Genetically barcoded influenza A virus libraries tracked through cell culture, embryonated eggs, guinea pigs, ferrets and mice show that transmission bottlenecks differ by route and recipient, with airborne transmission reducing a diverse inoculum to as few as two founder clones.
Varble A, Albrecht RA, Backes S, Crumiller M, Bouvier NM, Sachs D, García-Sastre A, tenOever BR. Influenza A Virus Transmission Bottlenecks Are Defined by Infection Route and Recipient Host. Cell Host & Microbe. 2014. Volume 16, issue 5, pages 691-700.
DOI 10.1016/j.chom.2014.09.020. PMID 25456074. PMCID PMC4272616.
Genetically barcoded influenza A virus libraries tracked through cell culture, embryonated eggs, guinea pigs, ferrets and mice show that transmission bottlenecks differ by route and recipient, with airborne transmission reducing a diverse inoculum to as few as two founder clones.
Influenza A virus circulates as a genetically diverse population, and the size of the viral population that actually founds a new infection sets limits on how quickly the virus can move through genetic space. Before this study the physiological constraints on that founder population had been inferred largely from sequence surveys of natural infections rather than measured directly under controlled conditions. The authors built a library of more than one hundred influenza A viruses of the 2009 pandemic H1N1 background, each carrying a unique 22 nucleotide barcode in an engineered intergenic region of the NS segment, and confirmed that barcoding did not alter replication. Deep sequencing of the barcode pool then allowed the composition of the viral population to be read out at any point in propagation or transmission. Amplification in canine and human cell lines imposed no detectable bottleneck. Amplification in embryonated chicken eggs collapsed the library to between five and thirteen clones, and hemagglutinin sequencing of the survivors showed convergent changes at residues associated with avian receptor binding, indicating a genetically driven bottleneck during host adaptation. Transmission between susceptible mammals also collapsed the population, but without convergent hemagglutinin selection, and separate recipients exposed to the same donor received different clone sets. Airborne transmission was more restrictive than direct contact. The work separates adaptation bottlenecks from transmission bottlenecks and gives an experimental floor for how few virions initiate a natural influenza infection.
RNA virus populations carry standing genetic variation generated by an error prone polymerase, and that variation is a substrate for adaptation. Bottlenecks that repeatedly sample only a few genomes constrain that substrate, and severe repeated bottlenecks had been shown in other systems to reduce viral fitness. Natural bottlenecks had been described for human immunodeficiency virus, Venezuelan equine encephalitis virus, poliovirus, hepatitis C virus and several plant viruses, and genetic tagging approaches had been used to track poliovirus and Venezuelan equine encephalitis virus populations. For influenza A virus the practical parameters of transmission had been approached from other directions, including estimates of viral copies in exhaled breath and estimates of minimal infectious dose in human volunteers and in ferrets. What remained unresolved, as the paper states, was the physiological dynamics of the viral population during transmission as it occurs in an animal host, including whether observed restrictions reflect genetic selection on particular variants or a sampling process, and whether the restriction acts in the donor or in the recipient.
How severe are the population bottlenecks that influenza A virus encounters during amplification and transmission, and are those bottlenecks driven by genetic selection on particular viral variants or by stochastic sampling, and do they act at the level of the donor or the recipient?
The design rests on making viral genotype trackable without making it consequential. The NS segment was split so that NS1 and NS2 occupy separate reading frames with a noncoding intergenic region between them, following an approach the laboratory had established previously, and a 22 nucleotide GC content matched barcode was inserted into that region. Over one hundred such viruses were rescued individually by reverse genetics in the A/California/04/2009 background, then pooled at equivalent titers. Because the barcodes are intended to be neutral, any change in the barcode distribution between an input population and an output population reports on population sampling rather than on the fitness of the tag. Multicycle growth curves against wild type virus, pairwise comparison of amplified against nonamplified clones, and duplicate sequencing of the same population were used to test that assumption. The library was then passed through a graded series of settings, namely canine and human cell monolayers, embryonated chicken eggs, guinea pig transmission under conditions permitting droplet and aerosol spread, guinea pig cocaging that allowed one donor to seed three recipients, ferret transmission partitioned into direct contact and airborne arms in the same experiment, and mouse infection comparing intranasal instillation against nebulized delivery. Hemagglutinin sequencing was used to distinguish selection on receptor binding from sequence independent sampling. A Monte Carlo simulation was used to ask whether the observed distribution of successful transmissions is compatible with a model in which the only determinant of transmission is a clone's starting proportion in the donor.
Barcode insertion into the split NS segment did not detectably alter replication. Multicycle growth of barcoded A/California/04/2009 matched wild type virus in human lung epithelial cells (Figure 1B), and the pooled library was stoichiometrically balanced with no clone above five percent of the population. Duplicate deep sequencing of the same population returned comparable barcode profiles, which the authors take as evidence that the readout is an accurate surrogate measure of the quasispecies.
Propagation in cell culture imposed no detectable bottleneck. Infection of MDCK cells at low multiplicity gave uniform and reproducible barcode distributions across three independent experiments, and the same was seen in a human lung epithelial line (Figures 2A and 2B).
Propagation in embryonated chicken eggs imposed a severe bottleneck, with only five to thirteen clones recovered per egg and different clone sets emerging in different replicates (Figure 2C). Clones that were and were not amplified showed no intrinsic replication difference when tested individually, so the barcode itself does not explain the outcome.
The egg bottleneck was accompanied by convergent hemagglutinin change. Every virus sequenced from egg passage carried amino acid changes at residues previously implicated in the switch from mammalian to avian receptor specificity, whereas hemagglutinin sequences from MDCK passage showed no divergence from wild type (Figure S2B). A library pre-passaged once in eggs retained about fifty percent of its members on subsequent egg amplification against about ten percent for the original library. The authors read this as selective pressure for entry into the avian host rather than a sampling effect, with the caveat they state that beneficial mutations in other segments cannot be excluded.
Transmission between guinea pigs in neighboring cages imposed a stringent bottleneck. Three of four exposed animals became infected, and those animals carried only two to five clones at day six (Figure 3B), against roughly three quarters of the library recoverable from the nasal wash of inoculated donors.
The transmission bottleneck acts at the recipient and is not explained by viral genetics. When three naive guinea pigs were cocaged with a single donor, all three became infected and all three carried markedly different barcode profiles despite identical exposure (Figures 4C and 4D). The authors note that some selection at the point of secretion cannot be ruled out. Barcode diversity in recipients rose over time to an average of about twenty five barcodes per animal, which is interpreted as ongoing transmission or amplification after initial seeding.
Route of transmission sets bottleneck stringency. In ferrets, contact recipients carried seven to twenty four clones where donors carried seventy one to one hundred (Figure 5B). Airborne transmission occurred in two of three exposed ferrets and reduced the recovered population to as few as two barcodes (Figure 5D). Mice infected with equivalent doses by nebulizer rather than by intranasal instillation likewise showed stronger bottlenecks (Figures S3C to S3E).
Unlike the egg bottleneck, the ferret transmission bottleneck did not show convergent hemagglutinin selection. Viruses from directly inoculated, contact infected and airborne infected animals carried disparate mutations rather than the complete penetrance of particular residues seen in eggs (Figure S3A).
The observed distribution of successful transmissions is statistically consistent with a stochastic bottleneck. Transmission probability correlated positively with a clone's starting proportion in the donor, and when that relationship was fitted and used in a Monte Carlo simulation, the observed distribution fell at the thirtieth percentile of simulated outcomes (Figures 6A to 6C), which the authors take as consistent with initial proportion being the only contributing factor.
Transmitted virus tracks with the upper respiratory tract. Comparing barcode populations in donor nasal wash against donor bronchus tissue, only nasal wash proportions correlated significantly with transmission for both contact and airborne routes (Figure 7). There were eight nasal wash only transmission events and no examples of a bronchus only population transmitting. Simulation using bronchus proportions as the predictor made the observed transmissions unlikely, while nasal wash proportions gave consistent rates.
The study does not establish a single molecular mechanism for the transmission bottleneck, and the authors state their account of it as a hypothesis. What the data do constrain is a separation of two distinct kinds of restriction. During adaptation to a new host, here modeled by a mammalian virus amplifying in the avian egg, the restriction is genetically driven, with viruses that first acquire avian receptor binding changes in hemagglutinin outcompeting the rest of the inoculum. During transmission between already susceptible mammalian hosts, the restriction is sequence independent by the evidence available, since no convergent selection appears in hemagglutinin and since identically exposed recipients receive different clones. The authors propose that donors excrete virus with roughly equal opportunity per virion, that the founder population is therefore drawn in proportion to each clone's abundance at the site of shedding, and that the restriction is imposed at the recipient. The correlation of transmission with nasal wash rather than bronchus proportions is interpreted as indicating the upper respiratory tract as the source of transmitted virus for both contact and airborne routes. The data do not resolve whether an additional selection step occurs at secretion, do not identify a physical or immune barrier in the recipient that performs the sampling, and do not explain how viral fitness is maintained across repeated bottlenecks. The authors raise two alternatives for that last point, namely purifying selection favoring the fittest transmitted viruses, or heterogeneity among hosts such that some individuals impose weaker restriction and act as local reservoirs.
A neutral barcode library in a replication competent influenza A virus, read out by deep sequencing, converts the size and composition of a founder population into a directly measurable quantity in live animals. That makes it possible to compare bottleneck stringency across routes and hosts within a single experimental frame rather than inferring it from consensus sequence divergence. The separation of genetically driven adaptation bottlenecks from sequence independent transmission bottlenecks is the conceptual result, and it reframes the pandemic emergence problem as two sequential and differently governed constraints. The observation that airborne transmission can be founded by very few genomes bears directly on estimates of how likely it is that a multi mutation avian phenotype present at low frequency in a donor would be carried into a new host, and the authors connect this to why an airborne transmissible H5N1 has not emerged despite numerous human infections.
The barcoding approach builds directly on the laboratory's engineered split NS segment, which was developed to allow foreign sequence to be carried in the influenza A virus genome without disrupting NS1 and NS2 expression. That platform recurs across the corpus as a general tool for making the influenza genome carry a readable payload. The study also connects to the laboratory's interest in how the NS segment coordinates the timing of infection through suboptimal splicing, since the one poorly growing barcoded clone is attributed to a possible splicing effect. Category 3 synthesis, across the corpus rather than from this paper alone, is that the recurring strategy is to add a genetically tractable readout to a replicating virus and then let animal physiology rather than cell culture define the result. Confirming that synthesis requires setting this paper beside the other engineered influenza A virus work in the corpus, and it is not asserted from this paper in isolation.
The findings rest on one viral background, the 2009 pandemic H1N1 strain A/California/04/2009, carrying an engineered split NS segment rather than the wild type segment, so generalization to other subtypes and to unmodified viral genomes is an extrapolation. Animal numbers are small, with three or four cages per transmission arm, and the airborne arm produced only two infected ferrets. The barcodes report clone identity and not full genome sequence, so within clone variation and reassortment are invisible to the readout, and only the hemagglutinin segment was sequenced to test for selection. The claim that the transmission bottleneck is sequence independent therefore rests on the absence of convergent hemagglutinin change rather than on genome wide evidence, and the authors explicitly allow that selection at the point of secretion may still occur and that beneficial mutations in segments other than hemagglutinin cannot be excluded in the egg experiments. The limit of detection was set at one hundred reads, which sets a floor on the minority clones that can be counted. The egg experiments model adaptation across a host species barrier in a specific and artificial setting rather than natural avian infection. Guinea pig, ferret and mouse models each recapitulate only parts of human influenza transmission, and no human data are presented. Finally, the stochastic model is a statistical consistency argument, since showing that observed outcomes fall within the range simulated under a proportion only model does not exclude other contributing factors.
Barcoded influenza viruses tracked through animals show that transmission passes only a handful of virus particles, with airborne spread the most restrictive and the recipient imposing the limit.
Influenza A virus circulates as a mixed population, but how much of that mixture survives a transmission event was unclear. Inserting neutral genetic barcodes into more than a hundred otherwise identical viruses allowed the population to be read by sequencing at each step. Cell culture passed the population intact, egg passage selected avian adapted variants, and animal to animal transmission collapsed it to a few founders, with airborne transmission the narrowest route.
Over one hundred influenza A viruses in the 2009 pandemic H1N1 background were rescued individually with unique neutral barcodes in an engineered intergenic region of the NS segment, pooled, and tracked by deep sequencing. Propagation in canine and human cells preserved the population. Propagation in embryonated chicken eggs collapsed it to five to thirteen clones, and all survivors carried hemagglutinin changes at residues associated with avian receptor binding, indicating genetically driven selection during host adaptation. Transmission between guinea pigs and between ferrets also collapsed the population, but without convergent hemagglutinin change, and three recipients exposed to one donor received different clone sets, placing the restriction at the recipient. Airborne transmission was more stringent than direct contact, with as few as two clones establishing infection. Transmission probability tracked with a clone's abundance in donor nasal wash rather than bronchus, implicating the upper respiratory tract as the source of transmitted virus.
lab-led for Varble 2014. McCune 2020 is collaborative and led by the Pfeiffer laboratory at the University of Texas Southwestern, where all mouse work was done, and Muñoz-Moreno 2019 is collaborative and led by the García-Sastre laboratory, with the tenOever contribution in both being the barcoded library method and supervision