A strong SARS-CoV-2 specific T cell response is predictive of milder disease

Ever since SARS-CoV-2 first appeared, researchers have been trying to understand whether sometimes the immune system does more harm than good during the acute phase of COVID-19.

The latest study by researchers at La Jolla Institute for Immunology clearly argues in favor of the immune system.

Their work, published in the Sept. 16, 2020, online issue of Cell, confirms that a multi-layered, virus-specific immune response is important for controlling the virus during the acute phase of the infection and reducing COVID-19 disease severity, with the bulk of the evidence pointing to a much bigger role for T cells than antibodies.

A weak or uncoordinated immune response, on the other hand, predicts a poor disease outcome.

The findings suggest that vaccine candidates should aim to elicit a broad immune response that include antibodies, helper and killer T cells to ensure protective immunity.

“Our observations could also explain why older COVID-19 patients are much more vulnerable to the disease,” says senior author Shane Crotty, Ph.D., who co-led the study with Alessandro Sette, Dr. Biol.Sci., both professors in LJI’s Center for Infectious Disease and Vaccine Research.

“With increasing age, the reservoir of T cells that can be activated against a specific virus declines and the body’s immune response becomes less coordinated, which looks to be one factor making older people drastically more susceptible to severe or fatal COVID-19.”

Adds Sette, “What we didn’t see was any evidence that T cells contribute to a cytokine storm, which is more likely mediated by the innate immune system.”

When SARS-CoV-2 (or any other virus) infiltrates the body, the innate immune system is first on the scene and launches a broad and unspecific attack against the intruder.

It releases waves of signaling molecules that incite inflammation and alert the immune system’s precision forces to the presence of a pathogen.

Within days, the so-called adaptive immune system tools up and moves with pinpoint precision against the virus, intercepting viral particles and killing infected cells.

The adaptive immune system consists of three branches: antibodies; helper T cells (Th), which assist B cell to make protective antibodies; and killer T cells (CTL), which seek out virus-infected cells and eliminate them.

For their latest study, the researchers collected blood samples from 50 COVID-19 patients and analyzed all three branches of the adaptive immune system – SARS-CoV-2 specific antibodies, helper and killer T cells – in great detail.

“It was particularly important to us to capture the whole range of disease manifestation from mild to critically ill so we could identify differentiating immunological factors,” says co-first author and infectious disease specialist Sydney Ramirez, M.D., Ph.D., who spearheaded the sample collection.

What the team found was that similar to their previous study all fully recovered individuals had measurable antibody, helper and killer T cell responses, while the adaptive immune response in acute COVID-19 patients varied more widely with some lacking neutralizing antibodies, others helper or killer T cells or any combination thereof.

“When we looked at a combination of all of our data across all 111 measured parameters we found that in general, people who mounted a broader and well-coordinated adaptive response tended to do better.

A strong SARS-CoV-2 specific T cell response, in particular, was predictive of milder disease,” says co-first author and postdoctoral research Carolyn Moderbacher, Ph.D. “Individuals whose immune response was less coordinated tended to have poorer outcomes.”

The effect was magnified when the researchers broke down the dataset by age.

“People over the age of 65 were much more likely to have poor T cell responses, and a poorly coordinated immune response, and thus have much more severe or fatal COVID-19,” says Crotty.

“Thus, part of the massive susceptibility of the elderly to COVID-19 appears to be a weak adaptive immune response, which may be because of fewer naïve T cells in the elderly.”

Naïve T cells are inexperienced T cells that have not met their viral match yet and are waiting to be called up.

As we age, the immune system’s supply of deployable naïve T cells dwindles and fewer cells are available to be activated to respond to a new virus.

“This could either lead to a delayed adaptive immune response that is unable to control a virus until it is too late to limit disease severity or the magnitude of the response is insufficient,” says Moderbacher.

In line with what other research teams had found before, antibodies don’t seem to play an important role in controlling acute COVID-19.

Instead, T cells and helper T cells in particular are associated with protective immune responses. “This was perplexing to many people,” says Crotty, “but controlling a primary infection is not the same as vaccine-induced immunity, where the adaptive immune system is ready to pounce at time zero.”

If a vaccination is successful, vaccine-induced antibodies are ready to intercept the virus when it shows up at the doorstep.

In contrast, in a normal infection the virus gets a head start because the immune system has never seen anything like it.

By the time the adaptive immune system is ready to go during a primary infection, the virus has already replicated inside cells and antibodies can’t get to it.

“Thus, these findings indicate it is plausible T cells are more important in natural SARS-CoV-2 infection, and antibodies more important in a COVID-19 vaccine,” says Crotty, “although it is also plausible that T cell responses against this virus are important in both cases.”

---

Innate Immune Sensing of SARS-CoV-2

Innate immune sensing serves as the first line of antiviral defense and is essential for immunity to viruses. To date, our understanding of the specific innate immune response to SARS-CoV-2 is extremely limited. However, the virus-host interactions involving SARS-CoV-2 are likely to recapitulate many of those involving other CoVs, given the shared sequence homology among CoVs and the conserved mechanisms of innate immune signaling.

In the case of RNA viruses such as SARS-CoV-2, these pathways are initiated through the engagement of pattern-recognition receptors (PRRs) by viral single-stranded RNA (ssRNA) and double-stranded RNA (dsRNA) via cytosolic RIG-I like receptors (RLRs) and extracellular and endosomal Toll-like receptors (TLRs).

Upon PRR activation, downstream signaling cascades trigger the secretion of cytokines. Among these, type I/III interferons (IFNs) are considered the most important for antiviral defense, but other cytokines, such as proinflammatory tumor necrosis factor alpha (TNF-α), and interleukin-1 (IL-1), IL-6, and IL-18 are also released.

Together, they induce antiviral programs in target cells and potentiate the adaptive immune response.

If present early and properly localized, IFN-I can effectively limit CoV infection (^{Channappanavar et al., 2016, Channappanavar et al., 2019}). Early evidence demonstrated that SARS-CoV-2 is sensitive to IFN-I/III pretreatment in vitro, perhaps to a greater degree than SARS-CoV-1 (^{Blanco-Melo et al., 2020, Lokugamage et al., 2020, Mantlo et al., 2020, Stanifer et al., 2020}).

However, the specific IFN-stimulated genes (ISGs) that mediate these protective effects are still being elucidated. Lymphocyte antigen 6 complex locus E (LY6E) has been shown to interfere with SARS-CoV-2 spike (S) protein-mediated membrane fusion (^{Pfaender et al., 2020, Zhao et al., 2020c}).

Likely, the IFN-induced transmembrane family (IFITM) proteins inhibit SARS-CoV-2 entry, as demonstrated for SARS-CoV-1 (^{Huang et al., 2011b}), although their action in promoting infection has also been described for other CoVs (^{Zhao et al., 2014, Zhao et al., 2018}).

Evasion of Innate Sensing by Coronaviruses

As these cytokines represent a major barrier to viral infection, CoVs have evolved several mechanisms to inhibit IFN-I induction and signaling. Numerous studies have demonstrated that SARS-CoV-1 suppresses IFN release in vitro and in vivo (^{Cameron et al., 2012, Minakshi et al., 2009, Siu et al., 2009, Wathelet et al., 2007}).

SARS-CoV-2 likely achieves a similar effect, as suggested by the lack of robust type I/III IFN signatures from infected cell lines, primary bronchial cells, and a ferret model (^{Blanco-Melo et al., 2020}).

In fact, patients with severe COVID-19 demonstrate remarkably impaired IFN-I signatures as compared to mild or moderate cases (^{Hadjadj et al., 2020}).

As is often the case, there are multiple mechanisms of evasion for CoVs, with viral factors antagonizing each step of the pathway from PRR sensing and cytokine secretion to IFN signal transduction (Figure 1 ).

CoV-mediated antagonism of innate immunity begins with evasion of PRR sensing. ssRNA viruses, like CoVs, form dsRNA intermediates during their replication, which can be detected by TLR3 in the endosome and RIG-I, MDA5, and PKR in the cytosol. ssRNA may also be detected by TLR7 or TLR8 and potentially RIG-I and PKR.

CoVs are known to avoid PRR activation by either avoiding recognition altogether or antagonizing PRR action (^{Bouvet et al., 2010, Chen et al., 2009, Deng et al., 2017, Hackbart et al., 2020, Ivanov et al., 2004, Knoops et al., 2008}).

To evade PRRs, dsRNA is first shielded by membrane-bound compartments that form during viral replication of SARS-CoV-1 (^{Knoops et al., 2008}).

In addition, viral RNA is guanosine-capped and methylated at the 5′ end by CoVs non-structural proteins (NSPs) 10, 13, 14, and 16 (^{Bouvet et al., 2010, Chen et al., 2009, Ivanov et al., 2004}), thereby resembling host mRNA to promote translation, prevent degradation, and evade RLR sensing.

Finally, CoVs also encode an endoribonuclease, NSP15, that cleaves 5′ polyuridines formed during viral replication, which would otherwise be detected by MDA5 (^{Deng et al., 2017, Hackbart et al., 2020}). CoVs have evolved additional strategies to impede activation of PRRs. SARS-CoV-1 N-protein prevents TRIM25 activation of RIG-I (^{Hu et al., 2017}).

Likewise, MERS-CoV NS4a, which itself binds dsRNA, impedes PKR activation (^{Comar et al., 2019, Rabouw et al., 2016}) and inhibits PACT, an activator of RLRs (^{Niemeyer et al., 2013, Siu et al., 2014}).

Additionally, MERS-CoV NS4b antagonizes RNaseL, another activator of RLRs (^{Thornbrough et al., 2016}). The role of other PRRs remains unclear.

For example, SARS-CoV-1 papain-like protease (PLP) antagonizes STING, suggesting that self-DNA may also represent an important trigger (^{Sun et al., 2012}). The extent to which SARS-CoV-2 homologs overlap in these functions is currently unknown.

Following activation, RLR and TLRs induce signaling cascades, leading to the phosphorylation of transcription factors, such as NF-kB and the interferon-regulatory factor family (IRF), ultimately leading to transcription of IFN and proinflammatory cytokines.

Although no experimental studies have delineated the precise functions of SARS-CoV-2 proteins, proteomic studies have demonstrated interactions between viral proteins and PRR signaling cascades. SARS-CoV-2 ORF9b indirectly interacts with the signaling adaptor MAVS via its association with Tom70 (^{Gordon et al., 2020}), consistent with prior reports that SARS-CoV-1 ORF9b suppresses MAVS signaling (^{Shi et al., 2014}).

Furthermore, SARS-CoV-2 NSP13 interacts with signaling intermediate TBK1, and NSP15 is associated with RNF41, an activator of TBK1 and IRF3 (^{Gordon et al., 2020}).

Similarly, SARS-CoV-1 M protein is known to inhibit the TBK1 signaling complex (^{Siu et al., 2009}), as does MERS-CoV ORF4b (^{Yang et al., 2015}). Other proteins, including SARS-CoV-1 PLP, N, ORF3b, and ORF6, block IRF3 phosphorylation and nuclear translocation (^{Devaraj et al., 2007, Kopecky-Bromberg et al., 2007}).

NF-kB is also inhibited by CoV proteins. These include SARS-CoV-1 PLP (^{Frieman et al., 2009}) and MERS-CoV ORF4b and ORF5 (^{Canton et al., 2018, Menachery et al., 2017}).

Finally, SARS-CoV-1 NSP1 (^{Huang et al., 2011a, Kamitani et al., 2009}) and MERS-CoV NSP1 (^{Lokugamage et al., 2015}) initiate general inhibition of host transcription and translation, thus limiting antiviral defenses nonspecifically.

To prevent signaling downstream of IFN release, CoV proteins inhibit several steps of the signal transduction pathway that bridge the receptor subunits (IFNAR1 and IFNAR2) to the STAT proteins that activate transcription.

For SARS-CoV-1, these mechanisms include IFNAR1 degradation by ORF3a (^{Minakshi et al., 2009}), decreased STAT1 phosphorylation by NSP1 (^{Wathelet et al., 2007}), and antagonism of STAT1 nuclear translocation by ORF6 (^{Frieman et al., 2007, Kopecky-Bromberg et al., 2007}).

However, SARS-CoV-2 ORF6 shares only 69% sequence homology with SARS-CoV-1, suggesting this function may not be conserved. In support of this notion, SARS-CoV-2 infection fails to limit STAT1 phosphorylation, unlike in SARS-CoV-1 infection (^{Lokugamage et al., 2020}).

Imbalance between Antiviral and Proinflammatory Responses

Taken together, the multiplicity of strategies developed by pathogenic CoVs to escape immune sensing, particularly the IFN-I pathway, suggests a critical role played by the dysregulation of IFN-I response in COVID-19 pathogenicity.

Concordantly, animal models of SARS-CoV-1 and MERS-CoV infection indicate that failure to elicit an early IFN-I response correlates with the severity of disease (^{Channappanavar et al., 2016}).

Perhaps more importantly, these models demonstrate that timing is key, as IFN is protective early in disease but later becomes pathologic (^{Channappanavar et al., 2016, Channappanavar et al., 2019}).

Perhaps interferon-induced upregulation of ACE2 in airway epithelia may contribute to this effect (^{Ziegler et al., 2020}).

Furthermore, while pathogenic CoVs block IFN signaling, they may actively promote other inflammatory pathways contributing to pathology. For instance, SARS-CoV-1 ORF3a, ORF8b, and E proteins enhance inflammasome activation (^{Chen et al., 2019, Nieto-Torres et al., 2015, Shi et al., 2019, Siu et al., 2019}), leading to secretion of IL-1β and IL-18, which are likely to contribute to pathological inflammation.

Similarly, SARS-CoV-2 NSP9 and NSP10 might induce IL-6 and IL-8 production, potentially by inhibition of NKRF, an endogenous NF-kB repressor (^{Li et al., 2020a}).

Collectively, these proinflammatory processes likely contribute to the “cytokine storm” observed in COVID-19 patients and substantiate a role for targeted immunosuppressive treatment regimens. Moving forward, a clear understanding of the delicate balance between antiviral and inflammatory innate immune programs will be essential to developing effective biomarkers and therapeutics for COVID-19.

Myeloid Cells

Mucosal immune responses to infectious agents are orchestrated and regulated by myeloid cells with specialized functions, which include conventional dendritic cells (cDCs), monocyte-derived DCs (moDCs), plasmacytoid DCs (pDCs), and macrophages (^{Guilliams et al., 2013}).

A growing body of evidence points to dysregulated myeloid responses that potentially drive the COVID-19 hallmark syndromes, such as acute respiratory distress syndrome (ARDS), cytokine release syndrome (CRS) and lymphopenia (^{Mehta et al., 2020}).

Myeloid Characterization in COVID-19

Flow cytometric analyses of peripheral blood mononuclear cells (PBMCs) from symptomatic COVID-19 patients have shown a significant influx of granulocyte-macrophage colony-stimulating factor (GM-CSF)-producing, activated CD4⁺ T cells and CD14⁺HLA-DR^lo inflammatory monocytes (IMs) (^{Giamarellos-Bourboulis et al., 2020, Zhang et al., 2020c, Zhou et al., 2020b}).

This matches single-cell transcriptomic (scRNA-seq) data demonstrating CD14⁺IL-1β⁺ monocytic expansion (^{Guo et al., 2020, Wen et al., 2020}), interferon-mitogen-activated protein kinase (MAPK)-driven adaptive immune responses (^{Huang et al., 2020c}), and IL-1β-associated inflammasome signatures (^{Ong et al., 2020}) in peripheral blood of COVID-19 patients, although systemic levels of IL-1β detected are conspicuously low (^{Del Valle et al., 2020}).

Importantly, these immune signatures track with progression of clinical disease. scRNA-seq studies performed on pulmonary tissues of patients with severe COVID-19 disease have revealed an expansion of IMs and Ficolin-1⁺ monocyte-derived macrophages at the expense of tissue-resident reparative alveolar macrophages (AMs) (^{Liao et al., 2020}).

The aforementioned study also observed signatures of IFN signaling and monocyte recruitment that likely contribute to the rapid decline in alveolar patency and promote ARDS. Although most of the clinical focus has been on pulmonary damage and mononuclear phagocyte (MNP) dysfunction therein, it is increasingly clear that COVID-19 likely presents systemic challenges in other organ sites, such as the ileum and kidneys. Understanding the role of non-pulmonary myeloid cells in tissue-specific pathology associated with COVID-19 will be important.

Prior Knowledge from SARS-CoV-1, MERS-CoV, and Murine Coronaviruses

While data on COVID-19 patients continues to rapidly emerge, studies of myeloid cell dysfunction in SARS-CoV-1 and MERS-CoV can provide an important roadmap to understanding COVID-19 pathogenesis (Figure 2 ).

SARS-CoV-1 infection in mouse models results in an aberrant AM phenotype that limits DC trafficking and T cell activation (^{Zhao et al., 2009}).

Additionally, YM1⁺ FIZZ1⁺ alternative macrophages can increase airway hypersensitivity, thus exacerbating SARS-associated fibrosis (^{Page et al., 2012}). Further, as described above, murine SARS-CoV-1 studies have demonstrated that delayed IFN-I signaling and inflammatory monocytes-macrophages promote lung cytokine and chemokine levels, vascular leakage, and impaired antigen-specific T cell responses, culminating in lethal disease (^{Channappanavar et al., 2016}).

The role played by prominent IFN-producing pDCs in SARS-CoV-2 control or pathogenesis warrants investigation, as they have been shown to be critical in murine CoV (MHV) control (^{Cervantes-Barragan et al., 2007}).

Longitudinal studies in SARS-CoV-2 models are awaited, but initial phenotypic studies in humanized hACE2 mice have shown the characteristic alveolar interstitial pneumonia, with infiltration of lymphocytes and monocytes and accumulation of macrophages in the alveolar lumen (^{Bao et al., 2020a}), which recapitulates patient findings (^{Xu et al., 2020c}).

Lastly, non-human primate (NHP) studies and patient data on SARS-CoV-1 have also shown that virus spike-specific immunoglobulin G (IgG) responses can exacerbate acute lung injury due to repolarization of alveolar macrophages into proinflammatory phenotypes and enhanced recruitment of inflammatory monocyte via CCL2 and IL-8 (^{Clay et al., 2012, Liu et al., 2019}).

However, the extent to which the antibody response contributes to disease pathophysiology remains to be confirmed.

Myeloid Cells Contribution to Pathogenic Inflammation

The initial mode of viral pathogen-associated signal (PAMP) recognition by innate cells has a major impact on downstream myeloid signaling and cytokine secretion (^{de Marcken et al., 2019}).

While macrophages are somewhat susceptible to MERS-CoV and SARS-CoV-1 infection (^{Perlman and Dandekar, 2005, Zhou et al., 2014}), data do not suggest that they are infected by SARS-CoV-2, although one study reported ACE2 and SARS-CoV-2 nucleocapsid protein is expressed in lymph nodes and spleen-associated CD169⁺ macrophages of COVID-19 patients producing IL-6 (^{Chen et al., 2020h}).

Significantly elevated systemic levels of proinflammatory cytokine IL-6 have been reported in several COVID-19 patient cohorts and shown to correlate with disease severity (^{Mehta et al., 2020}). Increased IL-6 can also be associated with higher levels of IL-2, IL-7, IFN-ɣ, and GM-CSF, as seen in secondary hemophagocytic lymphohistiocytosis.

In response to viral infections, MNPs drive IL and IFN-I and IFN-III production resulting in inflammasome activation, induction of pathogenic Th1 and Th17 cell responses, recruitment of effector immune cells, and CRS pathology (^{Prokunina-Olsson et al., 2020, Tanaka et al., 2016}).

Independently, in vitro studies have demonstrated SARS-CoV-1 infection can induce intracellular stress pathways, resulting in NLRP3-dependent inflammasome activation and macrophage pyroptosis (^{Chen et al., 2019, Shi et al., 2019}). Functional studies are required to implicate these myeloid inflammasome pathways in COVID-19 lung pathology and to assess other immunogenic pathways such as RIPK1/3-dependent necroptosis (^{Nailwal and Chan, 2019}).

In conclusion, the strength and duration of myeloid ISG)signaling potentially dictate COVID-19 disease severity, but rigorous studies are warranted to confirm this.

Lastly, more work is needed to ascertain the mechanistic role played by lung-resident and recruited granulocytes in SARS-CoV-2 control and pathogenesis (^{Camp and Jonsson, 2017, Flores-Torres et al., 2019}). In contrast to their early protective role, neutrophil NETosis and macrophage crosstalk can drive later-stage inflammatory cascades (^{Barnes et al., 2020}), underscoring the overall pathogenic nature of damage-sensing host responses (Figure 2).

Collectively, the current knowledge of CoVs and SARS-CoV-2 infection, in particular, points to an inadvertent collusion involving myeloid cells in COVID-19 pathogenesis, despite their critical role in early sensing and antiviral responses.

Innate Lymphoid Cells

Innate lymphoid cells (ILCs) are innate immune effector cells that lack the expression of rearranged antigen receptors (T cell receptor [TCR], B cell receptor [BCR]). The ILC family is divided into two main groups: the cytotoxic natural killer (NK) cells and the non-cytotoxic helper ILCs, which include ILC1, ILC2, and ILC3 (^{Vivier et al., 2018}). Conventional NK cells include CD56^brightCD16⁻ NK cells and CD56^dimCD16⁺ cells, which are specialized in cytokine production or cytotoxicity, respectively.

NK Cells Are Decreased in the Peripheral Blood of COVID-19 patients

Multiple studies have reported reduced numbers of NK cells in the peripheral blood of COVID-19 patients, which is associated with severity of the disease (^{Song et al., 2020, Wang et al., 2020f, Yu et al., 2020, Zheng et al., 2020b}).

A recent scRNA-seq analysis revealed a transcriptomic signature for NK cells that was equally represented in lungs from patients and healthy donors (^{Liao et al., 2020}). The majority of lung NK cells are non-resident (^{Gasteiger et al., 2015, Marquardt et al., 2017}), and CXCR3 has been shown to mediate NK cell infiltration upon influenza infection (^{Carlin et al., 2018}). 

In vitro, CXCR3 ligands (CXCL9-11) are increased in SARS-CoV-2-infected human lung tissue (^{Chu et al., 2020}), and CXCR3-ligand-producing monocytes are expanded in the lungs of COVID-19 patients (^{Liao et al., 2020}). This suggests that the CXCR3 pathway might facilitate NK cell recruitment from the peripheral blood to the lungs in COVID-19 patients (Figure 2).

NK Cell Activation Pathways in Antiviral Immunity

NK cells express inhibitory and activating receptors that regulate their cytotoxicity. They are therefore able to induce the lysis of virus-infected cells that upregulate virus-derived proteins, as well as stress-inducible ligands, which are then recognized by NK-cell-activating receptors, such as NKp46 (^{Cerwenka and Lanier, 2001, Draghi et al., 2007, Duev-Cohen et al., 2016, Glasner et al., 2012}).

Future studies should investigate the expression of NK receptor ligands on SARS-CoV-2-infected cells in order to better understand the mechanisms underlying NK cell activation in COVID-19 disease. Further, secretion of IgG1 and IgG3 antibodies during SARS-CoV-2 infection (^{Amanat et al., 2020}) may induce CD56^dim CD16⁺ NK cell activation through Fc receptor recognition of antibodies either bound to surface antigens expressed on infected cells or to extracellular virions as immune complexes (Figure 2).

This interaction might trigger both cytokine production by NK cells and lysis of infected cells through antibody-mediated cellular cytotoxicity (ADCC), as shown in influenza infection (^{Von Holle and Moody, 2019}). Emerging data highlight the capacity for NK-mediated ADCC in response to naturally isolated SARS-CoV-1 anti-S IgG that crossreacts with SARS-CoV-2 S glycoprotein when transfected into Chinese hamster ovary (CHO) cells (^{Pinto et al., 2020}).

These findings suggest that triggering NK cell activation may not only contribute to the resolution of infection, but also contribute to the cytokine storm in ARDS.

Impairment of NK Cell Function in SARS-CoV-2 Infection

Ex vivo NK cells from peripheral blood of COVID-19 patients have reduced intracellular expression of CD107a, Ksp37, granzyme B, and granulysin, suggesting an impaired cytotoxicity, as well as an impaired production of chemokines, IFN-ɣ, and TNF-α (^{Wilk et al., 2020, Zheng et al., 2020b}).

Several pathways may contribute to the dysregulation of NK cells. While influenza virus infects NK cells and induces apoptosis (^{Mao et al., 2009}), lung NK cells do not express the entry receptor for SARS-CoV-2, ACE2, and are therefore unlikely to be directly infected by SARS-CoV-2 (^{Travaglini et al., 2020}).

The majority of NK cells found in human lung display a mature CD16⁺KIR⁺CD56^dim phenotype and are able to induce cell cytotoxicity in response to loss of human leukocyte antigen (HLA) class I or through Fc receptor signaling, although to a lower extent than their peripheral blood counterpart (^{Marquardt et al., 2017}).

Killer-immunoglobulin receptors (KIRs) are acquired during NK cell development alongside CD16 (FcRγIIIA) and are essential for NK cell licensing and subsequent capacity for cytolytic function (^{Sivori et al., 2019}). Frequencies of NK cells expressing CD16 and/or KIRs are decreased in the blood following SARS-CoV-2 and SARS-CoV-1 infection, respectively (^{Xia et al., 2004, Wang et al., 2020d}). Collectively, the data suggest either an impaired maturation of the NK compartment or migration of the mature, circulating NK cells into the lungs or other peripheral tissues of SARS-CoV-2-infected patients.

The immune checkpoint NKG2A is increased on NK cells and CD8 T cells from COVID-19 patients (^{Zheng et al., 2020b}). NKG2A inhibits cell cytotoxicity by binding the non-classical HLA-E molecule (^{Braud et al., 1998, Brooks et al., 1997}), and this interaction is strongly correlated with poor control of HIV-1 infection (^{Ramsuran et al., 2018}).

Genes encoding the inhibitory receptors LAG3 and TIM3 are also upregulated in NK cells from COVID-19 patients (^{Wilk et al., 2020, Hadjadj et al., 2020}). Thus, increased immune checkpoints on NK cells might contribute to viral escape. Additionally, COVID-19 patients have higher plasma concentrations of IL-6 (^{Huang et al., 2020b}), which significantly correlate with lower NK cell numbers (^{Wang et al., 2020d, Wang et al., 2020f}). 

In vitro stimulation by IL-6 and soluble IL-6 receptor has previously revealed impaired cytolytic functions (perforin and granzyme B production) by healthy donor NK cells, which can be restored following addition of tocilizumab (IL-6R blockade) (^{Cifaldi et al., 2015}).

TNF-α is also upregulated in the plasma of COVID-19 patients (^{Huang et al., 2020b}), and ligand-receptor interaction analysis of peripheral blood scRNA-seq data suggests that monocyte-secreted TNF-α might bind to its receptors on NK cells (^{Guo et al., 2020}). TNF-α is known to contribute to NK cell differentiation (^{Lee et al., 2009}), which includes downregulation of NKp46 (^{Ivagnès et al., 2017}), though no effect of TNF-α or IL-6 on NK cell-mediated ADCC has been reported so far. Collectively, these data suggest that crosstalk with monocytes might impair NK cell recognition and killing of SARS-CoV-2-infected cells, and antibodies targeting IL-6 and TNF-signaling may benefit enhanced NK cell functions in COVID-19 patients (Figure 2).

Relevance for Helper ILCs in SARS-CoV-2 Infection

No studies, to date, have reported ILC1, ILC2, or ILC3 functions in SARS-CoV-2 infection. All three subsets are present in healthy lung (^{De Grove et al., 2016, Yudanin et al., 2019}).

ILC2s are essential for the improvement of lung function following influenza infection in mice through amphiregulin-mediated restoration of the airway epithelium and oxygen saturation (^{Monticelli et al., 2011}).

However, ILC2s also produce IL-13, contributing to the recruitment of macrophages to the lung and influenza-induced airway hyperreactivity (^{Chang et al., 2011}).

Indeed, ILCs are involved in the polarization of alveolar macrophages, either toward a M1-like phenotype (ILC1 and ILC3) or a M2-like phenotype (ILC2) (^{Kim et al., 2019}). Given the increased IL-13 concentrations (^{Huang et al., 2020b}) and the dysregulation of the macrophage compartment observed in COVID-19 patients, the role played by ILCs in SARS-CoV-2 infection warrants further investigation.

T Cell Responses

T cells play a fundamental role in viral infections: CD4 T cells provide B cell help for antibody production and orchestrate the response of other immune cells, whereas CD8 T cells kill infected cells to reduce the viral burden. However, dysregulated T cell responses can result in immunopathology.

To better understand the role of T cell responses in SARS-CoV-2 infection, the pursuit of two major questions is imperative:

(1) what is the contribution of T cells to initial virus control and tissue damage in the context of COVID-19, and

(2) how do memory T cells established thereafter contribute to protective immunity upon reinfection?

Some tentative answers are beginning to emerge.

Overall Reduction of CD4 and CD8 T Cell Counts in Peripheral Blood

Similar to earlier observations about SARS-CoV-1 infection (^{He et al., 2005}), several current reports emphasize the occurrence of lymphopenia with drastically reduced numbers of both CD4 and CD8 T cells in moderate and severe COVID-19 cases (Figure 3 ) (^{Chen et al., 2020c, Nie et al., 2020b, Wang et al., 2020d, Zeng et al., 2020, Zheng et al., 2020b}).

The extent of lymphopenia—most striking for CD8 T cells in patients admitted to the intensive care unit (ICU)—seemingly correlates with COVID-19-associated disease severity and mortality (^{Chen et al., 2020c, Diao et al., 2020, Liu et al., 2020b, Liu et al., 2020c, Tan et al., 2020a, Wang et al., 2020d, Wang et al., 2020f, Zeng et al., 2020, Zhou et al., 2020c}).

Patients with mild symptoms, however, typically present with normal or slightly higher T cell counts (^{Liu et al., 2020a, Thevarajan et al., 2020}). The cause of peripheral T cell loss in moderate to severe COVID-19, though a phenomenon also observed in other viral infections, remains elusive, and direct viral infection of T cells, in contrast to MERS-CoV (^{Chu et al., 2016}), has not been reported.

Several mechanisms likely contribute to the reduced number of T cells in the blood, including effects from the inflammatory cytokine milieu. Indeed, lymphopenia seems to correlate with serum IL-6, IL-10, and TNF-α (^{Diao et al., 2020, Wan et al., 2020a}), while convalescent patients were found to have restored bulk T cell frequencies paired with overall lower proinflammatory cytokine levels (^{Chen et al., 2020f, Diao et al., 2020, Liu et al., 2020a, Liu et al., 2020b, Zheng et al., 2020b}).

Cytokines such as IFN-I and TNF-α may inhibit T cell recirculation in blood by promoting retention in lymphoid organs and attachment to endothelium (^{Kamphuis et al., 2006, Shiow et al., 2006}). However, in an autopsy study examining the spleens and hilar lymph nodes of six patients who succumbed to COVID-19, Chen et al. observed extensive cell death of lymphocytes and suggested potential roles for IL-6 as well as Fas-FasL interactions (^{Chen et al., 2020h}).

In support of this hypothesis, the IL-6 receptor antagonist tocilizumab was found to increase the number of circulating lymphocytes (^{Giamarellos-Bourboulis et al., 2020}).

T cell recruitment to sites of infection may also reduce their presence in the peripheral blood compartment.

scRNA-seq analysis of bronchoalveolar lavage (BAL) fluid of COVID-19 patients revealed an increase in CD8 T cell infiltrate with clonal expansion (^{Liao et al., 2020}). Likewise, post-mortem examination of a patient who succumbed to ARDS following SARS-CoV-2 infection showed extensive lymphocyte infiltration in the lungs (^{Xu et al., 2020c}).

However, another study that examined post-mortem biopsies from four COVID-19 patients only found neutrophilic infiltration (^{Tian et al., 2020a}). Further studies are therefore needed to better determine the cause and impact of the commonly observed lymphopenia in COVID-19 patients.

Induction of Antiviral T Cell Responses

Available information about SARS-CoV-1-specific T cell immunity may serve as an orientation for further understanding of SARS-CoV-2 infection. Immunogenic T cell epitopes are distributed across several SARS-CoV-1 proteins (S, N, and M, as well as ORF3), although CD4 T cell responses were more restricted to the S protein (^{Li et al., 2008}).

In SARS-CoV-1 survivors, the magnitude and frequency of specific CD8 memory T cells exceeded that of CD4 memory T cells, and virus-specific T cells persisted for at least 6–11 years, suggesting that T cells may confer long-term immunity (^{Ng et al., 2016, Tang et al., 2011}).

Limited data from viremic SARS patients further indicated that virus-specific CD4 T cell populations might be associated with a more severe disease course, since lethal outcomes correlated with elevated Th2 cell (IL-4, IL-5, IL-10) serum cytokines (^{Li et al., 2008}).

However, the quality of CD4 T cell responses needs to be further characterized to understand associations with disease severity. Few studies have thus far characterized specific T cell immunity in SARS-CoV-2 infection. In 12 patients recovering from mild COVID-19, robust T cell responses specific for viral N, M, and S proteins were detected by IFN-γ ELISPOT, weakly correlated with neutralizing antibody concentrations (similar to convalescent SARS-CoV-1 patients; ^{Li et al., 2008}), and subsequently contracted with only N-specific T cells detectable in about one-third of the cases post recovery (^{Ni et al., 2020}).

In a second study, PBMCs from COVID-19 patients with moderate to severe ARDS were analyzed by flow cytometry approximately 2 weeks after ICU admission (^{Weiskopf et al., 2020}). Both virus-specific CD4 and CD8 T cells were detected in all patients at average frequencies of 1.4% and 1.3%, respectively, and very limited phenotyping according to CD45RA and CCR7 expression status characterized these cells predominantly as either CD4 Tcm (central memory) or CD8 Tem (effector memory) and Temra (effector memory RA) cells.

This study is notable for the use of large complementary peptide pools comprising 1,095 SARS-Cov-2 epitopes (overlapping 15-mers for S protein as well as computationally predicted HLA-I- and -II-restricted epitopes for all other viral proteins) as antigen-specific stimuli that revealed a preferential specificity of both CD4 and CD8 T cells for S protein epitopes, with the former population modestly increasing over ∼10–30 days after initial onset of symptoms.

A caveat, however, pertains to the identification of specific T cells by induced CD69 and CD137 co-expression, since upregulation of CD137 by CD4 T cells, in contrast to CD154, may preferentially capture regulatory T cells (Treg) (^{Bacher et al., 2016}). Further analyses of S protein-specific T cells by ELISA demonstrated robust induction of IFN-γ, TNF-α, and IL-2 concomitant with lower levels of IL-5, IL-13, IL-9, IL-10, and IL-22.

A third report focused on S-specific CD4 T cell responses in 18 patients with mild, severe, or critical COVID-19 using overlapping peptide pools and induced CD154 and CD137 co-expression as a readout for antiviral CD4 T cells.

Such cells were present in 83% of cases and presented with enhanced CD38, HLA-DR, and Ki-67 expression indicative of recent in vivo activation (^{Braun et al., 2020}). Of note, the authors also detected low frequencies of S-reactive CD4 T cells in 34% of SARS-CoV-2 seronegative healthy control donors.

However, these CD4 T cells lacked phenotypic markers of activation and were specific for C-terminal S protein epitopes that are highly similar to endemic human CoVs, suggesting that crossreactive CD4 memory T cells in some populations (e.g., children and younger patients that experience a higher incidence of hCoV infections) may be recruited into an amplified primary SARS-CoV-2-specific response (^{Braun et al., 2020}).

Similarly, endemic CoV-specific CD4 T cells were previously shown to recognize SARS-CoV1 determinants (^{Gioia et al., 2005}). How previous infections with endemic CoV may affect immune responses to SARS-CoV-2 will need to be further investigated.

Finally, in general accordance with the above findings on the induction of SARS-CoV-2-specific T cells, using TCR sequencing (TCR-seq), Huang et al. and Liao et al. reported greater TCR clonality of peripheral blood (^{Huang et al., 2020c}) as well as BAL T cells (^{Liao et al., 2020}) in patients with mild versus severe COVID-19.

Moving forward, a comprehensive identification of immunogenic SARS-CoV-2 epitopes recognized by T cells (^{Campbell et al., 2020}), as well as further studies on convalescent patients who recovered from mild and severe disease, will be particularly important.

T Cell Contribution to COVID-19 Hyperinflammation

While the induction of robust T cell immunity is likely essential for efficient virus control, dysregulated T cell responses may cause immunopathology and contribute to disease severity in COVID-19 patients (Figure 3).

This is suggested in a study by Zhou et al., which reported a significantly increased PBMC frequency of polyclonal GM-CSF⁺ CD4 T cells capable of prodigious ex vivo IL-6 and IFN-γ production only in critically ill COVID-19 patients (^{Zhou et al., 2020c}).

Of note, GM-CSF⁺ CD4 T cells have been previously implicated in inflammatory autoimmune diseases, such as multiple sclerosis or juvenile rheumatoid arthritis, and high levels of circulating GM-CSF⁺ CD4 T cells were found to be associated with poor outcomes in sepsis (^{Huang et al., 2019}).

Additionally, two studies observed reduced frequencies of Treg cells in severe COVID-19 cases (^{Chen et al., 2020c, Qin et al., 2020}). Since Treg cells have been shown to help resolve ARDS inflammation in mouse models (^{Walter et al., 2018}), a loss of Tregs might facilitate the development of COVID-19 lung immunopathology. Similarly, a reduction of γδ-T cells, a subset of T cells with apparent protective antiviral function in influenza pneumonia (^{Dong et al., 2018, Zheng et al., 2013}), has been reported in severely sick COVID-19 patients (^{Guo et al., 2020, Lei et al., 2020b}).

Phenotype and Function of T Cell Subsets in COVID-19

Currently, little is known about specific phenotypical and/or functional T cell changes associated with COVID-19.

In the majority of preprints and peer-reviewed studies, there are reports of increased presence of activated T cells (Figure 3) characterized by expression of HLA-DR, CD38, CD69, CD25, CD44, and Ki-67 (^{Braun et al., 2020, Ni et al., 2020, Guo et al., 2020, Liao et al., 2020, Thevarajan et al., 2020, Yang et al., 2020a, Zheng et al., 2020a}).

Generally, independent of COVID-19 disease severity, CD8 T cells seem to be more activated than CD4 T cells (^{Qin et al., 2020, Thevarajan et al., 2020, Yang et al., 2020a}), a finding that echoes stronger CD8 than CD4 T cell responses during SARS-CoV-1 (^{Li et al., 2008}).

Furthermore, in a case study of 10 COVID-19 patients, Diao et al. showed that levels of PD-1 increased from prodromal to symptomatic stages of the disease (^{Diao et al., 2020}). PD-1 expression is commonly associated with T cell exhaustion, but it is important to emphasize that PD-1 is primarily induced by TCR signaling; it is thus also expressed by activated effector T cells (^{Ahn et al., 2018}).

In addition, several studies reported higher expression of various co-stimulatory and inhibitory molecules such as OX-40 and CD137 (^{Zhou et al., 2020c}), CTLA-4 and TIGIT (^{Zheng et al., 2020a}), and NKG2a (^{Zheng et al., 2020b}). Reduced numbers of CD28⁺ CD8 T cells (^{Qin et al., 2020}) as well as larger frequencies of PD-1⁺ TIM3⁺ CD8 T cells in ICU patients were also reported (^{Zhou et al., 2020c}).

Expression of most of these markers was found to be higher in CD8 than in CD4 T cells, and levels tended to increase in severe versus non-severe cases, which may be due to differences in viral load.

Cellular functionality was shown to be impaired in CD4 and CD8 T cells of critically ill patients, with reduced frequencies of polyfunctional T cells (producing more than one cytokine) as well as generally lower IFN-γ and TNF-α production following restimulation with phorbol myristate acetate (PMA) and ionomycin (^{Chen et al., 2020c, Zheng et al., 2020a, Zheng et al., 2020b}).

Similarly, Zheng et al. reported that CD8 T cells in severe COVID-19 appear less cytotoxic and effector-like with reduced CD107a degranulation and granzyme B (GzmB) production (^{Zheng et al., 2020b}). In contrast, a different study found that both GzmB and perforin were increased in CD8 T cells of severely sick patients (^{Zheng et al., 2020a}).

In accordance with the latter observation, when compared to a moderate disease group, convalescent patients with resolved severe SARS-CoV-1 infection had significantly higher frequencies of polyfunctional T cells, with CD4 T cells producing more IFN-γ, TNF-α, and IL-2 and CD8 T cells producing more IFN-γ, TNF-α, and CD107a, respectively (^{Li et al., 2008}).

However, given the vigorous dynamics of acute T cell responses and potential differences in sample timing throughout disease course, these observations are not necessarily mutually exclusive. Accordingly, RNA sequencing (RNA-seq) data by Liao et al. showed that CD8 T cells in the BAL fluid of severe COVID-19 patients express cytotoxic genes such as GZMA, GZMB, and GZMK at higher levels, while KLRC1 and XCL1 are enriched in mild cases (^{Liao et al., 2020}).

In summary, T cells in severe COVID-19 seem to be more activated and may exhibit a trend toward exhaustion based on continuous expression of inhibitory markers such as PD-1 and TIM-3 as well as overall reduced polyfunctionality and cytotoxicity. Conversely, recovering patients were shown to have an increase in follicular helper CD4 T cells (T_FH) as well as decreasing levels of inhibitory markers along with enhanced levels of effector molecules such as Gzm A, GzmB, and perforin (^{Thevarajan et al., 2020, Yang et al., 2020a, Zheng et al., 2020b}). Collectively, these studies provide a first glimpse into T cell dynamics in acute SARS-CoV-2 infection, but any conclusions have to be tempered at this stage on account of significant limitations in many of the current investigations.

B Cell Responses

Acute B Cell and Antibody Responses

The humoral immune response is critical for the clearance of cytopathic viruses and is a major part of the memory response that prevents reinfection. SARS-CoV-2 elicits a robust B cell response, as evidenced by the rapid and near-universal detection of virus-specific IgM, IgG and IgA, and neutralizing IgG antibodies (nAbs) in the days following infection.

The kinetics of the antibody response to SARS-CoV-2 are now reasonably well described (^{Huang et al., 2020a}).

Similar to SARS-CoV-1 infection (^{Hsueh et al., 2004}), seroconversion occurs in most COVID-19 patients between 7 and 14 days after the onset of symptoms, and antibody titers persist in the weeks following virus clearance (Figure 4 ) (^{Haveri et al., 2020, Lou et al., 2020, Okba et al., 2020, Tan et al., 2020b, Wölfel et al., 2020, Wu et al., 2020b, Zhao et al., 2020a}).

Antibodies binding the SARS-CoV-2 internal N protein and the external S glycoprotein are commonly detected (^{Amanat et al., 2020, Ju et al., 2020, To et al., 2020}). The receptor binding domain (RBD) of the S protein is highly immunogenic, and antibodies binding this domain can be potently neutralizing, blocking virus interactions with the host entry receptor, ACE2 (^{Ju et al., 2020, Wu et al., 2020b}).

Anti-RBD nAbs are detected in most tested patients (^{Ju et al., 2020, To et al., 2020, Wu et al., 2020b}). Although crossreactivity to SARS-CoV-1 S and N proteins and to MERS-CoV S protein was detected in plasma from COVID-19 patients, no crossreactivity was found to the RBD from SARS-CoV-1 or MERS-CoV. In addition, plasma from COVID-19 patients did not neutralize SARS-CoV-1 or MERS-CoV (^{Ju et al., 2020}).

RBD-specific CD19⁺IgG⁺ memory B cells were single-cell sorted from a cohort of eight COVID-19 donors between days 9 and 28 after the onset of symptoms (^{Ju et al., 2020}). From their antibody gene sequences, 209 SARS-CoV-2-specific monoclonal antibodies were produced.

The monoclonal antibodies had a diverse repertoire, relatively low or no somatic mutations, and variable binding reactivity, with dissociation constants reaching 10⁻⁸ to 10⁻⁹, similar to antibodies isolated during acute infections.

Two potent neutralizing SARS-CoV-2 RBD-specific monoclonal antibodies were characterized that did not crossreact with the RBD of SARS-CoV-1 or MERS-CoV (^{Ju et al., 2020}). Together, these results demonstrate that antibody mediated neutralization is virus specific and likely driven by binding of epitopes within the RBD.

B Cell Memory: Development and Lifespan

The B cell response to a virus serves not only to protect from the initial challenge, but also to offer extended immunity against reinfection. Following resolution of an infection, plasma cells formed during the acute and convalescent phases continue to secrete antibodies, giving rise to serological memory.

Memory B cells that are also formed during the primary infection constitute the second arm of B cell memory. Memory B cells can quickly respond to a reinfection by generating new high-affinity plasma cells. Long-term protection is achieved through the induction of long-lived plasma cells and memory B cells.

There is great interest in understanding the lifespan of B cell memory responses to SARS-CoV-2. Protection from reinfection has direct medical and social consequences as the world works to develop vaccination strategies and resume normal activities.

In COVID-19 patients, evidence of near-universal seroconversion and the lack of substantial descriptions of reinfection point to a robust antibody response, which, along with the T cell memory response, would offer protection to reinfection. Indeed, a case study of a single patient described induction of CD38^HiCD27^Hi antibody-secreting cells (ASCs), concomitant with an increase in circulating follicular T helper cells (Tfh) cells (^{Thevarajan et al., 2020}), and a scRNA-seq study of PBMCs from critically ill and recently recovered individuals revealed a plasma cell population (^{Guo et al., 2020}).

In addition, IgG memory cells specific to the RBD have been identified in the blood of COVID-19 patients (^{Ju et al., 2020}). Consistent with the development of immunity after COVID-19 infection, a recent study of SARS-CoV-2 infection in rhesus macaques found that two macaques that had resolved the primary infection were resistant to reinfection 28 days later (^{Bao et al., 2020b}).

Due to the timing of this outbreak, it is not yet possible to know the nature and extent of long-term memory responses, but lessons may again be learned from other human CoVs. In the case of the human CoV 229E, specific IgG and nAbs are rapidly induced but wane in some individuals around a year after infection, with some residual protection to reinfection (^{Callow et al., 1990, Reed, 1984}).

The lifespan of the humoral response following SARS-CoV-1 infection is also relatively short, with the initial specific IgG and nAb response to SARS-CoV-1 diminishing 2–3 years after infection and nearly undetectable in up to 25% of individuals (^{Cao et al., 2007, Liu et al., 2006}).

A long-term study following 34 SARS-CoV-1-infected healthcare workers over a 13-year period also found that virus-specific IgG declined after several years, but the authors observed detectable virus-specific IgG 12 years after infection (^{Guo et al., 2020}). In the case of MERS-CoV, antibodies were detected in six of seven volunteers tested 3 years after infection (^{Payne et al., 2016}).

IgG specific to SARS-CoV-2 trimeric spike protein was detectable in serum up to 60 days after symptom onset, but IgG titers began decreasing by 8 weeks post symptom onset (^{Adams et al., 2020}). Long-term protection from reinfection may also be mediated by reactive memory B cells.

A study that analyzed SARS-CoV-1 S protein-specific IgG memory cells at 2, 4, 6, and 8 months post infection found that S-specific IgG memory B cells decreased progressively about 90% from 2 to 8 months after infection (^{Traggiai et al., 2004}). A further retrospective study of 23 individuals found no evidence of circulating SARS-CoV-1-specific IgG⁺ memory B cells 6 years after infection (^{Tang et al., 2011}). This is in contrast to the memory T cell response, which was robustly detected based on induced IFN-γ production (^{Tang et al., 2011}).

Studies of common CoVs SARS-CoV-1 and MERS-CoV indicate that virus-specific antibody responses wane over time and, in the case of common CoVs, result in only partial protection from reinfection. These data suggest that immunity to SARS-CoV-2 may diminish following a primary infection, and further studies will be required to determine the degree of long-term protection (Figure 4).

Consequences of the B Cell Response: Protection versus Enhancement

Several studies have demonstrated that high virus-specific antibody titers to SARS-CoV-2 are correlated with greater neutralization of virus in vitro and are inversely correlated with viral load in patients (Figure 4) (^{Okba et al., 2020, Wölfel et al., 2020, Zhao et al., 2020a}).

Despite these indications of a successful neutralizing response in the majority of individuals, higher titers are also associated with more severe clinical cases (^{Li et al., 2020b, Okba et al., 2020, Zhao et al., 2020a, Zhou et al., 2020a}), suggesting that a robust antibody response alone is insufficient to avoid severe disease (Figure 4).

This was also observed in the previous SARS-CoV-1 epidemic, where neutralizing titers were found to be significantly higher in deceased patients compared to patients who had recovered (^{Zhang et al., 2006}). This has led to concerns that antibody responses to these viruses may contribute to pulmonary pathology via antibody-dependent enhancement (ADE) (Figure 4).

This phenomenon is observed when non-neutralizing virus-specific IgG facilitate entry of virus particles into Fc-receptor (FcR) expressing cells, particularly macrophages and monocytes, leading to inflammatory activation of these cells (^{Taylor et al., 2015}).

A study in SARS-CoV-1-infected rhesus macaques found that anti-S IgG contributed to severe acute lung injury (ALI) and massive accumulation of monocytes and macrophages in the lung (^{Liu et al., 2019}).

Furthermore, serum containing anti-S Ig from SARS-CoV-1 patients enhanced the infection of SARS-CoV-1 in human monocyte-derived macrophages in vitro (^{Yip et al., 2014}).

ADE was also reported with a monoclonal antibody isolated from a patient with MERS-CoV (^{Wan et al., 2020c}). Somewhat reassuringly, there was no evidence of ADE mediated by sera from rats vaccinated with SARS-CoV-2 RBD in vitro (^{Quinlan et al., 2020}) nor in macaques immunized with an inactivated SARS-CoV-2 vaccine candidate (^{Gao et al., 2020c}).

As of now, there is no evidence that naturally developed antibodies toward SARS-CoV-2 contribute to the pathological features observed in COVID-19. However, this possibility should be considered when it comes to experimental design and development of therapeutic strategies.

Importantly, in all of the descriptions of ADE as it relates to CoV, the FcR was necessary to trigger the antibody-mediated pathology. High-dose intravenous immunoglobulin (IVIg), which may blunt ADE, has been trialed in COVID-19 patients (^{Cao et al., 2020b, Shao et al., 2020}), but further studies are needed to determine the extent to which IVIg is safe or beneficial in SARS-CoV-2 infection.

Vaccine trials will need to consider the possibility of antibody-driven pathology upon antigen rechallenge; strategies using F(ab) fragments or engineered Fc monoclonal antibodies may prove particularly beneficial in this setting (^{Amanat and Krammer, 2020}).

Predictors of COVID-19 Disease Risk and Severity

With the rapidly growing number of cases in the first few months, numerous reports on predictors of COVID-19 severity with small cohorts were released. These offered clinicians and immunologists the first understanding of the clinical course and pathological processes that are associated with the novel SARS-CoV-2 infection. This section highlights key findings from those studies, with a major focus on the immune factors associated with disease risk or severity.

Susceptibility and Risk Biomarkers

There are currently limited known risk factors for susceptibility to COVID-19, although this has been evaluated in several studies. Zhao et al. compared the ABO blood group distribution in a cohort of 2,173 COVID-19 patients to that of healthy controls from the corresponding regions (^{Zhao et al., 2020b}).

They found blood group A to be associated with a higher risk for acquiring COVID-19 when compared to non-A blood groups; blood group O had the lowest risk for the infection. Another study demonstrated an identical association (^{Zietz and Tatonetti, 2020}), and similar results have been previously described for other viruses (^{Lindesmith et al., 2003}), including SARS-CoV-1 (^{Cheng et al., 2005a}).

Several large collaborative efforts are currently underway to generate, share, and analyze genetic data to understand the links between human genetic variation and COVID-19 susceptibility and severity, the most prominent of which is the COVID-19 Host Genetics Initiative (covid19hg.org).

These studies are supported by previous observations on SARS-CoV-1 that followed the 2003 outbreak, which have identified significant associations between genetic variants and immune phenotypes (^{Chan et al., 2007, Wang et al., 2011, Zhao et al., 2011}).

Although identifying such polymorphisms and their associated genes and pathways for SARS-CoV-2 will require large cohorts, several studies, which remain to be tested in clinical trials, have already highlighted genetic polymorphisms that may potentially impact susceptibility.

These studies have focused on genetic variants that may impact the expression or function of genes important in viral entry, namely ACE2 (SARS-CoV-2 receptor) and TMPRSS2 (spike protein activator) (^{Asselta et al., 2020, Cao et al., 2020c, Renieri et al., 2020, Stawiski et al., 2020}). Cao et al. identified variants that are potentially expression quantitative trait loci (eQTL) of ACE2 (i.e., they may potentially alter ACE2 gene expression) and analyzed their frequencies in different populations (^{Cao et al., 2020c}). Stawiski et al. listed variants that may be critical in ACE2 binding and thereby its function and compared the frequencies of these variants within different populations (^{Stawiski et al., 2020}).

While there are several limitations to these studies, the major question is whether the utility of these biomarkers is replicable in large populations with COVID-19 clinical outcomes data and in targeted or large-scale genomic analyses that are currently underway. In addition, these studies will reveal the potential associations between genetic variants and susceptibility in a gene or loci agnostic fashion.

Routine Bloodwork Biomarkers

Several routine blood and serological parameters have been suggested to stratify patients who might be at higher risk for complications to aid in allocation of healthcare resources in the pandemic (Table 1 ). Serologic markers from routine bloodwork were reported by comparing patients with mild or moderate symptoms to those with severe symptoms.

This includes different acute phase proteins, such as SAA (serum amyloid protein) and C-reactive protein (CRP) (^{Ji et al., 2020}). Interestingly, elevations in CRP appear to be unique to COVID-19 patients when compared to other viral infections. Other consistently reported markers in non-survivors are increased procalcitonin (PCT) and IL-6 levels (^{Huang et al., 2020d}), as well as increased serum urea, creatinine, cystatin C, direct bilirubin, and cholinesterase (^{Xiang et al., 2020a}).

Overall, inflammatory markers are common in severe cases of COVID-19 and appear to correlate with the severity of the symptoms and clinical outcome. Moreover, the extensive damage that occurs in specific organs of severe COVID-19 patients is possibly related to differences in the expression of ACE2 (Figure 5 ) (^{Du et al., 2020}).

Table 1

Routine Blood and Immunological Prognostic Biomarkers in COVID-19 Patients

	Biomarker	Purpose
Routine Bloodwork	Lymphocyte count	Predicted the disease severity and the outcomes of hospitalized patients (Tan et al., 2020a). Prognostic value was confirmed in numerous studies (Huang et al., 2020d, Liu et al., 2020b, Song et al., 2020, Wang et al., 2020f, Wynants et al., 2020, Yan et al., 2020b, Yang et al., 2020b, Zhang et al., 2020c, Zhao et al., 2020d). Decreased continuously in non-surviving patients (Wang et al., 2020b).
N/L	Patients with N/L ≥3.13 were reported to be more likely to develop severe illness and to require ICU admission (Liu et al., 2020c). N/L on admission was a risk factor for short-term progression of patients with moderate pneumonia to severe pneumonia (Feng et al., 2020). Confirmed to be of prognostic value in COVID-19 in several studies (Song et al., 2020, Wynants et al., 2020, Zhou et al., 2020d).
CRP	Proposed as an early biomarker of disease progression (Ji et al., 2020). Even in early stages, CRP levels were positively correlated with lung lesions and reflected disease severity (Wang, 2020). Confirmed in numerous studies (Fu et al., 2020, Huang et al., 2020d, Wynants et al., 2020, Yan et al., 2020b, Zhao et al., 2020d, Zhou et al., 2020b). Predicted the risk of acute myocardial injury (Liu et al., 2020g, Xu et al., 2020a).
LDH	Higher in severe cases than in mild cases (Xiang et al., 2020a). Widely proposed to have prognostic value in COVID-19 (Huang et al., 2020d, Wynants et al., 2020, Yan et al., 2020b, Zhao et al., 2020d).
D-dimer (and coagulation parameters)	Predicted severity independently of other variables (Zhou et al., 2020d). Elevated levels and disseminated intravascular coagulation are found in non-survivors (Wang et al., 2020b). Identified patients at risk for acute cardiac injury (Liu et al., 2020g). Other coagulation parameters, such as fibrin degradation product levels, longer prothrombin time, and activated partial thromboplastin time, were also associated with poor prognosis (Tang et al., 2020a).
SAA	SAA was proposed to be used as an auxiliary index for diagnosis as it was elevated in 80% of the patients in a small cohort (Ji et al., 2020).
NT-proBNP (N-terminal pro B type natriuretic peptide)	NT-proBNP was an independent risk factor of in-hospital death in patients with severe COVID-19 (Gao et al., 2020b).
Platelet count	High platelet-to-lymphocyte ratio was associated with worse outcome (Qu et al., 2020). Thrombocytopenia was associated with poor outcome and with incidence of myocardial injury in COVID-19 (Liu et al., 2020h, Shi et al., 2020).
Immunological	CD4+, CD8+, and NK cell counts	Lower CD4+, CD8+, and NK cells in PBMCs correlated with severity of COVID-19 (Nie et al., 2020b). Validated by several studies (Wang et al., 2020f, Zheng et al., 2020b).
PD-1 and Tim-3 expression on T cells	Increasing PD-1 and Tim-3 expression on T cells could be detected as patients progressed from prodromal to overtly symptomatic stages (Diao et al., 2020). Expression was higher in infected patients versus healthy controls and in ICU versus non-ICU patients in both CD4 and CD8 T cells (Zhou et al., 2020b).
phenotypic changes in peripheral blood monocytes	The presence of a distinct population of monocytes with high forward scatter (CD11b+, CD14+, CD16+, CD68+, CD80+, CD163+, and CD206+, which secrete IL-6, IL-10, and TNF-α) was identified in patients requiring prolonged hospitalization and ICU admission (Zhang et al., 2020c). CD14+CD16+IL-6+ monocytes are increased in ICU patients (Zhou et al., 2020b).
IP-10, MCP-3, and IL-1ra	IP-10, MCP-3, and IL-1ra were, among 48 examined cytokines, the only ones that closely associated with disease severity and outcome of COVID-19 in a study by Yang et al. (Yang et al., 2020b).
IL-6	Associated with disease severity (hospitalization and ICU admission) and poor prognosis (Chen et al., 2020g, Huang et al., 2020b, Liu et al., 2020b, Liu et al., 2020f, Wang et al., 2020b). Increased levels were associated with higher risk of respiratory failure (Yao et al., 2020b).
IL-8	Positively correlated with disease severity (Chen et al., 2020e, Gong et al., 2020), with severe cases showing the highest IL-8 levels.
IL-10	Increased in severe or critical patients as compared to mild patients (Gong et al., 2020, Zhou et al., 2020d) without a statistically significant difference between severe and critical cases (Gong et al., 2020).
IL-2R	Associated with disease severity in a study that, among other cytokines, also associated ferroprotein levels, PCT levels, and eosinophil counts with COVID-19 severity (Gong et al., 2020).
IL-1β	CD14+IL-1β+ monocytes are abundant in early-recovery patients as shown in a single-cell RNA-seq analysis and thought to be associated with cytokine storm (Wen et al., 2020). IL-1β did not correlate with disease severity in a cross-sectional study with mild, severe, and critical patients (Gong et al., 2020).
IL-4	IL-4 was associated with impaired lung lesions (Fu et al., 2020), but some reports point to a potential mediator effect (Wen et al., 2020).
IL-18	In modeling immune cell interaction between DCs and B cells in late recovery COVID-19 patients, IL-18 was found to be important in B cell production of antibodies, which suggests its importance in recovery (Wen et al., 2020).
GM-CSF	GM-CSF+IFN-γ+ T cells are higher in ICU than in non-ICU patients. CD14+CD16+GM-CSF+ monocytes are higher in COVID-19 patients as compared to healthy controls (Zhou et al., 2020b).
IL-2 and IFN-γ	IL-2 and IFN-γ levels were shown to be increased in severe cases (Liu et al., 2020b).
anti-SARS-CoV-2 antibody levels	Prolonged SARS-CoV-2 IgM positivity could be utilized as a predictive factor for poor recovery (Fu et al., 2020). Higher anti-SARS-CoV-2 IgG levels and higher N/L were more commonly found in severe cases (Zhang et al., 2020a).

Lymphopenia is the most frequently described prognostic marker in COVID-19 (Table 1), and it appears to predict morbidity and mortality even at early stages (^{Fei et al., 2020}).

Tan et al. proposed a prognostic model based on lymphocyte counts at two time points: patients with less than 20% lymphocytes at days 10–12 from the onset of symptoms and less than 5% at days 17–19 had the worst outcomes in this study (^{Tan et al., 2020a}).

Wynants et al. compared predictors of disease severity across seven studies (>1,330 patients), highlighting CRP, neutrophil-to-lymphocyte ratio (N/L), and lactate dehydrogenase (LDH) as the most significant predictive biomarkers (^{Wynants et al., 2020}). Furthermore, a meta-analysis of 30 COVID-19 studies with a total of 53,000 patients also attempted to identify early-stage patients with poor prognosis (^{Zhao et al., 2020d}).

The most consistent findings across the different studies were elevated levels of CRP, LDH, and D-dimer, as well as decreased blood platelet and lymphocyte counts (^{Yan et al., 2020b, Zhou et al., 2020d}). Systemic and pulmonary thrombi have been reported with activation of the extrinsic coagulation cascade, involving dysfunctional endothelium and monocytic infiltration (^{Poor et al., 2020, Varga et al., 2020}); thrombocytopenia and elevated D-dimer levels may be indicative of these coagulopathies in COVID-19 patients with important therapeutic implications (^{Fogarty et al., 2020, Poor et al., 2020}).

Immunological Biomarkers in the Peripheral Blood

Immunological biomarkers are particularly important, as immunopathology has been suggested as a primary driver of morbidity and mortality with COVID-19. Several cytokines and other immunologic parameters have been correlated with COVID-19 severity (Table 1). Most notably, elevated IL-6 levels were detected in hospitalized patients, especially critically ill patients, in several studies and are associated with ICU admission, respiratory failure, and poor prognosis (^{Chen et al., 2020g, Huang et al., 2020b, Liu et al., 2020f}).

Increased IL-2R, IL-8, IL-10, and GM-CSF have been associated with disease severity as well, but studies are limited, and further studies with larger cohorts of patients are needed to indicate predictive power (^{Gong et al., 2020, Zhou et al., 2020b}).

Conflicting results regarding IL-1β and IL-4 have been reported (^{Fu et al., 2020, Gong et al., 2020, Wen et al., 2020}). Although elevated cytokine concentrations have been widely described in COVID-19 patients, the vast majority (including IL-6, IL-10, IL-18, CTACK, and IFN-γ) do not seem to have prognostic value, because they do not always differentiate moderate cases from severe cases (^{Yang et al., 2020b}).

This stratification was possible with IP-10, MCP-3, and IL-1ra. While there are reports that levels of IL-6 at first assessment might predict respiratory failure (^{Herold et al., 2020}), other publications with longitudinal analyses demonstrated that IL-6 increases fairly late during the disease’s course, consequently compromising its prognostic value at earlier stages (^{Zhou et al., 2020a}).

Liu et al. developed a web-based tool using k-means clustering to predict prognosis in terms of death or hospital discharge of COVID-19 patients using age, comorbidities (binary), and baseline log helper T cell count (TH), log suppressor T cell count (TS), and log TH/TS ratio (^{Liu et al., 2020e}). Total T cell, helper T cell, and suppressor T cell counts were significantly lower and the TH/TS ratio was significantly higher in patients who died from infection as compared to patients who were discharged.

Importantly, most serological and immunological changes observed in severe cases are associated with disease severity but cannot necessarily serve as predictive factors, as they may not have utility in early identification of patients at higher risk. Discovery of truly predictive biomarkers and potential drivers of hyperinflammatory processes requires comprehensive profiling of asymptomatic and mild cases and longitudinal studies that are limited to date.

Confounding variables including age, gender, and comorbidities may dramatically affect associations observed. In addition, direct correlation with patient viral load will be important to provide a greater understanding of underlying causes of morbidity and mortality in COVID-19 and the contribution of viral infectivity, hyperinflammation, and host tolerance (^{Medzhitov et al., 2012}).

In summary, lymphopenia, increases in proinflammatory markers and cytokines, and potential blood hypercoagulability characterize severe COVID-19 cases with features reminiscent of cytokine release syndromes. This correlates with a diverse clinical spectrum ranging from asymptomatic to severe and critical cases.

During the incubation period and early phase of the disease, leukocyte and lymphocyte counts are normal or slightly reduced. After SARS-CoV-2 binds to ACE2 overexpressing organs, such as the gastrointestinal tracts and kidneys, increases in non-specific inflammation markers are observed. In more severe cases, a marked systemic release of inflammatory mediators and cytokines occurs, with corresponding worsening of lymphopenia and potential atrophy of lymphoid organs, impairing lymphocyte turnover (^{Terpos et al., 2020}).

---

More information: “Antigen-specific adaptive immunity to SARS-CoV-2 in acute COVID-19 and associations with age and disease severity” Cell (2020). DOI: 10.1016/j.cell.2020.09.038