Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by brain atrophy, cognitive decline, and memory loss1,2. The pathological changes of AD involve the accumulation of amyloid-beta (Aβ) plaques and neurofibrillary tangles, synaptic dysfunction, and neuronal loss3,4. In addition to these classic pathological changes, impaired neurogenesis has been reported to be implicated in the pathogenesis of AD, but not without controversy5,6,7. Neurogenesis is regulated by a variety of factors, such as transcription factors and neurotrophic factors6,8. Neurogenin-1, a member of the neurogenin family of transcription factors, has been thought to play a crucial role in neuronal development, differentiation, and synaptic plasticity9,10. Therefore, understanding the interplay between neurogenin-1 and cognitive decline and neurodegeneration could provide valuable insights for the development of novel therapeutic strategies aimed at preserving cognitive function in individuals affected by this debilitating condition. To the best of our knowledge, however, the association of CSF neurogenin-1 levels with cognitive decline and changes in structural magnetic resonance imaging (MRI) features in older adults remains unknown.

The primary objective of this study was to investigate the relationship between CSF neurogenin-1 levels and changes in cognitive performance over time in older adults, with or without cognitive impairment. Additionally, this study explored the association between CSF neurogenin-1 levels and changes in brain volumes, as measured by MRI, specifically focusing on hippocampal and ventricular volumes, which are known to be affected in AD. Given its potential neuroprotective effect, we hypothesized that higher neurogenin-1 levels would be associated with a slower rate of cognitive decline and neurodegeneration. By examining these relationships, we sought to gain insights into the potential role of neurogenin-1as a biomarker for tracking cognitive and neurodegenerative changes and as a novel target for therapeutic interventions. Understanding the interplay between neurogenin-1 levels, cognitive decline, and structural brain changes may provide valuable information for the development of treatment strategies aimed at slowing the progression of AD and improving patient outcomes.

Results

Baseline comparison of demographic and clinical variables by cognitive status

At baseline, there were a total of 666 participants, comprising 161 CU participants and 505 CI participants. The comparison of demographic and clinical variables between cognitive status (CU vs. CI) is summarized in Table 1. CU participants were older than CI participants, while there was no difference in years of education or percentage of female participants. As expected, there were significant differences in the distribution of APOE4 carriers, MMSE, CDR-SB, ADAS-Cog-13, RAVLT total score, aHV, aVV, and the percentage of amyloid positivity. However, no significant difference in CSF neurogenin-1 levels was found between the CU and CI participants.

Table 1 Sample characteristics by cognitive status.
Full size table

Association of baseline CSF neurogenin-1 levels with changes in cognitive measures over time

Linear mixed-effects models were performed to examine the association of baseline CSF neurogenin-1 levels with changes in cognitive measures over time. To explore whether cognitive status might modify this relationship, separate models were conducted for the CU and CI groups. In the models where MMSE was treated as the primary outcome, we found that CSF neurogenin-1 levels were not associated with changes in MMSE in the CU participants (coefficient = 0.008; 95% CIs =  − 0.012 to 0.027; p value = 0.439; see Table 2 and Fig. 1A), while a significant relationship was observed in the CI participants (coefficient = 0.045; 95% CIs = 0.004 to 0.085; p value = 0.03; see Table 2 and Fig. 1B). This finding suggested that higher baseline CSF neurogenin-1 levels were associated with a slower cognitive decline, as measured by MMSE, in the CI participants.

Table 2 Summary of the linear mixed-effect model with MMSE as the outcome.
Full size table
Fig. 1
figure 1

Association of CSF Neurogenin-1 levels with changes with MMSE over time. (A) Shows the association in the CU group. (B) Shows the association in the CI group.

Full size image

Several additional analyses were conducted using other cognitive measures, such as CDR-SB, ADAS-Cog-13, and RAVLT total score, as secondary outcomes in the linear mixed-effect models. In the models with CDR-SB and ADAS-Cog-13, the results from the analyses using MMSE as the outcome were generally replicated (see Table 3). Briefly, in the models where CDR-SB was treated as the outcome, we found that CSF neurogenin-1 levels were not associated with changes in CDR-SB in the CU participants (coefficient =  − 0.008; 95%CIs =  − 0.022 to 0.005; p value = 0.206), while a significant relationship was observed in the CI participants (coefficient =  − 0.036; 95%CIs =  − 0.063 to -0.008; p value = 0.011). In the models where ADAS-Cog-13 was treated as the outcome, we found that CSF neurogenin-1 levels were not associated with changes in ADAS-Cog-13 in the CU participants (coefficient =  − 0.031; 95%CIs =  − 0.085 to 0.023; p value = 0.255), while a significant relationship was observed in the CI participants (coefficient =  − 0.088; 95%CIs =  − 0.172 to -0.005; p value = 0.039). However, in the models where RAVLT total score was treated as the outcome, there was no significant relationship between CSF neurogenin-1 levels and changes in RAVLT total score over time either in the CU (coefficient = 0.044; 95%CIs =  − 0.036 to 0.124; p value = 0.281) or CI (coefficient = 0.053; 95%CIs =  − 0.005 to 0.112; p value = 0.075) group.

Table 3 Summary of linear mixed-effect models with CDR-SB, ADAS-Cog-13, and RAVLT total score as the outcomes.
Full size table

Association of baseline CSF neurogenin-1 levels with changes in structural MRI features over time

To examine the relationship between CSF neurogenin-1 levels with changes in aHV and aVV over time, several linear mixed-effect models were performed in the CU and CI groups separately. In the models where aHV was treated as the outcome, there was no significant relationship between CSF neurogenin-1 levels and changes in aHV over time either in the CU (coefficient =  − 0.002; 95%CIs =  − 0.006 to 0.001; p value = 0.203) or CI (coefficient =  − 0.002; 95%CIs =  − 0.004 to 0.001; p value = 0.237) group (see Table 4 and Fig. 2 A, B). In the models where aVV was treated as the outcome, we found that CSF neurogenin-1 levels were not associated with changes in aVV in the CU participants (coefficient =  − 0.030; 95%CIs =  − 0.072 to 0.011; p value = 0.154), while a significant relationship was observed in the CI participants (coefficient =  − 0.062; 95%CIs =  − 0.104 to -0.020; p value = 0.004; see Table 4 and Fig. 2C, D).

Table 4 Summary of linear mixed-effect models with MRI features as the outcomes.
Full size table
Fig. 2
figure 2

Association of CSF neurogenin-1 levels with changes in structural MRI features over time in CU and CI participants. (A) Shows the relationship between CSF neurogenin-1 levels and changes in aHV in the CU participants. (B) Shows the relationship between CSF neurogenin-1 levels and changes in aHV in the CI participants. (C) Shows the relationship between CSF neurogenin-1 levels and changes in aVV in the CU participants. (D) Shows the relationship between CSF neurogenin-1 levels and changes in aVV in the CI participants. CU cognitively unimpaired, CI cognitively impaired, aHV adjusted hippocampal volume, aVV adjusted ventricular volume.

Full size image

Supplementary analysis

To examine whether the associations differed between individuals with MCI and AD, six additional linear mixed-effects models were fitted within the CI group, using MMSE, CDR-SB, ADAS-Cog, RAVLT, aHV, and aVV as outcome variables. The predictor of interest was the three-way interaction between CSF neurogenin-1 levels, cognitive status (MCI vs. AD), and time. However, none of the three-way interaction terms were statistical significance (all p > 0.05), suggesting that the longitudinal associations did not significantly differ between MCI and AD participants.

We further examined the association between CSF neurogenin-1 and p-tau181 levels in the CU and CI groups. Positive correlations were observed in both the CU (r = 0.21, p = 0.006; Fig. 3A) and CI (r = 0.21, p < 0.001; Fig. 3B) groups.

Fig. 3
figure 3

Relationship between CSF p-tau and neurogenin-1 levels in the CU and CI groups. CU cognitively unimpaired, CI cognitively impaired.

Full size image

Discussion

This study has several major findings. CSF neurogenin-1 levels at baseline were associated with changes in global cognitive and functional measures, such as MMSE, CDR-SB, and ADAS-Cog-13, in the CI individuals, but not in the CU individuals. Specifically, higher CSF neurogenin-1 levels were associated with a slower rate of cognitive decline over time in the CI individuals. However, we did not find a relationship between CSF neurogenin-1 levels and memory decline, as measured by RAVLT total score, in either the CU or CI individuals. For structural MRI features, we found that higher CSF neurogenin-1 levels at baseline were associated with a slower rate of ventricular enlargement in the CI individuals, but not in the CU individuals. No association between CSF neurogenin-1 levels and hippocampal atrophy was observed in either the CU or CI individuals. Our findings may provide useful information for the future development of treatment strategies aimed at slowing the progression of AD and improving patient outcomes.

To our knowledge, this is the first study to report that higher baseline CSF neurogenin-1 levels were associated with a slower rate of cognitive decline in CI individuals. This association was observed across various cognitive and functional measures, including the MMSE, CDR-SB, and ADAS-Cog-13, which are widely used tools for assessing cognitive function in AD11,12,13. The lack of this association in CU individuals suggested that neurogenin-1 may exert a protective effect in the context of existing cognitive impairment, potentially by promoting neuronal survival or enhancing synaptic plasticity9,10. The absence of a relationship between CSF neurogenin-1 levels and memory decline, as measured by the RAVLT total score, may indicate that the effects of neurogenin-1 on cognitive performance are not memory-specific. It is possible that neurogenin-1’s influence on memory, specifically episodic memory assessed by the RAVLT, is less pronounced. Further research is needed to explore the specific cognitive processes influenced by neurogenin-1 and to identify the underlying neural substrates.

The association between higher CSF neurogenin-1 levels and a slower rate of ventricular enlargement in CI individuals is another key finding of this study. Ventricular enlargement is a well-established marker of brain atrophy in AD and is associated with the progression of the disease14. The observed association suggested that neurogenin-1 may play a role in modulating the neurodegenerative processes that lead to brain atrophy. The lack of an association between neurogenin-1 levels and hippocampal atrophy is consistent with the absence of an association between CSF neurogenin-1 levels and memory decline. While hippocampal atrophy is a hallmark of AD, our findings suggested that neurogenin-1 may be more closely related to global structural changes, such as ventricular volumes. Ventricular enlargement may reflect broader neurodegenerative processes, which could be modulated by neurogenin-1 through indirect mechanisms, such as regulating inflammation or promoting compensatory plasticity. In contrast, hippocampal atrophy may represent a more localized pathological process that is less responsive to neurogenin-1–mediated impact. In addition, the adult hippocampus may demonstrate limited neurogenin-1 expression or activity, thus decreasing its effect on this specific structure.

Exact mechanisms of the potential neuroprotective effects of neurogenin-1 observed in this study remain unclear. There are several possible explanations. Neurogenin-1 is known to regulate neuronal survival, differentiation, and synaptic plasticity9. It is possible that higher neurogenin-1 levels promote the survival of vulnerable neurons, enhance synaptic connectivity, or modulate neuroinflammatory processes, thereby slowing cognitive decline and structural brain changes. In addition, it is also possible that the positive correlation between CSF neurogenin-1 and p-tau181 levels may reflect an adaptive response to tau pathology, potentially suggesting a compensatory mechanism aimed at mitigating neurodegeneration (Fig. 3). Further research is needed to elucidate the precise mechanisms by which neurogenin-1 influences cognitive and brain health in AD. The findings also have important implications for the development of novel therapeutic strategies for AD. Targeting neurogenin-1 or its signaling pathways could represent a promising approach to slow disease progression and improve patient outcomes. Future studies should explore the potential of neurogenin-1-based interventions, such as pharmacological modulation or gene therapy, to enhance the neuroprotective effects of neurogenin-1 in AD.

This study has several limitations. First, the observational nature of the study design limits our ability to establish causality between neurogenin-1 levels and cognitive or structural changes. Interventional studies with larger sample sizes and diverse populations are needed to confirm these findings and to explore the dynamic interplay between neurogenin-1, cognitive decline, and brain atrophy over time. Likewise, this study cannot fully rule out the possibility of reverse causation, and further interventional studies would be needed to determine the directionality of the relationship. Second, the specific cognitive processes influenced by neurogenin-1 and the underlying neural substrates require further investigation. Third, several unobserved variables or comorbidities could potentially affect the current findings. These latent variables could independently influence both cognitive decline and changes in MRI markers, thereby introducing potential bias in effect estimates. Further studies are needed to address this issue. Fourth, we assumed the outcome data to be missing at random (MAR). While this is a reasonable assumption given the lack of scientific evidence against it, caution should be exercised when interpreting the findings of the current study. Fifth, the presence of statistical significance does not guarantee clinical relevance. For example, in the CI model with MMSE as the outcome variable, the coefficient of the CSF neurogenin-1 × Time interaction term was 0.045 (Table 2). This suggests that, compared to a participant with a neurogenin-1 level of 20 pg/mL, a participant with a level of 30 pg/mL exhibited a slower annual decline by 0.45 MMSE points. However, this study offers a first step in exploring the potential role of neurogenin-1 as a novel target for intervention. Sixth, the findings of the current study should be validated in future studies. Particularly, more commonly used immunoassay methods, rather than the SomaScan platform, should be used to validate the findings.

In conclusion, our findings suggest that neurogenin-1 may play a neuroprotective role in CI individuals, potentially slowing cognitive decline and structural brain changes associated with the disease. Further research is needed to fully understand the mechanisms by which neurogenin-1 influences cognitive and brain health in AD and to explore its potential as a therapeutic target.

Methods

Alzheimer’s Disease Neuroimaging Initiative (ADNI) database

The ADNI is a longitudinal, observational, multicenter study that aims to test whether clinical, neuropsychological, neuroimaging, and other biological markers can be integrated to track disease progression on the AD spectrum. The longitudinal dataset used in the current study was obtained from the ADNI database (adni.loni.usc.edu). ADNI participants are aged between 55 and 99 years and are assigned to the diagnostic classifications of normal cognition, mild cognitive impairment (MCI), and mild AD dementia. The ADNI study was approved by the institutional review board in each of the over 60 ADNI recruitment centers (names of all participating centers can be found at: https://adni.loni.usc.edu/wp-content/uploads/ADNI%20Acknowledgement%20List_Feb2025_clean.pdf), and all ADNI participants provided written informed consent. All methods at each participating center were performed in accordance with the relevant guidelines and regulations (https://adni.loni.usc.edu/wp-content/uploads/2024/02/ADNI_General_Procedures_Manual.pdf). In addition, as this study utilized fully de-identified data, the Institutional Review Board of Wenzhou Seventh People’s Hospital concluded that it was exempt from submission for ethical review.

Participants

We selected ADNI participants who had a baseline assessment of the Mini-Mental State Examination (MMSE)11 and at least one follow-up assessment, and had baseline CSF neurogenin-1 levels available. In the current study, there were a total of 666 participants, comprising 161 cognitively unimpaired (CU) participants and 505 cognitively impaired (CI; comprising MCI and mild AD dementia) participants. The diagnostic criteria have been described elsewhere15 and can be found at the ADNI website (https://adni.loni.usc.edu/wp-content/themes/freshnews-dev-v2/documents/clinical/ADNI-1_Protocol.pdf). In brief, for the CU group, the criteria included a score of 24 or above on the MMSE and a score of 0 on the Clinical Dementia Rating (CDR)16. For the MCI group, the criteria included an MMSE score of 24 or higher, a CDR of 0.5, presence of a subjective memory complaint, and objective evidence of memory impairment as measured by the Wechsler Memory Scale Logical Memory II, along with the maintenance of daily life activities. In the case of mild AD dementia, the criteria included an MMSE score ranging from 20 to 26, a CDR score of 0.5 or 1, and fulfillment of the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association criteria for probable AD17. In the ADNI study, ‘baseline’ refers to the initial visit during which the first cognitive assessments and MRI scans are conducted. After the baseline (year 0), participants are typically reassessed at 0.5, 1, 2 years, and beyond. At baseline, the time interval between cognitive assessments and MRI scans or CSF measurements is required to be within 2 weeks.

Cognitive measures

In the current study, the primary cognitive outcome examined was the MMSE score. Additionally, we assessed three secondary cognitive measures: the Clinical Dementia Rating Sum of Boxes (CDR-SB)12, the Alzheimer’s Disease Assessment Scale Cognitive Subscale 13-item version (ADAS-Cog-13)13, and the Rey Auditory Verbal Learning Test (RAVLT) total score18. The MMSE Score ranges from 0 to 30, with higher scores indicating better cognitive abilities. CDR-SB scores range from 0 to 18, where 0 signifies no impairment and 18 represents severe impairment. ADAS-Cog-13 scores span from 0 to 85, with higher scores indicating more significant cognitive impairment. The RAVLT total score varies from 0 to 75, with 0 representing no words recalled and 75 indicating perfect recall of all words.

Structural MRI features

Structural MRI scans were collected and processed according to a standardized procedure that was validated across centers19. Structural MRI variables used in the current study were extracted from the file “ADNIMERGE.csv”. We focused on two MRI features, including hippocampal and ventricular volumes, due to their early involvement on the AD spectrum2,14. To adjust for the sex differences in head size, adjusted volumes were used rather than raw volumes. Adjusted hippocampal volume (aHV) was calculated by dividing hippocampal volumes by intracranial volume (ICV) and multiplying the result by 1000. Similarly, adjusted ventricular volume (aVV) was calculated by dividing ventricles by ICV and multiplying the result by 1000.

CSF neurogenin-1 levels

The levels of CSF neurogenin-1 were analyzed as part of proteomic assessments using SomaLogic’s SomaScan platform by the Neurogenomics and Informatics Center at Washington University. SomaLogic undertook initial standardization procedures for quantifying the proteins20. Specifically, hybridization normalization was carried out for each sample individually. Subsequently, aptamers were grouped into three distinct normalization cohorts—S1, S2, and S3—based on the signal-to-noise ratio observed in both technical replicates and samples. This classification was essential to prevent the merging of aptamers with varying protein signal intensities for further normalization stages. After this categorization, a median-based normalization approach was employed to address diverse assay-related inconsistencies, such as fluctuations in protein concentration, pipetting, reagent concentration, and assay timing. The CSF neurogenin-1 levels are presented in relative fluorescence units (RFU).

Statistics

Baseline comparisons of demographic and clinical variables between cognitive status (CU vs. CI) were conducted using two-sample t tests for continuous variables and Pearson’s x2 tests for categorical variables. To address our primary research question—whether baseline CSF neurogenin-1 levels are associated with changes in MMSE over time—a linear mixed-effects model with baseline CSF neurogenin-1 levels as the predictor of interest and MMSE scores as the dependent variable was performed. The model included main effects of age, gender, education level, APOE4 status, amyloid status, CSF neurogenin-1, and their interactions with follow-up time (in years). The model included a random intercept and a random slope for each study participant. Separate models were conducted for the CU and CI participants. The rationale for our stratified modelling approach lies in the hypothesis that CU and CI participants may undergo distinct neurobiological processes, leading to heterogeneous patterns of variability (e.g., differential random intercepts or slopes). In contrast, interaction models typically assume shared random effects across groups. To further validate our findings from the MMSE model, we conducted several secondary analyses where CDR-SB, ADAS-Cog-13, and RAVLT total score were treated as our cognitive outcomes. To study the question of whether baseline CSF neurogenin-1 levels are associated with changes in structural MRI features over time, several linear mixed-effect models with baseline CSF neurogenin-1 levels as the predictor of interest and aHV and aVV as the dependent variables were performed. Several core assumptions were tested, including the normality of residuals and random effects. Most of the tested models did not substantially violate these assumptions. Encouragingly, fixed-effects and variance estimates for linear mixed-effects models appear to be largely robust to violations of distributional assumption21. All statistical work was conducted using R software (version 4.3.3). The significant level was set at p < 0.05.