Two groups of patients hospitalized for COVID-19 were assessed: 1) patients treated with IVIG (COVID-19 IVIG; n=18) and 2) untreated controls (COVID-19 controls; n=17). In addition, 13 healthy donors were included, who donated blood once a week for 1–4 weeks after passing daily COVID-19 screening protocols. Inclusion criteria for the COVID-19 patients were confirmed or suspected diagnosis of COVID-19, rapid hypoxemic respiratory failure from acute respiratory distress syndrome (ARDS), and mechanically ventilated status. Patients were excluded from consideration of off-label IVIG for the purposes of this study as well as in the overall off-label allocation of IVIG in our healthcare system during the COVID-19 pandemic if they had pre-existing chronic organ failure, history of stroke, dementia, or active malignancy. Both groups received standard of care diagnostic and therapeutic management contemporary for the period in the pandemic (July 2020-March 2021). No patients in this study were vaccinated for SARS-CoV-2.

A total of 18 COVID-19 IVIG patients were included in the study with intention to treat with off-label IVIG therapy (0.5 g/kg adjusted body weight/day for four days). IVIG was administered ≤72 hours after the onset of mechanical ventilation. After successful therapeutic application of IVIG in five patients, the protocol “IVIG in Patients with Severe COVID-19 Requiring Mechanical Ventilation” (clinicaltrials.gov, NCT04616001) was initiated, wherein the same treatment regimen was accompanied by patient consent for collection and subsequent analysis of blood samples. All untreated COVID-19 patients, or designated family member, and healthy controls provided informed consent in accordance with UCSD Rady Children’s Hospital and VASDHS institutional review board (IRB) approval (190699 and B200003, respectively) and within the Helsinki Declaration of the World Medical Association.

Patient sex, age, body mass index (BMI), pre-existing conditions, time of hospitalization, time of intubation for mechanical ventilation, and length of stay were obtained from the electronic medical record. APACHE II scores were calculated on admission to the ICU and O₂ saturation (SaO₂) to fractional oxygenation (FiO₂) ratios were calculated on days 1, 3, 5, and 8 after the first dose of IVIG.

Blood samples used in this study were residuals from samples drawn as part of routine bloodwork in the ICU; no additional blood was required for inclusion in this study. Plasma was isolated from whole blood by conventional methods and was stored at -80°C. For IVIG-treated (COVID-19 IVIG) patients, samples were collected a day before treatment (referred to as pre-treatment), on day 3–6 after first IVIG was given (referred to as early treatment), and on day 8-13 (referred to as late treatment). Similarly, samples from COVID-19 controls were taken within the same range of time points based on initiation of mechanical ventilation to match their IVIG-treated counterparts.

Plasma was depleted from HSA/Immunoglobulin using Thermo Scientific™ High Select™ HSA/Immunoglobulin Depletion Mini Spin Columns (cat.#: A36366) prior to proteomics sample prep. 10 μl of plasma was added onto the HAS/lgG depletion column containing a resin slurry. Column was capped, incubated for 10 min at room temperature while vortexed to ensure complete coverage in resin. After incubation, columns were centrifuged in a 2 mL collection tube at 1,000 x g for 2 minutes. Collected samples were dried down in a speed-vac for proteomic sample prep. For mass spectrometry analysis, plasma samples were sonicated in buffer containing 6 M Urea, 7% SDS, 50 mM TEAB, and protease inhibitor/PhosStop tablet (Roche), with pH adjusted to 8.1 with phosphoric acid. Samples were reduced, alkylated, and acidified before being mixed per manufacturer’s instructions and loaded on S-Trap columns (Protifi). On-column digestion with trypsin for 3 hours at 47°C followed by elution in 50 mM TEAB, 5% formic acid, and 50%/5% acetonitrile and formic acid. Samples were desalted (Waters, C-18 Seppak), and 50 μg aliquots of each sample dried using a speed-vac. Samples were labeled per manufacturer’s instructions (TMT10plex, Lot VL312003, ThermoFisher).

Basic pH reverse-phase LC, followed by data acquisition through LC–MS²/MS³, was performed as described (46, 47). Linear gradients of 22% to 35% acetonitrile and 10 mM ammonium bicarbonate were passed on HPLC C18 for 75 minutes columns and the resulting fractions concatenated. Fractions were analyzed using tandem mass spectrometry (MS³) on an Orbitrap Fusion MS (ThermoFisher) with an in-line EASY-nLC 1000 (ThermoFisher). Separation and acquisition settings were performed using previously defined methods (48).

MS data was screened against the reference proteome for Homo sapiens downloaded from Uniprot.com on 1/30/2019 using Proteome Discoverer 2.1, and SEQUEST was used to align MS² spectra against theoretical peptides generated in silico. Static modifications included TMT labels on N-termini and lysine residues, and carbamidomethylation of cysteines. Dynamic modifications included oxidation of methionine. A 1% false discovery rate was specified for the decoy database search. Peptide spectral match abundances were first summed to the protein level then normalized against the average for each protein divided by the median of all average protein values. A second normalization step was performed whereby the abundance value for each protein per sample was divided by the median value for each channel which had itself been divided by the overall dataset median. Differentially abundant proteins were identified using π score as determined through a Student’s t-test with or without Welch’s correction. Gene ontology and molecular networks were created using Enrichr (49).

One way ANOVA or Fisher exact tests were used to analyze categorical data, where appropriate. Mann Whitney-U was used to compare medians. T-test were performed in unpaired and paired analyses. In all instances, statistical significance was defined as two-tailed p<0.05, unless otherwise noted.

In this study, sex was not considered as a biological variable for the multiple assessments performed. Three cohorts were analyzed: healthy controls, COVID-19 controls (non-IVIG-treated) and COVID-19 IVIG-treated patients. The 13 healthy control subjects included 9 males and 4 females, with a mean age of 41.93 (range 25-70). No significant differences were observed between COVID-19 control and COVID-19 IVIG-treated patients in terms of sex, age, BMI, or APACHEII scores. However, IVIG-treated patients had significantly longer durations of mechanical ventilation and hospital stays ( Table 1 , Supplementary Figure 1 ). Despite this, in-hospital survival was similar between the groups; 84.2% in COVID-19 controls (3 deaths out of 19 patients) and 83.3% for COVID-19 IVIG-treated patients (3 deaths out of 18 patients).

Prior studies suggest that IVIG’s benefit in COVID-19 depends on early administration. We therefore examined clinical outcomes based on the timing of IVIG initiation. Eighteen patients were treated with IVIG (5 before protocol initiation and 13 as part of the clinical trial). IVIG was typically administered within 3–5 days after intubation (mean: 1.9 days post-intubation, range 0–5 days). Median time from admission to intubation for mechanical ventilation was 4.5 days, with a range of 1–19 days. All but one patient received IVIG within 72 hours of mechanical ventilation. Two pre-protocol patients received 3 doses (total 1.5 g/kg), while all others received 4 doses (total 2g/kg). The Type 2 diabetes mellitus was the most common comorbidity (28%) followed by hypertension (22%), morbid obesity (50%, defined as BMI >40 kg/m²) ( Table 1 ).

Earlier IVIG administration was associated with improved clinical outcomes as evidenced by improved functional status at the time of hospital discharge (p=0.0175, Figure 1A ). Patients discharged home received IVIG at a median of 3 days after hospital admission (range 3-10), with a median hospital stay of 21 days. Those discharged to a rehabilitation received IVIG at a median of 7 days (range 4–11 days), with a median hospital stay of 34 days. For the 4 patients who died, IVIG was administered at a median of 10 days after admission (range 8–20 days). All patients discharged to rehabilitation survived at one year follow-up. There was no difference in age between survivors and non-survivors. Notably, among the patients discharged to a rehabilitation facility were two individuals aged ≥65-years and a 27-year-old with a BMI 90 kg/m² ( Figure 1B ). All patients who initiated IVIG therapy within ≤5 days admission survived, and 67% were discharged home. In contrast, those who received IVIG >5 days after admission, only 11% were discharged home, 44% required rehabilitation, and 44% died. This time-dependent association with mortality reached statistical significance (p=0.041, Fisher exact test).

Lung function also correlated with outcome. Five patients showed rapid improvement in oxygenation after IVIG, defined as stable O₂ saturation and ≥15% reduction in FiO₂ within 48 hours of the first dose; all were discharged home ( Figure 1C ). These patients received IVIG earlier (median 3 days, range 3-5) compared to those without improvement (median 8 days, range 3-20, p=0.0146, Mann Whitney-U test, Figure 1B ). Moreover, patients with improved SaO₂/FiO₂ ratios by day 8 all survived, while 4 of the 5 with worsened ratios died. ( Figure 1C , p=0.002, two-tailed Fisher exact test). One patient was extubated by day 4.

To determine how IVIG treatment influences the circulating proteome, we performed proteomic analysis of plasma from IVIG-treated COVID-19 patients at multiple time points, as well as from COVID-19 controls and healthy donors. First, we confirmed that disease severity was comparable between COVID-19 control and IVIG-treated groups based on similar APACHE II scores at admission ( Supplementary Figure 1 ). Figure 2A illustrates the workflow for plasma sample acquisition during hospitalization at early and late time points, including processing steps such as IgG depletion, which was necessary to reduce the dominance of exogenous IVIG in the plasma and allow for detection of other proteins. In addition, Figure 2A also outlines the full proteomic pipeline.

We then performed a principal component analysis (PCA) to visualize group differences. As expected, healthy controls clustered distinctly from all COVID-19-infected subjects ( Figure 2B ). Notably, the proteomes of patients receiving early IVIG treatment were different from those of both pre-treatment samples and early COVID-19 controls, which clustered closely together. By contrast, at later timepoints, we observed convergence in the proteomic profiles of all patient groups, with both IVIG-treated and untreated COVID-19 subjects trending closer to healthy controls ( Figure 2C ), consistent with the general clinical improvement seen by that stage.

When we restricted the PCA to only IVIG-treated patients, comparing early versus late treatment time points, we observed a clear temporal evolution in the proteomic profiles: patients sampled at later time points tend clustered closer to healthy controls than those sampled earlier in hospitalization ( Supplementary Figure 2 ). These data suggest that IVIG treatment modulates the plasma proteome early after administration, resulting in a profile that differentiates treated patients from untreated COVID-19 controls. However, it is important to highlight that IVIG was only administered for only 4 days, which may explain why some proteomic changes were not sustained at later time points.

Heatmap analyses further highlighted these trends. When comparing proteomic profiles across healthy donors, COVID-19 controls, and IVIG-treated patients, we observed considerable heterogeneity within both COVID-19 groups. Nonetheless, IVIG-treated subjects exhibited proteomic signatures that differed from both healthy donors and COVID-19 controls ( Figure 2D ). This was further supported by a heatmap analysis comparing early versus late samples from IVIG-treated patients, which showed dynamic evolution over time, although some subjects exhibited similar proteomic signatures at both time points ( Figure 2E ). This suggests that while IVIG treatment modulates the plasma proteome in many patients, the response is variable and not uniform across all individuals.

Next, we evaluated specific pathways impacted by IVIG treatment, focusing on differences relative to untreated COVID-19 controls. As a reference point, we first assessed how COVID-19 affected the plasma proteome by comparing early samples from COVID-19 controls (collected day 5 after intubation) with those from healthy donors.

Out of 239 identified proteins, 58 were downregulated and 51 upregulated in COVID-19 controls compared to healthy donors ( Supplementary Figure 3A ). Pathway analysis of these proteins revealed upregulation of inflammatory responses, coagulation and complement activation pathways, as well as disruptions in lipid metabolism ( Supplementary Figure 3B )—consistent with previous reports of COVID-19 pathophysiology (35, 37, 42, 44, 50, 51). Network analysis highlighted how certain proteins participate in multiple pathways central to COVID-19 pathogenesis, including immune signaling, complement, and coagulation ( Table 2 , Supplementary Figure 3C ).

We then compared early samples from IVIG-treated COVID-19 patients to those from early COVID-19 controls. In this comparison, 28 proteins were downregulated and 45 upregulated in the IVIG group ( Figure 3A ). Notably, the differentially modulated proteins were predominantly associated with coagulation pathways, though changes were also observed in immune and complement-related proteins, including downregulation of complement-binding proteins and factors involved in wound healing ( Figure 3B ). This pattern was further supported by molecular network analysis ( Figure 3C ). Remarkably, pathway analysis revealed that several up- and downregulated pathways were the same, likely reflecting the significantly higher or lower abundance of specific proteins within these shared pathways in IVIG-treated patients compared to COVID-19 controls ( Figure 3B ). This becomes more apparent when examining a network pathway analysis focused specifically on the coagulation pathway ( Figure 3D ). In contrast, analysis of the same network for complement activation showed that most of the proteins with significant altered abundance are downregulated in IVIG-treated COVID-19 patients relative to COVID-19 controls ( Figure 3D ). Altogether, these findings suggests that IVIG modulates the coagulation cascade through both up and downregulation of several components, whereas its effect on the complement system appears to be more uniformly suppressive.

We decided to focus on both coagulation and complement pathways, as there is substantial evidence of the involvement of both in determining clinical outcomes in COVID-19 (35, 37, 42, 44, 50, 51). For both pathways, we analyzed the effects of IVIG treatment at early ( Figures 4A–C ) and later ( Figures 4D–F ) time points, which is relevant since IVIG was administered only over a 3–4 days course. At early time points following IVIG treatment, we found that patients upregulated 10 proteins and downregulated 8 proteins relative to COVID-19 controls ( Figure 4A ). A heatmap including the pre-treatment group showed some variability within each cohort, although early IVIG-treated patients tend to cluster with certain pre-treatment samples ( Figure 4B ). To better visualize the abundance differences in coagulation-related proteins, we constructed radar plots using healthy donors as a reference and recognized protein vertices based on hierarchical clustering. This approach more clearly highlighted group-level differences.

The radar plots showed that early after treatment, IVIG-treated patients exhibited increased levels of TNL1, FLNA, ACTB, and VCL, and decreased levels of FGG, FGB, FGA, VWF, F13B, FBLN1, CBP2, and KNG1 ( Figure 4C ). Interestingly, the relative abundance of FGG, FGB, FGA, F5, and F9 in IVIG-treated patients approximated levels found in healthy controls ( Figure 4C ). These trends were further illustrated in Supplementary Figure 4A , which compared relative abundance against pre-treatment values and reflected the same findings observed in the radar plots. Table 2 details the differentially modulated coagulation-related proteins and their known pro- or anti-coagulant functions. Collectively, these results suggest that IVIG treatment modulates key coagulation-related proteins early in the disease course, in some cases restoring levels toward those observed in healthy controls.

When we analyzed samples collected at later time points (8–14 days post-treatment), we found 9 upregulated and 7 proteins downregulated in IVIG-treated patients relative to COVID-19 controls ( Figure 4D ). The heatmap analysis showed that coagulation-related protein abundances in IVIG-treated patients at this stage tended to cluster both with each other and with pre-treatment samples ( Figure 4E ). Similar to our observations at the early time point, the radar plot at later time points showed increased levels of TLN1, ACTB, FLNA, F5, VCL, and THBS1, and decreased levels of KRT1, F13B, CPB2, VWF, and SERPINA1 ( Figure 4F , Table 2 ). When analyzing these profiles in the context of pre-treatment levels (supplemental Figure 4B ), we found that IVIG-treated patients at later time points had relative abundances that remained closer to their pre-treatment state. Overall, these findings suggest that the impact of IVIG on coagulation-related proteins is most pronounced shortly after treatment, with a partial return to baseline levels observed at later time points. While significant modulation persists in some proteins, the overall magnitude of change diminishes over time.

For complement-related proteins, we found that IVIG-treated patients have the abundance of one protein increased and three proteins decreased compared to COVID-19 controls at early post-treatment time points ( Figure 5A ). Heatmap analysis showed high variability within each treatment group ( Figure 5B ). However, radar plots revealed a clear decrease in the average abundance of several complement proteins in IVIG-treated patients compared to both COVID-19 controls and healthy donors, including C1RL, CFD, CD5L, C8G, and CFHR5 ( Figure 5C , Table 2 ). Similar findings were observed when comparing against pre-treatment samples: protein levels in IVIG-treated patients were lower than in both COVID-19 controls and pre-treatment samples ( Supplementary Figure 5A ).

At later time points following treatment, two complement proteins were increased and one was decreased in IVIG-treated patients ( Figure 5D ). Heatmap analysis again did not reveal a clear clustering pattern associated with treatment ( Figure 5E ). Unlike early time points, the radar plot shows similar average protein abundances across IVIG-treated subjects, COVID-19 controls, and healthy donors, with CD5L as the only protein showing a significant difference compared to COVID-19 controls ( Figure 5F ). This trend was also evident when comparing relative abundances, although the magnitude of differences was lower than at early time points ( Supplementary Figure 5B ). Interestingly, at early time-points, the complement protein profile of IVIG-treated patients resembles their pre-treatment baseline more closely than that of COVID-19 controls. These results suggest that IVIG modulates complement-related proteins transiently after treatment, with effects that are not sustained at later stages.

Since the most pronounced changes following IVIG treatment occurred early, likely due to the short 3–4 days treatment course—we focused on paired analyses comparing pre-treatment and early post-treatment samples. In the coagulation pathway, we found that IVIG treatment decreases KNG1 and ACTB and increased FGA relative to COVID-19 controls ( Figure 6A ), although the variance in FGA levels was similar between groups. Additionally, F13B and CPB2 increased in COVID-19 controls but not in IVIG-treated patients. No significant changes were observed for FGB, FGG, VCL, FBLN1, VWF, TLN1, and FLNA at the individual level ( Figure 6A ), though some of these proteins were significantly reduced in unpaired group-level analyses ( Figure 5C , Supplementary Figure 4A ).

Longitudinal analysis of relative protein abundance across pre-treatment, early, and late time points showed that reductions in coagulation-associated proteins were most evident early after IVIG treatment and largely returned to baseline by the later time point ( Figure 6B ). However, the magnitude and direction of the response varied between individuals. These data suggest that IVIG transiently reduces proteins known to promote coagulation ( Table 2 ), though this effect may not persist into later stages of illness.

In the complement pathway, IVIG treatment decreased levels of C1RL, C8G, and CFD compared to COVID-19 controls ( Figure 7A ). These proteins are known to promote complement cascade activation ( Table 2 ). In contrast, CD5L and CFHR5 did not differ significantly between groups ( Figure 7A ). As with the coagulation proteins, longitudinal analysis showed that complement-related proteins were reduced shortly after IVIG treatment, but these effects were not sustained at later time points ( Figure 7B ). The pattern varied among patients, again suggesting a transient and heterogeneous effect of IVIG on complement activation.