Familial Mediterranean fever mutations lift theobligatory requirement for microtubules in Pyrininflammasome activationHanne Van Gorpa,b,1, Pedro H. V. Saavedraa,b,1, Nathalia M. de Vasconcelosa,b, Nina Van Opdenboscha,b,Lieselotte Vande Wallea,b, Magdalena Matusiaka,b, Giusi Prencipec, Antonella Insalacoc, Filip Van Hauwermeirena,b,Dieter Demona,b, Delfien J. Bogaertd,e,f, Melissa Dullaersd,g, Elfride De Baereh, Tino Hochepieda,i, Joke Dehoornej,Karim Y. Vermaelenb,k, Filomeen Haerynckd,e,f, Fabrizio De Benedettic, and Mohamed Lamkanfia,b,2

aInflammation Research Center, VIB, Zwijnaarde, B-9052, Belgium; bDepartment of Internal Medicine, Ghent University, Ghent, B-9000, Belgium;cRheumatology Unit, Bambino Gesù Children’s Hospital, Rome, I-00146, Italy; dClinical Immunology Research Laboratory, Centre for PrimaryImmunodeficiency Ghent, Ghent University Hospital, Ghent, B-9000, Belgium; eDepartment of Pediatric Immunology and Pulmonology, Centre for PrimaryImmunodeficiency Ghent, Ghent University Hospital, Ghent, B-9000, Belgium; fJeffrey Modell Diagnosis and Research Centre, Ghent University Hospital,Ghent, B-9000, Belgium; gLaboratory of Immunoregulation, Inflammation Research Center, VIB, Zwijnaarde, B-9052, Belgium; hCenter for Medical GeneticsGhent, Ghent University, Ghent, B-9000, Belgium; iDepartment of Biomedical Molecular Biology, Ghent University, Ghent, B-9000, Belgium; jDepartment ofPediatric Rheumatology, Ghent University Hospital, Ghent, B-9000, Belgium; and kTumor Immunology Laboratory, Department of Pulmonary Medicine,Ghent University Hospital, Ghent, B-9000, Belgium

Edited by Vishva M. Dixit, Genentech, San Francisco, CA, and approved October 28, 2016 (received for review August 8, 2016)

Familial Mediterranean fever (FMF) is the most common monogenicautoinflammatory disease worldwide. It is caused bymutations in theinflammasome adaptor Pyrin, but how FMF mutations alter signalingin FMF patients is unknown. Herein, we establish Clostridium difficileand its enterotoxin A (TcdA) as Pyrin-activating agents and show thatwild-type and FMF Pyrin are differentially controlled by microtubules.Diverse microtubule assembly inhibitors prevented Pyrin-mediatedcaspase-1 activation and secretion of IL-1β and IL-18 from mousemacrophages and human peripheral blood mononuclear cells(PBMCs). Remarkably, Pyrin inflammasome activation persisted uponmicrotubule disassembly in PBMCs of FMF patients but not in cells ofpatients afflicted with other autoinflammatory diseases. We furtherdemonstrate that microtubules control Pyrin activation downstreamof Pyrin dephosphorylation and that FMF mutations enable microtu-bule-independent assembly of apoptosis-associated speck-like proteincontaining a caspase recruitment domain (ASC) micrometer-sizedperinuclear structures (specks). The discovery that Pyrin mutationsremove the obligatory requirement for microtubules in inflamma-some activation provides a conceptual framework for understandingFMF and enables immunological screening of FMF mutations.

FMF | Pyrin | inflammasome | colchicine | microtubules

Inflammasomes are multiprotein complexes that culminate inprocessing of caspase-1, thereby promoting maturation of proIL-

1β and proIL-18 into their active forms (1). Several inflammasomeplatforms have been described, and the concerted actions ofinflammasomes frequently are of utmost importance for effectiveprotection of the host against harmful environmental agents andinfections (1). Conversely, mutations in genes coding for inflam-masome components and regulators cause debilitating systemicautoinflammatory diseases, of which cryopyrin-associated periodicsyndromes (CAPS; NLRP3 mutations), autoinflammation withinfantile enterocolitis (AIFEC; NLRC4 mutations), hyperimmu-noglobulinemia syndrome (HIDS; MVK mutations) and familialMediterranean fever (FMF; MEFV mutations) are notable ex-amples (2, 3).FMF is the most common monogenic autoinflammatory dis-

ease worldwide, affecting an estimated 150,000 patients (4). Ittypically has an autosomal recessive inheritance, and the clinicalpresentation is characterized by periodic fevers with childhoodonset, frequently accompanied by serositis and joint pain (3, 4).The disease is highly prevalent in populations of the EasternMediterranean basin and the Middle East and has spread to therest of the world with the extensive migrations of these pop-ulations (4, 5). More than 80% of FMF patients are homo- or

(compound) heterozygous for mutations in MEFV, the gene thatcodes for the inflammasome adaptor Pyrin (4, 5). More than 310disease-associated variants in MEFV have been reported to datein the InFevers registry (6), with most residing in the C-terminalB30.2 (PRY/SPRY) domain of human Pyrin. Importantly, how-ever, how FMF mutations regulate Pyrin signaling has remainedenigmatic, and mouse studies of FMF are complicated by theabsence of the B30.2 domain in murine Pyrin.FMF alleles occur in as many as one of every four individuals

of non-Ashkenazi Jew, Arab, Armenian, and Turkish descent (7–10). In addition, a subset of FMF patients is heterozygous fordisease-associated MEFV alleles, and the clinical/functional rel-evance of some MEFV alleles is debated. Consequently, geneticanalysis of FMF is sometimes inconclusive, and FMF diagnosismay be delayed for years (11). Although FMF is a systemic im-munological disease, immunological diagnosis of the disease iscurrently not available and is likely to require further insight intohow FMF mutations modulate Pyrin activation. The work presented

Significance

Familial Mediterranean fever (FMF) is an autoinflammatory dis-ease caused bymore than 310mutations in the geneMEFV, whichencodes Pyrin. Pyrin recently was shown to trigger inflammasomeactivation in response to Rho GTPase-modifying bacterial toxins.Here we report that Clostridium difficile infection and intoxicationwith its enterotoxin TcdA engage the Pyrin inflammasome. More-over, activation of the Pyrin inflammasome, but not other inflam-masomes, was hampered by microtubule-depolymerizing drugs inmouse and humans. Unexpectedly, we found that FMF mutationsrender Pyrin activation independent of microtubules. Thus, ourfindings provide a conceptual framework for understanding Pyrinsignaling and enable functional diagnosis of FMF.

Author contributions: H.V.G., P.H.V.S., and M.L. designed research; H.V.G., P.H.V.S., N.M.d.V.,N.V.O., L.V.W., M.M., F.V.H., and D.D. performed research; G.P., A.I., D.J.B., M.D., E.D.B., T.H.,J.D., K.Y.V., F.H., and F.D.B. contributed new reagents/analytic tools; H.V.G., P.H.V.S., N.M.d.V.,N.V.O., L.V.W., M.M., F.V.H., D.D., and M.L. analyzed data; H.V.G., P.H.V.S., and M.L. wrotethe paper; and M.L. coordinated the project.

Conflict of interest statement: H.V.G., P.H.V.S., and M.L. are listed as inventor on a patentapplication on immunological FMF diagnosis.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1H.V.G. and P.H.V.S. contributed equally to this work.2To whom correspondence should be addressed. Email: mohamed.lamkanfi@irc.vib-ugent.be.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1613156113/-/DCSupplemental.

14384–14389 | PNAS | December 13, 2016 | vol. 113 | no. 50 www.pnas.org/cgi/doi/10.1073/pnas.1613156113

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

here expands the set of Pyrin inflammasome-activating agents tolive C. difficile infection and its enterotoxin TcdA. We furthershow that among the different known inflammasome sensors,wild-type Pyrin of both humans and mice relies selectively onmicrotubules for inflammasome activation. Microtubules controlPyrin signaling downstream of Pyrin dephosphorylation. Sur-prisingly, however, we found that FMFmutations lift the obligatoryrequirement for microtubules in activating the Pyrin inflamma-some, providing a conceptual framework for understanding FMFand enabling immunological segregation of FMF from relatedautoinflammatory disorders.

ResultsTcdA Activates the Pyrin Inflammasome in Mouse Macrophages andHuman Monocytes. The Pyrin inflammasome responds to infectionwith Burkholderia cenocepacia (12, 13) and the Rho GTPase-targeting toxins Clostridium botulinum toxin C3 and C. difficilecytotoxin B (TcdB) in mouse macrophages (13–15). Importantly,the B30.2 domain is absent in the murine Pyrin ortholog, whichprompted us to characterize regulation mechanisms of Pyrin acti-vation in peripheral blood mononuclear cells (PBMCs) of healthydonors in parallel with studies in murine bone marrow-derivedmacrophages (BMDMs). As reported (14, 15), TcdB activated thePyrin inflammasome in our studies (Fig. S1 A and B). C. difficileproduces a second enterotoxin, C. difficile toxin A (TcdA), whichexerts markedly different functions in disease models (16). In ex-amining the inflammasome response to TcdA, we found thatapoptosis-associated speck-like protein containing a caspase recruit-ment domain (ASC), but not caspase-11, was required for caspase-1maturation in LPS-primed BMDMs that had been incubated withTcdA (Fig. S1C). Also, proteolytic conversion of proIL-1β into themature cytokine required ASC and caspase-1 but not caspase-11(Fig. S1 C and D). Consistent with TcdA activating a canonicalinflammasome, extracellular release of IL-1β continued unabatedin BMDMs lacking caspase-11 but was blunted in Asc−/− macro-phages, in caspase-1−/−caspase-11−/− BMDMs, and in macrophagesof a newly generated caspase-1–deficient mouse strain (Fig. S1Dand Fig. S2). TcdA paralleled TcdB in engaging the Pyrin inflam-masome because TcdA-induced caspase-1 activation and subse-quent cleavage of proIL-1β were abolished in Pyrin (Mefv−/−)BMDMs but not in Nlrp3−/− macrophages (Fig. S1E). Consistentwith these findings, TcdA-intoxicated macrophages required Pyrin,but not Nlrp3, for secreting IL-1β (Fig. S1F). These results dem-onstrate that TcdA selectively activates the Pyrin inflammasome inmurine BMDMs. To examine whether this selective activation alsooccurs in humans, PBMCs from three healthy donors were incu-bated with the cell-permeable caspase-1 inhibitor Ac-YVAD-cmkor the Nlrp3 inflammasome-selective inhibitor MCC950/CRID3(17, 18) before intoxication with TcdA. Pharmacological inhibi-tion of caspase-1 significantly reduced TcdA-induced cleavage ofproIL-1β and the extracellular release of IL-1β and IL-18 fromhuman PBMCs (Fig. S1 G–I). In contrast, TcdA-induced inflam-masome activation was insensitive to MCC950/CRID3 at drugconcentrations that effectively blunted Nlrp3-relayed IL-1β and IL-18 secretion from nigericin-treated PBMCs (Fig. S1 G–K). Thus,our results confirm recent findings that purified TcdA selectivelyengages the Pyrin inflammasome in mouse BMDMs (19) and ex-tend these results to human PBMCs.

The Pyrin Inflammasome Is Engaged by C. difficile Infection.Both TcdAand TcdB contribute critically to C. difficile-induced pseudomem-branous colitis (16), although the pathogen also activates additionalimmune mechanisms independently of TcdA and TcdB (20). Be-cause microbial pathogens may express several virulence factorsthat engage multiple inflammasomes in parallel (21, 22), and be-cause the mechanisms of inflammasome activation induced byC. difficile infection are unknown, we next studied inflammasomeresponses in C. difficile-infected macrophages. Caspase-1 was acti-vated in wild-type BMDMs infected with C. difficile, and this acti-vation resulted in substantial cleavage and extracellular release ofmature IL-1β (Fig. S3 A and B). These responses required live

bacteria and the expression of the bacterial toxins TcdA and TcdBbecause inflammasome-dependent cytokine processing and secre-tion were blunted when BMDMs were exposed to heat-killedC. difficile or were infected with the TcdA/B-deficient (VP11186)C. difficile strain (Fig. S3 A and B). As with the purified toxins (Fig.S1), C. difficile infection-induced caspase-1 activation required Pyrinand the inflammasome adaptor ASC, whereas Nlrp3 and caspase-11 were dispensable (Fig. S3 C andD). Likewise,Mefv−/− and Asc−/−

BMDMs failed to secrete IL-1β in the culture supernatants,whereas the supernatants of C. difficile-infected Nlrp3−/− and cas-pase-11−/− macrophages contained significant levels of IL-1β (Fig.S3 E and F). Notably, caspase-1 was responsible for the grossamount of IL-1βmaturation and secretion, but proIL-1βmaturationand IL-1β secretion were not fully inhibited in caspase-1−/−caspase-11−/− macrophages (Fig. S3 C and E). These results suggest thatadditional proteases may, to a limited extent, contribute to Pyrin-and ASC-dependent IL-1β secretion in C. difficile-infected macro-phages. Regardless, C. difficile-infected PBMCs of healthy individ-uals also secreted high levels of IL-1β and IL-18, and theseresponses were inhibited by the cell-permeable caspase-1 inhibitorAc-YVAD-cmk (Fig. S3 G and H). As in murine macrophages,C. difficile-induced inflammasome activation in human PBMCs waselicited by the bacterial toxins because IL-1β and IL-18 secretionwas blunted upon infection with the TcdA/TcdB-deficientC. difficile mutant (Fig. S3 I and J). Together, these results dem-onstrate that TcdA and TcdB are fully responsible for C. difficile-induced inflammasome activation in rodents and humans alike.

LPS Priming Is Dispensable but Enhances Pyrin Inflammasome Activation.Having established that the enterotoxins are required for inflam-masome activation induced by C. difficile infection, we next set outto investigate the mechanisms by which the enterotoxins engage thePyrin inflammasome. We first asked whether Pyrin activation re-quired priming, an important hallmark of the Nlrp3 inflammasome(23). Both TcdA and TcdB triggered caspase-1 maturation inunprimed macrophages, indicating that Toll-like receptor (TLR)priming is not a prerequisite for the Pyrin inflammasome (Fig.S4A). ProIL-1β is produced only in response to NF-κB cues suchas TLR ligands, whereas unprimed BMDMs express a constitu-tive pool of proIL-18 that is readily available for secretion uponinflammasome activation (24). In agreement, culture medium ofunprimed BMDMs secreted IL-18, but not IL-1β, when intoxicatedwith TcdA or TcdB (Fig. S4 B and C). As expected, LPS primingsupported the secretion of mature IL-1β by the two toxins (Fig.S4B) and further enhanced IL-18 secretion (Fig. S4C). Notably,LPS priming also markedly enhanced the upstream activation ofcaspase-1 relative to levels seen in unprimed BMDMs (Fig. S4A).These results suggested that LPS may increase Pyrin expression inBMDMs. Indeed, both a meta-analysis of public (25–29) micro-array data (Fig. S4D) and a longitudinal quantitative RT-PCRanalysis of Pyrin expression levels showed that LPS increased Pyrinand Nlrp3 transcript levels by ∼10-fold in the first hours (Fig. S4 Dand E). At 6 h after LPS stimulation, Pyrin levels had increasedfurther to ∼30-fold over baseline levels, whereas Nlrp3 remainedstable (Fig. S4E). Collectively, these findings indicate that, althoughpriming is not essential for Pyrin activation, it significantly enhancesPyrin inflammasome activation by TcdA and TcdB.

Microtubule-Depolymerizing Drugs Selectively Inhibit the PyrinInflammasome. Ectopically expressed Pyrin partially associateswith cytoskeletal structures (30), but how this localization relatesto its physiologic role in inflammasome activation is unclear.We found that neither inhibition of actin polymerization withcytochalasin D nor small-molecule targeting of downstream actineffectors (c-Abl kinase with STI-571, myosin II with myostatin,and Rock1/2 with Y-27632) interfered substantially with Pyrin-induced IL-1β secretion from murine BMDMs and humanPBMCs (Fig. S5 A and B). Also the microtubule-stabilizing agentpaclitaxel (taxol) failed to modulate the secretion of IL-1β andIL-18 from TcdA-treated PBMCs (Fig. S5 C and D). In markedcontrast, however, the microtubule polymerization inhibitor

Van Gorp et al. PNAS | December 13, 2016 | vol. 113 | no. 50 | 14385

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

colchicine abolished caspase-1 maturation as well as downstreamcleavage and secretion of IL-1β from TcdA-treated BMDMs (Fig. 1A and B). Lumicolchicine, a structurally related colchicine photo-isomer that does not bind tubulin (31), did not affect Pyrin acti-vation (Fig. 1 A and B), demonstrating the specificity of theseresults. Colchicine also abolished Pyrin-mediated IL-1β secretionfrom C. difficile-infected BMDMs (Fig. 1C). Consistent with theseresults, colchicine, but not lumicolchicine, inhibited Pyrin-inducedproIL-1β maturation in human PBMCs (Fig. 1D) and thus pre-vented the secretion of IL-1β and IL-18 from these cells (Fig. 1 Eand F). Importantly, a set of structurally unrelated microtubulepolymerization inhibitors (nocodazole, ABT-751, CA4P, and

CYT997) also abolished the maturation of caspase-1, therebyhampering the ensuing cleavage and secretion of IL-1β from TcdA-treated BMDMs (Fig. 1 G and H). Paralleling these results, thesetubulin polymerization inhibitors prevented Pyrin-dependent IL-1βmaturation in human PBMCs (Fig. 1I). Similarly, secretion of IL-1βand IL-18 from TcdA-treated PBMCs was significantly reduced(Fig. 1 J and K), establishing that microtubules are essential forhuman and murine Pyrin activation. These findings led us toexamine the role of microtubule polymerization in other inflam-masomes. Anthrax lethal toxin engaged Nlrp1b-dependent auto-maturation of caspase-1 and cleavage of proIL-1β regardless ofwhether BMDMs had been pretreated with colchicine (Fig. 1L).Nlrc4-driven caspase-1 activation and intracellular IL-1β cleavagein Salmonella enterica serovar Typhimurium (S. Typhimurium)-in-fected macrophages also were normal in the presence of colchicine(Fig. 1M), as was activation of the AIM2 inflammasome by trans-fected dsDNA (Fig. 1N). Likewise, colchicine failed to modulateextracellular IL-1β release by each of these inflammasomes (Fig. S6A–C). Nigericin-induced Nlrp3 inflammasome activation in murineBMDMs and human PBMCs was insensitive to colchicine in-hibition (Fig. 1O and Fig. S6 D–F), and nigericin-induced caspase-1activation along with downstream maturation and release of IL-1βcontinued unabated in the presence of the microtubule poly-merization inhibitors nocodazole, ABT-751, CA4P, and CYT997(Fig. S7). In conclusion, polymerized tubulin selectively and

A

E F G H

I J K

L M N O

B C D

Lum

i-col

chic

ine

Lum

i-col

chic

ine

Lum

i-col

chic

ine

Lum

i-col

chic

ine

Unt

reat

ed

Moc

k Lu

mic

olch

icin

e C

olch

icin

e

LPS+TcdA

Fig. 1. Microtubule depolymerizing drugs specifically inhibit the Pyrin in-flammasome. (A and B) Wild-type LPS-primed BMDMs were pretreated withlumicolchicine or colchicine before stimulation with TcdA. Samples wereimmunoblotted for caspase-1 and IL-1β (A), and supernatants were analyzedfor IL-1β (B). (C) Unprimed BMDMs were pretreated with colchicine before in-fection with C. difficile followed by supernatant collection and analysis of IL-1β.(D–F) PBMCs from healthy donors (n = 3) were pretreated with lumicolchicineor colchicine before stimulation with TcdA. Samples were immunoblotted forIL-1β (D), and supernatants were analyzed for IL-1β (E) and IL-18 (F). (G–K) LPS-primed BMDMs (G and H) and PBMCs from healthy donors (n = 3) (I–K) werepretreated with colchicine, nocodazole, ABT-751, CA4P, or CYT997 beforestimulation with TcdA. Samples were immunoblotted for caspase-1 and IL-1β(G), and supernatants were analyzed for IL-1β (H). PBMC samples were immu-noblotted for IL-1β (I), and supernatants were analyzed for IL-1β (J) and IL-18(K). (L–O) LPS-primed BMDMs were pretreated with colchicine before beingstimulated with activators of the Nlrp1b (anthrax lethal toxin; LeTx) (L), Nlrc4(S. Typhimurium; STm) (M), AIM2 (dsDNA) (N), and Nlrp3 (nigericin; Nig) (O)inflammasomes followed by immunoblot to detect caspase-1 and IL-1β. Blackarrowheads indicate procaspase-1 and proIL-1β, and white arrowheads indicatethe p20 and p17 subunits. Luminex data are shown as mean ± SD, and all dataare representative of at least three independent experiments.

A B C

D E

GF

Fig. 2. Differential microtubule regulation of Pyrin inflammasome activationidentifies FMF patients. (A) PBMCs from healthy donors (n = 7) and FMF patients(n = 2) were infected with C. difficile, and the supernatant was analyzed for IL-1β.(B) PBMCs from healthy donors (n = 3) and FMF patients (n = 2) were pretreatedwith Ac-YVAD-cmk before being stimulated with TcdA, and the supernatant wasanalyzed for IL-1β. (C andD) PBMCs from healthy donors (n= 7) and FMF patients(n = 2) were stimulated with LPS for 5 h (C) or were pretreated with colchicinebefore stimulation with TcdA (D), and the supernatant was analyzed for IL-1β. (E)PBMCs from healthy donors (n = 3) and FMF patients (n = 2) were pretreatedwith colchicine, nocodazole, ABT-751, CA4P, or CYT997 before stimulation withTcdA, and the supernatant was analyzed for IL-1β. (F and G) PBMCs from healthydonors (n = 5) and from CAPS (n = 4), JIA (n = 7), and FMF (n = 9) patients werepretreated with colchicine before stimulation with TcdA, and the supernatantwas analyzed for IL-1β (F) and IL-18 (G). Luminex data are shown as mean ± SD,and all data are representative of at least three independent experiments. ns,non-significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

14386 | www.pnas.org/cgi/doi/10.1073/pnas.1613156113 Van Gorp et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

critically controls activation of the Pyrin inflammasome in humansand rodents alike. Because murine Pyrin lacks the C-terminalB30.2 domain of human Pyrin, these results also imply that thisregion is dispensable for inflammasome activation and colchicineregulation. Nevertheless, the great majority of FMF mutations inhuman Pyrin localize to the C-terminal B30.2 domain (4).

Differential Microtubule Regulation of Pyrin Inflammasome ActivationIdentifies FMF Patients. We therefore sought to determine how mi-crotubules relate to Pyrin inflammasome signaling in FMF PBMCs.We reasoned that if microtubules relay an activating signal up-stream of Pyrin, colchicine would halt TcdA-induced inflamma-some activation in PBMCs of FMF patients. To test this hypothesis,we collected PBMCs from seven healthy controls and two FMFpatients with confirmed MEFV mutations in the C-terminal B30.2domain [Table S1, patients FMF1 (M694V/R761H) and FMF2(M694I/M694I)]. We did not detect secreted IL-1β in culturesupernatants of untreated PBMCs of healthy donors or FMFpatients (Fig. 2A). C. difficile infection triggered a substantial butcomparable release of IL-1β in wild-type and FMF PBMCs (Fig.2A). FMF PBMCs that had been intoxicated with TcdA also se-creted normal levels of IL-1β, a response that was efficiently

blocked by the caspase-1 inhibitor Ac-YVAD-cmk (Fig. 2B).Moreover, IL-1β levels secreted by FMF PBMCs in response tothe NLRP3 inflammasome stimulus LPS (32, 33) were compara-ble to those of healthy donors (Fig. 2C). These results indicatethat FMF mutations are not hypermorphic for inflammasomeactivation relayed by either Pyrin or NLRP3. Moreover, theysuggest that FMFmutations differ from CAPS-linked mutations inNLRP3 that significantly enhance LPS- and cold-induced Nlrp3inflammasome activation (33, 34). Remarkably, however, althoughcolchicine pretreatment abolished TcdA-induced IL-1β secretionfrom PBMCs of healthy individuals (Fig. 2D), it augmented theTcdA-induced IL-1β secretion from FMF PBMCs (Fig. 2D). Themicrotubule assembly inhibitors nocodazole, ABT-751, CA4P, andCYT997 also had opposite influences on TcdA-induced IL-1β se-cretion from FMF PBMCs and from PBMCs of healthy donors(Fig. 2E). Furthermore, IL-18 release from TcdA-treated wild-typePBMCs was blunted by colchicine and other microtubule poly-merization inhibitors, whereas FMF PBMCs resisted suppression(Fig. S8). To validate these results further and to test whetherresistance to colchicine inhibition was a defining feature of FMFPBMCs, we repeated our studies with PBMCs from a larger groupof healthy donors (n = 5) and from an additional cohort of FMF

A D

C

B

IG

H J

E

F

Fig. 3. FMF-associated Pyrin binds tubulin but doesnot require microtubules for ASC speck assembly. (A)Schematic representation of different Flag-taggedPyrin domain constructs. (B–D) Expression of differ-ent Flag-tagged Pyrin constructs in HEK293T cellsfollowed by immunoprecipitation with Flag-beadsand immunoblotting for tubulin and Flag. HEK293Tcells were transfected with empty vector (EV), Flag-tagged wild-type (WT FL), or FMF (M694V) full-length human pyrin (FL) and its different domains,PYD, linker, PYD+linker and C-term, followed byimmunoprecipitation with Flag-beads and immuno-blotting for tubulin and Flag. (E and F) LPS-primedBMDMs (E) or HEK293T cells transfected with wild-type or FMF (M694V) full-length human pyrin (F)were pretreated with colchicine before stimulationwith TcdA and then were immunoblotted to de-tect phosphorylated Pyrin (S241), IL-1β, and β-actin(BMDMs), and phosphorylated Pyrin (S242), Flag,and tubulin (293T). PBMCs from healthy donorswere pretreated with colchicine or CYT997 beforestimulation with TcdA or nigericin (G and H). PBMCsfrom healthy donors and from CAPS, JIA, and FMFpatients were pretreated with colchicine beforestimulation with TcdA (I and J). ASC specks wereanalyzed by confocal microscopy as shown in mi-crographs (G and I) and automated quantification (Hand J). Immunoblots and confocal images are rep-resentative of at least three independent experiments.ns, non-significant; ***P < 0.001; ****P < 0.0001.

Van Gorp et al. PNAS | December 13, 2016 | vol. 113 | no. 50 | 14387

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

patients with a variety of defined MEFV mutations (n = 9; TableS1). PBMCs from patients afflicted with CAPS disease resultingfrom heterozygous mutations inNLRP3 (n = 4; Table S1) and frompatients diagnosed with juvenile idiopathic arthritis (JIA, systemicand nonsystemic; n = 7) were tested also. As with PBMCs fromhealthy donors, colchicine blocked TcdA-induced IL-1β and IL-18secretion from PBMCs of CAPS and JIA patients, indicating thattheir Pyrin inflammasome responses were regulated identically tothose of healthy individuals (Fig. 2 F and G). In marked contrast,however, all FMF patients continued to secrete significant IL-1βand IL-18 levels after colchicine pretreatment (Fig. 2 F andG). Weverified that colchicine efficiently disrupted assembled microtubulesin the PBMCs of healthy individuals and FMF patients alike (Fig.S9), ruling out the remote possibility that microtubules of FMFPBMCs resisted microtubule disassembly by colchicine. Together,these results suggest that FMF mutations converge on lifting the crit-ical requirement for microtubules in Pyrin inflammasome activation.

FMF-Associated Pyrin Binds Tubulin but Does Not Require Microtubulesfor ASC Speck Assembly. Dephosphorylation of Pyrin’s intermediatelinker domain was recently shown to be required for TcdA- andTcdB-induced inflammasome activation (14, 15, 19), and intro-duction of FMF mutations in the B30.2 domain of ectopicallyexpressed Pyrin did not interfere with this process (15). In agreementwith these reports, we showed that endogenous Pyrin was not con-stitutively active in FMF PBMCs but was engaged only after TcdAstimulation or C. difficile infection (Fig. 2 and Fig. S8). To charac-terize further how FMF mutations alter Pyrin-dependent inflam-masome activation, we examined the binding of ectopically expressedPyrin domains to endogenous tubulin (Fig. 3A). Flag-fused con-structs corresponding respectively to the N-terminal PYD domain,the intermediate linker sequence, the PYD+linker combination, orthe C-terminal tripartite motif (TRIM) and B30.2 domains wereexpressed in 293T cells and were immunoprecipitated with Flag-beads. Constructs containing either the PYD or the C-terminal re-gions of wild-type Pyrin coprecipitated endogenous tubulin, but theintermediate linker sequence did not (Fig. 3B). This result suggeststhat Pyrin interacts with tubulin through both its N and C termini.However, introduction of the FMF-associated M694V mutation didnot prevent tubulin binding in either full-length Pyrin (Fig. 3C) or inthe isolated carboxyl-terminal region (Fig. 3D). As reported (19),TcdA also triggered Pyrin dephosphorylation in wild-type BMDMsthat had been pretreated with colchicine (Fig. 3E). Furthermore,colchicine failed to prevent TcdA-induced dephosphorylationof ectopically expressed wild-type Pyrin and the FMF-associ-ated M694V Pyrin mutant in 293T cells (Fig. 3F). In agreementwith our other studies showing that microtubule polymerizationinhibitors prevented inflammasome activation selectively in wild-type, but not in FMF, PBMCs, this finding positions microtubulesin the pathway downstream of Pyrin dephosphorylation.PYD-based inflammasome sensors such as Pyrin nucleate ASC

filaments through a biphasic mechanism in which their PYD do-main first nucleates ASC PYD to oligomerize into prion-like fi-bers that then condense into ASC specks through ASC–caspaserecruitment domain (CARD) interactions that also enable therecruitment of caspase-1 zymogens (35). Interference with thesefirst and second stages of ASC speck assembly therefore results ineither the complete absence or the formation of atypical fila-mentous fibers, respectively (35). We exposed PBMCs of healthydonors to TcdA or nigericin to trigger ASC speck assembly throughthe Pyrin and NLRP3 inflammasomes, respectively (Fig. 3 G andH). Colchicine and the unrelated microtubule polymerization in-hibitor CYT997 selectively prevented ASC speck formation in re-sponse to TcdA but not in nigericin-treated PBMCs (Fig. 3 G andH), demonstrating that microtubules were selectively required fornucleation of ASC specks by Pyrin but not by NLRP3. As in healthydonors, colchicine abolished TcdA-induced ASC speck assemblyin PBMCs of CAPS and JIA patients (Fig. 3 I and J). In markedcontrast, however, the assembly of ASC specks by FMF PBMCswas unabated in the presence of colchicine (Fig. 3 I and J).Similar results were obtained with the unrelated microtubule

polymerization inhibitor CYT997 (Fig. S10). Thus, FMF muta-tions in Pyrin remove the critical dependency on microtubules forASC speck assembly and inflammasome activation downstreamof Pyrin dephosphorylation.

DiscussionThe observation that Pyrin inflammasome activation by TcdA,TcdB, and live C. difficile infection required intact microtubules inboth human PBMCs and murine macrophages implies that theC-terminal B30.2 domain, which harbors most FMF mutations inhumans but is absent in mouse Pyrin, is dispensable for inflamma-some activation. Although the B30.2 domain is dispensable forPyrin inflammasome activation, we established here that FMFmutations in this domain nonetheless remove the critical reliance onintact microtubules for Pyrin-based nucleation of ASC specks andinflammasome signaling. Microtubules were recently proposed tocontrol inflammasome activation apically of Pyrin dephosphorylationin response to bacterial RhoA inactivation (14). However, thissuggestion is difficult to reconcile with the observation that TcdA-induced Pyrin dephosphorylation continued unhampered incolchicine-pretreated macrophages and 293T cells (Fig. 3 E andF and ref. 19). Moreover, we showed that FMF mutations renderPyrin activation independent of microtubules. Thus, our resultsprovide a conceptual framework for understanding FMF basedon a mechanistic model of Pyrin signaling in which microtubulescontrol inflammasome activation downstream of Pyrin dephos-phorylation (Fig. 4). In this model, microtubules relay an activatingsignal to dephosphorylated wild-type Pyrin that shifts autorepressedPyrin into an open conformation. FMF mutations in the humanB30.2 domain may force dephosphorylated Pyrin in an open con-formation that readily binds the inflammasome adaptor ASC, ef-fectively replacing microtubule-relayed signals (Fig. 4).Paradoxically, although we show here that FMF mutations render

Pyrin inflammasome activation insensitive to colchicine, this drug isan effective treatment that suppresses periodic inflammatory attacksin the majority of FMF patients and prevents amyloidosis, a majorlong-term complication of the disease that may result in renal failureand death (5, 36). However, the clinical efficacy of colchicine treat-ment is likely associated with its ability to decrease leukocyte motilityand phagocytosis during inflammation (37, 38). Recent studies sug-gested that defects in the mevalonate pathway seen in the hereditaryautoinflammatory disease mevalonate kinase deficiency (MKD), alsonamed “hyperimmunoglobulinemia D syndrome” (HIDS), may alsotrigger unwarranted activation of the Pyrin inflammasome (14, 39).

WildtypePyrin

FMFPyrin

RhoA glucosylation

Inflammasome assembly

pro-IL-1βpro-IL-18

IL-1βIL-18

C. difficile TcdA

Pyrin Ser208/Ser242 dephosphorylation

microtubulesmicrotubules

Fig. 4. Schematic model of Pyrin inflammasome activation by RhoA-modifyingtoxins. C. difficile TcdA and TcdB inactivate RhoA, thereby triggering de-phosphorylation of Pyrin and its release from inhibitory 14-3-3 proteins. Mi-crotubules are critical for activation of wild-type Pyrin in human PBMCs andmurine macrophages. In contrast, FMF-associatedmutations in Pyrin render ASCspeck assembly and inflammasome activation independent of microtubules.

14388 | www.pnas.org/cgi/doi/10.1073/pnas.1613156113 Van Gorp et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020

In contrast to FMF patients, however, MKD patients generally donot benefit from colchicine treatment, whereas blockade of IL-1 hasshown promising results (40). Given our observations that FMFmutations render Pyrin activation resistant to colchicine blockade, itwould be interesting to investigate the role of microtubules in MKD-associated inflammasome activation.Akin to colchicine, certain pathogens express toxins that manip-

ulate microtubule dynamics, as exemplified by the CDT toxin ofhypervirulentC. difficile strains (41). It therefore is tempting to speculatethat the high frequency of heterozygous MEFV mutations in endemicFMF regions (42) might be related to their rendering Pyrin acti-vation insensitive to microtubule manipulations by such pathogens.Given the key role of inflammasomes in antimicrobial host defense(43), the ability to engage the Pyrin inflammasome in the presenceof microtubule dynamics blockade is likely to have offered het-erozygous individuals a selective advantage in clearing such in-fections. Finally, the insight that inflammasome activation by FMFPyrin resists colchicine blockade enables functional/immunologicalscreening of the disease among clinically overlapping autoin-flammatory patients and thus may contribute to timely diagnosisand commencement of therapy in the future.

Materials and MethodsAll reported patients and healthy controls provided written informed consentfor participation in the study, in accordance with International Conferenceon Harmonization of Technical Requirements for Registration of Pharmaceu-ticals for Human Use/Good Clinical Practice (ICH/GCP) guidelines. The researchprotocol was approved by the ethics committee of Ghent University Hospitalunder number 2012_593 and the protocols of Bambino Gesù Children’s Hospital.

All mice were kept in specific pathogen-free conditions within the animalfacilities of Ghent University. All animal experiments were approved by theethics committee on laboratory animal welfare of Ghent University.

Detailed methods used in all experiments throughout this work are de-scribed in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank the patients and their families who providedspecimens for this study; Vishva Dixit and Nobuhiko Kayagaki (Genentech) forgenerously supplying mutant mice; and Amelie Fossoul (VIB-University of Ghent)and the VIB Bio Imaging Core for technical support. F.V.H., L.V.W., and N.V.O.are postdoctoral fellows with the Fund for Scientific Research-Flanders. This workwas supported by Ghent University Concerted Research Actions Grant BOF14/GOA/013, European Research Council Grant 281600, and a Baillet Latour MedicalResearch Grant (to M.L.).

1. Van Gorp H, Kuchmiy A, Van Hauwermeiren F, Lamkanfi M (2014) NOD-like receptorsinterfacing the immune and reproductive systems. FEBS J 281(20):4568–4582.

2. McGonagle D, McDermott MF (2006) A proposed classification of the immunologicaldiseases. PLoS Med 3(8):e297.

3. Masters SL, Simon A, Aksentijevich I, Kastner DL (2009) Horror autoinflammaticus: Themolecular pathophysiology of autoinflammatory disease. Annu Rev Immunol 27:621–668.

4. Portincasa P (2016) Colchicine, biologic agents and more for the treatment of familialMediterranean Fever. The old, the new, and the rare. Curr Med Chem 23(1):60–86.

5. Ozen S, Bilginer Y (2014) A clinical guide to autoinflammatory diseases: FamilialMediterranean fever and next-of-kin. Nat Rev Rheumatol 10(3):135–147.

6. Milhavet F, et al. (2008) The infevers autoinflammatory mutation online registry:Update with new genes and functions. Hum Mutat 29(6):803–808.

7. French FMF Consortium (1997) A candidate gene for familial Mediterranean fever.Nat Genet 17(1):25–31.

8. The International FMF Consortium (1997) Ancient missense mutations in a new memberof the RoRet gene family are likely to cause familial Mediterranean fever. Cell 90(4):797–807.

9. Daniels M, Shohat T, Brenner-Ullman A, Shohat M (1995) Familial Mediterraneanfever: High gene frequency among the non-Ashkenazic and Ashkenazic Jewishpopulations in Israel. Am J Med Genet 55(3):311–314.

10. Rogers DB, et al. (1989) Familial Mediterranean fever in Armenians: Autosomal re-cessive inheritance with high gene frequency. Am J Med Genet 34(2):168–172.

11. Toplak N, et al.; Paediatric Rheumatology International Trials Organisation (PRINTO),Eurotraps and Eurofever Projects (2012) An international registry on auto-inflammatory diseases: The Eurofever experience. Ann Rheum Dis 71(7):1177–1182.

12. Gavrilin MA, et al. (2012) Activation of the pyrin inflammasome by intracellularBurkholderia cenocepacia. J Immunol 188(7):3469–3477.

13. Xu H, et al. (2014) Innate immune sensing of bacterial modifications of Rho GTPasesby the Pyrin inflammasome. Nature 513(7517):237–241.

14. Park YH, Wood G, Kastner DL, Chae JJ (2016) Pyrin inflammasome activation andRhoA signaling in the autoinflammatory diseases FMF and HIDS. Nat Immunol 17(8):914–921.

15. Masters SL, et al. (2016) Familial autoinflammation with neutrophilic dermatosis re-veals a regulatory mechanism of pyrin activation. Sci Transl Med 8(332):332ra45.

16. Smits WK, Lyras D, Lacy DB, Wilcox MH, Kuijper EJ (2016) Clostridium difficile in-fection. Nat Rev Dis Primers 2:16020.

17. Coll RC, et al. (2015) A small-molecule inhibitor of the NLRP3 inflammasome for thetreatment of inflammatory diseases. Nat Med 21(3):248–255.

18. Perregaux DG, et al. (2001) Identification and characterization of a novel class ofinterleukin-1 post-translational processing inhibitors. J Pharmacol Exp Ther 299(1):187–197.

19. Gao W, Yang J, Liu W, Wang Y, Shao F (2016) Site-specific phosphorylation and mi-crotubule dynamics control Pyrin inflammasome activation. Proc Natl Acad Sci USA113(33):E4857–E4866.

20. Hasegawa M, et al. (2011) Nucleotide-binding oligomerization domain 1 mediatesrecognition of Clostridium difficile and induces neutrophil recruitment and pro-tection against the pathogen. J Immunol 186(8):4872–4880.

21. Wu J, Fernandes-Alnemri T, Alnemri ES (2010) Involvement of the AIM2, NLRC4, andNLRP3 inflammasomes in caspase-1 activation by Listeria monocytogenes. J ClinImmunol 30(5):693–702.

22. Rathinam VA, et al. (2010) The AIM2 inflammasome is essential for host defenseagainst cytosolic bacteria and DNA viruses. Nat Immunol 11(5):395–402.

23. Bauernfeind FG, et al. (2009) Cutting edge: NF-kappaB activating pattern recognitionand cytokine receptors license NLRP3 inflammasome activation by regulating NLRP3expression. J Immunol 183(2):787–791.

24. Van Opdenbosch N, et al. (2014) Activation of the NLRP1b inflammasome in-dependently of ASC-mediated caspase-1 autoproteolysis and speck formation. NatCommun 5:3209.

25. Carlson BA, et al. (2009) Selenoproteins regulate macrophage invasiveness and ex-tracellular matrix-related gene expression. BMC Immunol 10:57.

26. Chen X, et al. (2012) Requirement for the histone deacetylase Hdac3 for the in-flammatory gene expression program in macrophages. Proc Natl Acad Sci USA 109(42):E2865–E2874.

27. Litvak V, et al. (2009) Function of C/EBPdelta in a regulatory circuit that discriminatesbetween transient and persistent TLR4-induced signals. Nat Immunol 10(4):437–443.

28. Rodgers MA, et al. (2014) The linear ubiquitin assembly complex (LUBAC) is essentialfor NLRP3 inflammasome activation. J Exp Med 211(7):1333–1347.

29. Rowley SM, et al. (2014) Tumor progression locus 2 (Tpl2) kinase promotes chemokinereceptor expression and macrophage migration during acute inflammation. J BiolChem 289(22):15788–15797.

30. Mansfield E, et al. (2001) The familial Mediterranean fever protein, pyrin, associateswith microtubules and colocalizes with actin filaments. Blood 98(3):851–859.

31. Bryan J (1972) Definition of three classes of binding sites in isolated microtubulecrystals. Biochemistry 11(14):2611–2616.

32. Netea MG, et al. (2009) Differential requirement for the activation of the in-flammasome for processing and release of IL-1beta in monocytes and macrophages.Blood 113(10):2324–2335.

33. Lamkanfi M, et al. (2009) Glyburide inhibits the Cryopyrin/Nalp3 inflammasome. J CellBiol 187(1):61–70.

34. Hoffman HM, et al. (2004) Prevention of cold-associated acute inflammation in fa-milial cold autoinflammatory syndrome by interleukin-1 receptor antagonist. Lancet364(9447):1779–1785.

35. Lu A, et al. (2014) Unified polymerization mechanism for the assembly of ASC-dependentinflammasomes. Cell 156(6):1193–1206.

36. Giancane G, et al. (2015) Evidence-based recommendations for genetic diagnosis offamilial Mediterranean fever. Ann Rheum Dis 74(4):635–641.

37. Goldfinger SE (1972) Colchicine for familial Mediterranean fever. N Engl J Med287(25):1302.

38. Lidar M, et al. (2004) Colchicine nonresponsiveness in familial Mediterranean fever:Clinical, genetic, pharmacokinetic, and socioeconomic characterization. Semin ArthritisRheum 33(4):273–282.

39. Akula MK, et al. (2016) Control of the innate immune response by the mevalonatepathway. Nat Immunol 17(8):922–929.

40. Ter Haar N, et al.; Paediatric Rheumatology International Trials Organisation (PRINTO) andthe Eurofever/Eurotraps Projects (2013) Treatment of autoinflammatory diseases: Resultsfrom the Eurofever Registry and a literature review. Ann Rheum Dis 72(5):678–685.

41. Schwan C, et al. (2009) Clostridium difficile toxin CDT induces formation of micro-tubule-based protrusions and increases adherence of bacteria. PLoS Pathog 5(10):e1000626.

42. Schaner P, et al. (2001) Episodic evolution of pyrin in primates: Human mutationsrecapitulate ancestral amino acid states. Nat Genet 27(3):318–321.

43. Lamkanfi M, Dixit VM (2011) Modulation of inflammasome pathways by bacterial andviral pathogens. J Immunol 187(2):597–602.

44. Kayagaki N, et al. (2011) Non-canonical inflammasome activation targets caspase-11.Nature 479(7371):117–121.

45. Vande Walle L, et al. (2014) Negative regulation of the NLRP3 inflammasome by A20protects against arthritis. Nature 512(7512):69–73.

Van Gorp et al. PNAS | December 13, 2016 | vol. 113 | no. 50 | 14389

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Oct

ober

1, 2

020