Influenza A Virus Compromises anti-Streptococcal Innate Immunity

Seasonal influenza epidemics pose a considerable hazard for global health. In the past decades, accumulating evidence revealed that influenza A virus (IAV) renders the host vulnerable to bacterial superinfections which in turn are a major cause for morbidity and mortality. However, whether the impact of influenza on anti-bacterial innate immunity is restricted to the vicinity of the lung or systemically extends to remote sites is underexplored. We therefore sought to investigate intranasal infection of adult C57BL/6J mice with IAV H1N1 in combination with bacteremia elicited by intravenous application of Group A Streptococcus (GAS). Co-infection in vivo was supplemented in vitro by challenging murine bone marrow derived macrophages and exploring gene expression and cytokine secretion. Our results show that viral infection of mice caused mild disease, led to persistent pulmonary immune response in the lung and induced the depletion of CCL2 in the periphery. Influenza preceding GAS infection promoted the unopposed dissemination of bacteria and their invasion into remote tissues like lung and joints and was accompanied by exacerbated sepsis. In vitro co-infection of macrophages led to significantly elevated expression of TLR2 and CD80 compared to bacterial mono-infection, whereas CD163 and CD206 were downregulated. The GAS-inducible upregulation of inflammatory genes, such as Nos2, as well as the secretion of TNFα and IL-1β were notably reduced or even abrogated following co-infection. Our results indicate that IAV primes an innate immune layout that is inadequately equipped for bacterial clearance.


33
Seasonal influenza is a major cause of respiratory disease that affects 5 -10% of the global 34 population annually with an estimated death toll of up to 500,000 [1,2]. The segmented genome 35 of influenza A virus (IAV) combined with an error-prone RNA polymerase enables the periodical 36 emergence of new strains with elevated pandemic capacities, which annually challenge human-37 kind yet devoid of adequate adaptive immunity [3,4]. The most prominent paradigm for the dra-38 matic consequences of an influenza pandemic is the 1918/1919 flu that caused roughly 50 Mil-39 lion casualties [5]. Notably, the vast majority of fatal cases were attributed to secondary bacterial 40 infections predominantly caused by pneumococci and hemolytic streptococci [6,7]. Along these 41 lines, excess morbidity due to bacterial superinfection with the nasopharyngeal colonizers S. 42 pneumoniae, S. aureus and S. pyogenes (Group A Streptococcus, GAS) was confirmed for the 43 most recent influenza pandemic in 2009 [8]. As of yet, there is neither a licensed vaccine against 44 S. aureus nor against S. pyogenes that would help contain invasive infections with these patho-45 gens during future influenza pandemics [9][10][11]. 46 Several modes by which an immune response against IAV supports viral clearance yet fails 47 software v10.7.1. Supplementary Fig. 1 illustrates the gating strategy. Live macrophages were 239 identified as 7-AAD − F4/80 + singlets. This population was used for the subsequent determination 240 of expression levels based on median fluorescence intensity (MFI) values and as a parent for 241 measuring the proportions of subpopulations expressing different combination of the above-listed 242 surface antigens. For dimension reduction, 10,000 macrophage events were down-sampled, con-243 catenated and submitted to the algorithm t-distributed stochastic neighbor embedding (t-SNE) 244 using an automated learning configuration (opt-SNE combined with the exact KNN algorithm 245 and the Barnes-Hut gradient algorithm) with a perplexity of 50 and a maximum of 1000 iterations 246 [35]. Unsupervised clustering of subpopulations expressing any combinations of the analyzed 247 surface proteins was conducted by FlowSOM [36]. 248 249

Macrophage gene expression 250
After aspirating cell culture supernatants, 700 µL of a chaotropic agent solution (Qiagen) 251 was added to individual wells and cells were lysed by scraping and vigorous shaking. RNA was 252 subsequently isolated using the RNeasy Plus Mini Kit (Qiagen) after the manufacturer's instruc-253 tions. Quantification of RNA contents were determined photometrically and 200 ng RNA was 254 submitted to reverse transcription as described above. Amplification of cDNA was then per-255 formed by TaqMan Gene Expression Master Mix, primer pairs and probes for the relative quanti-256 control groups, respectively, or by feature scaling into a 0 -1 range by the formula x i ' = (x i -x min )/(x maxx min ). Heatmaps and hierarchical clusters were generated with the "pheatmap" pack-287 age that incorporated feature scaling by standardization applying the formula z i = (x i -x̄)/σ. Cali-288 bration curves were fitted and samples values were estimated by n-parameter logistic regression 289 using the "nplr" package. Two-sided statistical tests were used for the comparisons of group me-290 dians or means. Repeated measures (body weight trajectories) were compared by one-way and 291 two-way analyses of variance (ANOVA), respectively. Probabilities of survival and incidences 292 were compared by the logrank test. Bivariate interdependencies were evaluated by the Pearson 293 product-moment correlation coefficient (r). Data sets were tested for normality by the Shapiro-294 Wilk test. Normal distribution of within-group raw or normalized variables was rejected when the 295 test resulted in a p-value of < 0.05. Depending on the outcome of this test, univariate statistical 296 analyses on variables that were normalized to respective controls were performed by the one-297 sample Wilcoxon signed rank test and the one-sample t-test, respectively. Two independent sam-298 ples were compared with the Mann-Whitney U test or the t-test. Comparisons of variables be-299 tween multiple groups were performed with the Dunn's test and the Tukey HSD test, respective-300 ly, in combination with type I error correction using the Bonferroni-Holm method. A p-value of < 301 0.05 was considered statistically significant. 302 3

Infection with influenza A virus H1N1 caused mild symptoms and induced a persistent im-304
mune reaction in the lung 305 In order to examine clinical manifestations of influenza, we used a model of intranasal in-306 fection with 2009 pandemic H1N1 IAV in adult mice (Fig. 1A). Intranasal application of PBS 307 served as a control. Mice were monitored for relative weight loss post infection as a proxy for 308 disease severity and indeed, exhibited minor reductions in body weight as early as two days after 309 virus application (Fig. 1B). This trend continued until day seven after infection and resulted in a 310 maximum weight loss of 5.5% ± 2.1% (mean ± SEM). Thereafter, body weight continuously in-311 creased and returned to starting values by day 14 suggesting robust recovery from infection. 312 When comparing weight trajectories over the entire observation period using two-way analysis of 313 variance (ANOVA), we found a statistically significant difference between infected mice and 314 uninfected controls (p < 0.001). In accordance with the observed mild disease courses, we did not 315 measure quantifiable amounts of the inflammatory cytokines TNFα and IFNγ in plasma samples 316 from IAV infected animals (not shown). We did, however detect significant reductions of plasma 317 CCL2 concentrations by 12.8% and 13.6% at days two and four after infection, respectively, rela-318 tive to uninfected controls (Fig. 1C). By day seven, CCL2 plasma levels equalized between both 319 groups (p = 0.65, not shown). 320 In order to examine immune responses in the lower respiratory tract, we further performed 321 gene expression analyses on whole lung homogenates that were obtained 16 days after IAV ap-322 plication. For this, we focused on mRNA expression levels of Ccl2 and Ifnb1, as the former was 323 altered in the periphery and the latter can be indicative of an anti-viral response. Protein data 324 were not collected because of limited sample quantities. We found no meaningful differences in 325 the expression of Ccl2 between the IAV and control groups (Fig. 1D, left panel). Interestingly, Ifnb1 expression was found to be significantly increased in lungs from infected mice (Fig. 1D, 327 right panel). Given this prolonged immune activation, we consequently utilized primer pairs for 328 the detection of viral genes in lung samples that code for hemagglutinin, matrix protein and nu-329 cleoprotein (supplementary Fig. 2). We indeed detected IAV specific RNA in 38% (3/8) of in-330 fected animals by quantitative PCR (Fig. 1E). False positive detection of unspecific targets was 331 ruled out by confirming the expected amplicon melting temperatures (supplementary Fig. 3). 332 However, the quantities of all three viral genes were generally low (C t > 32) and might rather 333 indicate residual viral antigen. 334 In summary, we here show that an infection with IAV H1N1 in mice induced minor clinical 335 manifestations that were accompanied by reductions of plasma CCL2 levels. Our data further 336 show that residual genetic material of the virus persisted in the lungs, which was accompanied by 337 an ongoing type I IFN immune response. 338 339

Group A Streptococcal sepsis was aggravated following Influenza A virus infection 340
As CCL2 is integral to bacterial control yet reduced during respiratory tract infection with 341 IAV, we sought to investigate the clinical features of IAV superimposed bacteremia. To this end, 342 we compared infection with bacteria only to co-infection models combining intranasal virus ap-343 plication with intravenous GAS infection in alternating succession ( Fig. 2A). By monitoring for 344 macroscopic symptoms following bacterial infection, we observed the occurrence of localized 345 paw inflammation (Fig. 2B). Of note, the emergence of these edemas was accelerated and more 346 frequent in post-influenza bacteremia (IAV+GAS) compared to bacterial infection only (GAS, p 347 = 0.01) and pre-influenza bacteremia (GAS+IAV, p = 0.045), respectively (Fig. 2C). In detail, 348 80% (8/10) of mice in the IAV+GAS group exhibited signs of paw inflammation already one day 349 after bacterial infection. In contrast, the incidence of paw edemas was increased to only 40% 350 (4/10) in the GAS+IAV group as opposed to 20% (2/10) in the GAS only group, and this differ-351 ence did not reach statistical significance (p = 0.33). We further analyzed eicosanoids from paw 352 extracts and found that these immunologically active lipid metabolites were upregulated in some 353 animals irrespective of the (co-)infection regimen (supplementary Fig. 4). 354 Additionally, more blood smears and knee joint capsule swabs were positive for β-355 hemolytic bacteria in co-infected mice from the IAV+GAS group (Table I), which suggested that 356 preceding influenza promoted bacterial dissemination and invasion into synovial tissues. By as-357 sessing macroscopic signs of burden as a proxy for sepsis severity (see Materials and Methods), 358 we found a significantly increased median disease score when comparing post-influenza bacte-359 remia with monocausal GAS infection (Fig. 2D). Interestingly, when correlating sepsis scores 360 with eicosanoids from paw homogenates, we found a significant relationship between the indi-361 vidual disease severity and the corresponding amounts of prostaglandins D 2 and E 2 as well as 5-362 and 12-Hydroxyeicosatetraenoic acid (supplementary Table III). Furthermore, elevated disease 363 severity in the IAV+GAS group was paralleled by a reduction in survival probability to 40% 364 compared to 80% in the GAS only group (Fig. 2E). In contrast, mice from the GAS+IAV group 365 had an only marginally decreased survival chance of 70%. However, the overall probability for a 366 fatal outcome was, according to logrank statistics, not significantly different between groups (p = 367 0.13) and this was likely due to low sample sizes and a high degree of uncertainty. 368 For our further analyses, we focused on the IAV+GAS co-infection sequence because our 369 data suggested that the clinical outcome was not different between the GAS+IAV and GAS 370 groups. We next aimed to investigate whether post-influenza GAS infection impacted on the im-371 mune activation in the lower respiratory tract. To this extent, we analyzed lung homogenates for 372 the expression of Ccl2 and Ifnb1, and compared the data from co-infected mice to GAS mono-373 infection or uninfected controls. We found that neither GAS nor IAV+GAS infection resulted in a meaningful alteration of the Ccl2 expression in the lung (Fig. 2F, left panel). Of note, lungs 375 from both mono-and co-infected mice exhibited a median 2-fold upregulation of Ifnb1 relative to 376 lungs from uninfected animals (p = 0.008 for GAS and p = 0.039 for IAV+GAS; Fig. 2F, right 377 panel). Yet, when comparing the infection regimens with each other, we found that Ifnb1 overex-378 pression was comparable between both infection groups (p = 0.93).We were curious whether the 379 bacteria are capable of disseminating from the blood into the lower respiratory tract and therefore 380 analyzed lung homogenates for the presence of GAS specific genes using quantitative PCR (sup-381 plementary Fig. 5). Indeed, we detected genomic speB in four out of nine lungs from the 382 IAV+GAS group whereas only one out of nine lungs from the GAS group was positive for this 383 bacterial gene (Fig. 2G, left panel). However, when analyzing for spy2158, only two lung ex-384 tracts from the IAV+GAS group were positive (Fig. 2G, right panel). Specific amplification was 385 again confirmed by melting curves (supplementary Fig. 6). As whole lungs were submitted to 386 chaotropic agent assisted homogenization and PCR analysis, we were not able to confirm wheth-387 er there were any vital bacteria present in these samples. 388 Collectively, our in vivo data demonstrated that a preceding IAV infection of the respiratory 389 tract aggravated bacteremia by promoting bacterial dissemination into remote tissues, localized 390 inflammation and a dysregulated host response as shown by an elevated sepsis activity. In con-391 trast, application of the virus following an already established bacteremia did not influence dis-392 ease progression and outcome. 393 394

Preceding influenza A virus infection impacted on the Group A Streptococcus induced di-395
versification of macrophage surface expression profiles 396 As our in vivo co-infection model implicated a preceding IAV infection to cause impaired 397 control of the bacterial burden following a superimposed GAS infection, we sought to explore any modification of anti-bacterial innate immunity. To this end, we chose in vitro (co-)infection 399 models of primary macrophages. In detail, murine macrophages were differentiated from bone 400 marrow cells by M-CSF stimulation and were subsequently infected with IAV, GAS or IAV and 401 GAS (Fig. 3A). We then analyzed the expression patterns of immunologically relevant surface 402 antigens by flow cytometry. In order to gain insight into differentially expressed macrophage 403 markers, we performed dimension reductions on our multiparametric data sets by t-distributed 404 stochastic neighbor embedding (t-SNE).  Fig. 7). 412 In an effort to obtain a more detailed picture of IAV-and GAS-induced immune responses, 413 we next focused on the individual expressions of macrophage surface antigens. Given the inter-414 experimental variance of macrophage cultures, median fluorescence intensities (MFI) of (co)-415 infected cells were normalized to their corresponding uninfected controls that were acquired from 416 the same donor animal (supplementary Fig. 8). Notably, expression patterns were similar within 417 each group, which resulted in a robust hierarchical clustering for IAV, GAS and IAV+GAS in-418 fected macrophages (Fig. 3C). In detail, apart from a significant upregulation of CD163 com-419 pared to both bacterial infection and co-infection, IAV had hardly any impact on the expression 420 of the investigated surface proteins (Fig. 3C, 3D). Conversely, GAS infection induced the over-421 expression of TLR2, which was even amplified following co-infection (Fig. 3D). Both the appli-cations of GAS and IAV+GAS comparably prompted an elevated production of MHCII. Alt-423 hough not statistically significant, GAS infection led to a slight downregulation of CD80, which 424 was reversed to an upregulated expression in the IAV+GAS group. Similarly, co-infection trig-425 gered a minor overexpression of CD86 that was short of reaching statistical significance due to a 426 high within-group variance (p = 0.067, compared to uninfected). The downregulation of CD163 427 as well as the attenuation of the GAS-induced CD206 upregulation in the IAV+GAS group fur-428 ther supports the notion that a preceding IAV infection led to a distinct immune response in mac-429 rophages during co-infection (Fig. 3D). 430 As a result of differentially affected expression landscapes, the proportions of distinctive 431 macrophage subpopulations shifted depending on the infection regimen (Fig. 3E). We found a 432 minor depletion of CD80 + CD86 + cells following GAS infection (p = 0.1), whereas co-infection 433 caused a significantly increased proportion of this population when compared to uninfected con-434 trols (Fig. 3F). Both bacterial mono-infection and co-infection induced an enrichment of MHCII + 435 macrophages, suggesting a retained ability of these immune cells to inform and coordinate an 436 adaptive immune response. In accordance with the altered expression profiles shown in Fig. 3D, 437 the proportions of CD163 + and CD206 + cells, respectively, were decreased upon co-infection 438 relative to GAS infection only (Fig. 3F). 439 Collectively, our data on the diversification of surface antigen expression demonstrated that 440 the immune response of macrophages towards co-infection with IAV and GAS was considerably 441 distinct from the effects that were induced by either mono-infection. Although we encountered 442 some similarities between the GAS and IAV+GAS groups, the preceding viral infection seeming-443 ly manipulated or obliterated the macrophages´ reaction towards the bacterial pathogen. 444

A preceding influenza A virus infection impaired the inflammatory capacity of macrophag-446 es during co-infection 447
In order to gain further insights into the immune response triggered by infected macrophag-448 es, we next performed analyses on mRNA expression of immune mediators by quantitative PCR 449 and measured cytokine secretion by ELISA or bead-based multiplex analysis (Fig. 4A). As illus-450 trated in Fig. 4B, the different (co-)infection regimens triggered distinct expression patterns of 451 immunomodulatory agents that resulted in strong within-group associations as shown via hierar-452 chical clustering. IAV infection was specifically characterized by a relatively higher expression 453 of Mgl2 and Tgfb1 (Fig 4C, supplementary Fig. 9). Bacterial infection, on the other hand, com-454 prehensively stimulated the overexpression of several genes that mediate an inflammatory re-455 sponse (Fig. 4B). Strikingly, co-infected macrophages mostly failed to induce a similar magni-456 tude of GAS inducible overexpression, yet upregulated Arg1 (Fig. 4B, 4C). 457 By further examining individual expressions, we found that Mgl2 was significantly reduced 458 following bacterial mono-infection and co-infection by 1.8-and 3.9-fold, respectively, compared 459 to uninfected controls (Fig. 4C). Remarkably, GAS application induced an approximately 3,000 460 fold overexpression of Nos2 that was impeded during co-infection to a mere, yet statistically sig-461 nificant, 10-fold overexpression. Furthermore, co-infection triggered the upregulation of Ccl2, 462 Cxcl2 and Tnf, which were significantly less pronounced in comparison to GAS infection only 463 (Fig. 4D). Secretion of these cytokines was mostly comparable between these groups, however 464 TNFα production by co-infected macrophages was reduced (Fig. 4E). Although both Il6 and Il10 465 were increased in the GAS and IAV+GAS group, respectively, only co-infection caused a signifi-466 cant secretion of the protein products (supplementary Fig. 9). While Ifnb1 was only upregulated 467 following GAS infection, Tgfb1 was downregulated after GAS infection as well as co-infection 468 (supplementary Fig. 9A). Of note, although the GAS-induced overexpression of Il1b was also observed in the IAV+GAS group, co-infection entirely abrogated the secretion of mature 470 which suggests that a preceding IAV infection compromised innate immune sensing of Strepto-471 cocci (Fig. 4D, 4E). 472 In summary, the expression patterns of immunologically active mediators were noticeably 473 different between GAS mono-infection and IAV+GAS co-infection, implying that prior virus 474 infection modifies anti-streptococcal immunity. 475 476 4

477
In this study we demonstrated that influenza promoted subsequent GAS-induced bactere-478 mia and allowed for an unopposed dissemination of the bacterial pathogen within the blood and 479 its migration into lungs as well as synovial tissues. Although we did not assess any alterations in 480 bone or cartilage morphology, we would like to argue that an invasion of articular tissue by GAS 481 is reminiscent of septic arthritis [37]. Indeed, we previously demonstrated that the occurrence of 482 paw edemas, which in the present study was more likely during IAV and GAS co-infection, was 483 due to bacterial colonization of both, subcutaneous and periarticular tissues and was paralleled by 484 immune cell infiltration [32]. Hence, we here show for the first time, that a preceding IAV infec-485 tion predisposes the host to severe complications during GAS blood infection. Conversely, IAV 486 infection elicited subsequent to GAS bacteremia did not aggravate disease severity, suggesting 487 that immune priming events in response to a prior viral encounter incapacitate an otherwise com-488 petent anti-bacterial immune response. 489 Influenza in humans is usually characterized by mild-to-moderate disease that is rarely le-490 thal and resolves shortly after infection [38], which was also shown in our animal model of IAV 491 inoculation. Upon entry into nasopharyngeal cavities, the virus trespasses into the mucus, invades 492 the epithelium and spreads to immune cells [39,40]. The host then recognizes parts of the viral 493 RNA genome by intracellular pattern recognitions receptors, which triggers the production of 494 several inflammatory cytokines, among them type I IFNs, that establish an anti-viral immune 495 state [41][42][43]. We have demonstrated that residual viral genes persisted for 16 days in the lungs 496 of some infected mice, which was paralleled by a continuous upregulation of Ifnb1. However, we 497 believe it to be unlikely that replicative viral particles were still present in the lungs up to this 498 point because IAV is typically cleared within a couple days following infection [44][45][46]. Type I 499 IFN can have beneficial effects during bacterial infection by promoting host resilience and by 500 preventing systemic hyperinflammation [47][48][49][50]. However, several studies advocated that the 501 consequences of type I IFN expression are detrimental for the containment of a secondary bacte-502 rial insult subsequent to influenza [20,51,52]. 503 By using a mouse strain that lacks the common IFNα/β receptor (IFNAR) in a model of 504 pneumococcal superinfection, Shahangian and colleagues demonstrated that the IAV-induced 505 IFNAR signaling led to an impaired production of the neutrophil attractants CXCL1 and CXCL2 506 [22]. They argued that, in agreement with a complementary study by Didierlaurent et al.,type I 507 IFNs desensitize subsequent TLR-mediated recognition of bacterial components by macrophages, 508 which are major producers for these chemokines [22,23]. Another work on IFNAR -/mice by 509 Nakamura and colleagues had some contrasting results concerning the impact of type I IFN sig-510 naling on pneumococcal superinfection [24]. In their study, they found that the virus and the bac-511 teria were capable of synergistically inducing an overproduction of type I IFNs, which led to an 512 impaired production of CCL2 while CXCL1/2 production was unaltered [24,53]. CCL2 supports 513 bacterial clearance by the attraction of CCR2 + monocytes to the infected tissue [54,55]. Along 514 these lines, we found in our study that CCL2 was significantly reduced in the plasma of IAV-515 infected mice and that both, monocausal bacterial infection and co-infection featured Ifnb1 over-516 expression in the lung. Hence, although the role of CCL2 during GAS infections is not yet fully 517 elucidated, we propose a mode in which a preceding influenza restricts anti-bacterial immunity 518 by limiting monocyte homing and their differentiation to macrophages not only in pulmonary 519 tissues but also in remote host compartments that would be affected during bacteremia. 520 Apart from the ramifications due to an impaired chemokinogenesis, we suspected other 521 means by which IAV dampens innate immune sensing of GAS. We hence focused on macro-522 phage immunobiology in the context of co-infection and found that the virus comprehensively 523 altered GAS-induced gene expression patterns and cytokine layout. In detail, we detected that the immune sensors CD163 and CD206 were markedly downregulated in co-infected compared to 525 GAS only infected macrophages. CD163 is an acute phase-regulated scavenger receptor that is 526 exclusively expressed by cells of the monocyte lineage and aids in the removal of potentially tox-527 ic iron complexes during intravascular hemolysis [56][57][58][59] Strikingly, a preceding IAV inoculation notably reduced the GAS-induced upregulation of 535 Nos2 while boosting Arg1 expression. Both genes code for enzymes that compete for the sub-536 strate L-Arginine, yet induce opposed immune mechanisms [70][71][72]. While nitric oxide synthase 537 2 (NOS2) provides inflammatory and bactericidal metabolites [73][74][75], arginase (ARG1) supports 538 tissue repair and immune resolution [73]. Thus, our data hint at a distortion of anti-bacterial pro-539 cesses due to a prior IAV infection. This is further corroborated by an inadequate sensing of the 540 bacterial pathogen indicated by the reduced and abolished production of TNFα and IL-1β, respec-541 tively, which was similarly shown in a model of pneumococcal superinfection [19]. Interestingly, 542 we detected for both, GAS mono-infection and superinfection an upregulation of Il1b, which 543 suggests that the incapacity of co-infected macrophages to process and secrete IL-1β is due to a 544 failure in the GAS-inducible activation of the NLRP3 inflammasome [76][77][78][79]. In fact, it was 545 shown that different variants of IAV, including a 2009 pandemic strain, were capable of thwart-546 ing IL-1β maturation by interfering with NLRP3 inflammasome assembly [80][81][82], which is cru-547 cial for innate immune sensing and coordination [83]. An IAV-mediated nullification of IL-1β 548 secretion would be of dramatic consequences during streptococcal superinfection. The absence of 549 signaling via the IL-1 receptor (IL-1R) was in fact associated with an increased susceptibility to 550 systemic GAS infection in both mice and humans [76,84,85]. Remarkably, rheumatoid arthritis 551 patients that received the IL-1R antagonist Anakinra exhibited a roughly 330-fold increased rate 552 of invasive GAS infections which included an elevated likelihood of life-threatening complica-553 tions such as necrotizing fasciitis and sepsis [85]. 554 In summary, we here describe in complementary in vivo and in vitro co-infection model 555 that IAV infection paralyses anti-streptococcal innate immunity. This finding warrants further 556 investigations on the mechanisms underlying this phenomenon that sets the stage for post-557 influenza superinfection. As an important side issue, our work underscores the importance of 558 regular vaccinations against influenza in order to avert bacterial superinfection and prevent fatal 559 invasive GAS complications [10,[86][87][88][89][90]. 560 577 Infect Dis. 2008;14: 1193-1199