T Lymphocyte–Dependent and –Independent Regulation of Cxcl8 Expression in Zebrafish Intestines

The potent chemokine CXCL8 is produced by a range of cell types including myeloid and nonmyeloid cells, such as endothelial and epithelial cells (1–4). Epithelial CXCL8 is induced upon microbial stimulation via TLR signaling. CXCL8 signals through its receptors, CXCR1 and CXCR2, and is a strong chemoattractant for neutrophils, basophils, resting T cells, and stimulated eosinophils (4–7). Previously, we reported that T cell inhibition by the NFAT inhibitor Tacrolimus (FK506) diminished CXCL1 and CXCL2 (functional mouse homologs of CXCL8) in the early phases of enterocolitis induction, suggesting a role for adaptive immune cells in the regulation of innate chemokine expression (8). Although much is known about the regulation of adaptive immunity by innate immune cells, less is known about the role of activated intestinal T lymphocytes in regulating epithelial and innate cell responses (9, 10). Recently, Strutt et al. (11) showed that memory CD4⁺ T lymphocytes enhance production of multiple innate inflammatory cytokines and chemokines in the murine lung independently of pathogen. In vitro studies have shown that activated T cell clones derived from human PBMCs were able to induce cytokine production, including CXCL8, in human airway epithelial cells (12). Likewise, T cell–derived IL-17A and IL-17F were capable to either directly or indirectly induce CXCL8 expression in airway and retinal pigment epithelial cells (13, 14). Moreover, evidence exists that human T cells themselves are capable of producing cytokines such as CXCL8 after TLR stimulation (15–17). In vivo research into CXCL8 function and regulation has been hampered by the lack of appropriate animal models. CXCL8 is identified in other mammals but not in mouse and rat (18, 19). However, CXCL8 has been identified in teleost fish, including the zebrafish (20, 21).

Using zebrafish as a model, it was discovered that Cxcl8 establishes tissue-bound gradients in vivo by binding to heparan sulfate proteoglycans. In this way, Cxcl8 establishes a robust surface-bound gradient that recruits and retains leukocytes at sites of infection (21). Recently, a second zebrafish ortholog of human CXCL8 was identified. Both zebrafish orthologs, named cxcl8-l1 and cxcl8-l2, recruit neutrophils and are induced after wound-elicited acute inflammation (22). CXCL8 receptors cxcr1 and cxcr2 are expressed in the zebrafish gut epithelium, and cxcl8-l1 is expressed by myeloid and intestinal epithelial cells (20). Because of its conserved function in zebrafish, we used this animal model to learn more about CXCL8 regulation. In this study, we specifically addressed intestinal cxcl8 regulation by adaptive immunity. We took advantage of the fact that in contrast to mice and human, in which adaptive immunity is already present at birth, zebrafish adaptive immunity does not mature until 3 wk of age (23). This enables analysis of intestinal cxcl8 expression in the absence (<3 wk) and presence (>3 wk) of adaptive immunity. We show that Cxcl8-l1 and Cxcl8-l2 are differentially regulated during homeostasis and inflammatory conditions via both T lymphocyte–dependent and –independent mechanisms. Cxcl8-l1 but not cxcl8-l2 expression is regulated by T lymphocytes during development. In contrast, during intestinal inflammation especially Cxcl8-l1 expression is significantly upregulated in a T lymphocyte–independent manner.

Zebrafish were maintained under standard husbandry conditions at the Hubrecht Institute (14-h light, 10-h dark regimen). Fish are fed twice daily with a combination of flakes and live artemia. The Rag1-deficient zebrafish (rag1^t26683) were previously generated at the Hubrecht Institute and for all analyses, except for the 1 wk postfertilization (wpf) data (larvae had to be pooled before RNA and DNA extraction), clutch mates of Rag1^+/− in-crossed fish were used (24). The lck:GFP zebrafish were purchased from ZIRC (Oregon) (25). Neutrophil reporter Tg(mpx::eGFP)ⁱ¹¹⁴ zebrafish were provided by S. Renshaw (University of Sheffield, Sheffield, U.K.) (26). All animal experiments were approved by the Animal Experimentation Committee (DEC) of University Medical Centre and were carried out following guidelines of the Animal Experimentation Committee and Dutch Law on Animal Experimentation.

Total intestinal tissue was collected in TriPure (Roche), and RNA was extracted by phenol/chloroform extraction. cDNA was synthesized from RNA by iScript reverse transcriptase, according to the manufacturer’s instructions (Bio-Rad). Primers were designed using the oligo 6.22 program and blasted against the European Molecular Biology Laboratory database. Primers and gene information are listed in Table I. Real-time quantitative PCR was performed using SyBr Green amplification (Bio-Rad). The PCR program used: 95°C, 3 min, 40 × (95°C 10 s, 60°C 10 s, and 72°C 30 s), followed by a melting curve consisting of 95°C 30 s, 65°C 5 s, and increase to 95°C in 0.5°C steps. Relative expression was assessed by calculating relative expression compared with β-actin; 2^٨ − (Cq^target − Cq^β-actin).

Upon necropsy, the intestines were removed and flushed with PBS. Next, the intestinal tissue was strained over a 40-μm filter (Greiner). The cell suspension was collected in L-15 medium (Leibovitz; Sigma-Aldrich) supplemented with 0.3 mg/ml glutamine, 50 U/ml penicillin, 0.05 mg/ml streptomycin, 0.8 mM CaCl₂, 10% embryo extract, and 1% carp serum (specific pathogen-free) (adapted from ZFIN, zebrafish handbook [http://zfin.org/zf_info/zfbook/zfbk.html]). Cells were washed twice with L-15 medium with supplements and filtered over 40-μm filters. Samples were measured on the FACSCanto (BD Biosciences), and results were analyzed with FlowJo. For analysis of neutrophils, mpx:eGFP zebrafish were used. The eGFP⁺ population (as seen in Figs. 1D, 6E) are back-gated to ensure these cells appear in the granulocyte gate. This population gate was previously determined in our laboratory by FACS sorting and cytospin analysis. As a positive control, kidney suspensions were made. The kidney is the main site for granulopoiesis in zebrafish, so large amounts of eGFP⁺ cells are present. As a negative control, nontransgenic fish were used. For analysis of lymphocytes, lck:GFP zebrafish are used. The GFP⁺ population as seen in Fig. 2D are back-gated to ensure these cells appear in the lymphocyte gate. This population gate was previously determined by FACS sorting and cytospin analysis in our laboratory. As a negative control, nontransgenic zebrafish were used.

T cell reporter zebrafish (Lck:GFP) were used to isolate GFP⁺ T lymphocytes. Intestines were pooled and strained over a 40-μm filter (Greiner), and the cell suspension was collected in supplemented L-15 medium (as above). Cells were washed twice with medium and strained again over 40-μm filters. Subsequently, T lymphocytes were FACS sorted in PBS on the basis of forward light scatter and side scatter and GFP expression (FACSAria; BD Biosciences). Cells were washed with PBS and were adoptively transferred into 5-wpf-old Rag1-deficient zebrafish and wild-type siblings. A total of 15,000 cells or PBS was injected i.p. under anesthesia with MS222 (Tricaine) in a volume of 10 μl. One week after injection, zebrafish were sacrificed, and intestinal RNA was isolated.

T cell reporter zebrafish (Lck:GFP) were used to isolate and sort GFP⁺ T lymphocytes as described above. After sorting, 100,000 T lymphocytes were seeded per well containing L-15 medium (as described above). After 4 h, cells were collected in TriPure, and the supernatant was harvested and filtered through a 0.2-μm filter. FACS-sorted T and non-T lymphocytes were also checked for B and T cell markers by quantitative PCR (qPCR) (Supplemental Fig. 1).

Intestines of 8-wk-old Rag1-deficient zebrafish and wild-type siblings were removed upon necropsy, were embedded in Matrigel (BD Biosciences), and were immersed in supplemented L-15 medium. Intestines were left in medium alone or exposed to filtered T lymphocyte supernatant at 27°. Six hours after exposure intestines were collected in TriPure for RNA isolation.

Sections of 7 μm were collected on RNase-free slides, dried, and deparaffinized. After PBT (PBS with 0.1% Tween) wash, slides were incubated at room temperature with 5 μg/ml proteinase K. After proteinase K treatment slides were washed with PBT. Slides were prewarmed with a hybridization mix containing 50% formamide, 5× SSC, 0.1% Tween, citric acid (pH 6), heparin, and tRNA (yeast) for 2 h at 60°C. A Cxcl8-l1 probe was constructed using a digoxigenin (DIG) labeling kit (Roche) to hybridize the 5′-untranslated region and region 109789-109923 of clone DKEY-151B16 CT826376.9 (ZDB-GENE-081104-317; cxcl8-l1). Cxcl8-l1 and nonsense DIG probe (Roche) were applied at 200 ng/slide in hybridization buffer and incubated overnight at 60°C. Slides were washed in washing buffer containing 50% formamide, 5× SSC, 0.1% Tween, and citric acid (pH 6). Next slides were blocked with 2% BSA in PBT for 1 h at room temperature and washed with PBT. Next, slides were incubated with 1:1000 anti-DIG Fab-AP (Roche) in blocking buffer for 3 h. After Ab incubation, slides were washed with PBT four times 5 min and transferred to AP⁻ solution containing 10% 1 MTris HCl (pH 9.5), 2% NaCl, and 0.1% Tween for 15 min. Next, slides were transferred to AP⁺ solution (AP⁻ solution + 5% MgCl₂) twice 10 min. Slides were incubated with 5-bromo-4-chloro-3-indolyl phosphate/NBT staining solution (MP Biomedicals) for 25 min. Slides were washed in PBT, transferred to PBT plus 1 mM EDTA, dipped in 100% methanol, and mounted with Polymount (Polysciences) and a coverslip.

Rag1-deficient and wild-type zebrafish were anesthetized with MS-222 (Tricaine; 160 μg/ml). Next, 50% ethanol was administered intrarectally in a volume of 10 μl. This administration leads to mild inflammation in the intestines as described previously (27). Six hours after administration of ethanol, intestines were removed upon necropsy and kept in TriPure (Roche) for RNA isolation.

Wild-type zebrafish (Tg(mpx::eGFP)ⁱ¹¹⁴) or Rag1-deficient zebrafish were anesthetized with MS-222 (Tricaine; Sigma-Aldrich) and intrarectally injected 10 μl of a 1 μg/ml solution of monocyte-derived human recombinant CXCL8 in PBS (PeproTech). Six hours after injection fish were sacrificed, the intestines were removed for FACS analysis or RNA isolation. As a control, human recombinant IL-4 was injected at the same concentration (10 μl of a 1 μg/ml solution; Immunotools).

Upon necropsy, kidneys were removed from Tg(mpx::eGFP)ⁱ¹¹⁴ zebrafish. Kidneys are strained and pooled as described above. Neutrophils were FACS sorted on basis of forward light scatter and side scatter and eGFP expression (FACSAria; BD Biosciences). Migration was assessed by administration of the chemoattractant in the lower well and 5.0 × 10⁵ neutrophils in 500 μl/insert (pore size, 8 μm) (Greiner BioOne). As a positive control, H₂O₂ (0.0003% solution) was used. Migration was assessed by counting the number of neutrophils migrated to the lower well over time.

Data were tested for normal distribution by Kolmogorov–Smirnov test. Mann–Whitney tests or unpaired t tests (normal distribution) were used to calculate statistical significance. Tests used and p values are stated in the figure legends.

To investigate cxcl8 expression over time in zebrafish during innate and adaptive phases of immune development, we performed real-time PCR on RNA isolated from the intestines of zebrafish at 1 wpf (only innate immunity), 5 wpf (development adaptive immunity), and 14 wpf (innate and adaptive immunity). Primer sequences are listed in Table I. We found that cxcl8-l1 expression increased significantly from 1 to 5 wpf of age, whereas at 14 wpf, cxcl8-l1 expression levels were low and comparable to 1 wpf larvae (Fig. 1A). In contrast, cxcl8-l2 expression decreases over time (Fig. 1B). By using mpx:eGFP transgenic neutrophil reporter zebrafish, we observed that intestinal neutrophil percentages increase significantly from 1 to 5 wpf, remaining stable ∼1.0–1.5% of intestinal cells from 5 to 14 wpf (Fig. 1C, 1D).

To investigate whether the increase in cxcl8-l1 expression at 5 wpf under homeostasis depends on adaptive immunity, we analyzed intestinal expression of cxcl8-l1 and cxcl8-l2 in Rag1-deficient zebrafish. The elevated intestinal cxcl8-l1 expression seen in wild-type zebrafish at 5 wpf is absent in Rag1-deficient zebrafish (Fig. 2A). In contrast, no difference in cxcl8-l2 expression is observed between Rag1-deficient zebrafish and wild-type siblings at 1 and 5 wpf. However, at 14 wpf, a significantly higher cxcl8-l2 expression is seen in Rag1-deficient zebrafish compared with wild-type siblings (Fig. 2B). By using the T lymphocyte reporter zebrafish (lck:GFP), we observed lckGFP⁺ T lymphocytes at 5 wpf in the intestines, confirming the presence of adaptive immune cells at that time point (Fig. 2C, 2D). In fact, we already detect lck:GFP⁺ cells in the intestines as early as 3 wpf (Supplemental Fig. 2).

To address whether T and/or B lymphocytes are responsible for the elevated cxcl8-l1 expression at 5 wpf under homeostatic conditions, we performed adoptive transfer experiments. Sorted T lymphocytes from transgenic T lymphocyte reporter (lck:GFP) zebrafish were transferred into Rag1-deficient zebrafish. As a control, the GFP⁻ fraction (non-T lymphocytes) was injected in a second group of fish. Flow cytometric analysis on intestinal cell suspensions 1 wk after transfer confirmed the presence of GFP⁺ cells within the intestines of Rag1-deficient zebrafish (Fig. 3A, 3B). In Rag1-deficient zebrafish that received T lymphocytes, cxcl8-l1 expression was partially restored (Fig. 3C) and correlated with the expression of T cell marker lck (Fig. 3E). In contrast, cxcl8-l2 expression was low, did not increase after adoptive transfer of T lymphocytes, and did not correlate with lck expression (Fig. 3D, 3F). Furthermore, adoptive transfer of the GFP⁻ cells in the lymphocyte gate (non-T lymphoid cells; B and NK-like cells) did not result in an induction of cxcl8-l1 or cxcl8-l2 expression in Rag1-deficient recipients. These data show that adoptive transfer of T lymphocytes increases cxcl8-l1 expression in the intestines of Rag1-deficient zebrafish.

Previously, we and others demonstrated that human CD4⁺ T lymphocytes themselves produce CXCL8 upon activation (8, 16). Therefore, we investigated whether sorted zebrafish T lymphocytes also express cxcl8-l1 and/or cxcl8-l2. Indeed, sorted T lymphocytes are capable of cxcl8-l1 and cxcl8-l2 expression, although expression levels of cxcl8-l2 are very low (Fig. 4A, 4B). Cxcl8-l1 is also expressed in sorted neutrophils and nonneutrophils within the granulocyte population, whereas cxcl8-l2 was not detected in either population (Supplemental Fig. 3). Recently, Strutt et al. (11) showed that murine memory CD4⁺ T lymphocytes can enhance innate cytokines and chemokines in the lung independently of the presence of pathogen. Along the same lines, we investigated whether T lymphocyte–secreted factors can induce non–lymphocyte-derived cxcl8-l1 and/or cxcl8-l2 expression in fish. Exposure of Rag1-deficient intestinal explants of 8 wpf were exposed to supernatant of sorted T lymphocytes of 14-wpf-old lck:GFP fish. At this time point, both intestinal cxcl8-l1 and cxcl8-l2 expression is low providing a window to observe possible upregulation. As shown in Fig. 4C, T lymphocyte supernatant induces cxcl8-l1 expression in the intestines of Rag1-deficient but not wild-type zebrafish. In contrast, expression levels of cxcl8-l2 in intestinal explants of either wild-type and Rag1-deficient zebrafish were very low and did not respond to T lymphocyte supernatant (Fig. 4D).

In situ hybridization for cxcl8-l1 on intestinal paraffin sections revealed that in wild-type zebrafish at 5 wpf staining was seen in immune cells (Fig. 4E, Supplemental Fig. 4) and the epithelial cell lining (n = 2). This latter staining was absent in Rag1-deficient intestines (n = 2) (Fig. 4E, Supplemental Fig. 4). In this study, some immune cells stain positive; however, the epithelial cells are devoid of staining. After adoptive transfer of T lymphocytes into Rag1-deficient zebrafish, the epithelial cells appear to be positive for cxcl8-l1 (n = 3; Fig. 4E, Supplemental Fig. 4). In situ hybridization for cxcl8-l1 on intestines of (adult) 14 wpf Rag1-deficient and wild-type zebrafish mainly localizes to immune cells (Supplemental Fig. 4).

Recently, de Oliveira et al. (22) demonstrated that after tail wounding both Cxcl8-l1 and Cxcl8-l2 were upregulated. To investigate whether this upregulation is also occurring in the intestines under inflammatory conditions, we induced mild intestinal inflammation by intrarectal ethanol administration in 8-wk-old fish. Previously, we have shown that intrarectal administration of 50% ethanol induces mild damage to the epithelium on histology (27). In addition, ethanol administration also induces both il1b and mmp9 expression in wild-type zebrafish (data not shown). We observed that ethanol induced inflammation induces cxcl8-l1 expression 6 h after administration in the intestines of both wild-type and Rag1-deficient zebrafish (Fig. 5A). Expression levels of cxcl8-l2 are low and variable but tend to increase after ethanol injection, albeit not significantly (Fig. 5B).

Recently, Stoll et al. (28) reported that human recombinant CXCL8 could rescue knockdown of zebrafish endothelial cxcl8. Furthermore, it has been shown by Yang et al. (29) that human recombinant CXCL8, when injected in the ear of larval zebrafish, is able to recruit neutrophils but not macrophages. Given this apparent functional homology, we used human recombinant CXCL8 to investigate whether intrarectal injection of human recombinant CXCL8 would be able to induce cxcl8-l1 and or cxcl8-l2. We performed qPCR for cxcl8-l1 and cxcl8-l2 6 h after injection. Cxcl8-l1 expression in the intestines of Rag1-deficient fish was increased compared with the PBS-injected control and human recombinant IL-4–injected zebrafish (Fig. 6A). Expression levels of cxcl8-l2 were low and more variable but also tended to increase upon human recombinant CXCL8 injection (Fig. 6B). To investigate whether human recombinant CXCL8 is able to recruit neutrophils to the intestines of adult fish, we intrarectally injected human recombinant CXCL8 into 8-wk-old mpx:eGFP reporter zebrafish. Six hours after administration of human recombinant CXCL8, the percentage of neutrophils in the intestines was significantly increased compared with PBS-injected or human recombinant IL-4–injected zebrafish (Fig. 6C, 6E) Moreover, sorted neutrophils from transgenic neutrophil reporter fish (mpx:eGFP) are recruited by human recombinant CXCL8, whereas they do not respond to human recombinant IL-4 (Fig. 6D). These experiments indicate that exogenous human CXCL8 can induce also innate cxcl8-1 expression and induce neutrophil recruitment in the intestines of zebrafish.

In this paper, we report that cxcl8-l1 and cxcl8-l2 are differentially regulated by both T lymphocyte–dependent and –independent mechanisms in the zebrafish intestines. Although cxcl8-l1 expression is regulated by T lymphocyte–secreted factors during development, upregulation of cxcl8-l1 expression under inflammatory conditions occurs in a T lymphocyte–independent manner. In this study, we show that in homeostasis Cxcl8-l1 expression levels are specifically increased at 5 wpf, associating with the presence of adaptive immunity in the intestines. Indeed, adoptive transfer of T lymphocytes is able to increase cxcl8-l1, but not cxcl8-l2 expression, in homeostasis. Our in situ hybridization experiment suggests that in Rag1-deficient zebrafish epithelial cxcl8-l1 is absent and can be induced after adoptive transfer of T lymphocytes. These findings indicate that T lymphocytes are able to induce epithelial cxcl8-l1. Our data are in agreement with a recent report that shows that memory (influenza virus specific) CD4⁺ T lymphocytes are capable of enhancing the expression of several cytokines and chemokines in the lung of mice, even in the absence of pathogen (11). The authors conclude that this induction together with other aspects of CD4⁺ T lymphocyte activation results in nonspecific induction of a transient protective state during the initial phase of infection. In this study, we speculate that the increase of (epithelial) cxcl8-l1 expression at 5 wpf induced by appearance of T lymphocytes in the intestines provides additional neutrophil recruitment to this site of high antigenic pressure. This ensures baseline levels of neutrophils in the intestines able to protect the epithelial barrier. What cannot be explained, however, is the fact that cxcl8-l1 expression decreases from 5 wpf to low levels at 14 wpf, whereas the neutrophil percentages do not decrease from 5 to 14 wpf. Cxcl8-l2 expression is increasing over time, which might indicate that both Cxcl8-l1 and Cxcl8-l2 are involved in neutrophil regulation in a time-dependent fashion in homeostasis. This relation between different chemokines and neutrophil recruitment under homeostasic conditions, however, will require further investigation.

The elevated cxcl8-l1 expression seen in Rag1-deficient zebrafish after adoptive transfer might arise in part from the transferred T lymphocytes themselves. However, our in vitro studies suggest that next to T lymphocyte contribution to intestinal cxcl8-l1 expression after adoptive transfer, T lymphocyte–secreted factors can induce cxcl8-l1 expression in Rag1-deficient zebrafish, most likely in epithelial cells. Indeed, in situ hybridization seems to corroborate that epithelial cxcl8-l1 is induced. Furthermore, Oehler et al. (20) have shown that FACS-sorted epithelial cells from transgenic ifabp:GFP zebrafish are able to express cxcl8-l1. Likewise, several studies provide evidence that both T lymphocytes and epithelial cells are capable to express CXCL8 (15, 16). For example, in vitro studies have shown that activated T cell clones derived from human PBMCs were able to induce cytokine production in human airway epithelial cells including CXCL8 (12). Likewise, T cell–derived IL-17A and IL-17F were capable to either directly or indirectly induce CXCL8 expression in airway and retinal pigment epithelial cells (13, 14). Future studies will investigate which cytokines or secreted factors are involved in the induction of chemokine expression in epithelial cells. Furthermore, until now it has not been possible to discriminate different lymphocyte subsets in zebrafish. We anticipate that further advancement in this research area is to be expected in the near future.

Like cxcl8-l1 expression, cxcl8-l2 expression is also dynamic over time. Cxcl8-l2 expression is significantly higher in 14 wpf Rag1-deficient zebrafish compared with wild-type siblings, suggesting T lymphocyte dependency. However, cxcl8-l2 expression is not upregulated after adoptive transfer of T lymphocytes into Rag1-deficient zebrafish. In addition, next to cxcl8-l2, IL1β expression is also elevated in adult Rag1-deficient zebrafish intestines (data not shown). We hypothesize that this upregulation of innate cytokines can be the result of higher antigenic exposure and increased innate immune activation because of the lack of adaptive immunity. Indeed, it has been shown by adoptive transfer experiments in Rag1-deficient mice that T lymphocytes are both necessary and sufficient to temper the early innate response (30). Furthermore, like Rag1-deficient mice, Rag1-deficient zebrafish are more susceptible to infectious disease especially at older age (>14 wpf) (S. Brugman and E. Nieuwenhuis, unpublished observations).

Recently, investigation into the role of zebrafish Cxcl8-l1 and Cxcl8-l2 in neutrophil recruitment after tail wounding revealed that both Cxcl8-l1 and Cxcl8-l2 are upregulated in the wound area (22). In this study, we observe that 6 h after intestinal inflammation especially cxcl8-l1 expression is upregulated in both wild-type and Rag1-deficient zebrafish, indicating that expression of this chemokine under inflammatory conditions does not require T lymphocytes. Intestinal cxcl8-l2 expression is very low and highly variable and is upregulated in some wild-type and Rag1-deficient zebrafish albeit not significantly. Furthermore, intrarectal injection of human recombinant CXCL8 increases cxcl8-l1 and tends to increase cxcl8-l2 expression in Rag1-deficient intestines.

In conclusion, our data indicate that intestinal cxcl8-l1 and cxcl8-l2 expression are differentially regulated in homeostasis and inflammatory conditions via T cell–dependent and T cell–independent mechanisms. Furthermore, as evidenced by the functional homology of human recombinant CXCL8 in also recruiting zebrafish neutrophils to the intestines, zebrafish are an excellent model to study CXCL8 function and regulation in vivo.

We thank Dr. Schulte-Merker at the Hubrecht Institute for providing the necessary infrastructure for the experiments, Dr. Maria Forlenza of Wageningen University for providing specific pathogen-free carp serum, Dr. Renshaw for the mpx:eGFP transgenic line, and Dr. Sabine Middendorp for critically reading the manuscript. In addition, we thank the animal caretakers at the Hubrecht Institute for excellent care of the zebrafish.

The authors have no financial conflicts of interest.