The hallmark brain changes underlying Alzheimer’s disease (AD), the accumulation of amyloid plaques and tau neurofibrillary tangles, are known to begin years before the onset of clinical symptoms (Jack et al., 2024). For example, abnormal levels of cerebrospinal fluid (CSF) biomarkers of amyloid beta and tau among cognitively unimpaired individuals are associated with increased risk of developing mild cognitive impairment (MCI) and AD dementia many years later (Greenberg et al., 2022, Moghekar et al., 2013, Ossenkoppele et al., 2022, Vos et al., 2013). It has been known for decades that synaptic alterations and loss occur in the early symptomatic phase of AD and are recognized as a main contributor to cognitive impairment (Masliah et al., 2001, Scheff et al., 2006, Terry et al., 1991). Evidence from animal models also indicates that pathological amyloid and tau can promote synaptic dysfunction and loss in early disease stages, which contribute to cognitive decline (Martínez-Serra et al., 2022, Wu et al., 2021).

Considering the importance of synapses in the evolution of AD-related cognitive impairment, there has been a significant focus on identifying synaptic markers in CSF that change during the symptomatic phases of AD (for a review, see (Camporesi et al., 2020)). Relatively little, however, is known about synaptic alterations among cognitively unimpaired individuals and how these may contribute to cognitive decline and future clinical progression. This is an important topic for research because the preservation of synapses or synaptic signaling may be an important cellular mechanism of resilience to AD (de Vries et al., 2024). The current study therefore examined how CSF levels of three synaptic proteins, previously shown to change during the symptomatic phases of AD, relate to long-term cognitive trajectories among individuals with unimpaired cognition at baseline, both alone and in combination with CSF AD biomarker levels. These synaptic proteins included neuronal pentraxin 2 (NPTX2), AMPA-type glutamate receptor subunit A4 (GluA4), and neurosecretory protein VGF (VGF).

NPTX2 is a secreted protein that is expressed by excitatory neurons in response to behavior-related neural activity (Tsui et al., 1996). It is secreted as a presynaptic factor at excitatory glutamatergic synapses where it functions to recruit postsynaptic AMPA receptors, particularly GluA4 (Chang et al., 2010, Pelkey et al., 2015). NPTX2 thereby strengthens and stabilizes excitatory drive of inhibition, supporting synaptic plasticity, and helping to maintain the homeostatic balance of excitation and inhibition in cortical networks (Chang et al., 2010, Chapman et al., 2019, Gómez de San José et al., 2022, Pelkey et al., 2015, Xiao et al., 2017). Cross-sectional studies have reported that CSF levels of NPTX2 are lower in individuals with AD dementia and mild cognitive impairment (MCI) compared to those with unimpaired cognition (Galasko et al., 2019, Nilsson et al., 2021, Soldan et al., 2019, Watson et al., 2023, Xiao et al., 2017). Longitudinal declines in NPTX2 are associated with declines in cognitive performance among dementia-free individuals (i.e., combining cognitively unimpaired and MCI participants; (Libiger et al., 2021)). Moreover, lower CSF levels of NPTX2 among cognitively unimpaired, middle-aged and older participants have been associated with an increased risk of progressing to MCI, independent of CSF AD biomarker levels (Soldan et al., 2023), as well as an increased risk of progression from MCI to AD dementia (Galasko et al., 2019, Llano et al., 2023, Spellman et al., 2015).

Less is known about the relationship of GluA4 to cognitive decline during the symptomatic phase of AD. Like NPTX2, GluA4 is required for maintaining circuit homeostasis (Fuchs et al., 2007, Pelkey et al., 2015). GluA4 expression is dependent on NPTX2 in mouse models (Pelkey et al., 2015, Xiao et al., 2017), and therefore GluA4 expression may provide an indication of NPTX2 function. Levels of GluA4 are lower in post-mortem tissue among individuals with dementia due to AD compared to cognitively unimpaired older adults (Canal-Garcia et al., 2025, Xiao et al., 2017). Additionally, GluA4 levels and NPTX2 levels are positively correlated in both postmortem cortical tissue (Xiao et al., 2017) and in CSF (Belbin et al., 2020), consistent with the view that GluA4 expression is regulated by NPTX2 (Xiao et al., 2017). Although the relationship of GluA4 to cognitive performance and decline in humans is largely unknown, animal models suggest that low GluA4 levels are reflective of inhibitory circuit dysfunction (Pelkey et al., 2015, Xiao et al., 2017) and that both NPTX2 and GluA4 are important for maintaining the balance of excitation and inhibition in cortical circuits (Fuchs et al., 2007, Pelkey et al., 2015, Xiao et al., 2017).

Neurosecretory protein VGF (non-acronymic, also known as secretogranin VII) is a secreted protein and neuropeptide precursor that supports synaptogenesis, neurogenesis, learning and memory, and energy metabolism (for reviews, see (Colín-Martínez and Arias, 2025; Quinn et al., 2021)). VGF is widely expressed in the central and peripheral nervous system, particularly the hippocampus and hypothalamus, and is also recognized as a potential therapeutic target for AD (Beckmann et al., 2020). Previous studies have reported decreased VGF in brain tissue (Askenazi et al., 2023, Bai et al., 2020, Beckmann et al., 2020) and CSF (Bai et al., 2020, Duits et al., 2018, Hölttä et al., 2015, Khoonsari et al., 2019, Watson et al., 2023) from AD dementia patients compared to controls. Lower VGF levels in brain tissue in individuals with AD dementia at death were also associated with greater cognitive decline prior to death, independent of AD neuropathological (Aβ and tau) burden (Wingo et al., 2019). Additionally, low CSF VGF levels were predictive of progression from MCI to dementia when accounting for CSF AD biomarker levels (Llano et al., 2019, Llano et al., 2023).

Taken together, prior research suggests that NPTX2, GluA4, and VGF play essential roles in synaptic processes and neuronal signaling, and their decline appears to track the progression of cognitive impairment in AD. It is unclear, however, whether these measures are associated with long-term cognitive decline among individuals who are initially cognitively unimpaired and whether this association differs as a function of baseline AD biomarker levels and across cognitive domains. To address these gaps, the present study examined the association of baseline CSF levels of NPTX2, GluA4, and VGF with longitudinal cognitive performance among 269 middle-aged and older adults who were cognitively unimpaired at baseline and have undergone on average 16 years of cognitive follow-up. Cognitive performance was evaluated using both a global cognitive composite score and domain-specific scores for episodic memory and executive functions.