Randomized phase 2a trial assessing a novel septin molecular glue in Alzheimer's disease

Abstract

INTRODUCTION

Pharmacological restoration of septin filament integrity has the potential to provide symptomatic benefit and disease modification in Alzheimer's disease (AD).

METHODS

REM127, a septin modulator, was assessed in mild‐to‐moderate AD (EudraCT: 2022‐000080‐43) in a phase 2a trial (n = 14). Primary endpoints: safety and tolerability; exploratory endpoints: pharmacokinetics, cerebrospinal fluid (CSF) biomarkers, electroencephalography (EEG), and functional outcomes.

RESULTS

In participants on active therapy, dose‐dependent increases in serum aminotransferase were observed, leading to study discontinuation. CSF hyperphosphorylated tau (P‐tau181), endpoints reflecting synaptic function and cognitive outcomes, were changed significantly (p < 0.05) to normal compared to placebo.

DISCUSSION

REM127 triggers off‐target liver adverse effects. Anticipated on‐target outcomes suggest septin modulation has symptomatic benefit and modifies processes underlying AD. Results are considered exploratory as statistical power is constrained due to the small sample size caused by early termination. Further investigation of the therapeutic concept using an optimized septin molecular glue with an improved safety profile is warranted.

Highlights

• Septin 6/7 molecular glue REM127 was assessed in symptomatic participants with Alzheimer's disease (AD).

• REM127 triggers off‐target effects suggesting liver adverse effects.

• REM127 brain exposure was consistent with saturated target engagement.

• Biomarker and cognitive outcomes were changed consistent with therapeutic benefit.

• Septin modulation may restore synaptic function and mitigate pathology in AD.

1. BACKGROUND

Alzheimer's disease (AD) is the most common form of dementia, afflicting more than 50 million patients worldwide. The symptoms are linked to synaptic dysfunction followed by degeneration of nerve cells, ¹ most prominently affecting those participating in neuronal circuits underlying memory and cognitive behaviors. Pathological hallmarks in the brain of patients include amyloid‐beta protein (Aβ) plaques and neurofibrillary tangles comprised of aggregated Aβ (Aβ) peptides and hyperphosphorylated tau (P‐tau), respectively. ² Neurons affected by Aβ pathology phosphorylate and secrete tau, leading to increased cerebrospinal fluid (CSF) and plasma tau protein concentrations. ³

The causes and risk factors for AD are heterogeneous, entailing nongenetic risk factors such as ageing, lifestyle choices, or traumatic brain injury, as well as genetic mutations controlling the production or clearance of toxic Aβ‐peptides. Despite their diverse nature, these risk factors ultimately contribute to a common pathophysiology and symptomatology. Such common processes include calcium (Ca²⁺) dyshomeostasis, ⁴ a condition characterized by inappropriately elevated concentrations of Ca²⁺ in the cytosol. ⁵ As a key second messenger regulating fate and function of neurons, excessive cytosolic Ca²⁺ levels adversely impact downstream pathways toward repressed synaptic function, accelerated formation of Aβ and tau pathology, and activation of cell death pathways. ⁵ Hence, deregulated Ca²⁺ homeostasis plays a central role in driving neurodegeneration in AD, and restoration of Ca²⁺ homeostasis may be an effective therapeutic intervention.

Recently, septin filaments were identified as a promising AD therapeutic target that acts by normalizing cytosolic Ca²⁺ levels, ⁶ without disturbing physiological functions of Ca²⁺. Septin filaments bind to the inner side of the plasma membrane where they control, among other activities, the function of Store‐Operated Ca²⁺ (SOC) channels. In the presence of pathological tau, the integrity of these filaments is impaired, resulting in excessive activity of SOC channels and elevated influx of Ca^{2+ 6}. Small molecule drug candidates were developed to restore the integrity of the septin filaments by “glueing” together the key subunits Sept6 and Sept7. ⁶ Administration of these “septin glues” in both tau‐ and Aβ‐driven model systems restored Ca²⁺ homeostasis and resulted in rapid beneficial effects on synaptic function (within 1–2 weeks) and disease‐modifying effects, including reduction of tau and Aβ pathology.

Here we report results of a phase 2a proof‐of‐concept study assessing the neuroprotective activity of a novel septin glue REM0046127 ⁶ (referred to in the remainder of the text as REM127) on trial participants with mild‐to‐moderate AD dementia. The study entailed a double‐blind, safety and exploratory efficacy trial of REM127, administered orally for 28 days. The primary objective was to assess safety and tolerability of REM127. Exploratory objectives entailed evaluation of the REM127 pharmacokinetic (PK) and pharmacodynamic relationship of fluid‐based biomarkers, electroencephalography (EEG), and cognitive endpoints.

2. METHODS

2.1. Trial design

This study was a 28‐day (with 14 days placebo run‐in), phase 2a multicenter, randomized, placebo‐controlled, double‐blind clinical trial of REM127 in participants with mild to moderate AD dementia (EU Clinical Trials identifier: 2022‐000080‐43; Clinical Trials.gov identifier: NCT05478031). Access clinical study report: https://www.clinicaltrialsregister.eu/ctr‐search/search?query=2022‐000080‐43. REM127 was administered orally, twice daily (BID). The initial REM127 dose levels were selected based on PK and safety data of a previously conducted phase 1 trial in healthy volunteers, including a panel of healthy elderly participants of both sexes (unpublished data) entailing the highest tested safe dose of 1400 mg/day (700 mg BID) to maximize target occupancy and 350 mg/day (175 mg BID) to explore dose dependency of possible pharmacodynamic effects. Based on emerging safety data indicating liver adverse effects, the protocol was amended to include a dose level of 88 mg/day, given as a 44 mg BID treatment. This dose was selected to decrease the likelihood of detecting liver toxicity‐related adverse events (AEs) (based on the “rule‐of‐two” principle ⁷ , ⁸ ); however, still high enough to ensure adequate central nervous system (CNS) exposure, as extrapolated from cerebrospinal fluid concentrations in previous dose groups.

The trial was initiated in four site locations in two different European countries: The Netherlands and Spain. The trial was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice of the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH‐GCP). The first participant enrolled in September 2022. The last participant completed treatment in March 2024. The last safety follow‐up visit was performed in July 2024. Informed consent was obtained by the participants before the screening visit. The primary trial outcome was safety (incidence of treatment‐emergent AEs within the 2‐month study period), assessed through AE/serious AE (SAE) reporting and participant physical evaluations, including vital signs, blood pressure, 12‐lead electrocardiogram, hematology, blood biochemistry, and urinalysis. Clinical safety evaluations included the Columbia Suicide Severity Rating Scale. Exploratory outcomes included REM127 PK in plasma and CSF, mean changes from baseline at Day 29 in cognition, qEEG power, and CSF biomarkers reflecting AD pathophysiology, including Tau, Aβ, and neuroinflammation.

RESEARCH IN CONTEXT

1. Systematic review: Traditional (PubMed) sources for publications on the role of the septin cytoskeleton in relation to neurodegeneration were reviewed. These sources are acknowledged appropriately.

2. Interpretation: The results of this study show that a molecular septin 6/7 glue administered to participants with mild‐to‐moderate Alzheimer's disease (AD) improves outcomes reflecting pathophysiology, synaptic function, and cognition after 28 days of treatment. Hence, interventions aimed to restore the septin cytoskeleton may have therapeutic potential in AD. These study results confirm and extend existing literature data on the importance of the septin cytoskeleton for neuronal function in the human brain.

3. Future directions: The results of this study offer a basis for additional investigations to assess the therapeutic potential of septin 6/7 molecular glues with improved safety profiles.

2.2. Eligibility criteria

The trial included participants 50 to 85 years of age, with a Mini‐Mental State Examination (MMSE) above 12 and a maximum of 24 with either mild or moderate AD related dementia on the basis of National Institute on Aging–Alzheimer's Association criteria (NIA‐AA) ⁹ based on CSF measurement of Aβ142 and P‐tau concentrations (Aβ42 < 1000 pg/mL; P‐tau181 > 19 pg/mL or ratio P‐tau181/Aβ42 > 0.020). All the participants had clear slowing of EEG as assessed by the central EEG reader. If treated with an approved cholinesterase inhibitor, the treatment must be stable for at least 6 months prior to the screening visit with a dose which is not expected to change during the study, as per the investigator's judgment, or the participant must be off such medication for a period of 8 weeks prior to screening.

2.3. Blinding and randomization

Under protocol versions 1.0 to 3.0, participants were randomized 1:1:1 into placebo, 350 mg REM127, or 1400 mg REM127. Under protocol version 4.1, participants were randomized 1:2 into placebo and 88 mg REM127. The randomization list was developed by an independent statistical group (Pentara Corporation, USA). A total of 14 participants were randomized and treated in the safety population (Figure 1). Eurofins (France) packaged, labeled, and distributed medication kits for the study. Medication kits received by the participants were labeled only with randomization numbers. The sponsor's personnel, study sites’ personnel, participants, and caregivers were blinded to the assigned treatment. Participants self‐administered medication twice daily (morning and evening) as a liquid formulation dosed with a dosing syringe, under supervision of the caregiver. A single administration of medication consisted of 10 mL of microsuspension under protocol versions 1.0 to 3.0 and 2.5 mL of microsuspension under protocol version 4.1. All microsuspensions containing REM127 or placebo were identical.

FIGURE 1.

Participant flow diagram of REMAD‐02. *Eight participants did not meet the EEG criteria. EEG, electroencephalography.

2.4. Sample size determination

In in‐vivo models, drug‐related efficacy was observed within 2 weeks of treatment with effect sizes to almost healthy control levels ⁶ . Based on the nonclinical CSF tau and EEG treatment outcomes, sample size was determined prospectively based on power calculations that assumed the drug effect was equal to half the difference between healthy controls and AD patients, for total tau, which corresponds to Cohen's d = 1.24. With a type 1 error rate of 0.05, 18 subjects would provide 80% power to detect an effect size of d = 1.24. These 18 subjects would be split into groups of 6 subjects per arm (two active dose arms and a placebo arm). Nonclinical data for EEG theta power indicated that a large effect size might be present for detecting group differences. With a group mean difference of 0.082 and a pooled standard deviation of 0.053, the Cohen's d is equal to 1.55. With 80% power and an alpha level of 0.05, 15 subjects would be necessary to detect an effect size of 1.55 (5 subjects per arm).

2.5. PK evaluation

PK samples were collected on Day 8 pre‐dose and at 0.5, 1, 2, 3, and 6 h after the morning dose, on Days 15 and 29 at 3 h after the morning dose, and on Day 36. For the 88 mg dose group, PK samples were collected on Days 8 and 15 pre‐dose, on Day 22 pre‐dose, and at 0.5, 1, 2, 3, and 6 h after morning dose, on Day 29 pre‐dose and 3 h after morning dose, and on Days 36, 43, 50, and 57. CSF was collected on Day 29, 3 h after morning dose. Drug concentrations were determined using a validated LC‐MS/MS assay.

2.6. Exploratory pharmacodynamic outcome measures

2.6.1. CSF biomarker and neurotransmitter measurements

Baseline and end‐of‐study CSF samples were analyzed side‐by‐side to reduce batch assay variability. All analyses were performed blinded to the clinical information. The samples were analyzed for total‐tau (T‐tau), tau phosphorylated at amino acid 181 (P‐tau181), and 40 and 42 amino acid‐long Aβ (Aβ40 and Aβ42) using Lumipulse technology (Fujirebio, Japan). ¹⁰ For neurofilament light (NfL) and glial fibrillary acidic protein (GFAP), single molecule array (Simoa) technology was used on an HD‐X platform (Quanterix, USA). For chitinase 3‐like 1 (YKL40/CHI3L1), MicroVue technology was used from Quidel Corporation, USA. For neurogranin (NRGN), soluble triggering receptor expressed on myeloid cells 2 (sTREM2), neuronal pentraxin 2 (NPTX2), and vesicle‐associated membrane protein 2 (VAMP2), ELISA assays were used from Fujirebio (ADx Neurosciences, Belgium). The synaptopathophysiology mass spectrometry‐based biomarker panel was analyzed according to the methods described in ref [11]. The neurotransmitters dopamine and gamma‐amino butyric acid (GABA) were measured with an Agilent 1260 Infinity system coupled to a triple quadrupole Sciex API 4500 LC/MS/MS detector (Agilent, USA).

2.6.2. EEG measurements

EEG recordings were performed at screening, on Day 1, Day 15, and Day 29 or early termination with a Sienna Ultimate Amplifier. All site personnel performing EEG recordings were trained and qualified by The Siesta Group (Austria). Quantitative EEG measurement consisted of a 4‐min vigilance‐controlled EEG with eyes closed (ECV), a 4‐min resting EEG with eyes closed (ECR), and a 4‐min resting EEG with eyes open (EOR). All data were reviewed for quality and analyzed by Siesta. This analysis included artefact minimalization and detection, re‐referencing to common average reference, spectral analysis, determination of individual alpha frequency (IAF) and theta/alpha transition frequency (TF), as well as statistical analysis. The IAF was determined as the maximum peak in the Gaussian fit model after subtraction of the estimated aperiodic component from the power spectral density (PSD) averaged over all 4 time points and all parietal and occipital leads for the eyes closed condition. ¹² The TF between the theta and the alpha band was defined as the first local minimum below the IAF in the Gaussian fit. ¹³ Per‐subject individual frequency bands were determined according to Babiloni et al. ¹⁴ : TF‐4 ≤ δ < TF‐2; TF‐2 ≤ θ < TF; TF ≤ α1 < IAF; IAF ≤ α2 < IAF+2; IAF+2 ≤ β1 < 18.5; 18.5 ≤ β2 < 21.5; 21.5 ≤ β3 < 30; 30 ≤ γ < 40.

2.6.3. Cognitive testing

Cognitive outcomes included the computerized cognitive assessment (PROTECT Cognitive Test System [PCTS]), complemented with standard tests, including MMSE, selected tests from ADAS‐Cog (Comprehension, Orientation, Spoken Language Ability, Word Finding Difficulty, Word Recall, Delayed Word Recall), a variant of the digit symbol substitution test (WAIS‐IV coding), letter fluency (D‐KEFS Verbal), and Amsterdam‐instrumental activities of daily living (A‐iADL) scale. The PCTS ¹⁵ is a neuropsychological assessment battery that is completed online. In this study, participants completed four PCTS tests (Simple Reaction Time, Choice Reaction Time, Digit Vigilance, Delayed Picture Recognition) to enable assessment of attention, accuracy, and episodic memory. For each participant, geometric means of the outcome measures for each cognitive domain were calculated and used for statistical analysis.

3. STATISTICAL ANALYSIS

Multiplicative or fold change from baseline of each individual patient was calculated to correct for biological variation by dividing the measurements by the corresponding baseline value (no change = 1). Corresponding percent change was calculated as follows: (multiplicative change‐1)*100%. The primary statistical analysis compared the active therapy and placebo groups using a Mann–Whitney test. Supplementary, secondary analysis included a Wilcoxon signed‐rank test within the active therapy group for select outcomes to assess treatment effects relative to baseline. This supplementary analysis should not to be interpreted as direct evidence of treatment effectiveness. Consistent with an exploratory trial approach, statistical analyses presented in this paper were not corrected for multiple comparisons. p < 0.05 was set as threshold to indicate statistical significance. Missing data were handled using forward imputation of the last known datapoint—or imputation using “last observation carried forward.”

Topographical EEG analysis: variables were transformed appropriately to approximate normal distribution (log transformation for absolute power variables and indices; log(x/(100‐x)) transformations for relative power variables; no transformation for frequency variables). The results of the t‐tests per EEG target variable were displayed as statistical probability maps (SPMs). The thresholds were set, according to the degrees of freedom, to p < 0.01, p < 0.05, and p < 0.10. EEG power analysis: A linear mixed model was fit to model the log‐transformed fold change as a function of treatment group, electrode, and group X electrode interaction, allowing a separate treatment effect to be estimated for each electrode. In each model, the log‐transformed baseline measurement was included as a covariate, and a random participant intercept was included to adjust for clustering of measurements within participants. For each combination of frequency variable and electrode, geometric mean fold change under placebo and treatment was estimated from the fitted mixed model, and the effect of treatment was quantified as the multiplicative effect of treatment on fold change.

Using the R package npmv, which implements nonparametric multivariate tests for outcome variables with different scales and distributions, ¹⁶ a multivariate hypothesis test was carried out to compare the placebo and active therapy groups on fold change from baseline for Delayed Picture Recognition Score, Absolute individual EEG power at electrode Cz, and P‐tau181.

4. RESULTS

4.1. Participant disposition

The disposition of participants is shown in Figure 1. A total of 34 participants were screened for inclusion, and 15 participants were enrolled in the trial (1 discontinued treatment on Day 1 as the trial was terminated for safety reasons as described below). Fourteen participants were included in the safety population. Two participants discontinued treatment on Day 23 upon sponsor request due to observed increased alanine aminotransferase (ALT) levels in other participants, and 1 participant received quetiapine treatment (an atypical antipsychotic) from Day 21 until Day 23. Eleven participants completed the study as outlined in the protocol and were included in the per‐protocol population. Baseline characteristics of the trial cohort are outlined in Table 1.

TABLE 1.

Demographic and clinical characteristics of safety population.

Parameter	Placebo (N = 4)	88 mg/day (N = 5)	350 mg/day (N = 3)	1400 mg/day (N = 2)
Age (years)	69 ± 12 (54, 82)	74 ± 5.3 (66, 80)	76 ± 5.6 (71, 82)	72 ± 19 (58, 85)
Male (%)	2 (50%)	2 (40%)	2 (67%)	0 (0%)
Female (%)	2 (50%)	3 (60%)	1 (33%)	2 (100%)
MMSE at screening	18 ± 1.7 (16, 20)	19 ± 2.3 (15, 21)	19 ± 1.2 (18, 20)	12.5 ± 0.7 (12, 13)
BMI	28 ± 3.0 (24, 31)	23 ± 2.0 (20, 25)	28 ± 6.3 (21, 33)	26 ± 1.0 (25, 26)
Aβ42 (pg/mL)	533 ± 227 (390, 868)	479 ± 113 (278, 550)	689 ± 253 (454, 957)	567 ± 46 (534, 599)
P‐tau181 (pg/mL)	28 ± 15 (18, 50)	41 ± 21 (26, 78)	30 ± 4 (25, 32)	62 ± 36 (36, 87)
P‐tau181/Aβ42	0.052 ± 0.011 (0.038, 0.063)	0.089 ± 0.040 (0.050, 0.14)	0.046 ± 0.011 (0.033, 0.055)	0.11 ± 0.055 (0.067, 0.15)
Using AchEIs (%)	3 (75%)	5 (100%)	1 (33%)	1 (50%)
Using memantine (%)	1 (25%)	0 (0%)	1 (33%)	1 (50%)

Note: Continuous data are represented as the mean +/‐ SD and (min, max). Categorical data are represented as the number of participants (percentage).

Abbreviations: Aβ, amyloid‐beta protein; AchEI, acetylcholinesterase inhibitor; BMI, body mass index; MMSE, Mini‐Mental State Examination; P‐tau, hyperphosphorylated tau.

4.2. Tolerability and safety

Ten participants experienced AEs considered to be related to study medication (Table 2). Of these participants, two received placebo, three received 88 mg, three received 350 mg, and two received 1400 mg of REM127. AE reporting to the system organ class hepatobiliary dysfunction were observed in participants on active treatment and not in any of the placebo participants. One SAE occurred in the study in one participant in the 350 mg dose group. This SAE was considered related to the study treatment and was reported as toxic hepatitis. Hepatobiliary dysfunction was reported in six participants leading to study discontinuation. The most frequently observed AEs were increased serum lactate dehydrogenase (LDH) levels, increased serum transaminase levels and post lumbar puncture syndrome. All participants in the 1400 mg (2/2) and the 350 mg group (2/2) (except for a third participant who did not complete the treatment) and 1 out of 5 participants in the 88 mg group had increased serum transaminases versus none in the placebo group. Hepatic‐biliary dysfunction was observed only in participants treated with REM127. The increase in transaminases typically started in the fourth week—usually noticed either at last day of study drug intake, or at the follow‐up visits, peaking days to weeks after study discontinuation (Table S1). One case fulfilled Hy's Law criteria and led to hospitalization of the participant and was reported as suspected unexpected serious adverse reaction (SUSAR). Because of this event, administration of the study drug was discontinued in 2 participants (1 placebo and 1 in the 350 mg group) who were in the study at Day 23 when the decision to discontinue the trial was made. The increase in transaminases returned to normal levels in all participants. The time course to resolution was several weeks to several months after peak transaminase values. Among participants treated with REM127, a decrease in total cholesterol was noted (Table S2). No significant abnormalities within the placebo or REM127 groups were identified for participant vital signs (blood pressure, heart rate, respiratory rate, and body temperature) and 12‐lead electrocardiogram.

TABLE 2.

Safety of REM127 in mild‐to‐moderate AD.

Category	Placebo (N = 4)			88 mg (N = 5)			350 mg (N = 3)			1400 mg (N = 2)			All subjects (N = 14)
	n	%	freq	n	%	freq	n	%	freq	n	%	freq	n	%	freq
TEAE	3	75	6	5	100	20	3	100	16	2	100	14	13	92.9	56
Drug relationship
Related	2	50	2	3	60	4	3	100	7	2	100	9	10	71.4	22
Not related	3	75	4	5	100	16	3	100	9	2	100	5	13	92.9	34
Intensity
Mild	3	75	6	5	100	18	3	100	12	2	100	10	13	92.9	46
Moderate	0	0	0	2	40	2	2	67	3	2	100	4	6	42.9	9
Severe	0	0	0	0	0	0	1	33	1	0	0	0	1	7.1	1
SAEs	0	0	0	0	0	0	1	33	1	0	0	0	1	7.1	1
AE leading to death	0	0	0	0	0	0	0	0	0	0	0	0	0	0.0	0
Most common TEAEs
Hepatobiliary disorders
Total	0	0	0	1	20	2	3	100	5	2	100	3	6	42.9	10
Alanine aminotransferase increased	0	0	0	0	0	0	1	33	1	0	0	0	1	7.1	1
Blood lactate dehydrogenase increased	0	0	0	0	0	0	2	67	2	1	50	1	3	21.4	3
Drug‐induced liver injury	0	0	0	1	20	1	0	0	0	0	0	0	1	7.1	1
Hepatic enzyme increased	0	0	0	0	0	0	0	0	0	1	50	1	1	7.1	1
Hepatitis toxic	0	0	0	0	0	0	1	33	1	0	0	0	1	7.1	1
Transaminases increased	0	0	0	1	20	1	1	33	1	1	50	1	3	21.4	3
Injury, poisoning, and procedural complications
Post lumbar puncture syndrome	1	25	1	2	40	2	1	33	2	1	50	1	5	35.7	6

Abbreviations: AD, Alzheimer's disease; AE, adverse event; freq, the total number of events (multiple events may occur per participant); N, the number of participants exhibiting an event as reported by the investigator; SAE, serious adverse event; TEAE, treatment emergent adverse event.

4.3. PK

REM127 reached maximum observed plasma concentration (t_max) around 2 h post‐dose in all dose groups. The mean terminal elimination half‐life (t_1/2) was 320, 438, and 396 h for the 88, 350, and 1400 mg/day groups, respectively (Figure 2A and Table S3). A dose‐dependent relationship in circulating drug levels across the tested dose levels was observed.

FIGURE 2.

REM127 exposure in plasma and CSF at indicated tested doses. (A) Pharmacokinetic profiles. (B) Concentrations in CSF at end of treatment (EoT). Values above the bars indicate fold margin relative to the affinity constant of septin 6. For the 88, 350 and 1400 mg doses the number of patients analyzed per dose group was n = 5, n = 3, n = 2, respectively. CSF, cerebrospinal fluid.

REM127 exposure in the CNS was measured in CSF samples taken on Day 29 at 3 h after the last dose. REM127 concentrations in CSF were dose‐dependent (Figure 2B) and were 8/44/118‐fold above the dissociation constant (Kd) of the target septin 6 ⁶ for the dose levels 88/350/1400 mg/day. The unbound CSF‐to‐plasma partition coefficient (K_{p, uu, brain}) was 1.4.

4.4. Analysis of pharmacodynamic endpoints

Data collected from participants after 15 days and 28 days were used to study potential drug‐related pharmacodynamic effects. At all tested doses, CSF exposures were consistent with full target engagement (Figure 2B). Therefore, the data of the three dose groups were combined and analyzed as one active treatment (verum) group.

4.5. Exploratory outcomes

4.5.1. CSF biomarkers of pathophysiology and neurotransmitters

A nominally statistically significant difference of verum versus placebo was found for CSF GFAP, P‐tau181, but not T‐tau (Figure 3). In the verum group, a nominally statistically significant longitudinal decrease of T‐tau and P‐tau181 concentrations of minus 9% and minus 8%, respectively, was observed (Figure S1A). No other tested biomarkers of pathophysiology were changed significantly between the groups. No dose‐dependent effects were observed (Figure S1B).

FIGURE 3.

Pathophysiological biomarker outcomes. Change from baseline after 28 days of treatment of indicated analytes in CSF of the placebo and verum group (*p < 0.05). CSF, cerebrospinal fluid.

Retrospective analysis of CSF neurotransmitter concentrations in the verum group revealed a nominally statistically significant increase of dopamine relative to the placebo (Figure 3). Compared to baseline, CSF dopamine was nominally statistically significantly increased by 9% in the verum group (Figure S1C). The effects were not dose‐dependent, and GABA levels in CSF were not significantly changed (Figure S1C left panel).

4.5.2. EEG

IAF and theta/alpha TF were determined (Table S4). An overall increase of individual (i)EEG power in the active treatment group, particularly at central electrodes (Figure 4 and Figure S2), was observed, most prominently of the alpha1 and beta/gamma frequencies. The increases of total iEEG power and of the beta2, beta3, and gamma frequencies were nominally statistically significant compared to placebo (Figure 4A, S2A). Nominally statistically significant decreases of relative i‐alpha2 power as well as a slowing of the i‐alpha1 centroid across most electrodes were observed in the active treatment group relative to placebo (Figure S2A). Within‐group analysis of the active therapy group revealed nominally statistically significant increases relative to baseline of iEEG total power and of the theta, alpha1, and beta1‐3 frequencies mostly at the central electrodes (Figure S2B). The power of the individual total and alpha1 frequency bands was elevated in REM127‐treated participants at Day 15 and Day 29, up to an approximately 50% change from baseline (Figure 4B).

FIGURE 4.

The iEEG power analysis (ECR). (A) Change from baseline after 28 days of treatment of iEEG power at indicated frequency bands of the placebo or verum group. Each of the dots indicates geometric mean change from baseline of iEEG power of the placebo and verum group for each of the 19 electrodes (* p < 0.05). (B) Geometric mean change from baseline of individual (i) total power and i‐alpha1 power for all 19 electrodes in function of the indicated days of treatment. Green dotted line and percentage value indicate the projected improvement required to fully restore late‐onset AD to healthy controls as reported in literature. ³⁰ ECR, resting EEG with eyes closed; EEG, electroencephalography; iEEG, individual electroencephalography.

4.5.3. Cognition

Participants on active therapy recognized pictures nominally statistically significantly faster in the PROTECT memory test and had an improved score in the ADAS‐Cog orientation task compared to placebo (Figure 5). In the verum group, a nominally statistically significant faster response time was observed for PROTECT picture recognition and attention tests by −26% and −19%, respectively, and an improved performance of +40% in the coding test, a measure of executive function relative to baseline (Figure S3A).

FIGURE 5.

Cognitive outcomes. Change from baseline after 28 days of treatment period of indicated cognitive tests of placebo and verum group (*p < 0.05).

None of the observed significant effects were dose‐dependent (Figure S3B). No significant between‐group difference was found for the PROTECT attention and accuracy scores, MMSE, the Letter Fluency test, and the ADAS‐Cog coding, word recall, and delayed word recall tasks (Figure 5).

4.5.4. Global statistical test

A multivariate hypothesis test was carried out to compare the placebo versus the active treatment group on change from baseline for three outcomes: delayed picture recognition score, absolute alpha1 power at electrode Cz, and CSF P‐tau181. Drug placebo difference on this measure was significant in favor of active therapy (p < 0.0001).

4.5.5. Biomarker associations

An association analysis of cognitive and CSF biomarker outcomes of all participants was performed retrospectively (Figure 6). Treatment effects of dependent analytes which share a common mechanistic origin (T‐tau vs. P‐tau181 and Aβ40 vs. Aβ42) were found highly correlated. The mean response time to picture recognition associated statistically significantly associated with the change of pathological markers T‐tau, P‐tau181, and Aβ40, neuroinflammatory marker GFAP, and with synaptic marker NPTX2. T‐tau and P‐tau181 changes were found strongly associated with the effects on synaptic markers NPTX2 and NRGN. No statistically significant associations were found between changes in memory and sTREM2, NfL, NRGN, and YKL40.

FIGURE 6.

Association matrix of treatment outcomes. Association matrix of the individual longitudinal effect sizes of all participants (placebo and verum group combined) between indicated endpoints across participants.

4.5.6. Synaptopathophysiology panel

No statistically significant change between the groups was found for CSF‐based biomarkers reflecting AD synaptopathophysiolology (Figure S4A). In the verum group, a nominally significant change from baseline was observed for peptides representing A4, a non‐amyloidogenic amyloid precursor protein (APP) processing product, chromogranin‐A (CMGA) and secretogranin‐2 (SCG2), calsyntenin 1 (CSTN1), VGF nerve growth factor inducible (VGF), and phosphatidylethanolamine‐binding protein 1 (PEBP‐1) (Figure S4B). Other analytes quantified were not significantly changed from baseline.

5. DISCUSSION

We conducted a phase 2a, double‐blind, randomized placebo‐controlled clinical trial to assess safety and exploratory efficacy of REM127, a novel modulator of septin filament function, in mild‐to‐moderate AD dementia participants. Signs of liver adverse effects as evidenced by increases in serum aminotransferase in the REM127 treatment groups were the most common AE reported and led to discontinuation of the study. All patients at the highest doses tested (350 and 1400 mg) who had received the study drug for the intended treatment period had elevated plasma ALT levels three times above upper‐limit‐of normal, whereas at the lowest tested dose (88 mg), the frequency was lower (1/5), suggesting a dose‐dependency of the hepatic effect. These ALT elevations returned to normal after the treatment period, indicating these effects are reversible. The precise mechanism underlying the increased ALT levels is not known. Retrospective nonclinical studies have revealed REM127 accumulates in hepatocytes by an active transport mechanism independent of septin modulation (data not shown), which possibly underlies the observed liver adverse effects and the relatively long observed half‐life of plasma REM127. Non‐clinical chronic toxicology studies and a 7‐day phase 1 study in healthy volunteers did not reveal liver adverse effects comparable to those observed in the current study (data not shown). The delayed onset (3–4 weeks of treatment) and the human‐specific nature of these effects suggest a mechanism of toxicity not evident in shorter‐duration studies or animal models. Mechanistic studies designed to characterize the mechanism of transport to human hepatocytes could have been informative to better anticipate and mitigate potential liver liabilities.

CSF T‐tau and P‐tau181 levels were significantly lowered after active treatment suggesting a hallmark of AD pathology is modified by REM127 treatment. Exocytosis of tau to the extracellular environment is driven by cytosolic Ca²⁺ in neurons ¹⁷ in line with elevated levels of tau in CSF in AD patients. The reduced levels of T‐tau and P‐tau181 in CSF are consistent with pharmacodynamic lowering of cytosolic Ca²⁺ levels toward normal physiological levels, ⁶ expected with the proposed on‐target mechanism of REM127. Other tested biomarkers of AD pathophysiology for which the study was not powered were not significantly changed. Larger group sizes and/or longer treatment duration may be more appropriate to assess impact on these outcomes by septin modulatory drugs.

CSF dopamine levels were significantly increased in the active treatment group with a meaningful effect size based on reference data from existing literature, suggesting improved function of presynaptic dopaminergic terminals to release dopamine. Also, a subset of tested CSF‐based biomarkers reflecting AD synaptic pathophysiology were significantly changed relative to baseline in the active therapy group, confirming a rapid impact of REM127 on synaptic integrity/function.

EEG activity reflects the summation of the synchronous electrical activity of spatially aligned cortical neurons. The power and synchronization of different frequency bands reflect functional connectivity between neuronal circuits underlying cognitive and sensory functions. ¹⁸ , ¹⁹ The alpha frequency band is a key feature of waking EEG, associated with cortical excitability ²⁰ and linked to processes such as perception and memory. ²¹ , ²² , ²³ In AD, a reduction of EEG activity, particularly in the alpha1 frequency band, is observed, ²⁴ , ²⁵ which is thought to be linked to cognitive impairment. ²⁶ , ²⁷ , ²⁸ REM127 treatment led to increased EEG power mostly of the alpha1 frequency band, indicating improved synaptic activity of corresponding neurons, possibly linked to cognitive functioning. The effect on EEG alpha1 power had a rapid onset (15 days) and a meaningful effect size based on reference data from the available literature. These data suggest at least a partial reversal of disease‐associated reductions in brain network activity in REM127‐treated participants. Further in line with the nonclinical data ⁶ quantitative EEG appears to be a sensitive, noninvasive pharmacodynamic biomarker to evaluate effects of septin modulation in AD dementia trial participants.

The impact of REM127 on outcomes reflecting synaptic integrity and function corresponded with better performance in cognitive tests, suggesting REM127 treatment leads to a symptomatic benefit.

Collectively, the data from this small and interrupted study suggest that targeting the septin cytoskeleton in AD participants with a Sept6/7 molecular glue restores synaptic activity and enhances cognition within 28 days. This conclusion is based on four independent outcomes: EEG, CSF dopamine levels, synaptic markers in CSF, and cognitive performance in favor of active therapy. Further support is provided by multivariate testing using three outcomes: delayed picture recognition score, absolute individual alpha1 power for electrode Cz, and CSF P‐tau181‐ demonstrating a robust outcome of REM127 effect.

In addition to outcomes of a symptomatic benefit, CSF concentrations of markers reflecting core AD pathology were also mitigated by REM127, indicating a possible effect on canonical features of AD. The decrease of the A4 peptide (Figure S4), a non‐amyloidogenic APP‐derived peptide, suggests lowered APP processing consistent with an impact on Ca²⁺ homeostasis. ⁵ , ²⁹ These results confirm the non‐clinical efficacy data, ⁶ which demonstrated REM127 lowered both tau and Aβ pathology and rescued survival of neuronal cells challenged with pathological tau or Aβ.

An obvious limitation of this study is the relatively small group sizes due to the interruption of the study. Not all outcomes in the verum group were directionally positive, and additional trials with larger sample sizes are required to understand the repertoire of outcomes characteristic of septin 6/7 modulatory drugs. Strengths of the observations that allow preliminary exploration of treatment effects include: (1) all key outcomes were guided by nonclinical in‐vivo data; (2) all participants had biologically confirmed AD; (3) the directionality of independent outcomes (pathology, EEG, CSF dopamine, CSF synaptic markers and cognition) are congruent; (4) the effect sizes of the changes from baseline of the different outcomes were (a) anticipated in the protocol and (b) associated significantly across participants consistent with common underlying mechanism; (5) global statistical testing using independent anticipated outcomes revealed a consistent drug treatment effect unlikely to be observed under the null hypothesis (no drug effect).

In conclusion, the primary endpoint of this study was not met for REM127, primarily attributed to an unfavorable safety profile related to liver adverse effects. At all tested doses, oral treatment with REM127 led to brain exposures consistent with saturated target engagement. Analysis of exploratory endpoints indicated potential symptomatic and disease‐modifying effects in participants with biologically confirmed mild‐to‐moderate AD in favor of active therapy. These pharmacodynamic effects were dose‐independent, consistent with full target occupancy in all REM127 treatment groups. The liver adverse effects, however, unlike all anticipated on‐target efficacy outcomes and despite drug levels in CSF indicating full target occupancy, were dose‐dependent, indicating an off‐target mechanism. Assessment of an optimized septin modulator with improved safety profile to interrogate the therapeutic effects of septin modulation in AD is warranted.

CONFLICT OF INTEREST STATEMENT

P.A. is a paid consultant of The Siesta Group Schlafanalyse GmbH. J.E.H. is an employee and shareholder of Metis Cogniton Ltd and has previously received consultancy fees from remynd. K.D.W. has received consultancy fees from remynd and owns remynd warrants and shares. S.R. is paid consultant for remynd. G.G. is paid consultant for remynd and Babylon Biosciences and owns remynd warrants and shares. M.F., E.D. and K.P. own remynd warrants. H.Z. has served at scientific advisory boards and/or as a consultant for Abbvie, Acumen, Alector, Alzinova, ALZpath, Amylyx, Annexon, Apellis, Artery Therapeutics, AZ Therapies, Cognito Therapeutics, CogRx, Denali, Eisai, Enigma, LabCorp, Merry Life, Nervgen, Novo Nordisk, Optoceutics, Passage Bio, Pinteon Therapeutics, Prothena, Quanterix, Red Abbey Labs, remynd, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave, has given lectures sponsored by Alzecure, BioArctic, Biogen, Cellectricon, Fujirebio, Lilly, Novo Nordisk, Roche, and WebMD, and is a co‐founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). J.C. received consultancy fees from Acadia, Acumen, ALZpath, Annovis, Aprinoia, Artery, Axsome, Biogen, Biohaven, BioXcel, Bristol‐Myers Squib, Cervomed, Eisai, Fosun, GAP Foundation, Green Valley, Hummingbird Diagnostics. IGC, Janssen, Kinoxis, Lighthouse, Lilly, Lundbeck, LSP/eqt, Mangrove Therapeutics, Merck, MoCA Cognition, New Amsterdam, Novo Nordisk, NSC Therapeutics, Optoceutics, Otsuka, Oxford Brain Diagnostics, Praxis, Prothena, Remynd, Roche, Scottish Brain Sciences, Signant Health, Simcere, sinaptica, T‐Neuro, TrueBinding, and Vaxxinity pharmaceutical, assessment, and investment companies.  E.V. received consultancy fees for New Amsterdam Pharma, Treeway, remynd, Vivoryon, Biogen, Vigil Neuroscience, ImmunoBrain Checkpoint, Muna Therapeutics, Esai, Eli Lilly, CogRX, Therini, UCB and Roche. C.E.T. has research contracts with Acumen, ADx Neurosciences, AC‐Immune, Alamar, Aribio, Axon Neurosciences, Beckman‐Coulter, BioConnect, Bioorchestra, Brainstorm Therapeutics, Celgene, Cognition Therapeutics, EIP Pharma, Eisai, Eli Lilly, Fujirebio, Instant Nano Biosensors, Novo Nordisk, Olink, PeopleBio, Quanterix, Roche, Toyama, Vivoryon. C.E.T. received consultancy fees from Aribio, Biogen, Beckman‐Coulter, Cognition Therapeutics, Eisai, Eli Lilly, Merck, Novo Nordisk, Novartis, Olink, Roche, Sanofi and Veravas. XM received research grants from Novo Nordisk and Kern Pharma. M.B. received consulting fees from Grifols, Araclon Biotech, Roche, Biogen, Lilly, Merck, Zambon, Novo‐Nordisk and received payments for lectures sponsored by Roche, Biogen, Grifols, Nutricia, Araclon Biotech, Servier, Novo‐Nordisk. C.B. received grants, contracts or consultancy fees from Novartis, Johnson and Johnson, Novo Noridsk, Roch, remynd, Acadia, AARP, Eli Lily, BMS, Janssen, Orion, Exciva, Sunovion, Suven, Roche, Biogen, TauRx. A.C. received contracts, grants or consultancy fees from remynd, Novo Nordisk, Therini Bio, Suven, Johnson and Johnson. K.P. and G.G. are inventors on patent WO2013/004642 held by remynd; K.P., M.V., M.F., and G.G. are inventors on patent WO 2024/0033479 held by remynd; K.P., M.V., and G.G. are inventors on patent WO 2025/104092 held by remynd. M,N,, M.V., E.D., K.P., L.P., M.F. are employees of remynd. E.B. is a former employee of remynd., J.P.T., A.L., E.F.M., J.A.M.G., V.S., V.H., M.W., A.B.W., S.D., C.M., S.H. have nothing to disclose. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All participants in the present study provided signed informed consent.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Nuytten M, Voets M, Debroux E, et al. Randomized phase 2a trial assessing a novel septin molecular glue in Alzheimer's disease. Alzheimer's Dement. 2025;21:e70537. 10.1002/alz.70537