Genetic Disruption of Adenosine Kinase in MousePancreatic b-Cells Protects Against High-FatDiet–Induced Glucose IntoleranceGuadalupe Navarro,1 Yassan Abdolazimi,1 Zhengshan Zhao,1 Haixia Xu,1,2 Sooyeon Lee,1

Neali A. Armstrong,1 and Justin P. Annes1

Diabetes 2017;66:1928–1938 | https://doi.org/10.2337/db16-0816

Islet b-cells adapt to insulin resistance through increasedinsulin secretion and expansion. Type 2 diabetes typicallyoccurs when prolonged insulin resistance exceeds theadaptive capacity of b-cells. Our prior screening effortsled to the discovery that adenosine kinase (ADK) inhibitorsstimulate b-cell replication. Here, we evaluated whetherADK disruption in mouse b-cells affects b-cell massand/or protects against high-fat diet (HFD)–induced glu-cose dysregulation. Mice targeted at the Adk locus werebred to Rip-Cre and Ins1-Cre/ERT1Lphi mice to enableconstitutive (bADKO) and conditional (ibADKO) disru-ption of ADK expression in b-cells, respectively. Weightgain, glucose tolerance, insulin sensitivity, and glucose-stimulated insulin secretion (GSIS) were longitudinallymonitored in normal chow (NC)–fed and HFD-fed mice.In addition, b-cell mass and replication were measuredby immunofluorescence-based islet morphometry.NC-fed adult bADKO and ibADKOmice displayed glucosetolerance, insulin tolerance and b-cell mass comparableto control animals. By contrast, HFD-fed bADKO andibADKO animals had improved glucose tolerance and in-creased in vivo GSIS. Improved glucose handling was as-sociated with increased b-cell replication and mass. Weconclude that ADK expression negatively regulates theadaptive b-cell response to HFD challenge. Therefore,modulation of ADK activity is a potential strategy for en-hancing the adaptive b-cell response.

Diabetes is a pathologic state of disrupted glucose homeo-stasis characterized by an absolute or relative insulindeficiency and a loss of insulin-producing b-cells. In type2 diabetes (T2D), b-cell failure results from a multifactorialprocess initiated by insulin resistance, often in the setting

of obesity (1–3). In T2D, a variety of insults contribute toprogressive b-cell failure, including endoplasmic reticulumstress, inflammatory cytokines, excess reactive oxygen spe-cies, and glycolipid toxicity (2). b-Cell loss occurs through acombination of increased apoptosis and dedifferentiation,although the relative contribution of these outcomes re-mains unclear (3–6). Presently, a major research goal is tounderstand the molecular mechanisms of b-cell failure anddevise strategies to reverse this process.

Although T2D is accompanied by reduced insulinsecretion in late disease, increased insulin secretion isan early adaptation to insulin resistance (7,8). Of note,individuals without diabetes with a high genetic risk fordiabetes have a reduced glucose-stimulated insulin re-sponse (9), but whether this is a consequence of defectiveb-cell function or deficient b-cell mass is unclear. T2D-associated risk alleles implicate genes that participate inboth processes (e.g., CDKN2A, KCNQ1 [10–12]). Murinestudies have demonstrated a central role for b-cell massplasticity in the accommodation of obesity-associatedinsulin resistance (13). Although adaptive b-cell expansionis less evident in humans, obese humans without diabeteshave a 1.5-fold increase in b-cell mass and increased b-cellnumber (14). Hence, human b-cell mass possibly exhibitsmodest plasticity that influences an individual’s susceptibil-ity to T2D (15).

In mature animals, numerous potential sources of newb-cells have been identified (16); however, the usual sourceof new b-cells is previously existing b-cells (17,18). Conse-quently, understanding the signals that control self-dupli-cation is critical to understanding how b-cell mass iscontrolled. To identify molecular mechanisms that regulate

1Department of Medicine and Division of Endocrinology, Stanford University,Stanford, CA2Department of Endocrinology and Metabolism, Third Affiliated Hospital of SunYat-Sen University, Guangzhou, People’s Republic of China

Corresponding author: Justin P. Annes, jannes@stanford.edu.

Received 5 July 2016 and accepted 24 April 2017.

© 2017 by the American Diabetes Association. Readers may use this article aslong as the work is properly cited, the use is educational and not for profit, and thework is not altered. More information is available at http://www.diabetesjournals.org/content/license.

1928 Diabetes Volume 66, July 2017

ISLETSTUDIES

b-cell growth, we developed a primary islet cell–basedsmall-molecule screening platform (19,20). With this plat-form, we uncovered the b-cell replication-promoting activ-ity of adenosine kinase inhibitors (ADKIs). ADK is a broadlyexpressed metabolic enzyme that controls extracellular andintracellular adenosine pools through its enzymatic activity:conversion of adenosine to AMP (21). Several lines of evi-dence indicate that ADKIs promote b-cell replication, inpart through ADK inhibition: Multiple structurally dissim-ilar ADKIs promote b-cell replication, ADK-directed RNAinterference triggers cell autonomous b-cell replication,and an independent screen for b-cell regeneration-promot-ing compounds identifies distinct ADKIs (19,22). Addi-tional activities of some ADKIs contribute to their b-cellreplication-promoting activity. For example, 5-iodotubercidin(5-IT) has been shown to promote human b-cell replica-tion through inhibition of the dual-specificity tyrosine phos-phorylation-regulated kinase 1A (DYRK1A) (23–25). Toinvestigate the function of ADK in b-cells, we generatedmice conditionally targeted at the ADK locus and testedthe hypothesis that ADK acts as a negative regulator ofb-cell replication and limits the adaptive response ofb-cells to diabetogenic challenge.

RESEARCH DESIGN AND METHODS

Generation, Genotyping, and Feeding ofADK-Targeted MiceAll animal work was approved and carried out in accordancewith our institutional animal care and use committee andthe Guide for the Care and Use of Laboratory Animals. Adk-targeted mice were generated from International KnockoutMouse Consortium clone EUCE0154a03 (26). Genotypingof the following ADK-targeted mice was performed by PCR:wild-type allele (ADK_F [59-AGCCTAGACTACACAACAAG-39] and ADK_R [59-GCTCAATCACCTAGATGGCC-39]), Adk-targeted allele (ADK1) (ADK_F and B32 [59-CAAGGCGATTAAGTTGGGTAACG-39]), and nonmutagenic orienta-tion (ADK2) (B32 and ADK_6898 [59-TCAAGCCCTTTGTACACCCTAAG-39]) (Fig. 1A). The FLP deleter strain(FLPo-10, JAX 011065) was used to convert ADK1 to ADK2.Two Cre-expressing strains were used for constitutive[Tg(Ins2-Cre)25Mgn, JAX 0035731] and conditional (Ins1-Cre/ERT1Lphi, JAX 024709) recombination. For Cre/ERTactivation, tamoxifen (100 mg/kg, #13258; Cayman Chem-ical) was dissolved in 100% ethanol solubilized in sterilewarm corn oil (10 mg/mL, C8267; Sigma) and adminis-tered by intraperitoneal injection on 4 consecutive days.b-Galactosidase activity was used to confirm efficient ta-moxifen-induced recombination. One week after tamoxifeninjection, pancreata were excised and fixed (4% paraformal-dehyde/PBS at 4°C for 30 min), washed with PBS, andplaced in 30% sucrose/PBS for 48 h at 4°C before freezing inoptimal cutting temperature compound (Tissue Tek). Ani-mals were fed normal chow (NC) (2018; Harlan Teklad) or ahigh-fat diet (HFD) (D12492; Research Diets) as indicated.Experiments were conducted in female mice unless stated

otherwise. Mice were of mixed background (129P2/OlaHsdand C57BL/6J).

Western BlottingIsolated islets and homogenized liver tissue lysates wereused for SDS-PAGE (RIPA Lysis System, sc-24948; SantaCruz Biotechnology). Islet-restricted ADK disruption wasassessed by using liver and islet lysates from adult bADKOand control animals. Islets were isolated as previously de-scribed (20). Anti-ADK (1:200, sc-32908) and anti-enolase(1:500, sc-15343) were used for protein detection.

Glucose Physiology ExperimentsAll glucose physiology experiments were performed onage- and sex-matched cohorts. Glucose measurements weremade at the indicated times with a TRUEresult glucometerand TRUEtest strips. For fasting glucose measurementand intraperitoneal glucose tolerance testing (IPGTT),mice were weighed and fasted overnight (14 h, 6:00 P.M.–8:00 A.M.). Glucose levels were obtained at 0 min beforeinjection of glucose 1.5–2.0 g/kg i.p. (10 mL/g) and at thesubsequent indicated time points. Random glucose levelswere obtained for fed mice at 10:00 A.M. Intraperito-neal insulin tolerance testing (IPITT) was performed byusing 0.5–1.5 units/kg Humulin (R-100; Eli Lilly) as indi-cated after a 4-h fast starting at 8:00 A.M. In vivo glucose-stimulated insulin secretion (GSIS) was measured inblood collected from overnight fasted mice at time 0 andafter injection of glucose 3 g/kg i.p. Insulin was mea-sured with Alpco Mouse Ultrasensitive Insulin ELISA kits(80-INSMSU-E01). ADK2/+ Rip-Cre animals displayed nodetectable phenotype and were included with ADK+/+ Rip-Cre as control animals.

HistologyTissue b-galactosidase staining was performed on fresh fro-zen 25-mm sections that were fixed for 5 min with cold 4%paraformaldehyde/PBS and washed in cold PBS. Slides wereincubated in X-gal staining solution containing PBS withpotassium ferricyanide (5 mmol/L, P8131; Sigma), potassiumferricyanide trihydrate (5 mmol/L, P3289; Sigma), magne-sium chloride (2 mmol/L; Promega), and X-gal (20 mg/100mL; Promega) at 37°C. Slides were counterstained as indi-cated. Antigen retrieval was performed by heating slides to90°C for 10 min in 10 mmol/L sodium citrate (pH 6.0)solution. Primary antibodies were incubated overnight at4°C. To measure islet size distribution, pancreatic sectionswere insulin (1:300, A0564; Dako) or glucagon (1:1,000,A0565; Dako) stained along with DAPI. Affinity-purifiedsecondary antibodies (donkey) were obtained from JacksonImmunoResearch (1:300). Entire sections (eight per mouse)were scanned with an inverted Nikon spinning disk confo-cal microscope. Images were stitched together and analyzedwith Volocity 6 image analysis software (PerkinElmer). Cel-lular clusters of .900 mm2 were counted. Total pancreaticarea was determined by the DAPI-positive area. The b-cellmass was calculated by multiplying pancreatic mass by thepercentage of pancreatic area that was insulin positive.

diabetes.diabetesjournals.org Navarro and Associates 1929

b-Cell replication indices were determined by BrdU incor-poration (1:50, M074401-8; Dako) and/or ki67 expression(1:300, 556003; BD Bioscience); ki67 detection was per-formed by using biotin amplification (1:200; Jackson Immuno-Research). b-Cells were identified by the expression ofinsulin or PDX-1 (1:200, AF2419; R&D Systems); use ofinsulin staining to confirm findings obtained with PDX-1 isnecessary given that d-cells also express PDX-1 (27). A min-imum of 2,000 b-cells from nonconsecutive sections(.50 mmol/L apart) were used to determine the b-cellreplication rate for each animal. Replication events were

adjudicated in a blinded fashion. Mice were providedBrdU-containing water (0.8 mg/mL, 228590100; Acros) for1 week in opaque bottles that were changed every 2 days.

In Vitro b-Cell ReplicationIslets were isolated from 36-week vehicle-injected (n = 4)and tamoxifen-injected (n = 4) ibADKO mice. Injectionswere performed at age 18 weeks to avoid any developmen-tal impact, and islets were harvested after an 18-week delayto avoid potential short-term tamoxifen and recombina-tion effects. Islets were dispersed, plated (allowed 60 h to

Figure 1—Conditional disruption of ADK gene expression. A: Schematic representation of the Adk locus and targeting construct: mutagenicorientation (ADK1), Flp recombinase–dependent nonmutagenic orientation (ADK2), and Cre recombinase–dependent mutagenic orientation(ADK3). Forward primer (Fwd), reverse primer (Rev), B32 primer, recombination sequences (Frt and LoxP), SA, and the b-galactosidase/bgeoare shown. B: Histochemical staining of pancreatic sections from Adk-targeted mice for b-galactosidase activity (blue); counterstaining witheosin (pink, top row) or insulin (brown, bottom row) are shown. C: ADK-directed Western blot of hepatic tissue lysates from Adk-targeted andcontrol mice. A nonspecific upper band reflects sample loading (loading control [LC]). D: Liver and islet lysates probed for ADK and enolase (LC).

1930 ADK Negatively Regulates b-Cell Expansion Diabetes Volume 66, July 2017

attach), treated (60 h duration), fixed, and assayed (PDX-1and ki67 staining) as previously described (19,20).

StatisticsExperimental data are presented as scatter plots withan adjacent representation of the statistical mean thatincludes an error bar representing the SD. Statisticallysignificant differences were determined by using Studenttwo-tailed t test where P# 0.05 was taken to be significant.Experimental results were confirmed in independent exper-imentation in all cases except for the in vivo b-cell replica-tion experiments.

RESULTS

Adk-Targeted Mouse ModelTo study the in vivo function of ADK in b-cells, we generatedmice conditionally targeted at the Adk locus (Fig. 1A). Theintegrated mutagenic orientation (ADK1) placed a highlyefficient splice-acceptor sequence (SA)/bgeo expression cas-sette (b-galactosidase-neomycin resistance fusion gene) inframe with exons 1a and 1b, the ADK transcriptional startsites that encode ADK-short (cytoplasmic) and ADK-long(nuclear) isoforms, respectively (28). b-Galactosidase activ-ity was used to assess ADK expression in the pancreas ofAdk1/+ mice (Fig. 1B, left panels). ADK was highly expressedin the exocrine pancreas and modestly expressed in theendocrine pancreas where the nuclear-located isoform(ADK-long) predominates. Newborn litters from heterozy-gous (Adk1/+ 3 Adk1/+) breeding pairs contained Adk1/+

pups at Mendelian frequency. However, consistent withthe published phenotype of ADK-null animals, no ADK1/1

mice were identified at weaning (.150 mice screened) (29).Furthermore, ADK protein was nearly undetectable in he-patocyte lysates from Adk1/1 newborn pups (Fig. 1C). Theseresults indicate efficient disruption of ADK expression bythe targeting strategy.

We disrupted ADK expression in b-cells (bADKO mice)by using the Rip-Cre driver strain (30). We confirmedb-cell–restricted disruption by measuring b-galactosidaseactivity in pancreatic tissue sections from 1) mice harboringthe gene-trap vector in the nonmutagenic orientation(Adk2/+ mice), 2) Rip-Cre mice, and 3) Rip-Cre Adk2/+ mice(bADKO-Het) (Fig. 1B). As anticipated, pancreatic sectionsfrom Adk2/+ and Rip-Cre animals lacked b-galactosidase ac-tivity, whereas bADKO-Het sections displayed b-cell–restricted b-galactosidase activity. Of note, Rip-Cre micehave recombination activity in the brain and exhibit glucoseintolerance (31,32). Consequently, all studies included Rip-Cre control animals. Next, we assessed ADK disruption byWestern blot of liver and islet lysates from Rip-Cre,bADKO-Het, and bADKO mice (Fig. 1D). Whereas hepaticADK expression was similar among genotypes, islet ADKexpression was reduced in bADKO animals. Residual ADKexpression in islet samples might reflect exocrine and/orinsulin-negative islet cell contamination. Taken together,these data indicate successful disruption of ADK expressionin pancreatic b-cells.

bADKO Mice Are Protected Against HFD-InducedGlucose IntoleranceYoung bADKO animals were heavier than Rip-Cre controllittermates (Fig. 2A). At 8 weeks of age, female bADKOmice weighed ;2 g more than Rip-Cre control animals(18.3 6 0.3 vs. 16.1 6 0.6 g; P , 0.01). As mice aged,the body weights of bADKO and Rip-Cre mice converged;no statistical difference was present beyond 14 weeks (P .0.05). By 52 weeks, the body weights of Rip-Cre (28.9 61.4 g) and bADKO (29.2 6 1.5 g) were indistinguishable.HFD-fed 13-week bADKO and Rip-Cre mice demonstratedincreased weight gain compared with genotype-matchedNC-fed mice beginning 2 weeks after HFD initiation (P ,0.05 for weeks 15–21). HFD-induced weight gain in Rip-Creand bADKO mice was similar (P . 0.05 at every timepoint) (Fig. 2B). Thus, bADKO animals were transientlyheavier than control animals but not predisposed to in-creased HFD-induced weight gain.

We next assessed the impact of b-cell–targeted ADKdisruption on glucose tolerance by performing IPGTTs ev-ery 4–8 weeks on NC-fed Rip-Cre and bADKO mice. Nodifference was observed until 52 weeks when subtle im-provement in the IPGTT of bADKO mice emerged (Fig.2C and D). To assess a potential protective role of ADKdisruption against diabetogenic challenge, IPGTTs were per-formed on Rip-Cre and bADKO mice placed on an HFD.After only 2 weeks of HFD, female bADKO mice displayedlower fasting glucose levels but similar glucose tolerance(Fig. 2E). After 6 weeks and 18 weeks of HFD, femalebADKO mice demonstrated substantial improvement inIPGTT (Fig. 2F and G). A similar improvement in IPGTTwas also observed in HFD-fed male bADKO mice (Fig. 2H).Fasting and random-fed glucose values were also signifi-cantly improved in HFD-fed bADKO mice (Fig. 2I and J).Thus, bADKO animals were resistant to HFD-dependentimpairment of glucose homeostasis.

HFD-Fed bADKO Mice Demonstrate Enhanced InsulinSensitivity and Insulin SecretionTo assess the physiologic basis of improved IPGTT in HFD-fed bADKO mice, we performed IPITT (Fig. 3A and B). Asanticipated, bADKO fasting glucose values were lower thanthat of controls (154 6 19 vs. 214 6 34 mg/dL; P , 0.01).Consequently, we compared glucose clearance by bADKOand Rip-Cre animals with (Fig. 3A) and without (Fig. 3B)normalizing to starting glucose values. bADKO mice dis-played more-robust insulin responsiveness at 30 and60 min after insulin injection. However, the absolute dropin blood glucose levels in response to insulin was greater forRip-Cre mice than for bADKO mice (156 6 29 vs. 124 618 mg/dL, respectively; P , 0.05). These results indicatethat HFD-fed bADKO mice had modestly enhanced insulinsensitivity, but given the subtlety of this difference, otherfactors were likely contributing to the improved glucosetolerance of the bADKO mice.

We next evaluated whether enhanced insulin secretioncontributed to the improved glucose tolerance of bADKO

diabetes.diabetesjournals.org Navarro and Associates 1931

Figure 2—Constitutive deletion of ADK in b-cells enhanced glucose tolerance. A: Body weights of NC- and HFD-fed Rip-Cre and bADKO femalemice (n = 8 mice/group). Statistical comparisons: Rip-Cre (NC) vs. bADKO (NC), P < 0.05 for weeks 8 and 13 only; Rip-Cre (HFD) vs. bADKO(HFD), P > 0.05 for all time points; Rip-Cre (NC) vs. Rip-Cre (HFD), P < 0.05 for weeks 15–20; bADKO (NC) vs. bADKO (HFD), P < 0.05 forweeks 15–20. B: Weight gain of Rip-Cre and bADKO mice on HFD (no statistical differences detected). C: IPGTT of 13-week NC-fed Rip-Creand bADKO mice (n = 15 mice/group; P > 0.05 for all time points). D: IPGTT of 52-week NC-fed Rip-Cre and bADKO mice (n = 8 mice/group).*P < 0.05 at 120 min (P > 0.05 for all other time points). E: IPGTT of Rip-Cre and bADKO mice after 2 weeks of HFD (age 15 weeks;n = 7 mice/group). *P < 0.05 at 0 min (P > 0.05 for all other time points). F: IPGTT of female Rip-Cre and bADKO mice after 6 weeks ofHFD (age 19 weeks; n = 7 mice/group). *P < 0.05 at 30, 60, and 90 min. G: IPGTT of female Rip-Cre and bADKO mice after 18 weeks of HFD(age 31 weeks; n = 7 mice/group). *P< 0.05 at 30 and 60 min. H: IPGTT of male Rip-Cre and bADKOmice after 6 weeks of HFD (age 19 weeks;n = 7 mice/group). *P < 0.05 at 30, 60, and 120 min. I: Fasting glucose values of female Rip-Cre and bADKO mice after 23 weeks of HFD (age30 weeks; n = 8 mice/group). *P < 0.01. J: Random glucose values of female Rip-Cre and bADKO mice after 12 weeks of HFD (age 25 weeks;n = 8 mice/group). *P < 0.01.

1932 ADK Negatively Regulates b-Cell Expansion Diabetes Volume 66, July 2017

mice by measuring GSIS in NC- and HFD-fed Rip-Cre andbADKO mice (Fig. 3C). As anticipated HFD-fed mice dis-played increased insulin secretion relative to NC-fed mice.Both NC- and HFD-fed bADKO mice displayed increasedinsulin secretion compared with Rip-Cre controls (Fig. 3Cand D). The increased GSIS of NC-fed mice was unexpectedbecause no difference in IPGTT was detected. However, thisdifference was subtle (NC area under the curve [AUC] Rip-Cre 8.1 6 2.1 vs. bADKO 13.6 6 1.9; P , 0.05) comparedwith HFD-fed mice (HFD AUC Rip-Cre 32.31 6 3.7 vs.bADKO 61.4 6 9.9; P , 0.05).

b-Cell Expansion Protects bADKO Mice FromHFD-Induced Glucose IntoleranceIslet isolations from HFD-fed bADKO mice demonstratedincreased yield (1.3 6 0.2-fold; P , 0.01) compared withRip-Cre mice (Fig. 4A). This observation led us to considertwo potential explanations: 1) HFD-fed bADKO mice hadmore islets, or 2) HFD-fed bADKO mice had larger isletsthat were more efficiently isolated. To assess these possibil-ities, islet size and number were analyzed in pancreaticsections from HFD-fed Rip-Cre and bADKO animals. Visualinspection of insulin-stained sections gave the impressionthat bADKO islets were enlarged (Fig. 4B). To quantitate

this impression, we measured the insulin-stained area. Con-sistent with visual observation, bADKO animals demon-strated a right shift in insulin cluster size (Fig. 4C). Inaddition, the average insulin-positive cluster size was in-creased in bADKO mice (average insulin area, bADKO7,790 6 694 vs. Rip-Cre 3,857 6 952 mm2; P , 0.01)(Fig. 4D). Furthermore, the insulin-positive area andb-cell mass but not pancreatic area or pancreatic masswere increased in bADKO mice (Fig. 4E–I). No increase inthe number of insulin-positive clusters was detected (Rip-Cre 6.34 6 1.9 vs. bADKO 7.6 6 0.4; P = 0.4) (Fig. 4J). Nochange in a-cell area was detected (Fig. 4K), and b-cell pro-liferation in HFD-fed mice, as measured by BrdU incorpo-ration, demonstrated a trend toward an increase in thebADKO mice (Fig. 4L). Taken together, these results indi-cate that b-cell expansion rather than increased islet num-ber was present in HFD-fed bADKO mice.

Disruption of ADK Expression in Mature b-Cells ProtectsAgainst HFD-Induced Hyperglycemia and Promotesb-Cell ReplicationTo mitigate deficiencies of the Rip-Cre transgene Tg(Ins2-Cre)25Mgn, we reevaluated our findings with the inducibleand b-cell–specific Cre-expressing Mip-Cre/ERT mouse line

Figure 3—HFD-fed bADKOmice have enhanced insulin tolerance and GSIS in vivo. A and B: Normalized and raw glucose values from 20-weekHFD-fed female Rip-Cre and bADKOmice subjected to an IPITT (n = 7 mice/group). C: In vivo GSIS of 24-week NC- or HFD-fed female Rip-Creand bADKO mice (n = 5–8 mice/group). Significant differences are for comparisons made between Rip-Cre and bADKO mice on the same dietindicated. Insulin levels are significantly higher at all time-points in HFD-fed mice (P < 0.05). D: The total insulin secretion (AUC) is calculatedfrom C. *P < 0.05.

diabetes.diabetesjournals.org Navarro and Associates 1933

Figure 4—HFD-fed bADKOmice demonstrate increased b-cell but not a-cell mass. A: Relative number of islets isolated from 28-week (21 weeksof HFD) female HFD-fed Rip-Cre and bADKOmice (n = 8). The average number of islets isolated per mouse was 107.56 10.1 and 141.76 17.0from HFD-fed Rip-Cre and bADKO mice, respectively. B: Representative images of 24-week female HFD-fed Rip-Cre– and bADKO-derivedpancreatic sections stained for DAPI (blue) and insulin (red). C: Size distribution of insulin-positive area obtained from Rip-Cre and bADKOpancreatic sections (minimum of six per mouse) stained as in B (n = 5 mice/group). D and E: Average and total insulin-positive cluster area (n =5 mice/group). F: Total pancreas area measured on the basis of DAPI staining. Data are mean 6 SEM. G: Percentage of pancreatic area (DAPI)that is insulin-positive (n = 5). H: Average pancreatic weight (n = 5). I: Calculated b-cell mass by using data obtained from G and H (n = 5).J: Number of insulin-positive clusters per pancreatic section (no statistical difference detected). K: Percentage of total pancreatic area (DAPI)that costains for glucagon. L: b-Cell replication in HFD-fed Rip-Cre and bADKO mice measured as BrdU-positive cells per square micrometerof insulin staining (n = 5 mice/group, 10 sections per mouse; P = 0.06). *P < 0.05.

1934 ADK Negatively Regulates b-Cell Expansion Diabetes Volume 66, July 2017

(Ins1-Cre/ERT1Lphi) (31,32). Because Mip-Cre/ERT micehave tamoxifen-independent growth hormone expression,we used vehicle-treated ibADKO mice as control subjects(33).

We evaluated the impact of tamoxifen-induced b-cell–specific ADK disruption on glucose homeostasis andb-cell replication in mature animals (Fig. 5A). Tamoxifen-dependent disruption of ADK expression was confirmedby induction of b-galactosidase activity (Fig. 5B). Similar toNC-fed bADKO mice, NC-fed vehicle- and tamoxifen-treat-ed ibADKO mice demonstrated comparable IPGTTs (Fig.5C). Consequently, tamoxifen treatment had no impacton the IPGTT. Also consistent with the bADKO phenotype,HFD-fed tamoxifen-treated mice demonstrated significantlylower fed glucose values (Fig. 5D) and enhanced IPGTT after3 (Fig. 5E) and 11 (Fig. 5F) weeks of HFD. By contrast,insulin sensitivity measured by IPITT was unchanged (Fig.5G), perhaps indicating an impact of ectopic recombinaseactivity in bADKO mice. Finally, we assessed whether dis-ruption of ADK enhanced in vivo GSIS. Indeed, tamoxifen-treated mice demonstrated increased insulin secretion afterglucose challenge (15-min vehicle treatment 0.156 0.02 vs.tamoxifen-treatment 0.24 6 0.04 ng/mL; P = 0.01) (Fig.5H). These data indicate that b-cell–specific disruption ofADK in adult mice was protective against HFD-inducedglucose intolerance and enhanced GSIS in vivo.

We next assessed the impact of ADK disruption onmature b-cell proliferation. We determined the percentageof insulin-positive cells that coexpressed ki67 in vehicle-and tamoxifen-treated mice fed either NC or HFD. Lossof ADK expression had no effect on b-cell replication inNC-fed mice (replication of vehicle 0.36 6 0.1% vs. tamox-ifen 0.50 6 0.10%; P = 0.38); however, ki67 expression wassignificantly increased in HFD-fed tamoxifen-treated mice(vehicle 1.59 6 0.10% vs. tamoxifen 2.70 6 0.32%; P ,0.05) (Fig. 5I). We confirmed this finding by using distinctb-cell (PDX-1) and replication (BrdU incorporation) mar-kers (vehicle 1.00 6 0.26% vs. tamoxifen 2.33 6 0.26%;P , 0.05) (Fig. 5J). Representative images used to quantifyb-cell replication are shown in Fig. 5K.

Finally, we compared the in vitro basal and compound-induced replication index of b-cells obtained from vehicle-and tamoxifen-injected ibADKO mice (Fig. 5L). Islet ADKdeletion was confirmed by Western blot (Fig. 5M). ADK-deficient b-cells (tamoxifen-injected) displayed increasedbasal replication compared with ADK-expressing b-cells(Fig. 5L). In addition, ADK-expressing b-cells displayed anapproximately twofold replication increase after treatmentwith 5-IT, an ADK and DYRK1A/B inhibitor (Fig. 5L)(19,20). Similarly, 5-IT treatment of ADK-deficient b-cellsdemonstrated an ;1.5-fold increase in b-cell replication.Therefore, disruption of ADK expression in b-cells increasesthe basal b-cell replication index in vitro but does not elim-inate 5-IT–dependent replication induction. These resultsindicate that ADK negatively regulates b-cell replicationin vitro and that the b-cell replication-promoting activityof 5-IT was not mediated entirely through ADK inhibition.

DISCUSSION

Identification of novel strategies to enhance in vivo b-cellproliferation and mass while retaining optimum function isan attractive therapeutic strategy for diabetes. Previously, wediscovered that short-term treatment with ADKIs stimulatedrodent and porcine b-cell replication (19). In the currentstudy, we used a novel genetic mouse model to study thefunction of ADK in b-cells and to contemplate the potentialutility of ADK inhibition. This study addressed two criticalquestions: 1) is long-term disruption of ADK detrimental tob-cell function, and 2) are mice lacking ADK expression intheir b-cells protected against diabetogenic insults, such asaging and HFD? By using two different mouse models, wefound b-cell–selective disruption of ADK expression to bewell tolerated and protective against HFD challenge.

Is Long-term Disruption of ADK Detrimental to b-CellFunction?Currently available ADKIs are not amenable to long-termin vivo treatment because of associated toxicity (19,34). Inaddition, global disruption of ADK expression results inperinatal lethality (29). These findings raise concern aboutpotential detrimental effects of ADK disruption on b-cellviability and function. We found that ADK expression inthe islet was low relative to the exocrine pancreas and thatdisruption of ADK had a limited impact on b-cell develop-ment, function, and growth. NC-fed mice lacking b-cell ex-pression of ADK, either constitutively or acutely, displayedglucose homeostasis parameters similar to control mice,with no detectable difference in b-cell replication or mass.Hence, disruption of ADK in b-cells was well tolerated.

Are Mice Lacking ADK Expression in Their b-CellsProtected Against Diabetogenic Insults, Such as Agingand HFD?Although NC-fed ADK-deficient mice were essentiallyindistinguishable from control animals, a subtle but con-sistent improvement in IPGTT emerged as animals aged be-yond 1 year. To better gauge the potential benefit of ADKdisruption on glucose homeostasis, bADKO mice were chal-lenged with an HFD. Indeed, HFD-fed mice harboring ADK-deficient b-cells displayed significantly enhanced IPGTT,enhanced in vivo GSIS, and increased b-cell mass. No changein a-cell mass was observed, indicating that the impact wascell autonomous and/or lineage specific. In addition, wefound no evidence for islet neogenesis because islet density(insulin clusters per section) was unchanged in HFD-fedbADKO mice. Although bADKO mice maintained on HFDfor several months displayed increased b-cell mass, activeb-cell replication was not significantly increased in theseanimals. By contrast, b-cell replication was increased1 week after conditional disruption of ADK expression.These observations are consistent with prior work demon-strating that b-cell replication rates initially increase in re-sponse to HFD but decline with prolonged exposure (35).Of note, ADK-deficient b-cells demonstrated an increasedin vitro replication index compared with control b-cells. TheADKI 5-IT that promotes human b-cell replication, in part

diabetes.diabetesjournals.org Navarro and Associates 1935

Figure 5—Conditional deletion of ADK in the b-cells of HFD-fed mice enhances glucose tolerance and b-cell replication. A: The temporalrelationship of experiments performed on ibADKO mice (red text indicates male mice only). B: Representative b-galactosidase histochemistry inpancreatic sections obtained from vehicle- and tamoxifen-injected animals 1 week postinjection. C: IPGTT of NC-fed female mice treated withvehicle or tamoxifen (n = 8 mice/group; no significant differences observed). D: Fed glucose values obtained from 22-week NC- and HFD-fed(4 weeks) mice (n = 8 mice/group). E and F: IPGTT of 21-week (3 weeks of HFD) and 29-week (11 weeks of HFD) female mice (n = 8 mice/group),respectively.G: IPITT of HFD-fed female mice treated with vehicle or tamoxifen (n = 8 mice/group; no significant differences observed). H: In vivoGSIS measurements of HFD-fed female mice treated with vehicle or tamoxifen (n = 8 mice/group). I: b-Cell replication index of male NC- andHFD-fed ibADKO mice that received vehicle or tamoxifen treatment (n = 8 mice/group). J: b-Cell replication index of vehicle- and tamoxifen-treated HFD-fed male ibADKO mice (n = 8 mice/group). K: Representative images of pancreatic sections used for b-cell replication analysis.Images were obtained from vehicle- and tamoxifen-treated ibADKO mice stained for DAPI (blue), insulin (red), and ki67 (green) (top panels) orDAPI (blue), PDX-1 (red), and BrdU (green) (bottom panels). L: In vitro b-cell replication index of DMSO-treated and 5-IT–treated (2 mmol/L) isletcultures obtained from vehicle-injected (ADK-expressing) and tamoxifen-injected (ADK-deficient) mice (n = 3–4). M: Western blot of islet lysatesfrom vehicle- and tamoxifen-injected mice for ADK and b-actin. *P < 0.05.

1936 ADK Negatively Regulates b-Cell Expansion Diabetes Volume 66, July 2017

through DYRK1A/B inhibition (25), further enhanced thereplication of ADK-deficient b-cells, suggesting that inhibi-tion of both ADK and DYRK1A/B contribute to rodentb-cell replication control. In summary, disruption of ADKexpression in b-cells was protective against HFD-inducedglucose intolerance in mice in part as a result of an in-creased b-cell replication response, b-cell mass expansion,and enhanced insulin secretion.

Numerous mouse models of b-cell–specific gene disrup-tion yield altered b-cell mass and/or glucose homeostasis(36). However, the majority of these models demonstratean impact under basal conditions. Disruption of ADK ex-pression in b-cells is distinct in that essentially no pheno-type has been identified in unchallenged animals. Thesefindings suggest that ADK plays a role in dampening theadaptive b-cell response to metabolic challenge. Definingfactors that specifically modulate the adaptive b-cell re-sponse is an important avenue for uncovering therapeuticstrategies for T2D. Similar to ADK, connective tissuegrowth factor (CTGF) contributes to the adaptive responseof b-cells. Overexpression of CTGF in mature b-cells has noimpact on b-cell proliferation but enhances b-cell expan-sion under diabetogenic conditions (37). Similarly, b-cell–specific deletion of the prolactin receptor has no impact onb-cell mass or function under basal conditions but predis-poses female mice to gestational diabetes mellitus by damp-ening the b-cell replication response during pregnancy(38). Although the molecular links among CTGF, prolactinsignaling, and ADK are not immediately obvious, thesemodels highlight the potential to therapeutically manipu-late the adaptive response of b-cells. Perhaps more rele-vant, b-cell–specific deletion of the adenosine receptor 2a(Adora2a) in mice was recently shown to impair glucoseregulation and b-cell proliferation in pregnancy while hav-ing no impact on these parameters under basal conditions(39), indicating a role for adenosine signaling in promotingthe adaptive b-cell response to an insulin-resistant state. Aprimary function of ADK is to reduce extracellular adeno-sine levels through adenosine phosphorylation; hence, dis-ruption of ADK in b-cells might promote an adaptive b-cellresponse in vivo by increasing extracellular adenosine levelsand augmenting Adora2a-dependent signaling.

The current study has notable limitations. First, cautionmust be taken with use of the Rip-Cre and MIP-Cre/ERTtransgenic lines (32,33). Prior studies have shown detri-mental effects of the Rip-Cre cassette on glucose toleranceand insulin secretion, findings opposite to what we ob-served in the bADKO mice. Consequently, impacts of theRip-Cre line are anticipated to bias results away from theobserved protective effect of ADK deletion on HFD-inducedglucose intolerance. In addition, complementary findingsobtained with the constitutive and inducible Cre driverstrains excluded developmental impacts and ectopic genedeletion as the probable basis of the enhanced b-cell repli-cation, mass, and function we observed. Second, theapplicability of the findings to human b-cells is unknownbecause adult human b-cells have limited regenerative

capacity (14,40). Finally, the potential applications ofin vivo ADK inhibition may be limited by the function ofADK in other tissues, such as the liver. Indeed, identifyingmethods for lineage-restricted drug delivery is likely to be acritical hurdle for developing regenerative therapies for di-abetes. Future efforts will focus on better understandingthe mechanism by which ADK disruption enhances theadaptive response of b-cells to HFD challenge.

Acknowledgments. The authors thank Fredric Kraemer (Stanford UniversitySchool of Medicine) for generous support of this work. The authors also thank SaraSun (Stanford University) for technical contributions to this work.Funding. This research was supported by the Friedenrich BII Diabetes Fund, theSPARK Translational Research Program, and the Child Health Research Institute at StanfordUniversity (National Institutes of Health National Center for Advancing Translational SciencesClinical and Translational Science Award UL1-TR-001085 and National Institute of Diabetesand Digestive and Kidney Diseases grant R01-DK-101530).Duality of Interest. No potential conflicts of interest relevant to this articlewere reported.Author Contributions. G.N. conceived and designed experiments, performedexperiments, analyzed data, and approved the final version of the manuscript. Y.A.conceived and performed experiments, analyzed data, and assisted in writing themanuscript. Z.Z., H.H., and S.L. performed experiments, analyzed data, and approvedthe final version of the manuscript. N.A.A. analyzed data and assisted in writing of themanuscript. J.P.A. conceived and designed experiments, analyzed data, and wrote themanuscript. J.P.A. is the guarantor of this work and, as such, had full access to allthe data in the study and takes responsibility for the integrity of the data and theaccuracy of the data analysis.Prior Presentation. Parts of this study were presented at the 97th EndocrineSociety’s Annual Meeting and Expo, San Diego, CA, 5–7 March 2015.

References1. Weir GC, Bonner-Weir S. Five stages of evolving beta-cell dysfunction duringprogression to diabetes. Diabetes 2004;53(Suppl. 3):S16–S212. Alejandro EU, Gregg B, Blandino-Rosano M, Cras-Méneur C, Bernal-Mizrachi E.Natural history of b-cell adaptation and failure in type 2 diabetes. Mol Aspects Med2015;42:19–413. Butler AE, Janson J, Bonner-Weir S, Ritzel R, Rizza RA, Butler PC. Beta-celldeficit and increased beta-cell apoptosis in humans with type 2 diabetes. Diabetes2003;52:102–1104. Butler AE, Dhawan S, Hoang J, et al. b-cell deficit in obese type 2 diabetes, aminor role of b-cell dedifferentiation and degranulation. J Clin Endocrinol Metab2016;101:523–5325. Jonas JC, Sharma A, Hasenkamp W, et al. Chronic hyperglycemia triggers lossof pancreatic beta cell differentiation in an animal model of diabetes. J Biol Chem1999;274:14112–141216. Talchai C, Xuan S, Lin HV, Sussel L, Accili D. Pancreatic b cell dedifferentiationas a mechanism of diabetic b cell failure. Cell 2012;150:1223–12347. Polonsky KS, Given BD, Hirsch L, et al. Quantitative study of insulin secretionand clearance in normal and obese subjects. J Clin Invest 1988;81:435–4418. Kulkarni RN, Jhala US, Winnay JN, Krajewski S, Montminy M, Kahn CR. PDX-1haploinsufficiency limits the compensatory islet hyperplasia that occurs in responseto insulin resistance. J Clin Invest 2004;114:828–8369. Pimenta W, Korytkowski M, Mitrakou A, et al. Pancreatic beta-cell dysfunctionas the primary genetic lesion in NIDDM. Evidence from studies in normal glucose-tolerant individuals with a first-degree NIDDM relative. JAMA 1995;273:1855–186110. Gaulton KJ, Ferreira T, Lee Y, et al.; DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) Consortium. Genetic fine mapping and genomic annotationdefines causal mechanisms at type 2 diabetes susceptibility loci. Nat Genet 2015;47:1415–1425

diabetes.diabetesjournals.org Navarro and Associates 1937

11. Porat S, Weinberg-Corem N, Tornovsky-Babaey S, et al. Control of pancreaticb cell regeneration by glucose metabolism. Cell Metab 2011;13:440–44912. Helman A, Klochendler A, Azazmeh N, et al. p16(Ink4a)-induced senescence ofpancreatic beta cells enhances insulin secretion. Nat Med 2016;22:412–42013. Cox AR, Lam CJ, Rankin MM, et al. Extreme obesity induces massive beta cellexpansion in mice through self-renewal and does not alter the beta cell lineage.Diabetologia 2016;59:1231–124114. Saisho Y, Butler AE, Manesso E, Elashoff D, Rizza RA, Butler PC. b-cell massand turnover in humans: effects of obesity and aging. Diabetes Care 2013;36:111–11715. Tyrberg B, Eizirik DL, Hellerström C, Pipeleers DG, Andersson A. Humanpancreatic beta-cell deoxyribonucleic acid-synthesis in islet grafts decreases withincreasing organ donor age but increases in response to glucose stimulation in vitro.Endocrinology 1996;137:5694–569916. Nichols RJ, New C, Annes JP. Adult tissue sources for new b cells. Transl Res2014;163:418–43117. Dor Y, Brown J, Martinez OI, Melton DA. Adult pancreatic beta-cells are formedby self-duplication rather than stem-cell differentiation. Nature 2004;429:41–4618. Meier JJ, Butler AE, Saisho Y, et al. Beta-cell replication is the primary mech-anism subserving the postnatal expansion of beta-cell mass in humans. Diabetes2008;57:1584–159419. Annes JP, Ryu JH, Lam K, et al. Adenosine kinase inhibition selectively pro-motes rodent and porcine islet b-cell replication. Proc Natl Acad Sci U S A 2012;109:3915–392020. Zhao Z, Low YS, Armstrong NA, et al. Repurposing cAMP-modulating medi-cations to promote b-cell replication. Mol Endocrinol 2014;28:1682–169721. Park J, Gupta RS. Adenosine kinase and ribokinase—the RK family of proteins.Cell Mol Life Sci 2008;65:2875–289622. Andersson O, Adams BA, Yoo D, et al. Adenosine signaling promotes re-generation of pancreatic b cells in vivo. Cell Metab 2012;15:885–89423. Wang P, Alvarez-Perez JC, Felsenfeld DP, et al. A high-throughput chemicalscreen reveals that harmine-mediated inhibition of DYRK1A increases human pan-creatic beta cell replication. Nat Med 2015;21:383–38824. Litovchick L, Florens LA, Swanson SK, Washburn MP, DeCaprio JA. DYRK1Aprotein kinase promotes quiescence and senescence through DREAM complex as-sembly. Genes Dev 2011;25:801–81325. Dirice E, Walpita D, Vetere A, et al. Inhibition of DYRK1A stimulates humanb-cell proliferation. Diabetes 2016;65:1660–167126. Schnütgen F, De-Zolt S, Van Sloun P, et al. Genomewide production of mul-tipurpose alleles for the functional analysis of the mouse genome. Proc Natl Acad SciU S A 2005;102:7221–7226

27. Serup P, Petersen HV, Pedersen EE, et al. The homeodomain protein IPF-1/STF-1is expressed in a subset of islet cells and promotes rat insulin 1 gene expressiondependent on an intact E1 helix-loop-helix factor binding site. Biochem J 1995;310:997–100328. Cui XA, Singh B, Park J, Gupta RS. Subcellular localization of adenosine kinasein mammalian cells: the long isoform of AdK is localized in the nucleus. BiochemBiophys Res Commun 2009;388:46–5029. Boison D, Scheurer L, Zumsteg V, et al. Neonatal hepatic steatosis bydisruption of the adenosine kinase gene. Proc Natl Acad Sci U S A 2002;99:6985–699030. Postic C, Shiota M, Niswender KD, et al. Dual roles for glucokinase in glucosehomeostasis as determined by liver and pancreatic beta cell-specific gene knock-outs using Cre recombinase. J Biol Chem 1999;274:305–31531. Lee JY, Ristow M, Lin X, White MF, Magnuson MA, Hennighausen L. RIP-Crerevisited, evidence for impairments of pancreatic beta-cell function. J Biol Chem2006;281:2649–265332. Wicksteed B, Brissova M, Yan W, et al. Conditional gene targeting in mousepancreatic b-cells: analysis of ectopic Cre transgene expression in the brain. Diabetes2010;59:3090–309833. Oropeza D, Jouvet N, Budry L, et al. Phenotypic characterization of MIP-CreERT1Lphi mice with transgene-driven islet expression of human growth hormone.Diabetes 2015;64:3798–380734. Boison D. Adenosine kinase: exploitation for therapeutic gain. Pharmacol Rev2013;65:906–94335. Mosser RE, Maulis MF, Moullé VS, et al. High-fat diet-induced b-cell pro-liferation occurs prior to insulin resistance in C57Bl/6J male mice. Am J PhysiolEndocrinol Metab 2015;308:E573–E58236. Ackermann AM, Gannon M. Molecular regulation of pancreatic beta-cellmass development, maintenance, and expansion. J Mol Endocrinol 2007;38:193–20637. Riley KG, Pasek RC, Maulis MF, et al. Connective tissue growth factor modu-lates adult b-cell maturity and proliferation to promote b-cell regeneration in mice.Diabetes 2015;64:1284–129838. Banerjee RR, Cyphert HA, Walker EM, et al. Gestational diabetes mellitus frominactivation of prolactin receptor and MafB in islet b-cells. Diabetes 2016;65:2331–234139. Schulz N, Liu KC, Charbord J, et al. Critical role for adenosine receptor A2a inb-cell proliferation. Mol Metab 2016;5:1138–114640. Okada T, Liew CW, Hu J, et al. Insulin receptors in beta-cells are critical for isletcompensatory growth response to insulin resistance. Proc Natl Acad Sci U S A 2007;104:8977–8982

1938 ADK Negatively Regulates b-Cell Expansion Diabetes Volume 66, July 2017