Mapping structure-function relationships in the brain

Abstract

Mapping the structural and functional connectivity of the brain is a major focus of systems neuroscience research and will help identify causally important changes in neural circuitry responsible for behavioral dysfunction. Several methods for examining brain activity in humans have been extended to rodent and monkey models where molecular and genetic manipulations exist for linking to human disease. In this review paper, which is part of a special edition focused on bridging brain connectivity information across species and scale, we address mapping brain activity and neural connectivity in rodents using optogenetics in conjunction with either functional magnetic resonance imaging (Opto-fMRI) or optical intrinsic signal imaging (Opto-OISI). We chose to focus on these techniques because they are capable of reporting spontaneous or evoked hemodynamic activity most-closely linked to human neuroimaging studies. We discuss the capabilities and limitations of blood-based imaging methods, usage of optogenetic techniques to map neural systems in rodent models, and other powerful mapping techniques for examining neural connectivity over different spatial and temporal scales. We also discuss implementing strategies for mapping brain connectivity in humans with both basic and clinical applications, and conclude with how cross-species mapping studies can be utilized to influence preclinical imaging studies and clinical practices alike.

Introduction

Representation of function within the brain classically has been understood in terms of focality. For example, expressive language function has been attributed to the left inferior frontal cortex because focal damage to this part of the brain (Broca’s area) commonly results in expressive language deficits(1). However, it is increasingly recognized that performance deficits are best understood in terms of functional systems distributed over multiple parts of the brain(2). These functional systems are efficiently studied using “resting state” functional magnetic resonance imaging (fMRI), i.e., fMRI acquired without imposed tasks(3, 4). The topographies revealed by such analyses are equivalently known as either resting state networks (RSNs)(5), or intrinsic connectivity networks(6). The physiology underlying RSNs remains incompletely understood. While anatomical connectivity (white matter(7–15)) explains many features of functional connectivity(16), attempts to model whole-brain functional connectivity on the basis of connectional anatomy (all known white matter tracts) have been only partially successful(10, 17). Thus, anatomical and functional connectivity are related, but not in a simple one-to-one fashion. We shall shortly introduce a third type of connectivity, effective connectivity, which exhibits its own relations to the other connectivity measures. Studying these relations (brain connectomics) and dissecting the complex neural circuitry responsible for brain function (normal or abnormal) has emerged as a major focus of systems neuroscience research(18).

Classically, anatomic connectivity is studied by ex-vivo tracing of white matter tracts in brain slices(19). Today, structural connectivity can be non-invasively estimated in-vivo using diffusion tensor imaging(8) in combination with computational methods for reconstructing major fiber tracts(7). The basis of resting state functional connectivity (RS-FC) is that spontaneous fluctuations of blood oxygen level dependent (BOLD) signals, measured with either fMRI(20) or optical techniques(21), are temporally correlated within RSNs. RSNs commonly are mapped by extracting signals from regions-of-interest (see below) or computed using spatial independent component analysis(5). Both methods yield reliable and broadly similar results provided that the data are relatively uncorrupted by artifact and acquired over a sufficiently long time(22). The functional significance of RSNs derives from the observation that they topographically correspond to known sensory, motor, and “cognitive” functional systems (23, 24). Compared to task-based measures, RS-FC analyses have provided an efficient method for mapping the whole brain, and can be performed in patients incapable of performing tasks(25). Changes in RS-FC have been reported in a variety of neuro-psychiatric disorders(26) including Alzheimer’s disease(27), depression(28), schizophrenia(29), Parkinson’s disease(30) and stroke(31).

BOLD-fMRI is the dominant modality for measuring RS-FC in humans, and is becoming increasingly more adopted for examining the brain’s structure-function relationship in animals(9, 32, 33), including mice(10, 34). Importantly, whole-brain functional imaging in mice reveals large-scale functional network architecture corresponding to known murine structural connectivity patterns (34, 35). Moreover, basic functional connectional topology is conserved across mice, rats, primates, and humans(36, 37). Thus, bridge measurements can be made across animal models to enrich findings in human populations. Given the power of mice to aid in our understanding of human disease, longitudinal mapping techniques that probe brain circuitry more directly could profoundly illuminate pathology-related changes in brain organization.

Effective connectivity (EC) is distinct from, but related to, both anatomical and functional connectivity. EC measures the influence (direct or indirect) that one brain region exerts on another(38). The crucial distinction between RS-FC vs. EC is that RS-FC characterizes shared spontaneous (ongoing, intrinsic) signals. By definition, pairwise RS-FC is symmetric and uninformative regarding the directionality of propagated signals. Axonal propagation in living animals is physiologically uni-directional (from cell body to axon terminal). But, this directionality cannot be recovered either by histological tract tracing or DTI-based tractography(39). In contrast, EC reports how activity in an identified part of the brain affects other regions. Thus, measures of EC are not necessarily symmetric. In other words, (and neglecting ethical considerations), analyzing RS-FC is analogous to modeling social networks on Facebook by observation of friend relations; studying EC would be analogous to sending provocative material to known Facebook users and analyzing to whom the material spreads.

A frequently used non-invasive mode of studying EC in humans is measuring the influence of task-induced fMRI responses on the BOLD signal in other parts of the brain(40). Transcranial magnetic stimulation (TMS), the other major non-invasive technique for studying EC(41), is discussed at the end of this review. fMRI and TMS aside, the study of EC generally involves invasive techniques, e.g., direct cortical stimulation in patients undergoing surgical evaluation of intractable epilepsy (see(42, for recent reviews). The responses evoked by such stimuli (cortico-cortical evoked potentials; CCEPs) map out functionally connected regions on the surface of the brain. CCEP EC exhibits correspondence with anatomical connectivity as determined by DTI-based tractography(44). Recording electrophysiological responses to injected current has a long history in animal experimentation(11). More recently, the mode of stimulation and/or the means of recording the response have been replaced with more modern techniques such as calcium fluorescence imaging(12) and BOLD fMRI(45). As discussed below, combining optogenetic targeting with wide-field functional neuroimaging represents an even more powerful strategy for examining structure-function relationships in the mammalian brain.

In this review, we examine recently developed methods for EC mapping in animals in combination with optogenetics. We focus on techniques employing hemoglobin as the imaging contrast for linking with human fMRI findings. We use the general term “optogenetic effective connectivity (Opto-EC) mapping” to refer to these strategies for consistency with prior definitions in the literature(38, 46). We examine the spatiotemporal capabilities of hemodynamic mapping, as well as other mesoscopic mapping techniques that probe brain activity more directly. Finally, we discuss implementing strategies for mapping EC in the human brain using TMS for both basic and clinical applications, with attention to how cross-species mapping studies can be utilized to influence preclinical imaging studies and clinical practice.

Examining brain circuits with optogenetics

Optogenetic methods have revolutionized neuroscience by allowing researchers to control genetically-defined cell populations through optical stimulation of light-gated, opsin-expressing neurons(47). The most frequently used protein in optogenetics is channelrhodopsin-2 (ChR2), a non-selective cation channel that opens when illuminated with blue light(48) causing the targeted neuron to depolarize. Light-induced hyperpolarization is also possible though illumination of inhibitory opsins (e.g., archaerhodopsin or halorhodopsin) with yellow light(49). Opsin expression can be achieved through a variety of techniques including viral vector injection for retrograde or anterograde labeling(50), in utero electroporation(51), and transgenic approaches(52). The optogenetic toolbox is constantly evolving(26, 53). Red-shifted opsin variants(54) have been developed to allow for deeper tissue penetration or combinatorial excitation or depolarization of different populations of neurons(55). Other developments include the engineering of opsins that exhibit faster kinetics(56) to evoke ultra-fast firing frequencies in fast-spiking neurons(57), or “step function” opsins that cause long-term hyper- or hypo-excitability with a single light pulse but revert to prestimulus membrane potentials after a second pulse of light(58). Although the vast majority of optogenetic-based circuit mapping studies have been performed in the rodent, use of these techniques has been extended into non-human primates(59, 60).

Several optogenetic strategies have been implemented to examine neural connectivity over a range of temporal and spatial scales. At the micron scale, optogenetic stimulation in conjunction with multiple whole cell recording is able to map synaptic connections of individual cells. Functional connectivity within a microcircuit can be mapped using ChR2-assisted circuit mapping(51). For this method, a single neuron is used as a recording site while multiple regions are stimulated. This method can be used to map local circuitry within a cortical column, for example in somatosensory barrel cortex(61), or over longer distances spanning several millimeters(62). Maps of specific motor representations can be determined using ChR2-assisted, light-based motor mapping(63). Here, the site of photostimulation is controlled using mirrors, while motor outputs are recorded for quantifying limb(64–66) or whisker(67) movements. This technique can be applied longitudinally making it an attractive method for tracking cortical reorganization following injury(65). Systems-level connectivity (mm to cm) can be mapped using optogenetics in conjunction with wide-field mapping of brain activity. Voltage sensitive dye imaging(68, 69), calcium imaging(70, 71), and hemodynamic imaging(72–74) have all been successfully applied to measuring Opto-EC in the brain.

Chemogenetic technologies represent another powerful platform for probing genetically-targeted, neuro-anatomical circuits. The most widely used method, designer receptors exclusively activated by designer drugs (DREADDs), has been successful in mapping the influence of local circuit activity on global functional network organization across multiple species including mice(75), rats(33) and non-human primates(76). Due to space limitations, chemogenetics are not discussed in detail here. The interested reader is directed to(77) for review.

Hemodynamic mapping of brain circuitry

BOLD-fMRI and optical intrinsic signal imaging (OISI) have been used extensively to indirectly examine neural activity in the mammalian brain. Understanding how underlying electrophysiological activity, either in single or multiple cell populations, relates to the local hemodynamic response is essential for understanding the spatiotemporal capabilities of hemodynamic mapping. Please see Supplementary Material for an expanded discussion on using optogenetics to study neurovascular coupling and spatial resolution limits of hemodynamic mapping.

Activity-based mapping of brain circuitry requires examination of stimulus-evoked or spontaneous activity in the brain. We and others have developed methods for mapping local and distant EC using photostimulus-evoked activity with hemodynamic readout in rodents using either Opto-fMRI(73, 74, 78) or Opto-OISI(72, 79–82). For example, in transgenic Thy1-ChR2 mice, photostimulation of left motor cortex (Fig. 1A, blue ROI) results in local hemodynamic responses in oxygenated (HbO), deoxygenated (HbR), and total hemoglobin (HbT) at stimulated site measured with multispectral OISI (Fig. 1B). In the absence of an overt stimulus, intrinsic activity can also be measured (Fig. 1C). While each contrast can be used for mapping, HbO exhibits the largest signal changes, while HbT exhibits the highest contrast-to-noise(72), and is more closely linked to underlying neural activity(83). Single-wavelength OISI takes advantage of the differential absorption profiles of HbO and HbR to either collect measurements where HbO and HbR absorb light equally (so measurements are sensitive to changes in blood volume(13)) or at wavelengths where one contrast dominates (for collecting measurements sensitive to oxygenation(81, 84, 85)).

Figure 1. Hemodynamic mapping of brain activity.

A) Dorsal view of the exposed, intact, mouse skull and regions of interest located in left motor (blue), left barrel (green), and right motor (red) cortex for demonstrating Opto-EC and RS-FC mapping strategies. B) Left plot: In transgenic Thy1-ChR2 mice, photostimulation of left motor cortex (473nm laser light, 0.5 mW, 5 ms pulses delivered at 10Hz for 5 seconds) results in local hemodynamic responses in oxygenated (HbO, red), deoxygenated (HbR, blue), and total hemoglobin (HbT, black). A single contrast is chosen for further examination of brain connectivity. In this example, we choose HbT as it is most closely related to underlying neural activity and exhibits the highest contrast to noise. Right plot: In left motor cortex, photostimulus evoked activity (blue trace) simultaneously evolves with distant hemodynamic activity in left barrel (green trace) and contralateral right motor cortex (red trace). C) Left plot: In the absence of any overt stimulus, spontaneous fluctuations of all 3 contrasts can also be measured in the same region. Right plot: Co-fluctuations in the same 3 regions are also observed under resting-state conditions. Notice, however, that spontaneous activity in left (blue) and right (red) motor cortex are approximately in phase while fluctuations in barrel cortex (green) are distinctly less coherent. D) Approximately 5 seconds after photostimulation, a map of peak HbT activity reveals wide spread activity in cortical regions surrounding the site of stimulation (blue dot). Additionally, activity in barrel cortex as well as evoked responses in the homotopic contralateral hemisphere suggests that these satellite regions are functionally-connected to the site interrogated. E) Zero-lag correlation between the region stimulated and all other brain regions produces an Opto-EC map (based on Thy1 excitatory stimulation) for left motor cortex. While topographically similar to the map of peak response in D, the connections with right motor and left barrel are more pronounced. The time courses in left/right motor and left barrel are all approximately in phase, but of different amplitudes; zero-lag correlation analysis is only sensitive to signal timing and immune to amplitude differences (within noise limits). F) Performing the same zero-lag correlation analysis with spontaneous HbT fluctuations in left motor cortex produces a resting state functional connectivity map for the blue region. Distinct topographical differences exist between the RS-FC map and the Opto-EC map. Unlike the Opto-EC map, the RS-FC map is symmetric about midline, and covers large portions of motor cortex and lateral somatosensory cortex. RS-FC between left motor and left barrel is notably absent. Maps in panels D, E, and F are thresholded at a Fisher z(r) value of 0.3. See reference (72) for more methodological details.

The most straightforward mapping method examines induced activity at a particular time after stimulation. Following photostimulation of left motor cortex, a map of peak activity (Fig. 1D) reveals a response surrounding the site of stimulation. Additionally, activity in left barrel and right motor cortices suggests that these regions are functionally-connected to the site interrogated. We can evaluate these connections using zero-lag correlation between activity in left motor and all other brain regions, to produce a left motor Opto-EC map (Fig. 1E). While similar topographically to the map of peak response, connections between left motor and left barrel are more pronounced. A different picture is observed when examining spontaneous fluctuations (Fig. 1C). Performing the same analysis of intrinsic activity shows a symmetric map covering large portions of motor cortex and lateral somatosensory cortex (Fig. 1F). Implementing these strategies for all sites interrogated yields a set of Opto-EC and RS-FC maps over the cortex (Fig. 2).

Figure 2: Cortical patterns of Opto-EC and RS-FC.

Opto-EC: Maps of Thy1-based, optogenetically defined effective connectivity in 5 mice calculated by zero-lag correlation of the evoked time course at the stimulated site and activity over the rest of the brain during over a 25 second epoch. RS-FC: Resting state functional connectivity maps in 8 mice calculated by zero-lag correlation of spontaneous time courses at the coordinates of stimulated sites and the rest of the brain (20–30 min/mouse) during a separate imaging session. All connectivity images were created using total hemoglobin as contrast. While both maps share common features, the RS-FC maps are both qualitatively and quantitatively different from their Opto-EC counterparts. For example, secondary motor cortex demonstrates ipsilateral EC with barrel cortex that is not present in the RS-FC map, and RS-FC maps are largely bilaterally symmetric. Further details can be found in (72). Data and figures modified from (72).

Patterns of Opto-EC are distinct from RS-FC maps

Several differences between RS-FC and Opto-EC maps are apparent. Stimulation of some regions, e.g., primary and secondary motor, and primary barrel cortex, generates distant responses that are not present in the corresponding RS-FC map (Fig. 2). Other major differences include the degree to which homotopic regions share coordinated activity. Bilaterally-symmetric RS-FC is very well documented in multiple species on the basis of fMRI and optical techniques(9, 13, 14, 20, 21, 86). A profound lack of homotopic EC is observed in barrel, visual, and posterior parietal regions (Fig. 2). These Opto-EC vs. RS-FC differences may provide insights into the nature of spontaneous activity. Synchronous activity in homotopic cortical regions, as reflected in RS-FC maps, may reflect monosynaptic, transcallosal connections or coordination by subcortical structures. RS-FC in the visual cortex of monkeys most likely involves polysynaptic pathways(9), and indirect cortico-cortical structural connectivity predicts interhemispheric RS-FC between human visual cortices(87). Additionally, the thalamus can regulate cortical excitability(88), and may do so selectively so that homotopic regions are excitable at the same time(15, 89, 90).

Opto-EC matches axonal projection connectivity more closely than RS-FC.

Given the differences between Opto-EC and RS-FC patterns, it is reasonable to hypothesize that one method reflects more directly underlying anatomical connectivity. We recently compared Opto-EC to patterns of axonal projection connectivity (APC)(72) collected from the Allen Mouse Brain Connectivity Atlas(91). For most sites evaluated, the similarities between the APC images and the Opto-EC maps were significantly higher compared to the RS-FC and APC maps (Fig. 3A). The highest overlap between Opto-EC and APC occurred in sensorimotor cortices (where the RS-FC patterns overlapped least). Interestingly, hemodynamic measures of Opto-EC within sensorimotor cortices appear to reflect the known input-output organization of these regions (Motor and Barrel regions in Figs. 2, 3A). In the rodent, whisker-based tactile sensation and sensorimotor integration are mediated by a connection loop between somatosensory barrel and motor cortices(92). We observe appreciable connection strength asymmetry between the primary motor cortex (M1) and the lateral barrel region of somatosensory cortex (S1BL) agreeing with previous studies reporting directionality within this system using invasive anatomical tracing(92) and other local(51, 93) and global(64, 69) in vivo ChR2-based methods measuring neural activity more directly. The similarity between Opto-EC and APC suggests that Opto-EC patterns tend to reflect pairwise connectivity between neurons, much like electrical stimulation of the cortex(94) while RS-FC appears to largely reflect polysynaptic pathways.

Figure 3: Global Opto-EC mapping to probe brain anatomy and function.

A) Maps of Thy1-based Opto-EC and RS-FC are compared to axonal projection connectivity (APC) images from the Allen Mouse Brian Connectivity Atlas. APC images were acquired using the source search feature within the Mouse Connectivity Data Portal and projected onto the cortical surface using the cortical map signal viewer, co-registered to OISI data and normalized by maximum fluorescence intensity. Generally, Thy1-based, Opto-EC maps report monosynaptic cortical connectivity structure, but can be collected in awake, behaving animals. See reference (72) for more details. B) Nonlocal mapping of cells defined by location and genetic identity. Top left: Viral injection (AAV5-CaMKIIα::ChR2-EYFP) and optical stimulation in primary motor cortex results in local (Motor, image 1) and distant (thalamus, image 2) BOLD responses. Fluorescence/bright-field images show ChR2–EYFP expression in motor and thalamic regions while confocal images show that expression is limited to axons. Figures modified from (74). C) Hippocampal–cortical pathways contribute to global RS-FC in the rodent. Viral injection of ChR2::CaMKIIα in dorsal dentate gyrus (dDG) excitatory neurons. Arrows below brain image indicate approximate locations of Opto-EC activation maps numbered 1 (posterior) through 8 (anterior). Histology image shows mCherry fluorescence from viral expression in dorsal hippocampus. Brain-wide activation maps of visually-related regions during low-frequency optogenetic stimulation of dDG excitatory neurons. Top row: Averaged activation maps from 1Hz photostimuli delivered to dDG show robust positive BOLD responses detected in bilateral V1, V2, LGN, and SC, as well as Cg. Middle row: High-frequency optogenetic stimulation (40Hz) of dDG excitatory neurons in dorsal hippocampus does not evoke brain-wide cortical and subcortical activation in V1, V2, LGN, and SC. Bottom row: Regions of interest (ROIs): Cg, cingulate cortex; LGN, lateral geniculate nucleus; SC, superior colliculus; V1, primary visual cortex; V2, secondary visual cortex. Figures modified from (90). In all images, white dot with black outline indicates stimulation site.

Global Opto-EC mapping to probe brain anatomy and function

The recent integration of optogenetics with fMRI and OISI has allowed for examining the effects of precise optogenetic manipulations on global brain activity. The first Opto-fMRI study delivered light to motor cortex as well thalamic targets to reveal BOLD responses resulting from local excitatory neural activity(74)(Fig. 3B). Reciprocal stimulation of the anterior thalamic nucleus resulted in bilateral motor responses while stimulation of the posterior thalamic nucleus resulted in ipsilateral somatosensory responses. These findings demonstrate different cortical projections of thalamic nuclei, in line with neural tracing studies, but also a thalamic role in shaping coherent spontaneous activity across hemispheres(15).

Changes in infraslow (<0.1Hz) spontaneous activity (e.g., as reflected in patterns of RS-FC) facilitate and modulate a diverse set of motor, sensory, and cognitive processes. However, the neural basis of RS-FC remains unknown. Most studies have only correlated measures of neural activity with hemodynamics instead of directly probing the effects of modulated, region-specific electrical activity on global RS-FC. With its dense reciprocal APC to and from the cortex, the hippocampus is believed to mediate numerous cognitive functions and sensory processing(95). Further, cortical slow oscillations(1Hz), which are phase-locked with cellular activity in the dentate gyrus(96), resemble the spatiotemporal characteristics of infraslow activity in RSNs. Thus, low-frequency activity in hippocampal–cortical pathways could contribute to global RS-FC to integrate sensory information. Chan and colleagues examined large-scale effects of spatiotemporal-specific propagation of downstream hippocampal activity on global RS-FC(90). In that study, stimulation of excitatory neurons in the dorsal dentate gyrus at low frequency(1Hz), but not high frequency(40Hz), evoked robust cortical and subcortical brain-wide responses (Fig. 3C). Low-frequency stimulation enhanced interhemispheric hippocampal and cortical RS-FC. LFP recordings revealed an increase in slow oscillations in the hippocampus and visual cortex, interhemispheric visual RS-FC, and hippocampal– cortical RS-FC. Together, these results highlight the role of low-frequency activity propagating along the hippocampal–cortical pathway, and its contribution to interhemispheric cortical RS-FC.

Opto-EC mapping can also identify large-scale brain networks corresponding to specific behavioral and electrophysiological markers. For example, the thalamus plays a crucial role in coordinating brain signaling responsible for cognition and normal waking behavior(97). The central thalamus and intralaminar nuclei are postulated to regulate arousal, attention, and goal-oriented behavior(98). Using Opto-fMRI in conjunction with electrophysiology, Liu et al., examined the thalamic-driven global brain networks responsible for switching across brain states(99) with activity corresponding to electrophysiological metrics of arousal (e.g., spindle-like oscillations). In a separate study, stimulation of excitatory cells in intermediate hippocampus caused widespread BOLD signaling at high frequencies and predicted seizure-like after-discharges in EEG recordings and behavioral output(100).

EC Mapping in humans

Non-invasive strategies for mapping brain connectivity in the human brain have primarily employed TMS, which offers a less painful alternative to current injection through the scalp(101). The basic principle underlying TMS is that rapidly switched current though a coil placed over the scalp induces an electric potential around the coil which stimulates the brain. Repetitive TMS (rTMS) refers to pulse train stimulation at frequencies in the range of 1–50Hz. Pulse trains of duration on the order of few seconds often are delivered over minutes with the objective of inducting temporary changes in cortical excitability. Whether rTMS is inhibitory or excitatory has been thought to depend on pulse frequency (<1Hz: inhibitory; >5Hz: excitatory)(102). However the sign of observed effects does not always follow this rule(41, 103). Theta-burst TMS (brief pulse trains [3 pulses at 50Hz], rhythmically delivered at theta frequencies [5Hz]) over tens of minutes, is frequently used to induce plasticity(104).

Basic physiological investigations in animals have established that several seconds of rTMS initially increases cerebral blood flow (CBF) immediately under the coil; the initial response lasts on the order of a minute after which blood flow becomes depressed(105). Similar physiologic “activation” responses have been observed with [¹⁴C]-deoxyglucose autoradiography(106) and measurement of regional CBF using [¹⁵O]H₂O positron emission tomography (PET)(107). In humans, increases in CBF have been observed using arterial spin labeling MRI(108) and near infrared spectroscopy (NIRS)(109). Importantly, these physiological responses do not reflect enhanced neural function. Single-unit discharge under the coil might variably increase or decrease following rTMS(110) while neural responses to natural stimuli generally are disrupted(105). This effect commonly is referred to as a “virtual lesion”(111). For example, occipital rTMS disrupts RS-FC within the visual system and degrades performance on a visual discrimination task(112). A particularly informative example of the virtual lesion effect was observed in an experiment in which 10Hz occipital rTMS impaired Braille reading in lifelong blind individuals, thereby demonstrating that a part of the brain normally used for vision had been reconfigured to analyze haptic (tactile) information(113). A recent meta-analysis of 29 studies confirmed that continuous theta-burst rTMS reliably impairs executive function(114). However, exceptions to this general rule exist. In normal volunteers, 20Hz parietal rTMS (daily 2s bursts every 30s × 20min × 5days) increased FC within the hippocampal-parietal memory system and significantly improved associative memory performance(115).

To study EC, TMS is combined with additional techniques including EEG(116), magnetoencephalography(117), PET(118), NIRS(119), or fMRI(120). The variety of such “multimodal” TMS experiments is thoroughly covered in recent reviews(41, 121). Combining TMS with EEG arguably is methodologically most similar to the above-reviewed Opto-EC studies in rodents. Thus, single-pulse TMS delivered to left motor cortex reliably evokes electrophysiological responses that evolve over ~400ms to involve functionally-related parts of the brain(122). Most experiments combining fMRI with TMS stimulate with pulse trains (i.e., rTMS) and measure BOLD responses both under the coil as well as in functionally-related parts of the brain remote from the stimulated site(123, 124). The response may depend on whether the remote region is engaged by a concurrently administered behavioral task. For example, 10Hz right parietal rTMS (delivered alone) depresses CBF in left primary sensory cortex; however, in the presence of right median nerve shocks, CBF in left S1 is enhanced. Moreover, 10Hz right parietal rTMS enhances detection of near-threshold right median nerve stimuli(125). Similar “state-dependent” modulations of EC between dorsal premotor cortex and motor cortex have been demonstrated with motor tasks(126).

A wide variety of neurologic and psychiatric disorders have implemented TMS therapy(127). The rational basis for the therapeutic effect is specific to condition. Unilateral stroke might destroy the part of the brain controlling motor function on the opposite side of the body. It follows that weakness contralateral to the stroke is attributable to loss of function. However, interference from the intact hemisphere, which often is over-active, also is a factor. Thus, inhibitory(1Hz) TMS over the intact hemisphere as well as excitatory(10Hz) stimulation over the lesioned hemisphere both may improve motor function of the paretic hand(128). In Parkinson disease, the physiological abnormality responsible for rigidity and bradykinesia (inability to initiate action) is persistent, abnormally hypersynchronous beta(15–30Hz) activity throughout the motor system (cortex, basal ganglia, thalamus). High frequency stimulation of the subthalamic nucleus via implanted electrodes often restores mobility by disrupting the pathological hypersynchrony(129). High frequency rTMS of motor cortex offers a non-invasive (but less permanent) alternative that could work via the same mechanism(130).

From Man to Mouse to Man: Iterating between preclinical and clinical findings

Depression is the condition most often treated by rTMS. Converging neuroanatomical and post-mortem histological evidence points to VMPFC as a specific locus of dysfunction in depression(131). rTMS offers a non-invasive alternative to DBS, but the challenge is that VMPFC is a deep structure not directly accessible to TMS. However, experimental evidence indicates that rTMS can be optimized by targeting a specific region in dorsolateral prefrontal cortex (DLPFC) known to be maximally anticorrelated with VMPFC in normal individuals(132). This result represents a case in which RS-FC mapping in normal subjects has informed the clinical therapy of depression. Clinical efficacy may be further improved by targeting the specific DLPFC region maximally anticorrelated with VMPFC individually identified in each patient using information gleaned from RS-FC mapping(133).

Further dissection of PFC circuitry via EC mapping in animal models could also help improve the specificity of targeted therapy. Depression is linked to several neural pathways, and distinct neuromodulatory systems and receptors(53). In animal models of depression, the medial prefrontal cortex (mPFC) is the most typically examined, and has been dissected anatomically and functionally using optogenetics. For example, selective activation or suppression of VMPFC circuits in rodents has established the causal relationship between mPFC connectivity and depression-like behaviors(134, 135). Mesolimbic dopamine circuitry is also widely studied in models of depression(136). A key component of this circuitry are dopamine neurons in the ventral tegmental area (VTA) that project to the limbic regions, such as the nucleus accumbens, amygdala and the PFC(137). This mesolimbic system is thought to be functionally-distinct from dopamine-innervated basal ganglia regions. However, Lohani and colleagues(138) demonstrate that phasic stimulation of VTA dopamine neurons increased BOLD signaling in known VTA-innervated regions, but also in regions that receive little or no VTA dopaminergic input (thalamus, hippocampus and dorsal striatum)(138). Thus, Opto-EC mapping is capable of identifying novel circuits within RSNs for probing questions about causal changes in circuit communication responsible for behavioral dysfunction. Compared to other diseases the myriad symptoms that define clinical depression cannot all be accurately modeled in animals. One strategy for linking clinical findings in patients with experimental work (and vice versa) would be to focus on human behavioral symptoms having measurable correlates in animal models.

Summary and Conclusion

We have reviewed several related but distinct forms of mapping brain connectivity: anatomical connectivity (white matter tracts), effective connectivity (how activity in one part of brain affects function elsewhere) and “functional” connectivity (correlated infraslow spontaneous activity) (Figure 4). Owing to space limitations, we have omitted discussion of many important topics, for example, that infraslow spontaneous activity, as imaged with hemoglobin-dependent signals, propagates over centimeter-scale distances at velocities far slower than axonal propagation of action potentials(139). The physiology underlying this phenomenon currently is not understood.

Figure 4:

Tools for mapping brain connectomics. Mapping anatomical, or white matter, connections of the brain can be performed with ex vivo histological tract tracing(10, 19), or in vivo using diffusion tensor imaging (7, 8, 39) or manganese-enhanced MRI(150). Mapping the functional organization of the brain can be performed in vivo by invoking ecological (physiologically normal) stimuli as functional localizers (38, 40), or via synthetic (non-physiological) stimuli via invasive (12, 42, 45), non-invasive(41, 101, 102, 104), or genetics-based (33, 48, 50, 52, 72–76) strategies. In the absence of any overt task, ongoing spontaneous or intrinsic brain activity can be evaluated through electrophysiology(15, 62, 92), ROI-based (14, 21, 23, 25, 27, 86) or ICA-based(5, 24, 27) analyses, as well as through evaluating the relative timing differences in regional activity propagation (temporal lag analysis) (86, 139). Abbreviations: Mn²⁺-MRI: manganese enhanced magnetic resonance imaging; DBS: deep brain stimulation; TMS: transcranial magnetic stimulation; TDCS: transcranial direct current stimulation; TENS: transcutaneous electrical nerve stimulation; ROI: region of interest; Ca²⁺: Calcium; ICA independent component analysis; BOLD: Blood oxygen level dependent; FMRI: Functional magnetic resonance imaging.

Much of the presently-reviewed findings were obtained in the mouse but very likely apply to humans as basic principles of brain connectomics are conserved across a wide range of species(37). We have focused on recently-developed methods capable of specifically stimulating excitatory(69, 72, 73, or inhibitory(74, 79) neurons. New technologies for examining responses include calcium signaling within neurons(139–142), or astrocytes(143). Advances in genetically-encoded, fluorescent protein voltage indicators continues to progress(144, 145); this technology may eliminate some of the problems associated with voltage-sensitive dye signaling (indiscriminant labeling, cytotoxicity) and could allow for wide-field, systems-level all-optical electrophysiology(146, 147). These methods, perhaps in combination with red-shifted opsins, will undoubtedly reveal temporal processes and other aspects of neural interactions important for sustaining normal brain function.

We concluded this review with brief overview of human rTMS with an emphasis on treatment of depression. As discussed above, multiple lines of evidence derived from animal experimentation support rTMS as a treatment for depression. However, notwithstanding its widespread use, the efficacy of rTMS as a treatment for depression has not been established(148). A recent meta-analysis of 61 published studies suggests that a significant fraction of positive clinical responses is attributable the placebo effect(149). Thus, there remains a gap between the rich body of accumulated connectomics knowledge and practical clinical application. This is not cause for discouragement as basic science eventually leads to practical benefits.