Phosphatidylinositol 3-Kinase p110δ Isoform Regulates CD8⁺ T Cell Responses during Acute Viral and Intracellular Bacterial Infections

CD8⁺ T cells are a critical component of the adaptive immune response, regulating immunity to neoplastic cells and intracellular microbial pathogens. During viral or intracellular bacterial infections, Ag-specific naive CD8⁺ T cells are activated, which then proliferate rapidly into differentiated effector CD8⁺ T cells (1). These effector CD8⁺ T cells subsequently clear infected cells through mechanisms involving production of cytokines such as IFN-γ as well as cytotoxic molecules such as perforin and granzymes (1). Multiple molecular mechanisms may regulate these processes, some of which may potentially involve molecules such as class I PI3Ks.

Class I PI3Ks belong to a family of lipid kinase heterodimers consisting of a catalytic and a regulatory subunit. These enzymes can phosphorylate phosphatidylinositol 4,5-bisphosphate into phosphatidylinositol 3,4,5-triphosphate (PIP₃) (2, 3). PIP₃ provides binding sites for the pleckstrin homology domains of different signaling proteins, such as the serine-threonine kinase B or Akt (2). This in turn activates signaling pathways that can promote survival, proliferation, migration, and differentiation of cells (2, 3). Class I PI3Ks are further divided into two groups: class IA and class IB. There are three catalytic isoforms of class IA PI3K (p110α, p110β, and p110δ) and one catalytic isoform of class IB PI3K (p110γ) (3).

The p110δ isoform of class IA PI3K is highly expressed in immune cells and is an important signaling molecule in lymphocytes (4, 5). PI3K can be activated in T cells by TCR and CD28 signaling, with p110δ being the main PI3K isoform responsible for accumulation of PIP₃ at the immunological synapse during TCR activation (6, 7). T cell development is not visibly affected in mice with p110δ deletion or the kinase-dead (KD) version of p110δ (8, 9). However, mice lacking both p110δ and p110γ showed a profound blockade at the pre-TCR selection step during T cell development (10, 11) when compared with mice deficient in p110γ alone (12), suggesting a level of redundancy between these two kinases. Additionally, the lymph nodes (LNs) of p110δ KD mice had normal ratios of CD4⁺ and CD8⁺ T cells; however, CD44 expression was reduced, indicating a possible role for p110δ in effector/memory T cell differentiation or survival (6).

The proliferation of p110δ KD CD4⁺ T cells is impaired and shows reduced production of IFN-γ, IL-2, and IL-4 (6, 13). Additionally, there is defective Th1, Th2, and T follicular helper cell differentiation in p110δ KD CD4⁺ T cells, as determined from in vitro studies (13, 14). A PI3K p110δ inhibitor, IC87114, can block proliferation and cytokine production of naive and effector/memory human T cells (15). This inhibitor can also impair the recall response of human memory T cells from allergic and rheumatoid arthritis patients (15). Pharmacological inhibition through administration of IC87114 or genetic inactivation of p110δ can reduce disease in in vivo models of asthma (16, 17), allergy (18), inflammatory arthritis (19), and contact-hypersensitivity reactions (15). During immune responses to Leishmania major, which involve CD4⁺ Th1 cells, p110δ KD mice have impaired T cell responses during both primary and secondary infections (20, 21). PI3K p110δ can play a critical role in promoting Th17 differentiation and in driving the pathogenesis of experimental autoimmune encephalomyelitis, a murine model of multiple sclerosis (22). The p110δ isoform of PI3K was also determined to regulate the suppressive function of CD4⁺CD25⁺Foxp3⁺ regulatory CD4⁺ T cells and their production of IL-10 (23).

Whereas PI3K p110δ has been determined to play a critical role in CD4⁺ T cells, less is known about the role of this enzyme in CD8⁺ T cells, particularly in the context of in vivo antimicrobial responses. IC87114 can inhibit CD8⁺ T cell proliferation and cytokine production in vitro (15). CD62L shedding and transcriptional repression could also be regulated by p110δ in CD8⁺ T cells through MAPKs and mammalian target of rapamycin (mTOR), respectively, at least in vitro (24). Importantly, in vitro–activated CTLs pretreated with IC87114 preferentially traffic to lymphoid tissue when injected into naive mice, as a consequence of inhibiting Akt-dependent expression of trafficking and differentiation molecules (25). In the same study, p110δ chemical inhibition, through reduced Akt activation, could also inhibit in vitro IFN-γ production in CTLs. However, the same group determined that Akt was dispensable for T cell metabolism, which was more dependent on mTORC1 activity that was not regulated by PI3K and Akt (26). Additionally, p110δ-deficient mice were found to develop larger tumors when challenged with MC38 colon adenocarcinoma cells, potentially as a consequence of impaired activation and cytotoxicity of CD8⁺ T cells (27). In contrast, a recent study has indicated that p110δ KD mice show increased protection against a broad range of cancers as a consequence of impaired regulatory T cell (Treg) function, allowing for enhanced CD8⁺ T cell responses (28). In this study, however, the direct effect of p110δ signaling on CD8⁺ T cells was not clear. In vitro p110δ KD OT-I cells exhibited reduced target cell killing, but adoptive transfer of large numbers of these cells in vivo before tumor challenge revealed no defect (28). Thus, the above CD8⁺ T cell studies do not conclusively address the intrinsic role for p110δ in regulating CD8⁺ T cell responses.

In our present study, we show that in vivo CD8⁺ T cells intrinsically require the p110δ catalytic subunit of PI3K for mounting optimal adaptive immune responses against influenza virus and Listeria monocytogenes infection. The p110δ deficiency impairs early proliferation, protein kinase B (Akt) activation, and metabolic activity of CD8⁺ T cells. This study thus identifies an important role for PI3K p110δ in regulating CD8⁺ T cell responses to microbial pathogens and demonstrates that PI3K p110δ signaling is critical for the optimal CD8⁺ T cell response.

Specific pathogen-free 8- to 12-wk-old female C57BL/6J, B6.PL-Thy1^a/CyJ (Thy1.1⁺), C57BL/6-Tg(TcraTcrb)1100Mjb/J (OT-I mice), B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1⁺), and CD4-cre mice were purchased from The Jackson Laboratory. p110δ^−/− mice and p110δ^fl/fl mice (9, 14, 29) were a gift from Dr. Martin Turner (Babraham Institute, Cambridge, U.K.). OT-I mice were backcrossed with p110δ^−/− mice to generate p110δ^−/−OT-I mice (all on C57BL/6 background). The Drexel University and Drexel University College of Medicine Institutional Animal Care and Use Committee reviewed and approved the animal care and use protocol for this study. All mice were maintained in American Association for the Accreditation of Laboratory Animal Care–certified barrier facilities at Drexel University College of Medicine. Mice were anesthetized with 2-2-2-tribromoethanol (avertin; 250 mg/kg i.p., Acros) and infected intranasally with a sublethal dose (ten 50% tissue culture-infective dose [TCID₅₀]) of A/Puerto Rico/8/34 (PR8 H1N1) viral strain (gift of Dr. W. Gerhard, Wistar Institute, Philadelphia, PA) or with a sublethal dose (1.87 × 10³ TCID₅₀) of the OVA_257–264-expressing influenza A/WSN/33 (WSN-OVA H1N1) virus strain (gift from Dr. David Topham, University of Rochester, Rochester, NY). Some mice were i.v. infected with 10⁴ CFU OVA-expressing L. monocytogenes (Lm-OVA) (gift from Dr. Hao Shen, University of Pennsylvania, Philadelphia, PA).

Lungs were harvested at the indicated time points and RNA was extracted by TRI Reagent (Molecular Research Center). Viral loads were determined by RT-PCR using influenza virus–specific primers as previously described (30). Cells were stained as previously described (31) using mAbs against CD69 (clone H1.2F3), Thy1.1 (clone HIS51), Thy1.2 (clone 53-2.1), CD4 (clone L3T4), CD45.1 (clone A20), CD45.2 (clone 104), CD62L (clone MEL-14), CD44 (clone 1M7), CD127 (clone A7R34), KLRG1 (clone 2F1), IFN-γ (clone XMG1.2), TNF-α (clone MP6-XT22), and granzyme B (clone NGZB) (all from eBioscience, San Diego, CA), CD8α (clone 53-6.7, BD Biosciences, San Jose, CA), CD25 (clone PC61, BD Biosciences), and H-2^b MHC class I tetramers loaded with immunodominant influenza virus nuclear protein (NP) epitope NP_366–374 or OVA_257–264 peptide. Anti-CD16/32 (Fc Block, clone 2.4G2; BD Biosciences) was used in all stains. For intracellular cytokine and granule staining, GolgiPlug (BD Biosciences), Cytofix/Cytoperm buffer (eBioscience), and Perm/Wash buffer (eBioscience) were used according to the manufacturers’ instructions. Samples were analyzed using a FACSAria flow cytometer (BD Biosciences). All data were analyzed using FlowJo software (Tree Star).

For all in vitro and in vivo assays, splenic naive CD8⁺ T cells were isolated by negative selection using magnetic beads (EasySep mouse naive CD8⁺ T cell isolation kit, Stemcell Technologies, Vancouver, Canada). The purity of isolated CD44⁻CD62L⁺CD8⁺ T cells was >92% as determined by FACS analysis.

Equal numbers (10⁴) of naive Thy1.1⁺/CD45.2⁺ OT-I or p110δ^−/−OT-I cells were transferred i.v. into Thy1.2⁺ or CD45.1⁺ wild-type recipient mice. One day later, the recipient mice were infected with WSN-OVA or Lm-OVA. For CD8⁺ T cell memory studies, 0.8 × 10⁶ Thy1.1⁺ OT-I or p110δ^−/−OT-I cells were transferred i.v. into Thy1.2⁺ wild-type recipient mice, and a day later these mice were infected with WSN-OVA.

For all coculture experiments, CD45.1⁺ splenocytes at 10⁷ cells/ml were loaded with 10 μM OVA_257–264 peptide for 2 h at 37°C, irradiated with 1500 rad, and then washed twice. Purified naive OT-I or p110δ^−/−OT-I CD8⁺ T cells were cocultured with OVA_257–264-pulsed splenocytes for the indicated time. For in vitro BrdU assays, 10 μM BrdU was added at the time of coculture. Cells were stained with FITC-conjugated anti-BrdU Ab (clone BU20a, eBioscience) using Foxp3/transcription factor staining buffer as per the manufacturer’s instructions, and BrdU labeling was determined by flow cytometry. For phospho-AKT assays, BD Phosflow fix buffer I, BD Phosflow perm buffer III, and BD Phosflow perm wash (BD Biosciences) were used as per the manufacturer’s instructions. Cells were stained with Alexa Fluor 488–conjugated anti–phospho-AKT (Ser⁴⁷³) (clone M89-61) or Alexa Fluor 488–conjugated mouse IgG1k isotype control and phospho-Akt (Ser⁴⁷³) was determined by flow cytometry.

Wild-type mice were first infected with WSN-OVA for 48 h and then received 2 × 10⁶ CFSE (CellTrace, Thermo Fisher)–labeled purified naive p110δ^−/−OT-I or OT-I cells. Three days after adoptive transfer, mice were euthanized and the mediastinal LNs, lungs, and spleen were harvested to track the proliferation of CFSE-labeled OT-I and p110δ^−/−OT-I cells (as determined by the dilution in CFSE intensity by flow cytometry).

Purified naive OT-I or p110δ^−/−OT-I CD8⁺ T cells were cocultured with CD45.1⁺ OVA_257–264-pulsed splenocytes (as described above) for 20 h and sorted to >98% purity using a FACSAria flow cytometer (BD Biosciences) (activated OT-I or p110δ^−/−OT-I CD8⁺ T cells), or splenic naive OT-I or p110δ^−/−OT-I CD8⁺ T cells were isolated by negative selection using magnetic beads (Stemcell Technologies) with >92% purity (unactivated OT-I or p110δ^−/−OT-I CD8⁺ T cells). Activated or resting CD45.2⁺ OT-1 cells were subsequently plated in XF assay media (Seahorse Bioscience) containing 11 mM glucose, 2 mM l-glutamine, and 1 mM sodium pyruvate. The basal oxygen consumption rate (OCR) and extracellular acidification rates (ECAR) were assessed using an XF-24 extracellular flux analyzer (Seahorse Bioscience). Changes in OCR and ECAR were measured over time in response to the synchronous addition of 1 μM oligomycin, 1.5 μM fluoro-carbonyl cyanide phenylhydrazone, and 100 nM rotenone plus 1 μM antimycin A (XF Cell Mito Stress kit, Seahorse Bioscience) at time points specified in the figure.

The Student t test, the Mann–Whitney U test, ANOVA, and the Shapiro–Wilk W test for normality were used with the JMP statistical analysis program (SAS). A p value <0.05 was considered significant.

To investigate the role of the p110δ isoform of PI3K for T cell responses in vivo, we used CD4-cre/p110δ^fl/fl animals (14). These animals lack the p110δ in all cells that express, or have expressed, CD4. This includes peripheral CD8⁺ T cells, which express CD4 during T cell development (32). We then infected CD4-cre/p110δ^fl/fl and p110δ^fl/fl control mice with a sublethal dose of influenza type A virus strain A PR/8/34 (H1N1). At the peak of the CD8⁺ T cell response, 10 d postinfection, lungs, mediastinal LNs, and spleens were harvested and the magnitude of the influenza-specific CD8⁺ T cell responses was determined by flow cytometry. Ag-specific CD8⁺ T cells were identified by surface staining with MHC class I tetramers loaded with the NP_366–374 immunodominant peptide and by intracellular staining for IFN-γ upon ex vivo stimulation with NP_366–374 peptide. Whereas the total number of CD4⁺ T cells in the lungs was not affected by p110δ deficiency (Fig. 1A), there was a significant reduction in the total number of CD8⁺ T cells (Fig. 1B). Additionally, CD4-cre/p110δ^fl/fl mice had greatly reduced frequencies and numbers of lung NP_366–374-specific CD8⁺ T cells compared with p110δ^fl/fl control mice (Fig. 1C, 1D). There was a 2-fold reduction in the overall percentage and absolute number of pulmonary NP_366–374-specific CD8⁺ T cells in CD4-cre/p110δ^fl/fl mice when compared with p110δ^fl/fl control animals (Fig. 1D). A similar reduction was also observed using intracellular IFN-γ staining for NP_366–374-specific CD8⁺ T cell responses in CD4-cre/p110δ^fl/fl mice, relative to the p110δ^fl/fl control animals (Fig. 1E). Additionally, the frequency of IFN-γ⁺ NP_366–374-specific CD8⁺ T cells was not significantly different between p110δ^fl/fl and CD4-cre/p110δ^fl/fl mice (Fig. 1F), indicating that p110δ deficiency in CD8⁺ T cells does not affect differentiation into IFN-γ–producing CD8⁺ T cells during influenza virus infection. However, the total numbers of spleen and mediastinal LNs NP_366–374-specific CD8⁺ T cells in CD4-cre/p110δ^fl/fl mice were not different from those of p110δ^fl/fl control animals (data not shown). Impaired viral clearance was also associated with the defective pulmonary CD8⁺ T cell response of CD4-cre/p110δ^fl/fl mice (Fig. 1G), indicating a requirement for p110δ in T cells for mediating protective responses against influenza virus.

In addition to CD8⁺ T cells, the function or development of various CD4-expressing cell types, such as CD4⁺ T cells, Tregs, and iNKT cells, can also be affected in CD4-cre/p110δ^fl/fl mice. The reduced CD8⁺ T cell response to influenza virus could thus potentially be due to an indirect effect of p110δ on CD8⁺ T cells because of defects in other immune cell types. Additionally, recent studies have shown that p110δ KD mice show increased resistance to tumor challenges, where p110δ deficiency impaired Treg functions, but had minimal effects on CD8⁺ T cell responses (28). Therefore, to determine whether the absence of p110δ conferred an intrinsic defect in the CD8⁺ T cell ability to respond to microbial pathogens, we generated p110δ^−/− Thy1.1⁺ TCR-transgenic OVA_257–264-specific CD8⁺ T cells (hereafter referred to as p110δ^−/−OT-I) by breeding Thy1.1⁺ OT-I mice with p110δ^−/− mice. Uninfected p110δ^−/−OT-I and OT-I mice showed similar frequencies of CD44⁻CD62L⁺ naive CD8⁺ T cells and activation in spleens (Supplemental Fig. 1A, 1B). We adoptively transferred 10⁴ naive Thy1.1⁺ p110δ^−/−OT-I or control Thy1.1⁺ OT-I CD8⁺ T cells into congenically mismatched Thy1.2⁺ wild-type mice. These animals were then infected with a modified H1N1 influenza virus WSN-OVA that expresses the cognate peptide of OT-I. At 10 d postinfection, the frequencies and absolute numbers of donor Thy1.1⁺ CD8⁺ T cells were measured in lungs, mediastinal LNs, and spleens. Specificity of donor Thy1.1 cells was confirmed with OVA_257–264 peptide–loaded MHC class I tetramers. The frequency and absolute number of donor Thy1.1⁺ CD8⁺ T cells were greatly reduced in the lungs of hosts transferred with p110δ^−/−OT-I cells, compared with hosts transferred with wild-type OT-I cells (Fig. 2A, 2B). There was a 3-fold reduction in the frequency and a 4-fold reduction in the absolute number of pulmonary donor Thy1.1⁺ CD8⁺ T cells in recipients transferred with p110δ^−/−OT-I cells when compared with those given wild-type OT-I cells (Fig. 2B). Whereas this reduction in donor Thy1.1⁺ CD8⁺ T cells was also found in the spleens (Fig. 2C), only a trend towards reduction was observed in the mediastinal LNs of hosts transferred with p110δ^−/−OT-I at day 10 postinfection (Fig. 2D). Importantly, this reduction in donor Thy1.1⁺ p110δ^−/−OT-I cells correlated with impaired viral clearance (Fig. 2E). Additionally, the kinetics of the response was not altered, as transfer of p110δ^−/−OT-I cells resulted in reduced responses during early (day 7) and later (day 14) stages of the lung CD8⁺ T cell response (Fig. 2F). Similar kinetic responses were also observed in the spleens and mediastinal LNs of recipient mice transferred with p110δ^−/−OT-I cells (Fig. 2G, 2H). Importantly, at day 7 of influenza WSN-OVA infection, during the peak of the donor OT-I CD8⁺ T cell response in the mediastinal LNs, there was a greatly reduced number of donor Thy1.1⁺ OT-I cells in the mediastinal LNs of mice transferred with p110δ^−/−OT-I cells (Fig. 2H). This indicates that expansion rather than trafficking of CD8⁺ T cells is affected by p110δ deficiency. These kinetic studies also suggest that rapid expansion of effectors was the most affected phase of the CD8⁺ T cell response.

To determine whether the requirement for p110δ in CD8⁺ T cell responses is restricted to influenza virus infections, we examined the effect of p110δ deficiency on CD8⁺ T cell responses to an intracellular bacterial infection. For this, we infected mice with a recombinant strain of the intracellular bacterium L. monocytogenes, which expresses the OVA protein (Lm-OVA). We performed similar naive OT-I adoptive transfer experiments as described above and 1 d after adoptive transfer, recipient mice were i.v. infected with 10⁴ CFU Lm-OVA. On day 7 postinfection, splenocytes and mesenteric LNs were harvested and the antibacterial CD8⁺ T cell response was evaluated by flow cytometry. As with influenza virus infection, hosts that received p110δ^−/−OT-I cells had a significantly reduced frequency and number of donor CD8⁺ T cells on day 7 postinfection in the spleens (Fig. 2I, 2J). There was an ∼3-fold reduction in both the frequency and absolute number of donor Thy1.1⁺ CD8⁺ T cells in the spleens of recipients given p110δ^−/−OT-I cells when compared with those given wild-type OT-I cells (Fig. 2J). Whereas a 3-fold reduction in the frequency of donor p110δ^−/−OT-I cells was observed in the mesenteric LNs of the day 7 Lm-OVA–infected recipients, there was also a similar trend toward reduction in terms of absolute number of donor p110δ^−/−OT-I cells (Fig. 2K). The above results therefore confirm that PI3K p110δ deficiency directly impairs CD8⁺ T cells and that it plays an important intrinsic role in regulating the magnitude of CD8⁺ T cell responses against viral and intracellular bacterial infections.

Previous studies on PI3K p110δ have shown that it regulates T cell cytokine production (6, 15), particularly when cells are treated with IC87114 (15). Additionally, CD8⁺ T cells from p110δ^−/− mice exhibited reduced perforin, granzyme A, and granzyme B mRNA expression when activated with anti-CD3 Ab (27). To evaluate whether p110δ deficiency impairs CD8⁺ T cell effector functions, we stimulated Ag-specific CD8⁺ T cells from day 10 influenza virus and day 7 L. monocytogenes–infected mice ex vivo with OVA_257–264 peptide and examined intracellular cytokine and cytotoxic granule production. We observed that the total number of IFN-γ–, TNF-α–, and granzyme B–producing donor OT-I CD8⁺ T cells was greatly reduced in recipients transferred with p110δ^−/−OT-I cells, when compared with hosts given wild-type OT-I cells in both influenza virus (Fig. 3A) and L. monocytogenes infections (Fig. 3B). However, this reduction is expected, given the impaired total donor OT-I CD8⁺ T cell numbers seen in mice transferred with p110δ^−/−OT-I cells. In day 10 influenza virus–infected animals, p110δ deficiency did not affect the frequencies of IFN-γ– or granzyme B–producing donor OT-I cells (Fig. 3C, 3E, 3F). However, the frequency of TNF-α production was significantly impaired in donor p110δ^−/−OT-I CD8⁺ T cells (Fig. 3D, left panels, 3D). This reduction in TNF-α production by donor p110δ^−/−OT-I CD8⁺ T cells was also seen in day 7 L. monocytogenes infection (Fig. 3D). Interestingly, p110δ deficiency did not affect the frequency of IFN-γ or granzyme B production in OT-I CD8⁺ T cells during Listeria infection (Fig. 3D). These results suggest that PI3K p110δ expression is critically required for amplifying/sustaining TNF-α production in CD8⁺ T cells, particularly during microbial infections, but it does not affect the differentiation of naive CD8⁺ T cells into IFN-γ– or granzyme B–producing effector cells.

To investigate the role of PI3K p110δ in the establishment of memory CD8⁺ T cells, we i.v. transferred Thy1.1⁺ p110δ^−/−OT-I or Thy1.1⁺ OT-I cells into Thy1.2⁺ recipient mice that were then infected with influenza virus WSN-OVA. We observed that on day 60 postinfection, memory donor p110δ^−/−OT-I cells were significantly reduced in terms of both frequency and absolute numbers in the spleens relative to memory wild-type OT-I cells (Fig. 4A, 4B). There was a 6-fold reduction in the average frequency as well as a 2.5-fold reduction in the absolute number of spleen donor Thy1.1⁺ CD8⁺ T cells in recipients transferred with p110δ^−/−OT-I cells when compared with those given wild-type OT-I cells (Fig. 4A, 4B). Additionally, we found that this reduction in memory p110δ^−/−OT-I CD8⁺ T cells was also apparent in the lungs as well as in the mediastinal LNs of recipients (Fig. 4B). CD44⁺CD62L⁻ effector and CD44⁺CD62L⁺ central memory populations did not differ between p110δ^−/−OT-I and OT-I cells in spleens, lungs, and mediastinal LNs after influenza virus infection (Fig. 4C, Supplemental Fig. 1C). Because the reduced memory CTLs could result from effects of p110δ signaling on memory differentiation or maintenance but also could simply reflect the reduced primary response, we calculated the ratio of the day 10 primary CTL response over the day 60 memory CTL population in p110δ^−/−OT-I and OT-I cells recipients. Indeed, we found that the ratio of day 10 effector versus day 60 memory p110δ^−/−OT-I and OT-I cells does not differ (Fig. 4D). This indicates that memory CTLs are not directly affected by p110δ signaling but is reduced as a consequence of p110δ decreasing the primary response. To determine whether the differentiation of naive CD44⁻CD62L⁺ CD8⁺ T cells into CD44⁺CD62L⁻ effectors or the generation of short-lived effector cells (SLECs) and memory precursor effector cells (MPECs) was affected by p110δ deficiency, we analyzed the donor p110δ^−/−OT-I and OT-I cells on day 10 of WSN-OV influenza virus A infection. The effector phenotype of donor p110δ^−/−OT-I cells did not differ from OT-I cells (Fig. 4E, 4F). SLEC (KLRG1⁺CD127⁻) and MPEC (KLRG1⁻CD127⁺) differentiation of donor p110δ^−/−OT-I cells also did not differ during the primary response (Fig. 4E, 4F), indicating that the generation of memory precursors is not specifically affected. These findings therefore indicate that p110δ deficiency results in a reduced CD8⁺ T cell memory pool and this is a consequence of its effect on the primary effector response.

Sustained PI3K activation, which results in rapid turnover of PIP₃, is required for primary T cell proliferation for up to 9 h upon interaction with Ag-pulsed APCs (33). To test whether the diminished in vivo CD8⁺ T cell responses demonstrated above were a consequence of impaired proliferation of p110δ-deficient CD8⁺ T cells, purified splenic p110δ^−/−OT-I or wild-type OT-I CD8⁺ T cells were stimulated with OVA_257–264-pulsed irradiated splenocytes, and at 20 h postincubation BrdU incorporation was measured. We observed that p110δ^−/−OT-I CD8⁺ T cells displayed significantly reduced proliferation when compared with wild-type OT-I CD8⁺ T cells (Fig. 5A, 5B). A similar reduction in proliferation was seen when p110δ^−/−OT-I cells were stimulated with solid-phase anti-CD3 Ab (data not shown). Thus, p110δ plays an important role in regulating the early proliferative capacity of CD8⁺ T cells in vitro.

To examine whether PI3K p110δ is also required for optimal CD8⁺ T cell activation, we measured the expression of activation markers CD25 (IL-2Ra) and CD69 by flow cytometry. At 20 h poststimulation, p110δ^−/−OT-I CD8⁺ T cells exhibited a marked reduction in CD25 and CD69 expression when compared with wild-type OT-I CD8⁺ T cells (Fig. 5C, 5D). The overall reduction in expression of activation markers suggests that this was a global effect on T cell activation and not specific to IL-2Rα expression and signaling. The above results suggest that p110δ is also required at the initiation of CD8⁺ T cell activation and proliferation.

We next examined whether a reduced proliferative response of p110δ^−/−OT-I CD8⁺ T cells could also be found during in vivo infection. For this, we first infected recipient mice with the influenza virus WSN-OVA for 48 h and then transferred CFSE-labeled naive OT-I and p110δ^−/−OT-I CD8⁺ T cells. Mediastinal LNs were harvested at day 3 of transfer and the CFSE dilution of donor cells was analyzed. In the mediastinal LNs, p110δ^−/−OT-I cells had fewer divisions as determined by dilution in CFSE intensity when compared with OT-I cells, (Fig. 5E). This was accompanied by a concomitant increase in CFSE^high donor p110δ^−/−OT-I CD8⁺ T cells relative to control OT-I CD8⁺ T cells (Fig. 5E). These results suggest that p110δ plays an important role in regulating the early proliferative capacity of CD8⁺ T cells.

Akt signaling pathways exist in all mammalian cells and can regulate different processes such as cellular proliferation, differentiation, metabolism, and survival (34). Activated Akt can promote CD28-independent cell growth, cytokine production, and proliferation in T cells (35). Earlier studies have shown that TCR-induced activation of Akt is highly susceptible to p110δ inhibition in both human and mouse T cells (6, 7, 13, 15). Because the p110δ isoform of PI3K was found to regulate early proliferation and activation of CD8⁺ T cells (Fig. 5), we wanted to determine whether these defects were potentially a consequence of impaired signaling through the Akt pathway. Accordingly, we stimulated p110δ^−/−OT-I and wild-type OT-I CD8⁺ T cells with OVA_257–264 peptide–pulsed irradiated splenocytes for 5 h. We then determined the level of activation of Akt by detecting Akt phosphorylation at Ser⁴⁷³ by intranuclear flow cytometry. Following activation of OT-I CD8⁺ T cells for 5 h, we observed that p110δ^−/−OT-I CD8⁺ T cells had significantly reduced phosphorylated Akt Ser⁴⁷³ when compared with wild-type OT-I CD8⁺ T cells (Fig. 6). Therefore, the expression of the p110δ isoform of PI3K is critically required for activating Akt signaling during early activation of Ag-specific CD8⁺ T cells.

T cell activation with concomitant CD28-mediated costimulation requires the enhanced utilization of extracellular glucose, directed by PI3K signaling through the activation of the kinase Akt, which in turn results in increased effector cytokine production, as well as cellular proliferation (15, 36, 37). Because pan-PI3K or Akt inhibition abolishes early glycolytic flux in effector memory CD8⁺ T cells (38) and p110δ is the main transducer of PI3K signals through Akt in TCR-stimulated T lymphocytes (15), we reasoned that the p110δ isoform of PI3K may be responsible for the induction of TCR-mediated glycolysis in CD8⁺ T cells, possibly in part through Akt. Using metabolic flux analysis, we assessed mitochondrial respiration in purified wild-type OT-I and p110δ^−/−OT-I CD8⁺ T cells by examining the OCR and ECAR under basal conditions and following drug-induced mitochondrial stress (39). Ex vivo p110δ^−/−OT-I CD8⁺ T cells exhibited a decreased basal OCR relative to wild-type OT-I CD8⁺ T cells, and this decreased oxygen consumption profile remained constant following the addition of mitochondrial inhibitors (Fig. 7A). ECAR levels, which positively correlate with glycolysis, were similar in both wild-type OT-I and p110δ^−/−OT-I CD8⁺ T cells (Fig. 7B). The ECAR of wild-type OT-I CD8⁺ T cells responded as expected to the pharmacological manipulators of mitochondrial respiration (Fig. 7B). However, the p110δ^−/− CD8⁺ T cells were incapable of recovering after treatment, suggesting impaired glycolytic activity in these cells.

We next used a system involving OVA_257–264 peptide–stimulated OT-I cells to investigate whether p110δ^−/−OT-I CD8⁺ T cells differed from wild-type OT-I CD8⁺ T cells in the context of their metabolic phenotypes after TCR-mediated activation. Following 20 h of Ag-specific stimulation, p110δ^−/− OT-I CD8⁺ T cells had a significantly lower OCR (Fig. 7C) and ECAR (Fig. 7D) than did wild-type OT-I CD8⁺ T cells, indicating decreased oxidative phosphorylation and aerobic glycolysis, respectively, and correlating with slower proliferation and reduced activation of p110δ^−/− OT-I effector T cells (Fig. 5). Taken together, these data demonstrate that p110δ-deficient CD8⁺ T cells possess lower levels of metabolic activity, including the impaired ability to undergo aerobic glycolysis after TCR-specific activation, something that would contribute to proliferative impairment as well as reduced effector cytokine production.

PI3K signaling is known to activate diverse signaling pathways that can in turn promote various cellular processes such as growth, proliferation, survival, differentiation, and migration of cells (2, 3). Despite the well-described effect of the p110δ isoform of PI3K on B cell development and function, such as B cell proliferation, the differentiation or survival of the marginal zone (MZ) and B1 B cell subsets, and a role in affinity maturation (6, 8, 9, 14), its role in T cells and particularly CD8⁺ T cells remains much less understood. In Tregs, the role of p110δ signaling for the development and function of these cells is well characterized (23). Previous studies have shown that blocking p110δ primarily affects cytokine production by CD4⁺ T cells and the proliferation of naive CD4⁺ T cells, although memory CD4⁺ T cell proliferation is not affected (15). Accordingly, signaling defects or blockade of p110δ results in impaired proliferation and cytokine production by CD4⁺ T cells (6, 13) as well as defective differentiation of Th1, Th2, T follicular helper, and Th17 cells (13, 14, 40).

Much less, however, is known about the role of the p110δ isoform of PI3K in CTL responses and how it could affect CTL immunity during microbial infections. Pharmacological inhibition of p110δ of in vitro human CD8⁺ T cells will reduce CD3/CD28-induced phosphorylation of Akt and Erk (15). OT-I CD8⁺ T cell cytokine production is also affected when p110δ is blocked; however, in vitro proliferation is only inhibited by ∼50%. These findings suggest that p110δ primarily affects effector function rather than expansion of CTL responses (15).

Our studies show that, in the absence of signaling mediated by p110δ, CD8⁺ T cell responses to pathogens are greatly impaired in terms of both magnitude and also possibly effector functions such as TNF-α production. We show that intrinsic p110δ signaling is required in CD8⁺ T cells, as demonstrated by our adoptive transfer experiments, where the host is p110δ sufficient and only the CD8⁺ T cells are p110δ deficient. In these experiments, the magnitude of the CTL response is greatly reduced in the absence of p110δ. In contrast to previous observations where p110δ inhibition was found to reprogram CTL trafficking by regulating expression of homing markers such as CD62L, S1P1, and CCR7 (24, 25), we did not observe any preferential trafficking of p110δ-deficient CD8⁺ T cells to lymphoid tissues when infected with microbial pathogens. One possible explanation could be that the previous study performed adoptive transfers of IC87114-pretreated and long-term in vitro–activated CTLs into naive uninfected mice (25). In contrast, our study involved transferring naive Ag-specific p110δ-deficient CD8⁺ T cells into mice that were challenged with microbial pathogens, where the inflammatory response alone could modulate CD8⁺ T cell trafficking. Additionally, recent studies have indicated that blocking p110δ in vivo inhibits Tregs, which allows for better tumor control through release of tumor-specific CD8⁺ T cell immunity from Treg suppression (28). However, this study also concluded that p110δ signaling may also directly affect in vitro cytotoxicity of CTLs, but when tested in vivo against an OVA-expressing tumor, p110δ KD OT-I cells provided equivalent protection as for wild-type OT-I cells (28). A major caveat, however, of this in vivo study is that wild-type and p110δ KD OT-I cells were adoptively transferred in large numbers (2 × 10⁷) prior to tumor inoculation and such large numbers of CTLs would most likely mask any limiting effect of p110δ deficiency on the CTL response. Indeed, dose responses with numbers of adoptively transferred cells in these tumor models may reveal the requirement of p110δ signaling in antitumor CTL responses.

Our studies investigating the primary response showed an intrinsic requirement for p110δ signaling by performing adoptive transfers of naive p110δ^−/− OT-I cells. These experiments show the impact of p110δ deficiency during Ag stimulation of a naive CD8⁺ T cell. However, p110δ signaling may also impact the precursor frequency and TCR repertoire by effects on thymic selection or peripheral tolerance. Such effects could be direct on T cells but could also be indirect by affecting Ag presentation, costimulation, and cytokine production by APCs. In the present study, we did not address whether these effects of p110δ signaling on the TCR repertoire could have additional impacts on CTL immunity. Furthermore, other intrinsic effects on CD8⁺ T cells could involve their ability to interact with APCs and respond to cytokines. Although Ag-specific and anti-CD3 proliferation is affected by p110δ deficiency (data not shown and Ref. 41), suggesting that TCR signaling is affected, we cannot exclude such additional effects.

We found that in addition to a reduced primary CTL response, p110δ-deficient CTLs also result in a decreased memory pool. However, the MPEC/SLEC populations during the primary response, the effector and central memory populations generated at day 60, and the ratio of the day 10 effector response to the day 60 memory response did not differ in p110δ-deficient CTLs. Our findings thus suggest that the reduced memory we find with p110δ-deficient CTLs is not due to an effect on memory differentiation or maintenance but is directly linked to the decrease in the size of the primary response.

Importantly, we also find that PI3K p110δ is specifically required for the production of the effector cytokine TNF-α, which plays an important role in host defense against microbial pathogens (42–44). This observation is similar to that made with macrophages, where p110δ was found to be critical for TNF-α secretion, as a consequence of facilitating fission of trans Golgi network tubules carrying TNF-α on Golgi membranes (45). This does not appear to be an effect on differentiation of naive CD8⁺ T cells into effector cells, as the frequency of IFN-γ–and granzyme B–producing cells, the MPEC/SLEC ratio, and the CD44⁺CD62L⁻ phenotypes did not differ with p110δ deficiency. We found that granzyme B production by CD8⁺ T cells was not dependent on p110δ expression during in vivo infections in terms of protein expression. This contrasts with previous observations made with in vitro–activated p110δ-deficient CTLs, which were determined to have lower cytotoxic granule expression (granzymes A and B and CD107a), at least in terms of mRNA (27, 28). These differences suggest a requirement for PI3K p110δ in effector functions of CD8⁺ T cells that may be modulated by levels of inflammation as well as mode of infection. Future studies are needed to further understand the role of p110δ on CD8⁺ T cell effector functions such as cytotoxicity and the production of TNF-α and other TNF family members.

In the absence of p110δ, we found that Ag-specific activation of CD8⁺ T cells resulted in reduced proliferation and impaired phosphorylation of Akt. This serine/threonine protein kinase is usually activated downstream of PI3K and can translocate from the cell membrane, through the cytoplasm, and into the nucleus, where it affects a large number of signaling pathways (34). Akt causes mTORC1 activation that can in turn control protein synthesis, cell growth, and metabolism (46). Akt can also inhibit glycogen synthase kinase 3, which increases cellular glycogen synthesis, and inactivates forkhead box O (FOXO) transcription factor proteins, which are involved in proliferation, apoptosis, motility, and metabolism (34). Activated T cell subsets have an increased metabolic activity as they undergo rapid proliferation, growth, and production of effector molecules such as cytokines (47). Additionally, T cells switch their glucose metabolism from oxidative phosphorylation to aerobic glycolysis, that is, glucose is metabolized to produce lactate even though oxygen is readily available (48). Thus, limiting glucose availability during T cell activation can compromise TCR-induced growth, proliferation, and expression of effector molecules such as IFN-γ (37, 49). Although Akt has been implicated to be important for controlling cell metabolism in many cell types (50), its role in CD8⁺ T cells has not been clearly identified.

Numerous experiments using PI3K inhibitors and Akt overexpression in different T cell subsets have supported that PI3K signaling through Akt could be important for protein synthesis and/or glucose uptake (35–37, 51). However, many of these studies made use of PI3K inhibitors that could have off-target effects on other kinases. A few studies did indicate that whereas Akt could direct the differentiation and expression of cytolytic effector molecules in CD8⁺ T cells, p110δ and Akt were not required for in vitro mTORC1-mediated glucose uptake or glycolysis in CTLs (25, 26). These studies showed that IL-2 maintains high levels of glucose uptake and glycolysis by a p110δ- and Akt-independent pathway of mTORC1 activation that depends on PDK1 (25, 26). These studies were performed on CTLs that were in vitro activated and then conditioned with IL-2 for up to 7 d. In contrast, we found that p110δ is required for oxidative phosphorylation and aerobic glycolysis in in vitro–activated CD8⁺ T cells. This suggests that signaling through the p110δ isoform of PI3K may be important for the glycolytic switch early after activation, which later is regulated independently of PI3K/Akt, especially in the presence of high levels of cytokines. Such an effect of p110δ signaling on metabolism could contribute to the decreased proliferation and activation of p110δ-deficient CTLs. A question that remains is whether p110δ signals via Akt to affect the glycolytic switch of TCR-activated CTLs. PI3K signaling through increased Akt activation can phosphorylate and inactivate FOXO proteins, which in turn can allow for increased c-myc expression in cells (52). Importantly, strength of Ag stimulation through the TCR determines the frequency of T cells expressing c-myc, whereas IL-2 signaling fine-tunes c-myc levels within individual T cells (53). It has been shown that c-myc could be important for the initial TCR-induced glycolytic switch in naive T cells (54). PI3K p110δ could therefore potentially regulate c-myc expression in CD8⁺ T cells through increased Akt activation. However, the question still remains whether Akt and oxidative phosphorylation and aerobic glycolysis are directly associated in our studies. As mentioned above, earlier in vitro studies using IL-2–maintained CTLs have argued against Akt being important for glucose metabolism in CTLs (25, 26). Whether this is also true for the TCR activated CTLs we examined is unknown, but it is possible that in our studies p110δ is also regulating oxidative phosphorylation and aerobic glycolysis in CD8⁺ T cells in an Akt-independent manner. One such mechanism could potentially involve the direct activation of mTORC2 by PIP₃ (55), independently of Akt, which causes the acetylation of FOXO proteins and leads to c-myc upregulation (56), thereby also regulating glucose metabolism. Future studies are needed to directly address whether and how Akt activation, oxidative phosphorylation, and aerobic glycolysis are directly or indirectly linked to p110δ signaling in TCR-activated CD8⁺ T cells.

The use of p110δ inhibitors has been pursued recently in clinical trials to treat patients with neoplastic malignancies such as refractory chronic lymphocytic leukemia, B cell non-Hodgkin’s lymphoma, or solid tumors as well as patients with inflammatory diseases such as asthma, chronic obstructive pulmonary disease, and rheumatoid arthritis (57). Additionally, p110δ blockade may also potentially be used to treat patients with activated PI3K delta syndrome (APDS), who carry an activating mutation in the PIK3CD gene and suffer from immunosuppression and recurrent respiratory infections (58). However, our findings may suggest the need to exercise caution in this approach, as p110δ inhibition may prevent newly activated CD8⁺ T cells from mediating immunity to microbial Ags, tipping the balance toward increased pathogen susceptibility. Indeed, the recent study by Pearce et al. (41) reported that inhibition of p110δ by IC87114 in mice impairs their CD8⁺ T cell responses to L. monocytogenes.

Pearce et al. (41) also reported a requirement for p110δ in CD8⁺ T cell responses to L. monocytogenes. These studies made use of p110δ^D910A mice (in which a catalytically inactivate p110δ is expressed). Pearce et al. observed reduced primary, memory, and secondary Ag-specific CD8⁺ T cell responses in complete p110δ^D910A mice, but observed enhanced clearance of pathogen, which they attributed to an enhanced innate immune response. The authors performed adoptive transfers of p110δ^D910A OT-I cells in Listeria infections to show a requirement for p110δ signaling in primary and secondary CTL responses. We similarly found an intrinsic requirement for p110δ in CD8⁺ T cells in L. monocytogenes infections, but also confirmed this in influenza virus infections. In agreement with Pierce et al., we found the frequency of IFN-γ–and granzyme B–producing donor CD8⁺ T cells to be unaffected by p110δ deficiency, but we report a selective effect on TNF-α production. Both Pearce et al. and we found that MPEC frequency is not affected in p110δ^−/− CD8⁺ T cells during microbial infections. Importantly, similarly to the reduced pAkt in response with anti-CD3 and anti-CD28 stimulation seen in Pearce et al., we also observed reduced phosphorylation of Akt in p110δ^−/−OT-I CD8⁺ T cells using Ag-specific–stimulated in vitro cultures. In contrast to Pierce et al., who did not report effects on pathogen control of p110δ signaling deficiency in CTLs, we present data demonstrating impaired viral clearance in p110δ^fl/flCD4cre and p110δ^−/−OT-I transferred recipient mice infected with influenza virus. Importantly, we were also able to identify a defect in early proliferation of CD8⁺ T cells with p110δ deficiency. Additionally, we have also implicated a potential role for p110δ in regulating the metabolic activity of early activated Ag-specific CD8⁺ T cells (oxidative phosphorylation and aerobic glycolysis), which may in turn regulate their proliferative capacity as well as effector functions.

Our study has demonstrated a critical role for the p110δ isoform of PI3K in the development of optimal primary and memory CD8⁺ T cell responses during acute viral and intracellular bacterial infections. We show that p110δ signaling is intrinsically required in CD8⁺ T cells, as its absence impairs their proliferation and effector cytokine production potentially as a consequence of reduced TCR-activated oxidative phosphorylation and aerobic glycolysis. These findings have thus revealed a novel and direct role for p110δ in controlling CD8⁺ T cell antimicrobial immunity.

The authors have no financial conflicts of interest.