PI3Kδ Is Essential for Tumor Clearance Mediated by Cytotoxic T Lymphocytes

Background PI3Kδ is a lipid kinase of the phosphoinositide 3-kinase class 1A family and involved in early signaling events of leukocytes regulating proliferation, differentiation and survival. Currently, several inhibitors of PI3Kδ are under investigation for the treatment of hematopoietic malignancies. In contrast to the beneficial effect of inhibiting PI3Kδ in tumor cells, several studies reported the requirement of PI3Kδ for the function of immune cells, such as natural killer and T helper cells. Cytotoxic T lymphocytes (CTLs) are essential for tumor surveillance. The scope of this study is to clarify the potential impact of PI3Kδ inhibition on the function of CTLs with emphasis on tumor surveillance. Principal Findings PI3Kδ-deficient mice develop significantly bigger tumors when challenged with MC38 colon adenocarcinoma cells. This defect is accounted for by the fact that PI3Kδ controls the secretory perforin-granzyme pathway as well as the death-receptor pathway of CTL-mediated cytotoxicity, leading to severely diminished cytotoxicity against target cells in vitro and in vivo in the absence of PI3Kδ expression. PI3Kδ-deficient CTLs express low mRNA levels of important components of the cytotoxic machinery, e.g. prf1, grzmA, grzmB, fasl and trail. Accordingly, PI3Kδ-deficient tumor-infiltrating CTLs display a phenotype reminiscent of naïve T cells (CD69lowCD62Lhigh). In addition, electrophysiological capacitance measurements confirmed a fundamental degranulation defect of PI3Kδ−/− CTLs. Conclusion Our results demonstrate that CTL-mediated tumor surveillance is severely impaired in the absence of PI3Kδ and predict that impaired immunosurveillance may limit the effectiveness of PI3Kδ inhibitors in long-term treatment.


Introduction
The common catalytic function of phosphoinositide 3-kinases (PI3Ks) is the phosphorylation of the D3-position of phosphatidylinositol. The PI3K family consists of three classes based on their primary structure, regulation, and in vitro liquid substrate specificity. Class I PI3Ks catalyze the phosphorylation of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) and thereby generate phosphatidylinositol 3,4,5-triphosphate (PIP 3 ). PIP 3 is selectively recognized by some pleckstrin homology domains and thus provides a membrane docking site for many different proteins, e.g. the serine-threonine-kinase AKT and its upstream activator the phosphoinositide-dependent kinase-1 (PDK1), the guanine nucleotide exchange factors for ARF6 ARNO (ARF nucleotide-site opener), the general receptor of phosphoinositide-1 (GRP1), and non-receptor tyrosine kinases of the BTK and TEC-family. Accordingly, class I PI3Ks impinge on many cellular signaling cascades, which affect cell growth and survival, trafficking of vesicles and dynamics of the actin cytoskeleton. As a consequence, the PI3K/AKT/mTOR pathway has been shown to play an important role in apoptosis and cancer [1].
Class I PI3Ks are heterodimeric molecules comprising a catalytic and a regulatory subunit. There are four catalytic isoforms of class I PI3Ks (class IA p110a, p110b, p110d and class IB p110c). The isoforms p110a and p110b are ubiquitously expressed, whereas p110d and p110c are predominantly expressed in the hematopoietic system [2,3]. Currently, tools to study PI3K signaling range from genetically modified mouse strains either lacking individual class I PI3K isoforms or harboring point mutations giving rise to catalytically inactive proteins, to PI3K isoform-specific small-molecule inhibitors [4].
T lymphocytes are of particular interest because they express all four catalytic isoforms. The enzymes can therefore be envisaged to have both, redundant and unique functions. In fact, T cells develop normally in mice with engineered deletions or kinase-dead (KD) versions of PI3Kd [5,6], but in PI3Kc-deficient mice T cells show partial defects in b-selection [7]. In contrast, mice deficient in both, PI3Kd and PI3Kc, suffer from a profound block at the pre-T cell receptor (pre-TCR) selection step of thymus development. In these mice the numbers of splenic CD4+ and CD8+ T cells are significantly reduced and the majority of peripheral CD4+ T cells display a memory phenotype [8,9]. Using small-molecule inhibitors, Ji et al [8] demonstrated that in mature T cells PI3Kd, but not PI3Kc, controls Th1 and Th2 cytokine secretion.
PI3Kd is a key component of the signaling machinery downstream of the TCR and CD28 [10] and it is the most relevant isoform responsible for PIP 3 accumulation at the immunological synapse upon TCR activation [10,11]. Hence, PI3Kd-deficient CD4+ T helper (Th) cells display defects in antigen-presenting cell-mediated stimulation and clonal expansion in vivo and in vitro [12]. PI3Kd-KD CD4+ T cells proliferate moderately slower in response to anti-CD3 (aCD3) stimulation, but this defect is rescued by antibody-dependent co-stimulation of CD28 [10]. Upon physiological stimuli PI3Kd-KD Th cells show reduced differentiation along the Th1 and Th2 lineages [12]. As a consequence of reduced Th2 responses, PI3Kd-deficient mice are protected from experimentally-induced airway inflammation [13]. Additionally, a study by Haylock-Jacobs et al [14] showed that PI3Kd is a key player in the pathophysiology of experimental autoimmune encephalomyelitis (EAE), a Th17-driven model of multiple sclerosis. Furthermore, loss of PI3Kd was also shown to compromise the function of regulatory T cells [15], another CD4+ T cell lineage.
While the role of PI3Kd in CD4+ T cells is understood in considerable detail, it is not clear whether the enzyme is important for the function of CD8+ cytotoxic T lymphocytes (CTLs). Thus, the aim of the present study was to determine the consequences of PI3Kddeficiency on CTL functions in vitro and in vivo. Our observations clearly show that PI3Kd is indispensable at several stages of CTL biology. PI3Kd-deficiency impedes the activation of CTLs and gives rise to inactive and quiescent CTLs, whose composition of the lytic machinery required for degranulation and target cell lysis is altered and functionally impaired. This defect severely curtails CTLmediated antigen-specific cytotoxicity and impairs tumor surveillance. PI3Kd-deficient mice develop significantly bigger solid tumors after inoculation with MC38 colon carcinoma cells. These results evoke safety concerns and challenge the use of PI3Kd inhibitors in cancer treatment. Impaired CTL-mediated immunosurveillance might limit the effectiveness of PI3Kd inhibitors by counteracting intended treatment effects on tumor target cells. However, we suggest that PI3Kd inhibitors might be of therapeutic relevance in areas where suppression of CD8+ T cells is useful, e.g. in transplantation medicine or in the treatment of autoimmune diseases and chronic obstructive pulmonary disease (COPD).

PI3Kd is Required to Induce T Cell Responses to Allogeneic Lymphocytes
As PI3K p110 isoforms might have redundant functions, we determined expression levels of PI3Ka, b, and c in PI3Kd2/2 CTLs. Figure S1 illustrates no major alterations or compensations of other PI3K p110 isoforms in PI3Kd2/2 CTLs.
The mixed lymphocyte reaction (MLR) shall induce T cell activation and enhanced proliferation. We challenged carboxyfluorescein succinimidyl ester (CFSE)-labeled wild type and PI3Kd2/2 splenocytes (C57BL/6 background) in vitro with allogeneic lymphocytes from BALB/c mice (different MHC haplotype). Over a period of 84 hours the proliferation of CD8+ CFSE-labeled cells was assessed via flow cytometry according to the incremental reduction in CFSE-intensity due to cell division ( Figure 1A). PI3Kd2/2 CTLs failed to react with enhanced proliferation upon challenge with allogeneic antigens, as their growth rates were comparable to unstimulated CD8+ T cells. Additionally, we performed pharmacological inhibition of PI3Kd with CAL-101. This potent and selective inhibitor of PI3Kd has already paved its way into human clinical trials [16], for review see [17,18]. The efficient inhibition of PI3Kd by CAL-101 was confirmed by abrogated phosphorylation of the downstream target AKT. Further, we ruled out that in vitro treatment of CTLs with CAL-101 evoked toxic effects ( Figure S2). Pharmacological inhibition of PI3Kd resulted in a reduced proliferative response of CD8+ CFSE-labeled cells towards allogeneic lymphocytes, mimicking the phenotype of PI3Kd2/2 CTLs ( Figure 1B). In contrast, wild type CTLs were rapidly activated and the cells proliferated significantly faster upon co-incubation with allogeneic splenocytes. These differences were not related to a general inability of PI3Kd2/2 CTLs to divide and grow, as proliferation in response to anti-CD3e ( Figure 1C) and interleukin-2 (IL-2, Figure 1D) was unaltered. Similarly, we failed to detect any alterations in the apoptotic behavior of PI3Kd2/2 CTLs after IL-2 withdrawal: apoptosis rates in wild type and PI3Kd2/2 CTLs were comparable, which was assessed via cell cycle staining with propidium iodide (data not shown). These experiments led us to the conclusion that PI3Kd is dispensable for CTL proliferation per se, but required in the activation process of CD8+ CTLs when challenged with allogeneic lymphocytes.
Calcium (Ca 2+ ) is released in response to TCR activation and is the final step to trigger CTL response. It was shown previously that Ca 2+ influx into PI3Kd2/2 CD4+ T cells was impaired [19]. Accordingly, we examined, whether a defect in early TCR signaling or altered Ca 2+ -concentration in intracellular stores accounted for the missing response of PI3Kd2/2 CTLs to foreign antigens. Intracellular Ca 2+ -flux was monitored after preloading cells with the fluorescent indicator dye Indo-1. Importantly, PI3Kd-deficiency did not interfere with Ca 2+ -influx triggered by TCR-crosslinking, indicating that early TCR signaling is intact in PI3Kd2/2 CD8+ T cells ( Figure 1E). We also analyzed the size of the intracellular releasable storage pool in the endoplasmic reticulum by blocking the sarco/endoplasmic reticulum Ca 2+ -ATPase with thapsigargin. The ensuing increase in intracellular Ca 2+ was virtually identical in PI3Kd2/2 and wild type CTLs ( Figure 1F). These experiments ruled out that altered Ca 2+mobilization accounted for any functional failure of PI3Kddeficient CTLs.

PI3Kd is Needed to Arm CTLs
The major task of CTLs is the eradication of cells containing non-self-antigens, i.e. cancer cells or virally infected cells. For their cytotoxic action CTLs rely on lytic granules filled with proteolytic enzymes such as granzymes and perforin. We evaluated the expression of these components in aCD3-activated CTLs and found significantly reduced mRNA levels for grzmA, grzmB and prf1 in PI3Kd2/2 CTLs (Figure 2A). This indicates that PI3Kd is required to arm CTLs for efficient lysis. Moreover, CTLs are WT and PI3Kd2/2 splenocytes were CFSE-labeled and cultured in the absence and presence of allogeneic (BALB/c), mitomycin C-treated splenocytes. At the indicated time points, cells were harvested and proliferation of responding CTLs was assessed by flow cytometry. Percentages of proliferating CFSE+CD8+ T cells with and without the stimulus of mixed lymphocytes are illustrated. Proliferating CD8+ T cells were discriminated from undivided T cells by the reduced levels of CFSE in daughter cells. B. WT splenocytes were CFSE-labeled and cultivated in analogy to A. Pharmacological inhibition of PI3Kd was achieved by treatment with indicated concentrations of CAL-101 during the experimental procedure. DMSO-treatment served as negative control. C. Proliferation of WT and PI3Kd2/2 CTLs in response to aCD3e treatment was assessed in a CFSE proliferation assay. D. Proliferation of WT and PI3Kd2/2 aCD3-activated T cells was assessed under standard T cell medium conditions (in the presence of IL-2) and after deprivation from IL-2 by performing an [ 3 H]-thymidine incorporation assay over 48 hours (with IL-2: WT: 120976491cpm; versus PI3Kd2/2: 124136501cpm; without IL-2: WT: 13926381cpm; versus PI3Kd2/2: 11406160cpm, n = 6, values represent mean6SD). E., F. WT and PI3Kd2/2 equipped with the death receptors TNF-related apoptosis-inducing ligand (TRAIL) and Fas ligand (FasL), which can directly induce apoptosis in target cells upon linkage to the respective counterparts. As shown in Figure 2B, PI3Kd-deficient CTLs express significantly lower mRNA levels of trail and fasl. Apart from granule-exocytosis and the death-receptor cytotoxicity pathway, activated CTLs produce prominent amounts of interferon-c (IFNc) which supports the antiviral response of the host. Again, also ifng mRNA was substantially reduced in the absence of PI3Kd ( Figure 2C). Accordingly, even when stimulated with the plant lectin concanavalin A for 48 h, PI3Kd2/2 splenocytes released less IFN-c protein than wild type splenocytes ( Figure 2D).

Granule Release by CTLs Depends on PI3Kd
So far, our data indicate that PI3Kd is required to equip the lytic granules of CTLs with cytotoxic components. We used two approaches to test whether PI3Kd is involved in the degranulation process. The first approach was performed on the single cell level and relied on the recording of cellular membrane capacitance by electrophysiological measurements. Upon degranulation lytic granules fuse with the cell membrane and lead to an increase in cellular surface area, which corresponds to a quantifiable rise in membrane capacitance. This alteration in membrane capacitance can be recorded with the whole cell patch clamp technique at the single cell level. Degranulation was induced by providing activating signals, i.e. stimulation of protein kinase C isoforms by the phorbol ester PMA and calcium influx by the calcium ionophore ionomycin. Figure 3A summarizes the membrane capacitance recordings of wild type and PI3Kd2/2 CTLs under basal conditions and after superfusion with PMA and ionomycin. Corresponding to differences in cell size, wild type and PI3Kd2/2 CTLs displayed a distinct range of basal membrane capacitances. However there was no significant difference in basal cellular capacitance between wild type and PI3Kd2/2 CTLs. Stimulation of wild type cells triggered a significant increase of approximately 1.5 fold cell capacitance, independent of their basal cell size ( Figure 3B). In contrast, no significant increase in cellular capacitance was detectable after superfusion of PI3Kd2/2 CTLs. Accordingly, upon pharmacological inhibition of PI3Kd in wild type cells no significant increase in cellular capacitance could be induced, as compared to untreated DMSO controls.
The second approach was based on flow cytometry and allowed for the analysis of a large cell population: the lysosomal marker CD107a is expressed on the cell surface at significant levels only after degranulation and thus has been used as a surrogate marker of cytotoxicity [20]. We therefore analyzed CD107a levels by flow cytometry under basal conditions and after stimulation with PMA and ionomycin. This combination elicited a prominent increase in surface expression of CD107a in wild type cells. In contrast, CD107a expression was not significantly increased upon PMA/ ionomycin stimulation of PI3Kd2/2 CTLs ( Figure 3C). Additionally, the impact of pharmacological inhibition of PI3Kd in wild type CTLs was investigated. The induction of degranulation was significantly decreasing proportional to the applied CAL-101 concentrations as compared to DMSO-treated cells ( Figure 3D). Treating cells with pharmacological compounds may interfere with more than one reaction or induce off-target effects, e.g. phorbol esters bind to MUNC13 (mammalian homologous of the C-elegans unc-13 gene), a component of the vesicle fusion machinery [21]. To confirm our data we also employed an independent stimulus to trigger degranulation again via elevation of intracellular Ca 2+ : electrical field stimulation (chronic low frequency stimulation, CLFS, 1 Hz). As electrical field stimulation has not been used in this context before, we employed several control experiments to prove the specific nature of the stimulatory effect. No significant increase in CD107a expression was induced, when Ca 2+ was removed from culture media (data not shown). Importantly, PI3Kd2/2 CTLs did not display any detectable degranulation upon electrostimulation, whereas wild type CTLs expressed high levels of CD107a on their surface ( Figure 3E). When CTLs were pre-treated with 10 nM concanamycin A, an inhibitor of perforin-mediated cytotoxicity [22], both CLFES ( Figure 3E) and pharmacological stimulation (not shown) did not lead to increased CD107a-surface expression. To sum up, both, the single cell-as well as the population-based approach verified that fusion of lytic granules with the cellular membrane was impaired in PI3Kd2/2 CTLs or due to pharmacological inhibition of PI3Kd.

PI3Kd-deficiency Impairs Antigen-specific Response in vitro and cytotoxicity of CTLs in vivo
So far we demonstrated that PI3Kd2/2 CTLs do not respond to foreign antigens with enhanced proliferation, that they express reduced levels of important components of the lytic machinery, and that they have a severe degranulation defect. Hence, all these data reveal an essential role of PI3Kd for CTL-mediated cytotoxicity at different stages in the canonical killing pathway. To investigate the antigen-specific cytotoxicity of CTLs in vivo, wild type and PI3Kd2/2 mice were immunized subcutaneously with the SIINFEKL peptide in combination with an adjuvant (CpG). One week thereafter mice received syngeneic target cells containing CFSE low splenocytes, CFSE mid splenocytes pulsed with an irrelevant peptide and CFSE high splenocytes pulsed with SIINFEKL mixed in a 1:1:1 ratio intravenously. Following 18 hours, the draining inguinal lymph nodes and spleens of wild type and PI3Kd2/2 mice were analyzed by flow cytometry. Figure 4A depicts a representative histogram plot obtained from lymph nodes of immunized wild type and PI3Kd2/2 mice: whereas the splenocytes loaded with SIINFEKL peptide were drastically reduced in lymph nodes from wild type mice after immunization, this effect was observed to a significantly lower extent in PI3Kd2/ 2 mice (summarized in Figure 4B). These experiments document a reduced cytotoxic ability of CTLs in the absence of PI3Kd. To further support this finding, we additionally challenged PI3Kd2/2 CTLs in an in vitro experimental set-up. We again immunized wild type and PI3Kd2/2 mice with SIINFEKL peptide plus adjuvant over a period of two weeks. Thereafter, splenocytes were prepared and co-cultivated with SIINFEKLloaded splenocytes for five days. The thymoma cell lines EL4 and its corresponding ovalbumin (OVA)-expressing cell line EG7 were used to test the cytolytic capacity of the splenocytes. EL4 cells should not be recognized and were used as targets to determine unspecific background killing, which was comparable between wild type and PI3Kd2/2 CTLs ( Figure 4C). In contrast, OVAexpressing EG7 cells were efficiently recognized and killed by primed wild type CTLs, whereas primed PI3Kd2/2 CTLs failed to induce any detectable antigen-specific cytotoxicity above background levels ( Figure 4D). To confirm the impact of PI3Kd splenocytes were stained with 1 mM Indo-1 AM. Ca 2+ flux in response to aCD3e followed by crosslinking with streptavidin (E) or thapsigargin (F) was measured in CD8+ T cells using flow cytometry. Treatment with ionomycin served as positive control. Three independent experiments were carried out and one representative experiment is shown, respectively. doi:10.1371/journal.pone.0040852.g001 on CTL-mediated antigen-specific cytotoxicity we isolated splenocytes from OT-I mice and co-cultivated them with SIINFEKLloaded splenocytes as described above. The CTLs were treated with the PI3Kd inhibitor CAL-101 2 hours prior and during the in vitro cytotoxicity assay. Again, OVA-expressing EG7 cells were efficiently lysed by DMSO-treated OT-I CTLs, whereas CAL-101-treated OT-I cells showed significantly reduced cytotoxicity ( Figure S3).

PI3Kd is Required for CTL Activation and Tumor Surveillance
At this point we speculated that PI3Kd inhibition -besides beneficial effects on tumor growth restriction -might entail adverse effects on the immune system. Thus, we wanted to challenge the concept of PI3Kd inhibition in the treatment of malignancies and tested whether impaired CTL-mediated cytotoxicity observed in PI3Kd2/2 mice was indeed relevant for tumor surveillance. We made use of the colon adenocarcinoma cell line MC38, which is recognized and lysed in a CTLdependent manner [23]. Fifteen days after subcutaneous injection of 10 6 cells into the flanks of wild type and PI3Kd2/2 mice, the animals were sacrificed and the tumors analyzed. As depicted and summarized in Figure 5A and 5B significantly larger tumors had developed in PI3Kd2/2 mice compared to wild type recipients. When we characterized the tumor-infiltrating lymphocytes, we found comparable numbers of CD3+CD8+ effector cells irrespective whether a tumor evolved in wild type or PI3Kd2/2 recipients ( Figure 5C). Investigating resting CTLs from untreated healthy mice, we found comparable expression profiles of the activation markers CD45RB and CD69 on wild type and PI3Kd2/2 lymphocytes, whereas the expression of CD44 was reduced in PI3Kd2/2 lymphocytes ( Figure 5D). Pharmacological inhibition of PI3Kd in mature wild type CTLs over a time period of 3 days resulted in reduced expression of CD44 while no difference was observed in the expression of CD69 compared to DMSO-treated wild type controls ( Figure S4). On tumor-infiltrating CTLs the expression of CD44 and CD45RB was comparable ( Figure 5E, F); no consistent and significant differences in their expression levels could be detected. In contrast, we found a significantly reduced expression of the activation marker CD69 on PI3Kd2/2 tumor-infiltrating CTLs. Thus, although PI3Kd2/2 CTLs migrated to the tumor, they were activated to a lower extent. Further proof for this concept was obtained by analyzing CD62L, which is down-regulated upon activation. CD62L was consistently higher in PI3Kd2/2 CTLs (compare Figure 5D depicting expression profiles of unchallenged, resting CTLs and Figure 5E, F showing expression levels on tumor-infiltrating CTLs). Moreover, prolonged pharmacological inhibition of PI3Kd in splenic wild type CTLs resulted in significantly increased expression of CD62L compared to DMSO-treated wild type controls ( Figure  S4). Splenic CD8+ T cells comprise different subsets, amongst which naïve T cells are the most prevalent ones. We could not observe any toxic effect of CAL-101 on the entity of CD8+ T cells. Hence, we are convinced that the differences in CD44 and CD62L are not the result of different CD8+ subset susceptibilities to CAL-101.
In summary, these experiments clearly verified reduced CTLmediated tumor surveillance in PI3Kd2/2 mice accompanied by decreased activation of PI3Kd2/2 CTLs ( Figure 5D-F).

Discussion
T cells express all four catalytic isoforms of PI3K. Robust networks are engineered to be resilient by built-in redundancies: because of redundant trajectories, removal of one nod does not affect the flow of signals through the network. In each instance, the product of PI3Ks' enzymatic catalysis is PIP 3 , a lipid second messenger that supposedly activates identical effectors. It was therefore surprising to observe that removal of PI3Kd had such a profound effect on the function of CD8+ cytotoxic cells supported by the following key findings: (i) The components required for their eponymous action were only expressed at low levels in PI3Kd-deficient CTLs (granzyme A and B, perforin, FasL, TRAIL, and IFN-c). (ii) Granule release and (iii) antigen-induced clonal expansion was impaired in PI3Kd-deficient CTLs. (iv) Predictably, in the absence of PI3Kd, CTLs failed to restrict MC38 tumor cell growth in vivo. Our findings suggest that the absence of PI3Kd in CD8+ T cells phenocopies the deficiencies that have been observed in PDK1-deficient CTLs and CTLs treated with an inhibitor of AKT1 [24], suggesting that low expression levels of granzymes, perforin and FasL might be due to disrupted AKT-dependent phosphorylation of Foxo3 (forkhead box O3). Similarly, CTLs treated with an AKT1-inhibitor or deficient in PDK1 also failed to mount a proliferative response upon antigen challenge. PIP 3 , the second messenger produced by PI3Ks, activates clearly more effectors than only AKT and PDK1. However, PI3Kd-deficiency, AKT-inhibition and abrogation of PDK1 expression result all in similar phenotypic consequences, i.e. impaired expression of cytotoxic proteins. Hence we speculate that the pertinent signaling module that mediates extracellular input to transcriptional control is composed of PI3Kd, PDK1, AKT and Foxo3. In this module, the requirement for PI3Kd is absolute; the other PI3K-isoforms cannot compensate for its absence. Similarly, antigen-induced, TCR-dependent cell proliferation is contingent on the module comprising PI3Kd, PDK1 and AKT. Our observations suggest that this is not necessarily the case for other stimuli such as IL-2, a conclusion that was reached by McIntyre et al [24] in a similar way.
CTLs can kill infected or malignant cells via degranulation. We demonstrate here for the first time that PI3Kd is indispensable for degranulation of CTLs. Our observations extend comparable findings with mast cells [25] and Natural Killer (NK) cells [26]. Whereas the indispensable role of PI3Kd in degranulation and cytokine secretion is beyond dispute, its relevance for NK cell cytotoxicity is seen controversially [26][27][28][29][30][31]. It is very likely that the deficiency in PI3Kd impairs degranulation by effector mechanisms other than downstream signaling via AKT. The following arguments support this conjecture: (i) Capacitance measurements 160.076; versus PI3Kd2/2: 0.41960.075, n = 6, p,0.0001) and prf1 (WT: 0.043360.004; versus PI3Kd2/2: 0.01360.003, n = 6, p,0.0001) was quantified by qRT-PCR and normalized to the house-keeping gene gapdh. B. Similarly, under standard culturing conditions mRNA levels of trail (WT: 0.004960.0012; versus PI3Kd2/2: 0.002660.0007, n = 4, p = 0.0178) and fasl (WT: 0.013860.0015; versus PI3Kd2/2: 0.003960.0001, n = 4, p,0.0001) were measured. C. Ifng mRNA was quantified by qRT-PCR under standard culturing conditions (WT: 0.01960.0005; versus PI3Kd2/2: 0.00460.001, n = 6, p,0.0001) and after stimulation with 5 ng/ml IL-12 for 4 h (WT: 0.24160.06; versus PI3Kd2/2: 0.08760.016, n = 6, p = 0.0002). D. To quantify IFN-c protein levels, WT and PI3Kd2/2 splenocytes were stimulated with ConA. After 48 h supernatants were harvested and IFN-c release was measured by ELISA (WT: 651562061 pg/ml; versus PI3Kd2/2: 135961147 pg/ml, n = 3, p = 0.0193). IFN-c release of unstimulated controls of WT and PI3Kd2/2 splenocytes was below detection limit of the assay (,10 pg/ml). Statistics were calculated with an unpaired Student's t-test, and values represent mean6SD. One out of two independently performed experiments with comparable results is shown. doi:10.1371/journal.pone.0040852.g002  [26,32]. The absence of PI3Kd impaired this very rapid fusion event, suggesting that the depletion of PIP 3 interferes with the fusion per se. (ii) Electrical field stimulation triggers the fusion event directly, i.e. by calcium influx and presumably by calcium-dependent activation of SNARE proteins (soluble NSF-attachment receptors) [33]. PI3Kd-deficient CTLs failed to exocytose and to expose CD107a despite unaltered calcium responses. (iii) Phosphoinositides are known to be asymmetrically distributed. Due to their acidic nature they have been proposed to contribute to SNARE-dependent fusion [34]. In addition, Low et al [35] demonstrated a contribution of PI3Kd to membrane trafficking at the trans-Golgi network. This resulted in impaired cytokine-secretion in macrophages and might also contribute to the reduced secretion of IFN-c of PI3Kd-deficient CTLs that we observed.
The disclosure of the indispensable role of PI3Kd for CTL function led us to envision potential limitations or safety concerns regarding the clinical applicability of PI3Kd inhibitors in humans. Thus, we explored how PI3Kd-deficiency affected CTL-mediated tumor surveillance in vivo, thereby challenging the therapeutic concept of PI3Kd inhibition. PI3Kd2/2 CTLs infiltrated evolving tumors with unaltered frequency, but CD69 expression was severely reduced. Increase in CD69 expression is directly linked to T-lymphocyte activation [36]. Additionally, we observed a concomitant enhanced CD62L expression, which is indicative of reduced activation [37]. The failure of PI3Kd2/2 cells to downregulate CD62L has also been observed by others, who documented the importance of CD62L as regulator of lymphocyte recirculation [38]. Whereas defective T cell recirculation may e.g. explain the increased susceptibility of PI3Kd2/2 mice towards Leishmania major second infection [39], it is unlikely that it accounts for the increased tumor growth we observed. We conclude that several deficiencies contribute to impaired immunosurveillance by PI3Kd-deficient CTLs: (i) In vivo, PI3Kd2/2 CTLs acquire a phenotype reminiscent of naïve cytotoxic T cells. In contrast to terminally differentiated effector CTLs (CD62L low , full equipment with lytic machinery), naïve cells (CD62L high ) are only partially equipped and do not function properly. Indeed, the cytolytic capacity of PI3Kd-deficient T cells was found to be severely impaired both in vivo and in vitro. (ii) An additional factor leading to this naïve-like phenotype is the inability of PI3Kd2/2 CD8+ T cells to react to foreign antigens via clonal proliferation. This was substantiated by the results obtained in the mixed lymphocyte reaction. (iii) Impaired secretion of IFN-c by PI3Kd2/2 CTLs is likely to contribute to defective tumor surveillance [40], given the observation that PI3Kd regulates IFN-c production in several lymphocytic lineages [26,28] and is required for TCR-induced IFN-c production [39,41].
The therapeutic concept of PI3Kd inhibition in the treatment of hematopoietic malignancies is based on the knowledge that the PI3Kd isoform is important in sustaining growth of leukemic cells [4,16,26,42]. Accordingly, the specific PI3Kd inhibitor CAL-101 have already entered clinical trials for the treatment of hematopoietic malignancies [17,18,43,44]. Similarly, inhibitors for AKT are under development [45]. However, AKT is a questionable target. Due to its wide expression pattern, its inhibition is predicted to affect many cellular functions. Isoform-specific inhibitors are not yet available and their applicability is considered doubtfully [46]. At the current stage, specific PI3Kd-inhibitors can be assumed to be superior to AKT-inhibitors. We believe that these compounds will have fewer side effects (due to restricted expression pattern of PI3Kd) and be more efficacious.
Nevertheless, PI3Kd-specific inhibition has unintended consequences on the functional activity of CTLs at all levels: activation, antigen-induced clonal expansion, degranulation and cytotoxicity. Therefore, unintended side effects such as impaired CTLmediated tumor surveillance might counteract or even reverse intended treatment effects of PI3Kd-inhibitors. As a consequence, in the clinical development of PI3Kd-inhibitors, rigorous monitoring of tumor development should be implemented. Additionally, contraindications should be carefully defined, especially with respect to pre-existing immunodeficiency and/or tumor development in a patients history. However, while impaired CTL-function may be a drawback in cancer therapy, it is evident that the suppression of CTLs would be of profound benefit in the treatment of autoimmune diseases and other CD8+ T cellassociated diseases, such as COPD (chronic obstructive pulmonary disease) [47] and even more, in transplantation medicine. This concept is supported by a recent study of Ying et al [48], who investigated the therapeutical potential of pharmacological PI3Kd inactivation in murine models of heart and skin transplantations.

Ethics Statement
All experiments were conducted in accordance with protocols approved by the Animal Welfare Committee of the Medical University of Vienna (66.009/0155-II/3b/2011) and the Austrian Federal Ministry of Science and Research.

Pharmacological Inhibition of PI3Kd
The selective PI3Kd inhibitor CAL-101 was purchased from Selleck Chemicals, IC-87114 was provided by ICOS (Bothell, Figure 4. PI3Kd-deficiency leads to severely impaired CTL cytotoxicity in vivo and in vitro. A., B. WT and PI3Kd2/2 animals were immunized with SIINFEKL and CpG. Seven days later mice received CFSE-labeled targets and peptide-specific CTL activity in draining inguinal lymph nodes was analyzed by flow cytometry. A. A representative FACS histogram for each genotype is depicted. B. SIINFEKL-specific target cell killing was calculated from three independent experiments as described in Materials and Methods (WT: 8066% antigen-specific lysis; versus PI3Kd2/2: 52610% antigen-specific lysis, n = 12; values represent mean6SD, unpaired t-test, p,0.0001). No specific killing was observed in control mice (data not shown). C., D. WT and PI3Kd2/2 animals were immunized with SIINFEKL peptide and CpG and boosted 7 days later. To generate effector cells, splenocytes were isolated at day 14 and co-cultured for 5 days with irradiated SIINFEKL-pulsed splenocytes. To determine peptide-reactive CTL cytotoxicity in vitro, CFSE-labeled EL4 (C) and OVA-expressing EG7 (D) target cells were co-cultured with effectors in ratios of 30:1, 15:1, 5:1 and 1:1. Specific in vitro target cell killing was quantified by flow cytometry (EG7, E:T = 30:1: WT: 63% specific lysis; versus PI3Kd2/2: 15% specific lysis). One representative experiment out of three is shown. doi:10.1371/journal.pone.0040852.g004 PI3Kd Is Indispensable for CTL-Mediated Immunity PLoS ONE | www.plosone.org

Generation of Splenocytes and Expansion of aCD3activated T Cells in vitro
Spleens from WT and PI3Kd2/2 animals were collected and forced through a 70 mm cell strainer. The resulting single cell suspension was treated with red blood cell lysis buffer. T cells were activated in vitro by stimulation with aCD3e antibody (clone 145-2C11; 0.5 mg/ml) and expanded in culture for 3 days in T cell medium composed of RPMI-1640 containing L-glutamine (PAA) and supplemented with 10% FCS, 50 mM 2-mercaptoethanol, 1 mM sodium pyruvate (Gibco), non-essential amino acids (PAA), 100 U/ml penicillin, 100 mg/ml streptomycin and 100 U/ml rhIL-2 (ProleukinH, Novartis).

Mixed Lymphocyte Reaction (MLR)
BALB/c splenocytes were growth arrested by treatment with mitomycin C (50 mg/ml, 20 min, Sigma). C57BL/6 WT and PI3Kd2/2 splenocytes were labeled with 2.5 mM carboxyfluorescein succinimidyl ester (CFSE). 1610 5 CFSE+ responding splenocytes (C57BL/6) were co-cultured in a ratio of 1:1 and 1:4 with stimulating splenocytes (BALB/c). Pharmacological inhibition of PI3Kd was achieved by treatment of C57BL/6 WT splenocytes with 0.1 mM, 0.5 mM and 1 mM CAL-101 during the entire experimental procedure. After 24 h, 36 h, 48 h, 60 h, 72 h and 84 h co-cultures and controls were harvested and stained with aCD8-APC to determine CD8+-specific T cell proliferation by flow cytometry. Dead cells were excluded from analysis via gating in the FSC/SSC dot plot. % proliferating CD8+ cells define the percentage of cytotoxic T cells showing reduced CFSE intensity compared to the initial CFSE staining intensity.

CFSE Proliferation Assay
Splenocytes were labeled with 2.5 mM CFSE and cultured in T cell medium supplemented with 0.5 mg/ml aCD3e antibody (clone 145-2C11) at a concentration of 5610 5 cells/ml. After 12 h, 48 h, 62 h and 86 h splenocytes were harvested and stained with aCD8-APC antibody to determine CD8+-specific T cell proliferation characterized by the incremental loss of CFSE intensity in a flow cytometer. Dead cells were excluded from analysis via gating in the FSC/SSC dot plot. After washing twice, cells were again incubated for 30 min at 37uC in cell loading buffer. Indo-1 AM loaded cells were stained for CD8 (clone 53-6.7). After staining, cells were resuspended in cell loading buffer. Cells were gated for CD8 expression and analyzed at 37uC on a LSRII flow cytometer (BD Bioscience, Heidelberg, DE). Ca 2+ flux was either induced by using 1 mM thapsigargin (Sigma) or by stimulation with a biotinylated aCD3 antibody (10 mg/ml) followed by crosslinking with streptavidin (Sigma) (10 mg/ml). 3.3 mM ionomycin (Sigma) was used to induce full-scale deflection of internal Ca 2+ . The median fluorescence intensity of the Indo-1 AM violet versus Indo-1 blue ratio, obtained using FlowJo software (Tree Star), of at least three samples were combined using GraphPad PrismH and plotted against time.

Capacitance Measurements
Capacitance measurements were performed in the whole-cell mode of the patch-clamp technique using the time domain method of Lindau and Neher [51]. As described for NK cells [26], as soon as stable whole-cell conditions were established, recordings of aCD3-stimulated and in vitro expanded T cells were started under superfusion (DAD-8-VC superfusion system) and PI3Kd2/2 animals. C. After removal, tumors were minced and digested with collagenase D and DNase I. Tumor-infiltrating CD3+CD8+ CTLs were with control solution (140 mM NaCl, 2.5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES NaOH, pH 7.4) with a sample rate of 330 kHz and continued under superfusion with Ca 2+ -ionophore (1 mM ionomycin dissolved in DMSO) and 500 nM phorbol myristyl acetate (PMA, Sigma). IC-87114 was used at 1 and 4 mM concentrations added to standard medium 1 hour prior to the analysis; DMSO treatment served as control. Cell capacitance was calculated by integrating the area under the capacitive current transiently elicited by a 20 ms voltage step from 2120 to 280 mV. This area was divided by the applied change in voltage (40 mV). Capacitance-measurements were performed at room temperature (2261.5uC) using an Axoclamp 200B patch clamp amplifier (Axon Instruments, Foster City, CA). Pipettes were pulled from aluminum silicate glass (AF150-100-10, Science Products, Hofheim, Germany) with a P-97 horizontal puller (Sutter Instruments, Novato, CA), heat-polished on a microforge (MF-830, Narishige, Japan), and had resistances between 1 and 2 MV when filled with the recording pipette solution (105 mM CsF, 10 mM NaCl, 10 mM EGTA, 10 mM HEPES, pH 7.3). Voltage-clamp protocols and data acquisition were performed with pclamp 6.0 software (Axon Instruments) through a 12-bit A-D/D-A interface (Digidata 1200; Axon Instruments).

Flow Cytometric-based Degranulation Assay
Degranulation of cytotoxic cells was quantified by monitoring the expression of CD107a on the cell surface as described previously [20]. Degranulation of WT and PI3Kd2/2 CTLs was triggered by 100 nM PMA and 0.33 mM ionomycin, or due to chronic low-frequency electrical stimulation. The PI3Kd inhibitor CAL-101 was applied at different concentrations (0.1 mM, 0.5 mM, 1m M and 5 mM) two hours prior to the stimulation with PMA/iono.

Chronic Low-frequency Electrical Stimulation (CLFES)
WT and PI3Kd2/2 aCD3-activated CTLs were seeded at a cellular density of 3610 5 cells/ml T cell medium. Electrical stimulation was performed using the C-Pace EP from IonOptix Corporation (Milton, USA). Stimulation pulses of 5V amplitude and 5 ms duration were delivered via two carbon electrodes to 3.5cm dishes at a frequency of 1Hz. Three stimulation time periods (30 min, 10 min and 1 min) were tested. After the stimulation CTLs were directly subjected to the FACS-based analysis: stimulation-induced surface expression of CD107a was compared to the unstimulated situation (see above, flow cytometric-based degranulation assay). The most pronounced degranulation was obtained if CTLs were subjected to continuous stimulation for 10 minutes.

In vivo CTL Assay
In vivo cytotoxicity was measured according to Schellack et al [52]. In brief, mice were immunized by subcutaneous injection of 0.1mg/mouse SIINFEKL (Bachem) in combination with the adjuvant CpG-ODN 1668 (Eurofins). Seven days later control mice and immunized mice received syngeneic splenocytes labeled with three different concentrations of the intracellular dye CFSE: 2.5, 0.25 and 0.025 mM. The CFSE high population was pulsed with the relevant SIINFEKL peptide (10 mg/ml), the CFSE mid population was pulsed with the irrelevant peptide m-TRP2 181-188 (10 mg/ml, Bachem) and the CFSE low population remained untreated. The three CFSE+ populations were mixed in a 1:1:1 ratio and 3610 7 cells were injected via the tail veins of recipient WT and PI3Kd2/2 mice. After 18 hours spleens and draining lymph nodes were removed. Single cell suspensions were analyzed by flow cytometry. Specific killing was calculated as [1-(% CFSE high /% CFSE low )] x 100.

Generation of Peptide-reactive T Cells
Mice were immunized by subcutaneous injection of the OVA agonist peptide SIINFEKL (0.1 mg/mouse) in combination with the adjuvant CpG-ODN 1668. After seven days, mice were boosted by a second injection. Spleens were removed at day 14 and suspensions of splenocytes were obtained from control mice and immunized mice as described above. In parallel, splenocytes from WT mice were prepared, irradiated (30 Gy -gamma ray) and pulsed with SIINFEKL (10 mg/ml in RPMI-1640 medium containing L-glutamine and supplemented with 10% FCS, 50 mM 2-mercaptoethanol, 100 U/ml penicillin, and 100 mg/ml streptomycin). 5610 6 splenocytes of immunized or control mice were cocultured with 2610 6 irradiated, SIINFEKL-pulsed splenocytes in 1 ml T cell medium for 5 days.

In vitro Cytotoxicity Assay
Splenocytes of control mice and splenocyte suspensions containing peptide-reactive T cells were co-cultured with 5610 4 CFSE-stained (2.5 mM) EL4 or EG7 target cells at effector-totarget (E:T) ratios of 30:1, 15:1, 5:1 and 1:1 in triplicates in 96-well plates. In parallel, tumor cells were incubated in the absence of splenocytes to assess the extent of spontaneously occurring apoptosis. After 18 h, 5610 4 PKH26-stained control cells were added to each well as internal control and cytotoxicity was quantified via flow cytometry. 10 4 labeled cells (either CFSE+ or PKH26+) were counted and the CFSE+ target cells were calculated as percentage of labeled tumor cells. Dead target cells were discriminated from living cells in a control staining with propidium iodide (Sigma) and further distinguished by determination of forward and sideward scattering. The percentage of specific lysis was calculated as [1-(%CFSE+ target cells after coincubation)/(%CFSE+ cells without co-incubation)]6100.

MC38 Tumor Model and Analysis of Infiltrating Lymphocytes
The flanks of WT and PI3Kd2/2 animals were depilated and 1610 6 MC38 cells were injected subcutaneously in each flank. After 15 days tumors were removed and weighed. For flow cytometric analysis the tumors were minced and digested with 1 mg/ml collagenase D (Roche Applied Sciences) and 0.05 mg/ ml DNase I (Roche Applied Sciences) for 1h at 37uC. The digested tissue suspension was squeezed through a 70 mm cell strainer and washed twice with PBS prior to antibody staining.

Cell Cycle Analysis
Cell cycle and apoptosis analysis were conducted according to Hoelbl et al [53]. Dead cells were defined as cells with DNA contents lower than 2 n (sub-G0/G1).

Whole Cell Extracts and Western Blot Analysis
Cells were lysed in 50 mM Tris/HCl pH 8.0, 10% (v/v) glycerol, 25 mM EDTA, 150 mM NaCl (all from ROTH), 2 mM DTT, 0.5% NP40 (Igepal CA-630), 25 mM NaF, 1 mM sodium vanadate, 0.5 mM PMSF, SIGMAFAST Protease Inhibitor (all from Sigma Aldrich Austria), and cell debris removed by centrifugation and re-suspended in 50 ml 26 Laemmli sample buffer. Proteins were separated with SDS-PAGE and blotted onto nitrocellulose membranes (Hybond, GE Healthcare Austria). PageRulerH Prestained Protein Ladder (Fermentas ThermoScientific Austria) was used as molecular weight standard. Membranes were probed and analyzed using the ECL Western blotting detection system (GE Healthcare Austria). The antibodies PhosphoDetect TM aAKT1-pThr 308 (Calbiochem) and aAKT (Cell Signaling Technology, New England Biolabs GmbH Germany) were a kind gift from J. Werzowa (Department of Clinical Pharmacology, Medical University of Vienna, Vienna, Austria). The peroxidase-conjugated secondary rabbit antibody was from GE Healthcare (Austria).