Autoantibodies from Long COVID-19 and Their Possible Impact on Fertility

To identify pentapeptides that share sequence identity between Spike protein and spermatogenesis-linked proteins, a library of human proteins (in)directly linked to spermatogenesis was assembled from the UniProtKB database (https://www.uniprot.org↗) (24) using "spermatogenesis" as a keyword. Additionally, the Spike protein primary sequence (ancestral "Wuhan" virus isolate sequence) was split into pentapeptides offset by one–amino acid (aa) residue (that is, MFVFL, FVFLV, VFLVL, and so forth), and the resulting pentapeptides were analyzed for occurrences within spermatogenesis-related proteins. PIR Peptide Match (research.bioinformatics.udel.edu/peptidematch/index.jsp) and Peptide Search (www.uniprot.org/peptidesearch/↗) programs were used. Occurrences and the corresponding proteins were annotated.

Information on the Spike protein coded by the viral open reading frame 2 and its three‐dimensional (3D) structure coordinates file were obtained (25). 3D structure coordinate files were downloaded from the Protein Data Bank (PDB) site (https://www.rcsb.org↗). Protein molecular modeling and personalization were performed on the basis of the PDB file coded 6vxx, which represents the Spike protein closed conformation. Peptides 2, 3, and 4 were highlighted on the Spike protein chain A for their localization in its 3D structure. For peptide 1 located at the Spike protein C-terminus, a predictive 3D structure model was generated from the PepFold remote server as described (26, 27). Therefore, PDB files for all peptide sequences were obtained, fulfilling the Ramachandran plot structure requirements, as well as the geometrical constraint and restraint parameters for a valid protein structure. Hence, personalized molecular modeling for displaying all PDB 3D structure coordinate files was allowed using the VMD 1.9.3 version software (NIH Biomedical Research Center for Macromolecular Modeling and Bioinformatics, University of Illinois) (28).

We included 182 adult individuals in this study. Sera from 163 infected or vaccinated patients from different cohorts were obtained from donors after written consent. The main characteristics of infected patients acquired by retrospective chart review are summarized in. The most common effects present in patients with LCS included breathlessness, chronic fatigue, "brain fog" anxiety, and stress. No information in the patient's records regarding the possible signs of autoimmune infertility in men and women (anti-sperm antibodies in men, and premature ovarian failure and ovarian autoantibodies in women) was available. Due to the retrospective nature of the study, data on the pre-existing immunity to SARS−CoV−2, which is possibly linked to past hCoV strains or other sources of cross-reactive immunity to SARS-CoV-2, were also unavailable. As patients with LCS had been infected with SARS-CoV-2 in late summer to winter of 2020, they had not previously been vaccinated. Our study examined the sera of men and women to determine whether the data could differ as a function of gender (the sex of donors is listed in). 1 1

Eighty-six healthcare workers vaccinated in January to February 2021 with BNT162b2 (Comirnaty/Pfizer-BioNTech) and 45 with ChAdOx1 (Vaxzevria/Oxford-AstraZeneca) were included. Inclusion/exclusion criteria and demographic characteristics of vaccinated subjects have been reported previously (29). Adverse side effects or any clinical manifestation potentially correlated with vaccination were recorded for all the investigated subjects (29). The follow-up of the vaccinated subjects was carried out for 12 months starting from the first injection of the vaccine.

Control sera (n = 19;) were randomly selected from a bank of healthy blood donations collected before the worldwide outbreak of COVID-19 (Etablissement Français du Sang, Strasbourg, France). As anonymous blood donation involves no disclosure of the donor's identity or personal information, no information was available regarding gender and age distribution of voluntary, non-remunerated donors. All serum samples were stored at −20°C or −80°C until use. 1

Peptides were synthesized and purified as described (30). Their homogeneity was checked by analytical high-performance liquid chromatography, and their purity was ≥97%. Peptide identity was verified by liquid chromatography–mass spectrometry. Spike peptides were mixed with maleimide-activated keyhole limpet hemocyanin (Sigma, K0383) and given subcutaneously (first administration) and intramuscularly (booster, after 2 weeks) to healthy, inbred strain, BALB/c female mice (100 µg injection/mouse; six mice/peptide). Peptide solutions were mixed 1:1 with complete Freund's adjuvant (Sigma, ref. F5881) for the first injection and with incomplete Freund's adjuvant (Sigma, ref. F5506) for subsequent injections. Mice were bled 7 days after the second injection. A control blood sample was collected from each mouse before the first injection. The blood withdrawal was performed under gas anesthesia (O₂/isoflurane, 4%–5% induction, and 1%–3% maintenance). At the end of the experiment, mice were euthanized by cervical dislocation. Serum anti-Spike peptide antibodies were purified using protein G magnetic beads (Magne Protein G, Promega, G7472) following the manufacturer's instructions. Purified antibodies were stored without preservatives at −80°C until use.

Standard enzyme-linked immunosorbent assay (ELISA) procedures measured antibody reactivity in human and mouse sera. Spike peptides (2 µM) were directly coated onto polyvinyl plates (Thermo Fisher, ref. 2801; 100 µL/well) and incubated overnight at 37°C. Plates were post-saturated with 0.8% (w/v) bovine serum albumin (BSA; Roche Diagnostics, ref. 10735094001) in phosphate-buffered saline (PBS) containing 0.05% (v/v) Tween (PBS-T) for 1 h at 37°C. After washing plates with PBS-T, human or mouse sera, diluted 1:1,000 and 1:500, respectively, in PBS-T–0.8% BSA, were added to plates and incubated for 2 h at 37°C (100 µL/well). Plates were washed before applying (100 µL/well) goat anti-human or goat anti-mouse IgG Fcγ–horseradish peroxidase (HRP) conjugates (Jackson ImmunoResearch Laboratories, refs. 109-035–008 and 115-035-008, respectively) diluted in PBS-T (1:20,000 and 1:15,000, respectively) to detect antibody binding. After incubation for 1 h, plates were washed twice with PBS-T and twice with distilled water. The final reaction was visualized with H₂O₂ used as substrate and 3,3′,5,5′-tetramethyl benzidine (Sigma, ref. T2885-SG) used as the chromogen (75 µL/well). After 30 min at 37°C, the enzymatic reaction was stopped by adding 25 μL of 1 N HCl. Absorbance was measured at 450 nm on a Thermo Scientific Multiskan. Absorbance values were corrected by subtracting the extinction values from the plate material measured at 535 nm. All serum samples were systematically tested in at least two independent assays. Each plate contained several samples from healthy individuals to determine baseline absorption and possible false reaction. The threshold for positivity was determined as the mean absorbance value +2 SD of the control sera.

The full-length recombinant SARS-CoV-2 (2019-nCoV) Spike (S1 + S2) protein containing a histidine (His)–tag at its C-terminal end (BioLegend, ref. 793706) was mixed with Laemmli buffer (Bio-Rad, ref. 1610737) and denatured for 10 min at 95°C. The protein (1–2 µg) and molecular weight marker (Bio-Rad, ref. 1610393) were run on 4%–20% Mini-PROTEAN TGX Gels and transferred onto Trans-Blot Turbo nitrocellulose membrane (Bio-Rad). Rabbit IgG (1 µg) specifically recognized by HRP-conjugated anti-rabbit IgG antibody was included as a negative control. Following the saturation with blocking agent, blots were incubated with the primary antibodies raised in mice or anti-His monoclonal antibody (0.3 µg/mL; Invitrogen, MA1-21315) overnight at 4°C. HRP-conjugated goat anti-mouse IgG (1:15,000; Jackson ImmunoResearch, 115-035-008) or HRP-conjugated goat anti-rabbit IgG (1:15,000; Jackson ImmunoResearch, ref. 111-035-008) secondary antibody and Clarity Western ECL Substrate (Bio-Rad, ref. 1705060) were used to reveal the signal through the ChemiDoc XRS System (Bio-Rad).

Testes from a healthy 10-week-old C57BL/6 mouse were collected, fixed in 4% (v/v) paraformaldehyde for 24 h, incubated in 30% (w/v) sucrose overnight, snap-frozen, and stored at −80°C. Cryosections (7 µm) were cut, permeabilized in PBS–0.1% (v/v) Triton X-100 for 5 min, briefly rinsed in PBS, and blocked in PBS–0.1% Triton X-100–3% BSA for 1 h. Sections were incubated with primary antibodies: anti–peptide 2 or mouse IgG control (Sigma, ref. I5381), diluted 1:20 in PBS–Triton X-100–BSA blocking buffer overnight at 4°C. After washing, AF488-labeled goat anti-mouse IgG (Invitrogen, ref. A11029) was applied for 1 h at room temperature, and nuclei were stained with Hoechst 33258 for 10 min (2 µg/mL in PBS; Invitrogen, ref. H3560) before mounting cover slips with Fluoromount Aqueous Mounting Medium (Sigma, ref. F4680). Images were acquired using a Leica DMRA2 microscope.

Statistical tests were performed using GraphPad Prism 6 (GraphPad Software Inc.). The statistical tests applied are described in the corresponding figure legends. The statistical significance was assessed by the Mann–Whitney non-parametric test. All data presented (mean ± SD) were obtained from at least two independent experiments. Heatmap graphs were constructed using the ComplexHeatmap R package (https://www.bioconductor.org/packages/release/bioc/html/ComplexHeatmap.html↗) and the Morpheus tool (https://software.broadinstitute.org/morpheus↗). Regression analysis was done using GAMLSS R packages with ZAGA (Zero Adjusted Gamma Distribution; https://search.r-project.org/CRAN/refmans/gamlss.dist/html/ZAGA.html↗).

Initially, we defined a list of pentapeptides common to human proteins linked to spermatogenesis and SARS-CoV-2 Spike protein. A library of 667 human proteins linked to spermatogenesis was assembled randomly from the UniProtKB database using "spermatogenesis" as a keyword (Supplementary Table S2). In parallel, the primary sequence of Spike protein was dissected into pentapeptides offset by one-aa residue. The resulting pentapeptides were analyzed for occurrence within spermatogenesis-related proteins using the PIR Peptide Match and Peptide search programs. Peptide matching analyses led to establishing a list of 217 proteins sharing 245 pentapeptides with the Spike protein for 324 pentapeptide occurrences (Supplementary Table S3). Given the size of the viral versus human pentapeptide overlap, analyses were limited to 20 protein alterations that had been documented to be related to spermatogenesis disorders using Online Mendelian Inheritance in Man and/or PubMed (Supplementary Table S4). This list was further sorted using the Immune Epitope DataBase resource (www.iedb.org/↗), which includes experimentally validated immunoreactive, Spike protein epitopes (Supplementary Table S5) (26).

From the 58 segments encompassing contiguous five-mer-long sequences shared between Spike protein and spermatogenesis-related proteins shown in Supplementary Table S5, four peptides named 1 to 4 were finally selected for synthesis and testing (Figure 1). Peptide 1 (16 aa; Pi = 4.2) encompassed a long sequence of nine residues (DEDDSEPVL) shared between the C-terminus of Spike protein and spermatogenesis-related proteins. Peptide 2 (13 aa; Pi = 10.6) contained a stretch of five shared residues (PSKPS) in the testis-specific serine/threonine-protein kinase 1, essential for male fertility. The sequence of peptide 3 (13 aa; Pi = 10.9) was contiguous to a motif shared between Spike protein and a spermatogenesis-related protein but did not encompass shared residues. Likewise, peptide 4 (14 aa; Pi = 9.9) did not contain a sequence shared between Spike protein and a spermatogenesis-related protein but encompassed the motif IEDLL that is present in histone-lysine N methyltransferase 2D (KMT2D), which is linked to oogenesis (IEDB ID 1309578) (31). Peptides 4 and 2 were overlapping by four residues (KRSF) in the primary sequence of Spike protein. Peptide 2, 3, and 4 are located on the Spike protein chain A. The 3D molecular model of Spike protein and the localization of peptides 1 to 4 are depicted in Figures 1A, B.

To evaluate the antigenicity of peptides 1–4, we recruited cohorts of adult individuals from different countries, either patients presenting COVID-19 infection proved by nasopharyngeal swabs or polymerase chain reaction test (n = 163) or healthy donors (n = 19) (Figure 2A, Supplementary Table S1). Infected patients were categorized into two groups corresponding to acute COVID-19 (n = 87) or LCS (n = 76), on the basis of infection severity and length. In these two groups, the mean age was 53 and 42 years, respectively (Supplementary Tables S1A, B). The repartition of men and women was found to be different in the two groups, namely, 78% men/22% women in the acute group and 37% men/63% women in the LCS group (Supplementary Tables S1A, B). The sera from healthy donors were used to calibrate the test conditions, determine the threshold for positivity [cutoff optical density (OD) values], measure the inter- and intra-assay coefficients of ELISA variability, and verify the absence of unwanted reactivity. Following these calibration steps, the peptide reactivity of sera was accurately evaluated. IgG reactivity was investigated because IgM data are more prone to false positivity, mainly due to test sensitivity and misinterpretation.

As expected, sera of healthy subjects showed low and infrequent (<5% of subjects) reactivity with peptides 1–4, while reactivity with these peptides was found in patients' sera (Figure 2). Patient's sera reacted with variable levels of intensity (Figure 2B); the highest antibody reactivity was observed in the German cohort, mostly with peptide 4, followed by that of sera collected in Singapore, primarily with peptide 1 (Figures 2C, D). Although statistically significant, the ELISA absorbance values were generally low with peptide 2, in contrast to absorbance values measured with peptide 4.

Next, the sera reactivity with peptides 1–4 was examined in acute or long COVID-19 patients. Infected patients showed the highest serum reactivity with peptide 4 both in acute and long form (mean OD = 0.15, frequency of 25%; and mean OD = 0.093, frequency of 21%, respectively). Additionally, we found that sera of LCS patients showed significantly higher levels of antibodies against peptide 2 compared to acute COVID-19 individuals' sera (mean OD = 0.012, frequency of 15.8%, versus mean OD = 0.007, frequency 9%, P = 0.008, for acute disease and LCS, respectively; Figure 2E).

Based on these results with peptides 2 and 4, we further characterized certain features of COVID-19 patients with and without peptide-reactive antibodies (Figure 3, Supplementary Figures S1, S2). Acute COVID-19 patients with positive antibodies to peptides 2 and 4 were mainly over 60 years old (69% and 58%, respectively), at variance with LCS patients expressing positive antibodies, who were younger (Figures 3A, B). Considering the gender of our patients' cohort and the severity of viral infection, we found that women affected by LCS significantly presented a higher rate of peptide 2 antibodies compared to those with acute forms of COVID-19 (mean OD = 0.017, frequency of 24%, versus mean OD = 0.005, frequency 5%, for LCS and acute COVID-19, respectively; P = 0.04). In contrast, we observed only a trend of a higher level of peptide 4 antibodies in men affected by acute COVID-19 compared to LCS (mean OD = 0.18, frequency of 28%, versus mean OD = 0.15, frequency of 21%, P = 0.35, in patients with acute disease and LCS, respectively; Figure 3C).

The possible association of peptide-reactive antibodies with some pathological characteristics of SARS-CoV-2–infected patients was also analyzed. Detailed information was not systematically listed in the patient's clinical records to which we had access for our analyses. Based on the available data, no statistically significant correlations emerged concerning to typical biological parameters, e.g., the rate of reactive protein C, levels of total IgG and IgM antibodies, and pre-existing serum autoantibodies to classical autoantigens (e.g., double-stranded DNA, small nuclear ribonucleoproteins, and extractable nuclear antigens Ro/La). No correlation was found with documented clinical features, such as respiratory or neurological impairment (loss of memory), headache, and fever.

No correlation could be established when we considered patients with LCS presenting or not chronic fatigue syndrome (CFS) as a consequence of SARS-CoV-2 infection (Figure 4A), regardless of gender (Figure 4B). Interestingly, when LCS patients with CFS were compared to COVID-19 negative subjects presenting idiopathic CFS, a significant correlation could be made about the presence of antibodies reacting with peptides 2 and 4 (peptide 2: non COVID-19 mean OD = 0.0053, frequency of 5%, versus long COVID-19 mean OD = 0.012, frequency of 14.5%, P = 0.018; peptide 4: non-COVID-19 mean OD = 0.015, frequency of 0%, versus long COVID-19 mean OD = 0.10, frequency 21%, P = 0.04) (Figure 4C). Furthermore, the sera of women infected by SARS-CoV-2 possessed significantly increased levels of antibodies reacting with peptide 2 compared to women with idiopathic CFS (15% versus 0%; P = 0.012; Figure 4D).

A similar analysis was also performed using the sera from 131 vaccinated Italian individuals. We tested 90 sera from 45 healthcare workers pre- and post (1 month)–Oxford/AstraZeneca vaccination and 172 sera from 86 healthcare workers pre- and post (3 months)–Pfizer/BioNTech vaccination (Table 1). No antibody reactivity was detected in the sera from 131 pre-vaccinees with peptides 1 and 2. Some reactivity with peptides 3 and 4 was measured pre-vaccination in healthcare workers. However, post-vaccination, the percentage of sera reacting with any of the peptides 1–4 was low, and their level of reactivity was extremely weak (Table 1). Results lead to conclude that the pattern of sera's reactivity against Spike protein peptides 1–4 was not found in vaccinees at least 1–3 months post-vaccination with either Oxford/AstraZeneca or Pfizer/BioNTech vaccines.

To characterize the potential functional role of cross-reactive antibodies, we finally immunized healthy mice with peptides 2 and 4. No macroscopic alterations were observed in organs, tissues, or body cavities post-immunization, indicating that neither the peptides nor the induced antibodies exhibited apparent toxicity in BALB/c mice. The generated antibodies were affinity-purified, and their purity was validated through biochemical analyses (Figures 5A, B). The specific binding of purified IgG antibodies to peptides 2 and 4 with Spike protein was also confirmed by western blot (Figure 5C). The reactivity of IgG antibodies to peptide 2 was further studied by immunochemistry using mouse testicular tissues that we incubated with these antibodies or with normal mouse IgG used as control (Figure 5D, Supplementary File 1). While no specific signal was observed with control IgG, histological images pointed out a specific one in testicular tissue treated with IgG antibodies to peptide 2, indicating that these antibodies readily react with antigens expressed by mouse testicular tissues.