The COVID-19 pandemic began in December 2019 with the emergence of a novel member of the coronavirus family, SARS-CoV-2. As of July 14, 2024, more than 778 million people have been infected globally, resulting in more than seven million deaths worldwide (World Health Organization, 2024). The infection spans the spectrum from asymptomatic to severe respiratory disease, leading to hospitalization, admission to the ICU and death (Ziehr et al., 2020; Qi et al., 2021; Nanwani-Nanwani et al., 2022). The devastating impact of COVID-19 has led to intense interest in the host’s defense against the infection, involving organ-specific defense and innate and adaptive immunity, including natural killer (NK) cells.

Natural killer (NK) cells are innate lymphoid cells that serve as one of the first lines of defense against tumor cells and microbial pathogens, including bacteria, fungi, parasites, and viruses (Abel et al., 2018; Garcia-Peñarrubia et al., 1989; Kiessling et al., 1975; Lodoen and Lanier, 2006; Schmidt et al., 2013). Unlike other cytotoxic lymphocytes, natural killer cells do not require previous exposure to recognize and deploy their cytotoxic products to attack and kill pathogens, pathogen-infected or malignantly transformed cells (Kiessling et al., 1975). Viruses and tumors can cause a reduction in inhibitory MHC-I molecules on the target cells, making them susceptible to NK cell targeting through activating receptors (See et al., 1997).

Some viruses stimulate NK cell activity, while others suppress their function, leading to contrasting effects on host defense. For instance, the human immunodeficiency virus and respiratory syncytial virus have been shown to enhance NK cell activity (Richard et al., 2010; van Erp et al., 2019). By contrast, other viruses, such as influenza A, human pegivirus, and vaccinia, reduce NK cell function in infected individuals (Kirwan et al., 2006; Mao et al., 2010; Chivero et al., 2015). Different viruses can manipulate various pathways of activation and inhibition in NK cells. For example, human cytomegalovirus and human T lymphotropic virus type 1 enhance NK cell function through augmented degranulation and increased IFN-γ production. However, they can also inhibit NK cell function by altering receptor expression, including checkpoint proteins (Béziat et al., 2013; Tschan-Plessl et al., 2016; Amorim et al., 2019). Consequently, it is necessary to understand the sequence of events during NK cell function required to predict the effect of viruses under different conditions.

When an NK cell encounters a tumor or virus-infected cell, it kills the target cell by releasing cytotoxic effector molecules. These molecules, such as perforin, granulysin, and granzymes, are released from the NK cells’ granules, also called vesicles or secretory lysosomes. This degranulation event requires the receptor activation signal to be stronger than the inhibitory signal. This triggers cytoskeletal rearrangement, which is necessary for the proper trafficking of vesicles within the NK cells toward the immunological synapse, where they can release their cytotoxic proteins.

Perforin damages or creates pores in the target cell membrane, allowing the entry of granzymes (Lieberman, 2003; Trapani and Bird, 2008). Additionally, granulysin associates with the plasma membrane of tumor cells, causing mitochondrial damage and calcium overload, leading to cell death (Okada et al., 2003). Granzymes, including Granzyme A, B and K, are a family of enzymes that induce caspase-dependent cell death (Beresford et al., 2001; Chowdhury and Lieberman, 2008; Martinvalet et al., 2005; Zhao et al., 2007). While studies have described the intracellular levels of granzymes A, B and perforin in patients with COVID-19, it still needs to be clarified whether the extracellular concentration of these molecules is altered upon stimulation.

NK cells are dysfunctional in severe COVID-19. In hospitalized patients with severe COVID-19 (grade 6 to 9 on the WHO progression scale), the number of NK cells is reduced, and the expression of IFN-stimulated genes and genes involved in IFN-α signaling is increased, leading to defective NK cell cytotoxicity (Marshall et al., 2020; Zheng et al., 2020; Krämer et al., 2021). IFN-α stimulates NK cell function and can downregulate IFN-γ production (Krämer et al., 2021; Lee et al., 2019). When considering intracellular effector molecule content, there is less consensus about whether granzyme B and perforin expression increase or decrease (Maucourant et al., 2020; Zheng et al., 2020; Witkowski et al., 2021). There is also an impairment in cytotoxic degranulation assessed by CD107a presentation (Krämer et al., 2021; Witkowski et al., 2021).

Independent studies have looked at limited panels of activating and inhibitory receptors, and the conclusions drawn so far show an exhausted phenotype based on the expression of NKG2D, NKG2A, PD-1, CTLA-4, and TIM-3 (Zheng et al., 2020; Varchetta et al., 2021; Ustiuzhanina et al., 2023). Transcriptional analysis has proposed that these defects are due to a transforming growth factor-β (TGF-β) signature leading to reduced cell-cell adhesion, granule exocytosis and cytotoxicity (Witkowski et al., 2021). However, a systematic assessment of the sequence of intracellular events leading to NK cell cytotoxicity and the expression of inhibitory receptors as possible mechanisms of defective cytotoxicity in patients hospitalized with COVID-19 (PHWC-19) has not been performed.

Demographic and clinical factors can affect NK cell intracellular mechanisms. Sex differences are present across many aspects of the immune system and are essential to understand and treat infections and immunologic disorders fully. Males are more likely to become hospitalized when infected with COVID-19 and comprise the majority of all age and ethnic groups in hospitalized patients (Qi et al., 2021). There are differences in receptor and cytokine expression, which could be the basis for sex differences in the immune response to SARS-CoV-2. Despite the relationship between sex differences and the severity of COVID-19, there is a lack of information about NK cell function in males and females hospitalized with COVID-19.

The ancestral strain of SARS-CoV-2 predates the subsequent variants of concern (VOCs) that evolved later in the pandemic. These variants, including Alpha, Beta, Gamma, Delta, and Omicron, are progressively more transmissible due to genetic adaptation and external pressures such as vaccine-induced and natural immunity (Carabelli et al., 2023; Jacobs et al., 2023). It has been shown that the CD56^dim CD16^bright NK cells from patients infected with Omicron BA.2 demonstrated greater IFN-γ secretion and cytotoxicity in previously vaccinated individuals compared to unvaccinated controls (Peng et al., 2023). However, there was no comparison of NK cell functioning in patients infected with earlier VOCs.

Dexamethasone, a corticosteroid, was introduced as an anti-inflammatory medication to reduce the progression of organ injury and failure that could result in death (The Recovery Collaborative Group, 2021). Moderate administration of dexamethasone benefits PHWC-19 with respiratory support (supplemental oxygen or ventilation), and no advantages have been found in patients without respiratory aids (The Recovery Collaborative Group, 2021). There have been studies regarding the effects of corticosteroids, with some studies showing an enhancement of proliferation and cytokine production of NK cells while others showing reduced cytotoxicity (Piccolella et al., 1986; Morgan and Davis, 2017; Chen et al., 2018). The role of dexamethasone as an independent variable on NK cell function during COVID-19 infection has not been investigated. One study found no difference in the cytotoxicity of NK cells from patients before and after the establishment of dexamethasone as a standard treatment for COVID-19, but did not investigate further (Witkowski et al., 2021).

We utilized antibody labelling and mass cytometry to determine the percentages of different blood NK cell subsets and other lymphocytes reduced in PHWC-19. NK cells from PHWC-19 were co-cultured with a tumor cell target to compare cytotoxicity with NK cells from healthy subjects. This led to our investigation of the mechanisms involved in NK cell cytotoxicity. Using flow and mass cytometry, we found that NK cells from PHWC-19 had reduced effector molecule expression, degranulation, and release using flow cytometry and ELISA. We used live cell imaging to establish that granule trafficking and polarization were defective in NK cells from PHWC-19, which is critical for granule release and effectiveness. We used a panel of antibodies of activating and inhibitory receptors to show that inhibitory receptor expression was enhanced, which could affect downstream granule trafficking, granule content, and degranulation. Lastly, we explored the role of demographic and clinical factors like age, sex, VOCs, prior vaccination and dexamethasone treatment on NK cell functioning to understand how NK cells are affected by COVID-19.

Peripheral blood samples were collected from forty-four patients hospitalized in a medical ward at Foothills Medical Centre, Calgary, Canada, after informed consent was obtained between November 2020 and September 2023. The patients were categorized within the moderate range of 4 or 5 on the WHO clinical progression scale for COVID-19 disease severity (Marshall et al., 2020). These patients presented with clinical features requiring hospitalization but were not admitted to the intensive care unit. Patients were excluded if they had major comorbidities that might affect study outcomes, were immunosuppressed, including having an active malignancy or receiving immunosuppressive drugs. Characteristics such as age, sex, vaccination status, dexamethasone treatment, and variants at the time of hospitalization are summarized in Table 1 . When patient samples were obtained, peripheral blood samples were collected on the same day from forty-five healthy donors at the Cumming School of Medicine to act as the paired negative controls.

Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood using the Ficoll-Paque^® separation technique (Millipore Sigma, Germany). PBMCs were used for flow cytometry studies, while primary NK cells (pNK) were isolated by negative selection using magnetic beads (Miltenyi Biotec, Germany). Three mL of whole blood was put aside for mass cytometry.

NK cell cytotoxicity was assessed using 721.221 tumor cells (a gift from Dr. Nicole Bernard, McGill University, Montreal). The 721.221 cells were cultured in RPMI 1640 (Gibco, Thermo Fisher Scientific Inc, Waltham, MA) with 10% fetal bovine serum (FBS) and 1% Penicillin-Streptomycin-Glutamine (PSQ). The 721.221 cells were stained with the carboxyfluorescein succinimidyl ester (CFSE) (Millipore Sigma, Germany) as a cell tracking dye to distinguish them from the primary cells. The cells were co-cultured with NK cells in RPMI 1640 with 10% FBS, 1% non-essential amino acids, 1% Na pyruvate, 1% PSQ at 5:1 and 10:1 effector-to-target (E:T) ratios for 4 hours before adding 7-amino actinomycin D (7AAD) (Invitrogen, Thermo Fisher Scientific Inc, Waltham) and analyzed by flow cytometry (Guava EasyCyte, EMD Millipore, Danvers, MA). The positive control for cell death was obtained by fixing 721.221 cells with 4% paraformaldehyde (PFA) followed by Triton-X100 (0.1% in PBS). The percentage of cell death is analyzed using the FlowJo LLC program by examining the percentage of cells that are double positive for the CFSE and 7AAD stains.

PBMCs were fixed with 4% PFA and permeabilized with saponin (0.1%). Intracellular effector proteins granulysin, perforin, and granzymes A, B, and K, and checkpoint proteins LAG-3, PD-1, TIGIT, and CD94 were labelled with fluorescent antibodies or their respective isotype control antibodies for 30 minutes along with anti-CD56 and anti-CD3 to define NK cells ( Supplementary Table 1 ). Geometric mean fluorescence intensity was assessed by flow cytometry (LSRII, BD Biosciences, New Jersey). The healthy and COVID-19 samples were collected and processed simultaneously to limit interexperimental variability. Initially, PBMCs were gated on CD56+ CD3- cell population for NK cells. Histograms were then utilized to determine the percentage of NK cells positive for each marker. Flow cytometry data was analyzed using the FlowJo LLC program.

NK cell degranulation was assessed as previously described (Islam et al., 2013). The Anti-CD107a fluorescent antibody and the intracellular protein transport inhibitor cocktail Brefeldin A/Monensin were added to the NK cell plus 721.221 cell co-culture for four hours. Cells were fixed with 4% PFA, labelled with CD56 and CD3, and analyzed by flow cytometry (Guava EasyCyte, EMD Millipore, Danvers, MA). These experiments were conducted concurrently in the healthy and COVID-19 samples to pair the results.

Supernatants from the co-cultures of NK and tumor cells were collected at 4 hours and stored at -80°C prior to measuring levels of released perforin, granzyme A and granzyme B by ELISA. Effector molecule release was measured using the manufacturer’s reagents and instructions (Abcam, Cambridge).

Live-cell imaging of primary NK cells in the presence of 721.221 target cells was used to assess spatial and temporal changes in conjugate formation and granule trafficking. The NK cells were treated with Cresyl violet to label granules (Millipore Sigma, Germany) and tubulin tracker green to label microtubules (Invitrogen, Thermo Fisher Scientific Inc, Waltham, MA). A Ti Eclipse microscope (Nikon Instruments Inc., Melville, NY) was used with an objective lens with a magnification of 60X, a numerical aperture of 1.2 and oil immersion. Images were analyzed using the ImageJ Fiji program following the protocol described previously to measure granule distances to the microtubule-organizing center (MTOC) and the synapse (Hsu et al., 2017). The synapse is defined as the center of the area of contact between the NK cell and the 721.221 tumor cell. The width of the synapse was determined when the plasma membrane of the NK cell and the tumor cell was deformed by the contact and defined as the linear distance between the points of contact of both cells’ plasma membranes. The granule distance is defined as the mean distance from the granules to the closest point on the immunological synapse. Distances of the MTOC to synapse and granule to synapse were determined over time using the Trackmate plugin in ImageJ (Ershov et al., 2022). Imaging in the healthy donor was performed at a frame rate of 1 image every minute for 60 minutes, and imaging in the PHWC-19 was performed at a frame rate of 1 image every minute for 80 minutes. The time to initial polarization was defined as the time between the start of a sustained contact between the NK cell and the tumor cell and the initial decrease in distance of the granules to the synapse.

The whole blood was diluted at a ratio of 1:1.4 of blood to the PROT1 stabilizer for ten minutes and stored at -80°C until the samples were prepared for mass cytometry (HELIOS, Standard BioTools Inc, San Francisco). The cells were thawed, and the red blood cells were lysed using the PROT1 lysis buffer and washed with a cell staining media (CSM) consisting of phosphate-buffered saline (PBS) (Gibco, Thermo Fisher Scientific Inc, Waltham, MA) with 1% bovine serum albumin (BSA). The samples were fixed, permeabilized and barcoded with the Cell-ID 20-Plex Pd Barcoding Kit (Standard BioTools Inc, San Francisco) according to the manufacturer’s recommendations. After barcoding, the cells were adjusted to have 5 x 10⁵ cells per sample for 3 x 10⁶ cells per tube. Cells were washed, and samples were combined and stained with a surface staining cocktail of lanthanide-labelled antibodies ( Supplementary Table 2 ) for 30 mins at room temperature. Cells were reconstituted in 0.3% Saponin/PBS with iridium and 1.6% paraformaldehyde in the fridge overnight. The next day, the cells were washed with CSM and PBS before being aspirated into a pellet and brought to the HELIOS machine for acquisition. All the data was uploaded to the Cytobank program to generate all gates and run all analysis methods, including viSNE for dimensionality reduction and FlowSom for clustering cells. GraphPad Prism (GraphPad Software Inc, CA) was used to conduct the PCA analysis of the data from mass cytometry. Gating strategies for cell populations have been provided in Supplementary Figure 1 .

Statistical analysis was performed using GraphPad Prism (GraphPad Software Inc, CA). A paired t-test was used to assess cytotoxicity and degranulation after ensuring normal distribution based on a Shapiro-Wilk test. The remaining analysis was done with Mann-Whitney, Wilcoxon matched-pairs signed rank, or Kruskal-Wallis ANOVA based on paired or unpaired samples and the number of comparisons. P<0.05 was considered statistically significant.

NK cells play an essential role in the innate immune system during many respiratory diseases. These cells can kill tumors, virus-infected cells, and bacterial and fungal pathogens (Garcia-Peñarrubia et al., 1989; Lodoen and Lanier, 2006; Schmidt et al., 2013). Since the number of NK cells could affect the severity and degree of immune dysfunction in PHWC-19, the peripheral whole blood of PHWC-19 was analyzed using mass cytometry. We found a significant decrease in the percentage of NK cells and, indeed, all mononuclear cells in PHWC-19 compared to healthy donors ( Figure 1A , gating strategy in Supplementary Figure 1 ). Using flow cytometry, we found the percentage of CD3- CD56+ NK cells in PHWC-19 was lower compared to healthy controls processed on the same day (9.8 ± 3.8% in PHWC-19, 22.8 ± 8.0% in healthy subjects) ( Figure 1B ). We further assessed different NK cell subsets based on the expression of CD56 and CD57. We found that the mature, differentiated, and cytotoxic NK cell subset, CD56^dim CD57+, was decreased in PHWC-19, while the immature cytotoxic NK cells, CD56^dim CD57- increased ( Figure 1C ) (Fan et al., 2008; Vivier et al., 2008). The less abundant cytokine-producing CD56^bright CD57+/- population remained unaltered due to COVID-19 ( Figure 1C ) (Lopez-Vergès et al., 2010; Béziat et al., 2011). Although the reduced numbers and maturity of NK cells may contribute to the observed reduction in cytotoxicity, we were interested in determining whether there is a defect at the cellular level of NK cell function.

To test the cytotoxic ability of NK cells from PHWC-19, we used an established method of tumor target killing (Epling-Burnette et al., 2007; Lisovsky et al., 2015; MacFarlane et al., 2017). NK cells were co-cultured with the 721.221 cell line, and killing was assessed using 7-AAD uptake and flow cytometry. There was a decrease in 721.221 cell death in response to NK cells from PHWC-19 at E:T ratios of 5:1 and 10:1 compared to healthy control subjects ( Figure 2A ).

Since the mean age of PHWC-19 was older than that of our healthy control subjects, and aging has been reported to affect some aspects of immune cell function, we tested the effect of age on NK cell cytotoxicity (Solana and Mariani, 2000; Solana et al., 2006). Linear regression analysis of the healthy donors showed no decline in the percentage of 721.221 cells killed when co-cultured with NK cells from older healthy subjects ( Figure 2B ). Surprisingly, there was a significant increase in the percentage of 721.221 cells killed with increasing age of the PHWC-19 ( Figure 2B ). Although there were fewer young PHWC-19, results were significant, revealing that COVID-19 caused a defect in cytotoxicity that was greater in younger patients than in older patients who required hospitalization.

After establishing that cytotoxicity was reduced in NK cells from PHWC-19, we were interested in the sequence of intracellular events that might account for the reduced cytotoxicity. Effector molecules are preformed and stored in granules before their release. Granzymes A, B, and K are serine proteases, while perforin and granulysin are membrane-disrupting or pore-forming proteins involved in the defense against tumors and viruses (Krensky and Clayberger, 2009; van Domselaar and Bovenschen, 2011; Thiery and Lieberman, 2014). Using intracellular antibody labelling and flow cytometry, we compared the relative abundance of intracellular granzymes A, B, K, perforin and granulysin in NK cells from PHWC-19 and healthy adults. We found that only the intracellular levels of granzyme B were significantly decreased in the NK cells of PHWC-19 compared to those from healthy subjects ( Figure 3A ), and the fluorescent intensity of granzyme B correlated with cytotoxicity ( Supplementary Figure 2A ), revealing that the level of granzyme B was important for cytotoxicity. Intracellular granzyme A showed a lower trend in NK cells from PHWC-19 but did not reach statistical significance using a Wilcoxon matched-pairs signed rank test ( Figure 3A ).

In contrast, granzyme K showed no significant differences between the healthy and COVID-19 groups ( Figure 3A ). Among the pore-forming effector molecules, intracellular perforin and granulysin were not significantly different between the healthy and COVID-19 samples ( Figure 3A ). The intracellular content of granzymes and pore-forming proteins reflects the stores available. However, since only a fraction of the stores are utilized during each round of killing, the effect would potentially manifest only after multiple rounds of killing, leading to the depletion of the stores (Gwalani and Orange, 2018). Because of this, we were interested in whether degranulation was affected.

Degranulation occurs when the effector-containing granules fuse with the NK cell plasma membrane and release their contents. To assess degranulation in NK cells from patients hospitalized with COVID-19, flow cytometry was used to detect the cell surface membrane expression of CD107a (LAMP1) during the killing of tumor cells. There was a significantly lower expression of CD107a after co-culture of NK cells from PHWC-19 with 721.221 cells, and the fluorescent intensity of CD107a correlated with cytotoxicity ( Supplementary Figure 2B ), revealing the importance of reduced degranulation of NK cells from hospitalized SARS-CoV-2 infected patients compared to healthy NK cells ( Figure 3B ).

The release of effector proteins during degranulation increases the concentration of effectors in the extracellular compartment. Since degranulation was reduced in NK cells from patients hospitalized with COVID-19, we then determined if there was a difference in the amount of effector molecules released. The supernatants of the co-cultures were collected, and the effector molecules, granzyme A, B, and perforin, were measured by ELISA. There was a significant decrease in the concentration of granzymes A and B released from the NK cells obtained from PHWC-19 ( Figure 3C ). Meanwhile, extracellular concentrations of perforin were not significantly different between the healthy and COVID-19 cohorts ( Figure 3C ). This data reveals that intracellular stores of critical effector molecules were reduced, and degranulation and the release of granzymes were impaired in NK cells from PHWC-19, which was associated with reduced cytotoxicity.

To release the effector molecules at the target cell interface, the intracellular cytolytic granules must move along the microtubules and reach the area of contact with the target cell. We sought to determine if there was impaired trafficking of the granules and the microtubule-organizing center (MTOC). Granule trafficking requires the granules to converge on the MTOC via microtubule motor proteins and for the converged granules and MTOC to polarize to the NK cell synapse with the target cell (Ogbomo et al., 2018). Since enveloped viruses disrupt many forms of cellular transport (Hassan et al., 2021), we asked whether the defect in granule release was due to deficiencies in trafficking or failure to reach the plasma membrane. To examine granule movement, we used live-cell imaging of NK cells from healthy subjects and PHWC-19 in the presence of 721.221 target cells. NK cells from healthy subjects rapidly mobilized the granules and MTOC to be adjacent to the tumor cell, indicating successful granule trafficking ( Figure 4A , Supplementary Video 1A ). By contrast, the cytolytic granules and MTOC in NK cells from PHWC-19 did not polarize to the synapse with the target cell, indicating failed granule trafficking, which would lead to defective granule release and cytotoxicity ( Figure 4A , Supplementary Video 1B ). The imaging of NK cells from hospitalized patients with COVID-19 was continued for an additional 20 minutes to capture any events that may have taken longer to occur. Quantifying the mean granule-to-synapse distance indicated decreased localization of granules towards the synapse in NK cells from PHWC-19 ( Figure 4B ), although the time to initial polarization was similar in healthy and PHWC-19 ( Supplementary Figure 3 ). Similarly, reduced localization of the MTOC was also seen in the NK cells of PHWC-19 ( Figure 4C ).

NK cells from healthy subjects formed a broad, concave base of attachment between the NK cell and the tumor cell. By contrast, the NK cells from PHWC-19 formed a synapse with the tumor cells, but the synapse was not as intimate as the synapse formed by NK cells from healthy control subjects, and the width of the synapse was greater in NK cells from healthy control subjects than in NK cells from PHWC-19 ( Figures 4A, D , Supplementary Video 1B ). Thus, NK cells from PHWC-19 demonstrated impaired granule trafficking and a less intimate synapse formation that would contribute to defective cytotoxicity.

Because of the suboptimal conjugate formation and granule polarization in NK cells from PHWC-19, we hypothesized that there might be changes in inhibitory receptor expression, which could affect granule trafficking and degranulation. The inhibitory receptors include a subset called checkpoint inhibitory proteins. These proteins regulate the NK cell function by reducing their cytolytic activity. We performed experiments using samples of NK cells obtained from healthy and PHWC-19 on the same day to assess PD-1, LAG-3, TIGIT, and CD94. Only PD-1 was significantly increased in NK cells from PHWC-19 ( Figure 5A ). Meanwhile, LAG-3 expression was reduced in PHWC-19, and TIGIT and CD94 showed no difference between the healthy and COVID-19 groups ( Figure 5A ). Since different cell subsets are present, we examined the expression patterns of inhibitory and activating receptors on different NK cell subsets defined by CD56 and CD57 expression. Using mass cytometry, we studied the expression of the inhibitory and checkpoint receptors PD-1, LAG-3, TIGIT, CTLA4, TIM-3, Siglec 9, NKG2A, NKB1, CD161, CD158b, CD47 and the activating receptor NKG2D on these subsets. The differential expression of each receptor within the subsets was visualized via a log2 heatmap ( Figure 5B ). The more mature and most cytotoxic CD56^dim CD57+ NK cells in PHWC-19 showed a higher expression of PD-1, CTLA-4, and CD158b and a decrease in CD47 compared to healthy controls, suggesting that the net effect is to dampen cytotoxicity ( Figure 5C ). Similarly, the more abundant and slightly less cytotoxic CD56^dim CD57- NK cells showed a significant upregulation of PD-1, LAG3, CTLA4, and CD158b, along with a downregulation of NKG2D, NKG2A, and CD161 in PHWC-19 compared to healthy controls. This also suggests a net negative effect ( Figure 5C ). Overall, these cytotoxic NK cell populations exhibit an increased expression of inhibitory receptors in PHWC-19, which is consistent with reduced killing. In the less prevalent CD56^bright CD57- NK cells, which are associated with cytokine production and lower cytotoxicity, there was an increased expression of PD-1, CTLA4, CD158b, and decreased expression of NKG2D, CD161, and CD47 in PHWC-19 ( Figure 5C ), revealing a receptor profile consistent with a negative effect on signaling.

We identified major clusters within CD56^dim CD57- and CD56^dim CD57+ NK cells based on the co-expression of activating and inhibitory receptors ( Figure 6A ). This aided us in identifying alterations in NK cells with specific receptor co-expression between healthy donors and PHWC-19. The major clusters for CD56dim CD57- NK cells were 1, 4, 7, 9, and 10 ( Figure 6B ). Among these, cluster 4 was significantly increased in PHWC-19, while cluster 7 was significantly decreased in PHWC-19 ( Figure 6B ). Cluster 4 of CD56^dim CD57- NK cells was characterized by higher co-expression of CTLA4, CD161, and CD47, while cluster 7 showed higher co-expression of NKG2D, CTLA4, and CD161 ( Figure 6C ). Dendrogram clustering revealed three major groups of receptors co-expressed on CD56^dim CD57- NK cells, which showed similarities between healthy and PHWC-19 groups ( Figure 6C ).

The major clusters for CD56dim CD57+ NK cells were clusters 1, 9, and 10 ( Figure 6B ). Cluster 9 was significantly decreased in PHWC-19 ( Figure 6B ). This cluster showed higher co-expression of NKG2D, TIGIT, CTLA4, CD161, and CD47 ( Figure 6C ). Dendrogram clustering revealed four major groups of receptors co-expressed on CD56^dim CD57+ NK cells, which were similar between healthy and PHWC-19 donors ( Figure 6C ). We also investigated the clusters within CD56^bright CD57- and CD56^bright CD57+ NK cells, but these populations did not show major differences between healthy donors and PHWC-19 ( Supplementary Figures 4A–C ). The dendrograms of CD56^bright CD57- and CD56^bright CD57+ NK cells showed similar groupings in the healthy and PHWC-19 groups ( Supplementary Figure 4C ). Additional PCA analysis was performed on CD56^bright and ^dim and CD57+ and – subsets ( Supplementary Figure 5 ). These results reveal that specific combinations of inhibitory receptors cluster into major subsets and that these clusters differ in their expression in PHWC-19 compared to control subjects.

After establishing the intracellular mechanistic defects responsible for impaired cytotoxicity in PHWC-19, we wondered how other biological, therapeutic and virologic factors might influence the defect. These characteristics include the different variants of the SARS-CoV-2 virus, the treatment of patients with dexamethasone, and the sex-related differences. The SARS-CoV-2 virus underwent genetic mutations, producing variants of concern. The severity of the disease and immune responses differed depending on the variant causing the infection (Lin et al., 2021; Arabi et al., 2023). The major variants of concern sequentially evolved from Alpha to Beta, Gamma, Delta, and Omicron. These variants have shown increased transmissibility, decreased virulence, and antigenicity (defined by antibody binding) (Alefishat et al., 2022). We sought to determine the cytotoxicity of NK cells obtained from PHWC-19 infected with different SARS-CoV-2 variants compared to healthy control subjects. Unfortunately, the sample size was small, and while the differences in the degree of impaired NK cell killing did not reach statistical significance, it is worth noting that there is a trend suggesting that the reduction in NK cell cytotoxicity was greater with the Alpha variant compared to NK cells from PHWC-19 infected with Delta and Omicron ( Figure 7A ), which correlates with the lack of effective host defense and worsened pathology in patients infected with earlier variants of concern.

During the transition from the time when the Alpha variant was dominant to the time when the Delta variant was dominant, dexamethasone was routinely added to therapy for the management of hypoxemic hospitalized patients (The Recovery Collaborative Group, 2021). It is possible that dexamethasone diminished the function and the number of NK cells (Chen et al., 2018). To address this possibility, we performed experiments to determine whether dexamethasone affected cytotoxicity in PHWC-19. In these experiments, multiple comparisons were made. Control PHWC-19 who were enrolled before the use of dexamethasone or where the care team chose not to use it served as the comparator for PHWC-19 that received dexamethasone, and NK cells from healthy subjects were compared to their NK cells treated in vitro with dexamethasone. Previous studies suggested that the serum concentration after a dose of 6mg would be 0.1μM (Weijtens et al., 1998; Spoorenberg et al., 2014; Morgan and Davis, 2017). To evaluate the effect of the equivalent of 6mg dexamethasone on NK cells in vivo, NK cells from healthy control subjects were pretreated with 0.1 μM in vitro. There was no significant reduction in NK cell cytotoxicity between the NK cells from PHWC-19 treated or not treated with dexamethasone, and no reduction in cytotoxicity among healthy subjects treated with dexamethasone compared to the control in vitro ( Figure 7B ).

Differences in biological sex influence many aspects of the immune system, and we are beginning to learn their effect on infections. Males were more likely to become hospitalized when infected with SARS-CoV-2, and males comprise the majority of all age and ethnic groups in hospitalized patients (Qi et al., 2021). To determine whether biological sex affected NK cell cytotoxicity, the responses from NK cells from male PHWC-19 were compared to those of healthy male control subjects, and responses of NK cells from female PHWC-19 were compared to those of healthy female control subjects. There was a significant decrease in the killing of NK cells in male PHWC-19 compared to the healthy male subjects ( Figure 7C ). By contrast, we did not detect a difference in NK cell cytotoxicity in female PHWC-19 compared to healthy control females.

A multivariate analysis was performed to examine the influence of age, healthy vs PHWC-19, the influence of dexamethasone, and sex. The regression formulation for an E:T ratio of 10:1 was Cytotoxicity = 31.898 + 0.340*Age + 22.722*Status (healthy) + 9.143*Dexamethasone - 14.046*Sex (males). Where P = 0.010 (Age), 0.001 (Status-healthy), 0.015 (Sex-Male).

To determine whether vaccine status affected NK cell function in patients who had been hospitalized with COVID-19, the cytotoxicity was compared between NK cells from PHWC-19 who had been vaccinated and those who had not. This demonstrated no difference in the cytotoxicity ( Supplementary Figure 6 ), suggesting that once hospitalized with COVID-19, vaccine status did not influence the reduction in NK cell function.

Having shown that the defects in cytotoxicity resulting from COVID-19 disease were correlated with the male but not female sex, we were interested in exploring whether the intracellular defects also exhibited a similar pattern of sex-related differences. Unfortunately, the sample size was small, and while trends emerged, the intracellular expression of effector molecules, degranulation or extracellular free effector molecules released into the supernatant failed to demonstrate statistically significant differences between males and females ( Supplementary Figures 7A–C ).

We also assessed the expression of activating and inhibitory receptors in male versus female participants to determine whether there were differences. NK cells from male PHWC-19 showed an increased expression of PD-1 compared to healthy subjects, while females did not ( Figure 8A ). By contrast, there were no differences in TIGIT and CD94 expression in males or females ( Figure 8A ).

Further examination of the NK cell subsets revealed similar changes in males and females, including upregulation of PD-1, CTLA4, and CD158b, compared to sex-matched controls for the CD56^dim CD57- NK cells ( Figure 8B ). Meanwhile, CD56^dim CD57+ NK cells from male PHWC-19 showed a higher expression of CTLA4 and CD158b and NK cells from female PHWC-19 had a higher expression of CTLA4, which was balanced by a lowered CD47 expression ( Figure 8B ). Lastly, in the CD56^bright CD57- NK cells, male PHWC-19 exhibited more of an inhibitory phenotype with increased CD158b and decreased NKG2D expression ( Figure 8B ). In females, more inhibitory receptors are expressed in the healthy donors or evenly expressed between the healthy and PHWC-19 cohorts ( Figure 8B ). In summary, there were significant differences in the male patients, including reduced activating receptor expression and increased inhibitory receptors.

Two studies have shown the sexually dimorphic nature of dexamethasone treatment that is more beneficial in men than in women (The Recovery Collaborative Group, 2021; Sinha et al., 2022). We were interested in whether this effect extended beyond neutrophils and involved NK cells. While there was no difference at a 5:1 ratio, at the 10:1 ratio, males who did not receive dexamethasone had a trend towards a more pronounced NK cell cytotoxicity defect than male PHWC-19 who received dexamethasone ( Supplementary Figure 8 ). While dexamethasone treatment may not show overall effects on NK cell function, dexamethasone may have a role in recovering NK cell cytotoxicity specific to males.