How do apoptotic cells advertise their presence?
It has long been observed by many investigators that even in tissues with known high turnover of cells (e.g. thymus, bone marrow, or testes) few apoptotic cells are seen in the steady state. However, when the apoptotic cell clearance process is disrupted, for example through genetic ablation of engulfment genes or pharmacological approaches that inhibit cell clearance, there is an accumulation of uncleared corpses that becomes readily apparent (
Elliott et al., 2010;
Henson, 2005;
Lu et al., 2011;
Nagata et al., 2010). This suggests that in the steady state the apoptotic cell clearance machinery is quick and efficient, such that the dying cells were sensed, recognized and cleared quickly. To use an analogy, we do not appreciate the garbage collector until there is a disruption in trash collection and we begin to notice the unpleasant odor. Similarly, as long as the engulfment machinery is intact and functional, uncleared apoptotic cells are not evident; however, disruption of cell clearance leads to accumulation of secondarily necrotic corpses in tissues often associated with autoimmunity toward self antigens derived from the uncleared corpses (
Franz et al., 2006). This implies that the apoptotic cells must somehow ‘advertise’ their presence at the earliest stages of death to promote clearance. The very close link between cells beginning to die and their quick and efficient removal came first came from genetic studies in
C. elegans. Using a mutant background where the caspase-mediated apoptosis was partially impaired, two groups (
Hoeppner et al., 2001;
Reddien et al., 2001) elegantly demonstrated that even under these partial death circumstances, if the engulfment machinery was intact, these ‘partially dead’ cells are efficiently recognized and removed; however, when the engulfment machinery is also impaired, then the partially dead cells abort the death program. This strongly suggests that the phagocytes are extremely efficient in sensing and detecting the dying cells at the earliest stages of apoptosis, and contribute to the ‘final stages’ of death.
In the above mentioned
C. elegans studies, the phagocytes are often the healthy neighbors. In the context of mammalian cells, this is often not the case. For example, in the thymus, a dying thymocyte is unlikely to be engulfed by a neighboring healthy thymocyte (since immature thymocytes lack the cytoplasmic volume and cytoskeletal machinery to ingest another cell of the similar size); rather a resident macrophage or a dendritic cell within the thymus is more likely to mediate the clearance (
Ravichandran, 2003). In such a case, the dying cell must be able to advertise its state of death and in turn recruit a phagocyte to its proximity. This led to the concept of soluble ‘find-me’ signals being released by apoptotic cells. Such soluble mediators would then set up a gradient within the tissue to attract the phagocytes to the proximity of the dying cells. The current notion is that the find-me signals, released by the dying cells at the earliest stages of apoptosis, would then be sensed by the phagocytes via receptors, and subsequent signaling within the phagocytes would induce the migration to the proximity of the dying cells (
Peter et al., 2010).
Specific find-me signals, their release from apoptotic cells, and sensing by phagocytes
In the past few years, four possible find-me signals released from apoptotic cells have been reported. These include: the lipid lysophosphatidylcholine (LPC); sphingosine 1-phosphate (S1P); the fractalkine CX3CL1; and the nucleotides ATP and UTP (). Remarkably, these are different types of molecules, yet they are all linked to monocyte or macrophage recruitment toward apoptotic cells. The discussion below focuses on what we currently know about the generation and release of these find-me signals, how phagocytes sense them, and whether or not they have been proven to be of relevance in vivo.
Lauber et al identified LPC as a find-me signal released by MCF-7 breast cancer cells based on the ability of their apoptotic cell supernatants to attract the THP-1 monocytic cell line (
Lauber et al., 2003). Few other cell lines and different modes of apoptosis could lead to the release of LPC, but the MCF-7 cells were the best characterized. The authors also went on to show elegantly that the calcium-independent phospholipase A2 (iPLA2), upon cleavage by caspases, is likely responsible for the release of LPC from the dying cells (
Lauber et al., 2003). This study brought the idea of find-me signals to the fore and that caspase-dependent release of soluble mediators could act as find-me signal to recruit monocytes from circulation in to tissues with apoptotic cells. However, a number of interesting questions were raised by this work. The first is the challenge of LPC setting up a concentration gradient. The concentration of LPC released by MCF-7 cells (at least as detected in the assay system) is small (
Lauber et al., 2003). Since there is a high concentration of LPC in the serum or plasma (100μM), it is unclear how LPC can set up a gradient to attract phagocytes specifically to the proximity of apoptotic cells. While high concentration of LPC could be present locally near a dying cell and thereby still provide the gradient, higher LPC concentrations cause lysis of many cells. Perhaps LPC is not recognized in its native form, but rather is bound to other serum components (and therefore effectively unavailable), further modified in tissues, or LPC may function together with another soluble mediator(s). The second challenge with LPC as a find-me signal is that G2A (
Peter et al., 2008), the receptor that is linked to LPC recognition by phagocytes, has been controversial and the validity of the original identification of G2A as the receptor LPC is questionable (
Witte et al., 2005). Therefore, the specific receptor on phagocytes that mediates LPC dependent movement toward apoptotic cells remains to be determined. The third challenge is that the role of LPC as a find-me signal
in vivo is not yet established. While the original studies clearly show that LPC can act as a find-me signal
in vitro (
Lauber et al., 2003), complementary
in vivo studies of LPC dependent recruitment of phagocytes to apoptotic cells remain to be established.
The soluble molecule sphingosine 1-phosphate (S1P)has also been proposed as a find-me signal (
Gude et al., 2008;
Weigert et al., 2010). Gude et al suggested that induction of apoptosis results in upregulation of S1P kinase 1 (SphK1) (
Gude et al., 2008). The increased SphK1 was then linked to generation of S1P, and in turn S1P-dependent recruitment of macrophages to the apoptotic cell supernatants. However, a second study suggests that S1P kinase 2 (SphK2) itself is a target of caspase 1, and that the cleaved fragment of SphK2 could be ‘released’ from dying cells into the extracellular space where it would generate S1P (
Weigert et al., 2010). Given the well-known role of S1P as a migratory signal, and the pharmacological interest in modifying S1P based signals, this observation is potentially very interesting. However, a number of key issues need to be resolved. First, all of the work to date suggesting S1P as a find-me signal has been done
in vitro, and the relevance of S1P in recruiting phagocytes to apoptotic cells
in vivo has not been determined. Staurosporine-induced cell death has been shown to induce caspase-1 and in turn cleavage of SphK2 (
Weigert et al., 2010). However, caspase-1 is normally not induced during other forms of apoptosis and the mechanism for S1P generation remains to be better defined. With respect to sensing of S1P, it can be recognized by the receptors S1P
1 through S1P
5, yet which of these G-protein coupled receptors (GPCRs) are relevant for phagocytic recruitment to apoptotic cells is not known. Both sphingosine kinase 1 and 2 have been linked to S1P generation during apoptosis (
Gude et al., 2008;
Weigert et al., 2010), but presumably by distinct mechanisms: SphK1 protein level being increased during apoptosis versus cleavage of SphK2 by caspases. Since genetically targeted mice exist for both
Sphk1 and
Sphk2, the specific enzyme question as well as the relevance of S1P as find-me signal should be addressable in an organismal context. Collectively, while S1P could serve as a find-me signal
in vitro, its relevance
in vivo, the receptors on phagocytes that help mediate recruitment to apoptotic cells, and the relative importance of S1P compared to other find-me signals need to be determined.
In addition to the two lipid mediators referred to above, a soluble fragment of fractalkine (CX3CL1) protein can also serve as a find-me signal for monocytes (
Truman et al., 2008). Chris Gregory and colleagues demonstrated that a soluble chemokine fragment of fractalkine, which is normally on the plasma membrane and serves as an intercellular adhesion molecule, is released as a 60kDa fragment during apoptosis and can act as a chemoattractant (
Truman et al., 2008). CX3CL1 release is caspase dependent, although it may not be direct (the cleavage being more likely done by extracellular ADAM protease family members). Truman et al also suggested that fractalkine (CX3CL1) is released as part of microparticles from early stages of apoptotic Burkitt Lymphoma cells (which are the malignant version of the germinal center B cells). CX3CR1, the receptor for CX3CL1, appeared important in sensing the chemokine and for inducing monocyte migration both
in vitro and
in vivo. Mice lacking CX3CR1 show a defect in the migration of macrophages to the germinal centers, where a high rate of apoptosis takes place. While the authors clearly acknowledge that the fractalkine might be more specific for Burkitt Lymphoma or germinal center B cells, this work through its
in vivo studies using CX3CR1 establish fractalkine (CX3CL1) as a
bona fide find-me signal.
The latest to join the find-me signal mediators are the nucleotides found in the supernatants of apoptotic cells. Elliott et al. demonstrated that the regulated release of triphosphate nucleotides ATP and UTP from early apoptotic cells can potently attract monocytes both
in vitro and
in vivo (
Elliott et al., 2009). Nucleotide release was shown to occur in Jurkat cells, primary thymocytes as well as MCF-7 cells and lung epithelial cells after apoptosis induction via multiple modalities (crosslinking Fas, UV-treatment, etoposide etc.) and is dependent on caspase activity. Interestingly, less than 2% of the cellular ATP is released from early stage apoptotic cells when their plasma membrane is still intact. The released nucleotides in the apoptotic cell supernatants are ‘chemotactic’ for monocytes (i.e. inducing directional migration) rather than promoting ‘chemokinesis’ (i.e. random migration).
Importantly, the apoptotic cell supernatants injected in to a mouse dorsal airpouch could preferentially attract monocytes and this recruitment is abolished when the triphosphonucleotides from the supernatants were degraded (
Elliott et al., 2009). Furthermore, using clearance of dying thymocytes (after Dexamethasone-induced apoptosis in the thymus) as a readout, enzymatically ablating the nucleotides, or blocking nucleotide recognition by phagocytes results in uncleared corpses in this
in vivo model. The sensing of the released extracellular nucleotides was shown to involve the P2Y family of nucleotide receptors; more specifically, the P2Y2 member is relevant for sensing of nucleotides by monocytes both
in vitro and in the context of P2Y2 genetically ablated mice. Collectively, these data identified release of the nucleotides ATP and UTP by early stage apoptotic cells and their sensing via P2Y2 on monocytes, and that this find-me signal circuit is relevant for apoptotic cell clearance
in vivo.
As with other studies, a number of challenges remain in defining the role of nucleotides as a find-me signal. Extracellular nucleotides are often degraded by nucleotide triphosphatases (NTPases) (Knowles;
Schetinger et al., 2007). Given the small amount of ATP released (about 100nM or ~2%), it is unclear how a nucleotide gradient would be established and what distances such a gradient may be able to ‘travel’ to attract phagocytes. The
in vivo data in the airpouch model and the thymocyte clearance model suggest that a relevant gradient can be established by this amount of ATP and recognized by monocytes (
Elliott et al., 2009), but the distance for this ‘final call’ remains unknown. Also, the NTPdases are expressed at different amounts in various tissues (Knowles;
Schetinger et al., 2007), and therefore, may further regulate the distance for the attraction signal. It is noteworthy however, that diphosphate nucleotides (ADP or UDP), monophosphate (AMP) or adenosine show very poor chemotactic activity, suggesting that it is the triphosphates that were acting as find-me signals (
Elliott et al., 2009). Another challenge is that while
P2ry2 ablated mice clearly show a defect in attracting monocytes (
Elliott et al., 2009), there was still residual migration in these mice, suggesting other P2Y family members may also play a role. Furthermore, the signaling downstream of the P2Y receptors and the specific signaling that induces migration of monocytes remains to be determined. Nevertheless, nucleotides represent one of the better-understood find-me signals at this point.
In addition to the above four, the release of ribosomal protein S19 (which dimerizes during apoptosis) and binds to the complement receptor C5a r on monocytes has been suggested as a find-me signal (
Yamamoto, 2007). However, it appears that S19 is likely released at very late stages of apoptosis. Also, a fragment of the tyrosyl tRNA synthetase (EMAPII) released from apoptotic cells has been shown to attract monocytes (
Shalak et al., 2001). Surprisingly, EMAPII has inflammatory properties and also attracts (and activates) neutrophils. The neutrophil attraction and why this would be the case is unclear.
How are the find-me signals released?
One of the key features of find-me signals is that they are released when the apoptotic cells are still intact such that the released mediators can attract the phagocytes, which in turn would clear the apoptotic cells before they become secondarily necrotic. This implies that there has to be mechanisms in place that allow the direct or indirect generation and release of find-me signals that are coupled to induction of apoptosis. At present, we do not know how LPC is released from dying cells. The generation of S1P as reported by Weigert et al appears to involve caspase-1 dependent release of a fragment of Sphingosine kinase 2 (SphK2), which in turn would generate S1P. During apoptosis, a specific cleavage site in SphK2 results in a fragment of this enzyme retaining its enzymatic property. It was also suggested that this cleaved fragment of SphK2, presumably by binding to phosphatidylserine (PtdSer) on the inner leaflet of the plasma membrane, could be ‘released’ from dying cells when the PtdSer is flipped to the outer leaflet during apoptosis. SphK2 release into the extracellular space would then generate S1P. The timing of caspase dependent cleavage of SphK2 versus the exposure of phosphatidylserine (PtdSer), and the generality of this mechanism for other forms of apoptosis remain to be determined. For fractalkine release, Truman et al suggested that the 60kDa fragment of fractalkine was released as microparticles from early stages of apoptotic Burkitt Lymphoma cells. The microparticles were annexin V positive and suggested that they may be generated at the same time as PtdSer exposure.
The mechanism of nucleotide release during apoptosis has been better defined (
Chekeni et al., 2010). Nucleotides are released through PANX1, a member of the pannexin family of channels on the plasma membrane. PANX1 is a four transmembrane protein (likely functioning as homo-hexamers) forming large pores that allow molecules up to 1 kDa to pass through the plasma membrane(
D'Hondt et al., 2009). Silencing of PANX1 expression results in loss of nucleotide release from apoptotic cells, while overexpression of PANX1 strongly enhances the nucleotide release during apoptosis (
Chekeni et al., 2010), and results in better recruitment of monocytes. Moreover, PANX1 currents are detectable upon induction of apoptosis, and this correlates with plasma membrane permeability. The uptake of dyes such as YO-PRO1 (and TO-PRO3), which have long been used to track plasma membrane permeability of early apoptotic cells, is dependent on PANX1 (
Chekeni et al., 2010). Mechanistically, it was demonstrated that the C-terminal tail of PANX1 itself is a target of caspsase dependent cleavage (by caspases 3 and 7). Mutation of the cleavage site within PANX1 results in failure of nucleotide release, while truncation of PANX1 at the cleavage site leads to a constitutively active protein (
Chekeni et al., 2010). In subsequent work, Vishva Dixit and colleagues demonstrated that cells from mice with a genetic ablation of
Panx1 are defective in nucleotide release during apoptosis as well as in recruiting macrophages (
Qu et al., 2011). Thus, a model has emerged where apoptotic cells release the nucleotides ATP and UTP at the earliest stages of apoptosis via caspase dependent cleavage and opening or activation of PANX1 (). These nucleotides are then be sensed by P2Y2 (and possibly other P2Y family members) and lead to migration of monocytes to the proximity of apoptotic cell clearance.
Other roles for find-me signals besides phagocyte attraction?
In addition to establishing a chemotactic gradient to aid in the location of the dying cells, find-me signals may also modulate the phagocytic activity of cells in the direct vicinity of the apoptotic cells. This becomes relevant when one considers situations where the dying cell is likely eaten by a neighbor (not a recruited phagocyte), yet releases find-me signals. For example, apoptotic airway epithelial cells release nucleotides (
Elliott et al., 2009), yet neighboring epithelial cells can engulf other dying epithelial cells without any migration to the prey. This raises the question whether find-me signals have other roles besides recruiting phagocytes. An interesting possibility is that the find-me signals such as nucleotides or S1P could ‘activate’ or ‘prime’ phagocytes and improve their phagocytic capacity. In flies, apoptosis in the context of a tissue enhances the capacity of engulfment by the neighboring cells – for example, by inducing the upregulation of the engulfment machinery (
MacDonald et al., 2006;
Ziegenfuss et al., 2008). In mammals, the ‘bridging’ molecule MFG-E8 (which binds to PtdSer on apoptotic cells and facilitate engulfment by engaging the integrin α
vβ
3 on phagocytes, see below), is expressed by activated but not resting macrophages (
Hanayama et al., 2002) and it has been demonstrated that the find-me signal fractalkine can induce production of the bridging molecule MFG-E8 by macrophages (
Miksa et al., 2007). One possibility is that the find-me signals are sort of like a ‘smell of food’ that increases the ‘appetite’ of the phagocyte by upregulating the engulfment machinery components. If this were to be applicable to other cell types beyond macrophages, such as epithelial cells, perhaps the find-me signals do not quite act as the final call for a phagocyte, but rather help boost the engulfment machinery in the neighbor and thereby actively promote its own clearance.
ATP has been typically thought of as a danger signal (
Trautmann, 2009), and has been linked to recruitment of neutrophils during inflammatory conditions and for inducing anti-tumor immunity (
Aymeric et al., 2010). Yet, apoptotic cell supernatants appear to preferentially recruit monocytes over neutrophils
in vivo (
Elliott et al., 2009), and the apoptotic cell clearance is usually anti-inflammatory and immunologically silent (
Fadok et al., 1998b;
Voll et al., 1997). There are perhaps several differences between the regulated release of nucleotides during apoptosis, and cytolysis (such as necrosis). Less than 2% of total cellular ATP is released via PANX1 by early apoptotic cells (
Chekeni et al., 2010;
Elliott et al., 2009), while damage-induced loss of membrane integrity may release a much higher quantity of nucleotides. Also, the notion of ATP as an inflammatory molecule derives from its ability to activate the ionotropic nucleotide receptor P2X7, which in turn results in activation of the inflammasome and release of pro-inflammatory cytokines (
Bours et al., 2006;
Di Virgilio, 2007). In fact, ATP derived from necrotic cells has been shown to result in sterile inflammation via inflammasome activation (
Iyer et al., 2009). However, much higher ATP concentrations are necessary for P2X7 activation (EC
50 > 100μM) than those necessary for activation of the receptors mediating monocyte chemotaxis (such as P2Y2; EC
50 < 1μM) (
Bours et al., 2006;
Trautmann, 2009). Although early in vitro studies link ATP release via Panx1 to P2X7 activation, Dixit and colleagues have now clearly shown (using
Panx1 ablated mice) that Panx1 is dispensable for P2X7 and inflammasome activation (
Qu et al., 2011). Interestingly, lower concentrations of ATP have been shown to have an anti-inflammatory effect by suppressing the secretion of inflammatory cytokines, while promoting the release of anti-inflammatory cytokines (
Di Virgilio, 2007;
Hasko et al., 2000;
la Sala et al., 2001;
Wilkin et al., 2002). Therefore, the concept of ATP as a universal danger signal might be too simplistic. How nucleotides and other find-me signals may influence immunogenic versus non-immunogenic responses to apoptotic cells remains to be better examined (
Green et al., 2009; Zitvogel et al.).
The publisher's final edited version of this article is available at 
me signals – the altruistic suicide notes
