INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Porcine reproductive and respiratory syndrome virus (PRRSV) is a recently emerged pathogen of domesticated swine. The virus, which belongs to the Arteriviridae family, has a 15-kb positive-sense, single-stranded RNA genome. PRRSV encodes an approximately 4,000-amino-acid large replicase polyprotein (open reading frame [ORF] 1a and 1b) and six structural proteins of 130 to 265 amino acids (ORFs 2 to 7) (reviewed in references 6, 36, and 45). The replicase polyprotein is processed by autoproteolytic cleavage into nonstructural protein fragments (Nsps). The replicase polyprotein processing cascade has recently been reviewed by Ziebuhr et al. (45), and the Nsp and protease domain nomenclature suggested by Ziebuhr et al. is used throughout this article. Two main PRRSV genotypes exist, the American (US) and European (EU) types, which are only approximately 60% identical at the nucleotide level. For reasons currently not understood, these two distantly related PRRSV types emerged virtually simultaneously on their respective continents in the late 1980s. Since then, intermingling of the genotypes has occurred through the use of a live, US-type PRRSV vaccine in Europe (2, 18).

PRRSV infection poses a challenge to current serodiagnostic and vaccination strategies. Although live PRRSV vaccines provide protection against homologous challenge, the genetic diversity of field PRRSV isolates is very high, and vaccine effect against heterologous challenge may be limited (39). Also, live PRRSV vaccines have been observed to revert to virulence (2, 38), and the safer, killed vaccines have so far proved less effective (28). Finally, PRRSV can persist in some animals despite high levels of antiviral antibodies; in boars, this is associated with long-term, intermittent seminal excretion of the virus (3, 13). Serological tests cannot discriminate between seropositive animals which have cleared PRRSV infection and carrier animals. Addressing these problems might involve improving the design of antigens for vaccines and diagnostic tests. For example, PRRSV envelope glycoproteins are candidates for use in subunit vaccines (11), and the nonstructural ORF 1 polyprotein could be a possible candidate antigen for the development of serological tests to identify carrier animals (24). Production of full-length recombinant proteins to explore such potential applications may be hampered by the presence of hydrophobic regions and, for the ORF 1 polyprotein, a very large size and the ability to undergo autoproteolytic cleavage. This could be overcome by expression of protein subunits, guided by prior knowledge of naturally antigenic regions. However, no knowledge currently exists about epitopes in the PRRSV ORF 1 nonstructural polyprotein, and only two epitopes have been mapped in the envelope glycoproteins (23, 30), one of which was mapped using monoclonal antibodies (MAbs) and thus is of unknown significance for PRRSV infection in vivo (23). To address these questions, we have in the present study screened 97% of the PRRSV protein mass for linear porcine B-cell epitopes using phage display.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of phage libraries of random PRRSV fragments. Routine M13 methods, such as phage amplification in liquid cultures of F' Escherichia coli, purification of phages from E. coli culture supernatants by polyethylene glycol (PEG)-NaCl precipitation, titration of phages, and preparation of phage DNA for sequencing were done as described in reference 35 and the Ph.D-7 phage display kit manual from New England Biolabs (Hitchin, Hertfordshire, U.K.).

Experimental infections and characterization of PRRSV-immune sera. Six specific-pathogen-free pigs, 4 weeks of age, were intranasally inoculated with 10^5.4 50% tissue culture-infective doses (TCID₅₀) of PRRSV 111/92, the same Danish European-type isolate used for phage library construction. Serum samples were prepared prior to infection (0 days postinfection [dpi]) and at 3- to 7-day intervals thereafter for up to 56 days. The pigs were euthanized at 42 dpi (pigs 9 and 10) or 56 dpi (pigs 11, 12, 13, and 14), at which time tonsil homogenates, lung homogenates, and lung washings were tested for PRRSV by RT-PCR (29).

Selection of phage libraries with porcine sera and characterization of selected phages by sequencing. Serum to be used for selection of phage libraries was filtered (0.45-µm pore size), sodium azide was added to 0.02%, and UV-inactivated M13KO7 helper phage (New England Biolabs) was added to 10¹² particles/ml. Sera were stored at 4°C. Prior to being used in the selection described below, protein A- and G-Sepharose particles (Pharmacia) were blocked for 30 min at room temperature (or 4°C overnight) in bovine serum albumin (BSA) (100 mg/m). The selection protocol below was modified from the PhD-7 kit manual (New England Biolabs). First, to remove nonspecifically binding phage, 10 µl of phage library (typically 10¹¹ PFU) was added to 190 µl of PBS-0.5% Tween 20-20% protein A-Sepharose in a polypropylene tube and incubated for 20 min at room temperature with vigorous shaking. The Sepharose particles were centrifuged for 20 s in a tabletop microcentrifuge and discarded. The supernatant was transferred to a new polypropylene tube, 0.25 to 2 µl of whole serum was added, and the phage-serum mix was incubated for 20 min at room temperature with shaking. To isolate antibody-bound phages, protein A-Sepharose was added to 20%, and the tubes were incubated for a further 15 min at room temperature with shaking. The Sepharose particles were washed 15 times with 1 ml of PBS-0.5% Tween 20 and once with 1 ml of PBS without detergent. During washing, the Sepharose suspensions were twice transferred to new polypropylene tubes to reduce the background of nonspecifically binding or adsorbing phage. Finally, phages were eluted through a 7-min incubation of the Sepharose in 1 ml of 200 mM glycine-1 mg of BSA per ml (pH 2.2). The Sepharose particles were centrifuged for 60 s in a table-top microcentrifuge and discarded. The eluate was neutralized by adding 150 µl 1 M Tris (pH 9.1) and stored at 4°C. At this stage, the phage content of the eluate (the output from the selection) and of the phage library (the input to the selection) were determined by titration on E. coli ER2537 (New England Biolabs). Eluates were amplified in 30 ml of 100-fold-diluted overnight culture of E. coli ER2537 in LB medium. Eluates thus amplified were subjected to further cycles of selection with porcine sera, as described above. Typically, a total of three or four cycles of selection and amplification were carried out. To reduce the occurrence of background phage, protein A and protein G-Sepharose particles were used alternately in successive selection cycles.

Phage ELISA. Phages for ELISA were plaque purified and underwent two rounds of PEG-NaCl precipitation. The particle content of phage preparations was determined spectrophotometrically by measurement of the absorbance at 269 nm (5). A total of 4 × 10¹⁰ phage particles in 100 µl of PBS were added per well to Maxisorp U-96 ELISA plates (Nunc, Life Technologies, Tåstrup, Denmark) and incubated at 4°C overnight. Prior to addition of the antibody layers, plates were blocked for 1 h at room temperature with 10% dried skimmed milk in PBS (Blotto). In the following, reagents were used at 100 µl/well, incubations, unless otherwise stated, were for 1 h at room temperature (with shaking), and washes between antibody layers were done with 0.5 M NaCl-15 mM Na₂HPO₄-2 mM KH₂PO₄-0.05% Tween 20 (ELISA buffer).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Selection with PRRSV-immune sera enriches for phages displaying PRRSV epitopes. Three segments of the PRRSV genome were RT-PCR amplified (Table 1), cloned, fragmented by DNase treatment, and used to construct three separate phage display libraries. The RT-PCR amplicon used for the ORF 1 library contained 94% of the ORF 1 coding sequence, lacking only 485 nt of the 3' end of ORF 1b. The two RT-PCR amplicons used for ORF 2-3 and ORF 4-7 library construction provided complete, overlapping coverage of all the structural genes of PRRSV. The cloning strategy described in Materials and Methods resulted in the surface display of random PRRSV fragments of 10 to 50 amino acids, fused to the N-terminus of the phage pIII protein, which is present in three to five copies at one end of the filamentous M13 virion. The libraries were selected in parallel with porcine sera taken at 42 days post-PRRSV infection (42-dpi sera) and with preinfection sera from the same animals (0-dpi sera). The 42-dpi sera used for selection had high (>1:1,000) antibody titers against PRRSV in routine diagnostic assays and contained neutralizing antibodies. In 0-dpi sera, no anti-PRRSV antibodies could be detected. The phage display libraries underwent three or four rounds of selection, where the eluted phage population from one selection was amplified and used for a successive selection (Table 2). To bias the screening towards commonly recognized epitopes, sera from two individual pigs were used alternately in successive selection cycles. Also, to bias the screening towards identifying epitopes which induce higher-affinity antibodies, progressively lower levels of serum were used for each selection (Table 2). For each selection, we measured the total amount of phage added and the total amount of phage recovered and calculated an output-input ratio by dividing these two values. Reproducibly, the output-input ratio increased through successive selections with 42-dpi sera, indicating that the phage population was increasingly enriched for antibody binders (Table 2). To assess how much of this enrichment was due to nonspecific reaction of porcine antibodies against M13 virions, the output-input ratios obtained with 42-dpi sera were divided by the output-input ratios obtained in parallel selections with 0-dpi sera to yield specific enrichment ratios. The specific enrichment ratios also increased through successive selections (Table 2), indicating that our selection strategy was effective in specifically enriching the libraries for phages that reacted with anti-PRRSV antibodies, ie, phages that likely expressed PRRSV epitopes. After three or four rounds of selection, the specific enrichment ratios were at least 10, which indicated that the majority of phages selected with 42-dpi sera might express PRRSV epitopes. At this stage, the inserts displayed by individual phages were determined by DNA sequencing.

A total of 83% of putative epitope sites identified by sequencing are real and are not unique to the sera used for library selection. Plaque-purified phage clones representing each putative ES were used as the antigen in total Ig ELISA (Fig. 1A). If more than one phage clone identified an ES, the phage clone displaying the longest insert was used for ELISA in all cases. Of the 12 putative ES identified by sequencing, 10 could be verified serologically (Fig. 1A). Experimental pig sera (42-dpi sera) reacted with these ES, whereas a panel of known negative sera did not. Also, experimental 0-dpi sera did not react with the ES (not shown in Fig. 1A, but see Fig. 4). A phage representing one of the epitope sites that could not be verified serologically (ES10) was used as a negative-control antigen. Experimental 42-dpi sera did not react with this phage, which displayed a 14-amino-acid nonepitope fragment of the PRRSV helicase (Fig. 1A, hel). Importantly, the serologically confirmed 10 ES reacted not only with the two pig sera that were used for library selection but also with one or more of the other experimental sera (Fig. 1A), indicating that our selection procedure (Table 2) had been successful in isolating B-cell epitopes that are commonly recognized by pigs. For most ES (except the weak ES1 and ES11), an IgA response could also be detected (Fig. 1B). Unsurprisingly, a strong reaction in total Ig ELISA generally correlated with a strong reaction in IgA ELISA. More remarkably, the opposite was also seen, with ES12 exhibiting a strong reaction with all experimental sera in total Ig ELISA but a very weak (albeit positive) reaction with all experimental sera in IgA ELISA (Fig. 1, compare A and B).

View larger version (48K): [in this window] [in a new window]   FIG. 1.   Serological confirmation of ES by ELISA. ELISA to detect total porcine Ig or porcine IgA directed against the different ES was done as described in Material and Methods. Hel is a phage displaying a 14-amino-acid fragment from the PRRSV helicase which is not an epitope, i.e., a negative-control antigen. To estimate the specific reactivity of porcine sera against the phage-displayed PRRSV peptides and to correct for any unspecific reactivity against the M13 virion, the OD450 ratio (y axis) was calculated as described in Materials and Methods. The 19× neg (A) and 14× neg (B) columns show the mean reaction of negative serum panels, with bars indicating 1 standard deviation. Experimental sera (pigs 9, 10, 11, 12, 13, and 14) were from 42-dpi PRRSV 111/92 infection. (B) Isotype control bars indicate the reactivity of the indicated pig serum using an isotype-matched MAb instead of the anti-IgA MAb. Sera from pigs 10 and 12 were used for phage library selection, as described in Table 2, footnote a.

View larger version (68K): [in this window] [in a new window]   FIG. 2.   Sequences of serologically confirmed ES. For each ES, phage-displayed sequences are shown in the top box. The phage-displayed sequences were derived from PRRSV isolate 111/92, which was used for library construction. The sequences shown were fused to the phage pIII protein, in the context MKKLLFAIPLVVPFYSHS*A?AQTVQSCLA, where the pIII leader peptide is underlined, * is the predicted leader peptidase cleavage site, ? is the phage-displayed PRRSV sequence shown in the figure, and naturally phage-encoded residues immediately neighboring the displayed sequence in the mature pIII protein are in italics. The published PRRSV sequence is shown in the bottom box, annotated and aligned to the phage-displayed sequences. For ES in ORF 1, only published Lelystad PRRSV sequence is shown. For ES in the structural genes (sites 11 and 12), published 111/92 and Lelystad PRRSV sequences are shown (111/92 top, Lelystad bottom). The sequences for two ES which could not be confirmed by ELISA (ES8 and ES10) are not shown. Amino acid (aa) homologies were calculated between the phage-displayed sequences and Lelystad PRRSV, in all cases omitting the first and last of the phage-displayed residues, which might have been corrupted during the cloning procedure. For ES12, the 5 C-terminal residues in the first phage sequence and the three N-terminal residues in the second sequence likely represented a ligation artifact and were omitted from the homology calculation. For each phage clone, the number of times the clone was observed (numerator) in the total number of sequenced phages from a given library (denominator) is indicated. Altogether 57, 95, and 61 phage clones were sequenced from the ORF 1, ORF 2-3 and ORF 4-7 libraries, respectively. Where more than one phage identified an ES, the phage displaying the longest PRRSV sequence was used as the ELISA antigen in all cases.

View larger version (32K): [in this window] [in a new window]   FIG. 3.   Positions of serologically confirmed ES in the PRRSV genome. The positions of the serologically confirmed ES are shown by black boxes. Box width is proportional to the total length of the ES, as defined in Fig. 2. Only ORF 1 features of relevance to the identified ES are shown, using the protease domain and Nsp nomenclature suggested by Ziebuhr et al. (45). Arrowheads indicate predicted autoproteolytic cleavage sites in the ORF 1 replicase polyprotein. For the PCP1 / cleavage site, the question mark indicates that the exact position is not known. PCP1, PCP1, and CP2 denote accessory protease domains; the Nsp4 fragment is the main arteriviral protease. RKASLSTS is a previously identified ORF 3 epitope (30). Hopp-Woods antigenicity predicitions were made using the built-in option of the Omiga software and a 17-amino-acid sliding window.

Examination of seroconversion kinetics and titers. Pigs seroconverted to the 10 ES at different times postinfection in what appeared to be a consistent pattern (Fig. 4). Seroconversion was first noted towards ES12 (ORF 4 envelope glycoprotein) and ES7 (Nsp2). Very late seroconversion was seen towards ES1 (ORF1a) and ES11 (ORF 3 envelope glycoprotein). Thus, the kinetics of seroconversion did not discriminate between ES in structural and nonstructural parts of the viral genome. This is the first report of ES in the PRRSV replicase polyprotein, and it is noteworthy that some of the nonstructural ES (ES7, Fig. 4) were targeted very rapidly by the pig antibody response. To examine the relationship between the speed and the magnitude of the porcine antibody response, total Ig titers towards all ES were determined at 42 dpi (Table 3). We found that the speed with which ES were recognized during infection in vivo (Fig. 4) was not generally predictive of the 42-dpi titers (Table 3), yet the 42-dpi titers in Table 3 are unlikely to represent a "final" antibody response; it is possible that ES exhibiting very late seroconversion such as ES1, ES5, and E11 (Fig. 4), attained titers higher than those shown in Table 3 at later times postinfection. In support of this, we observed that seroconversion to ES11 did not occur untill 56 dpi in one pig (not shown). It is an intriguing possibility that the seroconversion and antibody titer towards such very late epitopes might be informative about late phases of PRRSV replication.

View larger version (18K): [in this window] [in a new window]   FIG. 4.   Seroconversion kinetics of six experimentally infected pigs against phage-displayed PRRSV ES. Serum samples taken at 0, 14, 21, and 42 dpi from six experimentally PRRSV-infected pigs were examined in total Ig ELISA against all the ES identified in this study. Seropositive/seronegative scoring was done as described in Materials and Methods. Hel is a phage displaying a 14-amino-acid fragment from the PRRSV helicase which is not an epitope. No pigs seroconverted to this negative-control antigen. Viremia levels were examined on more closely spaced serum samples than seroconversion, as indicated.

Fine mapping of ES12. ES12 corresponded to an epitope identified by murine MAbs and peptide scan in the Lelystad virus ORF 4 protein by Meulenberg et al. (23). Based on the reaction of 15 of 15 MAbs raised against recombinant ORF 4 protein with this site, the authors noted that the region was immunodominant in mice but did not examine its relevance in pigs. The anti-ORF 4 MAbs made by Meulenberg et al. were neutralizing in vitro (23). We observed that high anti-ES12 titers coexisted with viremia in individual animals (not shown) as well as in the experimental group as a whole (Table 3). Thus, the protective effect of anti-ES12 antibodies appeared to be limited in vivo. Interestingly, a recent study showed that anti-ORF 4 MAbs are inefficient at neutralization in vitro compared to anti-ORF 5 MAbs (42). Six of six experimentally infected pigs consistently produced very high antibody titers towards ES12 (Table 3 and Fig. 5). In one pig, half-maximal ELISA reaction was not reached untill a serum dilution of 1:12,500 to 1:62,500 (Table 3), and in two pigs, weak positive ELISA reactivity was still evident at a 1:62,500 serum dilution (not shown). Thus, our study confirmed the immunodominance of ES12 in pigs. In addition to inducing high antibody titers, ES12 was special in that it appeared to be deficient in IgA induction (Fig. 1, compare A and B). It remains to be determined whether this deficiency is one of IgA amount, affinity, or both. The deficiency might also be influenced by host factors, because serum from 1-year-old experimentally PRRSV-infected boars in some cases produces a stronger anti-ES12 signal in the IgA ELISA (M. B. Oleksiewicz, unpublished).

View larger version (28K): [in this window] [in a new window]   FIG. 5.   Fine mapping of the immunodominant ES12. Serum samples taken at 42 dpi from six experimentally infected pigs were titrated in ELISA on each of the five phage clones defining the ORF 4 epitope site (ES12). Each data point represents the average ELISA reactivity of the six experimental sera, and bars indicate 1 standard deviation. The boxed residues in the Lelystad PRRSV ORF 4 sequence indicate a core epitope previously determined by Meulenberg et al. using murine MAbs and Pepscan (23). The boxed residues in the 111/92 ORF 4 sequence indicate a core epitope defined by the sequence overlap between the three most strongly reacting phage clones (6-1, 6-2, and 6-62). Arrows indicate two highly conserved cysteine residues, and the heavy horizontal black bar indicates PRRSV sequence which is hypervariable and deletion prone in field isolates (30). The five C-terminal residues in phage 6-2, the three N-terminal residues in phage 6-1, and the C-terminal residue in phage 8-05 do not match the 111/92 sequence and likely represent a ligation or cloning artifact during library construction.

In PRRSV, the antigenicity of linear epitopes correlates with genetic variability. We and others have previously found, based on the analysis of two epitope sites, that linear epitopes in the envelope glycoproteins of PRRSV may exhibit a very high level of genetic variability (10, 14, 23, 30). In the present study, we were able to examine the variability-antigenicity relationship on a larger panel of epitope sequences and to extend the analysis to the replicase polyprotein. The amino acid homology between the Lelystad and 111/92 PRRSV strains was plotted against the maximal antibody titer at 42 dpi for each of the 10 ES (Fig. 6). Interestingly, in this plot two main clusters of ES were apparent: one cluster exhibited a low level of sequence conservation and induced higher levels of antibodies (cluster A, Fig. 6), and one cluster exhibited a high level of sequence conservation and induced lower levels of antibodies (cluster B, Fig. 6).

View larger version (18K): [in this window] [in a new window]   FIG. 6.   For linear ES in PRRSV, antigenicity in vivo correlates with genetic variability. The amino acid homology values (x axis) are from Fig. 2, and the maximal antibody titers (y axis) are from Table 3.

Antibody responses of short-term and longer-term viremic pigs are different. It is well known that pigs may exhibit various periods of viremia following PRRSV infection, but the mechanisms responsible are unknown. Since our pigs were experimentally infected by the nasopharyngeal route and the initial viremic peak was likely to be derived from acute PRRSV replication in alveolar pulmonary macrophages (16, 17), we examined whether local mucosal humoral responses might have influenced viremia levels. We estimated viremia duration by the number of viremia-positive samples among 11 serum samples taken at 3, 7, 10, 14, 17, 21, 24, 28, 31, 35, and 42 dpi (all pigs were viremia negative at 0 dpi). Two pigs had only three positive samples, the latest at 14 dpi, and remained viremia negative thereafter. One pig had four positive samples, the latest at 14 dpi, and remained viremia negative thereafter. One pig had four viremic samples, the latest at 21 dpi, and remained viremia negative therafter. The remaining two pigs had seven and eight viremic samples, respectively, the latest at 35 dpi, and were viremia negative at 42 dpi (summarized for the group as a whole in Fig. 4). Thus, our group of animals contained short-term as well as longer-term viremic pigs. For analysis of IgA responses, the ES were grouped as follows: ES11 was excluded from the analysis because seroconversion in IgA ELISA was not consistently seen against this weak epitope (Fig. 1 and not shown). Also, owing to large interanimal variability in which the ORF 1 ES were targeted by the IgA response (not shown in detail, but see Fig. 1B), the ORF 1 ES (ES1 to ES9) were analyzed collectively. Finally, ES12 was analyzed separately because, in comparison to the ORF 1 ES, ES12 appeared to be deficient in IgA induction relative to total Ig induction (Fig. 1, compare ES12 in A and B). Interestingly, viremia duration correlated with the kinetics of the IgA response in individual pigs: short-term viremic pigs also exhibited the quickest IgA response to the "virtual antigen" consisting of the combined ORF 1 epitopes (Fig. 7). For ES12, the kinetics of IgA production did not correlate with viremia duration (Fig. 7).

View larger version (15K): [in this window] [in a new window]   FIG. 7.   Short-term and longer-term viremic pigs exhibit differences in humoral immune responses. Serum samples taken from six experimentally infected pigs at 0, 14, 21, 31, and 42 dpi were examined in IgA ELISA. The time of seroconversion (the earliest time when ELISA reactivity exceeded the cutoff determined by examining a negative serum panel, as described in Materials and Methods) was plotted against the duration of viremia for ES12 as well as for the ORF 1 ES (ES1 to ES9) collectively. Each data point represents a single animal. All six experimentally infected pigs, irrespective of viremia duration, were positive for PRRSV by RT-PCR on lung or tonsil material at euthanasia at 42 to 56 dpi.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, we describe an approach based on long RT-PCR and phage display for the screening of whole viral genomes for linear B-cell epitopes. We show that, given an effective library selection strategy and an appropriate monitoring of enrichment ratios (Table 2), sequencing of phage clones is more than 80% accurate in identifying B-cell epitopes. Sequencing phages is technically simple and well suited to high-throughput formats. In contrast, downstream serological confirmation of epitopes is more labor intensive and requires larger panels of characterized sera. Thus, although the phage display method identifies only linear epitopes and may exhibit residue bias in the displayed peptides (32), the relatively high degree of accuracy in epitope identification based on sequencing alone makes it attractive for rapid first-pass epitope screening of viral genomes using minimal amounts of immune sera and pathogen material. The phage display method identified 10 ES in total, 8 of which were in the replicase polyprotein. The results from the serological characterization of the 10 ES appeared to have implications for the diagnostics as well as immunopathology of PRRSV infection, as discussed below.

Underestimate of the total number of linear ES in the structural genes. Although seven ES were identified in ORF 1a, none were identified in ORF 1b (Fig. 3). In ORF 1a, six unique phage clones sometimes clustered at an ES (Fig. 2), indicating that the phage display library provided at least a sixfold coverage of ORF 1. Also, by using appropriate ligands, we have isolated phage displaying ORF 1b functional domains from the ORF 1 library (unpublished data). Thus, we believe that the failure to identify linear ES in ORF 1b was not due to poor library quality. The paucity of ES in ORF 1b may reflect a low level of protein production and hence poor stimulation of porcine B cells: translation of ORF 1b is thought to require a ribosomal frameshift event at a "slippery knot" RNA feature just upstream of the ORF 1a stop codon. The frameshift efficiency has been estimated at 15 to 20% in vitro (8), suggesting that significantly more ORF 1a protein than ORF 1ab protein may be produced in vivo. Additionally, the lack of linear ES in ORF 1b may reflect the antigenicity-conservation relationship for short linear sequences described in this study (Fig. 6). ORF 1b is generally more conserved than ORF 1a and contains several very highly conserved functional domains (22, 25). Interestingly, the Nsp2 fragment found to contain most linear ES (Fig. 3) has been noted by others to be the most variable part of the PRRSV genome, not excluding the structural genes (6, 34). In short, we believe that the absence of linear ES from ORF 1b and the clustering of linear ES in Nsp2 represent meaningful rather than artifactual results (Fig. 3). The biological mechanisms behind the clustering of ES in Nsp2 are unknown, but the protein is produced in quantity in Arterivirus-infected cells, where it carries out several functions, such as participating in the induction of double lipid membrane vesicles, as well as acting as a membrane anchor for the assembly of multiprotein viral replication complexes on the surface of the vesicles (31, 41).

Antigenicity-variability relationship for linear ES in PRRSV. We observed that the ES exhibited an unexpectedly marked clustering in antigenicity-variability plots (Fig. 6). Only one of the 10 ES did not exhibit clustering (Fig. 6, ES2). The very strong ES12 localized to cluster A (Fig. 6). ES12 induced very high antibody levels in vivo (Table 3), is known to mediate neutralization in vitro (23), and is hypervariable in field PRRSV isolates (10, 14, 23, 30). Thus, although our results suggested that the protective effect of anti-ES12 antibodies may be limited in vivo (see results above and Fig. 4), it nevertheless seems likely that the porcine antibody response exerts a selective pressure, which results in the sequence diversity at the ES12 site, as has also been suggested by others (10, 14, 23, 30). The genetic mechanisms regulating ES12 sequence diversity may be quite complex: the stretch of the viral genome which encodes ES12 in ORF 4 also codes for an RKASLSTS epitope in the overlapping ORF 3 (Fig. 3) (30). Nucleotide changes in the first codon position in ORF 4 result in third-codon position changes in the overlapping ORF 3. This suggests the presence of a selective pressure amplifying and quenching mechanism, where radical (first codon position) changes in ES12 are linked with more conserved (third codon position) changes in the RKASLSTS ORF 3 epitope. This mechanism, if it exists, is not essential; we have previously reported that whereas the genetic locus encoding ES12 and the ORF3 RKASLSTS epitope was intact in 70% of Danish field PRRSV isolates from 1992, it is deleted in 63% of present-day Danish field PRRSV isolates (30).

Implications of the identified ES for PRRSV diagnostics. ELISA using phage-displayed ES as the antigen exhibited characteristics which might be attractive for diagnostic tests. (i) The wide range in seroconversion times against the different ES in effect made seroconversion predictive of the occurrence of viremia in the group of animals (Fig. 4, compare serology and virus isolation data). This might be exploited to develop serological tests with improved ability to measure disease progression at the herd level. (ii) Since preliminary experiments indicate that the ES described in Fig. 2 are not recognized by antibodies from animals infected with US-type PRRSV (unpublished data), they may be applicable for highly discriminatory serodiagnostics in areas where US- and EU-type PRRSV coexist, such as Denmark (2, 18, 38) and Canada (9). (iii) M13 phage particles are resistant to proteolysis, allow very simple and reproducible coating of ELISA plates with peptide antigen, and combine multivalent display with a good presentation of peptides for antibody binding (46). All of these features may be attractive for detection of low-level antibodies in protease-containing sources such as boar semen or meat juice. In fact, ELISA using phage-displayed ES as the antigen can be used to detect antibodies in boar semen (M. B. Oleksiewicz, submitted for publication).

Implications of the identified ES for PRRSV immunopathology. For PRRSV, the demonstration of antibody-dependent enhancement of infection in macrophages in vitro has caused interest in the effect of humoral responses on viral replication in vivo (44). A recent study showed a correlation between the humoral immune response and quenching of PRRSV replication in the lungs of pigs (16). In our study, we observed a correlation between the kinetics of the IgA response towards ES in the replicase polyprotein and the duration of viremia (Fig. 7). Although the presence of neutralizing ES in the replicase polyprotein cannot be excluded, as mentioned above, it seems more likely at present to assume that the IgA response per se, rather the IgA response specifically against ORF 1 ES, might be of importance in shortening viremia duration. Also, the IgA levels at late times postinfection did not correlate with viremia duration (not shown). This indicates that the early IgA response may directly affect viral replication, as opposed to IgA being a marker for another, unknown protective immune parameter. Importantly, in all six experimentally infected pigs, PRRSV could be detected by RT-PCR on lung or tonsil material at euthanasia at 42 to 56 dpi (not shown). Thus, the IgA response did not correlate with the clearance of virus from solid tissues. It is also noteworthy that no correlation existed between the IgA response kinetics and viremia duration for ES12 (Fig. 7). While further examination of the correlation between virus replication and serological parameters is warranted, it appears that the use of "precise" antigen, such as the ES identified in the present study, may have inherent advantages in examining the interactions between PRRSV and the humoral immune system.