Mouse Models and Techniques for the Isolation of the Diabetic Endothelium

Understanding the molecular mechanisms underlying diabetic endothelial dysfunction is necessary in order to improve the cardiovascular health of diabetic patients. Previously, we described an in vivo, murine model of insulin resistance induced by feeding a high-fat diet (HFD) whereby the endothelium may be isolated by fluorescence-activated cell sorting (FACS) based on Tie2-GFP expression and cell-surface staining. Here, we apply this model to two new strains of mice, ScN/Tie2-GFP and ApoE(−/−)/Tie2-GFP, and describe their metabolic responses and endothelial isolation. ScN/Tie2-GFP mice, which lack a functional toll-like receptor 4 (TLR4), display lower fasting glucose and insulin levels and improved glucose tolerance compared to Tie2-GFP mice, suggesting that TLR4 deficiency decreases susceptibility to the development of insulin resistance. ApoE(−/−)/Tie2-GFP mice display elevated glucose and cholesterol levels versus Tie2-GFP mice. Endothelial isolation by FACS achieves a pure population of endothelial cells that retain GFP fluorescence and endothelial functions. Transcriptional analysis of the aortic and muscle endothelium isolated from ApoE(−/−)/Tie2-GFP mice reveals a reduced endothelial response to HFD compared to Tie2-GFP mice, perhaps resulting from preexisting endothelial dysfunction in the hypercholesterolemic state. These mouse models and endothelial isolation techniques are valuable for assessing diabetic endothelial dysfunction and vascular responses in vivo.


Introduction
Vascular complications are the main causes of morbidity and mortality associated with diabetes. Diabetics have a 2-4fold increased risk for developing cardiovascular disease [1]. Progressive degeneration of microvascular beds is a major contributing factor to many complications of diabetes including retinopathy, nephropathy, and neuropathy. Macrovascular complications associated with diabetes include coronary heart disease, stroke, and peripheral vascular disease. The relationship between diabetes and vascular complications emphasizes the importance of understanding the mechanisms underlying this association.
The increased risk of cardiovascular disease observed in diabetes is primarily due to a damaged or dysfunctional endothelium. In the diabetic state, numerous inflammatory, hormonal, and metabolic influences impinge upon the endothelium and alter its function. Elevated glucose and fatty acids lead to endothelial nitric oxide synthase (eNOS) uncoupling [2]. This in turn leads to reduced nitric oxide bioavailability and generation of reactive nitrogen and oxygen species as well as increased glycation of proteins and lipids [3][4][5]. Pathological levels of hormones and cytokines in the diabetic state alter endothelial signaling pathways promoting inflammation and atherogenesis and impairing vasoregulation and vascular hemostasis [6,7]. Endothelial dysfunction precedes symptomatic vascular disease and is often detected before structural changes in the vascular wall [2]. A deeper understanding of the mechanisms underlying the endothelial response to diabetes should lead to new therapeutic interventions to inhibit these changes before they progress to vascular disease. It may also identify useful biomarkers of diabetic vasculopathy.

ISRN Endocrinology
Endothelial isolation has been difficult, in part due to the lack of specific markers available for the endothelium. Column purification techniques often lead to contamination by other cell types. A freeze-fracture technique [8] and scraping of the endothelium off the vessel wall [9] usually do not achieve pure populations of endothelial cells and may cause damage leading to RNA degradation. These techniques may be acceptable if further selection is possible, such as altering cell culture conditions to eliminate contaminating cell types. However, in vitro responses of endothelial cells are highly dependent on the choice of growth medium and flow conditions [10]. In vitro studies are further complicated by the altered expression of endothelial cell markers and function commonly observed in both primary isolates and immortalized cultures [11]. Most importantly, monoculture of endothelial cells is a terrifically artificial environment lacking the hemodynamic stimuli, intercellular interactions, and hormonal influences of the intact, perfused, in vivo blood vessel. Therefore, assessment of the endothelial response to diabetes is optimally obtained with an in vivo murine model of diabetes accompanied by rapid isolation of pure populations of endothelial cells.
We have previously described a high-fat diet-induced model of insulin resistance using Tie2-GFP mice wherein the endothelium can be reliably isolated by fluorescence-activated cell sorting based on Tie2-driven GFP expression and cell-surface staining for endothelial markers [12]. Here we have applied this high-fat model to two new strains of mice, ScN/Tie2-GFP and ApoE(−/−)/Tie2-GFP, created by interbreeding with Tie2-GFP mice. ScN/Tie2-GFP lack a functional toll-like receptor 4 (TLR4), which serves an important role in innate immunity and is the principal sensor of LPS [13]. TLR4 has also been implicated in promoting vascular insulin resistance by binding free fatty acids followed by subsequent activation of inflammatory pathways [14]. ApoE(−/−) mice lack the lipid carrier, apolipoprotein E, and consequently display hypercholesterolemia and increased susceptibility to the development of atherosclerosis. The crossing of TLR4-and APOE-deficient mice with Tie2-GFP mice allows the isolation of their endothelium in order to elucidate the molecular biology underlying these observed vascular phenotypes. Here we characterize the metabolic profiles of both strains after feeding a high-fat diet and compare their endocrine responses to that of Tie2-GFP mice. We also provide a detailed description of our endothelial isolation technique and other important considerations when studying the diabetic vasculature.

Measurement of Metabolic
Characteristics. After 2, 4, 6, and 8 weeks on their respective diets, glucose levels were determined by glucometry of the tail blood following an overnight fast (OneTouch Ultra, Lifescan, Milpitas, CA, USA). A glucose tolerance test (GTT) was performed after 6 weeks on the diet regimen. Glucose (1 mg/g of body weight) was administered i.p. following an overnight fast. Glucometry of the tail blood was performed prior to glucose injection and every 20 minutes afterwards for 2 hrs. The area under the curve (AUC) was determined using the statistical software in Graphpad. After 8 weeks of HFD, insulin levels were measured in serum from overnight-fasted mice using a Mercodia Mouse Insulin ELISA kit (Uppsala, Sweden, cat. number 10-1149-01). Total cholesterol in the sera of overnightfasted HFD-fed and chow-fed ApoE(−/−)/Tie2-GFP mice and both parental strains were measured by ELISA performed according to manufacturer's instructions (Cayman Chemical, Ann Arbor, MI, USA, cat. number 10007640).

Endothelial
Isolation. Animals were sacrificed by CO 2 asphyxiation and perfused with ice cold PBS. Aortae were dissected from the aortic root to the iliac bifurcation. Leg muscles consisting of the plantaris, gastrocnemius, and biceps femoris (which are readily dissected as a single group) were excised. Tissues were placed in cold PBS, and adherent fat was removed. The aortic and skeletal muscle tissues from 3 animals were each pooled, minced into 1 mm fragments, and dispersed in PBS with 100 mM CaCl 2 and MgCl 2 containing 5 mg/mL collagenase type 2 (Worthington Biochemical, Lakewood, NJ, cat. number LS004174), 2 mg/mL glucose, and 30 U/mL DNAse I (Worthington Biochemical, cat. number LS006331) under constant agitation (180 rpm at 37 ∘ C). Aortae were collagenolytically digested for 40 min, while the muscle was digested for 1 h, with mechanical dissociation obtained by triturating every 10 min. Enzyme activity was stopped by the addition of a 10% solution of FBS in PBS. Cells were pelleted by centrifugation (5 min, 1000 rpm, 4 ∘ C), aspirated, resuspended in PBS, and filtered through a 40 m cell strainer (BD Biosciences, San Jose, CA, USA, cat. number 352340). Red blood cells were lysed by resuspension in 1 mL RBC Lysis Solution (Applichem, St. Louis, MO, USA, cat. number A4617) overlaid on 4.5 mL FBS followed by centrifugation. In each experiment, pooled cells from 3 experimental and 3 control mice were collected for each tissue.

Matrigel Tube Formation and Acetylated-Low Density
Lipoprotein (ac-LDL) Uptake Assays. GFP + endothelial cells from the skeletal muscle were isolated by FACS and sorted into tissue culture medium for fluorescence microscopy and functional assays to confirm their endothelial origin. Aliquots containing 1000 cells were deposited onto lysine-coated slides by centrifugation at 450 rpm for 5 minutes with a Cytospin (Shandon). Cells were fixed in 10% formalin for 30 min and mounted in DAPI-containing medium. Cells were viewed under the fluorescent microscope and images were collected with an Axiophot system (Zeiss, Oberkochen, Germany).
For tube formation assays, approximately 8,000 GFP + , sorted endothelial cells were plated per well in 48-well plates coated with Matrigel Basement Membrane Matrix (BD Biosciences, cat. number 354234) and allowed to grow in endothelial growth medium. After 5 days, tube formation was assessed. For ac-LDL uptake assays, the sorted endothelial cells were plated on Nunc Lab-Tek 8-well chamber slides (Thermo Scientific, Billerica, MA, USA, cat. number 154941) and incubated with 50 g/mL Dil-Ac-LDL (Biomedical Technologies Inc., Stoughton, MA, USA, cat. number BT-902) for 4 hours, washed, fixed in 2% paraformaldehyde, stained with DAPI, and imaged using the Axiophot system (Zeiss).

Flow Cytometry Analysis of Endothelial Populations
Stained for Monocyte Markers. Collagenolytically separated cell suspensions derived from aortic and muscle tissues as previously described were stained for 25 min at 4 ∘ C with Alexa Fluor Ⓡ 647 anti-mouse CD11b (eBiosciences, cat. number 51-0112), and flow cytometry was performed on a FACSAria. The sorting gates were established by initially gating for mononuclear cells based on forward scatter. This was followed by gating for GFP fluorescence, which was considered to be any signal in the FITC channel above that of an unstained, non-GFP sample. After applying the sorting gate for the GFP + endothelial cells, the number of GFP + /CD11b + events within that gate was analyzed using FlowJO software.

Microarray Analysis.
Microarray analyses were performed to determine the transcriptional responses in aortic and skeletal muscle endothelium from ApoE(−/−)/GFP mice exposed to high-fat diet for 4 weeks compared to ApoE(−/−)/GFP mice fed a chow diet. 10,000 endothelial cells positive for both GFP and phycoerythrin-CD31 staining were isolated with a FACSAria directly into TRIzol. RNA was purified with RNeasy columns (Qiagen, Valencia, CA, USA) to yield <80 ng, which was then amplified using an Ambion Amino Allyl MessageAmp kit (Life Technologies, Carlsbad, CA, USA, cat. number AM1753) according to the manufacturer's protocol to produce approximately 100 g of aminoallyl modified cRNA. This was labeled with Cy3 and Cy5 CyDye Post-Labeling Reactive Dye Pack (GE Healthcare, Waukesha, WI, USA) according to the manufacturer's instructions. Following purification, 200 pmoles of Cy3 and Cy5 dye-labeled cRNA, as measured by NanoDrop 2000c spectrophotometer (Thermo Scientific), were combined and fragmented with Ambion 10X fragmentation reagents at 70 ∘ C for 15 min. Yeast tRNA (4 g), polyA RNA (4 g), mouse cot-1 DNA (1 g), and Slidehyb III hybridization buffer (Ambion) were added for a total sample volume of 35 L. Triplicate experiments were performed for both tissues. For each of the biological replicate experiments, two array hybridizations were performed, each of which included a dye reversal. Samples were hybridized overnight to glass slides spotted with the Operon Murine V4 oligo set produced by the Duke University Microarray Core. Arrays were scanned using a Genepix 4000B system (Molecular Devices, Union City, CA, USA) and analyzed with Genepix and Acuity software.
Statistical analysis of microarray results was performed using the functions available within Acuity. Results were normalized using the ratio of medians method and filtered to exclude any exhibiting the following characteristics: a percentage of saturated pixels >3, a signal/noise ratio <3, a (regression ratio 635/532) 2 <0.6, or a Genepix flag. Transcripts exhibiting upregulation (a log 2 (fold change) >0.75) or downregulation (log 2 (fold change) <−0.75) were tabulated

Murine Model of Type II Diabetes.
In order to provide a more authentic model of the pathophysiological response observed in the human population, we have chosen to use a nutritional model of type II diabetes induced by feeding a high-fat diet consisting of 60% fat calories. The compositions of the high-fat and chow diets are compared in Table 1.
The high-fat diet contains nearly 60% of its calories from fat, while the normal chow diet derives 60% of its calories from carbohydrates and only 12% of its calories from fat ( Table 1). The total caloric intake per gram is much greater in the HFD compared to the chow diet (Table 1). Therefore, mice consuming equal quantities of the diet receive more calories from the HFD than the chow diet, and this is further exacerbated by the hyperphagia induced by the HFD [15]. Studies have also shown that calories derived from fat are more diabetogenic than the same caloric intake from carbohydrates [16]. In humans, dietary fat intake has been linked to obesity, decreased insulin sensitivity, and progression to type II diabetes [17][18][19]. High-fat diets also raise LDL cholesterol and increase LDL particle size compared to low-fat diets in human subjects [20]. This diabetic model allows us to evaluate the endothelial response to the diabetic milieu as a whole without selecting for particular biochemical attributes of diabetes. However, this also means that individual responses of the endothelium cannot be definitively ascribed to one aspect of diabetes or another, such as those changes due to insulin resistance versus hyperglycemia versus weight gain, and so forth. However, we can try to assign causation based on our current understanding of endothelial cell biology and the effects of diabetes in other cell types. The use of knockout models, however, does allow us to determine differential vascular responses due to a specific gene deletion. Analysis of ScN/GFP and ApoE(−/−)/GFP endothelial responses to this high-fat diet provides insight into the effect of TLR4 and APOE deficiency.

Metabolic Characteristics of ScN/GFP Mice on a High-Fat Diet
Regimen. As we have previously described, Tie2-GFP mice have accelerated weight gain and hyperglycemia by 2 weeks, marked hyperinsulinemia and impaired glucose tolerance by 6 weeks, and aortic vascular insulin resistance observed after 5 weeks of HF feeding [12]. ScN/GFP mice on chow diet have lower fasting glucose levels compared to Tie2-GFP mice ( Table 2). Over the time course of the feeding, the weights and fasting glucose levels of ScN/GFP mice increase compared to chow-fed controls ( Table 2). Although elevated compared to chow-fed controls, the fasting glucose levels of ScN/GFP mice remain significantly lower than that of Tie2-GFP mice even after 8 weeks of HFD (Table 2). Insulin levels of ScN/GFP mice are also much lower compared to Tie2-GFP mice and are not different from chow-fed controls ( Figure  1(a)). Glucose tolerance testing of ScN/GFP mice after 6 weeks of HFD reveals a normal response to glucose challenge compared to Tie2-GFP HFD mice where glucose tolerance is impaired (Figures 1(b) and 1(c)).
Obese and type II diabetic patients display increased expression of TLR4 in their skeletal muscle, implicating its role in diabetic pathophysiology [21]. Following 8 weeks of high-fat diet, whole aorta lysates from C57BL/6 mice have been shown to display decreased phosphorylation of protein kinase B (AKT) and endothelial nitric oxide synthase (eNOS) after insulin stimulation, suggesting a state of vascular insulin resistance and impaired vasoregulation, which were not observed in TLR4(−/−) mice [14]. Our studies also suggest that TLR4 serves a role in the development of insulin resistance as its deficiency in ScN/GFP mice lowers glucose and insulin levels and improves glucose tolerance. The ability to isolate the endothelium from these mice allows molecular studies to elucidate the mechanisms underlying the role of TLR4 in vascular insulin resistance and endothelial dysfunction.

Metabolic Characteristics of ApoE(−/−)/GFP Mice on a
High-Fat Diet Regimen. ApoE(−/−)/GFP mice also display increased weight gain after HFD feeding (Table 2). Unlike ScN/GFP mice, the fasting glucose levels of ApoE(−/−)/GFP mice are elevated compared to Tie2-GFP mice ( Table 2). Over the time course of the feeding, the fasting glucose levels of ApoE(−/−)/GFP mice on HFD increase slightly compared to chow-fed controls. ApoE(−/−)/BL/6 mice are hypercholesterolemic due to defective lipoprotein metabolism. We therefore assessed the total cholesterol levels in the sera of ApoE(−/−)/GFP mice and found that they exhibit similar cholesterol levels to ApoE(−/−) mice on the C57BL/6 background. After 5 months of HF feeding, these levels doubled (Figure 2).
endothelial populations mostly derived from microvessels. To date, there are no acceptable antibodies to specifically isolate these endothelial subtypes although the differential expression of a few genes has been identified [22]. The heterogeneity of endothelial populations among the different vascular beds suggests that the responses of any one vascular bed may be not be reflective of those of the entire circulatory system. We have previously performed gene expression analyses on the endothelium derived from two vascular beds of Tie2-GFP mice and have identified both the specific and common responses of the large vessel endothelium and the microvasculature to the diabetic state [12]. The transcriptional responses of the macrovascular and microvascular endothelial cells to diabetes differ in both the degree and onset of dysregulation. While some transcripts are commonly regulated among the 2 vascular beds, others are specific to one vascular bed. Therefore, the choice of endothelial population is an important consideration when studying endothelial responses and vascular complications of diabetes.

Endothelial Isolation by Fluorescence-Activated Cell Sorting.
In Tie2-GFP mice, GFP expression is driven by the endothelial promoter for the angiopoietin receptor, Tie2, which is important in angiogenesis and vasculogenesis [23]. Therefore, all endothelial cells of Tie2-GFP mice fluoresce at a peak wavelength of 509 nm when excited at a wavelength of 488 nm. This allows for endothelial cell isolation by fluorescence-activated cell sorting (FACS).
Fat in tissue digests is highly autofluorescent and is particularly abundant in diabetic models. The identification of GFP + endothelial cells is thus enhanced by double staining for CD31/Platelet endothelial cell adhesion molecule 1 (PECAM-1), which does not label fat cells. CD31 is highly expressed by endothelial cells and has long been used as a marker for endothelial cell isolation [24]. However, it is also expressed by leukocytes and therefore, may not be used alone. As shown in Figure 3, all GFP + cells also exhibit phycoerythrin-CD31 staining, whereas an additional population exhibits CD31 staining without GFP fluorescence. GFP + /CD31 + endothelial cell populations represent approximately 1-2% of the total population of cells derived from muscle and aortic tissue (Figure 3). Approximately 3,500 endothelial cells may be isolated from one mouse aorta and up to 30,000 cells from the skeletal muscle prep described in this study. We routinely collect at least 10,000 GFP + /CD31 + cells from each pooled tissue sample for transcriptional analysis.

Confirmation of Endothelial Identity and Assessment of
Monocyte Contamination. The endothelial identity of the sorted cells is confirmed by fluorescence microscopy. After FACS, we see only GFP + endothelial cells and no other contaminating, non-GFP cells (Figure 4(a)). Endothelial identity is further confirmed by functional assays. GFP + cells sorted from the skeletal muscle retain the unique endothelial ability to form tubes when plated on Matrigel (Figure 4(b)). In addition, the sorted cells metabolize acetylated-LDL as demonstrated by Dil fluorescence, a property that is specific to endothelial cells and macrophages (Figure 4(c)).
Tie2-expressing myeloid cells have been reported to account for 2-7% of human blood mononuclear cells [25], which also express CD11b. We analyzed the number of CD11b + events within our sorting gates by flow cytometry of muscle tissue suspensions from 3 independent experiments following 8 weeks of diet. In Tie2-GFP mice, FACS revealed that only 0.8 ± 0.4% of GFP + cells derived from HFD animals also display the monocyte marker CD11b with similar low percentages observed in control animals (0.3 ± 0.5%). Similarly, analysis of GFP + cells derived from ScN/GFP mice reveals few GFP + cells staining for CD11b in pooled samples from either control or HFD mice ( Figure 5).
In diabetes, there exists a state of inflammation characterized by an increase in activated monocytes in the peripheral blood of diabetic patients [26]. Analysis of the CD11b + / CD115 + inflammatory monocyte population in peripheral blood reveals a nonsignificant increase from 9±2% of CD45 + leukocytes in chow-fed mice to 11 ± 3% in Tie2-GFP mice fed HFD for 8 weeks (Table 3). We therefore investigated whether the number of monocytes appearing to have a positive signal for GFP was altered by HFD. As shown in Table 3, chow and HFD mice display similar, low levels of GFP + cells in the peripheral blood. Importantly, this "GFP + " signal is much lower than that obtained from the endothelial population and would therefore be excluded by gating.  Table 3: Flow cytometry analysis of Tie2-GFP expression in peripheral blood mononuclear cells (PBMCs) derived from Tie2-GFP mice fed a high-fat or chow diet for 8 weeks.
in the skeletal muscle are presented in Table 4(b). Similar transcriptional analyses may be performed on ScN/GFP mice to elucidate the role of TLR4 in vascular insulin resistance and endothelial dysfunction upon high-fat feeding.

Potential
Limitations. The use of the Tie2-GFP transgene, which is expressed at moderately high levels, may potentially affect endothelial function, either by nonspecific effects of a novel protein or, perhaps, by specific squelching effects of the Tie2 promoter on expression of the endogenous Tie2 gene. We have not seen any gross effects upon the phenotyping described in this study; however, we cannot exclude subtle effects. It is thus ultimately useful to confirm results from the Tie2-GFP studies in nontransgenic models. For example, transcriptional regulation identified in FACS-sorted cells can be confirmed by immunohistology or in situ hybridization on nontransgenic vascular samples.

Conclusions
We have demonstrated the feasibility of isolating vascular endothelial cells from murine models of diabetes in order to study responses of the diabetic vasculature. We have shown that the endothelium may be isolated from these In this study, we describe the metabolic responses and endothelial isolation of ScN/GFP and ApoE(−/−)/GFP mice. ScN/GFP mice display decreased susceptibility to the development of insulin resistance by HFD as demonstrated by lower fasting glucose and insulin levels and improved glucose tolerance compared to Tie2-GFP mice. TLR4 has been shown to play a role in vascular insulin resistance due to its ability to bind free fatty acids and activate inflammatory pathways [14]. Therefore, this model of high-fat feeding and endothelial isolation may be useful to explore the metabolic and vascular changes due to TLR4 deficiency. ApoE(−/−)/GFP mice display elevated glucose and cholesterol levels compared to Tie2-GFP mice. Transcriptional analyses of the endothelium isolated from the aorta and skeletal muscle of these mice demonstrate reduced endothelial response to HFD, perhaps resulting from endothelial dysfunction in the preexisting hypercholesterolemic state in chow-fed ApoE(−/−)/GFP mice. Similar studies to evaluate the effect of gene loss on the diabetic endothelium and vascular responses may be performed by utilizing the techniques described in this study.