CORD-19:ff6bd29c8ecd325c4edddb2b050d5f4919db6112 JSONTXT 8 Projects

Deep sequencing of primary human lung epithelial cells challenged with H5N1 influenza virus reveals 1 a proviral role for CEACAM1 2 3 Abstract 15 Current prophylactic and therapeutic strategies targeting human influenza viruses include 16 vaccines and antivirals. Given variable rates of vaccine efficacy and antiviral resistance, 17 alternative strategies are urgently required to improve disease outcomes. Here we describe the 18 use of HiSeq deep sequencing to analyze host gene expression in primary human alveolar 19 epithelial type II cells infected with highly pathogenic avian influenza H5N1 virus. At 24 hours 20 post-infection, 623 host genes were significantly up-regulated, including the cell adhesion 21 molecule CEACAM1. The up-regulation of CEACAM1 was blocked in the presence of the 22 reactive oxygen species inhibitor, apocynin. H5N1 virus infection stimulated significantly higher 23 CEACAM1 protein expression when compared to low pathogenic PR8 H1N1 virus, suggesting a 24 key role for CEACAM1 in influenza virus pathogenicity. Furthermore, silencing of endogenous 25 CEACAM1 resulted in reduced levels of proinflammatory cytokine/chemokine production, as 26 well as reduced levels of virus replication following H5N1 infection. Our study provides 27 evidence for the involvement of CEACAM1 in a clinically relevant model of H5N1 infection and 28 may assist in the development of host-oriented antiviral strategies. 29 30 IMPORTANCE 31 Highly pathogenic avian H5N1 influenza virus continues to pose a potential pandemic 32 threat. Therapeutic strategies that enhance host immune responses to promote clearance of 33 influenza virus infection are urgently needed. CEACAM1 was significantly up-regulated in 34 primary alveolar epithelial type II cells infected with H5N1 virus. It was one of 623 human genes 35 identified using HiSeq deep sequencing analysis that were up-regulated following H5N1 36 infection. Although CEACAM1 is known for its multifaceted immune regulatory role, little is 37 known about its contribution to host immunity. The results presented here show that CEACAM1 38 All rights reserved. No reuse allowed without permission. regulates the innate immune response in a clinically relevant in vitro human model of H5N1 39 infection. 40 Influenza viruses cause acute and highly contagious seasonal respiratory disease in all age 43 groups. Between 3-5 million cases of severe influenza-related illness and over 250 000 deaths are 44 reported every year. In addition to constant seasonal outbreaks, highly pathogenic avian influenza 45 (HPAI) strains, such as H5N1, remain an ongoing pandemic threat with recent WHO figures showing 46 454 confirmed laboratory infections and a mortality rate of 53%. It is important to note that humans 47 have very little pre-existing immunity towards avian influenza virus strains. Moreover, there is no 48 commercially available human H5N1 vaccine. Given the potential for H5N1 viruses to trigger a 49 pandemic (1, 2), there is an urgent need to develop novel therapeutic interventions to combat known 50 deficiencies in our ability to control outbreaks. Current seasonal influenza virus prophylactic and 51 therapeutic strategies involve the use of vaccination and antivirals. Vaccine efficacy is highly variable 52 as evidenced by a particularly severe 2017/18 epidemic and frequent re-formulation of the vaccine is 53 required to combat ongoing mutations in the influenza virus genome. In addition, antiviral resistance 54 has been reported for many circulating strains, including the avian influenza H7N9 virus that emerged 55 in 2013 (3, 4) . Influenza A viruses have also been shown to target and hijack multiple host cellular 56 pathways to promote survival and replication (5, 6). As such, there is increasing evidence to suggest 57 that targeting host pathways will influence virus replication, inflammation, immunity and pathology (6, 58 7). Alternative intervention strategies based on modulation of the host response could be used to 59 supplement the current prophylactic and therapeutic protocols. 60 While the impact of influenza virus infection has been relatively well studied in animal models 61 (8, 9) , the human cellular responses are poorly defined due to the lack of available human autopsy 62 material, especially from HPAI virus-infected patients. In the present study, we characterized influenza 63 virus infection of primary human alveolar epithelial type II (ATII) cells isolated from normal human 64 lung tissue donated by patients undergoing lung resection. ATII cells are a physiologically relevant 65 infection model as they are a major target for influenza A viruses when entering the respiratory tract 66 (10). Human host gene expression following HPAI H5N1 virus (A/Chicken/Vietnam/0008/04) 67 infection of primary ATII cells was analyzed using Illumina HiSeq deep sequencing. In order to gain a 68 better understanding of the mechanisms underlying modulation of host immunity in an anti-69 inflammatory environment, we also analyzed changes in gene expression following HPAI H5N1 70 infection in the presence of the reactive oxygen species inhibitor (ROS), apocynin, a compound known 71 to interfere with NADPH oxidase subunit assembly (5, 6). 72 Our HiSeq analysis described herein focused on differentially regulated genes following H5N1 73 infection and identified carcinoembryonic-antigen (CEA)-related cell adhesion molecule 1 74 (CEACAM1) as a key gene of interest. CEACAM1 (also known as BGP or CD66) is expressed on 75 epithelial and endothelial cells (11), as well as B cells, T cells, neutrophils, NK cells, macrophages and 76 dendritic cells (DCs) (12) (13) (14) . Human CEACAM1 has been shown to act as a receptor for several 77 human bacterial and fungal pathogens, including Haemophilus influenza, Escherichia coli, Salmonella 78 typhi and Candida albicans, but has not as yet been implicated in virus entry (15) (16) (17) . There is however 79 emerging evidence to suggest that CEACAM1 is involved in host immunity as enhanced lymphocyte 80 expression was detected in pregnant women infected with cytomegalovirus (18) and in cervical tissue 81 isolated from patients with papillomavirus infection (19) . 82 Eleven CEACAM1 splice variants have been reported in humans (20). CEACAM1 isoforms 83 (Uniprot P13688-1 to -11) can differ in the number of immunoglobulin-like domains present, in the 84 presence or absence of a transmembrane domain and/or the length of their cytoplasmic tail (i.e. L, long 85 or S, short). The full-length human CEACAM1 protein (CEACAM1-4L) consists of four extracellular 86 domains (one extracellular immunoglobulin variable-region-like (IgV-like) domain and three 87 immunoglobulin constant region 2-like (IgC2-like) domains), a transmembrane domain, and a long (L) 88 cytoplasmic tail. The long cytoplasmic tail contains two immunoreceptor tyrosine-based inhibitory 89 motifs (ITIMs) that are absent in the short form (20) Table S1 ). HPAI H5N1 infection of ATII cells 124 activated an antiviral state as evidenced by the up-regulation of numerous interferon-induced genes, 125 genes associated with pathogen defense, cell proliferation, apoptosis, and metabolism (Table 1; Table 126 S2). In the apocynin alone-treated (HA) group, a large number of genes were also significantly up-127 regulated (509 genes) or down-regulated (782 genes) ( Fig. 1A ; Table S1 ) relative to the control group. 128 Whilst a subset of genes was differentially expressed in both the HD and HA groups, a majority of 129 Genes down-regulated by apocynin include those that are involved in cell adhesion (GO:0007155), 135 regulation of cell migration (GO: 0030334), regulation of cell proliferation (GO:0042127), signal 136 transduction (GO: 0007165) and oxidation-reduction processes (GO:0055114) (Table S2 , "ND vs HA 137 Down"; Down, down-regulation). 138 A total of 623 genes were up-regulated following H5N1 infection ("ND vs HD Up", Fig. 1D ). 139 By overlapping the two lists of genes from "ND vs HD Up" and "HD vs HA Down", 245 genes were 140 shown to be down-regulated in the presence of apocynin (Fig. 1D ). By overlapping three lists of genes 141 from "ND vs HD Up", "HD vs HA Down" and "ND vs HA Up", 55 genes out of the 245 genes were 142 present in all three lists (Fig. 1E ), indicating that these 55 genes were significantly inhibited by 143 apocynin but to a level that was still significantly higher than that in uninfected cells. The 55 genes 144 include those involved in influenza A immunity (hsa05164; DDX58, IFIH1, IFNB1, MYD88, PML, 145 STAT2), Jak-STAT signaling (hsa04630; IFNB1, IL15RA, IL22RA1, STAT2), cytokine-cytokine 146 receptor interaction (hsa04060; CX3CL1, IFNB1, IL15RA, IL22RA1) and RIG-I-like receptor 147 signaling (hsa04622; DDX58, IFIH1, IFNB1) (Table S3 and S4) . Therefore, critical immune responses 148 induced following H5N1 infection were not dampened following apocynin treatment. The remaining 149 190 of 245 genes excluded from the "ND vs HA Up" list were those significantly inhibited by apocynin 150 to a level that was similar to uninfected control cells (Fig. 1E ). The 190 genes that were reduced by 151 apocynin to basal levels were those involved in PI3K-Akt signaling (hsa04151; CCND1, GNB4, 152 IL2RG, IL6, ITGA2, JAK2, LAMA1, MYC, IPK3AP1, TLR2, VEGFC), cytokine-cytokine receptor 153 interaction (hsa04060; VEGFC, IL6, CCL2, CXCL5, CXCL16, IL2RG, CD40, CCL5, CCL7, IL1A), 154 TNF signaling (hsa04668; CASP10, CCL2, CCL5, CFLAR, CXCL5, END1, IL6, TRAF1, VEGFC) 155 and apoptosis (hsa04210; BAK1, CASP10, CFLAR, CTSO, CTSS, PARP3, TRAF1) (Table S3 and 156 S4) . This is consistent with the role of apocynin in reducing inflammation and apoptosis (29). By overlapping the three lists of genes from "ND vs HD Up", "HD vs HA Down" and "ND vs 158 HA Down", eleven genes were found in all three comparisons. This suggests that these 11 genes are 159 upregulated following H5N1 infection and are significantly reduced by apocynin treatment to a level 160 lower than that observed in uninfected control cells (Fig. 1F ). Among these were inflammatory 161 cytokines/chemokines genes, including CXCL5, IL1A, AXL (a member of the TAM receptor family of 162 receptor tyrosine kinases) and TMEM173/STING (Stimulator of IFN Genes) ( Table S4) . and qRT-PCR analysis (Fig. 2E ). Apocynin treatment further increased SOCS1 mRNA expression (Fig. 172 2E) . Although HiSeq analysis did not detect a statistically significant increase of SOCS1 following 173 apocynin treatment, the Log2 fold-changes in SOCS1 gene expression were similar between the HD 174 and HA groups (4.8-fold vs 4.0-fold) (Fig. 2E) . SOCS3 transcription was also found to be significantly 175 increased following H5N1 infection alone following HiSeq analysis and was further increased after 176 apocynin treatment (Fig. 2F ). In contrast, SOCS3 mRNA was only slightly increased when samples 177 were analyzed by qRT-PCR (Fig. 2F) . 178 Apocynin, a compound that inhibits expression of ROS, has been shown to influence influenza-179 specific responses in vitro (5) and in vivo (6). Although virus titers are not affected by apocynin 180 treatment in vitro (5), some anti-viral activity is observed in vivo when mice have been infected with a 181 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint low pathogenic A/HongKong/X31 H3N2 virus (6). Our HiSeq analysis of HPAI H5N1 virus gene 182 transcription ( Fig. 2G ) showed that although there was a trend for increased influenza virus gene 183 expression following apocynin treatment, with only influenza non-structural (NS) gene expression 184 statistically significant. 185 186 Enrichment of antiviral and immune response genes in HPAI H5N1-infected ATII cells. 187 GO enrichment analysis was performed on genes that were significantly upregulated following HPAI 188 H5N1 infection in ATII cells in the presence or absence of apocynin to identify over-presented GO 189 terms. Many of the H5N1-upregulated genes were broadly involved in defense response 190 (GO:0006952), response to external biotic stimulus (GO:0043207), immune system processes 191 (GO:0002376), cytokine-mediated signaling pathway (GO:0019221) and type I interferon signaling 192 pathway (GO:0060337) ( Table 1; Table S2 ). In addition, many of the H5N1-upregulated genes mapped 193 to metabolic pathways (hsa01100), cytokine-cytokine receptor interaction (hsa04060), Influenza A 194 (hsa05164), TNF signaling (hsa04668) or to Jak-STAT signaling (hsa04630) (Table S3 ). However, not 195 all the H5N1-upregulated genes in these pathways were inhibited by apocynin treatment as mentioned 196 above ( Table 2; CEACAM5, were not affected by H5N1 infection (Fig. 3B ). It is also worth noting that more reads were 203 obtained for CEACAM5 (>1000 FPKM) than CEACAM1 (~ 7 FPKM) in uninfected ATII cells (Fig. 3A 204 and B), which is consistent with their normal expression patterns in human lung tissue (30). Although 205 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint CEACAM1 forms heterodimers with CEACAM5 (23), the higher basal expression of CEACAM5 in 206 ATII cells may explain why its expression was not altered by H5N1 infection. Apocynin treatment 207 significantly reduced transcription levels of both CEACAM1 (Fig. 3A) and CEACAM5 (Fig. 3B ). 208 Expression of 11 CEACAM1 variants in ATII and/or A549 cells was further characterized using qRT-209 PCR using primer pairs designed to specifically detect each variant (Table S5) difference in CEACAM1 protein levels were observed at various MOIs (2, 5 or 10), CEACAM1 222 protein expression at 48 hpi was significantly higher than that at observed at 24 hpi (Fig. 3E) . 223 After confirming the upregulation of CEACAM1 protein expression following infection with 224 the low pathogenic PR8 virus in A549 cells, CEACAM1 protein expression was then examined in 225 primary human ATII cells infected with HPAI H5N1 and compared to PR8 virus infection (Fig. 3F) . 226 Lower MOIs of 0.5, 1 and 2 HPAI H5N1 were tested due to the strong cytopathogenic effect it causes 227 at higher MOIs. Endogenous CEACAM1 protein levels were significantly elevated in H5N1-infected 228 ATII cells at all three MOIs tested. Similar levels of CEACAM1 protein expression were observed at 229 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint MOIs of 0.5 and 1 and were higher at 48 hpi when compared to 24 hpi (Fig. 3F ). HPAI H5N1 virus 230 infection at MOIs of 0.5, 1 and 2 stimulated higher endogenous levels of CEACAM1 protein 231 expression when compared to low pathogenic PR8 H1N1 virus infection at a MOI of 2 (a maximum 232 23-fold increase induced by H5N1 at MOIs of 0.5 and 1, 48 hpi when compared to PR8 at MOI of 2), 233 suggesting a possible role for CEACAM1 in influenza virus pathogenicity (Fig. 3F) . (data not shown). CEACAM1 protein expression was reduced by approximately 50% in both ATII and 250 A549 cells following siCEACAM1 transfection when compared with siNeg-transfected cells (Fig. 4C) . 251 It is important to note that the anti-CEACAM1 antibody only detects L isoforms based on 252 epitope information provided by Abcam. Therefore, observed reductions in CEACAM1 protein 253 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint expression can be attributed mainly to the abolishment of CEACAM1-4L isoform. Increasing doses of 254 siCEACAM1 (10, 15 and 20 pmol) did not further down-regulate CEACAM1 protein expression in 255 A549 cells. As such, 15 pmol of siCEACAM1 was chosen for subsequent knockdown studies in both 256 ATII and A549 cells (Fig. 4C) . 257 The functional consequences of CEACAM1 knockdown were then examined in ATII and A549 258 cells following H5N1 infection. IL-6, IFN-β, CXCL10, CCL5, and TNF production was analyzed in 259 H5N1-infected A549 and ATII cells using qRT-PCR. ATII (Fig. 5A ) and A549 cells ( The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint modulation of reactive oxygen species (ROS) to reduce inflammation (5) and inhibition of NFκB and 278 mitogenic Raf/MEK/ERK kinase cascade activation to suppress viral replication (34, 35). These host 279 targeting strategies will offer an alternative to current interventions that are focused on targeting the 280 virus. In the present study, we analyzed human host gene expression profiles following HPAI H5N1 281 infection and treatment with the antioxidant apocynin. As expected, genes that were significantly 282 upregulated following H5N1 infection were involved in biological processes including cytokine 283 signaling, immunity and apoptosis. In addition, H5N1-upregulated genes were also involved in 284 regulation of protein phosphorylation, cellular metabolism and cell proliferation, which are thought to 285 be exploited by viruses for replication (36). Apocynin treatment had both anti-viral (Table S2 , S3, S4) 286 (6) and pro-viral impact (Fig. 2G) , which is not surprising as ROS are potent microbicidal agents, as 287 well as important immune signaling molecules at different concentrations (37). In our hands, apocynin 288 treatment reduced H5N1-induced inflammation, but unavoidably impacted the cellular defense 289 response, cytokine production and cytokine-mediated signaling. Importantly, critical antiviral responses 290 were not compromised, i.e. the expression of pattern recognition receptors (e.g. DDX58 (RIG-I), TLRs, 291 IFIH1 (MDA5)) was not down-regulated (Table S1 ). Given the significant interference of influenza 292 viruses on host immunity, we focused our attention on key regulators of the immune responses. 293 Through HiSeq analysis, we identified the cell adhesion molecule CEACAM1 as a critical regulator of 294 immunity. Knockdown of endogenous CEACAM1 not only reduced H5N1-stimulated inflammatory 295 cytokine/chemokine production but also inhibited virus replication. 296 H5N1 infection resulted in significant upregulation of a number of inflammatory 297 cytokines/chemokines genes, including AXL and STING, which were significantly reduced by 298 apocynin treatment to a level lower than that observed in uninfected cells (Table S4 ). It has been 299 previously demonstrated that treatment of PR8 H1N1-infected mice with anti-AXL antibody 300 significantly reduces lung inflammation and virus titers (38). STING has been shown to be important 301 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint for promoting anti-viral responses, as STING-knockout THP-1 cells produce less type I IFN following 302 influenza A virus infection (39). Reduction of STING gene expression or other anti-viral factors (e.g. 303 IFNB1, MX1, ISG15; Table S1 ) by apocynin, may in part, explain the slight increase of the 304 transcription levels of all influenza genes (NS gene was significantly upregulated) following apocynin 305 treatment (Fig. 2G) . These results also suggest that apocynin treatment may reduce H5N1-induced 306 inflammation and apoptosis. Indeed, the anti-inflammatory and anti-apoptotic effects of apocynin have 307 been shown previously in a number of disease models, including diabetes mellitus (40) could suggest the possibility that upregulation of CEACAM1 (to inhibit NK activity) may be a novel 324 and uncharacterized immune evasion strategy employed by influenza viruses. Our laboratory is now 325 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. In support of this notion, we did not observe a difference in apoptosis of PR8-infected A549 cells 338 following transfection with siCEACAM1, siNeg, or mock-transfected cells using flow cytometry-based 339 Annexin V/7AAD staining (data not shown). Our study, together with the CEACAM6 findings, serve 340 to highlight the emerging role of CEACAMs in influenza virus infection. 341 The molecular mechanism of CEACAM1 action following infection has also been explored in 342 A549 cells using the low pathogenic PR8 virus (51) (18) and 359 papillomavirus (19), it will be important to determine whether a common mechanism of action can be 360 attributed to CEACAM1 in order to determine its functional significance. If this can be established, 361 CEACAM1 could be used as a target for the development of a pan-antiviral agent. Furthermore, it has 362 been shown that CEACAM1-S isoforms and soluble isoforms (CEACAM1-4C1, -3, -3C2) also have 363 regulatory effects (stimulatory or inhibitory) on CEACAM1-L isoforms (20, 52). Although 364 CEACAM1-Ls are the dominate isoforms, the involvement of other CEACAM1 isoforms in influenza 365 immunity should also be further investigated, especially as H5N1 infection increased mRNA 366 expression levels of several CEACAM1 variants (Fig. 3C) . 367 In summary, molecules on the cell surface such as CEACAM1 are particularly attractive 368 candidates for therapeutic development, as drugs do not need to cross the cell membrane in order to be 369 effective. Targeting of host-encoded genes in combination with current antivirals and vaccines may be 370 a way of reducing morbidity and mortality associated with influenza virus infection. Our study clearly 371 demonstrates that increased CEACAM1 expression is observed in primary human ATII cells infected 372 with highly pathogenic H5N1 influenza virus. Importantly, knockdown of CEACAM1 expression 373 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. DNase I (0.5 mg/mL; Roche Diagnostics) for 60 min at 37°C. Single cell suspensions were obtained by 389 filtration through a 40 μm cell strainer and cells were allowed to attach to a petri dish in a 1:1 mixture 390 of DMEM/F12 medium (Gibco) and small airway growth medium (SAGM) medium (Lonza) 391 containing 5% fetal calf serum (FCS) and 0.5 mg/mL DNase I for 2 hours at 37°C. The non-adherent 392 cells, including ATII cells, were collected and subjected to centrifugation at 300 g for 20 min on a 393 discontinuous Percoll density gradient (1.089 and 1.040 g/mL). Purified ATII cells from the interface 394 of two density gradients was collected, washed in HBSS, and re-suspended in SAGM media 395 supplemented with 1% charcoal-filtered FCS (Gibco) and 100 units/mL penicillin and 100 µg/mL 396 streptomycin (Gibco), prior to plating on polyester Transwell inserts (0.4 μm pore; Corning) coated 397 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint with type IV human placenta collagen (0.05 mg/mL; Sigma) at 300,000 cells/cm 2 . Cells were then 398 cultured under liquid-covered conditions in a humidified incubator (5% CO 2 , 37°C) and growth 399 medium was changed every 48 hours. These culture conditions encourage ATII cells to grow and 400 differentiate into alveolar epithelial cells. trypsin (Worthington) was included in media post-inoculation to assist replication. Virus titers were 420 determined using standard plaque assays using MDCK cells as previously described (56). 421 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint RNA extraction, quality control (QC) and HiSeq analysis. ATII cells from three individual 422 donors were used for HiSeq analysis. Total RNA was extracted from cells using a RNeasy Mini kit 423 (Qiagen). Influenza-infected cells were washed with PBS three times and cells lysed with RLT buffer 424 supplemented with β-mercaptoethanol (10 μL/mL; Gibco). Cell lysates were homogenized with 425 QIAshredder columns followed by on-column DNA digestion with the RNase-Free DNase Set 426 (Qiagen), and RNA extracted according to manufacturer's instructions. Initial QC was conducted to 427 ensure that the quantity and quality of RNA samples for HiSeq analysis met the following criteria; 1) 428 RNA samples had OD260/280 ratios between 1.8 and 2.0 as measured with NanoDrop TM 429 Spectrophotometer (Thermo Scientific); 2) Sample concentrations were at a minimum of 100 ng/μl; 3) 430 RNA was analyzed by agarose gel electrophoresis. RNA integrity and quality were validated by the 431 presence of the sharp clear bands of 28S and 18S ribosomal RNA, with a 28S:18S ratio of 2:1, along 432 with the absence of genomic DNA and degraded RNA. As part of the initial QC and as an indication of 433 consistent H5N1 infection, parallel quantitative real-time reverse transcriptase PCR (qRT-PCR) was 434 performed as previously described (5) to measure mRNA expression of IL-6, CXCL10, CCL5, 435 suppressor of cytokine signaling (SOCS) 1 and SOCS3 (Fig. 2) , which is known to be up-regulated 436 following HPAI H5N1 infection of A549 cells (5). RNA samples were stored in 0.1 volumes of 3 M 437 Sodium Acetate (pH7.5 in DEPC-treated water) and 2 volumes of 100% Ethanol and submitted to 438 Macrogen Inc. (Seoul, Republic of Korea) for HiSeq analysis. 439 Sequencing analysis and annotation. After confirming checksums and assessing raw data 440 quality of the FASTQ files with FASTQC, RNA-Seq reads were processed according to standard 441 Tuxedo pipeline protocols (57), using the annotated human genome (GRCh37, downloaded from 442 Illumina iGenomes) as a reference. Briefly, raw reads for each sample were mapped to the human 443 genome using TopHat2, sorted and converted to SAM format using Samtools and then assembled into 444 transcriptomes using Cufflinks. Cuffmerge was used to combine transcript annotations from individual 445 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint samples into a single reference transcriptome, and Cuffquant was used to obtain per-sample read 446 counts. Cuffdiff was then used to conduct differential expression analysis. All programs were run using 447 recommended parameters, however the reference gtf file provided to cuffmerge was first edited using a 448 custom python script to exclude lines containing features other than exon/cds, and contigs other than 449 chromosomes 1-22, X, Y. 450 Gene ontology (GO) and KEGG enrichment. Official gene IDs for transcripts that were 451 differentially modulated following HPAI H5N1 infection with or without apocynin treatment were 452 compiled into three target lists. Statistically significant differentially expressed transcripts were defined 453 as having ≥ 2-fold change with a Benjamini-Hochberg adjusted P value < 0.01. A background list of 454 genes was compiled by retrieving all gene IDs identified from the present HiSeq analysis with FPKM > 455 1. Biological process GO enrichment was performed using GOrilla comparing unranked background 456 and target lists (58) and redundant GO terms were removed using REVIGO (59). The target lists were 457 also subjected to KEGG pathway analysis using a basic KEGG pathway mapper (60). (Table S5 ) were designed to estimate the expression of CEACAM1 467 variants in ATII and A549 cells using iTaq Universal SYBR Green Supermix (BioRad) according to 468 manufacturer's instruction. The absence of nonspecific amplification was confirmed by agarose gel 469 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint electrophoresis of qRT-PCR products (15 μL). Gene expression was normalized to β-actin mRNA 470 using the 2 -ΔΔCT method where expression levels were determined relative to uninfected cell controls. 471 qRT-PCR primers and primer pairs used to identify each CEACAM1 variant are listed in Table S5 cell adhesion molecule of the immunoglobulin superfamily, on human lymphocytes: structure, 555 expression and involvement in T cell activation. Eur J Immunol 28:3664-3674. 556 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint and chemokine production in response to H5N1 infection (MOI of 0.01, 24 hpi) by qRT-PCR com-775 pared to siNeg-transfected control cells. One sample t test compared to siNeg-transfected cells that 776 have a fold-change of 1, *p < 0.05 and **p < 0.01. Plaque assay of H5N1 virus titres in ATII (C) and 777 A549 (D) that were either mock-transfected or transfected with siCEACAM1 or siNeg. *p < 0.05 and 778 **p < 0.01. 779 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. . https://doi.org/10.1101/324723 doi: bioRxiv preprint TABLE 2. Representatives of over-represented KEGG pathways and the number of genes contributing 784 to each pathway that is significantly up-regulated following H5N1 infection ("ND vs. HD Up"). The 785 number of genes in the same pathway expressed at a significantly lower level in the HA group are also 786 listed (Fig. 1D , "ND vs. HD Up and HD vs. HA Down"). The full list of over-represented KEGG 787 pathways is presented in Table S3 . hsa04622 RIG-I-like receptor signaling pathway 11 5 hsa04514 Cell adhesion molecules (CAMs) 11 4 All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder.