A caspase-2 inhibitor blocks tau truncation and restores excitatory neurotransmission in neurons modeling FTDP-17 tauopathy

Abstract

Synaptic and cognitive deficits mediated by a severe reduction in excitatory neurotransmission caused by a disproportionate accumulation of the neuronal protein tau in dendritic spines is a fundamental mechanism that has been found repeatedly in models of tauopathies, including Alzheimer’s disease, Lewy body dementia, frontotemporal dementia, and traumatic brain injury. Synapses thus damaged may contribute to dementia, among the most feared cause of debilitation in the elderly, and currently there are no treatments to repair them. Caspase-2 (Casp2) is an essential component of this pathological cascade. Although it is believed that Casp2 exerts its effects by hydrolyzing tau at aspartate-314, forming Δtau314, it is also possible that a non-catalytic mechanism is involved, since catalytically dead Casp2 is biologically active in at least one relevant cellular pathway, i.e., autophagy. To decipher whether the pathological effects of Casp2 on synaptic function are due to its catalytic or non-catalytic properties, we discovered and characterized a new Casp2 inhibitor, compound 1 (pK_i (Casp2) = 8.12), which is 123-fold selective versus Casp3 and >2000-fold selective versus Casp1, Casp6, Casp7, and Casp9. In an in vitro assay based on Casp2-mediated cleavage of tau, compound 1 blocked the production of Δtau314. Importantly, compound 1 prevented tau from accumulating excessively in dendritic spines and rescued excitatory neurotransmission in cultured primary rat hippocampal neurons expressing the P301S tau variant linked to FTDP-17, a familial tauopathy. These results support the further development of small molecule Casp2 inhibitors to treat synaptic deficits in tauopathies.

Graphical Abstract

INTRODUCTION

Dementia is the most common clinical manifestation of tauopathies, defined as neurodegenerative diseases with tau histopathology. Tauopathies are the most prevalent form of neurodegenerative diseases. Tau is a neuronal cytoskeletal and scaffold protein best known for its ability to bind tubulin and stabilize microtubules; it is also known for forming pathological, hyper-phosphorylated cellular inclusions in over two dozen neurodegenerative disorders, called tauopathies, and it accumulates excessively in dendritic spines in multiple models of tauopathies^1–7. In models of tauopathies expressing mutant human tau, this accumulation depends on Casp2⁴ and tau phosphorylation⁸.

Caspases comprise a family of cysteine proteases found only in multicellular organisms. Some caspases (initiators) are activated by intra- or extracellular signals and cleave other caspases (executioners) which degrade vital structural proteins resulting in cell death. Other caspases both sense the triggers and implement the responses of the functions they serve. Caspase-2 (Casp2) belongs in the latter category, being capable of transducing a variety of incoming signals into proteolytic cascades⁹. In the brain, only two physiological roles for Casp2 are known. During development, Casp2 likely mediates neuronal apoptosis^10,11. In the mature brain, Casp2 facilitates long-term depression, a synapse-weakening form of long-lasting synaptic plasticity¹².

Casp2 also performs pathological roles. Lowering Casp2 levels restores memory function in mouse models of frontotemporal dementia⁴, Alzheimer’s disease¹³, and Huntington’s disease¹⁴. While it is assumed that lowering Casp2 rescues synaptic and cognitive deficits by reducing its ability to hydrolyze its substrates, one of which is believed to be tau⁴, it is also possible that it exerts its effects via a non-catalytic mechanism. For example, dead Casp2 has been shown to negatively regulate autophagy¹⁵. Therefore, another mechanism by which Casp2 might mediate cognitive deficits is by inhibiting autophagy, thereby causing misfolded proteins and pathological aggregates to accumulate in the brain.

We sought to resolve whether the pathological effects of Casp2 on synaptic function are due to its catalytic or non-catalytic properties using a Casp2 inhibitor. Because there have been no reports of a neuroactive, selective Casp2 inhibitor, we designed a new, reversible covalent Casp2 inhibitor based on published structure-activity relationships. Previous studies have shown that the placement of sterically demanding groups in the S2 subsite of the canonical pentapeptide structure increases the Casp2-selectivity of reversible, covalent inhibitors¹⁶. In addition, specific groups in the S4 and S5 subsites were shown to increase the selectivity of irreversible, covalent inhibitors of Casp2¹⁷.

RESULTS AND DISCUSSION

Design and characterization of Casp2 selective inhibitors

We prepared new, potent, and, reversible pentapeptide Casp2 inhibitors (Figure 1) by employing a rational-design strategy. That is, we combined the structural elements known to lead to selectivity and potency which had been identified in previous structure-activity-relationship (SAR) studies. In 2011, Maillard et al. reported a moderately selective Casp2 inhibitor 2 by introducing racemic 6-methyltetrahydroisoquinoline-1-carboxylic acid (6-Me-THIQ-COOH) at P2 position of the previously known non-selective Casp2 inhibitor AcVDVAD-CHO¹⁶. We prepared 2a and 2b (pure P2 epimers of compound 2) by employing enantiopure (R)-6-Me-1-THIQ-COOH and (S)-6-Me-1-THIQ-COOH, respectively (Figure 1). In a fluorescence-based in vitro biochemical assay¹⁸ compound 2a containing (R)-THIQ did not show any notable affinity for Casp2 and Casp3 (pK_i (Casp2) = 4.98, pK_i (Casp3) < 4; cf. Table 1). In contrast, 2b bearing (S)-THIQ showed significant affinity for Casp2 with 70-fold Casp2/Casp3 selectivity (pK_i (Casp2) = 7.89, pK_i (Casp3) = 6.04; cf. Table 1).

Figure 1.

Structures of caspase-2 inhibitors 1, 2a, 2b, and 3.

Table 1.

Binding data (pK_i values) of pentapeptides 1, 2a, 2b, and 3 at Casp2 and Casp3. Selectivity expressed as K_i (Casp3) / K_i (Casp2).

Cmpd.	pKi ± SEM		Casp2 Selectivity
	Casp 2	Casp 3
1	8.12 ± 0.10	6.03 ± 0.06	123
2a	4.98 ± 0.20	< 4	> 9
2b	7.89 ± 0.02	6.04 ± 0.06	70
3	7.40 ± 0.11	< 4.5	794

Data shown are mean values ± SEM of N ≥ 3 independent experiments, each performed in duplicate or triplicate. Data were analyzed by nonlinear regression and were best fitted to sigmoidal concentration−response curves using four parameter logistic (4PL) regression using GraphPad Prism.

Poreba et al. have shown that pentapeptides with an extended side chain carboxylic acid-based residue at P4 (h-L-glutamic acid) and a cyclic non-acylated amino acid at P5 (indoline-2(S)-carboxylic acid), show significant affinity and selectivity for Casp2¹⁷. Taking (S)-THIQ-based 2b as our lead structure, we further replaced P4 L-aspartic acid moiety with h-L-glutamic acid and introduced (S)-indoline-2-carboxylic acid in place of N-acetyl-L-valine at P5. These modifications resulted in compound 3 with 794-fold Casp2 selectivity versus Casp3 (Figure 1; Table 1). Reversible peptide- and peptidomimetic protease inhibitors featuring aldehyde (-CHO) reactive groups are popular for early SAR studies since they are readily prepared by solid-phase synthesis and used in known drugs. Notably, the -CHO warhead is a prominent feature of the recently approved drug voxelotor¹⁹, a treatment for sickle cell anemia that is dosed at 1.5 g QD. Typically, however, the R-CHO reactive group is seen as undesirable due to its high electrophilicity and its susceptibility to oxidative metabolism. On the other hand, nitrile-based covalent inhibitors are also clinically validated and known for their safety and efficacy in humans (²⁰, a covalent inhibitor of the SARS-CoV-2 main protease) and have been used successfully in inhibitors of other caspases²¹. Therefore, we prepared 1, the nitrile analog of 3, using solution-phase chemistry (Scheme 1, Figure 1). This compound was fully characterized and used in all subsequent experiments. Compound 1 showed high-affinity at Casp2 (pK_i = 8.12; cf. Table 1), a promising selectivity profile in an in vitro caspase panel assay (123-fold selective versus Casp3 and >2000-fold selective versus Casp1, 6, 7, 9; cf. Table 2) and, evinced only weak activity at thrombin and cathepsin B proteolytic enzymes (See Table 2).

Scheme 1.

Synthesis of caspase-2 inhibitor 1. (i) 2-chlorotrityl chloride resin, collidine, DCM; (ii) 20% piperidine in DMF; (iii) Fmoc-hGlu(OtBu)-OH, HATU, DMF; (iv) 20% piperidine in DMF; (v) (S)-Fmoc-IdcOH, HATU, DMF; (vi) 20% HFIP in DCM; (vii) H₃PO₄; (viii) (S)-6, HATU, collidine, ACN; (ix) 30% Diethylamine in ACN; (x) 9, HATU, collidine; (xi) H₃PO₄; (xii) 30% Diethylamine in ACN; (xiii) 10% Pd/C.

Table 2.

Binding data of compounds 1, 2a, 2b, and 3 at Casp1, Casp2, Casp3, Casp6, Casp7, Casp9 (pK_i values), Cathepsin B and Thrombin a (IC₅₀ values in nM).

Target	pKi ± SEM
	1	2a	2b	3
Casp1	< 4	< 4	5.56 ± 0.07	6.41 ± 0.12
Casp2a	8.12 ± 0.10	4.98 ± 0.20	7.89 ± 0.02	7.40 ± 0.11
Casp3a	6.03 ± 0.06	< 4	6.04 ± 0.06	< 4.5
Casp6	< 4	< 4	< 4	< 4
Casp7	4.80 ± 0.06	< 4	< 4.5	< 4
Casp9	< 4.5	< 4	< 4.5	5.12 ± 0.04
Cathepsin B	18,620b	n.d.	38,000b	14,300b
Thrombin a	> 100,000b	n.d.	n.d.	n.d.

Data shown are mean values ± SEM of N ≥ 2 independent experiments, each performed in duplicate or triplicate. Data were analyzed by nonlinear regression and were best fitted to sigmoidal concentration−response curves using four parameter logistic (4PL) regression using GraphPad Prism.

^aData at Casp2 and Casp3 are also shown in Table 1.

^bData shown are IC₅₀ values (nM) of N = 1 independent experiments.

Casp2-selective inhibitors block enzyme-catalyzed tau cleavage at D314

To understand the effects of Casp2-selective inhibitors on enzyme-catalyzed site-specific cleavage of tau, a naturally occurring substrate in the brain, we performed in vitro Casp2 cleavage assays using purified recombinant proteins. The presence of Δtau314, the soluble truncated tau fragment ending C-terminally at D314⁴, was verified by mass spectrometry (MS) (Figure 2A). Both Casp2-selective compounds, 1 and 2b, inhibited enzyme-catalyzed Δtau314 production at 10 μM but with different efficacies (96% for 1 versus 68% for 2b). In contrast, 2a, the epimer of 2b, did not inhibit Δtau314 production at 10 μM (Figure 2B). 1 inhibited tau-cleavage at D314 with an IC₅₀ of 2.02 μM (Figure 2C).

Figure 2.

Casp2–selective inhibitors block enzyme-catalyzed Δtau314 production in vitro A) Mass spectrometry (MS) analysis confirmed the identity of a ~37-kDa tau cleavage product as Δtau314. A representative Western blot (WB) of Casp2-cleaved tau proteins immunoprecipitated by the Δtau314-specific 4F3 antibody was probed using biotin-conjugated pan-tau antibody tau-5 (tau-5-biotin). Liquid chromatography-MS/MS analysis identified the presence of Δtau314 in the ~37 kDa product as indicated by a representative MS spectrum of its trypsinized signature peptide at the C-terminus. B) Hybrid compounds inhibited Casp2-mediated Δtau314 production with different efficacies. Upper panel, a representative immunoprecipitation (IP, 4F3)/Western blotting (WB, tau-5-biotin) showing Δtau314 in the presence of various hybrid compounds at 10 μM. Δtau314 produced in the absence of any compound (No cmpd.) and a reaction in the absence of enzyme (No Casp2) served as positive and negative controls, respectively. Lower panel, quantification. The level of Δtau314 produced in the absence of any tested compound was defined as 100% (dotted line). At 10 μM, compounds 1 and 2b inhibited tau cleavage at D314 by 96% and 68%, respectively, whereas compound 2a showed no inhibition; compound 1 inhibited Δtau314 production more efficiently than compound 2b. Experiments were repeated six times. Individual values (open circles), means (histograms), and standard deviations (SDs, error bars) are shown. A single-sample t-test was performed to compare the effect of each compound to that of no compound (####, p < 0.0001). One-way ANOVA was performed to compare effects of tested compounds (F_{(2, 15)} = 48.50, P < 0.0001), followed by Tukey’s post hoc test (*, p < 0.05; ****, p < 0.0001). C) Determining the IC₅₀ of compound 1. Upper panel, a representative IP (4F3)/WB (tau-5-biotin) showing Δtau314 in the presence of various concentrations of compound 1. Lower panel, levels of Δtau314 (normalized to the no-compound control) and compound concentrations were fitted to a dose-response curve; the IC₅₀ of compound 1 was determined to be 2.02 ± 0.19 μM. Experiments were performed in duplicate. Means (open circles) and SDs (error bars) are shown.

Compound 1 prevents the disproportionate accumulation of P301S tau in dendritic spines of cultured rat hippocampal neurons

The disproportionate accumulation in dendritic spines of mutant tau linked to FTDP-17 depends upon tau-truncation by Casp2⁴. As shown in Figure 2, of the compounds tested the one that blocks tau-truncation most effectively is 1. Here, we examined the P301S tau mutation linked to FTDP-17^6,22. To determine whether 1 can prevent the accumulation of tau in dendritic spines, we cultured dissociated hippocampal neurons from postnatal day 0–1 rat pups. After 5–7 days in vitro (DIV), we co-transfected the neurons with plasmids encoding DsRed (to label neuronal morphology) and GFP-tagged wild-type or P301S mutant tau proteins (Figure 3A), as previously described¹. Neurons at 17–18 DIV were incubated with 10 μM compound 1 for 3 days and live neurons were imaged at 20–21 DIV.

Figure 3.

Compound 1 blocks the hyper-accumulation of tau caused by the P301S mutation in dendritic spines. A) Untreated neurons co-expressing DsRed and GFP-tagged wild-type tau (top row) were compared with neurons co-expressing DsRed and GFP-tagged P301S tau in the absence (middle row) or presence of compound 1 (bottom row). Arrows denote dendritic spines containing tau. Triangles denote spines devoid of tau. B) Comparison of the average percentage of dendritic spines labeled by DsRed containing GFP-tagged tau between the three groups. n = 8 neurons in each group; ANOVA, ***, p < 0.001. C) Comparison of the density of dendritic spines (the number of DsRed-labelled dendritic spines/100 μm length of dendrites) between the three groups. n = 8 neurons in each group; ANOVA, NS (not significant), P > 0.05.

As shown in Figure 3A, in neurons co-expressing DsRed and wild-type tau, relatively few dendritic spines contain GFP-tagged tau proteins (top row). In contrast, in neurons co-expressing DsRed and GFP-tagged P301S tau, most of the dendritic spines contain GFP-tagged tau (middle row). When 1 was applied to neurons expressing GFP-tagged P301S tau, a minority of dendritic spines contain GFP-tagged tau (bottom row). The percentage of dendritic spines containing GFP-tagged tau proteins in neurons significantly increased from 21.9% in neurons expressing wild-type tau to 75.5% in neurons expressing P301S tau; 1 reduced the percentage to 24.8% (Figure 3B; n=8 neurons in each group; ANOVA, F = 82.1; Bonferroni post-hoc test: P < 0.01 comparing untreated wild-type and P301S tau, and then comparing untreated and treated P301S tau). There were no significant differences in the density of dendritic spines between the three groups (Figure 3C; ANOVA, F = 0.13, P>0.05), indicating that neither the P301S mutation nor 1 caused loss of dendritic spines. Overall, these results indicate that 1 blocks the excessive accumulation of tau induced by the P301S mutation in dendritic spines, supporting its efficacy in reversing the molecular pathology of dendritic spines in a cellular model of tauopathy.

Compound 1 rescues P301S tau-induced functional deficits in dendritic spines

Our previous studies in models of several different tauopathies demonstrated that the excessive accumulation of tau in dendritic spines is associated with a reduction in postsynaptic excitatory neurotransmission due to the internalization of AMPA receptors^1–5,7,8. Therefore, we compared the ability of 1, 2a, and 2b to rescue functional synaptic deficits in a cellular model of tauopathy.

For these studies, we used PS19 transgenic mice expressing P301S mutant tau, one of the most used models of FTDP-17. To detect functional deficits, we cultured primary hippocampal neurons from heterozygous transgenic mice overexpressing P301S tau^23,24 and littermate controls, (denoted Tau −/+ and Tau −/− respectively in Figure 4), and recorded miniature excitatory synaptic currents (mEPSCs). 1, 2a (inactive analog of 2b), or 2b (final concentrations = 10 μM) were directly applied to the culture media (Figure 4A) and incubated for 3 days. In the two groups treated with 2a, the average amplitude of mEPSC responses in Tau −/+ neurons is significantly smaller than that in Tau −/− neurons (Figure 4B, ANOVA, F = 3.3; Bonferroni post hoc test: P < 0.05), indicating that the expression of P301S tau impaired postsynaptic function of the excitatory synapses by reducing the number of functional AMPA receptors on post-synaptic membranes. In contrast, the average amplitudes of mEPSC responses in Tau −/+ and Tau −/− neurons treated with 1 do not differ significantly, suggesting that 1 rescues impaired mEPSC amplitudes (ANOVA, F = 3.3, P>0.05). The average amplitudes of mEPSC responses of Tau −/+ neurons treated with 1 are significantly larger than those treated with 2a (Figure 4B, F = 3.3; Bonferroni After Test: P < 0.01) or 2b (Figure 4B, F = 3.3; Bonferroni post-hoc test: P < 0.05), indicating that 1 rescues post-synaptic neurotransmission more effectively than 2a or 2b. There were no significant differences in the average frequency of mEPSCs between the six groups (Figure 4C, ANOVA, F = 0.6, P > 0.05), indicating no changes in presynaptic function.

Figure 4.

Treatment with 1 rescue synaptic dysfunction caused by the P301S mutation. A. Representative traces of mEPSCs recorded in Tau−/+ and Tau−/− neurons that have been treated with 2a, 2b, and 1. B and C. Comparisons between the means of mEPSC amplitude (B) or frequency (C) between the above six groups. n = 31 neurons in Compound 1 treated groups (−/− and −/+); n = 41 in Compound 2b treated groups; n = 39 in Compound 2a treated groups; ANOVA, *, p < 0.05; **, p < 0.01. D. Comparisons of cumulative curves of mEPSC amplitudes recorded in the above six groups. E. A comparison between the rescuing efficacy between 1 and 2b in Tau −/+ neurons. For D and E: Mann-Whitney U Test, *, p < 0.05; ***, p, 0.001. For B-E, orange circles, blue circles, and blue triangles denote treatment with 2a, 2b, and 1, respectively. Closed shapes denote Tau−/+ mice and open shapes denote Tau−/− mice.

To confirm and extend the results in Figure 4B, we analyzed the cumulative curves of mEPSC amplitudes. The cumulative curve of mEPSC amplitudes of 2a-treated neurons from Tau −/+ compared with Tau −/− mice is shifted to the left (Figure 4D, Mann-Whitney U test: P < 0.001), indicating that the P301S mutation decreases the amplitudes of mEPSCs. The cumulative curve of mEPSC amplitudes of 2b-treated neurons from Tau −/+ compared with Tau −/− mice is shifted to the left but less markedly (Figure 4D, Mann-Whitney U test: P < 0.001), indicating that 2b partially blocks the synaptic impairment caused by the P301S mutation. In contrast, the cumulative curve of mEPSC amplitudes of 1-treated neurons from Tau −/+ compared with Tau −/− mice is not significantly shifted (Figure 4D, Mann-Whitney U test: P > 0.05), which indicates that 1 completely blocks the synaptic impairment caused by the P301S mutation. Comparison of the cumulative curves of Tau −/+ neurons treated with each of the three compounds indicates that 1 shifted the cumulative curve of mEPSC amplitudes more to the right than 2b, further supporting that 1 is the more effective compound (Figure 4E, Mann-Whitney U test: P < 0.001).

Discussion and Summary

We designed and prepared 1, a new, potent, and selective Casp2 inhibitor, and found that it is significantly more effective than 2b, the P2 pure epimer of the previously reported Casp2 inhibitor 2, in restoring excitatory postsynaptic neurotransmission in a cellular model of tauopathy. We designed 1 with the following features in mind: potency at Casp2, selectivity at Casp2 versus Casp3 (discussed in Maillard, et al.¹⁶), and reversibility of the covalent reactive group (CRG), anticipating that we could fine-tune the warhead in later optimization work. Additionally, we sought physicochemical properties that were consistent with cell entry and an overall structure that contained moieties that would allow more careful optimization in future structure-activity studies. Thus, we incorporated several known selectivity elements in the design of 1 and were delighted to discover that our initial selection of features for the pentapeptide inhibitor²⁵ yielded a potent, selective compound. While ultimately the Casp2/Casp3 selectivity of 3 (CRG = -CHO) was several-fold greater than that of 1 (CRG = -CN), the 10-fold increase in potency offered by 1 outweighed its slightly diminished selectivity.

The first studies of Casp2 in the nervous system focused on its apoptotic roles in development¹⁰ and neurodegeneration caused by the withdrawal of trophic factors and the application of amyloid-β peptides^26,27. The presence of Casp2 in post-mitotic neurons in the adult brain indicated that it presumably also performs non-apoptotic roles, but it was not until quite recently that its role in the endocytosis of AMPA receptors in response to the expression of tau variants was discovered⁴. Subsequently, Xu et al. demonstrated the requirement of Casp2 for the endocytosis of AMPA receptors in response to the induction of long-term depression¹², echoing its pathological effects in tauopathy.

Although it was the search for genes mediating apoptosis that led to the discovery of caspases²⁸, most of the functions that the family of caspases perform are non-apoptotic⁹. Casp2 is unusual in that it can mediate some non-apoptotic processes through a non-catalytic mechanism. For example, dead Casp2 mutated at its catalytic cysteine negatively regulates autophagy¹⁵. The precise mechanism by which this occurs is unknown. One possibility involves the role of catalytically dead Casp2 as a translational co-factor to produce cyclin-dependent kinase inhibitor p21²⁹. p21, in turn, has been shown to negatively regulate autophagy in various cell types under different stress conditions^30,31. Therefore, it is possible that lowering the levels of Casp2, which was how Zhao et al. demonstrated its role in tauopathy⁴, might have rescued neurotransmission by non-catalytically enhancing autophagy and thereby lowering the levels of misfolded and aggregated forms of tau. Distinguishing between a catalytic versus non-catalytic mechanism underlying the pathological effects of Casp2 had not been done until the present studies. It is important to eliminate the possibility that its effects were non-catalytic, to solidify the rationale for developing small-molecule Casp2 inhibitors. Our current results, in combination with previous studies⁴, clearly and definitively demonstrate that Casp2 exerts its pathological effects on synaptic function catalytically, by hydrolyzing tau at aspartate-314.

Compound 1 rescues tau-induced synaptic dysfunction by blocking the excessive accumulation of tau in dendritic spines (Figures 3 and 4). This mechanism has been shown to mediate synaptic dysfunction repeatedly in models of tauopathy ranging from Alzheimer’s disease, Lewy body dementia, and FTDP-17 to chronic traumatic encephalopathy^1–5,7,8. The dysfunctional process begins when tau redistributes from axons to dendrites in response to various pathological triggers^32,33. Our current working model stipulates that the phosphorylation of serine-396 or serine-404 in the C-terminal tail of tau by GSK3β and/or CDK5 concomitantly with the cleavage of tau by Casp2 at aspartate-314 in the third microtubule domain^4,5,8 causes tau to accumulate in dendritic spines. A minority of dendritic spines normally contains tau³⁴, but under pathological conditions tau is found in the vast majority of dendritic spines. This occurs presumably because non-physiological, pathological triggers activate or elevate the levels of Casp2 and relevant kinases. The presence of tau in dendritic spines stimulates calcineurin-dependent endocytosis of AMPARs, but only if one or more of five specific residues in the proline-rich domain of tau are phosphorylated^1,2,5.

Elevated levels of Δtau314 and Casp2 found in brain specimens of individuals with Alzheimer’s disease^4,35, Lewy body dementia³⁶, and Huntington’s disease³⁷ underscore the relevance of the Casp2-tau signaling pathway in tauopathies. The results reported here using 1 establish the feasibility of using a small-molecule Casp2 inhibitor to restore excitatory neurotransmission and justify efforts to develop brain-penetrant Casp2 inhibitors to treat patients with dementia due to tauopathies.

Based on structural attributes of the previously reported Casp2 inhibitors, we identified a reversible Casp2 inhibitor 1 with an excellent selectivity profile. Subsequently, we confirmed in an in vitro assay that 1 effectively inhibits Casp2-mediated cleavage of tau and blocks the production of Δtau314, the N-terminal truncation product formed by hydrolysis at aspartate-314. Using cultured rat and mouse hippocampal neurons expressing P301S tau, we showed that 1 prevents the disproportionate accumulation of P301S tau in dendritic spines. Finally, we showed that the expression of P301S tau impairs postsynaptic function of excitatory synapses, determined by recording miniature excitatory synaptic currents (mEPSCs), and that 1 completely normalizes the mEPSCs and rescues synaptic function.

METHODS AND MATERIALS

General Methods and Materials (Chemistry)

All solution phase experiments were carried out under nitrogen in oven-dried glassware. All starting materials, solvents, and reagents were purchased from commercial sources and used without further purification. Aspartic acid-derived semi-carbazide Merrifield resin was prepared by using the previously reported procedure³⁸. The intermediates were purified by flash chromatography using 230–400 mesh silica gel. ¹H NMR spectra were recorded on Bruker 400 MHz spectrometer in deuterated solvents. Final compounds for the biological assays were purified by preparative HPLC. Chemical shifts are reported in parts per million (ppm). Final compounds 1 and 3 were purified by preparative-HPLC, utilizing a Waters Atlantis Zoraba × SB C18 column, gradient 0–100% B (A = water, B = acetonitrile) over 18 min, injection volume 1 mL, flow = 20 ml/min. UV spectra were recorded at 215 nm using a Gilson 119 UV/vis detector system. LCMS purity was determined on Acquity UPLC (Waters Corporation) equipped with an Acquity BEH UPLC C18 column (2.1 × 50 mm, 1.7 μm) for separation. Compound absorbance was detected at 214 nm using a photo diode array detector. Mass data was acquired using a Micromass ZQ mass spectrometer. LC was completed using a gradient method with each of the following mobile phase systems: Mobile Phase A2: 95% Water, 5% Acetonitrile, 0.1% Formic Acid; Mobile Phase B2: 95% Acetonitrile, 5% Water, 0.1% Formic Acid; Mobile Phase A1: 95% Water, 5% Methanol, 0.1% Formic Acid; Mobile Phase B1: 95% Methanol, 5% Water, 0.1% Formic Acid. Analytical HPLC-MS was performed on Shimadzu LCMS-2010 EV systems using reverse phase Atlantis C18 columns, gradient 5–100% B (A = water/0.1% formic acid, B = acetonitrile/0.1% formic acid) over 6 min, flow = 0.6 mL/min. Mass spectra were obtained over the range m/z 150–850 at a sampling rate of 2 scans per second using Waters LCT or analytical HPLC-MS systems using reverse phase Water Atlantis C18 columns gradient 5–100% B (A = water/0.1% formic acid, B = acetonitrile/0.1% formic acid).

For the compounds 2a and 2b, preparative HPLC was performed on a Knauer system (consisting of two K-1800 pumps, a K-2001 detector) using a Phenomenex Gemini C18 column (250 × 21 mm, 5 μm) at a flow rate of 15 mL/min. As a mobile phase, mixtures of MeCN and 0.1% TFA in water were used. The UV detection was carried out at 220 nm. Purity analysis of the compounds 2a and 2b was performed on an Agilent 1100 HPLC system (equipped with an Instant Pilot controller, a G1312A binary pump, a G1329A ALS autosampler, a G1379A vacuum degasser, a G1316A column compartment, and a G1315B diode array detector) using a Phenomenex Kinetex XB-C18 column (250 × 4.6 mm, 5 μm) at 30 °C oven temperature using mixtures of MeCN and 0.1% TFA in water as mobile phase. Absorbance was detected at 220 nm. Purity control for 2a and 2b was performed as recently reported by Bresinsky et al.¹⁸.

Racemic 6-methyl-tetrahydroisoquinoline-1-carboxylic acid (6-Me-1-THIQCOOH) was obtained from a commercial source and was outsourced for chiral resolution by supercritical fluid chromatography (SFC) (Figure S1). The enantiopure amino acids P1 and P2 were subjected to N-Fmoc protection to obtain (S)-6 and (R)-6 respectively (the absolute configurations of (S)-6 and (R)-6 were determined by comparison of the sign of the reported specific rotation of the commercially available enantiopure N-Fmoc-tetrahydroisoquinoline-1-carboxylic acids. Compounds 2a, 2b, and 3 were prepared via solid-phase peptide synthesis using aspartic acid modified semi-carbazide resin³⁸ as described by Maillard and Bresinsky^16,18.

Experimental data of compound 1 and 2b

(S)-6-(((S)-1-((S)-1-(((S)-2-carboxy-1-cyanoethyl)carbamoyl)-6-methyl-3,4-dihydroisoquinolin-2(1H)-yl)-3-methyl-1-oxobutan-2-yl)amino)-5-((S)-indoline-2-carboxamido)-6-oxohexanoic acid (1).

To a solution of tripeptide carboxylic acid 9 (267 mg, 0.39 mmol) in 5 mL acetonitrile were added HATU (102 mg, 0.39), 2,4,6-collidine (1.56 mmol, 0.21 mL), and 11 (100 mg, 0.26 mmol). The reaction mixture was stirred at rt for 6 h and the solvent was evaporated. The crude residue was diluted with water (20 mL) and extracted with ethyl acetate (4 × 20 mL). The combined organic layer was sequentially washed with sat. aq. NaHCO₃ (20 mL), ice-cold 10% aq. KHSO₄ (20 mL), and brine (20 mL). The organic layer was collected over anhyd. Na₂SO₄, filtered, and concentrated under reduced pressure to obtain the crude peptide. The crude residue was dissolved in a minimum volume of dichloromethane (3 mL) and transferred to vigorously stirring phosphoric acid (85%, 3 mL). The completion of the reaction was monitored by UPLC-MS. The reaction mixture was diluted with water (5 mL) and extracted with ethyl acetate (4 × 10 mL). The organic layer was dried over anhyd. Na₂SO₄, filtered, and concentrated to obtain the crude carboxylic acid. The crude product was dissolved in acetonitrile (10 mL) and diethyl amine (3 mL) was added to it. The reaction mixture was stirred at rt for 3 h and was concentrated under reduced pressure. The crude residue was triturated sequentially with hexane, diethyl ether, and the solvent decanted. The residue was dried under reduced pressure and dissolved in DMF (3 mL). Palladium (10%, 10 mg) was added, and the reaction mixture was purged with hydrogen gas for 5 minutes and stirred under hydrogen (1 atm). The completion of the reaction was monitored by UPLC-MS. The reaction mixture was filtered over a thin layer of celite. The celite layer was washed with 2 mL of DMF. The solvent was evaporated and the residue was purified by preparative HPLC. Lyophilization of the pure fractions provided the product 1 as a colorless amorphous powder (38 mg, 22% yield). ¹H NMR (400 MHz, MeOD) δ 9.02 − 8.72 (m, 0.5H), 8.43 (m, 0.7H), 8.13 (m, 1H), 7.64 − 7.29 (m, 1H), 7.25 − 6.43 (m, 7H), 5.58 (s, 0.6H), 5.40 (s, 0.4H), 4.63 − 4.15 (m, 3H), 4.07 − 3.83 (m, 1H), 3.81 − 3.61 (m, 1H), 3.60 − 3.3 (m, 1H), 3.13 − 2.45 (m, 5H), 2.38 − 1.96 (m, 5H), 1.95 − 1.08 (m, 5H), 0.99 − 0.67 (m, 5H), 0.55 (d, J = 6.8 Hz 1H). HRMS (ESI) m/z [M+H⁺] calculated for C₃₅H₄₃N₆O₈⁺: 675.3137, found 675.3. HPLC purity 99%.

(S)-3-((S)-2-acetamido-3-methylbutanamido)-4-(((S)-1-((S)-1-(((S)-1-carboxy-3-oxopropan-2-yl)carbamoyl)-6-methyl-3,4-dihydroisoquinolin-2(1H)-yl)-3-methyl-1-oxobutan-2-yl)amino)-4-oxobutanoic acid (2b)

The title compound was synthesized according to the general procedure described from Bresinsky et al.³ using 300 mg Asp-resin (0.188 mmol) and yielding a fluffy white solid (43 mg, 36% yield). ¹H NMR (400 MHz, D⁶-DMSO) δ9.51 − 8.73 (m, 1H), 8.29 − 8.22 (m, 1H), 7.89 (d, J = 8.4 Hz, 1H), 7.73 − 7.60 (m, 1H), 7.47 − 7.16 (m, 1H), 7.09 − 6.92 (m, 2H), 5.97 − 5.16 (m, 2H), 4.80 − 3.33 (m, 6H), 3.06 − 2.43 (m, 5H), 2.75 (s, 3H), 2.10 − 1.876 (m, 2H), 1.85 (s, 3H), 0.97 − 0.58 (m, 12H). HRMS (ESI) m/z [M+H⁺] calculated for C₃₁H₄₄N₅O₁₀⁺: 646.3083, found 646.3096. HPLC purity >99%.

Animal Care and Usage

Rat and mouse pups in postnatal day 0 or 1 were harvested to make primary hippocampal cultures. Sprague-Dawley timed-pregnancy adult rats were housed in facilities of Research Animal Resources (RAR) at the University of Minnesota (UMN) after being delivered from Envigo (www.envigo.com). The rats were fed a diet of regular chow before giving birth. Rat pups were euthanized by decapitation to harvest brain tissues in postnatal day 0 or 1. The appropriate transgenic mice were also housed and bred in RAR facilities at UMN. All work was conducted in accordance with the American Association for the Accreditation of Laboratory Animal Care and Institutional Animal Care and Use Committee (IACUC) at the University of Minnesota (protocol #1211A23505). We performed all procedures of euthanasia strictly according to the guidelines of the IACUC at the UMN.

Materials (Biology)

All common chemical reagents and cell culture supplies were purchased from Sigma-Aldrich, Fisher Scientific, Promega, and Thermo-Fisher Scientific/Invitrogen/Life Technologies unless otherwise indicated.

In vitro caspase-2-catalyzed tau cleavage assay

Purified recombinant Casp2 (final concentration during pre-incubation: 67 μM) was pre-incubated with various concentrations of inhibitory compounds at 4 °C for 72 h. Pre-treated enzyme was then incubated with purified recombinant tau at a molar ratio of 1:1 in 37 °C water bath, 1x reaction buffer (25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 0.1% (w/v) 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS), 10 mM dithiothreitol (DTT), pH 7.5) for 7 h. The final volume of each reaction was 100 μL. At the end of the 7-h incubation, 0.1% (v/v) protease inhibitor cocktail (MilliporeSigma) was added to stop the reactions.

Immunoprecipitation (IP) / Western blotting (WB)

Immediately after the reaction was stopped, the 100 μL enzyme-product mixture was diluted in 400 μL IP buffer (50 mM Tris-HCl (pH 7.4) and 150 mM NaCl, containing 0.1 mM phenylmethylsulfonyl fluoride, 0.2 mM 1,10-phenanthroline monohydrate, and protease inhibitor cocktail (MilliporeSigma)), and then incubated with 10 μg of Δtau314-specific monoclonal antibody 4F3 and 50 μL of Protein G Sepharose 4 Fast Flow resin (GE Healthcare) at 4 °C for 14–16 h. Subsequent resin wash and protein elution were performed as described previously³⁹. WB was performed according to a previously published protocol³⁷.

Mass spectrometry (MS)

In-gel trypsin digestion, liquid chromatography-MS/MS, mass spectral database search, and data interpretation were performed as previously described³⁵, except that Peaks Studio Xpro (Bioinformatics Solutions, Inc, Waterloo, Ontario, Canada) was used for interpretation of mass spectra.

Electrophysiology

Miniature excitatory postsynaptic currents (mEPSCs) were recorded from cultured dissociated mouse hippocampal neurons at 17–21 DIV with a glass pipette (resistance ~ 5 MΩ) at holding potentials of −65 mV on an Axopatch 200B amplifier (output gain = 0.5; filtered at 1 kHz) as we previously described⁵. Input and series resistances were assessed and found to have no significant difference before and after recording time (5–20 mins). Recording sweeps lasted 200 ms and were sampled for every 1 s (pClamp, v10, RRID:SCR_011323). Neurons were bathed in bubble-oxygenated artificial cerebrospinal fluid (ACSF) at 23 °C with 100 μM APV (NMDA receptor antagonist), 1 μM TTX (sodium channel blocker), and 100 μM picrotoxin (GABAa receptor antagonist). Passive oxygen perfusion was established with medical-grade 95% O₂–5% CO₂. ACSF contained (in mM) 119 NaCl, 2.5 KCl, 5.0 CaCl₂, 2.5 MgCl₂, 26.2 NaHCO₃, 1 NaH₂PO₄, and 11 D-glucose. The internal solution of the glass pipettes contained (in mM) 100 cesium gluconate, 0.2 EGTA, 0.5 MgCl₂, 2 ATP, 0.3 GTP, and 40 HEPES. The pH of the internal solution was normalized to 7.2 with cesium hydroxide and diluted to a trace osmotic deficit in comparison to ACSF (~300 mOsm). All analysis of recordings was performed using an automated detection software suit (Clampfit, 11.0.3, Molecular Devices, San Jose, CA, USA). Minimum analysis parameters were set at greater than 1 min stable recording time and events with amplitudes greater than 3 pA and smaller than 40 pA were included. A mEPSC event was identified by using a template which included a distinct fast-rising depolarization and slow-decaying repolarization. Combined individual events were used to form relative cumulative frequency curves; whereas the means of all events from individual recordings were treated as single samples for further statistical analysis. Example traces were exported from Clampfit and live-traced, simplified, and united in vector editing software (Adobe Illustrator CS5 and Affinity Designer).

DATA AVAILABILITY STATEMENT

All raw and quantified data will be made available by the authors upon reasonable request.