12 VOLUME 9 NUMBER 1 JANUARY 2008 NATURE IMMUNOLOGY

time between injection and analysis increases the number of labeled marginal zone B cells, and in spleen sections, brightly labeled cells are present in the follicles. Using extensive controls, the authors rule out the possibili-ties that the antibody could diffuse into the follicles, that the labeling effects were depen-dent on the type of receptor used for labeling, and that the specificity of the B cell receptor was involved. They therefore conclude that marginal zone B cells continuously shuttle between the marginal zone and the splenic follicles and that most marginal zone B cells shuttle back and forth within a few hours.

These are notable findings. They demon-strate continuous interaction of marginal zone B cells with the follicular environment involving nontrivial cell numbers and kinet-ics. Furthermore, this shuttling seems to be independent of antigenic activation and must be regarded as a homeostatic process whereby blood-borne antigens, presumably opsonized by complement, can be delivered to the follicles. The cellular shuttling itself is independent of complement receptors, although the involvement of complement receptors for antigen transport is suggested, and it is possible that other receptors are involved. The continuity of this process is particularly noteworthy and could be based on a rather simple ‘up and down’ regulation of chemokine receptors, in particular S1P1 (Fig. 1). Not only do the marginal zone B cells form a bridge between innate and adaptive immunity because of their ability to rapidly respond to antigens, but in this way they also form a physical bridge by supplying antigens

to the follicles. Thus, published findings on the ability of memory B cells to return to follicles to undergo further mutation also indicate mechanisms whereby continuous scanning for antigens leads to refinement of the immune response6. Pathogenic antigens can be presented on follicular dendritic cells to sustain somatic hypermutation of the follicular B cells. In addition, self antigens can be delivered to the follicles, and this way the shuttling process may be involved in the maintenance of B cell tolerance, as suggested by the authors4. Future research might char-acterize the antigens delivered so that more insight into the regulation of autoantibodies can be generated.

Is this process unique to the spleen? There is ample evidence for cellular transport of antigens into lymph node follicles by B cells7, but it remains to be determined whether this is also associated with a similar cellu-lar ‘treadmill’ and, if so, from what location the cells migrate. The shuttling occurs as a homeostatic process and is independent of antigenic stimulation4. The transport most likely involves binding to complement recep-tors, and it is assumed that in the follicles, the antigens are carried over to the follicular dendritic cells with their high expression of complement receptors. So the issue is why B cells are involved in this process, as it makes no use of the specific B cell receptor recogni-tion capacity of the B cell. In fact, when mar-ginal zone B cells are activated through their B cell receptors, they do not migrate into the follicles but instead upregulate CCR7, which makes them migrate into the T cell zone,

where they function as antigen-presenting cells or differentiate into plasma cells.

Could it be that the continuous shuttling of marginal zone B cells works in two ways? First, the follicular B cells recognize and react to the deposited antigens, leading to immu-noglobulins of higher affinity. Second, the marginal zone B cells scan for all antigens that have entered the relatively large mar-ginal zone. Having missed the antigen when it first entered the marginal zone, these mar-ginal zone B cells can have a ‘second chance’ to inspect the repertoire of antigens for B cell receptor specificity by moving along the follicular dendritic cells, where the antigen is deposited by one of their neighbors. This process resembles, on a small scale, the way naive lymphocytes continuously migrate from blood to lymph nodes, to scan for antigens on dendritic cells in the T cell zones, and back to blood. These findings collectively indicate another level of complexity of the cellular interactions during immune surveillance. More experiments are needed to fully delin-eate the molecular basis of these movements and the cellular interactions involved.

1. Mebius, R.E. & Kraal, G. Nat. Rev. Immunol. 5, 606–616 (2005).

2. Gray, D., Kumararatne, D.S., Lortan, J., Khan, M. & MacLennan, I.C. Immunology 52, 659–669 (1984).

3. Ferguson, A.R., Youd, M.E. & Corley, R.B. Int. Immunol. 16, 1411–1422 (2004).

4. Cinamon, G., Zachariah, M., Lam, O.M., & Cyster, J.C. Nat. Immunol. 9, 54–62 (2008).

5. Cinamon, G. et al. Nat. Immunol. 5, 713–720 (2004).

6. Bende, R.J. et al. J. Exp. Med. 204, 2655–2665 (2007).

7. Phan, T.G., Grigorova, I., Oka da, T. & Cyster, J.G. Nat. Immunol. 8, 992–1000 (2007).

Bring out your deadDavid A Hume

Phagocytosis of apoptotic cells by macrophages occurs through recognition of specific molecules on the surfaces of apoptotic cells by receptors on macrophages. Three recent studies add to the understanding of this process.

David A. Hume is with the The Roslin Institute and

Royal (Dick) School of Veterinary Studies, University

of Edinburgh, Roslin EH25 9PS, Scotland, UK.

e-mail: d.hume@imb.uq.edu.au

Each cell in a multicellular organism has a ‘use-by’ date. As a cell approaches the end

of its useful life, it initiates the programmed death pathway, or apoptosis. One chief purpose of programmed cell death is to ensure that the cell can be ‘painlessly’ removed (that is, without

eliciting inflammatory responses) through a process of engulfment by another cell, permit-ting efficient recycling of its constituents and replacement. In some organisms, the dying cell is engulfed by its immediate neighbors, whereas in others the task is accomplished by ‘professional’ phagocytes. Three recent papers, one by Silva et al. in Immunity1 and two oth-ers by Miyhanishi et al.2 and Park et al.3 in Nature, evaluate various aspects of phagocy-tosis of apoptotic cells in mice and drosophila,

considerably extending the understanding of such processes.

A ‘professional’ phagocyte accomplishing the task of engulfing dying cells must leave the bloodstream, migrate toward its target, distinguish it from neighboring cells, engulf it, digest it and leave again. Although there are circumstances in which the phagocyte (in mammals, usually a macrophage) actually initiates the cell death process, in most cases the cell ‘commits suicide’ and normal cells

NEWS AND V IEWS©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

NATURE IMMUNOLOGY VOLUME 9 NUMBER 1 JANUARY 2008 13

are spared. Indeed, there is evidence that they generate a signal saying “I am not dead yet!” The processes of macrophage recognition and removal of apoptotic cells have been reviewed4. The removal of apoptotic cells is so rapid and efficient that dying cells can be difficult to see in tissue sections even when cell death is exten-sive. However, the ‘smoking gun’ is the mas-sive accumulation of macrophages in areas of cell death, such as that seen during embryonic development5 (Fig. 1). The macrophages that infiltrate developing organs in mice have been found to have a novel gene expression profile and a trophic as well as a clearance function6.

Considerable progress in the understanding of programmed cell death has been provided by studies of development in model organisms. In the worm Caenorhabditis elegans, the rather prosaically named C. elegans death (CED) genes identified by high-density mutation screens have led us to many new insights, not least fundamental understanding of the apop-tosis signaling pathways7. Among the CED mutations are those that do not affect death itself but instead lead to the accumulation of cellular ‘corpses’. CED mutations of this class may affect genes that must be expressed by the dying cell to signal to the phagocyte or by the phagocyte to enable it to recognize and remove the dying cell. In this class, CED-1 is a receptor on the phagocytic cell, and CED-7 is an ABC transporter expressed in the target cell that somehow contributes to making it ‘edible’.

In C. elegans, apoptotic cells are gener-ally cleared by neighboring cells, whereas in drosophila this is more commonly done by ‘specialist’ cells. As in C. elegans, many inde-pendent genome-scale mutation screens have identified genes in drosophila that are involved in cell death, usually with rather more colorful and morbid names such as grim, reaper, sickle, draper, morgue and croquemort8. Among these is a class of mutations associated with clearance defects. The paper in Immunity by Silva et al.1 demonstrates the power of this approach. The authors identify a deletion mutation that leads to less efficient uptake of ‘corpses’ by macro-phages. They name the gene ‘pallbearer’ and show that it encodes a so-called ‘F-box’ pro-tein. This immediately suggests an association with the SCF (Skp1–cullin–F box) complex, which regulates ubiquitin-dependent degrada-tion of proteins through the 26S proteasome. Although this same complex has been linked to control of cell death9, the authors provide evidence that the effect of the mutation is mac-rophage autonomous.

The authors demonstrate the macrophage-autonomous function by complementing the phagocytosis-defect phenotype with a macrophage-specific expression transgene

(using the croquemort promoter) encoding pallbearer. In subsequent experiments, they show that the pallbearer mutation interacts genetically with mutations in other compo-nents of the SCF-proteasome pathway. The authors also demonstrate that macrophages with the pallbearer mutation are not defective in phagosome formation and maturation or in degradation of the few apoptotic corpses that are engulfed. These data indicate a requirement for pallbearer and ubiquitination in the uptake of apoptotic cells.

Exactly where the pallbearer protein fits in drosophila macrophage biology is not appar-ent, and the extension to mammalian macro-phage biology is not yet evident, as the authors do not discuss whether there is a mammalian ortholog of pallbearer. The most similar gene in mammalian genomes, the gene encoding F-Box-only 28 (Fbox28), does not have detect-able expression in mouse macrophages (D.A.H., unpublished data). However, proteasome func-tion and ubiquitination have been associated with phagosome maturation in mammalian macrophages, where there is a crucial link to

antigen presentation to T lymphocytes10. The protein encoded by pallbearer could also con-tribute to intracellular signaling in phagocytes taking up apoptotic cells. The apoptosis inhibi-tor DIAP2, which contains a ubiquitin-ligase domain, has been shown to function in control of the NF-κB pathway9, indicating a conver-gence of immune and apoptosis signals.

The gene croquemort, previously identified by the same mutation-screening approach, encodes a membrane receptor with some homology to mammalian CD36, the first receptor linked to recognition of apoptotic cells in mammals4. From this beginning, the search for the mechanism(s) and signals that initiate engulfment of apoptotic cells has progressed rapidly, but the picture has become less clear with time. The unifying idea is that the dying cell puts onto its surface and/or secretes mol-ecules that are normally contained within, the so-called ‘eat me’ signals (Fig. 1).

Among such surface molecules are the membrane phospholipid phosphatidylserine (normally restricted to the inner leaflet of the bilayer11) and the endoplasmic reticulum

Opsonins

Integrins (CD11b)CD14CD36CD44CD93BAI1TIM1TIM4StabilinMBPGalactin3Scavenger receptor(s)LRP-1PS receptorMer, Axl

calRDNAPSCHO

FicolinsMBPsComplementGAS6MGF-E8Annexin 1Surfactants

Macrophage Apoptoticcell target

Figure 1 The array of receptors known to contribute to the recognition and internalization of apoptotic cells. The macrophage receptors induce signals in the macrophage to ensure that the apoptotic ‘corpse’ is efficiently degraded without undesirable inflammatory responses. Apoptotic cells display lipids (phosphatidylserine (PS)), carbohydrates (CHO), proteins (calreticulin calR) and DNA on their surfaces. The surface markers of apoptotic cells may be ‘seen’ directly by macrophage receptors or indirectly by means of opsonins, many of which are secreted by macrophages. Below, the ‘smoking gun’ of the clearance of dying cells during embryonic development. ‘MacBlue’ mice, in which all embryonic macrophages express cyan fluorescent protein5, show the massive infiltration of macrophages into the developing footpad as progressive apoptosis generates polarity of the foot and clears the space between the digits (left to right: 10.5, 11.5 and 12.5 days after coitus).

NEWS AND V IEWS©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y

14 VOLUME 9 NUMBER 1 JANUARY 2008 NATURE IMMUNOLOGY

protein calreticulin1. The list of macrophage surface or secreted proteins linked to the selec-tive recognition of apoptotic cells is substantial and is growing rapidly; these include CD14, CD36, CD44, CD93, ficolins, LRP-1 and other scavenger receptors, membrane tyrosine kinase mer and axl, the so-called phosphatidyl serine receptor, various integrins including the com-plement receptors, stabilin(s), soluble and mem-brane-bound C type lectins (mannose binding protein, galectin 3), annexin 1, GAS6, MGF-E8 (also called lactadherin) and surfactant proteins (Fig. 1). Of all these candidates, the definitive receptor for phosphatidylserine has remained elusive, and there have been previous false leads11.

Two recent papers in Nature address this problem. Miyanashi et al.2 begin from the observation that mouse peritoneal macro-phages that lack production of the putative phosphatidylserine opsonin MGF-E8 are still able to take up apoptotic cells. They identify a monoclonal antibody that blocks the inter-nalization of apoptotic T cells by peritoneal macrophages and identify the target antigen as TIM-4 (T cell immunoglobulin and mucin domain–containing molecule 4). A series of gain-of-function experiments, with expression of the genes in cells that are not macrophages, suggests that TIM-4 and the related TIM-1 can function as phosphatidylserine-specific cellular uptake receptors. The difficulty with this study is that the evidence that these genes are expressed specifically in macrophages and in substantial amounts is not compelling; conversely, they are expressed in several other cell types (nota-bly T cells) that do not phagocytose apoptotic cells. No loss-of-function data are provided to demonstrate that TIM-1 and TIM-4 provide a nonredundant mechanism of phosphatidyl-serine recognition in macrophages other than the effect of the original antibody, which might conceivably cross-react with other phosphati-

dylserine-binding proteins.Park et al.3 begin from a different position:

the ELMO–Dock180–Rac complex is a mem-brane signaling complex required for the uptake of apoptotic cells; ELMO1 is the mammalian homolog of CED-12. The authors ‘go fishing’ for binding partners for ELMO1 with a yeast two-hybrid approach and find the cytoplasmic tail of the plasma membrane receptor BAI1, the product of the Bai1 gene. The gene prod-uct is a rather obvious candidate receptor for phosphatidylserine, as it has the seven-trans-membrane-pass architecture of a G protein–coupled receptor and its extracellular domain contains five thrombospondin type 1 repeats, a module known to bind phosphatidylser-ine. The authors find that as with TIM-1 and TIM-4, enforced expression of BAI1 in nonphagocytes permits uptake of apop-totic cells, and the protein binds phospha-tidylserine on the surface of apoptotic cells. And although they show that the soluble BAI1 extracellular domain is able to block uptake of the apoptotic cells by macro-phages in the peritoneum, this finding really only indicates that the molecule can bind phosphatidylserine (not that it serves as the functional phosphatidylserine receptor on macrophages). As with TIM-1 and TIM-4, therefore, compelling evidence is not pro-vided that BAI1 is a functional receptor that is expressed in substantial amounts by macro-phages and is absolutely required for uptake of apoptotic cells in physiological settings, and neither receptor is linked to phosphati-dylserine recognition by human phagocytes.

So it would seem that there are a couple of new candidate ‘players’ in the rather crowded field of receptors for dying cells. No doubt efforts to ‘knock out’ of each of these and stud-ies of the subsequent effect on phagocytosis of apoptotic cells are underway, but given that, for example, TIM-4 is part of a multigene family

and that there is strong evidence of redundancy in the apoptotic cell recognition pathways used by phagocytic cells (such as pathways using the other receptors mentioned above), it would not be unexpected to find a mini-mal distinctive phenotype resulting from the knockout of a single one of these new partici-pants. This does not mean that the individual receptors are truly redundant. As Miyanishi et al. point out, TIM-1 and TIM-4 map to a human chromosomal region associated with susceptibility to certain autoimmune diseases, and specific failures of macrophage-mediated clearance could contribute to a wide range of chronic inflammatory states, as exemplified by a study of the effect of MGF-E8 deficiency in atherosclerosis12.

Perhaps there is not a ‘one size fits all’ solution to the issue of apoptotic cell recognition or even of phosphatidylserine binding. Instead, differ-ent cell types, in different locations, with dif-ferent precise mechanisms of death (including autophagic death13) and at different stages in the death pathway may use different solutions to the problem of orchestrating their own ‘funerals’.

1. Silva, E., Au-Yeung, H.W., Van Goethem, E., Burden, J. & Franc, N.C. Immunity 27, 585–596 (2007).

2. Miyanishi, M. et al. Nature 450, 435–439 (2007).3. Park, D. et al. Nature 450, 430–434 (2007).4. Henson, P.M. & Hume, D.A. Trends Immunol. 27, 244–250

(2006).5. Ovchinnikov, D.A. et al. J. Leukoc. Biol. advance online

publication, doi:10.1189/jlb.0807585 (30 October 2007).

6. Rae, F. et al. Dev. Biol. 308, 232–246 (2007).7. Lettre, G. & Hengartner, M.O. Nat. Rev. Mol. Cell. Biol.

7, 97–108 (2006).8. Hay, B.A. & Guo, M. Annu. Rev. Cell. Dev. Biol. 22,

623–650 (2006).9. Huh, J.R. et al. J. Biol. Chem. 282, 2056–2068

(2007).10. Houde, M. et al. Nature 425, 402–406 (2003).11. Schlegel, R.A. & Williamson, P. Sci. STKE 2007, pe57

(2007).12. Ait-Oufella, H. et al. Circulation 115, 2168–2177

(2007).13. Klionsky, D.J. Nat. Rev. Mol. Cell. Biol. 8, 931–937

(2007).

NEWS AND V IEWS©

2008

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eim

mun

olog

y