Measurement of protein turnover rates by heavy water labeling of nonessential amino acids

1760 (2006) 730–744http://www.elsevier.com/locate/bba

Biochimica et Biophysica Acta

Measurement of protein turnover rates by heavy water labelingof nonessential amino acids

Robert Busch a,1, Yoo-Kyeong Kim b,1, Richard A. Neese b, Valerie Schade-Serin b,Michelle Collins b, Mohamad Awada a, James L. Gardner a, Carine Beysen a, Michael E. Marino a,

Lisa M. Misell a, Marc K. Hellerstein b,c,⁎

a KineMed, Inc., Emeryville, CA 94608, USAb Department of Nutritional Sciences and Toxicology, University of California at Berkeley, CA 94720, USA

c Department of Medicine, San Francisco General Hospital, University of California, San Francisco, CA 94110, USA

Received 25 August 2005; received in revised form 23 November 2005; accepted 21 December 2005Available online 24 January 2006

Abstract

In vivo measurements of protein synthesis using isotope-labeled amino acids (AAs) are hampered by the heterogeneity of AA pools and, forslow turnover proteins, the difficulty and expense of long-term labeling. Continuous oral heavy water (2H2O) labeling can safely maintain stablebody water 2H enrichments for weeks or months. 2H is metabolically incorporated into C–H bonds of nonessential AAs (NEAAs) and hence intoproteins. No posttranslational label exchange occurs, so 2H incorporation into protein NEAAs, in principle, reports on protein synthesis. Here, weshow by mass isotopomer distribution analysis (MIDA) of 2H2O-labeled rodent tissue proteins that metabolic 2H flux into C–H bonds of Ala, Gly,or Gln used for protein synthesis is nearly complete. By 2H2O labeling of rodents, turnover of bone and muscle mixed proteins was quantified andstimulation of liver collagen synthesis by CCl4 was detected. Kinetics of several human serum proteins were also measured, reproducing publishedt1/2 estimates. Plateau enrichments in Ala varied among different proteins. Moderate amounts of protein, isolated chromatographically orelectrophoretically, sufficed for kinetic analyses. In conclusion, 2H2O labeling permits sensitive, quantitative, operationally simple measurementsof protein turnover in vivo by the rise-to-plateau approach, especially for proteins with slow constitutive turnover.© 2006 Elsevier B.V. All rights reserved.

Keywords: Protein synthesis/turnover; Stable isotope labeling; Deuterated water; Mass isotopomer distribution analysis; Gas chromatography/mass spectrometry

Abbreviations: 2H2O, heavy water; 2H, deuterium; MIDA, mass isotopomerdistribution analysis; tRNA, transfer ribonucleic acid; GC/MS, gas chromatog-raphy/mass spectrometry; NEAAs, nonessential amino acids; PMSF, phenyl-methylsulfonyl fluoride; EDTA, ethylenediamine tetraacetic acid; LDL, low-density-lipoprotein; PTFE, polytetrafluoroethylene; PFBBr, pentafluorobenzylbromide; PCI, positive chemical ionization; NCI, negative chemical ionization;SEC, size exclusion chromatography; ELISA, enzyme-linked immunosorbentassay; m/z, mass to charge ratio; AAs, amino acids; SDS-PAGE, sodiumdodecyl sulfate-polyacrylamide gel electrophoresis; PBS, Dulbecco's phosphatebuffered saline without divalent cations; ApoB, apolipoprotein B; Ig,immunoglobulin; BSA, bovine serum albumin; GCRC, General ClinicalResearch Center; PFB, pentafluorobenzyl N,N-di(pentafluorobenzyl); VLDL,very low density lipoprotein⁎ Corresponding author. Department of Nutritional Sciences and Toxicology,

University of California at Berkeley, CA 94720, USA. Tel.: +1 510 642 0646;fax: +1 510 642 0535.

E-mail address: [email protected] (M.K. Hellerstein).1 Contributed equally to this work.

0304-4165/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.bbagen.2005.12.023

1. Introduction

Protein synthesis and turnover are essential for life, and mucheffort has gone into quantifying the rates of these processes invarious biological settings by use of isotopic labels. Each of thecommonly used labeling approaches, however, has drawbacks,especially for applications in vivo. Radioisotopes are too hazardousfor routine use in humans, so non-radioactive, stable isotope labelsare preferable. Most commonly, deuterium- (2H-) or 13C-labeledessential amino acids (AAs) are used. Their incorporation intoproteins of interest, via protein synthesis, is tracked following bolusor continuous administration, using mass spectrometric (MS)analysis of protein hydrolyzates or peptide fragments [1].

A central challenge in any biosynthetic labeling experiment is todetermine or infer the amount of label in the precursor pool [2], i.e.,in the relevant AA-tRNAswhen protein synthesis is measured. Theintracellular pools of AAs and AA-tRNAs used for protein

731R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

synthesis often are not in isotopic equilibriumwith plasmaAA, anda complex subcellular organization exists that is usually inacces-sible to direct measurement [1–7]. One way to overcome thisproblem of the true precursor pool enrichment is to administer thelabeled AA precursor continuously over several half-lives (t1/2) ofthe protein of interest. Eventually, label enrichment in the protein-bound AA approaches a plateau that equals the enrichment in theprecursor pool [2]. This obviates the need to measure intracellularAA label content directly. In the case of protein biosyntheticlabeling, the plateau isotopic enrichment, at 100% replacement ofthe protein by newly synthesized molecules, reflects labeling of thetRNA-AA pool during the labeling period [2,6].

MS analysis of stable isotope incorporation into proteins alsopermits another approach for determining precursor enrichment.The label distributes into different mass variants (massisotopomers), which differ in mass, due to different isotopecontent, but not in chemical structure. The pattern of massisotopomers after a labeling experiment contains informationabout the precursor pool enrichment [2,8–11]. This informationcan be extracted by comparing the results of biosynthetic labelingexperiments to predictions of computational modeling by massisotopomer distribution analysis (MIDA) [2,8–11]. MIDA hasbeen used successfully in measurements of protein synthesis[1,5,7,9]. The combinatorial probability algorithms used forMIDA calculations have been reviewed in detail previously [9].

Particular difficulties in establishing the precursor/productrelationship arise when isotope-tagged AAs are used to labelslow-turnover proteins. In this setting, short-term labeling oftenresults in insufficient isotope incorporation for MIDA. Contin-uous labeling to plateau can sometimes be achieved with stableisotope-labeled essential AAs [1], but is usually too expensiveor impractical. We therefore looked for a novel biosyntheticlabeling approach for applying the rise-to-plateau principle toslow-turnover proteins.

Labeling with heavy water (2H2O) has been used safely fordecades as a tracer for physiologic studies of body waterturnover [12–14] and the synthesis and turnover of lipids[8,11,14–16]. More recently, synthesis rates of nucleic acidshave been measured with 2H2O [11,17–22]. No toxicities areobserved below about 20 mol% 2H in body water [13,14]. Atlower doses (≈1–2% 2H in body water), no adverse effects havebeen observed in humans labeled continuously for severalmonths [12,18,21]. Rodents have been kept at 3–20% 2H inbody water for months or years without adverse effects to thelabeled animals or their progeny [11,17,21].

The dynamics of 2H label in body water makes oral 2H2O aparticularly useful biosynthetic label. In humans given oral 2H2O,the 2H label equilibrates in water throughout all tissues within≈1 h, and when 2H2O administration is discontinued, the labeldecays with the half-life of body water (≈1 week) [23–25].Constant 2H levels in body water (precursor pool enrichment) cantherefore be achieved for extended periods through simple oraladministration protocols. From body water, 2H label rapidlyenters free nonessential AAs (NEAAs) through intermediarymetabolic pathways (Fig. 1A); moreover, all free AAs are labeledat their α carbon (C2) positions via transamination. The 2H-labeled AAs then are incorporated into newly synthesized

proteins via AA-tRNAs (Fig. 1B). Importantly, the enzymes ofintermediary metabolism through which 2H-label enters free AAsdo not act on AA residues after incorporation into the proteinbackbone. Unlike labile N–H or O–H bonds, C–H bonds do notexchange hydrogen spontaneously with body water. Thus, 2Hfrom 2H2O does not enter into C–H bonds of AA posttransla-tionally. These features of NEAA metabolism make 2H2O apromising label for measuring protein synthesis and turnover.

Here, we address the general applicability of 2H2O labelingfor measuring protein synthesis in vivo by the rise-to-plateauapproach.

2. Materials and methods

2.1. Animal studies

Sprague–Dawley rats (female, 7–8 weeks old, 200–250 g, Simonsen Inc.,Gilroy, CA) and C57Bl/6J male mice, 4–6 weeks old, 10–15 g, JacksonLaboratories, Bar Harbor, ME) were housed individually (rats) or in groups of 5(mice) in a specific pathogen-free facility with a 12 h light/12 h dark cycle,controlled temperature and humidity. Feeding (with Purina® rodent chow) andphysical activity were ad libitum. All studies received prior approval fromAnimal Care and Use Committees at UC Berkeley and KineMed, Inc.

Continuous 2H2O labeling of animals was initiated with one or two primingintraperitoneal bolus injections of 99.9% 2H2O (Cambridge Isotope Labs,Andover, MA) containing 0.9% w/v NaCl, which raised initial body water 2Henrichment to 2.5% or 5%. This enrichment was then maintained byadministration of 4% or 8% 2H2O in drinking water, given ad libitumthroughout the labeling period [10,21]. Rat pups were labeled in utero bystarting breeding pairs on 4% 2H2O in drinking water prior to mating; labeling ofdams was continued throughout the pregnancy until sacrifice, which wasperformed within 24 h of delivery. In order to track the decay of 2H in bodywater and proteins, animals were first labeled with 2H2O for 8–10 weeks asabove, then returned to ordinary drinking water and sacrificed at various timesthereafter. To analyze toxin-stimulated liver collagen synthesis, Sprague–Dawley rats were labeled with 8% 2H2O for 3 weeks, during which time theyreceived twice-weekly intraperitoneal injections of carbon tetrachloride (CCl4; 1ml/kg in olive oil vehicle) [26,27]. No adverse effects of 2H2O labeling wereobserved in any studies. For measurements of body water 2H enrichment, urinewas collected longitudinally from some animals; alternatively, blood wasobtained by cardiac puncture at sacrifice.

2.2. Human studies

Healthy volunteers were recruited by advertisement. Protocols received priorapproval from the Committee on Human Research at UC San Francisco andfrom the Committee for the Protection of Human Subjects at UC Berkeley.Written informed consent was given for all procedures.

In order to study VLDL apolipoprotein B (ApoB) turnover, two subjectswere labeled. One was a 53 year old obese man (112 kg, BMI 35.2; subject 1),the other was a 45-year-old hypercholesteremic woman without a history ofdiabetes or impaired glucose tolerance, who was on multiple medicationsincluding atorvastatin (91 kg; subject 2). Subjects were admitted overnight to theGeneral Clinical Research Center (GCRC) at San Francisco General Hospital,UC San Francisco, for a bolus labeling protocol. Subjects received four doses of70% 2H2O (36 ml each), at 20-min intervals, over the course of 60 min, withoutadverse effects. Heparinized plasma was collected every 30–60 min over a 10h period for measurements of body water 2H enrichment, and EDTA plasma wascollected hourly for serum protein isolation.

The long-term 2H2O administration protocol was performed as describedpreviously [11,18,21]. In brief, four healthy subjects entered the GCRC andwere given aliquots of 70% 2H2O (50–70 ml each) every 3–4 h for a total of 7–8 doses over 24 h. None of these subjects experienced any adverse effects. Overthe following 10 weeks, the subjects received aliquots of 70% 2H2O to drink athome (50–60 ml, once or twice a day), in order to achieve and maintain

Fig. 1. 2H labeling of protein AAs using 2H2O. (A) Biosynthetic pathways that introduce hydrogen atoms from water are shown for two NEAAs (Ala, Gly) and anessential AA (Leu). Abbreviations: TA, transaminase; PEP-CK, phosphoenolpyruvate carboxykinase; TCAC, tricarboxylic acid cycle; STHM, serine tetrahydrofolatemethyl transferase. Other reactions may also contribute. (B) Simplified scheme for 2H2O labeling of protein NEAAs. 2H introduced into body water by oral intake of2H2O is incorporated into NEAAvia reactions of intermediary metabolism (step 1; cf. A); labeled NEAA are then used to charge the appropriate tRNAs (step 2), whichare used for protein biosynthesis (step 3). Free NEAAs are regenerated by protein catabolism (step 4). Potential sources of label dilution and disequilibrium (not shownhere) are described in the Web supplement.

732 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

enrichments of about 1.0–2.0% 2H in body water. Saliva, urine, and bloodsamples were collected every 1–2 weeks for the 10 weeks of 2H2O intake. Noadverse effects or symptoms were reported during the outpatient protocol.

2.3. Materials

[2H]2-glycine and [2H]-alanine were purchased from Isotec (Miamisburg,OH). Other reagents, unless mentioned otherwise, were from Sigma (St. Louis,MO).

2.4. Protein isolation

In order to isolate mixed bone proteins, the rear left femur of 2H2O-labeled animals was obtained at sacrifice and dissected free of soft tissue.Bone marrow and trabecular bone were removed using a needle with sharpcutting surface. After washing 3 times with water, the bone was splintered

and powdered under liquid N2 in a Spex mill (Metuchen, NJ), delipidatedwith chloroform/methanol (1:1, v:v), and dried. For isolation of mixedmuscle proteins, labeled or de-labeled animals were sacrificed, and skeletalhindlimb muscle (quadriceps femoris) and heart were dissected and frozenin liquid N2.

Human sera, obtained by venipuncture, or animal sera, obtained bycardiac puncture, were stored at −20 °C. For purification of human IgG,thawed serum was spun (10,000×g, 20 min), diluted in 20 mM sodiumphosphate buffer, pH 7.0, 0.1% sodium azide, 3 mM PMSF, and 5 mMEDTA, and filtered (0.22 μm). Diluted sera (450 μl) were injected onto a 1-ml HiTrap Protein A HP column (Amersham Biosciences, Piscataway, NJ).After washing in 20 mM sodium phosphate buffer, pH 7.0, bound materialwas eluted using 100 mM sodium citrate buffer, pH 2.5. The methodcaptured IgG1, IgG2, and IgG4 quantitatively (not shown). The eluted peakwas concentrated using Centricon-30 ultrafiltration devices (Millipore,Billerica, MA) and further purified on a 300 mm×7.8 mm Phenomenex

Table 1Mass-to-charge (m/z) ratios for NEAA derivatives used

AA Derivative

N-acetyl, n-butylester (PCI)

N-propyl carboxyl,n-propyl ester (PCI)

PFB (NCI)

Ala 188–190 130–132 448–450Gly 174–176 116–118 434–436Gln/Glu 302–304 N.D. N.D.

PCI, positive chemical ionization; NCI, negative chemical ionization;m/z, mass-to-charge ratio; N.D., not done; PFB, pentafluorobenzyl-N,N-di(pentafluor-obenzyl) derivative.

733R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

(Torrance, CA) BioSep SEC-S 3000 size exclusion chromatography (SEC)column, equilibrated in 200 mM sodium phosphate buffer, pH 6.8, 150 mMNaCl. Fractions around the principal IgG1/2/4 peak (MW ca. 150,000) werescreened for contamination with unwanted, short-lived Ig isotypes bysandwich ELISA. Peak and tail SEC fractions containing ≤1% IgG3, b0.4%IgM, ≤3% IgA (i.e., near or at ELISA sensitivity limits) were processed forGC/MS analysis with indistinguishable results.

Human serum IgA1 was purified by affinity chromatography on Jacalin-sepharose (Pierce, Rockford, IL) with elution in 0.1 M melibiose in PBS. Freshresin was used for each patient, and carryover between samples from anindividual patient was b3%. Protein-containing fractions were pooled,concentrated, and resolved by SEC as described above, except that a BioSuite450 column (Waters, Milford, MA) was used. Fractions around the principalpeak, lacking detectable long-lived IgG isotypes and contaminated ≤1% withIgG3 as judged by ELISA, were processed for GC/MS analysis; no systematiceffect of fraction number on 2H enrichments was noted.

Human serum albumin was initially captured on HiTrap Blue HP columns(Amersham), with 50 mM potassium phosphate buffer, pH 7.0, as bindingbuffer, and eluted with 2 M KCl. Eluates were purified further by SEC.Alternatively, HiTrap Blue HP eluates were exchanged into 20 mM Tris–HCl,pH 7.5, and subjected to ion exchange chromatography on Resource Q columns(Amersham), with elution on a linear gradient of 0–1MNaCl. Peak and tail SECfractions, peak Resource Q fractions, and HiTrap Blue HP eluates all gaveindistinguishable 2H enrichments.

For isolation of rat albumin and IgG, serum was passed sequentially overHiTrap Blue and HiTrap protein G columns, using protocols similar to thosedescribed above for human proteins, and eluates were further purified by SEC.

Human low density lipoprotein (LDL)- and very low density lipoprotein(VLDL)-derived apolipoprotein B (ApoB) were isolated by ultracentrifugationand isopropanol precipitation as described previously [28], except that thesecond VLDL ultracentrifugation step was omitted.

Rat liver collagen was purified from 10 to 20 mg samples of fresh or frozenliver (J.G. et al., submitted). The tissue was homogenized in 0.5 ml 100 mMNaOH, using a Polytron (Norcross, GA) model PT 10–35 homogenizer, whichsolubilized most proteins but not collagen [29,30]. A collagen-containing pelletwas obtained by centrifugation at 7000×g for 10 min at 4 °C, washed once in 0.5ml H2O, and dissolved in 20 μl of SDS-PAGE sample buffer (Bio-Rad,Hercules, CA). After boiling for 3 min, the dissolved material was fractionatedby SDS-PAGE (10% acrylamide) and transferred onto PVDF membranes.Protein bands corresponding to monomeric and dimeric collagen werevisualized with Coomassie Brilliant Blue R250 and excised.

All chromatography steps were monitored by SDS-PAGE on 10%acrylamide or gradient gels under reducing conditions, followed by CoomassieBlue staining. All proteins, or bands excised from PVDF blots, were subjected tototal acid hydrolysis in 6 N HCl (110 °C, 24 h) essentially as described [28].Hydrolysates were dried under N2 gas or in vacuo prior to derivatization.

2.5. Sandwich ELISA for human IgG subclasses

Mouse monoclonal antibodies against human IgG1, IgG2, IgG4, IgM, IgA1/2 (all from Becton Dickinson), or IgG3 (Zymed) (10 μg/ml in 50 mM Tris–HCl,pH 9.0) were immobilized on Nunc (Rochester, NY) Maxisorp ELISA plates.Plates were blocked overnight at 4 °C using 3% (w/v) BSA in PBS, 0.05%Tween-20. Immunoglobulins were captured from sera or purified proteinfractions diluted in blocking buffer and detected using polyclonal rabbit anti-human IgG/IgM/IgA antibodies conjugated to horseradish peroxidase(1:15,000) using ABTS substrate. All binding steps were followed by extensivewashes in PBS, 0.05% Tween-20. Monoclonal IgG1, IgG2, IgG3, IgG4,polyclonal IgA from colostrum, and polyclonal serum IgM (the latter fromCalbiochem, San Diego, CA) were used to estimate sensitivity limits for eachisotype.

2.6. Measurement of 2H enrichment in body water

Aliquots of serum or urine (100 μl), placed into the caps of tightly sealed,inverted screw-capped vials, were kept at 60 °C in a heating block overnight. Inmost studies, the 2H enrichment of the condensate collected in the vial was

measured using GC/MS after conversion to tetrabromoethane, as described indetail elsewhere [8,17,21].

Low levels of 2H in body water of human subjects after bolusadministration were measured using isotope ratio mass spectrometry, usingan approach similar to that developed independently by Richelle et al. [31].Five aliquots (1 μl each) of water distilled from each serum sample wereinjected into a High Temperature Conversion/Elemental Analyzer, coupledto a MAT253 Isotope Ratio Mass Spectrometer via a Conflo-III interface(all from Thermo Finnigan, Bremen, Germany). The reactor containedglassy carbon granules for pyrolysis at 1450 °C, and helium (120 ml/min)was used as the carrier gas. Data analysis was performed using the FinniganMAT Isodat data system (version 2.0). The deuterium isotope abundancewas first calculated in δ2H values relative to the international VSMOWstandard and then converted to atomic percent excess using a calibrationcurve spanning the enrichment values of the blood samples. The first twoinjections of each sample were not considered, so as to minimize hysteresis(memory) effects from the previous sample, and the results from the lastthree injections were averaged.

2.7. Measurement of mass isotopomer abundances of AAs by GC/MS

The mass isotopomer abundances of AAs were determined by GC/MS usingone of two different derivatives for analysis in positive chemical ionization(PCI) mode, or a third derivative for analysis by negative chemical ionization(NCI; Table 1). GC retention times of all NEAA derivatives were established byuse of unlabeled standards, and the M0, M1, and M2 mass isotopomers wereanalyzed by selected ion monitoring, using a model 5973 mass spectrometerattached to a 6890 gas chromatograph (Agilent, Palo Alto, CA). Table 1 lists themass-to-charge ratios monitored for each derivative. Injection volumes wereadjusted to maintain abundances of each derivative within a range that allowedaccurate measurement of isotope abundances [9,22].

N-acetyl, n-butyl ester derivatives (prepared as described, [32]) or N-propylcarboxyl n-propyl ester derivatives (Easy:Fast kit, Phenomenex) wereused for PCI analysis. The N-acetyl, n-butyl NEAA esters were resolved ona DB225 GC column (15 m, 0.25 mm i.d., 0.25 μm film thickness, J andW Scientific, Folsom, CA) at 120–220 °C, with methane CI analysis. N-propylcarboxyl, n-propyl NEAA esters (1 μl) were injected in a splitlessmode onto a ZB-PAAC-MS column (Phenomenex, 10 m×0.25 mm) withthe injector port kept at 250 °C, and the transfer line was kept at 280 °C.The helium gas flow rate was adjusted to 1.0 ml/min, and the temperaturewas programmed, following an initial hold at 70 °C for 2 min, to rise from70 °C to 125 °C at 15 °C/min, then from 125 °C to 320 °C at 15 °C/min.

In order to prepare NEAA derivatives for NCI analysis, dried proteinhydrolysates were suspended in 1 ml of 50% acetonitrile, 50 mM K2HPO4,pH 11. Pentafluorobenzyl bromide (Pierce; 20 μl) was added, and the sealedmixture was incubated at 100 °C for 1 h. Ethyl acetate (0.6 ml) was added,and the solution was mixed. The top layer was transferred to a fresh tubeand dehydrated using Na2SO4. The resultant pentafluorobenzyl-N,N-di(pentafluorobenzyl)-NEAA derivatives (PFB derivatives) were resolved ona DB-225 column. The injector temperature was held at 220 °C. After 2min at 100 °C, the oven temperature was increased to 220 ° at a rate of 15°C/min. MS was performed in NCI mode with helium as the carrier gas andmethane as the reagent gas.

734 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

2.8. Calculations

MIDA calculations were performed according to algorithms that have beendescribed in detail elsewhere [9], as implemented in a program available fromthe authors on request. The application to 2H2O labeling of protein NEAA isdescribed in Results.

GC/MS data were used to determine, for derivatives of NEAA preparedfrom unlabeled NEAA standards and experimental samples, the relativeproportions of the M0, M1, and M2 mass isotopomers. Isotope incorporationinto the M1 and M2 mass isotopomers of 2H-labeled samples (EM1 and EM2,respectively, where E stands for excess over natural abundance) was calculatedby subtracting the corresponding fractional abundances of unlabeled standards.Measured mass isotopomer distributions depend slightly on the amount of theanalyte in each GC/MS sample [22]. To correct for this bias, baseline subtractionwas performed using standards that were abundance-matched to experimentalsamples, as described [22].

Fractional protein synthesis (f, the fraction of protein that was newlysynthesized during the labeling period) was determined for each sample as theratio of the measured EM1 to the calculated asymptotic or maximal value,EM1,max [9]. The asymptotic value was obtained from MIDA calculations inwhich p was taken to be equal to the measured body water 2H enrichment and nwas calculated from measured EM2/EM1 ratios by comparison to MIDA tables.Alternatively, the maximal integral values of nwere assumed (n=2 for Gly, 4 forAla, and 5 for Gln/Glu), and pwas calculated frommeasured EM2/EM1 ratios byMIDA. In experiments where accurate EM2 data were unavailable, EM1,max wascalculated by assuming maximal values of n and using measured body waterenrichments to represent p.

Fractional synthesis rate constants of proteins (k) were estimated by non-linear least squares fitting of plots of measured values of f or EM1 againstlabeling time (i.e., rise-to-plateau kinetics), assuming a single-exponential rise tothe asymptotic or maximal value:

ft ¼ fmax�ð1�e�ktÞ; or EM1;t¼EM1;max�ð1�e�ktÞ: ð1ÞFor de-labeling experiments, rate constants for label decay were estimated

by non-linear least squares fitting of plots of remaining enrichment in M1

(EM1,t) against de-labeling time, after wash-out of 2H from body water pools,assuming single-exponential decay:

EM1;t¼EM1;t¼0�e�kt : ð2ÞHalf-lives (t1/2) for protein synthesis or turnover were calculated as:

t¼0:693=k: ð3Þ

3. Results

3.1. Experimental approach

We wished to test whether continuous biosynthetic labelingwith oral 2H2O, measured by GC/MS analysis of 2Hincorporation into NEAA residues of proteins and byapplication of the rise-to-plateau kinetic approach, allowsreliable measurement of turnover rates of long-lived proteinsin vivo. The general experimental strategy consisted of thefollowing steps: (1) Oral administration of 2H2O to animals orhumans, to achieve stable 2H enrichments in body water; (2)isolation of proteins of interest at various time points afterinitiation of labeling; (3) total hydrolysis of proteins to freeAAs using 6 N HCl; (4) chemical derivatization of thereleased NEAAs for GC/MS analysis; (5) analysis of NEAAderivatives for 2H incorporation (label content and pattern),using GC/MS and MIDA calculations; and (6) application ofrise-to-plateau kinetic analysis.

3.2. Lack of nonenzymatic deuterium exchange into or out ofC–H bonds of NEAA derivatives

In order to analyze 2H incorporation into protein-boundNEAA, three different sets of derivatives were used (see Table 1and Materials and methods). An important methodologicalassumption was that there is no spontaneous 1H/2H exchangeinto or out of C–H bonds of NEAAs in vivo or during samplepreparation for GC/MS. Indeed, after incubation of an unlabeledprotein (human serum albumin) in 70% 2H2O for 24 h at roomtemperature and subsequent hydrolysis, no 2H incorporationwas observed in any of the derivatized AA (not shown).Moreover, when synthetic [2H]-alanine or [2H]2-glycine (bothlabeled at C2) was subjected to acid hydrolysis conditions, noloss of 2H label was observed (not shown). Thus, nonenzymaticdeuterium exchange at C–H bonds, in vivo or during sampleprocessing, does not confound this technique.

3.3. Stable body water 2H enrichments in 2H2O-labeledrodents

Continuous heavy water labeling in vivo requires that thelabeling protocol should achieve stable 2H enrichments in bodywater. As in previous studies [8,17,21], this was in fact achievedin all experiments described below (Web supplement, Fig. S1,and data not shown).

3.4. MIDA calculations for 2H2O labeling of protein NEAA

Important questions for application of heavy water labelingare how many C–H bonds of NEAA exchange with 2H in bodywater through intermediary metabolic reactions (Fig. 1A), andto what extent 2H-labeled NEAA are diluted in the trueprecursor pool for protein synthesis (NEAA-tRNA). Weaddressed these questions by comparing patterns of labelincorporation in rodents to the predictions of MIDA calcula-tions [9,10]. Like other biomolecules, NEAAs, even at naturalabundance, exist as mixtures of mass variants (called massisotopomers), due to the random presence of different isotopes(1H or 2H; 12C or 13C; 16O or 18O, etc.). For example, the PFBderivative of unlabeled Ala generates an [M-HF] fragment uponGC/MS analysis, with the formula C17H8O2NF10 and nominal(monoisotopic) MW of 448 Da (Table 1). Due to the naturalabundance of stable isotopes, only about 80% of this fragmentexists as the M0 (all-

1H/12C/16O etc.) mass isotopomer with m/z=448 Da; additional peaks of m/z=449 (M1) and m/z=450(M2) comprising the remaining 20% or so are present in themass spectrum, as calculated by MIDA (Fig. 2A–C) andverified to within ±0.2% by GC/MS analysis (not shown). M3

and heavier mass isotopomers are too rare for accuratemeasurement at natural abundance and are not used in thesestudies.

During a 2H2O labeling experiment, 2H atoms are incorpo-rated covalently into C–H bonds of free Ala through reactionsof intermediary metabolism (Fig. 1A), and hence, via Ala-tRNA, into Ala residues of newly synthesized proteins (Fig.1B). The replacement of 1H atoms by 2H shifts the mass

Fig. 2. MIDA calculations for the PFB derivative of Ala. (A–C) Mass isotopomer distributions as a function of varying n, the number of C–H bonds accessible to 2Hfrom 2H2O (A), p, the 2H enrichment in the precursor pool (B), or f, the fraction of Ala molecules newly synthesized (or incorporated into protein) during the labelingexperiment (C). The other parameters were held constant at the values indicated above each panel. The abscissa represents the mole fraction of the Ala PFB derivativespresent as each mass isotopomer (M0, M1, and M2). (D–F) The same M1 and M2 abundances are expressed as excess mole fractions (EM1 and EM2, respectively) aftersubtraction of baseline abundances (unlabeled, n=p= f=0), as a function of n (D), p (E), or f (F). (G) The EM2:EM1 ratio is shown as a function of p at different valuesof n, illustrating that both parameters influence the relative label incorporation into the M1 and M2 mass isotopomers.

735R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

isotopomer distributions of protein-bound Ala towards in-creased proportions of M1 and M2 at the expense of M0 (Fig.2A–C). MIDA calculations show that the shift depends on thenumber of C–H bonds at which the label can be incorporated(designated n; Fig. 2A and D), the proportion of labeled vs.unlabeled H-atoms in the precursor pool (i.e., the 2Henrichment, designated p; Fig. 2B and E), and the percentageof Ala that was newly synthesized (designated f; Fig. 2C and F).The excess abundance of M1 and M2 over baseline (EM1 andEM2, respectively; Fig. 2D–F) is a function of these threeparameters. Importantly, n and p uniquely determine the relativefrequency of single and double labeling events, and thus the

ratio of EM2 (reflecting double labeling) to EM1 (Fig. 2G). Theexact mass isotopomer distribution for each NEAA derivative(Table 1) depends on its elemental composition (Fig. 2; WebSupplement, Table S1; and data not shown).

Conversely, the measured EM2:EM1 ratios contain informa-tion about the metabolic pathways that introduce the label from2H2O into protein-bound Ala and can be used experimentally toconstrain n, p, or both. In principle, the value of n cannot exceedthe number of C–H bonds chemically present in the NEAAstudied (2 for Gly, 4 for Ala, 5 for Gln/Glu), but it can be less ifsome C–H bonds are metabolically inaccessible or partiallyaccessible to 2H2O. Similarly, the value of p, i.e., the effective

736 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

2H enrichment in 2H2O that feeds the precursor pool used forprotein synthesis, cannot exceed the 2H enrichment in bodywater, but it can be less if the label is diluted at any step duringprotein synthesis (Web Supplement, Fig. S2).

3.5. Labeling in utero and for prolonged periods in adultanimals

In order to characterize 2H labeling of protein NEAA from2H2O, we first tested the hypothesis that Ala and Gly residues ofproteins are labeled to the maximal possible extent. If all the C–H bonds are metabolically accessible to body water (n=numberof C–H bonds) and the label is fully equilibrated between theseC–H bonds and body water (p=body water enrichment), thenmeasured EM2/EM1 ratios will be consistent with thesetheoretical values of n and p, according to relationshipspredicted by MIDA (Fig. 2G, Table S1) [9].

Testing of these predictions is easiest if the all molecules areknown to have been newly synthesized during the period oflabel exposure (i.e., the fraction of labeled molecules, f, is100%). In order to create this situation experimentally, we firstlabeled rat pups in utero. A dam was maintained at 2.7–3% 2Hin body water through continuous oral 2H2O administration,starting before mating and continuing throughout pregnancy.Tissue proteins were obtained from 3 newborn pups, hydro-lyzed, derivatized, and analyzed for label incorporation into Alaand Gly derivatives by GC/MS (Table 2). The value of n wascalculated from the EM2/EM1 ratio, by assuming that themother's measured plasma 2H2O enrichment represents p. Bythis approach, values of n were close to 2 for Gly and 4 for Ala

Table 2Determination of p and n for glycine and alanine residues of fully turned-over tissu

Animal Tissue EM1 EM2

Glycin1 Liver 0.0403 0.0044

Muscle 0.0396 0.0042Brain 0.0410 0.0044

2 Liver 0.0403 0.0044Muscle 0.0388 0.0042Brain 0.0412 0.0045

3 Liver 0.0425 0.0046Brain 0.0427 0.0047

Mean±SD 0.0408±0.0013 0.0044±0.00

Alanin1 Liver 0.0821 0.0124

Muscle 0.0824 0.0126Brain 0.0711 0.0103

3 Liver 0.0827 0.0126Brain 0.0691 0.0099

Mean±SD 0.0783±0.0068 0.0116±0.00

Prior to mating, a female rat was started on 4% 2H2O in drinking water, and labelingmother and 3 pups were sacrificed. Blood was collected from the mother for measumuscle and brain tissue were collected. Mixed proteins from these tissues were precipanalyzed by GC/MS in PCI mode. Isotope enrichments (excess fractional abundance iValues of p were calculated from EM2/EM1 ratios, assuming n=2 for Gly and n=4 fosacrifice) was assumed to represent the precursor pool enrichment, and nwas calculateinterpolation betweenMIDA predictions for integral values of n. Low EM2 abundancefrom pup #3 and of Ala in tissues from pup #2.

(Table 2). Conversely, when the number of C–H bonds wasassumed to represent n, and p was calculated from the EM2/EM1 ratio, the values of p were in excellent agreement with theexperimentally determined body water 2H enrichment in themother's plasma at sacrifice (Table 2). Finally, based on eitherthe calculated values of p or n (Table 2), protein fractionalsynthesis (f) was calculated and was indeed complete (99.8%for protein-bound Gly, 106% for protein-bound Ala). Inconclusion, all C–H bonds in the Ala and Gly residues oftissue proteins were accessible to 2H2O labeling in utero, and nodilution of 2H label was detectable during its flux from themother's plasma water into the pups' tissue protein NEAA. Thisfinding also implied that no label dilution occurred in any of thebiosynthetic intermediates (free cellular NEAA or NEAA-tRNA; Fig. 1B), although this was not examined directly.

In a separate experiment, we administered 8% 2H2O indrinking water to an adult rat for 6 months to achieve completeturnover of serum proteins. Serum albumin and IgG wereisolated from replicate blood samples, hydrolyzed, andanalyzed for label incorporation into Ala by GC/MS (Table3). Calculated values of p (assuming n=4 for Ala) were onlyslightly lower than the measured body water 2H enrichment atthe time of sacrifice. Alternatively, calculated values of n(assuming that p=measured 2H2O enrichment) were 3.6–3.7.The fractional synthesis (f) was 100–103% for both serumproteins, based on these MIDA calculations, confirmingcomplete turnover (Table 3). Thus, in serum proteins of adultanimals, C–H bonds of protein-derived Ala are almost fullyaccessible to body water-derived 2H. The slight decrease in p orn, or both, seen in adults suggests that exchange is incomplete at

e proteins in rat pups exposed to 2H2O in utero

EM2/EM1 Calc. p (%) Calc. n

e0.1088 2.60 2.020.1071 2.30 1.990.1071 2.30 1.990.1088 2.60 2.020.1082 2.55 2.010.1104 2.83 2.050.1074 2.38 2.000.1071 2.25 1.99

02 0.1081±0.001 2.48±0.20 2.01±0.02

e0.1510 2.57 4.210.1529 2.67 4.350.1449 2.22 3.770.1524 2.64 4.310.1433 2.33 3.65

13 0.1489±0.0045 2.49±0.20 4.06±0.32

was continued throughout pregnancy and delivery. Within 24 h of delivery, therement of body 2H enrichment. The pups were dissected and samples of liver,itated and hydrolyzed to free amino acids, and N-acetyl, n-butyl derivatives weren the M1 andM2 mass isotopomers over background) are shown for Gly and Ala.r Ala. Alternatively, the mother's body water 2H enrichment (2.4% at the time ofd from the measured EM2/EM1 ratio. Non-integral values of nwere estimated bydue to variable sample losses prevented accurate analysis of muscle Ala and Gly

Fig. 3. Time course of 2H label incorporation into selected NEAAs of mixedbone proteins after continuous heavy water labeling of growing adult femalemice. After continuous 2H2O labeling of mice for the indicated times, Ala (A)and Gly (B), obtained by acid hydrolysis of bone proteins, were derivatized tothe corresponding N-acetyl, n-butyl esters and analyzed by GC/MS in PCImode. Data are shown as average % EM1 enrichment (percent molar excess ofM1 mass isotopomer over natural abundance) for 2 animals per time point. Rateconstants for bone protein synthesis are based on single-exponential curve fits(Eq. (1)) to the data.

Table 3Turnover of serum proteins after 6 months of 2H2O labeling in an adult rat

Serumprotein

EM1 EM2 EM2/EM1 Calc.p (%)

Calc. n EM1, max f (%)

Albumin 0.1057 0.0319 0.3016 4.50 3.7 0.1038 1020.1066 0.0322 0.3016 4.50 3.7 0.1038 103

IgG 0.1016 0.0304 0.2996 4.41 3.6 0.1020 1000.1021 0.0306 0.2995 4.40 3.6 0.1019 100

An individual adult male rat was maintained on 8% 2H2O in drinking water for 6months. Blood was collected at the time of sacrifice. Replicate samples of serumproteins were isolated and hydrolyzed, and the alanine NCI derivative wasanalyzed by GC/MS. Body water 2H enrichment in the rat was 4.9% (postmortem plasma). Values for p and n were calculated as in Table 2. EM1, max,maximal (plateau) EM1 value, calculated from the measured EM2/EM1 ratiosusing either calculated n or p. f, fractional synthesis of protein, calculated asEM1/EM1,max.

737R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

one or both positions, perhaps due to dilution of thecorresponding Ala-tRNA pool(s). Possible sources of labeldilution are discussed further in the Web Supplement.Nonetheless, these data collectively indicate that for theNEAAs studied, C–H bonds are fully (or, in the case of Alain adult rats, almost fully) metabolically accessible to 2H labelfrom 2H2O.

3.6. Kinetics of a slow turnover protein in rodents

In order to apply 2H2O labeling to proteins with a slowturnover rate, we performed a 2H2O labeling time course formixed bone proteins (mostly collagens). In young adult mice(Fig. 3), there was a continuous rise in label incorporation, withsimilar apparent rate constants for Ala and Gly, (kAla=0.178week−1, kGly=0.163 week−1). Similar data were obtained inrats, but the rate of bone protein turnover was slower (0.084week−1, corresponding to t1/2 ≈8.25 weeks; not shown). MIDAcalculations showed that the values of n and p remainedconstant and maximal (Gly) or near-maximal (Ala) for boneproteins, even at earlier times when the protein had not turnedover fully (Web supplement, Tables S2 and S3). We inferredfrom these calculations that NEAA precursors used for boneprotein synthesis were equilibrated with 2H in plasma water,and that their isotopic enrichment did not detectably drift duringthe labeling time course, as might happen, for instance, ifrecycling of labeled vs. unlabeled NEAAs changed substan-tially over time.

3.7. Comparison to Die-away curves

We also measured protein turnover rate constants duringlabeling and de-labeling for a second long-lived proteincompartment, mixed proteins in muscle (Fig. 4). Duringcontinuous labeling with 2H2O,

2H label incorporation intoAla residues of murine skeletal muscle proteins followed singleexponential kinetics with k=0.23 week−1 (Fig. 4A). Results inrat muscle were similar, both for Ala (k=0.21 week−1) and Glu/Gln (k=0.23 week−1). Replacement rates of mixed proteinsfrom rat heart were somewhat higher (k=0.31 week−1, protein-bound Ala). When 2H2O administration to rats was discon-

tinued, wash-out of 2H2O from body water pools took about 2weeks (Fig. 4B). The time course of the subsequent loss of 2Hlabel for Ala residues of muscle proteins was consistent withsingle-exponential decay kinetics. The rate constants for loss oflabel were similar to the corresponding incorporation rateconstants measured during continuous labeling: kd=0.19week−1 for Ala in rat skeletal muscle and kd=0.37 week−1 forAla in rat heart (Fig. 4C and D). The internal consistency ofthese results suggests that heterogeneous labeling of NEAA-tRNA pools does not confound the interpretation of labelincorporation data by the rise-to-plateau approach.

3.8. Synthesis of crosslinking variants of rat liver collagen inresponse to a fibrogenic stimulus

We next wished to demonstrate the ability of 2H2O labelingto detect experimentally induced changes in synthesis ofparticular proteins isolated from complex tissue sources. Tothis end, we exposed adult rats to carbon tetrachloride (CCl4), awell-defined animal model of hepatic toxicity, featuringupregulation of collagen synthesis and subsequent fibrosisamong the responses to free radical damage [27]. Mice werelabeled with 2H2O for 3 weeks and received either a fibrogenicdose of CCl4 (J.G., unpublished; 3 animals) or vehicle (2animals) twice weekly throughout the labeling period. Collagenwas isolated from their livers at sacrifice. SDS-PAGE analysis(Fig. 5A), tryptic peptide mapping, and Western blots (notshown) demonstrated that these preparations were highlyenriched in type I collagen. Moreover, the collagen isolatescontained electrophoretically separable species of monomers,

Fig. 4. Label incorporation and decay in rodent mixed muscle proteins. (A) Time course of 2H label incorporation into mixed skeletal muscle proteins in micemaintained on 8% 2H2O in drinking water. Ala, obtained by total acid hydrolysis of muscle proteins, was derivatized to the N-propyl carboxyl, n-propyl ester andanalyzed by GC/MS in PCI mode. Data are shown as fractional synthesis (f), assuming n=4 for Ala. Error bars are SD for groups of 3–5 mice per time point. (B–D)Rat muscle protein turnover measured in de-labeling experiments. (B) Washout of 2H from body water in rats at various times after discontinuing 2H2O intake indrinking water (groups of 3–4 rats per time point, as indicated; SDs are shown). (C and D) Loss of 2H label in rat skeletal (C) and cardiac (D) muscle mixed proteinsafter cessation of oral 2H2O intake. The loss of 2H enrichment in the M1 mass isotopomer of Ala (N-acetyl, n-butyl ester derivative, PCI analysis) is shown (average,groups of 3 rats per time point). Time zero is 2 weeks after discontinuing intake of 2H2O (i.e., arrow in B). Rate constants for de-labeling were determined by curvefitting to a single exponential decay model (Eq. (2)).

738 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

distinguished by their isoform content, as well as covalentlycrosslinked dimers and trimers (Fig. 5A). To examine labelincorporation, monomer and dimer regions were excised fromPVDF blots, hydrolyzed, and analyzed for 2H incorporation intothe released Ala residues by GC/MS (Fig. 5B). The resultsclearly show that collagen monomers incorporated 2H-Ala morerapidly than crosslinked dimers, consistent with the notion thatturnover in the more extensively crosslinked species is slower[33,34]. Moreover, CCl4 administration increased incorporationof 2H-Ala into collagen monomers and, to a lesser extent, intodimers, compared to vehicle-treated controls (Fig. 5B). Thus,this analytical method was capable of measuring 2H incorpo-ration by modest quantities of a purified protein excised fromblots (Coomassie Blue-stained bands, representing ca. 0.1–1 μgper lane) and detected the interacting effects of an experimental

stimulus and a posttranslational modification on proteinsynthesis.

3.9. Application to human serum proteins

The well-documented safety of 2H2O labeling in humans (cf.Introduction) allowed us to investigate the feasibility ofmeasuring slow protein turnover rates in humans by this method.We were unable to use MIDA because at the lower body water2H enrichments attained in humans (1–2%), label incorporationinto the M2 mass isotopomer is insufficient for accuratemeasurements (not shown). Instead, we asked whether 2H2Olabeling reproduced the known half-lives of well-studied humanserum proteins and looked for evidence of label dilution afterlabeling to plateau. To this end, healthy human subjects (3 men,

Fig. 5. Fractional synthesis of crosslinking variants of liver collagen in a rat model of CCl4-induced fibrogenesis. (A) SDS-PAGE analysis of liver collagen (mainlytype I) and of type I and type III collagen standards. The chain composition of monomer (M) and dimer (D; covalently crosslinked) bands of collagen is indicated [49].(B) Comparison of 2H label incorporation into liver collagen monomers and dimers. Collagen was purified from livers of rats that had been labeled with 8% 2H2O indrinking water and given twice-weekly injections of 1 ml/kg CCl4 (or vehicle) for 3 weeks. Collagen preparations were resolved by SDS-PAGE and transferred toPVDFmembranes. Monomer and dimer bands (as indicated in C) were excised and hydrolyzed, and the PFB derivative of Ala from each hydrolyzate was analyzed for2H incorporation (expressed as % EM1). Paired results from 3 CCl4-treated and 2 control rats are shown.

739R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

1woman; age 41±4 years) were given 2H2O daily for a period of10 weeks. Body water 2H enrichments after 1-week of labelingwere in the range of 1.8–2.5% and remained stable over 2months (not shown), as observed previously [11,18,21]. Asexamples of slow-turnover proteins, we purified immunoglob-ulin G (Fig. 6A, lane 1) and albumin (lane 2) from serum samplesobtained at various times after initiation of labeling. For bothproteins, the time course of label incorporation was consistentwith a single-exponential rise to plateau (Fig. 6B), with t1/2values that agreed with those obtained by standard methods(21.5 and 17 days for IgG1/2/4 and albumin, respectively, vs. 21and 17 days in the literature [35–37]). Also in line with previousestimates [37], albeit less well estimated from our data, t1/2 forIgAwas ≈6 days. LDL apoB turnover was essentially completeby day 11, the first time point sampled, as expected (t1/2≈2 daysin the literature [38]); the curve fit did not constrain turnover ofthis protein well.

In Fig. 6, fractional turnover (the fraction of protein-boundAla that was newly synthesized) was calculated assuming thatall C–H bonds of Ala-tRNAs were fully accessible to 2H labelfrom 2H2O (i.e., n=4) and that 2H in Ala-tRNAs was fullyequilibrated with body water (i.e., p=body water 2H enrich-ment). If true, then the label incorporation curves for fullyturned-over proteins should approach asymptotes of fmax=100%. This was indeed observed for LDL apoB (fmax=96%),but for other proteins the asymptotes were somewhat lower(83% for albumin, 68% for IgG, and 70% for IgA; Fig. 6B).These results suggest that de novo synthesized Ala-tRNAs canbe diluted with unlabeled sources of Ala, even after prolonged

labeling, and that the dilution depends on the protein studied,possibly reflecting different metabolic environments at theirdistinct sites of biosynthesis (liver for albumin and LDL apoB;bone marrow plasma cells for Ig). Mechanisms for labeldilution are discussed further in the Web Supplement (WebSupplement, Fig. S2). Importantly, however, the observedlabel dilution did not alter the shape of the label incorporationcurve or otherwise bias measurements of fractional turnoverrates.

These experiments demonstrated the utility of 2H2O labelingfor proteins with half-lives of N1 week, but left open thepossibility that the analysis of proteins with faster turnoverrates, on the order of a few hours to a few days, might beconfounded by the kinetics of label equilibration between bodywater and NEAA-tRNA. To address this concern, we studiedthe kinetics of 2H incorporation into VLDL ApoB in twosubjects after bolus labeling with 2H2O (Fig. 7). Following fourdrinks of 2H2O, distributed over 60 min, body water 2Henrichments reached stable values of 0.2–0.3% within anotherhour and remained at this level for the subsequent 8 h (Fig. 7A).Starting 2 h after the first dose of 2H2O (i.e. after body water 2Henrichments had stabilized; arrowheads), label incorporationinto VLDL ApoB was consistent with single-exponentialkinetics. Half-lives were about 2.5 h for a normolipidemic,obese volunteer (subject 1, Fig. 7B), and about 8.3 h for adyslipidemic (but normoglycemic) patient on the hypolipidemicagent, atorvastatin (subject 2), within the range of literaturevalues for VLDL ApoB turnover in similar human subjectsbased on labeling with deuterated Leu [38]. As with LDL, near-

Fig. 6. Turnover of human serum proteins. (A) Purity of proteins isolated from 2H2O-labeled humans, as determined by SDS-PAGE (examples). Lane 1, IgG1/2/4purified by protein A affinity chromatography and SEC; positions of heavy (H) and light (L) chains and of incompletely reduced species (*) are shown. Lane 2, albumin(HSA) purified by HiTrap Blue sepharose chromatography and SEC. (B) Time course of label incorporation into HSA, IgG1/2/4, LDL apoB, and IgA1 in healthysubjects. Sera were obtained at the indicated times after initiation of continuous 2H2O labeling, proteins were isolated, and hydrolyzates were derivatized for GC/MSanalysis in NCI mode. Body water 2H enrichment was determined from urine samples starting around 1 week, and averaged separately for each individual to calculatetheoretical asymptotic EM1,max in alanine (assuming n=4), for estimation of fractional synthesis (f). Lines represent curve fits to the data using (Eq. (1)). Calculatedturnover constants (k), half-lives (t ½) and asymptotes of each protein are shown; errors are SEM for these parameters based on the curve fits. Analysis of N-acetyl, n-butyl ester derivatives prepared from IgG1/2/4 hydrolyzates in PCI mode (not shown) gave similar kinetic parameters as did NCI analysis: Asymptotic f=0.69±0.08;k=3.0±0.7%/days; t1/2 = 23.3 days.

740 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

maximal label incorporation was observed at plateau in VLDLApoB (fmax=96% for subject 1). Thus, heavy water labelingappeared capable of measuring protein half-lives as short as afew hours; measurement of even shorter protein half-liveswould be limited by the time scale of 2H equilibration in bodywater (b30 min; Fig. 7A and R.N., E. Parks, M.K.H.,unpublished observations).

4. Discussion

The attractiveness of 2H2O as a metabolic label for long-termbiosynthetic studies and its proven utility for measuring lipidand DNA turnover led us to examine its general utility formeasuring protein turnover. Recently, we [39] used 2H2Olabeling of Ala to measure dynamic exchange between tubulinα/β dimers and polymerized microtubules. Previs et al. trackedthe time course of 2H incorporation from body water into Alaresidues of human albumin [40,41]. However, 2H2O labelingpotentially is subject to the same concerns about heterogeneousAA pools as is labeling with isotope-tagged AAs [2–7].Moreover, earlier studies had highlighted particular concernsover the use of 2H2O for measuring protein synthesis,discouraging use of this label [42–45]. One key issue thatneeded to be resolved is the precursor/product relationship for2H incorporation into proteins from 2H2O. To this end, weperformed MIDA of rodent proteins and plateau labelingexperiments in humans, two approaches that allow precursorpool enrichments to be assessed even if they are not accessible

to direct measurement [2]. Taken together, the data show that2H2O is a broadly useful biosynthetic label for measuringprotein turnover.

First, 2H label incorporation from 2H2O into C–H bonds ofNEAA residues of a protein represents protein synthesis. Weconfirmed that C–H bonds in protein-bound AAs did notexchange 2H spontaneously with 2H2O. Moreover, 2H atomschemically attached at these sites were fully retained duringsample processing. Different derivatives and GC/MS modalities(NCI, PCI) could be used interchangeably, showing theconsistency of the method. Thus, analytically, the methodmeasures biosynthetic label incorporation at C–H bonds ofprotein-bound NEAAs.

2H atoms from 2H2O can attach to labile N–H and O–Hbonds in protein-bound NEAA via nonenzymatic exchange, aprocess that is independent of polypeptide synthesis and maydepend on the folded state of the protein. Analytic retention ofthese labile 2H atoms would confound turnover measurements[44]. In our method, however, labile 2H atoms are first lostduring total acid hydrolysis of proteins to free NEAA (whichalso destroys any secondary structure that may otherwiseprotect exchangeable 2H) and, in addition, are then replaced byfunctional groups during chemical derivatization reactions.

A second potential concern with 2H2O labeling is that themass difference between 2H and 1H may affect the physico-chemical properties of 2H-labeled molecules, and thus theirphysiology. However, there is no evidence that, in the amountsused for labeling (body water enrichments of 1–2% in humans,

Fig. 7. Turnover of VLDL ApoB measured after bolus labeling of healthydonors with 2H2O. Time course of body water 2H enrichment (A) and labelincorporation into VLDL ApoB (B) in two human subjects after oral bolusadministration of four doses of 2H2O, given at 20 min intervals over 60 min(arrows in A). The arrowheads indicate the 2 h time point after the first dose of2H2O, which was taken to be time zero for the protein labeling time course.Values of f (fraction of Ala that is newly synthesized) were calculated as in Fig.5, using average body water 2H enrichment after the 2 h time point to representthe precursor pool enrichment (p). Label incorporation at time zero of the proteinlabeling time course (2 h after the initial 2H2O bolus) represents proteinsynthesis during the initial rise in body water 2H. Closed diamonds andcontinuous line represent data and single-exponential curve fit for anormolipidemic, obese subject (k=27.7%/h, or t1/2=2.5 h; fmax=96.1%).Open circles and dashed line represent data and curve fit for a hypercholes-teremic subject on atorvastatin (k=8.3%/h, or t1/2=8.3 h; since no plateau wasreached, fmax was constrained to 100%).

741R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

3–6% in rodents), 2H exerts physiologic effects that couldconfound kinetic studies. The safety and lack of long-termadverse effects of 2H2O (see Introduction) argue againstphysiologic abnormalities in labeled animals, as does theagreement between 2H2O-based and classical measurements ofprotein turnover. In our previous study of tubulin dynamics, wefailed to detect any effects of 4% 2H2O in cell culture on theabundance of monomeric and polymeric tubulin [39]. Similarcontrols can easily be designed for other applications.

3H incorporation from 3H2O into proteins is less efficientthan from 3H-labeled food proteins, which had raised concernsthat incorporation from labeled water might be insufficient formeasurement [42–45]. In practice, this was not a problem,however. One advantage of measuring 2H incorporation by GC/MS was that the 2H label is incorporated into a given NEAA atmultiple sites (e.g., 4 C–H bonds in Ala). Therefore, the amountof label entering the M1 mass isotopomer of the amino acidderivative is several-fold higher than the corresponding bodywater 2H enrichment (Fig. 2E, Table S1), which increases

sensitivity. Accurate protein labeling curves can thus be obtainedby GC/MS, even at the low body water 2H enrichments used inhumans.

Another concern was that, even though we attained stable 2Henrichments in body water, 2H enrichment in precursor poolsused for protein synthesis (i.e., NEAA-tRNA) might be variableor drift over time. In earlier work, brief 3H2O or 2H2O exposureprimarily labeled α carbons (C2), while longer label exposureresulted in incorporation into other positions [44]. However, asexpected from previous studies [43], this did not appear to be aproblem with the method described here. In rodents, measure-ments of precursor pool enrichment (p), using MIDA, showedno label dilution in Gly or Gln/Glu incorporated into tissueproteins, compared to body water enrichment, so the issue ofvariable or drifting label dilution was moot. The slight (≈10%)dilution detected for Ala in adult rodent proteins was consistentover time. Similarly, Previs et al. showed that label dilution atC3 of Ala was consistent during a labeling time course [40].Moreover, label incorporation curves in humans and rodentsand comparison of labeling and delabeling curves showed noevidence of delayed label equilibration, which, if present, mighthave caused deviations from single-exponential labelingkinetics.

The extent of label dilution in precursor pools, however, canbe experimentally assessed for each tissue, if not for eachprotein of interest, even in humans. Although we have notmeasured 2H enrichments in NEAA-tRNA directly, asymptoticfmax values reflect, in principle, label dilution in the relevantprecursor pools [2]. Measured values of fmax ranged from almostno dilution (relative to the theoretical maximum for the bodywater enrichment that was present), for liver-derived VLDL andLDL ApoB, to ~30% dilution for plasma cell-derived IgG andIgA. Dilution of 2H label might occur at each step of the labelincorporation pathway outlined in Fig. 1B. In practice, MIDAstudies do not permit these alternative sources of label dilutionto be distinguished (Web Supplement, Fig. S2). In any case,label dilution of a constant magnitude should not alter rise-to-plateau kinetics and did not, in fact, appear to bias measurementof t1/2, in that the kinetics of 2H incorporation from 2H2O intovarious serum proteins agreed well with protein half-livesreported in the literature.

The time required for 2H2O to equilibrate in body water andfor 2H to distribute into the relevant NEAA pools does define alower limit on the half-lives that can be measured using 2H2O.In rodents, body water equilibrates within 15 min after an i.p.priming bolus. In humans, water in urine, plasma, and salivaequilibrate within an hour after a single dose of oral 2H2O (Fig.7A; unpublished observations). Moreover, although 2H2Olabeling was not compared side by side with other methods,measured half-lives for VLDL ApoB (a few hours) agreed withtypical literature values, suggesting that label disequilibrium didnot confound turnover measurements on this time scale. It istherefore likely that 2H exchange between 2H2O and NEAAprecursor pools is fast enough to allow protein half-lives asshort as a few hours to be measured. There appears to be noabsolute upper limit to the range of measurable half-lives, otherthan the life-span of the animal.

742 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

In terms of maintaining constant 2H enrichment in theprecursor pools used for protein synthesis, 2H2O labeling viaNEAAs is an extremely effective approach. 2H2O labeling is atleast comparable to a continuous intravenous infusion of labeledessential AAs, a standard approach for fast-turnover proteins.Labeled AAs must be taken up by tissues and mix withintracellular AA pools before being incorporated into proteins,leaving opportunities for incomplete isotope mixing inprecursor pools and complicating measurements of proteinsynthesis. 2H2O, in contrast, diffuses rapidly into tissues andacross cell membranes and exchanges 2H rapidly withintracellular NEAA pools via reactions of intermediarymetabolism. Rapid equilibration with the large 2H pool inbody water may explain why neither we nor Previs et al.observed substantial fluctuations in the precursor pool enrich-ment over time, or evidence of drift due to label recycling, whenusing long-term 2H2O labeling. The only exception was thebrief perturbation of plasma Ala enrichments after a meal [41].

2H2O is inexpensive, allowing consistent 2H enrichments tobe maintained for weeks or months. As a result, labelincorporation into proteins can readily be tracked to plateau,even for proteins with low turnover rates. Similar experimentsthat required prolonged administration of 2H-labeled essentialAAs would be difficult and expensive. These features of 2H2Olabeling enable long-lived proteins to be studied by the rise-to-plateau labeling approach, which has several advantages overinitial rate measurements using labeled AAs. First, rise-to-plateau labeling kinetics are straightforward to interpret, even insituations where protein levels change over time [6]. Second,long-term labeling designs permit short-term fluctuations ofsynthesis rates to be averaged out (although this could be adisadvantage when such fluctuations are of physiologicalinterest). Third, kinetically distinct subpopulations, includingprotein isoforms with distinct labeling kinetics, are more easilydetected when the entire labeling curve can be followed. Fourth,as we have shown above, labeling to plateau with 2H2Orepresents a practical way to address the problem of variablelabel dilution in precursors of protein synthesis [2–7] for slow-turnover proteins, even when isotopic enrichment AA-tRNApools cannot be measured directly and when label incorporationis too low for MIDA.

Some data presented here hint at kinetic complexities. Whenliver collagens with different degrees of crosslinking werecompared, the label was preferentially incorporated into theleast extensively crosslinked species, highlighting the relation-ship between posttranslational modifications and proteinturnover. Moreover, turnover rates for human albumin basedon 2H2O labeling (Fig. 6 and [40]) agreed with the literature[35,36], but faster rates have recently been reported by short-term labeling with essential AAs [46,47]. It may proveinteresting to trace the source of the latter discrepancy (e.g.,by comparing concurrent labeling with 2H2O and 2H-labeledessential AAs).

The convenience of 2H2O labeling, compared to infusion of2H-labeled AAs, cannot be overemphasized. In animal studies,giving 2H2O in drinking water is considerably easier than theformulation of diets or continuous infusions with labeled AAs.

In clinical studies, daily or twice-daily oral intake of 50–60 ml70% 2H2O by out-patients obviates the need for intravenousinfusions, medical supervision, sterility concerns, specialhandling of tracers, or complex instructions. Although in thepresent study, 2H2O labeling was initiated at a clinical researchcenter, we have performed 2H2O labeling protocols entirely onan out-patient basis. These features make field studies practical,even for long-lived proteins.

2H2O labeling need not be limited to abundant proteins.Using NCI analysis of Ala PFB derivatives, we obtainedexcellent data starting with 0.1–1 μg of purified protein (Figs.5 and S3). These sample requirements should allow manyproteins of moderate abundance to be analyzed, whileallowing abundant proteins to be processed from smallamounts of blood or tissue biopsies. The sensitivity of NCI-GC/MS analysis is similar to the sample needs of recentkinetic proteomics studies, in which proteins were isolated asabundant 2D-PAGE spots and their turnover was measured byLC-MS analysis of tryptic fragments [1,48]. Despite theanalytical sensitivity of NCI-GC/MS, applications to proteinsof very low abundance may prove challenging. High proteinpurity and rigorous exclusion of contamination by ambientproteins or free NEAA are critical for generating reliablekinetic data.

Many preclinical and clinical applications can be envisagedfor the 2H2O labeling, rise-to-plateau approach. In animals,2H2O labeling may simplify studies of disease and drug effectson turnover of various proteins (Figs. 3–5) [39]. Clinical studiesthat could benefit from 2H2O labeling include, for example,altered lipoprotein turnover in metabolic diseases, collagendeposition in fibrosis, hemoglobin production in anemia, or Igoverproduction in multiple myeloma (Figs. 6 and 7; R.B., C.B.,M.M., P.P. Lee, C. Chen, M.K.H., unpublished data). Anotheruseful feature of the method is that other physiologic processesof interest, such as cell proliferation, can be measuredconcurrently using the same 2H2O label [39] (R.B., M.M., P.P. Lee, C. Chen, unpublished data). In conclusion, rise-to-plateau labeling with 2H2O is an easy to use, versatile, androbust method with many practical advantages for measuringsynthesis and turnover rates of proteins with turnover timesexceeding a few hours.

Acknowledgments

These studies were supported in part by NIH grants AI44767and AI41401 (to M.K.H.) and by an unrestricted gift fromKineMed, Inc. We thank Glen Lindwall, Abraham J. Bautista,Tim Riiff, and Kristen L. LaPrade for assistance with proteinpurification, Ablatt Mahsut for assistance with mass spectrom-etry, and Glen Lindwall for critically reading the manuscript.Marc K. Hellerstein is Chief of the Scientific Advisory Boardand owns stock in KineMed, Inc.

Appendix A. Supplementary data

Supplementary data associated with this article can be foundin the online version at doi:10.1016/j.bbagen.2005.12.023.

743R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

References

[1] R.J. Beynon, J.M. Pratt, Metabolic labeling of proteins for proteomics,Mol. Cell Proteomics 4 (2005) 857–872.

[2] M. Hellerstein, Methods for measuring polymerisation biosynthesis: threegeneral solutions to the problem of the “true precursor”, Diabetes Nutr.Metab. 13 (2000) 46–60.

[3] J. Airhart, A. Vidrich, E.A. Khairallah, Compartmentation of free aminoacids for protein synthesis in rat liver, Biochem. J. 140 (1974) 539–545.

[4] E.A. Khairallah, G.E. Mortimore, Assessment of protein turnover inperfused rat liver. Evidence for amino acid compartmentation fromdifferential labeling of free and tRNA-gound valine, J. Biol. Chem. 251(1976) 1375–1384.

[5] C. Papageorgopoulos, K. Caldwell, C. Shackleton, H. Schweingruber,M.K. Hellerstein, Measuring protein synthesis by mass isotopomerdistribution analysis (MIDA), Anal. Biochem. 267 (1999) 1–16.

[6] J.C. Waterlow, P.J. Garlick, D.J. Millward, Protein Turnover inMammalian Tissues and in the Whole Body, North Holland, Amsterdam,1978.

[7] C. Papageorgopoulos, K. Caldwell, H. Schweingruber, R.A. Neese, C.H.Shackleton, M. Hellerstein, Measuring synthesis rates of muscle creatinekinase and myosin with stable isotopes and mass spectrometry, Anal.Biochem. 309 (2002) 1–10.

[8] S.M. Turner, E.J. Murphy, R.A. Neese, F. Antelo, T. Thomas, A. Agarwal,C. Go, M.K. Hellerstein, Measurement of TG synthesis and turnover invivo by 2H2O incorporation into the glycerol moiety and application ofMIDA, Am. J. Physiol.: Endocrinol. Metab. 285 (2003) E790–E803.

[9] M.K. Hellerstein, R.A. Neese, Mass isotopomer distribution analysis ateight years: theoretical, analytic, and experimental considerations, Am. J.Physiol. 276 (1999) E1146–E1170.

[10] W.N. Lee, S. Bassilian, H.O. Ajie, D.A. Schoeller, J. Edmond, E.A.Bergner, L.O. Byerley, In vivo measurement of fatty acids and cholesterolsynthesis using D2O and mass isotopomer analysis, Am. J. Physiol. 266(1994) E699–E708.

[11] A. Strawford, F. Antelo, M. Christiansen, M.K. Hellerstein, Adipose tissuetriglyceride turnover, de novo lipogenesis, and cell proliferation in humansmeasured with 2H2O, Am. J. Physiol.: Endocrinol. Metab. 286 (2004)E577–E588.

[12] S.K. Peng, K.J. Ho, C.B. Taylor, Biologic effects of prolonged exposure todeuterium oxide. A behavioral, metabolic, and morphologic study, Arch.Pathol. 94 (1972) 81–89.

[13] P.D. Klein, E.R. Klein, Stable isotopes: origins and safety, J. Clin.Pharmacol. 26 (1986) 378–382.

[14] P.J. Jones, S.T. Leatherdale, Stable isotopes in clinical research: safetyreaffirmed, Clin. Sci. (Lond.) 80 (1991) 277–280.

[15] C.A. Leitch, P.J. Jones, Measurement of human lipogenesis usingdeuterium incorporation, J. Lipid Res. 34 (1993) 157–163.

[16] R.L. Jungas, Fatty acid synthesis in adipose tissue incubated in tritiatedwater, Biochemistry 7 (1968) 3708–3717.

[17] M.L. Collins, S. Eng, R. Hoh, M.K. Hellerstein, Measurement ofmitochondrial DNA synthesis in vivo using a stable isotope-massspectrometric technique, J. Appl. Physiol. 94 (2003) 2203–2211.

[18] M.K. Hellerstein, R.A. Hoh, M.B. Hanley, D. Cesar, D. Lee, R.A. Neese,J.M. McCune, Subpopulations of long-lived and short-lived T cells inadvanced HIV-1 infection, J. Clin. Invest. 112 (2003) 956–966.

[19] J. Kim, F. Antelo, R.A. Neese, M.K. Hellerstein, A new stable isotope-mass spectrometric method to measure proliferation rates of colonepithelial cells, FASEB J. 14 (2000) A718.

[20] L.M. Misell, J. Thompson, F. Antelo, Y.C. Chou, S. Nandi, R.A. Neese,M.K. Hellerstein, A new in vivo stable isotope-mass spectrometricmethod for measuring mammary epithelial cell proliferation, FASEB J.14 (2000) A786.

[21] R.A. Neese, L.M. Misell, S. Turner, A. Chu, J. Kim, D. Cesar, R. Hoh, F.Antelo, A. Strawford, J.M. McCune, M. Christiansen, M.K. Hellerstein,Measurement in vivo of proliferation rates of slow turnover cells by 2H2Olabeling of the deoxyribose moiety of DNA, Proc. Natl. Acad. Sci. U. S. A.99 (2002) 15345–15350.

[22] R.A. Neese, S.Q. Siler, D. Cesar, F. Antelo, D. Lee, L. Misell, K. Patel, S.Tehrani, P. Shah, M.K. Hellerstein, Advances in the stable isotope-massspectrometric measurement of DNA synthesis and cell proliferation, Anal.Biochem. 298 (2001) 189–195.

[23] H. Shimamoto, S. Komiya, Comparison of body water turnover inendurance runners and age-matched sedentary men, J. Physiol. Anthropol.Appl. Hum. Sci. 22 (2003) 311–315.

[24] A. Raman, D.A. Schoeller, A.F. Subar, R.P. Troiano, A. Schatzkin, T.Harris, D. Bauer, S.A. Bingham, J.E. Everhart, A.B. Newman, F.A.Tylavsky, Water turnover in 458 US adults 40 to 79 years of age, Am. J.Physiol.: Renal. Physiol. 286 (2003) F394–F401.

[25] B.T. Messmer, D. Messmer, S.L. Allen, J.E. Kolitz, P. Kudalkar, D. Cesar,E.J. Murphy, P. Koduru, M. Ferrarini, S. Zupo, G. Cutrona, R.N. Damle, T.Wasil, K.R. Rai, M.K. Hellerstein, N. Chiorazzi, In vivo measurementsdocument the dynamic cellular kinetics of chronic lymphocytic leukemia Bcells, J. Clin. Invest. 115 (2005) 755–764.

[26] S.W. Luckey, D.R. Petersen, Activation of Kupffer cells during the courseof carbon tetrachloride-induced liver injury and fibrosis in rats, Exp. Mol.Pathol. 71 (2001) 226–240.

[27] L.W. Weber, M. Boll, A. Stampfl, Hepatotoxicity and mechanism of actionof haloalkanes: carbon tetrachloride as a toxicological model, Crit. Rev.Toxicol. 33 (2003) 105–136.

[28] E.J. Parks, R.M. Krauss, M.P. Christiansen, R.A. Neese, M.K.Hellerstein, Effects of a low-fat, high-carbohydrate diet on VLDL-triglyceride assembly, production, and clearance, J. Clin. Invest. 104 (1999)1087–1096.

[29] K. Sato, Y.H. Guo, J. Feng, S. Sugiyama, M. Ichinomiya, Y. Tsukamasa, Y.Minegishi, A. Sakata, K. Komiya, Y. Yamasaki, Y. Nakamura, K. Ohtsuki,M. Kawabata, Direct fractionation of proteins in particle-containingfeedstocks by a filter paper pieces-based DEAE-cellulose columnchromatography. Rapid, robust and low-cost capturing procedure forprotein, J. Chromatogr. A 811 (1998) 69–76.

[30] K. Sato, T. Tanahashi-Shiina, F. Jun, A. Watanabe-Kawamura, M.Ichinomiya, Y. Minegishi, Y. Tsukamasa, Y. Nakamura, M. Kawabata,K. Ohtsuki, Simple and rapid chromatographic purification of type Vcollagen from a pepsin digest of porcine intestinal connective tissue,an unmanageable starting material for conventional column chroma-tography, J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 790(2003) 277–283.

[31] M. Richelle, C. Darimont, C. Piguet-Welsch, L.B. Fay, High-throughputsimultaneous determination of plasma water deuterium and 18-oxygenenrichment using a high-temperature conversion elemental analyzer withisotope ratio mass spectrometry, Rapid Commun. Mass. Spectrom. 18(2004) 795–798.

[32] R.R. Wolfe, Radioactive and Stable Isotope Tracers in Biomedicine:Principles and Practice of Kinetic Analysis, Wiley-Liss, New York, 1984.

[33] P.K. Mays, R. McAnulty, G.J. Laurent, Age-related changes in total proteinand collagen metabolism in rat liver, Hepatology 14 (1991) 1224–1229.

[34] G.J. Laurent, Dynamic state of collagen: pathways of collagen degradationin vivo and their possible role in regulation of collagen mass, Am. J.Physiol. 252 (1987) C1–C9.

[35] T. Peters Jr., Serum albumin, Adv. Protein Chem. 37 (1985) 161–245.[36] A. Wagner, U. Herzfeld, Determination of the biological half life of 131-I-

albumin by means of the whole body counter for clinical diagnosis, Med.Klin. 65 (1970) 248–252.

[37] T.A. Waldmann, W. Strober, Metabolism of immunoglobulins, Prog.Allergy 13 (1969) 1–110.

[38] S. Bilz, S. Wagner, M. Schmitz, A. Bedynek, U. Keller, T. Demant, Effectsof atorvastatin versus fenofibrate on apoB-100 and apoA-I kinetics inmixed hyperlipidemia, J. Lipid Res. 45 (2004) 174–185.

[39] P. Fanara, S. Turner, R. Busch, S. Killion, M. Awada, H. Turner, A.Mahsut, K.L. LaPrade, J.M. Stark, M.K. Hellerstein, In vivo measurementof microtubule dynamics using stable isotope labeling with heavy water:effect of taxanes, J. Biol. Chem. 279 (2004) 49940–49947.

[40] S.F. Previs, R. Fatica, V. Chandramouli, J.C. Alexander, H. Brunengraber,B.R. Landau, Quantifying rates of protein synthesis in humans by use of2H2O: application to patients with end-stage renal disease, Am. J. Physiol.:Endocrinol. Metab. 286 (2004) E665–E672.

744 R. Busch et al. / Biochimica et Biophysica Acta 1760 (2006) 730–744

[41] D.A. Dufner, I.R. Bederman, D.Z. Brunengraber, N. Rachdaoui, F. Ismail-Beigi, B.A. Siegfried, S.R. Kimball, S.F. Previs, Using 2H2O to study theinfluence of feeding on protein synthesis: effect of isotope equilibration invivo vs. in cell culture, Am. J. Physiol.: Endocrinol. Metab. 288 (2005)E1277–E1283.

[42] S. Bonotto, I.O. Ndoite, G. Nuyts, E. Fagniart, R. Kirchmann, Study of thedistribution and biological effects of 3H in the algae Acetabularia, Chla-mydomonas and Porphyra, Curr. Top. Radiat. Res. Q. 12 (1978) 115–132.

[43] S.L. Commerford, A.L. Carsten, E.P. Cronkite, The distribution of tritiumamong the amino acids of proteins obtained from mice exposed to tritiatedwater, Radiat. Res. 94 (1983) 151–155.

[44] T.J. Humphrey, D.D. Davies, A sensitive method for measuring proteinturnover based on the measurement of 2–3H-labelled amino acids inprotein, Biochem. J. 156 (1976) 561–568.

[45] R. Mathur-De Vre, J. Binet, Molecular aspects of tritiated water and naturalwater in radiation biology, Prog. Biophys. Mol. Biol. 43 (1984) 161–193.

[46] H.M. van Eijk, D.R. Rooyakkers, B.A. vanAcker, P.B. Soeters, N.E. Deutz,Automated isolation of high-purity plasma albumin for isotope ratiomeasurements, J. Chromatogr., B, Biomed. Sci. Appl. 731 (1999) 199–205.

[47] H. Barle, A. Januszkiewicz, L. Hallstrom, P. Essen, M.A. McNurlan, P.J.Garlick, J. Wernerman, Albumin synthesis in humans increases immedi-ately following the administration of endotoxin, Clin. Sci. (Lond.) 103(2002) 525–531.

[48] J.M. Pratt, J. Petty, I. Riba-Garcia, D.H. Robertson, S.J. Gaskell, S.G.Oliver, R.J. Beynon, Dynamics of protein turnover, a missing dimension inproteomics, Mol. Cell Proteomics 1 (2002) 579–591.

[49] E.J. Miller, S. Gay, Collagen: an overview, Methods Enzymol. 82 (Pt. A)(1982) 3–32.