All cell culture reagents were from Invitrogen (Paisley, UK), unless otherwise indicated. Culture treatments were procured as follows: staurosporine (from Streptomyces sp.) and recombinant human calreticulin were from Abcam (Cambridge, UK); peptide synthesized human amyloid-beta (1-42) was purchased from Anaspec (Fremont, CA); adenosine 5’-triphosphate (ATP) disodium salt hydrate, lipopolysaccharide (LPS; from Salmonella Enterica serotype typhimunium), polymyxin B sulfate salts and tunicamycin (from Streptomyces sp.) were from Sigma-Aldrich (St Louis, MO, USA); recombinant mouse LRPAP-1 protein was from R&D Systems (Minneapolis, USA).

The immortalized cell lines BV-2 (ECACC Cat# 0356, RRID: CVCL_0182) and CHME3 (ATCC Cat# CRL-3304, RRID: CVCL_II76) were maintained as previously described (22, 23). Neither cell line is listed as a commonly misidentified cell line by the International Cell Line Authentication Committee. Primary microglial and mixed neuron-glial cultures were respectively prepared from the cortex and cerebellum of 3-5 day old Wistar rats (Charles River, RRID: RGD_2312511), following procedures described elsewhere (24, 25). All animal experiments were approved by the Cambridge University Local Research Ethics Committee and undertaken in accordance with the UK Animals (Scientific Procedures) Act (1986).

Cells were treated as follows: calreticulin was used at 1 nM, 10 nM, 100 nM or 170 nM for 20 hours; 10 nM, 17 nM or 50 nM for 24 hours; 10 nM or 50nM for 48 hours; and 2.5 nM, 10 nM or 50nM for 72 hours as indicated. Monomeric Aβ was used at 250 nM over 72 hours. LPS was added at 100ng/mL over 20 or 24 hours where indicated. Calreticulin or LPS were pre-treated with 10 U/mL polymyxin B for 1 hour where indicated. Polymyxin B was used at 10 U/mL for 20 hours. LRPAP-1 was used at 100 nM for 20 hours. Tunicamycin was used at 0.2 and 2 µg/mL for 24 hours. Staurosporine was used at 10 nM and 100 nM for 24 hours, and at 1000nM for 6 or 24 hours as specified. ATP was added at 10 µM for 24 hours.

Cell viability, defined as the percentage of non-necrotic cells, was measured at indicated endpoints by differential dye uptake of propidium iodide (identifying necrotic cells) and Hoechst 33342 (identifying all cells) using a fluorescent microscope (EVOS M5000). Alexa Fluor 488-tagged isolectin B4 was used to identify microglia. Apoptotic cells were recognized as non-necrotic cells with nuclear condensation (identified with Hoechst 33342). For pure microglial cultures, the entire well was imaged, and the number of cells quantified using QuPath (version 0.3.0). Two wells were quantified per experiment, with the number of experimental replicates indicated in the figure legends. For neuronal-glial co-cultures, four microscopic fields were quantified per well, with two wells per experiment, and three experimental replicates.

BV-2 migration was measured by transwell (Boyden) chamber assay using COSTAR 24-well plate 8.0µm pore membrane inserts (Corning). Briefly, 2 x 10⁴ BV-2 cells were seeded into serum-free DMEM on the top chamber membrane. In the bottom chamber, serum-free DMEM, supplemented with calreticulin or ATP, was added. Migration was assessed across 24 hours under incubating conditions (37°C, 5% CO₂). At experimental endpoints, residual microglia found on top of the membrane were removed using a cotton bud. Microglia that had migrated to the underside of the membrane were fixed in 4% paraformaldehyde, washed in PBS, and stained with Hoechst 33342. Migrated microglia were imaged using the EVOS M5000 (4x Magnification, DAPI Channel).

Calreticulin release from BV-2 microglia and mixed neuron-glia were measured by quantitative western blot. Briefly, 5 x 10⁴ BV-2 cells were cultured in serum-free DMEM and treated accordingly. Neuron-glial culture medium was replaced with PBS supplemented with proteinase inhibitors, and cultures were compressed with the flat end of a syringe plunger for one minute to rupture the cells. Proteins from the conditioned mediums of treated cultures were resolved on NuPAGE 4-12% Bis-Tris gels (Invitrogen) and transferred onto PVDF membranes. Blots were probed with a polyclonal antibody to calreticulin (1:500, EnzoLifeSciences) followed by IRDye^® 800CW donkey anti-rabbit IgG secondary antibody (1:5000, Licor). Antibody binding was detected using the LI-COR Odyssey^® CL_X. Quantification of calreticulin loaded was attained by comparing band intensity values for each condition, quantified by Licor Image Studio™, to a standard curve created using 1, 2.5 and 5 ng of recombinant calreticulin loaded within each gel.

Calreticulin release from BV-2 microglia treated with LPS was measured by ELISA. Briefly, 5 x 10⁴ BV-2 cells were cultured in low-serum DMEM. After 24 hours, supernatants were assessed for calreticulin using a calreticulin ELISA (Abbexa), following the supplied protocol. Absorbances were read at 450nm by the FLUOstar OPTIMA spectrophotometer (BMG Labtech, Ortenberg, Germany).

Release of inflammatory molecules from BV-2 microglia and primary rat microglia were measured by ELISA. Briefly, 5 x 10⁴ BV-2 cells were seeded in serum-free DMEM, 5 x 10⁴ primary rat microglia were seeded in appropriate medium, and microglia were treated accordingly for 20 hours. Following treatments, supernatants were removed, and protein detection was achieved using mouse ELISAs for TNF-α, IL-6 and CCL2, and a rat ELISA for TNF-α, as per the manufacturer’s instructions (all BioLegend). Absorbances were read at 450nm by a spectrophotometer.

Release of inflammatory molecules from CHME3 microglia were assessed using a custom pre-coated meso scale discovery plex human pro-inflammatory panel 2 (Meso Scale Discovery). Briefly, 3 x 10⁴ CHME3 cells were seeded in serum-free DMEM and treated with calreticulin. After 20 hours, supernatants were removed, and protein detection was achieved by electrochemiluminescence for human TNF-α, IL-6 and IL-1β as per the manufacturer’s instructions.

Endotoxin (LPS) levels were measured from recombinant calreticulin preparations using the Pierce™ limulus amebocyte lysate (LAL) chromogenic endotoxin quantification kit (Thermo Fisher Scientific), as per manufacturer’s protocol. Endotoxin levels are expressed as endotoxin unit per milliliter (EU/mL).

2,000 BV-2 cells per well were seeded in DMEM and treated with ± calreticulin. After 24, 48 and 72-hour periods, microglial count and viability were measured through differential dye uptake of Hoechst 33342 and propidium iodide by fluorescence microscopy.

Peptide synthesized human amyloid-beta 1-42 was prepared as previously described (26). Briefly, Aβ was dissolved in 1,1,1,3,3,3-hexafluoroisopropanol (HFIP, Sigma) and dried under a stream of nitrogen. Prior to the aggregation assay, Aβ was resolubilized by DMSO (Sigma) and resuspended in DMEM.

Aβ aggregation was measured by thioflavin T fluorescence as previously described (27). Briefly, thioflavin T (Sigma) and Aβ (10 µM final concentration for both) were added to DMEM ± 0.1 μM or 1 μM calreticulin. Aβ was let to aggregate at 31.5°C, with orbital shaking for 10 seconds, every 10 minutes and fluorescence (440 nm absorbance, 480 nm emission) was measured at 10 minute intervals within the plate reader. At the end of the assay, aggregates in wells were imaged with a fluorescence microscope (DMI6000; Leica) using 480/40 nm excitation and 527/30 nm emission filters with a 40x objective.

To generate oligomers, monomeric Aβ (10 μM) was incubated ± 1 μM calreticulin [18 h at 4°C using a thermocycler (Thermo Fisher Scientific)]. These oligomeric samples (10 μL) were applied to carbonate coated grids for 1 min and negatively stained with 1% uranyl acetate for 1 minute. Micrographs were obtained on a Tecnai G2 transmission electron microscope.

Statistical analyses were conducted using GraphPad Prism v9. Statistical differences between two groups were analyzed by unpaired and paired t-tests where indicated. Statistical differences between three or more groups were analyzed by one-way repeated measures or mixed model ANOVA followed by Dunnett’s or Šidák multiple comparisons post-hoc test. Normality of acquired data was tested by Shapiro-Wilk test. Error bars represent the standard error of the mean of experiments (SEM). p-values refer to the probability of the null hypothesis that the means do not differ. p<0.05 was considered significant, and p≥0.05 not significant.

We have previously shown that primary mouse microglia release calreticulin into the medium when inflammatory activated with lipopolysaccharide, LPS (13), and we confirmed here that BV-2 microglia release calreticulin in response to LPS (Supplementary Figure 1A), without inducing cell death (Supplementary Figure 1B). We then tested whether BV-2 microglia would release calreticulin, measured by western blots of cell culture supernatants, in various other conditions. Tunicamycin, which induces ER stress by inhibiting protein glycosylation (28), induced significant calreticulin release (Figures 1A, B), without inducing significant cell death (Supplementary Figure 2). Staurosporine, which induces apoptosis by inhibiting protein phosphatases (29), killed the cells (Supplementary Figure 2), and induced release of calreticulin into the medium (Figures 1A, B).

We then tested whether necrotic neurons would release calreticulin, and found that crushed cerebellar granule neurons significantly released calreticulin into the culture medium, measured by western blots (Figures 1C, D). The reason for crushing the neurons was to test whether calreticulin would be released into the medium when rupturing their plasma membrane. Thus, dying microglia and neurons, and ER-stressed or inflammatory-activated microglia, can release calreticulin, resulting in nanomolar levels in the extracellular culture medium.

Calreticulin is known to be present in human cerebrospinal fluid, but the levels are unknown (20). Having found that microglia and neurons can release nanomolar levels of calreticulin, we tested whether the addition of nanomolar levels of calreticulin can activate microglia to release pro-inflammatory cytokines. Human, recombinant calreticulin was used here and in all subsequent experiments. First, we added 0 nM, 1 nM, 10 nM or 100 nM concentrations of extracellular calreticulin to mouse BV-2 microglia and measured the amount of extracellular TNF-α (tumor necrosis factor alpha) in the medium 20 hours later by ELISA, and compared the release to that induced by the positive control, LPS. Untreated BV-2 microglia released no detectable TNF-α, while 1 nM, 10 nM and 100 nM calreticulin induced the release of 0.05, 0.12 and 0.25 ng/ml TNF-α, and LPS induced the release of 0.35 ng/ml TNF-α (Figure 2A). Thus, nanomolar concentrations of calreticulin can induce TNF-α release from BV-2 microglia, at a similar level to that induced by LPS.

As mouse microglia may differ from human microglia (30), we tested whether calreticulin can activate human microglia to release cytokines, using a human microglial cell line, CHME3 [also known as HMC3 (31)]. By meso scale discovery, we measured TNF-α, IL-6 (interleukin 6) and IL-1β release from CHME3 microglia treated for 20 hours with calreticulin. Calreticulin induced significant release of TNF-α, IL-6 and IL-1β from this human microglial cell line (Figures 2B–D).

As the calreticulin we used was recombinant, and recombinant proteins are potentially contaminated with LPS (32), we tested whether Polymyxin B, which binds and inactivates LPS (33), could prevent the release of TNF-α induced by calreticulin in BV-2 microglia. Polymyxin B prevented the TNF-α release from BV-2 microglia induced by LPS, but had no effect on the TNF-α release induced by calreticulin (Figure 3A). This suggests there was little or no LPS contamination of the calreticulin used. This was confirmed by limulus amebocyte lysate (LAL) assay, which measured LPS levels in the calreticulin preparation to be negligible (0.22 ± 0.01 EU/mL (<1 ng of LPS)).

One potential means by which extracellular calreticulin might activate microglia is via the receptor LRP1, as this receptor mediates some actions of extracellular calreticulin (34, 35). LRP1 is inhibited by LRPAP1 (LRP-associated protein (36);), so we tested whether added LRPAP1 could prevent calreticulin inducing TNF-α release from BV-2 microglia, but found that it had no effect on this release (Figure 3A).

In order to test whether calreticulin can induce the release of other cytokines or chemokines from BV-2 microglia, we added calreticulin to BV-2 microglia and measured IL-6 and CCL2 in the extracellular medium 20 hours later by ELISA. IL-6 is a pro-inflammatory cytokine, and CCL2 (chemokine (C-C motif) ligand 2) is a chemokine, also known as monocyte chemoattractant protein 1 (MCP1). Calreticulin induced the release of both IL-6 and CCL2, and this was not prevented by Polymyxin B or LRPAP1 (Figures 3B, C). Thus, calreticulin induces the release of multiple pro-inflammatory cytokines and a chemokine from BV-2 microglia, and this does not appear to be due to LPS contamination or activation of LRP1.

As BV-2 are a cell line, which may diverge from primary cells as a result of transformation and/or mutations (37), we also tested whether calreticulin could induce TNF-α release from primary microglia isolated from rat brain. Calreticulin did induce TNF-α release from primary microglia, and this was not prevented by Polymyxin B or LRPAP1 (Figure 3D), just as BV-2 microglia (Figure 3A).

We conclude that nanomolar extracellular calreticulin can activate microglia (mouse, rat and human, primary and cell lines), as measured by the release of pro-inflammatory cytokine and chemokines, and this is not mediated by LPS or LRP1.

Having found that damaged or activated brain cells release calreticulin, and that added calreticulin can activate microglia, we went on to test whether extracellular calreticulin can do anything else relevant to neuroinflammation and neurodegeneration, in particular whether calreticulin can act as an alarmin, extracellular chaperone or neuroprotectant. If extracellular calreticulin acts as an alarmin, then it should chemoattract microglia. We tested whether extracellular calreticulin can chemoattract microglia by placing BV-2 microglia on top of a transwell (Boyden) chamber and culture medium ± calreticulin or ATP in the lower chamber. ATP is known to chemoattract microglia (38), and acted as a positive control here. After 24 hours, the number of cells that had migrated through the transwell membrane (with 8 µm pores) were counted, and both calreticulin and ATP were found to significantly stimulate microglial migration (Figures 4A, B). Thus, extracellular calreticulin at nanomolar concentrations can chemoattract microglia.

If extracellular calreticulin activates microglia, then we might expect it to affect microglial proliferation. Calreticulin can alter the proliferation of a variety of cell types, and is implicated in tumor growth (39), tumor suppression (40, 41), and wound healing (42). We therefore tested whether calreticulin affected the proliferation of microglia. To do this, we seeded the same number of BV-2 cells in culture, to which we added either 0, 10 or 50 nM calreticulin, and then counted the numbers of cells after 24, 48 and 72 hours. Across all timepoints, calreticulin dose-dependently reduced the proliferation of microglia (Figures 5A–C), without inducing cell death (Supplementary Figure 3). 10 nM calreticulin significantly inhibited proliferation at all times, and 50 nM calreticulin inhibited proliferation more, but, note that even in the presence of 50 nM calreticulin, the BV-2 microglia still proliferated substantially (Figure 5). Thus, exogenous calreticulin mildly reduces microglial proliferation.

Calreticulin is an intracellular chaperone, which can bind Aβ (21), but it is not known whether calreticulin can act as extracellular chaperone for Aβ, which would be particularly relevant for Alzheimer’s disease. We investigated whether calreticulin affected Aβ aggregation using an in vitro fibrillization assay, incubating pure monomeric Aβ with thioflavin T, which becomes fluorescent when bound to β-sheet structures within Aβ aggregates (43). In the absence of calreticulin, Aβ (10 μM) fibrillized with standard lag-phase kinetics at 31.5°C (Figure 6A). In the presence of 1 μM calreticulin, fibrillization of Aβ was almost completely prevented (Figure 6A). While in the presence of 0.1 μM calreticulin, the kinetics of Aβ fibrillization was little affected, but the maximum fluorescence was reduced (Figure 6A). This suggests that calreticulin may help fold Aβ into forms with less β-sheet structure, consistent with calreticulin acting as a chaperone for Aβ. Imaging the Aβ aggregates after the fibrillization assay confirmed that calreticulin reduced the thioflavin T fluorescence of the Aβ aggregates (Figure 6B).

As oligomeric Aβ is widely considered to be more neurotoxic than Aβ fibrils (44), we next assessed the effect of calreticulin on Aβ oligomer formation. We incubated 10 μM monomeric Aβ ± 1 μM calreticulin for 18 hours at 4°C [conditions known to result in Aβ oligomers (43)], and imaged the resulting Aβ oligomers by transmission electron microscopy. Aβ oligomerized in the presence of calreticulin resulted in larger oligomeric Aβ structures compared to Aβ oligomers formed in the absence of calreticulin (Figures 6C, D).

Taken together, this data suggests that calreticulin inhibits Aβ fibrillization, or folds Aβ into aggregates with less β-sheet structure, and promoted the formation of larger oligomeric structures.

As calreticulin affects fibrillization and oligomerization of Aβ, we next tested whether calreticulin affected the neurotoxicity induced by Aβ. We have previously shown that nanomolar concentrations of Aβ induce neuronal loss mediated by microglia in mixed neuron–glial co-cultures (26). Here, we found that 250 nM Aβ induced loss of about 35% of the neurons over 72 hours (Figures 7A, B). 50 nM calreticulin alone (in the absence of Aβ) had no effect on neuronal or microglial numbers (Figures 7A–C), but did dramatically affect microglial morphology, as microglia became flatter with increased area (Figures 7A, D). Thus, calreticulin activates microglia in a way that is not detrimental to neurons. Indeed, addition of 10 nM calreticulin significantly inhibited the Aβ-induced neuronal loss, and 50 nM calreticulin almost completely prevented the Aβ-induced neuronal loss (Figures 7A, B). Thus, extracellular calreticulin is neuroprotective at concentrations similar those released by brain cells in a variety of circumstances.