Sequence searches were performed using BLAST on relevant online genetic databases, with Genomescan (Massachusetts Institute of Technology, Cambridge, MA) used to predict protein coding sequences from genomic DNA (Yeh et al., 2001). Sequence analysis, manipulation and assembly were carried out using Sequencher version 4.10.0 (Gene Codes). ClustalX 2.1 (Jeanmougin et al., 1998) was used to generate sequence alignments, from which phylogenetic trees were generated using the Neighbor-Joining algorithm (Saitou and Nei, 1987) with replicates of 1000 and viewed with NJ plot (Perriere and Gouy, 1996) and Treeview 1.6.6 (Page, 1996). Synteny analysis was performed using Ensembl.

Wild-type and Tg(mpeg1.1::GFP) (Ellett et al., 2011) zebrafish were maintained using standard husbandry practices (Lawrence, 2007). This included feeding thrice daily with a mixture of live feed (artemia and rotifers) and a dry granulated foodstuff (Otohime Hirame Japan). Embryos were obtained from spawning tanks, and in some cases were injected with either control morpholino (5’-CCTCTTACCTCAGTTACAATTTATA) or anti-sense bcl6aa morpholino targeting the intron 2/exon 3 boundary (5’-AGAGCCCACTGTGGAGAAATTATGA) at 0.5 mM. All experiments were approved by the Deakin University Animal Welfare Committee.

The zebrafish bcl6aa gene was targeted using genome editing with CRISPR/Cas9. Embryos were injected with guide RNA (gRNA), designed to a region of exon 3 encoding the BTB/POZ domain using the zifit protocol (Hwang et al., 2013) with the primers 5’-TAGGTCCAGACTGATGGCGTTC and 5’-AAACGAACGCCATCAGTCTGGA, along with Cas9-encoding mRNA and raised to adulthood. Founders were identified with high-resolution melt (HRM) analysis of PCR products with Precision Melt Suremix and Analysis Software (BioRad) (Garritano et al., 2009) using primers spanning the targeted region (5’-CACAGTGGGCTCTTCTACTCTATC and 5’-GGATTGCGAAACCCTCTGG). These fish were outcrossed two times to wild-type fish to remove off-target mutations before in-crossing. Sequence analysis was performed with primers 5’-GCGACCTAAAAAGTTGACTAAAATC and 5’-CCTGGACTTTATGAATCTGTGGC to identify a bcl6aa mutant allele (mdu21), which was also crossed onto the Tg(mpeg1.1::GFP) background.

Embryos were dechorionated and fixed in 4% (w/v) paraformaldehyde (PFA) at 4°C prior to WISH with DIG-labeled anti-sense probes, as described (Thisse and Thisse, 2008). Imaging was performed using Olympus MVX10 fluorescence microscope and DP72 camera using Cellsens Dimension 1.6 software, with ImageJ used for quantitation, as required (Abramoff et al., 2004).

Total RNA was extracted from whole embryos or juvenile zebrafish using an RNeasy Mini Kit (Qiagen) according to the manufacturer’s protocol for animal tissues. This was subjected to quantitative real-time reverse-transcription PCR with immune cell gene specific primers (cd4, cd8, cd79a, ighm, mpeg1.1, mpo, nklb, nkld and tcr) (Sertori et al., 2022) along with ccr2 (5’-TGGCAACGCAAAGGCTTTCAGTGA; 5’-TCAGCTAGGGCTAGGTTGAAGAG), cxcr4b (5’-CCCATCACAAGCACCACAAG; CGATAGCATCATTTTAGACAACAG), il1b (5’-GGACTTCGCAGCACAAAATG; 5’-GTTCACTTCACGCTCTTGGATG), tgfb1 (5’-AAATAGCAGGTTTGTCCCGC; 5’-CACTTCCAGCCCAGGTCTT) and tnfa (5’-GACTGAGGAACAAGTGCTTATGAG; 5’-TGCCCAGTCTGTCTCCTTCTC). Data were normalized to β-actin (actb) and fold change calculated using the ΔΔCt method (Livak and Schmittgen, 2001).

Wounding assays were performed on 3 dpf embryos (n>20 mixed progeny) by excising the end of the caudal tail fin with a scalpel after anesthesia with 0.1 mg/mL benzocaine (Hall et al., 2007) in a conservative manner as described (Meier et al., 2022), with the number of migrating cells and the number of embryos with migrating cells counted up to 8 h after wounding using fluorescence microscopy.

Embryos at 4 dpf were injected with 2-5 nl ~ 5×10⁹ CFU/mL E. coli expressing GFP (#25922GFP, ATCC) into the venous return, with bacteria visualized by fluorescence microscopy, as described (Basheer et al., 2020).

Statistical analyses were performed using Graph Pad Prism (Version 8) software. To determine the statistical significance of various treatments, the unpaired independent student’s t test was employed, with Welch’s correction, where appropriate.

Bioinformatic analysis identified putative zebrafish homologues for both the BCL6A and BCL6B genes, as well as an additional related sequence, with all three genes also being present in another teleost fish, torafugu (Takifugu rubripes). One of these showed conserved synteny with human and mouse BCL6A and their adjacent genes LPP, TPRG1 and TP63 ( Figure 1A ), with the encoded proteins forming a clade with mammalian BCL6A ( Figure 1B ), and so was designated bcl6aa. The fish-specific gene showed conserved synteny across fish genomes, but not with bcl6aa or bcl6b genes ( Figure 1A ), but the encoded proteins formed a larger clade with the BCL6A sequences ( Figure 1B ), and so was named bcl6ab. The final gene showed conserved synteny with human and mouse BCL6B and their adjacent SLC16A13, ACADVL and DVL2 genes ( Figure 1A ), with the encoded fish and mammalian proteins divergent from the other BCL6 proteins ( Figure 1B ), and was designated bcl6b. Collectively, this suggests zebrafish bcl6aa and bcl6b are functional orthologues of mammalian BCL6A and BCL6B, respectively, while bcl6ab represents a fish-specific duplicate of the BCL6A gene.

Alignment of the human BCL6A and zebrafish Bcl6aa proteins confirmed the presence of conserved BTB/POZ, PEST and zinc finger domains, which showed 77%, 35% and 96% identity, respectively ( Figure 1C ). Notably, the latter domain included a stretch of 126 identical amino acids that encompassed the last four of the six C2H2-type zinc fingers. Comparison of the genomic and mRNA ( Figure 1D ) revealed a strongly conserved splicing pattern between human BCL6A and zebrafish bcl6aa genes across the coding exons, with both also possessing a non-coding exon(s) in the proximal promoter region.

The embryonic expression pattern of zebrafish bcl6aa was investigated by high resolution whole-mount hybridization (WISH) on staged wild-type embryos using an anti-sense bcl6aa probe. Expression was observed from 10 hours post-fertilization (hpf) in the anterior lateral mesoderm (ALM) and the posterior lateral mesoderm (PLM), sites of early myeloid cell development (Bertrand and Traver, 2009), which continued until 24 hpf ( Figures 2A, C–F ). From 36 hpf, bcl6aa was expressed in the retina, cerebellum and medulla ( Figures 2G, H ) that continued until 7 dpf although declining after 4 dpf ( Figures 2I–N ). From 4 dpf bcl6aa expression was also detected in the developing thymus ( Figures 2J–N ), which houses T cell development (Gore et al., 2018). No staining was observed with a control sense bcl6aa probe ( Figure 2B and data not shown).

The zebrafish bcl6aa gene was mutated using genome editing with CRISPR/Cas9 to target a region of exon 3 encoding the BTB/POZ domain ( Supplementary Figures 1A, B ). Potential founders were identified with high-resolution melt analysis of PCR products spanning the targeted region, with these outcrossed two times to wild-type fish to remove potential off-target mutations before in-crossing. Sequence analysis identified a bcl6aa allele (mdu21) that harbored a combined large deletion and insertion, predicted to encode a protein that shared just the first 70 amino acids with the wild-type protein, and then encodes 27 amino acids of unrelated sequence before a stop codon is reached ( Supplementary Figure 1C ). Since this represents only part of the BTB/POZ domain and none the PEST or zinc finger domains, it is anticipated that the encoded mutant protein would be non-functional.

The progeny of bcl6aa^wt/mdu21 in-crosses were imaged by light microscopy, with no evidence of overt developmental perturbation during embryogenesis observed in mixed groups, which should contain 25% bcl6aa^mdu21/mdu21 embryos ( Figures 3A–C ), or in individually genotyped embryos ( Figure 3D and data not shown). However, this changed dramatically during the juvenile phase, such that by 21 dpf there were large and distinct differences in size that were in roughly Mendelian ratios. Genotyping of individual fish confirmed homozygote bcl6aa^mdu21/mdu21 mutants were the smallest, with bcl6aa^wt/mdu21 heterozygotes intermediate in size compared to the larger wild-type bcl6aa^wt/wt individuals ( Figure 3E, F ). Additionally, the bcl6aa^mdu21/mdu21 mutants showed a clearly under-developed dorsal fin, abdominal fin, tail fin, swim bladder and eye, with the bcl6aa^wt/mdu21 mutants again showing an intermediate phenotype.

It was also noted that the proportion of smaller fish decreased over time, with none surviving to adulthood. Genotyping of adult fish confirmed that no bcl6aa^mdu21/mdu21 fish were present (data not shown). Close analysis of the relative proportion of bcl6aa^mdu21/mdu21 fish at by genotyping across multiple timepoints revealed that while present at an expected Mendelian ratio at 7 dpf, this steadily decreased, with none observed at 60 dpf ( Figure 3G ). The proportion of bcl6aa^wt/wt fish also increased to above the expected Mendelian ratio, indicating that heterozygote fish also had a milder survival defect, although a good proportion survived to adulthood and showed robust fecundity.

The effect of bcl6aa ablation on immune cells was investigated by WISH analysis with specific markers during embryogenesis, before any growth or survival defects were present. Homozygote bcl6aa^mdu21/mdu21 embryos showed a significant decrease in expression of ikzf1, a marker of T cell progenitors in the developing thymus (Willett et al., 2001), compared to bcl6aa^wt/wt and bcl6aa^wt/mdu21 siblings at both 3.5 dpf ( Figures 4A, B ) and 5 dpf ( Figures 4C, D ). Expression of rag1, a marker of mature T cells (Willett et al., 1997) was significantly decreased in both bcl6aa^wt/mdu21 and bcl6aa^mdu21/mdu21 embryos compared to bcl6aa^wt/wt siblings at both 3.5 dpf ( Figures 4E, F ) and 5 dpf ( Figures 4G, H ), but to a much greater extent in bcl6aa^mdu21/mdu21 embryos across both timepoints ( Figures 4E–H ). In contrast, no significant difference was observed in the number of cells expressing mpo, a marker of neutrophils (Lieschke et al., 2001), in either bcl6aa^wt/mdu21 or bcl6aa^mdu21/mdu21 compared to bcl6aa^wt/wt embryos at 5 dpf ( Figures 4I, J ). However, the number of cells expressing lcp1, a marker of leukocytes including macrophages (Bennett et al., 2001), was significantly decreased in both bcl6aa^wt/mdu21 and bcl6aa^mdu21/mdu21 compared to bcl6aa^wt/wt embryos, although again the quantity in bcl6aa^mdu21/mdu21 embryos was also significantly reduced compared to heterozygotes ( Figures 4K, L ). To facilitate further analysis of macrophages the bcl6aa^mdu21 allele was crossed onto the Tg(mpeg1.1:GFP) background, in which macrophages are marked with GFP (Wittamer et al., 2011). Tg(mpeg1.1:GFP) bcl6aa^wt/mdu21 fish were in-crossed and visualized by fluorescence microscopy that revealed a significant decrease in GFP⁺ cells at 4 dpf in bcl6aa^mdu21/mdu21 compared to the bcl6aa^wt/wt and bcl6aa^wt/mdu21 embryos ( Figures 4M, N ). To confirm the effects of bcl6aa on macrophages, embryos were injected with an anti-sense morpholino targeting the intron 2/exon 3 splice site. This also resulted in a decrease of lcp1+ cells at 22 hpf in wild-type embryos ( Figures 4O, P ) and in GFP+ cells in Tg(mpeg1.1:GFP) embryos at 3 dpf ( Figures 4Q, S ) in comparison to those injected with a control morpholino. Macrophage morphology was also altered in the bcl6aa morpholino-injected embryos with a statistically significant decrease in those with an amoeboid morphology ( Figure 4R ).

To further understand the effect of bcl6aa ablation on macrophages, a wounding assay was performed on the progeny of Tg(mpeg1.1:GFP) bcl6aa^wt/mdu21 fish as described (Hall et al., 2007), with individual embryos subsequently imaged at various times, after which they were genotyped. This revealed that in bcl6aa^wt/wt embryos GFP+ cells peaked at the wound at 4 hours post wounding (hpw) before decreasing at later time points ( Figures 5A, B ). In both bcl6aa^wt/mdu21 and bcl6aa^mdu21/mdu21 embryos, GFP+ cells peaked at 8 hpw, however, the numbers observed in bcl6aa^wt/mdu21 embryos were equivalent to or exceeded those of bcl6aa^wt/wt embryos, whereas they were significantly reduced in bcl6aa^mdu21/mdu21 embryos. When normalized to total GFP+ cells, the number of GFP+ cells remained reduced to a statistically significant level at both 4 and 24 hpw (data not shown). The experiment was repeated but examining embryos at 0.25 h intervals until the first GFP+ cell reached the wound site, after which embryos were genotyped. For almost all bcl6aa^wt/wt and bcl6aa^wt/mdu21 embryos this occurred by 1 hpw, but for bcl6aa^mdu21/mdu21 embryos this was around 3 hpw ( Figure 5C ).

The bcl6aa mutants were next investigated in a bacterial infection assay (Basheer et al., 2020). The progeny of bcl6aa^wt/mdu21 in-crosses were subjected to injection with GFP+ E. coli at 4 dpf. Surviving embryos were imaged by fluorescent microscopy until 24 hour post infection (hpi) and subsequently genotyped, with the fluorescence intensity used as an indicator of bacterial load. The bcl6aa^wt/wt , bcl6aa^wt/mdu21 and bcl6aa^mdu21/mdu21 embryos showed no difference in fluorescence at 0.5 hpi ( Figures 6A, B ). However, at 12 and 18 hpi the fluorescence intensity was increased in bcl6aa^mdu21/mdu21 compared to bcl6aa^wt/wt and bcl6aa^wt/mdu21 embryos. No difference in mortality was observed in the injected embryos at 0.5 hpi, but from 12 hpi decreased survival of bcl6aa^mdu21/mdu21 embryos was observed, reaching zero survival at 24 hpi, while bcl6aa^wt/wt and bcl6aa^wt/mdu21 showed similar high survival rate ( Figure 6C ) and fluorescence intensity ( Figure 6B and data not shown). Analysis of a set of inflammatory genes indicated increased basal expression of ccr2 in bcl6aa^mdu21/mdu21 compared to bcl6aa^wt/wt embryos ( Figure 6D ). Infection resulted in increased expression of il1b and tnfb1b in both bcl6aa^wt/wt and bcl6aa^mdu21/mdu21 embryos (data not shown), but il1b was significantly enhanced in bcl6aa^mdu21/mdu21 compared to bcl6aa^wt/wt embryos ( Figure 6D ).