iRhoms 1 and 2 are essential upstream regulators ofADAM17-dependent EGFR signalingXue Lia,b,1, Thorsten Maretzkya,1, Gisela Weskampa, Sébastien Monettec, Xiaoping Qingd, Priya Darshinee A. Issureea,e,Howard C. Crawfordf, David R. McIlwaing,h, Tak W. Maki,2, Jane E. Salmond, and Carl P. Blobela,j,2

aArthritis and Tissue Degeneration Program and dAutoimmunity and Inflammation Program, Hospital for Special Surgery, New York, NY 10021;Departments of bBiochemistry, Cell and Molecular Biology, eImmunology, and jPhysiology, Biophysics and Systems Biology, Weill Cornell Graduate School ofMedical Sciences, New York, NY 10021; cTri-Institutional Laboratory of Comparative Pathology, New York, NY 10021; fDepartment of Cancer Biology, MayoClinic Cancer Center, Jacksonville, FL 32224; gDepartment of Gastroenterology, Hepatology, and Infectious Diseases, Heinrich Heine University, D-40225Dusseldorf, Germany; hBaxter Laboratory in Stem Cell Biology, Department of Microbiology and Immunology, Stanford University, Stanford, CA 94305; andiCampbell Family Institute for Breast Cancer Research, Princess Margaret Cancer Center, University Health Network, Toronto, ON, Canada M5G 2M9

Contributed by Tak W. Mak, March 20, 2015 (sent for review January 30, 2015)

The metalloproteinase ADAM17 (a disintegrin and metallopro-tease 17) controls EGF receptor (EGFR) signaling by liberating EGFRligands from their membrane anchor. Consequently, a patient lack-ing ADAM17 has skin and intestinal barrier defects that are likelycaused by lack of EGFR signaling, and Adam17−/− mice die perina-tally with open eyes, like Egfr−/− mice. A hallmark feature ofADAM17-dependent EGFR ligand shedding is that it can be rapidlyand posttranslationally activated in a manner that requires itstransmembrane domain but not its cytoplasmic domain. This sug-gests that ADAM17 is regulated by other integral membrane pro-teins, although much remains to be learned about the underlyingmechanism. Recently, inactive Rhomboid 2 (iRhom2), which hasseven transmembrane domains, emerged as a molecule that controlsthe maturation and function of ADAM17 in myeloid cells. However,iRhom2−/− mice appear normal, raising questions about howADAM17 is regulated in other tissues. Here we report thatiRhom1/2−/− double knockout mice resemble Adam17−/− andEgfr−/− mice in that they die perinatally with open eyes, mis-shapen heart valves, and growth plate defects. Mechanistically,we show lack of mature ADAM17 and strongly reduced EGFRphosphorylation in iRhom1/2−/− tissues. Finally, we demonstratethat iRhom1 is not essential for mouse development but regulatesADAM17 maturation in the brain, except in microglia, whereADAM17 is controlled by iRhom2. These results provide genetic,cell biological, and biochemical evidence that a principal functionof iRhoms1/2 during mouse development is to regulate ADAM17-dependent EGFR signaling, suggesting that iRhoms1/2 couldemerge as novel targets for treatment of ADAM17/EGFR-dependent pathologies.

inactive Rhomboid proteins | a disintegrin and metalloprotease 17 |epidermal growth factor receptor | heparin-binding epidermal growthfactor | transforming growth factor alpha

ADAM17 (a disintegrin and metalloprotease 17) is a mem-brane-anchored metalloproteinase that controls two major

signaling pathways with important roles in development anddisease, the EGF receptor (EGFR) pathway and the proin-flammatory tumor necrosis factor α (TNF-α) pathway (1–5).Mice lacking ADAM17 resemble mice with defects in EGFRsignaling in that they have open eyes at birth, enlarged semilunarheart valves, and enlarged hypertrophic zones in long bonegrowth plates, most likely caused by a lack of ADAM17-dependent release of the EGFR ligands transforming growthfactor α (TGF-α) and heparin-binding epidermal growth factor(HB-EGF) (3, 6–14). In humans, defects in skin and intestinalbarrier protection have been reported in a patient lackingADAM17 (15) and in patients treated with EGFR inhibitors (16,17), and similar skin defects were recently identified in a pa-tient with defective EGFR signaling (18). Mouse models ofADAM17/EGFR signaling appear to recapitulate these mecha-nisms, because defects in skin barrier protection can be observed

by inactivating either ADAM17 or the EGFR in keratinocytes(19), as well as in mice expressing very low levels of ADAM17,which also have increased susceptibility to intestinal inflammation(20). A hallmark feature of ADAM17 is its rapid response tovarious activators of cellular signaling pathways (21–23), which ispresumably important to allow a rapid response to injury and tomaintain the skin and intestinal barrier. The rapid activation ofADAM17 is controlled by its transmembrane domain whereas thecytoplasmic domain is dispensable in this context (22), suggestingthat ADAM17 is regulated by one or more other membraneproteins, yet the underlying mechanism has remained enigmatic.Recent studies have shown that the maturation and function

of ADAM17 in myeloid cells depend on inactive Rhomboid 2(iRhom2), a catalytically inactive member of the Rhomboidfamily of seven membrane-spanning intramembrane serine pro-teinases (24–28). Myeloid cells lacking iRhom2 release very littleTNF-α in response to activation of Toll-like receptor 4 by li-popolysaccharide (LPS) (24, 26, 28). Therefore, mice lackingiRhom2 are protected from the detrimental effects of TNF-α inmouse models for septic shock and inflammatory arthritis, sim-ilar to conditional knockout mice lacking ADAM17 in myeloidcells (11, 26, 29). However, iRhom2−/− (iR2−/−) mice are viablewith no evident spontaneous pathological phenotypes (26, 29),whereas Adam17−/− (A17−/−) mice die shortly after birth (3). A

Significance

The skin and intestinal barrier are controlled by signaling scis-sors, termed ADAM17 (a disintegrin and metalloprotease 17),that reside in the membrane on the surface of cells. The mainpurpose of these signaling scissors is to liberate growth factorsfrom their membrane anchor, allowing them to activate theirreceptors, including the epidermal growth factor receptor(EGFR). The ADAM17/EGFR signaling axis is tightly regulated,yet little is known about the underlying mechanism. Here weuse genetic, cell biological, and biochemical approaches toidentify two membrane proteins termed iRhoms 1 and 2 (in-active Rhomboid-like proteins) as crucial upstream regulatorsof ADAM17-dependent EGFR signaling. This uncovers theiRhoms as attractive novel targets to treat ADAM17/EGFR-dependent diseases such as cancer.

Author contributions: X.L., T.M., G.W., S.M., X.Q., P.D.A.I., D.R.M., T.W.M., J.E.S., and C.P.B.designed research; X.L., T.M., G.W., S.M., X.Q., and P.D.A.I. performed research; X.L., T.M.,G.W., H.C.C., D.R.M., T.W.M., and C.P.B. contributed new reagents/analytic tools; X.L., T.M.,G.W., S.M., X.Q., P.D.A.I., H.C.C., D.R.M., T.W.M., J.E.S., and C.P.B. analyzed data; and X.L.,T.M., G.W., S.M., X.Q., H.C.C., D.R.M., T.W.M., J.E.S., and C.P.B. wrote the paper.

The authors declare no conflict of interest.1X.L. and T.M. contributed equally to this work.2To whom correspondence may be addressed. Email: tmak@uhnresearch.ca or blobelc@hss.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1505649112/-/DCSupplemental.

6080–6085 | PNAS | May 12, 2015 | vol. 112 | no. 19 www.pnas.org/cgi/doi/10.1073/pnas.1505649112

Dow

nloa

ded

by g

uest

on

May

19,

202

0

major unresolved question has therefore been whether iRhom2and the related iRhom1 are the long-sought-after regulators of thefunction of ADAM17-dependent EGFR signaling in vivo. Herewe generate iRhom1−/− (iR1−/−) mice, which are viable andhealthy, and report that iR1/2−/− double knockout mice closelyresemble mice lacking ADAM17 or the EGFR, providing the firstgenetic evidence, to our knowledge, that the principal function ofiRhoms1/2 during mouse development is to control ADAM17/EGFR signaling.

ResultsGeneration and Characterization of Mice Lacking iRhom1. Mice lack-ing iRhom1 were generated using embryonic stem cells providedby the European Conditional Mouse Mutagenesis Program con-sortium. Animals carrying the targeted allele of iR1 were matedwith mice expressing Cre under the control of the adenovirusEIIa promoter to delete exons 4–11 of iR1 in the germ line (Fig. 1Aand Fig. S1A). Matings of the resulting heterozygous iR1+/− miceproduced iR1−/− offspring at the expected Mendelian ratio (Fig.1B), as assessed by genotyping through genomic PCR (Fig. 1C)and confirmed by Southern blot analysis (Fig. 1D). Character-ization of the iR1 mRNA produced in these animals demon-strated that exons 1/2 could be detected in iR1−/− mouseembryonic fibroblasts (mEFs) by quantitative (q)PCR, whereasexons 4/5, 12/13, and 16/17 were not detectable (Fig. S1 B–E).Thus, any protein fragment, if produced in the mutant mice,would lack all transmembrane domains and most of the N-terminalcytoplasmic domain of iRhom1. iR1−/− mice were indistinguishablefrom their wild-type littermate controls during routine handling,appeared normal, survived up to at least 6 mo of age (Fig. 1 E andF), and had no evident spontaneous pathological phenotypes asadults (see Materials and Methods and SI Materials and Methodsfor details).AWestern blot analysis of ADAM17 in different tissues showed

largely normal levels of mature ADAM17 in the heart, liver, lung,and spleen of iR1−/− mice compared with wild-type controls butslightly lower levels in the kidney and a strong reduction in thebrain (Fig. 2A; quantification is shown in Fig. S2A). A qPCRanalysis of iRhom 1 and 2 transcripts in different tissues of wild-type mice demonstrated that both iRhoms were expressed acrossall tissues examined, with the exception of the brain, whereiRhom1 was easily detected but levels of iRhom2 were very low(Fig. 2 B and C) (biogps.org/#goto=genereport&id=217344) (30).Western blots of different parts of the brain of 8-wk-old iR1−/−

mice showed barely detectable mature ADAM17 in the cortex,cerebellum, brainstem, olfactory bulb, and hippocampus, in sharpcontrast to wild-type mice (Fig. 2D; quantification is shown in Fig.S2B). Because iRhom2 is required for the maturation and functionof ADAM17 in myeloid cells, we isolated CD45high leukocytesfrom the brain to assess the function of ADAM17 by measuringthe phorbol ester (phorbol 12-myristate 13-acetate; PMA)-stimu-lated shedding of CD62L (L-selectin) (31). We found that the PMA-stimulated down-regulation of CD62L was comparable in wild-typeand iR1−/− brain leukocytes (Fig. 2E) but almost completelyabolished in iR2−/− brain leukocytes (Fig. 2F). Similar experimentswith resident microglia [CD45lowCD11bhighF4/80+Ly6Clow cells (32)]showed normal LPS-stimulated release of the ADAM17 sub-strate TNF-α from iR1−/− microglia compared with controls (Fig.2G), whereas TNF-α shedding was almost undetectable fromiR2−/− microglia (Fig. 2H), demonstrating that the functions ofADAM17 in microglia and circulating leukocytes depend oniRhom2 and not iRhom1.

Generation and Characterization of iR1/2−/− Double Knockout Mice.The normal appearance and behavior of iR1−/− mice (this study)or iR2−/− mice (24, 26, 28) and the finding that treatment of iR2−/−

mEFs with iRhom1 siRNA reduced the function of ADAM17 (27,29) led us to examine possible compensatory or redundant rolesof iRhoms 1 and 2 during mouse development by simultaneouslyinactivating both iRhoms. Matings of iR1+/−iR2−/− mice yielded

iR1/2−/− double knockout offspring at the expected Mendelian ratio(Fig. S3A). However, iR1/2−/−mice were born with open eyes (Fig. 3A and B) and suffered perinatal lethality, just like A17−/− mice (3,11). A histopathological analysis of iR1/2−/− mice showed failure ofeyelid closure (Fig. 3C), enlarged aortic, pulmonic, and tricuspidheart valves (but normal mitral valves), and an enlarged zone ofhypertrophic chondrocytes in the femoral and humoral growthplates, closely resembling defects in newborn A17−/− mice (Fig. 3D–G and Fig. S3 B–E) (3, 10, 12). In iR1−/− or iR2−/− singleknockout mice, these and other tissues were indistinguishablefrom wild-type controls (Fig. 3 and Fig. S3 B–E).

Analysis of the Maturation of ADAM17 in Tissues and Cells fromNewborn iR1/2−/− Mice. In a Western blot analysis of several tis-sues, we found no detectable mature ADAM17 in the heart, kid-ney, lung, skin, liver, or brain of newborn iR1/2−/− mice, althoughthe proform was always present (Fig. 4; tissues from newborn wild-type or Adam17−/− mice served as positive and negative controls,respectively). Mature ADAM17 was also present in the heart,kidney, lung, and skin of iR1−/− and iR2−/− mice (Fig. 4 A–D) butstrongly reduced in the liver of iR2−/− mice (Fig. 4E) and the brainof iR1−/− mice (Fig. 4F), consistent with similar findings in adultiR1−/− and iR2−/− mice (Fig. 2A; see also ref. 29; quantification ofFig. 4 is shown in Fig. S4). Taken together, these results demon-strate that either or both iRhoms 1 and 2 regulate the maturationand function of ADAM17 in all tissues examined duringmouse development.

A

null

Exon4 Exon11 Wild-type

LacZ Frt site

targeted Exon4 Exon11

probe

S S

20.7kb

probe

S S

12.2kb

probe

S S

7.3kb

S

iR1F iR1R

LacZ LoxR

iRhom1 LoxR

Lox Psite Neo SSspI

E

0

6

4

Time (month)

2

8

10

Mic

e nu

mbe

r

WT

iR1-/-

0 1 2 3 4 5 6

FWT iR1-/-

polyA

WT

llun

t

+/+ t -/- 449

236

414

C

LacZ/LoxR

iRhom1/LoxR

iR1F/iR1R

D

20 12

7

WT WT/T WT Null SspI

BWT Het null

24.86% (44)

50.85% (90)

24.29% (43)

Fig. 1. Generation and initial characterization of iR1−/− mice. (A) Diagramof the wild-type mouse iRhom1 locus from exon 4 to exon 11 (Top), thetargeted (iR1t/+) allele with loxP sites flanking exon 4 to exon 11 (Middle),and the null allele resulting from recombination by EIIa-Cre (Bottom).(B) Ratio of offspring from iR1+/− x iR1+/− matings. (C) PCR genotyping resultsusing DNA generated from wild-type, iR1t/+, and iR1−/− mice. t, targeted.(D) Southern blot analysis of genomic DNA from mice carrying the targetedallele of iRhom1 (A, Middle) or the null allele (A, Bottom). (E) The survival of10 wild-type and iR1−/− mice was monitored over 6 mo, during which timeone wild-type mouse died whereas all iR1−/− mice survived and appearedindistinguishable from their wild-type controls. (F) Photo of a 6-mo-old adultmale iR1−/− mouse next to an age-matched male wild-type control.

Li et al. PNAS | May 12, 2015 | vol. 112 | no. 19 | 6081

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

May

19,

202

0

EGFR Ligand Shedding from iR1/2−/− mEFs and EGFR Phosphorylationin iR1/2−/− Tissues. Previous experiments had uncovered a role foriRhom2 in regulating the substrate selectivity of stimulatedADAM17-dependent shedding events in mEFs (27). To test howinactivation of iRhom1 or both iRhoms 1 and 2 affects thefunction of ADAM17 in cell-based assays, we performed shed-ding experiments with iR1−/− mEFs, which have marginally re-duced levels of mature ADAM17 compared with wild-typecontrols, or iR1/2−/− mEFs, which have no mature ADAM17(Fig. S5 A and B; see also refs. 27 and 30). Immortalized iR1−/−

mEFs showed comparable constitutive or PMA-stimulated shed-ding of the EGFR ligands epiregulin and HB-EGF and a slightdecrease in TGF-α and amphiregulin shedding compared withwild-type controls (Fig. S6 A–D). However, iR1/2−/− mEFs re-sembled Adam17−/− mEFs (4, 27) in that they showed almost noconstitutive or PMA-stimulated shedding of amphiregulin, epi-regulin, HB-EGF, or TGF-α (Fig. 5 A–D). Moreover, constitutiveand PMA-stimulated shedding of endogenous TGF-α from pri-mary iR1/2−/− keratinocytes was also strongly reduced comparedwith wild-type controls (Fig. 5E). The ionomycin-stimulatedshedding of the ADAM10 substrate betacellulin was normal iniR1/2−/− and iR1−/− mEFs, arguing against a role of iRhoms 1 and2 in ADAM10-dependent ectodomain shedding (Fig. 5F and Fig.S6E). Because a principal function of ADAM17 is to promoteEGFR signaling during development, we determined how thelack of mature ADAM17 affects the phosphorylation of theEGFR in different tissues of iR1−/−, iR2−/−, and iR1/2−/− micecompared with wild-type controls. We found no significant dif-ferences in EGFR phosphorylation in tissues of iR1−/− or iR2−/−

mice compared with controls (Fig. S6F). However, the levels ofphosphorylated EGFR were strongly reduced in all tissues har-vested from viable newborn iR1/2−/− mice compared with wild-type tissues, providing direct evidence for an in vivo defect inEGFR signaling in iR1/2−/− mice (Fig. 5 F and G).

DiscussionThe ADAM17/EGFR signaling pathway can be rapidly activatedwithin minutes in a manner that depends on the transmem-brane domain of ADAM17, which is distal from its catalytic site(22, 23). This raises interesting questions about putative inter-acting regulatory membrane proteins. The discovery of iRhom2as a crucial regulator of ADAM17 in hematopoietic cells (24,26, 29) and in vitro studies in mouse embryonic fibroblastsdemonstrating that iRhom2 controls the substrate selectivity ofADAM17-dependent shedding (27) and that iRhoms 1 and 2are required for the function of ADAM17 in mouse embryonicfibroblasts (27, 30) raised the possibility that iRhom2 and therelated iRhom1 could be the long-sought-after membraneregulators of ADAM17-dependent EGFR signaling. However,this hypothesis had not been previously corroborated by in vivogenetic studies. The close resemblance of iR1/2−/− doubleknockout mice and Adam17−/− mice reported here and the lackof functional ADAM17 as well as lack of EGFR phosphoryla-tion in iR1/2−/− tissues provide genetic, cell biological, andbiochemical evidence that the primary function of iRhoms 1 and2 during development in vivo is to control ADAM17-dependentactivation of the EGFR. Therefore, these studies strongly sup-port a model in which the iRhoms are crucial upstream regu-lators of ADAM17 and the EGFR signaling pathway.We also report that mice lacking iRhom1 appear normal and

have no evident pathological or histopathological defects. In-terestingly, iRhom1 is essential for the maturation of ADAM17in the brain, likely because iRhom2 expression in the brain istoo low to independently support the maturation of ADAM17(biogps.org/#goto=genereport&id=217344) (30). We have pre-viously found that iRhom1/ADAM17-dependent shedding has asignificantly more limited substrate repertoire compared withiRhom2/ADAM17-dependent shedding (27). The unique im-portance of iRhom1 for ADAM17 maturation and function inbrain cells may have evolved because the more limited substrate

repertoire of iRhom1/ADAM17 could have advantages in thecentral nervous system. Very little is currently known about thefunction of ADAM17 in the brain, and a patient who is deficient

A

A9

p A17

KDa

100

75

brain heart lung liver kidney spleen

m

D

A9

pKDa

100

75

A17

cortex CRBL HPC OB BS

m

iRho

m1

mR

NA

(rel

. exp

. lev

els)

iRho

m2

mR

NA

(r

el. e

xp. l

evel

s)

1

100

10

B

1

100

10

C

n.s.

TNF

(pg/

ml)

0

20

LPS:

40

60

n.s.

0

20

40

60

Cel

l sur

face

C

D62

L (%

)

PMA:- + - +

TNF

(pg/

ml)

0 20

LPS:- + - +- + - +

80 100

60 40

Cel

l sur

face

C

D62

L (%

)

0

H

F

G

E

PMA:- + - +

60

20

40

80 leukocytes

* **

microglia microglia

leukocytes

* *

*

Fig. 2. ADAM17 Western blots and qPCR for iRhom1 and iRhom2 expressionin different tissues of iR1−/− mice and controls, and analysis of the role of iR1and iR2 in regulating the function of ADAM17 in microglia. (A) RepresentativeWestern blots of ADAM17 expression in tissues of iR1−/− and control mice.(B and C) Representative qPCR analysis of the expression of iRhom1 (B) or iRhom2(C) in different tissues (representative of three experiments; please note that thescale is logarithmic). (D) Representative Western blots of ADAM17 in extractsfrom different brain areas from iR1−/− mice and controls (BS, brainstem; CRBL,cerebellum; HPC, hippocampus; OB, olfactory bulb). Open arrowheads inA andDindicate pro-ADAM17 (p); black arrowheads indicate mature ADAM17 (m).ADAM9 (A9) Western blots served as loading control in A and D; blots arerepresentative of three separate experiments (quantification is shown in Fig. S2A and B). (E–H) Circulating CD45high leukocytes (E and F) and CD45lowCD11bhigh

F4/80+Ly6Clow resident microglia (G and H) were isolated from brains of iR1−/−

mice or iR2−/− mice or their wild-type control littermates. (E and F) The down-regulation of the substrate CD62L following stimulation with PMAwas normal incirculating CD45high leukocytes from iR1−/− mice compared with controls (E), butwas abolished in circulating CD45high leukocytes from iR2−/−mice compared withcontrols (F). (G and H) The function of ADAM17 in microglia was monitored bymeasuring LPS-stimulated release of TNF-α, which was normal in microglia iso-lated from iR1−/− mice but strongly reduced in microglia from iR2−/− mice com-pared with controls. *P < 0.05; ±SD (n = 3). n.s., not significant.

6082 | www.pnas.org/cgi/doi/10.1073/pnas.1505649112 Li et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

0

in ADAM17 has no reported cognitive deficits (15). However,the targeted deletion of ADAM17 in oligodendrocytes leads todefects in myelination and in exploratory behavior in mice (33),and inactivation of the EGFR results in progressive neurodegen-eration (34). Because iR1−/− mice are indistinguishable from theirwild-type littermates during routine handling and do not displayevident histopathological brain defects, future studies will benecessary to learn more about the function of ADAM17 andiRhom1 in the brain and whether iR1−/− mice could have subtledefects in the brain or cognitive or behavioral abnormalities thatwere not evident in the analysis performed here.A recent study identified a strong association between Alz-

heimer’s disease and changes in methylation of iRhom2 in pa-tient brains (35). Presumably, the change in methylation ofiRhom2 (RHBDF2) leads to an increased expression of iRhom2in microglia or brain leukocytes or both cell types, which couldenhance their capacity to trigger proinflammatory injury in thebrain. Our results suggest that the function of ADAM17 in residentmicroglia and circulating leukocytes infiltrating the brain dependson iRhom2, just as in peripheral myeloid cells (29). DecreasingiRhom2 activity in the brain would selectively inhibit ADAM17in microglia and infiltrating leukocytes, likely without alteringthe function of ADAM17 in other cell types in the brain. Thera-peutically targeting iRhom2 in the brain could thus be beneficialfor prevention or treatment of Alzheimer’s disease, particularly inpatients with increased expression of iRhom2 (35).

The phenotype of the iR1−/− and iR1/2−/− mice described hereis different from the more severe phenotype of iR1−/− miceand the early embryonic lethality of iR1/2−/− mice reported byChristova et al. (30). This could be caused by the larger deletiongenerated by Christova et al. (exons 2–18 versus exons 4–11 inour iR1−/− mice), which could potentially affect regulatory ele-ments for other genes or the presence of unidentified proximalmutations segregating with the targeted allele. We cannot ruleout a contribution of subtle differences in genetic background, orthat the first two exons in our iR1−/− mice (encoding amino acidresidues 1–82 of iRhom1) could have functions that are lackingin the iR1−/− mice generated by Christova et al. (see SI Materialsand Methods for details). However, because all transmembranedomains of iRhom1 are absent in the iR1−/− mice used here andwe find striking similarity between iR1/2−/− mice and ADAM17−/−

mice, the difference in phenotypes is most likely not related toadditional iRhom1/2-regulated client membrane proteins that in-teract with these transmembrane domains (30).The identification of iRhoms 1 and 2 as key regulators of

ADAM17-dependent EGFR signaling in mice highlights an im-portant difference compared with EGFR signaling in Drosophilamelanogaster, which depends on active Rhomboid proteinases tocatalyze EGFR ligand release (36, 37). Because other compo-nents of the EGFR signaling pathway are largely conserved be-tween Drosophila and mammals (38), it is interesting to considerthe potential evolutionary basis for why the release of EGFR

cm

iR2-/- iR1-/- iR1/2-/- Adam17-/- WT A

ortic

val

ve

Gro

wth

pla

te

Eye

lid

G

0

400

300

200

*

100

*

A

F

0

80

120

60

x th

ickn

ess

(μm

)

40

Aortic valve

20

100

Growth plate

B

C

D

E

* *

x th

ickn

ess

(μm

)

Fig. 3. iR1/2−/− double knockout mice closely re-semble Adam17−/− mice. (A) A comparison of new-born (P1) wild-type, iR1/2/−/−, Adam17−/−, iR1−/−, oriR2−/− mice, which appeared similar except for theopen eyes at birth (OEB) in iR1/2−/− and Adam17−/−

mice (arrows). (B and C) The OEB phenotype (arrows)is shown on macroscopic images of the head (B) andhistological sections (C). (D) Aortic valve sections(arrows). (E) Sections through the femoral growthplate (the thickness of the hypertrophic zone is in-dicated by yellow lines). In all cases, the defects ob-served in iR1/2−/− mice (OEB, enlarged aortic valve,thickened zone of hypertrophic chondrocytes) closelyresembled those in Adam17−/− mice, whereas iR1−/−

and iR2−/− mice appeared normal. (F and G) Quanti-fication of the average width relative to the lengthof individual aortic valve leaflets (F) or of the lengthof the growth plates (G) for each genotype (seeMaterials and Methods for details). All sections arerepresentative of three mice examined. [Scale bars,200 μm (C and E) and 100 μm (D).] x, average. *P <0.05; ±SD (n = 3).

Li et al. PNAS | May 12, 2015 | vol. 112 | no. 19 | 6083

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

May

19,

202

0

ligands, which is crucial for activating the EGFR (reviewed inref. 21), is accomplished by active Rhomboids in flies, whereas itrequires one or both inactive Rhomboids and ADAM17 inmammals. The principal functions of ADAM17 in vivo appear tobe protection of the skin and intestinal barrier through activatingthe EGFR (15, 19, 20). A unique property of ADAM17 (notobserved for Drosophila Rhomboids) is its ability to be rapidlyand posttranslationally activated within minutes by many sig-naling pathways (22, 23, 27, 39, 40). Therefore, it is tempting tospeculate that this dedicated system, consisting of iRhoms 1 and2 and ADAM17, evolved because of the advantage provided by arapid EGFR activation to protect the skin and intestinal barrier(17) as well as release of TNF-α to activate innate immune re-sponses, two major functions of ADAM17 (3, 11, 15, 19, 20).Taken together with previous studies demonstrating that muta-tions in the cytoplasmic domain of iRhom2 result in activation ofADAM17, leading to esophageal cancer with palmar-plantarkeratosis (41), and our previous observation that iRhom2 controlsthe substrate selectivity of ADAM17-dependent shedding (27),our findings support a model in which the principal purpose ofiRhoms 1 and 2 is to control the maturation and function ofADAM17, which specifically and rapidly engages the EGFR- andTNF-α–dependent signaling pathways. The interaction betweeniRhoms and ADAM17 thus provides novel therapeutic opportu-nities for selective and simultaneous inactivation of two majorsignaling pathways with important roles in development and dis-ease, the EGFR signaling pathway, a well-established target fortreatment of cancer, and the TNF-α pathway, which is a target fortreatment of autoimmune diseases such as rheumatoid arthritis.

Materials and MethodsAdditional details are provided in SI Materials and Methods.

Materials. All materials were from Sigma-Aldrich unless noted otherwise.Restriction enzymes were from New England BioLabs, and dNTPs for qPCRwere from Qiagen. The rabbit antibodies against phospho-EGFR werefrom Cell Signaling Technology and the secondary anti-rabbit HRP-labeled antibodies were from Promega.

Genetically Modified Mice. Adam17−/−mice (A17−/−) and iRhom2−/−mice (iR2−/−)have been described previously (11, 26) and were of mixed genetic background(129Sv,C57BL/6). iRhom1−/− mice (iR1−/−) were generated from a conditionallytargeted C57BL/6N ES cell line obtained from the European Conditional MouseMutagenesis Program (EPD0577_2_H04). iR1/2−/− double knockout mice were ofmixed genetic background (129Sv,C57BL/6N,FVB). All animal experiments wereapproved by the Internal Animal Use and Care Committee of the Hospital forSpecial Surgery.

Southern Blot Analysis. Genomic DNA isolated from mouse tails was purifiedand subsequently digested overnight with SspI (New England BioLabs),separated on a 1% agarose gel, and transferred to a Biodyne B membrane(Pall). Immobilized DNA was detected with a 32P-labeled DNA probe (seeFig. 1A for localization of the probe and the expected band size for thewild-type, targeted, and null alleles). Hybridization was performed for 14 hat 65 °C, and then the membranes were washed in 0.1× SSC/0.1% SDS andsubsequently exposed to HyBlot CL film (Denville).

Western Blot Analysis. For Western blots of the pro- and mature forms ofADAM17 in different tissues of adult iRhom1−/− mice and wild-type controls,the heart, liver, lung, spleen, kidney, brain, or different areas of the brainwere isolated and processed as previously described (29), and the same ap-proach was used to prepare samples for Western blots of different tissues innewborn mice [postnatal day (P)1].

Quantitative PCR Analysis. Total RNAwas extracted from heart, liver, lung, spleen,kidney, andbrainusing theRNeasyMiniKit (Qiagen). RNAwas reverse-transcribedbyM-MuLV Reverse Transcriptase (New England BioLabs). qPCR (SYBR Green;ABI PRISM 7900HT; Applied Biosystems) was normalized to GAPDH.

Isolation of Microglia for TNF-α ELISA and Circulating Brain Leukocytes forCD62L FACS Analysis. Mice were euthanized and perfused with PBS beforeisolation of whole brains. Brains were minced and digested with 1 mg/mLcollagenase D (Roche Life Science) at 37 °C for 30 min. After passage througha cell strainer, cells were centrifuged in 40% (vol/vol) Percoll at 1,000 × g atroom temperature for 15 min with low acceleration and no brake.Enriched microglia were collected at the bottom of the tube. Cells were

Fbrain

KDa

100 75

Bkidney

KDa

100

75

Clung

KDa

100

75

Dskin

KDa

100

75

Eliver

KDa

100

75

Aheart

KDa

100

75

A9

A17 m

p A17

A9

pm

A17

A9

pm

A9

A17 m

p

A9

A17 m

p

A9

A17 m

p

Fig. 4. Western blots of ADAM17 in different tissues of newborn mice.Different tissues were harvested from newborn (P1) mice of the indicatedgenotypes [heart (A); kidney (B); lung (C); skin (D); liver (E); brain (F)] andsubjected to Western blot analysis to determine the levels of pro- and ma-ture ADAM17 in each sample (the open arrowheads point to pro-ADAM17;the closed arrowheads point to mature ADAM17), with ADAM9 serving asloading control. All blots are representative examples of at least three sep-arate experiments (quantification is shown in Fig. S4).

A B C

D E F

G

H

mEFs/AREG mEFs/Epiregulin

Fig. 5. ADAM17-dependent EGFR ligand shedding is abolished in iR1/2−/−

mEFs, and EGFR phosphorylation is strongly reduced in iR1/2−/− tissues. (A–F)The constitutive and PMA-stimulated shedding of amphiregulin (AREG) (A),epiregulin (B), HB-EGF (C), and TGF-α (D) is almost completely abolished in iR1/2−/−

mEFs compared with wild-type controls, as is the constitutive and PMA-stimulatedshedding of endogenous TGF-α from primary iR1/2−/− keratinocytes (E), whereasthe constitutive and ionomycin (IO)-stimulated shedding of the ADAM10 sub-strate betacellulin (BTC) from iR1/2−/− mEFs is normal (F). (G and H) The levels ofphosphorylated EGFR are strongly reduced in extracts of heart, liver, lung, kidney,skin, and brain of newborn iR1/2−/− mice compared with wild-type controls(quantification is shown in H). *P < 0.05; ±SD (n = 3).

6084 | www.pnas.org/cgi/doi/10.1073/pnas.1505649112 Li et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

0

stained with FITC-conjugated anti-mouse CD45 (clone 30-F11; BioLegend),PE-conjugated anti-mouse CD11b (clone M1/70; BioLegend), APC-conjugatedanti-mouse F4/80 (clone BM8; BioLegend), and PerCp/Cy5.5-conjugated anti-mouse Ly6C Abs (clone HK1.4; BioLegend) for 30 min on ice. CD45lowCD11bhigh

F4/80+Ly6Clow microglia (32) were sorted on a FACSVantage cell sorter (BectonDickinson) with >95% purity. Microglia (1 × 105) were cultured in the presenceor absence of LPS (1 μg/mL) in a 96-well U-bottom plate at 37 °C for 2 h. Se-creted TNF-α levels in the culture supernatant were measured with a TNF-αELISA Kit (eBiocience). For analyzing CD62L shedding on circulating leukocytesin the brain, leukocytes enriched from whole mouse brain were left untreatedor were treated with 25 ng/mL PMA for 20 min at 37 °C. Cell-surface expressionof CD62L on CD45high circulating leukocytes was analyzed by FACS on anupgraded 11-color FACSCalibur machine (Becton Dickinson).

Histopathological Analysis. Following euthanasia of 13.5-wk-old iR1−/− andwild-type mice (two of each genotype) by carbon dioxide (according to theguidelines of the American Veterinary Medical Association), all organs wereexamined grossly and fixed in 10% neutral buffered formalin or 4% para-formaldehyde. Tissues were processed routinely for histology and embedded inparaffin, sectioned, stained with hematoxylin and eosin, and examined.

Generation of Mouse Embryonic Fibroblasts. Embryonic fibroblasts were iso-lated from embryonic day (E)13.5 iR1−/− and iR1/2−/− embryos to generateprimary mEFs for immortalization as previously described (4).

Transfection and Ectodomain Shedding Assays. iR1−/−, iR1/2−/−, and wild-typecontrol mEFs were grown to 80% confluency and then transfected with

plasmids encoding alkaline phosphatase-tagged EGF receptor ligands (4)using Lipofectamine 2000 (Invitrogen). One day after transfection, cells werewashed in Opti-MEM (Gibco) for 1 h. Then, fresh Opti-MEM with or without25 ng/mL phorbol 12-myristate 13-acetate or 2.5 μM ionomycin was added tothe cells for 45 min to stimulate shedding, as indicated. After incubatingwith the alkaline phosphatase substrate 4-nitrophenyl phosphate, the alkalinephosphatase activity in the supernatant and lysate was measured at A405.Three identical wells were prepared, and the ratio of alkaline phosphataseactivity in the supernatant and that of the cell lysate plus supernatant wascalculated. Each experiment was conducted at least three times.

Statistical Analysis. All values are expressed as means ± SD (with data fromat least three independent experiments). The assumptions for normality(Kolmogorov–Smirnov test) and equal variance (Levene median test) wereverified with SigmaStat 3.1 software (SYSTAT). The analysis of variance wasperformed with one-way analysis of variance. Statistics following a Student’st distribution were generated using a two-tailed Student’s t test. Multipleparametric statistical comparisons between experimental groups versus acontrol group were accomplished by Dunnett’s method. P values of <0.05were considered statistically significant.

ACKNOWLEDGMENTS. We thank Elin Mogollon for technical assistance, andAlex Wolujuczyk for preparing the sections for histological analysis. Thiswork was funded by NIH GM64750 (to C.P.B.), NIH CA159222 (to H.C.C.),Heinrich Heine University Grant 9772555 and Canadian Institutes of HealthResearch (CIHR) fellowship 201210MFE-289576-150035 (to D.R.M.), CIHR-MOP123276 (to T.W.M.), and NIH CA008748 (to S.M.).

1. Black RA, et al. (1997) A metalloproteinase disintegrin that releases tumour-necrosisfactor-α from cells. Nature 385(6618):729–733.

2. Moss ML, et al. (1997) Cloning of a disintegrin metalloproteinase that processesprecursor tumour-necrosis factor-alpha. Nature 385(6618):733–736.

3. Peschon JJ, et al. (1998) An essential role for ectodomain shedding in mammaliandevelopment. Science 282(5392):1281–1284.

4. Sahin U, et al. (2004) Distinct roles for ADAM10 and ADAM17 in ectodomain sheddingof six EGFR ligands. J Cell Biol 164(5):769–779.

5. Sunnarborg SW, et al. (2002) Tumor necrosis factor-alpha converting enzyme (TACE)regulates epidermal growth factor receptor ligand availability. J Biol Chem 277(15):12838–12845.

6. Miettinen PJ, et al. (1995) Epithelial immaturity and multiorgan failure in mice lackingepidermal growth factor receptor. Nature 376(6538):337–341.

7. Threadgill DW, et al. (1995) Targeted disruption of mouse EGF receptor: Effect ofgenetic background on mutant phenotype. Science 269(5221):230–234.

8. Sibilia M, Wagner EF (1995) Strain-dependent epithelial defects in mice lacking theEGF receptor. Science 269(5221):234–238.

9. Sibilia M, et al. (2003) Mice humanised for the EGF receptor display hypomorphicphenotypes in skin, bone and heart. Development 130(19):4515–4525.

10. Hall KC, et al. (2013) ADAM17 controls endochondral ossification by regulatingterminal differentiation of chondrocytes. Mol Cell Biol 33(16):3077–3090.

11. Horiuchi K, et al. (2007) Cutting edge: TNF-alpha-converting enzyme (TACE/ADAM17)inactivation in mouse myeloid cells prevents lethality from endotoxin shock.J Immunol 179(5):2686–2689.

12. Jackson LF, et al. (2003) Defective valvulogenesis in HB-EGF and TACE-null mice isassociated with aberrant BMP signaling. EMBO J 22(11):2704–2716.

13. Mine N, Iwamoto R, Mekada E (2005) HB-EGF promotes epithelial cell migration ineyelid development. Development 132(19):4317–4326.

14. Iwamoto R, et al. (2003) Heparin-binding EGF-like growth factor and ErbB signaling isessential for heart function. Proc Natl Acad Sci USA 100(6):3221–3226.

15. Blaydon DC, et al. (2011) Inflammatory skin and bowel disease linked to ADAM17deletion. N Engl J Med 365(16):1502–1508.

16. Lacouture ME (2006) Mechanisms of cutaneous toxicities to EGFR inhibitors. Nat RevCancer 6(10):803–812.

17. Lichtenberger BM, et al. (2013) Epidermal EGFR controls cutaneous host defense andprevents inflammation. Sci Transl Med 5(199)199ra111.

18. Campbell P, et al. (2014) Epithelial inflammation resulting from an inheritedloss-of-function mutation in EGFR. J Invest Dermatol 134(10):2570–2578.

19. Franzke CW, et al. (2012) Epidermal ADAM17 maintains the skin barrier by regulatingEGFR ligand-dependent terminal keratinocyte differentiation. J ExpMed 209(6):1105–1119.

20. Chalaris A, et al. (2010) Critical role of the disintegrin metalloprotease ADAM17 forintestinal inflammation and regeneration in mice. J Exp Med 207(8):1617–1624.

21. Blobel CP (2005) ADAMs: Key components in EGFR signalling and development. NatRev Mol Cell Biol 6(1):32–43.

22. Le Gall SM, et al. (2010) ADAM17 is regulated by a rapid and reversible mechanism

that controls access to its catalytic site. J Cell Sci 123(Pt 22):3913–3922.23. Maretzky T, et al. (2011) Migration of growth factor-stimulated epithelial and en-

dothelial cells depends on EGFR transactivation by ADAM17. Nat Commun 2:229.24. Adrain C, Zettl M, Christova Y, Taylor N, Freeman M (2012) Tumor necrosis factor

signaling requires iRhom2 to promote trafficking and activation of TACE. Science

335(6065):225–228.25. Lichtenthaler SF (2012) Sheddase gets guidance. Science 335(6065):179–180.26. McIlwain DR, et al. (2012) iRhom2 regulation of TACE controls TNF-mediated pro-

tection against Listeria and responses to LPS. Science 335(6065):229–232.27. Maretzky T, et al. (2013) iRhom2 controls the substrate selectivity of stimulated

ADAM17-dependent ectodomain shedding. Proc Natl Acad Sci USA 110(28):11433–11438.28. Siggs OM, et al. (2012) iRhom2 is required for the secretion of mouse TNFα. Blood

119(24):5769–5771.29. Issuree PD, et al. (2013) iRHOM2 is a critical pathogenic mediator of inflammatory

arthritis. J Clin Invest 123(2):928–932.30. Christova Y, Adrain C, Bambrough P, Ibrahim A, Freeman M (2013) Mammalian

iRhoms have distinct physiological functions including an essential role in TACE reg-

ulation. EMBO Rep 14(10):884–890.31. Li Y, Brazzell J, Herrera A, Walcheck B (2006) ADAM17 deficiency by mature neu-

trophils has differential effects on L-selectin shedding. Blood 108(7):2275–2279.32. Hashimoto D, et al. (2013) Tissue-resident macrophages self-maintain locally through-

out adult life with minimal contribution from circulating monocytes. Immunity 38(4):

792–804.33. Palazuelos J, et al. (2014) TACE/ADAM17 is essential for oligodendrocyte de-

velopment and CNS myelination. J Neurosci 34(36):11884–11896.34. SibiliaM, Steinbach JP, Stingl L, Aguzzi A,Wagner EF (1998) A strain-independent postnatal

neurodegeneration in mice lacking the EGF receptor. EMBO J 17(3):719–731.35. De Jager PL, et al. (2014) Alzheimer’s disease: Early alterations in brain DNA meth-

ylation at ANK1, BIN1, RHBDF2 and other loci. Nat Neurosci 17(9):1156–1163.36. Urban S, Lee JR, Freeman M (2001) Drosophila Rhomboid-1 defines a family of pu-

tative intramembrane serine proteases. Cell 107(2):173–182.37. Urban S, Lee JR, Freeman M (2002) A family of Rhomboid intramembrane proteases

activates all Drosophila membrane-tethered EGF ligands. EMBO J 21(16):4277–4286.38. Blobel CP, Carpenter G, Freeman M (2009) The role of protease activity in ErbB bio-

logy. Exp Cell Res 315(4):671–682.39. Prenzel N, et al. (1999) EGF receptor transactivation by G-protein-coupled receptors

requires metalloproteinase cleavage of proHB-EGF. Nature 402(6764):884–888.40. Fischer OM, Hart S, Gschwind A, Ullrich A (2003) EGFR signal transactivation in cancer

cells. Biochem Soc Trans 31(Pt 6):1203–1208.41. Blaydon DC, et al. (2012) RHBDF2 mutations are associated with tylosis, a familial

esophageal cancer syndrome. Am J Hum Genet 90(2):340–346.

Li et al. PNAS | May 12, 2015 | vol. 112 | no. 19 | 6085

DEV

ELOPM

ENTA

LBIOLO

GY

Dow

nloa

ded

by g

uest

on

May

19,

202

0