Direct Administration of Insulin Into Skeletal Muscle Reveals That the Transport of Insulin Across the Capillary Endothelium Limits the Time Course of Insulin to Activate Glucose Disposal
- Jenny D. Chiu,
- Joyce M. Richey,
- L. Nicole Harrison,
- Edward Zuniga,
- Cathryn M. Kolka,
- Erlinda Kirkman,
- Martin Ellmerer and
- Richard N. Bergman
- Department of Physiology and Biophysics, University of Southern California, Keck School of Medicine, Los Angeles, California
- Address correspondence and reprint requests to Richard Bergman, PhD, Keck School of Medicine, 1333 San Pablo St., MMR 626, Los Angeles, CA 90033. E-mail:
OBJECTIVE—Intravenous insulin infusion rapidly increases plasma insulin, yet glucose disposal occurs at a much slower rate. This delay in insulin's action may be related to the protracted time for insulin to traverse the capillary endothelium. An increased delay may be associated with the development of insulin resistance. The purpose of the present study was to investigate whether bypassing the transendothelial insulin transport step and injecting insulin directly into the interstitial space would moderate the delay in glucose uptake observed with intravenous administration of the hormone.
RESEARCH DESIGN AND METHODS—Intramuscular injections of saline (n = 3) or insulin (n = 10) were administered directly into the vastus medialis of anesthetized dogs. Injections of 0.3, 0.5, 0.7, 1.0, and 3.0 units insulin were administered hourly during a basal insulin euglycemic glucose clamp (0.2mU · min−1 · kg−1).
RESULTS—Unlike the saline group, each incremental insulin injection caused interstitial (lymph) insulin to rise within 10 min, indicating rapid diffusion of the hormone within the interstitial matrix. Delay in insulin action was virtually eliminated, indicated by immediate dose-dependent increments in hindlimb glucose uptake. Additionally, bypassing insulin transport by direct injection into muscle revealed a fourfold greater sensitivity to insulin of in vivo muscle tissue than previously reported from intravenous insulin administration.
CONCLUSIONS—Our results indicate that the transport of insulin to skeletal muscle is a rate-limiting step for insulin to activate glucose disposal. Based on these results, we speculate that defects in insulin transport across the endothelial layer of skeletal muscle will contribute to insulin resistance.
Insulin resistance is an important risk factor for type 2 diabetes, as well as other chronic ailments, including cardiovascular disease, hypertension (1–3), and certain cancers (4). Understanding the cause(s) of insulin resistance is an important goal to improve public health. While much is known about insulin's action on its target cells, the primary mechanism(s) of insulin resistance remains controversial (5–7). The prevalent approach to understanding insulin action has been to define the intracellular steps in insulin signaling (8,9). Yet, before activating intracellular pathways in skeletal muscle, secreted insulin must first distribute through capillaries and cross the endothelial barrier, diffuse through the interstitial space, and bind to insulin receptors on the target tissue. Although all three steps are critical for insulin to act peripherally, transendothelial transport requires further characterization (10,11). Such characterization is critical to understanding insulin action, as changes in access of insulin to the tissues have been shown to be altered in states of insulin resistance (12,13). Access of insulin to skeletal muscle is reduced in diet-induced insulin-resistant states (14). To implicate changes in transendothelial transport as causing subsequent changes in insulin resistance will require better understanding of the transport process itself.
That insulin resistance may reside at the level of endothelial transport is supported by data suggesting a sluggish transport rate between plasma and interstitial fluid. Plasma insulin levels per se are not correlated on a moment-to-moment basis with glucose uptake; glucose utilization increases slowly even in the face of a rapid increment in plasma insulin. Unlike with plasma insulin concentration, there exists an intimate “hand-in-glove” relationship between interstitial insulin (as reflected in lymph) and insulin-mediated glucose uptake (15). Others have shown a similar difference between plasma and interstitial insulin using alternative methods to access interstitial fluid (16,17). Thus, previous studies suggested that insulin concentration in the interstitial rather than the plasma compartment dictates insulin action. However, there has not been a direct demonstration that the delay observed in the appearance of insulin in interstitial fluid with insulin infusion reflects the time for insulin to cross the capillary wall. Nor has it been shown whether the correlation between interstitial insulin and glucose uptake is retained when the interstitial space is faced with rapidly changing levels of insulin.
The mechanism of insulin transendothelial transport and the explanation for the observed delay in interstitial insulin remain unclear. Our laboratory and others have reported that the transcapillary transport of insulin is not saturable in vivo (15,18). Barrett et al. (19) have recently imaged the endothelial transport process and showed that fluorescently labeled insulin is rapidly taken up and concentrated in endothelial cells, which suggests that transit to the interstitium may be rate limiting. However, Clark and colleagues have suggested that insulin mediates multiple hemodynamic changes in the vasculature before acting on skeletal muscle, which could also account for the delay in insulin action (20,21).
The present study was undertaken to examine in a direct manner whether the transport of insulin across the capillary endothelial barrier can explain the time delay observed with intravenous insulin infusions and whether this transport limits the increase in insulin action. Here we have, for the first time, bypassed the transendothelial transport process. If transport is responsible for the delay in insulin action by providing insulin directly into the interstitial space of muscle we should rapidly increase interstitial insulin concentrations and thereby eliminate the delay for insulin-mediated glucose uptake. In addition, the sensitivity to insulin of skeletal muscle can be measured in the living animal by comparing interstitial insulin to glucose disposal. Here, we ask whether the sluggish transport of insulin or the time for diffusion within the interstitial space is rate limiting for insulin action. If transport limits the rate of insulin action, transendothelial transport of the hormone could be an important locus for insulin resistance.
RESEARCH DESIGN AND METHODS
Experiments were performed on male mongrel dogs under anesthesia (n = 13). Animals were housed in the University of Southern California Medical School vivarium under controlled kennel conditions (12 h light:12 h dark) and were fed standard chow (49% carbohydrate, 25% protein, and 9% fat; Alfred Mills, Chicago, IL) once per day. Dogs were used for experiments only if they were judged to be in good health, determined by visual observation, body weight, hematocrit, and body temperature. Protocols were approved by the University of Southern California institutional animal care and use committee.
Animals were fasted 15 h before the morning of the experiment at 0600 h. Dogs were preanesthetized with acepromazine maleate (0.22 mg/kg Prom-Ace; Aueco, Fort Dodge, IA) and atropine sulfate (0.11 ml/kg; Western Medical, Arcadia, CA). Anesthesia was induced with sodium pentobarbital (0.44 ml/kg; Western Medical) and maintained with isofluorane (Western Medical). Dogs were placed on heating pads to maintain body temperature. Indwelling catheters were implanted in the right jugular vein for a continuous saline drip (∼1 L for the first 60 min of surgery and a slow drip thereafter) and in the left carotid artery for sampling and blood pressure monitoring (model 90603A; Space Labs, Issaquah, WA). Intracatheters were inserted into the left cephalic vein for variable glucose infusion and the right cephalic vein for insulin and somatostatin infusion. Indwelling catheters were also placed into both the right and left femoral arteries and veins for sampling. Two perivascular ultrasonic flowprobes (2 mm diameter; Transonic, Ithaca, NY) were placed around both right and left femoral arteries proximal to the femoral catheter for measuring rates of blood flow. Left and right hindlimb lymphatic vessels were cannulated by placing polyethylene catheters (PE10) into the afferent lymphatic vessels of the deep inguinal lymph node. Lymph was collected by gently massaging the leg directly above the popliteal area; massage has been shown to instantaneously increase lymph drainage without affecting lymph or plasma oncotic pressures (22). Blood pressure, heart rate, CO2, and O2 saturation were monitored continuously. At the conclusion of these experiments, animals were killed with an overdose of sodium pentobarbital (65 mg/kg Eutha-6; Western Medical).
Immediately after starting surgical procedures, a basal insulin eugylcemic clamp was started (t = −180 min). Somatostatin was infused (1ug · min−1 · kg−1; Bachem) and basal insulin replaced systemically (0.2m Ul· min−1 · kg−1; Novo Nordisk, Bagsvaerd, Denmark) for the entire study (Fig. 1A). Plasma and lymph samples were taken every 10 min after each injection. Exogenous 20% glucose was infused at variable rates to clamp arterial glucose to the basal level (t = −60 to 0 min) throughout the entire experimental period (t = 0 to 300 min). Euglycemia was maintained based on online glucose measurements from the injected leg's femoral artery plasma. Samples were taken simultaneously from the right and left femoral arteries and veins. Left and right hindlimb lymph vessels were sampled from 2 min before to 2 min after each blood sample point.
At times 0, 60, 120, 180, and 240 min, porcine insulin (0.3, 0.5, 0.7, 1, and 3 units, represented as I1, I2, I3, I4, and I5, respectively (n = 10); 31.3 ± 1.3 kg) or saline (S1, S2, S3, S4 and S5, respectively (n = 3); 25.9 ± 1.0 kg) was injected into the vastus medialis of the quadriceps femoris with two 30-gauge needles containing 0.5 ml per syringe (Fig. 1B). The injection rate was ∼0.6 ml/min. For individual experiments, right or left hindlimb was chosen at random for injections while the contralateral limb served as a control.
Plasma and lymph samples were assayed for insulin and glucose. Arterial, venous, and lymph samples were collected in microtubes precoated with lithium heparin (Becton Dickinson, Franklin Lakes, NJ). Plasma tubes also contained 50 uL EDTA (Sigma Chemicals). Blood samples were centrifuged immediately, and the supernatant was stored at −20°C until further assay. Femoral artery plasma samples taken from the injected hindlimb were immediately assayed for glucose with a YSI 2700 autoanalyzer (Yellow Springs Instrument, Yellow Springs, OH). Lymph samples were immediately stored at −20°C after sampling. Insulin was measured in plasma and lymph with an enzyme-linked immunosorbent assay adapted for dog plasma.
Clamp stability was assessed by the coefficient of variation (CV) between the six femoral artery glucose samples taken each h after injected doses. The possibility existed that some of the injected insulin would exit the injected muscle, enter the systemic circulation, and affect glucose uptake by the contralateral limb. Therefore, to correct for systemic effects of insulin on the injected limb, we calculated net insulin or local glucose uptake (LGU) as injected leg minus noninjected leg values. Glucose uptake across the limb was calculated based on Fick's principle: LGU = (GA [mins] GV) × BF, where GA = arterial glucose, GV = venous glucose, and BF = femoral blood flow. Area under the curve (AUC) was calculated using the trapezoidal method. To calculate the percentage of total injected insulin per h in the femoral vein and lymph vessel, preliminary studies were conducted with 2 h between each injection (0.5, 0.7, and then 1.0 units insulin) to allow insulin levels to decline from its maximal peak (data not shown; n = 3). From these data, measured insulin concentration times blood flow was plotted against time. Calculated area under the curve (AUC) demonstrated that if insulin was injected at t = 0 min it nearly cleared in both sampled lymph and venous vessels by 120 min. AUC60–120min was determined to be approximately half of total AUC0–120 min for both lymph and femoral vein insulin. Thus, for the present protocol of hourly insulin injections, AUC of the first injection hour was subtracted from the AUC of the second injection hour to negate the additive effect of injecting each hour before the insulin was fully cleared:
AUCbasal represents the insulin from basal insulin infusion. These corrected hourly AUCs for venous and lymphatic insulin were individually divided by the amount of insulin administered during the respective hours to determine the percentage of insulin in each compartment per h after injection. The sum of lymph and venous insulin was subtracted from the administered dose of insulin to determine the percentage of insulin that remained within the muscle interstitium to either be taken up, used, or degraded.
All experimental data are expressed as means ± SE. Statistical analyses were performed with paired or unpaired Student's t tests or one-way ANOVAs with Tukey's pairwise comparisons as appropriate.
Glucose clamp stability.
Femoral artery glucose was successfully clamped at each animal's basal glucose concentration throughout each experiment (CV between femoral artery glucose samples per h was 2.4 ± 0.5 for saline injections [Table 1] and 2.9 ± 0.4 for insulin injections [Table 2]).
Blood pressure and blood flow.
Throughout the experiments, stable vital signs were maintained, with an average systolic blood pressure of 108.7 ± 0.4 mmHg, diastolic 47.6 ± 0.7 mmHg, and pulse 112.7 ± 0.5 bpm. Blood flow between injected and noninjected hindlimbs was not significantly different, and there were not significant changes from basal flow (160.0 ± 13.9 ml/min) with sequential saline or insulin injections.
Lymph glucose was 107% of arterial or venous concentrations (P < 0.01, lymph versus arterial and lymph versus venous) but did not differ comparing injected and noninjected legs or between the five injections S1–S5 per leg (Table 1). As previously reported, lymph insulin was only one-half of arterial and venous insulin concentrations (P < 0.01, lymph versus arterial and lymph versus venous) (23–25) but was not different between hindlimbs (P > 0.2, injected versus noninjected). With saline, glucose uptake comparing the two legs was also not different (P > 0.7) and did not change between saline injections (P > 0.1, between S1 and S5 in either leg). Saline injections also did not affect glucose or insulin concentrations in arterial or venous blood (Table 1).
Similar to those in the saline injection study, lymph glucose levels were 110% of plasma glucose levels and remained so in the control leg (P < 0.01, lymph versus arterial and lymph versus venous glucose [Table 2]). However, increasing insulin concentrations in the injected leg due to injection resulted in a tendency for lymph glucose to decline toward arterial glucose levels (P < 0.01, injected versus noninjected leg with I5), presumably reflecting insulin-mediated cellular glucose uptake out of the interstitial space. Femoral vein glucose decreased with I2–I5 (from 89.0 ± 2.8 to 81.4 ± 2.8 mg/dl; P < 0.01 venous versus arterial glucose with I2–I5), consistent with increasing glucose uptake across the muscle bed.
Basal arterial insulin was 12.08 ± 0.7 mU/l. I1–I4 caused arterial insulin to incrementally rise ∼20% similarly in both legs (Fig. 2A). I5 resulted in an additional increase of ∼29% in both legs (Fig. 2A).
We observed a significant difference in femoral vein insulin concentrations between the injected and noninjected leg. With I1, venous insulin in the injected limb rose 57% to 19.56 ± 0.80 mU/l, while the insulin in the control leg equaled arterial insulin, which increased to 14.20 ± 0.28 mU/l. Each subsequent intramuscular injection resulted in higher venous insulin in the injected leg, while venous insulin of the control leg equaled arterial insulin (Fig. 2B).
Interstitial insulin increased with each injection. Basal lymph insulin in both hindlimbs averaged 5.86 ± 0.45 mU/l (Fig. 2C). As expected, with I1 lymph insulin in the injected limb increased ∼39% to 9.95 ± 0.86 mU/l, while the control hindlimb lymph did not change (6.31 ± 0.13 mU/l). Subsequent injections resulted in further rises in interstitial insulin of the injected leg to achieve a maximum concentration of 82.81 ± 2.92 mU/l with I5 (P < 0.05, injected versus noninjected for I2–I5 [Fig. 2C]). Elevations in lymph insulin in the injected leg were apparent within 10 min.
With doses I2–I5, net femoral vein and net lymph insulin were significantly higher than net femoral artery insulin. With each intramuscular injection, 15.1 ± 1.9% of injected insulin appeared on the venous side; net amount of insulin exiting via the lymphatic vessel was negligible (0.2 ± 0.03%). Thus, 84.7 ± 1.9% of the injected insulin remained in the muscle interstitium to presumably bind to receptors and mediate glucose uptake and to undergo degradation.
Insulin-mediated glucose uptake.
As expected, glucose uptake across the injected leg was significantly greater than that of the noninjected leg with doses I2–I5 (Fig. 3B). At basal, glucose uptake was almost the same in both legs and not different from zero. With each successive injection, LGU in the injected leg increased to an average maximum of 26.0 ± 4.5 mg/min over the last h while reaching only 5.2 ± 3.4 mg/min in the noninjected leg (P < 0.05, injected versus noninjected [Fig. 3B]). Similar to insulin appearance in lymph, glucose uptake reached maximum rates within 10 min of each insulin injection.
Relationship between lymph insulin and glucose uptake.
While the kinetics of insulin in plasma and net LGU were markedly different from each other, the nearly superimposable time course between net lymph insulin and net LGU was striking (Fig. 4A–C). The correlation between net LGU and net lymph insulin (Fig. 4D–F) is much stronger than correlation of LGU with net femoral vein or artery insulin at every dose (r = 0.19, r = 0.77, and r = 0.92 for femoral artery, femoral vein, and lymph insulins, respectively).
Insulin sensitivity from lymph insulin.
We calculated skeletal muscle insulin sensitivity by plotting LGU as a function of lymph insulin averaged for each insulin dose. LGU occurs in the insulin-injected hindlimb with saturable kinetics; the Vmax for local hindlimb glucose uptake was ∼22 mg/min per leg (Fig. 5). The ED50, or insulin concentration at which half-maximal stimulation is achieved, was ∼20 mU/l.
Insulin is released directly from the pancreas into the blood stream where it must distribute among skeletal muscle capillaries, pass the endothelial barrier, diffuse through the interstitial space, and bind to receptors to mobilize GLUT4 transporters to the cell surface. All of these processes must occur before insulin is able to increase glucose uptake into the muscle tissue. The latter step of insulin binding and signaling is well characterized (26). The transport of insulin to the target site and diffusion of insulin through the interstitium has received much less focus even though it may play an important role in determining insulin sensitivity. Previous conclusions regarding the time course of insulin diffusion through the endothelial barrier resulted from studies in which the time course of insulin in plasma was compared with the time course of glucose disposal (27). However, the delayed activation of glucose disposal may be due to slow uptake of insulin by endothelial cells (or movement across those cells), slow release into intersititum, and/or retarded diffusion within the interstitial compartment. Insulin resistance may occur in one or more of these steps. To determine the importance of transport versus diffusion in the interstitium, we administered insulin directly into the interstitial compartment. The present results prove that it is the transport process, and not the diffusion within the interstitial space, that accounts for the slow insulin effect on glucose disposal.
In vitro skeletal muscle insulin receptor binding and insulin receptor kinase activation have been shown to occur within seconds, and GLUT4 translocation in skeletal muscle has been shown to reach maximum rates within 15 min (28–30). Similarly, in vivo studies that examine the kinetics of signaling and transporter mobilization show that the two processes occur within 10 min of insulin binding (31,32). However, even before receptor binding and signaling, insulin must diffuse through the gel-like matrix of skeletal muscle interstitium to distribute itself among its receptors (33). To our knowledge, this is the first in vivo study to examine the kinetics of insulin arriving at its receptor sites once it has entered the interstitial matrix. Here, we demonstrate that providing insulin directly into the interstitium of skeletal muscle results in simultaneously increasing interstitial (lymph) insulin and glucose uptake within 10 min. These data suggest that the time for insulin diffusion within the interstitial space, insulin receptor binding, signaling, and GLUT4 translocation are not temporally limiting for insulin to initiate glucose uptake. Although the process of injecting a small volume of fluid into the space may itself contribute to faster distribution of insulin through the interstitium, saline injections given at the same volume and rate did not change either glucose uptake or insulin concentrations. This suggests that the effects of fluid volume changes on redistributing the substances within the interstitial space were minimal.
It is interesting to compare the time courses of insulin's microvascular actions with insulin's traversal of the capillary endothelium. Our current study demonstrates that once insulin arrives in the interstitial space, all remaining steps for insulin to mediate glucose uptake are not temporally limiting. In support of this, Miles et al. (32) estimated similar half times for insulin transport from the vasculature to the interstitium (∼28 min), insulin receptor kinase activation (∼32 min), and peripheral glucose disposal (∼20 min) in dogs. Thus, we can narrow the possible mechanisms for sluggish insulin action to the steps before interstitial delivery. However, insulin recruitment of capillaries occurs very early, within 5 to 10 min of insulin infusion (34). This suggests that it is most likely the time delay associated with insulin transport across capillary endothelium that renders insulin action sluggish.
Nevertheless, the mechanism of transendothelial transport remains unclear; it occurs through either a receptor-mediated process or simple diffusion, and it is either paracellularly or transcellularly transported (35). Using bovine aortic endothelial cells in vitro, King et al. (10) reported the transport to be a receptor-mediated, unidirectional process. However, in vivo transport of insulin is not saturable even at supraphysiological insulin infusions (18), and studies in which the insulin receptor was knocked out in vascular endothelial cells showed no change in glucose metabolism (36), suggesting that insulin receptors are not required for transport. Others have also reported a nonsaturable transport process, both in vivo and in vitro, which argues for a non–receptor-mediated process (37). Little is known about the transport of insulin from the interstitium into circulation observed in this study. We demonstrated an increase in venous insulin with interstitial injections, indicating an exit of insulin from the interstitium into the venous circulation. Barrett and colleagues employed confocal immunohistochemical methods to elucidate a time course for insulin's traversal of the endothelial cell. Their results suggested that insulin rapidly enters the vascular endothelial cell of skeletal muscle (within 10 min) and is then concentrated in the cell (60 min) to be slowly released from the abluminal side (19). Their results demonstrated a slow traversal of insulin across the vascular wall, which supports the delayed transendothelial transport hypothesis. When we allowed insulin to access the interstitium without crossing the vascular wall (i.e., by direct administration into muscle tissue), the delay in insulin action was attenuated, suggesting that the time involved in transendothelial transport could account for delays in peripheral glucose disposal.
Although transendothelial transport seems the likely culprit for delaying insulin action, we cannot discount hemodynamic or vascular changes as limiting the provision of insulin to its target site. Baron et al. (38–40) suggested that insulin itself can increase blood flow; neither we nor others have been able to support this result. Interestingly, Ellmerer et al. observed insulin's ability to augment the volume of distribution in the interstitial space of sensitive tissues whether by increasing the number of dilated capillaries in the area or by another unidentified mechanism (41). However, the time courses of these changes remain to be determined. The time associated with insulin-mediated redirection of flow to skeletal muscle (21) or changes in capillary permeability may help account for sluggish glucose disposal. Thus, we cannot discount the contribution of insulin's hemodynamic changes in limiting insulin-mediated glucose uptake. Additional studies are required to differentiate between these potential mechanisms.
By bypassing the transendothelial transport process, we found much higher calculated skeletal muscle sensitivity to insulin than previously reported from clamp studies. Others have reported an ED50 of 70 mU/l for skeletal muscle in vivo based on plasma insulin and LGU measurements (42). However, whole-body glucose uptake measurements with interstitial samples for insulin showed that the ED50 estimated from plasma insulin was two times greater than that measured by lymph insulin (70 vs. 40 mU/l) (43). Our present study observed an ED50 for skeletal muscle insulin sensitivity of only 20 mU/l, four times more sensitive than observed using glucose clamp and whole-body techniques. The realization that sensitivity to directly intramuscularly administered insulin is greater than plasma insulin suggests that there is a dose-dependent step in the endothelial transport process to limit access of insulin to the interstitium, thus increasing the apparent ED50. Further studies will be necessary to examine what aspect of transport limits insulin action and whether this limitation is increased in insulin-resistant states.
Studies have yet to substantiate any defect in insulin's ability to cross the capillary endothelium under insulin-resistant states such as obesity (44,45) and type 2 diabetes (46) and/or other insulin-resistant models (41,47,48). However, these studies have been conducted under hyperinsulinemic-euglycemic clamp conditions wherein steady state was achieved and possibly displayed the effects of a saturated system. We believe our local in vivo method, where comparatively little insulin is administered, more accurately assesses whether the transport of insulin across the vascular wall is altered with insulin resistance. Studies have shown a decrease in insulin-mediated capillary recruitment with various insulin-resistant models (49,50) and a decrease in the distribution volume (41). Thus, a decreased distribution of insulin to sensitive tissues may contribute to insulin resistance with or without affecting the ability of insulin to cross the endothelium. Further experiments will examine this theory using intramuscular injections in insulin-resistant models. The maintenance of anesthesia may induce some sort of sympathetic effect to change insulin dynamics, so conscious studies would be beneficial, although difficult to perform. We cannot discount that the infusion of somatostatin or meal feeding may alter our results, although we anticipate that these would not have a great affect on the local dynamics of insulin. It would be interesting to compare these results in additional species to ensure that this is not a species-specific effect.
In conclusion, our data demonstrate that steps before insulin's arrival at skeletal muscle interstitium are rate limiting for insulin action. We have directly shown that the processes following this arrival, including insulin diffusion through the interstitium, receptor binding, insulin signaling, and GLUT4 translocation, do not temporally limit insulin's ability to increase glucose disposal. Therefore, insulin transport appears to be a rate-limiting factor for insulin action and, thus, comparing the different processes of such transport and how they may change in different physiologic or pathophysiologic conditions requires further investigation.
This study was supported by grants from the National Institutes of Health (NIH) (DK 029867 and DK27619). C.M.K was supported by a Mentor-Based Postdoctoral Fellowship from the American Diabetes Association, and L. N. H. was supported by a training grant from the NIH.
We thank Elza Demirchyan and Rita Thomas for technical assistance.
Published ahead of print at http://diabetes.diabetesjournals.org on 25 January 2008. DOI: 10.2337/db07-1444.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Received October 9, 2007.
- Accepted January 16, 2008.