Danio rerio, the zebrafish, has been a rising star within neuroscience since it was championed by George Streisinger starting in the 1970s (Varga, 2018). Though it was just one of many bony fish models used at the time, it has left the world of “emerging model organisms” to become an example of successful widespread adoption of a “new” model. The optical clarity of zebrafish during embryonic and larval stages offers unprecedented access to imaging for a vertebrate, and its reduced size means that imaging whole brain volumes is increasingly achievable. Their potential as a disease model is good, with 82% of human disease genes have a zebrafish ortholog (Howe et al., 2013). Zebrafish spawn in large clutches of ∼100 embryos, making them popular for large scale pharmacological, genetic, and behavioral screening. To this end, multiple well-established methods exist for genetic manipulation in zebrafish, including CRISPR/cas9 editing and transgenesis via Tol2 mediated insertion (Kawakami, 2005; Kwan et al., 2007; Don et al., 2017; Liu et al., 2019). However, it is worth noting that zebrafish have ∼30% of their genome duplicated, which can complicate the creation of mutant or knockout animals as multiple genes may need to be targeted (Woods et al., 2000, 2005; Howe et al., 2013). Existing zebrafish mutants and transgenic lines can be acquired easily through an international repository (ZIRC), and a large library of phenotypically interesting mutants already exist thanks to an important large scale ENU mutagenesis effort (Driever et al., 1996).

Aiding in their utility as a genetic model, zebrafish are quick to develop. They reach juvenile status at 30 days post-fertilization and become sexually mature adults at ∼90 days post-fertilization. Mice, in comparison, reach sexual maturity at ∼55 days post-fertilization, making them slightly faster to mature but relatively comparable. Zebrafish are generally kept for 2 years in lab housing facilities but can survive for as long as 5 years. Their lifespan along with detectable aging-associated changes make them a model with great potential for addressing neurodegenerative and aging-associated diseases (Keller and Murtha, 2004; Kishi, 2004; Adams and Kafaligonul, 2018).

Zebrafish have already proven useful models for neuroscience because of their small and accessible brains. The structure and composition of the zebrafish nervous system shares much in common with that of humans, including a general architecture of a spinal cord, hindbrain, midbrain, and forebrain and predominantly similar neurochemical components (Figure 1; Kimmel, 1993; Bae et al., 2009; Panula et al., 2010). One notable anatomical difference between mammalian and teleost fish brain structure is that fish lack a cerebral cortex. All mammalian CNS glial cell types are also found in zebrafish, namely, oligodendrocytes, astrocytes, and microglia (Kawai et al., 2001; Yoshida and Macklin, 2005; Avila et al., 2007; Lyons and Talbot, 2014; Xu et al., 2015; Chen et al., 2020).

In addition to brains that are largely comparable to mammals, adult zebrafish also have a host of complex behaviors that parallel many established murine tests of learning and memory, anxiety, and social preference (Orger and de Polavieja, 2017). The similarities between zebrafish and mouse behavioral assays allow comparisons to be made between the substantial body of literature in mice and new work in zebrafish.

Pairing behavioral assays with live imaging has immense potential for the study of neurological disease, especially neurodegenerative diseases that are known to cause motor, learning, and memory deficits. These kinds of experiments are increasingly common, though technical challenges limit their widespread adoption. One such caveat is that imaging behaving larvae generally requires stabilization, either through mounting or paralysis through genetic or chemical means. This stabilization reduces much of the natural feedback received by the fish while enacting the behavior. Stabilized or partially stabilized larvae have been used with multiphoton and LSFM microscopy to image genetically encoded calcium indicators (GECIs) in opto-motor control, auditory and visual response, olfaction, and prey-capture decisions (Förster et al., 2020; Wang et al., 2020; Bruzzone et al., 2021; Herrera et al., 2021). The introduction of adaptive imaging techniques and real-time correction for movement allow for reduced levels of restraint while improving image quality (Wang et al., 2014; Rodríguez and Ji, 2018; Griffiths et al., 2020). This leaves the ability to image reaction to virtual stimuli, but without natural feedback from the environment or sensory systems such as the lateral line or vestibular system. Real-time virtual feedback stimulation can be provided in response to active behaving (Portugues and Engert, 2011; Ahrens et al., 2013; Trivedi and Bollmann, 2013; Bianco and Engert, 2015; Jouary et al., 2016; Kawashima et al., 2016; Naumann et al., 2016; Huang et al., 2020), but great interest still remains in functional neuronal imaging of freely swimming fish.

Currently, multiple solutions for imaging of freely behaving larvae exist, most with widefield fluorescence illumination and detection combined with methods to perform tracking and predictive adjustments of optical elements (Cong et al., 2017; Kim et al., 2017; Symvoulidis et al., 2017). 2-photon and LSFM solutions to image true freely behaving larvae are technically difficult to implement as they require precise direction of the illumination area. Despite their relative speed compared to other techniques, acquiring whole brain volumes using 2-photon or LSFM takes additional time compared to a single widefield acquisition, making it slow in comparison to the speed of a moving larva. Light field microscopy presents an alternative approach, providing speed at the sacrifice of spatial resolution (Levoy et al., 2006). Confocal light field microscopy has recently been implemented to improve resolution for volumetric imaging of neuronal GECIs in freely swimming larvae (Zhang Z. et al., 2021).

Both scanning and spinning disk confocal microscopy have been standard methods in the field for imaging cellular dynamics in development and disease in the zebrafish. Since its introduction in 2004, however, LSFM has revolutionized imaging cellular dynamics in embryonic and larval zebrafish, particularly during development. The use of mounting methods that preserve the 3D integrity of the sample is an important principle in LSFM, and zebrafish embryos can be mounted for LSFM imaging purely in water using FEP tubes and an agarose plug (Power and Huisken, 2017). The ability to mount and image without using a constrictive mounting medium allows for a much more natural progression of shape change and morphogenesis during development. This can be combined with adaptive imaging techniques to keep embryos in focus and within the field of view over many hours or days of development (McDole et al., 2018; Daetwyler et al., 2019; He and Huisken, 2020). Little work has been done specifically using LSFM to look at cell-scale phenomenon, like glia and the NVU in neurodegenerative disease in zebrafish. However, foundational work has been completed in characterizing lymphatics, vasculature, astrocytes, and microglia in the zebrafish brain using both LSFM and standard confocal microscopy.

Vascular imaging in zebrafish is a relatively mature field with established transgenic lines and existing protocols for the segmentation and quantification of complex vascular networks, including cerebellar vasculature (Chen Q. et al., 2012; Daetwyler et al., 2019; Kugler et al., 2020, 2021). A newly published zebrafish transgenic line marking astrocytes will prove a valuable resource to illuminating the role of astrocytes in development and disease (Chen et al., 2020) and could be combined with existing transgenic lines (Lawson and Weinstein, 2002) marking vasculature to better understand interplay between vascular systems and glia in the NVU.

Use of the zebrafish model has also enabled imaging of the glymphatic network of the CNS, a traditionally difficult imaging subject due to its complexity and location. Castranova et al. (2021) achieved live imaging of immune cells within meningeal lymphatics in larval and juvenile zebrafish brain. Zebrafish have thin skulls, and when coupled to a pigmentless strain they were able to image the meningeal lymphatics in intact animals through the skull using spinning disk confocal. This work shows the promise of non-invasive cellular-level imaging in the intact juvenile zebrafish, and positions zebrafish as a model to understand the interplay of the glymphatic network with neurodegenerative diseases.

Microglia in zebrafish are a rapidly expanding topic of interest–more papers have been published between 2019 and 2021 on microglia in zebrafish that in the preceding 12 years combined. Zebrafish microglia are generated in two waves, with a primitive wave derived from the primitive macrophages in the yolk sac, and a second wave that replaces the original population and originates from the hematopoietic stem cells of the dorsal aorta (Xu et al., 2015; Ferrero et al., 2018; Wu et al., 2020; Silva et al., 2021). Zebrafish microglia share a core transcriptional signature with mammalian microglia and are morphologically and functionally similar to microglia of other species (Oosterhof et al., 2017; Geirsdottir et al., 2019). Imaging studies in zebrafish have been used to show cross talk between neuronal activity and microglial morphology, phagocytosis of myelin sheaths and dying neurons, and response to long-range signaling cues (Peri and Nüsslein-Volhard, 2008; Li et al., 2012; Sieger et al., 2012; Mazaheri et al., 2014; Hughes and Appel, 2020). Two populations of zebrafish microglia have been observed: ameboid microglia, which have a higher capacity for phagocytosis, and the white-matter enriched ramified microglia, which actively remodel their protrusions and may play a more regulatory role (Wu et al., 2020). This is an interesting distinction because humans have heterogenous microglia populations, while other studied model organisms have been found to have more homogeneous microglia populations (Böttcher et al., 2019; Geirsdottir et al., 2019; Masuda et al., 2019; Sharma et al., 2021). If zebrafish microglia more closely resemble the heterogeneity of human microglia, they could fill an important niche in our understanding of microglia in human disease.

Embryonic and larval zebrafish have multiple subtypes of neuron with axons that grow superficially enough to allow for detailed imaging of sub-cellular structures at excellent resolution. The posterior lateral line (pLL) axons and the Rohon-Beard (RB) sensory neurons are two such commonly imaged populations. Imaging of microtubules (Stepanova et al., 2003; Asakawa and Kawakami, 2010; Lee et al., 2017; Haynes et al., 2021), mitochondria (Plucińska et al., 2012; Campbell et al., 2014; Paquet et al., 2014; Auer et al., 2015; Drerup et al., 2017; Mandal et al., 2018), ER-Mitochondria contact sites (Cieri et al., 2018; Vallese et al., 2020), endosomes (Clark et al., 2011; Ponomareva et al., 2014, 2016), and lysosomes (Drerup and Nechiporuk, 2013) is common in these superficial neurons in zebrafish embryos and larvae. When paired with genetic manipulation of disease-associated genes, zebrafish are an excellent tool to discern the molecular mechanisms behind basic disease processes.

Spinning disk and swept-field confocal microscopes have traditionally been the tool of choice to image fast-moving sub-cellular structures like microtubule comets and organelles. Zebrafish sensory axons, for example, can have axon arbors spread through a volume of 20–50 μm. That volume must be obtained in seconds or less to achieve adequate speed for measuring sub-cellular dynamics. However, progressive improvements in high NA light-sheet microscopes are enabling them to break into the sub-cellular niche. Lattice light sheet is difficult to implement but able to resolve objects of 200 nm, meaning that structures like Golgi and ER were able to be resolved in the brain of a living zebrafish embryo (Liu et al., 2018). Single objective light sheet as well as super-resolution techniques like STED have yet to be applied for sub-cellular imaging in zebrafish, however, the potential for their utility is high.

With an array of tools and techniques available for multiple ages, zebrafish are growing into their potential as an excellent model for neurodegenerative disease research. Disease-relevant mutations can be easily introduced through CRISPR/Cas9 editing or transgenic insertion, allowing the study of both the native function of the mutated gene as well as the disease-relevant consequences of its loss or mutation. Numerous disease models for neurodegenerative disorders including PD, AD, and hereditary spastic paraplegia (HSP) have been generated in zebrafish through genetic or non-genetic means (Fontana et al., 2018; Vaz et al., 2018; Naef et al., 2019; Wang and Cao, 2021).

Many of these diseases are complex and polygenic, making them difficult to model. For example, there are multiple genes associated with familial risk for PD, though most cases are spontaneous. In humans, PD results in neurodegeneration of the dopaminergic neurons and is accompanied by aggregation of the protein α-synuclein. PD can be modeled in the zebrafish through drug treatments that decrease dopamine levels or numbers of dopaminergic neurons or through genetic knockdown or knockout (Vaz et al., 2018). The role of synuclein genes and other PD associated genes have been studied in zebrafish and have been associated with defects in the dopaminergic system, mitochondrial dysfunction, and neurodegeneration (Milanese et al., 2012; O’Donnell et al., 2014; Vaz et al., 2018).

Alzheimer’s disease has been more difficult to model in zebrafish. A standard model of AD in mice is the 5xFAD mouse, which uses transgenic insertion of amyloid precursor protein (APP) and presenilin1 (PSEN1) with a constellation of 5 mutations associated with human familial AD (Oakley et al., 2006). An equivalent to this model has not been generated in zebrafish, though perturbation of presenilin-1 and −2 genes and zebrafish APP genes appa and appb have been widely used to test the role of these genes in development and disease (Kaiser et al., 2012; Luna et al., 2013; Fontana et al., 2018; Banote et al., 2020). Aβ peptide, the product of APP cleavage, has also been used directly in zebrafish to study the role of different length Aβ on biological processes and Aβ toxicity. Cerebroventricular injection of Aβ-42 showed AD-like phenotypes including neuronal cell death and microglial activation in both young and aged animals (Bhattarai et al., 2016, 2017). Microglia have been implicated in pathways controlling neuroproliferation in the zebrafish in response to Aβ injection (Bhattarai et al., 2020), though no live cell imaging of glia or neuronal cell interaction with plaques have been performed in zebrafish to date.

A strength of the zebrafish model is the ability to correlate phenotypes at the cellular and sub-cellular resolution with adult behavior and pathology. For example, a genetic zebrafish model of the progressive neurodegenerative disease hereditary spastic paraplegia showed defects in microtubule dynamics and axon stability during development as well as adult behavioral defects, demonstrating the ability to link early onset phenotypes to adult outcomes (Haynes et al., 2021). Further work to follow the consequences of genetic disruption of neurodegenerative disease risk genes from the sub-cellular level to specific circuits to whole-animal behavioral phenotypes could establish zebrafish as a revolutionary disease model.

Danionella are a genus of danionin fish species related to zebrafish, but smaller in size and with the peculiar advantage of being nearly completely transparent as adults. This remarkable characteristic is owed in part to their neoteny, or retention of juvenile traits. Danionella species range in length from 11 to 18 mm, which is roughly comparable to an older zebrafish larvae/a young juvenile (∼12 mm) (Conway et al., 2021). Many previously published papers on Danionella have stated the species used in their work is Danionella translucida. However, recently published morphological and molecular analysis showed that the widely used species in current research is not D. translucida, but a new species currently designated as Danionella cerebrum (Britz et al., 2021). Thus, we will use the designation D. cerebrum to discuss the species used in Penalva et al. (2018), Schulze et al. (2018), and Kadobianskyi et al. (2019). Danionella cerebrum possess the smallest adult vertebrate brain known, only eight times larger than a Drosophila brain (Schulze et al., 2018). Both D. cerebrum and Danionella dracula lack a bony roof on the top of the skull, making them highly amenable to brain imaging (Schulze et al., 2018; Conway et al., 2021). This allows the study of more established and mature circuits and brain regions with excellent resolution and relatively unobstructed field of view. Danionella cerebrum reach sexual maturity at 6–8 weeks, and spawn in clutches of around 12 eggs per female (Rajan et al., 2022). Danionella can be raised and maintained similarly to zebrafish, and their embryos develop at similar rates, making comparative studies between zebrafish larvae and Danionella adults straightforward, and avoiding additional investment in equipment (Penalva et al., 2018). Transgenesis in Danionella uses the same techniques applicable to zebrafish and other teleosts. Transgene insertion using Tol2 to express GCaMP6f and CRISPR/Cas9 mutagenesis of the tyr gene are both demonstrated successfully in Schulze et al. (2018). While the same tools for zebrafish transgene insertion can be used with Danionella, CRISPR gRNAs must be designed to be specific to available sequence for Danionella genes (Schulze et al., 2018; Kadobianskyi et al., 2019). Currently, no demonstration of genetic mutation of human-disease associated genes has been published.

Danionella exhibit measurable social behavior and socially reinforced learning and have an additional surprising characteristic: males of both D. translucida and D. dracula species can vocalize. These vocalizations are short multi-pulse bursts and made primarily during displays of male-directed aggression (Penalva et al., 2018; Schulze et al., 2018; Tatarsky et al., 2021). Vocalization is not known to occur in zebrafish, so study of vocalization represents new possibilities in neurobiological research with this model. The ability to pair behavior with adult imaging, as discussed previously, is a huge strength of the fish systems presented here. Adult Danionella can be mounted and imaged in agarose as long as their gills remain exposed, unlike zebrafish larvae which do not yet depend on their gills for oxygenation (Schulze et al., 2018). This opens the possibility of live-imaging of neuronal calcium indicators in vocalizing fish presented with a visual cue of an intruding male, an exciting new way to assess behavioral circuits and sex differences in the brain (as females do not vocalize).

Danionella have immense potential for visualization of cellular and sub-cellular structures at high resolution within the brain of an adult animal. Though efforts have only begun recently, mosaic expression of GFP driven by the pan-neuronal promotor elavl3 has been demonstrated to visualize sparsely labeled neurons in 5 days post-fertilization larval D. cerebrum using 2-photon microscopy. A region of the brain of a living adult D. cerebrum has been imaged from its dorsal surface to a depth of 376 μm, corresponding to the ventral surface (Penalva et al., 2018), and the brain vasculature of an adult D. dracula has been imaged to a depth of ∼800 μm using 2 and 3-photon microscopy (Akbari et al., 2021). These studies combined indicate that whole-brain live imaging is achievable in an adult vertebrate species.

The lack of a skull roof also means that these fish are likely amenable to both LSFM and confocal microscopy for shallow to mid-range depths, though neither technique has been thoroughly explored in Danionella species. It is unknown how well these animals tolerate long-term imaging, or whether they could be successfully intubated for long-term experiments. The ability to image fast events such as calcium dynamics, as well as longer-term events such as neurodegenerative processes, glia-neuron interactions, or neuronal development would increase their attractiveness as a model. Though additional work remains to establish protocols and best practices for imaging in Danionella species, the ease of imaging Danionella at both the larval and adult stages enables previously impossible or impractical longitudinal studies of development, aging, and neurological disease. This underscores their importance as an emerging vertebrate model.

A well-established animal model with largely unexplored imaging potential is the zebra finch, T. guttata. Zebra finches are a songbird species from Australia that are easy to breed and reach sexual maturity in 70–80 days, making them one of the fastest maturing bird species. The greatest potential of the zebra finch for modeling neurobiology is its song: Finches and other songbirds reflect human speech patterns in their capacity to learn, recall, and perform complex songs. The circuits directing songbird vocalization have been used to investigate the neurobiology behind practice and performance, sexual differences in vocal learning, motor control of speech, and the interplay between motor control and vocal learning.

Juvenile male finches learn their song from a male tutor, typically the father, and their vocal learning shares much in common with human vocal learning. The subsong produced by young zebra finch parallels human infant “babbling,” where both generate random syllables to dial in the correct notes that will eventually form their adult song or speech (Aronov et al., 2008). The adult song is highly stereotyped, though differences between individuals exist, making it easy to quantify changes in song pattern (Mello, 2014).

Many neurological diseases such as Huntington’s, Parkinson’s, and Fragile X syndrome result in changes in speech. Yet, the underlying mechanisms behind loss of normal speech patterns in neurological disease cannot be truly mirrored by any of the dominant research organisms outside of zebra finch. Rodents can vocalize, including audible squeaks and ultrasonic vocalizations (USVs), but are not known to imitate sounds or learn vocalizations. This can lead to unclear results when studying genes known to affect vocal learning, such as Foxp2. Global Foxp2 knockdown has been accomplished in mouse, and area-specific Foxp2 knockdown has been performed in mouse and zebra finch (Haesler et al., 2007; French and Fisher, 2014; Norton et al., 2019; Urbanus et al., 2020). In mice, results for whether USVs were affected in Foxp2 global heterozygous knockout pups was inconsistent, and region-specific knockouts in the cortex, striatum, or cerebellum did not produce a deficit in USVs (French and Fisher, 2014; Urbanus et al., 2020). In zebra finch, region-specific knockdown of Foxp2 using lentiviral shRNA in a region known to be important for song production (Area X) impaired the ability of young male finches to copy their tutor’s song (Haesler et al., 2007). This defect in song learning reflects the developmental vocal dyspraxia shown by humans with mutation of FoxP2 (Lai et al., 2001).

Currently, the relative lack of genetic tools in birds has been a hinderance for their use as models of neurological disease. Great efforts have been made to interrogate why traditional virus-mediated transfection and primordial germ cell (PGC) editing are more difficult in zebra finch compared to other birds like the chicken. Recent studies have shown that divergence in the low-density lipoprotein receptor of zebra finches is one of the factors behind inefficient lentiviral transfection, as it is the main receptor for the VSV G protein component of the lentiviral system (Velho et al., 2021). Significant differences exist between zebra finch and chicken PGCs, including different distribution and developmental gene-expression programs, which may further underly difficulties in transgenesis (Jung et al., 2019). Alternative strategies, such as electroporation and lentiviral injection into the brain have been successful at producing tissue-targeted zebra finch knockdowns and transgenics (Haesler et al., 2007; Ahmadiantehrani and London, 2017; Norton et al., 2019).

Germ line transgenesis of zebra finches have been achieved, but with low-efficiency (Agate et al., 2009; Abe et al., 2015; Liu et al., 2015). Recent reports have improved efficacy of transgenesis by direct lentiviral infection of cultured PGCs (Gessara et al., 2021), resulting in transgenic songbirds expressing eGFP in various tissues including neurons in the song circuits. CRISPR/Cas9 editing has been achieved in quail and chicken, and adenovirus-mediated CRISPR/Cas9 delivery has been used to edit zebra finch PGCs, but no adult mutant has yet been generated in zebra finch (Lee et al., 2019; Xu et al., 2020; Jung et al., 2021). Successful use of CRISPR/Cas9 in zebra finch would be a great leap forward in the tractability of this model for the study of neurodegenerative diseases.

Even with the current difficulty of transgenesis, a transgenic models for Huntington’s disease has been produced (Liu et al., 2015). Parkinson’s has also been modeled in the finch through reduction of presynaptic dopamine input to Area X of the finch brain using 6-hydroxydopamine injection (Miller et al., 2015). HD model finches had severe defects in song learning and adult song maintenance, as well as other pathology associated with human HD-like loss of striatal neurons and HTT protein aggregates (Liu et al., 2015). The dopamine-reduction PD model resulted in changes to normal vocal variability (Miller et al., 2015).

Current studies in finch mostly rely on electrophysiology and examination of the song pattern to understand how changes in neuronal function led to changes in song. Most imaging studies in finches are done in fixed tissue using confocal or widefield detection, and fluorescent markers or antibodies. A host of validated antibodies and their localization in the brain can be found at the Zebra Finch Expression Brain Atlas.² Recently, both expansion microscopy and LSFM have been used to image neurons and vasculature in the zebrafish brain, though use of LSFM in adult finch brains requires tissue clearing to reduce scattering. There remain some hurdles with tissue clearing in finch since existing clearing protocols have been primarily optimized for murine brain tissue. Rocha et al. (2019) report successful use of iDISCO+ and CUBIC clearing protocols in zebra finch adult brains and were able to visualize anatomical landmarks and the song systems of the finch via LSFM. Düring et al. (2019) have also used Expansion Microscopy to resolve individual neurons and their dendritic spines within the zebra finch brain using LSFM.

Fixed and sectioned tissues, rather than entire brains, still dominate studies of zebra finch song system projections. Though not a new technique in itself, organotypic brain slices have been recently used to image the finch brain at a higher resolution, including functional imaging using calcium sensors (Shen et al., 2017). The use of brain slices allows better access to tissues for specific labeling of structures and may prove useful for introducing other biosensors or looking at glial function. The initial use of biosensors and optogenetic tools has been successful in the zebra finch in vivo. Optogenetic control of song circuits using channel rhodopsin (hChR2-YFP) in vivo has been used to test mechanisms controlling song patterning and learning (Roberts et al., 2012). 2-photon microscopy with an installed cranial window has been used with head-fixed male zebra finches to record GECIs during female-directed song (Picardo et al., 2016), and to look at the differential response of GECI expressing high vocal center neurons and astrocytes during playback of the bird’s own song (Graber et al., 2013). Head-fixed 2-photon imaging has been successful in generating some higher-resolution data as well, notably to measure dendritic spine size in the high vocal center before and after a pupil bird listens to a tutor bird sing (Roberts et al., 2012). Further work to expand the toolkit for in vivo imaging with multi-photon and cranial window or thinned skull techniques would enhance the zebra finch’s potential as a major brain disease model. Additional ways to differentially label multiple cell types within the living zebra finch could also expand our understanding of how supporting cells like glia contribute to successful vocal learning and memory. Imaging at the level of cellular resolution can help to further refine our understanding projections within the song system, and how neurological disease processes can lead to changes in vocal learning and speech function.