Zebrafish B Cell Development without a Pre–B Cell Stage, Revealed by CD79 Fluorescence Reporter Transgenes

Jawed vertebrates, including the fish, express a number of innate and adaptive immune system receptors, such as TLR and NOD-like receptor for innate immunity, and recombinase activating gene (RAG) and TCR and BCR genes for adaptive immunity, initially found in mice and humans (1). The zebrafish is a bony fish, a teleost, with ancestry that was generated >300 million years ago as one of the early jawed vertebrates. The zebrafish has both an innate as well as an adaptive immune system, and it is thereby regarded as a good model organism for the study of immune responses (2–4). Presence of one of the major cell types in adaptive immunity, the T cells, has been identified in zebrafish and studied by detection of relevant mRNAs and use of a lymphocyte cell–specific protein tyrosine kinase (Lck)–GFP reporter transgenic line (5). It has been established that the thymus is a common primary site for T cell development, as confirmed by examination of Rag1 and Rag2 and TCR gene expression (6, 7). The Rag genes encode proteins necessary for rearrangement of both T and B cell Ag receptor chains (8, 9), and a Rag2:GFP reporter identified the presence of Rag2:GFP⁺ cells in thymus (10, 11).

B cells are the other major adaptive immune cell type. However, the details of the B cell development in zebrafish are still not well understood. In mice, B cells are generated from hematopoietic stem cells that reside in the liver before birth and in the bone marrow of adults (12, 13). Mouse B cell development is a highly orchestrated process, wherein precursors initiate Ig H chain rearrangement at the pro–B stage (14), then assemble the H chain with a surrogate L chain to form a pre-BCR that signals clonal expansion of pre–B stage cells, progression to later stages of development, and initiation of Ig L chain rearrangement (15). Upon successful completion of L chain rearrangement, the BCR is expressed on the surface of newly formed B cells that then undergo further maturation to become fully functional B cells. A similar process has been identified in the generation of B cells in humans (16) and in rabbits (17). However, not all vertebrate species construct B cells in this fashion. For example, chicken B cells are produced by simultaneous rearrangement of Ig H and L chains in the bursa of fabricius, with no distinct pre–B stage (18). VpreB together with λ5 form a surrogate L chain (19), also known as pseudo–L chain, to generate progression through the pre-BCR to pre–B cell stage in mammals. In general, existence of the pseudo–L chain has not been clearly established in nonmammals (20). In zebrafish, neither a pseudo–L chain nor pre-Tα that creates a pre-TCR has been detected (20, 21). Thus, it has not been clear whether the zebrafish generates pre–B and/or pre–T cell stage in development.

We have now investigated B cell development in the zebrafish model organism, seeking to determine similarities and differences from mammalian B cells. The zebrafish is a small fish where embryos develop most organs by 5 d after fertilization, allowing visual tracking of maturation (22). Although Abs to detect and distinguish adaptive immune cell types in zebrafish have not yet been developed, a powerful approach in this model organism has been the generation of fluorescent reporter transgenic fish, an approach that has been used to identify and characterize erythroid, myeloid, T cell, and early lymphoid cells (5, 23, 24). Fluorescent reporter lines have revealed early T cell development in zebrafish thymus (7, 25), and such reporters have been used to extensively study T cells (5). Thus, we sought to assess B cell development by generating CD79 fluorescent reporter transgenic zebrafish.

B cell Ag receptors use heterodimers of the adapter proteins, CD79a (Igα) and CD79b (Igβ), to couple membrane-bound Ig, the external Ig Ag-sensing molecule, with cytoplasmic signaling cascades (26–28). These CD79 proteins are essential for Ig cell surface expression and B cell signal transduction, including the pre-BCR and pre–B cell stage in mammals. Both CD79 peptides contain an ITAM, and this motif is also found in other transmembrane receptor signaling adapter proteins such as CD3 components, TCRζ and DAP12 (29). These CD79 molecules serve as robust B cell markers. Mammalian CD79 expression is first detected at the stage of B lineage commitment, including the pre–B cell stage (30), and continues until the plasma cell stage, when they are downregulated (31). Both CD79a and CD79b have been described in chickens (32, 33), and orthologous genes have been identified in some bony fish species, including fugu (34) and channel catfish (29), as well as in a boneless jawed vertebrate, the spiny dogfish (shark) (35). The presence of CD79a in zebrafish B cells has also been demonstrated by single-cell quantitative PCR (36). However, these CD79 genes have not been characterized in the context of zebrafish B cell development. Thus, to unambiguously identify and characterize B cells and B cell development in zebrafish, we cloned orthologs of mammalian CD79a and CD79b, then generated GFP reporter lines with transgenic constructs containing regulatory sequences for these genes driving fluorescence protein expression. We then used these lines to cross with a transgenic Rag2-mCherry line to assess B cell development and differentiation. In zebrafish B cells, three Ab classes have been identified, IgM, IgD, and IgZ (also named as IgT) (37–40). IgM is the main Ab class in all jawed vertebrates, and IgZ/T is specific in teleost fish (38). IgZ/T is specialized in mucosal immunity and present in gut mucus, closely resembling IgA in mammalian mucosal immunity (41). Three Ig L chain isotypes, named L1, L2, and L3, have been detected in zebrafish. L1 and L3 are Ig-κ orthologous (IgL-κ), and L2 is orthologous to Xenopus σ (38, 42). Thus, we focused on Rag, IgH-μ, IgL-κ, and CD79 expression in kidney marrow where Rag-expressing B cell development occurs as the major site of hematopoiesis in adult zebrafish (43, 44).

Zebrafish adults were bred and embryos were staged using standard practices (45). Rag2-mCherry and Lck-GFP transgenic lines were previously described (5, 46). All procedures using experimental animals were carried out under a protocol approved by the Fox Chase Cancer Center Institutional Animal Care and Use Committee.

Predicted zebrafish CD79a (XM_001339187) and CD79b (XM_002663902) mRNA sequences were found in the National Center for Biotechnology Information database. 5′-RACE was performed to detect and amplify the 5′ untranslated region of CD79a and CD79b mRNA. Primers included: 1) cd79a reverse 1, 5′-CTACGGCTTCTCCAGCTGAAT-3′ for the first round; cd79a reverse 2, 5′-CAGCTGAATGTCTTCCTCACA-3′ for second round; and 2) cd79b reverse 1, 5′-TCACTCTTGGCATGGAGACT-3′ for first round; cd79b reverse 2, 5′-CTCTTGGCATGGAGACTCAA-3′ for second round. RT-PCR and PCR reactions were performed according to the manufacturer’s recommendations (5′RACE kit; Invitrogen). CD79a and CD79b PCR products were amplified with PFU DNA polymerase, cloned into the pCR2.1-TOPO vector (Invitrogen), and subcloned into the PCS2⁺ vector. Clones were then characterized and validated by sequencing. CD79a mRNA was identical to NM_001326470, and CD79b was close to XM_009306430.2 (as a predicted sequence) in GenBank (see Supplemental Fig. 1).

Tissue RNA was prepared using TRIzol (Invitrogen) and phenol extraction, then precipitated with isopropanol and washed with 70% ethanol; pelleted RNA was resuspended in DEPC. In flow cytometry experiments, cells were sorted directly into RNA lysis buffer and RNA was prepared as previously described (47). RNA was reverse transcribed with random hexamers using the SuperScript II kit (Invitrogen). PCR reactions (RT-PCR and quantitative RT-PCR [qRT-PCR]) were performed with the primers listed in Table I.

The CD79a antisense RNA probe was prepared by 5′ linearization with NotI and synthesis with SP6 polymerase; the sense control was prepared by 3′ linearization with HindIII and synthesis with T7 polymerase. The CD79b antisense RNA probe was prepared by 5′ linearization with BamHI and synthesis with T7 polymerase; the sense control was prepared by 3′ linearization with XhoI and synthesis with SP6 polymerase. The IgH-μ clone (AY643753; GenBank) was obtained from Addgene (plasmid 11301); the antisense RNA probe was prepared by 5′ linearization with XbaI and synthesis with T7 polymerase; the sense probe was prepared by 3′ linearization with BamHI and synthesis with SP6 polymerase. Polymerases are from Roche Biotech, probes are labeled with DIG, and posttranscription cleanup was done with SigmaSpin columns (Sigma-Aldrich). In situ hybridization of zebrafish thin sections was carried out as described by Smith et al. (48). Staining in sections hybridized with antisense and sense (control) probes was developed with NBT/5-bromo-4-chloro-3-indolyphosphate, and then images were acquired using a spectral imaging camera and processed with Nuance software for visualization of the staining.

To generate CD79a -and CD79b-GFP transgenes, bacterial artificial chromosome (BAC) clones CH211-64K10 and CH211-93E11, containing CD79a and CD79b genes and an upstream genomic sequence, respectively, were obtained from the Children’s Hospital Oakland Research Institute and reengineered by galk recombineering (49), replacing the first exon from the ATG start codon of CD79a and CD79b by an enhanced GFP (EGFP)–polyA segment. CD79a and CD79b recombineered BAC plasmids (containing EGFP) were used as a PCR template for transgene generation with the following primers: CD79a (14k) (forward), 5′-GGCAGAAATTCCCAAAGACA-3′; CD79a (reverse), 5′-GGCGTCATACATGGACAGTG-3′; CD79b (5k) (forward), 5′-TAGGTTTTGCTCGCCTCTGT-3′; CD79b (reverse), 5′-GCCAGGTACACAACGTTCCT-3′. The restriction site RsrII was incorporated into both forward and reverse PCR primers for easy release of transgenes. Transgenes were amplified using the proofreading LA Taq (TaKaRa) enzyme and cloned into a TOPO TA vector (Life Technologies). Then the plasmids were digested with RsrII to release the transgene from the vector backbone; following agarose electrophoresis, transgenes were gel purified using a QIAquick gel extraction kit (Qiagen) and the DNA was resuspended in double-distilled H₂O at 200–500 ng/μl, then injected (30–50 pg per oocyte) into zebrafish eggs at the one- to two-cell stage. Larvae were monitored for transient GFP expression, discarding those that lacked expression. Founders were identified by PCR. This screen eventually yielded two CD79a (14k)–GFP lines and two CD79b (5k)–GFP lines.

Tissues were dissected from adult fish and dissociated by grinding between two frosted glass slides. Material was rinsed into a petri dish with staining medium (deficient RPMI 1640 [lacking phenol red, riboflavin and biotin], 3% newborn calf serum, 0.1% sodium azide) and this suspension was passed through a 70-μm Nitex cloth screen into a 15-ml conical centrifuge tube. Cells were pelleted by centrifugation (1200 rpm; 273 × g; 7 min; 4°C) and washed twice more with staining medium. Cell pellets were resuspended in staining medium that included propidium iodide (PI; 1 μg/ml) to facilitate exclusion of dead cells (also evaluation of DNA content in cell cycle) and transferred to sample tubes. Clutches of larval fish (day 7 to day 28) were dissociated by grinding between frosted glass slides as above, but after initial pelleting, the sample was resuspended in 1 ml of staining medium and layered over 1 ml of Lympholyte M (Accurate Chemical & Scientific) and centrifuged (2800 rpm; 1762 × g; 20 min; room temperature). Cells were recovered from the interface and washed twice with staining medium. Cell pellets were resuspended as above. Flow cytometry was performed with a FACSAria II (BD Biosciences).

Images of whole fish or larvae were acquired with a Nikon SMZ1500 stereomicroscope using a Nikon DS-Fi1 camera and NIS-Elements software. Fish frozen sections were mounted with Vectashield mounting media (Vector Laboratories), overlaid with coverslips, and images were acquired with a Nuance camera (CRi, Hopkinton, MA) on a Nikon Eclipse 80i microscope. WISH white-light images were also acquired with the same system. Nuance spectral images were unmixed using CRi software. Confocal images were acquired with a Nikon C1 spectral imaging confocal microscope, using 488 nm excitation for GFP and detecting fluorescence between 415 and 570 nm. mCherry fluorescence was excited at 561 nm and detected between 570 and 655 nm.

The bacterial strain used for infection is Pseudomonas aeruginosa PA14 (p66TDC1) strain with Tomato Red expression (50). A stab of this strain (obtained from Dr. J. Singer, Maine School of Marine Sciences, University of Maine) was plated out onto L agar with 750 μg of ampicillin per milliliter. Bacteria were grown in Luria–Bertani medium and concentration was determined by comparing OD600 of a dilution series, simultaneously plating dilutions in plates, and then counting colonies. For infection, adult CD79a-GFP and CD79b-GFP transgenic zebrafish were anesthetized with 160 μg/ml tricaine and injected i.p. with 10⁴ or 10⁵ CFU of P. aeruginosa (RFP labeled). A third (control) group was injected with PBS. Bacterial exposures were performed in triplicate per dose. After recovery from anesthesia, each group was moved to a small offline tank in a quarantine room and maintained for 28 d. Fish were observed daily for signs of disease and morbidity. Fish were anesthetized with 160 μg/ml tricaine and observed two times per week by microscopy (fluorescence detection of GFP and RFP) to track changes in B cell numbers and localization and bacterial growth. Three fish were selected from 7, 14, and 28 d postinfection for each group; dissected tissues, including liver, intestine, spleen, and kidney, were used for FACS analysis.

Zebrafish CD79a and CD79b cDNA sequences were amplified by 5′RACE using primers based on partial clones that showed homology to channel catfish CD79a and CD79b. The full-length CD79a and CD79b cDNA sequences and their deduced amino acid sequences are shown in Supplemental Fig. 1. Both CD79a and CD79b zebrafish sequences contain the ITAM domain, and strong homology around this domain was found between zebrafish, catfish, mouse, and human amino acid sequences (Fig. 1, bold line), indicating that zebrafish CD79a and CD79b mediate the same function as CD79a and CD79b in other vertebrates. These ITAM domains are known to be important in BCR signaling, as they provide two tyrosine phosphorylation sites (Fig. 1, asterisk marked) for docking Syk kinase and initiating downstream signaling pathways (51).

Fish lack bone marrow and lymph nodes, and B cell lymphopoiesis in zebrafish occurs mainly in the kidney, as a kidney marrow, after the larval stage. The pronephros is the first kidney to be formed during embryogenesis, then the mesonephros forms as an extended kidney, and pronephros head kidney is the major site of B cell lymphopoiesis in the adult (52). In the adult zebrafish, expression of CD79a and CD79b was strongly detected in kidney marrow sections by in situ hybridization using antisense RNA probes, localized in areas similar to IgH-μ (Fig. 2A). To further assess tissue distribution of zebrafish CD79a and CD79b, RT-PCR with primers of CD79a and CD79b was performed (Fig. 2B, Table I) (53). This showed predominant expression in head kidney marrow where B cells develop, and also in spleen, a site of mature B cells by reentry from circulation after B cell development in kidney. Although at lower transcript levels, CD79 and IgH-μ were also detected in intestine.

Next, two lines each of CD79a-GFP and CD79b-GFP transgenic reporter lines (Fig. 3A) were established. These four lines were bred to the F₃ generation and showed similar stable GFP expression patterns. In all lines, GFP expression was clearly detected in the primary immune site, head kidney marrow (Fig. 3B). These GFP-labeled cells in the kidney were distributed around the pronephric tubules, in a pattern similar to the antisense RNA probe staining originally detected in Fig. 2A. GFP-labeled cells showed a small round morphology, resembling lymphocytes (Fig. 3B, right panels), supporting their identification as B cells. Scattered GFP⁺ cells were also visualized microscopically in other tissues, including in intestine of both CD79a-GFP and CD79b-GFP cell lines (Fig. 3C), as originally found by RT-PCR in Fig. 2B.

To determine whether GFP in these transgenic lines labels the B lineage, we used flow cytometry to sort GFP⁺ and GFP⁻ cells from kidney marrow (Fig. 3D) and isolated RNA from the sorted samples. We also sorted GFP⁺ and GFP⁻ cells from Lck-GFP transgenic zebrafish from kidney to identify T cells for comparison. We then used RT-PCR to examine expression of CD79a, CD79b, IgH-μ, and Lck (Fig. 3E). The GFP⁺ fraction from both lines showed expression of CD79a, CD79b, and IgH-μ, but not Lck. In contrast, GFP⁺ cells from the Lck reporter line lack expression of the B cell genes, but they are strongly positive for Lck mRNA. Thus, CD79:GFP expression clearly distinguishes between B and T lymphocytes in zebrafish, not expressed by the T lineage. We conclude that GFP⁺ cells in the CD79a-GFP and CD79b-GFP transgenic lines label the B-lineage cells in kidney marrow.

In mice, CD79a and CD79b are expressed early in B-lineage cell development with the Rag1 and Rag2 genes before generation of BCR-expressed immature B cells (30). In zebrafish, Rag1 and Rag2 are also both present. Thus, to determine timing of CD79a and CD79b expression in zebrafish B-lineage cells, we crossed one of each CD79a-GFP and CD79b-GFP transgenic fish lines with the Rag2-mCherry transgenic zebrafish line, which bears a Rag2 regulatory segment driving expression of mCherry.

Fig. 4A shows the screening of early developing zebrafish from the first 7 d postfertilization (7 dpf) to the 28 dpf (4 wk) stage by flow cytometry analyses of cell suspensions prepared from clutches of dual transgenic line, in comparison with nontransgenic zebrafish as a control (wild-type AB). Before this dual transgenic line analysis, we noticed that offspring from female CD79:GFP transgenic carriers exhibited diffuse maternal GFP mRNA. Thus, experiments with larval clutches were established by crossing transgenic male parents from both CD79a-GFP and CD79b-GFP lines. Even under these conditions, we detected GFP in the midbrain, ventral branch arches, and otic vesicle during the first 6 d postfertilization in CD79a-GFP zebrafish by sequential screening. FACS analysis of macerated clutches of the CD79a lines at 1 wk to <3 wk postfertilization consistently showed 5- to 10-fold higher frequencies of GFP⁺ cells compared with the CD79b lines, and most of these CD79a:GFP⁺ cells found in the larval stage were negative for Rag2:mCherry (Fig. 4A). Thus, the expression of GFP in the B lineage was unclear in younger larvae. Then, in subsequent analyses of the 3-wk-old (21 dpf) larval–juvenile transition stage, fish showed low numbers of double-positive cells detected with either the CD79a-GFP or CD79b-GFP reporter, and double-positive cells became abundant by 4 wk postfertilization (28 dpf) as juveniles.

After 4 wk postfertilization (1 mo), CD79a- and CD79b-GFP/Rag2-mCherry analysis was performed with kidney marrow. As shown in Fig. 4B, CD79a:GFP⁺ cells in the CD79a-GFP/Rag2-mCherry transgenic line increased by 16 wk (4 mo) in adult kidney, and part of them showed the clearly decreased Rag2:mCherry expression, which was more prominent at 32 wk (8 mo). The CD79b-GFP/Rag2-mCherry transgenic line showed a similar increase in CD79a:GFP⁺ cells in kidney during 1–8 mo aging, as shown in a 5-mo adult CD79b-GFP/Rag2-mCherry transgenic line in Fig. 4D. Fluorescence microscopy of kidney and thymus sections of these adults revealed strikingly different patterns of expression and localization of these genes (Fig. 4C). In kidney marrow, we observed a broad dispersal of red-only (Rag2⁺CD79b⁻), yellow (Rag2⁺CD79b⁺), and green (Rag2⁻CD79b⁺) cells throughout the kidney tissue. In contrast, green fluorescent cells were only localized along the border of the thymus, and a large number of red cells were observed within the thymus organ. Fig. 4C (bottom) shows high-power images obtained with a confocal microscope that reinforced this difference. T-lineage cells rearrange TCR genes during their development in the thymus. As expected from the Fig. 4C microscope images, the flow cytometry pattern shown in Fig. 4D demonstrated fewer CD79b:GFP⁺ cells in thymus than in kidney, and such GFP⁺ cells mostly expressed decreased levels of Rag2:mCherry.

Our major question was whether the CD79⁺μ⁺ pre–B cell stage exists in zebrafish. In the CD79b-GFP/Rag2-mCherry line, RT-PCR of the few GFP⁺ cells isolated by sorting from 3 wk (21 dpf) larval clutches showed weak coexpression of IgH-μ and IgL-κ, but not IgH-μ alone (Fig. 5A). In adult kidney, GFP⁺Rag2-mCherry⁺ cells showed further increased IgH-μ and IgL-κ levels (Fig. 5A). We divided these adult kidney marrow cells into several fractions, with different CD79:GFP levels in Rag2:mCherry⁺ cells, including the Rag2:mCherry dimmest cell fraction, and characterized Ig gene expression (by RT-PCR, Fig. 5B), cell size (by forward light scatter, Fig. 5C), and cell cycle (by PI staining of cell DNA, Fig. 5D). This analysis revealed that Rag2:mCherry⁺ cells with dim to intermediate CD79b:GFP⁺ fractions (Fr. 2–3) expressed IgH-ζ and IgL-κ, the alternate fish IgH known as IgZ/T. This IgZ/T was predominant in cells with the lowest GFP levels as Fr. 2. Fr. 3 showed a mixture of IgH-ζ and IgH-μ, whereas IgH-μ was predominant in cells in Fr. 4 and later. It was striking that IgL-κ was continuously found through all CD79b:GFP⁺ cell fractions coexpressing with H chain. This pattern was very different from what is seen in the mouse, where H chain precedes L chain expression at a distinct cytoplasmic IgH μ⁺ pre–B cell stage. As shown in Fig. 5C, Fr. 4–6 cells with higher levels of CD79b:GFP together with middle to diminishing levels of Rag2:mCherry are small cells, and they have exited the cell cycle, because PI staining showed a <0.5% increased cell fraction, in contrast to 12–22% for Fr. 1–3 (Fig. 5D). Fr. 3 cells coexpressing IgH-μ and IgL-κ contained cycling cells with slightly increased cell size and DNA content. This suggested that Fr. 3 cells become a mature quiescent stage as Fr. 4.

Although Rag2:mCherry was consistently detected from a high to diminishing level in all CD79b:GFP⁺ cells, we anticipated the possibility that this long-term detection was due to the long half-life of fluorescent proteins after Rag2 expression is extinguished, as elongated stability. Previous work in mice found that the GFP⁺ level in Rag2-GFP immature B cells was higher than comparative mRNA levels (54). Similarly, analysis of the thymus in Rag2-GFP zebrafish also suggested longer persistence of GFP protein than Rag2 protein, because GFP was detected in both cortical and medullary regions of the thymus (5). In the zebrafish CD79-GFP/Rag2-mCherry lines shown in this study, although Rag2-mCherry was clearly detectable, mCherry levels were lower by Fr. 2 than Fr. 1 and further declined in later fractions (Fig. 5E). This decline of Rag2:mCherry and expression of both IgH and IgL in Fr. 2–5, distinct from Fr.1, was also found in the CD79a-GFP/Rag2-mCherry transgenic line (Fig. 5F). When qRT-PCR was performed on these fractions prepared from both CD79a- and CD79b-GFP/Rag2-mCherry lines, we consistently found strong downregulation of both Rag1 and Rag2 mRNAs after the Fr. 1 stage (Rag1 was reduced by >30-fold from Fr. 1 to Fr. 2) and undetectable by Fr. 5 (Fig. 5G, left panel). In opposition to Rag mRNAs, both CD79a and CD79b mRNAs were upregulated after the Fr. 1 stage, and increased by Fr. 3/4, together with IgM mRNA (Fig. 5G, right). This confirmed the absence of a Rag^hiCD79⁺IgH-μ⁺ pre–B cell stage, in contrast to the presence of Rag^hiCD79⁻IgH-μ⁻ cells and Rag^loCD79⁺IgH-μ⁺IgL-κ⁺ B cells. Interestingly, Fr. 5 lacked Rag mRNAs and consistently showed decreased CD79a and CD79b mRNA levels as compared with Fr. 3/4, and often showed further increased IgH-μ mRNA. This suggested that after generation of mature B cells, reduced expression of CD79 may be occurring during differentiation.

To confirm whether alterations in the level of CD79 occur during zebrafish B cell differentiation, we examined B cell responses to pathogen by exposing CD79a-GFP and CD79a-GFP lines to bacteria. As shown with the CD79a-GFP line, injection of adult fish i.p. with P. aeruginosa, which also contained a plasmid coding for the RFP Tomato Red (50), resulted in changes in GFP⁺ cell numbers in secondary tissues, including a prominent increase in intestine (Fig. 6A). We noted an increased frequency of both CD79a⁺⁺ and CD79a⁺:GFP cells in intestine, but also in spleen and liver, in contrast to kidney (Fig. 6B). Fig. 6C shows an increased incidence of CD79a⁺:GFP cells as compared with CD79a⁺⁺:GFP cells in intestine at 1 wk postinfection. We also found an increase in B cell size, particularly in the intestine early postinfection (1 wk), that resolved by 4 wk postinfection (Fig. 6D). At this 4 wk stage postinfection (and control), we then purified B cell fractions in kidney and assessed expression of membrane and secreted IgH-μ by RT-PCR. Clearly, infection-induced B cells with low to intermediate CD79a:GFP levels were present in kidney, showing decreased levels of membrane IgH-μ and greatly increased levels of secreted IgH-μ, indicating the presence of plasma cells in kidney that had been generated in response to infection (Fig. 6E). This confirmed that downregulation of CD79 occurs in zebrafish, in parallel with the differentiation of mature B cells to Ig-secreting plasma cells, as found in mammals.

By generation of CD79-GFP lines, we investigated zebrafish B cell development and differentiation, seeking to determine similarities and differences from mammalian B cells.

In adult mouse B cell development, a surrogate L chain is involved in progression to the pre–B cell stage. Although it has already been known that the zebrafish lacks a surrogate L chain (20), it was unclear whether an alternative pre–B cell stage exists, because early fetal/neonatal B-1 B cell development in mice can occur in the absence of a surrogate L chain (12, 55), and this surrogate L chain–independent mouse B-1 development still progresses through a Rag^hiCD79⁺IgH-μ⁺ pre–B cell stage. In this study, we demonstrate that the zebrafish does not require the pre–B cell developmental stage. Both H and L chains rearrange simultaneously and B cells expressing complete BCRs initially proliferate with slightly increased cell size, before entering a mature quiescent stage, more similar to chickens (18) than mammals.

Previously, it was reported that the pre–B cell stage exists in zebrafish by using a newly generated IgM1:eGFP line crossed with a Rag2:DsRed line (56). In the context of this dual IgM/Rag2 transgenic reporter line, it was concluded that IgM⁻, IgM^lo, and IgM⁺ cells all express a similar level of Rag2⁺ in kidney, as compared with the B cells that exhibit decreased levels of Rag2 with further increased IgM⁺. This result led to the conclusion that a Rag2⁺IgH-μ⁺ pre–B cell stage exists. However, our new generation of CD79a-GFP and CD79b-GFP lines then crossed with the Rag2-mCherry line allowed us to reveal further details of the B cell development process in zebrafish. Previous IgM/Rag2 data showed the presence of IgZ in the Rag2⁺IgM⁻ cell fraction. Now, our CD79b/Rag2 transgenic system revealed that the Rag2⁺IgM⁻ fraction consists of the mixture of two distinct cell fractions, Rag2^hiCD79⁻IgM⁻IgZ⁻ precursor cells (Fr. 1) and Rag2^loCD79^lo/+IgM^−/+IgZ⁺ cells (Fr. 2) (Fig. 5B). Thus, IgZ⁺ cells are already Rag^lo, as further confirmed by the presence of Rag^loIgZ⁺ cells in the CD79a/Rag2 line. Importantly, our data demonstrate that a strong reduction of Rag mRNAs occurs when B cells are initially generated, and a further reduction of Rag mRNAs results at differentiated B cell stages often in conjunction with increased IgM. Thus, our discovery of Rag2^hiCD79⁻ cells in our CD79/Rag2 transgenic system allowed us to clarify the Rag2 levels.

Analysis of our CD79/Rag2 transgenic lines also allowed us to find nonlymphoid cells with CD79:GFP expression without Rag2-mCherry at the larval stage (Fig. 4A). In the CD79a-GFP line, all CD79a:GFP⁺ cells showed more expression in non–B-lineage cells than in the CD79b-GFP line at the larval stage. Under microscopy, we also detected some tissue-restricted GFP at 24–48 h postfertilization in both CD79a-GFP and CD79b-GFP lines, with neural expression by CD79a-GFP lines and muscle expression by CD79b-GFP lines. An explanation for this nonlymphoid cell expression of GFP remains unclear. However, a previous report described expression of CD79a with the ITAM domain by immature myeloid cells in mouse bone marrow, and expression of CD79a has been found in some cases of acute myeloid leukemia in human (57). Thus, there is a possibility of CD79 expression in nonlymphoid cells at the larval stage, particularly CD79a. This requires further larval analysis to confirm the presence of CD79-expressing nonlymphoid cells in early zebrafish development.

A major remaining question in teleost fish is whether the distinct early generated B cell development occurs before adult B cell development. Although green⁺red⁺ (CD79⁺Rag2⁺) double-positive cells were difficult to detect at the larval stage, our finding of CD79^lo IgZ B cells in adults supports the possibility that distinct B cell development may exist in early life. IgZ, also known as IgT in teleost fish, has been known to be expressed in early B cells in mucosal immunity (37, 39, 41). Importantly, IgH sequencing of zebrafish discovered that the expression of zebrafish (DJC)ζ precedes to (DJC)μ, leading to early IgH-ζ transcript before IgH-μ, thereby generating IgZ B cells (37). This resembles TCRδ expression preceding TCRα in T cell development. In mice, early fetal γδTCR T cell generation occurs prior to αβTCR T cell generation, leading to the proposal that IgZ/T B cell generation may also be occurring prior to generation of IgM B cells. In rainbow trout, IgT is present 4 d before hatching, and IgT is the main B cell subset in GALT, including in gut mucosa as found with IgA in mammals (41). In zebrafish, at the 4 dpf larval stage, the presence of V(D)J rearrangement encoding B cell receptors (53), and all IgLC isotype mRNAs, including IgL-κ, were detected at low levels (20), and thus B cells are generated. The pancreas is one of the abdominal organs, an outpocketing of the posterior gut, and the presence of early B cells was previously found at the larvae stage in the pancreas (53). These observations led to the proposal that IgZ generation occurs earlier at the larval stage, along with IgM, as an initial component of innate immunity. In our adult kidney analysis, we found that IgZ/T B cells were often within the lowest CD79:GFP-expressing fraction. Because intestine is sensitive to infection and generates differentiated B cells with strongly decreased levels of CD79, CD79^lo IgZ B cells found in adult kidney may also include the early generated, differentiated B cells that are maintained in kidney.

Further evidence supports the existence of distinct B cell development in the larval stages of zebrafish that differs from adults. In mice, the fetal/neonatal B-1 lineage precursor is Lin28b⁺let7⁻, in contrast to the the adult Lin28b⁻let7⁺ B-2 lineage precursor (58). Lin28 was originally found in Caenorhabditis elegans, and Lin28 expression was also found during embryonic development in both chicken and zebrafish (59, 60). In zebrafish, lin28a and lin28b were found to be ubiquitously expressed early after embryonic fertilization, then expression decreased, and Let7a/let7b microRNA expression increased at 3 dpf (59). Ikaros is a marker for lymphoid progenitors, and higher expression is found in fetal hematopoietic stem cells than adult hematopoietic stem cells in mice (61). In zebrafish, elimination of the Ikaros gene resulted in absolute loss of IgH-ζ–bearing B cells (IgZ) but sparing of some IgH-μ B cells (IgM) (62). In mice, fetal/neonatal generated B1 B cells can enter the intestine and generate IgA, independent of T cell help (63, 64). Perhaps this issue of distinctive developmental origin of B cells in zebrafish can be further addressed by sorting early progenitors in culture. There is a report of a culture system using a zebrafish stromal cell line, ZKS (65), that might facilitate such work together with cells from our CD79-GFP lines. Developing a culture system would also facilitate direct testing of the proposed progression of B cell stages.

In addition to enabling tracking of B cell immune responses and a more precise examination of their development, the identification of B cell–specific regulatory elements may also provide an approach for the development of B cell cancer models. In contrast to mice, the development of B cell leukemia models in zebrafish has proven difficult to date, with one report of B cell acute lymphoblastic leukemia driven by a TEL-AML1 translocation fusion with an actin promoter, but not Rag (66). Both the CD79a-GFP and CD79b-GFP transgenic reporter lines drive high GFP expression, so these are candidates for expressing oncogenes selectively in B cells. In the future we expect that these lines and the regulatory elements defined in the course of this work will lead to the development of leukemic B cell models that may yield insights into their growth, with potential therapeutic implications.

We thank the staff from the Zebrafish Core Facility, as well as Dietmar Kappes and Kerry Campbell for discussion and comments on the manuscript.

The authors have no financial conflicts of interest.