Dual Tracer Test to Measure Tissue-Specific Insulin Action in Individual Mice Identifies In Vivo Insulin Resistance Without Fasting Hyperinsulinemia

Insulin resistance (IR), a defect in the ability of insulin to promote glucose uptake, is the most pernicious risk factor for chronic metabolic disease (1). Despite the underlying importance of IR, relatively few insulin-sensitizing agents are available to patients and clinicians (2). Genetic and environmental factors both interact to produce IR, and delineating the contributions of these two factors has been challenging, especially since certain genetic backgrounds are more susceptible to IR in specific tissues and this can manifest in divergent whole-body outcomes (3–5). Studies in mice offer increased environmental control compared with human studies and have identified genetic regulation of candidate IR drivers (6,7). Despite the potential benefits of understanding the mechanisms and consequences of IR in different tissues, a recent review conceded that analyzing all tissues relevant to the pathogenesis of IR is not feasible using existing techniques (8).

The gold standard technique for assessing insulin action in humans and animals is the hyperinsulinemic-euglycemic clamp (9), referred to here as the clamp. In mice, tight control of glucose and insulin levels during the clamp has enabled the physiological consequences of dietary and genetic perturbations to be assessed with exceptional rigor, and the translational capacity of the method to humans has galvanized significant progress in the assessment of therapies for type 2 diabetes (10–12). Studies in mice combining the clamp with glucose tracers, most commonly radiolabeled 2-deoxyglucose (2DG), provide insight into in vivo insulin action in individual tissues.

One limitation of current approaches that assess tissue insulin action is that a separate group of animals is required to measure basal glucose uptake, and, for this reason, many studies tend to overlook this important measure. However, this significantly impacts interpretation of the data; insulin action cannot be inferred solely from insulin-stimulated glucose uptake since this is also influenced by differences in basal glucose uptake. This provides a significant motivation for an approach that measures both basal and insulin-stimulated uptake in the same animal. Such a technique would also enable insulin action to be studied in outbred animals like Diversity Outbred mice. This could be advantageous as Diversity Outbred mice are increasingly being used as a powerful tool to identify genetic drivers of complex phenotypes in metabolic research (13–16). These populations demand methods capable of measuring both basal and insulin-stimulated glucose uptake within the same animal since each mouse in these populations is genetically unique.

To address these limitations, we developed an in vivo method to measure both basal and insulin-stimulated glucose transport into all tissues of the same animal, referred to as the Dual Tracer Test. This approach uses two different 2DG tracers to obtain paired basal and insulin-stimulated measures of in vivo glucose uptake across all tissues simultaneously. The approach was validated in a dietary model of IR and produced results in high-fat fed C57BL/6J mice consistent with established literature using other methods such as the clamp or single tracer approaches. We used the Dual Tracer Test to characterize a new model of IR in which mice are partially deficient in insulin-responsive GLUT4. GLUT4-deficient mice demonstrated IR in all tissues without changes in body composition, enabling the consequences of reduced glucose disposal to be studied in isolation of other defects. The Dual Tracer Test halves the number of mice required for studies using inbred mice, and power simulations revealed a significant reduction in the number of mice required to dissect the genetic regulation of insulin action in outbred populations. The Dual Tracer Test will not replace the gold standard clamp method but instead offers a technically simple and robust method to assess in vivo tissue insulin action that can be performed by any laboratory with access to mice. The most challenging aspect is the retro-orbital injection, although this is easily learned. The Dual Tracer Test could also be adapted to the clamp, with the administration of one 2DG tracer 40 min prior to the commencement of the clamp and administration of the second tracer once an insulin-stimulated steady state has been achieved.

Male C57BL/6J mice (10 weeks old) were obtained from the Animal Resources Centre (Perth, Western Australia, Australia). Mice were fed a standard laboratory chow diet containing 13% calories from fat, 65% calories from carbohydrate, and 22% calories from protein (Irradiated Rat and Mouse Diet, Specialty Feeds, Glen Forest, Western Australia, Australia), or a high-fat high-sugar diet (HFD) made in-house containing 45% calories from fat, 35% calories from carbohydrate, and 20% calories from protein and manufactured to closely resemble D12451 (Research Diets, New Brunswick, NJ). Specifically, the HFD diet contained 23% w/w casein, 0.3% w/w methionine, 2% w/w gelatin, 20.2% w/w sucrose, 17% w/w corn starch, 5% w/w bran, 3% w/w safflower oil, 22% w/w lard, 5.8% w/w AIN-93 mineral mix (MP Biomedicals), 0.4% w/w choline bitartrate, and 1.3% w/w AIN-93 vitamin mix (MP Biomedicals).

Experiments were performed in accordance with National Health and Medical Research Council (Australia) guidelines under the approval of The University of Sydney Animal Ethics Committee. Mice were monitored twice weekly and weighed weekly. Lean mass and adiposity were determined using an EchoMRI-900 (EchoMRI Corp. Ptd. Ltd.). Mice were maintained at 23°C on a 12-h light/dark cycle and given ad libitum access to food and water in individually ventilated cages with a density of five mice per cage. Mice were acclimatized to housing conditions and handling for 1 week prior to experimentation.

Mice fasted for 6 h (0800–1400) were dosed via oral gavage with 20% glucose in water at 2 g/kg lean mass. Blood glucose concentrations were measured directly from tail whole blood via a glucose monitor at 0, 15, 30, 45, 60, and 90 min postglucose administration (Accu-Chek, Roche Diabetes Care). Blood insulin concentrations were measured by collecting 5 μL of blood from tail veins at 0 and 15 min directly into an Insulin Mouse Ultra-Sensitive ELISA (Crystal Chem, Elk Grove Village, IL) containing sample diluent. The assay was then performed according to the manufacturer’s protocol, using an insulin standard curve to determine sample insulin concentrations.

Mice were fasted for 2 h (1000–1200) and then anesthetized with 65 mg/kg total body mass sodium pentobarbital by intraperitoneal injection. Anesthesia was confirmed by loss of toe-pinch reflex, and mice remained unconscious for the full duration of the Dual Tracer Test. Euthermia was maintained throughout the duration of the experiment using a combination of insulating foil and heating pads (17), and mice remained warm to the touch throughout.

Mice were injected retro-orbitally (18) with 2.5 μCi [¹⁴C]2DG in plasma replacement (B. Braun, Melsungen, Germany), and blood glucose was measured 2, 15, and 30 min postinjection using a glucose monitor. Whole blood (5 μL) was collected via tail vein bleed into 95 μL 0.9% NaCl on ice at each time point. A second retro-orbital injection containing 2.5 μCi [³H]2DG and 0.75 units/kg lean mass insulin in plasma replacement was conducted at 40 min. Glucose measurements and blood samples were taken 2, 15, and 30 min postinjection. A final blood sample was taken 40 min postinjection, such that both basal and insulin-stimulated uptake phases were each 40 min in duration. Mice were euthanized by cervical dislocation at 80 min. Tissues were rapidly excised, snap-frozen in liquid nitrogen, and stored at −80°C until further processing.

The area above the insulin-stimulated blood glucose curve (AOC) was calculated as:

Where G_i and t_i are the blood glucose concentration and time at the ith blood sample, respectively, and G₀ is the blood glucose concentration at the time of insulin administration. Differences in the AOC reflect differences in insulin response rather than altered fasting glucose levels (19).

The systemic kinetics of tracer uptake were determined by measuring the disappearance of the tracer from the circulation over time. Blood samples (whole blood) were centrifuged at room temperature for 5 min at 20,000g. The supernatant (70 μL) was mixed with 3 mL liquid scintillation cocktail (PerkinElmer) and measured using a Tri-Carb 2810R Liquid Scintillation Counter (PerkinElmer). Tissue-specific tracer uptake was quantified by the amount of phosphorylated 2DG (2DG-6P) per mg of tissue as this represents the tracer taken into the cell. Tissues were hand pulverized in liquid nitrogen using a prechilled mortar and pestle. Approximately 40 mg of crushed tissue was lysed in 2 mL 1% Triton X-100 (Sigma-Aldrich). Samples were vortexed and incubated for 1 h on ice before centrifugation at 10,000g for 10 min at 4°C to pellet tissue debris. Two portions of supernatant were used to determine the amount of total and unphosphorylated 2DG present within the tissue. For the total fraction, 500 μL supernatant was combined with 500 μL H₂O and 3 mL liquid scintillation cocktail and briefly vortexed. For the unphosphorylated fraction, 500 μL supernatant was loaded onto 300 μL 37.5% w/v AG 1-X8 anion exchange resin (Bio-Rad Laboratories). Columns were washed three times with 1 mL H₂O, and the flow-through was collected and vortexed. Flow-through solution (1 mL) was combined with 3 mL liquid scintillation cocktail to produce the unphosphorylated fraction. The amount of 2DG-6P was calculated by subtracting unphosphorylated from total 2DG, following adjustments for the volume of each solution and the mass of tissue originally lysed.

The following counting windows were used to separate emissions from the different radiolabeled tags: 0–12 kiloelectron volts (keV) [³H] and 12–156 keV [¹⁴C]. The effectiveness of this strategy was assessed by counting ∼5,000 counts per million (CPMs) of each radioactive nuclei. Negligible window spillover was detected (Table 1).

Tissue-specific tracer uptake rates were calculated using the method of Hom et al. (20) to adjust 2DG-6P accumulation for initial dose and changes in systemic availability of the tracer as follows:

Where k_i is the rate of uptake in tissue i, C_i is the amount of 2DG-6P in the tissue at the end of the experiment, k_c is the rate of tracer disappearance from circulation, d₀ is the initial concentration of tracer in circulation extrapolated from a single exponential curve, and t is the duration that tissues are exposed to the tracer (t = 80 and 40 for basal and insulin-stimulated tracers, respectively).

For simulations of studies using inbred strains to compare a single control group with a single experimental group, groups were randomly sampled with variances matching that of insulin-stimulated 2DG uptake alone or the fold increase in 2DG uptake with insulin. The phenotypic difference between experimental and control groups was equal to the diet-induced defect in insulin-stimulated uptake for that tissue shown in Fig. 2H. The number of mice required to obtain 90% power at a significance level of 0.05 was identified. The difference in required sample size between simulations using the variances from data collected using insulin in isolation or insulin/basal fold changes is reported.

For simulations in outbred populations, genotypes and phenotypes were randomly sampled from a population with a minor allele frequency of 5% and locus effect size explaining ∼10% of phenotypic variation by one-way ANOVA. Genetic and residual variation both contribute to total within-genotype variation. Within-genotype variation was fixed for all simulations, and residual variation matched to the variance of data obtained using single or dual tracer approaches to measuring insulin response in soleus muscle. The number of mice included in the study was varied, and each combination was simulated 1,000 times. For each simulation, a quantitative trait locus was detected if the simulated effect could be successfully detected using a linear model to relate genotype to phenotype. Power is reported as the proportion of times an effect was correctly identified. Results are consistent with previous simulations performed for Diversity Outbred populations (14).

pHluorin Glut4 was made by deletion of the tandem-dimer Tomato using site-directed mutagenesis on the pHluorin Glut4 tandem-dimer Tomato plasmid (21). Retrovirus was made by transfecting Plat-E cells using Lipofectamine. 3T3-L1 fibroblasts were retrovirally infected with empty vector or pHluorin Glut4 by addition of virus and polybrene (8 μg/μL), and then positive cells were selected using puromycin. 3T3-L1 cells were maintained in DMEM (Thermo Fisher Scientific) supplemented with 10% FCS (Thermo Fisher Scientific), 1% GlutaMAX (Thermo Fisher Scientific), and puromycin in a humidified atmosphere with 10% CO₂. Confluent 3T3-L1 cells were differentiated into adipocytes by the addition of DMEM containing 0.22 μmol/L dexamethasone, 100 ng/mL biotin, 2 µg/mL insulin, and 500 μmol/L 3-isobutyl-1-methylxanthine. Medium was replaced with DMEM/10% FCS/GlutaMAX containing 2 µg/mL insulin after 72 h. After a further 72 h, cells were cultured in DMEM/10% FCS/GlutaMAX. Medium was subsequently replaced every 48 h. On day 10, cells were serum-starved for 2 h in DMEM/0.2% BSA/1% GlutaMAX at 37°C. After 2 h they were washed three times in PBS and changed into 0.6 mmol/L NaH₂PO₄, 0.4 mmol/L NaH₂PO₄, 120 mmol/L NaCl, 6 mmol/L KCl, 1 mmol/L CaCl₂, 1.2 mmol/L MgSO₄, and 12.5 mmol/L HEPES (pH 7.4) at 37°C. Cells were stimulated with insulin for 20 min. During the last 5 min, 0.25 μCi [³H]2-deoxyglucose ([³H]2DG) (PerkinElmer) and 50 μmol/L 2DG were added to the cells. Cells were moved to ice and washed three times in ice-cold PBS. Cells were solubilized in 1% Triton X-100 in PBS, and tracer uptake was quantified by liquid scintillation counting. Data were normalized for protein content, which was determined using a bicinchoninic acid (BCA) assay.

Tissue was homogenized by addition of TRIzol (Thermo Fisher Scientific) and using bead homogenization on a Mixer Mill 400 (Retsch). Then, 0.1× volume of 1-bromo-3-chloropropane was added to the tubes, and the samples were centrifuged at 13,000g for 15 min for phase separation. The clear phase was transferred to a fresh tube, and an equal volume of isopropanol was added. Samples were centrifuged at 13,000g for 10 min to precipitate RNA. RNA was washed three times in 70% ethanol with centrifugation. RNA was resuspended in diethyl pyrocarbonate treated water and quantified on a NanoDrop 2000 (Thermo Fisher Scientific). RNA was reverse transcribed to cDNA using PrimeScript Reverse Transcriptase (Takara) according to the manufacturer’s instructions. Primers specific to Glut4 (5′ gacggacactccatctgttg 3′ and 5′ gccacgatggagacatagc3′) and normalized to cyclophilin B (5′ ttcttcataaccacagtcaagacc 3′ and 5′ accttccgtaccacatccat 3′) were used for quantitative (q)PCR. qPCR was performed using FastStart SYBR Green MasterMix (2×) (Bio-Rad Laboratories) according to the manufacturer’s instructions.

Tissues were snap frozen in liquid nitrogen and stored at −80°C until further processing. Tissues were lysed in lysis buffer containing 20 mmol/L HEPES (pH 7.4), 1 mmol/L EDTA, and 250 mmol/L sucrose containing 2% SDS with sonication and centrifugation at 13,000g for 10 min. Protein concentration for each sample was quantified using a BCA assay (Sigma-Aldrich), and samples were made up to 1 mg/mL in Laemmli buffer containing tris(2-carboxyethyl)phosphine) (TCEP; Thermo Fisher Scientific). Samples were heated at 37°C for 30 min, then 20 μg was loaded into an 8% SDS-PAGE gel and transferred onto a polyvinylidene fluoride membrane (Merck Millipore). Membranes were blocked in 5% skim milk powder in TBS-T (0.1% v/v Tween 20 in Tris-buffered saline) for 1 h, followed by an overnight incubation at 4°C with antibodies against GLUT4 (22), 14-3-3 (Santa Cruz Biotechnology), or α-tubulin (Merck) antibody. GLUT4 signal was developed using rabbit horseradish peroxidase (HRP) secondary antibody (Sigma-Aldrich) and Immobilon Western Chemiluminescent HRP Substrate (Merck Millipore) and imaged on a ChemiDOC (Bio-Rad Laboratories). Signals for 14-3-3 and α-tubulin were developed using LI-COR fluorescent secondary antibodies and imaged on a LI-COR Odyssey (LI-COR).

The GLUT4-pHluorin knock-in mouse was created by Ozgene Pty. Ltd. Targeting vectors (created by PCR from C57BL/6J genomic DNA) were designed to conditionally knock-in pHluorin (21,23) into exon 3 of the Slc2a4 gene in mouse embryonic stem (ES) cells with the aim of inserting pHluorin between residues G65 and P66 of GLUT4. To avoid dominant negative effects of the knock-in, floxed wild-type (WT) cDNA containing exon 2 to 11 of the Slc2a4 coding sequence was inserted in the intron between exon 1 and 2. As such, Cre-mediated deletion of this region would result in knock-in of the pHluorin into the first exofacial loop of GLUT4. The vector was introduced into ES cells by electroporation, and homologous recombination was confirmed by screening for neomycin-resistant ES cells. Positive ES cells were expanded and injected into blastocysts, which were then implanted in pseudopregnant C57BL/6J females. Breeding of chimeras resulted in germline transmission. Transgenic mice were crossed with mice on the same background ubiquitously expressing Cre. Filial 1 offspring of this cross were heterozygous for Cre and pHluorin-tagged GLUT4. Once mice that ubiquitously expressed pHluorin-tagged GLUT4 were established, a breeding strategy was conducted to remove the Cre gene. Heterozygous breeding pairs were used to generate homozygous transgenic mice and WT littermates for this study.

Mice fasted for 6 h (0800–1400) were dosed via intraperitoneal injection of insulin at 1 unit/kg lean mass. Blood glucose concentrations were measured directly from tail vein whole blood via a glucose monitor at 0, 5, 10, 20, and 30 min after insulin administration.

Tissues were snap frozen in liquid nitrogen and pulverized before resuspension in 250 µL of 1% sodium deoxycholate (SDC) in 100 mmol/L Tris pH 8.5 buffer and boiled at 95°C for 10 min. Samples were homogenized by sonication using a tip-probe sonicator for 30 s at 70% power. Lysates were centrifuged at 20,000g for 10 min to remove cell debris and clarified by high-volume chloroform-methanol precipitation. One-part (200 µL) supernatant was mixed with four-parts chloroform (800 µL) and eight-parts methanol (1.6 mL) and sonicated for 30 s at 90% power. Four-parts (800 µL) Milli-Q H₂O was added and vortexed until the solution was completely milk-white. The solution was centrifuged at 2,000g for 5 min, and the top aqueous phase was carefully removed. Then, 16-parts (2.4 mL) methanol were added, and the pellet was washed for 30 s at 800 rpm. Centrifugation was repeated at 2,000g for 5 min and all liquid carefully decanted. The pellet was resuspended in 1% SDC and 100 mmol/L Tris-HCl (pH 8.5) by sonication at 60% power for 30 s with a tip-probe sonicator. Lysate was clarified by centrifugation at 20,000g for 10 min, and the supernatant taken for subsequent steps.

Protein concentration was determined by BCA assay. A total of 20 µg of protein was aliquoted into 1.5 mL centrifuge tubes and adjusted to 100 µL with H₂O before reduction/alkylation (10 mmol/L TCEP, 40 mmol/L 2-chloroacetamide) buffer was added, and the samples were boiled for 10 min at 95°C. Once cooled to room temperature, 0.4 mg trypsin and 0.4 mg LysC were added to each sample and incubated overnight (18 h) at 37°C with gentle agitation. Then, 100 µL 1% trifluoroacetic acid (TFA) in ethyl acetate was added to stop digestion and dissolve any precipitated SDC. Samples were prepared for mass spectrometry (MS) analysis by StageTip clean up using styrenedivinyl benzene reverse phase sulfonate (SDB-RPS) solid-phase extraction material (24). Briefly, three layers of SDB-RPS material were packed into 200 µL tips and 200 µL of samples were loaded onto StageTips by centrifugation at 1,000g for 5 min. StageTips were washed with subsequent spins at 1,000g for 5 min with 100 µL 1% TFA in ethyl acetate, then with 1% TFA in isopropanol, and 0.2% TFA in 5% acetonitrile (ACN). Samples were eluted by addition of 100 µL 60% ACN with 5% NH₄OH. Samples were dried by vacuum centrifugation and reconstituted in 20 µL 5% formic acid.

Samples were analyzed using a Dionex UltiMate 3000 RSLCnano LC coupled to a Q-Exactive HF-X mass spectrometer (Thermo Fisher Scientific). Briefly, 1 µg of peptide sample was injected onto an in-house packed 75-µm 55-cm column (1.9-µm particle size, ReproSil Pur C18-AQ) and separated using gradient elution, with Buffer A consisting of 0.1% formic acid in water and Buffer B consisting of 0.1% formic acid in 80% ACN. Samples were loaded to the column at a flow rate 0.4 mL/min at 100% Buffer A for 12 min, before ramping to 19% Buffer B over 40 min, and then to 98% Buffer B over 20 min and held for 10 min. Eluting peptides were ionized by electrospray with a spray voltage of 2.4 kV and a transfer capillary temperature of 300°C. Mass spectra were collected using a data-independent acquisition (DIA) method with varying isolation width windows (widths of 27 to 589 charge/mass ratio [m/z]) between 350 and 1,650 according to Table 2. MS spectra were collected between m/z 350 and 1,650 at a resolution of 120,000. Ions were fragmented with a higher-energy C trap dissociation collision energy at 25% and tandem MS spectra collected between m/z 300 and 2,000 at resolution of 30,000, with an automatic gain control target of 3e⁶ and the maximum injection time set to automatic.

Proteomics raw data files were searched using DIA-NN 1.8.1 using a library free FASTA search against the reviewed UniProt mouse proteome (downloaded August 2022) with deep learning enabled (25). The protease was set to trypsin/P with 1 missed cleavage, N-terminal methionine excision, carbamidomethylation, and methionine oxidation options on. Peptide length was set to 7–30, precursor range 350–1,650, and fragment range 300–2,000, and the false discovery rate was set to 1%. Data for GLUT4 were isolated to verify reduced expression levels.

All analysis and data visualization were conducted using the R programming environment (26). Unpaired Student t tests were performed to determined significance, except for when comparing basal and insulin-stimulated 2DG uptake within the same mouse, where paired tests were conducted. The Welch correction was applied since sample variances were not assumed to be equal between groups. Note is made where other statistical tests are performed. Significance is represented with a P < 0.05.

All data are available upon request.

A pilot experiment was conducted to determine the feasibility of measuring basal and insulin-stimulated 2DG uptake rates in the same mouse by sequential injection of two radiolabeled 2DG tracers (Fig. 1A). Chow-fed male C57BL/6J mice (12 weeks old) were anesthetized by intraperitoneal injection of sodium pentobarbital to enable retro-orbital injections of tracers and insulin. An increase in blood glucose levels occurred immediately following injection of the anesthetic (Fig. 1B). Glucose levels stabilized 20 min postinjection, although they remained elevated compared with levels prior to anesthesia.

The first tracer ([¹⁴C]2DG) was administered 10 min after anesthesia. Injection of insulin 40 min later elicited a significant reduction in the blood glucose concentration and led to more rapid disappearance of the insulin-stimulated tracer ([³H]2DG) than observed under basal conditions (Fig. 1C), consistent with an increased rate of 2DG uptake by peripheral tissues in response to insulin. Quantification of tissue-specific abundance of each 2DG-6P tracer revealed more rapid accumulation of the insulin-stimulated tracer in all insulin-sensitive tissues (Fig. 1D). Data from metabolically active tissues were compared with lung to assess the ability of the dual tracer approach to identify insulin action in individual tissues independently of systemic factors such as blood flow. Consistent with data from an established single tracer approach (27), lung 2DG uptake increased 30% following insulin administration. Insulin responses in metabolically active tissues varied as expected (27,28) and greatly exceeded that of lung, illustrating the ability of our approach to quantify tissue-specific insulin action in vivo.

It was important to assess the impact of insulin administration on uptake of the basal tracer as ∼30% of the tracer remained in circulation at the time of insulin stimulation. Separation of the [¹⁴C]2DG disappearance curve between basal (0–40 min) and insulin-stimulated (40–80 min) phases revealed that the rate of tracer disappearance from circulation (k_c) was not significantly affected by subsequent insulin administration (k_c[basal] = −0.032 ± 0.004 ln[disintegration per minute (DPM)/µL]/min, k_c[insulin] = −0.033 ± 0.003 ln[DPM/µL]/min; P = 0.68, paired t test). A second pilot experiment was performed to assess the impact of insulin administration on basal 2DG uptake into individual tissues. Mice were anesthetized and administered basal tracer ([¹⁴C]2DG) for 40 min, with or without a second injection of insulin, and the insulin-stimulated tracer ([³H]2DG) for 40 min (Fig. 1E). The Dual Tracer Test progressively overestimated basal 2DG uptake in more insulin-sensitive tissues, and although this did not significantly affect calculation of the basal 2DG rates for most tissues, basal 2DG uptake was significantly overestimated in the heart (Fig. 1F). However, in all other tissues, these data show that tissue-specific basal 2DG uptake rates were not significantly influenced by subsequent insulin administration.

Diet-induced obesity is a workhorse of metabolic research used by numerous investigators in the field. We fed 10-week-old male C57BL/6J mice a chow or HFD for 6 weeks. The metabolic consequences of HFD exposure were confirmed at 4 weeks using an oral glucose tolerance test. HFD-fed mice demonstrated glucose intolerance (Fig. 2A and B) and fasting hyperinsulinemia (Fig. 2C). At 6 weeks, the Dual Tracer Test detected a trend toward attenuated systemic glucose disposal (Fig. 2D and E) and a significant reduction in insulin-stimulated tracer disappearance (Fig. 2F and G), indicative of IR in peripheral tissues. We observed a higher initial concentration of tracer in chow-fed mice, especially of insulin-stimulated [³H]2DG, despite injecting a constant 5 μCi per mouse, suggesting that HFD-fed mice have a greater volume of distribution or reduced blood flow. Nevertheless, this did not affect the results as the rate of tracer disappearance from circulation was not influenced by the initial tracer dose. The Dual Tracer Test detected significant HFD-induced IR in soleus muscle and subcutaneous and epididymal adipose tissues (Fig. 2H). Other muscles, including red quadriceps, extensor digitorum longus (EDL), heart, and diaphragm, displayed preserved insulin responses with HFD, demonstrating the capacity of our approach to identify tissue-specific changes in insulin action. These data are consistent with previous studies using the clamp in diet-induced obese rats, where HFDs induced IR in soleus but not in other tissues, such as the heart or EDL (29), emphasizing the capacity of the Dual Tracer Test to produce results comparable with those from established methods.

A primary advantage of the Dual Tracer Test is that it halves the number of mice required for experiments comparing a single experimental group with a single control group, as is typical of experiments in inbred mice, since both basal and insulin-stimulated measures are obtained from every mouse. Unexplained variation between genetically identical inbred animals may confound results when comparing basal and insulin-stimulated 2DG uptake between distinct animals. We hypothesized that paired measurement of basal and stimulated 2DG uptake would enable insulin action to be measured more consistently compared with approaches using a single-tracer. Insulin-stimulated 2DG uptake in chow-fed mice was compared with the insulin-stimulated fold change in 2DG uptake (data in Fig. 2H) to assess this potential benefit. Only tissues that showed a 2DG uptake defect in response to diet could be included in this analysis. Accounting for differences in basal uptake between mice reduced the coefficient of variation for insulin response by 30% on average (95% CI 11–47%) for chow-fed mice (Fig. 3A), suggesting that combined basal and insulin measures from the same mouse are more consistent than those between mice.

Phenotypic variation within an inbred mouse strain reportedly increases in response to HFD feeding. For example, male C57BL/6J mice have 3–6 g of fat when fed a chow diet but between 5 and 30 g following HFD feeding (30). Interestingly, we observed similar variation in tissue-specific insulin action using the Dual Tracer Test; four of eight HFD-fed C57BL/6J mice were IR in red quadriceps, while the remainder had a robust response to insulin (Fig. 2H). These four “nonresponsive” mice also demonstrated the smallest responses to insulin in soleus, EDL, and subcutaneous adipose tissue, indicating that this is likely a biologically meaningful observation. Chow-fed mice responded consistently to insulin, and this contributed to the lower variation between mice in chow-fed conditions. Despite this, correcting for basal 2DG uptake in HFD-fed mice still provided improved statistical power compared with a single tracer approach in epididymal and brown adipose depots (Fig. 3A). This reduced variation was not observed in soleus or subcutaneous adipose tissue since IR was more marked in a subset of HFD-fed mice and biological variation between mice is more readily detected using the Dual Tracer Test.

Power simulations were conducted to determine whether the Dual Tracer Test would outperform studies only measuring insulin-stimulated glucose uptake and thus reduce the number of required mice. The Dual Tracer Test exhibited little benefit for tissues with strong diet-induced defects, but significantly enhanced the ability to detect subtle defects in brown adipose tissue (Fig. 3B). This suggests a minor improvement in statistical power for studies using inbred mice to identify IR in specific tissues; however, paired collection of basal and insulin-stimulated data using the Dual Tracer Test greatly assists mechanistic interpretation of data without increasing the number of mice in the study.

We next evaluated the potential benefit of the Dual Tracer Test for experiments in Diversity Outbred mice, where very large sample sizes are typically required to identify genetic signals affecting phenotype, and measurement accuracy has greater potential to impact study size. Power simulations were conducted to detect quantitative trait loci using insulin-stimulated 2DG uptake alone or the insulin-stimulated fold change, as before. The power to detect a quantitative trait locus associated with a phenotype depends on the allele’s effect size and frequency as well as genetic and technical variances. The allele frequency was fixed at a minor allele frequency of 5% and an effect size explaining ∼10% of phenotypic variation. Simulations were performed over a range of minor allele frequencies and effect sizes, and this did not alter interpretation of the results. Simulations were performed 1,000 times for an increasing sample size fed a chow diet. Even though soleus displayed the smallest reduction in the coefficient of variation when using the Dual Tracer Test, this was sufficient to reduce the number of mice required to obtain 90% power by almost 100 (Fig. 3C). In subcutaneous adipose tissue, which demonstrated the largest reduction in contraction velocity for chow-fed mice, the Dual Tracer Test reduced the required sample size from 710 to 210 mice (Fig. 3D). While the benefit appears to vary by tissue, perhaps due to some tissues being more challenging to dissect precisely, these analyses suggest that the Dual Tracer Test will improve the capacity to interrogate genetic and environmental contributions to IR in individual tissues.

The Dual Tracer Test was further validated using a genetic model since these are essential to hypothesis-driven research and it was important to further assess the robustness of the dual tracer approach.

Our laboratory recently generated a unique GLUT4-deficient mouse model. This serendipitous model resulted from efforts to generate a reporter mouse to visualize in vivo GLUT4 trafficking by inserting a pHluorin tag into the first exofacial loop of the GLUT4 gene (21). Overexpression of the GLUT4-pHluorin construct in 3T3-L1 adipocytes significantly enhanced insulin-stimulated glucose uptake compared with an empty vector control (Fig. 4A), suggesting that the GLUT4-pHluorin product is functional. However, we were unable to detect a significant fluorescent signal in tissues of mice homozygous for the GLUT4-pHluorin construct (data not shown). To determine whether the construct was being expressed correctly, we performed qPCR in epididymal white adipose tissue (Fig. 4B). Transcription of pHluorin-tagged GLUT4 was dependent on copy number; however, insertion of the pHluorin tag also led to a decrease in total transcription of the GLUT4 gene. This suggested that mice homozygous for the pHluorin-tagged GLUT4 (G4KD_hom) should demonstrate ∼50% reduced GLUT4 expression compared with WT littermate controls. At the protein level, a minor product was detected in various metabolic tissues of the correct molecular weight for GLUT4-pHluorin (82 kDa). As expected, this product was lowly expressed compared with endogenous GLUT4 (Fig. 4C). Semiquantitative MS analysis of heart, soleus, and epididymal white adipose tissues revealed that GLUT4 was expressed at 25% of the level found in WT mice (Fig. 4D), implying that the GLUT4-pHluorin product is also repressed posttranslationally since the reduction in GLUT4 at the protein level exceeded that observed at the transcript level. Posttranslational repression of pHluorin-tagged proteins has been reported previously (31), where it was concluded that pHluorin is unable to fold correctly due to the oxidizing state of the endoplasmic reticulum. Thus, G4KD_hom mice exhibit consistent reductions in GLUT4 across a variety of metabolically relevant tissues, likely due to both transcriptional and posttranslational repression of the pHluorin-tagged construct.

Strikingly, G4KD_hom mice displayed metabolic abnormalities in the absence of changes in body composition compared with WT (Fig. 4E and F). This is notable, as previous studies have reported that GLUT4-null mice possess very little adipose tissue (32) and that the G4KD_hom model overcomes this confounding factor. GLUT4-deficient mice demonstrated systemic IR in an intraperitoneal insulin tolerance test (Fig. 4G), consistent with ubiquitous reduction of GLUT4 levels. Tissue-specific IR was confirmed in epididymal white adipose tissue (Fig. 4H) using an established technique for assessing insulin action ex vivo (6). Aberrant systemic glucose homeostasis was detected by oral glucose tolerance test (Fig. 4I–K). Intriguingly, hyperinsulinemia was not detected in GLUT4-deficient mice in the fasted state or within 15 min of an oral glucose dose (Fig. 4K), despite marked peripheral IR and hyperglycemia at the 15-min time point. Thus, these mice provided an ideal genetic model of systemic IR with which to test the utility of the Dual Tracer Test.

The Dual Tracer Test was performed in 15- to 20-week-old chow-fed male G4KD_hom and WT mice (Fig. 5). Consistent with the results of the intraperitoneal insulin tolerance test (Fig. 4G), G4KD_hom mice exhibited a reduced systemic response to a maximal dose of insulin, quantified as the area above the glucose excursion curve upon insulin administration (Fig. 5A and B). Insulin-stimulated tracer disappearance was significantly dampened in G4KD_hom mice relative to WT controls (Fig. 5C and D). The tissue-specific insulin response is reported as the fold change between basal and insulin-stimulated tracer uptake rates within individual mice (Fig. 5E and F). The Dual Tracer Test identified the presence of IR in all G4KD_hom tissues except for the lung negative control, as expected based on the IR observed ex vivo in epididymal white adipose tissue explants (Fig. 4H). Despite significant impairments in insulin action, basal 2DG uptake rates were not significantly different between genotypes (Fig. 5E). The extent of IR was similar across all insulin-sensitive tissues (32.6 ± 0.04% reduction compared with WT), consistent with a marked reduction in GLUT4 levels in all these tissues. The observation that some insulin-stimulated 2DG uptake was retained in G4KD_hom tissues is consistent with the GLUT4-pHluorin protein being functional (Fig. 4A), albeit lowly expressed compared with GLUT4 in WT mice. Interestingly, brain 2DG uptake was significantly increased by insulin in WT, and this effect was ablated in G4KD_hom mice, suggesting that glucose disposal into the brain is regulated by GLUT4. Collectively, these data demonstrate that GLUT4-deficient mice exhibit IR in all insulin-sensitive tissues and that the Dual Tracer Test robustly identified these differences.

Despite enormous investment in the generation of various mouse models, there remains a paucity of technically simple and rigorous methods for studying insulin sensitivity in vivo in rodents. Many studies infer insulin sensitivity from glucose and insulin tolerance tests, although these data are difficult to interpret given potentially different contributions from the liver compared with peripheral tissues and variable insulin release from the pancreas during glucose tolerance tests (19,33). Existing single tracer approaches are incompatible with significant heterogeneity between genetically identical mice in inbred strains and cannot be readily applied to the study of different genetic backgrounds or outbred populations. This report describes the Dual Tracer Test, a highly reproducible approach to quantify both basal and insulin-stimulated 2DG uptake rates in individual tissues in the same mouse. To assist critical interpretation of the validity and results of future studies using the Dual Tracer Test, a section describing the minimum data reporting standards of the technique is included toward the end of the Discussion.

The Dual Tracer Test builds upon an anesthetized single tracer approach originally devised by Hom et al. (20) and can be performed by any investigator proficient in injecting materials into mice and collecting tissues. The method is time-efficient and readily extendable to the study of 2DG uptake rates in other experimental contexts, just as many variations of the clamp exist. For example, the effects of a pharmacological agent on glucose uptake could be easily assessed by performing the second injection with that compound instead of insulin. Additionally, the dual tracer approach could easily be implemented within a clamp protocol to provide paired basal and insulin-stimulated glucose transport rates under controlled glucose and insulin levels. The ease, robustness, and flexibility of the Dual Tracer Test make it a valuable addition to the metabolic phenotyping armamentarium.

Many changes in glucose uptake measured in the presence of insulin are due to changes in basal rates and are independent of insulin stimulation (29,34). Thus, mechanistic interpretation of insulin action may be confounded if basal glucose uptake rates are not measured accurately. For example, many of the changes in insulin action observed in our dietary model of IR were accompanied by changes in basal 2DG uptake. Aside from the heart, basal 2DG uptake was not significantly altered by subsequent insulin stimulation in tissues in C57BL/6J (Fig. 1E) or in G4KD_hom mice, suggesting that any change in basal uptake is a consequence of HFD feeding rather than an artifact of the dual tracer approach or a consequence of reduced glucose disposal. Marked variation in insulin responsiveness exists between different tissues in both healthy and perturbed states (Figs. 1D, 2H, and 5 F) and the tissue-specific mechanisms by which IR manifests remain unclear. It is essential that methodologies able to capture this diversity are made available to gain insight into tissue-specific mechanisms of IR and to explain how IR in varying tissues can lead to distinct metabolic outcomes (5). The ability to quantify both basal and insulin-stimulated responses in all tissues from the same mouse simultaneously is, therefore, a major advance. The effects of the HFD feeding and reduced GLUT4 on insulin action, measured with the Dual Tracer Test, were consistent with those expected from a gold standard clamp, emphasizing the robustness of this new assay.

Comprehensively identifying the tissue-specific causes and consequences of IR will require leveraging phenotypic diversity across a wide array of genetic and dietary settings. Recombinant inbred mouse panels have successfully identified candidate genes affecting metabolism (7,35,36), and newer diversity outbred populations offer finer genetic resolution and a more complete understanding of the genetic architecture of complex traits (13). We suggest that the Dual Tracer Test is ideally suited to interrogate tissue-specific action in such populations due to its relative simplicity and sample size reductions.

All methods for assessing whole body metabolism in mice possess limitations, and the Dual Tracer Test is no exception. It is important for future users to be aware of these limitations to evaluate the relative utility of the method. These are as follows:

Anesthesia has been reported to cause suppression of hepatic glucose output and IR in peripheral tissues (37). Although pentobarbitone causes sustained hyperglycemia in 18-h fasted mice (38), the Dual Tracer Test is performed in 2-h fasted mice, where this effect is relatively modest (Figs. 1B, 2D, and 5,A). The mechanism by which pentobarbital triggers hyperglycemia likely requires both hepatic, and muscle IR, both of which have been reported (37,39). Interestingly, anesthesia was found to principally affect insulin action in tonic muscles, such as soleus, and brown adipose tissue, concomitant with reduced blood flow (40). Insulin action in other muscles and adipose depots was unaffected. Even though anesthesia impacts glucose uptake in the insulin-stimulated state, the Dual Tracer Test was clearly able to detect robust two- to threefold increases in insulin-stimulated 2DG uptake in a range of tissues, differences in insulin action between tissues analogous to that seen during conscious clamps, and effects of both dietary (Fig. 2F–H) and genetic (Fig. 5C–F) manipulation similar to that previously reported with using the clamp (41,42). While conscious clamps in nonrestrained animals have clear advantages over other methods, many of these, such as glucose or insulin tolerance tests in conscious animals, possess important limitations. Most notably, animal handling combined with administration of reagents either orally or intraperitoneally often causes stress and a period of increased activity. Since activity increases muscle glucose uptake much more than insulin in mice (43), this likely contributes to considerable noise in tissue-specific responses and was a central motivation for the development of the Dual Tracer Test.

Anesthesia was required to facilitate retro-orbital injection of the tracers. Retro-orbital injections (18), while similar to tail vein injections (44), are technically much easier and allow simultaneous blood sampling from the tail vein without concern of tracer contamination from the injection site. This is essential to establish the plasma disappearance curves from each mouse during the Dual Tracer Test. Delivery of tracers and insulin directly into the circulation enabled efficient distribution to peripheral tissues. Intraperitoneal injections were unsuitable as absorption from the intraperitoneal space is slow, and the mechanism of transport of 2DG from the intraperitoneal space into the circulation is ill-defined and could be affected by genetic variation. A single injection of pentobarbitone was sufficient to anesthetize mice for the full duration of the Dual Tracer Test (90 min), and the plane of anesthesia was constant during this time. This meant that a stress response was observed only with initial handling and not during subsequent tracer injections. Regardless, future users of the Dual Tracer Test should be aware that peripheral 2DG uptake is likely being measured under conditions of reduced total glucose uptake.

Glucose and insulin levels change throughout the Dual Tracer Test, just as in common insulin tolerance and single tracer tests. Importantly, the rate of 2DG uptake is not based on the specific activity of the tracer, and there is no attempt to convert this rate into a measure of absolute glucose uptake. 2DG is present in tracer quantities only, and it will decrease in the circulation at a rate comparable to the rate of glucose disappearance. More insulin sensitive animals experience hypoglycemia (Figs. 2D and 5 A) and counterregulatory responses are likely activated. The latter portion of insulin-stimulated 2DG uptake is therefore assessed under conditions of hypoglycemia. Future users should be aware that this may lead to an underestimation of insulin action in more insulin-sensitive groups and potentially generate false-negative results. For example, the Dual Tracer Test only identified a trend toward IR in brown adipose tissue from HFD-fed C57BL/6J mice (Fig. 2H), which the clamp would likely be better able to identify clearly.

Insulin clearance in mice varies significantly with genetics (45), diet, and exercise (46) and provides a challenge for investigating submaximal insulin responses. To circumvent this, we used a maximal dose of insulin to study insulin action in isolation. This method will therefore overlook changes in insulin sensitivity that can only be observed at submaximal insulin doses, and so this limitation must be recognized. IR in humans is evident at both submaximal and maximal insulin doses (47), and consistent with this, we observed IR in select tissues of C57BL/6J mice fed the HFD for 6 weeks. It is therefore unlikely that this limitation significantly impairs the ability of the Dual Tracer Test to detect IR.

The 2DG method was originally designed such that tissue measures would be acquired 40 min after tracer administration to ensure maximum removal of the tracer from the bloodstream. This is important for studies using fluorodeoxyglucose combined with positron emission tomography, where high levels of the tracer in the blood may significantly contribute to tissue signals. This problem is overcome here by specifically measuring tissue levels of 2DG-6P in tissues. However, it is conceivable that insulin delivered with the second tracer may also influence the uptake of the basal tracer in the Dual Tracer Test, thus artifactually elevating the estimate of basal 2DG uptake. Most of the basal tracer is removed from circulation within the first 15 min, and uptake of the basal tracer between 40 and 80 min is likely to be small as the rate of uptake of [³H]2DG will be greater than that of [¹⁴C]2DG during this time since it is present in vast excess, consistent with the principles of mass action. However, it was still expected that the Dual Tracer Test would progressively underestimate insulin action in more insulin-sensitive tissues. To address this, we conducted pilot studies and observed that insulin administration did not significantly affect accumulation of the basal tracer within most of the tissues, with the notable exception of the heart, where basal 2DG uptake was significantly overestimated (Fig. 1E). Investigators interested in assessing glucose uptake in the heart should therefore be aware of this limitation of the method.

As is the case with any new method, it is important to establish minimum data criteria for implementation of the technique in other laboratories so that the validity of future studies can be evaluated consistently and so that future users do not contribute to methodological drift. Figure 5 provides a visual overview of these requirements.

Blood glucose concentrations (preanesthesia, immediately after tracer injections, and routinely throughout the test) (Fig. 5A) need to be measured to assess the glycemic response to handling stress and the anesthetic. Glucose levels postinsulin administration should be presented to enable the extent of hypoglycemia to be considered, particularly when comparing different mice. Increasing glucose levels between 70 and 80 min indicate reactivation of hepatic glucose production and restoration of glucose homeostasis. These processes are inhibited if core body temperature is not maintained during anesthesia and results in prolonged hypoglycemia, elevated counterregulatory responses, and IR (48,49). Future users of the technique should refer to the Research Design and Methods for a simple approach to prevent these outcomes. The AOC in the insulin-stimulated phase provides insight into insulin action at the systemic level independently of fasting glucose levels (Fig. 5B). The AOC can also be used to corroborate the presence of muscle IR. The area under the glucose curve should not be reported as it is partly influenced by fasting glucose levels, and investigators should instead report these directly if they wish to discuss these. It is important to note that the Dual Tracer Test is chiefly concerned with measuring insulin responses in peripheral tissues and that other tests are required to comment on the systemic consequences of any defects detected. For example, glucose tolerance tests with concurrent measures of blood insulin levels would be required to establish an association between IR identified in the Dual Tracer Test and glucose intolerance; systemic glucose homeostasis cannot be inferred from tissue-specific IR.

Plasma tracer levels are required to calculate tissue-specific uptake rates and should be explicitly presented to demonstrate proficient injection of materials and provide insight into glucose uptake at the systemic level (Fig. 5C). As it can be challenging to infer rates of tracer disappearance directly from the tracer disappearance curves, decay constants should be presented alongside to aid comparison (Fig. 5D). Finally, tissue-specific data should be presented to assess the presence of defects in insulin action in individual tissues (Fig. 5E and F). Measures should be collected in the lung as it is an important negative control for insulin action and should be consistent between experimental groups. It is also essential that future authors demonstrate effective separation of [³H]2DG and [¹⁴C]2DG using their specific radioactive counting instrumentation. If future users wish to reduce the duration of the basal uptake phase, it is essential that they replicate the experiment shown in Fig. 1E with their desired time point to demonstrate that basal 2DG uptake remains unaffected by insulin stimulation in the presence of a greater amount of basal tracer in the circulation. Shortening of the insulin-stimulated phase is possible, assuming that the duration is long enough for sufficient 2DG-6P to accumulate within tissues.

The physiological implications of a complete loss of GLUT4 at a whole-body level are debatable as changes in body composition confound interpretation of the effects of GLUT4-mediated IR on systemic metabolism (50). A partial GLUT4 knockdown model (GLUT4 expressed at ∼25% of WT in G4KD_hom) was fortuitously generated by our laboratory, and several critical observations have emerged. The G4KD_hom mice exhibited profound changes in glucose and insulin tolerance (Fig. 4G and I), supporting a major homeostatic role of peripheral insulin sensitivity. Insulin action was ablated in all muscle and fat tissues as well as the brain, suggesting that GLUT4 is essential in all of these tissues (Fig. 5F). This is consistent with reduced 2DG uptake in brain-specific GLUT4 knockouts (51) and, importantly, the unique ability of the Dual Tracer Test to measure both basal and insulin-stimulated 2DG uptake in the same mouse enabled us to conclude that GLUT4 reductions selectively affect insulin-stimulated but not basal glucose metabolism in the mouse brain.

Ubiquitous peripheral IR was accompanied by glucose intolerance in G4KD_hom mice, suggesting that increased insulin secretion would be required to normalize blood glucose. However, hyperinsulinemia was not observed in fed or fasted states in GLUT4-deficient mice, contrary to the popular belief that peripheral IR is sufficient to drive the development of hyperinsulinemia over time. This is an exciting observation that warrants further investigation in our G4KD_hom model.

Acknowledgments. The authors thank technical staff in the Garvan Molecular Genetics facility, at Australian BioResources (Mossvale, New South Wales, Australia), and in the Laboratory Animal Services at the Charles Perkins Centre of The University of Sydney. EchoMRI was performed in the Sydney Imaging Facility, The University of Sydney. Proteomics was performed in the Sydney Mass Spectrometry Facility, Charles Perkins Centre, The University of Sydney.

Funding. This work was supported by an Australian Research Council Laureate Fellowship (to D.E.J.). This research was also supported by an Australian Government Research Training Program (RTP) Scholarship (to H.B.C.).

The content is solely the responsibility of the authors and does not necessarily represent the official views of the Australian Research Council.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.B.C. contributed to designing the study, performing experiments, analyzing results, and writing the manuscript. S.M., S.W.C.M., J.S., M.E.N., and G.J.C. contributed to designing the study, performing experiments, and editing the manuscript. K.C.C. and M.P. contributed to performing experiments. J.G.B. and D.E.J. contributed to designing the study and editing the manuscript. D.E.J. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.