From benchmarking alignment of genome assemblies to IMGT annotation: the paradigm of the bovine Bos taurus T cell receptor (TRG) locus

Abstract

T cell receptors (TR) are essential components of the adaptive immune system, typically classified into αβ and γδ types. In humans and mice, αβ T cells predominate, with γδ T cells comprising only a small percentage of the total T cell population. γδ T cells are mainly distributed in peripheral tissues rather than lymphoid organs and have limited diversity. However, in ruminant species, the proportion of γδ T cells is significantly higher. To better understand bovine γδ T cells, comprehensive annotation of the bovine TRG locus is essential. Recent advancements in sequencing technologies have led to the availability of high-quality chromosome-level genomes, enabling more precise annotation of TR loci. In this study, by using the LAST alignment tool and comparative genomic analysis, we identified previously unannotated TRG genes in the bovine genome, including 1 novel TRGV gene, 11 novel TRGJ genes and 1 novel TRGC gene. We compared and integrated information from three different assemblies of the bovine genome to provide an updated annotation of the bovine TRG locus. Expression of one newly identified TRGJ gene was experimentally validated through next-generation sequencing. This study expands our knowledge of the bovine TRG locus and repertoire through improved TRG locus annotation and expression data, providing a more complete picture of bovine γδ T cell diversity and function, which may help explain the unique immunobiology of cattle.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-025-12081-z.

Introduction

Antigen receptors of the adaptive immune responses include Immunoglobulins (IG) or antibodies and T cell receptors (TR), synthetized by the B cells and plasma cells and by the T cells, respectively [1]. The IG and TR antigen receptors achieve extensive diversity through somatic rearrangements of the Variable (V), Diversity (D) and Joining (J) genes which result in the synthesis of the variable domain (V-(D)-J-region) at the N-terminal end of each IG and TR chain [2, 3]. Through V(D)J rearrangements, junctional diversity, and—specifically for IG—somatic hypermutations [4], a vast repertoire of IG and TR is generated. This enables the recognition of diverse antigens, with theoretical diversity estimates in humans exceeding 10¹³ for TR and 10¹⁶ for IG [1, 5, 6]. The chain isotype and effector functions of IG and TR are determined by the constant (C) region encoded by C genes, with additional diversity in IG achieved through class switch recombination [7].

T cells are mainly divided into αβ T cells and γδ T cells based on the chains expressed on their cell surfaces. The genes encoding the α and δ chains are both located at the TRA/TRD locus, while the genes for β and γ chains are distributed at the TRB and TRG loci, respectively [8]. γδ T cells are among the earliest T cells to develop in the thymus. In humans and mice, γδ T cells represent only a small fraction (1–5%) of circulating T cells in blood and secondary lymphoid organs [9]. In contrast, the proportion of γδ T cells in ruminant species is significantly higher than in other mammals, with γδ T cells constituting 30–60% of peripheral blood mononuclear cells (PBMCs) in young sheep and cattle [10, 11]. This marked difference in abundance suggests distinct evolutionary pressures and functional importance of γδ T cells in ruminants. Studies have shown that ruminant γδ T cells express a diverse repertoire [12–14] and can directly respond to various pathogen-associated molecular patterns (PAMPs), with TLR 2, 3, 4, and 7, as well as NOD1 and NOD2, participating in bovine γδ T cell responses to LPS, peptidoglycan, and Poly I: C [15–17].

The Bos taurus TRG genes used to synthetize the bovine γ chain were previously reported [18, 19]. Cattle TRG locus is split in two loci, TRG1 and TRG2, at different regions of the same chromosome; it was later confirmed that TRG1 is located on chromosome 4 at 4q3.1 containing 8 TRGV, 4 TRGJ, and 3 TRGC [19], while TRG2 is positioned at 4q1.5-2.2 with 4 TRGV, 5 TRGJ, and 3 TRGC. A similar organization in two TRG loci has been found in other ruminant species such as sheep and goat [20–22]. The bovine TRG1 locus spans 250 kb and is in forward (FWD) orientation on the chromosome whereas the TRG2 locus spans 190 kb and is in reverse (REV) orientation [8]. They are approximately 32 Mb apart. The TRG locus 5’ and 3’ bornes are AMPH (amphiphysin) and STARD3NL (STARD3 N-terminal like), respectively [8]. AMPH is upstream of TRD1 most 5’ gene. Owing to the TRG2 reverse orientation, STARD3NL, downstream of the most 3’ gene of the TRG2 locus, is located on the chromosome upstream of the most 3’ gene of the TRG2 locus [8].

Both ruminant TRG1 and TRG2 loci have an organization in clusters with, within each cluster, several TRGV genes preceding a few TRGJ genes, themselves preceding a single C gene [21, 23]. Bovine γδ T cells are usually classified based on the expression or not of a pathogen-recognition receptor, Workshop Cluster1 (WC1), on their surface, with WC1 + being the predominant population in the periphery [24]. In bovine, goat and swine γδ T cells, the expression of WC1 demonstrates a unique pattern of gene usage: despite the presence of six TRGC clusters in the genome, experimental studies have shown that WC1 + γδ T cells exclusively utilize only one specific cluster—TRGC5 (TRGC1 for swine) [25–27]. Notably, the TRGC5 cluster has been shown to be an evolutionarily ancient gene cluster with an expression pattern which suggests a conserved immune function mechanism across species, whereas the other cluster may have derived from each other by successive replications which led to similar TRGC clusters [21, 28].

Earlier annotations of the Bos taurus loci may have missed some TR genes, given that most species have a relatively limited number of TRG genes. Recently, many high-quality chromosome-level genomes have become available, making more complete description of IG and TR loci possible in different species [29, 30]. Although several methods and tools have been developed for the identification of IG and TR genes [31–36], a complete and thorough identification is very challenging because the variable length of the V, D and J gene, the varying degree of conservation and even the evolutionary relationships between species can affect the final prediction.

In this study, we re-examined the genomic sequence of the TRG locus using the latest high-quality bovine genome assembly [37]. We employed a robust method by testing a wide range of parameters and reference species sequences to identify several previously unannotated TRGV/TRGJ/TRGC genes in the bovine TRG locus. Then, we compared the three different TRG loci from three assemblies and confirmed that that the bovine TRG loci exhibit structural variations among these three different genomic assemblies. Finally, we validated the functionality of one newly annotated TRGJ gene through high-throughput sequencing. Using the resulting gene expression data, we also found evidence of an association between the expression of the TRGJ genes and the recombination signal (J-RS) (12-RS) sequence. Our findings not only expand the known repertoire of bovine TRG genes but also provide insights into the potential functional implications of these newly discovered genes, contributing to a better understanding of γδ T cell diversity and function in cattle.

Results

Effect of parameters and reference on alignment

For TRGV gene identification, we analyzed the performance of six reference species combinations under the e = 30 parameter setting. Results revealed an average of 20.5 ± 1.0 hits, of which 18 overlapped with existing annotations and were classified as true positives (TP) (Supplementary Fig. 1a), while the remaining 2.5 ± 1.0 hits were initially categorized as false positives (FP) (Supplementary Fig. 1b). Manual validation of these FPs led to the identification of one previously unannotated novel TRGV gene. Notably, when more lenient thresholds (e = 25 and e = 20) were employed, despite significantly increasing the average number of hits to 30.8 ± 2.3 and 157.7 ± 10.9 respectively, no additional novel TRGV was found (Supplementary Fig. 1c).

For TRGJ genes, the e = 30 parameter setting yielded an average of 17.7 ± 2.9 hits (Supplementary Fig. 1 d). All reference species combinations except cat-mouse successfully identified all 10 known TRGJ genes, and through manual verification of the average 8.3 ± 1.9 FPs (Supplementary Fig. 1e), we identified up to 7 novel TRGJ genes (Supplementary Fig. 1f). When parameters were relaxed to e = 15, although FPs significantly increased to 110.3 ± 38.0, the number of novel genes discovered rose to 11 (Supplementary Fig. 1f).

By comparing the performance of different reference species combinations, we found that the cat-mouse combination consistently underperformed across all parameter settings, detecting the fewest known and novel genes (Supplementary Fig. 2a-b). In contrast, “human-monkey”, “human-mouse”, or “monkey-mouse” combinations significantly reduced the proportion of FPs in TRGJ analysis while maintaining efficient gene detection capabilities (Supplementary Fig. 2c).

Finally, we present the known and newly identified genes using human and monkey genomes as reference, with the optimal parameters set at e30 for TRGV (Fig. 1a) genes and e15 for TRGJ genes (Fig. 1b).

Fig. 1.

Identification of new TRGV/TRGJ genes. (a) TRGV genes under e30 condition and (b) TRGJ genes under e15 condition. Results were classified as: known (purple) for sequences overlapping with existing genes, unknown (yellow) for previously identified but unannotated sequences, and unknown-new (blue) for sequences with V/J gene characteristics but not overlapping with existing genes. The y-axis represents the alignment score, and the x-axis shows gene name from both TRG1 and TRG2 loci

Identification and naming of novel TRGV/TRGJ/TRGC genes

Using the optimized parameters, we identified one previously unannotated TRGV gene, eleven TRGJ genes, and one TRGC gene from the NCBA_BosT1.0 assembly.

The newly discovered TRGV gene (TRGV11) (Table 1) is a pseudogene due to the presence of a stop codon and the absence of the V-RS (23-RS). Nevertheless, it retained the V-region four conserved amino acids (the cysteine C23, the tryptophan W41, the hydrophobic amino acid 89 and the cysteine C104 [38]) and the associated structural elements that we identified: WY motif at positions IMGT 41–42, CALW sequence with ALW of the CDR3-IMGT region (105–107) following the FR3-IMGT C104 (Fig. 2a). Multiple sequence alignment analysis revealed that this gene (TRGV-new) shared less than 50% sequence identity with all previously known cow TRGV genes. However, the new TRGV gene has 79% similarity to human TRGV9, a unique representative of the TRGV2 subgroup in Homo sapiens. So far, it has been shown that the bovine TRGV genes identified belonged to six of the seven mammal TRGV subgroups, with the TRGV2 subgroup missing. With TRGV11 belonging to the TRGV2 subgroup, this shows that in terms of evolution, the Bos taurus species has representatives of the seven mammal TRGV subgroups and that TRGV11 may be found functional in other related species [39] (Fig. 2b).

Table 1.

Nomenclature and functionality of the novel Bos taurus (Bostau) bovine TRG genes

Temporary number	Official name	Functionality	Assembly
TRGV-new	TRGV11	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0
TRGV8-4	TRG8-4D	Functional	ARS-UCD2.0
TRGJ-new1	TRGJ5-2	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0
TRGJ-new2	TRGJ5-3	Functional	NCBA_BosT1.0, ARS-UCD2.0
TRGJ-new3	TRGJ5-4	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0
TRGJ-new4	TRGJ7-1	Functional	NCBA_BosT1.0, ARS-UCD2.0, AY644517
TRGJ-new5	TRGJ3-3	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0, AY644517
TRGJ3-2	TRGJ8-2	Functional	NCBA_BosT1.0
TRGJ-new6	TRGJ8-4	Pseudogene	NCBA_BosT1.0
TRGJ-new7	TRGJ4-3	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0, AY644517
TRGJ-new8	TRGJ6-2	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0
TRGJ6-2	TRGJ6-3	Pseudogene	AY644518
TRGJ-new9	TRGJ2-3	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0, AY644518
TRGJ-new10	TRGJ2-4	Functional	NCBA_BosT1.0, ARS-UCD2.0, AY644518
TRGJ-new11	TRGJ1-3	Pseudogene	NCBA_BosT1.0, ARS-UCD2.0, AY644518
TRGC-new	TRGC8	Functional	NCBA_BosT1.0

Fig. 2.

Cattle (Bos taurus) TRGV gene analysis: protein alingment and phylogenetic relationships. a. Protein display of the TRGV gene in cattle, New TRGV11 sequences from the NCBA−BosT1.0 and ARS-UCD2.0 assemblies are highlighted in bold red. Dots indicate gaps according to the IMGT unique numbering. Letters in red correspond to amino acids which are polymorphic in other alleles. Colors follow IMGT conventions for mutations, sequence alignments, and splicing types. b. Neighbor-Joining tree of all TRGV subgroups across species with one representative gene per subgroup (using V-REGION). Species abbreviations: Homsap: human, Macmul: rhesus monkey, Gorgor: western gorilla, Bostau: cattle, Orycun: rabbit, Hetgla: naked mole-rat. The two new-TRGV11 sequences from ARS-UCD2.0 and NCBA−BosT1.0 assemblies are highlighted in red. Phylogenetic analysis used 1000 bootstrap replicates; values higher than 95 are marked with pink dots. Analysis parameters: and visualized with iTOL v [60]. Data available in IMGT Repertoire (IG and TR) https://imgt.org/IMGTrepertoire/. TRGV subgroups are according to references [39]

The eleven newly identified Bos taurus (Bostau) TRGJ genes are listed in Table 1. The sequences of the J-REGION and the J-RS were identified. Sequence identity analysis revealed high identity among TRGJ5-2, TRGJ5-4, TRGJ4-3, TRGJ2-3, and TRGJ1-3, with sequence identities ranging from 70–100% (Fig. 3), with 2–3 and 1–3 being 100% identical to each other. TRGJ7-1 exhibited 84% sequence identity to the existing gene TRGJ3-1/2, while TRGJ2-4 showed 88% sequence identity to TRGJ4-1 (Supplementary Fig. 3). Furthermore, TRGJ6-2 shared 81% identity with TRGJ2-4. Although TRGJ3-3 and TRGJ8-4 lack the J-RS and the conserved J-MOTIF, they are 71% similar to human TRGJ2 [40]. Further analysis showed that TRGJ5-3, 7 − 1, and 2–4 contained the J-RS in 5’ (12RS), the conserved FGXG motif, the donor splicing site in 3’, and 12RSS; they were therefore predicted to be functional genes. In contrast, TRGJ5-2, 5 − 4, 4 − 3, 2–3, and 1–3, despite the conserved J-MOTIF, lacked the J-RS (12-RS)12RSS elements and were classified as pseudogenes. TRGJ3-3, TRGJ8-4, and TRGJ6-2 were also designated as pseudogenes due to the presence of stop codons in the J-REGION.

Fig. 3.

Sequence alignment of novel bovine TRGJ genes. a Alignment of nine newly identified TRGJ genes showing the J-RS (12-RS) comprising the J-NONAMER, the 12 nucleotides spacer and the J-HEPTAMER (left), J-REGION with the J-MOTIF (FGXG) and DONOR-SPLICE (right). TRGJ3 and TRGJ5-3, 7-1, and 1-3 contain complete functional elements and are predicted to be functional genes. b Sequence alignment of TRGJ3-3 and TRGJ3-4 compared with human TRGJ2, showing their sequence homology despite lacking functional elements. Dashes are only used to align the J-REGION. Genome assembles in which the bovine TRGJ were identified are indicated in Supplementary Table 2. IMGT/LIGM-DB accession number of Homo sapiens TRGJ2 is M12961

To our surprise, the TRGC-new gene was aligned to four TRGC of Heterocephalus glaber (Supplementary Fig. 4). The newly discovered TRGC8 gene shares 97.9% sequence identity with the first exon of TRGC3 and exhibits no obvious defects, indicating that it is likely a functional gene (Fig. 4).

Fig. 4.

Cattle (Bos taurus) TRGC gene analysis: protein alignment and phylogenetic tree. a. Protein display of the TRGC gene in cattle. New-TRGC8 from NCBA−Bost1.0 are highlighted in bold red. b. Method of tree constrcution as follow Fig. 3. Data available in IMGT Repertoire (IG and TR) https://imgt.org/IMGTrepertoire/

Comparison of three bovine TRG loci

The updated bovine TRG loci in the NCBA_BosT1.0 assembly are displayed in Fig. 5a with newly discovered genes indicated in red.

Fig. 5.

Cow TRG locus in three different assemblies. Genes are colored according to the IMGT color menu for genes. IMGT Locus representation: Bovine (Bos taurus) TRG locus on chromosome 4 (4q3.1 and 4q1.5-2.2) (https://www.imgt.org/IMGTrepertoire/index.php?section=LocusGenes&repertoire=locus&species=bovine&group=TRG). The boxes representing the genes are not drawn to scale. Triangular arrows indicate the transcriptional orientation of each gene. Transparent regions highlight the differences between the three assemblies. Newly identified genes are shown in red font. In panels (a), (b) and (c), TRG2 is shown as it appears on the chromosome, in reverse orientation (REV) relative to TRG1 (FWD). In panel (a), TRGJ6-1 is rearranged to TRGV6-2, and TRGJ2-1 is rearranged to TRGV5-2

After applying the same methodology to the ARS-UCD2.0 genome assembly, we identified 15 TRGV, 8 TRGJ, and 4 TRGC genes in the TRG1 locus, as well as 4 TRGV, 9 TRGJ and 3 TRGC genes in the TRG2 locus (Fig. 5b). Since the two TRGV8-4 found in ARS-UCD2.0 are completely identical, we renamed TRGV8-4 to TRGV8-4D (Table 2).

Table 2.

Summary of TRG repertoire data from three cattle samples

Metrics	Cow 1	Cow 2	Cow 3
Total sequencing reads	84,918,150	82,467,695	81,242,791
Successfully aligned reads	72,782,657 (85.71%)	66,969,415 (81.21%)	69,768,899 (85.88%)
Final clonotype count	150,698	82,910	90,812
TRG functional	82,614 (54.82%)	47,239 (56.98%)	48,487 (53.39%)

IMGT Repertoire (IG and TR) [1] show that the bovine TRG1 locus contains 13 TRGV genes, 4 TRGJ genes, and 4 TRGC genes, derived from two discontinuous sequences: NW_937068 for the TRGC5 cluster, and AY644517 for the TRGC7, TRGC3, and TRGC4 cluster, while the TRG2 locus contains 4 TRGV genes, 5 TRGJ genes, and 3 TRGC genes from the TRGC6, TRGC2, and TRGC1 cluster in AY644517 (Fig. 5c). Upon application of our methodology, we found that certain TRGJ and TRGV genes (including TRGJ5-2, 5 − 3, 5 − 4, and TRGV11) were absent from the NW_937068 scaffold. This is primarily attributed to insufficient length. Specifically, in NW_937068, TRGJ5-1 is situated approximately 600 bp from the 5’ end, whereas in the NCBA_BosT1.0 genome, the distance between TRGJ5-1 and the nearest new TRGJ gene (TRGJ5-2) is approximately 2,500 bp. In the NCBA_BosT1.0 genome, the distance between TRGV3-1 and new TRGV11 is approximately 13,000 bp, while in NW_937068, TRGV3-1 is positioned 11,500 bp from the 3’ end. We identified three novel TRGJ genes (TRG7-1, TRGJ3-3, TRGJ4-3) at the TRG1 locus of AY644517 which is same as ARS-UCD2.0. Although four novel TRGJ genes (TRGJ2-3, TRGJ2-4, TRGJ1-3, and “TRG6-2”) at the TRG2 locus of AY644518 had been identified, which is same as NCBA_BosT1.0 (Fig. 5c). Three of them (TRGJ2-3, TRGJ2-4, and TRGJ1-3) are identical to those in NCBA_BosT1.0 except for “TRGJ6-2”.

Comparative analysis among IMGT (AY644517, AY644518), NCBA_BosT1.0, and ARS-UCD2.0 revealed structural differences in both bovine TRG1 and TRG2 loci.

For the TRG1 locus, firstly regarding the TRGC7 cluster, IMGT (AY644517) lacks a TRGV8-4D gene that is completely identical to TRGV8-4 in ARS-UCD2.0, while in the NCBA_BosT1.0 TRGC7 cluster, neither TRGV8-4 nor TRGV8-4D genes can be found. Secondly, although the TRGC3 clusters in IMGT (AY644517) and ARS-UCD2.0 are identical, the TRGC3 cluster in NCBA_BosT1.0 contains an additional TRGV9-3 gene between TRGV9-1 and TRGV9-2, which is not found in either IMGT (AY644517) or ARS-UCD2.0. Finally, neither IMGT (AY644517) nor ARS-UCD2.0 includes the TRGC8 cluster that we identified in the NCBA_BosT1.0 assembly.

For the TRG2 locus, TRGV6-1 is classified as a pseudogene in the ARS-UCD2.0 genome due to a 78-base insertion causing a frameshift, whereas in NCBA_BosT1.0 and IMGT (AY644518), TRGV6-1 is a functional gene (Supplementary Fig. 5). Secondly, TRGJ6-2 in NCBA_BosT1.0 and ARS-UCD2.0 are completely identical, but TRGJ6-2 in IMGT (AY644518) differs entirely from them and instead shares a higher similarity of 77% with new TRGJ5-4; therefore, we named this gene TRGJ6-3 (Fig. 3).

All previous and newly annotated VDJC gene sequences and positions, from both ARS-UCD2.0 and NCBA_BosT1.0 assemblies, are provided for reference (Supplementary Tables 2–4).

The TRG repertoire of bovine PBMCs

After incorporating the three functional new TRGJ, TRGJ5-3, TRGJ7-1 and TRGJ2-4) into the IMGT reference files, we analyzed the raw fastq data of peripheral blood mononuclear cell (PBMC) TRG receptor Libraries from three cattle using MiXCR. The final clonotype counts for the three samples were 150698, 82910, and 90812, with functional sequences accounting for 54.82%, 56.98%, and 53.39%, respectively (Table 2). Clonotype abundance analysis revealed similar distribution patterns across all three samples, predominantly comprised of small-abundance clonotypes (< 0.01%), accounting for approximately 70% of the repertoire. Minor variations were observed between samples: cow1 had a higher proportion of medium-abundance clonotypes (0.1%−1%) at 11.73%; cow3 displayed the highest proportion of hyper-expanded clonotypes (1%−100%) at 3.62%; while cow2 showed a relatively balanced distribution across abundance categories (Fig. 6a). CDR3 lengths occurred with a Gaussian-like frequency distribution, with the most frequent lying in the range 12–14, the most common length being 13, which occupied 24.00 ± 0.35% of the total expression (Fig. 6b).

Fig. 6.

Three cattle PBMC TRG repertoire analysis. a. Distribution of clonotypes across three samples. b, Average CDR3 region lenght distribution. c, Expression of TRGJ5-3 within the TRGC5 cluster. d-e, Average expression of all TRGV and TRGJ genes across the three samples. f, Average expression of V-J gene pairings. g, Correlation between V-RS scores and expression levels of TRGV genes. h, Correlation between J-RS scores and expression levels of TRGJ genes

We observed distinct preferences in TRGV and TRGJ gene expression. Among them, we found that almost all newly identified TRGJ5-3 expression came from the TRGC5 primer, with TRGJ5-3 (83.46% ± 5.97) showing predominant expression in this cluster compared to TRGJ5-1 (16.54% ± 5.97) (Fig. 6c). For all TRGJ genes, TRGJ4-2 and TRGJ6-1 were dominant, representing 25.82 ± 0.18% and 24.80 ± 0.46%, respectively. The newly discovered TRGJ5-3 appeared at a frequency of 8.68 ± 1.65%, while newly identified TRGJ2-4 and TRGJ7-1 were rarely detected, with only a few counts (Fig. 6e). TRGV5-2, TRGV6-2, and TRGV1-1 were the most frequently observed V genes, accounting for 25.60 ± 2.14%, 23.93 ± 0.59%, and 22.58 ± 0.70% of total V gene usage, respectively (Fig. 6d). By exploring the V-J pairings, we found that TRGV6-2-TRGJ6-1 was the most common combination, accounting for 23.92% of the total pairings, followed by TRGV1-1-TRGJ4-2 (15.30%) and TRGV5-2-TRGJ2-2 (12.39%). The newly identified TRGJ5-3 gene was mainly paired with three V genes: TRGV3-1 (3.38%), TRGV7-1 (2.92%), and TRGV3-2 (2.36%), and the genes that were paired with TRGJ5-3 were within the same cluster (Fig. 6f).

We calculated the RIC score for each gene and performed correlation analysis with the average expression levels. Our analysis demonstrated that the J-RS influenced TRGJ gene expression but had minimal impact on TRGV gene expression. For TRGJ genes, a moderate correlation was observed between RIC scores and expression levels (R²=0.3387, p = 0.04712), indicating that J-RS sequence quality accounts for approximately 33.87% of the variation in TRGJ gene expression (Fig. 5h). In contrast, for TRGV genes, there was virtually no correlation between RIC scores and expression levels (R²=0.0098, p = 0.7060) (Fig. 6g).

Discussion

In this study, we developed and validated a simple alignment-based method to comprehensively update the bovine TRG locus, and used high throughput sequencing to verify the expression patterns of newly identified TRGJ genes. Additionally, we observed significant variations in the TRG loci across different bovine genome assemblies. Our findings extend our current understanding of TR gene diversity in ruminant species.

The earliest bovine TRG genes were discovered by Carolyn Herzig et al. in 2006 [18]. Subsequent work utilizing Bacterial Artificial Chromosome (BAC) libraries revealed the complete structure of two bovine TRG loci, providing valuable information for further research on the γ chain and loci in ruminants [19] (Fig. 3(a)). Unlike other mammals, bovine TRG1 and TRG2 loci are located on chromosome regions 4q3.1 and 4q1.5-2.2, approximately 32 Mb apart on chromosome 4. The AMPH 5’ borne is upstream of the most 5’ gene of TRG1 as expected but owing to the reverse orientation of TRG2 the STAR3NL 3’ borne is upstream of the TRG2 more 3’ gene, in contrast to what is commonly observed in other mammals [41, 42]. This unique arrangement likely originated from chromosomal rearrangement events in the ruminant ancestor, which may be related to the known extensive expansion of TR genes in ruminants [43, 44]. These genomic structural differences may reflect adaptations to specific environments and pathogen pressure.

In recent years, with the rapid development of sequencing technology at a significant reduction in cost, coupled with high-quality genome projects initiated by research teams worldwide, we can now access more precise genomic data for analysis [29, 30]. Although the emergence of high-quality genomes has resolved data quality issues that have long hampered gene annotation, accurately mapping adaptive immune receptor genes in genomes remains challenging. Despite the recent development of several tools for identification of the IG and TR genes [31–34, 36, 45], balancing specificity and sensitivity in bioinformatic analysis continues to be problematic: overly high thresholds prevent the discovery of evolutionarily divergent germline genes, while excessively low thresholds generate numerous false positives, increasing the burden of manual validation work.

TR genes consist of four types of genes: V, D, J, and C, which vary in length, characteristics, and conservation levels, thus requiring different annotation approaches for each type. Our research clearly demonstrates that bovine TRGV and TRGJ genes require different alignment parameters to achieve optimal results. Although using expression data to identify germline genes, such as employing 5’RACE methods to avoid directly designing primers through V genes or J genes [46], is feasible, this approach faces two major challenges. First, when sequence identity is high, it becomes difficult to accurately distinguish true germline genes. In this paper, TRGJ5-3 exemplifies this case—although Carolyn Herzig et al. amplified over 100 clones at the time [18], TRGJ5-3 was Likely overlooked due to its 75% high similarity to TRGJ4-1. Second, there are Limitations in current sequencing technology. Most mainstream HTS platforms currently use paired-end 150bp sequencing, which is insufficient for distinguishing highly similar sequences. In cases where sequence identity reaches 98% or more, as with TRGC8 and TRGC3 in cattle, TR full-length sequencing becomes the only reliable method for differentiation. In this study, while we cannot directly determine whether TRGC8 is expressed, the evidence is compelling. Considering that clonotypes amplified by TRGC8 and TRGC3 primers account for 44.87% ± 2.71% of total functional sequences, we have substantial reason to believe that TRGC8 is expressed.

We constructed TRG repertoires from peripheral blood samples of three cattle and analyzed their clonal distributions. Our analysis revealed that TRG chains are predominantly composed of rare clones (70%), which differ from previous findings by Alexandria E. Gillespie et al. regarding bovine peripheral TRG receptors [47]. Despite this difference, our study identified an average of 59,447 ± 15,445 clonotypes across samples. Although cow1 showed different clonotype numbers compared to cow2 and cow3, we believe that these data, which cover the majority of TRGV and TRGJ functional genes, are sufficient to characterize the majority of γ-chains. The observed differences in clonal distribution may be attributed to methodological variations in library construction—our study employed multiplex PCR whereas Gillespie et al. used 5’RACE.

Regarding expression data, TRGJ5-3 showed substantial expression (8.68 ± 1.65%), while TRGJ2-4 and TRGJ7-1 were represented by only a few clonotypes. The minimal expression of TRGJ7-1 is expected since it resides in the TRGC7 cluster, which contains a non-functional TRGC7 gene. TRGJ2-4 is located in TRGC2 with an atypical V-J-V-J structure rather than the standard V-J-C cluster arrangement, which might affect its expression. Additionally, our analysis revealed a significant correlation between RIC scores and TRGJ expression levels. Multiple studies have confirmed the correlation between J-RS and V-RS quality and expression levels of TR and IG light chains [48, 49]. TRGJ2-4 has a RIC score of −34, the second lowest among all TRGJ genes after TRGJ4-1 (−42). Since TRGJ4-1 shows virtually no expression, the poor RIC score of TRGJ2-4 likely contributes to its minimal expression as well.

Another possible explanation involves species or individual variation. Since our study utilized Japanese cattle, which may exhibit polymorphisms compared to the reference sequenced species, the impact of polymorphisms at TR loci might have been underestimated [50]. Martin Corcoran et al., after conducting detailed analysis of TRA, TRB, TRG and TRD genes expressed in 45 donors from four human populations (African, East Asian, South Asian, and European), discovered 175 additional TR variable and junction alleles [51]. Furthermore, My H. Hoang et al. detected numerous genes annotated as pseudogenes or open reading frame (ORF) across multiple cells and dogs; they demonstrated that in at least one canine subject, a germline SNP corrected a stop codon, rendering a pseudogene functional. In another instance, TRBJ1-3 (a pseudogene owing to an in-frame stop codon at position 2 in the J-region) may be part of productive V-D-J sequences with any TRBV gene when the V-D-J rearrangement eliminates the stop codon, an observation consistent across all dogs studied [52]. As described here , we were unable to detect two identical TRGV8-4 in the IMGT (AY644517) or TRGV8-4D in the IMGT (AY644517); nor could we identify TRGV9-3 in the NCBA_BosT1.0 assembly. Thus, the bovine TRG loci exhibit variations across different genome assemblies (two or even three), and these differences may be manifested through TRGJ2-4.

In conclusion, this study provides a simple approach for predicting bovine TRG gene loci and may be applicable to other families. The specific parameter choices are likely limited to ruminant TRG genes, as the conservation levels of TR and IG chains vary considerably. Future work will be required to explore how to comprehensively identify IG and TR genes in a wider range of species through the use of Machine Learning to automate parameter optimization.

Materials and methods

Sequence collection

The Bos taurus genome sequences were obtained from three different sources: ARS-UCD2.0 (NCBI assembly accession: GCF_002263795.3), NCBA_BosT1.0 (National Genomics Data Center assembly accession: GWHBISA00000000), and IMGT reference sequences. The IMGT reference included, for the TRG1 locus, sequences from NW_937068 (TRGC5 cluster) and AY644517 v.2 (TRGC7, TRGC3 and TRGC4 clusters) [53], and for the TRG2 locus, sequences from AY644518 (TRGC1, TRGC2 and TRGC6 clusters) [53]. Due to the large distance between bovine TRG1 and TRG2 loci (approximately 30 million base pairs), we extracted and merged sequences extending 100 kb upstream and downstream of either locus to use as query sequences. The merged file for ARS-UCD2.0 had a total of 644,025 bp and the merged file for NCBA_BosT1.0 had a total of 711,238 bp. The sequences of the IMGT reference directories [54] used for the identification of the genes in the different assemblies were collected from the IMGT/LIGM-DB through query of the IMGT/GENE-DB [55] (https://www.imgt.org/genedb/).

Sequence alignment and parameter optimization

We employed LAST (version 1454), a widely used local alignment tool, to locally align the sequences [56–58]. In order to properly set LAST alignment score thresholds, we selected pairs of reference species—Homo sapiens (human), Mus musculus (mouse), Felis catus (cat), and Macaca mulatta (rhesus monkey)-resulting in six different pairs: human-mouse, human-cat, human-monkey, mouse-cat, mouse-monkey, and cat-monkey.

The lastal command was used for sequence alignment and parameter optimization. Initially, we established a range (lenient to stringent) of parameters: the minimum alignment score (-e) ranged from 15 to 30 (with a step size 5). For alignment results from each parameter, we first performed a preliminary classification: hits overlapping with known annotated genes were classified as true positives (TP), while the remaining results were tentatively designated as false positives (FP). Considering that preliminary FPs might contain unannotated novel genes, we conducted comprehensive manual examination of all FPs, searching for the potential V genes, the characteristic four conserved amino acids of the variable region (cysteine C23, tryptophan W41, hydrophobic 89 and cysteine 104 [38]), the presence of the L-PART1 and in 3’ the canonical V recombination signal (V-RS) (23-RS) and for the potential J genes, in 5’ the canonical J-RS (12-RS), the J-motif (F-G-X-G) and in 3’ the DONOR-SPLICE. Each type of RS contains a conserved 7-mer and 9-mer separated by a less conservative 12- or 23-nucleotide spacer, forming 12-RS and 23-RS, respectively. In the TRG locus, the RS identification includes the J-RS (12-RS) and the V-RS (23-RS). Results exhibiting these features were reclassified as “Unknown-new” genes. Ultimately, based on the ability of each parameter ability to identify both known and novel genes, while simultaneously controlling FP rate, we determined optimal parameter settings for TRGV and TRGJ genes respectively.

Due to the high conservation of TRGC, we simply collected EXON1 sequences of 53 TRGC genes from 12 species managed in the IMGT/LIGM-DB database by querying IMGT/GENE-DB. We performed the “Map to Reference” function in Geneious Prime (version 2024.0.7) with the sensitivity set to ‘medium’ and the fine-tuning parameter set to ‘25x’. The identity of potential TRGC genes in the alignments was then manually confirmed. Neighbor-Joining trees of TRGV and TRGC genes were constructed using MEGA12, with 1000 bootstrap replicates [59]. Analysis parameters included pairwise deletion for 1 st + 2nd + 3rd + Noncoding positions. Phylogeny tree visualization was performed using iTOL v6 [58].

Bovine TRG repertoire analysis

Raw data were processed using the MiXCR software (version 4.6.0) to characterize the TRGV-J repertoire [53]. We first collected cattle TRGV and TRGJ nucleotide sequence files from IMGT/V-QUEST reference directory sets (released on 25 June 2009), https://www.imgt.org/download/V-QUEST/IMGT_V-QUEST_reference_directory/Bos_taurus/TR/TRGV.fasta, https://www.imgt.org/download/V-QUEST/IMGT_V-QUEST_reference_directory/Bos_taurus/TR/TRGJ.fasta, then incorporated our newly identified genes to create an updated reference library (Supplementary file1). The analysis was performed using the MiXCR "generic-amplicon" preset pipeline. The number of successful mapped reads, clonotype counts, and the functional ratio for each sample are presented in Table 1. The output files from MiXCR were used for downstream analysis. For TRGV and TRGJ usage, the count number. For For clone distribution analysis, we employed the Immunarch package (version 0.9.1) in R studio (version 2024.12.1)[60]. Additionally, we used the RSSite online tool (https://www.itb.cnr.it/rss/analyze.html) to calculate Recombination information content (RIC) scores for each TRGV and TRGJ gene, enabling us to evaluate the correlation between gene expression and V-RS and J-RS conservation [61].

Authors’ contributions

DMS. and KK.: funding acquisition, supervision M.P.L., D.M.S. and K.K.: review & editing M.P.L., HZ.: conceptualization, investigation, validation, formal analysis, writing - original draft CJ. and CG.: resources.

Funding

This work was supported by JST SPRING, Grant Number JPMJSP2138 and the Fujii Medical International Exchange Foundation, and supported by Platform Project for Supporting Drug Discovery and Life Science Research (Basis for Supporting Innovative Drug Discovery and Life Science Research [BINDS]) from AMED under Grant Number JP21am0101001.

Data availability

Three bovine peripheral blood TRG receptor library raw data have been uploaded to SRA (PRJNA1242064).

Declarations

Ethics approval and consent to participate

Blood samples were obtained postmortem from a licensed meat processing and storage center, and no live animal experimentation was involved. This procedure was conducted under the supervision of the Nishinomiya City Health and Welfare Bureau, Meat Hygiene Inspection Office in accordance with standard protocols for humane slaughter of food animals.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.