Modeling Brain Dysconnectivity in Rodents

Altered or atypical functional connectivity as measured with functional magnetic resonance imaging (fMRI) is a hallmark feature of brain connectopathy in psychiatric, developmental, and neurological disorders. However, the biological underpinnings and etiopathological signi ﬁ cance of this phenomenon remain unclear. The recent development of MRI-based techniques for mapping brain function in rodents provides a powerful platform to uncover the determinants of functional (dys)connectivity, whether they are genetic mutations, environmental risk factors, or speci ﬁ c cellular and circuit dysfunctions. Here, we summarize the recent contribution of rodent fMRI toward a deeper understanding of network dysconnectivity in developmental and psychiatric disorders. We highlight substantial correspondences in the spatiotemporal organization of rodent and human fMRI networks, supporting the translational relevance of this approach. We then show how this research platform might help us comprehend the importance of connectional heterogeneity in complex brain disorders and causally relate multiscale pathogenic contributors to functional dysconnectivity patterns. Finally, we explore how perturbational techniques can be used to dissect the fundamental aspects of fMRI coupling and reveal the causal contribution of neuromodulatory systems to macroscale network activity, as well as its altered dynamics in brain diseases. These examples outline how rodent functional imaging is poised to advance our understanding of the bases and determinants of human functional dysconnectivity.

Over the past 3 decades, progress in functional neuroimaging methods such as functional magnetic resonance imaging (fMRI) has been instrumental in assessing the landscape of brain network dysfunction in psychiatric disorders. The recent deployment of computational tools and resources for sharing and processing large datasets has further accelerated the transition from small-scale proof-of-concept studies in selected patient cohorts to initiatives aimed at assessing the extent and manifestation of network dysfunction in larger patient populations. The Autism Brain Imaging Data Exchange (ABIDE) (1), the ADHD-200 Consortium (2), ENIGMA (Enhancing Neuro Imaging Genetics through Meta Analysis) (3), and the UK Biobank (4) are examples of such large-scale endeavors. There is now great hope that these initiatives will help better pinpoint and categorize disruption of brain networks in psychiatric conditions (5), possibly providing objective imaging markers that can distinguish a diseased state from a normal one, a long-term quest of neuropsychiatric imaging (6).
However, understanding the biological and etiopathological significance of aberrant connectivity in mental illnesses is a nontrivial problem that is unlikely to be solved simply by aggregating more and more data. While it has been possible to demonstrate the presence of altered or atypical connectivity in most of these disorders at the population level, albeit with great variability at the individual level, the biological determinants and mechanistic significance of these deficits remain largely unknown. Moreover, human neuroimaging methods provide a snapshot of functional activity at the macroscale but do not allow us to probe pathophysiological mechanisms occurring at a finer investigational scale. This results in a wide explanatory gap between models of brain (dys)function at the cellular mesoscopic scale (i.e., receptor or neuronal dysfunction, excitatory/inhibitory imbalance, miswiring or circuit alterations) and the corresponding system-level measurement of brain function and connectivity. For these reasons, it is not yet clear what it means when specific brain functional connections in a psychiatric condition are weakened, are altered, or deviate from the corresponding measure in healthy control populations. The consequence is that imaging techniques in psychiatric and developmental disorders are still widely regarded as general imaging markers of endophenotypes, often devoid of true diagnostic, etiopathological, or mechanistic significance.
In this review, we argue that investigational approaches that allow for causally testing mechanistic hypotheses and bridging investigational scales beyond what is currently possible in human research can clarify the significance and multifactorial origin of brain dysconnectivity in mental disorders. Preclinical application of fMRI in animal models, such as rodents, gives a chance to fill this gap. By incorporating targeted and causally explainable perturbations into the same fMRI paradigms that are used in human investigations, rodent connectivity mapping is rapidly becoming a key tool for modeling, examining, and comparing network signatures found in human brain disorders (7,8). Furthermore, compared with humans, there is an abundance of high-resolution wholebrain physiological data (9), including direct tract tracing axonal connectivity data (10)(11)(12)(13), high-resolution gene expression maps (14), and cell density atlases (15,16). This allows examination and comparison of macroscopic network metrics with fundamental biological aspects of the brain, thereby addressing the need to reconcile data on brain cellular mechanisms in rodent models with theories of brain network function from human imaging.
This review is organized as follows. To begin, we briefly go over the technical considerations for acquiring resting-state connectivity fMRI signals in rodents, as well as current initiatives aimed at creating common platforms for data acquisition, processing, and sharing. We discuss strategies for building rodent-to-human translation based on encouraging correspondences in network organization across species. Next, we illustrate how rodent fMRI can be used to generate and test mechanistic hypothesis concerning the origin and significance of functional brain dysconnectivity in psychiatric illnesses. Finally, we review (pre)clinical fMRI research that has looked at ascending neuromodulatory systems (NMSs) and the impact of these systems on brain connectivity by offering the reader insight into how these mechanisms are relevant to the interpretation of connectivity aberrations in mental disorders. Our review is intended for both specialists and nonexperts in the fields of psychiatry, neuroscience, and neuroimaging who want to learn about the outstanding topics that rodent fMRI aims to investigate (Table 1). Reliable habituation protocols on multiple days (weeks) Control for stress levels (e.g., corticosterone) (54,57,63,64,66,67,154) Lack of consensus on rodent and human common data acquisition, preprocessing, and analysis pipelines Unclear effects of data (pre)processing and acquisition on fMRI connectivity and connectivity changes Large data sharing initiatives will engage standardization of best practices for acquisition and analyses of datasets (7,27,33,34,155) Rodent imaging studies tend to be minimally powered owing to strict ethics requirements (3Rs) May limit reproducibility of rodent research Limited brain-behavior inferences Multicenter data sharing initiatives Larger emphasis on independent replication of rodent fMRI studies Application of strict protocols to control for variables of no interest (standardization) (33,34,86) Unclear physiological meaning of fMRI connectivity changes in animal models fMRI connectivity changes often relegated to endophenotypes of unknown biological origin in animal studies Stronger focus on multimodal recordings and manipulations in rodents (49,86,96,133,139,156) MRI is an expensive and technically challenging method inaccessible to many groups Limited scope of application of the technique for technical or economic reasons Increased access to large-scale facilities Interlaboratory training groups and rotation Data and code sharing (33,34) fMRI, functional magnetic resonance imaging.

FUNCTIONAL CONNECTIVITY MRI IN RODENTS: TECHNICAL CONSIDERATIONS AND PERSPECTIVES
Recent reviews summarizing the history and development of rodent fMRI are an excellent resource for the interested reader (7,17,18). The first attempts to map fMRI connectivity in rodents date back to more than 15 years, and they were met with mixed and often contradictory results (19)(20)(21)(22). Leveraging advancements in MR hardware, optimized control of motion artifacts and physiological parameters (23), and improved image analyses pipeline, a second wave of investigations revealed the possibility of reliably mapping networks in rats (24)(25)(26) and mice (27)(28)(29)(30).
Fast-forwarding to 2022, the field has matured and begun to provide answers to initial uncertainties. For example, divergences between animal preparation, anesthesia, data acquisition, and processing were found to underlie a number of disagreements within the animal functional neuroimaging community, such as the nature of unilateral versus bilateral resting-state networks (RSNs) in mice (19,27,28,31), or the existence of RSNs of translational relevance, such as plausible rodent homologues of the human default mode network (DMN) and salience network (28,32).
Perhaps one of the community's most interesting actions occurred 2 years ago when 17 groups around the world openly shared their data in a joint effort to define a common image processing and analysis pipeline for mouse fMRI (33). Despite differences between laboratories in imaging equipment and procedures, this study identified multiple and reproducible large-scale RSNs in mice, including a DMN, in most datasets. This work also described several parameters, animal handling procedures, and equipment that can improve detection of RSNs. The experimental parameters associated with an improved spatial specificity of RSNs and an enhanced reproducibility of the functional connectivity parameter estimation between institutes include the use of dedicated cryoprobes, mechanical ventilation, and light sedation with medetomidineisoflurane or intrapulmonary administered gaseous anesthetics. This and other initiatives including a similar ongoing effort in rats (34) are critical to guide the design and analysis of future rodent fMRI investigations.

HOMOLOGIES AND DISSIMILARITIES IN RSNs BETWEEN HUMANS AND RODENTS
The discovery of reproducible and consistent RSNs in rodents has sparked fresh ideas about how this information may be exploited to assess commonalities and dissimilarities between animal models of humans and rodents, parallel to fMRI efforts conducted in nonhuman primates. Indeed, the establishment of comparable and homologous brain networks is required for direct comparison of rodent and human fMRI research. Testing this falls within the scope of a novel branch of neuroscience termed comparative functional neuroanatomy, which studies the brain's organization from an evolutionary perspective (35,36). The initial findings of these investigations revealed that several brain networks in rodents have a homologous architecture similar to that seen in human and primates, such as the salience (37,38), default mode (32,39), motor (40), and limbic (41) networks. Other investigations have expanded these analogies to include hierarchical organization of cortical connectivity as mapped with fMRI connectivity gradients (9,13,42,43) or the coactivation dynamics of blood oxygen level-dependent fMRI signals (44)(45)(46).
Indeed, cross-species variation exists in both functional and neuroanatomical network organization. Notably, rodents lack a clear neuroanatomical equivalent of the precuneus in their DMN, which serves as the most prominent hub in the human DMN, and its functional role may be transferred to the retrosplenial cortex (32,47). It is also possible that some of these disparities are the outcome of the evolution of functional networks with particular and distinctive capabilities for humans. For example, Balsters et al. (41) revealed that connectivity from the caudate nucleus and anterior putamen striatal regions in humans and macaques could not be matched to any mouse corticostriatal circuitry. Interestingly, the circuits formed by these areas are related to executive and sociolinguistic function and seem to be specific to nonhuman and human primates.
This information is extremely valuable and can assist the interpretation of connectivity fMRI recordings on animals and their cautious extrapolation to corresponding investigations of network dysconnectivity in humans. We believe that upcoming results from this active field of research will be pivotal in determining the translation potential of rodent models to humans so that the knowledge gained in animal research may be applied to understand the significance of clinical results. A review by Xu et al. (48) covers further outstanding topics and open questions about network homologies and dissimilarities between rodents and humans, from both a methodological and an evolutionary standpoint.

TOWARD AWAKE fMRI MAPPING IN RODENTS
Light anesthesia or sedation is commonly used in rodents to ensure animal immobilization and reduce stress related to image acquisition. Notwithstanding the possible confounding effects of anesthesia, this procedure also comes with a number of possible advantages. For one, optimized anesthesia protocols exist (23,25,29,49), and they enable the reliable detection of translationally relevant RSNs while mitigating physiological artifacts and reducing intersubject variability (50). In contrast, anesthetics may interfere with hemodynamic coupling under some circumstances, as well as regionally alter cortical and subcortical activity [reviewed by (51)], potentially confounding the results of manipulation or recording studies (36,52,53). These effects can be compounded by possible peripheral confounds affecting body temperature, heart rate, blood pressure, and respiratory rate, a set of parameters that can however be tightly controlled and monitored using advanced animal preparations (49,51).
In response to these concerns and in an attempt to increase the direct translatability of connectivity fMRI, the number of studies using fMRI in awake rodents has grown over the past few years (54,55). These investigations have shown the possibility of reliably mapping networks in awake restrained (56) or head-fixed (57) animals, with minimal stress and motionrelated artifacts. Notably, closely recapitulating analogous human and primate investigations (58,59), comparisons between network organization in awake and anesthetized rodents have shown that RSN organization is overall preserved across Brain Dysconnectivity in Rodents Biological Psychiatry --, 2022; -:---www.sobp.org/journal conditions (57,60,61), but the underlying network dynamics are profoundly altered, exhibiting stereotypical organization as a function of state. While these encouraging results support a transition to awake preparations, this procedure also comes with a few technical caveats. Awake recordings in rodents include physical body or head restraint and exposure to loud scanner-related acoustic noise. Humans are mindful of how loud the scanner is; thus, soundproof headphones are typically provided to subjects to reduce perceived noise levels. On the other hand, if animals are not thoroughly and repetitively habituated to this new environment, scanner noise may elicit strong stress responses from them. Scanner noise may not only increase the likelihood of head and body movement-a nemesis of fMRI recordings-but could also distort information processing and selective attention, preventing the onset of true resting states comparable to the quiet wakefulness attained in human imaging. The need for longer habituation regimens for animal models that are known to be more sensitive to sensory input from the environment, such as En2 knockout or Fmr1 knockout models (62), is another topic that is currently up for debate. The issue is still largely unexplored, and we advise carefully considering it and determining the appropriate practices on an individual basis.
Efforts are currently underway to create protocols aimed at optimizing habituation to the MRI environment and mitigating the stressful effects of head fixation, body restraint, and noise via gradual and incremental habituation protocols (63)(64)(65)(66). Some of these studies have shown that these procedures can result in stress-related corticosterone levels comparable to prehandling levels in both mice (57) and rats (67). In the latter species, owing to a sparse MRI sequence tuned to reduce acoustic stress, animals could even discriminate auditory stimulation from the background scanner noise (67).
Overall, this body of research points to the value of both awake and anesthetized imaging procedures provided that they are carried out with great methodological rigor and, where possible, controlling for the confounders and constraints of each technique. In our opinion, the field and methods have matured to the point where awake imaging should be regarded as the gold standard for rodent fMRI investigations and should be used in study design whenever possible. The use of anesthesia, on the other hand, can be accepted if done consistently and with proper verifications of physiological parameters and if motivated by a clear study goal, such as limitations resulting from the strain used, use of manipulations that may induce discomfort, stimulus-locked motion or physiological responses, and inability to habituate the animals (e.g., studies in pups). Consequently, the experimental choice of whether to use anesthesia-as well as the choice of the animal model-should be tailored to the research question being answered rather than the current trends

MODELING FUNCTIONAL DYSCONNECTIVITY IN RODENTS
Disrupted or atypical patterns of functional connectivity have been detected in all major psychiatric and brain disorders (68)(69)(70), supporting a conceptualization of brain pathology as the result, at least in part, of impaired brain communication (71)(72)(73). However, recent progress in human mapping of fMRI dysconnectivity has not been paralleled by increased knowledge of the mechanistic and etiological significance of these findings. The implementation of fMRI connectivity in rodents can strategically fill this knowledge gap, shedding light on the mechanistic and etiological significance of brain functional dysconnectivity (Figure 1).
Much of the added value of this field of research lies in the possibility of testing (or generating) mechanistically relevant hypotheses under tightly controlled experimental conditions that are unachievable in clinical settings. These include the control of 1) physiological and motion artifacts via sedation or head fixation (23,33,57), 2) environmental variability by breeding mice under controlled laboratory conditions, and 3) genetic variation via the assessment of genetic mutations or pathological determinants with respect to well-defined control groups composed of age-or sex-matched, genetically homogeneous littermate animals. The importance of these factors should not be understated because difficulties in controlling motion-related artifacts and in properly accounting for the genetic and demographic heterogeneity of both control and patient populations are recognized limitations of human fMRI mapping in mental illness (74)(75)(76).
Two dominant translational paradigms encompassing the use of rodent fMRI to unravel functional dysconnectivity in mental disorders have emerged over the past few years. The most widely used approach ( Figure 1A) relies on the isolation and modeling of disease-relevant genetic alterations in rodents to investigate whether and how these factors affect functional connectivity. The most advanced examples of this method have been described in the field of autism and related developmental disorders, a broad spectrum of conditions marked by high heritability and high genetic heterogeneity (77). Using fMRI in transgenic rodents, many studies have causally linked autism-associated etiological determinants to specific patterns of fMRI hypo/hyperconnectivity. The vast majority of examined factors are genetic mutations in autism risk genes such as Cntnap2 (78,79), Shank3 (80), Tsc2 (81), Fmr1 (82), Nf1 (83), Chd8 (84), and 16p11.2 microdeletion (85).
The translational relevance of this paradigm is high because it can be used to explain findings from clinical populations harboring the corresponding genetic alterations. Bertero et al. (85) used this approach to characterize similar patterns of prefrontal hypoconnectivity in a mouse model and in people with 16p11.2 microdeletion, revealing that this effect is linked to immature thalamoprefrontal wiring and diminished delta band coupling. Similarly, mTOR (mechanistic target of rapamycin)-related synaptic surplus has been shown to produce hyperconnectivity patterns that can be decoded in patients with idiopathic autism (81). Encouraging cross-species correspondences in dysconnectivity have also been reported for Nf1 deficiency (83). In our view, these investigations are critical because they can inform both preclinical and clinical researchers about the relevance of the mechanism studied and the translational value of the animal models used.
When carried out on a large scale, rodent fMRI can also help address fundamental questions related to the significance of functional dysconnectivity in brain disorders. In a recent study (86), we compared connectivity alterations across 16 distinct mouse models of autism, with the goal of assessing whether network alterations in autism converge onto a discrete network Brain Dysconnectivity in Rodents signature of dysfunction as previously hypothesized (87) or if instead they differ as a result of the etiological heterogeneity of the spectrum. The mouse models chosen for this work were based on 1) genetic modifications that resemble/relate to a genetic alteration found in individuals with autism spectrum disorder as listed in the Simons Foundation Autism Research Initiative (SFARI) gene database and 2) other autism spectrum disorder-associated etiologies such as environmental models or models for idiopathic autism spectrum disorder accompanied by an autistic-like behavioral phenotype. Our mapping revealed a broad array of connectional abnormalities in which diverse, even diverging, connectivity signatures were recognizable across models. These results reconcile highly conflicting findings in clinical populations (88), suggesting that etiological variability is a key determinant of heterogeneous dysconnectivity in autism. Moreover, they support a reconceptualization of autism dysconnectivity as the sum of distinct neurosubtypes, possibly reflecting common cross-etiological mechanisms (89). Future extensions of this paradigm to other complex mental disorders can be envisioned.
A second emerging research paradigm ( Figure 1B) relies on a broader modeling of basic pathophysiological processes associated with mental illness with the aim of identifying how each of these affects the organization of fMRI connectivity networks. Mechanistically relevant studies that can be ascribed to this category include investigations of the contribution of molecular, cellular, or environmental factors associated with brain disorders, such as impaired developmental pruning (30,90) and maternal immune activation (91), both of which linked synaptic dysfunction to fMRI dysconnectivity, or the role of chronic stress on brain network function (92). While the broad transdiagnostic nature of the mechanisms investigated with this approach prevents direct translation to clinical populations, the benefit of this paradigm ultimately lies in the mechanistic understanding of the cascade linked to altered functional connectivity and the possibility of conceptualizing fMRI dysconnectivity into a set of physiologically dysregulated components that may add and converge to produce clinical dysconnectivity ( Figure 1B).

DECODING THE SIGNIFICANCE OF DYSCONNECTIVITY VIA MULTIMODAL fMRI
Importantly, recent extensions of this approach to physiologically decode functional dysconnectivity potentially allow for reverse-translation of human fMRI datasets ( Figure 1C). Recent examples of this line of investigation entail the use of chemogenetic manipulations to probe how regional alterations in brain activity affect corresponding brainwide patterns of connectivity (93)(94)(95). Three recent studies epitomize the power Figure 1. Unraveling the determinants of functional dysconnectivity with rodent functional magnetic resonance imaging (fMRI). (A) Transgenic models can be used to isolate genetic alterations linked to psychiatric or developmental diseases (here referred to as A, B, and C). Here, rodent fMRI can serve to identify large-scale disconnection patterns associated with these mutations and, if possible, to compare the changes with corresponding human populations (85). This process forms the basis for establishing whether dysconnectivity can be further investigated in the animal model with more invasive or postmortem investigations. (B) Rodents can also be used to isolate and map the effects of known molecular, cellular, developmental, or environmental factors on brainwide patterns of connectivity. These investigations have high mechanistic relevance, and they can help conceptualize human functional dysconnectivity as the complex combination of multiple and distinct pathophysiological mechanisms. (C) Acute neuronal manipulation studies using optogenetics, chemogenetics, and concurring neural recordings can similarly help gain a basic understanding of the determinants of functional dysconnectivity in human disorders via a multimodal dissection of the basic cascade of events linking regional patterns of brain activity to brainwide fMRI coupling (96). When linked to appropriate physiological validations and computational modeling, this approach could potentially be used to reverse-translate (or decode) physiologically relevant fMRI signal metrics from patient populations into microcircuital dysfunction parameters, such as imbalances in excitatory:inhibitory (E:I) ratio (95,99). Brain renderings from panel (C) replicate design used in (146).

Brain Dysconnectivity in Rodents
Biological Psychiatry --, 2022; -:---www.sobp.org/journal of this approach. Rocchi et al. (96). recently showed that chronic or acute chemogenetic inactivation of the mouse cortex can counterintuitively lead to fMRI overconnectivity and increased delta band coupling between the inactivated regions and its terminals as a result of enhanced global oscillatory activity. This result suggests that the fMRI hyperconnectivity and increased delta power often observed in disorders characterized by loss of cortical function (i.e., stroke and degenerative disorders) may mechanistically reflect increased global oscillatory activity as opposed to rerouting of signals as previously hypothesized (97). Importantly, the same study also confirmed the prediction that inverse chemogenetic manipulations, i.e., those leading to increased excitatory-inhibitory ratio (E:I) in cortical areas, would lead to decreased fMRI connectivity via increased gamma and decreased delta activity. Similar results were previously described in another rodent study in which chemogenetically augmented E:I in somatosensory areas was found to reduce cortical fMRI connectivity (95). Notably, in the same study, the authors next trained a classifier on the recorded fMRI signals in mice and showed its ability to accurately classify cortical areas exhibiting increased E:I in a mouse model of autism. These important investigations link E:I imbalance [a postulated physiological correlate of cortical dysfunction in multiple brain disorders (98)] to a characteristic signature of fMRI dysconnectivity, thus offering additional opportunities to infer and possibly decode microcircuit abnormalities from macroscopic fMRI measurements. A compelling demonstration of the translational power of this approach has recently been described by Trakoshis et al. (99). Using chemogenetics to increase or decrease E:I in mouse cortical areas during fMRI recordings, the authors identified a time-series metric called the Hurst index, corresponding to 1/f relationship of blood oxygen level-dependent fMRI signal, that changes significantly in relation to the experimental manipulation. The same parameter could be used to decode regions with imbalanced E:I in a clinical population and revealed impairments in specific social brain regions including the medial prefrontal cortex. This research reveals the possibility, under certain circumstances, of using the fMRI signal to infer microcircuit properties of high pathophysiological significance.
Other important rodent studies have used chemogenetic or optogenetic manipulations to probe the contribution of subcortical NMSs (e.g., noradrenergic, cholinergic, serotonergic, or dopaminergic neurotransmission) to brainwide patterns of connectivity. These investigations may help disambiguate the physiological or maladaptive contribution of specific neurotransmitter systems to brainwide network dynamics with a precision unattainable in pharmacological studies, and they are a key component of research on brain dysconnectivity that we cover in greater detail in the next section.

NMS DYSFUNCTION AND BRAIN STATES IN MENTAL ILLNESSES
Ascending NMSs form the basis of many cognitive functions and endow the brain's relatively static structural architecture with flexibility, making it possible to support malleable neural dynamics required for adaptive behavior (100). Therefore, it is not surprising that NMS dysfunction is related to many psychiatric disorders characterized by persistent discomfort when adapting to new environmental, sensory, or social stimuli (101).
Until now, much of the experimental and theoretical work in this area has been devoted to determining how neuromodulatory activity is encoded in the firing patterns of target neurons (102,103). However, far less is known about how changes in single neuron firing pattern characteristics driven by NMSs propagate into large-scale phenomena, such as large-scale fMRI network activity. Recent studies have produced initial evidence that sustained neuromodulatory release exerts a powerful modulatory effect on coordinated neural activity and fMRI connectivity (104)(105)(106)(107). In humans, Shine et al. (105) have shown that brainwide fMRI responses across a range of cognitive tasks align with regional differences in the density of neuromodulatory receptors. Based on these findings, they theorized that a key function of NMSs is to coordinate the fluctuations between integration and segregation of functional networks with the aim of optimizing cognitive functioning as a response to a continuously evolving environment (108,109). Accordingly, alterations in NMSs could lead to inability to flexibly switch between states of connectivity, contributing to symptomatology or the emergence of neurological and psychiatric conditions.
Given the high level of interest in these processes coupled with the inability to dissect them in humans, research platforms that allow for controlled manipulation of NMSs are crucial for understanding their contribution to brain (dis)connectivity in brain disorders.
To date, preclinical neuroimaging research has combined fMRI and pharmacological NMS manipulations to examine the effects of various receptor agonists and antagonists on the brain's neuronal activity (110). This approach, termed pharmacological fMRI, was first used to map broad substrates of the brain that are directly activated by drugs of abuse such as nicotine (111), cocaine (112,113), amphetamine (114), ketamine (115), and psilocybin (116). These studies were later expanded to probe the receptor basis of these responses (112,117), thus laying the foundation for a fertile area of translational research (118)(119)(120). Leveraging the sensitivity of this approach, pharmacological fMRI has recently been expanded to study the functional connectivity profiles and substrates engaged by exogenously administered modulatory compounds and neuropeptides such as oxytocin (121,122), ghrelin (123), or orexin (124). Investigations of how drugs affect resting-state fMRI connectivity in rodent models have also been described (115,125). While useful in probing central engagement of drugs of interest in a fashion amenable to clinical translation, the mechanistic specificity of pharmacological fMRI to the investigation of modulatory transmission is inevitably limited by off-target pharmacological effects and possible direct vasoactive contributions of multiple drug agents (126,127). As a result, the central substrates engaged by pharmacological agents can differ from those modulated by endogenous transmitter release.
Optogenetic and chemogenetic tools, combined with advances in rodent imaging capabilities, now enable us to map the functional substrates of endogenous modulatory activity without the inconvenient contribution of vasoactive or peripheral pharmacological effects (Figure 2) (128). For example, Giorgi et al. showed that cell type-specific chemogenetic activation of 5-HT (serotonin) cells led to a specific pattern of fMRI activation of corticohippocampal, ventrostriatal, and cerebellar areas. In contrast, pharmacologically increasing serotonin levels resulted in widespread fMRI deactivation, reflecting a combination of central and peripheral vasoconstrictive effects (126). Other studies have used similar approaches in animal models to show that manipulation of dopaminergic neurons in the ventral tegmental area and substantia nigra (129)(130)(131)(132) and their targets in the striatum (133,134), serotonin neurons in the dorsal raphe nucleus (135), cholinergic neurons in the basal forebrain (128,136,137), and noradrenergic neurons in the locus coeruleus (138,139) can lead to brainwide activity changes measured by cerebral blood volume or blood oxygen level-dependent fMRI, which in turn could alter functional connectivity within specific networks and regions of interest.
One highly relevant mechanism through which NMSs could dynamically shape brainwide functional connectivity is via alterations of spontaneous neuronal ensemble dynamics from a synchronous to an asynchronous state and vice versa. High asynchronous dynamics would lead to lower connectivity but at the same time to stronger brain responses to incoming stimuli as background noise is reduced, thus increasing signalto-noise ratio. Conversely, a state of high global synchronicity would elevate functional connectivity but at the cost of reducing the selective response to external stimuli. Evidence for this modulatory role has been shown by Lottem et al. (140), who demonstrated that optogenetic activation of the serotonin dorsal raphe nucleus can rapidly inhibit spontaneous fluctuation in the olfactory cortex in mice, effectively increasing activity related to incoming sensory responses. This mechanism would be consistent with the data of Grandjean et al. (135) in which stimulation of dorsal raphe nucleus evoked a reduction in cortical blood volume mirrored by suppression of intrinsic delta oscillations. Other neurotransmitters may act in a similar way. For example, Meir et al. (141) showed that electrical or optogenetic activation of the cholinergic system is able to shift cortical activity to an asynchronous state, which improved sensory responses. These effects of NMS could at least partially explain the results of a systematic meta-analysis on working memory tasks in patients with schizophrenia or major depressive disorder, which found common stronger fMRI responses in prefrontal and anterior cingulate cortices, 2 regions belonging to the DMN (142). In contrast, 2 independent studies showed that activation of the locus coeruleus-norepinephrine system increases low-frequency synchronous fMRI connectivity within multiple cortical networks, including the DMN (138,139). The strength of this reconfiguration was found to be spatially correlated with a 1-2 and b 1 adrenergic receptor transcript levels and norepinephrine turnover levels, corroborating the hypothesis from human stress research that locus coeruleus activity and norepinephrine release are causally linked to fMRI network integration (143)(144)(145). Overall, these studies show that fMRI connectivity is strongly constrained by underlying neuromodulatory tone. Studying these links is an opportunity to shed light on the elusive contribution of maladaptive NMS function to brain dysconnectivity in psychiatric and neurological disorders.
CONCLUSIONS fMRI applied to rodents offers a privileged angle of investigation from which to explore the origin and significance of brain dysconnectivity at different levels of inquiry. We urgently need to break fMRI dysconnectivity down into a number of physiological processes that can be mechanically explained, such as the function of NMSs in mental illness. Multimodal imaging must be strongly promoted in this case. We further advise that the field moves toward longitudinal and awake imaging studies to examine the relationships between etiological factors and their chronobiological effects on brain connectivity. In addition, we must keep fostering research domains whereby animal and human networks can be comparable. Understanding which circuits and networks exist in both species is essential for choosing research questions and hypotheses that arise from clinical research on patients. Leveraging emerging correspondences in the organization of fMRI networks across the phylogenetic tree, the impact of this versatile research platform toward a better understanding of human brain function is  Figure 2. Multimodal rodent functional magnetic resonance imaging (fMRI) to study the contribution of neuromodulations to (dys)connectivity. One important process attributed to neuromodulatory systems (NMSs) is promotion of dynamic adjustments in behavioral states. For example, switching from a quiet, inattentive state to an aroused, vigilant state is attributed to a sustained increase in locus coeruleus-norepinephrine firing (147), while changes in motivational vigor have recently been correlated with tonic firing of dopaminergic neurons in the ventral tegmental area (148). Evidence from both human and rodents suggests that these behavioral states are paired with internal brain connectivity states, a term used to describe brainwide and timevarying patterns of global neural activity that can be captured by whole-brain functional imaging modalities, such as fMRI. Dynamic and flexible changes in connectivity states-under direct control of NMSs-allow anatomically defined circuits to give rise to many different patterns of (co)activity, which is crucial for adaptability of neural processing in different behavioral contexts but often impaired in psychiatric or neurological disorders. One of the key advantages of rodent fMRI is the ability to perform direct manipulations of NMSs using tools such as chemo-and optogenetics. Compared with behavioral or drug challenges in humans, this has the advantage of minimizing the influence of internal beliefs and perception biases that are present in traditional human fMRI studies and cause heterogeneity in the observed responses. In addition, it is possible to study the mechanisms that form network state changes using more invasive recordings. This gives rodent fMRI the unique ability to dissect the contribution and the role of NMSs to whole-brain (dys)connectivity patterns.

Brain Dysconnectivity in Rodents
Biological Psychiatry --, 2022; -:---www.sobp.org/journal substantial, and it is expected to grow rapidly in the coming years. We believe that the current research marks the beginning of a new chapter for functional rodent imaging and that new routes for integrating preclinical and clinical data through direct comparisons or computational models will lead to a better understanding of the mechanism(s) underlying dysconnectivity in mental illness.