L E T T E R S

CD1d is a major histocompatibility complex (MHC) class I–related molecule that functions in glycolipid antigenpresentation to distinct subsets of T cells that express naturalkiller receptors and an invariant T-cell receptor-α chain(invariant NKT cells)1–3. The acquisition of glycolipid antigensby CD1d occurs, in part, in endosomes through the function ofresident lipid transfer proteins, namely saposins4–10. Here weshow that microsomal triglyceride transfer protein (MTP), aprotein that resides in the endoplasmic reticulum ofhepatocytes and intestinal epithelial cells (IECs) and isessential for lipidation of apolipoprotein B11,12, associateswith CD1d in hepatocytes. Hepatocytes from animals in whichMttp (the gene encoding MTP) has been conditionally deleted,and IECs in which Mttp gene products have been silenced, areunable to activate invariant NKT cells. Conditional deletion ofthe Mttp gene in hepatocytes is associated with aredistribution of CD1d expression, and Mttp-deleted mice areresistant to immunopathologies associated with invariant NKTcell–mediated hepatitis and colitis. These studies indicate thatthe CD1d-regulating function of MTP in the endoplasmicreticulum is complementary to that of the saposins inendosomes in vivo.

CD1d-restricted T cells regulate many immune disorders throughthe ability of CD1d to present lipid antigens to invariant NKTcells13–15. Because CD1d assembly occurs within the endoplasmicreticulum, in a manner similar to but distinct from that associatedwith other MHC class I–related molecules, it has been predicted thatthese in vivo functions of CD1d involve acquisition of glycolipidantigens in this cellular locale5,16–18. MTP is an endoplasmic reticulum–resident lipid transfer protein involved in the formationof apolipoprotein B–containing chylomicrons and very-low-densitylipoproteins in hepatocytes and IECs, both of which are known toexpress CD1d19–22. Given these properties, as well as the ability ofMTP to transfer glycolipids such as phosphatidylinositol20 that are

known to bind CD1d18, we reasoned that MTP could also beinvolved in CD1d function.

To confirm a direct relationship between CD1d and MTP, we firstassessed whether CD1d and MTP associate with each other in hepato-cytes. We immunoprecipitated protein lysates of hepatocytes fromC57BL/6 (B6) mice with a monoclonal antibody to either CD1d orMHC class I (H–2Db). We resolved the immunoprecipitates by SDS-PAGE, followed by western blotting for the presence of MTP. Wedetected a specific band of 97 kDa, consistent with MTP, in the CD1d,but not the MHC class I, immunoprecipitates (Fig. 1a).

To further substantiate a direct biochemical relationship betweenCD1d and MTP, we examined CD1d expression and function in aconditional MTP-deficient mouse model. The mice, which had a‘floxed’ Mttp gene (Mttpfl/fl), had been intercrossed with miceexpressing Cre recombinase under the control of the Mx1 promoter(Mx1Cre)21,23. The Mx1 promoter is induced by interferon-inducingsubstances, such as polyinosinic-polycytidylic acid (pIpC)21,23,24,such that pIpC treatment leads to deletion of Mttp (Mttp–/–) prima-rily within the liver, spleen and intestine24. As previouslyreported21,23, treatment of Mttpfl/flMx1Cre mice with pIpC causedprogressive hepatic steatosis (Supplementary Fig. 1 online), adecrease in serum triglyceride and cholesterol (Supplementary Table 1 online), and a decrease in MTP mRNA and protein(Supplementary Fig. 1) in the liver and colon. Consistent with this,the specific biochemical association observed between CD1d andMTP in hepatocytes of B6 mice was also observed in Mttpfl/flMx1Cremice before, but not after, pIpC treatment and consequent deletionof Mttp (Fig. 1b).

Given the biochemical association observed between CD1d andMTP (Fig. 1a,b), we surmised that the MTP deficiency induced byMx-1Cre activity would affect CD1d expression, function or both inhepatocytes in vivo. We therefore examined the expression of CD1d inthe liver of Mttpfl/flMx1Cre mice before and after pIpC treatment.Before pIpC treatment, CD1d was present diffusely throughout hepa-tocytes, with a membranous staining pattern, in both B6 and

1Gastroenterology Division, Department of Medicine, Brigham and Women’s Hospital; 2Program in Immunology; 3Department of Pathology, Brigham and Women’sHospital; and 4Department of Anesthesia, Brigham and Women’s Hospital; Harvard Medical School, Boston, Massachusetts 02115, USA. 5Gladstone Institute ofCardiovascular Disease, University of California-San Francisco, San Francisco, California 94110, USA. 6Hematology-Oncology Division, Beth Israel DeaconessMedical Center, Harvard Medical School, Boston, Massachusetts 02215, USA. 7These authors contributed equally to this work. Correspondence should be addressedto R.S.B. (rblumberg@partners.org).

Published online 25 April 2004; doi:10.1038/nm1043

CD1d function is regulated by microsomal triglyceridetransfer proteinSuzana Brozovic1,7, Takashi Nagaishi1,7, Masaru Yoshida1, Stephanie Betz2, Azucena Salas1, Daohong Chen1,Arthur Kaser1, Jonathan Glickman3, Timothy Kuo1, Alicia Little1, Jamin Morrison1, Nadia Corazza1,Jin Yong Kim1, Sean P Colgan4, Stephen G Young5, Mark Exley6 & Richard S Blumberg1

NATURE MEDICINE VOLUME 10 | NUMBER 5 | MAY 2004 535

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

L E T T E R S

Mttpfl/flMx1Cre mice (Fig. 1c). After pIpC treatment, CD1d in hepa-tocytes from Mttp–/– mice, but not B6 mice, showed a prominent perinuclear staining pattern that overlapped significantly with anendoplasmic reticulum marker, and cell surface expression of CD1dwas reduced (Fig. 1c). pIpC treatment of B6 or Mttpfl/flMx1Cre micedid not affect the distribution of MHC class I expression (Fig. 1c).Consistent with this, after Mttp deletion, cell surface expression ofCD1d, but not MHC class I, was significantly decreased as assessed byflow cytometry (Fig. 1d). Taken together, these studies indicate thatthe intracellular trafficking behavior of CD1d is selectively influencedby the presence of MTP in hepatocytes in vivo, such that CD1d isreduced on the cell surface and increased in the endoplasmic reticu-lum in the absence of MTP. This suggests abnormal entry of CD1dinto the secretory pathway.

We next examined whether Mttp deletion also affects the ability ofhepatocyte CD1d to activate invariant NKT cells. Hepatocytesobtained from either pIpC-treated or untreated Mttpfl/flMx1Cre micewere cocultured with DN32.D3, a mouse CD1d-restricted invariantNKT cell hybridoma. Hepatocytes from untreated Mttpfl/flMx1Cremice stimulated DN32.D3 cells to secrete interleukin-2 (IL-2) in thepresence, but not the absence, of α-galactosylceramide (α-GalCer;Fig. 2a). Hepatocytes from Mttp–/– mice, in contrast, stimulated littleIL-2 production by DN32.D3 cells even in the presence of α-GalCer(Fig. 2a), providing direct evidence linking MTP in hepatocytes to theregulation of CD1d-restricted T-cell responses. This is consistent withthe decreased cell surface expression of CD1d in the absence of MTP.This effect was specific for CD1d, as MHC class II–restricted activationof ovalbumin-specific CD4+ T cells (from OT-II transgenic mice) by

CD11c+ dendritic cells obtained from the liver of Mttpfl/flMx1Cre micewas in fact increased after Mttp deletion (Fig. 2b). The latter result isconsistent with the effects of interferon-α induced by pIpC treatment.CD1d-restricted autoreactivity to hepatocytes was also defective in theabsence of MTP; this affected T cells bearing either an invariant (24.8)or noninvariant (14S.6) TCR-α chain (Fig. 2c).

Based on this inability of hepatocytes to effect CD1d-restrictedpresentation of exogenous and endogenous antigens in the absence ofMTP expression, we examined the response to a hepatocyte-mediated,CD1d-dependent response in vivo25. As expected, administration ofα-GalCer before pIpC treatment induced hepatitis in both B6 (Fig. 2d)and Mttpfl/flMx1Cre (Fig. 2e) mice. The absence of inflammation withpIpC treatment alone is consistent with the lack of hepatic inflamma-tion in either the absence of Mttp21 or the presence of Mx1 activityinduced by interferons (Fig. 2d,e)24.The lack of MTP function induceshepatic steatosis23. However, in contrast to what might be predicted,Mttp–/– mice were nearly completely protected from the inflammatoryeffects of α-GalCer administration at either 1 or 10 d after completionof the pIpC treatment regimen (Fig. 2e). The latter was evident fromthe absence of elevated transaminase levels (Fig. 2e), macroscopic livernecrosis (Fig. 2f) or microscopic evidence of hepatitis (Fig. 2g) in theMttpfl/flMx1Cre mice that received α-GalCer with prior pIpC treat-ment. In contrast, B6 mice that received α-GalCer both 1 and 10 dafter pIpC treatment had severe hepatitis (Fig. 2d). Thus, in theabsence of MTP, the liver is protected from CD1d-restricted andinvariant NKT cell–mediated hepatocyte injury.

To extend this observation to other MTP-expressing cell types, weexamined the relationship between MTP and CD1d in IECs. As

536 VOLUME 10 | NUMBER 5 | MAY 2004 NATURE MEDICINE

a b

c

d

Figure 1 CD1d expression after Mttp deletion. (a) Hepatocyte lysates from B6 mice were immunoprecipitated with the indicated antibodies (8F12, anti-H-2D; 1B1, anti-CD1d), followed by immunoblotting with either an antibody to MTP (left) or a rabbit antiserum to MHC class I (right). (b) Hepatocyte lysatesfrom B6 or Mttpfl/flMx1Cre mice before (–) or after (+) pIpC treatment were subjected to same analysis as in a. (c) Livers from B6 or Mttpfl/flMx1Cre micewere treated (+) or not (–) with pIpC, and stained for CD1d (green), H–2Kb (green), endoplasmic reticulum (red) and nuclei (blue). 1B1, anti-CD1d; CTKb,anti-H-2Kb. White arrowhead (bottom right) indicates colocalization of CD1d and endoplasmic reticulum marker. (d) Hepatocytes from Mttpfl/flMx1Cremice before (–) or after (+) pIpC treatment were stained for CD1d and MHC class I (H–2Db) expression after gating on live cells.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

L E T T E R S

predicted11, MTP was expressed by a mouse IEC line, MODE-K (Fig. 3a), which also functionally expresses CD1d26. Gene silencing ofMttp in MODE-K cells with Mttp-specific small interfering RNA(siRNA) oligomers caused a significant reduction in Mttp mRNA (Fig. 3b,c). In the absence of an exogenous source of the glycolipidantigen α-GalCer, MODE-K cells were unable to stimulate IL-2 pro-duction by DN32.D3 cells (Fig. 3d). In contrast, in the presence ofα-GalCer, MODE-K cells stimulated significant IL-2 production byDN32.D3 cells (Fig. 3d). Mttp silencing, but not silencing of an irrele-vant gene target, inhibited IL-2 production (Fig. 3d). These studiesindicate that, as shown above for hepatocytes, MTP regulates the ability of IECs to effect CD1d-restricted antigen presentation.

Given this in vitro evidence for a functional link between MTPand CD1d in IECs, and given the fact that Mx1 promoter activity isactive in intestine cells24 including IECs from the colon(Supplementary Fig. 1), we examined the effect of Mttp deletion onthe clinicopathologic outcome of oxazolone-induced colitis. Thismodel of colitis has recently been shown to be mediated by CD1dand CD1d-restricted invariant NKT cells27. Mttpfl/flMx1Cre mice(Fig. 4a), as well as B6 mice (Fig. 4b), experienced severe colitis inassociation with administration of the hapten oxazolone, as mani-fested by profound weight loss (Fig. 4a,b)and pathological evidence of mucosal ulcer-ations and infiltration of intestinal tissuesby inflammatory cells (Fig. 4c). In contrast,Mttp–/– mice experienced minimal weightloss, which was identical to that experiencedby Mttpfl/flMx1Cre mice exposed to theethanol control (Fig. 4a). Mttp–/– mice alsoshowed little histologic evidence of colitis(Fig. 4c). A semiquantitative estimate ofcolitis severity confirmed these results (Fig. 4d). Colitis was not ameliorated in B6mice exposed to oxazolone, when pIpC wasadministered on the same schedule used for

the Mttpfl/flMx1Cre mice (Fig. 4b,c,e). These results indicate that theprotection observed in the Mttp–/– mice was not simply an effect ofpIpC treatment.

In summary, these studies reveal that MTP regulates the ability ofCD1d-bearing cell types to effect CD1d-restricted antigen presenta-tion. This property of MTP has pathophysiological relevance, as abro-gation of MTP function leads to protection from CD1d-mediatedimmunopathology involving cell types—such as hepatocytes andIECs—that contain MTP and express CD1d. Whether other CD1d-bearing cell types, such as hematopoietic cells, express MTPand are similarly regulated remains to be established, but this wouldsignificantly expand the implications of the role of MTP in regulatingCD1d-restricted antigen presentation. These studies suggest that,similar to the relationship between MTP and apolipoprotein B28, thefunction of MTP in the lipidation and maturation of CD1d in theendoplasmic reticulum complements that of the saposins withinendolysosomal compartments8–10. This role of MTP in regulating theacquisition of glycolipid antigens is likely to be involved in the normalfunction of CD1d in vivo, as shown here. We predict that blockade ofMTP function will be of therapeutic benefit in diseases mediated byCD1d and related pathways.

NATURE MEDICINE VOLUME 10 | NUMBER 5 | MAY 2004 537

0

20

40

60

80

100

120

140

AL

T (

U/l)

pIpCα-GalCer + +_ _

_ _+ ++ +_ _

_ _+ +

0

20

40

60

80

100

120

140

160

IL-2

(p

g/m

l)

14S.614S.6 + 3C1114S.6 + rat IgM24.8

ND ND

pIpCα-GalCer + +_ _

_ _+ +

*

0

OVAVehicle

IL-2

(pg

/ml)

50

100

150

200

pIpC _ +

**

pIpC _ +

**

*

0

20

40

60

80

100

120

140

AL

T (

U/l)

pIpCα-GalCer

** ***

+ +_ __ _+ +

+ +_ __ _+ +

Day 1 Day 10Day 1 Day 10

IL-2

(pg

/ml)

0

1,200

600

900

300

a c

b

d e

f g

Figure 2 Deletion of Mttp inhibits CD1d-restricted presentation by hepatocytes and α-GalCer-induced hepatitis. (a) Hepatocytepresentation of α-GalCer to DN32.D3 cells fromMttpfl/flMx1Cre mice treated (+) or not (–) withpIpC, as assessed by IL-2 production. ND, notdetectable. *, P < 0.001. (b) Presentation ofovalbumin (OVA) by liver CD11c+ cells fromMttpfl/flMx1Cre mice, treated (+) or not (–) withpIpC, to CD4+ splenocytes from OT-II mice, asassessed by IL-2 production. **, P < 0.005. (c) Response of autoreactive mouse CD1d-restricted T-cell hybridomas 14S.6 or 24.8 tohepatocytes from Mttpfl/flMx1Cre mice treated (+)or not (–) with pIpC, as assessed by IL-2production with or without blocking antibody toCD1d (3C11) or rat IgM control. *, P < 0.05. (d) α-GalCer-induced hepatitis, as assessed byaminoleucine transferase (ALT) levels in C57BL/6mice before (–) or after (+) pIpC treatment. (e) α-GalCer-induced hepatitis in Mttpfl/flMx1Cremice, as described in d. **, P < 0.005; ***, P < 0.01. (f) Gross morphology of liversdescribed in d and e. White arrowheads indicatefoci of liver necrosis. (g) Histopathology of liversdescribed in d and e. White arrowheads indicatefoci of coagulative necrosis (×200).

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

L E T T E R S

METHODSCells. MODE-K, a mouse IEC line, has been previously described26. DN32.D3,a mouse Vα14Jα18 invariant TCR-positive T-cell hybridoma, was provided byA. Bendelac (University of Chicago)29. The autoreactive mouse Vα14Jα281invariant TCR-positive hybridoma 24.8 and the noninvariant TCR-positivehybridoma 14S.6 were provided by S. Behar (Brigham and Women’sHospital)2,16.

Animals. Mice with a ‘floxed’ Mttp allele (Mttpfl/fl) were bred with mice trans-genic for Mx1 promoter–driven Cre recombinase, as previously described21. Allanimal experimentation was done in accordance with institutional guidelinesand the review board of Harvard Medical School, which granted permission forthis study.

Assessment of hepatic and plasma lipids after Mttp deletion. FemaleMttpfl/flMx1Cre, Mttpfl/fl or wild-type C57BL/6 (B6) mice (6–8 weeks old) were

treated intraperitoneally with either PBS alone or 500 µg pIpC (Sigma) in PBS,four times daily every other day. Serum was obtained from peripheral blood formeasurement of total cholesterol and triglycerides, and liver was assessed forthe presence of intrahepatic lipid accumulation, as previously described21.

Cell isolation, antigen presentation assays and flow cytometry. Colonic epithe-lial cells were isolated as previously described26. Primary hepatocytes wereisolated as previously described23,25 and stained with FITC-conjugated anti-

bodies to CD1d (1B1, PharMingen), or with biotin-conjugated antibody toH–2Db (Caltag Laboratories) and phycoerythrin-conjugated streptavidin(PharMingen), in the presence of Via-Probe (PharMingen). Staining was ana-lyzed with a FACSort flow cytometer (Becton Dickinson). Primary hepatocytes(1 × 105) were loaded overnight with either vehicle or 100 ng/ml α-GalCer, fol-lowed by the addition of 5 × 104 DN32.D3 cells. Autoreactive CD1d-restricted T cells were tested in the absence or presence of antibody to mouse CD1d(3C11)22. CD11c+ cells (1 × 105) were purified from liver using CD11c

538 VOLUME 10 | NUMBER 5 | MAY 2004 NATURE MEDICINE

0

100

200

300

400

500

IL-2

(pg

/ml)

0

0.2

0.4

0.6

0.8

1

Mttp

/Act

B

siRNA _ Mock Silenced

*

ND

siRNAα-GalCer ++_

_ Mock Silenced

**

a b c d

Figure 3 Silencing of Mttp in IECs inhibits CD1d-restricted antigen presentation. (a) Assessment by RT-PCR of Mttp transcripts in MODE-K cells,compared with primary hepatocytes from B6 mice, relative to β-actin (ActB) expression. (b) Assessment of Mttp transcripts in MODE-K cells treated withmock or Mttp-specific (silenced) siRNA oligomers. (c) Ratios of Mttp to ActB mRNA for each experimental condition shown in b. *, P < 0.001. (d) Presentation of α-GalCer to DN32.D3 cells by MODE-K cells treated with mock or Mttp-specific (silenced) siRNA oligomers, as assessed by IL-2production. ND, not detectable. **, P < 0.01.

80

90

100

110

0 1 2 3 4Days

0

2

4

6

8

10

12

pIpC + +_ __ _+ +

0

2

4

6

8

10

12

Col

itis

scor

e

Col

itis

scor

e

*

pIpCOxazolone Oxazolone

+ +_ __ _+ +

80

90

100

110

0 1 2 3 4

% b

od

y w

eig

ht

% b

od

y w

eig

ht

Days

***

**

a b c

d e

Figure 4 Mttp deletion inhibits oxazolone-induced colitis. (a) Wasting, as defined by percentage of body weight from baseline, of Mttpfl/flMx1Cre micesubjected to oxazolone-induced colitis. Mice were treated with either vehicle (�,�) or pIpC (�,�) before skin was painted with oxazolone. This wasfollowed by rectal challenge with either 50% ethanol alone (control; �,�) or 50% ethanol containing oxazolone (�,�). n = 12 mice per group. *, P < 0.005; **, P < 0.001 for Mttpfl/flMx1Cre mice with or without pIpC treatment and oxazolone. (b) Results for C57BL/6 mice treated as described in a(n = 12 per group). (c) H&E-stained microscopic images of proximal colon from groups in a and b. Black arrowhead (lower right) indicates mildinflammation with minimal epithelial cell proliferation (×100). (d) Quantitative histopathologic assessment of colitis activity in a. *, P < 0.001. (e) Quantitative histopathologic assessment of colitis activity in b.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

L E T T E R S

antibody–labeled magnetic beads and MACS separation columns (MiltenyiBiotec). Purified cells were incubated with 5 × 105 CD4+ splenocytes from OT-II mice in the presence of 100 µg/ml ovalbumin or PBS. Mouse IL-2 productionwas assessed by ELISA (OptEIA, BD PharMingen) after 24 h. Western blottingwas done with standard methods using the following specific antibodies: ratmonoclonal antibody to mouse CD1d (1B1; IgG2b; PharMingen); mouse anti-body to mouse MHC class I (H–2D; 8F12; IgG1; PharMingen); mouse antibodyto mouse MTP (IgG2a; BD Biosciences); rabbit antiserum to mouse MHC class I (gift of H. Ploegh, Harvard Medical School); horseradish peroxidase–conjugated goat antibody to mouse IgG2a (Southern BiotechnologyAssociates), rat IgG2b (PharMingen) or mouse IgG2a (PharMingen); or horse-radish peroxidase–conjugated donkey antibody to rabbit IgG (Pierce).

RT-PCR. Total cellular RNA (100 ng) was subjected to reverse transcriptionusing the Advantage RT-for-PCR Kit (Clontech Laboratories). The sense andantisense primers for mouse Mttp were 5′-GGACTTTTTGGATTTCAAAAGT-GAC-3′ and 5′-GGAGAAACGGTCATAATTGTG-3′, respectively. cDNA bandswere normalized to the RT-PCR products of β-actin using the primers 5′-GTGGGCCGCTCTAGGCACCAA-3′ and 5′-CTCTTTGATGTCACGCAC-GATTTC-3′.

Immunofluorescence and confocal microscopy. Sections of liver frozen in OCTcompound (Sakura Finetek USA) were incubated with rat monoclonal antibodyto mouse CD1d (1B1) and rabbit antiserum to endoplasmic reticulum (provided by D. Meyer, University of California, Los Angeles). Sections werethen incubated with Alexa 488–conjugated goat antibody to rat IgG and Alexa568–conjugated goat antibody to rabbit IgG (Molecular Probes). For staining ofMHC class I, after blocking endogenous biotin with the Endogenous Biotin-Binding Kit (Molecular Probes), sections were incubated with biotin-conjugatedantibody to mouse H–2Kb MHC (Caltag Laboratories) and the rabbit antiserumto endoplasmic reticulum described above. Sections were then incubated withAlexa 488–conjugated mouse antibody to biotin (Molecular Probes) and Alexa568–conjugated goat antibody to rabbit IgG. Nuclei were stained with TO-PRO-3 (Molecular Probes) and tissues were mounted with ProlongAntifade reagent (Molecular Probes). All images were collected using aMRC1024 laser scanning confocal system (Bio-Rad Laboratories).

α-GalCer-induced hepatitis. Wild-type B6 or Mttpfl/flMx1Cre mice weretreated with pIpC, followed by intraperitoneal injection of α-GalCer (2 µg/mouse) 1 or 10 d later. After 24 h, sera from peripheral blood were sub-jected to aminoleucine transferase or aminoaspartate transferase assays(Stanbio Laboratory). Liver samples were collected for macroscopic and micro-scopic examination after staining with H&E.

Mttp silencing by siRNA. We generated Mttp-specific siRNA duplexes con-sisting of a sense strand (5′-AAGCUCUGGAACUACCAACGAdTdT-3′) andan antisense strand (5′-UCGUUGGUAGUUCCAGAGCUUdTdT-3′;Xeragon). The duplexes recognized nucleotides 480–500 of Mttp (GenBankaccession no. NM_008642). Mock siRNA oligomers (specific for SH2domain–containing protein tyrosine phophatase-1) and transfection proto-cols were previously described30.

Hapten-induced colitis. The hapten-induced colitis model using oxazolone (4-ethoxymethylene-2-phenyl-2-oxazolin-5-one; Sigma) as the hapten wasestablished in B6 or Mttpfl/flMx1Cre mice after pIpC treatment, as previouslydescribed27. Colonic tissue specimens for histological assessment of colitis wereembedded in paraffin, cut into 5-µm-thick sections and stained with H&E. Thestained paraffin sections were evaluated by J.G. for the severity of colitis, as previously described27.

Statistical analysis. Data are expressed as mean ± s.e.m. Statistical significancewas determined by Student t test. P < 0.05 was considered significant.

Note: Supplementary information is available on the Nature Medicine website.

ACKNOWLEDGMENTSR.S.B. was supported by National Institutes of Health grants DK44319, DK53056and DK51362; the Harvard Digestive Diseases Center; and the Broad MedicalResearch Program. A.K. was supported by the Max Kade Foundation. We thank

D. Bailey for excellent technical assistance, A. Bendelac for DN32.D3 cells, S. Beharfor CD1-restricted T-T hybridomas, H. Ploegh for MHC class I–specific antibodies,D. Meyer for endoplasmic reticulum–specific antibodies, and P. Libby, N. Davidsonand A. Bendelac for critical discussions.

COMPETING FINANCIAL INTERESTSThe authors declare that they have no competing financial interests.

Received 8 December 2003; accepted 7 April 2004Published online at http://www.nature.com/naturemedicine/

1. Lantz, O. & Bendelac, A. An invariant T cell receptor α chain is used by a unique sub-set of major histocompatibility complex class I-specific CD4+ and CD8+ T cells inmice and humans. J. Exp. Med. 180, 1097–1106 (1994).

2. Behar, S.M., Podrebarac, T.A., Roy, C.J., Wang, C.R. & Brenner, M.B. Diverse TCRsrecognize murine CD1. J. Immunol. 162, 161–167 (1999).

3. Kronenberg, M. & Gapin, L. The unconventional lifestyle of NKT cells. Nat. Rev.Immunol. 2, 557–568 (2002).

4. De Silva, A.D. et al. Lipid protein interactions: the assembly of CD1d1 with cellularphospholipids occurs in the endoplasmic reticulum. J. Immunol. 168, 723–733(2002).

5. Kang, S.J. & Cresswell, P. Calnexin, calreticulin, and Erp57 cooperate in disulfidebond formation in human CD1d heavy chain. J. Biol. Chem. 277, 44838–44844(2002).

6. Prigozy, T.I. et al. Glycolipid antigen processing for presentation by CD1d molecules.Science 291, 664–667 (2001).

7. Roberts, T.J. et al. Recycling CD1d1 molecules present endogenous antigensprocessed in an endocytic compartment to NKT cells. J. Immunol. 168, 5409–5414(2002).

8. Zhu, D. et al. Editing of CD1d-bound lipid antigens by endosomal lipid transfer pro-teins. Science 303, 523–527 (2004).

9. Kant, S.J. & Cresswell, P. Saposins facilitate CD1d-restricted presentation of anexogenous lipid antigen to T cells. Nat. Immunol. 5, 175–181 (2004).

10. Winau, F. et al. Saposin C is required for lipid presentation by human CD1b. Nat.Immunol. 5,169–174 (2004).

11. Gordon, D.A., Wetterau, J.R. & Gregg, R.E. Microsomal triglyceride transfer protein: aprotein complex required for the assembly of lipoprotein particles. Trends Cell. Biol.5, 317–321 (1995).

12. Wetterau, J.R., Combs, K.A., Spinner, S.N. & Joiner, B.J. Protein disulfide isomeraseis a component of the microsomal triglyceride transfer protein complex. J. Biol.Chem. 265, 9800–9807 (1990).

13. Cui, J. et al. Requirement for Vα14 NKT cells in IL-12-mediated rejection of tumors.Science 278, 1623–1626 (1997).

14. Nieuwenhuis, E.E. et al. CD1d-dependent macrophage-mediated clearance ofPseudomonas aeruginosa from lung. Nat. Med. 8, 588–593 (2002).

15. Vincent, M.S., Gumperz, J.E. & Gumperz, M.B. Understanding the function of CD1-restricted T cells. Nat. Immunol. 4, 517–523 (2003).

16. Gumperz, J.E. et al. Murine Cd1d-restricted T cell recognition of cellular lipids.Immunity 12, 211–221 (2000).

17. Joyce, S. et al. Natural ligand of mouse CD1d1: cellular glycosylphospha-tidylinosi-tol. Science 279, 1541–1544 (1998).

18. Park, J.J. et al. Lipid-protein interactions: biosynthetic assembly of CD1 with lipids inthe endoplasmic reticulum is evolutionarily conserved. Proc. Natl. Acad. Sci. USA101, 1022–1026 (2004).

19. Atzel, A. & Wetterau, J.R. Mechanism of microsomal triglyceride transfer protein cat-alyzed lipid transport. Biochemistry 32, 10444–10450 (1993).

20. Jamil, H. et al. Microsomal triglyceride transfer protein – specificity of lipid bindingand transport. J. Biol. Chem. 270, 6549–6554 (1995).

21. Raabe, M. et al. Analysis of the role of microsomal triglyceride transfer protein in theliver of tissue-specific knockout mice. J. Clin. Invest. 103, 1287–1298 (1999).

22. Bleicher, P.A. et al. Expression of murine CD1 on gastrointestinal epithelium. Science250, 679–682 (1990).

23. Bjorkegren, J., Beigneux, A., Bergo, M.O., Maher, J.J. & Young, S.G. Blocking thesecretion of hepatic very low density lipoproteins renders the liver more susceptible totoxin-induced injury. J. Biol. Chem. 277, 5476–5483 (2002).

24. Ralf, K., Schwenk, F., Aguet, M. & Rajewsky, K. Inducible gene targeting in mice.Science 269, 1427–1429 (1995).

25. Osman, Y. et al. Activation of hepatic NKT cells and subsequent liver injury followingadministration of α-galactosylceramide. Eur. J. Immunol. 30, 1919–1928 (2000).

26. van de Wal, Y. et al. Delineation of a CD1d-restricted antigen presentation pathwayassociated with human and mouse intestinal epithelial cells. Gastroenterology 124,1420–1431 (2003).

27. Heller, F., Fuss, I.J., Nieuwenhuis, E., Blumberg, R.S. & Strober, W. Oxazolone coli-tis, a Th2 colitis model resembling ulcerative colitis, is mediated by IL-13-producingNK-cells. Immunity 17, 629–638 (2002).

28. Davidson, N.O. & Shelness, G.S. Apolipoprotein B: mRNA editing, lipoprotein assem-bly, and presecretory degradation. Annu. Rev. Nutr. 20, 169–193 (2000).

29. Bendelac, A. et al. CD1 recognition by mouse NK1+ T lymphocytes. Science 268,863–865 (1995).

30. Chen, D. et al. Carcinoembryonic antigen-related cellular adhesion molecule 1 iso-forms alternatively inhibit and costimulate human T cell function. J. Immunol. 172,3535–3543 (2004).

NATURE MEDICINE VOLUME 10 | NUMBER 5 | MAY 2004 539

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine