Twenty participants were selected from the COVID-19 cohort at Virgen del Rocío University Hospital (VRUH, Seville, Spain) and eighteen Controls (without any documented SARS-CoV-2 infection) from Heliopolis nursing home (Seville, Spain). The Control population included a highly monitored group of elderly individuals, with regulated visits and routine analytical assessments. In fact, due to this rigorous monitoring, five active cases of SARS-CoV-2 were detected during the sample collection and were therefore excluded from the analysis. All the participants underwent the memory alteration test (M@T) (18). Inclusion criteria were: ≥50 years old, confirmed SARS-CoV-2+ PCR (at least 12 months before the M@T), and no SARS-CoV-2 reinfections. Exclusion criteria: drug and alcohol abuse, active infections, hospital admission, anti-tumor therapy, or any immunomodulatory therapy or treatment that could influence the immune system (mainly corticosteroids) at least 6 months before the beginning of the study. Demographic data, self-reported symptoms (COVID-19), and blood samples were obtained at the moment of M@T performance. The COVID-19 and Control groups were age- and sex-matched and had received at least one vaccine dose. Colon biopsies of five COVID-19 participants were collected during a colonoscopy at 30-56-65-90-297 days after SARS-CoV-2+ PCR. Written informed consent was obtained. The study was reviewed and approved by the Ethics Committee of Virgen Macarena and VRUH (C.P. NeuroCOVIH-C.I. 1155-N-21, C.P. S230054-C.I. 1518-N-23).

Peripheral Blood Mononuclear Cells (PBMCs) were isolated as previously (11) and cryopreserved (90% Fetal Bovine Serum (FBS), ThermoFisher Scientific, 10% dimethyl sulfoxide, PanReac AppliChem) in liquid nitrogen until further use. Plasma samples were collected as previously (8) and stored at -80°C.

Complete hemograms were determined using Epic XL-MCL or FC500 flow cytometer (Beckman Coulter Inc., California). All the assays were performed according to the manufacturers’ instructions.

For monocyte functional assays, 1x10⁶ PBMCs were in vitro stimulated with lipopolysaccharide (LPS, 0.5ng/mL, Invivogen) in R-10 medium (10% FBS, 1% L-glutamine, 1% penicillin-streptomycin in RPMI medium) (5h, 37°C), including a negative control without stimulation. PBMCs were incubated with monensin (Golgi Stop, BD Biosciences) following the manufacturer’s instructions and intracellular cytokines were analyzed by multiparametric flow cytometry.

Flow cytometry.

In general, after washing with PBS (625g, 5min, RT), PBMCs were incubated with the viability marker (LIVE/DEAD Fixable Aqua or LIVE/DEAD Fixable Violet Cell Dead Stain Kits, Invitrogen) and the extracellular antibodies (35min, RT) ( Supplementary Table S1 ). PBMCs were then washed, fixed and permeabilized with BD Cytofix/CytoPerm (BD Biosciences) (20min, 4°C) for cytoplasmic markers (Indoleamine 2,3-dioxygenase (IDO) and cytokines) or Fixation/Permeabilization Buffer Set (eBioscience) (45min, 4°C) for Ki67, following the manufacturer’s instructions. Cells were incubated with intracellular markers (30min, 4°C), washed and fixed with 4% paraformaldehyde solution (PFA, Sigma-Aldrich) (20min, 4°C). Extracellular and intracellular markers’ staining was performed adding the volumes/concentrations of each antibody recommended by manufacturers. For monocyte identification, HLA-DR⁺ cells were selected and classical (CD14⁺⁺CD16^-), intermediate (CD14⁺⁺CD16⁺) and non-classical (CD14⁺CD16⁺) monocytes were identified ( Supplementary Figure S1 ). Within CD8⁺ and CD4⁺ T-cells, different subsets were gated: naïve (CD45RA⁺CD27^-), central memory (CM, CD45RA^-CD27⁺), effector memory (EM, CD45RA^-CD27^-), terminally differentiated effector memory (TEMRA, CD45RA⁺CD27⁺) and total memory (Memory) ( Supplementary Figure S2 ). DCs were gated as Lin-2^- and HLA-DR⁺ cells, and DC subsets (myeloid, mDCs, and plasmacytoid DCs, pDCs) were gated based on CD11c and CD123 expression, respectively. mDCs subsets were gated based on CD16, CD1c and CD141 expression ( Supplementary Figure S3 ).

Multiparametric flow cytometry was performed on an LRS Fortessa flow cytometer using FACS Diva software (BD Biosciences), acquiring 0.5-1×10⁶ events. Data were analyzed using the FlowJo 10.7.1 software (TreeStar). Additionally, flow cytometry data were analyzed with dimension reduction tools. Only data from CD4⁺ T-cells are shown, since no significant differences were found in the rest of the populations. Total CD4⁺ T-cells were gated as explained before ( Supplementary Figure S2 ). Then, data was downsampled to 6.000 events (CD4⁺ T-cells) and concatenated. From the concatenated, unbiased hierarchical clustering was performed using Uniform Manifold Approximation and Projection (UMAP, data visualization), Phenograph (to determine the number of clusters), and FlowSOM (cluster analysis). Both concatenated and separated data in groups (Control and COVID-19) were analyzed ( Supplementary Figure S4 ).

Colon biopsies were fixed in 10% formalin, paraffin-embedded in blocks and sliced using a Leica RM2255 microtome (3 µm of thickness). Immunostainings against SARS-CoV-2 Nc and spike (S) proteins were performed as previously described with modifications (19). Samples were incubated with the primary antibodies (1:100, o/n, 4°C), and secondary antibodies (1:500, 1h, RT) ( Supplementary Table S1 ), and mounted using Prolong Diamond Antifade Mountant with DAPI (Thermo Fisher Scientific). Images were taken using a Leica Stellaris 8 laser-scanning confocal microscope using the 40x air objective (HCX PL APO 40x/0.95 W.D., 0.17mm) with constant acquisition parameters. A sample from the pre-COVID period was included as a negative control of Nc and S immunolabelling. The fluorescently labeled structures were analyzed using Fiji Image J software (20). Image background was first subtracted before setting of a brightness threshold for the measurement of Nc or S markers. Images were analyzed from a maximum intensity projection. 3D reconstructions were performed with Imaris software (Bitplane).

Total and neuronal-derived (NDEs) EVs were isolated, as previously reported (21) using the SmartSEC™ Single for EV Isolation (System Biosciences), following the manufacturer’s instructions. NDEs were obtained as previously described (16) by incubating with mouse anti-CD171 antibody in 50μL of 3% BSA (1h, RT). Samples were then incubated with 10μL of streptavidin-agarose Ultralink Resin (Thermo Fisher Scientific) in 3% of Bovine Serum Albumin (BSA) (1h, 4°C). Afterward, samples were centrifuged (800g, 10min, 4°C) and the pellet resuspended in 100μL of cold 0.05M Glycine-HCl (pH3, 5min) and centrifuged again (4000g, 10min). Supernatants were then gently mixed with 25μL of 10% BSA, 10μL of Tris-HCl (1M, pH8) and 370μL of the mammalian protein extraction reagent (M-PER, Thermo Fisher Scientific) and stored at -80°C.

EVs protein concentration was quantified using the BCA assay (BCA Protein Assay-Kit, Thermo Scientific), following the manufacturer’s instructions. EV proteins were separated using SDS-PAGE with pre-cast gels (4-20%, Bio-Rad) and transferred to nitrocellulose membranes using the TransBlot Turbo System (Bio-Rad). Blots were incubated with the primary antibody (1:500, o/n, 4°C), washed and incubated with the corresponding secondary antibody (1:5000, 2h, RT) ( Supplementary Table S1 ). Blots were developed using the Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific), visualized (ChemiDoc-Touch, BioRad) and analyzed by densitometry using Fiji ImageJ Software.

Pro-inflammatory cytokines in both plasma and EVs were analyzed using the MesoScale Discovery U-Plex multiplex (IFN-γ, IL-1β, IL-6, IL-8, IL-12p70, IL-18, IP-10, MIP-1α, MIP-1β, and TNF-α) following the manufacturer’s instructions. Neurofilament (NfL) levels were analyzed in NDEs using the NEFL ELISA Human Kit (OKEH02111, Aviva Systems Biology), according to the manufacturer’s guidelines. Samples below the limit detection were considered as zero.

Non-parametric statistical analyses were performed using SPSS (the Statistical Package for the Social Sciences software, SPSS v.25.0, Inc., Chicago, IL) and Prism (v.8.0, GraphPad Software, Inc.). In graphs, individual dots represent one participant (Blue: Control; Red: COVID-19). Continuous variables were expressed as medians with interquartile ranges [IQR] and categorical variables as percentages. The ROUT method was used to identify and discard outliers (Q= 0.1%). Differences between groups were analyzed by 2-tailed unpaired Mann-Whitney U test. The polyfunctionality index (P-Index) was calculated using Funky Cells software (22). P values <0.05 were considered statistically significant and are indicated in the figure legends.

Twenty participants who underwent COVID-19 at least 15 months before the study (15.30 [13.79-15.56]), and thirteen aged- and sex-matched Controls were enrolled. Time after SARS-CoV-2 infection ranged from 12.2 to 22.4 months. Clinical and demographical parameters are listed in Table 1 and Supplementary Table S2 . COVID-19 participants presented significant decreased numbers of leukocytes and neutrophils compared to Controls; no significant differences were found regarding the rest of the blood cell subsets. The percentage of participants with diabetes mellitus and hypertension was significantly higher in COVID-19 than in Control group ( Supplementary Table S2 ), both comorbidities described as risk factors for COVID-19 (23). Interestingly, most of the self-reported symptoms in the COVID-19 group were associated with neurological alterations, including headache or mood manifestations (30% and 45%, respectively).

PASC involves neurological disturbances (24) and brain structure alterations (25) associated with cognitive impairment. To analyze the long-term effects of SARS-CoV-2 infection related to memory dysfunction, we used the M@T (18). Remarkably, although no differences were observed in the total M@T score ( Figure 1A ), a higher concentration of NfL was found in NDEs from COVID-19 ( Figure 1B ), suggesting that neurodegenerative processes may still be taking place 15 months after SARS-CoV-2 infection. As previously reported in acute infection (26), the levels of CD81 (a protein of the tetraspanin complex) were significantly higher in the COVID-19 group than in Controls ( Figure 1C ). Immune alterations and elevated levels of inflammatory markers have been found in COVID-19 (27). Although we did not detect differences in the concentration of pro-inflammatory cytokines in plasma ( Supplementary Figure S5 ), significantly decreased concentration of IFN-γ ( Figure 1D ), but higher concentration of IL-18 ( Figure 1E ) and MIP-1β ( Figure 1F ) were detected in EVs of the COVID-19 group in comparison with Controls. No significant differences were detected for the other cytokines analyzed in EVs ( Supplementary Figure S5 ). Our data suggest a higher transport of some inflammatory molecules through EVs 15 months after SARS-CoV-2 infection.

As important players in inflammation and COVID-19 severity (28), we further characterized activation and homing markers in monocytes 15 months after SARS-CoV-2 infection. Our results showed that in COVID-19 participants, there was decreased expression of classical monocytes expressing tissue factor (CD142⁺) ( Figure 2A ) and both classical and intermediate monocytes expressing the activation marker TLR4 ( Figures 2B, C ). No differences were observed in the rest of the markers analyzed, except for non-classical monocytes expressing TLR2 ( Supplementary Figure S6 ). Additionally, monocytes were stimulated in a TLR4-dependent manner by adding LPS, similar to previously reported (11). Although no differences were found in the production of IL-1α ( Figure 2D ), the COVID-19 group exhibited a significantly higher production of IL-6 and TNF-α by monocytes upon LPS stimulation ( Figures 2E, F ) compared with Control group. However, the P-index tended to be lower in the COVID-19 group ( Figure 2G ), suggesting less polyfunctional monocyte response than in Control group.

We have previously reported an altered T cell pattern in previously hospitalized COVID-19 patients 7 months after SARS-CoV-2 infection (8). Fifteen months after SARS-CoV-2 infection, in CD4⁺ T-cells, we found a significant increased percentage of HLA-DR⁺ CM T-cells ( Figure 3A ) and in CD38⁺HLA-DR⁺ memory population in COVID-19 compared to Control ( Figure 3B ). In line with these results, UMAP analysis ( Figure 3C ) reported higher proportion of a specific CD4⁺ T cell population in COVID-19 (population 6, right bar graph); this population included CM and EM CD4⁺ T cells, with an activated (HLA-DR⁺⁺) and no senescent phenotype (CD57^-CD28⁺). A higher percentage of the bulk CD8⁺ T-cells ( Figure 3D ) was identified in the COVID-19 group. Interestingly, a lower percentage of senescent (CD28^-CD57⁺) CD8⁺ EM T-cells were observed in COVID-19 in comparison with Control group ( Figure 3E ). No differences were found in any of the other markers analyzed for both CD4⁺ and CD8⁺ T-cells ( Supplementary Figures S7A, B , respectively).

We previously reported deficiencies in DCs 7 months after SARS-CoV-2 infection (9). A deep immunophenotyping of both mDCs and pDCs revealed decreased levels of DCs expressing IDO⁺, including total mDCs ( Figure 4A ), CD16⁺ ( Figure 4B ), and CD1c⁺ ( Figure 4C ) mDCs and pDCs ( Figure 4D ) in COVID-19 compared to Control group. No differences were found in any of the other markers analyzed ( Supplementary Figure S8 ). Interestingly, the percentage of pDCs expressing β7-integrin was increased in the COVID-19 group ( Figure 4E ).

pDCs expressing β7-integrin migrate from peripheral blood to colorectum tissue in simian immunodeficiency virus (SIV) infection (29). The analysis of the SARS-CoV-2 Nc and S immunolabeling in colon samples ( Figure 5 ) revealed that the percentage of occupied area of both proteins decreased over the time. However, SARS-CoV-2 persistence was still observed 297 days after the infection in colon tissue.