Vδ Repertoire During Thymic Ontogeny Suggests Three Novel Waves of γδ TCR Expression

The γδ T cells can be detected in the thymus around day 13 of gestation, 3 days before αβ T cells are found. During fetal ontogeny successive waves of γδ cells arise bearing TCR composed of different variable (Vγ and Vδ) regions (reviewed in Refs. 1 and 2). The first wave comprises cells bearing GV1 (Vγ3; for nomenclature, see Ref. 3) paired with DV101 (Vδ1). On the basis of subtractive Ab staining, a second wave using GV2(Vγ4)/DV101 receptors has been postulated. Both fetal clonotypes, generated by canonical rearrangements of V, D (only in δ-chains), and J gene segments, colonize specific epithelia, where they persist in the adult mouse. The intraepithelial lymphocytes that populate the skin (s-IEL)⁴ bear the GV1/DV101 TCR (4), and both the intraepithelial cells of the female reproductive tract (r-IEL) (5, 6) and, during the perinatal period, the resident pulmonary lymphocytes (7) bear the GV2/DV101 TCR. The intestinal intraepithelial lymphocytes (i-IEL) exhibit some diversity in Vδ usage, employing predominantly DV104, DV105, and ADV7 (ADV7 designating the Vα7 subfamily that includes the homologous Vδ6 gene segments; for nomenclature, see Ref. 3), whereas they express GV4 (Vγ5) almost exclusively (8, 9). GV3 (Vγ2) and GV5S1 (Vγ1.1) predominate in the adult thymus and spleen in combination with a variety of Vδ gene segments. The γδ T cells that migrate to distinct epithelial tissues may be suited for different functions. The progenies of early fetal γδ thymocytes with restricted V gene usage may recognize autologous Ags from damaged cells rather than the agent inducing the damage, whereas γδ cells arising later in ontogeny and exhibiting junctional diversification may recognize foreign Ags associated with pathogens.

The first Vδ repertoire study was performed by screening of a cDNA library from CD4⁻CD8⁻ thymocytes with a constant (Cδ) gene probe (10). This study revealed six Vδ gene segments, originally designated Vδ1–Vδ6. Four of these defined novel subfamilies, whereas Vδ3 and Vδ6 were homologous to the previously described Vα6 and Vα7 subfamilies, respectively (75% identity at the nucleotide level being used as the cut-off between subfamilies) (11). Here we adhere to a nomenclature that has subsequently proposed to give the mixed subfamilies a common designation, ADV6 and ADV7, respectively, whereas the unique Vδ subfamilies were designated DV101–DV105 (3). Ever since the initial repertoire study, subsequent studies of polyclonal γδ T cell populations were based on PCR primers specific for the already known Vδ subfamilies and thus were confined to the originally characterized set of V gene segments. The δ locus is nested within the α locus on chromosome 14, between the germline Vα and Jα gene segments. This location should yield access to the large family of Vα gene segments. However, the δ-chain repertoire appears to be limited to the classical Vδ gene segments, DV101–DV105, ADV6, and ADV7. These map in vicinity to the Dδ and Jδ gene segments, with the exception of the ADV7 subfamily, which contains four or five different members that are scattered across the Vα locus (12, 13). Furthermore, rearrangements of a few Vα gene segments have occasionally been reported to occur in γδ T cell clones and hybridomas (summarized in Ref. 3). Moreover, Northern hybridizations of a large panel of neonatal thymocyte hybridomas with Vα subfamily-specific probes revealed expression of Vα gene segments from almost every Vα subfamily in γδ thymocytes (14). It has remained unclear to date whether just a particular subset of Vα gene segments from each subfamily was suited for δ rearrangement and expression in γδ receptors or whether in principle any Vα gene segment was accessible for δ rearrangement.

In this study we have undertaken the first systematic Vδ repertoire analysis throughout ontogeny. Sampling fetal thymocytes from days 14, 16, and 18 of gestation as well as thymuses from newborn and 4-wk-old mice, we have re-examined the thymic Vδ expression waves at the transcriptional level. We used the technique of inverse PCR (iPCR), which permits the rapid amplification and identification of unknown V segments flanking the constant gene Cδ. Moreover, this technique permits quantitative analysis of the Vδ repertoire, because amplification of all V gene segments is performed with the same pair of specific primers. By contrast, using panels of Vδ subfamily-specific primers that may differ in their amplification efficiency would result in biased representation of certain V gene segments. We were thus able to study the differentiation of the Vδ repertoire during ontogeny and to define several novel waves of δ-chain transcription.

Fetal thymuses were obtained from timed pregnancies, with day 0 being the day that a vaginal plug was observed. The newborn mice were used within 24 h of birth. Adult thymuses were obtained from 4-wk-old mice. Thymuses from day 16 of gestation and newborn and adult stages were prepared from BALB/c mice. Thymuses from day 14 of gestation were from C57BL/6 mice, and thymuses from day 18 of gestation were from (C57BL/6 × BALB/c)F₁ mice. BALB/c mice were bred at the Max Planck Institute for Immunobiology (Freiburg, Germany). C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). At all fetal stages analyzed as well as at birth, thymuses from several litters were pooled for isolation of RNA. At 4 wk of age, thymuses from two animals were pooled.

For amplification of flanking regions by iPCR, double-stranded cDNA is circularized such that the unknown Vδ, Dδ, and Jδ gene segments at the 5′ end of the molecule are ligated to the 3′ end of the known Cδ gene segment in a self-ligation step. Outwardly oriented Cδ-specific primers (a 5′ antisense and a 3′ sense primer) were used to amplify around the circle from the Cδ gene segment into the unknown flanking gene segments. The protocol of iPCR has been established for investigating TCR-αβ junctional diversity in human peripheral blood (15) and has been improved at certain steps to make up for the minimal amount of γδ message present in fetal murine thymus (16).

Briefly, RNA was extracted from thymocytes by the acid guanidinium thiocyanate phenol method (17). Five micrograms of total RNA was taken for oligo(dT)-primed double-stranded cDNA synthesis (Choice System, Life Technologies, Gaithersburg, MD). The cDNA was end-polished with 10 U of T4 DNA polymerase (Life Technologies) at 16°C for 5 min and extracted with phenol/CHCl₃. DNA was ethanol-precipitated with 2.5 M NH₄Ac. Circularization was performed in a total volume of 50 μl with 50 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 1 mM DTT, 0.5 mM ATP, and 5 U of T4 DNA ligase (Life Technologies) overnight at room temperature. The whole cDNA was used to perform five iPCR reactions for each stage of development in parallel. PCR was performed in a total volume of 50 μl, containing 67 mM Tris-HCl (pH 8.8), 16.6 mM (NH₄)₂SO₄, 0.1 mg/ml BSA, 2 mM MgCl₂, 0.2 mM dNTP mix, and 0.5 μM of each of the following primers: DCG (antisense primer), 5′-CGA ATC TCC ATA CTG ACC-3′ derived from Cδ exon 1; and DCB (sense primer), 5′-TTA ATG CTC TCC AAG CAG-3′ derived from Cδ exon 4. The samples were overlaid with 30 μl mineral oil. Amplification took place in a Biometra Trio-Thermoblock under the following conditions: 5 min at 95°C, 85°C for addition of 2.5 U of AmpliTaq (Perkin-Elmer, Palo Alto, CA; hot start), 5 min at 95°C, 40 cycles of 1 min at 95°C, 1 min at 55°C, and 1 min at 72°C, followed by a final elongation step of 15 min at 72°C. Five iPCR samples per developmental stage were pooled, treated with 12.5 U of Klenow fragment (New England Biolabs, Beverley, MA), extracted with phenol/CHCl₃, and ethanol precipitated with 2.5 M NH₄Ac. After digestion twice with 40 U of HindIII (New England Biolabs), 1/10th of each pool was size-fractionated on an analytical 1% agarose gel (UltraPure Agarose, Life Technologies) and probed with the internal Cδ-specific oligonucleotide GM11 (5′-CAT GAT GAA AAC AGA TGG-3′). The remaining pool was separated on a preparative 1% low melting agarose gel (SeaPlaque GTG; Biozym, Oldendorf, Germany), and the DNA was cut out. The one or two uppermost blocks were molten, extracted with phenol/CHCl₃, and ethanol precipitated with 2.5 M NH₄Ac.

Size-selected PCR products were ligated into HindIII-cut pBluescript II SK⁺ (Stratagene, La Jolla, CA). Transformed colonies (Escherichia coli strain XL2-Blue; Stratagene) were screened with the oligonucleotide probe GM11, and recombinant plasmid DNA from positive bacterial colonies were sequenced on an ABI automated sequencer with the Taq Dye-Deoxy Terminator Cycle sequencing kit (Applied Biosystems, Foster City, CA) using the Cδ-specific oligonucleotide primer GM11.

The designations used herein for TCR V genes follow the traditional numbering (3, 18) in compliance with the standard nomenclature (WHO-IUIS Nomenclature Subcommittee on TCR Designation) (19). DV101–DV105 designate Vδ gene segments that lack similarity to Vα gene segments. ADV refers to Vα subfamilies that include homologous Vδ gene segments. A/DV refers to germline V gene segments that are expressed in α- and/or δ-chain messages. GV1 and GV2 designate Vγ gene segments originally named Vγ3 and Vγ4 (1), or Vγ5 and Vγ6 (2), respectively. The numbering system for residues of the V segments was previously described (3).

On day 14 of gestation, 27 Cδ-containing cDNA clones were derived from transcripts of complete V-D-J rearrangements that exclusively contained the DV101 gene segment, and 22 (82%) of which were rearranged in-frame (Fig. 1). The majority, 16 of the productively rearranged clones, used the Jδ1 gene segment, whereas Jδ2 was found in only six clones. Half the Jδ1-containing clones had the Dδ2 and Jδ1 gene segments joined in their germline configuration, mediated by a microhomology of 4 nt (AGCT) including two palindromic (P) nucleotides. These clones displayed a 7-nt homology (GATATCG; rather than 5 nt, as suggested in Ref. 20) at the V-to-D joint, with the germline 3′ end of DV101 ending with ATC (3). This predominant rearrangement has previously been reported as the second most frequent rearrangement in s-IEL (8). The remaining Jδ1-containing clones have heterogeneous rearrangements. Five of the six productive Jδ2 rearrangements bear the canonical rearrangement that dominates in s-IEL (4, 8). It is mediated by a 3-nt homology between Dδ2 and Jδ2, resulting in deletion of the 3′ half of the Dδ2 gene segment and, at the V-D junction, by the same 7-nt homology described above. We found this canonical rearrangement in only 5 of 22 DV101 rearrangements (23%). This was unexpected, because it has been shown that the s-IEL precursors reside in the early fetal thymus (21).

In contrast, on day 16 of gestation, 10 of 27 in-frame DV101-containing clones (37%) carried the canonical Jδ2 rearrangement (Fig. 2). This increase on day 16 indicates that the expression of this invariant γδ receptor is not the first event in thymic ontogeny. Rather, it is preceded by a frequent DV101-D-Jδ1 rearrangement that is also present in the s-IEL population as well as heterogeneous Jδ1 rearrangements. This canonical rearrangement persists on day 16 (5 of 14 productive rearrangements involving Jδ1). On day 18, only 8 of 35 complete V-D-J in-frame rearrangements (23%) contained DV101. Of these, four had undergone the canonical rearrangement to Jδ2, and two had undergone the canonical rearrangement to Jδ1 (Table I; data not shown). Thus, the canonical DV101-D-Jδ2 rearrangement increases to reach maximal levels only on day 16, but diminishes dramatically by day 18. In fact, it was undetectable in neonatal thymus, where none of the 64 V-containing clones analyzed carried DV101, casting doubt on the existence of a consecutive, late fetal wave of invariant receptors bearing DV101 paired with GV2.

On day 16 of gestation, 39 of the 45 complete VDJ transcripts analyzed (87%) were derived from in-frame V-D-J rearrangements. About one-third of the productively rearranged transcripts used V gene segments other than DV101. Each three cDNA clones contained ADV7, the ADV17S3, and the DV104 gene segments. In addition, one cDNA clone each was found with a DV4S8-like gene segment, DV6S2, and DV105 (Fig. 2). By day 18, the majority (77%) of the productively rearranged (35 of 43, or 81%, of the V-D-J rearrangements were in-frame) cDNA clones had diverse non-DV101 V segments. Among these, several members of the ADV7 subfamily were found six times, ADV17S3 was found twice, and the DV104 gene segment was found 11 times (data not shown). In neonatal thymocytes, 43 of 64 V-bearing cDNA clones (67%) were rearranged in-frame. Among the productively rearranged cDNA clones, the ADV7 subfamily occurred most frequently (12 clones, or 28%). The members ADV7S1, ADV7S2, and DV7S4 each represented approximately one-third of this subfamily. Dominant expression of ADV7 was followed by DV104 (eight clones, or 19%) and ADV17S3 (five clones, or 12%; Fig. 3). These three subfamilies display a relatively high degree of sequence similarity (≥54% at the nucleotide level, whereas different Vα subfamilies usually are <40% similar) (3). Together, these closely related subfamilies account for 59% of the productive V-D-J rearrangements. Thus, this homologous subset of subfamilies has increased from 23% on day 16 and 54% on day 18 to greater than half of all neonatal cDNA clones (Table I).

The other half used diverse Vα gene segments from six different Vα subfamilies, including the larger subfamilies comprising 5–10 different members. Unexpectedly, the neonatal δ-chain transcripts preferentially used a single gene segment of these multimembered subfamilies. Thus, all three (two productive and one nonproductive) δ rearrangements of members of the Vα4 subfamily contain the same gene segment (Fig. 3). Moreover, the same sequence was identified in the two Vα4-to-δ rearrangements from days 16 and 18 of gestation and in one nonproductive rearrangement from adult thymocytes. The V exon sequence of this novel Vα4 gene segment differs from DV4S8 by 13 nt, and we therefore designate it DV4S8A2. A similarity search of the nucleotide sequence databases revealed that our cDNA does not differ by a single nucleotide from the corresponding V exon of the genomic sequence of the mouse Vα/δ locus. In this genomic sequencing project⁵ Lee and co-workers used cosmid clones that were derived from mouse strain 129/Sv. BALB/c and 129/Sv share the a haplotype of the TCR Vα locus (22). The absence of allelic polymorphism between these two mouse strains permitted us to unequivocally identify the germline counterparts of our cDNA clones. Thus, the DV4S8A2-matching germline gene segment, among all Vα4 subfamily members, maps the most proximal to the δ locus.

Similarly, just one member of the Vα10 subfamily, DV10S7, was used in four neonatal and three day 18 Vα10-to-δ rearrangements, including two nonproductive rearrangements. The one exception was a gene segment identical with ADV10S6 (clone NB55; Fig. 3). The DV10S7 segments of two cDNA clones (1839 and 1840; data not shown) from day 18 fetal thymus, which was derived from (BALB/c × C57BL/6)F₁ mice are identical with that of cDNA KN25-D4, previously isolated as δ-chain message from C57BL/6 mice (23) and therefore designated DV10S7 (Table I) (3). The third day 18 cDNA (clone 1841, out-of-frame) differs from the former two by four nucleotide exchanges in the V segment. The latter bears a V segment identical with those of the four cDNAs isolated from BALB/c neonates. We therefore attribute these minor differences to strain polymorphism and tentatively designate the BALB/c allele DV10S7A2. In fact, the BALB/c-derived DV10S7A2 segment showed a perfect match with one Vα10 germline gene segment from the genomic sequence of the Vα/δ locus (see Footnote 5). Again, the matching member of the Vα10 subfamily is located most proximal to the δ locus.

Among six neonatal Vα2-to-δ rearrangements, including one nonproductive rearrangement (clone NB50; not shown), the ADV2S6 gene segment was used three times (Fig. 3 and Table I). The dominance of ADV2S6 was further quantified by Vα2 subfamily PCR using a V-specific primer conserved in all members of the Vα2 subfamily and an antisense primer specific for Cδ. The result was that 15 of 19 productive transcripts (79%) contained the ADV2S6 gene segment. The second most frequently used segment, which is identical with Vα2.2 (24), was found in only two instances. Only one clone had an out-of-frame joint, also using ADV2S6 (data not shown). A database search identified ADV2S6 as the Cδ-proximal Vα2 subfamily member in the genomic sequence of the α/δ locus (see Footnote 5). This striking result raises the question of whether particular gene segments of this subfamily are targeted for δ rearrangement and others for α rearrangement. The Vα2 subfamily PCR was, therefore, also performed with a Cα-specific primer. The ADV2S6 gene segment was equally prevalent in α-chain message from neonatal thymocytes. Nine of fourteen in-frame rearrangements (64%) had joined ADV2S6 to diverse Jα gene segments, followed by only three Vα2.2 rearrangements (data not shown). These results demonstrate that the same V gene segment can be used for α and δ rearrangements. More importantly, the 3′-most gene segment of the larger Vα subfamilies is preferentially used in neonatal thymocytes.

The usage of the smaller Vα subfamilies containing a single or just two members further substantiates the bias for proximal Vα in neonatal thymus. Vα6 was found five times, four of five clones in-frame, each in neonatal (Fig. 3) and day 18 fetal thymus. BALB/c neonates exclusively expressed DV6S2, the first Vα upstream of the classical Vδ gene segments (see Footnote 5). The two members of the Vα9 subfamily map to the distal and proximal ends of the α locus (13). Our cDNA from neonatal thymus differs from AV9S2 (clone HY-A1) (3) by 4 nt. We, therefore, designate it DV9S2A2. It matches the AV9 gene segment from the proximal end with the exception of 1 nt (see Footnote 5; Table I). This is the one exception where we did not find a perfect match. Finally, we found a new member or allele of the AV18 subfamily, which we named DV18S3 (out-of-frame clone NB60; data not shown). It is identical with the most proximal germline gene segment of the AV18 subfamily (see Footnote 5). Thus, almost all the classical Vα gene segments that are located near the 3′ end were identified in the neonatal δ repertoire.

In the adult, among 50 V-bearing cDNA clones 35 (70%) were rearranged in-frame (Fig. 4). There was a striking dominance of the ADV7 subfamily, which was found in 14 (40%) of the productively rearranged transcripts. Whereas the second wave of diverse neonatal Vα-to-δ rearrangements is totally diminished in 4-wk-old thymus, the contribution of productive ADV7 transcripts steadily increases from 8% (3 of 39) on day 16 of gestation to 17% on day 18 (6 of 35), 28% (12 of 43) at birth, and 40% (14 of 35) in 4-wk-old thymocytes (Table I). On day 16, DV7S6 was found twice. It most likely represents the BALB/c allele of ADV7S1, and we refer to it as DV7S1A2. Another member DV7S4 was found in one clone. On day 18, DV7S1A2 was found once, ADV7S3 was found three times, and DV7S4 was found once. ADV7S3 was exclusively found in day 18 fetal (BALB/c × C57BL/6)F₁ mice, suggesting that it may represent the C57BL/6 allele of DV7S4. In neonates, we found four DV7S1A2, three ADV7S2, three DV7S4, and one DV7S5 clone. Based on sequence similarity, the ADV7 subfamily can be subdivided into two subsets, ADV7S1/2 and ADV7S3/4/5, with mean nucleotide sequence similarities of 95% within and 80% between subsets. Including nonproductive rearrangements, the subfamily members ADV7S1/2 and ADV7S3/4/5 were represented with equal frequency, whereas among the in-frame transcripts the S1/2 to S3/4/5 ratio was 7:5. By contrast, in day 18 fetal (BALB/c × C57BL/6)F₁ thymocytes the in-frame transcript S1/2 to S3/4/5 ratio was 1:5 (Table I). Thus, there is selective expression of different subsets in the two mouse strains. All members of the ADV7 subfamily are used, in contrast to the preferential usage of members from the proximal cluster in the neonatal wave of diverse Vα subfamilies. Moreover, the overexpressed ADV7S1/2 gene segments map to the more distal Vα clusters (12), whereas DV7S4 is identical in nucleotide sequence to the most proximal member in the genomic sequence of the α/δ locus (see Footnote 5). The ADV7 subfamily is thus unique in that even at early stages in ontogeny its proximal member is not preferentially used.

The two subfamilies, DV104 and ADV17, which share a high degree of similarity with ADV7, closely follow the expression pattern of the ADV7 subfamily. On day 16, each of these starts with a frequency of 8% (3 of 39 productive transcripts). In neonatal thymus DV104 has increased to 19% (8 of 43), reaching 23% (8 of 35) in the adult thymus. Of interest, in (BALB/c × C57BL/6)F₁ mice, DV104 has already reached 31% (11 of 35) by day 18. Thus, whereas the ADV7 subfamily dominates in BALB/c mice, DV104 appears to be overselected in C57BL/6 mice. Using a V region-specific mAb to DV104, variation in the relative levels of γδ splenocytes expressing DV104 was reported for these strains of mice (25). Selection of DV104⁺ cells appears to be tissue specific and is linked to the particular GV/DV104 pair expressed at the surface of these cells (26). In neonatal thymus we found two productive transcripts of a hitherto undescribed second functional member of the DV104 subfamily, which we refer to as DV104S2. It is located 10 kb upstream of the DV104S1 gene segment (see Footnote 5). ADV17 is the least frequent of the three homologous subfamilies. The exclusively used ADV17S3 segment amounts to 6% of the total productive rearrangements (2 of 35) on day 18, 12% (5 of 43) in newborn, and 9% (3 of 35) in 4-wk-old thymus. It may be worth noting that we found one pseudogene, DVX, transcribed at significant rates. Its 3′ end shares 68% identity with the ADV7S1 gene segment. DVX extends the sequence of the previously described cDNA Z78 (10). The major portion of the V gene segment, including Cys²², appears to be deleted, but its leader exon and 5′-flanking region are conserved. It shows an expression pattern similar to that of the clan subfamilies. It cannot be excluded that minor differences in gene expression levels or RNA stability will lead to pattern distortions. Therefore, eventual confirmation by cell surface staining will be required to confirm the observed pattern of TCR δ-chain transcription.

The first fetal thymic wave expresses GV1 paired with a δ-chain composed of DV101, Dδ2, and Jδ2 segments. Both chains are encoded by canonical rearrangements. Using an mAb directed to an epitope on the GV1 gene product, it has been shown that the total number of GV1⁺ cells peaks on day 16, and the cells have disappeared by day 18. By subtraction of this population from the total number of CD3⁺ cells, a second wave of γδ T cells using Vγ segments other than GV1 with a maximum on day 18 became apparent (27). However, the lack of GV2-specific Abs prevents a direct comparison between the GV1 and GV2 waves (28). Our results revealed that the transcription of the canonical DV101-Dδ2-Jδ2 rearrangement is confined to a narrow time window between days 14 and 18 of gestation, with a maximum on day 16 (Fig. 5). Thus, there is a good concordance between the extent of canonical DV101 rearrangement and the abundance of GV1⁺ cells. Unexpectedly, we did not observe a late fetal wave of DV101 expression that could provide the δ-chain for pairing with GV2. We cannot rule out the possibility that the changes in the DV101 expression level observed by us during fetal development might reflect strain-dependent differences. Thus, strain-specific developmental changes in the thymic environment may play a role in shaping the GV1/DV101 TCR repertoire (29). Yet, our data are in good agreement with more recent quantitative PCR analyses of genomic DNA rearrangements. GV1 and GV2 rearrangements were shown to be relatively infrequent on day 14, and their abundance increases about 20-fold by day 15. Both GV1 and GV2 rearrangements decrease by day 18, whereas GV3 rearrangements increase to reach a maximum on day 18 (30). Moreover, this study shows that the relative frequencies of the different Vγ rearrangements correspond reasonably well with the frequencies of Vγ⁺ cells at the different time points. Thus, our observation of a decline in canonical DV101 expression on day 18 is corroborated by the reported coincident decline in GV1 and GV2 rearrangements. It has also been demonstrated by others that all GV2⁺, but none of the GV2⁻, hybridomas carried GV1 rearrangements (31), indicating that GV1⁺ and GV2⁺ T cells belong to a common, distinct lineage that does not give rise to other γδ T cells. Most of the GV2⁺ cells carried out-of-frame GV1 rearrangements and, thus, appear to have a second chance to rearrange GV1 or GV2 on the other allele. These data are consistent with a major first wave consisting of two subsets expressing GV1 or GV2 and the same DV101 rearrangement. The onset of canonical DV101-D-Jδ2 rearrangement is on day 14. Yet, it is outnumbered by heterogeneous rearrangements to Jδ1 with one canonical DV101-D-Jδ1 rearrangement predominating (Fig. 5). Thus, the canonical DV101-D-Jδ2 rearrangement is not the first event in thymic ontogeny. In comparison, the canonical GV1 rearrangement amounts to 81% of the productive GV1 rearrangements on day 14 (32). Similarly, 87% of the productive GV2 rearrangements are of the canonical type (32), suggesting that the onset of canonical GV2 expression occurs as early as that of GV1. Taken together, we conclude that the previously postulated two consecutive waves of canonical GV1 and GV2 expression completely overlap and coincide with the short wave of canonical DV101 expression. This suggests that on day 16 of gestation the thymus harbors the maximal number of precursors of both s-IEL and r-IEL.

The genomic organization of the α/δ locus with the δ locus nested between the Vα and Jα gene segments raises the question of whether the Vα repertoire is completely or partially accessible for functional δ rearrangement and, consequently, expression in γδ receptors. We have shown for three large Vα subfamilies that just a single member of each subfamily preferentially undergoes δ rearrangement in neonatal thymocytes. Previously, occasional Vα-to-δ rearrangements have been described (for review, see Ref. 3). The prevalent Vα gene segments of our study precisely match those described in the literature. Thus, the exclusively used DV4S8A2 gene segment differs from DV4S8 by only five amino acid residues (3). This limited extent of variation may be ac- counted for by allelic polymorphism between the BALB/c mouse strain used in this analysis and the C57BL/6 strain from which the CD4⁻CD8⁻ γδ thymocyte hybridoma expressing DV4S8 was derived (33). Similarly, DV10S7A2 from BALB/c neonates in this study differs by four amino acid replacements from DV10S7 (3), previously isolated from a CD4⁻CD8⁻ γδ thymocyte hybridoma from C57BL/6 mice. Moreover, the occurrence of both forms in fetal thymus on day 18 of gestation in (BALB/c × C57BL/6)F₁ mice further argues for allelic variation and, therefore, prevalence of one Vα10 gene segment in seven of eight clones (Table I). Finally, the predominantly used Vα2 gene segment is identical with AV2S6 previously isolated from BALB/c mice (3). It differs from DV2S8 by four residues, indicating that these may represent allelic counterparts rather than two different subfamily members. DV2S8 was previously found to be expressed in γδ thymocyte hybridomas from C57BL/6 neonates (34). Taken together, if one takes into account allelic variation, our systematic analysis combined with occasional previous findings reveals that just a single V gene segment from a given Vα subfamily is used in γδ receptors.

There are several possibilities to explain the observed bias in Vα usage in γδ thymocytes. Regulatory sequences may dictate α-ness vs. δ-ness of individual subfamily members. Comparative analysis of the 5′- and 3′-flanking sequences of the highly homologous Vα2 subfamily genes, however, did not reveal any significant differences between ADV2S6 and its counterparts (24) (ADV2S6 is identical with the segment designated Tcra V2.6 herein). Therefore, differential activation of promoter regions or targeting of recombination signal sequences in the vicinity of the coding sequence can be excluded. This possibility was unlikely, given the fact that the same V gene segment, ADV2S6, undergoes α and δ rearrangements with equal relative frequencies. Instead, positive thymic selection may choose certain specificities of γδ receptors encoded by particular Vα subfamily members. The occurrence of the same Vα gene segments in nonproductive rearrangements, however, argues against positive selection being solely responsible for overexpression. Based on our comprehensive sequence analysis, we, rather, conclude that proximity determines Vα rearrangement to genes in the δ locus. A similar mechanism is thought to control the transcription of the globin genes that are closest to the locus control region early in development (35). Comparison of our δ-chain clones expressed in BALB/c mice with their genomic counterparts (see Footnote 5) in all instances revealed a perfect match, with only one exception. The absence of polymorphic nucleotide substitutions permitted us to identify their genomic location. Adjacent to the classical Vδ gene segments DV101, 102, 104, and 105 that appear to be exclusively associated with Cδ is located the δ-proximal Vα cluster, one of multiple duplicated clusters, each containing intermingled single representatives of different Vα subfamilies. Among the 10 furthest proximal Vα gene segments, pseudogenes not included, seven are represented in the neonatal thymic δ repertoire: DV6S2, DV9S2A2, ADV2S6, DV10S7A2, ADV17S3, DV18S3, and DV4S8A2, from 3′ to 5′ (ADV designating V segments that undergo α and δ rearrangement; Fig. 6). ADV11S5 was not identified in this study, but has previously been isolated from a cytolytic CD4⁻CD8⁻ γδ T cell clone from peripheral lymph nodes of BALB/c nu/nu mice (36). We thus observed an almost perfect coincidence of Vα subfamily members mapping to the proximal cluster and their expression in newborn thymus δ-chain message. By contrast, if one compares Vα usage in TCR-αβ repertoire development in human thymus and hemopoietic organs from 14- and 15-wk-old fetuses (37) with their genomic location in the human Vα/δ locus (see Footnote 6), Vα gene segments from the 5′ half of the Vα/δ locus appear to be preferentially used, in particular all Vα gene segments from the distal end. Together, this may reflect a bidirectional readout mechanism, from the 5′ end for α-chain rearrangements and from the 3′ end for δ-chain rearrangements, at distinct stages in ontogeny.

The functional significance of this shared repertoire of V genes is not known. The mouse α locus has evolved through several rounds of duplication, such that the proximal duplication unit includes a subset of genes representative of most subfamilies. Preferential readout of the proximal cluster may thus permit the neonate, despite its limiting number of T cells, to build up a diverse spectrum of specificities against a broad range of pathogens. This would require extensive cross-reactivity to fill the holes in the repertoire. In fact, the mean CDR3 length of the δ rearrangements from the proximal Vα cluster is nine amino acids, three residues shorter than that of the other neonatal δ rearrangements (Fig. 7; see below), which could reflect a low affinity repertoire of this subset. Increased promiscuity for antigenic peptides has been demonstrated in the neonatal αβ T cell repertoire (38). Amplification of α-chain transcripts from neonatal thymus with a Vα2 subfamily-specific primer also yielded a mean CDR3 loop length of nine amino acids (data not shown). Recently, crystal structure analysis of a δ-chain with a CDR3 length of 10 amino acids revealed that in terms of its relative position in the TCR combining site, its CDR3 resembles that of Vα and forms a flat binding surface, as observed in αβ TCRs (39). We have previously provided evidence that the neonatal δ-chain CDR3 regions, displaying limited junctional diversity, may be more flexible, changing conformation to associate with various Ags (40).

The neonatal wave of diverse Vα-to-δ rearrangements has totally vanished in 4-wk-old thymocytes (Fig. 5). Just a few V gene segments dominate in the adult thymus. Thus, the frequency of productively rearranged ADV7 transcripts increases continuously from day 16 of gestation, reaching a plateau of 40% at 4 wk of age (Fig. 5). In-frame rearrangements of the subfamily members of ADV7S1/2 outnumber those of ADV7S3/4/5 (see Results). We have previously demonstrated that positive selection plays a role in the overexpression of the ADV7S1/2 subset (40). The DV7S4 gene segment is located most proximal of all ADV7 subfamily members (see Footnote 5; Fig. 6). Its location in the proximal Vα cluster should favor preferential utilization in the δ repertoire and is thus in striking contrast with the observed counterselection. The other two members, ADV7S1 and ADV7S2, are expressed with equal frequency. They are interspersed with different subfamilies throughout the α locus (12). Thus, neonatal expression of the ADV7 subfamily members does not show the position-dependent bias of the other Vα subfamilies. A similar transcription pattern of the ADV7-like pseudogene DVX indicates that preferential expression of the clan subfamilies may be regulated at the DNA level before selection at the protein level. More studies need to be performed regarding the role of their flanking sequences in promoting preferential δ rearrangement. Biased expression of IgH J-proximal V genes has been demonstrated in neonatal mice (41), although the mouse model is controversial (42, 43). Interestingly, not all human V_H segments preferentially expressed during early stages of ontogeny map in the proximity of the J_H segments (44). A defect specific to the rearrangement of the J_H-distal V_H gene segments has been described in mice that lack the α-chain of the IL-7R (45). The defect correlates with reduced expression of Pax-5, a transcription factor that is required for V-to-DJ_H recombination in B cells (46). Furthermore, differential targeting of TCR Vγ genes for rearrangement has recently been shown to be controlled by sequences immediately upstream of the Vγ gene segments (47). The three subfamilies, ADV7, DV104, and ADV17, display high sequence similarity with one another (>54%) and clearly represent a distinct subset of Vα/δ subfamilies. Expression of this clan of homologous subfamilies increases from 23% on day 16 of gestation to 59% in neonatal and 72% in adult thymus (Fig. 5).

A comparison of the mouse and human Vα/δ subfamilies reveals a number of human homologues of the mouse clan subfamilies (hDV101, hADV6, hAV12, and hADV14) (48). The human clan counterparts are equally scattered across the Vα locus (see Footnote 6). All these are expressed in γδ thymocytes and in γδ cells of the intestinal mucosa, with prevalence of hDV101 (49, 50). The CDR1 and CDR2 lengths of the clan subfamilies are each increased by one residue, on the average, compared with the other Vα subfamilies. An extension of the 3′ ends of the clan germline V gene segments, encoding two additional residues, results in increased CDR3 length. In neonatal thymus, the clan members have an average CDR3 length of 12 residues as opposed to nine residues for the δ rearrangements from the proximal Vα cluster (Fig. 7). PCR with primers specific for the ADV2 and ADV7 subfamilies, which dominate the proximal Vα and clan repertoires, respectively, also yielded mean CDR3 lengths of 9 and 12 residues, respectively (data not shown). This is in part due to usage of the Dδ1 gene segment, while the neonatal δ rearrangements from the proximal Vα cluster lack Dδ1 and also to the presence of N-nucleotides and their virtual absence in the Vα-to-δ rearrangements. The shorter CDR3 of the Vα-to-δ rearrangements may establish a low affinity repertoire useful for a first defense, whereas the longer clan CDR3 may provide a more specific response against common pathogens. Within the highly variable CDR2 of most Vα subfamilies, there is a functionally conserved motif of residues with alternating charge, called the KEK motif (51). It has been proposed that the positive net charge of the KEK motif, which is located on the lateral surface of the Vα domain, may be responsible for binding to the CD8 coreceptor (51). The motif is absent in all clan V gene segments, with the exception of ADV7S1 (3). Instead, in the clan subfamilies ADV17 and DV104 (as in their human homologues hADV14 and hADV6) (18) this putative protein-protein interaction motif is replaced by a site (NXT) for potential N-linked glycosylation. A carbohydrate moiety would prevent interaction with the CD8 coreceptor. It is, therefore, conceivable that usage of the clan subfamilies defines a lineage of γδ T cells that lack CD8 coreceptors. Of note, γδ i-IEL expressing the human clan homologue DV101 were recently shown to recognize the MHC class I-related molecules MICA and MICB, which lack a CD8 binding site (52). We are currently investigating whether usage of the clan and the proximal Vα subfamilies correlates with the absence or the presence, respectively, of the CD8 coreceptor in γδ thymocytes as well as in γδ i-IEL.

We thank Martin Selbert for his invaluable help with automated sequencing, and Constanze Taylor for help with the manuscript. We also thank Profs. D. Petzoldt and H. Näher (Department of Dermatology, University of Heidelberg) for their support.