CUN. CHEM. 41/1, 62-68 (1995) #{149} Molecular Pathology Rapid Diagnosis of Maple Syrup Urine Disease in Blood Spots from Newborns by Tandem Mass Spectrometry Donald H. Chace,”3 Steven L. Hiliman,’ David S. Milhngton,’ and Edwin W. Naylor2 We report a new method for the diagnosis of maple syrup urine disease (MSUD) from dried blood spots on newborn screening cards based on tandem mass spectrometry (MS-MS). The mean ± SD concentration of Leu plus lie in normal newborns was 151 ± 47 pmol/L (n = 1096); for Val, 131 ± 58 pmol/L (n = 791). SDs were lower when the concentrations of these amino acids were expressed relative to that of Phe. The mean ratiofor Leu + lie to Phe was 2.5 ± 0.49; for Valto Phe, 2.18 ± 0.51. These results compare well with values previously reported in the literature. With these criteria, samples from a collection categorized by a bacterial inhibition assay as normal or falsely positive for MSUD were normal by MS-MS [(Leu + lle):Phe <5.0]. Samples from confirmed MSUD patients were categorized as abnormal [(Leu + lle):Phe >9.0] by MS-MS. IndexingTerms: aminoacids/screening/metabolism/heritable dis- orders Stephen G. Kahier,’ Charles R. Roe,’ the current methods, especially the BIA, the number of false positives is significant. We have developed a tandem mass spectrometric (MS-MS) method for the simultaneous analysis of Leu + He and Vai from blood spots. Because of the high selectivity and sensitivity of MS-MS, the numbers of potential false positives and false negatives should be minimized. We validated the measurement of Phe and Tyr with MS-MS in the diagnoses of phenylketonuria (PKU) and tyrosinemia, respectively, in previous publications (7-9). As is apparent from these publications, the methods can be used to simultaneously measure several amino acids and acylcarnitines (7-9). Given the numerin the application of multiple-component ous pitfalls profiles to detect multiple diseases, quantitative information for disease-specific components is required to develop diagnostic indicators. With such vigorous validation, the method will continue to expand, accurately both quickly and cost effecdetecting more diseases tively. Maple syrup urine disease (MSUD) is an inherited metabolic disorder characterized by the accumulation of the branched-chain amino acids Leu, Ile, Val, allolle, and their corresponding a-keto acids in blood and urine (1-3). This disorder arises from defects in the oxidative decarboxylation by branched-chain a-keto acid dehydrogenase, a multienzyme complex. Mental and physical retardation are likely in patients with MSUD who are diagnosed after severe decompensation (1). Presymptomatic diagnosis by newborn screening followed by dietary re8triction of branched-chain amino acids as the primary therapy has led to an improved outcome (4, 5). Newborns are screened for MSUD by the quantitative analysis of Leu from dried blood specimens on filter paper. Leu is the most important diagnostic marker and is measured by bacterial inhibition assay (BIA) in many laboratories. Blood concentrations of Leu >305 pinol/L are considered clearly abnormal with this (4 mg/dL) assay (2, 6). Ile and Val may also be increased. Leu is also measured by liquid chromatography and an amino (2) or by thin-layer chromatography. With acid analyzer of Pediatrics, DiviMedical Center, Box 14991, Research Triangle Park, NC 27709. 2Department of Genetics, Magee Women’s Hospital, 300 Halkett St., Pittsburgh, PA 15213. ‘Author for correspondence. Fax 919-549-0709. abbreviations: MSUD, maple syrup urine dis4Nonstandard ease; MS-MS, tandem mass spectrometry; BIA, bacterial inhibition assay; PKU, phenylketonuria; RMSE, root mean square error; and C, collision-induced dissociation. Received July 13, 1994; accepted September 28, 1994. ‘Mass Spectrometiy Facility, Department sion of Biochemical Genetics, Duke University 62 CLINICAL CHEMISTRY, Vol. 41, No. 1, 1995 Materials and Methods Solvents, Reagents, and Internal Standards High-purity-grade methanol was obtained from Burdick and Jackson (Muskegon, MI). Glycerol, sodium octyl sulfate, Leu, lie, allolle, Val, and hydroxyPro (hydroPro) (St. Louis, MO). Butanolic were obtained from Sigma HC1 (3 mol/L) was obtained from Regis (Morton Grove, IL). Stable isotopes were from Cambridge Isotopes (Andover, MA) and include: [‘5N,, ‘3C1]Giy, [2H4AIa, [2H8]VaI, [2H3}Leu, [2H3]Met, [2H5]Phe, [2H4]Tyr, and [2H3]Glu. Blood Specimen Collection This laboratory has received >10 000 newborn blood spots from the State of North Carolina Division of Laboratory Services Newborn Screening Program. A collection of blood spots from newborns was received from the Department of Genetics, Magee Women’s Hospital, Pittsburgh, PA, and was analyzed by the BIA. These blood spots were categorized as normal (n = 5), positive for MSUD (n = 7), and false-positive for MSU1) (n = 23). Original blood spots were received from two siblings with MSUD who were subsequently treated at Duke University Medical Center. The first was diagnosed with MSUD at age 3 months after an episode of coma; her brother was diagnosed prenatally and was treated from birth. Plasma samples spotted onto newborn screening cards from normal subjects and patients previously diagnosed with MSUD were also obtained. All specimens collected or prepared at Duke or Magee Wo- men’s Hospital involved Type 903 filter paper from Schleicher and Schuell (Keene, NH). Note that the MSUD patients presented in this paper are not all from a specific regional population. MSUD) were compared with results from analyses with a Beckman amino acid analyzer (2). Potential interferences from hydroPro were examined by adding 150 and 300 tmo1/L of this compound to 1 mL of blood, from which blood spots were prepared. Sample Preparation The semiautomated preparation of butyl ester derivatives of amino acids from blood spots consists of a simple solvent extraction and derivatization procedure that takes -2.5 h for a batch of 60 samples. Two 0.19-in. (4.8-mm)-diameter dots [equivalent to 12.6 LL of whole blood (10)] were excised from a 0.5-in. (12.7-mm)-diameter dried blood spot into a 0.6-mL conical plastic vial. A methanol stock solution of internal standards was prepared, containing 2.5 mol/L each of [2H4]Ala, [2H8]Val, [2H3]Leu, [2H3}Met, [2H5]Phe, [2H4]Tyr, and [2H3]Glu; and 12.5 moI/L of [‘5N,,’3C1]Gly. This stock solution (400 tLL) was added to each vial in the 60-sample rack with a Gilson (Middleton, WI) Model 222 sample changer. The sample rack was placed on an orbital shaker for 30 mm. With the sample changer, the supernates were transferred to 1-mL flat-bottom vials in another rack and evaporated to dryness at 50#{176}C under a gentle stream of dry nitrogen with a custom-designed warm air incubator (Grey Line Engineering, Silver Spring, MD). Next, 50 L of 3 mol/L HC1 in n-butanol was added to each vial with the sample changer. The 60 vials contained in the rack were sealed with a bilayer cover made from a sheet of septum material (bottom layer) and a steel weight (top layer), placed in a forced air oven, and incubated at 65#{176}C for 15 mm. After removal of the bilayer cover, the vials were placed in the warm air incubator and excess HC1-butanol was evaporated to dryness under dry nitrogen. These derivatized samples were reconstituted with 35 L of a 1:1 (by vol) methanol:glycerol solution containing 1 gIL sodium cctyl sulfate, and the vials sealed with Teflon-lined caps. The samples were then ready for analysis by MS-MS. To estimate the linearity of this assay, we prepared 500-L aliquots of whole blood containing 0, 25,50, 100, 200, and 500 nmol of added Leu and Val. After mixing, these samples were spotted onto filter paper and dried overnight. To estimate recovery (extraction efficiency) of Lou and Val from blood spots, we separately added 0, 100, 200 and 400 nmol of each analyte to four 2-mL aliquots of whole blood, then divided these into two identical groups of four 1-mL aliquots. One group was spotted onto filter paper and dried overnight. The samples were prepared by using the extraction and derivatization procedure described above. To the second group, 80 nmol of PH3]Leu and [2H8]Val was added and mixed well. These whole-blood samples were spotted on filter paper and dried overnight. The samples were prepared by extraction with pure methanol containing no internal standards. The remainder of the sample preparation procedure was followed as described above. Assay variability was measured by performing five replicate analyses on 1 day (intraday variability) and five replicate analyses over a period of several weeks (interday variability). Results for plasma samples (control and Mass Spectrometry A VG Quattro triple-quadrupole tandem mass spectrometer with Lab-base data system (Fisons Instruments, Danvers, MA) was used, operated in the static liquid secondary ionization mode. This mode incorporates an ion source containing a cesium ion gun operating at 10 keV and a manually operated insertion probe. Positively charged molecules are detected after separation in the first mass analyzer region (MS 1) and final mass analyzer region (MS2). An intermediate quadrupole located between the first and third quadrupoles is used as the coffision region into which argon gas is introduced (11). Tuning the instrument was optimized by using a solution of deuterium-labeled standards prepared as butyl esters as described above. Molecules were protonated in the source to form [M+H] molecular ions. The production of ions in the source and separation of these ions was optimized to produce maximum intensity of unit mass resolution at the intermediate (MS 1) detector. The intermediate collision cell and the final quadrupole were optimized similarly, to achieve maximum intensity of individual ions separated at unit mass resolution at the final detector. After optimization of the signal from protonated molecular ions (precursor ions) for Leu, Ile, and Val, argon was introduced into the coffision cell until the intensity of the molecular ions was decreased 50%. This addition of argon induces the maximum degree of fragmentation (collision-induced dissociation, CID) of the molecular ions to form product ions. The product ion ([M-i-H]-102), corresponding to a loss of a neutral (uncharged) molecular fragment of 102 Da from the protonated molecular ion, was optimized to provide maximum sensitivity. The coffision energy used in the analysis of amino acids was 25 eV. These tuning indicators were used in the acquisition of all subsequent samples analyzed on that day. Tuning was usually carried out once daily. Only fine adjustments, if any, were necessary to provide optimum instrument performance on a day-to-day basis. All tandem mass spectra (precursor ion, product ion, or neutral loss scans) were acquired in the multichannel analysis (“summed continuum”) mode. In this mode, the intensity of each 1/16th fraction of a mass unit in the desired mass range is acquired and summed every second for 60 s to produce a single raw spectrum. The raw data are processed by using algorithms to smooth peak shape, subtract background noise, and perform centroid calculations, resulting in a list of ion masses with their absolute intensities. These steps result in better reproducibility through improved signal-to-noise and ion ratios. Product ion scans were produced by focusing MS1 on the molecular masses [M+H] of the butyl esters of either Leu (m/z 188), Ile CLINICAL CHEMISTRY, Vol. 41, No.1,1995 63 (m/z 188), or Val (m/z 174), allolle (m/z 188), and hydroPro (m/z 188), whereas MS2 was used to scan product ions between m/z 25 and 200. Amass spectrum showing the fragmentation for each molecular ion of Lou, He, and Val was obtained. Neutral loss scans of 102 Da were produced by scanning MS1 from m/z 125 to 300 while simultaneously scanning MS2 at a mass range 102 Da lower, m/z 23-198. This resulted in a spectrum of prod102Y. The mass uct ions corresponding to (M + H scale of the final data was adjusted to show the masses of molecular ions that correspond to each product ion in Fig. IA-C. A common loss of the elements of butyl formate (102 Da) was demonstrated by the presence of product ions at m/z 86 (Lou), 86 (Ile), and 72 (Val). The mass spectra of labeled precursor ions for the internal standards, [2H3]Lou and [2H8]Val, exhibited a similar fragmentation pattern with the presence of deuterated productions at m/z 89 and mlz 80, respectively (data not shown). The fragmentation process is shown schematically in Fig. 2, and involves a proton transfer to generate a stable carbonium ion that is apparently specific for a-amino acids. Other fragments seen in the product ion spectra include m/z 57, derived from the butyl group - mass. To quantify Lou, Ile, and Val, we determined the ion of Leu:[2H3]Lou (m/z 188:191), Ile: [2H3]Leu (m/z 188: 191), and Val:[2H8]Val (m/z 174: 182). The list of abundances for each ion was obtained from processed raw data. Lou and Ile are not separated by this method because they have identical masses. Similarly, molecular ions at m/z 188 can represent both albIle and hydroPro. Thus the ion signal at m/z 188 represents a combination of these potential isomers. Concentrations of Lou and Va! were calculated by reference to the appropriate calibration curves, prepared from the analyses of blood spots containing serially added known concentrations of Lou and Val as described previously. (Fig. abundance ratios Results Analysisof Branched-ChainAminoAcids by MS-MS of the protonated molecular ions, The fragmentation [M+H]4 (precursor ions) of Lou, Ile, and Val, are shown 100j Leucine A 1). Since Lou, Ile, and Val share a common neutral loss of 102 Da from the molecular ion, the spectrometer is set up in a way such that MS1 and MS2 are synchronized to scan the mass range with a constant 102-Da difference between them. Product ions (which differ by 102 daltons from the parent ion) are detected in MS2. Note that the mass of the parent ion is shown, although the product ion is actually detected. Fig. iD-F shows the neutral of the molecular ions (precursor ions) for loss spectra Lou, Ile, and Val, respectively. In addition, the neutral loss scan spectra also detects most other common amino previously (7). acids, e.g., Phe and Tyr, as described The sensitivity of the method depends in part on the efficiencies of ionization and CII) and is different for each amino acid. Furthermore, isomeric amino acids, by MS-MS. such as Lou and Ile, cannot be differentiated As shown in Fig. 1, the fragmentation of {M+H] ions of 100] 188 D 188 Leucine (DAU 188) (NTL 102) 50 50 86 40 - B 60 80 100 120 140 160 180 188 200 140 1001 160 E 180 200 220 260 280 300 Isoleucine Isoleucine (NTL 102) (DAU 188) C 240 188 504 50 r 1 - - --‘ 0L._ 40 100 60 80 100 120 140 C 160 180 200 140 1001 174 F 160 180 200 220 240 260 280 300 280 300 174 Valine Valine (DAU 174) (NTL 102) 50 50 72 [-102) 57 40 60 80 100 120 140 160 180 200 “ 140 160 180 200 220 240 260 Mass/Charge Mass/Charge Fig. 1. Product ion(daughterIon,DAU) mass spectra from the CID of IM+HP ions of Leu (A), lie (B), and Val (C) and mass spectra for neutral loss of 102 Da (NTL 102) [M+H-1O2J of Leu (D), lie (E), and Val (F) in the tandem mass spectrometer. A-C, fragmentation of the (M+H] ionsof Leu (m/z 188), lIe (m/z 188), and Val (m/z 174) to smaller positively charged ions such as m/z 86, 86, and 72, respectively. Fora descnption of the fragmentatIon processes see Fig.2. Thereisa common 102-Da difference between the original lM+H1 ion and its respective fragment ions. in D-Fthe scan function detects only IM+H1 that have fragment ions that differ by 102 Da. 64 CLINICAL CHEMISTRY, Vol. 41, No. 1, 1995 H3N 100 A RI + - C - Ala Pro COOC4H9 Leu+lIe Normal amino acid profile (NTL 102) * CID HCOOC4H9 50 Val Gly 8 >- H2N = C - R Phel His I Set a’ a’ C C 0 Leu: lie: R R Val: A = CH(CH3)2 = = CH2CH(CH3)2 CH(CH3)CH2CH3 140 a, 160 180 200 Glu J1 tA .1 C Tyr I Meth 220 240 260 280 300 > 100 B a, Leu+Ile MSUD amino acid profile Fig.2. Schematic representation of the specific fragmentation in the (NIL 102) tandem mass spectrometer that characterizes amino acid butyl ester derivatives. HCOOC4H9 is butyltormate, a neutral fragment of mass 102 Da common to the fragmentation of the iM+H1 ionsof Leu, lie, and Val. Lou and Ile is very similar. Thus, the signal at mlz 188 represents the sum of Lou, Ile, and other possible isomers. One of these possibilities is allolle. Increased concentrations of alloIle have been reported in patients with MSUD (1-3). Another possibility is hydroPro, which also has [M+H] ions equal to mlz 188. The relative ratios (response factors) of Lou, Ile, allolle, and hydroPro to [2H3]Lou internal standard were 1:1, 1:3, 1:3, and 2:3, respectively. AlloIle is not normally present in newborns and the mean concentration of hydroPro is only 10 mo1/L compared with 150 moWL for the sum of Lou + Ile (1, 2); the signal at mlz 188 is denoted as Lou + lie, and therefore this sum is the ion signal at mlz 188 in most cases. Figure 3A shows a normal blood amino acid profile from a newborn screening card obtained by using the neutral loss of 102 scan function. Ion signals at reprosentative masses of several amino acids and internal standards include m/z 174 (Val), m/z 182 ([2H8]Val), m/z 188 (Lou + Ile), and mlz 191 [2H3]Lou. Fig. 3B is an amino acid profile of a blood spot from an original card of a patient subsequently diagnosed with MSUD. The ion signals of (Lou + lie) and Val are obviously increased when compared with their respective internal standard ion signals (i.e., mlz 188 relative to mlz 191 for Lou + Ile and m/z 174 relative to m/z 182). Because all spectra are normalized to the highest ion signal intensity, ions signals from each analyte (amino acid) must be compared with the internal standard ion signal. Note that Ala is the second largest peak in the normal profile in Fig. 3A. Although Ala is considerably smaller in ion intensity in the MSUD profile in Fig. 3B, its intensity standard at mlz 150 is relatively relative to its internal unchanged. 50 Val GIy 140 Pro Set 160 Met ‘Phe 180 200 220 Tyr 240 Glu 260 280 300 Mass/Charge Fig. 3. MS-MS amino acid profiles from a normal neonatal screening card (A) and from an original neonatal screeningcard of a patient with MSUD (B) obtainedby usingthe neutralloss of 102 Da scan function. Asterisks represent deuterated amino acid internal standards. Quantitation Is achieved by comparing the peak height of the amino acid of interest with their respective internal standards. Each spectrum is normalized to the largest Ion signalpresent. the masses of their respective deuterium-labeled internal standards were plotted as a function of the concentration of added Leu and Val to blood. The calibration curve for added Lou was linear over the concentration range 0-1000 j.tmoIIL: slope = 0.136 (SE 0.000175), intercept = 2.17 (SE 0.0395), and r2 = 0.998 [root mean square error (RMSE) = 0.1041; for added Val, linearity was observed over a concentration range of 0-1000 pinol/L: slope = 0.138 (SE 0.0002144), intercept = 2.99 (SE 0.0482), and r2 = 0.998 (RMSE = 0.127). The individual concentrations of Lou or Ile cannot be determined, because of the lack of separation by MSMS. However, certain assumptions described above let us more closely approximate the sum of the concentrations of Leu + Ile. The contribution of Ile to the ion signal at mlz 188 is one-third the contribution of Lou to that same ion signal; thus, for an equimolar amount of Lou and lIe, 25% of the total ion signal represents lie. Given that the relative ratios of the amounts of Lou and lie is 2:1 in normal healthy newborns, the best approximation for a corrected sum of Lou + lie is to apply a correction factor of 1.3 to the ratio of Lou + Ile to [2H3]Lou Assay Calibration and Umits of Detection Calibration curves for Lou and Val were generated by using standard isotope dilution techniques (12, 13). Accurate amounts of Lou and Va! were added to blood over the calibration range of interest with fixed amounts of internal standards. Ion signal ratios of Lou and Val to Ala Typical trations (internal standard). signal-to-noise ratios for endogenous concenof Lou + lie and Val in normal blood samples were 140:1 and 40:1, respectively. The estimated detection limits, based upon the signal-to-noise ratio of 3:1, correspond to concentrations of 2 mol/L for Lou + Ile and 6 jimol/L for Val. These are well below normal CLINICAL CHEMISTRY, Vol.41, No. 1, 1995 65 physiological concentrations of 150 mol/L and 150 imol/L for Val (2). for Lou + Ile Table 1. ComparIson of MS-MS wfth HPLC In analysis of plasma samples. Concentration, Analytical Recovery, Precision, and Accuracy The analytical recoveries of added Lou and Val to blood were determined in triplicate at concentrations of 0, 50, 100, and 200 tmoIJL. The respective mean ± SD values obtained for added Lou and Val were 101% ± 1% and 105% ± 1% at 0 moI/L, 92% ± 2% and 96% ± 4% at 50 moWL, 86% ± 3% and 90% ± 1% at 100 smoI/L, and 97% ± 4% and 98% ± 4% at 200 p.mol/L. Precision of the assay was calculated by replicate analyses of the same normal blood sample by performing the complete analytical procedure on the same day and on different days. The within-day CVs were 2.5% for Lou + lie, 3.6% for Val, 1.5% for the ratio of (Lou + Tie) to Phe, and 3.1% for the ratio of Val to Phe (n = 5). The between-day CVs, determined over 2 weeks, were 11.5% for Lou + Ile, 9.5% for Val, 5.6% for the ratio of (Lou + fle) to Phe, and 4.5% for the ratio of Va! to Phe (n = 5). The estimated concentrations of (Lou + Ile) and Val in respectively. these samples were 145 and 225 moI/L, The reliability of this method was determined by comparing the quantitation of (Lou + Tie) and Val by MS-MS with HPLC in control plasma and plasma obtained from two siblings with MSUD. (Data were not available for Va! in selected controls.) These results are presented in Table 1. An increase of m/z 188 due to the presence of increased blood hydroPro is possible. The addition of 75 hydroPro control blood resulted in apand 150 moJ/L parent Lou + Tie concentrations of 205 and 266, respectively. The control blood Lou + ile concentration was 176 LmoI/L. These increased concentrations of hydroPro still provide signals in the normal range for Lou + Tie. Higher crease of hydroPro could potentially into >300 mo1/L, resulting in a falsely positive result. Comparison of the product ion spectrum of hydroPro (Fig. 4) with the product ion spectrum for Lou (Fig. 1A) reveals a fragment ion at mlz 68 unique to hydroPro. This characteristic fragment represents loss of water and butyi formate from the parent molecular ion, a useful indicator for testing the contribution of hydroPro to the mlz 188 signal. In daughter ion spectra from a control patient, a patient with added (150 JLmoI/L), the ratios of mlz 68 to mlz 188 hydroPro were 0.0023, <0.0010, and 0.19, respectively. Analysisof BloodSpecimen Collections We obtained >10 000 fresh biood spots through the North Carolina State Screening Laboratory; most were from three regional hospitals. The samples had been analyzed for PKU by the State Screening Laboratory and were considered normal. In this group the mean ± SD (range, CV) for Lou + lie was 151 ± 47 mo1/L (30-673, 40%) (n = 1096); for Va!, 131 ± 58 (33-586, 44%) (n = 791). The (Lou + He) to Phe ratio was 2.5 ± 25%). The Va! to Phe ratio was 2.18 ± 0.49 (0.86-5.0, 0.51 (0.95-4.58, 23%). The ranges reported in literature 66 CLINICALCHEMISTRY, Vol. 41, No. 1, 1995 HPLC Days posttreatinent (Leu II.)e + Vii Leri+lie Vii Controls 1 266 2 188 3 201 4 286 221 5 6 na na 149 9 10 260 78 Patient A 229 174 174 na na na na 253 136 7 8 218 na 139 na na na na 144 na na 242 + 67 na 179 279 200 (MSUDdiagnosedprenatally) EstimatedLeu Leu n&’ na na na na na 222 235 2 199 167 225 3 344 4 304 312 5 101 512 6 148 981 Patient B (MSUD diagnosedat age 3 months) 1 2860 450 7 702 180 9 494 120 + 170 271 265 126 380 418 791 3116 324 426 389 79 89 106 11 414 142 290 12 13 296 202 153 295 157 201 183 1640 14 15 225 177 236 205 224 160 225 192 16 150 209 152 184 lie with correction factor of 1.3 from the original estimated This factor is necessary to approximate theotherwise lieconcentration. Leu + lIe. na, not available (see mean data for 1000 samples). underestimated total b concentrations the Lou + lie MSUD, and a control patient sample containing Mmoi/L MS-MS for neonates, including premature infants, are Lou, 48220 j.&mol/L; lie, 26-91 jmoI/L; (Lou + lie), 74-311 .Lmo1/L; and Val, 86-220 moI/L. The mean ± 5SD of the data, which includes >99.9% of the samples, compares well with the range of literature values with an 100 a’. 188 Hydroxyproline (DAU 188) a’ C a’ 86 50 a, > 5, 57 ___ 40 60 68 132 80 100 120 140 160 180 200 Mass/Charge Fig. 4. Product ion mass spectra from the CID of IM+HI ions of hydroPro. This spectrumrepresentsthe fragmentationof the [M+HJ ion of hydroPro (m/z 188) to smaller fragmentsat mhz 132, 86. 68, and 57. One fragment unique to hydroPro is m/z 68, representingthe 120-Da loss corresponding to water (18 Da) plusbutyl forrnate (102 Da). unknown saxnpie population. Of these 10 000 samples, one was found with a clearly abnormal Val concentration (717 tmol/L). Information obtained upon follow-up revealed that this neonate was prenatally diagnosed with MSUD and was under treatment. Further studies with this patient are described below. The collection of blood spots received from Magee in a blinded fashion at Women’s Hospital was analyzed the Duke Mass Spectrometry Facility. Results for (Lou + Ile), Val, and the ratios of (Lou + He) to Phe and Val to Phe are presented in Table 2. The samples were categorized at normal, false positive by BLA, and con- Table 2. QuantItative analysis of Leu + lie and Val In blood spots. Conc, imoI/L Sample Leu + Ilea Ratios Val (Lair + lls)IPheb Vat/Ph. includes five control samfirmed MSUD. This collection ples from normal patients, six samples from patients diagnosed with MSUD by BJA (one sample was collected at 17 h), reported normal and 23 samples in which Lou was falsely increased by BLA and subsequently found to be upon repeat analysis or follow-up. For the con- trol group and true MSUD-positive samples (BIA cutoff, 300 j.mo1/L Lou), both assays were in good agreement. The ratio of (Lou + lie) to Phe was also in the normal range. A sample obtained from a MSUD patient at only 17 h postpartum was also positive for MSUD. Amino acids of two siblings with MSUD were analyzed. The original PKTJ card was obtained from the sibling diagnosed with MSUI) after hospitalization. For a subsequent sibling diagnosed with MSUD prenatally and given close clinical and dietary management at birth, PKU cards were obtained on days 1, 2, and 3 postpartum. Results for plasma obtained from these patients during treatment are summarized in Table 1. Control group I 2 3 4 5 Control group 2 1 2 3 119 58 135 52 2.9 3#{149}3 Discussion 2.2 77 75 3.2 4.1 2.0 1.8 122 112 110 121 (false positive by BIA) 408 262 139 117 188 247 2.1 In the new use of MS-MS in neonatal screening for amino acid metabolic disorders, including MSUD, samples are prepared by a simple semiautomated extraction 2.9 3.0 and derivatization 2 20 instrument time required for each analysis is <3 min. The incorporation of isotope-dilution techniques provides quantitative information for specific components 4 145 142 5 6 7 8 9 10 11 157 205 148 130 145 145 140 134 3.9 2.3 4.4 2.9 3.6 161 2.1 28 31 16 105 131 118 107 2.8 3.9 2.0 2.2 2.3 12 142 115 1.9 2.7 1.7 2.0 13 14 15 16 17 18 130 3.6 3.9 2.4 104 83 3.5 2.2 234 191 2.7 2.2 213 114 135 146 3.8 115 136 2.2 2.6 2.2 2.5 166 131 19 89 77 20 103 21 138 74 103 22 143 129 23 192 117 2.5 3.0 3.7 2.5 2.8 4.1 3.6 2.4 1.8 1.8 2.1 2.1 2.2 MSUD-positive (collected 24 h posoartum) 1 765 174 2 748 404 3 4 5 6 756 186 647 359 15.2 11.7 18.3 21.6 569 1019 404 18.0 1162 18.2 MSUD-positive 1 303 9.0 + method is robust of samples of 60. Total and capable of because but not clearly abnor- 63 mal. We therefore consider the combined results for Lou + He ratio and Val to Phe ratio a better index for 120 13:0 20.8 diagnosis of MSUD than the measurement of blood Lou concentration only. The single abnormal sample detected in this popula- 8.0 + lie with correctionfactorof 1.3 from theoriginalestimated lie concentration. b (Leu + lle)/PheIsthe corrected concentration (zmoi/L) of Leu + lie divided by the concentration(moI/L) of Phe. Estimated Leu Leu This numbers in batches of its nonnature. Butyl esters of Lou, lie, and Va! are detected with excellent sensitivity and selectively in blood and plasma by MS-MS, as has been described previously for Phe and Tyr (7). This laboratory has analyzed >10 000 newborn blood sainples,with the mean ± 5SD concentrations being within the published range for newborns and prematures for Lou + Ile and Val. This method appears to be sensitive, precise, and accurate. In addition, 22 previously diagnosed false-positive samples, when analyzed and by MS-MS, had normal (Lou + He) concentrations normal (Lou + Ile):Phe ratios; another did have an increased blood (Lou + Ile) concentration but a normal ratio to Phe. Six patients previously diagnosed with MSUD were analyzed in a blinded fashion; all were detected as positive for MSUD. One sample in this group, collected at 17 h, was diagnosed as MSUD by of using the ratio of Lou + Ile to Phe. The concentration large analyzing chromatographic Lou in that sample was increased (collected 17 h posloartum) 344 of each sample. procedure tion revealed an abnormally high Va! concentration. Follow-up analysis determined that this patient had been diagnosed prenatally with MSUD and was already under dietary management, including Lou restriction, when the sample was collected. Newborn screening methods that measure only Lou concentration in blood, CLINICAL CHEMISTRY, Vol. 41, No.1,1995 67 i.e., BIA, would likely not have detected this abnormal result. This is a clear demonstration of the utility of using methods that measure several amino acids in one test. Quantitative amino acids analysis with most published HPLC techniques also have the ability to separate amino acids and diagnose multiple diseases in one test. However, the speed of the HPLC assay is a limitasince most assays average tion in newborn screening 15-30 mm per run. With MS-MS, the assay time is <4 min. The inability to distinguish between Lou and Ile is of this method. Further, the contribuone limitation tions of each species to the mass intensities is not equal. As the data reveal, for the vast majority of newborn samples a single correction factor applied to our final result provides a good estimate of the true Lou and Ile concentrations. Evaluation of potential interference from hydroPro showed it to be relatively insignificant for several reasons. First, the amount of hydroPro in <150 j.&mol/L (1, 2). Secnormal patients is generally ond, the response factor of hydroPro is -0.66% relative to Lou + lie. Thus, even an apparent increase of -150 tmoI/L of hydroPro would increase the false-positive rate only slightly if the MSUD diagnosis were based on the concentration of Lou + Ile only. Third, in those cases where there is an apparent increase of the Lou + lie concentration, a specific analysis can be performed that The absence detects a fragment ion unique to hydroPro. of this fragment confirms an increased Lou + Ile and the diagnosis of MSUD. thus validates MS-MS increases the specificity of newborn screening for MSUD. Its sensitivity appears good, and its accuracy is apparent by the lack of false positives. Success of these criteria should be judged on the basis of the very low false-positive rate, very low false-negative rate, and excellent correlation of the quantitation of Lou + Ile with other methods. We believe that MS-MS applied to newborn screening is a cost-effective methodology for the following reasons: (a) The potential exists for adding new tests without incurring additional costs because the same equipment is utilized; (b) efficiency is increased when only one method is used to test for multiple diseases; and (c) automated sample preparation, analysis, and interpretation, which are under development, will increase the per day, number of samples analyzed per instrument further reducing the cost per test. As the validation of the diagnostic indicators for each disease is accom- 68 CLINICAL CHEMISTRY, Vol. 41, No. 1, 1995 plished, as described previously MSUD, we are providing strong for PKU and here for evidence that the accuracy and precision of MS-MS results in very low falsepositive rates compared with existing methods. This reduce the overall costs of newinevitably will further born screening since the need for retesting of presumptive positives will be much reduced. We thank Sheryl Barker, Steven Worthy, and Stephanie To!liver for their expert technical assistance. Financial support was from the State of North Carolina, l)ivision of Maternal and Child Health, Department of Environmental Health and Natural Resources (Raleigh, NC, grant no. C-05070). References 1. Danner DJ, Elsas LI II. Disorders of branched chain amino acid and keto acid metabolism. In: Shriver CR, Beaudet AL, Sly WS, Valle D, eds. The metabolic basis of disease. New York: McGraw Hill, 1989:671-92. 2. Slocum RH, Cummings JG. Amino acid analysis of physiological samples. In: Hommes FA, ed. Techniques in diagnostic human biochemical genetics: a laboratory manual. New York: John Wiley & Sons, 1991:87-126. 3. Saudubray JM, Poggi F, Spada M, Billette de Villemeur T, Hubert P, Charpentier C, et al. A programmed clinical screening for inborn errors of metabolism in neonates. In: Farriaux JP, Dhont JL, eds. Proceedings, 9th international neonatal screening symposium. Excerpta Med mt Congr Ser 1994;1041:85-100. 4. Nyhan WL, Rice-Asara M, Acosta P. Advances in the treatment of amino acid and organic acid disorders. In: Desmck RJ, ed. Treatment of genetic diseases. New York: Churchill Livingstone, 1991:45-68. 5. Levy H. Nutritional therapy in inborn errors of metabolism. In: Desmck RJ, ed. Treatment of genetic diseases. New York: Churchifi Livingstone, 1991:1-22. 6. Naylor EW. Newborn screening for maple syrup urine disease. In: Therrell BL Jr, ed. Laboratory methods for neonatal screening. Washington, DC: Am Public Health Assoc, 1993:115-24. 7. Chace DH, Milhington DS, Terada N, Kahler CR, Hofman LF. Rapid diagnosis of phenylketonuria by quantitative analysis for phenylalanine and tyrosine in neonatal blood spots by tandem mass spectrometry. Clin Chem 1993;39:287-90. 8. Millington DS, Kodo N, Terada N, Roe D, Chace DH. The analysis of diagnostic markers of genetic disorders in human blood and urine using tandem mass spectrometry with liquid SIMS. Int J Mass Spectrum Ion Process 1991;111:211-28. 9. Millington DS, Chace DH, Hillman SL, Kodo N, Terada N. Diagnosis of metabolic disease. In: Matsumoto T, Caprioli R, Gross ML, Seyama Y, eds. Biological mass spectrometry: present and future. New York: John Wiley & Sons, 1994:559-79. 10. National Committee for Clinical Chemistry Standards. Blood collection on filter paper for neonatal programs: approved standard. NCCLS Publication LA4-A, Villanova, PA: NCCLS, 1988. 11. McLafferty FW. Tandem mass spectrometry. Science 1981; 214:280-7. 12. Markey SF. Quantitative mass spectrometry. Biomed Mass Spectrom 1981;8:426-30. 13. Fa]kner F. Comments on some common aspects of quantitative mass spectrometry. Biomed Mass Spectrum 1981;8:43-6.