Apolipoprotein C3 Deficiency Results in Diet-Induced Obesity and Aggravated Insulin Resistance in Mice

Male and female apoc3^−/− mice and their wild-type littermates (C57BL/6 background) were originally obtained from The Jackson Laboratories (Bar Harbor, ME) and further bred in our institution. The 4- to 5-month-old animals (n = 15) were individually housed, allowed free access to food and water, and kept on a 12-h light cycle (lights on at 7:00 a.m.), under standard conditions. After a standard rat-mouse chow diet (Standard Diet Services, Essex, U.K.), the mice were given a high-fat corn oil diet (Hope Farms, Woerden, the Netherlands) until the end of the experimental period. This diet contained 24% corn oil, 24% casein, 20% cerelose, 18% corn starch, and 6% cellulose by weight, resulting in 46.2% of calories derived from corn oil. Body weight and food intake were followed through the duration of the experiment. Food intake was assessed by determining the difference in food weight during a 7-day period to ensure reliable measurements. Food intake was assessed as food weight (g) per mouse per day. From these data, the “feed efficiency” was calculated as total body weight gained per week divided by the total amount of food consumed per week. All experiments were approved by the animal care committee of TNO Prevention and Health (Leiden, the Netherlands).

Plasma levels of cholesterol, free fatty acids (FFAs), triglycerides (without free glycerol), glucose, ketone bodies (β-hydroxybutyrate), insulin, and leptin were determined after an overnight fast in apoc3^−/− and wild-type littermates after 0 and 20 weeks of high-fat-diet feeding. Blood samples were taken from the tail vein in paraoxon-coated capillaries to prevent lipolysis (15). The plasma was collected via centrifugation, and plasma cholesterol, triglyceride, glucose, ketone body, and FFA levels were determined using standard commercial kits, according to the manufacturer’s instructions (Roche Diagnostics, Mannheim, Germany; Triglyceride GPO-Trinder, glucose Trinder 500 and β-hydroxybutyrate, Sigma Diagnostics, St. Louis, MO; and NEFA-C, Wako Chemicals, Neuss, Germany, respectively). Plasma insulin and leptin levels were measured by radioimmunoassays (RIAs), using rat insulin standards, that show 100% cross-reaction with mouse and human insulin, or mouse leptin standards (sensitive rat insulin RIA kit, mouse leptin RIA kit, Linco Research, St. Charles, MO).

Mouse carcasses (wet weight) were dehydrated at 65°C until a constant mass was achieved (dry weight). The bodies were hydrolyzed in 100 ml of ethanolic potassium hydroxide (3 mol/l in 65% ethanol) for determination of body lipid, using enzymatic measurement of glycerol (Sigma Diagnostics), and body protein content by the Lowry assay (16). Total water content was calculated as wet weight minus dry weight, and lean body mass (LBM) was calculated as wet weight minus total lipid weight.

Pieces of WAT from reproductive fat pads were fixed in formalin and embedded in paraffin. Sections of 3 μm were cut and stained with hematoxylin-phloxine-saffron. Adipocyte size was quantified using Leica Qwin v1.0 (Leica Microsystems, Wetzlar, Germany).

To exclude obesity-induced differences in adipose tissue fatty acid uptake, we used body weight–matched apoc3^−/− and wild-type mice in the fed state, which had been treated with the high-fat diet for 2 weeks. The mice were sedated by intraperitoneal injection of hypnorm (0.5 ml/kg; Janssen Pharmaceutical, Tilburg, The Netherlands) and midazolam (12.5 mg/kg; Roche Netherlands, Mijdrecht, The Netherlands) and equipped with a catheter for tail-vein intravenous infusion. Large (150 nm) [³H]triolein-labeled chylomicron-like particles, prepared as described by Rensen et al. (17), were mixed with a trace amount of [¹⁴C]oleic acid complexed to BSA (both isotopes obtained from Amersham, Little Chalfont, U.K.) and continuously infused for 2 h into the animals as described earlier (18). Blood samples were taken using chilled paraoxon-coated capillaries by tail bleeding at 1.5 and 2 h of infusion to quantify steady-state conditions. Subsequently, mice were killed and organs were quickly harvested and snap-frozen in liquid nitrogen. Analyses and calculations were performed as described by Teusink et al. (18).

For determining the total LPL activity level, body weight–matched apoc3^−/− and wild-type mice, treated with the high-fat diet for 1 week, were fasted for 4 h and received an intraperitoneal injection of heparin (1 unit/g body wt; Leo Pharmaceutical Products, Weesp, the Netherlands). Blood samples were taken after 20 min. Plasmas of these samples were snap-frozen and stored at −80°C until analysis. One week later, the animals in the fed state were killed, and liver, heart, skeletal muscle (quadriceps), and WAT samples were collected. The organ samples were cut into small pieces and put in 1 ml of Dulbecco’s modified Eagle’s medium that contained 2% BSA. Heparin (2 units) was added, and samples were shaken at 37°C for 60 min. After centrifugation (10 min at 13,000 rpm), the supernatants were taken and snap-frozen until analysis. Total LPL activity of all samples was determined as modified from Zechner (19). In short, the lipolytic activity of plasma or tissue supernatant was assessed by determination of [³H]oleate production upon incubation of plasma or tissue supernatant with a mix that contained an excess of both [³H]triolein, heat-inactivated human plasma as source of the LPL coactivator apoC2 and fatty acid–free BSA as FFA acceptor. Hepatic lipase and LPL activities were distinguished in the presence of 1 mol/l NaCl, which specifically blocks LPL.

To allow for studying the effect of apoC3 deficiency on the modulated LPL activity in plasma, postheparin mouse plasma (as a source of LPL, apoC2, and apoC3) was incubated with a mix of [³H]triolein-labeled 75-nm-sized VLDL mimicking protein-free emulsion particles (20) (0.25 mg triglyceride/ml) and excess fatty acid–free BSA (60 mg/ml). Hepatic lipase and LPL activities were distinguished as described above, and the LPL activity was calculated as the amount of FFA released per minute per milliliter.

After an overnight fast and after 20 weeks of high-fat feeding, the animals were anesthetized, as described earlier, and basal rates of glucose turnover were determined followed by a hyperinsulinemic-euglycemic phase (plasma glucose at ∼7.5 mmol/l) as described previously (5,21). After the last blood sample, mice were killed, and liver, cardiac muscle, skeletal muscle (quadriceps), and adipose tissue were immediately frozen in liquid nitrogen and stored at −20°C for subsequent analysis. Carcasses were stored at −20°C until body mass composition was analyzed. Whole-body insulin-mediated glucose uptake and insulin-mediated suppression of endogenous glucose production were calculated on the basis of LBM. The whole-body insulin sensitivity index was expressed as the ratio between insulin-induced whole-body glucose disposal and hyperinsulinemic plasma insulin concentration. The endogenous glucose production insulin sensitivity index was expressed as a ratio between insulin-mediated inhibition of endogenous glucose production and hyperinsulinemic plasma insulin concentration. Insulin clearance (ml/min) was calculated from steady-state insulin concentrations and insulin infusion rates:

Tissues were homogenized in PBS (∼10% wet wt/vol), and samples were taken to measure protein content by the Lowry assay (16). Lipid content was determined by extracting lipids using the Bligh and Dyer method (22) and by separating the lipids using high-performance thin-layer chromatography on silica gel plates as described before (23), followed by TINA 2.09 software analysis (Raytest Isotopen meßgeräte, Straubenhardt, Germany) (24).

For determining the effect of apoC3 deficiency on the hepatic VLDL-triglyceride production rate, animals that were fed the high-fat diet for 2 weeks were fasted for 4 h and were anesthetized, followed by an intravenous injection of 10% Triton WR1339 (500 mg/kg body wt) to inhibit lipolysis and hepatic uptake of VLDL-triglyceride. Blood samples were drawn at 0, 15, 30, 60, and 90 min after Triton injection, and triglyceride concentrations were determined in the plasma as described above.

The Mann-Whitney U test was used to determine differences between apoc3^−/− and wild-type mice. The criterion for significance was set at P < 0.05. All data are presented as mean ± SD. Statistical analyses were performed using SPSS 11.0 (SPSS, Chicago, IL).

Male apoc3^−/− and wild-type littermate mice were put on the high-fat diet for a period of 20 weeks. Already after 2 weeks, apoc3^−/− mice showed a significant increase in body weight on the high-fat diet compared with littermate controls, leading to a 22% higher body weight in apoc3^−/− mice compared with wild-type mice at the end of the experiment (week 20), as shown in Fig. 1A. A significant increase in body weight was also observed in female apoc3^−/− mice compared with wild-type littermates, although less extreme (data not shown).

Food intake of male apoc3^−/− and wild-type mice was comparable during the first 11 weeks. After 11 weeks until the end of the experiment, the food intake of apoc3^−/− mice was increased 5–15% compared with that of wild-type littermates (Fig. 1B). The calculated feed efficiency (Fig. 1C) was significantly increased in the apoc3^−/− mice compared with wild-type littermates. The greatest difference in feed efficiency between the genotypes was seen at week 4 of high-fat feeding (0.048 ± 0.009 vs. 0.019 ± 0.007 g wt gain/g food consumed, for apoc3^−/− and wild-type mice; P < 0.05). This difference between the groups gradually decreased toward the end of the experiment, although apoc3^−/− mice still had a higher feed efficiency at 20 weeks of high-fat feeding compared with wild-type littermates (0.031 ± 0.006 vs. 0.019 ± 0.007 g wt gain/g food consumed, respectively; P < 0.05).

To investigate alterations in body composition, we analyzed mouse carcasses after 20 weeks of high-fat diet. Body weight, LBM, and the proportion of water, protein, and lipid of apoc3^−/− mice and wild-type littermates are shown in Table 1. Although body weight was ∼7 g higher in apoc3^−/− compared with wild-type mice, the LBM was comparable for both groups of mice. The absolute amount of body lipid in apoc3^−/− mice was ∼6 g higher compared with wild-type mice. No differences were found in protein content and amount of body water between apoc3^−/− mice and wild-type littermates. Analysis of adipocyte size in the reproductive fat pads revealed that after 20 weeks of high-fat feeding, adipocytes of apoc3^−/− mice and of wild-type mice are comparable in size (Table 1).

To show that indeed the increased adipose tissue mass was due to increased LPL-dependent triglyceride-derived fatty acid uptake, we determined the tissue-specific uptake of fatty acids derived from either plasma triglyceride or albumin in several tissues of nonfasted, body weight–matched apoc3^−/− and wild-type mice that were fed the high-fat diet for 2 weeks (Fig. 2). The small difference in body weight after only 2 weeks of high-fat-diet feeding ensured the availability of body weight–matched apoc3^−/− and wild-type mice. We observed no differences in uptake of albumin-bound fatty acids in liver, heart, muscle, and adipose tissue between apoc3^−/− and wild-type littermates. It is interesting that triglyceride-derived fatty acid uptake was significantly increased in visceral, subcutaneous, and reproductive fat pads from apoc3^−/− mice compared with wild-type mice. No differences were found in triglyceride-derived fatty acid uptake in liver, heart, and skeletal muscle in apoc3^−/− mice compared with littermates.

Because LPL-mediated triglyceride-derived fatty acid clearance was increased in WAT of apoc3^−/− mice, we determined the total plasma and tissue LPL activity in body weight–matched apoc3^−/− mice and littermate controls that were fed the high-fat diet for 1 week (plasma LPL) and 2 weeks (tissue LPL). Total postheparin plasma LPL activity was similar in apoc3^−/− and wild-type mice (7.0 ± 5.6 vs. 5.4 ± 3.2 μmol FFA · h⁻¹ · ml⁻¹, respectively), demonstrating that the absence of apoC3 does not affect LPL expression. Likewise, tissue-specific LPL activity measured in liver, heart, skeletal muscle, and visceral, subcutaneous, and reproductive fat pads was not different between apoc3^−/− and control mice (Fig. 3). We next studied the LPL activity in postheparin plasma in the absence of excess heat-inactivated human plasma and in the presence of limited amounts of VLDL-like emulsion particles rather than excess solubilized triglycerides. Under these conditions, the LPL activity as modulated by endogenous mouse plasma factors (e.g., apoC3) can be studied. Indeed, apoc3^−/− mouse plasma showed 78% increased triglyceride hydrolase activity compared with wild-type littermates judging from [³H]oleate production (1.33 ± 0.20 vs. 0.75 ± 0.17 nmol oleate · ml⁻¹ · min⁻¹, respectively; P < 0.05). Collectively, these data clearly show that apoC3 modulates LPL activity by interfering with the interaction between LPL and triglyceride-rich lipoproteins (i.e., VLDL and chylomicrons), rather than by affecting total LPL levels.

High-fat feeding induced increased total plasma cholesterol in both groups (Table 2). In apoc3^−/− mice, plasma glucose, ketone bodies, and leptin were also increased after 20 weeks of high-fat feeding compared with t = 0 (chow diet). Plasma total cholesterol and FFAs were comparable between the mice that were on chow and on the high-fat diet (Table 2). After 20 weeks of high-fat feeding, plasma ketone bodies and leptin levels were increased in apoc3^−/− mice compared with littermates. Plasma triglyceride levels were significantly lower in apoc3^−/− mice compared with wild-type mice as reported earlier (12). No significant differences were found with respect to plasma glucose and insulin levels in apoc3^−/− versus wild-type animals before high-fat feeding (Table 2). At the end of the 20-week period of high-fat feeding, apoc3^−/− mice showed significantly higher plasma glucose and slightly but not significantly increased plasma insulin levels compared with control littermates (Table 2). Insulin sensitivity was tested using hyperinsulinemic-euglycemic clamp analyses (Table 3). Under fasted conditions, plasma glucose levels were higher in apoc3^−/− mice compared with wild-type littermates. During hyperinsulinemic conditions, no significant differences were observed in plasma glucose. Insulin levels were increased in apoc3^−/− mice as compared with wild-type mice (Table 3). This increase in plasma insulin levels was explained by decreased insulin clearance in apoc3^−/− mice (Table 3). The clamp results revealed that insulin-mediated whole-body glucose uptake was significantly lower in apoc3^−/− mice compared with littermate controls (15 ± 5 vs. 35 ± 9%, respectively; P < 0.05; Fig. 4A). Moreover, the endogenous glucose production was only slightly suppressed under hyperinsulinemic conditions in apoc3^−/− mice as compared with wild-type mice (7 ± 1 vs. 27 ± 10%, respectively; P < 0.05; Fig. 4B).

Because apoc3^−/− mice were insulin resistant compared with wild-type littermates, we analyzed liver and muscle lipid content. In apoc3^−/− mice, hepatic lipid content was significantly higher compared with that in wild-type littermates (Fig. 5A). Skeletal and cardiac muscle triglyceride content did not significantly differ between both mouse groups.

Because increased liver triglyceride levels were observed in apoc3^−/− mice, the ability of the liver to secrete VLDL-triglyceride was investigated. Mice that were fed the high-fat diet for 2 weeks were fasted 4 h and received an injection of Triton WR1339, and the accumulation of endogenous VLDL-triglyceride in plasma was determined over time (Fig. 5B). The VLDL-triglyceride production rate as determined from the slope of the curve was unchanged in apoc3^−/− mice compared with wild-type littermates (3.89 ± 0.54 vs. 3.73 ± 0.58 mmol triglyceride/h, respectively). Therefore, the hepatic triglyceride accumulation as observed in apoc3^−/− mice is not caused by reduced VLDL secretion.

LPL plays an important role in the delivery of fatty acids into peripheral tissues. Several mouse studies indicate that decreased LPL activity in adipose tissue decreases the propensity to develop obesity (6–9,25). However, it is unclear whether the opposite is true as well, i.e., whether activation of LPL leads to an enhanced susceptibility to diet-induced obesity and associated insulin resistance. Adipose tissue–specific LPL overexpression leads to only a slight increase in fat pad weight (10). However, this relatively minor effect can result from an altered muscle versus adipose tissue LPL activity ratio and/or the presence of natural regulators of in vivo LPL activity (11). ApoC3 is a known inhibitor of LPL activity, and disruption of the gene in mice increases triglyceride clearance (12,14).

In this study, we investigated the possible effects of enhanced whole-body LPL activity (resulting from deletion of the LPL inhibitor apoC3) on development of obesity and insulin resistance making use of the apoC3 knockout mouse (apoc3^−/−). The data show that apoc3^−/− mice developed a higher adipose tissue mass during high-fat feeding compared with wild-type littermates. This higher adipose tissue mass in these mice was caused by a higher uptake of fatty acids derived from plasma triglyceride, leading to insulin resistance of whole-body glucose uptake and production. These data indicate that increased LPL activity, resulting from the absence of apoC3-dependent attenuation of the LPL activity, indeed profoundly increases the propensity to develop obesity and insulin resistance on a high-fat diet.

Apoc3^−/− and wild-type littermates were fed a high-fat, high-calorie diet. Within 2 weeks of high-fat feeding, male apoc3^−/− mice developed a significantly higher body weight. Body composition analysis, after 20 weeks of high-fat feeding, revealed that the increase in body weight in apoc3^−/− mice compared with wild-type littermates was completely explained by the increase in body fat mass. LBM, body protein, and body water were not different between the two mouse groups (Table 1). Up to a period of 11 weeks, no difference in daily food intake was observed between apoc3^−/− and wild-type mice, although body weight already differed after a relatively short period of 2 weeks. This suggests that the increase in body weight and fat mass is primarily due to the enhanced LPL activity and not due to a higher food intake. As a consequence, feed efficiency, as expressed as total weight gained per week divided by the total amount of food consumed per week, was clearly higher in apoc3^−/− mice (Fig. 1C). The unchanged adipocyte cell size found in apoc3^−/− mice after 20 weeks of a high-fat diet suggests that more adipocytes should be present (Table 1). One might in fact expect the opposite, and we can only speculate on the effects of increased fatty acid flux to adipose tissue and adipocyte differentiation. One clue seems to come from studies using peroxisome proliferator–activated receptor-γ agonists. Treatment with peroxisome proliferator–activated receptor-γ agonists leads to increased adipose tissue fatty acid flux and accumulation, interestingly with increased adipose tissue LPL expression (26). Consequently, more adipocytes occur with similar or even smaller size (27).

The observation that apoC3 deficiency has no effect on the uptake of albumin-bound fatty acid by the liver, heart, muscle, or adipose tissue (Fig. 2) indicates that the effect of apoC3 on fatty acid uptake in the respective tissues occurs solely via the modulation of LPL activity. We showed recently that triglyceride-derived fatty acid uptake in the various tissues is dependent on LPL activity in those tissues (18). In the fasted state, skeletal muscle is relatively enriched in LPL, whereas adipose tissue is relatively rich in LPL in the fed state (2,6,28–31). Thus, in the fasted state, relatively more triglyceride-derived fatty acid is transported to the muscle, whereas in the fed state, transport of triglyceride-derived fatty acids to the adipose tissue is more pronounced. In the current study, we show that apoC3 deficiency does lead to increased uptake of triglyceride-derived fatty acids by adipose tissue. This implicates, therefore, that the effect of apoC3 on LPL-mediated fatty acid uptake is relevant only in the (high fat) fed condition. Despite that LPL activity is the highest in the heart (Fig. 3), we did not observe an effect of apoC3 deficiency on triglyceride-derived FFA uptake by this organ (Fig. 2). These data are in agreement with our previous observations that either reduced expression of LPL as in VLDL receptor–deficient mice (32) or increased LPL activity by overexpression of the LPL activator apoA5 (33) did not modulate the uptake of triglyceride-derived FFA by the heart. Apparently, the cardiac uptake of triglyceride-derived FFA is irrespective of LPL modulation, as related to its continuous need for vast amounts of fuel.

Apolipoproteins can alter LPL activity by interfering with the (physical) interaction between LPL and triglyceride-rich lipoproteins such as VLDL and chylomicrons, without regulating the LPL protein and/or gene expression in tissues. To exclude an effect of apoC3 deficiency on total LPL levels, we analyzed postheparin plasma and tissue-specific LPL levels in apoc3^−/− and wild-type littermates. In this assay, plasma and tissue samples are diluted into a very large pool of solubilized triglyceride and heat-inactivated human plasma, thereby abolishing the interaction of LPL with potential modulators of LPL activity (e.g., apoC3) contained in the samples. Indeed, no differences were observed in total plasma or tissue LPL activity between apoc3^−/− and littermate controls (Fig. 3). However, under conditions in which mouse plasma is added to limited amounts of VLDL-like particles in the absence of excess human plasma, the triglyceride hydrolase activity of plasma LPL from apoc3^−/− mice was indeed 78% higher than that of wild-type mice. Our data thus clearly show that the absence of apoC3 from lipoprotein particles is responsible for enhancing LPL activity in vitro as evident from increased triglyceride-derived FFA liberation and in vivo as evident from increased triglyceride-derived fatty acid flux into adipose tissue of apoc3^−/− mice. Furthermore, the plasma triglyceride half-life was only 6 ± 3 min in apoc3^−/− mice vs. 22 ± 6 min for littermate controls. This clearly demonstrates increased LPL-mediated triglyceride clearance in apoc3^−/− mice, which is in agreement with previous reports (12,14). Therefore, it would be interesting to investigate whether human apoC3-overexpressing mice could be protected from diet-induced obesity.

Although we cannot exclude that altered adipose tissue lipid handling, e.g., hormone-sensitive lipase activity, leads to a similar adipose tissue phenotype, the absence of hormone-sensitive lipase leads to decreased adipose tissue mass instead of an increase despite increased adipose tissue LPL activity (34).

Using hyperinsulinemic-euglycemic clamp analyses, we observed that the insulin-mediated stimulation of whole-body glucose uptake was strongly impaired in apoc3^−/− mice as compared with wild-type littermates (Fig. 4A), indicating decreased insulin sensitivity in apoc3^−/− mice. Insulin sensitivity was found to be negatively correlated with muscle triglyceride content in several studies (3–5). However, we could not observe a statistically significant increase in muscle triglyceride content in apoc3^−/− mice when compared with wild-type littermates (Fig. 5A), in line with the observation that there was no increase in muscle uptake of albumin-derived fatty acids or triglyceride-derived fatty acids in fed mice. Presumably, apoC3 deficiency leads to insulin resistance as a result of increased adipose tissue mass. For instance, apoC3 deficiency might affect secretion of various endocrine factors by adipose tissue, such as leptin, resistin, and adiponectin. These hormones are known to affect insulin sensitivity and are correlated with adipose tissue mass (35–38). In this study, we found that apoc3^−/− mice have increased plasma leptin levels (Table 2), in accordance with the increase in adipose tissue mass (Table 1). However, it is likely that the hyperleptinemia observed in these mice is the consequence rather than the cause of insulin resistance, as has been observed earlier in humans (35,39–42).

Suppression of the endogenous glucose production (largest contribution by liver) by insulin in apoc3^−/− mice on a high-fat diet was completely absent compared with wild-type littermates (Fig. 4B). In the liver, triglyceride content is inversely correlated with hepatic insulin sensitivity (4). Indeed, the decreased hepatic insulin sensitivity in apoc3^−/− mice coincides with an increased hepatic triglyceride content in these mice (Fig. 5A). The higher triglyceride levels in livers of apoc3^−/− mice seem not to be related to increased hepatic total fatty acid uptake in these mice (Fig. 2). Also, the hepatic VLDL-triglyceride production was not affected in apoc3^−/− mice on high-fat diet (Fig. 5B), which is similar to our previous observations on chow diet (14). Alterations in intrahepatic fatty acid metabolism and/or other factors, e.g., adipokines, might in fact be fundamental to the observed mild hepatic steatosis in these high-fat-fed apoc3^−/− mice.

In conclusion, enhanced LPL activity in mice as a result of absence of the natural inhibitor of LPL, apoC3, leads to enhanced susceptibility to diet-induced obesity. This enhanced obesity seems to be fully explained by the increased triglyceride-derived fatty acid flux toward adipose tissue rather than by differences in food intake. The increased obesity and probably the altered fatty acid partitioning result in severe development of peripheral and hepatic insulin resistance in apoc3^−/− mice. Absence of hepatic production of apoC3 decreases plasma triglyceride levels by enhancing adipose tissue triglyceride-derived fatty acid uptake. Because triglyceride-rich lipoproteins are atherogenic, the decrease in plasma triglyceride levels might be beneficial in the prevention of cardiovascular diseases. Adversely, the higher in vivo LPL activity in turn leads to higher susceptibility to diet-induced obesity and insulin resistance. It is intriguing that our results show that the liver can modulate plasma lipid levels and at the same time fat mass and insulin sensitivity through production of just one protein, apoC3. It warrants further investigation whether in humans, apoC3 is a potential therapeutic target for treatment of obesity and insulin resistance and at the same time for prevention of cardiovascular risk.

This study was conducted in the framework of the Leiden Center for Cardiovascular Research LUMC-TNO and supported by the Netherlands Organisation for Scientific Research (NWO Grant 903-39-179, NWO VIDI Grant 917-36-351 to P.C.R., NWO VENI Grant 916-36-071 to P.J.V., and program grant 903-39-291 to L.M.H.) and the LUMC (Gisela Thier Fellowship to P.C.R.).