Background
Most neurological diseases involving neuronal loss are accompanied by the appearance of activated microglia in the affected tissues [
1]. Microglia are resident brain macrophages that mediate the immune response against CNS infections and clear cellular debris following injury. However, there is growing evidence that inflammatory-activated microglia actively participate in the death of neurons during neurodegenerative processes, for example through release of reactive oxygen and nitrogen species (ROS/RNS) and release of pro-inflammatory neurotoxic cytokines and inflammatory mediators such as TNF-α [
2]. We have recently described a novel form of neuronal death mediated by inflammatory-activated microglia in which microglia phagocytose viable neurons, referred to as ‘primary phagocytosis’ or ‘phagoptosis’ [
3,
4]. Phagocytosis is normally thought to occur after the target cell has undergone cell death, but we found that in inflammatory conditions inhibition of phagocytic signalling rescues neurons both
in vitro and
in vivo, demonstrating that phagocytosis can be a direct cause of neuronal death in models of inflammatory neurodegeneration [
5‐
7].
Phagocytosis is controlled by a complex array of signals. The interaction between a number of ‘eat-me’ and ‘don’t-eat-me’ signals located on the target cell surface and their respective receptors on the phagocyte determine whether or not phagocytosis takes place [
8]. The best-characterised ‘eat-me’ signal is exposure of the phospholipid phosphatidylserine (PS) on the outer leaflet of the plasma membrane. In most viable cells that are not activated, PS is almost exclusively localised on the inner leaflet of the plasma membrane due to an aminophospholipid translocase that pumps PS from the outer to the inner leaflet. Upon induction of cell death by apoptosis or necrosis, PS becomes exposed on the cell surface due to inactivation of the translocase or activation of a scramblase, which randomises phospholipid distribution between the inner and outer leaflets thus resulting in net PS exposure. However, PS exposure also occurs on the surface of viable cells when ‘activated’, usually as a result of calcium stimulation of the scramblase and inhibition of the translocase, for example during activation of all leucocytes [
9‐
11], and on neurons exposed to oxidants from activated microglia [
5]. Exposed PS can be either bound directly by some phagocyte receptors, such as Tim4, stabilin-1 and −2 and BAI1, or bound by bridging proteins such as MFG-E8, which activates phagocytosis via the vitronectin receptor (α
vβ
3/5 integrin) [
8]. Indeed we have shown that primary phagocytosis of viable neurons by inflammatory-activated microglia is mediated by microglia-induced PS exposure on viable neurons, evoking microglial phagocytosis via MFG-E8 and the vitronectin receptor [
5,
7].
Surface-exposed calreticulin (CRT) has been demonstrated to act as an eat-me signal in a number of cell types [
12]. CRT, principally characterized as an endoplasmic reticulum (ER)-resident chaperone, is constitutively expressed at the surface of numerous cancer cell lines and its expression at the cell surface can be increased in the early stages of apoptosis induced by a subset of apoptotic stimuli including anthracyclins and UV irradiation [
13,
14]. CRT has been shown to act as an essential eat-me signal promoting phagocytosis of apoptotic cells and its activity can be modulated not only by increasing exposure at the cell surface but also potentially by rearrangement of existing exposed CRT [
15,
16]. Surface-exposed CRT is recognised by the phagocytic receptor LRP (low-density lipoprotein receptor-related protein) [
15,
17], although CRT is also found associated with LRP on the phagocyte membrane where it acts as a co-receptor for LRP ligands such as C1q and alpha-2-macroglobulin [
18]. The constitutive expression of CRT on the surface of a number of cell types does not necessarily result in their phagocytosis as don’t-eat-me signals have a dominant inhibitory effect on phagocytosis, for example CD47 and its receptor SIRPα [
13,
15,
19,
20]. The role and regulation of exposed CRT and LRP in mediating phagocytosis of neurons by microglia is unknown. We therefore sought to test the requirement for CRT- and LRP-mediated signalling for phagocytosis of dying neurons and primary phagocytosis of viable neurons in models of inflammatory neurodegeneration. Here we demonstrate that neuronally exposed CRT is required as an eat-me signal for phagocytosis of both apoptotic and viable neurons by microglia, and that CRT is constitutively exposed on the surface of neurons but this only promotes phagocytosis in specific contexts, and indeed released CRT can inhibit phagocytosis at microglia.
Methods
All experiments were performed in accordance with the UK Animals (Scientific Procedures) Act (1986) and approved by the Cambridge University Local Research Ethics Committee.
Cell culture and treatments
Mixed neuronal/glial cerebellar cultures were prepared from the cerebella of postnatal day 5 to 7 rats as previously described [
21] and were allowed to mature
in vitro for six to eight days prior to treatment. Pure microglia were prepared from mixed cortical astroglial/microglial cultures as previously described [
5]. BV2 microglial cells were grown in Dulbecco’s modified Eagle’s medium (DMEM, Invitrogen, Carlsbad, CA, USA)) supplemented with 10% fetal bovine serum (FBS, PAA Laboratories, Colbe, Germany). PC12 neuronal cells were grown in Roswell Park Memorial Institute medium (RPMI, Invitrogen) supplemented with 10% FBS and 20% horse serum (Sigma-Aldrich, St Louis, MO, USA). PC12 were plated on collagen-coated tissue culture plates (0.5 mg/ml collagen, Sigma-Aldrich). All tissue culture medium was supplemented with 100 units/ml penicillin G, 100 μg/ml streptomycin sulphate (Invitrogen). Reagents were procured as follows: lipopolysaccharide (LPS), calreticulin (CRT), cytochalasin D (CytoD), 5-(and-6)-carboxytetramethylrhodamine succinimidyl ester (TAMRA) were from Sigma-Aldrich. β 1–42 monomers (EZBiolab, Carmel, IN, USA) were prepared as previously described [
6]. Receptor-associated protein (RAP, R&D systems, Minneapolis, MN, USA), normal mouse IgG (mIgG, Santa Cruz Biotech, Santa Cruz, CA, USA), anti-CRT antibodies (Abcam, Cambridge, UK; Stressgen, Brussels, Belgium), anti-LRP (American Diagnostica Inc., Stamford, CT, USA), Alexa 488-labelled isolectin B4 (IB4, Molecular Probes, Eugene OR, USA). Neuronal and microglial cell survival was quantified three days after stimulation as previously described [
5]. Anti-CRT and anti-LRP blocking antibodies were Fc-blocked with an F(ab’)2 fragment antibody (Jackson Immunoresearch, West Grove, PA, USA). Nitrite levels in culture supernatants were measured as previously described [
5].
BV2 and PC12 phagocytosis assay
BV2 were plated in 6-well plates in DMEM plus 0.5% FBS and were at approximately 60% confluency when target cells were added. PC12 in suspension were stained for 10 minutes with 50 μM TAMRA, washed in warm PBS and then plated in 10 cm collagen-coated dishes at high density. UV-treated PC12 received 200 mJ/cm
2 irradiation. Untreated and UV-treated PC12 were harvested 16 hours after UV treatment by trypsinisation. PC12 target cells were counted and resuspended in DMEM plus 0.5% FBS. Some 200,000 PC12 target cells were added to each well of BV2 (approximate four-fold excess of target PC12 cells compared to BV2) followed by a two-hour incubation at 37°C. For FACS analysis, BV2 were stained with IB4 (1 μg/ml) for 15 minutes prior to washing in PBS and brief trypsinisation to detach cells. BV2 were then resuspended in 200 μl PBS and FACS analysis performed using an Accuri C6 Flow Cytometer (BD Services, San Jose, CA, USA). Alexa 488 IB4 fluorescence was detected in FL1 channel whilst TAMRA fluorescence was detected in FL2. For fluorescence microscopy, BV2 were labeled with IB4 as above and washed briefly in PBS prior to labelling of nuclear DNA with Hoechst 33342 [
5]. Cells were imaged on an Olympus Fluoview 300 microscope (Olympus, Tokyo, Japan).
Transwell and microglial reconstitution experiments
Following six to seven days in vitro microglia were selectively eliminated from cerebellar cultures by adding 50 mM L-leucine methyl ester (LME, Sigma-Aldrich). After three hours LME-containing medium was aspirated, neurons washed once in warm HBSS (Invitrogen) and then medium was replaced with conditioned medium from sister cultures. Twenty-four hours later, 6.5 mm 0.4 μm pore size polycarbonate transwell inserts (Corning, Sigma-Aldrich) that had been poly-L-lysine coated were inserted and 25,000 microglia were plated onto the insert. After 24 hours, LPS was added at 100 ng/ml as indicated in figure legends. Forty-eight hours later, transwell inserts containing microglia were removed. During this time microglia were purified and plated in 6-well plates, left for 24 hours and then incubated for a further 24 hours with 100 ng/ml LPS. LPS-activated pure microglia were gently blown from wells after a brief incubation in Versene solution (Invitrogen) at 37°C. At this point, blocking antibody was added to the neurons or to purified LPS-activated microglia in suspension for one hour at 37°C. Neurons were washed three times in warm HBSS before conditioned medium from untreated sister cultures was added back. LPS-activated microglia were washed three times in warm HBSS, collected by centrifugation and counted. A total of 25,000 LPS-activated microglia was then added directly back to neuronal cultures as indicated and plates were spun briefly to allow microglia to settle. Cerebellar cultures were then incubated for six hours at 37°C before neuronal survival and numbers were assessed as described above. For addition of exogenous CRT, microglia were eliminated from cerebellar granule cells (CGC) that had been in vitro for seven days using LME as before. After 24 hours, 1 μg/ml CRT (Sigma-Aldrich) was added directly to neurons and left to incubate for two hours, followed by three washes in warm HBSS. Pure microglia that had either been left untreated or LPS activated for 24 hours as described above were then added back to neurons as indicated at a density of 25,000 cells per well (24-well plate), plates incubated for six hours at 37°C prior to quantification of neuronal number and survival.
Externalised protein biotinylation and pull-down
Surface biotinylation was performed using the Pierce Cell Surface Protein Isolation Kit including Sulfo-NHS-SS-Biotin as the labelling reagent (Thermo Fisher Scientific, Waltham, MA, USA). Cerebellar cultures were seeded in 6-well plates and after six days in culture microglia were eliminated with LME as described above. After 24 hours, a poly-L-lysine-coated transwell was inserted and 200,000 purified microglia were plated on the transwell. Twenty-four hours later, 100 ng/ml LPS was added where indicated. After 48 hours, further incubation transwells were removed and neuron plates were transferred to ice and washed several times in HBSS prior to addition of Sulfo-NHS-SS-Biotin. All subsequent steps including streptavidin pull-down of biotinylated proteins followed the manufacturer’s instructions. Neurons were lysed in a volume of 500 μl and prior to pull-down a load sample of 40 μl was collected. Biotinylated proteins captured on streptavidin beads were eluted by boiling in SDS gel loading buffer. Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes for western blot detection as previously described [
22].
Statistical analysis
Statistical analysis was performed using SPSS software. All results represent the mean value from at least three separate experiments (see figure legends) with each individual experiment having two replicates per condition with four fields counted per replicate. In figure legends n = x refers to the number of separate experiments performed. Error bars represent the standard error of the mean of experiments (SEM). Normality of data was verified using the Shapiro-Wilk test. Data was analysed using one-way ANOVA and post hoc Bonferroni test. In figures * = P < 0.05, ** = P < 0.01, *** = P < 0.001.
Discussion
In this study we demonstrate that the CRT/LRP system is required for primary phagocytosis of viable neurons by microglia, so that inhibition of this system could prevent neuronal loss and death induced by LPS or Aβ. Given the increasing evidence supporting a neurodegenerative role for microglia, this system might potentially play a role in loss of neurons during inflammatory neurodegenerative processes such as brain infections (for example, AIDS dementia), ischaemia (for example, stroke), inflammation (for example, multiple sclerosis), trauma or neurodegeneration (for example, Alzheimer’s or Parkinson’s disease) [
3,
26]. However, this requires further investigation in relevant models of disease. We found that RAP can prevent neuronal loss induced by LPS or Aβ, and thus in principal might be therapeutically useful, at least in acute, life-threatening brain pathologies such as stroke, trauma or meningitis. RAP is normally expressed in the brain but declines in Alzheimer’s disease [
27], which might in principle contribute to the neuronal loss.
We have previously shown that microglia activated by LPS or Aβ induce neuronal loss and death by phagocytosis of otherwise viable neurons, and this ‘primary phagocytosis’ required microglia-induced PS exposure by neurons, bound by MFG-E8, which induced phagocytosis of the neurons via microglial vitronectin receptors [
5‐
7]. PS exposure on viable neurons was induced by peroxynitrite production by microglia and phagocytosis of neurons occurred independently of apoptosis [
5,
7]. In the present work, we have found that the CRT/LRP pathway plays a permissive role for this induced primary phagocytosis, such that if the CRT/LRP pathway is blocked primary phagocytosis can not proceed. The finding that blocking this phagocytic pathway with CRT antibodies, LRP antibodies, RAP or free CRT results in the accumulation of live rather than dead neurons, again supports the notion that the neurons are lost by primary phagocytosis, rather than phagocytosis secondary to the neurons dying by some other means.
We also found that CRT/LRP was important for BV2 microglial phagocytosis of apoptotic PC12 cells. If this is true for primary neurons and microglia
in vivo, then blocking this system may be detrimental in a variety of physiological and pathological conditions by allowing dead neurons to accumulate and promote inflammation. However, for cancer cells, it has been shown that macrophage phagocytosis of CRT-exposed cancer cells is immunogenic as the macrophages present antigens from these cells [
14]. This immunogenic phagocytosis of CRT exposing dead cells might be beneficial in the context of brain tumours or brain infections, but potentially detrimental in other contexts such as MS or development. It is possible that the CRT/LRP system contributes to the phagocytosis apoptotic and viable neurons arising during development. In
C.elegans mutation of a number of the
Ced genes (including
Ced1, a proposed functional homologue of LRP) involved in phagocytosis of dead cells in combination with a weak
Ced3 (caspase) mutation results in survival of cells normally eliminated in the presence of the weak
Ced3 mutation alone [
28,
29]. CRT knockout is lethal in mice, and intriguingly 16% of CRT-null mice develop exencephaly of the brain characterised by failure to close the neural tube, a process involving programmed cell death [
30,
31].
Our data are consistent with a model in which CRT acts as an eat-me signal on the neuronal surface and is recognised by LRP on the phagocytic membrane, as has been described in other non-neuronal systems [
15,
17]. LRP is known to be expressed and functional on microglia [
32]. Through use of surface biotinylation, we demonstrated that CRT is constitutively expressed on the surface of cerebellar neurons. A previous report from Hossain and colleagues demonstrated external localisation of CRT on rat hippocampal neurons where it co-localised with NMDA receptor and potentially played a role in modulating Ca
2+ influx into neurons [
33]. The amount of CRT was unchanged when neurons were primed for phagocytosis by inflammatory-activated microglia. Studies in non-neuronal cell types have shown that CRT-dependent phagocytosis does not necessarily require increased CRT exposure. In some cell types surface-exposed CRT accumulates in patches on the cell surface, often in association with exposed PS [
15,
16,
34]. CRT-dependent phagocytosis can also be triggered by a reduction in don’t-eat-me signalling by CD47/SIRPα signalling [
13,
15,
19]. However, we tested a CD47-blocking antibody and found that this had no effect on phagocytosis of viable or apoptotic neuronal cells in the presence or absence of inflammatory stimuli (data not shown). Gardai and colleagues demonstrated that CRT knockout prevented phagocytosis of PS-exposing apoptotic cells, and that readdition of exogenous CRT restored phagocytosis although this phagocytosis remained PS-dependent [
15]. Similarly, in our model we have shown that LPS-activated microglia induce PS exposure on viable neurons and that PS recognition by MFG-E8 is required for phagocytosis to proceed [
5,
7]. Thus whilst CRT is not increased on the surface of viable neurons during primary phagocytosis, the exposure of PS induced by LPS-activated microglia may be sufficient to allow phagocytosis that occurs in a CRT-dependent manner. We demonstrated that incubation of viable neurons with exogenous CRT allowed primary phagocytosis to proceed in the presence of activated but not unactivated microglia. It is possible that other types of neuropathologically relevant stimuli may cause increased neuronal CRT exposure and, therefore, may cause neurodegeneration by primary phagocytosis in this way.
We found that addition of CRT to microglia or to both microglia and neurons could block microglia-induced neuronal loss. This may be because free CRT can activate microglial phagocytosis/endocytosis via LRP in the absence of bound neurons/target cells (as occurs in macrophages [
15]), resulting in downregulation of surface LRP and associated phagocytic machinery. This type of mechanism might be involved in the neuroprotective effect of peptide Y-P30. Y-P30 can bind CRT and was reported to cause release of extracellular CRT in SH-SY5Y cells, as well as dissociation of CRT from membranes isolated from rat cortex. In conjunction with this activity, Y-P30 inhibited the appearance of microglia
in vivo following lesioning of the cortex in rat [
35].
Competing interests
The authors declare they have no competing interests.
Authors’ contributions
MF performed surface biotinylation, primary tissue culture and treatments, helped design the study and drafted the manuscript. MJOM performed apoptotic phagocytosis assays. GCB conceived and directed the study and helped draft the manuscript. All authors read and approved the final manuscript.