Megakaryocyte emperipolesis: a new frontier in cell-in-cell interaction

Abstract

Histology of bone marrow routinely identifies megakaryocytes that enclose neutrophils and other hematopoietic cells, a phenomenon termed emperipolesis. Preserved across mammalian species and enhanced with systemic inflammation and platelet demand, the nature and significance of emperipolesis remain largely unexplored. Recent advances demonstrate that emperipolesis is in fact a distinct form of cell-in-cell interaction. Following integrin-mediated attachment, megakaryocytes and neutrophils both actively drive entry via cytoskeletal rearrangement. Neutrophils enter a vacuole termed the emperisome which then releases them directly into the megakaryocyte cytoplasm. From this surprising location, neutrophils fuse with the demarcation membrane system to pass membrane to circulating platelets, enhancing the efficiency of thrombocytogenesis. Neutrophils then egress intact, carrying megakaryocyte membrane and potentially other cell components along with them. In this review, we summarize what is known about this intriguing cell-in-cell interaction and discuss potential roles for emperipolesis in megakaryocyte, platelet and neutrophil biology.

Summary sentence

Megakaryocyte emperipolesis is a distinct form of cell-in-cell interaction whereby cells transiting through the megakaryocyte cytoplasm can transfer membrane to megakaryocytes and their platelets.

Megakaryocytes – beyond platelet production

Megakaryocytes (MKs) are large polyploid cells, ranging up to 100μm in diameter and 128N. Extensively studied for their ability to elaborate platelets, evidence now shows that their portfolio extends beyond hemostasis (1). MKs regulate the bone marrow niche, supporting the development and function of plasma cells and hematopoietic stem cells (2–5). Expressing a range of Toll like receptors, Fc γ receptors and other immune sensors, MKs respond to immunologic danger signals, and even contribute to antiviral immunity (6–10). MKs present antigen and thereby engage adaptive immune responses (11–13). Finally, MKs produce proinflammatory microparticles and cytokines and can play a direct role in systemic inflammatory disease (8). Thus, MKs are immune cells as well as hemostatic cells (1).

In this review, we focus on another intriguing but understudied aspect of MK biology: their ability to internalize other hematopoietic cells. In 1970, studying fresh marrow from a patient with thrombocytopenia and anemia, Larsen visualized a neutrophil entering into, moving around within, and then leaving a MK without apparent harm to either cell (14). This phenomenon was termed emperipolesis, coined originally from the Greek (em inside, peri around, and polemai wander about) to describe the movement of lymphocytes within tumor cells (15). MK emperipolesis is now routinely observed in marrow, usually but not invariably with neutrophils as the internalized cells. Almost 50 years since its initial description, the cell biology, regulation, and function of MK emperipolesis remain largely unexplored.

Often considered a mere histological curiosity, emperipolesis bears hallmarks of a process of biological importance. It occurs constitutively in healthy marrow and is induced selectively across a range of conditions. Emperipolesis occurs in all mammals studied so far (mouse, rat, cat, dog, cow, monkey, human), indicating conservation across more than 50 million years of divergent evolution (16–20). In fact, ultrastructural, in vitro modelling, and in vivo studies now establish that emperipolesis is a distinct form of cell-in-cell relationship that challenges assumptions about how mammalian cells can interact with each other (21). Here, we provide an overview of what is known about MK emperipolesis and discuss how it might modulate neutrophil, MK, and platelet biology.

Emperipolesis as a novel form of cell-in-cell interaction

Discovery and definition

The term ‘emperipolesis’ was introduced by Humble et al. in 1956 to refer to the phenomenon, in live cell preparations, of cells entering and moving around within other cells (15). It is not restricted to MKs; indeed, Humble studied lymphocytes within tumor cells, as observed also by others (22–24). In Rosai-Dorfman disease, a pathologic histiocytosis sometimes associated with mutations affecting the MAPK pathway, ‘emperipolesis’ is used to describe the histological appearance, in fixed sections, of morphologically-intact leukocytes within histiocytes, and is considered a diagnostic hallmark of the disease (25). Some have proposed that ‘emperipolesis’ should be applied even more broadly, to encompass any cell-in-cell phenomenon (26). However, our preference is to restrict the term to a cell-in-cell interaction characterized at a minimum by preserved viability of the internalized cell. Our study of emperipolesis in MKs suggests that other features may further distinguish this interaction, including active cytoskeletal participation by both participants (we will term these “host” and “guest” cells), egress from the endocytic vesicle to enter the host cytoplasm, and exit from the host without injury to either cell (21) (Figure 1). Whether these additional features are shared by all cell-in-cell events that are currently termed emperipolesis remains to be determined.

Figure 1:

Cell-in-cell interactions

Types of cell-in cell interaction

Cell-in-cell interactions occur in several forms (Figure 1). In phagocytosis, the engulfing cell actively encompasses a passive target, leading to its destruction as the phagosome fuses with lysosomes; when the target cell is a dead or dying, this interaction is also termed efferocytosis (26, 31). In entosis, the target cell is alive and contributes actively to the engulfing event, although death of the internalized cell still remains the ultimate outcome (26, 42). In transcellular migration, leukocytes penetrate through the body of an endothelial cell to enter the tissues, a process to which both lineages contribute (36, 43). Endothelial cells develop an interconnected network of tubulovesicular structures called the lateral border recycling compartment (LBRC) that is directed to the apical surface to enable leukocyte entry, with additional recruitment of LBRC vesicles to complete the transcellular channel (37, 44). In each of these forms of cell-in-cell interaction, engulfed cells reside with a membrane-bound compartment within the host cell rather than in contact with its cytoplasm, thus remaining topologically outside the cell.

Emperipolesis is different, at least in MKs (21). In studies focused principally on the interaction between neutrophils and MKs, we observed that efficient emperipolesis required active cytoskeletal engagement by both participants. Indeed, videomicroscopy shows that neutrophils sometimes employ podosomes to “push” their way in. Electron microscopy complemented by immunofluorescence studies suggests a multi-step process. Neutrophils enter the MK in a vesicle we have termed an emperisome, initially larger than the engulfed cell but contracting over time such that the membranes of emperisome and neutrophil become closely apposed. In places, this two-membrane structure becomes indistinct, including patches that bear some ultrastructural similarity to tight junctions (Figure 2). The emperisome membrane then disappears, leaving the neutrophil with the MK cytoplasm. After a period of residence, neutrophils exit the host without evident harm to either cell.

Figure 2: Steps of megakaryocyte emperipolesis.

(1) ICAM-1/ezrin cluster on the MK surface upon neutrophil attachment to facilitate interaction with neutrophil β2 integrins. (2) Neutrophils enter into MK vacuole (“emperisome”) in an actin-dependent manner. (3) The emperisome contracts around the neutrophil to enable close contact between neutrophil and emperisome membranes. (4) Neutrophils exit the emperisome and translocate to the DMS, where they transfer membrane and proteins to MKs and platelets. (5) Egress of viable neutrophil from MK (model adapted from (21)).

Intracytoplasmic localization serves an important function. Neutrophils fuse their membranes with the MK’s demarcation membrane system (DMS), the cytoplasmic network that forms the plasma membrane of future platelets, enabling transfer of membrane from the neutrophil to the MK and from the MK to the neutrophil. Donation of membrane is not terminal for the neutrophil, which instead exits the MK intact. The mechanism and topology of this egress remain to be determined. Whether all cells engaged in “emperipolesis” as conventionall employed in histological descriptions pass through the cytoplasm and donate membrane is not known. In vivo data as well as studies in human MKs establish that the essential elements of emperipolesis occur in living animals and across species, such that emperipolesis has now emerged as a novel form of cell-in-cell interaction distinct from phagocytosis, entosis, and transcellular migration (21).

The cell biology of emperipolesis

Time course

In 1970, Larsen observed a single granulocyte move around within an MK for 30 minutes before exiting (14). Using time-lapse spinning disk fluorescent microscopy, we confirmed that cell transit during emperipolesis typically occurs over 10–40min. In some instances, however, we could observe neutrophils remaining immobile within an MK for several hours, without apparent morphological compromise. Emperipolesis can involve single neutrophils or multiple at once; in extreme cases we observed more than 50 neutrophils within the same MK in vitro (21). This intriguing heterogeneity suggests that emperipolesis may not be a single uniform process but instead could be modulated to serve distinct purposes.

Cytoskeleton control

Microscopic observation shows that emperipolesis is a dynamic process wherein both cells participate actively (21). Neutrophils polarize toward the point of entry and demonstrate directed migration toward the MK, sometimes protruding podosomes into the MK before entry. The MK itself forms an actin-lined cup at the site of neutrophil engagement, similar to the transmigratory cup observed during transcellular migration (45, 46). Prominent actin polymerisation is observed within the MK, layered under the surface membrane at the point of contact with the neutrophil and surrounding the emperisome containing an internalized neutrophil (Figure 2 and (21)). Surprisingly, the tubulin cytoskeleton seems to have no essential role, since tubulin polymerisation is not observed around the internalized neutrophil and the tubulin inhibitor nocodazole does not inhibit emperipolesis (21).

Receptor-ligand pairs mediating emperipolesis

A process with the organization and selectivity of emperipolesis is likely to be mediated by specific receptor-ligand interactions. At present, the only MK receptor with a recognized role is intracellular adhesion molecule 1 (ICAM-1). In rats, immunohistochemistry found enhanced ICAM-1 staining in MKs engaged in emperipolesis compared with MKs not engaged, while internalized neutrophils expressed the ICAM-1 counterligand, the lymphocyte function-associated antigen1 (LFA-1, or CD18/CD11a) (47). Immunogold staining suggested expression of ICAM-1 within the MK, potentially in the DMS. Consequently, LFA-1 blockade reduced lipopolysaccharide (LPS)-induced emperipolesis in vivo (47). In the murine system, we observe ICAM-1 clustered on the MK surface where neutrophils attach, together with ezrin, the linker between ICAM-1 and the actin cytoskeleton (21). ICAM-1-deficient MKs and neutrophils lacking CD18 are less efficient at emperipolesis, while blocking either ezrin or ICAM-1 decreases murine emperipolesis in vitro. However, in each case inhibition is partial, implicating additional receptor-ligand pairs still to be defined (21). The role of ICAM-1/β2 integrin binding may contribute to tropism of emperipolesis for neutrophils, as well as the marked increase in emperipolesis upon LPS treatment of rats (18, 47) and mice (21), since LPS activates ICAM-1 and increases the surface expression and/or affinity of its ligands LFA-1 and Mac-1 (CD18/CD11b) (36).

CD62P is another receptor of interest. Also known as P-selectin, CD62P is a ligand for the neutrophil surface protein PSGL-1. CD62P is usually restricted to the α granule membrane and thus exposed on the surface of MKs and platelets only with activation. Interestingly, in patients with idiopathic myelofibrosis, CD62P expression is increased and distributed abnormally to the DMS (48–50). A similar phenotype is observed in murine models of myelofibrosis (48, 51). Emperipolesis is a hallmark of myelofibrosis (19, 49, 51). Indeed it has been proposed that abnormal distribution of CD62P during myelofibrosis traps neutrophils in the DMS via PSGL-1 binding (48, 51). In agreement with this hypothesis, deletion of CD62P in myelofibrotic GATA-1^low mice abrogates enhanced emperipolesis and rescues myelofibrosis (48). Whether CD62P participates in physiologic emperipolesis, where it is not normally exposed on the cell surface, or in emperipolesis associated with non-myelofibrotic disease, remains unknown.

To our knowledge, no other receptor / ligand pair has yet been implicated in emperipolesis. However, the prevalence of emperipolesis in pathologic states and the fine coordination observed between MK and neutrophils suggest that other pathways will emerge. For example, CD42b, a component of von Willebrand factor receptor, is also a counterreceptor for Mac-1 (52), and CD41, a component of the glycoprotein IIb/IIIa, binds β2-integrins in the presence of fibrinogen (53). Finding such new receptor/ligands pairs and defining physiological contexts in which they become expressed or activated will be critical for understanding and potentially manipulating emperipolesis.

Frequency of emperipolesis in health and disease

Emperipolesis in physiological states

Emperipolesis is routinely observed in healthy marrow of humans and other mammals. Human estimates based on stained marrow sections range from 3 to 7% (51, 54). In mice and rats, rates of 1–4% are typical(18, 19, 21, 47, 55), though up to 12–15% has been reported in studies employing marrow flush (56, 57). In the absence of LPS or other stress, most emperipolesis events involve one or two guest cells (47, 57, 58). Importantly, since MKs are much larger in diameter than the typical 5μm histological section, it is likely that a cross-sectional enumeration underestimates the prevalence of emperipolesis. Consistent with this possibility, we undertook 3-D immunofluorescence imaging of whole-mount murine femurs and found a rate of emperipolesis of ~6% (21). Emperipolesis has also been reported in MKs in spleen, although it has not yet been sought in MKs in lung (19, 57). Importantly, these percentages represent only one moment in time. Since emperipolesis is a brief and transient event, lasting in most cases 30 minutes or less, many neutrophils and MKs will likely be involved sequentially. Accurate estimates of the fraction of MKs and neutrophils that have engaged in emperipolesis are not yet available.

Emperipolesis can increase strikingly in disease. In humans, emperipolesis is abundant in myelofibrosis, where 10–20% of marrow MKs enclose leukocytes (51). An even higher prevalence is observed in the inherited macrothrompocytopenia gray platelet syndrome (GPS), with 36–65% of MKs containing leukocytes, often 2–4 per cell (59). GPS results from deficiency of NBEAL2 (neurobeachin like 2), a protein involved in normal development and/or retention of platelet α granules. Intriguingly, as in myelofibrosis, GPS MKs exhibit misdirected CD62P expression (59). Beyond these genetic syndromes, emperipolesis is increased in states associated with increased platelet demand, including myeloproliferative disease, myelodysplastic syndromes, polycythemia vera, reactive thrombocytosis, immune thrombocytopenia, essential thrombocythemia, and blood loss (54, 60–63).

Rodent models enable further exploration of these associations. Enhanced emperipolesis is observed in murine models of GPS-like disease (Nbeal2−/− and the gunmetal mouse bearing a spontaneous mutation affecting Rabggta) (64–67). Similarly, emperipolesis is prominent in murine models of myelofibrosis, including GATA-1^low mice and animals engrafted with marrow infected with a retrovirus carrying TPO cDNA, wherein half of MKs contain leukocytes (19, 49, 51). Emperipolesis increases with irradiation (56) and phlebotomy mimicking acute or chronic blood loss (68). As noted, emperipolesis is induced by inflammatory stimuli, including LPS injection as well as thrombocytopenia mediated by injection of IgG immune complexes in mice with platelets expressing human FcγRIIa (18, 21, 47, 58). Intriguingly, administration of TPO sufficient to double the platelet count, or depleting platelets by antibodies targeting murine CD41, do not by themselves enhance the frequency of emperipolesis, confirming that platelet demand alone is not a sufficient drive, at least in mice (21).

Together, these observations reveal emperipolesis to be a complex process that can be provoked in several ways. Extreme levels of emperipolesis arise through genetic disturbance of normal MK development and function, in particular when associated with aberrant distribution of CD62P to the DMS. Whether this reflects “trapping” of leukocytes internalized normally, or also enhanced cell uptake, is unknown. Under more typical (patho)physiologic conditions, emperipolesis arises under conditions where enhanced platelet demand is paired with systemic inflammation, potentially reflecting the role of activated integrins in the neutrophil half of the “emperipolesis equation”.

Potential consequences for emperipolesis

The observations that emperipolesis is preserved across species and tightly-regulated with environmental context strongly suggests that cells are doing this for a reason. An early speculation was that MKs provide a “sanctuary” for neutrophils in an unfavorable marrow environment, though it is not clear under which conditions such shelter should be necessary (55, 69). Others have suggested that emperipolesis arises because the space constraints of the marrow force cells into MKs through channels between the DMS and the cell surface (63). Indeed, very limited data have suggested that some ‘emperipolesed’ cells may be in a compartment accessible to a membrane-impermeable tracer, indicating a topological location contiguous with the extracellular environment (70). However, at least as a general explanation of emperipolesis, this hypothesis no longer appears compatible with the known cell biology (21). A further suggestion is that emperipolesis could be a route for blood cells from the marrow to the circulation, in particular since erythrocytes and their precursors are sometimes observed within MKs (68). Focusing in particular on MK emperipolesis of neutrophils, two teleological questions arise: why would an MK seek to internalize neutrophils, and why would a neutrophil seek to pass through an MK?

Possible impact of emperipolesis on MK function

Once the emperisome membrane dissolves, neutrophils reside in the MK cytoplasm (21). They are thus poised to release mediators directly into the MK. Whether this occurs in a physiological context is unknown, but its plausibility is enhanced by observations in “pathological” emperipolesis. In myelofibrotic patients and in murine myelofibrosis, neutrophil-specific myeloperoxidase-positive granules have been observed within the MK cytoplasm near emperipolised neutrophils (19, 49, 51). How CD62P in the DMS membrane encounters neutrophil surface P-selectin glycoprotein ligand-1 (PSGL-1) remains to be clarified (49, 51, 71). In agreement with the hypothesis that neutrophils release lytic granules within MKs, destruction of MK α-granules is often observed at an ultrastructural level in myelofibrosis, and neutrophil proteases have been implicated (19, 49, 51, 71). Interestingly, pathological release of MK α-granule proteins and associated MK growth factors such as TGFβ are critical in the pathogenesis of myelofibrosis (48). Further investigations are needed to establish emperipolesis as a direct contributor to the disease. Nevertheless, the possibility that release of neutrophil proteins (or lipids, or nucleic acids, or organelles such as mitochondria) during “healthy” emperipolesis serves the ends of the MK and of the organism as a whole constitutes an exciting hypothesis for future research.

An interesting feature of emperipolesis is the bi-directional transfer of membrane between neutrophils and MKs. In our model system, we observed continuity between neutrophil and DMS by electron microscopy, and were able to employ fluorescent lipid strainers, time lapse imaging, and chimeric mice to demonstrate conclusively that neutrophil membranes transfer to the DMS during emperipolesis (21). What is the physiological relevance of this membrane transfer? Whole-cell patch clamping finds that the DMS capacitance is equivalent to that 500–5,500 platelets (~1,600–17,000μm² of membrane), comparable with the surface area of a skeletal muscle cell (72). How the MK generates this mass of membrane is not fully understood. The DMS may arise as a specialization of endoplasmic reticulum or Golgi apparatus, via do novo membrane formation, or as an invagination of plasma membrane (73). While MKs cultured alone do develop DMS and elaborate proplatelets, platelet production in vitro is underwhelming compared to MK yields in vivo. It may be that emperipolesis allows other marrow cells to donate membrane to the DMS and thereby to platelets, a function that would necessitate cytoplasmic localization. Remarkably, we found both in cultured MKs and in in vivo MKs that donor cell membrane transfers via emperipolesis to MKs and thereby to platelets, though the fraction of platelets bearing such donor membrane remains unknown (21). Further, MKs co-cultured with donor cells produced substantially more platelets than control MKs, a phenotype that was independent of soluble mediators and so plausibly reflects emperipolesis itself (21). These findings suggest that emperipolesis serves to maximize platelet production, a possibility that fits well with the observation that emperipolesis is enhanced in states of high platelet demand. The role of emperipolesis in platelet production and function will be a fascinating area for future research.

Possible impact of emperipolesis on neutrophils and other marrow cells

Emperipolesis is increased in pathological states characterized by increased demand for marrow cell delivery to the periphery. For instance, emperipolesis increases after LPS treatment (58), phlebotomy (63) or irradiation (56). Marrow cells reach the circulation via the sinusoidal endothelium. MKs closely associate with this blood-marrow interface, localizing to the abluminal side of sinusoidal endothelium where they protrude cytoplasmic extensions through the vascular wall to release platelets into the circulation. MKs contribute further to the bone marrow environment by producing extracellular matrix and basement membrane components (74), and are thus plausibly regarded as an integral part of the blood / bone marrow barrier (75). A “trans-MK” route could be an efficient path for some marrow cells to enter the circulation (or potentially to re-enter the marrow from the blood). Consistent with this possibility, emperipolesis shares many mechanistic similarities with transendothelial migration, including utilization of the β2-integrin/ICAM-1/ezrin pathway, the actin cytoskeleton, and the formation of a transmigratory cup (10, 21, 46, 76, 77). While transit of cells between blood and marrow via MKs remains unproven, the growing understanding of emperipolesis lends plausibility to the suggestion of Tavassoli more than 30 years ago that emperipolesis serves a role in cell transport (68).

Ultrastructural images of emperipolesis show striking membrane ruffling on the surface of internalized neutrophils characteristic of macropinocytosis, especially when in the emperisome (78). Indeed, these engulfed neutrophils sometimes appear deeply invaginated by MK projections and can be seen to internalize MK content ((78) and personal observations). MK cytoplasm and α-granules are rich in soluble mediators that have the potential to impact neutrophil behavior (1). For instance, platelet factor 4 (PF4), TGFβ, and keratinocyte chemoattractant (KC) are MK mediators that promote neutrophil migration and survival (79–83). MKs also produce potent neutrophil activators including IL-1 (both α and β) and IL-6 (1, 8). Indeed, our preliminary data suggest that emperipolesis can enhance neutrophil migration in vitro and in vivo (84). Thus emperipolesis will likely modulate the properties of the transiting cells (we refer to this as “grooming”), just as the MK and its daughter platelets are themselves altered.

Summary and unanswered questions

Emperipolesis is a new type of cell-in-cell interaction characterized by several distinct features, at least with respect to MKs. These include (1) active cytoskeletal engagement by host and guest cell; (2) transit of the guest cell through the cytoplasm of the host, potentially enabling exchange of membrane and other cell components; and (3) egress of the guest cell without impairing the viability of either participant. Whether all cell-in-cell events currently termed ‘emperipolesis’ conform to all three features is unknown, even for cells inside MKs. Aside from the DMS, none of the cellular “machinery” of emperipolesis is unique to the neutrophil-MK dyad, suggesting the possibility that the phenomenon could occur more broadly. While it has been generally assumed that the one mammalian cell within another must necessarily reside in a membrane-bound compartment, emperipolesis shows that this principle is not invariable, raising the broader question of when else it might give way.

The study of emperipolesis is just beginning, and many questions remain to be addressed. What determines which cells participate, as host and guest? Which receptor-ligand pairs drive the interaction? What distinguishes an emperisome from a phagosome? How does the emperisome membrane disappear, and do some emperipolesis cells remain encased? What mediates fusion of the cytoplasmic guest cell with the DMS, and are other intracellular organelles sometimes also targeted? What consequences do intracytoplasmic cells have for the host cell – in particular, what is the implication of emperipolesis for platelet function? Is there “physiologic” and “pathologic” emperipolesis, from the point of view of the host cell and of the organism? How does a cytoplasmic guest egress without causing injury to the host? What is the impact of emperipolesis on the guest cell after egress? These questions and others will occupy our group and hopefully others for years to come.