Pharmacodynamic effects of semorinemab on plasma and CSF biomarkers of Alzheimer's disease pathophysiology

Abstract

INTRODUCTION

Semorinemab, an anti‐tau monoclonal antibody, was assessed in two Phase II trials for Alzheimer's disease (AD). Plasma and cerebrospinal fluid (CSF) biomarkers provided insights into the drug's potential mechanism of action.

METHODS

Qualified assays were used to measure biomarkers of tau, amyloidosis, glial activity, neuroinflammation, synaptic function, and neurodegeneration from participant samples in Tauriel (NCT03289143) and Lauriet (NCT03828747) Phase II trials.

RESULTS

Plasma phosphorylated Tau 181 (pTau181) and CSF chitinase‐3‐like protein 1 (YKL‐40) increased following semorinemab treatment in both studies. In Lauriet, increasing plasma glial fibrillary protein (GFAP) concentrations stabilized with semorinemab, while this was not observed in Tauriel. Other AD pathophysiology biomarkers showed no consistent response to semorinemab.

DISCUSSION

Increases in CSF YKL‐40 suggest that semorinemab may stimulate microglia activation in the presence of AD‐associated Tau pathology, but not in healthy controls. Stabilization of plasma GFAP in Lauriet indicates a possible impact on reactive gliosis in mild‐to‐moderate AD.

Trial Registration: Tauriel ClinicalTrials.gov Identifier: NCT03289143. Lauriet ClinicalTrials.gov Identifier: NCT03828747. Phase 1 ClinicalTrials.gov Identifier: NCT02820896.

Highlights

• AD pathophysiology biomarkers were measured to assess the mechanism of action.

• Semorinemab increased CSF YKL‐40 in participants with AD but not in healthy controls.

• Semorinemab possibly stabilized plasma GFAP in the Lauriet trial.

• Semorinemab treatment may activate microglia and moderate reactive gliosis.

1. BACKGROUND

Alzheimer's disease (AD) is the most common form of dementia, representing 60%–80% of cases. It is progressive, with no known cure, and an estimated 6.7 million people in the United States are afflicted with the disease. ¹ AD is characterized by two main pathological hallmarks: extracellular amyloid plaques composed of aggregated amyloid‐beta (Aβ) peptides, and intracellular neurofibrillary tangles that consist of the protein tau. ² The spatial temporal burden of cortical tau neurofibrillary tangles tracks more closely with neuronal loss and ante mortem cognitive measures than amyloid ³ , ⁴ , ⁵ , ⁶ suggesting, along with mechanistic data from in vitro and in vivo models, ⁷ that tau is an active contributor to AD pathogenesis and a viable therapeutic target. Tau pathology spreads in a predictable sequence in AD ³ , ⁸ implying that it may be released extracellularly and propagates across interconnected brain networks in a prion‐like fashion. ⁹ Semorinemab, a humanized anti‐tau IgG4 monoclonal antibody under development for AD, is hypothesized to target the proliferation and accumulation of tau pathology by intercepting and clearing tau from the extracellular space.

After a Phase I study (NCT02820896), two Phase II studies in different AD patient populations were conducted: Tauriel (NCT03289143), a Phase II study of semorinemab in prodromal‐to‐mild (P2M) AD patients, and Lauriet (NCT03828747), a Phase II study in mild‐to‐moderate (M2M) AD patients. In Tauriel, semorinemab failed to demonstrate clinical efficacy at doses of 1500, 4500, and 8100 mg for 73 weeks. In contrast, a 4500 mg dose of semorinemab for 49 or 61 weeks in Lauriet was associated with a significant slowing in progression (vs. placebo) in a co‐primary endpoint of cognition (Alzheimer's Disease Assessment Scale‐Cognitive Subscale 11 (ADAS‐Cog11)) but it was not accompanied by improvements in daily function (Alzheimer's Disease Cooperative Study‐Activities of Daily Living (ADCS‐ADL)). In an attempt to reconcile divergent clinical outcomes with evidence of CSF tau target engagement in both studies, ¹⁰ , ¹¹ , ¹² exploratory post‐hoc assessments of AD pathophysiology biomarkers were performed to gain clearer insight into semorinemab's mechanism of action.

2. METHODS

2.1. Study participants: P2M and M2M AD (Tauriel and Lauriet) and Phase I healthy volunteers

In Tauriel (NCT03289143), 457 patients with prodromal or mild AD were recruited per criteria from the National Institute on Aging and Alzheimer's Association (NIA‐AA). Participants were aged 50–80 years old, with Mini‐Mental State Examination (MMSE) scores between 20 and 30, Clinical Dementia Rating (CDR) Global Score of 0.5 or 1, episodic memory deficit by Repeatable Battery for the Assessment of Neuropsychological Status Delayed Memory Index (RBANS DMI) ≤ 85, and positive for cerebral amyloid pathology by an Aβ positron emission tomography (PET) visual read or by reduced CSF amyloid‐β (1‐42) concentrations. Patients were randomized 2:3:2:3, and received intravenous (IV) doses of 1500, 4500, or 8100 mg semorinemab or placebo administered every 2 weeks (Q2W) for doses 1 through 3, and then every 4 weeks (Q4W) until the end of the double‐blind (DBL) study period (Week 73). ¹⁰ The 4500 mg and placebo cohorts had increased allocation to improve power for these groups since modeling suggested sufficient binding to extracellular Tau at 4500 mg. ¹⁰

In Lauriet (NCT03828747), 272 patients with mild‐to‐moderate AD were recruited; participants were aged 50–85, with MMSE scores between 16 and 21, CDR‐GS Global Score of 1 or 2, and positive for amyloidosis by PET imaging or reduced CSF amyloid‐β (1‐42) concentrations. Participants were randomized 1:1 to receive either placebo or 4500 mg semorinemab IV Q2W for doses 1 through 3, and then Q4W (every month) until the end of the DBL period (Cohort 1, Week 49). However, due to the COVID‐19 pandemic, a subset of patients (n = 76) missed more than two doses during the DBL treatment period. To account for the potential effect of missed doses, the DBL period was extended for an additional 12 weeks for these patients (Cohort 2, Week 61). Serum pharmacokinetics demonstrated that these patients had the same overall exposure as patients who missed none or 1 dose. Patients who completed the DBL treatment period had the option to enter into an open‐label extension, during which all patients received semorinemab 4500 mg Q4W for 96 weeks. Patients provided plasma samples for pharmacodynamic and pharmacokinetic analyses at baseline (pre‐dose) and Weeks 5, 25, 49, or 61. CSF collection was optional; of 272 patients, 48% (n = 131) provided CSF at baseline and 19% (n = 53) provided CSF at follow‐up. Identical serum pharmacokinetics between the Week 49 and 61 cohorts enabled the pooling of CSF results from each cohort to increase the statistical power. ¹¹ Exploratory CSF and plasma analyses were restricted exclusively to samples from participants who provided informed consent. Baseline demographics of the study populations are detailed in Table 1.

TABLE 1.

Baseline demographics.

Parameter	Tauriel	Lauriet
Study identifier	NCT03289143	NCT03828747
Diagnosis	Prodromal‐to‐mild AD	Mild‐to‐moderate AD
n	431	267
Age (mean (SD))	69.3 (7.09)	71.8 (8.02)
Sex, M/F (n, %)	F: 241 (55.9) M: 190 (44.1)	F: 173 (64.8) M: 94 (35.2)
APOE4 ± (n, %)	APOE4+: 316 (73.3) APOE4‐: 115 (26.7)	APOE4+: 167 (62.5) APOE4‐: 100 (37.5)
MMSE (mean (SD))	23.3 (2.68)	18.2 (2.04)

Abbreviations: AD, Alzheimer's disease; APOE4, apolipoprotein E4; MMSE, Min‐Mental State Examination.

In a Phase I trial to evaluate the safety, tolerability, pharmacokinetics, and activity of semorinemab (NCT02820896), 65 healthy volunteers were recruited and randomized in a 2:1 ratio (active:placebo) to receive 225 mg, or in a 6:2 ratio to receive 675, 2100, 4200, 8400, or 16800 mg IV doses of semorinemab or placebo. In addition, a single‐dose (SD) cohort received a 1200 mg subcutaneous dose of semorinemab (12:0). In two multiple‐dose (MD) cohorts, 10 healthy volunteers and 10 patients clinically‐diagnosed with M2M AD and confirmed amyloid PET positive by visual read were randomized 8:2 and received IV doses of 8400 mg semorinemab once weekly for four weeks (q1wx4). CSF samples for pharmacokinetic and pharmacodynamic measurements were collected only from healthy volunteers in the 2100 and 8400 mg SD IV cohorts and from healthy volunteers in the 8400 mg MD IV cohort at baseline, day 8, 15, or 29. Additional detail and information on the study design and results can be found in previously published research. ¹²

RESEARCH IN CONTEXT

1. Systematic review: We performed a comprehensive review of publicly available scientific databases and conference proceedings on Alzheimer's disease (AD), the role of tau protein in AD pathogenesis, pathophysiological biomarkers of AD, and therapies targeting AD pathology. These sources are acknowledged appropriately.

2. Interpretation: Our findings, derived from two Phase II trials for semorinemab, shed light on the mechanistic consequences of modulating extracellular tau levels with an immunotherapeutic within the context of AD. Semorinemab engaged tau in accordance with previously published data, and this was accompanied by increases in CSF YKL‐40, indicating a treatment‐specific activation of microglia. In mild‐to‐moderate AD patients who presented partially positive clinical outcomes, this was accompanied by stabilization of increasing plasma glial fibrillary acidic protein (GFAP) levels, suggesting that treatment may translate into a moderation reactive gliosis.

3. Future directions: The results of this study offer a basis for additional investigation into the effects of tau modulation by an immunotherapeutic. Future research should focus on an understanding of (i) how the interaction between anti‐tau antibodies and tau aggregates results in microglia activation; (ii) how increased microglia activity impacts reactive gliosis; and (iii) the mechanisms driving variations in response between different AD patient populations.

2.2. Cognitive and functional measures

The following cognitive and functional measures were collected in Tauriel and Lauriet. The Clinical Dementia Rating—Sum of Boxes (CDR‐SB) assesses six domains, each rated from 0 to 3, with higher total scores indicating greater severity of dementia. ¹³ The Alzheimer's Disease Assessment Scale—Cognitive Subscale (ADAS‐Cog) includes 11 or 13 tasks that evaluate cognitive performance, where higher scores reflect greater cognitive impairment. ¹⁴ The MMSE is a brief questionnaire covering various cognitive domains, with higher summed scores indicating better cognitive function. ¹⁵ Lastly, the Alzheimer's Disease Cooperative Study—Activities of Daily Living (ADCS‐ADL) evaluates a patient's ability to perform both basic and instrumental daily activities, with higher total scores indicating better overall function. ¹⁶

2.3. Biomarkers

Exploratory fluid biomarker analyses of participant CSF and plasma samples were conducted to examine various aspects of AD pathophysiology, including measures of tau, glial activity, neuroinflammation, synaptic integrity, synucleinopathy, neurodegeneration, and amyloidosis. A detailed list of all biomarkers, matrices, and the studies in which they were measured can be found in Table 2.

TABLE 2.

Summary of biomarker analyses.

Biomarker	Matrices	Biology	Studies	Assay type
Total tau	CSF, Plasma	Tau burden, target engagement	Tauriel, Lauriet	CE‐marked Elecsys (CSF), NeuroToolKit (plasma)
N‐Terminal tau	CSF	Target engagement	Tauriel, Lauriet	IPMS
pTau181	CSF, Plasma	Tau burden, target engagement	Tauriel, Lauriet	CE‐marked Elecsys (CSF), Elecsys RPA (plasma)
pTau217	CSF, Plasma	Tau burden, target engagement	Tauriel, Lauriet (Plasma only)	Simoa (CSF), NeuroToolKit (Plasma)
CXCL10	CSF	Glial activity, neuroinflammation	Tauriel, Lauriet	Simple Plex
GFAP	CSF, Plasma	Glial activity, neuroinflammation	Tauriel, Lauriet	NeuroToolKit
S100b	CSF	Glial activity, neuroinflammation	Tauriel, Lauriet	NeuroToolKit
sTREM2	CSF	Glial activity, neuroinflammation	Tauriel, Lauriet	NeuroToolKit
YKL‐40	CSF, Plasma	Glial and lysosomal activity	Tauriel (CSF only), Lauriet	NeuroToolKit
Neurogranin	CSF	Synaptic integrity and function	Tauriel, Lauriet	NeuroToolKit
SNAP‐25	CSF	Synaptic integrity and function	Tauriel, Lauriet	NeuroToolKit
Neuronal pentraxin‐2	CSF	Synaptic integrity and function	Tauriel, Lauriet	NeuroToolKit
Alpha‐synuclein	CSF	Synaptic integrity and function, synucleinopathy	Tauriel, Lauriet	NeuroToolKit
IL‐6	CSF	Neuroinflammation	Tauriel, Lauriet	NeuroToolKit
Lipocalin	CSF	Neuroinflammation	Tauriel, Lauriet	Simple Plex
β‐Amyloid (1‐42)	CSF	Amyloidosis	Tauriel, Lauriet	CE‐marked Elecsys β‐Amyloid (1‐42)
β‐Amyloid (1‐40)	CSF	Amyloidosis	Tauriel, Lauriet	Elecsys β‐Amyloid (1‐40)
Neurofilament‐light	CSF, Plasma	Neurodegeneration	Tauriel, Lauriet	NeuroToolKit
Neurofilament‐heavy	CSF	Neurodegeneration	Tauriel, Lauriet	Simple Plex

Abbreviations: CSF, cerebrospinal fluid; GFAP, glial fibrillary acidic protein; IL‐6, interleukin‐6; RPA, robust prototype assay.

2.4. Elecsys immunoassay measurements of CSF and plasma samples

The CSF β‐amyloid 1‐42 (Aβ1‐42), β‐amyloid 1‐40 (Aβ1‐40), phosphorylated tau T181 (pTau181), total tau (tTau) assays were validated to Clinical and Laboratory Standards Institute (CLSI) standards and demonstrated high intra‐ and inter‐assay precision, which enabled individual determinations and were implemented during the clinical trials at Covance CLS sites in Geneva (samples originating in Europe) and Indianapolis, IA (sample originating in the United States). Concentrations were calculated from 2‐point calibration curves, and two recombinant protein quality controls were required to quantify within 21% of their established concentration for data to be acceptable.

Elecsys electrochemiluminescence immunoassays were used to measure CSF and plasma samples. Plasma pTau181 was measured blinded on a cobas e 402 analyzer using a robust prototype assay (RPA) with a special extended measurement range. A suite of Elecsys assays collectively known as the Roche NeuroToolKit (NTK), which include both RPAs and CE‐marked in vitro diagnostic assays, was used to measure concentrations of α‐synuclein, glial fibrillary acidic protein (GFAP), interleukin‐6 (IL‐6), neurogranin (Ng), neurofilament light chain (NfL), neuronal pentraxin‐2 (NPTX2), synaptosomal‐associated protein 25 kDa (SNAP‐25), S100 calcium‐binding protein B (S100B), soluble triggering receptor expressed on myeloid cells 2 (sTREM2), and YKL‐40 in the CSF, and GFAP, NfL, tTau, pTau217 (early exploratory assay version) and YKL‐40 in plasma (Table 2). Blinded measurements were performed on cobas e 411 and e 601 instruments (Roche Diagnostics International Ltd, Rotkreuz, Switzerland).

2.5. Simple Plex immunoassay measurements of CSF samples

C‐X‐C motif chemokine ligand 10 (CXCL10), lipocalin (LCN), neurofilament heavy Chain (NfH) from CSF were measured with SimplePlex assays on the ProteinSimple Ella platform (Bio‐Techne, San Jose, CA). Baseline and follow‐up samples were measured in pairs in a single batch. Concentrations were calculated from factory calibration curves specific to each assay lot, and the intra‐assay coefficient of variation (CV) was required to be ≤20% for data acceptability.

2.6. Simoa immunoassay measurements of CSF samples

Concentrations of CSF pTau217 were measured with a custom single molecule array (Simoa) assay (Quanterix, Billerica, MA) developed in‐house. The analyte was captured with a monoclonal antibody specific for the pTau217 epitope (Roche Diagnostics, GmbH, Penzberg, Germany), and then detected with a biotinylated monoclonal antibody specific for the projection domain of tau (125B11H3, Genentech Inc., South San Francisco, CA). Samples were analyzed on the HD‐1 system (Quanterix, Billerica, MA) where the fluorescent signal in measured samples was converted to an average number of enzymes per bead (AEB). Concentrations were interpolated from a recombinant pTau calibration curve. Samples were analyzed in duplicate, with baseline and follow‐up measured in the same assay run. Recombinant protein quality controls were required to quantify within 20% of their established concentration for data acceptability.

2.7. IPMS for N‐term tau

CSF N‐terminal tau concentrations were measured by targeted LC‐MS analysis. Protein A cartridges were crosslinked with an anti‐N‐terminal tau antibody (Ab62, Genentech, Inc.) on the AssayMAP Bravo Platform (Agilent, Santa Clara, CA). Patient CSF samples were randomized, and all available patient timepoints were measured within an assay run. CSF was clarified with perchloric acid and a centrifugation step, denatured, and then digested with lysyl‐endopeptidase overnight. The peptides derived from this step were oxidized, purified on RP‐S cartridges (Agilent, Santa Clara, CA), eluated, and then spiked with a stable isotope peptide standard. The N‐terminal tau peptides were then captured with the Ab‐62 cross‐linked cartridges and eluted with 1% formic acid. The samples were injected into an M5 MicroLC (Sciex, Redwood City, CA) coupled to a 6500 QTRAP mass spectrometer (Sciex) operated in MRM mode. Four transition ions (y14, y15, y13, y9) for the precursor ion (A[+42]EPRQEFEVM[+16]EDHAGTYGLGDRK, 4+) were monitored for quantification. Raw data files were processed with Skyline software (University of Washington, USA), and concentrations were determined by using the normalized peak area of the endogenous peptide to the peak area of the stable isotope peptide standard against 8‐point calibration curves with a range of 0.78–200 fmol/mL. Each 96 well plate of samples included two replicates of a pooled QC CSF lot and calibration curve. At least one QC replicate was required to quantify within 20% of the N‐terminal tau established concentration for data acceptability.

2.8. Statistical analysis

Comparisons between baseline biomarker concentrations in Tauriel and Lauriet were assessed for significance with a Mann–Whitney U test. P‐values for comparisons between treatment groups within each study were calculated using a Student's t‐test. Spearman's rho values were calculated to represent the strength of the correlation between biomarkers and clinical scores at baseline. Unless otherwise noted as nominally significant, p‐values were adjusted for multiple comparisons using the false discovery rate (FDR) method, with the alpha set to 0.05 as a determination of significance. When conducting aggregated analyses of biomarkers across Weeks 49 and 61 in Lauriet, the data were annualized prior to analysis. Statistical assessments were performed with SAS statistical software version 9.4 (SAS Institute), R software version 4.3.3 (R Foundation), and Prism version 8.4.1 (GraphPad Software).

3. RESULTS

3.1. GFAP

Plasma and CSF GFAP levels were quantified in Lauriet and Tauriel participants to determine if semorinemab treatment was associated with changes in glial activity. In each trial, pre‐dose baseline plasma GFAP levels were comparable across the placebo and treatment groups (Figure 1A). In Lauriet, placebo plasma GFAP levels (expressed as the median percent change from pre‐dose baseline levels ± standard error) increased as the study progressed (Week 49: 9.52 ± 2.56%, Week 61: 12.5 ± 5.96%) (Figure 1B). In contrast, plasma GFAP levels stabilized near baseline levels following administration of 4500 mg semorinemab (Week 49: 4.48 ± 2.42%, Week 61: 1.89 ± 4.23%). The difference between the baseline normalized changes in the placebo and 4500 mg semorinemab groups at Weeks 49 and 61 was not significant (Figure 1B). An examination of unadjusted plasma GFAP concentrations was consistent with these results. The median (± standard error) post‐dose concentration of plasma GFAP in each treatment cohort increased in parallel until Week 25, and then separated at Weeks 49 and 61. Plasma GFAP levels stabilized in the 4500 mg semorinemab group while concentrations in the placebo group continued to accumulate for the duration of the study. In contrast to the baseline‐normalized percent change analysis, the difference between plasma GFAP concentrations in the 4500 mg semorinemab group (166 ± 14.0 pg/mL) and placebo group (191 ± 25.4 pg/mL) was nominally significant at Week 61 (p = 0.0428) (Figure 1C).

FIGURE 1.

(A) Individual, group median, and interquartile range of plasma GFAP concentrations at baseline in Tauriel and Lauriet participants across the placebo and semorinemab treatment groups. (B) Plasma GFAP median percent change from baseline (± standard error) at Weeks 5, 25, 49, and 61 in Lauriet M2M participants treated with placebo (gray) or 4500 mg semorinemab (teal). (C) Plasma GFAP median concentration (± standard error) at Weeks 5, 25, 49, and 61 in Lauriet M2M participants treated with placebo (gray) or 4500 mg semorinemab (teal). (D) Plasma GFAP median percent change from baseline (± standard error) at Weeks 5, 33, 49, and 73 in Tauriel P2M participants treated with placebo (gray), 4500 mg semorinemab (orange) and semorinemab (dark blue). (E) Plasma GFAP median concentration (± standard error) at Weeks 5, 33, 49, and 73 in Tauriel P2M participants treated with placebo (gray), 4500 mg semorinemab (orange) and semorinemab (dark blue). The asterisk denotes the significance of a nominal unpaired parametric Student's t‐test without FDR‐adjustment, where *p < 0.05. GFAP, glial fibrillary acidic protein.

Tauriel placebo plasma GFAP levels also trended higher, but not until the end of the study period (Week 73: 12.5 ± 4.11%). In contrast to Lauriet, there was considerable overlap between the placebo and semorinemab treatment groups at Weeks 49 and 73, and no significant differences were detected between the groups irrespective of whether the evaluation was based on percent change (Figure 1D) or absolute concentration (Figure 1E). Tauriel plasma GFAP levels were not quantified from the 1500 mg semorinemab dose group.

Baseline CSF GFAP levels in the placebo and treatment groups in each trial did not differ in a statistically significant manner (Figure S1A). In Tauriel, semorinemab had no effect on CSF GFAP across the dose groups. (Figure S1B). Similarly, when CSF GFAP measurements from Lauriet Cohorts 1 and 2 (Weeks 49 and 61) were annualized and aggregated for analysis, concentrations were not significantly impacted by the administration of 4500 semorinemab (Weeks 49/61: 0.841 ± 2.42%) or placebo (Weeks 49/61: 2.77 ± 5.00%) (Figure S1C).

To determine if there was a relationship between the plasma GFAP and cognition, an analysis of correlation between plasma GFAP concentrations and cognitive scores was performed. Baseline plasma GFAP in Tauriel participants correlated modestly, but significantly, with baseline MMSE (ρ_s = −0.118), RBANS (ρ_s = −0.105), and ADAS‐Cog13 (ρ_s = 0.0568), but not ADAS‐Cog11. However, baseline plasma GFAP did not significantly correlate with any of the baseline cognitive and functional assessments in Lauriet (Table S1). As the 4500 mg dose of semorinemab in Lauriet was associated with a significant slowing in ADAS‐Cog11, data from Lauriet Cohort 1 (Week 49) and Cohort 2 (Week 61) were annualized and then aggregated for an analysis of correlation between post‐dose changes in plasma GFAP concentrations and ADAS‐Cog11 scores. There was a modest correlation between reductions in plasma GFAP and ADAS‐Cog11 scores in the 4500 mg semorinemab group (R_s  = 0.235, p = 0.027) (Figure 2A) that was not present in the placebo group (R_s  = 0.026, n.s) (Figure 2B).

FIGURE 2.

Correlations between the change from baseline at follow‐up of plasma GFAP levels (pg/mL) and change from baseline at follow‐up ADAS‐Cog11 (points) in M2M participants treated with (A) 4500 mg semorinemab or (B) placebo. Linear regression lines fitted to the change in GFAP and ADAS‐Cog11 are displayed in blue. ADAS‐Cog11, Alzheimer's Disease Assessment Scale‐Cognitive Subscale 11; GFAP, glial fibrillary acidic protein.

3.2. YKL‐40

YKL‐40 levels were measured from participant CSF in Tauriel, from participant CSF and plasma in Lauriet, and from healthy volunteers in the Phase I trial. Baseline CSF YKL‐40 levels were statistically indistinguishable between placebo and treatment groups across trials (Figure 3A). In Tauriel, CSF YKL‐40 levels rose significantly after administration of semorinemab, irrespective of dose. The median percent change from baseline increased ∼29.6%–35.2% across dose groups and timepoints. YKL‐40 levels in the placebo group did not increase in this manner and were stable across the study period (∼3.2%–4.4%). (Figure 3B). In Lauriet, CSF data from Cohorts 1 and 2 (Weeks 49 and 61) were annualized and then aggregated as in the CSF GFAP analysis. Consistent with Tauriel, CSF YKL‐40 levels increased significantly from baseline following administration of 4500 mg semorinemab (40.7 ± 7.87%), while placebo had no effect (0.279 ± 2.66%) (Figure 3C). In Lauriet plasma, administration of 4500 mg semorinemab or placebo was not followed by significant changes in YKL‐40 concentrations across the study period (Figure S2). Plasma YKL‐40 concentrations were not analyzed in Tauriel. In Phase I healthy volunteers, a SD of 2400 mg or 8400 mg as well as four weekly doses of 8400 mg semorinemab did not result in a comparable increase in CSF YKL‐40 levels over placebo 8 and 15 days after administration (Figure S3).

FIGURE 3.

(A) Individual, group median, and interquartile range of CSF YKL‐40 concentrations at baseline in Tauriel and Lauriet participants across the placebo and semorinemab treatment groups. (B) Individual, group median, and interquartile range of CSF YKL‐40 percent change from baseline at Weeks 49 and 73 in Tauriel P2M participants treated with placebo (gray), 1500 mg (light blue), 4500 mg (orange), or 8100 mg (dark blue) semorinemab. (C) Individual, group median, and interquartile range of annualized CSF YKL‐40 percent change from baseline at Weeks 49 and 61 in Lauriet M2M participants treated with placebo (gray) or 4500 mg semorinemab (teal). Asterisks denote the significance of an unpaired parametric Student's t‐test with FDR‐adjustment, where ****p ≤ 0.0001, ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. CSF, cerebrospinal fluid; FDR, false discovery rate.

3.3. Plasma pTau181

Plasma pTau181 concentrations were measured from the 4500 mg semorinemab and placebo arms of the DBL treatment periods of Tauriel (Weeks 1–73) or Lauriet (Cohort 1: Weeks 1–49, Cohort 2: Weeks 1–61), and from the 4500‐mg semorinemab Open Label Extension (OLE) period of Lauriet (Cohort 1: Weeks 53–145, Cohort 2: Weeks 65–157). Baseline plasma pTau181 levels showed no statistical difference between placebo and treatment groups across the trials (Figure 4A). Plasma pTau181 levels increased dramatically (>30‐fold over baseline) following administration of semorinemab, starting at Week 5 in both studies. (Figure 4B,C). In Lauriet participants who received 4500 mg semorinemab in the DBL period and then continued into the OLE (23 additional Q4W doses of 4500 mg semorinemab), plasma pTau181 remained elevated for the duration of the study, demonstrating that extended exposure did not translate into reductions of this biomarker in the periphery (Figure 4C). This is consistent with previously published data demonstrating that semorinemab induces a peripheral target engagement response of similar rate and magnitude when plasma total or plasma pTau217 were evaluated. ¹⁰ , ¹¹ In Lauriet participants who received placebo during the DBL period and then transitioned to semorinemab in the OLE, plasma pTau181 concentrations increased to levels comparable to participants in the DBL period (>30‐fold over baseline), and then remained stable for the duration of the OLE (Figure S4).

FIGURE 4.

(A) Individual, group median, and interquartile range of plasma pTau181 concentrations at baseline in Tauriel and Lauriet participants across the placebo and semorinemab treatment groups. (B) Mean fold change in plasma pTau181 concentrations over baseline in placebo (gray) and 4500 mg semorinemab (teal) treatment groups at Weeks 1, 5, 33, 49, and 73 for Tauriel P2M participants and (C) at Weeks 1, 5, 25, 49, 53, 61, 65, 97, 109, 145, and 157 in Lauriet M2M participants. Measurements in the gray‐shaded region are from participants in the OLE. OLE, Open Label Extension.

3.4. Baseline biomarker and clinical intercorrelations

Baseline levels of AD pathophysiology biomarkers in the CSF and plasma from P2M and M2M participants enrolled in the Tauriel and Lauriet studies are shown in Table S2. CSF Aβ42 concentrations were in alignment with a clinical diagnosis of AD and the presence of amyloid pathology. Concentrations were significantly lower in Lauriet participants (588 ± 17.6 pg/mL) than in Tauriel participants (680 ± 15.5 pg/mL) (p < 0.0001). Elevated CSF total Tau and pTau181 concentrations were also consistent with a diagnosis of AD. ¹⁷ CSF and plasma levels of NfL were significantly higher in Lauriet than in Tauriel, consistent with a higher degree of neurodegeneration in the M2M population. Higher levels of CSF CXCL10, plasma GFAP, and CSF IL‐6 were also present in Lauriet and consistent with more glial activity and neuroinflammation in M2M disease. Of the synaptic biomarkers, NPTX2 was significantly lower in M2M AD, consistent with reports that NPTX2 decreases with the severity of AD and predicts AD‐related outcomes. ¹⁸ , ¹⁹ Of the remaining biomarkers, there were no significant or consistent differences between the disease populations.

We conducted an assessment of baseline correlations between plasma GFAP, CSF GFAP, and CSF YKL‐40 and biomarkers shared between each study. Plasma GFAP was moderately correlated with plasma tTau (ρ_s = 0.45), plasma pTau181 (ρ_s = 0.53), and plasma NfL (ρ_s = 0.54), and weakly correlated with Aβ1‐42 (ρ_s = −0.34), lipocalin (ρ_s = −0.21), and IL‐6 (ρ_s = 0.12) (Figure 5). Plasma GFAP did not correlate significantly with CSF GFAP (Figure S5). With the exception of IL‐6 and CXCL10, CSF YKL‐40 correlated moderately with all CSF biomarkers examined (Figure 5). As anticipated, the CSF tau species were strongly intercorrelated (ρ_s = 0.84–0.97), as were plasma tTau and plasma pTau181 (ρ_s = 0.70) (Figure S5).

FIGURE 5.

Heatmap of correlations between baseline plasma GFAP, CSF GFAP, and CSF YKL‐40 concentrations and all other CSF and plasma biomarkers jointly evaluated in Tauriel and Lauriet. All biomarkers are measured in CSF unless specified otherwise. Spearman's rho values are displayed and colored circles represent correlations where the FDR‐adjusted p‐value < 0.05. CSF, cerebrospinal fluid; FDR, false discovery rate; GFAP, glial fibrillary acidic protein.

In addition to biomarker intercorrelations, we also evaluated correlations between these analytes and cognitive/functional scores common to both studies in aggregate (ADAS‐Cog 11, CDR‐SB, MMSE, ADCS‐ADL). At baseline, a strong intercorrelation was observed between the shared cognitive and functional scores (ρ_s = −0.75 to 0.51). Modest correlations were also observed between the cognitive/function metrics and biomarkers representative of a variety of pathophysiological features. Weak correlations were noted with plasma GFAP (ρ_s = −0.39 to 0.29), IL‐6 (ρ_s = −0.30 to 0.25), lipocalin (ρ_s = −0.32 to 0.34), plasma NfL (ρ_s = −0.28 to 0.22), plasma pTau181 (ρ_s = −0.17 to 0.18), plasma tTau (ρ_s = −0.18), S100B (ρ_s = −0.08 to 0.18), and Aβ1‐42 (ρ_s = −0.22 to 0.37). (Figure S5).

3.5. Additional biomarkers

As outlined in Table 2, biomarkers related to neuroinflammation (Figure S6), amyloidosis (Figure S7), neurodegeneration (Figure S8), synucleinopathy, and synaptic function (Figure S9) were measured in Tauriel and Lauriet CSF and plasma. In both studies, these biomarkers did not exhibit any effects from semorinemab treatment that reached statistical significance.

4. DISCUSSION

The administration of semorinemab to P2M and M2M participants in the Tauriel and Lauriet Phase II trials demonstrated that while target engagement could be detected in the periphery and CNS of study participants, ¹⁰ , ¹¹ there were divergent clinical outcomes.

Tau biomarkers in the CSF, including total Tau and phosphorylated Tau (i.e., pTau181 & pTau217) are correlates of amyloid plaque and neurofibrillary tangle burden in the brain with demonstrated diagnostic and prognostic utility for AD. ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ Previously, we published data demonstrating that semorinemab induces a peripheral target engagement response of a similar rate and magnitude when plasma total and plasma pTau217 were evaluated, and this was similar to what we observed with pTau181. ¹⁰ , ¹¹ In agreement with our findings, a maximum increase of ∼25× in plasma Tau after treatment with tilavonemab, another N‐terminal anti‐Tau, has been reported. ²⁷ Subsequent to the preliminary response, plasma pTau181 levels remained elevated and did not differentiate between the P2M and M2M populations in each study. Furthermore, prolonged exposure to semorinemab did not lead to a reduction in plasma pTau181 levels in participants who elected to continue in the OLE.

This response is consistent with previously reported data detailing the kinetics and magnitude of the PD response of plasma total Tau and plasma pTau217 to semorinemab in Tauriel and Lauriet, but are in contrast to the reductions observed in CSF Tau species. ¹⁰ , ¹¹ An aggregated analysis of the post‐dose concentration changes in CSF and plasma Tau biomarkers measured across both studies demonstrated robust and significant correlations in the PD response within the CSF Tau species (ρ_s = 0.56–0.97) and plasma Tau species (ρ_s = 0.53), but not between them (Figure S10). These differences likely reflect distinct antibody‐tau kinetics in response to diverging clearance mechanisms of Tau proteoforms in the CNS and periphery. ²⁸ , ²⁹ While CSF Tau changes were reported by other anti‐Tau antibodies, they are unfortunately not directly comparable to our data since unbound Tau was reported by those studies while ours report total Tau. ²⁷ , ³⁰

In a post‐hoc examination of AD pathophysiology biomarkers in the plasma and CSF of trial participants, we observed noteworthy changes in glial activity biomarkers. YKL‐40 increased significantly in the CSF post‐dose in both studies, suggesting that the administration of semorinemab stimulated an increase in microglial activity in the CNS. The magnitude of this increase is similar to responses observed with AL002, a TREM2 agonist antibody that directly activates microglia, in a Phase I healthy volunteer study. ³¹ In the case of other drugs targeting AD pathology, no increases were reported in Crenezumab's CREAD studies, ³² and a relatively smaller increase of ∼6% was reported in Gantenerumab's GRADUATE studies. ³³ YKL‐40 is expressed in microglia, astrocytes, and peripheral immune cells, and is elevated in the CSF of AD patients. ³⁴ At the transcriptomic level, YKL‐40 expression is increased in human AD brains, but not with subjects carrying the loss of function R62H TREM2 variant which reduces microglia activation. ³⁵ Additional evidence of a semorinemab effect on microglia activity comes from proteomic analyses of Tauriel and Lauriet CSF, where the post‐treatment proteomic signatures indicated that the proteomic response in Lauriet was likely of microglial origin. This was noticeably distinct from the response signature observed in Tauriel, which suggested a more ubiquitous expression across neural cells. In addition, semorinemab‐dependent increases in GPNMB, a marker of activated microglia, were observed in Lauriet, but not Tauriel. ³⁶ , ³⁷ Together with the CSF YKL‐40 observations, these findings were unexpected because semorinemab was developed on an IgG4 backbone, which has been demonstrated to have lower binding affinity to fragment crystallizable gamma receptor (FcR) and expected to have reduced immune effector function. ¹² However, it is possible that the binding of semorinemab to tau aggregates could drive higher local concentrations of the antibody, cross‐link FcRs through an avidity effect, and trigger microglia activation. Data from the Phase I study of semorinemab in healthy volunteers support this hypothesis. Although the dosing regimen and follow‐up periods were shorter than Tauriel and Lauriet, ¹² single and multiple doses of up to 8400 mg semorinemab had no clear impact on CSF YKL‐40 concentrations relative to placebo in cognitively unimpaired participants expected to lack tau pathology (Figure S3).

Increases in plasma GFAP observed throughout the study periods were stabilized in response to 4500 mg semorinemab in the Lauriet trial, while no remarkable impact of treatment was observed between the dose groups in the Tauriel study. The response observed in Lauriet is in contrast to CSF GFAP, where no notable differences were observed between the placebo and treatment groups in either study. Plasma GFAP is an emerging biomarker for AD that is considered reflective of reactive astrocytes in a variety of neurodegenerative conditions. ³⁸ , ³⁹ It decreases in response to anti‐amyloid therapy, is associated with plaque removal, ³³ , ⁴⁰ , ⁴¹ and across our Ph2 studies, baseline levels demonstrate a degree of correlation with metrics of cognition and function. The robust CSF YKL‐40 response and the unique enrichment of microglia proteins observed in an analysis of the Lauriet CSF proteome ³⁶ suggest an intriguing possibility that the stabilization of plasma GFAP in Lauriet may be a response to microglia activity stimulated by semorinemab. Since semorinemab has no discernable effect on Tau aggregation, our post‐hoc discovery of microglia activation and variations in plasma GFAP between the trials presents a scientific hypothesis that this biology may account for the discrepancies observed in the trial outcomes. However, additional mechanistic studies need to be conducted to elucidate this hypothesis and analysis of future trials with semorinemab are needed to confirm the robustness of these findings. One limitation of our studies is the lack of diversity in our sample population. Future research should include adequate representation across demographic groups to ensure that insights gained from these investigations are broadly applicable.

AUTHOR CONTRIBUTIONS

Edmond Teng; Cecilia Monteiro; and Kristin R. Wildsmith: Study design. Anna Bayfield; Gwendlyn Kollmorgen; Julie Lee; Jenny Jiang; and Norbert Wild: Sample and data acquisition. Balazs Toth and Stephen P. Schauer: Data analysis. Balazs Toth; Cecilia Monteiro; Edmond Teng; Felix L. Yeh; Kristin R. Wildsmith; Lee A. Honigberg; Veronica Anania; and Stephen P. Schauer: Data interpretation. Felix L. Yeh and Stephen P. Schauer: Manuscript writing—original draft. All authors: Manuscript writing—review and editing.

CONFLICT OF INTEREST STATEMENT

A.B., G.K., and N.W. are full‐time employees of Roche Diagnostics GmbH, Penzberg, Germany. All other authors are full‐time employees of Genentech Inc. (member of the Roche group). Edmond Teng is listed as a co‐inventor on the patent for semorinemab. Author disclosures are available in the Supporting information.

CONSENT STATEMENT

All participants in the present study and/or their legally authorized representatives provided signed informed consent.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Schauer S, Toth B, Lee J, et al. Pharmacodynamic effects of semorinemab on plasma and CSF biomarkers of Alzheimer's disease pathophysiology. Alzheimer's Dement. 2024;20:8855–8866. 10.1002/alz.14346