NEUROPHYSIOLOGICAL TOOLS

The EEG, first described in the U.S. by Herbert Jasper, PhD (Bradley Hospital) in 1935, is a neurophysiological technique for the noninvasive measurement of brain electrical activity as inferred from superficial signals detected through electrodes on one’s head (Jasper & Carmichael, Reference Jasper and Carmichael1935). Oscillations detected by EEG are rhythmic patterns or waves that reflect the culmination of neuronal activity occurring in response to a direct event (e.g., cognitive processes in response to demands of a WM task) or at rest (Mathalon & Sohal, Reference Mathalon and Sohal2015). Oscillations are broken down into temporal frequency bands, including delta (1–3 Hz), theta (4–8 Hz), alpha (9–12 Hz), beta (13–30 Hz), gamma (31–80 Hz), fast (81–200 Hz), and superfast (201–600 Hz), and can further be characterized by cycles and amplitudes (Mathalon & Sohal, Reference Mathalon and Sohal2015). The EEG is the most often used method to measure oscillations given its excellent temporal resolution. The MEG measures magnetic (instead of electrical) signals, and while less sensitive to activity in gyri, its heightened sensitivity to sulci activity results in stronger spatial resolution. Thus, MEG with its combination of excellent temporal and spatial resolution is another technique that can also examine brain oscillatory activity (Luck, Reference Luck2005).

The TMS, first described by Anthony Barker, PhD in 1985, is a method for noninvasively stimulating human cortex with an electromagnetic device (Barker, Jalinous, & Freeston, Reference Barker, Jalinous and Freeston1985). A TMS stimulator discharges electrical current through a coil of conducting wire, in turn emitting a magnetic field from the coil that is capable, when placed on one’s head, of inducing neuronal depolarization and electrical current in the underlying cortical tissue (Barker et al., Reference Barker, Jalinous and Freeston1985; George, Lisanby, & Sackeim, Reference George, Lisanby and Sackeim1999; Lisanby, Luber, Perera, & Sackeim, Reference Lisanby, Luber, Perera and Sackeim2000). The TMS can be utilized as an experimental probe to transiently modify the activity of a targeted cortical region beneath the coil, or in a therapeutic manner [repetitive TMS (rTMS)], through repeated applications of stimulation pulse “trains” over time (Lisanby et al., Reference Lisanby, Luber, Perera and Sackeim2000). Neurophysiological research using EEG and TMS reflects a novel opportunity to specifically target and then modulate selected neural oscillations with a range of standard pulse frequencies or theta burst stimulation.

OSCILLATIONS–COGNITION

Many excellent reviews have already been written on oscillations and cognition. This is particularly true for the topic of gamma activity underlying cognition and WM, as well as theta and theta–gamma coupling (TGC) in WM (Basar, Reference Basar2013; Hsieh & Ranganath, Reference Hsieh and Ranganath2014; Jensen, Kaiser, & Lachaux, Reference Jensen, Kaiser and Lachaux2007; Kahana, Reference Kahana2006; Kaiser, Ripper, Birbaumer, & Werner, Reference Kaiser, Ripper, Birbaumer and Werner2003; Lisman, Reference Lisman2010; Mably & Colgin, Reference Mably and Colgin2018; Mathalon & Sohal, Reference Mathalon and Sohal2015; Miller, Lundqvist, & Bastos, Reference Miller, Lundqvist and Bastos2018; Roux & Uhlhaas, Reference Roux and Uhlhaas2014; Senkowski & Gallinat, Reference Senkowski and Gallinat2015). Regarding oscillatory dynamics, slow oscillations (e.g., theta) are able to recruit a large volume across greater area of cortex due to the longer time windows, while fast-/high-frequency oscillations (e.g., beta and gamma), and their shorter time windows, are limited to small volume and local cortex (Boudewyn & Carter, Reference Boudewyn and Carter2018; Buzsaki & Watson, Reference Buzsaki and Watson2012; Cavanagh & Frank, Reference Cavanagh and Frank2014). The phase of slow oscillations often then modulates the power of faster oscillations, which reflects the “cross-frequency coupling” that is a core foundation of neuronal communication and brain rhythm organization (Buzsaki & Watson, Reference Buzsaki and Watson2012). This general structure of oscillatory dynamics is particularly relevant for WM in the form of TGC, as theta oscillations (particularly midline, but originating in anterior cingulate cortex, hippocampus) reflect the need for cognitive control, while gamma (particularly frontal) reflects the execution of actions (Cavanagh & Frank, Reference Cavanagh and Frank2014; Roux & Uhlhaas, Reference Roux and Uhlhaas2014). Importantly, recent work has identified that it is gamma bursts, not sustained activity (as previously thought), that underlie complex cognitive functions such as WM (Miller et al., Reference Miller, Lundqvist and Bastos2018).

ANIMAL RESEARCH ON OSCILLATIONS AND WORKING MEMORY

Animal research has provided invaluable findings regarding the underlying role of neural oscillations in WM (seven animal model studies identified). The WM tasks utilized ranged by species/labs (rodents: delayed T or Y maze tasks and monkeys: delayed match-to-sample tasks), although all had visual-spatial WM demands (and largely involved maintenance stage of WM). Yamamoto and colleagues (Reference Yamamoto, Suh, Takeuchi and Tonegawa2014) found that transient bursts of high-gamma synchronization during memory retention were associated with correct (but not incorrect) responses during a spatial WM task in mice. As an example of neuromodulation techniques, the subsequent optogenetic (i.e., using light to modulate/control cells) inhibition of the medial entorhinal cortex – hippocampus circuit reduced the high-gamma synchrony and lowered test performance accuracy in the mice. In other animal work, brief bursts of gamma power and synchronization in prefrontal cortex (PFC) emerged or increased with WM demands at the initial encoding and retention/maintenance stages of WM (Bai, Xia, Liu, & Tian, Reference Bai, Xia, Liu and Tian2016; Lundqvist, Herman, Warden, Brincat, & Miller, Reference Lundqvist, Herman, Warden, Brincat and Miller2018; Lundqvist et al., Reference Lundqvist, Rose, Herman, Brincat, Buschman and Miller2016; Yamamoto, Suh, Takeuchi, & Tonegawa, Reference Yamamoto, Suh, Takeuchi and Tonegawa2014). This power/synchronization subsequently decreased upon the removal of WM demands. Importantly, the gamma bursts were identified during correct, but not incorrect, trials, and deviations in this oscillatory pattern were associated with incorrect responses (Bai et al., Reference Bai, Xia, Liu and Tian2016; Lundqvist et al., Reference Lundqvist, Herman, Warden, Brincat and Miller2018; Yamamoto et al., Reference Yamamoto, Suh, Takeuchi and Tonegawa2014). Of particular clinical relevance, decreased gamma activity/synchronization has been identified in both amyloid-B peptide-injected rats and mice with poor WM (Bai et al., Reference Bai, Xia, Liu and Tian2016; Zhang et al., Reference Zhang, Zhang, Yu, Yang, Li and Qian2017).

Although gamma appears to be the most prominent oscillatory band underlying WM, theta and beta also play a role. Beta and gamma bursts have been found to be inversely correlated, with increased gamma bursts associated with decreased beta bursts (Lundqvist et al., Reference Lundqvist, Herman, Warden, Brincat and Miller2018). Beta activity appears to reflect a default network state that may regulate and initiate gamma activity at the onset of WM demands, that is, when there is a need to shift out of the resting state or default mode (Lundqvist et al., Reference Lundqvist, Herman, Warden, Brincat and Miller2018; Lundqvist et al., Reference Lundqvist, Rose, Herman, Brincat, Buschman and Miller2016). In contrast, theta activity has been positively correlated with gamma and appears to modulate gamma activity, particularly within periods of TGC (Li, Bai, Liu, Yi, & Tian, Reference Li, Bai, Liu, Yi and Tian2012; Tamura, Spellman, Rosen, Gogos, & Gordon, Reference Tamura, Spellman, Rosen, Gogos and Gordon2017; Zhang et al., Reference Zhang, Zhang, Yu, Yang, Li and Qian2017). The TGC is particularly linked to WM, with increased levels of TGC associated with improved or successful WM performance (Li et al., Reference Li, Bai, Liu, Yi and Tian2012; Tamura et al., Reference Tamura, Spellman, Rosen, Gogos and Gordon2017). Further confirming the TGC–WM association, the experimental manipulation of WM demands, such as implementing longer WM maintenance/retention periods or optogenetically disrupting neural circuits, is associated with increased TGC between hippocampus and medial PFC (Tamura et al., Reference Tamura, Spellman, Rosen, Gogos and Gordon2017). Despite the inherent limitations in translating animal findings into clinical implications, this work establishes the role of theta/gamma oscillations within prefrontal–hippocampal regions in the manifestation of WM.

OSCILLATIONS AND WORKING MEMORY

We identified 12 studies, primarily utilizing MEG, that investigated neural oscillatory patterns in healthy participants (age range across studies: 18–37 years) during WM demands as measured via delayed match-to-sample, letter-memory manipulation, and n-back tasks (i.e., largely visual–spatial stimuli with maintenance and/or manipulation demands). Generally (high-) gamma oscillations, theta oscillations, and TGC increased in response to the onset and gradual increase of WM demands in central/occipital, frontal, and temporal/parietal regions (Holz, Glennon, Prendergast, & Sauseng, Reference Holz, Glennon, Prendergast and Sauseng2010; Honkanen, Rouhinen, Wang, Palva, & Palva, Reference Honkanen, Rouhinen, Wang, Palva and Palva2015; Itthipuripat, Wessel, & Aron, Reference Itthipuripat, Wessel and Aron2013; Jokisch & Jensen, Reference Jokisch and Jensen2007; Kaiser, Rieder, Abel, Peters, & Bledowski, Reference Kaiser, Rieder, Abel, Peters and Bledowski2017; Kaiser et al., Reference Kaiser, Ripper, Birbaumer and Werner2003; Lutzenberger, Ripper, Busse, Birbaumer, & Kaiser, Reference Lutzenberger, Ripper, Busse, Birbaumer and Kaiser2002; Park, Jhung, Lee, & An, Reference Park, Jhung, Lee and An2013; Rajji et al., Reference Rajji, Zomorrodi, Barr, Blumberger, Mulsant and Daskalakis2017; Roux, Wibral, Mohr, Singer, & Uhlhaas, Reference Roux, Wibral, Mohr, Singer and Uhlhaas2012; Senkowski, Schneider, Tandler, & Engel, Reference Senkowski, Schneider, Tandler and Engel2009). Subsequently, TGC and gamma oscillations were positively associated with WM performance (Park et al., Reference Park, Jhung, Lee and An2013; Roux et al., Reference Roux, Wibral, Mohr, Singer and Uhlhaas2012). Inconsistent with other findings, one study found that beta activity was positively correlated with WM demands, while gamma activity was negatively correlated with WM demands (von Lautz et al., Reference von Lautz, Herding, Ludwig, Nierhaus, Maess, Villringer and Blankenburg2017). Across studies, the authors concluded that the frontal/prefrontal activation may reflect the actual WM demands (i.e., maintaining or manipulating information), while frontotemporal and occipital activation may reflect specific type (i.e., visual) of stimuli/information that must be maintained/manipulated (Honkanen et al., Reference Honkanen, Rouhinen, Wang, Palva and Palva2015; Kaiser et al., Reference Kaiser, Ripper, Birbaumer and Werner2003).

Utilizing EEG (none utilized MEG), we identified 10 studies, all of which examined oscillatory patterns in clinical samples (typically with healthy controls) during WM demands as measured via n-back, Sternberg spatial WM, mental arithmetic, delayed match-to-sample, and Simon tasks. These include samples of schizophrenia/psychosis, epilepsy, attention deficit hyperactivity disorder (ADHD), Alzheimer’s disease/mild cognitive impairment, and those characterized by low WM performance [across studies age range: 18–49 years, excluding ADHD (8–14 years) and Alzheimer’s (60+ years)]. Abnormal gamma oscillations and/or abnormal TGC were identified in the clinical conditions across all studies, primarily in frontal brain regions, although some variability in the direction of these findings was noted. Gamma oscillations and TGC were found to increase as WM demands increased, particularly in healthy controls (Basar-Eroglu et al., Reference Basar-Eroglu, Brand, Hildebrandt, Karolina Kedzior, Mathes and Schmiedt2007; Chaieb et al., Reference Chaieb, Leszczynski, Axmacher, Hohne, Elger and Fell2015). Across clinical samples, TGC and gamma activity were positively correlated with WM performance, although this was more consistently identified in healthy controls (Barr et al., Reference Barr, Rajji, Zomorrodi, Radhu, George, Blumberger and Daskalakis2017; Chaieb et al., Reference Chaieb, Leszczynski, Axmacher, Hohne, Elger and Fell2015; Chen et al., Reference Chen, Stanford, Mao, Abi-Dargham, Shungu, Lisanby, Schroeder and Kegeles2014; Goodman et al., Reference Goodman, Kumar, Zomorrodi, Ghazala, Cheam, Barr, Daskalakis, Blumberger, Fischer, Flint, Mah, Herrmann, Bowie, Mulsant and Rajji2018; Kim et al., Reference Kim, Kim, Lee, Na, Kee, Min, Han, Kim and Lee2016). Among studies in schizophrenia, increased or excessive WM-related gamma oscillations were found in two studies, yet decreased or lowered WM-related gamma oscillations and TGC were found in two other studies (Barr et al., Reference Barr, Farzan, Tran, Chen, Fitzgerald and Daskalakis2010; Barr et al., Reference Barr, Rajji, Zomorrodi, Radhu, George, Blumberger and Daskalakis2017; Basar-Eroglu et al., Reference Basar-Eroglu, Brand, Hildebrandt, Karolina Kedzior, Mathes and Schmiedt2007; Chen et al., Reference Chen, Stanford, Mao, Abi-Dargham, Shungu, Lisanby, Schroeder and Kegeles2014). As expected, decreased WM-related TGC was associated with Alzheimer’s disease and ADHD (Goodman et al., Reference Goodman, Kumar, Zomorrodi, Ghazala, Cheam, Barr, Daskalakis, Blumberger, Fischer, Flint, Mah, Herrmann, Bowie, Mulsant and Rajji2018; Kim et al., Reference Kim, Kim, Lee, Na, Kee, Min, Han, Kim and Lee2016). However, increased WM-related gamma oscillations (albeit in occipital lobe) were identified in participants with ADHD in another study (Prehn-Kristensen, Wiesner, & Baving, Reference Prehn-Kristensen, Wiesner and Baving2015). Participants with first-episode psychosis displayed increased gamma variability and elevated time-dependent gamma activity changes, which may help explain the findings previously interpreted as relatively increased or decreased may represent variability or other types of abnormality (Missonnier, Curtis, Ventura, Herrmann, & Merlo, Reference Missonnier, Curtis, Ventura, Herrmann and Merlo2017).

In sum, theta and gamma oscillations are closely linked to WM demands and performance in healthy controls, particularly in frontal/prefrontal regions. However, the research in clinical samples is quite variable. It appears pathological patterns can at times manifest as too much or too little oscillatory activity, although more generally, deviation from typical patterns is associated with worse WM performance. This suggests that targeting these oscillatory dynamics with novel interventions may subsequently remediate WM deficits.

OSCILLATIONS AND TMS

Three identified studies examined the modulatory effect of a single session of rTMS on neural oscillations during WM demands. In these studies, TMS was administered at the dorsolateral prefrontal cortex (DLPFC) with either 20 Hz or intermittent theta burst stimulation protocols. Compared to sham stimulation, all three studies found that rTMS leads to increased gamma power during WM demands in healthy participants (Barr et al., Reference Barr, Farzan, Arenovich, Chen, Fitzgerald and Daskalakis2011; Barr et al., Reference Barr, Farzan, Rusjan, Chen, Fitzgerald and Daskalakis2009; Hoy et al., Reference Hoy, Bailey, Michael, Fitzgibbon, Rogasch, Saeki and Fitzgerald2016). Hoy and colleagues (Reference Hoy, Bailey, Michael, Fitzgibbon, Rogasch, Saeki and Fitzgerald2016) found this effect led to a large effect on WM test performance. In one study, participants with schizophrenia had excessive gamma power at baseline (compared to controls), and interestingly, rTMS in these patients led to decreased gamma power (Barr et al., Reference Barr, Farzan, Arenovich, Chen, Fitzgerald and Daskalakis2011). While this gamma modulatory effect was primarily detected in frontal brain regions, one study identified the effect on gamma power in parietal regions along with an effect on frontoparietal theta synchronization (Barr et al., Reference Barr, Farzan, Arenovich, Chen, Fitzgerald and Daskalakis2011; Barr et al., Reference Barr, Farzan, Rusjan, Chen, Fitzgerald and Daskalakis2009; Hoy et al., Reference Hoy, Bailey, Michael, Fitzgibbon, Rogasch, Saeki and Fitzgerald2016). In another study, a single application of 5 Hz TMS to the intraparietal sulcus in healthy participants resulted in WM improvement, with such performance improvement directly associated with the degree of TMS-induced theta entrainment (Albouy, Weiss, Baillet, & Zatorre, Reference Albouy, Weiss, Baillet and Zatorre2017).

Three additional studies examined the effect of rTMS on resting-state oscillations after 10–20 sessions of open-label rTMS treatments delivered to DLPFC (i.e., depression treatment studies). While these studies did not specifically examine WM, resting-state gamma and theta activity changes were found. Significant posttreatment increases were seen in resting-state activity, specifically frontal gamma power, central theta power, and centro-temporal TGC after 10 sessions (Noda et al., Reference Noda, Nakamura, Saeki, Inoue, Iwanari and Kasai2013; Noda et al., Reference Noda, Zomorrodi, Saeki, Rajji, Blumberger, Daskalakis and Nakamura2017). Regarding the association of oscillatory change to clinical variables, posttreatment theta power and TGC at rest were positively correlated with cognitive flexibility performance, while gamma power was positively correlated with change in depressive symptomatology (Noda et al., Reference Noda, Nakamura, Saeki, Inoue, Iwanari and Kasai2013; Noda et al., Reference Noda, Zomorrodi, Saeki, Rajji, Blumberger, Daskalakis and Nakamura2017). In another study which used MEG, 20 sessions of rTMS delivered to left DLPFC resulted in enhanced gamma power at DLPFC and decreased gamma connectivity between the DLPFC and subgenual anterior cingulate cortex (Pathak, Salami, Baillet, Li, & Butson, Reference Pathak, Salami, Baillet, Li and Butson2016). The power and connectivity changes also correlated with depressive symptom improvement (Pathak et al., Reference Pathak, Salami, Baillet, Li and Butson2016).

In sum, rTMS can directly target and modulate frontoparietal theta and gamma oscillations, particularly during WM demands, with such oscillatory changes associated to improved WM performance (as well as clinical variables such as depression).