Human amyloid-b synthesis and clearance rates as measured in cerebrospinal fluid in vivo Randall J Bateman1���3, Ling Y Munsell4, John C Morris1,2,5, Robert Swarm6, Kevin E Yarasheski4 & David M Holtzman1���3,7 Certain disease states are characterized by disturbances in production, accumulation or clearance of protein. In Alzheimer disease, accumulation of amyloid-b (Ab) in the brain and disease-causing mutations in amyloid precursor protein or in enzymes that produce Ab indicate dysregulation of production or clearance of Ab. Whether dysregulation of Ab synthesis or clearance causes the most common form of Alzheimer disease (sporadic, 499% of cases), however, is not known. Here, we describe a method to determine the production and clearance rates of proteins within the human central nervous system (CNS). We report the first measurements of the fractional production and clearance rates of Ab in vivo in the human CNS to be 7.6% per hour and 8.3% per hour, respectively. This method may be used to search for novel biomarkers of disease, to assess underlying differences in protein metabolism that contribute to disease and to evaluate treatments in terms of their pharmacodynamic effects on proposed disease- causing pathways. Protein production and clearance are important parameters that are tightly regulated and reflect normal physiology as well as disease states1���4. Previous studies of protein metabolism in humans have focused on whole-body or peripheral-body proteins, but not on proteins produced in the CNS. A technique to measure specific protein metabolism in the CNS could provide important insights into CNS protein physiology in health and disease. Certain disease states are characterized by disturbances in protein production, accu- mulation or clearance. In the CNS, disturbances in metabolism of proteins such as the prion protein5, alpha-synuclein6, tau7 or Ab8 can contribute to and, in some cases, cause neurodegenerative diseases such as Creuzfeldt-Jakob disease, Parkinson disease, frontotemporal dementia or Alzheimer disease, respectively. Biochemical, genetic and animal model evidence implicates Ab as a pathogenic peptide in Alzheimer disease. The neuropathologic and neurochemical hallmarks of Alzheimer disease include synaptic loss and selective neuronal death, a decrease in certain neurotransmitters and the presence of abnormal proteinaceous deposits in neurons (neurofibrillary tangles), in the cerebral vasculature (amyloid angiopathy) and in the extracellular space (diffuse and neuritic plaques). The main constituent of plaques is Ab, a peptide of 38���43 amino acids cleaved from the amyloid precursor protein (APP)9,10. Throughout life, soluble Ab is secreted mostly by neurons but also other cell types. In late-onset Alzheimer disease, the total amount of Ab that accumulates in brain is B100���200-fold higher in homoge- nates from Alzheimer disease brains than from control brains11. Disturbance of Ab production can lead to rare forms of Alzheimer disease in humans. Mutations in three different genes (APP, PSEN1 and PSEN2), which cause early-onset autosomal dominant Alzheimer disease, all result in overproduction of total Ab or Ab42 (ref. 9). In Down syndrome, three copies of APP result in increased production of Ab, and 100% of individuals with Down syndrome develop Alzheimer disease pathology by age 35 (ref. 12). In late-onset Alzheimer disease (B99% of cases), however, there is not strong evidence for over- production of Ab. Therefore, the underlying cause of deposition of Ab (increased production versus decreased clearance) is not known for most cases of Alzheimer disease. No methods were previously available to quantify protein synthesis or clearance rates in the human CNS. Such a method would be valuable to assess not only Ab synthesis and clearance rates in humans but also the metabolism of a variety other proteins relevant to diseases of the CNS. To address crucial questions about underlying pathogen- esis of Alzheimer disease and Ab metabolism, we developed a method for quantifying the fractional synthesis rate (FSR) and fractional clearance rate (FCR) of Ab in vivo in the human CNS. Our results indicate that by administering a stable isotope-labeled amino acid (13C6-leucine), sampling cerebrospinal fluid (CSF) and using high-resolution tandem mass spectrometry to quantify labeled Ab, reproducible rates of Ab synthesis and clearance can be quantified in humans. RESULTS In vivo labeling and quantification of Ab To determine whether labeled Ab (Fig. 1) could be produced and detected in vivo in a human, one individual underwent a 24-h infusion of labeled leucine followed by a lumbar puncture to obtain CSF. We immunoprecipitated Ab from the CSF sample with the Ab-specific Received 2 November 2005 accepted 24 January 2006 published online 25 June 2006 doi:10.1038/nm1438 1Departments of Neurology, 2The Alzheimer Disease Research Center, 3Hope Center for Neurological Disorders 4Medicine, 5Pathology and Immunology, 6Anesthesiology and 7Molecular Biology & Pharmacology, Washington University School of Medicine, 660 South Euclid Avenue, Box 8111, St. Louis, Missouri 63110, USA. Correspondence should be addressed to R.J.B. (batemanr@neuro.wustl.edu). 856 VOLUME 12 [ NUMBER 7 [ JULY 2006 NATURE MEDICINE TECHNICAL REPORTS �� 200 6 Nature Publishing Group http://www.nature.com/naturemedicine

antibody m266, digested it with trypsin and analyzed it using liquid chromatography���mass spectrometry (LC-MS). The m266 antibody is directed against the central domain of Ab and binds to all Ab species containing amino acids 13���28. The results showed that unlabeled and labeled Ab could be detected and measured in human CSF (Fig. 2). We conducted a pharmacokinetic study to optimize the labeling and sampling times, so that detectable 13C6-leucine labeling of Ab was achieved and maintained for an adequate period of time that permitted us to use steady-state equations to calculate Ab synthesis and clearance rates. We tested a range of 13C6-leucine intravenous infusion dosages (1.8���2.5 mg/kg/h), durations (6, 9 or 12 h) and CSF or blood sampling times (sampling at 12���36 h Table 1). We found that labeled Ab could be reliably quantified after 9 or 12 h of infusion of the label but not after 6 h of infusion of the label. The synthesis portion of the labeling curve could be determined in the first 12 h of sampling however, the clearance portion of the labeling curve could only be determined with 36 h of sampling. Based on these results, we defined optimal labeling parameters for Ab to be 9 h of intravenous infusion of the label and 36 h of sample collection. These parameters allowed for assessment of both the FSR and FCR portions of the labeling curve. In vivo labeling protocol For the last three individuals, we administered 13C6-labeled leucine with an initial bolus of 2 mg/kg over 10 min to reach a steady state of labeled leucine, followed by 9 h of continuous intravenous infusion at a rate of 2 mg/kg/h. We sampled blood and CSF for 36 h in the last three individuals. We took serial 12-ml blood samples and 6-ml CSF samples at 1- or 2-h time intervals (Fig. 3a). CSF has a production rate of B20 ml/h13 in a normal-sized adult and replenishes itself throughout the procedure. Over a 36-h study, the total amounts of blood and CSF collected were 312 ml and 216 ml, respectively. There were a total of ten individuals enrolled in the study, with eight completing the predefined protocols we stopped two studies before completion because of postlumbar puncture headache associated with the study. Two of the eight completed studies had a 6-h labeled leucine infusion, and labeled Ab levels in these two individuals were too low to accurately mea- sure and were not used for analysis. The findings from the remaining six studies are reported here. Quantification of labeled leucine We analyzed plasma and CSF samples to determine the amount of labeled leucine present in each fluid (Fig. 3b). We quanti- fied the labeled-to-unlabeled leucine ratios for plasma and CSF 13C6-leucine using capil- lary gas chromatography���mass spectrometry (GC-MS)14, which is more suitable than LC- MS for analysis of low-mass amino acids. Within 1 h, the 13C6-leucine reached steady- state levels in both plasma and CSF of 14% and 10%, respectively. This confirmed that leucine is rapidly transported across the blood-brain barrier through known neutral amino acid���transporter systems15. Dynamics of labeled Ab For each sample of CSF collected, we deter- mined the labeled-to-unlabeled ratio of Ab by immunoprecipitation���tandem mass spectro- metry (MS/MS) as described above. The number of MS/MS ions from labeled Ab*17��� 28 was divided by the number of MS/MS ions from unlabeled Ab17���28 to produce a ratio of labeled Ab to unlabeled Ab. The mean labeled Ab ratio and standard error (n �� 6) of each Cytosol Cytosol Lumen K K K ... ... APP-CTF APP-NTF APP A�� A��1-5 A��6-16 A�� 17-28 A��29-40/42 V V V K K K O D D D E E E E E T E V V V V G G N S S S V V G G G G V A A A K N S G A D E V G A A F F F H H R I H Y M M M K T T I I I I I L L A I V V V G G G M I I L L L K O E V V F F H H D A E F H R D G S Y L ��-secretase Amyloid �� fragments Amyloid �� sequence Trypsin cleavage sites ��-secretase ��-secretase Figure 1 The amino acid sequence of Ab is depicted in the amyloid precursor protein (APP) in the cell membrane with the leucines (L) labeled in red to indicate possible labeling sites. The sequence of Ab is shown below with the trypsin digest sites indicated to show the fragments that were analyzed by mass spectrometry. LVFFAEDVGSNK LV (m = 213) 212.9 219.0 348.2 366.2 405.2 10 20 Relative intensity (max 2.07 �� 10 4 ) 30 40 50 60 70 80 90 100 10 180 200 220 240 260 280 Mass / charge 300 320 340 360 380 400 20 Relative intensity (max 2.67 �� 10 3 ) 30 40 50 60 70 80 90 100 348.2 360.1 405.2 LVF (m = 360) L*VFFAEDVGSNK L*V (m = 219) L*VF (m = 366) a b Figure 2 Spectra of unlabeled and labeled Ab17���28. We collected human CSF after intravenous infusion of 13C6-leucine. Representative spectra of unlabeled (a) and labeled (b) Ab17���28 are shown. We obtained the spectra using MS/MS analysis of unlabeled parent ion Ab17���28 at m/z 663.3 or labeled parent ion Ab17���28 at m/z 666.3. The MS/MS ions containing leucine (Ab17) are mass shifted by 6 Da, indicating the labeled leucine. The MS/MS ions without leucine are not labeled and are not mass shifted by 6 Da (348 and 405). TECHNICAL REPORTS NATURE MEDICINE VOLUME 12 [ NUMBER 7 [ JULY 2006 857 �� 200 6 Nature Publishing Group http://www.nature.com/naturemedicine