Quiescence preserves the self-renewal capacity and, therefore, the long-term function of HSCs. The regulators of this dormant state include intrinsic pathways as well as soluble and contact-dependent factors present in the niche microenvironment [reviewed in Ref. (6)]. Dysregulated HSC cycling may contribute to AA by enhancing differentiation over self-renewal or by sensitizing HSCs to apoptosis (7–10). Interferons have been implicated in both driving proliferation (11) and impairing proliferation of primitive hematopoietic stem and progenitor cells (HSPCs) (12), and sensitizing cells to apoptosis (13), thus supporting the notion that IFNs directly impair hematopoiesis by compromising stem cell function.

An altered microenvironment may also contribute to the pathogenesis of AA. Analysis of BM mesenchymal stromal cells (MSCs) derived from AA patients revealed reduced proliferative capacity and adherence, and a propensity to differentiate into adipocytes at the expense of osteoblasts (OBs) (14, 15). Considering the essential survival and dormancy-enforcing cues provided by niche cells, it will be important to investigate more fully the defects in stromal cells in acquired AA, and the impact of IFNs, either directly or indirectly, on such cells.

The observation that patients with acquired AA exhibit increased levels of circulating IFNγ was made over 30 years ago (16). The presence of T cells containing intracellular IFNγ and positive for the prototypical Th1 transcription factor Tbet is an indicator of disease (17), and reduced frequencies of IFNγ positive T cells correlates with responsiveness to immunosuppressive therapy (4), suggesting that Th1 cells contribute to disease pathogenesis. Attempts to understand how IFNγ-mediated disease pathogenesis revealed that overexpression of IFNγ in vitro impairs long-term culture initiating cells LT-CIC (18), consistent with observations that neutralizing IFNγ in cultures derived from AA patients resulted in improved colony formation (16). Moreover, a polymorphism that results in enhanced stability of IFNγ transcripts is strongly associated with the risk of developing AA (5). However, the precise mechanisms whereby IFNγ drives BMF in vivo are still unclear and may involve multiple overlapping pathways and multiple cell types.

Type I IFNs (IFNα/β) are key regulators of innate and adaptive immunity. Although not directly implicated in AA pathogenesis, type I IFNs mediate host responses to most infections and contribute to autoimmunity in systemic lupus erythematosus [recently reviewed in Ref. (19)] and potentially in diabetes mellitus, Sjogren’s syndrome, autoimmune myositis, and rheumatoid arthritis (20, 21). Pegylated IFNα 2a (PEG-IFNα2a) is the standard of care in hepatitis C virus (HCV) patients, but is also a treatment option for melanoma (22), hairy cell leukemia (23), and multiple sclerosis (24–26). Type I IFN therapy is not well tolerated by all patients, however, and hematologic side effects are closely monitored. HCV patients receiving both PEG-IFNα2a and the nucleoside analog ribavirin are prone to hemolytic anemia due to ribavirin processing in erythrocytes as well as PEG-IFNα2a-mediated BM suppression (27, 28). Rarely, type I IFN therapies have also been linked to persistent BM suppression and the development of AA (24, 29, 30). BM suppression appears not to require exogenous or supraphysiologic levels of IFNα/β, as anemia and BM failure have also been associated with physiologic type I IFN responses to chronic viral infection (31). Of particular relevance to AA, the impact of type I IFNs on hematopoiesis is often not immediately suppressive, but requires secondary stress, such as exposure to subsequent IFNγ during the pathogenesis of lymphocytic choriomeningitis virus (LCMV) infection (12). Herein, we will discuss the potential for direct and niche-mediated type I IFN stimulation to impair HSCs and contribute to acquired AA.

Bone marrow suppression has been observed subsequent to a number of viral infections, including parvovirus (32, 33), human immunodeficiency virus [HIV; Ref. (34)], viral hepatitis (35), Epstein–Barr virus (36), and influenza (37), among others. The ability of viral infections to suppress the BM may be due to both the ability of viruses to actively infect cells of the hematopoietic system and the host response to the virus, likely involving production of interferons and other pro-inflammatory factors. BM suppression and severe cytopenias are also common after exposure to tick bites, and are associated with the rickettsial pathogens Ehrlichia chaffeensis and Anaplasma phagocytophilum (38). Though transient, cytopenias are often severe, and infection requires antibiotic treatment (39). Human monocytic ehrlichiosis has been associated with bone marrow hypoplasia (40) and hemophagocytic lymphohistiocytosis [HLH; (41)], and murine models implicate interferon responses in mediating bone marrow suppression in rickettsial infections (42–44).

Bone marrow failure pathogenesis was first modeled in mice using exposure to toxins, instigated by the association of benzene exposure with human disease (45). Observation that AA is a result of immune-mediated pathology prompted the development of donor lymphocyte infusion models relying on the adoptive transfer of lymph node or spleen-derived lymphocytes from histocompatibility mismatched strains of mice (46). This model recapitulates many observations in human AA patients as protection can be achieved with immunosuppressive treatment and abrogation of IFNγ (47, 48). A technical hurdle of the infusion-based model is that the use of F1 recipients precludes analysis of genetically targeted mice. Thus, it has been difficult to evaluate direct and indirect roles of specific cytokines on hematopoietic versus stromal cells. However, it has allowed a deeper understanding of T cell intrinsic mechanisms necessary for initiation of disease, including Notch signaling (49) and transcriptional regulators of Tbet (50). To model human patients carrying a mutation that renders a higher risk for developing AA, a mutation was introduced to the 3′ untranslated region of the Ifng gene, stabilizing IFNγ transcripts (51). Termed “ARE-delete,” this mouse model reproduces many features of human disease and is not associated with autoreactive T cells, suggesting that elevated IFNγ, independent of activated T cells, can drive disease by impairing progenitor cell function (51). In addition, a number of insights into bone marrow suppression have come from studying bacterial and viral pathogens. In ehrlichiosis, HSC loss requires IFNγ sensing by macrophages, demonstrating that interferon signaling reduces the HSC supportive capacity of niche cells during infection-induced BM suppression (42). In LCMV, phenotypic HSCs are reduced early in the course of infection, independent of IFNγ and likely through the actions of type I IFNs (12, 31). Together, the observations made in murine infection models and in a subset of patients undergoing PEG-IFNα2a treatment provide additional evidence that interferons impair HSCs, likely via multiple direct and niche-mediated mechanisms.

The negative impact of IFNγ on hematopoiesis is well documented [reviewed in Ref. (52)], but what is the evidence that there is a direct impact of IFNγ on the most primitive HSCs? HSC loss can occur via impaired self-renewal, increased differentiation, or induction of cell death, which may be results of both direct and/or indirect effects of IFNγ. While some studies suggest that IFNγ has an antiproliferative effect on HSCs (12, 53), evidence also suggests that IFNγ signaling promotes proliferation, and subsequent exhaustion of HSCs (11, 54). During infection with Mycobacterium avium or LCMV, IFNγ increased HSC proliferation and led to a reduction in transplantable myeloid potential (11, 55). Moreover, HSCs from a microenvironment deficient in IFNγ have more robust long-term potential, whereas excessive IFNγ signaling reduces transplantable HSPC activity (54, 56), further suggesting that tonic IFNγ signaling limits HSC function, perhaps through inducing proliferation. The discordant results with respect to whether IFNγ induces or suppresses proliferation is further confounded by the complex interaction with other cytokines, as IFNγ can augment the expansion of myelogenous leukemia cells when it signals in concert with IL-3, but can suppress proliferation in cells lacking IL-3 stimulation (57). In addition, TNFα stimulation is necessary for maximal IFNγ-induced suppression and proliferation of leukemia cell cultures (57), further emphasizing the potential for IFNγ to elicit distinct and even opposing effects dependent on the local cytokine milieu.

Stem cell proliferation can result in the generation of more stem cells (self-renewal) or more committed progenitors (differentiation) and IFNγ has also been implicated in impeding self-renewing divisions (Figure 1a) (12, 58). Notably, IFNγ was shown to directly reduce HSC self-renewal during recovery from viral infection where robust type I IFNs had ablated the HSC pool (12), suggesting that type I IFNs may potentiate the suppressive impact of IFNγ on hematopoiesis during viral infection. These data highlight the importance of the cellular and cytokine context in the impact of single cytokines. Whereas IFNγ may not impede self-renewal in the steady state, prior exposure to type I IFNs may sensitize HSCs to the effects of IFNγ; at the same time, the induction of cellular stress by type I IFN-induced HSC cycling could enhance the potential for IFNγ to provoke HSC apoptosis during immune-mediated BM failure (59, 60).

Hematopoietic stem cells require a variety of inputs from growth factors, chemokines, G protein-coupled receptors, and cytokines to maintain their dormant status, location, and capacity to self-renew. An intriguing role for IFNγ in limiting responsiveness to the growth factor thrombopoietin (TPO) via the increase in suppressor of cytokine signaling (SOCS1) (12) (Figure 1b) illustrates yet another direct mechanism whereby IFNγ can impede HSC function. Support for the role of TPO in HSC function comes from the promising clinical data using a TPO receptor (c-Mpl) agonist, Eltrombopag (61). When given in combination with immunosuppressive drugs, it can provide tri-lineage recovery in patients refractory to traditional therapies (62). Though precise mechanisms of Eltrombopag function have not yet been elucidated, one possibility is that the drug works by overcoming a direct impact of IFNγ on suppressing TPO signaling in HSCs.

Interferon gamma is elicited by many microbial infections and plays a critical role in host defense by sensitizing cells to undergo apoptosis, thus impeding pathogen growth (63, 64). IFNγ can induce apoptosis by increasing the expression of Fas on cells subsequently targeted by Fas ligand-expressing cells, such as T lymphocytes (60). Evidence that HSCs express Fas in response to IFNγ (Figure 1c) suggests Fas-mediated destruction of HSCs contributes to their loss in AA (60). It is important, however, to consider the question of HSC sensitivity to IFNγ. Indeed, whereas some cell types respond very rapidly to IFNγ in vitro, such as macrophages, HSCs exhibit a much more subtle response, as measured by activation of STAT1 (42). This may indicate that the ability of HSCs to respond to IFNγ in vivo may be concentration dependent, and it suggests that HSCs are likely not first responders to IFNγ during an initial exposure. Under prolonged conditions of chronic exposure, however, HSCs may become direct targets. Thus, there are temporal considerations when evaluating the direct impact of IFNγ on HSCs under different inflammatory conditions.

The idea that HSC activation can be achieved directly or as a result of demand implies that HSC loss may result from increased progenitor cell activity or loss. Several lines of evidence support a direct role for IFNγ in impacting murine progenitor cells in the context of infection (43, 65). IFNγ promotes the emergence of a phenotypically unique, hybrid progenitor population that expresses the IL-7 receptor and has both myeloid and lymphoid potential in vitro, but has primarily myeloid potential in vivo. Similarly, an intrinsic requirement for IFNγ was found to occur during bacterial infection, directing the production and terminal differentiation of myeloid cells (43). In a model of sterile inflammation, via adoptive transfer of activated effector T lymphocytes, IFNγ acted directly on progenitors, but not HSCs (66). The ability of IFNγ to act on downstream progenitors to drive proliferation, however, may indirectly call HSCs from a dormant state, which may explain observations suggesting that IFNγ acts directly on HSCs (Figure 1d).

Early observations made in LCMV-infected mice (31), and in IFNα-treated HSPC cultures (127), led to the conclusion that type I IFNs suppress progenitor cell proliferation and differentiation. Indeed, IFNα induces HSPC expression of cell cycle inhibitors in vitro (13). In vivo, however, the impact of IFNα/β differs. Acute administration of the double-stranded RNA mimetic polyinosinic:polycytidylic acid (polyI:C) causes rapid, IFNα receptor (IFNαR)-dependent HSC cycling (13, 128), and has been the model of choice for studying type I IFN-mediated HSPC activation (see Table 1 for a summary of relevant findings). The impact of polyI:C-induced sterile inflammation varies depending upon the duration of stimulation and the precise HSC subset analyzed, but acute stimulation is sufficient to decrease HSC expression of cyclin-dependent kinase inhibitors and quiescence-enforcing transcriptional programs, including FoxO3a, Notch, and TGFβ (13).

The function of type I IFNs in the context of physiologic induction, such as infection, may provide insight into pathogenic role(s) of type I IFNs in AA. Somewhat paradoxically given the BM suppressive impact of IFNα in viral infection, the type I IFN response to opportunistic Pneumocystis lung infection is protective in Rag^−/− mice (133). Since these mice lack all B and T lymphocytes, immunity depends entirely upon myeloid cells, which undergo greater apoptosis in the absence of type I IFNs (134). In Pneumocystis-infected Rag competent mice, IL-10 and IL-27 production by B lymphocytes is protective, and correlates with enhanced myelopoiesis (135). Thus, in the absence of lymphocytes, and the cytokines they produce, type I IFNs provide a survival signal for myeloid cells. These findings further support the notion that the complex cytokine milieu greatly impacts the outcome of IFN signaling on HSC function.

At homeostasis, HSC quiescence protects against replication stress and genomic instability. Long-term label retaining studies demonstrate that ~1% of phenotypic HSCs cycle per day (136) and that a subset of multipotent progenitors is maintained in a similarly dormant state (137). PolyI:C increases HSC cycling six- to sevenfold for up to 3 days, leading to the accumulation of reactive oxygen species and DNA damage in remaining HSCs (59). DNA damage itself induces type I IFN-mediated stem cell senescence (138), in addition to the activation of cellular checkpoints and tumor suppressor genes [recently reviewed in Ref. (139)]. Additionally, type I IFNs transcriptionally regulate p53 (140), through the interferon-stimulated signaling complex ISGF3 (141, 142). Therefore, type I IFNs have the capacity to induce both proliferation as well as DNA damage-induced p53 signaling, thus priming HSCs to undergo apoptosis upon cellular stress, such as in vitro culture (13). These findings, therefore, implicate type I IFNs in the induction of replication and oxidative stress in HSCs.

In vivo, repeated IFNα/β stimulation or uncontrolled type I IFN signaling is detrimental to HSCs exposed to chemotherapeutic injury or transplantation (13, 128, 143), likely by promoting cell cycle entry and heightened sensitivity to cellular stress. While this may be detrimental in some cases, complete molecular remission has been observed in several cases of chronic myelogenous leukemia (CML) where IFNα pre-treatment was followed by imatinib mesylate (144), suggesting that IFNα may induce CML stem cell exit from dormancy and subsequent sensitization to growth factor withdrawal. The sensitizing effect of type I IFNs on stem cells persists, as HSCs transplanted from mice 2 weeks after polyI:C stimulation remained functionally impaired in their repopulating capacity (13). Therefore, it is reasonable to expect that increases in endogenous IFNα/β during viral infections or chronic administration of type I IFNs may have long-term impacts on HSC response to subsequent inflammatory stimuli. Such a mechanism is consistent with HSC impairment and BM suppression during LCMV infection, which elicits initial IFNα/β followed by subsequent IFNγ (12, 31, 145). Further studies are necessary, however, to determine if apoptosis is the predominant mechanism by which HSCs are depleted upon type I IFN sensitization, or if IFNα/β also sensitizes HSCs to senescence, or to non-apoptotic cell death.

In addition to HSC proliferation and apoptotic sensitization, type I IFNs influence HSPC differentiation. IFNα/β increases the synthesis of proteins required for rapid hematopoietic progenitor cell differentiation in response to inflammation and demand. This occurs via a post-transcriptional mechanism, the targets of which include the c-Myc transcription factor (130), and megakaryocyte lineage proteins (104). Expression of the megakaryocyte lineage gene von Willebrand factor and the alphaIIb integrin protein CD41 have previously been attributed to the most primitive HSCs (146, 147), but Haas et al. identified the IFNαR-dependent emergence of highly proliferative, CD41^hi megakaryocyte-restricted progenitor cells within the phenotypic HSC pool upon stimulation of mice with polyI:C, TNFα, or lipopolysaccharide (104). Since CD41 was not interrogated in previous studies, it is unclear to what extent megakaryocyte-primed progenitor cells contributed to the observed effects of type I IFNs on HSPC proliferation, apoptotic sensitization, and multilineage repopulation (13, 59, 128).

Hematopoietic stem cell metabolism is exquisitely regulated to protect against metabolic stress and to regulate the nature of cell division upon entry into the cell cycle (148), such as occurs upon type I IFN stimulation. One mechanism by which HSC metabolism is regulated is through autophagy and the Foxo family of transcription factors (149). FOXO3A, in particular, has been implicated in the activation of autophagy gene expression programs in HSCs that are essential for HSC survival upon cytokine withdrawal or calorie restriction-induced stress (150). Additionally, Warr et al. found that HSCs derived from aged mice had greater autophagic flux and were more reliant on autophagy for colony formation in vitro. Sterile type I IFN stimulation reduces FOXO3A expression and signaling activity in HSPC subsets (13, 104). Moreover, infection-induced type I IFNs are linked to reduced autophagic flux in the liver (151). The impact of type I IFNs on HSC autophagy has not yet been assessed, but autophagic suppression was recently identified in CD34⁺ BM cells from AA patients, and persisted even upon amelioration of AA symptoms (152). Therefore autophagy inhibition could represent an additional mechanism by which interferon signaling impairs HSC stress responses and exacerbates pathology in AA.