| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Briefs |
Departments of
1
Nuclear and
2
Laboratory Medicine, Free University of Brussels, Brugmann University Hospital, Place van Gehuchten 4, B-1020 Brussels, Belgium
3
Englebienne & Associates, Strijpstraat 21, B-9750 Zingem, Belgium
a author for correspondence: fax 32-9-384-7250, e-mail patrick.englebienne{at}skynet.be
Serum ferritin concentrations, with some exceptions (1)(2)(3), reflect iron stores (4)(5). Ferritin assays must have a broad dynamic range because the serum concentrations can be <10 µg/L (6) or >1 mg/L in some types of malignancies (7). Radio- and enzyme immunoassays have been used routinely (8), but rapid, automated latex agglutination immunoassays have been developed and validated (9)(10)(11). These methods have drawbacks (12), of which the disturbance of colloidal stability by nonspecific bridging processes particularly should be avoided (13).
In our quest for new reporter reagents (14)(15), our attention was drawn to colloidal gold as a potential substitute for latex in particle-enhanced agglutination immunoassays. Colloidal gold was used by Leuvering et al. (16) in sol particle immunoassays (SPIAs) for several serum or urine analytes. Unfortunately, the technique was prone to interference when undiluted serum samples were used (17). Recently, we showed (18)(19) that the change in visible absorbance at 600 nm (A600) observed when colloidal gold particles coated with an antibody interact with the antigen results not only from agglutination but also from subtle changes in the refractive index at the particle surface [surface plasmon resonance effect (SPR)]. Thus, the unidentified random interferences noted by Leuvering et al. (16) could result from interactions between nonspecific reacting sites on the antibody molecule (distinct from the binding site) and several serum components (distinct from the analyte). According to this model, these interactions were transduced as a photometric signal by the SPR effect of gold (19)(20), which adds to the signal produced by the agglutination. With this in mind, we optimized the buffer to be used in the SPIA with colloidal gold and carefully selected the antibodies to be used for their lack of SPR effects with human serum components other than the analyte. We present here the results obtained with an assay for human serum ferritin.
We synthesized colloidal gold particles with a mean diameter of 50 nm by reducing a boiling aqueous hydrogen tetrachloroaurate solution (500 mL, containing 0.4 g/L hydrogen tetrachloroaurate) with 20 mL of a 10 g/L solution of sodium citrate. The gold sol was then adjusted to pH 9 and coated separately during mixing for 10 min under magnetic stirring with two murine monoclonal antibodies (057-10030 and 090-12710; OEM Concepts) against ferritin. The antibodies, previously diluted in water (50 mg/L), were added rapidly to the colloidal gold sol to a final antibody concentration of 15 mg/L. These antibodies were selected because of their lack of reactivity with nonspecific human serum components by SPR (19). The minimum amount of IgG and the optimal pH for coating the particles by charge adsorption were determined by isothermal titration (19)(20). Such preliminary optimization allowed us to use an IgG concentration sufficient to saturate the particle surface; consequently, neither subsequent centrifugations nor washings were required. The coated particles were then buffered at pH 9.0 and stabilized by the sequential addition of sodium tetraborate (50 mmol/L), sodium chloride (150 mmol/L), bovine serum albumin (5 g/L), sodium azide (0.9 g/L), and Tween 20 (2 mL/L). The two gold-antibody conjugates were then mixed at a ratio of three parts of 090-12710 conjugate per one part of 057-10030 conjugate. Once buffered, the reagent could be used immediately. By contrast, the preparation of latex reagents would require extensive centrifugation and washing steps to eliminate unbound or loosely bound IgG (21)(22). The further addition of bovine serum albumin and Tween after buffering at high pH prevented nonspecific bridging between particles.
On the Cobas Mira, we used a reaction mode with sequential additions of
sample (30 µL) and reagent [120 µL of accelerating buffer
containing 200 mmol/L Tris-HCl (pH 8.0), 70 g/L poly(ethyleneglycol)
6000, 15 g/L poly(ethyleneglycol) 35 000, 0.9 g/L sodium azide, and 10
mL/L Triton X-100], followed by mixing and the addition of 50 µL of
gold-antibody conjugate. Because the viscosities of the sample and
reagents were substantially different and to improve repeatability, we
added a small volume of water (2 µL by pipetting) immediately after
the addition of the gold-antibody conjugate, so that the instrument
mixed the cuvette content a second time. Using the optimal
sample:reagent ratio, we tested the addition of several components to
the buffer to improve the signal specificity
(18)(19). Magnesium chloride, mannitol, and EDTA
were each found to substantially improve the signal specificity at
concentrations up to 0.25, 0.11, and 0.01 mol/L, respectively. During
the kinetics of interaction, which occurred after addition of the
gold-antibody conjugate (Fig. 1A
), the A600 changes were quick
and linear during the first 2 min; consequently, we limited the
specific reaction (sample-reagent-gold conjugate) incubation time to
2.08 min (5 cycles of 25 s). Because of the good linearity of the
kinetics, we programmed the instrument in kinetic mode, which allowed
us to report the response as change in
A600 per min. Fig. 1B
displays the
mean (± SD) of three calibration curves performed in duplicate in the
optimized conditions.
|
With this optimized assay, the least detectable concentration
calculated at 3 SD above the zero calibrator was 4.5 µg/L. A
Heidelberger-Kendall curve performed up to 6000 µg/L indicated that
the equivalence point (hook effect) was >1200 µg/L. The kinetic
reading mode of the assay allowed the identification of samples with
very high ferritin contents. Such samples showed an earlier plateau and
a marked increase in A600 during the
first 25 s of incubation (Fig. 1A
), which was well above that of
the highest calibrator. Thus, the program included a
A600 threshold for the first
25 s above which the instrument reassayed samples automatically
after proper dilution to avoid hook effect errors encountered with
similar sandwich immunoassays for ferritin (23).
Assay precision was evaluated according to the NCCLS protocol (24) using three control sera (52, 123, and 334 µg/L ferritin, respectively). Within-run, between-run, between-day, and total imprecisions (CVs) were 1.5–4.3%, 2.3–3.8%, 3.5–5.3%, and 4.8–7.9%, respectively. These results were similar to those observed with latex immunoassays for ferritin (9)(10).
Recovery with liver ferritin added to serum (62–3239 µg/L; n = 8) was 90.7–109.5% (mean ± SD, 100.1 ± 6.3%), which compared favorably with latex (10) and ELISA (25) assays.
We evaluated interferences according to NCCLS and Société Française de Biologie Clinique (20) protocols at three ferritin concentrations. Bilirubin did not interfere up to the highest concentration tested (300 mg/L). Similarly, no interferences were observed from the common anticoagulants heparin (up to 0.5 g/L), EDTA, sodium citrate (10 g/L), and sodium fluoride (up to 5 g/L). Hemoglobin concentrations >2 g/L produced a positive linear interference. The major interference found with this assay was exerted by lipids. The interference was particularly marked for low-concentration samples. A similarly substantial lipid interference requiring the pretreatment of turbid samples by centrifugation or with a lipid-clearing agent has been reported for a commercially available latex-based reagent (10). IgM rheumatoid factor (RF) also exerted a substantial positive interference in the assay. This could be expected because the gold particles were coated with complete murine antibody IgG molecules. This latter type of interference reportedly could be avoided by either coating the gold particles with the F(ab')2 fragments instead of the complete IgG (26) or by use of a polyclonal antiserum against RF (27). The RF interference was greatest at low ferritin concentrations. This may indicate that, at higher analyte concentrations, the ferritin-antiferritin agglutination process impairs RF access to the Fc fragment of the antibodies.
We compared the results of unknown samples observed with this new
technique to those obtained with the ACS:180 technique from Chiron
Diagnostics (chemiluminescence) and the Unimate latex reagent from
Roche Diagnostics (Table 1
). The residual plot did not indicate a substantial difference
for the ACS. Results were higher with the colloidal gold than with the
Unimate (slope, 1.18; Sy|x = 0.035), mainly
because of a single sample that for which the measured concentration
was more than twice that observed with latex. This discrepancy was not
observed in the ACS:180 correlation, and both results for the same
sample were in agreement. Because the results obtained with the ACS:180
method, which was not based on particle agglutination, fully agreed
with the colloidal gold results, we suggest that the discrepancy
resulted primarily from differences in epitope recognition by the
antibodies used in the agglutination techniques or in the tissue
ferritin used in the different calibrators, as noted recently
(28), rather than from a nonspecific interaction with the
gold probes. This is further supported by the fact that identical
discrepancies were observed during an evaluation of this gold reagent
in an independent laboratory, involving comparative studies with the
Unimate latex and another latex reagent of commercial origin. With the
outlier excluded, the regression slope was 1.10
(Sy|x = 0.028) and the mean bias was 0.95 µg/L.
|
Despite the random and unidentified interference problems identified in the early days of its use as the reporter reagent in SPIA (17), colloidal gold is likely to constitute an interesting alternative to latex in homogeneous immunoassays. We have demonstrated that the reagent is applicable to a ferritin homogeneous immunoassay. This was made possible by reagent optimization, which eliminated the nonspecific interactions that could lead to a plasmon-resonance effect (18)(19) and interfere with the signal resulting from the specific agglutination of the antibody-sensitized particles. However, for the interference with RF to be circumvented (26)(27), further investigation and assay validation with samples containing other autoantibodies or myeloma paraproteins will be required. The colloidal particles are much easier to synthesize and to coat by charge adsorption with the suitable antibodies than latex (20). A recent report (21), which compared two covalent-coating procedures with one passive adsorption procedure for antibody immobilization on latex particles, concluded that, despite the centrifugation and washing steps involved, substantial aggregation occurred. Even when latex particles with sizes similar to colloidal gold were used for reagent preparation to decrease the sedimentation rate during storage, expensive synthetic materials and tedious covalent-coating procedures were involved, including several ultracentrifugation and sonication steps (22). Once coated and properly stabilized, the colloidal gold particles did not show any sign of nonspecific agglutination. To date, we have used reagent preparations stored at 4 °C for >20 months without any substantial loss in reactivity. The performance of the ferritin assay reported here compares at least equally with other immunoassay technologies, including latex, ELISA, and chemiluminescence.
Our experience with the colloidal gold reagent in immunoassay points to
other important advantages that further ease preparation compared with
latex. In particular, application of the technology to other analytes,
such as the human cardiac fatty acid-binding protein
(18)(19), has shown that the detection limit
that could be reached was 
1 log-unit lower than the detection limit
achievable with a matching latex reagent.
Acknowledgments
We acknowledge the support of Roche Diagnostic, which provided the latex reagent and a Cobas Mira instrument for the performance of the correlation studies. We are also grateful to Drs. Eisenwiener, Dessauer, Karl, and Kürtzinger of Roche Diagnostics Research and Development for the fruitful discussions and their permission to publish the results of the correlation studies.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
), 10 (
), 50
(•), 150 (
), 300 (
), and 600 (
) µg of human ferritin/L.
For comparison, the kinetics of a sample containing a ferritin
concentration above the equivalence point (2000 µg/L) is also
displayed (
). (B), the signal data are the means
± SD (bars) from three independent curves in
duplicate.