Protease-activated receptor-1 activation by granzyme B causes neurotoxicity that is augmented by interleukin-1β

Histochemical portions of this study were performed on formalin-fixed, paraffin-embedded archival material of eight (n = 4 patients with MS, n = 4 control) human brain autopsy tissue samples obtained from The Human Brain and Spinal Fluid Resource Center, University of California Los Angeles, CA. In the four subjects designated with a diagnosis of MS, all patients had a chronic course. The four patients serving as controls in histochemistry experiments did not have neurological diseases (Table 1).

For neurotoxicity and enzyme-linked immunosorbant assay studies, human CSF samples were obtained via lumbar puncture from patients diagnosed with secondary progressive MS (n = 7) involved in a clinical trial of lamotrigine as described previously [24, 25] or from NINDS patients without neuroinflammation (n = 7) who served as controls.

The Office of Human Subjects Protection and Research at the National Institutes of Health approved this research.

iCell neurons (Cellular Dynamics International, Madison, WI) which were derived from induced pluripotent stem cells were used for the viability and toxicity studies. Neurons were cultured according to the manufacturer’s instructions. In brief, 96-well plates were coated with 100 μl of 0.01% (w/v) poly-L-ornithine (Sigma-Aldrich, St. Louis, MO) for 1 h at room temperature, washed twice with sterile distilled water, and finally coated with a 3.3-μg/mL solution of laminin (Sigma-Aldrich) and molecular biology-grade water. The dishes with laminin solution were incubated in a 37 °C incubator for at least 1 h prior to plating of neurons. The laminin solution was aspirated immediately before addition of the cell suspension. A vial containing 2.5 million iCell neurons was removed from liquid nitrogen storage and immediately placed in a 37 °C water bath for exactly 3 min. The cells were transferred to a 50-ml conical tube and diluted to 10 ml total volume with iCell Complete Maintenance Media (Cellular Dynamics International) dropwise per manufacturer’s protocol. Contents of the flask were swirled gently throughout addition of media. Cell viability was assessed from the initial volume and then further dilution was performed to ensure that sufficient quantity of cell suspension was added to 96-well plates to ensure at least 80,000 viable cells were seeded per square centimeter. The cells were kept in a 37 °C incubator in 5% (v/v) CO₂. One day after initial plating, there was a 100% exchange of fresh iCell Complete Maintenance Media, and this media was changed (50–75% exchange of the total volume) every 1–2 days to ensure maximum viability prior to experimental use. Neurons were not used for experiments until at least 3 days after successful plating. Opti-MEM reduced-serum media (Life Technologies, Carlsbad, CA) supplemented with 2% (v/v) B27 (Life Technologies) served as the control and treatment media for all cultures. For multiple day experiments, there was a 100% replacement of media and any additional components every 24 h until analysis. If an additive such as DMSO was used to promote solubility of a compound being evaluated, then an equivalent amount of the additive was added to control media daily. In experiments using human CSF, the cultures were maintained in a 100-μl volume of undiluted thawed CSF from a single donor with a 100% daily exchange to replenish CSF and any additives such as protease inhibitors or interleukin-1 (IL-1) receptor antagonist. Recombinant human granzyme B, IL-1β, and thrombin as well as agonists/antagonists of PAR1 and second messenger components were obtained from R&D Systems/Tocris Bioscience (Minneapolis, MN). Any agonists, antagonists, or inhibitors were added at least 30 min prior to addition of granzyme B and IL-1β. Atopaxar (E5555) was prepared using a variant of the convergent synthesis described in the patent # US 2006/0058370 A1 [26]. The coupling of a fluorinated cyclic benzamidine (A) with a morpholine-substituted phenacyl bromide (B) in THF gave the HBr salt of Atopaxar as seen below.

The syntheses of (A) and (B) are shown in the following Schemes 1 and 2.

Starting with 3-fluorocatechol, the synthesis required essentially seven steps. Benzamidine (A) was ultimately prepared from phthalimide (2) which could be prepared from bromide (1).

Starting with 2-tert-butylphenol, the synthesis of (B) required essentially six steps. One slight variation from the Eisai patent procedure involved the hydrolysis of ketal (3) back to the methyl ketone (4). This material could be purified chromatographically before brominating, resulting in much cleaner phenacyl bromide (B). Atopaxar hydrobromide prepared in this manner gave satisfactory NMR, elemental analysis, and liquid chromatography/mass spectroscopy. The crystal retained 0.25 equivalent of THF as seen by nuclear magnetic resonance and elemental analysis.

Sections of 100-μm thickness were prepared from formalin-fixed and paraffin-embedded brain sections from control and from subcortical white matter containing known lesions in patients with MS. They were deparaffinized using standard protocols. Antigen retrieval was performed using sodium citrate buffer (10 mM sodium citrate, 0.05% (v/v) Tween 20). Endogenous peroxidase was inhibited using a dual enzyme block solution (Dako, Carpinteria, CA) and native protein block was achieved with a serum-free protein block solution (Dako). Following a TBS-Tris wash, sections were immunolabeled for 1 h at room temperature using mouse monoclonal anti-PAR1 (1:500, Abcam, San Francisco, CA) or rabbit polyclonal anti-IL1 receptor (1:100, Abcam) as primary antibodies. After a wash with TBS-Tris, sections were incubated with a Powervision poly-HRP anti-rabbit or anti-mouse secondary antibody solution (Dako). Development for visualization of antibody binding within the tissue was achieved using the chromagen 3,3'-diaminobenzidine (Sigma). Hematoxylin (10%; w/v) was used to provide a counterstain. Omissions of the primary or secondary antibodies were used as negative controls.

The total number of cells and immunolabeled cells were counted manually in a minimum of ten fields from within and immediately around subcortical white matter (healthy or containing a lesion) and the adjacent cerebral cortex (layer VI) of patients with MS and controls. Lesions were defined based on demyelination borders identified with Luxol Fast Blue staining of an adjacent section. For quantification of cell number within a tissue section, a 20-μm grid was aligned manually across the desired counting area and all cells fully within the section were counted using a microscope with a 20× objective configured to image in transmitted light mode. Optimal light and focal settings were established prior to counting. Serial sections stained with anti-NeuN antibody (1:1000, Abcam) were used to definitively identify whether a given counted cell nucleus belonged to a neuron. Cells were graded as “no,” “low”, or “high” expression based on comparative visual appearance. Regions with significant tissue artifacts or overlapping with a tissue edge were excluded from analysis. The total density of cells was estimated from the observed counts using a modified Abercrombie estimation [27] and is expressed as a calculated density (neurons/mm²).

Concentrations of granzyme B (eBioscience, San Diego, CA) and IL-1β (Abcam) were quantitated in CSF samples using enzyme-linked immunosorbent assay (ELISA) kits according to manufacturer’s instructions. CSF samples from patients with MS or control patients were incubated on microtiter strips with biotinylated monoclonal anti-granzyme B or IL-1β antibody. Streptavidin-peroxidase and substrate were added, and absorbance was measured at 450 nm. Quantitation of secreted granzyme B and IL-1β were determined based on a standard curve generated from known standards included in the kits.

Total protein concentration in CSF samples was analyzed using a BCA Protein Assay (Life Technologies) following the manufacturer’s instructions. Glucose concentration was estimated in CSF using a Glucose Colorimetric Assay Kit (Cayman Chemical, Ann Arbor, MI). A glutamate assay kit (Abcam) was used to measure CSF glutamate concentrations.

The MultiTox-Fluor Multiplex Cytotoxicity Assay (Promega, Madison, WI) was used for simultaneous evaluation of viability and toxicity in neuronal cultures for all conditions. Cultures were maintained in 96-well plates with black walls in a total volume of 100 μl of media with or without experimental additives such as granzyme B, IL-1β, or inhibitors. To inhibit granzyme B activity in CSF samples, these samples were pretreated with the selective cell-permeable protease inhibitor Z-Ala-Ala-Asp-chloromethylketone (Enzo Life Sciences, Farmingdale, NY). The optimal concentration for additives was determined empirically. The MultiTox-Fluor Multiplex Cytotoxicity Assay Reagent was prepared at two times concentration according to the manufacturer’s instructions and 100 μl of this assay reagent was added to each well. Each well’s contents were mixed by gentle pipetting. The plates were returned to the incubator for 1 h. Fluorescence intensity was then measured using a FlexStation 3 Microplate Reader and SoftMax Pro software (Sunnyvale, CA) with the following settings, for viability: excitation = 400 nm and emission = 505 nm and for toxicity: excitation = 485 nm and emission = 520 nm. All conditions were performed at least in duplicate and data were pooled from multiple independent experiments. For multiple day cumulative toxicity measurements, several wells were analyzed at each daily time point to generate a cumulative toxicity percentage for a given day. Data were collected as relative fluorescence units. Conditions included in each experiment included empty wells as well as wells without neurons containing just media and additives for background assessment, wells containing neurons in experimental media with appropriate additives which served as negative controls, and wells with sufficient saponin 0.1% (w/v) added prior to assessment to ensure 100% toxicity which served as the positive control condition for toxicity calculation purposes.

Neurons were fixed in 4% (w/v) paraformaldehyde for 10 min at room temperature, followed by three washes with PBS. After incubation in 0.1% (v/v) Triton X-100 in PBS (PBS-T) for 10 min at room temperature, the cells were incubated with blocking buffer (PBS-T containing 4% (v/v) goat serum and 1% (v/v) glycerol (Sigma-Aldrich) at room temperature for 20 min. Cells were immunostained with mouse monoclonal anti-PAR1 (1:500, Abcam) and rabbit polyclonal anti-IL1 receptor (1:100 Abcam), followed by corresponding secondary antibodies (anti-mouse Alexa Fluor 488 1:200 Abcam, anti-rabbit Alexa Fluor 647 1:200 Abcam) and 4′,6-diamidino-2-phenylindole (DAPI) nuclear staining. Images were acquired on a Zeiss LSM 510 META multiphoton confocal system (Carl Zeiss, Dublin, CA) or an EVOS fluorescence microscope (AMG, Bothell, WA).

The total number of DAPI-stained nuclei per field provided an estimate of the total number of neurons present. Degenerating nuclei were excluded from counts as were indistinct DAPI-positive structures. The number of PAR1-positive neurons was then counted after overlaying these images on the DAPI-positive images. Only neurons with distinct DAPI-positive nuclei were noted for analysis. The percentage of positive neurons for a given field was generated by dividing the number of PAR-stained neurons by the total of the counted DAPI-positive nuclei.

PAR1 expression was determined by quantitative real-time polymerase chain reaction (qPCR) as described elsewhere [28]. Briefly, cultured iCell neurons were treated with granzyme B, interleukin 1β, and/or siRNA designed to suppress PAR1 for the stated times and total RNA was extracted using a Qiagen RNeasy mini kit (Qiagen, Valencia, CA). The RNA samples were treated with DNase I (Invitrogen, Grand Island, NY) for 15 min followed by Turbo DNAse treatment (Life Technologies) according to the manufacturer’s instructions for “stringent” conditions. Then complementary DNA (cDNA) was synthesized from this purified total RNA using random primers (PAR1 sense: 5′-AGTAGGCTATTCCTGAGAGCTGCAT-3′; PAR1 antisense: 5′-ATGGCCCTGGCATGTGTCT-3′) and SuperScript III First-Strand cDNA kit (Invitrogen). The resulting cDNA was then mixed with gene-specific primers and SYBR Green PCR universal master mix (Life Technologies). RNA was amplified using an ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA) according to the manufacturer’s instructions. The relative levels of messenger RNA (mRNA) were calculated using the ΔΔCt method by normalization to an internal control, glyceraldehyde-3-phosphate dehydrogenase (GAPDH).

The Accell siRNA kit (Thermo Scientific, Waltham, MA) was used to reduce expression of PAR1. A pooled set of siRNA sequences, three directed against the human F2R (thrombin receptor, PAR1) Open Reading Frame and one against the 3′UTR of the gene were used. The siRNA introduction was performed according to the manufacturer’s instructions for “plated cells sensitive to media-free delivery conditions.” To maximize siRNA delivery to neurons, an aliquot of TransIT-X2 Dynamic Delivery System (Mirus Bio, Madison, WI) was added with the diluted siRNA to the Accell delivery buffer. After 72 h, the siRNA solution was replaced with experimental media following a wash with PBS. Reduction in PAR1 protein was confirmed by a Western blot analysis using anti-PAR1 antibody (1:1000) following a methodology otherwise as described elsewhere [18].

Human post-mortem brain tissue sections from patients with MS and control patients lacking neuroinflammatory disease were used to investigate whether there were qualitative or quantitative differences in neurons expressing PAR1 and the IL-1 receptor within healthy and inflammatory subcortical tissue samples.

Brain tissue from both control patients and patients with MS exhibited PAR1 staining; neurons and glia had detectable surface expression of PAR1 (Fig. 1a, b). IL-1 receptor staining was consistently rated as “low” or “no” expression without visual distinction between tissue from patients with MS or control patients (Fig 1c, d). Brain tissue originating from patients with MS had significantly fewer neurons overall owing to loss of cellularity within white matter lesions and in the cortex (Table 2). Despite the discrepancy in numbers of cells, there were approximately 30% more PAR1 high expression neurons within the cortex adjacent to, inside, and around lesions from brain tissue samples of patients with MS compared to healthy patients’ tissue. Regions around and within demyelinated lesions had concentrations of “high expression” rated cells (Fig. 1e). Additionally, there were prominent PAR1-positive neurons with darkly stained axons situated inside demyelinating lesions (Fig. 1f); similar neurons were not found in healthy control samples. Omitting the primary antibodies from the staining protocol yielded no visible PAR1 or IL-1 receptor staining in brain tissues (not shown).

We used ELISA kits to determine the levels of granzyme B and IL-1β within samples of human CSF to confirm the relative amounts of granzyme B and IL-1β in CSF samples from patients with MS. Granzyme B levels in CSF from patients with MS ranged from 10.11 to 12.22 pg/ml (mean = 14.22 pg/ml), and the IL-1β levels were detectable at values between 7.30 and 24.56 pg/ml (mean = 17.38 pg/ml). There was a significant positive correlation between granzyme B and IL-1β within CSF samples from patients with MS (r = 0.852, p < 0.015). In control samples, the granzyme B levels were near the assay’s detection threshold (range <0.20–1.10 pg/ml) and IL-1β levels in CSF samples from control patients were below the assay’s detection threshold (<6.5 pg/ml).

To assess whether CSF from patients with MS was neurotoxic, aliquots of the assayed CSF samples were then introduced into human neuron cultures derived from induced pluripotent stem cells for 3 days. Viability and toxicity were measured simultaneously within each sample well at the end of the third day. Following 3 days of CSF exposure, there was significant neurotoxicity (>70%) in cultures cultured with CSF from patients with MS compared with neurons in control media (Fig. 2a, b). When viability and toxicity were analyzed for the samples, there was a significant negative correlation (r ² = 0.882, p < 0.002) between the concentration of granzyme B within the sample well and the viability of the neurons within the well after 3 days (Fig. 2c). A reciprocal significant positive correlation (r ² = 0.853, p < 0.0004) between granzyme B concentration and toxicity was also observed (Fig. 2c). The correlations between IL-1β concentration and toxicity (r ² = 0.5759, p < 0.01) and viability (r ² = 0.5932, p < 0.009) were significant but less robust. The control patients’ CSF was as non-toxic to the neuronal culture as the control neuronal media (Fig. 2d). There was no difference between the CSF samples from patients with MS and control CSF with respect to concentrations of glucose, total protein, or L-glutamate (Table 3).

When CSF from patients with MS was pretreated with a granzyme B-specific protease inhibitor prior to being added into neuronal cultures, the observed neurotoxicity was reduced and not significantly different than control media with protease inhibitor. Addition of a selective IL-1 receptor antagonist (AF 12198) decreased toxicity by approximately 50%. Protease inhibition together with IL-1 receptor antagonism was as effective as protease inhibition alone in preventing neurotoxicity after 3 days of exposure (Fig. 2d).

To determine the effect of repeated chronic exposure of granzyme B on neurons, human neurons were plated at a known, fixed concentration and allowed to mature per manufacturer’s guidelines. Once neurons exhibited a mature phenotype, they were exposed to a range of concentrations of granzyme B for 7 days. To maintain potency throughout the duration of the experiments, granzyme B was replenished on a daily basis for 7 days with neuronal viability measured in a set of neurons reared in parallel at each 24-h time point. As in previous studies, a single 24-h exposure to granzyme B yielded a dose-dependent reduction in viability (Fig. 3a). However, when granzyme B was replenished on a daily basis, the effect of granzyme B in culture reached a plateau (Fig. 3b). By day 4, the granzyme B toxicity approached zero, and the toxicity converged on that of the control conditions and subsequent days’ exposures were not toxic relative to control media (Fig. 3c). The decline in viability noted on day 5 in culture was universal and is an expected outcome from depriving these neurons of proprietary media’s nutrients.

To investigate the basis for the observed stabilization in viability and toxicity, immunocytochemistry techniques were used to visualize the PAR1 receptor on the surface of cultured neurons, and we observed that consistently the overwhelming majority of neurons (91.66 ± 4.09%) in this specific neuron preparation had visible surface expression of PAR1 at baseline (Fig. 4a). Following several days of granzyme B exposure when toxicity returned to control media equivalence, we observed that the percentage of PAR1-positive neurons would sequentially decline to less than 10% (6.57 ± 0.64%) of the total number (Fig. 4b). The expression of IL-1 receptor increased after initial granzyme B exposure then remained stable for the duration of exposure (Fig. 4c, d). The timing of the decline in the percentage of PAR1-positive neurons in these cultures mirrors the fall in toxicity we observed with repeated presentation of granzyme B (Fig. 4e). Thus, it appeared that a finite number of neurons expressing PAR1 were at-risk to granzyme B and the ultimate total toxicity achieved corresponded to neurons expressing high levels of PAR1 on their surfaces.

We explored the role of neuronal surface PAR1 in granzyme B neurotoxicity PAR1 by both increasing and decreasing PAR1 expression. IL-1β increases PAR1 expression on neurons, and administering IL-1β with granzyme B had a dramatic effect on neurotoxicity by sustaining neural losses throughout the 5-day observation such that the cumulative loss of neurons by day 5 reached approximately 100% (Fig. 5a). The increased toxicity was not due to a direct toxic effect of IL-1β because IL-1β was not toxic when administered alone (Fig. 5b). We introduced a selective antagonist of the IL-1 type I receptor (AF12198) to block IL-1β signaling. The antagonist blocked the IL-1β component of granzyme B toxicity (Fig. 5b). IL-1β acting at the IL-1 receptor altered PAR1 production in the neurons. IL-1β caused PAR1 mRNA expression to increase twofold and PAR1 protein to increase fourfold (Fig. 5c). Antagonism of IL-1 receptors prevented a significant IL-1β-mediated increase in PAR1 mRNA and protein.

We used a siRNA preparation to reduce PAR1 expression in vitro. A siRNA cocktail directed against PAR1 mRNA markedly reduced PAR1 mRNA and protein; these results were confirmed by real-time qPCR and Western blot analysis (Fig. 5d). Inhibiting PAR1 using siRNA made neurons resistant to granzyme B toxicity. When neurons with inhibited PAR1 expression were chronically exposed to granzyme B, there was no decrease in viability (Fig. 5e). A scrambled siRNA sequence of PAR1 and siRNA for a housekeeping gene (GAPDH) both failed to prevent granzyme B toxicity (Fig. 5e).

To establish PAR1 as the definitive receptor for the granzyme B-mediated neurotoxicity, two different strategies were attempted, antagonism and the use of alternative PAR agonists. To determine if direct antagonism of PAR1 could prevent neurotoxicity due to granzyme B, an antagonist compound specific to PAR1, Atopaxar (also known as “E5555”), was placed into the media prior to the introduction and replacement of granzyme B. Atopaxar successfully prevented chronic neurotoxicity due to granzyme B in a dose-dependent fashion (Fig. 6a).

To determine if maximum activation of PAR1 had been achieved at the selected dose of granzyme B (10 nM), we added the native ligand for the PAR1 receptor, thrombin, with granzyme B and observed toxicity over the course of 3 days. Thrombin alone was neurotoxic but produced no additional toxicity after pre-treatment with granzyme B (Fig. 6b). As an additional confirmation, the peptide ligand TFLLR-NH₂, a non-protease agonist of PAR1 receptors, was added with and without granzyme B. This selective agonist proved as toxic as granzyme B when introduced alone, but the inactive reverse sequence of this peptide (RLLFT-NH₂) did not have any toxic effect (Fig. 6c). When TFLLR-NH₂ was added with granzyme B, as was the case with thrombin, there was no additional significant increase in toxicity. Therefore, we concluded that the observed toxicity was attributable only to optimized activation of PAR1 and not significantly influenced by off-target effects.

Protease activation of PAR1 can trigger several different second messenger entities including phospholipase Cβ, IP3, PI-3-kinase, protein kinase C, and ERK 1/2. We screened several second messenger inhibitors directed against targets within these pathways to investigate whether interference with a specific path could modify the toxicity of chronic granzyme B exposure.

We found that blocking phospholipase Cβ hydrolysis of PIP2, an intermediate in the IP3/diacyl glycerol pathway, using the inhibitor U73122 prevented chronic granzyme B-mediated neurotoxicity (Fig. 7a). Inhibition using U73122 had no significant impact on acute granzyme B toxicity as its effect was significant over a time course of several days and evident only when cumulative toxicity was analyzed (Fig. 7c).

The hydrolysis of PIP2 catalyzes the production of inositol trisphosphate (IP3). Two inhibitors of IP3, 2-APB and (−)-xestospongin C, prevented neurotoxicity due to chronic granzyme B (Fig. 7b). The coupling of PAR1 activation through IP3 production completes a clear and known mechanism of neurotoxicity (Fig. 7d). We incidentally observed a dose-dependent improvement in viability following an acute granzyme B toxicity for 2-APB exclusively (Fig. 7c).

Pre-treatment with inhibitors targeting protein kinase C (Ro 32–0432 hydrochloride), PI3 (LY 294002), ERK (FR180204), and src kinase (Src Inhibitor-1) had no effect on granzyme B neurotoxicity (Fig. 8a). Minocycline, a known neuroprotective agent, provided a mild reduction in neuronal death due to chronic granzyme B (Fig. 8b).