Journal of Immunological Methods, 134 (1990) 95-100
95
Elsevier JIM 05737
Comparison of an enzyme-linked immunosorbent assay (ELISA) with a radioimmunoassay (RIA) for the measurement of rat insulin Helen V. Webster, Adrian J. Bone, Katherine A. Webster and Terence J. Wilkin Endocrine Section, Medicine II, Southampton General Hospital, Southampton S09 4XY, U.K.
(Received 2 January 1990, revised received 4 June 1990, accepted 16 July 1990)
A recently developed competitive enzyme-linked immunosorbent assay (ELISA) was compared with a conventional competitive radioimmunoassay (RIA) for the measurement of rat insulin in culture medium. Fifty-six samples were analysed by both assays. There was a correlation coefficient of r = 0.783 between results obtained using the two assay systems. The binding curves of the two assays were differently shaped, so that the ELISA gave good reproducibility over the concentration range 5 - 5 0 / ~ U / m l insulin with interand intra-assay coefficients of variation < 14%, but poor reproducibility at higher concentrations. Conversely, the R I A showed excellent reproducibility at concentrations greater than 5 0 / ~ U / m l insulin, but poor sensitivity and high coefficients of variation below this level. The ELISA procedure offers practical advantages over the RIA, and performs well when measuring physiological concentrations of insulin. Key words: Insulin; Rat insulin; ELISA; Radioimmunoassay
Introduction
Insulin is responsible for the control of blood sugar, and its measurement is central to research in islet cell physiology and diabetes. The laboratory rat is widely used both for in vivo studies and as a source of isolated islets of Langerhans for in vitro investigation of pancreatic endocrine function. Yalow and Berson (1960) were the first to develop a chemical rather than a biological assay for insulin, and their competitive radioimmunoassay (RIA) has remained the basis for insulin measurement since. Radioimmunoassays, however, suffer inherent disadvantages, including notoriously poor repro-
Correspondence to: A.J. Bone, Medicine II, Level D, South Block, Southampton General Hospital, Southampton SO9 4XY, U.K.
ducibility at low insulin concentrations, and in 1988 Kekow and colleagues described an ELISA for rat insulin. The aim of the present study was to compare formally competitive ELISA and RIA systems in the measurement of rat insulin.
Materials and methods
Samples. The samples we used were supernatants from rat islet cell cultures, approximately 250/~1 in volume. Insulin standards. Gel purified rat insulin (Novo, Bagsvaerd, Denmark), supplied lyophilised with a biological potency of 19.2-23.6 I U / m g , was used to construct standard curves. A vial of 100/tg dissolved in 1 ml water was further diluted with buffer B (see RIA) or standard sample buffer (see ELISA) to an insulin concentration of 200
0022-1759/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
96 n g / m l (5,000/~U/ml) and aliquoted for storage at - 2 0 ° C . Standards were made up to concentrations of 150, 100, 75, 50, 25, 10 and 5 / ~ U / m l by diluting appropriate volumes of the 5000 /~U/ml solution with buffer B or standard sample buffer. Quality controls. Quality control samples consisted of the buffer and RPMI culture medium, in the same proportions to those in the culture supernatants, spiked with the following concentrations of rat insulin: 10, 20, 35, 45, 70 and 130/~U/ml. They were aliquoted and stored at - 2 0 ° C . Antisera. Guinea pig anti-porcine insulin serum (which cross-reacts 100% with rat insulin) was purchased from the Scottish Antibody Production Unit (SAPU, Glasgow) in 1 ml aliquots of freeze-dried guinea pig serum at a dilution of 1/100. The same insulin antiserum was used in both assays.
Radioimmunoassays Buffers. A stock buffer, pH 7.4, containing 31 g N a H 2 P O n / 2 H 2 0 , 1.2 g thiomersal and 25 g bovine serum albumin (Miles fraction V, BSA) per litre of distilled water was prepared. Assay buffers, A and B, were prepared from this stock as follows: buffer A was 200 ml of stock buffer diluted to 1 litre with distilled water. Buffer B was 100 ml of stock buffer containing 4.5 g NaC1 and 12.5 ml trasylol (2 × 104 k l U / m l , Bayer, U.K.) and then made up to 500 ml with distilled water. All three buffers were stored at 4°C. Method. The method was based upon that of Ashby and Speake (1975). Preincubation of sample and antiserum was carried out in LP3 tubes for 3 h at 4°C in a volume of 300 /~1 which comprised 100/~1 buffer B containing the sample at a 1 / 5 dilution, and 200 #1 buffer A containing the anti-insulin serum at a dilution of 1/15,000. 100 /~1 of buffer A containing about 12,000 cpm 125I labelled insulin (Amersham International, U.K.; specific activity > 300/~Ci//~g [IM38]) were added and the mixture (volume 400/~1, final sample dilution 1/20) left to incubate overnight at 4°C. A charcoal phase separant was made up from 1 g activated charcoal (Norit A, BDH), 10 ml filtered horse serum (Gibco, Paisley, U.K.) and 10 ml buffer A. The mixture was stirred for 20 min before use and 400 /~1 added to each tube. The
tubes were whirlimixed, incubated for 10 min on ice, and centrifuged for 30 min at 4°C. A 400 /~1 aliquot of the supernatant was carefully removed and counted in a gamma counter (LKB) serviced by Ria Calc software.
ELISA The method was similar to that described by Kekow et al. (1988), and embodied two key principles: (1) the insulin capture antibody was immobilised on the solid phase by means of a coating antibody; (2) the reaction was competitive rather than direct. Coating antibody. An affinity purified anti guinea pig IgG raised in rabbits, purchased from Sigma (Poole, U.K.). Buffers. The coating antibody buffer was 0.05 M carbonate/bicarbonate, p H 9.6, containing 1.59 g NaCO3, 2.93 g N a H C O 3 and 0.20 g NaN 3 in 11 distilled water. The insulin capture antibody was dissolved in phosphate buffer, p H 7.4, containing 5.77 g N a 2 H P O 4 / 2 H 2 0 , 1.19 g N a H 2 P O 4 / 2 H 2 0 , 1 g BSA and 0.24 g thiomersal in 11 distilled water. The incubation buffer for the standards, samples and quality controls contained the phosphate buffer modified by the addition of 0.6 g NaC1 and 5.9 g BSA per 100 ml. The washing buffer (0.15 M phosphate buffered saline, pH 7.2), contained 8.0 g NaC1, 0.2 g KC1, 1.15 g N a 2 H P O 4, 0.2 g K H 2 P O 4 and 0.5 ml Tween 20 per litre of distilled water. The substrate solution for the enzyme conjugate contained 17 mg o-phenylenediamine dihydrochloride (OPD) and 5 #1 30% H202 per 50 ml substrate buffer. The substrate buffer was made up from 4.20 g citric acid in 20 ml distilled water and disodium hydrogen phosphate buffer containing 8.9 g N a 2 H P O 4 / 2 H 2 0 in 500 ml distilled water, mixed in a ratio of I : 2, and adjusted to pH 5.0. The enzyme conjugate (peroxidase-labelled rat insulin) and substrate (OPD) were purchased from Sigma (Poole, U.K.). Method. All wells of a 96-well plastic microtitre plate (Nunc, Roskilde, Denmark) except two 'air blanks' were filled with 150 /~1 coating antibody in buffer at a dilution of 1/1000 and incubated for 2 h at 37°C. They were then emptied, tapped on to paper towels, and washed manually three times with washing buffer.
97
The wells were then filled with insulin capture antibody at a dilution of 1/15,000 and incubated for 2 h at 37°C. After washing three times, samples, diluted at 1 / 5 , standards and quality controls were added in duplicate, 100/~1 per well, and incubated overnight at 4°C. After further washing, 100/~1 insulin peroxidase conjugate were added at 1/10,000 and incubated for 4 h at 4°C. The plates were again washed and 100 #1 of O P D substrate were added to all wells, including air blanks. The colour reaction was allowed to proceed for 30 min in the dark, following which 100/~1 0.5 M H2SO 4 were added to halt the reaction. The plate was allowed to equilibrate for a further 90 min and the optical densities were read at 492 nm by an automatic ELISA reader (Kontron, St. Albans, U.K.) linked to Ria Calc software for data analysis.
TABLE I COMPARISON OF C O E F F I C I E N T S OF INTERASSAY VARIATION A N D RECOVERIES OF T H E SAME QUALITY C O N T R O L SAMPLES (n = 8) IN ELISA A N D RIA FOR RAT I N S U L I N Expected concen-
Mean +__SD recovery (/tU/ml)
% CV
tration (/LU/ml)
ELISA
RIA
10 20 35 45 70 130
10.2+ 1.3 17.3+ 2.3 40.2+ 5.0 39.7+ 4.1 209.0+ 76.3 311.4+111.7
16.4+ 9.0 16.9+ 2.7 29.0+ 6.8 31.3__12.4 78.8+ 7.3 131.9+ 6.0
ELISA
RIA
12.1 13.2 12.5 10.4 37.5 35.9
54.9 46.0 23.4 39.5 9.2 4.6
A ccuracy/prec•ion
Results Standard curves
Standard curves for R I A and ELISA, obtained from the same dilution series, are shown in Fig. 1. The dynamic range was greater for the ELISA, whose discrimination was greatest over the physiological range of insulin concentration (5-20 # U / m l ) in contrast to that of the R I A whose discrimination was least at physiological insulin concentrations.
This difference is reflected in the interassay and intraassay variations at various points on the binding curve shown in Tables I and II. The tables also give an indication of accuracy at different insulin concentrations, based on mean recoveries. The ELISA was both accurate and reproducible at low insulin concentrations, but considerably less so at higher concentrations. The RIA, on the other hand, was relatively inaccurate and imprecise at low insulin concentrations but impressively accurate and reproducible at high concentrations. RIA Standard Curve
ELISA Standard Curve C.P.M x 103
O.O.
9(
1000-
O
°
9007 800-
5
700-
3(
6005000
I
I
I
5
10
20
I
I
I
50 100 200
Insulin Concentration pU/ml
I
500
0
I
I
I
I
I
I
5
10
20
50
100
200
Insulin Concentration iJU/ml
Fig. 1. Standard curves for the ELISA and RIA for rat insulin based on the means of duplicate samples.
98 T A B L E II
T A B L E III
C O M P A R I S O N OF C O E F F I C I E N T S OF INTRA-ASSAY V A R I A T I O N A N D RECOVERIES OF T H E SAME Q U A L ITY C O N T R O L SAMPLES (n = 8) IN RIA A N D ELISA FOR RAT INSULIN
E F F E C T OF D I L U T I N G TEST SAMPLES ON ELISA PERFORMANCE
Expected
Mean _+SD recovery
% CV
concentration (/~U/ml)
(/~U/ml)
ELISA
10 20 35 45 70 130
EL1SA
RIA
11.6+ 1.6 20.5+ 2.8 46.8_+ 5.0 47.7_+ 4.6 172.6_+38.6 272.0_+48.3
11.1+5.5 13.2+7.7 17.3_+3.1 19.4_+1.4 69.8_+4.0 133.4_+6.9
13.7 13.9 10.6 9.7 22.4 17.9
Each result is the mean of duplicates.
RIA
49.3 58.3 17.8 7.4 5.7 5.2
Original concentration (~U/ml)
Dilution factor
Interpolated result (ttU/ml)
Adjusted for dilution factor (t~U/ml)
50 80 120 220 220 220 220
1 in 1 in 1 in 1 in 1 in 1 in 1 in
9.1 15.7 28.1 54.2 28.3 11.4 7.2
45.4 78.5 140.4 271.0 262.8 227.8 286.4
5 5 5 5 10 20 40
Sensitivity The sensitivity of the two assays, based on the mean plus two standard deviations of the signals given by 20 insulin-free samples, was < 2 # U / m l for the ELISA and approximately 10 /~U/ml for the RIA.
The shape of the ELISA standard curve meant that samples with high insulin concentration could be diluted to advantage for measurement in ELISA. The reverse would not readily apply to samples of low insulin concentration in the RIA. The results in ELISA of diluting quality control samples in 6% BSA phosphate buffer, shown in Table III, show close correspondence between the original value and the interpolated value corrected for its dilution factor.
Correlation between E L I S A and RIA The correlation between ELISA and RIA for 56 samples of widely differing insulin concentrations, measured at fixed dilution, is shown in Fig.
400 -
n=56
350 -
I
300 -
O
O
=_
.~
Q
O
250 O
:a_ 200
Q
Q
D
150 ~7
100
o~ •
50-
o
0
I
I
I
I
I
I
I
I
I
I
I
I
I
50
100
150
200
250
300
350
400
450
500
550
600
650
ELISA values pU/ml insulin Fig. 2. T h e correlation b e t w e e n the R I A and the E L I S A for rat insulin.
99
2. The coefficient, r, was 0.78 ( P < 0.001). There were, nevertheless, major differences in individual values ascribed to these samples, related almost certainly to the low precision of RIA at low insulin concentrations, and low precision of ELISA at high concentrations. The relatively limited dynamic range of the RIA is further evidenced in this plot.
Discussion We have found the ELISA for rat insulin first described by Kekow et al. (1988) is reproducible in a second laboratory and simple to carry out. The performance parameters (accuracy, reproducibility and sensitivity) at physiological concentrations of insulin are satisfactory and correspond well to those previously described. Samples containing higher concentrations of insulin can be diluted to fall on the optimum part of the standard curve without serious loss of precision. This manoeuvre is not appropriate for RIA, whose optimum reproducibility and accuracy fall at the higher concentration end of the standard curve. The relatively poor reproducibility of RIA at low concentrations of insulin is widely recognised. Any study such as this depends critically upon the quality of the assayist and the materials used. This is an experienced laboratory which has published widely on solid phase and liquid phase assays, and the insulin specific antiserum used is the only commercial preparation widely available in the U.K.. Furthermore, care was taken to use the same antibody (from the same batch) in both assay systems. The quality control data detailed here compares favourably with that reported elsewhere. There are practical advantages to ELISA. It is cheap and avoids the increasingly severe constraints placed upon laboratories using radiochemicals. ELISA can also deal with longer sample runs and the steps can all be automated. The components used in ELISA have a long shelf life and are not subject to decay in the same way as an isotope. Stored in glycerol at 4°C, for example, the peroxidase conjugate can last several months or even years (Montoya and Castell, 1987). Finally, the ELISA takes about half as long as the RIA to
complete a run of a larger number of samples. This is important since the major expense of any assay is in technician time. Two other ELISAs for the measurement of rat insulin have been published. Bank (1988) developed an indirect, competitive immunoassay where the unknown sample of insulin competes with insulin in the solid phase for anti-insulin antibodies, and a labelled anti-globulin antibody was used for detection. This labelled antibody had to be stored at - 7 0 ° C , and the incubation step for the anti-insulin serum and insulin sample with bound insulin was of 72 h duration. Quality control data were not provided and there was insufficient data for this report to be evaluated. The ELISA developed by Angel (1988) employed the principle of competitive saturation, but the anti-insulin antibody was bound directly to the microtitre wells by evaporation to dryness for 24-30 h, a lengthy procedure for which a typical laboratory will not have the necessary equipment. Although the dynamic range of the assay was greater than that of the ELISA used in this study, it was not as sensitive, and the reproducibility up to 50/~U/ml insulin was inferior. To date, the ELISA developed by Kekow and his colleagues (1988) would seem to be the ELISA of choice for assaying rat insulin. The technique correlates well with the conventionally used RIA, but appears to have greater reproducibility and sensitivity, and in practice is simpler, cheaper and safer to perform.
Acknowledgements Financial support is gratefully acknowledged from Lilly Research and the Wellcome Trust. Thanks to Wendy Couper for the preparation of the manuscript.
References Angel, I. (1988) The use of microtitre plates for the simple and sensitive determination of insulin by an ELISA method. Experientia 4, 877-879. Ashby, J.P. and Speake, R.N. (1975) Insulin and glucagon secretion from isolated islets of Langerhans. The effects of calcium ionophores. Biochem. J. 150, 89-96.
100 Bank, H.L. (1988) A quantitative enzyme-linked immunosorbent assay for rat insulin. J. lmmunoassay 9, 135-158. Kekow, J., Ulrichs, K., Muller-Ruchholtz, W. and Gross, W.L. (1988) Measurement of rat insulin: enzyme-linked immunosorbent assay with increased sensitivity, high accuracy and greater practicability than established radioimmunoassay. Diabetes 37, 321-326.
Montoya, A. and Castell, J.V. (1987) Long-term storage of peroxidase-labelled immunoglobulins for use in enzyme immunoassay. J. Immunol. Methods 99, 13-20. Yalow, R.S. and Berson, S.A. (1960) Immunoassay of endogenous plasma insulin in man. Clin. Invest. 39, 1157-1175.
95
Elsevier JIM 05737
Comparison of an enzyme-linked immunosorbent assay (ELISA) with a radioimmunoassay (RIA) for the measurement of rat insulin Helen V. Webster, Adrian J. Bone, Katherine A. Webster and Terence J. Wilkin Endocrine Section, Medicine II, Southampton General Hospital, Southampton S09 4XY, U.K.
(Received 2 January 1990, revised received 4 June 1990, accepted 16 July 1990)
A recently developed competitive enzyme-linked immunosorbent assay (ELISA) was compared with a conventional competitive radioimmunoassay (RIA) for the measurement of rat insulin in culture medium. Fifty-six samples were analysed by both assays. There was a correlation coefficient of r = 0.783 between results obtained using the two assay systems. The binding curves of the two assays were differently shaped, so that the ELISA gave good reproducibility over the concentration range 5 - 5 0 / ~ U / m l insulin with interand intra-assay coefficients of variation < 14%, but poor reproducibility at higher concentrations. Conversely, the R I A showed excellent reproducibility at concentrations greater than 5 0 / ~ U / m l insulin, but poor sensitivity and high coefficients of variation below this level. The ELISA procedure offers practical advantages over the RIA, and performs well when measuring physiological concentrations of insulin. Key words: Insulin; Rat insulin; ELISA; Radioimmunoassay
Introduction
Insulin is responsible for the control of blood sugar, and its measurement is central to research in islet cell physiology and diabetes. The laboratory rat is widely used both for in vivo studies and as a source of isolated islets of Langerhans for in vitro investigation of pancreatic endocrine function. Yalow and Berson (1960) were the first to develop a chemical rather than a biological assay for insulin, and their competitive radioimmunoassay (RIA) has remained the basis for insulin measurement since. Radioimmunoassays, however, suffer inherent disadvantages, including notoriously poor repro-
Correspondence to: A.J. Bone, Medicine II, Level D, South Block, Southampton General Hospital, Southampton SO9 4XY, U.K.
ducibility at low insulin concentrations, and in 1988 Kekow and colleagues described an ELISA for rat insulin. The aim of the present study was to compare formally competitive ELISA and RIA systems in the measurement of rat insulin.
Materials and methods
Samples. The samples we used were supernatants from rat islet cell cultures, approximately 250/~1 in volume. Insulin standards. Gel purified rat insulin (Novo, Bagsvaerd, Denmark), supplied lyophilised with a biological potency of 19.2-23.6 I U / m g , was used to construct standard curves. A vial of 100/tg dissolved in 1 ml water was further diluted with buffer B (see RIA) or standard sample buffer (see ELISA) to an insulin concentration of 200
0022-1759/90/$03.50 © 1990 Elsevier Science Publishers B.V. (Biomedical Division)
96 n g / m l (5,000/~U/ml) and aliquoted for storage at - 2 0 ° C . Standards were made up to concentrations of 150, 100, 75, 50, 25, 10 and 5 / ~ U / m l by diluting appropriate volumes of the 5000 /~U/ml solution with buffer B or standard sample buffer. Quality controls. Quality control samples consisted of the buffer and RPMI culture medium, in the same proportions to those in the culture supernatants, spiked with the following concentrations of rat insulin: 10, 20, 35, 45, 70 and 130/~U/ml. They were aliquoted and stored at - 2 0 ° C . Antisera. Guinea pig anti-porcine insulin serum (which cross-reacts 100% with rat insulin) was purchased from the Scottish Antibody Production Unit (SAPU, Glasgow) in 1 ml aliquots of freeze-dried guinea pig serum at a dilution of 1/100. The same insulin antiserum was used in both assays.
Radioimmunoassays Buffers. A stock buffer, pH 7.4, containing 31 g N a H 2 P O n / 2 H 2 0 , 1.2 g thiomersal and 25 g bovine serum albumin (Miles fraction V, BSA) per litre of distilled water was prepared. Assay buffers, A and B, were prepared from this stock as follows: buffer A was 200 ml of stock buffer diluted to 1 litre with distilled water. Buffer B was 100 ml of stock buffer containing 4.5 g NaC1 and 12.5 ml trasylol (2 × 104 k l U / m l , Bayer, U.K.) and then made up to 500 ml with distilled water. All three buffers were stored at 4°C. Method. The method was based upon that of Ashby and Speake (1975). Preincubation of sample and antiserum was carried out in LP3 tubes for 3 h at 4°C in a volume of 300 /~1 which comprised 100/~1 buffer B containing the sample at a 1 / 5 dilution, and 200 #1 buffer A containing the anti-insulin serum at a dilution of 1/15,000. 100 /~1 of buffer A containing about 12,000 cpm 125I labelled insulin (Amersham International, U.K.; specific activity > 300/~Ci//~g [IM38]) were added and the mixture (volume 400/~1, final sample dilution 1/20) left to incubate overnight at 4°C. A charcoal phase separant was made up from 1 g activated charcoal (Norit A, BDH), 10 ml filtered horse serum (Gibco, Paisley, U.K.) and 10 ml buffer A. The mixture was stirred for 20 min before use and 400 /~1 added to each tube. The
tubes were whirlimixed, incubated for 10 min on ice, and centrifuged for 30 min at 4°C. A 400 /~1 aliquot of the supernatant was carefully removed and counted in a gamma counter (LKB) serviced by Ria Calc software.
ELISA The method was similar to that described by Kekow et al. (1988), and embodied two key principles: (1) the insulin capture antibody was immobilised on the solid phase by means of a coating antibody; (2) the reaction was competitive rather than direct. Coating antibody. An affinity purified anti guinea pig IgG raised in rabbits, purchased from Sigma (Poole, U.K.). Buffers. The coating antibody buffer was 0.05 M carbonate/bicarbonate, p H 9.6, containing 1.59 g NaCO3, 2.93 g N a H C O 3 and 0.20 g NaN 3 in 11 distilled water. The insulin capture antibody was dissolved in phosphate buffer, p H 7.4, containing 5.77 g N a 2 H P O 4 / 2 H 2 0 , 1.19 g N a H 2 P O 4 / 2 H 2 0 , 1 g BSA and 0.24 g thiomersal in 11 distilled water. The incubation buffer for the standards, samples and quality controls contained the phosphate buffer modified by the addition of 0.6 g NaC1 and 5.9 g BSA per 100 ml. The washing buffer (0.15 M phosphate buffered saline, pH 7.2), contained 8.0 g NaC1, 0.2 g KC1, 1.15 g N a 2 H P O 4, 0.2 g K H 2 P O 4 and 0.5 ml Tween 20 per litre of distilled water. The substrate solution for the enzyme conjugate contained 17 mg o-phenylenediamine dihydrochloride (OPD) and 5 #1 30% H202 per 50 ml substrate buffer. The substrate buffer was made up from 4.20 g citric acid in 20 ml distilled water and disodium hydrogen phosphate buffer containing 8.9 g N a 2 H P O 4 / 2 H 2 0 in 500 ml distilled water, mixed in a ratio of I : 2, and adjusted to pH 5.0. The enzyme conjugate (peroxidase-labelled rat insulin) and substrate (OPD) were purchased from Sigma (Poole, U.K.). Method. All wells of a 96-well plastic microtitre plate (Nunc, Roskilde, Denmark) except two 'air blanks' were filled with 150 /~1 coating antibody in buffer at a dilution of 1/1000 and incubated for 2 h at 37°C. They were then emptied, tapped on to paper towels, and washed manually three times with washing buffer.
97
The wells were then filled with insulin capture antibody at a dilution of 1/15,000 and incubated for 2 h at 37°C. After washing three times, samples, diluted at 1 / 5 , standards and quality controls were added in duplicate, 100/~1 per well, and incubated overnight at 4°C. After further washing, 100/~1 insulin peroxidase conjugate were added at 1/10,000 and incubated for 4 h at 4°C. The plates were again washed and 100 #1 of O P D substrate were added to all wells, including air blanks. The colour reaction was allowed to proceed for 30 min in the dark, following which 100/~1 0.5 M H2SO 4 were added to halt the reaction. The plate was allowed to equilibrate for a further 90 min and the optical densities were read at 492 nm by an automatic ELISA reader (Kontron, St. Albans, U.K.) linked to Ria Calc software for data analysis.
TABLE I COMPARISON OF C O E F F I C I E N T S OF INTERASSAY VARIATION A N D RECOVERIES OF T H E SAME QUALITY C O N T R O L SAMPLES (n = 8) IN ELISA A N D RIA FOR RAT I N S U L I N Expected concen-
Mean +__SD recovery (/tU/ml)
% CV
tration (/LU/ml)
ELISA
RIA
10 20 35 45 70 130
10.2+ 1.3 17.3+ 2.3 40.2+ 5.0 39.7+ 4.1 209.0+ 76.3 311.4+111.7
16.4+ 9.0 16.9+ 2.7 29.0+ 6.8 31.3__12.4 78.8+ 7.3 131.9+ 6.0
ELISA
RIA
12.1 13.2 12.5 10.4 37.5 35.9
54.9 46.0 23.4 39.5 9.2 4.6
A ccuracy/prec•ion
Results Standard curves
Standard curves for R I A and ELISA, obtained from the same dilution series, are shown in Fig. 1. The dynamic range was greater for the ELISA, whose discrimination was greatest over the physiological range of insulin concentration (5-20 # U / m l ) in contrast to that of the R I A whose discrimination was least at physiological insulin concentrations.
This difference is reflected in the interassay and intraassay variations at various points on the binding curve shown in Tables I and II. The tables also give an indication of accuracy at different insulin concentrations, based on mean recoveries. The ELISA was both accurate and reproducible at low insulin concentrations, but considerably less so at higher concentrations. The RIA, on the other hand, was relatively inaccurate and imprecise at low insulin concentrations but impressively accurate and reproducible at high concentrations. RIA Standard Curve
ELISA Standard Curve C.P.M x 103
O.O.
9(
1000-
O
°
9007 800-
5
700-
3(
6005000
I
I
I
5
10
20
I
I
I
50 100 200
Insulin Concentration pU/ml
I
500
0
I
I
I
I
I
I
5
10
20
50
100
200
Insulin Concentration iJU/ml
Fig. 1. Standard curves for the ELISA and RIA for rat insulin based on the means of duplicate samples.
98 T A B L E II
T A B L E III
C O M P A R I S O N OF C O E F F I C I E N T S OF INTRA-ASSAY V A R I A T I O N A N D RECOVERIES OF T H E SAME Q U A L ITY C O N T R O L SAMPLES (n = 8) IN RIA A N D ELISA FOR RAT INSULIN
E F F E C T OF D I L U T I N G TEST SAMPLES ON ELISA PERFORMANCE
Expected
Mean _+SD recovery
% CV
concentration (/~U/ml)
(/~U/ml)
ELISA
10 20 35 45 70 130
EL1SA
RIA
11.6+ 1.6 20.5+ 2.8 46.8_+ 5.0 47.7_+ 4.6 172.6_+38.6 272.0_+48.3
11.1+5.5 13.2+7.7 17.3_+3.1 19.4_+1.4 69.8_+4.0 133.4_+6.9
13.7 13.9 10.6 9.7 22.4 17.9
Each result is the mean of duplicates.
RIA
49.3 58.3 17.8 7.4 5.7 5.2
Original concentration (~U/ml)
Dilution factor
Interpolated result (ttU/ml)
Adjusted for dilution factor (t~U/ml)
50 80 120 220 220 220 220
1 in 1 in 1 in 1 in 1 in 1 in 1 in
9.1 15.7 28.1 54.2 28.3 11.4 7.2
45.4 78.5 140.4 271.0 262.8 227.8 286.4
5 5 5 5 10 20 40
Sensitivity The sensitivity of the two assays, based on the mean plus two standard deviations of the signals given by 20 insulin-free samples, was < 2 # U / m l for the ELISA and approximately 10 /~U/ml for the RIA.
The shape of the ELISA standard curve meant that samples with high insulin concentration could be diluted to advantage for measurement in ELISA. The reverse would not readily apply to samples of low insulin concentration in the RIA. The results in ELISA of diluting quality control samples in 6% BSA phosphate buffer, shown in Table III, show close correspondence between the original value and the interpolated value corrected for its dilution factor.
Correlation between E L I S A and RIA The correlation between ELISA and RIA for 56 samples of widely differing insulin concentrations, measured at fixed dilution, is shown in Fig.
400 -
n=56
350 -
I
300 -
O
O
=_
.~
Q
O
250 O
:a_ 200
Q
Q
D
150 ~7
100
o~ •
50-
o
0
I
I
I
I
I
I
I
I
I
I
I
I
I
50
100
150
200
250
300
350
400
450
500
550
600
650
ELISA values pU/ml insulin Fig. 2. T h e correlation b e t w e e n the R I A and the E L I S A for rat insulin.
99
2. The coefficient, r, was 0.78 ( P < 0.001). There were, nevertheless, major differences in individual values ascribed to these samples, related almost certainly to the low precision of RIA at low insulin concentrations, and low precision of ELISA at high concentrations. The relatively limited dynamic range of the RIA is further evidenced in this plot.
Discussion We have found the ELISA for rat insulin first described by Kekow et al. (1988) is reproducible in a second laboratory and simple to carry out. The performance parameters (accuracy, reproducibility and sensitivity) at physiological concentrations of insulin are satisfactory and correspond well to those previously described. Samples containing higher concentrations of insulin can be diluted to fall on the optimum part of the standard curve without serious loss of precision. This manoeuvre is not appropriate for RIA, whose optimum reproducibility and accuracy fall at the higher concentration end of the standard curve. The relatively poor reproducibility of RIA at low concentrations of insulin is widely recognised. Any study such as this depends critically upon the quality of the assayist and the materials used. This is an experienced laboratory which has published widely on solid phase and liquid phase assays, and the insulin specific antiserum used is the only commercial preparation widely available in the U.K.. Furthermore, care was taken to use the same antibody (from the same batch) in both assay systems. The quality control data detailed here compares favourably with that reported elsewhere. There are practical advantages to ELISA. It is cheap and avoids the increasingly severe constraints placed upon laboratories using radiochemicals. ELISA can also deal with longer sample runs and the steps can all be automated. The components used in ELISA have a long shelf life and are not subject to decay in the same way as an isotope. Stored in glycerol at 4°C, for example, the peroxidase conjugate can last several months or even years (Montoya and Castell, 1987). Finally, the ELISA takes about half as long as the RIA to
complete a run of a larger number of samples. This is important since the major expense of any assay is in technician time. Two other ELISAs for the measurement of rat insulin have been published. Bank (1988) developed an indirect, competitive immunoassay where the unknown sample of insulin competes with insulin in the solid phase for anti-insulin antibodies, and a labelled anti-globulin antibody was used for detection. This labelled antibody had to be stored at - 7 0 ° C , and the incubation step for the anti-insulin serum and insulin sample with bound insulin was of 72 h duration. Quality control data were not provided and there was insufficient data for this report to be evaluated. The ELISA developed by Angel (1988) employed the principle of competitive saturation, but the anti-insulin antibody was bound directly to the microtitre wells by evaporation to dryness for 24-30 h, a lengthy procedure for which a typical laboratory will not have the necessary equipment. Although the dynamic range of the assay was greater than that of the ELISA used in this study, it was not as sensitive, and the reproducibility up to 50/~U/ml insulin was inferior. To date, the ELISA developed by Kekow and his colleagues (1988) would seem to be the ELISA of choice for assaying rat insulin. The technique correlates well with the conventionally used RIA, but appears to have greater reproducibility and sensitivity, and in practice is simpler, cheaper and safer to perform.
Acknowledgements Financial support is gratefully acknowledged from Lilly Research and the Wellcome Trust. Thanks to Wendy Couper for the preparation of the manuscript.
References Angel, I. (1988) The use of microtitre plates for the simple and sensitive determination of insulin by an ELISA method. Experientia 4, 877-879. Ashby, J.P. and Speake, R.N. (1975) Insulin and glucagon secretion from isolated islets of Langerhans. The effects of calcium ionophores. Biochem. J. 150, 89-96.
100 Bank, H.L. (1988) A quantitative enzyme-linked immunosorbent assay for rat insulin. J. lmmunoassay 9, 135-158. Kekow, J., Ulrichs, K., Muller-Ruchholtz, W. and Gross, W.L. (1988) Measurement of rat insulin: enzyme-linked immunosorbent assay with increased sensitivity, high accuracy and greater practicability than established radioimmunoassay. Diabetes 37, 321-326.
Montoya, A. and Castell, J.V. (1987) Long-term storage of peroxidase-labelled immunoglobulins for use in enzyme immunoassay. J. Immunol. Methods 99, 13-20. Yalow, R.S. and Berson, S.A. (1960) Immunoassay of endogenous plasma insulin in man. Clin. Invest. 39, 1157-1175.
