We and others have previously demonstrated that only a small fraction of IAV-infected cells express type I or type III IFN [31,34–36]. We further showed that different IAV strains can differ in the percentage of infected cells that eventually express IFN [31]. Similar cellular heterogeneity in IFN induction has been observed for other viruses as well [37–41]. Given the rarity of cells driving the IFN response, bulk methods of analyzing the host response may fail to effectively capture important dynamics within these cells. Further, due to the highly non-linear nature of cytokine responses [42,43], even small differences in the percentage of infected cells that produce IFN could have significant effects on downstream infection dynamics. Thus, we aimed to more carefully characterize patterns of single cell innate immune activity during IAV infection.

To examine the effects of infection on host gene expression across millions of individual cells while avoiding the high costs associated with single-cell RNAseq (scRNAseq), low sensitivity and noise associated with this approach [44,45], we adapted hybridization chain reaction-based detection of nucleic acid targets [46] for use with flow cytometry (HCR-flow). To evaluate the sensitivity and specificity of HCR-flow, we measured the expression of the human housekeeping genes GAPDH and ACTB (as positive controls) and murine Cd45(as a negative control) in the human lung epithelial cell line A549. We detected both GAPDH, ACTB in nearly 100% of cells but not murine Cd45, indicating the suitability of HCR-flow specific detection of transcripts within single cells (S1 Fig).

We next evaluated the ability of HCR-flow to measure the induction of innate immune-associated transcripts upregulated during viral infection [7,31]. We transfected A549 cells with the RIG-I ligand polyinosinic-polycytidylic acid (pIC) and measured expression of IFNL1 and the ISG IFIT3 at different times following transfection using HCR-flow. We observed a rapid increase in IFIT3 expression in A549s resulting in ~80–90% of cells expressing ISGs by 16–20 hrs. In parallel, we observed a slower increase in IFNL1 expression that plateaued with only ~20% of the cells expressing detectable levels of IFNL1 by 20 hrs (Fig 1A and 1B). We further observed that IFNL1 expression frequencies measured by HCR-flow agreed with both IFN transcript levels quantified by bulk qPCR and detection of secreted IFNL in supernatant of pIC treated A549s (S2 Fig). These data are consistent with previous reports that IFN induction typically only occurs within a subset of stimulated cells [31,35–37,47,48].

Finally, we assessed the ability of HCR-flow to detect viral infection compared with traditional antibody staining. We infected A549 cells with A/California/07/2009 (Cal07; 2009 pandemic H1N1 strain) and simultaneously stained for viral NP protein with antibody and the viral mRNA that encodes NP with HCR-flow. We observed clear positive signals for both NP protein and NP mRNA that were highly concordant with each other, indicating that HCR-flow performs comparably to antibody staining for detection of IAV infection (Fig 1C). Altogether, these data indicated that HCR-flow can be used to quantify the expression of both host and viral RNAs at the single cell level.

To assess IFN and ISG induction in infected and bystander cells, we infected A549 cells at a multiplicity of infection (MOI) of 0.1 based on NP-expressing units (NPEU) [49] with recombinant Cal07 or A/Perth/16/2009 (Perth09; human seasonal H3N2 isolate) under single cycle conditions. We used a low MOI to ensure that each infected cell was infected with a single virion, as cellular co-infection can alter IFN induction kinetics [50]. At 16 hpi, we collected the cells and examined viral and host gene expression by HCR-flow (S3 Fig). Based on detection of viral NP mRNA, we could clearly resolve infected and bystander cells (Fig 2A and 2B). We compared expression of IFNL1 transcript between infected (NP+) and bystander (NP-) cells (Fig 2C). We observed that the frequency of IFNL1 induction in Perth09-infected cells was consistently ~4-5-fold higher than that observed in Cal07-infected cells, consistent with previously published results comparing these strains using scRNAseq [31].

We next assessed induction of the ISG IFIT3 within infected and bystander cells. In contrast with the IFNL1 results, a significantly lower percentage of Perth09-infected cells expressed IFIT3 compared with Cal07-infected cells (Fig 2D). At least 50% of all bystander cells expressed IFIT3, consistent with widespread paracrine activation of ISG expression and suggesting that the lower levels of IFIT3 expression in infected cells are due to active suppression by the virus. While the pattern of expression differences between Cal07- and Perth09-infected cells was consistent across three independent experiments, the specific percentages of IFNL1+ and IFIT3+ cells varied between experiments as a function of cell passage number used. As a result, in this study we are showing a single representative of three independent experiments.

We next compared our HCR-flow results with scRNA-seq data obtained from A549 cells infected with either Cal07 or Perth09. We infected A549 with either Cal07 or Perth09 at MOI of 0.1 under single cycle conditions using ammonium chloride (NH₄Cl). We sorted cells at 8 and 16 hpi into bystander or infected populations based on the expression of viral HA protein (S4 Fig). We calculated percentages of infected cells with any detectable level of IFNL1 or IFIT3 in our scRNA-seq dataset and observed similar patterns of IFN/ISG expression using both single cell methods (Fig 2E and 2F). Furthermore, we observed similar differences between the strains in the induction of type I IFNs and larger group of diverse ISGs (S5 Fig). Additionally, we compared IFNL1 counts with either viral NP or NS counts for Cal07 and Perth09 libraries (S6 Fig). Perth09-infected cells with high IFNL1 tended to have lower NS counts while the opposite was observed for Cal07-infected cells. The relationships between IFNL1 and NP counts were similar for both viruses.

Finally, we compared innate immune activation kinetics between Cal07 or Perth09 by infecting cells and harvesting at different times following infection. We measured bulk IFNL1 expression by qPCR and secretion of functional type III IFN in the supernatant (Fig 3A and 3B). Perth09 infection was associated with faster IFN induction compared with Cal07. The frequencies of IFNL1+ cells as measured by HCR-flow were consistent with bulk measurements (Fig 3C). IFIT3 induction during Cal07 infection was delayed relative to Perth09; however, it accelerated after 12 hpi and surpassed Perth09-infected cells by 16 hpi (Fig 3D). This dynamic could reflect more potent inhibition of IFN signaling by Perth09. NP and NS1 mRNA expression kinetics were largely similar between Cal07 and Perth09, suggesting that their distinct IFNL1 and IFIT3 expression phenotypes are not simply the effect of differences in viral gene expression kinetics (Fig 3E and 3F). Altogether, these data indicate that Perth09 is more effective at suppressing ISG induction but less effective at avoiding IFN induction, compared with Cal07.

The viral NS segment encodes NS1, which plays a well-described role in modulation and antagonism of the innate immune response [51,52]. The NS segments encoded by the 2009pdm H1N1 and H3N2 lineages have distinct natural histories and sequence features and thus potentially differ in their immune antagonism effectiveness. While the Perth09 strain used here is part of the seasonal H3N2 lineage that has circulated in humans since emerging through the 1968 pandemic, the H3N2-associated NS segment has actually circulated in humans continuously since the 1918 pandemic [53]. This is presumably due to reassortment events that allowed the 1918-origin NA segment to persist through the subtype replacements associated with the 1957 H2N2 and 1968 H3N2 pandemics [53–55]. In contrast, Cal07 was isolated during the 2009 pandemic and encodes a classical swine-origin H1N1 NS segment (itself originally derived from the 1918 pandemic virus) that had circulated in swine for decades [55,56]. As a result of these distinct natural histories, Cal07 and Perth09 exhibit limited nucleotide (81.73%) and amino acid (NS1: 78.08%; NEP: 90.08%) sequence similarity. Based on these differences, we examined the specific contributions of the NS segment to the observed strain differences in IFN/ISG induction.

We generated reassortant viruses in which we swapped the NS segments between Cal07 and Perth09 using reverse genetics (Cal07:NS-Perth09 and Perth09:NS-Cal07). We infected A549 cells at an MOI of 0.1 NPEU/mL under single cycle conditions and collected cells at 16 hpi for HCR-flow analysis. We gated infected based on the presence of detectable NP mRNA signal (Fig 4A and 4B). Unlike our results with the parental viruses, we did not observe a significant difference in IFNL1 expression frequency between the reassortants (Fig 4C), suggesting that the different IFNL1 induction frequency phenotypes observed for Cal07 and Perth09 (Fig 2C) are only partially due to the NS segment. In contrast, the IFIT3 expression phenotype segregated cleanly with the NS segment, as Cal07:NS-Perth09 was associated with lower IFIT3 expression compared with Perth09:NS-Cal07 (Fig 4C). These data demonstrate that the strain differences in ISG expression frequency map to the NS segment.

The differences in IFNL1 and IFIT3 expression frequencies during Cal07 and Perth09 infection could be explained by differences in either (a) ability to actively suppress IFN/ISG induction within infected cells, or (b) ability to evade sensing by the cell, as NS1 has been shown to suppress dsRNA sensing [21,57,58]. To differentiate between these possibilities, we infected A549 cells with either Cal07 and Perth09, treated cells 4 hrs later with 10 ng/mL of the RIG-I agonist pIC, and then assessed IFNL1 and IFIT3 expression in mock and infected cells by HCR-flow at 16 hpi.

Uninfected A549 cells transfected with pIC (10 ng/mL) for 16 hrs exhibited relatively high expression frequencies of both IFNL1 (~20%) and IFIT3 (~85%), indicating the maximal expression frequencies of these transcripts in the absence of viral interference (Figs 5A, 5B and S7). In Cal07 or Perth09 infected cells treated with pIC, IFNL1 expression frequency was much lower than in mock controls shown in Fig 5A, indicating that both strains actively inhibit IFN induction in response to RNA sensing within the cell. Perth09 infected cells consistently exhibited modestly higher rates of IFNL1 induction compared with Cal07 infected cells, suggesting that Perth09 is slightly less effective at suppressing IFN induction (Fig 5C).

IFIT3 expression frequencies were much lower in Perth09 infected cells versus Cal07 infected cells following pIC treatment (Fig 5C). This suggests that Perth09 is more effective than Cal07 at suppressing ISG induction rather than simply avoiding ISG induction. Similar to our results with virus-mediated induction of ISG expression, the difference between Perth09 and Cal07 in suppression of pIC-induced ISG expression segregated entirely with the NS segment (Fig 5D).

Canonical ISG induction occurs downstream of signaling through the type I or type III IFN receptors [59]; however, ISGs can also be directly induced through IRF3 activation, independent of IFN signaling [60]. To determine whether the effects of Perth09 NS on ISG induction resulted from viral antagonism of IFN signaling, we infected A549 cells with either Cal07 or Perth09 and treated the cells with recombinant human IFNB1 or IFNL1 at 4 hpi. We first confirmed that neither IFNB nor IFNL treatment induced expression of IFNL1 but that both did induce IFIT3 expression (Figs 6A and S8). IFNL1 and IFNB1-driven IFIT3 expression frequencies were significantly lower in Perth09-infected cells compared to Cal07 (Fig 6B and 6C). These data suggest that Perth09 is better at inhibiting type I and type III IFN signaling compared to Cal07.

Cells expressing IFNs can upregulate ISGs through autocrine and paracrine IFN signaling, in which type I or type III IFN receptor engagement triggers the canonical JAK/STAT signaling pathway [61]. To investigate if JAK mediated IFN signaling is essential for the differential ISG antagonism between Perth09 and Cal07, we pre-treated cells with the JAK1/2 inhibitor ruxolitinib [62]. In the presence of ruxolitinib, we expect to lose all ISG induction triggered by IFN signaling. To test this, we treated cells with 10 μM ruxolitinib and then exposed them to recombinant IFNB1 or IFNL1 (Fig 6D). Under these conditions, ruxolitinib pretreatment completely prevented IFIT3 induction by recombinant IFNB1 (S9 Fig). To quantify IFN signaling-independent induction of IFIT3 by pIC, we pre-treated cells with 10 μM ruxolitinib, transfected with 10 ng/mL pIC, and then measured the percentages of cells that turned on IFIT3 expression (Fig 6E).

Next, we asked how inhibition of JAK/STAT signaling affected ISG induction during viral infection. As expected, controls pretreated with DMSO prior to Cal07 or Perth09 infection induced IFIT3 at frequencies similar to what we observed previously (Fig 6F). Pretreatment with ruxolitinib flipped the pattern of IFIT3 expression between Cal07 and Perth09, however, with a significantly higher percentage of Perth09-infected cells expressing IFIT3 compared with Cal07. Thus, in the presence of ruxolitinib, the IFIT3 induction phenotypes of Perth09 and Cal07 closely mirror their IFNL1 induction phenotypes in the absence of ruxolitinib. These data suggest that Perth09 is less effective than Cal07 at suppressing initial IRF3-mediated induction of IFN and ISGs, but more effective at antagonizing JAK/STAT-dependent ISG induction downstream of IFN signaling.

Overcoming innate immune defenses is thought to be one of the major factors limiting cross-species transmission [3,63]. Recent evidence from the SARS-CoV-2 pandemic has suggested that viruses may undergo selection for enhanced innate immune antagonism in humans following zoonotic emergence [64]. It is not clear whether seasonal IAVs have undergone similar selection to optimize innate immune antagonism phenotypes as they have adapted to humans over the decades.

We asked whether the single cell IFN/ISG induction phenotypes associated with the seasonal H3N2 and H1N1 lineage NS segments have evolved from their original pandemic emergence and subsequent years of human circulation. We used reverse genetics to generate recombinant viruses with NS segments from representative H3N2 isolates from 1968 to 2020 in the Perth09 background, and H1N1 isolates from 2009 to 2020 in the Cal07 background (Table 1). Representative H3N2 strains from 1968–1999 were chosen from vaccine strains. Representative H1N1 and H3N2 from 2012–2020 were selected from years where significant numbers of new coding substitutions appeared in the NS1 consensus sequence, specifically looking at years with more than five coding changes compared to Perth09 and accumulation of at least one substitution between them. For each year chosen, we selected a representative strain that exhibited high NS1 sequence similarity to the consensus sequence which were generated using available NS1 sequences from each year in the bacterial and viral bioinformatics resource center (BV-BRC) database.

We infected A549 cells at an MOI of 0.1 NPEU/mL with each recombinant virus under single cycle conditions and measured the percentages of infected cells expressing IFNL1 and/or IFIT3 using HCR-flow. To ensure fair comparison between the different viral genotypes, we confirmed that infected cell percentages were comparable for all viruses tested (Fig 7A and 7B). We first examined the effects of different historical H3N2-origin NS segments (1968, 1979, 1999, 2009, 2012, 2015, 2018, and 2020) on single cell IFNL1 and IFIT3 induction frequencies (Fig 7C and 7E). To account for technical variation between experiments, we normalized everything to the mean values measured for Perth09. We consistently observed that IFNL1 and IFIT3 induction frequencies were largely constant from 1968 through 1999, suggesting that the effect of the NS segment on IFN/ISG induction phenotypes was largely unchanged over the first 30+ years of H3N2 circulation.

Starting with Perth09, IFNL1 and IFIT3 induction frequencies decreased through 2015 before increasing again significantly by 2020. These data suggest that the ability of the H3N2-origin NS segment to antagonize IFN and ISG induction has fluctuated significantly at times since the 1968 pandemic; however, there has not been a constant trend towards increased antagonism potency over 50+ years of human circulation.

For the pdm2009 H1N1 lineage NS segment, all post-2009 representatives tested except for A/Arizona/16/2020 exhibited significant increases in IFNL1 induction frequencies, compared with Cal07 (Fig 7D). In contrast, IFIT3 induction frequencies were largely stable from 2009 to 2020 (Fig 7F). These data suggest that, if anything, the pdm2009 H1N1 lineage NS segment has evolved to become less capable of inhibiting IFN induction since its original emergence.

Finally, we examined whether the differences in IFNL1 and IFIT3 induction across the different H1N1 and H3N2 isolates tested could be simply explained by changes in relative levels of NS1 expression. We measured NS1 transcript levels in A549 cells infected with either H1N1 or H3N2 at MOI 0.1 by qPCR and normalized to NP mRNA levels to account for any differences in infection kinetics (S10A and S10B Fig). We observed a steady increase in relative NS1 expression levels as a function of year of isolation for the H3N2 lineage but not for the pdm2009 H1N1 lineage. When we directly compared relative NS1 expression levels and IFN/ISG expression phenotypes across the different viruses in our panel, we observed a significant negative correlation (p = 0.0034) between NS1 expression levels and IFIT3 induction frequencies for H3N2 viruses, but not for H1N1 (S10C and S10F Fig). This suggests that some of the changes in IFN signaling antagonism observed across the H3N2 lineage may be driven by changes in the kinetics and/or magnitude of NS1 expression. No significant correlation was observed between NS1 expression levels and IFNL1 induction frequencies for either lineage.

Altogether, our analyses suggest that the innate immune antagonism phenotypes of the seasonal H1N1 and H3N2 NS segments have varied significantly over the years; however, neither lineage has exhibited a consistent trend of evolving increasingly potent antagonism of IFN or ISG induction since emerging into humans.

A/California/04/09 plasmids were provided by Dr. Jonathan Yewdell, and substitutions A660G and A335G were introduced into HA and NP, respectively, to convert them into A/California/07/09 (NCBI accession numbers CY121680 and CY121683). Dr. Seema Lakdawala generously provided the plasmids encoding A/Perth/16/09 viral segments. Madin-Darby canine kidney cells (MDCK) and human lung epithelial cells (A549) were both obtained from Jonathan Yewdell. MDCK and A549 cells were maintained in Gibco’s minimal essential medium with GlutaMax (Life Technologies) and Gibco’s Dulbecco’s Modified Eagle Medium high glucose supplemented with GlutaMax and sodium pyruvate (Life Technologies), respectively. Human kidney embryonic cells (293T) were provided by Dr. Joanna Shisler and maintained in Gibco’s minimal essential medium with GlutaMax (Life Technologies). All cell lines were supplemented with 8.3% fetal bovine serum (Avantor) and grown at 37°C with 5% CO2.

Recombinant A/California/07/09 and A/Perth09/16/09 were rescued using a reverse genetics system by transfecting 60%-80% confluent 293T with 500ng of each plasmid using jetPRIME (Polyplus). MDCK cells were used to amplify the virus rescue and to generate working stock by infecting T-75 flasks at MOI of 0.01 based on TCID50/mL. Sequences for NS segments selected from historic strains (Table 1) were synthesized using gBlocks (IDT) and cloned into the pDZ dual-promoter reverse genetics vector using SapI sites. Constructs were used to rescue viruses using the reverse genetic system described previously.

Recombinant viruses were titered by infecting A549 under single cycle conditions (imposed by adding neutralizing antibodies EM4-C04 [2.75 μg/ml; for Cal07] or CO5 [6.5 μg/ml; for Perth09] to cell supernatant after binding and entry) and quantifying infected cell numbers by flow cytometry using the anti-NP monoclonal antibody HB65. By combining viral dilution factor and the number of infected cells, we were able to calculate NP-expressing units/mL (NPEU/mL) as previously described [73]. A549 were infected at MOI of 0.1 NPEU/mL by diluting virus in PBS + 0.1% BSA and incubating for 1 hr at 37°C, after incubation the monolayer was rinsed with PBS and growth media supplemented with 8.3% FBS and neutralizing antibodies EM4-C04 [2.75 μg/ml] or CO5 [6.5 μg/ml] to block viral spread and cells were collected at 16–18 hpi. Infectivity was tested by staining for viral protein NP using the anti-NP mouse monoclonal HB65. Cells were incubated with Foxp3/Transcription Factor Staining Buffer Set (eBioscience) for 30 mins at 4°C and rinse twice with PBS supplemented with 0.1% BSA and 0.1% Saponin before staining. Cells were incubated with HB65 conjugated to Alexa Fluor 647 diluted in PBS supplemented with 0.1% BSA and 0.1% Saponin for 30 mins at 4°C. After incubation cell were rinsed with PBS twice and analyze using BD FACSymphony A1.

Infections for scRNA-seq experiment were performed in A549 cells incubated with virus diluted in PBS + 0.1% BSA for 1 hr at 37°C at MOI of 0.1 based on NPEU/mL. After incubation, the cells were rinsed with PBS, and growthy media was added. At 3hpi, growth media was removed, and F-12 was supplemented with 50 mM HEPES and 20 mM NH₄Cl to block secondary spread. After incubation, cells were rinsed with PBS stained with HA stem-specific antibody FI6v3 and washed before being sorted into mock, bystander, and infected populations using a Bigfoot cell sorter.

Sorted A549s were counted, and viability was determined on a BD20 cell counter (BIO-RAD). Cells were diluted to desired concentrations (5000 cells/sample) and individual cDNA libraries were made using the 10x Chromium Single cell 3`(Pleasanton, CA) as per manufacturers protocol. Following cDNA synthesis, libraries were sequenced on an Illumina NovaSeq 6000 using S4 flowcells. In house analysis used for this study can be found in https://github.com/BROOKELAB/Influenza-virus-scRNA-seq.

Concentration of polyinosinic-polycytidylic acid (pIC) (Invivogen: tlrl-picw) was determine by transfecting A549 using lipofectamine 3000 transfection reagent (Invitrogen) following manufacturer’s instructions. After 16 hrs RNA was isolated using RNeasy Mini kit (Qiagen) and cDNA synthesis was performed using Verso cDNA synthesis kit (Thermo Scientific). Reaction were setup by adding 4 μL of 5x cDNA synthesis buffer, 2 μL of dNTPs, 1 μL Oligo-dT, 1 μL of RT enhancer, 1 μL of Verso enzyme, 5 μL of cDNA and 6 μL of nuclease-free water. Mixture was incubated at 45°C for 50 min and 95°C for 2 min and held at 4°C; cDNA was stored at -20 and subsequently used for quantitative pcr. TaqMan fast advanced master mix (Applied Biosystems) was used to carry out the qPCR in a QuantStudio 3 (Applied Biosystems). Reaction was setup as follows 10 μL of TaqMan fast advanced MM, 1 μL of IFNL1 probe (Applied Biosystems: Hs00601677_g1), IFIT3 probe (Applied Biosystems: Hs01922752_s1) or ACTB probe (Applied Biosystems: Hs01060665_g1), 1 μL of cDNA and 8 μL of nuclease-free water. Cycle conditions for qPCR were setup as 50°C for 2  min, 95°C for 2 mins, and 40 cycles of 95°C for 1 second followed by 60°C for 20 seconds.

RNA from infected A549s was isolated and cDNA synthesis was performed as described above. Expression of NS1 and NP was measured by qPCR using PowerUp SYBR Green Master Mix (Applied Biosystems) Reaction was setup as follows 10 μL of using PowerUp SYBR Green MM, 0.5 μL of 10 μM forward and reverse primers (H1-NS1-forward: AACACCCTTGGCCTCGATAT, H1-NS1-reverse: TGAGCATGAACCAGTCTCGT, H1-NP-forward: CCCAGGAAACGCTGAGATTG, H1-NP-reverse: GACCAGTGAGTACCCTTCCC, H3-NS1-forward: CCATTCCTTGATCGGCTTCG, H3-NS1-reverse: TCCTTCCATTTTCTGCTTGGG, H3-NP-foward: TCGGGACAATGGTGATGGAA, H3-NP-reverse: CCTGGGTTCCGACTTTCTCT), 1 μL of cDNA and 8 μL of nuclease-free water. Cycle conditions for qPCR were setup as 50°C for 2  min, 95°C for 10 mins, and 40 cycles of 95°C for 15 seconds followed by 60°C for 1 minute.

Dose of recombinant human IFNB1 and IFNL1 (PeproTech) was determine by diluting IFNs growth media and incubated with A549 cells. After 16 hrs RNA was isolated, cDNA synthesis was performed and IFIT3 expression was measured by qPCR as described previously. Inhibition of JAK 1/2 by Ruxolitinib or Baricitinib (MedChemExpress) was determined by diluting inhibitors in growth media and incubated with A549 for 4 hrs before inducing IFN signaling using recombinant hIFNB1. After 16 hrs post IFN treatment RNA was isolated, cDNA synthesis was performed and IFIT3 expression was measured by qPCR as described above.

Supernatant from A549 cells transfected with pIC or infected with IAV was collected and incubated with HEK-Blue IFN-λ reporter cells (Invivogen) for 24 hrs. After incubation, supernatant from HEK-Blue cells was incubated with QUANTI-Blue reagent (Invivogen) as per manufacturers protocol. Presence of IFNL in supernatant was determined by measuring the optical density at 620 nm.

Single cell suspension from A549 cells was prepared by incubating cells with 0.05% Trypsin-EDTA (Gibco) at 37°C for 5 mins. Cell suspension was washed with DPBS (Gibco) and fixed with 4% formaldehyde (Thermo Scientific) for 30 mins at 4°C. After incubation, cells were washed twice with DPBS and incubated overnight in 70% ethanol. Single cells suspension was washed twice with DPBS + 0.1% Tween 20 (Thermo Scientific Chemicals) and incubated with amplification buffer (Molecular Instruments) for 30 mins at 37°C. After incubation, 1-3 μM of each probe (Molecular Instruments) was added to cell suspension and incubated overnight at 37°C. Cells were resuspended in probe wash buffer (Molecular Instruments) and incubated at 37°C for 10mins and this was repeated three times. Cells were washed once with 5x sodium chloride sodium citrate + 0.1% Tween 20 (SSCT) and incubated in amplification buffer (Molecular Instruments) at RT for 30 mins. After incubation 3-6 μM of each snap-cooled hairpin (Molecular Instruments) was added to cells suspension and incubated at RT overnight. Cells were washed twice with SSCT and analyzed using BD FACSymphony A1.