Antibody Response to Rotavirus Functional Maturation of the Human

The Journal of Immunology

Functional Maturation of the Human Antibody Response to Rotavirus1 Nicole L. Kallewaard,* Brett A. McKinney,†‡ Yingqi Gu,¶ Annie Chen,¶ B. V. Venkataram Prasad,¶ and James E. Crowe, Jr.2*‡§ Infant Abs induced by viruses exhibit poor functional activity compared with those of adults. The human B cell response to rotavirus is dominated by use of the VH1– 46 gene segment in both adults and infants, but only adult sequences are highly mutated. We investigated in detail the kinetic, structural, and functional advantage conferred by individual naturally occurring somatic mutations in rotavirus-specific human Abs encoded by the immunodominant VH1– 46 gene segment. Adult Abs achieved enhanced binding through naturally occurring somatic mutations in the H chain CDR2 region that conferred a markedly prolonged off-rate and a desirable increase in antiviral potency. Three-dimensional cryoelectron microscopy studies of Ag-Ab complexes revealed the mechanism of viral inhibition to be the binding of high-affinity Abs at the viral RNA release pore in the double-layer particle. These structure-function studies suggest a molecular basis for the poor quality of Abs made in infancy following virus infection or immunization. The Journal of Immunology, 2008, 180: 3980 –3989.

R

otavirus (RV)3 is the most important viral cause of severe dehydrating diarrhea in infants and young children worldwide. The RV particle is composed of three concentric protein layers surrounding a genome of 11 segments of double-stranded RNA. The outer capsid layer consists of viral protein (VP) 4 and VP7, the intermediate layer is composed of VP6, and the viral core is made up of the shell protein VP2 and enzymes VP1 and VP3 (1). VP6 is the most antigenic RV protein in humans, but does not induce classically neutralizing Abs because it is covered by the outer capsid proteins in the virion particle. Some anti-VP6 IgA mAbs, however, protect nonimmune mice from infection and clear chronic infection in SCID mice (2). The mechanism of the protection mediated by such Abs is still being investigated, but it is thought that they block the virus life cycle at the step of transcription, probably inside virus-infected cells (3). After the virus has entered the cell, the outer capsid is released, exposing the VP6-coated double-layer particle (DLP), which then is activated for transcription. These anti-VP6 mAbs block viral transcription in vitro in a dose-dependent manner *Departments of Microbiology and Immunology, †Department of Mathematics, ‡Department of Pediatrics, and §Program in Vaccine Sciences, Vanderbilt University and Vanderbilt University Medical Center, Nashville, TN 37232; and ¶Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, TX 77030 Received for publication September 5, 2007. Accepted for publication January 8, 2008. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 This work was supported by National Institutes of Health Grants R01 AI57933 (to J.E.C.) and AI36040 (to B.V.V.P.). N.L.K. was supported by the National Institutes of Health Training Program in Cellular and Molecular Microbiology, T32 AI07611. J.E.C. holds a Clinical Scientist Award in Translational Research from the Burroughs Wellcome Fund. Cryo-EM studies were supported by the National Center for Macromolecular Imaging (P41 RR02250). 2 Address correspondence and reprint requests to Dr. James E. Crowe, Jr., Vanderbilt University, 1161 21st Avenue South T2220 Medical Center North, Nashville, TN 37232. E-mail address: [email protected] 3

Abbreviations used in this paper: RV, rotavirus; Bmax, maximal binding; DLP, double-layer particle; FR, framework region; Fv, fragment variable; HCDR2, H chain CDR2; koff, rate of dissociation; kon, rate of association; RSV, respiratory syncytial virus; VP, viral protein. Copyright © 2008 by The American Association of Immunologists, Inc. 0022-1767/08/$2.00

www.jimmunol.org

when introduced into a cell concurrently with DLPs or during transcytosis in the cell (3– 6). The specific role of such VP6 Abs in preventing or resolving human infection is still unclear. Recently, studies have shown that VP6 interacts with a large fraction of naive B cells from both adults and neonates (7). RV infection causes a T cell-independent expansion of B cells that are responsible for early secretion of anti-RV IgM and activation of APCs for intestinal IgA production (8, 9). Although there have been insights into the mechanisms of primary immunity and protection against RV, specific correlates of immunity are lacking. One of the central unexplained features in RV immunity is the poor quality of the Ab response in infants compared with that in adults. We previously investigated in detail the human Ab gene repertoire of RV-specific B cells from infected adults or infants derived from single B cells responding to RV VP6. We found that both adult and infant Ab gene sequences exhibited a distinct bias in the usage of H chain gene segments in the VH1 and VH4 families, which differed markedly from the VH3–23 dominant gene segment bias seen in randomly-selected B cells (10, 11). We also noted that the gene segment VH1– 46 was the dominant gene segment used in B cells responding to RV, whether in the systematic compartment or those that have an intestinal homing phenotype (10 –12). Although the infant Ab gene sequences used the same immunodominant VH gene segments as the adult sequences to respond to RV, a striking genetic difference between the B cells of the two groups was the marked lack of somatic mutations in the infant Ab gene sequences (13). We sought to investigate the structural and functional consequences of these naturally occurring somatic mutations, because this genetic profile was the distinguishing feature between adult and infant B cell responses. Detailed kinetic analysis of the contribution of somatic mutations has been performed primarily with haptens and model protein Ags such as hen egg lysozyme (14 –18). Hapten binding studies may not be relevant to microbial protein responses because the small size of haptens allows for high-affinity interactions with only a limited portion of the Ab-contacting surface. Studies are needed to determine how naturally occurring somatic mutations in human Ab genes affect the response to an important biologically relevant protein, particularly an immunodominant protein from a major viral pathogen. We investigate here in detail the kinetic advantage and impact on

The Journal of Immunology antiviral activity of individual, naturally occurring, somatic mutations in two human Abs specific for RV. We found that human Abs encoded by the immunodominant VH1– 46 gene segment bind primarily through interactions in and around the H chain CDR 2 (HCDR2) region of the H chain that are markedly enhanced by naturally occurring somatic mutations. These mutations account for the enhanced affinity of these Abs almost exclusively through a prolonged off-rate. Our studies also show that the potency of the antiviral function of these Abs correlated directly with the overall affinity.

Materials and Methods Expression and purification of recombinant Abs The RV6-25 and RV6-26 H and L chain Ab variable region genes studied here were obtained from single RV-specific IgD⫺ B cells isolated from the blood of a healthy immune donor as described (GenBank accession numbers AF452995, AF453155, AF452996, and AF453157) (19). Soluble Fabs were expressed in HB2151 Escherichia coli, a nonsuppressor strain, and purified by fast protein liquid chromatography under native conditions as described (19); the concentration of the purified Fab was determined by ELISA (10).

Generation of chimeric and mutant Fab Ab fragments We constructed chimeric Abs using portions of the Ab genes from respiratory syncytial virus (RSV) Fab19, a recombinant human Fab to a heterologous virus Ag (RSV F glycoprotein) that we previously isolated (20, 21). The plasmid vectors containing the chimeric combinations of RV6-26 and RSV Fab19 Ab variable gene sequences were prepared by inserting a specific restriction enzyme site and then using molecular cloning techniques. A SmaI site was inserted in the nucleotide sequence in the coding region of amino acids A105 and R106 by site-directed mutagenesis. This change to the nucleotide sequence was a silent change that did not disrupt the amino acid sequence. The SmaI site then acted as a cutting site in the H chain framework region (FR) 3 region just before the start of the HCDR3. Two chimeric Abs were produced by taking advantage of the new SmaI site along with the restriction sites already encoded in the expression vector. To produce the chimeric Abs, double digests with the restriction enzymes SmaI and ApaI were performed at 37°C for 2 h on both the RSV Fab19 and RV6-26 parental plasmids. Digested products were resolved on an agarose gel and excised and extracted. The RSV Fab19 insert was ligated into the RV6-26 plasmid; conversely, the RV6-26 insert was ligated into the RSV Fab19 plasmid. Plasmid vectors containing cDNA that specified mutant derivatives of the parental Fabs RV6-25 or RV6-26 were prepared by changing one of the naturally occurring mutations in the parent Fab gene sequence back to the germline amino acid. Oligonucleotides complementary to the site of interest were synthesized that contained the desired nucleotide back mutation flanked by an unmodified nucleotide sequence. These oligonucleotides were used in a PCR with the plasmid encoding the parental Ab to amplify daughter plasmids that had incorporated the mutation. The parental plasmid then was degraded by restriction enzyme digestion using DpnI at 37°C for 1 h. Plasmids containing the intended mutation were then transformed into the HB2151 nonsuppressor strain of E. coli for Ab expression. The numbering scheme used throughout this paper is derived from the ImMunoGeneTics IMGT VBASE Ab amino acid numbering system (22). The mutant Fabs were named by the original amino acid at the position indicated, followed by the single amino acid code for the germline residue to which it was changed.

Purification of virus particles Rhesus RV particles were harvested from the supernatant and cell fraction after the infection of MA104 cell monolayer cultures and purified by cesium-chloride density gradient centrifugation as described (19). DLPs were collected and dialyzed using PBS solution containing 20% glycerol buffer and stored at ⫺20°C until use.

RV-specific ELISA binding assay using recombinant Abs To verify that the recombinant RV-specific Fab bound to RV Ags, we performed an ELISA using purified RV DLPs as described (19). KD values were calculated using the nonlinear regression curve fitting from GraphPad Prism program; curves that did not have R2 values ⱕ95% were not used in data analysis.

Surface plasmon resonance affinity measurements All experiments were performed at 25°C on a BIAcore 2000 biosensor instrument. CM5 sensor chips were used for all of the analyses of the

3981 VP6-Fab interactions. Purified RV DLPs were treated for 5 min with glycine-HCl (pH 2.0) to disrupt the particles. This viral protein preparation containing monomeric VP6 then was diluted to a concentration of 50 ␮g/ml in sodium acetate buffer at pH 5.0 for immobilization. The Ag was coupled to the chip by amine coupling chemistry. All immobilizations were conducted at 25°C at a flow rate of 5 ␮l/min. The surface was activated by a 7-min injection of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC) and N-hydroxysuccinimide (NHS) mixture. The VP6 protein solution then was loaded onto the chip for 6 min. The surface was then quenched with ethanolamine for 7 min. This protocol resulted in an average of 868 resonance units after coupling. Uncoated surfaces were prepared for the determination of nonspecific binding and bulk refraction subtraction by treating the CM5 sensor chip surface as for Ag immobilization, except that Ag injection was omitted. Purified Fabs were tested in the biosensor at six different concentrations in HBS buffer (10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20 (pH 7.4)), and compared with buffer alone. Fab solutions were injected at a constant flow rate of 30 ␮l/min for 2 min over a blank chip surface and then over a VP6-coated surface. The Fab interaction then was allowed to dissociate for 10 min with the exception of samples RV6-26 wild-type Fab, S29T, and M91L, which were allowed to dissociate for 15 min because of their slow dissociation kinetics. Chip surfaces were regenerated with a 15-s pulse of glycine-HCl (pH 2.5) at a flow rate of 50 ␮l/min. The regenerated surfaces then were washed with running buffer for one additional minute. Coated chip surfaces were used for a maximum of 72 cycles before a new surface was prepared. Preliminary experiments showed that the Ag retained 84% of the original level of reactivity at this point (data not shown). Kinetic data were interpreted with the BIAevaluation 4.0 software (BIAcore) using a 1:1 Langmuir binding model. For each Fab dilution series, data were collected on the active ligand-containing flow cell and on the noncoated blank flow cell. The signal from the blank flow cell was subtracted from that of the active flow cell signal to remove any contributions of bulk response or nonspecific binding from the sample signal. These referenced data were aligned manually using the y-intercept average of a small sample before injection. All of the data curves were zeroed to the buffer alone and then a global analysis using the 1:1 Langmuir binding model calculated the rate of association (kon) and the rate of dissociation (koff) values. We only accepted data that fit the following criteria: 1) ␹2 values of ⬍10; and 2) T values for the Rmax, ka, and kd of ⬎10. All samples were run in triplicate in a full dilution series; values for triplicates were averaged and SD values were calculated. The KD values then were calculated for each Fab from the equation kon/koff ⫽ KD. These values are reported in Tables I and II.

Model of the structure of Ab fragments A model of the structure of the recombinant Abs was obtained using a structural algorithm program (Web Antibody Modeling (WAM)). This program uses an improved algorithm for modeling Abs that is based upon the AbM package and is used to construct three-dimensional models of Ab fragment variable (Fv) sequences using a combination of established theoretical methods together with the latest Ab structural data (23). Individual sets of structural coordinates were obtained for RV6-25 and RV6-26 from WAM. Molecular graphics images were produced using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, CA (24).

Intracellular neutralization assay Intracellular virus neutralization using DLPs, MA104 cells, and lipofectin was performed as described previously (4). The number of Ag-positive foci or fluorescent focus units in each well was counted by sweeping across one well of a 96-well plate using a ⫻10 objective. These numbers were then compared with the RV-infected control wells treated with medium alone. The endpoint titer at which 50% reduction of fluorescent focus units occurred for each Ab was determined and the triplicate values for each Fab were averaged and reported in Table III.

Three-dimensional cryoelectron microscopy (cryo-EM) Complexes of DLP with purified human Fabs were formed by mixing DLPs and Fab in a stoichiometric ratio equivalent to five Ag-binding regions per VP6 molecule to ensure complete saturation of all available epitopes. Mixtures were incubated for 2 h at room temperature and then concentrated to ⬃1012 particles per milliliter by ultrafiltration using a MicroCon-100 microconcentrator (Millipore). Specimens for microscopy were embedded in vitreous ice on holey carbon films according to standard procedures (25). Specimens were observed on a JEOL 1200EX transmission electron microscope operated at ⫻30,000 magnification using 100-kV electrons and a beam dose of ⬃5 electrons/Å2. Images were collected using Kodak SO-163

3982

HUMAN Abs TO ROTAVIRUS VP6

Table I. The high affinity of the RV6-26 wild-type Ab is mediated primarily by somatic mutations in the HCDR2 region that cause a decrease in dissociation kinetics

Antibody Type

Region of Mutationa

Wild type Single mutations

Fold kon (M⫺1 s⫺1) Differencec (⫻105)

RV6-26 CDR1 CDR1(⫹1) FR2 FR2 CDR2(⫺1) CDR2 CDR2(⫹1) FR3 FR3 FR3 FR3 FR3 FR3

Double mutations CDR2(⫹1)/CDR2(⫺1) CDR2(⫹1)/FR3 CDR2(⫹1)/FR3 CDR2(⫹1)/CDR1(⫹1) Triple mutation

Cloneb

S29T V39M E44Q E48Q M55I D60G Y66S R70K P74G T85S F88Y M91L G93S

6.6 (0.19) 6.4 (0.36) 7.2 (0.04) 4.4 (0.16) 6.0 (0.04) 2.7 (0.12) 1.5 (0.49) 2.6 (0.07) 4.0 (0.03) 4.9 (0.05) 4.6 (0.93) 6.4 (0.04) 13.0 (0.90) 3.3 (0.07)

Y66S/M55I Y66S/P74G Y66S/G93S Y66S/V39M

CDR2(⫹1)/CDR2(⫺1)/FR3 Y66S/M55I/R70K

Koff (s⫺1) (⫻10⫺5)

Fold Differencec

2.6 (1.64) 1.5 (0.32) 3.0 (0.35) 3.3 (0.23) 11 (0.40) 98 (2.00) 250 (6.00) 5.2 (0.48) 15 (3.10) 2.1 (0.33) 1.7 (0.28) 1.2 (0.32) 13 (0.40)

Fold Differencec

4.49 ⫻ 10⫺11

3.0 (0.92) ⴱ ⴱ ⴱ ⴱ ⫺2 ⫺4 ⫺2 ⴱ ⴱ ⴱ ⴱ ⫹2 ⫺2

KD (M)

ⴱ ⫺2 ⴱ ⴱ ⫹4 ⫹33 ⫹83 ⴱ ⫹5 ⴱ ⴱ ⫹3 ⫹4

4.11 ⫻ 10⫺11 2.14 ⫻ 10⫺11 6.93 ⫻ 10⫺11 5.43 ⫻ 10⫺11 4.05 ⫻ 10⫺10 6.63 ⫻ 10⫺9 9.80 ⫻ 10⫺9 1.30 ⫻ 10⫺10 3.01 ⫻ 10⫺10 4.52 ⫻ 10⫺11 2.59 ⫻ 10⫺11 9.00 ⫻ 10⫺12 4.08 ⫻ 10⫺10

ⴱ ⴱ ⴱ

⫹2

⫺9 ⫺147 ⫺218 ⫺3 ⫺7 ⴱ ⴱ ⫹5 ⫺9

0.45 (0.01) 9.7 (0.04) 1.8 (0.01) 7.2 (0.16)

⫺15 ⴱ ⫺4 ⴱ

4800 (20.0) 240 (2.00) 240 (3.00) 230 (3.00)

⫹1,600 ⫹80 ⫹80 ⫹78

1.06 ⫻ 10⫺6 2.46 ⫻ 10⫺9 1.34 ⫻ 10⫺8 3.20 ⫻ 10⫺9

⫺23,617 ⫺55 ⫺298 ⫺71

0.81 (0.13)

⫺8

580 (170)

⫹193

7.16 ⫻ 10⫺8

⫺1,586

a

The designations “⫹1” and “⫺1” in parentheses indicate the first flanking amino acid N-terminal or C-terminal, respectively, to the indicated CDR. b Designates the somatically mutated amino acid followed by position and then the germline reverted amino acid. c Fold differences were calculated by dividing the wild-type value by the selected mutant value for fold decreases (⫺) or dividing the selected mutant value by the wild-type giving fold increases (⫹). The asterisk (ⴱ) designates differences that were ⬍2-fold increased or decreased.

electron film (Eastman Kodak) with a 1-s exposure time. Micrographs were developed for 12 min using Kodak D-19 developer at 21°C and fixed for 10 min in Kodak Fixer. For each imaging field, two micrographs were collected, one at ⬃1 ␮m underfocus and the other at ⬃2 ␮m underfocus.

Three-dimensional structural analysis Micrographs suitable for three-dimensional structural analysis were digitized on a Nikon Super Coolscan 8000 scanner using a 6.35-␮m step size. Pixels were averaged to give a 12.7-␮m step size that corresponded to 4.23 Å per pixel in the object. Three-dimensional reconstructions were computed using software in the ICOS Toolkit suite (26) according to standard procedures (27–30). Particle centers and orientations were initially estimated from further-from-focus image data, and the parameters were then applied to the closer-to-focus image set for refinement and structural determination. The three-dimensional reconstruction was computed to a resolution of ⬃22 Å. The effects of the contrast transfer function were corrected as previously described (31). The resolution achieved in the final structure was confirmed by Fourier ring correlation analysis (32). The distribution of image data throughout the asymmetric unit was sufficient for the resolution achieved, as 97% of the mean inverse eigenvalue spectrum was ⬍0.01. Threshold values for the reconstructions were chosen to account for 780 molecules of VP6 between radii of 250 and 350 Å. The three-dimensional structure was viewed using IRIS Explorer 5.2 (Numerical Algorithms Group) and Chimera 1.23. Surface representations were displayed using a mass density threshold chosen to account for the expected number of VP6 molecules within the VP6 capsid region. The atomic model of the VP6 trimer (Protein Data Bank accession code 1qhd; Ref. 33) was fitted into the three-dimensional cryo-EM reconstructions using the Situs suite of programs (34). The solutions from Situs that corresponded to the highest correlation coefficient were chosen. The x-ray coordinates for RV6-25 and RV6-26 were fitted into their respective portions of the cryo-EM density maps by visual inspection using the graphical program O (35).

Results The VH1– 46 gene segment is the site of primary interaction with RV VP6 We previously found that a majority of human VP6-specific Abs were encoded by the immunodominant gene segment VH1– 46 (10, 11). Because of this immunodominant bias in gene usage, we hypothesized that the VH1– 46 gene segment-encoded HCDR1 and

HCDR2 regions of RV6-26 retained a strong binding activity considering the unnatural heavy and L chain pairing, having a KD of 2.46 ⫻ 10⫺7 M. The chimeric Ab containing the HCDR3 and the L chain of RV6-26 without the VH region exhibited a diminished affinity of 1.45 ⫻ 10⫺6 M. These results show that in the case of an immunodominant VH gene segment response, the HCDR1 and HCDR2 loops can dominate the interaction with Ag with only minor influence from the L chain and HCDR3 region. Binding affinities of the wild-type RV immunodominant Fabs In previous studies, we determined that both adult and infant B cells specific for RV used the same immunodominant gene segment, VH1– 46, but the Ab gene sequences from the infant clones lacked somatic mutations (10, 13). The use of an immunodominant VH gene segment (with differing D and J segments) suggested that dominant common structural features encoded by the VH gene segment (i.e., the HCDR1 and HCDR2 regions or FRs 1–3) mediate optimal binding to a common epitope on the virus. Some of the Ab sequences isolated, including the infant Abs using a germline (unmutated) VH1– 46 gene segment, bound Ag as B cell receptors or natural Ig but did not exhibit binding that could be detected in immunofluorescent or ELISA as monovalent Fabs (10) (and data not shown). Based on these observations, we hypothesized that infant germline-encoded Abs bind with very weak affinity but that somatic mutations in adult Abs facilitate high-affinity binding. To test this hypothesis, we determined the contribution of the amino acids that had been somatically mutated from the germline sequence in the VH region of two representative adult Abs. We chose to study two adult Abs from our panel, designated RV6-25 and RV6-26, which contained 21 or 13 somatic mutations in their VH regions, respectively. As stated above, RV6-26 exhibits high affinity whereas RV6-25 has a more moderate affinity. Therefore, a wide spectrum of affinity effects could be detected upon the back mutation of somatic mutations in these two VH1– 46 clones to the germline-encoded amino acid.


3983

Table II. Somatic mutations in the low affinity RV6-25 Fab result in limited changes in binding kinetics, most prominently in the HCDR2 region

Antibody Type

Region of Mutationa

Wild type Single mutations

FR1 FR1 FR1 FR1 FR1 CDR1 CDR1(⫹1) FR2 CDR2(⫺1) CDR2 CDR2 CDR2(⫹1) FR3 FR3 FR3 FR3 FR3 FR3 FR3 FR3 FR3

Cloneb

kon (M⫺1 s⫺1) (⫻104)

RV6-25

4.1 (0.77)

M12V R13K R20K I21V T25A T32S I39M L53M V55I K59S Y62S T66S E69Q T80R I87V I89M R92S G93S K95R D97E I101V

#d 1.2 (0.13) 8.0 (2.10) 1.1 (0.20) 3.7 (0.42) 2.5 (0.07) 8.4 (1.40) 7.7 (2.40) #d †e ##f §g 6.9 (3.20) 4.1 (0.45) #d 5.0 (0.40) #d 1.6 (0.24) 4.4 (0.72) 0.34 (0.22) ##f

Fold Differencec

koff (s⫺1) (⫻10⫺3)

Fold Differencec

KD (M)

Fold Differencec

9.5 ⫻ 10⫺8

3.9 (0.46) ⫺3 ⫹2 ⫺4 ⴱ ⴱ ⫹2 ⴱ

4.5 (2.30) 1.4 (0.90) 0.52 (0.03) 2.0 (0.50) 1.9 (0.45) 1.2 (0.10) 1.9 (0.01)

ⴱ ⫺3 ⫺8 ⴱ ⫺2 ⫺3 ⫺2

3.8 ⫻ 10⫺7 1.7 ⫻ 10⫺8 4.8 ⫻ 10⫺8 5.4 ⫻ 10⫺8 7.6 ⫻ 10⫺8 1.4 ⫻ 10⫺8 2.5 ⫻ 10⫺8

⫺4 ⫹6 ⴱ ⴱ ⴱ ⫹6 ⫹4

ⴱ ⴱ

2.9 (1.52) 1.9 (0.12)

ⴱ ⫺2

4.2 ⫻ 10⫺8 4.6 ⫻ 10⫺8

⫹2 ⫹2

ⴱ

1.7 (0.23)

⫺2

3.4 ⫻ 10⫺8

⫹3

⫺3 ⴱ ⫺12

2.8 (0.20) 2.2 (0.16) 1.3 (0.35)

ⴱ ⴱ ⫺3

1.7 ⫻ 10⫺7 5.2 ⫻ 10⫺8 4.0 ⫻ 10⫺7

ⴱ ⴱ ⫺4

a

The designations “⫹1” and “⫺1” in parentheses indicate the first flanking amino acid N-terminal or C-terminal, respectively, to the indicated CDR. Designates the somatically mutated amino acid followed by position and then the germline reverted amino acid. Fold differences were calculated by dividing the wild-type value by the selected mutant value for fold decreases (⫺) or dividing the selected mutant value by the wild-type giving fold increases (⫹). The asterisk (ⴱ) designates differences that are within 2-fold increased or decreased. d The single pound sign (#) represents Fabs that did not show binding at concentrations of 6.4 ␮M or less. e The cross (†) indicates that this Fab showed nonspecific binding to the blank flow cell as well as to the RV-coated flow cell, giving unreliable results. f The double pound sign (##) represents Fabs that did not show binding at concentrations of 1.6 ␮M or less. g The section symbol (§) indicates that this Fab exhibited a very low level of binding by visual inspection of the curves; however, accurate kinetics could not be determined at this low level of binding (i.e., T values were ⬍10). b c

Each of the 34 mutant recombinant human Fabs we generated contained each of the somatic mutations in the RV6-26 or RV6-25 parental Fab, except that one amino acid was changed back to the original germline configuration as found in the infant sequences. Because of this mutation strategy, a decrease in the affinity of the mutant Fab would suggest that this amino acid would be advantageous for a germline sequence to acquire. Each of the 34 mutants was created and expressed, and Fab was purified and screened in a RV-binding ELISA. The resulting binding curves were analyzed using a nonlinear regression program to define the binding curves for maximal binding (Bmax) and the concentration at which binding was 50% (KD). The calculated equilibrium binding affinities are shown in the gray bars and the individual Bmax of the Fabs is depicted as a line (Fig. 1). In the ELISA binding assay the wildtype Fab RV6-26 exhibited a very high KD of 8.6 ⫻ 10 ⫺1 M (Fig. 1A). Most of the mutants retained binding equivalent to that of the wild-type Fab except for S29T, M55I, D60G, Y66S, and G93S. Mutant S29T showed Bmax values equivalent to that of the wildtype Fab, although a curve with high R2 value could not be determined. In contrast, the M55I, D60G, Y66S, and G93S mutants exhibited binding to a much lower extent than that of wild-type Fab, resulting in poor curve fitting. Interestingly, three of the mutations that caused deficient binding clustered in or immediately adjacent to the HCDR2 region. This same distribution of effect was seen with the RV6-25 mutant binding results (Fig. 1B), although the overall affinities were much lower those than for RV6-26. The panel of mutants of RV6-25 contained many more Fabs that did not appear to bind DLPs in the ELISA. The ELISA and BIAcore binding findings were discordant in some cases because of the low amount of binding and the strict definitions of quality of data for the binding data. For example, biosensor curves for the R92S mutant showed no

binding when tested in triplicate on two different occasions with different Ag chips, but the ELISA analysis detected binding. Several reasons could account for these discrepancies, including the long incubation period in the ELISA, differences in the presentation of the Ag on the two solid substrates (gold vs plastic), or differences in treatment buffers (carbonate buffer in the ELISA vs glycine-HCl used in the biosensor assay). The mutations in the HCDR2 region, containing mutants V55I, K59S, and Y62S, showed no binding, with Bmax values that did not differ from the negative control. Clearly there were also mutations outside of the HCDR2 that affected binding, including both HCDR1 mutations and framework mutations. Together, the steady-state binding data suggested that the mutations in and around the HCDR2 region played a prominent role in the enhanced binding of both of these RV-specific human Fabs that share the immunodominant VH1– 46 gene segment, but the enhancing mutations were not restricted to this loop. Kinetic analysis of mutant Fabs derived from the Fab RV6-26 To gain a better understanding of the kinetic parameters of the binding and to increase the binding sensitivity of our assay, we performed a detailed study of each of the 34 Fabs using surface plasmon resonance. Comparison of the single mutation results showed that the mutant Fabs with somatic mutations reverted to germline that are located in or near the HCDR2 loop exhibited a lower affinity than the wild-type Fab RV6-26 (Table I). The mutations M55I, D60G, Y66S, R70K, and P74G all cluster near the HCDR2 region and each of these Fabs showed a decrease in overall affinity, up to a 218-fold difference. Comparison of the individual kinetic rates of association and dissociation showed that the mutated core HCDR2 residues M55I, D60G, and Y66S caused

3984


Table III. Somatic mutations in the RV6-26 HCDR2 region that cause decrease in dissociation kinetics also cause increase in antiviral function Antibody Type

Position of Mutationa

Wild type

Clone

Reciprocal Titer

Specific Activity (ng/ml)b

RV mouse mAb 7D9 RSV human Fab19 RV6-25 RV6-26

6,144 ⱕ2 ⱕ2 1,024

16 ⱖ100,000 ⱖ100,000 98

1,536 512 2,048 1,024 256 224 32 1,024 512 512 2,048 4,096 16

65 200 49 98 390 450 3,100 98 200 200 49 24 6,300

RV6-26 Single mutations

CDR1 CDR1(⫹1) FR2 FR2 CDR2(⫺1) CDR2 CDR2(⫹1) FR3 FR3 FR3 FR3 FR3 FR3

S29T V39M E44Q E48Q M55I D60G Y66S R70K P74G T85S F88Y M91L G93S

RV6-26 double mutations

CDR2(⫹1)/CDR2(⫺1) CDR2(⫹1)/FR3 CDR2(⫹1)/FR3 CDR2(⫹1)/CDR1(⫹1)

Y66S/M55I Y66S/P74G Y66S/G93S Y66S/V39M

ⱕ2 ⱕ2 6 4

ⱖ100,000 ⱖ100,000 17,000 25,000

RV6-26 triple mutation

CDR2(⫹1)/CDR2(⫺1)/FR3

Y66S/M55I/R70K

ⱕ2

ⱖ100,000

a

The designations “⫹1” and “⫺1” in parentheses indicate the first flanking amino acid N-terminal or C-terminal, respectively, to the indicated CDR. Specific activity is defined as the lowest concentration of antibody at which a 50% reduction in fluorescent focus units was detected in the antibody-DLP co-transfection assay. b

only a 2- to 4-fold slower association rate, but a dramatic difference was seen in the dissociation rates. The Y66S mutation exhibited a striking koff of 2.5 ⫻ 10⫺3 s⫺1 as opposed to the wild-

FIGURE 1. The equilibrium binding affinity of the wild-type or mutant Fabs determined by ELISA binding to viral double-layered particles. A, RV6-26. B, RV625. Each graph contains the calculated Bmax (line form) and KD (bars) for each Fab.

type RV6-26 koff of 3 ⫻ 10⫺5 s⫺1, meaning that the Y66S mutant had a 83-fold faster rate of dissociation. The D60G mutation, which is only four amino acids away from Y66S, also caused a


3985

FIGURE 2. Representative surface plasmon resonance binding curves showing the interaction between VP6 and the mutant RV-specific Fabs that were decreased in overall affinity in comparison to the wild-type Fab. A, Binding curves of RV6-26 wild-type and mutants at a concentration of 12.5 nM. B, Binding curves of RV6-25 wild-type and mutants at a concentration of 800 nM.

very fast dissociation rate of 9.8 ⫻ 10⫺4 s⫺1, which was 33-fold faster than that of the parental Fab. These differences can be appreciated visually by viewing the representative binding curves at a 12.5 nM concentration of each of the mutant RV 6 –26 Fabs that exhibited a decrease in the overall affinity (Fig. 2A). Although some of the curves differed from that of the wild-type Fab to a minor extent in the association phase (M55I and D60G), all differed strikingly in the dissociation phase. We next constructed selected mutant Fabs that contained two amino acids reverted back to the germline residues to test whether the affinity improvement caused by important residues was additive. Because of the large number of mutants that would have to be made to test all of the residues in every combination, we chose to focus on the amino acid associated with the most dramatic effect on affinity (Y66S) along with residues at positions that were shown to be somatically mutated in multiple, independently cloned Abs from our panel of 28 VH1– 46 RV-specific Abs (11). The Y66S/ M55I combination of mutations caused a striking effect on both the association and dissociation rates (Table I). This double mutation resulted in a 15-fold decrease in kon and ⬎1,000-fold faster koff, which exceeds the sum of the individual effects caused by Y66S and M55I. This dramatic effect suggested that these two amino acids likely are involved directly in the high-affinity binding mechanism of the RV6-26 Fab. The Y66S/G93S double mutation resulted in a pattern that appeared to be an additive effect of the single mutations. Interestingly, the Y66S/P74G and Y66S/V39M mutations rescued the small kon defect seen with the Y66S single mutation, yet they retained the 80-fold koff difference. We also produced a triple mutant Ab clone with the Y66S/M55I/R70K mutations, incorporating a FR3 mutation that was the closest mutated position to the dominant residue Y66S. Interestingly, the effect on affinity of the addition of the M55I mutation was partially reverted by the addition of this FR3 mutation. These data suggest that although the central finding that the mutations are cooperative in enhancing affinity holds true, the cooperative effects of combined back reversions is complex. Likely, there are contextual effects in which mutations exert differing effects on affinity depending on the entire ensemble of mutations present that modulate the fine topography of the combining sites. Given that we do not know the order in which the somatic mutations occurred, it is not possible to determine whether any particular double or triple mutant actually occurred naturally during the derivation of the Ab. Nevertheless, the studies suggest a dominant effect of particular residues likely to be in contact with Ag and a cooperative effect on affinity of the mutations rather than independent effects.

Kinetic analysis of the mutant Fabs derived from RV6-25 The effects of the RV6-25 mutations were not as dramatic as those in RV6-26, although similarities in mechanism were observed (Table II). In contrast to the RV6-26 mutations for which we could detect binding of all the mutants, some of the RV6-25 mutants did not exhibit RV-specific binding at the concentrations tested. The lack of binding likely was observed because those mutations caused a dramatic decrease in affinity that was below detectable levels in our system. Sixteen of the 22 mutations gave surface plasmon resonance results that were statistically relevant by our criteria. Most of the mutants had a KD that was very similar to that of the RV6-25 wild-type Fab. The mutants that still retained binding but exhibited an effect on the KD were decreased only slightly in affinity. Although the mutants D97E and R13K both had a 4-fold decrease in overall affinity, D97E maximal binding was greatly reduced (Fig. 2B). The R13K, I21V, G93S, and D97E mutants were slower in their association, but the dissociation rates of I21V and G93S were slower than that of the wild-type Fab, which resulted in no net change in the overall affinity. The mutants with HCDR2 region germline-reverted residues were all reduced in binding in some way. The residues that were located in the HCDR2 region showed varying results; V55I and Y62S did not bind VP6 in these assays, K59S bound nonspecifically, and T66S bound but with a marked decrease in maximal binding (Fig. 2B).

Structural implications To gain more understanding of the mechanism by which the somatically mutated residues of these Abs affect the binding kinetics, we examined them in the context of the topography of the Fab structure. We modeled the structure of the Fabs using a collection of Ab modeling algorithms. On both structures the RV6-25 (Fig. 3 right column) and RV6-26 (Fig. 3 left column) Fabs are shown as ribbon and surface representations and the mutated amino acids are color coded according to the results of the binding experiments. In the RV6-26 structure, the amino acids that affected binding the most (shown in red) were all surface exposed on the top face of the Ab, except for G93S (Fig. 3G). One region located in the middle of the top view contains the most potent mutations, Y66S, D60G, and M55I. This same topographic area in RV6-25 also was mutated, with the residues T66S, V55I, and Y62S that reduced or eliminated binding in our assays (colored orange) (Fig. 3H). These amino acids are located in the HCDR2 loop and are exposed on the surface of the Fab in the proposed RV-binding site. These data

3986


FIGURE 3. Location of the somatic mutations in the Fab structures that cause affinity changes. A and B show the ribbon diagrams of the Fv structures of RV6-26 and RV6-25, respectively. C and D show the spacefilling structure from the side, E and F show the opposite side of the Fab, and G and H show the top of the Fv structures. In all structures, the L chain (LC) is shown in blue and the H chain (HC) in gray. The germline reversion mutant Fab that had a better affinity than the wild type (i.e., deleterious germline mutations) is shown in green. Mutations that resulted in a lower affinity than the wild type (i.e., advantageous germline mutations) are shown in red, mutations that had no effect on affinity are shown in pink, mutations that resulted in loss of binding are shown in orange, and the mutation that caused nonspecific binding is shown in blue.

suggest that the area in and around the HCDR2 is the central domain of the RV-binding site for VH1– 46-encoded Abs. Although these two Fabs shared a discrete area in which somatic mutations enhanced the binding to RV, the surface topography of this area in the two Fabs differs dramatically primarily because of the different positions of the tyrosine mutations in the two Fabs. It is likely that RV6-26 mutations, including a tyrosine at position 66 and the aspartic acid at 60, confer more fitness for binding to VP6 than the mutations in RV6-25.

Viral inhibition mediated by VP6-specific Fabs Treatment with some VP6-specific mouse mAbs disrupts efficient transcription by RV particles in cells (4 – 6). Some of these VP6specific mAbs can clear infection in a SCID mouse model of RV infection when the Abs are delivered as IgA isotype molecules (2), suggesting a unique intracellular mode of Ab-mediated viral inhibition. We sought to determine whether our human VP6 Abs also mediate this antiviral activity and, if so, to what extent somatic


3987

FIGURE 4. Three-dimensional cryo-EM studies of DLP:Fab complexes reveal different patterns of binding of the low-affinity RV6-25 and high-affinity RV6-26 Fabs. A, Surface representation of the DLP:Fab complex. DLP is shown in gray, Fabs are shown in yellow. An instance of a type I channel along the 5-fold symmetry is indicated by a pink pentagon, whereas a type III channel is indicated by a blue hexagon. B, Close inspection of the DLP:Fab complex reconstructions at a single instance of the type III channel on the particle. C, The atomic resolution structure of the VP6 trimer and a model of the Fv portion of RV6-25 were fit into the cryo-EM density map.

mutations affect the potency of this function. To test antiviral activity, we used a previously established assay based on DLP neutralization inside cells (4). DLPs are rendered infectious if introduced into the cytoplasm of permissive cells by lipid-mediated introduction of particles (36). In this assay, VP6 Fabs can exhibit antiviral activity if introduced into the cytoplasm concurrently with DLPs. The RV-infected cells were stained, the fluorescent focus units were counted, and the reciprocal titer of the Fab dilution needed to inhibit 50% of infected cells was calculated (Table III). The low-affinity wild-type RV6-25 Fab and the RSV-specific Fab19 control Ab did not mediate neutralization even at high concentration. Conversely, the high-affinity wild-type RV6-26 Fab and some of its mutants effectively neutralized the virus. The RV6-26 Fab showed a neutralizing titer of 1/1,024, correlating to a neutralizing specific activity of 98 ng/ml. It was remarkable that the human Fab inhibited replication at a concentration in the same order of magnitude as the potent murine IgA mAb 7D9, which was reported previously to have a high level of activity in this assay (4) even though the IgA mAb has four binding sites but the human Fab is only monovalent. We investigated which amino acids of RV6-26 Fab were important for the inhibition of viral growth by testing mutant Fabs for antiviral function. We found that the mutants that were decreased in their overall affinity were less effective at neutralizing the virus. Mutants Y66S and G93S showed profound loss of neutralizing activity, with reciprocal titers of 16 or 8 respectively (Table III). Single mutants with affinities that were decreased 9-fold or more showed a corresponding decreased ability to neutralize the virus. The double mutants showed a more pronounced effect, with nearly a complete loss of inhibitory capacity. These data suggest that the antiviral function is mediated by high-affinity Abs with slow off-rates. Three-dimensional structural analysis of DLP:Fab complexes To directly visualize the interactions between human Fabs and the VP6 capsid layer, we imaged DLP:Fab complexes in vitreous ice by cryo-EM (data not shown) and computed the three-dimensional

structure to a resolution of ⬃22 Å from 90 and 95 particles from the cryoimages of DLP-RV6-26 and DLP-RV6-25 complexes, respectively (Fig. 4). The VP6 capsid is made up of 260 trimers arranged on a T ⫽ 13 (levo) icosahedral lattice. This arrangement of capsomers defines 132 channels that penetrate the VP6 layer (1). Twelve of these channels, designated type I, are located along the icosahedral 5-fold axes, surrounded by five VP6 timers and are the site of the mRNA release. The other 120 channels, designated type II and III, are surrounded by six VP6 timers. In the outer regions of the VP6 capsid layer the trimers are well separated from one another, while in the lower regions, near the VP2 layer, the trimers form a network of interactions with one another (33). Densities due to bound Fabs are observed in both the DLP:Fab complexes (Fig. 4). In the RV6-25 complex (Fig. 4A) Fab molecules are attached at the tops of the VP6 trimers with the epitope positioned along the edge. Well-defined densities are seen attached to a VP6 subunit in one of the six trimers surrounding each of the channels. The Fab domains had the expected bilobed morphology seen in the atomic model of a Fab (6). In other VP6 subunits the densities for the Fab are scattered and not as well defined. Thus, of the 780 possible Fab molecules, only 120 have well-defined shapes. The disrupted morphology of the Fab domains in the threedimensional structure is likely a consequence of steric hindrance that does not allow all of the Fab molecules to bind uniformly to all of the VP6 subunits. RV6-26 exhibited a different distribution. In this structure, large and significant density due to the bound Fab molecules is centered inside each of the 132 channels. This density runs deep into each of the channels but does not make contact with the VP6 subunits surrounding the channels, thus making it difficult to ascertain the exact location of the Fab-binding epitope on VP6. One possibility is that this density represents the CH/CL constant regions from one or two Fab molecules and the binding epitope is in a lower domain of the VP6 that is accessible through the channels in the VP6 layer. Because of the limited volume inside the channels, not all six (in the type II and III channels) or five (type I channels) Fabs can be accommodated. The volume of the observed density inside each

3988 channel is large enough to accommodate two constant domains, indicating that two Fab molecules perhaps are accommodated in each channel. Because any two of the available five (in type I) or six (type II and III) sites could be occupied, the density due to the Fv domain is averaged out and not visible. Thus, a maximum of ⬃264 molecules of RV6-26 bind to each virion. Based on the cryo-EM studies, it is evident that the epitopes for RV6-25 and RV6-26 are distinct. For RV6-25 it is clear that the epitope is on the upper domain of the VP6 subunit. The exact binding site of RV6-26 is less clear from the cryo-EM studies, but likely is lower on the VP6 molecule. The higher affinity RV6-26 Fab appeared to overcome the limited accessibility inside the channels. These data are also revealing because the mRNA release type I channels in transcriptionally-active DLPs occur at the 5-fold symmetry. The RV6-26 Fabs completely block the type I channel, providing a reason for why RV6-26 inhibited virus replication in the experiments described above while RV6-25 did not (Fig. 4A, pink pentagon). The atomic structure of the VP6 trimer (33) was fit into the cryo-EM density map, as was a model of the Fab RV6-25 (Fig. 4C). The fitting of the structures into the cryo-EM density suggested that the Fab binds to one VP6 subunit and is consistent with a mode of binding in which the HCDR2 region of the Ab is in closest contact with a surface loop of VP6.

Discussion In this study, we identified the structural features that determined the emergence of an immunodominant VH-type human B cell response to an important viral pathogen. Previously, we discovered that the human Ab responses of infant and adults to RV used variable gene segments (most commonly VH1– 46) that differed markedly from those used in the random Ab repertoire (10). We also found that the human infant B cell response to RV was distinguished by a marked lack of somatic mutations (13). Because Abs that are encoded by the VH1– 46 gene segment were not all capable of binding to RV to the same extent, we sought to determine what were the principal structural determinants of binding. These studies examined the largest panel of human Abs to date derived from a single VH region. By studying two highly mutated VH1– 46 Abs with differing affinities in detail, we found that in both cases critical somatic mutations in the HCDR2 region were involved in enhancing interaction with VP6. This effect of HCDR2 mutations was seen in both independently derived Fabs, but to a different extent. The RV6-25 Fab affinity of 9.5 ⫻ 10⫺8 was nearly three orders of magnitude lower than that of RV6-26. When the HCDR2 single amino acids that had been somatically mutated in the parental Ab were back mutated to their germline-encoded configuration, the resulting Fabs lost their ability to bind at a detectable level in our assays. Theoretically, this lack of binding could also be due to a Fab folding error; however, these germline residues are located on the surface in the Ag-binding site and thus are highly unlikely to cause structural folding problems. The mutations of the RV6-26 Fab were more valuable in determining the mechanism of the interaction. The wild-type RV6-26 Fab exhibited an affinity constant of 4.5 ⫻ 10 ⫺1 M for VP6. With this high affinity, a broad range of loss of affinity could be determined for the mutants in our surface plasmon resonance assay. We found that RV6-26 somatic mutations in and around the HCDR2 region were the major determinants of enhanced binding, with the single mutation of Y66S alone decreasing the affinity over two orders of magnitude to 9.8 ⫻ 10⫺9 M. Although HCDR2 mutations did slightly affect the on-rate compared with the wild-type Fab, the dramatic change in off-rate caused the overall marked affinity difference. For example, the Y66S mutant had a 83-fold

HUMAN Abs TO ROTAVIRUS VP6 faster off-rate than RV6-26. This effect was increased when two of these HCDR2 region mutations were reverted to germline together, with the most extreme case being the Y66S/M55I double mutation that resulted in a Fab with affinity of only 1.06 ⫻ 10⫺6 M. These mutations caused a 15-fold decrease in the on-rate and a 1,600-fold faster off-rate. The mutational data for the two Fabs show that the HCDR2 region is critical in the binding of immunodominant Abs to RV and that somatic mutations in the HCDR2 are sufficient to dramatically increase the affinity of the Ab for the Ag because of slow off-rates. Despite the critical role of mutations in the HCDR2 loop in enhancing the affinity of binding, other mutated residues contributed to enhanced binding, especially FR3 residues and a few HCDR1 mutations. Although the HCDR1 mutations likely affect Ag interactions directly, many of the FR3 mutations probably enhance binding by distal effects on the conformation of the CDRs. The complex findings on the effect of the reversion of mutations in both CDRs and FRs suggests that Ag combining sites evolve in an iterative fashion, resulting in an intricate elaboration of fine structural features that enhance binding through cooperative effects that cannot be easily discerned or recapitulated with single point mutations. Remarkably, most of the mutations that occurred in these Abs were outside the CDRs and did not enhance affinity. In fact, several mutations were deleterious for binding. In contrast to our findings, a previous detailed study of a virus-neutralizing Ab using engineered mutations selected for affinity suggested that improvements in the on-rate enhanced neutralizing activity (18). Abs directed against some model Ags that have been characterized previously at a kinetic level also show an importance of the off-rate in affinity maturation caused by somatic mutations (14, 15, 17, 37, 38). Because some murine VP6 mAbs have been shown to exhibit an antiviral function in vivo and in vitro, we investigated whether our human Fabs possessed this antiviral activity. We found that treatment with the moderate affinity Fab RV6-25 did not reduce the number of infected cells, but the high-affinity RV6-26 mediated a strong antiviral activity in concentrations that were similar to that of the previously described murine VP6 mAbs. To our knowledge, this is the first described human mAb to exert this unique function; however it should be noted that Abs using VH1– 46 and directed to VP6 are the dominant components of the human B cell response to RV infection. Our studies raise the possibility that this VP6-specific response may be a critical and dominant component of the human immune response that restricts RV infection and replication. Previously, most investigators have considered the human VP6 Ab response nonfunctional. We examined the ability of the mutant Fabs to block viral infection and found that the same germline-reverted residues that were responsible for decreased affinity also caused a decrease in the ability to inhibit the virus. Cryo-EM studies provided revealing structural information. The two Fabs bound to VP6 but exhibited different patterns of binding on the structure of the DLP. The high-affinity RV6-26 bound inside the channels in the VP6 layer of the capsid at both the 5-fold and 6-fold points of symmetry. In contrast, the moderate-affinity RV6-25 Fab recognized an epitope in an apparently more distal domain of the VP6 and did not obstruct type I channels at the 5-fold symmetry axis, which are the sites where RNA is extruded from the transcriptionally active DLP. This differential pattern of binding explains why RV6-26 inhibits replication more effectively. This differential pattern of binding on the DLP also is of interest because of the implications the data have for the evolution of Ab specificity. These two Abs began with a common germline configuration (and thus almost certainly bound initially to a common epitope), but each explored the surface of VP6 in a different way

The Journal of Immunology during the accumulation of somatic mutations that resulted in Abs of differing fine specificity. Although there have been many studies investigating how VP6 mAbs inhibit viral transcription, a clear mechanism has not been found to fit all cases. Most investigators think that the inhibition is caused by a conformational change in VP6 that is either prohibited or is induced upon Ab binding (4 – 6). Our studies using human Fabs show that in the case of the RV6-26 and its derived mutants, high-affinity binding due to prolonged off-rate is the major factor in enhancing antiviral activity. We propose that in the case of RV6-26, the Fab binds so tightly to the VP6 and the off-rate is so slow that the Fab rarely dissociates from the particle, likely physically blocking the mRNA release channels. We are not aware of any previous inhibitory Ab, murine or human, that binds VP6 in this particular position in the center of the channel. Taken together, these results suggest that the Abs specified by germline configuration of the RV immunodominant VH gene segments such as VH1– 46 are structurally fit for binding to RV VP6 at a low affinity, but somatic mutations in the HCDR2 region of the adult response increase the affinity of the Ab through a slower off-rate. This change in affinity causes a desirable increase in antiviral function. These structure-function studies using naturally occurring mutations from adult Abs that are not present in infant Abs also suggest a molecular basis for the poor quality of Abs made in infancy following virus infection or immunization.

Acknowledgments We thank Richard Ward and Monica McNeal for technical advice, Harry Greenberg for RV strains, and Raymond Mernaugh and the Vanderbilt Molecular Recognition Core Laboratory for BIAcore technical assistance. We thank Frances House and Ashley Long for technical assistance and reagent preparation.

Disclosures The authors have no financial conflict of interest.

References 1. Pesavento, J. B., S. E. Crawford, M. K. Estes, and B. V. Prasad. 2006. Rotavirus proteins: structure and assembly. Curr. Top. Microbiol. Immunol. 309: 189 –219. 2. Burns, J. W., M. Siadat-Pajouh, A. A. Krishnaney, and H. B. Greenberg. 1996. Protective effect of rotavirus VP6-specific IgA monoclonal antibodies that lack neutralizing activity. Science 272: 104 –107. 3. Corthesy, B., Y. Benureau, C. Perrier, C. Fourgeux, N. Parez, H. Greenberg, and I. Schwartz-Cornil. 2006. Rotavirus anti-VP6 secretory immunoglobulin A contributes to protection via intracellular neutralization but not via immune exclusion. J. Virol. 80: 10692–10699. 4. Feng, N., J. A. Lawton, J. Gilbert, N. Kuklin, P. Vo, B. V. Prasad, and H. B. Greenberg. 2002. Inhibition of rotavirus replication by a non-neutralizing, rotavirus VP6-specific IgA mAb. J. Clin. Invest. 109: 1203–1213. 5. Thouvenin, E., G. Schoehn, F. Rey, I. Petitpas, M. Mathieu, M. C. Vaney, J. Cohen, E. Kohli, P. Pothier, and E. Hewat. 2001. Antibody inhibition of the transcriptase activity of the rotavirus DLP: a structural view. J. Mol. Biol. 307: 161–172. 6. Lawton, J. A., M. K. Estes, and B. V. Prasad. 1999. Comparative structural analysis of transcriptionally competent and incompetent rotavirus-antibody complexes. Proc. Natl. Acad. Sci. USA 96: 5428 –5433. 7. Parez, N., A. Garbarg-Chenon, C. Fourgeux, F. Le Deist, A. Servant-Delmas, A. Charpilienne, J. Cohen, and I. Schwartz-Cornil. 2004. The VP6 protein of rotavirus interacts with a large fraction of human naive B cells via surface immunoglobulins. J. Virol. 78: 12489 –12496. 8. Coffin, S. E., S. L. Clark, N. A. Bos, J. O. Brubaker, and P. A. Offit. 1999. Migration of antigen-presenting B cells from peripheral to mucosal lymphoid tissues may induce intestinal antigen-specific IgA following parenteral immunization. J. Immunol. 163: 3064 –3070. 9. Blutt, S. E., K. L. Warfield, D. E. Lewis, and M. E. Conner. 2002. Early response to rotavirus infection involves massive B cell activation. J. Immunol. 168: 5716 –5721. 10. Weitkamp, J. H., N. Kallewaard, K. Kusuhara, E. Bures, J. V. Williams, B. LaFleur, H. B. Greenberg, and J. E. Crowe, Jr. 2003. Infant and adult human B cell responses to rotavirus share common immunodominant variable gene repertoires. J. Immunol. 171: 4680 – 4688. 11. Weitkamp, J. H., N. L. Kallewaard, A. L. Bowen, B. J. Lafleur, H. B. Greenberg, and J. E. Crowe, Jr. 2005. VH1– 46 is the dominant immunoglobulin heavy chain gene segment in rotavirus-specific memory B cells expressing the intestinal homing receptor ␣4␤7. J. Immunol. 174: 3454 –3460.

3989 12. Weitkamp, J. H., B. J. Lafleur, and J. E. Crowe, Jr. 2006. Rotavirus-specific CD5⫹ B cells in young children exhibit a distinct antibody repertoire compared with CD5⫺ B cells. Hum. Immunol. 67: 33– 42. 13. Weitkamp, J. H., B. J. Lafleur, H. B. Greenberg, and J. E. Crowe, Jr. 2005. Natural evolution of a human virus-specific antibody gene repertoire by somatic hypermutation requires both hotspot-directed and randomly-directed processes. Hum. Immunol. 66: 666 – 676. 14. De Genst, E., F. Handelberg, A. Van Meirhaeghe, S. Vynck, R. Loris, L. Wyns, and S. Muyldermans. 2004. Chemical basis for the affinity maturation of a camel single domain antibody. J. Biol. Chem. 279: 53593–53601. 15. England, P., R. Nageotte, M. Renard, A. L. Page, and H. Bedouelle. 1999. Functional characterization of the somatic hypermutation process leading to antibody D1.3, a high affinity antibody directed against lysozyme. J. Immunol. 162: 2129 –2136. 16. Lavoie, T. B., S. Mohan, C. A. Lipschultz, J. C. Grivel, Y. Li, C. R. Mainhart, L. N. Kam-Morgan, W. N. Drohan, and S. J. Smith-Gill. 1999. Structural differences among monoclonal antibodies with distinct fine specificities and kinetic properties. Mol. Immunol. 36: 1189 –1205. 17. Sagawa, T., M. Oda, M. Ishimura, K. Furukawa, and T. Azuma. 2003. Thermodynamic and kinetic aspects of antibody evolution during the immune response to hapten. Mol. Immunol. 39: 801– 808. 18. Toran, J. L., L. Sanchez-Pulido, L. Kremer, G. del Real, A. Valencia, and A. C. Martinez. 2001. Improvement in affinity and HIV-1 neutralization by somatic mutation in the heavy chain first complementarity-determining region of antibodies triggered by HIV-1 infection. Eur. J. Immunol. 31: 128 –137. 19. Weitkamp, J. H., N. Kallewaard, K. Kusuhara, D. Feigelstock, N. Feng, H. B. Greenberg, and J. E. Crowe, Jr. 2003. Generation of recombinant human monoclonal antibodies to rotavirus from single antigen-specific B cells selected with fluorescent virus-like particles. J. Immunol. Methods 275: 223–237. 20. Barbas, C. F., III, J. E. Crowe, Jr., D. Cababa, T. M. Jones, S. L. Zebedee, B. R. Murphy, R. M. Chanock, and D. R. Burton. 1992. Human monoclonal Fab fragments derived from a combinatorial library bind to respiratory syncytial virus F glycoprotein and neutralize infectivity. Proc. Natl. Acad. Sci. USA 89: 10164 –10168. 21. Crowe, J. E., Jr., B. R. Murphy, R. M. Chanock, R. A. Williamson, C. F. Barbas, III, and D. R. Burton. 1994. Recombinant human respiratory syncytial virus (RSV) monoclonal antibody Fab is effective therapeutically when introduced directly into the lungs of RSV-infected mice. Proc. Natl. Acad. Sci. USA 91: 1386 –1390. 22. Lefranc, M. P. 2002. IMGT, the international ImMunoGeneTics database: a highquality information system for comparative immunogenetics and immunology. Dev. Comp. Immunol. 26: 697–705. 23. Whitelegg, N. R., and A. R. Rees. 2000. WAM: an improved algorithm for modelling antibodies on the WEB. Protein Eng. 13: 819 – 824. 24. Pettersen, E. F., T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt, E. C. Meng, and T. E. Ferrin. 2004. UCSF chimera: a visualization system for exploratory research and analysis. J. Comput. Chem. 25: 1605–1612. 25. Dubochet, J., M. Adrian, J. J. Chang, J. C. Homo, J. Lepault, A. W. McDowall, and P. Schultz. 1988. Cryo-electron microscopy of vitrified specimens. Q. Rev. Biophys. 21: 129 –228. 26. Lawton, J. A., and B. V. Venkataram Prasad. 1996. Automated software package for icosahedral virus reconstruction. J. Struct. Biol. 116: 209 –215. 27. Crowther, R. A., D. J. DeRosier, and A. Klug. 1970. The reconstruction of a three-dimensional structure from projections and its application to electron microscopy. Proc. R. Soc. Lond. Ser. A 317: 319 –340. 28. Crowther, R. A. 1971. Procedures for three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Philos. Trans. R. Soc. Lond. B Biol. Sci. 261: 221–230. 29. Fuller, S. D. 1987. The T⫽4 envelope of Sindbis virus is organized by interactions with a complementary T⫽3 capsid. Cell 48: 923–934. 30. Olson, N. H., and T. S. Baker. 1989. Magnification calibration and the determination of spherical virus diameters using cryo-microscopy. Ultramicroscopy 30: 281–297. 31. Zhou, H., and A. van Oosterom. 1994. Application of the boundary element method to the solution of anisotropic electromagnetic problems. Med. Biol. Eng. Comput. 32: 399 – 405. 32. Van Heel, M. 1987. Similarity measures between images. Ultramicroscopy 21: 95–100. 33. Mathieu, M., I. Petitpas, J. Navaza, J. Lepault, E. Kohli, P. Pothier, B. V. Prasad, J. Cohen, and F. A. Rey. 2001. Atomic structure of the major capsid protein of rotavirus: implications for the architecture of the virion. EMBO J. 20: 1485–1497. 34. Wriggers, W., R. A. Milligan, and J. A. McCammon. 1999. Situs: A package for docking crystal structures into low-resolution maps from electron microscopy. J. Struct. Biol. 125: 185–195. 35. Jones, T. A., J. Y. Zou, S. W. Cowan, and M. Kjeldgaard. 1991. Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47: 110 –119. 36. Bass, D. M., M. R. Baylor, C. Chen, E. M. Mackow, M. Bremont, and H. B. Greenberg. 1992. Liposome-mediated transfection of intact viral particles reveals that plasma membrane penetration determines permissivity of tissue culture cells to rotavirus. J. Clin. Invest. 90: 2313–2320. 37. Mueller, C. M., and R. Jemmerson. 1996. Maturation of the antibody response to the major epitope on the self antigen mouse cytochrome c: restricted V gene usage, selected mutations, and increased affinity. J. Immunol. 157: 5329 –5338. 38. Patten, P. A., N. S. Gray, P. L. Yang, C. B. Marks, G. J. Wedemayer, J. J. Boniface, R. C. Stevens, and P. G. Schultz. 1996. The immunological evolution of catalysis. Science 271: 1086 –1091.