Resting-state quantitative EEG (qEEG) is a good candidate marker for intervention clinical trials in Alzheimer’s disease (AD) since it has shown good discriminatory power between AD patients and healthy aging in several studies (Brenner et al., 1988; Lehmann et al., 2007; Jelic and Kowalski, 2009; Babiloni et al., 2021). These studies report on an increase in the slow frequency activities including early changes in theta range and the later stage of the disease alterations in the delta range, in parallel with a decrease in the faster alpha and beta frequencies.

Already at the very beginning of acetylcholine esterase inhibitor (ChEI) treatment for AD, it was reported that chronic treatment with cholinergic drugs induces a specific pattern of electrical brain activity corresponding to both the treatment response and early decline in treatment efficacy (Jelic et al., 1998).

It has been shown repeatedly that the ChEI treatment increases the physiological alpha rhythm oscillations in EEG, particularly in the posterior regions (Jelic et al., 1998; Moretti, 2014). Furthermore, these positive effects on the resting EEG spectral alpha power induced by ChEIs, correlated well with improved cognitive function as measured by significant changes in Mini-mental state examination (MMSE) scores (Moretti, 2014). Recently an EEG-based acetylcholine (ACh) index has been developed using data from a scopolamine challenge study showing a reduced ACh index even in prodromal AD (Johannsson et al., 2015). Interestingly, even cholinergic stimulation of healthy subjects with the ChEI galantamine showed an effect on the alpha power and working memory performance (Eckart et al., 2016).

The symptomatic cholinergic treatment approach to AD is based on findings that Nucleus Basalis of Meynert (NBM) is a major source of cholinergic projections to the cortex (Mesulam and Geula, 1988) and reductions in levels of the cholinergic markers acetylcholinesterase (AChE) and choline acetyltransferase (ChAT) reflects the loss of cholinergic innervation that have been reported in the AD brain (Giacobini, 2003). Traditionally, ChAT activity is used as a marker of cholinergic neuronal loss in lesion studies (Wenk et al., 1994; Rossner et al., 1995). It has been shown that NBM lesions in animal models decrease cortical ChAT activity and modulate event-related oscillations in EEG by increasing power in low and decreasing power in high frequencies (Sanchez-Alavez et al., 2014).

Nerve growth factor (NGF) has emerged as a potential therapeutic agent in Alzheimer’s disease (AD) due to its regenerative effects on basal forebrain cholinergic neurons (Williams et al., 2006; Mufson et al., 2008). Exogenous NGF delivered to the cholinergic basal forebrain has shown regenerative effects correlating with improved cognition in animal models of AD (Hefti and Mash, 1989; Olson et al., 1992). Since NGF does not cross the blood-brain barrier, administration to the brain provides a challenge.

Nerve growth factor administration to the brain has previously been tested in clinical studies of AD patients. A first trial was performed in three AD patients where NGF was delivered by intracerebroventricular infusion (Eriksdotter Jönhagen et al., 1998). The results indicated an up-regulation of nicotinic receptors and glucose metabolism on positron emission tomography (PET), as well as a normalization of the electroencephalography (EEG) pattern. However, adverse effects of NGF (neuropathic pain being the most prominent), made this route of administration unacceptable for routine treatment. In a clinical trial using genetically modified fibroblasts secreting NGF injected into the basal forebrain in AD patients, Tuszynski et al. (2005) demonstrated the feasibility of an ex vivo gene therapy approach with results indicating a slowing of disease progression. Our group has shown that targeted delivery of NGF through encapsulated cell biodelivery (NGF-ECB) into the basal forebrain is safe and well-tolerated (Eriksdotter-Jönhagen et al., 2012; Wahlberg et al., 2012). Three of the six patients responded to the NGF-delivery with a decrease in brain atrophy (Ferreira et al., 2015), and an increase in cholinergic markers in cerebrospinal fluid (CSF), correlating with improved cognition and brain glucose metabolism (Karami et al., 2015). Importantly, the activity of the acetylcholine synthesizing enzyme, ChAT in CSF showed a significant increase in patients with stable cognition during the 12-month NGF delivery (responders), compared to those patients who declined cognitively (non-responders) (Koenig et al., 2011).

The present report aimed to explore the potential effects of NGF-ECB administration on qEEG parameters during a 12-month exploratory study in a small, but unique sample of AD patients that have undergone this advanced experimental treatment. Our findings corroborate the previously reported changes in the rate of brain atrophy (Ferreira et al., 2015) and cholinergic markers in CSF (Karami et al., 2015) and further investigate changes in the functional measure of synaptic brain activity such as EEG. Moreover, we wanted to investigate whether qEEG changes were related to the global measures of cognitive stabilization and the levels of cholinergic markers in the CSF by using an advanced statistical method to analyze multichannel EEG data.

Six patients were enrolled in the study, two men and four women, implanted with NGF-ECB implants in the basal forebrain for 12 months, when the implants were removed uneventfully. All six patients were treated with ChEI as concomitant medication for a median duration of 12 months (range 8–26) before enrollment and continued this therapy throughout the study (Table 1).

This exploratory study is the first to report a correlation between the cholinergic marker ChAT in CSF and EEG oscillations in the alpha and theta frequency range during encapsulated NGF delivery to patients with a definite AD diagnosis as confirmed by a cortical biopsy (Eriksdotter-Jönhagen et al., 2012). Associations between changes in ChAT activity and EEG in narrow windows of both the theta and alpha frequency bands were mainly observed during the first 3 months of NGF delivery and do not seem to increase further thereafter. These limited effects could be explained by the fact that all the patients already were on a stable long-term treatment with ChEI which is a requirement for all clinical trials with new experimental treatments.

This was also the first cohort of patients who received NGF delivery via the NGF-ECB implants, with safety and tolerability as primary endpoints. However, at implant retrieval at 12 months, the NGF release was low (Eriksdotter-Jönhagen et al., 2012). It would be of interest to study the delivery of NGF with NGF-ECB implants for a longer duration than 12 months at sustained NGF dose level, to see if there are similar or even more pronounced effects on EEG, cholinergic biomarkers, and cognition.

In our study, a cognitive stabilization in some patients, as measured by an increased or stable MMSE score during the NGF delivery was related to a significant effect in a narrow frequency of the fast alpha band 10–11.5 Hz. Although such a narrow window of significant correlations might be the result of small sample size and could be questioned as physiologically relevant, an alternative explanation for the selective effect on the fast alpha frequencies may be the different functional significance of slow and fast alpha oscillations (Klimesch, 2012). It has been suggested that activities in the lower alpha bands (slower alpha oscillatory frequencies) are related to attentional processes and that those of the higher alpha bands (faster alpha oscillatory frequencies) are related to retrieval from the long-term memory (Klimesch, 1996). Although this knowledge is generated in EEG studies during a cognitive activation paradigm, we could speculate that it could be extrapolated to the resting EEG during stimulation of the cholinergic system, which may preferentially activate neuronal networks involved in selective cognitive processes.

An alternative explanation may be that both thalamocortical and cortico-cortical networks are responsible for the generation of rhythmic alpha oscillations, assuming that the balance with other neurotransmitter systems during the disease process in AD affects the profile of EEG changes and cognitive response (Dringenberg, 2000).

While our earlier study showed that treatment with ChEI gave limited improvements of qEEG parameters up to 6 months (Jelic et al., 1998), this study showed a sustained association between positive changes in qEEG profile in the patients who showed clinical stabilization as suggested by MMSE change after a 12 month NGF delivery. The narrow windows of correlations between the EEG frequency spectra and cognitive and biochemical markers of cognitive deterioration and biological activity of the disease during the NGF-ECB treatment should be interpreted with a reservation as a converging pattern in the context of current knowledge.

There was no strong consistent relation with theta power and the MMSE change as was shown for alpha power. Theta activity is known to increase early during AD and reaches a plateau during the disease course (Musaeus et al., 2018). Thus, theta power may not be a single sensitive electrophysiological state marker for monitoring disease activity and effects of therapeutic stimulation of the cholinergic system but can rather be considered an indicator of the baseline disease severity and could therefore be a basis for the selection of patients for clinical trials.

One limitation of this study is the small number of patients, which increases the variance of the data and puts limits to the use of conventional statistics. We propose here the use of the Ragu program which by using assumption-free randomization statistics computes significance as a function of frequency, controls for type 2 error, and displays results in a user-friendly visual format with EEG spectral power covariance maps (Gudmundsson et al., 2007).

Another limitation is the invasiveness of the treatment method which induces a minor brain lesion during implantation of the NGF-releasing implants. This is reflected in the temporary increase in the CSF neurofilament light chain (NFL) protein at the 3-month follow-up, returning to baseline levels when measured at 12 months (Eriksdotter-Jönhagen et al., 2012). However, there was no correlation between the qEEG parameters and CSF NFL levels and no significant interaction with time, MMSE, or ChAT.

A possible confounding placebo effect of the implantation procedure is less likely but since it could not be ruled out future investigation requires a cross-validation design.

We have employed only selected conventional EEG parameters that have shown in previous studies reliable test-retest reliability (Gudmundsson et al., 2007; Näpflin et al., 2007) at the individual level, which was important for the present study with the limited number of patients. Since the molecular family of NGFs plays an important role in synaptic plasticity (Liu et al., 2014; Numakawa and Odaka, 2021) it would be of interest to further explore on a larger sample size EEG connectivity measures that are more closely related to the efficacy of neuronal networks.

Although this study is exploratory, the results give further support for the qEEG method as a possible marker of disease activity. Furthermore, it is a desirable outcome measure in clinical trials monitoring treatment efficacy over time since it does not put limits on serial recordings. As with many other treatment strategies in AD, there will probably always be responders and non-responders to therapy and the corresponding need to define these patients with an objective and validated functional measures. The present results motivate the use of qEEG as one of the outcome measures in the clinical trials of novel potential therapeutics in AD. However, future validation studies with different treatment approaches on a larger sample of patients are further warranted.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, upon request.

Ethical approval was obtained from the Regional Human Ethics Committee of Stockholm. The patients/participants provided their written informed consent to participate in this study.

HE, ME, and VJ: conceptualization. TK, PA, BL, GL, LW, TD-S, and VJ: methodology. HE, TK, and VJ: formal analysis. HE, AK, and ME: investigation. HE, ÅS, ME, and VJ: resources. HE and VJ: writing—original draft preparation. TK and VJ: visualisation. ÅS, ME, and VJ: supervision. ME: funding acquisition. All authors have been engaged in review and editing and have read and agreed to the published version of the manuscript.

LW is an employee of NsGene and owns shares in the company. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.