—
RTA 1-33/44-198, Q98A	E1	279,000	0.00105	3.76 × 10−9		—
RTA 1-33/44-198, Q112A	E1	323,000	0.00118	3.67 × 10−9		—
RTA 1-33/44-198, R114A	E1	414,000	0.00173	4.18 × 10−9		—
RTA	H1W12V	8,620,000	0.00447	5.18 × 10−10	—
RiVax	H1W12V	100,000,000	0.0408	4.08 × 10−10	—
RTA 1-33/44-198	H1W12V	DB	DB		≥100-fold	—
RTA 1-33/44-198, Q98A	H1W12V	DB	DB			≥100-fold
RTA 1-33/44-198, Q112A	H1W12V	DB	DB			≥100-fold
RTA 1-33/44-198, R114A	H1W12V	DB	DB			≥100-fold

Ab binding to the denatured RTA was consistent with binding to a linear epitope (Fig. 2A). However, binding to an N-terminally biotinylated-peptide that contains residues H94-T116 (94HPDNQEDAEAITHLFTDVQNRYT116) and the UNIVAX R70 epitope (underlined) was not detectable by SPR, suggesting that the Ab may make an important contact to residues outside of this region. A3C8 carries the same complementarity-determining region (CDR) sequences as C8, but has a different framework. A3C8 bound less well to RTA than C8 (Table 3) suggesting that the framework residues may favor particular side-chain conformations of CDR residues involved in binding or alter the mobility of the CDRs.

Epitope mapping of other anti-ricin sdAb: D12f, D10, E1, F6, C10neg and H1W12V

We also identified the general locations of the epitopes of the other sdAb tested. The Kd values measured for RTA, RiVax, RTA1-33/44-198, and the 3 variants are summarized in Table 3. The structure of the D10:RTA complex had been previously determined, and the Q98A, Q112A and R114A mutations did not have significant effects on binding affinity consistent with the structure (Fig. 3E). Like sdAb A3C8, the D12f sdAb had markedly reduced affinity (41-fold higher Kd) for the R114A variant, suggesting that the epitopes of the neutralizing D12f and A3C8 sdAb overlapped (n.b. sdAb A3C8 and C8 contain the same CDRs).

Figure 3.

(A) The B-cell and T-cell epitopes are shown for reference in this view and are colored as in Fig. 1. (B) Five conformations of modeled RTA1-33/44-198 from replica-exchange molecular dynamics simulations culled at 298 K with the previously identified regions involved in the early stages of unfolding are shown in blue, red and green. The 2 immunodominant B-cell epitopes are located within 2 of these regions. (C) Locations of 2 disulfide bonds previously engineered into these regions to thermostabilize the protein. A residue within the blue region was anchored to the core of the protein in the R48C/T77C variant (PDB 3LC9). (D) Two residues in the blue and green regions were anchored to one another in the V49C/E99C variant (PDB 3MK9). The R48C/T77C and V49C/E99C mutations both increased the Tm of the protein by 5 degrees relative to the parent construct RTA1-33/44-198, and by 9 degrees relative to RTA. (E) Comparison of the A3C8 structure (PDB 5SV3) with 5 other sdAb-RTA complex structures previously reported by Rudolph, et al. Neutralizing anti-ricin Abs were found bound to 2 of the 3 regions involved in the early stages of unfolding. The strongly neutralizing Abs (lime green) bound to the red (residues 140–170) or green (residues 90–110) regions of RTA. Weakly neutralizing and non-neutralizing sdAb are shown in pink. A weakly neutralizing Ab (G11, PDB 4LHJ) bound to a region distant from the red, green and blue regions of RTA and had an IC50 which was estimated to be higher than that of sdAb A3C8. (F) Four neutralizing and non-neutralizing Abs appeared to bind to a similar region, however the most potent toxin-neutralizing Abs (lime, E5 PDB 4LGP and A3C8 PDB 5SV3) were found to pack with the green helical region (residues 90–110) of RTA, while the non-neutralizing (“decoy”) and weakly neutralizing Abs (pink, PDB 4LGS, 4LHQ) left this helix exposed to solvent. Similarly, the D10 sdAb packed with the helix-turn-helix motif (red, residues 140–170) more than the non-neutralizing “decoy” Abs A7 and G12.

The H1W12V Ab, which is non-neutralizing, had significantly reduced affinity for RTA1-33/44-198 (>100-fold higher Kd) and the 3 RTA1-33/44-198 variants, suggesting that the epitope is within the C-terminal domain residues, 199–267, or within the loop (residues 34–43). The loop (residues 33–44) packs with the C-terminal domain (Fig. S1). The weakly neutralizing G11 sdAb, for which a structure is known, is an example of an sdAb that can bind the C-terminal domain (Fig. 3E). The non-neutralizing Ab, C10neg, did not bind to RTA alone. This Ab is thought to bind to regions of both the A- and B-chains of ricin. The C-terminal domain of RTA packs with RTB (Fig. 1). The F6 sdAb, which also is a non-neutralizing Ab, had slightly reduced binding affinity to RTA1-33/44-198 (5-fold), and showed no significant binding to the R114A variant, suggesting that the F6 epitope may be near that of sdAb A3C8. The epitope of F6 may be near the epitope recognized by the non-neutralizing sdAbs A7 and G12. The E1 sdAb, which is also non-neutralizing, was not significantly affected by the Q98A, Q112A or R114A mutations or by the C-terminal domain truncation.

Analysis of linear regression models and comparison of mean values

We assembled a set of neutralizing and non-neutralizing Abs. The sdAb was classified as neutralizing if >50% cell viability was obtained in the cell-based assays using the specified concentrations of ricin and sdAb. In cell-based assays using the ricin holotoxin, 50% cell viability corresponded to sdAb with IC50 values ≤ 80 nM. The sdAb were classified as non-neutralizing if <50% cell viability was obtained or the Ab had IC50 values > 80 nM. For the 4 non-neutralizing sdAb used in this study, the IC50 values exceeded the highest concentration of ricin tested (5 µM) (Table 4). The A3C8 sdAb is an engineered thermostabilized variant containing the CDRs of the C8 sdAb and the framework regions of a thermostable anti-SEB (Staphylococcal enterotoxin B) Ab called A3 (Tm = 85°C). The A3C8 sdAb has a Tm that is 10°C higher than that of C8. The A3C8 variant binds the same epitope as C8, but with 3-fold lower affinity than C8 (Table 2). This difference allowed us to examine the effects of Kd. The binding of D12f was affected by the R114A mutation and its epitope likely overlaps with that of C8 and A3C8. SdAb D10 is also a neutralizing Ab; its epitope is known and differs from that of A3C8, C8, and D12f (Fig. 3). For non-neutralizing Abs, we selected H1W12V, C10neg, F6 and E1. The E1 and H1W12V sdAbs recognize an epitope within the C-terminal domain residues 199–267, and the C10neg sdAb recognizes both RTA and RTB. To find correlations between the toxin-neutralizing activities of the sdAb and various biophysical parameters, we performed linear regression analysis and examined the r2-values (Fig. 5). We also compared the difference between the mean values of the biophysical parameters relative to the variation in the sample data, and p-values were calculated from T-values. In our cell viability assays, all of the sdAb concentrations were above the Kd values (i.e., stoichiometric binding regime). All experiments were carried out in triplicate, and the averages and standard deviations are summarized in Table 4. Melting temperatures were measured by monitoring changes in secondary structure as the temperature was raised using a CD spectrometer (Table 2, Fig. 4).

Table 4.

Neutralizing activities of anti-RTA sdAb in cell-based assays and in cell-free translation assays. Results are expressed as the percent of viable control cells ([experimental/untreated control] × 100) and are the average of 3 experiments. The standard deviations are shown. Antibodies yielding <50% viable cells were not considered to be ‘toxin-neutralizing’. For cell-free translation assays results are expressed as percent control ([experimental/untreated control] × 100) and are the average of 3 experiments; standard deviations are shown.

Antibody	Cell Viability (%)	IC50 (nM)	Ribosomal Activity in a Cell-free Translation Assay (%)	Tm Shift (°C)	Kd (M)	Tm Antibody (°C)
Neutralizing
C8	70 ± 4	40	49 ± 3	20.3	1.63 E-10	60.5
D12f	66 ± 4	30	60 ± 4	16.1	2.27 E-11	78.8
D10	63 ± 4	53	57 ± 2	14.8	1.44 E-10	66.8
A3C8	54 ± 2	80	44 ± 4	8.7	4.6 E-10	70.7
Non-Neutralizing
H1W12V	18 ± 4	>5000	27 ± 4	5.6	5.18 E-10	67
C10neg	15 ± 3	>5000	14 ± 3	—a	—a	77.9
F6	12 ± 3	>5000	24 ± 2	5.5	6.04 E-09	76.3
E1	7 ± 3	>5000	14 ± 4	6.8	1.25 E-09	66.4
Controls
Cells (No Toxin)	100 ± 3		100 ± 4	—	—	—
PBS	5 ± 2		5 ± 3	—	—	—

Antibody recognizes RTA/RTB and does not form a stable complex with RTA alone.

Figure 5.

Correlations between the Tm-shift, % Inhibition of translation, and Kd with cell viability or IC50. (A) Comparison of the % inhibition of translation by neutralizing (n = 4) and non-neutralizing (n = 4) sdAbs yielded a p-value = 0.0003 for a one-tailed hypothesis (T-value = 6.57). The significance threshold was set at 0.05. A linear fit of all of the data points yielded an r2-value = 0.910. While the difference was significant, a mechanism for Ab-bound RTA entry into the cytoplasm has not been reported, and an Ab which inhibits RTA's enzymatic activity but does not neutralize cytotoxicity has been reported. (B) Comparison of the Tm-shifts induced by neutralizing (n = 4) and non-neutralizing sdAb (n = 3) yielded a p-value = 0.0127 (T-value = 3.15) for a one-tailed hypothesis. The significance threshold was set at 0.05. For the neutralizing sdAb, a linear fit of the Tm-shifts vs. the % cell viability (n = 4) yielded an r2-value = 0.992 suggesting a strong correlation between the 2 parameters. Notably, when the stability of the RTA antigen was altered (i.e. RTA 1-33/44-198 was used in place of RTA), the correlation between the Tm-shift and the % cell viability was lost (r2-value = 0.18) suggesting that a property of the Ag-Ab complexes had been perturbed. (C) Comparison of the Kd vs. cell viability using a 2-tailed hypothesis yielded a p-value = 0.16 indicating that the Kd does not significantly correlate with the effects of the sdAb in cell-based assays. The significance threshold was set at 0.05. (D) The Tm-shift also correlated well with the IC50 (nM) values obtained from cell based assays.

Figure 4.

Toxin-neutralizing antibodies (shades of red, pink, and purple) shift the apparent Tm of the Ab-Ag complex more significantly than do non-neutralizing (shades of blue) Abs. Thermal denaturation in 1x PBS pH 7.4 was monitored using a CD spectrometer fitted with a Peltier temperature control unit. In (A), the thermal denaturation of the RTA-sdAb complexes are shown. In (B), the effect is perturbed when RTA is substituted by RTA1-33/44-198 in the sdAb-Ag complexes. RTA1-33/44-198 is a variant of RTA that has a reduced propensity to unfold under thermal stress (higher apparent Tm).

Cell viability vs. percent inhibition of translation

We observed a correlation between cell viability vs. percent inhibition of translation (r2 = 0.912). The T-value was 6.5747 and p-value 0.000297; however, the correlation is not necessarily related to the MOA of anti-RTA Abs (Fig. 5A). This coincidence has been noted by others, and may be due to the location of active site and substrate binding residues within the neutralizing epitopes (Fig. S3). Residues 90–110 contain the UNIVAX R70 epitope (N97-F108) and form a portion of the substrate binding site. Similarly, the B-cell epitope (L161-I175) forms a portion of the substrate binding site. However, while ricin can enter into the Golgi, ER and cytoplasm, the anti-ricin Abs likely do not unfold and translocate into the cytoplasm via retrograde transport (i.e., ERAD system) to inhibit the enzymatic activity of RTA. Antibodies are typically separated from their cargo and recycled via the neonatal Fc receptor. Notably, a non-neutralizing anti-RTA antibody known as mAb RAC23 that can inhibit RTA's enzymatic activity, but fails to protect in cell-based assays, has been reported. RAC23 actually enhanced the toxicity of ricin in a mouse model. The neutralizing anti-RTA mAb called RAC18 led to the accumulation of ricin at the cell surface, suggesting that this neutralizing anti-RTA Ab halts an early step(s) in retrograde transport, rather than RTA substrate binding or catalysis in the cytoplasm. Comparison of sdAb A3C8 with C8 showed that both sdAb comparably inhibited cell-free translation, and that inhibition in cell-free translation could not distinguish between the 2 sdAb. So, while neutralizing and non-neutralizing Abs could be distinguished from one another in this graph, a MOA involving the inhibition of the enzymatic activity in the cytoplasm by an Ab is not supported by Abs such as RAC23.

Tm shift vs. cell viability

A strong correlation between cell viability and the magnitude of the Tm shift was observed for neutralizing Abs and yielded a coefficient of determination r2 = 0.992 and a linear relationship with a positive slope of 0.65 ± 0.05 (ΔTm/Δ% cell viability) and a negative y-intercept (−26 ± 3); the x-intercept corresponded to 40% cell viability and was near the 50% cutoff for neutralizing vs. non-neutralizing Abs (Fig. 5B). A similar correlation was obtained with IC50 values measured by cell-based assays and Tm shifts (r2 = 0.846) (Fig. 5D). The ranges on the x- and y-axes were relatively large compared with the other parameters, and the points were well dispersed along the line. Excluding the engineered A3C8 sdAb, the mean values of the Tm-shifts for the neutralizing vs. non-neutralizing Abs were compared and produced a T-value = 6.48 and a p-value of 0.00146 (one-tailed), showing that the magnitude of the Tm-shifts for the neutralizing Abs was significantly higher than those of the non-neutralizing Abs. With the inclusion of A3C8, the T-value was 3.15 and the p-value was 0.0127 (one-tailed), showing that the mean value of the Tm-shifts obtained for the neutralizing Abs still was significantly different from that of the non-neutralizing Abs. Thus, the magnitude of the Tm-shift could be used to distinguish between neutralizing and non-neutralizing Abs. Notably, the magnitude of the Tm-shift rank-ordered the sdAb in the same manner as the cell-based assay data.

The apparent Tm values of the Ab-Ag complexes containing RTA (residues 1–267, 30 kDa) were well correlated with the ability of the sdAb to protect cells from intoxication during exposure to the 59 kDa A-B ricin toxin. However, a much poorer correlation was observed as the stability of the ricin A-chain antigen was altered (i.e., RTA1-33/44-198 was used in place of RTA), suggesting that a property of the Ab-Ag complexes had been perturbed (Fig. 4). No significant correlation between cell viability and Tm-shift was observed with the thermostabilized RTA1-33/44-198 variant (r2 = 0.2836). RTA1-33/44-198 had been engineered to have a reduced propensity to unfold, and lacks the C-terminal domain and a loop formed by residues 34–43 that pack with the C-terminal domain. The C-terminal domain packs with RTB (Fig. 1, Fig. S2).

Tm shift vs. percent inhibition of translation

Analysis of all of the data points for both neutralizing and non-neutralizing sdAb. This resulted in only a weak correlation (r2 = 0.667) between these parameters (Table 4).

Tm shift vs. Kd

For neutralizing and non-neutralizing Abs, no correlation between Kd and Tm shift was observed (r2 = 0.295) (Table 4). The Tm reflects a protein's propensity to unfold. This result suggested that the location(s) at which the sdAb bind, rather than their binding affinity, may affect RTA's propensity to unfold. Previously, we had identified 3 regions that became disordered during the early stages of unfolding as the protein was heated in silico (Fig. 3B). We showed that we could elevate the Tm of RTA1-33/44-198 by disulfide bonding these regions to the core of the protein or to one another. The Tm values of 2 disulfide-bonded RTA1-33/44-198 variants were increased by 5 degrees (Fig. 3C, 3D). Disulfide bonds introduced at other locations did not significantly increase the Tm of the RTA1-33/44-198 construct. Other groups have empirically identified regions that affect unfolding by mutagenesis alone. In regards to Ab binding, residues within these 3 loop regions were found to be engaged by neutralizing Abs (Fig. 3E). Non-neutralizing Abs recognizing so-called “decoy” epitopes adjacent to these regions did not pack well with the 90–110 or 140–170 residues compared with the neutralizing A3C8 (PDB 5SV3) and E5 (PDB 4LGP) sdAb. The exposed surface area of these 3 regions differed in the neutralizing vs. non-neutralizing Ab complexes (Fig. 3F).

Cell viability vs. Kd

For RTA and anti-RTA Abs, the Kd was not a strongly predictive parameter of neutralizing activity (Fig. 5C). However, neutralizing activity was affected by Kd in some cases (e.g., C8 vs. A3C8). The Kd value has been shown to be a predictive parameter for other antibodies, and the correlation between Kd and protection may largely depend upon the MOA of the protective Ab. From competition experiments we know that D12f and C8 bind the same epitope or overlapping epitopes (Table 3). The Kd values for D12f and C8 differ by a factor of 10; however, the percentage of viable cells obtained for the weaker binding C8 sdAb was higher than that of the tighter binding D12f sdAb. The trend was reversed for A3C8 and C8. A3C8 and C8 have the same CDR sequences and Kd values that differ by 3-fold. A lower percentage of viable cells was obtained with the weaker binding A3C8 sdAb compared with the tighter binding C8 sdAb. The p-value confirmed that Kd alone could not be used to distinguish between neutralizing and non-neutralizing anti-RTA sdAb. For the neutralizing and non-neutralizing sdAb, the T-value was −1.6554 and the p-value was 0.07937 (one tailed). The spread and overlap are shown in (Fig. 5C).

Tm of the antibody vs. Tm of the complex

Thermostabilized Abs have been developed as robust affinity reagents for refrigeration-free use. Thus, we examined whether a high Tm Ab would be advantageous in preventing the unfolding of its cognate antigen as temperature was varied. Interestingly, there was no advantage in using a thermostabilized sdAb. This was most clearly demonstrated with sdAb A3C8 and C8. SdAb A3C8 has a Tm that is ∼10 degrees higher than that of sdAb C8. The Tm of the A3C8:RTA complex was ∼11 degrees lower than the Tm of the C8:RTA complex. The Tm of the Ab did not correlate with the Tm of the RTA-sdAb complex for neutralizing (r2 = 0.064) or non-neutralizing Abs (r2 = 0.361). The Tm values of the free neutralizing and non-neutralizing Abs were similar (p-value = 0.30), while the Tm values of the Ab-Ag complexes differed (Tables 2, 4).

Effect of pH on the sdAb-RTA complexes

After the A-B ricin toxin is endocytosed, RTA is separated from RTB shortly (∼minutes) after internalization. Endosomes play an important role in transporting and sorting molecules to different compartments in the cell. Endosomes acidify as they transit through the cell and Abs release their antigens. Crystal structures of RTA have been determined at both low (pH 4.2) and high (pH 8.9) pH indicating that RTA can remain stably folded in both acidic and basic conditions. To determine if the neutralizing sdAb still could remain bound and reduce RTA's propensity to unfold at low pH, we measured the Tm values of the purified RTA-sdAb complexes at pH 4.5 to mimic the endosomal environment. At pH 4.5 micromolar concentrations of the RTA:A3C8, RTA:C8, RTA1-33/44-198:A3C8 and RTA1-33/44-198:C8 complexes still could be purified by gel-filtration. The Kd values were ∼10–110-fold higher at pH 4.5 than at pH 7.4 (Table 2). The shifts in the Tm of the RTA:A3C8 vs. RTA:C8 complexes still were detectable at pH 4.5 (6.1 degrees and 6.7 degrees, respectively) (Table 2). The elution volumes differed for the sdAb, RTA, and the RTA-sdAb complexes, and were additional confirmation that the complexes had been purified (Fig. S4). For sdAb C8, the magnitude of the Tm shift decreased significantly from 20.3 °C (pH 7.4) to 6.7 °C (pH 4.5), suggesting that the sdAb may predominantly exert their effects at ˜neutral pH, but can remain bound to the antigen at low pH. The stability of the sdAb-RTA complexes at both pH may be important to their MOA.

Discussion

For ribosome inactivating proteins (RIPs) prior attempts to inhibit the enzymatic activity of the A-chains have had limited success, and no post-exposure small molecule therapeutics are available to reverse the effects of intoxication. For most enzymes, in vitro kinetic methods as well as thermal shift assays and other binding assays have been developed to facilitate the separation of potent from weakly inhibitory compounds. Comparatively few cell-free assays are available for screening therapeutic anti-toxin Abs. Improved cloning, expression, and selection methods also have increased the number of Ab-based therapies undergoing expensive human clinical trials, but there is a shortage of validated cost-effective methods that rank-order different Abs. Expensive and time-consuming cell-based assays and in vivo studies involving pathogens or Select Agent toxins must be performed to differentiate among neutralizing and non-neutralizing Abs, followed by in vitro measurements of Kd values, but the Kd does not necessarily predict biological or clinical efficacy. For some antigens and Abs, the Kd is predictive of the protective effects of an Ab, but for others no correlation is observed. The correlation between Kd and protection may be indicative of a specific MOA that may not be generally applicable to protein toxins such as ricin, which must undergo conformational changes or partial unfolding in order to enter a cell.

Conformational changes occur in a variety of biochemical processes. Abs that stabilize one conformational state over another may be common. The MOA may be rarely recognized since, in these cases, no notable conformational changes would be observed upon Ab binding and alternate conformations of the antigen may elude structural determination. As this study shows, knowledge of the general location of an epitope derived from structures of antibody-bound RTA complexes is not sufficient to predict the neutralizing effects of an Abs, and “decoy epitopes” recognized by non-neutralizing Ab can be found adjacent to epitopes recognized by neutralizing Ab. For toxins, as well as viruses, neutralizing and non-neutralizing Abs recognizing the same antigen can be identified using cell-based assays, indicating that the 2 types of Abs are exerting different effects on the antigen that are protective to the cell. How and where an Ab binds appears to be significant.

Conformationally restrictive mutations in some but not all locations in a 3-dimensional fold can affect the Tm of a protein. Here, we used a structure-guided biophysical approach to examine Abs that appear to conformationally restrict 3 regions of the RTA protein that become disordered during the early stages of unfolding. Neutralizing sdAb packed with 2 of these regions, while non-neutralizing sdAb bound to regions adjacent to or away from these regions. Here we found that the magnitude of the Tm-shift induced by an Ab bound to its Ag reflected how well an Ab was able to neutralize the ricin toxin in cell-based assays, and that the correlation could be perturbed by altering the stability of either the Ag or Ab as we have done here (i.e., (A3C8 vs. C8) or (RTA vs. RTA1-33/44-198). Interestingly, the largest effect on the Tm-shift arose from sdAb C8, which recognizes a region of RTA containing an immunodominant epitope (N97-F108); its effects on the Tm-shift suggest that this RTA epitope may have functional significance.

Deciding which anti-toxin Abs are the most protective has always been empirically determined using cell-based assays or in vivo studies. This has been primarily due to the lack of a physical parameter that correlates with protection. By examining a set of neutralizing and non-neutralizing sdAbs, we found that the Kd of an Ab-Ag complex was not a discriminating feature of a neutralizing antibody. In fact, the difference in the Kd values of the neutralizing and non-neutralizing antibodies was not statistically significant for anti-RTA Abs. While further studies are warranted to determine if the Tm-shift of an Ab-Ag complex is generally useful for evaluating neutralizing Abs against other foreign proteins, we found that the Tm-shift rank-ordered the toxin-neutralizing Abs without the use of cell-based assays. These results indicate a potential MOA for anti-RTA Abs, and provide a predictive cell-free parameter. This MOA may also apply to Abs that inhibit viral fusion since envelope proteins also undergo conformational changes that are required for viral entry (e.g., influenza hemagglutinin).

As Abs now dominate the biologic therapeutics pipeline, methods that rapidly and quantitatively assess which Abs are more protective than others are of immense value. Understanding of the MOA of Abs may affect how we select and assemble Abs into therapeutic cocktails. Our results suggest that the measurement of the apparent Tm-shift induced by an antibody may be a useful cell-free method to quantify additive or synergistic effects by combinations of Abs. Moreover, using CD to measure the Tm-shift provides a label-free method that has relatively few limitations and can be applied to Abs recognizing linear or conformational epitopes.

Materials and methods

Materials. Q Sepharose, SP Sepharose, Superdex G-200 10/300 GL column, Superdex G-75 10/300 GL column, and PD-10 columns were purchased from GE Healthcare Life Sciences (Piscataway, NJ). General chemicals were purchased from Sigma (St. Louis, MO) or Fisher Scientific (Pittsburgh, PA). E. coli BL-21(DE3) cells were purchased from Invitrogen/Life Technologies (Grand Island, NY). BugBuster® Protein Extraction Reagent and TBST buffer was obtained from Millipore (Billerica, MA). SDS-PAGE gels and gel buffers were purchased from Thermo Fisher Scientific Inc. (Rockford, IL). RTA isolated from Ricinus communis was purchased from Vector Labs (L-1190, Burlingame, CA). Novagen's mouse anti-His-tag Ab (#70796) was obtained from Millipore (Billerica, MA). Western Blue® stabilized substrate for alkaline phosphatase (AP) and anti-mouse IgG (S3841), Cell Titer 96 reagent, luciferase mRNA, and luciferin substrate were obtained from Promega (Madison, WI). The ammonium sulfate grid screen for crystallization were obtained from Hampton Research (Aliso Viejo, CA).

Cloning of constructs

The sdAb sequences utilized in this work were derived from the variable domain of heavy chain Abs naturally produced by immunized llamas (C8, D12f, H1W12V, C10neg) or immunized alpacas (D10, F6, E1). The cloning of sdAb A3, an anti-SEB Ab with a high Tm (Tm = 85°C), was previously described and similar methods were used to construct the other sdAb clones. To construct the engineered A3C8 sdAb variant, the CDRs of the anti-RTA sdAb C8 were transferred into the framework regions of sdAb A3 to produce the thermostabilized A3C8 sdAb. Two pet22 constructs of A3C8 were prepared for periplasmic expression─ a C-terminal His-tagged plasmid and a tag-free construct (referred to as A3C8stop). The protein sequences are shown in the Suppl. Information.

Protein expression and purification of sdAb

The His-tagged sdAb were periplasmically expressed and purified using the methods previously described in George et al.

Protein expression and purification of RTA and RTA1-33/44-198 and its variants

RTA and RTA1-33/44-198 were purified as previously described. Briefly, E.coli BL-21(DE3) were grown in LB media supplemented with Kanamycin (50 µg/ml) at 37°C. Cultures at a cell density of 0.8–1.0 OD600 were induced by addition of IPTG (0.2 mM) and grown at 17°C for 18–20 hours before collection by centrifugation. Cell pellets were resuspended in lysis buffer (50 mM sodium phosphate buffer, pH 7.3, 2 mM EDTA, 30% BugBuster®) and sonicated, 30 seconds on/30 seconds off, for a total of 2 minutes. The lysate was clarified by centrifugation at 20,000 × g for 30 min at 4°C. The supernatant was applied to a Q Sepharose column equilibrated with 50 mM sodium phosphate buffer, 2 mM EDTA, pH 7.3, and the protein was collected in the flow through. Fractions containing the protein were combined and dialyzed overnight at 4°C against 1 L of 50 mM MES, 2 mM EDTA (pH 5.5 for RTA and pH 6.4 for RTA1-33/44-198). Dialyzed protein was loaded onto a SP Sepharose column equilibrated with 50 mM MES, 2 mM EDTA pH 5.5 for RTA and pH 6.4 for RTA1-33/44-198. Protein was eluted from the column using a 0–250 mM sodium chloride gradient and analyzed using SDS-PAGE. The fractions containing a single band on SDS-PAGE gels were combined, dialyzed into 50% glycerol, and stored at −20°C.

Preparation and purification of the RTA:sdAb complexes

RTA:sdAb complexes were prepared by incubating RTA or RTA1-33/44-198 with a 10% molar excess (1:1.1 molar ratio) of each sdAb at 4° C for >17 hours. The concentration of RTA or RTA1-33/44-198 in 0.5 mL was determined by UV-vis using calculated extinction coefficients, ϵ = 26,000 M−1cm−1 or ϵ = 18,005 M−1cm−1, respectively. The calculated extinction coefficients for each sdAb were the following: 36,565 M−1cm−1 for C10neg; 23,045 for H1W12V and F6; and 21,555 M−1cm−1 for the remaining sdAbs. Each complex was applied to a Superdex G-75 or G-200 gel-filtration column (as appropriate) equilibrated in 1x phosphate-buffered saline (PBS) pH 4.5 or pH 7.4 to separate the formed complex from excess sdAb. To verify the formation of each complex, each column was calibrated using 3 molecular weight (MW) standards: albumin (66,000 Da); carbonic anhydrase (29,000 Da); and cytochrome C (12,400 Da). Calibration curves for each column at pH 4.5 or pH 7.4 were made using a log-linear plot of molecular weight (kDa) vs. elution volume (mL) and fitting the points to a single exponential decay equation using Grafit 5.0 (Erithricus Software Ltd. West Sussex, UK). The MWs of each complex were calculated from the elution volumes and the protein sequence for comparison and confirmation of complex purification.

Circular dichroism and thermal denaturation curves

Thermal denaturation (2°C/min) of RTA, RTA1-33/44-198, the sdAbs, or their complexes, was monitored using a Jasco 810 Circular Dichroism (CD) spectropolarimeter fitted with a Peltier temperature controller. Concentrations of protein solutions (˜0.2 mg/ml in PBS, pH 4.5 or pH 7.4) were determined by UV-vis using the appropriate calculated extinction coefficient. Melting curves were measured between 10–95°C by monitoring the change in ellipticity at 207, 214, or 222 nm. The apparent melting temperature (Tm) was determined from a 4 parameter fit of data. The three Tm values were averaged and standard deviations (SD) were calculated.

Western analysis of site-directed mutants

Proteins were first separated by SDS-PAGE using an 8–16% gradient gel. Equal amounts of RTA, RTA1-33/44-198, or their variants were loaded (5 μg/lane). RTA isolated from ricin (glycosylated) was used as a control. The proteins in the gel were transferred to nitrocellulose and blocked for 1 h using 1x TBST containing 5% (w/v) dry milk. The primary Ab was incubated in 1x TBST and 5% dry milk for 17 hours at 4°C. The blot was washed 3 times with 1x TBST between subsequent Ab incubations (1 h incubation at 22 ± 3°C) and then developed. SdAb A3C8 containing a C-terminal His-tag was used as the primary Ab (0.1 mg/mL). A mouse anti-His-tag antibody was used as the secondary Ab (1:500 dilution). An anti-mouse AP- conjugated IgG (1:500 dilution) and stabilized AP substrate were used for detection as directed.

Cell-free translation assay

Ricin biological activity was determined using a microtiter cell-free translation assay that detects luciferase translation from luciferase mRNA. Dilutions of ricin are included in each assay for generation of a standard curve. Briefly, antibodies were diluted with PBS in 96-well microtiter plates. A constant amount of ricin (100 ng/mL final concentration) was added to the Ab dilutions, the plate left on a microplate shaker for 15 min (25° C), and then 5 μL was transferred to a v-shaped microtiter plate. The rabbit reticulocyte lysate, RNasin, amino acids complete, and luciferase mRNA were mixed together, and then 25 μL was added to each well. The plates were incubated (37°C) for 90 min, and then 5 μL of the translation lysate was transferred to a black microtiter plate. After the addition of the assay buffer, containing the luciferin substrate, luminescence was measured as relative light units (RLU) using a SpectraMax M5 Multi-Mode Microplate reader (Molecular Devices, LLC., Sunnyvale, CA). The average ± SD of 3 wells was reported as the percentage (%) untreated control, [(treated/untreated control) × 100].

In vitro cell cytotoxicity assay

The mouse lymphoma EL-4 cell line (ATCC, Manassas, VA) was maintained in RPMI-1640 medium supplemented with 5% fetal calf serum. Immediately prior to the assay, cells were pelleted (600 × g, 10 min, 4° C) and resuspended to 2 × 10 cells/mL in RPMI-1640 medium. Using RPMI-1640 medium, sdAbs were diluted (50 μL/well) in a 96-well microtiter plate. Ricin was diluted to yield a final concentration of 200 ng/mL and was added (50 μL/well) to the antibody dilutions. After the plates had been left on a shaker for 15 min, cells (100 μL) were added to each well. The plates were incubated (37° C, 5% CO2) overnight (˜17 h) after which 20 μL Cell Titer 96 was added to each well. The plate incubated for 4 additional hours. Dilutions of ricin were included in each assay and used to generate a standard curve. Optical density (490 nm) was measured on a Microplate reader. The average ± SD of 3 wells was reported as percentage (%) untreated control cells, [(OD of treated cells/OD untreated control cells) × 100].

Crystallization

SdAb A3C8 with a C-terminal His-tag (A3C8cterm) was crystallized in the ammonium sulfate grid screen condition B2 (0.1 M citric acid pH 5.0, 1.6 M ammonium sulfate) at 17°C. Protein in 50 mM Tris pH 7.6 was concentrated to 3.5 mg/mL. Hexagonal plates were obtained in hanging drops containing 1:1 mixtures of protein and precipitant.

RTA1-33/44-198 and A3C8cterm were mixed at a 1:1.1 molar ratio and was exchanged into 50 mM Tris pH 7.6, concentrated (12.5 mg/mL) and crystallized in the ammonium sulfate grid screen condition B4 (0.1 M HEPES pH 7.0, 1.6 M ammonium sulfate). Crystals grew at 17°C within a month.

X-ray crystallography

A3C8 crystals were cryo-protected in 30% glycerol, 70% ammonium sulfate grid screen condition B2. Crystals of the RTA1-33/44-198:A3C8cterm complex were cryo-protected in 30% glycerol and 70% precipitant (0.1 M HEPES pH 7.0, 1.6 M ammonium sulfate). The RTA1-33/44-198:A3C8 crystals contained 4 molecules (2 sdAb, and 2 RTA or RTA1-33/44-198 variants) per asymmetric unit.

Data were collected at 150 K using a Bruker Micro-STAR rotating anode equipped with Helios optics and a Bruker Platinum 135 CCD area detector. The structure of the complex was solved by molecular replacement using Phaser. The structures of RTA (PDB 1J1M), RTA1-33/44-198 (PDB 4IMV), a sdAb (PDB 4FHB), and the structure of the A3C8 sdAb (PDB 5SV4) were used as search models. Model-building was performed using Coot. Simulated annealing was carried out using CNS 1.1, and refinement with Refmac 5.

Surface plasmon resonance of site-directed mutants of RTA

The dissociation constants, Kd, for each of the Abs was determined using a ProteOn™ XPR36 Protein Interaction Array System (Bio-Rad) as described previously. All tests were performed on a GLC sensor chip, for general amine coupling and with a binding capacity of approximately a one protein monolayer. Following activation of the sensor chip with 0.1 M 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride and 0.25 M N-hydroxysulfosuccinimide, immobilization was performed in 10 mM acetate buffer pH 5.0 for immobilization of RTA or its variants at 20 µg/mL on all 6 lanes. The binding kinetics of each sdAb were determined by rotating the chip and flowing various concentrations (300, 100, 33, 11, 3.7, 0 nM) over the chip at 100 µL/minute in the orthogonal direction for 90 s over the RTA-coated chip (or RTA variants) and then monitoring dissociation for 600 s using PBS with 0.005% Tween-20 (pH 7.4) or PBS with 13 mM sodium citrate (pH 4.5) as the running buffer. The surface was regenerated at 50 µL/minute with 1% phosphoric acid for 36 s. Data were corrected by subtraction of the zero antigen concentration column as well as by interspot correction.