The In Vivo Source of Type I and Type III IFNs is Pathogen Dependent

Type I (-α, β) and type III (-λ) interferons (IFNs) are produced in response to virus infection and upregulate a largely overlapping set of IFN stimulated genes which mediate the protective effects of these antiviral cytokines. In vitro studies have demonstrated the redundancy of these two cytokine families which activate the same transcription factor, IFN stimulated gene factor 3 (ISGF3), via distinct ligands and receptors. However, in vivo, these IFN types do have distinct functions based on receptor distribution, but also ligand availability. Using a newly generated IFN-λ reporter mouse strain we have observed that both type I and type III IFNs are produced in response to respiratory tract infection by Newcastle disease virus (NDV) and influenza A virus (IAV). In the case of NDV these IFNs are synthesized by different cell types. Type I IFNs are produced primarily by alveolar macrophages, type III IFNs are made only by epithelial cells, and production of either is dependent on MAVS. While epithelial cells of the respiratory tract represent the primary target of IAV infection, we found that they did not significantly contribute to IFN-λ production, and IFN-λ protein levels were largely unaffected in the absence of MAVS. Instead we found that pDCs, a cell type known for robust IFN-α production via TLR/MyD88 signaling, were the major producers of IFN-λ during IAV infection, with pDC depletion during influenza infection resulting in significantly reduced levels of both IFN-α and IFN-λ. In addition, we were able to demonstrate that pDCs rely on type I IFN for optimal IFN-λ production. These studies therefore demonstrate that the in vivo producers of Type III IFNs in response to respiratory virus infection are pathogen dependent, a finding which may explain the varying levels of cytokine production induced by different viral pathogens.


Abstract 22
Type I (-α, β) and type III (-λ) interferons (IFNs) are produced in response to virus 23 infection and upregulate a largely overlapping set of IFN stimulated genes which 24 mediate the protective effects of these antiviral cytokines. In vitro studies have 25 demonstrated the redundancy of these two cytokine families which activate the same 26 transcription factor, IFN stimulated gene factor 3 (ISGF3), via distinct ligands and 27 receptors. However, in vivo, these IFN types do have distinct functions based on 28 receptor distribution, but also ligand availability. Using a newly generated IFN-λ reporter 29 mouse strain we have observed that both type I and type III IFNs are produced in 30 response to respiratory tract infection by Newcastle disease virus (NDV) and influenza 31 A virus (IAV). In the case of NDV these IFNs are synthesized by different cell types. 32 Type I IFNs are produced primarily by alveolar macrophages, type III IFNs are made 33 only by epithelial cells, and production of either is dependent on MAVS. While epithelial 34 cells of the respiratory tract represent the primary target of IAV infection, we found that 35 they did not significantly contribute to IFN-λ production, and IFN-λ protein levels were 36 largely unaffected in the absence of MAVS. Instead we found that pDCs, a cell type 37 known for robust IFN-α production via TLR/MyD88 signaling, were the major producers 38

Introduction 45
The antiviral activity of the type I interferons (IFN-α/β) has been appreciated since their 46 discovery in 1957 [1,2]. The more recent discovery of type III IFNs (IFN-λ)

Generation and Characterization of the IFN-λ reporter mouse 87
To identify IFN-λ-producing cell populations in vivo, we generated an IFN-λ reporter 88 mouse by replacing the coding region of ifnl2 with an eGFP-Neo cassette through 89 homologous recombination, while maintaining the ifnl2 promoter and UTR regions intact 90 ( Fig 1A). The FRT sites flanking the neo gene allowed its removal by crossing mice 91 carrying the targeted allele with mice expressing the FLP-recombinase to generate 92 animals lacking the Neo gene while simultaneously carrying the GFP-targeted allele. Mice 93 heterozygous for the eGFP-targeted allele were crossed to generate mice homozygous 94 for the targeted allele. Homozygous ifnl2 eGFP/eGFP knockin mice can no longer produce 95 IFN-λ2, but still retain functional ifnl3 gene loci. Genotypes were determined by 96 conventional PCR using primers targeting the IFN-λ locus and the eGFP transgene (Fig  97   1A, B). 98 Human pDCs secrete high levels of both type I and type III IFNs when stimulated by 99 TLR ligands or virus infection [11], so characterization of the reporter strain was initiated 100 by Newcastle disease virus (NDV) infection of Flt3-ligand (FLT3L) cultured bone marrow 101 derived dendritic cells. NDV is an avian paramyxovirus known to induce high levels of 102 type I IFNs in many cell types including murine pDCs [5,12], so we used this system to 103 ask whether GFP expression correlated with IFN-λ production upon induction by virus 104 infection. 105 GFP expression following virus inoculation was detected by fluorescence microscopy 106 (Fig 1C), as well as flow cytometry (Fig 1D). A Fluorescence Activated Cell Sorter (FACS) 107 was used to separate NDV-treated, FLT3L-cultured DCs into GFP-and GFP+ populations 108 for RNA extraction and subsequent qRT-PCR analysis of IFN-λ3 and eGFP transcripts 109 ( Fig 1D). This analysis confirmed that IFN-λ3 and eGFP transcripts were enriched only 110 in GFP+ cells, demonstrating that GFP fluorescence from reporter cells reflected  We also explored GFP induction following virus infection of epithelial cells. Primary 113 kidney epithelial cells derived from ifnl2 gfp/gfp reporter mice were cultured in vitro and 114 infected with NDV, respiratory syncytial virus (RSV), rhesus rotavirus (RRV) or influenza 115 A virus (Fig 1E). At 24 hours post-infection, all infected cultures contained GFP+ cells,116 but for all viruses the percentage of GFP+ cells was relatively small, even at higher MOIs. 117 A similar effect was observed in FLT3L cultured DCs (data not shown), suggesting that 118 during a virus infection, not all cells capable of producing IFN-λ go on to do so as has 119 been found for . 120 121 A small population of CD8α DCs produce IFN-λ in response to poly I:C treatment in 122 vivo 123 Production of IFN-λ by different cell types has been detected in response to a wide 124 variety of stimuli or pathogens in vitro [14][15][16]. Taken together, these publications 125 demonstrate that many cell types have the capacity to produce type III IFNs under the 126 appropriate conditions. However, reports investigating the cellular sources of IFN-λ in a 127 whole animal in response to TLR ligands or virus infections are few. One such study 128 identified splenic CD8α+ dendritic cells as the main producers of IFN-λ in mice that were 129 treated intravenously (i.v.) with the TLR3 agonist, poly I:C [17]. We carried out a similar 130 experiment using our IFN-λ reporter mouse model to determine whether we could confirm 131 that result by following GFP expression. Cohorts of Ifnl2 gfp/gfp reporter mice were injected 132 i.v. with 100 μg of poly I:C or PBS, and their splenocytes harvested six hours later. Using 133 antibodies to the cell surface markers CD11b, CD11c, CD8α, B220, and Siglec H, we 134 identified pDC, CD8α DC, and cDC subsets, and found, as expected, that GFP 135 expression was mainly confined to the CD8α DC populations (Fig 2) reporter mouse strain that allows visualization and quantification of IFN responses by 154 measuring luciferase activity driven by the promoter of the interferon-stimulated gene Mx2 155 [21]. By this approach, a robust interferon response was detected as early as 12 hours 156 post-infection, peaking at 24 hours, and subsiding after 48 hours ( Fig 3A&B). Consistent 157 with this observation, IFN-λ protein was detected in the small intestine of rotavirus-158 infected WT 129SvEv mice 24 hours post-infection by ELISA (Fig 3C). To determine the 159 source of IFN-λ induced by RRV, intestines were harvested from infected IFN-λ reporter 160 mice 24 hours post-inoculation and analyzed by immunohistochemistry. Formalin-fixed, 161 paraffin-embedded tissue sections from mock-infected or RRV-infected IFN-λ reporter 162 mice were stained with antibodies against GFP and RRV antigens ( Fig 3E). GFP+ cells 163 were present only in tissues harvested from RRV-infected IFN-λ reporter mice, and 164 absent from the intestines of mock-infected IFN-λ reporter mice, or RRV-infected WT 165 129SvEv animals. GFP expression in RRV-infected IFN-λ reporter mice was found 166 exclusively in the intestinal epithelial cells. RRV-infected epithelial cells were found in 167 both RRV-infected IFN-λ reporter mice and RRV-infected WT controls, confirming 168 successful infection of these animals. These results support the hypothesis that epithelial 169 cells are the primary source of IFN-λ during RRV infection. As IFN-λ reporter mice lack 170 the ifnl2 locus but have an intact ifnl3 locus, we compared RRV titers from the small 171 intestine of IFN-λ reporter and WT 129SvEv animals, and found no difference in the 172 amount of virus detected (Fig 3D)

Airway epithelial cells produce IFN-λ in response to NDV infection in vivo 178
Next, we sought to investigate IFN-λ production during virus infection of the respiratory 179 tract, another mucosal compartment that serves as a major portal for virus entry. Our lab 180 and others have reported that both type I and type III IFNs are produced in response to 181 respiratory virus infections [22,23]. In the lung, as in the intestine, both IFN types appear 182 to be important for antiviral protection of the respiratory tract, as mice deficient in both As with our rotavirus studies, we looked for GFP+ cells in tissue sections prepared from 197 IFN-λ reporter mice that were mock or NDV-infected. As before, lung tissue sections were 198 stained using antibodies against GFP and viral antigens ( Fig 4D). This histological 199 analysis confirmed our flow cytometry results, showing that GFP expression was limited 200 to the columnar epithelial cells that line the bronchi and bronchioles of the lung. The same 201 cell type expressed both NDV and GFP protein in infected lungs. Interestingly, while rare 202 cells stained with antibodies to both GFP and NDV proteins, most cells in the infected 203 airway showed either GFP or NDV staining. To be certain that elimination of IFN-λ2 in our 204 reporter mouse was not skewing our results, we compared total IFN-α and IFN-λ 205 production in BAL samples from NDV-infected WT and reporter mice and saw no 206 significant difference between the strains (Fig 5). 207 208

Immune cells are not a significant source of IFN-λ during NDV infection in vivo 209
IFN-λ synthesis by dendritic cells and monocytes exposed to viruses and pattern 210 recognition receptor (PRR) agonists in vitro has been reported [7,8,11,28]. So while 211 results from our flow cytometry and histology studies using the IFN-λ reporter mice clearly 212 demonstrated GFP expression from bronchial epithelial cells during NDV infection in vivo, 213 we wished to determine whether there was also a contribution from innate immune cells 214 in this model. To do this we used a panel of antibodies targeting cell surface markers to 215 assess IFN-λ production, or GFP expression, by neutrophils (CD45 + F4/80 -Ly6G + ), 216 eosinophils (CD45 + F4/80 + SiglecF + CD11c-), monocyte-derived macrophages (CD45 + 217 F4/80 + SiglecF -CD11c + ) and alveolar macrophages (CD45 + F4/80 + Siglec F + CD11c + ) 218 purified from the lungs of NDV-infected IFN-λ reporter mice 24 hours post-infection. GFP 219 expression was minimal in all of these cell types (Fig 6), and therefore it is unlikely that 220 they are contributing significantly to IFN-λ production in this setting. 221 We were particularly intrigued by the lack of GFP expression in alveolar macrophages 222 from NDV-infected IFN-λ reporter mice. Previously published reports using an IFN-α 223 reporter mouse have identified alveolar macrophages as the major IFN-α-producing cell 224 population during NDV and RSV infections in vivo [5,6]. The results of our experiment 225 demonstrate that alveolar macrophages do not contribute to IFN-λ production stimulated 226 by NDV infection in vivo, suggesting that alveolar macrophages preferentially upregulate 227 type I IFNs. To confirm this observation we carried out NDV infections in WT 129 SvEv 228 mice and looked for induced IFN-λ expression in both epithelial cells and alveolar 229 macrophages. Lungs from mock-or NDV-infected mice were harvested 24 hrs post-230 infection, and epithelial cells (CD45-EpCAM+) and alveolar macrophages (CD45+ 231 F4/80+ SiglecF+ CD11c+) were sort purified by FACS. RNA from the sorted cell 232 populations was used to synthesize cDNA for qPCR analysis which showed that IFN-λ 233 transcripts were present only in the epithelial cell fraction isolated from NDV-infected mice 234 (Fig 7). IFN-λ transcripts were not detected in the alveolar macrophage fraction from 235 either mock-infected or NDV-infected mice, further confirming the readout from the IFN-λ 236 reporter mouse strain. As an alternative approach to assessing IFN-λ production induced 237 by virus, we used in situ hybridization to detect IFN-λ mRNA in lung sections from infected 238 animals (Fig 8). IFN-λ RNA was found in bronchial epithelial cells of NDV-infected mice 239 and not mock-infected animals. Taken together, the results from our qRT-PCR and in situ 240 hybridization studies lead to the conclusion that only epithelial cells, and not alveolar 241 macrophages, are responsible for IFN-λ production during NDV infection. 242 To ensure that our study produced results consistent with the work of Kumagai et al. 243 [5], we also assayed RNA purified from the sorted epithelial cell and alveolar macrophage 244 populations ( Fig 8A) for the presence of IFN-α transcripts. As expected, induction of IFN-245 α mRNAs were observed only in alveolar macrophages ( Fig 8C&D). 246

IFN production by alveolar macrophages ex vivo 248
Noting the inability of the alveolar macrophage population to synthesize type III IFNs 249 in the course of NDV infection, we wished to determine whether these cells were 250 intrinsically unable to produce IFN-λ in response to NDV, or whether extrinsic factors in 251 the lung microenvironment were restricting their production of IFN-λ. To ask this question, 252 alveolar macrophages were purified from the lungs of naïve 129 SvEv mice by 253 collagenase digestion and FACS sorting ( Fig 9A) then mock-infected, or infected with 254 NDV at an MO1=10, for 24 hours. The cells were then collected for RNA extraction and 255 assayed by qPCR for the presence of IFN-λ transcripts. RNA isolated from a rodent cell 256 line constitutively expressing FLAG-tagged mIFN-λ2, was used as a positive control. This 257 analysis failed to detect IFN-λ expression from either mock-infected or NDV-infected 258 alveolar macrophages (Fig 9B), a result confirmed by ELISA assay of media from the 259 infected alveolar macrophage cultures ( Fig 9C). Assay of the same samples showed high 260 levels of IFN-α in media harvested from NDV-infected alveolar macrophages (Fig 9C). 261 These results demonstrate that the same stimulus results in the production of either type 262 I or type III IFN in a cell-type specific manner. 263 264

IFN-l during influenza A virus (IAV) infections in vivo 266
Intranasal administration of the non-replicating NDV allowed us to examine type I and 267 type III IFN production following initial exposure, without regard to virus spread. However 268 we also wished to use the Ifnl2 gfp/gfp reporter mouse strain to characterize IFN-λ induction 269

by influenza A virus (IAV), a virus capable of robust replication and spread in this species. 270
Published studies have reported co-production of type I and type III IFNs from both pDCs 271 [11] and epithelial cells [23, 29] exposed to IAV ex vivo, and we wished to characterize 272 the in vivo source(s) of these cytokines. Lung homogenates from Ifnl2 gfp/gfp mice infected 273 with the WSN strain of IAV (10 6 pfu) were prepared 48 hours post infection, and assayed 274 by flow cytometry for GFP expression. By this method, measurable GFP expression was 275 detected only in the pDC subset (CD45+/CD11c+/CD11b-/B220+/PDCA1+) and not in 276 epithelial cells (Fig 10A-C). 277 Given the data obtained with NDV, and a recent study which concluded that epithelial 278 cells are the primary source of IFN-λ in IAV infection [30], we considered the possibility 279 that the more numerous epithelial cells might produce lower levels of IFN-λ on a per cell 280 basis, below the level of detection by GFP expression, but still contribute to overall 281 production of this cytokine. We approached this question in two ways. Using the sorting 282 strategy shown in Figure 10 B & C, epithelial cells and pDCs from lungs of IAV-infected 283 WT mice and controls were obtained, and RNA harvested from these populations was 284 assayed by qRT-PCR for the presence of Ifnl2/3 transcripts. As shown in Figure 10D, 285 while IFN-λ2/3 mRNA could be detected in EpCAM+ cells from infected animals, pDCs 286 appeared to be the predominant source of both IFN-α and IFN-λ. This result was 287 consistent with ISH hybridization studies using formalin-fixed lung tissues form IAV-288 infected WT and reporter mice. Immunofluorescence studies showed diffuse positive 289 staining for influenza antigen in airway lining epithelium, and oligonucleotide probes did 290 detect rare IFN-λ2/3 expressing epithelial cells (Fig 10E) in animals infected with the WSN 291 strain of IAV. A similar GFP expression pattern in pDCs and epithelial cells was observed 292 by flow cytometry in animals infected with the PR8 strain, in the presence or absence of 293 the NS1 gene (Fig 11) 48 hrs post-infection, but we were not able to detect ifnl transcripts 294 by ISH in PR8-infected epithelial cells (data not shown). 295 To determine the relative contribution of these sources, we generated marrow chimeras 296 using bone marrow harvested from newly generated IFN-λ ligand deficient mice, ifnl-/-, 297 lacking both the ifnl2 and ifnl3 alleles (See Supplemental Fig. 2). Lethally irradiated WT 298 mice were reconstituted with bone marrow harvested from either WT or ifnl-/-animals. [32-34] results in robust type I IFN expression. Based on our finding that pDCs are also 332 the major source of type III IFNs in response to IAV infection, we suspected that this 333 pathway was also required for optimal IFN-λ production. To test this assumption, MAVS-334 /-mice on the C57BL6 background were infected with the WSN and PR8 strains of IAV, 335 and IFN-λ protein levels were measured in BALs. We saw no reduction in IFN-λ 336 expression following PR8 infection of MAVS-/-mice at 24, 48 and 72 hours post infection 337 (Fig 15A), but a ~ 40% reduction following WSN infection at 48 hours post infection, 338 consistent with data from our pDC depletion study. Conversely, in the absence of MyD88 339 ( Fig 15B), a more significant reduction in IFN-λ was seen in the BALs of WSN-infected 340 mice. Taken together, our data show that while both pathways can contribute to IFN-λ 341 production following IAV infection, the major contribution is pDC derived, particularly for 342 the PR8 strain of IAV. 343

344
We further explored this IAV strain dependence ex vivo using WT murine tracheal 345 epithelial cells (mTEC) cultured at an air-liquid interface on transwell filters. IFN-l protein 346 levels in culture supernatants from virus infected mTECs showed robust production by 347 NDV or WSN infected cultures, but no detectable IFN-l following PR8 infection (Fig 16). 348 This result is consistent with our inability to detect epithelial IFN-l production following 349 PR8 infection in vivo, and the absence of a requirement for functional MAVS protein. We and others have observed no significant difference in the resistance of wild type 353 and IFNAR-/-mice to IAV infection [22,35]. In the absence of the type I IFN pathway, 354 IFN-λ production is sufficient to inhibit virus replication and spread. Assuming that pDCs 355 were a major source of this cytokine, GFP expression was assayed in pDCs isolated from 356 the lungs of PR8-infected Ifnl2 gfp/gfp mice on WT or IFNAR-/-129SvEv strain backgrounds. 357 In this experiment, pDCs from IAV-infected Ifnl2 gfp/gfp reporter mice with intact type I IFN 358 signaling expressed GFP, but strongly reduced expression was detected in pDCs from 359 IFNAR deficient Ifnl2 gfp/gfp animals (Fig 17A). In vitro WSN infection of FLT3L cultured 360 pDCs derived from bone marrow of wild type or IFNAR-/-mice produced the same result, 361 with ELISA-detectable amounts of IFN-λ protein found only in medium from IAV-infected 362 IFNAR +/+ pDC cultures. 363 Plasmacytoid DCs are unique for their constitutive expression of interferon regulatory 364 factor 7 (IRF7), a transcription factor essential for the induction of IFN-α genes [36]. IRF7 365 has also been implicated as a driver for IFN-λ transcriptional activation [37] [38]. Given 366 the importance of IRF7 in the induction of IFN genes, we looked to see whether IRF7 367 levels were altered in the absence of IFNAR. As shown in Figure 17C, qPCR analysis 368 demonstrated a substantial reduction in basal levels of IRF7 mRNA in IFNAR-/-pDCs. 369 Since IRF7 is itself an ISG, we conclude that pDCs require some level of tonic activation 370 of the type I IFN signaling pathway to maintain sufficient IRF7 for type III IFN induction in 371 response to IAV infection. Interestingly, there was no decrease in GFP+ epithelial cells in 372 NDV-infected IFNAR-/-IFN-λ reporter mice (Fig 18) 30,41,42]. Until recently it was thought that all epithelial cells respond to both 384

IFN-α/β and IFN-λ, but recent in vivo studies have demonstrated that intestinal epithelial 385 cells become IFN-α insensitive soon after birth, and respond only to IFN-λ in situ [19]. 386
These observations point to distinct roles for type I and type III IFNs in innate immune 387 responses, determined primarily by the responsiveness and sensitivity of specific cell 388 types to IFN-α/β and IFN-λ, but also by the availability of their ligands. 389 Our goal in generating an IFN-λ reporter mouse was to investigate the source of this 390 cytokine as a viral infection of an epithelial surface progressed. This was done by 391 replacing the IFNL2 coding sequence with eGFP, a reporter gene whose expression is 392 then regulated by the IFNL2 promoter, UTR regions and other elements surrounding the 393 IFNL2 gene. Our first aim was to determine the extent to which GFP expression correlated 394 with IFN-λ production by various cell types. In vitro virus infection of both epithelial and 395 dendritic cells derived from the reporter mice induced GFP expression, and only GFP+ 396 cells were found to express IFN-λ transcripts. Equally important were studies to confirm 397 GFP and IFN-λ co-expression in vivo. Lauterbach et al. [17] had previously demonstrated 398 that splenic CD8α DCs were the major producers of IFN-λ following systemic treatment 399 with poly(I:C), a TLR3 ligand, and we established that GFP expression in the IFN-λ 400 reporter mouse was limited to that cell type (Fig 3) following i.v. administration of poly(I:C). 401 We next asked whether GFP expression by the Ifnl2 gfp/gfp reporter mouse would 402 recapitulate the pattern of IFN-λ expression previously described for heterologous 403 rotavirus infection. Hernandez et al. [43] reported that IFN-λ transcripts were present only 404 in CD45-EPCAM+ intestinal epithelial cells, but not expressed by cells isolated from 405 lamina propria during RV infection. We repeated this study in WT and IFN-λ reporter 406 suckling mice, harvesting tissue at the time point corresponding to maximal IFN 407 responsiveness post-infection. Elevated, and equivalent, levels of IFN-λ were detected in 408 intestinal homogenates from both strains following infection, with GFP expression 409 detectable only in the epithelial compartment (Fig 6). We concluded from these results 410 that GFP expression is an accurate indicator of IFN-λ expression in this reporter mouse 411 strain and, importantly, that IFN-λ expression levels are equivalent in WT and Ifnl2 gfp/gfp 412 mice where production of IFN-λ3 protein compensates for the lack of the functional IFNL2 peroxisomes induces type III, but not type I IFN production [9]. Preferential IFN-λ 452 synthesis following MAVS activation was found to depend upon the relative abundance 453 of peroxisomes in a given cell type, and polarization of intestinal epithelial cells resulted 454 in increased numbers of these organelles. While these data predict a shift in the relative 455 levels of type I and type III IFN production by different cell types, the complete absence 456 of IFN-λ synthesis by alveolar macrophages in infected animals was unexpected. 457 Further investigation is required to understand this compartmentalization of IFN 458 production that we observe in vivo. 459 As NDV infection is abortive in mammalian cells [49], it was of interest to repeat these 460 studies using IAV, which replicates and causes disease in the murine host. Based on 461 our NDV data, and published reports which suggested that infected epithelial cells were 462 the primary source of type III IFNs in IAV infection [30,45], we expected our influenza 463 studies to confirm this conclusion, but this was not the case. By all approaches taken it 464 appeared that, for IAV infection, the bulk of IFN-l was produced by pDCs rather than 465 the respiratory epithelium. When pDC-depleted mice were infected with IAV, there was 466 a 60% reduction in BAL levels of both type I and type III IFNs. Infection of bone marrow 467 chimeras generated by the transfer of Ifnl-/-bone marrow into WT mice showed a 468 similar decrease in type III IFN induction by IAV. In our study, ~ 40% of the IFN-l 469 induced by IAV was from non-pDC sources, but this was dependent on virus strain. 470 While GFP+ pDCs were detected using either the PR8 or WSN strains of IAV, only in 471 WSN infections was epithelial IFN-l synthesis detected in vivo or ex vivo. The 472 conclusion that type III IFN synthesis induced by PR8 came primarily from pDCs was 473 further supported by studies in MAVS-/-animals which showed no impact on IFN-l 474 levels in the absence of this pathway in PR8 infection. 475 Also of interest was our finding that IFN-l production by IAV-exposed pDCs is type I 476 IFN dependent. Our data suggest that the tonic signaling through the type I IFN receptor 477 may be required to maintain the elevated basal levels of IRF7 which is a hallmark of this 478 cell type [50]. While this result was unexpected, it is consistent with the observation that 479 human subjects with a deficiency of IRF7 expression are more likely to experience life-480 threatening infections of influenza [51] as well as SARS-CoV-2 [52]. As both type I and 481 type III IFNs are known to play a role in protection from respiratory viruses, this 482 observation is further support of the hypothesis that pDCs are the major source of both 483 of these cytokines during IAV infection. The demonstration that both IFN-a and IFN-l 484 are produced by IAV-exposed human pDCs [11] also supports this possibility. 485 In summary, we have used an IFN-l reporter mouse model to determine the source of 486 type III IFNs in two mouse models of respiratory virus infection. For NDV, type I and 487 type III IFNs are simultaneously induced through the engagement of the same virus 488 sensing pathway, but from two distinct cell types with respiratory epithelium as the sole 489 source of IFN-l. In IAV infection, pDCs appear to produce the bulk of type III IFN, and 490 its production by this cell type requires type I IFN signaling. This study suggests that 491 IFN induction by viruses in vivo is pathogen specific, as is the source of these cytokines. 492

Generation of the IFN-λ Reporter Mouse 495
Generation of the IFN-λ reporter mouse was carried out by ingenious Targeting 496 Laboratory (www.genetargeting.com). The targeting vector was constructed using an 497 11 kb region of a C57BL/6 BAC clone (RP23: 24B20), containing the Ifnl2 locus, sub-498 cloned into the pSP72 backbone vector (Promega). A GFP/FRT-flanked neomycin 499 cassette was inserted into the BAC subclone using Red/ET recombineering technology, 500 resulting in the complete replacement of the IFNλ2 coding region of exons 1-5 (including 501 intron sequences), with a long homology arm extending approximately 6.91 kb 5' to the 502 site of the cassette insertion and a short homology arm that extends approximately 2.68 503 kb 3' to the site of the cassette insertion. The GFP-Neomycin targeting vector was 504 linearized by Not I enzymatic digestion and electroporated into BA1 (C57B/6 x 505 129SvEv) hybrid embryonic stem cells. G418 resistant ES cell clones expanded for 506 PCR analysis to identify homologous recombinants. Clones found to carry the GFP 507 transgene were sequenced to confirm insertion, sequence fidelity, the genome/5' 508 cassette junction and the genome/3' cassette junction. Southern blot analysis of 509 positive clones was carried out using ApaI digestion of PCR products hybridized with a 510 probe targeted against the 5'external region. Targeted BA1 hybrid embryonic stem cells 511 were microinjected into C57BL/6 blastocysts and resulting chimeras with a high 512 percentage agouti coat color were mated to C57BL/6 FLP mice to remove the neomycin 513 cassette. Tail DNA from pups with agouti or black coat color was analyzed by PCR to 514 confirm presence of transgene. Animals heterozygous for the Ifnl2 +/gfp allele 515 were backcrossed onto the 129SvEv background for 10 generations, then bred to 516 produce homozygous animals. Genotyping was carried out by two PCR reactions: One 517 reaction using Primer A (5'-CAGAGCTGGAAACTCAGAGCC-3') and Primer B (5'-518 GACCGAGTCTGAGACCCACAAG-3') and another reaction using Primer C (5'-519 CAGAGCTGGAAACTCAGAGCC-3') and Primer B. Thermocycler conditions for those 520 reactions were as follows: 95° C for 15 minutes, followed by 35 cycles of (94° C for 45 521 minutes, 65° C for 1 minute, and 72° C for 1.5 minutes), with an end step of 72° C for 5 522 minutes. The amplicon generated by Primers A and B, which encompasses the Ifnl2 or 523 gfp region, was then digested with the NcoI restriction endonuclease. NcoI treatment of 524 the wildtype allele yields a 1100 bp band and a 700 bp band, while digestion of the 525 amplicon containing the Ifnl2 +/gfp allele, which lacks NcoI restriction sites, yields a single 526 1110 bp band. Bone-marrow derived FLT3-ligand dendritic cell cultures were generated from the 540 bone marrow of IFN-λ reporter mice, ifnl-/-mice and WT controls. Bone marrow from 541 femurs and tibia of 6-10 week old mice was washed, depleted of red cells with ACK 542 lysis buffer, and cultured in RPMI+10% FBS containing 100 ng/mL human FLT3-ligand 543 (Peprotech), 50 µM β-mercaptoethanol (Sigma), penicillin (100 IU/ml.) and streptomycin 544 (100 μg/ml) at a cell density of 3x10 6 cells/ml for 7 -8 days. 545 546 For preparation of kidney epithelial cells, kidney capsules were removed, renal 547 parenchyma was minced into 1 mm 3 pieces and digested with collagenase IV 548 (Worthington) at 37°C for 30 min. Following red blood cell lysis, digested tissue was 549 pressed through a100 μM, then a 40 μM cell strainer, and the recovered cells were 550 washed and then plated in DMEM/HamsF12 with 10% FBS. After 1-2 hours of 551 incubation, non-adherent cells were collected and plated on collagen-coated dishes. 552 Cells were allowed to reach 70% confluence prior to virus infection. 553

554
Murine tracheal epithelial cells were derived from WT or IFN-l reporter mice at 6 to 10 555 weeks of age following the procedures outlined by You et al. [54].

Virus infection and assay 570
Six-day-old suckling mice were infected with 4x10 6 PFU of the simian RRV strain by 571 oral gavage. Intestines were harvested 24 hours post-infection. These were either 572 formalin fixed or homogenized in medium for virus quantitation. RRV quantitation was 573 done using a focus forming assay as previously described [19]. Harvested tissues were fixed in 10% neutral buffered formalin, and submitted for 610 processing and paraffin embedding. Deparaffinized sections underwent antigen retrieval 611 and were subsequently blocked with Superblock (ScyTek) for 5 minutes at room 612 temperature, washed in wash buffer (PBS+0.05% Tween) and then stained using anti-613 NDV antibodies (US Biological), anti-RRV antibodies (Meridian), anti-eGFP (Life 614 Technologies) antibodies or anti-IAV antibodies (US Biological)overnight at 4°C. The 615 following day, slides were washed in wash buffer and stained using anti-DyLight488 616 (Abcam) or anti-DyLight564 (Abcam) secondary antibodies for 1 hour at room 617 temperature in the dark, washed in wash buffer, and incubated with DAPI for 6 minutes. 618 After washing in distilled water, mounting medium (Vectashield) was applied, and slides 619 were coverslipped. Images were obtained using an Axiovert 200M inverted 620 fluorescence microscope (Zeiss) using the AxioVision LE64 software (Zeiss). In situ detection of IFN-λ transcripts was carried out using the RNA Scope© HD 2.5 646 Detection Reagent-Red kit (Advanced Cell Technologies) according to manufacturer's 647 instructions. Briefly, formalin-fixed, paraffin-embedded tissue sections were heated at 648 60°C for 1 hour, deparaffinized in xylene and 100% alcohol, and incubated with 649 hydrogen peroxide for 10 minutes at room temperature. Antigen retrieval was then 650 carried out for 1 hour in RNA Scope antigen retrieval buffer, and slides were washed in 651 distilled water and 100% ethanol, and then left to dry overnight. The following day, 652 slides were washed in RNA Scope proprietary wash buffer and incubated in protease-653 plus buffer, washed again, and then incubated with mouse IFNλ2/3 DNA probes 654 (Advanced Cell Technologies), followed by a series of RNA Scope adaptor probes,