To analyze γδ T cell development by flow cytometry, we used a panel of antibodies that distinguish developmental stages and lineage fate choices in the mouse fetal thymus. γδ T cells develop from DN2/3 cells in the thymus from precursors with both αβ and γδ T cell potential, with upregulation of CD8 (immature single positive [ISP]) and CD4 marking commitment to the αβ-T cell lineage (Figure 1A). γδTe (early γδ T cell subsets prior to maturation; CD24⁺) cells arising from DN2/3 cells express γδTCR and CD24 on the cell surface but retain the ability to become either γδ T cells or αβ T cells (dotted arrow) (Coffey et al., 2014). Strong γδTCR signaling results in upregulation of CD73 (Nt5e) at the γδT1p stage (p=progenitor), followed by downregulation of CD24 (Cd24a) to yield mature CD27⁺CD24^-CD73⁺CD44^-γδT1 (IFNγ-producing) cells. Signaling through Vγ6 or Vγ4 TCRs results in downregulation of CD24 and upregulation of CD44 as γδ T cell precursors differentiate into immature γδT17p cells, which mature into CD27^-CD24^-CD73^-CD44⁺γδT17 (IL-17-producing) cells (Sumaria et al., 2017). CD27 is expressed on all immature γδ T cells and stays on in mature γδT1 cells but is downregulated during γδT17 maturation. Thus, detection of γδTCR, CD4, CD8, CD24, CD73, CD27, and Vγ chains, and the genes that encode them, provides a solid framework for analyzing the impact of gene perturbations on fetal γδ T cell development and maturation.

It is important to clarify the Vγ and Vδ chain nomenclature used in this study (Figure 1B). Cell surface proteins detected in flow cytometry assays were classified using the Tonegawa nomenclature (Heilig and Tonegawa, 1986). When referring to genes or gene transcripts, the International Immunogenetics Information System (IMGT) nomenclature (Lefranc, 2003) was used, except in scRNA-seq figures where R-Seurat program output gene names from the Mouse Genome Informatics (MGI) site were preserved (Baldarelli et al., 2024). The numbering is the same in all three systems except for Vδ1, which is equivalent to TRDV4 (IMTG) and Trdv4 (MGI).

To investigate how HEB loss affects fetal γδ T cell development, we crossed Tcf12^fl/fl mice with Tcf12^fl/fl Vav-iCre mice to generate littermates without Cre (wild-type [WT]) or with Cre (HEB cKO). HEB cKO mice lack HEB in all hematopoietic cells, as previously described (In et al., 2017; Welinder et al., 2011). At E18, HEB cKO thymocytes showed reduced cellularity (Figure 1C) and a developmental block at the ISP-to-DP transition (Figure 1—figure supplement 1), as previously reported in adult HEB-deficient mice (Barndt et al., 1999; Wojciechowski et al., 2007). Although γδ T cells were proportionally increased in HEB cKO mice (Figure 1D and E), their absolute numbers were comparable to WT mice (Figure 1F), indicating that the reduced cellularity was primarily due to the loss of DP thymocytes.

We next analyzed Vγ chain subsets. Vγ6 cells were identified using an exclusion strategy as cells lacking expression of Vγ1, Vγ4, and Vγ5 (Figure 1G and H). It should be noted that the flow cytometry plots show the percentage of Vγ5⁺ and Vγ5^- (Vγ6) cells out of the Vγ1^-Vγ4^- population (Figure 1H), whereas the bar graph shows the percentage of Vγ6 cells out of all γδTCR⁺ cells (Figure 1I). HEB cKO mice exhibited a marked reduction in the frequency (Figure 1G and I) and absolute numbers (Figure 1J) of Vγ4⁺ cells, and a corresponding increase in the other Vγ subsets (Figure 1G, H, I, and J). Immature (CD24⁺) cells were more prevalent across all Vγ subsets in HEB cKO mice, except for Vγ4⁺ cells, which were primarily CD24⁺ and thus had not yet matured, in either WT or HEB cKO mice (Figure 1K). Furthermore, fetal HEB cKO γδ T cells showed a major impairment in IL-17 production in response to PMA/ionomycin stimulation (Figure 1L), consistent with our prior fetal thymic organ culture results (In et al., 2017). These findings indicate that even in the intact E18 thymic environment, HEB is essential for γδT17 cell development.

To assess the transcriptomic impact of HEB deficiency, we performed scRNA-seq on γδTCR⁺ thymocytes sorted from E18 WT and HEB cKO littermates. After quality control, we merged the datasets, excluded myeloid cells, and regressed out cell cycle genes using Seurat (Hao et al., 2021). Our analysis identified eight distinct clusters (numbered 0–7) visualized as a UMAP (Figure 2A). WT cells were enriched in clusters 2 and 4, while HEB cKO cells dominated clusters 0, 1, and 3, whereas clusters 5–7 showed a more balanced distribution (Figure 2B and C). To annotate the clusters, we curated 90 γδ T cell subset-defining genes from prior studies (Inácio et al., 2025; Liu et al., 2020; Mistri et al., 2024; Narayan et al., 2012; Pokrovskii et al., 2020; Spidale et al., 2018; Yang et al., 2023; Figure 2—source data 1). The top 10 differentially expressed genes were used to generate a clustered dot plot (Figure 2D), which we used to identify each cluster. We classified clusters 2 and 6 as early γδT cells (randomly assigned as γδTe1 and γδTe2), clusters 1 and 4 as γδT17 progenitors (γδT17p), cluster 0 as γδT1 progenitors (γδT1p), cluster 7 as mature γδT1 cells (γδT1), cluster 5 as mature γδT17 cells (γδT17), and cluster 3 as αβ lineage-like cells (αβT).

Quantification of WT and HEB cKO cells per cluster (Figure 2E) revealed that the γδT17p clusters segregated by genotype into WT (γδT17pw) and HEB cKO (γδT17pk) cells. HEB cKO cells were also over-represented in the αβT cluster and under-represented in the γδTe1 cluster. An unbiased heatmap of the top 10 most differentially expressed genes across all clusters further validated our assigned identities and revealed additional genes associated with these subsets (Figure 2F). These findings suggest that the loss of HEB impairs the early γδ T cell program and allows diversion toward the αβ T cell lineage.

We next confirmed that WT and HEB cKO cells within each cluster represented equivalent developmental subsets (Figure 2—source data 1). All immature γδ T cell clusters expressed Cd24a and Cd27, while mature subsets lacked Cd24a (Figure 2—source data 1). Among mature cells, only γδT1 cells expressed Cd27 and Nt5e (CD73), whereas γδT17 cells were uniquely Cd44-positive, consistent with our assignments. We also analyzed TRGV and TRDV gene expression (Figure 2—figure supplement 1B and C). In WT cells, TRGV4 and TRDV5 were co-expressed in γδTe and γδT17p populations but were greatly diminished in HEB cKO cells. TRGV6 and TRGV5 transcripts were detectable in γδTe subsets from HEB cKO mice, consistent with a delay in Vγ6 and Vγ5 γδ T cell maturation, whereas TRDV4 was expressed more broadly and at higher levels in all immature HEB cKO γδ T cell subsets relative to WT counterparts. These results indicate that HEB plays an important role in maintaining the subset specificity and magnitude of TRGV5 and TRDV4 expression during γδ T cell development.

To further define differences between WT and HEB cKO cells at the transcriptomic level, we constructed gene modules composed of suites of signature genes for the γδTe1, γδTe2, αβT, γδT17p, γδT17, γδT1p, and γδT1 subsets, derived from the analysis of merged WT and HEB cKO cells (Figure 2D and F; see Materials and methods). These modules were used to assign scores in WT (left) and HEB cKO (right) cells, which were visualized using split dot plots.

This analysis revealed a pronounced loss of the γδTe1 gene signature in HEB cKO cells (Figure 3A), alongside a notable increase in the αβT signature (Figure 3B). The γδTe2, γδT17p, and γδT17 gene signatures were also diminished in HEB cKO cells relative to WT, whereas the γδT1p signature was enriched. We also examined the expression of individual genes that were diagnostic for γδ T cell subsets (Figure 3C) versus committed αβ T cells (Figure 3D). In the γδTe1 and γδTe2 subsets of HEB cKO mice, Sox13 and Etv5 were reduced and Cd8b1 and Dgkeos were elevated relative to their WT counterparts. In γδT17p cells, αβT lineage genes were not detectable, but Sox13 and Etv5 were still lower in HEB cKO cells than WT cells, potentially decoupling the specification and commitment events. A less dramatic reduction in Il1r1 was observed in HEB cKO γδT17 cells (Figure 3E), and the levels of Nrgn and Eomes were similar in γδT1 lineage cells (Figure 3F). These results indicated that the impact of HEB deficiency at the transcriptomic level was more pronounced in early γδ T cell subsets than in mature γδ T cells.

The γδT17p subset segregated into distinct clusters in WT and HEB cKO cells, highlighting a critical stage of HEB-dependent regulation. To explore this further, we performed an unbiased differential gene expression analysis in WT cells versus HEB cKO cells and visualized the results using an enhanced volcano plot with stringent thresholds (log₂FC>0.5, –log₁₀p>25) (Figure 4A). Many of the top differentially expressed genes were signature markers of γδT17 differentiation, including Blk, Sox13, and Etv5, all of which were downregulated in HEB cKO cells. Trdv4 was among the most upregulated genes in HEB cKO γδT17p cells.

We next generated a list of differentially expressed genes between WT and HEB cKO γδT17p cells using less stringent criteria (adj p-value>0.001, log₂FC >0.25) (Figure 4—source data 1). This list was subjected to gene ontology analysis using ShinyGO, which resulted in a list of Kyoto Encyclopedia of Genes and Genomics (KEGG) pathways with reduced gene representation in HEB cKO γδT17p cells. We observed significant depletion of gene pathways related to TCR signaling (PD-1 checkpoint, NF-κB signaling) and Th1, Th2, and Th17 differentiation in HEB cKO cells relative to WT cells (Figure 4—source data 2). Three key markers of TCR signal strength, Cd5, Cd69, and Egr1, were reduced in both γδTe2 and γδT17p subsets in HEB cKO mice (Figure 4C). Id3 expression was markedly decreased in HEB cKO γδTe cells and nearly absent in γδT17p cells but remained intact in γδT1p and γδT1 cells (Figure 4D and E). We also examined the expression of Maf and Rorc, two key regulators of γδT17 maturation, and found that they were expressed similarly between HEB cKO and WT cells in each cluster (Figure 4F). Therefore, HEB modulates TCR signaling during γδT17 specification but appears to be largely dispensable for the expression of γδT17 cell maturation factors.

To further elucidate the landscape of E protein and Id protein expression during γδ T cell development, we examined the expression of Tcf12 (encodes HEB), Tcf3 (encodes E2A), Id3, and Id2 in each cluster (Figure 4—figure supplement 1). Id1 and Id4 transcripts were not detectable in any of our datasets. It should be noted that the Tcf12 deletion occurs in one of the last exons of a 200 kb gene locus, and although the protein is absent (Barndt et al., 1999), some mRNA expression can still be detected. In WT mice, γδTe cells co-expressed Tcf12, Tcf3, and Id3 (Figure 4—figure supplement 1A–C, E), consistent with dynamic regulation of E protein-dependent target gene expression during the αβ/γδ fate choice (Murre, 2019). γδT17p cells co-expressed Tcf12 and Tcf3, whereas γδT1 cells co-expressed Tcf3 and Id3 (Figure 4—figure supplement 1B and D). Id2 expression was restricted to mature γδT17 and mature γδT1 cells in both WT and HEB cKO mice. Tcf3 expression was also unaffected by a lack of HEB, indicating that Tcf3 and Id2 expression are HEB-independent in this context (Figure 4—figure supplement 1E and F). Id3 expression was disrupted in HEB cKO cells in γδTe and γδT17 cells but not in γδT1p cells (Figure 4—figure supplement 1C and D). This analysis reveals that Id3 expression was disrupted only in cells that normally express Tcf12.

Given the clear dependence of Id3 expression on HEB in specific γδ T cell subsets, we next investigated how loss of Id3 itself affects γδ T cell development using E18 fetal thymocytes from Id3 knockout (Id3-KO) mice. Total thymic cellularity in Id3-KO mice was comparable to WT controls (Figure 5A), but the proportion of γδ T cells among total thymocytes was reduced (Figure 5B and C). Analysis of Vγ chain usage revealed no major differences between WT and Id3-KO mice, although there was a slight increase in the proportion of Vγ6 cells in Id3-KO γδ T cells (Figure 5E and F). Interestingly, very few Id3-KO γδ T cells expressed CD73 (encoded by Nt5e) (Figure 5G and H), including Vγ5 and Vγ1 cells (Figure 5—figure supplement 1). Nt5e expression is upregulated in response to strong TCR signaling during γδ T cell development, and it can also be induced in response to short-term stimulation (Buus et al., 2016; Coffey et al., 2014). To determine whether the deficiency in CD73 expression in Id3-KO mice reflected an expansion of CD73^- cells or a failure to induce Nt5e, we cultured E18 fetal thymocytes from WT and Id3-KO mice with PMA/ionomycin (P/I) for 4 hr and measured CD73 expression in γδ T cells by flow cytometry (Figure 5I and J). While CD73 was robustly induced in a substantial fraction of CD27⁺ γδ T cells from WT mice, it remained nearly undetectable in Id3-KO γδ T cells, indicating a direct role for Id3 in the regulation of Nt5e expression.

Mature CD73^- γδ T cells typically exhibit a bias toward the γδT17 lineage (In et al., 2017). We hypothesized that loss of Id3 might re-direct γδT1-fated cells into the γδT17 lineage. However, upon stimulation, Id3-KO γδ T cells failed to produce IL-17 (Figure 5K and L), indicating that Id3 deficiency does not promote γδT17 differentiation. These findings highlight a critical role for Id3 in the functional maturation of both CD73⁺ and CD73^- γδ T cells.

To further investigate how Id3 deficiency affects γδ T cell development, we measured intracellular expression of PLZF (encoded by Zbtb16) and MAF in E18 γδ T cells from WT and Id3-KO mice by flow cytometry (Figure 6). PLZF is expressed in innate fetal/neonatal γδ T cells and adult iNKT and IL-4-producing γδ T cells (Alonzo et al., 2010; Kreslavsky et al., 2009; Lu et al., 2015). To compare developmental stages within Vγ subsets between WT and Id3-KO γδ T cells, we gated on: (1) immature (CD24⁺) Vγ4 cells, which comprised the majority of Vγ4 cells, (2) immature (CD24⁺) Vγ6 cells, and (3) mature (CD24^-) Vγ6 cells (Figure 6A). This analysis yielded three populations, based on PLZF and MAF expression: PLZF⁺MAF⁺, PLZF⁺MAF^–, and PLZF^–MAF^– cells (Figure 6A and B).

In WT mice, most immature Vγ4 and Vγ6 cells expressed PLZF, with about half of the PLZF⁺ cells co-expressing MAF. Nearly all mature Vγ6 cells co-expressed PLZF and MAF. Immature Vγ4 and Vγ6 subsets in Id3-KO mice had reduced proportions of PLZF⁺MAF⁺ cells and increased percentages of PLZF^-MAF^- cells relative to WT. Mature Vγ6 cells in Id3-KO mice exhibited a more severe disruption, with only ~50% co-expressing PLZF and MAF, and significant increases within the PLZF⁺MAF^- or PLZF^-MAF^- subsets. Moreover, mean fluorescence intensity analysis revealed substantially lower PLZF protein levels in Id3-KO cells across all subsets, especially immature cells, suggesting that this defect is not solely due to delayed maturation (Figure 6C).

To investigate population dynamics and gene expression changes in Id3-KO fetal thymocytes, we performed scRNA-seq on CD4^-CD8^- E18 fetal thymocytes. This strategy was designed to capture γδ T-biased cells with low surface γδTCR that might be missed by flow cytometric cell sorting. After quality control, WT and Id3-KO datasets were merged and subjected to dimensionality reduction, which identified 11 clusters (0–10) (Figure 6—figure supplement 1A). Violin plots of lineage and subset-specific genes identified two γδ T cell clusters (Figure 6—figure supplement 1B), which were computationally isolated and re-clustered. This process yielded four new γδ T cell clusters (0–3), which we identified as γδTe, γδT17p, γδT17, and γδT1 cells (Figure 6D and E), using the same strategy shown in Figure 2 (Figure 6G). This scRNA-seq dataset lacked a γδ/αβ T lineage cluster, likely due to low cell numbers and/or exclusion of CD4⁺ and CD8⁺ cells in the enrichment strategy. However, flow cytometry of E18 fetal thymocytes revealed a significantly higher proportion of TCRγδ⁺ cells co-expressing CD4 and CD8 in Id3-KO mice than in WT mice, indicating a bias toward the αβ T cell program (Figure 6—figure supplement 2). Notably, our scRNA-seq analysis showed that γδ T cell specification genes (Sox13, Etv5) and Tcf12 were unaffected by the decrease in Id3 (Figure 6H). These results decouple the loss of Id3 from the disruption of early γδ T cell regulators in HEB cKO mice.

Compared to WT, the Id3-KO mice had increased numbers of γδT17p cells and fewer mature γδT17 cells (Figure 6F), consistent with a partial block in γδT17 maturation. Id2 expression was markedly increased in all Id3-KO γδ T cell subsets (Figure 6I), consistent with a previously reported compensatory role (Zhang et al., 2014). Maf and Zbtb16 transcripts were markedly reduced in Id3-KO cells at the γδTe stage but recovered to near WT levels in mature γδT17 cells (Figure 6I). We noted a similar impact on expression of Rora in Id3-KO mice. Rora is not required for γδT17 cell function (Barros-Martins et al., 2016), but does play roles in the development of Th17 (Hall et al., 2022) and ILC3 cells (Lo et al., 2016). We also examined expression of Rorc (encodes RORγt) (Figure 6—figure supplement 1). Given the decrease in other genes involved in γδT17 cell maturation and the dependence of Rorc expression on MAF in γδT17 cells (Zuberbuehler et al., 2019), we were surprised to find that Rorc was higher in Id3-KO γδT17p cells than in WT γδT17p cells (Figure 6—figure supplement 1D) and was also elevated in the Cd8a-expressing DN4 subset (Figure 6—figure supplement 1E). Although RORγt is a critical regulator of γδT17 development and function, it is also upregulated after β-selection and maintains DP cell survival (Xi et al., 2006), suggesting that the increase we observed in the γδT17p cells could in part be due to αβ-T cell lineage diversion.

Although our data showed that HEB is necessary for Id3 expression during γδT17 development, it remained unclear whether HEB directly regulates Id3. The Id3 gene locus is composed of three exons adjacent to the long non-coding RNA gene Gm42329, which lies in the opposite orientation (Figure 7A). To assess direct binding of HEB and/or E2A to the Id3 locus, we analyzed ChIP-seq datasets from Rag2^-/- DN3 thymocytes (Fahl et al., 2021). This analysis identified three major binding regions for both HEB and E2A upstream of the first exon of Id3, each containing multiple peaks. We also examined ChIP-seq datasets from DN3 cells stimulated with anti-CD3 or anti-TCRβ to mimic TCR signaling (Pekowska et al., 2011; Seiler et al., 2012). RNA polymerase II bound the Id3 promoter in Rag2^-/- mice stimulated with anti-CD3ε, revealing active Id3 transcription in cells that had experienced CD3-mediated signaling. Additionally, we observed binding of Egr2 to two sites that overlapped with the HEB/E2A bound regions in total thymocytes from mice that had been injected with anti-TCRβ (Seiler et al., 2012). Notably, H3K27me3 repressive marks were detectable across the Gm42329 locus but were absent from Id3, indicating that Id3 is epigenetically poised for activation before pre-TCR and/or γδTCR signaling occurs.

Given that the data from the Egr2 and HEB/E2A ChIP-seq analyses were generated using different thymocyte subsets, we analyzed additional ChIP-seq data for HEB and E2A binding in Rag2^-/- DN3 cells that had been transduced with retroviral constructs encoding the KN6γδTCR and cultured with stroma expressing the weak KN6 ligand T10 for 4 days (Fahl et al., 2021). These Rag2^-/- DN3 + γδTCR cells allowed us to examine HEB/E2A binding to sites in the Id3 locus in a cellular context more closely aligned to the Egr2 binding assay (Figure 7—figure supplement 1A). This analysis revealed that the binding of HEB/E2A on those sites persisted after weak γδTCR signaling, strengthening the likelihood that concurrent binding of HEB/E2A and Egr2 occurs during this developmental transition. We noted that HEB/E2A binding was slightly dampened in Rag2^-/- DN3 + γδTCR cells relative to Rag2^-/- DN3 cells, consistent with the induction of Id3 and subsequent Id3-mediated disruption of E protein binding.

We designated the regions of overlapping HEB, E2A, and Egr2 binding as Tcf12^HE1 and Tcf12^HE2. Sequence-level analysis of Tcf12^HE1 and Tcf12^HE2 allowed the identification of predicted E protein binding sites (E box) in close proximity to Egr binding sites (Egr) (Figure 7—figure supplement 1B), suggesting that HEB/E2A and Egr2 may participate in a gene regulatory protein complex. Together, these findings support the hypothesis that HEB and E2A work in concert with Egr2 to modulate Id3 transcription in thymocytes undergoing TCR-mediated selection.

To examine how the chromatin landscape of the Id3 locus might change across this transition, we generated ATAC-seq data from DN3 and DN4 cells, as well as Rag2^-/- DN3 cells, to provide a genuine pre-selection context (Figure 7—figure supplement 1A). Alignment of ATAC-seq and ChIP-seq peaks in the Id3 locus revealed accessibility of Tcf12^HE1 and Tcf12^HE2 in Rag2^-/-, WT DN3, and WT DN4 cells, strengthening their relevance. Given the known ability of E2A and HEB to induce chromatin remodeling (Emmanuel et al., 2018; Lin et al., 2010), we also examined accessibility in DN3 and DN4 cells from HEB cKO mice (Figure 7—figure supplement 1A). Alignment of ATAC-seq and ChIP-seq peaks in the Id3 locus revealed accessibility of Tcf12^HE1 and Tcf12^HE2 in Rag2^-/- DN3, WT DN3, and WT DN4 cells. However, accessibility of Tcf12^HE1 and Tcf12^HE2 was dampened in HEB cKO cells, especially at the DN3 stage, suggesting that HEB may be involved in remodeling the Id3 locus, resulting in a poised state that enables TCR-dependent transcription factors like Egr2 to induce Id3 proportionally to TCR signal strength.

To test whether HEB can cooperate with TCR/CD3 signaling to upregulate Id3, we used a gain-of-function approach. We took advantage of the SCID.adh cell line, which expresses a chimeric hIL-2Rα:CD3ε receptor that mimics pre-TCR signaling when stimulated with anti-TAC (hIL-2Rα) antibody (Carleton et al., 1999). Stimulation results in downregulation of CD25, Rag1, Rag2, and Ptcra, while inducing Trac germline transcripts, recapitulating pre-TCR activity. We transduced SCID.adh cells with either control or HEB-expressing retroviruses to generate control and HEB-overexpressing cells (Figure 7B). To avoid the growth arrest and cell death associated with full-length HEB (HEBCan) overexpression (Engel and Murre, 2004; Wang et al., 2010), we used a construct encoding HEBAlt, a truncated form that activates E protein target genes without impairing cell viability (Wang et al., 2010; Wang et al., 2006; Yoganathan et al., 2022).

Upon overnight stimulation with anti-TAC, both control and HEB-expressing cells downregulated CD25 to similar degrees, as assessed by flow cytometry, confirming effective CD3-mediated signaling (Figure 7C and E). Quantitative RT-PCR (qRT-PCR) analysis showed that Id3 expression was modestly elevated in unstimulated HEB-expressing cells compared to controls, and stimulation of control cells upregulated Id3 levels to a greater degree (Figure 7D). Notably, Id3 expression in stimulated cells expressing HEB was much higher than either HEB or stimulation alone, and exceeded the sum of these two conditions, indicating a synergistic interaction. These results demonstrate that HEB can amplify Id3 induction in response to TCR signaling, potentially through cooperative interaction with TCR-induced factors such as Egr2.

The overall goal of this study was to understand how HEB transcription factors regulate the transcriptional networks required for the development of IL-17-producing γδ T cells. The subjects of these studies were genetically modified mice. The experimental variable was the genotype of the mice. For a given genotype, experimental mice were chosen randomly based on the availability of animals and littermate controls. Mice of both sexes were used. No sex differences were apparent; therefore, data were pooled. Experiments were done two to three times, depending on the number of mice available for each experiment, and results were pooled for statistical analysis for a total n of ≥3. Each biological replicate (mouse) is depicted as an open circle on the bar graphs. The investigators were not blinded to allocation during experiments and outcome assessment, except when fetal thymocytes were analyzed by flow cytometry prior to genotyping. Groups were unpaired, with similar variances, and all comparisons between genotypes were made within a given subset. Therefore, we assessed the significance of each comparison of mean values using an unpaired two-tailed classic Student’s t-test, with p<0.05 considered statistically significant. Error bars represent standard error of the mean (SEM) values.

Genetically modified mice were used in this study. Tcf12^fl/fl mice have loxP sites flanking the helix-loop-helix dimerization domain, which is shared by all HEB isoforms (Wang et al., 2006; Wojciechowski et al., 2007). These mice were bred to Vav-iCre transgenic mice (JAX stock #008610) (de Boer et al., 2003), which deletes regions flanked by loxP sites in all hematopoietic cells (Siegemund et al., 2015), to generate Tcf12^fl/fl Vav-iCre (HEB cKO) mice, as previously described (In et al., 2017; Welinder et al., 2011). All experimental HEB cKO mice were obtained from timed matings of Tcf12^fl/fl x Tcf12^fl/fl Vav-iCre mice, enabling parallel analysis with WT (no Cre) littermates. Id3-deficient mice (Id3-KO) lacked Id3 in all cells due to a knock-in/knockout allele in which the Id3 coding sequence was replaced with red fluorescent proteins (RFP) (JAX strain # 010983, B6;129S-Id3^tm1Pzg/J). RFP was not detected in our analyses due to quenching during intracellular staining. Id3-KO mice were acquired on the E29/B6 background and bred onto the C57Bl/6 background for 8 generations. Id3^+/- mice were timed mated together to produce WT and Id3-KO littermates. Rag2^-/- mice were obtained from Jackson Labs (JAX strain # 008449, B6.Cg-Rag2tm1.1Cgn/J). Fetal thymocytes were analyzed at E18, and adult mice were analyzed at 6–9 weeks of age. Mice were bred and maintained in the Comparative Research Facility of the Sunnybrook Research Institute (Toronto, ON, Canada) under specific pathogen-free conditions. All animal procedures were approved by the Sunnybrook Research Institute Animal Care Committee under animal user protocol (AUP) 22204 and in compliance with the Canadian Council of Animal Care guidelines.

Timed matings were performed by setting up mating pairs (day 0) and separating them ~16 hr later. Embryos were harvested after 18 days (E18). Fetal thymuses were dissected and individually pressed through a 40 µm mesh with a syringe plunger to create single-cell suspensions. Cells were resuspended in 1× HBSS/BSA for sorting or flow cytometry. Embryos were processed and analyzed separately, and tail tissue was collected for genotyping. Genotyping was performed as previously described for HEB cKO mice (Wojciechowski et al., 2007). Id3-KO mice were genotyped using the protocol described for the B6;129S-Id3^tm1Pzg/J strain on the Jackson Laboratories website.

Antibodies were purchased from eBiosciences (San Diego, CA, USA), BioLegend (San Diego, CA, USA), and BD Biosciences (Mississauga, ON, Canada). For flow cytometry, cells were washed and incubated with Fc blocking antibody (BD Biosciences), followed by extracellular staining for surface CD4 (clone GK1.5), CD8α (clone 53–6.7), CD3 (clone 145-2C11), TCRγδ (clone GL3), Vγ4 (Vγ2; clone UC3-10A6), Vγ5 (Vγ3; clone 536), Vγ1 (Vγ1.1; clone 2.11), CD27 (clone LG3-1A10), CD24 (clone M1/69), CD73 (clone TY11.8), and CD25 (clone PC615). To assess functional capacities, cells were stimulated by incubation for 4 hr with PMA (50 ng/ml) and ionomycin (500 ng/ml) in the presence of Brefeldin A (5 mg/ml; eBioscience), washed with 1× HBSS/BSA, and incubated with Fc block before staining for surface epitopes. Cells were then fixed and permeabilized (Fix and Perm Cell Permeabilization Kit; eBioscience) and stained with antibodies against IL-17A (clone eBio17B7). For intracellular transcription factor staining, the cells were fixed, permeabilized (FoxP3 Staining Kit, eBioscience), and stained for MAF (clone sym0F1) and PLZF (clone R17-809). All flow cytometric analyses were performed using Becton-Dickinson (BD) LSRII, Fortessa, or SymphonyA5-SE cytometers. FACSDiva and FlowJo software were used for analysis. Sorting was performed on BD ARIA and BD Fusion sorters.

SCID.adh cells, which have been engineered to express a surface human IL-2Ra (TAC):CD3epsilon chimeric signaling molecule on the surface, were cultured as previously described (Anderson et al., 2002). SCID.adh cell line identity was confirmed by flow cytometry phenotyping of CD44 and CD25 expression and response to stimulation with downregulation of CD25, and the cells tested negative for mycoplasma. SCID.adh cells were transduced with MIGR1 encoding GFP only (control) or MIGR1 encoding GFP and HEBAlt. GFP⁺ cells were sorted, expanded, and cultured overnight on plates coated with anti-TAC antibody at 5 µg/ml in 500 µl of PBS or PBS only as control. Cells were analyzed by flow cytometry for expression of CD25 and for mRNA expression of Id3 by qRT-PCR.

Total RNA was extracted from cells using TRIzol Reagent (Invitrogen) and reverse-transcribed into complementary DNA using Superscript III (Invitrogen). Reactions for qRT-PCR were prepared using PowerUp SYBR Green Master Mix (Thermo Fisher) and 0.5 μM of primers. Primers (5′ to 3′) were as follows: β-actin forward: ATGGTGGGAATGGGTCAGAA, β-actin reverse: TCTCCATGTCGTCCCAGTTG, Id3 forward: CTGTCGGAACGTAGCCTGG, Id3 reverse: GTGGTTCATGTCGTCCAAGAG. The qRT-PCR was run and analyzed using an Applied Biosystems 7500 Fast Real-Time PCR System (Thermo Fisher) and a QuantStudio 5 Real-Time PCR System (Thermo Fisher). Values from the qRT-PCRs were normalized to β-actin values, and relative expression values were calculated by the delta Ct method.

E18 fetal thymocytes from HEB cKO mice were incubated with Fc block for 30 min on ice. Small aliquots from each sample were stained with CD4, CD8, and CD3 to assess genotypes by flow cytometry, with a lack of DP indicating HEB deletion, and genotypes were verified by PCR. Five WT and five HEB cKO littermates were pooled by genotype, stained, and sorted for TCRγδ⁺CD3⁺ cells. Samples were not hash-tagged as this technology was not available at the time of the experiment. Fifty thousand cells per sample were loaded onto the 10× Chromium controller, and barcoded libraries were generated using the Chromium Next GEM Single Cell 5' Kit v2 (Dual Index) (10x Genomics) at the Princess Margaret Genomics Facility (Toronto, ON, Canada). Next-generation Illumina sequencing was performed to a depth of ~30,000 reads per cell. The estimated number of cells sequenced was ~3300 for WT and ~5500 for HEB cKO samples.

E18 fetal thymuses from Id3-KO mice were dissected from embryos, pressed through mesh as above, and incubated with Fc block. Embryos were genotyped by PCR, and thymocytes were pooled according to genotype (3 WT and 3 Id3-KO mice). DN cells were enriched by magnetic sorting using anti-CD4 and anti-CD8 microbeads according to the manufacturer’s instructions (Miltenyi Biotech). Flow-through (CD4^-CD8^-) cells were processed in-house using the Chromium Next GEM Single Cell 3' Kit v2 (Dual Index) (10x Genomics). Next-generation sequencing was performed using the Illumina platform to a depth of ~100,000 reads per cell. The estimated number of cells sequenced for each sample was ~3700.

FASTQ raw data files were aligned to the mm10 genome using the STAR aligner (STAR v2.5.2b). Cell Ranger (v1.1.7; 10x Genomics) (Zheng et al., 2017) was used to generate matrix files, which were analyzed using programs in R-Seurat version 4.4 (Hao et al., 2021), as detailed in the R-Markdown files. WT and HEB cKO datasets were processed for quality control by excluding cells with more than 7% mitochondrial genes, less than 1000 unique genes, and/or less than 4000 transcripts, resulting in 1272 WT cells and 1951 HEB cKO cells for further analysis. Filtered WT and HEB cKO datasets were merged, and SCTransform was applied to the merged dataset. Cell cycle regression was performed to mitigate the influence of cell cycle heterogeneity on clustering, and cells expressing high levels of Lyz2 were excluded to remove most myeloid cells. PCA was performed (RunPCA) and used to compute a nearest neighbor graph (FindNeighbors) and identify clusters (FindClusters). UMAP plots were generated using RunUMAP and displayed using DimPlot_scCustom. FindMarkers was used to identify the top 10 most differentially expressed genes between clusters.

To construct an unbiased and comprehensive list of genes diagnostic for different stages and lineages of γδ T cell development, we collected lists of differentially expressed genes from eight publications characterizing γδ T cell subsets in the fetal thymus, adult thymus, and adult peripheral tissues (du Halgouet et al., 2024; Inácio et al., 2025; Liu et al., 2020; Mistri et al., 2024; Narayan et al., 2012; Pokrovskii et al., 2020; Spidale et al., 2018; Yang et al., 2023). Genes were filtered to remove those involved in cell cycle and metabolism, as well as NK cell receptor genes. Lists were combined, duplicates were removed, and a final list of 87 genes was obtained (Figure 2—source data 1). This list was used for computation of the top 10 genes per cluster that were most differentially expressed from all other clusters, visualized as a Clustered DotPlot. Comparisons with the source literature and the Immunological Genome Project (Heng et al., 2008) were used to categorize the clusters. γδTe1 and γδTe2 represent two types of early γδ T cell subsets with random assignments into ‘1’ and ‘2’.

As developmental stages are continuous, not discrete, some genes were present in more than one module; scores were based on all genes in each module. Genes in each module are listed below.

Module location and intensity in WT versus HEB cKO cells were visualized using split UMAPs. Comparisons of the expression of single genes between WT and HEB cKO clusters were shown using split violin plots. To compare clusters 1 and 4, we performed FindMarkers, using the Wilcox test. Differentially expressed genes were displayed using an Enhanced Volcano Plot, with significance set at log₂fold change>0.5 and –log₁₀p<10²⁶.

The following analyses were conducted on the Id3-KO dataset. FASTQ raw data files were aligned to the mm39 genome using the STAR aligner (STAR v2.5.2b), and Cell Ranger was used to generate matrix files, which were analyzed using programs in R-Seurat. Cells were computationally filtered as described above in the HEB cKO dataset, resulting in 2323 WT cells and 2893 Id3-KO cells for further analysis. Datasets were merged and subjected to SCTransform as above. PCA plots were generated and used to construct UMAP plots depicting clusters in merged datasets and distribution of WT versus Id3-KO cells within each cluster by UMAP (see R-markdown file). Feature plots and violin plots were used to identify clusters with γδ T cell characteristics, which were subsetted using FindClusters. ClusteredDotPlot was used to visualize the top eight most differentially expressed genes within the curated gene set described above. Split violin plots were generated to show relative expression of early and late γδ T cell genes in WT versus HEB cKO cells within each cluster.

Thymuses were dissected from adult Rag2^-/-, WT, and HEB cKO mice, pressed through mesh as above, and incubated with Fc block. WT and HEB cKO DN cells were enriched by magnetic sorting using anti-CD4 and anti-CD8 microbeads according to the manufacturer’s instructions (Miltenyi Biotech). Flow-through (CD4^-CD8^-) cells were stained and flow-sorted into two populations: DN3 (CD4^-CD8^-CD44^-CD25⁺) cells and DN4 (CD4^-CD8^-CD44^-CD25^-). Rag2^-/- DN3 cells were obtained from Rag2^-/- mice, which do not develop past the DN3 stage, restricting the Rag2^-/- DN3 population to pre-selection cells (Shinkai et al., 1992). Duplicates were generated for each subset, with each biological replicate derived from three mice. Both males and females were used. Sorted cells were cryopreserved in 50% FBS/40% growth media/10% DMSO in aliquots of 100,000 cells, which were subjected to ATAC-seq, as follows. After membrane permeabilization, a transposase loaded with sequencing adaptors was added, which mediated insertion of adaptors at accessible genomic locations. Libraries were amplified and subjected to paired-end next-generation sequencing using the Illumina platform. Paired-end ATAC-seq libraries were generated and sequenced by Active Motif (Carlsbad, CA, USA).

Bcl2fastq2 (v2.20) was used to process Illumina base-call data and perform demultiplexing, and bwa (v0.7.12) was used to align reads to the reference genome (mm10). Samtools (v0.1.19) was used to process BAM files. BEDtools (v2.25.0) was used to process BED files, and wigToBigWig (v4) was used to generate bigWIG files. bigWIG files were aligned using the Integrative Genomics Viewer (IgV) software (v2.16.2). Each sample yielded ~65 million mapped reads.

Genes differentially expressed between cluster 1 and cluster 4 with a significance of log₂FC>0.25 and adjusted p-value<0.001 (Figure 4—source data 1) were submitted to ShinyGO 8.0 (Ge et al., 2020) for gene ontology analysis. Pathway inclusion was set at a minimum of five genes with a false discovery rate (FDR) of 0.05. Genes were analyzed using the KEGG pathway database (Kanehisa et al., 2025; Figure 4—source data 1). Results were visualized as a bar graph showing fold enrichment (numbers) and –log₁₀(FDR) values (colors).

ChIP-seq data for Rag2^-/- DN3 and Rag DN3 + γδTCR cells binding to HEB and E2A were generated as follows: Rag2^-/- fetal liver hematopoietic cells were cultured on OP9-DL4 stroma to produce DN3 cells. Rag2^-/- DN3 cells were retrovirally transduced with the KN6 γδTCR or control (no TCR), and sorted GFP⁺ DN3 cells were cultured for 4 days on stroma expressing the weak KN6 ligand T10 to initiate γδTCR signaling. These cells were subjected to ChIP-seq using anti-E2A and anti-HEB antibodies as previously described (Fahl et al., 2021). All other files were obtained from the Cistrome database (Zheng et al., 2019). Peaks were aligned to the mouse genome (mm38) using the IgV (Robinson et al., 2011). Sources were as follows: Thy anti-TCRb-Egr2, GEO accession # GSM845900 (Seiler et al., 2012); Rag d7 aCD3-RNA pol II, GEO accession # GSM1340642 (Pekowska et al., 2011); DN3 H3K27me3, GEO accession # GSM1498423, and DN4 H3K27me3, GEO accession # GSM1498422 (Oravecz et al., 2015).