Understanding the complex mechanisms underlying brain function and dysfunction has been a central goal of neuroscience research. While fMRI offers unparalleled insights into the brain’s spatial and temporal dynamics, its full potential is unlocked when combined with complementary neuroscience technologies. Performing fMRI in rodent models offers a platform to investigate the topological and functional basis of behavior and neurological diseases. This review focuses on integrating fMRI with cutting-edge neuromodulation methodologies, including electrical DBS, optogenetics, and chemogenetics, which offer time-locked and/or cell-type-specific controls of information flow within the brain.

DBS modulates neural circuits, with its versatility in parameter selection, offering a direct bridge between animal studies and human clinical applications (1). Optogenetics provides an unmatched level of specificity, manipulating neurons with light to parse out neural circuitry dynamics and their consequent behavioral influences (2). Chemogenetics allows for non-invasive and remote cellular modulation through receptor signaling, enabling researchers to study the sustained effects of modulation (3). Each of these techniques contributes uniquely to interrogating distinct aspects of brain signaling and has the potential to precisely control behaviors or provide interventions for neurological and psychiatric conditions.

fMRI has revolutionized the field of neuroscience, offering a non-invasive window into the complex dynamics of brain activity patterns in health and disease across species. Though fMRI experiments can be conducted using several specialized acquisition techniques that measure cerebral blood flow, cerebral blood volume, and cerebral metabolic rate of oxygen (CMRO2), the technique primarily relies on the Blood Oxygen Level Dependent (BOLD) contrast (4). The BOLD contrast is a measure of magnetic field susceptibility effects based on the changes in the amounts of oxygenated and deoxygenated blood across the brain, serving as an indirect indicator of neuronal activity. This technique has evolved into various research strategies, including, but not limited to, stimulus-evoked (task-related or event-related) fMRI, which captures brain activity in response to task execution and discrete events or stimuli with precise temporal resolution (5), its variant for observing brain responses to drug administration (6), and resting-state fMRI for examining intrinsic coupling of activity dynamics across brain regions (7). Amidst these advances, naturalistic fMRI emerges as a promising frontier, reflecting a shift towards studies that mimic real-life scenarios to understand the brain's functioning in environments that closely resemble those encountered in the natural world (8). This approach is increasingly recognized for its potential in human neuroimaging research. However, the application of naturalistic fMRI in animal studies remains relatively unexplored but holds significant promise for future research. Techniques such as zero echo time (ZTE) (9) and multi-band SWeep Imaging with Fourier Transformation (MB-SWIFT) (10) have shown promises in advancing this direction, offering reduced acoustic noise and susceptibility/motion artifacts, which are useful for accurately mapping brain activity in more naturalistic settings.

As an extended method, pharmacological fMRI (phfMRI) extends the utility of fMRI into the realm of pharmacology, allowing for assessment of drug effects on brain function. phfMRI can elucidate the brain-wide dynamics and spatial site of action for various compounds, offering valuable data on the efficacy and potential side effects of drugs. In rodent studies, this approach is crucial for preclinical drug development, providing a bridge between molecular pharmacology and behavioral neuroscience (6, 11-13). Unlike direct brain stimulation methods such as electrical, optogenetic, or direct pharmacological injections (14, 15), which induce comparatively rapid, localized changes in neuronal activity, phfMRI often involves the systemic administration of pharmacological agents. This difference means that phfMRI generally observes slower, more widespread responses that cannot be repeated as rapidly due to the systemic distribution and metabolic clearance of drugs. Analytically, phfMRI emphasizes changes in functional connectivity or drug-induced network alterations, contrasting with the assessment of immediate, task-evoked brain activity (16). This requires models that account for the extended temporal effects of drugs on brain activity.

Analyzing spontaneous fMRI signals offers a unique window into the brain's intrinsic coupling of activity (or often referred as functional connectivity), without the need for specific tasks or stimuli (7, 17). During the "rest state", the spontaneous fluctuations over time can classically be used to examine temporal correlation focusing on low frequency (0.01–0.1 Hz) BOLD fluctuations between spatially distinct brain regions (18, 19). Emerging studies indicate that signals at frequencies higher than 0.1 Hz also contain useful information (18-21). The term "resting-state fMRI" may sometimes be a misnomer, particularly in studies where the brain is not in a typical "resting" condition. Therefore, we believe that associating fMRI with the terms such as functional connectivity, intrinsic activity, or coupling could be more encompassing, reflecting the broad spectrum of studies that explore changes in correlated brain activity, whether during various behaviors or under experimental neuromodulation, as described in this article. In both humans and rodents, fMRI has revealed large-scale brain networks involved in critical functions such as auditory processing, motor function, visual processing, memory, and executive functioning. In rodents, the default mode network (DMN), anchoring at the retrosplenial cortex, was first demonstrated by Lu et al. (22) with its constituent (23) in rats, and then also identified by Liska et al. in mice (24) with detailed anatomical architecture reported more recently (25). Its functional implications have since then been increasingly investigated (26-29). The salience network (SN), anchoring at the ventral anterior insular cortex, was also identified and investigated in rodent models (20, 28, 30, 31). Tsai et al. demonstrated that the rat SN reacts to conditioned cues and shows increased functional connectivity with the DMN during conditioned heroin withdrawal, emphasizing its crucial role in mediating attentional and emotional processes related to addiction (30).

With the advent of high spatial resolution fMRI (32, 33), researchers are now able to discern functional activity at the layer-specific level within the cortex, revealing patterns of neural engagement. This advancement, coupled with high-resolution studies at ultra-high magnetic fields (34), allows for detailed mapping of neural circuits and their contributions to overall brain function. On the temporal front, advancements in fast fMRI techniques (35) have significantly improved the ability to capture dynamic brain activities, offering a window into the rapid changes in neural communication. A pivotal shift in fMRI research methodologies is the transition from anesthetized to awake rodent studies (36-39). This transition addresses the confounding effects of anesthetics on brain activity measurements. Awake rodent fMRI studies present a more accurate reflection of natural brain states and behaviors, facilitating a closer approximation to human fMRI studies, which are predominantly conducted without anesthetics. As fMRI technology and analytical methods continue to evolve, especially in the realm of awake imaging, the future of rodent fMRI research is poised for significant breakthroughs. Emphasizing awake imaging as a key direction aligns with the imperative to mitigate anesthetic confounds, enabling a more authentic exploration of the brain's functional architecture and its relationship with behavior and cognitive processes. Below we review the technical evolution of DBS, optogenetics, and chemogenetics, highlight key progress made in the field of animal fMRI, and discuss technical considerations and future opportunities.

DBS allows testing our understanding of neural circuits and exploring potential circuit-based treatment for neurological and neuropsychiatric disorders. Initially approved by the FDA for Parkinson’s Disease in 1997 (40), DBS has expanded its reach to include dystonia (2003), obsessive-compulsive disorder (2009), and epilepsy (2018) (41). DBS offers easily adjustable stimulation parameters (e.g., frequency, pulse-width, and amplitude), enabling the study of brain function and facilitating more seamless translation of research finding between animals and humans (1). In fact, the concept of DBS was initially explored in rodent models, where its potential therapeutic effects on movement disorders were examined on the subthalamic nucleus (STN) based on knowledge of the basal ganglia circuitry in movement control (42, 43). These early successes laid the groundwork for its application in humans. This translation was significantly propelled by the work of Benabid and Pollak (44). Although their initial study did not focus directly on the STN, their seminal discoveries of the thalamic DBS effects have demonstrated the efficacy of high frequency DBS in alleviating movement disorders. Since the establishment of DBS as a clinical procedure in the 1990s (45), the integration of DBS with brain mapping technologies, such as fMRI, has emerged as a key area of interest. This synergy is particularly crucial for researchers aiming to decipher the brain’s information flow or to interrupt pathological signals at the frequency and fidelity that are difficult to achieve by other competing neuromodulation techniques.

The integration of DBS and fMRI technologies enables researchers to visualize and quantify the immediate effects of DBS on brain activity in action, enhancing our understanding of the underlying mechanisms of DBS therapy. The convergence of DBS with fMRI in small animals began in the early 2000s as researchers aimed to navigate the technical challenges inherent in integrating these methods (46-48). A notable step was made in 2004, when Shyu et al. reported the use of a bended glass-coated carbon fiber electrode underneath a surface coil at 4.7T to demonstrate reliable thalamic DBS-induced changes in BOLD activity and c-Fos expression within the downstream ipsilateral anterior cingulate cortex (46). Building on this, Dunn et al. introduced a protocol for coating carbon fiber electrodes with polymer solution for chronic applications at 9.4T (47), which was among the initial studies to illustrate the feasibility of using chronically implanted electrode to monitor evoked BOLD activities within the brain. Further research expanded the range of “MR-compatible” electrode materials, exploring alternatives such as platinum–iridium (Pt–Ir) alloy electrodes (48-53) , which, despite their translational practicality and accessibility, introduce comparatively stronger susceptibility artifacts. Tungsten microwire electrodes, available commercially (54-58) or as custom fabrications (46), present smaller susceptibility artifacts. Recent advances have introduced designs using graphene fiber electrodes (Fig. 1) (59) and thin-film polyimide-based gold-copper electrodes (60-63) further minimizing susceptibility artifacts and enabling the acquisition of more comprehensive fMRI maps without signal loss. The continuous advances of thin-film electrodes, notable for their superior biocompatibility, flexible manufacturing for variable channel counts, inherent MR-compatibility, and broad usage in research beyond MRI, indicate a promising increase in their use alongside fMRI in the foreseeable future.

From the perspective of imaging sequences, the innovative applications of ZTE (9) or MB-SWIFT techniques, as initially demonstrated by Lehto et al. (10) and subsequently elaborated in multiple follow-up studies (64-70), have shown significant potential. Owing to the sequences’ insensitivity to T2* effects, these methods provide fMRI maps nearly free of susceptibility artifacts, thereby enabling the use of a wider range of electrode materials in high magnetic field environment. Importantly, these techniques enable the visualization of activity changes near an implanted lead – a capability that proves challenging with conventional gradient-echo-based fMRI methods. They also facilitate studies of DBS effects in awake animals, free from confounding effects caused by anesthetics, thanks to the reduced acoustic noise.

While continuing advances in MR-compatible electrodes and imaging sequences have notably improved the ability to perform DBS during fMRI, with only a few exceptions (55), rats are predominantly used over mice due to their larger brain size in affording electrode susceptibility artifacts and more stable physiology. DBS-fMRI studies of the hippocampus and its related pathways (48, 49, 51, 67, 71-73) have shown significant modulation of brain networks involved in memory and cognition. This suggests potential applications in enhancing memory performance and offering therapeutic benefits for neurological conditions like Alzheimer’s disease and temporal lobe epilepsy. A seminal study by Canals et al. (51) demonstrated the pioneering use of hippocampal DBS to map long-term potentiation (LTP), showing that LTP induces a global reorganization of activity patterns beyond hippocampus, enhances interhemispheric communication, and engages limbic and neocortical circuits due to changes in hippocampal synaptic strength (Fig. 2). DBS of the STN and internal globus pallidus has demonstrated frequency-dependent responses in the rat brain, particularly affecting the ipsilateral motor cortex (74). DBS in the ventroposterior thalamus has allowed for repetitive sampling of thalamocortical responses, revealing consistent and specific BOLD responses in the somatosensory cortex. This highlights the potential of using DBS-fMRI to study brain plasticity and the dynamics of specific neural circuits over time (56). Similarly, DBS of the substantia nigra pars reticulata (SNr) and external globus pallidus (GPe) revealed extensive functional connectivity within the basal ganglia, uncovering unexpected connectivity patterns and evoked responses. These findings suggest a more complex role for the SNr and GPe in modulating neural networks (75). DBS of the nucleus accumbens (NAc) has been demonstrated to modulate the prefrontal cortex and limbic areas, potentially affecting emotional, cognitive, and reward processing (57). DBS in the infralimbic cortex leads to specific alterations in amygdala activity, showcasing the utility of precision neuromodulation in mapping and modulating brain responses (64). DBS of the lateral habenula (LHb) in rat models of depression demonstrated an immediate alleviation of depressive-like symptoms, with fMRI revealing DBS-induced activation of multiple regions within the LHb’s afferent and efferent circuitry (76). Furthermore, DBS-fMRI studies targeting the infralimbic cortex suggest sex-specific neural dynamics in anxiety, highlighting the need for gender-informed therapeutic strategies (77). DBS targeting the ventral tegmental area (VTA) has proven to be a consistent method for eliciting dopamine release. This neurostimulation approach has significantly contributed to the advancement and empirical validation of integrated techniques combining fast-scan cyclic voltammetry (FSCV) with fMRI. Such simultaneous FSCV-fMRI methodologies enable real-time monitoring and mapping of dopamine neurotransmission dynamics in response to VTA stimulation, offering valuable insights into dopaminergic signaling mechanisms (Fig. 3) (52). Central thalamic DBS (CT-DBS) has emerged as a promising intervention for enhancing cognitive functions and addressing social deficits in neurological and psychiatric disorders, including autism spectrum disorder (ASD), with fMRI findings illustrating CT-DBS’s role in modulating the corticostriatal, corticolimbic, and thalamocortical networks (60, 78). DBS in the STN of Parkinsonian rats revealed robust BOLD responses across the basal ganglia-thalamocortical network, indicating that STN-DBS modulates both motor and non-motor pathways, potentially through both orthodromic and antidromic signal propagation (Fig. 1) (59). These examples collectively underscore the pivotal role of integrating DBS and fMRI in advancing our understanding of neurological diseases in animal models, facilitating a deeper comprehension of neural circuitry and pioneering innovative treatment approaches, thereby setting the stage for future clinical applications.

Considering the future of DBS in neuromodulation research, it is imperative to address both the ethical and practical challenges associated with its use. The capacity of DBS to drive both incoming and outgoing axons as well as neurochemical activity raises important questions about the specificity and precision of DBS in modulating neural circuits (1, 79-82). This effect complicates the delineation of DBS-induced changes, necessitating advanced analytical methods to unravel the complex neural interactions at play. In addition, the ethical considerations of subjecting animal models to invasive procedures for the sake of research demand rigorous justification. The pursuit of less invasive neuromodulation approaches remains a critical area of research, promising to expand our toolkit for studying and modulating brain function with fewer ethical concerns and practical limitations.

In response to these challenges, the field is witnessing the emergence of ultrasound, a non-invasive technique that is promising for modulating activities deep in the brain. This approach leverages the use of focused ultrasound waves from multiple transducers, offering a potential solution to the limitations posed by traditional DBS and promising to expand toolkit for studying and modulating brain function (83-85). We anticipate more relevant studies to emerge in the small animal fMRI space.

The synergy between DBS and fMRI holds potential to offer insights into neurological and neuropsychiatric disorders that do not present immediate behavioral feedback such as that in movement disorders. The potential for personalized or closed-loop therapy based on fMRI activity and connectivity metrics is significant (86), urging continued efforts to overcome technical hurdles and foster interdisciplinary collaboration. Specifically, the individual differences in DBS treatment responses are broadly documented in humans but not typically accounted for in small animal fMRI experiments. Future studies examining how baseline brain state and connectivity patterns predict neuromodulation outcomes may shed light on this issue.

Optogenetics has emerged as an essential tool for understanding the roles of cell types, circuits, and systems under both normal and pathological conditions (2). This revolutionary technique originated from the ingenious integration of light-sensitive proteins, known as opsins, into cells, allowing for the precise control of these cells upon light exposure. The conceptual foundation for optogenetics was laid in a 1979 Scientific American article by Nobel laureate Francis Crick, who articulated the paramount challenge in neuroscience: the need to control specific types of cells within the brain, while leaving others unaffected (87). In 1970s, Oesterhelt and Stoeckenius made a breakthrough by discovering bacteriorhodopsin, a light-sensitive protein in microorganisms (88). Building on this foundation, Hegemann, Nagel, and colleagues identified and characterized channelrhodopsin in the early 2000s (89), a critical step towards the development of optogenetics. The field reached a seminal milestone with the work done in Karl Deisseroth’s lab (90, 91) which enabled the introduction of channelrhodopsin-2 (ChR2) into mammalian neurons. ChR2, when activated by light, opens channels in the membrane of neurons, leading to depolarization and the initiation of an action potential. The core technologies underpinning optogenetics include the genetic engineering methods used to express opsins in targeted cells, and the optical systems designed to deliver light of specific wavelengths to activate or inhibit cells. Notably, excitatory opsins like the ChR2 variants (90, 92) depolarize neurons to initiate firing, while inhibitory opsins such as archaerodopsin (Arch) (93) and halorhodopsin (NpHR) (94) hyperpolarize neurons, effectively silencing neuronal activity. These tools have been continually improved over the years and complemented by advances in fiber optics and light delivery systems (95-97), allowing precise spatial and temporal control of neuronal activity in vivo (98-101). Moreover, the development of a diverse array of opsin proteins, each with unique properties regarding light sensitivity, ion selectivity, and activation kinetics, enables more nuanced control over neuronal activity (95, 96).

The advent of optogenetics paired with fMRI technology has heralded a new era in the exploration of neural circuitry and functional connectivity in animal models. The pioneering work by Lee et al. initiated this journey, revealing how optogenetic stimulation of excitatory neurons targeted by CaMKIIα promoter leads to modulation of BOLD signals (102). This effort allowed for the mapping of complex brain networks and laid the foundation for subsequent explorations into the intricacies of brain function. Kahn et al. built on this, examining the linearity of BOLD responses, further cementing fMRI's reliability as a gauge of neuronal activity (103). In followed up studies, Kahn et al. correlated optogenetic activation of neocortical neurons with fMRI signals (104), suggesting a link between neural firing and BOLD responses, whereas Yu et al. advanced the measurement of optogenetically driven fMRI responses from mesoscopic to microscopic, single vessel scale (Fig. 4) (33). Chen et al. innovatively combined optogenetic manipulation of neuronal Ca²⁺ signaling with single-vessel BOLD and CBV-fMRI, uncovering precise hippocampal hemodynamic responses related to specific neuronal activities (105). This study significantly enhances fMRI resolution, offering detailed insights into the hippocampus, a region often limited by conventional fMRI techniques. Collectively, the studies above have paved the way towards a better understanding of neural activation and its coupling with local and downstream vascular responses (106-108). Beyond demonstration of the roles of excitatory neuron stimulation (15, 102-104, 109-112), the precise selectivity of optogenetics has further elucidated the role of GABAergic, dopaminergic, serotonergic, cholinergic, and norepinephrine systems in shaping brain-wide activity patterns and behaviors. Studies examining GABAergic neurons, where optogenetic stimulation reveals a generally lower oxygen demand compared to excitatory activities, suggesting a nuanced metabolic landscape (113, 114). While excitatory neuron stimulation typically induces positive BOLD responses, inhibitory GABAergic neuron stimulation can lead to positive, negative, or biphasic responses, revealing the intricate dynamics between neuronal activity and cerebral hemodynamics (15, 115-123). For example, Cerri et al. delineate the contrasting fMRI responses with optogenetically induced elevated striatal neuronal activity: negative responses from stimulating GABAergic neurons in striatum and several of their relevant projections, versus positive responses from dopaminergic neuron stimulation (Fig. 5), highlighting the critical role of neurochemicals in modulating fMRI signal polarity (15). Several studies focusing on optogenetic stimulation of dopaminergic neurons have collectively shed light on the precise topological landscapes of these neurons in striatal and prefrontal regions (124-126), linking their relevance to reward-seeking behaviors (124), as well as the impact of pairing dopamine stimulation on strengthening the representation of a discrete sensory stimulus (125). Recent studies employing optogenetics in conjunction with fMRI have significantly advanced our understanding of the serotonin, cholinergic, and norepinephrine systems. These studies have revealed the complex roles these neurotransmitter systems play in brain function and behavior. Research utilizing this combined approach has detailed the serotonergic system's brain-wide effects in response to acute stress and fluoxetine, a selective serotonin reuptake inhibitor (127). This exploration has unveiled the intricate, region-specific dynamics of the serotonergic system, illuminating its role in modulating brain function and stress responses through its extensive projections and receptor distributions. Significant advancements have been made in understanding the cholinergic system's role in modulating large-scale brain networks (39). Though cholinergic neurons comprise only a fraction of the basal forebrain's neuronal population, their activation via optogenetics significantly alters functional connectivity and BOLD signal amplitude within key neural networks such as the default-mode-like network. This alteration contrasts with effects observed from other neuron types in the basal forebrain, highlighting the cholinergic neurons' unique contribution to modulating attention, learning, memory, and emotional regulation. Their findings also reveal that cholinergic and non-cholinergic neurons collaboratively modulate global functional networks, driving behavior preferences towards either internally or externally oriented states. Further studies into the locus coeruleus-norepinephrine system have elucidated how its various firing patterns uniquely modulate brain activity, dynamics, and behavioral states (128). Tonic firing, with slower frequencies, affects prefrontal and limbic areas, influencing cognitive performance and arousal in a nonlinear manner. High-frequency bursts is linked to increased alertness and responsiveness to stimuli, enhancing network integration and facilitating brain state transitions. This study highlights the LC-NE system's critical role in brain flexibility and behavioral adaptation, demonstrating its complex influence on global brain dynamics through specific activity patterns. These findings highlight the regulatory role of the locus coeruleus in brain function, emphasizing its importance in facilitating transitions between brain states and optimizing cognitive performance.

The variety of viral expression strategies enables researchers to achieve multiplexed control of neural circuit activity with high specificity (96). For example, using retrograde vectors, recombinase-, or promoter-dependent strategies allows for the manipulation of selected cell types within a mixed population, whereas the point of light illumination at somata or axonal terminals in downstream brain regions offers different levels of pathway specificity. These approaches have been used to increase throughput in a single experimental session, where mixed populations can be manipulated or multiple regions can be independently targeted and controlled using distinct wavelengths of light. These multiplexing strategies can further incorporate naturalistic or pharmacological stimuli and enable simultaneous or sequential activation or inhibition of different cell populations, allowing for complex experimental designs. For example, Jung et al. exemplifies the controlled silencing of circuits downstream of somatosensory activity using fMRI. This study unveils a technique for suppressing specific brain regions and reveals how interconnected regions contribute to information flow (129). A study by Cerri et al. demonstrates the concurrent use of optogenetics with intracranial pharmacology and fMRI, and showcases the utility of optogenetics to initiate antidromic spiking through terminal stimulation, while studying the effects of various pharmacological compounds on manipulating the evoked fMRI activity (15).

Beyond research in neurovascular coupling and neurochemical influences of fMRI signals, several studies have leveraged optogenetic stimulation to map the causal relationship between selected nodes and functional network dynamics. Several studies have initially focused on the relation to sensory/motor functions and memory consolidation, highlighting the importance of large-scale spatiotemporal patterns in neural signal transmission, significantly advancing our understanding of systems-level brain network (101, 130-136). For example, frequency-dependent modulation of brain networks by distinct hippocampal stimulations has revealed how different stimulation frequencies can elicit varied network responses (137-140). Lee et al. provide a comprehensive analysis on how the anterior cingulate cortex contributes to the modulation of sensorimotor pathways in chronic pain, underscoring the role of this brain region in orchestrating movement avoidance behaviors in response to pain (141). Their findings elucidate the increased network strength of the anterior cingulate cortex in a chronic pain model, highlighting a critical aspect of how pain is not merely a sensory experience but deeply integrated with motor responses to form a cohesive defensive behavior. Adding to this, the work of Bernal-Casas et al. used dynamic causal modeling for optogenetic fMRI data to elucidate the cell-type-specific interactions within the basal ganglia-thalamocortical circuit, a key player in motor control. This circuit, comprising various brain regions and neuronal types, including D1- and D2-receptor-expressing medium spiny neurons, orchestrates complex motor behaviors through distinct pathways (142). A recent study by Menon et al. represents a reverse-translation case using rodent models to examine a hypothesis derived from human fMRI data modeling. Specifically, Menon et al. tackled the functional causal influence of anterior insular cortex on its related major network nodes, demonstrating feedforward optogenetic stimulation of insular neurons induces dynamic suppression and decoupling of the DMN (143). Interestingly, beyond the translational insights into anterior insular cortex’s role in network switching, this study suggests that prelimibic and cingulate cortices, commonly identified as the rodent DMN nodes, may also involve in the SN. This implies distinct levels of network segregation between rodents and humans that warrants further investigations.

As the field of optogenetics-fMRI advanced, the significance of addressing light-induced heating artifacts to avoid confounding fMRI signals has been underscored (144), alongside the necessity of recognizing and mitigating unspecific artifacts to preserve data integrity (145). Furthermore, the light-induced vasodilation, as indicated by Rungta et al. (146), introduces additional complexity, emphasizing the necessity to include essential controls. We expect the application of optogenetic fMRI to be extended to awake animals, as initially demonstrated by Liang et al. (110), to offer insights into the anesthetics-free brain dynamics. Along the same vein, Cover et al. (147) developed a method to image whole-brain dynamics during optogenetic self-stimulation in mice, enabling concurrent observation of brain activity related to motivated behavior. Further, the innovative work by Kim et al. on whole-brain mapping of effective connectivity via fMRI with cortex-wide patterned optogenetics marks a significant step forward (Fig. 6) (101). It represents an exciting opportunity for modulating large-scale networks at high spatiotemporal precision beyond single-node interrogations as seen in conventional studies.

Chemogenetics, also known as Designer Receptors Exclusively Activated by Designer Drugs (DREADDs), represents an innovative technique that leverages the integration of receptor engineering with pharmacology to remotely control neuronal activity in experimental animals (3). This concept was initially used to describe how mutations affect enzyme activity and substrate specificities, exemplified by the chalcone isomerase in Dianthus caryophyllus flowers (148). Distinguished from optogenetics, which uses light and light-sensitive opsins for neuronal modulation, chemogenetics, as described by Bryan Roth et al. (3, 149), enable control over selected cellular functions with less invasive procedures, offering prolonged and sustained modulations. This proves especially advantageous for studying behaviorally relevant neural circuits similar to pharmacological approaches but with improved specificity. The development of DREADDs technology was pioneered by Armbruster et al. in the laboratory of Bryan Roth’s (150), showing the engineered G-protein-coupled receptors (GPCRs) can be potently activated by an inert ligand. This innovation not only showcased the feasibility of neuromodulation using engineered receptors but also introduced clozapine-N-oxide (CNO) as a synthetic ligand that activates DREADDs. Subsequent enhancements in DREADD technology have broadened its application to various intracellular G-protein pathways (Gq, Gi, or Gs), enabling more sophisticated regulation of cellular responses beyond simple excitation or inhibition (151-153).

Similar to optogenetics, chemogenetics also allows cell body versus projection-specific targeting (154). For the later case, different levels of precision can be achieved by using viral vectors and Cre-recombinase strategies (155), such as injecting Cre-dependent DREADD vectors to the cell bodies of the targeted circuit and injecting retrograde Cre-expressing vectors in projection terminals, allowing DREADDs to be expressed in a specific projection. Alternatively, localized, intracranial injection of a DREADD actuator during experiments can also approach projection-specific effects. Similar to optogenetics, chemogenetics enables multiplexed modulation when distinct DREADDs are expressed in a cell-type-specific manner within different neuron populations. For example, (e.g., Gq-DREADDs (for activation) and KOR-DREADDs (for inhibition using a distinct actuator salvinorin B) can be expressed within the same subjects and be activated by different actuators (156). Importantly, at the time of data acquisition, the minimal invasive nature of chemogenetics allows its seamlessly integration with optogenetics or other neuromodulation methods. This integration facilitates the simultaneous or sequential activation or inhibition of multiple neural pathways with less space constraints, facilitating intricate experimental manipulations of brain circuits during behavior or during fMRI. These advancements in chemogenetics, alongside the continuous development of DREADD actuators with improved affinity and specificity, pave the way for a simplified integration of selective neuromodulation technique with fMRI and other neuroscience tools in animal models.

Integrating chemogenetics with fMRI offers the most simplified, and perhaps most efficient setting to examine cell type and circuit specific modulations on brain networks. This methodology marries the precise control over neuronal activity afforded by chemogenetics with the capability to visualize the spatiotemporal effects across the entire brain via fMRI. A foundational study in this field by Gozzi et al. showcased the potential of this integrative approach (157). Their work highlighted how the early variant of chemogenetic inhibition within the central nucleus of the amygdala impacts conditioned fear responses in mice, unveiling specific brain circuits involved in fear expression and modulation. This study established a direct connection between targeted neuronal inhibition and changes in brain-wide activity and behavior. Building on this foundational work, further research introduced chemogenetic fMRI to examine the brain-wide effects of serotonergic transmission (158). This design allowed for the selective activation of serotonin-producing neurons in mice, leading to the identification of specific regional brain activations. Such studies underscore the targeted action of endogenous neurotransmission systems and their unique functional signatures across the brain. Subsequently, Roelofs et al. showcased a chemogenetic and fMRI approach to map the mesocorticolimbic system in rats, revealing how activating VTA-to-NAc core or VTA-to-mPFC projections influences global brain dynamics and behavior (159). By using DREADDs combined with the concept pioneered in phMRI, their study demonstrated significant BOLD responses not just in targeted areas but also across remote neural circuits, aligning these changes with behavioral outcomes like locomotor activity. This work highlights the effects of specific neural pathway activations on broad neural network activity. A landmark paper demonstrated that the chemogenetic activation of the locus coeruleus (LC) noradrenergic neurons could rapidly reconfigure the functional connectome (Fig. 7)(160). This study highlights how neurotransmitter systems, specifically norepinephrine, critically modulate brain function and network dynamics, emphasizing the locus coeruleus’ role in enhancing certain connectivity patterns.

Building upon these insights, a deeper understanding has emerged on how manipulations of specific neural pathways can profoundly influence brain dynamics and behavior. The regulation of BOLD dynamics and brain-wide connectivity, for example, has been significantly advanced through studies targeting specific neural circuits. Research into the striato-thalamo-cortical circuit has unveiled its crucial role in regulating these aspects, particularly through the manipulation of D1 medium spiny neurons (161). The unilateral activation of D1 medium spiny neurons can influence bilateral brain networks, highlighting the striato-thalamo-cortical circuit's significant role in brain dynamics (162). Similarly, the exploration of the hippocampal regions' differential roles has been associated with pain management, revealing novel strategies that could potentially transform therapeutic approaches to neuropathic pain (163). Kargl et al. explored affective learning mechanisms through a cortico-limbic circuit involving the insular cortex and amygdala (164). Their research illuminates the integration of affective value, dissecting it into salience and valence within this circuitry, which plays a role in behavioral conditioning. This work addresses how specific neural pathways contribute to complex behaviors and also suggests avenues for therapeutic interventions. Li et al. demonstrated that chemogenetic suppression of specific brain hubs, such as the sensory cortex and subcortical areas, significantly impacts memory consolidation in mice, suggesting the role of distributed networks beyond the hippocampus in this process (165). The research into brain network dynamics and memory underscored the essential role of network integration in memory consolidation dynamics.

The role of the DMN and its nodes has been of particular interest in recent years, notably highlighted by Peeters et al. (2020). Their study on the anterior cingulate area of mice, using chemogenetic silencing, revealed the invovlvement of this area in modulating functional network activity and connectivity (166). Tu et al. discovered that inactivating the dorsal anterior cingulate cortex disrupts brain network organization, affecting resilience, segregation, and behavior without compromising overall network integration (167). This chemogenetics and resting-state fMRI study highlights the DMN's crucial role in complex behaviors and the interconnectedness of brain networks. A recent study revealed that inhibiting activity in specific brain regions like the medial prefrontal cortex can paradoxically enhance fMRI connectivity (Fig. 8) (168), suggesting a counterintuitive increase in network connectivity through altered oscillatory processes. The paradoxical phenomenon is further elucidated by in vivo electrophysiology findings, which show that such chemogenetic inhibition augments low-frequency (0.1–4 Hz) oscillatory power. These findings challenge conventional understanding and indicate that decreased regional brain activity can in fact enhance overall network connectivity by modifying the dynamics of oscillatory brain processes. Furthermore, the application of chemogenetic manipulations to study the default mode-like network and the locus coeruleus has provided critical insights into the interplay between neuromodulation, network connectivity, and behavior. Oyarzabal et al. (Fig. 9) (26) utilized a multifaceted approach, incorporating fMRI, PET, and spectral fiber photometry, to elucidate the locus coeruleus's role in modulating the DMN through chemogenetic modulation. Specifically, this study revealed that tonic activation of locus coeruleus noradrenergic neurons notably reduces cerebral blood volume and PET-derived glucose uptake while concurrently enhancing synchronous low-frequency fMRI activity as well as calcium-dependent neuronal activity within the DMN's frontal cortices. This observation of NE-induced vasoconstriction alongside increased neuronal firing and network connectivity highlights the importance to consider additional neurochemical-related vasomotor activity when interpreting fMRI data. Aside from the noradrenergic system, cholinergic modulation on DMN has also been examined by chemogenetic fMRI. Peeters et al. (169) provided insights into the interplay between cholinergic neuromodulation and the DMN-like network through unilateral stimulation of cholinergic neurons. More recently, Sanda et al. explored the neural mechanisms by which cholinergic modulation affects the resting-state activity in the brain, particularly focusing on the DMN (170). Utilizing computational models based on rodent and human connectomes, they demonstrated that alterations in excitability and synaptic connections—precipitated by cholinergic influences—can lead to significant changes in resting-state dynamics.

The chemogenetic technique has established itself as a broadly utilized approach for neuromodulation. Yet, understanding the procedural nuances of this technique is crucial to enhancing its application in fMRI. Rapidly toggling activity on and off with multiple repetitions, common in DBS and optogenetic fMRI studies, is not feasible with chemogenetics. Consequently, this prompted most studies to leverage the prolonged data sampling to examine functional connectivity changes. Drift in the fMRI signal can pose challenges in chemogenetic fMRI data analysis. This drift may be due to changes in animal physiology and/or MRI gradient heating and can be confounded with the slow pharmacodynamics of DREADD-induced activity changes. The sensitivity of fMRI to such drifts suggests the need for meticulous experimental design to discern changes from chemogenetic modulations, often necessitating a sufficiently long baseline period and appropriate controls. Control experiments are especially warranted due to the side effects of actuator drugs. For instance, CNO and its analogs may interact with endogenous receptors, representing a critical point of consideration. Off-target effects can obscure interpretations of results, as changes in neural activity or behavior may not solely indicate the activation of DREADDs but also the broader pharmacological impact of the actuators. A study by Gomez et al. investigated the use of CNO in chemogenetic studies, suggesting the potential confounding effect of its rapid in vivo conversion to clozapine, a metabolite with high affinity and potency for DREADD activation (171). Specifically, it was suggested that the assumed direct activation of DREADDs by CNO may be confounded by its rapid in vivo conversion to clozapine, a metabolite of CNO with high affinity and potency for DREADD activation. Over the past few years, the search for alternative actuators capable of more specific modulation of DREADDs has led to the identification and increasing use of ligands such as Compound 21 (C21) (172) and Deschloroclozapine (DCZ) (173), with the latter now being the most increasingly utilized. These advancements have laid the groundwork for enhancing the utility of this technique in animal fMRI, and it is conceivable that the ongoing development of DREADD actuators will further improve specificity and reduce the systemic effects associated with this technique.

It is important to note that the potential future applications of DREADDs are promising for studies extending beyond neurons, offering insights into other brain cell types essential for neurovascular coupling, such as astrocytes. Unlike neurons, astrocytes don't fire action potentials but signal through GPCRs. By utilizing DREADDs to selectively modulate astrocytic signaling, one can parse out their distinct contributions to fMRI signals and deepen our understanding of fMRI contrast mechanisms. Ultimately, the capacity for sustained neuromodulation afforded by chemogenetics epitomizes pharmacological mimicry, delivering a level of specificity superior to that of traditional drugs. This could lead to alterations in behaviorally relevant brain states, facilitating causal studies with a precision previously unattainable. We anticipate that the integration of chemogenetics and fMRI will be instrumental in elucidating the mechanisms underlying brain network transitions.

Causal inference models typically include mediation analysis (Fig. 10) (26, 134, 175), where the model can be set to investigate whether the relationship between a neuromodulatory input (the independent variable) and behavioral outcomes (the dependent variable) passes through an intermediate brain network variable (the mediator). These models excel in distinguishing direct effects from those mediated by alterations in brain network activity. Their capacity to handle scenarios with multiple mediators (176) enhances the understanding by allowing for the simultaneous consideration of various brain network measures as mediators. This capability is crucial for delving into whether behavioral changes induced by interventions like optogenetic stimulation are direct or mediated through modifications in specific fMRI-derived brain network measures, such as functional connectivity within and between multiple networks. Furthermore, these causal inference models are adaptable, enabling researchers to extend their analyses to consider network measures not just as mediators but also as outcomes, with activities from other network nodes acting as mediators. The application of mediation analysis goes beyond mere analytical choices; they are integral to weaving a comprehensive narrative that links disparate scales of data. Positioning behaviors or network measures as independent variables in the model allows researchers to address fundamentally different questions. For instance, when behaviors are set as the dependent variables, the focus is on elucidating the pathways through which neuromodulation influences behaviors via hypothesized mediators. Conversely, when network measures are deemed the dependent variables, the analysis shifts towards understanding how the mediators might alter network strengths or connectivity patterns. This approach not only enriches our insight into the specific mechanisms of action behind neuromodulatory interventions but also underscores the importance of carefully considering the theoretical and methodological underpinnings of our analytical models in neuroimaging research.

On the other hand, moderation analysis (Fig. 10) allows us to understand the conditions under which certain neuromodulatory effects on fMRI signals are strengthened or weakened, providing insights into the contextual dependencies of neural circuit manipulations (177). For instance, the efficacy of electrical deep brain stimulation may vary depending on the baseline activity level of the targeted network, a relationship that moderation analysis can quantify. Furthermore, considering factors like sex and age is increasingly recognized as crucial in brain studies, as these variables can significantly influence both the baseline state of neural networks and their response to neuromodulatory inputs . As an example, a recent work (178) demonstrates that the intrinsic FC between the anterior insula and the retrosplenial cortex before cocaine exposure serves as a moderator for the impact of cocaine self-administration on FC changes between these regions during abstinence in rats. This finding highlights the critical role of pre-existing network configurations in influencing the neural and behavioral outcomes following neuromodulatory interventions. Another example that demonstrates the influence of pre-existing conditions on neuromodulatory outcomes comes from the work of Lee et al (31). Their research into acute alcohol exposure reveals how it differentially affects the functional connectivity within the lateral cortical network in rats, depending on their age. Specifically, they found that acute alcohol induces a greater dose-dependent increase in FC within this network in adult rats compared to adolescents. This age-dependent sensitivity to alcohol underscores the critical role of developmental stages in determining the brain's response to neuromodulatory substances.

While both mediation and moderation analyses are useful toolkit for analyzing neuromodulatory effects, it is critical to recognize their unique contributions and requirements. Mediation analysis provides a framework for understanding causality, mapping the pathway from an independent variable through a mediator to a dependent variable. This approach assumes a specific sequence: that the independent variable can cause changes in the mediator, which in turn affects the dependent variable. Such a model requires a strong theoretical foundation to hypothesize and subsequently test the direct and indirect effects within the causal chain. This is particularly useful in neuromodulation studies for delineating the process by which neuromodulatory techniques influence behavior through changes in brain network activity, allowing researchers to pinpoint the role of specific neural circuits in mediating behavioral outcomes. On the other hand, moderation analysis focuses on the conditions or contexts that modify the effect of the independent variable on the dependent variable. A moderator variable is one that affects the strength or direction of the relationship between the independent and dependent variables, without being in the causal pathway. Moderation can reveal how different baseline characteristics, again, such as the sex and age of animals or the baseline activity level of targeted neural networks, can influence the outcomes of neuromodulatory interventions. This analysis does not imply causality between the moderator and the outcome; instead, it provides insights into the variability of effects across different conditions or subgroups. The distinct roles of mediation and moderation analyses in causal inference highlight the importance of carefully designing neuromodulation studies. Researchers must clearly define their hypotheses regarding causal pathways and moderating factors to apply the appropriate analytical models.

These analytical models underscore the importance of a holistic view of the data, one that recognizes the complexity of interactions between neural activity, brain networks, and behavior. It provides one of the worth noting structured methodology to dissect these complex interactions, paving the way for our use of fMRI technique towards a comprehensive understanding of the brain's modulatory landscapes and their behavioral correlates.