Introduction PhD Thesis

Back to PhD thesis

 Cardiovascular disease is the most frequent cause of death in the world (see WHO factsheet[1]), but its diagnosis is complex and can be improved. This PhD thesis contains a report of research aimed at improving the diagnosis of cardiovascular disease in an early stage.
         Currently, cardiovascular disease is diagnosed based on a range of risk factors, including age, gender, total cholesterol, HDL cholesterol, smoking, blood pressure and blood pressure medication. From this information, the ‘Framingham risk score’ can be calculated that indicates the 10-year risk for coronary heart disease outcome (1). Alternative risk scores based on other population studies have also been developed (see e.g. (2,3)). Although this approach identifies a sub-population at risk for cardiovascular disease, there is still room for improved risk prediction. To give a rough indication, statin treatment is started from approximately a 10% 10-year risk, and results in a risk reduction of approximately 30%. In a population of people at 10% risk, 97% would therefore take statins without benefit within a 10-year timeframe. Health insurance needs to pay for the unnecesary medication, and people taking statins may also suffer from side-effects. Improved diagnostics help to improve this situation.
         Lipoprotein biology is an area in which diagnostics can be improved. Lipoproteins are the particles that carry cholesterol and triglycerides through the bloodstream. Especially low-density lipoprotein (LDL) cholesterol and LDL particle concentrations are known to be positively associated with cardiovascular disease risk (4). LDL particles contain the protein apoB, and are to a large extent a metabolic product formed out of larger apoB-containing lipoproteins called VLDL and IDL. Technological advances allow the full size spectrum of  lipoproteins to be measured in increasing detail (5-9). Although such a detailed measurement contains much information, it does not lead to diagnostic values that both use all relevant information and that doctors can easily use. Instead, the detailed lipoprotein profile contains many measured parameters that need a further interpretation step to be useful for the clinic. One possible way of interpreting this data is by pooling all LDL particles, and reporting an ‘LDL particle number’. This diagnostic has proven to be successful at predicting cardiovascular risk (4). Still, the lipoprotein profiles contain more information that is discarded when only reporting LDL particles, for instance about the LDL precursor VLDL. Therefore, there exists an opportunity for a computational model that can interpret all measured information, transform it into new diagnostic markers, and in that way contribute to improving cardiovascular risk prediction.
         We have developed a mathematical model to aid the interpretation of detailed lipoprotein measurements. This model is able to derive information on the status of lipoprotein metabolism from a single lipoprotein measurement. This PhD thesis reports on the development of the model and model-based diagnostics, their application in a nutritional and a pharmacological human intervention study, and their validation as predictors for cardiovascular disease.

Scientific background questions


How are cardiovascular disease, cholesterol and lipoproteins associated?

The diagnosis of cardiovascular disease is complex. Usually, a range of 'risk factors' is evaluated, and treatment focuses on reducing all these risk factors. The risk factors included in the 'Framingham risk score', the most often used risk score model, are age, gender, total cholesterol, HDL cholesterol, smoking, systolic blood pressure, diabetes, left ventricular hypertrophy, and current blood pressure medication (1). Other risk score calculators, such as PROCAM, also include LDL cholesterol, triglycerides, family history and a regional prevalence adjustment factor (3). Many risk scores have been developed, a detailed comparison can be found in this review (10).
         All these risk scores show that an elevated plasma cholesterol concentration is a key risk factor for cardiovascular disease, and, in consequence, also a key treatment target. This observation is supported by the statistics on the number of people on cholesterol-lowering medication, which are usually statins. For example, in the Netherlands, in the period of 1997-2002 statin use among insured people was 47.28 defined daily doses per 1000 people per day (11). Statins have proven to be very effective cholesterol-lowering agents. They have proven to lower cardiovascular disease incidence, but a large number of patients at risk for cardiovascular disease need to be treated to prevent a small number of disease cases (12,13). Currently, all patients that have an elevated Framingham Risk score are treated. Of these patients, only a small percentage would actually get a cardiovascular event in a 10-year period if untreated, and statins prevent approximately 25% of these events. Taken together, only a small percentage of patients benefit from statin treatment, and many patients take either superfluous or insufficient drugs, leading to costs for healthcare and sometimes the discomforts of side-effects. Therefore, in order to reduce this overmedicalisation of society, there is a need for improved diagnostics to indicate exactly which patients would benefit most from treatment with statins or other cholesterol-lowering medication.
         Cardiovascular disease and lipoproteins are related in a similar way as cardiovascular disease and cholesterol are, since, functionally, lipoproteins transport cholesterol through the bloodstream. Yet, lipoprotein measurements may contain additional information that is relevant for cardiovascular disease prediction. For instance, LDL particle number measurements and apolipoprotein B measurements have been found to predict disease risk more accurately than LDL cholesterol measures in certain studies (4). This improvement has been attributed to the fact that especially small LDL particles, which contain less cholesterol but an equal amount of apoB, are thought to contribute to cardiovascular disease risk.
         Next to LDL particles, other types of lipoprotein measurements also hold promise for improving cardiovascular risk prediction. VLDL and HDL fractions can be measured in considerable detail, with many size or density subfractions. Since VLDL is the metabolic precursor of LDL and HDL is involved in reverse cholesterol transport from the periphery to the liver (see below), this subfraction information is potentially useful for refining risk prediction. Yet, adequate interpretation is necessary to unlock the potential of this data. This interpretation requires further understanding of lipoprotein physiology.

What is the role of lipoproteins and apoB in the human body and how does their physiology work?

Lipoproteins are particles that transport cholesterol and triglycerides through the blood. The particles contain triglycerides and cholesterol esters in their core, while the membrane around the core consists of phospholipids, free cholesterol and proteins called apolipoproteins that take care of many of the biochemical interactions of the lipoprotein (see Figure 1.1). There are two main types of lipoproteins: ApoB-containing lipoproteins and ApoA-containing lipoproteins. ApoB-containing lipoproteins include LDL and its precursors IDL and VLDL. These particles transport triglycerides from the liver to extrahepatic tissues and are considered atherogenic (14). ApoA-containing lipoproteins (also called HDL) are thought to be involved in reverse cholesterol transport from extrahepatic tissues to the liver and are thought to have additional anti-atherogenic functions (see (15)).
         Most lipoprotein kinetics studies have focused on apoB-containing lipoproteins, since they are easier to track than HDL. Each particle has a single apoB molecule, which remains fixed to the particle during its entire life cycle. In contrast, HDL particles may contain multiple apoA molecules, which can interchange between particles, thereby generating more complex kinetics. Therefore, apoB containing particles have been studied more extensively and are better understood, while apoA kinetics is less well understood. Since for constructing a computational model a large body of background knowledge is necessary, we chose to build forward on the existing knowledge concerning apoB metabolism. Therefore, the research in this thesis focuses on apoB-containing particles, and leaves apoA containing particles aside. In continuation, we will explain what is known about the metabolism of apoB-containing particles.

Figure 1.1: Schematic representation of the composition of a lipoprotein, and the processes involved in apoB-containing lipoprotein metabolism. Lipoproteins transport triglyceride and cholesterol through the bloodstream. These components are packed inside a phospholipid membrane. The membrane contains free cholesterol and various proteins known as ‘apolipoproteins’ that regulate the biological functions of the particle. The lipoprotein lifecycle consists of the following processes: production by the liver, lipolysis of large particles by extrahepatic tissue and lipolysis of small particles by the liver, and uptake by the liver. Uptake mainly concerns small particles, although larger particles are also taken up.

The three processes in the lifecycle of an apoB-containing lipoprotein particle are: production, remodeling and lipolysis, and uptake. ApoB-containing lipoproteins are usually produced as large, triglyceride-rich VLDL particles that become smaller during the process of losing their core triglycerides through lipolysis, subsequently becoming IDL and LDL. We will now give an overview of the biochemical background of each of the three main processes.
         The first process is lipoprotein production. ApoB-containing lipoprotein assembly takes place in the liver. It is thought to be a three-step process. The first step is the transcription of the apoB molecule, followed by a first lipidation step, in which triglycerides are added to the nascent particle. If insufficient triglycerides are available, the apoB molecule can be degraded. If there are sufficient triglycerides, the particle can either be secreted, usually as a VLDL2 particle, or a second lipidation step can occur. In this second lipidation step, the nascent lipoprotein fuses with a lipid droplet, which results in a VLDL1 particle that can be secreted into the plasma. More detail about the exact intracellular localization of these steps, and the biochemical entities involved can be found in the PhD thesis of Caroline Beck (16).
         The second process is the remodeling of apoB-containing lipoproteins in the plasma. During remodeling changes occur in triglyceride content, as well as in phospholipid, cholesteryl ester and protein composition. ApoB-containing lipoproteins go through a sequence of remodeling steps during which they become successively smaller. The driving process during remodeling is lipolysis, through which changes in the triglyceride content occur. Lipoprotein lipolysis takes place through two mechanisms. Lipolysis of lipoproteins in extrahepatic tissues is carried out mainly by lipoprotein lipase (LPL). This enzyme mainly lipolyzes larger lipoproteins such as VLDL1, while VLDL2 and IDL are lipolyzed to a subsequently lesser extent (17). The particle binds to cell-surface heparan sulfate proteoglycans (HSPGs) and GPIHBP1 (18) mainly through LPL itself, while ApoE modulates the binding affinity (19). Multiple LPLs which are already bound to the HSPGs can then be transferred to the lipoprotein, and mediate the lipolysis of the particle. What exactly determines the rate of this lipolysis is not known, although the available surface area, the biochemical composition of the particle (20), gene expression changes, activators (e.g. ApoCII, ApoE, ApoAV), inhibitors (e.g. ApoCI, ApoCIII, Angptl4) and modulators of LPL expression (e.g. VLDL receptor) (21) are all thought to influence this rate.
The second lipolysis mechanism is via hepatic lipase (HL), which mainly occurs in the liver. In this process, the particles first bind to liver HSPG via ApoE and can subsequently be lipolyzed by HL. This enzyme functions primarily on smaller ApoB and ApoE-containing lipoproteins such as IDL and LDL, and to a lesser extent on VLDL2 (22).
         Another candidate for lipolyzing apoB-containing lipoproteins is endothelial lipase (EL), an extracellular lipase mainly produced by, and bound to endothelial cells (23,24). This enzyme mainly affects HDL particles, and although this has been little researched, there is a possibility that it affects apoB-containing lipoproteins as well (25). Yet another enzyme affecting lipoprotein size is cholesterol ester transfer protein (CETP), which facilitates the transfer of CEs from CE-rich LDL and HDL toward VLDL. CETP promotes the reciprocal enrichment of LDL and HDL with TGs derived from VLDL (26,27). Treatment with the CETP-inhibitor torcetrapib was shown to have impact on apoB-containing lipoprotein kinetics. However, these effects were not exclusively related to CETP action, but were thought to relate to increased LPL activity resulting from CETP action (28). For both EL and CETP, determining their direct effect on lipoprotein lipolysis is complicated due to either lack of data or concurrent effects of other proteins.
         The third process in the lipoprotein life cycle is lipoprotein uptake or clearance, which occurs in the liver. In the liver, VLDL and IDL particles first bind to liver HSPG via ApoE, whereas LDL binds directly to LDL receptors via ApoB100. LDL particles are directly taken up, but ApoE-mediated binding of larger particles need not result in uptake and can lead to lipolysis via HL instead, as discussed above. Uptake of VLDL lipoproteins mainly takes place via the LDL receptor, but can also take place via low density lipoprotein receptor-related protein (LRP) (29). Roles for scavenger receptor class B type I (SR-BI) and direct incorporation via HSPGs have also been suggested (30,31).

As can be judged from the above, the lifecycle of apoB-containing lipoproteins has been studied extensively. But how straightforward is it to measure lipoproteins and their metabolic state in individual people? We will study this question in continuation.

What methods are available for measuring lipoproteins and to what data types do these methods lead?

For the description of lipoprotein measurement methods, we will distinguish between concentration and kinetics measurements, and discuss the interpretational challenges that the more recent data types present.

Concentration measurements

Historically, the first lipoprotein classes to be related to cardiovascular risk were separated based on their flotation characteristics(32,33). Ultracentrifugation of lipoproteins in plasma made a separation according to lipoprotein density possible (34,35). The VLDL, IDL, LDL and HDL fractions were defined based on this method. Although LDL concentrations were found to predict cardiovascular risk better than total cholesterol concentrations in plasma (36), practical objections to ultracentrifugation led to a more widespread use of total cholesterol and triglyceride measures in the clinic. Subsequently, a chemical method was devised to quickly measure HDL cholesterol (37), and dr. Friedewald devised his formula (38) to derive the LDL cholesterol concentration from total cholesterol, HDL cholesterol and triglyceride measurements. These rapid measurements became part of routine clinical chemistry assessments. In addition to the cholesterol-based risk factors, apolipoprotein measurements such as ApoB or the ApoB / ApoA-I ratio have been found to indicate atherosclerosis risk (39-41). Technical improvements for the measurement of lipoproteins have been around for some time, and new methods are under development. The NMR lipoprotein profile measures a range of lipoprotein subclasses (5,42). We will refer to such detailed concentration measurements in more size- or density-based subclasses than the traditional VLDL to HDL subclasses as “lipoprotein profiles”. Lipoprotein profiles can also be measured using HPLC (6), Ion Mobility Analysis (7,8),Vertical Auto Profile (43) or Asymmetric Flow Field-Flow Fractionation (9). These methods measure different lipoprotein subclasses, as shown in Figure 1.2. One company is planning a more widespread introduction of their technology into clinical practice. [2] A more elaborate historical review of the development of cholesterol diagnostics can be found here (44).



Figure 1.2: Diagram of different subfractionation schemes. Ultracentrifuge represents the classical density separation, by approximation converted to size separation. AF4 stands for Asymmetric flow field-flow fractionation, Lipofit and Liposcience are companies marketing NMR-based methods, Liposearch markets an HPLC based method.



Table 1.1: Table of different subfractionation schemes. Methods as in Figure 1.2. Figures represent subclass boundaries of lipoprotein particle diameter in nm.

Ultracentrifuge
AF4
Lipofit
Liposcience
Liposearch
Ion Mobility
LDL
15.0
15.0
15.0
15.0
15.0
15.0


16.3

15.3
17.7
18.0


19.8
19.0
19.8
19.7
20.2



21.0
21.3
21.9
21.1


23.7
22.0
23.5

22.0
IDL
25.0

25.0

24.3
23.8


27.8

27.0
27.1
26.8
VLDL 2
30.0
32.3
30.0
31.0
30.0
29.6
VLDL 1
36.0
37.0

35.5
34.1
33.5


42.0
40.0
44.0
40.7
42.4


47.2






52.6






58.3
60.0
60.0
58.8
52.0


64.1






70.0






76.1

75.0



80.0
82.3
80.0
80.0
80.0


Kinetics measurements

The kinetics of ApoB-containing lipoproteins has been investigated extensively. The most comprehensive investigations in humans have been done using isotopic tracer studies, which measure the kinetics of incorporation and loss of externally added  radioactive or stable-isotope tracers in lipoproteins (45). Such studies give a comprehensive view of lipoprotein metabolism, yet they are very expensive and time-consuming. Still, a large number of studies has been carried out, characterizing patients with many different medical conditions affecting lipoprotein metabolism. Therefore, the tracer methodologies have resulted in much insight into lipoprotein metabolism, as discussed by Parhofer et al. (46). Most studies have measured isotope enrichment in subfractions separated by density, the VLDL, IDL and LDL fractions (47,48), in some cases with an additional distinction between VLDL1 and VLDL2 (49,50) or only in the VLDL1 and VLDL2 fractions (51).
         The first lipoprotein kinetics studies were carried out using radioiodinated tracers exogenously attached to lipoproteins to trace VLDL metabolism (52,53). Later, endogenous labeling using stable isotopes was developed (54). Various stable isotope-labeled compounds (e.g., [15N]glycine, [13C]phenylalanine, [13C]leucine, and [2H3]leucine) have been used to study lipoprotein kinetics. All of these give similar results when used to determine apoB-100 production rates (55). Nevertheless, leucine has many advantages as a protein tracer (56) because it is an essential amino acid, is readily available, and is not converted into other amino acids. Other tracers, such as [2H5]glycerol, that label the triglyceride content of the particle have also been used (51), giving more insight into the kinetics of the different lipoprotein constituents. Barrett et al. have written a detailed review on the design and analysis of lipoprotein tracer kinetic studies (45).

The lipoprotein profile challenge

The potential of the detailed lipoprotein subfraction information disclosed by the newer lipoprotein concentration measurement methods is increasingly recognized. Still, it is hard to say which lipoprotein subclass contributes most to cardiovascular disease development, because of statistical difficulties that we will discuss in continuation. Therefore, in practice, subclasses are often pooled and a parameter such as ‘LDL particle number’ is reported. This procedure facilitates interpretation but in fact reduces the measured information, and even discards information that is likely to be relevant for cardiovascular risk prediction. For example, several aspects of the lipoprotein phenotype are known to be statistically associated with cardiovascular risk together, in a so-called ‘Atherogenic Lipoprotein Phenotype’ or ‘lipid triad’, which consists of raised triglycerides, low HDLc and small-sized LDL particles.(14) The complete lipoprotein profile contains detailed information on all these traits, while the ‘LDL particle number’ does not. Therefore, data analysis tools that can assist in integrating the profile’s full information content, thereby optimally using the data would be helpful. We will now first examine the power of statistical techniques for analyzing lipoprotein profile data before concentrating on mechanistic computational modeling approaches.

What is the power and what are the limitations of using statistics for analyzing lipoprotein profiles?

Several statistical approaches have been applied to detailed lipoprotein profiles. In continuation, we discuss three of them.

Univariate statistics

We speak of univariate statistics when a statistical analysis only takes into account one variable in the analysis. Based on some of the lipoprotein profile measurement methods such as the NMR lipoprofile test, the single-variable ‘LDL particle number’ diagnostic has been defined that has proven to be effective for cardiovascular risk prediction (4). But this diagnostic discards most of the measured information in the lipoprotein profile, which also contains information on various HDL, IDL and VLDL particle subclasses.  However, it is difficult to relate cardiovascular risk to so many different subclasses. As more potential diagnostic markers, in this case more lipoprotein subclasses, are measured, more predictors of cardiovascular disease are possible. For example, in a lipoprotein profile with 20 subclasses, risk can be related to the triglyceride, cholesterol and particle number content of each of these classes. In total, this would lead to 60 possible predictors for cardiovascular disease risk. It then becomes more likely that in a study, one of the predictors seem to be predictive for the diseased state by chance. This problem is known as the 'multiple comparisons’ problem in statistics, and is the main cause why univariate statistics is of limited use for defining diagnostics based on lipoprotein profile measurements.

Multivariate statistics

This multiple comparisons problem provides an incentive for new approaches to aid the interpretation of detailed lipoprotein profiles. One possible approach is to use multivariate statistics, as Musunru et al. showed (57). Multivariate techniques such as principal component analysis (PCA), allow inspection of how multiple variables co-vary in a specific patient population. Using this approach, Musunru et al. were able to identify three major independent components of CVD risk. The first component represents LDL-associated risk, the second represents HDL-associated protection, and the third component represents a pattern of decreased large HDL, increased small/medium LDL, and increased triglycerides. The last corresponds to the previously described “atherogenic lipoprotein phenotype.” (57)
         The main advantage of this method is its ability to decompose cardiovascular disease risk into three independent (i.e., zero correlation) risk components. Based on this analysis, causal mechanisms cause for each component can be studied. Musunru et al. studied these mechanisms by associating each of the risk components with SNPs. For the component corresponding to the “atherogenic lipoprotein phenotype”, several associations were found. These were partially known associations with the CETP, LIPC, APOA1/A5, and LPL genes, and partially unknown associations with the GALNT2 and MLXIPL genes. The multivariate approach, therefore, considerably improves the univariate analysis of lipoprotein profiles.
         The multivariate approach also has some limitations. First of all, since the method is directly data-based, the results can only be used for other patients that are measured with exactly the same measurement method. Other measurement methods have different ways of classifying lipoproteins, and are consequently not directly comparable. Second, the statistical method does not give direct information about the processes underlying the different risk axes. Although the authors reason that “the correlations observed among the lipoprotein subfractions may reflect hitherto unmeasured biological processes that may be more relevant to cardiovascular risk than the standard lipoprotein measures” (57), the relation between the identified risk axes and underlying biological processes is not directly clear from the analysis. Therefore, there is room for an analysis method that a) can take into account different measurement methods, and b) that can derive a more direct relation with underlying processes.

K-means clustering

K-means clustering is an unsupervised classification method that partitions a dataset into K non-overlapping clusters. Subjects are assigned to the same cluster when they have similar data patterns, while K distinctively different patterns are distinguished. The method returns a number  of ‘centroids’ (K centroids, to be exact) that are representative for the data in each of the clusters, and also shows which individuals belong to each cluster.
         When applied to lipoprotein profiles, K-means clustering methodology can give an impression of the types of distinctive lipoprotein profile patterns present in the population. The clusters thereby define subgroups based on the lipoprotein profile, which can be analyzed further to see whether other properties of the patients in the subgroups differ, such as their genetic makeup or drug response. Such research can, therefore, be used to generate hypotheses on how the lipoprotein profile relates to other relevant patient characteristics, which possibly provides links to biological mechanisms involved.
         The power of the clustering method for analyzing lipoprotein profiles is limited by the fact that it does not analyze the profiles in terms of biology. Also, it does not result in a continuous diagnostic, but only discriminates between K distinct groups. If a method would be able to analyze the lipoprotein profile biology and derive a continuous diagnostic from it, such a method would provide a more direct link to the disease mechanism. This mechanistic insight is interesting for diagnostic purposes. Furthermore, such a method would also allow setting various ‘cutoffs’ to best define groups with high and low CVD risk or high and low treatment response. Clustering does not allow such optimal group selection, since the groups are pre-defined based on the clustered variables. A continuous variable allows selecting subgroups to find the best possible balance between specificity and sensitivity of a diagnostic (58). These limitations of the clustering approach motivate searching for additional methods to interpret lipoprotein profiles. One candidate method is mechanistic computational modeling.

What perspective does mechanistic computational modeling offer for lipoprotein profile analysis?

Mechanistic models can include biological knowledge in the model formulation. The model analyzes the data and returns biological parameters that can subsequently be related to disease risk using statistical techniques. Examples of diagnostic models of this kind include a model for the blood coagulation process, which derives kinetic coagulation parameters from a measured time course of coagulation factor activities (59), and a model to derive insulin sensitivity parameters from an oral glucose tolerance test (OGTT), via simulation and interpretation of the measured time course of plasma glucose values after an oral glucose dose (60). Lipoprotein profiles would be a suitable data type to apply a modeling approach to. In fact, first attempts in this direction have already been undertaken, as we will discuss in continuation.
         Existing computational models dealing with apoB-containing lipoproteins can be divided roughly into two types. One type of models was developed for analyzing the isotopic tracer experiments mentioned earlier (50,51) (see Figure 1.3). These are mostly compartmental models that use isotope enrichment curves from different lipoprotein fractions over time, and return lipoprotein production, clearance and turnover rates. These models obtain the mentioned rates by iteratively changing the model parameters until the model outcome matches the data, a process known as fitting. The fitted model parameters contain the relevant information about the lipoprotein’s metabolic processes that are included in the model.
         The second model type was constructed to simulate detailed lipoprotein profiles, as measured by density separation. The one example of this model type was presented by Hübner et al. (61). The Hübner model attempts to include all relevant biochemical reactions underlying a lipoprotein profile, and then simplify to achieve numerical traceability. It was devised with the goal of deriving model-based diagnostic markers from a lipoprotein profile, but it faces an over-parameterization problem. This problem means that the model parameters cannot be determined unambiguously based on the lipoprotein profile data. The Hübner et al. paper reads: “there is no unambiguous relationship between the conventionally measured pattern of lipoprotein main classes (VLDL, IDL, LDL, HDL2, HDL3) and the kinetic parameters of the kinetic processes included in the model.” (61) In consequence, the parameter values of a model that includes all possible biological processes cannot be specified based on a single lipoprotein profile. Therefore, a model for analyzing lipoprotein profiles should integrate biological processes in such a way that its parameters can be specified based on a single profile, and that it extracts sufficient biological information to be interesting as a diagnostic.

The two types of modeling approaches for lipoprotein kinetics cannot be applied widely in the clinic. The first approach is not suitable because the isotopic tracer methodology is currently too expensive and time-consuming. Measuring a single lipoprotein profile, in contrast, is much less expensive and much faster. The second approach does not provide an unambiguous relation between the lipoprotein profile and the diagnostic parameter. When deriving diagnostic markers, an unambiguous relation should exist between the lipoprotein profile and the diagnostic marker, or the marker itself becomes ambiguous. This principle should be taken into account when designing a new model. One way of tackling this problem is by integrating biological processes that form a functional unity, and distinguishing only those processes that are thought to be relevant for the disease. The data may contain enough information to distinguish the integrated processes, whereas it may not contain enough information to distinguish the subprocesses.  Therefore, there is room for a new modeling approach that improves on the two mentioned aspects.



Figure 1.3: Model of apoB metabolism developed by Packard et al (50). The model was developed to explain 131I-labeled VLDL1 and 1251-labeled VLDL2 apoB metabolism, and had to fit two sets of decay curves and the measured apoB masses in VLDL1, VLDL2, IDL and LDL. In the model shown here 131I-labeled apoB radioactivity decay curves were modeled by compartments 1, 2, 4, 6, 8, 9, 11-13. The 125I-labeled VLDL2 model had to include the kinetics of apoB derived from VLDL1 as dictated by compartments 22-29, 31, and 34 and also allow for independent behavior of apoB input at the level of VLDL2 in compartments 5, 7, and 10. (from (50)).

What useful information do lipoprotein profiles contain that model-based diagnostics could indicate?

We have seen that lipoprotein profile measurements contain interesting information that is not optimally utilized by current data analysis techniques, and that mechanistic computational modeling is an option for improving the analysis. This raises the question what information can be obtained from such a profile. It clearly does not contain flux information and not all underlying biology can be quantified based on such a profile, as the Hübner model shows. In principle, it is possible to derive flux ratios, such as a lipolysis / production ratio or an uptake / production ratio from a lipoprotein profile. The model could integrate all the information from the lipoprotein profile in the process of deriving those ratios. Lipoprotein metabolic ratios are, therefore, candidates for becoming model-based diagnostics.
         The first projected use of lipoprotein metabolic ratios, is to improve cardiovascular risk prediction. Lipoprotein metabolic ratios can hope to improve CVD risk prediction, since they help to distinguish between liver and extrahepatic tissue function. The liver is responsible for lipoprotein production, uptake and lipolysis of smaller particles, while extrahepatic tissue lipolyzes the larger lipoprotein particles (see Figure 1.1). Dysfunction of these tissues not only influences the formation of small, dense LDL particles that are thought to confer risk (62), but also disturbances of triglyceride and HDL metabolism confer additional risk (14). Therefore, metabolic flux ratios integrate various aspects of cardiovascular risk by focusing on the functioning of tissues. This integration opens perspectives for improved risk prediction.
         A second projected use of lipoprotein metabolic ratios is to improve therapy selection. Cholesterol-lowering drugs primarily act on metabolic processes and the functioning of tissues, and affect plasma lipoprotein particle and cholesterol concentrations only in second instance. Different drugs affect processes in a different way and with different efficacy. For instance, statins block HMG CoA reductase, reducing cholesterol production in the liver, and leading to increased cholesterol uptake from the plasma (63). On the other hand fibrates activate transcription factors called peroxisome proliferator-activated receptors (PPARs).(64,65) This activation results in increased extrahepatic lipolysis of VLDL particles, increased removal of LDL particles by the liver and increased HDL production.(66) A diagnostic that gives more information on tissue function is likely to improve prediction of the response to therapy, and can also help to select the optimal therapy for the patient.
         If information about the actual fluxes proves to be important in addition to the flux ratios, possibilities exist to supplement the lipoprotein profile data with VLDL production data. For this purpose, a method called ‘Intralipid’ was developed (67). This method allows a determination of the VLDL production rate without the use of stable isotopes. Using this measurement in addition to the uptake / production ratio and the lipolysis / production ratio in VLDL, would allow full quantification of the production, lipolysis and uptake fluxes in the VLDL range. Nevertheless, this procedure requires infusion of lipids into the patient’s bloodstream and taking multiple blood samples, and so has the disadvantage of being more labor intensive than measuring a lipoprotein profile from a single blood sample. On the other hand, the combined approach has the advantage of being less labor intensive than stable isotope methodologies.
         Developing a model to analyze lipoprotein profiles and derive model-based diagnostics also faces specific technical challenges. We will discuss some of these challenges in continuation.

  

Why develop a modeling framework and subsequent model implementations within that framework?

As indicated above, compartmental modeling is the standard model type used for modeling the lipoprotein metabolism measured in isotopic tracer experiments. This model type is not suitable for the analysis of lipoprotein profiles, for the following reasons. First, since detailed lipoprotein profiles contain much information regarding cholesterol, triglyceride, and particle concentrations at different sizes, a model designed to analyze these profiles should be able to utilize the particle size information effectively. A promising way to utilize this information is to specify metabolic processes as a continuous function of particle size, which compartmental models have not been designed for. Second, different measurement methods subdivide the lipoprotein size spectrum in different subclasses. A standard compartmental model would need to be different for each of these measurement methods, if it defines one compartment per measured subclass. This is an unwanted situation because of the extra work involved, and because results obtained with the different models cannot be compared directly. Therefore, we need to have a new modeling methodology that can specify lipoprotein metabolic processes as a continuous function of particle size, and that can deal with all possible lipoprotein profile measurement methods.
         The two mentioned requirements for the new modeling methodology can be met by a class model. In this model type, particles in a given size range are put together in a single ‘class’, that can span any size range the user specifies. For a detailed calculation, classes as small as 0.01 nm intervals can be used, leading to a total of approximately 500 classes for a complete profile of apoB-containing lipoproteins. Instead of specifying the kinetics for single classes, which would be impossible for such a large number of small classes, each of the processes (production, lipolysis and uptake) is specified as a continuous function of particle size. Since the lipoprotein concentration is always the result of a balance between production, lipolysis, and uptake processes, the model can calculate the particle concentration in each class from the process specifications. A ‘modeling framework’ needs to be constructed for constructing this type of model. Various model implementations within the framework then correspond to different ideas about how the lipoprotein particle size relates to the underlying processes, but the modeling methodology remains the same in all these implementations.
         Once such a modeling framework and a first implementation are constructed, that model implementation needs to be validated. For the validation process, detailed data are necessary. We will examine the available data for validation in continuation.

What datasets are available for validating the model and the model-based diagnostic?

Detailed data about how the kinetics of lipoproteins depends on particle size is indispensible for validating the projected model of lipoprotein kinetics. Fortunately, a sustained research effort in the area of lipoprotein biology has made much material available to base the model on.

Stable-isotope flux studies

A large number of studies on lipoprotein metabolism have been carried out using tracer techniques. These studies include patients with a variety of genetic defects or polymorphisms, metabolic or other diseases, and different drug therapies. Since the model needs detailed data for its validation, only studies in which production, lipolysis and uptake processes were analyzed in the VLDL1, VLDL2, IDL and LDL classes are suitable for model validation. Table 1.1 gives an overview of the studies of this type that have been published, and in which data for individuals were reported. These studies provide a wealth of information to base a model on.

Table 1: list of lipoprotein kinetics studies in which VLDL1, VLDL2, IDL and LDL kinetics has been determined and reported for individual patients.

First Author
Year
Studied in relation to ApoB kinetics
Reference
Bilz
2004
Atorvastatin and fenofibrate in mixed hyperlipidemia
(68)
Demant
1988
Hepatic lipase deficiency
(22)
Demant
1991
Apolipoprotein E polymorphisms
(69)
Demant
1993
Familial hyperchylomicronemia
(17)
Demant
1996
Normolipidemic subjects
(70)
Demant
1998
Nephrotic patients
(71)
Forster
2002
Atorvastatin and simvastatin in moderate combined hyperlipidemia
(63)
Gaffney
2002
Familial hypercholesterolemia and familial defective apoB
(72)
Gaw
1993
Simvastatin
(73)
Gaw
1996
Colestipol alone and in combination with simvastatin
(74)
Lundahl
2006
Microsomal triglyceride transfer protein -493T variant
(75)
Nierman
2005
LPLS447X Carriers
(76)
Packard
1993
Thyroid replacement therapy in hypothyroidism
(77)
Packard
1995
Normal and hyperlipemic subjects
(50)
Packard
2000
VLDL and LDL subfraction pattern
(78)
Schmitz
2001
HIV-infected patients with antiretroviral combination therapy
(79)


GOLDN study

In order to test whether the model is able to deal with lipoprotein profile data and able to derive relevant information from these profiles, population studies in which lipoprotein profiles have been measured are important. Fortunately, several such studies exist, among which the Framingham Heart Study is the first and most well-known (1,80,81).[3] The Framingham Offspring cohort contains 3066 subjects who had their lipoprotein profiles measured (4). Another somewhat smaller study is the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study,[4] in which family members from three-generational pedigrees were re-recruited from two centers of the ongoing NHLBI Family Heart Study (Salt Lake City and Minneapolis). The population in this study can be described as genetically homogeneous (predominantly white), and encompassing a wide distribution of lipid phenotypes. To be included in the study, participants needed to have a fasting triglyceride level less than 1500 mg/dL. The population is highly interrelated, since the 1200 subjects in the present sample represent only 165 different families.
         All subjects in the GOLDN study underwent a 3-week treatment with a daily dose of 160 mg fenofibrate. Before and after the intervention, lipid phenotypes were measured using both biochemical measurements and NMR spectroscopy, after fasting as well as postprandially after a dietary fat challenge. This study, therefore, contains ample lipoprotein data and is interesting for examining what model-based diagnostics can contribute to understanding the effect of a large-scale drug intervention in a broad ranging population.

TNO nutritional intervention study

Next to medicinal interventions, also dietary interventions are known to affect lipoprotein metabolism. For instance, fatty acids of different chain lengths are thought to take different routes through the body to their final destination. Long-chain fatty acids are transported directly from the intestine via the lymph and possibly the circulation into the fat tissue (82). The transport occurs as triglycerides in chylomicrons. Medium-chain fatty acids are thought to first enter the liver via the portal circulation as free fatty acids (probably bound to albumin), before being transported to extrahepatic tissue as triglycerides in lipoproteins (82). Medium-chain fatty acids are, therefore, expected to affect lipoprotein metabolism in a different way than long-chain fatty acids.
         At TNO, human study 7261 was carried out, which included a dietary intervention with spreads containing different fatty acid chain lengths. The primary objective of this study was to examine the effect of body fat distribution on the physiological response to this dietary fat intervention. The two treatments were Medium Chain Fatty Acids (C8:0 and C10:0) and Poly Unsaturated Fatty Acids (C18:2n-6), administered on in a spread during two study periods of three weeks with a six-week washout period in between. With respect to lipoprotein kinetics, the apoB production in the VLDL class was studied using stable isotope tracers, and lipoprotein profiles were measured using HPLC technology (83). This study provides an opportunity to examine how dietary fatty acids and body fat distribution affect lipoprotein profiles and lipoprotein kinetics.

Framingham Heart Study Offspring Cohort

The Framingham Heart Study started in 1948, it is one of the most well known cohort studies investigating cardiovascular risk. In 1971, the study enrolled a second generation – 5124 of the original participants’ adult children and their spouses: the Framingham Offspring cohort (FOS).[5] All participants underwent extensive physical examinations and lifestyle interviews that the researchers later analyze for common patterns related to CVD development. In the fourth of the examinations that the subjects underwent at approximately 3-year intervals, lipoprotein profiles were measured using NMR methodology (4). Over the subsequent years, cardiovascular events were recorded. This study therefore allows relating lipoprotein profile-derived diagnostic markers to cardiovascular risk.

Questions about this research project


What is the purpose of the research project reported in this PhD thesis?

Since lipoprotein profiles contain information with diagnostic potential that is hard to disclose, and since mathematical modeling is a suitable methodology for interpreting detailed datasets, the aim of the research reported in this PhD thesis is to develop a computational model to aid the interpretation of detailed lipoprotein measurements. The new computational model should derive information about the metabolic status of lipoprotein metabolism from a single lipoprotein profile measurement in terms of ratios of metabolic production, lipolysis and uptake processes. Furthermore, the research presented here aims to show that these ratios are capable of distinguishing between subjects with various types of genetic variations in key genes affecting lipoprotein metabolism, and between patients with different types and intensities of metabolic disorders. Finally, the research aims to investigate whether lipoprotein metabolism indicators derived from the model improve cardiovascular risk prediction.

Why is this purpose relevant for society?

We anticipate that developing a computational model to aid interpretation will increase the diagnostic power of detailed lipoprotein profiles. The increased diagnostic power can be employed to give each patient the correct treatment. Therefore, such analysis potentially has added value for doctors, since they can profit from more detailed diagnostic information. Also, biotech companies offering lipoprotein profile measurements can profit from additional biological interpretation of their measurements, which adds value to their diagnostic. Most importantly, patients will benefit from more accurate diagnosis through a more personalized treatment advice.

What parties have collaborated to make this research possible?

This PhD thesis arises from collaboration between the Netherlands Bioinformatics Centre (NBIC), who financed the research, the Leiden-Amsterdam Centre for Drug Research (LACDR) at the University of Leiden, where the author was employed, and the TNO Quality of Life research institute, Department of Biosciences, where most of the work was done.

Outline of this thesis

Next to this introduction, this PhD thesis contains the following chapters. The first part, containing chapter 2 and 3 mainly concerns model development. Chapter 2 describes the development of a mathematical modeling framework within which different diagnostic models based on lipoprotein profiles can be developed, and a first validation of a model in this framework using data from a stable-isotope flux study (78). Chapter 3 describes a further development of the model, and the development of a new lipoprotein model-based candidate diagnostic and its validation using a range of stable-isotope flux studies. The second part, containing chapter 4, 5, and 6, concerns model applications. Chapter 4 contains the first application of the model-based candidate diagnostic using HPLC-measured lipoprotein profiles, in the context of a nutritional intervention study conducted at TNO. Chapter 5 contains an exploration of lipoprotein profile data from a population study (the GOLDN study), and examines how subgroups of patients, identified via K-means clustering based on lipoprotein profiles, differ in their lipid response to a fenofibrate intervention. It also examines how the model-based candidate diagnostic behaves in these subgroups. Chapter 6 contains the study that shows that lipoprotein metabolism indicators may have predictive value for cardiovascular risk on top of classical risk markers. Chapter 7 contains a methodological reflection on the development of computational models for clinical diagnostic use based on clinical chemistry data. Finally, chapter 8 contains a general discussion of the research presented in this PhD thesis and suggestions for further research.


[5] See www.framinghamheartstudy.org 

 Back to PhD thesis

 1.    Wilson, P. W. F., W. P. Castelli, and W. B. Kannel. 1987. Coronary risk prediction in adults (the Framingham Heart Study). Am J Cardiol 59:91G-94G.
2.    Conroy, R. M., K. Pyörälä, A. P. Fitzgerald, S. Sans, A. Menotti, G. De Backer, D. De Bacquer, P. Ducimetière, P. Jousilahti, U. Keil, I. Njolstad, R. G. Oganov, T. Thomsen, H. Tunstall-Pedoe, A. Tverdal, H. Wedel, P. Whincup, L. Witheimsen, and I. M. Graham. 2003. Estimation of ten-year risk of fatal cardiovascular disease in Europe: The SCORE project. Eur Heart J 24:987-1003.
3.    Assmann, G., P. Cullen, and H. Schulte. 2002. Simple Scoring Scheme for Calculating the Risk of Acute Coronary Events Based on the 10-Year Follow-Up of the Prospective Cardiovascular Munster (PROCAM) Study. Circulation 105:310-315.
4.    Cromwell, W. C., J. D. Otvos, M. J. Keyes, M. J. Pencina, L. Sullivan, R. S. Vasan, P. W. F. Wilson, and R. B. D'Agostino. 2007. LDL particle number and risk of future cardiovascular disease in the Framingham Offspring Study-Implications for LDL management. J Clin Lipidol 1:583-592.
5.    Otvos, J. D., E. J. Jeyarajah, D. W. Bennett, and R. M. Krauss. 1992. Development of a proton nuclear magnetic resonance spectroscopic method for determining plasma lipoprotein concentrations and subspecies distributions from a single, rapid measurement. Clin Chem 38:1632-1638.
6.    Usui, S., Y. Hara, S. Hosaki, and M. Okazaki. 2002. A new on-line dual enzymatic method for simultaneous quantification of cholesterol and triglycerides in lipoproteins by HPLC. J Lipid Res 43:805-814.
7.    Caulfield, M. P., S. Li, G. Lee, P. J. Blanche, W. A. Salameh, W. H. Benner, R. E. Reitz, and R. M. Krauss. 2008. Direct Determination of Lipoprotein Particle Sizes and Concentrations by Ion Mobility Analysis. Clin Chem 54:1307-1316.
8.    Caulfield, M. P., S. Li, G. Lee, P. A. Blanche, W. A. Salameh, W. H. Benner, R. E. Reitz, and R. M. Krauss. 2008. In Reply. Clin Chem 54:2088-2089.
9.    Rambaldi, D. C., A. Zattoni, S. Casolari, P. Reschiglian, D. Roessner, and C. Johann. 2007. An Analytical Method for Size and Shape Characterization of Blood Lipoproteins. Clin Chem 53:2026-2029.
10.   Beswick, A., and P. Brindle. 2006. Risk scoring in the assessment of cardiovascular risk. Curr Opin Lipidol 17:375-386.
11.   Walley, T., P. Folino-Gallo, U. Schwabe, and E. van Ganse. 2004. Variations and increase in use of statins across Europe: data from administrative databases. BMJ 328:385-386.
12.   Mills, E. J., B. Rachlis, P. Wu, P. J. Devereaux, P. Arora, and D. Perri. 2008. Primary Prevention of Cardiovascular Mortality and Events With Statin Treatments: A Network Meta-Analysis Involving More Than 65,000 Patients. J Am Coll Cardiol 52:1769-1781.
13.   Brugts, J. J., T. Yetgin, S. E. Hoeks, A. M. Gotto, J. Shepherd, R. G. J. Westendorp, A. J. M. de Craen, R. H. Knopp, H. Nakamura, P. Ridker, R. van Domburg, and J. W. Deckers. 2009. The benefits of statins in people without established cardiovascular disease but with cardiovascular risk factors: meta-analysis of randomised controlled trials. BMJ 338:b2376-b2383.
14.   Austin, M. A., M. C. King, K. M. Vranizan, and R. M. Krauss. 1990. Atherogenic lipoprotein phenotype. A proposed genetic marker for coronary heart disease risk. Circulation 82:495-506.
15.   Noto, H. M. 2009. Latest Insights Into High-Density Lipoprotein Functions. Endocrinologist 19:179-186.
16.   Beck C. 2008. Assembly and Secretion of Atherogenic Lipoproteins. PhD thesis. University of Gothenburg, Göteborg, Sweden
17.   Demant, T., A. Gaw, G. F. Watts, P. Durrington, B. Buckley, C. W. Imrie, C. Wilson, C. J. Packard, and J. Shepherd. 1993. Metabolism of apoB-100-containing lipoproteins in familial hyperchylomicronemia. J Lipid Res 34:147-156.
18.   Young, S. G., B. S. Davies, L. G. Fong, P. Gin, M. M. Weinstein, A. Bensadoun, and A. P. Beigneux. 2007. GPIHBP1: an endothelial cell molecule important for the lipolytic processing of chylomicrons. Curr Opin Lipidol 18:389-396.
19.   de Beer, F., W. L. Hendriks, L. C. van Vark, S. W. A. Kamerling, K. W. van Dijk, M. H. Hofker, A. H. M. Smelt, and L. M. Havekes. 1999. Binding of beta-VLDL to Heparan Sulfate Proteoglycans Requires Lipoprotein Lipase, Whereas ApoE Only Modulates Binding Affinity. Arterioscler Thromb Vasc Biol 19:633-637.
20.   Adiels M. 2004. Compartmental Models of Lipoprotein Kinetics. PhD Thesis, Chalmers University of Technology, Göteborg, Sweden.
21.   Voshol, P. J., P. C. Rensen, K. W. Van Dijk, J. A. Romijn, and L. M. Havekes. 2009. Effect of plasma triglyceride metabolism on lipid storage in adipose tissue: Studies using genetically engineered mouse models. Biochim Biophys Acta-doi:10.1016/j.bbalip.2008.12.015.
22.   Demant, T., L. A. Carlson, L. Holmquist, F. Karpe, P. Nilsson-Ehle, C. J. Packard, and J. Shepherd. 1988. Lipoprotein metabolism in hepatic lipase deficiency: studies on the turnover of apolipoprotein B and on the effect of hepatic lipase on high density lipoprotein. J Lipid Res 29:1603-1611.
23.   Jaye, M., K. J. Lynch, J. Krawiec, D. Marchadier, C. Maugeais, K. Doan, V. South, D. Amin, M. Perrone, and D. J. Rader. 1999. A novel endothelial-derived lipase that modulates HDL metabolism. Nat Genet 21:424-428.
24.   Hirata, K., H. n. L. Dichek, J. A. Cioffi, S. Y. Choi, N. J. Leeper, L. Quintana, G. S. Kronmal, A. D. Cooper, and T. Quertermous. 1999. Cloning of a Unique Lipase from Endothelial Cells Extends the Lipase Gene Family. J Biol Chem 274:14170-14175.
25.   Lamarche, B., and M. E. Paradis. 2007. Endothelial lipase and the metabolic syndrome. Curr Opin Lipidol 18:298-303.
26.   Morton, R. E., and D. B. Zilversmit. 1983. Inter-relationship of lipids transferred by the lipid-transfer protein isolated from human lipoprotein-deficient plasma. J Biol Chem 258:11751-11757.
27.   Yen, F. T., R. J. Deckelbaum, C. J. Mann, Y. L. Marcel, R. W. Milne, and A. R. Tall. 1989. Inhibition of cholesteryl ester transfer protein activity by monoclonal antibody. Effects on cholesteryl ester formation and neutral lipid mass transfer in human plasma. J Clin Invest 83:2018-2024.
28.   Millar, J. S., M. E. Brousseau, M. R. Diffenderfer, P. H. Barrett, F. K. Welty, A. Faruqi, M. L. Wolfe, C. Nartsupha, A. G. Digenio, J. P. Mancuso, G. G. Dolnikowski, E. J. Schaefer, and D. J. Rader. 2006. Effects of the cholesteryl ester transfer protein inhibitor torcetrapib on apolipoprotein B100 metabolism in humans. Arterioscler Thromb Vasc Biol 26:1350-1356.
29.   Krieger, M. 1994. Structures and Functions of Multiligand Lipoprotein Receptors: Macrophage Scavenger Receptors and LDL Receptor-Related Protein (LRP). Annu Rev Biochem 63:601-637.
30.   MacArthur, J. M., J. R. Bishop, K. I. Stanford, L. C. Wang, A. Bensadoun, J. L. Witztum, and J. D. Esko. 2007. Liver heparan sulfate proteoglycans mediate clearance of triglyceride-rich lipoproteins independently of LDL receptor family members. J Clin Invest 117:153-164.
31.   Van Eck, M., M. Hoekstra, R. Out, I. S. Bos, J. K. Kruijt, R. B. Hildebrand, and T. J. van Berkel. 2008. Scavenger receptor BI facilitates the metabolism of VLDL lipoproteins in vivo. J Lipid Res. 49:136-146.
32.   Gofman, J. W., and F. LINDGREN. 1950. The role of lipids and lipoproteins in atherosclerosis. Science 111:166-171.
33.   Lindgren, F. T., H. A. ELLIOTT, and J. W. Gofman. 1951. The ultracentrifugal characterization and isolation of human blood lipids and lipoproteins, with applications to the study of atherosclerosis. J Phys. Colloid Chem 55:80-93.
34.   Gofman, J. W., F. T. Lindgren, and H. ELLIOTT. 1949. Ultracentrifugal studies of lipoproteins of human serum. J Biol Chem 179:973-979.
35.   Havel, R. J., H. A. Eder, and J. H. BRAGDON. 1955. The distribution and chemical composition of ultracentrifugally separated lipoproteins in human serum. J. Clin. Invest 34:1345-1353.
36.   GOFMAN, J. W., M. A. R. T. HANIG, H. B. JONES, M. A. LAUFFER, E. Y. LAWRY, L. A. LEWIS, G. V. MANN, F. E. MOORE, F. R. E. D. OLMSTED, J. F. YEAGER, E. C. ANDRUS, J. H. BARACH, J. W. BEAMS, J. W. FERTIG, J. W. GOFMAN, M. A. LAUFFER, I. H. PAGE, J. A. SHANNON, F. J. STARE, and P. D. WHITE. 1956. Evaluation of Serum Lipoprotein and Cholesterol Measurements as Predictors of Clinical Complications of Atherosclerosis: Report of a Cooperative Study of Lipoproteins and Atherosclerosis. Circulation 14:689-741.
37.   BURSTEIN, M., and J. SAMAILLE. 1958. Determination of serum beta-lipoproteins after selective precipitation by heparin. Presse Med 66:974-975.
38.   Friedewald, W. T., R. I. Levy, and D. S. Fredrickson. 1972. Estimation of the Concentration of Low-Density Lipoprotein Cholesterol in Plasma, Without Use of the Preparative Ultracentrifuge. Clin Chem 18:499-502.
39.   Alaupovic, P. 1991. Apolipoprotein composition as the basis for classifying plasma lipoproteins: Characterization of ApoA- and ApoB-containing lipoprotein families. Progr Lipid Res 30:105-138.
40.   Alaupovic, P. 1996. Significance of apolipoproteins for structure, function, and classification of plasma lipoproteins. Method Enzymol 263:32-60.
41.   Walldius, G., A. H. Aastveit, and I. Jungner. 2006. Stroke mortality and the apoB/apoA-I ratio: Results of the AMORIS prospective study. J Intern Med 259:259-266.
42.   Kaess, B., M. Fischer, A. Baessler, K. Stark, F. Huber, W. Kremer, H. R. Kalbitzer, H. Schunkert, G. Riegger, and C. Hengstenberg. 2008. The lipoprotein subfraction profile: heritability and identification of quantitative trait loci. J Lipid Res 49:715-723.
43.   Kulkarni, K. R. 2006. Cholesterol Profile Measurement by Vertical Auto Profile Method. Clin Lab Med 26:787-802.
44.   Grundy, S. M. 2004. Richard Havel, Howard Eder, and the evolution of lipoprotein analysis. J Clin Invest 114:1034-1037.
45.   Barrett, P. H. R., D. C. Chan, and G. F. Watts. 2006. Design and analysis of lipoprotein tracer kinetics studies in humans. J Lipid Res 47:1607-1619.
46.   Parhofer, K. G., and P. H. Barrett. 2006. What we have learned about VLDL and LDL metabolism from human kinetic studies. J Lipid Res 47:1620-1630.
47.   Phair, R. D., M. G. Hammond, J. A. Bowden, M. Fried, W. R. Fisher, and M. Berman. 1975. Preliminary model for human lipoprotein metabolism in hyperlipoproteinemia. Fed Proc 34:2263-2270.
48.   Beltz, W. F., Y. A. Kesaniemi, B. V. Howard, and S. M. Grundy. 1985. Development of an integrated model for analysis of the kinetics of apolipoprotein B in plasma very low density lipoproteins, intermediate density lipoproteins, and low density lipoproteins. J Clin Invest 76:575-585.
49.   Barrett, P. H., N. Baker, and P. J. Nestel. 1991. Model development to describe the heterogeneous kinetics of apolipoprotein B and triglyceride in hypertriglyceridemic subjects. J Lipid Res 32:743-762.
50.   Packard, C. J., A. Gaw, T. Demant, and J. Shepherd. 1995. Development and application of a multicompartmental model to study very low density lipoprotein subfraction metabolism. J Lipid Res 36:172-187.
51.   Adiels, M., C. Packard, M. J. Caslake, P. Stewart, A. Soro, J. Westerbacka, B. Wennberg, S. O. Olofsson, M. R. Taskinen, and J. Borén. 2005. A new combined multicompartmental model for apolipoprotein B-100 and triglyceride metabolism in VLDL subfractions. J Lipid Res 46:58-67.
52.   Gitlin, D., D. G. Cornwell, D. Nakasato, J. L. Oncley, W. L. Hughes, and C. A. Janeway. 1958. Studies on the metabolism of plasma lipoproteins in the necrotic syndrome. II. The lipoproteins. J Clin Invest 37:720-729.
53.   Bilheimer, D. W., S. Eisenberg, and R. I. Levy. 1972. The metabolism of very low density lipoprotein proteins I. Preliminary in vitro and in vivo observations. BBA: Lipids & Lipid Metab 260:212-221.
54.   Cryer, D. R., T. Matsushima, and J. B. Marsh. 1986. Direct measurement of apolipoprotein B synthesis in human very low density lipoprotein using stable isotopes and mass spectrometry. J Lipid Res 27:508-516.
55.   Lichtenstein, A. H., J. S. Cohn, D. L. Hachey, J. S. Millar, J. M. Ordovas, and E. J. Schaefer. 1990. Comparison of deuterated leucine, valine, and lysine in the measurement of human apolipoprotein A-I and B-100 kinetics. J Lipid Res 31:1693-1701.
56.   Fisher, W. R., V. Venkatakrishnan, E. S. Fisher, P. W. Stacpoole, and L. A. Zech. 1997. The 3H-leucine tracer: Its use in kinetic studies of plasma lipoproteins. Metab Clin Exp 46:333-342.
57.   Musunuru, K., M. Orho-Melander, M. P. Caulfield, S. Li, W. A. Salameh, R. E. Reitz, G. Berglund, B. Hedblad, G. Engstrom, P. T. Williams, S. Kathiresan, O. Melander, and R. M. Krauss. 2009. Ion mobility analysis of lipoprotein subfractions identifies three independent axes of cardiovascular risk. Arterioscler Thromb Vasc Biol 29:1975-1980.
58.   Akobeng, A. K. 2007. Understanding diagnostic tests 1: Sensitivity, specificity and predictive values. Acta Paediatr 96:338-341.
59.   Panteleev, A., N. M. Ananyeva, F. I. Ataullakhanov, and E. L. Saenko. 2007. Mathematical models of blood coagulation and platelet adhesion: Clinical applications. Curr Pharm Des 13:1457-1467.
60.   Mari, A., G. Pacini, E. Murphy, B. Ludvik, and J. J. Nolan. 2001. A Model-Based Method for Assessing Insulin Sensitivity From the Oral Glucose Tolerance Test. Diab Care 24:539-548.
61.   Hubner, K., T. Schwager, K. Winkler, J. G. Reich, and H. G. Holzhutter. 2008. Computational lipidology: predicting lipoprotein density profiles in human blood plasma. PLoS Comput Biol 4:e1000079-doi:10.1371/journal.pcbi.1000079.
62.   Rizzo, M., and K. Berneis. 2006. Should we measure routinely the LDL peak particle size? Int J Cardiol 107:166-170.
63.   Forster, L. F., G. Stewart, D. Bedford, J. P. Stewart, E. Rogers, J. Shepherd, C. J. Packard, and M. J. Caslake. 2002. Influence of atorvastatin and simvastatin on apolipoprotein B metabolism in moderate combined hyperlipidemic subjects with low VLDL and LDL fractional clearance rates. Atherosclerosis 164:129-145.
64.   Issemann, I., and S. Green. 1990. Activation of a member of the steroid hormone receptor superfamily by peroxisome proliferators. Nature 347:645-650.
65.   Schoonjans, K., B. Staels, and J. Auwerx. 1996. The peroxisome proliferator activated receptors (PPARs) and their effects on lipid metabolism and adipocyte differentiation. BBA 1302:93-109.
66.   Staels, B., J. Dallongeville, J. Auwerx, K. Schoonjans, E. Leitersdorf, and J. C. Fruchart. 1998. Mechanism of Action of Fibrates on Lipid and Lipoprotein Metabolism. Circulation 98:2088-2093.
67.   Al-Shayji, I. A. R., J. M. R. Gill, J. Cooney, S. Siddiqui, and M. J. Caslake. 2007. Development of a novel method to determine very low density lipoprotein kinetics. J Lipid Res 48:2086-2095.
68.   Bilz, S., S. Wagner, M. Schmitz, A. Bedynek, U. Keller, and T. Demant. 2004. Effects of atorvastatin versus fenofibrate on apoB-100 and apoA-I kinetics in mixed hyperlipidemia. J Lipid Res 45:174-185.
69.   Demant, T., D. Bedford, C. J. Packard, and J. Shepherd. 1991. Influence of apolipoprotein E polymorphism on apolipoprotein B-100 metabolism in normolipemic subjects. J Clin Invest 88:1490-1501.
70.   Demant, T., C. J. Packard, H. Demmelmair, P. Stewart, A. Bedynek, D. Bedford, D. Seidel, and J. Shepherd. 1996. Sensitive methods to study human apolipoprotein B metabolism using stable isotope-labeled amino acids. Am J Physiol 270:E1022-E1036.
71.   Demant, T., C. Mathes, K. Gutlich, A. Bedynek, H. B. Steinhauer, T. Bosch, C. J. Packard, and G. L. Warwick. 1998. A simultaneous study of the metabolism of apolipoprotein B and albumin in nephrotic patients. Kidney Int 54:2064-2080.
72.   Gaffney, D., L. Forster, M. J. Caslake, D. Bedford, J. P. Stewart, G. Stewart, G. Wieringa, M. Dominiczak, J. P. Miller, and C. J. Packard. 2002. Comparison of apolipoprotein B metabolism in familial defective apolipoprotein B and heterogeneous familial hypercholesterolemia. Atherosclerosis 162:33-43.
73.   Gaw, A., C. J. Packard, E. F. Murray, G. M. Lindsay, B. A. Griffin, M. J. Caslake, B. D. Vallance, A. R. Lorimer, and J. Shepherd. 1993. Effects of simvastatin on apoB metabolism and LDL subfraction distribution. Arterioscler Thromb 13:170-189.
74.   Gaw, A., C. J. Packard, G. M. Lindsay, E. F. Murray, B. A. Griffin, M. J. Caslake, I. Colquhoun, D. J. Wheatley, A. R. Lorimer, and J. Shepherd. 1996. Effects of colestipol alone and in combination with simvastatin on apolipoprotein B metabolism. Arterioscler Thromb Vasc Biol 16:236-249.
75.   Lundahl, B., C. Skoglund-Andersson, M. Caslake, D. Bedford, P. Stewart, A. Hamsten, C. J. Packard, and F. Karpe. 2006. Microsomal triglyceride transfer protein -493T variant reduces IDL plus LDL apoB production and the plasma concentration of large LDL particles. Am J Physiol Endocrinol Metab 290:E739-E745.
76.   Nierman, M. C., B. H. C. M. Prinsen, J. Rip, R. J. Veldman, J. A. Kuivenhoven, J. J. P. Kastelein, M. de Sain-van der Velden, and E. S. G. Stroes. 2005. Enhanced Conversion of Triglyceride-Rich Lipoproteins and Increased Low-Density Lipoprotein Removal in LPLS447X Carriers. Arterioscler Thromb Vasc Biol 25:2410-2415.
77.   Packard, C. J., J. Shepherd, G. M. Lindsay, A. Gaw, and M. R. Taskinen. 1993. Thyroid replacement therapy and its influence on postheparin plasma lipases and apolipoprotein-B metabolism in hypothyroidism. J Clin Endocrinol Metab 76:1209-1216.
78.   Packard, C. J., T. Demant, J. P. Stewart, D. Bedford, M. J. Caslake, G. Schwertfeger, A. Bedynek, J. Shepherd, and D. Seidel. 2000. Apolipoprotein B metabolism and the distribution of VLDL and LDL subfractions. J Lipid Res 41:305-318.
79.   Schmitz, M., G. M. Michl, R. Walli, J. Bogner, A. Bedynek, D. Seidel, F. D. Goebel, and T. Demant. 2001. Alterations of apolipoprotein B metabolism in HIV-infected patients with antiretroviral combination therapy. J Acquir Immune Defic Syndr 26:225-235.
80.   Kannel, W. B., W. P. Castelli, T. Gordon, and P. M. McNamara. 1971. Serum cholesterol, lipoproteins, and the risk of coronary heart disease. The Framingham study. Ann Int Med 74:1-12.
81.   Freedman, D. S., J. D. Otvos, E. J. Jeyarajah, I. Shalaurova, A. Cupples, H. Parise, R. B. D'Agostino, P. W. F. Wilson, and E. J. Schaefer. 2004. Sex and age differences in lipoprotein subclasses measured by nuclear magnetic resonance spectroscopy: The Framingham study. Clin Chem 50:1189-1200.
82.   Metges, C. C., and G. Wolfram. 1991. Medium and Long Chain Triglycerides Labeled with 13C: A Comparison of Oxidation after Oral or Parenteral Administration in Humans. J Nutr 121:31-36.
83.   Pasman WJ, de Graaf AA. 2009. The effect of body fat distribution on the physiological response to a dietary fat intervention. Rep. 7261, TNO, Zeist, the Netherlands.

 Back to PhD thesis

No comments:

Post a Comment

Subscribe to: Posts (Atom)