The reference cohort was made up of 949 apparently healthy individuals, presenting for blood donation from June 2019 to February 2021 at the Transfusion Medicine unit of the Fondazione IRCCS Ca’ Granda Hospital (Liver-Bible-Cohort 2021). The Liver-Bible-Cohort 2021 has previously been described (29) and the clinical features of this cohort are summarized in Table 1.

Since the first COVID-19 cases were officially registered in Italy on January 30, 2020, we used this date to separate blood donors who donated before (pre-pandemic) and throughout (pandemic-era) the pandemic. The absence of active infections at the time of sample collection was confirmed in all Liver-Bible-Cohort 2021 subjects by serological screening.

The retrospective cohort of hospitalized COVID-19 patients (Hospitalized COVID Patients, Table 1) has been determined following the STROBE guidelines for observational cohorts. The cohort was selected starting from the Fondazione Genomic Study (FOGS) collection of data and biological samples of hospitalized people included in COVID-19 registries at the Fondazione IRCCS Ca’ Granda Hospital. We considered as main criteria for the enrolment in the study patients aged over 18 years, admitted between March 2020 and February 2022 to the Infectious Disease Unit and the Intensive Care Unit (ICU) (Registro COVID, nCoV-ICU), with a confirmed COVID-19 diagnosis (positive RT-PCR for SARS-CoV-2 and/or positive nasopharyngeal swab and/or positive BAS/BAL). Only patients who provided informed consent were included. Patients for whom data were not collected at least 48h after their hospitalization, or whose data were missing due to initial admission to other facilities or units, were excluded from the study. In conclusion, a total of 212 patients met these inclusion criteria for the study (Figure 1).

The study was approved by the Ethics Committee of Milano Area 2, Fondazione IRCCS Cà Granda of Milan (approval number 342_2020). All participants enrolled in the study provided written informed consent prior to participation.

Given the relevance of cardiometabolic comorbidities known to influence circulating IL-32 levels, we collected data on arterial hypertension and related antihypertensive treatments, corticosteroid therapies, diabetes, and previous episodes or known pre-existing cardiovascular diseases. Furthermore, we classified COVID-19 disease severity based on ICU admission status.

The neutrophil-to-lymphocyte ratio (NLR) was calculated as the ratio between absolute neutrophil and lymphocyte counts obtained from the complete blood cell count.

Interleukin-6 (IL-6) levels were quantitatively measured using Elecsys IL-6 test, following the manufacturer’s instructions.

Whenever possible, the plasma samples collected for each patient were clustered in two groups: acquired at 24 and 48 hours after hospitalization (Acute phase) and acquired during visits at 3, 6, and 12 months post-discharge (Follow-Up phase). Infection waves were delineated in accordance with the criteria established by Protezione Civile Italiana (30). For the purposes of this study, emphasis was placed on stratifying data from the initial wave (March 2020 to September 2020) in comparison to the other four subsequent waves.

A complete overview of the clinical features of the Hospitalized COVID Patients cohort is reported in Table 2, while Supplementary Table S1 describes follow-up patients for the same clinical characteristics, stratified for the first wave and subsequent waves.

From this cohort, a subgroup of 42 patients was also included in a parallel study investigating endothelial biomarkers.

For plasma collection, peripheral blood collected in EDTA tubes was centrifuged at 2000 x g for 15 min and immediately stored at -80°C at the Fondazione Biological Resource Center (POLI-MI Biobank, which is part of the Italian node of Biobanking and Biomolecular Resources Research Infrastructure, BBMRI). Plasma samples were thawed, maintaining the cold chain to obtain aliquots. IL-32 plasma levels were measured using Human IL-32 DuoSet ELISA kit (Cat. N° DY3040, R & D Systems, Minneapolis, MN, USA), following manufacturer’s instruction. The assay is designed to detect IL-32α, IL-32β, and IL-32γ with a detection range of 78.5–5000 pg/mL. Kit storage followed the manufacturer’s instructions. Briefly, Samples were diluted 1:2-1:128 in cold PBS and measured in duplicate. Analyses were repeated, increasing dilution ratios when IL-32 values outside the standard curve were detected. Optical density was measured at 450 nm and 540 nm (as reported in the manufacturer’s instructions) using TECAN Infinite F200 PRO instrument (Männedorf, Switzerland). The minimal detectable concentration was 39 pg/mL. We employed the same methodology and procedures for data normalization and variance analysis of raw ELISA measurements as outlined in prior publications (29). Additionally, in instances where IL-32 concentrations were undetectable, a default value of 10 pg/mL was assigned.

Soluble thrombomodulin (TM) was measured using a sandwich ELISA kit (Human Thrombomodulin/BDCA-3 Quantikine ELISA Kit, Cat. N° DTHBD0, R&D Systems). Vascular endothelial growth factor (VEGF) in plasma was assessed by means of the Human VEGF Quantikine ELISA Kit (Cat. N° DVE00, R&D Systems). Vascular cell adhesion molecule-1 (VCAM-1) was measured in plasma using the Human VCAM-1/CD106 Quantikine ELISA Kit (Cat. N° DVC00, R&D Systems). Endoglin was assessed by means of the Human Endoglin/CD105 Quantikine ELISA Kit (Cat. N° DNGD00, R&D Systems). All measurements have been performed following manufacturer’s instruction. Optical density was measured using a VersaMax microplate reader (Molecular Devices, San Jose, CA).

Since IL-32 values were not normally distributed, log₁₀-transformation was applied for normalization.

Analogously, IL-6 and NLR values were normalized by applying the Rint transformation.

Comparisons of continuous variables between independent groups (Liver-BIBLE pre-pandemic, Liver-BIBLE pandemic-era, and Hospitalized COVID cohort) were performed using the unpaired Wilcoxon test. For continuous variables within paired groups of patients, where samples were available both at hospitalization during the Acute phase and at the Follow-Up phase, the paired Wilcoxon test was used.

The Liver Bible Cohort 2021 served as the reference population to calculate the minimum sample size needed to detect significant changes of one log₁₀ unit in distribution comparisons using the Wilcoxon test (power=0.8, two-tailed alpha=0.05). A sample size of at least 114 subjects was determined to achieve the desired statistical power, consisting of 38 controls and 76 patients. Categorical subsets and proportional fractions for each tested group were defined by low or high levels of circulating IL-32, with high levels corresponding to values above the third quartile of the total IL-32 distribution (>3.237 log10 pg/ml).

Possible correlations between variables (sex [M, F], age, COVID-19 severity [ICU, Not ICU], corticosteroid treatment [yes, no], anti-hypertensive therapies [yes, no], wave of infection [first, second-fifth], previous pathologies [type 2 diabetes, or arterial hypertension, or cardiovascular diseases], survivability [outcome, dead], pro-inflammatory biomarkers [IL-6, NLR], endothelial biomarkers (TM, VEGF, VCAM-1, endoglin]) and IL-32 levels were evaluated performing generalized linear model (GLM). P-values less than 0.05 were considered significant. R (v. 4.2.2) was used to perform all the statistical analyses.

To test whether SARS-CoV-2 infection may affect circulating IL-32 levels irrespective of disease course, we first analysed cytokine levels in apparently healthy blood donors sampled before and after the onset of the first wave in Italy. Results are shown in Figure 2. Comparison of pre-pandemic and pandemic-era groups in Liver-BIBLE showed a significant general increase in circulating IL-32 in blood donors after January 2020 (Δ Mean ± SE: +0.78 ± 0.09 log₁₀ pg/ml, p=2.2x10-16; pre-pandemic: 2.14 ± 0.07 log₁₀ pg/ml vs. pandemic-era: 2.91 ± 0.05 log₁₀ pg/ml). These data suggest that the spread of infection was associated with subclinical inflammation and an increase in IL-32 even in apparently healthy individuals in the population.

The Hospitalized COVID-19 patients’ cohort had ultimately enrolled 212 patients, primarily over 60 years old (66.8 ± 14.9), with the majority being male (138 individuals, 65.1%) (Table 2). Among them, 23 patients (10.8%) were admitted to the ICU due to COVID-19 severity or complications. A total of 125 (59.0%) patients were hospitalized during the first period of the pandemic (FIRST wave). The most prevalent comorbidity was hypertension, reported in 131 patients (61.8%), of whom 108 were receiving antihypertensive treatment. Additionally, 6 patients (0.1%) were on corticosteroid therapy. Seventy-five patients (35.4%) died from COVID-19 during hospitalization, while follow-up was available for 96/137 of surviving patients (70.1%).

Figure 3 shows the distribution of circulating IL-32 in hospitalized patients with COVID-19 (2.43 ± 0.08 log₁₀ pg/ml). Consistent with previous data, we detected a highly variable range in IL-32 measurements, which spanned over almost 5 log₁₀ pg/ml (Maximum value: 5.78 log₁₀ pg/ml).

When we compared IL-32 levels to those of healthy blood donors before the pandemic, we detected significantly higher levels of IL-32 in patients compared to controls (+0.29 ± 0.11 log₁₀ pg/ml, p=0.016), supporting the notion that IL-32 levels increase following SARS-CoV-2 exposure (Figure 3).

Next, we examined the clinical correlates of circulating IL-32 in hospitalized COVID-19 patients, first using univariate linear models to test associations.

The strongest positive association was observed between IL-32 levels and the severity of inflammation, as detected by the clinical indication to start corticosteroid treatment (estimate 1.99 ± 0.50; p=7.71×10⁻⁵, Table 3). Conversely, IL-32 levels were significantly lower in COVID-19 patients infected during the subsequent waves compared to those infected during the first wave (estimate -0.56 ± 0.16; p=0.0005).

Additionally, advancing age was modestly associated with a decline in circulating IL-32 (estimate -0.01 ± 0.01; p=0.020).

Multivariable regression analyses (accounting for corticosteroid treatment, infection wave, age, Table 3) corroborated the independent positive correlation between IL-32 levels and corticosteroid treatment (estimate 1.86 ± 0.49; p=0.0002), as well as the inverse association with hospitalization during the first wave of infection relative to subsequent waves (estimate -0.54 ± 0.16; p=0.001). Although the negative correlation between age and IL-32 levels persisted, it did not reach statistical significance (estimate = -0.01 ± 0.01; p=0.074).

When considering IL-32 concentration as the dependent variable in COVID patients, no statistically significant associations were observed with the other covariates characterizing the patient cohort: sex, disease severity (ICU), hypertension, cardiovascular comorbidities, or mortality.

To explore the potential involvement of IL-32 in the inflammatory response, we initially evaluated its associations with two key inflammatory markers—IL-6 (n=84) and the neutrophil-to-lymphocyte ratio (NLR) (n=203)—in participants from the Hospitalized COVID Patients cohort for whom complete data were available (Supplementary Table S2). This approach allowed us to assess both cytokine-based (IL-6) and cellular (NLR) indicators of inflammation in relation to circulating IL-32 levels. Correlation analyses indicated no significant relationship between IL-32 and IL-6 (estimate -0.17 ± 0.13; p=0.21). In contrast, circulating IL-32 was significantly and inversely associated with NLR (estimate -0.23 ± 0.81; p=0.005).

Given that IL-32 has also been implicated in angiogenesis and endothelial biology, we further expanded our analysis to explore the potential relationship between circulating IL-32 concentrations and markers of endothelial activation and injury. For this purpose, we focused on a subgroup of 42 patients from the same Hospitalized COVID Patients cohort (Supplementary Table S2). In total, 72 plasma samples were collected from these patients at multiple time points throughout hospitalization and follow-up, allowing for a dynamic assessment of biomarker changes over time. We then performed correlation analyses between IL-32 and a panel of endothelial biomarkers, including thrombomodulin (TM), vascular endothelial growth factor (VEGF), vascular cell adhesion molecule-1 (VCAM-1), and endoglin. None of these markers exhibited a statistically significant correlation with IL-32 levels (Table 4).

To investigate the longitudinal trend of IL-32 levels in COVID-19 patients, we analyzed those with follow-up visits by comparing their IL-32 measurements taken at hospitalization with levels measured post-discharge. IL-32 levels remained stable over time in these patients (+0.03 ± 0.12 log₁₀ pg/ml, Figure 4), with no statistically significant change observed (p = 0.970).

This outcome suggests that elevated circulating IL-32 can persist for several months after recovery from COVID-19.