Locus Coeruleus in Non-Mammalian Vertebrates

The locus coeruleus (LC) is a vertebrate-specific nucleus and the primary source of norepinephrine (NE) in the brain. This nucleus has conserved properties across species: highly homogeneous cell types, a small number of cells but extensive axonal projections, and potent influence on brain states. Comparative studies on LC benefit greatly from its homogeneity in cell types and modularity in projection patterns, and thoroughly understanding the LC-NE system could shed new light on the organization principles of other more complex modulatory systems. Although studies on LC are mainly focused on mammals, many of the fundamental properties and functions of LC are readily observable in other vertebrate models and could inform mammalian studies. Here, we summarize anatomical and functional studies of LC in non-mammalian vertebrate classes, fish, amphibians, reptiles, and birds, on topics including axonal projections, gene expressions, homeostatic control, and modulation of sensorimotor transformation. Thus, this review complements mammalian studies on the role of LC in the brain.


Introduction
The locus coeruleus (LC) is a vertebrate-specific norepinephrinergic (NE) nucleus [1][2][3][4]. The LC locates deeply in the dorsal part of the brainstem [5], where a small number of neurons extensively branch their axons and provide the main source of NE to the brain [6]. Intriguingly, along the evolutionary path, animals develop more LC cells, probably due to demand for commanding more brain nuclei and influencing more complex physiological and behavioral responses. For example, as a member of the most ancient group of vertebrates, zebrafish LC has only 10-20 NE cells [7,8]. Songbirds have~700 and quails havẽ 1300 LC cells [9]. Mammals possess more LC neurons than birds, rats have 3000 LC cells, and monkeys have 7000 LC cells [10]. The number of LC neurons correlates with the overall number of neurons across non-mammalian species. As one might expect, humans have the most LC cells, up to 50,000, to cope with arguably the most sophisticated brains [11].
Comparative studies combine observations from different model organisms. Among vertebrate species, remarkable conservation at multiple scales has been reported in LC, including a relatively homogeneous cell identity [12], broad axonal projections [13], and general roles in enhancing arousal and alertness [14]. For example, the whole-brain screening in larval zebrafish guided the discovery in rodents that multiple parallel neuromodulatory systems, including the LC, coordinately modulate brain states [15]. Among vertebrates, the shared traits of LC across the molecular, synaptic, circuit, and system levels make this highly complex brain nucleus an ideal system for comparative analysis.

Telencephalon
In mammals, the LC broadly innervates multiple regions in the telencephalon to mediate attention, arousal, and sensory processing [12,[48][49][50]. Similar projection patterns exist in non-mammalian vertebrates. In fish Polypterus palmas, telencephalon receives bilateral LC projections [18]. Within the telencephalon of zebrafish Danio rerio, the olfactory bulb, subpallium, striatum, midline septum, and anterior commissure are targeted [19,20]. In newt Triturus cristatus, lizard Varanus exanthematicus, and frog Rana perezi, LC fibers extend from the isthmic segment to the striatum and septum [21][22][23]. Within the telencephalon of the pigeon Columba livia, projection targets of the striatum can be further divided into the paleostriatum augmentum, neostriatum, archistriatum, and hyperstriatum [31]. Striatum and septum in telencephalon are the common targets of the LC among vertebrates, whereas the projections to the cortex evolve as new ascending pathways to the telencephalon. The LC neurons project to the dorsal medial cortex (DCm) in lizard Gekko gecko [24], and the secondary auditory cortex (the caudomedial nidopallium, NCM) and the wulst in zebra finch Taeniopygia guttata and pigeon Columba livia [33,34], indicating their involvement in species-specific functions such as generating bird songs.

Diencephalon
The common targets of LC axons in the diencephalon are the thalamus and the hypothalamus in zebrafish Danio rerio, lizard Varanus exanthematicus, and pigeon Columba livia [20,25,32]. In pigeons, the LC projects to the limbic system and striatal complex, overlapping more with dopaminergic efferents than in mammals [32]. Moreover, all the major NE receptors are present in the amygdala and the hippocampus, confirming the impact of the LC-NE system on emotions, learning, and memory [41-43].

Mesencephalon
Optic tectum (OT) proves to be the biggest target region of LC NE projections in the mesencephalon of zebrafish Danio rerio, frog Rana perezi, frog Xenopus laeuis, and pigeon Columba livia [13,20,27,28,36]. These projections form pathways that contribute to the visual alertness in zebrafish and the dark/light adaptation in amphibians [15,27,28]. The arborization field of LC neurons in the OT expanded during the course of evolution. In fish, the LC only spreads a few collaterals extending caudally and ventrally to the OT [13]. In amphibians, the LC projects predominantly to the superficial, recipient tectal layers [28]. In birds, the LC projection arborizes densely in layers 2, 4, and 7, suggesting a potent modulation of visual processing [36]. Apart from the OT, the substantia nigra (SN) in frog Xenopus laevis and chicken Gallus domesticus receives LC projections [26,35]. A similar pathway from the ventral tegmental area (VTA)-DA system to the LC-NE system exists in mammals, supposedly conveying the reward prediction signal from the dopaminergic system to the NE system [51]. However, such DA to NE pathway is not identified in fish. Therefore, it might have appeared later in the course of vertebrate evolution.

Rhombencephalon
In the descending pathways, projections from the LC first enter the cerebellum. In zebrafish Danio rerio and turtle Pseudemys scripta elegans, LC axons enter the cerebellum, with little path or region specificity [13,20,29]. In chicken Gallus domesticus, however, the LC forms region-specific projections to the cerebellum: the caudal LC forms a latero-caudal bundle in the inferior cerebellar peduncle, while the rostral LC forms a medio-rostral bundle in the superior peduncle [35,38].

Spinal Cord
In all the vertebrates, the LC substantially projects to the spinal cord [20,30,37,[45][46][47]. In mammals, LC projections inhibit sensory neurons in the dorsal horn [52,53] and activate motor neurons in the ventral horn by acting on different types of NE receptors [2,39,44,54]. The LC also projects to sympathetic and parasympathetic preganglionic neurons located in the intermediolateral cell column (IML) of the spinal cord and exerts intricate effects involving both excitatory and inhibitory depending on the NE receptor types [55][56][57]. In fish, the TH-reactive fibers are observed to enter the spinal cord, predominantly through the ventrolateral pathway [58]. While there is little evidence on the impact of LC in any specific motor behaviors, those TH+ fibers may form a close connection with motor neurons in the spinal cord and influence any motor-related behaviors.

Future Directions for Mapping LC Wiring Diagram
The above studies show conserved projections of the LC among vertebrates. However, these projections have been mapped at the populational level by anterograde/retrograde dye tracing [32,37,41,44,45,[59][60][61], or immunostaining [13,[62][63][64]. Next, a single-cell level wiring diagram will be important. For example, although neuronal ensembles fire asynchronously in the LC [65], it is unknown if this heterogeneity is maintained in downstream areas by "specific" projections (groups of LC neurons preferentially target different brain regions), or erased by "uniform" projections (different LC neurons randomly project to brain regions) [12,[66][67][68]. Two promising technologies include sparse labeling and highthroughput sequencing. Examples are the MouseLight project carried out by Janelia Research Campus, reconstructing~1000 projection neurons in the mouse brain [69], and the Multiplexed Analysis of Projections by Sequencing (MAPseq) developed by Anthony Zador's lab at the Cold Spring Harbor Laboratory [68].
Neither the MouseLight project nor the MAPseq project can completely reconstruct the LC wiring at single-cell resolution due to low throughput or lost spatial details. However, studies in larval zebrafish may provide the first complete dataset for LC wiring. In larval zebrafish, within a short time period, hundreds of LC neurons can be reconstructed and mapped onto a common brain [70][71][72]. Because the number of neurons in larval zebrafish is two orders of magnitude less than that in mice, such datasets will oversample the LC and lay a solid foundation for further characterization. The projection profile of a single LC neuron will be covered as well. Next, with mature technology on brain-wide functional imaging, neuronal activities, and NE release can be integrated into this projection database [73][74][75][76]. Together, a complete LC functional projectome will substantially advance the understanding of LC systems in vertebrates.

LC Development
These complex projections are established from the morphogenesis of the LC during early development. In birds, the formation of the LC generally takes three steps-starting with loose neuroblasts, followed by the distinction between the dorsal part and ventral part during the incubation stages, and finally are equipped with rich and specific projections at the age of 6 weeks [5]. The branching pattern of dendrites and the volumetric alterations of individual LC neurons continue to increase during the last days of embryonic development and after hatching [5]. Similarly, in fish, amphibians, and birds, all the neuronal populations and most of the projections are detected by the end of the embryonic period [7,[77][78][79][80]. Notably, the LC cells in embryos are always observed in positions corresponding to their adult location, suggesting that they do not migrate much during development [81].

Unique Gene Expression in LC
A unique gene expression profile gradually unfolds in LC. In mammalian LC, more than 3000 genes are highly expressed and more than 100 genes show significant sex differences [82,83]. In zebrafish, a mutant line called soulless completely loses LC NE neurons, leading to the discovery of the master control gene for generating these neurons. The homeodomain protein Phox2a integrates the signal from morphogens such as bone morphogenic protein (BMP) and fibroblast growth factor-8 (FGF8), in turn, triggers the expression of NE synthesis enzymes, tyrosine hydroxylase (TH) and dopamine β-hydroxylase (DBH), and generates functional LC [84].
The soma size of LC NE neurons is generally larger than the neurons in the immediate vicinity [17,58,81,85]. This phenomenon could be related to a high expression level of nerve growth factor (NGF) receptors in the LC [86]. In chicken, LC neurons express a high level of neurotrophin receptors TrkA and p75, and in the presence of neurotrophin NGF, these neurons grow bigger and are rescued from the toxic effect of 6-hydroxydopamine [86]. Interestingly, two parallel routes were identified for trafficking the NGF from the richest source (such as the retina) to LC soma. One pathway is transporting NGF by axon tracts, anterogradely by other neurons (such as retinal ganglion cells), and retrogradely by LC axons. The second pathway is through a specialized glial/ependymal cell type, the tanycyte, which extracts NGF from the cerebrospinal fluid (CSF) and delivers it to the LC [87]. However, in mammals, the LC NE neurons do not transport NGF and lack neurotrophin receptors. Therefore, they do not respond to NGF [88]. Such differences in gene expression may cause divergence of projection patterns and circuit functions between species.
Unique gene expression profile gives rise to distinct cellular and circuit functions. Another example comes from metal transporters. The LC contains the highest concentration of copper than any other brain regions [89]. On the one hand, copper is required for metalating DBH to convert DA into NE [90]. On the other hand, copper could trigger cellular stress from its high redox activity and is associated with Parkinson's disease and other neurodegenerative diseases [91,92].
A recent study in zebrafish led to a more thorough and surprising finding of the physiological relevance of copper to the LC activity [93]. Copper Fluor-4, a synthesized copper sensor, was developed and used to detect the copper level in vivo. Using this tool, Xiao et. al. found that copper distribution was heterogeneous in the larval zebrafish brain, and a high level of copper in LC was critical for rest-activity balance and arousal responses. The concentrated copper depends on a protein highly and specifically expressed in LC-ATP7A, a copper transporter [93,94]. Phylogenetic analysis showed that the gene atp7a comes from gene duplication in early chordates. Interestingly, the NE transporter, NET, also appears first in Gnathostomata. Such parallel evolution provides evidence that the LC-NE system was restricted in vertebrates. In Parkinson's disease patients, decreased copper levels were found in both the substantia nigra and the LC [95], and in Wilson's disease, mutations in the copper transporter ATP7B may lead to copper accumulation in patients [96]. In both diseases, copper-related treatments such as a low-copper diet or copper supplementation are proposed as therapeutic strategies [36,97]. Due to the specificity of copper transporter expression and concentrated copper in the LC, tools monitoring copper or copper-related biomarkers will be useful for measuring LC functions and diagnosing neurodegenerative diseases.

LC Regulates Body Homeostasis
Through the extensive axonal projection network, the LC-NE system modulates a broad spectrum of brain functions in mammals and non-mammalian vertebrates alike. For instance, the LC mediates transitions of brain states between sleep/anesthesia and wake/alertness to sustain internal homeostasis. In larval zebrafish, the normal activity of the LC maintains the awake state and promotes the sleep to wake switch triggered by sensory stimuli [98]. A similar effect was observed during general anesthesia. General anesthetic drugs decrease LC neural activity by reducing both its intrinsic excitability and synaptic inputs. Impairing LC delays the recovery from the anesthetic state [99].
The LC NE neurons are directly thermosensitive. In slice recordings from amphibian (frog Lithobates catesbeianus) LC neurons, when the temperature fluctuates within a naturalistic range (from 10 • C to 30 • C), the basal firing rate decreases during warming and increases during cooling [100,101]. Hyperpolarization-activated current (I h ) is proposed to mediate cooling-induced firing rate increase in amphibians [101]. In mammals, though unclear in LC, neuronal firing rate from the slice recording of other brain regions, such as dorsal raphe and hypothalamus, is positively correlated with temperature fluctuation [102,103]. Moreover, although deep brain temperature is much less affected by environmental changes, pyrogen-induced fever or ambient heating could nonetheless increase the firing rate of the LC NE neurons in cats [104]. To address the difference between endotherms and ectotherms in brain thermo-environment, it would be informative to test if LC neurons in mammals are also thermosensitive and if the same I h current responds to temperature.
Furthermore, the LC NE neurons are chemosensitive [105,106]. In amphibians, reptiles, and mammals, the firing rates of LC neurons increase in more acidic environments [100,105,[107][108][109]. In response, enhanced LC activity promotes respiration [105,110], which reduces the CO 2 concentration and increases the body temperature [111][112][113]. Unlike in the endotherms, the CO 2 /pH in ectotherms fluctuates with body temperature. An intriguing phenomenon is that acid-base homeostasis is maintained even when the CO 2 /pH changes with temperature [114]. This control process is potentially mediated through a compensatory mechanism involving the LC. In amphibians, the chemosensitivity of LC is not constant but rather positively correlates with temperature [100], allowing for adaptive respiration regulation that better responds to the fluctuation of body temperature.
Further investigations should answer two key questions. First, from ectotherms to endotherms, how have these homeostatic regulation functions been preserved, lost, or transformed. Second, could the LC neurons receive environmental signals through either local sensors in LC neurons [102] or afferent connections carrying information from sensory organs [115]. Distinguishing and combining these two input sources would advance the understanding of LC NE function in homeostatic control.
Neural function, body temperature, and bodily fluid acidity depend on each other and together constitute a system that dynamically adapts to the environment [100,[115][116][117][118][119]. The LC NE neurons integrate multiple sources of homeostatic signals and provide feedback control through homeostatic mechanisms such as respiration (Figure 2). The above mechanisms allude to a broader role of LC in evaluating environmental feedback to optimize behavioral strategies and preserve body homeostasis. In both non-mammals and mammals, the decrease of pH enhances LC firing rate [108,109], which in turn promotes respiration and reduces retaining CO 2 [111]. Apart from this chemosensitivity-based negative feedback, the thermosensitivity of LC forms a regulatory pathway with respiration and temperature changes. In ectothermic non-mammals, the environmental temperature negatively regulates the basal firing rate of LC, but positively modulates the chemosensitivity of LC [100,101,105,107,120]. In mammals, body temperature change and LC activity reciprocally interact with each other [104,112,113,121,122]. Direct thermosensitivity of LC has been shown in non-mammals [100] but remains elusive in mammals. Created with BioRender.com.

LC Modulates Sensorimotor Transformation
During sensorimotor transformation, three major components-LC activity, sensory processing, and motor generation interact with each other and form a dynamic system. Behavioral paradigms involving continuous sensorimotor streams, such as songbird singing or zebrafish swimming, are especially suitable for delineating these interactions. In these paradigms, LC-NE signals modulate multiple sensorimotor areas to produce flexible and adaptive behaviors.
Songbirds recognize and prefer conspecific songs for mating and vocal tutoring. The song generation system in songbirds receives wide LC projections [9], which provide NE signals that modulate both sensory processing and motor generation. In the sensory processing pathway, the LC modulates the activity of the caudomedial nidopallium(NCM), a region analogous to the mammalian secondary auditory cortex. Upregulating the NE signal in NCM enhances signal detection by reducing spontaneous firing rates and increasing the signal-to-noise ratio [123]. On the other hand, downregulating the NE signal by inducing lesion at LC disrupted the selectivity between the bird's own song, conspecific songs, and heterospecific songs [124]. In the motor generation pathway, the NE signal modulates the singing states. Increasing NE signal in the nucleus of the arcopallium (RA), a motor nucleus critical for song generation, promotes the production of a robust song ("exploitation state", in the presence of females), whereas decreasing NE enhances the production of variable songs ("exploration state", singing alone) [125]. Similarly, directly stimulating the LC neurons promoted "exploitation state" [125], potentially by modulating the song variability pathway: basal ganglia-anterior nidopallium (LMAN)-RA [126,127]. In this pathway, the NE signal suppresses the firing of spiny neurons in the basal ganglia and the input from LMAN to RA [127,128].
These studies demonstrate the influence of LC-NE on sensory and motor circuitry. The system is further confounded by motor feedbacks on sensory input. Novel environment or unexpected sensory feedback triggers bursting activity in the LC [129]. Conversely, the LC modulates sensory processing, reflected by enhanced alertness or arousal and altered synaptic plasticity [15]. A similar reciprocal influence exists between the LC and motor activity. A theoretical framework was established in the last decade, proposing that the phasic activity of LC is an interrupt signal that represents "unexpected uncertainty", and, in turn, modulates sensorimotor transformation [130][131][132]. This process may be the foundation of various forms of adaptive behaviors. However, such as the "three-body problem", interactions that are easy to describe between two subjects could appear chaotic when three or more subjects are involved. Simultaneously recording from, and perturbing the sensory, motor, and the LC systems could enable quantitative modeling that would greatly enhance our understandings.
A study on swimming generation in zebrafish considered the three components in the same context. Larval zebrafish are placed in a virtual reality environment in which visual feedbacks reflect intentions to swim [75]. Appropriate feedback encourages zebrafish to continue swimming. The system allows for fast and repeated adjustments of feedback strength. Alternatively, there could be no feedback at all, in which case the swimming attempts would reduce or cease. During this sensorimotor loop, motor outputs and sensory feedback are constantly evaluated to generate "expected" and "unexpected" signals. When visual feedback is mismatched with motor command, the LC activity increases, representing an "unexpected" situation [75]. This phenomenon is consistent with the findings in mammalian studies that the LC is sensitive to the sensory feedback that deviates from expectation. Interestingly, the LC activation in zebrafish only encodes negative or "punishing" deviation when feedback is worse than expectation [75]. In contrast, in mammals, the LC NE neurons can encode reward prediction signals [133], which might be conveyed by projections from VTA to LC [51]. The VTA-LC pathway could have later evolved, but analogous circuits might exist in zebrafish. It would be informative to verify possible projections from DA neurons to the LC in zebrafish.

Conclusions
LC-NE, a vertebrate-specific neuromodulatory system, projects to the majority of brain areas. The broad LC projection forms a global network and regulates the neural activity locally. This local regulation targets both neurons and non-neuronal cells, such as astrocytes, and induces different functional changes depending on the difference of targeted cell types, brain regions, and animal classes. For example, astrocytes in the hippocampus of mammals, mediating NE induced place cell recruitment around the reward location [134][135][136], while astrocytes in the medulla oblongata of zebrafish, mediating NE induced regulation of the motor strategy [75]. The local specific modulation and its behavioral relevance can be highlighted and thoroughly addressed by utilizing different behavioral paradigms from different animals. Moreover, the widespread projections and local specific modulation of the LC-NE system can serve as a model for illustrating the organization principles of neuromodulation at the whole-brain scale.
We could imagine the LC-NE system as a tree (Figure 3). The widely projecting axons form branches broadcasting signals globally. Along each branch, the leaves are formed by local coordinating neural elements such as astrocytes, and the fruits of specific neural functions emerge from local coordinated NE neuromodulations. Under this framework, a whole-brain-wide survey from three aspects would be critical: (1) to characterize the organization principles of the "tree branches," mesoscale single-cell reconstruction needs to be conducted. For example, brain regions could be functionally clustered, depending on if they are targeted by the same or different group of LC neurons. To deduce such principles from single reconstructed neurons, an oversampled number of cells would be necessary. Sparse labeling technique in zebrafish is widely used (200 LC neurons, 10 times the total LC neuron number) [137], and newly developed high-throughput imaging and reconstruction methods could potentially reconstruct 20,000 LC neurons in mice.
(2) to decipher the communication rules between "branches" and "leaves," simultaneously monitoring multiple neural components could outline the dynamics emerging from global to local interaction. The targets to be monitored include activity of LC soma and axon terminals, NE release from axons, and activity of local neurons and astrocytes. (3) to identify the "fruits" emerging from each branch, relevant behaviors need to be measured with simultaneous neural imaging, as mentioned in (2). Evidence accumulated from these three aspects will lead to a comprehensive description of the LC. Elucidation of this "tree" structure will provide a thorough understanding of LC and provide new insights regarding how a brain nucleus works with its global targets (Table 1).  Ascending and descending pathways. Zebrafish, Danio rerio [13,19,20,79,137] Circuit: hypocretin(Hcrt) and melanin-concentrating hormone(MCH) neurons project to the vicinity of the LC.