Quantifying apoprotein synthesis in rodents: coupling LC-MS/MS analyses with the administration of labeled water.

Stable isotope tracer studies of apoprotein flux in rodent models present difficulties as they require working with small volumes of plasma. We demonstrate the ability to measure apoprotein flux by administering either 2H- or 18O-labeled water to mice and then subjecting samples to LC-MS/MS analyses; we were able to simultaneously determine the labeling of several proteolytic peptides representing multiple apoproteins. Consistent with relative differences reported in the literature regarding apoprotein flux in humans, we found that the fractional synthetic rate of apoB is greater than apoA1 in mice. In addition, the method is suitable for quantifying acute changes in protein flux: we observed a stimulation of apoB production in mice following an intravenous injection of Intralipid and a decrease in apoB production in mice treated with an inhibitor of microsomal triglyceride transfer protein. In summary, we demonstrate a high-throughput method for studying apoprotein kinetics in rodent models. Although notable differences exist between lipoprotein profiles that are observed in rodents and humans, we expect that the method reported here has merit in studies of dyslipidemia as i) rodent models can be used to probe target engagement in cases where one aims to modulate apoprotein production and ii) the approach should be adaptable to studies in humans.

H-leucine) and measured its incorporation into various apoproteins, including apoB100 and apoA1 ( 1,2 ). Although useful, the methods are not well suited for application in rodent models. For example, sample preparation methods often require multiple centrifugation steps to separate different lipoprotein fractions, followed by electrophoretic separation of the apoproteins; this typically requires a minimum of several hundred microliters of plasma ( 3 ). In addition, investigators determine the fl ux rates by collecting multiple samples from a given subject and fi tting the labeling curves; clearly this is not practical in studies conducted in rodent models in which sample volumes are often a limiting factor (e.g., mice have ‫ف‬ 1 ml of blood).
In addition to the concerns noted above regarding the analytical challenges that are associated with studies of lipoprotein kinetics in small animals, one needs to consider the route of administering the tracer. The administration of a labeled amino acid typically requires intravenous access; i.e., tracers are either given as a primed infusion or a bolus injection ( 1,2 ). Although it is feasible to catheterize small animals, the ability to circumvent this step has obvious advantages. Because labeled water (either 2 H or 18 O) can be given via an intraperitoneal injection, virtually no surgical expertise is needed to administer the tracer ( 4,5 ). In addition, as the t 1/2 of water is relatively slow in rodents (e.g., ‫ف‬ 2 days for 2 H in a mouse), it is possible to maintain a constant labeling of the precursor pool for several hours by administering a single bolus of either water tracer ( 6 ), and this can be extended by allowing the animals access to enriched drinking water ( 4,7 ).
Recent studies have demonstrated the ability to quantify protein synthesis using labeled water in rodents and humans (8)(9)(10)(11)(12)(13). In this report, we have examined the use of an LC-MS/MS-based method for quantifying the turnover of plasma apolipoproteins in mice given either 2

H 2 O or
Abstract Stable isotope tracer studies of apoprotein fl ux in rodent models present diffi culties as they require working with small volumes of plasma. We demonstrate the ability to measure apoprotein fl ux by administering either 2 H-or 18 O-labeled water to mice and then subjecting samples to LC-MS/MS analyses; we were able to simultaneously determine the labeling of several proteolytic peptides representing multiple apoproteins. Consistent with relative differences reported in the literature regarding apoprotein fl ux in humans, we found that the fractional synthetic rate of apoB is greater than apoA1 in mice. In addition, the method is suitable for quantifying acute changes in protein fl ux: we observed a stimulation of apoB production in mice following an intravenous injection of Intralipid and a decrease in apoB production in mice treated with an inhibitor of microsomal triglyceride transfer protein. In summary, we demonstrate a high-throughput method for studying apoprotein kinetics in rodent models. Although notable differences exist between lipoprotein profi les that are observed in rodents and humans, we expect that the method reported here has merit in studies of dyslipidemia as i ) rodent models can be used to probe target engagement in cases where one aims to modulate apoprotein production and ii ) the approach should be adaptable to studies in humans. In vivo turnover studies have contributed to our understanding of the physiology and pathophysiology of apolipoprotein metabolism. Traditionally, investigators have acetone was eluted at ‫ف‬ 1.4 min , and the mass spectrometer performed selected ion monitoring of m/z 58 and 59 (10 ms dwell time per ion) in the electron impact ionization mode. The 18 O-labeling of plasma water was determined using GC-MS as described by Brunengraber et al. ( 6 ). Briefl y, 5 µl of plasma was reacted with PCl 5 to generate phosphoric acid, and 150 µl of TMSdiazomethane (Sigma) was then added to generate the trimethylphosphate. The solution was evaporated under a stream of nitrogen and then dissolved in 150 µl chloroform. Samples were analyzed using an Agilent 5973 MS coupled to a 6890 GC oven fi tted with an Agilent DB5-MS column (30 m × 250 µm × 0.15 µm), the oven was initially set at 100°C then programmed to increase at 35°C per min to 250°C, and helium carrier fl ow was set at 1.0 ml × min Ϫ 1 (2 µl of sample was injected using a 40:1 split), trimethylphosphate was eluted at ‫ف‬ 1.9 min , the mass spectrometer was set to perform selected ion monitoring of m/z 140 and 142 (10 ms dwell time per ion) in the electron impact ionization mode.
Protein isolation and analyses. Measurements for protein kinetics and protein quantifi cation were performed separately as described below. ( Table 1 contains a list of peptide sequences and transitions.) For kinetic measurements, 40 µl of plasma was fi rst run over an immunodepletion column to remove albumin, immunoglobulin G, and transferrin. The fl ow-through from the immunodepletion was concentrated using a 5 kDa molecular weight cutoff device, and then it was denatured, reduced, and alkylated prior to tryptic digestion. To identify candidate peptides for the labeling measurements, the resulting digests were analyzed on a Thermo Orbitrap mass spectrometer coupled with Waters nanoAcquity UPLC using a data-dependent acquisition method. The gradient was 98% A (0.1% formic acid in water) / 2% B (0.1% formic acid in acetonitrile) ramped to 80% A at 16 min, 10% A at 19 min, 98% A at 19.8 min. The column was a Waters Symmetry NanoEase column, 3.5 µm, 300 µm × 50 mm maintained at 50°C, and the fl ow rate was 10 µl/min. The acquired LC-MS/MS data fi les were imported into Elucidator Data Analysis Suite, and the tandem mass spectra were searched against mouse protein database using SEQUEST. Proteins were identifi ed using the following fi lter criteria: 1+, Xcorr above 1.8; 2+, Xcorr above 2.2; 3+, Xcorr above 3.75; and 4+, Xcorr above 3.75; delta Cn above 0.1. Blast searches (NCBI) were performed to ensure that peptide sequences were unique to the target proteins. Isotope labeling was then determined using multiple reaction monitoring of M0, M1, and M2 of the product ions of a given peptide using a Thermo TSQ Quantum triple quadrupole mass spectrometer coupled with Waters nanoAcquity UPLC; the separation condition was identical to that used with the Orbitrap. The mass spectrometric acquisition parameters were as follows: Q1 = 0.7, Q3 = 0.7, scan width = 0.5, scan time = 150 ms, collision energy = 18 V, and S-lens = 120.
For protein quantifi cation measurements, 4 µl of plasma was diluted with 136 µl ammonium bicarbonate (pH 8.0), and a known amount of a stable isotope-labeled peptide standard O. Attention was focused on quantifying the isotope labeling and measuring the abundance of multiple proteolytic peptides, which, in total, yields an estimate of absolute rates of apoprotein fl ux in vivo. Acute labeling studies. Male C57BL/6J mice ( ‫ف‬ 12 weeks old) were fed a standard diet ad libitum. Animals were given an intraperitoneal injection of 2 H 2 O (20 µl × g Ϫ 1 of body weight) and then euthanized at various points up to 6 h after tracer administration. Plasma samples were frozen until analyses. In a second experiment, mice were studied under conditions that were expected to perturb apoB fl ux; i.e., control versus Intralipid challenge (50 µl Intralipid 20 per mouse, to stimulate apoB production) versus an inhibitor of microsomal triglyceride transfer protein (MTPi, Pfi zer compound CP-346086, at 50 mg per kg, to inhibit apoB production) ( 14 ). In those studies, we tested both 2 H 2 O-and H 2 18 O-labeled water in the respective groups (20 µl of either 2 H 2 O or H 2 18 O was administered × g Ϫ 1 of body weight), mice were euthanized 2 h later, and plasma samples were frozen until analyses. The injection of Intralipid was given at the same time as the water tracers (i.e., 2 h before the samples were collected), and the MTPi was given 1 h before the water tracers.

Analytical studies
Water labeling. The 2 H-labeling of plasma water was determined using GC-MS as described by Shah et al. ( 15 ). Briefl y, 2 H present in water was exchanged with hydrogen bound to acetone by incubating 10 µl of plasma or known standards in a 2 ml glass screw-top GC vial at room temperature for 4 h with 2 µl 10N NaOH (Fisher Scientifi c) and 5 µl of acetone (Sigma-Aldrich). The instrument was programmed to inject 5 µl of headspace gas from the GC vial in a splitless mode. Samples were analyzed using a 2.0 min isothermal run [Agilent 5973 MS coupled to a 6890 GC oven fi tted with an Agilent DB-5MS column (30 m × 250 µm × 0.15 µm), oven set at 170°C, and helium carrier fl ow set at 1.0 ml × min

RESULTS
Because measurements of protein synthesis require reliable quantitation of the isotopic labeling of a given analyte, a critical fi rst step in examining the LC-MS/MS approach centered on determining the precision of the assay(s). For example, Fig. 2 demonstrates the effect of peak intensity on the apparent natural isotopic labeling ratio of the peptide ARPALEDLR (an apoA1-derived tryptic peptide). Although it is clear that signals with lower intensity are generally less reproducible, the coeffi cient of variation in either the M1/M0 ratio or the M2/M0 ratio can by minimized by increasing the signal intensity. Note that here we have considered a worst-case scenario, as these tests were run by spiking known amounts of AR-PALEDLR into plasma digests. Although the precision can be affected by multiple variables, such as coeluting peptides, electronic drifts, etc., similar observations were made when analyzing pure standards (unpublished data), suggesting that the noise in the assay can be minimized by injecting more material. Figure 3 demonstrates the ability to simultaneously quantify changes in the isotopic labeling of multiple apoproteins following immunodepletion and tryptic digestion of plasma. As is shown, we were able to measure the incorporation of 2 H and 18 O into peptides derived from apoE, apoB, apoA4, apoC3, apoA2, and apoA1. Regardless of whether mice were given 2 H 2 O or H 2 18 O, we CA) was spiked into plasma; 10 µl 10% sodium deoxycholate was added prior to reduction, alkylation, and tryptic digestion ( 16 ). These conditions allowed for the complete digestion of apoB; therefore, the apoB peptide concentration refl ected the apoB protein concentration (apoB peptide concentration was calculated using peak area ratio between the sample and internal standard). Samples were analyzed using identical conditions described above for TSQ analyses.

Calculations
In mice given either 2 H 2 O or H 2 18 O, excess labeling is calculated by normalizing the M1 or M2 isotopomer of a peptide against the M0 isotopomer [i.e., the isotopically substituted species (singly or doubly, respectively) versus the monoisotopic species] and then subtracting the mean background ratio(s) observed in control mice (i.e., mice that did not receive any isotope) ( 17 ). In long-term studies, the fractional synthesis rate (FSR) is calculated from the exponential increase in protein labeling using equation 1: O is used, we model the change in the abundance of M+2-labeled molecules.
In short-term studies (e.g., following an acute perturbation with Intralipid or MTPi), the FSR is calculated using equation 2: where product labeling 2hour represents the total labeling of a proteolytic peptide 2 h postinjection of the respective tracers and n represents the number of copies of the precursor that are incorporated (17)(18)(19)(20). For the apoB peptide, we assumed n = 13 and n = 9 for 2 H and 18 O, respectively. As noted, the precursor labeling is infl uenced by the amino acid composition of the peptide and the equilibration of isotope in the respective amino acids ( Fig. 1 ). Although one can estimate the precursor labeling from the asymptotic (or steady-state) labeling of a protein that is obtained from a long-term study, in cases where 2 H 2 O is used acutely, it is diffi cult to predict the maximum labeling because 2 H is incorporated into various carbon-bound positions of amino acids ( 21 ). Therefore, one requires knowledge regarding the equilibration constants between hydrogen in water and peptide-bound amino acids. We estimated the precursor labeling for 2 H from previous investigations ( 21,22 ). However, when H 2 18 O is used, we expect that one 18 O atom will be stably bound per peptide bond and additional 18 O atoms for certain amino acid side chains (e.g., glutamine or glutamate) ( 23-26 ). We estimated the precursor labeling for 18 O by counting the number of peptide bond oxygens and the side chain oxygens ( 7 ). A second approach to estimating n is to run a long-term tracer study and experimentally determine the steady-state labeling. The asymptotic labeling refl ects the labeling of the water and n , and as we measured the water labeling, we could calculate n ( 5,9,27 ).
To estimate the absolute fl ux, one needs to measure the pool size. In studies of apoB metabolism, we estimated its concentration as described above. Therefore, total apoB fl ux was determined by multiplying equation 2 by the concentration of total apoB.

Statistics
Comparisons were either made using a two-tailed t -test (assuming equal variance) or ANOVA with Tukey's posthoc testing. Fig. 1. Generation of labeled amino acids in the presence of labeled water. In cases where 2 H 2 O is administered, the labeling of amino acids largely depends on the equilibration with keto-acids and/or de novo synthesis, which can lead to different levels of incorporation depending on whether the amino acid is essential (e.g., leucine) or nonessential (e.g., alanine). In cases where H 2 18 O is administered, one expects an instantaneous labeling of amino acids during proteolysis; 18 O can be incorporated into some side chain oxygens (e.g., glutamate), which may be back-exchanged during sample preparation. In addition, one expects fairly uniform labeling in cases in which H 2 18 O is administered, as the route of labeling is similar for most amino acids versus amino aciddependent labeling. In cases where 2 H 2 O is administered, the latter occurs because amino acids contain different numbers of carbonbound hydrogens and each is subject to differential labeling. Note that 2 H and D are used interchangeably in the fi gure.
not impact our fi ndings (as noted, the labeling profi les of apoE у apoB > apoA1, etc., in both of the tracer groups); however, this observation implies that studies that require a long-term administration of the tracer(s) should adjust the design(s) to ensure a steady-state water labeling ( 7 ) and/or account for changes in the water labeling in the modeling ( 28 ). As the labeling of most peptides approached steady-state labeling, it was possible to experimentally estimate the n for the different tracers ( O water was administered, the y axis represents the excess labeling in the M2 form of the peptide (solid circles). The legend contains the tryptic peptide used in the analyses and identifi es the fragment ion that is used to determine the labeling; note that the labeling of the fi rst several amino acids in each sequence (shown in parenthesis) was not determined via the transitions that were monitored (see Table 1 ). mating the kinetic constants (it simply requires that one have knowledge of the expected labeling), studies that use 2 H 2 O should consider peptides that predominantly contain nonessential amino acids to maximize the shift in the labeling.
On the basis of the data shown in Fig. 3 , we next designed a pilot experiment to determine the initial labeling profi les of the various apoproteins. For example, Foster et al. ( 29 ) have elegantly described the effects of different assumptions when modeling kinetic data. Although the data shown in Fig. 3 are less than ideal for determining the early changes in labeling, we found that the incorporation of 2 H in some apoproteins approaches a steady state within a few hours postinjection of the tracer, whereas the labeling of others remains in a pseudo steady state ( Fig. 4 ). Note that the time-dependent changes for most proteins are reasonably stable; however, in a few instances, there appears to be an outlying data point (e.g., the 1.5 h sample for apoA1 and the 2 h sample for apoA2). This is not unexpected as the aim of this pilot was to establish the general temporal response; therefore, we euthanized only one animal per time point. On the basis of these observations, we determined that studies of apoB fl ux, for example, should be run for ‫ف‬ 2 h. This would ensure that measurements are made during a period in which the change in protein labeling is pseudo-linear and minimizes the error when interpreting data using samples collected at a single time point (see equation 2) ( 29 ). Using the data contained in Fig. 4 , it was possible to estimate the t 1/2 for the various proteins; these values are ‫ف‬ 1.1 h for apoE, ‫ف‬ 1.4 h for apoB, ‫ف‬ 3.8 h for apoA2, ‫ف‬ 10.1 h for apoA4, ‫ف‬ 16.2 h for apoC3, and ‫ف‬ 27.7 h for apoA1. Note that the values for apoE, apoB, and apoA2 were determined by fi tting the labeling curves using equation 1, whereas the values reported well with previously published values for the labeling of individual amino acids ( 7,21,22 ).
The data shown in Fig. 3 emphasize another important concept regarding the incorporation of labeled water into a protein. For example, the labeling of apoE is rather striking in that there is a major difference between the steadystate enrichment in mice given 2 H 2 O versus those given H 2 18 O. This is expected given the amino acid sequence and the nature of how the different isotopes are incorporated. In cases where H 2 18 O is administered, one expects a more uniform incorporation of the label (e.g., primarily in the peptide bonds), whereas when 2 H 2 O is given, one expects the asymptotic labeling to be strongly affected by the presence of several glycine and alanine residues (each containing multiple 2 H from body water). Although the variable labeling does not constitute a problem when esti-  apoA1  29  11  34  11  apoB  12  9  13  9  apoC3  26  12  29  12  apoA2  16  8  18  7  apoA4  19  9  16  7  apoE  27  9  26  8 Theoretical values for the number of 2 H or 18 O were estimated using published data (see Refs. 7 ,21 ,and 22 ), whereas the experimentally determined values were derived by fi tting the data obtained from the chronic labeling study (i.e., those contained in Fig. 3 ), the asymptotic labeling of a given peptide, and the water labeling used for that purpose. In most cases, there is reasonable agreement between the different values. one can purchase catheterized animals for studies, many academic centers support mouse metabolic phenotyping activities, including several National Institutes of Health (NIH)-funded sites that are expected to serve as core facilities for the general scientifi c community (see www. mmpc.org). Although it is now somewhat easy to execute studies that require several hours of a tracer infusion, the ability to circumvent this step would allow for higher throughput studies by minimizing the cost/time associated with implanting catheters and limiting the need to ship mice to specialized labs.
We and others have demonstrated the ability to study protein synthesis by administering labeled water ( 7-9, 12, 31, 32 ). For example, because subjects will generate 2 H-or 18 O-labeled amino acids in the presence of either 2 H-or 18 O-labeled water, rates of protein synthesis can be determined by measuring the incorporation of the respective 2 H-or 18 O-labeled amino acids into a protein of interest ( Fig. 1 ). In cases where 2 H 2 O is administered, one expects that the amino acid composition of the protein of interest for apoA4, apoC3, and apoA1 were calculated using equation 2 and using the experimentally determined n ( Table  2 ) in the denominator.
Because the control of apoB production is affected by various factors, we designed an experiment to determine whether our method could detect increases and/or decreases in apoB production. Table 3 contains the requisite end points for estimating protein fl ux using a single sample; i.e., equation 2 requires the precursor labeling (water), the product labeling (tryptic-derived peptide), and the concentration. An additional parameter that is required is n , which we derived using previously published data ( 7,21 ). The data contained in Table 3 and Fig. 5 demonstrate that our approach has the ability to detect acute changes in protein fl ux. For example, the administration of an intravenous bolus of Intralipid led to a stimulation of apoB production, whereas treatment with MTPi led to a decrease in apoB production. These observations were consistently observed using 2 H 2 O and H 2 18 O.

DISCUSSION
Historically, most studies of lipoprotein kinetics were performed in humans and/or large animals. Although the use of stable isotope tracers has been limited in small animals (e.g., mice) ( 30 ), we have demonstrated an approach that should enable routine investigations in such models. The ability to translate knowledge regarding physiology across models may not be obvious in all cases. For example, although there are well-described differences in enzyme activities and lipoprotein profi les between rodents and humans, rodents can be used to address certain questions, especially those regarding the ability to modulate a target. Two central challenges that required attention included i ) how to implement a simple tracer protocol and ii ) how to measure changes in isotopic labeling of a protein. We consider the novelty of our approach in the context of those challenges.
First, considerable progress has been made regarding working with small animals. In addition to the fact that  Table 3 . To express the results as a percentage changes, the data obtained in an individual animal was compared with the average in the respective control groups for a given water tracer (average ± SEM, fi ve mice per group). All measurements were made 2 h postinjection of the tracers (data shown as average ± SEM, fi ve mice per group). Rates of apoB production were determined using Equation 2 (see Materials and Methods); n = 13 and n = 9 in cases where 2  or MTPi ( Table 3 and Fig. 5 ), it is necessary to consider how the absolute rate of protein synthesis depends on knowledge of the time-course of change in protein concentration. To this point, our interpretation has neglected any effect on protein clearance (or degradation), and this is especially important in mice given a bolus of Intralipid ( 35,36 ). For example, based on the increase in the concentration of apoB ( Table 3 ), it is clear that the Intralipid challenge led to an imbalance between production and removal; the concentration increased by ‫ف‬ 145 nM in 2 h (i.e., ‫ف‬ 330 nM in control versus ‫ف‬ 475 nM in Intralipid, regardless of whether animals were given 2 H 2 O or H 2 18 O). Assuming, as we have done, that the concentration rapidly changed to a new steady state suggests that ‫ف‬ 30 nM of protein was newly made (i.e., ‫ف‬ 50 nM in controls versus ‫ف‬ 80 nM in Intralipid, regardless of the tracer). From this simple calculation, it is obvious that there is a substantial gap between the change in concentration and the apparent change in the absolute rate of production, ‫ف‬ 145 nM and ‫ف‬ 30 nM, respectively. As we did not collect samples to describe the exact shape of the curve, it is not possible to make fi rm statements regarding explicit changes to the absolute rate of protein synthesis and/or degradation (e.g., did the concentration reach a new steady state within the fi rst 15 min of the Intralipid bolus, or did the concentration rise steadily over the 2 h?) We believe that in the absence of data which demonstrate the time-dependent change in apoB concentration, it is better to interpret the absolute protein fl ux results with caution. It is clear that Intralipid stimulated the fractional rate of synthesis, which presumably translated into a stimulation of the absolute rate of protein synthesis; however, the exact magnitude of the change in the absolute rate is not clear. In total, our observations suggest that it is possible to determine the acute impact of a perturbation on protein synthesis and protein breakdown provided that one collect the appropriate samples to defi nitively interpret the data, and we expect that the methods reported here can be used for such purposes.
Finally, it is of interest to consider the effect of timing on the studies reported here and differences between our design and many articles in the literature. Investigators often collect numerous samples from a given subject to determine protein fl ux. For example, during a typical human study, one may obtain more than 10 samples per subject ( 1,2 ). Clearly, this is not practical in small animals and may not be necessary in humans either ( 29 ). As we have demonstrated in rodent models, it is possible to use a single sample to estimate protein fl ux. Although this is appealing, especially as it minimizes the number of samples that are required for analyses, one must carefully choose when to collect a sample. As outlined by Foster et al. ( 29 ), the sampling interval can have substantial consequences on the apparent turnover rate of a given protein. Although we have demonstrated the ability to study apoB fl ux by sampling ‫ف‬ 2 h posttracer injection, studies focused on proteins with faster or slower half-lives would likely require additional consideration. The consequences are that one may underestimate the t 1/2 if a protein turns over rapidly (or proteolytic peptide) will affect the degree of labeling as individual amino acids reach different steady-state labeling. However, as we recently demonstrated ( 21 ), it appears that one can reliably interpret the data because in the presence of 2 H 2 O, the rate of 2 H-labeling of amino acids is relatively fast. Although one expects that exposure to H 2 18 O will also rapidly label amino acids, it is presumed that H 2 18 O will lead to more uniform labeling of proteolytic peptides, especially when considering the mechanism(s) by which 18 O labels amino acids ( Fig. 1 ) ( 23-26 ). 18 O is incorporated into amino acids during proteolytic cleavage, amino acid activation (not shown in the scheme), and de novo synthesis. In total, this should result in a rapid and stable 18 O-labeling of free amino acids. As we previously discussed, the degree of labeling that is observed in a given protein depends on i ) the amount of labeled water that is administered, ii ) the amino acid composition of a given peptide (or fragment ion), and iii ) the synthetic rate ( 21 ). In cases where 2 H 2 O or H 2 18 O are administered, one typically expects to observe the incorporation of multiple copies of the precursor in a product; i.e., because the protein acts as a biopolymer of the precursor, the labeling of the protein can exceed that of the precursor ( 18 ). This is most obvious in Figs. 3 and 4 , in which the labeling of apoE reaches ‫ف‬ 20% excess 2 H and yet the body water labeling is ‫ف‬ 2.5% excess 2 H. The apoE peptide that we have monitored contains several amino acids that carry more than one 2 H from body water (e.g., glycine ‫ف‬ 2, alanine ‫ف‬ 4, etc.) ( Table 2 ). As apoE is known to display a reasonably short t 1/2 , one expects that shifts in its labeling will be the most apparent under these conditions, in contrast to apoA1, which typically has a slower t 1/2 ( 1,33 ). In cases where one obtains a time course of labeling, it is possible to use equation 1 and neglect n (i.e., the number of labeled sites); however, in short-term studies and/or cases in which a single sample is obtained, it is necessary to account for n , so equation 2 is required (18)(19)(20).
Note that there is a discrepancy in the literature regarding apoB kinetics: it is not immediately clear whether the liver directly makes VLDL-apoB and LDL-apoB or whether LDL-apoB is made via delipidation ( 34 ). Our analysis is rather simplistic and circumvents these concerns; as noted, we are quantifying total apoB fl ux. In cases where Intralipid is acutely administered via intravenous injection, one might expect a sizeable increase in the concentration of circulating lipids. The dose of Intralipid administered in this study delivered ‫ف‬ 10 mg of lipid, yet the endogenous triglyceride pool was on the order of ‫ف‬ 1.5 mg. Presumably, a fraction of the Intralipid bolus was cleared by the liver, repackaged into VLDL particles, and then exported; therefore, the stimulation of apoB production that we observed in mice treated with Intralipid is consistent with what one might expect. Likewise, the administration of a known inhibitor of triglyceride secretion (Pfi zer compound CP-346086) led to a marked decrease in the production of apoB, again consistent with what one might expect ( Table 3 and Fig. 5 ).
Although we were intrigued to fi nd signifi cant changes in the FSRs of apoB following the administration of Intralipid and samples are collected later ( 29 ) and that the analytical window is compromised when measuring the labeling of a protein with a long t 1/2 if it is sampled too early. In addition, it is often necessary to add time delays in the modeling to account for packaging of newly made proteins before they are secreted ( 29 ). Although we recognize that we neglected to include such a delay, we do not believe that it has a serious impact on our data. For example, although the rates of apoB production may not refl ect the true rates of apoB production, we were able to detect the expected directional changes in apoB production ( Fig. 5 ).
In summary, we have demonstrated a simple and robust method for quantifying apoprotein fl ux in small animals. This method is well suited for high-throughput studies in model systems and can be used to evaluate the effi cacy of novel compounds capable of modulating specifi c targets (e.g., apoB production in mice). The fact that virtually identical results were obtained when using 2 H 2 O and H 2 18 O is of importance. Namely, considering Fig. 1 , H 2

18
O should rapidly label all amino acids. In contrast, 2 H is incorporated during amino acid metabolism, which could presumably be slower than the 18 O-labeling of amino acids. However, our data suggest that the 2 H-labeling of amino acids is rapid, and these observations are consistent with data that we recently obtained regarding the equilibration of 2 H-labeling in mice following the administration of O appear to yield comparable results regarding protein synthesis, 2 H 2 O offers advantages in cases where one also aims to examine lipid fl ux ( 37,38 ). As both types of labeled water can be given to humans ( 10,39 ), we suspect that it is possible to translate aspects of this work to clinical investigations.