αβ/γδ T cell lineage outcome is regulated by intrathymic cell localization and environmental signals

The level of γδ T cells varies between species because of differences in cellular dynamics and signals from thymic niche.

Abstract

αβ and γδ T cells are two distinct sublineages that develop in the vertebrate thymus. Thus far, their differentiation from a common progenitor is mostly understood to be regulated by intrinsic mechanisms. However, the proportion of αβ/γδ T cells varies in different vertebrate taxa. How this process is regulated in species that tend to produce a high frequency of γδ T cells is unstudied. Using an in vivo teleost model, the medaka, we report that progenitors first enter a thymic niche where their development into γδ T cells is favored. Translocation from this niche, mediated by chemokine receptor Ccr9b, is a prerequisite for their differentiation into αβ T cells. On the other hand, the thymic niche also generates opposing gradients of the cytokine interleukin-7 and chemokine Ccl25a, and, together, they influence the lineage outcome. We propose a previously unknown mechanism that determines the proportion of αβ/γδ lineages within species.

INTRODUCTION

The adaptive immune system consists of two T cell sublineages, defined according to the expression of αβ or γδ T cell receptors (TCRs). Their development starts with the entry of common progenitors into the thymus and their migration into distinct thymic niches to interact with thymic epithelial cells (TECs) (1, 2). Mechanisms regulating the αβ or γδ T cell lineage choice in mice and humans are explained using two models: precommitment and TCR instructional (1, 3). The precommitment model proposes that fate is determined in developing T cells (thymocytes) before TCR expression. This model is supported by studies showing that interleukin-7 (IL-7) signaling is crucial for TCRγ development (4–6), and that early murine thymocytes expressing a high level of IL-7 receptor (IL-7R) are thus biased toward the γδ T cell lineage (7). However, the TCR-instructed model proposes that the quantitative difference in TCR signals is critical for T cell fate choice. Murine thymocytes undergo simultaneous TCRγ, TCRδ, and TCRβ gene rearrangements and, depending on the outcome, may express either TCRγδ or the pre-TCR, a complex composed of a TCRβ chain and the pre-TCRα chain, encoded by PTCRA gene. A weak TCR signal favors development of the αβ lineage, while a strong signal biases toward the γδ T cell lineage (8, 9). In human and mice, most thymocytes are developmentally biased toward the αβ lineage, with only 1 to 5% developing into the γδ lineage. However, the proportion of γδ T cells is substantially higher (10 to 40%) in other species, including sheep, cattle, pigs, goats, elk, chickens, axolotl, zebrafish, and shark, which are termed γδ T cell high species (10, 11). How the T cell lineage decision proceeds in these species has not been studied.

Emerging studies reveal that genetic networks underlying T lymphocyte development are evolutionarily conserved among jawed vertebrates (12–15). For γδ T cells, however, studies to date are mainly restricted to the histological examination, highlighting that their thymic localization varies between species. While αβ and γδ T cells are scattered throughout the murine thymus, they do not appear to be intermixed in chickens, in which γδ T cells preferentially occupy the corticomedullary junction, where αβ T cells are less likely to localize (16). In nurse shark and amphibian Mexican axolotl, γδ T cells are exclusively located in the outer zone (OZ) (subcapsular region) of the thymic cortex and in the medulla, while αβ T cells are located inside the thymic cortex and in the medulla (17, 18). In sheep, early γδ T cells are mainly located in the cortex, including the subcapsular region (19). It remains unknown the extent to which cellular localization in the thymus and environmental signals could influence the lineage outcome. To study this process, we chose to use medaka teleost fish as a model. We report that medaka αβ and γδ T cells are spatially organized into two distinct thymic niches, resembling the situations described for nurse sharks and axolotl. Our in vivo findings reveal that chemokine receptor–mediated cell localization within the thymus influences the responsiveness of thymocytes to thymic IL-7 signal that bias toward the γδ T cell fate. By combining experimental approaches with computational modeling, we show how the interplay between intrathymic cell localization, thymic signals, and the onset of gene expression influences the T cell lineage outcome, revealing a mechanism that determines the frequency of thymic γδ T cells in species.

RESULTS

T cell sublineages are spatially organized into two distinct thymic regions

Whole-mount in situ hybridization (WISH) using probes specific for TCRα (tcra), TCRβ (tcrb), TCRγ (tcrg), and TCRδ (tcrd) constant regions revealed that tcra- and tcrb-expressing cells (hereafter called TCRαβ⁺ cells) were enriched in the center of the medaka thymus, whereas tcrg- and tcrd-expressing cells (hereafter called TCRγδ⁺ cells) were predominantly located at the ventral side of the thymus (Fig. 1A). Fluorescent double in situ hybridization (FISH) in the whole thymus, followed by confocal microscopy, showed that TCRαβ⁺ and TCRγδ⁺ cells were organized in a spherical core and in the surrounding shells (Fig. 1B). Thus, the larval thymus can be subdivided into the OZ and the inner zone (IZ). The spatial organization of TCRαβ⁺ and TCRγδ⁺ cells in the thymus was retained until adulthood (Fig. 1C and fig. S1), resembling that of the nurse shark (17) and axolotl (18), supporting the notion that TCRαβ⁺ and TCRγδ⁺ cells are not intermixed in the thymus of species that tend to produce high number of γδ T cells.

Fig. 1. Spatial distribution of T cell sublineages in the medaka thymus.

(A) Expression of TCR genes in the medaka larval thymus. The yellow dashed circles indicate the position of larval thymus. Scale bars, 30 μm. (B) Top: 3D rendering of a thymus stained with tcrg and tcrb probes, as revealed by FISH and imaged by confocal microscopy. Scale bars, 15 μm. Bottom: One plane (z = 1 μm) of a Z-stack spanning the whole thymus stained with tcrg and tcrb probes. Scale bar, 10 μm. The fluorescence intensities along the cyan line are shown. AFU, arbitrary fluorescent units. (C) Scheme depicting the distribution of TCRαβ⁺ and TCRγδ⁺ cells in the medaka larval and adult thymus. Note that thymic cortical (C) and medullary (M) regions were defined by the spatial expression of rag2 and autoimmune regulatory genes, respectively (12, 20).

Progenitors first enter the thymic niche where they potentially develop as γδ T cells

Since T cell development depends on the continual colonization of the thymus by progenitors derived from hematopoietic tissue, we investigated whether they enter each thymic zone to differentiate into a given lineage. In teleosts, ccr9a, the paralogous to mouse Ccr9, and its ligand ccl25a, the paralogous to mouse Ccl25, are required for migration of progenitors into the thymus (12, 20–22). Long-term in vivo imaging of the medaka transgenic reporter lines revealed that ccr9a⁺ progenitors migrate through the mesenchyme in a straight path toward the larval thymus (20). To examine the thymic entry sites of progenitors, we used a double-transgenic ccr9a:h2p-gfp; ccl25a:tag-rfp reporter fish, where green fluorescent protein (GFP) is constitutively expressed in progenitors and thymocytes (20), whereas TagRFP (a monomeric derivate of red fluorescent protein and hereafter called RFP) is expressed in TECs and stromal cells in the extrathymic region (fig. S2). Time-lapse in vivo imaging of the entire thymus region to track 88 migratory thymic cells in 10 freshly hatched larvae revealed four routes of migration (Fig. 2, A to C, and movie S1). We found that ~70% of the cells moved into either anterior/ventral or posterior/ventral part of the thymus (Fig. 2A, routes I and II), and the remaining cells entered the anterior or posterior region of the thymus (Fig. 2A, routes III and IV). Upon entry, thymic cells slowed down (20) and remained in the thymic OZ, where they were in close contact with TECs (Fig. 2D and movie S2). This indicated that the medaka larval thymus contains defined niches for the progenitors, and upon entry into thymus, these cells remain in the OZ. This observation is in agreement with our previous study reporting that early events of T cell development such as specification and proliferation of progenitors occur in this region (20).

Fig. 2. Four defined thymic niches for seeding by progenitors.

(A) 3D rendering of a transgenic ccr9a:gfp larva illustrating four migratory routes (arrows) for seeding the thymus. From 88 tracked cells, 35, 26, 15, and 12 cells migrated inward the thymus through routes I, II, III, and IV, respectively (data from 10 videos). Scale bar, 15 μm. (B and C) Still photographs from time-lapse recording showing the migration of four ccr9a⁺ cells (indicated by numbers) into the ventral side of the thymus (movie S1). The yellow dashed circles indicate the position of cell entry into the thymus. Numbers indicate time in minutes. Scale bars, 15 μm. (D) Left: Images of a double tg[ccr9a:gfp (green); ccl25a:rfp (cyan)] larva illustrating the migratory path (white arrow) of thymic immigrants. Right: Still photographs from a time-lapse recording illustrating the close contact between thymocytes (ccr9a⁺) and TECs (ccl25a⁺) in the ventral side of the thymus (movie S2). Numbers indicate time in minutes. Scale bars, 15 μm.

The Ccr9b-Ccl25a axis controls the intrathymic localization of cells and αβ/γδ outcome

Unlike αβ T cells, the development of medaka γδ T cells is not coupled with migration throughout thymic niches; progenitors first enter a thymic niche, where their proliferation, specification (20), and differentiation into γδ T cells are favored. Therefore, to test the hypothesis that intrathymic positioning of the cells decides the αβ/γδ lineage fate, we focused on members of chemokine receptor family, which regulate the positioning of immune cells within lymphoid organs, including the thymus (23, 24). Our expression survey of 28 chemokine receptors in medaka (25) showed that no candidate gene was coexpressed with tcrg in the thymic OZ, while in the thymic IZ, only ccr9b was coexpressed with tcra/b (Fig. 3A and fig. S1). This gene is the result of the teleost-specific duplication event of the ancestral ccr9 gene (24) and is mainly expressed in the thymus (12, 20, 21). To investigate its function, we generated a Ccr9-deficient fish using CRISPR-Cas9, deleting the entire exon 2 encoding the seven transmembrane regions (Fig. 3B). We carried out in vivo analysis of the offspring of ccr9b^+/− fish, carrying the ccr9b:rfp reporter (Fig. 3C). Live recording of the entire larval thymus revealed that the RFP⁺ cells were scattered throughout the ccr9b^−/− thymus, in contrast to their wild-type (WT) siblings (Fig. 3D). We also noticed an approximately twofold decrease in fluorescence intensity (Fig. 3E), which did not correlate with their intrathymic localization in the ccr9b^−/− larvae (Fig. 3F and movie S3). The number of RFP⁺ cells (Fig. 3G) and their average cell speed (Fig. 3H and movie S4) were also significantly reduced in the ccr9b^−/− thymus compared to WT tissue. These data reveal that ccr9b is required for the correct positioning of thymocytes in the thymic IZ.

Fig. 3. Ccr9b deficiency affects the correct positioning of thymocytes.

(A) Expressions of ccr9b (red), tcrg (magenta), and tcrb (yellow) in the larval thymus, as revealed by FISH and imaged by confocal microscopy (z = 1 μm). OZ and IZ indicate the positions of OZ and IZ of the thymus, respectively. Scale bars, 1 μm. Bottom: Fluorescence intensities along the yellow lines. (B) Top: Schematic of the ccr9b gene structure, indicating the deletion introduced by CRISPR-Cas9. A schematic description of the ccr9b:rfp construct is shown for comparison. Bottom: Polymerase chain reaction (PCR) analysis of the ccr9b locus in WT and ccr9b^−/− fish. L1 and L2 indicate 1-kb and 100–base pair (bp) DNA ladders. (C) Generation of Ccr9b-deficient siblings carrying the ccr9b:rfp reporter construct. (D) Spatial distribution of RFP⁺ cells in WT and Ccr9b^−/− fish carrying the ccr9b:rfp reporter. Dashed circles indicate the position of thymus. Scale bars, 30 μm. (E) Thymus-scale RFP intensity (box-and-whisker plot) revealed by confocal imaging of the entire thymus. AU, arbitrary units. (F) Spatially resolved mean RFP intensity profiles. (G) Number of RFP⁺ cells in the thymus; N indicates number of biological samples. (H) Average cell speed of RFP⁺ cells in the thymus; n indicates total number of cells examined from >3 samples per condition.

Next, we examined the thymic expression of TCR chains in the ccr9b^−/− larvae. In WT thymus, the ratio of tcrb to tcrg expression was 1.5 ± 0.4 (means ± SD; N = 4), as suggested by the surface area of fluorescent cells in the entire thymus (Fig. 4, A and B). This ratio was approximately four- to sevenfold less in the ccr9b^−/− thymus (0.2 ± 0.11; N = 6), while the total expression of tcrg and tcrb was unchanged (Fig. 4C), indicating that the development of TCRγδ⁺ cells was enhanced. This is further supported by the increased expression of rorc (encoding retinoic acid–related orphan receptor γ) in the ccr9b^−/− thymus (fig. S3A), which is essential for the differentiation of a subset of γδ T cells in mice (26). More TCRγδ⁺ cells were also seen in the adult thymus (fig. S3B) and spleen (fig. S3C). The reduced expression of bcl11b, a transcription factor required for αβ T cell commitment (12, 27), and foxp3, a marker for regulatory T cells, further supported the reduction of TCRαβ⁺ cells in ccr9b-deficient fish (fig. S3A). To determine whether ccr9b-deficient thymocytes shifted to the γδ lineage, we studied tcrg expression in ccr9b^−/− larvae with ccr9b:rfp reporter. In contrast to their WT siblings, the mutants had a group of tcrg-expressing RFP⁺ cells, located mainly in the thymic OZ (Fig. 4D). Together, these results strongly suggest that lack of Ccr9b triggers γδ T cell development. Given that ccr9b is expressed neither in progenitors before thymus seeding (20) nor in TCRγδ⁺ cells, the shift toward the γδ lineage is most probably due to the mislocalization of thymocytes before T cell lineage commitment.

Fig. 4. Manipulation of Ccr9b-Ccl25a axis affects the T cell lineage outcome.

(A) 3D rendering of WT, ccr9b^−/−, ccl25a:ccl25a⁺, and ccr9b:ccl25a⁺ thymuses stained with tcrg (magenta) and tcrb (yellow) probes using FISH. Scale bars, 10 μm. (B) Proportions of tcrb and tcrg expression in the thymus as revealed by FISH analysis. (C) Total tcrb and tcrg expression in the thymus normalized to the average of WT thymus at 10 dpf. N in (B) and (C) indicates number of biological samples. n.s., not significant. (D) Representative images of tcrg expressions in WT and ccr9b^−/− fish carrying the ccr9b:rfp reporter, as revealed by combined FISH against tcrg (magenta) and RFP immunostaining (white). Scale bars, 10 μm; inset shows a higher magnification of the indicated region. Scale bars, 3 μm. (E) Top illustrates constructs used to overexpress ccl25a using ccl25a and ccr9b promoters. Note that GFP was used to select embryos with strong expression in the thymus. Bottom describes the experimental rationale. T, thymocyte. (F) Representative images (left) and number (right) of mitotic cells [pH3⁺ (phospho–histone 3); red] in the WT, ccr9b^−/−-, ccl25a:ccl25a-, and ccr9b:ccl25a-injected thymus. The yellow dashed circles demarcate the thymus. Scale bars, 30 μm. Each dot represents an individual sample.

To examine the contribution of chemokine-mediated cell localization to the T cell lineage outcome, we next manipulated the thymic chemokine milieu. Previous studies have shown that Ccl25a, the ligand for Ccr9a and Ccr9b, is the sole chemokine that is expressed in the thymus of medaka and zebrafish larvae (12, 20, 21). FISH analysis showed that this chemokine is expressed more in TECs located in the thymic IZ than those in the OZ (fig. S2A). To increase the chemokine levels within the thymus, we designed two constructs where ccl25a expression was driven by the ccl25a or ccr9b promoter (Fig. 4E). Injection of both constructs resulted in a two- to fivefold increase in the ratio of tcrb:tcrg expression (Fig. 4, A and B). To exclude the possibility of Ccl25a triggering cell proliferation, we counted the mitotic cells in the entire thymus, after staining them with the M-phase marker phospho–histone 3 (pH3). Compared to WT, the number of mitotic cells remained unchanged in the injected thymus (Fig. 4F). However, the Ccr9b-deficient larvae showed an increased number of dividing cells in the thymus (Fig. 4F). This may be because mitotically quiescent thymocytes move to the thymic IZ and express ccr9b (20). Together, these results suggest that the Ccr9b-Ccl25a axis controls the intrathymic positioning of the cells, affecting the overall T cell lineage outcome.

Ccr9b-mediated intrathymic cell localization occurs before T cell commitment stage

Considering the role of Ccr9b-mediated cell localization in lineage outcome, we proposed that ccr9b expression may begin in thymocytes before T cell commitment. To test this hypothesis, we used the rag2^−/− larvae (fig. S4A), in which T cell development is blocked at this stage, and examined the ccr9b expression. Detection of ccr9b⁺ cells in the rag2^−/− thymus (fig. S4B) indicated that ccr9b expression starts before TCR rearrangement in thymocytes. We next examined whether ccr9b is regulated by Notch1 signaling during T cell specification. The medaka genome contains two genes encoding for notch1, but only notch1b is expressed in the larval thymus (12). First, the intracellular domain of notch1b (ICDN1b) was overexpressed in embryos using a heat-inducible promoter (fig. S5A), where ccr9b was found to be ectopically expressed in the ICDN1b-induced embryos (fig. S5, B and C). In the second approach, we used notch1b^−/− larvae that carry a stop codon in the third exon, leading to a truncated protein (Fig. 5A). WISH showed significantly reduced expression of ccr9b in the notch1b^−/− thymus (Fig. 5B), reflecting either a decrease in ccr9b expression or reduced number of ccr9b⁺ cells. Therefore, we incrossed notch1b^+/− fish carrying ccr9b:rfp and analyzed their progeny to distinguish between these possibilities. We found approximately a 4-fold decrease in RFP intensity (Fig. 5C and fig. S5, D and E) and a 10-fold reduction in the number of ccr9b⁺ cells (Fig. 5D) in the notch1b^−/− thymus compared with the WT siblings, which was probably due to reduced cell proliferation in the notch1b^−/− thymus (Fig. 5E). Further, we assessed the migratory behavior of ccr9b⁺ thymocytes through time-lapse in vivo imaging. The average speed and directionality of ccr9b⁺ cells were decreased approximately four- and twofold, respectively, compared with those in the WT siblings (Fig. 5, F and G, and movie S5). Hence, these results suggest that ccr9b expression starts at the precommitment stage and is regulated by Notch1b.

Fig. 5. Notch1b deficiency impairs T cell specification and the migratory behavior of thymocytes.

(A) Top: Schematic of the notch1b gene structure, indicating a single-nucleotide mutation from TAT to stop codon TAG in exon 3 (asterisk), which would yield a protein truncated after amino acid 245 of the 2488–amino acid predicted protein. Bottom: A representative gel image from PCR products of notch1b exon 3 digested with Bfa I restriction enzyme (L, 100-bp DNA ladder). (B) Expressions of ccr9b in WT and Notch1b^−/− larval thymuses. Scale bars, 40 μm. Dashed circle indicates the position of thymus. Arrow indicates the stained cells. (C) 3D rendering of WT and Notch1b^−/− carrying the ccr9b:rfp reporter. Scale bars, 30 μm. (D) Number of RFP⁺ cells in WT and Notch1b^−/− thymuses. N indicates the number of biological samples. (E) Number of pH3⁺ cells in the thymus. N indicates number of biological samples. (F and G) Average cell speed (F) and straightness (G) of RFP⁺ cells in WT and Notch1b^−/− larval thymuses carrying the ccr9b:rfp reporter. n indicates the total number of cells examined from >3 samples per genotype. (H) Expressions of tcrb and tcrg in WT and Notch1b^−/− thymuses. Dashed circles indicate the position of thymus. Arrows indicate stained cells in the notch1b mutant. Scale bars, 60 μm.

Notch1b is required for specification and differentiation of both lineages

We next investigated the role of Notch1b in T cell development. WISH showed that tcrb and tcrg expression levels were remarkably reduced in the notch1b^−/− thymus (Fig. 5H). Further, the expression of gata3, a transcription factor required for T cell specification (2), was reduced but not completely abolished (fig. S5F, top). However, bcl11b expression was absent in the notch1b^−/− thymus (fig. S5F, bottom), indicating that Notch1b acts upstream of bcl11b, which is consistent with previous findings in mice (28). Notch1b function in T cell development was confined to thymus because the expression of cmyb, a marker of hematopoietic stem and progenitor cells, was unchanged in the hematopoietic tissue of notch1b^−/− embryos (fig. S5G). The presence of thymocytes in the thymus of Notch1b-deficient larvae with ccr9a:gfp reporter supports this conclusion (fig. S5H). Most of the thymocytes were located predominantly in the thymic OZ (fig. S5I). Together, these results strongly suggest that Notch1b is required for T cell specification and commitment.

IL-7 in TECs triggers proliferation and development of γδ⁺ cells

αβ⁺ and γδ⁺ cell organization in two distinct thymic regions raised the question about the extent to which signals from the thymic niche could influence a given lineage. To address this, IL-7 was selected as a probe, based on four observations: First, it is supplied by TECs and promotes mammalian γδ T cell development (6, 29, 30). Second, the il7 gene has been identified in the genome of cartilaginous fish (31) and all other jawed vertebrates. Third, the lack of tcrd rearrangement in il7 receptor alpha (il7r)–deficient zebrafish (32) suggests the evolutionary conservation of IL-7 signaling in γδ T cell development. Fourth, we now show that in medaka, il7 and its receptor are expressed in cells strategically positioned at the thymic niche where TCRγδ⁺ cells are localized (Fig. 6A and fig. S1). To challenge the hypothesis that high levels of IL-7 are sufficient to trigger γδ T cell development, we manipulated the thymic cytokine level (Fig. 6B) by injecting the ccl25a:gfp-t2a-il7 construct into embryos, which strongly up-regulated the il7 level in the thymus (Fig. 6C, bottom inset). WISH of injected embryos showed enhancement of TCRγδ⁺ cells as compared to uninjected controls (Fig. 6, D and E). Ectopic TCRγδ⁺ cells were also seen in the thymic IZ (Fig. 6D, arrows), in the close vicinity of il7⁺ cells (Fig. 6F, arrows). Approximately a 1.6-fold increase in mitotic cells was observed in the injected embryos (Fig. 6, G and H), supporting the role of this cytokine in cell proliferation as was reported in mice (33). Similarly, overexpression of il7 in ccr9b⁺ thymocytes (fig. S6A) resulted in ectopic TCRγδ⁺ cells (fig. S6, B and C) and higher cell proliferation (fig. S6D). We also found thymocytes expressing il7 and tcrg in the thymic IZ (fig. S6C, arrows). Together, these findings confirm that il7 promotes cell proliferation and γδ T cell development.

Fig. 6. Overexpression of il7 in TECs enhanced the development of TCRγδ⁺ cells.

(A) Expressions of il7 and il7r in the larval thymus. Scale bars, 40 μm. (B) Top: The vector construct used to overexpress il7 in TECs. Bottom illustrates the experimental rationale. (C) Schematic outline (top) and image of an embryo injected with ccl25a:il7 construct (bottom). Inset shows il7 expression in the thymus of injected embryo. Arrowheads indicate autofluorescent pigments. (D) Expressions of tcrg, tcrd, tcra, and tcrb in WT and ccl25a:il7-injected embryos. Scale bars, 40 μm. Arrows indicate ectopic tcrg expression in the IZ of the thymus. (E) Left: Relative tcrg expression in the thymus compared to WT. Right: Relative thymus size normalized to the average of WT at 10 dpf. N indicates number of biological samples. (F) A representative image of a ccl25a:il7-injected embryo stained for il7 (green) and tcrg (magenta). Arrows indicate the close contact between il7⁺ and tcrg⁺ cells in the IZ of the thymus. OZ indicates the OZ of the thymus. Scale bars, 10 μm. (G) A representative image of a ccl25a:il7-injected embryo stained with GFP and pH3. Scale bars, 10 μm. (H) Number of pH3⁺ cells in the thymus. N indicates number of biological samples.

A cell-based in silico model suggests that the balance between environmental signal and cell location is critical for lineage outcome in a thymus

To better understand the process of T cell lineage decision, we created a spatial, cell-based model. A center-based framework (34, 35) was used to model discrete cells, moving in a three-dimensional slice of the thymus (Fig. 7A and fig. S7, A to C). On the basis of our experimental data and consistent with previous findings in murine models, TECs express Delta-like 4a (Dll4a), the Notch1b receptor ligand, and IL-7 in this model (Fig. 7, B and C). Each simulation starts with the entry of the progenitors that can respond to thymic signals through receptors into the thymus. After differentiation and proliferation stages (fig. S7D), they commit either to the αβ or γδ lineage, followed by purging by thymic selection or emigration (movie S6). To match the experimental conditions, we integrated spatial and temporal quantitative parameters, including cell proliferation (fig. S7, E and F) and motility (fig. S7G), derived from various time-lapse in toto imaging of transgenic reporter lines presented previously (20) and in this work (table S1 to S4). For other unknown parameters, qualitative rules were used (fig. S7, H and I; for details, see the Supplementary Materials).

Fig. 7. The interplay between environmental signals and intrathymic cell localization in an in silico model.

(A) A spatial cell-based model for the cortical region of a medaka larval thymus (movie S6). In the model, Notch (B) and IL-7 signaling (C) were applied. (D) Distribution of sublineages (top) and cell division (middle) according to IL-7 signaling (bottom) and IL-7R levels. Data shown are from three simulations. Proportions of sublineages (E) and cell proliferation (F) in scenarios a and b. (G) Extracellular distribution of IL-7 in the model. (H) Proportions of sublineages (left) and cell proliferation (right) in scenarios a and c as shown in (G). For both conditions, θ_IL-7 = 0.4, and IL-7R was randomly expressed. (I) Parameter scan over the Ccr9b-dependent free parameters shows how the proportions of sublineages vary. Conditions i to iv are shown in (J). Data shown are from two simulations. (J) Localization of ccr9b⁺ cells in the model for conditions i to iv are shown in (I). Each circle represents a cell. Data are projections of one time point from two simulations of each per condition. (K) Number of dividing cells under conditions i and iii from (I). Error bars in (E) and (H) (left) indicate 95% confidence interval, elsewhere SD.

The precommitment model proposes that the IL-7R level on progenitors influences the T cell lineage fate (1, 3). However, the extent to which the spatial distribution of IL-7 (29) contributes to the lineage decision is less clear. Therefore, we tested the interplay between IL-7R level on progenitors and the thymic IL-7 milieu in our in silico model. We set the strength of the IL-7 signal to favor γδ T cell differentiation. The free parameter θ_IL−7 was used as an IL-7 signaling activity threshold for commitment to the αβ lineage (Fig. 7, D and E). First, we compared the impact of two scenarios where progenitors before entry into the thymus had either a random (scenario a) or an equal (scenario b) amount of receptor (Fig. 7D). Consequently, scenario a models the precommitment hypothesis with the impact of spatial IL-7 distribution on lineage outcome, while scenario b considers only the impact of thymic IL-7 signal. Both scenarios were simulated with three different θ_IL−7 parameter sets, which resulted in an overall higher γδ lineage proportion (Fig. 7E) and cell proliferation (Fig. 7F) in scenario a compared with that in scenario b. This effect was consistent among all tested θ_IL−7 values, indicating that both scenarios could explain the lineage proportion observed in the in vivo experiments. As expected, most of the IL-7R^high cells differentiated into the γδ lineage; however, some of them also committed to the αβ lineage (Fig. 7D). This observation is consistent with the report that murine IL-7R^high progenitors have greater potential to develop into γδ T cells than IL-7R^neg-low progenitors (7). In addition, our model predicts that proliferation is increased when progenitors express different levels of IL-7R (Fig. 7D, middle). Next, to test the effect of cytokine milieu on lineage outcome, we compared two scenarios in which IL-7 was secreted either by a subset of TECs (scenario a) or by all TECs (scenario c). In scenario a, the extracellular IL-7 level generated a stable gradient within the thymic region (Fig. 7G, left, and fig. S8, A and B), while IL-7 was equally distributed in scenario c (Fig. 7G, right). The lineage proportion and cell proliferation in scenario a were closer to the in vivo situation (Fig. 7H), while scenario c mimicked the ccl25a:il7 experiment. Consistent with in vivo experiments, these simulations predict that the spatial distribution of IL-7 has an important role for thymocytes proliferation and lineage choice.

To challenge the effect of Ccr9b-mediated cell localization on lineage outcome, we integrated the tendency of the thymocytes expressing Ccr9b to migrate inward into the thymus in the model. We set two free parameters: The time of onset of Ccr9b expression was defined by t_ccr9b and the strength of directional migration by d_ccr9b. We then simulated 75 different combinations of θ_IL−7, t_ccr9b, and d_ccr9b parameters (Fig. 7I and fig. S8C). High variability in terms of lineage outcome was observed when θ_IL−7 = 0.4. Consistent with the situation in the Ccr9b-deficient thymus, a significantly higher proportion of γδ T cells was observed when d_ccr9b = 0, that is, thymocytes in the simulation displayed no directionality and distributed randomly within the thymus (Fig. 7, I and J, condition iii). Ectopic localization of thymocytes in the thymic OZ also increased IL-7 signaling and enhanced cell proliferation, particularly when t_ccr9b was low (Fig. 7K). However, this effect was not apparent when t_ccr9b = 1.0, which simulated a condition wherein the onset of ccr9b expression was immediately before the commitment stage. Under this condition, thymocytes probably did not have enough time to reenter the cell cycle. Low t_ccr9b increased the duration of Ccr9b expression, allowing thymocytes to “correct the course.” Conversely, with high d_ccr9b, cells moved with greater directionality deeper into the thymus, and consequently, the proportion of αβ lineage increased (Fig. 7I, condition iv), resembling the situation seen in the embryos injected with ccr9b:ccl25a. The condition most similar to the normal in vivo situation was when ccr9b expression started shortly before the commitment stage, where cells displayed an average directionality (Fig. 7, I to K, condition i). Thus, our in silico model demonstrates that to influence the T cell fate, Ccr9b expression must precede commitment, which we experimentally verified in the rag2^−/− larvae (fig. S4B).

Considering the central role of Notch signaling in T cell specification and commitment, we also modeled a scenario in which thymocytes exhibit slow differentiation and set the free parameter κ_diff = 0.3 (referred to as scenario d). Consistent with results from the notch1b mutants, both lineages (Fig. 8A) and the number of dividing cells (Fig. 8B) decreased significantly. By coupling differentiation to cell speed in the model, on the basis of our experimental data, we noticed a reduction in cell speed (Fig. 8C). Unexpectedly, in the simulation, we noted that a high proportion of undifferentiated thymocytes accumulate within the thymus (Fig. 8D), which was inconsistent with the situation in the notch1b mutant whose thymus size is reduced (fig. S5I). On the basis of these results, we assume that Notch1b also influences the process of thymus homing or the survival of thymocytes within the thymus. Notch has a role in progenitors before their thymic entry (36).

Fig. 8. Impaired Notch signaling affects T cell development in an in silico model.

(A) Distribution of T cell sublineages according to IL-7 signaling and IL-7R levels when differentiation was strongly reduced (κ_diff = 0.3), akin to the deficiency of notch1b (scenario d). Simulations were run at the same time as shown in Fig. 7C. (B and C) Number of dividing cells (B) and cell speed (C) in scenario d compared to condition i from Fig. 7H. (D) Render of a simulation with κ_diff = 0.3 at 24,000 simulation steps (100 hours).

DISCUSSION

The main focus of research on the mechanisms of divergence of the αβ/γδ lineage decision has been restricted to human and mouse, which are both γδ T cell low species (10). Here, in the first study of this process in γδ T cell high species, we combined experimental approaches with informatics and computational modeling to understand the mechanisms leading to the αβ/γδ lineage outcome. In a previous study, we provided quantitative data linking the migratory behavior of thymocytes to the stages of αβ T cell development (20), highlighting that the medaka thymus is a suitable in vivo model to study system-level regulatory mechanisms (13). Furthermore, we have shown that landmark events of T cell development including proliferation, T cell specification, recombination, and selection are spatially organized in the medaka larval thymus. Lymphoid progenitors enter through the anterior/ventral and posterior/ventral parts of the thymus to the thymic OZ, that is, in the region where they proliferate and specify as T cells (akin to the mouse cortex). The expression patterns of autoimmune regulator (aire) gene, a marker for mammalian medullary region, and the dynamic interaction between ccr9b⁺ thymocytes and cxcr3a⁺ dendritic cells strongly suggest that the process of thymic selection mainly occurs in the innermost part of the IZ and the dorsal side of the thymus (20). Note that the presence of thymic OZ and IZ appears to be specific to teleosts, as it was previously described in some other fish species (37, 38).

In medaka, T cell sublineages are not intermixed within the thymus, as they do not follow the same migratory paths during development. Upon entry, Ccr9a⁺ progenitors regulate their intrathymic localization through the expression of chemokine receptor Ccr9b. Different environmental signals that have an impact on their cell fate choice participate in this process. Contribution of chemokine receptors in lineage commitment is evolutionarily conserved, as the lack of CCR7 or CCR9, which typically guides the migration of thymocytes in opposing directions within the murine thymus (39, 40), leads to either reduced or enhanced γδ T cell development, respectively (41). How CCR9-deficient mice generate higher frequency of γδ T cells is not fully understood. It was speculated that intrathymic cell migration could influence the sensing of an IL-7 signal that supports γδ T cell development (41). In addition to spatial localization, the speed of migration of a cell can fine-tune the sensing outcome (42). Chemotaxis and migratory speed are two independent signaling modules that can be regulated by chemokine receptors (43, 44). The data presented here and in our previous study (20) also provide experimental support for the notion that cell migratory speed is coupled with ccr9b expression. At the T cell commitment stage, normal thymocytes can be divided into rag2⁺/ccr9b⁺ and rag2⁺/ccr9b⁻ subpopulations, in which the average migratory speed of the former subpopulation is higher than that of the latter. There is also a correlation between the level of ccr9b expression and the migratory speed of TCRαβ⁺ cells within the thymus (20). Reduced migratory speed of thymocytes in the Ccr9b-deficient thymus is almost comparable to that of WT ccr9b⁻ thymocytes located in the thymic OZ, suggesting that Ccr9b not only controls the intrathymic cell positioning but also regulates the speed of the thymocytes. Further supporting this notion, simulations of various scenarios in the in silico model showed that when thymocytes lost their directionality, αβ/γδ T cell lineage outcome was comparable to that in Ccr9b-deficient fish. In our simulations, the absence of Ccr9b-mediated homing to the thymic IZ contributed to three parts of the Ccr9b^−/− phenotype: (i) a high proportion of TCRγδ⁺ compared to TCRαβ⁺ cells, (ii) high-cell proliferation, and (iii) mislocalization of thymocytes. However, the in silico model could not explain the reduced RFP reporter level in the ccr9b^−/− fish thymus, indicating that Ccr9b signaling may not be directly involved in this process. It is likely that thymocytes that normally express high ccr9b levels undergo impaired development, similar to αβ T cells in ccr9b^−/− fish, down-regulating the ccr9b activity. One interesting additional aspect of chemokine receptor evolution revealed by our studies pertains to segregated roles of the paralogous ccr9a and ccr9b during T cell development. In mammals, the single-copy Ccr9 is involved in the processes of thymus homing (45) and positioning of thymocytes in the subcapsular zone (39). In teleosts, the former role appears to reside in the ccr9a paralog (20, 22), whereas the latter role has been assigned to the ccr9b paralog.

The impact of environmental signals in the thymus on lineage commitment is not completely understood. Contributions of Notch1 and TCR ligand–mediated signals in this process have been proposed (1). Comparative genome analysis suggests the emergence of PTCRA gene in the common ancestor of birds and mammals (46). Therefore, it is tempting to speculate that pre-TCR signaling might not occur in early jawed vertebrates, akin to pre-TCR–deficient mice (46, 47). In humans, Notch1 activation is more stringently required for the development of γδ than αβ T cell lineage (28), whereas in mice, moderate Notch signals are thought to synergize with pre-TCR signals to generate αβ T cell lineage (48). Studying the role of Notch signaling in lineage diversification is complicated by engagement of the Notch1 receptor by the environmental Notch ligand Dll4, in activation of T cell–specific developmental programs in progenitors (2) before lineage diversification. Here, we provide the first evidence that medaka Notch1b is indispensable for T cell specification. In addition, alterations in cell speed and directionality observed in the Notch1b-deficient thymus, together with the proof of Notch1b acting upstream of ccr9b, strongly suggest that the migratory behavior and differentiation of thymocytes are coupled. However, these findings raise the question of how ccr9b expression is switched on only in a subset of thymocytes. Notch1, Bcl11b, and Ccr9 form a gene regulatory network, required for αβ lineage development. In humans, NOTCH1 acts upstream of CCR9 (49) and BCL11b (28). NOTCH1 not only activates BCL11B expression during T cell lineage specification but also restricts its expression levels in TCRαβ⁺ cells through a microRNA-17–dependent negative feedback loop (28). In mice, Bcl11b, which is required for αβ but not for γδ T cell development (27), is regulated by Notch signaling and stochastic epigenetic events that take place on the Bcl11b locus (50). It is therefore possible that multiple mechanisms cooperatively regulate the ccr9b expression in a stochastic manner. Concerning αβ versus γδ lineage commitment, medaka Notch1b appears to be required for both lineages, and its ligand, dll4a, is uniformly expressed within the thymus (12). By contrast, IL-7 is an ancient molecule that plays a role in lineage decision. The emergence of IL-7 coincided with gnathostome evolution (31), and its expression is restricted to areas where immature thymocytes and TCRγδ⁺ cells reside in the murine thymus (29, 51), as in medaka. Previous models have shown that low-density cytokine-producing cells, together with cells expressing a high level of corresponding receptor, generate a local extracellular cytokine gradient (52). We incorporated such a local spatial IL-7 gradient in our in silico model, with the cytokine concentration around the secreting cells decaying sharply, across only a few cell diameters. Under this condition, when early thymocytes lose their ability to leave the IL-7–producing region, they receive sufficient stimuli that bias their development toward γδ lineage. This effect was independent of whether thymocytes express IL-7R differently, as in mice (7), or at the same level. However, the biologically more realistic scenario of differentially expressed IL-7R leads to higher average cell proliferation, as daughter cells in the model inherit their progenitor’s levels of expression. Thus, our model predicts that IL-7R^high thymocytes have a proliferative advantage and form large clones. This concurs with the findings that spontaneous overexpression of murine IL-7R has a selective advantage that contributes to proliferation and leukemogenesis (53, 54). Although IL-7 was a determining factor in lineage decision in our in silico model, we do not rule out the contribution of other nonredundant cytokines to the T cell lineage decision. When il7 was overexpressed in all TECs, TCRγδ⁺ cell expansion was predominantly found in the thymic OZ, indicating that the epithelial compartment in this region provides other unknown nonredundant cues required for γδ T cell development. Together, the in vivo experiments and in silico model suggest that the interplay between thymic signals, migratory behavior of thymocytes, and the onset of gene expression influences the T cell lineage outcome.

Overall, this study underlines the conservation of genetic pathways involving in T cell development, reveals that differences in cellular composition and environmental signals within the thymus influence the T cell lineage outcome, and explains how the number and frequency of γδ T cells vary between species. It is conceivable that a mechanism of this kind has been co-opted by convergent evolution in the γδ T cell high mammalian animals, including ruminants and pigs. The biological significance of high frequency of γδ T cells and their abundance at epithelial surfaces is still unclear. Given that these cells are responsible for sensing and responding immediately to bacterial, viral, and fungal infections without the need of antigen-presenting cells (55), it is reasonable to expect that the mechanism regulating their frequency is a species-specific immunological adaptation to the environment. Notwithstanding, the differentiation of other immune cells from hematopoietic stem cells in the bone marrow appears to be tightly regulated by intrinsic mechanisms and signals from the niche. Hence, our approach can be used to examine fundamental principles governing cell-fate choice and lineage commitment during hematopoiesis and in disease.

MATERIALS AND METHODS

Animals

Medaka husbandry were performed in accordance with the German animal welfare standards (Tierschutzgesetz §11, Abs. 1, Nr. 1, husbandry permit no. 35/9185.46/Uni TÜ and no. 35-9185.64/BH Wittbrodt). All experiments have been performed at stages before the legal onset of animal life. Generations of medaka transgenic reporter lines and mutants, as well as experimental protocols, were performed in accordance with relevant institutional and national guidelines and regulations and were approved by the Institutional Animal Care and Use Committee (IACUC nos. 2019-03-19ML and 35-9185.81/G-145/15).

Generation and genotyping of medaka mutants

The CRISPR-Cas9 technique was used to generate medaka ccr9b and rag2 knockout fish. Single guide RNAs (sgRNAs) for these genes were designed with CCTop website (56). Cloning and in vitro transcription of sgRNAs were performed as described previously (56). Briefly, oligos were annealed using 100 μM of each oligo and annealing buffer [10 mM tris-HCl (pH 7.5) and 30 mM NaCl] in a polymerase chain reaction (PCR) tube at high temperature. Oligos for sgRNAs were then ligated into the DR274 plasmid (Addgene). The resulting plasmid was then digested with Dra I restriction enzyme. The linearized DNA was used as a template for in vitro transcription using the MEGAscript T7 Transcription Kit (Life Technologies). For microinjection, 15 to 30 ng/μl of each gRNAs were mixed with Cas9 mRNA (150 ng/μl) and injected into the blastomere at one-cell stage embryos. Embryos were maintained at 28°C in an embryo-rearing medium (0.1% NaCl, 0.003% KCl, 0.004% CaCl₂ × 2 H₂O, and 0.016% MgSO₄ × 7H₂O). Adult ccr9b^−/− fish were crossed with the transgenic ccr9b:rfp animals (20), and their progeny at 2 to 3 months of age were incrossed to obtain ccr9b^−/− embryos carrying the reporter line. All ccr9b^−/− fish were viable and appeared normal till adulthood.

The medaka notch1b mutant was generated by target-selected mutagenesis approach as described previously (57). The region encompassing the third exon of medaka notch1b gene was screened for mutations using PCR and sequenced with 3730xl 96-capillary DNA analyzers (Applied Biosystems) using the reverse primer. The nonsense mutation in the exon 3 of medaka notch1b was detected as AGACCAGCTA (T > G) GAATGCTCCT. At least nine generations of notch1b mutant lines were backcrossed with Cab WT line as heterozygous fish to eliminate other possible background mutations in the genome. Genomic DNA was extracted using QuickExtract DNA extraction solution (Epicentro). As Notch1b^−/− larvae at 15 days post-fertilization (dpf) were lethal, our analysis was restricted to freshly hatched larvae at 10 dpf, which exhibited normal morphology. To develop Notch1b^−/− carrying various reporters for ccr9a:gfp (20), ccr9b:rfp (20), or ccl25a:rfp (this study), notch1b^+/− fish carrying reporter construct were identified by genotyping at 2 to 3 months of age and then incrossed. Each larva was genotyped after in vivo imaging or WISH analysis. In this study, both males and females were used.

Cloning and injection of expression constructs

To generate ccl25a:ccl25a and ccr9b:ccl25a constructs, a fragment containing GFP and full length of ccl25a complementary DNA (cDNA) separated by t2a, a short viral sequence, was cloned into vectors containing the upstream region of the ccl25a and ccr9b genes, respectively. To generate ccl25a:il7 and ccr9b:il7 constructs, a fragment containing GFP and full length of il7 cDNA separated by t2a was cloned into vectors containing the upstream region of the ccl25a and ccr9b genes. To overexpress notch1b, the intracellular domain from positions 5323 to 7443 nucleotides of full-length medaka notch1b cDNA was amplified by reverse transcription PCR (RT-PCR) and then cloned into the multicloning site of the heat-inducible promoter, pSGH2 (58), using restriction enzymes Bgl II and Xho I. For DNA injection, each construct at concentration (10 to 25 ng/μl), together with 1-μl I–Sce I meganuclease and NEB buffer (New England BioLabs), was coinjected into the blastomere at one-cell stage embryos. GFP signal was used to select positive embryos for WISH analysis.

Isolation and probe synthesis of target genes

Full-length medaka il7 and il7r transcripts, as well as partial fragments of tcrd, tcra, and foxp3, were amplified from a thymus or spleen cDNA library and cloned into the pCR-Blunt II-TOPO vector (Thermo Fisher Scientific). The identity of amplified PCR product was confirmed by sequencing. ClustalW alignment, neighbor-joining phylogenetic tree of protein sequences was deduced to determine the correct annotation of uncharacterized genes. For probe synthesis, plasmids were digested with a restriction enzyme, and the resulting linearized DNA was used as template for in vitro antisense probe synthesis using digoxigenin (DIG) or Fluorescein-RNA labeling mix kit (Thermo Fisher Scientific).

Whole-mount in situ hybridization

WISH was performed with antisense probes as described previously (12). Whole-mount FISH was carried out as described previously (20). Briefly, DIG- and fluorescein-labeled RNA antisense probes were hybridized to RNA in samples simultaneously. The fluorescein-labeled probe was detected first, with anti–fluorescein-POD antibody (1:500 dilution in blocking solution). It was revealed by staining with TSA Plus Fluorescein solution (Tyramide Signal Amplification Plus system) for 60 min in the dark. Samples were then treated with 1% H₂O₂ in methanol for 30 min. After stepwise rehydration and blocking, the DIG-labeled probe was detected by anti–DIG-POD antibody (1:1000 dilution in blocking solution) and revealed by the TSA Plus Cy3 solution. Samples were washed in PTW [1× phosphate-buffered saline (PBS) and 0.1% Tween 20] for several hours and photographed using an LSM 710 confocal microscope. To detect gene expression in the adult thymus, 4-week-old fish were fixed with 4% paraformaldehyde for 24 to 48 hours at 4°C and then washed in PBS. Samples were sectioned at a thickness of 15 μm using a microtome (Leica). Probes are listed in table S5.

Whole-mount immunostaining

Immunostaining was performed as described previously (59). Briefly, samples were fixed with 4% paraformaldehyde in 2× PBS and 0.1% Tween 20 for at least 24 hours at 4°C. For permeabilization, samples were treated with 100% acetone at −20°C for 20 min. After stepwise washing, they were incubated with blocking solution [10% fetal bovine serum (FBS), 0.8% Triton X-100, 1% bovine serum albumin (BSA), and PTW] at 4°C for 2 to 3 hours. For primary antibody incubation, rabbit anti–phosphohistone 3 antibody (Ser¹⁰, Millipore, 06-570; 1:500 dilution) was incubated with the sample in the incubation buffer (1% FBS, 0.8% Triton X-100, 1% BSA, and PTW) at 4°C in the dark for 3 days. As secondary antibody, the Cy3-donkey anti-rabbit immunoglobulin G (the Jackson laboratories, 711-165-152; 1:500 dilution) was used. Samples were incubated with the second antibody in the incubation buffer at 4°C in the dark for 2.5 days. To remove residual secondary antibody, samples were washed several times with PBS-TS (10% FBS and 1% Triton X-100 in PBS).

Live imaging of the entire thymus

Freshly hatched medaka larvae containing yolk sac were used for in vivo imaging as described previously (20). Briefly, time-lapse in vivo recording the entire larval thymus were carried out on a PerkinElmer Ultraview VoX or Ultraview ERS Spinning disk confocal using a 40× water-immersion objective (LD C-Apochromat, 1.1 numerical aperture, Corr, Zeiss). For accurate tracking of individual cells, z-stacks of 60- to 70-μm spanning the whole thymus area (z-space, 1 μm) were imaged with a time interval of <15 s.

Four-dimensional data analysis

Stacks of images were analyzed by Imaris Bitplane software as described previously (20). Briefly, the built-in spot detection algorithm (spot diameter, 3 μm; MaxDistance, 4 μm; and MaxGapSize, 3 μm) was used to identify the x, y, and z coordinates for each cell at each given time point. The Imaris software was used to calculate parameters such as average cell speed and directionality (displacement divided by path length). The obtained data were exported into the Excel file, and GraphPad Prism software (version 9) was used for graphing and statistical analysis.

Quantitative RT-PCR

Total RNA from the spleen and kidney was extracted using TRIzol (Life Technologies) following the manufacturer’s protocol. RNA samples were then treated with RQ1 ribonuclease-free deoxyribonuclease (Promega) before first-strand cDNA synthesis with random hexamer primers and SuperScript III Reverse Transcriptase (Thermo Fisher Scientific) following the manufacturer’s protocol. Quantitative PCR was carried out using the SYBR Green Kit (Applied Biosystems) on the LightCycler 480 (Roche). The data were analyzed in Microsoft Excel using the ΔC_t method with ef1a as a reference gene for normalization. Primers are listed in table S6.

Statistical analysis

Statistical details including the number of biological replicates (N) and P values are detailed in figure legends. Data in bar graphs are shown as an absolute number with means ± SD noted. Wilcoxon-Mann-Whitney test was used to calculate significant differences where indicated. P < 0.05 was considered statistically significant. All statistical analysis was carried out using Prism 9 (GraphPad).

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/7/29/eabg3613/DC1

View/request a protocol for this paper from Bio-protocol.