Macrophages Discriminate Glycosylation Patterns of Apoptotic Cell-derived Microparticles*

Background: Apoptotic cells release vesicles, which expose “eat-me” signals. Results: Vesicles originated from endoplasmic reticulum expose immature glycoepitopes and are preferentially phagocytosed by macrophages. Conclusion: Immature surface glycoepitopes serve as “eat-me” signals for the clearance of apoptotic vesicles originated from endoplasmic reticulum. Significance: Understanding the distinction by macrophages of apoptotic blebs may provide new insights into clearance-related diseases. Inappropriate clearance of apoptotic remnants is considered to be the primary cause of systemic autoimmune diseases, like systemic lupus erythematosus. Here we demonstrate that apoptotic cells release distinct types of subcellular membranous particles (scMP) derived from the endoplasmic reticulum (ER) or the plasma membrane. Both types of scMP exhibit desialylated glycotopes resulting from surface exposure of immature ER-derived glycoproteins or from surface-borne sialidase activity, respectively. Sialidase activity is activated by caspase-dependent mechanisms during apoptosis. Cleavage of sialidase Neu1 by caspase 3 was shown to be directly involved in apoptosis-related increase of surface sialidase activity. ER-derived blebs possess immature mannosidic glycoepitopes and are prioritized by macrophages during clearance. Plasma membrane-derived blebs contain nuclear chromatin (DNA and histones) but not components of the nuclear envelope. Existence of two immunologically distinct types of apoptotic blebs may provide new insights into clearance-related diseases.

During apoptotic death the surfaces of apoptotic remnants, including apoptotic blebs, are modified due to oxidative processes resulting in the appearance of immunologically "novel" autoantigens (1). An inefficient clearance of apoptotic cell material (shrunken apoptotic cells, apoptotic bodies or subcel-lular membranous particles (scMP) 2 ) will result in the accumulation of secondary necrotic remnants that release modified and potentially pro-inflammatory contents into the extracellular milieu. It is common belief that inappropriate clearance of apoptotic cells can lead to the accumulation of modified autoantigens in tissues that foster autoimmune diseases (2).
The mechanism(s) modifying the glycocalyx are still elusive. We propose that (i) apoptozing cells may change the glycosylation of de novo synthesized glycoconjugates. (ii) internal membranes containing immature glycoproteins (GP) may become exposed and (iii) mature GP may get modified by glycosidases (e.g. sialidases) in situ.
To avoid confusion we introduced the term scMP for blebs that obviously had been released into the supernatant from the surfaces of apoptotic cells. The membranous structure formed by late apoptotic cells will be referred to as the big membranous bubble (bMB) (supplemental Fig. S1).
Here we demonstrate that apoptotic cells release distinct types of scMP derived from endoplasmic reticulum (ER) or plasma membrane. Both exhibit desialylated glycotopes resulting from surface exposure of immature ER-derived GP or from surface-borne sialidase activity, respectively. We addressed molecular mechanisms of GP redistribution during apoptosis and observed that sialidase activity is activated by caspase-dependent mechanisms. Plasma membrane-derived bMB and scMP contain chromatin but not the nuclear envelope and ER-derived scMP are prioritized by macrophages during clearance. The understanding of the immunological distinction by macrophages of ER and plasma membrane derived apoptotic blebs may provide new insights into clearance-related diseases.

EXPERIMENTAL PROCEDURES
Cell Culture and Isolation of Cells-We employed human leukemia Jurkat T-cells, human HeLa cells, and CD95 positive human breast cancer cells MCF-7, as well as primary human polymorphonuclear leukocytes (PMN) and monocyte-derived macrophages from healthy volunteers for our experiments. Peripheral blood mono-nuclear cells (PBMC) were isolated from peripheral anticoagulated blood by LymphoPrep gradient centrifugation according to the manufacturer's recommendations. Plasticadherent PBMC were then cultured for 7 days in the presence of GM-CSF (100 units/ml) and autologous serum (days 1, 3, and 5) to generate monocyte-derived macrophages. After 7 days of differentiation, these phagocytes usually contain more than 85% phagocytes positive for CD11b, CD14, and CD89.
Transfection of Cells-Transfection of HeLa cells with KDEL receptor-GFP construct was carried out as described previously (10). Transduction with GFP-linked N-acetylgalactosaminyltransferase-2 and SV40 nuclear translocation signal were performed using Organelle Lights TM (Invitrogen) Golgi and nuclear envelope GFP kits, respectively.
Microscopy-Confocal microscopy was performed using PerkinElmer Ultra-View spinning disc confocal on Nikon TE2000-E Ti inverted microscope using ϫ100 oil-immersion lens (excitation at 488, 568, 647 nm, detection at 650 nm, shown in red and 488 nm, shown in green). Fluorescent microscopy was performed using Zeiss AxioImager A1 with Zeiss AxioCam MRm digital camera. Some images were deconvolved using AutoQuant X2 (MediaCybernetics Inc). ImageJ (NIH) software None of these inhibitors prevented the redistribution of the apoptotic cell-specific glycotopes (Eto), indicating that conventional de novo synthesis is unlikely to be responsible for the exposure of an altered glycotope during apoptosis. Dots represent data for cell populations detected by flow cytometry or cytochemical analyses.
was used for co-localization analysis and image processing. Infocus rendering of DIC images was done with Helicon Focus. In all cases, microscopy was done observing at least 5 Petri dishes (9 cm) with cells grown to 70 -90% confluency for each variant. For quantification at least 1000 cells were counted. Experiments were repeated at least 3 times.
Determination of Trans-sialidase and Sialidase Activity-Cell trans-sialidase activity toward RBCs was determined using a test based on RBC agglutination with PNA lectin described elsewhere (14) and by their co-incubation at 37°C for 3 h with cells/conditioned medium or Clostridium perfringens (Sigma) sialidase as controls and standard, respectively.
Phagocytosis Assays-Phagocytosis of scMP was assessed by incubation of PMN-derived scMP with human monocyte-derived macrophages and uningested scMP stained with lectin(s) were analyzed by flow cytometry.
In silico analysis for transmembrane regions was done with HMMTOP and TMHMM. For the prediction of cleavage sites FIGURE 2. Apoptotic cells produce distinct types of scMP. a, apoptotic cells form scMP that are derived from the PM (red) or from the ER (green). 120Ј after irradiation with UV-B, only PM-derived blebs are generated. Later, ER-derived membranes can be observed on the surface of the cell (arrow) and on scMP (arrowhead), respectively. Finally a big PM-derived bubble (bMB) is formed from PM. b, HeLa cells were stained with ER-tracker™, and apoptosis was induced; after 3 h apoptotic cells were counterstained for immature ER-related glycotopes with CEL. scMP, positive for both ER-tracker and CEL, are indicated by arrows. During apoptosis, ER-related glycoconjugates are exposed on the cell surfaces. The intensely stained cell is a secondary necrotic one, exposing internal glycoepitopes. c, apoptotic cells (like B) were incubated in the medium containing the sialidase substrate 4-MUNA. A scMP positive for ER and negative for sialidase activity is indicated by arrows; a scMP negative for ER and positive for sialidase activity is indicated by arrowheads. During apoptosis sialidase activity is present on the PM and on PM-related scMP, but not on ER-related scMP. d, bMB arises from the PM (DiI staining, red, left panel) and is filled with chromatinderived antigens (here: histone 2B (middle panel)) in most of the transfected cells. The nuclear envelope, visualized by GFP-linked nesprin-1␣, is excluded form the bMB (five preparations utilizing 1000 cells each; right panel). a-d, non-permeabilized non-fixed cells were imaged. by caspases of sialidases, we employed GrabCas, CASVM, Pep-tideCutter, and CasCleave and set the cut-off scores Ͼ5,0 (GrabCas).

RESULTS
The Exposure of Galactose/Mannose on the Surfaces of Apoptotic Cells Is Independent of Proteinneogenesis-To analyze the mechanisms modifying the glycocalyx we blocked the synthesis pathway for the N-glycans of GP and gangliosides at several steps. Inhibitors of transcription [actinomycin D], translation [cycloheximide], N-glycan synthesis in the ER [tunicamycin, 2-deoxy-D-glucose], and export from Golgi to plasma membrane (PM), [monensin] did not prevent the increased exposure of galactose/mannose and decreased exposure of sialic acid (Fig. 1).
ER-derived Membranes Are Exposed in Late Apoptosis-Next, we transfected HeLa cells with an ER-residing KDEL receptor-GFP fusion construct and stained the PM with the fluorescent hydrophobic dye DiI. When we induced apoptosis with UVB, blebbing of the PM started 120 min after irradiation. After 240 min, the PM constricted and formed a polar conglomerate still connected to the cells' bodies. Concomitantly, ER membranes got exposed and started blebbing. In the supernatant of these cells, distinct scMP were detected that had originated from PM (red) or from ER (green), respectively (Fig. 2a). After 360 min, a bMB was formed that originated from the former PM (red). All blebbing cells produced PM-derived scMP, and most (Ͼ90,0%) of the late apoptotic cells demonstrated ER-derived scMP. However, analyzing about 1000 cells from seven independent preparations, we never observed double positive scMP in all our microscopic studies. ER-tracker positive blebs bound NPL (Narcissus pseudonarcissus lectin) and CEL (Canavalia ensiformis lectin), which preferentially recognize ER-related high mannose N-glycans and ER-specific terminally glucosylated N-glycan intermediates (16), respectively (Fig. 2b). Both NPL and CEL signal were co-localized with ER-tracker signal, as shown for permeabilized cells in supplemental Fig. S2.

Sialidase Activity and ER-Markers Are Mutually Exclusive-
We screened the cells for sialidase activity and observed a reduced surface sialylation and an increased exposure of galactosyl residues on the surfaces of apoptotic cells and their surface-derived scMP. The non-penetrating sialidase substrate 4-MUNA detected sialidase activity on the surfaces of apoptotic cells and of some of the scMP. Sialidase activity and ER-tracker staining of the scMP were mutually exclusive (Fig. 2c). Chromatin-derived histone 2B-GFP fusion proteins were translocated to the cytoplasm during apoptosis and was finally located inside the bMB (Fig. 2d) originating from the former PM (observed in most of the apoptotic cell remnants). Moreover, during late apoptosis nuclear components can mainly be included in apoptotic scMP, (observed in some few of the blebbing cells (supplemental Fig. S3)). After induction with etoposide of apoptosis the trans-sialidase activity of Jurkat cells toward red blood cell surface, GP was increased more than 20-fold (Fig. 3).
The Increase of Sialidase Activity in Apoptotic Cells Is Caspase-dependent-Treatment with caspases 3 or 8 of cell lysates from viable cells increased the sialidase activities to a level similar to cell lysates generated from apoptotic cells (Fig. 4a). In contrast, the caspase inhibitor zVAD inhibited the generation of sialidase activity in Jurkat cells treated with etoposide or in aged PMN (Fig. 4b). Caspase 3-deficient MCF-7 cells displayed no increased surface sialidase activity during apoptosis (induced by UV-B, etoposide or anti-CD95 Ab; Fig. 4c). Transfection of MCF-7 cells with caspase 3 restored the increase during apoptosis of sialidase activity (Fig. 4d). MCF-7 cells, lacking caspase-3 and increased sialidase activity during apoptosis produced ER-derived scMP upon treatment with UV-B, They also produce two distinct populations of scMP that were positive for ER-tracker or Dil, respectively ( Fig. 4e and supplemental Fig. S4).
In silico analysis revealed transmembrane regions and a predicted caspase 3 cleavage sites for both Neu1/4 and Neu1 (Uni-Prot: Q5JQI0, cleavage at Asp-135 with score 12,0 by GrabCas), respectively. As Neu2 is exclusively expressed in muscle and Neu3/4 has no high-score cleavage site for executer caspases we focused on Neu1 as the best fitting candidate among the four known human sialidases. Western blot analysis of the lysates from viable or apoptotic PMN showed a band most likely representing a cleavage product of the Neu1 protein in the latter (Fig. 5a). A similar fragment was to be observed after in vitro treatment of cell lysates with caspase 3 (not shown). The depletion of Neu1 abrogated the increased sialidase activity of apoptotic human PMN (Fig. 5b).
Isoelectric focusing of PM-enriched fractions revealed a caspase-dependent sialidase activity in the acidic fractions. The peak of the sialidase activity contained a polypeptide with the molecular mass (ϳ35 kDa) compatible with the C-terminal fragment of Neu1 (136 -415; predicted pI-5.61), which was reactive with a polyclonal antiserum against Neu1 (Fig. 5c).
Macrophages Prefer ER-derived scMP-To analysis of phagocytosis of the distinct types of scMP we employed human monocyte-derived macrophages as a validated model for clearance studies (7). scMP derived from aged human PMN served as "prey." We detected the staining (judged by MFI) of "uncleared" scMP, since it is less reliable to evaluate the amount of ingested scMP, which instantly become degraded inside the

Distinct Glycosylation of Apoptotic Blebs
phagocytes. The scMP, that are produced by macrophages were also quantified and served as control (despite their amount was extremely low). scMP gated by FSc/SSc data as shown on Fig. 6a were analyzed for fluorescence intensity after staining with lectins. When we co-incubated apoptotic ER-and PM-derived scMP from aged human PMN with human macrophages, ERderived scMP, endowed with immature ER-related oligomannosidic glycoepitopes, detected with the lectin NPL, were significantly faster cleared by macrophages than the PM-derived ones, exposing PM-related desialylated glycolepitopes characterized by terminal galactose or subterminal fucose residues detected with the lectins VAA (Fig. 6c) or UEA, respectively (Fig. 6d). This was not due to the size difference as macrophages do not display any marked size preference for the engulfment of scMP (supplemental Fig. S5).

DISCUSSION
Blockage of the synthesis pathway for N-glycans of GP and gangliosides suggests that de novo synthesis is unlikely to be responsible for the apoptosis-related surface-neoglycotopes. Fluorescence microscopy revealed the formation of distinct apoptotic scMP that originated from the PM or the ER and possess characteristic patterns of glycosylation. PM-and ERderived scMP expose desialylated and immature mannose-rich glycotopes, respectively. We demonstrated that sialidase activity is focused on the PM and on PM-derived scMP. This finding was also corroborated by staining with VAA (Viscum album lectin 1), a galactosyl-specific lectin, which exclusively bound to ER-tracker negative scMP derived from the PM (supplemental Fig. S6).
Employing a GFP-linked resident Golgi enzyme, N-acetylgalactosaminyl transferase-2 (17), we did not observe any exposure on apoptotic cell surfaces of Golgi-derived membranes (not shown). Furthermore, we demonstrated that chromatin-  . ER-derived blebs are preferentially engulfed by macrophages. a, size distribution of scMP population used for further study. b, MFI of scMP population, stained for ER-related glycans (NPL) or desialylated glycans (VAA) before (0 min) and 120 min after incubation with macrophages. A representative histogram is shown. c, blebs generated from apoptotic human PMN cells (24 h aging period), were co-incubated for the indicated time with human monocytederived macrophages. Un-ingested scMP were stained with lectins specific for PM-derived desialylated glycoepitopes (VAA, dots) or ER-related immature glycotopes (NPL, squares). Dots and squares represent the MFI (mean fluorescence intensity) value of total population of scMP, n ϭ 3; asterisks represent statistical significance compared with VAA staining. The scMP, that could be produced by macrophages were also quantified and served as control (despite their amount was extremely low). d, relative change of MFI of lectin binding to scMP population during different time intervals of incubation with macrophages. Ulex Europeaus Agglutinin I (UEA) binds subterminal fucosyl residues, exposed after desialylation, while NPL preferentially binds ER-derived glycoepitopes. The MFI of NPL signal decays more rapidly then that of UEA. The dotted line is a symmetrical line representing absence of preference in scMP engulfment. MFI values were normalized to the control. derived histone 2B-GFP fusion proteins are translocated to the cytoplasm during apoptosis and end up in a bMB (Fig. 2d) that originated from the PM. During late stages of apoptosis most chromatin is translocated into a bMB (supplemental Fig. S3). However, we did not observe scMP containing nesprin-1␣ the marker protein of the nuclear envelop.
Analyzing the following organells: endoplasmic reticulum, endosomes, Golgi, lysosomes, mitochondria, nuclear envelope, nucleus, and peroxisomes (all visualized via transfection with corresponding fluorescent marker proteins) we observed that only endoplasmic reticulum, endosomes, and nuclei are redistributed into the scMP. Additionally, we have observed some organelles (discriminated by DIC microscopy), inside the bMB.
Induction of apoptosis considerably increased a surface-associated sialidase activity, which was not detected in the cells' supernatants. This suggests the action of a cell surface-linked sialidase like Neu-3 (18) or the reportedly surface associated Neu-1 (19). Antisera directed against peptides of Neu-1 and Neu-3 diminished surface-associated sialidase activity (not shown).
Immunoprecipitation studies suggest that Neu-1 may at least partially be responsible for the increased sialidase surface activity observed during apoptosis. As shown by Western blot analyses, Neu-1 is cleaved by caspase 3. However, we do not yet know how (directly or indirectly) this cleavage leads to the activation of Neu-1.
The ability of MCF-7 cells lacking caspase-3 (and thus lacking increased sialidase activity during apoptosis) to produce both ER-and PM-derived scMP, suggests distinct molecular mechanisms involved in the formation of ER-derived scMP and in desialylation of glycoepitopes of PM-derived scMP. The former process was previously reported to be dependent on ROCK (10).
A plethora of receptors and adaptor proteins are involved in the uptake by macrophages of apoptotic cells and their subcellular particles (reviewed in Ref. 20). In general they either recognize phosphatidylserine or altered carbohydrates. Just recently the RAGE receptor has been added to the list of phosphatidylserine-dependent receptors enhancing the efferocytosis of apoptotic cells (21).
Both nucleotides (22) and lipids (23) were shown to act as "eat-me" and "find-me" signals in promotion of apoptotic cell clearance. The surface exposure of the latter was shown to be dependent on caspase 3. Changes of the surface charge (24) as well as carbohydrate content (25) are essential for the clearance of apoptotic cells. Recently we have shown that artificial desialylation of viable cell surfaces creates an "eat-me" signal for macrophages (7).
In phagocytosis assays, human macrophages cleared ER-derived scMP (possessing immature ER-related glycoepitopes) significantly faster than the PM-derived ones (possessing PMrelated desialylated glycoepitopes) (Fig. 6). This may result in a differential immunological processing of these subcellular particles and their associated antigens, the size of scMP has had a minor effect on their clearance speed (supplemental Fig. S5).
Intriguingly, nucleosomes, a prototypic autoantigen targeted by sera of patients with systemic lupus erythematosus (SLE), were mainly concentrated in the bMB (Fig. 2D). This is of particular importance since SLE patients have been reported to frequently display a deficient clearance of apoptotic cells (2).
Vaccinia and other viruses have been shown to mimic PMderived, phosphatidylserine (PS) exposing scMP, utilizing the ability of surrounding cells for macropinocytosis of PS-exposing particles. This pathway augments the infectivity of the viruses, especially that for phagocytes (26,27).
In conclusion, we propose a mechanism including alterations of the glycocalyx during apoptosis and provide evidence that reduced sialylation of cells undergoing apoptosis can be caused by both surface exposure of immature ER-derived GP and surface-bound sialidase activity cleaving mature GP in situ. Our data indicate that these processes are mutually executed by previously undifferentiated types of apoptotic scMP (Fig. 7). ERderived scMP are prioritized by macrophages during apoptotic cell clearance. Our data are in a good agreement with previous observation where calreticulin exposed from ER during apoptosis has been found to dictate the immunogenicity of cancer cell death (28).