Runx3 controls multiple aspects of the CTL differentiation program, in part through induction of Eomes
Because Runx3 is highly expressed in peripheral CD8+ T cells, and because of the T-bet-Runx3 cooperation we observed earlier in CD4+ T cells (15), we examined the role of Runx3 in effector CTL differentiation. We isolated CD8+ T cells from Runx3-/- (KO) mice of the outbred ICR background and their WT Runx3+/+ littermates by positive selection with anti-CD8 magnetic beads (Figs. S1 and S2, available at http://www.jem.org/cgi/content/full/jem.20081242/DC1). Strikingly, Runx3-/- CD8+ T cells were strongly impaired in their ability to differentiate into effector CTLs, as judged by expression of perforin, granzyme B, and IFN-gamma (Fig. 3). Compared with WT T cells, perforin mRNA and protein expression were essentially undetectable in Runx3-/- T cells at day 6 of culture (Fig. 3, A and B). Runx3-/- T cells also had no detectable Eomes expression; in contrast, T-bet expression was unimpaired (Fig. 3 A). Furthermore, Runx3 was required for maximal production of IFN-gamma, but not TNF or IL-2, by CD8+ T cells restimulated at day 6 (Fig. 3 C). 
We previously reported that Th1 cell differentiation was regulated through a feed-forward loop in which T-bet is up-regulated early and induces Runx3, after which T-bet and Runx3 cooperate to induce IFN-gamma and silence IL-4, thus promoting stable differentiation toward the Th1 lineage (15, 22). Because (a) Runx3 appeared necessary for Eomes induction (Fig. 3 A), (b) the kinetics of Eomes expression paralleled those of perforin expression (Fig. 2), and (c) overexpression of Eo-VP16 in either WT or T-bet-deficient T cells led to an increase in both perforin and IFN-gamma expression (Fig. 2, B and D), we asked whether CTL differentiation was also potentially regulated by a feed-forward loop involving these same two classes of Runx and T-box transcription factors. Specifically, we asked whether Runx3, which was necessary for Eomes induction, then cooperated with Eomes to regulate transcription of the effector CTL markers perforin, IFN-gamma, and granzyme B. 
To test this hypothesis, we used chromatin immunoprecipitation (ChIP) assays to ask whether Eomes and Runx3 bound regulatory regions of the Prf1, Ifng, and Gzmb genes (Fig. 3 D). Both proteins associated with gene regulatory regions in differentiated CTLs. Runx3 bound to the Prf1 and Gzmb transcription start sites (TSS); to a known IL-2 responsive enhancer located near -1 kb of the Prf1 gene (23); to the distal CTL-specific DNase I hypersensitive site 9 in the Prf1 locus (24); to the Ifng promoter near the TSS, as previously reported for Th1 cells (10); and to several DNase I hypersensitive sites in the Ifng locus (Fig. 3 D and not depicted) (25). Eomes bound primarily to the Prf1 TSS and the -1 kb enhancer; this binding was substantially greater than that observed at the promoter of the Il2rb gene, a known direct target of Eomes (8), and comparable to that observed at the Ifng TSS, a known target of T-box proteins in both Th1 and CD8+ T cells (Fig. 3 D) (17). 
To determine whether Runx3 controlled the expression of CTL effector genes through its induction of Eomes, we retrovirally expressed Runx3 and Eo-VP16 in CD8+ T cells from Runx3-/- mice. Because of the limited number of CD8+ T cells in these mice, and because we saw no difference between Runx3-/- CD8+CD4- SP and CD8+CD4+ DP cells in our previous experiments, we used total Runx3-/- CD8+ T cells without further fractionation as recipients for retroviral transduction. Reconstitution of Runx3-/- CD8+ T cells with Runx3 restored expression of Eomes as well as perforin, granzyme B, and IFN-gamma (Fig. 4, A and B). In addition, Runx3-/- T cells showed a compensatory up-regulation of Runx1, which was suppressed upon reconstitution with Runx3, indicating that Runx1 is a target of repression by Runx3. Notably, Eo-VP16 did not up-regulate perforin expression when expressed in Runx3-/- cells, even though it restored the capacity to induce IFN-gamma expression upon TCR restimulation (Fig. 4, A and B). This result suggests strongly that perforin expression requires Runx3 and Eomes. 
As expected from their defect in perforin and granzyme B expression, Runx3-/- CD8+ T cells showed defective cytolytic activity in a mixed lymphocyte reaction (12). However, TCR-stimulated Runx3-/- CD8+ cells were as effective as WT cells in killing tumor cells in a redirected CTL assay (12). Furthermore, CD8+ cells from the peritoneal cavity of Runx3-/- mice immunized with certain tumor cells effectively killed these targets (13). Therefore, although activation of the perforin/granzyme B machinery is defective in Runx3-/- CD8+ cells, these cells are not entirely devoid of cytolytic activity and could still effectively kill targets, possibly by alternative mechanisms such as the Fas-Fas ligand pathway. 
