Reciprocal connections between brain and liver are apparent in various conditions, including encephalopathy and encephalitis after severe liver damage (Butterworth, 2013; Felipo, 2013), liver disruption after traumatic brain injury (Villapol, 2016; Lustenberger et al., 2011), and intracerebral injection of pro-inflammatory cytokines (Campbell et al., 2005). However, the routes of communication directly linking brain to liver remain poorly understood. To investigate one possible avenue of this, we sought to determine whether a brain perturbation such as inflammation could trigger macrophage activities corresponding to a response by the liver or other peripheral organs. To this end, we directly microinjected bacterial lipopolysaccharides (LPS) or E. Coli cells, well-established immune activators, into the brain tectum at four days post-fertilization (dpf) at a stage known to have a well formed brain with an established BBB (Jeong et al., 2008) and choroid plexus/ventricle system (Henson et al., 2014; Fame et al., 2016) as previously described (Earley et al., 2018). Stage-matched animals that were not injected or injected with the water vehicle in the brain were used as the control group (Figure 1 and Figure 1—figure supplement 1). To analyze a possible change in peripheral macrophages after brain-LPS injection at four dpf, we used whole-mount RNA in situ hybridization for a macrophage marker, mfap4, to characterize the macrophage distribution in the whole body. Strikingly, we found a robust and distinctive aggregation of macrophages in the liver and near the brain injection site, but not apparently elsewhere (Figure 1—figure supplement 1). By contrast, the uninjected and water injected controls at four dpf were devoid of macrophages in the liver (Figure 1—figure supplement 1). We also found at least some of these macrophages in the liver to be activated based on their expression of the mitochondrial enzyme gene irg1/acod1 (Figure 1—figure supplement 1) known to be highly upregulated during inflammation and specifically induced in inflammatory macrophages in zebrafish (Lampropoulou et al., 2016; Sanderson et al., 2015). Both injections of LPS or live E. coli cells into the brain led to macrophage presence in the liver, albeit LPS effects were consistently stronger (Figure 1—figure supplement 1).

To relate macrophage presence in the liver after brain-LPS injection to macrophages that normally reside in the liver (which we refer to as Kupffer cells), we analyzed the developmental timing of Kupffer cells, which had not been previously described, starting at the four dpf larval stage to the 40 dpf juvenile adult stage (Figure 1a). We found normally an absence of Kupffer cells at the stage of our brain microinjection at four dpf (0.2 ± 0.4 standard deviation (s.d.) per liver), few to none at 5–6 dpf (4.0 ± 2.2 s.d. per liver), and once larvae were fed and moved into the fish facility, Kupffer cell numbers grew substantially (78.2 ± 14.9 s.d. per area of liver at 26 dpf) reaching to hundreds per liver in the juvenile adults (26–40 dpf at 9–12.5 mm standard length) (Figure 1a). Prior to this work, Kupffer cells were thought to be missing or sparse in zebrafish and other teleost species (Goessling and Sadler, 2015; Sakano and Fujita, 1982), but recent work tracing adult zebrafish Kupffer cells to their hematopoietic origin (He et al., 2018), and the data presented here collectively provide the first evidence for the prevalence of Kupffer cells in the zebrafish liver akin to their mammalian counterpart. Since the brain-LPS injection and subsequent analysis were conducted at four dpf, before the establishment of Kupffer cells, the presence of macrophages in the liver was likely because of active recruitment of peripheral macrophages and monocytes.

We subsequently conducted a time-course analysis of macrophage activities after brain-LPS injection at four dpf to determine if macrophages were actively infiltrating the liver. Using in vivo static and time-lapse imaging in double transgenic zebrafish expressing both the macrophage reporter mpeg1:GFP and the liver hepatocyte reporter fabp10a:DsRed, we observed macrophages actively migrating or circulating into the liver, affirming the in situ results (Figure 1—figure supplement 1). We captured macrophage dynamics in the liver region at 8–10 hr post-injection (hpi) of LPS in the brain (Video 1 and Figure 1—figure supplement 2) when substantial numbers of infiltrating macrophages can be observed. Conversely, the brain-water injected controls mostly had zero macrophages in the liver (average of 1.4 ± 1.2 standard deviation (s.d.) macrophages per liver) compared with an average of 13.2 ± 6.6 s.d. macrophages per liver after brain-LPS injection (Figure 1b–e). Later at 24–30 hpi and 48 hpi, significant numbers of macrophages in the liver persisted even two days after brain-LPS injection (11.2 ± 6.3 s.d. and 13.1 ± 6.1 s.d. macrophages per liver, respectively) (Figure 1c, Video 2 and Figure 1—figure supplement 2). At these later timepoints, a few Kupffer cells may begin to emerge at less than five per liver (Figure 1c). In agreement with the timeline of Kupffer cell development (Figure 1a), only a few macrophages were detectable in uninjected and water-injected controls at the two later timepoints (3.1 ± 1.8 s.d. and 3.9 ± 2.8 s.d. macrophages per liver, respectively) (Figure 1c). Macrophages in the liver at 48 hpi appeared more stationary than earlier at 8–10 hpi (Videos 1 and 2, and Figure 1—figure supplement 2). In all timepoints of analysis, infiltrated macrophages in the liver were found in several locations, including inside the sinusoids (liver microvessels) similar to that described during mammalian liver injury (Iwakiri et al., 2014), and surprisingly also in the parenchyma intermingling with hepatocytes (Figure 1d–e, Videos 1 and 2, Figure 1—figure supplement 2), a macrophage behavior previously not known. Due to some examples of broad liver mfap4 in situ expression (Figure 1—figure supplement 1) in comparison to a discrete number of infiltrating macrophages by live imaging after brain-LPS injection (Figure 1), we assessed whether this could be explained by an induction of ectopic mfap4 expression in the liver upon LPS activation (Figure 1—figure supplement 3). Using a transgenic line mfap4:tdTomato to mark cells expressing the mfap4 gene, we found mfap4 restricted to macrophages and absent in liver cells (Figure 1—figure supplement 3), suggesting the broad liver mfap4 expression may be diffuse in situ signals coming from a liver more densely populated by infiltrated macrophages. To analyze the nature of physical contact of infiltrated macrophages to the hepatic sinusoids after brain-LPS injection, we imaged double transgenic zebrafish expressing the endothelial (kdrl:mCherry) and macrophage (mpeg1:GFP) reporters. Time-lapse imaging showed that macrophages can actively infiltrate the hepatic sinusoids, as well as be associated or entirely independent of the hepatic vasculature (Figure 1—figure supplement 4 and Video 3).

To determine whether this infiltration endured over time, we examined the number of liver-infiltrating macrophages over a two-day period after brain-LPS injection. Results indicate no significant change in macrophage presence in liver, suggesting that infiltrated macrophages either stayed in the liver, moved in and out of the liver at similar rates, or both. To distinguish these possibilities, we tracked these infiltrating macrophages directly in vivo over a continuous 10 hr period after brain-LPS injection at four dpf starting at 12 hpi. Using live cell tracking, we found a small number (~13%) of infiltrating macrophages that stayed in the liver longer than 2 hr, including occasional infiltrates which remained in the liver past the total duration of imaging (>9 hr) (Figure 1—figure supplement 5 and Video 4). Infiltrating macrophages were coined ‘residing’ when they occupied the liver for more than 2 hr, and these were on average moving slower at 1.1 ± 0.7 µm/min than the ‘transient’ population at 2.2 ± 1.1 µm/min which occupied the liver for more than one timepoint (2 min) but less than 2 hr (Figure 1—figure supplement 5). We found a significant fraction (~30%) of infiltrating macrophages to be of the ‘transient’ type, while the majority at 57.4% were circulating (detected only in a single timepoint) (Figure 1—figure supplement 5). Both the ‘residing’ and ‘transient’ populations were not truly stationary but rather moved dynamically back and forth across the liver parenchyma amounting to large total distances traveled over time (309.2 ± 285.9 µm s.d. and 62.8 ± 66.7 µm s.d. in 560 min of tracking, respectively) (Figure 1—figure supplement 5 and Video 4). These results indicated that both the presence of short- and long-term occupying macrophages accounted for the sustained large macrophage number in the liver even two days after brain-LPS injection (Figure 1c).

To understand how LPS in the brain may lead to liver effects, we first determined whether the injected LPS remained restricted to the brain or possibly transferred to the periphery over time. We used fluorescently tagged LPS to directly track the LPS molecules after brain tectum microinjection in the whole body for a continuous 24 hr period (Figure 2 and Video 5). This provides a means to visualize binding, transport, and internalization of LPS in the brain and body. Fluorescently tagged dextran was used as a control tracer to analyze the general molecular distribution independent of LPS (Figure 2 and Video 6) and to verify successful injections. Initially the LPS macromolecules injected into the brain parenchyma were restricted to the focal location of the injection site but quickly within seconds they filled the cerebral ventricles joining the cerebrospinal fluid (CSF) as they continue to flow into the spinal canal in an anteroposterior direction (Figure 2 and Video 7), thus our zebrafish brain microinjections are comparable to mammalian intracerebroventricular injections (Glascock et al., 2011).

To evaluate the dynamics and major routes of LPS passage to the general circulation, we used high-speed and high-sensitivity stereomicroscopy to trace the movement of fluorescently tagged LPS in real time at one frame per second starting before the brain tectum microinjection to almost one hour after the injection (Figure 2c–d and Video 7). While most of the LPS remained restricted within the ventricular system, we found the hindbrain ventricle (hbv) to be the key region from which LPS spread into the parenchyma and surrounding interstitial space, especially in the dorsal-most portion of the junction between the hindbrain and spinal cord encompassing the dorsal longitudinal anastomotic vessels (DLAVs) and the dorsal longitudinal lymphatic vessel (DLLV) (Figure 2—figure supplements 1 and 2, and Video 7). LPS at a low level were detected to exude out from the hbv into the hindbrain interstitial fluid (ISF) starting at around 1 min after injection (Video 7 and Figure 2c) and reached a maximum of about 30% of LPS injection level at 50 min after injection (Figure 2—figure supplement 3). After a delay of about 13 min after injection, the first LPS fluorescence signals were faintly measured in the pronephros, a region assessed as a proxy for general circulation (Figure 2c–d and Video 7). Peripheral LPS level increased over time but remained low at less than 15% of the LPS injection level (Figure 2—figure supplement 3).

To characterize LPS at the tissue and cellular level along its outflow path, we used high power confocal imaging on the same LPS-injected zebrafish which were tracked by stereomicroscopy (Figure 2 and Video 7) to carry out further imaging at later timepoints. Interestingly, LPS molecules were taken up by brain lymphatic endothelial cells (blec), also known as fluorescent granular perithelial cells (FGPs), as well as by facial and trunk lymphatic vessels as shown by co-localization of LPS with the lymphatic reporter mrc1a:GFP at 1.5 hpi (Jung et al., 2017; Figure 2—figure supplements 1 and 2), suggesting LPS exited through these lymphatic structures. These results were consistent with the previous LPS tracing within the first hour after injection (Figure 2 and Video 7), but shown more definitively by high-resolution confocal imaging. By contrast, LPS were not localized within the CNS vasculature using the endothelial reporter kdrl:mCherry, but only in the peripheral blood vessels (Figure 2—figure supplement 2), indicating that transport of LPS did not likely result from a disruption of the BBB or a direct transit through brain blood vessels. Tracing the movement of LPS from the brain to the periphery also revealed its transient flow through the liver sinusoids prior to immune cell infiltration (Figure 2, Video 5). However, LPS did not appear to accumulate or bind to cellular structures within the liver as we did not detect LPS there (Figure 2b, Figure 2—figure supplement 2 and Video 5), suggesting either the transient exposure of hepatic cells to LPS, or yet unknown extrahepatic signals trigger the hepatic response to systemic LPS. Taken together, LPS appeared to enter circulation via a mechanism directed by the lymphatics for clearing away excess substances from the brain parenchyma and interstitial fluids (Figure 2e).

In light of the broad LPS distribution, we sought to functionally test whether the liver response after brain-LPS activation was due to systemic LPS. To create a systemic LPS condition akin to mammalian models of sepsis or endotoxemia (Szabo et al., 2002; Mathison and Ulevitch, 1979), we directly injected LPS intravenously at the caudal plexus into the bloodstream and compared its resulting peripheral response to that after brain-LPS injection (Figure 2). These injections resulted in different outcomes for microglial activation at six hpi, whereby brain-LPS led to a strong activation of microglia but intravenous (IV)-LPS did not (Figure 2f). Both routes, however, led to the same robust macrophage infiltration of the liver (Figure 2g). To determine if circulation was required for the liver response after brain-LPS injection, we used a morpholino to knockdown tnnt2, a cardiac muscle troponin T gene, a well-established reagent for blocking circulation (Sehnert et al., 2002; Figure 2—figure supplement 4). We indeed found that inhibiting circulation prevented macrophage infiltration into liver after brain-LPS injection (Figure 2—figure supplement 4). Taken together, several lines of evidence show that circulating LPS that causes systemic inflammation was driving macrophage infiltration into the liver after brain-LPS injection: 1) systemic distribution of LPS, 2) sufficiency of IV-LPS injection to cause liver infiltration, and 3) blocking circulation prevented macrophage infiltration of the liver.

We next tested the effects of anti-inflammatory drugs on liver response after brain-LPS microinjection to address whether systemic inflammation was indeed responsible for the macrophage recruitment into the liver. We screened five small-molecule drugs (GW2580, 17-DMAG, celastrol, Bay 11–7082, and dexamethasone) known to effectively curb inflammation by attenuating NF-kB mediated transcription or activating glucocorticoid functions in zebrafish and other systems (Conway et al., 2005; Shimp et al., 2012; Sevin et al., 2015; Coffin et al., 2013; Lancet et al., 2010; Venkatesha et al., 2012; Yu et al., 2010; Aghai et al., 2006; Yamamoto and Gaynor, 2001; Lee et al., 2012). These small molecules are known to act through different mechanisms: GW2580, a selective inhibitor of cFMS kinase that blocks the receptor tyrosine kinase CSF1R function which can prevent NF-kB activation (Caescu et al., 2015; 17-DMAG (a water-soluble geldanamycin analog) and celastrol, both potent inhibitors of the heat-shock protein Hsp90 that cause disruption or degradation of its target proteins, including the NF-kB protein complex (Shimp et al., 2012; Lee et al., 2006; Bay 11–7082, an inhibitor of E2 ubiquitin (Ub) conjugating enzymes, which target NF-kB inhibitor, IkB-alpha, for proteasomal degradation (Lee et al., 2012; and dexamethasone, an agonist of the glucocorticoid receptor (GR) that activates a negative feedback mechanism to reduce inflammation (Aghai et al., 2006). We first assessed the effects these small molecules had on macrophage infiltration into the liver after brain-LPS microinjection using whole-mount RNA in situ hybridization with the macrophage marker mfap4 (Figure 3—figure supplement 1). As positive controls for liver infiltration, we used brain-LPS injected animals that were untreated (water only) for comparison with water-reconstituted drugs (17-DMAG and dexamethasone), and treated with only DMSO for comparison with DMSO-reconstituted drugs (GW2580, celastrol, and Bay 11–7082). Untreated animals without brain microinjection were also used as negative controls. By in situ analysis, we found that Bay 11–7082 and dexamethasone substantially reduced the frequency of macrophage infiltration after brain-LPS injection compared with control groups (Figure 3—figure supplement 1). These effects were further validated by in vivo imaging of macrophage recruitment into the liver, which enabled a precise macrophage count at a high cellular resolution (Figure 3—figure supplement 1). These results indicated that macrophage infiltration into the liver after brain-LPS stimulation can be prevented by suppressing inflammation via mechanisms inhibiting the NF-kB pathway or activating the glucocorticoid signaling.

To examine possible components affiliated with the NF-kB pathway that may drive macrophage recruitment into the liver, we examined whether an intracellular signal adaptor protein myeloid differentiation protein-88 (MyD88) of Toll-like receptors (TLRs) known for recognizing LPS and mediating cytokine production was essential (Medzhitov, 2001). We employed an effective myd88 specific splice-blocking morpholino to knockdown myd88 as previously described (van der Sar et al., 2006) during the liver response to brain-LPS stimulation (Figure 3a,b). The efficacy of myd88 morpholino to mediate splice-blocking was confirmed by RT-PCR analysis (Figure 3c). By in vivo imaging, we found that the number of liver-infiltrating macrophages after brain-LPS microinjection was significantly reduced when myd88 function was disrupted (Figure 3a,b). To further validate these results, we performed the brain-LPS and control injections in myd88 null mutants and their siblings derived from a heterozygous incross (Figure 3d). myd88 mutants showed either few or no macrophages in the liver after brain-LPS injection at 16–24 hpi similar to baseline brain-water injected animals (Figure 3d), demonstrating a much stronger effect in reversing macrophage infiltration than the partial myd88 knockdown by morpholinos. These results show that the inflammatory liver response depended on the MyD88 pathway and are consistent with LPS-triggered inflammation as the driver of macrophage infiltration into the liver.

As possible hepatic signals that recruit macrophages into the liver during systemic inflammation, we examined whether the interleukin-34 (IL-34) and colony stimulating factor-1 (CSF-1) that share a common receptor CSF1R have a role. IL-34 and CSF-1 are known to mediate various functions of macrophages including inflammatory processes and promoting production of pro-inflammatory chemokines (Caescu et al., 2015; Sauter et al., 2016; Masteller and Wong, 2014). They have recently been shown to be required for macrophage migration and colonization of the brain to form microglia in zebrafish (Oosterhof et al., 2018; Kuil et al., 2019; Wu et al., 2018). Interestingly, engineering an artificial expression of il-34 in hepatocytes has been shown to be able to recruit macrophages to the liver in zebrafish, but the physiological relevance was not known (Jiang et al., 2019). In light of these previous studies, il-34 and the two zebrafish orthologs of CSF-1 gene (csf1a and csf1b) were strong candidates for attracting macrophages to the liver after brain-LPS injection. To examine this possibility, we first determined whether these genes (il-34, csf1a, and csf1b) were upregulated in the liver after brain-LPS injection using quantitative PCR (qPCR) analysis on liver-specific and body-minus-liver tissues compared with control animals without the brain microinjection (Figure 3e). We found that while csf1a and csf1b were either not detected or had no difference in the liver with or without LPS injection, il-34 was significantly upregulated in the liver after brain-LPS microinjection (Figure 3e). This upregulation was specific to the liver as il-34 was not elevated in the body-minus-liver tissue after brain-LPS injection (Figure 3e). To test if il-34 was a required cytokine for recruiting macrophages in our experimental paradigm, we used CRISPR/Cas9 targeted mutagenesis as previously described (Earley et al., 2018; Oosterhof et al., 2018) to disrupt il-34 function to assess the immune infiltration (Figure 3f–g, Figure 3—figure supplement 2). To verify the efficacy of the gene knockdown in transient il-34 F₀ Crispr-injected animals, we determined whether they phenocopied a reduced microglia phenotype recently described in il-34 stable mutants (Kuil et al., 2019). Indeed we found about 40% of the transient il-34 F₀ Crispr-injected animals to have highly decreased microglial numbers (Figure 3f). By Sanger sequencing analysis, we verified that these F₀ Crispr-injected animals induced a high frequency of frameshift indels altering reading frames and introducing early stop codons in the il-34 locus (Figure 3—figure supplement 2). We found that transient il-34 F₀ Crispr-injected animals after brain-LPS injection indeed had a significantly reduced number of infiltrating macrophages compared with controls (no Crispr injection) after brain-LPS injections (Figure 3g). To confirm these results, we also tested the effects of brain injections in stable il-34 null mutants and their control siblings derived from a heterozygous incross (Figure 3h). il-34 mutants had either few or no macrophages in the liver after brain-LPS injection at 16–24 hpi similar to baseline brain-water injected animals (Figure 3h), providing clear evidence that il-34 is essential in recruiting macrophages to the liver after brain-LPS activation. Taken together, these results showed that macrophage infiltration into the liver was driven by inflammatory processes that require the myd88 pathway as well as il-34 signaling likely coming from the liver.

Since inflammation is typically a concerted response of the innate immune system, the liver response after brain-LPS stimulation may involve other immune cells besides macrophages. To examine this possibility, we investigated whether neutrophils, the other functional leukocytes prominent at early larval zebrafish stages (Bennett et al., 2001; Xu et al., 2012), could participate in infiltrating the liver along with macrophages. Using live imaging in transgenic zebrafish at four dpf prior to Kupffer cell development, we quantified the numbers of neutrophils and macrophages in the liver 8–10 hpi with control water or LPS, as well as in uninjected controls (Figure 4a). We performed the same brain injection experiments also at 8–10 dpf after Kupffer cell establishment (Figure 4b). Our data interestingly showed that neutrophils also significantly infiltrated the liver after brain-LPS but not in the control brain-water and uninjected groups. Similar to macrophages, neutrophils infiltrated the liver irrespective of the presence of Kupffer cells (Figure 4a–c). To test whether liver infiltration would still occur at juvenile adult stages, when the liver, blood and lymphatic vasculature, and brain structures are fully mature, we conducted brain tectum injections in 1 month-old zebrafish which had a standard length of 0.8–1 cm (Figure 4—figure supplement 1). We found at 16–18 hpi that the numbers of macrophages and neutrophils in the liver were significantly increased after LPS injection compared with water vehicle injections and uninjected controls (Figure 4—figure supplement 1), indicating that the liver infiltration persists even at young adult stages. These results implicate that the hepatic response to brain-LPS injection is independent of Kupffer cells, age, or maturity of the brain vasculature and architecture, and drainage of LPS from brain to periphery may still be evident in adulthood. They also raise the possibility that macrophages coordinate with neutrophils for moving into the liver, and signals other than from Kupffer cells can recruit these leukocytes into the liver during systemic inflammation induced by circulating LPS.

To interrogate a possible coordination between macrophages and neutrophils, we turned to gene perturbations to examine the functional consequence of eliminating one immune cell type on the other. We utilized a previously established morpholino targeting irf8, an essential transcription factor for macrophage formation, whose effect phenocopies the macrophage-lacking irf8 mutants for ablating all macrophages at embryonic and early larval stages (Li et al., 2011; Shiau et al., 2015). Assessment of neutrophil infiltration was conducted at 3–5 hpi, which is earlier than the timepoint of macrophage analysis, because we found neutrophils to infiltrate the liver first before macrophages (Figure 4l, Video 8). Strikingly, a highly significant 3-fold increase of liver-infiltrating neutrophils was found in macrophage-ablated irf8 morpholino-injected animals after brain-LPS injection at four dpf compared with the control brain-LPS animals, while brain-water injected controls had no neutrophil infiltration into the liver with or without macrophages (Figure 4d). To verify these findings, we used a complementary approach for macrophage depletion by using clodronate-containing liposomes to induce apoptosis in macrophages as previously described (van Rooijen and Hendrikx, 2010; Figure 4—figure supplement 2). Significantly increased neutrophil infiltration of liver after brain-LPS injection in clodronate-mediated macrophage depletion was also found compared with the control response to brain-LPS (Figure 4—figure supplement 2). Both methods of macrophage depletion corroborate to show striking increases in neutrophil recruitment to the liver after brain-LPS microinjection. These results suggest that neutrophils may compensate for macrophage loss, or that they are normally constrained by macrophages in part possibly by being phagocytosed as they become impaired or die, as has been described in an immune response (Prame Kumar et al., 2018). However, these infiltrating neutrophils do not appear to fully make up for macrophage functions as they do not exhibit the same dynamic movements or lasting occupation in the liver during the hepatic response to brain-LPS (Figure 4j). Most of the infiltrating neutrophils are circulating through the liver (>80%) regardless of macrophage presence in the liver during brain-LPS response, albeit a larger fraction is transient (22%) and long-term (1%) after macrophage ablation (Figure 4j). Our data therefore favor the latter explanation that macrophages may limit ongoing recruitment of inflammatory neutrophils.

Conversely, we employed a splice-blocking morpholino targeting the granulocyte colony-stimulating factor receptor (CSF3R/GCSFR) for reducing neutrophil numbers in zebrafish (Pazhakh et al., 2017). We verified by in situ gene expression pattern of the neutrophil marker, mpx, that csf3r morpholino-injected animals indeed had neutrophil numbers reduced on average ~30% with a maximum of a 65% decrease (Figure 4—figure supplement 3). Also, by in vivo imaging of larval zebrafish carrying both macrophage and neutrophil reporters, the csf3r morpholino-injected animals showed no significant change in macrophage number, but a substantial neutrophil reduction (Figure 4—figure supplement 3). Although we cannot completely rule out this possibility, these results indicate csf3r knockdown likely does not impact baseline macrophages, consistent with neutrophil-specific reduction and functional defects shown in zebrafish csf3r mutants (Pazhakh et al., 2017; Basheer et al., 2019). In these neutrophil-reduced csf3r morpholino-injected animals after brain-LPS injection, we found significantly reduced macrophage infiltration of liver by in situ and live imaging analyses (Figure 4 and Figure 4—figure supplement 3). As a complementary approach, we assessed the effect of csf3r gene disruption by Crispr injection on macrophage infiltration after brain-LPS injection and also found a significant decrease (Figure 4—figure supplement 4). Both morpholino- and F₀ Crispr injection- mediated csf3r knockdown corroborate to suggest a possible neutrophil role in recruitment of macrophages into the liver during inflammation. While neither method yields a total csf3r elimination to deplete most neutrophils, a stronger hindrance to macrophage infiltration may be possible from a complete csf3r depletion. In agreement, constant and dynamic intermingling of these two immune cell types were found around the liver normally (Figure 4k and Video 9). These results taken together show that during liver inflammation, macrophages and neutrophils coordinate intimately, and macrophages are at least in part driven by neutrophils to enter the liver as they infiltrate the liver first (Figure 4l, Video 8).

The functional consequence of macrophage and neutrophil infiltration on the liver during a systemic inflammatory response remains unclear. To examine this, we assessed transcriptional changes comparing immune infiltration and lack thereof in the absence of macrophages and neutrophils. Previous studies have implicated the role of infiltrating innate immune cells in causing liver injury and inflammation (Karlmark et al., 2008) but whether this effect is conserved in this model of brain-triggered systemic inflammation is unknown. We examined by qPCR analysis a set of 8 genes after brain-LPS injection with or without innate immune cells compared with the control brain-water injected group to gauge the long-term effect of immune infiltration at 48 hpi. Disruption in liver homeostasis can be defined by an upregulation of genes associated with inflammation (such as pro-inflammatory cytokines tnfa and il1b). Liver under stress, trauma and inflammation is also characterized by activating an acute phase response (Ramadori and Christ, 1999) (such as serum amyloid A (saa) and interleukin six signal transducer (il6st/gp130)). Disrupted liver can also exhibit altered expression level of genes associated with liver growth and function (such as alpha-2-macroglobulin-like (a2ml) (Hong and Dawid, 2008), aryl hydrocarbon receptor 1a (ahr1a) (Andreasen et al., 2002), and glutathione S-transferase pi 1 (gstp1) in zebrafish) (Abunnaja et al., 2017). To eliminate myeloid cells, we analyzed mutants with a loss-of-function in an essential myeloid transcription factor pu.1/spi1b (Figure 5) to eliminate macrophages/microglia and neutrophils (Li et al., 2011; Roh-Johnson et al., 2017). As a complementary approach, we also used an established translation-blocking morpholino targeting pu.1/spi1b (Figure 5—figure supplement 1) previously shown to eliminate macrophages/microglia and varying levels of neutrophils depending on dosage (Li et al., 2011; Su et al., 2007). The efficacy of pu.1 morpholino was phenotypically validated by a complete loss of brain macrophages (microglia) in a sub-sample of pu.1 morpholino-injected animals using a neutral red staining assay (n = 4/4) as previously described (Shiau et al., 2013). The qPCR results indicated that in the control situation with the full complement of macrophages and neutrophils at two days after brain-LPS injection in the four dpf zebrafish larvae, all genes associated with inflammation and acute phase response gene saa were significantly upregulated, while genes known to affect zebrafish liver development and function were either not changed or modestly increased (Figure 5 and Figure 5—figure supplement 1). By contrast, depletion of myeloid cells in pu.1 mutants and pu.1 morpholino-injected animals corroborated to show no significant upregulation in all inflammation genes analyzed (Figure 5 and Figure 5—figure supplement 1), indicating that the innate immune cells were responsible for the upregulation of inflammatory genes.

Furthermore, after brain-LPS injection compared with control siblings, saa remained mostly upregulated, but a2ml and ahr1a, two liver genes associated with growth or function had a net reduction in myeloid-deficient animals either by pu.1 knockout or knockdown (Figure 5 and Figure 5—figure supplement 1). The more significant reduction in liver genes as well as the larger change in acute phase response genes in pu.1 morpholino-injected animals than in pu.1 mutants may reflect a speedier recovery from the LPS injection in the mutants as they were not subjected to the early embryonic microinjection as the morpholino animals were. Alternatively, the possibility of not a total but partial depletion of macrophages, neutrophils, or both cell types in pu.1 morpholino-injected animals may also contribute to differences.

Since saa elevation after brain-LPS injection was not eliminated by an absence of macrophages, we further asked whether saa may be dispensable for macrophage infiltration. Comparing brain-LPS injections in saa^-/- mutants side-by-side with injected control siblings showed no difference between the genotypes (Figure 5—figure supplement 2), indicating that saa is not required for macrophage infiltration into the liver upon LPS activation. The downregulation of liver genes in myeloid cell-depleted animals after brain-LPS injection raised the question as to whether this reflected an actual change in liver growth. To address this, we used in vivo confocal 3D imaging to capture the whole liver at 48 hpi to measure liver size by volume (Figure 5—figure supplement 3). We found a decrease in liver size after LPS injection compared with water vehicle injection in both baseline and control-MO groups. Interestingly, by contrast, myeloid-lacking pu.1 morpholino-injected animals had an increase in liver size after brain-LPS injection (Figure 5—figure supplement 3), indicating liver growth was likely impeded by immune cell infiltration during inflammation. Since the expression of some liver genes decreased but the liver size increased in myeloid-deficient animals after brain-LPS injection, these liver genes may not be associated with growth but rather function. These results indicate that the liver infiltration by macrophages and neutrophils promoted inflammation and disrupted liver growth, but had minimal to no effect on the acute phase response gene saa, which was not essential for the infiltration (Figure 5—figure supplement 2). Taken together, they raise the possibility that immune cell infiltration leads to liver resources and functions being redirected from normal developmental growth to a full-fledged inflammatory response.

As a rapid means by which the brain can engage the periphery in response to possible danger, our study reveals how inflammatory molecules introduced into the brain parenchyma can circulate outside of the brain to initiate a robust hepatic response in zebrafish (Figure 6). It remains to be explored whether the transport of molecules or substances outside of the brain even at trace levels resulting from CNS trauma or injury can explain at least to some degree the associated hepatic damage described in mammals (Lustenberger et al., 2011; Catania et al., 2009; Louveau et al., 2016). The most likely route through which macromolecules can drain out of the brain is through the lymphatic system that removes interstitial fluids, waste products, and immune cells in zebrafish and mammals (Louveau et al., 2016; Jung et al., 2017; Okuda et al., 2012). Recent new knowledge in meningeal lymphatics and glymphatics indicate that macromolecules in the brain can flow from the parenchymal interstitial space into the cerebrospinal fluid (CSF) which drains into the lymphatic vessels or directly into the major veins (sinuses) to enter general circulation (Louveau et al., 2016; Jung et al., 2017; Iliff et al., 2012; Ma et al., 2017). In support of this, previous studies using a number of mammalian species (cat, rabbit, dog, and rat) have shown that injection of tracers at 5–100 ul volume such as horseradish peroxidase, India ink or dextran blue into the CSF or brain parenchyma gets drained rapidly into cervical lymph nodes within seconds as well as in entire vasculature within minutes (Zhang et al., 1992; Cserr and Ostrach, 1974; Casley-Smith et al., 1976; Bradbury and Cole, 1980; Bradbury et al., 1981; Szentistvanyi et al., 1984), while radiolabeled albumin microinjected into the brain parenchyma of different brain regions appear rapidly in the CSF and within 4 hr can be found in cervical lymph nodes and the common carotid arteries (Szentistvanyi et al., 1984; Harling-Berg et al., 1989). Furthermore, metabolites of glucose from the brain are found to be released in the cervical lymph nodes (Mergenthaler et al., 2013), supporting the possibility that drainage of endogenous macromolecules from brain to blood circulation can also happen. Taken together, studies in mammals indicate the ability for molecules to transport from the parenchyma or CSF-filled ventricles into general circulation either through the lymphatics, or blood vessels directly. However, the physiological impact of brain drainage on the periphery had not been investigated, nor directly visualized. Our study demonstrates that such drainage can be directly tracked in vivo to induce a robust hepatic response defined by an active recruitment of inflammatory macrophages and neutrophils into the liver.

Similarly, we found that brain tectum microinjection in zebrafish leads to effects comparable to that known of mammalian intracerebroventricular, intracisternal, and intraparenchymal injections (Glascock et al., 2011; Iliff et al., 2012; Ma et al., 2017), such that the injected LPS in the parenchyma rapidly flows into the CSF compartment (or ventricles) likely via bulk flow (Figure 2 and Video 7), followed by clearance of LPS and ISF predominantly by lymphatic drainage (Figure 2, Figure 2—figure supplements 1 and 2). After brain-LPS injection, we found a low-level spread of LPS/CSF from the hindbrain ventricle into the brain parenchyma and surrounding interstitial space prior to LPS accumulation in brain lymphatic cells and passage through the extracranial lymphatic vessels. Whether CSF flows to the brain more widely than the major perivascular spaces upon intracranial injection remains controversial in mice (Iliff et al., 2012; Ma et al., 2017), although this appears to occur easily in the larval zebrafish perhaps due to its small brain size conducive to simple diffusion or bulk flow of fluids. Previous work in mice have shown that injected tracers in the subarachnoid CSF or lateral ventricle can enter the brain parenchyma quickly in less than 30 min, and get cleared through paravascular or perineural pathways via the lymphatic system similarly to intraparenchymal injections (Iliff et al., 2012; Ma et al., 2017), indicating the process of drainage from the brain parenchyma may be conserved between zebrafish and mammals. On average, at about 13 min after tectum microinjection of fluorescent LPS in larval zebrafish, we begin to detect the LPS in circulation by fluorescence, albeit at a baseline intensity, far less than that at the injection site and in the ventricles. This is about two times faster than the transit time for a lateral ventricle injection in mice with a fluorescent tracer to reach blood circulation (around 25 min) (Ma et al., 2017), which is still relatively fast. Two key traits that may endow larval zebrafish a faster rate of drainage of solutes from brain ventricle or parenchyma than in rodents may be: 1) the short physical distances between anywhere in the parenchyma and the nearest ventricle and vasculature, which would span no farther than the width of the brain, approximately a few hundred microns, enabling simple diffusion as a primary mode of transport (Hladky and Barrand, 2018, and 2) the lack of lymph nodes, based on the prevailing understanding of the zebrafish anatomy (Yaniv et al., 2006; Küchler et al., 2006), which are lymphatic structures throughout the lymphatic network in mammals that capture and filter the fluid that exits the CNS before joining circulation. The content and concentration of the molecules leaving the CNS that reach blood circulation may be limited by lymph nodes in mammals, but these restrictions may be more lenient in zebrafish due to a possible lack of these structures. Instead of lymph nodes, we found that the intraparenchymally-injected LPS in larval zebrafish after spreading in the brain parenchyma and interstitial space is engulfed by brain lymphatic endothelial cells (blec), which are non-lumenized (Venero Galanternik et al., 2017; van Lessen et al., 2017; Shibata-Germanos et al., 2020), as well as collected by the facial and trunk lymphatic vessels (Figure 2—figure supplements 1 and 2). These non-lumenized blec also exist in mouse (Shibata-Germanos et al., 2020), but their contribution to the clearance of solutes from the CSF/ISF is yet unclear. Interestingly, while facial and trunk lymphatic vessels surrounding the brain were filled with LPS by 1.5 hpi, localization of LPS within the brain blood vessels was not observed (Figure 2—figure supplements 1 and 2), indicating the main route by which LPS exit the CNS to enter general circulation was via the lymphatic tracts. Additionally, in both zebrafish and mammals, it is also possible that macromolecules to some degree get eliminated by ingestion by perivascular and other phagocytic cells before they drain out to the periphery through the lymphatic-vascular system.

Whether the process of drainage differs during developmental stages when the tissue boundaries and structures of the brain are less mature, we provide evidence that the peripheral hepatic response to brain-LPS injection may be age-independent in zebrafish. The hepatic response was observed from early to late larval stages, and also in juvenile adults, suggesting passage of inflammatory molecules occurs even after establishment of a fully mature brain architecture (including the blood-brain and blood-CSF barriers). This study focused on the early larval stage at four dpf as it offers a timepoint at which infiltrating macrophage cells can be clearly assessed in the liver without the presence of liver-resident macrophages (Kupffer cells). At this age, they are also functionally mature with established functional blood-brain barriers and ventricles with choroid plexus and CSF (Jeong et al., 2008; Henson et al., 2014; Fame et al., 2016). Regardless of the route or quantities at which macromolecules can drain out from the brain, our results implicate the process of drainage from brain as an effective mechanism by which the brain can quickly transmit molecular information to the periphery to engage a prompt hepatic response to potential CNS threats.

Our data demonstrate that the most striking effect of LPS injection into the brain parenchyma was the recruitment of macrophages into the liver in the zebrafish model (Figure 6). In agreement with the liver being most susceptible to systemic inflammation, previous studies using intravenous injection of the endotoxin LPS in different mammalian models have shown an immediate localization of LPS to the liver and less to other organs by tracking radiolabeled LPS (Bikhazi et al., 2001; Mathison and Ulevitch, 1979; Braude et al., 1955; Herring et al., 1963). Prior to our work, it was not known that introduction of endotoxin into the brain parenchyma can result in the same impact on the liver as that from a systemic administration. These findings beg the question as to how the liver is the predominant and conserved target of circulating LPS and systemic inflammation in vertebrates from zebrafish to mammals. One overarching explanation could simply be related to the sheer blood volume coming from the hepatic artery and portal vein (Eipel et al., 2010) that the liver constantly processes, making it particularly sensitive to inflammatory cues in circulation.

Kupffer cells have been considered to be key immune cells responsible for producing cytokines and chemokines to recruit neutrophils and other leukocytes that may further propagate systemic inflammation (Bilzer et al., 2006; Dixon et al., 2013). In fact, tracking of radiolabeled LPS has shown that circulating LPS was mostly taken up by Kupffer cells (Bikhazi et al., 2001; Mathison and Ulevitch, 1979), implicating these cells as the main sink for endotoxins or microbes. By contrast, our study shows that the liver is capable of responding to systemic inflammation and recruiting neutrophils first and then macrophages, independent of Kupffer cells. We found that starting in the larval zebrafish at four dpf when the liver is already functional and differentiated with diverse hepatic cell types including hepatocytes, sinusoidal endothelial cells, biliary cells, and stellate cells (Yin et al., 2012), but before Kupffer cells are established, other hepatic cells besides the Kupffer cells are capable of signaling to recruit peripheral leukocytes into the liver. Furthermore, while the gut may provide a source of stimulating molecules through the portal vein, we show that the brain is another source of inflammatory cues that can result in immune infiltration of the liver that was not previously appreciated.

The receptor tyrosine kinase CSF1R (also known as Fms) pathway regulates migration, differentiation, proliferation, and survival at varying degrees of different tissue-resident macrophages and monocytes in mammals and zebrafish (Oosterhof et al., 2018; Kuil et al., 2019). Recent studies have implicated a role for this pathway in macrophage recruitment in disease, including the assembly of tumor-associated macrophages in various cancers (Noyori et al., 2019; Lin et al., 2019; Zins et al., 2018; Baghdadi et al., 2016; Baghdadi et al., 2018). Of particular interest IL-34 is elevated and implicated in the inflammation process of several inflammatory diseases including rheumatoid arthritis, inflammatory bowel disease, and Sjogren’s syndrome (Ciccia et al., 2013; Hwang et al., 2012; Moon et al., 2013; Chemel et al., 2012; Boulakirba et al., 2018). Interestingly, among the three known ligands of the zebrafish homolog of the receptor tyrosine kinase CSF1R, namely csf1a, csf1b, and il-34, we found that il-34 was singly upregulated in the liver upon brain-LPS microinjection. Moreover, we showed that disruption of il-34, either by antisense splice-blocking morpholinos or a stable loss-of-function mutation, significantly eliminated macrophage infiltration of liver induced by brain-LPS microinjection (Figure 3). The requirement for il-34 signaling presumably from the liver may be coupled with myd88 known to act downstream of TLR signaling upon LPS recognition, as myd88 was found to be highly expressed in the normal larval zebrafish liver (Koch et al., 2018). Whether these pathways act only in the liver, other cell types, or both, and whether other signaling mechanisms are involved, including CXCR3-CXCL11 known to affect macrophage chemotaxis to infection sites in human and zebrafish (Torraca et al., 2015), remain to be explored. Our data indicate an essential role for il-34 in recruitment of macrophages into the liver, implicating a new function for the il-34/csf1r pathway in conjunction with myd88 dependent mechanisms during hepatic inflammation.

Chronic active infiltration of leukocytes into the liver can cause liver damage and subsequent progression to fibrosis, cirrhosis or liver cancer (Karlmark et al., 2008; Mossanen and Tacke, 2013; Huang et al., 2016). This is a common feature shared among liver diseases caused by infection, toxic insults, or autoimmunity (Edwards and Wanless, 2013). Using zebrafish, we found that macrophages and neutrophils were co-dependent in the process of infiltrating the liver after brain microinjection of LPS. We further showed by in vivo time-lapse imaging an early recruitment of neutrophils in the first few hours was followed by migration or circulation of peripheral macrophages into the liver. This is consistent with previous studies showing early neutrophil recruitment followed by monocyte-derived macrophages in other contexts of inflammation (Soehnlein et al., 2009; Jenne et al., 2018; Kim and Luster, 2015). While secreted granule proteins from neutrophils that have already infiltrated the liver may directly recruit macrophages (Gregory et al., 2002; Soehnlein et al., 2008) in our experimental platform of liver infiltration, we cannot exclude the possibility that indirect effects by which neutrophils alter vascular permeability or signaling from endothelial or other hepatic cell types actually direct macrophage recruitment into the liver.

Conversely, the impact of monocytes and macrophages on neutrophil activity remains less understood. Using two complementary approaches to deplete macrophages either by irf8 genetic deficiency or liposomal clodronate treatment in zebrafish, we found that macrophages may be important to limit neutrophil infiltration into the liver upon brain drainage of LPS. An excessive level of neutrophil infiltration may lead to a heightened release of toxic metabolic products and proteolytic enzymes (Prame Kumar et al., 2018), thereby causing more damage to the liver. Macrophages in the liver are known to remove apoptotic or impaired neutrophils by phagocytosis, which in turn can modulate their own signaling based on the receptor(s) used during neutrophil clearance (Gregory et al., 2002). Since deficiency in irf8 causes an elimination of macrophages but also an increase in baseline neutrophil number (Shiau et al., 2015), we cannot completely rule out that a larger neutrophil pool may contribute to the significant increase in neutrophil infiltration after brain-LPS injection in irf8 morpholino-injected animals. Nonetheless, the agreement in effect from two distinct methods of macrophage ablation, and the observed intimate macrophage-neutrophil contacts in the hepatic region strongly support reciprocal interactions between macrophages and neutrophils.

Our study shows that the drainage following microinjection of LPS into the brain robustly caused peripheral macrophages and monocytes to infiltrate the liver, a process mediated by an upregulation of the cytokine IL-34 in the liver presumably downstream of activating the Toll-like receptor adaptor protein MyD88, and signaling from neutrophils. Treatment with a subset of anti-inflammatory drugs indicates that glucocorticoid activity or inhibiting NF-kB activation suppresses immune cell infiltration during hepatic response to systemic inflammation. These results prompt intriguing questions on how a hepatic response to a central disruption may reflect the presence and severity of CNS inflammation; whether a hepatic response contributes to the recovery or disruption of brain homeostasis; and to what extent does drainage of harmful or inflammatory molecules from brain to circulation explains for the peripheral problems associated with primarily CNS pathologies. LPS were used as an effective tool for brain-localized microinjection in this study to allow tracing of inflammatory molecules originating in the brain to assess peripheral consequences, but whether endogenous proteins and other molecules could be secreted by the brain and drained into circulation to cause similar effects as the LPS do on the liver remains to be investigated. Some examples that lend support to this possibility relate to hallmarks of Alzheimer’s disease that has a known association with liver dysfunction, such as brain tau secretion and its presence in the CSF as well as clearance of toxic accumulations of brain amyloid-beta (Pernègre et al., 2019; Bacyinski et al., 2017; Nho et al., 2019); however the link between the brain secretion or drainage and liver disruption is yet unclear. Understanding mechanisms regulating immune infiltration of the liver and brain-periphery interactions can offer new approaches for modulating or detecting central inflammation while limiting damage elsewhere in the body.

Embryos from wild-type (TL and AB), mutant and transgenic backgrounds: il34^re03 (Kuil et al., 2019), myd88^b1354 (Burns et al., 2017), saa^rdu60 (Murdoch et al., 2019), pu.1/spi1b^fh509 (Roh-Johnson et al., 2017), mpeg1:EGFP (Ellett et al., 2011), lyz:GFP (Hall et al., 2007), lyz:mCherry; cmlc2:GFP (Meireles et al., 2014), fabp10a:DsRed (Dong et al., 2007), kdrl:mCherry-CAAX (Fujita et al., 2011), mrc1a:egfp^y251 (Jung et al., 2017), mfap4:tdTomato (Walton et al., 2015), and nbt:DsRed (Peri and Nüsslein-Volhard, 2008) were raised at 28.5°C and staged as described (Kimmel et al., 1995). Stable transgenic mpeg1:BFP and fabp10a:BFP fish lines were generated using Tol2-mediated transgenesis based on cloned constructs combining the BFP coding sequencing (gift from Martin Distel) downstream of their respective regulatory sequences as published for mpeg1(1.86 kb) (Ellett et al., 2011) and fabp10a(2.8 kb) (Murdoch et al., 2019). 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).

Zebrafish larvae at 4–10 dpf and 1 month old were mounted dorsal side up for brain injections or on their sides for intravenous injections in 1–3% low-melting agarose. A pneumatic microinjector (WPI) with a fine capillary glass pipette was used to inject 1 nL for 4–10 dpf or 2 nL for 1 month old of stimulus into the targeted site (either tectum for brain injections, or caudal vein plexus for intravenous injections). Lipopolysaccharides (LPS) derived from 0111:B4 E. coli (L3024 Sigma) at 5 ng/nL, ultra-pure water, or live Escherichia coli cells were supplemented with fluorescently labeled dextran (Invitrogen, 10,000 MW at 1:100 dilution of a 5 ng/nL stock) for visualization. E. coli cells were prepared for injections as previously described (Earley et al., 2018). Fluorescently tagged Alexa 594 conjugated LPS from 055:B5 E. coli (L23353 Sigma) at 5 ng/nL was used for brain microinjection at 1 nL per fish. All fish after microinjection are carefully monitored for normal health and behavior, and nearly all injected fish remain healthy and viable, showing no signs of overt change. These healthy post-injection fish are used for further experimentation and analysis.

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. Dissected juvenile and adult liver tissues were mounted in fluoromount-G (Southern Biotech) for imaging.

At larval stages 4–10 dpf, the whole liver in the live transgenic larvae was captured in a z-stack that was used for counting total number of macrophages and neutrophils in the liver. Transgenic reporters labeling both the immune cells and hepatocytes were used to count cells through the optical sections. At the juvenile and adult stages, cell counts were made on whole liver dissected and imaged ex vivo using a maximum intensity projection of the z-stack using the ImageJ cell counter tool. Two representative z-stacks were taken for each liver counted.

Transgenic larvae were injected at three dpf intravenously with 1 nL clodronate liposomes (Liposoma) supplemented with Alexa 568 conjugated dextran (10 kDa, Invitrogen) used at 1:100 for visualization of injection. Larvae were incubated for two days to allow macrophage depletion to occur before they were subjected to brain microinjection with LPS or water vehicle at five dpf. Clodronate depletion of macrophages in the brain (microglia) was confirmed in a subset of larvae using the neutral red staining assay. After brain injections, analysis of neutrophil numbers in the liver was conducted in live transgenic zebrafish imaging at 3.5 hpi.

RNA in situ was performed using standard methods. Antisense riboprobes were synthesized from plasmids as described (Shiau et al., 2013) encoding mfap4, mpx (Open Biosystems clone 6960294), and a 739 bp coding fragment of irg1 (NM_001126456.1; pCES161) cloned from a cDNA library derived from a four dpf E. coli injected larva using primers Forward-5’- TCGTTCTGCCAGTAGAGATGTTA-3’ and Reverse-5’- GCGAGCTGAGATGCCTCTAAAC-3’.

RNA was isolated following the RNAqueous-Micro kit RNA Isolation Procedure (Ambion). Whole larvae or dissected livers and remaining body were lysed in 100–300 uL RNA lysis buffer. Larval liver dissections were performed on transgenic larvae Tg(fabp10a:DsRed) to aid in identifying the liver. cDNA was made from 150 or 200 ng of total RNA using oligo (dT) primer with SuperScript IV reverse transcriptase (Invitrogen) for qPCR or RT-PCR analysis. 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 and RT-PCR analysis are listed in Supplementary file 1.

Antisense morpholino oligos were purchased from Gene Tools and re-suspended in water to make 1 mM or 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 injecting into single-cell embryos at 0.5–1 nL.

Microglia were scored in live larvae by neutral red vital dye staining as previously described (Shiau et al., 2015; Shiau et al., 2013). In brief, 3–4 dpf larvae were stained with neutral red 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 2–3 hr later using a stereomicroscope.

The target genes were il-34 (NCBI accession: NM_001128701.1; Gene ID: 560193) and csf3r (NCBI accession: NM_001113377.1; Gene ID: 100134935). Co-injection of Cas9 mRNA and guide RNAs (gRNAs) was conducted in wild-type 1-cell stage zebrafish embryos. Cas9 mRNA was transcribed from XbaI linearalized pT3TS-nCas9n plasmid (Addgene #46757) using mMessage mMachine T3 Kit (Ambion) according to the manufacturer’s instructions. CRISPR targets for gRNA designs were identified using CHOPCHOP (http://chopchop.cbu.uib.no) (Gagnon et al., 2014). Gene-specific oligonucleotides using T7 promoter were used to make gRNAs as previously described (Gagnon et al., 2014). gRNA target sequences and genotyping primers are provided in Supplementary file 1. In vitro transcription of gRNAs from assembled oligonucleotides was conducted using the HiScribe T7 Quick High Yield RNA Synthesis Kit (NEB). To ensure high mutagenesis rate and large deletion mutations, three gRNAs were simultaneously injected with Cas9 mRNA for each gene. Injected clutches of embryos were validated to contain CRISPR mediated mutagenesis by a T7 endonuclease assay. Mutations were analyzed by TOPO TA cloning followed by Sanger sequencing.

Administration of different small-molecule chemicals, DMSO control, or no treatment were performed in parallel starting at three dpf through the time of brain microinjection at four dpf in clean multi-well dishes. Analysis of the effect on macrophages was performed at eight hpi either by RNA in situ hybridization or by live confocal imaging using the liver and macrophage transgenes (fabp10a:DsRed and mpeg1:GFP). 17-DMAG (5 µM) and dexamethasone (6.5 µM) were reconstituted in water and these treatments were compared to the water controls. GW2580 (25 µM), Celastrol (0.22 µM) and Bay 11–7082 (1 µM) were resuspended in DMSO so these treatments were compared to the DMSO-treated controls. List of small molecules is detailed in Supplementary file 1.

Whole liver was imaged 48 hr after LPS or water injection at four dpf in the brain tectum on a confocal Nikon A1R+ using an apochromat lambda 40x water immersion objective (NA 1.15). Four dpf larvae for the brain injections were derived from either uninjected, pu.1-MO injected, or negative control p53-MO injected wild-type embryos. Z-steps were taken at 1 µm thickness. Surface and 3D rendering of the z-stack to measure the liver volume was conducted using the Imaris 3D/4D Image Analysis Software.

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 pair-wise tests to determine the pair(s) showing significant differences. GraphPad Prism eight was used to run statistical tests and create graphs. Scatter bar plots show symbols representing biological replicates.