Preclinical and Clinical Studies of p38α MAP kinase inhibition to Treat Basal Forebrain Cholinergic Dysfunction and Degeneration


 The endosome-associated protein Rab5 is a central player in the molecular mechanisms leading to degeneration of basal forebrain cholinergic neurons (BFCN), a long-standing target for drug development. As p38α kinase is a Rab-5 activator, we hypothesized that inhibition of this kinase held potential as an approach to treat diseases associated with BFCN loss. Herein we report that treatment with an oral small molecule p38α kinase inhibitor reversed pathological disease progression in the basal forebrain in a mouse model that develops BFCN degeneration. Further, the preclinical results were successfully translated to the clinic, with improvement of clinical outcomes associated with cholinergic function in a clinical study in dementia with Lewy bodies (DLB), a disease in which BFCN dysfunction and degeneration is the primary driver of disease expression. The findings both advances a novel approach to treating DLB and validates the translational platform that provided the mechanistic rationale for advancing that approach.

Degeneration of the basal forebrain, the primary source of cholinergic innervation in the brain, occurs in aging-and neurodegenerative disease-related cognitive disorders, including in Alzheimer's disease (AD) where basal forebrain cholinergic neuron (BFCN) dysfunction may be a major driver of the neurodegenerative process elsewhere, including in the hippocampus 1-3 .
Further, recent publications demonstrate that BFCN loss underpins the gait dysfunction in Parkinson's disease (PD), suggesting that therapeutically targeting the cholinergic system could also address certain motor aspects of neurodegenerative disease 4,[50][51] . Recent evidence also indicates that correcting the functional deficit in the cholinergic system, thus amplifying the physiologic release of acetylcholine is likely to be more effective than the conventional approach of compensating for BFCN dysfunction by delaying clearance of released acetylcholine with cholinesterase inhibitors 5 .
A critical pathogenic event in development of BFCN dysfunction and degeneration is impaired nerve growth factor (NGF) signaling, depriving cholinergic neurons of the neurotrophic support necessary for proper functioning and survival 6,14 . NGF signaling is transduced by endocytosis and retrograde trafficking of a maturing Rab5-"signaling endosome" containing the NGF receptor, TrkA, to initiate a transcriptional program. Cholinergic neurons, with long axonal projections throughout the cortex, are particularly vulnerable to disruption of this retrograde signaling process from distant synaptic connections back to the cell body. From a pathogenic mechanistic standpoint, the protein Rab5, a GTPase and master signaling molecule regulating endocytosis and endosome function, is implicated in development of impaired NGF signaling.
Both Rab5 hyper-activation and lowered endosome recycling rates impair NGF signaling, partly due to enlargement of endosomes which slows their retrograde transport and trophic signaling 7,8 , leading to cholinergic atrophy in Down syndrome (DS) mouse models 7,[9][10][11] . These DS models essentially recapitulate adult-onset basal forebrain cholinergic neurodegeneration 12 . Further, Rab5 hyperactivation in vivo in Rab5 overexpressing transgenic mice causes BFCN degeneration 12 . These findings establish Rab5 as a therapeutic target for BFCN dysfunction.
Moreover, results from animal studies also indicate that the neurodegeneration of BFCNs is reversible, as the BFCNs do not die with age and disease; rather, they lose cholinergic phenotype and functional properties, and this loss can be reversed by direct NGF infusion to the basal forebrain 13,14 . Thus, pharmacologically restoring NGF signaling through targeting Rab5 has the potential to reverse disease progression by increasing numbers of functional cholinergic neurons.
As the alpha isoform of p38α kinase is a major Rab5 regulator and activator 15 , and it has been implicated in the regulation of retrograde axonal transport 16 , we hypothesized that p38α inhibition would be a pharmacological approach to treating diseases associated with BFCN dysfunction. Accordingly, we evaluated a specific p38α kinase inhibitor neflamapimod (NFMD) [17][18][19] , preclinically in a DS mouse model (Ts2 mouse) 12 that demonstrates BFCN degeneration and, subsequently, clinically in a placebo-controlled phase 2a study in patients with dementia with Lewy bodies (DLB), where BFCN dysfunction and degeneration is considered to be the primary driver of disease expression and progression [20][21][22] . Herein we report successful translation of the scientific understanding of mechanisms of cholinergic degeneration to beneficial effects on BFCNs both preclinically and clinically, with neflamapimod reversing pathological disease progression in the basal forebrain in the preclinical study and improving clinical outcomes associated with cholinergic function in the clinical study in DLB patients.

Neflamapimod treatment of Ts2 mice reverses endosomal pathology and restores number of cholinergic neurons
Ts2 mice, a DS mouse model that develops adult-onset Rab5+ endosomal pathology and cholinergic degeneration in the basal forebrain 12 , and control 2N (wild type) mice were treated at approximately six months of age with either neflamapimod (3mg/kg body weight, twice daily (BID)) or vehicle (1% Pluronic F108) by oral gavage for 28 days. The number and size of Rab5-GTP-positive endosomes, determined by Image J analysis, were greater in Ts2 than in wildtype mice, while, with neflamapimod treatment, both parameters were normalized to values seen in the wildtype mice ( Fig. 1A and B). Similarly, the intensity ratio of Rab5-GTP (activated form of rab5) to total Rab5 was significantly higher in vehicle-treated Ts2 than in wildtype mice ( Fig.   1C) collectively and reduced with neflamapimod treatment in Ts2 mice and not in wildtype mice; results which suggest neflamapimod impacts aberrant Rab5 activation, without impacting physiologic Rab5 activation. In Ts2 mice, basal forebrain neurodegeneration is well established by 6 to 7 months of age 12 and the number of medial septal choline acetyltransferase-positive (ChAT+) neurons (i.e., cholinergic neurons) continues to decline as the mice age (Fig. 1D).
Neflamapimod-treated Ts2 mice had significantly higher numbers of ChAT+ neurons at the end of the four-week treatment period than did vehicle-treated Ts2 mice (p<0.005), exhibiting numbers comparable to those in the wildtype mice (Fig. 1E,F). Notably, in vehicle-treated Ts2 mice, ChAT+ neurons had an abnormal morphology, with neurite swelling (yellow arrows) or atrophy (orange arrow), perikaryal atrophy (red arrows), and below normal ChAT immunoreactivity (green arrows) (Fig. 1H), while neflamapimod treatment appeared to normalize their morphology (Fig. 1G,H For the open field test, the performance at the end of treatment also represented a significant improvement from performance at the pretreatment baseline ( Figure 2E).
The positive effects of neflamapimod on pathology and function in Ts2 mice appear to be mediated by inhibition of the p38α kinase pathway Further biochemical analyses were conducted to understand the underlying mechanisms.
As reduction of p38α activity in APP/PS1 double transgenic mice leads to decreased BACE1 protein expression 26 , BACE1 protein levels were similarly evaluated in the brain cortex samples. Ts2 mice treated with neflamapimod, compared with vehicle, had significantly lower BACE1 protein levels (Figs. 3C and D; p<0.05 for neflamapimod vs. vehicle-treated Ts2 mice).
In addition, levels of bCTF, the product of APP cleavage by BACE1, which were significantly higher in Ts2 than in wildtype mice (p<0.001), were decreased significantly by neflamapimod treatment in the Ts2 mice (Figs. 3C and D, p<0.001).

Neflamapimod treatment in patients with DLB improves cognition within domains associated with cholinergic function
The AscenD-LB study (NCT04001517) was a 91-patient, 16 weeks). The aim of this exploratory study was to evaluate the cognitive effects of neflamapimod in this patient population, utilizing a six-test neuropsychological test battery (NTB) designed to assess attention and executive function. No formal sample size calculations were performed but based on prior experience with use of a NTB, 40 participants per arm was considered sufficient for the primary objective. Patients were randomized 1:1 to either neflamapimod 40mg capsules or matching placebo; and then, based on body weight, assigned to either a twice-daily (BID) [weight < 80 kg; 40mg BID (80mg/day) neflamapimod or placebo BID) or thrice-daily (TID) [weight ≥ 80 kg; 40mg TID (120mg/day) neflamapimod or placebo TID] regimen. The weightbased dosing regimen was utilized because pharmacokinetic data available at the time of study start indicated that 40mg BID would achieve the targeted 12-hour plasma drug exposure of 100 ng*hr/mL, but only in patients weighing less than 80kg (thus requiring 40mg TID in the higher weight range). However, during the study, additional data became available, including from a study in AD 20 , indicating that 40mg BID would not achieve the target plasma drug exposure and would likely be ineffective. Indeed, measured plasma drug concentrations in the current study were 50% higher in 40mg TID recipients [median steady-state concentration (CSS) of 10.2 ng/mL] than in 40mg BID recipients (median CSS=6.8 ng/mL). As a result, the efficacy endpoints were analyzed in two ways, (1) considering all neflamapimod recipients as one dose group as originally intended (an evaluation in which treatment effects are diluted by 40mg BID not achieving target plasma drug concentrations) and (2) considering 40mg TID recipients separately. As this was an exploratory study, p-values were not adjusted for multiple comparisons. There were no significant differences in baseline disease characteristics between groups, either comparing all neflamapimod recipients vs. placebo, or 40mg TID vs. either placebo or placebo TID (Supplemental Table 1; Supplemental Figure 1 shows Consort Flow diagram).
The mean NTB composite z-score, in which the individual tests are weighted equally, over the course of study is shown in Figure 4A. In the placebo group, this cognitive composite remained stable or worsened slightly over the 16 weeks of the study, indicating no placebo or learning effects in the data. For the 40mg BID group, the cognitive composite results were very similar to those for the placebo group while, beginning at week 4 and maintained to week 16, there was an apparent improvement in the 40mg TID group. With neflamapimod 40mg BID being similar to placebo, there were no significant differences (p>0.2) when the combined neflamapimod groups were compared with placebo. For the 40mg TID group, the cognitive composite improved from baseline and was significantly different over the course of the study from either placebo (p=0.049, mean difference=0.18 95%CI:0.00-0.35, Cohen's d effect size (ES)=0.47) or placebo TID (p=0.049, difference=+0.11, 95% CI 0.00-0.21, ES=0.49).
As attention is the cognitive domain regarded as most specific for the cholinergic system, performance on the two NTB tests that measure attention and processing speed (i.e., the Cogstate Detection and Identification Tests) were combined to yield an Attention Composite ( Figure 4B).
In this analysis there was a significant difference favoring 40mg TID over placebo (p=0.023, Gait dysfunction due to parkinsonism is very common in patients with DLB and recent evidence indicates that gait dysfunction is to great extent due to deficits in basal forebrain cholinergic function 4 . Gait in the AscenD-LB study was assessed with the Timed Up and Go (TUG) Test (Fig. 4D), measuring functional mobility by monitoring the time that a subject takes to rise from a chair, walk three meters, turn around 180 degrees, walk back to the chair, and sit down while turning 180 degrees. In the comparison of the combined neflamapimod groups (i.e., "all neflamapimod" participants) to placebo, there was significant reduction in the time required to complete the TUG test (i.e., improvement) with neflamapimod treatment (p=0.044,

Efficacy of Neflamapimod was greater in DLB than in mixed AD-DLB
In patients with DLB, elevated levels of the protein phosphorylated tau (ptau) have been reported to identify the presence of Alzheimer's disease (AD) co-pathology, particularly temporal lobe tau pathology 28 ; and patients with such co-pathology are considered to have mixed AD-DLB, rather than DLB. Accordingly, plasma ptau181 levels in baseline plasma samples were quantitated utilizing the Simoa® platform in a laboratory where the plasma ptau181 cut-off for AD co-pathology had been set at 2.2 pg/mL (i.e., plasma ptau181 ≥ 2.2 pg/mL associated with existence of AD co-pathology).
At baseline, 22 of 41 (53%) of placebo and 22 of 42 (54%) of neflamapimod participants in the efficacy analysis population had ptau181 < 2.2 pg/mL (i.e., had DLB, rather than mixed AD-DLB). When the efficacy analyses were repeated, after taking into account plasma ptau181 levels, the response in the patients with baseline plasma ptau181 < 2.2 pg/mL (i.e., those predicted to not have AD co-pathology and therefore "pure DLB") appears to be better than the response in patients with baseline plasma ptau181 levels ≥ 2.2 pg/mL (i.e., those predicted to have AD co-pathology, and therefore mixed AD-DLB); see Figure 5. Further, as shown in Table   1 (lower half), when individual endpoints showing positive treatment effects in the main analysis were analyzed in only patients with pure DLB (defined by ptau181 < 2.2 pg/mL at baseline), the magnitude of the treatment effect was high with the statistics particularly robust for the comparison of 40mg TID vs. placebo or vs. placebo TID for attention composite, TUG and CDR-SB. For the NTB, the magnitude of the treatment effect is also greater than in the main analysis but with the smaller sample size, did not reach significance. The magnitude of the treatment effect in patients with pure DLB for the comparison between 40mg TID vs. placebo (or placebo TID), as evaluated by effect size, was substantially higher than for the full efficacy population. In addition, in the 40mg TID vs. placebo TID comparison, the effect size was consistently greater than 0.6, the level above which is considered to be a clinically important treatment effect.

Neuropsychiatric Outcomes and Safety in the Clinical Study
Four specific domains within the 10-item Neuropsychiatric Inventory [depression (dysphoria), anxiety, hallucinations, and agitation/aggression] were secondary outcome measures in the study. Of these, the domain most specific for both DLB and the cholinergic system is hallucinations 29 , for which there was a significant effect favouring neflamapimod treatment for severity (Supplemental Figure 4; p=0.050 for all neflamapimod vs. placebo, p=0.019, for neflamapimod 40mg TID vs. placebo). There were no discernible treatment effects for the other three domains of the NPI-10 (Supplemental Figure 5).
Neflamapimod treatment was well tolerated with no study-drug associated treatment discontinuations or serious adverse events (SAEs) reported (See Supplemental Table 1 for all treatment discontinuations and SAEs). The only treatment-emergent adverse events occurring at an incidence >5%, in either neflamapimod-or placebo-treated patients, were falls (13% in neflamapimod versus 9% in placebo), headache (9%, 4%), diarrhea (7%, 11% ), nausea (7%, 7%), and tremor (0%,7%). Among neflamapimod-treated patients, the incidence (3/20, 15% for each) of diarrhea and headache were higher in those receiving 40mg TID than in those receiving 40mg BID (0/26, 0% for diarrhea; 1/26, 3% for headache); the incidence of falls was higher in those receiving 40mg BID (5/26, 19%) than in those receiving 40mg TID (1/20, 5%). The preclinical mechanistic findings were consistent with our original hypothesis that p38α inhibition is an approach to decrease Rab5 activation and reverse the block in endosomal trafficking and signaling. In addition, the data provide insight on additional relevant targets that may be contributing to drug activity. Specifically, neflamapimod treatment decreased Rab5-GTP levels, a marker of Rab5 activation, and reversed the Rab5+endosomal pathology that is a direct histopathological marker of the block in endosomal trafficking underlying the defect in NGF signaling in Ts2 mice 6,7 . In separate experiments, restoration of cholinergic neuron numbers was observed following neflamapimod treatment in Rab5-overexpressing transgenic mouse , which otherwise develop basal forebrain cholinergic degeneration 42 ; further supporting the activity of neflamapimod on Rab5-mediated cholinergic neuron loss. The beneficial effects were associated with p38α inhibition, because neflamapimod treatment normalized levels of activated, i.e., phosphorylated p38α (in addition to being activated by upstream kinases, p38α autoactivates 43,44 ) and lowered levels of those of its downstream substrates MK2 and MNK1 45 .

DISCUSSION
Furthermore, neflamapimod treatment decreased BACE1 protein levels, consistent with reports that neuronal p38α knockout in APP/PS1 ("Alzheimer's") transgenic mice led to decreased BACE1 protein levels, through increased autophagy-lysosome mediated degradation of the BACE1 protein 26,46,47 . Moreover, in a recent study, p38α activity decreased BACE1 levels in synaptic terminals by increasing retrograde axonal transport of BACE1 to lysosomes for degradation 48 . Thus, the finding that neflamapimod decreases BACE1 protein levels in Ts2 mice provides further evidence that the drug improves retrograde axonal transport of endosomes, the fundamental pathophysiologic defect being targeted with our therapeutic approach. Finally, the reduction in bCTF levels, resulting from decreased BACE1 protein levels, is also a likely contributor to the effects on cholinergic degeneration and function 7,12 .
With a mechanistic foundation provided by the animal studies, the results of the clinical study provided direct evidence that the preclinical results could be translated to the clinic, with neflamapimod treatment leading to significant dose-dependent improvements in cognition, motor function and potentially visual hallucinations in a neurodegenerative disease characterized by substantial basal forebrain cholinergic loss, DLB. Despite the use of a relatively small sample in this exploratory study, the consistency of the benefits with neflamapimod 40mg TID, relative to placebo, across multiple aspects of DLB endpoints that are related to cholinergic function but are otherwise uncorrelated, makes the results of this clinical trial compelling. Further, as disease expression in the patients with "pure DLB" (rather than mixed AD-DLB) would be expected to most specifically driven by BFCN dysfunction, the specificity for, and magnitude of the clinical efficacy in this patient sub-group further supports that neflamapimod has potent activity against BFCN dysfunction. Notably, all these improvements were observed in patients already receiving cholinesterase inhibitors and the effect size on cognition was, in fact, greater for neflamapimod than that reported for rivastigmine on a similar cognitive test battery 49 60 , a finding that is consistent with the notion that degeneration in the basal forebrain precedes, and is a major contributor to neurodegeneration and disease progression in the hippocampus in AD. Indeed, neflamapimod treatment in a 24-week placebo-controlled study in AD, while not improving episodic memory function, did improve, relative to placebo, CSF markers of neurodegeneration and there was evidence of a reduction in progression of disease by clinical endpoints, including the episodic memory endpoint 19 . The combined findings suggest that treating basal forebrain cholinergic degeneration has the potential to treat disease progression in the hippocampus in AD, but longer duration trials are required to evaluate that potential.
The clinical results provide a foundation for further evaluation of neflamapimod in patients with DLB, where the primary pathology is in the basal forebrain. However, a larger study is required to better define the magnitude of the benefit and only a longer duration trial can address whether neflamapimod treatment impacts DLB disease progression. Our results, combined with other clinical and preclinical data, also suggest that neflamapimod, by improving basal forebrain cholinergic function, has the potential to slow disease progression in the hippocampus in AD. Beyond implications specific to neflamapimod, our results provide validation of our pathogenic model of cholinergic degeneration and of the Ts2 DS mouse as a translational platform for target validation, drug discovery and development, and for obtaining preclinical proof-of-concept for novel therapeutic approaches to treating basal forebrain cholinergic dysfunction and degeneration.   Statistical significance is represented by asterisks *p<0.05, **p<0.01, ***p<0.005.     The test compound neflamapimod (manufacturer lot number M10140) and instructions for formulation preparation, route of administration and dosing volume were provided by EIP Pharma, Inc. as for previously described studies 17,61 . Both wild-type (2N) and Ts2 mice (n=8-10 per group, targeting an equal number of female and male mice) were treated for the duration of two to four weeks, twice-daily by oral gavage at a 5 ml/kg dosing volume with vehicle (1% (w/v) Pluronic F108, Sigma) or neflamapimod that was freshly dissolved in 1% (w/v/) Pluronic F108 at a final concentration of 0.6 mg/ml to administer the test compound at 3 mg/kg. Treatment was initiated at 4.7-6.4 months of age for 2N(control) and Ts2 mice, when endosomal pathology is evident and cholinergic neuronal loss is developing in Ts2 mice. To assess the health status of the animals throughout the studies and ensure their suitability for use in research, their general health status was monitored daily, and body weight was determined weekly.

Figure 2. Slower decline of LTP and normalized scores of behavioral tests in
One hour after either the terminal vehicle or neflamapimod treatment, mice were euthanized by displacement of air with 100% CO2 and decapitated within 5 minutes. Whole mouse brains were collected and split into two brain hemispheres. The hippocampal sections were dissected out for electric physiology testing, while the remainders of the hemibrains were stored at -80°C for further biochemistry analysis. For immunocytochemistry analysis, mice were anesthetized with a mixture of ketamine (100 mg/kg body weight) and xylazine (10 mg/kg body weight) and transcardially perfused with 4% paraformaldehyde and brain tissues were sectioned into 40 µm thick slices with a vibratome (Leica VT1000S).

Immunocytochemistry and Western Blot Analysis of Preclinical Samples
Immunohistochemistry was performed on vibratome sections 12 All the secondary antibodies for western blot analyses were used according to the manufacturer's recommendations (Jackson ImmunoResearch Laboratories, PA). A digital gel imager (Syn-gene G:Box XX9) was used to capture the ECL images and band intensities were quantified with Fiji/ImageJ (https://imagej.net/Fiji) using b-actin as an internal control. Sizes of Rab5+ endosomes were determined by Image J, as previously described 12,42,63 .

Stereological counting of MS-ChAT+ neurons in Ts2 and wild-type mice
The numbers of ChAT-immunoreactive BFCNs in MS region were determined using the optical fractionator method 67 using ImageJ software as previously described 12,68 . Cells were sampled with a grid of optical dissector sampling sites, on every third 40-micron thick coronal section through the medial septal area. At each sampling site a z-stack of 2-micron spaced images was collected using a 40x oil-immersion objective, and 160 x 116 micron counting box was drawn onto the z-stack, with an upper guard zone of 2 microns (1 slice), counting box of 4 microns, and lower guard zone of the remaining slices. All counts were adjusted for on-slide zaxis shrinkage.

Electrophysiology of Ts2 and wild-type mice
6-7-month-old 2N and Ts2 mice, either treated with vehicle or with neflamapimod for 4 weeks (n=4-5) were used to measure long term potentiation (LTP) in CA1 hippocampal regions, as previous described 36, 42,68 . Transverse hippocampal slices (400 µm) were cut and placed in a recording chamber which was filled with artificial CSF (ACSF) at 29°C and maintained consistent at 95% O2 and 5% CO2. Field EPSPs (fEPSPs) were recorded by placing both the stimulating and the recording electrodes in hippocampal CA1 stratum radiatum, while basal synaptic transmission (BST) was determined by plotting the stimulus voltages over the slopes of fEPSPs. LTP was induced using theta-burst stimulation and responses were recorded for 2 h and measured as fEPSP slope expressed as percentage of baseline 68 .

Preclinical evaluation of behavioral outcomes
Open field and NOR test were performed according to previous publications 42, 69  For open field, 6 mins exploration time was used for the analysis with ToxTrac 71 , and the following parameters including speed, travel distance and percentage of time spend in thigmotaxis (all edges of the chamber) was determined for each mice.

Statistical analysis of preclinical study
All quantitative data were subjected to two-tailed unpaired Student's t-test for single comparison, and one-way ANOVA analysis for multiple comparisons with post-hoc Tukey's analysis using and after all queries were resolved and database was locked, SAS datasets were provided to Anoixis corporation for statistical analysis. The data from the Cogstate tests were processed at Cogstate and provided directly to Anoixis.
Due to the various scales among different tests in the NTB and the need to have equal weights for deriving composite score metrics performance on each, the tests were standardized relative to baseline data from all randomized subjects (i.e., the individual test core was converted to a z-score by subtracting the study sample's mean at baseline from the score and dividing by the standard deviation (SD) of the study sample's baseline). The NTB composite z-score was calculated as the average of the z-scores from the component of the six individual tests in the composite. For the purposes of facilitating further understanding of the drug effects on the cholinergic system, an exploratory Attention composite endpoint made up of the two tests in NTB that evaluate information processing speed, the identification and detection tests, was calculated and evaluated.
As pre-specified in the protocol, the analyses of all efficacy endpoints (primary, secondary and exploratory) used a Mixed Model for Repeated Measures (MMRM) analysis method with change from baseline as the dependent variable, with a fixed effect on treatment extended to baseline composite score and study visit as covariates. There was a fixed effect on treatment that was extended to baseline composite score and study visit. The interaction term (i.e., scheduled visit by treatment) was considered. The random effect was modeled at subject level first and was extended to other covariates if there were no numerical integration issues.