H1N1 G4 swine influenza T cell epitope analysis in swine and human vaccines and circulating strains uncovers potential risk to swine and humans

Abstract Background Pandemic influenza viruses may emerge from animal reservoirs and spread among humans in the absence of cross‐reactive antibodies in the human population. Immune response to highly conserved T cell epitopes in vaccines may still reduce morbidity and limit the spread of the new virus even when cross‐protective antibody responses are lacking. Methods We used an established epitope content prediction and comparison tool, Epitope Content Comparison (EpiCC), to assess the potential for emergent H1N1 G4 swine influenza A virus (G4) to impact swine and human populations. We identified and computed the total cross‐conserved T cell epitope content in HA sequences of human seasonal and experimental influenza vaccines, swine influenza vaccines from Europe and the United States (US) against G4. Results The overall T cell epitope content of US commercial swine vaccines was poorly conserved with G4, with an average T cell epitope coverage of 35.7%. EpiCC scores for the comparison between current human influenza vaccines and circulating human influenza strains were also very low. In contrast, the T cell epitope coverage of a recent European swine influenza vaccine (HL03) was 65.8% against G4. Conclusions Poor T cell epitope cross‐conservation between emergent G4 and swine and human influenza vaccines in the US may enable G4 to spread in swine and spillover to human populations in the absence of protective antibody response. One European influenza vaccine, HL03, may protect against emergent G4. This study illustrates the use of the EpiCC tool for prospective assessment of existing vaccine strains against emergent viruses in swine and human populations.


| INTRODUCTION
The annual occurrence of influenza outbreaks causes considerable morbidity and mortality and poses a global public health challenge.
The causative agent of these outbreaks is influenza A virus (IAV), although influenza B also contributes to some outbreaks. IAV infects a wide range of hosts including mammals and avian species. Pigs are one of the most important hosts given their susceptibility to a broader range of avian and human influenza viruses, and they are often the source of novel reassortant viruses from avian-and human-origin strains. IAV infection of swine causes significant economic losses for commercial pork producers. In addition, swine IAV poses a threat to human health due to the potential for swine influenza A virus (swIAV) to spill over into the human population, as occurred in 2009 (for H1N1/pdm09). 1,2 After the outbreak of H1N1/pdm09 in human in 2009, the virus was reintroduced into pig herds around the world and reassorted with other swine influenza viruses, forming new swIAVs which spread within pig herds in the United States (US), Brazil, Europe (EU), Japan, China, and other countries. 3,4 Some of these reassortants harboring H1N1/pdm09 internal genes have gradually evolved and replaced previous strains of swine influenza, demonstrating antigenic drift due to genetic selection pressure exerted on H1N1/pdm09 and other strains worldwide. 5 6 Emergent G4 is a reassortment product of EA H1N1 virus, bearing H1N1/pdm09 and TR-derived internal genes. In the reassorted G4, the HA and NA genes are from the EA H1N1 lineage, and in particular, the HA gene falls within the 1C.2.3 lineage. The viral ribonucleoprotein (vRNP) genes and M gene are from the H1N1/pdm09 lineage, and the non-structural (NS) gene is derived from the TR lineage. 6 Given that as many as 20% of pork industry workers in China have been found to be seropositive for G4 antibodies, it appears that G4 has the potential to cross species barriers. 6 Vaccine efficacy evaluation usually involves assessment of cross-reactive influenza-specific antibodies generated by exposure or vaccination. Seasonal vaccination in humans does not generate antibodies that protect against G4 (hemagglutination inhibition; see Figure 1C in Sun et al. 6 ). In a separate study, monoclonal antibodies isolated from mice immunized with pandemic (A/California/07/09) hemagglutinin (HA) and a novel flu vaccine, computationally optimized broadly reactive antigen (COBRA) P1 HA, generated hemagglutination of G4 virus-like particles in vitro. 7 The relevance of this murine study to swine and human populations remains to be determined.
When cross-reactive antibodies are not present, cross-conserved T cell epitopes in IAV vaccines, and strains have been shown to play an important role in reducing morbidity and limiting the spread of IAV, even when vaccines and emergent strains are poorly matched. [8][9][10] There is strong evidence that (1) T cell responses generated by previous influenza exposure cross-reacts with novel IAV strains 11 and (2) T cell responses are critically important for protection against IAV infection in both humans and swine. 10,12,13 Thus, even in the absence of cross-reactive antibody to G4, T cell cross-reactivity might be protective.
Here, we apply an immunoinformatics tool to evaluate whether existing vaccines may have the capacity to prevent the spread of G4 in humans and swine. We developed a computational workflow that employs the Epitope Content Comparison (EpiCC) algorithm to measure the degree of epitope conservation between target vaccines and outbreak strains. In previous studies, we identified an EpiCC score that was correlated with protection in the absence of cross-reactive antibody. We used the same approach to establish thresholds for protective efficacy for vaccines against circulating strains in this study.
EpiCC's estimation of T cell epitope conservation between emerging viruses and vaccine strains may be useful as a potential surrogate measure of vaccine efficacy, in conjunction with other methods of pandemic risk assessment. platform.gisaid.org/epi3/; accessed in August 2020). 14 Strain information regarding commercial use EU swine influenza vaccine was based on literature review (Table 1A). 15 G4 genotype strains were used in this study. 6 Given that the 29 HA G4 strains were derived from a shorter time range (2016-2018) and that there is great similarity (percentage identity in the range of 95.8%-100.0%) between these sequences, eight sequences were randomly selected for analysis in this study ( Figure 1 and Table 1B)

| Data curation
Duplicate and partial sequences containing less than 1400 nucleotides were removed using a publicly available python script. 18 Phylogenetic analysis was performed following sequence alignment using MUSCLE 3.8.31. 19 Maximum-likelihood phylogenetic trees were constructed with RAxML.v8 using the GTR-GAMMA nucleotide substitution model. 20 To ensure computational tractability and to preserve representative clades, Phylogenetic Diversity Analyzer (PDA) was used to subsample 150 sequences each from large dataset that consisted of swIAV strains circulating in the US and EU, respectively, as well as 300 sequences from human IAV strains circulating in the US. 21 The final reduced dataset (supporting information Table S1A-C) was translated into amino acid sequences and combined with selected G4 strains and respective vaccine strains for three sets of analyses

| T cell epitope binding prediction
We initiated our analysis by focusing on HA, given its importance as the critical antigen that is most relevant to protective immunity to influenza. Additional antigens were also evaluated (see Section 3.3).
After compiling IAV sequences as described in Section 2 and illustrated in Figure 1, the translated HA protein sequences were screened using host-specific T cell epitope identification algorithms developed by EpiVax. Particularly, PigMatrix epitope prediction tools were used for the swine sequences, and EpiMatrix was used for identification of human epitopes. These prediction tools parse sequences into overlapping 9-mer frames to define the relative likelihood of binding to a set of prevalent swine leukocyte antigen (SLA) or supertype human leukocyte antigen (HLA) class I and II alleles. 23,24 More specifically, using a position-specific scoring matrix, predicted binding scores were computed by PigMatrix for the likelihood of each 9-mer in the HA antigens of IAV to bind to a panel of prevalent class I (SLA-1*01:01, 1*04:01, 1*08:01, 1*12:01, 1*13:01, T A B L E 1 A Source of H1 HA sequences used in the analyses ing preferences) that cover the genetic diversity of more than 95% of human populations globally. 23,27,28 For each 9-mer, i in each individual allele a of a set of MHC alleles A, PigMatrix or EpiMatrix raw scores, r, are normalized to Zscores using the average μ and the standard deviation σ of scores calculated for 100,000 random 9-mers using the formula below. 23,24 Nine-mers with Z-scores greater or equal to 1

| Analysis of T cell EpiCC
We applied EpiCC to facilitate the pairwise T cell EpiCC of protein sequences ( Figure 1). EpiCC enables large-scale sequence analysis for conservation of T cell epitopes between swine and human flu vaccines and circulating IAV and G4 strains, focusing only on shared T cell epitopes between the vaccines and the target IAV strains. 26,30 Once T cell epitope content is defined for each vaccine or strain, the set of conserved T cell epitopes that are shared between two strains can be enumerated.
In mathematical terms, EpiCC assesses the relatedness of T cell epitope, i, contained in a protein sequence of vaccine strain v and T cell epitope, j, contained in a protein sequence of a strain s based on respective PigMatrix SLA binding or EpiMatrix HLA binding score.
Since cross-reactive memory T cells can be stimulated by epitopes (i, j) with identical TCR-facing residues (TCRf) that may have different HLA binding pocket residues, as long as they bind to the same alleles, we searched for potentially cross-reactive epitopes that shared TCRf as follows: Cross-conserved class I epitopes were defined by identical residues at positions 4, 5, 6, 7, and 8, and class II epitopes were defined by identical residues at positions 2, 3, 5, 7, and 8. The score of cross-conserved T cell epitope shared between two strains s and v was calculated using predicted binding probabilities as follows: To normalize shared EpiCC score, the sum of shared epitope scores of each i, j, was normalized by the total number of compared pairs, p, and by the number of MHC alleles in A. This is to account for different epitope densities and for comparison of values of E determined using different numbers of MHC alleles. Therefore, the normalized shared EpiCC score, (termed as EpiCC score), can be computed by applying the following equation: Maximum EpiCC scores were calculated. These scores were derived from shared EpiCC scores computed from the comparison of any sequence to itself. The greater the maximum EpiCC score, the greater the total epitope content of the sequence. Since no sequence can be better matched to another sequence than itself, the maximum value for any comparison between any target sequence and a comparison sequence is always less than or equivalent to their maximum EpiCC scores. Both class I and class II EpiCC analyses were combined by summing class I and class II EpiCC scores (termed as total EpiCC score) for each vaccine-to-strain comparison.
When the shared T cell epitope content of a strain of influenza is highly related or "covered" by a given influenza vaccine sequence, the vaccine-to-strain's EpiCC score approaches the circulating strain's maximum EpiCC score (it approaches the maximum if nearly all the epitopes are identical, as defined above). To determine vaccine-tostrain EpiCC scores coverage, each vaccine-to-strain comparison was divided by that strain's maximum EpiCC score and expressed as a percentage. The greater the T cell epitope coverage (percentage), the better the vaccine matches or covers the T cell epitope content of the circulating strain sequences ( Figure 1).

| EpiCC scores and EU vaccine efficacy estimation
To identify a threshold of protective efficacy of existing swine vaccines against circulating field strains, we extrapolated from available data, using an approach similar to one that we have already published. 26 For HA sequences, we calculated the EpiCC scores for three H1N1 EU commercial vaccines and three experimental monovalent vaccines against EU circulating swine IAV strains from the same period. Once we had obtained the scores, we identified the minimum EpiCC scores that correlated with protective endpoint results in four published studies that used the commercial and/or experimental vaccine strains. Vaccines were protective if they significantly reduced lung virus titers. The EpiCC score protective threshold was defined as the lowest EpiCC for at which the vaccine strain was shown to be protective. This is the main criterion for evaluating protection in the EU. 16 Scoring was performed independently of and prior to obtaining information about the outcomes of the vaccination and challenge studies.  (Table 1A), and the eight G4 sequences (Table 1B)

| RESULTS
We set out to evaluate the potential for cross-conserved T cell immune responses to protect against G4 in swine and human populations using immunoinformatics methods. The results of this analysis are divided into two parts due to species-specific MHC binding preferences (swine and human) and the species-specific circulation patterns of influenza strains. The first part of the analysis focused on protection by swine vaccines against G4 in swine, and the second part focused on predicting protection against G4 influenza in case of spillover into human hosts.
For the swine IAV analysis, we evaluated two commercial H1N1 swine influenza strains used in the EU and three US commercial H1N1 vaccine strains against strains from their respective regions of the world. We also evaluated the experimental (swine) COBRA vaccines that were previously studied for cross-reactive antibodies, for T cell epitope conservation against G4 using swine epitope prediction tools (PigMatrix and EpiCC). In the second section, we evaluated human T cell epitope content relatedness of five seasonal (human) F I G U R E 2 This radar plot of the EpiCC analysis enables the comparison of T cell epitope relatedness of the US swine influenza vaccines (A: Commercial; B: COBRA) to circulating US swIAV and G4 strains. EpiCC scores are plotted in a radial fashion in order of chronological time while also grouping flu variants into strain families, and color-coded lines represent each of the vaccine strains (V) compared with each of the circulating strains. Each of the vaccine strain labels is shown in legend. The ring surrounding the radar plot identifies the swIAV sequences metadata using the US clade naming system, which includes alpha, beta, delta, gamma, and pandemic 2009 lineage (PDM-09). The two shaded gray circles near the center of the radius define vaccine efficacy thresholds as reported in Gutiérrez et al. 2017 [26]. Refer to Figure 1-Data interpretation for more details. As indicated in both A and B, the G4 strain has few cross-conserved T cell epitopes with vaccine strains (all vaccines' EpiCC scores fall below the protective thresholds).
H1N1 IAV vaccine strains and three experimental (human COBRA) IAV vaccine strains in the US human population to determine their potential to control G4 in the event of G4 spillover into the US human population, using EpiMatrix and EpiCC.
As described above, T cell epitope content relatedness was defined as the density of shared T cell epitope content between the vaccine of interest and targeted strain (EpiCC score). A percent coverage was used to normalize score and permit comparison to established protective thresholds. Figure 1 illustrates how radar plots are constructed; the results of the swine and human vaccine-tostrain analyses are shown in separate radar plots (Figures 2-4), combining MHC class I and II comparisons (SLA for swine; HLA for human).
In the sections below, we described the results of the EpiCC analysis for HA, as it is the principal target of protective immune responses in influenza infection and is the most variable sequence in the pathogen. 32 Changes to vaccine composition and changes in vaccine efficacy are primarily due to drift and shift in the sequence of HA antigen. In Section 3.3, we provided results for additional influenza proteins, to allow for a more comprehensive understanding of their potential to contribute to changes in T cell epitope content.

| EpiCC analysis of US vaccines and strains
EpiCC scores comparing the T cell epitope content of the HA of three commercial swine H1N1 vaccines used in the US (FSXP/IA00, FSXP/ OK08 and FSPDM/CA09) to 150 swIAV strains circulating in the US from 1939 to 2020 and eight G4 sequences were calculated ( Figure 1A). Overall, for these swine vaccine-to-strain comparisons tended to be clade-specific. Higher total EpiCC scores (greater crossconservation) were observed for FSXP/IA00 when it was compared with US swIAV circulating strains.  26 were extrapolated and applied to the swine vaccine-tostrain EpiCC analysis. A total EpiCC score of at least 0.076 was used to define complete protection; a score between 0.065 and 0.076 was used to define partial protection (light and dark gray circles, respectively, in Figure 2). Based on these protective thresholds, FSXP/IA00 was predicted to confer protection and partial protection against all swine influenza strains except Delta2 and G4. A summary of US swine influenza vaccines EpiCC scores and vaccine coverage can be found in Table 2. Statistical analysis was also performed to evaluate the potential of protection against G4, and FSXP/IA00 was predicted to have lower T cell epitope coverage and was therefore considered not likely to provide protection (p < 0.05; supporting information Figure S1).
Other US commercial swine influenza vaccines had low T cell epitope coverage as well and were predicted to confer no protection.
COBRA vaccines are computationally optimized antigens designed to provide broad antibody epitope coverage against a wide range of variable sequences, but their T cell epitope content is not currently optimized. We considered the T cell epitope coverage of three experimental COBRA vaccines (COBRA/P1, COBRA/SW1, and COBRA/SW2) 17 for the set of US swIAV strains used in this study.
COBRA/SW2 showed the highest T cell epitope relatedness to US field strains, with vaccine T cell epitope coverage of 66.7% (Table 2).
COBRA/SW1 and COBRA/P1 had lower vaccine T cell epitope coverage of 53.3% and 49.7%, respectively. Again, the vaccines were cladespecific: COBRA/SW2 had greater total shared T cell epitope relatedness to field strains from Alpha, Beta, Gamma2, Gamma2-beta-like, and PDM-09 clades ( Figure 2B). It also had less T cell epitope relatedness to Delta-like and Delta2 field strains. In contrast, COBRA/SW1 was predicted to confer protection only against Delta-like swIAV field strains, while COBRA/P1 was predicted to confer only partial protection against Delta-like field strains. These results suggest that similar patterns of T cell epitope relatedness are observed among circulating strains from the same lineage (classical swine influenza versus human seasonal influenza).

| EpiCC analysis of EU vaccines and strains
For EU strains, we established thresholds for full and partial protection using retrospective data. The commercial and experimental inactivated virus vaccine strains used in challenge studies were considered F I G U R E 4 EpiCC analysis of human influenza vaccines against US circulating human IAV and G4 strains (A: human H1N1 seasonal influenza vaccines; B: Human COBRA influenza vaccines). The outer ring surrounding the radar plot shows sequence metadata. The symbol "V" represents vaccine strains, and each vaccine strain is color-coded as indicated in the legend. A protective vaccine efficacy T cell epitope conservation (EpiCC) threshold for human influenza vaccine efficacy has not been defined, as most influenza vaccines generate cross-reactive antibody. Note that EpiCC scores for G4 are extremely low compared to scores for other strains.
protective in our analysis if lung virus titers were significantly lower than the challenge control group and pigs showed no and/or mild clinical signs (low mean temperature or low scores of respiratory diseases) compared with unvaccinated pigs. Partial protection was defined as a significant reduction of lung lesions coupled with a non-significant reduction of lung virus titer when compared with unvaccinated controls.
To define the protective thresholds, we evaluated four published vaccine efficacy studies 33-36 which used four different H1N1 challenge strains. Table 3 lists all the vaccine strains that conferred protection or partial protection against specific challenge strains and the relevant lung titer data. In six of the eight evaluations, protection was demonstrated by the absence of virus, or by significantly lower lung virus titers than the control group. A shared T cell epitope content EpiCC score was calculated for each vaccine-to-strain comparison (Table 3).
Based on these studies, the lowest total EpiCC score that was associated with protective efficacy was 0.0604, defined by comparing T cell epitopes from the swine/Belgium/1/83 challenge strain with the first generation of H1N1 European vaccine strain (NJ76). This threshold is represented by the white area in Figure 3. Based on this study and previous studies, swIAV strains that have total EpiCC scores above the threshold are likely to be protective. Similarly, a total EpiCC score of 0.0474 was associated with partial protective efficacy, as defined for the T cell epitope comparison between NJ76 and challenge strain GT/112/07. For this analysis, therefore, we defined total EpiCC score between 0.0474 and 0.0604 as partially protective.

| Protective efficacy of EU swine influenza vaccines against circulating and G4
Having established estimated vaccine efficacy thresholds for EU swine influenza vaccines, we could evaluate whether the T cell epitope content of additional swIAV vaccine strains that are commonly used in the EU commercial settings (BK00 and HL03) might protect against 150 swIAV strains circulating in the EU and eight emergent G4 sequences (Figure 3).

HL03 had the highest T cell epitope coverage (vaccine-to-strain
EpiCC scores) for circulating swIAV strains in the EU and was predicted to provide protection against 76% of the swIAV strains (total EpiCC scores greater than 0.0604). The BK00 strain had lower T cell epitope relatedness against EU swIAV circulating strains (only 8.7% of EU swIAV circulating strains had total EpiCC score of at least 0.0604).
EpiCC comparisons showed that T cell epitope cross-conservation was lineage specific: HL03's EpiCC score suggested that it may only confer protection against 1C lineages strains. In contrast, vaccine strain BK00's EpiCC score suggests that it may confer protection to field strains related to 1B and other avian or human lineages but not 1C lineages.
Notably, the HL03 vaccine strain had total EpiCC scores that exceeded the defined protective threshold for emergent G4. The average HL03 vaccine strain T cell epitope coverage (slate blue line in Figure 3) for the G4 HA sequences was 65.7%, which is above the protective threshold (Table 4). This observation suggests that existing European swine vaccines may have a protective effect against emergent virus G4.

| Human vaccine-to-strain EpiCC analysis results
To assess whether existing human vaccines would provide protection against potential spillover of G4 into US human population, we evaluated the T cell epitope relatedness of five commercial human H1N1 seasonal influenza vaccine strains to G4 using EpiCC (Figure 4 and Table 5).
One seasonal vaccine strain, BR07, had a distinctive pattern with high T cell epitope content relatedness when compared with Deltalike human IAV circulating strains such as strains that were circulating prior to the 2009 pandemic ( Figure 4A). As expected, CA09, which was introduced in the 2009 pandemic, demonstrates high T cell epitope relatedness when compared with circulating strains of PDM-09 ( Figure 4A). Newer H1N1 vaccine strains that were introduced after the CA09 pandemic also had high EpiCC scores when compared with PDM-09 circulating strains (approximately 70.0% T cell epitope coverage on average). T cell epitope coverage for human vaccine-to-G4-strains comparison was much lower, at 32.4% on average (Table 5).
Unlike the vaccine efficacy studies performed in swine where challenge studies data such as lung virus titers and lung lesion reduction are accessible, human vaccine efficacy is determined via clinical trials. 37 Human vaccine efficacy estimates vary among published efficacy studies, 37 and hence, defining an EpiCC protective threshold is not straightforward. Since the protective thresholds were not defined for human vaccines, we used an average vaccine-to-strain EpiCC scores coverage of 38.5% (represented by the dotted line in supporting information Figure S1) to evaluate the T cell epitope coverage for human vaccines (supporting information Figure S1). The human seasonal influenza vaccines (CA09, MI15, BR18, and GDMN19) showed T cell epitope coverage below the mean and had no significant T cell epitope relatedness to G4 (p < 0.05) when compared with G4.
Three novel subunit influenza vaccines (COBRA HA vaccines) that were designed to generate cross-protective B cell epitopes 17 were also compared with both G4, and circulating human IAV strains, to estimate whether they might also generate protective T memory response against emerging IAV strains. The experimental COBRA Vaccine efficacy threshold as defined in Table 3 that served as a proxy for this analysis.
vaccine strains only showed T cell epitope content relatedness to pre-PDM strains of human IAV ( Figure 4B). COBRA vaccine T cell epitopes were poorly conserved with G4, with an average of only 37.3% T cell epitope coverage, as was observed for H1N1 seasonal influenza vaccines (Table 5).

| T cell epitope conservation among antigens other than HA
Although HA has been the focus of antigenic studies for most influenza vaccines, there is evidence that internal genes such as PB2 and NP may be associated with milder clinical signs and decreased virus shedding. 38  All publicly available non-HA protein sequence data were retrieved for seasonal vaccine strains, G4 and circulating IAV strains.
Given that there is no complete proteome sequence for the US swine influenza vaccine strains, this full proteome analysis was only performed for EU swine and US human IAV strains. There were variations in terms of the degree of T cell epitope conservation among viral antigens (supporting information Figure S2A,B and (average vaccine coverage less than 65%, supporting information Figure S2A and Table S2A).
As compared with EU swine vaccine strains, the internal antigens of the human vaccine strains for the US, particularly CA09 and MI15, had higher T cell epitope conservation with human circulating strains and G4 (supporting information Figure S2B and Table S2B). Internal antigens PB2, PB1, PA, NP, M1, and M2 had total EpiCC scores of at least 0.097, that is, at least 85% T cell epitope coverage (supporting information Figure S2B and Table S2B). In contrast, internal proteins from the vaccine strain BR07 had lower T cell epitope conservation with G4, with T cell epitope coverage ranging between 12.1% and 75.7%. This suggests that the T cell epitopes from the internal antigens of CA09 and MI15 (H1N1/pdm09 lineage) are more highly conserved with T cell epitopes from the internal antigens of G4. While antibodies are usually considered to be the major correlate of protection following influenza vaccination, influenza vaccines containing highly cross-conserved T cell epitopes have also been shown to reduce morbidity and limit spread in the absence of antibodies, even when there is a mismatch between vaccines and emergent strains. [10][11][12] To assess the potential epidemic risk posed by the emergent G4 G4. This study also suggests that the emergent G4 may be a greater threat to the US pork industry than to the EU industry, due to the lack of commercial vaccines that could provide cross-protective immunity to G4. Improving vaccination systems by updating vaccine strains used in pork farms and transitioning to include G4 or EA lineage should be prioritized.

| DISCUSSION
There are limitations in this study that could be addressed in future research. First, while the data subsampling strategy is applied to deal with a large sequence dataset and to avoid overrepresentation of data for certain years or geographical areas, having more data subsampling replicates would better ensure results consistency. Second, the putative cross-protective thresholds determined for the EU influenza vaccines were based on commercial NJ76 vaccine strains and applied for BK00 and HL03 analyses. While protection thresholds for different vaccine strains may vary, having more experimental data available for BK00 or HL03 would help refine the current thresholds used.
As previously mentioned, T cell epitope conservation between circulating virus strains and seasonal vaccines may contribute to the efficacy of existing (human and swine) influenza vaccines. While this study does not absolutely confirm the relevance of the EpiCC tool for the prediction of human and swine influenza vaccine efficacy, a relationship between EpiCC scores and vaccine efficacy is observed and could be used to establish a threshold for vaccine efficacy in the context of European vaccine strains. In a separate study, EpiCC correctly predicted the efficacy of a novel porcine circovirus type 2 (PCV2) viral vaccine against circulating strains of PCV2 in swine. 40,41 That prospective study and this retrospective analysis of G4 influenza serve to illustrate the utility of EpiCC analysis for additional prospective studies of existing vaccine strains against emergent strains.

ACKNOWLEDGMENTS
We acknowledge Matt Ardito for his technical input and support. We also gratefully acknowledge all data contributors, that is, the Authors

PEER REVIEW
The peer review history for this article is available at https://publons.