Stringing Together a Universal Influenza Antibody

Graphical Abstract

Each year, between two and five million cases of severe disease are reported from seasonal influenza infection, resulting in up to 500,000 deaths worldwide. In the developed world, vaccinations provide protection for large percentages of the population, but current vaccines only immunize against strains that are most commonly infectious to humans. These subtypes include two influenza A viruses (H1N1 and H3N2) and two influenza B viruses (B/Yamagata and B/Victoria lineages). However, numerous other subtypes are prominent in other species, particularly birds and swine, and these pose a risk of human pandemics in the case of antigenic shift and cross-species infection. Different influenza subtypes have different isoforms of the key glycoproteins hemagglutinin and neuraminidase. For hemagglutinin (HA) in particular, there are 16 known subtypes. A universal vaccination or antibody capable of neutralizing all HA subtypes would lead to better seasonal flu coverage and, importantly, protection in the case of an influenza pandemic. [1]

There has been some progress in recent years in producing antibodies for passive immunization, a strategy that provides temporary protection from a pathogen following direct administration of neutralizing antibodies. To be effective against influenza, a passive immunization strategy would require broadly-neutralizing antibodies (bnAbs) which target highly conserved elements of the HA “stem” domain, rather than the more immunogenic but hypervariable “head” domain (Figure 1). However, while bnAbs exhibit some cross-subtype neutralization, they have lacked effectiveness against both influenza A and B viruses. [2] In work recently published by Laursen et al., the authors build on prior work on bnAbs by applying approaches from antibody engineering. The primary result is the development of new multivalent antibodies that, in mouse models, block infection and mortality for all classes of influenza. [3]

Figure 1. Binding sites of single-domain antibodies (sdAbs) to the influenza hemagglutinin trimer.

An alignment of four crystal structures^[3] of sdAb-HA complexes depicts the relative binding sites of the four sdAbs. SD84 binds to the head domain, while SD36, SD38, and SD83 have overlapping but distinct binding sites on the conserved stem domain. H3N2 HA^[2] was used to model relative binding sites.

The authors began by producing single-domain camelid antibodies (sdAbs) against influenza HA. sdAbs are single-chain, small in size (15 kDa), and have excellent pharmacological profiles, making them good starting points for antibody engineering. In prior work, these small, high-affinity “nanobodies” have been fused linearly to produce multivalent antibodies. Several examples of multivalent sdAbs are in clinical trials, including ALX-0171 (phase II for respiratory syncytial virus infection), which is a trivalent antibody made up of three identical sdAbs targeting the RSV F-protein, and ALX-0761 (phase IIb for psoriasis), which is a trivalent antibody made up of three different sdAbs, two of which target different disease-related proteins and a third that targets human serum albumin to improve pharmacokinetic properties. These examples highlight the viability of rationally engineered, multidomain sdAbs as a potential prophylactic or treatment for influenza infection. [4]

sdAbs were generated by immunizing llamas with influenza vaccine (H1N1, H3N2, and B/Brisbane-like viruses) as well as recombinant HA. After several rounds of immunization, peripheral blood mononuclear cells were isolated, and their genetic material was incorporated into a phage display library to screen for sdAbs capable of cross-subtype influenza neutralization. This process discovered four sdAbs: SD36 and SD38 which potently neutralized different influenza A viruses, and SD83 and SD84 which potently neutralized many influenza B viruses. X-ray crystallography revealed that three of the four antibodies bound to highly conserved residues of the HA stem domain (Figure 1). To achieve maximal breadth, they generated a tetramer of all four newly-discovered sdAbs (MD2407) as well as a multivalent antibody with two of these tetramers conjugated to a human Fc domain (MD3606, Figure 2). The multivalent antibody was effective at neutralizing all subtypes of influenza tested (except for one avian H12 strain) with low nanomolar IC₅₀ values. Interestingly, the binding modes revealed by X-ray crystallography suggest that the tetramer cannot bind multivalently to a single HA trimer. Cryo-electron microscopy of MD2407-bound HA trimers identified a subpopulation of HA trimers that appeared to be cross-linked by MD2407, suggesting an unusual mechanism for the enhanced potency of the multidomain antibody.

Figure 2. Domain structures of various antibodies developed to neutralize influenza hemagglutinin.

The single-domain antibody (sdAb) is a discrete domain of the natural camelid antibody. In Laursen et al. and other work, these are fused linearly to form multivalent assembles such as the tetramer MD2407, or the dimer of tetramers MD3606.

To demonstrate effectiveness for passive immunization, the authors tested MD3606 in mice challenged with different subtypes of influenza virus. Administration of 1 to 5 mg/kg of MD3606 one day prior to virus exposure offered complete protection from all subtypes of influenza A and B tested. These results identify MD3606 as the most broadly acting influenza antibody to date. In a final set of experiments, the authors addressed the challenge of providing longer-lasting protection. Because prophylaxis would require consistent administration of a universal antibody over the course of a flu season, the authors used an adeno-associated viral (AAV) vector to induce the expression of MD3606 in the nasal epithelium of mice. Delivery of up to 5×10⁹ genome copies of the AAV was sufficient to offer complete protection against three different subtypes of influenza (A1, A2, and B). All together, these results provide strong evidence that engineering multidomain antibodies from carefully selected sdAbs could be a powerful approach to producing broadly-neutralizing antibodies to combat highly variable pathogens.

Laursen and colleagues are the first to apply multivalent sdAbs to simultaneously target influenza A and B. The result, multivalent antibody MD3606, surpassed previous bnAbs in potency and breadth. To further optimize the structure and function of MD3606, a better understanding of the mechanism of action will be critical. If activity involves cross-linking of multiple HA trimers, this feature could be optimized by changing the relative locations of the sdAb domains and altering the linkers between them. Further, while clinical trials involving multidomain antibodies are on the rise, some have been terminated prematurely due to dose-limiting toxicities. A better understanding of the features responsible for these off-target effects will be crucial for the entire field. In parallel, AAV-mediated gene therapies continue to progress through the clinic (with one recent FDA approval), mostly for rare genetic diseases and other difficult-to-treat indications. [5] However, if these trials continue to be successful, AAV-mediated delivery of a broadly-neutralizing antibody such as MD3606 could emerge as a useful treatment for acute influenza, or even as a prophylactic for seasonal flu outbreaks.