Diagnosis of Human Caliciviruses by Use of Enzyme Immunoassays

Abstract

The application of molecular technologies, such as the expression of viral proteins in baculovirus, has provided a powerful approach to the diagnosis of human calicivirus (HuCV) infections. The baculovirus-expressed HuCV capsid protein self-assembles into virus-like particles, providing excellent reagents for immunologic assays, such as enzyme immunoassays (EIAs). Following the expression of the capsid protein of Norwalk virus, the capsid proteins of 8 other HuCV strains have been expressed in baculovirus. The unlimited supply of baculovirus-produced reagents for HuCVs allows these EIAs to be applied in large-scale clinical and epidemiological studies. Both the antigen and antibody-detection EIAs are highly sensitive. The antigen-detection EIAs are highly specific, but the antibody-detection EIAs are more broadly reactive. This article reviews baculovirus expression techniques used to produce HuCV capsid antigens, development of EIAs using these antigens, and application of these EIAs in studies of HuCV infection and illness.

Twenty-seven years have passed since Kapikian and co-workers [1] discovered Norwalk virus (NV), the prototype human calicivirus (HuCV). A review of the history of HuCV research highlights the important role that the development of diagnostic methods played in the research. The application of immune electron microscopy (IEM) resulted in the discovery of NV [1]; however, HuCVs could not be cultivated in cell culture nor passed in animal models. Therefore, a number of techniques based on reagents obtained from patients or experimentally infected volunteers were developed, including radioimmune assays (RIAs) [2–4], blocking RIAs [5], ELISA [6–8], and immune adherence hemagglutination assays [2].

A modified IEM technique, called solid-phase IEM [9, 10], also was developed. The IEM utilized antibody to coat an EM grid to capture virus. One monoclonal antibody was generated for a single prototype HuCV strain, and an EIA based on this monoclonal antibody was developed but not used widely [8].

A Western blot technique to detect HuCV proteins from fecal specimens was developed to detect viral antigens and antibodies [11]. These techniques greatly advanced our knowledge of HuCVs and associated illness; however, because of the limited supply of reagents from humans, these techniques were used only in a few research laboratories. Furthermore, it was difficult to determine the infection history of volunteers who were administered HuCVs before providing serum and stools as reagents, making the specificity of assays using these reagents uncertain. Studies of HuCVs suffered from the limitations of these diagnostic methods.

The cloning of the prototype NV [12–14] and many related strains [15–21] allowed the development of new diagnostic assays. Determination of the nucleic acid sequence of different HuCV strains allowed the design of primers to detect viral RNA by reverse transcription-polymerase chain reaction (RT-PCR) [22–26]. Sequence analysis of RT-PCR products of prototype and newly recognized strains allowed the genetic characterization and classification of HuCVs and design of second-generation primers for detection and classification of more strains [27–29].

Molecular engineering techniques also permitted the development of diagnostic assays based upon expression of caliciviral capsid antigens [30–39]. The baculovirus-expressed HuCV capsid proteins self-assemble into virus-like particles (VLPs) that are similar morphologically and immunologically to authentic virions [30], providing excellent reagents for use in developing immunologic methods.

Herein, we briefly review the recent development of the baculovirus expression techniques to produce HuCV capsid proteins and their use in EIAs for diagnosis of HuCVs. The application of these EIAs in epidemiologic studies of HuCVs, possible problems, and future directions also are discussed.

Baculovirus-Expressed HuCV Capsid Proteins and Related EIAs

Production of HuCV capsid proteins by using baculovirus expression systems

The first HuCV capsid protein expressed in baculovirus was the NV capsid [30]. Like other HuCVs, the NV genome contains a single capsid gene encoding a protein of ∼60 kDa [13]. A cDNA from the 3′ end of the genome, containing the viral capsid gene and ORF3, was used to construct a baculovirus recombinant. Recombinant virus containing only the capsid gene also resulted in expression of the protein. The viral capsid protein was initiated from its own AUG initiation codon, and a nonfusion protein of the capsid gene was expressed under the control of the baculovirus polyhedrin promoter.

After construction of the baculovirus recombinant, expression of the protein was accomplished by infection of insect cells (Spodoptera frugiperda, Sf9) with the recombinant. The baculovirus-expressed NV capsid protein was detected as a single band following electrophoresis in an SDS polyacrylamide gel, although a minor small peptide was observed in the insect culture after longer incubation. The small minor peptide is a cleavage product from the NV capsid protein. The intact capsid protein produced in the insect cell cultures self-assembled into VLPs that were morphologically and antigenically similar to authentic virions [30, 40, 41]. The VLPs appeared empty and did not contain nucleic acid when observed by negative-stain electron microscopy. Three-dimensional structures determined by cryo-electron microscopy showed that the NV VLP contains 90 dimers of the capsid protein that form a T = 3 icosahedral capsid morphology that is unique among human viral pathogens [42, 43].

The protein yield in insect cells was high—up to 12.5 mg of protein/100 mL of culture medium [30]. Most VLPs were released into the culture media and easily purified. Because VLPs are particulate, they can be separated from most cellular and baculovirus proteins by rate-zonal centrifugation in a sucrose gradient. VLPs recovered from sucrose gradients usually are of high purity and can be used directly as test antigen or for immunization. Sucrose gradient-purified VLPs can be further purified by equilibrium density gradient centrifugation in CsCl solution [30]. VLPs also can be purified by CsCl gradient followed by sucrose gradient to separate soluble proteins. The baculovirus-expressed HuCV capsids maintain their structure at 4°C and after freezing; however, the particles degrade over time at 4°C if they are not highly purified, potentially due to contamination with cellular or baculovirus proteins, especially proteases. Addition of protease inhibitors during purification may avoid such degradation. The recombinant Mexico virus (rMxV) capsid is more stable when stored at ⩽20°C. NV VLPs also maintain their structure and antigenicity following lyophilization, so they can be used for long-term storage and for shipment under ambient conditions.

Following the expression of NV, the capsid proteins from 8 other HuCVs, including 3 “typical caliciviruses” in the provisionally named “Sapporo-like virus” (SLV) genus, were produced using the baculovirus expression system (table 1). Yields of these proteins were generally high and the products, except for that of Houston 1986 strain (Hou/86), also self-assembled into VLPs. In a recent study of Hou/86 and Hou/90 expression, it was found that genomic sequence upstream from the capsid gene plays a role in the optimal transcription or translation of the capsid gene [45]. The mechanism by which this sequence affects function remains unknown. In summary, methods for purification, storage, and yields of VLP may vary among different HuCV strains.

Table 1

Currently available baculovirus-expressed human calicivirus capsid antigens and related assays.

According to the sequences of the capsid genes, these newly expressed strains belonged to 6 statistically distinct genetic clusters within the 2 genera of HuCVs (table 1). Within the provisionally named “Norwalk-like virus” (NLV) genus, MxV and Toronto virus belong to the same cluster, and Lordsdale virus (LV) and Grimsby virus (GrV) belong to another cluster [18, 35, 37, 38]. These clusters have been named genogroups I and II for convenience. Similarly, Sapporo 1982 strain (Sapporo/82) and Hou/86 are in the same cluster of SLVs [39, 45]. Other clusters within these genera now are known.

When tested by antigen-detection EIA, the recombinant NV (rNV) capsid was found to be antigenically distinct from the rMxV capsid after the expression of the two proteins [30, 33]. The recombinant GrV (rGrV) capsid also was shown to be distinct from rNV and rMxV after the establishment of the rGrV antigen EIA [37]. On the other hand, the rMxV capsid was found to share minor antigenic epitopes with Hawaii virus (HV) when stool specimens from volunteers infected with HV were tested with the rMxV antigen EIA [31, 32]. However, little cross-reactivity between rMxV and rHV was found when a recombinant antigen EIA for HV was developed (Jiang X, unpublished data). The rMxV antigen EIA detected a number of Mexico-like viruses [31, 32, 41], which are believed to include the Toronto virus. On the basis of predicted amino acid sequence similarity, LV and GrV may be antigenically related, although direct cross-testing of the two antigens is necessary to confirm this prediction. In addition, Sapporo/82 and Hou/86 share antigenic epitopes [45, 46]. HuCVs are genetically diverse, and the currently available recombinant HuCV capsid proteins do not cover all the genetic clusters of viruses within the 2 HuCV genera. In addition, new strains with distinct genetic identities are continually being discovered. A continued search for new strains of HuCVs and development of diagnostic assays for these new strains are necessary.

Development of antigen-detection EIAs

The baculovirus-expressed recombinant HuCV capsid antigens have been used to raise hyperimmune antibodies in laboratory animals [31, 47]. The antigens used as immunogen were purified by centrifugation on sucrose or CsCl gradients (or both) and were of high quality. High-titered antibodies against the antigens usually were obtained. To develop a sandwich EIA to detect viral antigen in stools, hyperimmune antisera from 2 species of animals usually are used (one for coating the plate and one as a detector), although similar tests can be developed from a single antibody. To ensure specificity of the assay, duplicates of each stool specimen should be tested in wells coated with pre- and postimmunization antisera, respectively. Such testing allows determination of a positive result based on a significant positive/negative ratio (e.g., ⩾2) between the optical density (OD) values in the two wells. This is important because some stool specimens give false-positive results when only hyperimmune serum is used to coat a single well. The cutoff point of each assay needs to be determined by testing a panel of stool specimens containing HuCVs, other enteric pathogens, or no pathogens [31, 47].

Following the development of rNV and rMxV antigen EIAs, an antigen EIA using hyperimmune sera against rGrV capsid antigen was developed [37]. We recently used the same rGrV capsid antigen to raise a hyperimmune antiserum in mice and developed an antigen EIA using a mouse antiserum (Jiang X, unpublished data) together with a rabbit hyperimmune antiserum raised earlier [37]. In addition, we developed an antigen EIA for the prototype HV, using hyperimmune antisera against the baculovirus-expressed rHV capsid antigen [36] (Jiang X, unpublished data). The rGrV and rHV antigen EIAs together with the rNV and rMxV antigen EIAs have been used successfully in our laboratory in surveys of HuCV infection, and some of the results are described later in this article. Finally, monoclonal antibodies have been generated for a few recombinant capsid antigens. An antigen EIA using these monoclonal antibodies has been described [48].

Development of antibody-detection EIAs

The first method used for detecting antibody in serum specimens was developed by using baculovirus-expressed HuCV capsid proteins directly to coat wells in a microtiter plate [30]. Because the VLPs used for the coating were highly purified, high assay sensitivity and specificity were obtained. This format initially was developed to measure total immunoglobulin in human sera. The same format later was adapted for detection of immunoglobulin isotypes (IgA, IgG, and IgM) [49–52] by changing the conjugates to detect each specific isotype. The assays for IgG or total immunoglobulins have been widely used in serosurveys of HuCV infection in many countries [18, 53–58]. The IgM EIA has been used to detect recent infection in a single serum for many infectious pathogens. The kinetics of serum IgM of patients following NV infection have been described [59], but the kinetics for IgA and IgG are lacking. The NV IgA EIA also has been used to detect secretory IgA in fecal specimens [60].

A capture-format IgM EIA was developed for rNV and rMxV [50, 51, 59, 61]. In this format, an anti—human IgM antiserum was used to capture the IgM molecules from human sera. HuCV-specific IgM then was detected by addition of recombinant capsid antigens, followed by detection with monoclonal or hyperimmune antibodies against the recombinant capsid antigens. This capture IgM EIA was designed to increase test sensitivity by removing excess IgG in serum samples. The monoclonal antibody-based IgM EIAs have been reported to be more sensitive than the direct antigen-coating IgM EIAs [59]. The IgM detection EIAs may distinguish the “type” of infecting viruses better than other assays. The direct antigen-coating method is simple and less time consuming, making it more suitable for large-scale studies.

Application of the Antigen Detection EIAs in Surveillance and Outbreak Investigations of Gastroenteritis

The sandwich-format, antigen-detection EIAs have been used successfully for the detection of homologous strains of HuCVs (i.e., for the detection of NV by the rNV EIA in volunteer studies) [47, 60]. The rNV antigen—detection EIA was as sensitive as dot-blot hybridization and RT-PCR and more sensitive than RIA [47]. The finding of large amounts of soluble NV antigens in stool specimens might explain the high sensitivity of the EIA. The high sensitivity of the recombinant EIAs allowed for prolonged detection (up to 2 weeks following challenge [60]) of excreted NV antigen in the stools of volunteers; however, when the assays were applied in surveys of HuCV infection in field specimens, unexpectedly low detection rates were obtained (table 2).

Table 2

Detection of human caliciviruses in stool specimens from sporadic cases of gastroenteritis by use of antigen-detection enzyme immunoassays (EIAs) using hyperimmune antisera against baculovirus-expressed HuCV capsid antigens.

The rNV and rMxV EIAs have failed to detect HuCV infections in several populations. The highest detection rates were 2.1% for rNV and 4.3% for rMxV, and detection averaged 0.3% for rNV and 0.9% for rMxVs in sporadic cases of diarrhea in children, which was lower than rates observed in the 1970s using EM. Using the rNV and rMxV antigen detection EIAs, we also obtained low detection rates (table 2) in our study of HuCV infection in a birth cohort of children in Mexico [69–71]. Samples from this population should have included stool specimens from children experiencing their first HuCV infection. In a total of 621 diarrhea stool specimens from children <2 years of age, only 4 (0.6%) were positive by the rMxV antigen EIA, and none was positive by the rNV antigen EIA. We also tested the stool collection with the rHV antigen EIA and detected no positive specimens.

The antigen-detection EIAs are highly specific, HuCVs are genetically diverse, and multiple strains can co-circulate; therefore, the antigenic types of the strains used in the assay and the strains circulating in a population may differ enough to affect detection rates. To understand this better, a subset of 115 diarrhea stool specimens from Mexican children was tested for HuCVs by RT-PCR using a primer pair broadly reactive for NLVs and SLVs [29]. Twenty-two (19%) of the 115 diarrhea stool specimens were positive by this RT-PCR, and 15 of the 22 PCR products were cloned and sequenced. Two strains, which were detected 1 month apart in each of 2 children, shared identical sequence with each other. The other 13 strains had unique sequences (table 3).

Table 3

Genetic variation of human calciviruses characterized by reverse transcription-polymerase chain reaction and sequencing of virus in 115 diarrheal specimens from a cohort of Mexican children followed between 1989 and 1991.

Comparison of these sequences showed that 9 (60%) of the 15 strains were NLVs; these 9 strains belonged to genogroup II, and the other 6 (40%) strains were SLVs. NLVs could be further divided into 3 clusters represented by the LV, MxV, and HV prototypes, and SLVs could be divided into 3 clusters represented by Sapporo/82, London 1992 strain (Lon/92), and an as-yet-unpublished strain (table 3). Only 1 of the 15 strains was positive in the rMxV antigen EIA, and that strain shared 195% nucleotide (nt) identity in the RNA polymerase region with the prototype MxV. No strain was positive in the rNV and rHV antigen EIAs, including 1 strain that shared 91% nt identity with the prototype HV in the RNA polymerase region. This result confirmed previous observations that the recombinant antigen EIAs detect strains that are genetically closely related to the test antigen. This situation has become more complicated because of the occurrence of a naturally occurring HuCV recombinant, in which the capsid and RNA polymerase genes were found to be derived from parent strains with distinct genetic identity [72]. Naturally occurring recombination also was suggested by comparison of the RNA polymerase and capsid genes among different strains of caliciviruses when partial sequences of the Snow Mountain virus genome were determined [73]. The identity of the RNA polymerase gene of such strains will not predict their capsid antigenicity.

Similarly low detection rates were found using rNV and rMxV antigen EIAs in outbreak investigations of acute gastroenteritis in adults in some studies, and high detection rates were found by RT-PCR in other studies (table 4). When the rNV, rMxV, and rHV antigen EIAs were used in a 3-year survey of outbreaks of acute gastroenteritis in Virginia, similarly low detection rates were obtained by antigen EIAs and high detection rates were obtained by RT-PCR. In a total of 216 stool specimens from 25 outbreaks in Virginia from 1996 to 1998, 10 stool specimens from three outbreaks were positive by the rMxV antigen EIA. Three diarrhea stool specimens from two outbreaks were positive for the rNV antigen EIA. None of the 216 stool specimens was positive by the rHV antigen EIA. However, when the stool specimens were tested with our recently developed GrV antigen—specific EIA (Jiang X, unpublished data), samples from 6 of the 25 outbreaks were positive. Samples from 11 (44%) of the 25 outbreaks were positive using the three EIAs (rNV, rMxV, rGrV). Most of the specimens positive by the rMxV and the rGrV antigen EIAs were high reactors (OD > 0.2), and the 3 rNV antigen EIA—positive specimens were low reactors (OD of 0.1–0.2). When the outbreak stool specimens were tested for HuCVs by RT-PCR, samples from 22 (88%) of the 25 outbreaks were positive using a combination of five primer pairs, including a recently designed primer pair that detects both NLVs and SLVs [29].

Table 4

Detection of human calciviruses from outbreaks of acute gastroenteritis in different populations by enzyme immunoassays (EIAs) and reverse transcription-polymerase chain reaction (RT-PCR).

RT-PCR products from 21 of the 22 outbreaks were cloned and sequenced. Two outbreaks yielded samples with a mixture of two sequences (Arlington, VA, 1997; Page County, VA, 1996), and samples from the other 19 outbreaks had unique sequences. Therefore, a total of 23 unique sequences were identified in the 21 outbreaks. As determined by sequence analysis, these strains all belonged to the NLV genus, in which 5 (22%) belonged to genogroup I and the other 18 (78%) belonged to genogroup II (table 5). The 5 genogroup I strains shared 75%–77% nt identities with the prototype NV; 1 of the 5 strains was close to the Southampton virus and the other 4 were distantly related to other strains in the genogroup (data not shown). In genogroup II, 10 strains belonged to the LV cluster, 4 to the MxV cluster, 1 to the HV cluster, and 3 to a new cluster.

Table 5

Genetic variation of 23 outbreak strains of Norwalk-like viruses from 21 outbreaks of gastroenteritis in Virginia (1996–1998) and their detection by recombinant antigen enzyme immunoassays (EIAs).

Stool specimens from 7 of the 10 Lordsdale-like virus outbreaks were positive by the rGrV antigen EIA, and the 6 involved strains had 89%–92% nt identity with the prototype LV in the RNA polymerase region. The 3 rGrV antigen EIA-negative strains had nt sequence identities of 94%, 92%, and 90%, respectively, with LV in the RNA polymerase region. This suggested that divergence of classification based on RNA polymerase and capsid sequences may also occur within the LV cluster, a possibility that should be confirmed by sequence analysis of the capsid region. Two of the 3 rMxV antigen—positive strains had >96% nt identity and the other strain had 73% nt identity with MxV in the RNA polymerase region; this strain is likely to be a recombinant virus [72].

One outbreak with mixed infection (Arlington, VA, 1997) contained a Mexico-like virus (98% nt with MxV), but the stool specimens from the outbreak were not positive using the rMxV EIA. Two strains weakly reactive in the rNV antigen EIA shared 76% and 77% nt identity with the prototype NV in the RNA polymerase region. Strains weakly reactive in the rMxV antigen EIA also were found in stool specimens of patients from outbreaks [77] and volunteers infected with HV and Snow Mountain virus [31]. The basis for this cross-reactivity remains unclear. Further work is required to understand completely the genetic and antigenic relatedness of HuCV strains.

Application of the Antibody-Detection EIAs in Serosurveys and Outbreak Investigation of Gastroenteritis

Different assay formats for antibody-detection EIAs have been used to assess HuCV infection and immunity in different populations. Direct coating of antigen has been used most widely to study the seroprevalence of HuCV infection in different populations in several countries [18, 53–57, 62–64, 66, 67, 78, 79]. When the rNV and the rMxV antibody EIAs were used, high seroprevalence against the two antigens was detected in both developing and developed countries (figure 1) [53–56, 64, 80, 81]. The age of acquisition of antibody to these 2 strains was older in developed than in developing countries, but all children were infected before reaching adulthood. The prevalence of antibody to rMxV was slightly higher than that to rNV. Seroprevalence to rNV and rMxV antigens was higher in developing than developed countries. One study in the United Kingdom tested serum samples against three (rNV, rMxV, rHV) antigens and found that seroprevalence to MxV was highest, seroprevalence to HV lower, and seroprevalence to NV lowest in that population (figure 1) [56].

Figure 1

Seroprevalence by age group (m = month; y = year) to human caliciviruses (HuCVs) in Japan, United Kingdom, China, and Kuwait as determined by recombinant (r) antibody—detection EIAs for Norwalk virus (rNV), Mexico virus (rMxV), and Hawaii virus (rHV). Percentage of individuals with antibodies against HuCV recombinant antigens in each age group was re-plotted on basis of published data from Hokkaido, Japan [54, 64], London [53, 56, 80], Beijing [81], and Kuwait [55]. For better comparison among populations, some age groups were deleted during re-plotting.

As noted above, the antibody EIAs are more broadly reactive. This means antibody to strains with a broad range of sequence identity will be detected. Therefore, the antibody detected by an EI A may not represent infection with the particular strain—it may represent instead infection with any of a group of strains with shared epitopes. Some recombinant antigens (e.g., rNV and rMxV) have been used in many serologic studies. Therefore, some study results are comparable; however, variation of results may be seen among studies done in different laboratories from different countries. To avoid variation, standard assay protocols and control reagents are needed.

The antibody-detection EIAs have been used to investigate outbreaks of acute gastroenteritis [49, 61, 67, 82, 83]. In general, the frequency and titer of antibody responses by patients correlate with the degree of genetic identity between the strains used in the assay and the strain causing the outbreak [49, 83]. For example, in the Virginia outbreaks, 100% of the patients involved in the three MxV-associated outbreaks had antibody responses detected in the rMxV antibody EIA (table 6).

Table 6

Antibody responses of patients involved in 12 outbreaks of acute gastroenteritis in Virginia and the genetic identities (RNA polymerase region) of the strains involved in these outbreaks..

The meaning of heterologous antibody responses is more difficult to interpret. Heterologous responses among genetically closely related strains are particularly difficult to differentiate because past exposure to different strains by the individuals involved in outbreaks is unknown. For example, in the Virginia outbreaks, six outbreaks were caused by Lordsdale-like virus strains, from which stools and paired acute and convalescent sera were collected. Stools from five of the six outbreaks were positive by the rGrV antigen EIA. While the antibody responses have not yet been determined using the rGrV antibody EIA, the antibody responses to rMxV and rHV of individuals involved in these outbreaks varied significantly, with 100% to 0% of individuals mounting a seroresponse (table 6). The strains detected in these outbreaks shared 71%–76% nt RNA polymerase identity with MxV and 80%–87% with HV, suggesting that they are types heterologous to MxV and HV. Therefore, development of recombinant antigens to detect “homologous” seroresponses to more antigenic types is a key need for future outbreak investigations using the antibody-detection EIAs. We predict that testing for seroresponses to rGrV would yield a high level of seroresponse in the third group of Virginia outbreaks caused by the Lordsdale-like viruses.

The antibody-detection EIAs have been used in other studies of infection and immunity and epidemiology of HuCVs. One study using these assays found differences in risk factors for NV and MxV infection in populations with different socioeconomic and geographic locations [84]. EIAs specific for detection of IgA and IgM have been used to describe infection and immunity in volunteers and in patients involved in outbreaks of acute gastroenteritis [51, 52, 61, 79]. These assays were more specific than the IgG detection EIAs and may be useful to define the type of a recent outbreak [51, 52, 61, 76]. With continued development and standardization of the methods, these assays will have more applications for use in clinical and epidemiologic studies of HuCVs.

Conclusions and Future Directions

In conclusion, both the antigen- and antibody-detection EIAs are highly sensitive. The antigen-detection EIAs also are highly specific and generally detect strains that are genetically closely related (>95% identity in the RNA polymerase region). This pattern has been complicated by recognition of a naturally occurring recombinant. The antibody-detection EIAs are more broadly reactive, in that seroresponses to a single antigen can be measured for strains that have genetic identities much more distantly related to the test strain (70% identity in the RNA polymerase region). Such heterologous seroresponses are frequently detected but usually have lower titers than homologous seroresponses. Because the exposure history of individuals to HuCV strains usually is unknown, diagnosis based upon heterologous antibody responses is difficult. The IgM- and IgA-detection EIAs may be more specific because these assays appear to measure recent infection and they may be formulated to distinguish infections between different virus types.

The sandwich format of antigen-detection EIAs using hyperimmune antibodies from laboratory animals is simple and useful for large-scale epidemiologic studies. Previous low detection rates of HuCVs by antigen EIAs were due to the high specificity of the assays and limited numbers of assays used. Therefore, a panel of hyperimmune antisera to multiple HuCV antigenic types will be needed before the assays are useful in clinical laboratories. To develop additional EIAs, further description of the genetic diversity of the HuCV family is necessary. This can be achieved by molecular epidemiology studies and continued searching for new strains. An EI A technology that has not been evaluated is assays utilizing monoclonal antibodies, which may be more broadly reactive than hyperimmune antisera.

The direct antigen-coating method to detect antibodies from human serum specimens is simple and useful for large-scale epidemiologic studies. More recombinant capsid antigens representing more HuCV antigenic types are needed for detecting homologous immune responses. With a better understanding of the antigenic relationships among different HuCV genetic types and host immune responses to these types, a reduced number of antibody EIAs may be required for testing a particular population and period. Isotype-specific assays are more specific but need more evaluation. With continued improvement of technology and the expression of more recombinant capsid proteins, the antibody EIAs should find more application in clinical and research laboratories.

Understanding the temporal and geographic distribution of predominant strains and the range of sequence variation will be important for the future development and application of EIAs for diagnosis and epidemiologic studies. It appears that the NLVs (genogroups I and II) play an important role in sporadic and outbreak cases of nonbacterial gastroenteritis in both children and adults, while the SLVs mainly infect children. At present, genogroup II NLVs appear to be more predominant than genogroup I NLVs, but both groups are circulating. The Lordsdale-like virus strains of genogroup II have been most predominant in almost all locations studied in the last several years. Therefore, development of methods to detect these predominant strains is most important. However, the predominant circulating strains may change. For example, in the United Kingdom, Mexico-like virus strains, which predominated during 1993/1994, have been replaced by Lordsdale-like virus strains. Selection of assays for clinical studies will depend upon the continued monitoring of changes of the prevailing types.

In summary, development of EIAs for HuCVs remains a big challenge. Further research needs to include continued surveillance to understand the genetic diversity and to search for new strains of HuCVs, and it must monitor changes among circulating strains in different populations and countries. In addition, research needs to include an understanding of the classification of HuCVs by genetic and antigenic criteria. This includes, when possible, characterization of strains previously characterized by the old methods, such as IEM, solid-phase IEM, and cross-challenge studies. Finally, international collaboration that includes exchange of information, reagents, and methods is needed. Proposals to establish a calicivirus sequence network and a central laboratory of shared recombinant calicivirus capsids were discussed at the International Workshop on Human Caliciviruses in Atlanta (March 1999). The establishment of such a network and a central laboratory would accelerate needed collaboration and, thus, these proposals should be pursued in an effort to extend our knowledge of HuCVs.