Interleukin 2 Receptor Signaling Regulates the Perforin Gene through Signal Transducer and Activator of Transcription (Stat)5 Activation of Two Enhancers

SAM-19 cells were derived from the autonomously growing mouse CTL/rat thymoma hybrid PC60 25 that cannot be activated by IL-2 on its own to express perforin 26. We noticed a weak IL-2 response of the perforin gene in a subline designated SA. Further analysis indicated that the SA line, but not the other lines, contained cells constitutively expressing the mouse IL-2Rβ chain. Constitutive expression of mouse IL-2Rγ was equivalent in all lines. After cloning these cells at the single-cell level, one clone, SAM-19, was obtained that expressed IL-2Rβ mRNA at levels equivalent to those in the CTL effector–like CTLL-2 line. Neither resting nor IL-2–stimulated cells expressed detectable levels of mouse IL-2Rα protein or mRNA. As described for T cells exclusively expressing IL-2Rβ and IL-2Rγ 27, SAM-19 responded to recombinant mouse IL-2 at concentrations >100 U/ml with the induction of the mouse perforin and granzyme B genes. Half-maximal levels of perforin mRNA were detected after ∼14 h of induction. Maximum levels were induced after ∼48 h. Cell counts of SAM-19 cultures maintained in the presence of 500 U/ml IL-2 versus its absence did not significantly differ from each other, suggesting that IL-2R signaling has little or no effect on the growth and survival of SAM-19. The cells were maintained in IMDM supplemented with 10% FBS, 5 × 10⁻⁵ M β-ME, and 50 μg/ml gentamycin.

Total RNA was extracted and analyzed as described 28. Purification and RNase treatment of nuclei as well as the elongation reactions and their purifications were performed as described 29 using ∼2.5 × 10⁷ nuclei and 240 μCi of α-[³²P]UTP (800 Ci/mmol) per reaction (15 min at 30°C). Hybridizations were carried out with 1.6 × 10⁷/ml TCA-precipitable counts for 36 h at 65°C in 10 mM Tris, pH 7.5, 10 mM EDTA, pH7.5, 0.3 M NaCl, 1% SDS, 250 μg/ml tRNA, 1× Denhardt's solution, and 0.5% nonfat dry milk. Filter washes, including an RNase A digestion, have been described 29. Target probes were obtained by PCR of genomic DNA and cloned as follows. The β₂-microglobulin probe (nucleotides 3,942–4,384 of sequence available from EMBL/GenBank/DDBJ under accession number M18837) and the 5S ribosomal RNA probe (nucleotides 268–449 of sequence available from EMBL/GenBank/DDBJ under accession number X51545) were cloned into pCRII (Invitrogen Corp.). The perforin 5′ probe containing the short first exon and the beginning of the first intron (nucleotides 830–1,156 available from EMBL/GenBank/DDBJ under accession number M95527) and the perforin 3′ probe containing part of the third exon (nucleotides 101–464 of sequence available from EMBL/GenBank/DDBJ under accession number X51446) were cloned into M13 mp18, and single-stranded phage DNA containing perforin antisense DNA was produced. 10 μg denatured DNA was slot blotted for each target. All constructs were confirmed by sequencing.

The two genomic clones depicted (see Fig. 2) were obtained by screening a commercially available human placenta pWE15 cosmid library (Stratagene Inc.) on duplicate filters with a human perforin exon III probe and a human perforin promoter probe. They were further analyzed by restriction mapping and Southern hybridization to these two probes. All analyzed fragments matched the results of genomic Southern blots. Also, the sequence of a promoter fragment of both clones was identical to the one we reported previously 30. The initial reporter gene construct was assembled by cloning a KpnI–SfiI fragment (−15,600 to −277) into the promoterless pGL3 basic reporter vector (Promega Corp.), in which the remaining human perforin promoter, exon I, intron I, and the untranslated sequences of exon II had been fused to the firefly luciferase gene. Constructs without the first intron used an EcoRV site at +59. The progressive deletion constructs (see Fig. 3) were obtained by exonuclease treatment of a KpnI (−15,600)- and HindIII (−13,300)-digested plasmid, followed by Klenow and T4 DNA polymerase treatment, gel purification, and religation. All constructs were transformed by electroporation into DH10B cells (Life Technologies). Other deletions, including internal deletions, were created by restriction enzyme digestion followed by T4 DNA polymerase treatment and religation of the vector. The analysis of the upstream enhancer (see Fig. 4 B) also involved the cloning of PCR fragments, all of which were verified by sequencing both strands of the enhancer. Mutations of the signal transducer and activator of transcription (Stat)¹ elements were introduced using the QuickChange site-directed mutagenesis kit from Stratagene Inc. After verification of the sequences on both strands of the enhancers, they were recloned into a fresh vector to eliminate second-site mutations outside of the sequenced DNA. The dominant negative Stat5a expression vector was generated similarly.

Vector backbone–free DNA for stable transfections employed NotI sites flanking the inserts of the pWE15 cosmid clones or the NotI and SalI sites of the pGL3 reporter vectors. For stable transfections, an ∼5:1 molar excess of the perforin transgene over a selection cartridge containing a TK promoter driving the neomycin gene in a total of 40 μg DNA was transfected into 400 μl of 2.5 × 10⁷/ml SAM-19 in DMEM supplemented with 4.5 g/liter glucose and 25 mM Hepes by electroporation of a 4-mm gap cuvette with an ECM600 system (BTX) set to 230 V, 3,000 μF, and 24 Ω. The washed cells were cultured for 15–17 h in 2 ml of complete growth medium, washed again, and selected as 2 × 10⁴/ml live cells (1.0 ml per well in a 48-well plate) in the presence of 1.3 mg/ml active G418 (Life Technologies). 2–3 wk later, G418-resistant clones were expanded and screened by PCR for the presence of human perforin promoter sequences. PCR-positive clones were further analyzed by genomic Southern blot of restriction-digested DNA side-by-side with digested human genomic DNA to determine the integrity of the transgenic constructs and to estimate the copy numbers of the transgene. Transgenic mice, generated as we described previously 24, were analyzed accordingly.

Electroporation conditions for transient transfections were identical but used less DNA (indicated in the figure legends) and included an internal control plasmid, pRL-CMV (Promega Corp.) in which a CMV promoter drives the renilla luciferase gene. After the electroporated cells were washed, they were split into two 500-μl aliquots (three aliquots for activations using pharmacological agents), rested for 1–3 h, and then supplemented with 500 μl of medium with or without recombinant mouse IL-2 (and/or pharmacological agents), providing a final concentration of 600 U/ml. After overnight incubation for 15–17 h, the cells were washed with PBS and lysed and extracted in 50–80 μl passive lysis buffer (Promega Corp.) by two rounds of freeze thawing. Reporter gene activities were determined in three 10-μl aliquots of each extract using the dual luciferase assay system from Promega Corp. and a ML2250 96-well plate luminometer (Dynex Technologies). Signals were integrated for 10 s for both luciferase activities. The ratios of firefly luciferase activity to renilla luciferase activity varied by <5% in the triplicate measurements. Their average was used to represent the analysis of each independently transfected sample.

An important limitation in the study of antigen- and/or growth factor–dependent CTL clones is that they downregulate their overall RNA and protein synthesis and begin to apoptose when deprived of stimuli. In fact, the time periods required to “downregulate” the perforin gene to study its “activation” considerably overlap with this generalized cell shutdown. Therefore, we established a novel CTL tissue culture model, designated SAM-19, to facilitate the study of perforin gene induction by IL-2R signaling. SAM-19 is an autonomously growing, growth factor–independent clone derived from a mouse CTL–rat thymoma hybrid (see Materials and Methods). These cells respond to high doses of mouse IL-2 with the induction of mouse perforin and granzyme B mRNAs with kinetics similar to those observed in primary cells 3. The induction of the perforin gene by IL-2 did not depend on newly synthesized proteins (Fig. 1 A), and the mRNA induction was not regulated at the posttranscriptional level (data not shown), properties analogous to those of primary CTLs 3. In addition, perforin mRNA induction in SAM-19 was not blocked by rapamycin (Fig. 1 B), as reported for primary cells 31. Taken together, SAM-19 appears to comprise a reasonable in vitro model with which to study perforin gene regulation by IL-2R signals.

Before undertaking more detailed studies, we considered the possibility that IL-2–unresponsive constitutive expression by the mouse perforin promoter in transgenic mice 24 could have been due to a regulation at the level of transcription elongation. As assayed by nuclear run-on analysis, however, transcription of the 5′ and 3′ ends of the perforin gene were induced by IL-2 to similar levels (Fig. 1 C). These data indicate that IL-2R signals regulate the perforin gene at the level of transcription initiation via regulatory domains other than its promoter.

An expression screening strategy was used to identify putative regulatory domains other than the promoter. This approach was facilitated by the compact dimension of the human perforin gene (Fig. 2), the ability to distinguish the human and the mouse perforin mRNAs from each other, and the transfectability of SAM-19. In initial experiments, two suitable human gene locus DNAs from a cosmid library were stably transfected into SAM-19. The expression of both the −16,500 and −20,700 DNAs was regulated by IL-2 in several independent clones (Fig. 2). This suggested that these DNAs contained IL-2–responsive cis-acting sequences in their overlapping regions. Further efforts focused on their 5′ flanks, because we failed to detect transcriptionally relevant DNA in transient transfections of the intragenic DNA and the 3′ flank of the −20,700 gene locus DNA. Indeed, a restriction fragment comprising most of the cloned 5′ flank fragment also drove a transgenic luciferase reporter gene in an IL-2–responsive manner (Fig. 2, −15,600), in contrast to a promoter construct (−277). Interestingly, an intermediate construct (−5,300) responded at intermediate levels, which was shown in further investigations to be due to the presence of two regulatory domains rather than one. Unlike the inducibility of the transgenes, their levels of expression in relation to the transgenic copy numbers varied considerably in all clones, suggesting that these DNAs do not contain additional constitutive enhancers, silencers, or a locus control region.

To extend the results from the stable transfections of SAM-19 to normal T cells, a −15,600 to +59 perforin/luciferase construct was analyzed in transgenic mice. Two founders did not express luciferase activity at all, neither in resting nor activated peripheral lymphocytes. The other two founders expressed the transgene, with its expression regulated by IL-2R as well as TCR signals, including the cross-linking of CD3 or PMA plus ionomycin (Table). The rates of induction for luciferase protein observed in the transgenic mice were higher than those in the SAM-19 system. Based on our experience with a transgenic CD4 reporter gene 24, this may reflect a global increase in the translation efficiency upon activation of freshly obtained primary T cells. More importantly and unlike the previously investigated promoter transgenes, the −15,600 to +59 perforin/luciferase-transgenic T cells clearly responded to activation signals.

To address exactly where the IL-2–responsive cis-acting DNA was located, progressive deletions of the construct analyzed in transgenic mice were assayed in transient transfections. A consistent pattern for their IL-2 responses was noted (Fig. 3 A). The most 5′ deletion (from −15,600 to −13,300) resulted in an impaired IL-2 response. This response was abolished once sequences downstream of −1,450 were deleted. This finding suggested the presence of two cis-acting DNAs, consistent with the intermediate levels of response seen previously in the stable transfections of the reporter construct of intermediate length (Fig. 2). This interpretation was also consistent with the reporter analysis of internal deletions (Fig. 3 A, bottom). The deletion of the sequences residing between the two putative domains as well as the spacing between the two domains did not impair the levels of regulation observed by the entire 5′ flank. Finally, the far-upstream regulatory domain on its own also conferred an IL-2 response to the unresponsive perforin promoter (Fig. 3 A, last construct). Both regulatory domains required the context of a promoter (data not shown), indicating that neither comprised a second promoter. The responses by each individual domain appeared to be additive, suggesting that their functions were independent of each other within the experimental context.

The far-upstream regulatory DNA identified in the transient transfections comprised the very end of the −15,600 construct, whereas both of the original gene locus DNAs extended farther upstream and appeared to respond somewhat more to IL-2 (Fig. 2). Therefore, the remainder of the cloned DNA was analyzed to delineate the 5′ and 3′ borders of the far-upstream regulatory DNA (Fig. 3 B). This led to an NsiI–BalI fragment that retained the maximal response of the far-upstream regulatory DNA (Fig. 3 B, bottom). This fragment also functioned in the context of the heterologous SV40 promoter (Fig. 4 A). IL-2 increased the transcriptional activity of the constructs over the promoter levels irrespective of the orientation of the regulatory DNA. These attributes were consistent with an IL-2–inducible enhancer, whose required core was contained within ∼150 bp (Fig. 4 A, bottom). The analogous experiments for the upstream regulatory domain recapitulated these findings and localized this enhancer within ∼130 bp (Fig. 4 B).

Consistent with the rapamycin insensitivity of perforin mRNA induction by IL-2 (Fig. 1 B), the activation of the enhancers was not blocked by rapamycin (Fig. 5 A) at concentrations that abolished granzyme B induction (Fig. 1 B).

Because perforin mRNA can also be induced in primary CTLs by TCR signals 2, SAM-19 was activated with pharmacological agents known to mimic TCR signals, namely phorbol ester and ionomycin. These agents induce perforin mRNA in SAM-19 in the absence of detectable IL-2 (data not shown). Regarding the involvement of the identified enhancers in this process, the far-upstream enhancer, but not the upstream enhancer, was significantly activated in SAM-19 (Fig. 5 B, top). This activation was sensitive to cyclosporin A (CsA),¹ indicating that calcineurin participated in the activation of the enhancer. These findings suggest that the far-upstream enhancer may also respond to TCR signals.

The T and NK cell–restricted expression of perforin raised the additional question of whether activation of the enhancers by pharmacological agents might be lineage specific. There was no substantial activation of either enhancer in the J588L B cell or the L929 fibroblast model (Fig. 5 B, center and bottom). These data suggest that the enhancers can be activated primarily in T cells.

Lastly, participation of the Jak/Stat signaling pathway from the IL-2R was investigated, because both enhancers contained Stat-like elements. Coexpression of a dominant negative signal transducer and activator of transcription (Stat)5 molecule 32 blocked the activation of both enhancers by IL-2R signals in a dose-dependent manner, in contrast to the cotransfections with the parental wild-type Stat5 expression vector (Fig. 5 C). These data imply that the activation of Stat5 plays an important direct or indirect role in the activation of the perforin enhancers by IL-2R signals.

Visual and computer-aided inspection 33 of the sequenced enhancer cores revealed several similarities to known binding sites or their cores (Fig. 6 A). Potential binding sites present in both enhancers at similar positions included two Ets cores, an activator protein (AP)-1 site and, most strikingly, two identically spaced Stat-like elements. One closely resembled the consensus dyad symmetry (Fig. 6 A, STAT). The other element was less well conserved (Fig. 6, Fig. 3′ STAT-n). A third less well conserved Stat-like element occurred only in the far-upstream enhancer (Fig. 6, Fig. 5′ STAT-n) and overlapped with the highly conserved element (STAT).

To determine whether these elements were indeed of functional relevance, mutant enhancers were analyzed. Minimal mutations were designed to selectively interfere with the potential Stat elements (Fig. 6 A, asterisks). Regarding both enhancers, a mutation of their 3′ STAT-n sites reduced their transcriptional activation by IL-2 nearly to the levels of unactivated cells (compare data in Fig. 6 B for constructs cotransfected with an empty expression vector). A mutation of their STAT sites or both sites together (double mutant) completely abolished their enhancer function. In contrast, mutation of the 5′ STAT-n site present only in the far-upstream enhancer did not impair enhancer activation by IL-2 (Fig. 6 B). These results demonstrate a cooperative requirement for both the STAT and the 3′ STAT-n site in each enhancer. Combined with the observed inhibition of enhancer activation by a dominant negative Stat5 molecule, these results suggest that Stat molecules may directly target these elements.

To address whether Stat molecules could transactivate the enhancers via the putative elements, a constitutively active Stat5 molecule was coexpressed with the perforin enhancer reporter constructs. We focused on Stat5 rather than Stat3 because we could not detect the activation of Stat3 in SAM-19 upon IL-2R signaling under conditions where activated Stat5 molecules were readily detectable (data not shown). To that end, a chimeric Stat5-VP16-Jak2 molecule that is autoactivated and leads to Stat5-specific DNA binding and strong transactivation was used 34. Cotransfection with the respective expression vector, but not the empty expression vector, strongly hyperactivated the wild-type enhancer, but not the SV40 promoter, in the absence of IL-2 (Fig. 6 B). Regarding the upstream enhancer, the transactivation was absolutely dependent on the intact STAT site and to a lesser extent on the 3′ STAT-n site (Fig. 6 B, left). Additional IL-2R signaling provided somewhat higher levels, suggesting that additional transcription factors were induced. The analogous transactivation studies of the far-upstream enhancer (Fig. 6 B, right) were complicated by its additional 5′ STAT-n site. As described above, this site was physiologically irrelevant because its mutation did not at all impair the IL-2 response. It may have served, however, as an alternative binding site for Stat5-VP16-Jak2 when the partially overlapping STAT site had been mutated, because all three sites had to be mutated to completely abolish the transactivation (triple mutant). Regardless, the data presented for both enhancers strongly suggest that their activation by IL-2R signals is dependent on the binding of activated Stat molecules to a tandem element, which may be comprised of a higher affinity site (STAT) and a lower affinity site (3′ STAT-n).

Scrutiny of the regulation of the perforin gene is biologically important and offers a valuable model to shed molecular light on the activation/differentiation of CTLs. This line of investigation comprises a largely unexplored area in comparison to analogous investigations of cytokine genes, which have become widely used as a paradigm toward an understanding of Th activation/differentiation. In our study toward deciphering the control of perforin expression by IL-2, >45 kb of the human perforin gene locus was surveyed in a transgenic tissue culture system (Fig. 2) that mimics important aspects of perforin gene induction in primary CTLs (Fig. 1). This analysis led to two IL-2–responsive enhancers in the perforin gene locus (Fig. 3 and Fig. 4), which also promoted inducibility in T cells derived from transgenic mice, at least in the context of their intervening genomic sequences (Table). The activation of both enhancers by IL-2 required not only the activation of the Stat pathway (Fig. 5 C) but also the participation of tandem Stat-like elements that could be transactivated by Stat5 molecules in our CTL differentiation model (Fig. 6). These findings suggest that the activation of Stats by IL-2R signals, in particular the activation of Stat5 molecules, plays an important role for the generation of CTLs.

The regulation of the perforin gene by IL-2R signals via Stat proteins is consistent with several observations besides the well documented activation of Stat5 and Stat3 by IL-2R signals 35. Stat proteins are latent transcription factors 36 and, accordingly, the onset of perforin mRNA induction by IL-2 does not require newly synthesized proteins in our model cell line (Fig. 1 A) nor in primary cells 3. Similarly, the activation of Stat5 37, the induction of perforin mRNA (Fig. 1 B), and the activation of its enhancers (Fig. 5 A), as well as perforin mRNA induction by IL-2 in primary cells 31, are all resistant to rapamycin. Conversely, TGF-β, which has been suggested to block the activation of the Jak/Stat pathway in T cells, including the activation of Stat5 38,39, also prevents the induction of perforin by IL-2 40.

Importantly, the analysis of Stat5 knockout animals, as well as the phenotypes of animals lacking the Stat docking sites of IL-2Rβ, have recently indicated that this IL-2R signaling pathway may target the perforin gene in vivo and may be essential for the generation of cytotoxicity. Stat5 exists in humans and mice as two closely related genes with overlapping expression patterns and, therefore, has an often redundant role in vivo for both nonlymphoid and lymphoid tissues 41,42. Nevertheless, immunologically relevant phenotypes, albeit perhaps rather discrete, have been reported for certain single-deficient animals. Stat5a-deficient animals fail to upregulate IL-2Rα in response to IL-2R signals 43. Stat5b-deficient splenocytes, on the other hand, poorly generate CTLs in response to IL-2 or IL-15. This defect, which is accompanied by profoundly reduced but still detectable levels of perforin mRNA, led to the proposal that perforin is a Stat5-regulated gene 44. This notion is supported by our results. Similarly, splenocytes of mice expressing an IL-2Rβ lacking the Stat docking sites fail to generate CTLs in response to IL-2 45.

The redundancy of Stat5a and Stat5b may relate to the incomplete block of perforin gene induction by IL-2 or IL-15 in Stat5b-deficient splenocytes inasmuch as Stat5a-deficient cells also expressed slightly reduced levels of perforin 44. The inability of double-deficient T cells to enter the cell cycle 42 does not allow us to readily address this issue experimentally due to the relatively long stimulations required for significant inductions. It is also possible that Stat5 molecules are facilitators rather than the essential players for perforin gene activation or that they are redundantly used with Stat3 molecules. Whereas the latter pathway was not functional in our SAM-19 model (see Results), another report published while this manuscript was in preparation indicates that the upstream enhancer of the perforin gene could also be activated by Stat3 molecules 46. The authors identified a constitutively enhancing fragment similar to the upstream enhancer of our analysis by transient transfections of ∼1,400 bp of the 5′ flank into YT cells, a constitutively perforin-expressing NK cell–like lymphoma. The function of this DNA in YT cells was dependent on what is referred to in Fig. 6 as the STAT element, whereas the 3′ STAT-n site was not investigated. Regardless, the STAT element was shown to bind constitutively activated Stat3 molecules present in the YT lymphoma and Stat5 molecules when extracts of primary NK cells exposed to IL-2 were applied, suggesting that Stat5 and Stat3 molecules may regulate the upstream enhancer of the perforin gene. It is conceivable that genes are regulated redundantly or activation stage specifically by Stat5a, Stat5b, and Stat3 molecules in T cells, because an impaired upregulation of IL-2Rα in response to IL-2R signals has been described not only for Stat5-deficient animals 42,43 but also for Stat3-deficient T cells 47.

Notably, Stat5 proteins may also play an essential role in the development and/or maintenance of lineages that constitutively express perforin, i.e., NK cells 2 and γ/δ T cells 48,49,50. NK cells are absent in animals deficient in both Stat5 proteins 42, as well as in animals deficient in the ability to activate Stats in response to IL-2R or IL-15R signals 45. The latter animals also lack γ/δ intraepithelial lymphocytes, a lineage that so far has not been investigated in the Stat5 double-deficient animals. The biological relevance of Stat5 molecules and the identified enhancers for the constitutive expression of perforin by these lineages remains to be analyzed. A constitutively active Stat signaling pathway in NK cells in vivo, and thereby also an activation of the perforin gene by the identified enhancers, could be envisioned based on the expression of IL-15 in virtually all tissues and the absolute requirement of IL-15R signals for NK development and/or maintenance 51,52.

The perforin enhancers and the enhancer of the human and mouse IL-2Rα genes 53,54,55,56,57 are presently the only suspected Stat5 and Stat3 targets in vivo among genes specifically expressed by lymphocytes. Interestingly, they are also similarly organized. Just like the perforin enhancers, the IL-2Rα enhancer requires a tandem Stat element that serves as a composite binding site for a tetrameric Stat5 complex binding individual sites of weaker affinity 55,57. We are presently addressing whether the perforin sites, each of which can bind Stat5 in vitro (Schindler, U., and M.G. Lichtenheld, unpublished observation), also serve as targets for a tetrameric Stat5 complex, because the spacing of the Stat-like elements in the perforin enhancers exceeds the spacing of those in the IL-2Rα enhancers (17 vs. 11 bp). The loss of enhancer function by an individual mutation of either site (Fig. 6 B) is consistent with a tetrameric assembly, but these experiments do not address the mechanism for the observed cooperation. The other element with a major function in the mouse and human IL-2Rα enhancers entails an Elf-1–binding Ets core located 14 bp 3′ of their tandem Stat elements 54,56. This organization is exactly like that of the far-upstream enhancer of the perforin gene (Fig. 6 A). An analogous element in the upstream enhancer is located 10 bp 3′ of the tandem Stat-like elements. The STAT sites of both perforin enhancers as well as the 3′ STAT-n site of the upstream enhancer incorporate an Ets binding motif sequence core similar to that of the human IL-2Rα enhancer. This element has been suggested to repress the enhancer in unactivated cells via an unidentified Ets family protein 54.

Unlike the IL-2Rα enhancer and the upstream enhancer of the perforin gene, the far-upstream enhancer may respond not only to IL-2R but also to TCR signals that have been implied as a second pathway for perforin gene induction 1,2,3, because this enhancer also responded to PMA and ionomycin in a CsA-sensitive manner in SAM-19 (Fig. 5 B). This observation could suggest a possible role for a nuclear factor of activated T cells (NFAT)-like element in the far-upstream enhancer whose sequence resembles the “IL140” element of the IL-3 enhancer, which is known to strongly activate transcription independently of AP-1 proteins 58. The physiological activation of the perforin enhancers is likely to involve transcription factors in addition to Stat5. The combined control of the perforin enhancers remains to be experimentally established and compared with that of the IL-2Rα enhancer. Based on the inspection of the perforin sequences (Fig. 6 A), other participating transcription factors could be IL-2 inducible, namely AP-1, cAMP-responsive element binding proteins, and nuclear factor κb 59,60,61, or differentially expressed during lymphoid development, namely Ikaros proteins 62 and proteins recognizing the E-box 63.

In summary, this study investigates the differentiation processes of CTLs by working backwards from the perforin gene. It indicates that the induction of the perforin gene by IL-2R signals involves at least two enhancers whose activation is dependent on Stat elements that can be targeted by Stat5 molecules. These results are consistent with and complement the ongoing findings of the reverse genetics analysis of the IL-2R Jak/Stat signaling pathway.

We are indebted to M. Nabholz for providing PC60 cells, B. Groner for providing the chimeric Stat5-VP16-Jak2 expression vector, and T. Malek for the Stat5a expression vector. We thank M. Gutierrez and E. Cepero for their technical assistance in the transgene experiments. We also thank L. Boise, T. Malek, and U. Schindler for helpful discussions and critical reading of the manuscript.

This work was supported by National Institutes of Health grant R01 CA55811 to M.G. Lichtenheld. The Transgene and Sequencing Facilities of the University of Miami School of Medicine are supported by the Sylvester Comprehensive Cancer Center.