Serological testing complements molecular diagnosis by documenting the humoral response to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1][2]. IgG generally becomes more reliably detectable after the early phase of infection, and its magnitude and persistence vary with disease severity, sampling time, host factors, and the antigen used [3]-[5]. Serology is therefore most informative for retrospective exposure assessment and population surveillance rather than for diagnosing acute infection [2][6].

SARS-CoV-2 is an enveloped, positive-sense single-stranded RNA betacoronavirus. Its virion contains four principal structural proteins: spike (S), membrane (M), envelope (E), and nucleocapsid (N). N associates with the viral RNA, whereas M and E contribute to virion assembly and envelope organisation. The trimeric S glycoprotein projects from the viral surface and is divided functionally into S1 and S2. S1 contains the N-terminal domain (NTD) and receptor-binding domain (RBD), including the site that engages angiotensin-converting enzyme 2; S2 contains the more conserved membrane-fusion machinery (Figure 1) [7][8].

Figure 1. Structural organisation of SARS-CoV-2 and location of the antigens assessed by CoViDiag^®. (A) Enveloped virion showing S, M, E, N, and the RNA genome. (B) Functional organisation of spike into S1, including NTD and RBD, and the fusion-mediating S2 subunit. Original synthesis based on recent structural reviews and intact-virion imaging [7][8].

Most assays target the nucleocapsid (N) protein or regions of the spike glycoprotein. The receptor-binding domain (RBD) lies within S1 and contains major neutralising epitopes, whereas N is abundant and immunogenic but is not itself a neutralisation target. Multiplex platforms can assess multiple antigens in a single well, potentially increasing information content and revealing antibody signatures that single-antigen assays cannot capture [9][10]. However, analytical thresholds, antigen composition, and the interval since infection strongly influence performance [8]-[11].

Interpretation is particularly important in sub-Saharan Africa. Early in the pandemic, Mbow and colleagues proposed that demographic structure, previous microbial exposures, immune background, and public-health responses might contribute to the discrepancy between projected and reported COVID-19 burden in Africa; these were hypotheses requiring direct epidemiological and immunological evaluation rather than evidence of protection [12]. In Senegal, a national survey conducted in October-November 2020 estimated anti-SARS-CoV-2 seroprevalence at 28.4%, showing that reported cases substantially underestimated exposure [13]. Studies also identified pre-pandemic reactivity to SARS-CoV-2 antigens in African samples [11][14]. Such signals may reflect cross-reactive antibodies induced by endemic pathogens or seasonal coronaviruses, assay-specific effects, or non-neutralising responses and should not be equated with protection [15]-[18]. We therefore compared CoViDiag^® IgG reactivity to N, S1, S2, RBD, and NTD in pre-pandemic controls, RT-PCR-confirmed cases, and occupationally exposed healthcare workers in Senegal [18]-[21].

In this study, we compared multiplex IgG reactivity against five SARS-CoV-2 antigens in three groups of participants with different epidemiological profiles in Senegal. CoViDiag^® positivity was significantly higher among participants with

Table 4. Antigen-specific IgG positivity by study group.

N, nucleocapsid; S1/S2, spike subunits 1/2; RBD, receptor-binding domain; NTD, N-terminal domain. Antigen categories are not mutually exclusive.

RT-PCR-confirmed infection (42.0%) than among pre-pandemic controls (2.5%) and occupationally exposed healthcare workers (5.4%). The antibody response was heterogeneous and strongly dominated by N: among RT-PCR-confirmed participants, 35.8% showed reactivity to N, compared with 21.6% to S2, 16.7% to S1, 12.3% to RBD, and 0% to NTD. These results indicate that the multiplex panel provides information not only through an overall positive or negative classification but also through antigen-specific IgG reactivity profiles. The observed positive agreement with RT-PCR was 42.0%; however, because the interval between infection and blood collection was unknown, this agreement should not be interpreted as an estimate of the assay’s intrinsic sensitivity. In addition, N-directed reactivity observed in pre-pandemic samples should be interpreted cautiously, as it may reflect cross-reactivity or nonspecific binding [24].

The five antigens are not interchangeable diagnostic markers. Each contributes differently to the serological profile [8][9]. N is highly abundant and strongly immunogenic and, in this study, was the most frequently reactive individual antigen. However, N is an internal viral protein, and anti-N binding antibodies are generally not expected to neutralise intact virions [25]. In addition, N contains conserved epitopes that may promote cross-reactivity with other coronaviruses [15][16][24], and anti-N antibodies may decline over time [2][6]. Thus, an isolated anti-N result should be considered evidence of antibody recognition of N, of epidemiological interest and potentially complementary to clinical data, but not definitive evidence of SARS-CoV-2 infection [2][16][25]. S1 contains both RBD and NTD and is generally more immunogenic than the more conserved regions of the spike protein. RBD is biologically important because it mediates ACE2 binding and contains major neutralising epitopes. However, CoViDiag^® measures IgG binding only and does not measure neutralising activity; therefore, isolated RBD positivity cannot establish the presence of protective immunity. S1 and RBD are therefore mainly useful as relatively specific components of a multi-antigen signature. NTD may provide complementary information on the breadth of the response directed against spike, but the absence of reactivity across the different groups indicates that, under the conditions of this study, it was neither sensitive nor discriminative. This observation could be explained by the antigen design, the assay threshold, the viral strain, the timing of sampling, or simply by a weak or waning anti-NTD response. It should not be interpreted as evidence that anti-NTD antibodies are absent in the Senegalese population [24]. S2 contains the most conserved fusion machinery of the spike protein. Its reactivity rate of 21.6% among RT-PCR-confirmed participants is potentially of interest as a complementary marker of positivity, but the risk of cross-reactivity between coronaviruses is high. Isolated S2 reactivity should therefore be interpreted more cautiously than a concordant response directed against S1 and RBD. N and S2 may contribute to detecting broader or waning responses, whereas S1 and RBD may contribute more to specificity. The value of multiplexing lies precisely in combining these signals rather than selecting the antigen with the highest positivity rate [9][10].

Based on the available data, no antigen in this panel can be considered a diagnostic reference standard. The independent reference for infection was RT-PCR confirmation, whereas N, S1, S2, RBD, and NTD were components of the index test. Each antigen should therefore be evaluated against the independent reference standard in a direct comparison at standardized time points after infection, using a prespecified threshold, receiver operating characteristic analysis, and an independent validation set [26].

In this study, RBD or S1 could constitute the most appropriate antigenic anchor for supporting recognition of a SARS-CoV-2-specific response, but not on its own. The strongest evidence of previous natural infection is a concordant profile combining several complementary viral targets, particularly RBD or S1 together with N. Serological profiles limited to N or S2 are more difficult to interpret because they may reflect cross-reactivity, nonspecific binding, or different antibody kinetics. This approach is broadly consistent with the manufacturer’s hierarchical algorithm, which assigns direct positivity to S1, RBD, or NTD and requires multiple or combined signals for N and S2 [22][23]. However, the algorithm cannot be considered independent of the classification rule because it is precisely the rule being applied.

The 42.0% positive agreement with RT-PCR was lower than the performance reported for CoViDiag^® in studies conducted during the late convalescent phase [9][10][22]. It should not be described as intrinsic diagnostic sensitivity because the interval between infection and blood collection was unavailable and the study was not designed as a prospective diagnostic-accuracy evaluation. Early sampling before IgG seroconversion, mild or asymptomatic infection, waning antibody levels, host-related variation, and specimen- or reader-related factors may have reduced detection. Likewise, the 97.5% negative agreement observed in pre-pandemic samples is more descriptive than definitive, as the controls were selected retrospectively and reactive specimens were not assessed with a second serological method [11][27].

The pre-pandemic samples provide a local estimate of background seroreactivity that can be used to assess the panel in Senegal. All four reactive pre-pandemic samples recognised N; only one also recognised S1, and none showed reactivity to RBD, S2, or NTD. Because these samples were collected before SARS-CoV-2 circulation was well established, their signals cannot be attributed to infection with the pandemic virus. The predominance of N-directed reactivity suggests the presence of antibodies against conserved coronavirus epitopes, antibodies induced by other endemic infectious agents, polyreactive antibodies, or nonspecific binding to the assay.

African studies of pre-pandemic samples have also reported appreciable and heterogeneous responses to N and, in some cases, to spike antigens [15]-[21]. Malaria-associated binding to carbohydrate epitopes on spike further demonstrates that apparent SARS-CoV-2 reactivity may occur without previous infection with this virus [19]. These controls therefore have a dual value: they challenge the specificity of isolated antigen signals, particularly those directed against N, and show that thresholds developed exclusively in other populations may not be directly applicable in West Africa. The four reactive controls do not constitute evidence of previous SARS-CoV-2 circulation, prior protection, or reduced susceptibility to COVID-19. No neutralisation assay, antibody-avidity assessment, epitope mapping, seasonal-coronavirus serology, or absorption experiment was performed to establish the origin or function of these antibodies. These findings are therefore interpreted as pre-pandemic cross-reactivity or nonspecific binding, requiring further mechanistic investigation [15][21].

Only 3 of the 56 healthcare workers (5.4%) met the overall positivity rule. At the antigen level, three participants showed reactivity to N, and two showed reactivity to S1, S2, and RBD, indicating that some positive samples had multi-antigen profiles. Concordant reactivity directed against spike and N is more compatible with previous natural infection than an isolated signal directed against N or S2. However, occupational exposure was defined by the work context and not by documented infection, and the low positivity rate does not constitute a measure of the occupational attack rate or evidence of protection. Concordant reactivity directed against spike and N may be more suggestive of previous natural infection than an isolated N or S2 signal, although this pattern is not definitive and should be interpreted in clinical context [2][16][28].

Serology cannot determine where or when infection occurred, and the absence of IgG does not exclude previous exposure, particularly if sampling occurred before seroconversion or after antibody levels had declined. Individual sampling dates and vaccination status were not available. If the samples were collected before vaccination began, profiles combining spike and N would be compatible with previous natural infection. If collection occurred after vaccination began, spike-only profiles could suggest vaccination, whereas N reactivity would be more compatible with natural infection, although waning anti-N antibodies and cross-reactivity would remain potential confounding factors. The exposed group should therefore not be considered an infected group, but rather a comparison group defined by epidemiological criteria [2][6][16].

The difference between RT-PCR-confirmed participants and pre-pandemic participants is compatible with infection-related seroreactivity, although the number of confirmed cases was limited. At the population level, Senegalese serosurveillance showed that cumulative exposure was more than 10 times the number of reported cases at the time of the study [13]. Mbow and colleagues previously suggested that demographic composition, microbial exposures, immune background, and public-health responses may have contributed to the initially lower-than-expected number of cases in Africa [12]. The results presented here neither confirm nor refute these hypotheses. They only show that background immune reactivity exists and can substantially influence the interpretation of SARS-CoV-2 serology, and that the presence of antibodies should not be equated with neutralising activity or protection against infection.

In sub-Saharan Africa, producing a single composite positive or negative result may conceal biologically and analytically significant profiles. Reporting N, S1, S2, and RBD signals separately, together with a predefined multi-antigen classification rule, could help distinguish strong concordant responses from potentially cross-reactive signals directed against a single antigen. Serology should be used primarily for retrospective exposure assessment and population surveillance; it should not replace molecular or antigen detection for the diagnosis of acute infection [2].

The main strengths of this study are the simultaneous assessment of five antigens and the comparison of a large number of pre-pandemic samples with RT-PCR-confirmed participants. Antigen-specific analysis highlights features that would be difficult to observe with a binary single-antigen assay, particularly the absence of NTD reactivity. Several limitations restrict the interpretation of the findings. The retrospective, convenience-based design and the absence of standardised intervals between infection and sampling make it difficult to estimate sensitivity [2][6][11].

Disease severity, clinical history, vaccination status, comorbidities, and antibody kinetics were not available. The healthcare-worker group was small, and exposure was not biologically confirmed. Reactive pre-pandemic samples and samples from exposed participants were not evaluated using an independent serological method or a neutralisation assay. Available individual-level data were insufficient for a more detailed analysis [2][16][28].

In addition, the AID ELISPOT Reader settings, software version, calibration procedure, and any use of an AID reader comparable to the instrument used during the manufacturer’s evaluation were not sufficiently documented. This uncertainty regarding the reader is important; instrument-comparison and repeatability studies would be required before drawing conclusions about diagnostic performance from numerical values obtained with the AID reader [22][23][29].

Future studies should use prospectively collected samples with documentation of symptom-onset dates, RT-PCR dates, vaccination status, disease severity, and repeated sampling. Antigen-specific thresholds should be reassessed using local pre-pandemic controls and confirmed cases, followed by blinded validation in an independent Senegalese sample. Samples showing discordant reactivity limited to N or S2 should undergo independent serology, testing for antibodies against seasonal coronaviruses, and neutralisation or inhibition assays. These studies would help determine whether a locally optimised multi-antigen rule can improve specificity without compromising sensitivity and, if not, whether these profiles reflect SARS-CoV-2-specific antibody responses, cross-reactivity, or non-functional antibody responses.

The CoViDiag multiplex assay revealed heterogeneous IgG recognition of SARS-CoV-2 antigens in Senegal, with N as the dominant target and lower reactivity to S2, S1, and RBD. Overall positivity was substantially higher in RT-PCR-confirmed participants than in pre-pandemic controls or exposed healthcare workers. Because sampling time was unknown, the 42.0% positive agreement with RT-PCR should not be treated as an intrinsic sensitivity estimate. The predominantly N-directed pre-pandemic signals reinforce the need for local validation, orthogonal confirmation, and cautious separation of antibody binding from protective immunity.

The authors thank the participants, the clinical teams at the epidemic treatment centers, and the laboratory and biobank staff at IRESSEF for their assistance in collecting and processing samples and analysis. We also thank the collaborators and partner organisations for their help in accessing the reagents and laboratory materials that were required for this study.

Conceptualization: Moussa Ndour, Moustapha Mbow, and Tandakha N. Dieye. Investigation: Moussa Ndour, Biscotine F. Makoutchouang Nzonde, and Joséphine K. Badiane. Sample collection: M. Cisse, M. Gningue, I. Diallo and A.A. Diallo. Formal analysis and interpretation: Moussa Ndour and Moustapha Mbow. Resources, materials, and analytical tools: Mady Ndiaye, Massamba Sylla, Tandakha N. Dieye, and Souleymane Mboup. Project administration and coordination: Babacar Mbengue and Tandakha N. Dieye. Writing-original draft: Moussa Ndour and Moustapha Mbow. Writing-review and editing: Babacar Mbengue, Moustapha Mbow, and Tandakha N. Dieye. All authors read and approved the final version of the manuscript.

This research received no external financial funding. CoViDiag^® reagents were provided in kind through a scientific collaboration coordinated by Professor Tandakha N. Dieye and Professor Souleymane Mboup.

The datasets generated and analysed during the current study are available from the corresponding author on reasonable request, subject to institutional and ethical restrictions.