Methods

5.1.1 Cell infection experiments

Generally accepted criteria that unambiguously identify a cellular protein to function as a viral receptor are (i) the high-affinity interaction with the virus, (ii) the location on the surface of host cells, (iii) the inhibition of viral infection by specific receptor-blocking antibodies, (iv) the resistance of susceptible cells to infection by knock-down of this protein, and (v) the permissiveness of initially non-permissive cells after receptor cloning and transfection [63, 155, 197, 231]. In practise, however, identification of a viral receptor is hampered by the fact that (i) a clear inhibition is difficult to achieve since the virus may use several receptors or pathways in parallel, (ii) the receptor density on the cell surface may be low and the affinity of virus to its receptor limited, (iii) in most cases virus entry into host cells comprises several steps, and (iv) the validity of available methods is limited.

A cell based system designed for the study of virus-receptor-interactions is best suited to evaluate the potential in vivo susceptibility of cells as it mimics all steps in virus entry. In this context, cell binding or infection assays are the most commonly used approaches for studies that aim at the definitive identification of viral receptors. A broad diversity of protocols has been described by numerous authors. Other authors use isolated or recombinant cell surface proteins, which have been suggested by preceding studies, to investigate the direct interaction with the virus. However, difficulties in isolation and relative insolubility of many of the cellular surface proteins are problems that must be overcome for these kinds of experiments [228, 493].

The determination of virus titres or genome containing particles in the supernatant after several replication cycles is a common method used by many authors to measure differences in the binding behaviour and/or endocytosis mechanisms. However, the approach to draw conclusions from the numbers of virus particles in the supernatant for virus internalisation efficiencies seemed to be the most critical point of findings in this study (see also 4.4.1). The experimental approach of the replication set used in the infection assays was not able to distinguish (i) whether integrins are required for either binding in a post-attachment step, or as functional receptors for virus entry, or (ii) whether they induce virus internalisation, rather indirectly, by integrin-mediated signalling (5.3.2). Though the involvement of integrins in another stage of the virus replication cycle cannot be absolutely excluded, the surface locali-sation of these cell adhesion molecules allows the general assumption that integrins respond

to virus binding, and are involved in virus internalisation rather than in the downstream replication cycle [163, 482].

The approach applied in this study to evaluate the WNV binding characteristics, i.e. the binding set, represents only a rough approximation to the amount of specifically bound virus particles as washing may be insufficient to quantitatively remove all non-bound virions. Other methods may also be convenient and adequate in measuring the number of bound virus, such as flow cytometry [257, 467], but were not available for the study. However, since the treatment of all cells within each infection experiment was accomplished in exactly the same way differences in particle numbers can be regarded to reflect the specific experimental design. As a result, this approach yielded significant differences for the GAG experiments (4.4.2.2) and the MAV-1 studies (4.3.4.3). In the latter marked differences in the binding properties of MAV-1 to the MEF cell lines were seen.

Mouse fibroblasts isolated from 12.5 days old embryos were regarded to better imitate in vivo conditions than continuous, artificially immortalised cell lines or tumor cells, since they have undergone a limited number of passages only and are closer to primary cells that still keep the main functional components of the tissue from which they are derived [127, 428]. The use of knock-out cells is advantageous since deletions in the gene coding for a specific integrin subunit result in the complete absence of all corresponding heterogeneous integrin pairings on the surface (Figure 9 in 2.7.5). In contrast, application of blocking antibodies does not result in an absolute functional blockage of the integrin considered since even at high antibody concentrations not all integrins are targeted. Similarly, down-regulation of a particular integrin subunit by means of small interfering RNA does not suppress expression completely, and low expression levels can still be detected [99]. These two aspects complicate the estimation of a potential involvement of integrins in virus binding and uptake, and hamper final predictions about their quantitative contribution to virus entry. Another important advantage of the MF cell culture model used here is that the different integrin deficient and wild type MFs are comparable in their cell physiological characteristics, apart from the lack of the particular integrin subunit. This fact allows the direct comparison of results in order to assess the impact of a particular integrin subunit on virus infection for this cell type. To enhance comparability of results on how β1 and β3 integrin expression affects WNV replica-tion, deficient cell lines were directly compared with their corresponding integrin expressing cell lines, integrin β3-rescue MEFs or β1-floxed MKFs, which allowed excluding any background differences between cell lines that arise from possible differences between mouse strains. An additional motive for the establishment of this model was to get the option to

introduce a transgenic rescue of the absent subunit to find out whether the taxon affects cell susceptibility as has been described for other viruses, e.g. Foot-and-mouth disease virus (FMDV) [347]. This, however, was not part of this work but can be addressed by succeeding studies using this cell model.

The assessment of viral RNA content and the calculation of genome containing particle numbers was considered to be the best method to quantify virus yields, since determination of infectious virus particles may underestimate the true number of virus particles bound or produced [531], and to be the most efficient method viewed in terms of the very high sample numbers (see also 4.4.1). Determination of infectious virus particles by TCID₅₀ calculation was accomplished for the long-term study only and did not yield different results (4.4.3.1).

The discrepancy between infectious virus particles and relative numbers of RNA containing particles was notably high for the Uganda strain (4.4.3.1). Van der Schaar et al. reported a ratio of infectious to genome containing particles to be 1:2.600 for Dengue virus serotype 2 [531]. Similar findings were reported for the human immunodeficiency virus (HIV) type 1 and attributed to limited binding affinities of the virus in titration assays which led to a severe underrating of the actual concentration of virus particles [477]. On the other hand, determina-tion of genome containing particles itself may underestimate the physical virus particle numbers based on protein determinations due to incomplete viral genomes, insufficient RNA extraction or inefficient RT-PCR [531].

Plaque assay as an alternative method applied directly on the MF cell lines was tested in order to evaluate possible differences in their susceptibility. Readability and interpretation were of no relevance in the way that plaque sizes differed not only between the WNV strains but also between cell lines infected by the same strain (data not shown). The latter finding may be explained by an altered receptor usage on these cells [266].

An additional method was established using confocal laser scanning microscopy as a tool to visualise cell surface binding of virus particles and for use in future studies designed to investigate the characteristics of WNV entry into living cells [530, 531]. The use of the chimeric vaccine strain ChimeriVax-WN01 restricts comparisons with WNV specific traits to steps upstream the release of the nucleocapsid into the cytoplasm (2.5.3). Binding features were considered comparable to those of the New York strain since their envelope protein sequences were found to be homologue (4.1). DiD labelling of virus particles showed promis-ing results (4.4.3.7) but was sensitive to TritonX treatment which in turn was necessary for imaging of integrin-specific focal contacts. Hence this method was not helpful in this context.

No fluorophore labelled anti-E protein antibodies were commercially available and those

which were available for immuno-fluorescence application were generated in mice. This, however, was inadequate for the application of secondary mouse IgG antibodies as mouse fibroblasts were to be used here. Fluorescence conjugation of the primary anti-E antibody worked but showed high background staining of cells so that the results obtained are inter-preted in the context with other findings of this study. A co-localisation of a receptor candi-date would not be a definitive proof of an interaction with the virus since associated mole-cules may be involved or the interaction may be of different nature. However, detection of other than a direct interaction between virus and a specific cellular protein is beyond the limits of this method. Conglomeration of virus particles which allows visualisation has been described by van der Schaar et al. [531]. It is questionable, though, as to whether these virus complexes show the natural binding behaviour of single particles.

In summary it can be stated that there is no one method which fully satisfies all the require-ments in order to unambiguously identify a receptor and, at the same time, its precise function in virus entry. The consecutive application of a set of promising methods provides a set of results, based on different properties, which at best may point in the same direction. This can be taken as a proof that the protein in question is indeed a virus receptor. The methods applied in this study are expected to be reliable within their limits in answering the specific questions outlined earlier in Introduction.

5.1.2 Recombinant WNV envelope protein and domain III

Soluble forms of viral proteins involved in receptor binding are crucial not only in defining the role of such proteins, that means their capacity to interact with cellular surface proteins, but are also often used as an important tool for receptor identification. Recombinant Fla-vivirus envelope proteins have been used instead of infectious virus for binding experiments [89, 213], interaction force measurements [41, 269], and for co-immunoprecipitation assays [398]. The ability to compete with the virus for receptor binding in a dose-dependent manner constitutes the most relevant property of viral binding proteins [93, 95, 173]. The inhibition of virus entry may be achieved by two non-mutually exclusive mechanisms: preventing (i) the virus from interacting with cellular membranes by interference with the virus particle, and (ii) the interaction with the target surface protein and competition with the virus for binding [173].

In this study, WNV E protein, and domain III which constitutes the principal binding domain were recombinantly expressed since they had been proposed to mediate binding of WNV to

the integrin receptor [269]. Strong interacting forces between DIII and integrin αvβ3 were reported which led to the assumption that binding was highly specific (2.8). Application of increasing amounts of recombinant DIII inhibited virus binding to αvβ3-expressing cells but not to non-expressing cells [269]. Such findings are generally taken as a proof that the viral protein interacts with the receptor in question.

In contrast, first findings in this study, obtained from experiments with the mouse fibroblast cell lines showed that the putative receptor protein, integrin αvβ3, was not necessary for WNV infection of cells (4.4.3.1). It was concluded that alternative unidentified receptors exist which may interfere with recombinant proteins. Based on this result a convincing decision about the integrin’s involvement in WNV binding was impossible. Therefore, binding assays with recombinant proteins were abolished. Basically, the use of recombinant WNV proteins instead of infectious virus should be regarded with scepticism as some disadvantages exist:

(i) Only binding properties can be investigated whereas infectious virus organises all steps of entry. Later results of this study show that integrins do not mediate, in a first step, vi-rus binding to the cell surface (4.4.3.4.1 and 4.4.3.4.4). Thus, the use of recombinant protein alone may misguide conclusions concerning the potential role of receptor pro-teins in virus entry. On the other hand, Chu and Ng postulated that recombinant DIII ac-tivated signalling pathways by induction of FAK autophosphorylation via interaction with integrins [99]. This conclusion was rebutted by Scherbik et al. who demonstrated that FAK activation was not triggered by integrins [423]. Hence, the approach of path-way activation induced by recombinant WNV proteins must be regarded carefully.

(ii) Application of DIII can be insufficient to activate all steps that lead to virus internalisa-tion depending on the molecules and mechanisms involved. Several authors doubt its exclusive importance and suggest all three domains to be involved in the interaction of the virus and its receptor(s) (2.6.1).

(iii) In prokaryotic systems there is an uncertainty about the exact secondary structure of recombinantly expressed proteins compared to that under natural conditions, especially of posttranslational modified (e.g. glycosylated) proteins. The ability of a soluble pro-tein to imitate the virus binding properties can be influenced by its conformation. Thus, the proper folding of a recombinant protein is thought to be of great importance. Unfor-tunately, a definitive proof of a certain folding structure is difficult to achieve. As an example, Chu et al. argued that the induction of neutralising antibodies in immunised mice is a strong indication that recombinantly expressed WNV DIII adopted a biologi-cal relevant conformation [95]. In this study expression, purification and dialysis

proto-cols were basically the same as used by Chu et al. Therefore, native folding of DIII was assumed but could not be evidenced as no monoclonal antibodies directed against con-formational epitopes of E protein were commercially available.

These critical points make the utilisation of recombinant protein instead of infectious virus problematic when WNV binding behaviour is addressed. Hence, together with the reason outlined before, this approach was not further pursued. Once the knowledge regarding WNV host cell binding and the steps following attachment will have been improved the recombi-nantly expressed proteins of this work will possibly find application.

The E protein genes of the four WNV strains used in the infection experiments of this study were sequenced in order to check their identity and to interpret possible differences in binding and entry patterns (4.1). Multiple amino acid differences were found, some of them located within the binding domain DIII which is largely exposed on the lateral surface of the E protein (2.5.4.3). Distinct replication efficiencies in wild type and integrin deficient MFs (4.4.3.1), respectively, within a strain may, at least in parts, be based on differences in the sequence of the E protein. Provided that the established MEFs do not differ in their entire protein configuration, discrepancies in virus particle yields were, therefore, attributed to the lack of certain integrins. Most noticeably, there was a marked reduction in the replication efficiency in αv-deficient MEFs of the Dakar strain compared to the efficiency of other strains. Supposed that DIII alone mediates entry the question arises on how the differences seen in the replication of the New York and the Dakar strain come about since both are homologous in the DIII amino acid sequence. This again points to an additional participation of envelope protein domains I and II, or differential effectiveness in replication.