Adenovirus-mediated Knockout of a Conditional Glucokinase Gene in Isolated Pancreatic Islets Reveals an Essential Role for Proximal Metabolic Coupling Events in Glucose-stimulated Insulin Secretion*

The relationship between glucokinase (GK) and glucose-stimulated metabolism, and the potential for metabolic coupling between β cells, was examined in isolated mouse islets by using a recombinant adenovirus that expresses Cre recombinase (AdenoCre) to inactivate a conditional GK gene allele (gk lox). Analysis of AdenoCre-treated islets indicated that the gk lox allele in ∼30% of islet cells was converted to a nonexpressing variant (gk del). This resulted in a heterogeneous population of β cells where GK was absent in some cells. Quantitative two-photon excitation imaging of NAD(P)H autofluorescence was then used to measure glucose-stimulated metabolic responses of individual islet β cells fromgk lox/lox mice. In AdenoCre-infected islets, approximately one-third of the β cells showed markedly lower NAD(P)H responses. These cells also exhibited glucose dose responses consistent with the loss of GK. Glucose dose responses of the low-responding cells were not sigmoidal and reached a maximum at ∼5 mmglucose. In contrast, the normal response cells showed a sigmoidal response with an K catS0.5 of ∼8 mm. These data provide direct evidence that GK is essential for glucose-stimulated metabolic responses in β cells within intact islets and that intercellular coupling within the islet plays little or no role in glucose-stimulated metabolic responses.

Glucose-stimulated insulin secretion by pancreatic ␤ cells is a multistep process that depends on increased metabolic flux (1). The rate-determining step in ␤ cell glucose metabolism is widely thought to be glucose phosphorylation, which is catalyzed to a large extent by glucokinase (GK) 1 at physiological glucose concentrations (2). Although many studies have implied an essential role for GK in glucose metabolism, most of this information has been correlative and has not precisely defined the role of this particular hexokinase isoform in ␤ cell glucose usage. GK gene knockout studies in transgenic mice, which offer the most direct and minimally ambiguous route to assessing the function of this enzyme in ␤ cells, have been performed by several different groups. Both global gene knockout studies, and more recent ␤ cell-specific gene knockouts indicate that GK is indispensable for glucose-stimulated insulin secretion (3)(4)(5)(6). While these studies clearly demonstrate an essential role for GK in glucose homeostasis, they have been unable to determine the precise role of GK, since studies in ␤ cells from adult animals are prevented by the perinatal mortality that occurs in GK-null mice.
A second, less well studied feature of glucose-stimulated insulin secretion from ␤ cells is its dependence on cell-cell interactions. The amount of insulin secreted from intact islets has long been known to be greater than that secreted from an equivalent number of dispersed ␤ cells (7). Models to explain this behavior generally include cooperative phenomena between islet cells, as suggested by the presence of synchronous electrical responses from clustered ␤ cells and intact islets (8,9). The pharmacological blocking of gap junctions reduces islet insulin secretion, thereby suggesting that conexins are involved in the intercellular cooperation (10 -12). Based on measurements of NAD(P)H in both intact islets, we previously proposed that metabolic uniformity arises from a uniform GK distribution and is not dependent on intercellular coupling (13). The concept of uniform cell-to-cell GK distributions is also supported by a recent immunofluorescence study of intracellular expression patterns in intact islets (14). Given that the role of cell-cell communication within intact islets during glucosestimulated metabolic flux has never been directly examined, and that glucose-stimulated metabolism in islets does not exhibit the NAD(P)H response heterogeneities observed in isolated ␤ cells (15), it remains possible that intercellular coupling may play a role in generating the metabolic uniformity.
To elucidate the roles of GK and intercellular communication in glucose metabolism within intact pancreatic islets, we utilized three recently developed technologies. First, the Cre/loxP strategy for inducible gene knockouts enables the elimination of specific genes in vitro in tissues from mature animals (16). To use this strategy, mice have been created with a conditional GK gene allele (gk lox ) in which exons 9 and 10 of the GK gene are flanked by loxP sites, thereby allowing for its inactivation by Cre recombinase (6). Second, specific inactivation of single genes in isolated islets has been enhanced by the development of a recombinant adenovirus expressing Cre (AdenoCre) (17). Adenoviruses have been shown previously to be an efficient means of introducing genes into ␤ cells in isolated islets (18). Third, we used two-photon excitation microscopy (TPEM) to study glucose-stimulated processes within intact islets (19,20). This quantitative optical sectioning technique has been demonstrated to be useful for assessing glucose-stimulated metabolic responses in intact islets and allows simultaneous determination of the glucose dose response in many individual ␤ cells (13).
The combination of these three recent methodological advances enabled us to directly examine the role of GK in glucosestimulated metabolism in single ␤ cells within intact pancreatic islets. By exposing cultured islets isolated from mice that are homozygous for the conditional gk lox allele to AdenoCre, we have been able to eliminate GK in a sizable fraction of ␤ cells and to examine interactions between cells with normal GK levels and those that are not expressing GK. These data provide direct evidence that GK is an important component of the ␤ cell glucose sensor. However, even cells that are presumed to fully lack GK retain an attenuated glucose response, thus suggesting that the glucose sensor is multicomponent and likely involves hexokinases other than GK. In addition, these data provide compelling evidence that individual ␤ cells function as independent glucose-sensing units at the metabolic level.

EXPERIMENTAL PROCEDURES
Animals-Mice that contain a conditional GK gene allele (gk lox ) have been described previously (6). All animals used in this study were homozygous at this allele (gk lox/lox ) in a mixed C57Bl/6 -129SvEvTac genetic background. This conditional allele is identical to the wild-type allele, but contains loxP sites flanking exons 9 and 10 of the gene. The neomycin resistance cassette used to perform the gene targeting is absent in the gk lox allele. PCR primer pairs for distinguishing the gk lox and gk del alleles have been described previously (6).
Preparation of AdenoCre-293 cells (21) were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. AdenoCre (obtained from F. Graham, McMaster University, Ontario, Canada) expresses Cre under the control of the cytomegalovirus immediate early promoter (17). This virus was grown and purified according to Becker et al. (22). High titer stocks of adenovirus (3-7 ϫ 10 12 plaque-forming units/ml, determined as 1 A 260 Ϸ 10 12 particles/ml) were prepared by equilibrium centrifugation in CsCl, stored in small aliquots at Ϫ80°C, and used immediately after thawing.
Islet Isolation and Culture-Islets were isolated from gk lox/lox mice by distention of the splenic portion of the pancreas followed by collagenase digestion (23,24). Islets were cultured at 37°C in RPMI 1640 medium containing 20 mM glucose, 10% fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin (Life Technologies, Inc.) with 5% CO 2 atmosphere. For each experiment islets were divided into two groups. The first group was incubated in 2 ml of culture medium containing recombinant adenovirus at a concentration of 1.5 ϫ 10 9 plaque-forming units/ml for 1 h at 37°C prior to culturing (18). Higher concentrations of adenovirus were also tried, but did not result in greater infection efficiency. The second group was not exposed to adenovirus. After infection islets were washed three times in culture medium. Both groups of islets were maintained in culture for 4 days before imaging. For the AdenoCre treated islets, culture in 20 mM glucose maintained their viability and morphological integrity better than those cultured at lower glucose levels.
Analysis of GK Protein Levels-Western blot analysis performed on preparations from isolated islets and GK activity was measured in isolated islets as described previously (6). Immunofluorescence staining of GK was also performed on isolated islets as described previously (13), except that an affinity-purified sheep anti-GK-IgG was used at a 1:10 dilution, and preimmune sheep serum was used as the control. The primary antibody was detected by donkey anti-sheep IgG-CY3 (Jackson ImmunoResearch, West Grove, PA) and imaged using a Zeiss LSM410 laser scanning confocal microscope (Vanderbilt Cell Imaging Resource).
Preparation of Islet DNA-Pooled islet preparations (ϳ500 islets each) were placed in a 0.5-ml Eppendorf tube with 50 l of 50 mM Tris (pH 8), 100 mM EDTA, 0.5% SDS, and 2.5 l of a 10 mg/ml solution of proteinase K, and this was incubated at 55°C for 2 h. Samples were extracted once each with equal volumes of phenol, phenol/chloroform, and chloroform, respectively. Islet genomic DNA was precipitated with 3 M sodium acetate and 100% ethanol and washed with 70% ethanol.
DNA Analysis-PCR reactions were performed on DNA from control and AdenoCre-treated islets. Two separate PCR reactions were used to detect the gk lox and gk del alleles respectively. Reaction mixtures contained 1 ϫ Perkin-Elmer PCR buffer, 0.2 mM each dNTP, 0.4 M amounts of each primer, and 0.5 l of Perkin-Elmer AmpliTaq Gold in 100 l total volume. Reaction products were visualized by agarose gel. DNA from control and AdenoCre-treated islets were further analyzed by Southern blot. 4 g of genomic DNA from control and AdenoCretreated islets was digested with BglII in a 25-l reaction for 4 h at 37°C. Digested samples were run in a 0.8% agarose gel, then blotted onto Zeta-Probe membrane overnight. A 0.5-b EcoRI/BamHI fragment of the gk gene located ϳ1.4 kilobases 3Ј of the targeted region was used as the hybridization probe to distinguish the gk lox and gk del alleles (6). The resulting autoradiogram was quantified by densitometry.
Measurement of NAD(P)H Autofluorescence-NAD(P)H imaging was performed by TPEM as described previously (13). For imaging, two islets (one from each group) were adjacently attached to a coverslip bottom dish (Mat-Tek Corp.). A 0.5-l drop of Cell-Tak (Collaborative Biomedical Products) was placed in the center of the dish and dried for 30 s at 42°C; the dish was rinsed with Hanks' balanced salt solution (Life Technologies, Inc.) and the islets placed on the Cell-Tak. During imaging, the islets were perifused at 1 ml/min with BMHH buffer (125 mM NaCl, 5.7 mM KCl, 2.5 mM CaCl 2 , 1.2 mM MgCl 2 , and 10 mM HEPES, and 0.1% bovine serum albumin (pH 7.4)). The sample was held at 37°C with a microincubator (TLC-MI, Adams & List Associates, Westbury, NY) that heated both the sample dish and incoming perifusate. An air stream incubator (Nicholson Precision Instruments, Gaithersburg, MD) heated the objective to eliminate heat transfer through the glass-oil-objective interface. All NAD(P)H autofluorescence measurements were made after a 15-min equilibration period on the microscope stage at 1 mM (basal) glucose. NAD(P)H glucose dose response images were acquired after a 5-min equilibration at each glucose concentration. Three consecutive 3-s scans were averaged to form the image for each concentration.
Image Analysis and Quantitation-To quantify the glucose response of NAD(P)H autofluorescence, digital image analysis was performed on Macintosh Power PC computers running NIH Image 1.61 (Bethesda, MD). Single cell data were taken in 25-pixel circular regions of interest that did not include the cell nucleus; the same regions of interest was used for all measurements (in images acquired with different glucose perifusion concentrations) on that cell. Only cells that remained in the same location within the image and maintained its same total area (to within 3%) were used in the analysis. This excluded cells that might have shifted into or out of the focal plane during the experimental procedures.

RESULTS
Both the ability and efficiency of AdenoCre to cause recombination in islets from mice that were homozygous for the gk lox allele was determined. PCR analysis of AdenoCre-treated islets revealed the conversion of the gk lox allele to the gk del allele only after virus treatment (Fig. 1A). To assess the efficiency of AdenoCre-mediated recombination, DNA was isolated from ϳ500 each of AdenoCre-treated and control islets and analyzed by Southern blot analysis. Densitometry of the resulting autoradiograms revealed that AdenoCre had converted ϳ30% (shown in Fig. 1B) of the gk lox alleles to the gk del allele (lane 2). Similar analysis of DNA from untreated gk lox/lox islets did not indicate any gk del allele (lane 1).
While the appearance of the gk del allele predicted a partial loss of GK expression in AdenoCre-treated islets, the actual reduction of GK levels was assessed by Western blot, GK activity measurements, and GK immunofluorescence. Western blot analysis yielded a 29.7 Ϯ 4.4% reduction in GK levels from AdenoCre-treated islets (n ϭ 3; shown in Fig. 1C), and activity measurements (n ϭ 3, data not shown) also showed reduced GK activity in AdenoCre-treated islets. Because adenovirus treatment disrupts islet integrity, GK immunofluorescence could not be accurately determined since none of the islets which were imaged for NAD(P)H autofluorescence responses survived the fixation and staining procedures (n ϭ 28 islets). However, four GK immunostained islets each showed reduced GK levels and exhibited heterogeneous GK immunofluorescence, similar to the NAD(P)H results presented below (data not shown).
To determine whether AdenoCre-mediated elimination of GK caused differences in glucose-induced NAD(P)H responses, control and treated gk lox/lox islets were examined side-by-side using TPEM. Representative NAD(P)H autofluorescence images of control and treated islet pairs are shown in Fig. 2. Control islets showed a very uniform autofluorescence signal during the 1 mM glucose perifusion, which was enhanced by perifusion with the higher glucose. In contrast, AdenoCretreated islets exhibited significant heterogeneity of NAD(P)H response. Under low glucose perifusion, the latter group of islets showed low but fairly uniform autofluorescence patterns, but in response to high glucose, many cells failed to show the expected rise in NAD(P)H.
To assess quantitatively the metabolic response of individual ␤ cells within whole islets, NAD(P)H response ratios (defined as (NAD(P)H signal at 25 mM glucose perifusion)/(NAD(P)H signal at 1 mM glucose perifusion) were determined for 220 nonperipheral ␤ cells from 5 AdenoCre-infected islets. A histogram of the resulting ratios (Fig. 3) shows two separate distributions of cellular response. Approximately two-thirds of the analyzed ␤ cells showed a response similar to that observed in wild-type mice. The other approximately one-third of the cells formed a population with a lower NAD(P)H response. The solid line in Fig. 3 indicates the response ratios of Ͼ 1000 ␤ cells in islets isolated from wild-type mice. The right-hand peak of the histogram shows that ␤ cells from gk lox/lox islets are ϳ5% less responsive than cells from wild-type mice. The one-third Ad-enoCre infection rate is consistent with the 35% of islet cells that were infected with another adenovirus, which expresses ␤-galactosidase.
Measurements of NAD(P)H levels over a range of different glucose concentrations were performed to further define the metabolic responses of single ␤ cells in AdenoCre-treated islets. Two resulting glucose dose response curves are shown in Fig. 4. The first curve (solid circles) was generated from 10 normal response ␤ cells in two islets (cells that fall near the right-hand peak of the histogram in Fig. 3). This curve is sigmoidal with an inflection point at ϳ8 mM glucose, consistent with K cat S 0.5 of GK, thus indicating that GK is the main determinant of glucose metabolism in these cells. The second curve (open circles) was generated from eight low response ␤ cells in the same two islets (these cells fall near the left-hand peak of the histogram in Fig.  3). This curve is not noticeably sigmoidal and reaches a maxi-

FIG. 2. NAD(P)H response to glucose of control and AdenoCretreated gk lox/lox islets. A, NAD(P)H autofluorescence images of both islets (control islet (left) and
AdenoCre-treated islet (right)) under 1 mM glucose perifusion. Both islets show a low, but fairly uniform, signal level. B, same two islets after 5 min of 25 mM glucose in the perifusion medium. In the control islet, the NAD(P)H signal is greatly elevated in all cells, but in the AdenoCre-treated islet, many cells show a weak response.

FIG. 3. Histogram of the NAD(P)H response ratio ((NAD(P)H signal at 1 mM glucose perifusion)/(NAD(P)H signal at 25 mM glucose perifusion)
). This histogram represents 220 nonperipheral ␤ cells in 5 AdenoCre-treated islets from three different preparations. The normal distribution curve of islet cells from wild-type mice is overlaid. The bar graph histogram contains two populations of cells, the righthand peak is consistent with what is observed in wild-type mice, although untreated gk lox/lox islets show a ϳ5% reduction in NAD(P)H response ratio from wild-type mice. About one-third of the ␤ cells analyzed fall in the left-hand peak that shows considerably less than normal NAD(P)H response, which is consistent with a knockout of GK activity by Cre infection. mal value at ϳ5 mM glucose. The kinetics of response in these cells is consistent with glucose usage being determined by one or more other hexokinase gene family members with lower K m values and not by GK. DISCUSSION Many previous studies have pointed to a rate-determining role for GK in glucose-stimulated responses of ␤ cells (2,(25)(26)(27)(28). While the evidence for GK as the ␤ cell glucose sensor is compelling, it has been difficult to functionally distinguish the role GK from other hexokinases in ␤ cells. Mannoheptulose, a competitive inhibitor of hexokinases, has been used to inhibit glucose phosphorylation in ␤ cells (e.g. Ref. 27), but it is not specific for GK. In fact, application of mannoheptulose to Ad-enoCre-treated islets caused a uniform decrease in autofluorescence from all ␤ cells (not shown). Thus, to distinguish the specific role of GK apart from other hexokinase isoforms, we have made use of the highly specific Cre/loxP system to inducibly eliminate GK within ␤ cells in intact islets. This approach allows GK gene expression to remain normal in the adult animal prior to islet isolation, and avoids the lethality caused by either a global or ␤ cell-specific deficiency in GK (3-6). By using a recombinant adenovirus to express Cre, conversion of the gk lox allele to gk del allele was delayed until after islets are isolated from the adult animal. While an infection efficiency of only ϳ30% was achieved using this approach, it offered the advantage in this instance of creating heterogeneities of glucose sensing within the islet. These heterogeneities allowed us to examine the role of cell-cell communication during glucosestimulated metabolic activity, in addition to determining the specific functional role of GK.
The results from these studies are important in at least two regards. First, they provide strong additional evidence that GK does indeed function as the glucose sensor in ␤ cells. TPEM analysis showed that approximately one-third of the cells in AdenoCre-treated islets lacked a normal metabolic response to glucose (Figs. 2 and 3). This percentage was closely correlated to the amount of Cre-mediated recombination determined by Southern blot. Because AdenoCre uses the potent immediate early cytomegalovirus promoter to drive Cre expression, infection of a cell probably results in efficient Cre expression and thus recombination of both copies of the gk lox allele into the gk del allele. Because of this, intermediate levels of GK gene expression within infected cells (i.e. recombination of only a single allele) are unlikely, and the amount of recombination probably reflects the percentage of cells without GK. Similar percentages were observed in Western blots for GK. Unfortunately, it was not possible to examine the AdenoCre-treated islets by immunofluorescence after NAD(P)H imaging because, unlike normal islets (13), they did not remain immobilized on the Cell-Tak during fixation (despite over 20 attempts). We have found this to be a limitation regardless of the recombinant adenovirus used. Consequently, we were unable to correlate NAD(P)H responses with either Cre infection or GK expression on a cell-by-cell basis in the islets. Nonetheless, a few Adeno-Cre-treated islets that did survive the immunostaining process in parallel preparations showed increased heterogeneity and an overall reduction in GK immunofluorescence compared with untreated gk lox/lox islets. Furthermore, cells that were presumed to lack GK showed a muted NAD(P)H response to glucose that saturated at a lower glucose concentration than the more highly responding cells, consistent with glucose phosphorylation by other hexokinases.
It is well established that other hexokinases are expressed within ␤ cells and that glucose responsiveness requires a high GK/hexokinase ratio (29). Indeed, transformed ␤ cells maintained in culture for extended times demonstrate a lower GK/ hexokinase ratio and exhibit diminished glucose-stimulated responses (30). Thus, the significant NAD(P)H response observed at the lower glucose levels in the AdenoCre-infected islet cells is consistent with the activity of hexokinases other than GK. The absence of GK may cause an up-regulation of these other hexokinases, all of which exhibit a more pronounced inhibition by glucose 6-phosphate than does GK (29). Because of this differential inhibition, however, it is difficult to assess precisely the contribution of other hexokinases to glucose sensing in a normal ␤ cell. To rigorously determine the role of each hexokinase isoform, mice with conditional knockouts of each gene would have to be created and examined.
Second, these studies demonstrate that individual ␤ cells within intact pancreatic islets show independent metabolic responses. Because gradients in NAD(P)H levels between adjacent ␤ cells persist even after 30 min, there does not appear to be any mechanism within the islet to generate metabolic uniformity among heterogeneous cells. Thus, we conclude that intercellular communication is not involved in glucose-stimulated metabolic flux and that each ␤ cell senses glucose independently at the metabolic level. Exactly how cell-cell communication, which clearly plays a significant role in augmenting insulin secretion, is involved in more distal signaling events (e.g. membrane depolarization, Ca 2ϩ influx, and insulin exocytosis) remains to be determined.
The detection of artificial heterogeneities among single ␤ cells within AdenoCre-treated islets helps to further validate the use of TPEM for high-resolution measurements of cellular metabolism in thick samples. Because the optical section in TPEM is ϳ1 m thick, information from single cells should be uncontaminated by background fluorescence that arises from other cells. In fact, TPEM has proved useful for measuring activity of single synapses in brain slices using exogenous fluorescent probes (31). Here we have shown that TPEM measurements of NAD(P)H accurately report the metabolic flux in individual cells. Since TPEM offers submicron resolution, it also opens the possibility to perform subcellular metabolic measurements, such as differentiating the NAD(P)H signals between the cytoplasm and mitochondria.
It may be important to note that islets from gk lox/lox mice are slightly less responsive than islets from wild-type mice (ϳ5% difference between the line and the right-hand peak of the histogram in Fig. 3). This was not unexpected because the FIG. 4. Glucose dose response of single ␤ cells within Adeno-Cre-treated islets. Normal cells (F, middle of right-hand peak of histogram; n ϭ 10) showed a sigmoidal shaped glucose concentrationdependent response that is consistent with the known GK properties. The impaired cells (E, left-hand peak of histogram; n ϭ 8) show most of their response below 5 mM, which is consistent with the loss of GK from Cre recombination. blood glucose levels in the gk lox/lox mice are slightly elevated (Ͻ10%) from those in wild-type mice (6). Both of these findings indicate that the gk lox allele may express slightly less GK than the wild-type gene GK gene. This slight difference in responsiveness is not likely to be a limitation of the present study, since the difference in NAD(P)H response is less than we have observed between different mouse strains. 2 The use of in vitro gene knockouts offers great potential for investigations of cellular signal transduction. Because the Cre/ loxP approach allows normal gene expression to continue while the animal develops, any deleterious effects of the knockout in the whole animal are avoided. Another advantage of targeted gene knockouts is that, unlike pharmacological approaches, they are not, at least in principle, prone to nonspecific effects in the cells. Although we were able to take advantage of the heterogeneity introduced by the adenovirus infection in this study, complete in vitro gene knockouts could often be advantageous. To obtain 100% efficient recombination of the loxP sites, methods other than adenovirus-mediated transfections could be used for the introduction of Cre. One such alternative is to use mice that contain inducible Cre transgenes (32).