T cells transduce T-cell receptor signal strength by generating different phosphatidylinositols

T-cell receptor (TCR) signaling strength is a dominant factor regulating T-cell differentiation, thymic development, and cytokine signaling. The molecular mechanisms by which TCR signal strength is transduced to downstream signaling networks remains ill-defined. Using computational modeling, biochemical assays, and imaging flow cytometry, we found here that TCR signal strength differentially generates phosphatidylinositol species. Weak TCR signals generated elevated phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) and reduced phosphatidylinositol (3,4,5)-trisphosphate (PIP3) levels, whereas strong TCR signals reduced PI(4,5)P2 and elevated PIP3 levels. A proteomics screen revealed that focal adhesion kinase bound PI(4,5)P2, biochemical assays disclosed that focal adhesion kinase is preferentially activated by weak TCR signals and is required for optimal Treg induction, and further biochemical experiments revealed how TCR signaling strength regulates AKT activation. Low PIP3 levels generated by weak TCR signals were sufficient to activate phosphoinositide-dependent kinase-1 to phosphorylate AKT on Thr-308 but insufficient to activate mTOR complex 2 (mTORC2), whereas elevated PIP3 levels generated by a strong TCR signal were required to activate mTORC2 to phosphorylate Ser-473 on AKT. Our results provide support for a model that links TCR signaling to mTORC2 activation via phosphoinositide 3-kinase signaling. Together, the findings in this work establish that T cells measure TCR signal strength by generating different levels of phosphatidylinositol species that engage alternate signaling networks to control cell fate decisions.

could potentially alter the balance of phosphatidylinositols that are generated during T-cell activation. One possibly is that the PI(4,5)P2/PIP3 ratio serves as a measure of TCR strength, which could differentially regulate the activation of downstream signaling networks including AKT. Herein, we provide a mechanism describing how T cells gauge TCR signal strength with phosphatidylinositol metabolism.
TCR signal strength was modeled by altering the amount of TCR-pMHC in the simulation. The resulting simulations captured that strong TCR signals decrease PTEN protein levels (5) (Fig. 1B). PDK1 activation via autophosphorylation and resulting phosphorylation of Thr-308 on AKT required lower TCR signaling levels than mTORC2 activation (Fig. 1C) and phosphorylation of Ser-473 on AKT (Fig. 1D). An unexpected feature of the model occurred at the level of phosphatidylinositol metabolism. Maintenance of PTEN during a weak TCR signal promoted the accumulation of PI(4,5)P2 and reduced PIP3 (Fig.  1, E and F). Conversely, PTEN reduction resulting from a strong TCR signal diminished PI(4,5)P2 accumulation and allowed for elevated PIP3 generation. Taken together, these simulations predicted that the regulation of PTEN abundance by TCR signal strength results in a redistribution of phosphatidylinositol species generated.
Based on the predictions from the computational model that TCR signal strength regulated the amounts of PI(4,5)P2 and PIP3 (Fig. 1, E and F), phosphatidylinositol abundance generated by T cells activated across a range of TCR signal strengths was measured biochemically (27,28). Weak TCR stimulation yielded robust generation of PI(4,5)P2 with peak intensity at 10 min (Fig. 1G) and diminished PIP3 generation (Fig. 1H). Conversely, a strong TCR signal generated reduced levels of PI(4,5)P2 (Fig. 1G) and robust PIP3 (Fig. 1H) relative to a weak TCR signal. Finally, a strong TCR signal increased PI(3,4)P2 relative to a weak TCR stimulus, which was significant at the 10-min time point (Fig. 1I).
To determine whether PTEN inhibition impacted the distribution of phosphatidylinositols generated from a weak or strong TCR signal, murine CD4 ϩ T cells were activated in the presence of a PTEN inhibitor (Fig. 1, G-I). In T cells activated with a weak TCR stimulus, PTEN inhibition markedly decreased PI(4,5)P2 (Fig. 1G) and significantly increased PIP3 to levels observed in T cells activated with a strong TCR signal (Fig. 1H). PTEN inhibition slightly increased PI(3,4)P2 in T cells stimulated with a weak TCR signal (Fig. 1I); however, the difference was not statistically significant. In T cells activated with a strong TCR stimulus, PI(4,5)P2 levels were reduced (Fig. 1G), whereas PIP3 and PI(3,4)P2 levels were increased by PTEN inhibition (Fig. 1, H and I).
To confirm the results from the PTEN inhibitor studies, PTEN was knocked down in CD4 ϩ T cells using siRNA (Fig. 1J), and levels of PI(4,5)P2 (Fig. 1K), PIP3 (Fig. 1L), and PI(3,4)P2 (Fig. 1M) were measured in T cells stimulated with either a weak or strong TCR signal. Knockdown of PTEN reduced PI(4,5)P2 levels and increased both PIP3 and PI(3,4)P2 levels in T cells activated with a weak TCR stimulus. PTEN knockdown did not impact PIP generation in T cells versus scrambled control in T cells stimulated with a strong TCR stimulus. This was expected because strong TCR signals result in the degradation of PTEN protein to promote PIP3 synthesis. Taken together, these data demonstrated that PTEN was essential for PI(4,5)P2 accumulation during a weak TCR stimulus.

Weak TCR signals generate more PI(4,5)P2 than strong TCR signals
The heightened generation of PI(4,5)P2 from a weak TCR stimulus was unexpected. Therefore, we performed a detailed dose-response time course study to better characterize the kinetics of PI(4,5)P2 generation in both murine CD4 ϩ and CD8 ϩ T cells. A flow cytometric assay was utilized to measure PI(4,5)P2 abundance using an antibody that specifically binds PI(4,5)P2 (29). T cells were activated with varying doses of plate-bound anti-CD3 antibody and constant amounts of soluble anti-CD28 antibody (1 g/ml). Following fixation, the cells were stained with antibodies that bound CD4, CD8, TCR, and PI(4,5)P2. The CD4 ϩ T-cell population was defined as being double positive for CD4 and TCR. Likewise, the CD8 population was positive for both CD8 and TCR.
Stimulation of CD4 ϩ T cells resulted in the synthesis of PI(4,5)P2 across multiple anti-CD3 doses ( Fig. 2A). In comparing PI(4,5)P2 generation as a function of anti-CD3 antibody dose, lower doses resulted in increased generation of PI(4,5)P2 across multiple time points ( Fig. 2A) as highlighted for the 2.5min time point (Fig. 2B). Time course data for PI(4,5)P2 for the 0.25 and 1 g/ml anti-CD3 antibody doses were plotted (Fig.  2C). These data further highlighted that weak TCR stimuli increased generation of PI(4,5)P2 and are consistent with the mass ELISA results (Fig. 1G). The generation of PI(4,5)P2-positive cells was plotted as a function of both anti-CD3 antibody dose and activation time (Fig. 2D), which further illustrated that PI(4,5)P2 generation peaks at early activation time points and is more sustained in CD4 ϩ T cells stimulated with weaker TCR signals. Similar to CD4 T cells, weaker TCR signals also generated elevated PI(4,5)P2 levels in CD8 ϩ T cells (Fig. S1).
We tracked the generation of PI(4,5)P2 (Fig. 2E), PIP3 (Fig.  2F), and PI(3,4)P2 ( Fig. 2G) with imaging flow cytometry to further confirm that different TCR stimuli generated different levels of PIPs. As measured by biochemical assays (Fig. 1G) and by flow cytometry (Fig. 2C), the imaging flow cytometry mea- A, a simplified model of T-cell activation focusing on AKT activation was constructed. Simulations in Matlab were performed where TCR signal strength was modulated by altering the amount of TCR-pMHC in the simulation. B-F, the abundance of PTEN (B), P-PDK1 (C), mTORC2 (C), phosphorylated AKT (D), PI(4,5)P2 (E), and PIP3 (F) were plotted as a function of TCR signal strength. G-I, mass ELISA assays were used to measure the amount of PI(4,5)P2 (G), PIP3 (H), and PI(3,4)P2 (I) generated in murine CD4 ϩ T cells isolated by negative selection stimulated using a low (0.25 g/ml) or high (1.0 g/ml) dose of plate-bound anti-CD3 antibody and a constant amount of soluble anti-CD28 antibody (1 g/ml) in the presence or absence of 10 M PTEN inhibitor (SF1670). J, CD4 ϩ T cells were nucleofected with either scrambled control (SC) or siRNA targeting PTEN (siRNA) for 48 h, Western blotting was utilized to monitor PTEN levels, and actin was utilized as a loading control. K-M, mass ELISA assays were used to measure the amount of PI(4,5)P2 (K), PIP3 (L), and PI(3,4)P2 (M) generated in murine CD4 ϩ T cells treated with the scrambled control or siRNA targeting PTEN that were activated for 10 min with either a low (0.25 g/ml) or high (1.0 g/ml) dose of plate-bound anti-CD3 antibody and a constant amount of soluble anti-CD28 antibody (1 g/ml). Each experiment was repeated three times, and error bars are Ϯ standard deviation. A two-way ANOVA statistical test was performed. ****, Ͻ0.0001; ***, Ͻ0.001; **, Ͻ0.01; *, Ͻ0.05. Symbols over data points are comparisons between the low-and high-dose groups, and symbols in the legend are between the untreated and SF1670-treated groups. surement demonstrated that a weak TCR stimulus generated elevated PI(4,5)P2 levels, whereas a strong TCR stimulus generated elevated PIP3 levels (Fig. 1H). Together, these results demonstrated that CD4 ϩ T cells generated different levels of PIPs in response to TCR signal strength.

TCR signal strength regulates the extent and duration of TCR capping
Previous reports demonstrated that generation of phosphatidylinositols occurred in close proximity to TCR clusters, and PIP3 sequestration to TCR clusters controlled the architecture of the immunological synapse to promote signaling (30). Other work illustrated that PI(4,5)P2 inhibits TCR signaling by interacting with the intracellular signaling domains of CD3, preventing LCK binding, which would dampen downstream signaling (31). Thus, the differential generation of phosphatidylinositols induced by TCR signal strength might impact proximal signaling by regulating the local membrane environment around the TCR.
Imaging flow cytometry was utilized to track the spatial distribution of the TCR relative to PI(4,5)P2, PIP3, and PI(3,4)P2 in T cells activated with a weak or strong TCR signal (32). As expected, PIP generation occurred in proximity to TCR cap structures (Fig. 3, A-C). The colocalization between the TCR, PI(4,5)P2, PIP3, and PI(3,5)P2 was analyzed in the IDEAS software package. T cells activated with a weak TCR signal had an increase in colocalization of PI(4,5)P2 with the TCR relative to T cells activated with a strong TCR signal (Fig. 3D). Conversely, a strong TCR stimulus increased the colocalization between PIP3 and the TCR relative to a weak TCR stimulus (Fig. 3D). The colocalization score between PI(3,4)P2 and TCR was lower for T cells stimulated with either a strong or a weak TCR signal (Fig. 3D). These results demonstrated that TCR caps generated in response to a weak TCR stimulus are enriched with PI(4,5)P2, whereas TCR caps generated in response to strong TCR signals are enriched with PIP3.
The imaging flow cytometry data were analyzed to measure TCR capping during activation (Fig. 3E). The IDEAS software package was utilized to calculate TCR capping using the Delta centroid function between the nuclear stain and the TCR. In resting CD4 ϩ T cells, few TCR caps were observed (Fig. 3F). Stimulation with either a weak or strong TCR signal induced TCR capping (Fig. 3F). At 30 min, a strong TCR stimulus induced more TCR caps than a weak TCR stimulus (Fig. 3F). At later time points, a strong TCR signal increased TCR capping, whereas TCR clustering was diminished in T cells stimulated with a weak TCR signal. These results demonstrated that TCR signal strength regulated TCR capping.

Proteomic profiling identifies proteins in T cells that bind different phosphatidylinositols
Our data demonstrated that TCR signal strength regulated the PI(4,5)P2/PIP3 ratio. Many proteins bind to specific phosphatidylinositols. Therefore, differential PIP generation in T cells could regulate different proteins and downstream signaling pathways. We adapted a proteomic profiling approach to identify proteins in a resting CD4 ϩ T cell that could bind PI(4,5)P2, PIP3, or PI(3,4)P2 (33) (Fig. 4A). A 3-fold enrichment cutoff was utilized to classify a protein as uniquely pulled down by a specific phosphatidylinositol bead (Fig. 4B). A pathway analysis was performed on the protein lists for each PIP pulldown sample group. As expected, pathways known to be regulated by phosphatidylinositols including PI3K/AKT signaling, MTOR, actin-based signaling and RHO signaling were enriched in the data sets (Fig. 4C). Unexpectedly, proteins involved in protein ubiquitination and metabolic pathways including glycolysis and the TCA cycle were enriched. Together, these data demonstrated that generation of different phosphatidylinositols could potentially regulate fundamental biological processes in a CD4 ϩ T cells.
The relative amount of specific proteins in the PIP pulldown experiments were plotted for proteins in specific functional )P2 (C) was measured in murine CD4 ϩ T cells that were purified by negative selection stimulated using low (0.25 g/ml), medium (0.5 g/ml), and high (1.0 g/ml) doses of plate-bound anti-CD3 antibody and a constant amount of soluble anti-CD28 antibody (1 g/ml) using imaging flow cytometry. The yellow scale bars correspond to 5 m. D, the colocalization score between the TCR and PI(4,5)P2, PIP3, and PI(4,5)P2 was calculated in the IDEAS software package from at least 1000 individual cells. E, different levels of TCR capping were observed in the imaging flow cytometry data. The yellow scale bars correspond to 5 m. F, the Delta centroid function in the IDEAS software package was utilized to calculate the level of TCR capping from the imaging flow cytometry data as a function of activation time in CD4 ϩ T cells that received a weak or strong TCR signal. A one-way ANOVA statistical test was performed to analyze the data in D. A two-way ANOVA statistical test was performed to analyze the kinetic profiles (F). For all statistical tests, p values were summarized as follows: ****, Ͻ0.0001; **, Ͻ0.01; *, Ͻ0.05. Each experiment was repeated three times, and error bars are Ϯ standard deviation.
Proteins involved with chromatin remodeling (Fig. 4E), splicing (Fig. 4F), and transcription (Fig. 4G) also bound A, a proteomics screen was utilized to identify phosphatidylinositol-binding proteins in resting murine CD4 ϩ T cells that were purified by negative selection. B, proteins that bound to beads coated with different PIPs were identified by MS. A label-free approach was utilized to quantitate the relative abundance of each protein bound to PI(4,5)P2, PIP3, or PI(3,4)P2 beads. A protein had to be 3-fold more abundant relative to the other groups to be classified as specifically binding to a particular PIP-coated bead. C, pathway analysis was performed using the lists of proteins that bound to specific PIP beads using the Ingenuity software package. p values were calculated using the right-tailed Fisher's exact test, and the default p value cutoff for significance was Ͻ0.05. D-G, the average levels of proteins in each PIP pulldown determined by MS were plotted Ϯ standard deviation from three experiments for proteins that function in signaling (D), chromatin remodeling (E), splicing (F), and transcription (G). A one-way ANOVA statistical test was performed (D-G). For all statistical tests, p values were summarized as follows: ****, Ͻ0.0001; ***, Ͻ0.001; **, Ͻ0.01.

Weak TCR signals activate FAK via elevated PI(4,5)P2 in CD4 ؉ T cells, which is essential for optimal FOXP3 induction
Previous signaling studies in T cells focused on the role of PIP3 in activating downstream signaling pathways including AKT. In T cells, the role function of PI(4,5)P2 has focused on controlling the actin cytoskeleton and dynamics of the TCR. An interesting observation from our proteomic screen is that FAK bound PI(4,5)P2 (Fig. 4D). Additionally, FAK has established roles in T-cell signaling (41,42). Previous reports demonstrated that PI(4,5)P2 binding activated FAK activity (35,43,44). Therefore, it is possible that PI(4,5)P2 generated by a weak TCR signal could activate FAK to serve as transducer to downstream signaling pathways. The average percentage of CD4 ϩ Foxp3 ϩ CD25 ϩ T cells was quantitated. E, murine CD4 ϩ T cells purified by negative selection were nucleofected with scrambled control or siRNA targeting FAK, and Western blotting was utilized to monitor FAK levels after 48 h following nucleofection. F-H, T cells were then activated with a low or high dose (0.25 and 1.0 g/ml, respectively) of plate-bound anti-CD3 antibody and soluble CD28 (1.0 g/ml; F), and Foxp3 and CD25 expression was measured by flow cytometry 48 h after activation to monitor the generation of Foxp3 ϩ CD25 ϩ (G) and Foxp3 Ϫ CD25 ϩ (H) cells. The data are from three independent experiments Ϯ standard deviation. A two-way ANOVA statistical test was performed for data in B. A one-way ANOVA statistical test was performed from data in C. The t tests were performed for data in G and H. For all statistical tests, p values were summarized as follows: ****, Ͻ0.0001; ***, Ͻ0.001; *, Ͻ0.01.

TCR strength is encoded with phosphatidylinositols
FAK autophosphorylation on Tyr-397 serves as a marker of FAK activation (45). Western blotting was used to monitor the phosphorylation of Tyr-397 on FAK in CD4 ϩ T cells that received a weak or strong TCR signal (Fig. 5A). Weak TCR signaling generated more Tyr-397 phosphorylation than a strong TCR signal (Fig. 5, A and B), corresponding to elevated PI(4,5)P2 generated from a weak TCR signal (Fig. 1G). Because PTEN was critical in maintaining PI(4,5)P2 levels (Fig. 1G), CD4 ϩ T cells were treated with a PTEN inhibitor and then activated with a weak or strong TCR signal. The prediction was that reduced PI(4,5)P2 levels via PTEN inhibition would diminish FAK activation. Indeed, PTEN inhibition reduced FAK Tyr-397 autophosphorylation in cells activated with a weak TCR signal but had little effect on Tyr-397 phosphorylation in T cells stimulated with a strong TCR signal (Fig. 5, A and B).
One functional outcome driven by a weak TCR signal is the induction of FOXP3 expression and the high affinity interleukin-2 receptor CD25 subunit (5,7). To determine whether FAK participated in FOXP3 induction, CD4 ϩ T cells were activated with a low-dose TCR stimulus (0.25 g/ml plate-bound anti-CD3 antibody and 1 g/ml soluble anti-CD28 antibody) for 72 h in the presence of multiple doses of a FAK inhibitor. T cells were stained with antibodies that recognized CD4, FOXP3, and CD25 and analyzed by flow cytometry. The generation of CD4 ϩ FOXP3 ϩ CD25 ϩ T cells was reduced with the FAK inhibitor in a dose-dependent manner (Fig. 5, C and D). These data demonstrated that FAK was necessary for the optimal induction of FOXP3 ϩ CD25 ϩ T cells.
To confirm the results from the FAK inhibitor studies, FAK was knocked down in CD4 ϩ T cells using siRNA (Fig. 5E) and activated with either a low or high dose of anti-CD3 antibody and constant amount of anti-CD28 antibody. Activation with a low-dose stimulus drove the generation of FOXP3 ϩ CD25 ϩ T cells in cells treated with the scrambled control, whereas cells treated with siRNA targeting FAK had reduced generation of FOXP3 ϩ CD25 ϩ T cells (Fig. 5, F and G). Conversely, FAK knockdown promoted the generation of FOXP3 Ϫ CD25 ϩ T cells in similar proportion to T cells activated with a high dose anti-CD3 stimulus (Fig. 5, F and H). Taken together, these data demonstrated that FAK is essential for optimal FOXP3 induction from a weak TCR signal.

Differential PDK1 and mTORC2 PIP3 thresholds regulate AKT activation in CD4 ؉ T cells
Our model predicts that only low PIP3 levels are necessary to activate PDK1 to phosphorylate Thr-308 on AKT, whereas mTORC2 activation required high PIP3 levels to phosphorylate Ser-473 on AKT (Figs. 1, A, C, and D, and 6A). The differential activation thresholds between PDK1 and mTORC2 are likely controlled by the difference in binding affinity for PIP3 (K d of 1.5 versus 141 nM, respectively) (23,26).
The relationship between TCR signal strength and PIP3 generation was first established by measuring PIP3 abundance as a function of anti-CD3 antibody dose and constant anti-CD28 antibody concentration (1 g/ml) using a mass ELISA kit at 10 min poststimulation from murine CD4 ϩ T cells. PIP3 generation yielded a sigmoidal response as a function of anti-CD3 antibody dose and plateaued at 1 g/ml of anti-CD3 antibody (Fig. 6B). One approach to alter the PIP3/PI(4,5)P2 ratio was to inhibit PTEN, which should increase PIP3 levels. In CD4 ϩ T cells pretreated with PTEN inhibitor, the amount of anti-CD3 antibody required to achieve maximal PIP3 was reduced to 0.25 g/ml (Fig. 6B). Previous work demonstrated that TCR signal strength modulated PTEN protein abundance, where weak signals maintained PTEN and strong signals reduced PTEN (5). We monitored PTEN levels as a function of TCR signal strength and confirmed that PTEN levels are inversely correlated to TCR signal strength (Fig. 6, C and D), further demonstrating that CD4 ϩ T cells modulate the balance of PI(4,5)P2/PIP3 by regulating PTEN.
T-cell activation assays were performed to determine activation thresholds for PDK1 and mTORC2. Autophosphorylation of Ser-241 is an activation marker for PDK1 (46). Phosphorylation Thr-1135 on the RICTOR subunit of mTORC2 is a repressive modification, so disappearance of pThr-1135 RIC-TOR is a marker of mTORC2 activity (47). PDK1 achieved maximal activation at the lowest anti-CD3 antibody dose tested, 0.15 g/ml (Fig. 6, C and E). Maximal phosphorylation of Thr-308 on AKT, which is phosphorylated by PDK1, was also achieved at 0.15 g/ml of anti-CD3 antibody (Fig. 6F). However, maximal mTORC2 activation required 1 g/ml of anti-CD3 antibody as measured by dephosphorylation of mTORC2 (Fig. 6, C and E) and phosphorylation of AKT Ser-473 (Fig. 6, C  and F). To characterize the relationship between PIP3 levels and activation thresholds for PDK1 and mTORC2, the levels of p-PDK1, MTORC2 p-T308-AKT, and p-S473-AKT were plotted versus PIP3 levels (Fig. 6, G and H). These data confirmed that mTORC2 has a higher PIP3 activation threshold than PDK1 to become activated in a CD4 ϩ T cell, which in turn regulates the proteoform of AKT generated.
We reasoned that if mTORC2 required more PIP3 for maximal activation, then PTEN inhibition would reduce the dose of anti-CD3 stimulatory antibody needed for mTORC2 activation (Fig. 6B). In experiments were T cells were pretreated with the PTEN inhibitor, the dose of anti-CD3 antibody required to achieve maximal RICTOR dephosphorylation was reduced to 0.25 g/ml (Fig. 6, C and E). Additionally, PTEN inhibition reduced the anti-CD3 dose required for mTORC2 to phosphorylate Ser-473 on AKT (Fig. 6, C and F). Addition of a PTEN inhibitor did not alter the threshold for phosphorylation of Thr-308 (Fig. 6, C and F), which is consistent with PDK1 having a lower PIP3 activation threshold. Together, these results supported the proposed model where low levels of PIP3 generated by a weak TCR signal were sufficient to activate PDK1 to phosphorylate Thr-308 on AKT, whereas stronger TCR signals were required to generate higher PIP3 levels to activate mTORC2 to phosphorylate AKT-S473.

Discussion
Herein, we identified that the balance between PI(4,5)P2 and PIP3 is a key determinant in grading the strength of a TCR signal (Fig. 7). In our measurements, generation of PIP3 and PI(4,5)P2 has sharp TCR signal strength thresholds that resemble a digital signaling circuit. A weak TCR signal is encoded by high PI(4,5)P2 and low PIP3, whereas a strong TCR signal is encoded by low PI(4,5)P2 and high PIP3. In cytotoxic lympho-TCR strength is encoded with phosphatidylinositols cytes, PIP5 kinases are expelled from the membrane around TCRs in the immunological synapse, which prevents PI(4,5)P2 replenishment (48). These data demonstrate that the coordination of PIPs around the TCR is a mechanism of controlling T-cell activation and effector functions. Our model predicts that the activation of proteins sensitive to PIP3 levels would also have a sharp activation threshold as a function of TCR signal strength. Indeed, we observe sharp activation thresholds for proteins that are activated by PIP3, including AKT and mTORC2. Additionally, PTEN protein levels demonstrated a sharp diminution as a function of TCR signal strength, which further demonstrates the commitment of a T cell to actively modulate the PI(4,5)P2 to PIP3 ratio to interpret TCR signal strength.
Although most T-cell activation mechanisms ascribe PI(4,5)P2 as an intermediary metabolite in the synthesis of PIP3, our work suggests that PI(4,5)P2 has an active function in encoding a weak TCR signal. T cells activated with weak TCR stimulation employ multiple mechanisms to sustain PI(4,5)P2 levels including maintenance of PTEN activity at the transcriptional and posttranscriptional levels (5). We find that maintenance of PTEN enzymatic activity is crucial for generating PI(4,5)P2 during a weak TCR signal and is involved with the activation of downstream kinases including FAK. PTEN inhibition diminishes Treg function (5,49,50). Additionally, we found that FAK inhibition also reduced Treg induction. In other studies that utilized microscopy, weak TCR stimulation resulted in a ring-like structure around TCR microclusters that contained the FAK-interacting protein paxillin (51). Possibly, the elevated PI(4,5)P2 production around the TCR that we observed could facilitate the organization of this structure. FAK inhibition could have therapeutic potential in the context of Figure 6. The balance of PI(4,5)P2 versus PIP3 is necessary for interpreting TCR signal strength and setting AKT activation thresholds. A, TCR signaling engages PI3K to generate PIP3, which in turn activates both PDK1 and mTORC2. Pdk1 phosphorylates Ser-308 on AKT, and MTORC2 phosphorylates Thr-473 on AKT, which are required for AKT activation. B, the abundance of PIP3 generated in CD4 ϩ T cells activated for 10 min with varying doses of plate-bound anti-CD3 antibody indicated in each panel and a constant amount of soluble anti-CD28 antibody (1 g/ml) in the presence or absence of 10 M PTEN inhibitor, SF1670, was determined by a mass ELISA assay from three independent experiments. C, murine CD4 ϩ T cells were activated for 10 min in triplicate with varying doses of plate-bound anti-CD3 antibody and 1 g/ml anti-CD28 antibody in the presence or absence of 10 M PTEN inhibitor (SF1670). D, the abundance of PTEN as a function of anti-CD3 antibody concentration was determined by densitometry. E and F, P-PDK1, p-RICTOR, p-AKT308, and p-AKT473 Western blots were quantitated by densitometry where all phosphorylated species were normalized to the total amount of the respective protein. G and H, the normalized abundance of p-PDK1, p-RICTOR, p-AKT308, and p-AKT473 determined by Western blotting was plotted versus the level of PIP3 generated.

TCR strength is encoded with phosphatidylinositols
tumors. In studies that utilized the VS-4718 FAK inhibitor, Tregs were depleted from the tumor microenvironment, which allows a more effective CD8 ϩ T cell anti-tumor response (52). However, it is not clear whether the FAK inhibitor was acting directly on Tregs or impacting the tumor cells. Together, our data support a model where elevated PI(4,5)P2 generated from a weak TCR signal activates downstream signaling networks to induce Treg differentiation.
One mechanism by which differential PIP generation could impact TCR signaling is at the level of TCR clustering/capping. In our imaging flow cytometry studies, we found that weak TCR signals resulted in TCR caps enriched with PI(4,5)P2, whereas strong TCR signals resulted in TCR caps enriched with PIP3, which could have implications for regulating signaling pathways and engagement of the cytoskeleton. One possibility is that signal strength regulates protein recruitment to TCR caps. Signal strength generates different amounts of PI(4,5)P2 and PIP3 around the TCR and could therefore regulate protein recruitment. Differential generation of PIPs could also directly control TCR triggering. PI(4,5)P2 can directly associate with the intracellular domains of CD3, which blocks kinases from binding and phosphorylating immunoreceptor tyrosine-based activation motifs and dampens TCR signaling (31). PI(4,5)P2 also restricts TCR translocation in the plasma membrane, which could reduce TCR capping as we observed. Future work will focus on defining mechanisms by which different phosphatidylinositols regulate TCR signaling.
The sharp thresholds observed for PI3K/AKT signaling have functional significance for CD4 ϩ T cells in the context of T-cell fate decisions. Weak TCR signals drive Treg induction, whereas strong TCR signals drive Th induction (5,11,53). We propose that there are distinct signaling networks engaged to drive Treg induction that are qualitatively different from those needed to drive Th induction, and the PI(4,5)P2 to PIP3 ratio is a key determinant in these distinct activation programs. When a T cell receives a weak TCR signal, PTEN levels are maintained promoting higher PI(4,5)P2 and lower PIP3 levels, which is suf-ficient to activate PDK1 to phosphorylate Thr-308 on AKT. However, strong TCR signals generate elevated PIP3 levels that activate both PDK1 and mTORC2, resulting in the phosphorylation of both Thr-308 and Ser-473 on AKT. These AKT proteoforms have different substrates (54 -56), and thus TCR signal strength differentially regulates downstream pathways via AKT to drive alternate T-cell fate decisions (11).
In conclusion, the work presented here provides a molecular mechanism illustrating how CD4 ϩ T cells measure TCR signal strength. In this model, T cells generate different levels of multiple phosphatidylinositols, which in turn engage different signaling pathways to drive alternate cell fate decisions. Generally, phosphatidylinositol-binding proteins display a wide range of binding affinities and specificities (57,58). Therefore, cells contain signaling networks that are responsive to qualitative and quantitative fluctuations in phosphatidylinositol metabolism, which could be manipulated by receptor signaling to drive alternate signaling programs and integrate multiple receptor signaling inputs. We propose that generation of different phosphatidylinositols is a driver of T-cell fate decisions. Because of the number of possible phosphatidylinositols that can be generated, T cells could utilize phosphatidylinositol metabolism to encode the type of TCR and cytokine stimuli received and possibly integrate receptor signaling inputs to generate coherent effector outputs.

Computational modeling of AKT activation in a CD4 ؉ T cell
A model of AKT activation in a T cell was constructed in the SimBiology application of Matlab R2017B. The model was simulated using the ode15s (stiff/NDF) solver. TCR strength was modeled by changing the initial amount of the TCR-pMHC complex in the simulation.

Murine CD4 ؉ T-cell isolation and activation assays
Spleens from C57BL/6 mice were a generous gift from the laboratory of Dr. Louise D'Cruz at the University of Pittsburgh. Stimulation of a T cell with a weak TCR signal results in maintenance of PTEN, elevated PI(4,5)P2, and lower PIP3 levels. In this mode, there is sufficient PIP3 generation to activate PDK1 to phosphorylate AKT on Thr-308. The elevated PI(4,5)P2 levels generated from a weak signal activate FAK. Stimulation with a strong TCR signal reduces PTEN levels, which allows for higher PIP3 levels and reduced PI(4,5)P2. Higher levels of PIP3 activate both PDK1 and mTORC2, which results in phosphorylation of AKT on both Thr-308 and Ser-473. Diminished PI(4,5)P2 results in weak FAK activation. The AKT proteoforms generated by a weak versus high TCR signal have different substrate specificities and activate divergent downstream signaling pathways to program different T-cell fate decisions.

TCR strength is encoded with phosphatidylinositols
The mice were housed at the University of Pittsburgh in a pathogen-free facility and handled under Institutional Animal Care and Use Committee-approved guidelines. CD4 ϩ T cells were isolated from C57BL/6 spleens using a CD4 ϩ -negative selection kit (Miltenyi Biotech), and CD25 ϩ T cells were removed using CD25 microbeads. Following isolation, T cells were incubated for 1 h at 37°C. T cell activation assays were performed with various doses plate-bound anti-CD3 mAb (clone 17A2 BioLegend) noted specified throughout the manuscript in the presence of 1 g/ml soluble anti-CD28 mAb (Clone 37.51 BioLegend). For experiments using inhibitors, isolated CD4 ϩ T cells were incubated with 10 M PTEN inhibitor (SF1670) for 1 h prior to activation.

Mass ELISA assay to measure phosphatidylinositol abundance during T-cell activation
Following activation, 10 million CD4 ϩ T cell pellets were washed with 1 ml of ice cold 0.5 M TCA. Neutral lipids were extracted by adding 750 l of MeOH:CHCl 3 :12 N HCL (80:40: 1), vortexing for 30 min, and centrifuging for 10 min at 3000 rpm. The supernatant was transferred to a new 2-ml tube to which 250 l of CHCl 3 and 450 l of 0.1 N HCl were added. The sample was then centrifuged to separate the aqueous and organic phases. The organic phase was collected and dried under a stream of nitrogen gas. The sample was reconstituted in PBS. Mass ELISA kits from Echelon Biosciences to measure PI(4,5)P2, PIP3, and PI(3,4)P2 following the manufacturer's protocol. The mass ELISA results were measured at 450 nM on a Molecular Devices SpectraMax i3 plate reader. For each sample, three biological samples were measured, and two technical replicates were performed per sample. The standard curve was fit assuming a sigmoidal dose response with variable slope, and the level of phosphatidylinositol in each sample was extrapolated in the GraphPad Prism 8 software package.

siRNA knockdown
A murine PTEN siRNA kit (Origene) was used to knock down PTEN expression. A murine FAK kit (Origene) was used to knock down FAK expression. The siRNAs were introduced into isolated CD4 ϩ T cells using a standard protocol (Lonzo Nucleofector kit for mouse T cells). Western blotting analysis for either PTEN, FAK, or PTEN was performed after 48 h of incubation.

Analysis of cells by flow cytometry
The samples were analyzed on LSR II flow cytometer, and the data were analyzed with the Flowjo 10 software package. For each sample, five thousand CD4 ϩ or CD8 ϩ T cells were acquired.

Analysis of cells by imaging flow cytometry
Samples were analyzed on an Image Stream MarkII imaging flow cytometer. 1000 cells/sample were collected. The IDEAS software package was utilized to compensate, process, and analyze all of the imaging flow cytometry data.

Mass spectrometric analysis of PI(4,5)P2-, PIP3-, and PI(3,4)P2binding proteins
20 million CD4 ϩ T cells were lysed in a buffer containing 1% Nonidet P-40, 50 mM Tris (pH 8.0), and 150 mM NaCl containing Complete C phosphatase inhibitor mixture. Lysates were incubated with beads coated with either PI(4,5)P2, PIP3, and PI(3,4)P2 from Echelon Biosciences at 4°C for 12 h. Filter-aided sample preparation was utilized to generate tryptic peptide fragments. The samples were analyzed by reverse phase LC in tandem with MS using a Waters nanoAcquity LC system using a New Objective PicoChip nanospray column in line with a ThermoFisher LTQ Velos Orbitrap Pro mass spectrometer. Raw spectra were processed using the PEAKS 8 software package. The PEAKS 8 software package was used to identify proteins in the IP using the UNIPROT mouse protein database and a 1% false discovery rate. The quantitative module of PEAKS 8 was utilized to determine the relative abundance of proteins in each IP.

Bioinformatics
The proteins that bound to specific phosphatidylinositolcoated beads were analyzed with the Ingenuity software package (Qiagen). A standard core analysis with default settings was utilized to analyze the data set and identify pathways that were overrepresented in each phosphatidylinositol IP. The righttailed Fisher test was utilized to calculate p values, and a cutoff for significance was set to Ͻ0.05.

TCR strength is encoded with phosphatidylinositols
Signaling Technology 7074) was used with the SuperSignal West Pico Plus chemiluminescent substrate for detection on a protein simple FluorChem M system.

Statistics
For all experiments, the statistics were calculated with Prism GraphPad 8 software. Two-way ANOVA tests were calculated using Bonferroni after analysis correction.