Ebola Immunity: Gaining a Winning Position in Lightning Chess

In 1976, the first documented case of Ebola infection was reported in sub-Saharan Africa (1), caused by the prototypical Zaire ebolavirus (EBOV), whose most recent emergence in 2014 caused a global public health emergency with more than 28,000 cases and 11,000 deaths (2). From 1976 to 2014, the recognized outbreaks of EBOV had a cumulative reported case count of ∼1500 people and an average fatality rate of 78% (2). Most of these EBOV outbreaks were rapidly contained because of the small, isolated populations involved, the limited population movement outside of those areas, and the effective implementation of contact tracing and quarantine, but the West African epidemic illustrated the potential for uncontrolled spread. Analysis of index cases from outbreaks suggests that apes and other mammals can serve as intermediate hosts, but the natural reservoir for EBOV may be bats (3). Following the initial jump into humans, transmission occurs from person to person via close contact with infected secretions or fluids (4–6). The assumption that EBOV persists sublethally in its natural reservoir but causes rapid disease progression and death in humans highlights the role of host immune system in determining infection outcomes.

Insights into the natural history of EBOV infection in humans have been obtained through contact tracing and exposure history and have established the EBOV incubation period to be from 2 to 21 d after initial contact with the virus (5, 7). The delay between self-reporting of symptoms (fever, cough, rash, abdominal pain) and urgent care admission hinders the collection of immediate postexposure samples, so the majority of human biological samples are not obtained until days after the onset of symptoms (on average 6 d postonset) (Fig. 1) (8–14), resulting in a dearth of data on early infection events and initial host immune responses. During the next phase of the disease, infected individuals display extreme fatigue, headache, vomiting, diarrhea, dyspnea, hypovolemic shock, and organ failure, with death generally occurring from 7 to 14 d after the onset of initial symptoms (9, 11, 12, 14, 15). Survival is broadly associated with a lower viral load at the time of admission (16) and, although previously underappreciated, postrecovery, long-term sequelae have been identified in follow-up studies in West African survivor cohorts (17, 18).

Although there are limited historical case reports suggesting that EBOV replicates in sites of immunological privilege, the 2014–2016 epidemic provided reproducible case data on EBOV in the CNS, eye, and gonads (4, 19–26). These observations highlighted the fact that EBOV can persist in the infected host long after viral clearance from the circulation (plasma and urine) in convalescent individuals (Fig. 1). This theory is further supported by case reports of possible sexual transmission events from survivors to their partners long after symptom resolution (6) and recurrence of viremia potentially seeded from immunologically privileged sites (22). The failure of the immune system to prevent the establishment of viral persistence and reinitiation of infection is a newly recognized immunological challenge for the clearance of EBOV, adding to the spectrum of events from very early infection to long-term sequelae that require elucidation.

Because of the difficulty of early case identification and sampling during outbreaks, most data from infected humans are limited to the symptomatic phase of the disease. As a result, early events that preceded the onset of symptoms, in which virus–host dynamics are key to determining infection outcome, remain a critical gap in knowledge. If EBOV behaves similarly to what has been observed for other RNA viruses (hepatitis C virus, HIV, influenza virus), transmission will occur by a tiny fraction of selected variants from the exposing virus swarm that can successfully establish infection in a new host (27–29). After the founding event, host immune selective pressure and mutations introduced by the viral polymerase shape a new virus swarm that replicates to high circulating titers (27, 30, 31). Although extensive sequencing has been performed to characterize circulating EBOV after systemic spreading of the infection, little is known about the transmitted founder virus, which may be the most relevant for vaccine targeting and evaluation in animal models (32–34).

Mouse EBOV models require species adaption of the virus away from the human strain to overcome the murine type I IFN response and generate lethal infection (35). Collaborative Cross inbred mice lethally infected with mouse-adapted virus develop some signs of hemorrhage, a hallmark of human EBOV pathogenesis (36). However, mouse-adapted EBOV in most other mouse models does not result in an infection that fully recapitulates the fever, rash, or hemorrhagic clinical manifestations observed in humans (37), nor the immune responses, such as TNF-α production (38). Early insights into protective immunity are derived primarily from vaccination in small animal models (39). Although mouse models are still used to screen preliminary vaccine candidates, observations about vaccine or therapeutic efficacy from these models are often not transferrable to large animal models (40, 41).

Ebola disease in macaques shares many of the same clinical features as human infection, but time to death is accelerated, and the mortality rate (100%) greatly exceeds that in humans, likely due to the high dose infectious challenge (more than 100 × LD₉₉) and i.m. route of exposure in experimental infections. However, similar to human infection, EBOV elicits disease characterized by fever, rash, anorexia, multiple organ failure, and coagulation dysfunctions in macaques (42). Although the challenge route may result in Ag recognition by different populations of immune cells than would be present at mucosal or skin surfaces exposed during natural human infection, similarities such as the kinetics of immune responses relative to phase of infection, lymphopenia, and a late-stage cytokine storm suggest this model can yield insights into the mechanisms of immunologic control of EBOV. In light of the similarities between the macaque model and human disease, this review will focus primarily on studies that characterize the immune responses to EBOV infection in these two hosts.

EBOV replicates in the spleen and lymph nodes prior to systematic dissemination, at which point circulating virus can be detected (43). Early infection of monocytes, macrophages, and dendritic cells (DCs) is the gateway for virus dissemination, and represents the initial assault on the host immune system by viral proteins (Fig. 2). Thus, the contest between host and virus starts within these immune cells, which not only constitute the first line of immune defense, but are also pivotal for initiating the adaptive immune response.

When EBOV infects cells in vitro, viral dsRNA formed in the cytoplasm during the replication cycle is recognized by the pattern recognition receptors, the retinoic acid–inducible gene I (RIG-I), and melanoma differentiation–associated protein 5 (MDA5), which leads to induction of type I IFN responses and the expression of IFN-stimulated genes (ISGs) that inhibit viral replication (44, 45). In vivo, plasma IFN-α and upregulation of ISGs become detectable after the onset of symptoms in both EBOV patients and infected macaques (43, 46–49). However, instead of the transient peak that is observed in nonlethal viral infections (50–52), plasma IFN-α remains high throughout the remainder of the EBOV disease course in macaques (43, 53). Despite the large amount of circulating IFN-α, EBOV viral load in macaques continues to rise to >10⁶ PFUs/ml, and the outcome is uniformly fatal (43).

There are several possible explanations for the failure of observed type I IFN responses to promote protective immunity. First, there is evidence that type I IFN responses could be blunted by EBOV antagonism at steps downstream of IFN production. For example, EBOV protein VP24 blocks the IFN signaling pathway by preventing nuclear localization of a transcription factor STAT1, which is essential for ISG expression (54). As another example, EBOV glycoprotein antagonizes the action of tetherin, one of the ISGs that prevents virus budding (54). Second, the timing and magnitude of the type I IFN responses affect its antiviral impacts. Early type I IFN responses induced by the administration of exogenous type I IFN protect animals against lymphocytic choriomeningitis virus, severe acute respiratory syndrome coronavirus, and SIV infection (55–57). In contrast, delayed type I IFN responses, due to either the ablation of early type I IFN producing cells or viral antagonism, are associated with rapid virus propagation (57, 58), reduced frequencies of virus-specific CD8⁺ T cells (55, 58), and pathogenic inflammation (55). Further, excessive type I IFN responses can be detrimental, as observed in chronic lymphocytic choriomeningitis virus infection, in which CD8⁺ T cell exhaustion is driven by immune activation and the upregulation of inhibitory receptors (59, 60). In EBOV infection, initial viral replication takes place in the spleen and lymph nodes, days before the virus spreads through the circulation and the rise of systematic type I IFN responses (43). It is possible that local type I IFN responses at the tissue level prior to the onset of viremia are critical to determining survival outcomes.

The strength of tissue-level type I IFN responses during the earliest phase of virus replication is likely determined by DCs and macrophages, as they are the first cell types associated with EBOV RNA in lymph nodes (43). Monocyte-derived DCs (mDCs) do not secrete IFN-α, nor do they undergo normal maturation when infected by EBOV in vitro (61, 62). This is largely due to suppression of the type I IFN response by VP24 (63). VP35 is another viral protein that interferes with the activation of RIG-I, prevents the recognition of viral dsRNA by RIG-I/MDA5, and disrupts signaling events triggered by RIG-I/MAD5 engagement (45). In monocyte-derived macrophages (MDMs), in vitro EBOV infection induces IFN-α production between 9 and 72 h (46, 64). However, the level of IFN-α induction is limited by viral IFN antagonists, as shown by reduced IFN-α production by EBOV-infected MDMs in response to a synthetic IFN-α stimulant (46). The ability of EBOV proteins to impair IFN-α production by MDMs upon viral infection is further illustrated with a recombinant Newcastle disease virus encoding VP35, which induces 8000 times lower IFN production in infected MDMs as compared with a Newcastle disease virus encoding no VP35 (65). Impairment of the local type I IFN response by EBOV before the onset of viremia could profoundly affect subsequent viral–host interactions through inefficient containment of local viral replication, failure to activate innate immune cells such as NK cells, and poor priming of T cell responses in the absence of type I IFN responses (Fig. 2). This hypothesis is indirectly supported by experiments in which treatment of macaques on the day of infection with IFN-α2b extends survival time by 2 d (66), and early treatment with IFN-β extends mean survival time from 8 to 13 d (67).

Systemic inflammation is a key feature of EBOV infection. Proinflammatory cytokines such as IL-6 and TNF-α, as well as anti-inflammatory chemokines such as IL-10, are detected in the circulation following the onset of symptoms in humans and macaques (43, 48, 68–72). In survivors of EBOV infection, the level of circulating cytokines peaks briefly and subsides as infection is controlled. However, in nonsurvivors, cytokine levels continue to increase despite the presence of high levels of IL-10 (68, 70), resulting in uncontrolled systemic inflammation analogous to that seen in bacterial and fungal sepsis (73).

In contrast to impaired type I IFN responses, inflammatory responses are effectively triggered by EBOV infection in early target cells. MIP-1α and MIP-1β chemokines are secreted by both mDCs and macrophages upon EBOV infection in vitro (62, 64). In addition, macrophages infected in vitro produce inflammatory cytokines such as TNF-α, reactive oxygen species, and reactive nitrogen species (61, 64, 74). These data suggest that EBOV is capable of inducing local inflammation early in infection (Fig. 2). The fact that the type I IFN response and inflammatory response are differentially regulated by EBOV within the same target cells is reminiscent of innate immune modulation by 1918 influenza virus infection (75). In contrast to a seasonal influenza virus that induces strong type I IFN responses and transient inflammation in the bronchi of infected macaques, pandemic 1918 influenza virus suppresses the type I IFN response but induces strong and sustained inflammation in the lung (75). High levels of chemokines and cytokines result in infiltration of inflammatory cells, eventually leading to severe hemorrhage (76). Dysregulation of innate responses at an early stage of infection may be a common characteristic shared between EBOV and 1918 influenza virus, contributing to the high virulence in both cases.

Following EBOV triggering of inflammatory responses, prominent neutrophilia is observed prior to the onset of viremia in macaques (69), indicating widespread recruitment to sites of infection. This increase in peripheral neutrophils is detectable in either relative or absolute counts, with the latter approaching what is observed with bacterial infections (77, 78). Neutrophils are activated by EBOV infection in vitro, resulting in the secretion of proinflammatory cytokines, which potentially contribute to the overall proinflammatory milieu (79). Through these proinflammatory contributions, it is possible that neutrophils play a role in EBOV immunopathogenesis.

Peripheral NK cells have been reported to transiently decline upon EBOV infection in humans and macaques, suggesting that either cell death or redistribution into infected tissues is occurring (47, 80). Although NK cells are enriched in liver and spleen, which are major replication sites in early infection (43), their contribution to the control of EBOV infection remains to be demonstrated. Genetic analyses among EBOV-infected patients could provide more information regarding the role of NK cells in vivo, as specific allelic combination of killer cell Ig-like receptors and HLA loci has been associated with slow disease progression in HIV infection (81) or virus clearance in hepatitis C virus (82). Furthermore, adoptive transfer of in vitro expanded NK cells or depletion of NK cell subsets in primates can be explored to directly test their importance in protecting the host against EBOV (83).

In EBOV-infected humans and vaccinated macaques, high titers of plasma IgM and IgG are observed in survivors (84, 85). In humans, Ag-specific IgM levels increase coincidentally with viral load decreases, and Ag-specific IgG first appears 1 wk after the appearance of IgM in the blood (Fig. 1) (8, 26, 47, 84). Although an early study administering convalescent whole blood from EBOV survivors to patients suggested a potential therapeutic benefit of Abs, subsequent studies administering either convalescent whole blood or plasma to EBOV-infected humans did not demonstrate a survival advantage in recipients (10, 86–90). In most cases, the amount of Ag-specific IgG in the donor samples was not quantified prior to infusion, so negative results could be explained by insufficient quantities of Ag-specific Ab transferred to recipients. Similar studies in macaques, delivering EBOV-convalescent whole blood to recipients, did not provide protection against challenge; however, as in the human studies, this could be due to the quantity of Ag-specific Ab present, as evidenced by the low IgG titers measured a few days after infusion (91). Follow-up studies used IgG purification and concentration to maximize levels in the recipient but still failed to protect, possibly owing to species mismatch between the donor and recipient, resulting in rapid clearance of Abs postinjection or inferior Fc-mediated effector functions (66, 92). Indeed, in other experiments, species-matched IgG, transferred from EBOV-convalescent to naive macaques, remained elevated in the recipients for several weeks and did provide a survival benefit (93). However, the failure of high-titer, species-matched, vaccine-generated polyclonal IgG to protect in the more stringent cynomolgus macaque model illustrates that the requirements for Ab-mediated protection are multifactorial (94). The requirements for successful passive immunotherapy with Ig must be defined by macaque studies in which the Ag history of the donor (vaccination and challenge survival versus vaccination alone), species, regimen, formulation, and importantly, Ab quality, an aspect not addressed by these studies but likely impacting on protection, are all controlled. The mixed success of Ig transfer is not surprising because EBOV has evolved a morphology and replication strategy that present special challenges for recognition and inhibition by the humoral immune system. EBOV particles are in some cases more than 10 μm in length with a high density of glycoproteins stably anchored to the virion surface, suggesting that a high local concentration of Ab may be required to efficiently coat the virion and block infection (85), with polymorphic virion shapes potentially impeding Ab access to some glyoprotein trimers (95). Perhaps more problematic is the promiscuous use of multiple factors to mediate initial attachment to cells, ensuring that no single Ab specificity will successfully block this critical first step in entry (Fig. 3). This nonspecific attachment is followed by rapid uptake into the cell by macropinocytosis (96–101). Once in the endosome (102), Ab access to the virion is limited (unless Abs have bound prior to virion uptake), and high affinity interactions will be necessary to remain associated under conditions of endosomal acidification (103). Throughout the initial entry steps, the EBOV receptor binding domain (RBD), a primary target for neutralization, is largely protected from Ab recognition because of its placement recessed within the glycoprotein core (Fig. 3). Exposure of the RBD occurs only after cathepsins cleave off ∼130 kDa of the 150 kDa trimer, leaving only a subset of the Ab binding epitopes that were present on the native trimer. This cleavage exposes the glycoprotein RBD for interaction with the intracellular receptor, Niemann–Pick C1 (NPC1), which triggers membrane fusion and ribonucleoprotein complex liberation into the cytoplasm (Fig. 3) (104–108). The success of some polyclonal IgG preparations to protect macaques supports a potential strategy of using mAb mixtures targeting multiple sites on the glycoprotein, for mAbs that do not completely protect when given individually (109). Preclinical studies in macaques have demonstrated that postexposure transfer of mAb mixtures improves survival outcomes (110, 111), and in some studies, complete protection was achieved by infusion of mAb mixtures administered 3–5 d after infectious challenge (109, 112). Moreover, a human clinical trial suggested a modest trend toward greater survival when administration of mAb mixtures in combination with standard of care was compared with standard of care alone (113).

A potential caveat of mAb mixtures is that the approach can be compromised if manufacturing multiple mAbs impedes production and regulatory approval for use, especially under emergency outbreak conditions (114). Ideally, mAbs can be identified to target specific sensitivities in virus entry or on glycoproteins to reduce the number of mAbs required for potent protection. Even in the case of mixtures, formulations may be simplified and potency may be enhanced by eliminating mAbs that compete for glycoprotein binding and defining mAb mechanisms of action (115, 116). For example, Abs that target epitopes in the portion of glycoproteins removed by cleavage, and which demonstrate potent neutralization in vitro, failed to protect animals against challenge (109). An Ab targeting the glycoprotein base that is retained after cleavage, and which shows near complete neutralization in vitro, still failed to control infection in vivo, potentially owing to a lack of stable binding at low pH (117). The most intuitive mechanistic Ab target is inhibition of the critical glycoprotein–NPC1 receptor binding interaction. However, Abs targeting the glycoprotein RBD must gain access to the recessed RBD prior to virion uptake, tolerate acidic conditions in the endosome, and retain high affinity binding after cathepsin cleavage (Fig. 2). Although such Abs appear to be rare, one mAb to date fulfills these criteria (103). Indeed, when delivered up to 5 d after lethal EBOV challenge, this monotherapy mediates uniform protection of macaques with no sign of illness (112).

A vexing challenge in the search for efficacious EBOV mAbs is the limitation of in vitro neutralization assays and small animal models to predict Ab efficacy in the macaque challenge model (109, 117, 118). Viewed as an opportunity, those limitations provide an impetus to define the relative importance in vivo of all potential Ab effector functions, including complement-dependent cytotoxicity, Ab-dependent cellular phagocytosis (ADCP), or Ab-dependent cellular cytotoxicity (ADCC) (Fig. 2) (119). For other viruses, the cooperative interactions between Abs and components of cell-mediated immunity have been demonstrated in vivo by abrogating the FcR-binding functions of neutralizing mAbs and showing a loss of protective capacity in, for example, simian/human immunodeficiency chimeric virus infected macaques or influenza-infected mice (120, 121). However, interpretation of studies using Fc-mutated mAbs should take into account the fact that Fc modification will affect multiple non-ADCC related functions such as mAb distribution in tissues, and Ag uptake for subsequent processing and presentation to the cellular immune system. Because NK cells are enriched in spleen and liver, which are major sites of EBOV replication, ADCC may play a role in controlling viral replication in these organs. Abs can mediate ADCC against EBOV in vitro (112), but its importance in vivo in controlling EBOV replication remains to be defined in primates.

Well-defined EBOV vaccines and monotherapeutic mAbs provide indispensable model systems to identify the role of NK cells and Fc-mediated Ab effector functions in vivo (112, 122), and to determine if there are negative consequences to the host of the FcR–Ab interaction. The rapid dissemination of EBOV in infected individuals and failure of some Ab therapies to protect even when they exhibit in vitro neutralization raises the question of whether EBOV entry is being aided by mechanisms such as Ab-dependent enhancement (ADE) (Fig. 2). ADE has been observed for dengue virus and respiratory syncytial virus infection in humans in which the presence of low levels of Ag-specific Abs contribute to disease severity (123–125). ADE is thought to occur when Ab–virus complexes interact with FcR present on immune cells (e.g., monocytes, macrophages, and neutrophils), or with components of the complement system, resulting in enhanced proximity of the virus to cellular membranes, thus facilitating infection (126). In vitro, the use of plasma derived from macaques and human survivors demonstrated enhancement in EBOV infectivity of an immortalized cell line (127, 128); however, more in vitro research and in particular in vivo studies are required to fully understand what role, if any, ADE plays in the course of natural EBOV infection.

The longstanding belief that immune suppression by EBOV thwarts an effective T lymphocyte response bears reconsideration in light of recent observations that activated CD8⁺ T cells are effectively induced during the symptomatic phase of EBOV infection. Single-cell immunophenotyping revealed that at peak response, 10–60% of CD8⁺ T cells display activation phenotype (72, 87, 129), similar to the level of CD8⁺ T cell activation induced by other virus infections or vaccination (130, 131). CD8⁺ T cells expressed markers associated with cell proliferation (87, 129), cytotoxicity (87), and immunoinhibitory receptors programmed death–1 (PD-1) and CTLA-4 (72, 87). CD8⁺ T cell activation phenotype was first reported among EBOV patients who received advanced care and treatments in the United States and Europe, which could have potentially impacted immune responses (87, 129). However, a similar observation was also made among patients with limited access to advanced care in West African clinics (72), suggesting that the activation of CD8⁺ T cells is not associated with treatments, but rather an intrinsic character of the immune response against EBOV.

The kinetics of CD8⁺ T cell activation could shed light on the possible role of CD8⁺ T cells in controlling EBOV infection. A previous study reported CD8⁺ T cell activation during the recovery phase in survivors (84). However, in more recent studies, CD8⁺ T cell activation was detected during the symptomatic phase (72, 129) and coincided with the decline of viral load (84, 87). The discrepancy in detection of CD8⁺ T cell activation during the symptomatic phase could result from the difference in sample time points and sets of markers analyzed, or the difference in detection methods used. In one study, CD8⁺ T cell activation was inferred from RT-PCR analysis of bulk immune cells (84), whereas the more recent studies monitored CD8⁺ T cell activation by flow cytometric analysis of specific marker-based subpopulations (72, 84, 87, 129). More importantly, CD8⁺ T cells recognizing MHC class I epitopes in EBOV nucleoprotein (NP) are detectable in four survivors early in the symptomatic phase of infection (72). Although this is a small sample size, the temporal association between the accumulation of virus-specific CD8⁺ T cells and a decline in viremia suggests CD8⁺ T cells contribute to the control of EBOV replication in humans (72). However, to identify the role of the CD8⁺ T cell response in controlling EBOV infection, the functionality of EBOV-specific CD8⁺ T cells in the symptomatic phase should be defined.

A similar level of CD8⁺ T cell activation is also observed in nonsurvivors, although the quality of these CD8⁺ T cells may be inferior to those in survivors. For example, a larger proportion of CD8⁺ T cells express the inhibitory receptors PD-1 and CTLA-4 in nonsurvivors than in survivors (72). Furthermore, although NP-specific CD8⁺ T cells are detected in nonsurvivors, they fail to increase in frequency over time, suggesting their prolonged presence in survivors may impact favorably on clinical outcomes (72). Higher viral loads and excessive inflammation found in fatal EBOV cases may drive the increase in PD-1/CTLA-4 expression and impair CD8⁺ T cell function and proliferation (72). It is also possible that failure of the immune system to overcome viral IFN antagonism early in infection and aberrant DC maturation may set the stage for subsequent defects in T cell responses. As a result, the virus replicates without control, leading to leukopenia and T cell death by intravascular apoptosis during the last 5 d of a fatal disease course (84, 132).

In contrast to limited data on human T cell responses against EBOV infection, animal models have provided useful insights for the role of CD8⁺ T cells in controlling EBOV infection, as CD8⁺ T cell responses can be either enhanced through vaccination or eliminated by immunodepletion. In recombinant adenovirus–vaccinated macaques, protection is associated with robust CD8⁺ T cell responses that include IFN-γ and TNF-α dual-expressing, cytotoxic CD8⁺ T cells, in some cases in the absence of a humoral response (133–135). Furthermore, treatment of recombinant adenovirus–vaccinated macaques with a CD8-α–depleting Ab immediately before challenge abolished vaccine-mediated protection (94), suggesting CD8⁺ T cells, and possibly CD8⁺ NK cells, are necessary for virus clearance. A different result was obtained in recombinant vesicular stomatitis virus–vaccinated macaques, in which CD8-α–depletion during the vaccination phase did not affect vaccine-mediated protection (136). The interpretation of the result, however, is complicated owing to the detection of CD8⁺ T cells at a time shown in other studies to immediately precede rapid rebound to near predepletion CD8⁺ T cell levels (137–139). Nevertheless, because immune correlates and mechanisms of protection will likely vary across different vaccine platforms, the relative importance of cellular and humoral responses for protection against EBOV should be determined empirically for each vaccine strategy (140).

CD8⁺ T cells may play an important role beyond viremia clearance in the control of residual replication following recovery and in the prevention of the recurrent infection by EBOV that has persisted in the body. Infectious EBOV has been recovered from breast milk, urine, ocular aqueous humor, and semen of recovering patients (141). In EBOV patients treated at intensive care units who experienced severe disease and high viral load, CD8⁺ T cells remained activated for up to 48 d following the resolution of viremia, suggesting a role for CD8⁺ T cells in clearing residual noncirculating viral Ag (87, 129, 142). More importantly, the majority of these CD8⁺ T cells store significant quantities of intracellular granzyme B (129, 142) and produce cytokines upon EBOV Ag stimulation (87, 142), suggesting that they are fully functional effector memory cells and capable of killing infected cells upon a future resurgence of viral replication. However, treatment such as immunotherapy and antiviral drugs received by these patients could have separately impacted immune responses, including memory CD8⁺ T cell development (87, 129, 142). Therefore, monitoring EBOV-specific cellular responses in recovered survivors who did not received advanced treatment will shed light on the potential role of CD8⁺ T cells in clearing residual EBOV infection.

Despite the high mortality rate of EBOV infection, survival in the absence of therapeutics demonstrates that some hosts are able to mount effective immune responses. Although intense efforts are focused on developing preventative vaccines and therapeutic measures, there is increasing interest in understanding the basic immunological mechanisms mediating survival. In addition, models for predicting survival based on EBOV viral loads or peripheral immune cell transcription profiles of infected patients are being developed (16, 49, 143). Before the onset of viremia, interactions with the innate immune system are likely playing an important role in determining outcome. Hosts that can overcome the type I IFN antagonism of EBOV will be able to inhibit viral replication, delay the spread of infection, and mount an adaptive immune response. An aggressive inflammatory response initiated by infected macrophages and mDCs must be tempered to avoid recruitment of additional inflammatory immune cells (e.g., neutrophils) and damage caused by excessive inflammation. Future animal studies focusing on tissue-based responses and early time points in infection will provide a better understanding of the innate responses against EBOV.

Although passive transfer of optimized mAb preparations provides a survival benefit in macaque studies, host-generated Ab responses in natural infection may not mature to sufficient affinity and potency to provide protection, thus perhaps explaining the inability of some subjects to resolve infection. The aggressive nature of EBOV infection may not allow adequate time for host-generated Abs to acquire desired qualities such as the capacity to block interaction with the EBOV receptor (103). On the other hand, even a nonoptimal Ab response, alone or combined with innate immunity, may contribute to the containment of viral replication and allow time for T cell responses to develop. Ab-mediated effector functions such as ADCC and ADCP have been studied in vitro, and their role in vivo in the clearance of EBOV infection of primates remains to be demonstrated.

Experimental data support a role of CD8⁺ T cells in immune protection against EBOV in macaques. Samples from the West African epidemic demonstrated for the first time, to our knowledge, the early activation and development of EBOV-specific CD8⁺ T cells, which associate with virus clearance from the circulation. CD8⁺ T cells maintained effector phenotype for more than a month after viremia becomes undetectable, suggesting that CD8⁺ T cells may be activated by persistent viral Ag and could contribute to elimination of residual EBOV in tissues.

Finally, the persistence of the virus in immune privileged locations is a particularly alarming observation confirmed by the 2014–2016 epidemic. Immune analysis of survivors will help to clarify whether the persistence of EBOV raises the risk of long-term transmission, and in turn, if that has the capability to increase a worldwide re-emergence of EBOV. Understanding these immunological mechanisms of virus clearance will inform the development of vaccines and therapeutics in preparation for future EBOV outbreaks.

We thank Ken Abe, Kendra Leigh, and John Misasi for comments and critical reading of the manuscript. We apologize to those in the field whose important work was not included in this Brief Review because of space limitations.

The authors have no financial conflicts of interest.