The mononuclear phagocyte system has been initially formulated by the late 1960. It consists of a network of cells, comprising monocytes, macrophages, and dendritic cells (DCs) that are dis-seminated throughout the organism. These cells are characterized by their morphology, their phenotypic characteristics (including phagocytic activity), and their roles in orchestrating the immune system. The majority of their committed progenitors are quiescent at homeostasis, although their very high proliferative potential provides them with the capacity to continuously maintain their numbers. Significant progresses in system biology have been made only recently in regard to understanding of the ontogeny and the function of mononucleated cells (referred to as myelopoiesis). This led to the discovery of committed precursors for adult-derived monocytes, conventional, plasmacytoid, or monocyte-derived dendritic cells (Mo-DCs), which are primarily described in the present perspective article. For more details on the embryonically derived phagocytes, we direct the reader to the following outstanding review (1).

Macrophage and DC precursor cells (referred to as MDP) does not constitute a homogeneous population but rather consists in a mixture of progenitors committed either to the DC lineage or the monocyte/macrophage lineage when they are transferred into the bone marrow (BM) of hosts that have previously been irradiated (2, 3). While less is known about the ontogeny of monocytes, macrophages, and DCs in humans than in mice, recent studies have allowed a link to be made with what has been observed in animal models. Notably, a homolog of murine MDP has been identified based on the in vitro differentiation of human CD34⁺ hematopoietic progenitors into type 1 conventional DC (cDC1) (4). There has since been a concerted effort to identify precursors restricted to either cDCs or those derived from the monocytic lineage. MDP express M-CSF-R (or CD115) and the Flt3 receptor (CD135), which are receptors for cytokines that play important roles in the development of monocytes or DCs, respectively. It is likely that the commitment shift of MDP depends on the balance between signals linked to the activation of these receptors (5). This hypothesis is bolstered by the fact that the expression of M-CSF-R decreases in the precursors of cDCs and plasmacytoid DCs (pDCs), although it is not detectable in mature cells. Conversely, Flt3 is not found in the precursors restricted to the monocytic lineage (6, 7). Signaling by the aforementioned growth factors could induce changes at the level of the expression of certain transcription factors. For example, the hematopoietic transcription factors PU.1 and MAFB (for MAF BZIP Transcription Factor B) are crucial for the development of DCs or monocytes, respectively, and they could be implicated in engagement in one of these lineages (8).

Apart from the MDP, the precursor CDP stands for common DC progenitor (Figure 1). Like the MDP, it expresses M-CSF-R and Flt3 (9–11). The CDP on the one hand generates pDCs, and on the other hand generates pre-cDCs, which are the direct circulating precursors of the cDCs in tissues. In parallel, other teams have elegantly shown that, as is the case with mice, the generation of cDC1 and cDC2 by common DC progenitor (hCDP) occurs by production of a circulating progenitor, namely the hPre-cDC, which is incapable of generating pDCs (12). Like their murine homologs, hPre-cDCs are heterogeneous and they comprise various fractions already committed to become cDC1 or cDC2 (13–15). Pre-cDCs leave the BM via blood circulation and then penetrate into lymphoid and non-lymphoid tissues in order to differentiate into cDCs (9–11). The factors that influence the differentiation of pre-cDCs into cDC1 or DC2 are still unknown. However, it appears that this decision is taken at the CDP stage, which can already exhibit a transcriptional signature similar to cDC1 or cDC2. Moreover, the pre-cDC population appears to be heterogeneous, comprising a mixture of pre-cDC1 and pre-cDC2 in mice (16) and in humans (15).

More recently, a progenitor restricted to monocytes and derived directly from MDP was identified and designated as cMoP, for common monocyte progenitor (Figure 1). It differs phenotypically from MDP by the loss of Flt3 expression. Consequently, cMoPs differentiate into monocytes and their descendants, but they do not generate cDCs (7). The development of cMoPs into monocytes also takes place as monoblast and then as pro-monocyte stages. They are characterized by the expression of stem cell antigen 1 (Sca-1) and they undergo very fast turn-over in the BM (20, 21). The monocytes generated in this manner then migrate from the BM to the tissues where they differentiate depending on the microenvironment (22). Furthermore, a recent study has shown that the generation of human monocytes by hMDP occurs by production of restricted precursors referred to as cMoPs (23), as in mice (7).

These novel concepts are not yet set in stone, however, as the single cell genomic era is already leading to refinements in ontogeny of each subsets of DCs and macrophages. For instance, Helft and colleagues recently demonstrated that human cDC1 are derived more efficiently from the multipotent lymphoid progenitor than from the common myeloid progenitor (CMP) (24). In parallel, it is proposed that the MDP does not constitute a homogeneous population but rather consists in a mixture of progenitors committed either to the DC lineage or the monocyte/macrophage lineage, with no or only very few individual cells able to yield both cell lineages in their progeny (25, 26).

Monocytopoiesis is a dynamic process that occurs in the BM as well as in other organs as an adaption to several physiological stresses that varies over time, while emergency myelopoiesis refers to the rapid generation of myeloid effector cells in response to purified lipopolysaccharide (LPS) (27). A hallmark of septicemia is a profound decrease of circulating DCs, which is also an indicator of a poor prognosis for septic patients (28–30). Two main types of murine models of sepsis or acute inflammation are generally used. On the one hand, the model of peritoneal or intravenous injection of purified endotoxins constitutes a simple model of acute inflammation. On the other hand, the other widely used murine sepsis model is based on cecal ligature and puncture (CLP). This chirurgical model induces intestinal bacterial translocation into the peritoneal cavity, generating a systemic infection and massive inflammation. Meanwhile, the extent to which murine models adequately reflect the complexity of human sepsis or acute inflammation is a matter of debate (31, 32). Although differences in TLR distribution among the various mononuclear phagocyte subsets exist between humans and mice (33), the latter are widely used to understand part of these complex disorders.

In order for this emergency myelopoiesis to be induced, TLR4 needs to be expressed by the radiation-resistant cells of the host. These cells then produce the growth factor G-CSF, which is sufficient to induce this phenomenon (34). G-CSF can also be produced following activation of inflammasomes, which depends on the cytokines IL-1beta and IL-1alpha, thereby inducing emergency myelopoiesis (35, 36). A recent study has also provided evidence for the production of IL-3 by B lymphocytes in a murine model of septicemia. This IL-3 allows for a significant increase in the production of monocytes and neutrophils, which are involved in systemic inflammatory respiratory syndrome (SIRS) and the “cytokine storm.” Furthermore, an elevated level of IL-3 in serum is predictive of a poor prognosis in septic patients (37). The hematopoietic progenitors can hence be indirectly activated in case of severe infection, so as to reorient the production of cells toward the myeloid lineage. However, the mechanisms causing this decrease in DCs during sepsis remain unclear, possibly encompassing both enhanced cell death of at least some subsets of DCs (and their defective reconstitution from their progenitor cells, e.g., originated from cMop and/or pre-cDC). Using the recently accepted nomenclature (38), we herein discuss the potential mechanisms causing this decrease in some DCs during sepsis that is linked to long-term sepsis-related mortality, especially in elderly and diabetic populations. In addition to studies of DCs in sepsis and endotoxemia models, we also review the contribution of macrophages, as these phagocytes have sometimes been incorrectly classified as Mo-DCs with the use of non-discriminating markers (38), such as CD64.

The different types of mononuclear phagocytes might be affected during sepsis by a reduction in their number, by cell death or precursor fate mechanisms, or by their resolutive functions. Numerous studies have shown that immune cell death contributes to immunosuppression and damage to organs during the development of septicemia (145). Apoptosis appears to at least partially explain the loss of DCs observed in a murine model of septicemia (146). For instance, sera from sepsis patients has been shown to induce death of circulating CD11c⁺ CD123⁻ DCs, CD14⁺ monocytes and of in vitro generated monocyte-derived DCs (147). However, the markers used do not allow the subset specificity to be determined. Moreover, the potentially lowered survival of pDC needs to be evaluated with more specific markers that preclude pre-DC contamination (15, 79). This programmed cell death is in part due to the engagement of some TLRs (89, 96, 148). For instance, cDC1 apoptosis in the spleen within 48 h following live E. coli injection is TLR4- and TRIF-dependent (96). Furthermore, phagocytosis of apoptotic cells by DCs renders them tolerogenic. Immunosuppression induced by an endotoxic shock is restrained by the expression of an anti-apoptotic protein by DCs, or by an increase in their number and their activation state by treatment with Flt3L (106, 107, 149). In conclusion, the increase in apoptotic cells with septicemia could contribute to immunosuppression by, on the one hand, the loss of effector cells, and, on the other hand, the induction of tolerance (150).

In addition to death-mediated modulation of the mononuclear phagocyte system, these cells can also be affected by their functions. For instance, during sepsis, monocytes are reprogrammed to enhance protective functions such as anti-microbial functions, which are dependent on hypoxia inducible factor-1α (151). In contrast to this beneficial modulation, the mononuclear phagocyte system can also be modulated during the course of sepsis to promote immunosuppression. For instance, the mononuclear phagocyte system might lose the ability to drive a suitable adaptive immune response. DCs of septic patients, for example, exhibit a decrease in the expression of HLA-DR, thereby reducing their capacity to interact with T lymphocytes (152). Similarly, DCs of septic mice also exhibit a decrease in the expression of MHC-II (91). Moreover, numerous studies in humans as well as in mice have provided evidence for a pronounced decrease in the production of pro-inflammatory cytokines such as IL-12 or TNF by septic DCs stimulated by several PAMPs, while DC dysfunction during sepsis is partly mimicked by the TLR2 agonist Pam3CSK4, rather than the TLR4 agonist LPS (95). Conversely, their capacity to produce the anti-inflammatory cytokines IL-10 or TGF-β is significantly increased (86, 94, 152, 153). Like DCs, monocytes isolated from the blood of septic patients exhibit decreased expression of HLA-DR molecules and lower production of the pro-inflammatory cytokine IL-12 following stimulation after an increase in the production of the anti-inflammatory cytokine IL-10. A high concentration of IL-10 is particularly associated with a poor prognosis for septic patients (93). In the same way, human blood cDC2 produced immunoregulatory molecules, such as IDO, upregulated PD-L1 (a ligand of the inhibitory co-receptor PD1 on T cells), produced high levels of IL-10, and were immunosuppressive in response to E. coli (33). Similarly, PD-1 or PD-L1 was expressed at higher levels in septic shock patients (154), and their functional blockade by antibodies restored monocyte functions (155). In terms of helper T cell (Th) polarization, in mice it appears that the interaction between septic DCs and CD4⁺ T lymphocytes induces preferential polarization of the latter toward a Th2 type or T regulatory profile (86, 91, 153, 156). In addition, GM-CSF-derived BM DCs from CLP- and Pam3CSK4-treated mice were less effective in vivo at Th1 priming compared to GM-CSF-derived DCs from LPS-treated mice (95). It is possible that the loss of the capacity to induce Th1 responses is due, at least in part, to a specific loss of cDC1 which appear to be crucial for the development of such responses (92, 94, 107, 112, 113). Although progress has been made in this regard, the exact molecular mechanism of such functional difference remains unclear. Moreover, it has been reported that the failure of DCs generated by post-septic mice to produce IL-12 with the CLP model was observed at least 6 weeks after this process. This failure to produce IL-12 appears to be due to epigenetic changes induced at the level of promoters for genes coding for this cytokine (86). In summary, these long lasting events might occur in myeloid progenitors as DCs are short lived. Future molecular investigations should consider their epigenetic regulation.

In contrast to modulation of the mononuclear phagocyte system at a functional level, sepsis may affect the developmental fate of myeloid progenitor cells. Monocytes can also acquire phenotypic and functional characteristics of DCs, although the factors influencing this differentiation are still unknown. In keeping with the high level of plasticity of monocytes, the differentiation of these cells depends on local mediators such as cytokines, PAMPs, or DAMPs (76, 157). Some of these cytokines are induced by these danger signals, such as type I IFN. Indeed, type I IFN gives rise to Mo-APCs by acting through the IFNAR receptor on direct monocyte progenitors (Figure 1) (18). Similar to DAMPs, microbiota-dependent metabolites affect the balance between Mo-DCs and macrophages. For instance, aryl hydrocarbon receptor (AHR) ligands, derived either from dietary food intake or from tryptophan catabolism at the mucosal barrier, shift the monocyte cell fate toward monocyte-derived DCs in a PRDM1- (also known as BLIMP1) and IRF4-dependent manner (99, 158). It appears that activation of IRF4 allows monocytes to differentiate into Mo-DC while it remains controversial that only monocytes re-expressing Flt3 can generate Mo-APCs (71, 99, 159). It might be of interest to understand the molecular mechanism of how reprogramming of each monocytes impacts their subsequent ability to differentiate into either DCs or macrophages within different microenvironment. In summary, the generation of monocytes at the expense of cDCs could limit the availability of innate immune effector cells that can counter the infection, and this process might be involved in the immunosuppression in septic animals and patients.

Recent studies have shown that hematopoietic progenitors themselves express PRR, such as TLR (160, 161). They can hence theoretically directly detect PAMP and react as a consequence. In vitro culture experiments of murine and human HSC stimulated by agonists of TLR have shown their preferential differentiation into phagocytes at the expense of cells of the lymphoid lineage (161–164). Furthermore, experiments with parabiotics have ele-gantly demonstrated that a low number of HSC continuously enter the blood circulation before returning to the BM (165). This phenomenon could allow HSC to locally generate effector cells, directly after encountering a circulating microorganism and in a way that is tailored to the molecular signature of the invading pathogens (166).

Type I IFNs can have an effect on hematopoiesis in vivo, particularly by induction of the proliferation of quiescent HSC following injection of the TLR3 agonist polyinosinic: poly cytidylic acid (poly I:C) into mice (167). However, excessive signaling by type 1 IFNs, induced for example by a deficiency in the negative regulator IRF-2, leads to attenuation of HSC proliferation over time, as evidenced by the low capacity of these hematopoietic cells to repopulate following transplantation (168, 169). Additionally, chronic administration of poly I:C induces a selective depletion of WT hematopoietic stem cells (HSCs) in chimeric mice with WT: Ifnar1^−/− BM cells. Excessive proliferation could, as a matter of fact, induce a state of attenuation of the function of stem cells by differentiation, senescence, or also apoptosis, thereby decreasing the risk of malignant transformation and of perturbation of the tolerogenic tissue architecture (167, 170). This attenuation of stem cells could, over time, lead to leukopenias and hematopoietic anomalies. There is still scant documentation regarding the influence of the infectious context on the potential for differentiation of myeloid precursors. In addition to their roles in regard to HSC, type 1 IFNs are involved in the differentiation of common myeloid precursors into macrophages, following the direct activation of TLR7 of these cells by R848 (19). Also, a study has shown that the precursor cells of common DC progenitors express several TLR in mice, including TLR4. In vitro activation of these TLR induces a reduction in the expression of the chemokine receptor CXCR4, which is involved in the retention of CDPs in the BM. In vitro activation of the TLR of CDP also induces an increase in the expression of the chemokine receptor CCR7, which is involved in migration of DCs toward the lymphoid organs. When these active CDP were transferred to mice, they were preferentially found at the level of lymph nodes rich in agonists of TLR, subsequent to local TLR agonist injection, where they underwent differentiation. As the various populations of DCs were not studied in detail in this study, it is hence not possible to draw conclusions regarding the potential selective differentiation of stimulated CDP (160). However, a study has shown that the in vitro differentiation by the cytokine GM-CSF of hematopoietic cells derived from CLP septic mice induced the generation of DCs with an immunosuppressive phenotype, aggravating the susceptibility to secondary infections with Pseudomonas aeruginosa when they were injected into post-CLP septic mice (91, 171). Conversely, the injection of DCs derived in vitro from the BM of healthy mice considerably increased the resistance of septic mice to secondary infections (171). Similarly, when injected intratracheally at day-5 after CLP surgery, cultures of DCs isolated from mouse BM with GM-CSF and IL-4 protected recipient mice from Aspergilus fumigatus-induced death (153). The exact contribution of each DC subset in cultures of mouse BM with GM-CSF remains to be investigated, as in vitro generated CD11c⁺MHC II⁺ cells are a heterogeneous population of cells, with some resembling macrophages more than DCs (172).

Similarly, IL-4 may favor monocyte development toward monocyte-derived DCs to the detriment of monocyte-derived macrophages (99, 172). Aside from a role for IL-4 in the ontogenic shift between DCs and macrophages, IL-4 plays an important role at the functional level as it is required for optimal cross-priming by GM-CSF-induced Mo-DCs (71). A side-by-side comparison of these in vitro generated DC subsets in the protection of septic mice needs to be undertaken. Moreover, the contribution of in vitro generated cDC1 and cDC2 needs to be investigated by using cultures of mouse BM with Flt3L (173). Finally, supplementation of mouse BM cultures with GM-CSF and IL-4 should be studied so as to determine the contribution of each DCsubset in the resolution of sepsis. As done recently in a cancer model, and because in vitro generated DCs might lack environmental cues, the benefit of directly ex vivo extracted DC in sepsis models might be of interest (174). Despite this comparison between in vitro and ex vivo generated DC subsets, their ability to reach the organs of interest, such as lung-draining lymph nodes in the case of intratracheally injected cells, remains to be verified in each sepsis model (153). Meanwhile, limitations of diphtheria toxin-mediated models of cell type depletion have been described such as for DC targeting in CD11c-hDTR mice where many other cell types are affected (175). This implies performing complementation studies with each DC subset obtained in vitro or ex vivo, for proper interpretation of the phenotype of diphtheria toxin-treated mice. For instance, adoptive transfer of GM-CSF-derived DCs into DC depleted mice prevents CLP-induced mortality (176).

It hence appears that DCs generated during septicemia have different effects compared to those produced under homeostatic conditions, and that they are involved in the immunosuppression observed in septic patients. In various models of bacterial infection, the chemokine CCR2 receptor-dependent mobilization of monocytes is crucial for the control of the pathology. Pasquevich and his collaborators have recently shown that the infection of mice with Gram-negative Y. enterocolitica bacilli induces a selective differentiation of myeloid progenitors toward the monocytic lineage (monocytopoiesis) at the expense of cDCs. This process depends on the activation of TLR4 and on the production and the detection of IFN-γ by non-hematopoietic cells (17).