K_v1.3 inhibition as a potential microglia-targeted therapy for Alzheimer’s disease: preclinical proof of concept

Abstract

Microglia significantly contribute to the pathophysiology of Alzheimer’s disease but an effective microglia-targeted therapeutic approach is not yet available clinically. The potassium channels K_v1.3 and K_ir2.1 play important roles in regulating immune cell functions and have been implicated by in vitro studies in the ‘M1-like pro-inflammatory’ or ‘M2-like anti-inflammatory’ state of microglia, respectively. We here found that amyloid-β oligomer-induced expression of K_v1.3 and K_ir2.1 in cultured primary microglia. Likewise, ex vivo microglia acutely isolated from the Alzheimer’s model 5xFAD mice co-expressed K_v1.3 and K_ir2.1 as well as markers traditionally associated with M1 and M2 activation suggesting that amyloid-β oligomer induces a microglial activation state that is more complex than previously thought. Using the orally available, brain penetrant small molecule K_v1.3 blocker PAP-1 as a tool, we showed that pro-inflammatory and neurotoxic microglial responses induced by amyloid-β oligomer required K_v1.3 activity in vitro and in hippocampal slices. Since we further observed that K_v1.3 was highly expressed in microglia of transgenic Alzheimer’s mouse models and human Alzheimer’s disease brains, we hypothesized that pharmacological K_v1.3 inhibition could mitigate the pathology induced by amyloid-β aggregates. Indeed, treating APP/PS1 transgenic mice with a 5-month oral regimen of PAP-1, starting at 9 months of age, when the animals already manifest cognitive deficits and amyloid pathology, reduced neuroinflammation, decreased cerebral amyloid load, enhanced hippocampal neuronal plasticity, and improved behavioural deficits. The observed decrease in cerebral amyloid deposition was consistent with the in vitro finding that PAP-1 enhanced amyloid-β uptake by microglia. Collectively, these results provide proof-of-concept data to advance K_v1.3 blockers to Alzheimer’s disease clinical trials.

Introduction

Although several drugs for symptoms of Alzheimer’s disease have now been approved, they do not alter disease progression and their benefits are at most modest, indicating an urgent need for new target discovery. Multiple lines of evidence have implicated a significant role of microglia and associated neuroinflammation in Alzheimer’s disease. A generally accepted hypothesis states that activated microglia release cytotoxic substances and pro-inflammatory cytokines that cause neuronal damage and aggravate Alzheimer’s disease pathology (Heneka et al., 2015). A network-based analysis of whole-genome gene-expression profiling and genotypic data obtained from 1647 Alzheimer’s disease and non-demented brain samples highlighted the immune/microglia module as the molecular system most strongly associated with the pathophysiology of Alzheimer’s disease (Zhang et al., 2013). An even stronger correlation was recently found between a similar immune/microglial module and Alzheimer’s disease with large-scale protein co-expression network analysis (Seyfried et al., 2017). Immune receptor genes expressed in microglia or myeloid cells, such as TREM2 and CD33, are among the few genes that impose moderate risk for the development of Alzheimer’s disease (Karch et al., 2014). Despite this progress, a microglia-targeting therapeutic agent has not been available clinically. Results from several clinical trials using generic anti-inflammatory agents have been disappointing, mainly due to inadequate brain penetration of existing non-steroid anti-inflammatory drugs, suboptimal doses, unknown molecular targets, and non-specific activities (Rogers, 2008). A large-scale prevention trial with naproxen and celecoxib was stopped early because of drug safety concerns. These setbacks, however, should prompt investigations to develop novel anti-inflammatory agents with known specific targets, better brain penetrance, and low toxicities.

The voltage-gated potassium channel K_v1.3 plays an important role in immune cell activation by modulating Ca²⁺ signalling (Wulff et al., 2007; Feske et al., 2015). Through K⁺ efflux, K_v1.3 helps maintain a negative membrane potential, which provides the driving force for Ca²⁺ entry through the store-operated inward-rectifier calcium channels like the Ca²⁺ release activated Ca²⁺ channel ORAI, or transient receptor potential cation channels. Accordingly, K_v1.3 blockers have been demonstrated to inhibit microglia-mediated neurotoxicity in culture (Fordyce et al., 2005) and to protect mice from microglia-mediated radiation-induced brain injury in vivo (Peng et al., 2014). Interestingly, blockade of K_v1.3 appears to specifically inhibit certain signal pathways but not others in the molecular cascade leading to microglia activation (Fordyce et al., 2005). A systems pharmacology-based study identified functional roles for K_v1.3 in pro-inflammatory microglial activation (Rangaraju et al., 2017) while electrophysiological studies by our group demonstrated upregulated K_v1.3 expression in microglia acutely isolated from the infarct areas of mice subjected to experimental stroke (Chen et al., 2016). In vitro studies with primary microglia further demonstrated association of high K_v1.3 expression with the lipopolysaccharide (LPS)-induced ‘M1-like’ activation and inflammatory cytokine production (Nguyen et al., 2017). Interestingly, in ‘alternatively activated M2-like’ microglia, such as those activated by interleukin 4 (IL-4), K_v1.3 expression was silenced and instead the inward-rectifier K_ir2.1, another K⁺ channel often expressed in myeloid cells and involved in regulating their function (Kettenmann et al., 2011; Feske et al., 2015), was highly upregulated (Nguyen et al., 2017). K_v1.3 inhibition, therefore, could be used to fine-tune the pattern/state of microglia activation as an anti-inflammatory approach. K_v1.3 expression has also been described in some neurons. However, in neurons K_v1.3 is typically part of heteromultimers with K_v1.1, K_v1.2 and K_v1.6 subunits (Helms et al., 1997) and has a different pharmacology distinct from the K_v1.3 homotetramers found in T cells, microglia and macrophages (Chandy et al., 2004; Wulff et al., 2009). Therefore, a relatively selective pharmacological targeting of K_v1.3 in microglia is feasible. We have developed a highly specific K_v1.3 blocker PAP-1 [5-(4-phenoxybutoxy)psoralen; Schmitz et al., 2005] and demonstrated that this small molecule effectively prevents autoimmune diabetes in a rat model and treats psoriasis in a humanized xenograft model in mice (Beeton et al., 2006; Kundu-Raychaudhuri et al., 2014). Moreover, we have shown that the expression of K_v1.3 is elevated in human Alzheimer’s disease brains and the elevation is limited to microglia, suggesting that K_v1.3 is a pathologically relevant microglial target in Alzheimer’s disease (Rangaraju et al., 2015). Here we examined the effects of K_v1.3 blockade in cellular and mouse models of Alzheimer’s disease, using the brain penetrant PAP-1 as a pharmacological tool.

Materials and methods

Study approval

All protocols involving mouse models were approved by the Institutional Animal Care and Use Committee of the University of California Davis. Human brain samples were drawn from the brain repositories of University of California Davis Alzheimer’s Disease Center and Emory University Alzheimer’s Disease Research Center. The collection and use of brain samples were approved by the respective Institutional Review Board.

Mice

C57BL/6 and APP/PS1 [B6.Cg-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax] mice were originally purchased from the Jackson laboratory. The line Tg6799 5xFAD mice was obtained from Dr Robert Vassar at Northwestern University.

Tissue cultures and acute isolation of microglia

Primary microglia were prepared from mixed glia cultures established from newborn mice with the ‘shaking-off’ method as described (Maezawa et al., 2011). Hippocampal slice cultures were prepared from 7-day-old C57BL/6 mice as described (Maezawa et al., 2011) and cultured for 10 days in vitro before use.

Microglia were acutely isolated from adult brains without culturing as described (Jin et al., 2015). Briefly, brains were dissociated enzymatically with a Neural Tissue Dissociation Kit (Miltenyi Biotec). Microglia were subsequently purified by the magnetic-activated cell sorting using anti-CD11b magnetic beads (Miltenyi Biotec).

Electrophysiology on cultured and acutely isolated microglia

Whole-cell patch-clamp experiments on cultured or acutely isolated microglia were performed as previously described (Chen et al., 2016; Nguyen et al., 2017). For details, see Supplementary material.

Production and administration of amyloid-β oligomer

Amyloid-β oligomer (AβO) was prepared following a standard procedure (Lambert et al., 1998) except that the amyloid-β_1-42 peptide (Bachem) was diluted with Opti-MEM® culture medium instead of the F12 medium originally described, before incubation at 4°C for 24 h to generate oligomers. Our preparation has been extensively characterized (Maezawa et al., 2006, 2008). To ensure consistency of quality, a random sample from each batch was imaged using atomic force microscopy to characterize the size and shape of the aggregates, and the biological activities were confirmed by determining AβO’s neurotoxic activity and ability to rapidly induce exocytosis of MTT formazan, as described (Maezawa et al., 2006, 2008; Hong et al., 2007). For treatment, primary microglia cells were plated onto 6-well plates (5 × 10⁵ cells/well) or 12-well plates (3 × 10⁵ cells/well) in Dulbecco’s modified Eagle medium (DMEM) with 10% foetal bovine serum. After 24 h, culture medium was changed to Opti-MEM® and AβO was freshly diluted and added at indicated concentrations with and without PAP-1. The cultures were incubated for another 24 h before used for quantitative polymerase chain reaction (qPCR) and western blotting.

Quantitative PCR

Total RNA from primary microglia, acutely isolated microglia, and tissue samples were extracted using RNeasy® Plus Mini Kit (Qiagen) or RNeasy® Plus Universal Mini Kit (Qiagen). RNA samples from acutely isolated microglia were further reverse-transcribed and pre-amplified as we previously described (Horiuchi et al., 2017). The result was normalized to β-actin. The primer sequences used are listed in Supplementary Table 1. For β-actin the commercially available primer set Mouse-ACTB (Applied Biosystems) was used. Relative cDNA levels for the target genes were analysed by the 2-ΔΔCt method using Actb as the internal control for normalization.

Immunohistochemical staining and quantification

Immunohistochemistry using Vectastain Elite ABC and diaminobenzidine (DAB) on formalin-fixed, paraffin-embedded human brain sections and quantification was performed as previously described (Rangaraju et al., 2015). Briefly, for quantification using ImageJ, the raw images were deconvoluted using the deconvolute plugin into brown (DAB) and haematoxylin channels. Then, the images were thresholded and converted to binary images for quantitative analysis. The K_v1.3 immunoreactivity in each image was estimated as the proportion of the image area occupied by Iba1 positive microglia. From the Alzheimer’s disease and non-Alzheimer’s disease control brains, three adjacent slides were immunostained, and from each section, five random images were obtained for analysis using the same settings.

Immunofluorescence staining and quantification was performed as previously described (Jin et al., 2015). Briefly, microglia cells cultured on cover slip, hippocampal slices and frozen brain sections (20 µm) were fixed in 4% paraformaldehyde and stained with anti-Aβ1-42 (1:200; Millipore), anti-CD68 (1:200; Serotec), anti-CD11b (1:200; AbD Serotec), anti-Iba-1 (1:200, Wako Chemical), anti-K_v1.3 (1:200, NeuroMab), or anti-NeuN (1:200, Cell Signaling) overnight at 4°C, followed by respective Alexa Fluor®-conjugated secondary antibodies (1:700; Invitrogen). The amyloid deposits were also detected by 400 nM FSB. Immunostained slides were imaged under a Nikon Eclipse E600 microscope and photographed by a digital camera (SPOT RTke, SPOT Diagnostics) or under an Olympus FV1000 spectrum confocal microscope. Photomicrographs were randomly taken in a blind fashion from each culture condition or within a defined anatomic region. The images were transformed to 8-bit grayscale and analysed by the ImageJ program. The photography and analysis of immunoreactivity were conducted in an investigator-blinded manner.

Formalin-fixed, paraffin-embedded human tissue sections were immunostained with anti-K_v1.3 (1:150; NeuroMab) overnight at 4°C, followed by anti-Iba-1 (1:200; Wako) for 3 h at room temperature, and then Alexa Fluor® 594-conjugated anti-rabbit antibody (1:700) and Alexa Fluor® 488-conjugated anti-mouse antibody (1:700), respectively. Sections were then incubated with 0.1% Sudan black for 10 min to eliminate autofluorescence, followed by 400 nM FSB for 20 min.

Propidium iodide uptake, Fluoro-Jade® C assay, ELISA, NFκB assay, and nitric oxide quantification

The propidium iodide (PI) uptake assay to assess neuronal damage, ELISA quantification of cytokines, nitric oxide (NO) quantification, and NFκB assay were conducted as previously described (Maezawa et al., 2011). For the Fluoro-Jade® C assay, fixed hippocampal slices were placed onto gelatin coated slides and stained with Fluoro-Jade® C (Histo-Chem Inc.) according to the manufacturer’s instruction.

Induction of hippocampal long-term potentiation by high frequency stimulation

The preparation of mouse hippocampal slices, and the induction of long-term potentiation (LTP) by high frequency stimulation of the Schaffer collateral afferents were conducted as we previously described (Maezawa, 2017) and are described in detail in the Supplementary material.

Cell/tissue homogenate preparation and western blot analysis

To obtain cell lysates, cells were washed with ice-cold PBS and incubated with a lysis buffer (150 mM NaCl, 10 mM NaH₂PO₄, 1 mM EDTA, 1% Triton^™ X-100, 0.5% SDS) with protease inhibitor cocktail and phosphatase inhibitor (Sigma). Equivalent amounts of protein were analysed by 4–15%Tris-HCl gel electrophoresis (Bio-Rad). Proteins were transferred to polyvinylidene difluoride membranes and probed with antibodies. Visualization was enabled using enhanced chemiluminescence (GE Healthcare Pharmacia). The following primary antibodies (dilutions) were used: anti-K_v1.3 (1:700, NeuroMab), anti-phosphor-p38MAPK (1:1000, Cell Signaling), anti-p38MAPK (1:1000, Cell Signaling), and anti-GAPDH (1:2000, Cell Signaling). Secondary antibodies were HRP-conjugated anti-rabbit or anti-mouse antibody (1:1000, GE Healthcare).

Brain tissue samples from APP/PS1 treated with PAP-1 or control diet were fractionated into Tris-buffered saline (TBS)-soluble and TBS-insoluble, SDS-soluble fractions to assess the solubility of amyloid-β species. Briefly brain tissues were homogenized in TBS with protease inhibitors, followed by centrifugation (100 000g, 30 min, at 4°C). The supernatants were collected as the TBS-soluble fraction. The pellets were homogenized in 2% SDS with protease inhibitors, followed by centrifugation (100 000g, 30 min, at 4°C). The supernatants were collected as the TBS-insoluble, SDS-soluble fraction.

Neurobehavioural tests

The neurobehavioural tests, including open field test, novel object recognition test, and step-through passive avoidance test, were conducted in a fashion in which the scorers were blind to the treatment of the mice. The detailed methods are described in the Supplementary material.

Stereotactic intrahippocampal injection

Mice were first anaesthetized using 3% isoflurane and then restrained onto a stereotaxic apparatus. A small incision was made to expose the skull and a 1.0 mm burr hole was drilled. AβO (40 μM in artificial CSF), or vehicle (artificial CSF), was injected using a Hamilton syringe with a 27-gauge needle into the hippocampus by using the following coordinates (mm from bregma): −1.8 anterior/posterior, ±1.7 medial/lateral, and −1.9 dorsal/ventral, at a rate of 0.2 μl/min, 2 μl each side. Before waking, mice received 0.05 mg/kg of buprenorfine subcutaneously and were allowed to recover on an isothermal pad at 36°C.

PAP-1 exposure levels

Total PAP-1 plasma and brain concentrations were determined by LC-MS analysis using a Waters Acquity UPLC (Waters) interfaced to a TSQ Quantum Access Max mass spectrometer (Thermo Fisher Scientific) as described in the Supplementary material.

Statistics

All results were expressed as the mean ± standard error (SE). Normality of the data distribution was determined using the Shapiro-Wilk test. For group comparisons in means, paired or unpaired Student’s t-test or one- or two-way ANOVA, as appropriate, were conducted using the SigmaStat 3.1 (Systat Inc.). When the overall ANOVA was significant, Bonferroni’s multiple comparison procedure was carried out to maintain the family-wise error rate at 0.05. The significance level for the two-sided analyses was set at P < 0.05.

Results

Enhanced K_v1.3 and K_ir2.1 expression and activity in AβO-stimulated microglia

We previously showed that AβO, composed of amyloid-β_1-42 peptide, at low (5–200) nanomolar concentrations is a robust microglia activator (Maezawa et al., 2011). We also showed that functional polarization of microglia induces a differential K⁺ channel expression pattern—while the classic M1-polarizing LPS induces high current densities of K_v1.3 and no K_ir2.1 current, the M2-polarizing IL-4 induces large K_ir2.1 currents and no K_v1.3 current (Nguyen et al., 2017). As these two major microglial K⁺ channels regulate Ca²⁺ signalling instrumental for differential activation profiles, we investigated their expression/activity in cultured primary microglia treated with AβO. Similar to LPS, AβO significantly upregulated K_v1.3 at both transcript (Fig. 1A) and protein (Fig. 1B and C) levels. Whole-cell voltage-clamp revealed outward rectifying currents induced by AβO, with biophysical properties characteristic of homotetrameric K_v1.3 channels typically expressed on immune cells (Fig. 1D), including a half-activation (V_1/2) value of −28.4 mV in the Boltzmann fit of normalized peak currents (Fig. 1E) and use-dependent inactivation elicited by repetitive depolarization steps (Fig. 1F). Pharmacological characterization showed that the currents were sensitive to the K_v1.3 specific blockers ShK-186 (Tarcha et al., 2012), margatoxin (MgTX; Garciacalvo et al., 1993), and PAP-1 (Fig. 1G), with IC₅₀ values of 68.5 pM, 79.7 pM, and 6.5 nM, respectively, in good agreement with existing literature. AβO induced a significant increase in K_v1.3 current density compared to unstimulated microglia (Fig. 1H), consistent with the increased K_v1.3 protein level as shown by western blotting (Fig. 1B) and immunocytochemistry (Fig. 1C). Interestingly, unlike LPS, AβO also induced an increase in Ba²⁺-sensitive K_ir2.1 current density (Fig. 1I) that was similar in magnitude to the IL-4-induced K_ir2.1 current increases (Nguyen et al., 2017), suggesting that AβO induces a microglial activation state that is more complex than previously defined M1- or M2-like polarization. Consistent with this notion, AβO induced microglial expression of IL-1β, TNF-α, and IL-6 traditionally associated with LPS-induced polarization as well as CD86, Arg1, and CD206 traditionally associated with IL-4-induced polarization (Supplementary Fig. 1).

K_v1.3 is required for amyloid-β oligomer-induced microglial activation and microglial neurotoxicity

Compared to K_ir2.1, the pharmacological tools for K_v1.3 are far better developed; therefore, we focused our study on the AβO-induced K_v1.3 upregulation. We previously showed that AβO induces an activated microglial profile, including proliferation, activation of p38MAPK and NF-κB, NO production, as well as microglia-mediated neurotoxicity (Maezawa et al., 2011). This profile can be induced at concentrations ∼200-fold lower than those required for direct neuronal killing, suggesting that early accumulation of AβO in the preclinical stages of Alzheimer’s disease could cause significant indirect, microglia-mediated neurotoxicity. To determine if K_v1.3 activity is required for AβO-induced microglial activation and associated neurotoxicity, we used the small-molecule K_v1.3 inhibitor PAP-1 as a pharmacological probe. In some experiments, the general microglial activation inhibitor doxycycline was used for comparison. PAP-1 blocked AβO-induced microglia proliferation (Fig. 2A) and expressions of pro-inflammatory mediators (Fig. 2D) in a dose-dependent manner. PAP-1 also blocked AβO-induced NF-κB activation (Fig. 2B), NO generation (Fig. 2C), and p38MAPK phosphorylation (Fig. 2E), suggesting that the microglial activating effects of AβO require K_v1.3 activity. We further tested K_v1.3 inhibition using hippocampal slice cultures, a model allowing microglia and neurons to interact. We previously showed that in this model system, AβO caused substantial microglial activation and associated neurotoxicity mediated by NO, released by activated microglia (Maezawa et al., 2011). Figure 2F shows that AβO treatment of hippocampal slices enhanced CD11b immunoreactivity as well as neurotoxicity, indicated by increased PI uptake and Fluoro-Jade® C labelling, and decreased NeuN labelling (Supplementary Fig. 2). PAP-1 co-incubation substantially mitigated the above changes, similar to doxycycline (Fig. 2F and Supplementary Fig. 2). PAP-1 treatment also significantly reduced the NO level in the hippocampal culture medium [AβO/vehicle: 16.1 ± 0.42 (means ± SE) versus AβO/PAP-1: 6.73 ± 0.50 nM/μg total protein, n = 3, P < 0.001], consistent with the result obtained from isolated microglia cultures (Fig. 2C).

It was reported that minocycline, a tetracycline inhibitor of microglial activation, and inducible nitric oxide synthase (iNOS) reduction, prevented hippocampal LTP disruption by amyloid-β (Wang et al., 2004), suggesting that microglial activation and NO production contribute to amyloid-β synaptotoxicity. However, the microglial contribution to hippocampal LTP is by no means conclusive in view of the pleotropic, non-microglial effects of minocycline (Griffin et al., 2010). Because PAP-1 selectively targets microglial K_v1.3 and inhibits AβO-induced microglial neurotoxicity as well as NO production in hippocampal slices, we tested whether PAP-1 was able to rectify hippocampal LTP deficit using the same induction paradigm as Wang et al. (2004). AβO application to the slices blocked hippocampal LTP induction at 20–100 nM, within the range of concentrations that enhanced the expression of microglial K_v1.3. Significantly, in the presence of PAP-1, hippocampal LTP induction was not impaired by AβO and was recovered to the control (non-AβO, vehicle-treated) level (Fig. 2G and H). This result is consistent with the notion that AβO at low nanomolar concentrations impairs hippocampal LTP through K_v1.3-dependent microglial activation.

Elevated expression of microglial K_v1.3 in brains of Alzheimer’s transgenic models and Alzheimer’s disease patients

To validate the in vivo significance of K_v1.3 in Alzheimer’s disease, we examined brains from the widely used Alzheimer’s mouse model 5xFAD (Oakley et al., 2006), in which AβO is the major toxic amyloid-β species able to activate microglia to contribute to neuroinflammation (Hong et al., 2009; Xiao et al., 2013; Heneka et al., 2015). Immunofluorescence showed enhanced K_v1.3 immunoreactivity in microglia in 5xFAD mice compared to non-transgenic wild-type littermates, especially in areas around amyloid plaques, highlighted by the amyloid dye FSB (Fig. 3A and Supplementary Fig. 3). The K_v1.3-positive areas co-localized with areas immunoreactive to CD11b and Iba-1, markers for myeloid cells (Fig. 3A and B), consistent with our previous observations that K_v1.3 is typically expressed highest in microglia in the brain (Rangaraju et al., 2015). Similar results were obtained using the other Alzheimer’s transgenic model called APP/PS1 (Jankowsky et al., 2004) (Supplementary Fig. 3).

To confirm that the enhanced K_v1.3 expression translates to enhanced channel activity, we acutely isolated microglia from adult 5xFAD and wild-type littermate mice and performed voltage-clamp recordings. The procedure to isolate microglia in general took ∼1 h, and the cells, without further culturing, were immediately used for recording. While microglia from wild-type littermates showed minimal K_v1.3 current density, those from 4, 6, and 10 month-old 5xFAD mice showed significantly increased K_v1.3 channel activity (Fig. 3C–E). K_ir2.1 in microglia of 5xFAD mice also showed sustained high current density at 4, 6, and 10 months of age, reminiscent of the result seen in AβO-treated cultured microglia. The K_ir2.1 activity was significantly higher in 5xFAD microglia than in wild-type microglia at 4 and 10 months but not at 6 months due to a comparable high activity in wild-type microglia at that age (Fig. 3C, D and F). Interestingly, at 15 months when amyloid plaques were still abundant, the K_v1.3 and K_ir2.1 current densities in 5xFAD microglia were both reduced and no longer significantly different from wild-type microglia (Fig. 3E and F). Quantification of the level of the K_v1.3 transcript in acutely isolated microglia also showed age-dependent changes in K_v1.3 expression parallel to its channel activity (Supplementary Fig. 4). A similar increase of microglial K_v1.3 transcript was also observed in APP/PS1 mice (Fig. 6E).

To verify the in vivo significance of microglial K_v1.3 in human Alzheimer’s disease, our previous immunohistochemical study showed that K_v1.3 is almost exclusively expressed in microglia in post-mortem human brains and that K_v1.3 expression is substantially increased and associated with amyloid-β plaques in pathologically confirmed Alzheimer’s disease cases in the frontal cortex (Rangaraju et al., 2015). We extended this observation to the entorhinal cortical region, where Alzheimer’s disease pathologies are robust in the early stage Alzheimer’s disease. Again, K_v1.3 immunoreactivities were significantly increased in Alzheimer’s disease patients (Fig. 4A and B). Combined immunofluorescence using FSB, anti-Iba1, and anti-K_v1.3 showed that K_v1.3 was localized to a majority of Iba1-immunoreactive microglial cell bodies or processes (Fig. 4C, arrowheads), and its immunoreactivities were accentuated around amyloid-β plaques (Fig. 4A and C). The increase in K_v1.3 protein level was in part due to increased K_v1.3 transcript level in Alzheimer’s disease brains (Fig. 4D).

In vivo effects of K_v1.3 blockade by PAP-1 in microglia activation, hippocampal long-term potentiation induction, and memory deficit

Because PAP-1 mitigated the effects of AβO on hippocampal LTP in acutely prepared slices (Fig. 2G and H), we asked if PAP-1 could restore the AβO-induced, hippocampus-dependent memory deficit in vivo. To administer PAP-1 to mice, we used oral regimens to mimic treatments in humans, but refrained from oral gavage to avoid undue stress to the animals. PAP-1 dissolved in a mix of the semisynthetic triglyceride miglyol and condensed milk was quickly consumed by mice and a dose of 100 mg/kg p.o. resulted in high micromolar PAP-1 levels in plasma and several brain regions (Fig. 5A). For example, total hippocampal tissue concentrations peaked at 23 µM at 2 h and exceeded 2 µM for 15 h following ingestion, which, considering PAP-1’s IC₅₀ of 2 nM for K_v1.3 inhibition and its protein binding of 98%, was sufficient to exert a pharmacodynamic effect (Schmitz et al., 2005; Beeton et al., 2006). We found that mice receiving bilateral intrahippocampal injection of AβO (IH-AβO, 2 μl of 40 μM solution) performed poorly in the step-through passive avoidance test given on post-injection Day 8, compared to mice receiving bilateral intrahippocampal injections of vehicle (IH-Vehicle) only. Daily oral PAP-1 (100 mg/kg p.o.) treatment recovered this memory deficit to the control level (Fig. 5B), suggesting that K_v1.3 blockade could potentially ameliorate AβO-induced memory decline in Alzheimer’s disease.

To further translate our findings into a clinically relevant context, we intended to test the effect of chronic PAP-1 treatment on the behavioural phenotype of Alzheimer’s transgenic models. Unfortunately, we could not demonstrate reproducible memory deficits in 5xFAD mice at multiple ages. Therefore, we used the APP/PS1 model, which demonstrated memory deficits after 8–9 months of age, for this trial. Although compared to 5xFAD mice, APP/PS1 mice showed less robust and a later onset of Alzheimer’s disease-like amyloid pathology, the microglia of APP/PS1 mice were shown to lose amyloid-β-clearing capabilities at around 8 months of age, preceding the increase in amyloid-β accumulation and coincident with the beginning of measurable behavioural deficits (Hickman et al., 2008). To test if K_v1.3 inhibition by PAP-1 may affect the ability of microglia to clear amyloid-β, we treated cultured primary microglia with 50 nM Alexa Fluor® 488-labelled AβO and determined AβO uptake by flow cytometry. As expected, AβO uptake was reduced by an antibody 2F8 that blocks scavenger receptor A (SRA), a known microglial amyloid-β receptor (El Khoury et al., 1996), and by blocking the conformational epitopes of AβO using Congo red (Maezawa et al., 2008) or an AβO-specific antibody A11 (Kayed and Glabe, 2006) (Fig. 6A). Interestingly, compared to doxycycline, a known microglia activation inhibitor, which substantially inhibited AβO uptake, PAP-1 enhanced the ability of microglia to uptake AβO (Fig. 6A). This result suggests that rather than globally suppressing microglial function, K_v1.3 inhibition not only preserves but may enhance the beneficial amyloid-β clearance capability. This result also suggests that PAP-1 may recover the ability of microglia to clear amyloid-β in APP/PS1 mice.

We next tested if PAP-1 affects the behavioural phenotype and amyloid deposition in APP/PS1 mice. PAP-1 was administered via ad libitum consumption of PAP-1-medicated diet (ppm = 1100), starting at 9 months of age for a course of 5 months. Mice receiving chronic PAP-1 treatment were well groomed and displayed no observable differences compared to wild-type littermates. PAP-1 treatment significantly improved the performance of APP/PS1 mice in novel object recognition (Fig. 6B) and step-through passive avoidance tests (Fig. 6C). The open field test showed that APP/PS1 mice had no deficits in general motor activity but had shortened centre time, which was not rectified by PAP-1 treatment (Supplementary Fig. 5), suggesting that PAP-1 does not alleviate anxiety. Immediately after conclusion of the neurobehavioural tests, brains were removed for quantification of PAP-1 levels by UPLC/MS (ultra performance liquid chromatography - tandem mass spectrometer). While mice that had received unmedicated diet had no detectable PAP-1 levels in their brains, PAP-1-treated wild-type mice had average total brain concentrations of 608 ± 439 nM [mean ± standard deviation (SD); n = 12] and APP/PS1 mice concentrations of 411 ± 317 nM (mean ± SD; n = 12) at the time of sacrifice, demonstrating that our medicated diet regiment achieved pharmacologically active brain concentrations. PAP-1 treatment curbed microglia activation, as evidenced by reduced Iba-1 immunofluorescence (Fig. 6D) in brain sections and reduced expression of pro-inflammatory mediators in acutely isolated microglia (Supplementary Fig. 6). As expected, the level of K_v1.3 transcript in microglia was increased in APP/PS1 mice consuming control diet, compared to the wild-type littermates. Long-term PAP-1 treatment reduced the expression of K_v1.3, consistent with a reduced pro-inflammatory profile (Fig. 6E). Interestingly, PAP-1 treatment resulted in a 55% reduction of cerebral amyloid load in treated APP/PS1 mice (Fig. 6D and F). Quantities of TBS-soluble and TBS-insoluble amyloid-β in brain homogenates as determined by ELISA were also significantly reduced (Fig. 6G), consistent with the notion that K_v1.3 inhibition recovers the capability of microglia to clear amyloid-β in vivo. This notion was further strengthened by the observation that PAP-1 treatment also reduced the amyloid load as well as microglial activation in 5xFAD mice (Supplementary Fig. 7). Furthermore, PAP-1 treatment rectified the hippocampal LTP deficit in APP/PS1 mice to the wild-type level (Fig. 6H and I), consistent with the improved memory performance.

Discussion

Several key conclusions can be drawn from our study: (i) soluble AβO, the highly toxic amyloid-β species implicated as one of the principal initiators of Alzheimer’s disease phenotypes (Selkoe and Hardy, 2016), enhances microglial K_v1.3 and K_ir2.1 activity, two K⁺ channels characteristic of LPS- and IL-4-induced microglial polarization, respectively; (ii) K_v1.3 is required for AβO-induced microglial pro-inflammatory activation and neurotoxicity; (iii) K_v1.3 expression/activity is upregulated in transgenic Alzheimer’s disease animals and human Alzheimer’s disease brains; and (iv) pharmacological inhibition of K_v1.3 with PAP-1 in mice mitigates some key Alzheimer’s disease-like phenotypes including memory deficit and amyloid pathology. Our study suggests that K_v1.3 is a candidate therapeutic target for Alzheimer’s disease and that our selective small molecule K_v1.3 inhibitor PAP-1 is a promising therapeutic agent. The observations that pharmacological targeting of microglial K_v1.3 channels can affect hippocampal synaptic plasticity and reduce amyloid deposition in APP/PS1 mice are particularly interesting and support the potential of neuro-immunomodulation in treating neurodegenerative disorders (Marin and Kipnis, 2017).

When stimulating microglia in vitro we previously found that increased K_v1.3 expression was characteristic of LPS-induced ‘M1-like’ polarization, while expression of the inward-rectifier K_ir2.1 was a feature of IL-4-induced ‘M2-like’ polarization (Nguyen et al., 2017). This is reminiscent of our previous finding that different T and B cell subsets significantly differ in their K⁺ channel expression profile. For example, CCR7⁻ effector memory T cells and IgD⁻CD27⁺ B cells upregulate K_v1.3 and rely on this channel for their Ca²⁺ signalling, while CCR7⁺ naïve and central memory T cells as well as IgD⁺ B cells upregulate a different K⁺ channel after activation, the calcium-activated K_Ca3.1 channel (Wulff et al., 2003, 2004; Beeton et al., 2006). However, in contrast to cells of the adaptive immune system, which acquire stable phenotypes after subtype differentiation or class-switching, cells of the innate immune system such as microglia are highly plastic and typically do not acquire a stable phenotype that can be readily categorized by specific markers or signature genes (Murray et al., 2014; Bohlen et al., 2017). The M1/M2 classification defined by a set of markers is therefore considered an oversimplification and could be a misleading characterization of microglia activation in vivo (Ransohoff, 2016). In this study, based on not only in vitro but also in vivo data, we concluded that microglial K_v1.3 expression/activity is enhanced by AβO, and is required for AβO-induced microglial pro-inflammatory responses and associated neurotoxicity. Amyloid-β aggregates, including AβO, stimulate microglia in part via a well-known LPS receptor, TLR4 (Toll-like receptor 4) (Capiralla et al., 2012; Ledo et al., 2016), and induce responses widely considered as pro-inflammatory. However, AβO, differing from LPS, induced enhancement of both K_v1.3 and K_ir2.1 current density. Moreover, our studies using whole-cell patch-clamp recording, a single-cell approach, clearly showed that K_v1.3 and K_ir2.1 were co-expressed in single AβO-treated microglia, rather than segregated into different cell populations. AβO treatment also induced simultaneous expressions of markers traditionally used to categorize M1 and M2 polarization. This result is consistent with a previous single-cell gene expression study showing that individual myeloid cells responding to traumatic brain injury concurrently adopted both inflammatory and reparative features with a lack of exclusivity (Kim et al., 2016; Ransohoff, 2016) and suggests that AβO-treated microglia adopted an activation state that is more complex than simply being ‘pro-inflammatory’ or ‘neurotoxic’, and may even entail some M2-like tissue repair activity.

The age-dependent changes in microglial K_v1.3 and K_ir2.1 expression in 5xFAD mice followed a similar trend—initially an age-dependent increase, then a substantial decrease between 10 and 15 months of age. We suspect that these changes in K⁺ channel expression form part of the age-related changes in microglial function, documented by several lines of investigation, such as altered responses to amyloid-β aggregates or downregulation of ‘sensome’ genes (Hickman et al., 2008, 2013; Cameron et al., 2012; Heneka et al., 2013; Johansson et al., 2015). Our findings may also reflect the age-associated microglial dysfunction observed in mouse models overproducing amyloid-β (Hickman et al., 2008; Cameron et al., 2012; Heneka et al., 2013). However, the downregulation of K_v1.3 was not observed in human brains with advanced Alzheimer’s disease pathologies, in which K_v1.3 expression remains quite robust in microglia, especially in those associated with amyloid plaques (Fig. 4) (Rangaraju et al., 2015). This result suggests that current transgenic models of Alzheimer’s disease do not completely replicate the patterns of microglia activation in human Alzheimer’s disease, in which multiple stimuli other than amyloid-β may mould the eventual microglial phenotypes. In any case, the human study supports that K_v1.3 could be a therapeutic target even at the late stage.

While in wild-type mice the basal levels of K_v1.3 and K_ir2.1 remained low, the K_ir2.1 level was high as measured at 6 months (Fig. 3E), suggesting an unknown physiological function in adult microglia. Although some extrapolation can be made from a previous study implicating K_ir2.1 activity in microglial Ca²⁺ signalling and migration under ‘resting’ and ‘anti-inflammatory states’ (Lam and Schlichter, 2015) and our own study showing that IL-4 induces large K_ir2.1 currents (Nguyen et al., 2017), these studies were performed on cultured neonatal microglia. In contrast to K_v1.3, an important limitation preventing us from gaining insight into the in vivo role of K_ir2.1 is a lack of good pharmacological or genetic tools to analyse the significance of AβO-induced K_ir2.1 upregulation in microglia in adult and aged mice. K_ir2.1 plays a vital role in the vasculature and in cardiac excitability; its knockout causes mice to die within hours after birth (Zaritsky et al., 2001). The only available K_ir2.1 inhibitors, high micromolar to millimolar Ba²⁺ or the small molecule ML133 (IC₅₀ 2 μM) (Wang et al., 2011), are both unsuitable for in vivo use and are even problematic in experiments with cultured microglia because of their high toxicity (Nguyen et al., 2017). Future developments of mice with microglia-targeted deletion of K_ir2.1 and specific, non-toxic K_ir2.1 inhibitors are required to illuminate the in vivo role of microglial K_ir2.1 in Alzheimer’s disease.

It is intriguing to hypothesize that selectively blocking microglial K_v1.3 activity without affecting K_ir2.1 function would mitigate the neurotoxicity accompanying the pro-inflammatory activation while preserving the anti-inflammatory and tissue repair functions associated with IL-4-like effects. Indeed, our data using pharmacological K_v1.3 targeting are consistent with the above hypothesis, and support K_v1.3 blockers as attractive therapeutic agents to mitigate amyloid-β-induced pro-inflammatory microglial responses that are highly relevant to Alzheimer’s disease pathogenesis (Go et al., 2016; Ledo et al., 2016; Rangaraju et al., 2017). Furthermore, the parallel hyper-expression of microglial K_v1.3 in transgenic Alzheimer’s disease mouse models and in human Alzheimer’s disease brain tissue validates K_v1.3 as a therapeutic target for Alzheimer’s disease and/or mild cognitive impairment preceding Alzheimer’s disease.

The significance of microglia in modulating Alzheimer’s disease-like amyloid pathology remains controversial. Our in vitro data presented in Fig. 6A and in vivo data of substantial reductions of cerebral soluble and insoluble amyloid-β species in mice chronically treated with PAP-1 (Fig. 6D–G and Supplementary Fig. 7) suggest that K_v1.3 inhibition enhances the microglial amyloid-β clearance capacity that is suppressed in Alzheimer’s disease transgenic mice (Hickman et al., 2008; Hellwig et al., 2015). On the other hand, reductions of pro-inflammatory cytokines, as a result of K_v1.3 blockade, might also lead to decreased cerebral amyloid-β levels (Tan et al., 2014). Our result is consistent with several reports showing that selective modulation of specific microglial signalling pathways alters amyloid-β deposition and clearance (El Khoury et al., 2007; Cameron et al., 2012; Heneka et al., 2013; Guillot-Sestier et al., 2015; Hellwig et al., 2015; Johansson et al., 2015; Krauthausen et al., 2015), but appears contradictory to reports showing that inducible microglia depletion by genetic or pharmacological means fails to affect amyloid pathology in transgenic Alzheimer’s disease models (Grathwohl et al., 2009; Spangenberg et al., 2016). Although further studies are needed to resolve this controversy, our data using K⁺ channel expression as a readout show that the microglial activation profiles in transgenic Alzheimer’s disease mice are complex and age-dependent, as discussed above. Therefore, differences in the timing and approaches of neuro-immunomodulation to alter microglia quantity or signalling may differentially influence the outcomes. In our experiments, PAP-1 was administered to APP/PS1 mice at an age when microglia amyloid-β clearance capacity was impaired (Hickman et al., 2008). We hypothesize that while eliminating microglia at this stage may not alter the amyloid-β homeostasis as they cease to function properly, interventions to ‘rehabilitate’ the phagocytic capacity of existing microglia will reduce amyloid-β deposition (Guillot-Sestier et al., 2015). Our data suggest that K_v1.3 inhibition could be a means for such rehabilitation.

One limitation of the current study is that we cannot completely exclude the possible contributions of peripheral monocyte-derived macrophages and effector memory T cells, which also express K_v1.3 (Wulff et al., 2003; Feske et al., 2015). Peripheral immune modulation may impact neuroinflammation or brain function via humoral or cellular factors (Villeda et al., 2014; Prinz and Priller, 2017). While it remains controversial whether there is an influx of peripheral immune cells into the brain in Alzheimer’s disease mouse models or in human Alzheimer’s disease brains (Bien-Ly et al., 2015; Prinz and Priller, 2017), many brain mononuclear phagocytes that surround amyloid-β plaques are CD45^high, suggesting their peripheral monocyte origin (Jay et al., 2015). Several groups also reported the presence of T cells in the brain of Alzheimer’s disease patients (Itagaki et al., 1988; Town et al., 2005; Bryson and Lynch, 2016). Myeloid cells and lymphocytes were also found in the newly identified meningeal lymphatics (Louveau et al., 2015), through which the trafficking of immune cells might influence brain pathology. It is important in future studies to use Cre recombinase-guided cell targeting to selectively knockout K_v1.3 in brain microglia, peripheral macrophages/monocytes, or T cells, in order to parse out their relative contributions to Alzheimer’s disease pathogenesis. Although the current study does not distinguish the respective contribution of peripheral immune cells and brain myeloid cells, this limitation, nonetheless, does not hamper the translational potential of K_v1.3 blockers in Alzheimer’s disease therapy.

For future clinical development, one may raise the concern that inhibiting K_v1.3 may make individuals susceptible to infection or cause immune dysregulation. However, in contrast to stronger immunosuppressants like calcineurin inhibitors and anti-TNF reagents, K_v1.3 inhibitors are ‘mild’ immunosuppressants and do not affect the ability of rodents or primates to clear bacterial or viral infections (Pereira et al., 2007; Matheu et al., 2008). Overall, K_v1.3 is regarded as a relatively safe drug target (Wulff et al., 2009). PAP-1 does not prevent rhesus macaques from developing a protective, central memory T cell response following intranasal application of a flu vaccine (Pereira et al., 2007). Kcna3/K_v1.3^−/− mice are viable, reproduce normally and exhibit very subtle phenotypes (Koni et al., 2003; Fadool et al., 2004). Toxicity studies with PAP-1 and the peptidic K_v1.3 blocker ShK-186 have so far not revealed any toxicity despite 6 months or 28 days of continuous administration in rats or rhesus macaques (Beeton et al., 2006; Pereira et al., 2007; Tarcha et al., 2012). PAP-1 has further completed IND (Investigational New Drug)-enabling toxicity studies, while ShK-186 has passed both IND toxicity studies and phase 1 safety studies without any adverse findings and has demonstrated efficacy in a small phase 1b study in plaque psoriasis (Tarcha et al., 2017). The current study, showing excellent brain penetrance, no apparent toxicity after chronic treatment, and satisfactory efficacy, provides necessary proof-of-concept data to advance K_v1.3 blockers such as PAP-1 to clinical trials on Alzheimer’s disease and/or mild cognitive impairment.

Funding

This work was supported by a grant from the Alzheimer’s Association NIRG-10-174150 to I.M., the U.S. NIH grants R21 AG038910 (NIA) to L-W.J., and R01 NS098328 (NINDS) to H.W., and in part by R01 AG043788 to I.M., P30 AG10129 to L-W.J., K08 NS099474 to S.R., P30 NS055077 (NINDS) and P50 AG025688 (NIA) to A.I.L, and U54 HD079125 to K.K. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

Abbreviations
 
• AβO

  amyloid-β oligomer

 
• FSB

  (E,E)-1-fluoro-2,5-bis(3-hydroxycarbonyl-4-hydroxy)styrylbenzene

 
• LTP

  long-term potentiation

 
• LPS

  lipopolysaccharide

 
• MAPK

  mitogen-activated protein kinase

 
• PAP-1

  5-(4-phenoxybutoxy)psoralen

 
• TBS

  Tris-buffered saline

References

Author notes