Information

How extensive is CD47?

How extensive is CD47?


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

CD47 aka the "don't eat me" signal has recently been claimed to be expressed on all tumor cells. This doesn't seem to corroborate with other cell-biology experiments. On what other cells is CD47 expressed?


I don't know how extensive. Let's run a simple data query and find out: Go to GEO at NCBI. In the "Gene profiles" window, type CD47, and hit enter to launch the query. At the top of the resulting page, use the link labeled "Limits" to restrict the 7000+ results to human by entering the term "human" in the "DataSet organism" window. So, there are 3700+ results. Look through them to get an idea of which cell types and under which conditions CD47 is expressed.

Let's try a second method. Go to the BioGPS portal. Enter CD47 in the appropriate window and run the query. From the results, select the row with ID # 961 as this represents the human gene CD47. The resulting gene expression/activity chart for Hs (human) will show you in more general terms where CD47 is expressed.


I can't give a comprehensive answer to this, but the CD proteins are found on white blood cells of various types. If you think about it its impossible that a surface protein could be only observed on cancer cells as cancer is a cell line which is a reproductive dead end. You can see on the geneatlas box on the lower right side of the page - its gene is expressed, at least to a small amount, on nearly every major tissue type.


Abstract

Macrophages recognize CD47 as a marker of “self” and phagocytose CD47 null hematopoietic cells. Using CD47 chimera models, here, we show that the phagocytic activity of macrophages against CD47 null hematopoietic cells is conferred by CD47 expression on nonhematopoietic cells, and this �ucation” process is hematopoietic cell-independent. Macrophages in the chimeras where nonhematopoietic cells express CD47 phagocytose CD47 null cells, whereas those in the chimeras lacking CD47 on nonhematopoietic cells are tolerant to CD47 null cells. However, macrophages in the latter chimeras retain phagocytic activity against CD47 null RBCs, demonstrating a split macrophage tolerance to CD47 null hematopoietic cells. The findings highlight the potential importance of nonhematopoietic cells in the regulation of macrophage function, and suggest a previously uncharacterized mechanism of macrophage tolerance.

Macrophages are essential to both innate and adaptive immune responses. They initiate the innate immune response by recognizing pathogens by means of a range of receptors with specificity for pattern-recognition molecules (1, 2). Macrophages also express inhibitory receptors that control macrophage activity within tissues (3). One of these receptors called signaling regulatory protein (SIRP)α (also known as CD172a, SHPS-1) is expressed on macrophages and dendritic cells (DCs), which recognizes CD47, a pentaspan membrane glycoprotein ubiquitously expressed in all tissues (4 𠄶). Previous studies have shown that CD47 is an essential marker of “self,” and its signaling through SIRPα prevents inappropriate self-phagocytosis (7). Hematopoietic cells not expressing CD47 are rapidly cleared from the bloodstream by macrophages and DCs (7, 8) [supporting information (SI) Fig. 6]. Because only CD47 KO cells are cleared in WT mice receiving a mixture of CD47 KO and WT cells, CD47 expression on an individual cell, but not on macrophages or other surrounding cells, is required for inhibition of phagocytosis, and such a protection is likely mediated by a direct interaction between CD47 on a target cell and SIRPα on macrophages (SI Fig. 7).

Interestingly, macrophages and DCs from CD47 KO mice do not phagocytose CD47 KO cells. Because the level of SIRPα expression on CD47 KO macrophages is virtually the same as that of WT macrophages, this tolerance to CD47 KO cells is unlikely induced at the level of SIRPα expression (7). It is possible that lack of CD47 expression on macrophages or in the environment may regulate macrophage function. Alternatively, such tolerance might be acquired during the early stage of prenatal development as compensation for the disrupted gene, and may not occur in normal macrophages. Furthermore, it remains unknown why CD47, but not the ligands for other inhibitory receptors, is used by macrophages to identify self in WT mice. To address these questions and to understand the mechanisms of macrophage tolerance, here, we have evaluated the phagocytic activity against CD47 KO cells of macrophages from various types of CD47 chimeras, in which CD47 is expressed on hematopoietic cells, nonhematopoietic cells, or on cells of both compartments.


INTRODUCTION

The extension of axons from neurons to their target cells during development of the nervous system is guided by a variety of environmental cues, which include diffusible chemoattractants and chemorepellents, extracellular matrix proteins, and cell adhesion molecules (Tessier-Lavigne and Goodman, 1996 Dickson, 2002). In response to these guidance cues, the growth cone of the axon produces and retracts filopodia and lamellipodia, processes that require the temporal and spatial regulation of the actin cytoskeleton. Members of the Rho family of small G proteins, including Rho, Rac, and Cdc42, are implicated as key mediators that link guidance signals to rearrangement of the actin cytoskeleton (Ridley et al., 1992 Nobes and Hall, 1995 Kozma et al., 1997 Luo et al., 1997 Hirose et al., 1998 Takai et al., 2001 Arakawa et al., 2003). Although Rac and Cdc42 promote neurite extension, Rho inhibits it or induces growth cone collapse. These proteins are also implicated in dendritic development (Threadgill et al., 1997). Although certain guidance factors, including slit, semaphorin, and ephrin, have been shown to regulate Rho family proteins (Wahl et al., 2000 Whitford and Ghosh, 2001 Wong et al., 2001), it remains unclear how other extracellular cues control neurite extension through these small G proteins.

CD47, also known as integrin-associated protein (IAP), was originally identified in association with the integrin αvβ3 (Brown et al., 1990). It is a member of the immunoglobulin (Ig) superfamily, possessing an Ig-V–like extracellular region, five putative transmembrane domains, and a short cytoplasmic tail (Brown and Frazier, 2001). CD47 is implicated in the regulation of multiple cellular processes including neutrophil migration (Cooper et al., 1995 Parkos et al., 1996), T-cell activation (Reinhold et al., 1997 Waclavicek et al., 1997), T- and B-cell apoptosis (Mateo et al., 1999 Pettersen et al., 1999), and platelet activation (Chung et al.,1997, 1999). The extracellular region of CD47 is responsible for its association with the integrin β3 subunit (Lindberg et al., 1996b). Although most CD47-mediated cellular responses likely involve the activation of integrins, in particular that of αvβ3 or αIIbβ3 (Brown and Frazier, 2001), the molecular mechanism of such activation is not fully understood. The extracellular region of CD47 also binds the putative ligands thrombospondin-1 (TSP1 Gao et al., 1996 Chung et al., 1997 Brown and Frazier, 2001) and SHP substrate-1 (SHPS-1 Jiang et al., 1999 Seiffert et al., 1999).

SHPS-1, also known as BIT or SIRPα, is a receptor-like transmembrane protein that contains three Ig-like domains in its extracellular region as well as putative tyrosine phosphorylation sites and binding sites for the SH2 domains of the protein tyrosine phosphatases, SHP-2 and SHP-1, in its cytoplasmic region (Fujioka et al., 1996 Ohnishi et al., 1996 Kharitonenkov et al., 1997). Through its formation of a complex with SHP-2, SHPS-1 promotes cell migration by regulating reorganization of the actin cytoskeleton (Inagaki et al., 2000). We and others have recently shown that CD47 and SHPS-1 constitute a cell-cell communication system (the CD47–SHPS-1 system) that plays an important role in a variety of cell functions. The engagement of SHPS-1 by CD47 inhibits formation of the SHPS-1–SHP-2 complex and thus contributes to inhibition of cell migration by cell-cell contact (Motegi et al., 2003). The binding of CD47 on red blood cells to SHPS-1 on macrophages also inhibits the phagocytosis of red blood cells by macrophages, a process that requires the interaction of SHP-1 with SHPS-1 (Oldenborg et al.,2000, 2001). Monoclonal antibodies to CD47 inhibit neutrophil transmigration (Liu et al., 2001), suggesting that the CD47-SHPS–1 system might mediate bidirectional inhibitory regulation of cell migration. The signaling pathway activated by the binding of SHPS-1 to CD47 remains largely unknown, however. In contrast, certain cellular responses triggered by the binding of TSP1 to CD47 appear to be mediated by the pertussis toxin (PTX)-sensitive heterotrimeric G protein Gi (Frazier et al., 1999 Brown and Frazier, 2001).

In the CNS, both CD47 and SHPS-1 are localized in synapse-rich regions such as the hippocampus and cerebellum (Jiang et al., 1999 H. Ohnishi and T. Matozaki, unpublished observation). In the retina, both molecules are localized at presynaptic or postsynaptic sites (or both) however, SHPS-1 fails to associate with synaptic sites in CD47 knockout mice, suggesting that the interaction of SHPS-1 with CD47 is necessary for its synaptic localization (Mi et al., 2000). In addition, the abundance of CD47 in the hippocampus has been associated with memory retention in rats (Huang et al., 1998), and long-term potentiation (LTP) is impaired in CD47 knockout mice (Chang et al., 1999), suggesting a role for CD47 in synaptic plasticity and memory formation in the hippocampus. Together, these observations suggest that, through its interaction with SHPS-1, CD47 might mediate neuronal cell-cell communication in a bidirectional manner and thereby modulate synaptic transmission. Four alternatively spliced isoforms of mouse CD47 mRNA have been identified, one of which, designated form 4, is most abundant in the brain (Reinhold et al., 1995).

Despite these various observations, however, the physiological functions of CD47 in neuronal cells and the mode of action of this protein remain mostly unknown. We have now examined the role of CD47 in the regulation of neurite remodeling in N1E-115 neuroblastoma cells.


CD47/SIRPα, AN IMMUNE CHECKPOINT FORINNATE IMMUNE SYSTEM

Among cells of the myeloid lineage, macrophage has prominent potentials as the mediator of anti-cancer therapeutics based on its robust phagocytosis ability [ 8, 9]. CD47, known as an integrin-associated protein, was first identified as a transmembrane protein from red blood cells (RBC) [ 10]. Signal-regulatory protein alpha (SIRPα), a transmembrane protein on macrophage, is the main receptor of CD47 [ 11]. CD47 binding to SIRPα triggers the coupling of SIRPα to these phosphatases, thereby delivering the ‘don’t eat me’ signals to macrophage then preventing their activation [ 8, 12, 13]. CD47 expression is regarded as a self-protective mechanism of normal cells, including transfused RBC, lymphocytes and platelets, to resist the elimination of macrophage phagocytosis [ 14]. It is well documented that kinetic hematopoietic stem cells (HSC) protect themselves from macrophage phagocytosis by upregulating the expression of CD47 as they pass across sinusoids then they decrease CD47 expression after relocating to the marrow [ 15]. In addition, the level of CD47 expression predicts the probability whether HSC could be phagocytized by macrophage while circulating [ 16]. While, the absence of CD47–SIRPα interaction could activate pro-phagocytic receptors to trigger macrophage phagocytosis of RBC [ 17].

Targeting CD47/SIRPα axis for the therapy of cancer. Anti-CD47 mAb, Anti-SIRPα mAb and SIRPα-Fc fusion protein enhance tumor phagocytosis by macrophages, enabling the presentation of tumor antigens to T cells. TCR, T cell receptor. MHC, major histocompatibility complex.

Targeting CD47/SIRPα axis for the therapy of cancer. Anti-CD47 mAb, Anti-SIRPα mAb and SIRPα-Fc fusion protein enhance tumor phagocytosis by macrophages, enabling the presentation of tumor antigens to T cells. TCR, T cell receptor. MHC, major histocompatibility complex.

In the process of carcinogenesis, overexpression of CD47 has been identified across most tumors ( Fig. 1). In hematological cancer, CD47 expression on acute myeloid/lymphoblastic leukemia, non-Hodgkin’s lymphoma cells and bone marrow of multiple myeloma samples was detected as several folds increase compared with normal tissues and CD47 level was predictive of the overall survival to primary treatment [ 16, 18]. Furthermore, studies also found that solid tumors, including breast, ovarian, bladder, colon, glioblastoma, prostate tumor and hepatocellular carcinoma, expressed about three- to five-fold more CD47 than the corresponding normal tissues [ 19]. When patients were assigned into ‘CD47 low’ and ‘CD47 high’ groups based on a univariate analysis, high CD47 mRNA expression level was shown to be associated with the decreased progression-free survival and overall survival [ 19]. Therefore, those evidences indicated that CD47/SIRPα axis was exploited by malignant cells to transmit ‘don’t eat me’ signal to evade macrophage surveillance and phagocytosis, which highlighted that blocking CD47/SIRPα axis could be used to promote the ability of macrophages to phagocytose and eliminate tumor cells.


3. Mechanisms Underlying B-Cell Lymphoma Refractoriness to Immune Checkpoint Blockade

Resistance to immune checkpoint blockade therapy in human cancers was extensively reviewed by Bellone M and Elia A [295] however, the mechanisms underlying B-cell lymphoma refractoriness to immune checkpoint blockade are still poorly understood.

Although major advances have been made in the last 20 years to overcome the refractoriness of B-NHL patients to standard therapies through the introduction of immune checkpoint blockades in the clinical setting, either as single agents or in combination therapies ( Table 1 and Figure 2 ), several parameters can impair the efficacy of these new approaches. Patient-intrinsic factors such as age, sex, HLA heterozygosity or loss of 㬢-microglobulin (B2M), amplification of oncogenic signalling pathways, immunosuppressive cells and molecules present in the TME may impair antigen recognition and contribute to the failure of immune checkpoint blockades [296,297] ( Figure 3 ).

Mechanisms of resistance to immune checkpoint blockade. The deregulation of MHC class I components such as the loss of 㬢 microglobulin (B2M) and the loss of human leukocyte antigen (HLA) heterozygosity as well as defects in IFN signalling pathways may impair antigen recognition by antitumor CD8+ T-cells. Amplification of oncogenic signalling pathways such as PI3K/AKT/mTOR, Wnt/β-catenin, and MAPK increases the production of immunosuppressive cytokines, trigger T-cell exclusion from TME and may also result in resistance to immune checkpoint blockade. Epigenetic (histone acetylation or DNA methylation) and genetic (deleterious mutations) alterations are crucial triggers of gene expression disorders related to sustained T-cell exhaustion that could eventually cause the failure of immune checkpoint therapy. Moreover, myeloid-derived suppressor cells (MDSCs), Tregs, tumour-associated macrophages (TAMs), and cancer-associated fibroblasts (CAFs) are major immunosuppressive cell types within the TME that may contribute to resistance to immune checkpoint blockade. Immunosuppressive molecules such as TGF-β and IFN-γ, secreted by tumour cells, myeloid cells and macrophages in the TME, may also suppress the functions of effector T-cells, rendering immune checkpoint blockade ineffective.

The frequent PD-L1 aberrant expression that is found among lymphoma patients results in the most responsive cancer type to anti-PD1 therapy. However, the PD-1/PD-L1 blockade can be strongly influenced by disease-specific factors, and its predictive value in clinical trials is still controversial. The inconsistent data could be attributable mainly to the variable PD-L1 resources (tumour cells, tumour microenvironment cells, peripheral blood), the differences in staining (including detecting antibodies) procedures, and positive/negative PD-L1 cut-offs. Additionally, it has been shown that PD-L1 can interact in cis with CD80 on APCs and then disrupt the binding between PD-1 and PD-L1 [298]. The functional effects of alternative binding partners also highlight the differences seen in the efficacy of PD-1/PD-L1 immunotherapy in different biological settings.

There are several mechanisms determining if a patient will respond or not to a PD-1/PD-L1 blockade. Weakly immunogenic tumours may have an insufficiently active T-cell population to respond to PD-1/PD-L1 blockade additionally, potentially immunogenic tumours will also be resistant to PD-1/PD-L1 blockade if they develop mechanisms to suppress the activation and infiltration of T-cells after the treatment. Moreover, patients may become resistant to PD-1/PD-L1 therapy if they have insufficient reinvigoration of exhausted tumour-specific CD8+ T-cells or if they have lost target antigens or the ability to present them. Finally, some patients might initially respond to PD-1/PD-L1 blockade but become resistant if the antitumour T-cells are short-lived [299].

There are also epigenetic mechanisms of B-lymphoma cells driving the resistance to immune checkpoint blockades. The in vitro and in vivo data from Zheng et al. [300] showed that miR-155 overexpression enhances PD-L1 expression, reduces peripheral blood immune cells, induces CD8+ T-cell apoptosis and dysfunction via AKT/ERK dephosphorylation, and decreases the survival of DLBCL patients. It has also been shown that histone deacetylase 3 (HDAC3) is another important epigenetic regulator of PD-L1 in B-cell lymphoma as its inhibition increases PD-L1 transcription, resulting in a better clinical response to PD-L1 blockade [301].

The molecular mechanisms of the immune environment in regulating the efficacy of immune checkpoint blockades in B-cell lymphoma is poorly understood. It has been shown that T-cell-inflamed tumours are enriched for sensitivity to PD-1 blockade therapy [302]. Conversely, T-cell noninflamed tumours present low-infiltrating immune cells and are typically resistant to immune checkpoint blockade therapy [303]. Inflamed lymphomas are characterised by the presence of prominent T-cell infiltration [304], genetic alterations that facilitate escape from immune surveillance [78,82,305,306], and frequent mutations, resulting in hyperactivity of the NF-kB signalling pathway [78,307].

Hyperprogressive disease has been found in 9% of patients who received anti-PD-1/PD-L1 therapy [308,309]. The hyperprogression is associated with the elderly age of patients but not tumour burden or cancer type [308]. MDM2/MDM4 amplification and EGFR aberration are correlated with a higher risk of hyperprogression in solid cancers [309], although it was little known in lymphoma. Recent data showed that patients experiencing an hyperprogression have a higher prevalence of PD-L1 − disease [308]. It is predicted to be because the engagement of PD-1 with anti-PD-1 mAb inhibits but does not augment T-cell activations. Therefore, anti-PD-1 mAbs might be PD-1 agonists rather than antagonists in PD-L1 − status. The disease might also rapidly progress through interaction between PD-L1 and CD80 instead of PD-1 for the blocking duration by anti-PD-1 mAbs in PD-L1+ status [310]. Some polymorphisms of PD-1 could also affect the action of anti-PD-1 mAbs, and, thus, hyperprogression could be possible after PD-1 blockade [311].


Results

Design and preparation of SIRPα and CD47 soluble domains

To recapitulate the biological interaction of CD47 and SIRPα, we designed expression constructs to produce their recombinant, soluble, extracellular interaction domains. CD47-CD4-6His was produced in a mammalian expression system and was previously shown to bind recombinant SIRPα [77]. As expected, recombinant CD47-CD4-6His was highly glycosylated, migrating as a diffuse band at

60 kDa, as compared to the molecular weight of 40 kDa predicted from the protein sequence (S1 Fig). The heterogeneity in glycosylation of CD47 has no reported effect on its ability to bind SIRPα [78].

Non-glycosylated SIRPα binds CD47 with high nM affinity [78]. However, recombinant SIRPα containing abnormally high and heterogeneous glycosylation showed impaired CD47 binding [79]. To reduce biological assay variability, we chose to produce SIRPα in a glycosylation-deficient system. SIRPα-Avi (Isoform 1 allele 1) is expressed in E. coli as a fusion construct that comprises an N-terminal purification tag followed by the extracellular CD47-binding fragment of SIRPα linked to a C-terminal Avi-Tag. After affinity capture and tag cleavage, SIRPα-Avi was biotinylated in vitro (SIRPα-biotin) using purified recombinant E. coli biotin ligase (BirA) to >98% incorporation, as measured by mass spectrometry (S2 Fig). The purified protein is monomeric as judged by SEC. Small-angle x-ray scattering (SAXS) studies (see S1 File) confirmed that SIRPα is monomeric in solution (up to at least 1.8 mM), with the C-terminal Avi-tag positioned away from the CD47-binding region, demonstrating its suitability for our assay (S3 Fig).

To validate the biologically relevant interaction between our PPI reagents, we used SPR spectroscopy to measure their steady state binding affinity (Fig 2). SIRPα-biotin was immobilized on a SAV-chip with CD47-CD4-6His injected at multiple concentrations to yield a Kd of approximately 200 nM. This is comparable to a previous report of 300 nM and 279 nM employing similar constructs and measurement via SPR [66,80] and 470 nM for soluble CD47 binding to SIRPα-expressing CHO cells measured by fluorescence intensity [81].

SIRPα-biotin was immobilized on a SAV-chip followed by injection of CD47-CD4-6His at varying concentrations. (A) The association and dissociation responses are shown as an overlay for each concentration tested. (B) The maximum response of the association phase was plotted against concentration and fitted to a steady-state binding curve to approximate the Kd of the interaction pair.

Assay development

We developed a TR-FRET assay to quantitatively measure the interaction of CD47 and SIRPα [82]. We started with the TR-FRET reagents from CisBio for feasibility and optimization studies. The donor reagent consists of an anti-6His monoclonal antibody tagged with a Terbium cryptate (61HI2TLA) to provide a long-lived fluorescent signal bound to the 6His tag of CD47 (Fig 3A). The acceptor consists of streptavidin tagged with an XL665 fluorochrome (SAV-XL665 610SAXLA) to associate with the biotin tag on SIRPα. Feasibility testing was conducted in 384-well format to assess signal to background and assay specificity. Using the experimentally determined Kd as a starting point (see Fig 2), CD47 and SIRPα concentrations of 100 nM resulted in a S/B of 5 compared to the same mixture lacking CD47 (Fig 4A). To find the optimal concentration of protein reagents for each assay, each component was titrated to saturation (Fig 4B and 4C). We then miniaturized the assay volumes for use in 1536-well format and proceeded to test feasibility and further optimize the assay (Fig 4D). Finally, using the adjusted ligand protein concentrations, we optimized assay buffer, donor and acceptor levels, plate type, order of reagent addition, and incubation time (Table 1 and S4 Fig). Briefly, we found that nM concentrations of CD47 (12.5 nM) and SIRPα (100nM) in PBS buffer with minimal detergent (0.005% IGEPAL CA-630) and BSA (0.1%) produced a robust assay.

TR-FRET based assays (upper panel) and ALPHAScreen assay (lower panel). CD47 and SIRPα represented as their x-ray crystal structures from PDB 2JJS, chain D and chain B, respectively.

(A) Initial assay feasibility performed in 384-well plate format using CisBio TR-FRET reagents. CD47 and SIRPα were tested at 2 concentrations (20 nM and 100 nM), and the FRET ratio of the complete reaction was compared to controls lacking either CD47 or SIRPα to calculate the S/B (n = 16). (B) CD47 or (C) SIRPα concentrations were titrated while the other was fixed as indicated in (A) in 384-well plates to find their optimal assay concentrations as judged by S/B (n = 4). (D) Assay performance in 384-well or 1536-well formats compared using optimized concentrations derived from panels B and C. Assay Z’ indicated for each plate type (n = 32).

Second, we developed the LANCE (LANthanide Chelate Excite) TR-FRET assay, which consists of an anti-6His monoclonal antibody linked to an Eu-chelate (Anti-6His-Eu AD0110) for the donor associated with the CD47 reagent and streptavidin linked to allophycocyanin (SAV-APC AD0201) as acceptor associated with SIRPα. Using the optimized CD47 and SIRPα concentrations from the CisBio assay optimization as a starting point, the LANCE assay reagents were titrated in 1536-well format to arrive at a S/B of 12 (Table 4 and S5 Fig) using optimal conditions outlined in Table 1. As with the CisBio assay (S4 Fig), the endpoint is stable (≥80% of initial signal) for at least 48 h (S6 Fig).

Third, we developed the AlphaScreen assay as a cross validation orthogonal assay. It uses a bead-based luminescent oxygen channeling [83] technology based on the transfer of energy from the laser-excited (680 nm) donor bead to acceptor beads in the form of singlet oxygen to produce a luminescent signal (520–620 nm) [84,85]. Singlet oxygen has a limited lifetime (4 μsec half-life) prior to relaxing back to ground state during which it can diffuse approximately 200 nm in solution. If an acceptor bead is within that distance, energy is transferred from the singlet oxygen to thioxene derivatives within the acceptor bead, subsequently culminating in light production at 520–620 nm. In the absence of an acceptor bead, singlet oxygen relaxes to ground state and no signal is produced. This proximity-dependent chemical energy transfer is the basis for ALPHAScreen's homogeneous format (Fig 3).

The ALPHAScreen assay was optimized using the CD47 and SIRPα concentrations from the CisBio assay development as a starting point. While the optimal CD47 concentration remained unchanged throughout the three assays, the optimal SIRPα concentration is 4-fold lower for the ALPHAScreen assay, possibly due to the differences in SAV-bead-binding capacity versus streptavidin in solution. Optimization also allowed for a 2-fold lower bead concentration than suggested by the manufacturer, reducing the cost per data point [86]. The resulting ALPHAScreen assay yielded robust HTS parameters (Z’ of 0.79, S/B = 75 Table 4), a slightly faster read time than the TR-FRET assays (9 min/plate vs. 14 min/plate), and a common assay buffer and plate type with the CisBio and LANCE assays (Table 1).

Assay validation

We evaluated the performance of all three assays in a qHTS format using SIRPα-cold (non-biotinylated) as a specific inhibitor of SIRPα-CD47 interaction. The IC50 values exhibited good agreement with those obtained from the TR-FRET assay (2.8 μM and 1.1 μM for CisBio and LANCE, respectively) and the ALPHAScreen assay (0.55 μM Fig 5A). Hill slopes indicate a near ideal bimolecular interaction (1.2–1.5, Fig 5A). We also evaluated the ability of a small molecule to act as an inhibitor in the assay, as the eventual goal of the HTS assay will be to screen large chemical libraries for small molecule antagonists of the protein-protein interaction between SIRPα-CD47. As there are no known small molecule inhibitors of this interaction, we chose to use free biotin to inhibit the interaction between SIRPα and the streptavidin-conjugated assay reagents. Free-biotin was added in advance to the SIRPα-CD47 complex prior to the addition of the streptavidin-conjugated reagent to overcome the extraordinarily slow off-rate of the complex once formed [87]. The activity of biotin was similar for all assays and exhibited the same inter-assay potency ranking as the large molecule SIRPα-cold (AlphaScreen>LANCE>CisBio Fig 5B). The IC50 values of 260, 170, 115 nM were comparable to results generated by Perkin Elmer for similar assays evaluating the interference of biotin in assay media [88]. The cooperativity indicated by the Hill slope of between 2 and 3 is likely due to the multivalency of streptavidin and reflects the need to block multiple sites before the proximity signal is eliminated. This contrasts with SIRPα-cold’s interaction with the bivalent CD47-CD4-6His/anti-His antibody shown above. Assay robustness indicated by Z’ factor was “excellent” [74] (see Table 4) and allowed for advancement to pilot screening using the 1280 compound LOPAC Library (Sigma).

(A) Assay specificity was validated using SIRPα-cold titrated to inhibit the CD47-binding activity of SIRPα-biotin in the screening assays. Hill slope and IC50 are indicated (n = 4). (B) Assay performance using a positive control small molecule (biotin) was assessed in all three assays (n = 1). Normalized activity calculated with neutral control (+CD47, +SIRPα, no inhibitor) as 0% and low control (-CD47 +SIRPα, no inhibitor) as -100% activity. Assay activity with tested agents was then compared and normalized to these controls.

Pilot screening assay comparison

The LOPAC library was tested in qHTS format using a 7-point inter-plate titration spanning 38 μM to 2.4 nM (final concentration) in 1536-well format. All assays displayed robust performance as judged by Z’ factor, a statistical measure of the separations of the positive and negative control distributions. The Z’ was determined using column 1 and 2 of each plate as positive and negative controls, respectively (Figs 6 and 7).

Results for the (A) CisBio, (B) LANCE, and (C) AlphaScreen assays shown. Individual signal data points from neutral control (+CD47, +SIRPα, no inhibitor) and low control (+CD47 -SIRPα, no inhibitor) plotted by well for each plate in the 10-plate pilot screening run (n = 32 points per plate). Dotted lines represent + and– 20% of average signal for each high and low control. The Z’ factor was calculated from the high and low controls on each plate and averaged across all plates.

Results for the (A) CisBio, (B) LANCE, and (C) AlphaScreen assays shown. SIRPα-cold was titrated as a control inhibitor in each assay (n = 2). Error bars represent the standard deviation of two replicates. The average IC50 and minimum significant ratio (MSR) was calculated over the 10-plate run for each screening assay. MSR is defined as the smallest ratio between the potencies of two compounds that is statistically significant and is calculated as MSR = 10 2√2s , where s is an estimate of the standard deviation of a log potency for one compound.

Activity of library compounds was analyzed according to the curve classification (cc) as previously described [89,90]. Briefly, compounds were considered active when they displayed either concentration-dependent inhibition with two asymptotes and incomplete activity (-50–90% cc -1.2), complete activity at the highest tested concentrations (<-90% cc -2.1), or complete inhibition with two asymptotes indicating a saturable effect (<-90% cc -1.1), as shown in Table 5. The highest overall activity was demonstrated in the AlphaScreen (17 out of 1280 1.3% Fig 8) followed by the CisBio TR-FRET assay (15 out of 1280 1.2%) and the LANCE TR-FRET assay (3 out of 1280 0.2%).

Results shown using (A) CisBio, (B) LANCE, and (C) AlphaScreen assays. Concentration response curves fit to data of active compounds using a 4-parameter logistic equation are included with solid lines. Black lines indicate activity only within the indicated assay. Blue lines indicate activity in both CisBio and AlphaScreen assays. Red lines indicate activity in all three assays.

We further characterized activity of these compounds with additional assay readouts, counter-screening assays, data characteristics, and reported compound promiscuity. Assay interference by chemical matter can take many forms in biochemical HTS assays [91,92] and can be mediated by compound fluorescence or aggregation, signal attenuation by absorbance, or singlet oxygen quenching and affinity tag-disruption in the case of the AlphaScreen.

In the TR-FRET assays, fluorescent compounds or attenuators were identified using an additional readout of donor channel fluorescence data. Changes in donor fluorescence that mirror or contribute to overall activity (ratio of FRET/donor) indicate assay interference and these compounds were removed from further consideration [93]. Donor channel artifacts were observed for 14 of the 15 CisBio assay active molecules (Fig 9), and for 2 of the 3 LANCE actives (Fig 10), resulting in 6-hydroxy-DL-DOPA as the only remaining candidate in both TR-FRET assays, which is known to have activity in a number of unrelated assays and is considered a promiscuous molecule. For AlphaScreen-active compounds, we employed an additional counter-screen assay (TruHits, Perkin Elmer) where the tagged SIRPα and CD47 interaction pair was replaced by a single dual-tagged peptide (Biotin-peptide-6XHis). Compounds displaying activity in this counter-screen do not specifically disrupt the SIRPα-CD47 interaction but are rather affinity tag disruptors, singlet oxygen or fluorescence quenchers, or possibly protein aggregators. Accordingly, 3 of 17 active molecules displayed counter-screen activity and were removed from further consideration (Fig 11).

Normalized activity signal for the donor and acceptor channels and calculated ratio signal for each active compound reported for the 7-point concentration response generated using qHTS. Normalized activity was calculated with neutral control (+CD47 +SIRPα, no inhibitor) as 0% and low control (+CD47 -SIRPα, no inhibitor) as -100% activity. Assay activity with tested compounds was then compared and normalized within this range. (A) Compounds that interfered with donor fluorescence indicated by activity in the donor channel. (B) Reactive compound active in a variety of other unrelated screening assays.

Normalized activity signal for the donor and acceptor channels and calculated ratio signal for each active compound reported for the 7-point concentration response generated using qHTS. Normalized activity was calculated with neutral control (+CD47 +SIRPα, no inhibitor) as 0% and low control (+CD47 -SIRPα, no inhibitor) as -100% activity. Assay activity with tested compounds was then compared and normalized within this range. (A) Compounds that interfered with donor fluorescence indicated by activity in the donor channel. (B) Reactive compound active in a variety of other unrelated screening assays (Pubchem).

Normalized activity signal for the specific AlphaScreen assay (Alpha) and the non-specific counter screen assay (counter) for each active compound reported for the 7-point concentration response generated using qHTS. Normalized activity in the AlphaScreen assay was calculated with neutral control (CD47 + SIRPα, no inhibitor) as 0% and negative control (CD47—SIRPα, no inhibitor) as -100% activity. Normalized activity in the counter screen assay was calculated with neutral control (+TruHits reagent, no inhibitor) as 0% and negative control (-TruHits reagent, no inhibitor) as -100% activity. Assay activity with tested compounds was then compared and normalized within this range. (A) Compounds that interfered with the assay as indicated by significant activity in the counter screen assay. (B) Reactive compounds identified as active in a variety of other unrelated screening assays.

The remaining active molecules were further scrutinized for non-ideal behavior that would indicate a mechanism other than CD47-SIRPα antagonism. A now well-recognized mechanism of assay interference involves molecules that tend to aggregate in aqueous assay buffers and adsorb assay reagents to result in apparent inhibitory activity [94–96]. As this phenomenon occurs at a threshold where the test molecule reaches its solubility limit, the concentration response curves will exhibit a steep Hill slope near the solubility limit [94]. Hill slopes are then an indicator of non-specific activity, and compounds were flagged as non-ideal inhibitors when they displayed Hill slopes outside of the -0.5 to -1.5 range, with the understanding that this does not definitively identify problematic molecules as aggregators [97]. To avoid compound aggregation and associated assay interference, all primary assays and follow-up tests were run with detergent present [97,98].

In the case of the lone active molecule in the TR-FRET assays (6-hydroxy-DL-DOPA), its Hill slope falls within tolerance for both the CisBio and LANCE assays. Of the AlphaScreen actives, aurintricarboxylic acid, cephalosporin C, Apresoline, Indirubin-3'-oxime, 2,2'-Bipyridyl, CGS-15943, L-Histidine hydrochloride, and ruthenium red all displayed Hill slopes outside of our acceptance criteria and were discarded. The remaining actives (morin, rotterlin, 1,10-Phenanthroline monohydrate, Quinolinic acid, Fusaric acid, and 6-hydroxy-DL-DOPA) have PAINS characteristics ([91,99,100] and the ChEMBL database) that would make them likely false positives and unsuitable for follow-up. 6-hydroxy-DL-DOPA is also suspected of assay interference due to its aggregator activity (see above) and in other studies [101], PAINS characteristics (as defined in the ChEMBL database), known redox reactivity, and assay promiscuity (active in 119 of 386 assays in Pubchem [102] at the time of manuscript preparation). Despite its well-recognized assay interfering activity, the activity of 6-Hydroxy-DL-DOPA in our assays was consistent with a well-behaved hit compound, and it was also the only consistently active molecule in all qHTS assays. It was thus selected for follow-up testing.

A fresh powder sample of 6-hydroxy-DL-DOPA was acquired from the supplier and authenticated via HPLC-MS from a 10 mM DMSO stock. Follow-up testing in the qHTS assays resulted in similar if not slightly more potent activity (Fig 12 CisBio and AlphaScreen). The compound was then evaluated for SIRPα-CD47 disrupting activity in the previously described SPR assay (vide infra) but failed to inhibit the protein-protein interaction under conditions similar to those of the qHTS assays.

Results shown for the (A) CisBio, (B) LANCE, and (C) AlphaScreen assays. Follow-up testing was conducted using freshly plated compound from powder stocks in an 11-point concentration response (n = 1).

As mentioned above, 6-Hydroxy-DL-DOPA is a known aggregator. This type of interference should lead to activity in the AlphaScreen counter screen assay. However, we found none in our assay, following the manufacturer’s protocol. In this experiment, the control peptide is premixed with both the acceptor and donor beads to form the complex and subsequently added to assay wells followed by the compounds. Under these conditions, the preformed bead-peptide complex may be kinetically and physically insensitive to aggregated compound. By comparison, in the primary assay, reagent proteins are exposed to the aggregated compound before the addition of beads. To unambiguously determine whether 6-Hydroxy-DL-DOPA showed interfering activity in the counter screen, we changed the protocol to more closely match the primary assay. Under these revised conditions (counter screen v.2), the peptide is added first to the wells, followed by the compound, then the beads (Fig 13). The revised conditions revealed activity in the counter screen assay v.2 that was similar to the primary screen assay and identified 6-Hydroxy-DL-DOPA as an interfering molecule. Therefore, it was eliminated from further consideration. In conclusion, no compounds were identified as specific inhibitors for the CD47-SIRPα interaction from the pilot screening study. However, we developed assays and counter assays that excluded many interfering compounds and that can be applied to larger chemically diverse collections and screening campaigns.

Activity shown using the CD47-SIRPα PPI assay (left panel) and the counter screen assay run under different reagent addition sequences (right panel).

Assay performance in fully-automated large library qHTS

We tested the performance of the above assays positioned as part of a large library, fully-automated, qHTS campaign. We used a primary screen with downstream counter and orthogonal screening assays capable of identifying assay artifacts from compounds active in primary screening. For the primary screen, we chose the LANCE TR-FRET assay as it had the best Z’ factor and lowest compound interference profile during pilot screening (Fig 14). As a counter screen we utilized the LANCE TR-FRET assay with a 6XHis-biotin peptide in place of the SIRPα-CD47 interaction (as in the AlphaScreen counter screen assay described above). This allowed for assessment of assay-interfering compounds with the same TR-FRET reagents as the primary screen. Orthogonal screening was used to confirm active compounds and was performed using the AlphaScreen SIRPα-CD47 interaction assay as described above. To test this assay configuration, we screened a large (94,965 compounds), diverse (1,000 scaffolds that vary in representation from 20 to 100 compounds per chemotype), highly curated (PAINS, Lipinsky), lead-like (sp3-enriched, spirocyclic, novel chemotypes) chemical library (Genesis, NCATS) formatted as a 6 and 7-point interplate concentration response in 1536-well source plates. Screening was performed with the NCATS automation group using the fully-automated Kalypsis Robotic platform employing the same liquid dispensing, compound pinning, and endpoint measurement equipment as was used offline in the pilot screening. In this format, primary screening was performed on 409 plates over the course of 4 screening days. Screening day 2 included 200 plates translating to a peak throughput of over 40,000 compounds per day in 6 and 7-point concentration responses. The primary assay was robust (S/B = 17.6 ± 2.2, CV = 2.0% ± 0.6, and a Z′ of 0.93 ± 0.3 Fig 14A) as calculated from the control wells of each plate in the series.

(A) Z’ factor calculated from the intraplate controls. (B) Normalized activity of all compounds tested in a 6-point concentration response. Rendered curves represent all compounds displaying an activity with negative curve class. The curves rendered in red have a maximum activity ≤-25%.

Due to the low level of activity in the screen, compounds were selected for follow-up studies if they exhibited a negative curve classification (all curves in Fig 14B) and a maximum normalized response of <-25% (red curves in Fig 14B). This resulted in 12 compounds selected for follow-up testing. These compounds, selected from chemical stocks, were used to make the library plates, plated in 11-point concentration response format ranging from 38 μM to 37 nM, and confirmed off-line using the TR-FRET primary screen assay (Fig 15). Offline testing confirmed 8 of the 12 actives identified in the primary screen. The 4 compounds that were not confirmed displayed curve class -4 and were active only at the highest concentration in the primary screen. These results provided another example of the value of qHTS format screening and the ability to use concentration-responsiveness as an early activity filter. The remaining active compounds were subjected to additional investigation using the TR-FRET-based counter screen assay to remove assay-interfering false positive compounds. Three compounds displayed unacceptable activity in the counter screen assay (same curve classification or maximum activity as the primary screen) and were removed from follow-up testing. The remaining compounds were then subjected to orthogonal assay testing using the AlphaScreen SIRPα-CD47 interaction assay and 5 were confirmed to have similar activity as in the primary screen. The chemical identities of the confirmed active compounds will be the subject of a follow-up report resulting from their ongoing characterization and optimization.

Left panel: schematic of screening process showing input library size, assay utilization, advanced compounds (green arrow), and non-advanced compounds (red arrow). Right panel: activity profiles for compounds selected for follow-up testing from the primary screen. Activities are shown for each assay performed and the compound profiles are segregated by level of assay advancement. (A) No activity on TR-FRET follow up. (B) Unacceptable counter screen activity. (C) Confirmed active with AlphaScreen.


Senescent cells and macrophages: key players for regeneration?

Over the last decade, our understanding of the physiological role of senescent cells has drastically evolved, from merely indicators of cellular stress and ageing to having a central role in regeneration and repair. Increasingly, studies have identified senescent cells and the senescence-associated secretory phenotype (SASP) as being critical in the regenerative process following injury however, the timing and context at which the senescence programme is activated can lead to distinct outcomes. For example, a transient induction of senescent cells followed by rapid clearance at the early stages following injury promotes repair, while the long-term accumulation of senescent cells impairs tissue function and can lead to organ failure. A key role of the SASP is the recruitment of immune cells to the site of injury and the subsequent elimination of senescent cells. Among these cell types are macrophages, which have well-documented regulatory roles in all stages of regeneration and repair. However, while the role of senescent cells and macrophages in this process is starting to be explored, the specific interactions between these cell types and how these are important in the different stages of injury/reparative response still require further investigation. In this review, we consider the current literature regarding the interaction of these cell types, how their cooperation is important for regeneration and repair, and what questions remain to be answered to advance the field.

1. Introduction

Tissue repair and regeneration are critical biological processes which occur following injury and are essential for survival. Injury can occur as a result of infection, toxic or mechanical assault, and results in a prominent activation of the immune system and the recruitment of a vast number and type of cells, which infiltrate the damaged area. These consist of natural killer cells, macrophages, neutrophils, B cells, T cells, fibroblasts, epithelial cells and endothelial cells. In a healthy environment such cells work together in a concerted effort to restore tissue function and to limit damage, a process which must be tightly regulated. In many pathological environments, these mechanisms become dysregulated and the recruitment of immune cells can instead initiate, amplify and even sustain tissue injury. This process of healing damaged tissue is known as ‘repair’ and encompasses the two separate processes of regeneration and replacement, where regeneration refers to the process in which new tissue growth restores areas of damaged tissue to their original state while replacement occurs in severely damaged tissue, often in the form of scarring.

While the inflammatory cascade serves to eliminate the noxious stimulus and clear the injured area from dead cells and matrix debris, the healing of injured tissues is dependent on the timely suppression of inflammation, setting the stage for the activation of reparative cells [1]. However, the effectiveness of the reparative response is dependent on the severity and type of injury, the organ affected and species-specific characteristics. While amphibians can regenerate limbs [2] and fish can regenerate myocardium [3], adult mammals fail to regenerate either of these. Furthermore, in adult mammals organs such as the liver retain some regenerative capacity [4], whereas the regeneration of the brain and spinal cord is extremely limited [5]. To add to this complexity, when tissues are exposed to prolonged injury the process of repair can become chronic or dysregulated, leading to pathological processes, including fibrosis or chronic inflammation, which ultimately impact organ function and can result in organ or organism death.

1.1. Cellular senescence

A common outcome of the injury process is cellular senescence, an irreversible but stable form of cell cycle arrest, defined by an altered transcriptome, which occurs in proliferating cells when they have reached the end of their replicative lifespan, or when subjected to stress. Senescent cells are often characterized by an enlarged and flattened shape [6], and exhibit hallmarks of senescence, including DNA and chromatin alterations and gene expression changes [7–11] mitochondrial dysfunction and the subsequent release of reactive oxygen species (ROS) [12,13] protein modifications [14] and accumulation of lipofuscin granules [15] expression of SA-β-galactosidase [16] and the release of SASP factors [14,17] (figure 1a). Moreover, while often found in injured tissues [27,28], senescent cells can also be present in uninjured organs, especially in organs that have previously experienced damage or disease [29], and particularly in older individuals. Cellular senescence was first discovered in primary cell culture, where cells grown for long periods of time, akin to ageing, reached a state where they were no longer able to replicate [30,31]. Subsequently, cells positive for senescence-associated (SA)-β-galactosidase were observed in aged tissues [16]. For many years following this, senescence was solely viewed as a result of organismal ageing however, in the last decade, our understanding has dramatically evolved, indicating that cellular senescence can occur in response to a range of stimuli, including cellular damage [18], oxidative stress [32], oncogenic signalling [19], telomere attrition [20], ionizing radiation [21] and some cancer drugs [22] (figure 1b), and is even seen during development [23,24] (figure 1c). Senescence has been reported in numerous cell types during natural ageing, and following injury or disease including epithelia [33], endothelia [34], immune cells [35], mesenchymal cells [36], bone [37], muscle [38] and adipose tissue [39]. An important role of senescence is therefore to prevent the spread of damage throughout the tissue and, in cancer, acts as a potent barrier against tumorigenesis (reviewed in [40]) (figure 1c). In general, the transient induction of senescence followed by senescent cell elimination promotes tissue remodelling and regeneration [25,26] (figure 1c) however, chronic injury can result in the long-term accumulation of senescent cells, driving persistent inflammation which ultimately impairs tissue function and can contribute to organ failure (figure 1c). For this reason, the fine balance of senescent cells and their presence/clearance is likely to play a pivotal role in tissue repair. In this review, we have concentrated on the literature describing the interplay between senescence in epithelial tissues and immune cells, particularly macrophages.

Figure 1. The hallmarks, causes and effects of cellular senescence. (a) The key features of a senescent cell, which include an enlarged and irregular/flattened shape [6], DNA segments with chromatin alterations reinforcing senescence (DNA-SCARS) [9] and foci [10], altered gene expression [8] and cell cycle arrest [7], mitochondrial dysfunction and release of reactive oxygen species (ROS) [12,13], protein modifications [14], lipofuscin granules, expression of SA-β-galactosidase [16] and release of SASP factors [14,17]. (b) Cellular senescence can occur in response to cellular or DNA damage [18], oncogene, mitogen and cytokine signalling [19], telomere attrition/shortening [20], ionizing radiation [21] and anti-cancer drugs [22]. (c) Cellular senescence plays a dual role during development [23,24] and throughout tissue repair and regeneration [25,26], where it can promote the clearance of cell debris, reduces fibrosis, elicits epigenetic alterations and acts as a potent barrier against tumorigenesis, while also leading to senescence spread, DNA damage, further tissue injury, and ultimately leading to age-associated tissue deterioration and pathologies. Created with Biorender.com.

1.2. Senescence-associated secretory phenotype

Senescent cells are often characterized by their ability to develop a senescence-associated secretory phenotype (SASP), a pro-inflammatory response which activates and reinforces the senescence phenotype in surrounding cells, modulates fibrosis and promotes regeneration [41] (figure 1c). The SASP consists of a complex mixture of extracellular matrix proteases, growth factors, chemokines and cytokines, which have a profound effect on the tissue microenvironment [42]. Such SASP components can trigger senescence in neighbouring cells in both an autocrine [43,44] and paracrine [41,45] manner, suggesting that senescence creates an inflammatory microenvironment which may lead to the elimination of senescent cells. The secretion of pro-inflammatory cytokines, such as interleukin-6 (IL-6) and interleukin-8 (IL-8) [46], chemokines, such as monocyte chemoattractant proteins (MCPs) and macrophage inflammatory proteins (MIPs) [42], and growth factors, such as transforming growth factor-β (TGFβ) [47], cause inflammation and recruit immune cells to clear senescent cells.

1.3. Macrophages in tissue repair

Among the variety of cell types which orchestrate repair, macrophages have been shown to exhibit critical regulatory activity at all stages of repair and fibrosis. Macrophages are recruited to the site of injury by chemokine gradients and various adhesion molecules, where they carry out their role as scavenger cells that phagocytose cellular debris and invading cells, alongside other apoptotic cells, in response to tissue injury. Importantly, macrophages are key for the clearance of senescent cells following injury [48,49], as well as an important source of chemokines, matrix metalloproteinases (MMPs) and other inflammatory mediators which drive the initial cellular response [50]. Current models suggest that senescence initiates tissue modelling/remodelling by recruiting immune cells through the SASP, where macrophages clear senescent cells, allowing for repopulation by progenitor cells and regeneration of the damaged tissue [25,26,51]. However, in the case of persistent damage or in aged tissues, clearance and regeneration may be compromised due to poor macrophage recruitment, increased senescent cells or even damage to the macrophages themselves. Indeed, if macrophages are depleted in the early stages of repair in a number of organs, the inflammatory response is diminished [52] and leads to less efficient repair and regeneration [53,54].

This review focuses on the recent findings that have advanced our understanding of senescent cells and macrophages in tissue injury, and the importance of the cooperation of these cells as key players in facilitating tissue regeneration and repair.

2. Evidence for the role of senescent cells in tissue injury

Senescence is a state of irreversible proliferative arrest, which cells undergo in response to a variety of detrimental stimuli, and is associated with changes in morphology, lysosomal activity, alternations in chromatin structure (H2Ax expression) and activation of the SASP [55] (figure 1a). Much of our current understanding of senescence has stemmed from studies of either disease or ageing however, a novel role for senescence in resolving tissue injury has recently emerged. Indeed, senescent cells have been identified in a variety of injured organs, including the liver [28,47], kidney [56,57], heart [58], skeletal muscle [27] and salivary glands [59], and have largely been associated with the loss of tissue function. Nonetheless, the presence of senescent cells has been reported to have both positive and negative effects in their resident organ, depending on their abundance and duration (figure 1c), and provides us with important insights into their physiological function. Indeed, if the function of senescence is the elimination of cells, why do cells not undergo the faster and more direct route of apoptosis, and why were senescent cells selected during evolution? This question has led to the emerging concept that senescent cells play important roles in tissue repair and remodelling, providing a final function before eventually undergoing elimination themselves [26].

2.1. The role of senescent cells in tissue repair

To date, two main approaches have been used to explore the role of senescent cells in repair: genetic depletion strategies, where senescent cells are deleted from the tissue, and through the use of senolytics, where compounds are used to induce senescent cell death.

2.2. Genetic depletion strategies

To distinguish the role of senescent versus apoptotic cells in tissue injury, Baker et al. [66] used an inducible ‘senescence-to-apoptosis' progeric mouse model, where transgenic mice express pro-apoptotic proteins under the expression of the p16 INK4a promoter. By administering the mice with a ‘chemical switch’, cells expressing the senescence-associated marker p16 INK4a were converted into apoptotic cells in vivo. Interestingly, as well as observing a decrease in the number of senescent cells, Baker et al. [66] observed a reversion in a number of age-associated pathologies (figure 2), indicating a role for senescent cells in disrupting tissue homeostasis. However, using a similar but mechanistically different mouse model in cutaneous wound healing studies, it was shown by Demaria et al. [60] that inducing mice to switch from senescence to apoptosis significantly delayed wound healing and caused the wounds to accumulate larger amounts of fibrotic tissue (figure 2). Interestingly, in young mice, a transient burst of P16 INK4a + senescent cells was found to occur during normal wound healing and disappeared following wound closure, indicating an early role for senescent cells in wound recovery and the negative impact of their removal [60]. Similarly, cellular senescence occurs in myofibroblasts in cutaneous wounds during the healing process, which is thought to minimize the extent of fibrosis [61].

Figure 2. Depletion of macrophages or senescent cells has diverse and opposing effects on organ regeneration. The effects of senescent cell depletion (a) or macrophage depletion or prevention of accumulation (b) are depicted. In many organs, the timing of cell depletion has a crucial role on the regenerative outcome. While senescent cell depletion delays cutaneous wound healing and exacerbates fibrosis [60,61], the effect of deletion of macrophages during wound healing is timing-dependent [54] similarly while senescent cell depletion leads to liver fibrosis [28,62], depletion of macrophages can lead to both reduced liver scarring and fibrosis, dependent on timing [52]. Moreover, senescent cell removal has no or largely positive effects on muscle and heart [63] however, macrophage depletion leads to detrimental effects on heart and muscle regeneration [63–65]. Created with Biorender.com.

In support of these observations, contrasting effects of senescent cell function have also been demonstrated in other models of tissue injury. When the liver is injured, hepatic stellate cells become senescent and produce a stable fibrotic scar [28] (figure 2). In vivo, these senescent cells are identified inside the fibrotic lesions however, mice deficient for p53 and p16 INK4A show increased fibrosis in both the liver and kidney [28,62]. Conversely, in a mouse model of oncogenic NRAS G12 V , where senescent cells are usually cleared by monocytes and macrophages, immunodeficient mice show reduced clearance, resulting in mature liver tumours [67]. Moreover, the p16-3MR transgenic mouse, a model that contains a p16 INK4a promoter which allows the tracing and removal of senescent cells has been used to demonstrate that senescent cell deletion reduces pain in an experimental model of osteoarthritis [68]. Crucially, a mouse model that contains the transgene, INK-ATTAC, which induces apoptosis in p16 INK4a -expressing cells, established that senescent cell clearance treatment extended lifespan in both male and female mice, delayed tumorigenesis and attenuated age-related deterioration of several organs, including kidney, heart and fat, without apparent side effects [69].

2.3. Senolytics

The use of senolytic compounds also provides a mechanism to elucidate the role of senescent cells and in particular, the specific timing of depletion. Evidence shows that senolytics can drive the expression of SA-β-gal in cell culture [70]. Moreover, the administration of the senolytic compounds ABT-737 or Dasatinib plus Quercetin (DQ) in vivo induces apoptosis in senescent cells and leads to clearance in mouse skin, lung and the haematopoietic system, and subsequently improves tissue repair [70–73]. Moreover, DQ administration promotes the survival of transplants from aged mice [74].

Taken together, these studies highlight the opposing roles of senescent cells in injury and repair, and the variation in their function as a result of timing, degree and type of injury. Indeed, increasing evidence suggests that cellular senescence is a multi-step, dynamic process, progressing from a transient to a stable state of cell cycle arrest, dictating the outcome [75].

2.4. The role of the senescence-associated secretory phenotype in tissue repair

Furthermore, beyond the direct effect on cell division and clearance, a key mechanism in which senescent cells influence injury is through the SASP. The SASP secretome includes a wide variety of soluble signals capable of influencing tissue inflammation, repair and fibrosis, including IL-1, IL-8, IL-6 and transforming growth factor beta (TGFβ). The SASP is a pro-inflammatory response which activates and reinforces the senescent phenotype in surrounding cells and therefore mediates the spread of senescence throughout the tissue, known as ‘bystander senescence’. Through a variable set of cytokines and chemokines, paracrine senescence is induced and maintained through mechanisms which generate reactive oxygen species (ROS) and the DNA-damage response (DDR) [76]. For example, paracrine senescence has been shown to exacerbate biliary injury and impair regeneration in the liver [47]. Upon the partial ablation of the SASP, through TGFβ inhibition, hepatocyte proliferation is increased, fibrosis is decreased, and overall liver function is improved. Opposingly, in the mouse model of cutaneous skin injury employed by Demaria et al., the authors identified a positive effect of the SASP [60]. Intriguingly, through the secretion of the SASP factor PDGF-AA, the closure of skin wounds was found to be accelerated, by promoting myofibroblast differentiation and granulation tissue formation.

Indeed, when the SASP was first studied, its role in injury and resolution was not well understood, as the components appeared to be mostly pro-inflammatory. It has since been suggested that senescence may impair regeneration through paracrine signalling, leaving neighbouring cells unable to compensate for damage, resulting in enhanced fibrosis and a diminished capacity of the regenerative response [47]. However, it appears that the prominent inflammatory components and detrimental consequences occur largely when the SASP is long-lived and may be beneficial when transient, thus playing a critical role in tissue remodelling at the early stages of injury. In fact, in 2017, Chiche et al. [29] showed that in acute and chronic injury, the release of the SASP factor IL-6 from senescent cells enabled the reprogramming of muscle satellite cells, indicating a role for the SASP in facilitating cellular plasticity and repair. Importantly, this points to a beneficial role of the SASP in promoting the cellular plasticity of stem cell populations during acute muscle injury, as well as in the pathological setting of muscle deterioration [29].

2.5. Cellular senescence in pluripotency

To add to the aforementioned findings, a recent study has demonstrated that the reprogramming factors OCT4, SOX2, KLF4 and cMYC can induce cellular senescence and IL-6 production in vivo, which leads to more efficient reprogramming [77]. Moreover, SASP can promote a pro-regenerative response through the induction of plasticity and stemness in somatic stem/progenitor cells [78]. Ritschka et al. [78] showed that a transient exposure to the SASP in primary mouse keratinocytes caused the increased expression of stem cell markers and regenerative capacity in vivo. However, prolonged exposure to the SASP triggered senescence arrest which countered the regenerative stimuli [78]. It has thus been proposed that in the case of injury, the senescent cell uses the SASP to induce plasticity and stemness in neighbouring cells, enabling the replacement of the senescent cell once it has been cleared and encouraging tissue regeneration [25]. However, it is important to note, when this process is uncontrolled, senescence-associated reprogramming can lead to tumour formation by promoting cancer stemness [79]. By contrast, SASP factors which are secreted by senescent cardiac progenitor cells (CPCs) via paracrine signalling result in the senescence of otherwise healthy CPCs. In this context, the elimination of senescent cells in aged mice or in mice treated with senolytics abrogated the SASP and resulted in the activation of resident CPCs and an increase in the number of proliferating Ki-67 EdU+ cardiomyocytes, thus indicating that the removal of senescent cells may alleviate deterioration following cardiac injury and contribute to the capacity of the heart to regenerate [58]. Moreover, the deletion of the senescence effectors p53 and p16 INK4a improves the reprogramming efficiency of human fibroblasts to iPSCs, suggesting, that in this environment, senescence has a negative effect on plasticity [80]. Importantly, these differences in whether senescent cells are beneficial/detrimental to regeneration highlight the importance of the timing in which the senescence program is activated and its changing role during different stages of the injury response.

To date the SASP is known to have important roles in embryonic development, wound healing and tumour growth, indicating that the SASP has more complex physiological roles than we currently understand (reviewed in [81,82]). Taken together these studies demonstrate that senescent cells play important roles in tissue injury and regeneration and can both promote and inhibit tissue repair. Simply, the evidence of the positive effects of senescent cell removal comes from the circumstances where senescent cells accumulate and lead to negative consequences. Conversely, a transient wave of senescent cells appears to play an important role in promoting repair in the early stages of injury. Overall, these findings support the understanding that the senescence programme can be a beneficial regenerative process however, when it is perturbed, it can play a detrimental role. For example, while acute senescence clearly plays an important role in preventing malignancy and promoting successful tissue repair, the accumulation of chronically senescent cells further contributes to injury, disease and ageing.

3. Role of macrophages in tissue injury

While a large variety of cell types have been shown to play important roles in injury and repair, in recent years, a particular interest in macrophages has developed, due to the contribution of different macrophage populations and their plasticity in the context of injury. Thus, in recent years, there has been a particular focus on the identification of different macrophage states and subsets in many different organ systems and their different roles in injury and repair. Macrophages are crucial for limb regeneration in the salamander [83] and tail fin regeneration in the zebrafish [84] (figure 3). Moreover, macrophages interact intimately with their surroundings integrating cues from invading pathogens, commensal bacteria, as well as tissue-specific functions, rendering macrophages extremely well adapted to their local environment and thus acquiring organ-specific functions [85,86]. The macrophage populations which are found in the many different tissues of the body are termed ‘tissue-resident’ macrophages and are largely derived from the yolk-sac during embryogenesis (reviewed in [87]). As long-lived cells, tissue-resident macrophages are particularly important, due to their witnessing and ‘memorizing’ past and present events in the tissue, which plays an important role in their plasticity. Furthermore, it has been shown that these macrophages are capable of readjusting their behaviours, probably through epigenetic modifications [88,89]. Therefore, tissue-resident macrophages are critical in maintaining tissue homeostasis.

Figure 3. Animal models of regeneration have provided evidence of the interplay between macrophages and senescent cells during tissue regeneration. The induction of cellular senescence leads to the initiation of the senescence-associated secretory phenotype (SASP) during salamander limb regeneration, zebrafish fin regeneration and post-partum uterus regeneration. In the absence of macrophages, senescent cells in the regenerating salamander limb are not cleared, which is a possible reason for impaired regeneration (dashed grey arrow) [48]. The removal of either senescent cells or macrophages during zebrafish regeneration has a deleterious effect on regeneration, hypothesized to be as a result of altering the tightly regulated balance of cell senescence (dashed grey arrows) [125]. The mammalian uterus undergoes extensive remodelling post-partum where senescent cells are normally cleared by macrophages. In the absence of macrophages, senescent cells accumulate in the uterus [126], presumably leading to dysregulated regeneration and function (dashed grey arrow). Created with Biorender.com.

In response to tissue injury, inflammation results in an initial influx of neutrophils, accompanied by monocyte-derived macrophages, which clear cellular debris and coordinate cellular processes to initiate tissue repair. Importantly, this process leads to an overall diluting of the tissue-resident macrophages in the macrophage pool, which is further exacerbated by the proliferation of infiltrating macrophages, which adapt their function to surrounding cues in the local microenvironment. A range of studies have thus identified the specialized roles of monocytes and macrophages and the timing of their activation as critical in the various steps of tissue repair, regeneration and remodelling [52,54].

3.1. Macrophage phenotypes

In the past, macrophages have been broadly separated into two categories: M1 (classically activated) and M2 (alternatively activated), based on their inflammatory and anti-inflammatory/reparative functions, respectively. Nowadays, the binary M1/M2 classification is generally considered an oversimplification of the large variety of macrophage populations that exist in vivo, and to date have been further subdivided, based on their gene expression profiles [90,91]. Nonetheless, pro-inflammatory macrophages are generally associated with the expression of high levels of pro-inflammatory cytokines, an ability to mediate resistance to pathogens, produce reactive nitrogen and oxygen intermediates, and promote Th1 responses [92]. On the other hand, reparative macrophages are characterized by their role in tissue remodelling and repair, regulation of the immune system, scavenging and phagocytic capabilities [93], thus, exerting mainly pro-tumoral and immunoregulatory functions (reviewed in [94]).

Unsurprisingly, the presence of pro-inflammatory macrophages has been shown to sustain tissue-damaging inflammatory responses, and the presence of these cells has been associated with a variety of inflammatory and fibrotic diseases. The role of pro-inflammatory macrophages has been particularly well-studied in models of spinal cord injury, where macrophages have been shown to readily accumulate at the site of injury. In these models, macrophage activation and polarization, depending on changes in the microenvironment, has shown that the sustained recruitment of pro-inflammatory macrophages facilitates axonal dieback and can substantially delay the regenerative response [95], and their death in situ further contributes to tissue damage [96]. Furthermore, the presence of axon growth inhibitors is significantly higher in pro-versus anti-inflammatory macrophages, suggesting that these cells can actively contribute to suppressing regeneration after spinal cord injury [97]. In addition, studies in the liver have also implicated inflammatory macrophages in exacerbating injury, where an increase in inflammatory macrophages is observed in areas of hepatic necrosis [98,99]. This has also been observed during acute kidney injury where inhibition of early, pro-inflammatory macrophages improves renal function [100] (figure 2). However, it is important to note that pro-inflammatory macrophages may also contribute to the processes which lead to recovery. This has been observed in models of skeletal muscle injury, where inhibition of monocyte/macrophage accumulation impairs muscle regeneration [63,64], and cardiac regeneration, where macrophage depletion leads to alterations in myofibroblast infiltration and neovascularization, and subsequent ventricular dilatation and mortality [65] (figure 2). Thus, a fine balance between pro/anti-inflammatory macrophages is probably needed for optimal repair following injury.

3.2. Macrophage depletion and reconstitution studies

Perhaps unsurprisingly, the complete depletion of macrophages has also been found detrimental for tissue repair. Depletion studies in models of tissue injury in the liver have shown that macrophage-depleted mice fail to exert a complete cytokine response, which subsequently compromised liver regeneration [52]. This is likely to be due to the loss of anti-inflammatory pro-regenerative macrophages which play critical roles in promoting tissue repair. These macrophage populations have been partially defined by their production of the anti-inflammatory cytokine IL-10, which functions as an important anti-inflammatory mediator essential for the maintenance of anti-inflammatory activity [101]. Interestingly, in a model of early-onset inflammatory bowel disease, the loss of the IL-10 receptor (IL-10R) resulted in the spontaneous development of colitis [102], indicating that IL-10R signalling in intestinal macrophages is an important factor for controlling intestinal inflammation. Moreover, while cutaneous wound healing is accelerated in mice deficient for IL-10, a result attributed to accelerated re-epithelialization and wound contraction, macrophage infiltration was significantly elevated [103], further implicating IL-10 signalling in inflammation.

Furthermore, a recent paper by Podaru et al. [104] showed that the transplantation of functional ‘reparative macrophages', acquired from bone marrow mononuclear cells, in a mouse myocardial infarction model, resulted in significant improvement in functional recovery. Here, the authors found that transplantation of the reparative macrophages enhanced myocardial tissue repair, by promoting the formation of the vasculature and reducing cardiomyocyte hypertrophy and interstitial fibrosis. Interestingly, the transplantation of such reparative macrophages was also found to increase the number of their host-derived counterparts, which was partly mediated by TGFβ secretion [104]. Moreover, following hepatocyte death during liver injury, the engulfment of debris by macrophages leads to the induction of Wnt3a, which subsequently leads to canonical WNT signalling in nearby hepatic progenitor cells, facilitating their differentiation into hepatocytes and therefore contributing towards the regenerative response [105]. Thus, inflammatory cell-mediated cytokine signalling plays an integral role in regeneration and tissue resolution.

It has recently been shown that small extracellular vesicles (sEVs) derived from M2 bone marrow-derived monocytes (BMDMs) can attenuate spinal cord injury (SCI). sEVs mediate paracrine signalling and are important for regulating cellular function [106]. Here, sEVs from M2 BMDMs were found to protect neurons in SCI mice, by inhibiting the mTOR pathway and enhancing the autophagy ability of neurons, therefore reducing apoptosis in vitro and in vivo. This was found to be due to the transfer of the microRNA miR-421-3p, which regulates the mTOR pathway inside the M2 BMDM-sEVs [107]. This study showed for the first time that M2-derived BMDMs are important in protecting neurons and facilitating recovery following SCI, as well as highlighting that this occurs via the transfer of sEVs. Furthermore, this study thus further elucidates the beneficial roles for M2 macrophages during injury and indicates a mechanism by which they carry out their protective/reparative function.

Interestingly, such anti-inflammatory macrophages have been shown to not only promote tissue repair but also antagonize the function of pro-inflammatory macrophages and fight against their pro-fibrotic capabilities [108,109]. Thus, the interaction/cooperation of macrophages with various cell types, including those involved in the initial phase of inflammation, is of vital importance. It has recently been shown that neutrophils also have a crucial function in liver repair, by promoting the phenotypic conversion of pro-inflammatory Ly6C hi CX3CR1 lo monocytes/macrophages to pro-reparative Lyc6 lo CX3CR1 hi macrophages. Furthermore, this conversion was found to be dependent on the expression of reactive oxygen species (ROS) from neutrophils [110,111]. Intriguingly, this study demonstrated the cooperation between neutrophils and macrophages and the importance of their interaction in the resolution of inflammation and tissue repair [111]. This supports the accumulating evidence that monocyte-derived macrophages can undergo both phenotypic and functional transition in order to promote tissue regeneration and healing [112,113].

Finally, macrophage activity during different phases of tissue injury is important for tissue repair. By selectively depleting macrophage populations Duffield et al. [52] showed that macrophages have distinct, opposing roles during injury and repair. Specifically, in a mouse model of Ccl4-induced reversible liver injury, the depletion of macrophages during advanced fibrosis resulted in reduced scarring. However, if macrophages were depleted during the repair period this resulted in the failure of matrix degradation and a persistent activation of the fibrotic response. Importantly, this showed that macrophages perform both injury-inducing and repair-promoting tasks (figure 2), and that functionally distinct subpopulations of macrophages exist within the same tissue that play important roles in different phases of injury/recovery [52]. Moreover, the depletion of macrophages in the early stages of cutaneous wound repair delayed re-epithelialization, leading to reduced scar formation, while depletion in the mid-phase of new tissue formation led to an impaired wound closure. Crucially, depletion in the late stages of repair had no effect on the overall repair response, suggesting that macrophages play different and distinct functions during the phases of skin repair [54,114] (figure 2). Ultimately, studies such as these demonstrate that macrophages exert different and distinct functions at different stages of the repair or regeneration processes a discovery that is crucial if we are to therapeutically manipulate macrophage function in the future to improve organ regeneration.

4. The interaction between senescent cells and macrophages in tissue regeneration and repair

Senescence is now known to play important roles in many different tissues during murine embryonic development [23,24]. This includes in the mesonephric tubules during mesonephros involution (development of the kidneys and testes), the endolymphatic sac of the inner ear, the apical ectodermal ridge of the limbs, the regressing interdigital webs of the hands and feet, and the closing of the neural tube [23]. Indeed, there is evidence that senescent cells in murine development are surrounded by macrophages at days E13.5–14.5, and that the infiltration of macrophages leads to senescent cell clearance and the promotion of tissue remodelling [23,97]. Importantly, the processes that are observed during development provide us with a unique insight into the mechanisms which drive regeneration and repair following injury in adulthood, and act as a starting point for the manipulation of such processes in vivo.

4.1. Senescent cell surveillance by macrophages

The role of macrophages in clearing senescent cells has been known for more than a decade [115]. Due to the substantial release of cytokines and chemokines by senescent cells via the SASP, it is unsurprising that an increasing number of studies appear to support a macrophage-dependent surveillance mechanism which operates in both normal and regenerating tissues. Indeed, it seems logical that the number of senescent cells must be closely monitored to maintain tissue homeostasis and to mitigate and/or prevent the negative impacts of senescent cell accumulation. The first evidence of the involvement of the immune system in the surveillance of senescent cells came in 2007 from Xue et al. [116], who revealed that the reactivation of p53 in p53-deficient tumours led to complete tumour regression. This was found to be mediated by the upregulation of inflammatory cytokines and the activation of the innate immune response [116]. After this, the role of immune cells, including macrophages, were shown to be important in the removal of senescent cells in models of liver injury, as well as for preventing excessive detrimental fibrosis and in resolving liver fibrosis [117]. Moreover, senescent hepatic stellate cells have been shown to secrete a SASP that attracts macrophages [118]. In 2013, Lujambio et al. showed that p53-expressing stellate cells release IFN-γ and IL-6, which promote resident Kupffer macrophages and infiltrated macrophages polarize towards a tumour-inhibiting M1 state, capable of targeting senescent cells in culture. However, in senescent stellate cells lacking p53, IL-4 was produced, causing macrophage polarization towards the pro-survival M2 phenotype [119]. Thus, this evidence suggests that senescent cells can elicit phenotypic changes in macrophages which can affect their functionality. Interestingly, mesenchymal stem cells (MSCs) also appear to play a regulatory role in shifting local macrophages from a pro-inflammatory to a tissue reparative phenotype [120,121], and in the absence of macrophages, neonates lose their ability to regenerate their myocardia following myocardial infarction [122].

To add to this, senescent cells are also capable of eliciting an adaptive immune response where they are cleared by CD4+ T cells and monocytes/macrophages. This has been shown in the liver, where pre-malignant senescent hepatocytes undergo clearance by CD4+ T cells, which require the presence of monocytes/macrophages, highlighting the importance of immune cell cross-talk in senescent cell clearance [123]. Furthermore, in a model of liver cancer, senescent cell surveillance was shown to require the recruitment and maturation of CCR2+ myeloid cells, while the ablation of CCR2 caused outgrowth of hepatocellular carcinomas [124].

4.2. Senescent cells and macrophage interactions during regeneration

To explore these interactions further, in 2015, Yun et al. [48] showed that macrophages were critical for the clearance of senescent cells, providing the first evidence that senescence-surveillance mechanisms operate during normal regeneration. Here the authors demonstrated that during salamander limb regeneration, there was a significant induction of cellular senescence, indicating its importance as a mechanism for regeneration in normal regenerating tissues. Furthermore, Yun et al. also reported that a strong SASP signature in blastemas coincided with a strong peak in the induction of senescent cells (figure 3), indicating that these could have paracrine effects on regeneration and the chemo-attraction of macrophages [48]. Crucially, macrophages and senescent cells are found in close proximity to each other in regenerating limbs. By contrast, clodronate-mediated deletion of macrophages results in the persistence of senescent cells during limb regeneration [48]. In support of this, reports in other model organisms such as the zebrafish have demonstrated the negative impacts of macrophage ablation during zebrafish fin regeneration [84]. This has been further investigated in a 2020 study by Da Silva-Álvarez et al., who showed that following injury in the zebrafish, senescent cells were present at the site of injury, and that their removal impaired regeneration [125] (figure 3). Moreover, senescent cells that accumulate in the post-partum mammalian uterus are efficiently cleared by macrophages after birth, while macrophage depletion leads to abnormal accumulation of senescent cells [126] (figure 2).

There therefore exists convincing evidence for the role of senescent cells and their interaction with macrophages in tissue injury and repair. SASP promotes macrophage proliferation [127], while P16INK4a+ macrophages accumulate with increasing age and exacerbate SASP [128]. The transcription factor GATA4 is stabilized during cellular senescence, which in turn activates NFκB to facilitate SASP [129] (figure 4). However, to date, questions remain regarding the mechanisms by which senescent cells interact with the immune system, including macrophages, and how this elicits or prevents a reparative response. It has previously been shown that senescent cells express NKG2D ligands MICA and ULBP2 on their cell surface, allowing their recognition and elimination by natural killer (NK) cells. Furthermore, the expression of these ligands has been found to be regulated by the DNA damage response and the ERK signalling pathway as a result of injury [131] (figure 4).

Figure 4. Senescent cells and macrophages communicate via complex bi-directional signals during tissue remodelling, repair, regeneration and tumorigenesis. Senescent cells are detected through immune surveillance by natural killer (NK) cells, mediated by cell–cell contact, protein transfer, actin polymerization and cytoplasmic bridges, which ultimately aids recognition by NK cells and T cells [130]. Senescent cells express the ligands MICA and ULBP2 on their cell surface in response to DNA damage and ERK signalling, which aids in their clearance by NK cells [131]. In addition, senescent cells often express MHCII and vimentin which allows recognition by T cells and macrophages, respectively [123,132,133]. Conversely, senescent cells can evade detection by expressing the ‘don't eat me’ signal CD47, which impairs phagocytosis by macrophages by binding to the inhibitory receptor SIRPα [111] and GATA4 is stabilized during cellular senescence, which in turn activates NFκB to facilitate SASP [129]. Macrophages exist as tissue-resident cells or are derived from blood-borne monocytes and can be polarized towards a pro-inflammatory or pro-reparative state. Expression of p16 Ink4a and SaβGal in macrophages has been implicated in a role for polarization, rather than as a consequence of paracrine senescence [128,134,135]. CCR2 expressed by macrophages is critical for senescent cell surveillance and clearance by T cells and infiltrating macrophages [124]. Under normal physiological conditions, p53+ stellate cells release the pro-inflammatory cytokines, IFN-γ and IL-6, which promote resident Kupffer cells and infiltrating macrophages to polarize towards a pro-inflammatory state, which can target senescent cells. Conversely, in senescent stellate cells lacking p53, IL-4 is produced, resulting in macrophage polarization towards the pro-survival phenotype [119]. Senescence-associated secretory phenotype (SASP) leads to elevated inflammatory cytokines, activation of the immune response and senescent cell clearance. In addition to this classical role, SASP also leads to extracellular remodelling (ECM) which contributes to altered cell recruitment and impaired access of immune cells to senescent cells [136]. The MHC molecule HLA-E, which is expressed by senescent cells and induced by IL-6, interacts with the inhibitory receptor NKG2A, expressed by NK cells and T cells, to evade surveillance, and blocking this interaction improves the immune responses against senescent cells [137], in a manner whereby sustained inflammation also contributes to the persistence of senescent cells. Created with Biorender.com.

4.3. Senescent cells and macrophage interactions with ageing

Another outstanding question is what allows senescent cells to accumulate during ageing/injury and how senescent cells manage to escape being cleared by the immune system. Progress in answering this question has shown that senescent dermal fibroblasts can evade clearance by the immune system by expressing the non-classical MHC molecule HLA-E. HLA-E functions by interacting with inhibitory NKG2 receptors, expressed by NK cells and CD8+ T cells, to inhibit immune responses against senescent cells (figure 4). The authors found that blocking the interaction between HLA-E and the receptor NKG2A boosted immune responses against senescent cells in vitro. Interestingly, they found that the SASP-related cytokine IL-6 induced the expression of HLA-E in non-senescent cells in a paracrine fashion [137]. The upregulation of HLA-E expression by IL-6 thus suggests that sustained inflammation may also contribute to the persistence of senescent cells in tissue, further contributing to the pathogenesis of injury or age-related diseases. This is supported by the fact that IL-6 has been well characterized in senescence [138] and is found in the serum of elderly patients, as well as in various injury models [139,140]. Furthermore, as the authors note, the expansion of CD8+ T cells which are NKG2C+, another isoform of the inhibitory NKG2 receptors, with age may offer an explanation as to why the immune system is less effective in clearing senescent cells in older individuals. Determining whether the abundance/role of CD8+ NKG2C+ T cells is altered during acute and/or chronic injury would therefore be of interest.

Another important component for tissue regeneration is the extracellular matrix (ECM) and its cross-talk with cells in the surrounding microenvironment. Interestingly, the soluble SASP is known to induce ECM remodelling and stiffening, which has previously been shown to alter immune cell recruitment during ageing [136]. In fact, alterations to the ECM as a result of changes in matrix stiffness impair the access of immune cells to senescence-enriched tissues [136]. Furthermore, stromal cells such as fibroblasts are responsible for regulating tissue structure through deposition of the ECM, as well as supporting homeostasis through the secretion of cytokines, chemokines, growth factors and other key signalling proteins [136] (figure 4). We previously discussed the importance of the SASP in the release of soluble factors which influence the surrounding tissue microenvironment and maintain tissue homeostasis. It has been reported that extensive cross-talk exists between senescence-associated stromal populations that are known to accumulate during ageing and an immunosuppressive phenotype.

4.4. The parallels between macrophages and senescent cells

As well as the well-documented phagocytic role of macrophages, it was recently shown, for the first time, that chemotherapy-induced senescent cells (CISCs) are capable of engulfing both neighbouring senescent cells or non-senescent tumour cells in a macrophage-like fashion. It is important to note is that this behaviour was noted after chemotherapy treatment. However, this behaviour is also triggered by the administration of nutlin, which activates p53 without causing genotoxic stress [141]. This fascinating discovery that senescent cells, in the correct environment, can acquire a phagocytic phenotype, is suggestive of a potential survival advantage.

This newly found ability of senescent cells brings to light parallels with macrophages, whereby both macrophages and senescent cells secrete factors that elicit matrix remodelling and immunomodulation [142,143], and express metabolic markers, such as CD38 [144]. Indeed, cannibalism by breast cancer cells has been suggested to play a role in the induction of senescence [141] thus, it will be important to investigate the role of cannibalism of senescent cells in a range of different contexts, including development, ageing and regeneration/repair. In particular, it will be interesting to determine the role of such cells in senescent cell accumulation, and whether cannibalism may play a role in cell clearance following injury. Furthermore, it will be pivotal to determine what cell types these cannibalistic cells preferentially engulf, and whether there are certain characteristics which lead to their engulfment, in order to determine whether, if such cells exist outside the context of cancer, they are beneficial or detrimental to the process of regeneration and repair.

4.5. The ageing immune system and immunosenescence

As the body ages, the innate immune system gradually declines, a phenomenon now termed immunosenescence [145], resulting in decreased effector immune cell function and healthy tissue renewal rate decreases dramatically [146]. As a result, the aged tissue microenvironment accumulates senescent cells, such as SASP-associated fibroblasts, and gains the infiltration of immune infiltrates such as immunosuppressive myeloid-derived suppressor cells (MDSCs) and T regulatory cells. MDSCs are a heterogeneous population of cells of myeloid origin which repress T cells via the secretion of arginase 1, TGFβ and reactive oxygen species (ROS), while T regulatory cells have a role in regulating or suppressing other cells in the immune system (reviewed in [147]). It has been suggested that this alteration in tissue microenvironment may encourage the development of pathological conditions such as cancer, and allows for the expansion of cancer cells unabated by the immune system. Indeed, in the event of chronic injury, it is certainly possible that this leniency of the immune system, or impaired immune surveillance, may also have similar effects, for example, by allowing the accumulation of senescent cells. In addition, the immunosenescence of effector T cells, NK cells, macrophages and dendritic cells is known to lead to a dramatic decrease in their cytotoxic activities and infiltration within an aged tumour-promoting microenvironment [148]. This has been supported by the observation that there are systemic increases in immunosuppressive M2 macrophages and N2 neutrophils in elderly people (reviewed in [149]), which may contribute towards an immunosuppressive phenotype. This has been shown in a model of senescence induction, using the fibroblasts accelerate stromal-supported tumorigenesis (FASST) mouse, where the authors observed an accumulation of senescent cells, as well as an increased number of MDSCs and regulatory T cells, which was not observed in a younger tissue microenvironment. Furthermore, this increase in the number of MDSCs and T reg cells in healthy mice was found adjacent to senescent cell populations and was found to primarily be due to the secretion of IL-6 [148]. Moreover, the deletion of perforin (Prf1), an essential component of immune surveillance of senescent cells, leads to the accumulation of senescent cells in the liver [150] and a general increase in the accumulation with advanced age, resulting in chronic inflammation, fibrosis and tissue damage [117]. Thus, this process of transient SASP and senescent cell clearance in young versus aged mice during the natural process of ageing shows very clear similarities with the occurrence of the senescence programme and the cycle of clearance/accumulation of senescent cells in acute and chronic injury, respectively.

A well-characterized mechanism by which senescent and apoptotic cells are regulated by macrophages is via their ‘eat me’ signals. However, senescent cells often express the ‘don't eat me’ signal CD47, a cell surface protein mediated by upregulation of NFκB, which impairs phagocytosis by macrophages by binding to the inhibitory receptor signal regulatory protein alpha (SIRPα) found on the cell surface of macrophages. Conversely, senescent cells also appear to upregulate their expression of MHC class II molecules on their cell surface, which aid in their recognition by CD4+ T cells [123,132]. Furthermore, it has recently been identified that senescent cells also express an oxidized form of vimentin on their cell surface, which eventually gets released into the bloodstream, suggesting that oxidized proteins are capable of being recognized by the humoral innate immune system. This suggests that such proteins form part of a senescence-surveillance mechanism by acting as ‘eat me’ signals, and this process may become impaired with age [133]. Modified vimentin is already known to be expressed on the surface of immune cells including apoptotic neutrophils [151–153] and T cells [154], which eventually leads to elimination by macrophages [155]. In addition, it has been found that radiation-induced proteins found on the surface of apoptotic cells can interact with vimentin found on the surface of neighbouring phagocytes, leading to their removal by macrophages [155]. Understanding the mechanisms which regulate macrophage phagocytic ability and subsequent senescent cell clearance will be vitally important for developing future therapeutic approaches to aid regeneration and improve healthspan.

Senescent cells are also capable of communicating with neighbouring cells through the direct transfer of proteins. In 2015, it was shown by Biran et al. [130] that proteins were transferred to NK and T cells, and that this process was facilitated by immune surveillance of senescent cells by NK cells, resulting in their activation and increased cytotoxicity. The protein transfer was dependent strictly on cell–cell contact and CDC42-regulated actin polymerization, and partially mediated by cytoplasmic bridges. This raises the possibility that senescent cells may use intracellular protein transfer to induce senescence in neighbouring cells or perhaps even to promote survival. Furthermore, it is possible that this process may be important for regeneration, by allowing the cell–cell transfer of molecular mediators between senescent and regenerative or supportive cells. Indeed, the authors found that compared with normal cells, senescent cells preferentially use intercellular protein transfer (IPT) to regulate their own self-elimination by immune cells and communicate with epithelial cells. To support this, a variety of other studies have shown that IPT is used to both initiate and modulate immune responses which ultimately support cell survival [156–158]. Further investigation is required to comprehensively determine the transfer of material between macrophages and other immune cells during senescence and regeneration/repair. The recent publication of a proteomic database of soluble proteins and exosomal cargo SASP factors will undoubtedly lead to a more thorough understanding of cell communication and senescence [14].

While more is becoming known regarding the role of senescent cells and macrophages or the immune system in regeneration and repair, there remain many open questions. For example, it is still unclear whether macrophages themselves become senescent in the context of injury and whether this may contribute to the accumulation of senescent cells. Indeed, while damage to the immune system leads to the failure to clear senescent cells, contributing to their persistence in tissues, such immune failure may be in part due to immunosenescence, leading to immune cell dysfunction. Hall et al. [128] previously reported the observation of p16 INK4a and Sa-β-Gal expression in macrophages suggesting that senescent cells could spread senescence to immune cells. However, the authors later reported that the expression of these markers was not due to a spread of senescence, but rather acquired as part of a physiological response to immune stimuli [134]. This is consistent with reports that p16 INK4a is involved in macrophage polarization [135]. Thus, whether immune cells become senescent during injury or disease requires further investigation however, it has been reported that immune cells undergo a progressive decline in function, known as immunosenescence, which has been reported to contribute to senescent cell accumulation [159,160]. Furthermore, if macrophages become senescent, do they still retain their ability to clear target cells, including other senescent cells? If this is the case, it is likely that macrophages contribute to the process of ‘inflammaging’, including immunosenescence, especially if senescent macrophages are found to promote a pro-inflammatory environment. Furthermore, understanding other mechanisms by which macrophages become damaged which cause changes in their function and or/phenotype during injury will also be important when considering therapeutic approaches.

4.6. Senescent cell accumulation and pathology

Other outstanding questions include whether there is a ‘tipping point’ at which senescent cells must accumulate to before the tissue becomes dysfunctional, and whether a similar process occurs at the point where macrophages are unable to clear senescent cells. As previously mentioned, the specific mechanisms by which macrophages interact with senescent cells remains to be elucidated. Understanding the molecular mechanisms which govern these interactions and which regulate macrophage function will be critical for the progression of therapies and the understanding of chronic diseases. In line with the interaction of senescent cells with CD4+ T cells to evade immune surveillance, it would be beneficial to further characterize whether similar mechanisms exist between macrophages and senescent cells. Finally, while various studies have individually indicated the role of different components of the senescence programme in the regenerative response, a comprehensive study linking together these processes remains to be undertaken. While the transient SASP is beneficial for regeneration and sustained SASP signalling promotes senescent cell accumulation, a comprehensive study of the ‘waves’ of the SASP in repair/regeneration, the point at which this process becomes detrimental, and the exact mechanisms which govern this response and how they interact with each other is still lacking.

5. Therapeutic manipulation of macrophages for tissue regeneration/repair

As covered in previous sections, the role of macrophages in maintaining tissue homeostasis, clearing senescent cells and promoting regeneration is beginning to be better established. Due to the innate role of macrophages in these processes, an emerging idea of treating diseases/age-related pathologies related to senescent cell accumulation falls within the realm of manipulating macrophage function. Such ideas include treating Alzheimer's disease [161], myocardial infarction [162], cancer [163] and inflammatory diseases [164] such as rheumatoid arthritis, with macrophage therapy.

5.1. Macrophage engineering approaches

One method of achieving a therapeutic approach is by engineering macrophages to ignore ‘don't eat me’ signals found on the surface of senescent cells. As previously mentioned, the CD47–SIRPα axis allows senescent cells to avoid removal by the immune system therefore, blocking this axis may be an effective treatment against the accumulation of senescent cells in ageing and injury. Molecules which target this axis have already been developed and include those that target CD47, as well as SIRPα specifically, alongside bispecific targeting agents. The use of these agents has been well characterized in cancer, where anti-CD47 antibodies are currently in clinical trials [111]. Taking into account the role of macrophages in blocking the CD47–SIRPα axis, engineering macrophages to target CD47 may be effective. For example, similar to CAR-T cells, Chimeric Antigen Receptors for Phagocytosis (CAR-Ps) are designed to phagocytose specific targets, and it has been suggested that engineering macrophages to target CD47 may be an effective anti-tumour therapy [111]. Indeed, the use of these macrophages may prove effective in the clearance of chronic senescent cell accumulation following injury. However, there remain concerns over the manipulation of the CD47–SIRPα axis for therapeutic use including, but not limited to, the consensus that the CD47–SIRPα interactions demonstrated in mouse studies may not completely or as efficiently translate to human [165] the discovery that the clustering of CD47 can also influence the interaction between CD47 and SIRPα, and that the anti-CD47 non-blocking antibody 2D3 increases CD47 clustering [166,167], and the considerable functions of CD47 that are independent of SIRPα, which instead act upon SIRPγ or Thrombospondin 1 (TSP1), leading to potential off-target effects on T cells [168,169]. Another possibility is to increase the phagocytic capability of macrophages, by increasing ‘eat me’ signals. Defective expression of the ‘eat me’ signal calreticulin, a ligand required for activation of engulfment receptors on phagocytic cells, results in cellular resistance to efferocytosis, and apoptotic cells fail to be cleared by neighbouring macrophages [170]. A proportion of aged and cancerous cells are susceptible to being ‘labelled’ by macrophage-secreted calreticulin and are subsequently cleared from tissue [171]. Thus, the overexpression of calreticulin on senescent cells may provide a manner to increase phagocytosis and clearance from the tissue by macrophages. A third option involves the use of allogenic macrophages or administration of induced pluripotent stem (IPS) cell-derived macrophages from young donors [172] however, whether young macrophages have better phagocytotic ability is not clear, with some studies reporting reduced ability with age [173], while others report no difference [174]. Moreover, differences appear to be related to the tissue from which they are derived and whether they are tissue resident or infiltrating [175]. Thus, macrophage targeting may be a therapeutic option for the clearance of senescent cells in vivo.

5.2. Senolytics

An alternative way to clear senescent cells is via the use of drugs that kill senescent cells, known as senolytics. It has been shown that the removal of senescent cells using senolytic drugs, such as ABT-263, which work by targeting proteins involved in the apoptosis pathway (such as BCL-xL and BCL-2), have been successful in removing senescent cells in vitro and in vivo [73,176]. In particular, ABT-263, which signals through the anti-apoptotic proteins BCL-xL and BCL-2, has proven effective in eliminating senescent cells in mouse models [73,177] and is currently undergoing phase 1 clinical trials in cancer patients [178]. Moreover, dasatinib plus quercetin (DQ), which not only targets BCL-2 family members, but also HIF-1α, PI3-kinase and p21-related anti-apoptotic pathways [179], has proved efficient at improving physical dysfunction in idiopathic pulmonary fibrosis (IPF) patients [180]. However, there remain limitations with the use of senolytics, largely due to lack of specificity, bioavailability and their route of administration. Indeed, if macrophages themselves are found to become senescent in the context of injury, it may be necessary to combine work on senolytics to include those that can target macrophages as well as other cells. Alternatively, a therapy that activates gene networks to promote a ‘younger’ phenotype in macrophages could provide a realistic option. This is especially important due to the critical role macrophages play in generating a regeneration-permissive environment needed for tissue repair.

As previously mentioned, while senolytics have proven effective in killing senescent cells, they retain broad-spectrum activity, especially considering the heterogeneous nature of the senescence programme. In 2020, Cai et al. [181] developed a new prodrug named SSK1 which is specifically activated by β-galactosidase (β-Gal) activity, believed to be a well-documented marker of senescent cells. The prodrug was shown to eliminate senescent cells, as well as re-establishing low-grade inflammation and restoring some measures of function. This novel drug therefore represents a more selective method of deleting senescent cells in a wide range of cell types and tissues [181]. In addition to reducing the number of senescent cells in general, the prodrug was found to decrease the number of SA-β-Gal positive macrophages in injured lungs and aged livers, which was accompanied by a reduction in inflammation-related cytokines. This suggests that the removal of macrophages themselves may be beneficial. Here it is important to note the previously mentioned study by Hall et al. in which the authors reported that senescent macrophages were evident in young mice, entirely due to their expression of the markers p16 INK4a and β-galactosidase [128]. However, the expression of these markers was later described to be markers of macrophage polarization and response [134]. Importantly, this highlights the need for better methods for the identification of senescence in immune cells. Moreover, since senescent cells do not have any one specific universal marker, and due to their dynamic nature, that their markers can alter over time, the specificity of any drug aimed at targeting senescent cells alone is problematic.

5.3. Genetic engineering approaches

As the removal of macrophages themselves tends to lead to severe consequences, an alternative approach is to alter their gene expression in order to induce certain phenotypes. This involves the expression of certain transcription factors, for example IRF5, which is involved in the polarization of macrophages towards an inflammatory phenotype, which in turn prevents healing and promotes inflammation. Thus, manipulating macrophages to express lower levels of IRF5 could be a promising strategy [182]. Furthermore, the delivery of microRNAs or other molecules through methods of delivery such as liposomes may be a promising method for genetic manipulation worthy of further investigation [183,184].

Regarding genetic manipulation, an interesting aspect of senescence that is yet to be thoroughly explored is the regulation of gene expression at the epigenetic level. An interesting idea in regenerative medicine is that pluripotency factors can be expressed in senescent cells, to promote re-entry into the cell cycle and modify gene expression profiles. In addition to this, there has been interest in the use of epigenetic factors to promote reprogramming, as senescent cells display a repressive chromatin configuration, which is thought to stably silence proliferation-promoting genes, while also activating the SASP. In particular, histone modifications such as H3K27me3 have been associated with upregulation of the SASP in senescent cells [185]. Interestingly, in a recent study, it was also shown that the activation of the senescence programme leads to the remodelling of the epigenetic landscape by recruiting BRD4, a transcriptional and epigenetic regulator, which activates newly activated super-enhancers located next to SASP genes [186]. In addition, BRD4 was found to be critical for the SASP and paracrine signalling, and in senescence immune surveillance. Importantly this study revealed how cells can activate immune-modulatory genes required for paracrine immune activation and a tumour-suppressive immune surveillance programme. This may indicate the way in which senescent cells are regulated during homeostasis and may provide targets for regeneration. Indeed, the inhibition of BRD4 disrupts the ability of immune cells to target and eliminate pre-malignant senescent cells in vitro and in vivo [186], suggesting the effectiveness of this approach.

5.4. Epigenetic targeting

To complement macrophage genetic engineering, targeting epigenetic enzymes acting on the chromatin in senescent cells may also be an effective approach to switch on gene regulatory networks associated with a ‘younger morphology’ and may contribute to regulating the SASP. This approach is promising, as through the identification of ‘master regulators’ it allows the regulation of a large number of intricately interlinked genes and gene networks, instead of the manipulation of small subsets. Indeed, senescent cells are known to remodel the epigenetic landscape in order to induce the expression of genes implicated in cellular defence and the inflammatory response, many of which are characterized as part of the SASP [17]. Furthermore, alterations in epigenetic regulation have been associated with immunosenescence, as Menin promotes histone acetylation at the Bach2 locus, thereby suppressing T cell senescence, and subsequently, immunosenescence [187]. Indeed, this approach also relates to macrophages and may be successful in altering macrophage polarization.

5.5. Altering macrophage behaviour

Phagocytosis in macrophages is regulated through activation and inhibition of receptor signals. Activating receptors of macrophages sends a phagocytic signal that induces the ‘eat’ process. Importantly, when targeting senescent cells/macrophages for the purpose of regeneration/repair, it is important to note that an optimal result will probably be achieved through the careful, timely and balanced manipulation of this process. Indeed, the literature has pointed towards an important role of senescent cells in the initiation of the early stages of repair and of macrophages in clearing these cells in the first transient ‘wave’ of repair. It was not until recently that the soluble SASP was identified to have two distinct functional stages, the first stage being a highly anti-inflammatory stage enriched by TGFβ. Thus, the inhibition of this process as demonstrated by previous studies [188] is likely to have negative impacts on repair. However, the activation of macrophages at later points where senescent cells are found to accumulate is likely to be beneficial. Indeed, identifying the time points at which intervention is needed/most effective in different organ systems will be one of the most challenging aspects of developing effective therapies.

6. Conclusion, open questions and perspectives

From the initial thoughts that senescence was merely a natural result of ageing, our understanding of the role of senescence in ageing and human disease has greatly evolved. We are now aware that the senescence programme has more complex roles, notably in tissue regeneration and repair, beyond our current scope of knowledge. Furthermore, the crucial role of macrophages in repair and regeneration has also started to be demonstrated in the last decade, due to their well-reported role during injury. However, to date many questions remain regarding their specific role in the regulation of senescent cells following injury in different organ systems, which is complicated by the highly heterogeneous nature of the senescence programme, which directly influences macrophage function. Thus, to be able to target senescence for therapeutic means, a number of open questions need to be addressed, such as the following. (i) How do senescent cells interact with macrophages, and how does this change in different phases or ‘waves’ of the regenerative response? (ii) How do changes in the microenvironment trigger epigenetic/genomic/phenotypic changes in macrophages? (iii) At what point do senescent cells have to accumulate to before macrophages are unable to clear them, or when do senescent cells start to evade clearance by the immune system? Importantly, answering these questions will better elucidate how senescent cells and macrophages act in concert in the context of injury and repair, and will guide the development of treatments targeting senescent cells for therapeutic means.


SIRP structures explain paired-receptor specificity

The complex immune system of vertebrates must be tightly controlled in order to respond to pathogenic challenge without inducing unnecessary side effects. While soluble markers such as cytokines enable immune cells to localise to the site of infection, specific protein:protein interactions formed between cells fine-tune the immunological response.

The signal-regulatory proteins (SIRPs) are a family of cell-surface proteins expressed primarily on myeloid cells comprising three members: SIRP, SIRPß and SIRP [1] . While the extracellular regions of all three are very similar, the intracellular regions of these proteins are not alike and they convey very different signals to cells. SIRP&alpha recognises CD47, a protein expressed on the surface of most cells that acts as a molecular &lsquomarker of self&rsquo, and formation of this CD47/SIRP interaction inhibits engulfment of cells by macrophages. SIRPß, on the other hand, does not recognise CD47 and interaction of SIRPß with its (unknown) extracellular ligand stimulates macrophage engulfment. While SIRP recognises CD47 with lower affinity than SIRP, it is expressed on the surface of T-cells rather than macrophages and is thought to have no direct signalling function. The SIRPs thus typify the class of membrane protein families called &ldquopaired receptors&rdquo, which comprise several proteins with similar extracellular regions but radically different transmembrane/cytoplasmic regions with distinct (activating, inhibitory or neutral) signalling potentials.

The extracellular regions of SIRPs comprise three immunoglobulin (Ig) domains in a linear array. We had previously determined the structure of the first (amino-terminal) Ig domain of SIRP [2] . Using site directed mutagenesis we showed that the interaction with CD47 was mediated by the N-terminal (apical) loops of SIRP, more closely resembling how antibodies recognise antigens rather than general Ig-domain mediated cell:cell contacts.

To determine the precise mechanism by which SIRP recognises its ligand, we solved the structure of the N-terminal SIRP domain in complex with the sole extracellular Ig domain of CD47 in two crystal forms by molecular replacement using data collected at BM14 and ID23-2. Figure 74 shows the CD47/SIRP interface, which is unlike those seen for other moderate-affinity Ig domain mediated cell-surface interactions as it is highly convoluted, well fitting, and extensive, with a mutual involvement of loops from the amino-terminal ends of both SIRP and CD47.

Fig. 74: The structure of human SIRP in complex with CD47. A schematic representation of the five CD47 transmembrane helices is shown.

In contrast to SIRP, SIRPß and SIRP bind CD47 weakly or not at all, despite sharing > 90% sequence identity with SIRP. We solved structures of the N-terminal domains of SIRP and two isoforms of SIRPß by molecular replacement using data collected at BM14. These structures demonstrate the exquisite selectivity of the SIRP interaction interface. Overall, the structures of SIRPs , ß and are very similar ( Figure 75 ). However, for both isoforms of SIRPß specific amino acid differences with respect to SIRP could be identified which result in the inability of SIRPß to bind CD47. Further, by introducing a single amino acid mutation in the sequence of one isoform of SIRP&beta we were able to confer the ability to bind CD47 with an affinity approaching that of SIRP.

Fig. 75: a) Superposition of SIRPß and SIRP on the CD47/SIRP complex shows that the SIRPs have very similar structures. However, only SIRP binds CD47 with high affinity. b) Sites of SIRP amino acid polymorphism (red spheres) partition away from the CD47/SIRP interaction interface.

The N-terminal ligand-binding domain of SIRP has much higher sequence variability than most other cell-surface receptors. Mapping this sequence variability onto the structure of the CD47/SIRP complex shows that most SIRP&betaa polymorphisms reside away from the interaction interface ( Figure 75 ), implying that the sequence variability does not modulate the strength of the CD47 interaction. We propose that the driving force for this sequence variability might be to avoid recognition of SIRP by pathogens, which would exploit the ability of SIRP to down-regulate macrophage activity. An extension of the idea that SIRP might be a target for pathogens is that SIRPß may have evolved to mimic the extracellular regions of SIRP, its similar structure but opposite (activating) signalling activity serving to frustrate attempts by pathogens to exploit the inhibitory potential of SIRP.

The structure of SIRP in complex with CD47, together with the structures of SIRPß and SIRP in isolation, has shown us in molecular detail how this paired receptor binds its ligand and how subtle amino acid variations in the ligand-binding regions account for the exquisite ligand selectivity of the SIRP family. The partitioning of SIRP polymorphisms away from the CD47 interaction interface suggests that pathogenic challenge may drive SIRP polymorphism. Pathogen pressure on the inhibitory receptor provides a plausible raison d&rsquoêtre for the existence of paired receptors, with their similar extracellular structures but opposing signalling potentials.


We thank Dr. Kun Dong for assisting BK with ligation of gCD47 into the CAS9GFP vector.

1. Soto-Pantoja DR, Kaur S, Roberts DD. CD47 signaling pathways controlling cellular differentiation and responses to stress. Crit Rev Biochem Mol Biol. (2015) 50:212�. doi: 10.3109/10409238.2015.1014024

2. Roberts DD, Frazier WA. Thrombospondins and their receptors: evolving functions. In: Keeley FW, Mecham RP, editors. Evolution of Extracellular Matrix. Berlin Heidelberg: Springer (2013). p. 221�. doi: 10.1007/978-3-642-36002-2_8

3. Matlung HL, Szilagyi K, Barclay NA, van den Berg TK. The CD47-SIRPalpha signaling axis as an innate immune checkpoint in cancer. Immunol Rev. (2017) 276:145�. doi: 10.1111/imr.12527

4. Hughes AL. Origin and evolution of viral interleukin-10 and other DNA virus genes with vertebrate homologues. J Mol Evol. (2002) 54:90�. doi: 10.1007/s00239-001-0021-1

5. Cameron CM, Barrett JW, Mann M, Lucas A, McFadden G. Myxoma virus M128L is expressed as a cell surface CD47-like virulence factor that contributes to the downregulation of macrophage activation in vivo. Virology. (2005) 337:55�. doi: 10.1016/j.virol.2005.03.037

6. Willingham SB, Volkmer JP, Gentles AJ, Sahoo D, Dalerba P, Mitra SS, et al. The CD47-signal regulatory protein alpha (SIRPa) interaction is a therapeutic target for human solid tumors. Proc Natl Acad Sci USA. (2012) 109:6662𠄷. doi: 10.1073/pnas.1121623109

7. Murata Y, Saito Y, Kotani T, Matozaki T. CD47-signal regulatory protein alpha signaling system and its application to cancer immunotherapy. Cancer Sci. (2018) 109:2349�. doi: 10.1111/cas.13663

8. Advani R, Flinn I, Popplewell L, Forero A, Bartlett NL, Ghosh N, et al. CD47 Blockade by Hu5F9-G4 and rituximab in non-Hodgkin's lymphoma. N Engl J Med. (2018) 379:1711�. doi: 10.1056/NEJMoa1807315

9. Lin GHY, Chai V, Lee V, Dodge K, Truong T, Wong M, et al. TTI-621 (SIRPalphaFc), a CD47-blocking cancer immunotherapeutic, triggers phagocytosis of lymphoma cells by multiple polarized macrophage subsets. PLoS ONE. (2017) 12:e0187262. doi: 10.1371/journal.pone.0187262

10. Kauder SE, Kuo TC, Harrabi O, Chen A, Sangalang E, Doyle L, et al. ALX148 blocks CD47 and enhances innate and adaptive antitumor immunity with a favorable safety profile. PLoS ONE. (2018) 13:e0201832. doi: 10.1371/journal.pone.0201832

11. Kaur S, Elkahloun AG, Singh SP, Chen QR, Meerzaman DM, Song T, et al. A function-blocking CD47 antibody suppresses stem cell and EGF signaling in triple-negative breast cancer. Oncotarget. (2016) 7:10133�. doi: 10.18632/oncotarget.7100

12. Lee TK, Cheung VC, Lu P, Lau EY, Ma S, Tang KH, et al. Blockade of CD47-mediated cathepsin S/protease-activated receptor 2 signaling provides a therapeutic target for hepatocellular carcinoma. Hepatology. (2014) 60:179�. doi: 10.1002/hep.27070

13. Roberts DD, Kaur S, Isenberg JS. Regulation of cellular redox signaling by matricellular proteins in vascular biology, immunology, and cancer. Antioxid Redox Signal. (2017) 27:874�. doi: 10.1089/ars.2017.7140

14. Nath PR, Pal-Nath D, Mandal A, Cam M, Schwartz AL, Roberts DD. Natural killer cell recruitment and activation are regulated by CD47 expression in the tumor microenvironment. Cancer Immunol Res. (2019) 7:1547�. doi: 10.1158/2326-6066.CIR-18-0367

15. Kaur S, Kuznetsova SA, Pendrak ML, Sipes JM, Romeo MJ, Li Z, et al. Heparan sulfate modification of the transmembrane receptor CD47 is necessary for inhibition of T cell receptor signaling by thrombospondin-1. J Biol Chem. (2011) 286:14991�. doi: 10.1074/jbc.M110.179663

16. Miller TW, Kaur S, Ivins-O'Keefe K, Roberts DD. Thrombospondin-1 is a CD47-dependent endogenous inhibitor of hydrogen sulfide signaling in T cell activation. Matrix Biol. (2013) 32:316�. doi: 10.1016/j.matbio.2013.02.009

17. Soto-Pantoja DR, Terabe M, Ghosh A, Ridnour LA, DeGraff WG, Wink DA, et al. CD47 in the tumor microenvironment limits cooperation between antitumor T-cell immunity and radiotherapy. Cancer Res. (2014) 74:6771�. doi: 10.1158/0008-5472.CAN-14-0037-T

18. Feliz-Mosquea YR, Christensen AA, Wilson AS, Westwood B, Varagic J, Melendez GC, et al. Combination of anthracyclines and anti-CD47 therapy inhibit invasive breast cancer growth while preventing cardiac toxicity by regulation of autophagy. Breast Cancer Res Treat. (2018) 172:69�. doi: 10.1007/s10549-018-4884-x

19. Soto-Pantoja DR, Miller TW, Pendrak ML, Degraff WG, Sullivan C, Ridnour LA, et al. CD47 deficiency confers cell and tissue radioprotection by activation of autophagy. Autophagy. (2012) 8:1628�. doi: 10.4161/auto.21562

20. Kaur S, Soto-Pantoja DR, Stein EV, Liu C, Elkahloun AG, Pendrak ML, et al. Thrombospondin-1 signaling through CD47 inhibits self-renewal by regulating c-Myc and other stem cell transcription factors. Sci Rep. (2013) 3:1673. doi: 10.1038/srep01673

21. Miller TW, Soto-Pantoja DR, Schwartz AL, Sipes JM, DeGraff WG, Ridnour LA, et al. CD47 globally regulates metabolic pathways that control resistance to ionizing radiation. J Biol Chem. (2015) 290:24858�. doi: 10.1074/jbc.M115.665752

22. Zoppoli G, Regairaz M, Leo E, Reinhold WC, Varma S, Ballestrero A, et al. Putative DNA/RNA helicase Schlafen-11 (SLFN11) sensitizes cancer cells to DNA-damaging agents. Proc Natl Acad Sci USA. (2012) 109:15030𠄵. doi: 10.1073/pnas.1205943109

23. Rees MG, Seashore-Ludlow B, Cheah JH, Adams DJ, Price EV, Gill S, et al. Correlating chemical sensitivity and basal gene expression reveals mechanism of action. Nat Chem Biol. (2016) 12:109�. doi: 10.1038/nchembio.1986

24. Gardner EE, Lok BH, Schneeberger VE, Desmeules P, Miles LA, Arnold PK, et al. Chemosensitive relapse in small cell lung cancer proceeds through an EZH2-SLFN11 axis. Cancer Cell. (2017) 31:286�. doi: 10.1016/j.ccell.2017.01.006

25. Lok BH, Gardner EE, Schneeberger VE, Ni A, Desmeules P, Rekhtman N, et al. PARP inhibitor activity correlates with SLFN11 expression and demonstrates synergy with Temozolomide in small cell lung cancer. Clin Cancer Res. (2017) 23:523�. doi: 10.1158/1078-0432.CCR-16-1040

26. Tang SW, Bilke S, Cao L, Murai J, Sousa FG, Yamade M, et al. SLFN11 is a transcriptional target of EWS-FLI1 and a determinant of drug response in Ewing sarcoma. Clin Cancer Res. (2015) 21:4184�. doi: 10.1158/1078-0432.CCR-14-2112

27. Sousa FG, Matuo R, Tang SW, Rajapakse VN, Luna A, Sander C, et al. Alterations of DNA repair genes in the NCI-60 cell lines and their predictive value for anticancer drug activity. DNA Repair. (2015) 28:107�. doi: 10.1016/j.dnarep.2015.01.011

28. Deng Y, Cai Y, Huang Y, Yang Z, Bai Y, Liu Y, et al. High SLFN11 expression predicts better survival for patients with KRAS exon 2 wild type colorectal cancer after treated with adjuvant oxaliplatin-based treatment. BMC Cancer. (2015) 15:833. doi: 10.1186/s12885-015-1840-6

29. Tang SW, Thomas A, Murai J, Trepel JB, Bates SE, Rajapakse VN, et al. Overcoming resistance to DNA-targeted agents by epigenetic activation of Schlafen 11 (SLFN11) expression with class I histone deacetylase inhibitors. Clin Cancer Res. (2018) 24:1944�. doi: 10.1158/1078-0432.CCR-17-0443

30. Nogales V, Reinhold WC, Varma S, Martinez-Cardus A, Moutinho C, Moran S, et al. Epigenetic inactivation of the putative DNA/RNA helicase SLFN11 in human cancer confers resistance to platinum drugs. Oncotarget. (2016) 7:3084�. doi: 10.18632/oncotarget.6413

31. Li M, Kao E, Malone D, Gao X, Wang JYJ, David M. DNA damage-induced cell death relies on SLFN11-dependent cleavage of distinct type II tRNAs. Nat Struct Mol Biol. (2018) 25:1047�. doi: 10.1038/s41594-018-0142-5

32. Murai J, Tang SW, Leo E, Baechler SA, Redon CE, Zhang H, et al. SLFN11 blocks stressed replication forks independently of ATR. Mol Cell. (2018) 69:371� e376. doi: 10.1016/j.molcel.2018.01.012

33. Reinhold MI, Green JM, Lindberg FP, Ticchioni M, Brown EJ. Cell spreading distinguishes the mechanism of augmentation of T cell activation by integrin-associated protein/CD47 and CD28. Int Immunol. (1999) 11:707�. doi: 10.1093/intimm/11.5.707

34. Kaur S, Singh SP, Elkahloun AG, Wu W, Abu-Asab MS, Roberts DD. CD47-dependent immunomodulatory and angiogenic activities of extracellular vesicles produced by T cells. Matrix Biol. (2014) 37:49�. doi: 10.1016/j.matbio.2014.05.007

35. Singh NP, McCoy MT, Tice RR, Schneider EL. A simple technique for quantitation of low levels of DNA damage in individual cells. Exp Cell Res. (1988) 175:184�. doi: 10.1016/0014-4827(88)90265-0

36. Kaur S, Elkahloun AG, Singh SP, Arakelyan A, Roberts DD. A function-blocking CD47 antibody modulates extracellular vesicle-mediated intercellular signaling between breast carcinoma cells and endothelial cells. J Cell Commun Signal. (2018) 12:157�. doi: 10.1007/s12079-017-0428-0

37. Ma D, Liu S, Lal B, Wei S, Wang S, Zhan D, et al. Extracellular matrix protein tenascin C increases phagocytosis mediated by CD47 loss of function in glioblastoma. Cancer Res. (2019) 79:2697�. doi: 10.1158/0008-5472.CAN-18-3125

38. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. (2012) 2:401𠄴. doi: 10.1158/2159-8290.CD-12-0095

39. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. (2013) 6:pl1. doi: 10.1126/scisignal.2004088

40. Maxhimer JB, Soto-Pantoja DR, Ridnour LA, Shih HB, DeGraff WG, Tsokos M, et al. Radioprotection in normal tissue and delayed tumor growth by blockade of CD47 signaling. Sci Transl Med. (2009) 1:3ra7. doi: 10.1126/scitranslmed.3000139

41. Gewirtz DA. A critical evaluation of the mechanisms of action proposed for the antitumor effects of the anthracycline antibiotics adriamycin and daunorubicin. Biochem Pharmacol. (1999) 57:727�. doi: 10.1016/S0006-2952(98)00307-4

42. de la Casa-Esperon E. From mammals to viruses: the Schlafen genes in developmental, proliferative and immune processes. Biomol Concepts. (2011) 2:159�. doi: 10.1515/bmc.2011.018

43. Nath PR, Gangaplara A, Pal-Nath D, Mandal A, Maric D, Sipes JM, et al. CD47 expression in natural killer cells regulates homeostasis and modulates immune response to lymphocytic choriomeningitis virus. Front Immunol. (2018) 9:2985. doi: 10.3389/fimmu.2018.02985

44. Li Z, Calzada MJ, Sipes JM, Cashel JA, Krutzsch HC, Annis D, et al. Interactions of thrombospondins with a4b1 integrin and CD47 differentially modulate T cell behavior. J Cell Biol. (2002) 157:509�. doi: 10.1083/jcb.200109098

45. Li SS, Liu Z, Uzunel M, Sundqvist KG. Endogenous thrombospondin-1 is a cell surface ligand for regulation of integrin dependent T lymphocyte adhesion. Blood. (2006) 108:3112�. doi: 10.1182/blood-2006-04-016832

46. Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. (2012) 483:603𠄷. doi: 10.1038/nature11003

47. Soto-Pantoja DR, Ridnour LA, Wink DA, Roberts DD. Blockade of CD47 increases survival of mice exposed to lethal total body irradiation. Sci Rep. (2013) 3:1038. doi: 10.1038/srep01038

48. Kaur S, Roberts DD. Divergent modulation of normal and neoplastic stem cells by thrombospondin-1 and CD47 signaling. Int J Biochem Cell Biol. (2016) 81:184�. doi: 10.1016/j.biocel.2016.05.005

Keywords: radioresistance, epigenetics, CD47, thrombospondin-1, DNA damage response, schlafen-11, prostate cancer

Citation: Kaur S, Schwartz AL, Jordan DG, Soto-Pantoja DR, Kuo B, Elkahloun AG, Mathews Griner L, Thomas CJ, Ferrer M, Thomas A, Tang S-W, Rajapakse VN, Pommier Y and Roberts DD (2019) Identification of Schlafen-11 as a Target of CD47 Signaling That Regulates Sensitivity to Ionizing Radiation and Topoisomerase Inhibitors. Front. Oncol. 9:994. doi: 10.3389/fonc.2019.00994

Received: 15 August 2019 Accepted: 16 September 2019
Published: 01 October 2019.

Ira Ida Skvortsova, Innsbruck Medical University, Austria

Jih-Hwa Guh, National Taiwan University, Taiwan
Katerina Smesny Trtkova, Palacký University, Czechia

Copyright © 2019 Kaur, Schwartz, Jordan, Soto-Pantoja, Kuo, Elkahloun, Mathews Griner, Thomas, Ferrer, Thomas, Tang, Rajapakse, Pommier and Roberts. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

† These authors have contributed equally to this work

‡ Present address: Anthony L. Schwartz, Morphiex Biotherapeutics, Boston, MA, United States
David R. Soto-Pantoja, Wake Forest School of Medicine, Winston-Salem, NC, United States


Abstract 5218: Development of Anti-CD47 therapy for pediatric brain tumors.

Introduction: Extensive molecular characterization of Brain tumors done in recent years has proven valuable for tumor classification, risk stratification and outcome prediction for current treatment modalities. However the current standard of care has not improved the prognosis and carries an increased risk of cognitive impairments for children with brain tumors. A characteristic feature of tumor progression and recurrence is its ability to evade the immune system. Our hypothesis is that by modulating the innate immune system we can enhance the ability of macrophages to ‘eat and kill’ brain cancer cells. Recent data suggests that the interaction between the cell surface antigen CD47 and its binding partner SIRPα is a mechanism by which non-solid and solid tumors can evade the innate immune system. To see if anti-CD47 therapy is a potential therapy in malignant pediatric brain tumors we first looked at CD47 expression in freshly isolated patient and postmortem samples from 4 different tumor types Diffused Intrinsic Pontine Glioma, Medulloblastoma, Ependymoma and ATRTs. We tested the hypothesis that blocking CD47- SIRPα interaction induced phagocytosis in an in-vitro phagocytosis assay. We next established orthotopic xenografts models in immune compromised mice and treated them with anti-CD47 humanized antibody, which is currently being developed for clinical trials in hematopoietic and non-CNS malignancies.

Result: CD47 expression was upregulated in all tumor types and was present in >90% of the cells in high grade tumors. Increased CD47 expression was observed in CD15+ and CD133+ putative cancer stem cell population. Blocking the CD47- SIRPα interaction increases tumor phagocytosis by macrophages in-vitro. Systemic treatment with anti-CD47 antibody significantly reduced tumor burden in an orthotopic xenograft setting.

Conclusion: Anti-CD47 therapy is a viable and effective treatment modality for pediatric high grade brain tumors.

Citation Format: Sharareh Gholamin, Siddhartha S. Mitra, Chase Elliott Richard, Achal Achrol, Doosik Kong, Jun Jae Shin, Michelle Monje-Deisseroth, Yoon-Jae Cho, Irving Weissman, Samuel H. Cheshier. Development of Anti-CD47 therapy for pediatric brain tumors. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research 2013 Apr 6-10 Washington, DC. Philadelphia (PA): AACR Cancer Res 201373(8 Suppl):Abstract nr 5218. doi:10.1158/1538-7445.AM2013-5218


Watch the video: Stanford Cancer Institute - CD47 Discovery (December 2022).