ComEA-mCherry protein, which is indicative ofDNA translocation into the periplasm (6) of a pred-ator cell (Fig. 4 and figs. S13 and S14). Similar gene-transfer events were never or only rarely observedin the T6SS-negative strain; in the presence ofextracellular deoxyribonuclease; if the predatorlacked the outer membrane secretin protein PilQ,which is required for efficient DNA uptake (5, 6);or in the absence of any prey (figs. S14 and S15).Our findings indicated that the T6SS of V.

cholerae is part of the competence regulon andis induced on chitinous surfaces in a TfoX-,HapR-, and QstR-dependent manner, therebyenhancing HGT (Fig. 5). HGT plays a majorrole in bacterial evolution and contributes tothe spread of antibiotic resistance cassettes andpathogenicity islands. Moreover, because chitinouszooplankton are thought to play an importantrole in cholera transmission in endemic regions (1),chitin-induced expression of the T6SSmight alsoenhance the virulence potential of the pathogendue to the killing of commensal bacteria withinthe human gut.

REFERENCES AND NOTES

1. E. K. Lipp, A. Huq, R. R. Colwell, Clin. Microbiol. Rev. 15,757–770 (2002).

2. I. Chen, D. Dubnau, Nat. Rev. Microbiol. 2, 241–249 (2004).3. C. Johnston, B. Martin, G. Fichant, P. Polard, J. P. Claverys,

Nat. Rev. Microbiol. 12, 181–196 (2014).4. K. L. Meibom, M. Blokesch, N. A. Dolganov, C.-Y. Wu,

G. K. Schoolnik, Science 310, 1824–1827 (2005).5. P. Seitz, M. Blokesch, Proc. Natl. Acad. Sci. U.S.A. 110,

17987–17992 (2013).6. P. Seitz et al., PLOS Genet. 10, e1004066 (2014).7. P. Seitz, M. Blokesch, FEMS Microbiol. Rev. 37, 336–363 (2013).8. S. Yamamoto et al., Mol. Microbiol. 91, 326–347 (2014).9. A. B. Dalia, D. W. Lazinski, A. Camilli, mBio 5, e01028-13 (2014).10. M. Blokesch, Environ. Microbiol. 14, 1898–1912 (2012).11. W. L. Ng, B. L. Bassler, Annu. Rev. Genet. 43, 197–222 (2009).12. G. Suckow, P. Seitz, M. Blokesch, J. Bacteriol. 193, 4914–4924

(2011).13. M. Lo Scrudato, M. Blokesch, PLOS Genet. 8, e1002778 (2012).14. M. Lo Scrudato, M. Blokesch, Nucleic Acids Res. 41,

3644–3658 (2013).15. B. T. Ho, T. G. Dong, J. J. Mekalanos, Cell Host Microbe 15,

9–21 (2014).16. A. B. Russell, S. B. Peterson, J. D. Mougous, Nat. Rev.

Microbiol. 12, 137–148 (2014).17. D. Unterweger et al., Nat. Commun. 5, 3549 (2014).18. E. Durand, C. Cambillau, E. Cascales, L. Journet, Trends

Microbiol. 22, 498–507 (2014).19. Materials and methods are available as supplementary

materials on Science Online.20. S. Pukatzki et al., Proc. Natl. Acad. Sci. U.S.A. 103, 1528–1533

(2006).21. M. M. Shneider et al., Nature 500, 350–353 (2013).22. G. Bönemann, A. Pietrosiuk, A. Diemand, H. Zentgraf, A. Mogk,

EMBO J. 28, 315–325 (2009).23. M. Basler, M. Pilhofer, G. P. Henderson, G. J. Jensen,

J. J. Mekalanos, Nature 483, 182–186 (2012).24. M. Basler, B. T. Ho, J. J. Mekalanos, Cell 152, 884–894

(2013).25. T. G. Dong, J. J. Mekalanos, Nucleic Acids Res. 40, 7766–7775

(2012).26. T. Ishikawa et al., Infect. Immun. 80, 575–584 (2012).27. R. L. Marvig, M. Blokesch, BMC Microbiol. 10, 155 (2010).28. D. P. Keymer, M. C. Miller, G. K. Schoolnik, A. B. Boehm, Appl.

Environ. Microbiol. 73, 3705–3714 (2007).29. M. C. Miller, D. P. Keymer, A. Avelar, A. B. Boehm,

G. K. Schoolnik, Appl. Environ. Microbiol. 73, 3695–3704 (2007).30. T. G. Dong, B. T. Ho, D. R. Yoder-Himes, J. J. Mekalanos, Proc.

Natl. Acad. Sci. U.S.A. 110, 2623–2628 (2013).31. J. P. Claverys, L. S. Håvarstein, Nat. Rev. Microbiol. 5, 219–229

(2007).32. S. T. Miyata, D. Unterweger, S. P. Rudko, S. Pukatzki, PLOS

Pathog. 9, e1003752 (2013).

ACKNOWLEDGMENTS

We thank A. Boehm, M. Miller, members of the Institut Nationalde Recherche Biomédicale of the Democratic Republic of theCongo, and M. Lo Scrudato for providing V. cholerae strains.We also acknowledge the service provided by Microsynth andmembers of the Blokesch laboratory for scientific discussions.This work was supported by the Swiss National Science Foundation(grant 31003A_143356) and the European Research Council(grant 309064-VIR4ENV). Supporting data are provided in thesupplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/347/6217/63/suppl/DC1Materials and MethodsFigs. S1 to S15Tables S1 to S5References (33–54)

15 August 2014; accepted 1 December 201410.1126/science.1260064

INNATE IMMUNITY

Dermal adipocytes protect againstinvasive Staphylococcus aureusskin infectionLing-juan Zhang,1 Christian F. Guerrero-Juarez,2,3 Tissa Hata,1 Sagar P. Bapat,4

Raul Ramos,2,3 Maksim V. Plikus,2,3 Richard L. Gallo1*

Adipocytes have been suggested to be immunologically active, but their role in hostdefense is unclear. We observed rapid proliferation of preadipocytes and expansion ofthe dermal fat layer after infection of the skin by Staphylococcus aureus. Impairedadipogenesis resulted in increased infection as seen in Zfp423nur12 mice or in mice giveninhibitors of peroxisome proliferator–activated receptor g. This host defense functionwas mediated through the production of cathelicidin antimicrobial peptide from adipocytesbecause cathelicidin expression was decreased by inhibition of adipogenesis, andadipocytes from Camp−/− mice lost the capacity to inhibit bacterial growth.Together, thesefindings show that the production of an antimicrobial peptide by adipocytes is animportant element for protection against S. aureus infection of the skin.

Host defense against microbial infection in-volves the participation of several cell types.Owing to the rapid doubling time of manymicrobes, immediate protection providedby local resident cells—such as epithelial

cells, mast cells, and resident leukocytes—is es-sential to restrict the spread of infection duringthe lag period before recruitment of additionalcells, such as neutrophils and monocytes (1, 2).The production of antimicrobial peptides (AMPs)by local resident cells and recruited leukocytes isa keymechanism to limit pathogen growth (3–5).Staphylococcus aureus is a major cause of skin

and soft-tissue infections in humans, causing bothlocal and systemic disease (6, 7). We observed thata large and previously unrecognized expansionof the subcutaneous adipose layer was evidentduring the early response to S. aureus skin in-fection (Fig. 1A). The response to infection wasconfirmed with quantification of the abundanceof adipocytes (Fig. 1B and fig. S1A), observationsof an increase in lipid staining (fig. S1B), and

increased activation of the adiponectin promoteras measured in AdipoQcre;R26R mice (Fig. 1C)(8). Adipocytes progressively increased in sizeafter S. aureus infection (Fig. 1B), suggesting thatthe expansion of dermal adipose tissue occurs atleast in part through hypertrophy of mature ad-ipocytes. PREF1 and ZFP423 mark committedpreadipocytes required for adipose tissue de-velopment and expansion (9–11). Proliferationof these preadipocytes at the site of infectionwas further confirmed with colocalization ofPREF1 and ZFP423 with proliferation markersBrdU (Fig. 1D and fig. S1C) and Ki67 (fig. S1D).Additionally, dermal cells isolated from S. aureus–infected skin exhibited greater adipogenic po-tential than that of cells isolated from the sameamount of uninfected skin, as indicated by lipidproduction and induction of adipocyte markersAdipoq and Fabp4 in response to adipocyte dif-ferentiation medium (Fig. 1E and fig. S1E). Alsosupporting the conclusion that infection resultsin an increase of cells within the dermis with thepotential to differentiate into adipocytes wereobservations of an increase ofmRNAand proteinfor transcription factors driving preadipocytedifferentiation, includingCebpb,Pparg, andCebpa(Fig. 1F and fig. S1, D and F) (12, 13). Peroxisomeproliferator–activated receptor–g (PPARg)–positivecells at the infected sites were negative for CD11b(fig. S1G), confirming that they were not myeloidcells. To test that cell proliferationwas associatedwith adipocyte formation, we examined BrdU

SCIENCE sciencemag.org 2 JANUARY 2015 • VOL 347 ISSUE 6217 67

RESEARCH | REPORTS

1Division of Dermatology, University of California, San Diego(UCSD), La Jolla, CA 92093, USA. 2Department ofDevelopmental and Cell Biology, Sue and Bill Gross StemCell Research Center, University of California, Irvine, Irvine,CA 92697, USA. 3Center for Complex Biological Systems,University of California, Irvine, Irvine, CA 92697, USA.4Nomis Foundation Laboratories for Immunobiology andMicrobial Pathogenesis, The Salk Institute for BiologicalStudies, San Diego, La Jolla, CA 92037, USA.*Corresponding author. E-mail: rgallo@ucsd.edu

incorporation within the nuclei of adipocytesafter multiple injections of BrdU (14) during thefirst 3 days after infection. A significant increasein the number of BrdU-positive nuclei was seenwithin cells from S. aureus–infectedmice stained

with the adipocyte surface protein Caveolin (Fig.1G), thus confirming that adipocytes at the siteof infection were at least partially derived fromproliferative precursors. Together, these datashowed that infection by S. aureus triggers pre-

adipocyte proliferation and expansion of localdermal adipocytes.Wenext testedwhether adipocyte activationwas

essential for protection against S. aureus infec-tion using Zfp423 reporter mice and Zfp423nur12

68 2 JANUARY 2015 • VOL 347 ISSUE 6217 sciencemag.org SCIENCE

Fig. 1. Skin infection stimulates an increase indermal adipocytes. (A) Hematoxylin and eosinstaining of mouse skin injected with phosphate-buffered saline (PBS) control (Ctrl) orS. aureus (SA).Red brackets indicate subcutaneous adipose layer.Scale bars, 200 mm. (B)Quantification of the numberand size distribution of Caveolin+/Perilipin+ adipo-cytes 3 days after Ctrl or S. aureus injection (n = 3to ~5 mice/group and 3 microscopy fields/mice).(C) Increase of adiponectin positive cells seen bystaining for b-Gal (red) 3 days after S. aureus orCtrl injection in AdipoQcre;R26R mice. Wild-typemice injectedwithS. aureus are shownas a stainingcontrol. Scale bars, 200 mm. Nuclei were stainedwith 4ʹ,6-diamidino-2-phenylindole (DAPI) (blue).(D) Ctrl or S. aureus–infected mice (day 1) were in-jected with BrdU 4 hours before analysis. Proliferative preadipocytes wereidentified by means of yellow staining showing colocalization of PREF1 orZFP423 (red) and BrdU (green). Scale bars, 100 mm. (E) Dermal cells were iso-lated from Ctrl or S. aureus (SA)–infected skin then treated with adipocyte dif-ferentiationmedium for 5 days. Lipid production was shown byOil-Red-O (ORO)staining.Scale bars, 200 mm. (F) RelativemRNAexpression forCebpb andPparg

(n=3mice/group) in skin afterS. aureus infection. (G) (Top) Schematic of 3-dayBrdU labeling experiments during S. aureus infection. (Left) Representativeimages for caveolin (red) and BrdU (green) staining of skin sections. Arrowsindicate Caveolin+BrdU+ adipocytes. Scale bars, 50 mm. (Right) Quantification ofthe number of Caveolin+BrdU+ adipocytes (n = 3~5 mice/group and 3 micros-copy fields/mice). All error bars indicate mean T SEM; ** P < 0.01 (t test).

Fig. 2. Adipocytes are essential for host defenseagainst S. aureus infection. (A) b-gal stainingof the underside of Zfp423lacZ/+ skin 3 days afterinjection with PBS (Ctrl) or S. aureus. Injected areais indicated by dotted line. Scale bars, 1mm (left),200 mm(right). (B)GFP immunostaining (red) of skinsections from Ctrl or S. aureus–injected Zfp423GFP

mice. Scale bars, 100 mm. (C) Infected lesion size inZfp423nur12 or Zfp423+/+ mice (n = 7~8/group) and(D) bacteria in adipose observed by Gram stain-ing (dark blue). Scale bars, 100 mm. (E) Systemicbacteremia detected 3 days after S. aureus skininjection in spleens isolated from Zfp423nur12 micebut not Zfp423+/+ mice (n = 3 mice/group). Scalebars, 200 mm. (F toH) BADGEor GW9662 increasedsusceptibility to S. aureus as observed by (F) in-creased lesion size and [(G) and (H)] increasedS. aureus (CFU) from the skin calculated withplate counting (n = 8 to ~10/group). All error barsindicatemeanT SEM.**P<0.01, ***P<0.001 (t test).

Ctrl BADGE

Ctrl BADGELesi

on s

ize

(mm

)

**

0

5

10

15

Zfp423 nur12

**

u.d.C

FU

/gm

Zfp423 +/+

600

400

200

0

1000

800

S. a

ureu

s (C

FU

/gm

) **

Ctrl BADGE101

102

103

104

105

106

107

Ctrl GW9662S. a

ureu

s(C

FU

/gm

)

103

104

105

106

107***

Zfp

423

+/+

Zfp

423

nur1

2

Zfp423+/+ Zfp423 nur12

Lesi

on s

ize

(mm

)

***15

10

5

0

Zfp423 +/+ Zfp423 nur12

Ctr

lS

A

Zfp423 lacZ/+

β-Gal

Ctr

lS

A

GFP

Zfp423GFP

GRAM

RESEARCH | REPORTS

mice in which adipogenesis is prominently im-pared (11, 15). Activation of Zfp423 during infec-tion was confirmed by visualizing b-Gal stainingon the underside of skin from infected Zfp423lacZ

reporter mice (Fig. 2A) (16). Immunostainingof infected Zfp423GFP reporter mice (17) showedgreen fluorescent protein–positive (GFP+) cellslocalized within infected dermal adipose tis-sue and mostly colocalized with a fibroblastmarker [platelet-derived growth factor receptor–a(PDGFRa)], but notwith an endothelial cellmarker(CD31) (Fig. 2B and fig. S2, A andB). After S. aureusinfection, dermal adipose tissue in Zfp423nur12

mice expanded less than in control mice (fig.S2C). Immunostaining with the adipocyte markerPerilipin (PLIN) further confirmed that adipocyteformation was reduced in the Zfp423nur12 micecompared with control (fig. S2D). Impaired adipo-genesis in Zfp423nur12 mice was accompaniedby increased susceptibility to S. aureus skin infec-tion at the site injectedwith bacteria (Fig. 2, C andD) and a subsequent systemic bacteremia thatwas not detectable in controls (Fig. 2E). Increasedsusceptibility to S. aureus in Zfp423nur12 micewas associated with decreased activation ofPDGFRa+Sca1+ skin preadipocytes (fig. S2E) (18),but not with defective leukocyte recruitment be-cause there was no decrease in CD11b staining orneutrophil infiltration in the skin of Zfp423nur12

mice (fig. S2F).

To complement the observations made inZfp423nur12 mice, we chemically inhibited adipo-genesisbypretreatingwild-typemicewithbisphenolA diglycidyl ether (BADGE) or GW9662, bothpharmacological inhibitors of PPARg and acutechemical inhibitors of adipogenesis (14, 19, 20).Similarly toZfp423nur12mice, BADGE- orGW9662-treated mice showed decreased adipose expansionafter S. aureus infection (fig. S3A) and becamemore susceptible to infection (Fig. 2, F to H, andfig. S3B). In BADGE-treated mice, PDGFRa+Sca1+

preadipocyte numbers appeared normal (fig. S3C),but PPARg expression decreased (fig. S3D), sug-gesting as in a previous report (14) that inhibi-tion of PPARg impaired dermal adipose growth byblocking preadipocyte differentiation into adipo-cytes rather than by inhibiting the preadipocytecommitment process. Increased susceptibility toS. aureus in BADGE-treatedmicewas not causedby defective leukocyte recruitment (fig. S3E) norby any impaired ability of blood neutrophils tokill S. aureus (fig. S3F). Because neutrophil func-tion is a major factor in resisting S. aureus skininfection (6), this finding confirmed that the in-creased susceptibility to infection observed withPPARg inhibitors was caused by their effect onadipocyte formation.3T3-L1 cells are a well-characterized primary

mouse preadipocyte line that can be inducedto differentiate into mature adipocytes (21). To

determine how adipocytes might protect againstinfection, a comparison of antimicrobial pep-tide (AMP) gene expression was done in 3T3-L1 undifferentiated preadipocytes (pAds) anddifferentiated adipocytes (Ads). This showedthat mouse cathelicidin antimicrobial peptide(Camp) was strongly induced in Ads comparedwith pAds, whereas other AMPs belonging tothe a- and b-defensin families did not show anincrease in expression (Fig. 3A). Camp mRNAincreased at day 1 after differentiation then de-creased during later stages of differentiation(fig. S4A). Camp expression preceded expres-sion of adipokines Resistin (Retn), Leptin (Lep),and Fabp4, whereas the relative expression ofanother AMP, Defb14, decreased during differ-entiation (fig. S4, B to H).Cathelicidin protein peaked at day 2 to ~4 after

differentiation and was abundantly detectablewith Western blotting of cell extracts from differ-entiating pAds (fig. S4I). The size of cathelicidindetected bymeans ofWestern blot was consistentwith the mCAP18 cathelicidin precursor protein(22). Cathelicidin was also found with Westernblotting in adipocyte-conditioned medium (Fig.3B). Secretion of cathelicidin protein precededsecretion of Resistin, Leptin, and FABP4 at days2 to 4 after differentiation (Fig. 3B) and was sim-ilar to that of Collagen type-IV (COL4A1), whichwas produced 1 to 2 days after differentiation

SCIENCE sciencemag.org 2 JANUARY 2015 • VOL 347 ISSUE 6217 69

Adipocytes (Ads)preadipocytes (pAds)

Rel

ativ

e m

RN

A a

bund

ance

Campα-defensins β-defensins

pAds Ads

0

20

40

60

80

100

0

20

40

60

BMI<25 BMI>25

80

*

Ser

um C

AM

P (

ng/m

l)

CAMP

RETN

LEP

FABP4

COL4A1

LFD HFD

CAMP

ApoAI

CAMP

FABP40

20

40

60

Rel

ativ

e C

amp

mR

NA

abu

ndan

ce

CtrlAd-difM

*

CAMP/LipidORO

Mou

se A

T IgG ctrl

Hum

an A

T LL37Bright field IgG ctrl

CAMP / Lipid

pAds AdAds3T

3-L1

Ctrl SA0

5

10

15

20

Fol

d In

crea

se i

n C

amp

mR

NA

exp

ress

ion **

Ctrl SA

***

Fig. 3. Adipocytes produce cathelicidin antimicrobial peptide. (A) 3T3-L1preadipocytes (pAds) were differentiated to adipocytes (Ads). Expression ofmRNA for a-defensins, b-defensins, and cathelicidin (Camp) is shown (n = 3cultures/group). (Inset)Oil-Red-O staining. (B)Western-blottingof conditionedmedia from differentiating 3T3-L1 cells. (C) Staining of 3T3-L1 cells or mouseand human subcutaneous adipose tissue (AT) by using lipid dye (Bodipy;green), mouse CAMP (red),Oil-red-O, or the human cathelicidin peptide LL37(red) as indicated. Nuclei are counterstainedwithDAPI. Scale bar, 5 mm(3T3-L1)or 50 mm(adipose tissue). (D) C57BL6mice fed with low-fat diet (LFD) or high-fat diet (HFD) for 4 weeks, and mouse serum was collected for Westernblotting analysis by using CAMP and ApoAI antibodies. (E) Human serum

cathelicidin measured in overweight (/BMI >25) or normal-weight human sub-jects (n = 10 subjects/group). (F) Camp mRNA expression levels were exam-ined by quantitative reverse transcription PCR (RT-PCR) in 3T3L1 preadipocytestreated with or without adipocyte differentiation (Ad-difM)medium and withS. aureus–conditioned medium (SA-CM) or UV-killed S. aureus (SA). (G) Pri-mary cultures of dermal cells isolated fromCtrl orS. aureus–injected skin weretreated with Ad-difM. Relative Camp mRNA fold induction at day 2 comparedwith undifferentiated control were examined bymeans of quatitative (RT-PCR)(left); CAMP and FABP4 protein were detected in conditioned medium bymeans of Western blot (right). All error bars indicate mean T SEM; *P < 0.05,**P < 0.01, ***P < 0.001 (t test).

RESEARCH | REPORTS

(23). Immunostaining of 3T3-L1 cells showedcathelicidin appearing simultaneously with lipiddroplets in the cytoplasm of Ads (Fig. 3C) andlocalized in the early endosome, but not in thelysosome (fig. S4J).To verify that cathelicidin was also produced

by human adipocytes, we evaluated cultured pri-mary humanpreadipocyteswith reagents directedagainst human cathelicidin. The CAMP gene wasstrongly induced during differentiation (fig. S5A),and Western blot showed bands consistent withthe cathelicidin precursor protein hCAP18 and ashorter peptide (fig. S5B). This peptide appearedlarger than LL-37, the mature peptide form ofcathelicidin detected from human neutrophils(24), thus suggesting adipocytes may have an en-zymatic processing system differing from thatknown in other cell types (25).Cathelicidin mRNA was detected in the sub-

cutaneous fat pad from 1- to ~2-week-old mice(fig. S6, A and B), a stage in which the tissue isundergoing early adipogenesis (26). CampmRNAwas more abundant in white adipose tissue thanbrown adipose tissue (fig. S6C). Cathelicidin pro-tein immunostaining was detected in fat tissuefrom mice and humans (Fig. 3C) and was local-ized in the early endosomeof adipocytes (fig. S6D).In a mouse model of diet-induced obesity, CampmRNA measured with quantitative polymerasechain reaction (PCR) and cathelicidin proteinlevels estimated by means of Western blot (fig.S7, A and B) were more abundant in the subcu-taneous fat pad from mice fed high-fat diet ascompared with low-fat diet controls. High-fatdiet increased both bone marrow adiposity andcathelicidin protein expression in bone marrowadipocytes (fig. S7C). Serum cathelicidin levelswere also elevated in the high-fat diet group(Fig. 3D and fig. S7, D and E), similar to sam-

ples from overweight humans [body mass index(BMI) > 25] (Fig. 3E). The level of cathelicidindetected with enzyme-linked immunosorbentassay in serum (~50 ng/ml) was well below theamount of LL-37 necessary to inhibit S. aureusgrowth, which is typically above 2 to 5mg/ml (27).Thus, the high local expression of cathelicidinis more likely to provide antimicrobial functionthan is the small increase in systemic cathelicidinfound in obesity.Cathelicidin mRNA production during 3T3-L1

differentiation was significantly enhanced whenS. aureus–conditioned medium or ultraviolet(UV)–killed S. aureus was added to adipocytedifferentiation medium (Fig. 3F), revealing thatS. aureus directly enhances AMP production dur-ing adipocyte differentiation. Similar to 3T3-L1cells, dermal cells isolated from S. aureus–infectedskin, which were shown in Fig. 1E to exhibit highadipogenic potential, produced CampmRNA andabundant cathelicidin protein preceding the pro-duction of FABP4 (Fig. 3G).We next examined cathelicidin production

during S. aureus infection by immunostaininginfected mouse skin. Cathelicidin protein andmRNA were strongly induced in the fat layer atthe site of infection (Fig. 4A and fig. S8A).Cathelicidin protein colocalizedwith expression ofthe adiponectin reporter b-Gal in AdipoQ-cre;R26Rmice (fig. S8B) and COL4A1 (fig. S8, C and D),showing that cathelicidin was expressed inadipocytes in vivo during S. aureus infection,correlating with the increase in dermal adipo-cytes (Fig. 1) and its expression during earlyadipogenesis (Fig. 3B).The precursor (~18 kD) and a peptide form

(5 to ~10 kD) of cathelicidin were detected indifferentiated 3T3-L1 adipocytes (Fig. 4B) and inthe cell lysate of mouse adipose tissue (Fig. S9A).

Similarly to cultured human adipocytes, themouseadipocyte-derived cathelicidin peptide was slight-ly larger than mature cathelicidin-related anti-microbial peptide (CRAMP) derived frommouseneutrophils. We tested conditionedmedium fromdifferentiated 3T3-L1 adipocytes and observedthat conditionedmedium that contained secretedcathelicidin potently inhibited the growth ofS. aureus, whereas conditioned medium frompAds that lacked cathelicidin did not (Fig. 4C).Brefeldin-A treatment blocked cathelicidin se-cretion from adipocytes and abolished the anti-microbial activity of the conditioned mediumderived from adipocytes (fig. S9B). Furthermore,extracts from wild-type mouse adipose tissue in-hibited growth of S. aureus (Fig. 4D and fig. S9C)and Pseudomonas aeruginosa (Fig. S9D), butadipose tissue extracts from Camp−/− mice didnot possess antimicrobial properties. Camp−/−

mice did not differ from wild-type in their ca-pacity to develop metabolic syndrome after20 weeks of high-fat diet (fig. S10).Mice in which adipose expansion was inhib-

ited showed reduced defense against S. aureusinfection, and this correlated with reducedcathelicidin. BADGE treatment of 3T3-L1 dur-ing differentiation inhibited Pparg and CebpamRNA and blocked Camp (fig. S11A). The ex-pression of cathelicidin was less in mice treatedwith BADGE (fig. S11B) and in Zfp423nur12mice(Fig. 4, E and F), although the expression ofcathelicidin seen in cells recruited to the site ofinfection was maintained. Zfp423nur12 did notinhibit neutrophil recruitment as seen by a sim-ilar increase in abundance of the neutrophilmarker (Mpo) after infection (fig. S11D). As ex-pected, Camp−/− mice showed increased sus-ceptibility to infection (3). However, Camp−/−

mice treated with BADGE or GW9662 showed

70 2 JANUARY 2015 • VOL 347 ISSUE 6217 sciencemag.org SCIENCE

Fig. 4. Adipocytes kill bacteria by producingcathelicidin. (A) CAMP immunostaining (left) ofadipose in mice infected with S. aureus. Scale bar,100 mm. (B)Western-blotting of CAMP fromwhole-cell extracts (WCE) or conditioned medium (CM)from 3T3-L1 pAds or Ads. Red arrow indicatesmature peptide form, and red asterisks indicateprecursor forms of cathelicidin. Synthetic CRAMPpeptide is shown as standard for the processedform of CAMP in neutrophils. (C) Growth curve ofS. aureus in Ctrl medium or conditioned medium(CM) from pAd or Ad as indicated. (Inset) CAMPimmunoblot in conditioned medium. (D) S. aureusgrowth in media alone (Ctrl) or with addition ofprotein extracted from subcutaneous adipose tissuefrom Camp+/+ or Camp−/− mice (n = 3). (E) CAMPexpression is suppressed in adipose in Zfp423-deficient mice. Scale bars, 100 mm. (F) CampmRNAexpression in S. aureus– or PBS-injected Zfp423+/+

and Zfp423nur12 mice (n = 5 mice/group). (G)Camp+/+ or Camp−/− mice were pretreated withBADGE or dimethyl sulfoxide (ctrl) before infect-ing with S. aureus and S. aureus survival in skinmeasured 3 days later (n = 7 to ~10 mice/group).All error bars indicate mean T SEM. *P < 0.05, **P <0.01, ***P < 0.001 (one-way analysis of variance). Red brackets indicate subcutaneous adipose layer.

0.10

0.05

0.00

0.25

0.20

0.15

O.D

. 595

Hours

MediumMedium w SApAd-CM w SAAd-CM w SA

CAMP

pAd Ad

* **6

4

2

0

S. a

ureu

s(1

06 C

FU

/ml)

0

1x105

2x105

3x105

Ctrl BADGE

S. a

ureu

s(C

FU

/gm

)

*

**

n.s.n.s.

Ctrl SA0

2

4

6

8

Zfp423+/+

Zfp423nur12

***

SA

PBS

CAMP

20

15

10

5

CR

AM

P

MW (kDa)

WCE

CA

MP

GAPDH

CM

**

CAMP/PLIN

WT

Zfp

423nu

r12

CAMP

Rel

ativ

e C

amp

mR

NA

abu

ndan

ce

RESEARCH | REPORTS

no further increase in infection (Fig. 4G andfig. S12), suggesting that the effects of BADGEon cathelicidin production from adipocytes wasresponsible for the increase in S. aureus in-fection in vivo.These results show that a local increase in

subcutaneous adipocytes is an important hostdefense response against skin infection. Thisobservation is consistent with prior observa-tions that adipocytes secrete a variety of bio-active adipokines and cytokines that mediateimmune responses after injury (28) and nowshows that adipocytes produce an AMP that candirectly kill bacteria. Local expansion of dermalfat produces cathelicidin in response to infec-tion, but this response appears to decline asadipocytes mature. The expansion of dermal fatin response to infection may also indirectly ben-efit immune defense by influencing other pro-cesses such as neutrophil oxidative burst, thusfurther amplifying the importance of the sub-cutaneous preadipocyte pool in preventing in-fections. Defective AMP production by matureadipocytes may explain observations of elevatedsusceptibility to infection during obesity and in-sulin resistance (29). Cathelicidin has also beenshown to be proinflammatory (30). Therefore,the production of cathelicidin by adipocytes mayalso participate in the chronic, low-level inflam-mation observed in obesity (28).

REFERENCES AND NOTES

1. F. O. Nestle, P. Di Meglio, J. Z. Qin, B. J. Nickoloff, Nat. Rev.Immunol. 9, 679–691 (2009).

2. R. L. Gallo, L. V. Hooper, Nat. Rev. Immunol. 12, 503–516(2012).

3. V. Nizet et al., Nature 414, 454–457 (2001).4. P. Y. Ong et al., N. Engl. J. Med. 347, 1151–1160 (2002).5. K. Pütsep, G. Carlsson, H. G. Boman, M. Andersson, Lancet

360, 1144–1149 (2002).6. L. S. Miller, J. S. Cho, Nat. Rev. Immunol. 11, 505–518

(2011).7. J. M. Musser et al., Proc. Natl. Acad. Sci. U.S.A. 87, 225–229

(1990).8. J. Eguchi et al., Cell Metab. 13, 249–259 (2011).9. H. S. Sul, Mol. Endocrinol. 23, 1717–1725 (2009).10. C. S. Hudak et al., Cell Reports 8, 678–687 (2014).11. R. K. Gupta et al., Nature 464, 619–623 (2010).12. T. Tanaka, N. Yoshida, T. Kishimoto, S. Akira, EMBO J. 16,

7432–7443 (1997).13. M. I. Lefterova et al., Genes Dev. 22, 2941–2952 (2008).14. E. Festa et al., Cell 146, 761–771 (2011).15. W. A. Alcaraz et al., Proc. Natl. Acad. Sci. U.S.A. 103,

19424–19429 (2006).16. L. E. Cheng, J. Zhang, R. R. Reed, Dev. Biol. 307, 43–52

(2007).17. R. K. Gupta et al., Cell Metab. 15, 230–239 (2012).18. R. R. Driskell et al., Nature 504, 277–281 (2013).19. O. Naveiras et al., Nature 460, 259–263 (2009).20. B. A. Schmidt, V. Horsley, Development 140, 1517–1527

(2013).21. H. Green, O. Kehinde, Cell 5, 19–27 (1975).22. R. A. Dorschner et al., J. Invest. Dermatol. 117, 91–97

(2001).23. T. Sillat et al., J. Cell. Mol. Med. 16, 1485–1495 (2012).24. O. E. Sørensen et al., Blood 97, 3951–3959 (2001).25. O. E. Sørensen et al., J. Biol. Chem. 278, 28540–28546

(2003).26. N. Kawaguchi et al., Am. J. Pathol. 160, 1895–1903

(2002).27. H. G. Boman, J. Intern. Med. 254, 197–215 (2003).28. A. Schäffler, J. Schölmerich, Trends Immunol. 31, 228–235

(2010).29. W. P. Cawthorn, E. L. Scheller, O. A. MacDougald, J. Lipid Res.

53, 227–246 (2012).30. K. Yamasaki et al., Nat. Med. 13, 975–980 (2007).

ACKNOWLEDGMENTS

This study was supported by NIH R01 AI083358, R01AI052453,and AR052728 (R.L.G.). Human serum collections were funded byThe Atopic Dermatitis Research Network (HHSN272201000020C).S.P.B. was supported by the NIH (DK096828). M.V.P. is supportedby the Edward Mallinckrodt Jr. Foundation Research Grant, theDermatology Foundation Research Grant, and NIH NIAMSR01-AR067273; C.F.G.-J. is supported by the NIH MBRS-IMSDtraining grant (GM055246) and NSF Graduate Research Fellowshipnumber DGE-1321846; and R.R. is supported by CaliforniaInstitute for Regenerative Medicine training grant (TG2-01152).The authors declare no competing financial interests. We thankT. Nakatuji for advice relating to bacterial infections and C. Aguilerafor mouse technical expertise, UCSD Bio-Core facility for reagents,UCSD mouse hematology core laboratory for serum studies, and

J. M. Olefsky for comments on the manuscript. Zfp423-GFP miceare available from R. Gupta under a material transfer agreementwith Dana-Farber Cancer Institute. All the data reported in thismanuscript are presented in the main paper and in thesupplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/347/6217/67/suppl/DC1Materials and MethodsFigs. S1 to S12Table S1References (31–37)

8 September 2014; accepted 21 November 201410.1126/science.1260972

VIRUS STRUCTURE

Structure and inhibition of EV-D68,a virus that causes respiratoryillness in childrenYue Liu,1 Ju Sheng,1 Andrei Fokine,1 Geng Meng,1 Woong-Hee Shin,1 Feng Long,1

Richard J. Kuhn,1 Daisuke Kihara,1,2 Michael G. Rossmann1*

Enterovirus D68 (EV-D68) is a member of Picornaviridae and is a causative agent ofrecent outbreaks of respiratory illness in children in the United States. We report here thecrystal structures of EV-D68 and its complex with pleconaril, a capsid-binding compoundthat had been developed as an anti-rhinovirus drug. The hydrophobic drug-binding pocketin viral protein 1 contained density that is consistent with a fatty acid of about 10 carbonatoms. This density could be displaced by pleconaril. We also showed that pleconarilinhibits EV-D68 at a half-maximal effective concentration of 430 nanomolar and might,therefore, be a possible drug candidate to alleviate EV-D68 outbreaks.

Picornaviruses constitute a large family ofsmall icosahedral viruseswitha singlepositive-stranded RNA genome and an externaldiameter of about 300 Å. The Enterovirusgenus includesmedically important human

pathogens, such as human rhinoviruses (HRVs),polioviruses (PVs) and coxsackieviruses (CVs)(table S1) (1, 2). Many of these enteroviruses (EVs)have been well characterized structurally andfunctionally (3–10). However, the species EV-Dremains poorly characterized.An upsurge of EV-D68 cases in the past few

years has shown clusters of infections worldwide(11). In August 2014, an outbreak of mild to se-vere respiratory illnesses occurred among thou-sands of young children in the United States, ofwhich 1116 cases have been confirmed to be causedby EV-D68. This virus has also been associatedwith occasional neurological infections (12). Al-though EV-D68 has emerged as a considerableglobal public health threat, there is no availablevaccine or effective antiviral treatment.The capsids of EVs consist of 60 copies of each

of four different viral proteins: VP1, VP2, VP3,

and VP4 (Fig. 1A). Of these, VP1 (about 300amino acids), VP2 (about 260 amino acids), andVP3 (about 240 amino acids) each have a “jellyroll” fold arranged in the capsid with pseudo T =3 icosahedral symmetry, where T represents thetriangulation number (13). Their organization inthe capsid is similar to that of the T = 3 RNAplant viruses, except that the three subunits re-lated by quasi-threefold symmetry have differentamino acid sequences in picornaviruses (6, 10).Each roughly 70–amino–acids-long VP4 mole-cule forms an extended peptide on the internalsurface of the capsid shell. The jelly roll foldconsists of two antiparallel b sheets, which faceeach other to form a b barrel with a hydrophobicinterior (Fig. 1B).EVs have a deep surface depression (“canyon”)

circulating around each of the 12 pentamericvertices (Fig. 1A). The canyonwas predicted to bethe site of receptor binding, because the aminoacids outside the canyon that form the externalsurface of the virus were more exposed and wereshown to be accessible to neutralizing antibodies(Fig. 1A) (10, 14). The virus thus could remainfaithful to a specific receptor molecule that bindsinto the canyonwhile evading the host’s immunesystem (10, 15). The prediction that the canyonwould be the site of binding to cellular receptormoleculeswas subsequently confirmed for numer-ous different EVs (16, 17). All of the receptor

SCIENCE sciencemag.org 2 JANUARY 2015 • VOL 347 ISSUE 6217 71

1Department of Biological Sciences, Hockmeyer Hall ofStructural Biology, 240 South Martin Jischke Drive, PurdueUniversity, West Lafayette, IN 47907, USA. 2Department ofComputer Science, 305 North University Street, PurdueUniversity, West Lafayette, IN 47907, USA.*Corresponding author. E-mail: mr@purdue.edu

RESEARCH | REPORTS