Early-stage apoptotic cells have been injected in arthritis experimental models before the disease is fully established or at time of immunization with the autoantigen (13–15, 48). This is not relevant to the clinical situation, and contrasts with our recent study in which early-stage apoptotic cells are infused when CIA reaches a clinical score of 8 out of 16 (20). Therefore, one may distinguish the prophylactic versus the therapeutic effect of apoptotic cell infusion (Table 1). To date, one limitation is that only one experimental model has been tested for the therapeutic effect (20). For the prophylactic effect, several experimental models have been used (13–15). These models recapitulate differently RA pathophysiology. An absence of prevention has been reported in the serum transfer-induced arthritis (STIA) (13) in which arthritis is induced by the intraperitoneal (i.p.) injection of K/BxN serum in C57BL/6 mice (49) (Table 1). This STIA model recapitulates the effector phase of human RA, but is independent of the adaptive immune response (49, 50). Thus, this suggests that the prophylactic effect of apoptotic cell infusion modulates rather the adaptive immune response, such as CD4⁺ T cell polarization. The model consisting in injecting streptococcal cell wall (SCW) in Lewis rats is induced by a single i.p. injection of SCW fragments (51). This results in a first T cell-independent phase followed by a chronic inflammatory phase that is T cell-dependent and associated with the production of high levels of inflammatory cytokines (51). This results in erosive cartilage damage in the joints (51). In the prophylactic approach using early-stage apoptotic cell infusion, both phases were significantly reduced but the effect was more impressive or pronounced on the chronic phase (14). This is consistent with an impact of apoptotic cell infusion on inflammatory cytokine secretion by macrophages affecting the first phase and on Treg increase modulating the second chronic phase (14) (Table 1). Methylated bovine serum albumin (mBSA)-induced arthritis in C57BL/6 mice belongs to antigen-induced arthritis. In this model, arthritis results from IC-mediated inflammation followed by articular T cell-mediated responses. However, this model does not recapitulate the endogenous breach of tolerance that is typical of RA pathogenesis. This represents a limitation in applicability to RA (50). The prophylactic effect of apoptotic cell infusion has been observed in this model (15, 48) (Table 1). This effect is dependent on natural IgM and IL-10 secretion (15). Finally, the CIA model in DBA/1 mice has been used to evaluate the prophylactic and therapeutic effect of apoptotic cell infusion (13, 20) (Table 1). This mouse model shares with human RA several clinical (i.e., erythema and edema), histopathological (i.e., synovitis, pannus formation, cartilage, and bone erosion), as well as immunological features (51). These features consist in the breach of tolerance with the implication of pathogenic T cells associated with the production of inflammatory cytokines (e.g., TNF), as well as the production of auto-antibodies against self-antigens and collagen (50). Some drawbacks have been evoked for this CIA model. The main drawback is that CIA constitutes only an acute model in contrast to the SCW model (52). Nevertheless, all these models are relevant to some features of RA (49–51), and most of them have been used to test drugs now in clinical development for RA (51). Now, we will highlight some immune mechanisms (Figures 1A–C) and propose future investigations.

One of our hypotheses concerning the use of early-stage apoptotic cell infusion to treat ongoing arthritis was that reintroducing apoptotic cells in a context of non-resolving inflammation—a key feature in RA (38)—may force macrophage reprogramming after apoptotic cell uptake and stimulate the resolution of inflammation with a decrease of pro-inflammatory cytokines (Figure 1). Macrophage reprogramming following efferocytosis stimulates anti-inflammatory factors (e.g., TGF-β or IL-10) and reduces the pro-inflammatory ones (e.g., TNF or IL-1β) (9). The importance of anti-inflammatory cytokines, such as TGF-β (14, 20, 48) and IL-10 (13, 15), was shown in the prevention or treatment of experimental arthritis (Table 1). In addition, reduction of TNF after apoptotic cell infusion in the SCW model has been also reported (14) (Table 1). In the therapeutic approach using intravenous (i.v.) apoptotic cell infusion, macrophages sorted from the spleen [i.e., the main site where blood-borne apoptotic cells are eliminated (53, 54)] induce the polarization of naive CD4⁺ T cells toward a Treg phenotype. Altogether, this sustains our initial hypothesis.

Now, we want to discuss the implications of macrophages in the beneficial effects of apoptotic cell infusion in the light of recent data from the literature. The critical macrophage subset for this beneficial effect can be: (i) splenic macrophages, as identified after i.v. apoptotic cell infusion (20), (ii) peritoneal macrophages, as shown after i.p. apoptotic cell infusion (14), or perhaps (iii) macrophages present in the joint (Figure 1). This may concern tissue-resident macrophages or monocyte-derived macrophages (26) (Figure 1).

Here, the interactions of apoptotic cells with cDC and pDC in the settings of arthritis will be discussed. The implication of pDC after i.v. apoptotic cell infusion has been initially shown in the bone marrow transplantation (BMT) model (79), and found again in the CIA model with the capacity of ex vivo sorted splenic pDC (20) to generate pTreg (Table 1; Figure 1A). Data on the interactions of early-stage apoptotic cells and pDC are certainly easier to interpret than data on cDC. cDC represent different heterogeneous cDC subsets (80), and tools used so far to analyze the impact of apoptotic cell infusion on cDC functions do not allow researchers to separate each subset. For instance, in CD11c/diphtheria toxin (DT) receptor (DTR) transgenic mice, all CD11c^high cells are depleted after DT infusion (81). These CD11c^high cells consist of cDC (81), but also of other APC subsets having the ability to eliminate apoptotic cells, such as marginal zone and metallophilic macrophages in the spleen (82), sinusoidal macrophages in the lymph node (82), or alveolar macrophages (83). In contrast, pDC have been shown to be spared by depletion after DT administration in CD11c/DTR mice (79). It is known that depending on the considered cDC subsets, the response against apoptotic cells can be the opposite, with splenic lymphoid-resident cDC implicated in tolerance induction (73) while a particular subset of cDC localized at barrier surfaces (e.g., the intestine, the lungs, and the skin) boosts inflammatory responses via a PtdSer receptor CD300a (84). Thus, the role of cDC in the therapeutic effect of apoptotic cell infusion in the setting of arthritis has to be further explored. Nevertheless, in the therapeutic approach using the CIA model, we observed that the addition of anti-TNF therapy to apoptotic cell infusion is able to generate ex vivo sorted splenic CD11c⁺ cDC stimulating the polarization of Treg (20). The activation of draining lymph node cDC by NET exacerbated Th1-, but not Th17-, mediated autoimmune responses in CIA (29). This was confirmed in vitro by the maturation of human monocyte-derived DC or mouse bone marrow-derived DC in response to NET isolated from CIA mice or RA patients, respectively (29). Interestingly, it was reported that apoptotic cell clearance by neutrophils reduced NET formation (85). Apoptotic cell infusion may, therefore, limit cDC activation and subsequent Th1 responses by limiting NET formation.

One of the salient consequences following apoptotic cell clearance is the induction of pTreg. This has been demonstrated after i.v. apoptotic cell infusion (86) or local apoptotic death of epithelial cells (87). The increase of Treg in the spleen following i.v. apoptotic cell infusion has been shown to require TGF-β (79, 86), splenic macrophages, and donor pDC in the setting of BMT (79). TGF-β is also required for Treg polarization after intestinal epithelial cell apoptosis (87). In arthritis models, the induction of pTreg after apoptotic cell infusion is also TGF-β-dependent (14, 20) (Table 1; Figures 1A,B). In the therapeutic CIA model, we took advantage of the presence of an infectious antigen, Mycobacterium tuberculosis (MBT), mixed with collagen in the complete Freund’s adjuvant used for the induction of arthritis, to analyze T cell responses against another antigen than the autoantigen (i.e., bovine type II collagen). In contrast to the response observed with collagen, a similar cell proliferation against MBT antigen is found between cells from apoptotic cell-treated and untreated CIA mice. Moreover, the suppressive activity of Treg sorted from apoptotic cell-treated arthritis mice is restricted to collagen and not extended to MBT (20). This strongly demonstrated that the infusion of apoptotic cells allows the induction of pTreg in vivo with an antigenic specificity restricted to the collagen autoantigen. The same effect (i.e., induction of autoantigen-specific pTreg but not infectious antigen-specific Treg) has previously been reported in a similar therapeutic approach based on to the in vivo generation of apoptosis (88). Further works are necessary to explain why apoptotic cell infusion favors the induction of autoantigen-specific Treg. Nevertheless, other teams have reported the induction of IL-10-dependent Treg (13, 15) after the prophylactic infusion of apoptotic cells in arthritis models. Thus, this confirms the induction of pTreg after apoptotic cell infusion and may explain the anti-inflammatory effect in the joint whatever the administration route since the generated pTreg may migrate to inflamed joints.

Concerning safety reasons, one has to evoke the high plasticity of CD4⁺ T cells, and in particular, pTreg that have in common with pro-inflammatory Th17 cells the requirement of TGF-β (36, 89). Apoptotic cell-induced pTreg polarization can be influenced by a simultaneous microbial infection providing IL-6 necessary for Th17 differentiation. The coincident production of IL-6 and TGF-β in response to bacteria and apoptotic epithelial cell death, during orogastric bacterial infection, leads to the generation of both bacteria-specific and autoreactive Th17 cells (87, 90). Similarly, the production of IL-6 together with TGF-β has been shown when “activated” apoptotic cells or apoptotic cells containing high amounts of demethylated DNA have been infused in mBSA arthritis model instead of “resting” apoptotic cells rather containing methylated DNA (48, 91). In the same model, the infusion of “resting” apoptotic thymocytes decreases Th17 cells in the inguinal draining lymph nodes (15). This dichotomy between anti-inflammatory pTreg and pro-inflammatory Th17 cells is not so simple, since different Th17 cell subsets have been now described including pro-inflammatory and anti-inflammatory Th17 cells (89). To date, these subsets have not been studied in the settings of apoptotic cell infusion.

Here, we will evoke the clinical perspectives of apoptotic cell infusion. This is based on the preclinical data (Table 1), but also on data obtained in the field of cancer research. There is an extensive literature on the immunomodulation by dead and dying cells in the setting of cancer (92). We propose to discuss the critical points to achieve a beneficial therapeutic effect (73). These are the following: (i) peripheral blood leukocytes are the easiest and major source of apoptotic cells to consider in human settings, while apoptotic cells used in the preclinical studies were other apoptotic leukocytes [i.e., thymocytes (13–15, 20) or DC (48), Table 1] from a practical point of view. In RA patients, cytapheresis has to be considered to achieve a sufficient number of apoptotic cells as it has been done in the clinical trial in the setting of hematopoietic cell transplantation (16). The highest number of apoptotic leukocytes planned to be infused is 210 million of cells per kilogram. This will require to pool two sequential cytaphereses. Donor-derived apoptotic cells (i.e., allogeneic cells) will not be considered in the first instance for ethical/regulatory purposes; patient (i.e., syngeneic) apoptotic leukocytes are considered as a cell-based product by the French regulatory agency, while apoptotic cells from healthy volunteers correspond to advanced therapy medicinal products. Nevertheless in experimental models, prevention of arthritis is observed independently of the apoptotic cell origin (i.e., syngeneic, allogeneic, or even xenogeneic) (2); (ii) a tolerogenic signal inducing early-stage apoptotic cells, that is, leukocytes stained by annexin-V but little or no staining with PI or 7-AAD dyes (Table 1). These stimuli correspond to γ- or ultraviolet B (UVB)-irradiation (73, 93). Stimuli inducing apoptotic cell death have been particularly studied in the field of cancer research in order to generate immunogenic dying tumor cells to favor tumor rejection. A recent study performed in vivo using melanoma cells is particularly informative on these stimuli (93). The authors have compared three different apoptotic signals and have confirmed that UVB-irradiation generates non-immunogenic apoptotic cells. Furthermore, the authors have identified that the production of IL-27 and IL-1α by bone marrow-derived macrophages after in vitro incubation with apoptotic tumor cells predicts immunogenicity. In addition, this work shows that primary necrotic cells induced by tuberculosis-necrotizing toxin in vivo are also non-immunogenic (93). This confirms that necrotic cells induced by a repeated freeze/thaw procedure or obtained by incubating apoptotic cells for 24 h before infusion are very poor inducers of CD8⁺ T cell responses in vivo (94). The immunogenicity of necrotic cells remains, however, a matter of debate (92) that we do not want to comment further here; (iii) one infusion appears sufficient whereas multiple infusions may expose to a risk of immunization against apoptotic cell-derived antigens (95), as discussed in Ref. (92). The question arises as to how long the therapeutic effect will last. In the CIA model, the therapeutic effect of early-stage apoptotic cell infusion is transient but prolonged when associated with anti-TNF therapy (20). Only clinical studies will allow to answer to this question; and (iv) a systemic administration route can be considered while local administration is also possible. In experimental arthritis models, two distinct systemic routes of administration [i.e., i.p. (13, 14) versus i.v. (13, 15, 20, 48)] have been tested with a similar efficacy (Table 1). However, no local administration has been evaluated so far. Another lesson coming from cancer research on dying/dead cells is the ability of apoptotic tumor cells to stimulate the proliferation of nearby viable tumor cells (96, 97). This apoptotic cell-mediated proliferation is not restricted to tumor cells (98). Relevant to the present review, primary human synovial fibroblasts isolated from knee joints of RA patients are also able to proliferate in vitro when these fibroblasts are in close contact with apoptotic tumor cells (97). Considering the therapeutic use of apoptotic cell infusion, it is, however, reassuring to see that when the number of apoptotic tumor cells is increased, the proliferative effect is limited (97). The percentage of infused early-stage apoptotic cells planned to be infused to RA patients is higher than 50%. However, one has to be cautious on this apoptosis-induced proliferative effect, since it is mainly mediated by a soluble factor, the nucleoside inosine (97).

Finally, one has to remain cautious, since a different effect can be obtained depending on the infusion of “resting” versus “activated” apoptotic CD11c⁺ cDC (48). This may be related to the methylation status of DNA from apoptotic cells (91). This work found that, as apoptotic cDC, apoptotic CD4⁺ T cells from RA patients exhibit a DNA demethylated status, suggesting a pro-inflammatory effect after infusion associated with IL-6 secretion (91). Whether this may impact on the therapeutic efficacy of apoptotic cell infusion remains to be determined. We used apoptotic splenic cells from arthritic mice (i.e., containing multiple activated leukocytes) in the therapeutic CIA model, and we observed the same effects as apoptotic thymocytes (Bonnefoy F., Perruche S., unpublished results). An additional security can be also provided by the addition of csDMARD, such as low-dose MTX or bDMARD, such as TNF inhibitors. Indeed, these treatments do not inhibit the beneficial therapeutic effects of apoptotic cell infusion (20). MTX (at high-dose) has been also used as prophylaxis of graft-versus-host disease in the clinical trial testing the effects of donor early-stage apoptotic cell infusion (16). This confirms our experimental data in CIA: MTX does not affect the therapeutic effect of apoptotic cell infusion, and allows to preserve its beneficial effect on collagen (autoantigen)-specific Treg (20). In addition, the capacity of splenic pDC and macrophages to induce ex vivo pTreg polarization is not inhibited by MTX (20). Thus, MTX can be continued if an apoptotic cell-based therapy has to be proposed to patients. Compared with MTX, anti-TNF therapy has the advantage to synergize with apoptotic cell infusion to control ongoing arthritis (20). However, the exact mechanism responsible for this synergy has not been identified (20). In the future, IL-6 inhibitors, such as tocilizumab, can be also envisaged to be associated with apoptotic cell infusion in order to prevent the antagonistic effect of IL-6, previously reported in mBSA-induced arthritis (48, 91). This can be a way to neutralize the effects of the methylation status of DNA from apoptotic T cells obtained from RA patients (91).

Apoptotic cell infusion can potentially be associated with corticosteroids without any risk. Indeed, corticosteroids enhance apoptotic cell removal by inducing the expression of the PtdSer-binding protein, milk fat globule-EGF factor 8 (MFG-E8) selectively in human and mouse monocytes and macrophages (whatever their differentiation profile, M1 or M2) (99).