Peritoneal carcinomatosis is an advanced form of metastatic disease characterized by cancer cell dissemination onto the peritoneum. It is commonly observed in ovarian and colorectal cancers and is associated with poor patient survival. Novel therapies consist of cytoreductive surgery in combination with intraperitoneal chemotherapy, aiming at tumor cell death induction. The resulting dying tumor cells are considered to be eliminated by professional as well as semi-professional phagocytes. In the present study, we have identified a hitherto unknown type of ‘amateur’ phagocyte in this environment: human peritoneal mesothelial cells (HMCs). We demonstrate that HMCs engulf corpses of dying ovarian and colorectal cancer cells, as well as other types of apoptotic cells. Flow cytometric, confocal and electron microscopical analyses revealed that HMCs ingest dying cell fragments in a dose- and time-dependent manner and the internalized material subsequently traffics into late phagolysosomes. Regarding the mechanisms of prey cell recognition, our results show that HMCs engulf apoptotic corpses in a serum-dependent and -independent fashion and quantitative real-time PCR (qRT-PCR) analyses revealed that diverse opsonin receptor systems orchestrating dying cell clearance are expressed in HMCs at high levels. Our data strongly suggest that HMCs contribute to dying cell removal in the peritoneum, and future studies will elucidate in what manner this influences tumor cell dissemination and the antitumor immune response.
Peritoneal carcinomatosis is frequently observed in patients suffering from colorectal and ovarian cancers (van Grevenstein et al., 2007; Slack-Davis et al., 2009). It is initiated by tumor cells exfoliating into the abdominal cavity and subsequently adhering to human peritoneal mesothelial cells (HMCs) that line the intra-abdominal cavity and organs (Mutsaers, 2004). This can result in the formation of intraperitoneal metastases and underlines the importance of HMCs in the development of peritoneal carcinomatosis (van Grevenstein et al., 2007; Slack-Davis et al., 2009; Alkhamesi et al., 2005). The median survival period of patients with peritoneal carcinomatosis rarely exceeds 6 months (Levine et al., 2007). Hence, novel therapeutic strategies are currently being developed that target the prolongation of life expectancy. Often combined with surgery, all these strategies have the same goal: induction of tumor cell death. Cytoreductive surgery followed by hyperthermic intraperitoneal chemotherapy, during which chemotherapeutics are applied locally into the abdominal cavity, has been reported to increase the median survival period substantially (to an average of 22 months) (Levine et al., 2007). For example cisplatin derivates, commonly applied against ovarian tumors, are known to induce cell death and enhance the basic rates of apoptosis and necrosis in cancer cells (Vakifahmetoglu et al., 2008; Bryant et al., 2010).
The current notion is that dying cells in the abdominal cavity are eliminated by professional phagocytes: the peritoneal macrophages (Park et al., 2009; Frey et al., 2009; Gregory and Devitt, 2004). However, aside from macrophages, as the prototype of professional phagocytes, numerous examples of semi-professional, so-called ‘amateur’ phagocytes exist, including fibroblasts, endothelial, epithelial and mesenchymal cells (Gregory and Devitt, 2004; Dini et al., 1995; Cao et al., 2004; Dogusan et al., 2004). These are neighboring cells and have a rather limited capacity of phagocytosis but do contribute to the removal of cell corpses from their surroundings because they commonly outnumber professional phagocytes (Gregory and Devitt, 2004; Monks et al., 2005).
The phagocytic clearance of apoptotic cell corpses is known to involve a multitude of engulfment receptor systems (reviewed by Lauber et al., 2004; Mevorach et al., 1998). On the one hand, apoptotic cells can be recognized directly by surface receptors of the phagocyte. This is typically mediated by externalized phosphatidylserine on the apoptotic cell surface and phagocytic phosphatidylserine receptors, including BAI1, Tim1, Tim4 and Stab2, or scavenger receptors, such as SR-A1, SR-B1, LOX1, CD68 and CD36. On the other hand, there are phagocytic receptor systems that require the participation of soluble bridging proteins functioning as opsonins. Examples are the vitronectin receptor (αvβ3 integrin) with the bridging proteins TSP1 or MFG-E8, the Mer tyrosine kinase family (Mer, Tyro3 and Axl) with the bridging proteins Gas6 or protein S (ProtS), the low density lipoprotein-related proteins LRP2 and LRP8 with the bridging protein β2GPI, and the members of the collectin or complement receptor family [calreticulin (CD91) and αMβ2 or αXβ2 integrin] with the corresponding complement and collectin proteins (C1q, C3, MBL2, SP-A and SP-D). Bridging proteins often derive from body fluids, including serum and surfactant, but they can also be autonomously produced by the phagocytes themselves.
In the present study, we have analyzed the role of HMCs as semi-professional phagocytes in an in vitro model of peritoneal carcinomatosis in which confluent HMC monolayers were incubated with ovarian and colorectal tumor cell suspensions at varying ratios over different time periods. By employing flow cytometry, and confocal and electron microscopy, we show for the first time that HMCs phagocytose corpses of apoptotic and necrotic cancer cells. We show that HMCs internalize fragments of dying cells, by engaging an active actin cytoskeleton, and subsequently traffic the internalized prey to late phagolysosomes. HMCs clearly distinguished between viable, apoptotic and necrotic tumor cells; the highest rate of phagocytosis was observed when HMCs were co-incubated with apoptotic prey cells. Regarding the underlying mechanisms of prey cell recognition, our results show that HMCs engulf apoptotic prey cells in a serum-dependent as well as -independent fashion. Finally, quantitative real-time PCR (qRT-PCR) was employed to analyze the expression levels of different engulfment genes.
In conclusion, our study demonstrates that HMCs can act as amateur phagocytes that internalize dying tumor cell corpses. Given that HMCs are far more predominant, in terms of numbers, than peritoneal macrophages, we propose that HMCs play a pivotal role in the removal of dead cells from the abdominal cavity.
HMCs engulf tumor cell particles
In order to study the role of HMCs in peritoneal carcinomatosis, with special regard to the interaction between HMCs and cancer cells, we have established an in vitro model, in which primary HMCs [CellTrace Far Red DDAO-succinimidyl ester (FR)-labeled; red] were co-cultured with ovarian SKOV-3 or colorectal HT29 tumor cells [carboxyfluorescein succinimidyl ester (CFSE)-labeled; green]. The adherent tumor cells were enzymatically detached and added, in suspension, to confluent HMC monolayers to mimic peritoneal metastasis at the point of tumor cell spread. After 48 hours of incubation in the presence of 15% fetal calf serum (FCS) and 5% human serum (HS), cells were fixed and analyzed by confocal microscopy (Fig. 1). Cancer cells and HMCs could be clearly distinguished in the single-channel measurements (Fig. 1A) and displayed close cell-to-cell contacts (Fig. 1B,C). Surprisingly, we observed scattered green SKOV-3 and HT29 cell fragments within the red HMCs (Fig. 1B,C). The z-stacks unambiguously demonstrated that these green fragments (1 to 25 μm in diameter and generally smaller than whole tumor cells) indeed were inside the HMCs, suggesting that they had been engulfed. By contrast, uptake of HMC fragments into tumor cells was not detected.
Preparations of primary HMCs always contain small fractions of contaminating peritoneal macrophages (in the present study consistently <5%) (Stylianou et al., 1990). To exclude the possibility that these macrophages were responsible for the observed phagocytic phenomenon, we examined the uptake of SKOV-3 cell fragments by HMCs after magnetic macrophage depletion by live cell imaging (supplementary material Movie 1). SKOV-3 cells were again incubated with HMCs for 24 hours and photographs were taken at 15-minute intervals for a period of 14 hours. Over time, several SKOV-3 cells rounded up and displayed blebbing features. These structural changes typically appear during apoptosis (Mills et al., 1999). Most importantly, the video sequence clearly revealed two red HMCs engulfing green tumor cell fragments, which were constricted from the SKOV-3 cytoplasm, supporting the notion that in our co-culture system some SKOV-3 cells undergo apoptosis (~10–15%; data not shown) and subsequently are phagocytosed by HMCs.
Electron microscopy analysis of pseudopod formation by HMCs
Our confocal microscopy data strongly favor the idea that HMCs phagocytose (dying) tumor cell fragments (Fig. 1). Therefore, we next analyzed the anatomy of this engulfment process. During phagocytosis, scavenger cells form protrusions of the plasma membrane, so-called pseudopods (Booth et al., 2001). Hence, we examined the cell-to-cell contacts between HMCs and HT29 cells, after 48 hours of co-incubation, by transmission electron microscopy (Fig. 2A,B) and compared these contacts with intermesothelial cell contacts (Fig. 2C,D). As clearly shown, HMCs formed cellular extensions, which reached far into the cytoplasm of the HT29 cells, similar to the formation of pseudopods (Fig. 2B; HMC, red; HT29, green). The intercellular contacts formed between two HMCs were fundamentally different, as the cellular extensions only touched the cell surfaces (Fig. 2C,D).
Internalized tumor cell fragments colocalize with phagosome markers in HMCs
The degradation of ingested cellular material involves phagosome maturation through homotypic and heterotypic fusion of early endosomes, late endosomes and lysosomes (Zhou and Yu, 2008). The classical markers employed for the discrimination of these different phases are Ras-related protein 5 (Rab5), a small GTPase associated with early phagosomes (Kinchen et al., 2008; Desjardins et al., 1994), and annexin I (Anx1) or lysosomal-associated membrane protein 1 (LAMP1), two proteins that are utilized to show later phagosomal stages and phagosomes after lysosomal fusion, respectively (May and Machesky, 2001; Alvarez-Dominguez et al., 1997; Tjelle et al., 2000). We examined the colocalization of phagocytosed tumor fragments with the above-mentioned phagosomal marker proteins by confocal microscopy (Fig. 3). Tumor cells were co-incubated with macrophage-depleted HMCs, and intracellular staining for Rab5 (Fig. 3A,B), annexin I (Fig. 3C,D) or LAMP1 (Fig. 3E,F) was performed after 24 hours (Rab5) or 48 hours (annexin I, LAMP1). The micrographs clearly showed that HMCs (stained with CellTracker Orange CMRA; blue) internalized tumor cell fragments (FR; green) and that these fragments intracellularily colocalized with the Rab5 signal [fluoresceinisothiocyanate (FITC)-labeled; red] (Fig. 3A,B). These findings were illustrated even more clearly by the histograms displaying the distribution of tumor-cell- and Rab5-derived fluorescence intensities. In the case of annexin I and LAMP1, we observed an accordingly convincing association of the marker proteins with the internalized tumor cell material (Fig. 3C–F). Here, the signals for annexin I (FITC; red) and LAMP1 [phycoerythrin–cyanine 5 (PE–Cy5); red] in fact surrounded the ingested tumor cell fragments, as the maximal fluorescence intensities for both protein stainings always paralleled the increase or decrease in the tumor-cell-derived CFSE fluorescence. These results reflect the anatomical structure of phagosomes where the membrane-associated proteins surround the tumor cell fragments in the phagosomal lumen. Note, all the antibodies employed specifically stained their target proteins; in stainings with the corresponding isotype controls no considerable fluorescence signal was detected (data not shown).
Phagocytosis of tumor cells by HMCs is inhibited by cytochalasin D
After having examined the structural characteristics of HMC-mediated phagocytosis, we utilized a flow cytometric assay to quantify the process of phagocytosis. Prior to co-incubation, HMCs and tumor cells were labeled with FR and CFSE, respectively. Both cell populations could be clearly discriminated by FACS analysis (HMCs, gate 1, red; SKOV-3, gate 3, green; Fig. 4A, upper row). After 48 hours of co-incubation, FR-labeled HMCs were positive for tumor-cell-specific CFSE-fluorescence as reflected by their considerable shift on the x-axis. Assuming that this shift displays HMCs that have internalized tumor cell fragments (Figs 1 and 3), this population was defined as phagocytosing HMCs (gate 2, highlighted in turquoise; Fig. 4A, lower row). From the FACS data alone, it cannot be ruled out that the right-hand half of gate 2 also contained SKOV-3 cells that had phagocytosed HMCs. However, we never observed uptake of HMC fragments into tumor cells in our confocal microscopy analyses (Fig. 1). Note that doublet exclusion was performed on the basis of FSC-A and FSC-H. Hence, the detection of FR and CFSE double-positive cells could not be due to an artifact of cell pairing. The population of double-positive cells could be further subdivided into phagocytosing HMCs (gate 4, turquoise) and peritoneal macrophages positive for CD11c (PE) and/or CD86 [Pacific Blue (Pablue)] (gate 5, purple) (Fig. 4A, dot plot on the lower right-hand side). Notably, the percentage of these contaminating macrophages was always below 5% and they were excluded from further analyses. The gating strategy, and color display of the different cell populations (red, HMCs; green, tumor cells; turquoise, phagocytosing HMCs; and purple, phagocytosing macrophages), described here, was accordingly applied to the HT29 and HMC co-culture system (Fig. 4C) and was used throughout the whole study.
As displayed in Fig. 4B, ~30% of HMCs phagocytosed SKOV-3 cell fragments (Fig. 4B). Phagocytosis was significantly and dose-dependently reduced in the presence of cytochalasin D, an actin polymerization inhibitor widely used to substantiate the involvement of active cytoskeletal rearrangements in the context of phagocytosis (P<0.001) (May and Machesky, 2001; Akya et al., 2009). Similar effects were detected for the co-incubation of HMCs with HT29 cells (Fig. 4C), although the total percentage of phagocytosing HMCs was slightly lower.
The above data confirm that HMCs indeed phagocytose tumor cell fragments. Nevertheless, the ‘quality’ of the ingested tumor material still remained elusive, although the video data (supplementary material Movie 1) suggested that it was apoptosing, rather than living, tumor cells that were being engulfed. To clarify this issue, viable and apoptotic [staurosporine-treated with ~60% of cells annexin V (AnxV)–FITC-positive and propidium iodide (PI)-negative] tumor cells were co-incubated with HMCs, and the percentage of phagocytosing HMCs was assessed by flow cytometry. Fig. 4D depicts unambiguously that HMCs preferentially ingested fragments of apoptotic rather than viable HT29 cells; the same was true for SKOV-3 (dot plots not shown). Again, the addition of cytochalasin D strongly inhibited phagocytosis (Fig. 4D–F).
Note that, irrespective of the nature of prey cells (i.e. apoptotic or viable), HMCs internalized profoundly less prey cell fluorescence than CD11c- and/or CD86-positive macrophages (Fig. 4A,D; turquoise population compared with the purple population). This parallels our microscopical findings that the phagocytic capacity of HMCs is limited to subcellular fragments, whereas macrophages are well-known to engulf whole prey cells (Aderem, 2002). However, in the co-incubations with SKOV-3 cells a substantial number of HMCs were detected in the right-half of gate 2 suggesting that more prey cell fluorescence was taken up. Hence, it is reasonable to assume that more HMCs phagocytosed whole SKOV-3 cells than was the case with HT29 prey cells. Nevertheless, the percentage of HMCs that internalized only fragments of dying cells was always higher than the percentage of HMCs that engulfed whole cells.
HMC-mediated phagocytosis of viable, apoptotic and necrotic tumor cells: target ratio analysis
The results presented in Fig. 4 have shown that HMCs engulf fragments of apoptotic tumor cells far more potently than fragments of viable ones, suggesting that HMCs contribute to efferocytosis, the process of removing dying cells (Thorp, 2010). Under physiological and pathophysiological conditions, different forms of cell death occur (Vakifahmetoglu et al., 2008; Byrant et al., 2010). In this scenario, apoptosis and necrosis represent the extremes of programmed and accidental cell death, respectively, on a scale with a multitude of possibilities in between (Galluzzi et al., 2007; Kinchen and Ravichandran, 2007). Therefore, we next analyzed, by flow cytometry, the engulfment of viable, dying and dead cells by HMCs at different target ratios. AnxV–FITC and propidium iodide (PI) staining was employed to confirm the viable, apoptotic and necrotic status of untreated, staurosporine-stimulated or heat-treated tumor cells, respectively (Fig. 5A). FACS analysis of prey cell engulfment convincingly revealed a positive correlation between the percentage of phagocytosing HMCs and the amount of target cells applied (Fig. 5B,C). Most importantly, fragments of apoptotic prey cells were much more efficiently engulfed than fragments of necrotic or viable cells. It is reasonable to assume that, in the case of living HT29 cells, the contaminating 12% of spontaneously dying cells, rather than the living cells, were the ones that were being phagocytosed (Fig. 5A). Notably, the ‘quality’ of viable, apoptotic, and necrotic cells did not significantly change over the time of co-incubation (48 hours) as measured by FACS analysis of the FSC/SSC profiles and PI exclusion (data not shown). Hence, HMCs apparently discriminate between living, dying and dead tumor cells and display a strong predilection for apoptotic cells in terms of engulfment.
We next examined further the phagocytosis of other different apoptotic cell types by HMCs. In these studies, we included ovarian (OvCar-29), colorectal (CaCo-2) and leukemic tumor cell lines (Jurkat and THP-1), as well as primary human neutrophils. As displayed in supplementary material Fig. S1, all types of apoptotic cells were internalized. However, there seemed to be a preference of HMCs for ovarian cancer cells and neutrophils, as they were more efficiently engulfed than the other cell types tested (supplementary material Fig. S1).
Phagocytosis of apoptotic tumor cells: time dependency
So far, the data shown have demonstrated that HMCs can contribute to efferocytosis. However, our finding that their phagocytic capacity is limited to subcellular fragments rather than whole cells (Figs 1and 4) underlines their ‘semi-professional’ nature (Dini et al., 1995). Therefore, we next analyzed the kinetics of apoptotic corpse internalization by HMCs in comparison with that of peritoneal macrophages over a period of up to 48 hours. The percentage of phagocytosing HMCs increased continuously over time (Fig. 6A,B). However, in contrast to peritoneal macrophages, which phagocytosed apoptotic cells very quickly (Fig. 6C,D), HMCs engulfed apoptotic tumor cells with a rather slow kinetics – an observation that further substantiates their amateur phagocyte character (Parnaik et al., 2000). The number of phagocytosing HMCs in the co-incubation with apoptotic SKOV-3 cells was consistently higher than in the co-culture system with apoptotic HT29 cells (see also Figs 4, 5), as was the case for phagocytosing macrophages. Because the percentage of apoptotic cells in both prey cell populations did not differ substantially (70–90%, as determined by AnxV–FITC and PI staining; Fig. 5A and data not shown), these data indicate that the phagocytic potential of HMCs also depends on the target cell type.
Serum dependency of HMC-mediated efferocytosis
The phagocytic clearance of apoptotic cells is known to involve a multitude of engulfment receptor systems. Among them, there are receptors recognizing apoptotic cells directly (typically mediated by externalized phosphatidylserine on the apoptotic cell surface), whereas others require the participation of soluble bridging proteins (often serum-derived), functioning as opsonins (Lauber et al., 2004; Mevorach et al., 1998). The intriguing question that needed to be addressed at this stage was which receptor systems would be engaged in HMC-mediated efferocytosis. To this end, we examined the internalization of apoptotic tumor cell fragments in the absence or presence of serum. HMCs engulfed apoptotic cells under serum-free conditions, suggesting that either opsonin-independent receptor systems play a role or that HMCs autonomously produce the required bridging proteins (Fig. 7A,B). However, because addition of serum significantly enhanced the uptake of apoptotic corpses (P<0.001), HMCs apparently also utilize receptor systems involving serum-derived opsonins. Decomplementation, by heat-inactivation of serum, did not profoundly interfere with its phagocytosis-enhancing effect. Hence, serum-derived complement proteins are obviously dispensable for HMC-mediated efferocytosis. In order to exclude that the serum-free culture conditions negatively affected HMC viability, we monitored HMC viability by AnxV–FITC and PI staining. Importantly, cultivation in the absence of serum did not substantially induce HMC apoptosis within 48 hours (Fig. 7C–E). These results indicate that HMCs employ serum-dependent, as well as serum-independent, engulfment receptor systems for efferocytosis, with serum-derived complement proteins apparently playing a minor role.
Expression profiling of engulfment genes in HMCs
Finally, we examined the expression levels of different engulfment genes in HMCs (obtained from four different donors) by qRT-PCR. Intriguingly, mRNA transcripts for various known receptors could be detected. The most abundantly expressed transcripts, with average threshold cycle (Ct) values <25, belonged to the scavenger receptors (SR-B1, LOX1 and CD68), the MFG-E8 and TSP1 receptors (αvβ3 or αvβ5 integrin), the Gas6 and ProtS receptors (Axl and Tyro3), the β2-GPI receptors (LRP8), and the collectin receptor (CD91, also known as calreticulin). High transcript levels (Ct values <25) for the corresponding bridging proteins MFG-E8, TSP1, ProtS and Gas6 were also found in HMCs (supplementary material Fig. S2). For relative quantification, target gene mRNA levels were normalized to that of 18S rRNA and calibrated to the normalized expression levels of the respective gene in macrophage colony-stimulating factor (M-CSF)-differentiated peripheral blood monocyte-derived macrophages, which served as a positive control. This analysis yielded three different groups of engulfment genes (Fig. 8A): genes expressed in HMCs in comparable amounts to macrophages, such as the collectins (MBL2, SP-A and SP-D) and the collectin receptor (CD91); genes expressed in HMCs less frequently (factor 2−5 to less than 2−10) than in macrophages, like the complement receptors or the Fc gamma receptors; and genes expressed in HMCs at much higher levels (factor 22 to more than 210) than in macrophages, including Axl, Tyro3, αvβ3 integrin and the respective bridging molecules Gas6, ProtS, TSP1 and MFG-E8. We conclude, from these results, that it is particularly the members of the latter group that are the most probable candidates to orchestrate HMC-mediated efferocytosis.
To analyze whether HMCs, in principle, might be able to degrade and/or export the internalized cell fragments, we also examined DNaseII, the cathepsins B and D and ATP-binding cassette, sub-family A, member 1 (ABCA1) in our qRT-PCR analyses. As demonstrated in supplementary material Fig. S2, and Fig. 8B, all these genes were expressed at high levels (average Ct values <25), which is approximately comparable to the levels found in macrophages.
The timely and efficient removal of dying cells, efferocytosis, is crucial for tissue development, homeostasis and protection against chronic inflammation and autoimmunity. Macrophages, as the prototypical professional phagocytes, play a central role in this process (Gregory and Devitt, 2004; Lauber et al., 2004). Nevertheless, the contribution of amateur phagocytes to dying cell clearance can be demonstrated by two examples. First, in lower-order macrophage-deficient organisms, such as the nematode worm Caenorhabditis elegans, dying cells are engulfed by neighboring cells (Gregory and Devitt, 2004). Second, mice lacking macrophages develop ostensibly normally (Wood et al., 2000), strongly supporting the notion that non-professional phagocytes assist in the clearance of apoptotic cells when macrophages are absent or when their phagocytic capacity is overwhelmed (Parnaik et al., 2000).
Here, we have identified a hitherto unknown type of amateur phagocyte in the context of dying cell removal: human peritoneal mesothelial cells (HMCs). By utilizing an in vitro model of peritoneal carcinomatosis, we show that HMCs phagocytose corpses of apoptotic and necrotic ovarian SKOV-3 and OvCar-29 cancer cells, and colorectal HT29 and Caco-2 cancer cells, as well as other types of apoptotic cells. Flow cytometry, confocal and electron microscopy have revealed that this internalization process is dependent on an active actin cytoskeleton and that it is followed by the typical sequence of phagosome maturation (Figs 1, 2, 3 and 4). These observations add a new facet to HMC biology. Being derived from the mesoderm and lining the serous cavities of the body, mesothelial cells are known to exert a variety of functions: the production of lubricating fluid, thus providing a non-adhesive and protective surface to facilitate intracoelomic movement; the transport of fluid and cells across the serosal cavities; and acting as a physical barrier and first-line of defense against invading pathogens and harmful substances (Mutsaers, 2004; Mutsaers, 2002). In the context of the latter process, it has been shown that HMCs are able to phagocytose pathogens, such as Staphylococcus aureus, as well as asbestos fibers (Brinkmann and Muller, 1989; Haslinger-Löffler et al., 2006). Hence, our observation that HMCs internalize fragments of dying tumor cells complements these studies and extends the spectrum of HMC function to the field of efferocytosis.
Although the present study has clearly revealed that HMCs can distinguish between viable, apoptotic and necrotic tumor cells (Fig. 5), the large proportion of these cells ingested only subcellular fragments and not whole cells, in contrast with the action of peritoneal macrophages (Figs 1, 4). This is in line with the observation that other non-professional phagocytes, including liver endothelial cells, also preferentially engulf apoptotic bodies (Dini et al., 1995). Moreover, the internalization process followed a far slower kinetics than macrophage-mediated efferocytosis (Fig. 6), a finding that has accordingly been reported for other types of amateur phagocytes (Parnaik et al., 2000). The molecular mechanisms underlying this delay remain elusive. Parnaik et al. suggested that non-professional phagocytes, upon encountering dying cells, first have to acquire the competency to phagocytose, which might be accomplished during an extended period of binding to their targets (Parnaik et al., 2000). Nevertheless, our qRT-PCR data have clearly revealed that (at least on the mRNA level) different engulfment genes are readily expressed before HMCs encounter their prey. Hence, it can be hypothesized that either the translation of different engulfment systems or their trafficking to the cell surface are the crucial steps that are responsible for the delayed engulfment kinetics. Concerning the limited phagocytic capacity of HMCs in terms of engulfable prey size, one could speculate that it was related to a restricted availability of degradative enzymes, such as DNase II and members of the cathepsin family (Yoshida et al., 2005), or ABCA1, a transporter that is vital for the export of internalized prey-cell-derived cholesterol (Kiss et al., 2006). However, our qRT-PCR analyses have demonstrated that HMCs express these enzymes at mRNA levels that are higher than the levels found in macrophages (Fig. 8B). Hence, insufficient expression of degradative enzymes is unlikely to be the cause for the amateur phagocyte character of HMCs.
The question that arises at this point is: what is the contribution of HMCs to dying cell removal in vivo when, compared with peritoneal macrophages, their capability to engulf dying cells is much more limited and follows a rather slow kinetics? Obviously, in the peritoneal cavity HMCs are far more predominant in terms of numbers. Furthermore, Gregory et al. have suggested that macrophages serve exclusively as a cellular ‘backup’ compartment for professional efferocytes, which will only be recruited should the local semi-professional phagocytic system be overwhelmed (Gregory and Devitt, 2004). In line with this hypothesis, our data support the assumption that HMCs do contribute to the clearance of dying cell corpses in the peritoneal cavity and future in vivo studies will clarify this issue in detail.
Macrophages identify apoptotic prey cells by employing a variety of sensor systems. Within the engulfment synapse, ‘eat-me’ signals displayed on the apoptotic cell surface are recognized by corresponding phagocyte receptors, either directly or indirectly under participation of soluble bridging proteins (Gregory and Devitt, 2004; Lauber et al., 2004). Regarding the mechanisms underlying HMC-mediated prey cell recognition, our results show that HMCs potently engulf apoptotic cell corpses in the absence of serum, suggesting that they either utilize opsonin-independent engulfment receptors or that they autonomously produce soluble bridging factors themselves. qRT-PCR analyses in fact revealed that HMCs express high levels of several scavenger receptors (SR-B1, LOX1 and CD68), as well as the receptors for MFG-E8 and TSP1 (αvβ3 integrin), and Gas6 and ProtS (Axl and Tyro3), with the corresponding opsonins MFG-E8, TSP1, Gas6 and ProtS (Fig. 8A). Our observation that the addition of serum significantly enhanced the uptake of apoptotic cells by HMCs (Fig. 7) is in line with the qRT-PCR finding that HMCs express high levels of the ProtS receptors (Axl and Tyro3), the collectin receptor CD91-calreticulin and one of the β2-GPI receptors (LRP8). Decomplementation of serum did not considerably reduce the serum-mediated phagocytosis-promoting effect, which can be explained by the qRT-PCR observation that complement receptors are only weakly expressed by HMCs. Taken together, our data support the notion that serum-independent HMC-mediated efferocytosis is coordinated by the MFG-E8-TSP1-αvβ3 or αvβ5 integrin system, the Gas6-ProtS-Axl-Tyro3 system and several scavenger receptors (SR-B1, LOX1 and CD68), whereas serum-dependent HMC-mediated efferocytosis is governed by Axl, Tyro3, the collectin receptor CD91 and the β2-GPI receptor LRP8. Of note, previous reports have identified αvβ3 and αvβ5 integrin, CD36 and LOX1 as crucial contributors in amateur phagocyte-mediated efferocytosis (Hughes et al., 1997; Oka et al., 1998; Walsh et al., 1999; Sexton et al., 2001).
In summary, our findings provide proof-of-principle evidence that HMCs indeed phagocytose dying ovarian and colorectal tumor cells and thereby might contribute to the clearance of dead cancer cells in the peritoneal cavity. This is of special importance as HMCs are known to secrete various chemokines and cytokines involved in leukocyte recruitment and/or activation, and there is accumulating support for a role of HMCs in antigen presentation (Mutsaers and Wilkosz, 2007). Thus, future studies are required in order to clarify how far HMC-mediated efferocytosis influences tumor cell dissemination into the peritoneal cavity and how it modulates the anti-tumor-immune response, especially during chemotherapeutic intervention.
Materials and Methods
Cell isolation and culture
HMCs were isolated from greater omentum tissue obtained from consenting patients undergoing elective surgery. This method was approved by the ethics committee of the Faculty of Medicine and University Hospital at the University of Tübingen, Germany. The greater omentum samples were stored in isotonic saline solution for 30 minutes transportation time, followed by directly continued processing. The isolation procedure was based on the method described by Stylianou et al. (Stylianou et al., 1990). After enzymatic disaggregation with type II collagenase (Invitrogen), HMCs were grown on fibronectin-coated dishes (Sigma–Aldrich) in medium containing M199 supplemented with 20 mmol/l HEPES pH 7.4, 2 mmol/l glutamine (Invitrogen), 5% (v/v) human serum (HS), endothelial cell growth supplement (150 μg/ml) (both PromoCell, Heidelberg, Germany), 15% (v/v) fetal calf serum (FCS), 100 IU/ml penicillin, 100 μg/ml streptomycin (all PAA, Pasching, Austria) and 5 IU/ml heparin (Ratiopharm, Ulm, Germany). HMCs displayed a uniform cobblestone-like appearance at confluence and were tested for expression of intracellular cytokeratin 18 (mouse anti-human-cytokeratin-18–FITC; clone DC10, Abcam), surface staining of ICAM-1 (mouse anti-human-ICAM-1–FITC; clone 1H4, Abcam) and VCAM-1 (mouse anti-human-VCAM-1–PE; clone STA, BioLegend, San Diego, California, CA) using flow cytometry (data not shown). Mouse IgG2b and mouse IgG1 (clones MPC-11 and MPC-21, respectively; BioLegend) served as isotype controls.
The following tumor cell lines were chosen for this study: SKOV-3 and OvCar-29 (subclone of OvCar-3) ovarian cancer cells, and HT29 and Caco-2 colorectal cancer cells, as well as the leukemic tumor cell lines Jurkat and THP-1. The cell lines were purchased from the American Type Culture Collection (ATCC; HTB-77, HTB-161, HTB-38, HTB 37, TIB-152 and TIB-202, respectively). They were cultured in Iscove's modified Dulbecco's medium (Lonza) supplemented with 10% FCS (v/v), penicillin (100 IU/ml) and streptomycin (100 μg/ml). If not stated otherwise, all experiments were performed in M199 medium supplemented with 15% FCS and 5% HS but lacking endothelial cell growth supplement, which did not affect cell viability as monitored by Trypan Blue staining.
Magnetic depletion of CD11b-positive macrophages was performed with mouse anti-human-CD11b magnetic micro beads (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instructions. Successful depletion was ensured by anti-human-CD11b–PE staining (clone ICRF44, BD Biosciences) and flow cytometry. HMCs were only used if the CD11b-positive fraction was <0.5%. Prior to further processing HMCs were allowed to recover for 24 hours.
HMCs of the first passage were grown to confluence on Lab-Tek chambered 1.0 borosilicate coverglass systems (Thermo Scientific) after labeling with CellTrace Far Red DDAO-SE (FR) or CellTracker Orange CMRA (CMRA, both from Invitrogen) as recommended by the manufacturer. A total of 1×106 HMCs were stained in 1 ml PBS containing 8 μM or 25 μM dye, respectively.
Tumor cells were labeled with CFSE or FR (both from Invitrogen) according to the manufacturer's recommendations. Briefly, CFSE staining was conducted in 0.5 ml PBS containing 12 μM or 16 μM dye for 1×106 HT29 or SKOV-3 cells, respectively. FR staining was performed in 1 ml PBS containing 5 μM dye (HT29) or 10 μM dye (SKOV-3) for 1×106 cells in each case. Subsequently, labeled tumor cells were added in suspension to the confluent HMC monolayers (ratio 1:1). After 24–48 hours co-incubation, cells were gently washed, fixed in PBS containing 1% formaldehyde (Fluka) and, in some experiments, subjected to further antibody staining. After permeabilization with Cytofix/Cytoperm solution (BD Biosciences) for 20 minutes antibody staining was conducted at 4°C for 30 minutes in Perm/Wash buffer (BD Biosciences) with anti-human-LAMP-1–PE–Cy5 antibody (clone eBioH4A3; eBioscience, San Diego, CA), anti-human annexin I antibody (clone 29), anti-human-Rab5 mouse IgG2a antibody (clone 1/Rab5) (both from BD Biosciences) and the corresponding secondary antibody (goat anti-mouse-IgG–FITC antibody, Jackson ImmunoResearch). Isotype controls were stained in parallel. Confocal microscopy was performed using an inverted LSM 510 confocal laser scanning microscope (Carl Zeiss) endowed with a Plan-Apochromat 63× 1.4 NA oil objectives. CFSE and FITC were exited at 488 nm with an argon-ion laser and emission was detected from 505 to 530 nm using bipolar filter sets. CMRA and PE–Cy5 were exited at 543 nm and FR at 633 nm with helium-neon lasers. Emission was detected with appropriate filtersets: PE–Cy5 and FR at wavelengths longer than 650 nm and CMRA at 585 nm. The photographs were taken at 630× magnification (ocular ×10, objective ×63). z-stacks were taken in slices of 1–2 μm. Two-color images were acquired by single-track measurement. Three-color images were taken in the multi-track mode. All images were analyzed using LSM 5 Image Examiner (Carl Zeiss).
HMCs from the first to third passage were used for the experiments. HMCs were stained with the FR dye (4 μM), and SKOV-3 and HT29 cells with the CFSE dye (6 μM or 5 μM, respectively) as described above. HMCs and tumor cells were either incubated separately or in co-culture at different ratios over the indicated times. In some incubations, cells were pretreated with 1, 10 or 100 μM cytochalasin D (Calbiochem) for 1 hour. Dimethyl suloxide (DMSO) served as a vehicle control.
After co-incubation all cells were detached with accutase (PAA) and subjected to antibody staining at 4°C for 20 minutes with anti-human-CD86–Pacific-Blue (clone IT2.2) and anti-human CD11c-PE (clone 3.9, both from BioLegend). After staining, cells were fixed in 1% formaldehyde and examined using the FACSCanto II flow cytometer. The data were analyzed with FACSDiva software 6.0 (both BD Biosciences). Contaminating macrophages were identified by CD11c and/or CD86 positivity (see Fig. 4).
Induction of tumor cell death
Tumor cells were stimulated to undergo apoptosis by incubation with 5 μM staurosporine (Sigma–Aldrich) for 12 hours. Necrosis was induced by heat-treatment at 56°C for 20 minutes. Phosphatidylserine (PtdSer) exposure (with AnxV–FITC, BD Biosciences) and loss of membrane integrity (with PI, Sigma-Aldrich) were stained in accordance with the manufactures' guidelines and examined using a FACSCalibur flow cytometer (BD Biosciences). The data were analyzed with Summit version 4.1 software (Dako Cytomation, Glostrup, Denmark). Apoptotic cell populations comprised 70–90% of the AnxV–FITC-positive, PI-negative cells, whereas necrotic cell populations comprised 70–90% of the PI-positive cells (Fig. 5).
HT29 cells were incubated for 48 hours on confluent HMC monolayers. Owing to the clear structural differences between HMCs and HT29, only the colorectal cancer cell line was used for this experiment. Cells grown on coverslips were fixed in 2.5% glutaraldehyde for 2 hours and treated with 1% osmium tetroxide (both Sigma–Aldrich) in PBS for 90 minutes on ice. Thereafter, samples were stained with 1% aqueous uranyl acetate (Sigma–Aldrich) and dehydrated in a graded series of ethanol before infiltration and embedding in epoxy resin (Sigma–Aldrich). Before ultrathin sectioning, coverslips were removed from the cell layer by dipping the polymerized sample into liquid nitrogen. The sections were viewed in a LEO 906 transmission electron microscope.
Quantitative real-time PCR analysis
Quantification of engulfment gene mRNA levels was performed by qRT-PCR analysis with an ABI Prism 7000 Sequence Detection System (Applied Biosystems) and Maxima SYBR Green qPCR Mastermix (Fermentas St. Leon-Rot, Germany) as described previously (Peter et al., 2008). The primer pairs employed (Sigma–Aldrich) are listed in supplementary material Table S1.
Total RNA from approximately 2×106 HMCs from four different donors was extracted with the NucleoSpin RNA II-kit (Macherey & Nagel, Dueren, Germany). 1 μg of total RNA was reversely transcribed with 200 units RevertAid reverse transcriptase (Fermentas) in the presence of 50 μM random hexamers (GE Healthcare), 400 μM dNTPs (Promega, Heidelberg, Germany), and 1.6 units/μl RibolockRNase inhibitor (Fermentas). 20 ng of the resulting cDNA was applied for the following qRT-PCR analyses (20 μl final volume) with 300 nM primers. Relative quantification was performed according to the ΔΔCt algorithm with 18S rRNA serving as housekeeping control and cDNA from M-CSF-differentiated monocyte-derived macrophages as calibrator. Results are presented as log2 expression ratios.
Flow cytometry experiments were conducted at least three times in triplicate. HMCs were always obtained from three different individuals. The experimental group was analyzed using Student's t-test. Findings were considered statistically significant at P≤0.001 (*).
We thank Bernd Knöll (Department of Molecular Biology, University of Tübingen, Germany) and Karin Blume (Department of Internal Medicine I, University of Tübingen, Germany) for excellent technical advice. This work was supported by the fortune funding (1793-0-1) of the Medical Faculty, University of Tübingen, Germany and the SFB 685, Department of Immunology, University of Tübingen, Germany. The authors declare no conflict of interests.