Sleep as a predictor of tDCS and language therapy outcomes

Abstract

Study Objectives

To determine whether sleep at baseline (before therapy) predicted improvements in language following either language therapy alone or coupled with transcranial direct current stimulation (tDCS) in individuals with primary progressive aphasia (PPA).

Methods

Twenty-three participants with PPA (mean age 68.13 ± 6.21) received written naming/spelling therapy coupled with either anodal tDCS over the left inferior frontal gyrus (IFG) or sham condition in a crossover, sham-controlled, double-blind design (ClinicalTrials.gov identifier: NCT02606422). The outcome measure was percent of letters spelled correctly for trained and untrained words retrieved in a naming/spelling task. Given its particular importance as a sleep parameter in older adults, we calculated sleep efficiency (total sleep time/time in bed x100) based on subjective responses on the Pittsburgh Sleep Quality Index (PSQI). We grouped individuals based on a median split: high versus low sleep efficiency.

Results

Participants with high sleep efficiency benefited more from written naming/spelling therapy than participants with low sleep efficiency in learning therapy materials (trained words). There was no effect of sleep efficiency in generalization of therapy materials to untrained words. Among participants with high sleep efficiency, those who received tDCS benefitted more from therapy than those who received sham condition. There was no additional benefit from tDCS in participants with low sleep efficiency.

Conclusion

Sleep efficiency modified the effects of language therapy and tDCS on language in participants with PPA. These results suggest sleep is a determinant of neuromodulation effects.

Clinical Trial: tDCS Intervention in Primary Progressive Aphasia https://clinicaltrials.gov/ct2/show/NCT02606422

Statement of Significance.

Sleep plays a pivotal role in cognitive functioning. Given this notion, and the prevalence of sleep disorders in neurodegenerative diseases, we investigated whether baseline sleep efficiency (before therapy) was a predictor of transcranial direct current stimulation (tDCS) therapy and language outcomes in primary progressive aphasia (PPA), a neurodegenerative disorder primarily affecting language. Our data reveals sleep efficiency as a positive predictor of both tDCS and language therapy outcomes (i.e. better sleep resulted in more significant tDCS effects and an increase language therapy outcomes up to two-month post-therapy). The present study fills a critical gap in the literature by investigating sleep as a modulator of therapeutic efficacy (with and without neuromodulation) in PPA.

Introduction

Older adults are vulnerable to sleep disturbances, and those with neurodegenerative disorders are disproportionately affected [1–3]. Among individuals with neurodegenerative disease, underlying pathology and atrophy in sleep-related regions of the brain lead to poor sleep quality. For example, amyloid accumulations disrupt sleep in individuals with Alzheimer’s disease (AD) [4], and the loss of hypocretin cells within the hypothalamus may be responsible for sleep disturbances in individuals with Parkinson’s disease (PD) [5, 6]. In a less common form of dementia, known as frontotemporal lobar degeneration (FTLD), atrophy in the prefrontal cortex and diencephalic structures is believed to result in poor sleep patterns [7, 8]. Beyond pathology and atrophy, sleep disturbances can emerge in neurodegenerative disorders as a consequence of related medications, fatigue, depression and pain.

Primary progressive aphasia (PPA), a subtype of FTLD, principally affects speech and language. PPA is divided into three distinct variants: the semantic variant (svPPA), characterized by fluent speech, impaired semantic memory and single word comprehension and anterior temporal lobe atrophy; the nonfluent variant (nfvPPA), characterized by effortful, nonfluent speech with grammatical errors and omissions, spared comprehension of grammatically simplistic sentences, possible motor speech disorders (e.g. dysarthria or apraxia of speech [AOS]) and atrophy in the left inferior frontal gyrus (IFG), insula and supplementary motor areas; and the logopenic variant (lvPPA), in which individuals present with impaired repetition and sentence comprehension and frequent word finding pauses due to anomia with atrophy in the posterior temporal cortex and left parietal lobule atrophy [9]. The vast majority of individuals with PPA have been found to have tau-positive, ubiquitin/TDP43-positive FTLD or AD pathology mostly in the left hemisphere (see Gorno-Tempini et al. for review [10]). There is currently no treatment for the neuropathology of affected individuals, and disease-modifying pharmacological options have not yet yielded favorable outcomes [11].

Research groups have focused extensively on the development of behavioral therapies to slow linguistic decline in PPA. Numerous studies (mostly targeting verbal naming) have reported improved therapy outcomes [12–16]. After promising results in post-stroke aphasia [17–19], transcranial direct current stimulation (tDCS) has recently been used to augment language therapy in PPA. This neuromodulatory approach involves the safe, easily applied, and well-tolerated application of low voltage electrical current to the brain through surface electrodes (anode and cathode). Initial results for PPA are promising [20].

Critical questions remain regarding factors that predict how effectively patients with neurodegenerative disorders respond to language therapies more broadly and tDCS in particular, which is pertinent in a population in which the goal of therapy is to slow progression. Studies from our group have revealed that brain volumes [21], white-matter integrity [22] and cognitive/language performance at baseline [23] predict effects of tDCS combined with speech-language therapy in patients with PPA, but no studies to date have assessed sleep as a predictor. By investigating modulators of therapy efficacy, researchers will better understand the characteristics of patients who respond to language therapy and electrical stimulation and ultimately determine the patients for whom augmentative interventions may be efficacious in the spirit of precision medicine.

Sleep can be viewed as a relevant modulator to therapy in this population because those with PPA are likely to experience sleep-related problems over the course of their disease. Individuals with FTLD pathology report developing excessive drowsiness, narcolepsy-like behaviors (i.e. involuntary sleep attacks), insomnia and other sleep-related symptoms [8, 24–26] and up to two-thirds of individuals with AD pathology experience sleep-related problems [27, 28], including frequent daytime napping and nighttime wakefulness [29]. Within FTLD, individuals with PPA were shown to be more likely to experience poor sleep quality compared to those with behavioral variant FTLD [30, 31]. While the prevalence of sleep disorders in individuals with PPA remains unclear, research has shown that individuals with FTLD pathology have a similar rate of disturbed sleep as those with AD pathology [28], with onset occurring earlier in FTLD [8]. Given the likelihood of problematic sleep patterns in PPA, it is valuable to understand how sleep may impact therapy efficacy, particularly language therapy.

Sleep quality measures, including sleep efficiency (i.e. percentage of time in bed spent asleep) and sleep continuity (i.e. consolidation of sleep versus wakefulness in a given sleep period) have been correlated with impairments in cognitive functions in healthy adults. For example, in a group of healthy community-dwelling elders (ages 65–80) sleep quality (as measured by responses to the PSQI) positively correlated with performance on tests of working memory, attentional set shifting and abstract problem-solving [32]. Additionally, Wilckens et al. found that high sleep continuity (i.e. higher sleep efficiency and lower wake time after sleep onset [WASO]) was associated with better memory recall and verbal fluency in a group of older adults (ages 55–77) [33]. Given that poor sleep patterns play a role in cognitive abilities for healthy adults, it is reasonable to speculate that such abilities are even more susceptible to impairment, and perhaps decline, in individuals with neurodegenerative disorders. This may be especially true in disorders that principally affect linguistic domains, like PPA.

Numerous studies have revealed efficacy of tDCS application for improving sleep quality in healthy adults and individuals with neurodegenerative diseases. For example, anodal tDCS over the motor cortex and dorsolateral prefrontal cortex (DLPFC) during waking hours has been shown to improve overall sleep quality in individuals with fibromyalgia [34], and bilateral anodal tDCS stimulation applied simultaneously over the left and right prefrontal and motor areas during wake hours has improved overall sleep quality and decreased sleep onset latency in persons with Parkinson’s disease [35]. No studies, however, have investigated whether self-reported sleep measures in neurodegenerative populations are a predictor of tDCS therapy outcomes in language or other cognitive domains.

We aimed to determine whether sleep efficiency at baseline (before therapy) predicted of tDCS and language therapy outcomes in patients with PPA. Data reported in this study are from a clinical trial (ClinicalTrials.gov identifier: NCT02606422) that examined augmentative effects of tDCS combined with written naming/spelling therapy. We hypothesized that participants with high sleep efficiency at baseline would show more tDCS benefits in language therapy outcomes compared to those with low sleep efficiency.

Methods

Participants

Individuals were eligible to participate if they were native English speakers with premorbid proficiency in spelling, completed a high-school education, did not have developmental disorders (e.g. dyslexia) or other neurological conditions (e.g. stroke), were right-handed, and had a formal criteria-based diagnosis of PPA [9]. Participants were referred by physicians from specialized PPA and FTD centers across the United States. Referrals were generally based on neurological examination, cognitive-linguistic testing, and neuroimaging measures, including magnetic resonance imaging (MRI) and positron emission tomography (PET). Of the 23 participants (mean age 68.13 ± 6.21) from the main clinical trial (NCT02606422) who qualified for this study (see Figure 1), 7 were diagnosed with nfvPPA, 11 with lvPPA, and 5 with svPPA.

Figure 1.

CONSORT diagram.

Design of therapy protocol

In our double-blind randomized within-subjects crossover design protocol (NCT02606422), participants received anodal tDCS over the left inferior frontal gyrus (IFG) coupled with written naming/spelling therapy [36]. Stratified randomization of stimulation condition within each PPA variant determined whether participants received tDCS or sham condition in Period 1. In the present study we report only on Period 1 since we found a possible carryover effect in the main trial [36]. It should be noted that there are some inequalities among variants and the condition type in the present paper compared to the main clinical trial (NCT02606422). This can be explained by the fact that the PSQI was not administered during the initial stages of the clinical trial and only participants who completed the parts of PSQI required to calculate sleep efficiency are included here. tDCS and sham groups were matched for demographics and other clinical characteristics, including age, sex, years post onset of symptoms and language and overall severity measured using the Fronto-Temporal Dementia Clinical Dementia Rating (FTLD-CDR) scale [37] (see Table 1). Of the 23 participants, 12 received tDCS condition and 11 received sham condition in Period 1 with an average of 12.0 (±1.76) sessions (see Figure 1). Factors such as comorbidities and aging-related health issues introduced some variability in the number of sessions.

Table 1.

Means and standard deviations for baseline demographics grouped by Period 1 condition (n = 23)

Participant	Condition in Period 1	Variant	Sex	Onset (years)	Age (years)	LangSev	TotSev
P01	tDCS	n	F	6	60	2	8
P02	tDCS	n	F	4	69	2	6
P03	tDCS	n	F	2	69	2	10
P04	tDCS	s	M	3	75	3	14
P05	tDCS	s	M	7.5	59	2	5.5
P06	tDCS	l	F	9.5	70	3	10
P07	tDCS	n	M	2.5	80	2	3
P08	tDCS	s	M	7	73	3	8
P09	tDCS	s	F	2.5	64	1	1
P10	tDCS	l	M	1	54	2	8
P11	tDCS	l	M	4	63	1	8.5
P12	tDCS	s	M	10	71	2	4
Mean				4.92	67.25	2.08	7.17
SD				3.01	7.44	0.67	3.52
P13	sham	l	M	3.5	69	1	3.5
P14	sham	l	F	2.5	73	2	7
P15	sham	n	M	6	64	3	15
P16	sham	l	F	8	66	3	19
P17	sham	l	F	3	71	2	5
P18	sham	n	M	26	62	3	9.5
P19	sham	l	M	7.5	74	1	3
P20	sham	l	F	1.5	69	3	17
P21	sham	n	F	4	71	3	9
P22	sham	l	M	10	77	2	6.5
P23	sham	l	F	2.5	64	1	3.5
Mean				4.95	69.09	2.09	8.91
SD				2.72	4.70	0.83	5.68
p-value				0.49	0.24	0.49	0.20

Demographics for each participant. Variant: n = nonfluent PPA, l = logopenic PPA, s = semantic PPA. Onset = years since onset of language symptoms. LangSev = language severity per Fronto-Temporal Dementia Clinical Dementia Rating (FTD-CDR) scale (language domain only) [38]; TotSev = total severity per FTD-CDR scale (all eight domains: memory, orientation, judgement and problem-solving, community affairs, home and hobbies, personal care, behavior/comportment/personality, and language). Mean and standard deviation (SD) were calculated based on Period 1.

The two periods of therapy (anodal tDCS + language therapy versus sham tDCS + language therapy) were separated by two months, and all participants received both conditions (see Figure 2). Evaluations were conducted before, immediately after, two weeks and two months post-therapy. All evaluation and therapy sessions took place at the Johns Hopkins Hospital. The study was approved by the Johns Hopkins Institutional Review Board and all participants provided written consent.

Figure 2.

Study design model.

Participants were asked to complete a packet of questionnaires at baseline, including the Pittsburgh Sleep Quality Index (PSQI) [39]. The PSQI was developed as a self-report measure of sleep quality. Seven components are assessed by the participant based on the last month: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication and daytime dysfunction. We focused on Component 4, Habitual Sleep Efficiency, which is calculated by manually dividing Question #4 (During the past month, how many hours of actual sleep did you get at night?) by Question 3 (During the past month, when have you usually gotten up in the morning?) minus Question 1 (During the past month, when have you usually gone to bed?) × 100. Based on a skewed distribution of the data across 23 participants, we applied a median split (median = 95.2%). In the present paper, we considered “high” sleep efficiency to be above the median (11/23 participants) and “low” sleep efficiency to be less than or equal to the median (12/23 participants). Demographic differences between high and low sleep efficiency groups are reported in Table 2. To increase objectivity of the PSQI, we asked participants’ spouses, partners and individuals who live with our participants to verify their responses and provide guidance while completing the questionnaire. Therefore, we did not use separate questionnaires for caregivers.

Table 2.

Means and standard deviations for baseline demographics grouped by sleep efficiency (n = 23)

Participant	SE group	SE (%)	Variant	Sex	Onset (years)	Age (years)	LangSev	TotSev
P1	High	100	n	F	6	60	2	8
P3	High	100	n	F	2	69	2	10
P7	High	100	n	M	2.5	80	2	2
P9	High	100	s	F	2.5	64	1	1
P12	High	100	s	M	10	71	2	4
P13	High	100	l	M	3.5	69	1	3.5
P14	High	100	l	F	2.5	73	2	7
P15	High	100	n	M	6	64	3	15
P18	High	100	n	M	6	62	3	9.5
P22	High	100	l	M	10	77	2	6.5
P23	High	100	l	F	2.5	64	1	3.5
Mean	–	100	–	–	4.86	68.45	1.91	6.45
SD	–	0	–	–	2.85	6.08	0.67	3.84
P2	Low	83.7	n	F	4	69	2	6
P4	Low	95.2	s	M	3	75	3	14
P5	Low	88.9	s	M	7.5	59	2	5.5
P6	Low	86.5	l	F	9.5	70	3	10
P8	Low	87	s	M	7	73	3	8
P10	Low	85.3	l	M	1	54	2	8
P11	Low	94.1	l	M	4	63	1	8.5
P16	Low	95.2	l	F	8	66	3	19
P17	Low	63.6	l	F	3	71	2	5
P19	Low	87.5	l	M	7.5	74	1	3
P20	Low	75	l	F	1.5	69	3	17
P21	Low	94.1	n	F	4	71	2	9
Mean	–	86.34	–	–	5	67.83	2.25	9.41
SD	–	8.84	–	–	2.67	6.05	0.72	4.68
p-value	–	–	–	–	0.45	0.41	0.14	0.06

Demographics for each participant. SE = sleep efficiency. Variant: n = nonfluent PPA, l = logopenic PPA, s = semantic PPA. Onset = years since onset of language symptoms. LangSev = language severity per Fronto-Temporal Dementia Clinical Dementia Rating (FTD-CDR) scale [38]; TotSev = total severity per FTD-CDR scale. Mean and standard deviation (SD) were calculated based on sleep efficiency grouping: high versus low.

tDCS

To deliver stimulation, we used the Soterix Transcranial Direct Current Stimulation 1 × 1 Clinical Trials device, Model 1500. Specifications of tDCS setup are described in our main trial [36]. In short, the anode was placed over the left frontal lobe, centered on F7 in the 10–20 electrode placement system [40] and the cathode was placed over the right cheek. We aimed to excite the left frontal lobe with anodal stimulation, which has been shown to augment language rehabilitation in other studies [18]. Nonmetallic, conductive rubber electrodes, 5 cm × 5 cm, were fitted with saline-soaked sponges to limit skin-electrode reactions. The full left IFG was covered due to the size of the electrodes. During tDCS condition, current was delivered with an intensity of 2 mA (estimated current density 0.08 mA/cm²) for a total of 20 min in each therapy session. Delivery of tDCS was simultaneous with the start of language therapy, which continued for an additional 25 min beyond the cessation of stimulation. During sham condition, current ramped up to 2 mA for 30 s and back down to 0 mA in the beginning and middle of the 20 min neuromodulation session; such procedures have successfully blinded participants to the stimulation condition [41]. Additionally, participants were connected to electrodes for the same duration in both conditions. To monitor any adverse effects, each participant was asked to rate his or her pain level once or twice during each session, using the Wong Baker FACES Pain Rating Scales (www.WongBakerFACES.org). Some participants reported tingling and itching that lasted approximately 30 s at the onset of stimulation.

Behavioral language therapy

Participants received written naming/spelling therapy from a licensed speech-language pathologist based on the principals outlined in Copy and Recall Training (CART) [42]. Given the differences in spelling deficits among the three PPA variants, word sets were individualized while procedures and outcome measures remained consistent. Each participant was presented a picture on the computer and asked to produce the target word verbally and then in written form. If there were difficulties retrieving the target word, semantic knowledge was evaluated using the principles of Semantic Feature Analysis [43]. If semantic knowledge did not cue the target word, the participant was verbally presented the word and asked to repeat it three times. Likewise, if the participant expressed difficulties writing the word correctly, the correct spelling was presented in a spell-study-spell procedure (i.e. correct word was presented, each letter was reviewed, participant copied the word three times) [15].

Treated and untreated word sets (10–30 single words) depending on case severity were matched in length and frequency. Eleven participants were evaluated and treated on nouns, eleven individuals on verbs and one participant on both nouns and verbs. Four therapy evaluations were conducted for each period of therapy: before, immediately after, two weeks post, and two months post-therapy. Although therapy aimed to recruit both semantics and naming, the focus was on spelling and the main outcome was spelling performance. Letter accuracy was determined based on a scoring system that considered letter deletions, additions, substitutions, and movements [44]. To account for interrater reliability, each item was scored by one trained individual and then by second trained individual who noted as discrepancies. Discrepancies in scoring were resolved by a consensus score. Interrater reliability was 95% [38].

Of note, both nouns and verbs were involved in the primary outcome measure. Previous results from our group have shown that tDCS over the left IFG improves written naming accuracy more for both nouns [36] and verbs [45] compared to sham. While attenuated effects were observed by our group in verbs compared to nouns [45], we postulated that the richness of information carried by verbs may make them more challenging to treat, as shown in post-stroke aphasia [46]. Regardless, tDCS-related gains were observed in our studies and the left IFG triangularis and the left inferior temporal gyrus have been shown to be neural substrates for single level-word retrieval of both nouns and verbs in PPA [47]. Given this overlap, we included nouns and verbs in the present study as we did in the main clinical trial [36].

Statistics

Demographics and other patient characteristics were compared between treatment order (tDCS or sham first) and sleep efficiency groups (high or low), using t-tests for continuous variables. The primary outcome measure was the absolute difference in the percentage of letters correctly identified for target words (trained and untrained) retrieved in the written naming/spelling task before therapy and follow-up evaluations (immediately after, two weeks and two months post-therapy) limited to Period 1. We compared outcomes between sleep efficiency strata for each follow-up time and for the average over follow-up times. Outcomes were also compared between treatments (tDCS versus sham) within strata of sleep efficiency. Estimates of these effects, standard errors, and confidence intervals were obtained using the Generalized Estimating Equations (GEE) method with robust estimation of the variance of the estimates [48].

Data availability statement

We will make the data and associated documentation available to users only under a data-sharing agreement that provides for: (1) a commitment to using the data only for research purposes and not to identify any individual participant; (2) a commitment to securing the data using appropriate computer technology; (3) proper acknowledgment of data source; and (4) a commitment to destroying or returning the data after analyses are completed.

Results

Demographics among groups

There was no statistically significant difference in demographic characteristics between groups based on treatment order (tDCS versus sham) or sleep efficiency group (high versus low) (see Tables 1 and 2).

Effects of sleep efficiency on therapy outcomes

Participants with high sleep efficiency benefited more from therapy irrespective of condition (tDCS or sham) than did participants with low sleep efficiency in learning therapy materials at all evaluation time points (see Table 3, Figure 3, A). The effect sizes (raw difference as a percentage of standard deviation; 100 × raw difference/ standard deviation) of sleep efficiency predicting therapy outcomes for trained items were 82% immediately after therapy, 62% at two weeks post-therapy, and 76% at two months post-therapy. There was no effect of sleep efficiency in generalization of therapy materials (see Table 4, Figure 3, B).

Table 3.

Comparison of participants with high versus low sleep efficiency for trained items

	Immediately after	2-weeks post	2-months post	All times
#High/#low participants	11/12	11/10	9/11	–
Raw difference in means	24	19	23	22
Standard error	10	9	9	8
p-value	0.011	0.031	0.01	0.006
Effect size (100 × raw difference/standard deviation)	82%	62%	76%	99%

Figure 3.

Effects of sleep efficiency on language therapy outcomes. Blue represents high sleep efficiency. Red represents low sleep efficiency. The numbers in blue and red represent the number of participants with high versus low sleep efficiency. * p ≤0.05, ** p ≤0.01, *** p ≤0.005.

Table 4.

Comparison of participants with high versus low sleep efficiency for untrained items.

	Immediately after	2-weeks post	2-months post	All times
#High/#low participants	11/11	11/10	9/11	–
Raw difference in means	2	0	2	1
Standard error	3	5	6	4
p-value	0.459	0.977	0.775	0.738
Effect size (100 × raw difference/standard deviation)	10%	-1%	8%	14%

Effects of tDCS on participants with high and low sleep efficiency

Among participants with high sleep efficiency, those who received tDCS benefitted even more from therapy than those who received sham, especially over time (see Table 5, Figure 4, A). The effect sizes as a percentage of standard deviation of sleep efficiency predicting therapy outcomes for trained items at 2-months post-therapy was 352%. There was no additional benefit for tDCS in participants with low sleep efficiency (Table 6, Figure 4, B).

Table 5.

Comparison of tDCS versus sham on trained items, among all participants with high sleep efficiency

	Immediately after	2-weeks post	2-months post	All times
#High/#low participants	5/6	5/6	3/6	-
Raw difference in means	15	10	37	21
Standard error	12	13	11	10
p-value	0.226	0.413	0.001	0.044
Effect size (100 × raw difference/standard deviation)	68%	53%	352%	115%

Figure 4.

Effects of sleep efficiency and tDCS on language therapy outcomes. Pink represents tDCS condition. Orange represents sham condition. The numbers in blue and red represent the number of participants in each condition. * p ≤0.05, ** p ≤0.01, *** p ≤0.005.

Table 6.

Comparison of tDCS vs sham on trained items, among all participants with low sleep efficiency

	Immediately after	2-weeks post	2-months post	All times
#High/#low participants	5/6	4/6	4/6	-
Raw difference in means	10	-3	11	6
Standard error	14	11	10	11
p-value	0.489	0.814	0.293	0.582
Effect size (100 × raw difference/standard deviation)	37%	-15%	39%	26%

Discussion

The aim of the present study was to determine whether baseline sleep efficiency predicts tDCS and language therapy outcomes for individuals with PPA. Language therapy involved a written naming/spelling protocol and was coupled with either tDCS or sham condition. Sleep efficiency percentages for 23 participants were extracted from the PSQI. Our results reveal that sleep efficiency moderated the effects of language therapy (tDCS and sham conditions combined). This suggests that individuals with high sleep efficiency may benefit significantly more from language therapy than those with low sleep efficiency. Furthermore, among participants with high sleep efficiency, those who received tDCS benefitted even more from therapy than those who received sham (at the longest follow-up point, two months post-therapy), whereas those with low sleep efficiency showed no evidence for this additional benefit. This suggests that individuals with high sleep efficiency may benefit significantly more from tDCS in the long run compared to those who receive language therapy alone.

Sleep-induced neuroplasticity for learning, memory, and now for language therapy

The present results, showing that sleep efficiency predicts language therapy outcomes, are in line with previous results in the literature highlighting the importance of sleep on learning and memory [49–52]. Neuroplasticity involves the functional and/or structural modification of neuronal circuits [53] and it is based on modulations of synaptic plasticity. There are two basic mechanisms by which sleep modulates synaptic plasticity: (1) it downscales overall synaptic strength that becomes saturated during wakefulness (homeostatic plasticity), and (2) it affects transmission across single synapses by inducing long-term potentiation (LTP), or LTP-like activity (associative plasticity) [54, 55]. While the explicit effects of sleep on the interrelation between these two mechanisms has yet to be specified, we highlight each mechanism below to emphasize the impact of sleep on plasticity.

• (1) The first mechanism, also referred to as the synaptic homeostasis hypothesis, postulates that during sleep (particularly slow-wave sleep), synapses are downscaled to counteract the strengthening of synapses that occurs throughout the brain during waking hours (e.g. during learning) [56–58]. Slow-wave sleep is argued to facilitate the downscaling of synaptic weight to a baseline level that is more sustainable. The most compelling evidence of the synaptic homeostasis hypothesis comes from Vyazovskiy et al. who showed in a rat model that network connectivity is increased following wakeful experiences and decreased following sleep [59].

• (2) The second mechanism, referred to as associative synaptic plasticity, is the activity-dependent strengthening of transmission across single synapses, also referred to as long-term potentiation (LTP). Long-term potentiation is a Hebbian-like mechanism that induces the strengthening of synapses due to repeated exposure, leading to a prolonged increase in synaptic transmission [60]. The downscaling of synaptic weight (first mechanism above) seems to be a stipulation for LTP that also happens during sleep [54]. This mechanism of associative plasticity (LTP) is also a molecular correlate for both learning and memory [61–65].

In terms of sleep deprivation and behavior (i.e. learning), it has been shown that suppressing SWS, during which synaptic downscaling occurs [58], in older adults reduced their ability to encode new information the subsequent morning [66]. For language learning specifically, Sterpenich et al. showed that sleep (in comparison to sleep deprivation) following learning sessions improved memory for newly learned nonwords and resulted in improved outcomes on a lexical decision task targeting the same set of learned nonwords [67]. While there are numerous studies proposing that consolidation of memories during sleep might be a key component for language learning in healthy children and adults [68], there is no study to our knowledge that refers to the role of sleep in language learning during rehabilitation in either post-stroke aphasia or a neurodegenerative disease affecting language, such as PPA. The goal of language rehabilitation for individuals with PPA, specifically, is to relearn linguistic material that has been lost due to brain disease. In the present study we showed that those with higher sleep efficiency at baseline performed better on the language therapy task (spelling) than those who did not sleep as well, suggesting that poor sleep may counteract the strengthening of memory traces encoded during therapy sessions, resulting in poorer therapy outcomes. The present study is the first to our knowledge suggesting that higher-quality sleep enhances language learning related to rehabilitation outcomes. Of note, higher-quality sleep also enhanced tDCS-related outcomes for participants in our study.

Evidence for homeostatic and associative (LTP) plasticity as tDCS mechanisms

Aside from being a cellular mechanism for sleep [54], as well as learning and memory [61–65, 69], there is evidence of associative, Hebbian-type plasticity as a tDCS mechanism across the tDCS literature [70–73]. Kronberg et al. [70], in particular, documented LTP facilitation at the cellular level as an effect of tDCS. They interpreted this tDCS effect as a cellular tDCS mechanism that augments learning when tDCS is applied in conjunction with behavioral training that promotes plasticity, as it is the case in rehabilitation. Intriguingly, electrical stimulation in the motor cortex of rats impeded the induction of LTP after prolonged wakefulness and reinstated the induction of LTP following sleep [59], pointing to a synergistic effect of sleep and tDCS.

Further, this synapse strengthening (i.e. LTP) takes place on glutamatergic neurons [74], which bind to their receptors in the post-synaptic terminal. One of those receptors is the ionotropic receptor called n-methly-d-asparate (NMDA). Learning and consolidation are thought to be dependent on NMDA [75], and tDCS is also be dependent on NMDA to enhance learning and memory [76]. Other studies have shown that occluding NMDA receptors in humans reduces the effect of anodal tDCS [76, 77] and inhibits sleep in rats [78]. Therefore, better sleep could lead to improved outcomes for language therapy, as well as for tDCS as a neuromodulatory adjunct to therapy.

The hypothesis that homeostatic plasticity is a mechanism for tDCS, as it is for sleep, has not been explicitly articulated in the tDCS literature. However, both our group [79, 80] and others [81] who have reported on neural effects of tDCS in neurodegenerative conditions, have documented that tDCS decreases functional connectivity between the stimulated areas and other areas of the brain. Importantly, we and others have also found that these decreases downregulated abnormally high overall functional connectivity in neurodegeneration [82, 83]. Therefore, decreases in overall functional connectivity as a mechanism for tDCS effects in neurodegenerative conditions, align well with and may correspond at a macrolevel, to the synaptic downscaling hypothesis.

The question that has been raised is whether sleep modulates tDCS effects. In a paper advocating for consecutive days of tDCS application during training, Reis et al. speculated that off-line tDCS effects are based on sleep consolidation of learned information [84]. However, in a consecutive paper this was not confirmed to be the case and time post-learning had the same effect on tDCS effects regardless of whether it was spent sleeping or awake [85]. It should be noted that the follow-up study did not compare sleep during the night to sleep deprivation but instead a 3-h interval of sleep versus wake period. The present study does, however, argue that the quality of sleep efficiency during night sleep is a predictor of tDCS effects for elderly individuals with compromised brains due to neurodegeneration.

tDCS during sleep or wake?

It is noteworthy that the majority of tDCS research in the field of sleep investigates the effects of tDCS applied prior to or during sleep cycle (see Annarumma et al., 2018 for a review) [86] and the majority of outcome measures for tDCS therapy effects in the field of sleep are specific sleep parameters (e.g. sleep duration) [87, 88] or memory-related tasks [89, 90]. The present study differs in that tDCS was applied during a period of wakefulness with and without evidence-based behavioral therapy targeting a language task. Importantly, the mechanism in tDCS-during-sleep studies (e.g. modulation of sleep-related neural oscillations [90–93]) may not be the same as tDCS-during-wake studies like ours. Indeed, it would be very informative to compare these two different types of tDCS application in rehabilitation.

Limitations and future directions

A limitation to the present study is that sleep measures were self-reported and approximately half of our sample reported sleep efficiency to be 100% (i.e. they were asleep for the entire duration they spent in bed). This is unlikely, given that the general population of older adults typically have a sleep efficiency at or below 80% [94]. However, the unusually high mean of sleep efficiency presented in this paper may suggest that even small disturbances at sleep are a particularly sensitive predictor of therapy outcomes. Future studies should include objective sleep measures (e.g. actigraphy, polysomnography) and larger sample sizes in order to examine the extent to which this differs among PPA variants. Variants have different pathologies although sleep has been found to be equally, if not more, disturbed in FTLD than AD [8]. Our study showed that those individuals with high sleep efficiency benefit more from tDCS. But, could tDCS be applied to enhance sleep which could then improve therapy outcomes? While the present results do not address this hypothesis, tDCS application during both wakefulness and sleep should be studied as adjunct options to language recovery in neurodegenerative diseases.

Conclusion

To our knowledge, the present study is the first of sleep quality effects in behavioral and neuromodulatory language rehabilitation, in which the outcome was performance on an evidence-based language therapy task. Sleep efficiency modified the effects of language therapy with and without tDCS and of tDCS itself in participants with PPA. Advocating for the role of sleep as a modifier of language therapy outcomes may prompt referrals to sleep specialists and discussions of pharmacologic options for individuals prior to beginning language therapy both in post-stroke aphasia and in neurodegenerative conditions with language deficits. Addressing disrupted sleep may also lead to improved language performance and overall quality of life, which emphasizes further investigation of sleep in relation to behavioral, pharmacological, and neuromodulatory treatments in aphasia.

Disclosure Statements

Financial disclosure: This work was supported by grants from the National Institutes of Health (National Institute of Deafness and Communication Disorders) through award R01 DC014475 to K.T. and National Institute on Aging award R01 AG050507 to A.P.S. K.T. receives consultant fees for other NIH awards and from YBrain. A.P.S. received an honorarium as a consultant to Merck and honoraria from Springer Nature Switzerland AG for guest editing special issues of Current Sleep Medicine Reports.

Nonfinancial disclosure: None.

Preprint Repositories

This paper is not submitted in a preprint repository or any other form of media.