Distinct Stage-Specific cis-Active Transcriptional Mechanisms Control Expression of T Cell Coreceptor CD8α at Double- and Single-Positive Stages of Thymic Development¹

The αβ-TCR expressed by the majority of peripheral T lymphocytes recognizes foreign Ag associated with self-MHC molecules (1, 2, 3). The mutually exclusive pattern of expression of the coreceptor molecules, CD8 and CD4, distinguishes the two sublineages of αβ-TCR T cells: CD8⁺CD4⁻ cytotoxic T cells, which recognize Ag bound to class I MHC, and CD8⁻CD4⁺ helper T cells, which recognize Ag bound to class II MHC (4, 5, 6). Numerous studies have shown that CD8 and CD4 play crucial roles in the development and function of αβ-TCR T cells as a result of binding of nonpolymorphic regions of class I and class II proteins, respectively, and possibly through signaling (5, 7, 8, 9, 10, 11).

Developing thymocytes, which give rise to mature CD8⁺CD4⁻ and CD8⁻CD4⁺ T cells, go through progressive stages of CD4 and CD8 expression in the thymus, from initially expressing neither (the double-negative (DN)⁴ stage), to coexpressing both (the double-positive (DP) stage), to expressing one or the other (the single-positive (SP) stage) (5, 12, 13). Loss of CD8 or CD4 expression during the DP to SP transition is associated with thymic selection for recognition of self-MHC, further differentiation into mature thymocytes, which express increased levels of the αβ-TCR/CD3 complex, commitment to the cytotoxic or helper sublineage, and induction of sublineage-specific (i.e., CD8⁺/cytotoxic vs CD4⁺/helper) patterns of gene expression (12, 13, 14, 15). As the pattern of coreceptor expression is so strongly correlated with differentiation and selection, delineation of the underlying transcriptional control mechanisms should contribute to our understanding of T cell development and commitment to the helper or cytotoxic sublineage. However, the requirement that the CD8 and CD4 genes be activated and coexpressed in the same cell during the early to intermediate stages of thymic differentiation (i.e., the DP stage) yet be expressed in mutually exclusive sublineages at a later stage (i.e., the SP stage) presents a unique problem in gene regulation. Further, it is unclear at what level the underlying regulatory mechanisms are influenced by the events of T cell differentiation and selection, although a role for the Notch signaling pathway has been suggested (16).

Studies of mouse CD4 regulation have identified transcriptional enhancers at ∼13 kb (17, 18) and 24 kb (19) upstream and a negative-acting element within the first intron (20, 21). At least some of these elements are also associated with the human CD4 gene, although their locations may not be absolutely conserved (22, 23, 24). While initial identification of most of these elements involved DNase I-hypersensitive site (HSS) mapping and transfection assays, their functional roles in vivo have been confirmed in transgenic mice. DNase I hypersensitivity and transfection studies have also been used to identify candidate cis-active elements located nearby or within the CD8α gene (25, 26, 27). However, what role these elements may play in vivo is unclear.

Several aspects of CD8α expression suggest that its regulatory mechanisms may be more complex than those of CD4. For example, in thymus-derived αβ-TCR cells, CD8 is invariably expressed as a heterodimer composed of the products of two linked genes, CD8α (or Lyt2) and CD8β (or Lyt3), while CD4 is expressed only as a homodimer encoded by a single gene (5). Coexpression, combined with the close chromosomal linkage, suggests that both genes may share cis-active regulatory mechanisms. However, coexpression is not always the case, as most mouse intestinal intraepithelial T lymphocytes (iIELs) which express the γδ-TCR express CD8α in the absence of CD8β (i.e., CD8α⁺β⁻ (28, 29)). Also, within αβ-TCR iIEL cells, both CD8α⁺β⁻ and CD8α⁺β⁺ cells are present in the CD4-negative subpopulation, while CD4-positive cells are either CD8α⁺β⁻ or CD8α⁻β⁻ (28, 29). Thus, differential regulation of CD8α and CD8β does occur in some cell types.

To begin to delineate the mechanisms underlying stage-specific regulation of CD8α/Lyt2 expression in thymic differentiation, we have studied a series of CD8α/Lyt2 gene constructs in thymocytes and T cells of conventional and CD8α-deficient/“knock-out” transgenic mice. While these studies were in progress, Hostert et al. (30) reported that the CD8α gene in a P1 bacteriophage clone containing about 80 kb of the CD8 locus was expressed in a tissue- and sublineage-specific manner in transgenic mice. Rather than starting with such large DNA segments, our approach toward identifying important CD8 cis-active regulatory elements was to examine the effect on expression of adding defined segments of native 5′ and/or 3′ flanking DNA to a cloned genomic CD8α gene construct, which on its own was not expressed in transgenic mice. Here, we show that while constructs with up to 12 kb of 5′ and 4.5 kb of 3′ flanking DNA were not expressed in T or non-T cells of transgenic mice, stage- and sublineage-specific expression of CD8α in thymic-derived αβ-TCR T cells was observed upon inclusion of an additional 5′ 40-kb segment of the CD8 locus. Further studies aimed at determining the location and mechanism of action of the relevant cis-active element(s) within this 40-kb 5′ region uncovered at least two distinct stage-specific CD8α cis-acting regulatory mechanisms, one that mediates expression in developing thymic αβ-TCR T cells up to an intermediate DP stage and another that acts after this stage in SP thymocytes and peripheral T cells.

The structure of the mouse CD8 locus, including the genes for CD8α/Lyt2 and CD8β/Lyt3, is shown schematically in Figure 1. The fragments used for microinjection were generated as described below. Fragment A was a 5.1-kb HindIII-SalI fragment containing a hybrid Lyt2.1/Lyt2.2 gene with ∼0.3 and 0.7 kb, respectively, of 5′ and 3′ flanking DNA (31). This fragment also contained 0.28 kb of plasmid vector DNA at the 3′ end, which was used as a transgene-specific probe for identifying transgenic mice. The fragment A construct was generated by first subcloning the 5.2-kb HindIII fragment of the mouse Lyt-2.2 gene into pBR322. A 2.3-kb 5′ HindIII-XbaI fragment containing the first three exons of the Lyt-2.2 gene was replaced with a 2-kb 5′ HindIII-BamHI fragment of the Lyt-2.1 gene (which also contained the first 3 exons) after blunt-ending BamHI and Xbal sites of each fragment. This manipulation generated a hybrid DNA molecule consisting of the first three exons of the Lyt-2.1 gene and the last two exons of the Lyt-2.2 gene. Since the first three exons of the gene encode the protein region conferring the serologic difference between the Lyt-2.1 and Lyt-2.2 molecules, the protein made from the hybrid DNA is recognized by a mAb against Lyt-2.1 (32, 33). As an additional means of distinguishing the transgenic and endogenous genes, a fragment of ∼300 bp was deleted from intron 3 of the hybrid construct. PCR of tail or PBL DNA with primers flanking this region amplified a fragment of ∼600 bp from the endogenous gene and 300 bp from the transgene. The primers in intron 3 used for this purpose were 5′-GGGCGTTCCAGCTGACCTATAG and 5′-TGGAAGGCAGAGGCAGGCGGAT (32).

Fragment B was a 12-kb BamHI fragment with ∼7 and 0.7 kb of 5′ and 3′ flanking DNA. This fragment was generated from the fragment A construct by replacing the 2-kb Lyt2.1 HindIII-BamHI fragment with a 9-kb Lyt2.1 BamHI-BamHI fragment. This 9-kb fragment was inserted at the same XbaI site by blunt-end ligation. The injected fragment retained the 300-bp deletion in intron 3, but contained no vector DNA. Transgenic offspring were identified by PCR analysis of tail or PBL DNA.

Fragment C was a 22-kb SalI-SalI fragment with ∼12 and 4.5 kb of 5′ and 3′ flanking DNA. The generation of this construct involved multiple steps, the first one being addition of a 4.5-kb 3′ flanking fragment, which extended to a 3′ PvuII site, to the 5-kb Lyt2.1/Lyt2.2 hybrid gene (from fragment A above), through a shared SacI site in the 3′ untranslated region. A 0.6-kb SacI fragment from the 3′ untranslated region of the human CD4 cDNA was subsequently inserted into this SacI site for use as a specific probe for transgenic DNA and RNA. Finally, the 5′ segment up to an internal KpnI site of this construct was replaced with a SalI/KpnI 5′ flanking/coding fragment, originally derived from a phage clone from a Lyt2.1 (DBA/2) genomic library (32), by ligation through the shared KpnI site. The resulting purified fragment used for microinjection contained no vector DNA.

Fragment D was a SalI-linearized cosmid (cos Ly3–17) isolated from a genomic library from mice expressing Lyt3.2 (34). This fragment contained the complete Lyt3/CD8β gene, ∼1.5 kb and 25 kb of 5′ and 3′ flanking DNA and overlapped the Lyt2.1 fragment C construct by 2 to 3 kb. Fragment D contained vector pTL5 DNA at its 3′ end. Fragment E was a 23-kb ScaI fragment purified from the above cos Ly3–17 clone. Fragment F was a 17-kb BamHI fragment subclone from a genomic library from mice expressing Lyt3.1.

DNA fragments of the above described constructs were purified and microinjected into fertilized one-cell (B6SJL)F₂/J embryos for generation of transgenic mice as described (35). Fragments A, B, or C were microinjected at 2 to 5 ng/μl. For coinjection of fragment C together with fragment D, E, or F, both fragments were mixed at a final concentration of ∼1 to 2 ng/μl each. Transgenic mice and progeny were named as described previously (36). When more than one integration was identified among offspring of a single founder, the individual integrations were maintained separately in offspring and represented as A, B, etc. (e.g., 24A). Transgenic lines were established by breeding with nontransgenic C57BL/6J or (B6/SJL)F₁ (Lyt2.2) mates. The fragment (C+D) line 24A transgenic/CD8α knock-out line was generated by repeated crossing of line 24A offspring with mice that were homozygous for the Lyt2/CD8α mutation (9).

Tail skin DNA was prepared from 3-wk-old mice (37) and analyzed directly by hybridization for transgenic DNA by dot blotting and/or by Southern blot hybridization (36, 38).

Total RNA was prepared from mouse tissues by extraction with guanidinium isothiocyanate as described (35, 38, 39). RNA concentrations were determined by measuring the absorbance of 260 nm. Purified RNAs were denatured in 2.2 M formaldehyde and 50% formamide at 65°C, electrophoresed in 1% agarose gels containing 1.1 M formaldehyde and (1 × MOPS), and transferred to nitrocellulose (38). Hybridization of DNA and RNA blots was at 42°C in 50% formamide/5× SSC, as described (36), for 24 to 48 h with DNA probes labeled by the random primer technique to a sp. act. of 4 to 8 × 10⁸ cpm/μg. The final blot washing conditions were 0.1× SSC/0.1% SDS at 55 to 60°C. Blots were exposed to Kodak XAR-5 film with intensifying screens at −70°C.

The method of detection and hybridization probes used for identification of transgenic DNA were as follows: for fragment A, dot and Southern blot hybridization with a pBR322-derived probe; for fragment B, PCR analysis of tail or PBL DNA; for fragment C, dot and Southern blot hybridization with the human CD4 0.6-kb SacI fragment inserted into the 3′ untranslated region of the Lyt2.1 gene; for fragment D, dot and Southern blot hybridization with a vector-derived probe; for fragment E or F, single copy probes located within these fragments were used to confirm cointegration with fragment C by the appearance of bands of increased intensity and/or unique junction fragments. Lyt2.1 (fragment C+D) transgenic mice that were homozygous or heterozygous for the disrupted or wild-type endogenous CD8α gene were distinguished by the unique band pattern obtained for Southern blot hybridization of EcoRI-digested tail DNA with a Lyt2 cDNA probe.

For fragments A, B, and C, transgenic RNA was tested for by RT-PCR using primers flanking the position of a polymorphic KpnI site in the transmembrane-encoding exon. The primers, located in exon 1 (5′-ACAACAAGATAACGTGGGACGA) and in the 3′ untranslated region of exon 5 (5′-GTAGTAGTTGTAGCTTCCTGGCG), are conserved between both the transgenic Lyt2.1 and endogenous Lyt2.2 alleles (32) and amplify a band of 516 bp from reverse-transcribed RNA products from thymus and lymphoid tissues from nontransgenic Lyt2.1 (DBA/2J) and Lyt2.2 (B6 or B6/SJL) mouse strains. As some Lyt2.1 strains, including the Lyt2.1 transgene constructs A, B, and C, contain a single nucleotide substitution that creates a KpnI recognition sequence, the amplified band can be cut to two smaller fragments of 339 and 177 bp. Since the strains used for transgenic production and breeding, B6/SJL and B6, do not contain this polymorphism, RT-PCR products from lymphoid tissues of these mice are not cleaved by KpnI. For transgenic mice, endogenous Lyt2.2 and transgenic Lyt2.1 KpnI-treated RT-PCR products were distinguished by size on 2.2% agarose gels by ethidium bromide staining and by blot hybridization with an oligonucleotide probe end-labeled with [γ-³²P]ATP (NEN, Boston, MA; 6000 Ci/mmol) and T4 polynucleotide kinase (Life Technologies, Gaithersburg, MD) (38). This oligonucleotide (5′ ATCAAGGACAGCAGAAGGGCCA) is from exon 3 and is conserved between both the Lyt2.1 and Lyt2.2 alleles (Lyt2.1/2.2 probe) (32). Hybridization was at 53°C in 50% formamide/5× SSPE with final blot washing conditions of 0.1× SSC/0.1%SDS at 65°C.

Transcription of the fragment C transgene was also analyzed by Northern blot hybridization using the human CD4 3′ untranslated “tag” as a transgene-specific probe. In some cases, blots were stripped as described (36) and analyzed for endogenous Lyt2 expression using a full-length Lyt2 cDNA probe. Quantitation of Northern blot and RT-PCR blot hybridization signals was done by PhosphorImaging (Molecular Dynamics, Sunnyvale, CA).

Flow cytometry was conducted with cells prepared from the indicated tissues. Thymocytes were prepared as described (36, 40), and splenic and lymph node lymphocytes were purified by passage over Lympholyte-M (Cedarlane, Hornby, Ontario, Canada). Indirect immunofluorescence was conducted by incubating 1 to 2 × 10⁶ cells with primary mAbs, followed by appropriate secondary reagents. Cells (10⁴) for each sample were analyzed on a FACScan flow cytometer (Becton Dickinson). The percentage of positive cells was calculated using FACScan research software programs. mAbs used were as follows: anti-Lyt2.1 (Cedarlane; or American Type Culture Collection (ATCC), Manassas, VA, ATCC HB129 (33)); anti-Lyt-2.2 (ATCC TIB210/mAb 2.43 (41)); anti-Lyt2/CD8 (biotin conjugate from Becton Dickinson, reacts with both Lyt2.1 and Lyt2.2); anti-CD4 (phycoerythrin conjugate from Becton Dickinson; biotin conjugate from PharMingen, San Diego, CA); anti-CD3ε (biotin conjugate from PharMingen; Quantum Red conjugate from Sigma, St. Louis, MO). Staining with unconjugated mAbs was detected with fluorescently labeled secondary Ig reagents (Accurate Scientific, Westbury, NY), as indicated. Biotin-conjugated Abs were detected with streptavidin-labeled-FITC (Life Technologies) or phycoerythrin (Becton Dickinson).

To identify cis-acting transcriptional control mechanisms that mediate expression of CD8α in developing thymocytes and T cells, we studied the expression in transgenic mice of a series of mouse CD8α gene constructs containing increasing amounts of native 5′ and/or 3′ flanking DNA. To distinguish transgenic and endogenous CD8α, we took advantage of the existence of two CD8α alleles, using transgene constructs encoding the Lyt2.1 allele for microinjection of F₂ hybrid embryos from breedings of C57BL/6J and SJL/J strains, both of which carry the Lyt2.2 allele (42). Although the Lyt2.1 and Lyt2.2 alleles can be distinguished with mAbs (32), additional modifications were engineered into the Lyt2.1 transgene constructs to facilitate specific detection of transgenic DNA and RNA (see below, Materials and Methods and legend to Fig. 1).

Figure 1 gives a schematic representation of the CD8 locus on chromosome 6 in the mouse (34). Initial studies were conducted with a 5.1-kb and a 12-kb construct (fragments A and B in Fig. 1). Both contained the intact Lyt2.1 gene and had ∼0.7 kb of native 3′ flanking DNA, but differed in the amount of flanking 5′ DNA. Fragment A contained 0.3 kb of native 5′ DNA, while fragment B contained ∼7 kb of 5′ flanking DNA. Seven founder mice were generated with fragment A, while 10 were generated with fragment B. Founders were bred with C57BL/6J or (B6/SJL)F₁ mates, and transgenic offspring were identified and characterized. The transgene copy number for fragment A and B lines ranged from ∼1 to 30 copies per cell (not shown). Flow cytometric analysis of thymus, spleen, and lymph node cells from offspring of each line with a Lyt2.1-specific mAb failed to detect expression in any of these tissues (not shown). As Lyt2.1 mRNA expressed from either construct would not be distinguishable from endogenous Lyt2.2 mRNA by Northern blot hybridization, we used RT-PCR in conjunction with an allelic polymorphism to test for transcription of the Lyt2.1 transgene (see details in Materials and Methods). A nucleotide substitution in exon 3 of the Lyt2.1 allele generates a KpnI restriction site that is absent from the Lyt2.2 allele (32). While the amplified product of 516 bp observed for lymphoid tissues from a nontransgenic DBA/2J mouse (endogenous allele is Lyt2.1) could be cut by KpnI to give two bands of the expected sizes of 339 and 177 bp, the 516-bp RT-PCR band observed for lymphoid tissues from nontransgenic B6/SJL or B6 mice (endogenous allele is Lyt2.2) was not cut with KpnI (not shown). Using this assay, we were unable to detect transgenic Lyt2.1 expression in lymphoid or nonlymphoid tissues from any of the fragment A and B lines, even when the products were analyzed by blot hybridization with a ³²P-labeled oligonucleotide probe conserved in both alleles (not shown).

Failure of the fragment A and B Lyt2.1 transgenes to be expressed was not due to a problem in transcription, processing, or translation, because both constructs were expressed at the level of RNA and surface protein in transfected fibroblasts (not shown) and T cell lines (Ref. 43, and unpublished results). Further, in additional transgenic experiments, a Lyt2.1 construct consisting of the same coding region but with the 5′ flanking region replaced with that of the β-actin gene (44) led to high level constitutive expression in most tissues of all lines (31). Thus, the likely explanation for lack of expression of fragment A or B in transgenic mice was that essential cis-acting transcriptional control sequences located further 5′ and/or 3′ were not present within these fragments. To test for such sequences, a 22-kb Lyt2.1 construct with ∼12 kb of 5′ and 4.5 kb of 3′ flanking DNA (fragment C in Fig. 1) was used to produce transgenic mice. Eight transgenic lines were generated and analyzed for expression by flow cytometry, Northern blot hybridization, and RT-PCR. As for fragments A and B, we were unable to detect expression in either lymphoid or nonlymphoid tissues of any of these lines. The absence of even low levels of transgenic RNA indicated that essential cis-active transcriptional regulatory sequences were located further than 12 kb upstream and/or 4.5 kb downstream of the gene.

As shown schematically in Figure 1, the gene for mouse CD8α lies ∼36 kb 3′ of the gene for CD8β (34). As CD8α and CD8β are frequently coexpressed, this close chromosomal proximity suggests that these genes may share cis-active regulatory sequences and that these elements may be located some distance away from either gene. Until relatively recently, it was not possible to clone the entire CD8 locus as a single contiguous DNA molecule because the size exceeded the capacity of conventional cosmid cloning vectors. While larger capacity cloning systems are now available (45, 46) and have been used to introduce larger segments of DNA into transgenic mice, the large size can complicate subsequent attempts to identify and study regions of interest. As an alternate approach, we and others have found that coinjection of two or more different DNA molecules into fertilized embryos frequently leads to cointegration at the same chromosomal site (unpublished observations). Homologous recombination between overlapping coinjected fragments has also been reported (47). To test for cis-active regulatory elements located further 5′ of fragment C, this fragment was coinjected with a 40-kb linearized cosmid clone that extended in the 5′ direction (fragment D in Fig. 1). This clone contained the intact mouse CD8β gene (Lyt3.2 allele) with ∼1.5 kb of 5′ and 25 kb of 3′ flanking DNA, all intergenic DNA between the CD8α and CD8β genes, and overlapped the 5′ end of fragment C by ∼2 to 3 kb. The rationale for coinjecting these two constructs was that if fragment D contained cis-acting regulatory information required for transcription of the CD8α gene, then these sequences might be able to activate expression of the fragment C CD8α/Lyt2.1 transgene if, in the process of chromosomal integration, fragments C and D became linked in a cis-orientation, regardless of whether this occurred through homologous recombination involving the overlapping region.

Four transgenic founders (Nos. 6, 24, 31, and 33) were identified, by hybridization of tail DNA with the fragment C-specific human CD4 probe followed by the fragment D-specific vector probe, as carrying both coinjected fragments. Upon breeding, founders 6, 31, and 33 transmitted fragment C to ∼50% of their offspring. In all cases, fragment D cosegregated with fragment C, indicating that both fragments were cointegrated on a single chromosome. A higher than expected number of founder 24 offspring was found to inherit both fragments, which was subsequently shown to be due to two transgene integrations that segregated among offspring. Both integrations contained fragments C and D and were readily distinguished by hybridization band pattern and intensity. These two integrations were maintained independently and referred to as 24A and 24B. Fluorescent in situ hybridization (FISH) chromosomal mapping analysis of lines 24A and 6 confirmed that both fragments were located at a single chromosomal position (not shown). Together, these findings strongly suggest that in these two (and probably all five) lines, both fragments C and D were cointegrated in a cis orientation at a single locus. DNA blot hybridization analyses for each line indicated a copy number of 2 to 10 copies per cell. Although both fragments appeared to be integrated without rearrangement, we did not attempt to determine their relative orientation at a single integration site.

Expression of the Lyt2.1 transgene was analyzed for each fragment (C+D) line by flow cytometry. Figure 2,A shows the results relative to endogenous Lyt2 expression for cells from thymus ((i) Thy), lymph node (Fig. 2,A(ii), L.N.), and spleen (Fig. 2,A(iii), Spl.) for line 24A (Lyt2.1 Tg(C+D), left panels). Cells from nontransgenic (B6/SJL)F₁ (Non-Tg; right panels) or B6 (data not shown) mice did not react with the Lyt2.1-specific mAb, as they are homozygous for the Lyt2.2 allele. In contrast, Lyt2.1 expression was detected for each of these tissues for line 24A (left panels). For lymph node (ii) and spleen (iii) cells, essentially all Lyt2.1-expressing cells were also positive for endogenous Lyt2 (detected with either a Lyt2.2-specific mAb or Lyt2 monomorphic mAb), although there was some heterogeneity in the level of Lyt2.1 expression (Fig. 2,A). Furthermore, as shown in Figure 2 B(ii) for lymph node (spleen not shown), all Lyt2.1⁺ cells were negative for CD4 expression, and all CD4⁺ cells were negative for Lyt2.1 expression, similar to endogenous Lyt2.2. Altogether, these results indicate that the fragment C+D construct contains cis-active sequence information able to direct Lyt2.1 expression in peripheral T cells in a CD8-sublineage-specific pattern similar to the endogenous Lyt2.2 allele.

Similar flow cytometric analyses of thymocytes from line 24A mice showed that transgenic Lyt2.1 was also coexpressed on a significant fraction of cells that expressed endogenous Lyt2.2 (Figure 2,A(i)) or CD4 (Fig. 2,B(i)). Approximately 17.5 and 20.9% of thymocytes expressing endogenous Lyt2 (Figure 2,A(i)) or CD4 (Figure 2,B(i)) were clearly positive for Lyt2.1 expression. This percentage of transgene-expressing, presumed “double-positive” (i.e., CD4⁺CD8⁺) thymocytes is ∼3.6-fold less than for Lyt2.2⁺CD4⁺ cells (i.e., 20.9 vs 75.6%, Fig. 2,B(i)) and probably results, in part, from the low and heterogeneous level of expression of the transgene product (see below) combined with less intense staining obtained with available Lyt2.1-specific mAbs. Figure 2 B(i) also shows the presence of Lyt2.1⁺CD4⁻ thymocytes (0.5–1.0% of total), the phenotype of the more mature “single-positive” population also observed upon staining for endogenous Lyt2.2 (2.5–3.5%).

While the above pattern of expression in thymic, splenic, and lymph node cells was observed for all transgenic offspring of line 24A, expression of transgenic Lyt2.1 was not observed for offspring of lines 24B, 31, or 33 (not shown). Approximately half of the transgenic offspring analyzed from line 6 showed a similar pattern of Lyt2.1 expression as observed for line 24A, although the apparent level of expression was less (not shown).

To assess the tissue-specific distribution of expression of the Lyt2.1 transgene at the level of RNA, Northern blot hybridization analyses were conducted (Fig. 3). Panels A and C were obtained by hybridization of tissue RNA samples from line 24A offspring with the Lyt2.1 transgene-specific 0.6-kb human CD4 3′ tag probe (see Fig. 1 and Materials and Methods). This probe detected the native CD4 transcript in the human CD4⁺ T cell line Jurkat and did not cross-hybridize with transcripts in lymphoid or nonlymphoid tissue RNAs from nontransgenic mice (Fig. 3, A and C, Non-Tg spleen/Spl, thymus/Thy). In contrast, a band of ∼2 to 2.3 kb, the size expected for the transgenic Lyt2.1 transcript, was detected in thymus (Thy in Fig. 3, A and C) and lymph node (not shown) RNA from 8 of 8 line 24A transgenic offspring analyzed (Lyt2.1 Tg(C+D)). While no band was apparent for the 24A spleen RNA sample shown in A, a faint band was observed with a 2.5-fold greater loading of the same sample (not shown), a sample from a transgenic sibling (Fig. 3,C), and when poly(A)⁺ RNA was analyzed (not shown). Other than thymus, lymph node, and spleen, Lyt2.1 transcripts were not detected in any other tissues of line 24A mice (Fig. 3, A and C). Similar analyses for lines 24B, 31, and 33 failed to detect Lyt2.1 transcripts in total RNA from any lymphoid or nonlymphoid tissues (not shown). A low level of Lyt2.1 transcripts was observed for one of two thymus RNA samples analyzed for line 6 (not shown).

To compare the tissue distribution of Lyt2.1 fragment (C+D) expression with endogenous Lyt2, the blot in Figure 3,A was stripped and rehybridized with an Lyt2 cDNA probe (Fig. 3,B). Comparison of Figure 3, A and B, indicates that the relative tissue-specific patterns of expression of transgenic and endogenous Lyt2 were very similar. Although the transgenic Lyt2 mRNA was, as expected, slightly larger than the endogenous CD8α mRNA (2.0–2.3 kb vs 1.6–1.8 kb) due to the tag inserted into the 3′ untranslated region, the pattern and level of expression detected for transgenic and nontransgenic (Non-Tg) offspring were similar, indicating that the level of Lyt2.1 expression was probably considerably less than that of Lyt2.2.

The elevated level of Lyt2 RNA in thymus compared with spleen (Fig. 3,B) is expected because the majority of cells in the thymus (i.e., 85%) are at the DP stage, while in the spleen only ∼30% of lymphoid cells are T cells, and of these, about 1/3 are CD8⁺. The observation that the ratio of the level of transgenic Lyt2.1 RNA in spleen relative to thymus (determined by PhosphorImaging; see Fig. 3,C) is similar to that for endogenous Lyt2 for transgenic and nontransgenic mice (from Fig. 3 B; i.e., ∼0.1) is consistent with the flow cytometry results showing that the fragment C+D construct is expressed in a similar distribution of cells as endogenous Lyt2 (see below).

To estimate the level of RNA derived from the Lyt2.1 transgene relative to the endogenous allele, a transgene-specific KpnI polymorphism was used in conjunction with RT-PCR blot hybridization analysis of thymus, spleen, and lymph node cDNA from a line 24A mouse. The specificity of this approach (see Materials and Methods for details) for distinguishing the two alleles is shown in Figure 3 D for RT-PCR products from nontransgenic strains that carry the Lyt2.1 allele and the KpnI polymorphism (DBA/2J) or the Lyt2.2 allele without the KpnI polymorphism (B6/SJL)F1. Hybridization of blots of splenic RT-PCR products from DBA/2J and (B6/SJL)F₁ mice with a radiolabeled oligonucleotide probe conserved between both strains (Lyt2.1/2.2 probe) revealed the single expected 516-bp band (left (KpnI −) tracks of DBA/2J and B6/SJL lanes). While KpnI treatment did not alter the size of the (B6/SJL)F₁ Lyt2.2 band (compare KpnI + and −), the 516-bp DBA/2J band was cleaved by KpnI to two bands of about 339 and 177 bp (detected by ethidium bromide staining, data not shown), of which only the smaller one hybridized with the oligo probe. When a similar analysis was conducted for line 24A transgenic spleen (right lanes, Lyt2.1/T4 Tg), the probe detected both the 516-bp band (corresponding to expression of the endogenous Lyt2.2 gene) and the 177-bp band (corresponding to expression of the Lyt2.1 transgene). Both bands were also observed for line 24A thymus and lymph node but not for other tissues (not shown). Band quantitation showed that the level of Lyt2.1 (transgenic) RNA in line 24A spleen, thymus, and lymph node was ∼10% of that from the endogenous (Lyt2.2) locus.

To determine whether expression of the Lyt2.1 C+D transgene construct was regulated appropriately during thymic differentiation, three-color flow cytometry was conducted. CD3 is a useful marker for distinguishing thymic subpopulations as the levels of the TCR/CD3 complex increase with maturation (48). FigureF1 4 shows the results of this analysis for expression of Lyt2.1 (x-axis, left panels) and endogenous Lyt2.2 (x-axis, right panels) with respect to CD4 (y-axis, all panels) and CD3. Figure 4,A shows the results for the total thymocyte population (All), while B through D shows the results for cells expressing low (CD3 low), intermediate (CD3 int.), and high (CD3 hi) levels of CD3. As already noted in Figure 2,B(i), significant populations of Lyt2.1⁺CD4⁺ (i.e., DP; 17.2%) and Lyt2.1⁺CD4⁻ (i.e., SP; 1.0%) cells were detected in total thymocytes (Fig. 4,A, All/left panel). While these phenotypes and the approximate ratio of DP to SP cells for Lyt2.1 (17.2:1.0%) and endogenous Lyt2.2 (80.4:3.9%) are appropriate, the frequency of cells expressing the transgene is less than that of cells expressing the endogenous allele at both stages. The observation that Lyt2.1 is apparently not expressed in all Lyt2.2 thymocytes is reflected in the increased number of cells that appear to be Lyt2.1⁻CD4⁺ (i.e., 79%). Given the normal distribution of CD4 and Lyt2.2 staining for thymocytes of the same animal (Fig. 4 A, right; 14.5% Lyt2.2⁻CD4⁺), this apparent increase probably results from the lack of detectable expression of Lyt2.1 on a proportion of thymocytes that in other respects are at the DP stage.

In the most immature population (CD3 low; Fig. 4,B), ∼91.5% of cells coexpressed Lyt2.2⁺ and CD4⁺ (right), while 17.1% of cells were Lyt2.1⁺/CD4⁺ (left). Although the frequency of Lyt2.1⁺DP cells is less than for Lyt2.2⁺DP cells, these results show that the C+D construct contains cis-active information able to mediate expression at this early stage. As discussed below, it is most likely that the lower than expected number of cells expressing the transgene is due to the lower quantitative level of expression. As expected, there were essentially no Lyt2.2⁺CD4⁻ or Lyt2.2⁻CD4⁺ (right) SP CD3^low cells. Similarly, there were no Lyt2.1⁺CD4⁻ SP cells detected. However, regarding total thymocytes, there was a larger than expected number of cells that appeared to be Lyt2.1⁻CD4⁺ SP. An overall similar result was observed when cells expressing intermediate levels of CD3 (Fig. 4 C, CD3 int.) were analyzed. Again, Lyt2.1⁺CD4⁺ and Lyt2.2⁺CD4⁺ DP cells were detected but the frequency of transgene-expressing cells appeared to be less than for the endogenous allele (19.9 compared with 95.7%). A very small number of CD8⁺CD4⁻ SP cells was detected for both Lyt2.1 (0.36%) and Lyt2.2 (1.2%). While the number of Lyt2.2⁻CD4⁺ SP cells was also low, as expected (i.e., 1.54%), as for CD3^low cells, the apparent frequency of Lyt2.1⁻CD4⁺ SP cells was very high (77.9%). As these cells gave a normal distribution when stained for CD4 and endogenous Lyt2.2 (right), most of these cells are actually at the DP stage but just did not express detectable levels of Lyt2.1.

At the CD3^hi stage (CD3 hi; Fig. 4 D), similar to expression of Lyt2.2, there was a significantly increased frequency of Lyt2.1⁺CD4⁻ SP cells and a decreased frequency of Lyt2.1⁺CD4⁺DP cells compared with the CD3^low and CD3^int populations. While these changes between the CD3^int and CD3^hi stages were similar for both the Lyt2.1- and Lyt2.2-expressing populations, the frequency of transgene-expressing cells was still less than for the endogenous allele. Finally, the relative frequencies of Lyt2.1⁻CD4⁺ and Lyt2.2⁻CD4⁺ SP cells were much more similar (i.e., 75.2 and 65%) than at earlier stages.

Although these results indicate that the fragment C+D construct contains cis-active sequence information able to mediate stage-specific expression in a significant proportion of DP and SP thymocytes, it was possible even in these cells that there were subtle differences in the timing of activation or extinction of Lyt2.1 expression compared with the endogenous locus. This might be the result if there are additional cis-elements beyond the region covered by fragments C+D or from the influence of DNA sequences at the site of transgene integration (i.e., position effects). A more rigorous test of the appropriateness of the observed expression pattern would be provided by demonstrating that the distribution of expressed transgenic CD8α contributed functionally to thymic differentiation and selection and the generation of mature CD8⁺ CTL. However, as transgenic Lyt2.1 expression was low compared with the endogenous allele, it would be difficult to address this issue in normal transgenic mice. Therefore, line 24A transgenic mice were crossed multiple times to CD8α-deficient knock-out mice (9) to examine expression and function of the Lyt2.1 transgene in the absence of endogenous CD8α expression. Offspring that were homozygous for the endogenous CD8α/Lyt2.2 mutant allele and that carried the Lyt2.1 transgene were identified by DNA blot hybridization (not shown).

Figure 5 shows two-color CD4/Lyt2.1 flow cytometry results obtained for thymocytes (Fig. 5,A, Thy.) and lymph node cells (Fig. 5,B, L.N.) for nontransgenic/CD8α homozygous knock-out (Non Tg/CD8^−/−; right panels, A and B) and Lyt2.1 transgenic (line 24A)/CD8α homozygous knock-out (Lyt2.1Tg(C+D)/CD8^−/−; left panels, Fig. 5, A and B) mice. As expected (9), neither Lyt2.1- or Lyt2.2-expressing cells were detected in thymocytes or lymph nodes from nontransgenic/knock-out mice (Fig. 5, A and B, right profiles of upper/Lyt2.1 and lower/Lyt2.2 panels). In contrast, Lyt2.1⁺CD4⁺ DP (24.1% of total) and Lyt2.1⁺/CD4⁻ SP (0.6% of total) thymocytes (Fig. 5,A, upper panel, left profile) were detected in Lyt2.1 transgenic/CD8α knock-out mice. As for Lyt2.1 transgenic/wild-type mice (Fig. 4), there was a greater than expected number of cells that appeared to be Lyt2.1⁻CD4⁺ SP. In the transgenic/wild-type mice, these cells were actually at the DP stage, as they expressed the endogenous Lyt2.2 allele. Despite the absence of Lyt2.2 expression in the transgenic/knock-out mice, we assume that most of these CD8⁻CD4⁺ SP cells are also at the DP stage with respect to other phenotypic markers. Both lymph node (Fig. 5,B, left profile, upper panel) and spleen (not shown) cells of the transgenic/knock-out mice also contained significant populations of Lyt2.1⁺CD4⁻ (6.3%) and Lyt2.1⁻CD4⁺ (54.4%) SP cells. Taken together, these results demonstrate that the fragment C+D construct contains cis-regulatory information able to direct a temporal and spatial pattern of Lyt2.1 expression sufficient to mediate the development of thymic and peripheral CD8⁺ SP T cells. The reduced relative number of Lyt2.1-expressing cells in lymph nodes and spleen (not shown) of transgenic mice deficient for endogenous CD8α (Fig. 5,B) compared with mice that are not (Fig. 2 B(ii)) probably results from a decreased percentage of cells that survive thymic selection due to the lower overall level of coreceptor expression.

The above-described results indicate that fragment D contains cis-acting transcriptional control sequences able to direct expression of CD8α in both DP and SP thymocytes as well as in peripheral CD8⁺ T cells. To sublocalize this cis-active information and to investigate whether there is one or more than one element responsible for stage- and sublineage-specific expression, we used the same approach as described above to test the activity of two subfragments of D (i.e., E and F) in mediating expression of the fragment C CD8α transgene construct. Fragment E was a 23-kb ScaI fragment covering the intergenic region between CD8α and CD8β (i.e., the 3′ portion of D), while fragment F was a 17-kb BamHI fragment covering most of the CD8β gene region (i.e., the 5′ portion of D; see Fig. 1). Each of these two fragments was separately coinjected into fertilized eggs with fragment C to generate transgenic mice carrying fragments C+E and C+F. Four fragment C+E and three fragment C+F transgenic founders were identified by Southern blot hybridization and bred to produce lines. In each case, as both fragments were coinherited, as shown above for coinjection of fragments C+D, it is very likely that both fragments were cointegrated in cis at a single integration site.

Flow cytometric, RNA blot hybridization, and RT-PCR analyses of tissues of transgenic offspring for all four C+E lines failed to detect expression in thymocytes, splenic or lymph node T cells, or any other tissue (not shown). In contrast, expression of transgenic Lyt2.1 was detected in thymocytes but not peripheral T cells of offspring of two of the three C+F lines. No expression in any tissue could be detected in the third line. Figure 6,A shows the 2-color staining results (CD4 vs Lyt2.1) for thymocytes from an offspring of C+F line 42 (Lyt2.1 Tg(C+F), left) compared with a nontransgenic sibling ((Non Tg, right). Similar to the results obtained for fragment C+D transgenic mice (Fig. 2 B(i)), there was a significant fraction of CD4-expressing cells that coexpressed transgenic Lyt2.1 (i.e., ∼64%). All of these cells coexpressed endogenous Lyt2.2 (not shown) and thus appear to be typical DP thymocytes. There was also a very small population of cells that was negative for CD4 and expressed low levels of Lyt2.1 (i.e., 0.58%). As for thymocytes from C+D mice, the level of transgenic Lyt2.1 expression was heterogeneous, giving rise to a higher than expected number of cells that appeared to be SP CD4⁺Lyt2.1⁻ (∼32.9%). As staining of this same population for endogenous Lyt2.2 expression gave a normal distribution of DP and SP cells (not shown), this apparently increased number of CD4 SP cells was due to undetected surface expression of Lyt2.1 in a fraction of cells that were actually DP rather than a distortion in T cell development resulting from expression of the transgene.

In contrast to the above-described results, transgenic Lyt2.1 expression could not be detected in lymph node (Fig. 6,B, line 42) or spleen (not shown) cells of offspring of any of the C+F lines, notably the two that showed clear expression in thymocytes. This lack of Lyt2.1 expression in peripheral T cells of C+F mice is in contrast to the appropriate sublineage-specific pattern of expression observed in spleen and lymph node T cells of fragment C+D mice (Fig. 6,B, right; Lyt2.1 Tg(C+D)). To determine whether this lack of surface expression in T cells of C+F mice was due to transcriptional or posttranscriptional mechanisms, Northern blot hybridization analyses were conducted. Figure 7 shows the results for an offspring of C+F line 42. Fig. 7,A was derived by hybridization with the Lyt2.1 transgene-specific human CD4 probe (Lyt2.1-Tg), while Fig. 7,B was derived with a Lyt2 cDNA probe (Lyt2-Endog.). Strong signals were observed for both transgenic (Fig. 7,A) and endogenous (Fig. 7,B) Lyt2 in thymus RNA. However, although endogenous Lyt2 RNA was readily detected in spleen (Fig. 7,B) and lymph node cells (not shown), transgenic Lyt2.1 RNA could not be detected in total (Fig. 7,A) or poly(A)⁺ (not shown) RNA from either of these tissues, even by RT-PCR analysis (not shown). The second C+F line found to express transgenic Lyt2.1 in thymocytes, but not spleen or lymph node cells, gave a RNA tissue distribution similar to line 42, which is detectable expression in thymus only, although the level was about fivefold less (not shown). This thymus-specific pattern of Lyt2.1 expression in C+F mice contrasts with the results obtained for C+D mice in which the relative level of Lyt2.1 expression in spleen and lymph node compared to thymus was similar to that for endogenous Lyt2.2 expression (Fig. 3, A–C). The third fragment C+F line failed to show expression of Lyt2.1 RNA in any lymphoid or nonlymphoid tissue (not shown).

Expression of transgenic Lyt2.1 RNA and protein in peripheral T cells of C+D but not C+F mice indicates that there are distinct cis-acting transcriptional regulatory sequence requirements for CD8α expression in DP thymocytes and SP peripheral CD8⁺ T cells. While fragment D contains regulatory sequences able to mediate Lyt2.1 expression in both developing and mature T cells, fragment F contains cis information able to mediate expression in DP thymocytes only (i.e., not in SP CD8⁺ peripheral T cells). These results suggest that during T cell development there is a “switch” in transcriptional control of CD8α expression from a cis-mechanism(s) active in DP cells, which is dependent on cis sequences within fragment F, to another mechanism(s) active in SP CD8⁺ cells, which depends on cis sequences within fragment D but absent from F. To determine at what stage(s) in thymic development this switch occurs, three-color flow cytometry was conducted for thymocytes of C+F mice using CD3 expression level to distinguish different stages of maturation. Figure 8 shows the profiles of CD4 vs Lyt2.1 expression for cells that express low (CD3 low, top), intermediate (CD3 inter, middle), and high (CD3 hi, lower) levels of CD3 in Lyt2.1 transgenic C+D (right panels) and C+F (left panels) mice. For C+F and C+D mice, Lyt2.1 expression was detected in CD4⁺ cells (i.e., DP cells) in both the CD3^low (31.44% for C+F; 16.0% for C+D) and CD3^int (46.4% for C+F; 20.6% for C+D) populations. As all Lyt2.1⁺ cells coexpressed the endogenous Lyt2.2 allele (not shown), these Lyt2.1⁺CD4⁺ cells are typical DP thymocytes. The approximately twofold greater number of Lyt2.1⁺ cells in CD3^low and CD3^int cells from C+F compared with C+D mice is probably due to the quantitatively higher level of expression in C+F mice, as revealed by RNA analyses. In the more mature CD3^hi population, expression of the C+D construct is detected in both CD4⁺ (i.e., DP) cells (6.26%) and CD4⁻ (i.e., SP) cells (7.02%). As in Figure 4, the reduced number of Lyt2.1⁺DP cells and increased number of Lyt2.1⁺SP cells in the CD3^hi population is very similar to the distribution observed for the endogenous allele. In contrast to the C+D construct and the endogenous Lyt2.2 allele, expression of the C+F construct is detected in only a very small number of CD4⁺ (i.e., DP) cells (1.98%) and not at all in CD4⁻ cells (i.e., the SP quadrant). As normal distributions were observed for endogenous Lyt2.2 expression in both the DP and SP CD3^hi populations of C+F mice (not shown), these results indicate that expression of the Lyt2.1 C+F construct has been lost between the CD3^int and CD3^hi stages of T cell development.

The results presented in this paper reveal several important features of the mechanisms underlying stage- and sublineage-specific expression of CD8 in developing and mature T cells. First, it is clear that regulation of this coreceptor depends on multiple cis-active transcriptional control sequences, at least some of which are located a considerable distance from the CD8α gene itself. Based on the lack of even low level RNA or surface protein expression of Lyt2.1 constructs containing up to 12 kb of 5′ and/or 4.5 kb of 3′ native flanking DNA (fragments A, B, and C in Fig. 1), we conclude that expression of CD8α in thymus-derived αβ-TCR T cells depends on cis-active regulatory elements located beyond this region.

Previous studies from others identified candidate cis-regulatory elements within the CD8α gene (27) and in the immediate 5′ flanking region (25, 26) by homology with known elements, by transfection, or by mapping of DNase I-HSS. More recent DNase I-HSS mapping of a much larger segment of the CD8 locus identified four clusters of HSS as additional candidate regulatory elements (30). Non-tissue-specific cluster I was located ∼20 kb downstream of the CD8α gene, while tissue-specific clusters II and III were located within ∼6.5 kb 5′ and between ∼14 to 22 kb 5′ of the CD8α gene, respectively. Cluster IV was within and 3′ of the CD8β gene (30). Lack of expression of CD8α transgene fragments A, B, or C indicates that candidate cis elements identified within the CD8α gene or in the immediate 5′ flanking region (25, 26), including cluster II HSS (30), are not sufficient for directing expression in differentiating and mature CD8⁺ αβ-TCR T cells.

Our finding that inclusion of an additional 40-kb segment of upstream flanking DNA (fragment D) together with the fragment C CD8α gene led to a pattern of transgenic Lyt2.1 expression in DP and SP thymocytes and peripheral CD8⁺CD4⁻ and CD8⁻CD4⁺ T cells, which essentially paralleled the endogenous Lyt2.2 allele, indicates that this upstream region contains transcriptional control sequences able to direct stage- and sublineage-specific expression of CD8α in thymus-derived αβ-TCR T cells. As fragment D includes the CD8β/Lyt3 gene, it is possible that these cis sequence(s) normally serve to mediate expression of both genes. To assess fragment D-derived CD8β (Lyt3.2) expression, we are breeding the Lyt2.1 fragment C+D transgene onto a background that carries the alternate endogenous Lyt3.1 allele (AKR) or with CD8β knock-out mice.

Although an appropriate distribution of Lyt2.1 expression was observed in a proportion of fragment C+D transgenic lines, the level of expression was less than the endogenous locus. In lymph node and spleen of line 24A, for example, in which virtually all Lyt2.2⁺ cells coexpressed the Lyt2.1 C+D transgene (Fig. 2,A, ii and iii), RT-PCR analysis indicated that the level of transgenic RNA was no more than ∼10% of that of the endogenous allele (Fig. 3D). Transgene expression level in the thymus also appeared to be quantitatively less than for Lyt2.2, although the complexity of CD8-expressing thymic subpopulations makes precise quantitation difficult (see below).

One possible explanation for the reduced level of transgenic CD8α expression in C+D mice is that additional cis-regulatory information, such as a locus control region (LCR), which influences the quantitative level of expression, may lie beyond the 60-kb C+D region. LCRs have been identified for several genes, including the human β-globin locus (49), the αδ-TCR locus (50), and HLA class I genes (35, 51), and have been shown to direct RNA expression at a level comparable with the chromosomal locus, regardless of the integration site. Hostert et al. reported that an 80-kb mouse CD8α P1 clone was expressed in transgenic mice at a level similar to or greater than the endogenous allele in four of six transgenic lines (30). However, in some of these lines expression was observed in only 80 to 97% of the expected CD8⁺ subset (30). This result, combined with lack of expression in the other two lines, was taken to indicate that this P1 clone lacked LCR-like elements, either because there is no CD8 LCR or because it lies beyond the region included. This P1 clone extended from just 5′ of the CD8β gene, similar to fragment D in our studies, to more than 20 kb 3′ of the CD8α gene and included DNase I-HSS clusters I through IV (30). As fragment C+D (as well as a CD8 P1 clone we have studied (52); see below) appears to differ from this P1 clone only in the amount of 3′ DNA flanking the CD8α gene, HSS cluster I or other downstream elements may be responsible for the increased expression observed by this other group.

A second possible explanation for the lower than expected level of fragment C+D expression is that optimal transcription may be possible only when the 2 fragments are integrated relative to each other in their native chromosomal upstream/downstream configuration, possibly thereby permitting efficient formation of transcription factor-dependent enhancer/promoter interactions. Arguing against this view, however, is our recent finding that a 85-kb transgenic P1 phage clone, containing the entire region covered by fragments C+D as well as additional upstream and downstream DNA as a single contiguous molecule (52), gave results similar to the C+D construct (unpublished results).

Other than the reduced level, the only other difference in the distribution of transgenic and endogenous CD8α-expressing cells in C+D mice was an apparent increase in Lyt2.1⁻CD4⁺ thymocytes at the CD3^low and CD3^int stages. As most of these cells expressed endogenous Lyt2.2 and thus were actually at the DP Lyt2.2⁺CD4⁺ stage, it is likely this results from a combination of the low level of expression of the Lyt2.1 transgene and naturally heterogeneous CD8 levels between cells during the DN to DP to SP transitions. RT-PCR analysis of purified thymocytes with this Lyt2.1⁻Lyt2.2⁺CD4⁺ surface phenotype is being conducted to quantitate the level of transgenic Lyt2.1 RNA expression in these cells.

The ability of the Lyt2.1 fragment C+D transgene to “rescue” the development of CD8⁺ DP and SP T cells in the CD8α knock-out background provides functional confirmation that this construct contains cis-active regulatory information sufficient to mediate appropriate stage- and sublineage-specific CD8α coreceptor expression. The reduced level of total CD8α expression (i.e., transgenic plus endogenous) is probably responsible for the decreased frequency of transgene-expressing rescued cells in the knock-out vs wild-type backgrounds. As the rescued CD8⁺ SP T cells are responsive to alloantigen and foreign Ag (not shown), even low level expression of this coreceptor is sufficient to mediate development of fully functional CTLs.

Compared with transgene fragment C+D, the results obtained for fragment C+F were quite different. First, in contrast to C+D, the C+F construct was not expressed in peripheral CD8⁺ αβ-TCR T cells. Second, while C+F was expressed appropriately in DP thymocytes bearing low and intermediate levels of CD3, it was not expressed in more mature DP or SP thymocytes that expressed high levels of CD3. The C+D construct was expressed at all appropriate stages of thymic development, including the CD3^hi DP and SP populations.

In contrast, we were unable to detect expression in any cell type of mice derived from microinjection of fragments C+E. While this manuscript was under review, two papers were published (57, 58) indicating that an intergenic subregion containing DNase I-HSS cluster III and that overlapped with the 3′ end of fragment E was able to direct expression in mature SP CD8⁺ T cells, but not in immature DP cells. Possible explanations for this apparent difference include alteration of the cis activity of resident regulatory elements due to the specific structure of fragment E; an unfavorable pattern of cointegration of fragments C and E; partial disruption of DNase I-HSS cluster III, as it is located close to the end of fragment E; and/or analysis of only a limited number of mice. We are currently analyzing the activity of additional subregions of fragments D and E.

The results presented in this paper indicate that stage-specific expression of CD8α in DP as opposed to SP thymus-derived αβ-TCR T cells is mediated by the differential activity of distinct stage-specific cis-active regulatory mechanisms. One cis mechanism (i.e., mechanism I) depends on sequence information within fragment F and is able to direct expression of transgenic CD8α in developing thymocytes up to the CD3^int stage but not in CD3^hi DP or SP thymocytes or SP CD8⁺ peripheral T cells. As transgenic CD8α expression was detected in CD8⁺CD3^hi DP and SP T cells in the thymus and the periphery of C+D but not C+F mice, the results imply that a second cis-active mechanism (i.e., mechanism II), presumably involving DNase I-HSS cluster III (57, 58), mediates expression in these more mature cells. Thus, around this intermediate to late DP stage of T cell development, a “molecular switching” of the cis-active control of CD8α expression occurs from mechanism I to II. An important question is whether the switch occurs before (i.e., leading up to and possibly involved in the triggering of) or subsequent to (i.e., possibly as a result of) thymic selection (12, 14, 15, 53, 54, 55, 56). Switching before selection will indicate that the DP to SP transition in transcriptional control of CD8α is not determined by TCR-dependent selection-associated signaling events, while switching after selection is consistent with this possibility. Experiments in progress should answer this question.

The authors thank Dr. Tak Mak for providing the CD8α knock-out mice.