Comprehensive RNAi-based screening of human and mouse … › content › sigtrans › 9 › 409 › ra3.full.pdfin which the original identification of involvement in the TLR response

R E S E A R C H A R T I C L E

S Y S T E M S I M M U N O L O G Y

Comprehensive RNAi-based screening of humanand mouse TLR pathways identifiesspecies-specific preferences in signaling protein useJing Sun,1 Ning Li,1 Kyu-Seon Oh,1 Bhaskar Dutta,2 Sharat J. Vayttaden,1 Bin Lin,1

Thomas S. Ebert,3 Dominic De Nardo,4,5,6 Joie Davis,7 Rustam Bagirzadeh,8

Nicolas W. Lounsbury,1 Chandrashekhar Pasare,8 Eicke Latz,4,9,10

Veit Hornung,3 Iain D. C. Fraser1*

Toll-like receptors (TLRs) are amajor class of pattern recognition receptors, whichmediate the responsesof innate immune cells tomicrobial stimuli. To systematically determine the roles of proteins in canonicalTLR signaling pathways, we conducted an RNA interference (RNAi)–based screen in human and mousemacrophages. We observed a pattern of conserved signaling module dependencies across species, butfoundnotable species-specific requirements at the level of individual proteins. Among these,we identifiedunexpected differences in the involvement of members of the interleukin-1 receptor–associated kinase(IRAK) family between thehumanandmouseTLRpathways.WhereasTLRsignaling inmousemacrophagesdependedprimarily on IRAK4 and IRAK2, with little or no role for IRAK1, TLR signaling andproinflammatorycytokine production in human macrophages depended on IRAK1, with knockdown of IRAK4 or IRAK2having less of an effect. Consistent with species-specific roles for these kinases, IRAK4 orthologs failedto rescue signaling in IRAK4-deficient macrophages from the other species, and only mouse macro-phages required the kinase activity of IRAK4 tomediate TLR responses. The identification of a critical rolefor IRAK1 in TLR signaling in humans could potentially explain the association of IRAK1 with several auto-immune diseases. Furthermore, this study demonstrated how systematic screening can be used to identifyimportant characteristics of innate immune responses across species, which could optimize therapeutictargeting to manipulate human TLR-dependent outputs.

INTRODUCTION

The mammalian innate immune system plays a vital role in host defensebecause it continually engages microbial, viral, and parasitic stimuli. Thisengagement is mediated through the recognition of invariant pathogen-associated molecular patterns (PAMPs) by a range of pattern recognitionreceptors (PRRs) expressed by the host cell (1–5). Toll-like receptors(TLRs) are a major class of PRRs, which respond to a range of PAMPsto induce a complex signaling response and the transcription of hundredsof genes required for an effective immune response (6–8). TLRs are foundin multiple host cell types that encounter PAMPs, and consequently, knowl-edge of the canonical components that constitute TLR signaling pathwayshas been derived from studies encompassing different cells in differentcontexts (9). To date, there has been no systematic study to evaluate the

1Signaling Systems Unit, Laboratory of Systems Biology, National Institute ofAllergy and Infectious Diseases, National Institutes of Health, Bethesda, MD20892, USA. 2Bioinformatics Team, Laboratory of Systems Biology, National In-stitute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda,MD 20892, USA. 3Institute of Molecular Medicine, University Hospital, Universityof Bonn, 53127Bonn,Germany. 4Institute of Innate Immunity, UniversityHospital,Biomedical Centre, University of Bonn, 53127Bonn,Germany. 5InflammationDivi-sion, Walter and Eliza Hall Institute, Parkville,Victoria 3052, Australia. 6Depart-ment of Medical Biology, The University of Melbourne, Parkville, Victoria 3010,Australia. 7Laboratory of Clinical Infectious Diseases, National Institute of Aller-gy and Infectious Diseases, National Institutes of Health, Bethesda, MD 20892,USA. 8Department of Immunology, University of Texas Southwestern MedicalCenter, Dallas, TX 75390, USA. 9German Center for Neurodegenerative Diseases(DZNE), 53175 Bonn, Germany. 10Department of Infectious Diseases and Immu-nology, University of Massachusetts Medical School, Worcester, MA 01605, USA.*Corresponding author. E-mail: [email protected]

w

contributions of the primary TLR pathway components to different ligandsin a single cell type, no less in critical cells of the innate immune system,such as macrophages. RNA interference (RNAi) is a valuable research toolwith which to systematically evaluate gene product contributions in a givenbiological system through the targeted inhibition of gene expression (10, 11).However, there are substantial challenges in the use of this technology ininnate immune cells, including the efficient delivery of small RNAs into cellsand the occurrence of nonspecific immune responses to double-strandedRNA (dsRNA) (12), particularly small interfering RNA (siRNA) (13, 14).The difficulty of siRNA delivery can be circumvented by the viral expres-sion of short hairpin RNAs (shRNAs), and this approach has been used inprimary dendritic cells to identify important roles for transcriptional regu-lators and signaling proteins whose production is induced by TLR stimula-tion (15, 16). However, there are very few reports of siRNA-based screensin innate immune cells, as opposed to the more easily used fibroblast ormesenchymal cell lines, which do not have the same crucial roles in hostdefense as those of hematopoietic cells, such as macrophages.

Beyond this lack of comprehensive analyses of the TLR pathway in asingle, highly relevant cell type, a substantial proportion of our understand-ing of pathway architecture has come from mouse studies; knockout (KO)mice have been described for almost all of the components in the TLRpathway that were identified by in vitro studies with isolated cells (17, 18).This has shed considerable light on the similarities and differences in theresponse to different PAMPs engaging distinct TLRs, as well as on thebroader role of TLR responses in the immune response to pathogen infec-tion (19). Human patients deficient in several TLR pathway componentshave been described (20), and although the cellular response phenotypesof these patients generally support the pathway structure established from

ww.SCIENCESIGNALING.org 5 January 2016 Vol 9 Issue 409 ra3 1


mouse studies, the broader immune phenotypes of these human patientsoften show considerable differences to the corresponding KO mousemodel, which suggests that there is possible divergence between the spe-cies in the regulation of TLR-dependent immune outputs at the physio-logical level. A direct comparison of TLR signaling responses in humanand mouse innate immune cells could provide important insight into thecomparative dependencies of the two species on specific pathway com-ponents. Moreover, reports have debated the question of whether miceprovide a good model for human physiology or disease, especially withrespect to innate immunity (21, 22). In this regard, a careful comparisonof individual gene contributions at the cellular pathway level has a veryimportant role in addressing this issue.

We previously developed a highly optimized platform for the siRNA-based screening of pathogen responses in human andmousemacrophages,with stably expressed reporters that provide dynamic readouts for activationof the transcription factor nuclear factor kB (NF-kB) and expression fromthe TNF/Tnf promoter (23). Here, we used this platform to interrogate thesimilarities and differences in the responses of human and mouse macro-phages to different TLR ligands. We screened 126 canonical componentsof the human and mouse TLR pathways with six independent siRNAs pergene, and assessed signaling and inflammatory cytokine responses. We ob-served both conserved and distinct patterns of pathway component use fordifferent TLR ligands within each species. Moreover, the comparison ofpathway dependencies between human and mouse macrophages suggestsconservation at the level of signaling modules, but substantial divergence atthe level of individual signaling proteins. Through investigation of the usagepatterns for different interleukin-1 (IL-1) receptor–associated kinase(IRAK) family members, we identified considerable differences in howhuman and mouse macrophages signal through IRAKs in the responseto individual TLR ligands, which may have important consequences forstudies of human diseases that are influenced by this protein family.

RESULTS

Targeting the canonical TLR pathway in human andmouse macrophages with a comprehensive siRNAlibrary leads to low false-positive ratesWe previously developed an assay platform for the high-throughput anal-ysis of human and mouse macrophage activation (23). Using lentiviralconstructs containing dual expression cassettes, we generated a mouseRAW264.7-derived cell line (RAWG9 cells) that had a green fluorescentprotein (GFP)–tagged reporter of the nuclear translocation of the p65subunit of NF-kB (also known as RelA) and an mCherry-based red fluo-rescent reporter of the activity of the mouse Tnf promoter (fig. S1A). Bothof these reporters showed comparable responsiveness to multiple TLRligands (fig. S1, B and C). Using a similar approach, we also generated ahuman THP1-derived cell line (THP1 B5 cells) that had a dual luciferasereporter for activation of the human TNF promoter (fig. S1D) and that wasalso responsive to TLR ligand activation (fig. S1E). We previously showedthat these stably expressed reporter genes can be used as targets to develophighly optimized siRNA delivery protocols to achieve maximal gene si-lencingwithminimal nonspecific activation of the dsRNA-stimulated inter-feron response in macrophages (23). This cell system provides a platformfor systematic functional testing of putatively canonical components ofthe TLR signaling pathways in a relevant innate immune cell type.

As a first step toward a future full genome-wide screen of the effectsof siRNA-mediated knockdown of signaling pathway components onthe responses of macrophages to TLR ligands, we curated the publishedliterature and innate immune databases to identify a comprehensive set

w

of 126 genes whose products have reported function in TLR signalingpathways, either as direct signal propagation components or as feedbackmodulators (Fig. 1A). Inclusion in this list was irrespective of the cell typein which the original identification of involvement in the TLR responsewas made or of the species origin of the cells involved. The key goals ofthe present study of these putatively canonical components of TLR path-ways were to determine which of these molecules were functionally impor-tant for the TLR response in a key innate immune cell type and to examinewhether mouse and human macrophages required the same pathway pro-teins for their responses.

Target gene products could begrouped intovarious classes and broadlyclassified as either initial signal encoders, mediators of signal transmis-sion, or regulators of signal decoding at the transcriptional and feedbacklevels (Fig. 1A). This gene set was targeted with six independent siRNAsequences per gene, which were acquired from different commercial ven-dors (table S1), with previously described and optimized siRNA deliveryprotocols (23). Studies have shown that off-target effects in siRNA screen-ing can bemitigated by the use of at least three individual siRNAs per gene(24, 25). Our use of six individual siRNAs in this screen thus greatly re-duced the probability of off-target phenotypes and led to a more reliableevaluation of the relative contribution of each of the 126 gene products tothe TLR response (Fig. 1B). To address false-positive and false-negativerates arising from assay variability in the screens, we used data from the con-trol siRNAs included on each screening plate (Fig. 1B and seeMaterials andMethods). Comparing the mouse and human assay readouts, we calculateda false-positive rate in both screens of less than 5% for TLR signal propa-gators with z scores of <−1.0 (Fig. 1C). This suggests that genes with me-dian z scores below this threshold had a high probability of being true hits(table S2). Calculation of false-negative rates from the z scores of positivecontrol siRNAs showed that TLR pathway gene products with a mediansiRNA score above −0.75 had a <2% likelihood of being false negatives ineither the mouse or human screens (Fig. 1D), providing an important valid-ation of the macrophage reporter cell lines and the RNAi screening platform.

To evaluate the relative contribution of the 126 TLR pathway geneproducts in both human and mouse macrophages, we used species-specificsiRNA libraries. To assess the responses to different TLR ligands, we stim-ulated the THP1 B5 and RAW G9 reporter cells with ligands from the fol-lowing group andwemeasured the reporter assay readouts: lipopolysaccharide(LPS; for TLR4), Pam3CSK4 (P3C; for TLR1/TLR2), Pam2CSK4 (P2C;for TLR2/TLR6), peptidoglycan (PGN; for TLR2/TLR6), flagellin (FLG;for TLR5), and resiquimod 848 (R848; for TLR7 or TLR8). Data fromreplicate screens were averaged for each single siRNA, and z scores werecalculated from the median value of the six individual siRNAs tested foreach of the 126 genes (Fig. 1B, table S2, and see Materials and Methods).

Human and mouse TLR pathways show commondependencies in modules required forsignal propagationThe receptor dependency of the response to the tested TLR ligandsshowed the expected phenotypic patterns in the human THP1 macrophagecell line, with knockdown of TLR4, TLR2 or TLR1, TLR8, TLR2, TLR2 orTLR6, and TLR5 showing specific perturbation of tumor necrosis factor–a(TNF-a) reporter activation by LPS, P3C, R848, P2C, PGN, and FLG,respectively (Fig. 2, A and B). Beyond this anticipated outcome of TLRknockdown on the TNF-a reporter response, the strongest siRNA-mediatedeffects resulted from the perturbation of components in receptor-proximalsignaling, includingNF-kB,mitogen-activated protein kinase (MAPK), andtranscription factor modules (Fig. 2A and fig. S2A). As expected, manycomponents played important roles in responses to most of the TLR ligandstested, consistent with a broadly shared pathway structure downstream of



the different TLRs (6, 9). To develop a metric for identifying the mostconserved pathway dependencies across the TLR ligand set in humanmacrophages, we ranked the genes on the basis of their TNF-a reporterperturbation effect for each ligand and then summed the ranks for each gene(table S3). The top four genes (lowest summed rank values) that emergedfrom this analysis (Fig. 2C) encode products that are all from the proximalsignaling module (TRAF6, MAP3K7, IRAK1, and TAB1), followed by

www.SCIENCESIGNALING.org 5

geneswhose products are from theMAPK(MAP3K8 andMAPK14),NF-kB (NFKBIEand RELA), and transcription factor (EGR1and ELF4) modules.

We then assessed the effects of per-turbing the same 126-gene set in mousemacrophages using the RAW G9 reporterclone. In these cells, we tested three TLRligands—LPS, P3C, andR848—andmea-sured two downstream readouts: the initialactivation of NF-kB by measurement ofthe nuclear translocation of RelA and thesubsequent activation of the murine Tnfpromoter (Fig. 2D and fig. S1, B and C).TheTLR specificity of the responses to thedifferent ligands in mouse macrophagesshowed the expected inhibition of the LPSresponse with Tlr4 knockdown and of theR848 response with Tlr7 knockdown(Fig. 2E). The effect of knockdown ofTlr7 was in contrast to what occurred inhuman macrophages, which respond toR848 through TLR8 (Fig. 2B), and isconsistent with previous studies (26, 27).Knockdown of Tlr2 reduced the NF-kBresponse to P3C; however, unexpectedly,we did not see a strong Tlr2 dependencefor the TNF-a response to P3C, despiteusing the same siRNA set that gave astrong NF-kB reporter effect. This is sug-gestive of some possible redundancy in themurine TNF-a response to this ligand or tothere being different quantitative require-ments for NF-kB activation versus Tnftranscription. Among nonreceptor targets,the greatest effects on the NF-kB reporterwere found upstream of NF-kB activationin the receptor-proximal signaling modulesand in components of the NF-kB familyitself (Fig. 2D, top, and fig. S2B). Thegreatest perturbations of the mouse mac-rophage TNF-a reporter response wereobserved from knockdown of receptor-proximal signaling components and tran-scription factors (Fig. 2D, bottom, andfig. S2C).

Asdescribedearlier for thehumangenes,we calculated ranks for the 126mouse genesfrom the effects of their knockdown on theresponse to each of the three TLR ligandstested for both the early NF-kB readout andthe later TNF-a readout, and then we calcu-lated the average rankacross theTLR ligands

for each output (table S3 and Fig. 2, FandG).We found that the products ofhalf of the top 10 ranked genes that controlled the mouse NF-kB responsewere in the receptor-proximal signaling module (Traf6, Irak4, Cd14, Irak2,andMyd88), whereas genes encoding NF-kB (Nfkbia and Ikbkg) and phos-phatidylinositol 3-kinase (PI3K) pathway members (Akt1 and Pik3r1) alsoranked highly (Fig. 2F). The ranking of those genes with the most influenceon the mouse TNF-a response across the different TLR ligands (Fig. 2G)

–3 –2 –1 00

20

40

60

80

100

Z score threshold

Fal

se-n

egat

ive

rate

(%

)

–3 –2 –1 0 1 2 30

5

10

15

20

25

30

Z score threshold

Fal

se-p

ositi

ve r

ate

(%)

A

6 siRNA/gene

+ Cells 48–72 h

Hit selection

Z s

core

–1

–3

+1

+3

1 126 TLR pathway genes

RAW G9:

THP1 B5:

+ TLR ligands

40 min

16 h

4 h

High- content imaging

Dual luciferase

assay

B

TLR TLR

Encoder module

Transmission module

Decoder module G1 G2 G3 G4 Gn

Adaptors/IRAKs/TRAFs

C D

–3 –2 –1 0 1 2 30

5

10

15

20

25

30

Z score threshold

Fal

se-p

ositi

ve ra

te (%

)

–3 –2 –1 00

20

40

60

80

100

Z score threshold

Fal

se-n

egat

ive

rate

(%

)

Mouse screen Human screen Mouse screen Human screen

MAPK NF- PI3K/Akt κB

Fig. 1. A comprehensive siRNAscreenof theTLRpathway in humanandmousemacrophages. (A) Schematicof signal flow in the TLR pathway from initial signal encoding by TLR receptors, through complexes of TIRadaptor proteins, IRAKs, and TRAFs, to signal transmission through NF-kB, MAPK, and PI3Kmodules, andthen signal decoding through gene transcription, cytokine secretion, and feedback regulation. Right: 126gene transcripts targeted in the screen by siRNAs. (B) Workflow for the RNAi screen using six siRNA se-quences per gene distributed in separate regions of a 384-well plate. Red circles show examples of siRNAlocations for a single gene. The blue region of the plate shows the locations of gene-specific siRNAs,whereas the orange region shows the positions of control siRNAs. Forty-eight to 72 hours after they werereverse-transfected with siRNAs, the mouse (RAW G9) and human (THP1 B5) reporter cell lines were as-sayed for their responses to simulation with different TLR ligands, and the effects of individual gene pertur-bations were measured (see fig. S1 and Materials and Methods for further details). (C andD) Calculation offalse-positive rates (C) and false-negative rates (D) from negative and positive control siRNAs, respectively,in the mouse and human macrophage siRNA screens (seeMaterials and Methods). Data in (C) and (D) arepresented as median z scores from six siRNAs per gene. Individual siRNA scores were averaged from two(C) or three (D) independent experiments (see Materials and Methods for details).

January 2016 Vol 9 Issue 409 ra3 3


confirmed the enrichment of genes encoding products in the proximal sig-nalingmodule (Irak4, Irak2,Cd14, andMyd88) and key transcription factorsfor Tnf expression (Egr2, Sp1, Elk1,Crebbp, and Egr1). This pattern is sim-ilar to that observed in the humanmacrophage reporter cells and is consistent

w

with a pathway architecture in which single gene perturbations have thegreatest effect either before the signaling system branches out throughmultiplemodules (NF-kB,MAPK, andPI3K-Akt) or after the branches con-verge on the components of the Tnf transcriptional enhanceosome.

A

LPS P3C R848

na

muH

TN

F-

reporter

P2C PGN FLG

–3.0 3.0

TLR MAPK TFs lamixorP tkA/K3IPCytokines NF-κB TRIF Regulators

TLR1

LPS P3C R848 P2C PGN FLG

TLR2 TLR3 TLR4 TLR5 TLR6 TLR7 TLR8 TLR9

B C

D

–3.0 3.0

TLR MAPK TFs lamixorP tkA/K3IPCytokines NF-κB TRIF Regulators

LPS

E F

TLR1

P3C R848

TLR2 TLR3 TLR4 TLR5 TLR6 TLR7 TLR8 TLR9

LPS P3C R848

Mouse TNF-reporter Mouse NF- B

Human TNF- reporter

G

esu

oMN

F-

B

LPS P3C R848 M

ouse TN

F-

reporter

LPS P3C R848

126 genes

126 genes

Fig. 2. Effects of siRNA gene perturbations across the human and mouseTLR pathways. (A to G) Analysis of the effects of gene perturbations onTLR signaling. TLR pathway gene perturbation effects on (A) the TNFpromoter–driven transcriptional response of the human THP1 cell line to LPS(10 ng/ml), 100 nMP3C,R848 (10 mg/ml), 1 nMP2C, 1PGN (10 mg/ml), or FLG(10 ng/ml) and (D) the NF-kB– and Tnf promoter–dependent responsesof the mouse RAW264.7 cell line to LPS (10 ng/ml), 100 nM P3C, or R848(10 mg/ml). (B and E) Effects of the siRNA-mediated knockdown of theindicated TLRs on (B) the TNF promoter–dependent transcriptional re-sponse of the human THP1 cell line and (E) the NF-kB– and Tnf promoter–

dependent responses of the mouse RAW264.7 cell line to the indicated TLRligands. Top 10 genes with the most substantial effects on (C) the humanTNF promoter–dependent response, (F) the mouse NF-kB response, and(G) the mouse Tnf promoter–driven response across all tested TLR ligands.Rankings for individual TLR ligands are shown, with overall gene orderdeterminedby theaverage rankcalculatedover all ligands (the full 126-generanking scores are shown in table S3). Data in (A), (B), (D), and (E) arepresented as median z scores from six siRNAs per gene. Individual siRNAscores were averaged from three (for A and B) or two (for D and E)independent experiments (see Materials and Methods).



Human and mouse TLR pathways show distinct generequirements within signaling modulesWhereas we observed similar requirements in signaling pathway modulesin our analysis of human and mouse TLR pathway dependencies, we alsofound that the specific genes whose knockdown resulted in the strongestphenotypes were often different between macrophage cells from the twospecies. For example, although genes encoding proximal TLR signalingcomponents and transcription factors were both strongly enriched amongthe top 10 ranked genes in human andmouse TNF-a activation by TLR lig-ands, only one gene, EGR1, which encodes the transcription factor EGR1,ranked highly in the screens for both species (Fig. 2, C andG). Thus, althoughthe overall findings of the screen confirmed the expected roles ofmanymole-cules in responses to multiple TLR ligands within a species, a number of un-expected cross-species differences were discovered through this systematicanalysis.

To further analyze the degree of overlap between the human and mousemacrophage responses to TLR stimulation, we compared the pathway com-ponents required for TNF-a reporter activation by the three TLR ligandsthat were tested in both human and mouse cells (LPS, P3C, and R848). Hi-erarchical clustering of the gene perturbation effects across the three TLRligands in the two cell types showed that the gene dependencies often clus-tered separately for human and mouse cells (Fig. 3A and fig. S3). Whereastherewere a group of shared genes required for human andmouse responsesto all three TLR ligands, we identified distinct clusters of geneswith strongerrequirements for human or for mouse TLR signaling. Furthermore, withineach species, the P3C and R848 gene dependencies were more similar, with aseparate cluster of genes showing specificity for theLPS response (Fig. 3A).

To assess these species-specific patterns in the context of the canon-ical TLR pathway, we took the median value of the z scores for each geneacross the three TLR ligands and calculated the median score differencebetween human and mouse cells (table S4, “Median hTNF-mTNF”). Weoverlaid these differences onto the KEGG (Kyoto Encyclopedia of Genesand Genomes) TLR pathway to highlight pathway nodes that showed abias toward greater signaling dependency in a particular species (Fig. 3B).This analysis highlighted several patterns. Mouse cells were more sensitivethan were human cells to the knockdown of Cd14, Irak4, PI3K signalingcomponents, Ikbkb, Mek1, and Mek2. Human cells were more sensitive tothe knockdown of IRAK1, TRAF6, components of the TAK1/TAB com-plex, and Tpl2 (MAP3K8). Because the KEGG pathway does not containall of the individual genes tested in our screen, we highlighted the 15 geneproducts with the strongest differential dependencies in the human andmouse TLRpathways (Fig. 3C). To assesswhether the observed differencescould simply be a result of variation in the efficiency of knockdown, wetested the expression of each of the 15 genes in the mouse and human mac-rophage reporter cells after theywere transfectedwith the appropriate siRNAs(Fig. 3, D and E). The similar degrees of knockdown observed across the15-gene panel in both human and mouse cells suggest that differentialknockdown efficiency is unlikely to be the primary cause of the phenotypicdifferences observed in the screen.

Of the differences between human and mouse cells that we empha-sized here, we considered one of the most striking patterns to be theapparently differential usage of IRAK family proteins in the two species.Mouse cells showed a strong requirement for Irak2 and Irak4, but a weakerrequirement for Irak1 (particularly in the TNF-a readout), whereas the hu-man TNF-a readout had avery strong IRAK1 requirement, but showed verylittle phenotypewhen either IRAK2 or IRAK4was knocked down (Fig. 3C).Given that these findings suggest a very different view of the participationof IRAKs in the TLR responses of mouse and human cells than is present-ly thought, we focused further efforts on this specific aspect of the overallscreen findings.

w

IRAK4 is essential for both mouse and human TLRsignaling, but the human and mouse proteins are notfunctionally interchangeablePrevious studies showed that upon ligand activation, TLR signaling path-ways that depend on the adaptor protein myeloid differentiation primaryresponse gene 88 (MyD88) initially recruit IRAK4, which binds toMyD88through death domain interactions (28, 29). This complex then requireseither IRAK1 or IRAK2 to signal to TRAF6 (TNF receptor–associatedfactor 6) because IRAK4 lacks the TRAF6-binding domain that is presentin both IRAK1 and IRAK2 (30). In light of the apparently essential require-ment for IRAK4 to form the initial contact with MyD88, the weak pheno-type exhibited by IRAK4 knockdown cells in our human screen data wasunexpected, especially considering the strong dependence on IRAK4 inthe mouse data set (Fig. 3C). We assessed whether this difference was a re-sult of the more efficient knockdown of IRAK4 in the mouse cell line, butwe found that the knockdown of IRAK4 was comparable in the THP1 andRAW264.7 reporter cell lines (Fig. 4, A andB). Furthermore, we saw a sim-ilarly weak effect of the knockdown of IRAK4 in primary humanmonocyte-derived macrophages (hMDMs) (Fig. 4C) despite there being a robustknockdown of IRAK4 mRNA and protein in these cells (Fig. 4, A and B).

Human patients with IRAK4 deficiency have been previously de-scribed (31), and various immune cell types from these patients have de-fective TLR-induced cytokine responses, which are comparable to thoseof IRAK4-deficient mouse cells (32). We sought to determine whetherhMDMs from an IRAK4-deficient patient might retain some level of TLRsignaling response in light of the lack of phenotype observed in IRAK4-depleted human macrophages in our siRNA screen. However, consistentwith previous studies, we found an almost complete loss of TLR-inducedcytokine secretion in peripheral blood mononuclear cells (PBMCs) iso-lated from an IRAK4-deficient patient (fig. S4, A to C). We then tested theTLR response of MDMs from the same patient but found a similar lack ofTLR ligand–induced cytokine secretion in these cells (Fig. 4, D to F). Thesedata suggest that human macrophages require IRAK4 to propagate a signalthrough the TLR pathways but that they are less sensitive than mouse cellsto IRAK4 depletion, because both THP1 cells and primary hMDMs stillsignaled effectively through TLR pathways when IRAK4 was depleted to<20% of its normal quantity (Fig. 4, A to C). It is possible that low amountsof human IRAK4 are simply required for the recruitment of IRAK1, whichthen assumes the primary function of downstream signal transduction.

To assess whether the different sensitivities of mouse and humanIRAK4 to perturbation might indicate functional differences between theseIRAK4 orthologs, we attempted to rescue the TLR response in IRAK4-deficient immortalized mouse macrophages (IMMs) by retroviral expres-sion of either mouse or human IRAK4. Although expression of mouseIRAK4 in these cells restored TLR-stimulated cytokine secretion to anextent similar to that of wild-type cells, human IRAK4 was completely un-able to rescue the cytokine response (Fig. 4G), despite there being similaramounts of mouse Irak4 and human IRAK4mRNA in the cells (fig. S5A).To assess whether this pattern of selective rescue of signaling was also ob-served in human cells, we used CRISPR/Cas9-mediated genome editing togenerate a THP1 cell line deficient in human IRAK4 (fig. S5, B and C), andwe retrovirally expressed either human or mouse IRAK4 in this cell line(fig. S5,D andE).We found that expression of human IRAK4 fully restoredTLR-stimulated TNF-a secretion, whereas the expression of mouse IRAK4had no effect (Fig. 4H).

To further investigate the failure of the human and mouse IRAK4 pro-tein orthologs to rescue signaling inmacrophages from the other species, weimmunoprecipitated endogenousMyd88 from lysates of mouse and humanmacrophages expressing different IRAK4 proteins and that were stimulatedwith LPS for various times. In mouse macrophages, mouse IRAK4 formed



Contro

lIra

k1

Map

3k7Tab

2

Nfkbie

Map

3k8

Map

k3

Map

k14

Irf9Ira

k2Ira

k4Cd1

4Ikb

kb

Map

k1

Crebb

p

Ep300

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Mou

se m

RN

A

Exp

ress

ion

Contro

l

IRAK1

MAP3K

7TAB2

NFKBIE

MAP3K

8

MAPK3

MAPK14

IRF9

IRAK2

IRAK4

CD14

IKBKB

MAPK1

CREBBP

EP300

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Hum

an m

RN

A

Exp

ress

ion

C

-3.0

3.0

LPS

P3C

R84

8

LPS

P3C

R84

8

Mouse TNF

reporterMouse NF- B

LPS

P3C

R84

8

P2C

PG

N

FLG

Human TNF reporter

Human TLR pathway bias Mouse TLR pathway bias

B

A

D

E

Strong human hits

Weak human hits

Weak mouse hits

Strong mouse hits

LPS P3C R848 n

amu

HTN

F

reporter

esu

oMT

NF

reporter

LPS P3C R848

Human Mouse Shared LPS

126 genes

-3.0 3.0

Fig. 3. Human andmousemacrophages showboth shared anddistinct gene had the most substantial effects on the human andmouse TLR pathways.
product dependencies in TLR signaling. (A) Hierarchical clustering analysis(Pearson uncentered, average linkage) of the TNF/Tnf promoter–driven re-sponses to LPS, P3C, andR848 in human THP1 cells andmouseRAWcells.(B) Average z scores calculated from the responses to the three TLR ligandsshown in (A)werecomparedbetween thehumanandmouse reporter cell lines(see table S4). The z score differences between the species were overlaidonto the canonical TLR pathway from KEGG. (C) Genes whose perturbation
w

(D andE) Relative abundances of themRNAs corresponding to the genesshown in (C) in human THP1 B5 cells (D) or mouse RAWG9 cells (E) aftersiRNA-mediated knockdown.Data in (D) and (E) aremeans±SD from twoindependent experiments. Data in (A) and (C) are presented as median zscores from six siRNAs for each gene. Individual siRNA z scores were aver-aged from three (for the human TNF-a readout) or two (for the mouse NF-kBand TNF-a readouts) independent experiments (see Materials and Methods).



a sustained interaction with Myd88 for several hours (Fig. 5A), whereashuman IRAK4 showed no detectable interaction with mouse Myd88(Fig. 5B). In contrast, Myd88 immunoprecipitated from LPS-stimulatedhuman macrophages had little detectable association with mouse IRAK4(Fig. 5C), but clearly associated with human IRAK4 (Fig. 5D). We also ob-served different kinetics of Myd88 association with IRAK4 in mouse andhuman macrophages, with the association reaching a maximal extent at


30 min in mouse cells, but at 2 hours in hu-man cells (Fig. 5, A and D). We then eval-uated downstream signaling through theMAPK-dependent phosphorylation and nu-clear accumulation of the AP1 transcriptionfactor ATF2 in the mouse and humanIRAK4 rescue cell lines. We found that inmouse macrophages, human IRAK4 wasunable to restore the TLR-dependent phos-phorylation of ATF2 (Fig. 5E), whereas inhuman macrophages, mouse IRAK4 wasunable to restore TLR-dependent ATF2phosphorylation (Fig. 5F).

The data described earlier suggest thatthe differential sensitivity to IRAK4 deple-tion that we observed in our screen of theresponses of human and mouse cells toTLR signalingmight be a result of specificproperties of the IRAK4 protein that arenot conserved between the species, whichcould underlie the failure of the human andmouse IRAK4orthologs to rescue functionin IRAK4 KO cells from the other species.Previous studies suggested that the kinaseactivity of IRAK4 is required for TLR re-sponses inmouse, but not human, fibroblasts(33,34). To determinewhethermacrophagesfrom the two species exhibited a differen-tial dependence on the kinase activity ofIRAK4, we expressed kinase-deficient mu-tants (Kin−) of mouse (35) and human (36)IRAK4 in their respective KO macrophagecells (fig. S6). We found that the kinase-deficient mutant mouse IRAK4 was un-able to restore TLR-activated cytokinesecretion in IRAK4KOmousemacrophages(Fig. 5G), whereas the kinase-deficienthuman IRAK4 fully restored the cyto-kine response in IRAK4-deficient humanmacrophages (Fig. 5H).

Mouse and human primarymacrophages replicate thedifferent IRAK1 and IRAK2dependencies identified throughsiRNA screeningWe next sought to determine whether thedifferent patterns of IRAK1 and IRAK2dependency that were exhibited by thehuman and mouse macrophage reportercell lines used in our screen were observedin primary cells. We first tested bonemarrow–derived macrophages (BMDMs)

from different IRAK KO mice (37–41). We compared the cytokine re-sponses of BMDMs from wild-type, IRAK1 KO, and IRAK2 KO mice toa range of TLR ligands. We observed an almost complete loss of TNF-aand IL-6 secretion in response to P3C, PGN, or R848 in IRAK2-deficientBMDMs, with a partial reduction in the response to LPS, consistent withthe expected IRAK independence ofLPS-inducedTRIF (Toll–IL-1 receptordomain–containing adaptor inducing interferon-b) signaling (Fig. 6,A andB)

RAW G

9

THP1 B5

hMDM

s0.0

0.2

0.4

0.6

0.8

1.0

1.2Control si-IRAK4

UntLP

SP3C

PGNR84

8FLG

0

2

4

6

8304050

UntLP

SP3C

PGNR84

8FLG

0.00

0.04

0.08

0.121.52.02.5

IL-6

(ng

/ml)

E TNF- IL-6

A B hMDM

IRAK4

RhoGDI

THP1 B5 RAW G9

**** *** ****

****

****

****

****

****

****

***

****

****

***

UntLP

SP3C

PGNR84

8FLG

05

10152025

200250300

IL-8

(ng

/ml)

***

*** ***

****

****

LPS

P3CPGN

R848

0

2

4

6ControlmIRAK4 KOmIRAK4 KO + mIRAK4mIRAK4 KO + hIRAK4

*** ***

**** ** ****

**** ** **** **

**

ControlhIRAK4

F IL-8 D

G

LPS

P3CPGN

R848

FLG0

1

2

3 Control si-IRAK4C

**

**

***

LPS

0

2

4

6ControlhIRAK4 KOhIRAK4 KO + hIRAK4hIRAK4 KO + mIRAK4

H THP1 IRAK4 KO IMM IRAK4 KO

** **

Irak

4/IR

AK

4

mR

NA

abu

ndan

ce

TN

F-

(ng

/ml)

TN

F-

(ng

/ml)

TN

F-

(ng

/ml)

TN

F-

(ng

/ml)

–/–

Fig. 4. IRAK4 is required for human TLR responses, but the human andmouse IRAK4 proteins are not func-tionally interchangeable. (AandB) Analysis of the relativeabundancesof IRAK4mRNA (A) andprotein (B) inRAW G9 cells, THP1 B5 cells, and hMDMs transfected with control or IRAK4-specific siRNAs. The relativefractions of IRAK4protein abundance in cells treatedwith IRAK4-specific siRNA compared to that in controlcells were as follows: 0.30 ± 0.07 for RAWG9cells (P< 0.001), 0.28 ± 0.10 for THP1B5 cells (P< 0.01), and0.49 ± 0.06 for hMDMs (P < 0.01). Statistical analysis was by two-tailed t test. Western blots are represent-ative of three independent experiments. (C) hMDMs transfectedwith control siRNAor IRAK4-specific siRNAwere stimulated with LPS (10 ng/ml), 100 nM P3C, PGN (10 mg/ml), R848 (10 mg/ml), or FLG (10 ng/ml) for24hours, and the amounts of TNF-a secretedby the cellsweremeasuredbyenzyme-linked immunosorbentassay (ELISA). (D to F) hMDMs isolated from control and IRAK4-deficient patients were left untreated (Unt)or were stimulated with LPS (10 ng/ml), 100 nMP3C, PGN (10 mg/ml), R848 (10 mg/ml), or FLG (10 ng/ml) for24 hours. The amounts of TNF-a (D), IL-6 (E), and IL-8 (F) secreted by the cells were measured by Bio-Plexassay. (G) Immortalized bonemarrow–derivedmouse macrophages (IMM) from IRAK4 KOmice were stablytransducedwith retrovirus expressing eithermouse IRAK4-mCitrine or human IRAK4-mCherry. The cellswerethen stimulated for 24 hours with LPS (10 ng/ml), 100 nM P3C, PGN (30 mg/ml), or R848 (10 mg/ml) before theamount of TNF-a secreted by the cells was measured by ELISA. (H) IRAK4 KO THP1 cells were stably trans-ducedwith retrovirus expressingeither human IRAK4-mCherry ormouse IRAK4-mCitrine. The cellswere thenstimulated for 24 hours with LPS (10 ng/ml) before the amount of TNF-a that they secreted was measured byELISA. Data aremeans ± SD of two representative experiments for (A), (C), (G), and (H) or aremeans ± SD oftwo independent biological replicates from one set of patient blood samples for (D) to (F). **P < 0.01, ***P <0.001, ****P < 0.0001 by two-tailed t test.




(42–44). In contrast, we observed no sub-stantial reduction in single TLR ligand–induced TNF-a and IL-6 secretion byIRAK1-deficient mouse BMDMs (Fig.6, A and B). The marked effect of IRAK2deficiency on TLR-induced cytokine se-cretion by mouse macrophages was dem-onstrated previously (39) and is partly aresult of the role of IRAK2 in stabilizingproinflammatory cytokine transcripts (40).These data emphasize that in response tosingle TLR ligands, mouse macrophagesshow little dependence on IRAK1, as com-pared to IRAK2, for TNF-a and IL-6 se-cretion. These data mirror the effects ofIRAK1and IRAK2depletion in themousemacrophage reporter cell line inour siRNAscreen (Fig. 3C) and validate our mousemacrophage siRNA screening platform.

To examine further the dependency onIRAK1thatweobserved in thehumanmac-rophage reporter–based siRNA screen, wefirst tested the effect of depletion of IRAK1or IRAK2 on the secretion of TNF-a bythe THP1B5 reporter cell line. TNF-a se-cretion induced by LPS (Fig. 6C) or P2C,P3C, or PGN (fig. S7A) was markedly re-duced by IRAK1 knockdown but IRAK2depletion had little effect, which is con-sistent with the effects observed in thesiRNA screen with the TNF-a transcrip-tional reporter. We tested whether differ-ential extents of gene knockdown mightcontribute to the different IRAK effectsobserved in the human cells; however, weobserved substantial and comparable re-ductions in the amounts of mRNAs andproteins of both IRAKs in the knockdowncells (fig. S7, B and C).

To determine whether the greater re-quirement for IRAK1 for TLR signalinginTHP1cellswas also observed in humanprimary cells, we prepared hMDMs fromthe peripheral blood of a normal donor andtransfected them with IRAK1- or IRAK2-specific siRNAsbynucleofection.Weagainfound that the secretion of TNF-a in re-sponse to a range of TLR ligands dependedmostly on IRAK1, with much less depen-dence on IRAK2 (Fig. 6D).We further val-idated this observed pattern in hMDMswith independent siRNA sequences againstIRAK1 or IRAK2 (fig. S7D), andwe alsoconfirmed that the amount of knockdownof IRAK1 and IRAK2 mRNAs and pro-teins in the siRNA-treated hMDMs wassimilar (fig. S7, E and F). These data con-firm that the pattern of IRAK1 and IRAK2dependency that we observed in ourscreen was also found in primary human

0.0

0.2

0.4

0.6

0.8

1.0

pAT

F2

nucl

ear

inte

nsity

(n

orm

aliz

ed to

max

)

G H

P3C0

1

2

3

4ControlmIRAK4 KOmIRAK4 KO + mIRAK4 WTmIRAK4 KO + mIRAK4 Kin–

ControlhIRAK4 KOhIRAK4 KO + hIRAK4 WThIRAK4 KO + hIRAK4 Kin–

P3C0

1

2

3

THP1 IRAK4 KO IMM IRAK4 KO

** **

**

A mIRAK4 KO + mIRAK4

IP: Myd88 IB: mIRAK4

RhoGDI

LPS treatment (min) Input

mIRAK4 0 30 60 120 180

B mIRAK4 KO + hIRAK4

IP: Myd88 IB: hIRAK4

RhoGDI


hIRAK4 0 30 60 120 180

C hIRAK4 KO + mIRAK4

IP: Myd88 IB: mIRAK4

Myd88


mIRAK4 0 30 60 120 180

D hIRAK4 KO + hIRAK4

IP: Myd88 IB: hIRAK4

Myd88


hIRAK4 0 30 60 120 180

– + – + – +0.0

0.2

0.4

0.6

0.8

1.0

pAT

F2

nucl

ear

inte

nsity

(n

orm

aliz

ed to

max

)

P3C:

E mIRAK4 KO + mIRAK4

mIRAK4 KO + hIRAK4

WT IMM F hIRAK4 KO

+ mIRAK4 hIRAK4 KO

+ hIRAK4 WT

THP1

TN

F-

(ng

/ml)

TN

F-

(ng

/ml)

– – –P3C:

Fig. 5. Differential characteristics of human and mouse IRAK4 signaling properties. (A and B) Mouse IRAK4
KO IMMs were stably transduced with retrovirus expressing either (A) mouse IRAK4-mCitrine or (B) humanIRAK4-mCherry before being treated with LPS (10 ng/ml) for the indicated times. Cell lysates were then sub-jected to immunoprecipitation (IP) with an antibody against Myd88, and samples were analyzed byWesternblotting (IB) with an antibody specific for IRAK4. RhoGDI was used as an input control for the Myd88 immu-noprecipitation. (C and D) Human IRAK4 KO THP1 cells were stably transduced with retrovirus expressingeither (C) mouse IRAK4-mCitrine or (D) human IRAK4-mCherry before being treated with LPS (10 ng/ml) forthe indicated times. Cell lysates were then subjected to immunoprecipitation with an antibody specific forMyd88 and analyzed by Western blotting with an antibody specific for IRAK4. Myd88 was used as an inputcontrol for the immunoprecipitation. (E and F) The nuclear intensity of phosphorylated ATF2 (pATF2) wasmeasured by high-content imaging in either (E) mouse IRAK4 KO IMMs or (F) human IRAK4 KO THP1 cellsstably transduced with retrovirus expressing either mouse IRAK4-mCitrine or human IRAK4-mCherry. Cellswere left untreated orwere stimulated for 20min (IMM) or 60min (THP1)with 500 nMP3C.Data are shown forthe central 98th percentile of cells. (G) IRAK4 KO IMMs were stably transduced with retrovirus expressingeither wild-type (WT) or kinase-deficient (Kin−) mouse IRAK4. Cells were stimulated with LPS (10 ng/ml) for24 hours, and the amount of TNF-a secreted was measured by ELISA. (H) IRAK4 KO THP1 cells were stablytransducedwith retrovirus expressing eitherWT or kinase-deficient (Kin−) human IRAK4. Cells were stimulatedwith LPS (10 ng/ml) for 24 hours, and the amount of TNF-a secreted by the cells was measured by ELISA.Western blots in (A) to (D) are representative of at least two independent experiments. Single-cell data in (E)and (F) are representative of two independent experiments (100 to 500 cells imaged per condition, error barsindicate mean ± 95% confidence interval). Data in (G) and (H) are means ± SD of two independent experi-ments. **P < 0.01, ***P < 0.001 by Kolmogorov-Smirnov test (for E and F) or by two-tailed t test (for G and H).


macrophages, and they further indicate that our analysis of the human andmouse TLR pathways identified substantial differences in how human andmouse macrophages decode bacterial PAMPs.

IRAK1 and IRAK2 have different functionalities in thehuman and mouse macrophage TLR signaling pathwaysAs described earlier, signaling through the MyD88-dependent TLR path-way requires the TRAF6-binding property of either IRAK1 or IRAK2to propagate signals downstream to the NF-kB and MAPK signal trans-mission modules (Fig. 1A) (30). Previous studies in mouse macrophagessuggested that IRAK1 is required for early signaling through the MyD88-dependent pathway, whereas IRAK2 is thought to be required for sus-tained signaling (39, 40, 45). Data from our screen suggested that the earlyNF-kB response to P3C andR848 at 40min after stimulation was reducedslightly in IRAK1-depleted mouse cells, whereas the TNF-a reporter re-sponse at 16 hours was unaffected by IRAK1 perturbation (Fig. 3C). Incontrast, IRAK2-depleted mouse cells had markedly reduced TNF-a re-porter activity at 16 hours after stimulation, consistent with an importantrole for IRAK2 in sustained activation. However, we also found thatknockdown of IRAK2 had a more substantial effect on LPS- and R848-induced early NF-kB activation than did knockdown of IRAK1 (Fig. 3C),suggesting a dominant role for IRAK2 in both initial and sustained re-sponses to these TLR ligands in mouse cells.


To determine whether IRAK1 andIRAK2 contributed differently to the initialand sustained responses of human macro-phages, we used CRISPR/Cas9-mediatedgenome editing togenerate THP1 cell linesdeficient in human IRAK1 or IRAK2 (Fig.7A and fig. S8). We first measured the ef-fects of the loss of IRAK1 and IRAK2 onTLR ligand–induced TNF-a secretion bythe different IRAK KO THP1 cell lines.We found that there was a marked reduc-tion in cytokine secretion by the IRAK1KO cells, but there was a much lesser ef-fect of IRAK2 KO (Fig. 7B). We thentested the NF-kB and MAPK phospho-protein responses to P3C in the IRAK KOcells by phosphoflow cytometry and sawa similarly marked reduction in signalingin the IRAK1 KO cells, but again minimaleffects in the IRAK2 KO cells (Fig. 7C).This perturbation of early TLR signaling inhuman macrophages lacking IRAK1 wasin stark contrast to the behavior of IRAK1KO mouse macrophages, which showedrelatively normalMAPKandNF-kBphos-phoprotein responses to TLR stimulation(Fig. 7D).

We next determined the effect of lossof human IRAK1 and IRAK2 on the ex-pression of genes that show different LPSresponse kinetics, namely, early-transient(Nfkbiz), early-intermediate (Tnf), andlate-sustained (Il6) expression. We foundthat the loss of human IRAK2 had no ef-fect on the initial response kinetics of theexpression of any of the three genes; how-ever, IRAK2 was required to maintain

mRNA amounts at times greater than 2 hours after LPS stimulation(Fig. 7E). This finding is consistent with a previously reported role forIRAK2 in maintaining the stability of cytokine mRNAs in both mouseand human macrophages (40, 46), which suggests that this is a conservedfunction of IRAK2between the two species. In contrast, the loss of IRAK1had a substantial effect on the induction of all gene classes in humanmac-rophages (Fig. 7E), consistent with a major role for IRAK1 in the humanTLR response; however, the loss of IRAK1 had little effect on the inductionof the same genes in mouse macrophages (Fig. 7F). These data suggest apreviously unappreciated major difference in the use of IRAK1 betweenhuman and mouse TLR signaling pathways.

To determine whether the differential effects of IRAK KO on the TLR-dependent responses of human andmousemacrophageswere also observedwhen these cells encountered an intact pathogen, we infected cells with thegram-negative bacterium Burkholderia cenocepacia and measured the re-sulting cytokine responses. We observed a similar pattern to the effectsseen with individual TLR ligands, with defective cytokine secretion observedin mouse macrophages lacking IRAK2 or IRAK4 (Fig. 8A) and in humanmacrophages lacking IRAK1 or IRAK4 (Fig. 8B). To further assess whetherthe dependencies of mouse macrophages on IRAK2 and of human macro-phages on IRAK1were reflected in the signaling complexes formed after thecells were stimulated with TLR ligands, we immunoprecipitated Myd88from both cell types after they were treated with LPS. We observed the

Contro

l

si-IR

AK1

si-IR

AK20

2000

4000

6000

LPS

P3CPGN

R848

FLG0

1000

2000

3000Controlsi-IRAK1si-IRAK2

LPS

P3CPGN

R848

0

500

1000

1500

2000

IL-6

(pg

/ml)

WTIRAK1 KOIRAK2 KO

LPS

P3CPGN

R848

0

1000

2000

3000

4000A B

C D

**** ***

**** **

**** ****

* ***

***

**** ****

*** ****

BMDM BMDM

THP1 hMDM

*** **** *

****

*** ****

*** ****

TN

F-

(ng

/ml)

TN

F-

(ng

/ml)

TN

F-

(ng

/ml)

Fig. 6. Differences in IRAK1 and IRAK2 protein use in mouse and human TLR responses in primary macro-phages. (A and B) BMDMs fromWT, IRAK1 KO, and IRAK2 KO mice were stimulated with LPS (10 ng/ml),100 nM P3C, PGN (30 mg/ml), or R848 (10 mg/ml) for 24 hours before the amounts of TNF-a (A) and IL-6 (B)that were secreted by the cells were measured by ELISA. (C) THP1 cells transfected with control, IRAK1-specific, or IRAK2-specific siRNAs were stimulated with LPS (10 ng/ml) for 24 hours before the amounts ofTNF-a that they secreted weremeasured by ELISA. (D) hMDMs transfected with control, IRAK1-specific, orIRAK2-specific siRNAswere stimulatedwith LPS (10 ng/ml), 100 nMP3C, PGN (10 mg/ml), R848 (10 mg/ml),or FLG (10 ng/ml) for 24 hours before the amounts of TNF-a that they secreted were measured by ELISA.Data are means ± SD of four (for A and B) or three experiments (for C and D). *P < 0.05, **P < 0.01, ***P <0.001, ****P < 0.0001 by two-tailed t test.



LPS-dependent coimmunoprecipitation of IRAK2 with Myd88 in mouseIMMs throughout a 3-hour time course, with only weak and transient coim-munoprecipitation ofMyd88 with IRAK1 (Fig. 8C). In contrast, we did notobserve any IRAK2 inMyd88 immunoprecipitates fromhumanTHP1 cells

ww

stimulated with LPS for 3 hours, but we detected human IRAK1 in Myd88immunoprecipitates throughout the same time course (Fig. 8D).

Finally, we sought to determinewhether the apparent species-specificpreferences in IRAK use were reflected in the extent of expression of

0 2 4 6 80

10

20

30

40

50

Time (hours)

Nfk

biz

mR

NA

cha

nge

(fold

)

0 2 4 6 80

20

40

60

80

Time (hours)

Nfk

biz

mR

NA

cha

nge

(fold

)

0 20 40 600

2

4

6

Time (min)

Pho

spho

-p38

MA

PK

(fo

ld c

hang

e)

0 20 40 60

1

2

3

Time (min)

Pho

spho

-p65

/Rel

A

(fold

cha

nge)

WTIRAK1 KOIRAK2 KO

0 20 40 600

1

2

3

4

5

Time (min)

Pho

spho

-p38

MA

PK

(fo

ld c

hang

e)LP

SP3C

PGNR84

80

2

4

6

8ControlIRAK1 KOIRAK2 KO

C A

B

THP1 KO

D IMM

THP1 KO

E THP1 KO

F IMM

IRAK

RhoGDI

IRAK2

THP1 W

T

IB: IRAK1

THP1 IR

AK1 KO

THP1 W

T

THP1 IR

AK2 KO

**** **

**** **

***

****

0 20 40 600

1

2

3

4

Time (min)

Pho

spho

-p65

/Rel

A

(fold

cha

nge)

WTIRAK1 KO

0 2 4 6 80

25

50

75

100

125

Time (hours)

Tnf m

RN

A c

hang

e (fo

ld)

0 2 4 6 80

2

4

6

8

10

12

14

Time (hours)

Il6 m

RN

A c

hang

e (lo

g 2)

WTIRAK1 KOIRAK2 KO

0 2 4 6 80

5

10

15

20

25

Time (hours)

Tnf m

RN

A c

hang

e (fo

ld)

0 2 4 6 80

2

4

6

8

10

12

Time (hours)

Il6 m

RN

A c

hang

e (lo

g 2)

WTIRAK1 KO

TN

F-

(ng

/ml)

Fig. 7. IRAK1 and IRAK2 have different functionalities in the human andmouse macrophage TLR signaling pathways. (A) The WT and indicatedIRAK1KOand IRAK2KOTHP1 cell lines were analyzed byWestern blottingwith antibody against IRAK proteins. RhoGDI was used as a loading con-trol. Western blots are representative of two independent experiments. (B)The indicated control and IRAK KO THP1 cell lines were stimulated for4 hours with LPS (10 ng/ml) before the amounts of TNF-a that they secretedweremeasured by ELISA. (C andD) TheWT and indicated IRAK KO THP1cell lines (C) and mouse IMMs (D) were stimulated with 500 nM P3C for the

indicated times before the relative abundances of phosphorylated p38aMAPK (left) and p65 (right) were measured by phosphoflow cytometry.(E and F) The WT and indicated IRAK KO THP1 cell lines (E) and mouseIMMs (F) were stimulated with 500 nM P3C for the indicated times beforethe relative abundances ofNfkbiz (left), Tnf (middle), and Il6 (right) mRNAswere measured by quantitative reverse transcription polymerase chain re-action (qRT-PCR) analysis. Data in (B) to (F) are means ± SD of two inde-pendent experiments. **P< 0.01, ***P< 0.001, ****P< 0.0001 by two-tailedt test.

w.SCIENCESIGNALING.org 5 January 2016 Vol 9 Issue 409 ra3 10


these genes in human andmouse cells.We used the numerous expressiondata sets available in Gene Expression Omnibus (GEO) to determinewhether there were any species-specific trends in the abundances ofIRAK1, IRAK2, and IRAK4 mRNAs. We used data from the Illuminaplatforms GPL6884 (Illumina HumanWG-6 v3.0 expression beadchip)and GPL6885 (IlluminaMouseRef-8 v2.0 expression beadchip) becausethey used the same beadchip technology. We downloaded all availablenormalized gene expression data for these platforms (>6000 samples forhuman, >8000 samples for mouse) and calculated the expression distri-bution for the IRAK probes (fig. S9). We observed that IRAK1 was ex-pressed to a greater extent than was IRAK2 in human cells, whereas inmouse cells, the trend was reversed. The expression of human IRAK1and mouse Irak2 also showed greater variability (higher SD) acrossthe thousands of mRNA expression data sets available in GEO. Thisfinding suggests that the expression of human IRAK1 and mouse Irak2is specifically regulated across different cell types, which may be relatedto the species-specific roles in TLR signaling of their gene products thatwe have described in this study.

DISCUSSION

We previously described the development of an assay platform in humanand mouse macrophages for the high-throughput, siRNA-based screeningof responses to pathogenic stimuli (23). As a first step toward genome-wide

ww

screens, we described here the targeting of putatively canonical pathwaycomponents to assess TLR response requirements in a key innate immunecell type, and we also compared human and mouse pathway dependencies.Although the screen may have identified some characteristics that are dis-tinct to the transformed macrophage cell lines used, we demonstrated vali-dation of the findings from these cell line screens in primary human andmouse cells.Moreover, the broad use of the RAW264.7 and THP1 cell linesas model systems for the study of mouse and human macrophage biologymakes the presented data set a valuable resource for future studies.

At the level of signaling modules used by the TLR pathway, we ob-served considerable conservation between mouse and human cells in themodules containing the strongest phenotypes, namely, the ligand-specificreceptors, proximal signaling proteins, and the regulators of human TNFand mouse Tnf transcription. The slightly weaker phenotypes observedafter branching of the pathway into the NF-kB, MAPK, and PI3K-Aktmodules would suggest a degree of redundancy in the routing of TLR sig-nals that may provide robustness to individual gene perturbations or tocompetition for these widely used signaling components from other path-ways. One unexpected outcome in the screen was the relative lack of in-creased signals observed among the module of negative regulators. Theassay system was capable of identifying known negative regulatory com-ponents, such as the protein TANK in the TRIFmodule (fig. S2); however,aside from a few cases of increased reporter output through the knock-down of the genes encoding A20, Cyld, andMkp1 (Dusp1), we observedfew strong negative regulator hits. Because many of these regulators op-erate through a delayed feedback mechanism that requires their transcrip-tional induction, it may be that our screen readouts of NF-kB activation andprimary gene response were measured too early to fully capture the effectof perturbing these feedback regulators.

In comparing the hits for the different TLR ligands, we observed that thestrongest hits tended to be common across all of the ligands tested, suggest-ing a number of nonredundant, sharedTLRpathway components downstreamof the different receptors; however, there were also numerous instances ofgene knockdowns with ligand-selective effects. This suggests that althoughdifferent TLR pathways may use a number of shared core components, thereis a preferred use of certain signaling proteins in the pathways engaged byindividual TLR ligands, which may lead to the activation of transcriptionalprograms and macrophage responses that are customized to the PAMPspresented by different pathogens. Clustering analysis showed that P3C andR848 were more similar to each other in their pathway component require-ments than they were to LPS, which is likely reflective of their signaling ex-clusively through MyD88, whereas LPS signals through both MyD88 andTRIF.This is supported by the presence of TRIF in thegenephenotype clusterthat was shared by human and mouse cell lines in response to LPS (fig. S3).

Beyond the species conservation of strongest hit enrichment amongthe TLR receptors, proximal signaling proteins, and human TNF/mouseTnf transcription modules, we observed substantial cross-species differ-ences in the individual gene requirements for TLR signal propagation inour human and mouse macrophage screens. In addition to the selectivityfor the IRAK family that we have highlighted in this study, notable addi-tional findings include a strong requirement in mouse cells for CD14, IkBkinase b (IKKb), and the transcriptional regulating complex CBP/p300.Human cells appeared to be more sensitive to knockdown of TAK1, Tpl2,and ERK1 (extracellular signal–regulated kinase 1), and they had an un-expected positive regulatory role for the NF-kB inhibitory protein IkBe.This latter observation may warrant further investigation because we alsofound that knockdown of this protein rendered THP1 cells more sensitiveto infection by gram-negative bacteria. The specific gene dependencies forhuman and mouse TLR pathways identified in this study may provide in-sight to facilitate future systematic analyses of these pathways.

B. cenocepacia0

1

2

3

4

IL-6

(ng

/ml)

IMM WTmIRAK1 KOmIRAK2 KOmIRAK4 KO

B. cenocepacia0.0

0.2

0.4

0.6

0.8

IL-6

(ng

/ml)

THP1 WThIRAK1 KOhIRAK2 KOhIRAK4 KO

A B

C Mouse IMM

IP: Myd88 IB: IRAK2

RhoGDI

LPS treatment (min)

Input 0 30 60 120 180

IP: Myd88 IB: IRAK1

***

*** ***

***

Human THP1 Mouse IMM

D Human THP1

IP: Myd88 IB: IRAK2

RhoGDI

LPS treatment (min)

Input 0 30 60 120 180

IP: Myd88 IB: IRAK1

Fig. 8. Differential human and mouse IRAK dependencies are reflected inresponses to gram-negative bacterial infection and in ligand-induced
signaling complexes. (A and B) IMMs fromWTmice and the indicated IRAKKOmice (A) and theWT and indicated IRAK KO THP1 cell lines (B) were in-fected with B. cenocepacia at a multiplicity of infection of 1. Twenty hourslater, the amounts of IL-6 secreted by the cells were measured by ELISA.(C andD) IMMs from IRAK4KOmice (C) and IRAK4KOTHP1 cells (D) werestably transducedwith retrovirus expressingeithermouse IRAK4-mCitrine orhuman IRAK4-mCherry, respectively. The cells were then treated with LPS(10 ng/ml) for the indicated times before cell lysates were subjected to im-munoprecipitation with antibodies against Myd88 and analyzed by Westernblotting with antibodies against either IRAK1 or IRAK2. RhoGDI is shown asan input control for theMyd88 immunoprecipitation.Western blots are repre-sentative of three independent experiments. Data in (A) and (B) aremeans±SD of two independent experiments. ***P < 0.001 by two-tailed t test.


Our in-depth study of the different patterns of IRAK dependency inhuman and mouse TLR signaling, in addition to identifying a clear differ-ence in pathway architecture between species, also validated our screeningplatform for systematic analysis of macrophage responses to pathogens.We showed that the responsiveness of primary macrophages from IRAKKO mice to TLR ligands was in broad agreement with the results of ourscreen, and the effects of knockdown of the human IRAKs in blood-derivedprimary cells were also consistent with the effects that we observed in thehuman screen. Althoughwe found that both primary humanmacrophagesand THP1 cells were resistant to substantial depletion of IRAK4, humancells with a complete IRAK4 deficiency showed a loss of responsiveness toTLR ligands. This was in contrast to mouse cells in which siRNA-mediatedIRAK4 knockdown led to a strong defect in responsiveness. The differencein sensitivity to IRAK4 knockdownmay therefore reflect different species-specific properties of the human andmouse IRAK4 proteins, which is sup-ported by our finding that both human andmouse IRAK4were unable torescue TLR responses in IRAK4-deficient cells from the other species.Previous studies demonstrated that kinase-inactive mouse IRAK4 can-not rescue IL-1 or TLR signaling in IRAK4-deficient mouse fibroblasts(34) or macrophages (47), respectively, but that kinase-inactive humanIRAK4 can rescue IL-1 signaling in IRAK4-deficient human fibroblasts(33), and inhibition of IRAK kinase activity with a broad-specificity IRAKinhibitor has a minimal effect on TLR responses in several human immunecell types (48). We confirmed that kinase-inactive human IRAK4 rescuedTLR responses in THP1 cells (Fig. 5H), which suggests that the functionof IRAK4 in humans might be limited to its ability to recruit IRAK1 andthat the kinase activity of IRAK1 and its TRAF6 recruitment domain mayplay amuchmore critical role in TLR signaling in human cells than inmousecells. In light of our siRNA screen findings, the greater dependence of mousecells on the kinase activity of IRAK4may render themmore sensitive to thedepletion of this protein, whereas human cells may tolerate substantial re-ductions in the abundance of IRAK4protein, because they are less reliant onthis property. Furthermore, studies of mouse IRAK4 have suggested that ithas important scaffolding functions for the establishment of a myddosomestructure after TLR activation (28). Our data suggest the possibility thathuman IRAK1 compensates for a substantial depletion of IRAK4 in thehuman myddosome, but that mouse IRAK1 or IRAK2 is unable to simi-larly compensate for the depletion of mouse IRAK4.

Our data also demonstrate that the relative roles of IRAK1and IRAK2 in theprimary propagation of the TLR signal may be different between human andmouse, with there being a much stronger requirement for IRAK1 in humancells.We used the CRISPR/Cas9 genome editing technology tomake humanTHP1 cells completely deficient in IRAK1or IRAK2, andwe showed that theinitial signaling, transcriptional, and cytokine responses to TLR ligands and in-tact bacteria were particularly dependent on IRAK1 in those cells. Consistentwith previous studies, human IRAK2 has an important role in maintainingTLR-induced gene transcription beyond 2 hours after ligand stimulation (46).

It has been known for some time that the TLR-induced cytokine re-sponse of IRAK1 KO mice is comparable to that of wild-type mice, incontrast to the markedly reduced response observed in IRAK2 KO mice(38, 39, 41). Our cytokine data confirm that responses to individual TLRligands and to intact bacteria in mouse IRAK1KO cells remained relativelyintact, in stark contrast to those in cells lacking IRAK2, which showed al-most no cytokine secretion. Moreover, NF-kB and MAPK signaling andearly transcriptional responses in TLR2-stimulated mouse macrophageswereminimally affected by IRAK1deficiency. Previous studies showed thatmouse IRAK1 plays a critical role in interferon responses to endosomalTLR ligands in plasmacytoid dendritic cells (49), and mouse IRAK1 facili-tates the rapid, TLR ligand–dependent priming of the NLRP3 (NLR family,pyrin domain–containing 3) inflammasome (50, 51). Although these func-

ww

tions of IRAK1 have not been carefully compared inmouse and human cells,they suggest that mouse and human macrophages could have evolved dif-ferent functions for IRAK1 that are tuned to the differential innate immunechallenges faced by each organism. It is also noteworthy that among thehuman IRAKs, polymorphisms in IRAK1, in particular, have been linkedwith susceptibility to a range of immune disorders (52–54), supporting anespecially important role for this IRAK family member in human cells. Fur-thermore, in mouse models of human diseases in which IRAK1 polymorph-isms are implicated, there are examples in which experimentally inducedautoimmunity is defective when mouse IRAK1 is absent (55, 56); however,there are no examples inwhich the expression of a human IRAK1variant thatis implicated in a disease can drive disease in a mouse model. Our study es-tablishes a critical role for IRAK1 in humanTLRsignal transduction thatmayhelp explain the immune consequences of genetic alterations at the IRAK1locus. In addition, human IRAK1 is ubiquitously expressed, whereas mouseIRAK1shows amore selective expressionpattern (57,58), again supporting amore prominent role for IRAK1 in human cells.

Other studies have compared gene expression data sets from inflam-matory disease states in humans and mice and have arrived at contradictoryconclusions as to how accurately mousemodels canmimic human disease(21, 22). Notably, one aspect in which both of these studies agree is in theobservation that genes and pathways regulating the innate immune response,such as the TLR signaling pathway, are highly enriched during responses totrauma, burns, or sepsis in both species. Despite this pathway enrichment,these studies conclude that individual gene expression changes within apathway are often not predictive of a corresponding change in the ortho-logous gene in the other species, and it is proposed that the failure of nu-merous anti-inflammatory drugs in clinical trials can often be attributed tothe targeting of an individual gene product that does not show the same be-havior between human and animal models. The findings from our siRNAscreens are broadly supportive of these previous studies, because we ob-served enrichment within the same TLR pathway modules in human andmouse cells, but relatively poor conservation in dependencies at the indi-vidual gene level. Comprehensive genetic screening of individual gene con-tributions between human and mouse innate immune cells thus providesimportant insight to this gene-level variation and could identify more opti-mal anti-inflammatory drug targets in human cells.

In summary, we have performed a comprehensive siRNA-based screenof the canonical TLR pathway in human and mouse macrophages to gen-erate a valuable resource data set that systematically identified those pro-teins that are most required for the response to different TLR ligands in akey innate immune cell type. Beyond the species-conserved use of certainligand-specific receptors and core signaling modules, we report considera-ble differences in human and mouse TLR pathway dependencies at thelevels of individual genes and their products. In particular, we identifiedand experimentally validated a previously unappreciated difference inIRAK family use between mouse and human macrophages, which mayhave relevance for ongoing clinical efforts to target this kinase family to reg-ulate TLR-dependent inflammatory responses or for the development oftherapeutics for autoimmune disorders that are influenced by polymorph-isms in IRAK family genes.

MATERIALS AND METHODS

Cell culture and stimulation with TLR ligandsRAW264.7 G9 cells and IMMs derived from wild-type, Irak1−/−, andIrak4−/− mice were maintained in Dulbecco’s modified Eagle medium(glucose, 4.5 g/liter), 10% fetal bovine serum (FBS), 20 mMHepes, and2 mM glutamine. BMDMs from wild-type, Irak1−/−, and Irak2−/− mice



were prepared by differentiation for 6 days in the same culture mediumcontaining macrophage colony-stimulating factor (M-CSF; 60 ng/ml,R&D Systems). THP1 B5 and the various THP1 IRAK KO cell lines weremaintained in RPMI 1640, 10% FBS, and 2 mM glutamine [containingG418 (500 mg/ml) for the THP1 B5 cells]. After written informed consentwas provided, normal donor and IRAK4-deficient patient blood was ob-tained at the National Institutes of Health (NIH) under Institutional Re-view Board (IRB)–approved protocols. The IRAK4-deficient patient hasa heterozygous C-to-T mutation at nucleotide 877 in the IRAK4 codingsequence, which causes premature termination at Gln293. ComplementaryDNA (cDNA) sequence analysis showed that only the mutant allele is ex-pressed. PBMCswere prepared by a 1:1 dilution of a human blood samplewith Hanks’ balanced salt solution (HBSS) and layering onto the surfaceof lymphocyte separation medium (MP Biomedicals, LLC, 50494) at a2:1 ratio. After centrifugation at 625g for 10 min, mononuclear cells wererecovered from the opaque interface, erythrocytes were lysed with ACKlysing buffer (Quality Biological, 118-156-721), and the remaining cellswere washed with HBSS to provide purified PBMCs. hMDMs were de-rived from PBMCs cultured in cRPMI (RPMI 1640 medium containing10% FBS, 10 mM Hepes, 0.1% b-mercaptoethanol, and 2 mM glutamine)with granulocytemacrophage colony-stimulating factor (GM-CSF; 10 ng/ml,R&D Systems, catalog no. 215-GM-010) for 7 days. LPS was from AlexisBiochemicals (Salmonella minnesota R595 TLRgrade, catalog no. ALX-581-008-L002), P3C was from InvivoGen (catalog no. tlrl-pms), PGN wasfrom Sigma (Staphylococcus aureus PGN, catalog no. 77140), R848 wasfrom InvivoGen (catalog no. tlrl-r848), P2Cwas fromEMCMicrocollections(catalog no. L2020), and FLGwas from InvivoGen (FLA-STultrapure, cat-alog no. tlrl-epstfla).

Dual luciferase assays in THP1 B5 cellsThe THP1 B5 cell clone was generated as described previously (23). Cellswere differentiated into a macrophage-like state with phorbol 12-myristate13-acetate (PMA; 5 ng/ml, Sigma, P1585) for 72 hours. The differentiatedcellswere treatedwith TLR ligands for activation or with culturemedium asa control. After stimulation, firefly and Renilla luciferase activities in the celllysates were determined with the Dual-Luciferase Reporter Assay System(Promega, E1960), and the firefly/Renilla luminescence ratiowas calculatedto reflect the response of the cells to stimulation.

High-content imaging of RAW G9 cellsThe RAW G9 clone was generated as described previously (23, 59). TheGFP-p65 and Tnf promoter–mCherry reporters in RAW G9 cells were im-aged with a BD Pathway 855 bioimager (BD Biosciences). BDAttoVisionsoftwarewas used to automatically identify and quantify nuclei stainedwithHoechst 33342 (Invitrogen, H3570), GFP-p65, and mCherry fluorescence.GFP located within the area of nuclear staining (eroded by 2 pixels) wasdefined as nuclear NF-kB, whereas GFP within a 2-pixel-wide ring aroundthe area of nuclear staining was defined as cytosolic NF-kB. To determinethe extent of nuclear translocation of NF-kB after ligand stimulation, theratio of the intensities of nuclear to cytoplasmic GFP-p65 was calculatedwith BD Image Data Explorer software. For mCherry expression, nuclearmCherry was quantified with the same method that was used for NF-kB.Background red fluorescence was subtracted, and the average intensity wasused as a measure of Tnf promoter activity.

siRNA screenThe TLR pathway 126-gene set (Fig. 1A) was targeted with six uniquesiRNA sequences (three each from Ambion and Qiagen) per gene (tableS1), and individual siRNAs were arrayed in separate regions of 384-wellplates (Fig. 1B). Three central columns in each plate contained replicatewells

ww

of negative controls [nontargeting siRNA; NTC#2, 3, and 5 (Dharmacon),lipid only and cells only] and positive controls (mouse: Myd88, Irak4, andIkbkg; human:TLR4/TLR8,TIRAP, and IKBKG). To control for cell passagevariability, reporter cells were always thawed from a common stock 2weeksbefore transfection. After the cells were transfected with siRNAs accordingto previously described optimized protocols (23), THP1B5 cells were stim-ulated with TLR ligands for 4 hours and then were subjected to dual lucif-erase assays. Duplicate siRNA plates were run for RAWG9 cells to enableone plate set to be treated with TLR ligand for 40 min (for NF-kB readout)and the second plate set to be treated for 16 hours (for TNF-a reporter read-out). RAWG9 cells were fixed in 4% paraformaldehyde (PFA) (PolyScienceInc.), and nuclei were stained with 600 nM 4′,6-diamidino-2-phenylindole(Invitrogen) before high-content imaging was performed. Replicates wererun for all screens with separate batches of reporter cells. Raw data from eachreporter assay were normalized by the following intraplate median/MAD(median absolute deviation) calculation: (sample − median (all samples))/MAD (all samples). Plate comparisons between replicates all showedSpearman correlations >0.7. Robust z scores relative to negative controlswere calculated, and average values from replicates were calculated foreach single siRNA (see table S2). The median z score of the six individualsiRNAs for each of the 126 screened genes was used as the readout forspecific gene effects on the response to each TLR ligand (table S2). Esti-mations of false-positive and false-negative rates in the human and mousesiRNA screens were calculated with negative and positive siRNA controlsthat were reproducible between experiments and showed a reliable signalperturbation phenotype. Because false-positive and false-negative ratesare dependent on the selection of the screen score threshold, we estimatedthem over a wide range of thresholds. In this screen, we expected that thepositive control wells would have high negative z scores. Hence, the false-negative rate can be calculated as the percentage of individual positivecontrol wells with scores greater than the chosen threshold (ranging from−0.5 to −3), that is, not passing the threshold. We used negative controlwell scores for estimating the false-positive rates. Because the screen siRNAscores were normalized to the negative controls, we expected that the in-dividual negative controlwellswould have both positive and negativevalues(with a median value of 0), so false-positive rates were estimated in bothdirections. The percentage of wells with scores deviating from 0 beyond athreshold (ranging from −0.5 to −3 and 0.5 to 3) was used to estimate false-positive rates. The following individual human IRAK-specific siRNAswereused in postscreen experiments (all fromAmbion): IRAK1 (s323 and s322),IRAK2 (s7497 and s7495), and IRAK4 (s27527 and s27528).

Quantitative RT-PCRTotal RNAwas extracted with the RNeasy Micro Kit (Qiagen, 74004) andRNeasy Mini Kit (Qiagen, 74106). cDNA was synthesized from the ex-tracted RNA with the iScript cDNA Synthesis Kit (Bio-Rad, 170-8891).qRT-PCR assays were performed with Thermo Scientific Solaris qPCRROXMaster Mix (AB-4351/C) or with Power SYBR Green PCR MasterMix (Applied Biosystems, catalog no. 43091655) with a Mastercycler re-alplex4 real-time PCR detection system (Eppendorf), according to themanufacturers’ protocols. The abundances of themRNAs of interest in eachsample were normalized to that of ACTB/Actb orHPRT1/HprtmRNA, andfold changes in the abundances of target mRNAs relative to their basalabundances were calculated with the 2−DDCt method. The primers andprobes for human and mouse IRAK1/Irak1,MAP3K7/Map3k7, TAB2/Tab2,NFKBIE/Nfkbie, MAP3K8/Map3k8, MAPK3/Mapk3, MAPK14/Mapk14,IRF9/Irf9, IRAK2/Irak2, IRAK4/Irak4, CD14/Cd14, IKBKB/Ikbkb,MAPK1/Mapk1, CREBBP/Crebbp, EP300/Ep300, TNF/Tnf, NFKBIZ/Nfkbiz, and IL6/Il6 were purchased from GE Healthcare, Dharmacon(Solaris Mouse or Human qPCR Gene Expression Assay).



Cytokine ELISABMDMs, IMMs, hMDMs, or differentiated THP1 cells (2 × 105/ml) weretreated with the appropriate TLR ligands, the cell culture medium wascollected, and the concentrations of TNF-a or IL-6 secreted by the cellsweredetermined by ELISA. The mouse TNF-a ELISA kit was purchased fromR&D Systems (DY410), whereas the mouse IL-6 (555240) and humanTNF-a (555212) kits were purchased from BD Biosciences.

Cytokine Bio-PlexPBMCs were plated into a 96-well plate at 1 × 106 cells/ml and stimulatedwith the appropriate TLR ligands for 16 hours before cell culture mediumwas collected. PBMCs (1 × 107) from the IRAK4-deficient patient or ahealthy control were cultured in 10 ml of cRPMI containing GM-CSF(10 ng/ml; R&DSystems, 215-GM-010) for 7 days to induce the generationof hMDMs. The hMDMs were then plated at 1.5 × 105 cells/ml and stimu-lated with the appropriate TLR ligands for 16 hours before the cell culturemedium was collected. The amounts of secreted TNF-a, IL-6, MIP-1a (macrophage inflammatory protein–1a), and IL-8 were determinedwith the Bio-Plex assay (Bio-Rad).

Western blottingRAW cells, THP1 cells, or hMDMs were lysed in radioimmunoprecipita-tion assay buffer (Sigma, R0278) containing a protease inhibitor cocktail(Roche). Cell lysates were quantified by protein assay (Bio-Rad), and equalprotein amountswere resolved with a 4 to 12%Bis-Tris Gel/MOPSRunningBuffer System (Invitrogen) and transferred to nitrocellulose membranes. Themembraneswere analyzed byWestern blottingwith the following antibodies:rabbit anti-IRAK1 (Cell Signaling, 4504S), rabbit anti-IRAK4 (Abcam,ab13685 or ab32511), rabbit anti-IRAK2 (Abcam, ab62419, and ProSci,3595), rabbit anti–Rho-GDI (Sigma, R3025), and horseradish peroxidase(HRP)–conjugated donkey anti-rabbit secondary antibody (GE Healthcare).

Expression of mouse or human IRAK4 fluorescentprotein fusions in IRAK4 KO IMMs and THP1 cellsThe pR-IRAK4-mCitrine and pR-IRAK4(Kin−)-mCitrine retroviral plas-mids, which express wild-type or kinase-deficient (35) murine IRAK4,respectively, with an mCitrine fusion at the C terminus, were generated byamplifying the cDNA encoding murine IRAK4 from expression vectorswith the following specific primers: forward, 5′-TTTCGCGCGCGAT-GAACAAGCCGTTGACACCATC-3′; reverse, 5′-AAACTCGAGAGC-AGACATCTCTTGTAGCAGC-3′. The resulting PCR product wasdigested with Mau BI and Xho I and subcloned in-frame into the corre-sponding sites of the pR-mCitrine retroviral vector. Entry vectors containingthe human wild-type IRAK4 sequence (Addgene #23749: pDONR223-IRAK4) or the human IRAK4 kinase–deficient mutant (36) were recom-bined with pDS-FBhyg_X-mCH by Gateway cloning to generate retroviralplasmids that expressed either wild-type or kinase-deficient (Kin−) humanIRAK4 with an mCherry fusion at the C terminus. A control retrovirus wasalso generated that expressed mCherry alone. Immortalized IRAK4 KOmouse macrophages or human IRAK4 KO THP1 cells expressing murineIRAK4-mCitrine (either wild type or Kin−), human IRAK4-mCherry (eitherwild type or Kin−), or mCherry alonewere generated by retroviral transduc-tion as previously described (60). Cells were sorted on the basis of positiveexpression of mCitrine or were selectedwith hygromycin B (Clontech), andthen single-cell clones were expanded in 96-well tissue culture plates bylimiting dilution and were functionally tested by ELISA for reconstitutionof cytokine secretion in response to TLR activation. The abundances of themurine IRAK4-mCitrine and human IRAK4-mCherry fusion proteinsweredetermined either by flow cytometric analysis or by qRT-PCR assays withthe following primer pairs: mouse Irak4 forward, TGCATGAGAAGAA-

ww

AAACAGACG;mCitrine reverse,GGACACGCTGAACTTGTGG; humanIRAK4 forward, GACATTAAGAAGGTTCAACAGC; mCherry reverse,GCCATGTTATCCTCCTCG.

Generation of THP1 CRISPR/Cas9 KO linesTHP1 cells were plated at a density of 2 × 105/ml. After 24 hours, 2.5 × 106

cells were resuspended in 250 ml of Opti-MEM, mixed with 5 mg ofCRISPR/Cas9 plasmid DNA, and electroporated in a 4-mm cuvette withan exponential pulse at 250 V and 950 mF with a Gene Pulser electro-porating device (Bio-Rad Laboratories). We used a plasmid encoding aCMV-mCherry-Cas9 expression cassette and a human IRAK-specificguide RNA (gRNA) (figs. S5B and S8) driven by the U6 promoter (61).Critical exons of the IRAK genes were targeted with the following gRNAconstructs: IRAK1, 5′-GCTGAGGATGGCAACTTCCGGGG-3′;IRAK2, 5′-GCAGGGTGTGAGCATCACGCGGG-3′; IRAK4, 5′-GAT-GAACGACCCATTTCTGTTGG-3′. Cells were allowed to recover for 2days in six-well plates filled with 4 ml of medium per well before be-ing fluorescence-activated cell sorting (FACS)–sorted for mCherry ex-pression with a BD FACSAria III sorting device (BD Biosciences). Forsubsequent limiting dilution cloning, cells were plated at a density of 5,10, or 20 cells per well in nine round-bottomed 96-well plates and culturedfor 2 weeks. Plates were scanned for absorption at 600 nm, and growingclones were identified with customized software and were picked and du-plicated by a Biomek FXp (Beckman Coulter) liquid handling system.One duplicatewas used to recover gDNAand characterize the gene editingas previously described (62).

Immunoprecipitation of TLR signaling complexesFor immunoprecipitation experiments, whole-cell extracts from 1 × 107

cells per condition were prepared in 600 ml of cell lysis buffer containing1% NP-40, 50 mM tris-HCl (pH 7.5), and 150 mM NaCl. Lysates werecentrifuged at 18,000g for 10 min at 4°C. The cleared supernatants werecollected, and 50 ml of the supernatant was saved as the input sample.Onemicrogram of anti-MyD88 antibody (R&DSystems; AF3109, goat)for immunoprecipitation from mouse IMMs or 10 ml of anti-MyD88 anti-body (Cell Signaling Technology; 3699, rabbit) for immunoprecipitationfrom humanTHP1 cellswas added to 50 ml of Dynabeads ProteinG (Novex,10003D) or Dynabeads ProteinA (Novex, 10001D) in 200 ml of phosphate-buffered saline (PBS) containing 0.1% Tween 20 (PBST) for 20 min atroom temperature to form bead-antibody complexes. The remaining celllysate supernatants (about 550 ml) were added to the bead-antibodycomplexes and incubated with rotation overnight at 4°C. The beads werewashed three times with PBST, and then the immunoprecipitated com-plexes were eluted in 25 ml of a 1:1 dilution of elution buffer (Invitrogen)and 2× SDS buffer (Quality Biological, catalog no. 351-082-661) by heat-ing at 70°C for 10 min. Twenty microliters of eluted protein complexeswas resolved with a 10% Bis-Tris Gel/MOPS Running Buffer System(Invitrogen) and visualized by Western blotting analysis with the followingantibodies: rabbit anti-IRAK1 (Cell Signaling, 4504S), mouse anti-IRAK4(Pierce, MA5-15883), rabbit anti-IRAK2 (Abcam #ab62419 or ProSci#3595), mouse anti-MyD88 (Santa Cruz Biotechnology, sc-74532), rabbitanti–Rho-GDI (Sigma, R3025), and HRP-conjugated ECL anti-rabbitimmunoglobulin G (IgG) or anti-mouse IgG secondary antibody (GEHealthcare).

Imaging of the TLR-activated phosphorylated ATF2response in IMMs and THP1 cellsIMMs were seeded in 96-well plates 24 hours before they were treatedwith TLR ligands. THP1 cells were differentiated with PMA in 96-wellplates 3 days before they were treated with ligands. After treatment with



500 nM P3C, the cells were fixed in 4% PFA and blocked with PBSTcontaining 5% (w/v) bovine serum albumin. The fixed cells were treatedwith primarymouse antibody against pATF2 (1:200 dilution, Santa CruzBiotechnology, sc-8398) followed by Alexa Fluor 647–conjugated goatanti-mouse IgG secondary antibody (1:1000 dilution, Life Technologies,catalog no. A-21235) or Alexa Fluor 700–conjugated goat anti-mouse IgGsecondary antibody (1:1000 dilution, Life Technologies, catalog no. A-21036).The cells were imaged on either a Thermo Scientific CellInsight NXTor aLeica SP8.

Phosphoflow cytometryTHP1 cells or IMMs were treated with 500 nM P3C (InvivoGen, catalogno. tlrl-pms) for the times indicated in the figure legends. The cells werefixed with 1% PFA, detached, and permeabilized with ice-cold methanol.The cell suspensions were incubated with purified rat anti-mouse CD16/CD32 (BD Biosciences, Mouse BD Fc Block, clone 2.4G2, catalogno. 553141) to block nonspecific binding and stained with Alexa Fluor555–conjugated anti–phosphorylated p38 MAPK (Cell Signaling, 9836S),Alexa Fluor 647–conjugated anti–phosphorylated p65 NF-kB (CellSignaling, 4887S), or Alexa Fluor 488–conjugated anti–phosphorylatedp65 NF-kB (BD Biosciences, 558421) in PBS containing 1% fetal calfserum (FCS), 2 mM EDTA.

Bacterial infection of macrophagesInfection of both THP1 cells and IMMcell lineswithB. cenocepaciawasperformed with the infection protocol described previously (63). Afterincubation of cells with the bacteria, the cell culture medium was collected,and the amounts of secreted cytokines were measured by ELISA as de-scribed earlier.

Calculation of expression distributions from GEOdata setsAll available normalized gene expression data for platforms GPL6884(Illumina HumanWG-6 v3.0 expression beadchip, >6000 samples) andGPL6885 (Illumina MouseRef-8 v2.0 expression beadchip, >8000samples) were downloaded from GEO (www.ncbi.nlm.nih.gov/geo/).To make the samples comparable and eliminate potential batch effects,expression data from each sample were transformed by calculating therobust z score from the median andMAD of individual samples. Expres-sion z scores for probes from all available samples corresponding to hu-man IRAK1, IRAK2, and IRAK4 and mouse Irak1, Irak2, and Irak4werebinned into 500 equal bins between −10 and 10 to generate probabilitydistributions.

SUPPLEMENTARY MATERIALSwww.sciencesignaling.org/cgi/content/full/9/409/ra3/DC1Fig. S1. Design and TLR ligand response of mouse and human macrophage reporter celllines for siRNA screening applications.Fig. S2. Effects of siRNA-mediated gene perturbations across the human and mouse TLRpathways.Fig. S3. Human and mouse macrophages show both shared and distinct gene dependenciesin TLR signaling.Fig. S4. IRAK4 is required for the responses of human PBMCs to TLR ligands.Fig. S5. Generation of IRAK4 rescue cell lines from IRAK4 KO IMMs and THP1 cells.Fig. S6. Expression of kinase-deficient mouse and human IRAK4 in IRAK4 KO mouseIMMs and THP1 cells.Fig. S7. Analysis of the IRAK1 dependency of human macrophages for TLR responses.Fig. S8. CRISPR/Cas9-mediated targeting of the human IRAK1 and IRAK2 loci in THP1 cells.Fig. S9. Expression distributions of IRAK1/Irak1, IRAK2/Irak2, and IRAK4/Irak4 mRNAs inhuman and mouse GEO data sets.Table S1. Details of the siRNAs used to target the 126 human and mouse TLR pathwaygenes.

ww

Table S2. siRNA z scores from screens of the 126 human and mouse TLR pathwaygenes.Table S3. Ranking scores of human and mouse TLR pathway genes from the siRNAscreens.Table S4. Human versus mouse z score differences across matched TLR ligands andassay readouts.

REFERENCES AND NOTES1. T. Kawai, S. Akira, The roles of TLRs, RLRs and NLRs in pathogen recognition. Int.

Immunol. 21, 317–337 (2009).2. E. M. Creagh, L. A. J. O’Neill, TLRs, NLRs and RLRs: A trinity of pathogen sensors

that co-operate in innate immunity. Trends Immunol. 27, 352–357 (2006).3. R. Barbalat, S. E. Ewald, M. L. Mouchess, G. M. Barton, Nucleic acid recognition by

the innate immune system. Annu. Rev. Immunol. 29, 185–214 (2011).4. J. Wu, Z. J. Chen, Innate immune sensing and signaling of cytosolic nucleic acids.

Annu. Rev. Immunol. 32, 461–488 (2014).5. B. K. Davis, H. Wen, J. P.-Y. Ting, The inflammasome NLRs in immunity, inflammation,

and associated diseases. Annu. Rev. Immunol. 29, 707–735 (2011).6. S. Akira, S. Uematsu, O. Takeuchi, Pathogen recognition and innate immunity. Cell

124, 783–801 (2006).7. B. A. Beutler, TLRs and innate immunity. Blood 113, 1399–1407 (2009).8. S. L. Foster, R. Medzhitov, Gene-specific control of the TLR-induced inflammatory

response. Clin. Immunol. 130, 7–15 (2009).9. R. Ostuni, I. Zanoni, F. Granucci, Deciphering the complexity of Toll-like receptor

signaling. Cell. Mol. Life Sci. 67, 4109–4134 (2010).10. A. Fire, S. Xu, M. K. Montgomery, S. A. Kostas, S. E. Driver, C. C. Mello, Potent and

specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature391, 806–811 (1998).

11. M. Boutros, J. Ahringer, The art and design of genetic screens: RNA interference.Nat. Rev. Genet. 9, 554–566 (2008).

12. L. Alexopoulou, A. C. Holt, R. Medzhitov, R. A. Flavell, Recognition of double-strandedRNA and activation of NF-kB by Toll-like receptor 3. Nature 413, 732–738 (2001).

13. L. Cekaite, G. Furset, E. Hovig, M. Sioud, Gene expression analysis in blood cells inresponse to unmodified and 2′-modified siRNAs reveals TLR-dependent andindependent effects. J. Mol. Biol. 365, 90–108 (2007)

14. A. D. Judge, V. Sood, J. R. Shaw, D. Fang, K. McClintock, I. MacLachlan, Sequence-dependent stimulation of the mammalian innate immune response by synthetic siRNA.Nat. Biotechnol. 23, 457–462 (2005).

15. I. Amit, M. Garber, N. Chevrier, A. P. Leite, Y. Donner, T. Eisenhaure, M. Guttman,J. K. Grenier, W. Li, O. Zuk, L. A. Schubert, B. Birditt, T. Shay, A. Goren, X. Zhang,Z. Smith, R. Deering, R. C. McDonald, M. Cabili, B. E. Bernstein, J. L. Rinn, A. Meissner,D. E. Root, N. Hacohen, A. Regev, Unbiased reconstruction of a mammaliantranscriptional network mediating pathogen responses. Science 326, 257–263(2009).

16. N. Chevrier, P. Mertins, M. N. Artyomov, A. K. Shalek, M. Iannacone, M. F. Ciaccio,I. Gat-Viks, E. Tonti, M. M. DeGrace, K. R. Clauser, M. Garber, T. M. Eisenhaure,N. Yosef, J. Robinson, A. Sutton, M. S. Andersen, D. E. Root, U. von Andrian, R. B. Jones,H. Park, S. A. Carr, A. Regev, I. Amit, N. Hacohen, Systematic discovery of TLR signalingcomponents delineates viral-sensing circuits. Cell 147, 853–867 (2011).

17. O. Takeuchi, S. Akira, Pattern recognition receptors and inflammation. Cell 140, 805–820(2010).

18. K. Hoebe, B. Beutler, Forward genetic analysis of TLR-signaling pathways: An evalua-tion. Adv. Drug Deliv. Rev. 60, 824–829 (2008).

19. Y. Tan, J. C. Kagan, A cross-disciplinary perspective on the innate immune responses tobacterial lipopolysaccharide. Mol. Cell 54, 212–223 (2014).

20. C. Picard, J.-L. Casanova, A. Puel, Infectious diseases in patients with IRAK-4,MyD88, NEMO, or IkBa deficiency. Clin. Microbiol. Rev. 24, 490–497 (2011).

21. K. Takao, T. Miyakawa, Genomic responses in mouse models greatly mimic humaninflammatory diseases. Proc. Natl. Acad. Sci. U.S.A. 112, 1167–1172 (2014)

22. J. Seok, H. S. Warren, A. G. Cuenca, M. N. Mindrinos, H. V. Baker, W. Xu, D. R. Richards,G. P. McDonald-Smith, H. Gao, L. Hennessy, C. C. Finnerty, C. M. López, S. Honari,E. E. Moore, J. P. Minei, J. Cuschieri, P. E. Bankey, J. L. Johnson, J. Sperry, A. B. Nathens,T. R. Billiar, M. A. West, M. G. Jeschke, M. B. Klein, R. L. Gamelli, N. S. Gibran,B. H. Brownstein, C. Miller-Graziano, S. E. Calvano, P. H. Mason, J. P. Cobb, L. G. Rahme,S. F. Lowry, R. V. Maier, L. L. Moldawer, D. N. Herndon, R. W. Davis, W. Xiao,R. G. Tompkins, Genomic responses in mouse models poorly mimic human inflam-matory diseases. Proc. Natl. Acad. Sci. U.S.A. 110, 3507–3512 (2013).

23. N. Li, J. Sun, Z. L. Benet, Z. Wang, S. Al-Khodor, S. P. John, M.-H. Sung, I. D. C. Fraser,Development of a cell system for siRNA screening of pathogen responses in human andmouse macrophages. Sci. Rep. 5, 9559 (2015).

24. S. Marine, A. Bahl, M. Ferrer, E. Buehler, Common seed analysis to identify off-targeteffects in siRNA screens. J. Biomol. Screen. 17, 370–378 (2012).



25. R. König, C.-Y. Chiang, B. P. Tu, S. F. Yan, P. D. DeJesus, A. Romero, T. Bergauer,A. Orth, U. Krueger, Y. Zhou, S. K. Chanda, A probability-based approach for theanalysis of large-scale RNAi screens. Nat. Methods 4, 847–849 (2007).

26. H. Hemmi, T. Kaisho, O. Takeuchi, S. Sato, H. Sanjo, K. Hoshino, T. Horiuchi,H. Tomizawa, K. Takeda, S. Akira, Small anti-viral compounds activate immunecells via the TLR7 MyD88-dependent signaling pathway. Nat. Immunol. 3, 196–200(2002).

27. M. Jurk, F. Heil, J. Vollmer, C. Schetter, A. M. Krieg, H. Wagner, G. Lipford, S. Bauer,Human TLR7 or TLR8 independently confer responsiveness to the antiviral compoundR-848. Nat. Immunol. 3, 499 (2002).

28. P. G. Motshwene, M. C. Moncrieffe, J. G. Grossmann, C. Kao, M. Ayaluru, A. M. Sandercock,C. V. Robinson, E. Latz, N. J. Gay, An oligomeric signaling platform formed by the Toll-like receptor signal transducers MyD88 and IRAK-4. J. Biol. Chem. 284, 25404–25411(2009).

29. K. Burns, S. Janssens, B. Brissoni, N. Olivos, R. Beyaert, J. Tschopp, Inhibition ofinterleukin 1 receptor/Toll-like receptor signaling through the alternatively spliced,short form of MyD88 is due to its failure to recruit IRAK-4. J. Exp. Med. 197, 263–268(2003).

30. H. Ye, J. R. Arron, B. Lamothe, M. Cirilli, T. Kobayashi, N. K. Shevde, D. Segal,O. K. Dzivenu, M. Vologodskaia, M. Yim, K. Du, S. Singh, J. W. Pike, B. G. Darnay,Y. Choi, H. Wu, Distinct molecular mechanism for initiating TRAF6 signalling. Nature418, 443–447 (2002).

31. C. Picard, A. Puel, M. Bonnet, C.-L. Ku, J. Bustamante, K. Yang, C. Soudais, S. Dupuis,J. Feinberg, C. Fieschi, C. Elbim, R. Hitchcock, D. Lammas, G. Davies, A. Al-Ghonaium,H. Al-Rayes, S. Al-Jumaah, S. Al-Hajjar, I. Z. Al-Mohsen, H. H. Frayha, R. Rucker,T. R. Hawn, A. Aderem, H. Tufenkeji, S. Haraguchi, N. K. Day, R. A. Good,M.-A. Gougerot-Pocidalo, A. Ozinsky, J.-L. Casanova, Pyogenic bacterial infectionsin humans with IRAK-4 deficiency. Science 299, 2076–2079 (2003).

32. C.-L. Ku, H. von Bernuth, C. Picard, S.-Y. Zhang, H.-H. Chang, K. Yang, M. Chrabieh,A. C. Issekutz, C. K. Cunningham, J. Gallin, S. M. Holland, C. Roifman, S. Ehl, J. Smart,M. Tang, F. J. Barrat, O. Levy, D. McDonald, N. K. Day-Good, R. Miller, H. Takada,T. Hara, S. Al-Hajjar, A. Al-Ghonaium, D. Speert, D. Sanlaville, X. Li, F. Geissmann,E. Vivier, L. Maródi, B.-Z. Garty, H. Chapel, C. Rodriguez-Gallego, X. Bossuyt, L. Abel,A. Puel, J.-L. Casanova, Selective predisposition to bacterial infections in IRAK-4–deficient children: IRAK-4–dependent TLRs are otherwise redundant in protectiveimmunity. J. Exp. Med. 204, 2407–2422 (2007).

33. J. Qin, Z. Jiang, Y. Qian, J.-L. Casanova, X. Li, IRAK4 kinase activity is redundant forinterleukin-1 (IL-1) receptor-associated kinase phosphorylation and IL-1 responsiveness.J. Biol. Chem. 279, 26748–26753 (2004).

34. E. Lye, C. Mirtsos, N. Suzuki, S. Suzuki, W.-C. Yeh, The role of interleukin 1 receptor-associated kinase-4 (IRAK-4) kinase activity in IRAK-4-mediated signaling. J. Biol.Chem. 279, 40653–40658 (2004).

35. D. De Nardo, T. Nguyen, J. A. Hamilton, G. M. Scholz, Down-regulation of IRAK-4 is acomponent of LPS- and CpG DNA-induced tolerance in macrophages. Cell. Signal.21, 246–252 (2009).

36. J. Fraczek, T. W. Kim, H. Xiao, J. Yao, Q. Wen, Y. Li, J.-L. Casanova, J. Pryjma, X. Li,The kinase activity of IL-1 receptor-associated kinase 4 is required for interleukin-1receptor/Toll-like receptor-induced TAK1-dependent NFkB activation. J. Biol. Chem.283, 31697–31705 (2008).

37. P. Kanakaraj, P. H. Schafer, D. E. Cavender, Y.Wu, K. Ngo, P. F. Grealish, S. A. Wadsworth,P. A. Peterson, J. J. Siekierka, C. A. Harris, W.-P. Fung-Leung, Interleukin (IL)-1 receptor–associated kinase (IRAK) requirement for optimal induction of multiple IL-1 signalingpathways and IL-6 production. J. Exp. Med. 187, 2073–2079 (1998).

38. J. L. Swantek, M. F. Tsen, M. H. Cobb, J. A. Thomas, IL-1 receptor-associatedkinase modulates host responsiveness to endotoxin. J. Immunol. 164, 4301–4306(2000).

39. T. Kawagoe, S. Sato, K. Matsushita, H. Kato, K. Matsui, Y. Kumagai, T. Saitoh, T. Kawai,O. Takeuchi, S. Akira, Sequential control of Toll-like receptor–dependent responses byIRAK1 and IRAK2. Nat. Immunol. 9, 684–691 (2008).

40. Y. Wan, H. Xiao, J. Affolter, T. W. Kim, K. Bulek, S. Chaudhuri, D. Carlson, T. Hamilton,B. Mazumder, G. R. Stark, J. Thomas, X. Li, Interleukin-1 receptor-associated kinase 2 iscritical for lipopolysaccharide-mediated post-transcriptional control. J. Biol. Chem. 284,10367–10375 (2009).

41. N. Suzuki, S. Suzuki, G. S. Duncan, D. G. Millar, T. Wada, C. Mirtsos, H. Takada,A. Wakeham, A. Itie, S. Li, J. M. Penninger, H. Wesche, P. S. Ohashi, T. W. Mak,W.-C. Yeh, Severe impairment of interleukin-1 and Toll-like receptor signalling inmice lacking IRAK-4. Nature 416, 750–756 (2002).

42. M. Yamamoto, S. Sato, H. Hemmi, K. Hoshino, T. Kaisho, H. Sanjo, O. Takeuchi,M. Sugiyama, M. Okabe, K. Takeda, S. Akira, Role of adaptor TRIF in theMyD88-independent Toll-like receptor signaling pathway. Science 301, 640–643(2003).

43. K. Hoebe, X. Du, P. Georgel, E. Janssen, K. Tabeta, S. O. Kim, J. Goode, P. Lin, N. Mann,S. Mudd, K. Crozat, S. Sovath, J. Han, B. Beutler, Identification of Lps2 as a key transducerof MyD88-independent TIR signalling. Nature 424, 743–748 (2003).

ww

44. H. Oshiumi, M. Matsumoto, K. Funami, T. Akazawa, T. Seya, TICAM-1, an adaptormolecule that participates in Toll-like receptor 3–mediated interferon-b induction. Nat.Immunol. 4, 161–167 (2003).

45. E. Pauls, S. K. Nanda, H. Smith, R. Toth, J. S. C. Arthur, P. Cohen, Two phases ofinflammatory mediator production defined by the study of IRAK2 and IRAK1 knock-inmice. J. Immunol. 191, 2717–2730 (2013).

46. S. M. Flannery, S. E. Keating, J. Szymak, A. G. Bowie, Human interleukin-1 receptor-associated kinase-2 is essential for Toll-like receptor-mediated transcriptional andpost-transcriptional regulation of tumor necrosis factor a. J. Biol. Chem. 286,23688–23697 (2011).

47. T.Kawagoe,S.Sato, A. Jung,M.Yamamoto,K.Matsui, H.Kato,S.Uematsu,O. Takeuchi,S. Akira, Essential role of IRAK-4 protein and its kinase activity in Toll-like receptor–mediatedimmune responses but not in TCR signaling. J. Exp. Med. 204, 1013–1024 (2007).

48. E. Y. Chiang, X. Yu, J. L. Grogan, Immune complex-mediated cell activation fromsystemic lupus erythematosus and rheumatoid arthritis patients elaborate differentrequirements for IRAK1/4 kinase activity across human cell types. J. Immunol. 186,1279–1288 (2011).

49. S. Uematsu, S. Sato, M. Yamamoto, T. Hirotani, H. Kato, F. Takeshita, M. Matsuda,C. Coban, K. J. Ishii, T. Kawai, O. Takeuchi, S. Akira, Interleukin-1 receptor-associatedkinase-1 plays an essential role for Toll-like receptor (TLR)7- and TLR9-mediatedinterferon-a induction. J. Exp. Med. 201, 915–923 (2005).

50. K.-M. Lin, W. Hu, T. D. Troutman, M. Jennings, T. Brewer, X. Li, S. Nanda, P. Cohen,J. A. Thomas, C. Pasare, IRAK-1 bypasses priming and directly links TLRs to rapidNLRP3 inflammasome activation. Proc. Natl. Acad. Sci. U.S.A. 111, 775–780 (2014).

51. T. Fernandes-Alnemri, S. Kang, C. Anderson, J. Sagara, K. A. Fitzgerald, E. S. Alnemri,Cutting edge: TLR signaling licenses IRAK1 for rapid activation of the NLRP3 inflamma-some. J. Immunol. 191, 3995–3999 (2013).

52. H. Zhang, J. Pu, X. Wang, L. Shen, G. Zhao, C. Zhuang, R. Liu, IRAK1 rs3027898 C/Apolymorphism is associated with risk of rheumatoid arthritis. Rheumatol. Int. 33, 369–375(2013).

53. J. Arcaroli, E. Silva, J. P. Maloney, Q. He, D. Svetkauskaite, J. R. Murphy, E. Abraham,Variant IRAK-1 haplotype is associated with increased nuclear factor–kB activation andworse outcomes in sepsis. Am. J. Respir. Crit. Care Med. 173, 1335–1341 (2006).

54. K. M. Kaufman, J. Zhao, J. A. Kelly, T. Hughes, A. Adler, E. Sanchez, J. O. Ojwang,C. D. Langefeld, J. T. Ziegler, A. H. Williams, M. E. Comeau, M. C. Marion, S. B. Glenn,R.M. Cantor, J. M.Grossman, B. H. Hahn, Y.W. Song, C.-Y. Yu, J. A. James, J.M. Guthridge,E. E. Brown, G. S. Alarcón, R. P. Kimberly, J. C. Edberg, R. Ramsey-Goldman, M. A. Petri,J. D. Reveille, L. M. Vilá, J.-M. Anaya, S. A. Boackle, A. M. Stevens, B. I. Freedman,L. A. Criswell, B. A. Pons Estel, J.-H. Lee, J.-S. Lee, D.-M. Chang, R. H. A. Scofield,G. S. Gilkeson, J. T. Merrill, T. B. Niewold, T. J. Vyse, S.-C. Bae, M. E. Alarcón-Riquelme,C. O. Jacob, K. Moser Sivils, P. M. Gaffney, J. B. Harley, A. H. Sawalha, B. P. Tsao, Finemapping of Xq28: Both MECP2 and IRAK1 contribute to risk for systemic lupus erythe-matosus in multiple ancestral groups. Ann. Rheum. Dis. 72, 437–444 (2013).

55. C. Deng, C. Radu, A. Diab, M. F. Tsen, R. Hussain, J. S. Cowdery, M. K. Racke,J. A. Thomas, IL-1 receptor-associated kinase 1 regulates susceptibility to organ-specificautoimmunity. J. Immunol. 170, 2833–2842 (2003).

56. C. O. Jacob, J. Zhu, D. L. Armstrong, M. Yan, J. Han, X. J. Zhou, J. A. Thomas, A. Reiff,B. L.Myones, J.O.Ojwang,K.M.Kaufman,M.Klein-Gitelman,D.McCurdy,L.Wagner-Weiner,E.Silverman, J. Ziegler, J. A. Kelly, J. T.Merrill, J. B. Harley, R.Ramsey-Goldman, L.M. Vila,S.-C. Bae, T. J. Vyse, G. S. Gilkeson, P. M. Gaffney, K. L. Moser, C. D. Langefeld,R. Zidovetzki, C. Mohan, Identification of IRAK1 as a risk gene with critical role in thepathogenesis of systemic lupus erythematosus. Proc. Natl. Acad. Sci. U.S.A. 106,6256–6261 (2009).

57. M. Trofimova, A. B. Sprenkle, M. Green, T. W. Sturgill, M. G. Goebl, M. A. Harrington,Developmental and tissue-specific expression of mouse pelle-like protein kinase. J. Biol.Chem. 271, 17609–17612 (1996).

58. Z. Cao, W. J. Henzel, X. Gao, IRAK: A kinase associated with the interleukin-1 receptor.Science 271, 1128–1131 (1996).

59. M.-H. Sung, N. Li, Q. Lao, R. A. Gottschalk, G. L. Hager, I. D. C. Fraser, Switching ofthe relative dominance between feedback mechanisms in lipopolysaccharide-inducedNF-kB signaling. Sci. Signal. 7, ra6 (2014).

60. E. A. Wall, J. R. Zavzavadjian, M. S. Chang, B. Randhawa, X. Zhu, R. C. Hsueh, J. Liu,A. Driver, X. R. Bao, P. C. Sternweis, M. I. Simon, I. D. C. Fraser, Suppression of LPS-induced TNF-a production in macrophages by cAMP is mediated by PKA-AKAP95-p105.Sci. Signal. 2, ra28 (2009).

61. A. Ablasser, J. L. Schmid-Burgk, I. Hemmerling, G. L. Horvath, T. Schmidt, E. Latz,V. Hornung, Cell intrinsic immunity spreads to bystander cells via the intercellulartransfer of cGAMP. Nature 503, 530–534 (2013).

62. J. L. Schmid-Burgk, T. Schmidt, M. M. Gaidt, K. Pelka, E. Latz, T. S. Ebert, V. Hornung,OutKnocker: A web tool for rapid and simple genotyping of designer nuclease editedcell lines. Genome Res. 24, 1719–1723 (2014).

63. S. Al-Khodor, K. Marshall-Batty, V. Nair, L. Ding, D. E. Greenberg, I. D. C. Fraser,Burkholderia cenocepacia J2315 escapes to the cytosol and actively subverts autoph-agy in human macrophages. Cell. Microbiol. 16, 378–395 (2014).



Acknowledgments: We thank R. Germain, R. Gottschalk, and colleagues in the Laboratoryof Systems Biology for helpful discussions and critical reading of the manuscript; S. Hollandfor assistance in obtaining blood from an IRAK4-deficient patient; X. Li for provision of thehuman IRAK4 kinase–deficient mutant; A. Miller for assistance with the bacterial infectionassay; and R. Stahl for assistance in generating the mouse IRAK4-mCitrine retroviralplasmids. Funding: This work was supported by the Intramural Research Program ofthe National Institute of Allergy and Infectious Diseases (NIAID) (to J.S., N.L., K.-S.O., B.D.,S.J.V., B.L., J.D., N.W.L., and I.D.C.F.) and the BONFOR research commission at theUniversity of Bonn (to D.D.N.). V.H. and T.S.E. are supported by grants from the GermanResearch Foundation (SFB704 and SFB670) and the European Research Council (ERC-2009-StG 243046). E.L. is supported by grants from the NIH (R01HL112661), the DeutscheForschungsgemeinschaft (SFB670), and theEuropeanResearchCouncil (ERC InflammAct).E.L. and V.H. are members of the ImmunoSensation cluster of excellence. C.P. is supportedby grants from the NIAID (AI082265) and The Welch Foundation (I-1820). R.B. is supportedby an NIH Integrative Immunology Training program grant (5T32-AI005284-35). Authorcontributions: J.S., N.L., and I.D.C.F. designed the study; J.S. and N.L. performed thesiRNA screen; B.D., J.S., N.L., and I.D.C.F. analyzed the siRNA screen data; J.S. performedall aspects of the IRAK studies; K.-S.O. performed immunoprecipitation of TLR signalingcomponents; S.J.V. performed high-content imaging analysis of TLR pathway components;

ww

B.L. performed qRT-PCR analysis of siRNA-mediated knockdowns; T.S.E. and V.H. gen-erated the THP1 IRAK KO cell lines; D.D.N. and E.L. generated the mouse IRAK4 rescuecell lines; J.D. acquired and processed the IRAK4-deficient patient samples; R.B. and C.P.acquired and processed the IRAK1 and IRAK2 KOBMDMs; N.W.L. and B.D. performed theanalysis of IRAK expression in the GEO data; J.S. and I.D.C.F. wrote the paper; and allauthors commented on the manuscript.Competing interests: The authors declare that theyhave no competing interests. Data and materials availability: Data from the RNAi-basedscreens have been deposited in the PubChem BioAssay database (http://pubchem.ncbi.nlm.nih.gov; AID 1159581 and AID 1159582).

Submitted 27 March 2015Accepted 11 December 2015Final Publication 5 January 201610.1126/scisignal.aab2191Citation: J. Sun, N. Li, K.-S. Oh, B. Dutta, S. J. Vayttaden, B. Lin, T. S. Ebert, D. De Nardo,J. Davis, R. Bagirzadeh, N. W. Lounsbury, C. Pasare, E. Latz, V. Hornung, I. D. C. Fraser,Comprehensive RNAi-based screening of human and mouse TLR pathways identifiesspecies-specific preferences in signaling protein use. Sci. Signal. 9, ra3 (2016).


Documents

Comprehensive RNAi-based screening of human and mouse … › content › sigtrans › 9 › 409 › ra3.full.pdfin which the original identification of involvement in the TLR response