Levels of atherogenic lipoproteins are unexpectedly reduced in interstitial fluid from type 2 diabetes patients.

At a given level of serum cholesterol, patients with T2D have an increased risk of developing atherosclerosis compared with nondiabetic subjects. We hypothesized that T2D patients have an increased interstitial fluid (IF)-to-serum gradient ratio for LDL, due to leakage over the vascular wall. Therefore, lipoprotein profiles in serum and IF from 35 T2D patients and 35 healthy controls were assayed using fast performance liquid chromatography. The IF-to-serum gradients for VLDL and LDL cholesterol, as well as for apoB, were clearly reduced in T2D patients compared with healthy controls. No such differences were observed for HDL cholesterol. Contrary to our hypothesis, the atherogenic VLDL and LDL particles were not increased in IF from diabetic patients. Instead, they were relatively sparser than in healthy controls. The most probable explanation to our unexpected finding is that these lipoproteins are more susceptible to retainment in the extravascular space of these patients, reflecting a more active uptake by, or adhesion to, tissue cells, including macrophages in the vascular wall. Further studies are warranted to further characterize the mechanisms underlying these observations, which may be highly relevant for the understanding of why the propensity to develop atherosclerosis is increased in T2D.

phenomenon, including the frequent presence of hyperglycemia, dyslipidemia, and infl ammation in this condition ( 1 ). Although such patients often display elevated serum TGs, reduced HDL, and increased small dense LDL particles, the level of serum LDL cholesterol is generally not markedly elevated. However, LDL cholesterol levels represent a strong predictor of coronary heart disease and stroke in T2D (2)(3)(4), and lowering of LDL by pharmacological treatment clearly prevents cardiovascular disease in this condition (5)(6)(7). At a given LDL cholesterol level, patients with T2D have an increased risk for atherosclerosis compared with normal ( 3,4,8 ), a fi nding that is still poorly understood. Further mechanistic knowledge of this phenomenon should be of great importance for our possibilities to improve prevention of vascular disease, particularly facing the global acceleration of the prevalence of T2D.
Atherosclerosis is thought to be initiated after LDL particles have been translocated from plasma into the subendothelial space of the vascular wall, where they are modifi ed and initiate an infl ammatory response ( 9 ). In this process, modifi ed LDL particles are engulfed by resident macrophages that when overloaded transform into foam cells. One explanation for why the relative risk of atherosclerosis is increased at a given concentration of LDL in T2D patients compared with nondiabetic subjects may be that there is an increased leakage of lipoproteins over the vascular wall due to endothelial dysfunction in T2D ( 1 ). This would enhance the exposure of structures in the subendothelial space, including macrophages, to angioplasty were excluded. Patients on hypolipidemic and/or aspirin therapies prior to the trial had these withdrawn 4 and 2 weeks before the fi rst treatment period, respectively. They were randomized in a double-blind, crossover fashion to receive either aspirin 320 mg with placebo or aspirin 320 mg with atorvastatin (Lipitor®, Pfi zer) 80 mg, once daily. The treatments were given for 4 weeks each, separated by a 4-week washout period when only aspirin were administered. In this pilot study, IF and serum were collected in the fasting state in the same way as in the main diabetes study (see below). The study was approved by the ethics committee at Karolinska Institute, Stockholm, and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained at inclusion.

T2D patients and controls
Altogether, 35 patients with T2D, and 35 age-and gendermatched healthy controls were studied; their clinical characteristics are given in Table 1 . The study was designed to allow the detection of a 25% increase in IF-to-serum gradient for LDL cholesterol in the patients, with a 95% power at a signifi cance level of 0.05. The patients were recruited from the outpatient clinic of the Department of Endocrinology, Metabolism and Diabetes, Karolinska University Hospital Huddinge, and healthy controls by local advertisements. Laboratory assessment included routine fasting analyses of hemoglobin, glucose, HbA1c, insulin, aspartate aminotransferase, alanine aminotransferase, cystatin C, TGs, and total, HDL, and LDL cholesterol. Blood pressure was measured, and all subjects fi lled in a form to document their health status and medical history. Control subjects with clinical or laboratory evidence of reduced kidney function, infl ammatory disease, vascular disease, thyroid disease, high blood pressure, or treatment for any of those conditions were excluded. For the patients with T2D, exclusion criteria were infl ammatory disease and/or oral glucocorticoid treatment. Written informed consent was obtained at inclusion; the study was conducted in accordance with the Declaration of Helsinki and approved by the regional ethics review board in Stockholm.
As expected, diabetic patients had higher body mass index than the controls ( Table 1 ). The average duration of disease was 14 years. The systolic blood pressure was >140 mmHg in 20 of the patients, and cystatin C levels were increased (>1.25 mg/l) in 6 of them. All 35 patients were on long-term antidiabetic treatment (metformin, 23; sulfonylurea, 7; rosiglitazone, 2; pioglitazone, 1; dipeptidyl peptidase inhibitor, 1; glucagon-like peptide 1 (GLP-1) receptor agonists, 3; and/or insulin, 22). Twenty-fi ve of the patients were on lipid lowering therapy (statins, 24; fi brate, 1; ezetimibe, 2). atherogenic lipoproteins derived from the circulation. As the permeability of capillaries in skin and other tissues is increased in T2D ( 10 ), and because the transfer of macromolecules has been reported to be the same across capillaries and aorta ( 11 ), it is reasonable to assume that the permeability of the arteries to lipoprotein particles is likely increased in T2D and that this can be monitored in peripheral interstitial fl uid (IF). In accordance with that concept, it has been reported that the initial fractional escape rate of intravenously injected radiolabeled autologous LDL is increased by >30% in patients with T2D compared with healthy controls ( 12 ). If this is the case, such patients would be predicted to have a higher IF-to-serum concentration gradient of LDL compared with healthy controls. Due to the technical diffi culties to adequately access the IF compartment ( 13,14 ), studies in the metabolism of apoBcontaining lipoproteins in IF have been very limited so far, however, and we have not been able to fi nd any published information on the situation in T2D.
In this work, we have tested the hypothesis that the transvascular IF-to-serum gradient of lipoproteins, in particular LDL, is increased in T2D patients compared with healthy controls.

Pilot study of the effect of statin treatment on lipoprotein IF-to-serum gradient
It was early realized that it would not be practically or ethically feasible to study a statin-free population of T2D patients. Thus, we fi rst performed a pilot study on nondiabetic individuals to explore whether such therapy had any infl uence on lipoprotein gradient measurements. For this purpose, 19 patients (14 women and 5 men), aged 65 ± 26 years, with peripheral vascular disease and hypercholesterolemia (cholesterol у 6.5 mM and/or LDL cholesterol у 4.16 mM) were studied. They were treated at the Department of Surgery, Karolinska University Hospital Huddinge, because of intermittent claudication verifi ed by duplex ultrasound or angiography. Patients with angina pectoris, previous acute myocardial infarction, ischemic stroke, or previous coronary artery bypass grafting or percutaneous transluminal coronary Ϫ / Ϫ 25/24 Suffered from one or more myocardial infarctions (n) -10 Pathological cystatin C, >1.25 mg/l (n) -6 Pathological HbA1c, >46 mmol/mol (n) -28 HbA1c, glycosylated hemoglobin; hsCRP, high-sensitivity C-reactive protein. Data are means ± SD. P values indicate the signifi cance of difference between the study groups (Student's unpaired t -test). One value for HbA1c and one for duration of diabetes are missing from the patients.

RESULTS
In the pilot study, we evaluated the effect of statin treatment on the IF-to-serum gradients for cholesterol and TGs in different lipoprotein particles, using a crossover design. As expected, statin treatment reduced serum levels of VLDL and LDL cholesterol by 60% and 50%, respectively ( Table 2 ). In IF, the reductions were 60% and 54%, respectively. Likewise, serum levels of VLDL and LDL TGs were reduced by 45% and 12%, respectively, while their IF levels were reduced by 54% and 28%. Consequently, statin treatment did not change the IF-to-serum gradients for cholesterol or TGs, indicating that the lipoprotein permeability of the vascular wall was unaltered during statin treatment.
On average, serum total cholesterol was 28% lower in diabetic patients, while LDL and HDL cholesterol were 30% and 40% lower, respectively ( Table 3 ). A similar pattern, but with more pronounced reductions, was seen in IF, where total and LDL cholesterol was 40% and 44% lower, respectively, in T2D. HDL cholesterol was 36% lower than in healthy controls. Interestingly, and in opposition to the prediction following our original hypothesis, the IF-to-serum gradients for cholesterol were 29% lower for VLDL and 18% lower for LDL in T2D patients compared with controls ( Fig. 1A ). However, the corresponding gradient for HDL did not differ between patients and controls. The diabetic patients on statin treatment had lower serum total and LDL cholesterol levels than those without such treatment, while the groups did not differ in IF-to-serum gradients (supplementary Table 1).
As anticipated, serum total and VLDL TGs were significantly higher (72% and 130%, respectively) in patients than in healthy controls, while total and VLDL TG levels in

Collection of IF and serum
All subjects arrived to the clinic in the morning after overnight fast. Height, weight, and bioimpedance were assessed, and blood pressure was measured in the supine position. Serum hsCRP and cystatin C levels, together with urine protein were determined by routine clinical assays in all subjects. Additional measurements corresponding to the initial screening program were performed in the patients. Suction-blister fl uid was generated according to Kiistala ( 15 ). Briefl y, two plastic chambers (Ventipress Oy, Lappeenranta, Finland) were placed on the skin 5 cm bilaterally to the umbilicus. A mild suction-pressure of 28-32 kPa was generated by a pump (Ventipress Oy) and sustained until blisters were formed (1.5-2 h). Five blisters were formed under each chamber from where IF (250-500 µl) was aspirated with a syringe and stored at Ϫ 80°C until analysis. Venous blood was collected after 1 h of blister generation, and stored as whole blood, serum, and plasma at Ϫ 80°C.

Fast performance liquid chromatography analyses
Lipoprotein cholesterol, TGs, and free cholesterol profi les were analyzed by fast performance liquid chromatography (FPLC), as described ( 16 ). Briefl y, serum (1 µl for cholesterol and free cholesterol, and 2 µl for TGs) and IF (5 µl for cholesterol and free cholesterol, and 20 µl for TGs) lipoproteins were separated on a sepharose column. Reagents (Roche Diagnostics, Mannheim, Germany; and WAKO Diagnostics, Richmond, VA) were continuously added to the eluate online, and absorbance was measured. In the statin study, the volumes added were 2.5 µl serum and 6.7 µl IF. The lipid content in lipoprotein fractions was calculated from the areas under the curves in the elution profi les.

Statistical analysis
Values are presented as means ± SD and were log transformed prior to statistical analysis. Statistical analyses were performed found between the subject groups in either serum or IF levels ( Table 3 ), or in the IF-to-serum gradients ( Fig. 1D ).
The LDL cholesterol/apoB ratio was calculated as an indicator of LDL particle size. The lower ratio in serum in T2D than in healthy controls indicates that their particles are smaller ( Table 3 ). In contrast, the LDL cholesterol/ apoB ratio in IF was larger in the T2D patients than in healthy controls, indicating larger particles in T2D in IF ( Table 3 ). When calculating the HDL cholesterol/apoAI ratio as an indicator of HDL size, the ratio was lower in T2D serum than in healthy controls, indicating smaller HDL in serum in T2D patients ( Table 3 ). There was no difference in HDL cholesterol/apoAI ratio in IF between T2D patients and healthy controls, indicating similar HDL size ( Table 3 ).

DISCUSSION
T2D is a common cause of premature vascular morbidity and death, and there seems to be an inherent increased risk of cardiovascular disease in relation to the presence of established risk factors ( 3 ). Thus, although the risk is IF did not differ between the study groups ( Table 3 ). The level of LDL TGs in IF was 23% lower in T2D patients than in controls, while HDL TGs in IF were 19% lower. As for cholesterol, the IF-to-serum gradients for VLDL and LDL TGs were lower in T2D patients compared with controls, being reduced by 43% and 24%, respectively ( Fig. 1B ).
Total free cholesterol, and free cholesterol in LDL and HDL were signifi cantly reduced in samples from diabetic patients, in both serum (28,29, and 44%) and IF (38, 44, and 33%) ( Table 3 ). Furthermore, the IF-to-serum gradients for free cholesterol in VLDL and LDL were reduced by 21% and 20%, respectively, in the patients ( Fig. 1C ), similar to the IF-to-serum gradients for cholesterol.
The level of apoAI was 19% lower in serum from diabetics compared with controls, and the corresponding level in IF was 34% lower ( Table 3 ). The level of apoB in serum did not differ between patients and controls, while the IF level was 56% lower in the patients. In T2D patients, the IF-to-serum gradient for apoAI was reduced by 19% compared with controls, and the gradient ratio for apoB was reduced by 60%, again indicating a different transcapillary gradient in T2D ( Fig. 1D ). For Lp(a), no differences were Data are means ± SD. P values indicate the signifi cance of difference from the same fl uid in control group (Student's unpaired t -test). In serum, values for free cholesterol are missing from three healthy controls, and in IF from one healthy control. One value for serum apoAI in healthy controls is excluded because of being an outlier; consequently, the HDL chol/apoAI ratio is missing. One value each for IF apoB and Lp(a) in healthy controls is missing; consequently, one value for LDLchol/apoB ratio is missing. be the critical location where to fully explore our hypothesis. We used a technique that generates IF in blisters formed by mild suction ( 15 ). Such blister fl uid has been shown to represent IF ( 17 ), but it is clear that our fi ndings have to be taken with caution. There are few reports on lipoprotein levels in IF ( 13 ), and the IF-to-serum gradients seen by us are in accordance with those described by others previously ( 16,(18)(19)(20). Our results are also in agreement with the concept that there is a relationship between particle size with IF-toserum gradient, where the smallest particles display the highest gradient ( 15,16,18,19 ). Thus, the relative proportion of the lipoproteins decreases in the order HDL > LDL > VLDL. The apoAI gradient follows that of HDL, and that of apoB essentially that of LDL. The corresponding gradient for Lp(a) has not been described previously, but the fi nding that its IF-to-serum gradient is considerably lower than for LDL or apoB is reasonable, considering that Lp(a) represents a larger particle than ordinary LDL ( 21 ). enhanced with increasing levels of serum LDL cholesterol, the relative risk at a given LDL level is disproportionately elevated in diabetic compared with nondiabetic subjects ( 3,4,8 ). This is true also when taking the potential infl uence of other aspects of diabetic dyslipidemia (such as hypertriglyceridemia, low HDL, or the presence of small dense LDL) into account. We here tested the hypothesis that there is an unproportional accumulation of circulating lipoproteins in the subendothelial space in T2D. Such a "leakage" would result in an increased exposure of macrophages in the vascular wall to atherogenic lipoprotein cholesterol, which could explain this apparent anomaly. Unexpectedly, we found that the levels of VLDL and LDL lipoproteins (measured as cholesterol, TG, apoB, or free cholesterol) were instead relatively lower in diabetics than in matched controls.
It is obviously diffi cult to specifi cally assess the concentration of lipoproteins in the IF of the vascular wall, which would In the diabetes group, the IF-to-serum gradient for VLDL cholesterol and LDL TG are missing for one subject each due to a zero concentration in serum, which makes calculation of the gradient impossible. One value for IF-to-serum gradient for apoAI in healthy controls is excluded because of being an outlier. One value each for IF-to-serum gradient for apoB and Lp(a) in healthy controls is missing. For free cholesterol, the gradients are missing from three subjects. Data are presented as means ± SD. * P < 0.05; ** P < 0.01; *** P < 0.001 (Student's unpaired t -test). of LDL that display increased proteoglycan binding are also more easily taken up by macrophages ( 27 ).
As mentioned previously, T2D patients often display smaller LDL particles. This was confi rmed in this study when analyzing LDL cholesterol/apoB ratios, as this ratio in serum in the T2D patients was reduced compared with healthy controls. Interestingly, when analyzing the LDL cholesterol/ apoB ratio in IF, the opposite pattern was found (i.e., indications of larger particles in T2D). As smaller particles more effi ciently bind proteoglycans ( 28 ) one may speculate that small particles in the interstitial compartment in T2D more frequently are bound to proteoglycans and that this could be the reason for why we do not fi nd indications of more smaller particles in T2D than in healthy controls in IF. Moreover, the HDL cholesterol/apoAI ratio in serum was reduced in T2D compared with controls, indicating reduced HDL size in T2D. Because large HDL particles have been shown to reduce the affi nity of LDL to proteoglycans ( 29 ), the indications of smaller HDL particles in T2D in this material may cause increased proteoglycan binding and contribute to the lower-than-expected IF-to-serum gradient for LDL cholesterol. However, the HDL/apoAI ratio in IF did not differ between T2D patients and healthy controls.
In conclusion, apoB-containing, but not apoAI-containing, lipoprotein particles are sparser in IF relative to their serum levels in patients with T2D. The reduced levels of apoB-containing lipoproteins in IF may mirror an increased uptake of them from IF. The subsequent increased load of cholesterol to peripheral tissues could contribute to the markedly higher propensity to develop atherosclerosis in this disease. Further identifi cation of the mechanisms behind these unexpected fi ndings may lead to important insights that could be relevant in the prevention of accelerated cardiovascular disease in T2D.
Following current treatment recommendations ( 22 ), a majority of patients with T2D are on statin therapy. It was therefore essential to fi rst determine if such treatment alters the IF-to-serum gradient of LDL. This was done in a pilot study of hypercholesterolemic patients with claudication, but without symptomatic coronary heart disease. While atorvastatin therapy reduced serum and IF LDL and VLDL cholesterol as well as TGs, there was no change in the concentration gradients of any of the fractions in response to treatment. These data indicate that the exchange between serum and IF is not infl uenced by variation in the absolute levels of circulating lipoproteins and would also argue against the concept that statins exert some of their beneficial effects on cardiovascular health through changes in this aspect of endothelial function ( 23,24 ).
When comparing the cholesterol levels in T2D patients with those in controls, we found lower levels of total and LDL cholesterol in both serum and IF. These differences are most likely due to concurrent statin treatment, which was present in 24 of the 35 patients. Separate analysis showed that the patients taking statins indeed had lower serum total and LDL cholesterol levels than those who were untreated, but there was no signifi cant difference in IF cholesterol. Of particular importance was the fact that the IF-to-serum gradient did not differ between T2D patients with or without statin treatment (supplementary Table 1), in accordance with the results of our pilot study. Although the fact that a majority of our patients were on lipid-lowering or antidiabetic therapy is a limitation of our study, these fi ndings strongly indicate that our observations are primarily related to the diabetic state itself. The lower serum HDL cholesterol as well as the higher VLDL TGs observed in T2D patients in our study are in agreement with previous observations ( 25 ).
The present study demonstrated a clear reduction in the relative concentration of LDL (and VLDL) cholesterol and apoB in IF of T2D patients. In previous studies by Kornerup et al. ( 12 ), the early clearance of intravenously injected 131 I-labeled LDL was evaluated in T2D, and the disappearance rate of radioactivity was found to be 32% higher compared with controls. This was suggested to be due to an increased transvascular import of LDL particles into the interstitial compartment ( 12 ). To explain their fi ndings in relation to our current results, it is plausible to assume that the transfer of LDL into the IF is followed by an enhanced elimination of LDL particles from this compartment in T2D. As the vascular permeability of T2D patients is increased in the whole body, we expect lipoproteins to accumulate in several tissues. The parallel changes seen for VLDL TGs and cholesterol, as well as for apoB, would be compatible with that there is an increased catabolism of apoB-containing lipoproteins within the IF space in T2D. Whether this refl ects an enhanced consumption by activated tissue macrophages as part of the atherogenic process, or a stronger retention by the matrix structure within the vascular wall, can only be speculated on at this stage. In this context, it is of interest to note that LDL particles isolated from T2D patients display increased proteoglycan binding ( 26 ). Furthermore, it is also shown that subclasses