Human seasonal influenza under COVID-19 and the potential consequences of influenza lineage elimination

Annual epidemics of seasonal influenza cause hundreds of thousands of deaths, high levels of morbidity, and substantial economic loss. Yet, global influenza circulation has been heavily suppressed by public health measures and travel restrictions since the onset of the COVID-19 pandemic. Notably, the influenza B/Yamagata lineage has not been conclusively detected since April 2020, and A(H3N2), A(H1N1), and B/Victoria viruses have since circulated with considerably less genetic diversity. Travel restrictions have largely confined regional outbreaks of A(H3N2) to South and Southeast Asia, B/Victoria to China, and A(H1N1) to West Africa. Seasonal influenza transmission lineages continue to perish globally, except in these select hotspots, which will likely seed future epidemics. Waning population immunity and sporadic case detection will further challenge influenza vaccine strain selection and epidemic control. We offer a perspective on the potential short- and long-term evolutionary dynamics of seasonal influenza and discuss potential consequences and mitigation strategies as global travel gradually returns to pre-pandemic levels.

S easonal influenza viruses evolve to evade pre-existing immunity and gain competitive advantage via surface protein mutations which yield new antigenic variants 1 . Natural selection acts on a global scale due to rapid and widespread global circulation 2 . This effectively eliminates previously dominant antigenic variants and results in limited circulation of antigenically similar viruses within each subtype/lineage at a given point in time. However, the pace of antigenic selection varies over time for influenza A virus (IAV) subtypes and influenza B virus (IBV) lineages due mainly to population-level fluctuations in immune pressure. This confounds vaccine strain selection, which relies on the prediction of antigenic evolution 3 . To facilitate biannual selection of candidate vaccine viruses, the WHO Global Influenza Surveillance and Response System (GISRS) coordinates influenza surveillance from 138 National Influenza Centers (NICs) and diagnostic and reference laboratories in 108 countries 4 . Current seasonal vaccine formulations are either trivalent or quadrivalent, with either three or four representative strains including IAV subtypes A(H1N1) and A(H3N2) and either one or both IBV lineages, B/Victoria and B/Yamagata. Seasonal influenza viruses exhibit stronger seasonal cycles in temperate zones, with surges of infections in winter. Seasonal trends are weaker in tropical zones, with increased circulation evident in both the rainy season due to increased humidity and in cooler, drier months 5,6 . Seasonal temperate cycles are maintained through continuous reintroduction from tropical regions and opposing hemispheres, causing local transmission chains to emerge and perish in community settings 7,8 . Transmission chains arising from a single introduction (transmission lineages 9 ) dissipate at a greater frequency outside of peak seasonal circulation, although some may persist from one season to the next 2, 10 . In tropical regions, influenza viruses exhibit more complex multipeak dynamics, impacted by patterns of global circulation and evolution 11 . The interplay between the different seasonal influenza virus subtypes and lineages varies temporally and geographically, leading to significant variation in population immunity to each influenza virus.
Analysis of global sequence data has shown that (i) tropical and subtropical regions in Asia sustain transmission lineages for a longer duration than temperate regions, providing more opportunities for antigenic drift 2 , and (ii) A(H3N2) lineages do not generally persist between seasons in temperate regions but are reseeded annually 7,8,12 , whereas transmission lineages of A(H1N1), B/Yamagata, and B/Victoria can circulate for several years in temperate and sub-tropical regions 2 . Population density and regional interconnectedness play an important role in maintaining viral metapopulations 8,12 . However, the genetic and antigenic diversity of seasonal influenza has been severely impacted by dramatic changes in global migration and travel since the onset of the COVID-19 pandemic in March 2020 ( Fig. 1).
Since April 2020, most countries have seen historically low seasonal influenza virus circulation 13,14 attributable to nonpharmaceutical interventions (NPIs) such as travel restrictions, quarantine on arrival, social distancing, school and workplace closures, mask wearing, surface disinfection, and enhanced hand hygiene. NPIs have similarly disrupted the circulation of other common respiratory viruses such as respiratory syncytial virus and human metapneumovirus [15][16][17][18] by limiting opportunities for reintroduction and local transmission. Prolonged suppression of seasonal influenza virus circulation, compounded by regional inequities in vaccine distribution 19 and potential vaccine complacency, supply chain disruptions and misinformation 20 amid fewer cases, will reduce population immunity and increase severity of future influenza virus epidemics. Accumulating evidence indicates protection against influenza infection, acquired through infection or vaccination, wanes over the course of a single season 21 . At the individual level, circulating antibodies decline over six months 22 , and the half-life of T-cells for cellular responses lasts eight to 14 years 23 . Accumulation of susceptible individuals during milder seasons results in more intense subsequent seasonal epidemics 24 . The consequences may be most dire for children with lack of exposure to influenza, as immunological imprinting (also referred to as "original antigenic sin") during childhood influenza A and B infections [25][26][27] impacts patterns of susceptibility and circulation in subsequent years 18 . Epidemiological studies, corroborated by multiple modeling and immunological studies 27,28 , show lifelong immune memory to first childhood influenza infection confers lifelong homosubtypic protection at the cost of heterosubtypic protection. Prolonged suppression of seasonal influenza circulation during the 2020s will lead to greater susceptibility in this birth cohort due to lack of exposure by natural infection.
As COVID-19 vaccination rates increase in the coming months, the use of NPIs to limit transmission will gradually decline. Domestic and international travel will eventually return to pre-pandemic levels 29 , enabling a resurgence of influenza virus circulation. Through phylogenetic analysis of available influenza sequence data and case reports submitted to WHO GISRS we consider the short-and long-term implications of COVID-19 control measures on the epidemiology and evolution of seasonal influenza viruses.

Results
A global reduction in seasonal influenza virus case detection. Analysis of the GISRS FluNet database 4 to 1 August 2021 shows an unprecedented global reduction in seasonal influenza cases since the beginning of the COVID-19 pandemic in March 2020 (Fig. 2). Routine influenza testing was disrupted during the initial stages of the pandemic amid the high demand for SARS-CoV-2 testing. Nevertheless, many countries continued or resumed influenza testing and reporting by mid-2020 15   between 1500 and 3500 positive specimens per week, but the 2020 season was notably absent and the expected rise in seasonal influenza cases has yet to occur in 2021. Remarkably, <12 influenza positive cases per week were reported from May 2020 to July 2021 in the Southern Hemisphere (Fig. 2).
Reduction in seasonal influenza virus diversity and the likely elimination of B/Yamagata. Co-circulation of diverse A(H3N2) and B/Victoria viruses and antigenic drift within some cocirculating clades have necessitated frequent updates to the vaccine strain components in recent years. Since their emergence in 1968, A(H3N2) viruses have, on average, evolved distinct antigenic variants every 3-7 years with rapid elimination of previous antigenic variants 32,33 . However, leading up to the COVID-19 pandemic, a major A(H3N2) genetic bottleneck had not occurred for a number of years 34 (Figs. 1 and 3). The continued circulation of an A(H3N2) clade 3c3.A, a lineage which dates back to 2013, has been implicated in reduced production of neutralizing antibodies in adults with childhood exposure to A(H3N2) 28 , and it has been hypothesized that further accumulation of antigenic changes may result in A(H3N2) divergence 35 Fig. 1), while clade A1b/94N was detected across Asia and Oceania ( Supplementary Fig. 2). In particular, A1b/94N viruses were frequently detected in India since May 2021. The phylogeny of clade A1b/94N reveals six related clusters that originated independently prior to March 2020. First detected in Cambodia, one lineage circulated in the Greater Mekong subregion of Southeast Asia from July to February 2020. A second distinct A1b/94N lineage was detected in the Australian Northern Territory from individuals returning to Australia and in quarantine during February-March 2021 and from neighboring Timor-Leste during July 2020-March 2021, suggesting regional circulation during 2020/2021 ( Supplementary Fig. 2 Few A(H1N1) cases have been detected since April 2020 (Fig. 2), mainly in Ghana (n = 235), Togo (n = 226), the United States (n = 170), and Russia (n = 165) (Supplementary Dataset 1). Nevertheless, the 254 available sequences in GISAID reflect cryptic circulation of all previously circulating A(H1N1) clades into early 2021. Three independent lineages of clade 6b1.A/187A viruses circulated in West Africa (Ghana, Nigeria, and Togo) during 2020, along with a few 6b1.A/156K and 6b1.A/183P-5a viruses ( Fig. 4 and Supplementary Fig. 3). Since May 2021, clade 6b1.A/156K viruses were primarily detected in India, while the other A(H1N1) clades in circulation have been detected sporadically around the world (Fig. 4).
An  Fig. 4)  sequences from Cameroon, Niger, Nigeria, and the Democratic Republic of Congo to January 2021 cluster by country, suggesting containment of virus circulation within each country (Supplementary Fig. 1). In contrast, most Southeast Asian nations maintained relatively stringent domestic and international travel restrictions during 2020/2021, except for Laos and Cambodia, where COVID-19 suppression was followed by A(H3N2) influenza virus outbreaks in 2020. While international travel measures impacted influenza virus migration patterns, an analysis of control-measure stringency and influenza case reports in fourteen countries in Asia and Africa with substantial influenza activity showed no correlation between the stringency of public health interventions and domestic transmission of influenza (see Methods section, Supplementary Fig. 6). A caveat to note is that the stringency index does not reflect the efficacy of control measures or population compliance.

Discussion
Since the start of the COVID-19 pandemic, WHO influenza surveillance data reflect a substantial reduction in global influenza   virus circulation. Lack of exposure to influenza will lower population immunity and increase the severity of large epidemics upon a future global resurgence. Notably, countries in North America and Europe with strong influenza surveillance have only sporadically reported the influenza viruses in circulation, including several that have caused outbreaks in Africa and Asia, and B/Yamagata lineage viruses appear to have become extinct around mid-2020. Despite an overall increase in influenza surveillance, WHO reports 39 variable effects due to COVID-19. Influenza surveillance benefited due to rapid capacity building and training efforts to respond to SARS-CoV-2, however disruptions occurred at national or regional levels due to healthcare resource allocation and health care-seeking behavior. Generally, demographic details such as age are not available for all countries, many cases are missed due to timing of infection, and severe cases requiring primary care or hospitalization are more likely to be detected and reported. Furthermore, reference laboratories use various detection methods and may only submit a representative subset of their surveillance data alongside any cases that cannot be subtyped by conventional methods. Not all cases are confirmed by isolation or genomic characterization, and of those sequenced, only a portion are submitted to sequence databases such as GISAID. However, to offset issues with completeness of submitted record 40 , our analysis was limited to the primary data fields, and regional circulation was inferred using a combination of case numbers and virus genetic relationships.
Roughly one-quarter or more of seasonal influenza cases are caused by IBVs 41 , and in recent decades the two IBV lineages have caused comparable proportions of influenza cases. Historically, B/Yamagata viruses have caused a greater rate of infection in temperate regions and have infected adults at a greater rate than children, whereas B/Victoria viruses have infected more children than adults 41 . However, the long-term impact of B/ Yamagata elimination on the evolutionary dynamics of IBV is uncertain. Recently, Vieira et al. 24 examined historical patterns of IBV lineage frequencies in New Zealand using statistical modeling and showed that fluctuations in lineage dominance and lineage cross-protection explains contrasting age distributions of B/ Yamagata versus B/Victoria lineages. As IBV lineages offer some cross-protection 42,43 , the extinction of B/Yamagata will leave a higher proportion of individuals susceptible to IBV, enabling faster B/Victoria antigenic evolution.
It is important to note, the threat of re-introduction of apparently extinct influenza virus lineages could still pose a risk in coming years, as happened with the reemergence of A(H1N1) in 1977 44 following a 19-year hiatus since the 1958 A(H2N2) pandemic (Fig. 6). If B/Yamagata does not reemerge in the next year or so, it may need to be treated as a high consequence pathogen to prevent reintroduction, similar to A(H2N2) viruses which have not circulated since 1968 and are now held and handled in the higher level BSL-3 laboratory biosecurity levels 45 . Future B/ Yamagata positive samples will require urgent confirmation and characterization to be able to better determine the mechanisms that could sustain such low levels of virus circulationfor example, immunocompromised individuals can carry infection for several weeks or months and potentially accumulate additional mutations 46-48or to rule out the possibility that these were in fact false-positive test results.
Although two IAV subtypes and two IBV lineages have cocirculated in recent decades, prior to the re-emergence of A(H1N1) in 1977, only a single IAV subtype and a single IBV lineage circulated among humans (Fig. 6). In the early 1980s, IBV diverged from the ancestral B/Lee lineage into two antigenically distinct lineages 49 . The survival of two IBV lineages is attributed to the geographic isolation of B/Victoria in China in the 1990s, followed by a global resurgence during 2000-2002 50 . The continued endemicity of geographically disparate transmission lineages of A(H3N2), A(H1N1), and B/Victoria (compounded by limited availability of clinical isolates) confounds the accuracy of candidate vaccine virus selection, and further accumulation of antigenic changes could lead to long-term co-circulation of antigenically distinct lineages, as occurred for IBV. However, the concomitant reduction in population-level immunity towards seasonal influenza suggests global resurgence of any residual viruses could occur in the future and continued vigilance is required.
The emergence of pandemic influenza viruses A(H2N2) in 1958 and A(H3N2) in 1968 from animal reservoirs resulted in the rapid and complete elimination of previously circulating seasonal influenza A subtypes (Fig. 6) in part due to a lack of pre-existing immunity, which enables novel strains to out-compete their predecessors. However, while the recent emergence of the 2009 A(H1N1) pandemic virus caused elimination of A(H1N1)77 viruses (last detected in 2010; GISAID), seasonal A(H3N2) and IBV viruses sustained transmission throughout the pandemic. Mild and short-lived NPIs temporarily suppressed circulation of A(H3N2) and IBVs, and a combination of immune-driven selection and relatively slower antigenic evolution of A(H1N1)77 36 likely contributed to its elimination. By 2011, seasonal circulation of all four subtypes had resumed, though A(H3N2) evolutionary patterns were significantly altered following co-circulation with 2009 A(H1N1) viruses 35 .
A previous analysis of global sequencing data highlighted the propensity for sub-tropical regions in Asia to sustain transmission lineages and act as source populations in the emergence of influenza antigenic variants 2 . However, limited sequence and surveillance data were available from Africa at that time. Surveillance capacity in West Africa has since increased with direct support from the WHO and US CDC. In the context of pandemic disruptions to influenza circulation, surveillance in West Africa highlights the potential importance of this region for sustained transmission of influenza and suggests that this region may play a key role in the circulation and maintenance of seasonal influenza lineages along with larger population centers located in India, China, and Southeast Asia. Furthermore, recent studies that showed correlation between stringency and national-level influenza transmission were from countries with no influenza activity since pandemic emergence 31,[51][52][53] , with the exception of Cambodia that showed A(H3N2) activity during 2020 54 . Our study in contrast analyzed countries with significant influenza activity to find no correlation between the stringency of community measures and domestic influenza transmission indicating effectiveness of community measures in these countries was low.
We speculate that heterogeneity in COVID-19 vaccination rates and control policies will slow the global resurgence of influenza, delay competition among existing influenza lineages and enable further divergence of spatially separated lineages, but these individual influenza lineages will eventually expand, compete, and once again circulate more widely. Upcoming influenza seasons could therefore be compounded in severity as immunity wanes over time for all age groups 24 . Moreover, the continued evolution of regionally distinct lineages increases the risk that the antigens included in the vaccine will not be representative of the viruses that ultimately circulate, thereby reducing vaccine effectiveness.
Knowledge gained from influenza epidemiology and evolution under COVID-19 epidemic control underscores the importance of heightened vigilance and continued influenza vaccination programs as we emerge from the COVID-19 pandemic, as well as the potential consequences of recent changes in seasonal influenza virus lineage diversity. Based on observed genetic diversity and endemicity of circulating lineages, continued travel restrictions will limit the number of regional introductions, and prolonged pandemic mitigation strategies could further impact future seasonal influenza virus circulation and evolution. Ongoing global COVID-19 vaccination rates indicate that middleincome countries may be sufficiently vaccinated by the start of 2022; thus, continuation of mitigation strategies may become impractical, and global travel could return to pre-COVID-19 levels in the near future. As international travel is important for sustaining seasonal influenza transmission 36,55 , genomic surveillance at border crossings (using the infrastructure developed for COVID-19) could monitor importation from regions that maintain endemic circulation of seasonal influenza. As illustrated by influenza sequence and surveillance data from 2020 and 2021, East, South, and Southeast Asia have had sustained A(H3N2) and B/Victoria transmission lineages, and West Africa has maintained A(H1N1) circulation.
The uncertainty in future seasonal influenza circulation provides further incentive for rapid advancement of universal influenza vaccines that confer broad protection against multiple IAV or IBV lineages [56][57][58] . Indeed, the mRNA vaccine technology used against COVID-19 could be rapidly produced, modified, and deployed 59,60 with the potential to alleviate many of the concerns presented in this manuscript. Ultimately, regardless of the influenza vaccine technologies deployed and their coverage, surveillance remains the key to better understanding and controlling influenza infections in the immediate future.

Methods
Epidemiological trends of seasonal influenza-positive cases and samples tested between January 2015 and July 2021 ( Fig. 2 and Supplementary Fig. 6) were inferred from influenza notifications submitted to the WHO Global Influenza Surveillance and Response System (GISRS) 4 , obtained using FluNet-Scraper (https://github.com/MagnusBook/flunet-scraper). All human seasonal influenza hemagglutinin (HA) sequences collected from December 2018 to July 2021 were downloaded from GISAID (Supplementary Dataset 4) and aligned by HA subtype/ lineage using MAFFT v.7.22 61 . Preliminary maximum-likelihood phylogenies were estimated with FastTree v.2.1 62 . Root-to-tip regression analyses of phylogenetic branch lengths and sampling dates were used to control phylogenetic data quality in TempEst v.1.5.3 63 , and sequences <900 nt were excluded. After adding HA reference sequences (recommended vaccine strains from 2010 to 2021), the final dataset included 15,526 A(H3N2), 16,020 A(H1N1), 9,743 B/Victoria, and 1029 B/ Yamagata sequences.
Phylogenetic relationships and divergence times of seasonal influenza HA genes were estimated using IQ-TREE v.2 64 and the least-square dating method 65 . Largescale maximum likelihood analyses using all available HA sequence data were generated by FastTree v.2.1 62 with the generalized time reversible nucleotide substitution model. Branch support was assessed by Shimodaira-Hasegawa test 66 , and lineages were labeled according to WHO clade designations. Residual influenza virus lineages were estimated by counting individual monophyletic clades that derived from branches prior to March 2020. The R package 'ggstream' v.0.1 was used to map temporal changes in sampling of seasonal influenza clades, and 'rworldmap' v.1.3 was used to plot world maps.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The seasonal influenza gene sequences and associated metadata utilized in this study were downloaded from GISAID (accession numbers and acknowledgements are provided in (Supplementary Dataset 4). Details of confirmed influenza cases are available from the web based tool for influenza virological surveillance FluNet (https://www.who.int/tools/flunet).

Code availability
Code used for the above analysis is available on https://doi.org/10.5281/zenodo.6321464.