In this study, we investigated the roles of DENV E protein on the expression, stability and conformation of prM protein and how prM protein and C-terminal truncation of E protein affect the conformation of E protein. In the absence of E protein, prM protein did not express well. In the presence of E protein, the expression of prM protein increased in a dose-dependent manner. Pulse-chase experiment suggested EH2 is important for maintaining the stability of prM protein. Moreover, E protein affected the recognition of prM protein by anti-prM mAbs. To our knowledge, this is the first study reporting that DENV prM protein alone expresses poorly and a domain of E protein (EH2) can affect the stability and expression of prM protein, a chaperone of E protein. This is in contrast to what has been reported for TBEV prM protein, which was stable and expressed well by itself 
. Dot blot binding assay revealed that prM protein and C-terminal truncation of E protein affect the recognition of E protein by several mouse and human anti-E mAbs. These findings not only add to our understanding of the interaction between DENV prM and E proteins but also have implication for future design of subunit dengue vaccines.
A recent crystallographic study of a recombinant DENV2 prM/E protein complex revealed the heterodimeric interaction between the N-terminus of prM protein (residues 1 to 81 of pr peptide) and E protein ectodomain at high resolution 
. Our radioimmunoprecipitation and sucrose gradient sedimentation analysis of a series of C-terminally truncated prME constructs of DENV4 revealed that ET1 and EH2 rather than the ectodomain of E protein were critical for prM-E interaction (). Similar finding has been reported for TBEV by radioimmunoprecipitation analysis of C-terminally truncated prME constructs 
. These observations suggest that ET1 and EH2 probably contribute to prM-E interaction greater than the ectodomain of E protein. Cryo-EM study of DENV2 virions at high resolution revealed that the corresponding double membrane anchors of E protein (ET1, ET2) and M protein (MT1, MT2) were packed together and ET1 was next to MT1 
, it is conceivable that the interaction between ET1 and MT1 within the membrane contributed significantly to the prM-E interaction. In addition, the EH2 and MH are partially buried in the outer leaflet of membrane 
. It is possible that EH2 could interact directly with MH or other region of prM protein (such as residues 82 to 112) not depicted clearly by the crystal structure of recombinant prM/E protein complex 
In this study, we employed a dot blot binding assay using cell lysates (of different prME or E transfectants) prepared in 1% NP40 lysis buffer without SDS or boiling, a condition similar to RIPA lysis buffer containing non-ionic detergent, to preserve the conformation of membrane protein. In addition to the controls to exclude the possibility of overexposure (, column B of each membrane), we also prepared mixtures containing different amounts of native E protein (in 1% NP40 lysis buffer) and denatured E protein (in reducing buffer) to assess the recognition by different anti-E mAbs (, column A of each membrane). It is worth noting that the conformation of E protein examined in our dot blot binding assay was that of E protein in heterodimer with prM protein and might not be the same as that in the context of particles. Since most of the C-terminal truncation constructs (except prMEd470) did not form VLPs well (data not shown), a finding consistent with a previous report in TBEV 
, the binding of these C-terminal truncated E proteins in the context of particles was not examined.
summarizes the effect of prM protein and C-terminal E truncations on the recognition of E protein by mAbs based on dot blot binding assay and capture-ELISA. Truncation of EH2 or EH1 greatly affected the binding of E protein by two mouse GR mAbs (4G2 and DEN2-12), which recognized fusion loop residues of domain II 
, whereas truncation of EH2 but not EH1 greatly affected E protein binding by five human GR mAbs (DVD19.4, DVD19-13, DVD23.3, DVD23.4 and DVD26.11), which recognized similar fusion loop residues (). The reduced binding to prMEd421 and Ed421 suggests that EH2, buried in the outer leaflet of membrane, is important for proper folding or maintaining the conformation of E protein (prMEd421, Ed421) required for recognition by these mAbs. Notably, our previous study found that proline substitutions introduced to EH1 and EH2 did not affect the recognition by several mouse anti-E mAbs, suggesting differential effects of substitution and truncation 
. The reduced binding to prMEd395 by mouse GR mAbs but not by human GR mAbs suggests that mouse mAbs is sensitive to different conformations of Ed395 in the presence or absence of prM protein. Alternatively, prM protein may interfere with the binding of mouse anti-E mAbs to prMEd395 through steric hindrance. In the absence of prM protein the binding of E protein (except Ed395) by four anti-E TS mAbs was reduced, suggesting that prM protein is important for maintaining the conformation of E protein (except Ed395) required for recognition by these anti-E mAbs.
Schematic drawing of prM/E proteins after synthesis and summary of the effect of C-terminal E domains on the recognition of E protein by mAbs.
Interestingly, the ectodomain of E protein alone (Ed395) could be recognized well by all the anti-E mAbs tested, suggesting that the ectodomain of E protein alone can fold well by itself and preserve the conformation and epitopes recognized by different anti-E mAbs. Whether Fd395 can also be recognized by other newly discovered anti-E mAbs remains to be tested; our findings suggest that Ed395 among different C-terminal truncated E proteins is a potential subunit immunogen mimicking the native conformation of E protein. Consistent with this interpretation, a previous study comparing the immunogenicity of truncated and full-length E protein with or without prM protein (D1E80, D1ME80, D1ME92 and D1ME100) by DNA vaccines in mice revealed that only D1ME100 (corresponding to our WT prME) and D1E80 (corresponding to our Ed395) induced neutralizing antibodies 
. Another study used a series of C-terminally truncated DENV4 E proteins derived from recombinant vaccinia virus to immunize mice and reported that 79%-RKG construct (corresponding to truncation at residue 394) was most immunogenic, whereas 81% and 100% constructs (corresponding to truncation at residues 399 and 436, respectively) induced very low or no antibodies 
. Similarly, recombinant WNV E protein with truncation at residue 406 (corresponding to our DENV Ed395 construct) can induce neutralizing antibodies and showed protection in mice and hamsters 
. Other studies expressing recombinant DENV E protein with truncation at residue 421, 424, 442 or 437 (in DNA vaccine) showed only partial protection in mice or monkeys 
. It should be noted that a previous study reported that inactivated whole TBEV virus and DNA vaccine encoding prM/E proteins induced higher neutralizing antibodies and better protection in mice than DNA vaccine encoding truncated E or truncated prM/E proteins (corresponding to our DENV Ed395 or prMEd395) 
. A recent study of DENV suggested the importantce of mAbs that recognize quaternary structure 
. Therefore, further studies to compare the potency of Ed395 and inactivated whole DENV as vaccines are warranted.
While several tetravalent live-attenuated candidate DENV vaccines have moved to Phase II or III clinical trials, a major challenge is the difficulties in achieving balanced neutralizing antibodies against all four serotypes due to dominant viremia by one or two serotypes resulting from inter-serotype interference and the risk of ADE mediated by cross-reactive non-neutralizing antibodies 
. Several subunit vaccines including different recombinant proteins and DNA vaccines are under development to avoid viral interference and/or better present the neutralizing epitopes 
. Previously, cross-reactive non-neutralizing or poorly neutralizing anti-E antibodies were thought to be the major player of ADE; recent studies revealed that anti-prM mAbs did not neutralize DENV well and potently promote infectivity by ADE 
. These studies suggest that anti-prM responses should be minimized in future dengue vaccines. Since prM protein was reported as a chaperone for proper folding of E protein 
, how to design subunit vaccines presenting E protein in its native conformation in the absence of prM protein is critical. Because secreted E protein is preferred for subunit vaccine preparation, various C-terminal truncations of E protein to remove transmembrane anchor have been designed 
. Based on the analysis of binding of mAbs to a series of C-terminally truncated E proteins in the presence or absence of prM protein, our findings that the ectodomain of E protein alone (Ed395) can be recognized well by all the anti-E mAbs tested suggest it could be a potential subunit immunogen that preserves the conformation of E protein without inducing anti-prM response.