Preanalytical Stability of Antibodies to Pathogenic Antigens

Biochemical analyses of blood-based analytes are increasingly being recognized as valuable enhancements to epidemiological studies. Such analyses are typically used to measure exposures to risk factors, biochemical responses to interventions, or potential confounders of other study factors. For epidemiological studies, the collection, transportation, fractionation, aliquoting, and freezing of large numbers of blood samples are onerous tasks. Moreover, these tasks become increasingly difficult and expensive if it is necessary to transport samples at a specific temperature, in a particular type of vacutainer, and to fractionate, aliquot and freeze samples all within hours of being collected. Although guidelines for handling blood samples often suggest that plasma and serum be separated from cells within two hours of collection (1), such nonspecific recommendations may place unnecessary burdens on epidemiological studies for two reasons. First, the decay curves of serum or plasma biomarkers are known to vary (2) with some blood analytes being relatively unaffected by several days delay in fractionation (3, 4). Second, the degradative effects of delays in fractionation are, for some analytes, modified by blood storage temperature (5) and/or the type of vacutainer used for storage (6) prior to fractionation. Understanding these potentially complex interactions for specific analytes may indicate that blood collection is feasible in studies where it was thought otherwise.

Serological testing for antibodies to viral proteins has proven to be a valuable tool for identifying and confirming epidemiological associations between various viral infections and chronic diseases such as cancer (7–9). For example, the presence of circulating antibodies to the human papillomavirus (HPV) late capsid protein (L1) is considered a marker of lifetime exposure to HPV (10) and seropositivity for HPV16 L1 has been shown to be associated with increased risks of cancer of the cervix (7), oral cavity (8), and oropharynx (9). The available evidence suggests that antibodies against infectious agents are generally robust to variations in postfractionation, preanalytic conditions, such as freeze-thawing (11), freeze drying (12), or type of storage tube for the frozen samples (13, 14). In addition, some evidence exists that IgM isotype antibodies are more labile than IgG ones, at least when tested from filter paper (15). However, there has been no formal study to identify whether delays in fractionation and/or blood storage conditions prior to fractionation have a significant effect on antibody measurements. The aim of this study was to quantify and compare the degree of variability in antibody seropositivities to a selected set of antigens within blood samples from the same donors, with different time/temperature precentrifugation conditions for the isolated plasma, different time/temperature postcentrifugation conditions for the isolated serum, and with different vacutainer types.

The Cancer Council New South Wales Stability Study was approved by the Cancer Council NSW Ethics Committee (Ref. #257). Thirty-four employees of Cancer Council (ages between 24 and 62 years) agreed to participate in the study and provided blood samples for analyses. Twenty-nine participants were included in the antibody substudy (Fig. 1 and Table 1). Participants were not required to fast prior to blood collection. Blood was collected by the same trained phlebotomist on five separate collection days. Returning participants were requested to attend for blood collection at the same time on each collection day to minimize intraindividual variation.

On the first four collection days (each separated by a period of approximately 7 days), blood was drawn from each available participant and stored in 14 ethylenediaminetetraacetic acid (EDTA) vacutainers (collection day 1), 14 acid citrate dextrose (ACD) vacutainers (collection day 2), 13 serum separator tubes (SST; collection day 3), and 14 lithium heparin (LH) vacutainers (collection day 4; Fig. 1). Twenty-three of the 29 antibody substudy participants completed all of the first 4 collection days. On collection day 5, blood was drawn from 17 substudy participants and stored in 4 EDTA, 4 ACD, and 3 SST vacutainers. One substudy participant did not provide an ACD sample. Sixteen participants completed all 5 collection days.

Immediately after blood collection, vacutainers were randomly allocated into one of two Styrofoam boxes (kept at 4°C or room temperature inside) and transported to the Cancer Council Biobank, with the exception of SST vacutainers which were only transported at room temperature. The industry standard for SST fractionation requires clot formation prior to refrigeration (16). All vacutainers were received at the Biobank within 15 minutes of blood collection. Once at the Biobank, samples in EDTA, ACD, and LH vacutainers were retained at their storage temperatures until fractionation by centrifugation, followed by aliquoting and freeze storage. Samples from the first four collection days were freeze stored without delay (processing day 0) and at 1, 2, 3, 4, 5, and 6 days (processing days 1, 2, 3, 4, 5, and 6). SST samples were allowed to stand for 20–30 minutes to allow for clot formation followed by fractionation by centrifugation (all within 1 hour of blood collection). Serum from one vacutainer per participant was then aliquoted and freeze stored without delay (processing day 0), whereas the remaining SSTs were randomly allocated to 4°C and room temperature storage conditions prior to aliquoting and freeze storage at 24-hour intervals (processing days 1, 2, 3, 4, 5, and 6). Samples from collection day 5 (EDTA, ACD, and SST vacutainers only) were freeze stored on days 0 and 3 only (processing days 0 and 3).

Time of fractionation was determined by the time of day an individual participant's collections were received at the biobank. The timing of fractionation was consistent across all freeze-storage days, that is, exactly 24-hour intervals. For all samples, centrifugation was performed at 2,500 rpm for 10 minutes, aliquots were 200 μL and freeze-storage temperature was −80°C. Aliquots containing samples from all test conditions for each participant were air freighted on dry ice to the German Cancer Research Center (Deutsches Krebsforschungszentrum, DKFZ) in Heidelberg, Germany, for serological analyses. Aliquots were stored in 1 mL Greiner Bio-one Cryo.s cryovials.

Multiplex serology was used to analyze antibody reactivities to 35 antigens derived from different infectious agents, that is, 11 types of HPV, 9 human polyomavirus types (HPyV), Epstein–Barr virus (EBV), and Helicobacter pylori (H. pylori). The multiplex method is based on a glutathione S-transferase (GST) capture immunosorbent assay combined with fluorescent-bead technology (17, 18). Recombinant GST-X-tag fusion proteins including viral and bacterial proteins were used as antigens, loaded and affinity purified on spectrally distinct glutathione-casein-coupled fluorescence-labeled polystyrene beads (SeroMap, Luminex). Bead sets of different colors each carrying a different antigen were mixed with human serum or plasma at 1:1,000 dilutions. Antibodies bound to the beads via the GST-X-tag fusion proteins were stained with biotinylated goat anti-human IgA, IgM, IgG (Dianova), and the reporter conjugate R-phycoerythrin-labeled streptavidin. A Luminex 200 analyzer identified the internal bead color and thus the antigen carried by the bead. The quantity of bound antibodies was determined as the median reporter fluorescence intensity (MFI) of at least 100 beads per bead set per serum. In the current study, antibody reactivities were analyzed as dichotomous variables (seropositive vs. seronegative) consistent with their use in epidemiological studies. Algorithms used for deriving MFI cutoff values defining seropositivity have been described elsewhere (17–24).

Antigens included in the current analyses were the major capsid protein L1 of 11 (HPV) types (1, 4, 6, 8, 10, 11, 16, 18, 41, 48, and 49) as well as 5 early proteins of HPV 16 (E1, E2, E4, E6, and E7; refs. 17, 19); viral capsid proteins 1 (VP1) of 9 HPyV types (BKV, JCV, LPV, MCV, TSV, KIV, HPyV6, and HPyV7) as well as large T antigen of JCV (20, 21, 25); four proteins from EBV: full-length EA-diffuse (EA-D), viral capsid antigen (VCA) p18, ZEBRA, and a fragment of EBNA-1 (C-terminal part AA 325–641; ref. 22) and from H. pylori 6 full-length or partial proteins (GroEl, UreA, HcpC, Omp, VacA-C, and CagA-N; ref. 23).

The standard clinical practice for samples containing antibodies involves transportation of the samples in SST vacutainers at room temperature and fractionation without delay (referred to hereafter as the “standard clinical practice conditions”; ref. 16). Antigen-specific “referent” measurements were defined for each participant as their seropositivity status from the analysis of samples collected on collection day 3 and then exposed to standard clinical practice conditions (Fig. 1). Cohen's κ statistic was then used as a measure of chance corrected agreement between referent and nonreferent measurements for each individual (note that nonreferent measurements also include measurements from samples exposed to standard clinical practice conditions but collected on collection day 5; see Fig. 1). Cohen's κ can be interpreted as the probability of agreement beyond chance (i.e., adjusted for the probability of agreement at random). As a guideline for κ values, Landis and Koch proposed: Perfect (κ = 1.00), Almost Perfect (0.81–1.00), Substantial (0.61–0.80), Moderate (0.41–0.60), Fair (0.21–0.40), Slight (0.00–0.20), and Poor (<0.00; ref. 26).

To allow tests of various hypotheses related to the effects of preprocessing conditions on agreement, regression analyses were used to model κ with general estimating equation (GEE) adjustment to account for the repeated measurements on the same individuals (27). In these regression models, agreement between referent and nonreferent measurements was defined as the dichotomous dependent variable with identity link and normal distribution approximation, as described elsewhere (28). Independent variables were processing day (modeled continuously using a log function), storage temperature, and vacutainer type with terms added for the interaction between these variables where appropriate (i.e., dependent on the hypothesis being tested). The probabilities of agreement at random for each combination of independent variables were obtained through logistic regression and included as the offset variable in the GEE models (27, 28). Robust standard errors and independent working correlation structures were used in the GEE analyses. Estimates of κ obtained from the GEE models were validated by comparing model point estimates with empirical point estimates calculated using traditional methods (26). In order to increase statistical power and also reduce the complexity of the vast amounts of data generated by this study, results are presented for the following five groups of markers: (i) HPV L1 markers, (ii) HPV E markers, (iii) HPyV markers, (iv) EBV markers, and (v) H. pylori (the conclusions inferred by the grouped and individual marker results were essentially the same).

Thirty-four participants were recruited to the CCNSW Stability Study. Of these, 5 were excluded from the antibody substudy because they did not contribute blood on collection day 3 and thus did not have the defined referent measurements (Fig. 1). Of the 29 substudy participants, 22 (76%) were female and 18 (62%) were under 40 years of age (Table 1). Fourteen (0.8%) of 1,638 aliquots collected from the 29 participants were excluded from all analyses due to handling errors, leaving 1,624 aliquots in the final sample.

Across all five groups of markers, observed agreement between referent and nonreferent measurements was greater than 90% for all four vacutainer types and two storage temperatures (Table 2). Chance corrected agreement (κ) ranged from 0.76 (“substantial” agreement) to 0.89 (“almost perfect” agreement). Kappa was not associated with storage temperature (P value range, 0.23–0.95) or vacutainer type (P value range, 0.35–0.89). There was no evidence of interactions between vacutainer type and storage temperature (P tubeXtemp range, 0.34–0.69).

For all five groups of markers, observed agreement between referent and nonreferent measurements was no less than 89% for any preprocessing conditions involving processing day, storage temperature, and vacutainer type. Kappa ranged from 0.67 (“substantial” agreement; SST, room temperature, processing-day 6 for HPV E markers) to 1.00 (“perfect” agreement; SST, 4°C, processing day 6 for EBV markers; Figs. 2 and 3). Average (per day) change in κ did not significantly decline over increasing processing days for any vacutainer type storage temperature combination (P slope range, 0.06–1.00). Change in κ over processing days did not differ according to vacutainer type (within the same storage temperatures; P slopeXtube range, 0.25–1.00) or storage temperature (within the same vacutainer types; P slopeXtemp range, 0.50–0.82). Because the effects of processing delays were not found to differ according to vacutainer type or storage temperature, we analyzed the data after combining across vacutainer types and storage temperatures (Fig. 4). In these analyses, κ did not significantly decline over for any of the five groups of markers (P slope range, 0.21–1.00).

In the current study, we were unable to explicitly quantify the amount of disagreement that was due to analytic variability because replicate measurements of seropositivity were not performed (with replicates being two or more measurements from each individual's blood taken on the same collection day and exposed to the same preprocessing conditions). Nonetheless, we can broadly quantify the combination of analytic and intraindividual variability through the κ values corresponding to standard clinical practice conditions (but with blood samples taken on separate collection days 3 and 5). This is because, for these κ values, departure from perfect agreement can only be due to analytic and intraindividual variation and not due to samples being exposed to different preprocessing conditions (because the preprocessing conditions were the same). Table 2 shows that κ values corresponding to standard clinical practice conditions were not significantly different (and were similar in magnitude) to κ values corresponding to other preprocessing conditions (P value range, 0.35–0.89). This suggests that for all preprocessing conditions, the observed departures from perfect agreement appear to be primarily driven by analytic and intraindividual variability rather than due to the effects of preprocessing conditions.

In the current study, we found that delays of up to 6 days in the processing of whole blood samples, preprocessing blood storage temperature (room temperature or 4°C) and vacutainer type (SST, EDTA, ACD, or LH) did not significantly affect measurement of antibody seropositivities to 35 antigens. Hence, in so far as seropositivities to these antigens can be reliably measured from samples stored under standard clinical practice conditions, our findings suggest that these antigens can also be reliably measured from samples stored in SST, EDTA, ACD, or LH vacutainers at either room temperature or 4°C and for up to 6 days before processing.

The standard clinical practice for samples containing antibodies involves transportation of the samples in SST vacutainers at room temperature and fractionated without delay (16). The rationale for choosing SST vacutainers over other vacutainer types is that antibody titers are usually higher in serum than in plasma. The rationale for room temperature transportation is that serum blood coagulation for serum extraction is more efficient at room temperature than at cooler temperatures. Although the logic underpinning the standard clinical practice is seemingly sound, there is little empirical evidence to support the practice. It was recently shown that antibody-based functional assays were inhibited by the presence of EDTA and ACD, but not by heparin (29). In the same study, it was shown that anti-V. cholera–specific antibody levels were not significantly affected by anticoagulants, although IgA antibodies in heparinized plasma did not show a good correlation with serum (29). Serum has also been the specimen of choice in hemagglutination inhibition assays, but it was shown that plasma is equally fit for purpose and that citrated plasma showed the lowest variability and the highest agreement with serum samples (30). Many biobanks currently collect and store both serum and plasma samples because EDTA plasma is the sample of choice for proteomic and metabolomic analyses whereas serum is considered the sample of choice for immunologic analyses. Our experimental evidence showing that plasma provides similar results for antibody assays might allow biobanks to streamline their operations by collecting EDTA plasma samples only. Furthermore, in relation to biobanking blood products for future unspecified use in research, the relatively nonspecific collection and processing requirements for serological testing could be combined with assay requirements for other types of biomarkers that utilize the same vacutainer for several assays, reducing collection costs and burden on the participant. There is one exceptional class of IgG whose measurement is biased by the use of EDTA anticoagulant, and this is the platelet IgG (31–33). It has indeed been shown that EDTA anticoagulant significantly interferes with the quantification of platelet associated IgG.

As far as we are aware, our study is the first to directly assess the effects of processing delays, blood storage temperature and vacutainer type on measurements of antibody seropositivities. Our finding that antibody seropositivities can be reliably measured from samples stored at either room temperature or 4°C for up to 6 days before processing could have important practical implications for many blood-based epidemiologic studies. Often the collection of blood samples in large studies is performed at many remote sites over wide geographic areas before transportation to a central laboratory. For these studies, a perceived need to fractionate whole blood samples without delay means that samples must arrive at the central laboratory within hours of blood collection or that they be fractionated at the remote site before transportation. The task becomes even more complex if there is also a perceived need to chill samples (either whole blood or plasma/serum) during transportation. Our results suggest that for studies analyzing the antibodies that we assessed standard courier (or even mail) services could be viable options for the transportation of whole blood to a central facility for processing (3, 4).

This study has several limitations. First, the antibodies used were research not clinical antibodies; conclusions relating to the latter are beyond the scope of this study. Second, these analyses did not distinguish between different antibody isotypes and, therefore, we cannot draw conclusions relative to the robustness of IgG, IgA, or IgM antibodies. Third, because replicate measurements for each aliquot were not performed, we were unable to explicitly quantify disagreement due to analytic variability. Nonetheless, we were able to approximately quantify the combination of analytic and intraindividual variability using the κ values corresponding to standard clinical practice conditions.

In conclusion, we have shown that the measurements of individual's antibody seropositivities to 35 antigens are not impacted by processing delays of up to 6 days or by the choice of vacutainer type (SST, EDTA, ACD, or LH) or storage-temperature (room temperature or 4°C). For epidemiological studies planning to collect blood for serological testing of certain antibodies, our results suggest that blood collection methods might be considerably simpler and cheaper than had been expected.

No potential conflicts of interest were disclosed.

Conception and design: V.S. Hodgkinson, S. Egger, F. Betsou, T. Waterboer, M. Pawlita, M.S. Baker, E. Banks, F. Sitas

Development of methodology: V.S. Hodgkinson, F. Betsou, T. Waterboer, M. Pawlita, M.S. Baker, F. Sitas

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V.S. Hodgkinson, T. Waterboer, M. Pawlita, A. Michel, F. Sitas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V.S. Hodgkinson, S. Egger, T. Waterboer, M. Pawlita, A. Michel, E. Banks, F. Sitas

Writing, review, and/or revision of the manuscript: V.S. Hodgkinson, S. Egger, F. Betsou, T. Waterboer, M. Pawlita, A. Michel, M.S. Baker, E. Banks, F. Sitas

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V.S. Hodgkinson

Study supervision: V.S. Hodgkinson, M. Pawlita

We are grateful to Douglass Hanly Moir Pathologists for donation of BD Vacutainers and the services of a trained phlebotomist for the study. We would like to thank Cancer Council staff for providing blood samples for the study.

The Cancer Council NSW Stability Study is funded by Cancer Council NSW.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.