Hyperinsulinemia does not cause de novo capillary recruitment in rat skeletal muscle

Abstract Objective The effect of insulin on blood flow distribution within muscle microvasculature has been suggested to be important for glucose metabolism. However, the “capillary recruitment” hypothesis is still controversial and relies on studies using indirect contrast‐enhanced ultrasound (CEU) methods. Methods We studied how hyperinsulinemia effects capillary blood flow in rat extensor digitorum longus (EDL) muscle during euglycemic hyperinsulinemic clamp using intravital video microscopy (IVVM). Additionally, we modeled blood flow and microbubble distribution within the vascular tree under conditions observed during euglycemic hyperinsulinemic clamp experiments. Results Euglycemic hyperinsulinemia caused an increase in erythrocyte (80 ± 25%, P < .01) and plasma (53 ± 12%, P < .01) flow in rat EDL microvasculature. We found no evidence of de novo capillary recruitment within, or among, capillary networks supplied by different terminal arterioles; however, erythrocyte flow became slightly more homogenous. Our computational model predicts that a decrease in asymmetry at arteriolar bifurcations causes redistribution of microbubble flow among capillaries already perfused with erythrocytes and plasma, resulting in 25% more microbubbles flowing through capillaries. Conclusions Our model suggests increase in CEU signal during hyperinsulinemia reflects a redistribution of arteriolar flow and not de novo capillary recruitment. IVVM experiments support this prediction showing increases in erythrocyte and plasma flow and not capillary recruitment.


| INTRODUC TI ON
The skeletal muscle vasculature delivers insulin and glucose to muscle cells, and skeletal muscle capillarization is an important contributing factor to insulin sensitivity. 1 Blood flow distribution within the muscle microvasculature might therefore be important for glucose metabolism. Over the last two decades, a number of studies have suggested that insulin affects the distribution of blood flow within the muscle microvasculature [2][3][4][5][6][7] It is believed that insulin increases the number of capillaries that receive blood flow and thereby enhances the delivery of glucose and insulin to the myocyte. This phenomenon has been termed "capillary recruitment." Whether capillary recruitment actually occurs has frequently been debated. [8][9][10][11] The proponents of the capillary recruitment hypothesis rely mainly on indirect data collected using contrast-enhanced ultrasound (CEU). 2,3,[5][6][7] Opponents of the capillary recruitment hypothesis tend to point to direct observation of the microvasculature using intravital video microscopy (IVVM). Indeed, most [12][13][14][15][16][17][18][19] but not all 20,21 IVVM studies show that 80%-95% of all capillaries have red blood cells or plasma flowing through them at baseline and that there is no capillary recruitment of additional vessels with contraction 14,16,19 or vasodilator stimulation. 17 IVVM techniques have been criticized for possibly creating a hyperemic state that is unable to exhibit capillary recruitment in response to normal physiological stimuli due to the potential for surgical trauma or the use of superfusate solutions. However, there is strong evidence that IVVM preparations do reflect the normal physiological response to stimuli since these preparations have been instrumental in uncovering fundamental mechanisms of microvascular blood flow regulation (Segal; 22 Murrant and Sarelius 23 ).
Although IVVM does not typically show "de novo" capillary recruitment, that is, a reserve of capillaries with no red blood cell or plasma flow that can be recruited when needed, IVVM does show a more uniform distribution of RBCs among these perfused capillaries which reduces the heterogeneity of capillary hematocrits and supply rates. This network flow redistribution is primarily governed by rheological factors that passively occur in response to an increase in blood flow, not to the active "opening" of new capillary flow paths. 24 Some have also proposed that the use of thin muscle preparations such as the spinotrapezius or cremaster might not be representative of larger muscles involved in locomotion or lifting. 8,9 For this study, we chose to use the extensor digitorum longus (EDL) muscle, a relatively thick (>1 mm), locomotor muscle in the rat hind limb that can be surgically reflected without damage to the muscle belly or disruption of tissue blood flow, and minimal manipulation of surrounding tissues. 19 The available literature also suggests that surgical exteriorization does not affect vascular function. 25,26 We also have extensive experience with this IVVM preparation of the EDL muscle. Under baseline conditions, 85%-90% of capillaries containing red blood cells (RBC) are continuously perfused with RBCs. The remaining 10%-15% have either intermittent or stopped RBCs flow. Capillaries with plasma flow only cannot be consistently detected with our microscopy setup, nor can we detect "closed" capillaries if they exist. Perfused capillaries have a mean RBC velocity, RBC supply, and hemoglobin O 2 saturation of 0.150 mm/s, 7.5 RBC s −1 , and 65%, respectively. 13 These results are consistent with a healthy microvascular bed at physiological baseline conditions and should therefore serve as a good model for testing the effect of insulin on the microcirculation. Thus, the primary objective of the present study was to perform euglycemic hyperinsulinemic clamp experiments on rats for the first time combined with IVVM analysis of the microcirculation in the EDL muscle to test whether hyperinsulinemia leads to capillary recruitment and redistribution of microvascular blood flow in skeletal muscle. Blood flow was determined by measuring capillary RBC velocity, hematocrit, and diameter, and using these values to calculate RBC, plasma and blood flow rates applying established relationships. Since glucose and insulin are carried in the plasma, we were thus able to determine whether there was evidence of their redistribution with hyperinsulinemia.
We also defined a secondary objective, which was to attempt to reconcile CEU and IVVM results.
The increase in ultrasound signal with microbubbles during hyperinsulinemia may be explained without de novo capillary recruitment in a way that is consistent with IVVM data. The size of the rat EDL muscle makes it impractical to make a direct comparison between the two techniques, and our previous attempts to visualize fluorescently labeled microbubbles in the EDL were unsuccessful, likely due to the low density of microbubbles relative to RBCs. Hence, we chose to develop a mathematical model of blood flow and microbubble distribution within the microvasculature.
Although microbubbles have been proposed to distribute in a manner similar to erythrocytes, 27 it is also possible that they are more sensitive to the flow distribution at bifurcations than erythrocytes due to their rigidity. 28 This would account for the enrichment of microbubbles relative to erythrocytes in some flow paths as reported by Lindner and co-workers. 27 Our mathematical model examines the effect of a redistribution of blood flow on microbubble distribution to investigate whether it is possible to reconcile CEU and IVVM results.

| Animals
Intravital video sequences of capillary networks in the EDL muscle of eight rats at basal conditions and during a hyperinsulinemic euglycemic clamp were used as the basis for the present work. Animal protocols were approved by the Animal Care and Use Committee of the University of Western Ontario. Male Sprague Dawley rats, seven weeks of age (n = 8; Charles River), were housed in dedicated animal quarters at the University of Western Ontario with free access to food (Purina LabDiet RMH 3000) and water on a 12/12-h light/dark cycle. The rats were acclimatized for 1 week after delivery. On the day of the experiment, the rats weighed 165 ± 4 g.

| Surgical preparation of rats
The animals were anesthetized with an intraperitoneal, IP, bolus of αchloralose (80 mg/kg, Fluka Analytical #23 120-100 g), and urethane (500 mg/kg, Sigma-Aldrich U2500-500g) in saline. All rats were instrumented with catheters (Tygon S-54-HL Microbore Tubing, inner diameter: 0.41 mm, Norton Performance Plastics) in the right jugular vein (for infusion of insulin, glucose, and anesthesia) and left carotid artery (for blood sampling and blood pressure measurements) under aseptic conditions. Tracheotomy was performed, and a tracheotomy tube (plastic end of IV catheter, 18 G) was inserted to facilitate spontaneous ventilation.
Throughout the experiment, the animal was kept anesthetized by infusing a mixture of α-chloralose (80 mg/kg, Fluka Analytical #23 120-100 g) and urethane (500 mg/kg, Sigma-Aldrich U2500-500g) in saline at a rate of 0.8-1.2 mL/h. Depth of anesthesia was assessed by continuous monitoring of mean arterial blood pressure which was kept within a range of 90-100 mm Hg by varying the infusion rate of the anesthetic.
The EDL muscle, a bellied muscle in the hind limb, was chosen for this study since the muscle is easily exposed with minimal trauma and the collagen sheath covering the muscle is very thin, permitting good visualization of the microcirculation. The EDL muscle was prepared for in vivo microscopy using blunt dissection and externalized as previously described. 19 Briefly, a small section of the skin was removed from the lateral side of the right lower hind limb, exposing the fascia capsule of the underlying muscles. Superficial dissection of the capsule and blunt separation of the surrounding muscles allowed the EDL to be isolated. Silk ligature was threaded under the intact muscle and secured with a square knot on the distal portion of the EDL tendon. The tendon was then severed between the ligature and muscle insertion, leaving the ligature securely attached to the free end of the EDL tendon. After EDL dissection, the animal was transferred to the microscope stage and placed on its right side in a semi-prone position. The ligature secured to the EDL tendon was then taped to the stage such that the lateral side of the muscle was facing the objectives and the muscle maintained a length approximate to the resting position in vivo. The muscle was moistened with 37°C saline and covered on the medial side with a small square of plastic film (~2 × 2 cm, polyvinylidene chloride, Saran) and a glass coverslip to isolate the muscle from the external environment and ensure that the microvasculature is the only O 2 source for the tissue.

| Baseline and hyperinsulinemic euglycemic clamp measurements
Immediately following the surgical preparation, the animal rested on the microscope stage for 30 minutes prior to the start of the experiment. This pre-clamp period consisted of a 15-minute infusion of anesthesia followed by 15-minute infusion of anesthesia plus saline.
The experiment started with infusion of saline (~0.6 mL/h, Baxter Canada) for 30 minutes with arterial blood samples drawn every 5 minutes as described below. The volume of the saline infusion matched the combined insulin and glucose infusion during the last 30 minutes of the hyperinsulinemic euglycemic clamp.
Immediately following the baseline measurements, a bolus (300 ρmol/kg) of recombinant human insulin (Actrapid, Novo Nordisk) diluted to 1500 pmol/L in a buffer (pH 7.4) consisting of 140 mmol/L NaCl, 5 mmol/L Na 2 HPO 4 , and, in order to prevent protein adsorption, 70 ppm Tween20 was infused over 2 minutes followed by a constant rate (30 ρmol/kg/min) infusion for 70 minutes.

| Dual spectrophotometric intravital video microscopy
The muscle was transilluminated with a 75-W Xenon lamp and viewed through a Olympus IX-81 inverted microscope equipped with X10 and X20 objectives and a DualCam (parfocal beam splitter) fitted with 442 nm and 454 nm interference filters (10 nm bandpass) for absorption spectroscopy measurement of hemoglobin oxygen saturation. Simultaneous frame-by-frame video was captured at each wavelength using two identical Rolera XR digital video cameras streaming video sequences (696 X 520, 21 frames s -1 ) directly to a single acquisition computer using custom capture software (Neovision, Czech Republic). The two cameras were temporally synchronized and aligned such that video images were in register. Video sequences of capillary networks were acquired from the same ten 20X fields of view (FOV) at baseline with saline infusion and 45 minutes following initiation of the hyperinsulinemic euglycemic clamp.
Each FOV was captured for one minute (1260 video frames for each camera). The total acquisition time from capture of the first FOV to the last was twenty to thirty minutes. The same two overlapping 10X FOV were also acquired at baseline with saline infusion and 20 minutes after starting hyperinsulinemic euglycemic clamp. In all cases, each FOV was acquired for one minute (1260 frames) and the acquisition time for both was less than five minutes.

| Microvascular measurements
Hemodynamic measurements were made offline from video sequences of individual, in-focus capillaries within each field of view.
Automated measurements for RBC velocity (mm/s), lineal density (RBC mm −1 ), and supply rate (RBC s −1 ) were made on a frame-byframe basis from each 60s sequence using custom analysis software described elsewhere. 13,30,31 Briefly, segments of in-focus capillaries within each FOV were selected and outlined from functional images generated by processing the intravital video sequences. 32 Functional images provide high contrast delineation between tissue and the red blood cell column (luminal space swept out by the passage of RBCs) and were used to determine vessel diameter and segment length.
Functional capillary density in the EDL muscle was determined using the method described by Ellis et al 33  RBCs from plasma in order to measure RBC lineal density and to quantify the spatial displacement of the RBC column from frame-toframe to measure the RBC velocity. Capillary (tube) hematocrit for each video frame was calculated as the product of lineal density and RBC volume divided by the volume of the vessel segment. Vessel segment volume was calculated using mean vessel diameter over the segment length selected and by assuming the vessel had a circular cross section. Frame-by-frame RBC supply rate (RBC s −1 ) was calculated as the product of the RBC velocity (mm/s) and RBC lineal density (RBC mm −1 ). RBC oxygen saturations were determined from the ratio of RBC optical density at the two wavelengths based on an in vivo calibration. 34 Optical density was determined using the Lambert-Beer law, OD = log (I O /I RBC ) where I O is the plasma light intensity and I RBC is the RBC light intensity obtained from the space-time image.
The volumetric flow rates of erythrocytes, plasma, and blood (ρL/s) were calculated from the measured hemodynamic data above. RBC flow was calculated simply as the RBC supply rate (RBCs s −1 ) multiplied by the mean RBC volume for rat (65 μm 3 ; 35 ). Since we are unable to measure plasma velocity directly, plasma and blood flow rates were estimated using a published relationship for the Fahraeus effect. 36 This relationship uses the measured capillary (tube) hematocrit and capillary diameter to calculate the discharge hematocrit (discharge hematocrit is the hematocrit one would measure if the blood flowing through the vessel was collected in a reservoir). Blood velocity was calculated as the ratio of measured capillary (tube) hematocrit divided by calculated discharge hematocrit times the measured RBC velocity. Plasma velocity was calculated from the weighted blood and RBC velocity based on the capillary hematocrit. Blood and plasma flows were calculated assuming a circular cross section for each capillary.
The 10X overlapping FOV were analyzed for capillary hemodynamic data within each of the discrete capillary networks. Networks with measurements from at least four capillaries at baseline and 20 minutes following start of insulin infusion resulted in a pooled sample from seven animals of 17 discrete networks with a total of 147 capillaries. One animal was excluded from this analysis, because there were fewer than four capillaries in each capillary network that could be analyzed at both time points.

| Plasma protein concentrations
Rat C-peptide, rat insulin, and human insulin (used for infusion during hyperinsulinemic euglycemic clamp) were measured at Novo Nordisk facilities in Maaloev, Denmark, using in-house luminescent oxygen channeling immunoassays. 37 The lower limits of quantification for these assays are 18 pmol/L (Rat C-peptide), 20 pmol/L (rat insulin), and 15 pmol/L (human insulin in rat plasma).

| Microvascular network geometry and arteriolar asymmetry at bifurcations
Our baseline microvascular network geometry is based on published data on the structure of arteriolar trees in the rat EDL muscle. 38 Starting from a first-order arteriole (1A) with an inner diameter (D) of 75 μm, our network bifurcates a number of times (depending on the particular flow path) until reaching terminal arterioles with a mean diameter of ~11 µm. To model the redistribution of flow with hyperinsulinemia, we simulated a modified network with decreased asymmetry of diameters at each bifurcation (average of 5.5% vs 7% in the baseline network) and a 50% increase in total flow. The selection of parameters for the simulated hyperinsulinemia network reflected the IVVM measurements. We sought to simulate the effect this asymmetry would have on both blood and microbubble flow distribution.

| Blood flow model
Our steady-state two-phase blood flow model 39 is based on the model originally described by Pries and co-workers. 36 Since RBCs travel closer to the vessel center and have a higher average velocity than blood as a whole, the discharge (flow-averaged) hematocrit is larger than tube (volume-averaged) hematocrit. To compute blood and RBC flow in our modeled microvascular network, we assume fixed pressure drops (12 and 18 mm Hg for the baseline network and the simulated hyperinsulinemia network, respectively) between the inlet node (start of 1st order arteriole) and all outlet nodes (ends of terminal arterioles), and we also assume a physiological inflow hematocrit of 0.42 and that all arterioles and capillaries in the model are perfused with erythrocytes and plasma.

| Microbubble flow
Encapsulated microbubbles used in CEU imaging are nearly rigid spheres with diameters in the range of ~1-10 micrometers. 45 Since the ratio of microbubbles to RBCs is typically ~1:6000, 4 microbubbles travel as isolated particles and are expected to be more concentrated toward the center of microvessels than are RBCs. The more numerous deformable RBCs are more uniformly distributed across the lumen with a reduced concentration near the wall. Although the exact radial distribution of microbubbles is not known, it should depend on the ratio of microbubble diameter to vessel diameter. 46 However, since the biophysical properties of microbubble distribution in microvessels are not precisely known, we assume a simple model based on the data in Keller et al 28   The distribution of all data was evaluated using probability plots and Kolmogorov-Smirnov tests. The distribution of difference derived from the paired t test was evaluated using Q-Q plots. All data were normally distributed. Data are presented as mean ± standard error (SE) or percentage change (calculated as [(mean post-mean pre)/ mean pre] × 100). Mathematical modeling data are presented as generated by the model. Significance for all tests was set at P < .05.

| Characteristics of baseline and hyperinsulinemic euglycemic clamp
Blood pressure and core temperature remained unchanged throughout the experiment (  ). pH, partial pressure of carbon dioxide (pCO 2 ), and oxygen (pO 2 ) were measured at the end of the hyperinsulinemic euglycemic clamp, but were not measured (n.m.) during baseline to reduce the amount of blood collected. The hemoglobin oxygen saturation (sO 2 ) was calculated based on the measured pH, pCO 2, and pO 2 values. Rat insulin and human insulin were measured at baseline and hyperinsulinemic euglycemic clamp, respectively. *P < .05, compared to baseline. **P < .01, compared to 0. from 37.5 ± 1.0 to 36.0 ± 1.1% (P < .05, Table 1). At the end of the experiment, the systemic arterial pH, pCO 2 , pO 2, and sO 2 were within normal range (Table 1).
Arterial glucose concentration did not change throughout the experiment ( Figure 1A). At baseline, plasma concentration of rat insulin and C-peptide was 198 ± 15 ρmol/L and 346 ± 38 ρmol/L, respectively (Table 1). At the onset of the hyperinsulinemic euglycemic clamp, recombinant human insulin was infused and the concentration rose to 497 ± 29 ρmol/L (P < .01, Table 1) whereas C-peptide concentration fell by 51 ± 9% from baseline to the steady-state phase of the hyperinsulinemic euglycemic clamp (P < .01, Table 1). In the same space of time, GIR rose to 12.6 ± 0.4 mg/kg/min (P < .01, Figure 1B).

| Insulin increases capillary velocities and flow rates
Mean RBC velocity in the observed capillaries was 49 ± 12% greater during the steady-state portion of the hyperinsulinemic euglycemic clamp compared with baseline (P < .01, Figure 2A). In line with this, calculated plasma and blood velocity increased by 47 ± 12% and 48 ± 12%, respectively (P < .01, Figure 2B&C), whereas capillary hematocrit remained unchanged (P = .58, Figure 2D).

| Insulin does not increase functional capillary density
Although the mean RBC velocity increased from baseline to the steady-state portion of the hyperinsulinemic euglycemic clamp, there was no change in the total number of capillaries containing RBCs (P = .32, Figure 4A). To allow for a more refined analysis, we distinguished between capillaries with continuous, intermittent, and stopped RBC flow (Figure 4). and Suppinfo S1 442-nmVideos). The Supplemental Videos were recorded at 10X magnification at baseline and 20 minutes after initiating hyperinsulinemic euglycemic clamp protocol.

| Insulin makes RBC flow more homogenous
Comparisons of RBC flow using the 10X overlapping FOV revealed an increase in RBC flow from baseline to hyperinsulinemia that was not uniform ( Figure 6A). The capillaries with the lowest baseline flows had the largest increases in RBC flow during hyperinsulinemia, whereas RBC flow did not increase or in some instances decreased in the capillaries with the highest baseline flow rates (P < .01, Figure 6A).
Thus, insulin appeared to result in a redistribution of RBCs that increased the homogeneity of RBC flow in the capillary bed. However, plasma flow increased more uniformly and the increase in plasma flow as a response to hyperinsulinemia was not affected by baseline plasma flow (P = .24, Figure 6B). Insulin did not result in a redistribution of plasma flow.

| Modeling results
All arterioles have RBC flow under both simulated conditions ( Figure 7A,B). In contrast, at baseline, microbubble flow is zero in the majority of terminal arterioles and a large number of higher order arterioles as well ( Figure 7C). Although it is challenging to perceive the subtle redistribution of RBC flow in the simulated hyperinsulinemia network (comparing Figure 7A,B), the redistribution of microbubble flow in terminal arterioles is clear (comparing Figure 7C,D). In Figure 7D, only a few branches have zero

| D ISCUSS I ON
In this study, hyperinsulinemia at a constant glucose concentra- The data are presented as mean ± SE. **: P < .01, compared to baseline flow. There is no fundamental reason for assuming the increase in CEU signal must be due to de novo capillary recruitment.

| Redistribution of erythrocyte flow but not plasma flow
There are numerous publications reporting that insulin plays a significant role in regulating the distribution of microvascular blood flow to enhance the delivery of both insulin and glucose to skeletal muscle. 2,3,[5][6][7] Our experiments, using IVVM to quantify microvascular blood flow in the rat EDL muscle, confirm that an increased insulin concentration while infusing glucose to maintain a fixed blood glucose level does result in increased RBC velocity, RBC supply rate, and RBC flow (Figures 2A and 3F,A; but in contrast to previous reports using indirect methods, 2-7 we found no evidence of de novo capillary recruitment, either within individual capillary networks ( Figure 4) or among networks supplied by different terminal arterioles ( Figure 5). However, we did find a redistribution of RBC flow as shown in Figure 6A. The significant negative slope for the change in RBC flow from baseline vs hyperinsulinemia indicates a redistribution of RBC flow from high to low flow capillaries with hyperinsulinemia. In contrast, the slope for plasma flow in these same capillaries ( Figure 6B) was not significantly different from zero. Although hyperinsulinemia did increase plasma flow ( Figure 3B), we saw no evidence for a controlled redistribution of insulin and glucose as has been proposed. 47 Our observation that the distribution of plasma might be affected differently by hyperinsulinemia than erythrocyte distribution is not surprising. Plasma as a Newtonian fluid distributes proportionally to the downstream hemodynamic resistance.

| Computational model shows increased microbubble numbers with flow redistribution
Our computational model addressed the question of whether a redistribution of microbubbles within the microvasculature without de novo capillary recruitment could result in an increased CEU signal.
Experiments show that an increased flow rate such as we measured in capillaries does not increase the CEU ultrasound signal, whereas a redistribution of microbubbles to a larger number of flow paths does. 50 The model shows that a few terminal arterioles corresponding to the paths with the highest flow rate have hematocrits approaching systemic hematocrit ( Figure 8C and Table 1

| Evidence that EDL preparation reflects normal physiology
Proponents of capillary recruitment have criticized IVVM muscle preparations as not reflecting normal physiological function. They argue that surgery to expose these thin muscles likely result in hyperemia thus masking the capillary recruitment which one would observe using non-invasive techniques such as CEU on thicker muscles which play a more important role in physical activity. 8,9 To address these potentials criticisms, we used the rat EDL muscle, a bellied muscle with mixed fiber type, which is involved in locomotion by controlling the movement of the hind paw. Tyml 19  with resting RBC velocities for up to 5-6 hours 13,33,51,52 similar to those reported by Tyml. 19 Two independent research groups have also shown that exteriorization does not alter muscle flow. 25,26 Further evidence that our EDL preparation reflects the normal physiological response comes from the 80% increase in RBC supply rate ( Figure 3F) and corresponding increase in RBC O 2 saturation from 50 ± 3 to 67 ± 11% ( Figure 3E)   De-recruitment of perfused capillaries also occurs if the IVVM preparation is exposed to elevated O 2 levels 57 or to pathological conditions such as sepsis. 58 The baseline RBC O 2 saturation of 50 ± 3% and RBC supply rate of 6 ± 1 RBCs s −1 were lower than we have previously reported 13  Overall, this quantitative microvascular data provide evidence that the flow response to hyperinsulinemia measured in the rat EDL muscle reflects a normal physiological response to insulin.

PER S PEC TIVE
Hyperinsulinaemia under euglycaemic conditions results in an increase in mean blood flow and a more homogeneous distribution This study highlights that IVVM is a powerful tool for studying how the microvasculature redistributes blood flow in response to a stimulus, change in tissue demand, or potentially how it may fail in acute or chronic disease.

CO N FLI C T O F I NTE R E S T
We have no competing interests to declare. (1) ∑ i D,i Q i = 0