Optimized methods to measure acetoacetate, 3-hydroxybutyrate, glycerol, alanine, pyruvate, lactate and glucose in human blood using a centrifugal analyser with a fluorimetric attachment

Optimized methods are described for the analysis of glucose, lactate, pyruvate, alanine, glycerol, D-3-hydroxybutyrate and acetoacetate in perchloric acid extracts of human blood using the Cobas Bio centrifugal analyser. Glucose and lactate are measured using the photometric mode and other metabolites using the fluorimetric mode. The intra-assay coefficients of variation ranged from 0.7 to 4.1%, except with very low levels of pyruvate and acetoacetate where the coefficients of variation were 7.1 and 12% respectively. All seven metabolites can be measured in a perchloric acid extract of 20 μl of blood. The methods have been optimized with regard to variation in the perchloric acid content of the samples. These variations arise from the method of sample preparation used to minimize changes occurring in metabolite concentration after venepuncture.

Introduction effect of pH and perchlorate on the activity of each of the enzymes using spectrophotometric methods. In the light of these findings existing fluorimetric methods were modified and their performance evaluated. The use of spectrophotometric rather than fluorimetric methods for lactate and glucose was also investigated. From these studies optimized automated methods have been provided whereby unneutralized undiluted PCA extracts of blood can be applied directly to the analyser.
Initially, the effect of pH and perchlorate on the spectrophotometric assay each of the enzymes with each substrate was examined. This established that any effect on the enzyme activity was not simply due to changes in fluorescence. The LKB Reaction Rate Analyser was preferred for these experiments due to the ease in which large changes in sample and reagent volumes can be made.
Secondly, the reaction conditions were altered for the fluorimetric metabolite analyses on the Cobas Bio, taking account of the effects of pH and perchlorate on the spectrophotometric methods.
Economical and specific measurement of the blood concentrations of intermediary metabolites such as acetoacetate, 3-hydroxybutyrate, glycerol, alanine and pyruvate, as well as the more frequent measurement of lactate and glucose, is of interest to the clinical chemist. As some of these metabolites are present in low concentrations in human blood their accurate measurement in small samples (50 1) requires the sensitivity of fluorimetric methods, the required specificity being achieved by using enzymatic reactions. Using both these techniques with a centrifugal fast analyser, economic handling of large numbers of blood samples can be obtained. Methods for this analyser have been described which were adapted from continuous flow methods [2]. For all methods, the blood samples are deproteinized by treatment with perchloric acid (PCA). Perchlorate, however, interferes with the activity of the enzymes used in the methods and as the perchlorate concentration may vary in the samples, this in turn has an effect on the observed concentration of the metabolite [3]. This paper reports on an attempt to overcome the perchlorate interference and optimize the centrifugal fast analyser methods. To do this, it was necessary to examine the Thirdly, it was demonstrated that photometric methods were preferable to fluorimetric methods with metabolite concentrations in excess of 500 M. The latter could have been overcome by using dilutions of the PCA extracts but the aim was to establish a protocol whereby all seven metabolites could be analysed automatically on a single untreated PCA extract. (Lewes, East Sussex, UK); with glycerol (AR), glucose (50 mmol/1, 5 g/1 benzoic acid), perchloric acid-600 g/kg (AR) and all other chemicals from BDH (Poole, UK).

Sample preparation
About 1"5-2 ml of whole blood is added to a preweighed plastic tube containing 5 ml of 0"77 mol/1 perchloric acid previously cooled to 0 C. After mixing, the tubes were reweighed, centrifuged at 2500 g for 5 min and the protein-free supernatant removed. The dilution of the blood is determined by the weight changes measured. Alternatively, 0"5 ml whole blood may be added volumetrically to 2 ml 0"77 mol/1 perchloric acid and the weighing omitted. For the metabolites, other than acetoacetate, analysis of the samples is carried out on the same day or the samples are stored at -20C. Acetoacetate must be measured within 3 h of collection of blood or the samples stored at -70 C.
Enzyme assay using the LKB Reaction Rate Analyser To examine the possibility that the apparent difference in metabolite concentration in the presence of variable concentrations of perchlorate is due to interference in fluorescence, rather than by a decreased enzyme activity, measurements were made photometrically at 340 nm using a LKB 8600 Reaction Rate Analyser at 37C (Pharmacia LKB Biotechnology Division, Central Milton Keynes, Buckinghamshire, UK). To determine interference by perchlorate, various concentrations were added to cuvettes containing coenzyme in buffered substrate. To initiate the reaction the enzyme or enzymes were injected, usually in a volume of 100 tl. In all cases, the total volume in the cuvettes was ml and the experimental conditions the concentrations referred to are the final concentrations in the cuvettes.
Both these enzymes were diluted from stock in 0"04 mol/1 triethanolamine buffer pH 7"4 and injected together.
The results for each metabolite are shown in table where it can be seen that the glycerol, acetoacetate and glucose assays are most susceptible to perchlorate.

pH Optima
When the blood is added to the perchloric acid the acid concentration, as well as the perchlorate concentration of the extract, varies from sample to sample. The extent to Perchloric acid effects in the methods using the centrifugal analyser The extent to which variations in the perchlorate acid concentrations of the sample combined to alter the apparent concentration of the metabolites were then measured by conventional centrifugal analyser methods. A series of standard solutions of the metabolites were prepared, one set in 0.3 M PCA, one set in 0.4 M PCA, one set in 0"3 M sodium perchlorate, another in 0"3 M hydrochloric acid and a final set in water. The apparent concentrations of the metabolites in these solutions were measured by the methods previously described [1].  table 3 are therefore those which substantially reduce the effect of variations in PCA concentration and enable alanine concentrations to be measured relatively economically.
As shown in figure 8, fluorescence measurements of NADH and NADPH, at the levels present in the methods used for measuring lactate and glucose, gave non-linear results with the Cobas Bio. The computer within the instrument has a curve-fitting program, but as this nonlinearity could contribute to the overall imprecision in these methods, it was decided to use spectrophotometry rather than fluorimetry for the lactate and glucose analyses.
For optimized direct.automated analysis of the metabolites in PCA extracts of blood, the methods are as follows.  Figure 7. pH-activity curve for hexokinase/glucose-6-phosphate dehydrogenase using glucose as substrate.
To set the PM voltage use water as sample and a coenzyme reagent containing: 650 tl 0"75 mg/ml NADH in 0"3 mol/1 sodium phosphate buffer pH 7"4 added to 20 ml of the same buffer. The results are expressed as a percentage increase or decrease in the apparent concentration of the metabolite, as compared with that found in aqueous solution of the same concentration.
The working reagent is prepared by adding 0"7 ml of this solution to 8"8 ml of the buffer. This dilution is freshly prepared for each batch of analyses and is warmed to 25 C before use.
To set the PM voltage use a standard of 140 tmol/1. All concentrations are in tmol/1 except glucose which is in mmol/l. * The setting 340'2 is used in fluorescence modes to denote excitation wavelength 340 nm with emission filter position 2 (470 nm). The lactate and glucose methods are photometric hence no filter position is required. ** PM voltages are for guidance only and may vary depending on the state of the instrument.
Both coenzyme reagent and start reagent are prepared immediately before use.
To set the PM voltage, use water as sample and a NADH solution 20% more concentrated than used in the assay proper, i.e. 38 tl g/1 NADH per 10 ml buffer.
To set the PM voltage use a standard of 140 tmol/1.
For the assays for alanine, glycerol and 3hydroxybutyrate, the relatively high blank fluorescence makes it advisable to set the PM voltage using standard solutions of concentrations up to 20% higher than the respective top standards.

Standard solutions
Stock standards: 100 mmol/1 lithium lactate, 10 mmol/1 sodium pyruvate, 10 mmol/1D,L-3-hydroxybutyrate, 100 mmol/1 glycerol, 10 mmol/1 acetoacetate, 250 mmol/1 and 50 mmol/1 glucose alanine are prepared in deionizeddistilled water and diluted to the appropriate working level with 0"46 mol/1 perchloric acid. All standards are stored in a refrigerator at 4 C for seven days. Working standards containing all seven metabolites at the concentrations shown in table 3 are prepared on a daily basis.
Cobas Bio parameters for optimized methods It was found preferable to use print-out mode for all analyses apart from alanine. With print-out mode 1, the changes in fluorescence are related to the fluorescence of the reaction mixture prior to addition of the start reagent; with print-out mode 2 the changes in fluorescence are related to the fluorescence of the reaction mixture 0"5 s after addition of the start reagent. Ifvery rapid changes in fluorescence occur, such as with samples of high pyruvate concentration, the change in fluorescence in the first 0"5 s Although hydrazine has been used in the optimized method it may be omitted on safety grounds without a significant loss of performance.
The results are expressed as a percentage increase or decrease in the apparent concentration of the metabolite as compared with that found in aqueous solution of the same concentration. is significant and with print-out mode 2 an error is introduced. This and the other changes in parameters are included in table 3 where the parameters used for the optimized methods are listed.

Evaluation of optimized methods
The intra-assay imprecision is shown in table 4. For alanine, the parameters of the method are the same as those described previously [1 ], consequently the data for precision were not reassessed.
The effect of perchlorate, PCA and hydrochloric acid in the sample were examined using the optimized methods and the parameters listed in table 3. The concentrations used were within the range likely to be found in extracts using the standard method of sample collection where a variable quantity of blood is added to 5 ml of 0"77 mol/1 perchloric acid. The results are shown in table 5.

Discussion
To determine the in vivo concentrations of metabolites accurately it is necessary to make measurements immediately after venepuncture. This can be done with the biosensor systems now available to measure glucose [4] and lactate [5] in whole blood. More usually, blood samples are sent to a laboratory, which is often sited some distance away, so there can be a delay between venepuncture and analysis, incurring changes in metabolite concentrations. Even when fluoride is added to blood, stabilization is neither instantaneous nor complete, although after the first hour the level can be kept relatively stable at 4C for about 24 h, followed by a further slow decrease [6]. Complete stability of glucose and other metabolites is obtained if the blood proteins are precipitated and the enzymes activated with PCA.
Storage at -20 C of metabolites in PCA ensures a long period of stability except for acetoacetate where the lower temperature of-70 C may be required [2].
The main disadvantages of such preparation of samples are that perchlorate and acidity interfere with the activity of the enzymes used in the subsequent methods. As perchlorate affects both photometric and fluorimetric measurement modes, interference in fluorescence characteristics is not the explanation. One way to overcome the problem of PCA can be to neutralize the samples by titration with potassium carbonate or potassium hydroxide, which precipitates the perchlorate. Precipitation may not be complete and the neutralization procedure is in any case time-consuming and increases the imprecision of the technique. In addition it would require a significant amount of sample preparation prior to automated analysis, whereas the present aim was to reduce sample handling to a minimum.
The differences in the perchlorate ion and acid concentration of the perchloric acid extracts can be minimized by pipetting a measured volume of blood into the PCA.
However, such a procedure is time-consuming and leads to a delay in stabilization. In addition, the use of pipettes requires some degree of expertise, which those people most likely to be taking blood may not have. Substituting end-point methods for initial-rate methods overcomes the inaccuracy due to the presence of different amounts of PCA, except for alanine. With alanine the large amount of enzyme and the long incubation time required makes end-point analysis unacceptably expensive and timeconsuming. Fortuitously, the effects of acidity counteract the effects of the perchlorate ion, so that in practice the initial rate technique is the method of choice for alanine.
In general, end-point methods are less precise than initial-rate methods. This is because the calculations in the former are based on only two measurement, whereas, in the latter, multiple measurements are made. Initial-rate methods also use less enzyme and are quicker, but these advantages can only be utilized when the PCA concentration is exactly the same in samples and standards. In addition, it is necessary to be very precise about the activity of the enzyme used, and this should be determined at frequent intervals. Wide variations can occur in the activities of different batches of nominally identical enzymes provided by the same supplier.
The concentrations of glucose and lactate in blood are such that photometric methods can be used to measure these two metabolites. This is preferred to the alternative of diluting the sample and using fluorescence.
When large numbers of samples are being handled by several persons, the method chosen to measure an analyte is a compromise of various analytical ideals and objectives. These optimized methods have been developed in response to a first priority of minimizing changes in blood metabolites occurring after venepuncture, and, secondly, to improve accuracy. The former requires that sample preparation can be made by persons who do not necessarily have laboratory training and that blood is added to PCA without delay. All the analyses can be carried out without the need of further dilution. For glucose, the imprecision is within the limit now regarded as necessary for contemporary methods [7]. The other methods give a within-batch precision which makes them very useful. Some are slightly more prolonged and costly than those previously described ], but they now provide more reliable automated assays.