Immunological Fingerprinting Method for Differentiation of Serum Samples in Research-Oriented Biobanks

Abstract

An immunoenzymatic serum fingerprinting method was developed to establish a serum sample fingerprint based on IgG titers obtained with three different antigens. Three widely expressed antigens were selected for their capacity to induce long-lasting humoral immune responses. This fingerprinting method may be used to differentiate between two serum samples and to determine whether they come from the same primary blood specimen. The method showed a specificity of 99.5%. This method is suitable as a quality control method for biobanked serum samples.

Biobanks are infrastructures specialized in the acquisition, processing, validation, conservation, management, and distribution of renewable (cells and nucleic acids) and nonrenewable (serum, plasma, solid tissues, and proteins) biospecimens for research purposes. Serum and plasma samples stored in biobanks are often used in simple or multiplex serological or high-throughput proteomic comparative case-control studies. The scientific validity of such studies depends on a number of preanalytical, analytical, and postanalytical parameters. Indeed, preanalytical variations can introduce significant bias into molecular profiles for diseases (18).

The importance of the traceability of serum samples in biobanks is now recognized (9). However, the fact that two serum samples may originate from the same whole-blood specimen, leading to potential errors in the identification/differentiation of such samples, is less commonly acknowledged. It is important that samples that are considered different (or as coming from different donors) are indeed different and are not aliquots of the same original biospecimen which have inadvertently been mislabeled. Such errors in labeling may take place during sample collection, processing, or aliquoting. The statistical validity of research results based on the use of mislabeled samples supposedly derived from different patients but in reality derived from the same primary biospecimen may then be compromised severely, with the impact being particularly marked in studies based on relatively small numbers of cases and controls. A serum fingerprinting tool determining whether two serum samples are derived from different or the same original whole-blood specimens may be useful as a quality control tool. Mass spectrometry methods can be used to establish individual apolipoprotein CI, CII, and CIII profiles, but these methods are not widely available (19). We aimed to develop a reliable yet widely applicable and easy-to-implement method.

We established a serological fingerprint on the basis of the titers of polyclonal IgG antibodies specific for three different antigens. Widely distributed and persistent serological markers were needed for high resolution and specificity. These three antigens were selected for their immunogenicity and capacity to induce long-lasting humoral immunity. We selected synthetic peptides from the EBNA1 (EBV nuclear antigen 1) protein of Epstein-Barr virus (EBV), the Bordetella pertussis toxin (PT), and the outer membrane protein 2 (OMP2) of Chlamydia pneumoniae.

Most of the human population becomes infected by EBV, the agent of infectious mononucleosis. More than 90% of infected individuals remain infected throughout their lifetime, with EBV establishing nonproductive, quiescent, persistent infections of B lymphocytes. The genomic persistence of the virus is associated with the production of proteins such as EBNA1. Specific anti-EBNA1 antibodies are produced during primary infection. They usually persist throughout an individual's lifetime but may disappear in immunodeficient patients (1, 14). EBNA1 contains an immunodominant epitope containing the sequence Ser-Ser-Ser-Gly-Ser-Pro-Pro-Arg-Arg-Pro-Pro-Pro-Gly-Arg-Arg- Pro-Phe-Phe-His-Pro-Val, corresponding to amino acid positions 389 to 409 of the protein sequence (GenBank accession number PO3211) (3, 11).

B. pertussis is the causative agent of whooping cough. Most populations are vaccinated against it, with either the inactivated whole-cell B. pertussis vaccine or the acellular vaccine, composed of a combination of recombinant antigens, including PT. Vaccinal anti-PT antibodies persist for many years after infection or after vaccination (10, 12). PT contains an immunodominant epitope containing the sequence Leu-Thr-Trp-Leu-Ala-Ile-Leu-Ala-Val-Thr-Ala-Pro-Val-Thr-Ser-Pro-Ala-Trp-Ala-Asp-Asp, corresponding to amino acid positions 16 to 36 of the protein sequence (GenBank accession number PO4977) (16, 21).

C. pneumoniae is a causative agent of respiratory infections varying from asymptomatic infections to bronchitis and pneumonia. The majority of the population comes into contact with C. pneumoniae and acquires long-lasting humoral immunity (5, 7, 13, 23, 24). The cysteine-rich OMP2 protein (GenBank accession number P23700) is immunodominant (20) and contains an N-terminal immunoreactive epitope with the sequence Met-Thr-Ala-Lys-Lys-Val-Arg-Leu-Val-Arg-Arg-Asn-Lys-Gln-Pro-Val-Glu-Gln-Lys-Ser. This fragment was found to have a hydrophobicity index of −2.5 and was thus considered a potential immunodominant epitope (17).

The method is based on the principle that for at least one of the three serological markers (pepPT, pepOMP2, and pepEBNA1), the difference in IgG levels between two different samples will be significantly larger than the intra-assay coefficient of variation (CV%) for each marker. In contrast, overlapping confidence intervals for the IgG levels of all three markers suggest that the two samples were derived from the same blood specimen, i.e., from the same donor at a given time.

MATERIALS AND METHODS

Triplex peptide fingerprinting enzyme-linked immunosorbent assay (ELISA).

The three peptides pepEBNA1, pepPT, and pepOMP2, coupled to an N-terminal biotin molecule through a Gly-Gly bridge, were synthesized by Neosystem (Strasbourg, France). The purity of each peptide preparation was checked by mass spectrometry. Each peptide was diluted to 2 μg/ml in 0.1 M carbonate-bicarbonate buffer, pH 9.6, 4% NaCl, and 0.1% Tween 20 and added to streptavidin-precoated Nunc microplates for 1 h at 37°C. Microplates were dried and used on the same day. Aliquots of sera diluted 1:40 in phosphate-buffered saline (PBS) containing 0.1% Tween and 4% NaCl were added to the wells of microplates coated with pepPT and pepOMP2. Aliquots of sera diluted 1:400 in PBS containing 0.1% Tween and 4% NaCl were added to microplates coated with pepEBNA1. All plates were incubated at 37°C for 60 min. Plates were then washed three times with PBS-Tween 20 and incubated with a 1:1,000 dilution of alkaline phosphatase-conjugated goat antibody to human IgG (Sigma) for 60 min at 37°C. The wells were washed three times, and the alkaline phosphatase substrate para-nitrophenylphosphate was added. Plates were incubated at room temperature for 20 min for the pepEBNA1 reaction and for 30 min for the pepPT and pepOMP2 reactions. Reactions were stopped by adding 3 M NaOH, and the plates were read at 405 nm.

Serum fingerprint analysis software.

An analysis program was developed to allow iterative comparison of the triplex fingerprints of numerous serum samples. The program compares each sample with all other samples (one by one), using two different levels of confidence. It thus calculates the mean optical density at 405 nm (OD₄₀₅) and the corresponding standard deviation (SD) and iteratively compares them for each of the three serological markers. For each pair of samples and for each serological marker analyzed, it tests whether the mean OD₄₀₅ + 2 SD for the sample with the lower value is lower than or equal to the mean − 2 SD for the sample with the higher value; if this is the case, the samples are considered to be different, with 95% confidence (risk of error is 0.975 × 0.975). If this is not the case, then the program tests whether the mean OD₄₀₅ + 1 SD for the sample with the lower value is lower than or equal to the mean − 1 SD for the sample with the higher value. Samples for which this is true are considered different, with 71% confidence (risk of error is 0.84 × 0.84). If this is still not the case, there is a >30% risk that the samples are identical.

Results are given as lists showing serum sample pairs considered to be different from all other samples with 95% confidence, serum samples considered to be different from the other samples with 71% confidence, and sample pairs with a >30% risk of being identical.

The analysis software was developed as an Access application. Input data were batch number, serum sample number, the three OD₄₀₅ values for pepEBNA1, the three OD₄₀₅ values for pepPT, and the three OD₄₀₅ values for pepOMP2. Successive queries established the mean values and the 2-SD and 1-SD confidence intervals and compared each sample with the others. The output includes the list of sample pairs with no overlapping within the 2-SD confidence interval, the list of sample pairs with overlapping for the 2-SD interval but no overlapping for the 1-SD interval, and the list of sample pairs with overlapping within the 1-SD confidence interval. Overlapping of both the 1-SD and 2-SD confidence intervals for the OD values of all three peptides indicates that the corresponding samples are identical.

Validation of serum fingerprinting method precision, specificity, and robustness.

To validate the precision of the method, three different serum samples from the Picardie Biobank were used as reference samples. They were tested at different dilutions against pepEBNA1, pepPT, and pepOMP2 to determine the working dilutions that gave low (L), intermediate (I), and high (H) OD₄₀₅ values for each of the three peptide antigens. Precision was determined using the CLSI method (6): for each peptide ELISA, serum samples corresponding to L, I, and H OD₄₀₅ values were tested on five different days in the following order, without modification or interruption: I, H, L, I, I, L, L, H, H, and I.

To validate the specificity of the method, we used 91 serum samples from the Picardie Biobank which were known to be from different patients. These samples were tested against pepEBNA1, pepPT, and pepOMP2 in triplicate.

To validate the robustness of the method, we used 3 serum samples from the Picardie Biobank which were aliquoted in cryotubes and stored at −80°C. For each serum, we tested one aliquot after 1 freeze-thaw cycle, one aliquot after 2 freeze-thaw cycles, and one aliquot after 10 freeze-thaw cycles. These aliquots were tested against pepEBNA1, pepPT, and pepOMP2 in triplicate.

Study population.

Serum samples were collected from 144 female patients who had been included in a previous case-control study (median age, 28 years [range, 16 to 50 years]). The previous study was conducted to determine the diagnostic value of Chlamydia trachomatis-associated anti-Chsp10 and/or anti-Chsp60 antibodies in the detection of secondary infertility. It showed that combined detection of C. trachomatis-associated anti-Chsp10 and anti-Chsp60 antibodies allowed specific diagnosis of secondary infertility (8). Informed consent was obtained from all patients. Aliquots of serum samples were stored at −80°C. We selected 30 samples for which similar anti-C. trachomatis, anti-Chsp10, and anti-Chsp60 antibody titer values had been obtained by ELISA to determine, using the fingerprinting method, whether these samples differed. Serum samples were tested in a blinded fashion and in a randomized order. Each serum sample was tested in triplicate for each of the three serological markers, and the mean OD₄₀₅ value was calculated for each sample and for each marker.

RESULTS

Peptide ELISA validation.

To validate our method, we determined the working dilutions that gave “low,” “intermediate,” and “high” OD₄₀₅ values, using reference serum samples against each of the three peptide antigens (Table 1).

TABLE 1.

IgG OD₄₀₅ values obtained with dilutions of serum reference samples tested with pepEBNA1, pepPT, and pepOMP2

Peptide and reference serum no.	Dilution	OD405a
pepEBNA1
1	1/2,560	0.11-0.17 (L)
	1/320	0.76-0.95 (I)
	1/80	1.92-2.16 (H)
pepPT
3	1/160	0.11-0.17 (L)
1	1/80	0.81-0.91 (I)
	1/20	1.98-2.16 (H)
pepOMP2
2	1/160	0.04-0.08 (L)
3	1/80	0.14-0.22 (I)
	1/20	0.36-0.66 (H)

^aL, low; I, intermediate; H, high.

The combined uncertainty of measurement for each of the three serological markers and each of the three different OD₄₀₅ levels for each marker (low, intermediate, and high OD₄₀₅ values) was established using the CLSI method (6) (Table 2). Combined uncertainty of the mean of triplicate determinations included the day-to-day variation, interbatch variation, the spectrophotometer, and the spectrophotometer calibrator uncertainties. We applied a coverage factor of kp = 1.96, producing a 95% confidence interval and assuming a normal distribution (15).

TABLE 2.

Intra-assay CV% for individual OD₄₀₅ values and expanded uncertainties and CV% for the mean OD₄₀₅ for each serological marker and for each OD₄₀₅ level tested

Parameter	Value for OD405 level
	pepPT		pepOMP2			pepEBNA1
	0-0.18	>0.18	0-0.12	0.12-0.35	>0.35	0-0.5	0.5-1.4	>1.4
Intra-assay CV%	38.6	16.3	37.5	14.3	33.5	22.3	17.2	3.9
Combined μa	0.028	0.023	0.008	0.02	0.11	0.02	0.08	0.046
Combined 2μa	0.055	0.045	0.016	0.04	0.22	0.04	0.16	0.09
Combined CV%	42.5	20.8	27.1	22.1	42.4	28.5	19.1	4.4

^aμ, standard deviation.

Method specificity was defined as the risk of falsely concluding that two samples are identical when in reality they are different. The number of different possible binary combinations was (91² − 91)/2 = 4,095, of which only 22 gave erroneous results, falsely identifying serum samples as identical. Therefore, the discriminatory power and specificity of the fingerprinting method were 99.5% (22/4,095 × 100).

Method robustness against multiple freeze-thaw cycles was shown. No significant differences were found when serum samples were tested against either pepEBNA1, pepPT, or pepOMP2 after 1, 2, or 10 freeze-thaw cycles (Fig. 1).

FIG. 1.

Variation in serum fingerprint results after 1, 2, and 10 freeze-thaw cycles. The impact on each of the three serum fingerprint components, pepEBNA1, pepOMP2, and pepPT, and specific intra-assay variations at the observed OD₄₀₅ levels are shown.

Application of the serum fingerprinting method.

Among 30 highly “suspicious” serum samples tested with the serum fingerprinting method, the following sera were found to be identical: sera 042, 043, 047, and 049, corresponding to the “parent” blood specimen 042; sera 102, 103, 106, 108, 109, 110, 117, 118, 119, and 120, corresponding to the “parent” blood specimen 102; and sera 112, 114, 115, and 126, corresponding to the “parent” blood specimen 112. Therefore, 18 of the 144 serum samples (12.5%) on which the original research study was based did not correspond to distinct research subjects.

Of the (30² − 30)/2 different binary combinations, 86.4% had at least one serological marker showing mean OD₄₀₅ values that were considered to be significantly different using a 2-SD confidence interval, 6.7% had at least one serological marker showing mean OD₄₀₅ values considered to be significantly different using a 1-SD confidence interval, and 6.9% showed overlapping for all three serological markers, thus corresponding to identical samples (Fig. 2). The majority (92.8%) of combinations considered to be different serum samples had a confidence level of 95% (2 SD), with only 7.2% having the more ambiguous confidence level of 71% (1 SD) (Table 3).

FIG. 2.

Distribution of observed fingerprinting profiles for all possible binary combinations of serum samples. Peptide names in capitals indicate OD₄₀₅ values considered to be significantly different using a 2-SD confidence interval. Peptide names in capitals and parentheses indicate OD₄₀₅ values significantly different using a 1-SD confidence interval. The underscore for one category indicates overlapping OD₄₀₅ values for all three peptide serological markers.

TABLE 3.

Comparison profiles observed among the (30² − 30)/2 binary combinations corresponding to different samples

Profile	P value or result of comparisona			Occurrence (%)b
	pepEBNA1	pepPT	pepOMP2
1	<0.05	<0.05	<0.05	41
2	<0.05	<0.05	Overlapping	4.7
3	<0.05	<0.05	<0.3	9.6
4	Overlapping	<0.05	<0.05	3.7
5	<0.3	<0.05	<0.05	3.2
6	<0.05	Overlapping	<0.05	12.4
7	<0.05	<0.3	<0.05	10.6
8	<0.05	Overlapping	Overlapping	1.5
9	Overlapping	<0.05	Overlapping	3.5
10	Overlapping	Overlapping	<0.05	0.7
11	<0.05	<0.3	<0.3	0.7
12	<0.05	Overlapping	<0.3	1.2
13	<0.3	Overlapping	Overlapping	0.7
14	Overlapping	Overlapping	<0.3	2.5
15	Overlapping	<0.3	Overlapping	3
16	<0.3	<0.3	Overlapping	0.5
17	Overlapping	<0.3	<0.3	0.5

^aP < 0.05, mean OD₄₀₅ values were considered significantly different using a confidence interval of 2 SD; P < 0.3, mean OD₄₀₅ values were considered significantly different for a confidence interval of 1 SD.

^bPercentages in bold correspond to samples with at least one marker showing a significant difference in OD₄₀₅ values with a P value of <0.05.

We further examined these 29 ambiguous serum samples. The majority of these samples were shown to be identical by deductive reasoning, using binary relationships between serum samples (if a = b and b = c, then a = c). Serum 045 was thus included in the cluster with sera 042, 043, 047, and 049, and sera 113 and 127 were included in the cluster with sera 102, 112, 114, 115, and 126. Only two serum sample combinations remained ambiguous after this analysis. These were the 117-120 and 118-120 combinations, both of which showed mean OD₄₀₅ values for pepOMP2 that were considered significantly different using the 1-SD confidence interval, with overlapping values for pepEBNA1 and pepPT.

Table 4 shows the analysis program output for representative examples of serum sample comparisons.

TABLE 4.

Representative output of serum fingerprinting analysis program^a

^aMean OD₄₀₅ values considered significantly different using a 2-SD confidence interval are highlighted in green, mean OD₄₀₅ values considered significantly different using a 1-SD confidence interval are in yellow, and overlapping mean OD₄₀₅ values are in red.

DISCUSSION

We tested, at a specified level of confidence, whether two serum samples that were supposedly different were really different or may have inadvertently been mislabeled. The biobank holds records of samples from x different individuals; however, in reality, blood was collected from fewer than x individuals and serum from one individual was aliquoted in tubes corresponding to more than one identification number. This presents a challenge in biobanking that is existent but is rarely acknowledged. Such an anomaly may completely invalidate the statistical significance and interpretation of research results. We demonstrate a method to overcome this challenge by analyzing human serum samples for the presence of specific antibodies for three antigens, pepPT, pepOMP2, and pepEBNA1, and comparing the corresponding serological fingerprints.

We increased the resolution of the serum fingerprinting method by selecting antigens that induce the production of long-lasting infectious or vaccine antibodies. We thus used three synthetic peptides, each from an antigen inducing humoral immunity lasting at least 10 years after infection or vaccination. These three antigens were sufficient to develop a serum fingerprinting method with 99.5% specificity.

The method described provides a qualitative assessment of the identities of two serum samples. This qualitative method does not depend on absolute IgG antibody titers but is based on comparisons between the relative OD₄₀₅ values for three serological markers. Since the method is not designed to provide quantitative data, it cannot be validated in terms of accuracy, linearity, and sensitivity. We validated the precision of the method by determining intra- and interassay reproducibility. Assigning a “positive” result to two identical samples, the “diagnostic” specificity of the method is equivalent to the risk of obtaining a false conclusion that two samples are identical when these samples are actually different.

In using this fingerprinting method for the quality control of a panel of supposedly different serum samples, care should be taken to ensure that the samples compared have been maintained under the same storage conditions (time, temperature, and storage container type) and that no differential evaporation has occurred. It has previously been shown that IgG antibodies are extremely stable during storage at −80°C and when exposed to multiple freeze-thaw cycles (4). Our results suggest that the fingerprinting method described herein is robust despite these preanalytical conditions.

Serum fingerprinting does not allow discrimination between individual donors of serum samples, as the same donor may have different IgG titers against a specific antigen over time. The method allows discrimination either between individual donors from whom samples were collected at a given time or between primary blood specimens.

The method described here offers a novel, easy-to-use, and specific tool for quality control and quality assurance, overcoming a rarely acknowledged anomaly in biobanks (22).