Background
Globally influenza remains a significant cause of morbidity/mortality in human populations [
1], with 3–5 million severe clinical infections yearly resulting in approximately 250,000–500,000 fatalities [
2,
3]. Antibodies (Abs) that are specific to the viral envelope protein hemagglutinin (HA) prevent binding of virus to target cell sialic acid residues and are largely thought to mediate immune protection against influenza viruses [
4]. Viral HA will cross link erythrocytes, leading to increased rates of sedimentation. Hemagglutination-inhibition (HAI) assays measure antibody titers, by measuring the inhibition of the agglutination of erythrocytes [
5,
6]. An additional serological assay that is employed to assess influenza Abs includes the microneutralization (MN) assay, with an enzyme-linked immunosorbent assay (ELISA)-based endpoint assessment, that is used to detect neutralizing Ab titers through inhibition of virus infection [
7,
8]. Although a variety of additional serological assays to assess functional anti-influenza antibody titers exist [
9‐
11], the MN and, in particular, the HAI assays remain gold-standard assays to evaluate immunogenicity of influenza vaccines and to measure serological responses to natural infection [
12]. However, there is increasing awareness of the variability of influenza serological assay results [
12] and the need to improve inter-laboratory agreement on serological assay standards [
13]. This is especially relevant for: (i) assessment of influenza vaccine immunogenicity; (ii) epidemiological studies seeking to catalog newly emerging viral strains in an affected region over time; and (iii) assessment of clinical samples for anti-influenza Ab activity.
Developing effective vaccines against seasonal and pandemic influenza is a public health priority. Variation across laboratories, including operator inexperience influencing reproducibility and lack of common standardized neutralization and HAI assays, in serological procedures limits comparisons of vaccine strategies and vaccine efficacy [
12,
14]. Protocols are also not standardized in regards to expression of endpoint and this contributes to variable reporting [
14,
15]. An additional variable that can affect serological results is the choice of sera versus plasma for measuring samples, and the presence of anticoagulants in plasma.
Historically, serum has been the sample of choice for HAI assessments [
16,
17]. However, plasma has been gaining popularity for human subject research studies [
18]. Plasma has the potential disadvantage compared to sera in diagnostic assays, as anticoagulants have been demonstrated to interfere with antibody-antigen interaction and some enzyme reagents [
19,
20]. However, in regards to the HAI assay, we have previously determined that both serum and plasma (serum-citrate plasma and serum-heparinized plasma) can be used in serodiagnostic assays (seroconversion rates remaining unaffected by sample type), but that the use of plasma samples may underestimate HAI titers [
21]. Reasons that anticoagulants may affect assay readout are varied but include: (i) sodium citrate or EDTA acting as chelating agents and binding enzyme cofactors affecting enzyme activity in assays (increasing or decreasing titers); (ii) heparin being shown to interfere with antibody-antigen reactions [
19,
20] potentially biasing towards decreased titers; and (iii) plasma clots mimicking agglutination patterns in affected wells biasing towards a lower titer readout. A more detailed analysis of the influence of sample type (serum versus plasma) and the sample collection method (including various anticoagulant agents) on HAI and MN assays is warranted as plasma becomes a more routinely obtained specimen. This is especially relevant as sample collection methods may either overestimate or underestimate seroprevalence rates and this can impact assessment of vaccine efficacy, as new influenza vaccine are expected to meet specific seropositive and seroconversion rate thresholds, such as those outlined by the United States Food and Drug Administration (FDA) guidelines on licensure [
22].
The purpose of this study was to evaluate the influence of sample collection methods on influenza virus serological assays to inform on guidelines for sample collection and assay standardization. In particular, we assessed the effect of the use of various collection tubes and anticoagulants on the HAI and MN variability compared to standard serum samples collected using serum separator tubes. Blood samples collected from thirty donors previously vaccinated against influenza were collected using six different sample collection tubes (Serum: (1) serum separator tubes (SST); and (2) Plus Plastic serum “red-top serum” tubes. Plasma: (3) spray-coated K2 ethylenediaminetetraacetic acid (EDTA) tubes: (4) Sodium Heparin tubes; (5) Citrate tubes with 3.2% sodium citrate solution; and (6) Glass Blood Collection tubes with acid citrate dextrose (ACD) and serum or plasma samples were prepared. Three influenza virus strains were used in this study to evaluate potential differences in HAI and MN titer values associated with these variables.
Discussion
The serological identification of antibodies/antigens in serum/plasma is a rapidly progressing field with utility for both clinicians and scientist. Clinically, serology is useful for: (i) diagnosis; and (ii) monitoring vaccine/treatment efficacy in order to aid clinical management decisions. Scientifically, serology is useful for: (i) conducting seroepidemiological investigations; and (ii) to inform on the breadth of host immune responses to a particular pathogen [
28]. Although serological assays have improved in recent years, variability in assay methods across laboratories remains an issue impeding greater standardization in result reporting. This is particularly true in the field of influenza. It has become increasingly recognized that blood collection methods remain a frequently overlooked aspect of protocols that may be responsible for uncontrolled variability [
29]. As such, we sought to assess the influence of different blood collection methods, specifically different blood tubes, for influenza HAI and MN reporting.
In general, intra-laboratory titer values that are within two dilution factors (four-fold difference) from duplicate and/or repeated testing of a particular sample is considered acceptable and the values considered to be comparable [
12,
30]. In this study we compared HAI titers results from various collection tube methods to SST and demonstrated that there was no alternative sample that across strains had full concordance with the HAI obtained from a SST sample. Agreement between the test sample and SST for single strains ranged between 30 and 100%. Matched samples from the various collection methods were within the two-dilution criteria when assessed for type A and B influenza viruses and for HAI and MN 93% of the time for RT Sera, Hep plasma, and Cit plasma. EDTA plasma tested for type B influenza HAI titers demonstrated ≥ two-dilution variation in 70% of samples. Similarly, but to a lesser degree, 10% of samples from ACD plasma were different to SST for type B influenza MN titers. Based on these observations, we believe that RT serum, Hep plasma, ACD plasma, and Cit plasma could be used in place of or in addition to SST to measure HAI and MN titers if needed, though SST remains the gold standard and should still be used whenever possible.
MN assays are often reported to have greater inter- and, and importantly to the present study, intra-laboratory variability compared to HAI assays [
12,
14,
31]. We did not find this to be the case in the present study. Despite difference in starting material (sera and plasma based on the collection method) we found that only 2% (6 out of 300) of samples deviated from the SST reference for the MN assay. Alternatively, we saw 6.9% (31 out 450) of samples deviate from the SST reference for the HAI assay. Deviation from the SST values occurred more frequently for influenza B assays for both MN (3.3%; 5 out of 150) and HAI (18.7%; 28 out of 150), then for influenza A assays. The greatest variation in the MN assay was for influenza B when using ACD plasma as the starting material. ACD plasma resulted in a greater than two-dilutions higher titer two times compared to SST and a lower than two-dilutions titer once compared to SST. All of the higher titer results for ACD plasma occurred when MN titers for the SST were low (titers of 10–40). At titers above 40 there was more agreement on titer amongst sample collection methods, including ACD plasma. This may indicate that ACD plasma has either greater sensitivity for influenza B MN titer assessment at lower titer than SST or that ACD plasma overestimates titers.
For the HAI assay the most interesting finding was the observation that the EDTA plasma collection method resulted in a ≥ two-dilutions change compared to SST in 21 out of 30 samples assessed for influenza B. Additionally, in no cases was the influenza B HAI titer result for EDTA plasma lower than for SST. The observation that titers for influenza B HAI were higher for EDTA plasma collection tubes compared to SST might indicate that they are potentially overestimating the real titer. It should be noted that the SST titers for influenza B were routinely low, ranging from 10 to 160, and this may have influenced variation in the sample set.
Assessment of antibody responses to specific influenza subtypes is an important diagnostic, epidemiological, and immunological tool [
32]. The observation that we saw higher HAI titers for EDTA plasma for influenza B compared to SST has implications related to seroprevalence and seroconversion studies. Overestimating and underestimating the seroprevalence rates in response to a newly emerging influenza virus is of concern. Additionally, as new influenza vaccines are expected to meet specific seropositive and seroconversion rate thresholds, such as those outlined by FDA guidelines on licensure [
22], the method for assessing these effectiveness rates needs to be carefully considered. Based on our observation of a consistent bias in HAI titers for EDTA plasma for influenza B, compared to SST methods, we recommend that the use of EDTA plasma be considered with extreme care, or avoided, for studies related to seroprevalence and seroconversion following vaccination.
The mechanism of the effect of sample collection method on differences in HAI and MN titers for influenza viruses remains to be determined. Previously it has been hypothesized that anticoagulants in plasma may affect HAI (and MN) titers by lowering titers overall. Indeed we have previously seen lower HAI titers for temporally matched sodium citrate plasma or heparin plasma compared to sera [
21]. In the present study we found that RT sera, ACD plasma, heparin plasma, and 3.2% sodium citrate solution-plasma resulted in generally lower HAI and MN titer values (17/20 had a negative bias compared to SST), but statistically similar, compared to SST sera sample collection. EDTA plasma collection resulted in higher HAI titer values for influenza A and B, with the values being significantly higher for influenza B, compared to SST. However, EDTA plasma collection did not significantly change MN assay results compare to SST. This finding could be potentially due to more carry over of residual anticoagulants for HAI assays that is not seen for MN assays. Anticoagulants may bias towards higher titer results particularly at lower overall titers (as assessed by SST). EDTA is a chelating agent and could hypothetically influence the ability to detect antibody/antigen interactions. Indeed, EDTA treatment of serum has been shown to improve human leukocyte antigen detection [
33]. Anticoagulants in plasma collection tubes may also introduce other interferences related to enzyme inhibitors, fibrinogen, and cations [
34] that may influence HAI and MN titers. Variations could also occur for concentrations of anticoagulants present in tubes, perhaps particularly for EDTA and acid citrate dextrose. However, how EDTA presence may affect Type B influenza HAI titers in particular remains to be elucidated.
The present study has a number of limitation that need to be considered, including limitations related to the viruses used and limitations related to the choice of blood samples used. In the present study, we focused on three strains of influenza representing two type A (an H1N1 and an H3N2) and one type B virus, and representing strains capable of seasonal or pandemic influenza. We were unable to test one of the viruses (Type A H3N2) in both the HAI and MN assay for all collection methods. A limitation of this study is that we did not: (i) investigate additional types of influenza such as C or D; (ii) further investigate additional strains/lineages for influenza B; and (iii) further investigate additional subtypes/strains for influenza A based on hemagglutinin and neuraminidase subtypes. The potential differences in titer values for HAI and MN assays amongst different collection tube methods should, in particular, be assessed for additional A and B influenza viruses as these results would have broader implications for assessing seasonal influenza vaccines and assessing seroprevalence. Similarly, additional viral factors that may influence assay results could be investigated in future studies including: (i) comparing cell-derived versus egg-derived viral strains; and (ii) the effect of ether splitting influenza B viruses prior to conducting serological assays. As newer cell-based influenza vaccines become more common, properly assessing seroconversion rates will have implications for licensure. Therefore both egg-derived and cell-derived viruses should be assessed for influence upon different sample collection methods and serological results. An additional viral factor for consideration in future similar studies would be to assess the effect of ether splitting influenza B viruses prior to assessing serological results. This is important as ether splitting the virus prior to conducting the HAI assay might increase the dynamic range of the assay (and may result in higher titer values). This has implications for the present study as many of the blood samples collected had low HAI titer values (all below 320) but not MN titer values for influenza B, perhaps due to not conducting an ether split.
An additional limitation of this study is related to the choice of samples collected. In this study we utilized blood samples, collected six different ways, from thirty healthy donors previously immunized against influenza viruses. As such this study lacked seronegative controls. In the present study the closest to a seronegative control was sample #9, were all titer values for each virus and each assay for SST were < 40 and all additional titer values for each collection method were ≤ 80. Additionally, the spread of titer values could have ideally been greater. As a result of these two sample selection factor limitations it is more difficult to definitively interpret the findings of the present study. Future investigations into the effect of sample collection methods on serological assays should incorporate susceptible/seronegative controls, and should ideally include more samples with greater titer spread in order to better assess potential non-specific reactivity at lower antibody levels. Further investigation of serological blood collection methods, as well as additional factors that influence assay results, is warranted as we seek to standardize influenza serological assays and improve inter-laboratory cooperation globally in order to better interpret influenza serological assays, estimate influenza severity and attack rates, and inform on public health policy.