Protective Immunoglobulin A and G Antibodies Bind to Overlapping Intersubunit Epitopes in the Head Domain of Type 1 Reovirus Adhesin {sigma}1

Nonfusogenic mammalian orthoreovirus (reovirus) is an entericpathogen of mice and a useful model for studies of how an entericvirus crosses the mucosal barrier of its host and is subjectto control by the mucosal immune system. We recently generatedand characterized a new murine immunoglobulin A (IgA)-classmonoclonal antibody (MAb), 1E1, that binds to the adhesin fiber, ${sigma}$ 1, of reovirus type 1 Lang (T1L) and thereby neutralizes theinfectivity of that strain in cell culture. 1E1 is producedin hybridoma cultures as a mixture of monomers, dimers, andhigher polymers and is protective against peroral challengeswith T1L either when the MAb is passively administered or whenit is secreted into the intestines of mice bearing subcutaneoushybridoma tumors. In the present study, selection and analysisof mutants resistant to neutralization by 1E1 identified theregion of T1L ${sigma}$ 1 to which the MAb binds. The region bound bya previously characterized type 1 ${sigma}$ 1-specific neutralizing IgGMAb, 5C6, was identified in the same way. Each of the 15 mutantsisolated and analyzed was found to be much less sensitive toneutralization by either 1E1 or 5C6, suggesting the two MAbsbind to largely overlapping regions of ${sigma}$ 1. The tested mutantsretained the capacity to recognize specific glycoconjugate receptorson rabbit M cells and cultured epithelial cells, even thoughviral binding to epithelial cells was inhibited by both MAbs.S1 sequence determinations for 12 of the mutants identified ${sigma}$ 1 mutations at four positions between residues 415 and 447,which contribute to forming the receptor-binding head domain.When aligned with the ${sigma}$ 1 sequence of reovirus type 3 Dearing(T3D) and mapped onto the previously reported crystal structureof the T3D ${sigma}$ 1 trimer, the four positions cluster on the sideof the ${sigma}$ 1 head, across the interface between two subunits. Threesuch interface-spanning epitopes are thus present per ${sigma}$ 1 trimerand require the intact quaternary structure of the head domainfor MAb binding. Identification of these intersubunit epitopeson ${sigma}$ 1 opens the way for further studies of the mechanisms ofantibody-based neutralization and protection with type 1 reoviruses.

INTRODUCTION

Top

Introduction

Antibodies are key elements in protective immunity against mostviruses (reviewed in references 7, 27, 28, 45, and 59). Bothantibodies in serum (primarily immunoglobulin G [IgG]) and antibodiessecreted onto mucosal surfaces (primarily dimeric or polymericIgA) (27, 59) make important contributions to protection fromviruses that invade via mucosal routes. Serum IgG within mucosaltissues may protect by the same mechanisms as during neutralizationof viruses in cell culture (e.g., by interference with receptorbinding) (27, 28, 45, 59), but other serum and tissue factorsthat recognize virus-bound IgG (e.g., complement and phagocytes)participate as well (59). On mucosal surfaces, secretory IgAantibodies are thought to provide a first line of protectionby facilitating entrapment of viral particles in mucus, blockingviral adherence to epithelial cell surfaces, and/or preventingviral entry across the epithelial barrier (38).

Reoviruses provide useful models for studies of many aspectsof viral pathogenesis and host immunity (reviewed in reference50). Type 1 and type 3 reoviruses, including the prototype straintype 1 Lang (T1L), adhere selectively to apical surfaces ofM cells in the intestinal epithelium associated with Peyer'spatches in adult mice and exploit the transepithelial transportactivity of these cells (22, 60). M-cell adherence is a prerequisitefor reovirus T1L infection in the adult intestine since it isrequired for initial entry of this virus into the intestinalmucosa (1, 22, 24, 61; reviewed in reference 34). We have focusedour recent studies on the respective roles of secretory IgAand serum IgG in protecting the intestinal mucosa from reovirusT1L invasion and replication (24, 48) and on the viral and cellularcomponents that mediate adherence of type 1 reoviruses to epithelialcell surfaces (22).

We have recently shown that the binding of type 1 reovirusesto rabbit M cells is mediated by the viral ${sigma}$ 1 protein (22). The $~$ 50-kDa ${sigma}$ 1 protein is the hemagglutinin and attachment proteinthat also mediates binding of reoviruses to nonepithelial cells,including cultured fibroblasts and neurons (reviewed in references31, 48, and 50). It is the serotype-determining protein, againstwhich a neutralizing antibody response is most strongly directed(57). The ${sigma}$ 1 protein is sometimes visible by electron microscopyas long, 40- to 50-nm, bowed or bent fibers that extend fromthe fivefold axes of viral particles (19). ${sigma}$ 1 may also be capableof adopting a more contracted conformation in virions (16, 19).The ${sigma}$ 1 fiber is a homotrimer (13, 49) and is topped by a globularhead domain formed by C-terminal sequences primarily in ß-sheets(13, 18, 36). The fibrous tail of ${sigma}$ 1 is thought to comprise both ${alpha}$ -helical coiled-coil and ß-spiral domains (5, 13,18, 19, 36). The globular head domain of ${sigma}$ 1 in both type 1 andtype 3 reoviruses is required for efficient binding to fibroblastsand other cell types in culture (4, 10, 11), but the ${sigma}$ 1 region(s)involved in binding to M cell apical surfaces is not specificallyknown. One binding mechanism that appears common to many reovirusstrains involves interaction of the ${sigma}$ 1 head domain with the recentlyidentified protein receptor, junctional adhesion molecule 1(JAM1) (4, 41). JAM1 is a cell surface glycoprotein that ispresent on nonpolarized cells and on the basolateral cell surfacesof intact epithelial layers (32). However, since it is not normallyexpressed on the apical surfaces of epithelial cells (32), JAM1seems unlikely to mediate initial binding of reovirus to M cellsin the intestinal epithelium.

Our recent report on reovirus binding to intestinal epithelialcells showed that specific glycoconjugates containing ${alpha}$ 2,3-linkedsialic acid on the apical surfaces of rabbit M cells and polarizedCaco-2_BBe cells in culture are important for binding of type1 reoviruses to these cells (22). In contrast, type 3 reovirusesfail to bind rabbit M cells and do not require ${alpha}$ 2,3-linked sialicacid epitopes for binding to polarized Caco-2_BBe cells (22).This difference suggests that the interaction of ${sigma}$ 1 with particularsialic acid-containing glycoconjugates on epithelial cell apicalsurfaces is serotype specific and is consistent with previousevidence that the ${sigma}$ 1 proteins of type 1 and type 3 reoviruseshave distinct carbohydrate-binding regions. In the ${sigma}$ 1 of reovirusT1L, a region required for hemagglutination, which is thoughtto involve carbohydrate-receptor binding, is located in theT(iv) distal tail region (8, 11). The presence of the head domainappears to be required for this activity of T1L ${sigma}$ 1, but the type3 Dearing (T3D) ${sigma}$ 1 head can effectively substitute (11). In the ${sigma}$ 1 of T3D, a region required for binding to ${alpha}$ -linked sialic acid(39) is located in the T(iii) middle tail region (8, 11, 12,15, 35, 42), and the head domain is not required for this activity(11, 12, 35).

We have also recently characterized a new murine IgA monoclonalantibody (MAb), 1E1, which was raised against reovirus T1L andbinds to T1L ${sigma}$ 1 (24). 1E1 is the first IgA MAb specific for the ${sigma}$ 1 protein. Its hybridoma clone was obtained by fusion of Peyer'spatch cells after oral inoculation of mice with T1L, and itproduces a mixture of IgA monomers, dimers, and higher polymers.When either passively administered or secreted into the intestinefrom subcutaneous hybridoma tumors, 1E1 protects mice from intestinalentry and mucosal replication of T1L after peroral challenge(24). The fact that mice with 1E1 in their intestinal secretionshave no detectable virus in Peyer's patches after peroral challengewith T1L implies that this anti- ${sigma}$ 1 IgA prevents infection atan early stage, most likely by blocking entry to the mucosathrough M cells. 1E1 neutralizes T1L infection of cultured murinefibroblasts (L929 cells) in vitro, blocks viral adherence toapical membranes of polarized Caco-2 cells, and therefore likelyprotects mice from peroral challenge by blocking the ${sigma}$ 1-mediatedadherence of viral particles to intestinal M cells (24). Theepitope recognized by 1E1 appears to be discontinuous (24),but the regions of ${sigma}$ 1 recognized by 1E1 are not known. Definingthe binding site of this protective, mucosally derived IgA isimportant for understanding potential mechanisms of IgA-mediatedimmune protection. For example, the secreted antibody mightbind directly to the carbohydrate-binding region of ${sigma}$ 1 involvedin M cell adherence. Alternatively, it might interfere withthis binding region by steric hindrance or might disrupt ${sigma}$ 1 functionin some other way.

The importance of the head domain of T1L ${sigma}$ 1 for host cell infectionand immune protection has been demonstrated with IgG MAb 5C6,which was raised against reovirus T1L, binds to the ${sigma}$ 1 proteinof type 1 isolates (9, 51, 54, 55), and recognizes the ${sigma}$ 1 headdomain (11). The exact location of the 5C6 epitope on ${sigma}$ 1 is notknown. 5C6 blocks T1L binding and uptake in several culturednon-epithelial cell systems (4, 53, 54). When introduced intothe blood circulation of mice either by intravenous injection(51) or from subcutaneous hybridoma tumors (24), 5C6 mediatesprotection against the systemic spread of T1L but does not preventits entry into and infection of Peyer's patches after peroralchallenge. This lack of mucosal protection by IgG MAb 5C6 invivo is likely attributable to the incapacity of circulatingIgG to undergo receptor-mediated secretion into the intestinallumen of mice and the resulting lack of IgG in intestinal secretionsat the time of challenge with virus (21). Indeed, when 5C6 isadministered perorally to mice along with T1L, it is capableof preventing Peyer's patch infection (44). Thus, IgG MAb 5C6and IgA MAb 1E1 share certain protective capacities. However,their respective binding sites on ${sigma}$ 1 could be distinct becauseinduction of serum IgG and secretory IgA occur in differentimmune-sampling environments, function in different compartments,and may protect by different specific mechanisms.

In this study, we sought to identify the epitope on ${sigma}$ 1 that isrecognized by IgA MAb 1E1, to compare it to that recognizedby IgG MAb 5C6, and to determine the relationships of theseepitopes to the putative epithelial carbohydrate-binding regionof type 1 ${sigma}$ 1. We first confirmed the neutralizing activity of1E1 against representatives of the three established reovirusserotypes in L929 cells. We then selected and characterized1E1-resistant and 5C6-resistant mutants obtained from reovirusT1L. The results show that 1E1, like 5C6, recognizes the globularhead domain of T1L. Both 1E1 and 5C6 recognize overlapping epitopesthat span adjacent ${sigma}$ 1 monomers in the trimeric ${sigma}$ 1 head, and otherevidence suggests these epitopes are not directly involved incarbohydrate-receptor binding. IgA MAb 1E1 thus most likelyprotects in vivo by sterically hindering interaction of a ${sigma}$ 1carbohydrate-binding site with glycoconjugate receptors on apicalsurfaces of M cells in the intestinal epithelium. The resultsexpand our understanding of the protective mucosal and systemicimmune response against this enteric viral pathogen.

MATERIALS AND METHODS

Top

Materials and Methods

L929 cells, reoviruses, and antibodies. Murine L929 cells were a laboratory stock maintained in spinnercultures in Joklik modified minimal essential medium (Irvine)supplemented to contain 2% fetal bovine serum and 2% calf bovineserum (Invitrogen) in addition to 2 mM glutamine, 100 U of penicillin/ml,and 100 µg of streptomycin/ml (Mediatech). The same mediumwas used in growing virus stocks in L929 monolayers. For plaqueassays, 2x 199 medium (Irvine) supplemented to contain 4 mMglutamine (Mediatech) plus 200 U of penicillin/ml, 200 µgof streptomycin/ml, and 0.5 µg of amphotericin B (Fungizone;Invitrogen)/ml, with (standard assay) or without (protease overlayassay) 2% fetal bovine serum and 2% calf bovine serum (Invitrogen),was used.

Reoviruses T1L, T3D, 3HA1, 1HA3, T2/Human/Ohio/Jones/1955, andT3/Murine/France/Clone 9/1961 were laboratory stocks derivedfrom ones from the B. N. Fields laboratory. Reoviruses T1/Human/Netherlands/1/1984,T1/Human/Netherlands/1/1985, T2/Simian/Maryland/SV59/1958, T2/Human/Netherlands/1/1984,and T3/Human/Netherlands/1/1983 were laboratory stocks derivedfrom ones from the T. S. Dermody laboratory (20). Purified virionswere prepared with second-passage L929 cell lysates, and infectioussubvirion particles (ISVPs) were obtained after digestion withTLCK (N ${alpha}$ -p-tosyl-L-lysine chloromethyl ketone)-treated ${alpha}$ -chymotrypsin(Sigma) (16, 22).

IgA MAb 1E1 was obtained as a clarified hybridoma culture supernatant,diluted to an approximate concentration of 2 mg of MAb/ml, andstored frozen at –20°C since preparation (24). The1E1 hybridoma supernatant contained a mixture of IgA monomers,dimers, and higher polymers as previously described (58). Purifiedstocks of IgG MAb 5C6 were obtained as a gift from K. L. Tyler,H. W. Virgin IV, and M. A. Mann. These stocks were obtainedin phosphate-buffered saline (PBS) at an approximate concentrationof 1 mg/ml and stored frozen at –80°C after purification(55).

Plaque reduction assays. Virus in a second- or third-passage cell lysate stock was dilutedin PBS supplemented with 2 mM MgCl₂ and 0.4% (wt/vol) bovineserum albumin (Sigma) (PBS⁺/BSA) to an infectious concentrationof 150 or 200 PFU per 100 µl. Undiluted MAb or PBS⁺/BSA(no-MAb control) was then added to the diluted virus stock as1/20 of the final volume. The mixtures were incubated for 1h at room temperature or 37°C (consistent for all sampleswithin each experiment). The medium was removed from L929 monolayersin the wells of a six-well cluster plate, each monolayer wasinoculated with 100 µl of the virus-MAb or control mixture,and virus-cell absorption was allowed to proceed for 1 h atroom temperature. For each mixture, three wells were inoculatedwithin each experiment. After absorption, the wells were overlaidwith a 1:1 (vol/vol) mixture of completed 2x 199 medium and2% agar (Difco) and then analyzed by standard plaque assay (19).Plaques were counted after staining with neutral red (Sigma).The mean value from the three wells of each mixture was usedin calculating the relative plaque number in the matched MAband no-MAb samples. The results from one to four such experimentswere then combined to give the results shown. In some otherexperiments not shown, different fivefold serial dilutions ofthe MAbs were tested for their neutralization capacities againstdifferent viruses. In those experiments, MAb dilutions weremade in PBS⁺/BSA and mixed in a 1:1 (vol/vol) ratio with a virusdilution containing 300 or 400 PFU per 100 µl.

Selections for neutralization-resistant mutants. A second- or third-passage cell lysate stock of T1L was mixedwith MAb (1E1 or 5C6) at a 10:1 (vol/vol) ratio, followed byincubation at room temperature for 1 h. The medium was removedfrom an L929 cell monolayer in a 25-cm² flask, the monolayerwas inoculated with the virus-MAb mixture, and viral adsorptionwas allowed to proceed for 1 h. At the end of this period, 5ml of medium was added to the flask, and the flask was placedin a 37°C CO₂ incubator to allow viral growth. When 80 to100% of the cells had detached from the flask, viral particleswere released from the cells by freezing them at –80°Cand then thawing them. This cell lysate was then used for thenext passage. The different passage lysates in each series weresubjected to protease-overlay plaque assay (as described inreference 25 except that ${alpha}$ -chymotrypsin [Sigma] was used in placeof trypsin) to determine their infectious titers before theywere tested in a plaque reduction neutralization assay. Viralclones (plaques) that resisted neutralization in this assaywere picked for further isolation and characterization. Allclones characterized here were obtained from the first-passagelysate of each series. Each chosen plaque (putative 1E1- or5C6-resistant mutant) was subjected to two more rounds of plaqueisolation in protease overlay assays to promote clonal purity.The final plaque isolates were then subjected to serial passagein L929 cell monolayers to generate second-passage cell lysatestocks. The infectious titers of these stocks were again determinedby protease overlay plaque assay before use in further experiments.

Enzyme-linked immunosorbent assay (ELISA) for MAb binding to virions. Maxi-Sorp 96-well plates (Nunc) were coated with wild-type T1Lor neutralization-resistant mutants at 5 x 10⁹ purified virionsper well in virion buffer (150 mM NaCl, 10 mM MgCl₂, 10 mM Tris[pH 7.5]) at 4°C overnight. After a blocking step with 3%(wt/vol) BSA (Sigma) in PBS, the plates were incubated witheither MAb 1E1 at a starting concentration of 100 mg/ml or MAb5C6 at a starting concentration of 50 mg/ml in serial threefolddilutions for 2 h at room temperature. The dilutions were performedin PBS containing 0.3% (wt/vol) BSA and 0.05% (vol/vol) Tween20 (Sigma). After several washes with the dilution buffer, theplates were incubated with biotinylated goat anti-mouse anti-IgAor -IgG antibodies, respectively (Southern BioTech). The plateswere then washed, and the bound MAb was detected by using streptavidin-horseradishperoxidase and tetramethyl benzidine Microwell substrate (Kirkegaardand Perry). Absorbances were read at 650 nm.

S1 nucleotide sequence determinations. Genomic reovirus RNA was isolated from purified virions by usingTRIzol-LS reagent (Invitrogen). Briefly, 2 x 10¹² virions wereadded to TRIzol-LS, mixed by pipetting, and incubated at roomtemperature for 5 min. Samples were then extracted with chloroformand incubated at room temperature for an additional 15 min.After centrifugation at 12,000 x g, the aqueous phase was removedand mixed 1:2 (vol/vol) with isopropanol, and the precipitatedRNA was pelleted by centrifugation at 12,000 x g. RNA precipitateswere washed with 75% ethanol, pelleted again by centrifugationat 7,500 x g, allowed to air dry, and resuspended in diethylpyrocarbonate-treated water. The genomic RNA was converted toDNA and amplified by using S1 forward primer 5'-GCTATTCGCGCCTATGGATGC-3'and S1 reverse primer 5'-GATGATTGACCCCTTGTGCCGAGGGTTCG-3' (Invitrogen).Reverse transcription (RT) was done in two steps. First, 1 µl(1 µg) of genomic RNA, 1 µl (0.1 µg) of eachprimer, and 7 µl of diethyl pyrocarbonate-treated waterwere heated to 95°C for 5 min and then snap-cooled on ice.A mixture containing 1 µl of 10 mM deoxynucleoside triphosphates,2 µl of 10x reverse transcription (RT) buffer (Invitrogen),4 µl of 25 mM MgCl₂, 2 µl of 0.1 mM dithiothreitol,and 1 µl of RNaseOUT (Invitrogen) was next added to theprimer-template mixture on ice. The new mixture was incubatedat 42°C for 2 min before 1 µl of reverse transcriptasewas added. The RT reaction was allowed to proceed at 42°Cfor an additional 48 min and then shifted to 70°C to terminatethe reaction. Amplification of the RT products was carried outby mixing 5 µl of each reaction, 1 µl (0.1 µg)of each primer, 5 µl of 10x Thermopol buffer (New EnglandBiolabs), 1 µl of 10 mM deoxynucleoside triphosphates,36 µl of water, and 1 µl of Vent polymerase (NewEngland Biolabs) and then subjecting the mixture to the followingsteps: 95°C for 5 min; 36 cycles of 95°C for 45 s, 53°Cfor 45 s, and 70°C for 90 s; and finally 70°C for 10min. The final DNA products were gel purified before sequencing,which was performed at the Dana-Farber/Harvard Cancer CenterHigh-Throughput DNA Sequencing Facility according to their standardprotocols (http://dnaseq.med.harvard.edu/default.html). Sequenceswere determined from both strands in full, except for primer-proximalsequences at each 5' end.

Molecular graphics. The crystal structure of the T3D ${sigma}$ 1 C-terminal half was displayedwith RasMac 2.7.1 with coordinates obtained from The ProteinData Bank (entry 1KKE). The figure was assembled by using Photoshop5.5 and Illustrator 8.0 (Adobe Systems).

Cell-binding studies. Experiments to characterize virus binding to M cells in tissuesections of rabbit Peyer's patches or to Caco-2_BBe cells inconfluent monolayers were performed as previously described(22). Briefly, to detect virus binding to cells, purified ISVPswere biotinylated with EZ-Link Sulfo-NHS-Biotin (Pierce). Deparaffinizedsections of fixed Peyer's patch tissue were overlaid with 200µl of biotinylated ISVPs (3 x 10¹¹ to 4 x 10¹¹ particles/ml)in PBS containing 0.2% gelatin (Sigma) and incubated in a humidifiedchamber. Caco-2_BBe cell monolayers grown on glass coverslipswere washed and incubated with virus by inversion of coverslipsonto 50-µl droplets of biotinylated ISVPs (3 x 10¹¹ to4 x 10¹¹ particles/ml) in gel saline solution (137 mM NaCl,270 µM CaCl₂, 840 µM MgCl₂, 19 mM H₃BO₃, 130 µMNa₂B₄O₇, 0.3% gelatin [pH 7.4]). Sections or coverslips werethen washed and incubated with streptavidin-TRITC (tetramethylrhodamine isothiocyanate) to label viral particles.

RESULTS

Top

Results

Serotype-restricted neutralization of type 1 reoviruses by IgA MAb 1E1. The plaque reduction assay in cell culture is the classicalmethod for demonstrating antibody-mediated neutralization ofviral particles (57). We used this assay to test the neutralizingcapacity of 1E1 with several additional reovirus isolates (20,23) than previously tested (24) in L929 cells. The results demonstratedthat 1E1 neutralizes the two newly tested type 1 isolates, butneither of the two newly tested type 2 isolates nor the twonewly tested type 3 isolates (Fig. 1). We conclude that 1E1is capable of mediating serotype-specific neutralization oftype 1 reoviruses, as recently shown to be determined by the ${sigma}$ 1-encoding S1 genome segment in the case of T1L (24). This isconsistent with previous evidence that ${sigma}$ 1 is the serotype-determiningprotein of reoviruses (57), that 1E1 binds to ${sigma}$ 1 (24), and that1E1 blocks binding of reovirus T1L particles to L929 cells (24).

View larger version (17K):
[in this window]
[in a new window]
FIG. 1. Serotype-specific neutralization of type 1 reoviruses by IgA MAb 1E1. Two additional field isolates representing each of the three established reovirus serotypes (six new isolates total) were evaluated for neutralization by 100 µg of 1E1/ml in parallel with matched samples containing no MAb. The number of plaques in each 1E1-treated sample was expressed as a percentage of that in the matched no-MAb sample. Each bar represents the mean of three or four determinations ± the standard error. The newly tested isolates were T1/Human/Netherlands/1/1984 (T1N84), T1/Human/Netherlands/1/1985 (T1N85), T2/Human/Ohio/Jones/1955 (T2J), T2/Human/Netherlands/1/1984 (T2N84), T3/Murine/France/Clone 9/1961 (T3C9), and T3/Human/Netherlands/1/1983 (T3N83) (20, 23). The results for reoviruses T1L, T2/simian/Maryland/SV59/1958 (T2S59), and T3D, evaluated for neutralization by 1E1 in an identical manner, were previously reported (24) and were added to this figure for comparison. Representatives of a putative fourth serotype (2) were not tested.

Selection of T1L mutants that resist neutralization by 1E1. Passage of reovirus T1L in the presence of 1E1 rapidly selectedfor a population of viruses resistant to neutralization by thisMAb (Fig. 2A). Clones were isolated from the first-passage stock(i) because a large increase in neutralization resistance wasalready evident, (ii) to increase the chance of isolating cloneswith distinct mutations before the fittest mutants overwhelmedthe stocks, and (iii) to increase the chance of isolating cloneswith single mutations. Seven clones were isolated by plaquepurification from a 1E1-neutralized sample of the stock (Fig.2A), serially plaque purified to promote clonal purity, andsubjected to serial passage to generate high-titer stocks. Plaquereduction assays with these new stocks revealed that three ofthe clones are indeed 1E1-resistant mutants (Fig. 2B and Table1). Six additional 1E1-resistant mutants were identified amongseven clones selected from a second, clonally distinct stockof reovirus T1L by using the same protocol (Fig. 2D and E andTable 1). This second stock of T1L was used to guarantee theselection of mutants with independent origins which possiblycontained distinct mutations that can interfere with 1E1-mediatedneutralization.

View larger version (34K):
[in this window]
[in a new window]
FIG. 2. Generation and analysis of 1E1-resistant mutants. Virus stocks were evaluated for neutralization by 100 µg of 1E1/ml or 50 µg of 5C6/ml in parallel with matched samples containing no MAb. The number of plaques in each 1E1- or 5C6-treated sample was expressed as a percentage of that in the matched no-MAb sample. (A and D) Two clonally independent stocks of reovirus T1L [designated (a) and (b)] were subjected to serial passage in the presence of IgA MAb 1E1, and the resulting passages were evaluated for neutralization by 1E1. Each bar represents the value of a single determination. Each original T1L stock is designated passage 0. For each passage series, the arrows indicate when 1E1 was added, and the asterisk indicates the sample from which seven putative 1E1-resistant mutants were isolated by plaque purification. (B, C, E, and F). The total of 14 putative 1E1-resistant mutants described in panels A and D were evaluated for neutralization by 1E1 (B and E) or 5C6 (C and F) in parallel with the original T1L stock from which each mutant was, respectively, derived. The stacked bars represent the results of two independent determinations. The names of the clones examined for MAb binding in Fig. 4 are boxed.

View this table:
[in this window]
[in a new window]
TABLE 1. Properties of 1E1- and 5C6-resistant mutants for which S1 sequences were determined

Selection of T1L mutants that resist neutralization by IgG MAb 5C6. Seven of the nine 1E1-resistant mutants also showed resistanceto neutralization by 5C6 (51, 54, 55) in initial screens (Fig.2C and F and Table 1). This finding suggested that 1E1 and 5C6bind to overlapping regions of ${sigma}$ 1 (see below). In an effort toidentify additional mutations that affect binding to 5C6, weisolated seven putative 5C6-resistant mutants of reovirus T1Lby using the same protocol as that used to generate the 1E1-resistantmutants above, except that 5C6 was added during passage (Fig.3A). Plaque reduction assays revealed that six of these clonesare indeed resistant to 5C6 neutralization (Fig. 3B and Table1). All six of these 5C6-resistant mutants also showed resistanceto 1E1 (Fig. 3C and Table 1).

View larger version (21K):
[in this window]
[in a new window]
FIG. 3. Generation and analysis of 5C6-resistant mutants. Viral stocks were evaluated for neutralization by 1E1 or 5C6 as described in Fig. 2. (A) A stock of reovirus T1L [same stock (a) as in Fig. 2] was subjected to serial passage in the presence of IgA MAb 5C6 (see Materials and Methods), and the resulting passages were evaluated for neutralization by 5C6. Each bar represents the value of a single determination. The original T1L stock is designated passage 0. The arrow indicates when 5C6 was added in the passage series, and the asterisk indicates the sample from which seven putative 5C6-resistant mutants were isolated by plaque purification. (B and C) The seven putative 5C6-resistant mutants described in panel A were evaluated for neutralization by 5C6 (B) or 1E1 (C) in parallel with the original T1L stock from which they were derived. The stacked bars represent the results of two determinations. The names of the clones examined for MAb binding in Fig. 4 are boxed.

MAb-binding phenotypes of the mutants. Neutralization-resistant mutants almost always have mutationswithin the epitope to which the neutralizing MAb binds, andthe mutations thus act directly to reduce MAb binding ratherthan through some indirect mechanism (7, 43, 45). Indirect mechanismsare nonetheless conceivable (45). We used an ELISA to determinewhether the 1E1- and 5C6-selected mutants showed reduced bindingto either 1E1 or 5C6. Five of the mutants selected for 1E1 resistanceand two of the mutants selected for 5C6 resistance were tested.All seven of the mutants showed reduced binding of both MAbs(Fig. 4 and data not shown). As for relative effects on 1E1and 5C6 binding, mutants could be grouped into two categories:one for which 1E1 binding was more strongly affected than 5C6binding (L1E1-a2 and L1E1-a7) and another for which the oppositewas true (L1E1-b5, L1E1-b6, L1E1-b7, L5C6-a3, and L5C6-a5) (Fig.4, Table 1, and data not shown). These binding results correlatedwith those from neutralization experiments (Fig. 2, 3, and 4)and provided further evidence that 1E1 and 5C6 bind to overlappingregions of ${sigma}$ 1.

View larger version (24K):
[in this window]
[in a new window]
FIG. 4. Binding of 1E1 or 5C6 to the ${sigma}$ 1 protein in purified virions of T1L or T1L-derived MAb-resistant mutants. A microplate ELISA was performed by using different concentrations of the MAbs 1E1 and 5C6, as indicated, to bind to plate-bound virions. Values are expressed as optical density (OD) units from the detection system described in Materials and Methods. Each symbol represents the mean of three or four determinations ± the standard error. In each panel, values for MAb 1E1 are shown as open symbols connected by solid lines, whereas values for MAb 5C6 are shown as symbols with cross-hairs connected by dashed lines. The symbols represent T1L(a) (squares), L1E1-a2 (diamonds), and L1E1-a7 (circles) in panel A and T1L(b) (squares), L1E1-b5 (diamonds), and L1E1-b6 (circles) in panel B. The results for L1E1-b7 were also obtained in these experiments and were indistinguishable from those shown for L1E1-b5 and L1E1-b6. The symbols represent T1L(a) (squares), L5C6-a3 (diamonds), and L5C6-a5 (circles) in panel C. These samples were analyzed in the same experiments as for panel A, and so the data for T1L(a) are the same as presented in panel A.

Sequence substitutions in the S1 genome segments and encoded ${sigma}$ 1 proteins of the 1E1- and 5C6-resistant mutants. To localize the probable regions to which 1E1 and 5C6 bind inthe ${sigma}$ 1 protein, we determined the nucleotide sequences of theS1 genome segments of the two parental T1L clones, the sevenmutants characterized by ELISA, and five of the eight othermutants (Table 1). Sequences were directly determined by cyclesequencing after amplification of DNA copies of the respectiveS1 genomic RNA segment. Relative to the S1 sequence of its respectiveparent, each mutant contained either one or two nucleotide changesthat caused either one or two amino acid changes in the deduced ${sigma}$ 1 protein sequence (Table 1). Each mutant selected for 1E1 resistanceshowed a change at one of three amino acid positions (415, 417,and 445), and each mutant selected for 5C6 resistance showedchanges at one or two of three amino acid positions as well(417, 447, and 251 and 417) (Table 1). A mutation at one ofthese sites (Gln417Lys) was shared between the two sets of mutants,providing further evidence that 1E1 and 5C6 bind to overlappingepitopes. All but one of the mutations was found in the regionbetween amino acids 415 and 447, which constitutes part of thehead domain of T1L ${sigma}$ 1 (13, 18, 19, 36). The one mutation outsidethis region, Ile251Thr, was found in a mutant (L5C6-a3) thatalso contained a Gln417Lys mutation, and we consider it likelythat the Ile251Thr mutation arose by chance in this clone anddoes not contribute to its MAb resistance phenotypes (Fig. 3and 4; see also below). The two parental T1L sequences wereidentical but differed from the published T1L sequences (17,36) at two nucleotide positions (C393A and A1416G), which causedtwo changes in the deduced ${sigma}$ 1 protein sequence (Ser127Tyr andAsn468Ser) and one change in the deduced ${sigma}$ 1s protein sequence(Pro107Thr). All 12 of the 1E1- and 5C6-resistant mutants forwhich S1 sequences were determined contained these two additionalnucleotide changes relative to the published T1L S1 sequencesand were thus identical to their parents in that regard.

Locations of the 1E1- and 5C6-selected mutations in the ${sigma}$ 1 structure. The structure of the C-terminal half of the T3D ${sigma}$ 1 trimer (aminoacids 246 to 455) was recently determined at 2.6 Å byX-ray crystallography (13). Alignment of the T1L and T3D ${sigma}$ 1 sequences(17, 36) allowed conserved regions to be identified and thenlocated in the T3D structure (13). The conserved regions includea surface loop in the head domain that is proposed to interactwith the cell surface protein receptor JAM1 (13, 41) (Fig. 5).Although the T1L ${sigma}$ 1 structure is not yet known, the T3D ${sigma}$ 1 trimerstructure can be used as a model to locate the approximate locationsof the changes that impart 1E1 and 5C6 resistance to the currentset of T1L mutants. Amino acids 246 to 455 in the T3D ${sigma}$ 1 crystalstructure correspond to amino acids 258 to 470 in the alignedT1L ${sigma}$ 1 sequence (17, 36). The resistance-associated changes atamino acids 415, 417, 445, and 447 in T1L ${sigma}$ 1, respectively, alignwith amino acids 400, 402, 430, and 432 in T3D ${sigma}$ 1 (17, 36). Allfour of these residues are partly exposed on the side surfaceof the T3D ${sigma}$ 1 trimer head and cluster within 25 Å of eachother across an interface between two adjacent subunits (Fig.5). There are three such clusters per trimer. Within each ofthese putative MAb-binding regions, residues 415 and 417 arecontributed by one subunit and residues 445 and 447 by the adjacentsubunit (Fig. 5). Since mutations on both sides of each interfacereduce binding of both 1E1 and 5C6 (Fig. 4), we conclude thatthese MAbs bind to substantially overlapping regions that spanthe subunit interface. These findings suggest that binding of1E1 and 5C6 is likely dependent on the intact quaternary structureof the trimeric ${sigma}$ 1 head domain, a finding consistent with previousevidence that both 1E1 and 5C6 recognize a complex epitope thatis sensitive to ${sigma}$ 1 folding (24, 55). Amino acid 251 in T1L ${sigma}$ 1,which was identified as the site of a second mutation in 5C6-resistantmutant L5C6-a3 (Table 1), is outside the region visualized inthe T3D ${sigma}$ 1 crystal structure (13) and is thus distant from theother neutralization resistance mutations in the protein structure.This finding supports the conclusion that the mutation at residue251 does not contribute to the resistance phenotype of thatmutant.

View larger version (103K):
[in this window]
[in a new window]
FIG. 5. Approximate locations of the 1E1 and 5C6 resistance mutations based on the T3D ${sigma}$ 1 crystal structure. The crystal structure of the reovirus T3D ${sigma}$ 1 trimer (13) is shown in wall-eyed stereo view and space-filling format, with each subunit colored a different shade of blue. The positions of ${sigma}$ 1 mutations in the A2- and G5-resistant mutants of T3D (6, 46) are shown in yellow in each subunit. The positions of ${sigma}$ 1 mutations in the 1E1- and 5C6-resistant mutants of reovirus T1L (the present study) are shown in red in each subunit. The pattern and colors of the mutant positions are replicated in the space between the two trimer images and labeled with the position numbers of the mutations. For the 1E1- and 5C6-resistant mutants, each mutation is indicated by position in the T1L ${sigma}$ 1 sequence (upper), as well as by position in the aligned T3D ${sigma}$ 1 sequence (lower) (17, 36). The latter positions are the ones mapped on the T3D ${sigma}$ 1 trimer structure. A red oval indicates the approximate footprint (13 to 30 amino acids [7, 40]) expected to be covered or contacted by one epitope-binding Fab region centered around the identified mutations. Two other subunit interfaces flanked by MAb resistance mutations are hidden on the back of the trimer in this view. Residues 429 and 442 are shown in stick format to reveal residues 430 (T1L 445) and 402 (T1L 417) more fully. The putative JAM1 binding loop (4, 13) is shown in dark green in each subunit. The region of T1L ${sigma}$ 1 that contains type-specific determinants of hemagglutination (i.e., carbohydrate receptor binding) (11) is shown in light green for the subunit otherwise shown in light blue.

Cell adherence phenotypes of the mutants. In an effort to link other defects or changes with the neutralizationresistance mutations in T1L ${sigma}$ 1 identified in the present study,we analyzed the 1E1- and 5C6-resistant mutants with regard toseveral other properties. In hemagglutination assays, the mutantswere indistinguishable from their wild-type parents (data notshown). They were also indistinguishable from the parents withregard to relative infectivity in L929 cells (Table 1), whichis not surprising since these are the cells in which the mutantswere selected and amplified. ISVPs of representative mutantswere next compared to those of the T1L parents for attachmentto M cells in an overlay assay on tissue sections of rabbitPeyer's patches (22). Each mutant mimicked its parent in thateach showed selective adherence to M cells in the sections (Fig.6, Table 2, and data not shown). In addition, each mutant mimickedits parent with regard to inhibition of this adherence whenthe sections were preincubated with the lectins MAL-I and MAL-II,which specifically recognize oligosaccharides containing ${alpha}$ 2,3-linkedsialic acid, but not when the sections were preincubated withthe lectin SNA, which specifically recognizes oligosaccharidescontaining ${alpha}$ 2,6-linked sialic acid (Fig. 6, Table 2, and datanot shown) (22). As a final test, ISVPs of the same representativemutants were compared to those of their parents for attachmentto confluent, polarized Caco-2_BBe cell monolayers (22), andno differences were observed between mutants and parents withregard either to adherence to these cells or to the patternof adherence inhibition by the preceding lectins (Table 2 anddata not shown). Thus, the defined alterations in the ${sigma}$ 1 proteinsof the 1E1- and 5C6-resistant mutants did not affect their capacityto recognize specific glycoconjugates containing ${alpha}$ 2,3-linkedsialic acid and to adhere to epithelial cell apical surfaces.

View larger version (70K):
[in this window]
[in a new window]
FIG. 6. Binding to rabbit M cells by reovirus 1E1-resistant mutant L1E1-a7. Deparaffinized 5-µm sections of rabbit Peyer's patch were incubated with or without MAL-II lectin, washed, and overlaid with biotinylated ISVPs, which were then visualized with streptavidin-TRITC. In all micrographs, a representative follicle-associated epithelium (FAE) and the epithelium of an adjacent villus are shown. (A) ISVPs of the parental T1L strain adhered to M cell apical surfaces (arrows) in the follicle-associated epithelium (FAE; arrowhead). (B) In the absence of MAL-II, L1E1-a7 ISVPs adhered to M-cell apical surfaces in a pattern indistinguishable from that of the parent strain. (C) M cells in adjacent sections of Peyer's patch mucosa preincubated with MAL-II, in contrast, did not bind L1E1-a7 ISVPs. Bar, 25 µm.

View this table:
[in this window]
[in a new window]
TABLE 2. Binding profiles of representative 1E1- and 5C6-resistant mutants^a

DISCUSSION

Top

Discussion

During its pathogenesis in the intestine of adult mice, reovirusT1L first interacts with the apical surfaces of epithelial Mcells to enter transcytotic vesicles and cross the epithelialbarrier; once in the mucosa, the virus then binds to targetcells, such as the basolateral membranes of epithelial cells,to initiate a productive infection (60, 61; reviewed in references34 and 37). Our recent work has suggested that IgA MAb 1E1,directed against the reovirus T1L ${sigma}$ 1 protein, can protect theintestine potentially at both stages, by blocking attachmentto epithelial cell apical membranes and thereby preventing entryinto the mucosa and also by blocking infection of target cells,as shown by neutralization of infectivity in L929 cells (24).The results of the present study show that this protective andneutralizing IgA antibody and a protective and neutralizingIgG antibody, 5C6, recognize closely related epitopes that spanadjacent ${sigma}$ 1 monomers in the trimeric ${sigma}$ 1 head domain.

Binding of type 1 reoviruses to apical membranes of epithelialcells, where the receptor protein JAM1 is not normally present(32), involves particular carbohydrate structures (22). We haverecently shown that glycoconjugates containing ${alpha}$ 2,3-linked sialicacid on apical M cell surfaces are important for the ${sigma}$ 1-basedbinding of type 1 reoviruses to those cells (22). The part(s)of ${sigma}$ 1 required for this binding activity remain to be determinedbut could involve the distal region of the type 1 ${sigma}$ 1 tail [T(iv)]that is required for hemagglutination (11) or another, as-yet-unknowncarbohydrate recognition domain in this protein. The 1E1 and5C6 results reported here are consistent with the hypothesisthat the head domain of type 1 ${sigma}$ 1 is important for binding toM cells in vivo. The 1E1 and 5C6 epitopes in the ${sigma}$ 1 head maynot appear to be well situated for steric hindrance of the carbohydrate-bindingregion in the ${sigma}$ 1 tail, but it is important to note that antibodymolecules are relatively large. Indeed, IgA dimers visualizedby electron microscopy are 30 to 35 nm in length (14), and thereforeantibody binding to the 1E1 or 5C6 epitope could theoreticallyblock interaction of both the head and tail domains of ${sigma}$ 1 withcell surface receptors. Moreover, the carbohydrate structuresthat contribute to the receptors for type 1 reoviruses on intestinalepithelial cells may not be well exposed on the lumenal cellsurface but rather may be located on the oligosaccharide sidechains of large, complex membrane glycoproteins (29, 30) oron glycolipids (26) sequestered under the membrane glycoproteincoats of intestinal epithelial cells. This idea is supportedby the observation that the lectin MAL-II, which blocks bindingof type 1 reoviruses to epithelial cells (22), binds to rabbitM cells when it is applied to the mucosa in soluble form butnot when it is conjugated to 1-µm microparticles (33).We have previously observed that proteolytic conversion of T1Lvirions to ISVPs is required for M-cell binding in mice (1)and have hypothesized that extension of the ${sigma}$ 1 trimer from thesurface of ISVPs may promote M cell binding by allowing thehead region to penetrate the apical surface glycoprotein coatsof epithelial cells and reach sequestered binding sites nearthe plasma membrane (33). Attachment of IgA antibodies to the ${sigma}$ 1 head might therefore not only sterically hinder receptor bindingbut also impede access of the extended ${sigma}$ 1 trimer to sequesteredreceptor sites on epithelial apical surfaces.

To infect target cells within the mucosa, type 1 reovirusesmay exploit both carbohydrate structures and membrane proteinsas receptors. Through analysis of antibody escape mutants inL929 cell infectivity assays, we have shown that the epitopesrecognized by IgG MAb 5C6 and IgA MAb 1E1 overlap. Both antibodiesblock binding of T1L to fixed L929 cells and neutralize T1Linfectivity, and although their affinities for ${sigma}$ 1 may differ,both antibodies most likely protect target cells by the samemechanism. Neutralizing antibodies often act at the surfacesof target cells, by hindering virus-receptor interactions (28,45, 59). A conserved loop in the ${sigma}$ 1 head is proposed to be animportant part of the JAM1 binding site on target cells includingL929 cells (13). 1E1 and 5C6 binding near the top of the ${sigma}$ 1 head,as revealed by the present results, appear well situated toeffect steric hindrance of JAM1 binding at the proposed sites(13), especially if the contact regions are larger than currentlysuggested and extend laterally toward the ${sigma}$ 1 subunit interfaces(13). However, stable attachment of reoviruses leading to endocyticuptake from cell surfaces may be a multistep process involvingsuccessive or concurrent interactions with lower- and higher-affinitycarbohydrate and protein receptor molecules (3), and a particularneutralizing antibody may interfere with one or more of thesesteps. We have suggested that type 1 reoviruses may bind tospecific glycoconjugates containing ${alpha}$ 2,3-linked sialic acid onM-cell surfaces (22), and these carbohydrate structures maybe present on target cells as well. Both 5C6 (44, 55) and 1E1(24; data not shown) are effective at blocking agglutinationof human erythrocytes and attachment to M cells, suggestingthey do indeed interfere with carbohydrate-receptor bindingby some mechanism. However, binding of type 1 reoviruses toL929 cells is clearly not dependent on the specific carbohydratestructures that contribute to the receptors for these viruseson epithelial cells, because L929 cell binding is not inhibitedin the presence of MAL-II lectin (22). In any case, if carbohydraterecognition sites on type 1 ${sigma}$ 1 play a role in L929 cell infection,their capacity for such function is greatly reduced in the presenceof 1E1 and 5C6. Taken together, the data suggest that theseantibodies sterically hinder the binding of type 1 reovirusesto both JAM1 and carbohydrate receptors on target cells.

Neutralizing antibodies can hinder attachment by other mechanismsas well. For example, they can induce conformational changesin the viral surface proteins that reduce affinity for receptorbinding at an adjacent or more distant site (28, 45). This mechanismseems plausible for the adjacent, proposed JAM1-binding sitesin T1L ${sigma}$ 1 but possibly less plausible for the putatively moredistant carbohydrate-binding sites. Nonetheless, by bindingacross the subunit interfaces in the ${sigma}$ 1 head, 1E1 and 5C6 mightcontort the intratrimer interactions and the receptor-bindingregions in turn, even at some distance from the MAb-bindingsites. The discovery that the overlapping 1E1 and 5C6 epitopesspan subunit interfaces in the ${sigma}$ 1 head suggests an additional,distinctive mechanism of neutralization, namely, that 1E1 and5C6 may "staple" together the trimeric heads such that theycannot dissociate into their constituent monomers. Dissociationof the heads may be required, for example, to allow each trimerto mediate multivalent attachment in which two, or even allthree, of the dissociated heads concurrently bind to separateprotein or carbohydrate receptor molecules. Multivalent attachmentto cell surface receptors may be required for reovirus infection(3), so that blocking it may be an effective means of neutralization.In considering this hypothesis, it is notable that mutants ofreovirus T3D resistant to the neutralizing IgG MAb G5 (8, 46)have a change at ${sigma}$ 1 residue 340 or 419 (6), which also approacheach other from opposite sides of a subunit interface in thetrimeric T3D ${sigma}$ 1 head (13) (Fig. 5). Because residues 340 and419 in T3D ${sigma}$ 1 are near the proposed JAM1-binding sites in thestructure (13), it is reasonable to hypothesize that G5 mediatesneutralization by sterically hindering JAM1 binding or by inducingconformational changes in nearby regions that reduce affinityfor the receptor. Because the G5 epitope also spans the subunitinterface, however, it may be that neutralization is enhancedor wholly mediated by G5 stapling together the trimeric headof T3D ${sigma}$ 1, as we suggest above for 1E1 and 5C6 on T1L ${sigma}$ 1. Thecapacity of the ${sigma}$ 1 heads to undergo monomer-trimer transitionsduring the course of cell entry has been previously predictedbased on several features of the T3D ${sigma}$ 1 crystal structure (13,48).

The mutations identified in the present study indicate thatthe two type 1 ${sigma}$ 1-specific neutralizing MAbs 1E1 and 5C6 bindto essentially the same region of T1L ${sigma}$ 1. This was corroboratedby the capacity of 5C6 to compete with 1E1 binding to viralparticles in dot blot assays (data not shown). Not only werethese two MAbs generated at different times and in differentlabs, but they were generated with cells from different lymphaticorgans (Peyer's patch versus spleen) and are of different classes(IgA versus IgG), proving their independent origins. The discoverythat they bind to the same region of ${sigma}$ 1 suggests that (i) antibodiesof these two distinct classes are mediating neutralization bythe same mechanism, (ii) antibody binding to that region isespecially effective at neutralization, and/or (iii) this regionis especially immunogenic. A mutation at residue 419 in T3D ${sigma}$ 1, which is present in most of the G5-resistant mutants, isalso present in the two T3D mutants selected for resistanceto another type 3 ${sigma}$ 1-specific neutralizing IgG MAb, A2 (6, 8).In addition, A2 and G5 compete for binding to viral particlesin radioimmunoassays (47). A2 and G5 therefore appear to bindto very similar regions, which in turn are near the regionsbound by 1E1 and 5C6 (Fig. 5). Intraperitoneally administeredG5 is protective against reovirus type 3 infections of targetcells outside the mouse intestine (51, 52, 56). Thus, four stronglyneutralizing and protective IgA or IgG MAbs—A2, G5, 1E1,and 5C6—all bind to adjacent regions on the side of the ${sigma}$ 1 head and across the interface between ${sigma}$ 1 subunits. Becausethese epitopes all span subunit interfaces in the ${sigma}$ 1 head, theyare likely to require the intact quaternary (trimer) structureof the head for binding. Additional ${sigma}$ 1-specific neutralizingMAbs and the capacity to map their binding epitopes should contributeto further understanding the mechanisms of neutralization andprotection in the reovirus system.

ACKNOWLEDGMENTS

We thank Jason Dinoso, Stephanie Ferrant, Elaine Freimont, andMalen Link for technical assistance and other members of ourlaboratories, especially Teresa Broering, for helpful discussionsand suggestions about RT, DNA amplification, and DNA sequencing.Melina Agosto, Michelle Arnold, and John Parker provided carefulreviews of the manuscript.

Funding was provided by the Harvard Digestive Diseases Centerthrough NIH center grant DK34854, as well as by NIH R01 grantsAI46440 (M.L.N.) and HD17557 (M.R.N.). C.L.M. received additionalsupport as an Ernst Fellow of the Department of Microbiologyand Molecular Genetics. K.S.M. received additional support asa Harold Amos Fellow of the Division of Medical Sciences andthrough a Research Supplement for Underrepresented Minoritiesattached to NIH grant AI46440.

FOOTNOTES

* Corresponding author. Mailing address: Department of Microbiology and Molecular Genetics, Harvard Medical School, 200 Longwood Ave., Boston, MA 02115. Phone: (617) 645-3680. Fax: (617) 738-7664. E-mail: mnibert{at}hms.harvard.edu.

Present address: Department of Histology, Microbiology, andMedical Biotechnologies, University of Padua, 35121 Padua, Italy.

REFERENCES

Top