Abstract
Genetic defects in the ability to deliver effective perforin have been reported in patients with hemophagocytic lymphohistiocytosis. We tested the hypothesis that a primary perforin deficiency might also be causal in severe SARS-CoV-2 infection. We recruited 54 volunteers confirmed as being SARS-CoV-2–infected by RT-PCR and admitted to intensive care units or non–intensive care units and age- and sex-matched healthy controls. Compared with healthy controls, the percentage of perforin-expressing CD3−CD56+ NK cells quantified by flow cytometry was low in COVID-19 patients (69.9 ± 17.7 versus 78.6 ± 14.6%, p = 0.026). There was no correlation between the proportions of perforin-positive NK cells and T8 lymphocytes. Moreover, the frequency of NK cells producing perforin was neither linked to disease severity nor predictive of death. Although IL-6 is known to downregulate perforin production in NK cells, we did not find any link between perforin expression and IL-6 plasma level. However, we unveiled a negative correlation between the degranulation marker CD107a and perforin expression in NK cells (r = −0.488, p = 10−4). PRF1 gene expression and the frequency of NK cells harboring perforin were normal in patients 1 y after acute SARS-CoV-2 infection. A primary perforin defect does not seem to be a driver of COVID-19 because NK perforin expression is 1) linked neither to T8 perforin expression nor to disease severity, 2) inversely correlated with NK degranulation, and 3) normalized at distance from acute infection. Thus, the cause of low frequency of perforin-positive NK cells appears, rather, to be consumption.
Introduction
Natural killer cells are type 1 innate lymphoid cells playing a key role in defense against tumors and intracellular pathogens, particularly via perforin- and granzyme B–mediated cytotoxicity (1). A panel of activating and inhibitory receptors displayed at their surface and sensing molecules at the surface of infected and tumoral cells determine their propensity to kill. The signals delivered by these receptors provoke the release of cytotoxic factors by NK cells, including perforin, resulting in the apoptosis of the target cells. Moreover, NK cells regulate T cell–mediated antiviral immune response (2). NK cells are particularly well known for being important actors in the immune defense against respiratory viral infections, including coronaviruses (3).
Severe forms of COVID-19 have many similar features to hemophagocytic lymphohistiocytosis (HLH) (4–6). The frequency of HLH symptoms in COVID-19 is discussed, with some authors reporting it as low (7–9) but others reporting it as higher than in non–COVID-19 sepsis (10). Various articles (8, 11), but not all (9), established the prognostic value on mortality of these symptoms. Differences in the level of different markers have also been observed between severe COVID-19 and HLH. For instance, IL-12, IL-15, IL-18, IL-21, IFN-γ, and soluble Fas ligand were shown to be lower, and IL-8, IL-1 receptor antagonist, and ICAM-1 were higher in the former than in the latter (12, 13). HLH is defined by the presence of at least five of the following criteria: fever, splenomegaly, cytopenia of at least two lineages, hyperferritinemia, an increase in soluble CD25, high triglyceridemia and low fibrinogenemia, decrease in NK activity, and hemophagocytosis (14). HLH may be primary, caused by a genetic deficiency in perforin or in factors involved in exocytosis, or secondary to malignant hemopathies, autoimmune disorders or infections, particularly viral infections. Even in secondary HLH, deficiencies in perforin or mutations in proteins mediating exocytosis have been found in up to 40% of patients (15). This raises the possibility that cytotoxic deficiency is a key driver of HLH and, potentially, of the severity of COVID-19. This may be explained by the fact that perforin-deficient cytotoxic cells produce more cytokines than wild-type cytotoxic cells (16). In fact, it is the target cell that triggers the detachment of the cytotoxic cell via a caspase-dependent signal. When perforin is deficient, the caspase pathway is not fully activated in the target cell, so the contact time is drastically increased. This increase results in an overproduction of IL-2, TNF-α, and IFN-γ by the cytotoxic cell (17). IFN-γ induces the release of large amounts of IL-6 by myeloid cells, a marker of HLH and COVID-19 (4). In line with this observation, NK cells secrete greater quantities of inflammatory cytokines in perforin-deficient mice infected with mouse CMV (18). In addition, NK cells may kill T cells (19). Consequently, in mice infected with the lymphocytic choriomeningitis virus, perforin deficiency has been shown to result in the accumulation of exhausted CD8+ T cells, responsible for immune-mediated damage and death, as in severe forms of COVID-19 (20, 21). NK cells may also kill NK cells and thereby downregulate immune activation (22). Finally, perforin has also been involved in NK and CD8+ T cell killing by regulatory T cells (23). Altogether, these findings argue for perforin’s role in reducing the intensity of immune responses. Moreover, the hypothesis of defective cytotoxicity as a causal factor of COVID-19 ties in with the fact that aging (24) and comorbidities like diabetes (6), conditions known to reduce perforin expression (25), are predictive of a poor prognosis in SARS-CoV-2 infection. To better understand the pathogenic mechanisms of this disease, we analyzed perforin expression in the NK cells of COVID-19 patients, seeking arguments in favor of perforin deficiency having an etiologic role in the severity of this disease. In the current study, we show a low frequency of perforin-expressing NK cells in COVID-19 patients, but we report arguments against a causal role of this impairment in COVID-19.
Materials and Methods
Study design
All patients were diagnosed as being SARS-CoV-2-infected by RT-PCR. Their blood was drawn on the first day of hospitalization at an intensive care unit (ICU) or a non-ICU in the University Hospital of Nîmes, France. ICU participants presented with oxygen saturation of <90% in ambient air or <95% with 5 l/min of oxygen therapy and/or arterial oxygen tension of less than 60 mm Hg. Non-ICU participants presented with oxygen saturation of <96% in ambient air. The French Île-de-France 1 Ethics Committee approved this study, and all volunteers provided written informed consent. The trial was registered on ClinicalTrials.gov under the reference NCT04351711.
Flow cytometry
Perforin expression was quantified by intracellular PBMC labeling. Frozen cells were first thawed and washed twice. A total of 200,000 cells were surface-labeled with the following Ab panel: CD3-allophycocyanine/Alexa750 (Beckman Coulter) + CD16-allophycocyanine (Beckman Coulter) + CD56-PE/cyanine 5.5 (Beckman Coulter) + CD107a-PE (BioLegend). IgG1-FITC (clone REA293, Miltenyi Biotec) and IgG1- PE (clone MOPC-21, BioLegend) were used as isotypic controls. The cells were then fixed using the Immunoprep reagent system kit and TQ Prep automate (Beckman Coulter) and permeabilized for perforin labeling (perforin-FITC, Miltenyi Biotec) using a Cytofix/Cytoperm kit (Becton Dickinson). A minimum of 20,000 lymphocytes were run on a Navios flow cytometer, and the results were analyzed by using Kaluza software (Beckman Coulter).
Biomarkers
The plasma levels of IL-6, IL-12p70, and IFN-α were determined by Luminex/xMAP immunoassay (ProcartaPlex, Thermo Fisher Scientific, Saint Aubin, France), and those of C-reactive protein and lactate dehydrogenase were determined by turbidimetry. PBLs and monocytes were counted by a hemocytometer (Sysmex XN-10).
Perforin mRNA expression
RNA was extracted with TRIzol on a frozen dry pellet of 100,000 PBMCs from volunteers and resuspended in 20 μl of diethyl pyrocarbonate–treated water. Using the Superscript III kit (Invitrogen), 8 μl were then reverse-transcribed with random hexamers. The cDNA was then amplified on a LightCycler 480 (Roche) with the Roche SYBR green master kit with the appropriate primers as technical triplicate. Perforin expression was estimated relative to GAPDH. The primer sequences were as follows: perforin, AACTTTGCAGCCCAGAAGACC and GTGCCGTAGTTGGAGATAAGCC; and GAPDH, AGTTAAAAGCAGCCCTGGTG and AGTTAAAAGCAGCCCTGGTG.
Statistical analysis
Normality was assessed by the d’Agostino and Pearson test. A two-sided unpaired Student t test or Mann–Whitney U test was used to compare groups, as appropriate. A two-sided Pearson or Spearman test evaluated correlations, as appropriate. A p value of < 0.05 was considered statistically significant.
Results
Patient characteristics
Twenty-eight non-ICU volunteers (15 females and 13 males, 67.5 ± 20.8 y old, symptomatic for 6.3 ± 9.5 d) and 26 ICU volunteers (13 females and 13 males, 70.6 ± 13.4 y old, symptomatic for 12.3 ± 7.3 d) were recruited. Their bioclinical characteristics had been previously reported (26). Twenty-nine age- and sex-matched healthy controls (HCs) were recruited in parallel (12 females and 17 males, 63.7 ± 19.0 y old).
NK cell perforin expression in the course of severe SARS-CoV-2 infection
Perforin expression was quantified in permeabilized CD3−CD56+ NK cells by flow cytometry (Fig. 1). As compared with HCs, we observed a decrease in the percentage of perforin-positive NK cells in patients (69.9 ± 17.7% versus 78.6 ± 14.6%, p = 0.026), particularly in ICU patients (67.5 ± 19.9% versus 78.6 ± 14.6%, p = 0.037), as shown in Fig. 2A. The frequency of perforin-expressing NK cells was independent of corticotherapy, either dexamethasone, prednisolone, methylprednisolone, or hydrocortisone (Supplemental Fig. 1A), and oxygenotherapy, either supplemental oxygen therapy or mechanical ventilation (Supplemental Fig. 1B).
If this perforin deficiency was genetically determined, one would expect patients with low NK perforin expression to also have low T cell perforin expression. We previously measured perforin expression in the T8 cells of patients we recruited (26), yet we observed no correlation between the proportion of perforin-positive NK cells and the proportion of perforin-positive T8 cells in ICU and non-ICU patients (r = 0.195, p = 0.167; Fig. 2B). This observation is a first argument against the hypothesis of a primary perforin deficiency as a driver of COVID-19.
In addition to the frequency of perforin-positive NK cells, we evaluated the amount of perforin per NK cell using the MFI of NK cells labeled with the anti-perforin mAb. As shown in Fig. 2C, perforin expression in NK cell was elevated in ICU patients (59,599 ± 63,168 arbitrary units versus 54,380 ± 86,750 arbitrary units, p = 0.059) and even more in non-ICU patients (120,130 ± 75,926 arbitrary units versus 54,380 ± 86,750 arbitrary units, p < 10−4), as compared with healthy controls. These findings, in line with previous data (27), are a second argument against the hypothesis of a primary perforin deficiency as a driver of COVID-19.
Perforin expression and COVID-19 severity
Firstly, we tested the hypothesis of a causal link between the ability to produce perforin and the severity of SARS-CoV-2 infection. As shown in Fig. 2A, there was no difference in the percentage of perforin-positive NK cells between ICU and non-ICU patients (67.5 ± 19.9% versus 72.0 ± 15.5%, p = 0.497), whereas COVID-19 was more severe in the former than in the latter. Secondly, we looked for correlations between perforin expression and established markers of severity (28). The frequency of perforin-expressing NK cells was linked neither to C-reactive protein (r = 0.231, p = 0.136; Fig. 3A), nor to lactate dehydrogenase (r = 0.131, p = 0.507; Fig. 3B), nor to lymphocyte count (r = 0.025, p = 0.867; Fig. 3C), nor to monocyte count (r = 0.093, p = 0.531; Fig. 3D). There was not any negative correlation between the median level of perforin expression in NK cells and C-reactive protein (r = 0.086, p = 0.588; Supplementary Fig. 2A), lactate dehydrogenase (r = −0.240, p = 0.228, Supplemental Fig. 2B), lymphocyte count (r = 0.270, p = 0.069; Supplemental Fig. 2C), or monocyte count (r = 0.088, p = 0.560; Supplemental Fig. 2D).
Finally, we rationalized that if a primary defect in perforin production was involved in the severity of COVID-19, low perforin expression should be predictive of an adverse prognosis. However, as shown in Fig. 3E and 3F, the percentage of perforin-positive NK cells (63.8 ± 21.9% versus 83.4 ± 12.1%, p = 0.046) and the median perforin expression per NK cell (34,139 ± 31,258 arbitrary units versus 124,770 ± 137,681 arbitrary units, p = 0.110) were not higher in ICU patients who survived than in those who did not.
Perforin expression and cytokines
Because we did not find any argument for a primary deficiency in perforin production in patients, we looked for other potential causes of the low frequency of NK cells harboring perforin we had observed. IFN-α and IL-12 are known to induce perforin expression in NK cells (25, 29). We therefore looked for, but did not find, any positive correlation between the plasma level of each cytokine and the frequency of perforin-positive NK cells (r = −0.283, p = 0.049; Fig. 4A; and r = −0.190, p = 0.196; Fig. 4B, respectively) or the NK cell content in perforin (r = 0.021, p = 0.890; Fig. 4D; and r = −0.255, p = 0.084; Fig. 4E, respectively). By contrast, IL-6 is known to inhibit perforin expression in NK cells (30). Therefore, we sought a link between NK cell perforin expression and IL-6 production. To this aim, we measured IL-6 in plasma. However, we did not find a significant negative correlation between the percentage of perforin-positive NK cells (r = −0.198, p = 0.181; Fig. 4C) or the density in perforin in NK cells (r = 0.129, p = 0.397; Fig. 4F) and IL-6 plasma levels.
Perforin expression and perforin consumption
Another explanation for the low frequency NK cells expressing perforin we unveiled might be consumption. To test this possibility, we determined cell surface expression of the degranulation marker CD107a. Indeed, we observed a clear negative correlation between the percentage of CD107a-positive NK cells and the percentage of perforin-positive NK cells (r = −0.488, p = 10−4; Fig. 5A). There was also a negative correlation between CD107a cell surface expression and the intracellular perforin level (r = −0.269, p = 0.051; Fig. 5B). This is a strong argument in favor of perforin release as a cause of the low frequency of perforin-positive NK cells. Of note, as compared with controls, the frequency of perforin degranulation, as evaluated by CD107a expression, was elevated in ICU patients (31.7 ± 25.3% versus 17.3 ± 20.2%, p = 0.014) but not in non-ICU patients (6.2 ± 9.7% versus 17.3 ± 20.2%, p < 10−4), in whom it was even low (Fig. 5C). The same tendency, but less significant, was observed for T8 cells (Fig. 5D). In our cohort, the mean duration of symptoms was 7 d for non-ICU patients and 12 d for ICU patients. Therefore, the low percentage of CD107a-positive NK cells that we observed in non-ICU participants is in line with the previous report of a defect in NK degranulation during the first week of severe disease (31).
Perforin expression in NK cells at distance from infection
To definitively rule out the hypothesis of a constitutive quantitative deficiency in perforin as a cause of severe COVID-19 in cases of SARS-CoV-2 infection, we measured PRF1 expression in six volunteer PBMCs 1 y after the acute episode of infection. At that time, their PRF1 mRNA was no different from that of five HCs (102.3 ± 60.6 arbitrary units versus 59.4 ± 54.7 arbitrary units, p = 0.248; Fig. 6A). Likewise, there was no difference in the frequency of perforin-positive NK cells (78.0 ± 10.0% versus 78.6 ± 14.6%, p = 0.893; Fig. 6B) between 12 patients 1 y after the acute infection and HCs.
Discussion
SARS-CoV-2 appears to be able to block IFN production and signaling during the early phase of infection (32). Consequently, COVID-19 patients present with an impaired IFN type 1 (IFNI) activity (33). This allows its replication during an asymptomatic phase. In some patients, this results in an acute immune activation, including a cytokine storm that might be coresponsible for the lung lesions that determine the prognosis (6). A key question is to unravel the reasons why individuals who develop a severe form of COVID-19 fail to control this virus-induced immune response. Paradoxically, among the possibilities is a faint cytotoxic activity. Perforin impairment may result in inflammatory cytokine overproduction by the cytotoxic cells, insufficient clearance of infected cells and activated immune cells, and a regulatory T cell deficiency contributing to an acute immune activation (25). Accordingly, a deficiency in perforin production or release may be responsible for the HLH syndrome. As COVID-19 has fever, cytopenia, hyperferritinemia, increase in soluble CD25, high triglyceridemia and low fibrinogenemia, and hemophagocytosis in common with HLH (5), it is logical to propose cytotoxic deficiency as a driver of severe forms of COVID-19.
In the current study, we did observe a decrease in the frequency of perforin-expressing NK cells in patients hospitalized for COVID-19. Now, contradictory data on perforin expression in NK cells and NK cytotoxicity have previously been reported in COVID-19 patients. Most authors observed a decrease in NK perforin protein (34, 35) or mRNA (36) expression, as well as a decrease in NK cytotoxicity (Refs. 37–40), whereas Jiang et al. found an increase (41). However, in this latter study, mild COVID-19 was included, and the percentage of perforin-positive NK cells was very low in healthy controls.
The low percentage of perforin-expressing NK cells that we observed does not seem to be a primary etiologic factor in severe forms, because it was 1) not concurrently present in T8 cells and even contrasted with the high frequency of perforin-positive T4 cells that we previously reported (26), 2) negatively correlated with a marker of degranulation, an argument in favor of perforin consumption, 3) neither linked to bioclinical markers of severity nor predictive of death, and 4) not observed 1 y after the episode of acute infection. Interestingly, earlier on we described a decrease in NK perforin expression in HIV-1–infected individuals of approximately the same magnitude as what we are now reporting in COVID-19 patients (42), yet HIV patients do not suffer from cytokine storms. Altogether, these data argue against a genetic background impairing perforin expression as a cause of severe COVID-19. In line with this, various authors have underlined differences in COVID-19 and HLH immune activation profiles. In particular, IL-6 and HScore (43) are usually lower in the former than in the latter. Nonetheless, the perforin gene variant A91V, which encodes for a perforin protein with impaired processing, was observed three times more frequently in 22 young patients with severe forms of COVID-19 (44). Moreover, the two A91V-positive patients who presented with a high HScore, a score of severity in HLH, progressed rapidly and died. It is therefore possible that perforin deficiency plays a role in young patients but not in older ones.
Lymphocyte DNA damage and apoptosis, which we have previously described in severe COVID-19, might contribute to the low frequency of perforin-positive NK cells that we observed here (45–47). This low frequency might also be due to the TGF-β overproduction (39), to anti-IFNI autoantibodies (48), or to the impaired IFNI production and signaling observed in COVID-19 patients (32, 33, 49), because IFNI is a major inducer of perforin expression (25). However, in opposition to this scenario, we observed a negative—rather than positive—correlation between IFN-α plasma levels and perforin expression in NK cells. Because IFN-α plasma levels have been linked to SARS-CoV-2 viral load (50) and because a high viral load should provoke NK activation, this anticorrelation might be explained by the fact that low perforin expression is a consequence of NK degranulation.
Consequently, the most probable hypothesis is indeed that the reduced number of NK cells harboring perforin in COVID-19 is the consequence of the release of this cytolytic factor in the course of NK cytotoxicity. A major way in which NK cells kill is by delivering perforin and granzyme B to inside the target cells, thereby triggering apoptosis (25). Prager et al. (51) reported that target cell contact reduced perforin in NK cells over time. Moreover, exposure to target cells has been reported to downregulate perforin mRNA in NK cells (52). Consequently, within 8 h of their contact with the target cell, NK cells lose their cytotoxic efficiency (53). Accordingly, we have shown here an inverse correlation between perforin expression and degranulation in COVID-19 NK cells. It is therefore logical to expect that in vivo SARS-CoV-2–infected cell elimination also results in a decrease in intra-NK perforin level. Strikingly, the amount of perforin in COVID patients’ NK cells is increased. This might be a consequence of NK cell activation reported by various authors (54–56), yet the inverse correlation between perforin load in NK cells and their degranulation argues for the consumption of this cytotoxic mediator.
Whatever the causes of low perforin expression might be, our data argue for the administration of IFNI at the early stage of SARS-CoV-2 infection, in order not to increase the acute immune activation that may occur at later stages. In fact, we have already demonstrated that pegylated IFN-α2 administration restores NK perforin expression in people living with HIV-1 (42). It may also be noted that the positive effects of IFN treatment on the discharge rate and mortality have been reported (57, 58).
Disclosures
The authors have no financial conflicts of interest.
Acknowledgments
We are grateful to Teresa Sawyers, Medical Writer at the Laboratoire de Biostatistique, Epidémiologie clinique, Santé Publique, Innovation et Méthodologie, Nîmes University Hospital, France, for expert assistance in editing this paper and also to the Centre de Ressources Biologiques, Nîmes University Hospital, France.
Footnotes
This work was supported by Nîmes University Hospital Grant NIMAO/2020/COVID/PC-0, the Fondation pour la Recherche Médicale, the Agence Nationale de la Recherche (COVID-I2A) Grant 216261, and AbbVie. J.E. was supported by the Canada Research Chair Program.
The online version of this article contains supplemental material.