This study was conducted in accordance with current legislation including the Declaration of Helsinki and Resolution No. 466/2012 of the Conselho Nacional de Saúde do Brasil. Ethical approval was obtained from the institutional review board of the Instituto René Rachou, Fundação Oswaldo Cruz, CAAE: 31.919520.8.0000.5091, approval numbers: 4177931; 4291836; 4343318; 4624187; 5294423. Written informed consent was obtained from all participants before any study procedure was undertaken.

The population selected for this study was composed of HCW who worked in at least one of the public hospitals: Hospital das Clínicas (HC) at the Universidade Federal de Minas Gerais, the Unidade de Pronto Atendimento (UPA) Centro-Sul, and the Hospital Metropolitano Dr. Célio de Castro (HMDCC). All three health centers are in the city of Belo Horizonte, the state of Minas Gerais, Brazil. As inclusion criteria, in addition to what has already been mentioned above, all participants had to (i) present with at least one of the following symptoms within the previous seven days: fever (equal to or greater than 37.5°C), cough (dry or productive), fatigue, dyspnea, sore throat, anosmia/hyposmia, and/or ageusia; and (ii) have a positive result for the detection of SARS-CoV-2 RNA by RT-qPCR. Participants were invited by telephone, and those who met the inclusion criteria were included in the study. Individuals who reported volunteering in COVID-19 vaccine clinical trials, prior diagnosis of COVID-19, or reported pregnancy, were excluded. Hospitalization also resulted in loss of follow-up due to inability to perform the tests. The enrollment of participants took place between October 2020 to April 2021. Individuals were followed up for 450 days, with capillary blood samples collected on days 7, 10, 15, 20, 25, 30, 60, 90, 180, 270, and 450 after being enrolled in the study (referred to as D7, D10, D15 and so on). In addition to capillary blood collections, participants were contacted every 30 days after the 60th day (from December 2020 to June 2022) to assess possible SARS-CoV-2 reinfection. Besides, all participants were instructed to contact the research team when they present any symptoms that constitute suspicion of COVID-19. The predetermined symptoms which were monitored as evidence for reinfection were: fever (equal to or greater than 37.5°C), cough (dry or productive), sore throat, fatigue, dyspnea, and diarrhea. Also, any suspicious symptom established by medical criteria was monitored. All suspected cases were tested for the presence of viral RNA by RT-qPCR and serological status. In January 2021, during the study period, public roll-out of the vaccination scheme using the CoronaVac vaccine was started in Brazil. Fifteen days after the second dose, the initial two-dose vaccination scheme is considered “complete”. However, in September 2021, a booster dose was rolled-out (6). The workflow for our study is shown in Figure 1 .

All symptomatic healthcare professionals were tested for SARS-CoV-2 by RT-qPCR using the Charite Institute protocol (26). During the isolation period, kits were sent to participants to perform a self-collection of saliva or nasopharyngeal and oropharyngeal swabs on D1, D3, and D5 after enrollment in the study. Swab or saliva was placed in a tube containing a lysis buffer with guanidine (4 M Guanidine Isothiocyanate, 25 mM Sodium Citrate, pH 7.4) that were stored at 4°C. Samples were collected in participants’ houses within a 5-day interval and were transported under refrigeration to the central laboratory for RNA extraction. As demonstrated by Carvalho et al., 2021 (27) this buffer, besides decreasing the virus infectivity, maintains RNA integrity for up to 16 days at room temperature. In addition to the material for saliva/swab sample collection, the kits included materials, together with an instruction manual, for performing a rapid test to detect IgM and IgG antibodies against SARS-CoV-2 antigens (TR DPP^® COVID-19 IgM/IgG - Bio-Manguinhos). Participants characterized as infected with SARS-CoV-2 must have had at least one positive result for detection of SARS-CoV-2 RNA on D1, D3 and/or D5. Participants characterized as non-infected presented negative results in all three samples analyzed and were not included in the follow-up period of the study.

RNA was extracted from the nasopharyngeal and oropharyngeal swabs or saliva samples using the QIAamp^® Viral RNA kit (QIAGEN^®, USA) according to the manufacturer’s instructions. Briefly, 140 μL of sample was added to 560 μL of AVL buffer containing carrier RNA, and after 10 minutes at room temperature, 560 μL ethanol were added. This solution was applied to an RNA affinity column and this column was centrifuged at 6,000 x g for 1 minute. Then, the column was washed with AW1 and AW2 buffer solutions in that order. After the washing process, RNA was eluted using AVE solution and used in the RT-qPCR test.

The RT-qPCR reactions were performed using the ViiA™ 7 Real-Time PCR System of the communal Real-Time PCR platform at the Instituto René Rachou. The RT-qPCR assays were performed using 5 uL of sample RNA, and the 200 nM GoTaq^® Probe 1-Step RT-qPCR System Kit (Promega). This kit uses GoTaq Probe qPCR Master Mix with dUTP (10 uL), GoScript RT Mix for one-step RT-qPCR (0,4 uL), sense and antisense primers (400 nM) and nuclease-free water to a 20.0 uL final volume. The conditions for the amplification were: 45°C for 15 minutes and 95°C for 2 minutes, followed by 40 cycles of denaturation at 95°C for 15 seconds and hybridization at 60°C for 1 minute. The results were analyzed using the Thermo Cloud platform, according to the following criteria: samples with amplification of the E gene (Ct<37) and the RNAse P gene (RP) (Ct<35) were considered positive; samples without E gene amplification or with detection above Ct 37, with RP amplification (Ct<35), were considered negative. Samples with RP amplification above Ct 35 were considered invalid, and the test was performed again using RNA obtained from another extraction of the samples collected.

In order to detect IgM and IgG antibodies against SARS-CoV-2, the DPP^® COVID-19 IgM/IgG kit supplied by Bio-Manguinhos (FIOCRUZ, Brazil) was used according to the manufacturer’s instructions. Briefly, a digital puncture was performed, and a blood sample was diluted in the buffer provided in the kit. The sample was then applied to the cassette. After 5 minutes, 9 drops of the buffer were added to the cassette, and the results were read after an additional ten-minute period. The interpretation of the test was performed with the aid of the DPP^® Micro Reader, which provides the intensity of the reactive line. The DPP^® COVID-19 IgM/IgG system uses an algorithm that includes assay-specific cutoff values to determine test result. Values equal to or greater than 30 for the IgM and IgG antibodies were considered reactive. In order to assess antibody levels, the test was performed on scheduled days and/or when the participant was suspected of reinfection ( Figure 1 ).

Positive samples, with cycle threshold (Ct) lower than 36, were sequenced by Next-Generation Sequencing (NGS) on the Illumina MiSeq Platform using the Illumina COVIDSeq Kit for library construction (Illumina, San Diego, USA) generating paired-end reads 150 bp long. The raw reads were trimmed using Trimmomatic version 0.39 (28) with a sliding window of 4 nucleotides with a minimum average Phred score of 20. Trimmed reads smaller than 50 bp were removed. The filtered reads were mapped to the SARS-CoV-2 reference genome (NC_045512) using BWA version 0.7.17 (Li 2009) with the default parameters. The nucleotide variants were identified using iVar version 1.3 (29)., with a minimum frequency of 40% and depth of 30 reads. The consensus sequences generated by iVar were submitted to Pangolin version 4.2 (30) to identify the coronavirus lineage. Sequences that met the GISAID criteria were submitted to the EpiCoV database and are deposited at The European Nucleotide Archive under project number PRJEB49204 (https://www.ebi.ac.uk/ena/browser/view/PRJEB49204).

GraphPad Prism v.8.0.1., Jamovi 2.3.18.0 (https://www.jamovi.org/), and the statistical software R (https://www.r-project.org/) version 4.1.2 (31) were used for data analysis and generation of figures. Data organization and pre-processing of some graphs and figures were done using Excel and PowerPoint (Microsoft 365). The chi-square test or the Fisher’s exact test were used to assess the association of categorical demographic variables and the infection status of individuals, as well as the correlation of symptoms reported during the first infection, reinfection and in suspected cases of reinfection. Paired and unpaired t-tests were used to analyze the difference in viral load between infection vs reinfection, and females vs males, respectively.

Generalized Estimating Equations (GEE) (version 1.3.9) were used to evaluate the longitudinal data. Proposed by Lian and Zeger (1986) (32), the proposed model jointly estimates an average effect and intra-individual variations, considering the structure of correlation or dependence between the repeated measures. The outcome of interest was the numeric scale of IgM and IgG levels and the covariates were vaccine status, follow-up time, booster dose and reinfection. In the GEE model, the peak of IgM and IgG levels in the interval from D7 to D30 was considered. A model segmented in time, 180 and 270 for IgM and IgG, respectively, was adjusted to make the average structure more accurate.

In order to assess whether there is an association between antibody levels and reinfection, differences in the levels of IgM and IgG antibodies against SARS-CoV-2 among individuals with a confirmed case of reinfection and those not reinfected were evaluated. This analysis only considered samples taken during the period of the study when reinfection occurred (i.e., between July 2021 and July 2022), which was also the period when the delta and omicron variants circulated in Belo Horizonte. For this evaluation, the Mann-Whitney U-test was used. For individuals who did not become reinfected, the arithmetic means of all of the values of the antibody levels determined during the period stated above were considered. For individuals who became reinfected, we considered only the values of the antibody levels determined when reinfection was confirmed.

A total of 163 health professionals aged between 19 to 68 years were included in the initial study cohort between October 2020 to April 2021 ( Figure 1 ). Fifty-four (33%) of these individuals subsequently had positive test results by RT-qPCR. Two individuals were excluded from further involvement in the study due to pregnancy or hospitalization, while two other participants withdrew from the study during the follow-up period, but allowed the use of their data and the samples already collected. Therefore, our final study cohort comprised 52 individuals, with 50 of these participants remaining until the end of the 450-day follow-up period. Among the starting 52 participants, four (7.7%) worked at UPA, 16 (30.8%) at HMDCC, and 32 (61.5%) at HC. Overall, the average age of the subjects was 37.38 ± 6.99 years, and 55.8% were female. The final study cohort consisted of 22 (42.3%) physicians, 14 (26.9%) nursing technicians, 10 (19.2%) nurses, and 4 (7,7%) physiotherapists ( Table 1 ).

In January 2021, the vaccination schedule started in Brazil. Thus, 30 (58%) participants were infected with SARS-CoV-2 before starting the vaccination schedule, 11 (21%) during the schedule interval, and 11 (21%) after the complete vaccination, that is 15 days after the second dose of the initial vaccination scheme. During the 450 days of follow-up, 46 (88%) participants had suspected reinfections; of these, 19 (37%) cases were confirmed by RT-qPCR ( Figure 2 ). The proportion of women who became reinfected was higher than that of men (p=0.011). Among the participants who became reinfected 58% had already received the booster vaccine dose when the reinfection was confirmed.

The antibody peak observed between the seventh and thirtieth day (D7-D30) of each participant was chosen to be included in the analysis and represented a reaction average of 72.6 ( ± 68.0) for IgM ( Figure 3A ) and 193.7(± 121.2) for IgG ( Figure 3B ). Nine (17.3%) of the 52 participants did not IgG seroconvert within 30 days, and two (3.8%) individuals did not seroconvert at any time during the study to either of the two monitored immunoglobulins. In addition to these two latter individuals, 12 (23%) did not seroconvert to IgM throughout the study. The reactivity rate for IgM varied from 65%, 36%, 24%, 10%, 15%, and 24%, ( Figure 3A ) whereas for IgG it was 83%, 85%, 90%, 76% 58%, and 72% ( Figure 3B ) on days D7-30, D60, D90, D180, D270, and D450, respectively. Immunoglobulin levels decreased over time, reaching their lowest level on day 270 after infection when the mean level for IgM was 13.8 (± 17, p <0.0001) ( Figure 3A ) and for IgG was 68.5 (± 72.3, p<0.0001) demonstrating a significant reduction in comparison to the values observed at D7-D30. Figures 3C, D show the individual profile for anti-SARS-CoV-2 IgM and IgG in the study participants, respectively. Although the mean levels of anti-SARS-CoV-2 IgG antibody did not differ significantly between days 270 and 450, representing the period when the majority of reinfection and vaccination booster doses occurred, a tendency for IgG levels to increase was observed ( Figures 3A-F ). There was no difference between biological sex in the dynamics of antibodies ( Figures 4A, B ). However, significant differences in the dynamics of antibody levels were observed, with older individuals presenting the highest levels of IgM and IgG ( Figures 4C, D ) and a slower decrease in anti-SARS-CoV-2 IgG levels ( Figure 4D ) compared to IgM ( Figure 4C ). No significant differences in the Ct value obtained in the RT-qPCR at baseline were observed between the different sex and the age groups ( Figures 4E, F ).

The dynamics of IgM and IgG levels over time are shown in Figures 3E, F . For IgM and IgG immunoglobulins, there were two behaviors: the drop in the mean antibody levels up to 180 and 270 days, respectively, followed by a slight increase after this time point. Tables 2 and 3 demonstrate the behavior of IgM and IgG estimated by the segmented model. The GEE results indicate that the average IgM and IgG values decreased by 0.606 and 0.645 units each day until days 180 and 270, respectively. After this time, the slope is positive for both IgM (0.363) and IgG (0.224). As shown in Figure 3E , the observed mean IgM levels were below the cutoff despite this upward trend. We can also observe that 60 days after symptom onset, the mean anti-SARS-CoV-2 IgM levels are below the cutoff, while anti-SARS-CoV-2 IgG levels drop, but not below the cutoff, around 270 days after symptom onset ( Figures 3E, F ).

No significant difference in IgM and IgG antibody levels with regard to the vaccination status of the participants was observed ( Tables 2 , 3 ). However, after the booster dose, the anti-SARS-CoV-2 IgM and IgG levels both significantly increased. The IgM levels after the booster dose were 38.388 units higher. In addition, IgM levels were significantly higher after reinfection by 20.378 units. The interaction between these two covariables was not significant. In contrast, when we analyzed the IgG levels, the interaction between the booster dose and reinfection was significant, with the mean IgG levels 130.3 [95%CI: 82.19-178.30] units greater in those that received the booster dose and got reinfected. Before the booster dose, there was no difference in the mean IgG level when comparing reinfected and non-reinfected participants (p = 0.348). After the booster dose, the mean level of IgG for reinfected individuals was 87.8 [27.3; 148.0] units, which was higher than for non-reinfected individuals. For non-reinfected individuals, after the booster dose, the mean IgG was 52.22 units higher than before.

Of the 19 participants were reinfected ( Figure 5 ), 13 (68.4%) were after and 6 (31.6%) before the booster dose. In four (21.0%), increased levels of IgM and IgG antibodies were not observed, even after booster dose and re-infection, these individuals being non-reactive to SAR-CoV-2 antigens at day 450 for both immunoglobulin classes ( Figures 5F, G, J, O ). One individual ( Figure 5F ) did not seroconverted during the whole period of follow-up. Of the remaining 15 individuals were reinfected, considering IgM dynamics, four of them never seroconverted during the follow-up period ( Figures 5B, H, K, S ). Four individuals ( Figures 5C–E, R ) who were negative during visits D7-D30 only had IgM seroconversion after reinfection or the booster dose. For four other individuals ( Figures 5L, M, P, Q ) that became IgM reactive during period D7-D30, but turned non-reactive during the longitudinal assessment, the booster dose and reinfection were not able to stimulate production of this class of antibody. For the remaining three participants ( Figures 5A, I, N ), IgM levels declined to non-reactive, but the booster dose or reinfection seroconverted them again. For IgG, the dynamics are different, since all 15 reinfected participants seroconverted at some point during the study. Of these, one only seroconverted after the booster dose ( Figure 5K ). The other 14 participants had reactive IgG antibodies at baseline. For all of them, anti-SARS-CoV-2 IgG antibody levels declined, with seven participants becoming non-reactive to SARS-CoV-2 antigens for this class of antibody. For the remaining seven seroreactive participants, six showed increased IgG levels after reinfection or booster dose ( Figures 5H, I, L, M, P, R ). For the individuals who became IgG seronegative over the course of the study, the booster dose or reinfection was able to cause secondary IgG seroconversion of all such participants ( Figures 5A–E, N, S ).

Every 30 days or when participants presented symptoms that indicated a suspicion of COVID-19 reinfection, saliva or naso-oropharyngeal swab collection was performed to detect viral RNA. At the same time, a capillary blood sample was collected to evaluate immunoglobulins. The interval between symptoms onset and swab/saliva collection or antibody assessment varies from 1 to 15 days, with 75%of the participants being tested until 6 days of symptoms onset ( Supplementary Figure S1 ). Of the 52 participants, 46 (88%) described at least some symptoms compatible with suspected COVID-19, with 9 (17%) reporting it once, 19 (36%) twice, 9 (17%) three times, five (10%) four times, and four (8%) participants reporting five times ( Figure 6A ). In total, 114 suspected episodes of COVID-19 were recorded, all suspected cases were tested by RT-qPCR and of these only 19 were confirmed positive by RT-qPCR. Of all the symptom episodes/types (n = 185) reported by participants during the study, cough (61.6%) followed by congestion or runny nose (49.2%) and sore throat (41.1%) were the most frequent. The most common symptoms reported during the suspected and confirmed reinfections were congestion or runny nose (57.0% and 74%), cough (40.4% and 63%), and sore throat (45.6% and 47%), respectively. The profile of symptoms presented by reinfected individuals differs from those reported during the first SARS-CoV-2 infection, with statistical significance (p<0.001) for headache, fever and chills, myalgia, anosmia, and ageusia between groups were observed ( Table 4 ). Ageusia and anosmia were not reported in confirmed cases of reinfection (p<0.001). Other symptoms reported during the first infection in the suspected and confirmed cases of reinfection can be seen in Table 4 .

The first five (26%) cases of reinfection occurred between August and September 2021 ( Table 5 ). At this time, the Delta and Gamma variants were circulating in Belo Horizonte ( Figure 7 ). Between December 2021 and January 2022, 12 cases (63%) of reinfection were confirmed, a period that overlaps with the new wave of transmission caused by the Omicron variant ( Figure 7 ). Of the reinfection samples collected during this period that were sequenced, nine were identified as the Omicron variant. Sequencing samples with low viral load (Ct value > 36) was not possible. The mean CT of infected participants was 30.17 (± 6.5) ( Table 5 ). There was no significant difference between Ct values detected during infection and reinfection ( Figure 6B ). When assessing the humoral response of reinfected and non-reinfected individuals, we did not observe differences in IgM and IgG levels between these groups ( Figures 6C, D ). The intervals between infection and vaccination and seroconversion status are given in Tables 5 , 6 .