Decision letter: Host proteostasis modulates influenza evolution

Abstract

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods References Decision letter Author response Article and author information Metrics Abstract Predicting and constraining RNA virus evolution require understanding the molecular factors that define the mutational landscape accessible to these pathogens. RNA viruses typically have high mutation rates, resulting in frequent production of protein variants with compromised biophysical properties. Their evolution is necessarily constrained by the consequent challenge to protein folding and function. We hypothesized that host proteostasis mechanisms may be significant determinants of the fitness of viral protein variants, serving as a critical force shaping viral evolution. Here, we test that hypothesis by propagating influenza in host cells displaying chemically-controlled, divergent proteostasis environments. We find that both the nature of selection on the influenza genome and the accessibility of specific mutational trajectories are significantly impacted by host proteostasis. These findings provide new insights into features of host–pathogen interactions that shape viral evolution, and into the potential design of host proteostasis-targeted antiviral therapeutics that are refractory to resistance. https://doi.org/10.7554/eLife.28652.001 eLife digest Influenza viruses, commonly called flu, can evade our immune system and develop resistance to treatments by changing frequently. Specifically, mutations in their genome cause influenza proteins to change in ways that can help the virus evade our defences. However, these mutations come at a cost and can prevent the viral proteins from forming functional and stable three-dimensional shapes – a process known as protein folding – thereby hampering the virus’ ability to replicate. In human cells, proteins called chaperones can help our other proteins fold properly. Influenza viruses do not have their own chaperones and, instead, hijack those of their host. Host chaperones are therefore crucial to the virus’ ability to replicate. However, until now, it was not known if host chaperones can influence how these viruses evolve. Here, Phillips et al. used mammalian cells to study how host chaperones affect an evolving influenza population. First, cells were engineered to either have normal chaperone levels, elevated chaperone levels, or inactive chaperones. Next, the H3N2 influenza strain was grown in these different conditions for nearly 200 generations and sequenced to determine how the virus evolved in each distinctive host chaperone environment. Phillips et al. discovered that host chaperones affect the rate at which mutations accumulate in the influenza population, and also the types of mutations in the influenza genome. For instance, when a chaperone called Hsp90 was inactivated, mutations became prevalent in the viral population more slowly than in cells with normal or elevated chaperone levels. Moreover, some specific mutations fared better in cells with high chaperone levels, whilst others worked better in cells with inactivated chaperones. These results suggest that influenza evolution is affected by host chaperone levels in complex and important ways. Moreover, whether chaperones will promote or hinder the effects of any single mutation is difficult to predict ahead of time. This discovery is significant, as the chaperones available to influenza can vary in different tissues, organisms and infectious conditions, and may therefore influence the virus' ability to change and evolve in a context-specific manner. The findings are likely to extend to other viruses such as HIV and Ebola, which also hijack host chaperones for the same purpose. More work is now needed to systematically quantify these effects so that we can better predict how specific chaperones will affect the ability of viruses to adapt, especially in pathologically relevant conditions like fever or viral host-switching. In the future, such insights could help shape the design of treatments to which viruses do not evolve resistance. https://doi.org/10.7554/eLife.28652.002 Introduction Minimalist pathogens like RNA viruses survive dynamic host environments by virtue of their extreme adaptability. This adaptability is driven by a high rate of genetic variation, mediated by error-prone genome replication (Sanjuán et al., 2010). Most missense mutations have deleterious consequences for protein function, often owing to either thermodynamic (reduced stability of the native state or enhanced stability of unfolded/misfolded states) or kinetic (slow folding or enhanced misfolding/aggregation) effects on folding (DePristo et al., 2005). In the context of viruses, these phenomena may underpin the observation that the distribution of mutational fitness effects can be largely accounted for by considering protein folding biophysics (Wylie and Shakhnovich, 2011; Chéron et al., 2016; Tokuriki and Tawfik, 2009a). Indeed, stable proteins tend to be more evolvable, as any given missense mutation is less likely to severely disrupt protein folding or structure (Bloom et al., 2006; Gong et al., 2013). In cells, protein folding challenges are addressed by proteostasis networks composed of chaperones and quality control factors that work in concert to shepherd nascent proteins to folded, functional conformations (Balch et al., 2008; Hartl et al., 2011; Powers and Balch, 2013). Important work focused primarily on the Hsp90 chaperone has suggested a critical role for chaperones in modulating the evolution of their endogenous clients, (Cowen and Lindquist, 2005; Queitsch et al., 2002; Lachowiec et al., 2015; Sangster et al., 2007, 2008a, 2008b; Rohner et al., 2013; Whitesell et al., 2014; Geiler-Samerotte et al., 2016; Rutherford and Lindquist, 1998) in part by buffering deleterious effects of non-synonymous mutations. The consequences of Hsp90 activity for protein evolution may be due to Hsp90 directly engaging an evolving client protein (termed a primary effect). Alternatively, the effects of Hsp90 may be secondary, mediated indirectly by Hsp90 influencing the folding of other endogenous clients that themselves engage relevant evolving proteins. For instance, Hsp90-dependent azole resistance in Candida albicans is mediated by secondary effects of Hsp90 on calcineurin, an Hsp90 client that regulates responses to environmental stimuli (Cowen and Lindquist, 2005). Efforts to look beyond Hsp90 to understand how other components of the metazoan proteostasis machinery modulate evolution (e.g., Hsp40/70 chaperones or protein misfolding stress responses like the heat shock response) have been slowed by the paucity of chemical biology tools to perturb the activities of these systems. However, Tawfik and coworkers have shown that the GroEL/ES chaperonin system can govern the fitness of certain client protein variants in bacteria (Tokuriki and Tawfik, 2009b), and computational modeling suggests that other chaperones may also have roles in evolution (Bogumil and Dagan, 2012; Cetinbaş and Shakhnovich, 2013). Chaperones and other proteostasis mechanisms are theoretically well-positioned to address the biophysical challenges created by high mutation rates in viruses. Intriguingly, most RNA viruses lack autonomous chaperones or other co-factors to assist their proteins with folding. Instead, viral proteins engage host chaperones, (Melville et al., 1999; Momose et al., 2002; Naito et al., 2007; York et al., 2014; Watanabe et al., 2010) and host chaperone inhibitors have been shown to limit the viability of certain RNA viruses (Geller et al., 2007; Heaton et al., 2016; Taguwa et al., 2015; Chase et al., 2008; Geller et al., 2012). However, the possibility that host chaperones can shape the evolution of viral pathogens has not been investigated. In summary, it is clear that: (1) high genetic variability is essential to support RNA virus adaptability; (2) missense mutations important for viral adaptation are often biophysically deleterious, constraining the accessible mutational landscape; and (3) many viruses engage host chaperones to fold their proteins. An important but still untested hypothesis is that host proteostasis modulates RNA virus evolutionary trajectories. Experimentally testing this hypothesis requires methods to regulate the host cell’s proteostasis network without significantly perturbing cell health or the ability of an RNA virus to propagate. Here, we achieve this goal in the context of long-term influenza propagation by using small molecules to either modulate the heat shock response in a stress-independent manner (Shoulders et al., 2013; Moore et al., 2016) or to inhibit Hsp90 at sub-lethal concentrations (Ying et al., 2012). We find that the resulting perturbations to host proteostasis mechanisms significantly impact both the extent and the nature of selection pressure on the influenza genome. We conclude that host proteostasis is a critical, under-appreciated player in influenza evolution, with significant implications for our ability to predict and prevent the evolution of influenza and other RNA viral pathogens. Results Small molecule-based strategies create three distinctive host proteostasis environments for influenza evolution experiments Eukaryotic cells dynamically match proteostasis network capacity to demand via compartment-specific stress responses. Thus, one biologically relevant strategy to create an altered proteostasis environment is to induce such responses. We focused on the heat shock response, (Åkerfelt et al., 2010) because numerous influenza proteins must fold and/or function in the cytosol and nucleus (Watanabe et al., 2010). Typical methods to induce heat shock factor 1 (HSF1), the master regulator of the heat shock response and thus of cytosolic and nuclear chaperone and quality control protein levels, involve treatment with toxins or acute heat stress. These methods are not useful for our studies because they engender massive protein misfolding stress in host cells that rapidly become apoptotic, preventing influenza propagation. An alternative strategy is to over-express a constitutively active form of HSF1 lacking amino acids 186–202 (termed cHSF1), (Voellmy, 2005) but high levels of cHSF1 over-expression outside the physiologically relevant regime are typically toxic (Ryno et al., 2014). Instead, we took a chemical genetic approach, fusing cHSF1 to a destabilized variant of E. coli dihydrofolate reductase (DHFR) (Shoulders et al., 2013; Iwamoto et al., 2010). The DHFR.cHSF1 fusion is targeted for rapid proteasomal degradation and is therefore non-functional (Figure 1A), unless cells are treated with the DHFR-stabilizing pharmacologic chaperone trimethoprim (TMP). We created a stable, clonal Madin Darby canine kidney (MDCK) cell line expressing DHFR.cHSF1. In these cells, termed MDCKHSF1 cells, we can dosably induce cHSF1 transcriptional activity by TMP treatment in a stress-independent manner within the biologically relevant regime (Figure 1B), avoiding cytotoxicity that would be induced by stressors or cHSF1 overexpression and yet still providing robust access to cells expressing enhanced levels of HSF1 targets (Figure 1—figure supplement 1A–C). To control for any possible unintended consequences of TMP treatment or expression of a DHFR-fusion protein in our evolution experiments, we also created a control MDCK cell line expressing DHFR.YFP (MDCKYFP). This MDCKYFP cell line does not display TMP-dependent upregulation of HSF1-dependent chaperones, indicating that TMP induces HSF1 activity in our MDCKHSF1 cells specifically by stabilizing DHFR.cHSF1 (Figure 1—figure supplement 1B). Figure 1 with 2 supplements see all Download asset Open asset Chemical biology methods to modify the host cell’s proteostasis environment. (A) Destabilized domain technology for stress-independent control of HSF1 activity with trimethoprim (TMP). (B) Dosable induction of HSF1 activity by increasing concentrations of TMP shown by increases in Hsp70 transcripts up to physiologically relevant levels; arsenite is a positive control for endogenous HSF1 activation. Transcript levels normalized to vehicle-treated MDCKYFP cells; error bars represent SEM between biological triplicates. (C) 10 nM STA-9090 does not induce a compensatory heat shock response (representative blot shown; N = 3). Figure 1—figure supplement 1. Validation of chemical biology tools used to perturb proteostasis. Figure 1—figure supplement 2. Heat shock protein transcript expression during influenza infection in modulated proteostasis environments. https://doi.org/10.7554/eLife.28652.003 A second approach to create an altered folding environment is to inhibit individual chaperones. Here, we employed the Hsp90 inhibitor STA-9090, a small molecule capable of targeting multiple Hsp90 isoforms (Ying et al., 2012). Notably, Hsp90 inhibition using high concentrations of STA-9090 can cause an undesirable compensatory heat shock response (Moore et al., 2016), resulting in activation of HSF1 and upregulation of Hsp70, Hsp40, and other HSF1 targets. Such compensatory HSF1 activation would convolute interpretation of our results. Thus, we treated with the highest possible STA-9090 concentration that does not induce a compensatory heat shock response in MDCK cells. We selected the concentration used via a functional assay for all experiments, examining Hsp70 protein levels by immunoblotting to ensure the absence of any heat shock response signature (Figure 1C). To confirm that we are still engaging Hsp90 at the low STA-9090 concentration used in our serial passaging experiments, we performed a cellular thermal shift assay (Martinez Molina et al., 2013) and observed a small but highly reproducible increase in Hsp90 thermal stability upon STA-9090 treatment (Figure 1—figure supplement 1D). Because Hsp90 inhibition can be employed in our MDCKHSF1 cell line, these methods access three distinctive host proteostasis environments in a single cell line dependent only on small molecule treatment. The use of just a single cell line for experiments (MDCKHSF1) and for controls (MDCKYFP) minimizes any possible cell line-dependent bias in our evolution experiments. Importantly, monitoring chaperone levels in the context of influenza A/Wuhan/1995 (H3N2) infection shows that, under our infection conditions, influenza itself neither induces heat shock protein transcripts nor interferes with our method to activate HSF1. Thus, we have full user control of the host proteostasis environment during a progressing infection (Figure 1—figure supplement 2). To further characterize these perturbed host environments, we performed RNA-Seq for each treatment in the MDCKHSF1 and MDCKYFP cell lines. We observed a total of only 118 transcripts whose expression is altered ≥2 fold with a p-value <10–5 in any one or more of the treatments employed, indicating that we are remodeling only a small portion of the transcriptome (comprehensive quality control and RNA-Seq results are provided in Figure 2—source data 1–2). A heat map for these 118 genes highlights that two distinctive cellular environments are indeed created by HSF1 activation and Hsp90 inhibition in MDCKHSF1 cells (Figure 2A). Moreover, only two transcripts meet these cutoffs upon TMP treatment in MDCKYFP cells, confirming that the remodeled proteostasis environment upon TMP treatment of MDCKHSF1 cells is specifically due to HSF1 activation. Figure 2 Download asset Open asset Transcriptomic analysis of perturbed host cell proteostasis environments. (A) Differential expression analysis of MDCKHSF1 cells treated for 24 hr with 10 μM TMP (+HSF1) or 10 nM STA-9090 (+Hsp90 inhibitor) and MDCKYFP cells treated for 24 hr with 10 μM TMP (+YFP), normalized to vehicle treatment in the corresponding cell line. Transcripts displaying ≥2 fold changes in expression with p-values <10–5 for any of the treatments are included in the heat map (118 transcripts total). (B) Volcano plots showing the global distribution of expressed transcripts upon HSF1 activation or Hsp90 inhibition as in Figure 2A. Transcripts displaying ≥2 fold changes in expression with p-values <10–5 are shaded red. Outliers and transcripts encoding proteostasis network components, stress response genes, and transcription factor modulators are labeled. *=unannotated canine genes homologous to the indicated gene across multiple species; **=unannotated canine genes that fell within the indicated paralog gene family with partial homology. (C) Heat map showing the differential expression of heat shock response and unfolded protein response gene transcripts upon HSF1 activation or Hsp90 inhibition as in Figure 2A. Figure 2—source data 1. RNA-Seq characterization of MDCKHSF1 and MDCKYFP cells: quality control metrics. Figure 2—source data 2. RNA-Seq characterization of MDCKHSF1 and MDCKYFP cells: differential transcript expression. Figure 2—source data 3. RNA-Seq characterization of MDCKHSF1 and MDCKYFP cells: list of all transcripts displaying ≥2 fold changes in expression with p-values <10–5 for each treatment. https://doi.org/10.7554/eLife.28652.006 Figure 2—source data 1 RNA-Seq characterization of MDCKHSF1 and MDCKYFP cells: quality control metrics. https://doi.org/10.7554/eLife.28652.007 Download elife-28652-fig2-data1-v1.xlsx Figure 2—source data 2 RNA-Seq characterization of MDCKHSF1 and MDCKYFP cells: complete RNA-Seq datasets upon vehicle treatment, HSF1 activation, or Hsp90 inhibition in MDCKHSF1 cells and vehicle treatment or YFP activation in MDCKYFP cells. https://doi.org/10.7554/eLife.28652.008 Download elife-28652-fig2-data2-v1.xlsx Figure 2—source data 3 RNA-Seq characterization of MDCKHSF1 and MDCKYFP cells: list of all transcripts displaying ≥2 fold changes in expression with p-values <10–5 for each treatment. https://doi.org/10.7554/eLife.28652.009 Download elife-28652-fig2-data3-v1.xlsx The volcano plots in Figure 2B show the distribution of differentially expressed genes for HSF1 activation and Hsp90 inhibition in MDCKHSF1 cells, relative to the basal environment, with selected transcripts labeled (for a list of all transcripts meeting these thresholds see Figure 2—source data 3). As expected, stress-independent HSF1 activation upregulates numerous classic heat shock response genes (Ryno et al., 2014), including HSP70, BAG3, HSP90AA1, DNAJB1, and FKBP4 (Figure 2B). This remodeling of the cellular proteostasis network is limited to nuclear and cytosolic proteostasis mechanisms, as ER proteostasis network components are not induced by TMP treatment of the MDCKHSF1 cells (Figure 2C). Moreover, transcript-level proteostasis network remodeling is not observed upon STA-9090 treatment, with the exception of a 2.2-fold induction of HSPB1, as expected given the low concentration of inhibitor employed. We observe modest upregulation of several transcripts involved in transcription factor regulation and DNA damage upon HSF1 activation, such as CAMK1, RELB, and PARP3, consistent with the intimate role of HSF1 in cytoprotection (Morimoto and Santoro, 1998). STA-9090 slightly upregulates several interferon response-related transcripts, including OAS2 and MX2 (2.9 and 2.2 fold-change, respectively), a phenomenon that has been previously reported upon treatment with other Hsp90 inhibitors (Yang et al., 2006; Donzé et al., 2001; Shang and Tomasi, 2006). In summary, our RNA-Seq data are fully consistent with the chemically-controlled creation of three unique host cell proteostasis environments in MDCKHSF1 cells. Serial passaging to emulate influenza evolution We next serially passaged influenza A/Wuhan/1995 in our three distinctive, small molecule-controlled host environments to test the hypothesis that host proteostasis impacts influenza evolution. Prior to each infection, we split a population of MDCKHSF1 cells and treated with TMP to activate HSF1, STA-9090 to inhibit Hsp90, or vehicle (Figure 3A). We infected at a low multiplicity of infection (MOI), ranging from 0.001 to 0.04 infectious virions/cell, to minimize non-viable variants hitchhiking with functional variants owing to co-infection of a single cell (Figure 3—figure supplement 1A). We performed 23 serial passages in biological triplicate in each proteostasis environment. We also performed identical passaging experiments ± TMP in our MDCKYFP cells to control for any possible off-target effects of TMP treatment or expression of a DHFR-fusion protein in our evolution experiments (Figure 3B). Figure 3 with 1 supplement see all Download asset Open asset Serial passaging of Influenza A/Wuhan/1995 H3N2. (A) Serial passaging workflow in modified proteostasis environments in MDCKHSF1 cells. (B) Serial passaging workflow in control MDCKYFP cells. (C) Hemagglutination titers at intermittent passages for each folding environment; error bars represent SEM for biological triplicates. (D) Dn/Ds ratios for each viral population, normalized to the ratio of non-synonymous sites to synonymous sites in the influenza genome (3.5). Figure 3—figure supplement 1. Multiplicity of infection and hemagglutination titers during serial passaging. https://doi.org/10.7554/eLife.28652.010 Following each passage, clarified viral supernatant was harvested for hemagglutination titering and subsequent infection to initiate the next passage (Figure 3C, Figure 3—figure supplement 1B). Importantly, infectious titering at intermittent passages confirmed that the MOI did not vary systematically between the three distinctive proteostasis environments (Figure 3—figure supplement 1A). Thus, differences observed in evolutionary trajectories cannot be attributed to either systematically altered rates of viral growth in the three different host proteostasis environments studied here or to gross differences in cell health. Next, we sought to evaluate whether our serial passaging strategy provides a valid platform for modeling influenza evolution. In the absence of a strong exogenous selection pressure, such as an antiviral drug or antibody, we would predict that the influenza genome experiences purifying selection, meaning that most amino acid substitutions result in reduced fitness relative to the wild-type consensus sequence. We assessed the extent of purifying selection by determining the ratio of non-synonymous to synonymous substitutions at each passage, normalized to the ratio of non-synonymous sites to synonymous sites in the influenza genome (3.5: 1), defined as Dn/Ds. As expected, non-synonymous mutations are selected against in all folding environments, as indicated by a Dn/Ds < 1 throughout our serial passaging (Figure 3D) (Breen et al., 2012). Nature of selection pressure differs in modified proteostasis environments To comparatively evaluate the selection pressure placed on influenza by our three distinctive host proteostasis environments, we constructed variant frequency distributions (site frequency spectra; SFS) for each viral population (Nielsen, 2005). In the SFS (Figure 4A–B), we separate mutations present above our sequencing error threshold (1.5%) in a given viral population into either non-synonymous or synonymous groups. The bars in the SFS charts represent the portion of variants in a viral population that fall within a given frequency bin, averaged across biological triplicates. As expected, the resulting SFS for non-synonymous mutations (Figure 4A) show that early passages consist entirely of low frequency variants. As the passaging experiment progresses, the high mutation rate of influenza maintains a substantial population of low frequency variants. However, the proportion of low frequency variants in the viral population decreases as selected mutations increase in frequency and become fixed. This phenomenon is highlighted in later passages as variants begin to occupy high frequency bins and a bimodal, U-shaped distribution emerges. Figure 4 with 1 supplement see all Download asset Open asset Site frequency spectra show frequency distribution of non-synonymous (A) and synonymous (B) mutations in a given folding environment at a particular passage. Average between biological triplicates is plotted; error bars represent SEM. The Mann-Whitney test was used to compare the distributions of the SFS and assess statistical significance of these differences; resulting p-values are shown for passage 11 SFS. Figure 4—figure supplement 1. Trajectories for non-synonymous and synonymous mutations that increase in frequency during serial passaging. Figure 4—source data 1. List of ten highest % frequency non-synonymous mutations for each proteostasis environment and passage. Figure 4—source data 2. List of ten highest % frequency synonymous mutations for each proteostasis environment and passage. https://doi.org/10.7554/eLife.28652.012 Figure 4—source data 1 List of ten highest % frequency non-synonymous mutations for each proteostasis environment and passage. https://doi.org/10.7554/eLife.28652.014 Download elife-28652-fig4-data1-v1.pdf Figure 4—source data 2 List of ten highest % frequency synonymous mutations for each proteostasis environment and passage. https://doi.org/10.7554/eLife.28652.015 Download elife-28652-fig4-data2-v1.pdf The time resolution afforded by our serial passaging and sequencing strategy provides the opportunity to assess the strength of selection pressure on the influenza genome. Non-synonymous mutations fix latest in the Hsp90-inhibited environment and earliest in the HSF1-activated environment (Figure 4A; outlined in green). For instance, in the Hsp90-inhibited environment, no mutations have exceeded 60% frequency by passage 9 or 80% frequency by passage 11. In contrast, in the HSF1-activated environment, mutations exceed 80% frequency as early as passage 9 and become fixed (>90%) by passage 11. The basal environment lies between these two extremes, with mutations exceeding only 70% frequency by passage 9 and with no mutations yet being fixed at passage 11. As mutations in each environment increase in frequency, all the distributions achieve similar U-shaped, bimodal distributions by passage 23. For the basal and HSF1-activated environments, the distributions become U-shaped by passage 11, whereas for the Hsp90-inhibited environment, this distribution emerges only after passage 19. To assess the significance of these differences in the shape of the SFS, we applied the Mann-Whitney test, which is a statistical test for comparing two distributions (Mann and Whitney, 1947). Indeed, by passage 11, the first passage at which we observe fixed mutations in any environment, the shape of the non-synonymous SFS for influenza evolved in the Hsp90-inhibited environment is significantly different from that of influenza evolved in the basal and HSF1-activated environments (Figure 4A). This reduced rate of adaptation in Hsp90-inhibited versus basal and HSF1-activated environments is also evident from plotting individual mutation trajectories (Figure 4—figure supplement 1), revealing specific mutations that increase in frequency more gradually when Hsp90 is inhibited. In contrast, the overall shapes of the basal and HSF1-activated SFS are not significantly different. As an internal control, we also examined the SFS and mutation trajectories for synonymous variants (Figure 4B). We find that synonymous mutations for each environment have similar SFS within each passage, and that the Mann-Whitney test cannot distinguish between the synonymous SFS in different proteostasis environments at any stage of the serial passaging experiment (Figure 4B). Moreover, unlike non-synonymous mutations (see below and also see Figure 4—source data 1), specific synonymous mutations do not fix reproducibly in one particular environment (Figure 4—source data 2). This observation is consistent with our expectation that, although the influenza genome does undergo selection at the RNA level (Air et al., 1990), the host proteostasis-based perturbations of HSF1 activation and Hsp90 inhibition should primarily affect influenza evolution by modulating protein-level folding/function. The similar SFS for synonymous mutations also confirm that differences between folding environments for non-synonymous mutations cannot be attributed to altered viral growth rates. Taken together, these observations draw an intriguing picture of the impact of host proteostasis on viral evolution. On the one hand, there is little effect on strongly deleterious mutations, as illustrated by the similar Dn/Ds ratios across environments (Figure 3D). On the other hand, the rate of adaptation by accumulation of advantageous mutations is significantly reduced in the Hsp90-inhibited environment, as compared to the basal and HSF1-activated environments. This signature of altered selection pressure in distinctive host proteostasis environments highlights the importance of this under-appreciated factor impacting viral evolution. Influenza protein mutational landscapes are modulated by host proteostasis The data in Figure 4 show that the host proteostasis environment is a determinant of the nature of selection pressure placed on the influenza genome. Next, we examined how the mutational landscapes of individual influenza proteins are influenced by these same environments. We anticipated perturbations of