Modification of ghrelin receptor signaling bysomatostatin receptor-5 regulates insulin releaseSeongjoon Park1,2, Hong Jiang1,3, Hongjie Zhang4, and Roy G. Smith5

Department of Metabolism and Aging, Scripps Research Institute Florida, Jupiter FL 33458

Edited by Michael O. Thorner, University of Virginia Health Sciences Center, Charlottesville, VA, and accepted by the Editorial Board October 2, 2012 (receivedfor review June 11, 2012)

Both ghrelin and somatostatin (SST) inhibit glucose-stimulated in-sulin secretion (GSIS) from pancreatic β-cells, but how these inde-pendent actions are regulated has been unclear. The mechanismmust accommodate noncanonical ghrelin receptor (GHS-R1a)–G-pro-tein coupling to Gαi/o instead of Gαq11 and dependence on energybalance. Here we present evidence for an equilibrium model of re-ceptor heteromerization that fulfills these criteria. We show thatGHS-R1a coupling to Gαi/o rather than Gαq11 requires interactionsbetween GHS-R1a and SST receptor subtype 5 (SST5) and that in theabsence of SST5 ghrelin enhances GSIS. At concentrations of GHS-R1a and SST5 expressed in islets, time-resolved FRET and biolumi-nescence resonance energy transfer assays illustrate constitutiveformation of GHS-R1a:SST5 heteromers in which ghrelin, but notSST, suppresses GSIS and cAMP accumulation. GHS-R1a–G-proteincoupling and the formation of GHS-R1a:SST5 heteromers is depen-dent on the ratio of ghrelin to SST. A high ratio enhances heteromerformation and Gαi/o coupling, whereas a low ratio destabilizes het-eromer conformation, restoring GHS-R1a–Gαq11 coupling. The [ghre-lin]/[SST] ratio is dependent on energy balance: Ghrelin levels peakduring acute fasting, whereas postprandially ghrelin is at a nadir,and islet SST concentrations increase. Hence, under conditions oflow energy balance our model predicts that endogenous ghrelinrather than SST establishes inhibitory tone on the β-cell. Collectively,our data are consistent with physiologically relevant GHS-R1a:SST5heteromerization that explains differential regulation of islet func-tion by ghrelin and SST. These findings reinforce the concept thatsignaling by the G-protein receptor is dynamic and dependent onprotomer interactions and physiological context.

GPCR oligomers | glucose homeostasis

The growth hormone secretagogue receptor type 1a (GHS-R1a), was identified in 1996 as an orphan G-protein–coupled

receptor (GPCR) that regulates the action of a family of smallsynthetic molecules designed to rejuvenate the growth hormone(GH) axis in humans (1, 2). Three years later GHS-R1a wasdeorphanized by discovery of an endogenous agonist made in thestomach called “ghrelin” (3). Ghsr−/− mice are refractory to theGH-releasing and orexigenic properties of ghrelin, confirmingthat GHS-R1a is a physiologically relevant ghrelin receptor (4).The study we describe was prompted by the need for a modelthat includes endogenous ghrelin as well as somatostatin (SST)as a regulator of pancreatic islet function and that elucidatesthe paradox that ghrelin inhibits rather than enhances glucose-stimulated insulin secretion (GSIS).Experiments in ghrelin−/− and ghsr−/− mice led to the conclusion

that endogenous ghrelin is a physiologically important regulator ofGSIS (5–7). In fed mice endogenous ghrelin concentrations are ata nadir; hence, the majority of GHS-R1a binding sites are un-occupied. In this context, treating WT mice with exogenous ghrelinsuppresses GSIS. When WT mice are fasted, endogenous ghrelinlevels reach a maximum, and the mice are refractory to exogenousghrelin. In contrast, fasted ghrelin−/− mice are fully responsive tothe suppression of GSIS by exogenous ghrelin. Collectively, theseresults indicate that in fasted WT mice GHS-R1a binding sitesare fully occupied by endogenous ghrelin, suggesting that under

conditions of negative energy balance endogenous ghrelin acts byestablishing inhibitory tone on insulin secretion.Endogenous ghrelin is fundamentally important for maintaining

euglycemia and for survival under conditions of acute food re-striction (7, 8). Mice lacking the essential enzyme ghrelin octa-noylacyltransferase (GOAT), which converts inactive ghrelinpeptide into its active octanoylated form, are severely compro-mised by a 60% reduction in dietary intake (9–11). However, thephenotype can be rescued by restoring ghrelin or GH to the con-centrations measured in control WT mice subjected to 60% caloricrestriction, indicating that ghrelin is critically important for pre-venting neuroglycopenia (11, 12). Ironically, expression cloning ofGHS-R1a that led to the discovery of ghrelin was achieved usinga synthetic molecule that augmented episodic GH release (2, 3).GHS-R1a is expressed in pancreatic β-cells, and ghrelin inhibits

GSIS (13, 14). Inhibition of insulin release appears paradoxical,because generally activation of GHS-R1a ghrelin results in cou-pling to Gαq11, which would enhance insulin secretion. However,in the islet and β-cells, instead ofGαq11-mediated signal trans-duction, ghrelin suppresses GSIS via GHS-R1a coupling to Gαi/o,thereby reducing cAMP accumulation (14). Traditionally; SSTactivation of somatostatin receptor subtype-5 (SST5) is consideredthe major inhibitor of insulin secretion from β-cells (15–18). SSTreleased from islet δ-cells also inhibits glucagon secretion by ac-tivating SST2 on α-cells (19–22).How and under what physiological conditions are the in-

hibitory actions of ghrelin and SST on insulin secretion in-dependently controlled? Any model that explains ghrelin actionon pancreatic β-cells in vivo must include a mechanism for theswitch in canonical GHS-R1a–G-protein coupling from Gαq11 toGαi/o, a role for SST5 and SST, and dependence on reciprocalchanges in relative concentrations of ghrelin and SST as a con-sequence of changes in glucose concentrations. This reportprovides evidence for a model incorporating GHS-R1a:SST5heteromers that meet these criteria.

ResultsGhrelin Inhibition of GSIS Is Dependent on both GHS-R1a and SST5.We first measured expression of ghrelin, GOAT, GHS-R1a,SST5, and SST2 in rat pancreatic islets (Fig. 1A) and used these

Author contributions: R.G.S. designed research; S.P., H.J., and H.Z. performed research;S.P., H.J., and R.G.S. analyzed data; and S.P. and R.G.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. M.O.T. is a guest editor invited by theEditorial Board.1S.P. and H.J. contributed equally to this work.2Present address: Unit of Basic Medical Science, Department of Investigative Pathology,Graduate School of Biomedical Sciences, Nagasaki University, Nagasaki City 852-8523,Japan.

3Present address: Texas Therapeutics Institute at Brown Foundation, Institute of Molecu-lar Medicine, University of Texas, Health Science Center at Houston, Houston, TX 77030.

4Present address: Internal Medicine Residency Program, York Hospital, York, PA 17403.5To whom correspondence should be addressed. E-mail: rgsmith@scripps.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1209590109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1209590109 PNAS | November 13, 2012 | vol. 109 | no. 46 | 19003–19008

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

June

2, 2

020

results to develop a system to model ghrelin and SST signaling.Pharmacology studies indicate that SST5 rather than SST2 isprimarily involved in inhibiting GSIS from β-cells (15–18). Wehypothesized that noncanonical GHS-R1a coupling resulting inghrelin inhibition of GSIS involves molecular interactions be-tween GHS-R1a and SST5. To test the relative contributions ofGHS-R1a and SST5, we sought a subclone of INS-1 cells thatwould allow independent manipulation of GHS-R1a and SST5to match concentrations present in pancreatic islets. A subclonewas identified, INS-1SJ, that expresses SST2, ghrelin, and GOATat levels similar to those found in rat pancreatic islets with lowexpression of GHS-R1a and undetectable SST5 (Fig. 1A).Increasing GHS-R1a levels alone in INS-1SJ cells by trans-

ducing the cells with a GHS-R1a–expressing lentivirus enhancedinsulin secretion in either the presence or absence of ghrelin,consistent with GHS-R1a canonical Gαq11 signaling and pro-duction of endogenous ghrelin (Fig. 1B). Because INS1-SJ cellsdo not express SST5, this result suggested that SST5 is requiredfor ghrelin suppression of GSIS. We then established conditionsfor expressing GHS-R1a and SST5 at levels closely matchingthose in pancreatic islets (Fig. 2A). INS1-SJ cells expressingGHS-R1a and SST5 (INS1SJ-GHSR-SST5) were treated with

ghrelin siRNA to knockdown endogenous ghrelin production(Fig. S1). Under these conditions exogenous ghrelin inhibitedinsulin release induced by 15 mM glucose (Fig. 2B). WhenINS1SJ-GHSR-SST5 cells were treated with 3 mM and 15 mMglucose, high glucose increased GSIS, and expression of ghrelinsiRNA to knock down production of endogenous ghrelin furtherenhanced GSIS (Fig. 2C); hence, ghrelin siRNA unmasks thesuppression of GSIS caused by endogenous ghrelin. Importantly,enhanced GSIS is accompanied by an increase in cAMP (Fig.2D), consistent with ghrelin regulation of GSIS reported inpancreatic islets by noncanonical Gαi/o coupling (14).

Expression of Equivalent Concentrations of GHS-R1a Prevents SST5Inhibition of GSIS. We next asked if coexpression of equivalentlevels of GHS-R1a andSST5 modifies SST5 function. To avoidactivating SST2 in INS-1SJ cells, we selected the SST5-selectiveagonist Bim23052. INS-1SJ cells expressing either SST5 aloneor with an equal concentration of GHS-R1a were treated withBim23052. When SST5 was expressed alone, Bim23052 inhibitedGSIS, but coexpression of GHS-R1a blocked Bim23052 action(Fig. 3A). Native SST5 exists as a monomer, but signal trans-duction requires agonist-induced formation of homodimers (23);therefore, antagonism of SST5 signaling could be explained byconstitutive formation of GHS-R1a:SST5 heteromers. To testthis possibility, we used time-resolved (Tr)-FRET SNAP-tagtechnology (24, 25), which has the high sensitivity necessary todetect the association of GHS-R1a with SST5 at the concen-trations found in pancreatic islets.In isolation, GHS-R1a exists as homodimers. Previously, we

showed by Tr-FRET that dopamine receptor-2 (DRD2) is a com-petitive inhibitor of GHS-R1a homodimerization, resulting in theformation of GHS-R1a:DRD2 heteromers (26). Using similarmethods employing SNAP-GHS-R1a, Tr-FRET illustrates theformation of GHS-R1a:GHS-R1a homodimers on the plasmamembrane of INS-1SJ cells (Fig. 3B). Concentration-dependentinhibition of the Tr-FRET signal is observed when SST5 is coex-pressed with SNAP-GHS-R1a at ratios of 1/1 and 2/1 (Fig. 3B).Control experiments confirmed that attenuation of the Tr-FRETsignal was not a consequence of coexpressed SST5 inhibitingtransport of SNAP-GHS-R1a to the plasma membrane (Fig. S2);therefore, the marked inhibition of GHS-R1a homodimerization

Rat

Islets INS-1SJ

GHSR1a

SST2

SST5

Ghrelin

GOAT

ACTBGlucose (mM)

GHSR1a

Ghrelin (nM)

In

su

lin

se

cre

tio

n (%

c

on

te

nt)

0

40

80

120

160

C1 Co15 TC1 TCo15G15 15 15 15

- - + +

- 100 - 100

*****

A B INS1-SJ cells

Fig. 1. (A) Comparison of GHS-R1a, SST2, SST5, ghrelin, and GOAT expressionin rat islets and INS-1SJ cells by PCR. (B) INS1-SJ cells transduced with GHS-R1a–expressing lentivirus causes augmentation of GSIS without suppression byghrelin. Data represent means ± SEM (n = 4); **P < 0.01, ***P < 0.001.

020406080

100120 Vehicle

Ghrelin

Insu

lin s

ecre

tion

(% c

onte

nt)

**+ Ghrelin siRNA

pcDNA3.1 GHSR1a + SST5

Glucose

**

***

0

100

200

300

400

3 mM 15 mM

GhrelinsiRNA

***

A

0.0

0.4

0.8

1.2

3 mM 15 mM

GhrelinsiRNA**

*

Glucose

cAM

P(nM

)

*

DC

Insu

lin s

ecre

tion

(% c

onte

nt)

Rat islet

0.0

0.4

0.8

1.2

expr

essi

on Q

PCR

ElectroporatedINS-1SJ cells

B

GHSR1a/ACTB

SST5/ACTB

Ghrelin/ACTB

Con-siRNA

Con-siRNA

Fig. 2. Suppression of insulin secretion by ghrelinis dependent on coexpression of GHS-R1a and SST5in INS-1SJ cells electroporated with GHS-R1a andSST5 to match levels expressed in rat pancreaticislets. (A) Quantitation by real-time PCR of GHS-R1a,SST5, and ghrelin expression in rat islets and elec-troporated INS-1SJ cells. (B) Ghrelin inhibits insulinsecretion induced by 15 mM glucose in INS-1SJ cellsin the presence but not in the absence of GHS-R1aand SST5 following suppression of endogenousghrelin production with ghrelin siRNA. (C) Knock-down of endogenous ghrelin production increasesGSIS. (D) Knockdown of endogenous ghrelin pro-duction increases intracellular cAMP. Data repre-sent means ± SEM (n = 4); *P < 0.05, **P < 0.01,***P < 0.001.

19004 | www.pnas.org/cgi/doi/10.1073/pnas.1209590109 Park et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020

by SST5 is a consequence of high-affinity interactions betweenGHS-R1a and SST5.

SST5-Dependent Modification of GHS-R1a–G-protein Coupling IsControlled by the [Ghrelin]/[SST] Ratio. Both ghrelin and SST arecapable of inhibiting insulin secretion by a mechanism requiringGαi/o suppression of cAMP accumulation. Above we showed thatghrelin inhibition of GSIS and suppression of cAMP accumula-tion, as well as GHS-R1a–dependent antagonism of inhibition ofGSIS by a selective SST5 agonist, are dependent on both GHS-R1a and SST5. To elucidate the mechanism more completelyand to deduce the relative contributions of SST2 and SST5, acell system was needed in which the relative concentrations ofghrelin, SST, GHS-R1a, SST5, and SST2 could be controlled.Precise control requires a null background; therefore, we selectedHEK293 cells. Because inhibition of GSIS by ghrelin and SST inpancreatic β-cells and in the INS-1SJ β-cell line is dependent on

the inhibition of cAMP production, we selected cAMP accu-mulation as a surrogate for GSIS.We next tested whether interactions between GHS-R1a and

SST5 affected SST suppression of cAMP accumulation. InHEK293 cells expressing SST5, SST dose-dependently suppressesforskolin-induced cAMP accumulation with maximum inhibitionat 1–10 nM (Fig. S3A). Remarkably, as in the INS-1SJ cells,coexpressing GHS-R1a with SST5 made the cells refractory toSST (Fig. S3B); however, treatment with ghrelin, but not with des-acyl ghrelin (Fig. S3C), inhibits forskolin-induced cAMP accu-mulation by a pertussis toxin (PTX)-sensitive mechanism (Fig.S3D). Hence, the noncanonical GHS-R1a–G-protein coupling toGαi/o observed in pancreatic β-cells and INS-1SJ β-cells is re-capitulated in HEK293 cells coexpressing GHS-R1a and SST5.Tr-FRET analysis in INS-1SJ cells used fluorophor tagging at

the GHS-R1a N terminus. As an additional test of heteromerformation, we used bioluminescence resonance energy transfer(BRET) with GHS-R1a and SST5 tagged at the C terminus with

0

20

40

60

80

100

120

1:1 1:1 1:2

SNAP-GHSR1a : SST5F

re

t e

ffic

ie

nc

y (%

)SNAP-GHSR1a

: pcDNA3.1(+)

******

*

BIn

su

lin

se

cre

tio

n (%

co

nte

nt)

0

20

40

60

80

100

120

1 2 3 415 15 15 15

+ + - -

+ + + +

- - + +

- 2 - 2

Glucose (mM)

pcDNA3.1(+)

Bim23052 nM )

SST5

GHSR1a

*****A

(

Fig. 3. Expression of an equivalent concentrationof GHS-R1a blocks SST5 agonist-induced inhibitionof GSIS in INS-1SJ cells by a mechanism consistentwith the formation of GHS-R1a:SST5 heteromers. (A)Coexpression of SST5 with an equivalent concen-tration of GHS-R1a blocks suppression of GSIS by theSST5 agonist Bim23052. (B) Tr-FRET using N-terminalSNAP-tagged GHS-R1a illustrates competitive in-hibition of GHS-R1a:GHS-R1a homomer formationby equal (1/1) or excess (1/2) concentrations of SST5.In each case, data represent means ± SEM (n = 4);*P < 0.05, **P < 0.01, ***P < 0.001.

0

10

20

30

40

SST (nM)

cA

MP

nm

ol/

L/p

er

we

ll GHS-R1a + SST5

0 0.1 1 10 100 10000.000.00

0.04

0.08

0.12

0 4 8 12 18RLuc::GFP

BR

ET

Rat

io

SST5-RLuc/SST5-GFP

GHSR1a-RLuc/SST5-GFP

R² = 0.84

GHSR1a-RLuc/SST2-GFP

BR

ET

Rat

io

0.000.00

0.04

0.08

0.12

0 4 8 12 18RLuc::GFP

A

B

5

10

15

20

SST (nM)

cA

MP

nm

ol/

L/p

er

we

ll

0 0.1 1 10 1000

GHS-R1a + SST2

C

D

Fig. 4. BRET analyses and cAMP assays illustratingthat GHS-R1a forms constitutive heteromers withSST5 but not with SST2 and blocks SST/SST5-inducedsuppression of cAMP accumulation. The BRET ratiois expressed as a function of the acceptor/donorDNA ratio and is defined as [(emission at 515 nm) −(background emission at 515 nm)]/[(emission at410 nm) − (background emission at 410 nm)]. (A)The BRET ratio in HEK293 cells cotransfected with0.1 μg GHS-R1a-Rluc and increasing amounts (0.1–1.8 μg) of SST5-GFP (upper curve; BRET50= 2) or 0.1 μgSST5-Rluc DNA and increasing amounts (0.1–1.8 μg)of SST5-GFP DNA (lower curve). (B) The BRET ratiomeasured in HEK293 cells cotransfected with 0.1 μgGHS-R1a-Rluc and 0.1–1.8 μg ST2-GFP. (C) HEK293cells coexpressing GHS-R1a and SST5 are refractoryto SST suppression of forskolin-induced cAMP ac-cumulation. (D) SST inhibition of forskolin-inducedcAMP accumulation in HEK293 cells coexpressingGHS-R1a and SST2.

Park et al. PNAS | November 13, 2012 | vol. 109 | no. 46 | 19005

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

June

2, 2

020

either the energy donor luciferase or GFP as acceptor (27). Afixed DNA concentration of donor and increasing concentrationsof acceptor at ratios of 1/1–1/18 were transfected into HEK293cells, and the BRET ratio was plotted as a function of the relativeconcentrations of donor and acceptor. The hyperbolic functionobtained with the GHS-R1a-Rluc and SST5-GFP pair and theBRET50 ratio indicates high-affinity interactions (upper curve inFig. 4A; BRET50 ∼2). An identical result is obtained when donor-and acceptor-tagged receptors are reversed (SST5-Rluc and GHS-R1a-GFP). In contrast, titration of SST5-Rluc + SST5-GFP(lower curve in Fig. 4A) and GHS-R1a-Rluc + SST2-GFP (Fig.4B) produces linear BRET functions indicating random collisionsrather than heteromer formation. We next asked whether het-eromer formation correlated with resistance to SST inhibition ofcAMP accumulation. Cells coexpressing GHS-R1a with eitherSST5 or SST2 were treated with forskolin and SST. With SST5coexpression, SST did not inhibit cAMP accumulation (Fig. 4C),whereas with SST2 coexpression SST inhibited forskolin-inducedcAMP accumulation (Fig. 4D). Hence, as observed in INS1SJ-GHSR-SST5 cells, formation of GHS-R1a:SST5 heteromersblocks SST responsiveness in favor of ghrelin responsiveness.

[Ghrelin]/[SST] Ratio Modifies GHS-R1a–G Protein Coupling and IsAssociated with Heteromer Formation. In vivo, the concentrationof ghrelin is increased during an acute fast and is suppressedafter a meal. The postprandial increase in blood glucose sup-presses ghrelin and releases SST in pancreatic islets. To de-termine if GHS-R1a coupling to Gαi/o vs. Gαq11 is dependent onthe relative concentrations of ghrelin and SST, we asked if the[ghrelin]/[SST] ratio (1,000/1–1/1) influences signal transduction.High [ghrelin]/[SST] ratios result in ghrelin suppression of cAMPaccumulation, whereas lowering the ratio antagonizes ghrelininhibition of cAMP accumulation (Fig. 5A). To test if SST an-tagonism of ghrelin inhibition of cAMP accumulation is ac-companied by restoration of canonical GHS-R1a–Gαq11 coupling,HEK293 cells expressing GHS-R1a, SST5, and aequorin wereused (1, 28). Consistent with an equilibrium between GHS-R1acoupling to Gαi/o and Gαq11, lowering the [ghrelin]/[SST] ratioproduces a dose-dependent increase in the internal calciumconcentration ([Ca2+]i) mobilization as measured by aequorinbioluminescence (Fig. 5B). Next we performed BRET assaysin the absence or presence of ghrelin with increasing SST con-centrations to test whether the specificity of GHS-R1a–G-pro-tein coupling correlates with heteromer formation. Ghrelinmarkedly increased the BRET ratio which was dose-dependentlyattenuated by SST (Fig. 5C). These results are consistent withagonist concentration-dependent equilibrium between GHS-R1a:SST5 and GHS-R1a:GHS-R1a dimers, which in turn regu-late noncanonical and canonical GHS-R1a signal transduction,respectively.

DiscussionBased on our results, we propose a model of regulation of pan-creatic islet function incorporating endogenous ghrelin and SST.Although GHS-R1a normally couples to Gαq11, both ghrelin andSST inhibit GSIS secretion via GHS-R1a and SST5 coupling toGαi/o. To test conclusively for GHS-R1a and SST5 molecularinteractions and to deduce how combinations of ghrelin and SSTmight regulate signal transduction differentially, we identifieda clone of INS-1 cells (INS-1SJ) that formed a basis for ourstudies. By matching concentrations of GHS-R1a and SST5 inINS-1SJ cells to those expressed in pancreatic islets, we concludethat, contrary to current belief, ghrelin rather than SST suppressesGSIS and cAMP accumulation through a PTX-sensitive mecha-nism. This outcome is surprising, because canonical ghrelin sig-naling through GHS-R1a (via Gαq11) would mobilize [Ca2+]i andstimulate rather than inhibit insulin secretion. Indeed, transducingINS-1SJ cells with lentivirus encoding GHS-R1a enhanced insulin

secretion. However, when GHS-R1a is coexpressed with SST5 inINS-1SJ cells at concentrations and relative ratios commensuratewith those measured in native rat islets, ghrelin inhibits insulinsecretion, supporting our hypothesis of physiologically relevantmolecular interactions between GHS-R1a and SST5.To determine the mechanism by which ghrelin inhibits insulin

secretion via noncanonical GHS-R1a–G-protein coupling, we testedfor GHS-R1a:SST5 heteromer formation using highly sensitiveTr-FRET and BRET assays. Tr-FRET used receptors tagged atthe N terminus, and for BRET receptors were tagged at the Cterminus. Each method used relative and absolute expression levels

05

152535

45

0 0.1 1 10 100SST (nM)

cAM

P nm

ol/L

/per

wel

l

- ghrelin100 nM ghrelin

* * **

A

0

0.05

0.1

0.15

0.2

0.25

100SST (nM)

Frac

tiona

l Lum

ines

cenc

e

10 nM ghrelin

*

*B

101

0

0.1

0.20.30.40.5

BR

ET

Rat

io Control 100 nM ghrelin*

**

SST (nM)

*

C

10 10 100

0

Fig. 5. In HEK293 cells coexpressing GHS-R1a and SST5, GHS-R1a couplingto Gαi/o vs. Gαq11 is dependent on the [ghrelin]/[SST] ratio, which correlateswith GHS-R1a:SST5 heteromer formation. (A) Decreasing the [ghrelin]/[SST]ratio antagonizes ghrelin suppression of forskolin-induced cAMP accumu-lation. (B) HEK293-AEQ17 cells coexpressing GHS-R1a and SST5. Decreasingthe [ghrelin]/[SST] ratio increases ghrelin-induced [Ca2+]i. (C) HEK293 cellscotransfected with GHS-R1a-Rluc (0.2 μg) and SST5-GFP (1.8 μg). A high[ghrelin]/[SST] ratio increases the BRET ratio; a low [ghrelin]/[SST] ratiodecreases the BRET ratio. The data represent the means ± SD of three in-dependent experiments, each performed in triplicate. *P < 0.05 comparedwith absence of both agonists.

19006 | www.pnas.org/cgi/doi/10.1073/pnas.1209590109 Park et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020

of GHS-R1a and SST5 closely matching those in rat pancreaticislets. The results of the two different experimental methods arein complete agreement and illustrate high-affinity interactionsbetween GHS-R1a and SST5 at equimolar concentrations. In-deed, these results support our hypothesis that the atypical GHS-R1a–G-protein coupling essential for ghrelin suppression of GSISis a consequence of the formation of GHS-R1a:SST5 heteromers.Most importantly, we show that differential GHS-R1a–G-proteincoupling and signal transduction are influenced by the relativeconcentrations of ghrelin and SST. In cells coexpressing GHS-R1a and SST5, when the relative [ghrelin]/[SST] ratio is high,ghrelin suppresses cAMP accumulation; when the ratio is low,the cells are refractory to ghrelin inhibition of cAMP production,and ghrelin increases [Ca2+]i mobilization, consistent with res-toration of GHS-R1a–Gαq11 coupling. Furthermore, reducingthe [ghrelin]/[SST] ratio lowers the BRET ratio, consistent withdissociation of GHS-R1a:SST5 heteromers, thereby implicatingan agonist concentration-dependent equilibrium model of het-eromerization. Of course, we cannot preclude the possibility thatchanges in the BRET ratio also might be explained by a changein conformation of the GHS-R1a:SST5 heteromer resulting inaltered G-protein coupling.Constitutive formation of GHS-R1a:SST5 heteromers explains

why GHS-R1a antagonizes SST5 suppression of GSIS and cAMPaccumulation. Because SST-induced formation of SST5:SST5homomers is required for signal transduction (23), the insensitivityto SST argues that formation of SST5:SST5 or GHS-R1a:GHS-R1a homomers is energetically unfavorable compared with GHS-R1a:SST5 formation. Indeed, we speculate that the heteromeracts as a buffer preventing ghrelin-induced insulin secretion byGHS-R1a–Gαq coupling and oversuppression of insulin release byunopposed SST5 coupling to Gαi/o.We propose a model that defines a role for ghrelin and GHS-

R1a for controlling glucose homeostasis according to energybalance (Scheme 1). We show that when GHS-R1a and SST5are coexpressed, they are functionally and physically associated,resulting in coupling to Gαi/o. The physical association of GHS-R1a and SST5 allows signal transduction to be regulated tightlyaccording to the relative concentrations of ghrelin and SST.Under conditions of low energy balance, ghrelin concentrationsare elevated, resulting in GHS-R1a:SST5–mediated inhibitionof insulin secretion. Ghrelin also stimulates glucagon secretionby direct effects on α-cells and indirectly by disinhibiting insulin

suppression of α-cell activity (29–31). As well as inhibiting GSIS,ghrelin inhibits arginine-induced SST secretion (32). In the con-text of positive energy balance, high glucose lowers circulatingghrelin (14, 33), thereby relieving the ghrelin suppression of in-sulin release from β-cells and stimulating the release of SST fromδ-cells (34); SST, in turn, activates SST2 on α-cells, suppressingglucagon secretion (34). Therefore, only in states of low energybalance does ghrelin play a dominant role in regulating glucosehomeostasis.Recent studies show that glucose inhibits glucagon secretion in

WT mice but not in sst−/− mice, as is consistent with a direct rolefor glucose on SST secretion (33). The rapid cessation of insulinrelease upon glucose removal is the same in WT and sst−/− mouseislets, indicating that termination of the insulin-secretory responseis not regulated by SST from δ-cells. This result is consistent withour model, which proposes that when glucose levels fall in vivo, itis ghrelin rather than SST that establishes an inhibitory set pointon the β-cell to control insulin secretion (Scheme 1). This effect isnot observed in isolated mouse islets after glucose removal because,unlike rat and human islets, mouse islets do not produce ghrelin.In summary, we provide evidence for an equilibrium model in

which the formation of GHS-R1a:SST5 heteromers is controlledby the [ghrelin]/[SST] ratio; hence, ghrelin and SST differentiallyestablish a set point for β-cell insulin secretion. Of course, im-plicit in our equilibrium model is dependence on the relativeconcentrations of GHS-R1a and SST5. In rat islets, we show bothare expressed at approximately equal concentrations. Clearly theresults obtained from model systems do not necessarily equate toprimary β-cells; nevertheless, the proposed mechanism is con-ceptually consistent with observations made in isolated islets. Fi-nally, our results reinforce the notion that regulation of intracellularsignaling by GPCRs is not simply a linear relationship but isdependent on modulation via receptor signaling networks thatcan be modified according to relative agonist concentrations andrelative GPCR concentrations.

Experimental ProceduresExpression Constructs. GHS-R1a-GFP was constructed as described previously(27). The GHS-R1a-Rluc:human GHS-R1a cDNA fragment was inserted in-frame into the EcoRI/EcoRV sites of the pRluc-N1 vector (Perkin Elmer). Hu-man SST5 and SST2 cDNA constructs were purchased from Missouri S&TcDNA Resource Center (www.cdna.org). The SST5-GFP:SST5 cDNA fragmentwas inserted in-frame into the EcoRI/EcoRV sites of the pGFP-N3 vector

ββ-cell α -cell δ -cell

LowGlucose

High

InsulinGhrelin

InsulinGhrelin

Somatosta�n Glucagon

Insulin

Ghr

Gαi/o

GHS-R :SST5SST5GHS-R

Gαq11

High [Ghr]/[SST]Low [Ghr]/[SST]Low Energy BalanceHigh Energy Balance

Low Energy Balance

High Energy Balance

Low [SST]

High [SST]

High [Ghr]

Low [Ghr]

Scheme 1. Equilibrium model of ghrelin regula-tion of islet function mediated by GHS-R1a:SST5heteromers. Our data support a model of ghrelinsignaling dependent on the ratio of ghrelin to SSTaccording to energy balance. Low glucose increasesghrelin concentrations; ghrelin suppresses insulinsecretion from β-cells and increases glucagon re-lease from α-cells. High glucose reduces ghrelin butincreases SST release from islet δ-cells. Hence, whenglucose is low, the [ghrelin]/[SST] ratio is high. Inthis context, GHS-R1a activation by ghrelin results innoncanonical coupling to Gαi/o that lowers cAMPaccumulation and inhibits insulin secretion, whichis dependent on GHS-R1a:SST5 heteromer forma-tion. Conversely, by lowering the [ghrelin]/[SST]ratio, high glucose destabilizes GHS-R1a:SST5 het-eromers, reducing GHS-R1a coupling to Gαi/o; as aconsequence, insulin secretion no longer is regu-lated by ghrelin.

Park et al. PNAS | November 13, 2012 | vol. 109 | no. 46 | 19007

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

June

2, 2

020

(Perkin Elmer). The SST5-Rluc:SST5 cDNA fragment was inserted in-frameinto the EcoRI/EcoRV sites of the pRluc-N1vector (Perkin Elmer). The integrityof all tagged GHS-R1a, SST5, and SST2 expression constructs was confirmedby nucleotide sequencing and tested to check that they displayed functionalcharacteristics identical to those of WT receptors in transfected cells. SNAP-GHS-R1a was prepared as described previously (26).

INS-1SJ Cell Line and GSIS. INS-1 832/13 cells that exhibit robust GSIS derivedfrom a rat insulinoma were a kind gift from Chris Newgard (Duke UniversityMedical Center, Winston Salem, NC). A subclone of these cells, which wenamed “INS-1SJ,” was selected specifically for our experiments because itallowed testing of the hypothesis that ghrelin attenuation of GSIS is de-pendent on GHS-R1a and SST5 by independently matching expression levelsof GHS-R1a and SST5 to match levels in rat pancreatic islets. Eighteen hoursbefore GSIS experiments, the standard tissue culture medium for INS-1 cellscontaining 11.1 mM glucose was changed to fresh medium containing5 mM glucose (35). Insulin secretion was assayed in HBSS (114 mM NaCl,4.7 mM KCl, 1.2 mM KH2PO4, 116 mM MgSO4, 20 mM Hepes, 2.5 mM CaCl2,25.5 mM NaHCO3, and 0.2% BSA, pH 7.2). At confluence, cells were washedwith 1 mL HBSS containing 3 mM glucose followed by a 2-h preincubation in2 mL of the same buffer. Then insulin secretion was measured after staticincubation or 2 h in 0.8 mL of HBSS containing the glucose concentrationsand/or ghrelin, SST, or Bim23052 as indicated in the figures. Insulin wasmeasured using the rat/mouse insulin ELISA kit (Linco Research). Acid alcoholwas added to extract total protein, which was assayed using the Coomassieprotein assay reagent kit (Pierce Biotechnology). Secreted insulin was nor-malized to total protein.

Measurements of Intracellular cAMP. cAMPwas measured in INS-1SJ cells usingthe cAMP dynamic 2 kit (Cisbio) using homogeneous time-resolved fluores-cence technology. cAMP levels were calculated by fluorescence ratio (665 nm/620 nm). Forty-eight hours after transfection, the transfected HEK293 cells were

detached using nonenzymatic cell dissociation solution (Sigma), were washedwith HBSS, and were resuspended in stimulation buffer [1× PBS with 0.04%BSA (pH 7.4) and 0.5 mmol/L 3-isobutyl-1-methylxanthine]. cAMP accumu-lation also was determined using the Lance cAMP 384 kit from Perkin Elmer.

Live-Cell Tr-FRET and BRET. Cells were labeled 48 h after being transfected byelectroporation as described previously (26). Briefly, SNAP-GHS-R1a–trans-fected cells were incubated in the presence of 25 nM of donor benzyl guanine-conjugated terbium cryptate (BG-TbK) (SNAP-Lumi4-Tb; Cisbio) and 250 nM ofacceptor benzyl guanine-conjugated d2 fluorophore (BC-647) (SNAP-surface647; New England BioLabs) for 1 h at 37 °C in 0.5% FBS containing RPMImedium. Tr-FRET signals were measured at 665 nm and 620 nm after excita-tion at 337 nm with 50-μs delay and integration time of 400 μs by using anEnVision plate reader (Perkin Elmer). The ratio of fluorescence (665 nm/620nm) was calculated for each sample. BRET experiments were based on previousstudies (27). Details are given in SI Experimental Procedures.

Aequorin Bioluminescence Assay. The aequorin bioluminescence assay wascarried out as described in ref. 28. Human ghrelin and SST (Phoenix) werediluted with modified HBSS (25 mM Hepes at pH 7.3) and distributed into96-well plates. Assays were performed with the Luminoskan Luminometer(ThermoFisher Labsystems). The fractional luminescence for each well wascalculated by taking the ratio of the integrated response to the initialchallenge to the total integrated luminescence, as well as the Triton X-100lysis response. Fractional luminescence data for each point represent theaverage of triplicate measurements. Statistical analyses were performedusing the Student t test.

Additional methods are described in SI Experimental Procedures.

ACKNOWLEDGMENTS. This work was supported by National Institutes ofHealth/National Institute of Aging Grant R01 AG019230 (to R.G.S.).

1. Howard AD, et al. (1996) A receptor in pituitary and hypothalamus that functions ingrowth hormone release. Science 273(5277):974–977.

2. Smith RG, et al. (1997) Peptidomimetic regulation of growth hormone secretion.Endocr Rev 18(5):621–645.

3. Kojima M, et al. (1999) Ghrelin is a growth-hormone-releasing acylated peptide fromstomach. Nature 402(6762):656–660.

4. Sun Y, Wang P, Zheng H, Smith RG (2004) Ghrelin stimulation of growth hormonerelease and appetite is mediated through the growth hormone secretagogue re-ceptor. Proc Natl Acad Sci USA 101(13):4679–4684.

5. Sun Y, Asnicar M, Saha PK, Chan L, Smith RG (2006) Ablation of ghrelin improves thediabetic but not obese phenotype of ob/ob mice. Cell Metab 3(5):379–386.

6. Sun Y, Asnicar M, Smith RG (2007) Central and peripheral roles of ghrelin on glucosehomeostasis. Neuroendocrinology 86(3):215–228.

7. Sun Y, Butte NF, Garcia JM, Smith RG (2008) Characterization of adult ghrelin andghrelin receptor knockout mice under positive and negative energy balance. Endo-crinology 149(2):843–850.

8. Szentirmai E, Kapás L, Sun Y, Smith RG, Krueger JM (2009) The preproghrelin gene isrequired for the normal integration of thermoregulation and sleep in mice. Proc NatlAcad Sci USA 106(33):14069–14074.

9. Gutierrez JA, et al. (2008) Ghrelin octanoylation mediated by an orphan lipid trans-ferase. Proc Natl Acad Sci USA 105(17):6320–6325.

10. Yang J, Brown MS, Liang G, Grishin NV, Goldstein JL (2008) Identification of theacyltransferase that octanoylates ghrelin, an appetite-stimulating peptide hormone.Cell 132(3):387–396.

11. Zhao TJ, et al. (2010) Ghrelin O-acyltransferase (GOAT) is essential for growth hor-mone-mediated survival of calorie-restricted mice. Proc Natl Acad Sci USA 107(16):7467–7472.

12. Nass RM, Gaylinn BD, Rogol AD, Thorner MO (2010) Ghrelin and growth hormone:Story in reverse. Proc Natl Acad Sci USA 107(19):8501–8502.

13. Gromada J, Franklin I, Wollheim CB (2007) Alpha-cells of the endocrine pancreas: 35years of research but the enigma remains. Endocr Rev 28(1):84–116.

14. Dezaki K, Kakei M, Yada T (2007) Ghrelin uses Galphai2 and activates voltage-de-pendent K+ channels to attenuate glucose-induced Ca2+ signaling and insulin releasein islet beta-cells: Novel signal transduction of ghrelin. Diabetes 56(9):2319–2327.

15. Brazeau P, et al. (1973) Hypothalamic polypeptide that inhibits the secretion of im-munoreactive pituitary growth hormone. Science 179(4068):77–79.

16. Patel YC, Greenwood MT, Warszynska A, Panetta R, Srikant CB (1994) All five clonedhuman somatostatin receptors (hSSTR1-5) are functionally coupled to adenylyl cyclase.Biochem Biophys Res Commun 198(2):605–612.

17. Mitra SW, et al. (1999) Colocalization of somatostatin receptor sst5 and insulin in ratpancreatic beta-cells. Endocrinology 140(8):3790–3796.

18. Hoyer D, et al. (1995) Classification and nomenclature of somatostatin receptors.Trends Pharmacol Sci 16(3):86–88.

19. Rossowski WJ, Coy DH (1994) Specific inhibition of rat pancreatic insulin or glucagonrelease by receptor-selective somatostatin analogs. Biochem Biophys Res Commun205(1):341–346.

20. Rohrer SP, et al. (1998) Rapid identification of subtype-selective agonists of the so-matostatin receptor through combinatorial chemistry. Science 282(5389):737–740.

21. Yang L, et al. (1998) Synthesis and biological activities of potent peptidomimeticsselective for somatostatin receptor subtype 2. Proc Natl Acad Sci USA 95(18):10836–10841.

22. Zambre Y, et al. (1999) Inhibition of human pancreatic islet insulin release by re-ceptor-selective somatostatin analogs directed to somatostatin receptor subtype 5.Biochem Pharmacol 57(10):1159–1164.

23. Rocheville M, et al. (2000) Subtypes of the somatostatin receptor assemble as func-tional homo- and heterodimers. J Biol Chem 275(11):7862–7869.

24. Gautier A, et al. (2008) An engineered protein tag for multiprotein labeling in livingcells. Chem Biol 15(2):128–136.

25. Maurel D, et al. (2008) Cell-surface protein-protein interaction analysis with time-resolved FRET and snap-tag technologies: Application to GPCR oligomerization. NatMethods 5(6):561–567.

26. Kern A, Albarran-Zeckler R, Walsh HE, Smith RG (2012) Apo-ghrelin receptor formsheteromers with DRD2 in hypothalamic neurons and is essential for anorexigeniceffects of DRD2 agonism. Neuron 73(2):317–332.

27. Jiang H, Betancourt L, Smith RG (2006) Ghrelin amplifies. dopamine signaling bycrosstalk involving formation of GHS-R/D1R heterodimers. Mol Endocrinol 20(8):1772–1785.

28. Feighner SD, et al. (1998) Structural requirements for the activation of the humangrowth hormone secretagogue receptor by peptide and nonpeptide secretagogues.Mol Endocrinol 12(1):137–145.

29. Qader SS, Lundquist I, Ekelund M, Håkanson R, Salehi A (2005) Ghrelin activatesneuronal constitutive nitric oxide synthase in pancreatic islet cells while inhibitinginsulin release and stimulating glucagon release. Regul Pept 128(1):51–56.

30. Chuang JC, et al. (2011) Ghrelin directly stimulates glucagon secretion from pancre-atic alpha-cells. Mol Endocrinol 25(9):1600–1611.

31. Kawamori D, et al. (2009) Insulin signaling in alpha cells modulates glucagon secre-tion in vivo. Cell Metab 9(4):350–361.

32. Hauge-Evans AC, et al. (2009) Somatostatin secreted by islet delta-cells fulfills mul-tiple roles as a paracrine regulator of islet function. Diabetes 58(2):403–411.

33. Briatore L, Andraghetti G, Cordera R (2003) Acute plasma glucose increase, but notearly insulin response, regulates plasma ghrelin. Eur J Endocrinol 149(5):403–406.

34. Egido EM, Rodriguez-Gallardo J, Silvestre RA, Marco J (2002) Inhibitory effect ofghrelin on insulin and pancreatic somatostatin secretion. Eur J Endocrinol 146(2):241–244.

35. Hohmeier HE, et al. (2000) Isolation of INS-1-derived cell lines with robust ATP-sen-sitive K+ channel-dependent and -independent glucose-stimulated insulin secretion.Diabetes 49(3):424–430.

19008 | www.pnas.org/cgi/doi/10.1073/pnas.1209590109 Park et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020