Transcranial Direct Current Stimulation Modulates Activation and Effective Connectivity During Spatial Navigation

Brain Stimulation 7 (2014) 314e324

Contents lists available at ScienceDirect

Brain Stimulation journal homepage: www.brainstimjrnl.com

Transcranial Direct Current Stimulation Modulates Activation and Effective Connectivity During Spatial Navigation Benjamin M. Hampstead a, b, *, Gregory S. Brown b, Justin F. Hartley a a b

Rehabilitation R&D Center of Excellence, Atlanta VAMC, Decatur, GA, USA Department of Rehabilitation Medicine, Emory University, Atlanta, GA, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 23 September 2013 Received in revised form 6 December 2013 Accepted 7 December 2013

Background: Allocentric navigation declines with age and neurologic disease whereas egocentric navigation does not; differences that likely arise from maladaptive changes in brain regions mediating spatial (parietal cortex; hippocampus) but not procedural processing (caudate nucleus). Transcranial direct current stimulation (tDCS) holds promise for treating such decline given its ability to modulate neuronal excitability, but its effects have yet to be examined on spatial navigation. Objectives/hypotheses: Using healthy young adults as a model, Study 1 intended to validate a novel spatial navigation paradigm using functional magnetic resonance imaging (fMRI). Using these data to determine targets for tDCS, Study 2 aimed to determine if 1) stimulation modulates activation in a polarity-specific manner; 2) stimulation results in global and/or task-specific activation changes; 3) activation changes are accompanied by changes in effective connectivity. Methods: All participants underwent fMRI while learning allocentric and egocentric environments. Twelve participants completed Study 1. In Study 2, 16 participants were randomized to 20 min of tDCS (2 mA) using a montage with the anode over PZ and cathode over AF4 (PþF) or the reverse montage (PFþ). Results: Study 1 revealed that distinct networks preferentially mediate allocentric and egocentric navigation. Study 2 revealed polarity-dependent changes in activation and connectivity. The PþF montage increased these measures in spatial regions, especially during allocentric navigation, and the caudate nucleus. Conversely, the PFþ montage increased activation and connectivity in lateral prefrontal cortices and posterior hippocampus. Conclusions: These findings support the neuromodulatory effects of tDCS in non-motor areas and demonstrate proof-of-principle for ameliorating age- and disease-related decline in navigational abilities. Published by Elsevier Inc.

Keywords: tDCS fMRI Learning Memory Granger causality analysis Allocentric Hippocampus

Learning and remembering how to travel from one location to another is critical in everyday life and involves the interaction between the allocentric and egocentric navigation systems. During allocentric navigation, individuals use spatial information about an external reference system (e.g., landmarks) to create a cognitive map of the environment and determine their position within it. This navigational approach relies on medial temporal lobe (e.g., The Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Rehabilitation Research and Development Service supported this study through grant B6366W (to BMH). Financial disclosures/conflicts of interest: There are no financial disclosures or conflicts of interest for any author. * Corresponding Author. Department of Rehabilitation Medicine, Emory University, 1441 Clifton Rd. NE, Room 150, Atlanta, GA 30087, USA. Tel.: þ1 404 712 5667; fax: þ1 404 712 1652. E-mail address: [email protected] (B.M. Hampstead). 1935-861X/$ e see front matter Published by Elsevier Inc. http://dx.doi.org/10.1016/j.brs.2013.12.006

hippocampus, parahippocampal gyrus) [1e5] and posterior/medial parietal regions [3], which is consistent with roles these regions play in spatial and associative processing. Additionally, Jerde and Curtis [6] recently presented evidence that the frontal eye fields (i.e., meeting point of the superior precentral sulcus and superior frontal sulcus) and intraparietal sulcus were critical for cognitive map formation and use. Conversely, egocentric navigation is based on an internal reference system and relies on a series of stimuluseresponse relationships that are highly specific and inflexible. This reliance on “habit” memory is most commonly associated with the striatum (caudate nucleus, putamen) [4]. While distinct, these cognitive approaches (and their corresponding brain regions) frequently interact. For example, the egocentric system is primarily engaged when traveling the same, overlearned route to work each day; however, a detour along this standard route would require allocentric processing wherein a

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

cognitive map of the environment is accessed and a new route is planned. Allocentric navigation is known to decline with age [3,7,8] (also see review [2]) and to become impaired in a number of neurological populations, including those with mild cognitive impairment and Alzheimer’s disease (see Refs. [9,10]). Thus, identifying methods of enhancing allocentric navigation may ultimately result in functional improvement in such populations. The current report establishes proof-of-principle for the use of non-invasive brain stimulation in this regard. Transcranial direct current stimulation (tDCS) has been shown to modulate cognitive functioning in a manner consistent with stimulation polarity (e.g., see Ref. [11] for a review). A recent study demonstrated that tDCS effectively modulates more basic aspects of allocentric and egocentric processing [12]. Similarly, low frequency repetitive transcranial magnetic stimulation (rTMS) to the parietal cortex has been shown to disrupt spatial processing [13]. Several studies have paired tDCS and functional magnetic resonance imaging (fMRI) to investigate its effects on various motor [14] and cognitive systems [15]. While such studies have validated the physiological effects of tDCS, they have generally relied on data acquired during the resting-state. Navigation is an active process so the neurophysiological effects of stimulation may be most accurately assessed via task-based fMRI. The current two-part report is the first, to our knowledge, to combine tDCS and fMRI in the assessment of spatial navigation. In Study 1, we used fMRI to validate a novel spatial navigation paradigm. As discussed below, the results of this study revealed functionally related yet highly distinct neural networks mediating allocentric and egocentric navigation. In Study 2, we used these results to determine anodal and cathodal electrode placements and randomized participants to 20 min of tDCS using either a “parietal enhancing” montage with the anode over site PZ and the cathode over AF4, or the reverse “prefrontal enhancing” montage. Participants underwent fMRI immediately after stimulation and the results yield encouraging evidence for the utility of tDCS in enhancing activation within, and effective connectivity between, brain regions involved in allocentric navigation. We enrolled healthy young individuals in both studies in order to (1) validate our paradigm and approach and (2) establish the “normal” effects of stimulation without age- or disease-related confounds. Study 1 As discussed above, the allocentric and egocentric systems are known to interact and this is presumably facilitated by shared brain regions/networks. Thus, our first goal was to identify these shared regions. Despite any potential overlap, we predicted that allocentric processing would preferentially engage brain regions that mediate spatial and relational/associative processing (e.g., parietal regions, parahippocampal cortex, hippocampus) to a greater extent than egocentric processing. Conversely, we expected that egocentric processing would engage brain regions involved in “habit” or procedural knowledge, such as the caudate nucleus. Therefore, our second goal was to examine difference in activation during allocentric and egocentric navigation. Methods Participants A total of 13 healthy, right-handed adults completed a brief neuropsychological protocol to ensure they were cognitively intact (see Table 1). One participant could not tolerate the fMRI environment and was excluded from the study. Thus, data from 12 participants (7 male) were included in Study 1. Exclusion criteria

315

Table 1 Demographics and neuropsychological performances of the participants from Study 1. Standard deviations shown in parentheses. Age (years) Education (years) RBANS (Standard scores) Immediate memory Visuospatial/constructional Language Attention Delayed memory total score Trail Making Test (T-scores) Part A Part B Beck Depression Inventory II (raw scores) Beck Anxiety Inventory (raw scores)

25.8 16.8 103.7 106.8 100.8 101.3 105.7 105.0

(4.0) (2.7) (6.0) (14.2) (11.3) (11.8) (7.2) (10.5)

49.7 54.1 3.1 3.1

(12.8) (8.0) (3.4) (2.6)

included a history of (1) neurological disease or injury (e.g., epilepsy, stroke, moderate e severe traumatic brain injury) (2) mental illness (e.g., depression, bipolar disorder, schizophrenia), (3) sensory (especially visual) impairments that limit the ability to take part in the study (4) learning or attentional disorder; (5) history of (or current) alcohol or drug abuse/dependence; (6) any MRI contraindication as defined by the American College of Radiology. The Emory University Institutional Review Board and the Atlanta VAMC R&D Committee approved the study and all participants provided informed written consent. Stimuli We created a total of eight 3-dimensional environments using MazeSuite 2.0 (www.mazesuite.com), which is a publicly-available program designed specifically for navigational neuroimaging research [16]. There were four allocentric (apartment, movie theater complex, city-scape; English garden) and four egocentric environments (forest, supermarket, office, desert). Two environments (allocentric ¼ English garden; egocentric ¼ desert) served as practice stimuli in both Studies 1 and 2. The remaining 6 environments were used during fMRI scanning (described below). Although virtual environments have been used to examine spatial navigation in previous fMRI research [3,17e19], several factors may have confounded those studies. First, previous research used tasks in which the participant could use one or both navigational approaches and also typically did not instruct participants on how to perform the task (i.e., allocentric vs. egocentric processing). Therefore, we developed environments that were specifically designed to engage either allocentric or egocentric navigation (Fig. 1). To reinforce the particular approach, we provided explicit instructions on how participants should attempt to remember each environment. For the allocentric environments, participants were instructed to create a cognitive map using the spatial relationship between the landmarks (or key features) within the environment and to ignore the sequence of turns. For the egocentric environments, participants were instructed to remember the sequence of left vs. right turns and were told that forming a cognitive map would be useless since each of these environments used the same map. A second limitation of previous research is that participants have been able to freely explore the environment; which could result in different processing strategies and exposure times, especially in those who are less familiar with virtual navigation. Therefore, we created videos that standardized the path through each environment; thereby ensuring comparable exposure across participants. These 60" videos were assembled into four runs that were 8’6" in duration (324 volumes) and consisted of six e 60" active blocks (one video comprised the entire block) that alternated

316

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

Figure 1. Examples of allocentric (A) and egocentric (B) environments. Note that participants are not shown these overhead view but, rather, must form a cognitive map of the environment (for allocentric) and remember the sequence of turns (for egocentric) as they viewed the videos.

with seven e 18" rest blocks. An extra 9" (6 volumes) of fixation time was added to the beginning of each run to allow for scanner stabilization; these volumes were removed prior to data analysis. Each of 6 target environments was shown in a given run. Across runs, however, participants saw a different route through each allocentric environment in order to facilitate cognitive map formation and avoid the encoding of a specific route. Conversely, the exact same path through the egocentric environments was used in each run since this form of navigation relies on overlearned and inflexible memory. Block order was initially randomized for each run but then held constant across participants. The order of the 3 runs was randomized for each participant. To ensure participants were attending to the task, they were instructed to press a button on an MRI compatible response pad at the start of each block. fMRI acquisition (Studies 1 and 2) All scans were performed on a Siemens Trio 3T scanner using a 32-channel head coil. High-resolution anatomic images were acquired using a 3D MPRAGE sequence with repetition time (TR) 2300 ms, echo time (TE) 3.9 ms, inversion time (TI): 1100 ms, flip angle (FA) 8 , 176 sagittal slices of 1 mm thickness, in-plane resolution (IPR) 1  1 mm, in-plane matrix (IPM) 256  256, field of view (FOV) 256 mm. All fMRI data were collected using T2*-weighted functional images acquired with a multi-band slice accelerated gradient-recalled echoplanar imaging (EPI) sequence with BOLD contrast and the following parameters: TR: 1500 ms, TE: 27 ms, FOV: 220 mm, FA: 45 , 64 axial slices of 2 mm thickness, IPR: 2.0  2.0 mm, IPM: 110  110. fMRI data analysis Image processing and analysis used BrainVoyager QX v2.4.2 (Brain Innovation, Maastricht, The Netherlands). For each subject,

the functional images were realigned to the first image of the series. Images were pre-processed using trilinear-sinc interpolation for intra-session alignment of functional volumes, sinc interpolation for slice scan time correction, and high-pass temporal filtering to 2 cycles/run to remove slow drifts in the data. They were then coregistered with anatomic images and transformed into Talairach space [20]. For group analysis, transformed data were spatially smoothed with an isotropic Gaussian kernel (full-width halfmaximum ¼ 4 mm) and normalized across runs and subjects using the z-baseline normalization option in BrainVoyager, which uses data where the predictor values are at or near zero (the default of 0.15 was used). We created a participant-specific anatomic template by first selecting a representative (target) brain and then aligning each of the remaining participants’ brains to this target (linear alignment with sinc transformation). These transformed brains were then combined using the “average vmr” option in BrainVoyager and the resulting template was used to display the activations for this group. Using this template, we also created an anatomically defined mask that was applied to subsequent fMRI data analysis in order to exclude non-brain voxels. To examine brain regions associated with both allocentric and egocentric navigation, we performed a conjunction analysis using the pairwise contrasts of allocentric > baseline and egocentric > baseline using random effects, general linear model (GLM). Correction for multiple comparisons within the anatomically defined mask was achieved by imposing a cluster-volume threshold for contiguous voxels passing a voxel-wise significance threshold of P < 0.05, using a 3D extension (implemented in BrainVoyager QX) of the 2D Monte Carlo simulation procedure [21]. This contrast will be referred to as the conjunction analysis hereafter. To test our primary hypothesis that allocentric navigation would preferentially engage spatial and associative/relational processing,

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

317

we performed a pairwise contrast for the allocentric > egocentric environments using random effects, GLM. Results were cluster corrected using the same method as in the conjunction analysis. This contrast will be referred to as the task analysis hereafter. Memory testing Memory was assessed outside the MRI scanner approximately 30 min after scanning. For allocentric environments, participants were asked to draw a map of each environment and to place the key features within the correct spatial location. The dependent variable was the number of correctly placed features. This design is consistent with that of previous studies [22,23]. For the egocentric environments, participants were instructed to draw the series of turns for a given environment (e.g., supermarket) on a blank sheet of paper. The dependent variable was the number of correct turns until the first error. We calculated the percent correct for both memory tests in order to facilitate direct comparisons. Results and discussion Behavioral There were no significant differences between allocentric (mean ¼ 58.3%, SD ¼ 17.3) and egocentric recall (mean ¼ 58.78%, SD ¼ 35.43) (t(11) ¼ 0.05, P ¼ 0.96), which is not surprising given the young age and intact cognitive status of this group. fMRI Conjunction analysis An extensive, shared network involving ventral (e.g., inferior temporal) and dorsal (e.g., parietal) visual stream areas, cerebellum, frontal eye fields, and dorsolateral prefrontal cortices was activated during task vs. baseline (Supplemental Fig. 1; Supplemental Table 1). Most of these areas are known to be part of the dorsal attention network [24], implicated in working memory [25] and other cognitive processes [6] that are critical for spatial navigation. Medial temporal lobe regions were also involved in this shared network and likely reflect the novel learning (hippocampal) and spatial processing (parahippocampal cortex) demands of both tasks. Overall then, the conjunction analysis revealed that a common and distributed network mediates the general process of spatial navigation within novel environments. Negative BOLD signal (i.e., rest greater than task) was evident in the lateral temporal and parietal as well as medial prefrontal regions. These regions generally comprise the so-called default mode network that mediates internally directed rather than the externally driven cognition [26]. Decreased signal in these regions is consistent with the fact that participants were actively attending to an externally based task. Task analysis Despite the comparable behavioral performances, significant differences in activation were evident when comparing the allocentric and egocentric blocks (Fig. 2; Supplemental Table 2). Allocentric navigation was associated with greater activation throughout regions of both the dorsal and ventral visual streams. Dorsal regions included the superior parietal lobule (including the precuneus), presumably due to the spatial processing requirements, and the intraparietal sulcus e likely due to the working memory, attention, and intentional demands of the task [6]. Ventral regions included the inferior temporal gyrus, fusiform gyrus, collateral sulcus, and parahippocampal gyrus, which are known to play a critical role in the processing of objects, places,

Figure 2. Activation map for the allocentric > egocentric contrast. IFG ¼ inferior frontal gyrus; IPS ¼ intraparietal sulcus; PHG ¼ parahippocampal gyrus; SFS ¼ superior frontal sulcus; SPL ¼ superior parietal lobule.

and integrative landscapes [27]. Whereas the right hippocampus appears to be involved in general spatial-based learning given its activation in the conjunction analysis, the left hippocampus (especially the anterior portion) was more active during allocentric learning. The left cerebral hemisphere is believed to mediate relational spatial processing [28], which was explicitly required in the allocentric task and may explain this hemispheric difference in hippocampal activation. All of these differences were due to

318

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

Table 2 Demographics, neuropsychological performances, and responses to the tDCS side effects questionnaire of the groups from Study 2. Standard deviations shown in parentheses. An FDR corrected P-value of 0.027 was required for statistical significance for the demographic and neuropsychological variables (12 comparisons) and for the tDCS questionnaire (11 comparisons). PþF (n ¼ 8) Age (years) Education (years) RBANS (standard scores) Immediate memory Visuospatial/constructional Language Attention Delayed memory Total score Trail Making Test (T-scores) Part A Part B Beck Depression Inventory II (raw scores) Beck Anxiety Inventory (raw scores) tDCS side effects questionnaire (0 ¼ none; 1 ¼ mild; 2 ¼ moderate; 3 ¼ severe) Headache Neck pain Scalp pain Tingling Itching Burning sensation Skin redness Sleepiness Trouble concentrating Acute mood change Other

PFþ (n ¼ 8)

24.6 (2.4) 16.9 (2.2)

24.4 (5.1) 16.0 (2.4)

Between-group differences t(14) ¼ 0.13; P ¼ 0.9 t(14) ¼ 0.78; P ¼ 0.46

111.3 111.4 106.8 105.1 106.3 109.3

(9.5) (13.6) (5.3) (14.0) (7.6) (12.0)

107.3 113.9 100.0 107.0 108.3 113.8

(10.6) (11.2) (6.7) (12.8) (9.5) (16.3)

t(14) t(14) t(14) t(14) t(14) t(14)

¼ ¼ ¼ ¼ ¼ ¼

0.8; P ¼ 0.44 0.4; P ¼ 0.7 2.2; P ¼ 0.043 0.28; P ¼ 0.78 0.47; P ¼ 0.65 0.63; P ¼ 0.54

55.1 59.5 2.5 3.4

(9.6) (10.0) (1.4) (2.6)

51.9 56.0 5.6 4.0

(8.8) (10.0) (4.7) (4.0)

t(14) t(14) t(14) t(14)

¼ ¼ ¼ ¼

0.71; P ¼ 0.49 0.70; P ¼ 0.49 1.8; P ¼ 0.09 0.37; P ¼ 0.72

Percent reporting (max severity)

Percent reporting (max severity)

0 0 0 75% (1 ¼ mild) 25% (1 ¼ mild) 0 0 0 0 0 12.5% (mild “pulsating”)

0 0 0 75% (1 mild) 12.5% (1 mild) 12.5% (1 mild) 37.5% (1 mild) 0 0 0 0

greater positive BOLD signal for the allocentric than egocentric environments. Several areas demonstrated significantly greater activation for egocentric than allocentric environments; however, the nature of the BOLD signal differences depended on the particular brain region/structure. For example, differences within the right caudate nucleus, right angular gyrus, and left anterior insula were all associated with positive BOLD signal during egocentric navigation but negative BOLD during allocentric navigation. Therefore, these particular areas appear to be vital for egocentric processing, which replicates previous research for the caudate [4,19]. The angular gyrus is known to mediate aspects of both allocentric and egocentric processing [29] as well as in the bottom up capture of memory-related aspects of attention [30]. The repeated exposure to the exact same route through the egocentric environments may have increased these attentional aspects of memory. Similarly, the anterior insula has been implicated in selfawareness [31], which may arise in this case from an increased knowledge about their position and/or prediction of upcoming turns within the route. Differences in the remaining regions (see Supplemental Table 2) were all due to less negative BOLD signal (i.e., closer to 0) for the egocentric than the allocentric environments. Most of these regions were associated with the default mode network and are frequently found to have an inverse relationship with activation in dorsal attention network areas [26]. Thus, the reduced positive BOLD in attention-related areas during egocentric processing also appears to have resulted in concomitant reduction in negative BOLD in default mode regions. Overall then, Study 1 validated our experimental paradigm by revealing that distinct brain regions preferentially mediate allocentric and egocentric processing. Additionally, the egocentric task appears to be an especially well-matched control condition through which functioning in allocentric-specific regions can be examined. These results formed the basis of Study 2.

c2(1,N ¼ 16) ¼ 0, P ¼ 1.0 c2(1,N ¼ 16) ¼ 0.4, P ¼ 0.52 c2(1,N ¼ 16) ¼ 1.1, P ¼ 0.3 c2(1,N ¼ 16) ¼ 3.7, P ¼ 0.055

c2(1,N ¼ 16) ¼ 1.1, P ¼ 0.3

Study 2 Having validated our paradigm and identified the brain regions mediating allocentric and egocentric navigation, the primary goal of Study 2 was to examine whether tDCS modulates functioning within the respective networks. Given the extensive, bilateral parietal activation during allocentric navigation, we chose site Pz as our target for this network. We predicted that anodal stimulation would enhance parietal activation while cathodal stimulation would suppress it. If true, then connectivity with other taskrelevant brain regions, including the hippocampus, should be modulated accordingly. We used site AF4 as our other target for two primary reasons. First, this site lies over the right dorsolateral/superior frontal region that showed greater activation (albeit due to less negative BOLD) for the egocentric environments in Study 1. Second, stimulating this region may have downstream effects on the caudate nucleus given the well-established and extensive connections between these areas [32]. The key questions for this study were (1) does tDCS modulate activation in the expected direction during spatial navigation? (2) Does stimulation result in a global signal change irrespective of task demands and/or does it interact with task demands and result in a relatively specific pattern of activation change? (3) Are changes in activation accompanied by changes in connectivity? Methods Participants and randomization We recruited a new sample of 17 healthy, right-handed adults. The screening measures (neuropsychological; inclusion/exclusion criteria) from Study 1 were again used. Table 2 shows that these groups were comparable in all respects. All participants provided informed, written consent. Randomization used the sealed envelope method, where group assignment was revealed immediately

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

before the tDCS session began. Participants were randomized to either a “parietal enhancing, prefrontal suppressing (PþF)” (4 male; 4 female) montage in which the anode was placed over Pz and the cathode over AF4. The other participants were randomized to a “parietal suppressing, prefrontal enhancing (PFþ)” (5 male; 3 female) montage that used the opposite anode (AF4) and cathode (Pz) placements. One participant from the PþF group fell asleep during fMRI scanning and was excluded from the study. Thus, data from 16 participants were included in Study 2. We elected not to include a sham control group in Study 2 for three reasons. First, we were primarily interested in the neuromodulatory effects of tDCS and believed that comparing groups that received the opposite polarity over the same brain regions had a higher probability of demonstrating such modulation. Second, we believed the non-stimulated group from Study 1 served as a reasonable reference at this stage of the research. Third, it was difficult to justify the additional analytic complexity and costs associated with a sham group without first establishing the modulatory effects on this particular task (point #1). Several steps were taken to minimize intersubject variability. First, all participants were stimulated/scanned between 10ame3 pm (most between 1e3pm). Second, participants who consumed coffee/tea or energy drinks (generally 1e2 drinks/day) were instructed to avoid it for 2 hours before their session since caffeine blood levels peak about 1 hour after consumption. Only two of the 16 participants reported using medication of any kind: 1 used birth control (PFþ group) and 1 used an antibiotic for acne (PþF group). Third, participants were instructed not to use any hair gel/ products since we noticed higher levels of electrical resistance in those who used such products during pilot testing. tDCS Stimulation was performed using a NeuroConn DC-Stimulator Plus (model #0021; Ilmenau, Germany) within a quiet room that was approximately 50 feet from the MRI scanner. Two 7 cm  5 cm rubber electrodes were placed within saline soaked sponge pads and centered over the target locations (current density of 0.057 mA/ cm2). The electrode over Pz was always placed perpendicular to the midline while the electrode over AF4 was always placed parallel to the midline. Stimulation was performed at 2 mA for 20 min, which has been found to modulate cognitive functioning in previous studies (see Ref. [11]). After stimulation, participants completed a brief questionnaire about the nature and severity of any side effects as recommended by Brunoni and colleagues [33]. There were no differences in the number of participants reporting side effects (see Table 2). They were then escorted to the MRI suite and scanning began approximately 10 min after stimulation ended. Participants remained blinded to their stimulation group throughout the study. In order to ensure blinding of the study team, one co-author (GSB) performed the stimulation but no behavioral testing or data analysis. Another co-author (JFH), who was blinded to group assignment, administered and scored the memory tests. The primary study author (BMH) performed all fMRI analyses and was blinded to group during the first level of analysis. We broke the blind at the time of behavioral data and second level fMRI analyses. fMRI The exact same navigation paradigm and memory test, scan parameters, and preprocessing steps were used as during Study 1. Likewise, we created a new anatomical template and mask (for fMRI analyses) using participants from Study 2. We performed two primary contrasts using the ANOVA module in BrainVoyager (both random effects analyses). We first examined the main effect of

319

group and, second, the group by task (allocentric vs. egocentric) interaction. Resulting activation maps were cluster corrected using the same methods as in Study 1. For exploratory purposes, we used the Granger Causality Analysis (GCA) plugin (v2.5) within BrainVoyager to examine the effective connectivity between selected seed-regions and the remainder of the brain. We restricted the GCA to the allocentric blocks since this was our primary condition of interest. After creating connectivity maps at the individual level, we generated the average group maps using the “combine (average) maps” option. These group maps were directly compared using t-tests. Given the exploratory nature of these analyses and our relatively small sample sizes, we set an uncorrected threshold of P  0.05 but required a minimum cluster size of 15 voxels. Results and discussion Behavioral There were no significant main effects of task (F1,14 ¼ 0.22, P ¼ 0.65; h2P ¼ 0.015) or group (F1,14 < 0.001, P ¼ 0.97; h2P < 0.001) nor was the group  task interaction significant (F1,14 ¼ 0.22, P ¼ 0.65; h2P < 0.015). As with Study 1, these results likely relate to the young age and intact cognitive status of both groups. It is possible that our measure of performance (percent correct) was too gross for this population since other groups have demonstrated stimulation-induced changes in spatial processing using more finegrain measures like reaction time [12]. We suspect that our approach will be more sensitive in older adults and those with cognitive impairment since these groups demonstrate objective deficits as discussed above. In this respect, it may be relevant to note that stimulation appears to have facilitated performance to some degree since both groups in Study 2 performed somewhat better than the non-simulated participants in Study 1 (PþF: allocentric recall: 70.83 (16.2); egocentric recall: 70.83 (33.6); PFþ: allocentric recall: 66.67 (16.8); egocentric recall: 74.17 (27.0)). However, it is also possible that these relatively better performances resulted from participants’ expectations that tDCS would have an effect. Thus, a sham group and the use of a withinsubject (pre- post-) design will be critical for controlling for this possibility in future studies. Regardless, the fMRI analyses indicate that the groups in Study 2 achieved similar behavioral performances by preferentially engaging different neural networks. As discussed in detail below, the PþF montage appears to have facilitated “bottom-up” approach whereas the PFþ montage facilitated a “top-down” approach to learning the environments. fMRI analyses Main effect of group As seen in Fig. 3, tDCS induced general changes in activation (i.e., regardless of task demands) in a manner that was generally consistent with stimulation polarity. These differences were almost always due to positive BOLD signal in one group and negative BOLD in the other group. Importantly, the magnitude of mean BOLD signal for participants in Study 1 almost always fell between that of the groups from Study 2. This finding reinforces the neuromodulatory effects of tDCS in non-motor areas. As predicted, the PþF demonstrated greater activation in several parietal regions (left superior parietal lobule, bilateral angular gyrus) when compared to the PFþ group. GCA revealed greater effective connectivity between the left superior parietal lobule and some prefrontal regions in the PFþ group; however, there were robust differences that favored the PþF group between this seed region and multiple neocortical and subcortical structures

320

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

Figure 3. Activation map and coordinate table for the main effect of group fMRI analysis. Note that the column graphs demonstrate beta-weights that are in percent signal change (y-axis). Participants from Study 1 are included as a non-stimulated reference group. AG ¼ angular gyrus; CN ¼ caudate nucleus; Pre-SMA ¼ pre-supplementary motor area; IFS ¼ inferior frontal sulcus.

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

321

Figure 4. Activation map for the group  task interaction fMRI analysis. Note that the column graphs demonstrate beta-weights that are in percent signal change (y-axis). Participants from Study 1 are included as a non-stimulated reference group. IFS ¼ inferior frontal sulcus; IPS ¼ intraparietal sulcus; SPL ¼ superior parietal lobule.

(see Supplemental Table 3). Similarly, GCA revealed greater effective connectivity between the right angular gyrus and a number of posterior visual and cognitive control (e.g., IPS) regions and between the left angular gyrus and occipital regions in the PþF group. Conversely, the PFþ group demonstrated greater

connectivity between the left angular gyrus and a number of prefrontal and subcortical regions. Note that these angular gyrus seed regions were posterior to those involved in egocentric processing in Study 1; findings that may further support some degree of specialization or multi-approach processing within this general

322

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

Table 3 Areas showing a group  task interaction effect (see also Fig. 4). Allo ¼ allocentric; Ego ¼ egocentric. Hemisphere

Peak Talairach coordinates

t-max

x

y

z

19 12 3 9 6 23 22 24 16 9 37

10 46 39 64 60 52 61 48 63 50 78

13 50 55 54 57 46 45 44 47 47 11

8.38 8.12 10.71 10.28 7.54 11.13 31.67 18.80 12.63 9.15 11.13

Left 35 Right 9 Right 21 Right 15 Left 8 Right 6 Mixed * PþF: Ego  Allo; PFþ group: Allo > Ego Inferior frontal sulcus Right 30 Both groups showing negative BOLD Rostral prefrontal cortex Left 20 Medial rostral prefrontal Left 5 Insula Right 31 Anterior cingulate cortex (rostral) Left 5 Right 9

49 31 33 60 85 85

5 8 3 5 17 14

12.14 7.86 13.75 21.63 10.92 19.50

24

22

16.81

48 41 4 35 38

25 7 17 3 1

18.03 15.37 21.53 30.22 11.04

Greater increases in PþF group Caudate nucleus Paracentral lobule Superior parietal lobule Intraparietal sulcus

Right Left Right Left Right Left Right

Precuneus Middle occipital gyrus Greater increases in PFþ group Inferior frontal sulcus Thalamus (pulvinar) Hippocampus (posterior) Lingual gyrus (superior) Cuneus

Left Right

region (see review [29]). Together, these findings reveal that anodal stimulation over the parietal cortex not only increases activation but also enhances the modulatory role this region holds over a more distributed neural network relative to cathodal stimulation. It will be important to examine whether this translates into improved behavioral performance in those experiencing difficulty with allocentric processing, such as older adults or patients. Greater left caudate activation in the PþF group was an unexpected finding but is reasonable given the long-established structural connections between it, premotor regions, and the parietal cortex [32], all of which showed enhanced activation that favored this group. Thus, stimulation of the parietal cortex appears to have direct effects on these structurally connected regions. This relationship was further supported by GCA analyses as the PþF group showed greater effective connectivity between the left caudate and the left superior parietal lobule, left precuneus, and right superior occipital gyrus. It is worthwhile to note that the left superior parietal lobule also showed increased connectivity with the left caudate (Supplemental Table 3), so this relationship appears to be reciprocal in nature. Given these findings, several possibilities emerge for the role of the left caudate that will need to be examined in future studies. For example, the caudate has been associated with increased confidence in memory [34] and greater cognitive control over memory related processes [35]. Such findings could indicate that the PþF montage enhanced participants’ confidence in their learning/memory and/or allowed them to more effectively attend to the environments. We did not, however, find any differences in activation or effective connectivity with the hippocampus using this analysis. Group  task interaction In addition to the task-independent changes described above, tDCS also promoted task-specific effects that were most robust in the parietal and lateral prefrontal cortices (see Fig. 4 and Table 3).

The PþF group showed greater activation in several parietal regions that was specific to the allocentric environments. Some of the most robust differences were within the IPS, a region that has been posited to mediate cognitive map formation/use in concert with the FEF [6]. Using the IPS as a seed region for GCA revealed significantly greater effective connectivity in the PþF than PFþ group. These differences were evident between the left IPS and FEF (bilaterally) while the right IPS showed greater connectivity with the thalamus and parahippocampal gyrus (see Supplemental Table 4). Additionally, connectivity between the left IPS and the parietal cortex, inferior temporal, and occipital regions (bilaterally) was greater in the PþF group. The fact that many of these regions are involved in allocentric processing further support the potential benefits of the PþF montage and may help mitigate the well documented posterior to anterior shift in activation with aging [36]. Our findings suggest that tDCS may be especially effective in combination with cognitive training paradigms since intensive spatial training increased cortical thickness in the paracentral lobule and precuneus in healthy young but not older adults [37]. The PþF montage had no effect on hippocampal activity for either allocentric or egocentric environments whereas the right caudate was enhanced during egocentric processing. These findings again support a potential hemispheric specialization in caudate functioning (e.g., left mediates general spatial navigation or related cognitive abilities while right mediates egocentric navigation) and suggest that this structure is closely related to functioning of the parietal cortex. Consistent with anodal stimulation over the prefrontal cortex, the PFþ group demonstrated greater activation in the inferior frontal sulcus (bilaterally) relative to the PþF group; an effect that was specific to the allocentric environments. The ventrolateral prefrontal cortex is known to be important in memory encoding [38,39] and is posited to interact with the hippocampal memory system [40]. In this context, it may be especially meaningful that the PFþ group also demonstrated increased activation within the right hippocampus during allocentric navigation as well as increased effective connectivity between the inferior frontal sulcus and this same hippocampal region compared to the PþF group (Supplemental Table 4). We previously demonstrated that mnemonic strategy training increases activation in the inferior frontal sulcus [41] and hippocampus [42]. Other groups have shown that cognitive training can have neuroprotective effects against agerelated hippocampal decline [43]. Such findings again argue in support of the combination of tDCS and cognitive training, with the PFþ montage perhaps being most effective. Overall then, the results of Study 2 support our hypotheses of polarity-specific neurophysiological effects on the brain regions underlying spatial navigation and provide novel evidence that these effects can also be measured using effective connectivity (GCA). Contrary to our hypotheses, however, anodal stimulation over the prefrontal cortex actually facilitated hippocampal activation and effective connectivity during allocentric navigation whereas anodal stimulation over the parietal cortex facilitated caudate activation and effective connectivity. The combination of findings raises the possibility that the PþF montage promotes a “bottom up” approach to navigation that relies more on sensory/ spatial processing whereas the PFþ montage facilitates a “top down” approach. Potential implications The results of Study 2 yielded competing alternatives for treating age- and disease-related decline in allocentric spatial navigation. Previous research has demonstrated reduced posterior neocortical and hippocampal activation during spatial navigation in cognitively healthy older adults [3], which may lead to a compensatory increase

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

in the role of the prefrontal cortex [44]. The current data set suggests that the PþF montage may help mitigate the posterior cortical decline while also enhancing the role of the left caudate. This could ultimately mitigate the age-related decline in these areas or engage additional regions in a compensatory manner. Patients with mild cognitive impairment (MCI) and Alzheimer’s disease demonstrate characteristic decline in posterior cortices and substantial hippocampal impairment [45] beyond the age-related effects discussed above. We previously reported on whole-brain correlations in activation that suggest MCI patients rely on more basic sensory processes, which comes at the expense of deeper levels of processing, while learning information [46]. Thus, the PþF may actually further bias these patients toward this more basic, bottom-up approach. Therefore, the PFþ montage may be preferable in patients with MCI since it could enhance cognitive training programs that utilize top-down processes [37,41,42] and also engage the more posterior regions of the hippocampus that are preserved early in the disease course [47]. These possibilities are serving as the basis for our ongoing follow-up studies. Limitations Despite the novelty of the study and the encouraging findings, future studies should replicate our results using larger samples. Although justified at this stage, the lack of a sham control group and pre-post design preclude definitive discussion about the behavioral impacts of tDCS. We consider it highly unlikely that the lack of a sham group affected the fMRI results of Study 2 for several reasons. First, participants were blinded to their stimulation condition and each member of the study team performed specific duties without knowing the behavioral performances, fMRI results, or stimulation grouping. Second, we typically found a double dissociation in BOLD signal wherein anodal stimulation enhanced activation (and effective connectivity) whereas cathodal stimulation suppressed it. It is difficult to imagine how participants could have consciously modulated activation/connectivity in this manner even if they had known their stimulation montage (which they did not). Third, the mean BOLD signal from the non-stimulated participants in Study 1 generally fell between that of the groups from Study 2 e again arguing against any type of bias. The functional significance of these fMRI-based changes is unclear at this point given the lack of behavioral change. However, we are addressing these limitations in our ongoing studies by using a within-subject design wherein healthy young and older participants receive anodal, cathodal, and sham stimulation over the parietal cortex (counterbalanced across 3 separate sessions) and their memory is tested using a more sensitive behavioral paradigm (novel stimuli are used during each session). These modifications should yield more definitive evidence supporting the benefits of tDCS for spatial navigation. Conclusions This two-part study demonstrated the utility of fMRI-based targeting for tDCS. Our results supported the neuromodulatory effects of tDCS by revealing polarity-dependent changes in both the magnitude of BOLD signal and patterns of effective connectivity, measured via GCA. We posit that this multi-method approach will be useful for identifying stimulation sites that more effectively target age- and disease-related changes in the brain, thereby ameliorating the associated cognitive dysfunction. Acknowledgments The contents of this manuscript do not represent the views of the Department of Veterans Affairs or the United States

323

Government. We wish to thank Ms. Kimberly Lang for her assistance creating the experimental environments and Dr. Simon Lacey for his comments on earlier versions of the manuscript. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.brs.2013.12.006. References [1] Bird CM, Burgess N. The hippocampus and memory: insights from spatial processing. Nat Rev Neurosci 2008;9(3):182e94. [2] Moffat SM. Aging and spatial navigation: what do we know and where do we go? Neuropsychol Rev 2009;19:478e89. [3] Moffat SD, Elkins W, Resnick SM. Age differences in the neural systems supporting human allocentric spatial navigation. Neurobiol Aging 2006;27(7):965e72. [4] Poldrack RA, Packard MG. Competition among multiple memory systems: converging evidence from animal and human brain studies. Neuropsychologia 2003;41(3):245e51. [5] Wolbers T, Hegarty M. What determines our navigational abilities? Trends Cogn Sci 2010;14(3):138e46. [6] Jerde T, Curtis C. Maps of space in human frontoparietal cortex. J Physiol Paris 2013;107(6):510e6. [7] Moffat SD, Zonderman AB, Resnick SM. Age differences in spatial memory in a virtual environment navigation task. Neurobiol Aging 2001;22(5):787e96. [8] Rodgers KM, Sindone JA, Moffat SD. Effects of age on navigation strategy. Neurobiol Aging 2012;33. http://dx.doi.org/10.1016/j.neurobiolaging.2010.07.021. [9] Gazova I, Vlcek K, Laczo J, Nedelska Z, Hyncicova E, Mokrisova I, et al. Spatial navigation-a unique window into physiological and pathological aging. Front Aging Neurosci 2012;4. [10] Lithfous S, Dufour A, Despres O. Spatial navigation in normal aging and the prodromal stage of Alzheimer’s disease: insights from imaging and behavioral studies. Ageing Res Rev 2013;12(1):201e13. [11] Nitsche MA, Paulus W. Transcranial direct current stimulation e update 2011. Restor Neurol Neurosci 2011;29:463e92. [12] Medina J, Beauvais J, Datta A, Bikson M, Coslett HB, Hamilton RH. Transcranial direct current stimulation accelerates allocentric target detection. Brain Stimul 2013;6:433e9. [13] Trojano L, Conson M, Maffei R, Grossi D. Categorical and coordinate spatial processing in the imagery domain investigated by rTMS. Neuropsychologia 2006;46(2):399e408. [14] Sehm B, Schafer A, Kipping J, Margulies D, Conde V, Taubert M, et al. Dynamic modulation of intrinsic functional connectivity by transcranial direct current stimulation. J Neurophysiol 2012;108(12):3253e63. [15] Stagg C, Lin RL, Mezue M, Segerdahl A, Kong YZ, Xie JY, et al. Widespread modulation of cerebral perfusion induced during and after transcranial direct current stimulation applied to the left dorsolateral prefrontal cortex. J Neurosci 2013;33(28):11425e31. [16] Ayaz H, Allen SL, Platek SM, Onarai B. Maze Suite 1.0: a complete set of tools to prepare, present, and analyze navigational and spatial cognitive neuroscience experiments. Behav Res Methods 2008;40(1):353e9. [17] Bohbot VD, Lerch J, Thorndycraft B, Iaria G, Zijdenbos AP. Gray matter differences correlate with spontaneous strategies in human virtual navigation task. J Neurosci 2007;27(38):10078e83. [18] Etchamendy N, Bohbot VD. Spontaneous navigational strategies and performance in the virtual town. Hippocampus 2007;17(8):595e9. [19] Iaria G, Petrides M, Dagher A, Pike B, Bohbot VD. Cognitive strategies dependent on the hippocampus and caudate nucleus in human navigation: variability and change with practice. J Neurosci 2003;23(13):5945e52. [20] Talairach JT, Tournoux P. Co-planar stereotaxic atlas of the human brain. New York: Thieme Medical Publishers; 1988. [21] Forman SD, Cohen JD, Fitzgerald M, Eddy WF, Mintun MA, Noll DC. Improved assessment of significant activation in functional magnetic resonance imaging (fMRI) e use of a cluster size threshold. Magn Reson Med 1995;33:636e47. [22] Sjolinder M, Hook K, Nilsson LG, Andersson G. Age differences and the acquisition of spatial knowledge in a three dimensional environment: evaluating the use of an overview map as a navigation aid international. J Hum Comput 2005;63:537e64. [23] Moffat SD, Resnick SM. Effects of age on virtual environment place navigation and allocentric cognitive mapping. Behav Neurosci 2002;116(5):851e9. [24] Spreng RN, Sepulcre J, Turner GR, Stevens WD, Schacter DL. Intrinsic architecture underlying the relations among the default, dorsal attention, and frontoparietal control networks of the human brain. J Cogn Neurosci 2012;25(1):74e86. [25] Curtis CE. Prefrontal and parietal contributions to spatial working memory. Neuroscience 2006;139(1):173e80. [26] Buckner RL, Andrews-Hanna JR, Schacter DL. The brain’s default network e anatomy, function, and relevance to disease. Ann N Y Acad Sci 2008;1124:1e38.

324

B.M. Hampstead et al. / Brain Stimulation 7 (2014) 314e324

[27] Cabeza R, Nyberg L. Imaging cognition II: an empirical review of 275 PET and fMRI studies. J Cogn Neurosci 2000;12(1):1e47. [28] Postma A, Kessels RPC, van Asselen M. How the brain remembers and forgets where things are: the neurocognition of object-location memory. Neurosci Biobehav Rev 2008;32(8):1339e45. [29] Molenberghs P, Sale MV, Mattingley JB. Is there a critical lesion site for unilateral spatial neglect? A meta-analysis using activation likelihood estimation. Front Hum Neurosci 2012;6. [30] Cabeza R, Ciaramelli E, Moscovitch M. Cognitive contributions of the ventral parietal cortex: an integrative theoretical account. Trends Cogn Sci 2012;16(6):338e52. [31] Craig AD. How do you feel e now? The anterior insula and human awareness. Nat Rev Neurosci 2009;10(1):59e70. [32] Alexander GE, DeLong MR, Strick PL. Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annu Rev Neurosci 1986;9:357e81. [33] Brunoni AR, Amadera J, Berbel B, Volz MS, Rizzerio BG, Fregni F. A systematic review on reporting and assessment of adverse effects associated with transcranial direct current stimulation. Int J Neuropsychopharmacol 2011;14:1133e45. [34] Kim H. Differential neural activity in the recognition of old versus new events: an Activation Likelihood Estimation Meta-Analysis. Hum Brain Mapp 2013;34(4):814e36. [35] Risius UM, Staniloiu A, Piefke M, Maderwald S, Schulte FP, Brand M, et al. Retrieval, monitoring, and control processes: a 7 Tesla fMRI approach to memory accuracy. Front Behav Neurosci 2013;7. [36] Davis SW, Dennis NA, Daselaar SM, Fleck MS, Cabeza R. Que PASA? The posterior-anterior shift in aging. Cereb Cortex 2008;18(5):1201e9. [37] Wenger E, Schaefer S, Noack H, Kuhn S, Martensson J, Heinze HJ, et al. Cortical thickness changes following spatial navigation training in adulthood and aging. Neuroimage 2012;59(4):3389e97.

[38] Kuhl BA, Bainbridge WA, Chun MM. Neural reactivation reveals mechanisms for updating memory. J Neurosci 2012;32(10):3453e61. [39] Kohler S, Paus T, Buckner RL, Milner B. Effects of left inferior prefrontal stimulation on episodic memory formation: a two-stage fMRI-rTMS study. J Cogn Neurosci 2004;16(2):178e88. [40] Baddeley A. Working memory: looking back and looking forward. Nat Rev Neurosci 2003;4(10):829e39. [41] Hampstead BM, Stringer AY, Stilla RF, Deshpande G, Hu XP, Moore AB, et al. Activation and effective connectivity changes following explicitmemory training for face-name pairs in patients with mild cognitive impairment: a pilot study. Neurorehabil Neural Repair 2011;25(3): 210e22. [42] Hampstead BM, Stringer AY, Stilla RF, Giddens M, Sathian K. Mnemonic strategy training partially restores hippocampal activity in patients with mild cognitive impairment. Hippocampus 2012;22(8):1652e8. [43] Lovden M, Schaefer S, Noack H, Bodammer NC, Kuhn S, Heinze HJ, et al. Spatial navigation training protects the hippocampus against age-related changes during early and late adulthood. Neurobiol Aging 2012;33(3). [44] Moffat SD, Kennedy KM, Rodrigue KM, Raz N. Extra hippocampal contributions to age differences in human spatial navigation. Cereb Cortex 2007;17(6):1274e82. [45] Jack CR. Alzheimer disease: new concepts on its neurobiology and the clinical role imaging will play. Radiology 2012;263(2):343e60. [46] Hampstead BM, Stringer AY, Stilla RF, Amaraneni A, Sathian K. Where did I put that? Patients with amnestic mild cognitive impairment demonstrate widespread reductions in activity during the encoding of ecologically relevant object-location associations. Neuropsychologia 2011;49(9): 2349e61. [47] Martin SB, Smith CD, Collins HR, Schmitt FA, Gold BT. Evidence that volume of anterior medial temporal lobe is reduced in seniors destined for mild cognitive impairment. Neurobiol Aging 2010;31(7):1099e106.