Analysis of Human Cerebrospinal Fluid by Capillary Electrophoresis

Analysis of Human Cerebrospinal Fluid by Capillary Electrophoresis with Laser-Induced Fluorescence Detection. Jonas. Bergquist, S. Douglass. Gilman ...
0 downloads 0 Views 848KB Size
Download PDF
Anal. Chem. 1994,66, 3512-3518

Analysis of Human Cerebrospinal Fluid by Capillary Electrophoresis with Laser- Induced Fluorescence Detection Jonas Bergquist,t S. Douglass Gilman,* Andrew G. Ewing,'** and Rolf Ekman7 Department of Clinical Neuroscience, Section of Psychiatry and Neurochemistry, Goteborg University, Molndal Hospital, S-43 1 80 Molndal, Sweden, and Department of Chemistry, 152 Davey Laboratory, Penn State University, University Park, Pennsylvania 16802

Capillary electrophoresis with laser-induced fluorescence detection is used to analyze 10 pL samples of human cerebrospinalfluid. Primary amine-containing compounds in untreated cerebrospinal fluid are labeled with 3-(4-carboxybenzoyl)-2-quinolinecarboxaldehydeprior to analysis, producing fluorescent isoindoles. Electropherograms containing approximately 50 peaks are obtained in less than 35 min from cerebrospinal fluid samples. Ten peaks in the electropherograms have been identified and quantitated: arginine, glutamine, threonine, valine, y-amino-*butyric acid, serine, alanine, glycine, glutamic acid, and aspartic acid. Detection limits for these 10 amino acids range from 0.29 nM for y-amino-* butyric acid to 100 nMfor threonine, and separationefficiencies as high as 190 000 theoretical plates are obtained for these analytes. Electropherograms of cerebrospinal fluid samples from patients with Alzheimer's disease and from children with different neurological disorders are compared to those of healthy controls. Differences in individual amino acid levels are observed between the patient groups, and these differences appear to be disease and age related. These results indicate that analysis of cerebrospinal fluid by capillary electrophoresis will be useful as a selective, rapid, and sensitive tool for both diagnosis of central nervous system disorders and for study of the mechanisms of these disorders. Capillary electrophoresis offers several advantageous characteristics for the analysis of biological fluids, and its emergence as an effective tool for clinical analysis has been the subject of recent reviews.14 The high separation efficiencies obtained with capillary electrophoresis result in large peak capacities, high resolving power, and very short analysis times. Analysis of substances in biological fluids requires high selectivity in order to quantitate particular analytes without interference from other species, and large peak capacities are desirable in order to analyze many species of interest simultaneously. The small sample volumes required for capillary electrophoresis (femtoliter to nanoliter), along with the ability of capillary electrophoresis to determine multiple analytes, reduce the amount of sample consumed, t Gdteborg University.

* Penn State University.

(1) Chen, F.-T.A.; Liu, C.-M.;Hsieh, Y.-Z.;Sternberg, J.C. Clin. Chem. 1991, 37, 14-19.

(2) Guzman, N. A.; Gonzalez, C. L.; Hernandez, L.; Berck, C. M.; Trebilccck, M. A.; Advis, J. P. In Capillary Electrophoresis Technology; Guzman, N. A., Ed.; Marcel Dekker, Inc.: New York, 1993; Chapter 22. (3) Shihabi, 2. K. Ann. Clin. Lab. Sci. 1992, 22, 398-405. (4) Thormann, W.; Molteni, S.;Caslavska, J.; Schmutz, A. Electrophoresis 1994, 15, 3-12.

3512

Analytical Chemistry, Vol. 66, No. 20, October 15, 1994

leaving more sample for other types of analysis. Also, analysis of biological fluids with capillary electrophoresis often requires minimal sample processing before injection, and many biological samples can be directly injected into the capillary without adverse effects.14 Finally, the short analysis times obtained with capillary electrophoresis and the ease of automation of the technique can savevaluable time and labor. One potential disadvantage of the small volumes used with capillary electrophoresis is that concentration sensitivity for detection is compromised by extremely small detection volumes. This is particularly true for optical detection methods such as UV absorbance and refractometric detection where the short optical path length provided by the capillary results in poor concentration sensitivity relative to larger volume techniques. While the sensitivity offered by UV detectors commonly used with capillary electrophoresis is sufficient to detect a variety of analytes, many important species present in biological fluids are found in only trace quantities or do not absorb strongly enough and will go undetected. Laser-induced fluorescence (LIF) detection is one of the most successful methods used to overcome concentration detection insensitivity for capillary electrophoresis. Concentration detection limits on the order of 10-l2M can be obtained with LIF detection of fluorescently labeled analytes using a number of detector configuration^.^-^ Native fluorescence with UV lasers can also be used to obtain very low detection limits for a variety of a n a l y t e ~ ; ~however, -l~ many interesting analytes are not natively fluorescent, and most reports of LIF detection with capillary electrophoresis involve the use of visible lasers and labeling of analytes with fluorescent functional groups. One derivatization reagent successfully used with capillary electrophoresis and LIF detection is 3-(4-carboxybenzoyl)2-quinolinecarboxaldehyde (CBQCA), which has recently been developed by Novotny and co-workers.13 When allowed to react with a suitable nucleophile and compounds containing ( 5 ) Sweedler, J. V.; Sheer, J. B.; Fishman, H. A.; Zare, R. N.; Scheller, R. H. Anal. Chem. 1991,63, 496-502. (6) Hernandez, L.; Escalona, J.; Joshi, N.; Guzman, N. J. Chromatogr. 1991, 559, 183-196. (7) Wu, S.;Dovichi, N. J. Talanra 1992, 39, 173-178. (8) Amankwa, L. N.; Albin, M.; Kuhr, W. G. Trends Anal. Chem. 1992, 11, 114-120. (9) Lee, T. T.; Yeung, E. S . J. Chromatogr. 1992, 595. 319-325. (10) Milofsky, R. E.; Yeung, E. S . Anal. Chem. 1993.65, 153-157. (1 1) Chan, K. C.; Janini, G. M.; Muschik, G. M.; Issaq, H. J. J. Lig. Chromatogr. 1993, 16, 1877-1890. (12) Nie, S.;Dadoo, R.; Zare, R. N. Anal. Chem. 1993,65, 3571-3575. (13) Liu, J.; Hsieh, Y.-2.;Wiesler, D.; Novotny, M. Anal. Chem. 1991,63,408-

412.

0003-2700/94/0366-3512$04.50/0

0 1994 Amerlcan Chemical Society

primary amines, CBQCA forms highly fluorescent isoindole products in a manner analogous to the behavior of ophthaldialdehyde and naphthalene-2,3-dicarboxaldehyde.14 There are several important advantages realized by the use of CBQCA, including detection limits in the nanomolar range, excellent linearity, compatibility with micellar electrokinetic capillary chromatography, fast reaction time, andgood product ~tabi1ity.l~Background fluorescence is minimal because CBQCA itself does not fluoresce, and, therefore, no purification of CBQCA derivatives is required prior to analysis. Also, CBQCA can be used with both HeCd (442 nm) and argon ion lasers (488 nm) which are available with commercial capillary electrophoresis-LIF instruments. Because CBQCA is readily available, is easy to use, and provides very low detection limits for primary amines, it is an excellentprecolumn derivatization reagent for biological fluids. Cerebrospinal fluid is secreted in the brain, primarily by the choroid plexus, and cerebrospinal fluid is in a steady state with the extracellular fluid surrounding brain cells. Cerebrospinal fluid plays a critical role in providing a constant chemical environment for neurons and glia, and it is the body fluid most likely to reflect disturbances of the central nervous system. The molecules found in cerebrospinal fluid cover a wide spectrum including primary amine-containing species such as amino acids, amino sugars, short peptide fragments, neuropeptides, and larger proteins. Using HPLC, ionexchange chromatography, and isotachophoresis in large capillaries, changes in amino acid levels in cerebrospinal fluid have been found for a number of diseases including chronic schi~ophrenia,'~ multiple sclerosis,16Parkinson's syndrome,17 Alzheimer's disease (AD),18 and other important central nervous system disorders. 19-24 These reports emphasize the need for the development of simple and rapid techniques for the study of primary amine-containing compounds in cerebrospinal fluid. Analysis by HPLC and ion-exchange chromatography are very effective tools for cerebrospinal fluid analysis, but separations can be time consuming, and substantial sample treatment is often used prior to analysis.15-17q20-23Isotachophoresis in large capillaries has been used to determine only one amino acid quantitatively in cerebrospinal If suitable methods were available, patients with a variety of neurological disorders could be more rapidly diagnosed, and the roles of particular molecules in these diseases could be more easily studied. As discussed earlier, capillary electrophoresis with LIF detection is well suited to the analysis of biological fluids, including cerebrospinal fluid. A few reports on the analysis of human (14) Roach. M. C.: Harmonv. M. D. Anal. Chem. 1987. 59.411415. ( l 5 j Korpi, E. R.;Kaufmann,E. A.; Marnela, K.-M.; Weinberger, D. R.Psychiatry Res. 1987, 20, 337-345. (16) Qureshi, G. A.; Baig, M. S. J . Chromatogr. 1988, 459, 237-244. (17) Gjessing, L. R.; Gjesdahl, P.; Dietrichson, P.; Presthus, J. Eur. Neurol. 1974, 12,33-37. (18) Basun, H.; Forssell, L. G.; Almkvist, 0.;Cowburn, R. F.; Eklof, R.; Winblad, B.; Wetterberg, L. J . Neural Transm. 1990, 2, 295-304. (19) Perry,T. L.; Krieger, C.;Hansen, S.; Eisen, A. Ann. Neurol. 1990.28, 12-17. (20) Hagberg, H.; Thornberg, E.; Blennow, M.; Kjellmer, I.; Lagercrantz, H.; Thiringer, K.; Hamberger, A,; Sandberg, M. Acta Paediafr. 1993,82,925929. (21) Manyam, B. V.; Giacobini, E.; Ferraro, T. N.; Hare, T. A. Arch. Neurol. 1990, 47, 1 1 9 6 1199. (22) Kaakkola, S.; Marnela, K.-M.; Oja, S.S.; Icbn, A,; Palo, J. Acta Neurol. Scand. 1990,82, 225-229. (23) Halawa, I.; Baig, S.;Qureshi, G. A. Biomed. Chromarogr. 1991,5,216-220. (24) Hiraoka, A,; Miura, I.; Tominaga, I.; Hattori, M. Clin. Biochem. 1989, 22, 293-296.

cerebrospinal fluid by capillary electrophoresis using UV and LIF detectionZ5have been published, but very few peaks in the resulting electropherograms have been identified, and little or no quantitative data have been presented in these studies. Capillary electrophoresis has been used to quantitatively determine compounds in dialysates of cerebrospinal fluid from rats using UV electrochemical dete~tion,~' and LIF detection.28 In this paper we present data from a quantitative, sensitive, and rapid analysis of human cerebrospinal fluid obtained using capillary electrophoresis with LIF detection and precolumn CBQCA derivatization. Ten of the approximately 50 peaks obtained in separations of these cerebrospinal fluid samples have been identified and quantitated. Separation efficiencies, detection linearity, and detection limits are reported for the identified peaks. Values obtained for amino acids in cerebrospinal fluid from healthy adult patients, patients with Alzheimer's disease, healthy elderly patients, and children with different neurological disorders are compared. These results are used to assess the potential of this method as a diagnostic tool for neurological disorders and for the study of the mechanisms of these central nervous system disturbances.

EXPERIMENTAL SECTION Apparatus and Separation Conditions. A PIACE 2100 instrument (Beckman Instruments; Fullerton, CA) equipped with a P/ACE-LIF detector was used to obtain all capillary electrophoresis separations. The LIF detector used an argon ion laser for excitation at 488 nm (4 mW), and a 560 f 40 nm band-pass filter provided with the LIF detector was used to reject scattered laser light for fluorescence detection. Fused silica capillary (75 pm i.d., 360 pm 0.d.) was obtained from Polymicro Technologies (Phoenix, AZ), and all separations were performed in 57 cm lengths (50 cm to detector) of untreated capillary. The separation buffer consisted of 50 mM borate and 30 mM sodium dodecyl sulfate (SDS) in ultrapure water (Milli-Q UFPlus; Millipore Corp., Milford, MA). The pH was adjusted to 9.0 by addition of sodium hydroxide, and dimethyl sulfoxide (DMSO) was added to this solution (20% DMSO vol/vol) to form the final separation buffer. The instrument was programmed to rinse the capillary with 0.1 M NaOH at high pressure for 2.0 min and then with separation buffer at high pressure for 3.0 min prior to each injection in order to obtain consistent migration times from run to run. Samples were injected automatically by pressure injection for 1.0 s (5.9nL volume). A separation potential of 20 kV was used, and the capillary was held at 25 "C. The electrophoretic current was typically 33 pA. Data Collection and Analysis. Data collection, processing, and analysis were performed using the System Gold software package (Beckman). Data were collected at a sampling rate of 5 Hz. Peaks were identified by spiking cerebrospinal fluid samples with standard solutions of amino acids. Amino acids (25) Jansson, M.; Roeraade, J.; Laurell, F. Anal. Chem. 1993, 65, 2766-2769. (26) Tellez, S.;Forges, N.; Roussin, A.; Hernandez, L. J . Chromalogr. 1992,581, 257-266. (27) OShea, T. J.; Weber, P. L.; Bammel, B. P.; Lunte, C. E.; Lunte, S. M.; Smyth, M. R. J . Chromatogr. 1992, 608, 189-195. (28) Hernandez, L.; Escalona, J.; Verdeguer, P.; Guzman, N. A. J . Liq. Chromatogr. 1993, 26,2149-2160.

Analytical Chemistry, Vol. 66, No. 20, October 15, 1994

3513

were quantitated using linear calibration curves based on peak area. Each calibration curve contained data points at a minimum of six different concentrations, and each curve spanned the range of concentrations found in cerebrospinal fluid samples. Somatostatin (3-6) was used as an internal standard. Because the internal standard contains two primary amine groups due to the presence of a lysine residue, two peaks are observed after derivatization with CBQCA. Peak areas were normalized to the combined area of the two peaks, and this was experimentally determined to give consistent values over a large number of injections. All errors are reported as the standard error of the mean. Linearity was assessed using standard least-squares analysis of peak area vs concentration plots. Detection limits were estimated at 2 times the peak to peak noise by extrapolation from plots of peak height vs concentration. Separation efficiencies were calculated using peak widths at 10% of the peak height. Comparisons of amino acid levels in different samples were made using the standard two-tailed t test. DerivatizationProcedures. Untreated human cerebrospinal fluid was directly derivatized for capillary electrophoresisLIF analysis. One microliter of internal standard [somatostatin (3-6), 198 p M in HzO], 5 pL of CBQCA (3 mg/mL in MeOH), and 2 pL of KCN (50 mM in 50 mM borate buffer, pH 9.0) were added to 10 pL of cerebrospinal fluid in 100 pL plastic vials provided for use with the capillary electrophoresis instrument. The vial housings were filled with HzO in order to reduce solvent evaporation during the reaction time. Solutions were left to react for 2.0 hat room temperature before direct injection from the reaction vial without purification of the reaction products. Solutions containing standards were made in H2O and were derivatized under conditions identical to those used for cerebrospinal fluid samples. Samples used for examination of amino acid recoveries from cerebrospinal fluid samples were derivatized as described for other cerebrospinal fluid samples except that 5 pL of a standard solution of amino acids was added to 5 pL of cerebrospinal fluid (from CCSF), and 5 pL of H2O was added to the control cerebrospinal fluid sample (5 p L from CCSF). These solutions were allowed to equilibrate for 15 min at room temperature before addition of CBQCA and KCN. Chemicals. All chemicals were used as received. AttoTag CBQ amine derivatization kit (CBQCA) was purchased from Molecular Probes, Inc. (Eugene, OR). Amino acid standards and y-amino-n-butyric acid (GABA) were obtained from Sigma Chemical Co. (St. Louis, MO). The synthetic peptide fragment used as an internal standard [somatostatin (3-6), H-Cys-Lys- Asn-Phe-OH] was a gift from Professor VBcsei, Department of Neurology and Psychiatry, SzentGyorgi University Medical School, Szeged, Hungary. All other chemicals used were from E. Merck (Darmstadt, Germany). Human Cerebrospinal Fluid Samples. The investigation was approved by the Human Ethics Committee at the Faculty of Medicine, Goteborg University, Sweden. All patients gave their informed consent to participate in these studies after the nature of the procedure was fully explained. Lumbar punctures were performed in the L3-4 or L4-5 interspace in the morning and with the patient in a recumbent position. The first 12 mL of the cerebrospinal fluid (3 mL for children) 3514

Analytical Chemistry, Vol. 66, No. 20, October 15, 1994

was collected in plastic tubes and gently mixed to avoid gradient effects. Blood samples were collected at the same time. Samples of cerebrospinal fluid and serum were stored at -20 "C until analysis. Quantitative determination of albumin in cerebrospinal fluid and serum was performed by rocket immunoele~trophoresis.~~ The albumin ratio [cerebrospinal fluid albumin (in mg/L) divided by serum albumin (in g/L)] was used as a measure of blood-brain barrier function30 in order to exclude samples from patients with pathologically increased levels of compounds from serum in their cerebrospinal fluid. A pooled control sample (pool CCSF) consisted of a pool of 80 human cerebrospinal fluid samples sent to the laboratory for protein analysis. These samples did not exhibit pathological values for albumin or immunoglobulins G and M. All samples in the pool were from individuals over 15 years of age. The AD age-matched healthy control pool (pool AMCCSF) consisted of five samples (four males and one female; 66-70 years of age; mean age, 67.4 f 0.7 years) from patients admitted for minor surgery involving spinal anesthesia. None of these patients had histories, symptoms, or signs of psychiatric or neurological disease or systematic disorders. The cognitive status of each patient was evaluated using the Mini-Mental State e ~ a m i n a t i o n .Individuals ~~ with scores below 28 were excluded. The Alzheimer's type I patient pool (pool ADCSF) consisted of five samples (two males and three females; 60-68 years of age; mean age, 62.8 f 1.6 years; duration of dementia, mean, 5.8 f 1.5 years). All patients underwent thorough clinical investigation, including a medical history, physical examination, neurological and psychiatric examinations, screening laboratory tests, electrocardiography, chest roentgenography, electroencephalography, and computer tomography of the brain. Patients with a history of psychiatric or neurological disorders, alcoholism, or major head trauma were excluded, as were those patients with other manifest or suspected primary causes and secondary causes of dementia. Patients with a history of transient ischemic attacks or stroke episodes or occurrence of pronounced vascular factors or with computed tomographic scan findings of infarcts were also excluded. Thus, the diagnosis of "probable AD" was made in accordance with the NINCDS-ADRDA criteria.32 All of the AD patients belonged to the subgroup AD type I.33 Four cerebrospinal fluid samples were obtained from children admitted for investigation of suspected neurological disease. All four children were determined to have normal blood--brain barrier function, and no pathological values were found for routine protein or monoamine metabolite analysis of each child. Patient A was a 4-year-old female with undiagnosed muscular pain. She showed no oligoclonal bands or change in IgG index for her cerebrospinal fluid. No pathological values for albumin, IgG, or IgM were found in her cerebrospinal fluid. She had normal values for @-endorphin and neuropeptide Y. She may suffer from a brain stem lesion. (29) Laurell, C.-B. Scand. J . Clin. Lab. Itwesf. 1972, 29, Suppl. 124, 21-37. (30) Link, H.; Tibbling, G. Scand. J . Clin. Lab. Itwesf. 1977, 37,391-396. (31) Folstein, M. F.; Folstein, S. E.; McHugh, P. R. J. Psychiatry Res. 1975, 12, 189-198. (32) McKhann, G.; Drachman, D.; Folstein, M.; Katzman, R.; Price, D.; Stadlan, E. M. Neurology 1984, 34, 939-944. (33) Blennow, K.; Wallin, A. J. Geriafr.Psychiatry Neurd. 1992, 5, 106-113.

E D

I i

10

12

I

I

14

16

Time (min)

10

15

20

25

30

Time (minl Flguro 1. Electropherogram of a CBQCAderivatized human cerebrospinalfluid sample from a pool of healthy adults (pool CCSF). Peaks that have been identified are R, arginine: Q, glutamine; T, threonine; V, valine; GABA: S, serine: A, alanine; 0, glycine; E, glutamic acid: D, aspartic acid. The internal standard (IS)is somatostatin (3-6).

Patient B was a 3-year-old female diagnosed with progressive epilepsy. She exhibited no intrathecal production of IgG, and no oligoclonal bands were found in her cerebrospinal fluid. She might, however, suffer from some neuroimmunological/ neurodegenerative disturbance. Patient C was a 7-year-old female with suspected infantile autism and mental retardation. Patient D was a 9-year-old male who suffered from a headache and a swollen forehead. He exhibited normal values for IgG, 8-endorphin, neuropeptide Y, and somatostatin. He had high values for eosinophile cells and IgE in his blood and was diagnosed as having a parasitic infection. Based on these examinations, patients A and D were determined to be “brain healthy.”

RESULTS AND DISCUSSION Separation and Detection of Human Cerebrospinal Fluid. An electropherogram of cerebrospinal fluid from a pool of 80 healthy humans is shown in Figure 1. The sample has been derivatized with CBQCA for LIF detection, and the separation is performed in uncoated fused silica capillaries using a separation buffer consisting of 20% DMSO and 80% 30 mM SDS, 50 mM borate, pH 9.0. Sodium dodecyl sulfate is used in the buffer to enhance the electrophoretic separation by formation of a micellar pseudostationary phase,34 and the organic modifier, DMSO, is added to increase peak capacity by reduction of electroosmotic Ten amino acid peaks in the electropherogram have been identified by spiking small aliquots of the cerebrospinal fluid sample with the individual amino acids. Figure 2 shows an expanded view of a small region of the electropherogram shown in Figure 1. A large number of smaller peaks and shoulders are apparent in this view of the separation, and approximately 50 peaks are present in the electropherogram in the time window from 10 to 35 min. Buffer modification with an organic additive has been used to improve peak capacity in the separation, leading to the (34) Terabe, S. In Capillary Electrophoresis Technology; Guzman, N. A,, Ed.; Marcel Dekker, Inc.: New York, 1993; Chapter 2. (35) Schwer, C.; Kenndler, E. Anal. Chem. 1991, 63, 1801-1807.

Figure2. Expandedview of the electropherogram in Figure 1 showing some of the smaller peaks from CBQCAderlvatlzed human cerebrospinal fluid (pool CCSF).

ability to separate a large number of analytes in a short time period; however, reduction of electroosmostic flow does result in increased analysis time. The final composition of the separation buffer represents a compromise between the amount of information obtained from the separation and the total analysis time. Initial separations using borate buffer at pH 9.0 resolved only a few species with an analysis time of 15 min. Addition of 30 mM SDS to the buffer has been found to greatly enhance the separation of CBQCA-derivatized cerebrospinal fluid, and, as expected, the elution order of the peaks in the separation changed dramatically. Modification of this buffer with DMSO has been found to increase the peak capacity for the separation without dramatically altering the separation selectivity. The final DMSO concentration used (20%) allows a large number of peaks to be resolved in a reasonably short period of time (35 min). The selectivity and analysis time of this separation can be optimized for specific applications by modification of the buffer pH, the micellar phase composition, the organic modifier concentration, and the potential field strength. An electropherogram of a standard solution of the 10amino acids identified in the cerebrospinal separation and the internal standard, somatostatin (3-6), is shown in Figure 3. All 10 amino acids, with the exception of GABA and serine, are separated with near base line resolution in less than 35 min. Separation efficiencies obtained range from 120 000 theoretical plates for glutamic acid and aspartic acid up to 190 000 theoretical plates for arginine (Table 1). Concentrations of amino acids in this electropherogram range from 0.2 pM for GABA to 140 pM for threonine. Detection sensitivity varies significantly with amino acid structure as described previously.’3 Detection limits for these 10 amino acids range from 0.29 nM for GABA to 100 nM for threonine as shown in Table 1. The final derivatization time of 2.0 h has been chosen after examination of reaction times for cerebrospinal fluid samples between 5 min and 7 h. The reaction of amino acids with CBQCA should be nearly complete in 1 h at room temperature,I3 but cerebrospinal fluid contains a large number of different amine-containing compounds at a wide range of concentrations. Structurally different primary amines should react with CBQCA at different rates, and reaction rates for a specific compound will change as the concentration of that Analytical Chemistry, Vol. 66, No. 20, October 15, 1994

3515

Table 2. Amlno Acld Values' In Pooled Samples of Human Cerebrospinal Fluid E D

I 1

amino acid

CCSF

Pool AMCCSF

ADCSF

Art2 Gln Thr Val GABA Ser Ala GIY Glu ASP

27.0 f 5.7 670 f 140 151 f 46 37 f 12 0.1045 f 0.0077 36.7 f 5.3 59.3 f 8.2 20.5 f 4.6 8.3 i: 1.5 0.127 f 0.023

14.6 f 1.5 391 f 43 45.7 f 8.7 10,18 f 0.74 0.0919 f 0.0025 21.52 f 0.41 45.1 f 4.2 9.53 f 0.94 14.7 f 1.7 0.172 f 0.018

18.1 f 2.2 375 f 50 88 f 34 11.1 f 4.6 0.0396 i 0.0053 23.1 i 2.1 43.6 f 4.0 19.0 f 1.9 19.9 & 2.3 0.162 f 0.043

In pmol/L i SEM, N = 3. 10

15

20

25

30

Time (min)

Flgure 3. Electropherogram of a standard solution containing the 10 CBQCA-labeled amino acids identified in cerebrospinal fluid samples. Peak identities are as in Figure 1. Analyte concentrations (In pM) are arginine, 6.0; glutamine, 65; threonine, 1 4 0 valine, 30; GABA, 0.2; serine, 20; alanine, 2 0 glycine, 2.5; glutamic acid, 65; aspartic acid, 14; internal standard, 11.

Table 1. Separatlon Efflclencles, Detectlon LlmHs, and Linear Correlation Coefflclents for CBQCA-Labeled Amlno Acld Standard Solutlons se aration detection linear amino efgciency, limit (nM), 2 X correlation acid theoretical plates peak to peak noise coefficient, r

Arg Gln Thr Val GABA Ser Ala GlY Glu ASP

190 000 150 000 150 000 180 000 160 000 140 OM) 160 000 180 000 120 000 120 000

68 10 100 4.3 0.29 19 16 5.2 58 1.1

0.999 0.999 0.983 0.999 0.999 0.999 0.971 0.999 0.998 0.999

compound changes. Individual peaks in the cerebrospinal fluid electropherogram have been observed to slowly change throughout derivatization times from 5 min to 7 h. The final time of 2.0 h has been chosen because at this time, the response for the 10 peaks that have been identified is sufficiently stable for quantitative analysis. As additional peaks in the cerebrospinal fluid electropherograms are identified and targeted for more detailed analysis, the reaction times for these species will have to be assessed. Calibration curves have been obtained for standard samples of each of the 10 identified amino acids in order to quantitate these species in human cerebrospinal fluid samples. The curves have been constructed by derivatizing amino acid standards at each concentration and normalizing the corresponding peak areas to the total area of the internal standard peaks. As shown in Table 1, correlation coefficients of 0.971 or greater are obtained over concentration ranges spanning the range of values found for each amino acid determined in human cerebrospinal fluid. The internal standard, somatostatin (36 ) , contains two labeling sites due to the presence of lysine and gives two peaks in the electropherogram. Although this internal standard is not ideal, the total area of the two peaks has been found experimentally to give a consistent response for repeated derivatization and injection, and the two peaks 3516

Analytical Chemistry, Vol. 66, No. 20, October 15, 7994

are well separated from any major peaks in the cerebrospinal fluid electropherograms. The precision of the 1.0 s pressure injection has also been assessed by making repeated injections of 5 X le7M fluorescein, and a relative standard deviation of 6.3% has been obtained for five injections. Five 5.0 s injections of the same sample give a relative standard deviation of 1.9%, indicating that the precision of the analysis can be improved by increasing the injection time from 1.0 s. Amino Acid Levels in Different Patient Groups. Amino acid levels have been determined for pools of cerebrospinal fluid from three different patient groups, and the results are shown in Table 2. Pool CCSF consists of cerebrospinal fluid samples from 80 healthy individuals, all more than 15 years of age. Pool AMCCSF consists of cerebrospinal fluid from five healthy individuals showing no signs of dementia and aged 66-70. Pool ADCSF consists of cerebrospinal fluid from five individuals aged 60-68 and diagnosed with Alzheimer's disease. The amino acid levels reported for pool CCSF are similar to values obtained previously for adult human cerebrospinal fluid controls by ion-exchange chromatography and HPLC.22,23,36,37 The values of threonine and alanine for pool CCSF are considerably higher than those previously reported, and valine and glycine levels are higher than some previously reported values. Threonine and valine are poorly resolved in the cerebrospinal fluid separations, which may lead to inaccurate quantitation, but glycine and alanine appear to be well separated from other large peaks. A wide range of control values has been reported previously for amino acids in cerebrospinal Recoveries of amino acids from human cerebrospinal fluid samples have been examined by spiking cerebrospinal fluid samples (from CCSF) with the 10 amino acids that have been identified and quantitated in this study (data not shown). Standard solutions of amino acids have been allowed to equilibrate for 15min with cerebrospinal fluid prior to addition of CBQCA and KCN, and the amino acid peaks in the electropherograms from these samples have been compared to control samples of cerebrospinal fluid and to samples of amino acid standards. Glutamine, alanine and glycine recoveries are within 5% of 100%recovery. The recovery for glutamic acid is 107%, and recoveries for threonine, valine, (36) McGale, E. H. F.; Pye, I. F.;Stonier, C.; Hutchinson, E. C.; A h , G.M. J . Neurochem. 1977, 29, 291-297. (37) Spink, D. C.; Swann, J. W.; Snead, 0. C.; Waniewski, R. A.; Martin, D. L. Anal. Biochem. 1986, 158, 79-86.

GABA, serine, and aspartic acid range from 74% to 87%. It is difficult to determine the significance of these results. The reduced recoveries of amino acids may be due to adsorption to other molecules in the cerebrospinal fluid, reaction with other molecules in the samples, or incomplete reaction with CBQCA. Levels of a number of amino acids measured in cerebrospinal fluid including glutamine, GABA, glutamic acid, and aspartic acid have been shown to change with time as a result of storage at room temperature and deproteinization with sulfosalicyclic a ~ i d . 3 ~It- ~is ~clear that a number of different factors can alter recoveries of amino acids in the analysis of cerebrospinal fluid, and uniform sample storage, sample preparation, and sample derivatization are critical in order for comparisons to be made using values determined for amino acids in different samples. y-Amino-n-butyric acid determination in cerebrospinal fluid has been particularly challenging in the past due to both the instability of GABA levels in cerebrospinal fluid over time and the low level at which this neurotransmitter is found. Degradation of homocarnosine has been shown to cause increases in free GABA levels when untreated cerebrospinal fluid is stored at room t e m ~ e r a t u r e and , ~ ~a very wide range of GABA levels has been reported in human cerebrospinal fluid.39 The GABA levels reported here fall in the lower end of the range of previously reported values. The ability to determine an amino acid involved in neurotransmission at nanomolar levels in cerebrospinal fluid emphasizes the sensitivity advantage gained by using LIF detection with capillary electrophoresis. Several differences in amino acid levels are observed between the three different pooled human cerebrospinal fluid samples. The most interesting differences are between the Alzheimer’s patient pool and its age-matched control pool, ADCSF and AMCCSF, respectively. Figure 4 shows electropherograms from these two pools. Substantially lower levels of GABA are found in the Alzheimer’s pool (P< 0.001), and glycine levels are increased in the Alzheimer’s pool compared to AMCCSF (P< 0.02); however, the glycinelevel in ADCSF is similar to that for the younger pool CCSF. Basun et al. have reported decreased levels of valine, glycine, and leucine in cerebrospinal fluid from Alzheimer’s patients relative to controls, but they did not examine GABA levels.18 Our results do not agree with those of Basun et al., and it is unclear why this is so. Comparison of the pool of healthy elderly patients (AMCCSF) with the control pool (CCSF) shows significant differences in levels of serine (decreased in AMCCSF, P < 0.05) and glutamic acid (increased in AMCCSF, P < 0.05). These differences in amino acid levels may be age related, but these age correlations have not been found in other studies.1722.36 Table 3 shows amino acid levels found in the cerebrospinal fluid of four children. The amino acid levels for these children agreevery well with recently reported values for cerebrospinal fluid from children with the exception of GABA, which is found at much lower levels in our study.20 Other reports of GABA levels for children agree with our results.40 Levels of (38) Grove, J.; Schechter, P. J.; Tell, G.; Rumbach, L; Marescaux, C.; Warter, J.-M.; Koch-Weser, J. J . Neurochem. 1982, 39, 1061-1065. (39) Schechter, P. J.; Sjoerdsma, A. Neurochem. Res. 1990, 15, 419423. (40) Jaeken, J.; Casaer, P.; Haegele, K. D.; Schechter, P. J. J . Inherited Metob. Dis. 1990, 13, 793-801.

R

I

E D

IS IS

I1

I i

I

I

1

I

15

20

25

30

Time [minl

Figure 4. Comparlson of electropherogramsobtalned from separations of CBQCAderivatlzed cerebrospinal fluM from a pool of Alzheimer’s patients (pool ADCSF, top) and from an age matched control pool (pool AMCCSF, bottom).

each amino acid appear to be similar for each child with the exception of patient D. This patient is a 9-year-old male suffering from a headache and a swollen forehead and has been diagnosed as having a parasitic infection. Compared to the other three children, patient D shows significantly reduced levels of glutamine (P< 0.05 for D vs A, B, C, and EA-C) and threonine (P< 0.02 vs A, B, C, and EA-C). Patient C seems to have an amino acid profile similar to the apparent norms for the other three children. Patients A and B exhibit relatively unexceptional amino acid levels in their cerebrospinal fluid; however, patient A exhibits elevated levels of serine (P