TIPE proteins control directed migration of human T cells by directing GPCR and lipid second messenger signaling

Abstract

Tissue infiltration by circulating leukocytes via directed migration (also referred to as chemotaxis) is a common pathogenic mechanism of inflammatory diseases. G protein–coupled receptors (GPCRs) are essential for sensing chemokine gradients and directing the movement of leukocytes during immune responses. The tumor necrosis factor α-induced protein 8-like (TIPE or TNFAIP8L) family of proteins are newly described pilot proteins that control directed migration of murine leukocytes. However, how leukocytes integrate site-specific directional cues, such as chemokine gradients, and utilize GPCR and TIPE proteins to make directional decisions are not well understood. Using both gene knockdown and biochemical methods, we demonstrated here that 2 human TIPE family members, TNFAIP8 and TIPE2, were essential for directed migration of human CD4+ T cells. T cells deficient in both of these proteins completely lost their directionality. TNFAIP8 interacted with the Gαi subunit of heterotrimeric (α, β, γ) G proteins, whereas TIPE2 bound to PIP2 and PIP3 to spatiotemporally control immune cell migration. Using deletion and site-directed mutagenesis, we established that Gαi interacted with TNFAIP8 through its C-terminal amino acids, and that TIPE2 protein interacted with PIP2 and PIP3 through its positively charged amino acids on the α0 helix and at the grip-like entrance. We also discovered that TIPE protein membrane translocation (i.e. crucial for sensing chemokine gradients) was dependent on PIP2. Collectively, our work describes a new mechanistic paradigm for how human T cells integrate GPCR and phospholipid signaling pathways to control directed migration. These findings have implications for therapeutically targeting TIPE proteins in human inflammatory and autoimmune diseases.

TIPE family proteins are essential for sensing chemokine gradients and guiding human T cell migration.

1. Introduction

Tissue infiltration by circulating leukocytes via directed migration (also referred to as chemotaxis) is a common pathogenic mechanism of inflammatory diseases, as observed in multiple sclerosis,¹ atherosclerosis,² inflammatory bowel diseases, and others.^3–5 The Food and Drug Administration–approved drugs, such as natalizumab and fingolimod, which block the access of leukocytes to the tissues, have been shown to be effective therapeutics for treating multiple sclerosis.⁶ By preventing leukocytes from reaching the central nervous system, they exert a strong anti-inflammatory effect that limits the formation of focal lesions and irreversible tissue loss.⁷

G protein–coupled receptors (GPCRs) are the largest family of cell surface receptors that contain 7 transmembrane helices and are essential for the spatiotemporal control of immune cell migration in health and disease.⁸ Binding of chemokines to GPCRs activates the receptors through the conformational alteration, which in turn activates downstream heterotrimeric G proteins bound on the intracellular side of the cell membrane. Upon activation, the βγ subunit of heterotrimeric G protein dissociates from the Gα subunit, each subunit further stimulates distinct downstream signaling pathways. The current understanding is that the released Gβγ pair controls leukocyte chemotaxis through PI3K and Rac-GTPase signaling, while the Gα subunit controls adenylyl cyclase, phospholipase C activity, and Rho kinase.^9,10 It has long been implied that leukocytes utilize GPCR signaling to sense chemokine gradients and guide chemotaxis.⁹ However, the molecular mechanisms of chemoattractant sensing and linkage between chemokine sensing and downstream actin-dependent immune cell polarization, leading-edge formation, and motility remain poorly characterized.

The TIPE (tumor necrosis factor-α-induced protein 8-like or TNFAIP8L) family of proteins are newly described regulators of inflammation and cancer.^11,12 In mammals, the family consists of 4 highly homologous members: TIPE (also known as TNFAIP8), TIPE1 (TNFAIP8L1), TIPE2 (TNFAIP8L2), and TIPE3 (TNFAIP8L3). The crystal structures of all 4 proteins show a unique centrally located hydrophobic cavity formed by 6 alpha helices and constitutively occupied by phosphoinositides.^11,13,14 We and other groups have shown that TIPE proteins are able to facilitate PIP3 generation through interacting with PIP2, extracting PIP2 from lipid vesicles, and transferring it to PI3K.^11,15–17 TNFAIP8 promotes tumor metastasis and inhibits tumor cell proliferation;¹⁸ it is also a risk factor for non-Hodgkin's lymphoma, inflammatory bowel diseases, and bacterial infections.^19–21 Similarly, TIPE2 regulates both carcinogenesis and inflammation.^12,22 Unlike ubiquitously expressed TNFAIP8 and TIPE1, TIPE2 is preferentially expressed in bone marrow–derived cells.¹² In addition to interacting with phosphoinositide, TIPE2 and TNFAIP8 can also directly bind and inhibit RAC proteins.²³ TIPE3 was also discovered as a co-activator of PI3Ks promoting tumor growth and transfer protein of lipid second messengers.¹¹ Like other TIPE family proteins, TIPE3 extracts PIP2 from lipid bilayer, assumably, accommodates the acyl chains of this lipid in the hydrophobic cavity; and promotes PIP2 phosphorylation by PI3K. TIPE1, the least studied member of this family, may regulate cell death²⁴ and promote tumor progress.¹⁶

Recently, we discovered that loss of both TIPE2 and TNFAIP8, but not either protein alone, blocks directed migration of lymphocytes in mice. Murine T cells doubly deficient for TIPE2 and TNFAIP8 are defective in directed migration.²⁵ In addition, mice deficient in these proteins are resistant to myelin oligodendrocyte glycoprotein–induced experimental autoimmune encephalomyelitis, a mouse model for human multiple sclerosis, due to the impaired immune cell migration to the central nervous system.²⁵ However, whether TIPE proteins play a similar role in human immune cell migration with similar underlying molecular mechanisms is not clear. We report here that TNFAIP8 and TIPE2 work collaboratively to sense environmental cues and control migration directionality in human T cells. We posit that without them human T cells cannot integrate extracellular chemotactic cues with intracellular signaling networks, leading to a failed chemotaxis.

2. Methods

2.1. Antibodies, reagents, and cell culture

The antibodies used in this study were obtained from the following commercial sources: anti-TNFAIP8 and anti-TIPE2 antibodies were purchased from Proteintech; anti-α-tubulin, anti-Na⁺/K⁺ ATPase, anti-Histone-H3, anti-phospho-AKT (S473 and S308), anti-AKT, anti-phospho-cofilin (S3), anti-cofilin, anti-Gαi, anti-GAPDH, anti-GST and anti-HA antibodies were purchased from Cell Signaling Technology; anti-Gαi for co-immunoprecipitation (co-IP) was obtained from Santa Cruz Biotechnology; anti-FLAG and anti-β-actin antibodies and anti-FLAG M2 beads were obtained from Sigma-Aldrich; and anti-CXCR4 antibody was obtained from Invitrogen. Human CCL21 (hCCL21) and human CXCL12 (hCXCL12) were purchased from BioLegend. Protein A and G agarose beads were from Thermo Fisher Scientific. The 293 T cells were cultured in Dulbecco’s modified Eagle medium (Gibco) supplemented with 10% fetal bovine serum (FBS) (Invitrogen) and an antibiotic/antimycotic solution (Gibco) at 37 °C in a 5% CO₂ atmosphere. Human CD4+ T lymphocytes were obtained from the Human Immunology Core at the University of Pennsylvania (RRID:SCR_022380) and cultured in RPMI 1640 (Gibco) containing 10% FBS (Invitrogen), antibiotic/antimycotic solution (Gibco), 50 μM β-mercaptoethanol (Gibco), and 1% L-glutamine (Gibco) at 37 °C in a 5% CO₂ atmosphere. Small interfering RNA (siRNA) oligos specific against human TNFAIP8 or TIPE2 were purchased from OriGene Technologies Inc. Cu beads and wheat germ agglutinin (WGA) beads were purchased from PerkinElmer, and ADP-Glo kinase assay with PIP2:3PS was purchased from Promega.

2.2. Plasmids

Plasmids for the full-length murine and human TNFAIP8, TIPE1, TIPE2, and TIPE3 were constructed by polymerase chain reaction (PCR) amplification of the complementary DNA clones and inserted in-frame with an N-terminal FLAG tag into vector pRK5. Gαi1-3 and Myc-tagged CXCR4 DNA constructs were purchased from Addgene. The full-length HA-tagged Gαi1-3 was constructed by PCR amplification and inserted in-frame with an N-terminal HA-tag into vector pRK5. Truncated versions of HA-tagged Gαi2 lacking the N-terminal amino acids, designated as Gαi2_[55–355] and Gαi2_[198–355], and lacking C-terminal amino acids, designated as Gαi2_[1–198] and Gαi2_[1–55], were generated from a complementary DNA clone by PCR with appropriate primers and cloned in-frame with an N-terminal HA tag into vector pRK5. The primers used for amplifying HA-tagged Gαi2 lacking N-terminal or C-terminal amino acids were as follows: Gαi2 (sense) 5′-GAA TTC GCC ACC ATG TAC CCA TAC GAT GTT CCA GAT TAC GCT GGA TCC ATG GGC TGC ACC GTG AGC GC-3′, (antisense) 5′-AAG CTT TCA GAA GAG GCC ACA GTC CTT CA-3′; Gαi2_[55–355] (sense) 5′-GGA TCC TAC TCA GAA GAG GAG TGC CG-3′, (antisense) 5′-AAG CTT TCA GAA GAG GCC ACA GTC CTT CA-3′; Gαi2_[198–355] (sense) 5′-GGA TCC ACG GGC ATC GTG GAA ACA CA-3′, (antisense) 5′-AAG CTT TCA GAA GAG GCC ACA GTC CTT CA-3′; Gαi2_[1–198] (sense) 5′- GAA TTC GCC ACC ATG GGA TCC ATG GGC TGC ACC GTG AGC GC-3′, (antisense) 5′-GTC GAC CTA GGT CTT CAC ACG GGT CCG CA-3′; Gαi2_[1–55] (sense) 5′-GAA TTC GCC ACC ATG GGA TCC ATG GGC TGC ACC GTG AGC GC-3′, (antisense) 5′-TCT AGA CTA GCC ATC TTC ATG GAT GAT CT-3′. Plasmids expressing Lyn11-FRB-CFP, RFP-FKBP12-PJ, and FKBP-containing constructs (PJ-INPP5E and PJ-DEAD) were a gift from Dr. Robin Irvine (University of Cambridge, United Kingdom). Lyn-PTEN and Lyn-PTEN-Kras plasmids were a gift from Dr. Jin Zhang (Johns Hopkins University, Baltimore, MD, USA).

2.3. siRNA electroporation

Human CD4+ T lymphocyte were cultured in complete RPMI 1640 (Gibco) culture media at 37 °C in a 5% CO₂ atmosphere for 24 h. The cultured cells were collected, washed with phosphate-buffered saline (PBS) containing 1% bovine serum albumin (BSA), and electroporated with 100 nM siRNA using the human T cells Nucleofector kit (Lonza) and Amaxa Nucleofector II (Lonza) according to manufacturer's protocol. Briefly, 5 × 10⁶ cells were resuspended in 100 μL of human T cells Nucleofector solution containing 82 μL Nucleofector solution plus 18 μL supplements, then combined with 100 nM siRNA. The mixture of cells and siRNA mixture was incubated at room temperature for 5 min and transferred to a 4 mm cuvette (Lonza) and electroporated using program V-024. Cells were then transferred to a 12-well culture plate containing 1 mL of human T cell culture medium as described previously and allowed to recover for 24 to 48 h prior to collection for the experiments.

2.4. Quantitative real-time PCR

Total RNA from 2 × 10⁶ human CD4+ T cells was purified using the RNeasy kits (Qiagen) according to the manufacturer's protocol. Reverse transcription was performed with random hexamer oligonucleotides and the Invitrogen reverse transcriptase II kit. Real-time PCR was carried out in an Applied Biosystems 7500 system with Fast SYBR Green PCR Master Mix. Relative levels of gene expression were determined with 18S RNA as the control. QuantiTect primers for 18S, human TNFAIP8, TIPE1, TIPE2, and TIPE3 used for real-time PCR amplification were purchased from Qiagen.

2.5. Transmigration assay

Human CD4+ T cells were cultured and electroporated with specific siRNA oligos against TNFAIP8 or TIPE2, or both as described previously. The chemotaxis assay was performed using 96-well Neuro Probe ChemoTx transwell system with 3 μm pore size (Neuro Probe) according to the manufacturer's protocol. Briefly, 40 µL of 1.5 × 10⁶ CD4+ T cells/mL was placed onto the upper wells precoated with 10 μg/mL fibronectin (Sigma-Aldrich); the bottom wells were filled with 30 μL of migration buffer with or without 100 ng/mL CXCL12 or CCL21. After migrating at 37 °C for 4 h, the migrated cells were collected from the bottom wells and quantified using a hemocytometer. For migration assays with PI3K inhibitors (LY294002) (Thermo Fisher Scientific), Rac inhibitors (NSC23700) (Abcam), and Gαi/o inhibitors (pertussis toxin) (Thermo Fisher Scientific), CD4+ T cells were preincubated with 30 μM LY294002, 100 μM NSC23766, and 100 ng/mL pertussis toxin, respectively, for 30 min before the start of the transmigration assay.

2.6. Ibidi μ-slide migration assay

Ibidi μ-slide migration assays were performed according to the manufacturer's protocol. Briefly, human CD4+ T cells were resuspended in migration buffer (RPMI containing with 5% FBS and 1% HEPES) at a concentration of 3 to 5 × 10⁶/mL and loaded into collagen IV–coated Ibidi μ–slides. Slides were placed at 37 °C for 45 min to allow for adhering, 200 ng/mL of CXCL12 was added to one of the reservoirs, and cell migration was recorded every 40 s for 4 h using a Leica DMI4000 microscope with Yokagawa CSU-Z1 spinning disk confocal at ×10 magnification. Images were analyzed by Volocity software version 7.0.0 (PerkinElmer) using the automated tracking protocol. Objects <16 μm³ and static were excluded, as were tracks <50 μm. Cell velocity and vector angle between each track's starting and end points were obtained from Volocity software using these settings. Velocity was defined as a cell's centroid movement in μm/min along the total path. Cell directionality was defined as the sine of the bearing angle. A value of 1 indicates migration directly toward the chemoattractant, while a value of −1 indicates migration away from the chemoattractant. The tracks were sorted by migration length, and tracks were selected from the middle of the videos (track time was ∼60 min).

2.7. Western blots

The single knockdown and double knockdown (DKD) human CD4+ T and control cells were plated at 1 to 2 × 10⁶ cells/mL in a 12-well plate treated with 100 ng/mL CXCL12 for the indicated times. Cells were collected, washed, and lysed in RIPA lysis buffer (Thermo Fisher Scientific) in the presence of protease inhibitor cocktail and phosStop (Roche), and the protein concentration was determined using BCA protein assay kits (Pierce). Samples were separated by 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), transferred to PVDF membranes, and immunoblotted with specific antibodies. For confirmation of phosphorylation of AKT and cofilin, blots were first analyzed with anti-phospho-AKT (S473 or S308), anti-phospho-LIMK1, or anti-phospho-cofilin (S3) antibodies, washed with stripping buffer (1% SDS, 62.5 mM Tris HCl [pH 6.8], and 0.8% β-mercaptoethanol) for 30 min at 50 °C, then reanalyzed with anti-AKT or anti-cofilin antibody. GAPDH was used as the loading control.

2.8. Subcellular fractionation

To separate the membrane, cytosolic, and nuclear proteins, human CD4+ T cells were lysed using the Qproteome Cell Compartment Kit (Qiagen) according to the manufacturer's protocol. Briefly, 5 × 10⁶ human CD4+ T cells were treated with or without 100 ng/mL CXCL12 for the indicated times, collected, and washed with 1× cold PBS. Pellets were resuspended in 1 mL of Extraction Lysis Buffer and incubated for 10 min at 4 °C. After centrifugation, the supernatant containing cytosolic proteins were collected, while the pellet was resuspended in 1 mL of Extraction Buffer CE2 and incubated for 30 min at 4 °C. After centrifugation, the supernatant containing membrane proteins were collected, while the pellet was again resuspended in 7 μL of benzonase nuclease and 13 μL of distilled water and incubated for 15 min at room temperature. Next, this mixture was treated with 500 μL of Extraction Buffer CE3 and incubated for 10 min at 4 °C on an end-over-end shaker. After centrifugation, the supernatant containing nuclear proteins were collected. An equal amount protein of each lysate was resolved by SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted with specific antibodies to TNFAIP8 or TIPE2. α-tubulin was used as the loading control for the cytosolic fraction, Na⁺/K⁺ ATPase was used as the loading control for the membrane fraction, and Histone-H3 for the nuclear proteins.

2.9. Immunoprecipitation

To detect the interaction between TIPE proteins and Gα subunits, 293 T cells (3.5 to 6 × 10⁵ cells/mL) were seeded in a 6-well plate and cultured for 24 h, cells were cotransfected with FLAG-tagged TIPE constructs (1.0 μg) and HA-tagged Gα constructs (1.0 μg) using Fugene 6 (Promega). Twenty-four hours after transfection, cells were lysed in RIPA lysis buffer (Thermo Fisher Scientific) in the presence of protease inhibitor cocktail and phosStop (Roche), and protein concentration was determined using BCA protein assay kits (Pierce). Cell lysate was precipitated with 5 μL of anti-FLAG M2 beads (50% slurry) or anti-HA antibody (1:100) and 15 μL of protein G beads (50% slurry) with 10 mM NaF and 30 μM AlCl₃ (Sigma-Aldrich) for 16 h at 4 °C on an end-over-end shaker. Beads were washed 3 times with 1 mL lysis buffer, then samples were separated by 12% SDS-PAGE, transferred to PVDF membranes, and immunoblotted with specific antibodies against FLAG, HA or GST. To detect the endogenous protein-protein interaction, 5 × 10⁶ human T cells were stimulated with 100 ng/mL CXCL12 for 30 min or left unstimulated. Endogenous TNFAIP8 in human T cells lysate was immunoprecipitated with anti-TNFAIP8 antibody (1:100) or control Ig for 4 h at 4 °C on an end-over-end shaker, then treated with 15 μL of protein G beads (50% slurry) with 10 mM NaF and 30 μM AlCl₃ (Sigma-Aldrich) for 16 h at 4 °C on an end-over-end shaker. Then, beads were collected and washed 3 times with 1 mL lysis buffer. Samples were separated by 12% SDS-PAGE, transferred to PVDF membranes, and immunoblotted with specific antibodies against Gαi antibodies. β-Actin was used as the loading control.

To test if TNFAIP8 competes for Gαi2-CXCR4 interaction, 293 T cells (3.5 to 6 × 10⁵ cells/mL) were seeded in 6-well plates and incubated for 24 h, cells were then co-transfected with HA-tagged Gαi2 (1.0 μg), Myc-tagged CXCR4 (1.0 μg) and FLAG-tagged TNFAIP8 constructs (0.2 or 1.0 μg) using Fugene 6 (Promega). Twenty-four hours after transfection, cells were lysed in RIPA lysis buffer (Thermo Fisher Scientific) in the presence of protease inhibitor cocktail and phosStop (Roche); cell lysate was precipitated with anti-HA antibody (1:100) for 4 h at 4 °C on an end-over-end shaker; lysate was treated with 25 μL of protein A/G beads (50% slurry) in presence of 10 mM NaF and 30 μM AlCl₃ (Sigma-Aldrich) as described above. The beads were collected and washed, and samples were separated by 4% to 12% SDS-PAGE, transferred to a PVDF membrane, and immunoblotted with specific antibodies against Myc, FLAG, and HA. To further test if TNFAIP8-Gαi-CXCR4 interaction occurs endogenously, human T cells (5 × 10⁶ cells) were stimulated with 100 ng/mL CXCL12 for 5 min or 30 min or were left unstimulated. The endogenous Gαi in cell lysate was immunoprecipitated with anti-Gαi antibody (Santa Cruz Biotechnology) and protein A/G agarose beads. Briefly, cell lysate was incubated with 2ug anti- Gαi antibody or control IgG as control for 4 h at 4 °C on an end-over-end shaker. Then, cell lysate was treated with 25 μL of protein A/G beads (50% slurry) for 16 h at 4 °C on an end-over-end shaker. Beads were collected, washed, and analyzed as described previously but using specific antibodies against TNFAIP8, CXCR4 or Gαi proteins. β-Actin was used as the loading control.

2.10. PIP2 and PIP3 depletion by PJ/Lyn system

The 293 T cells were electroporated with the plasmids expressing TIPE2-FLAG, Lyn11-FRB-CFP, and RFP-FKBP12-PJ. Upon the addition of rapamycin (1 µM) to the culture medium for 5 min, these 2 protein domains (FKBP and FRB) heterodimerize, leading to a rapid recruitment of INPP5E to the plasma membrane (PM) and a depletion of PIP2. To deplete PIP3, 293T cells were electroporated with the construct expressing TIPE2-FLAG, Lyn-PTEN, and Lyn-PTEN-Kras as a negative control. Overexpression of Lyn-PTEN depletes PIP3 on membrane by targeting PTEN to the membrane. Colocalization was performed by manually counting the number of cells with a ring-like blue staining structure for the TIPE2 (membrane bound) vs a diffuse blue TIPE staining pattern (evenly distributed), using an Echo Revolve microscope (Discover Echo Inc.). Data were analyzed by Fisher's exact test. Three-dimensional reconstructions were performed with Imaris (version 9.6) (Oxford Instruments).

2.11. I-TASSER TIPE2 protein structure prediction and TIPE2 structure studies

Full-length TIPE2 models were constructed by threading from the PDB and then fragment re-assembly simulation following the I-TASSER protocol (http://zhanglab.ccmb.med.umich.edu/I-TASSER). The TIPE2 structural studies to determine position of hydrophobic, positive, and negative amino acid charges were performed using the program PyMOL version 2.3 (https://pymol.org).

2.12. Protein production

The coding regions for each construct (mouse TIPE2, mouse TIPE2 mutants, PLCD1-PH domain, GRP1-PH domain, and Sumotag domain [smt3]) were expressed in Escherichia coli strain BL21(DE3) and purified. The coding region for each protein was amplified by PCR, subcloned into pE-6XHisSumo vector (LifeSensors, Inc.) and overexpressed as a hexahistidine fusion in E. coli strain BL21(DE3). The TIPE2 construct was used to generate different TIPE2 mutants using the Quick Change site-directed mutagenesis kit (Stratagene). Cells were lysed with lysis buffer (25 mM Tris, 500 mM NaCl, 10% glycerol, 5 mM BME, 10 mM imidazole, 1 mg/mL lysozyme, 0.1% CHAPS, adjusted pH 7.2) containing protease inhibitor cocktail (Thermo Fisher Scientific; cOmplete, Mini Protease Inhibitor Cocktail) and clarified by centrifugation, and the supernatant was loaded onto a nickel-chelating Sepharose column (Qiagen Ni-NTA agarose beads) followed by elution with an imidazole gradient. Fractions rich in respective proteins were pooled and used in experiments. Protein bands was visualized by Coomassie Blue staining R250 (Bio-Rad). For Fat-Blot screens, proteins were expressed as Histag fusions. For surface plasmon resonance (SPR) assays, the 6XHisSumo tag fusions were cut using purified Ulp1 enzyme and native proteins were used in experiments. Protein concentrations were determined by A ₂₈₀ using calculated extinction coefficients or densitometry using standard albumin concentration on Coomassie stained gels.

2.13. Protein-Lipid overlay (fat blot) assay

Phospholipids (PIP2, PIP3, and DOPC) were solubilized in chloroform:methanol:HCL (1:1:0.1%) to reach the indicated concentration. Then 1 µL of lipids was spotted onto a nitrocellulose membrane strip and probed with anti-His tag antibody. Membranes were blocked with 5% milk in 1X Tris-buffered saline with Tween 20 (TBS-T) for 1 h followed by overnight incubation at 4 °C in TBS and 1 μM of the relevant protein. Lipid strips were washed with 1× Tris-buffered saline with Tween 20 (3 times, 5 min each), incubated with primary antibody for 2 h, again washed, and incubated with the appropriate IRDye secondary antibody for 1 h. Fat Blot signals were analyzed with an Odyssey Infrared Imager 9120 (LI-COR Biosciences).

2.14. Vesicle preparation

Phospholipids were purchased from Avanti and Cell Signals, Inc. Small unilamellar vesicles (SUVs) were prepared by bath sonication of a dispersion of lipids into aqueous 20 mM HEPES (pH 7.2) and 150 mM NaCl. SUVs contained (by molar fraction) 80% dioleoyl phosphatidylcholine and 20% of a phosphotidylinositide [PtdIns(4,5)P₂, or PtdIns(3,4,5)P₃] when indicated.

2.15. Surface plasmon resonance

SUV binding was monitored by surface plasmon resonance using a BIAcore T200 instrument. An HPA chip (BIAcore) was mounted in the instrument, and SUV surfaces were created in each flow cell by capturing the SUVs hydrophobic interactions. A DOPC flow cell was maintained to control for nonspecific binding to the carboxymethylated dextran chip. Equal amounts of SUV were immobilized on each respective flow cell as judged by increases in response units displayed on the sensorgrams. Typically, SUVs were loaded onto flow cell surfaces until 5,000 response units were achieved. The surfaces were stable and did not decay significantly throughout the titrations. Experiments were performed at 25 °C with a flow rate of 30 μL/min as the flow rate. TIPE proteins were dialyzed against 20 mM HEPES (pH 7.2) and 150 mM NaCl, filtered by centrifugation. Each surface of the biosensor was then exposed to 300-s injections of protein solution (association phase) followed by 600 s of buffer (dissociation phase) via the automatic command. Between injections, a 5-μL pulse of 5 M NaCl and 50 mM NaOH regenerated the SUV surfaces. Each protein was injected at concentrations ranging from 0.016 to 10 μM. Raw sensorgrams from each titration were aligned, and the signal due to binding the DOPC flow cell was considered as background binding levels. The data were then globally fit to a 1:1 Langmuir binding isotherm using BIAevaluation 3.0 software. The resulting dissociation constants (K_D) were obtained from best fitting curve for each experiment. Representative titrations are plotted as the steady state binding response (R_eq) over the range of floating protein concentrations.

2.16. Scintillation Proximity Assay

Scintillation proximity assays were performed using the PI3K Kinase kit (Promega) according to the manufacturer's protocol. Reactions were performed in a final volume of 500 µL. Briefly, purified 1 μM of SUMO, TNFAIP8, and TIPE2 were mixed with or without PI3K. Radioactive ATP stock (250 μM ATP and 1.7 μCi ATPγ33) was added to protein and PI3K mixtures and shaken for 1 h. After incubation, the quench solution (50 μM cold ATP alone or with 2 mg of copper beads or with 1 mg of WGA beads in 0.5% BSA in PBS solution) were added to the samples and incubated for 30 min. Samples were counted on a Beckman Coulter LS 5000 TD Scintillation Counter. The counts per minute value of experimental samples was determined by comparing with quench/bead alone control.

3. Results

3.1. TNFAIP8 and TIPE2 control human CD4+ T cell migration by directing GPCR, PI3K, and Rac signaling

Previously, we showed that murine leukocytes deficient in TIPE proteins are defective in directed migration; they are unable to migrate along chemoattractant gradients or into nervous tissue to cause encephalomyelitis in mice despite their normal activation status.^15,25 However, if and how TIPE proteins control immune cell migration in humans is not known. We hypothesized that the TIPE family proteins function as a cellular pilot, guiding human T cell migration through sensing the external chemokine gradients and defining the leading-edge location. To test this hypothesis, we first examined the expression of TIPE family genes in human CD4+ T cells and found that TNFAIP8 and TIPE2 were highly expressed in these cells but TIPE1 and TIPE3 were not (Supplementary Fig. 1). Therefore, we knocked down TNFAIP8 and TIPE2 using siRNA technology in these cells. Specifically, we electroporated human CD4+ T cell with the specific siRNA oligos against TNFAIP8 and TIPE2, either individually or in combination with scrambled siRNA oligos side by side as a control. Both TNFAIP8 and TIPE2 messenger RNA (Supplementary Fig. 2A) and protein (Supplementary Fig. 2B) were significantly and specifically reduced by the knockdown.

Furthermore, using transwell and μ-slide migration assays, we found that the human CD4+ T cells deficient in both TNFAIP8 and TIPE2 (DKD) but not in either single knockdown alone, were defective in directed migration. In the μ-slide migration assay, time-lapse video microscopy that tracked the trajectory of individual migrating cells (Fig. 1A) following treatment with chemokine CXCL12 revealed that the control T cells (electroporated with scrambled siRNA) had a directionality of 0.69 ± 0.031 (with 1.00 being the highest value of directionality); this was reduced to 0.03 ± 0.54 in DKD T cells (Fig. 1B, Supplementary Videos 1 and 2). However, knockdown of TNFAIP8 and TIPE2 only slightly affected the migration speed. The average velocity of 3.77 ± 0.86 μm/min in the control cells was reduced to 2.52 ± 0.46 μm/min in DKD human CD4+ T cells (Fig. 1C). Similarly, in transwell migration assays, DKD T cells exhibited a severe defect in transmigration; the individual knockdown of either TNFAIP8 or TIPE2 only marginally reduced T cell transmigration induced by CXCL12 (Fig. 1D) and CCL21 (Fig. 1E). These results indicate that TNFAIP8 and TIPE2 play essential but redundant roles in controlling human T cell directionality and a minor role in migration speed during chemotaxis.

Fig. 1.

Complete loss of directionality in TNFAIP8/TIPE2 DKD human CD4+ T cells. (A–C) Human CD4+ T cells were electroporated with individual or combined siRNA oligos specific for TNFAIP8 and/or TIPE2. Individual or combination knockdown samples together with controls (electroporated with scrambled siRNA) were loaded into a collagen IV–coated Ibidi μ-slide and recorded using a Leica DMI4000 microscope with Yokogawa CSU-X1 spinning disk confocal attachment. Migration tracks (A), directionality (B), and speed (C) of CD4⁺ T control cells (scrambled siRNA) (n = 41) and DKD cells (n = 41) in response to CXCL12, as determined in the μ-slide migration assay. Loss of TNFAIP8 and TIPE2 impairs human CD4+ T cell transmigration induced with CXCL12 (D) or CCL21 (E). The transmigration assays were carried out using Neuro Probe ChemoTx transwell system. Bottom wells were filled with migration buffer with or without CXCL12 or CCL21 (100 ng/mL). Cells were loaded into top wells and allowed to migrate at 37 °C for 4 h. Transmigrated cells were collected from bottom wells and quantified using a hemacytometer. Data are presented as mean ± SEM and pooled from at least 3 independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001 (Student's t test). ns = not significant.

To determine the potential involvement of heterotrimeric G proteins, PI3K, and Rac signaling, we first used the inhibitors of these signaling molecules (i.e. pertussis toxin, LY294002, and NSC27366) to block the pathways respectively, and then studied the migration of the control and TIPE DKD human CD4+ T cells. To assess possible inhibitor-caused cell death, which could affect migration, we stained cells with 7-AAD staining. None of the inhibitors induced cell death under any of our experimental conditions (Supplementary Fig. 3). As expected, pertussis toxin, the inhibitor widely used to block heterotrimeric G protein signaling, completely blocked the transmigration, whereas treatment with PI3K inhibitor (LY294002) or Rac1 inhibitor (NSC27366) partially but significantly blocked the transmigration induced by CXCL12 or CCL21 (Fig. 2A and 2B). Importantly, we discovered that pertussis toxin did not further reduce transmigration of TIPE DKD compared with the control cells (Fig. 2C and 2D), indicating that TIPE proteins may control human CD4+ T cell migration through interaction with heterotrimeric G proteins.

Fig. 2.

TNFAIP8 and TIPE2 promote chemotaxis of human CD4+ T cells by directing GPCR, PI3K, and Rac-GTPase signaling. Pertussis toxin abolishes chemotaxis of human CD4+ T cells induced by CXCL12 (A) or CCL21 (B), while PI3K and Rac1 inhibitors only partially block it. Human CD4⁺ T cells were preincubated with various inhibitors—pertussis toxin (100 ng/mL), LY294002 (50 μM), and NSC27366 (100 μM)—for 30 min prior to the transmigration assay. DKD human CD4⁺ T cells are defective in chemotaxis induced by CXCL12 (C) or CCL21 (D) to a similar level of Gαi blockage by pertussis toxin. Transfected human CD4⁺ T cells were preincubated with pertussis toxin (100 ng/mL) for 30 min before the start of the transmigration assay. Experiments were repeated independently at least 3 times and values are mean ± SEM *P < 0.05; **P < 0.01; ***P < 0.001 (Student's t test). ns = not significant.

3.2. The C-terminal domain of Gαi is important for interaction with TNFAIP8

TNFAIP8 was shown to preferentially interact with Gαi in yeast hybridization assays.²⁶ To explore the possibility that TIPE family proteins regulate human CD4+ T cell migration through directly interacting with the Gαi of heterotrimeric G protein, we first examined the interaction between TNFAIP8 and different Gα subunits using co-IP. We found that the interaction occurred only between TNFAIP8 and Gαi isoforms but not with Gαs or Gαq (Fig. 3A). We next expressed FLAG-tagged TIPE family proteins and HA-tagged Gαi2 in 293 T cells and immunoprecipitated TIPE family proteins with anti-FLAG antibody followed by detection of coprecipitated Gαi. Only TNFAIP8, but not other TIPE family proteins, interacted with HA-tagged Gαi2 (Fig. 3B). The specific interaction was confirmed using a reverse co-IP (first pulling down HA-tagged Gαi with anti-HA, then checking for TIPE proteins with anti-FLAG antibody). Again, TNFAIP8 coprecipitated with Gαi (Fig. 3C). To determine if the interaction between Gαi and TNFAIP8 occurs endogenously, we conducted a co-IP assay with anti-TNFAIP8 antibody in human CD4+ T cells treated with CXCL12 and without. We detected Gαi only in the precipitates treated with CXCL12 (Fig. 3D, lane 4), indicating that the TNFAIP8 preferentially interacts with the active form of the Gαi subunit.

Fig. 3.

Gαi interacts with TNFAIP8 through its C-terminal domain. (A) TNFAIP8 specifically binds to Gαi subunits but not Gαq and Gαs subunits. 293 T cells were cotransfected with HA-tagged Gαi isoform expression constructs and FLAG-tagged TNFAIP8 expression construct as indicated. Twenty-four hours after transfection, cell lysates were prepared and immunoprecipitated with anti-FLAG-M2 beads. Lysates and precipitates were Western blotted with anti-HA antibody. (B) TNFAIP8, but not other TIPE proteins, interacts with Gαi. The 293 T cells were cotransfected with HA-tagged Gαi2 expression construct and FLAG-tagged TNFAIP8, TIPE1, TIPE2, or TIPE3 expression construct as indicated. Twenty-four hours after transfection, cell lysates were immunoprecipitated and analyzed as described in (A). (C) Gαi2 is reversely co-immunoprecipitated with TNFAIP8. 293 T cells were cotransfected with HA-tagged Gαi2 and FLAG-tagged TNFAIP8 expressing constructs, and cell lysates were prepared and immunoprecipitated as described in panel B except using anti-HA beads. Precipitates were blotted with anti-FLAG antibody. (D) TNFAIP8 and Gαi interaction occurs endogenously. Human CD4+ T cells were stimulated with CXCL12 for 30 min or left unstimulated. Total cell lysates were immunoprecipitated with anti-TNFAIP8 antibody or control immunoglobulin, then Western blotted with anti-Gαi antibody. (E) Schematic diagrams of the Gαi2 domain structure, full-length Gαi2, and truncated TNFAIP8 mutants. (F) The C-terminal domain of Gαi mediates interaction with TNFAIP8. The 293 T cells were transiently transfected with GST-Gαi2 or N-terminal truncated Gαi2 mutants, and full-length FLAG-TNFAIP8 construct. Twenty-four hours after transfection, the cells were immunoprecipitated with anti-FLAG antibody and Western blotted to detect the region of Gαi2 that bound to TNFAIP8. All experiments were repeated at least 3 times with similar results. IB = immunoblot; IP = immunoprecipitation.

To map the region within Gαi2 (i.e. responsible for TNFAIP8 interaction), we generated full-length Gαi (amino acids 1–355) and a series of deletion constructs in frame with an N-terminal GST tag (Fig. 3E). Among the 4 truncated Gαi mutants, 2 C-terminal truncated constructs (with a deletion of amino acids 198–355 or 55–355) were unable to interact with TNFAIP8 (Supplementary Fig. 4). In contrast, Gαi mutants with the C-terminus (amino acids 198 to 355) or without the N-terminus (amino acids 1 to 55 deletion) were able to interact with TNFAIP8. Therefore, the C-terminus of the Gαi subunit is required for the interaction between TNFAIP8 and Gαi (Fig. 3F).

3.3. Chemokine stimulation promotes TNFAIP8 and TIPE2 membrane translocation and TIPE deficiency compromises downstream AKT activation induced by chemokines

To gain mechanistic insights into how TIPE proteins mediate human CD4+ T cell migration, we studied the subcellular localization of TNFAIP8 and TIPE2 in human CD4+ T cells upon chemokine stimulation. The subcellular fractionation efficacy of the cytosolic, PM, and nuclear components was validated by Western blot using antibodies against α-tubulin, Na⁺/K ⁺ ATPase, and histone H3, respectively. Although TNFAIP8 and TIPE2 proteins were mainly localized in the cytoplasm under resting conditions, upon CXCL12 treatment, increased TNFAIP8 and TIPE2 membrane translocation was observed (Fig. 4A and 4B). CXCR4 and CXCR7 are 2 known CXCL12 receptors that couple with pertussis toxin–sensitive heterotrimeric G proteins Gi/Go. Activation of Gαi leads to PIP2 hydrolysis at the PM,²⁷ indicating TNFAIP8 might make directional decision through the Gαi-dependent phospholipid dynamic change. As mentioned previously, PI3K-AKT is one of the major signal pathways downstream of G proteins. We first investigated the effect of TIPE protein deficiency on AKT signaling by knocking down TNFAIP8 or TIPE2, either singly or in combination in human CD4+ T cells, and stimulating them with CXCL12 for up to 15 min (Fig. 4C). As compared with the control (scrambled), although AKT473 and AKT308 phosphorylation was reduced in the respective single-knockdown cells, the strongest reductions were seen in DKD cells, indicating a collaborative effect of TNFAIP8 and TIPE2 in AKT signaling. In contrast, an increased phosphorylation of LIMK1 at the 30 min time point was observed, which corresponded to the enhanced phosphorylation of cofilin in the later time points, suggesting that TIPE proteins may control the leading-edge formation through regulating LIMK-mediated cofilin dynamic signaling.

Fig. 4.

CXCL12 promotes TNFAIP8 and TIPE2 membrane translocation and AKT activation, whereas deficiency in TNFAIP8 and TIPE2 leads to dysregulated AKT and cofilin signaling. (A) Increased membrane translocation of TNFAIP8 and TIPE2 upon CXCL12 stimulation. Human CD4⁺ T cells were treated with CXCL12 for the indicated times. The cytoplasmic, membrane, and nuclear fractions were then prepared and immunoblotted with anti-TNFAIP8 or anti-TIPE2 antibody. Anti-α-tubulin was used as the cytosolic loading control. Na⁺/K ⁺ ATPase was used as the membrane loading control. Histone H3 was used as the nuclear loading control. (B) Densitometric quantification of TNFAIP8 or TIPE2 membrane translocation shown in panel A. (C) The effect of TNFAIP8 and TIPE2 knockdown on CXCL12-induced signaling in human CD4+ T cells. Control cells were treated with scrambled siRNA and knockdown cells were treated with siRNA against TNFAIP8 and TIPE2. Control, single knockdown, and DKD human CD4+ T cells were treated with 100 ng/mL CXCL12 for up to 15 min and lysed with RIPA buffer and resolved by SDS-PAGE, transferred to a membrane, and analyzed by immunoblotting with antibodies to pAKT (S473 or T308), total AKT, pCofilin, total cofilin, pLIMK1, TNFAIP8, TIPE2, and GAPDH. (D) TNFAIP8 competes with CXCR4 for Gαi binding. The 293 T cells were cotransfected with HA-tagged Gαi2, Myc-tagged CXCR4, and FLAG-tagged TNFAIP8 expression constructs as indicated. Twenty-four hours after transfection, cell lysates were prepared and immunoprecipitated with anti-HA antibody and protein A/G beads. Precipitates and lysates then were resolved by SDS-PAGE, transferred to a membrane, and Western blotted with anti-FLAG, Myc, and HA antibody to detect TNFAIP8, CXCR4, and Gαi2, respectively. (E) Time-dependent interactions of endogenous Gαi with TNFAIP8 and CXCR4. Human CD4⁺ T cells were stimulated with 100 ng/mL CXCL12 for 5 or 30 min or were left unstimulated. Total cell lysates of human CD4⁺ T cells were immunoprecipitated with anti-Gαi or control immunoglobulin. Ips and cell lysates were prepared as described in panel D and blotted with anti-TNFAIP8 or CXCR4 antibodies to visualize the interactions among TNFAIP8, CXCR4, and Gαi. All the experiments were repeated 2 times with similar results. IB = immunoblot.

To examine if Gαi binding to CXCR4 is affected by TNFAIP8, we expressed HA-tagged Gαi2 (1.0 μg), Myc-tagged CXCR4 (1.0 μg), and a low and a high amount of FLAG-tagged TNFAIP8 (0.2 or 1.0 μg) in 293 T cells. Twenty-four hours later, cells were collected, and the lysates were precipitated with anti-HA (Gαi2) antibody. As shown in Fig. 4D, TNFAIP8 reduced the binding of CXCR4 with Gαi, in a dose-dependent manner, indicating that TNFAIP8 may compete with CXCR4 for binding to Gαi (Fig. 4D). Additionally, we also have new evidence that endogenous Gαi interactions with CXCR4 and TNFAIP8 peaked at different times after receptor activation. Thus, human CD4+ T cells were stimulated with 100 ng/mL CXCL12 for 5 or 30 min or left unstimulated. Cell lysates were immunoprecipitated with anti-Gαi antibody and protein agarose beads, and Western blotted for endogenous CXCR4 and TNFAIP8. As shown in Fig. 4E, CXCL12 stimulation increased the binding between CXCR4 and Gαi (lane 5) compared with the weak binding in lane 4 without stimulation (Fig. 4E). The strongest interaction between CXCR4 and Gαi was observed at 5 min after CXCL12 stimulation. By contrast, the strongest interaction between Gαi and TNFAIP8 was observed at 30 min after CXCL12 stimulation.

3.4. Specific amino acids in the α0 helix and at the entrance of the hydrophobic cavity of TIPE2 are required for TIPE2 interaction with phospholipid second messengers

On the other hand, all TIPE family proteins can directly bind to phosphatidylinositol 4,5-bisphosphate or PtdIns(4,5)P₂, also known as PIP2 or PI(4,5)P2 and phosphatidylinositol 3,4,5-bisphosphate or PtdIns(3,4,5)P₃ (PIP3 or PI(3,4, 5)P3.^{11,15–17,25} In the following experiments, we used TIPE2 as an example to characterize the specific TIPE sequences required for its binding to PIP2 and PIP3.

The crystal structure of human TIPE2 at 1.7-Å resolution has been resolved (PDB ID: 3F4M).¹⁴ Unfortunately, the structure of amino acids 1 to 30 was missing following crystallization. To understand how TIPE2 protein interacts with PIP2 and PIP3, we first rebuilt the intact TIPE2 protein structure using I-TASSER software (Fig. 5A). The murine TIPE2, which comprised of 184 amino acids (555nt), consisted of 6 antiparallel α-helices (Fig. 5A). Of these, helix α5 contained a kink, caused by Pro153, and thus broke into 2 helices, α5a and α5b. α5b, the shorter of the 2 helices, forms the base of the helical bundle. One notable feature of the TIPE2 structure was the centrally located, cylindrical cavity that measured approximately 20 Å depth, 7 Å in diameter at the bottom, and 10 Å at the entrance (Fig. 5B). Although the outer surface of TIPE2 was highly charged, the central cavity was predominantly hydrophobic (Fig. 5C). The volume of the cavity was 822 ų (≍1 nm³). Therefore, hydrophobic cofactors or substrates (e.g. PIP2, PIP3) were expected to bind inside the cavity. human and mouse TIPE2 share 94% sequence identity with conserved amino acids at the entrance pocket and inside cavity (Fig. 5D).

Fig. 5.

Identification of TIPE2 amino acids that are crucial for PIP2 and PIP3 binding. (A) Ribbon representation of the structure of mouse TIPE2 helices obtained from I-TASSER, containing the α0 helix missing from the crystal structure. (B) Representation of TIPE2 hydrophobic cavity size (≍ 1 nm³). (C) Distribution of hydrophobic and positive- and negative-charge amino acids on TIPE2 helices—positive charge (blue), negative charge (red), and hydrophobic (gold). (D) Amino acid sequence similarity between mouse and human TIPE2 (94% sequence identity with identical amino acids marked by an asterisk [*] and similar amino acids marked by a period [.]). (E, F) Positively charged amino acids in the α0 helix of TIPE2 (T2Δα) and at the entrance of the central hydrophobic cavity (T2Δent) are crucial for TIPE2 binding with PIP2 and PIP3. A total of 1 µM of TIPE2, T2Δα, T2Δent mutant, and SUMO proteins were added to immobilized lipid samples on a nitrocellulose membrane. Bound proteins were detected using anti-His primary antibody. The graphs shown are representative of at least 2 independent experiments. (G–K) Specificity of TIPE2 for phosphoinositide-containing SUVs. TIPE2, T2Δα, T2Δent, PLCD-PH, and GRP1-PH solutions were injected over various SUV sensor surfaces. Representative binding curve and K_D for the interaction between TIPE2 and liposome containing DOPC, PIP2, or PIP3 (G), comparison between binding levels of TIPE2 and PLCD1^PH to PIP2 containing liposome (H), and comparison between binding levels of TIPE2 and GRP1^PH to PIP3 containing liposome (I). (J, K) Representative binding levels for the interaction between TIPE2, T2Δα, and T2Δent with liposome containing PIP2 (J) or PIP3(K). The sensorgrams were aligned and globally fit to a 1:1 binding model using BIAevaluation software as described in Experimental Procedures. The resulting binding curves are presented as the steady state binding response (R_eq) as a function of respective protein concentration. The graphs shown are representative of at least 2 independent experiments.

To screen the possible domains for phosphoinositide binding on TIPE2, we utilized a protein lipid overlay assay or fat blot. TIPE2 and its mutants were expressed and purified from bacteria. These proteins were next applied to nitrocellulose membranes that were already spotted with DOPC, PIP2, and PIP3. The full-length TIPE2 recognized both immobilized PIP2 and PIP3 on nitrocellulose but did not interact with DOPC (Fig. 5E and 5F). The Sumo tag (smt3) alone (used as a negative control) did not display binding signals. Residues 15 K/16 K/20 K/24R labeled by red letters (Fig. 5D) and 28H/74H/75R/91R/95R/183 K labeled by blue letters (Fig. 5D) are positively charged amino acids located at the α0 helix and at the grip-like entrance of the TIPE2 cavity, respectively (Supplementary Fig. 5). To confirm if these positively charged residues of TIPE2 influence phospholipid binding, they were replaced with neutral charged glutamine (Q). Two TIPE2 mutations: K15Q/K16Q/K20Q/R24Q (T2Δα0), and H28Q/R75Q/R91Q/K183Q (T2Δent) were used to evaluate TIPE2-phosphoinositides interactions. Both mutations significantly blocked the binding of PIP2 and PIP3 to TIPE2 (Fig. 5E and 5F), suggesting that TIPE2 may first interact with the polar head of PIP2 and PIP3 through its positively charged amino acids, and then form strong hydrophobic interactions with the lipid tail of PIP2 and PIP3 in the central cavity.

Given the ability of TIPE2 to bind PIP2 and PIP3 on dot-blot, we sought to assess this interaction in a more quantitative manner using a SPR assay. SUVs composed of 20% phosphoinositide in a DOPC lipid background were immobilized on an HPA chip. Purified untagged TIPE2, T2Δα0, and T2Δent were injected over SUV surfaces followed by a wash to promote dissociation. Binding was monitored in real time by SPR assay over a range of concentrations to obtain quantitative binding constants. PLCD1-PH domain (PLCD1^PH) and GRP1-PH domain (GRP1^PH), expressed and purified in a similar way as TIPE2, were used as positive controls for PIP2 and PIP3 binding, respectively. Several negative controls including BSA, trypsin inhibitor and yeast Sumo (smt3) were used. Dissociation constants (K_D) for each TIPE2-SUV binding were calculated by globally fitting the sensorgrams to a 1:1 Langmuir binding model, and the resulting binding isotherms were presented as the steady state binding signal as a function of floating protein concentration (Fig. 5G–5K). TIPE2 displayed little affinity for vesicles that lack phosphoinositides (100% DOPC) (Fig. 5G). In contrast, TIPE2 bound with nanomolar K_D values to vesicles containing 20% PIP2 or PIP3, similar to the affinity of PLCD1^PH and GRP1^PH, the positive controls for PIP2 and PIP3 bindings, respectively (Fig. 5H and 5I). TIPE2 mutation in α0 helix and its entrance pocket caused significant reductions in its affinity for SUVs containing PIP2 or PIP3, consistent with the dot-blot results (Fig. 5J and 5K).

3.5. PIP2 guides TIPE2 protein membrane translocation which in turn promotes PIP3 generation

PIP2/PIP3 dynamic change is critical for directed migration and lamellipodial protrusion in immune cells.²⁸ We next asked whether TIPE2 protein PM translocation was PIP2 dependent. We turned to a rapamycin-inducible dimerization system of FKBP and FRB domains, which can be used to recruit INPP5E to the PM, therefore deplete PI(4,5)P2 specifically.²⁹ We overexpressed FLAG-tagged TIPE2 protein with Lyn11-FRB-CFP– and RFP-FKBP12-PJ–containing constructs INPP5E in 293 T cells. Upon treatment with 1μM rapamycin for 5 min and relocalization of PIP2 phosphatase (INPP5E) in membrane, PIP2 could be depleted.²⁹ We checked TIPE family localization using immunofluorescent microscopy before and after rapamycin treatment. As shown in Fig. 6, PIP2 depletion strongly changed TIPE family protein membrane occupancy. PIP2 depletion decreased the occupancy of TIPE2 on the PM under the experimental condition and led to an increase of TIPE2 in the cytosol (Fig. 6A). We also investigated PIP3 depletion effect on TIPE2 membrane localization using PIP3 deleting PJ/Lyn system.³⁰ However, PIP3 depletion did not affect TIPE2 membrane translocation (Fig. 6B).

Fig. 6.

TIPE protein membrane translocation is PIP2 dependent. (A) TIPE2 localization and relocation is PIP2 dependent. The 293 T cells transfected with TIPE2 construct and the PJ/Lyn system (red/green, respectively), with the addition of 1 μM rapamycin for 5 min, results in recruitment of both components to the membrane and depletion of PIP2. The TIPE2 membrane translocation was analyzed by confocal microscopy. (B) PIP3 depletion does not affect TIPE2 localization and relocation. PTEN-Kras and PTEN-Lyn constructs, which deplete only PIP3, were used in lieu of the PJ/Lyn system and relocation of TIPE2 with loss of PIP3 was investigated. The pie charts show quantitative analysis of the localization of TIPE2, TIPE2 mutants, and TNFAIP8 isoforms ± PIP2 or ± PIP3. (C) Both TNFAIP8 and TIPE2 can extract PIP2 from lipid vesicle and convert it to PIP3 by PI3K while bound to the protein. All values shown were subtracted from no kinase control samples. N = 12 technical replicates/group, representative of 2 independent experiments. For all graphs, ***P < 0.001; ****P < 0.0001 (Student's t test). The graphs shown are representative of at least 2 independent experiments. 3D = 3-dimensional; ns = not significant.

We reported previously that TNFAIP8 binds to PIP2 and PIP3 with K_D of 3.16 and 4.06 μM, respectively.²⁵ To assess if free PIP2 was able to be converted to PIP3 while bound to a TIPE protein, we performed a PI3K kinase assay where we tracked phosphorylation occurred using a scintillation proximity assay to measure radioactive ATP. His-tagged purified proteins were incubated with lipids, PI3K and ATPγ33. Radioligand signals associating with the protein fraction (including protein-bound phospholipid) were assayed with Cu beads while radioligand signals that were associated with the lipid vesicle fraction was assayed with WGA beads. PLC-PH and SUMO were used as control proteins. The experiment demonstrates that both TNFAIP8 and TIPE2 could promote PIP2 conversion to PIP3 (Fig. 6C, Cu beads), and that PIP2 conversion to PIP3 by PI3K occurs on PIP2-TIPE complexes (as TIPE proteins cannot extract PIP3). We also saw that TNFAIP8 could facilitate PIP3 reinsertion back to the lipid vesicles while TIPE2 was unable to do so (Fig. 6C, WGA beads), indicating that TNFAIP8 and TIPE2 might function differently in this aspect.

4. Discussions

Results reported here prompt us to propose a working model in which TNFAIP8 and TIPE2 function collaboratively to integrate external signal cues with intracellular PI3K and Rac-GTPase signaling network to control human T cell migration through Gαi-dependent and/or Gαi-independent mechanisms (Fig. 7). It was previously shown that TIPE proteins were cellular pilots that drive cell polarization and leading-edge formation in murine immune cells.²⁵ We demonstrated here that TNFAIP8, through interacting with Gαi subunit of heterotrimeric G proteins, was recruited to the site of receptor activation with the highest concentration of the gradients, thus spatiotemporally promoted the leading-edge formation (Fig. 7A). We also showed that other TIPE family members do not interact directly with the Gαi subunit. Interestingly, TNFAIP8 only interacted with Gαi upon chemokine stimulation, suggesting that TNFAIP8 might bind to the active form of Gαi, presumably should also regulate the Gαi activity. Gαi controls the cellular cAMP levels. Therefore, through interaction with Gαi, TNFAIP8, besides dictating the site of the leading-edge, may regulate cAMP level in the cells. However, if and how remains unknown and will be investigated in the future. For the TIPE family proteins that do not interact with G proteins, they could affect the leading-edge formation through a PIP2-dependent mechanism (i.e. by binding to PIP2 microdomain on the membrane and promoting PIP3 formation) (Fig. 7B). Therefore, TIPE proteins could relay the external environmental cues to the intracellular signaling pathways, such as PI3K and Rac signaling, to facilitate cytoskeleton rearrangement at the leading edge in Gαi-dependent and/or Gαi-independent mechanisms. In this study, we also identified for the first time the amino acids that are crucial for TIPE binding with PIP2 and PIP3 using TIPE2 as a model.

Fig. 7.

TNFAIP8 functions as a pilot protein sensing chemokine gradient and guiding chemotaxis via heterotrimeric Gαi-dependent signaling. (A) Upon ligation of chemokines with GPCRs, TNFAIP8 is translocated to the plasma membrane where the receptors are activated via interacting with Gαi subunit of heterotrimeric G proteins, causing the dynamic change in PIP2 level at the site of receptor activation and facilitating the initial PIP3 generation, which defines the leading-edge location. (B) The dynamic change in PIP2/PIP3 level further accelerates the other TIPE proteins membrane translocation and facilitates a functional leading-edge formation in a Gαi-independent but PIP2-dependent manner. Collaboratively, they define the leading-edge location and promote cell migration.

Blocking Gαi function with pertussis toxin could significantly impair human T cell–directed migration to the similar level as observed in the DKD human T cells (Fig. 2C and 2D), while knocking down just one of them only partially affected human T cell migration (Fig. 1D and 1E), suggesting that human TNFAIP8 and TIPE2 play a redundant role in human T cell migration as what they do in mice.²⁵ Interestingly, TNFAIP8 and TIPE1 are expressed in all cell types, while TIPE2 is only expressed in the hematopoietic cells, suggesting that TIPE2 might specifically intensify the migrating ability in these cells by collaborating with TNFAIP8. This should be a direction of future studies.

GPCRs relay extracellular signals to intracellular signal molecules to control many aspects of cell functions including cell migration and polarization. The extracellular signal is mainly relayed to various downstream intracellular pathways via heterotrimeric G proteins, as well as arrestins, kinases, ion channels, and various scaffolding proteins.³¹ Heterotrimeric G protein is composed of α, β, and γ subunits. Binding of chemokine to GPCRs leads to dissociation of the Gα subunit from the Gβγ pair, which in turn activates downstream signaling events. The current understanding is that leukocyte chemotactic migration is largely controlled by the Gβγ, functioning as a main signal transducer of chemotaxis by activating downstream PI3K, PLC/IP3/diacylglycerol effector pathway, and Rac-GTPase signaling,¹⁰ while Gαi is mainly controlling the release and reassociation of the Gβγ, thus regulating the dynamics of G protein activity.³² As shown here, TNFAIP8 interacted with the Gαi subunits but not with the Gβγ subunit²⁶ or Gαs/q (Fig. 3A). These results indicate that TNFAIP8 could initiate the leading-edge formation by extracting PIP2 from the plasma membrane, generating PIP3, and reinserting PIP3 back to the site of receptor activation through interacting with the Gαi subunit and PI3K (Fig. 6C), suggesting the direct involvement of the Gαi in human T cell migration. On the other hand, it is also possible that TNFAIP8 may function as a docking protein to hold the Gαi preventing its reassociation with the Gβγ when chemokine gradients are present. Interestingly, we also found that TNFAIP8 competed with CXCR4 to bind to Gαi (Fig. 4D and 4E). CXCR4 binds to inactive G proteins under nonstimulatory conditions. Chemokine stimulation facilitates the replacement of GTP with GDP on the inactive Gαi to form an active Gαi. Our data show that TNFAIP8 only binds to Gαi upon stimulation, suggesting that it might prevent the reassociation of Gαi with Gβγ, thus indirectly regulating Gβγ activity. However, we do not know if TNFAIP8 executes it positively or negatively. It will be a subject of the immediate future studies.

It is well established that the Gβγ subunit controls the major signaling pathways that mediate chemotaxis. The downstream signaling pathways include PI3K-AKT, MAPK, and Rac/CDC42 (the Rho family small GTPase family proteins). PtdIns(3,4,5)P3 generated at the leading edge by PI3K provides a binding site for AKT and stimulates Rac signaling, which eventually leads to actin cytoskeleton remodeling through cofilin. In this study, we tested cofilin activity and its upstream kinase LIMK. Knockdown of TNFAIP8 and TIPE2 inhibited AKT activation induced by CXCL12, while LIMK activity was reduced but prolonged, which resulted in an extended phosphorylation of Cofilin. Cofilin-1, a member of the actin depolymerizing factor/cofilin family, plays essential roles in setting the direction and speed of leukocyte migration.^33,34 Cofilins regulate actin dynamics by promoting actin treadmilling and thereby driving membrane protrusion and cell motility. Cofilins are inactivated by phosphorylation on residue serine 3 by LIM kinases. Our data indicate that TIPE proteins might control leading-edge formation through cofilin. However, the interplay between these signal molecules is not clear currently, which is a subject of investigation of our future studies.

The data presented here showed that the founding member of the TIPE family, TNFAIP8, can be recruited to the plasma membrane either through a Gαi-dependent and/or PIP2-dependent pathway. By interacting with the Gαi subunit of heterometric G protein, TNFAIP8 can sense the site of the receptor activation and decide where the leading edge should be formed (directly under the receptor clustering or ligand binding where the receptors are activated) through initiating PIP3 generation at the receptor activation site. This could also introduce a change in PIP2 microdomain or clustering, which further affects TIPE protein membrane occupancy, and therefore dynamically regulates chemotaxis in a refined mode. Because TNFAIP8 is the only TIPE family protein detected to interact with Gαi, it is most likely the pilot protein that initiates the directed migration (by dictating the site of the leading edge), while other TIPE proteins, TIPE1, TIPE2, and TIPE3, work collaboratively with TNFAIP8 to generate a fully functional leading edge. It is plausible that TNFAIP8 is recruited first through the Gαi-dependent mechanism to the specific site on the membrane of the receptor activation and then promotes PI3K enzyme activity to form a fully functional leading-edge through a PIP2/PIP3-dependent mechanism.^11,15,25 Results reported here may not only advance our current understanding of the TIPE biology in relation to leukocyte trafficking in health and inflammation, but also shed light on novel TIPE-based therapeutic strategies for inflammatory diseases.

Author contributions

H.S. and Y.H.C. conceived and supervised this study. H.S. wrote the article. J.Y. performed the human cell experiments and analyzed the data. A.Z. and J.R.G designed and performed the biochemical experiments. Z.E. performed and analyzed surface plasmon resonance data. C.L. helped with siRNA experiments.

Supplementary material

Supplementary materials are available at Journal of Leukocyte Biology online.

Funding

This research was funded by a grant from the National Institutes of Health (R01-AI136945-04 to YH Chen and H Sun.).