It has been described that a number of viruses including, CMV, Hepatitis C virus (HCV), and human herpes viruses can directly infect HSPCs (38–41). For some of these viruses, it has been documented that they can suppress hematopoiesis after direct infection of HSPCs. For example, Simmons et al. demonstrated that CMV infection can mediate myelosuppression in vitro (42). Furthermore, human herpes viruses 7 (HHV7) has the potential to infect HSPCs as well as impair HSPC survival and proliferation, presumably via lysis, or induced cell death (41). Parvovirus B19, the only known human pathogenic parvovirus, has a selective tropism for the erythroid lineage in the BM, where productive infection induces a block in erythropoiesis that can be manifested as a transient or persistent erythroid aplasia (43).

Moreover, the consequence of direct viral infection of HSPCs, such as changes in the expression of intracellular factors e.g., microRNAs, may impact hematopoiesis as well. It was shown that retroviruses such as HIV and HTLV target microRNAs to manipulate key cellular pathways, which can result in the development of hematopoietic malignancies such as B cell and Hodgkin’s lymphomas (44, 45). Furthermore, it was observed that both HIV and SIV can induce large hematopoietic defects. Prost et al. demonstrated that the inhibitory effects on hematopoiesis of SIV depend entirely on the presence of the viral protein Nef. They showed that SIV affected HSPCs by downregulating STAT5a and STAT5b via Nef in vivo and demonstrated that SIV strongly downregulates early hematopoiesis in this manner (46).

Overall, direct viral infection in HSPCs has been shown to reduce the hematopoietic output. However, the exact underlying mechanisms after direct virus infections in HSPCs have not been fully elucidated. It is conceivable that the halt of translation in host cells (or “shut off”) that occurs in many cells following viral infection contributes to this process, though the impact of this antiviral mechanism on HSPC function is not clear.

It has been proposed that direct infection of HSCs does not commonly occur, as they usually reside in protected BM niches (4). Because of this protected microenvironment, quiescent HSCs are thought to be resistant against bacterial infections (47). In fact, all the evidence described in the preceding paragraphs does not distinguish between infection of HSCs or downstream progenitors. It remains to be determined whether there are any differences in virus susceptibility between these cell types. It is likely that quiescent HSCs are less likely to become infected with a virus compared to their non-quiescent counterparts, and evidence supporting this notion is described in the following section.

Pathogen recognition receptors are receptors that detect pathogen-associated molecular patterns (PAMPs) within the body (48, 49). PRRs include toll-like receptors, retinoic acid-inducible gene 1 (RIG-I)-like receptors (RLRs), and cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) synthase (cGAS) (50). Four TLR members seem to play a critical role in recognition of viral nucleic acids: TLR3 recognizes dsRNA (dsRNA constitutes the genome of one class of viruses but is also generated during the life cycle of many viruses), TLR7 and 8 recognize single-stranded RNA (ssRNA), and TLR9 responds to dsDNA viruses recognizing non-methylated viral CpG-containing DNA. Although the majority of TLRs sense pathogen components on the cell surface, TLR3, TLR7, TLR8, and TLR9 sense nucleic acids in endosomal compartments. Other TLRs are also involved in viral recognition; TLR2 and TLR4 were shown to detect viral components such as envelope glycoproteins (51) and other components of viruses such as HIV, HBV, vaccinia virus (VV), and Dengue (52–55).

Interestingly, several intracellular and extracellular TLRs have been found on HSCs (13). It was shown both in vitro and in vivo that direct TLR ligation triggered cell cycle entry in quiescent HSCs, bringing them into a proliferative state (14, 56, 57). This indicates an active role for HSCs in immune sensing, and the modulation of early hematopoiesis during infection. De Luca et al. reported that human HSCs and lineage-committed progenitors express TLR3 (58), whereas Sioud et al. showed that HSPCs particularly express TLR7 and TLR8 (57). These different observations could be explained by different sources of cells that were used, as De Luca et al. isolated HSPCs from cord blood, whereas Sioud et al. isolated these cells from BM of healthy donors. Although side-by-side validation is lacking, differential TLR expression in fetal vs. adult HSCs suggests that these cells may be equipped to detect and respond to different viral infections. Furthermore, HSPCs also express TLR2 (14, 59). Finally, exposure of murine HSPCs to TLR ligands has also been shown to modulate their chemokine receptor expression (6) suggesting that TLR triggering may even regulate their migratory and homing capacities. In line with this, a comparison of TLR expression patterns between BM resident and mobilized HSPC could reveal interesting differences as to their response to viral infection in terms of lineage commitment and adaptation to demand. It might be that mobilized HSPCs in tissues have an increased surveillance capacity than those in the BM, but this remains to be demonstrated.

While signaling through TLR7 and TLR8 resulted in HSPC differentiation along the myeloid lineage (57), in the presence of ligands for TLR9, the in vitro generation of DCs dramatically decreased in favor of the production of macrophages (60). Correspondingly, it was demonstrated that the potential of myeloid progenitors to produce DCs was reduced upon TLR9 ligation (61). However, lymphoid progenitors from mice with a herpes virus infection were biased toward DC differentiation, which was dependent on TLR9, and that treatment of uninfected mice with the TLR9 ligand CpG ODN resulted in an increased DC generation (61). These findings indicate that TLR9 triggering differentially affects the generation of myeloid vs. lymphoid lineage-derived DCs. Furthermore, in the setting of acute leukemia, Dorantes-Acosta et al. found that leukemic HSPCs respond differently to various TLR ligands (62). Their findings suggested that B cell development is only marginally influenced by infectious agents, whereas they observed an increased production of myeloid and NK cell types in response to infections and disease-associated cell damage. For example, they observed an increase in the development of mature CD56⁺CD11c⁺ NK cells after stimulation of TLR8 and TLR9 in chronic acute leukemia-derived BM cells (62).

Besides affecting lineage commitment, viral sensing might also directly affect HSPC survival. Ligation of the RLRs RIG-I/MDA-5, but not TRL3, by poly I:C triggered apoptosis of human CD34⁺ cells, through a type I IFN-independent, caspase-dependent mechanism (63). It has, in fact, been shown that RLRs can trigger a dual response in virus-infected cells, which are independent and both contribute to viral control (64, 65). RLR-signaling activates IRF-3, which then migrates to the nucleus and binds to the IFN-stimulated response elements (ISRE) in the promoters of the target genes to induce type I IFN expression (64). In parallel, IRF3 can also activate the RLR-induced IRF-3-mediated pathway of apoptosis (RIPA) (65). It has long been reported that not every infected cell produces type I IFN (66, 67); there is heterogeneity in the type I IFN production within populations of virus-infected cells, which may be linked to expression levels of RIG-I signaling components or heterogeneous chromatin status of IFNβ genomic locus (67). It might be that the quiescent HSCs respond differently to RLR triggering than downstream progenitors. It is to be expected, given the relevance of maintaining the HSC pool for life, that quiescent HSC are more prone to inducing type I IFN production, which is followed by activation of numerous antiviral mechanisms, instead of inducing the RIPA pathway that kills the cells. There is potentially less biological cost in inducing RIPA in multipotent progenitors to limit viral replication, because these cells can be replaced by long-term HSCs later on. In line with this, a comprehensive proteomics analysis (68) revealed differences between HSPCs (Lineage^-Sca-1⁺c-kit⁺ cells) and myeloid progenitors (Lineage⁻ Sca-1⁻c-kit⁺ cells). Strikingly, 2 ′–5′ oligoadenylate synthetase 3 (Oas3), RIG-I, and Ifit1, three different cytoplasmic sensors for viral RNA, were strongly upregulated in the multipotent state as well as a number of antiviral interferon-stimulated proteins. These results suggest that, compared to downstream progenitors, HSPCs might actually be more prone to activating the IFN antiviral response, although differences between quiescent HSC and multipotent progenitors remain to be shown. In fact, HSPCs are poorly permissive to both retroviral- and lentiviral-based gene transfer. Lentivirus has the advantage that they do not require cycling cells to integrate their genome and could potentially be more suitable for transformation of quiescent HSCs. Nevertheless, genetic manipulation of HSPCs still requires the use of multiple hits of high vector doses and prolonged ex vivo culture, suggesting that permissiveness is not only dependent on cell-cycle but also on the activation of multiple innate immune sensors and restriction factors that limit viral infection (69).

Altogether, it is clear that HSPCs can respond to viral infections through direct recognition of several viral PAMPs, and that the activation of different PRRs can result in different biological outcomes, ranging from changes in chemokine receptor expression and lineage-specification to induction of apoptosis.

In response to PRR ligation by viruses, immune and non-immune cells produce a number of cytokines and chemokines. These include IL-1β, IL-6, IL-8, IL-10, TNFα, GM-CSF, TGFβ, CCL3 (MIP-1α), CCL4 (MIP-1β), CXCL10 (IP-10), and type I IFNs. Different viruses induce slightly different patterns that depend on the interaction of the particular viral PAMP with the specific PRR and the cell type involved (70, 71). In general, pro-inflammatory cytokines like IL-1β, IL-6, and TNFα together with type I IFN production are common to all viral, and many bacterial, infections (70). Many viruses that persist, such as EBV, HIV, and HTLV-I, signal innate cells such as dendritic cells, NK cells, and macrophages to produce anti-inflammatory molecules such as IL-10 and TGFβ (70, 71).

High levels of IFNα directly induces HSCs to exit quiescence and transiently proliferate in vivo (72, 73). Since most cell types stop proliferating in response to IFNα, the wiring of this signaling pathway must be fundamentally different in HSCs (74). According to Pietras et al. (75), type I IFN-driven HSC proliferation is a transient event resulting from a brief relaxation of quiescence-enforcing mechanisms in response to acute type I IFN exposure, which occurs exclusively in vivo. This proliferative burst fails to exhaust the HSC pool, which rapidly returns to quiescence in response to chronic type I IFN exposure, achieved by repeated polyI:C administration, because of the presence of intrinsic regulatory mechanisms. Type I IFN-exposed HSCs with re-established quiescence are not fully functional but are also largely protected from the killing effects of IFNs unless forced back into the cell cycle due to culture, transplantation, or myeloablative treatment, at which point they activate a p53-dependent proapoptotic gene program (75). These ideas are further supported by independent findings showing that, when genes that normally suppress IFN signaling are disrupted, mice have increased levels of IFN signaling, and their HSC populations are depleted over time (73, 76, 77). In chimeric mice transplanted with both wild-type HSCs and IFNα-receptor-deficient (IFNαR^−/−) HSCs and later exposed to IFNα, only the IFNαR^−/− HSCs survived (72). In fact, exit from quiescence, induced by polyI:C or other stimuli, leads to generation of DNA damage that activates a DNA damage response in HSCs. Repeated activation of HSCs out of their dormant state provoked their attrition, and this was exacerbated in mice with a defect in DNA repair, suggesting that inefficient repair of replicative DNA damage may result in HSC depletion (78). Additionally, one single dose of IFNα or polyI:C induced an increase in the production of megakaryocyte-related proteins by acting on a stem-cell-like megakaryocyte progenitor that is contained within the phenotypically characterized HSC pool. Again, chronic exposure to polyI:C triggered exhaustion of these stem-cell-like megakaryocyte progenitors and a delayed repletion of platelet counts (79). Recent work also proposes that chronic exposure to IFNα drives medullar lymphopoiesis toward T-cell differentiation, while impairing the generation of B, NK, myeloid cells, erythrocytes, and platelets (80). Collectively, these findings suggest that BM aplasia associated with chronic exposure to type I IFN could arise from a depletion or loss of function of progenitors, together with enforced quiescence of HSCs, which become less functional. Moreover, if subsequent inflammation or infections force these quiescent HSC chronically exposed to type I IFNs back into cell cycle or if control mechanisms such as those regulating IFN signaling or DNA damage repair are deficient in certain susceptible individuals, this might lead to rapid depletion of the HSC quiescent pool and ensuing BM failure.

After the first wave of type I IFNs produced by infected cells, type II IFN (IFNγ), produced by stimulated T cells and NK cells, also contributes to the impairment of HSC self-renewal (81, 82). IFNγ triggering on HSPCs enhances monocyte formation, but this is at the expense of the production of neutrophilic (83) and eosinophilic granulocytes (84), B cells (85, 86), and erythrocytes (87, 88). These data illustrate that IFNγ has a wide-ranging effect on the hematopoietic process in the BM (19). It seems clear that type I and type II IFNs have unique features, with type I IFNs leading to megakaryopoiesis on the short term and quiescence and T-cell-biased lymphopoiesis upon chronic exposure, while IFNγ favors myelopoiesis at the expense of other branches of differentiation. This might link the temporal expression of the two types of IFN to the changing needs of the immune response as it progresses from initial innate control to adaptive mechanisms.

As mentioned before, PRR stimulation by viruses can also lead to the production of pro-inflammatory cytokines and chemokines, which can in turn affect the proliferation and differentiation of HSCs. Several cytokines such as IL-1β, IL-6, TNFα, TGFβ, M-CSF, and GM-CSF have been identified to have the capacity of regulating the proliferation and differentiation of HSCs as reviewed by Mirantes et al. (33). Overall, these results indicate that cytokines can instruct the differentiation and proliferation of HSCs, with BM output being a complex integration of signals from the microenvironment. As with prolonged type I IFN exposure, overproduction of inflammatory cytokines is often associated with hematopoietic failure such as chronic inflammatory diseases and hematopoietic malignancies (33), but the individual contribution of each player, and their potential synergies and antagonisms, remains to be determined. Finally, apart from inflammatory cytokines, costimulatory molecules may also play a significant role in altering hematopoiesis during viral infection (5). One example is the interaction between CD27 and CD70, which is important in the control of T cell immunity against Influenza (89) and CMV (90). This is of interest for the BM, as we previously showed that CD27-triggering on HSPCs may serve as a negative feedback mechanism that can regulate hematopoiesis during inflammatory conditions (91). This provides another layer of complexity by which immune activation can modulate the BM output upon viral infection.

As described for sterile inflammation and bacterial infections, viral infections may also affect hematopoieisis indirectly via the BM microenvironment. Apart from hematopoietic cells, the BM also contains non-hematopoietic components, such as osteoblasts, mesenchymal stromal cells (MSCs), adipocytes, perivascular cells, endothelial cells, and non-myelinating Schwann cells. It has long been recognized that non-hematopoietic stromal cells in the BM are capable of supporting long-term hematopoiesis in vivo, and that the integrity of several populations of cells is crucial for the long-term maintenance of the quiescent HSC pool (92, 93).

Infection of mice with different strains of LCMV leads to transient type-I IFN-dependent BM aplasia that reaches a minimal BM cellularity within 3 days and recovers within 10 days after infection, coinciding with viral clearance. Although the mechanisms have not been clearly dissected, LCMV can infect both stromal cells and megakaryocytic and myelocytic precursors (36). It is likely that the aplasia is a result of apoptosis of the hematopoietic progenitors, which might be modulated by type I IFN signaling, as described above. However, the effects of viral infection on the stromal cells have not been dissected. It would be very interesting to address whether these cells die from apoptosis or respond by secreting cytokines that can affect hematopoiesis. If different populations of stromal cells are depleted and the hematopoietic niches are destroyed, the mechanisms by which these niches are reconstituted after viral clearance are still unknown. In this line, an important role for stromal cells in the hematopoietic stress-induced response during LCMV infection has been demonstrated by Schürch et al. (94). They described that IFNγ produced by CD8 T cells during the adaptive response to LCMV can also affect HSCs in an indirect manner by inducing IL-6 production in non-hematopoietic stromal cells (94), which in turn enhances myelopoiesis by downregulating the expression of Runx-1 and Cebpα in hematopoietic progenitor cells. Importantly, this increase in myelopoiesis was not only seen upon LCMV infection but also upon infection with vesicular stomatitis virus (VSV) or VV, suggesting that it is a general response to viral infection (94). These observations indicate that IFNγ not only acts directly on HSPCs, as described above, but also in an indirect manner via non-hematopoietic BM cells.

Mesenchymal stromal cells have also been reported to be affected by HIV, by a mechanism involving Tat and Nef (95). MSCs chronically treated with Tat and/or Nef reduced their proliferative activity, underwent early senescence, and showed decreased potential for osteoblastic differentiation, which could explain why HIV infected individuals present a higher prevalence of osteopenia/osteoporosis. Although no direct link was made, loss of the BM MSC functionality may have consequences on hematopoiesis.

Additionally, BM MSCs have been proposed to be a natural reservoir for human CMV (96). A recent retrospective study has shown that the CMV status of both donor and host are relevant for the overall survival of patients receiving allogeneic stem cell transplantation (HSCT) (97). As expected, CMV-negative recipients benefit from receiving HSCT from a CMV-negative donor, since the virus transmitted from a seropositive donor into an immunosuppressed seronegative recipient can have devastating effects. More intriguingly, CMV-positive recipients have an increased overall survival after receiving HSCT from a CMV-positive donor after myeloablative conditioning but not after reduced-intensity conditioning. Interestingly, there were differences in the causes of death between patients with CMV-seropositive or CMV-seronegative donors. Patients who received grafts from CMV-seronegative donors were more likely to die from viral infection given as the only cause of death or where a viral pathogen was included as a part of a mixed infection with different organisms. This effect was abrogated by T-cell depletion (97), which suggests that memory T cells from the donor can help control viral infections in situations of more aggressive conditioning. Since recipient MSCs are not depleted during either of the conditioning regimes (98), and they can harbor CMV, it would be interesting to study the interaction between donor CMV-specific memory T cells and host CMV-infected stroma, and how this interaction is modulated by the conditioning treatment.

Another virus that has a significant impact on the BM microenvironment in mice is the type B coxsackievirus (CVB). Althof et al. observed that after 3–4 days of infection, the femoral BM stroma was largely destroyed. While granulocyte and macrophage progenitors showed a relatively normal proliferative capacity, there was a marked decrease in colony-forming capacity in both erythroid and lymphoid progenitors, indicating a differential impact on the various HSPC subsets (99). Production of type I IFNs was shown to contribute to the development of lymphopenia upon CVB infection (99), which could also further contribute to the hematopoietic defects of the other hematopoietic lineages.

Although the contribution of the BM microenvironment to stress-hematopoiesis can be mediated through direct recognition of the pathogen and/or the ensuing production of inflammatory mediators, it may also simply be driven by the absence of particular cell types. Scumpia et al. observed that HSPC expansion upon bacterial infection can occur even without sensing of the bacterial PAMPs (using MyD88^−/− TRIF^−/− mice). They suggest that reduction in BM cellularity alone is sufficient to induce HSPC expansion, possibly mediated by supporting stromal cells that can provide the necessary signals for HSPC expansion within the BM space left void following infection or chemotherapeutic BM ablation (100). Similarly, it has been suggested that granulopoiesis can be driven by the number of neutrophils present in the BM. At low neutrophil numbers, macrophages and DCs produce IL-23, which stimulates IL-17 production by T cells, which in turn increases granulopoiesis via G-CSF. On the other hand, at high neutrophil numbers, the production of IL-23 in DCs and macrophages is inhibited by the phagocytosis of apoptotic neutrophils, resulting in a negative feedback loop (101). Large numbers of neutrophils are predominantly important in fungal and bacterial infections, and whether similar interactions are at play in viral infections is currently unknown. However, it is an interesting concept that the sheer presence or absence of particular cell types in the BM can also contribute to the virally induced hematopoietic stress response.

Overall, the link between the BM microenvironment and hematopoiesis is not easy to define. Multiple pathways can play a role in the regulation of hematopoiesis via the BM microenvironment, and these pathways may further complement each other. Yet, understanding the underlying cellular and molecular mechanisms by which the BM responds to viral infections may help us to counteract the ensuing pathogenic consequences, which were discussed at the beginning of this review.