1. Introduction
Apraxia of speech (AOS) is a condition that affects oral motor speech planning and production. It results in impaired speech fluency due to inhibition of the neural programming of articulation [
1]. It can occur in the absence of dysarthria (i.e., a language impairment characterized by paralysis or paresis and muscular control problems) [
2] and aphasia (a multimodal language impairment affecting language comprehension and production) [
3,
4]. Usually, AOS results from stroke, but neurodegeneration, traumatic brain injury, genetic disorders, or syndromes (e.g., childhood apraxia of speech) may also trigger AOS [
1,
5,
6,
7,
8,
9]. In this study, we will refer to AOS in the context of primary progressive aphasia (PPA), a neurodegenerative condition with speech and language deficits as its primary symptoms [
10,
11,
12]. According to the consensus criteria for subtyping of PPA [
13], AOS and agrammatism are key symptoms for identifying patients with the non-fluent PPA (nfvPPA) variant from patients with other PPA variants. However, since agrammatism occurs without AOS in some patients [
14] and AOS is the only symptom in others [
15], a number of studies suggested a clinicopathological presentation of AOS as a distinct PPA variant, primary progressive apraxia of speech (PPAOS) [
16,
17,
18].
The primary characteristics of AOS are articulatory and prosodic deficits with different degrees of severity (mild to severe) [
19,
20], resulting in effortful, slow speech, manifested by longer consonants and vowels [
21,
22,
23,
24]. For example, Duffy and colleagues (2017) [
21] argued that slow speech rate and abnormal lexical stress are primary characteristics of progressive AOS that make speech in patients with progressive AOS effortful, slow, labored. Thus, longer segmental duration (vowels and consonants) [
25,
26] constitute a primary deficit of patients with AOS that distinguishes them from patients with other PPA symptoms [
21,
22,
23,
24]. These measures may also serve as an objective and ecologically valid measure of AOS and an excellent outcome measure to estimate the effects of treatment(s) and symptom progression. Patients with AOS exhibit inconsistent and non-systematic speech articulatory errors and irregular insertions, distortions, deletions, substitutions, and transpositions of sounds [
27,
28,
29,
30]. They often produce consonants with irregular voicing [
31], stop consonants (e.g., /p/ and /t/) with irregular plosive distortions and increased voice onset time (VOT) [
24,
26,
32] or fricative consonants (e.g., /f/ and /θ/) with misplacing and/or misshaping the active articulator (tongue) relative to the passive articulator (a place along the palate) [
33]. AOS results in reduced coarticulation of adjacent sounds, a slowing down of syllable transitions, and non-canonical syllable segmentation [
26]. Further, irregular prosody and rhythm have been reported as characteristics of speech in patients with AOS [
34], affecting lexical (e.g., stress) and post-lexical prominence patterns and tonalities.
AOS symptoms have been associated with the left inferior frontal gyrus (IFG), an area involved in kinematic and sound representations of speech production [
6,
16,
17,
35,
36]. Patients with AOS show subtle structural and functional irregularities in the IFG, including other areas of the frontal operculum: the posterior frontal gyrus (i.e., pars opercularis [BA44]), which enables the cognitive selection of vocal and orofacial actions [
35,
37], the pre-supplementary motor area (pre-SMA), which controls vocalization [
17], and the insula under the left IFG, which facilitates articulatory planning [
20,
38]. BA 44 (pars opercularis) is proximal to the premotor cortex, an area involved in articulation and is bi-directionally connected with BA 40 via the ventral component of the superior longitudinal fasciculus (SLF III). This interaction with BA 40 provides BA 44 (which organizes speech production by selecting the phonemes, the words, and their order to form the sentence to be spoken) with critical phonological information. Thus, this cortico-cortical circuit appears to be the phonological loop in the left hemisphere. This parieto-frontal circuit formed by the SLF III (and perhaps the arcuate fasciculus) is involved in phonological processing [
39]. The anterior insula is found to be involved in motor speech planning [
40]. Other proximal and distal brain regions have also been associated with AOS, such as the parietal lobe [
4], the basal ganglia, and the cerebellum [
41]. Additionally, there are few behavioral studies with encouraging results targeting AOS symptoms in nfvPPA/AOS [
42].
Transcranial direct current stimulation (tDCS) is a non-invasive brain stimulation that modulates neuronal excitability by modifying neural cells’ resting membrane potential either by hyperpolarizing or depolarizing. The placement of the anode (positive electrode) and cathode (negative electrode), intensity, and the duration of stimulation are known to affect the efficacy of tDCS. Recent studies by our group [
43] and others [
44,
45] provided novel insights into the mechanisms of tDCS, showing that changes in functional connectivity (FC) and gamma aminobutyric acid (GABA) concentrations and may be important tDCS mechanisms. We found that tDCS modulated (decreased) functional connectivity (FC) between the stimulated area and the functionally, or structurally, connected temporal areas of the language network, as well as the homologous area in the right hemisphere (but not the default mode network (DMN)); these FC changes were maintained up to 2 months. These results, which are in line with similar decreases in connectivity observed after tDCS over the left IFG in aging [
46] and other neurodegenerative conditions, may be interpreted as an indication that fewer resources are needed after tDCS than before for related language tasks. We also tested the hypothesis that tDCS reduces GABA in the stimulated tissue in PPA. We applied GABA-edited magnetic resonance spectroscopy (MRS) to quantify GABA levels before and after a sham-controlled tDCS intervention with language therapy in PPA. Participants receiving tDCS had significantly greater language improvements than those receiving sham immediately after the intervention and at 2 months follow-up. GABA levels in the targeted tissue decreased after the intervention and remained so for 2 months [
43].
The association of motor planning and speech articulatory deficits to the left IFG has motivated neuromodulatory studies with tDCS that targeted this area. Specifically, in Marangolo, Marinelli, Bonifazi, Fiori, Ceravolo, Provinciali and Tomaiuolo [
19], three subjects with post-stroke aphasia with AOS participated in a randomized double-blinded experiment involving articulatory training in tDCS and sham conditions. Each subject participated in five consecutive daily sessions of anodal tDCS (20 min, 1 mA) and sham stimulation over left IFG. tDCS resulted in more improvement than sham condition. Chesters, et al. [
47] tested the effect of tDCS in adults who stuttered and found that anodal tDCS did not improve sentence reading, although, they observed a trend towards a reduction in stuttering when tDCS was coupled with a fluency intervention. In a follow up study, Chesters, et al. [
48] tested 30 individuals who stuttered, in which 15 had tDCS and 15 had sham and speech fluency intervention using choral and metronome-timed speech. The authors showed a significant fluency improvement in individuals with tDCS measured one week after the intervention compared to intervention without tDCS. The effects of tDCS were maintained six weeks after therapy during reading but not during conversation. Chester and colleagues concluded that tDCS may be effective for improving speech articulation in other patient populations. Furthermore, the positive effects of tDCS in speech production are supported by studies showing that tDCS improves speech production in typical speakers [
49].
Although previous tDCS studies in PPA, including our group’s largest--to our knowledge—double-blind, sham-controlled, cross-over trial of tDCS efficacy in PPA have shown positive effects of tDCS on spoken and written naming and spelling [
50,
51,
52,
53,
54,
55], there is no tDCS study demonstrating the potential of tDCS in reducing AOS symptoms despite the fact that AOS is a prominent feature in patients with nfvPPA or PPAOS (here referred to as nfvPPA/AOS to avoid classification debates). To our knowledge, this is the first study to provide preliminary, proof-of-concept evidence of tDCS efficacy as an adjuvant to speech therapy in PPA patients with AOS symptoms. The present study does not intend to suggest any criterion about AOS diagnosis or test the reliability of perceptual judgments for PPAOS diagnosis (e.g., Dabul et al.’s, questionnaire), or provide evidence for the presence or absence of AOS derived from perceptual judgments. The present study tests a simple hypothesis: if temporal acoustic measures, such as longer sound durations, are a characteristic of AOS [
19,
20,
21,
22,
23,
24,
25,
26,
27], and the left IFG is a critical area for motor planning in speech production [
6,
16,
17,
36,
37,
38,
39,
40], then tDCS over the left IFG may normalize these sound durations significantly better than speech therapy alone in nfvPPA/AOS.
In the present study, we hypothesized that tDCS over the left IFG coupled with speech production therapy will reduce AOS symptoms in patients with nfvPPA/AOS more than sham, i.e., speech production treatment alone. We used sound duration as a measure of AOS symptoms and reduced sound duration as an improvement of speech production in these patients. As slow speech production is a distinguishing characteristic of speech for patients with nfvPPA/AOS, a decrease in sound duration was considered as a therapeutic improvement corresponding to faster speech articulation. We asked three questions: (1) is tDCS more effective than sham in reducing sound duration in patients with nfvPPA/AOS? (2) Are tDCS effects sustainable over a two-months period? (3) Do tDCS effects generalize to untrained items? To answer these questions, we designed an experimental study where patients with nfvPPA/AOS received anodal tDCS over the left IFG or sham stimulation for the same duration paired with a word repetition task. Patients were evaluated three times: before treatment, immediately after treatment, and two months post-treatment. All words produced were segmented into vowels and consonants and we measured their temporal properties. Changes in syllable length do not affect equally their constituents, namely the vowels and consonants that make up these syllables [
26,
56,
57], as changes in length primarily involve vowels. Therefore, the effects of tDCS on vowel and consonant duration may be different, which motivated us to study the two sound categories separately.
4. Discussion
In this study, we investigated whether tDCS over the left IFG coupled with speech therapy improves sound duration in patients with nfvPPA/AOS more than sham, i.e., speech therapy alone. First, we evaluated whether tDCS is more effective than sham in improving sound duration in patients with nfvPPA/AOS and whether effects sustained for 2 months post-treatment. Second, we evaluated whether the effects of tDCS generalized to untrained items. Third, we evaluated whether effects differed between vowels and consonants. Our findings show that (1) tDCS in conjunction with speech therapy reduces sound duration significantly more than speech therapy alone (sham). Furthermore, tDCS effects sustained over time, i.e., the tDCS advantage was maintained for up to 2 months post-treatment. (2) The effects of tDCS generalized to untrained items immediately after treatment but this improvement was not maintained at 2 months post-treatment. (3) Patients who received tDCS coupled with speech therapy produced shorter vowels and consonants than patients who received speech therapy alone (sham). Below, we discuss the findings in detail, the contribution and limitations of this study, and future directions.
The most important finding of this study is that tDCS reduced sound duration immediately after and up to 2 months post-treatment with respect to baseline for trained and untrained items. Furthermore, in trained items, sound duration approached the sound duration of healthy controls, although sounds produced by patients with nfvPPA/AOS were still significantly longer than those produced by healthy controls. In sham condition, sound duration slightly increased (1.2%) immediately after treatment with respect to baseline and remained the same at 2 months post treatment. This study shows that combining speech training with tDCS induces more sustaining effects. Such sustaining effects of tDCS were observed in other studies related to speech fluency and articulation. For example, Marangolo, Marinelli, Bonifazi, Fiori, Ceravolo, Provinciali and Tomaiuolo [
19] also found improvement in response accuracy 2 months post-treatment in three patients with stroke-induced speech apraxia. Chesters, Mottonen and Watkins [
48] showed that the tDCS effect on stuttering severity sustained for six weeks post-treatment in reading (but not in conversation). Furthermore, tDCS showed significant generalization of improvement in sound duration relative to sham. Taken together our findings suggest that tDCS has the potential to improve AOS symptoms. This is particularly important for nfvPPA/AOS since some patients may only present with AOS symptomatology at least in initial stages [
13,
21].
The tDCS montage in the present study targeted the left IFG. As discussed in the Introduction, the left IFG, and in particular the left IFG opercularis, is associated with articulatory motor planning and is adjacent to the primary motor areas of the mouth and tongue [
84,
85]. Given the size of our electrodes (2 × 2 inches), we cannot claim that we targeted only the left IFG or the IFG opercularis, although this area would be functionally related to AOS symptoms. Recent evidence of the principle of ‘functional targeting’ in the tDCS literature, concurs with the opinion that the current flows only on active cells, those related to the function that is trained [
86]. Our previous study has shown that a possible mechanism for tDCS effects is through changes in functional connectivity of the stimulated area, the left IFG, in particular [
19]. Although stimulation over the left IFG improved speech production, our findings do not exclude a speech improvement due to stimulation over homologue areas in the right hemisphere or other adjacent areas of the premotor cortex or the insula [
87]. A subsequent functional connectivity study would need to provide evidence that this particular stimulation montage caused the present effects of segmental duration of vowels and consonants.
TDCS resulted in shorter vowels and consonants, yet the effects were greater on vowels than consonants. This is not surprising, since vowels and especially stressed vowels, are intrinsically longer than most consonants [
22,
57], and this is the case even for geminate consonants in languages that have geminates, such as Finnish and Estonian. Therefore, this effect may not reflect a selective effect on vowels but rather opportunities for shortening. There are several underlying causes for these intrinsic differences between vowels and consonants, such as stress, post-lexical prominence (nuclear or pronuclear pitch accents), or phonetic distribution of lengthening over the syllable onset nucleus and coda, which are language specific effects and further discussion would be beyond the scope of this paper. Sound duration is affected by both articulatory and linguistic parameters. Articulatory factors that affect sound duration may be related to articulatory planning, coordination, and timing of neural commands, execution of articulatory movements, control of the airflow from the lungs towards the oral cavity and the vocal fold vibration in the larynx [
56,
88,
89,
90,
91]. Additionally, phonemic factors that affect sound duration may be related to lexical stress, accentual prominence, lengthening effects demarcating the boundaries of words and phrases, speech fluency, and other communicative effects, such as emphasis [
92]. In other words, sound duration is better seen as an integral measure of different processes affecting speech production. The fact that sound duration is improved means that it could be the effect of a multidomain improvement either articulatory or linguistic (lung air pressure, vocal fold vibration, articulatory target approximation, etc.). The additional effects of articulatory deficits in nfvPPA/AOS, may explain why temporal properties of speech have been shown to distinguish patients with AOS from other patients with PPA [
8,
20,
21].
One remaining question is whether tDCS effects transfer to post-lexical coarticulation level phenomena and prosodic phenomena, such as phrasing, intonation, speech fluency, and speech rate that involve post-lexical processes. Word repetition provides very limited information on phrasing, partly because phrasing here would be defined as a measure between clinician-patient-clinician productions (which is partly determined by the clinician). With respect to intonation, it is difficult to study pitch accents (a nuclear pitch accent, a phrase accent, and a boundary tone) at the word level [
93]. By studying only F0, it would be very difficult to explain what constitutes an amelioration of the deficit vs. normalization. Furthermore, speech fluency and speech rate require sentence level productions. Nevertheless, segmental duration should be highly correlated with these sentence-level measures, as reduced segmental duration would indicate faster sentence production. Future studies should also incorporate connected speech productions.
The main limitation of this study is the small number of participants, and therefore it can only be considered as a preliminary, proof-of-concept study. A related possible limitation is the matching of participants between the two stimulation groups. We matched the patients with respect to the language component of the FTD-CDR. The participants in the sham group seemed to have a little higher overall severity score, although the difference was not very large (4.67 out of possible 27 in the sham group and 1.5 out of 27 in the tDCS group). The overall severity of the FTD-CDR includes the language component but also provides additional scores for memory, orientation, judgement, community affairs, home and hobbies, personal care, and behavior. Although it is possible to entertain that overall severity differences in other than language sections of the FTD-CDR may impact AOS treatment and tDCS effects, the two stimulation groups were matched at baseline on the AOS outcome measure (sound duration). This, in conjunction with their matched language severity, suggests that the overall severity differences did not affect the outcome measures.
Similarly, patient differences in letter fluency (FAS), and syntactic comprehension (SOAP) were not reflected on sound duration at baseline (the dependent variable of this study) as both groups exhibited approximately the same mean sound duration as shown in
Figure 3. If differences in performance on letter fluency or syntactic comprehension tasks influence the neuromodulatory effect of tDCS on sound duration as a primary AOS symptom, it remains an empirical question. Such a finding would rather speak against the consensus classification, i.e., against the fact that nfvPPA is a unitary variant. Rather, it should be split in two as Duffy et al., 2017 have argued: one with AOS symptoms (PPAOS) but without initial fluency or syntactic deficits and another with initial fluency and syntactic deficits and no AOS symptoms. Nevertheless, we acknowledge these differences in the statistics we run, by considering the participant as a random slope.
Another possible limitation is the inherent diffusivity in tDCS methodology, including the lack of specific current flow estimation for each of the present participants. Nevertheless, previous current modeling in
Figure 1 showed that the current distribution was centered in the left IFG. The choice for the 5 × 5 cm
2 electrode patches in our tDCS montage in the present study as well as in most previous clinical studies is driven by the premise and ease of transferring this methodology to clinic, if shown to be efficacious. Although not as precise as other tDCS methodologies, such as high-definition tDCS, the inherent large spread of electrical current in the present and other clinical studies, may actually be the very reason of their efficacy as the current affects larger brain regions. Future studies comparing these methods are needed to determine their clinical efficacy.