Selective pressure mediated by influenza virus M158–66 epitope-specific CD8+T cells promotes accumulation of extra-epitopic amino acid substitutions associated with viral resistance to these T cells

Highlights • Use of a unique co-culture system to study influenza virus evolution in HLA-A*02:01 transgenic A549 cells under selective pressure by cloned human T cells specific for the conserved M158–66 CTL epitope.• Upon serial co-culture passages, virus evolution was assessed by next generation sequencing using molecularly cloned recombinant viruses containing the M1 gene of an avian or a human influenza A virus.• Accumulation of previously described extra-epitopic amino acid substitutions associated with reduced recognition by M158–66 –specific CD8+ T cells is promoted by selective pressure mediated by these cells.


Introduction
Influenza viruses are known for their capacity to evade host immunity induced by previous infections.For example, these viruses accumulate amino acid substitutions in proximity to the receptor binding site of the viral hemagglutinin (HA), allowing them to evade recognition by virus-neutralizing antibodies induced in the human population by infections with ancestral viruses and/or vaccination (Rambaut et al., 2008;van de Sandt et al., 2012).Selective pressure by these antibodies drives this antigenic drift, which necessitates updating the composition of influenza vaccines almost annually to ensure that the vaccine strains match the circulating epidemic strains antigenically, and optimal effectiveness of the vaccines (Koel et al., 2013;Smith et al., 2004).In addition, new strains of influenza A virus (IAV) with novel antigenic properties can emerge after the reassortment of gene segments between IAVs of human and animal origin or following zoonotic transmission (Krammer et al., 2018).Virus-neutralizing antibodies induced by infection with seasonal influenza viruses afford little or no protection against these novel strains, increasing their pandemic potential (Kim et al., 2018).
Virus-specific CD4 + and CD8 + T cells contribute to protective immunity by releasing cytokines and recognizing and eliminating virusinfected cells (Jansen et al., 2019).It has been shown that the presence of pre-existing cross-reactive CD8 + T cells, mainly directed to the conserved nucleoprotein (NP) and the matrix-1 protein (M1) induced after infection with seasonal influenza viruses, limit severity and duration of disease caused by pandemic influenza A viruses (Gras et al., 2010;Sridhar et al., 2013;Hayward et al., 2015).
Although some IAV CTL epitopes display variation, others are conserved, like the HLA-A*02:01 restricted M1 58-66 epitope.Despite the high prevalence of HLA-A*02:01 in the human population and the immunodominant nature of the epitope, it remained conserved in seasonal IAVs despite strong immune pressure by specific CD8 + T cells.It was shown that the epitope does not tolerate amino acid substitutions and is under functional constraints (Berkhoff et al., 2005), which is explained by overlap with a nuclear export signal of the M1 protein (Terajima and Ennis, 2012;Cao et al., 2012).Of particular interest, it was recently demonstrated that M1 58-66 -specific CD8 + T cells displayed reduced lytic activity and delayed activation after stimulation with M1 protein derived from human seasonal IAV compared to stimulation with M1 protein derived from an avian influenza virus (van de Sandt et al., 2016).Also, recombinant influenza virus carrying the M1 gene of a human influenza virus displayed reduced replication than those with an M1 gene of an avian influenza virus in the presence of M1 58-66 -specific CD8 + T cells (van de Sandt et al., 2018a).It was hypothesized that the human seasonal influenza viruses display signs of immune adaption and that extra-epitopic amino acid residues play a role in this phenomenon.Five of these residues were found to be different between human and avian IAVs (I15V, K27R, K101R, V115I, T121A), affecting antigen processing and presentation of the M1 58-66 epitope (van de Sandt et al., 2016;van de Sandt et al., 2018a).
In the present study, we hypothesized that CD8 + T cells directed to the HLA-A*02:01/M1 58-66 epitope exert selective pressure and drive the accumulation of extra-epitopic amino acid substitutions associated with reduced activation of these T cells in the M1 protein originating from an avian influenza virus.To this end, isogenic influenza viruses with M1 genes from a human or an avian virus were serially passaged in A549 cells that express HLA-A*02:01 or not, in the presence or absence of M1 58-66-specific CD8 + T cells.Viral fitness and nucleotide sequences were analyzed of the progeny virus, we showed that influenza virus M1 58-66 epitope-specific T cells exert selective pressure and promote the prior described accumulation of extra-epitopic amino acid substitutions (van de Sandt et al., 2016;van de Sandt et al., 2018a).
One day before each experiment, the cells were seeded at 0.1 × 10 6 / well in a 48well plate (Corning®) without puromycin.Expression of HLA-A*02:01 was confirmed by staining with a FITC-labelled antibody directed to the HLA-A*02 molecule (FITC Mouse Anti-human HLA-A2; BD Pharmingen; 1:50), and analysis was done by flow cytometry (BD FACS Fortessa; FACS Diva software).
Madin-Darby canine kidney (MDCK) cells were cultivated in DMEM (Dulbecco's Modified Eagle Medium; Gibco™) supplemented with P/S, G, NEAA (all at 1 % v/v, Thermo Fisher), 10 % FBS and they were seeded into 96 well plates one day prior to use in infection experiments.
After the incubation, the supernatant was transferred into a 1.5 ml Eppendorf tube, and cells were removed by centrifugation at 300xg for 10 min at 4 • C. The subsequent blind passages were performed using 250 µl/well of supernatant from the previous passage.The remaining supernatants were stored at − 80 • C until further usage.In total, 10 serial passages were carried out.

Virus titration
One day before virus titration, 20.000/well MDCK cells were seeded into 96 well plates.A 1:10 dilution series of each sample was prepared in DMEM (Dulbecco's Modified Eagle Medium; Gibco™) supplemented with P/S, glutamine, non-essential amino acids (all at 1 % v/v, Thermo Fisher), 2 % sodium-bicarbonate and 0.1% BSA and Trypsin-TPCK (1 μg/ ml).After washing the cells twice with DPBS, 200μl/well of the dilution series was added.The plates were incubated for 72 h at 37 • C and 5 % CO2 before collecting the supernatant, which was tested for hemagglutinating activity.

Hemagglutination assay
Fifty μl of each sample was transferred into v-bottom plates before adding 50μl/well of 1 % turkey red blood cells (RBC).After 30 min incubation on ice, the assay was visually examined, and the highest dilution in which agglutination occurred was used to calculate the 50 % tissue culture infection dose (TCID50).

Isolation of viral RNA and RT-PCR
Viral RNA was isolated from cell culture supernatants using the QiaAmP Viral RNA kit (Qiagen) following the manufacturer's protocol.The Superscript IV First Strand Kit (Thermo Fisher Scientific) was used for cDNA synthesis, following the manufacturer's protocol using the following primer (5′ AGCAAAAGCAGG).Subsequently, PCR was performed using M1-specific primers (M1_fw 5′ GCAGGTAGATGTTGAAA-GATG; and M1_rev 5′ CGCTGCCTGCTCACTTGATC) to generate amplicons spanning the first 600 bp of the M1 gene according to the protocol of the manufacturer of AmpliTaq Gold® DNA Polymerase (ThermoFisher Scientific).The PCR program consisted of 39 cycles of 30 s at 95 • C, 30 s at 45 • C, and 3 min at 72 • C.

Agarose gel electrophoreses and gel extraction
PCR products were loaded on a 1.3% agarose gel, visible amplicons of 600 bp were cut from the gel, and the DNA was extracted using the GeneJet Gel Extraction Kit (Thermo Fisher Scientific).The samples were eluted in 30 µl of elution Buffer and stored at − 80 • C until further analysis.

NGS library preparation
NEBNextra Ultra II kit (New England Biolabs) was used for NGS library preparation following the manufacturer's protocol.To 100 ng in 25 µl of the PCR products, 1.5 µl Enzyme Mix and 3.5 µl reaction buffer were added and then placed into the thermocycler (20 • C for 30 min, 65 • C for 30 min, 10 • C until taken).15 µl ligation master mix, 0.5 µl enhancer, and 1.25 µl adaptor were added to the samples and then incubated for 1 h at 20 • C. 1.5 µl of USER enzyme was added to the ligation mix and then placed in the thermocycler (37 • C for 20 min, 10 • C hold).Afterward, the samples were purified.50 µl sample were added to 43 µl AmpureBeads.After 4 min of binding at RT, the tubes were put into a magnetic rack and washed twice with 80 % EtOH.The sample was taken off the rack, resuspended in 17 µl low TE buffer, and binding was allowed for 2 min.The tube was placed on the rack again to place the supernatant into a fresh tube.12.5 µl Ultra II Q5 Master Mix and 5 µl of a unique barcode was added to 7.5 µl of the purified samples and then run in the thermocycler following the manufacturer's protocol.The samples were purified again, using 22.5 µl AmpureBeads/sample and washing 3x with 80 % EtOH.The pellet was resuspended in 25 µl, transferred into a fresh tube, and its concentration was measured using a Qubit (Thermo Fisher).Next Generation Sequencing was performed using the Illumina MiSeq system at the Research Core Unit Genomics (RCUG) of the Hannover Medical School.

NGS analysis using Geneious
Sequencing on the Illumina MiSeq system with a paired-end sequencing strategy was performed to achieve a target depth of at least 1000x coverage per amplicon.The sequencing run generated ~250 bp reads.Paired-end reads were set using Geneious Prime software (2022.2.1; Dotmatics).Then, low-quality reads were removed, and the sequences were trimmed using the BBDuk, with a quality score cut-off of Q20.After merging the paired reads, the De Novo Assemble feature generated contingents of similar sequences.Contingents were then aligned to the reference sequence of the M segment of A/Netherlands/ 178/1995 (H3N2) or avian IAV A/Vietnam/1194/2005 (H5N1), and variant calling was performed using the Geneious Prime Software.Variants were filtered based on a minimum cut-off of 0.05 % and a minimum depth of 100x to ensure high-confidence variant calls.Statistical analyses were performed using Geneious Prime Software.

Serial passage of virus in A549 tg-A2 cells in the presence of M1 58-66specific CD8 + T cells
To assess selective pressure exerted on the M1 58-66 epitope, a viruspassaging system was established in co-cultures of A549 cells and M1 58-66 -specific CD8 + T cells, which has been described previously (van de Sandt et al., 2018a).To optimize conditions for serial passaging of the virus, we tested various MOIs of the virus carrying the human M1 gene in A549 wt and A549 tg-A2 cells in the presence of a fixed number of M1 58-66 -specific CD8 + T cells.Only inoculation of A549 cells at MOI 1 resulted in virus titers high enough for subsequent passaging.In contrast, with MOI 0.01, no infectious virus was detected 24 h post-infection (data not shown).Next, the optimal number of M1 58-66 -specific T cells was determined for serial passaging of the virus.Both A549 wt -and A549 tg-A2 cells were infected with the IAVs with the M1 gene of a human (WSN-M-hH3N2) or an avian (WSN-M-aH5N1) IAV.As expected, no differences in virus production could be detected using A549 wt cells in the presence or absence of M1 58-66 -specific T cells since the M1 58-66 peptide cannot be presented by these cells (Fig. 1A and 1C).Virus WSN-M-aH5N1 replicated faster during the first 24 h than WSN-M-hH3N2, but both reached similar virus titers 48 h post-infection.However, in A549 tg-A2 cells, a reduction in virus titers was observed with increasing numbers of M1 58-66 -specific T cells (Fig. 1B and 1C).For example, a 10,000-fold reduction was observed with 100,000 M1 58-66 -specific T cells per well 72 h post-infection (Fig. 1B).With 50, 000 T cells per well, virus titers between 10 3.5 and 10 4.5 TCID 50 /ml were observed 48 h post-infection.Additionally, this number resulted in a significant reduction of viral replication (Fig. 1C) and was selected for subsequent passaging experiments.Forty-eight hours post infection, replication of virus WSN-M-hH3N2 with a M gene segment of a human IAV was 6.7-fold higher in the presence of 12,500 M1 58-66 -specific T cells, compared to replication of virus WSN-M-aH5N1 with an M gene segment of an avian IAV.These data confirms that the virus with the M1 gene of a human seasonal influenza virus shows signs of immune adaptation and is more resistant to the action of M1 58-66 -specific T cells, as was demonstrated previously (van de Sandt et al., 2016; van de Sandt et al., 2018a) (Fig. 1B).
Using the optimized conditions, WSN-M-aH5N1 and WSN-M-hH3N2 viruses were serially passaged ten times in A549 wt and A549 tg-A2 cells.The resulting viral titers were determined for the first four passages.In the absence of HLA-A*02:01 expression, the presence of M1 58-66 -specific T cells did not affect the production of both viruses (Fig. 1D).In contrast, passaging of the viruses in the presence of 50.000M1 58-66specific T cells in A549 tg-A2 cells resulted in 150-and 60-fold mean reduction in virus replication during four passages of WSN-M-aH5N1 and WSN-M-hH3N2, respectively (Fig. 1E).This again reflects the relative resistance of WSN-M-hH3N2 to the action of M1 58-66 -specific T cells.The virus titers of WSN-M-aH5N1 and WSN-M-hH3N2 remained at the same level during subsequent passages, and we passaged the viruses further up to passage 10.Culture supernatants were collected, and selected samples were submitted for NGS analysis.

Next generation sequencing to screen for adaptive mutations
The samples obtained from passaging were subsequently prepared for NGS.Due to the limitation of read length to 600 base pairs, only the initial 600 base pairs of the M1 protein gene, which contain the coding region of the extra-epitopic positions of interest (human (M1-H3N2): Val15, Arg27, Arg101, Ile115, Ala121; avian (M1-H5N1): Ile15, Lys27, Lys101, Val115, Thr121), were amplified by PCR.Following the preparation of the NGS libraries, progeny virus samples collected after passages 1, 5, and 10, were sequenced and subjected to data analysis.Specifically, the sequences were examined for conversion at the previously mentioned positions encoding the avian-to-human amino acid substitutions I15V, K27R, K101R, V115I, T121A, and vice versa (van de Sandt et al., 2016;van de Sandt et al., 2018a) (Fig. 2A).All mutations of interest were found in both viruses passaged in A549 WT or A549 tg-A2 cells, irrespective of the presence of M1 58-66 -specific T cells.A distinct pattern of codon usage was found for all five amino acid substitutions of interest, that was statistically significant (variant p-values <10 − 27 ) (Fig. 2B).The occurrence of all five avian-to-human mutations (Fig. 2A) were detected at significant frequencies starting from passage 5 in both A549 WT or A549 tg-A2 cells.Interestingly, no other changes in nucleotide sequence that could lead to the same amino acid changes were found at a statistically significant frequency.
All five mutations relevant for T cell evasion, as well as other distinct avian-to-human mutations such as T137A and the A166V, T167A, and I168T motif (data not shown) occurred at similar rates in all samples (Fig. 1C).We observed an overall higher rate of sequence conversion, resulting in avian-to-human amino acid substitutions than vice versa (Fig. 2C).By passage 10, the average rate of changes from the avian-tohuman amino acid sequences in A549 WT cell reached 28.4 % and 15.7 % in the absence and presence of M1 58-66 -specific T cells, respectively.Interestingly, the highest avian-to-human conversion rate (93.2 %) was observed in the sequence of progeny virus obtained at passage 10 from A549 tg-A2 in the presence of M1 58-66 -specific T cells, while the lowest average conversion rate (1.2 %) was found in A549 tg-A2 in the absence of M1 58-66 -specific T cells.The highest proportion of mutations from human-to-avian signature (4 %) was observed in passage 5 (A549 tg-A2 /TC + ), but these were not fixed and did not persist until passage 10 (Fig. 2C).

Discussion
In the present study, we tested whether M1 58-66 -specific CD8 + T cells could exert selective pressure and thus drive the accumulation of extraepitopic amino acid substitutions associated with reduced recognition of the M1 protein of seasonal human influenza viruses by these T cells, compared to M1 of avian viruses (van de Sandt et al., 2016;van de Sandt et al., 2018a).The data from the present study confirmed that indeed, M1 58-66 -specific CD8 + T cells control replication of a virus with the M1 gene of an avian virus better than one with the M1 gene of a seasonal human influenza virus (van de Sandt et al., 2018a).
We observed a relatively low rate of accumulation of all five avianto-human amino acid substitutions associated with reduced recognition by M1 58-66 -specific CD8 + T cells in the absence of selective pressure by these cells.However, under conditions allowing the M1 58-66 -specific CD8 + T cells to exert selective pressure, we observed levels of avian-tohuman amino acid substitutions close to fixation after ten serial passages of the WSN-M-aH5N1 virus.
In order to test the selective pressure of M1 58-66 -specific CD8 + T cells in vitro, serial passaging of the virus in co-cultures of A549 cells and cells of an M1 58-66 -specific CD8 + T cell clone was established.Conditions were selected at which inhibitory effects on virus replication were apparent but still allowed sufficient replication of WSN-M-hH3N2 and WSN-M-aH5N1 to reach viral titers high enough for subsequent passaging.A549 cells lacking HLA-A*02:01 expression and cultures without CD8 + T cells served as negative controls.As expected and described previously (van de Sandt et al., 2018a), the inhibitory effect of M1 58-66 -specific T cells on replication of WSN-M-aH5N1 was more significant than that on WSN-M-hH3N2, confirming the relative resistant phenotype of viruses with an M1 gene of human seasonal influenza viruses.
To monitor the emergence of virus variants upon serial passaging in the absence or presence of potential selective pressure by M1 58-66 -specific T cells, the culture supernatants obtained after passages 1, 5, and 10 were subjected to Next Generation Sequencing (NGS) to detect amino acid substitutions at extra-epitopic residues 15, 27, 101, 115 and 121, known to affect recognition by M1 58-66 -specific T cells or in the epitope itself (van de Sandt et al., 2016;van de Sandt et al., 2018a).The M1 58-66 epitope remained 100 % conserved in all analysed progeny virus preparations.This confirms the highly conserved nature of this epitope, which has been shown to be under functional constraints (Berkhoff et al., 2005;Terajima and Ennis, 2012), most likely due to overlapping with a nuclear export domain in the M1 protein (Cao et al., 2012).Mutations at the extra-epitopic residues that converted the signatures from avian-to-human, and vice versa, were readily detected even in passage 5, although conversion from avian-to-human signature was observed at higher frequencies.However, these amino acid substitutions were also observed in A549 wt cells that lack HLA-A*02:01 expression and in the absence of M1 58-66 -specific T cells.
The passaging of a molecularly cloned virus possessing an M1 gene originating from an avian virus in human lung epithelial cells, generates a quasi-species containing variants of human signature with all five amino acid substitutions of interest.Thus, the emergence of variants with substitutions at amino acid residues 15, 27, 101, 115, and 121 may not be dependent per se on selective pressure mediated by M1 58-66specific CD8 + T cells, but may also reflect adaptive changes of avian virus to mammalian host environment.Of interest, the presence of selective pressure mediated by M1 58-66 epitope-specific T cells drove further expansion of the variants close to fixation.This observation resembles earlier findings obtained with variation in another influenza virus epitope NP 383-391 .The rapid fixation of a mutation of the HLA-B27:05 restricted epitope was explained by small selective advantages in the human population and strong bottleneck and founder effects (Gog et al., 2003).Similar processes may have been at the basis of our findings in the in vitro co-culture system used in the present study.IAV infection of mice, transgenic for a T cell receptor specific for the NP 366-374 epitope, readily resulted in the emergence of virus variants with mutations in the epitope that abrogated recognition by specific T cells (Price et al., 2000).The use of HLA-A*02:01 transgenic mice would enable further investigation of the effect of the extra-epitopic residues on the M1 58-66 -specific T cell response and the emergence of virus variants with a human signature.
Interestingly, we identified specific nucleotide substitutions that led to the observed the amino acid changes (Fig. 2B).These nucleotide substitutions were the only significantly occurring changes that led to the described changes at the extra-epitopic residues.The observed consistency of the variant frequency, including those encoding the more distant amino acid substitutions (T137A and the A166V, T167A, and I168T motif), contrasts the variable nature of influenza viruses.This may be associated with secondary RNA structure involved in virus replication (Spronken et al., 2017).
Human seasonal (and pandemic) viruses that circulated since the pandemic of 1918 all had an M1 gene of human signature (van de Sandt et al., 2012).In contrast, the influenza A virus that caused the pandemic of 2009 ((H1N1)pdm09) had an M1 gene of avian signature during the first seven years of circulation in the human population.This virus did not accumulate amino acid substitutions in the extra-epitopic residues, despite selective pressure by M1 58-66 -specific T cells in the human population (van de Sandt et al., 2018b).It may take some time to acquire adaptive changes in the M1 protein and the associated immune evasive phenotype under field conditions.Investigating within-host IAV evolution may shed light in these adaptive changes (Han et al., 2021).
In the present study, we used a unique co-culture system with HLAtransgenic A549 cells, cloned T cells specific for the M1 58-66 epitope and molecular cloned isogenic viruses.The approach enabled to investigate viral evolution under isolated selective pressure mediated by CD8 + T cells.The data suggest that the presence of amino acid residues that affect recognition of IAV by M1 58-66 -specific T cells, most likely by affecting antigen processing and epitope presentation (van de Sandt et al., 2016;Hofmann et al., 2001), is the result of host cell adaption on the one hand and selective pressure by specific T cells on the other.Continuous monitoring of the evolution of H1N1pdm09 viruses may advance our knowledge on evolution of influenza viruses with an M gene segment of avian origin.

Fig. 1 .
Fig. 1.Establishment of efficient virus passaging in A549/ M1 58-66 -specific CTL co-cultures.Virus replication kinetics of WSN-M-hH3N2 and WSN-M-aH5N1 in (A) A549wt and (B) A549tg-A2 cells in the presence of different numbers of M1 58-66 -specific CTL as indicated in the legend.(C) Total virus replication during 72 h of cocultures.Area under the curve (AUC) is depicted of the grow curves shown in A and B. (D) Viral titers after passaging of WSN-M-hH3N2 and WSN-M-aH5N1 in A549 wt (black symbols) and A549 tg-A2 (blue/red symbols respectively) cells in the presence of 50.000CTL/well, co-cultured with A549 wt cells or A549 tg-A2 cells expressing HLA-A*02:01.(E) Mean fold-reduction of viral titers in the presence of 50.000CTL/well, co-cultured with A549 wt cells or A549 tg-A2 cells expressing HLA-A*02:01, relative to the respective wells without CTL (as shown in D).For all data points, median of n = 3 (A-C) or n = 2 (D-E) and SD are shown.

Fig. 2 .
Fig. 2. Sequence comparisons and frequency of mutations.(A) M1 sequences of hH3N2 (A/Netherlands/18/1994) and aH5N1 (A/Vietnam/1194/2004).Highlighted in yellow is the M1 58-66 epitope.The amino acid variations of interest highlighted in red.(B) Mutations occurring at significant frequencies resulting in the depicted amino acid (AA) variations of interest (shown in A). (C) Proportion of the detected NGS reads coding for the aH5N1 AA sequence relative to the hH3N2 AA sequence at the 5 AA positions of interest.The hH3N2 virus and its progeny are indicated in blue colors, the aH5N1 virus and its progeny is indicated in red colors.The shade of the colors reflects individual AA substitutions as indicated.Depicted are the results obtained after passages 1, 5 and 10 of co-culture.