Simultaneous multielement analysis of microliter quantities of serum

Veillon , Susan A. Lewis , Kristine Y. Patterson , Wayne R. Wolf , James M. Harnly , Jacques. Versieck , Lidia. .... James M. Harnly , Donita L. Garla...
1 downloads 0 Views 415KB Size
1851

Anal. Chem. 1984, 56, 1651-1654

Simultaneous Multielement Analysis of Microliter Quantities of Serum for Copper, Iron, and Zinc by Graphite Furnace Atomic Absorption Spectrometry 5.A. Lewis’ and T. C. O’Haver* University of Maryland, College Park, Maryland 20742

J. M. Harnly Nutrient Composition Laboratory, USDA,Beltsville, Maryland 20705

A method Is descrlbed for the slmuitaneous detennlnatlon of Cu, Feyand Zn In 25 ML of serum, uslng slrnultaneous, multlelement atomic absorptlon spectrometry (SIMAAC) and graphite furnace atomization. Peak height measurementsare more adversely affected than peak area measurements by both the organlc and Inorganic condltuents In serum. Peak area measurements are signlficantly more preclse and accurate than peak helght measurements. Determlnatlons are carrled out at a 1:21 dllutlon of serum, In the concentration range wlth the optimum preclslon and mlnlmum carryover contamlnatlon and memory effects for these elements, uslng SIMAAC. The serum specimens are dlluted with Triton X-100 (10 mL/L), 10 mM nitrlc acld, and 30 mM ammonlum nitrate. The use of peak area measurements and atomlzatlon from a L’vov platform permlts callbratlon wlth acldlfled aqueous standards. A commercially available, proteln-based reference materlal, Catlon-Cal (Amerlcan Dade) and a bovlne serum pool are used as controls. The method Is shown to have good correlatlon to comparatlve methods (single-element flame AAS after deprotelnlzatlon) for the analysls of 37 serum samples.

In a previous report ( I ) , we described a method for the simultaneous determination of Ca, Cu, Fe, K, Mg, Na, and Zn using a simultaneous multielement atomic absorption spectrometer with a continuum source (SIMAAC) (2) and an &acetylene flame. This required a sample pipetting volume of 0.5 mL. Recent work (3-5) has shown that the SIMAAC can be successfully used with graphite furnace atomization. Furnace atomization permits the use of much smaller sample volumes and, potentially, the determination of metals in serum at the mg/L and pg/L levels. Single element micro methods exist for the determination of the above elements in serum. However, even given the extended calibration capabilities of SIMAAC (6),it is not feasible to determine all of these elements simultaneously, on a single dilution in a graphite furnace, because of the high concentrations of Ca, K, Mg, and Na (Table I). The development of a method to simultaneously determine Cu, Fe, and Zn on a small sample (25 pL) should be of particular interest to the pediatric clinical laboratory, especially to monitor neonatology specimens. In this report, we describe the successful use of the SIMAAC system and graphite furnace atomization to simultaneously determine Cu, Fe, and Zn in 25 pL of blood serum, diluted 1:21 with Triton X-100 (10 mL/L), 10 mM nitric acid, and 30 mM ammonium nitrate. Current address: Vitamin and Mineral Nutrition Laboratory, USDA, Beltsville, MD 20705. 0003-2700/84/0356-1651$01.50/0

Table I. Average Concentration of Metals in Blood Serum metal

concn, mg/L

metal

concn, mg/L

Na K Ca Mg

3500 200

Fe

100

Zn

1 1 1

cu

30

Table 11. Furnace Conditions”

dry ash atomize clean cool down

ramp, s

hold, s

temp, “C

20 20 0 1

20 20

1

14

200 500 2700* 2700 20

10 4

L’vov platform and pyrolytically coated graphite tube. bInternal flow, 50 mL/min (argon).

EXPERIMENTAL SECTION Instrumentation. The SIMAAC system described in detail elsewhere (2) is a prototype instrument, not currently available commercially. It consists of a continuum source, an atomizer (either flame or furnace), a 20-channel Echelle polychromator which produces a two-dimensional spectral array, a quartz refractor plate for wavelength modulation, and a dedicated minicomputer. Wavelength modulation provides for double-beam performance with background correction at all wavelengths. A sample volume of 20 pL was used for each determination. This was delivered by using an AS1 autosampler (Perkin-Elmer, Norwalk, CT), from a 0 . 5 dacid-washed autoanalyzercup (Elkay Products, Shrewsbury,MA). Nitric acid (50 mL/L) was used in the rinse cycle to reduce carry-over contamination. A PerkinElmer HGA 500 graphite furnace, with a pyrolytically coated graphite tube (Perkin-Elmer) and a L’vov platform, made from the same type of tube (7), was employed. The furnace program is shown in Table 11. Argon was used as a sweep gas. The atomization time was 10 s; data for Fe and Cu determinations were taken for the entire time. Data for Zn determinations were only taken for the fiist 3.5 s, because of ita volatility and to reduce background problems. Reagents. A custom-made standard containing 100 mg/L of Fe, Cu, and Zn (Spex Industries, Metuchen, NJ) was diluted with nitric acid (50 mL/L) (Ultrex, J. T. Baker Chemical Co., Phillipsburg, NJ)to give standard concentrations of 10,40,70,100, and 200 wg/L. The sample diluent consisted of Triton X-100(10 mL/L) (Sigma Chemical Co., St. Louis, MO), 10 mM Ultrex brand nitric acid and 30 mM ammonium nitrate. Ammonium nitrate was prepared from Ultrex brand nitric acid and isothermally distilled ammonium hydroxide (8). All reagents were prepared in 18 MQ deionized water (Millipore Corp., Bedford, MA). Procedure. Samples were prepared by adding 25 pL of serum to 0.5 mL of the sample diluent. A commercially availablequality control material, Cation-Cal (American Dade, Miami, FL) and 0 1984 Amerlcan Chemical Society

1852

ANALYTICAL CHEMISTRY, VOL. 56, NO. 9, AUGUST 1984

n

AREA

100

100

101

86

II

HEIGHT

100

107

126

89

io

13

II

GI L Figure 2. Effect of inorganic serum constituents on peak shape and percent recovery, calculated using peak area and peak height, on a 100 pg/L Cu solution. SALT

Figure 1. Effect of organic constituents of 10-fold dilutlons of serum on peak shape and concentration of analytes: (-) ashed; (- - -)

unashed.

a bovine serum pool (91,diluted in a manner identical with the serum specimens, were used aa controls. Specimens, controls, and appropriateblanks were analyzed in triplicate against standards prepared in nitric acid (50 mL/L), at a rate of 25-30 per hour. Comparative Methods. The methods used for the single element determination of Cu, Fe, and Zn were air-acetylene AAS after deproteinization (10, 11),using a Perkin-Elmer 603.

RESULTS AND DISCUSSION Matrix Effects. Serum is a complex matrix, consisting of high levels of both organic and inorganic material. During the initial work on this method, it was found that both of these constituents affect the peak shape. The absorbance-time tracings for Cu and Fe atomized from a L’vov platform are shown in Figure 1. Two serum specimens, one unashed and one externally dry-ashed a t 480 O C overnight, were diluted 10-fold with water. Concentrations of Cu, Fe, and Zn in these samples were obtained by using flame AAS. When graphite furnace atomization is employed against acidified aqueous standards, both peak area and peak height calculations give answers in reasonable agreement with the flame values for the ashed serum. The concentrations obtained by using flame AAS are 0.68 k 0.04 mg/L and 1.80 f 0.10 mg/L for Cu and Fe, respectively, on ten determinations. The concentrations for the ashed serum for graphite furnace AAS on triplicate determinations are 0.64 mg/L for Cu and 1.71 mg/L for Fe using peak area calculations and 0.63 mg/L for Cu and 1.79 mg/L for Fe using peak height calculations. However, only peak area calculations give reasonable answers for the ashed serum, 0.66 mg/L and 1.78 mg/L for Cu and Fe, respectively. The results obtained by using peak height calculations are erroneously high, 1.36 mg/L for Cu and 4.00 mg/L for Fe. It was found necessary to dilute the unashed serum 50-fold to obtain the correct values for Fe and Cu for peak height determinations. Zn did not behave in a similar manner. The inorganic constituents also influence peak shape. Typical levels of total inorganic salta in serum are about 9 g/L. Increasing amounts of sodium, potassium,calcium, magnesium as chlorides, in a ratio appropriate to serum levels, were added to a 100 pg/L solution of Cu. The resulting absorbance-time tracings are shown in Figure 2. Area measurementsgave good recoveries, until at a high salt concentration (13 g/L), a suppression was obtained, probably due to chloride interferences. The peak height measurements gave a good recovery in the absence of salts, but showed an initial increase with increasing amounts of salts, until a low recovery was obtained a t the highest salt concentration; again the suppression is probably due to chloride interferences. The added salt concentrations had a similar effect on Fe and Zn. At salt concentrations typical of normal sera (9 g/L), there should be no effect if peak area measurements are used. Figures 1and 2 illustrate that peak area is much less affected by serum matrix than peak height. Peak height will be altered by any matrix component which changes the

0

3

Table 111. Percent Relative Standard Deviationsnof a 50 pg/L Solution in “OS (SO mL/L)

cu Fe

Zn an

area

height

1.9 3.7 6.9

4.0 7.5 10.4

= 10.

atomization rate and/or vaporization temperature. Area is independent of both of these parameters (theoretically) and should be dependent only on the total number of analyte atoms present (12,13). Both area and height are suppressed by the formation of monohalides which may diffuse from the furnaces before sufficiently high temperatures are reached to break down the molecules. The use of a platform is a partial remedy to reduce matrix interferences ( 1 4 ) , but matrix modification was found to be necessary. Nitric acid and magnesium nitrate, with a high ashing temperature, have been used to reduce matrix interferences from chloride (15);however, due to the volatility of Zn,a low ashing temperature must be used. Nitric acid and ammonium nitrate have both been reported to reduce chloride interferences on analytes in serum (16). Nitric acid (50 mL/L) cannot be used to dilute serum as it causes precipitation of proteins, so we decided to use 10 mM nitric acid and 30 mM ammonium nitrate (approximately a 10-fold excess of NO; over C1-). Even with the use of matrix modification, it was not possible to eliminate all of the background signal; however, it was reduced by about 60%. The background signal, due for the most part to the molecular absorbance of sodium chloride, is greater at shorter wavelengths but is still within the correction capabilities of the SIMAAC system (2). The use of this matrix modification improved the precisions of the determination of Cu, Fe, and Zn in serum, by 0.4%, 1.2%, and 1.4%, respectively. The addition of the wetting agent, Triton X-100, eliminated the problem of carbon buildup in the graphite tube due to the organic constituents in serum. Precision Studies. The concentrations of Cu, Fe, and Zn in normal serum are sufficiently similar (about 0.5-2.0 mg/L) to allow for a single dilution factor. From previous work (5) it has been shown that optimum precision is obtained in concentrations above about 50 pg/L, and carryover contamination and memory effects are not a problem at less than 200 pg/L for these elements. Thus, a 20-fold dilution of most sera will give concentrations of Cu, Fe, and Zn that will fall in this range. The results of within-run relative standard deviations (RSD’s) for a 50 rg/L acidified aqueous solution of Fe, Cu, and Zn are shown in Table 111. Peak area measurements are approximately a factor of 2 more precise than peak height. The means, within-run RSD’s and day-to-day RSD’s of Cation-Cal and the bovine serum pool, for peak area measurements are shown in Table IV. The RSD’s range from

ANALYTICAL CHEMISTRY, VOL. 56, NO. 9, AUGUST 1984

1653

Table IV. Results of Analysis of Quality Control Sera by SIMAAC (n = 10)

cation-cal % RSD

mean

within-run

day to day

bovine serum % RSD

manufacturer's value

1.89 1.2 1.9 1.92 cu, mg/L 2.08 2.8 4.7 2.18 Fe, mg/L 3.05 7.5 4.7 3.02 Zn, mg/L Mean (standard deviation) of results obtained by six independent methods.

mean

within-run

day to day

assay valuen

0.75 2.04 1.13

3.0 3.7 5.4

5.9 1.6 7.9

0.76 (0.08) 1.84 (0.25) 1.08 (0.13)

(I

Table V. Regression Statistics Comparing Results Obtained by SIMAAC with Those by Reference Method

n

area height

cu

area

Fe

height area height

Zn

Deming regression equation y y y y y y

31 37 36 36 37 37

= 0.00 + 1.012 = 0.11 + 1.092 = 0.01 + 0.98% = 0.07 + 1.492 = -0.01 + 1.032 = 0.85 + 1.412

HE IGHT

.5

std error

0.983 0.939 0.995 0.901 0.977 0.697

0.07 0.17 0.06 0.45 0.11 0.65

HE 1GHT

1

1 1.5 CU MC/L FLAME OAS

2

3

FE HWL FLAHE

nas

4

AREA

AREA

0

r

" 3.2 a E 2.4 -

2-

a

u)

-I

2 1.6w (L

.e-

l / * .s

.E

1.5 2 MWL FLAME ACIS 1

CU

I 1.6 2.4 3.2 FE RC/L FLfiHE kfiS

c with

Figure 4. Correlatlon plots of values obtained for Fe by SIMAAC with values obtained by comparative method.

1.2% for Cu to 7.9% for Zn. The precision of Zn determinations is poorer because of the lack of intensity of the EIMAC lamp in the ultraviolet region. Reference Materials. The mean values obtained by SIMAAC for Cation-Cal are all within 5% of the manufacturer's values, and the mean SIMAAC values for the bovine serum pool are within one standard deviation of the assay values. Comparison to Other Methods. Correlation plots of sera analyzed by the SIMAAC system and comparative methods are shown in Figures 3-5, for both peak area and peak height measurements. Regression analyses were carried out by using the Deming method (17, 18)which takes into account error in the ordinate or reference values as well as error in the experimental values, as opposed to conventional regression analyses which assume no error in the reference values. The correlation statistics are shown in Table V. The peak area

measurements show correlation coefficients (r) close to 1, intercepts close to zero, small standard errors (SE), and slopes varying from the ideal by less than 3%. Peak height measurements give poorer correlation coefficients, positive intercepts, larger standard errors, and slopes that differ from the ideal by 49%) 41%, and 9% for Fe, Zn, and Cu, respectively. It is therefore imperativewhen using AAS and graphite furnace atomizationvs. standards to only use values calculated from peak area measurements. Peak area measurements are both more precise and more accurate than peak height measurements. It was found necessary to use standards with low blanks due to the nonlinearity of the standard curves. Sample blanks were subtracted in the concentration mode rather than in the absorbance mode. Because of the ubiquitous problem of Zn contamination, it was found that for the analyses of real-life

Figure 3. Correlatbn plots of values obtained for Cu by SII values obtained by comparative method.

1884

ANALYTICAL CHEMISTRY, VOL. 56, NO. 9,AUGUST 1984 HE I GHT

I

e.



-

v

A

SIMAAC system, and to run ten specimens, five standards, and appropriate blanks in triplicate was about 3 h. Single element determinations, using the comparative methods described in this text, took approximately 3 h for each analyte. These conditions have been optimized for the SIMAAC system, but many are applicable to conventional single element graphite furnace AAS. Registry No. Cu, 7440-50-8; Fe, 7439-89-6; Zn, 7440-66-6.

LITERATURE CITED I

I

.8

ZH

1.6 2.4 3.2 MC/L FLAME AAS

AREA

.6

1.2

1.8

2.4

ZH HCiL FLClME AClS

Flgure 5. Correlation plots of values obtained for Zn by SIMAAC wlth values obtained by comparative method.

specimens, single determinations from triplicate dilutions rather than triplicate determinations from a single dilution should be made. Although graphite furnace AAS is slower than flame AAS, it is easily automated. The S W C system has an advantage over conventional systems in that the analytes are measured simultaneously from each atomization and that only a single set of standards is needed. The time needed to optimize the

(1) Lewls, S.; O’Haver, T. C.; Harnly, J. M. Anal. Chem, 1984, 56, 1066- 1070. (2) Harnly, J. M.; O’Haver, T. C.; Golden, 8.; Wolf, W. R. Anal. Chem. 1979, 57, 2007-2014. (3) Lewis, S. A.; O’Haver, T. C.; Harnly, J. M. Presented paper, FACSS, 1983. (4) Harnly, J. M.; Kane, J. S. Anal. Chem. 1984, 56, 48-52. (5) Harnly, J. M.; Mlller-Ihll, N. J.; O’Haver, T. C. Spectrochlm. Acta, Part E 1984, 398. 305. (6) Harnly, J. M.; O’Haver, T. C. Anal. Chem. W81, 53, 1291-1298. (7) Koirtyohann, S. R.; Kaiser, M. C. Anal, Chem. 1082, 54, 1515A1524A. (8) Vellkm, C.: Reamer, D. C. AMI. Chem. W E $ , 53,549-550. (9) Veillon, C.; Patterson, K. Y. Presented paper, FACSS, 1983. (IO) “Methods for Atomic Spectroscopy”; Perkin-Elmer: Norwalk, CT, 1979. (11) Henry, R. J. “Clinical Chemistry Principles and Techniques”, 2nd ed.; Harper and Row: New York, 1974; pp 698-705. (12) L’vov, 8. V. Spectrochim. Acta, PadB 1978, 338, 153-193. (13) Sturgeon, R. E.; Chakrabarti, C. C. Prog. Anal. At. Spectrosc. 1978, 7 , 9-189. (14) Slavln, W.; Manning, D. C.; Carnrlck, 0. R. At. Spectrosc. 1981, 2, 137-145. (15) Slavln, W.; Carnrlck, 0. R.; Manning, D. C. Anal. Chem. 1984, 56, 163- 168. (16) Graf-Harsanyl, E.; Langmyhr, F. J. Anal. Chim. Acta 1980, 7f6, 105-1 70. (17) Deming, W. E. “Statlstlcal Adjustment of Data”; Wlley: New York. 1943: D 184. r .- -. (18) Cornbleet, P. J.; Gochman, N. Clln. Chem. (Wlnston-Salem, N E . ) 1979, 25, 432-438.

.-.

RECEIVED for review February 13,1984. Accepted April 18, 1984. This paper was presented, in parts, at the 1983 FACSS meeting and at the 1984 Pittsburgh Conference. This paper is from a dissertation to be submitted to the Graduate School, University of Maryland, by S. A. Lewis in partial fulfillment of the requirements for a Degree of Doctor of Philosophy in Chemistry.