There’s a ton of super-interesting transcranial magnetic stimulation work coming out these days (e.g., here, here, here, here, here, and here) and much of it pertains to a very particular “paired-pulse” form of TMS (ppTMS). Before diving into the new work, I wanted a basic crash course on what we know (and what we don’t) about how and why ppTMS works.
If you’re not familiar with the basic behavioral effects of TMS, check out this video.
Unfortunately, the crash course I was looking for doesn’t exist. So I’ve made one. Except where noted, the crash course below is largely derived from a fantastic 2008 review by Reis et al in the Journal of Neurophysiology. We’ll get into what ppTMS is, how it seems to work, and the numerous complex and interacting phenomena that have been documented in the ppTMS literature.
The basic idea of TMS is that one can use a magnetic stimulator to induce current within the brain; the difference in ppTMS is, naturally, that this can be done in a pairwise fashion (first a pulse induces current in locus A, and after some number of ms, a pulse to induce current in locus B). Why is this such a big deal? For one thing, you get a causal window into neuronal dynamics that goes far beyond the traditional “loss of function” logic typically used to understand the effects of TMS. For example, by stimulating two areas of the brain in a paired pulse fashion, one can actually induce a form of Hebbian learning (known aspaired associate stimulation). Another example: one can see how the influence of one brain area on another may be dynamic according to task demands (e.g., when performing on Go or NoGo trials in a Go/NoGo task). The kinds of inferences one could, in theory, draw from this type of data are pretty much unparalleled in the human literature.
The problem is that it becomes very difficult to know what TMS, and especially ppTMS, is actually doing to the underlying neuronal dynamics. See here for my previous discussion of TMS effects on neurovascular coupling – in this post I’ll focus on ppTMS.
The classic ppTMS paradigm is a series of paired magnetic pulses provided to either the same region of motor cortex or to homologous sites above the motor cortex on either side of the two cerebral hemispheres. The strength of the pulse is typically determined in proportion to the amount of electrical activity elicited on the muscle controlled by that area of cortex, as determined through electromyography, known as the motor evoked potential (MEP). This “suprathreshold” stimulation is called the test stimulus (TS), and can be preceded by a certain interval with another pulse, which is either also suprathreshold or subthreshold, and delivered either to the same site or to another site. This first stimulus is called the conditioning stimulus (CS).
There are several components to the MEP – which is usually the dependent measure in ppTMS. The first component is a “short latency direct wave” (or D-wave), reflecting the depolarization of the axons induced by TMS. This is followed by “indirect” waves (I-waves), with a period of about 1.5ms, the first of which reflects the synaptic discharge of the stimulated axon; subsequent I-waves are thought to reflect interaction among multiple local circuits. Critically, the type and number of the D- and I-waves elicited by a TS is dependent on the type and location of a preceding CS, as well as the type of task that a subject is engaged in.
There numerous phenomena that have been documented within the ppTMS paradigm:
Intracortical facilitation (ICF): when a subthreshold CS, induced in a posterior-anterior direction, is followed by a suprathreshold TS between 6-23ms later, the amplitude of I-waves is increased relative to cases where only the TS is delivered. Facilitation is stronger with increasing CS intensity.
Short-interval intracortical facilitation (SICF): when a suprathreshold CS is followed by a subthreshold TS, or when both stimuli are roughly equivalent to the motor threshold, at an interval of either 1.1-1.5ms, 2.3-2.9ms, or 4.1-4.4ms, the second I-wave is enhanced in the MEP. Interestingly, SICF is suppressed if the subject had first experienced sensory stimulation of the relevant peripheral nerves.
Short-interval intracortical inhibition (SICI): subthreshold CS followed by suprathreshold TS at intervals of 1-6ms leads to a suppression of I-waves. This effect is thought to actually reflect stimulation of inhibitory interneurons (as opposed to the induction of a refractory state by the CS) because SICI increases with TS intensity.
Long-interval intracortical inhibition (LICI): if both the CS and TS are suprathreshold but separated by intervals of 50-200ms, then the second MEP is reduced relative to the first (i.e., “inhibition”).
Interhemispheric facilitation (IHF): if the CS and TS are given not to the same site but sites on homologous regions of the two hemispheres, the inter-stimulus interval is short (approximately 4-6, but maybe as high as 10ms), the CS intensity is between 60-110% of active motor threshold, and the TS induces current in a posterior-anterior direction, then I3-waves to the TS are enhanced.
Interhemispheric inhibition (IHI): As above, except the inter-stimulus interval is between 6-50ms, and the CS and TS are both suprathreshold, then reductions in MEP are observed. This phenomenon is fairly robust to variations in CS and TS intensity.
These phenomena are fairly complex, seem to strongly interact, and are poorly understood. Moreover, the situations in which these different phenomena are observed may depend on what areas are being stimulated. Above, I’ve concentrated mostly on stimulation of primary motor cortex. Below, we’ll get into interactions among different areas, for example primary and premotor cortical areas.
Primary motor – dorsal premotor interactions. At intervals of 4-20ms, with the CS delivered to dorsal premotor cortex at 90% or 110% of resting motor threshold above either the contralateral or ipsilateral hemisphere as the primary-motor TS, inhibition of the MEP is observed. In contrast, with less intense CS’s (e.g., 80% of resting threshold) delivered 8ms prior to an anterio-posterior oriented TS, the MEP is actually enhanced.
Primary motor – pre-supplementary motor area (pre-SMA) interactions. CS to Pre-SMA with an interval of 6ms reduces the excitability of primary motor cortex by the TS. Interestingly, the opposite effect is observed following 5Hz suprathreshold (110% of active motor threshold) repetitive TMS of the pre-SMA, in which case the excitability of the primary motor cortex is increased.
Primary motor – ventral premotor interactions. A recent paper by Davare, Montague, Olivier, Rothwell & Lemon demonstrates that with a subthreshold CS over PMv, 6-8ms before a suprathreshold TS over m1, the MEP of the to-be-used muscle is increased. This can be contrasted with the reduction in MEP observed during a resting state task with either sub or suprathreshold CS. Here, PMv was the caudal portion of pars opercularis of the inferior frontal gyrus, with a euclidean distance from M1 of about 61mm.
There are numerous other documented interactions, including those involving somatosensory cortex as well as the cerebellum. Interestingly, all of these interactions may be modulated by the subject’s current behavior or task, as described below.
Movement. If a muscle is voluntarily contracted, then suprathreshold TMS to the contralateral hemisphere induces a “silent period” in the ongoing electromyographic activity.
Movement Preparation. Essentially, M1 is more excitable in the time leading up to a movement, but less excitable in the time leading up to a voluntary suppression of movement (e.g., in a Go/NoGo task, thought to be mediated by SICI). Immediately prior to movement onset, IHI is reduced and IHF is enhanced, a pattern that is stronger for the dominant hand.
Relaxation vs. Task performance.. The Davare paper mentioned above was a follow-up to a 2008 study, also from Davare. In the 2008 paper, they demonstrated that the effect of a CS to PMv on the TS over M1 was to reduce the MEP during a resting state, but reversed to increase the MEP during a motor task. The authors concluded that the task induced an activation of task-specific neuronal populations supporting movement, such that the PMv CS preferentially activated the neuronal population supporting good task performance.
Of course, all of this complexity leads to numerous pitfalls in the interpretation of results from ppTMS paradigms. This includes:
Even short-latency effects seem to be multisynaptic in origin. One might assume that IHI is due to a direct inhibitory transcallosal projection, but animal work suggests that transcallosal projections are primarily excitatory. This means that ppTMS effects observed at intervals as short as 8ms should likely be interpreted in terms of excitatory transcallosal projections synapsing onto local excitatory or inhibitory neurons in the opposite hemisphere.
Partial vs. complete depolarization. Increased excitability may reflect the partial depolarization of an axon via the CS, whereas decreased excitability could reflect the complete depolarization (and subsequent refractoriness) of axons via CS. In this way it is problematic to assign an apparent reduction in MEP directly to the CS-induced activation of inhibitory interneurons.
The neural populations affected by ppTMS may be dependent on top-down and bottom-up influences. In the presence of top-down input to a particular neuronal population in a CS-targeted site, one can expect a different effect of ppTMS than in cases where no such top-down input is provided (or is substantially weaker). The same appears to be true of bottom-up input, because the effect of TMS on activity seems dependent on the activity level prior to the magnetic stimulation.
The intensity problem. Even when CS and TS are separated by a constant interval (e.g., 2.5ms), the observed effect depends crucially on their relative intensities: IHI can turn into IHF if the relative intensities are changed. Thus, multiple intensities should be used to determine the bounds on the phenomenon of interest.
The preparation problem. Observed ppTMS effects can depend strongly on how close subjects are to providing a response. Therefore, apparent modulations in ppTMS effects as a function of task or condition might be ascribed to the cognitive demands of the task or condition when in fact they are an epiphenomenon related to the different degree of motor preparation occuring at a particular time in those conditions.
The spread problem. One of the difficulties of any form of TMS is that the spatial resolution of the technique is unknown, and is likely dependent on numerous factors (including technical things like the size of the TMS coil and the intensities used, as well as factors endogenous to the subjects, like what task they are performing). So, if you observe an effect of TMS at different times at different site, you might posit different underlying mechanisms. Alternatively, you might posit that the induced current merely spreads from one site to the other and leads to a variable pattern in the timing of the observed effects due to variable induced current in some crucial locus. Take a gander at this image from this paper:
What you should see is that TMS is not terribly focal at all, such that most of the cortex will experience fully 50% of the stimulation that your “focal” target region will undergo.
Mediating between loss-of-function interpretations and induction-of-function interpretations. Traditionally single-pulse TMS was used as a “virtual-lesion” technique, and changes in behavioral performance resulting from TMS were thought to reflect the effect of disrupting the function of the stimulated region. However, newer TMS studies rely on the assumption that by stimulating various areas, we are actually inducing them to perform their normal function upon stimulation. Clearly, these two interpretations are not mutually compatible, unless bounds can be placed on when one should assume that they are inducing as opposing to disrupting a region’s function. See here for a fascinating discussion of these and additional issues.