Makino et al. showed that fJAM-A is a receptor for FCV (28
). Here, we have confirmed and extended those results. We have shown that binding of FCV to fJAM-A requires the membrane-distal D1 domain; furthermore, mutations to three residues within D1 (D42, K43, and S97) decreased viral binding. The D1 domain of fJAM-A is predicted to contain two antiparallel β-sheets (strands ABED and GFCC′C"). Structural and biochemical analyses of human and murine JAM-A molecules revealed dimer formation through extensive interactions between the GFCC′ faces of two D1 domains (dimerization interface); at the center of the dimer interface are several charged residues critical in mediating JAM-A-JAM-A dimer formation by forming four salt bridges (dimerization motif) (19
). Residues D42 and K43 are predicted to reside in a short turn between β-strands A and B on the β-sheet opposite the dimerization face of hJAM-A, in close proximity to the short linker region between D1 and D2. Interestingly, these residues are also contained within a peptide that blocks the binding of MAb (F11) to human JAM-A (F11R) on the surfaces of platelets. This MAb induces platelet secretion and aggregation, and these stimulatory activities are inhibited by a peptide that contains residues 28 to 50 of human JAM-A (1
We hypothesize that the decreased binding observed in mutations of fJAM-A residues D42 and K43 is due to the importance of these residues in the FCV-fJAM-A interaction. However, the possibility exists that the mutations D42N and K43N introduce N-linked glycosylation sites in the fJAM-A receptor. Although the mutated fJAM-A constructs do not contain a classical Asn-X-Ser/Thr glycosylation motif at the position of the mutation, we cannot rule out atypical glycosylation. The presence of a glycan, and not the actual residue, could then be responsible for the decreased binding. Another possibility that we cannot completely exclude is that the lower levels of surface-expressed fJAM-A mutants D42N and K43N than of the wild-type fJAM-A (Fig. ) were responsible for the decreased virus binding. However, a distinct population of CHO-S cells expressed these mutants at levels comparable to those in cells expressing the wild-type receptor, yet these cells bound significantly less virus (Fig. ). Furthermore, mutating residue 97, which resides on a loop between β-strands E and F in very close proximity to residues 42 and 43, to an alanine resulted in significantly reduced viral binding, with expression levels similar to those of the wild-type fJAM-A. Thus, taken together, our data support the hypothesis that FCV binds to the D1 domain of fJAM-A in the vicinity of these residues.
Makino et al. reported that the FCV isolates they examined (which included the vaccine strain F9) could bind human 293T and monkey Vero cell lines and that anti-hJAM-A antibodies decreased that binding (28
). Confounding this observation, Stuart and Brown (48
) showed negligible binding of radiolabeled F9 virions to Vero and 293T cells. In agreement with the findings of Stuart and Brown, we were unable to detect an interaction between FCV and hJAM-A expressed on the surfaces of CHO cells. Furthermore, we did not detect an interaction between FCV and soluble recombinant hJAM-A ectodomain by ELISA or surface plasmon resonance studies (data not shown). Therefore, the binding detected by Makino et al. seems unlikely to have been mediated by hJAM-A.
The mammalian reovirus σ1 protein interacts with hJAM-A via residues E61 and K63 present at the dimer interface of D1, and it is the charge on these residues that is important for binding (19
). Mutations to just one of the residues of the dimerization motif prevented the formation of hJAM-A-hJAM-A dimers (19
). In contrast, FCV binding to cells expressing the E60R/K62E dimerization mutant of fJAM-A was similar to that of the wild-type fJAM-A (79% and 75%, respectively). The E60R/K62E mutations reverse the charges of two residues involved in forming the four predicted salt bridges important in dimerization (R58-E60, E60-R58, K62-E120, and E120-K62) (43
). Thus, we conclude that the charges of fJAM-A residues E60 and K62 (corresponding to hJAM-A E61 and K63) are not important in FCV binding. In addition, we predict that the E60R/K62E mutant is unable to form fJAM-A-fJAM-A dimers. Therefore, the observed binding of FCV to the dimerization mutant suggests that FCV can interact with both monomeric and dimeric forms of fJAM-A.
The FCV-fJAM-A receptor interaction is also different from the adenovirus-CAR interaction. The adenovirus attachment fiber protein is similar to the reovirus σ1 protein. The fiber protein is a trimer with a globular C-terminal knob that engages the CAR at its amino-terminal D1 domain (16
). Like the σ1 head domain, the fiber knob attaches at the dimer interface to residues important in CAR-CAR interactions at the dimer interface (7
Similar to caliciviruses, coxsackieviruses (Picornaviridae
) are small, nonenveloped, positive-strand RNA viruses. CVB utilize CAR as a host cell receptor (6
). CVB also engage the amino-terminal Ig-like domain D1 of CAR. However, unlike adenoviruses, the CVB interaction with CAR does not involve critical residues in the dimerization motif, although the binding site is on the distal end and the lateral side (A-G face) of CAR D1 and lies close to the dimer interface (20
). Residues D42, K43, and S97, which we found were important for FCV binding, are predicted to also be part of the orthologous lateral A-G face of fJAM-A; however, these residues lie much closer to the D1-D2 linker than the dimerization interface.
The ectodomain of fJAM-A was sufficient to mediate FCV binding, but when expressed on the cell surface with a GPI anchor, it could not mediate infection. In contrast, a GPI-anchored CAR mediated CVB and adenovirus binding and infection (50
). This finding suggests that the transmembrane and/or the cytosolic domain of fJAM-A is required for productive FCV infection of nonpermissive cells. We first hypothesized that signaling via the consensus C-terminal type II PDZ domain-binding motif at the carboxyl terminus of the cytoplasmic tail of fJAM-A might be required for infection; however, a construct that lacked the predicted PDZ binding motif (the last three amino acids of fJAM-A: Phe-Leu-Val) was as efficient as the wild-type receptor in mediating FCV infection of CHO cells (data not shown). Taken together, our data suggest a model in which FCV binding to the fJAM-A ectodomain transduces signals through its transmembrane and cytosolic domains that are required for infection. Evidence suggesting such a model for CVB was reported by Coyne and Bergelson, where interaction of CVB with the cellular receptor DAF activates kinases that trigger actin rearrangements and phosphorylation events necessary for mediating viral infection (10
Makino et al. (28
) reported that fJAM-A expression on nonpermissive hamster lung cells conferred susceptibility to FCV infection. Here, we show that expression of fJAM-A on the surfaces of five different cell lines was not necessarily sufficient to support productive infection for all FCV isolates. Not all of the cell lines that we investigated supported productive infection by all of the isolates; furthermore, the abilities of FCV isolates to productively infect each cell line tested varied. Only Vero cells were able to support productive infection by all of the FCV isolates we examined. While native nontransfected Vero cells have been reported to support infection by some FCV isolates (28
), none of the isolates we tested were capable of infecting Vero cells in the absence of fJAM-A.
The capacities of different FCV isolates to infect CHO-K1 cells that expressed fJAM-A varied. Indeed, we found that some isolates (FCV-131 and Kaos) were unable to productively infect these cells. Stuart and Brown reported that cell binding and infection by FCV are partially mediated by an N-linked glycoprotein containing α-2,6-linked, but not α-2,3-linked, sialic acids (48
). CHO cells lack the sialyltransferase necessary for generating α-2,6-sialic acid linkages and express predominantly α-2,3-linked sialic acid glycans (26
). Therefore, our results imply that at least a subset of FCV isolates are capable of binding and infecting fJAM-A-expressing cells independently of α-2,6-sialic acid linkages. fJAM-A contains a single putative N-linked glycosylation site at residue N184 in D2. The glycosylation status of this site, as well as the role of glycosylation in FCV binding and infection, warrants further investigation. It is also possible that there is an additional surface receptor for FCV. The findings that not all FCV isolates were capable of infecting nonpermissive cell lines expressing fJAM-A lends support to the idea that fJAM-A alone is not sufficient for FCV infection by all isolates and that additional coreceptors may be required.
In the last 10 years, there have been sporadic reports of highly virulent outbreaks of FCV disease in cats (11
). The virus isolates responsible for these outbreaks have been termed virulent systemic (VS) FCV (21
). The pathology associated with VS-FCV outbreaks indicates breakdown of the epithelial and endothelial barrier functions (41
). These observations may reflect the capacity of FCV to target fJAM-A, as it is known that JAM-A is required for maintenance of epithelial and endothelial tight junctions. It is possible that FCV can disrupt the homophilic interactions that occur between JAM-A molecules on apposing cells and thus disrupt tight junctions and barrier integrity, an effect proposed for secreted adenovirus fiber protein (49
JAM-A is also expressed on platelets, including feline platelets (T. Stokol and J. S. L. Parker, unpublished data). In some cases of VS-FCV, thrombocytopenia has been noted, together with disseminated intravascular coagulation (14
). It is possible that large amounts of FCV in the blood might disrupt normal platelet function. A monoclonal antibody (F11) that binds hJAM-A induces platelet secretion and aggregation. Binding of F11 to the surfaces of platelets is blocked by a peptide that incorporates the first 23 residues of hJAM-A D1 (residues 28 to 50) (1
). As we have shown, FCV likely interacts with residues D42 and K43 on fJAM-A. Therefore, it is possible that FCV binding to fJAM-A might induce secretion and aggregation of feline platelets and that this in turn might be responsible for some of the pathogenic sequelae seen with VS-FCV disease. We are currently examining these and other possibilities.