Abdelhadi
Djaileb‡
ab,
Maryam
Hojjat Jodaylami‡
c,
Julien
Coutu
c,
Pierre
Ricard
c,
Mathieu
Lamarre
d,
Léa
Rochet
a,
Stella
Cellier-Goetghebeur
a,
Devin
Macaulay
d,
Benjamin
Charron
c,
Étienne
Lavallée
a,
Vincent
Thibault
c,
Keisean
Stevenson
c,
Simon
Forest
c,
Ludovic S.
Live
b,
Nanouk
Abonnenc
e,
Anthony
Guedon
e,
Patrik
Quessy
e,
Jean-François
Lemay
e,
Omar
Farnós
f,
Amine
Kamen
f,
Matthew
Stuible
g,
Christian
Gervais
g,
Yves
Durocher
g,
François
Cholette
hi,
Christine
Mesa
h,
John
Kim
h,
Marie-Pierre
Cayer
j,
Marie-Joëlle
de Grandmont
j,
Danny
Brouard
j,
Sylvie
Trottier
k,
Denis
Boudreau
d,
Joelle N.
Pelletier
*a and
Jean-Francois
Masson
*c
aDepartment of Chemistry, Department of Biochemistry and PROTEO, The Québec Network for Research on Protein Function, Engineering and Applications, Université de Montréal, CP 6128 Succ. Centre-Ville, Montreal, Québec H3C 3J7, Canada. E-mail: joelle.pelletier@umontreal.ca; Tel: +1-514-343-2124
bAffinité Instruments, 1250 rue Guy, Suite 600, Montréal, Québec H3H 2L3, Canada
cDepartment of Chemistry, Quebec Centre for Advanced Materials (QCAM), Regroupement Québécois sur les Matériaux de Pointe (RQMP), and Centre interdisciplinaire de recherche sur le cerveau et l'apprentissage (CIRCA), Université de Montréal, CP 6128 Succ. Centre-Ville, Montreal, Québec H3C 3J7, Canada. E-mail: jf.masson@umontreal.ca; Tel: +1-514-343-7342
dDepartment of Chemistry and Centre for Optics, Photonics and Lasers (COPL), Université Laval, 1045, av. de la Médecine, Québec City, Québec G1V 0A6, Canada
eCNETE and PROTEO, The Québec Network for Research on Protein Function, Engineering and Applications, Cégep de Shawinigan, 2263 Avenue du Collège, Shawinigan, Québec G9N 6 V8, Canada
fDepartment of Bioengineering, McGill University McConnell Engineering Building, 3480 University Street, Montreal, Québec H3A 0E9, Canada
gMammalian Cell Expression, Human Health Therapeutics Research Centre, National Research Council Canada, Montréal, Québec, Canada
hNational Laboratory for HIV Reference Services, National Microbiology Laboratory at the JC Wilt Infectious Diseases Research Centre, Public Health Agency of Canada, Winnipeg, Canada
iDepartment of Medical Microbiology and Infectious Diseases, University of Manitoba, Winnipeg, Canada
jHéma-Québec, Affaires médicales et innovation, 1070, avenue des Sciences-de-la-Vie, Québec City, G1V 5C3, Québec, Canada
kCentre de recherche du Centre hospitalier universitaire de Québec and Département de microbiologie-infectiologie et d'immunologie, Université Laval 2705, boulevard Laurier, Québec City, Québec, Canada G1V 4G2
First published on 1st July 2021
We report on the development of surface plasmon resonance (SPR) sensors and matching ELISAs for the detection of nucleocapsid and spike antibodies specific against the novel coronavirus 2019 (SARS-CoV-2) in human serum, plasma and dried blood spots (DBS). When exposed to SARS-CoV-2 or a vaccine against SARS-CoV-2, the immune system responds by expressing antibodies at levels that can be detected and monitored to identify the fraction of the population potentially immunized against SARS-CoV-2 and support efforts to deploy a vaccine strategically. A SPR sensor coated with a peptide monolayer and functionalized with various sources of SARS-CoV-2 recombinant proteins expressed in different cell lines detected human anti-SARS-CoV-2 IgG antibodies in clinical samples. Nucleocapsid expressed in different cell lines did not significantly change the sensitivity of the assays, whereas the use of a CHO cell line to express spike ectodomain led to excellent performance. This bioassay was performed on a portable SPR instrument capable of measuring 4 biological samples within 30 minutes of sample/sensor contact and the chip could be regenerated at least 9 times. Multi-site validation was then performed with in-house and commercial ELISA, which revealed excellent cross-correlations with Pearson's coefficients exceeding 0.85 in all cases, for measurements in DBS and plasma. This strategy paves the way to point-of-care and rapid testing for antibodies in the context of viral infection and vaccine efficacy monitoring.
The immune system produces antibodies to SARS-CoV-2 within days to a few weeks following viral infection.5 Antibodies are expected to remain at a high level for months following infection, as previously shown following the 2003 outbreak of SARS-CoV-16,7 which has also been reported in sera and saliva of COVID-19-positive individuals.8 The immune reaction to coronaviruses generally provides acquired immunity via neutralizing antibodies9 in the event of a second exposure to the virus and also provides the basis for vaccine development. Vaccine development, clinical trials and immunity studies require assessing antibody titers or concentrations in animal and human subjects. As such, serological antibody testing is essential to assess the fraction of the population that is immune to a virus10 following infection or vaccination. On the longer term, the persistence of immunity to SARS-CoV-2 infections may need to be periodically assessed to ensure public health and prevent or monitor the resurgence of the virus.
Antibody detection is typically performed using serology immunoassays (IAs), with automated chemiluminescent IA (CLIA) and ELISA, and rapid lateral flow IA (LFIA) being the most prominent. In-house or commercially available diagnostic tests have been rapidly developed for SARS-CoV-2 antibodies (Table S1†).11–15 These IAs typically detect the immunoglobulin M (IgM) and immunoglobulin G (IgG) produced in response to SARS-CoV-2 infection or vaccination.16 While a highly valuable tool in the context of a viral epidemic, antibody tests, similarly to other tests, also have limitations such as false positives or false negatives related to technological or biological origin (for example, too early following an infection) that one should be aware of17 and the factors defining their performance.18,19
Central to the development of these tests is access to viral antigenic proteins. The SARS-CoV-2 recombinant proteins necessary for the development of the IAs assays are produced in prokaryotic or eukaryotic cell lines that are genetically modified to encode the viral proteins. The choice of production strain has an impact on capacity of the cognate human antibodies to bind to the recombinant proteins, depending on folding and glycosylation in different cell lines. In most cases, SARS-CoV-2 serology tests involve the detection of IgGs antibodies against nucleocapsid protein and the spike protein ectodomain, or its receptor binding domain (RBD). Persistence of the IgG antibodies in sera of COVID-positive individuals is more prolonged than IgM antibodies, such that IgG detection should be prioritized for immunity detection,20,21 although additional data may be needed to draw definitive conclusions. While antibodies against nucleocapsid and spike proteins are expected in sera of individuals who have been infected with SARS-CoV-2, only antibodies against spike are expected for individuals in whom immunity has been acquired by vaccination only. As such, developing tests for antibodies against nucleocapsid and spike protein are needed to gain knowledge on immunity of populations and whether this immunity was acquired following an infection or through vaccination.
Whereas ELISA offers high-throughput capacity, it requires several hour-long steps that lengthen overall assay time. Alternatively, faster and portable sensing technologies can decrease assay time and be employed at the point-of-care for infectious diseases.22 Lateral-flow assays have often been proposed to address this for IgM and IgG antibody detection and numerous are now commercially available for SARS-CoV-2 (Table S2†), but they can suffer from reliability issues and they are at best semi-quantitative. As such, a series of sensors will be needed to provide reliable and quantitative data on antibodies found in clinical samples of cohorts of individuals, to enable public health authorities to assess the evolution of the pandemic as well as vaccination efficacy.
Various platforms have been proposed for detection of SARS-CoV-2-related genetic material or viral load and for protein/antibody sensing,23–25,26 including nanophotonics,27 magnetic28 and electrochemical29 sensors. Among them, surface plasmon resonance (SPR) sensing is a label-free sensing technique30 that is particularly sensitive for large biomolecules such as antibodies. SPR sensing has been reported for the detection of antibodies to the first SARS-CoV,31 albeit in phosphate-buffered saline solution (PBS). Since then, SPR sensors have been reported to work in crude biofluids,32 illustrating their applicability potential for the direct detection of antibodies in clinical samples.33 Furthermore, portable SPR platforms have been reported and field-deployed.34
SPR sensing is thus well suited for quantitative analysis of SARS-CoV-2 protein–protein35,36 or protein–antibody37,38 interactions, detecting SARS-CoV-2 RNA39,40 or sensing antibodies associated to SARS-CoV-2.41,42 A recent study showed that the use of SPR imaging can be used to detect antibody isotypes in sera of clinical patients,43 paving the way for the use of SPR in clinical investigations. That initial study focused on the detection of antibodies only for the spike protein on an instrument confined to a centralized laboratory. A complete investigation of the clinical applicability of portable SPR for antibody detection for the nucleocapsid and spike proteins in sera of individuals and correlation to ELISA is necessary. In addition, the use of SPR has seldom been demonstrated with dried blood spots (DBS),44 and its use in the current context could be a game-changer due to the ease of collecting and shipping DBS. Here, we report the cross-validation of ELISA and SPR sensors for anti-nucleocapsid, anti-RBD, and anti-spike. We compare different sources of SARS-CoV-2 antigenic proteins and demonstrate their application to the detection of antibodies in COVID-19 positive individuals with control groups. We further demonstrate effective antibody detection in human plasma and DBS in addition to human serum (Scheme 1).
Alternatively, RBD was produced by Lemay and coworkers in Pichia pastoris SuperMan5. RBD (MN908947.3) was codon-optimized for this host and was C-terminally fused to the TEV cleavage site and to a hexa-His tag (see ESI† for detailed protocols). Briefly, the recombinant RBD protein was expressed in P. pastoris SuperMan5, harvested from the culture supernatant and purified using IMAC (Fig. S1†). The protein was aliquoted at a concentration of 0.16–0.17 mg mL−1 in storage buffer (PBS pH 7.4 with 10% glycerol) and stored at −80 °C.
For ELISA assays, heat inactivated sera were diluted 1:50 for use, unless otherwise specified. For SPR validation assays, no heat treatment was applied to the sera, which were diluted 1:5 with the running buffer before analysis for the detection of IgGs. The fluidic cell with 4 independent channels was used to collect this data and the SPR instrument was placed in a laminar flow cabinet in a biosafety Level 2 (BSL2) laboratory.
A panel of 20 contrived dried blood spots (DBS)52 and matching plasma were collected from different donors at Mount Sinai Hospital (Toronto, Canada) and at the National Microbiology Laboratory (Winnipeg, Canada). Ethical approval was obtained from the Health Canada and Public Health Agency of Canada Research Ethics Board (REB 2020-022P). Plasma samples were diluted 1:5 in running buffer for SPR analysis and 1:20 for the in-house ELISA tests. For the DBS, four 6 mm disks were punched and resuspended in 300 μL of Dulbecco's PBS (DPBS) supplemented with 0.5% BSA and 0.05% Tween20 overnight at 4 °C with agitation (400 rpm). The DBS samples were then diluted 1:5 in the same PBS buffer supplemented with BSA and Tween20. Of those 20 DBS samples, two positive and two negative samples, as determined by a commercial SARS-CoV-2 IgG ELISA for anti-spike (EuroImmun), served to optimize the dilution conditions with the anti-spike SPR assay (1:2.5 and 1:5 were tested), where 1:5 was found optimal. The remaining 16 samples were tested on the EuroImmun SARS-CoV-2 ELISA, the in-house ELISA and with SPR for anti-spike and anti-nucleocapsid. All clinical samples were diluted in the running buffer used for the experiments.
Antigen-down colorimetric ELISA assays were developed according to standard methods adapted for these antigens.47,49,53 Commercial sources of hexa-histidine-tagged rN antigen expressed in E. coli (MyBiosource, cat. no. MBS569934; SinoBiological, cat. no. 40588-V08B) served for initial method set-up but were rapidly substituted by the locally-sourced hexa-His-tagged rN antigen expressed in E. coli (Lemay and coworkers). Comparison of the two sources of hexa-His-tagged rN in an ELISA assay using commercial murine anti-rN diluted in human serum as a primary antibody and a murine HRP-conjugated secondary antibody gave indistinguishable results (Fig. S3A†), confirming the validity of the locally-sourced hexa-histidine-tagged rN antigen. Substitution with a different lot of the same murine anti-rN antibody yielded indistinguishable results, confirming robustness of the method (Fig. S3B†). The sensitivity of the ELISA assay was determined under the same conditions by performing a serial dilution of the murine anti-rN antibody; the limit of detection was calculated at 0.016 μg mL−1 using three standard deviations above the mean of blank measurements (Fig. S3C†). Hexa-histidine-tagged RBD expressed in HEK cells (Kamen and coworkers) was successful in ELISA assays. However, the ELISA signal was significantly lower than that obtained for rN, with a maximum absorbance (OD450) near 0.4 (Fig. S3D†). Expression of hexa-histidine-tagged RBD was also accomplished by Lemay in Pichia pastoris SuperMan5 strain in a bid to obtain faster production. Despite the modified glycosylation of that strain, it appears that RBD was not sufficiently humanized to afford reactivity in ELISA assays (not shown). Hexa-histidine-tagged Spike ectodomain expressed in CHO cells by Durocher was also successful in ELISA assays (Fig. 1). The sensitivity of the ELISA assay for the spike ectodomain was determined by performing a serial dilution of the rabbit anti-S1/RBD antibody; the limit of detection was calculated at 0.0125 μg mL−1 (Fig. S3E†), in the same range as that obtained for detection of murine anti-rN antibody (Fig. S3B†). However, it is important to note that the concentrations reported here are for a surrogate model using antibodies elicited in animals and detection limits or concentrations should be interpreted as indicative of a relative performance only.
One of the challenges for the design of antibody tests for human clinical samples is finding an appropriate source of antibodies to calibrate and optimize the sensor construction as human antibodies are not available and clinical samples difficult to obtain especially early on in a pandemic. We relied on the use of different antibodies elicited in animals as a surrogate (Table S4†). Only one source of surrogate murine antibody gave a nearly null signal, while others gave responses between 241 and 983 RU for 10 μg mL−1 anti-rN. This signal is significantly greater than the noise level of the instrument, which is on the order of 1–2 RU. The surrogate murine antibody with the highest SPR response (monoclonal, MyBiosource, MBS569903) was used for the remaining optimization experiments unless otherwise noted. Calibration was then performed in running buffer and in undiluted human serum to compare the performance of the SPR sensor in each condition (Fig. 2). While the SPR sensor performed similarly in both conditions, it was observed that the SPR signal was larger for lower antibody concentrations in running buffer than serum, but larger SPR signals were obtained at higher murine anti-rN concentrations in serum.
The increase in SPR signal for high antibody concentration in serum was also observed with rS (Fig. 4). While the phenomenon is under further investigation, some preliminary observations can be highlighted. The nonspecific adsorption or bulk refractive index effect of serum can be ruled out. All serum measurements were conducted on a surface passivated with a blank serum (serum containing no murine anti-rN antibody) prior to analysis, minimizing nonspecific adsorption. Finally, blank serum was injected into the reference channel of the SPR instrument at the same time as samples and the remaining background was subtracted from the measurement channels (Fig. S4†), confirming the enhancement of sensitivity in serum relative to PBS. We hypothesize that the enhancement results from adsorption of serum proteins on the captured antibodies, previously reported for abundant proteins such as albumin,54 increasing their mass and refractive index shift. We note that any remaining nonspecific adsorption on the SPR surface, albeit minimized, may help stabilize the surface-bound rN protein and improve binding of its cognate antibodies.
The matrix effect (refractive index, protein concentration, pH, etc.) can be significant in clinical sera and impede on the ability of SPR to perform direct detection of antibodies.55 As such, we implemented a secondary detection step with goat anti-mouse IgG (H + L), which is performed in running buffer following the detection of murine anti-rN in human serum (Fig. 2). In addition to being insensitive to the bulk RI variations of clinical sera, the secondary detection step improved the response by a factor of 2–3 times at lower concentrations and by about 50% at higher murine anti-rN concentrations. The limit of detection was calculated at 3 nM for the direct assay and at 2 nM for the secondary assay. The smaller gain by the secondary detection step at higher concentrations is likely due to steric effects. This is supported from the experiment performed with a secondary antibody with HRP (used for the ELISA test), which led to lower sensitivity (Fig. S5†). Henceforth, anti-mouse IgGs without HRP will be used for the secondary detection step in SPR sensing. The matching ELISA, of course, worked well with the HRP anti-IgG in serum diluted 1:5 and 1:10 (Fig. 1), and showed higher response than SPR with a dynamic range in the ng mL−1 compared to low μg mL−1 for SPR (Fig. 2). As a greater serum dilution is necessary with ELISA to avoid high background response, a different dilution factor was used for ELISA and SPR with clinical samples.
Finally, we also tested the use of humanized anti-rN as a surrogate antibody. However, we observed no signal in ELISA. The SPR calibration curve showed strong interaction of the humanized anti-rN (Fig. S6†), indicating that the lack of ELISA response is likely due to poor interaction of the secondary antibody with the humanized anti-rN. In summary, several sources of rN and murine anti-rN were evaluated to establish the ELISA and SPR tests for anti-rN antibodies. While both commercial and locally produced rN worked in SPR and in ELISA, we observed a stronger SPR signal for the locally produced rN. Murine anti-rN were a better surrogate to optimize the ELISA and SPR assays than a humanized anti-rN. As such, detection of murine anti-rN was achieved in diluted serum for ELISA with a dynamic range in the μg mL−1 range for both the direct and secondary detection assays with SPR in running buffer and in undiluted serum.
We then focused on the use of the RBD domain for the construction of the SPR sensors and ELISA tests for a rabbit anti-rS monoclonal antibody (Sino Biological, cat. no. 40150-R007). RBD expressed in HEK293SF cells and in Pichia pastoris SuperMan5 were compared. While the former led to excellent performance in SPR sensing, RBD expressed in Pichia pastoris SuperMan5 unfortunately did not lead to measurable signal in SPR or in ELISA. Immobilization of RBD expressed in HEK293SF led to SPR shifts of 1592 ± 222 RU for a concentration of 20 μg mL−1 in acetate pH 5.5 buffer. The SPR sensors provided promising outcome with responses of 1036 ± 96 RU and 1859 ± 96 RU respectively for the direct detection and secondary assay of rabbit anti-rS at a dilution of 1:125 (no concentration was provided by the commercial supplier, SinoBiological). Calibration for the rabbit anti-rS antibodies was then carried in running buffer, 10% serum and undiluted serum, with very similar performance for the direct and secondary detection in SPR (Fig. 3). In all cases, the SPR response was essentially within the error for the running buffer, diluted and undiluted serum. The dynamic range of the SPR sensor was in the 102–103 dilution range (Fig. 3), while the one for ELISA was in the 103–105 range (Fig. 1), providing assays for antibodies targeting the RBD domain of spike. Unfortunately, as the concentration of the antibody is unknown, calculation of a limit of detection was inappropriate.
Calibrations were then performed with the SPR assays for the detection of anti-rS, using the same surrogate rabbit antibody as for the RBD assay. The direct detection of rabbit anti-rS worked well in running buffer and in serum, and the detection in serum led to larger response (Fig. 4). This is might be due to the stabilization of rS in such a high concentration of human albumin, as the enhanced sensitivity in serum was not observed for this antibody with RBD. Secondary detection led to a significant improvement of the SPR response, especially for the detection in running buffer where the improvement was several folds. Comparatively, ELISA performed well in 1:5 and 1:10 serum (Fig. 1), with a 10× greater sensitivity than SPR (104 dilution for ELISA vs. 103 for SPR). Hence, the performance of the SPR and ELISA assays were demonstrated for murine anti-rN and rabbit anti-rS (both the ectodomain and the RBD domain) using these surrogate antibodies in human serum.
First, optimal conditions for secondary anti-human IgG antibodies were selected to increase the SPR response and to decrease the impact of the sample matrix, sera were first diluted 1:10. Using serum from individual 4901, we observed that using a goat polyclonal antibody to human IgG coupled to HRP (from Abcam) led to the highest amplification of the response between direct and secondary detection (Fig. S10†), but the AffiniPure goat anti-human IgG (H + L) was employed thereafter as the SPR response was nearly equivalent and a larger supply could be secured. This also shows that direct detection of human antibodies against spike could be detected using 1:10 serum dilution, with a signal between 200 and 500 RU, significantly larger than the background signal (mean = 13 RU, range from −40 to 110 RU). Secondary signal detection could exceed 1000 RU, providing the increased SPR response that is expected to be necessary for lower antibody concentration in individuals with a weaker immune response. Different serum dilutions were then tested to find the optimal dilution factor. It was found that the signal was relatively linear in direct detection with dilutions between 1:40 and 1:2.5, while the signal started to saturate at a dilution of 1:5 with the secondary detection (Fig. S11†). As such, a dilution of 1:5 was then used for the following experiments as it provides the largest signal in the secondary assay while using smaller volumes of the clinical samples. Error bars were also smaller with the secondary detection, which is anticipated as the background effect of refractive index and nonspecific adsorption from serum is minimal with secondary detection in a buffer. Thus, secondary detection will be used for screening purposes.
We further extended the validation of SPR sensing in serum samples with the measurement of human anti-rN, human anti-rS and human anti-RBD for a larger number of COVID-positive and negative individuals. For these experiments, human sera were collected 4 weeks post-infection from five PCR-positive adults (males and females). These samples were first analyzed with ELISA to confirm the presence of human IgG antibodies targeting rN and the ectodomain of rS (Table S7†). Four of the five positive samples showed a significantly stronger ELISA OD using rN antigen (range between 0.9 to 1.4 OD) for the PCR-positive individuals than the negative controls (range of 0.13 to 0.28 OD), and all 5 positive samples were correctly assigned in ELISA using rS antigen. As shown in Table 1, SPR sensing using rN, RBD or rS antigen has the sufficient sensitivity to detect human SARS-CoV-2 antibodies in all 5 positive clinical samples (diluted 1:5) targeting different proteins. Results from these positive sera were compared to 5 control sera from negative individuals (never tested positive for SARS-CoV-2), to ensure the response was specific. In some cases, the sera from the negative controls resulted in small negative shifts due to small refractive index mismatches. In all cases, the detection of human anti-rN, human anti-rS and human anti-RBD led to SPR responses at least one order of magnitude greater than those of the control samples, demonstrating the suitability of SPR sensing for human antibody detection in clinical samples.
Sample ID | Anti-rN (RU) | Anti-RBD (RU) | Anti-rS (RU) | |
---|---|---|---|---|
Positives | 4907 | 1106 | 648 | 1370 |
4911 | 368 | 555 | 856 | |
5905 | 299 | 438 | 765 | |
6902 | 339 | 643 | 1604 | |
7001 | 421 | 1047 | 1189 | |
Negatives | C002 | 1 | 36 | 52 |
C005 | 10 | 96 | 110 | |
C007 | 20 | 16 | −6 | |
C008 | −3 | −27 | −4 | |
C009 | −29 | −35 | −40 |
To confirm the validity of the SPR methods with a larger cohort of serum samples, the analysis of 32 SARS-CoV-2 positive and 8 SARS-CoV-2 negative control serum samples was undertaken with a point-mutated variant of the B.1.1.7 variant of concern (with the native D614 residue). ELISA response of the positive samples (OD450: 1.6 ± 0.5, n = 32) was significantly higher (p < 0.0001) than that of the controls (OD450: 0.14 ± 0.03, n = 8). Similarly, the SPR response of the positive samples (1.2 ± 0.5 kRU, n = 32) was significantly higher (p < 0.0001) than that of the controls (0.3 ± 0.2 kRU, n = 8), demonstrating the ability of SPR to reliably detect the seroconversion of SARS-CoV-2 individuals (Fig. 5).
Fig. 5 Validation of the SPR and ELISA methods with a point-mutated B.1.1.7 spike protein. The responses are expressed in OD450 for the ELISA and in kRU (1000 RU) for the SPR data respectively. |
The analysis of DBS was first optimized for human antibodies against rS, which is assessed more routinely than human antibodies against rN. The DBS samples were assayed at different dilution factors in the same elution buffer. Direct detection of human antibodies in DBS did not lead to conclusive results as the response from the positive and negative DBS samples was nearly equivalent, at approximately 700 RU. This is surely related to matrix effect from a mismatch of the refractive index of the DBS and running buffer, and from some nonspecific adsorption of the eluted biomaterial contained in DBS. DBS contains various proteins and cellular debris increasing the refractive index to a value above the ones of the running buffer. Dilution to 1:2.5 led to significant nonspecific response of the secondary detection step for negative controls, whereas a 1:5 dilution of the DBS gave essentially no response from the secondary antibody with the same negative DBS samples (RU slightly around or below 0). Positive DBS samples led to a response of the secondary detection step on the order of a few hundred RUs, clearly demonstrating the potential of detecting the human anti-spike antibodies in DBS with SPR (Tables S8 and S9,†Fig. 6). The shift values were calculated when equilibrium was reached for the secondary detection step. The slightly negative response of some of the controls was due to washing off nonspecifically adsorbed material from DBS samples. Larger SPR responses were obtained with secondary antibody at a concentration of 40 μg mL−1 (>300 RU for samples #1 and #3, both positive). The results from this dilution analysis were directly applied to human nucleocapsid antibodies, as the nonspecific adsorption will be identical for the same samples.
The optimized SPR assays for human anti-rN and anti-rS in DBS with samples 1 to 4 reported above were then compared to ELISA on a panel of 16 samples (labeled #5 to #20 in Tables S8 and S9†), containing an equal number of SARS-CoV-2 positive and negative samples as assessed with the commercial EuroImmun ELISA platform run by the NML in Winnipeg. Applying a threshold value (RU or OD450 larger than the highest of the controls) of 100 RU and 0.2 OD for positive detection of human anti-rS in DBS, all negative and all positive samples were correctly assigned with SPR sensing and our in-house colorimetric ELISA (Table S8†). Thresholds of 75 RU and 0.2 OD applied for the detection of human anti-rN in DBS led to the correct assignation of all samples with SPR and with the in-house ELISA (Table S9†). It must be noted that the response for human anti-nucleocapsid was rather low in SPR and ELISA, such that the difference between the negative and the positive values was small (Table S9† and Fig. 6) and misassignment can be expected in larger sampling campaigns even though the sensitivity and the specificity was 100% in this data set. This is not really limiting, as human anti-spike antibody detection is more common in clinical monitoring.
Plasma being a more concentrated biofluid explains the relatively higher SPR response compared to DBS. In these tests, the SPR assay was performed with four 6 mm DBS punches, where the total capillary blood volume is estimated at about 65 μL. As the DBS samples were eluted in 300 μL, the DBS samples were already diluted approximately 1:5. Factoring the further dilution prior to analysis, the effective dilution of the DBS samples is about 1:25. Plasma was only diluted 1:5 or 1:10 in comparison, explaining the larger response in plasma.
The results are robust as the data presented here was collected on two instrumental platforms (SPR and ELISA), three different instruments (SPR and 2 different ELISAs), in two labs across Canada (Université de Montreal and PHAC in Winnipeg) and run by at least 4 independent users. The average response from the positive and negative samples were statistically significant with p-values of <0.01 in all cases. The response was also cross-correlated between the SPR, in-house ELISA and the commercial EuroImmun SARS-CoV-2 platforms and led to excellent collinearity of the different methods (Fig. 7). Pearson's coefficient exceeded 0.85 and were as high as 0.95 for the different collinearities measured for every cross-correlation with the various platforms, sample type (DBS and plasma), and antibodies (human anti-spike and human anti-nucleocapsid). This implies that the magnitude of the response for all three platforms was relatively proportional and quantitative.
These results demonstrate that SPR sensing on a portable platform performs equally well to ELISA for the detection of prior SARS-CoV-2 infections, with a detection time under 30 minutes. We therefore envision that the platform could be deployed to different locations as a consequence of instrument portability, in addition to use in centralized laboratories. Preliminary results suggest that the nucleocapsid chips can be conserved for at least one week in the freezer with excellent retention of activity towards anti-nucleocapsid (Fig. S14†), further providing evidence of the field-deployability of the SPR sensor. Detection can be carried in various blood-based products, including serum, plasma and DBS, providing versatility in the applications of the SPR sensor for SARS-CoV-2 antibody detection. The analysis of DBS is especially interesting as they can be easily collected by individuals at home, stored at room temperature (especially interesting for remote locations where a cold chain may not be accessible) and/or shipped to central laboratories using regular post. As such, the use of DBS should facilitate larger cohorts to be screened and access to individuals residing far from urban centers, providing a better picture of the epidemiologic situation in populations worldwide. The current study also constitutes one of the largest multi-center studies about the use of SPR sensors with clinical samples, further advancing the use of SPR towards clinical applications.
Footnotes |
† Electronic supplementary information (ESI) available: Detailed experimental protocols for the expression of antigens and for the optimization of the SPR protocol. Supplementary tables on the different methods used for SARS-CoV2 antibody monitoring, for the optimization of the SPR sensors and of the SPR shifts of human blood-based products. Figures for the purification of the antigens, for the optimization of the ELISA tests and SPR sensors with antibodies elicited in animal models, for the optimization of the SPR sensor with human antibodies in serum samples and for the stability of the SPR sensor stored up to a week (PDF). See DOI: 10.1039/d1an00893e |
‡ These authors that contributed equally to the work. |
This journal is © The Royal Society of Chemistry 2021 |