The discovery of broadly neutralizing Abs against the conserved stalk domain has spurred the development of stalk domain-based UIVs. The first attempt to produce immunosubdominant headless HA was made by Graves et al. in 1983 by the chemical removal of the immunodominant HA global head domain [24]. However, this treatment results in conformation changes in the HA2 stalk domain and removal of conformational epitopes in the HA1 stalk domain, which are needed for the binding of broadly reactive anti-stalk Abs [24]. Other efforts, including genetic treatment and expression of headless HA in E. coli or insect cells, have also been made; however, the conformation of the headless HA is likely far from optimal despite the fact that it provided homologous, heterologous, and heterosubtypic protection after virus challenges [25,26,27]. In 2015, Impagliazzo et al. [28] and Yassine et al. [29] independently generated stable and correctly folded headless HA by introducing a stabilizing element in the membrane-proximal end of the molecule. The headless HA induced stalk-specific Abs and protected animals against heterosubtypic challenge with H5N1 viruses via Fc-dependent mechanisms, such as antigen-dependent cellular cytotoxicity (ADCC), without robust neutralizing Ab titers [28,29]. These findings not only show the potential importance of headless HA in UIV development but an approach to overcome conformation changes in this molecule. However, constructs based on the H1 HA stalk only provided protection against group 1 IAVs. In 2016, Poon’s group first reported that a group 1 headless HA mini-stem conferred protection against both group 1 (H1 and H5) and group 2 (H3) influenza viruses [30]. Moreover, a recent study by Deng et al. showed the use of a headless HA formulated with a tetrameric M2e into protein nanoparticles induced long-lasting immune responses and protected mice from challenges with divergent IAVs of the same group or of both groups [19]. While continuous efforts have been made to develop UIVs based on headless HA, there are some concerns about the induction of antibody-dependent enhancement (ADE) by HA stem-specific Abs [31,32]. Although the above studies warrant further exploration of headless HA-based UIVs, it is a prerequisite to carefully evaluate whether headless HA-induced Abs could enhance ADE, thereby increasing disease severity.

Aside from headless HA, another strategy to refocus the immune response from the HA head towards the HA stalk is to use a chimeric HA (cHA). This approach involves the induction of anti-stalk Abs through sequential vaccination with cHAs sharing the same stalk domain, but different head domains (Figure 2). The cHA consists of the HA stalk domain from current seasonal IAVs, such as H1, H3, or influenza B, and an HA head domain from an exotic avian virus [33]. The first vaccination with a cHA induces Abs specific to both the head and stalk domains. Boosting with another cHA that contains the same stalk domain as the first vaccination and a different head domain leads to the induction of a primary response to the new head domain and a recall response to the fixed stalk domain. The anti-stalk response can be further boosted by additional immunization with yet another cHA consisting of the fixed HA stalk domain and a different HA head domain [34]. Vaccination with cHA in various vaccine formulations, such as recombinant protein, viral vector, or nucleotide-based vaccines, has been shown to induce highly stalk-reactive Abs and to consequently protect mice against challenge with divergent IAVs [35,36,37,38,39,40]. Further, sequential administration of the cHA vaccines (cH8/1 and cH5/1) was demonstrated to be locally and systemically well-tolerated in rabbits [41]. Since this strategy induces both anti-head and anti-stalk-reactive Abs, the protection conferred by cHA is mediated by both neutralizing and non-neutralizing Abs (such as ADCC activity) [35,36,37,38,39,40]. Treatment of ferrets with cHA-based vaccines conferred protection against pandemic H1N1 virus infection [42]. Sequential administration of cHA-based vaccines with live-attenuated vaccines also induced HA stalk-specific specific humoral responses, conferred protective heterosubtypic immunity against different IAVs in the ferret model [43,44], and induced long-lived immune responses against pandemic H1N1 virus infection [43]. Recently, phase I clinical trials (NCT03300050) using cHA-based UIV revealed that the vaccine was safe for human use (Table 1). Healthy adults immunized with the vaccine exhibited remarkably high anti-stalk Ab titers and long-lasting immune responses [43,45].

Although cHA-based vaccines could induce Abs against anti-stalk HA and provide protection against both homologous and heterologous viral challenges, the protective efficacy of this vaccine might still be limited depending on the chimeric antigen subtypes. Further, multiple-dose vaccinations of different chimeric vaccines are required to induce vaccine efficacy with a broad spectrum. Therefore, planned antigen selection and simplified vaccine strategies are needed for further study.

Mosaic HAs (mHAs) were designed with the idea of eliciting antibodies not only against the conserved stalk domain but also against epitopes in the head domain outside of the major antigenic sites. Given that some monoclonal Abs showed cross-reactivity with the HA head domain of different influenza strains [46,47,48], mHA was developed to maximize cross-reactive responses and minimize genetic differences. The mHA can induce immune responses toward both the head and stalk domain of HA (Figure 3). This differs from cHA in which the entire HA head region was replaced. Specifically, mHA is constructed by replacement of only major antigenic sites of the head HA with those from exotic avian HA subtypes. Thus, the conserved head and stalk domains of mHA are derived from H1, H3, or influenza B virus HA. Priming with mHA induces high amounts of Abs against antigenic sites of the HA head domain and low amounts of Abs specific for conserved epitopes in the HA stalk domain. However, boosting with another mHA (differing in antigenic sites) induces Abs against both antigenic sites and conserved epitopes in the HA head and stalk domains [34]. In the mouse model, sequential immunization with mosaic influenza B HA was found to induce cross-protection against both homologous and heterologous influenza B virus strains [49]. Vaccination with mosaic H5 HA enhanced antibody and T cell responses, providing protection against H5N1 and H1N1 viruses in mice [50,51]. Mice vaccinated with H1 mosaic HA or H3 mosaic HA also exhibited broadly protective immune responses against divergent influenza virus strains [52,53]. Kingstad-Bakkle and colleagues showed that a mosaic H5 vaccine antigen delivered via polyanhydride nanoparticles or modified virus Ankara induced humoral and cellular immune responses in both specific-pathogen-free (SPF) and commercial chicks [54]. A study by Florek et al. revealed that immunization with mosaic H5 HA increased virus clearance and elicited cross-reactive antibodies capable of mediating ADCC activity in rhesus macaques [55]. Recently, a mosaic receptor-binding domain (RBD) nanoparticle displaying multiple copies of eight different HA head domains from H1 HA antigens was developed at the Vaccine Research Center of the National Institute of Allergy and Infectious Diseases (United States). Vaccination with these mosaic RBD nanoparticles elicited cross-reactive B cell responses in mice [56].

Although mHA-based vaccines could provide broader cross-reactive protection against both homologous and heterologous viral challenges, similar to cHA-based vaccines, the protective efficacy of this vaccine might still be limited depending on multi-dose vaccination and selected antigen subtypes. Therefore, planned antigen selection and simplified vaccine strategies are needed for further study.

The COBRA strategy applies a computational method to generate a consensus sequence for all strains from a certain HA subtype [57]. The main purpose of this vaccine strategy is to generate a broad anti-HA-head response with enhanced HI [hemagglutination inhibition] activity. The COBRA HA approach has been demonstrated to be an effective strategy for the development of influenza vaccines against panels of H1N1, H5N1, and H3N2 viruses by eliciting broadly reactive Abs in mice, ferrets, and non-human primates [57,58,59,60]. Carter et al. reported that COBRA H1 HA antigens could generate anti-head Abs against 17 human seasonal and pandemic H1N1 viruses isolated since 1918 [58]. A recent study revealed that the use of COBRA HA antigens designed using both human and swine H1 HA sequences protected mice against viruses of both swine and human origin [61]. In addition, animal studies by Giles BM et al. have shown that COBRA H5 HA antigens elicit broadly reactive Abs with HI activity against 25 highly pathogenic avian influenza H5N1 virus strains [57,59]. COBRA H3 HA antigens have been demonstrated to induce HI antibodies to protect mice against a panel of H3N2 influenza virus cocirculating variants [60]. More recently, Ross et al. developed a COBRA H3 HA vaccine by generating a consensus sequence from 22,144 human A(H3N2) viruses collected from 2002 to 2015 [62]. This vaccine candidate showed an increase in HI activity compared to the wild-type HA vaccine. COBRA H2 HA antigens were also developed using human and avian H2 HA sequences, and the use of this vaccine generated broadly cross-reactive Abs against both historical human H2N2, avian, and mammalian H2 viruses recently isolated in mice [63].

Overall, the COBRA HA approach is one of the promising strategies to develop UIVs by generating broadly reactive anti-head HA Abs to protect against a broad range of influenza viruses. However, sometimes the generated consensus sequences do not generate fully functional HA proteins, and thus it needs to use alternative vaccine platforms, such as VLP. Currently, COBRA HA vaccines are in the late preclinical stage of development, and clinical trial studies are needed to investigate their vaccine effects in humans.

More recent advances in HA-based vaccines utilize targeting of hidden, conserved epitopes in the HA domain, called the “breathing” HA strategy [64]. Studies by Bangaru et al. [65] and Watanabe et al. [66] have shown that Abs directed against concealed, conserved epitopes in the HA head domain interface could provide protection against broad-spectrum IAV subtypes. For instance, a human monoclonal antibody, FluA-20, recognizes epitopes positioned in the 220 loop and the adjacent 90 loop, which is usually buried in the native HA trimer. This antibody protects mice from sublethal and lethal challenges with human H1N1, H5N1, H3N2, and H7N9 viruses [65]. The collective breadth of Abs (S5V2-19, H2214, S1V2-58, S8V2-17, and S8V2-37) isolated from memory B cells of inactivated seasonal influenza vaccine-immunized donors target novel sites at the interface of the HA domain and provide protection against both group 1 (H1, H2, H5, and H9) and 2 (H3, H7, and H14) viruses in mice [66]. However, a method to expose the hidden epitopes will require further exploration. Glycosylation of the HA domain is reportedly involved in the antigenic drift of IAVs, allowing escape from the host immune response [67,68,69]. Interestingly, the glycosylation site of the HA head domain can be modified to expose the hidden conserved epitopes in order to direct immune responses to the conserved HA domain. Previous studies have reported that mice vaccinated with the hyperglycosylated HA head domain exhibit enhanced, conserved stalk domain-directed antibodies and have better protection against lethal virus challenge than those vaccinated with wild-type HA [70]. Studies by Lin et al. have shown that vaccination of hyperglycosylated HA in various formulations, such as VLPs, recombinant protein, or adenoviral vector, elicits broadly neutralizing Abs against avian influenza H5N1 viruses [71,72]. Another study revealed that vaccination of glycan-shielded HA antigens stably expressed by CHO cell clones increased potential neutralizing Abs against heterologous H5N1 viruses [73]. Hyperglycosylation of the HA domain does not dampen overall humoral responses, but rather changes patterns of immunodominance and elicits broadly protective Abs, as reported by Bajic et al. [74]. Collectively, these studies indicate that hyperglycosylation could be a promising strategy to find hidden epitopes in the HA head and stalk domains, facilitating the development of novel UIVs in the future. Currently, glycosylation in influenza vaccine design is still under debate. Studies showed that glycosylation was important for virus infectivity and host immune responses. Wu et al. reported that glycosylation of HA at N-142 promoted H1N1 influenza virus infectivity and the glycosite 142 was important for the virus in evading immune responses in humans [75]. The addition of N-link glycosylation modification to the head HA region also decreased Ab titers in HI assay and plaque-reduction neutralization test [76]. Thus, it is necessary to determine the exact glycosites needed for influenza vaccine development. Various studies should also be performed to have a better understanding of the influence of glycosylation on vaccine efficacy.