Influenza is a global public health problem, causing approximately 300,000–650,000 global deaths each year [1
]. Influenza A and B are the major virus types that infect humans. Antibodies directed against the head domain of the surface glycoprotein hemagglutinin (HA) of influenza virus have proven to be the major source of protective immunity, blocking viral binding to the receptors on the target human cell surface and inhibiting viral entry to target cells. In response to human immunity pressures, antigenically distinct influenza viruses emerge frequently, caused by continual mutation (antigenic drift) [2
], or reassortment among viruses from different species (antigenic shift) that can lead to a pandemic with high mortality [3
To date, seasonal influenza vaccines composed of three or four inactivated virus strains are the only licensed vaccines to elicit or boost protective immunity against influenza viruses in the United States. However, both antigenic drift and shift necessitate that the flu vaccine be reformulated and re-administered annually [5
]. It is a formidable challenge to select the strains each year to protect against current circulating viruses based on viral surveillance data of the previous year [6
], and to produce a large amount of antigenically matched vaccine. Developing a “universal flu vaccine” that induces broadly cross-protective immunity is one strategy to overcome this challenge [7
Antibody mediated immune responses against influenza HA are multi-dimensional, targeting multiple antigenic determinants (epitopes) within the HA molecule. Antibody mediated responses are also incredibly complicated, as they are influenced and altered by an individual’s prior influenza exposure history. This includes factors such as “original antigenic sin” (OAS) [9
] (also known as HA imprinting [10
]) and the shared epitopes between proteins from different influenza strains that induce cross-strain immunity, such as heterosubtypic immunity [11
]. The effects of pre-existing antibodies on the B cell response to vaccine strains that contain HA antigenic sites similar to those from prior exposures are still unclear. Systems serology, the application of bioinformatics to multidimensional data regarding anti-influenza IgG binding specificity and repertoire in response to vaccination, has emerged as a way to understand these responses, and to aid in vaccine design.
Because of the complex interplay between pre-existing, circulating, anti-HA antibodies and human IgG-mediated influenza responses, the first step in comprehensive analysis is measurement of anti-influenza HA IgG binding patterns against multiple influenza strain HAs. Such measurement is referred to as multi-dimensional, referring to the multiplicity of influenza strain binding reactions quantified. Such measurements are critical for understanding how IgG recognition of shared epitopes across influenza strains can lead to cross-strain protection, and for better defining the functional host anti-HA influenza repertoire.
Various assays exist to measure the host anti-HA influenza antibody response. The assays currently used to estimate the HA IgG antibody binding to single HA proteins, such as hemagglutinin inhibition (HAI) [13
], micro-neutralization (MN) [15
] and enzyme-linked immunosorbent assay (ELISA), all require a large amount of serum sample in order to test the cross-reactivity against an array of virus strains. These assays are also expensive and time consuming, limiting their usefulness in unraveling the complexity of cross-reactive antibody patterns to influenza viruses. In contrast, the novel technology of array-based high throughput multiple dimensional assay (MDA) provides a powerful tool to comprehensively analyze the presence and effects of broad cross-reactive antibodies (bcAbs) against the influenza HA protein.
Here, we review the genetic foundations of shared epitopes leading to IgG cross-reactivity between antigenically similar influenza virus strains, and the contributions of these cross-reactivities to OAS and subsequent host immune responses to influenza infection and vaccination. The remainder of the review discusses the technology of MDA, and highlights the application of MDA as a powerful tool for future influenza immunity studies and universal vaccine development.
2. Hemagglutinin (HA) and Its Antibodies
Influenza viruses, the pathogens that cause flu, belong to the Orthomyxoviridae
family, a group of negative-sense single strain RNA viruses [17
]. Influenza type A has two phylogenetic groups based on amino acid sequence and, to date, 18 HA subtypes: Group 1 (H1, H2, H5, H6, H8, H9, H11, H12, H13, H16, HA-like H17, and HA-like H18) and Group 2 (H3, H4, H7, H10, H14, and H15) [18
]. Influenza A viruses are further named based on the composition of major surface glycoproteins HA and neuraminidase (NA) (e.g., H1N1 or H3N2). Influenza type B also has two phylogenetically distinct lineages called Yamagata and Victoria [19
]. The major source of human protective immunity is the antibodies directed against the head domain of the HA of influenza virus [20
]. HA is the most abundant influenza viral surface glycoprotein and mediates binding to sialic acid expressed on the surface of target host cells. HA is synthesized as a polypeptide (HA0) before being cleaved into HA1 and HA2 subunits, which fold into a trimeric spike. The membrane distal globular head region of HA is composed of HA1 and contains the receptor binding site (RBS) that the virus uses to bind to host cell sialic acid. The stalk region then mediates virus fusion into host cells through structure transformation [21
Protective antibody-mediated immunity against HA is the first line of defense in preventing influenza virus infection. Such immunity is elicited by prior influenza exposure: infection or vaccination [22
]. Anti-head HA antibodies typically target epitopes in and around the RBS. Five major B cell epitopes have been identified for H1 (Sa, Sb, Ca1, Ca2, and Cb) [24
] and H3 (Eptitopes A–E) influenza strain HAs [25
]. The HA head region, formed by HA1, is strongly immunodominant, highly mutable, and strain-specific [26
]. The HA stalk region, formed by HA2 as well as the N- and C-terminal ends of HA1 in an alpha-helical structure, supports the head region of HA [27
]. The highly conserved nature of the HA stalk makes it a promising target for universal influenza vaccines [28
The goal of universal vaccines is to elicit protective broad cross-reactive antibodies (bcAbs), especially broad neutralizing antibodies (bnAbs). Most head-reactive antibodies are not bcAbs or bnAbs but rather strain-specific. However, more and more head domain recognizing bnAbs have been identified, such as KBm2, 5J8 and CH65, which neutralize a broad spectrum of H1 strain viruses in the MN assay [31
], and 8M2, which neutralizes many H2 strains [34
]. Several head-reactive bnAbs demonstrate heterosubtypic reactivity, such as C05, F045-92 and S139/1, which recognize the conserved receptor binding pocket on the HA head [35
]. These three bnAbs can neutralize H1, H2, and H9, while C05 can also weakly neutralize the Group 2 H3 influenza virus [35
Recently, an increasing number of bnAbs have been isolated and identified from the B cell repertoire after influenza virus infection and vaccination [31
], targeting both head and stalk regions of HA. Some bnAbs that target stalk region of HA neutralize a wider range of influenza types and subtypes [18
]. Human monoclonal antibodies CR6261, F10 and A06 were isolated from recently vaccinated donors and shown to neutralize nearly all Group 1 viruses [40
], while CR8020 and CR8043 neutralize a wide breadth of Group 2 viruses [43
]. Some monoclonal antibodies show broad cross-group influenza A reactivity, including MEDI8852, 27F3, FI6v3 and CR9114 [45
]. Notably, there are other in vivo mechanisms involved in antibody mediated broad-protection, such antibody-dependent cell-mediated cytotoxicity (ADCC) and antibody-mediated cellular phagocytosis (ADCP) [49
]. For example, the novel isolated human mAb FluA-20 was shown to protect mice against lethal challenge with H1, H3, H5 and H7 influenza A subtype viruses [51
]. It binds to an extremely well conserved epitope in the peripheral interface of the HA trimer, a novel epitope on the head region of HA, with extra high affinity. After binding to HA, it quickly interferes with the trimeric structure of HA, which blocks viral cell-to-cell spread. This mAb offered protection from influenza virus infection in vivo, but it did not display neutralizing activity in HAI or MN assay in vitro studies [51
]. This suggests that there are some bcAbs that are able to protect against influenza virus that would likely be disregarded by traditional assays. BcAbs can be detected by ELISA and other binding assays, including MDAs, thus highlighting the need for more sensitive assays in the search for broad cross-reactive antibodies.
One major issue with the evaluation of bnAb activities by traditional HAI and MN assays is that they can only measure the magnitude of bnAb against specific virus strains, and cannot determine the breadth of bnAbs against a large panel of influenza strains. In addition, during antibody screening, these assays most likely omit bcAbs, such as the human mAb FluA-20 [51
], as we discussed above, did not display neutralizing activity in MN assay in vitro.
3. Complexity of Human Immune Responses against Influenza Virus
The complexity of the human immune response to repeated influenza virus exposure is another major obstacle to the development of a universal vaccine. Because of the frequent antigenic drift in circulating influenza strains, humans have more complicated immune responses than can be modeled in naive animals. Each person has a unique history of influenza virus exposure, leading to pre-existing immune repertoires that are activated in the event of an immune challenge with an antigenically similar flu strain.
In 1960, Thomas Francis Jr. reported that antibodies against the first H1N1 flu strain encountered in life would be produced at high levels throughout a person’s lifespan, to the detriment of future specific responses to new strains [9
]. He coined the term “Original Antigenic Sin” (OAS), now referred to as imprinting, to describe how a specific immune response to a flu strain can be preferentially directed at a previously encountered strain. This phenomenon also relates to the cross-protection provided by pre-existing bcAbs and how the breadth and protective potency of cross-reactive immunity is enhanced by infection or vaccinations. For example, the lower mortality of older individuals during the 2009 H1N1 pandemic is attributed to the structural similarity between the pandemic 2009 “Swine” flu virus HA and the pandemic 1918 “Spanish” flu virus HA, suggesting within-subtype cross-strain protection [52
]. Importantly, in 2016, using all known fatal human cases of H5N1 and H7N9, Gostic et al. [10
] found that childhood H1 and H3 imprinting provided 75% and 80% protection against death from H5N1 and H7N9, respectively. Because H1 and H5 are found in phylogenetic Group 1, and H3 and H7 are found in Group 2, these results suggest that antigenic seniority boosts can offer cross-protection against HA subtypes of the same group [10
]. The mechanism of such immune imprinting is unclear, but it has been hypothesized that after a large number of memory B cells (MBCs) are activated during first influenza virus exposure, the next exposure to an influenza strain with some mutated and some shared epitopes will show lower de novo naive memory B cells activation against the new
epitopes. This suggests that preexisting antibodies could play a role in the MBC response, such as accelerating the clearance of influenza antigens or sterically blocking MBCs from accessing specific epitopes [55
]. Meanwhile, the MBCs specific for epitopes present in the first strain would proliferate more since they have been activated again [57
]. OAS [58
], “antigenic seniority” [59
] and HA imprinting [10
] try to describe the effects of pre-existing antibodies on the antibody response to similar or related influenza virus strains. Due to the multi-dimensional nature of human immune repertoires, single-dimensional assays are extremely limited in their ability to measure the breadth of pre-existing bcAb and MBC responses. MDAs, on the other hand, are an ideal tool for measuring pre-existing bcAb profiles and broad influenza immunity.
4. Multidimensional Assays (MDAs) for Anti-Influenza Antibodies
The gold standard and most widely used assays to evaluate antibody activity against HA and protection in clinical trials are HAI [13
] and MN [15
] assays. Both assays are semi-quantitative with a single target virus strain providing a discrete ranked readout of one of 8–14 titer values based on two-fold dilutions of serum samples. Including ELISA, which is less frequently used in influenza studies, all these common methods are single dimensional assays, which require the user to perform antibody testing for each strain of interest separately. This process is not only time-consuming and labor-intensive, but also requires large sample volumes. In addition, these assays are limited in their ability to show the breadth of cross-reactive anti-influenza antibody response.
To overcome the limitations of single dimensional assays, novel multidimensional assays (MDAs) have been developed over the last decade. MDAs are high throughput assays that use protein array technology to simultaneously measure antibodies against a panel of the HA proteins and peptides of multiple influenza virus strains in a single test with minimum amount of sample. They can measure the magnitude and breadth of antibody response against HAs of influenza virus. In general, the purified HA proteins are immobilized on a solid surface such as microchips, membranes or beads, to keep the native structure and provide their maximal binding properties. Then, the reactive antibodies are characterized by binding to the protein, followed by a fluorescent probed secondary antibody that is read by an array chips reader as median fluorescent intensity (MFI). The HA protein or peptide array offers the advantage of multiplex capabilities to generate statistically powerful data while conserving time, money, and requiring minimal sample compared to the traditional assays. While not a functional assay, multiple studies have confirmed that MDA results correlate well with HAI titers [60
]. Critical for the understanding of OAS, “back-boosting”, and the effects of pre-existing cross-strain immunity on current vaccine responses, such methods allow testing reactivity against a large number of antigenically related and disparate influenza proteins (generally HA at the moment).
Based on the immobilizing materials, there are two major types of array-based assays currently used for evaluation of HA antibodies: protein microarray and Luminex assays. The first method involves printing HA protein on chips to estimate the binding antibodies. The first report of HA protein array assay was published in 2010 [65
], and since then more than 10 publications have shown its powerful potential to study the breadth of cross-reactivity of HA antibodies on the population level (see the list in Table 1
). At present, 283 HA proteins can be printed on one microchip for maximum efficiency [66
]. However, this process requires expensive and specialized equipment, including a micro-printer and dedicated scanner.
In contrast to peptide arrays, the Luminex-based MDA method, which involves coupling HA protein to color coded Luminex beads, allows the user increased flexibility to customize the panel by easily combining multiple strain-specific beads without reprinting the entire panel [63
]. In addition, Luminex readers are more widely available now than chip scanners. However, Luminex-based MDAs support fewer analytes per assay. For example, the Luminex 200 can detect 100 color-coded beads, and the Magpix analyzer can read 50 coded beads [68
The first Luminex-based MDA, mPLEX-Flu, was developed to characterize the breadth and magnitude of the IgA, IgM and IgG antibodies against a large panel of whole HA proteins of multiple influenza virus types and subtypes in 2015 [67
]. Our recent comprehensive studies, with novel statistical methods and a continuous readout across a 4.5 log range, indicated that MDA highly correlated with HAI and MN results, and with substantially better sensitivity and precision on account of continuous readout [64
]. Furthermore, another study showed that using individual standard curves for each influenza HA strain in the mPlex-Flu assay to independently calculate IgG concentrations against each virus strain enables the direct comparison of serum anti-HA IgG concentrations against different influenza HA subtypes [76
]. This ability addresses an essential issue for estimation and comparison of cross-reactivities of influenza antibody against multiple strains that has always plagued single-dimensional assays including HAI, MN and ELISA. The principle of the mPLEX-Flu assay is shown in Figure 1
Based on the above characteristics of MDA, we use the example of a Luminex-based MDA, the mPLEX-Flu assay [67
], to discuss the application of MDA on influenza vaccine development and basic viral immunity research. The major applications are summarized in Figure 2
7. Limitations of MDAs
As a novel technique, mPLEX-Flu and other MDAs also face some challenges and limitations. First, because MDAs detect the binding between the antibodies and HAs of influenza viruses, they are not able to directly reflect antibody protective function. By contrast, MN assays can estimate titers of neutralizing antibodies that inhibit influenza virus infection. However, our studies showed that data generated from mPLEX-Flu assay are highly correlated to HAI and MN data [31
]. On the other side, HAI and MN assays have the chance to omit bcAbs that offer influenza virus protection through ADCC and ADCP, as discussed above. MDAs assay can be used to perform large scale screening, which can then be combined with other assays to identify and clarify the characters of antibodies.
Other limitations of MDAs are caused by the HA antigen coupled on the Luminex beads or microchips. First, most HA proteins used in the MDAs are expressed and purified by an insect cell baculovirus system that has to be glycosylated during post-translational modifications in infected insect cells. However, the insect cell glycosylation pathway is far simpler than those of human cells [99
], even when both occur on the same N-glycosylation sites. Moreover, it has been reported that glycosylation of HAs is a crucial factor that needs to be considered when studying influenza infection and antigenic mutation (see review [100
]). At present, it is not clear if differences in antigenic binding could be caused by the differences between insect cell-resourced HA proteins and those from human cells. Second, the density of HA proteins on the bead surface can vary, even when the coupling protocol and protein concentrations are kept consistent. This leads to error when comparing the binding between different influenza virus strains, which is similar to the error when comparing the HAI and MN titers between two viruses. To minimize this error, we established a unique method for generating standard curves for IgG binding to each virus strain [76