To analyze the role of ssRNA and other PAMPs in the response to influenza vaccines in detail, we immunized wild-type C57BL/6 mice, TLR7 knock-out mice, and MyD88/TRIF double knock-out mice with different vaccine formulations. MyD88 (myeloid differentiation factor 88) is an adaptor molecule which functions downstream of all known TLRs and IL1R family members with the exception of TLR3, which instead recruits a MyD88-related adapter molecule, TRIF (TIR domain-containing adaptor protein inducing interferon β) 
. Consequently, a deficiency of both MyD88 and TRIF excludes signalling by all TLRs. Mice were immunized intramuscularly with β-propiolactone-inactivated H5N1 (NIBRG-14) WIV, SV, or SU vaccine. Quantitative PCR using primers specific for segment 7 of the viral genome revealed that WIV contained per vaccine dose at least 5×108
copies of viral RNA, the natural ligand of TLR7. In SU or SV vaccine the amount of RNA was 500 and 5,000 times lower than in WIV, respectively. Four weeks after immunization, serum and spleen cells were collected for evaluation of humoral and cellular immune responses.
Serum HI titres in WIV-immunized TLR7−/−
mice and MyD88−/−
mice were found to be significantly lower than in WIV-immunized wild-type mice (; p
0.021 and p
0.001, respectively). Although sera from TLR7−/−
mice immunized with WIV showed a higher geometric mean titre (GMT) than sera from WIV-immunized MyD88−/−
mice, this difference was not significant (p
0.053). Most of the HI titres of SV- and SU-immunized wild-type mice were below detection level, precluding evaluation of the effect of the knock-out mutations on the HI responses to these vaccines.
TLRs contribute to the efficacy of H5N1 WIV vaccine.
Similar to the HI titres, virus neutralization (VN) titres of pooled serum samples from mice immunized with WIV were lower in the knock-out groups than in the wild-type group (). These results clearly show that TLR signalling is critically involved in the response to WIV immunization. Yet, in the knock-out groups, VN titres obtained after immunization with WIV were still modestly higher than those obtained after vaccination of wild-type mice with the other vaccines. This points to TLR-independent pathways contributing to the superior antibody response to WIV vaccine.
TLR-dependent and -independent mechanisms contribute to virus neutralization titres induced by WIV.
Serum titres of H5N1-specific IgG were determined by ELISA. In accordance with the HI and VN results, IgG titres were significantly decreased in WIV-immunized TLR7−/−
mice compared to wild-type mice (; p=
0.010 and p=
0.001, respectively). However, like the VN titres, the IgG titres in the WIV-immunized mutant mice were still significantly higher than those induced by SV (TLR7−/−
0.005) or SU (TLR7−/−
0.021) immunization again indicating involvement of TLR-independent pathways. The relative contributions of TLR-dependent and -independent mechanisms to the superior IgG response to WIV can be estimated by comparing the difference in geometric mean titre (GMT) between WIV-immunized wild-type and MyD88/TRIF-deficient mice with the difference between WIV-immunized wild-type mice and SV- or SU-immunized wild-type mice. Using this procedure the TLR-dependent contribution was calculated to be 73% and 83% for WIV versus SV and WIV vs SU, respectively (for calculation, see Text S1
). The IgG responses to SV and SU vaccine in both TLR7−/−
mice did not differ from those in wild-type mice, except for the IgG response to SU in TLR7−/−
mice, which was slightly but significantly decreased (p=
0.038; ). Together with the HI and VN results, these findings demonstrate that the superior antibody response to WIV is predominantly regulated by TLRs, TLR7 in particular, while TLRs do not seem to play a prominent role in SV and SU antibody responses.
We next investigated the role of TLRs in the Th1 polarization of the response characteristically found after WIV vaccination. We first assessed numbers of IFNγ- and IL4- producing T cells (Th1 and Th2 cells, respectively) in a cytokine-specific Elispot assay, after re-stimulation of spleen cells from immunized mice with H5N1 SU vaccine. Numbers of Th1 cells were significantly decreased in WIV-immunized knock-out mice compared to wild-type mice (p=
0.003 and p=
0.010 for TLR7−/−
mice, respectively), and matched those found in SV- and SU-immunized wild-type mice (). No difference was found between TLR7−/−
mice. Numbers of influenza-specific IL4-producing cells were extremely low in all animals for all vaccine formulations without significant differences between knock-out and wild-type mice (not shown). These data indicate that stimulation of TLR7 by ssRNA is the predominant determinant of the strong Th1-type cellular response induced by WIV.
Induction of IFNγ-producing T cells by H5N1 WIV vaccine depends on TLR7 signalling.
We further determined the subtype profiles of H5N1-specific serum IgG by ELISA (). IFNγ is known to stimulate production of IgG2a subtype antibodies by activated B cells, while IL4 stimulates IgG1 secretion 
. In C57BL/6 mice, however, the IgG2c subtype is produced instead of IgG2a 
. Hence, a predominance of IgG2c or IgG1 is indicative of a Th1- or Th2-type response, respectively. WIV immunization of TLR7−/−
mice as well as MyD88−/−
mice resulted in significantly reduced IgG2c levels as compared to wild-type mice (; p=
0.001 for both types of knock-out mice), supporting a role for TLR7 in Th1 polarization. IgG1 was increased in WIV-immunized TLR7−/−
0.050), adding to the preponderance towards a Th2-type response to WIV in these mice. The average of ratios of serum IgG2c and IgG1 concentrations (determined with appropriate IgG subtype protein standards) was 17.82 (SD 8.44) for the wild-type mice immunized with WIV, compared to 0.53 (SD 0.41) for TLR7−/−
mice immunized with WIV. SV and SU vaccines induced predominantly IgG1 and low levels of IgG2c, consistent with a Th2-type response (). For reasons unknown, SU vaccine induced lower IgG1 titres in both types of knock-out mice compared to the wild-type mice (TLR7−/−
0.014). Whether the presence of some residual RNA in SU vaccine might play a role remains to be shown.
H5N1 WIV vaccine induces Th1-type antibody responses via TLR7 signalling.
The response characteristics of the different H5N1 vaccines in wild-type mice were well in line with those previously found for other influenza subtypes 
. This consistency is supportive of a general mechanism underlying the differences in responses to WIV, SV and SU vaccine, which operates irrespective of the virus subtype used to vaccinate.
The above results demonstrate that TLR signalling plays an important role in the magnitude and Th1 skewing of the response to WIV influenza vaccines. Yet, in TLR-ko mice, WIV remained more immunogenic than SV and SU vaccines, inducing significantly higher titres of total IgG () and Th1-type antibody subtypes (IgG2b, IgG2c, IgG3; ; p
<0.05 for all comparisons). Thus, next to TLR-dependent mechanisms, a (minor) TLR-independent factor seems to contribute to the superior magnitude and Th1-skewing of the immune response to WIV. Type I interferons, including IFNα, have been shown to stimulate antibody responses and isotype switching to IgG2a when added to influenza subunit vaccine or other protein antigens 
, even without the need for additional TLR stimuli. We have previously shown for an H3N2 influenza virus strain that, unlike SU vaccine, WIV vaccine efficiently induced interferon α (IFNα) production in plasmacytoid dendritic cells (pDCs) in vitro 
. We therefore evaluated the induction of IFNα by the H5N1 influenza vaccine formulations used in this study and its TLR7 dependency in vitro
. In pDCs of wild-type mice cultured from bone marrow cells (, black bars) or enriched from splenocytes (, black bars) WIV but not SV or SU induced IFNα production. In bone marrow-derived pDCs from TLR7−/−
mice, IFNα production upon incubation with WIV was strongly decreased as compared to wild-type DCs (), confirming the results of others 
. However, spleen-derived pDCs from TLR7−/−
mice exposed to WIV produced similar amounts of IFNα as compared to pDCs from wt mice (). Thus, while in pDCs cultured from bone marrow induction of IFNα production by WIV is strictly dependent on TLR7, in pDCs enriched directly from spleen cells it is independent of TLR7. This implies that bone marrow pDCs and spleen pDCs are not completely identical. In line with this notion, bone marrow pDCs and spleen pDCs were earlier found to respond differently to HSV virus infection with respect to the TLR9 dependency of the IFNα response 
. Our results show that WIV is indeed able to induce IFNα in a TLR7-independent way. This may also be the case in the in vivo situation, where in accordance with its well-described adjuvant functions IFNα may lead to the production of Th1 type antibodies in TLR-deficient mice 
Induction of IFNα by WIV is TLR7-dependent in bone-marrow derived pDCs, but not in spleen-derived pDCs.
Possible TLR-independent pathways activated by WIV may involve the retinoic acid-inducible gene (RIG-I) 
is a cytoplasmic RNA-helicase that recognizes influenza virus by binding viral ssRNA bearing 5′-triphosphates which leads to IFNα production 
. The inactivated virus particles in WIV vaccine retained their membrane-fusion property (Text S2
) and part of the viral genomes could therefore have entered the target cell cytoplasm to be sensed by RIG-I.
Taken together our observations show that the superior immune response to WIV, relative to that to SV or SU vaccines, is driven primarily by TLR-dependent mechanisms. Herein the presence of the viral RNA in the vaccine seems to play a crucial role. In contrast to SV and SU vaccines WIV contains substantial amounts of viral RNA. Removal of ssRNA from WIV by detergent solubilization and ultracentrifugation followed by reconstitution of the viral membrane envelopes to virosomes abolishes the capacity of the vaccine to induce production of IFNα by pDCs in vitro
and Figure S1A
) and type 1 immune responses in vivo 
. On the other hand, ssRNA purified from WIV and condensed with polyethylenimine (PEI) did induce IFNα production in vitro (Text S3
and Figure S1B
). Obviously, exposure of the viral RNA to β-propiolactone in the course of virus inactivation leaves the RNA intact to trigger TLR7-mediated signaling pathways (), which translates into a strong and Th1-skewed antibody response to WIV in wild-type mice. In addition, the viral RNA may contribute to the TLR-independent part of the response to WIV since TLR7-independent production of IFNα could only be induced in pDCs by WIV and not by formulations (SV, SU, or reconstituted viral envelopes) which lack viral RNA () 
. These lines of evidence point to the ssRNA in WIV as the key component that enhances and steers the adaptive immune response by involvement of innate immune mechanisms.
IFNα induction in pDCs clearly discriminates WIV from SV and SU vaccines but seems to occur independent of TLR7. The fact that the immune response to WIV is predominantly dependent on TLR7 then suggests that other TLR7-mediated mechanisms, possibly involving conventional DCs and B cells, critically contribute to the immune reaction. Recently, an in vitro
study on B cells showed that TLR7 stimulation or CD40-CD40L binding by itself triggers IgG1 antibody production, but when simultaneously present induce proliferation and a switch to IgG2a production 
. Additional stimulation of IFNα/β receptors on the same cells further drives the production of IgG2a at the expense of IgG1 antibodies 
. Although this model might represent an over-simplification of the in vivo
situation, it is in line with our data. The different scenarios encountered upon immunization of wild-type and mutant mice with WIV, SV, or SU are summarized in . WIV provides the ssRNA for direct triggering of TLR7 in B cells as well as the CD40 ligand for CD40 stimulation on B cells through strong T helper cell induction, which was shown also to depend on TLR7 signalling. Together with IFNα produced by TLR7-mediated and/or TLR7-independent mechanisms, these signals will lead to the enhanced and strongly polarized Th1-type antibody responses characteristic for WIV. In the absence of TLR7, WIV-induced IFNα can still stimulate moderate production of Th1 type antibodies and increase the total IgG. In contrast, SV and SU vaccines are poor inducers of T helper cells and IFNα, and cannot stimulate B cells directly via TLR7. Consequently, SV and SU vaccines induce lower and more Th2-polarized antibody responses.
Table 2 Putative vaccine effects contributing to different adaptive immune responses based on the model proposed by Heer et al. .
Our data provide mechanisms which explain the superiority of WIV vaccine to prime HA-specific immune responses in mice. Whether similar mechanisms are operational in humans and contribute to the stronger immunogenicity of WIV compared to SV or SU in unprimed individuals remains to be elucidated. Despite the favourable immunogenic properties of WIV, recent clinical trials performed in the context of pandemic vaccine development show that even with WIV at least two immunizations with a substantial amount of antigen (15–30 µg) and/or the addition of adjuvants will probably be required to achieve immune responses that comply with the CPMP criteria. If TLRs are involved in the priming of humans with WIV, their role during recall responses may be less critical, given the fact that in general WIV, SU, and SV induce similar HI titres in primed populations 
. Use of WIV derived from wild-type virus instead of recombinant vaccine strains resulted in good antibody titres even without the addition of adjuvants and might thus be an option to obtain satisfying immune responses 
. Evaluation of adjuvants in combination with WIV in clinical trials is so far restricted to aluminium salts. However, where adjuvanted and non-adjuvanted WIV were compared side-by-side, effects of this Th2 adjuvant on vaccine efficacy were absent, poor, or inconsistent 
. So, better adjuvants have to be found that work synergistically with WIV in order to exploit the full potential of intact inactivated virus particles as vaccines.
In conclusion, our data reveal, for the first time to our knowledge, that TLRs play an eminent role in the immune responses to a classic influenza vaccine. Of the three influenza vaccine formulations studied here, only WIV efficiently triggered TLR7-mediated mechanisms leading to superior immune responses. Processing of inactivated whole virus particles into SV or SU eliminates the immuno-potentiating effect of the viral ssRNA, the primary PAMP in WIV vaccine, and results in a loss of quantity and shift in the quality of the immune response. Thus, TLR-dependent mechanisms appear to form the basis for WIV's antigen-sparing quality and hence its recognized strong potential as a pandemic vaccine candidate 
. Optimizing TLR7-signalling by rational vaccine design may produce even more potent vaccines, which are urgently needed in the face of the current influenza pandemic threat.