Abstract
Inflammation is known to disrupt normal behavior, yet the underlying neuroimmune interactions remain elusive. Here, we investigated whether inappropriate macrophage-evoked inflammation alters CNS control of daily-life animal locomotion using a set of zebrafish mutants selected for specific macrophage dysfunction and microglia deficiency. Large-scale genetic and computational analyses revealed that NOD-like receptor nlrc3l mutants are capable of normal motility and visuomotor response, but preferentially swim less in the daytime, suggesting low motivation rather than physical impairment. Examining their brain activities and structures implicate impaired dopaminergic descending circuits, where neutrophils abnormally infiltrate. Furthermore, neutrophil depletion recovered daytime locomotion. Restoring wild-type macrophages reversed behavioral and neutrophil aberrations, while three other microglia-lacking mutants failed to phenocopy nlrc3l mutants. Overall, we reveal how peripheral inflammatory macrophages with elevated pro-inflammatory cues (including il1b, tnfa, cxcl8a) in the absence of microglia co-opt neutrophils to infiltrate the brain, thereby enabling local modulation of neural circuits affecting spontaneous locomotion.
Introduction
The impact of the immune system on brain function is seemingly self-evident when recognizing that behavioral changes coined as “sickness behaviors” are triggered by the body’s response to infection. These are a collection of behaviors evolutionarily conserved from zebrafish to humans, which include fatigue, social withdrawal, and reduced physical activity intended to conserve energy to meet the demands of fighting off pathogens for survival1–5. In response to an infection or injury, immune cells collectively instigate inflammation that serves two main purposes: to eliminate pathogens or harmful debris and restore homeostasis6, 7. To this end, innate immune cells are thought to release pro-inflammatory cytokines predominantly IL-1β and TNF-alpha in the periphery that subsequently can reach the brain parenchyma via passage through the extracellular fluids or through the peripheral nerves3, 8–11. The brain is thought to be able to recognize these pro-inflammatory cytokines as cues to implement the sickness behavior circuit2, 9. Besides infection, immune dysregulation can also lead to unwarranted inflammation causing high elevation of these cytokines systemically6, but whether this can cause sickness behavior remains unclear. Irrespective of the infection context, the underlying mechanisms by which inflammation triggers behavioral changes remain poorly understood, including the dynamic neuroimmune interactions and brain targets involved.
Macrophages are key players that orchestrate inflammation to promote tissue resolution after an infection, injury or disease 12–15, but how their dysregulation in the periphery may alter brain function and its behavioral outcome remains underexplored. A recent study indicated that peripheral macrophages along the peripheral nerve were responsive to spinal cord neurodegeneration and able to repress proinflammatory microglial activation16, implicating more intimate communication between peripheral macrophages and central nervous system (CNS) cells than previously thought. In terms of routes through which peripheral immune signals may be interpreted by the brain, microglia, brain-resident macrophages, may act as major CNS facilitators for such communications as they can respond rapidly to environmental changes17. Furthermore, intimate ties between peripheral adaptive immunity and CNS functions have been well demonstrated through active engagement of T-cells in the brain meninges at homeostasis and in disease states capable of disrupting CNS circuits that affect animal behaviors18–22. Therefore, the interplay of the peripheral immune system, both the innate and adaptive arms, and the CNS is emerging as an important factor shaping brain function and homeostasis.
Genetic basis for immune crosstalk on neural circuits and behavioral output leading to sickness phenotypes remains limited. Most experiments have relied on administration of lipopolysaccharide (LPS), pro-inflammatory cytokines and activators, or inhibitors of inflammatory cytokine production to gauge the effects of immune disturbance on animal and human behavior as associated with sickness 1, 3, 8, 23–26. Other studies have used transgenic mice to show that CNS-driven overexpression of specific pro-inflammatory cytokines was sufficient to cause structural and behavioral changes mimicking human neurological disorders or sickness27. Here, we use an unbiased large-scale genetic approach to examine whether a null mutation in the NOD-like receptor nlrc3l causing peripheral macrophage dysregulation and systemic inflammation28 can modify homeostatic zebrafish behavioral outcomes. Free and exploratory whole-body movements, coined as “spontaneous locomotion”, represent an innate behavior that does not require learning or prior experience. Because they are arguably the most fundamental and evolutionarily conserved behavioral readout29–35, we focused on this mode of behavior. Spontaneous locomotion may also reflect the motivational state and survival instincts of animals to explore, search for food, and respond to environment35–38, which are conditions that may be altered by sickness. To this end, we examined daily routine spontaneous swimming continuously over several days in larval zebrafish and assessed the impact of innate immune perturbations.
The larval zebrafish model system offers clear advantages for investigating complex interactions between the immune and nervous systems affecting brain circuits and behavioral outcomes. Due to its optical transparency, small-size, rapid development, and genetic tractability in the intact organism, larval zebrafish is amenable to in-depth in vivo analysis of molecular and cellular processes, allowing mechanisms at the single-cell level to be revealed in the context of the whole animal physiology and behavior. Furthermore, zebrafish genes and immune system share high homology with those of humans39–41. It also presents a unique opportunity to study the innate immune system independent of adaptive immune contributions, since adaptive immunity does not become functionally mature until juvenile adult stages42.
In this study, we leverage the strengths of the zebrafish to interrogate the ability for peripheral inflammatory macrophages to modulate steady state brain circuits controlling spontaneous locomotion. Zebrafish larvae naturally make short bursts of movement referred to as swim bouts throughout a 24-hour day, albeit movement is minimized in the dark period at night for these diurnal animals43, 44, providing an ideal system to monitor spontaneous locomotion. We hypothesized that modulation of neural circuits by the immune system may be one of the intrinsic factors that govern spontaneous behaviors particularly in unprovoked locomotion, which has remained unexplored. Furthermore, we focused on the analysis of a member of the cytoplasmic nucleotide binding oligomerization domain (NOD)-like receptors (NLRs), which are pattern recognition receptors that respond to microbial molecules and danger- and stress-associated molecular patterns45. While most NLRs activate pro-inflammatory signaling, some members function as negative regulators of immune response45, including nlrc3l, a particularly interesting NLR as its deletion causes spontaneous unwarranted systemic inflammation in zebrafish28. Microglia, the brain-resident macrophages, are absent7, 28, because peripheral primitive macrophages from which they are derived from46, 47 are inappropriately activated and diverted from their intended developmental course. Therefore, the study of nlrc3l mutants in addition to three other independent microglia-lacking mutants (gene deletions of interferon regulatory factor irf848, the xenotropic and polytropic retrovirus receptor xpr1b49, and myeloid transcription factor pu.1/spi1b50) enabled a more systematic analysis of the impact of microglia on steady state locomotor circuitry. Taken together, the collective results from this study indicate negligible impact of microglia loss alone, but critical influences of peripheral macrophage-evoked inflammation compounded by microglia deficiency on triggering behavioral changes in nlrc3l mutants.
Results
A genetic paradigm for assessing impact of inappropriate inflammatory macrophage activation
Previous forward genetic screens have identified loss-of-function mutations in a novel NOD-like receptor, nlrc3l, that cause microglia loss due to abnormal macrophage activation in zebrafish28, 51. To address possible neuroimmune interactions in motivating locomotor behaviors, we therefore sought to use nlrc3lst73 null mutants as a genetic model to interrogate whether disruptions in macrophages and microglia can impact the locomotor circuitry.
To characterize the immune dysfunction in nlrc3l mutants more comprehensively, we conducted an RNA-seq based transcriptome analysis to define differentially expressed genes in the nlrc3l mutants compared with their co-housed wild-type and heterozygous siblings at steady state (Fig. 1a). Gene ontology analysis of the 245 genes significantly upregulated in the nlrc3l mutants revealed highly significant enrichment of pathways associated with innate immune activation and inflammation, including cytokines (il1b, cxcl8a), chemokines (cxcl8a, cxcl18b, cxcl19), enzymes (mmp9, mmp13a, timp2b, irg1, irg1l, adam8a, irak3, ncf1), receptors (il6r, gpr84, nlrp16), and transcription factors (irf8, irf1b)52 known to be elevated during pro-inflammatory conditions (Fig. 1a). tnfa was also found to be elevated in nlrc3l mutants but not above the threshold of a 1.5 fold increase as previously shown28. Conversely, 113 genes were significantly downregulated in nlrc3l mutants that largely affect metabolic processes, or notably, are known to be microglia gene markers (mrc1b, lgals3bpb, havcr1)53 consistent with microglial cells being absent in nlrc3l mutants (Fig. 1a).
Of particular interest, acod1/irg1, aconitate decarboxylase 1 or immune-responsive gene 1, was one of the most highly upregulated genes by RNA-seq analysis in nlrc3l mutants. This gene encodes for a conserved mitochondrial enzyme that catalyzes the formation of itaconate, known to be highly elevated in inflammatory macrophages 54–56. Similar to mouse, irg1 is specifically expressed in activated macrophages after bacterial infection, LPS activation, and cancer cell exposure in zebrafish, but not normally in unchallenged wild-type 54, 57, 58. By RNA in situ hybridization, we found a strong transcriptional induction of irg1 in macrophages of all nlrc3l mutants analyzed, in stark contrast to siblings devoid of any irg1 expression (Fig. 1b). The spontaneous induction of irg1 in nlrc3l mutants in the absence of infection, injury, or any overt challenge shows that macrophages are indeed inappropriately activated.
Furthermore, dysregulated inflammatory macrophages were implicated to be responsible for the systemic inflammation observed in nlrc3l mutants28, but direct evidence was not demonstrated. To examine this connection, we created a stable transgenic line to constitutively restore wild-type nlrc3l expression in a macrophage-specific manner using regulatory sequences for a macrophage gene mpeg1 in nlrc3l mutants (subsequently referred to as macrophage rescue)(Fig. 1c-d). This macrophage rescue cassette included a cardiac myosin cmlc2 gene promoter controlled GFP reporter (cmlc2:GFP) to enable selection of the transgenic embryos based on heart GFP (GH+) expression. nlrc3lst73 heterozygous intercrosses were used throughout the study to breed homozygous nlrc3l mutants and their siblings. Using this conditional macrophage rescue strategy, we show that indeed restoring wild-type macrophages was sufficient to eliminate the significant elevation of pro-inflammatory genes (il1b, mpx, irg1, gpr84), including the neutrophil gene marker mpx, in nlrc3l mutants, thereby reversing systemic inflammation (Fig. 1c). This was associated with a full recovery of microglia in all nlrc3l mutants carrying the macrophage rescue transgene (GH+) (Fig. 1d). By contrast, nlrc3l mutant siblings without the rescue transgene showed the expected inflammation and loss of microglia phenotypes (Fig. 1c-d).
Taken together, the results from RNA-seq, acod1/irg1 in situ expression, and conditional macrophage rescue collectively provide direct evidence that the mutant macrophages are inappropriately activated and responsible for systemic inflammation and microglial absence in nlrc3l mutants as previously implicated28. These mutants thereby provide an effective genetic paradigm for assessing the impact of inappropriate macrophage-evoked inflammation on other body systems, particularly the brain.
NOD-like receptor nlrc3l mutants capable of normal motility and visuomotor response appear less motivated to move
In the absence of any overt stimuli, locomotion of freely behaving animals represents a major form of spontaneous behavior, possibly reflecting some degree of intrinsic motivation for exploration and physical activity59–61. To examine whether inappropriately activated macrophages could disrupt spontaneous behaviors in larval zebrafish, we analyzed nlrc3l mutants using a large-scale automated tracking system equipped with a high-speed infrared CCD camera to simultaneously monitor freely swimming individually-housed zebrafish larvae at 5 days post-fertilization (dpf) under normal day-night cycles in 96-well (Fig. 2a and Supplementary Movie 1) and 24-well platforms (about 5× larger volume than the 96-well) (Fig. 2a and Supplementary Movie 2). The larger housings enabled analysis of not only motility but also spontaneous swimming patterns (Fig. 2a) that may be exploratory in nature. Animals were tracked simultaneously in their chambers for three continuous days over the normal 14-hour light and 10-hour dark cycle with all genotypes blinded (Fig. 2a).
Using this long-term tracking, we found a striking reduction in spontaneous daytime swimming in nlrc3l mutants in both 96-well and 24-well housings. These mutants exhibited the expected high activity during the day and periods of resting at night of diurnal animals (Fig. 2b). In the 96-well format, we found significantly large reductions of 35-50% in median daytime distance travelled and 30-40% in median frequency of swimming activity in nlrc3l mutants, while no difference in average swimming speed compared with heterozygous and wild-type siblings (Fig. 2b-c). When comparing large numbers between heterozygous and wild-type control animals (each n>50), small significant locomotor differences could be detected for distance travelled (Fig. 2c), but not apparent in smaller sample sizes (Fig. 2f), suggesting the difference, if any, may be trivial. Large arena 24-well analysis also showed reduced spontaneous daytime swimming in nlrc3l mutants (Fig. 2d-h). Both housing platforms indicate that a lower number of swimming bouts (daytime activity) accounted for the reduced locomotion as the average speed of swimming was unchanged.
While nlrc3l mutants are not visibly different from wild-type either by gross physical or behavioral traits (no obvious defect in skeletal muscles, response to touch, or general motility) or by traces of movement at any given timepoint, we sought to analyze the quality of their swimming in more detail. Swimming traces dynamically varied within and between individuals (Fig. 2d and Supplementary Movie 3), so analyzing a large number of animals over a long time period was imperative to acquire reliable trends. We developed a quantitative method to characterize swimming patterns in the large 24-well arenas based on image segmentation of swimming traces from 14-hour light periods on days 2 and 3 (d2 and d3) using a 3 x 3 grid system (Fig. 2i). A heatmap of the perimeter and center scores of individuals tracked over time conveyed the inter- and intra-individual variation in their swimming pattern (Fig. 2j-k and Supplementary Fig. 1). In general, locomotor activity was high at the beginning of the day and declined toward the end of the day (Fig. 2j-k). By unsupervised hierarchical clustering of perimeter (Fig. 2j) and center (Fig. 2k) scores, we found nlrc3l mutants to be sorted and grouped together with control siblings, thereby indicating similar swimming patterns for all genotypes. Based on the fraction of the daytime spent on full or no perimeter laps, and high or near zero center locomotion, nlrc3l mutants were similar to control siblings except for a significant 30% reduction in swimming around the full perimeter (Supplementary Fig. 1). We found anti-correlated trends between frequency of full perimeter swimming and low center activity in all genotypes, indicating that more perimeter swimming coincided with more center swimming (Supplementary Fig. 1). Taken together, behavioral tracking in small and large housings and the swimming pattern analysis indicate that nlrc3l mutants are capable of normal swimming, but conduct less perimeter swimming than their wild-type counterparts.
To determine whether impairment in visuomotor coordination may contribute to the nlrc3l mutant behavior, we tested these fish on a rapid light-to-dark switch assay known to activate a stereotypical hyper-swimming response62–64 (Supplementary Fig. 2). We found that the nlrc3l mutants responded to a light-to-dark switch with an equivalent rapid increase in swimming as their wild-type counterparts (Supplementary Fig. 2), indicating normal motility and no visuomotor impairment. Interestingly, when we subjected the same fish cohort to long-term 72-hour tracking after the visuomotor test, a reduction in daytime locomotion resumed in nlrc3l mutants (Supplementary Fig. 2). Collectively, our data demonstrate that nlrc3l mutants are capable of normal motility and visuomotor coordination, but appear less motivated to swim when unprovoked. These results implicate a disruption of the upstream supraspinal circuits in nlrc3l mutants underlying the brain control of locomotion rather than the effector spinal circuits that execute locomotion.
Aberrant inflammatory macrophages causing systemic inflammation lead to reduced daytime locomotion
In light of the known macrophage defects in nlrc3l mutants, we asked whether the inappropriately activated peripheral macrophages were responsible for the reduced locomotor activities. We turned to using the stable macrophage rescue line expressing mpeg1-nlrc3l as described above (Fig. 1c) to determine whether restoring wild-type macrophages in nlrc3l mutants was sufficient to re-establish normal levels of spontaneous swimming. A large-scale 96-well platform was used to test 48 individuals with the macrophage rescue transgene (GH+) and the remaining 48 wells were used for siblings without the transgene (non-rescued, no GH) as negative controls (Fig. 3a). All fish were derived from a heterozygous intercross for the nlrc3lst73 mutation with one parent carrying the macrophage rescue transgene. We analyzed the locomotor behaviors in the normal light-dark cycles over a continuous 72-hour period and found the expected reduction in daytime swimming in control nlrc3l mutants without the rescue construct (no GH)(Fig. 3b-c). By contrast, the macrophage-rescued nlrc3l mutants showed a significant recovery of daytime locomotion akin to wild-type and heterozygous animals (Fig. 3b-c). We did not detect any difference in locomotor behaviors in wild-type or heterozygous individuals due to expression of the macrophage-rescue mpeg1:nlrc3l transgene (Fig. 3b-c). Furthermore, we found that the macrophage-rescued nlrc3l mutants also significantly recovered exploratory behaviors in the large-arena 24-well tracking (Supplementary Fig. 3). In the large-arena environment, the rescue transgenic expression also did not alter swimming levels of control wild-type and heterozygous individuals (average distance (cm) travelled per 10 min ± standard deviation was 42.0 ± 11.5 for GH+ WT and 46.3±18.8 cm/10 min for GH+ Het which overlapped in range as non-transgenic WT at 35.6 ± 9.9 and non-transgenic Het at 36.4 ±12.3) (Supplementary Fig. 3 and Fig. 2f). By contrast, macrophage-rescued GH+ mutants restored wild-type levels of swimming in the large arena at 36.7 ± 8.9 cm per 10 min compared with their significantly reduced baseline swimming at 15.5 ± 9.6 cm per 10 min in the non-transgenic mutants (Supplementary Fig. 3 and Fig. 2f). Taken together, these results show that restoring wild-type macrophages in nlrc3l mutants was not only sufficient to recover microglia and reverse inflammation (Fig. 1c-d), but also significantly improves levels of daytime swimming in nlrc3l mutants (Fig. 3 and Supplementary Fig. 3). The data indicates that the inappropriately activated peripheral macrophages were responsible for the reduced locomotor behaviors in nlrc3l mutants, underscoring a possible innate immune modulation of the brain (or supraspinal) circuits that regulate spontaneous locomotion.
Absence of microglia cannot solely account for the reduced daytime locomotion
Because microglia are implicated in synaptic pruning and modulation of neuronal circuit function17, 65, 66, we investigated whether microglia loss in nlrc3l mutants may be another major cause of their locomotor deficit, in addition to the inappropriate macrophage activation. We found microglia widely distributed throughout the brain in wild-type zebrafish at 5 dpf at the time of our behavioral analyses, including in the deeper regions of the telencephalon and diencephalon that are known to control locomotor circuits67–69, besides their well-known tectal localization70 (Fig. 4a and Supplementary Movie 4). In addition to constant interactions with tectal neurons, microglia intimately intermingled with neurons in the ventral regions of the telencephalon and diencephalon (Fig. 4b and Supplementary Movie 5). The broad distribution of microglia is consistent with the possibility that these cells can impact development or function of various brain circuits.
Since uncoupling microglia loss from inflammatory processes in nlrc3l mutants is unattainable because both are consequences of the early inappropriately activated macrophages, we investigated whether an absence of microglia alone is sufficient to phenocopy nlrc3l mutant locomotor behavior. To this end, we extended our behavioral analysis to three other microglia-lacking mutants (irf8−/−, xpr1b−/−, and pu.1/spi1b−/−), which do not have the inflammation phenotypes48–50, 71(Fig. 4c). All three mutants lack microglia due to developmental disruptions distinct from nlrc3l mutants (Fig. 4c-d); both irf8 and pu.1/spi1b are conserved transcription factors required for myeloid and macrophage cell development48 whereas xpr1b is a phosphate exporter required for migration and differentiation of microglial precursors49. In common, all four microglia-lacking mutants including nlrc3lst73 do not exhibit any gross morphological defects with normally sized swim bladders, and are anatomically indistinguishable from wild-type fish (Fig. 4d). Time plot analysis and calculations of daytime locomotion surprisingly indicate that irf8 (Fig. 4e), xpr1b (Fig. 4f), and pu.1 (Fig. 4g) mutants were largely normal relative to control siblings, showing no significant difference in locomotion based on distance of movements, although a rather modest decrease in daytime activity count (average median reduction of 14%) for irf8 and xpr1b mutants (Fig. 4e-f). This is a relatively small change in comparison to the large average median reduction of 36.5% in nlrc3l mutant daytime activity (Fig. 2c), indicating that an absence of microglia alone could not explain the locomotor deficit in nlrc3l mutants.
Pro-inflammatory conditions are causal to disrupted daytime locomotion
Given the implications of pro-inflammatory factors on causing reduced locomotion during sickness3, 5 and inflammation as possibly a main driver of the locomotor deficiency in nlrc3l mutants, we tested whether reversing systemic inflammation by administering anti-inflammatory drugs could significantly restore normal behaviors in nlrc3l mutants. We examined three small-molecule drugs (17-DMAG, Bay 11-7082, and dexamethasone) that are water-soluble and compatible for administering in the fish water (Supplementary Fig. 4). They have been shown to curb inflammation through different mechanisms by attenuating NF-kB mediated transcription or activating glucocorticoid functions in zebrafish and other systems. 17-DMAG is a water-soluble geldanamycin analog that can inhibit the heat-shock protein Hsp90 and cause degradation of target proteins, including the NF-kB protein complex72, 73; Bay 11-7082 can inhibit E2 ubiquitin (Ub) conjugating enzymes and prevent degradation of IkB-alpha, an NF-kB inhibitor74; and dexamethasone is an agonist of the glucocorticoid receptor (GR) that activates a negative feedback mechanism to reduce inflammation75. We used the 96-well platform to enable a large sample size to test the different drugs at concentrations previously characterized to reduce inflammatory processes71. Each drug administered was tested side-by-side on the same plate with a control (no treatment) condition on fish derived from the same clutch of a nlrc3lst73 heterozygous intercross (Supplementary Fig. 4). We found that treatment with dexamethasone, but not 17-DMAG or Bay 11-7082, was effective in significantly restoring daytime locomotion in nlrc3l mutants to levels similar to untreated control siblings (Supplementary Fig. 4). The difference in the effect of the three drugs suggest that the mutants may be deficient in glucocorticoid regulation (which dexamethasone activates) rather than excessive NF-kB activation (which 17-DMAG and Bay-11-7082 inhibit). Interestingly, because dexamethasone appears to increase daytime spontaneous locomotion in control siblings consistent with another previously published study76, dexamethasone may have a stimulatory effect in addition to suppressing inflammation. It is, however, important to point out that dexamethasone increased locomotion in mutants more so than in controls as shown by the overlapping range of locomotor activity after dexamethasone treatment, instead of a proportional increase. Furthermore, additional means of suppressing inflammation by restoring macrophage nlrc3l function also significantly restored locomotor activity in nlrc3l mutants. Taken together, these results indicate that administering dexamethasone, a synthetic glucocorticoid, without restoring microglia or reversing macrophage inflammation was effective to reverse a significant level of the nlrc3l mutant locomotor deficit, lending support for pro-inflammatory signaling being a predominant factor in disrupting the corresponding locomotor circuits.
Reduced locomotion is associated with dorsal thalamic calcium imbalance
To define brain regions that may be disrupted in nlrc3l mutants, we used genetically encoded calcium indicator GCaMP6 fast variant (GCaMP6f) to assess four brain regions known in vivo to either contain circuits controlling locomotion in vertebrates (cerebellum, diencephalon (dorsal thalamus), and dorsal telencephalon (pallium))34, 77, 78 or a high density of microglia (tectum)70, 79 (Fig. 5a). Due to the implicated link between microglia and neuronal development17, 65, 66, we hypothesized that brain regions in or near the proliferative zone or ventricular surface may be more susceptible to immune interference or modulation. We therefore focused our analysis on a subset of newly differentiated neurons using the basic helix-loop-helix (bHLH) transcription factor neurod regulatory sequence80–82 to drive GCaMP6f in these neurons for in vivo calcium imaging. We assessed spontaneous neural activity in these regions to determine possible differences underlying the reduced locomotion in nlrc3l mutants compared with their control siblings using resonant scanning microscopy at 6 dpf under normal daytime conditions. Interestingly, we found no significant difference in number of calcium transients in the telencephalon, diencephalon, or cerebellum, but a significant increase in tectal neuronal activity in nlrc3l mutants compared with control siblings (Fig. 5b-f). To determine whether this increase was due to impaired macrophages in the nlrc3l mutants, we asked whether restoring normal macrophages in the nlrc3l mutants could reverse this tectal activity increase (Fig. 5b-f). Using the stable macrophage rescue line as described in Fig. 1c, we found indeed this was the case, while neuronal activity in the other brain regions had no apparent changes (Fig. 5b-f).
Since the macrophage rescue strategy cannot distinguish recovery of microglia from reversal of systemic inflammation, because both are direct consequences of the mutant macrophages, we next asked whether a sole microglia loss using the microglia-lacking irf8 mutant48 could account for the increased tectal neuronal activity. To address this, we conducted calcium imaging of tectal neurons in the irf8 mutants, and found an increase in tectal calcium transients compared with sibling controls at both the animal level (p=0.03) and individual neuron level (p=0.06) (Supplementary Fig. 5). These results suggest that microglia loss alone could be responsible for the increase in tectal activity. Consistent with these results, microglia have been shown in vivo to negatively regulate neuronal activity in the zebrafish tectum83 and mouse dorsal striatum84. Since a strongly reduced locomotor behavior was not found in irf8 mutants, we reason that the common increase in tectal activity in nlrc3l mutants and irf8 mutants was likely not responsible for the nlrc3l mutant behavior, but merely reflecting a consequence of their mutual microglia loss.
While the frequency of calcium spikes was not changed, we found a robust decrease in baseline intracellular GCaMP6f fluorescence level in the diencephalic ventricular neurons corresponding to the dorsal thalamus that stood out in the nlrc3l mutant brains compared with the siblings (Fig. 5g-h). We quantified the relative change in total cell fluorescence for a random selection of individual neurons in either the telencephalon (as a control) or the diencephalon over a 10-minute period of in vivo calcium imaging (Fig. 5g-h). Our analysis shows that nlrc3l mutants had on average ∼4-fold decrease in diencephalic ventricular calcium level compared with their sibling controls, while there was no difference in calcium levels in the telencephalic neurons (Fig. 5g-h). To determine whether this decrease was linked to the mutant macrophages in nlrc3l mutants, we used the macrophage-rescue construct and found that the calcium level decrease was reversed after recovery of normal macrophages (Fig. 5g-h), attributing the neuronal calcium phenotype to the mutant macrophages. We then also analyzed whether another microglia-lacking mutant (irf8−/−) had reduced diencephalic neuronal calcium but did not find a significant decrease, although a marked reduction can be quantified (Supplementary Fig. 6). Taken together, the significantly large decrease in dorsal thalamic intracellular calcium was unique to the nlrc3l mutants, raising the possibility that this neuronal disruption was linked to their altered behavior. This significant variation in baseline intracellular calcium level may reflect changes to intra-neuronal calcium signaling and homeostasis, thereby altering excitability of the neurons85.
Disruption of diencephalic dopaminergic TH levels is consistent with reduced locomotion
The severely reduced baseline calcium level of diencephalic neurons in nlrc3l mutants points to a possible disruption of the diencephalic circuit as a cause for the behavioral deficit. In support of this, dopaminergic neurons in the diencephalon are known to regulate brain and spinal cord motor circuits in zebrafish and mammals86, 87. To assess whether dopaminergic neurotransmission in the diencephalon may be altered in nlrc3l mutants, we analyzed possible changes to expression of dopaminergic system genes, namely the two tyrosine hydroxylase TH genes (th and th2) encoding the rate-limiting enzyme for dopamine synthesis and the dopamine transporter dat/slc6a3 essential for reuptake of dopamine to terminate neurotransmission88. Interestingly, using qPCR, we found that nlrc3l mutants had significantly reduced levels of th but not th2 or dat (Fig. 6a), raising the possibility that these mutants may have a defect in dopamine synthesis or reduction in number of dopaminergic neurons. Since expression of th2 is restricted to a small subset of dopaminergic neurons whereas dopaminergic neurons broadly express th89, the specific reduction in th corroborates the possible alteration in the dopaminergic diencephalic clusters that are known to regulate locomotion69, 90 in nlrc3l mutants. Furthermore, the dat gene expression pattern largely overlaps with th but was not reduced in nlrc3l mutants suggesting that the number of dopaminergic neurons or recycling of dopamine may not be altered.
To further distinguish whether th downregulation signifies a decrease in number of dopaminergic neurons or dopamine synthesis, we turned to investigate the TH protein level using a commercially available antibody specific to TH that does not bind TH289, 91. Using confocal brain imaging and immunofluorescence for the TH protein, we quantified the relative TH protein level and TH-expressing neuronal cluster size in three prominent brain regions (Fig. 6b-f). Co-localization analysis of TH protein expression with the dorsal thalamic neurons (as labeled by neurod:GFP+) showed that the neurod+ neurons imaged for calcium dynamics are not the same as the TH+ neurons, but may directly interact (Supplementary Fig. 7). To assess activity of the TH+ neurons in the diencephalon, we turned to using a pan-neuronal calcium indicator elavl3:GCaMP6s for in vivo imaging at two focal planes, through the dorsal thalamus and the presumptive DDC (Supplementary Fig. 8 and Supplementary Movies 6-7). Consistent with neurod+ calcium imaging results, the basal intracellular calcium levels in both areas of the diencephalon (dorsal thalamus and the DDC) were strikingly reduced in nlrc3l mutants compared with control siblings, while no difference was found in the telencephalon nor in number of calcium transients with the exception of a possible small increase in dorsal thalamic calcium activity (Supplementary Fig. 8). We then focused on two regions that consist of dopaminergic neurons for TH expression analysis: pretectum and dopaminergic diencephalic clusters (DDC), with the latter containing dopaminergic diencephalospinal neurons that directly project to the spinal motor circuits69, and the locus coeruleus (LC) which contains TH+ noradrenergic neurons as a comparison group. In nlrc3l mutants, the DDC, which has descending projections to the motor circuits, showed a significant reduction in TH protein level alongside a modest and nearly significant TH reduction in the LC (Fig. 6e-f), but there was no change in the other brain regions analyzed. The sizes of the neuronal clusters in all regions of analysis were also not altered, suggesting that the defect in nlrc3l mutants was likely in neural function rather than structure. These results in combination with aberrant calcium levels in the diencephalic region suggest that nlrc3l mutants may have disrupted diencephalic functions that cause reduced spontaneous locomotion.
Since the RNA and protein levels of TH were reduced but the sizes of the neuronal clusters did not significantly change, dopamine synthesis but not neural structure may be affected in nlrc3l mutants. Whole mount brain RNA in situ analysis further affirms a grossly normal brain structure in nlrc3l mutants by probing a diverse set of neuronal genes: nuclear receptor nr4a2b/nurr1 marking a subset of dopaminergic neurons92, serotonin transporter serta/slc6a4a showing serotoninergic neurons93, dopamine beta-hydroxylase dbh labeling noradrenergic neurons94, vesicular-glutamate transporter vglut2a/slc17a6b showing excitatory glutamatergic neurons95, and glutamate decarboxylase gad1b marking inhibitory GABA neurons95 (Supplementary Fig. 9). While the broad expression of vglut2a and gad1b indicate gross overall similarity between nlrc3l mutants and control siblings, we cannot exclude the possibility that detailed differences in the connection or pattern of the excitatory and inhibitory neurons exist. Taken together, given the locomotor deficit in nlrc3l mutants being linked to systemic inflammation compounded by an absence of microglia, these results implicate a role for the innate immune system in modulating the diencephalic dopaminergic system that regulates locomotor behaviors.
Abnormal infiltration of neutrophils in the diencephalic brain disrupts locomotor control
How mutant macrophages may modify CNS circuits that regulate locomotion remains an open question. Because brain infiltration and increased circulation of neutrophils were attributed to the systemic inflammation in the developing nlrc3l mutant embryo28, we sought to determine whether neutrophils are abnormally trafficked to the mutant larval brain. Consistent with this, neutrophils were found to be both residing in and moving through the diencephalon and other brain regions at 5-6 dpf, notably through the brain ventral vasculature surrounding the diencephalic choroid plexus (cp) and in close contact with diencephalic neurons (Fig. 7a-d and Supplementary Movies 8-9). By contrast, as is known under normal conditions96, slow moving and infiltrating neutrophils in the brain are not found in heterozygous or wild-type siblings, only rare occasions of individual neutrophils that rapidly circulate through the brain28 (Fig. 7a-d, Supplementary Movie 10). Compared with control siblings, nlrc3l mutants had a significant and unusual presence of brain lingering neutrophils and an increased number of peripheral neutrophils (Fig. 7c) at the time of our locomotor tracking at 5-6 dpf. Furthermore, the interaction of mutant neutrophils with the CNS begins early during development by 2 dpf (Fig. 7d and Supplementary Movies 11).
The significant presence of neutrophils in the larval brain raises the possibility that inflammatory macrophages may indirectly modulate CNS circuits through their interaction with neutrophils in nlrc3l mutants. To examine the impact of macrophages on neutrophil activity in the larval nlrc3l mutants, we determined whether infiltration of neutrophils into the brain was dependent on nlrc3l function in macrophages, or conversely in neutrophils or both cell types. We generated stable transgenic lines to express wild-type nlrc3l in a macrophage-specific (Fig. 7a-d) or neutrophil-specific (Supplementary Fig. 10) manner in nlrc3l mutants and analyzed lingering, circulating, and total peripheral neutrophil numbers in or around the brain using in vivo time-lapse confocal imaging (Supplementary Movies 9-10 and 12-13). We found that a neutrophil-specific rescue of nlrc3l function in nlrc3l mutants did not reverse either the brain lingering neutrophil or reduced locomotor phenotype (Supplementary Fig. 10), whereas restoring wild-type nlrc3l expression in macrophages led to rescue of these phenotypes (Figs. 3 and 7, Supplementary Fig. 3, Supplementary Movies 9-10 and 12-13). The data together indicate that nlrc3l has non-cell-autonomous functions in affecting neutrophils and their impact on locomotor behavior through its action in macrophages.
We next tested the possibility that activated neutrophils in nlrc3l mutants may directly interfere with the locomotor circuit by using well established morpholino combination-based gene knockdowns against the granulocyte colony-stimulating factor receptor G-CSFR (gene name of csf3r) and myeloid transcription factor PU.1 (gene name of spi1b) that effectively removes myeloid cells97, 98, or against csf3r alone, which reduces neutrophil numbers but not macrophages as pu.1/spi1b knockdown does71, and assessed subsequent locomotor behaviors (Fig. 7e). We confirmed the high efficacy of the morpholino approaches to deplete neutrophils (90% average reduction)(Supplementary Fig. 11). Using the large-scale 96-well locomotor assay over the normal 14-hour day and 10-hour night cycles, we compared routine daytime swimming in nlrc3l mutants with their wild-type and heterozygous siblings after neutrophil ablation by treatment of csf3r/spi1b morpholinos (MOs) or single csf3r MO, and in control conditions (uninjected and control standard morpholino injected) on the same plate for each experiment (Fig. 7e). We found that in contrast to baseline or control morpholino-injected nlrc3l mutants, nlrc3l mutants after neutrophil depletion recovered significant swimming levels that are closer or similar to that of control siblings (Fig. 7e), indicating an essential role for neutrophils in causing the locomotor deficit. Furthermore, the effect of the morpholino treatments was specific to nlrc3l mutants, as neither wild-type nor heterozygous individuals had significantly altered locomotor behaviors (Fig. 7e). These results further substantiate the possibility that instead of direct signaling to neurons in the brain locomotor center, peripheral inflammatory macrophages activate neutrophils to infiltrate the brain to enable neutrophils to locally interact with and alter the locomotor circuits in nlrc3l mutants.
Discussion
Immune modulation of brain circuitry controlling innate spontaneous locomotion
While the wiring of the nervous system for implementing sensation and perception such as in vision, audition, olfaction, gustation, and touch has received much attention99, 100, the neural basis of unprovoked spontaneous behaviors remains underexplored. Spontaneous or exploratory locomotion of animals at homeostasis may reflect some degree of motivation in response to internal or external cues for survival; these animals may be moving to explore, forage for food, or find a mate35–37, 59, 60. Since changes to the immune system have been implicated in sickness and abnormal social behaviors characterized by reduced physical activity and motivation1, 5, we sought to determine whether dysregulated macrophages may alter brain circuits controlling normal locomotor activity. We leveraged investigation of the nlrc3l loss-of-function mutants as it provided an effective genetic paradigm for assessing the impact of inappropriate macrophage inflammation on brain circuit control of locomotion.
Using a large-scale and blinded platform to perform long-term tracking of spontaneous movements, we show that nlrc3l mutant larval zebrafish, which have aberrant inflammatory macrophages and no microglia, exhibit reduced frequency of exploratory swimming, yet maintain normal patterns of swimming (Fig. 2). When they are subjected to a series of light-to-dark visual stimuli, the mutants are fully capable of swimming at the same frequency and pattern as their wild type and heterozygous siblings, indicating no overt physical disability but seemingly decreased motivation for locomotor activity (Supplementary Fig. 2). In fact, after a series of normal locomotor responses to the visual stimuli, these mutants when placed back into their normal day-night cycles revert to displaying reduced daytime swimming (Supplementary Fig. 2). These results implicate that inappropriately activated peripheral macrophages during development or over a prolonged period are capable of perturbing brain locomotor circuits at the high level where the decision to move is made, rather than the downstream execution of the body movements as conveyed in our working model (Fig. 8). The precise aspects of the neural circuits controlling spontaneous swimming that may be modulated by pro-inflammatory innate immune functions remain to be determined by dissection and manipulation of the relevant circuitry.
Inflammatory macrophages as a major cause of reduced locomotor drive in nlrc3l mutants
We show that the reduced locomotor drive phenotype in nlrc3l mutants has a myeloid origin. Using stable transgenics to restore wild-type expression of nlrc3l in a macrophage-specific manner in nlrc3l mutants, the locomotor deficit was significantly reversed, pointing to mutant macrophages as a major source of the behavioral phenotype. Since these mutants lack microglia and only have macrophages in the periphery, the mechanism by which the mutant macrophages could interfere with brain circuit function is expected to be indirect or mediated by soluble factors. Consistent with this, reversing infiltration of neutrophils in the brain by neutrophil ablation, restoring macrophage nlrc3l function, or inhibiting inflammatory signaling by dexamethasone 101, 102, as distinct means of suppressing inflammation were sufficient to significantly restore normal locomotor activity in nlrc3l mutants (Fig. 7 and Supplementary Figs. 4 and 10). Taken together, we show that inflammatory macrophages may indirectly, through yet unknown cues that may involve cytokines and neuromodulating molecules or recruitment of other leukocytes, alter the brain neural circuits regulating spontaneous behaviors at homeostasis (Fig. 8).
Trafficking of neutrophils as an effective means for peripheral immune system to modulate CNS circuit functions
The view that the central nervous system (CNS) is immune privileged and immunologically inert and separated from the peripheral immune system has been revised in recent decades18, 21. Under normal conditions, neutrophils are rarely found in the CNS as they are restricted from trafficking into the brain parenchyma and ventricular regions due to the blood-brain barrier96, 103. However, upon a brain perturbation or pathology, such as trauma, infection, ischemia, neurodegeneration or autoimmunity, neutrophils are recruited into the brain and found to be lingering in the CNS environment96, 103. Interestingly, while we found the expected absence of infiltrating neutrophils in wild-type and heterozygous larval zebrafish brain, the presence of infiltrating and lingering neutrophils in the brain of nlrc3l mutants was striking (Fig. 7 and Supplementary Movies 9-10). Beginning at embryonic development and continuing through mature larval stages, nlrc3l mutants exhibit the unusual activities of neutrophils passing through and lingering in the brain, as shown previously28 and in this study (Fig. 7 and Supplementary Movie 9), a phenomenon associated with inflammatory brain conditions96, 103. The neutrophil infiltration into the larval brain is secondary to the inappropriate activation of macrophages in nlrc3l mutants as it can be reversed by restoring wild-type macrophages but not by wild-type neutrophils in nlrc3l mutants (Fig. 7 and Supplementary Fig. 10), indicating a requirement for the mutant macrophages. These results indicate that the effects of the nlrc3l mutation on neutrophils were non-cell-autonomous (Supplementary Fig. 10). While the precise nature of the neutrophil interaction with the CNS neurons in nlrc3l mutants is yet unknown, it is possible that known neutrophil production of effector molecules (such as free oxygen radicals and extracellular traps composed of proteins and DNA called NETs)96, 103 may impact the surrounding neurons to affect locomotor control in the diencephalon. We further show that depletion of neutrophils by a combined morpholino-mediated knockdown of csf3r and spi1b, or single csf3r knockdown was sufficient to partially restore normal locomotion in nlrc3l mutants, providing evidence that modulation of the locomotor circuitry in the ventral larval zebrafish brain was in part mediated by activated neutrophils in the nlrc3l mutants (Fig. 7e). Partial recovery of the behavioral deficit may be due to incomplete neutrophil ablation, or other neutrophil-independent factors that disrupt normal behavioral circuit function stemming from other mutant cell types, in particular but not exclusively macrophages. Collectively, these results indicate that the inflammatory macrophages trigger neutrophil infiltration into the brain, allowing activated neutrophils to directly interact with the brain circuitry that controls innate spontaneous locomotion (Fig. 8).
Microglia loss has negligible impact on spontaneous locomotion
Besides macrophages in the periphery, we also examined the role of microglia, which are brain-resident macrophages that have been widely anticipated and implicated to affect CNS circuitry and functions due to their role in synaptic pruning and surveillance during development17, 65, 66, 104. Microglia are absent in nlrc3l mutants, because they are derived from primitive macrophages, which are inappropriately activated and thereby diverted from their normal developmental course 28. In order to delineate the causal factor(s) leading to the behavioral deficit in nlrc3l mutants, we investigated three other microglia-lacking mutants, gene knockouts of irf8, xpr1b, and pu.1/spi1b, to determine whether a sole microglia loss without the inflammatory phenotype could result in reduced daytime locomotion. Our results revealed minimal to no effect of microglia loss on locomotor behavioral change in irf8, xpr1b, and pu.1/spi1b mutants (Fig. 4c-g).
Our finding that an absence of microglia in larval zebrafish at 5-8 dpf did not cause a significant disruption to routine swimming behaviors seems consistent with murine studies indicating no significant behavioral or cognitive impairment with removal of microglia at baseline105. Understanding the full extent of microglia impact on behavioral outcomes remains incomplete, while murine studies raise the possibility that microglia may be primary responders to homeostatic perturbations in the CNS, rather than main drivers of CNS maintenance and homeostasis21, 105–108. Furthermore, in contexts of CNS pathology or inflammation, a role for microglia in modifying CNS circuits and behavior remains supported and possible21, 104. Consistent with this, our study underscores the possibility that systemic inflammation stemming from dysregulated peripheral macrophages intensifies the effects of microglia loss, thereby causing reduced locomotion and sickness-like behavior. Microglia may normally have a role in dampening inflammatory effects in the CNS, thereby limiting potentially negative effects on circuit functions and behavioral outcomes. Similarly, nlrc3l mutants may be more severely impacted by the chronic systemic inflammation due to an absence of microglia which normally limit pro-inflammatory signaling in the CNS.
Dopaminergic control of spontaneous locomotion may be susceptible to immune modulation
We found that the reduced locomotion in nlrc3l mutants was associated with a disruption of the diencephalic dopaminergic circuit previously known to control swimming34, 69, 86, 87 (Fig. 8). nlrc3l mutants at larval stages were found to have reduced intracellular calcium in the dorsal thalamus and DDC region at baseline (Fig. 5 and Supplementary Fig. 8) and a significant reduction in the rate-limiting enzyme TH for dopamine synthesis in the diencephalic dopaminergic clusters (DDC) (Fig. 6). These results are consistent with the known importance of dopamine neurotransmission and brain dopaminergic circuits for initiation and control of locomotion that is evolutionarily conserved from fish to humans109–112. DDC encompass neurons that directly control spinal motor neurons essential for swimming69 and provide the primary source of dopamine to the spinal cord neurons including the central pattern generator (CPG) that control body movements30, 34, 86. We therefore propose that appropriate levels of dopamine neurotransmission from the diencephalon is essential to regulate locomotion either directly from the DDC to the spinal CPG or via the mesencephalic locomotor region (MLR) to the reticulospinal neurons (RS) in the hindbrain (Fig. 8) known to convey motor commands from the brain to the spinal circuits in zebrafish69, 90, 113–115. This study implicates that the dopaminergic circuits underlying locomotor control may be more susceptible or responsive to inflammatory signaling, thereby subject to innate immune modulation, although the precise mechanism remains to be determined.
In summary, this study leverages genetic and large-scale approaches in zebrafish using nlrc3l and a set of microglia-lacking mutants to reveal the capability of peripheral inflammatory macrophages in the absence of microglia to alter function of the locomotion-controlling brain circuits through recruitment of neutrophils to the brain. By contrast, we found that microglia loss alone had negligible consequence on spontaneous locomotor behaviors. Overall, we reveal novel interactions by which chronic peripheral inflammation evoked by inflammatory macrophages can intercept central neural circuits through co-opting neutrophils to infiltrate the brain and lead to spontaneous locomotor behavioral changes.
Materials and Methods
Zebrafish
Embryos from wild-type (TL and AB), mutant and transgenic backgrounds: nlrc3lst73 28, irf8st95 48, xpr1bst89 49, pu.1/spi1bfh509 50, mpeg1:EGFP 116, lyz:GFP 117, neurod:GCaMP6f 80, and nbt:DsRed 118 were raised at 28.5°C and staged as described 119. Stable transgenic mpeg1:nlrc3l line was generated using Tol2-mediated transgenesis based on a previously cloned construct 28 combining full-length wild-type nlrc3l coding sequence downstream of mpeg1(1.86 kb)116 regulatory sequence. Stable elavl3:GCaMP6s line was generated using Tol2-mediated transgenesis. This study was carried out in accordance with the approval of UNC-Chapel Hill Institutional Animal Care and Use Committee (protocols 16–160 and 19–132).
Whole mount RNA in situ hybridization
RNA in situ was performed using standard methods. Antisense riboprobes for mfap4, mpx, and irg1 were synthesized as described 28, 71, 120. Additional antisense riboprobes were synthesized using primers targeting coding sequences of dbh, gad1b, nr4a2b/nurr1, serta/slc6a4a, and vglut2a/slc17a6b (sequences detailed in Supplementary File 1).
RNA isolation and qPCR
RNA was isolated using the RNAqueous-Micro kit RNA Isolation Procedure (Ambion). Whole larvae were lysed in 100–300 uL RNA lysis buffer. cDNA was made from 150 or 200 ng of total RNA using oligo (dT) primer with SuperScript IV reverse transcriptase (Invitrogen). qPCR was performed on the QuantStudio 3 Real-Time PCR System (Applied Biosystems) using SYBR Green. The delta-delta ct method was used to determine the relative levels of mRNA expression between experimental samples and control. ef1a was used as the reference gene for determination of relative expression of all target genes. Primer sequences for qPCR analysis are listed in Supplementary file 1. All qPCR experiments were conducted using 3 technical replicates and a minimum of 3 biological replicates.
RNA-seq analysis
Paired-end RNA-Sequencing was performed on two independent biological replicates of 10 or 12 pooled 4-dpf larval zebrafish per genotypic group (nlrc3l mutants versus their siblings) by Illumina HiSeq 2×100 basepairs. The mutants were pre-sorted by neutral red for lack of microglia while the siblings, consisting of heterozygotes and wild-types, were sorted for presence of normal microglia. Genotypes were verified using PCR-restriction enzyme assay prior to library preparation. RNA-seq analysis followed previously developed protocol121, which consists of modules and packages including FastQC for quality control, Trimmomatic for data trimming, and HISAT2 for sequence alignment. The differential expression analysis was performed using DESeq2. Gene ontology analysis was performed on the differentially expressed genes with bioMaRt in R. All RNA-seq data has been submitted to the GEO repository to be accessed under the accession GSE179883.
In vivo time-lapse and static confocal imaging
All time-lapse and static z-stack imaging were performed using a Nikon A1R+ hybrid galvano and resonant scanning confocal system equipped with an ultra-high speed A1-SHR scan head and controller. Images were obtained using an apochromat lambda 40x water immersion objective (NA 1.15) or a plan apochromat lambda 20x objective (NA 0.75). Z-steps at 1–2 µm were taken at 40x and 3–5 µm at 20x. Different stages of zebrafish were mounted on glass-bottom dishes using 1.5% low-melting agarose and submerged in fish water supplemented with 0.003% PTU to inhibit pigmentation. All image acquisition parameters (including laser power, gain, scan mode/speed, and resolution) were set constant for experiments requiring fluorescence quantification. Raw data was used for quantification, and brightness/contrast adjustments may be independently applied to whole images only for the purpose of clarity in figure presentation.
96-well or 24-well behavior experiment setup
Heterozygous intercrosses were conducted to generate homozygous mutants that were co-housed and monitored in identical conditions as their heterozygous and wild-type siblings. Embryos were raised with daily water change and health checks at a density of 50-100 per dish at 28.5 C without the light:dark cycles until they reached 5 dpf. Fish were checked for normal gross morphology including an inflated large swim bladder and normal motility before and after behavioral tracking. Although infrequent, any larvae that appeared sick or have deflated swim bladder were eliminated from further analyses. Individual 5 dpf larvae were placed alone in each well of a 96-well (Fig. 2a) or 24-well plate (Fig. 3a, Supplementary Fig. 2), and simultaneously tracked in the customized ZebraBox built by ViewPoint Behavior Technology. At 5 dpf we are able to avoid possible confounding effects from feeding since the larval zebrafish can survive on their yolk contents. The automated tracking system recorded the location of every animal on the plate continuously using an ultrasensitive infrared CCD camera. Because larvae undergo spontaneous burst and glide swimming interspersed with short periods of inactivity, data was automatically collected at 10-minute integrations43 for 72-hour tracking under 14-hour light: 10-hour dark cycles (432 total timepoints)(Figs. 2, 3), or 1-minute integrations for the 6-hour light-dark visuomotor assay (360 total timepoints)(Supplementary. Fig. 1). The larvae were housed in the ZebraBox at room temperature to maintain them at around 6 dpf so the impact of development is minimal while preserving their active and healthy state. 96 deep well polypropylene plates (0.7 ml) with flat bottom and square wells (EK-2074, E&K Scientific; 201242-100, Agilent), or 24 well cell culture clear plates with flat bottom and circular wells (T1024, Thomas Scientific) were used. For 72-hour experiments, water evaporation was minimal in the 24-well setup but significant in the 96-well plate, which required a daily replenishment of 50-100 ul of fish water per well. Animals were genotyped after completion of behavioral tracking to ensure unbiased data collection.
Large-scale locomotor behavior analysis
High-throughput locomotor behavior data collected was mined and quantified in Matlab and R. The outputs from ZebraBox for each experiment include a spreadsheet of raw data for all animals at all timepoints by variables (such as counts, distance, duration of movements) and the images of movement traces, which were integrated over a 10-minute window for the entire 3-day duration of tracking. Extraction of features such as distance and duration of movements at time points and periods of interest was achieved in Matlab. Key locomotor metrics (distance, frequency, and speed) for each individual fish were determined as an average across time periods of interest (such as d2 and d3 light periods for 72-hour tracking, unless otherwise noted). We did not include the first day given its variability during acclimation.
Locomotor pattern analysis
Matlab was used to process images of movement traces recorded at every timepoint corresponding to daytime d2 and d3 periods to digitally segment each well into 9 equal squares in a 3 x 3 grid to determine pixel values for each square of the grid for all fish at all timepoints of interest (Fig. 3g). R program was used to compute the perimeter and center scores from the outputs of the pixel-based pattern analysis in Matlab. The perimeter swimming was quantified using all but the center square for each individual at every timepoint; each outer square was assigned 0 for pixel value < 9 (baseline no movement) or 1 for pixel value > 9, then these numbers were summed to generate a single perimeter score per individual per time point, ranging from 0 (no perimeter swimming) to 8 (full perimeter swimming) (Fig. 3g). The center square was evaluated directly by its pixel value (<100 for little to no center activity and >550 for high center activity). Downstream statistical analyses of the quantified locomotor patterns, such as hierarchical clustering, were performed and visualized using various R packages, including heatmap.2 and hclust, or GraphPad Prism 9.
Visuomotor assay using light-dark conditions
For the visuomotor assay, the animals were subjected to a series of defined timed intervals of alternating 100% light or 0% light (for dark) conditions implemented by the built-in light control program in the Zebrabox and locomotor behaviors were recorded every minute for a total tracking time of 6 hours (Supplementary Fig. 1). We programmed our system light control to begin with a 3-hour full lighting for an acclimation period followed by alternating dark and light periods dispersed over four 30-minute intervals and six 10-minute intervals to allow assessment of 5 repeated light-to-dark switches (Supplementary Fig. 1). For analysis, we extracted data from specific periods of response to visual stimuli, and calculated locomotor metrics for each individual animal.
In vivo brain calcium imaging to measure spontaneous daytime neural activity
Double transgenic larval zebrafish carrying neurod:GCaMP6f and nbt:dsRed at 5 dpf were acclimated for ∼ 24 hours in 14h light: 10h dark cycle prior to brain imaging at 6 dpf. Two z-planes (tectum at ∼50 microns below skin and diencephalon at ∼ 55 microns below tectum slice) were imaged at 1 scan per second for 10 minutes in ambient light using resonant scanning at 16X line integration on a Nikon A1R+ confocal microscope with a 40x water immersion objective (NA 1.15). Genotyping ensued imaging. GCaMP6f fluorescence was extracted for ROIs in the telencephalon, diencephalon, tectum and cerebellum over the entire imaging period using ImageJ. R program was used to calculate spontaneous neural activity of the ROIs by ΔF/F0, either from discrete regions of interest or random sampling of individual neurons. Baseline calcium (F0) of each ROI was determined by the average fluorescence intensity over all time points. Number of calcium spikes was determined using Matlab by setting a threshold ΔF/F0 value for each brain region ROIs to define the number of discrete events at which ΔF/F0 is above the set threshold with at least one timepoint gap from the next spike event. Relative somatic calcium fluorescence for each ROI was calculated as an average of the relative changes in GCaMP6f fluorescence from baseline fluorescence over all timepoints of the 10-minute calcium imaging per fish. Due to the varying neurod:GCaMP6f reporter expression intrinsic to Tol2-mediated transgenesis and random sampling, we found GCaMP6f signals consistently strong in the telencephalon but not in the diencephalon that already has low baseline fluorescence. Animals with GCaMP6f signals that cannot be delineated in at least a subpopulation of the diencephalic neurons were excluded from diencephalon analysis. Pan-neuronal elavl3:GCaMP6s transgenic zebrafish were acclimated, imaged at 1 Hz for a minimum of 5 minutes in ambient light using confocal imaging with a 40x objective, and analyzed using the same pipeline as above.
Small-molecule anti-inflammatory drug treatment
Administration of different small-molecule drugs was conducted by adding each individual chemical into the fish water starting at ∼24 hours prior to behavioral tracking at 5 dpf. Fish water freshly supplemented with the chemical was continuously used in the behavioral plate in which the larvae were individually housed in each well. Daily replenishment of chemical-supplemented water of about 50-100 ul was applied to refill the wells in the 96-well format plate due to evaporation. Bay 11–7082 (1 µM), dexamethasone (3.25 µM), and 17-DMAG (5 µM) were used as listed in Supplementary file 1. Behavioral tracking ensued for 72 hours in normal day-night cycles.
Gene knockdown by morpholino microinjection
Morpholino oligomers were injected into one-cell stage embryos from nlrc3l heterozygous intercrosses. Antisense morpholino oligos were purchased from Gene Tools and re-suspended in water to make 1-3 mM stocks. Morpholino sequences are listed in Supplementary file 1. Morpholinos were heated at 65°C for 5 min and cooled to room temperature before microinjection at 1 nL. Efficacy of the morpholinos were validated based on the known phenotype. Quantification of neutrophils at 2 dpf was based on a cell count and at 5-6 dpf based on total area of neutrophils in a ROI using ImageJ. Area was used at later stages due to an increased presence of clustering and interconnected neutrophils, which makes cell counting less reliable.
Neutral red staining
Microglia were examined in live larvae by neutral red vital dye staining as previously described28, 48. Larvae at 3-4 dpf were stained by immersion in fish water supplemented with 2.5 μg/mL neutral red and 0.003% PTU at 28.5°C for 1 hr, followed by 1–2 water changes, and then analyzed using a stereomicroscope.
TH immunostaining
Standard whole mount tissue immunostaining was followed using the rabbit anti-Tyrosine Hydroxylase antibody (AB152, Sigma Aldrich) at 1:100 dilution. Briefly, zebrafish larvae were fixed overnight in 4% paraformaldehyde followed by methanol permeabilization, and incubated with primary antibody in blocking solution overnight at 4°C and then secondary antibody, goat anti-rabbit Alexa Fluor 594 (ThermoFisher), at 1:3000 dilution in blocking solution for detection.
Statistical analyses
Unpaired two-tailed t-tests were performed unless otherwise noted. F test was used to compare variances. For unequal variances, Welch’s correction was used on the two-tailed t-test. For multiple comparisons of 3 or more groups, one-way ANOVA test was applied followed by multiple pair-wise comparisons to determine the pair(s) showing significant differences using FDR-adjusted p-values. All graphical plots and statistical tests were generated using GraphPad Prism 9 unless otherwise noted. All box-whisker plots show the median, first and third quartiles, minimum, and maximum of the data plotted.
Author contributions
CES: conceptualization, data analysis and visualization, supervision, funding acquisition, behavior experiments and computational analyses, generation of rescue transgenic line, writing-original draft; VK: experimentation (behavior, RNA in situ, imaging, immunohistochemistry, genotyping, morpholinos, qPCR) and data analysis; PC: computational pipeline for behavior analyses and RNA-seq with CES, Matlab and R scripts, data analysis; CTD: experimentation (behavior, RNA in situ, calcium and confocal imaging, genotyping, qPCR); VH: experimentation (behavior, imaging, genotyping, qPCR); CGS: experimentation and analysis of calcium imaging; AMR: experimentation (macrophage rescue characterization, qPCR) and all schematic illustrations. All authors reviewed and edited the manuscript.
Competing interests
None.
Acknowledgements
We thank Ana Meireles, William Talbot, John Rawls, Cecilia Moens and Claire Wyart for the generous sharing of fish lines and reagents. We are grateful for the substantial insights and discussions from Jonathan Gable on our manuscript. Michelle Altemara and her staff at the UNC Zebrafish Aquaculture Core Facility conducted critical support for zebrafish housing and care. Pilot behavior experiments prior to this study benefited from critical insights and inputs given by Constance Richter and Alex Schier. This work was supported by NIH NIGMS grant 1R35GM124719 to C.E.S.
Footnotes
↵# Authors of equal contribution.
References
- 1.↵
- 2.↵
- 3.↵
- 4.
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.
- 11.↵
- 12.↵
- 13.
- 14.
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.
- 20.
- 21.↵
- 22.↵
- 23.↵
- 24.
- 25.
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.
- 32.
- 33.
- 34.↵
- 35.↵
- 36.
- 37.↵
- 38.↵
- 39.↵
- 40.
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
- 96.↵
- 97.↵
- 98.↵
- 99.↵
- 100.↵
- 101.↵
- 102.↵
- 103.↵
- 104.↵
- 105.↵
- 106.
- 107.
- 108.↵
- 109.↵
- 110.
- 111.
- 112.↵
- 113.↵
- 114.
- 115.↵
- 116.↵
- 117.↵
- 118.↵
- 119.↵
- 120.↵
- 121.↵