The Coronavirus disease-19 (COVID-19) is a respiratory illness caused by the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), a novel beta-coronavirus firstly described in Wuhan city, China, on December 2019 [1
]. SARS-CoV-2 spreading has been declared pandemic in mid-March 2020 by WHO [2
]. At present, the virus has infected more than 33 million people worldwide with an associated case mortality rate of 1% to 15%, depending on the country [3
COVID-19 is associated with a broad range of mild-to-severe symptoms, potentially leading to hospitalization in Intensive Care Units (ICU) for the most severe cases. The respiratory tract is initially involved with possible development of severe interstitial pneumonia [4
], albeit the gastrointestinal tract can also significantly participate in disease pathogenesis as a consequence of the expression of the ACE2 receptor, that mediates SARS-CoV-2 viral entry [6
], on both alveolar and enteric epithelial cells [7
]. Infected subjects manifest a complex clinical pattern appearing as early as two days post exposure and lasting several weeks [1
Infection with SARS-CoV-2 induces a prompt activation of the immune system, finalized to the clearance of infected cells [8
]. Innate and adaptive immune cells accumulate at the site of infection, where production of cytokines and inflammatory mediators may result in patient recovery or, in case of ineffective viral clearance, in hyperactivation of the immune system and development of severe complications, such as acute respiratory distress syndrome ARDS [4
]. Overexpression of pro-inflammatory cytokines (i.e., IL-1 beta, IL-2, IL-6, IL-17, TNFα etc.) and impairment of humoral immunity have been described in patients with the most severe form of the disease [5
]. Antibodies against SARS-CoV-2 proteins are produced as a consequence of the activation of the humoral arm of the immune system. Virus-specific IgM antibodies are secreted as first class of immunoglobulins, followed by the more specific IgG [10
]. Among the latter, those specific for the viral Spike Receptor Binding Domain (RBD) when expressed at higher titer manifest direct neutralizing activity towards the viral entry into cells, as they prevent effective engagement of surface ACE2 receptors by the Spike protein [11
]. The IgA response against SARS-CoV-2 has been shown to be rapid and persistent [13
] and is associated with mucosal responses, including respiratory and gastrointestinal responses.
Serological testing is a valuable tool to monitor viral spreading throughout the population [15
]. Furthermore, serological assays allow the identification of past infection in individuals with viral RNA levels undetectable by RT-PCR for epidemiological purposes [16
]. Various commercial and in-house assays that utilize distinct viral antigens and detect different antibody classes are currently available. However, SARS-CoV-2 serological tests available on the market do not always allow systematic simultaneous detection of a wide antibody spectrum for several antigens in a reliable manner, and this may hamper a proper population testing for clinical or epidemiological purposes [17
]. Conversely, serological enzyme-linked immunosorbent assays (ELISA) to detect immunoglobulins raised against the viral Spike Soluble Ectodomain (Spike) and its highly immunogenic receptor binding domain (RBD), or against the Nucleocapsid protein (N), provide promising results in terms of accuracy and reproducibility [11
]. Recently, these ELISA assays have been used to show that neutralizing antibodies (nAb) against different viral antigens may decline after 20–30 days post symptoms onset, and that the magnitude of nAb response may be associated with disease severity in COVID-19 patients [18
In order to measure the presence and variation of antibody responses against different viral proteins, we set up and validated an in-house direct ELISA assay based on three distinct SARS-CoV-2 viral antigens, i.e., eukaryotically-expressed RBD and Spike and bacterially-expressed Nucleocapsid protein. Using this assay, we simultaneously measured IgM, IgG and IgA anti-viral antibodies titers in the sera of COVID-19 patients, as well as levels of pro-inflammatory cytokines. In addition, we longitudinally collected the sera of 16 convalescent healthcare workers who tested positive for SARS-CoV-2 by nasopharyngeal (NF) swabs, and were symptomatic but not hospitalized. Our data show that humoral immune responses against SARS-CoV-2 correlated with disease severity in terms of both antibody titers, persistence over time and serum levels of pro-inflammatory cytokines. Notably, 90% of COVID-19 mildly symptomatic patients halved their anti-RBD IgG titers after 4 weeks from viral negativization, thus confirming the short lifespan of humoral immune responses against SARS-CoV-2.
2. Experimental Section
2.1. Human Subjects
Health care workers of two different COVID Hospitals in Milan (n
= 16) with documented COVID-19 infection (by NF swab), not hospitalized but with manifested COVID-19 symptoms (Supplementary Table S2
) were monitored for seroconversion by IgM, IgG and IgA serum levels at two time points after viral clearance between April and June 2020. The study has been conducted in accordance with the Standards of Good Clinical Practice, with the ethical principles deriving from the Helsinki Declaration and the current legislation on observational studies. Clearance from the Ethical Committee has been obtained (IEO1271). Additional study populations were ICU hospitalized severe COVID-19 patients (n
= 24) and (n
= 58) COVID-19 negative subjects whose sera were collected between April and June 2020. In total, 436 pre-COVID subjects enrolled in IEO studies between 2009 and 2015 were used to calculate the ROC curves for the assays.
The exclusion criterion was, for all subjects involved in the study, the inability to provide informed consent. The inclusion criteria were, for those not hospitalized with COVID-19, (i) being health care workers (medical doctors, practitioners, post-graduate students, nurses), potentially exposed to SARS-CoV-2 between February and June 2020, (ii) documented SARS-CoV-2 infection by NF swab, (iii) not being hospitalized for COVID-19; for those hospitalized with COVID-19: (i) documented SARS-CoV-2 infection by NF swab, (ii) being admitted in the ICU between February and June 2020 for COVID-19; for negative controls: (i) sera being collected before 2019.
2.2. Antigen Proteins Production
The recombinant Spike SARS-CoV-2 glycoprotein receptor binding domain (RBD) and the soluble full-length trimeric ectodomain have been produced in mammalian HEK293F cells as glycosylated proteins by transient transfection with pCAGGS vectors generated in Prof. Krammer’s laboratory [11
]. The constructs were synthesized using the genomic sequence of the isolated virus, Wuhan-Hi-1 released in January 2020, and contain codons optimized for expression in mammalian cells.
Briefly, HEK293F cells were seeded at a final concentration of 0.5 million/mL in freestyle medium (Thermo Fisher Scientific, Milano, Italy), incubated at 37 °C, 5% CO2 at 120 rpm O/N in an Eppendorf New Brunswick s41i incubator. The day after HEK293F cells were transfected using 1 µg of DNA per 1 × 106 cells and a DNA: PEI MAX ratio of 1:5 in OptiMEM medium. Four hours post-transfection, the medium was supplemented with Peptone Primatone RL (Merck) to a final concentration of 0.6% w/v. Cells were then incubated for 6 days, checking cell viability daily if needed (a mortality higher than 30% is indicative of a toxic protein).
For protein purification, the culture supernatant was transferred to conical centrifuge tubes, cleared by centrifugation at 1000×·g for 15 min and filtered with 0.22 µm Stericup Filters. The filtered medium addition with 1:10 volume of 570 mM NaH2PO4 pH 8.0, 300 mM NaCl and loaded on a HisPrep Fast Flow 16/60 column (GE-Healthcare) equilibrated in 57 mM NaH2PO4, 300 mM NaCl. His-tagged protein was eluted with step gradients of 50-100-150-235 mM imidazole. Peak fractions were pooled, dialyzed overnight against PBS and concentrated to 0.4 mg/mL (Spike soluble) or 1.0 mg/mL (RBD) in 10 kDa-MWCO Amicon filter units.
Retrieved proteins were quantified, flash frozen in liquid nitrogen in aliquots and stored at −80 °C.
His-tagged SARS-CoV-2 full length N-protein plasmid (kind gift of David D. Ho, MD, Columbia University, New York, NY, USA) was transformed in E.coli BL21 pLysS cells. Protein expression was induced with 0.5 mM IPTG and carried on at 18 °C overnight. Cells were harvested by centrifugation in lysis buffer (25 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 mM DTT, 5% glycerol, 10 mM imidazole, with Calbiochem protease inhibitor Cocktail III). All following steps were carried out at 4 °C or using ice-cold buffers. Cells were lysed by sonication; lysate was cleared by centrifugation at 20,000× g for 40 min, then PEI (pH 7.5—final concentration 0.02%) was dropwise added, under stirring, and lysate was then further cleared by centrifugation at 20,000× g for 40 min. Next, 5 mL Ni-NTA beads per liter of culture, pre-equilibrated in lysis buffer, were added to the cleared lysate and protein binding was continued for 1 h in gentle agitation at 4 °C. Beads were washed with at least 20 column volumes of 25 mM Tris-HCl pH 7.4, 500 mM NaCl, 1 mM DTT, 5% glycerol, 50 mM imidazole and His-tagged protein was eluted with 4 column volumes of 25 mM Tris-HCl pH 7.4, 200 mM NaCl, 1 mM DTT, 5% glycerol, 250 mM imidazole. The eluted fractions containing protein were diluted with heparin buffer (25 mM Tris-HCl pH 7.4, 1 mM DTT) to reach final a NaCl concentration of 0.1 M and were subsequently loaded onto a Hi-Trap Heparin HP column (GE Healthcare) equilibrated in 25 mM Tris-HCl pH, 50 mM NaCl, 1 mM DTT (buffer A). A linear gradient reaching 100% buffer B (25 mM Tris-HCl pH, 1 M NaCl, 1 mM DTT) in 30 column volumes was applied and fractions containing His-tagged N-protein were pooled, concentrated and loaded onto a Superdex 200 16/60 size exclusion chromatography. Fractions containing N-protein were pooled. A 4 L culture yielded 12.5 mL of 1.12 mg/mL pure N-protein, which was flash frozen in liquid nitrogen in aliquots and stored at stored at −80 °C.
The ELISA assay to detect immunoglobulins (Ig) uses fragments of the SARS-CoV-2 Spike glycoprotein (S-protein) and the Nucleocapsid (N) as antigens based on the protocol published in [11
]. After binding of the proteins to a Nunc Maxisorp ELISA plate, and blocking aspecific bindings with PBS-BSA 3%, patients’ sera to be analyzed were applied to the plate to allow antibody binding at a final dilution of 1:200, revealed with secondary anti-human-IgG (BD, clone G18-145), IgM (Merck, Polyclonal code A6907), IgA (Biolegend, Poly24110) antibody conjugated to HRP. Samples are read on a Glomax reader at 450 nm. This ELISA test is not intended for commercial use and is currently under evaluation at the Italy’s Ministry of Health (Aut.Min.Rich. 15.05.2020) for emergency use approval. The assay has been validated with a cohort of n
= 56 COVID-19 subjects (severe, moderate and mild disease) and n
= 463 (subjects collected in pre-COVID era (between 2012 and 2015)). ROC curves have been implemented to determine the sensitivity and specificity of the assay (Supplementary Figure S1
2.4. Multiplexing Analysis of Sera Cytokines
Quantification of soluble biomarkers was performed in sera of patients collected immediately after virus clearance (2 consecutive negative NF swabs) and one month post virus clearance using a Luminex Immunoassay (Human Cytokine/Chemokine/GF ProcartaPlex 45plex, Thermo Fisher) with MAP technology according to manufacturer’s protocol. Samples were acquired on a Luminex 200SD and analyzed with Xponent software 4.2. The sera of healthy subjects (n = 20) collected between April and June 2020 as well as ICU COVID-19 patients (n = 24) were used as control groups.
2.5. Statistical Analysis and Sample Size
The categorical variables were described as absolute frequency and percentage. The continuous variables with normal distribution were described as median ± standard deviation (SD), whereas the continuous variables without normal distribution were given as median and range. Normality of continuous variables was checked with D’Agostino–Pearson omnibus normality test. The Mann–Whitney test or Student’s t-test for continuous variables, and the Chi-square or Fisher’s exact tests for categorical variables, were used to associate clinical variables with the result of SARS-CoV-2 serological test (positive or negative). The p values lower than 0.05, two-tailed, will be considered statistically significant. GraphPad Prism software was used for all statistical analyses.
During the last months many key aspects of the immune response to SARS-CoV-2 have been elucidated. However, given the complexity and diversity of the clinical manifestation of COVID-19 disease, several outstanding questions remain still to be addressed.
Here we show that humoral immune responses against SARS-CoV-2 correlated with disease severity in terms of both antibody titers, persistence over time and serum levels of pro-inflammatory mediators. Moreover, we showed that the vast majority of COVID-19 mildly symptomatic patients analyzed in the study halved their anti-RBD antibody titers after 4 weeks from viral negativization, thus confirming the short lifespan of humoral immune responses against SARS-CoV-2.
Humoral immune response against SARS-CoV-2 proteins leads the production of antibodies against the portions of the viral proteins [10
]. In this sense, serological tests, based on the search of specific anti-SARS-CoV-2 antibodies, represent a useful tool aimed at identifying patients who contracted the infection and, consequently, comparing the clinical course and eventual complications between the general population and population at risk, such as health care workers [15
]. Importantly, measurable variations in the humoral response might account for a re-activation of the immune system as a consequence of viral re-exposure, both in healthcare workers and in the general population. Serological monitoring of antibody levels can thus provide information on the actual circulation of the virus, which can be used by decision makers to adapt safety and restriction measures according to the real presence of the virus within the population.
Nonetheless, the specificity and sensitivity of the different assays greatly vary among kits taking into consideration the different techniques implemented (ELISA, CLIA, lateral flow) and the antigens used (Spike ectodomain, S1-S2 of the Spike, Spike RBD, Nucleocapsid). Thus, only highly sensitive tests can detect with high accuracy whether people, including mildly symptomatic or asymptomatic subjects, have specific anti-SARS-CoV-2 antibodies present in their blood.
The test utilized in this study is a robust ELISA assay imported from the laboratory of Prof. Krammer at Mount Sinai, that has been approved for emergency use by the FDA [11
]. We reproduced its excellent performance in our lab, that allowed us to detect a broad range of antibody levels, spanning form those measured in the blood of severe hospitalized patients and not hospitalized mild COVID-19+ individuals. The ELISA assay has been validated with a cohort of more than 500 positive and negative subjects, giving rise to extremely high performance values. Specificity and sensitivity of the ELISA assays were high for anti-RBD IgG and IgA (92–97%) and slightly lower for IgM and the Spike and N proteins (70–85%). These performances are in line with those published for both in-house and commercial assays [20
]. For this reason, this test is also being currently evaluated by the Italy’s Istituto Superiore di Sanita’ (ISS) for its emergency use approval.
One additional key strength of this assay as compared to other types of serological assays is its flexibility, i.e., the possibility to simultaneously assess different classes of antibodies against a broad panel of SARS-CoV-2 antigens within the same assay. Thus, this ELISA assay gave us a comprehensive understanding of the magnitude and persistence of antibody titer against different viral proteins and their correlation with the strength of the immune response, as measured by the serum levels of pro-inflammatory mediators.
The presence of few false positives among the COVID-negative population tested with the viral nucleocapsid protein as compared to the RBD might be a consequence of a mistakenly detection of anti-N antibodies previously raised against common cold coronaviruses which cross-react with the SARS-CoV-2 nucleocapsid [23
]. The nucleocapsid protein is the more conserved protein among different coronaviruses. It is possible to speculate that antibodies produced against previous common cold coronaviruses (and cross-reacting with the SARS-CoV-2 antigens) might still be present in the sera at high levels, and therefore be detectable. As a consequence, when analyzed longitudinally, we observed that only the antibodies specific to SARS-CoV-2 decline while those aspecific and possibly reacting to previous coronaviruses remain detectable at the same levels over time. A similar observation was recently published by a large longitudinal study [24
]. Moreover, a recent paper evaluated the persistence of anti-N specific antibodies raised against four different common cold coronaviruses in a cohort of HIV+ individuals followed longitudinally for more than 10 years [25
]. The study confirmed that N-specific antibodies undergo fluctuations in their detection levels as a consequence of seasonal re-infections with a kinetic of 6–12 months. Interestingly, the authors reported that 2 out of 10 patients (20% of the individuals enrolled in the study) showed cross-reactive antibodies against the viral N-proteins of the four viruses, and in one of them these cross-reactive antibodies persisted over the years.
The duration of circulating IgG antibodies is still unclear and might depend on several factors, including the type and extent of immune response elicited upon the encounter with the virus [17
]. In this study, non-hospitalized subjects showed lower antibody titers and blood pro-inflammatory cytokine profiles compared to patients in Intensive Care Units (ICU), irrespective of the antibodies tested. This finding is in accordance to what published for asymptomatic [22
] and paucisymptomatic [14
] patients whose antibody titers were detected using commercial ELISA or chemiluminescence assays against either the Spike or the N-protein.
Anti-RBD IgA antibodies manifested a similar kinetic compared to that of IgG. IgA response against SARS-CoV-2 has been reported to be rapid and persistent [18
] and possibly associated with mucosal immune response in the gut and lungs. Notably, IgA production has been associated with disease severity, suggesting that IgA production might occur locally at the mucosal sites, possibly correlating with the viral load, the duration of the viral exposure and the virus entry route [13
]. Consistently, a recent communication [14
] confirmed that the highest levels of IgG and IgA antibodies against the Spike S1 domain, encompassing the N-terminal half of the protein with the RBD, were associated with severe disease [13
Severe hospitalized COVID-19 patients overexpressed pro-inflammatory cytokines (i.e., IL-1 beta, IL-2, IL-6, IL-17, TNFα). In one of the very first reports of the clinical course of COVID-19 patients, as early as March 2020, serum increase in interleukin (IL)-2, IL-7, GMCSF, IP-10, MCP 1, MIP1-α, and TNF-α was associated to disease severity [5
]. Elevated IL-6 levels were detected in hospitalized patients and have been associated with ICU admission, respiratory failure, and poor prognosis in several studies [5
]. Presently, conflicting results regarding IL-1b and IL-4 have been reported [29
]. The elevation of pro-inflammatory cytokines, albeit being widely described in COVID-19 patients, does not seem presently to have prognostic value, because they do not always differentiate moderate cases from severe cases [32
]. Levels of IL-6 at first assessment might predict respiratory failure [33
], other publications with longitudinal analyses demonstrated that IL-6 increases fairly late during the disease’s course, consequently compromising its prognostic value at earlier stages [34
]. Moreover, serum concentrations of KL-6, a molecule elevated in serum of patients with interstitial lung diseases (ILDs), such as idiopathic pulmonary fibrosis and hypersensitivity pneumonitis, was recently proposed to be capable of differentiating between severe and mild COVID-19 patients, being mainly produced by damaged or regenerating alveolar type II pneumocytes [35
Conversely, IP-10, MCP-3, and IL-1ra were capable of differentiating between severe and mild COVID-19 patients [32
]. Interestingly, MIP 1 alpha, IL8 and Eotaxin, similarly to the results published by Long et al. [22
], were expressed to a greater extent by healthy subjects compared to COVID-19 patients. Human MIP 1 alpha and Eotaxin were reported to be potent inhibitors of M-tropic HIV-1 infection, and were therefore considered as potential HIV-1 inhibitors [37
]. A similar protective mechanism of action might be envisaged in SARS-CoV-2 infection.
We also observed that during non-severe COVID-19 infections, pro-inflammatory cytokines are produced and correlate with the severity of the disease. Similarly to anti-SARS-CoV-2 antibodies, pro-inflammatory mediators also decreased within a month after viral clearance, as expected upon the resolution of the disease.
Overall, we suggest that the decline in antibody titer and pro-inflammatory cytokines is a common characteristic of SARS-CoV-2 infection. This study therefore has important implications for the use of serological testing for the monitoring of infection outbreaks against re-infection with SARS-CoV-2. Our results indicate that the detection of antibodies with serological assays for epidemiological and monitoring purposes in non-hospitalized seroconverted COVID-19+ subjects, who most likely represent the majority of people who encountered the virus, is only highly reliable within a limited window of time after viral clearance.