Signaling by vitamin A and retinol-binding proteinregulates gene expression to inhibit insulin responsesDaniel C. Berrya,b, Hui Jina, Avijit Majumdara, and Noa Noya,b,1

Departments of aPharmacology and bNutrition, Case Western Reserve University School of Medicine, Cleveland, OH 44106

Edited* by Joseph Schlessinger, Yale University School of Medicine, New Haven, CT, and approved February 1, 2011 (received for review July 28, 2010)

It currently is believed that vitamin A, retinol, functions throughactive metabolites: the visual chromophore 11-cis-retinal, and ret-inoic acids, which regulate gene transcription. Retinol circulates inblood bound to retinol-binding protein (RBP) and is transportedinto cells by a membrane protein termed “stimulated by retinoicacid 6” (STRA6). We show here that STRA6 not only is a vitamin Atransporter but also is a cell-surface signaling receptor activatedby the RBP–retinol complex. Association of RBP-retinol with STRA6triggers tyrosine phosphorylation, resulting in recruitment andactivation of JAK2 and the transcription factor STAT5. The RBP–retinol/STRA6/JAK2/STAT5 signaling cascade induces the expres-sion of STAT target genes, including suppressor of cytokine signal-ing 3 (SOCS3), which inhibits insulin signaling, and peroxisomeproliferator-activated receptor gamma (PPARγ), which enhanceslipid accumulation. These observations establish that the parentalvitamin A molecule is a transcriptional regulator in its own right,reveal that the scope of biological functions of the vitamin isbroader than previously suspected, and provide a rationale forunderstanding how RBP and retinol regulate energy homeostasisand insulin responses.

retinoids | adipokine | obesity | insulin-resistance

Vitamin A was recognized as an essential factor in foods abouta century ago (1, 2), and a substantial body of knowledge about

its biological functions has accumulated since then. It usually isbelieved that the parental vitamin A molecule, retinol (ROH),functions through active metabolites: 11-cis-retinal, which medi-ates phototransduction, and retinoic acids, which regulate genetranscription by activating specific nuclear receptors (3, 4). Themain storage pool forROH in the body is the liver, and the vitaminis secreted from this tissue bound to a 21-KDa polypeptide termed“serum retinol-binding protein” (RBP) (5). Association with RBPallows the poorly soluble vitamin to circulate in plasma, but retinoldissociates from the protein before entering cells. Interestingly, ithas been reported that serum levels of RBP aremarkedly elevatedin obese mice and humans, and that high serumRBP levels induceinsulin resistance (6). How RBP suppresses insulin signaling andwhether this activity is related to the function of the protein asa retinol carrier is unknown.It was reported recently that an integral plasma membrane

protein termed “stimulated by retinoic acid 6” (STRA6) mediatesuptake of ROH from RBP into cells (7). Mutations in STRA6in humans lead to defects in embryonic development, resultingin multiple malformations collectively termed “Matthew–Woodsyndrome” (8, 9). Interestingly, one of the identified disease-causing mutations in STRA6, T644M, is located at the cytosolic Cterminus of the protein within a sequence recognizable as an SH2domain-binding motif (8). Such motifs, which contain a phos-photyrosine, often are used by cytokine receptors as a docking sitefor the transcription factors STATs. Binding of cognate extra-cellular ligands to cytokine receptors triggers tyrosine phos-phorylation, leading to recruitment of STATs and their activatingJanus kinases (JAKs). These receptors thus propagate a signalingcascade that culminates in mobilization of STATs to the nucleuswhere they induce transcription of specific target genes (10). Thesignificance of the presence of a putative SH2 domain-bindingmotif in STRA6 and the basis for the critical need for this proteinduring development are unknown.

We showhere that binding of retinol-boundRBP (holo-RBP) toSTRA6 induces STRA6 phosphorylation, leading to recruitmentand activation of JAK2 and STAT5. We show further that thissignaling cascade, initiated by RBP-ROH, results in up-regulationofexpressionofSTATtarget genes including suppressorof cytokinesignaling 3 (SOCS3), which inhibits cytokine signaling mediatedby the JAK/STAT pathway (11, 12), and peroxisome proliferator-activated receptor γ (PPARγ), which controls adipocyte lipidhomeostasis. The observations reveal that STRA6 functions asa cytokine receptor to transduce signalling by holo-RBP, and pointat a novel mode by which retinol regulates insulin responses.

ResultsRBP-ROH Triggers STRA6 Phosphorylation, Leading to Recruitmentand Phosphorylation of JAK2 and STAT5. HepG2 hepatocarcinomacells, which endogenously express STRA6, were used to in-vestigate whether this protein is a signaling receptor activated byRBP-ROH. Human RBP lacking its N-terminal secretion signal(hRBPΔN), which corresponds to circulating RBP (13), wasexpressed in Escherichia coli and purified in the presence ofROH to obtain holo-RBP (14) (Fig. S1A). Non-liganded RBP(apo-RBP) was obtained by extracting ROH from the holo-protein, and its viability was verified by fluorescence titrations(15) (Fig. S1B). As with native RBP (16), the equilibrium dis-sociation constant characterizing the interactions of recombi-nant RBP with ROH is ∼100 nM. HepG2 cells were transfectedwith an expression vector for STRA6. Twenty-four hours later,cells were treated with precomplexed RBP-ROH (1 μM, corre-sponding to serum levels) and were lysed at different time points.STRA6 was immunoprecipitated, and its tyrosine phosphoryla-tion level assessed by immunoblots. The data (Fig. 1A) show thatRBP-ROH triggered a transient STRA6 phosphorylation thatpeaked at ∼30 min. HepG2 cells then were transfected withexpression vectors for STRA6 or STRA6 carrying mutations inthe putative SH2 domain-binding motif, Y643F, which lacks thetyrosine residue of the motif, and T644M, lacking the threonineresidue of the YTLL SH2 domain-binding motif of STRA6 andcorresponding to a Matthew–Wood mutation. Cells were treatedwith precomplexed RBP-ROH, STRA6 was immunoprecipi-tated, and its tyrosine phosphorylation level was assessed (Fig.1B). RBP-ROH increased the level of tyrosine phosphorylationof STRA6 but not of the mutant counterparts. RBP-ROH alsoinduced an association between STRA6 and STAT5 (Fig. 1 C–E). As with growth hormone and insulin, which are known toactivate STAT5, RBP-ROH increased STAT5 phosphorylation(Fig. 1F). Neither ROH nor RBP alone was effective (Fig. 1G),establishing that the activity was exerted specifically by holo-RBPand verifying that it did not emanate from the presence of bac-terial contaminants in the recombinant protein preparations.RBP-ROH treatment did not alter the phosphorylation level ofeither STAT3 or STAT1 (Fig. S1 C and D). Similar to the pat-

Author contributions: D.C.B., H.J., and N.N. designed research; D.C.B., H.J., and A.M.performed research; D.C.B., H.J., and N.N. analyzed data; and N.N. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1To whom correspondence should be addressed. E-mail: noa.noy@case.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1011115108/-/DCSupplemental.

4340–4345 | PNAS | March 15, 2011 | vol. 108 | no. 11 www.pnas.org/cgi/doi/10.1073/pnas.1011115108

Dow

nloa

ded

by g

uest

on

May

29,

202

1

tern of STRA6 phosphorylation, and as expected from a bonafide signaling event, phosphorylation of STAT5 induced by RBP-ROH was transient (Fig. 1H). Ectopic overexpression of STRA6enhanced both basal and RBP-ROH–induced STAT5 phos-phorylation, whereas expression of the STRA6-T644M mutantabolished the effect (Fig. 1I). RBP-ROH also induced associa-tion of STRA6 with JAK2 (Fig. 1J) and triggered JAK2 phos-phorylation (Fig. 1K). The phosphorylation of JAK2 in responseto RBP-ROH was enhanced upon overexpression of STRA6(Fig. 1K), and decreasing the expression level of JAK2 inhibitedRBP-ROH–induced STAT5 phosphorylation (Fig. S1F). Theseobservations indicate that RBP-ROH activates a signaling cas-cade mediated by STRA6, JAK2, and STAT5.

RBP-ROH Enhances the Transcriptional Activity of STAT5. Trans-activation assays using a luciferase reporter driven by STAT5response elements were carried out to examine the effect ofRBP-ROH on the transcriptional activity of STAT. HepG2 cellswere treated with ROH, RBP, RBP-ROH, or RA, an ROHmetabolite that regulates transcription by activating nuclearreceptors (4, 17, 18). Insulin, an established STAT activator (19),was used as a positive control. Insulin and RBP-ROH induced

reporter activity, but RBP, ROH, and RA did not (Fig. 1L). Theactivation was enhanced upon overexpression of STRA6 (Fig.1M) and was inhibited upon decreasing the expression level ofthe receptor (Fig. S1H). STRA6 variants lacking the SH2 do-main-binding motif (Y643F, T644M, ΔC) inhibited the ability ofRBP-ROH to activate STAT (Fig. 1M).The ability of RBP-ROH to activate STAT was examined

further by monitoring the expression of two endogenous STATtarget genes: SOCS3 (20) and PPARγ (21). RBP-ROH, but notROH or RBP, up-regulated the expression of both genes (Fig.2A and Fig. S2B). RA efficiently induced Cyp26a, a target genefor the RA receptor RAR, but did not affect the expression ofthe STAT targets (Fig. 2B), demonstrating that RA is tran-scriptionally active in these cells but is not involved in the RBP-ROH response. In addition to ROH, RBP also can bind thevitamin A metabolites retinal (RAL) and RA (16). However,complexes of RBP with either RAL or RA did not activateSTAT (Fig. 2C), nor was STAT activated by RAL alone (Fig.S2C). To examine the location from which RBP-ROH exerts itsactivity, cells were transfected with expression vectors for RBP,which is secreted into the medium in the form of holo-RBP andcan function extracellularly, or with RBP lacking its secretion

STAT5

STRA6 T644M Y643F

RBP-ROH - + - + - + - +

IP: STRA6e.v

0 5 15 30 45 60time (min)

STRA6-pY STRA6-pY

RBP-ROH - + - + - + - +STRA6 T644M Y643Fe.v

IP: STRA6

A B C

F

G

H I J K

L M

STRA6

ERBP-ROH - +

IP: STAT5- insulin GH RBP-ROH

pSTAT5

STRA6

2

2.5/S

TR

A6

pSTRA6

STAT5

* *

STRA6

D

- ROH RBP RBP-ROH

STAT5

STRA6

1.5

2

2.5

A6/S

TA

T5

**

pSTAT5

STAT5

0.5

1

1.5

old

phospho-p

rote

in/

e.v STRA6 T644M e.v. STRA6IP: STRA6

STRA6

pSTAT5

STAT50

0.5

1

- RBP-ROH

fold

ST

R0fo

STRA6e.vRBP-ROH - + - +

0 10 20 40 60

pSTAT5

time (min)

STAT5

pSTAT5

RBP-ROH

STAT5

- + - + - +

pJAK2

JAK2

RBP-ROH - + - +

JAK2

STRA6

RBP-ROH

STRA6

- +

4

6

8

10

STRA6 C

e.v.

STRA6

STRA6-T644M

STRA6-Y643F

cife

rase/b

-gal

1.5

2

2.5

3

cife

rase/b

-gal

**

*

0

2

4

- RBP-ROH

fold

luc

*

- ROH RBP RBP-ROH insulin RA

0

0.5

1

fold

luc

##

D

Fig. 1. RBP-ROHinducesSTRA6phosphorylation, triggeringrecruitmentandactivationof JAK2andSTAT5. (A)HepG2cellsweretransfectedwithavectorharboringSTRA6. Cellswere treatedwithRBP-ROH (1 μM)24hafter transfectionand lysedatdenoted times. STRA6was immunoprecipitated, andprecipitateswereblotted forphosphor-tyrosine. (B) HepG2 cells were transfected as denoted. Similar overexpression of proteins was verified for (Fig. S2A). Cells were treated with RBP-ROH (1μM; 15 min), STRA6-immunoprecipitated, and blotted for phospho-tyrosine. (C) HepG2 cells were transfected as denoted. STRA6 precipitated, and precipitateswereblottedas indicated. (D) Quantification ofbands inBandC in three independent experiments. (*P=0.05 vs. STRA6-transfectednontreated cells.) (E) (Upper)HepG2 cells were treated as denoted, STAT5 immunoprecipitated, and precipitates blotted for STRA6 and total STAT5. (Lower) Quantification of bands in threeindependent experiments. (**P = 0.01.) (F) HepG2 cells were treated with insulin (15 min, 25 nM), growth hormone (GH, 500 ng), or RBP-ROH. Lysates wereimmunoblotted for phosphorylated STAT5 (pSTAT5, Y694) and total STAT5. (G) HepG2 cellswere treatedwith denoted ligands (1 μM;15min), and lysates blottedfor denoted proteins. (H) Cells were treated with RBP-ROH and lysates blotted for pSTAT5 (pY694) and total STAT5. (I) Cells were transfected as denoted andtreated with RBP-ROH (1 μM; 15min) and lysates were immunoblotted as denoted. (J) HepG2 cells were transfected with a STRA6-encoding vector, treated withRBP-ROH (1 μM; 15 min), lysed, and STRA6 was immunoprecipitated. Precipitates were immunoblotted for JAK2. (K) HepG2 cells were transfected as denoted,treated with RBP-ROH (1 μM; 15 min), lysed, and immunoblotted for pJAK2 (Y1007/1008) and total JAK2. (L and M) HepG2 cells were cotransfected with a lu-ciferase reporter driven by STAT response elements and an expression vector for β-galactosidase. (L) Cells were treatedwith denoted ligands (1 μM), or insulin (5nM). Cellswere lysed25h later, and luciferaseactivitywasmeasuredandnormalized to β-galactosidase. (mean± SEM;*P<0.02 vs. nontreated controls;n=3.) (M)Transactivation assays using cells transfected with denoted vectors. Cells were treated with RBP-ROH for 24 h, and luciferase activity was normalized to β-ga-lactosidase. (mean ± SEM; *P < 0.02 vs. nontreated controls; #P < 0.05 vs. RBP-ROH–treated cells transfected with an empty vector; n = 3.)

Berry et al. PNAS | March 15, 2011 | vol. 108 | no. 11 | 4341

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

May

29,

202

1

signal (RBPΔN), which is trapped in the cells. Ectopic expres-sion of RBP induced SOCS3, and treatment with ROH furtherenhanced the response. In contrast, no effect was observed incells that expressed RBPΔN (Fig. 2D), indicating that RBP-ROH activated STAT through extracellular activities. In agree-ment with these results, treatment of HepG2 cells with condi-tioned medium collected from cells that ectopically overexpressedRBP induced SOCS3, whereas medium from RBPΔN-expressingcells did not (Fig. S2D). Attesting to the specific involvement ofSTRA6 in transcriptional activation by RBP-ROH, decreasing theexpression level of the receptor inhibited activation of STAT byRBP-ROH but not by another JAK/STAT activator, IL-6 (Fig.2E). In addition, overexpression of STRA6 increased the basalexpression and enhanced the RBP-ROH response of both SOCS3(Fig. 2F) and PPARγ (Fig. S3A). Interestingly, the expressionlevels of these genes in cells overexpressing STRA6 mutantslacking the SH2 domain-binding motif were lower than in controlcells, indicating the mutants exert dominant negative activities

(Fig. 2F and Fig. S3A). The molecular mechanism by which thesemutants interfere with STRA6 signaling remains to be clarified.Overexpression of STAT5 up-regulated SOCS3 and PPARγ (Fig.2G and Fig. S3B). Two synthetic JAK inhibitors, AG490 andZM449289, which efficiently inhibited up-regulation of SOCS3by IL-6 (Fig. S3C), also abolished the ability of RBP-ROH toactivate STAT (Fig. 2H and Fig. S3D). Decreasing the expressionlevel of JAK2, but not of JAK1, inhibited RBP-ROH–induced up-regulation of SOCS3 (Fig. 2I) and PPARγ (Fig. S3E). Taken to-gether, these observations establish that RBP-ROH activatesSTAT5 through a pathway mediated by STRA6 and JAK2.

RBP-ROH/STRA6/STAT5 Pathway Enhances Triglyceride Accumulationand Impairs Insulin Responses. The ability of RBP-ROH to activatea JAK/STAT pathway also was examined in adipocytes. To thisend, the well-established cultured adipocyte cell model NIH 3T3-L1 was used. Cells were induced to differentiate by a standardprotocol (22), and differentiation was verified by monitoring lipid

A

7

8

9

/18s SOCS3

PPARγ **

7

8

9

8s

SOCS3

B

PPAR

*

14

16

18

18

s PPARγSOCS3

C

*

0

1

2

3

4

5

6

ROH RBP RBP ROH

fold

mR

NA

/

1

2

3

4

5

6

fold

mR

NA

/18

CYP26a

*

*

2

4

6

8

10

12

fold

mR

NA

/1

*

3

3.5

4

s m

RN

A e.v

RBP

RBPΔN

Dhis-RBP N

RBP

β- tubulin

*

- ROH RBP RBP-0

RA RBP-ROH-0

- RBP-ROH RBP-RA RBP-RAL

8

10e.v.

STRA6shRNA

RN

A/1

8s

E

STRA6 T644M

F

6

7

8

mR

NA e.v.

STRA6

*

*

* *

0

0.5

1

1.5

2

2.5

fold

SO

CS

3/1

8

RBPΔN β

*

0

2

4

6

fold

SO

CS

3 m

STRA6-T644M

1

2

3

4

5

fold

SO

CS

3/1

8s m

STRA6ΔC

STRA6-Y643F

****

*

## #

0- ROH

0

- ROH RBP RBP-ROH IL-6 0- RBP-ROH

6

7

/18s

e.v

G

**

H-

5

6

18s

ZM449829* 8

9

8s

e.v.

JAK1shRNA

IJAK2

ev shRNA

1

2

3

4

5

6

ld S

OC

S3

mR

NA

/

+STAT5

+STRA6

**

*

**AG490

1

2

3

4

5

d S

OC

S3 m

RN

A/ ZM449829

3

4

5

6

7

SO

CS

3 m

RN

A/1

8

JAK2shRNAJAK2

ev shRNA

JAK1

ev shRNA

β-tubulin

0

- ROH RBP RBP-ROH

fo

0

1

fold

RBP-ROH-0

1

2

- RBP-ROH

fold

β-tubulin

JAK1*

γ

Δ

Fig. 2. RBP-ROH activates STAT in a STRA6- and JAK2-mediated fashion. (A) HepG2 cells were treated with denoted ligands (1 μM; 4 h). SOCS3 and PPARγmRNA were measured by qPCR. (mean ± SEM; *P < 0.01 vs. nontreated controls; n = 3.) (B) HepG2 cells were treated with RBP-ROH or RA (1 μM; 4 h). mRNAfor SOCS3, PPARγ, and CYP26a were measured by qPCR. (mean ± SEM; *P < 0.02 vs. nontreated controls; n = 3.) (C) Cells were treated with denoted ligands(1 μM; 4 h). mRNA for SOCS3 and PPARγ were measured by qPCR. (mean ± SEM; *P < 0.01 vs. nontreated controls; n = 3.) (D) Cells were transfected withexpression vectors for RBP or histidine-tagged RBP lacking its secretion signal (his-RBPΔN). Cells were treated with vehicle or ROH (1 μM; 4 h), and SOCS3mRNAwas measured by qPCR. (mean ± SEM; *P < 0.01 vs. corresponding nontreated controls; n = 3.) (Inset) Immunoblots demonstrating overexpression of denotedproteins. (E) Cells were transduced as denoted. Down-regulation of STRA6 was verified by qPCR (Fig. S1G). Forty-eight hours later, cells were treated withROH, RBP, RBP-ROH (1 μM), or IL-6 (5 ng) for 4 h. SOCS3 mRNA was measured by qPCR. (mean ± SEM; *P < 0.01 vs. nontreated controls; **P < 0.02 vs. RBP-ROH–treated cells transfected with empty vector; n = 3.) (F) HepG2 cells were transfected as denoted. Similar overexpression of proteins was verified byimmunoblots (Fig. S2A). Cells were treated with RBP-ROH (1 μM; 4 h). (mean ± SEM; *P < 0.01 vs. nontreated controls; **P < 0.05 vs. nontreated cellstransfected with empty vector; #P < 0.05 vs. RBP-ROH–treated cells transfected with empty vector; n = 3.) (G) Cells were transfected as denoted. Over-expression was verified by qPCR (Fig. S1F). Cells were treated with denoted ligands (1 μM; 4 h) 48 h after transfection. SOCS3 mRNA was measured. (Mean ±SEM; *P < 0.05 vs. nontreated cells transfected with empty vector; **P < 0.05 vs. RBP-ROH–treated cells transfected with empty vector; n = 3.) (H) Cells werepretreated with the JAK inhibitors AG490 or ZM449829 (50 μM) for 24 h and then were treated with RBP-ROH (1 μM; 4 h). SOCS3 mRNA was measured.(Mean ± SEM; *P < 0.01 vs. nontreated cells; n = 3.) (I) Cells were transfected with an empty lentiviral vector or a lentiviral vector harboring denoted shRNAs.Cells were treated with RBP-ROH (1 μM; 4 h) 48 h later. SOCS3 mRNA was measured (mean ± SEM; *P < 0.01 vs. RBP-ROH–treated cells transfected with emptyvector; n = 3.) (Insets) Immunoblots demonstrating down-regulation of JAKs.

4342 | www.pnas.org/cgi/doi/10.1073/pnas.1011115108 Berry et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

1

accumulation (Fig. S4A) and examining the expression of the ad-ipocyte marker FABP4 (22). Preadipocytes displayed very lowlevels of STRA6 and RBP, but both genes were markedly inducedduring differentiation (Fig. S5A). In accordance with the expres-sion of STRA6, RBP-ROH had little effect in preadipocytes (Fig.S5B) but effectively triggered phosphorylation of STAT5 andJAK2 (Fig. S5C) and up-regulated the expression of SOCS3 andPPARγ (Fig. S5D–F) in mature adipocytes. The protein synthesisinhibitor cycloheximide (CHX) did not inhibit the induction of theSTAT target genes by RBP-ROH (Fig. S5 D and E), demon-strating that the activity represented a direct response. Similar tothe behavior in HepG2 cells, decreasing the expression of eitherSTRA6 or STAT5 abolished the ability of RBP-ROH to up-reg-ulate SOCS3 and PPARγ in adipocytes (Fig. S5G andH). RA didnot affect SOCS3 expression, further demonstrating that thismetabolite is not involved in the pathway (Fig. S5G).The STAT target gene PPARγ is a key regulator of adipose

lipid storage (23). In accordance with up-regulation of thisprotein, treatment of adipocytes with RBP-ROH increased tri-

glyceride accumulation by the cells and did so in a STRA6-dependent manner (Fig. S5I). Down-regulation of RBP in thesecells reduced their lipid content (Fig. S6A). Similar effects wereobserved in HepG2 cells (Fig. S6 B and C). SOCS3 is a potentinhibitor of insulin receptor-mediated signaling (24, 25). Indeed,treatment of adipocytes with RBP-ROH inhibited insulin-inducedactivation of insulin receptor, monitored by insulin receptorautophosphorylation and by phosphorylation of the insulin re-ceptor downstream effector Akt1 (26) (Fig. 3 A–C). Decreasingthe expression levels of STRA6 or STAT5 (Fig. S4 B–D), whichabolished the ability or RBP-ROH to induce SOCS3 (Fig. 2Fand Fig. S4F), or decreasing the expression level of SOCS3 (Fig.S4E) abrogated the inhibition (Fig. 3D). Similar effects were ob-served in HepG2 cells (Fig. 3E). Another hallmark of insulinactivity is the mobilization of the glucose transporter GluT4 toplasma membranes. Adipocytes were pretreated with vehicle orwith RBP-ROH for 8 h and then were treated with insulin (20nM, 25 min), and a fraction enriched in plasma membranes wasisolated (27). Immunoblotting the fraction for GluT4 demon-

pIR

insulin - + - + - +

CROH RBP

A

pIR

IR

pAkt1

Akt1

Binsulin - + - +

RBP-ROH

insulin - + - + RBP-ROH

IR

0.8

1.2

pIR

/tota

l IR

0.4

0.8

1.2

fold

pIR

/IR

*

Akt1

0.4

0.8

1.2

old

pA

kt/A

kt

*

0

0.4

insulin ROH+insulin RBP+insulin

fold

0

insulin RBP-ROH+insulin

0

insulin RBP-ROH+insulin

fo

D NIH3T3-L1 adipocytes

RBP-ROH

pIR

β-tubuline.v.

STRA6-shRNApIR

insulin - + - +

E

HepG2 cells

e.v.pIR

β-tubulin

insulin - + - + RBP-ROH

1.2

1.4

1.6

1.8

2e.v

STRA6-shRNA

STAT5-shRNA

SOCS3-shRNA

β-tubu

lin

β-tubulin

β-tubulinSTAT5-shRNA

pIR

pIR

STRA6-shRNApIR

β-tubulin

SOCS3-shRNApIR

β tubulin0.2

0.4

0.6

0.8

1

foldph

osph

oIR/β

SOCS3-shRNAp

β-tubulinβ-tubulin

F totalinsulin - + +

RBP-ROH

0

- RBP-ROH

0.8

1

1.2

1.4e.v

STRA6-shRNASOCS3-shRNA

oIR/β-tub

ulin

CPMlysate

Na-K ATPase

Na-K ATPase

GluT4

insulin - + +

0

0.2

0.4

0.6

0 8

- RBP-ROH

foldph

osph

o

Fig. 3. The RBP-ROH/STRA6/STAT5 pathway impairs insulin responses. (A and B) Differentiated adipocytes were treated with insulin (20 nM; 25 min) or RBP-ROH (1 μM; 8 h) or were pretreated with RBP-ROH for 8 h before treatment with insulin. (Upper) Lysates were immunoblotted for phosphorylated insulinreceptor (pIR-Y1146) and total insulin receptor (IR) (A) or phosphorylated Akt (pAkt1-S473) and total Akt1 (B). (Lower) Quantification of phosphorylated/total proteins. (Data shown are mean ± SEM; #P < 0.02 vs. controls not pretreated with RBP-ROH; n = 3. (C) Differentiated adipocytes were treated withROH or RBP (1 μM; 8 h) and then were treated with insulin (20 nM; 25 min.). (Upper) Lysates were immunoblotted as denoted. (Lower) Quantification ofimmunoblots. (Data shown are mean ± SEM; n = 3.) (D and E ) Differentiated adipocytes (D) or HepG2 cells (E) were transduced for 5 d with lentivirusesharboring the denoted shRNAs. Decreased expression of target proteins was verified by qPCR and/or immunoblots (Figs. S3E, S4B, and S5 C and D). Cellswere treated with insulin (20 nM; 25 min) or RBP-ROH (1 μM; 8 h) or were pretreated with RBP-ROH for 8 h before treatment with insulin. Lysates wereimmunoblotted as denoted in left panel in D and upper panel in E. Experiments were performed twice with similar results. Right panel in D and lowerpanel in E show quantification of data, mean of two independent experiments. (F) Differentiated adipocytes were treated with vehicle or insulin (20 nM;25 min) or were pretreated with RBP-ROH (1 μM; 8 h) before treatment with insulin. Crude plasma membrane fractions (cpm) were obtained (27). (Left)Immunoblots showing enrichment of Na-K ATPase in cpm. (Right) Immunoblots of GluT4 in cpm. The plasma membrane marker Na-K ATPase served asa loading control.

Berry et al. PNAS | March 15, 2011 | vol. 108 | no. 11 | 4343

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

May

29,

202

1

strated that RBP-ROH inhibited insulin-induced mobilizationof GluT4 to plasma membranes (Fig. 3F).

RBP Induces the Expression of STAT Target Genes and Inhibits InsulinSignaling in STRA6-Expressing Tissues in Vivo. In mice, STRA6 isexpressed in white adipose tissue and skeletal muscle but not inliver (Fig. 4A) (28). We injected 10-wk-old C57BL/6 mice withRBP (12.5 mg/mL, 80 μL) intraperitoneally three times at 2-hintervals. Mice were killed 1 h after the last injection. Bloodlevels of RBP were assessed in lean mice, lean mice injected withRBP, and mice that were fed a high fat/high carbohydrate dietfor 18 wk and were obese and insulin resistant (22). As previouslyreported (6), serum RBP levels were about twofold higher inobese mice than in lean animals (Fig. 4B). Serum of RBP-injected mice displayed two bands, corresponding to endogenousand recombinant RBP, with total RBP approximating the levelobserved in obese animals. Hence, the injection elevated serumRBP within a physiologically meaningful range. Treatment ofmice with RBP enhanced the phosphorylation of STRA6, STAT5,and JAK2 in adipose tissue (Fig. 4 C–E), up-regulated the ex-pression of SOCS3 in adipose tissue and muscle (Fig. 4 E and F),

and induced the expression of adipose PPARγ (Fig. 4G). RBPtreatment also decreased the levels of phosphorylation of insulinreceptor and Akt1 in both adipose tissue and muscle (Fig. 4 H andI). Strikingly, although RBP activated the JAK2/STAT5 pathwayand inhibited insulin signaling in these STRA6-expressing tis-sues, it had no effect on the phosphorylation status of STAT5(Fig. S7A), the expression of SOCS3 and PPARγ (Fig. S7B), orthe level of phosphorylation of insulin receptor or Akt1 (Fig. 4J)in the liver, which does not express the receptor (Fig. 4A).

DiscussionThe findings described here establish that, in addition to beinga ROH transporter, STRA6 also functions as a surface sig-naling receptor. As depicted in Fig. 5, the data demonstratethat RBP-ROH serves as an extracellular ligand to trigger ty-rosine phosphorylation in the cytosolic domain of STRA6. Inturn, activated STRA6 propagates a JAK2/STAT5 cascadethat culminates in induction of STAT target genes. RBP-ROHthus joins the more than 30 cytokines, hormones, and growthfactors that signal through surface receptors associated withJAKs and STATs (10). The identification of RBP-ROH as an

1

1.2

1.4

1.6

8s m

RN

A

A B lean +RBP obese

endog RBP

his-RBP

2

n

STRA6-pY

STRA6

buffer RBPIP: STRA6CSTRA6

WAT SM liver

-actin

**

DpSTAT5

STAT5

buffer RBP

0

0.2

0.4

0.6

0.8

fold

ST

RA

6/1

8

0

0.4

0.8

1.2

1.6

fold

RB

P/lean

2

3

8s m

RN

A

WAT

SM

F*

*

E

1 5

2

2.5

s m

RN

A

G

0

WAT SM liverlean RBP inj obese

8

10

12

n/

-actin pJAK2

PPARγSOCS3

*

pJak2

buffer RBP*

0

1

2

buffer RBP

fold

SO

CS

3/1

0

0.5

1

1.5

buffer RBP

fold

PP

ARγ/

18s

0

2

4

6

buffer RBP

fold

pro

tein

**

pJak2

PPARγ

SOCS3

buffer RBP

pIR

WATH skeletal muscle I

buffer RBPbuffer RBP

liverJbuffer RBP

pIR

buffer RBP

pIR

pAkt1

Akt1

1.2

ein

/Akt1

pIR

pAkt1

pAkt1

Akt1

1.2

ein

/AK

T1

pIR

pAkt1

pAkt1

Akt1

1.2

ein

/AkT

1

0

0.4

0.8

buffer RBPfold

phos.pro

te

p

*

*

0

0.4

0.8

buffer RBPfold

phos.pro

te

* *

0

0.4

0.8

buffer RBPfold

phos.pro

te

Fig. 4. RBP triggers phosphorylation of STRA6, STAT5, and JAK2, up-regulates the expression of STAT target genes, and represses insulin signaling in vivo. (A)STRA6 mRNA in white adipose tissue (WAT), skeletal muscle (SM), and liver, measured by qPCR. (Data shown are mean ± SEM in three mice.) (Inset)Immunoblots of STRA6 in white adipose tissue, skeletal muscle, and liver in two mice. (B) (Upper) Immunoblots of serum RBP in lean and obese mice and inlean mice injected with RBP. Results from two animals per group are shown. (Lower) Quantification of RBP in plasma. (Data shown are mean ± SEM; *P < 0.03vs. lean mice; n = 3 mice per group.) (C) STRA6 was immunoprecipitated (IP) from white adipose tissue of three control (buffer) and three RBP-injected miceand immunoblotted for phospho-tyrosine. (D) Immunoblots of pSTAT5 and total STAT5 in white adipose tissue from four control (buffer) and four RBP-injected mice. (E) (Left) Immunoblots of pJAK2, PPARγ, and SOCS3 in WAT from three control (buffer) and three RBP-injected mice. (Right) Quantification ofdenoted protein normalized to β-actin. (*P < 0.05 vs. buffer-injected controls.) (F) SOCS3 mRNA in white adipose tissue and skeletal muscle of control (buffer)and RBP-injected mice. (Data shown are mean ± SEM; *P < 0.05 vs. buffer-injected mice; n = 3 per group.) (G) PPARγ mRNA in white adipose tissue of miceinjected with buffer or RBP. (Data shown are mean ± SEM; *P < 0.05 vs. buffer-injected mice; n = 4 per group.) (H–J) (Upper) Phosphorylation levels of insulinreceptor and Akt1 in white adipose tissue (H), skeletal muscle (I), and liver (J) in mice injected with buffer or RBP. (Lower) Quantification of band intensitiesnormalized to total Akt1. (Data shown are mean ± SEM; *P < 0.05; n = 3 per group.)

4344 | www.pnas.org/cgi/doi/10.1073/pnas.1011115108 Berry et al.

Dow

nloa

ded

by g

uest

on

May

29,

202

1

activator for a STRA6/JAK2/STAT5 cascade which induces theexpression of the well-known inhibitor of insulin signalingSOCS3, provides a molecular basis for the observations that RBP

inhibits insulin responses (6) as well as for the recent report thatSNPs in STRA6 are associated with type 2 diabetes (29). Thepathway also may be involved in other biological functions. Forexample, the observations that a T644M STRA6 mutation, whichimpairs RBP-ROH signaling, results in defects in embryonicdevelopment (8) suggest that the pathway may be involved inembryogenesis. In addition, given that STATs are key regulatorsof cell growth, migration, and survival (30, 31) and that the ex-pression of STRA6 is up-regulated in multiple types of cancers(32), RBP-ROH may be involved in cancer development. Theseobservations indicate that the spectrum of the biological activi-ties of vitamin A is much broader than previously suspected. Thecomplete scope of these activities, as well as the relationshipbetween the function of STRA6 as a vitamin A transporter andits role as a signaling receptor, remain to be explored.

Materials and MethodsDetails concerning reagents and plasmids and cells, and protocols used inquantitative PCR analyses, mutagenesis, and transactivation assays are de-scribed in SI Materials and Methods. hRBPDN was expressed in E. coli andpurified as described (14), and protein viability was ascertained by fluores-cence titrations (15). Crude plasma membranes were isolated as in ref. 27.For additional details, see SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Yin Huang for assistance in cloning. Weare grateful to W. Blaner, J. Schwartz, S. Vogel, and L. Yu-Lee for con-structs and antibodies. This work was supported by National Institutes ofHealth Grants DK060684 and CA107013 (to N.N.) and KO1-DK077915 (toH.J.). D.C.B. was supported in part by National Institutes of Health GrantDK073195T32. The Mouse Metabolic Phenotyping Center of the CaseWestern Reserve University is supported by National Institutes of HealthGrant DK59630.

1. McCollum EV, Davis M (1913) The necessity of certain lipins in the diet during growth.J Biol Chem 15:167–175.

2. Osborne TB, Mendel LB (1919) The vitamins in green foods. J Biol Chem 37:187–200.3. Noy N (2006) Vitamin A Biochemical, Physiological, and Molecular Aspects of Human

Nutrition, ed Stipanuk MH (Saunders Elsevier, St. Louis), 2nd Ed.4. Schug TT, Berry DC, Shaw NS, Travis SN, Noy N (2007) Opposing effects of retinoic acid

on cell growth result from alternate activation of two different nuclear receptors. Cell129:723–733.

5. Noy N (2000) Retinoid-binding proteins: Mediators of retinoid action. Biochem J 348:481–495.

6. Yang Q, et al. (2005) Serum retinol binding protein 4 contributes to insulin resistancein obesity and type 2 diabetes. Nature 436:356–362.

7. Kawaguchi R, et al. (2007) A membrane receptor for retinol binding protein mediatescellular uptake of vitamin A. Science 315:820–825.

8. Pasutto F, et al. (2007) Mutations in STRA6 cause a broad spectrum of malformationsincluding anophthalmia, congenital heart defects, diaphragmatic hernia, alveolarcapillary dysplasia, lung hypoplasia, and mental retardation. Am J Hum Genet 80:550–560.

9. Golzio C, et al. (2007) Matthew-Wood syndrome is caused by truncating mutationsin the retinol-binding protein receptor gene STRA6. Am J Hum Genet 80:1179–1187.

10. Darnell JE, Jr., Kerr IM, Stark GR (1994) Jak-STAT pathways and transcriptionalactivation in response to IFNs and other extracellular signaling proteins. Science 264:1415–1421.

11. Starr R, et al. (1997) A family of cytokine-inducible inhibitors of signalling. Nature 387:917–921.

12. Croker BA, Kiu H, Nicholson SE (2008) SOCS regulation of the JAK/STAT signallingpathway. Semin Cell Dev Biol 19:414–422.

13. Soprano DR, Pickett CB, Smith JE, Goodman DS (1981) Biosynthesis of plasma retinol-binding protein in liver as a larger molecular weight precursor. J Biol Chem 256:8256–8258.

14. Xie Y, Lashuel HA, Miroy GJ, Dikler S, Kelly JW (1998) Recombinant human retinol-binding protein refolding, native disulfide formation, and characterization. ProteinExpr Purif 14:31–37.

15. Cogan U, Kopelman M, Mokady S, Shinitzky M (1976) Binding affinities of retinol andrelated compounds to retinol binding proteins. Eur J Biochem 65:71–78.

16. Noy N, Slosberg E, Scarlata S (1992) Interactions of retinol with binding proteins:Studies with retinol-binding protein and with transthyretin. Biochemistry 31:11118–11124.

17. Chambon P (1996) A decade of molecular biology of retinoic acid receptors. FASEB J10:940–954.

18. Shaw N, Elholm M, Noy N (2003) Retinoic acid is a high affinity selective ligandfor the peroxisome proliferator-activated receptor beta/delta. J Biol Chem 278:41589–41592.

19. Chen J, Sadowski HB, Kohanski RA, Wang LH (1997) Stat5 is a physiological substrateof the insulin receptor. Proc Natl Acad Sci USA 94:2295–2300.

20. Krebs DL, Hilton DJ (2001) SOCS proteins: Negative regulators of cytokine signaling.Stem Cells 19:378–387.

21. Dentelli P, et al. (2009) Formation of STAT5/PPARgamma transcriptional complexmodulates angiogenic cell bioavailability in diabetes. Arterioscler Thromb Vasc Biol29:114–120.

22. Berry DC, Noy N (2009) All-trans-retinoic acid represses obesity and insulin resistanceby activating both peroxisome proliferation-activated receptor beta/delta andretinoic acid receptor. Mol Cell Biol 29:3286–3296.

23. Kersten S, Desvergne B, Wahli W (2000) Roles of PPARs in health and disease. Nature405:421–424.

24. Emanuelli B, et al. (2000) SOCS-3 is an insulin-induced negative regulator of insulinsignaling. J Biol Chem 275:15985–15991.

25. Emanuelli B, et al. (2001) SOCS-3 inhibits insulin signaling and is up-regulated inresponse to tumor necrosis factor-alpha in the adipose tissue of obese mice. J BiolChem 276:47944–47949.

26. Wing SS (2008) The UPS in diabetes and obesity. BMC Biochem 9(Suppl 1):S6.27. Garcia-Souza EP, et al. (2008) Maternal protein restriction during early lactation

induces GLUT4 translocation and mTOR/Akt activation in adipocytes of adult rats. AmJ Physiol Endocrinol Metab 295:E626–E636.

28. Bouillet P, et al. (1997) Developmental expression pattern of Stra6, a retinoic acid-responsive gene encoding a new type of membrane protein. Mech Dev 63:173–186.

29. Nair AK, Sugunan D, Kumar H, Anilkumar G (2010) Case-control analysis of SNPs inGLUT4, RBP4 and STRA6: Association of SNPs in STRA6 with type 2 diabetes in a SouthIndian population. PLoS ONE 5:e11444.

30. Leeman RJ, Lui VW, Grandis JR (2006) STAT3 as a therapeutic target in head and neckcancer. Expert Opin Biol Ther 6:231–241.

31. Quesnelle KM, Boehm AL, Grandis JR (2007) STAT-mediated EGFR signaling in cancer.J Cell Biochem 102:311–319.

32. Szeto W, et al. (2001) Overexpression of the retinoic acid-responsive gene Stra6 inhuman cancers and its synergistic induction by Wnt-1 and retinoic acid. Cancer Res 61:4197–4205.

33. Newcomer ME, et al. (1984) The three-dimensional structure of retinol-bindingprotein. EMBO J 3:1451–1454.

RBP-ROH

extracellular

insulin

cytosolIR IR

STRA6P

P

SOCS3JAK2STAT5

PAKT GluT4 translocation

PP

PPARSTA

T5

STA

T5

SOCS3lipid accumulationP

P

nucleus

PPARSS

pP γ

Fig. 5. Model of the RBP-ROH/STRA6/JAK/STAT pathway. Binding of RBP-ROH to the extracellular moiety of STRA6 triggers tyrosine phosphorylationwithin an SH2 domain-binding motif in the receptor’s cytosolic domain.Phosphorylated STRA6 recruits and activates JAK2, which, in turn, phosphor-ylates STAT5. Activated STAT5 translocates to the nucleus to regulate theexpression of target genes, including SOCS3, which inhibits insulin signaling,and PPARγ, which enhances lipid accumulation. The model of STRA6 (GeneID64220RBP) was generated using software http://bp.nuap.nagoya-u.ac.jp/sosui. The 3D structure of holo-RBP (GenBank accession no. DAA14765.1) wassolved as described in ref. 33.

Berry et al. PNAS | March 15, 2011 | vol. 108 | no. 11 | 4345

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

May

29,

202

1