Seventy-eight subjects from the Barcelona Brain Health Initiative cohort (BBHI; Cattaneo et al., 2018) aged 60 or above were contacted via telephone. First, a pre-recruitment screening was conducted. Subjects were pre-selected if they answered “agree” or “absolutely agree” in at least one of the following items included in a previously online administered Patient-Reported Outcomes Measurement Information System-Cognitive Function (PROMIS-CF; Ader, 2007) questionaire: “I had problems reasoning or recalling things”; “I had to read something several times to understand it”; “I had trouble remembering new information, like phone numbers or simple instructions”; “I had to work really hard to pay attention or I would make a mistake”; “I had trouble remembering whether I did things I was supposed to do.” Then, pre-selected participants underwent a screening call, whereby they were asked if they felt their memory was becoming worse. If the answer was positive, subjects were invited to enroll in the study. Upon initial consent, exclusion criteria were reviewed, which contained any of the following: neurologic or psychiatric diagnosis, contraindication to NIBS (Rossi et al., 2009; Antal et al., 2017), as well as any contraindication for MRI (including claustrophobia or metal and electronic implants). After applying these criteria, a total of 51 participants were selected for the first visit to our center, where a battery of neuropsychological tests assessing the main cognitive domains was administered (for further details see Supplementary Material Section 1.1). To ensure that all participants had a normal cognitive profile, final inclusion criteria were Mini-Mental State Examination (MMSE) score above 25 (Mitchell, 2009) and performances on neuropsychological tests no more than 1.5 standard deviations (SD) below normative scores adjusted by age and years of education (Peña-Casanova et al., 2009). In addition, clinical data were obtained, and we included participants if they had a previous diagnosis of hypertensive, diabetic, or hyperlipidemic conditions, and they were being treated with antihypertensive, antidiabetic, cholesterol-lowering agents, respectively. Six volunteers were excluded due to abnormal performance in neuropsychological assessment, two for brain MRI abnormalities (aneurysm and meningioma) and four due to methodological problems during the experimental design. Finally, one participant decided to withdraw from the study. Overall, 38 tDCS naïve individuals with mean age of 62.29 years (SD = 1.56) were finally included.

All study procedures were approved by the Institutional Review Board (IRB 00003099—amendment at the University of Barcelona) and Comité d’Ètica i Investigació Clínica de la Unió Catalana d’Hospitals in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). Written informed consent was obtained from each participant prior to study enrollment.

The present work was a randomized, double-blind, sham-controlled study. Participants and the study team members were not aware of the tDCS condition applied at any point of the experiment. Subjects were randomized into active or sham tDCS by using a simple randomization procedure with MATLAB (version R2019a, The MathWorks, Inc., Natick, MA, United States). The study protocol was accurately set based on previous publications (Manenti et al., 2017, 2018; Sandrini et al., 2019). The reconsolidation experiment per se began 1 month after the neuropsychological assessment (Supplementary Material Section 1.1) to avoid possible interferences with the verbal episodic memory test performed on the first visit (see Figure 1).

The episodic memory task was based on the Spanish-validated version of the Rey Auditory Verbal Learning Test (RAVLT; Schmidt, 1996). Subjects were instructed to try to remember as many words as possible from the cards that they had to transfer from one bag to another. Specifically, participants took each card (containing one written word) from a blue cloth bag, read them aloud, and put them in a black cloth bag. Following this, they were asked to recall as many words as possible. For the next learning trial, the 15 cards were placed in the blue bag again. This procedure was repeated five times as performed elsewhere (Manenti et al., 2017, 2018). At the end of this session, participants were asked to complete a memory strategies questionnaire (based on Manenti et al., 2010), which comprised 10 possible tactics commonly used to enhance the learning process.

Twenty-four hours later, the same experimenter involved in DAY-1, within the same experimental room, encouraged participants to describe the procedure followed the day before. This was done by showing the blue and black cloth bags. Participants were interrupted if they started recalling any specific word. Between 5 and 10 min after this contextual reminder (Monfils et al., 2009), tDCS (active or sham) was administered (Sandrini et al., 2014; Manenti et al., 2017). Subjects completed a Visual Analog Scale (VAS; Crichton, 2001) mood questionnaire immediately before and after tDCS induction. Also, at the end of the stimulation session, participants were asked to complete a NIBS-related adverse events questionnaire (adapted from Brunoni et al., 2011) to measure the perceived discomfort induced by tDCS.

On DAY-3, the delayed recall was measured in 4 recorded and timed attempts (i.e., free delayed recall 1). Each attempt was alternated with a copy of 3 simple sets of figures to avoid verbal interference and subvocal repetition of words: (i) circle + overlapping rectangles and (ii) diamond + cube extracted from the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuropsychological battery, and (iii) the interlocking pentagons from MMSE. Following this, a timed recognition-task was conducted, where participants had to identify the 15 words from DAY 1, out of a total of 30 cards (i.e., recognition 1). At one-month follow-up (DAY-30) subjects returned and delayed recall was again measured with only a timed and recorded attempt (i.e., free delayed recall 2). Then, a second timed recognition-task was performed (i.e., recognition 2), with 30 cards of mixed words from DAY-1 and 15 new words (different set of cards from DAY-3). Of note, participants were never informed that words would be asked again on DAY-3 and DAY-30. In addition, at both retrieval sessions, the experimenter requested for any review of the words at home or any cognitive tasks (such as crosswords in-between experimental days) that might have caused either facilitation or interference. Furthermore, at DAY-30, participants underwent additional questionnaires to assess factors such as memory complaints and functional status: cognitive reserve (CR), memory complaints (MiCog), Cognitive Difficulty Scale (CDS), and Functional Activity Questionnaire Pfeffer (FAQ). Additionally, as sleep quality is linked to learning and memory processes (Diekelmann and Born, 2010), the Pittsburgh Sleep Quality Index (PSQI) was administered at DAY-3 and DAY-30.

Following the same procedure described in previous studies (Manenti et al., 2017, 2018; Sandrini et al., 2019), tDCS was applied at 1.5 mA through two saline-soaked sponge electrodes of 7 × 5 cm (current density: 0.043 mA/cm²; Antal et al., 2017). Stimulation was delivered using a DC-Stimulator Plus (neuroConn GmbH, Ilmenau, Germany) and the same montage was used in both experimental groups (active and sham). According to the international 10–10 system of measurement, the anodal electrode was positioned over the F3 (l-DLPFC) and the cathodal electrode was placed over the FP2 (right supraorbital area). In all groups, the current was initially increased and finally decreased in a ramp-like fashion of 10 s. In the sham condition, the current delivery was terminated after 15 s of stimulation with no further blinding processes. In the active group, the current was supplied for 15 min. The stimulation was applied with a double-blind fashion and all the parameters adhered to safety criteria guidelines (Rossi et al., 2009; Antal et al., 2017).

MRI data were acquired in a 3T Siemens scanner (MAGNETOM Prisma) with 32-channel head coil at the Unitat d’Imatge per Ressonància Magnètica IDIBAPS (Institut d’Investigacions Biomèdiques August Pi i Sunyer) at Hospital Clínic de Barcelona, Barcelona.

For all participants, a high-resolution T1-weighted structural image was obtained with a magnetization prepared rapid acquisition gradient-echo (MPRAGE) three-dimensional protocol [repetition time (TR) = 2,400 ms, echo time (TE) = 2.22 ms, inversion time = 1,000 ms, field of view (FOV) = 256 mm, 0.8-mm isotropic voxel]. They also underwent resting-state fMRI (rs-fMRI) multiband (anterior-posterior phase-encoding; acceleration factor = 8) interleaved acquisitions (T2^∗weighted EPI scans, TR = 800 ms, TE = 37 ms, 595 volumes, 72 slices, slice thickness = 2 mm, FOV = 208 mm). All the MRI images were examined by a senior neuroradiologist [N.B.] for any clinically significant pathology (none found) and all study participants had Fazekas Scale scores lower than 1. Then, all acquisitions were visually inspected before analysis by one of the first co-authors [L.M.-P.] to ensure that they did not contain MRI artifacts or excessive motion.

Statistical analyses were performed using IBM SPSS Statistics (Statistical Package for Social Sciences, version 24.0. Armonk, NY: IBM Corporation). Demographic, and neuropsychological variables, functional assessment (i.e., CDS, MiCog, Pfeffer FAQ), depression measures (i.e., HDRS), CR and quality of sleep (i.e., PSQI) questionnaires, VAS scores, tDCS aftereffects, quality of sham, strategies used during the verbal episodic memory task, review the words learned at home, and possible interfering cognitive tasks, were compared between groups using independent-sample t-test. For categorical data, differences between groups were evaluated using the chi-squared (χ²)-test.

As regards the verbal episodic memory task, performance was measured as accuracy scores (hits minus intrusions) in DAY-1 (5 trials mean), DAY-3 (4 trials mean) and DAY-30 (1 trial), and, recognition scores (total hits minus intrusions) at DAY-3 and DAY-30. Repeated measures ANOVA was used to investigate differences between group trajectories (DAY-1 vs. DAY-3; DAY-3 vs. DAY-30; DAY-1 vs. DAY-30) in retrieval accuracy, recognition, time of retrieval and time of recognition. As post-hoc pairwise analyses, differences between tDCS conditions (active vs. sham) were performed with independent-sample t-test at each time-point (DAY-1; DAY-3; DAY-30). Likewise, differences between time-points (DAY-1 vs. DAY-3; DAY-3 vs. DAY-30; DAY-1 vs. DAY-30) for each group (active and sham independently) were measured using paired-samples t-tests.

Similar to the neuroimaging approach, we tested differences between responder and non-responder subgroups among the active-tDCS group (see section “tDCS Effects in Episodic Memory Performance” for the sample stratification details), in order to identify baseline cognitive and/or demographic characteristics associated with a greater tDCS-related modulation. Further, the rs-FC average strength measures obtained [see section “Functional Connectivity (FC) Measures”] were compared between subgroups (responder vs. non-responder). In addition, these rs-FC values were used to test correlations with the cognitive change.

Data distribution was tested for normality with Shapiro-Wilk test (p > 0.05). For those variables not normally distributed, non-parametric tests were used: Mann-Whitney U was carried out for group comparisons and Spearman correlations were performed to test correlations. Differences were considered significant at p < 0.05. Graphical representations were performed using IBM SPSS Statistics and GraphPad Prism (version 6.00, GraphPad Software, La Jolla, CA, United States).

No baseline differences were found between sham and active tDCS groups for demographic variables, neuropsychological and functional assessment, depression, CR and sleep quality. It should be noted that the number of on-treatment controlled hypertension, diabetes, and hyperlipidemia in both groups were equivalent (for further details see Supplementary Table 1). There were no group differences regarding the use of strategies during the verbal episodic memory task (χ² = 1.027, p = 0.311), reviewing the learned words at home (DAY-3: χ² = 0.792, p = 0.374; DAY-30: χ² = 0.175, p = 0.676) or performing possible interfering cognitive tasks (χ² = 0.230, p = 0.631). Furthermore, VAS pre-stimulation scores (all p > 0.05, see Supplementary Table 2) and quality of sham questionnaire results (χ² = 1.250, p = 0.535) did not differ between groups. All participants tolerated the stimulation well, and only heat and pinching (categorized as mild or moderate) were reported as tDCS-related adverse events in a similar proportion for both groups (heat: χ² = 1.254, p = 0.534; pinching: χ² = 1.067, p = 0.587).

Episodic memory performance decreased with time for the whole sample (DAY-1 vs. DAY-3: t = 4.934, p < 0.001; DAY-3 vs. DAY-30: t = 4.164, p < 0.001). Both active and sham groups significantly reduced accuracy performance on DAY-3 as compared with DAY-1 (active: t = 3.314, p = 0.004; sham: t = 3.697, p = 0.002). Critically, the active group maintained episodic memory accuracy in the follow-up session (DAY-30) compared to DAY-3 (t = 1.929, p = 0.070), while the sham group significantly decreased the number of words retrieved at DAY-30 (t = 4.149, p = 0.001; see Figure 2A). However, no significant group differences were detected at any experimental session as regards retrieval accuracy (DAY-1: t = 1.094, p = 0.281; DAY-3: t = 0.153, p = 0.879; DAY-30: t = 1.352, p = 0.185).

Focusing on recognition scores, we identified a time^∗group interaction (F = 7.399, p = 0.010). Specifically, performance in recognition significantly declined between the third session (DAY-3) and the 1-month follow-up measure (DAY-30) for the sham group (t = 4.003, p = 0.001). On the other hand, the active group maintained a stable recognition performance across measures (t = 0.900, p = 0.380; see Figure 2B). However, there were no significant differences between groups in recognition at DAY-3 (t = 1.168, p = 0.250) or DAY-30 (t = −1.686, p = 0.100).

There were no group differences regarding the time spent for completing the learning session (DAY-1: t = 0.992, p = 0.328), the retrieval (DAY-3: t = 0.848, p = 0.402; DAY-30: t = 0.058, p = 0.954), or recognition (DAY-3: t = 0.634, p = 0.530; DAY-30: t = −0.416, p = 0.680). There were no significant time^∗group interactions considering DAY-3 and DAY-30 time-related scores (retrieval: F = 0.429, p = 0.517; recognition: F = 0.684, p = 0.414). However, the active-tDCS group significantly increased the time employed for the recognition task at DAY-30 compared with DAY-3 (t = −2.585, p = 0.019). It should be noted that the time^∗group interaction identified for the recognition performance between DAY-3 and DAY-30 (Figure 2B) remained significant after adjusting by time of recognition (data not shown). For further details regarding episodic memory performance see Supplementary Table 3.

Finally, due to the tDCS effect in recognition (Figure 2B), we split the active group into responders and non-responders according to the difference between DAY-3 and DAY-30 in recognition scores. As a result of this, 13 out of 19 subjects were classified as responders (change equal or above 0; i.e., stable progression) and 6 out of 19 as non-responders (change below 0; i.e., longitudinal decline). We did not identify differences between these subgroups regarding demographic variables, clinical data (i.e., hypertension, diabetes, hyperlipidemia), other questionnaires (i.e., CDS, MiCog, Pfeffer FAQ, CR, PSQI), or cognitive performance (considering the neuropsychological evaluation and memory scores at DAY-1).

Regarding the behavioral effects, this investigation showed that anodal tDCS applied over the l-DLPFC is capable of strengthening existing episodic memories by reducing recognition performance loss up to 1 month, relative to sham stimulation, in individuals with SCD. Further, consistent with the previously achieved work (Manenti et al., 2017), we also identified clear tDCS-related cognitive effects on recognition, but not on free recall. This is in line with previous evidence showing that the familiarity component of recognition is relatively preserved in the aging process, whereas recollection does show age-related loss (Danckert and Craik, 2013). Moreover, no group differences were detected on the learning curve and immediate free recall (DAY-3), which is consistent with the notion that certain tDCS effects might take place after rather than during stimulation (Flöel et al., 2012; Abellaneda-Pérez et al., 2020). Anatomically, the current work and previous reports in the field (reviewed in Sandrini et al., 2020) are in accordance with the claim that l-DLPFC might have a causal role in the strengthening of existing episodic memories in a reconsolidation framework. Beyond the importance of this replication in terms of beneficial effects of tDCS on memory function, the present findings support the view that the memory reconsolidation process is susceptible to be strengthened by applying tDCS over the l-DLPFC after a contextual remainder (Sandrini et al., 2014, 2019; Manenti et al., 2017).

Concerning the neural basis of the tDCS-induced cognitive effects, our findings suggested that responsiveness to brain stimulation in this population could be partially explained by differences in structural and functional brain characteristics, as previous research suggested (Kim et al., 2014; Abellaneda-Pérez et al., 2018). At a structural level, CTh differences detected between cognitive responders and non-responders among the active group over the left middle temporal cortex are in line with the hypothesis that individual variability in behavioral outcomes of tDCS might be partly explained by individual anatomical differences (Kim et al., 2014). In particular, our results suggest that greater structural integrity in key memory-related areas (i.e., middle temporal lobe) predicted better cognitive gains induced by tDCS. Further, we assessed functional connectivity to evaluate if individual cognitive responsiveness to the tDCS-reconsolidation approach was predicted by rs-fMRI measures. We found that DMN intrinsic connectivity was associated with the magnitude of individual tDCS-induced memory enhancement in recognition. These results support that an adequate memory functioning, and particularly, a NIBS-induced enhancement in this cognitive domain, also depends on the integrity of brain functional systems associated with preserved cognitive functioning in aging (Abellaneda-Pérez et al., 2018; Antonenko et al., 2019). Therefore, the integrity of the DMN, which is crucial in cognitive aging and in SCD (Wang et al., 2013), is also critical to determine the individual capability to respond to stimulation protocols aimed to modulate cognition (Vidal-Piñeiro et al., 2015; Abellaneda-Pérez et al., 2018; Antonenko et al., 2019). Overall, these MRI-derived results highlight that a detailed subject-specific brain characterization might be used in forthcoming trials to select those individuals that may benefit the most from NIBS-based interventions. Finally, it is relevant to consider that the responder and non-responder groups classified following our tDCS-reconsolidation protocol, which entailed structural and functional neuroimaging differences, did not differ in baseline cognition, demographic variables, or clinical data. Thus, the stated imaging group differences are not likely explained by potential dissimilarities regarding vascular risk factors, which might have implied potential cortical excitability differences in response to NIBS, as previously reported (reviewed in Cantone et al., 2020).

Despite the relatively small number of participants could represent a limitation, the sample was larger than previous similar studies (Sandrini et al., 2014, 2019; Manenti et al., 2017, 2018). However, future research applying multiple sessions of tDCS after the contextual reminder should be conducted to define the potentially long-lasting effects of this non-invasive intervention. The present neuroimaging findings suggesting that greater structural and functional brain preservation may represent a key factor in the selection of those individuals with greater chances of interventional success might be efficiently used to boost the global outcomes of cognitive-NIBS therapeutical approaches in individuals with self-perceived cognitive decline. Notwithstanding, it should be noted that these brain-behavioral associations are only linked to the specific protocol used, and thus, our data may also suggest that distinct cognitive settings or different stimulation doses might be more appropriate. In the same line, further studies are needed to determine optimal stimulation designs and to shed light on the specific neurobiological mechanisms underlying the cognitive effects induced by this intervention before conducting clinical trials applying this memory reconsolidation protocol with tDCS.