Signalling through class I PI3K plays important roles at multiple stages of T cell development. Loss of both p110δ and p110γ results in near complete ablation of thymocyte β-selection, while individual loss of p110δ and p110γ individually only causes minor perturbations of T cell development (9–11). These findings revealed an unexpected redundancy between p110δ and p110γ in developing T cells and was explained by cooperative signalling from the chemokine receptor CXCR4 via p110γ and pre-TCR signalling via p110δ, either of which is sufficient to generate PIP₃ required during thymocyte β-selection (12). Consistent with this, deleting the PIP₃ phosphatase PTEN bypasses the requirement for pre-TCR stimulation during thymic β-selection, presumably by enabling sustained CXCR4-dependent PIP₃ levels (13). Beyond thymocyte β-selection, mice lacking PTEN show impaired thymocyte negative selection and evidence of autoimmunity, suggesting a role for PI3K activity in maintaining central tolerance (14). In addition, mice expressing a kinase-dead p110δ show increased numbers of Treg within the thymus, highlighting a role for p110δ in thymic T_reg selection (15). Together these findings underpin an important role for PI3K signalling in the development of T cells.

Following thymic development, class I PI3K signalling is also crucial for the peripheral development of multiple subsets of T cells; T follicular helper cells (T_FH cells) require ICOS signalling during development, and ICOS mutant mice that do not activate downstream PI3Kδ, as well as mice lacking p110δ in T cells, lack T_FH cells (16, 17). The loss of T_FH cells is the main reason underpinning the lack of germinal centres and immunoglobulin class switching after immunisation of PI3Kδ-deficient mice (16). PI3Kδ also regulates the differentiation of other T_H subsets, including T_H1, T_H2, T_H17, and T_reg, as well as production of cytokines and granzymes in both CD4⁺ and CD8⁺ T cells (15, 18–26). Human patients with loss of function mutations in PI3Kδ have also been identified (27–30). These have a more profound defect in B cell development than observed in mice. By contrast, profound defects in T cell development have not been observed so far in patients lacking the p85α or p110δ subunits of PI3Kδ; but this has been difficult to evaluate systematically in such immune-deficient patients who also suffer from inflammatory disease.

More recently, gain-of-function mutations that lead to increased PI3Kδ activity have also shown to be the cause of a novel immunodeficiency syndrome called Activated PI3K delta Syndrome (APDS) (31–35). APDS patients show increased susceptibility to airway infections (e.g. with Streptococcus Pneumoniae), chronic virus infections (CMV and EBV) and pertinent to this review, have enlarged LNs and spleens as well as signs of autoimmunity, mainly manifested as cytopenia (34). Remarkably after a 12-week trial of the PI3Kδ inhibitor Leniolisib, the LNs and spleens of these patients reduced in size by up to 50% (36). This may reflect in part the potential of PI3Kδ inhibitors to cause redistribution of lymphocytes in addition to the inhibitory effect on lymphocyte proliferation. Several groups have generated mouse models of APDS which recapitulate many of the features of the patients, including increased susceptibility to airway infections, enlarged LNs and spleen and production of autoantibodies (37–41). Altogether, these studies shine light on the paradox that both loss-of-function and gain-of-function of PI3Kδ leads to immunodeficiency, and highlight how this pathway needs to be dynamically regulated for optimal lymphocyte development and function (42, 43). This, as we will see, is also key for the control of lymphocyte trafficking.

Currently four different PI3Kδ inhibitors are approved for the treatment of B cell malignancies (44). A detailed description of these is beyond the remit of this review, however two concepts learned from the treatment of these patients are worth noting. Chronic lymphocytic leukaemia (CLL) patients treated with PI3Kδ inhibitors such as Idelalisib initially experience a dramatic redistribution of the malignant B cells from the LNs (45). This phenomenon is referred to as lymphocytosis and is now recognised as a beneficial clinical feature of this class of drugs. Lymphocytosis is thought to be secondary to the interference with BCR-dependent integrin activation and chemokine responsiveness (46). CLL cells that are purged from their protective LN environment are more susceptible to undergo apoptotic cell death which can be accelerated with chemotherapy or drugs such as rituximab (anti-CD20) (45). Immune-mediated colitis and hepatitis are common adverse effects of PI3Kδ inhibitors, but skin inflammation is also seen in some studies (47). These are thought to be caused by the selective depletion or inactivation of Tregs, especially from tissues with high exposure to microbial antigens, such as the gut, liver and skin (44, 48). In this context, by targeting Treg, PI3Kδ inhibitors can unleash potent antitumour immune responses (49). Recent evidence suggests that PI3Kδ inhibitors can purge Treg from the tumour microenvironment and into the circulation (50). Hence the capacity of PI3Kδ inhibitors to not only affect lymphocyte function, but also to cause redistribution out of lymphoid tissues may underpin the therapeutic effects of PI3Kδ inhibitors.

T cells recirculate between LNs through the blood, probing antigen-presenting cells for their cognate antigen. To exit blood vessels, selectins, integrins, and chemokines are required to halt the T cells at the right place and resist the shear stress in the blood (1-70 dyn/cm²) (56). This is a tightly regulated process. Selectin-mediated binding of their ligands facilitates rolling along the endothelial membrane which slows down the T cells. This allows the T cells to respond to chemokines secreted from the endothelia and immobilised on glycosaminoglycans (GAGs) on the surface of the endothelial cells Reviewed in (57). As a result, integrins (such as LFA-1 and VLA-4) are activated. LFA-1-mediated binding of endothelial ICAMs (ICAM-1 and -2) leads to firm adhesion to the endothelial barrier. This allows the T cells to crawl against the flow towards chemotactic gradients until the cell will undergo transendothelial migration (TEM, also termed diapedesis) through the endothelial barrier into the underlying tissue.

Recirculation and homing of naïve T cells to secondary lymphoid organs (SLOs), including LNs, requires expression of the chemokine receptor CCR7 and CD62L, both which are downregulated following PI3K activation as discussed later. CD62L interacts with peripheral node addressins (PNAds) expressed on high endothelial venules (HEVs) which are formed by specialised endothelial cells lining post-capillary venules associated with lymph nodes (LNs). The interaction between CD62L and PNAd causes T cells to start rolling along the HEVs. After slowing down, CCR7 on the T cells binds CCL21 presented by HEVs (58, 59), which rapidly induces LFA-1 activation, leading to arrest and transendothelial migration (60, 61). The process of LN entry is highly dependent on LFA1; LFA-1-deficient mice have greatly reduced adhesion to HEVs, particularly in peripheral LNs (pLNs) and therefore elicit limited to no migration to the LNs (62, 63). Similarly, LFA-1-blocking antibodies block adhesion to HEVs and prevent repopulation of LNs (64). Together, HEVs thus function as a selective gateway to the LNs, attracting naïve and resting memory T cells, but largely blocking entry of other leukocytes such as neutrophils under steady state (65). Migration to gut-associated lymphoid tissues, spleen and inflamed lymphoid tissues are governed by other mechanisms and molecules, such as α4β7/MAdCAM-1 interactions, and this integrin seems to be regulated by different pathways than LFA-1 and VLA-4 (66). After entering LNs, the role of LFA-1 is less clear; studies of LFA-1-deficient T cells indicate that LFA-1 is required for retention of T cells in the parenchyma (67). However, other studies using LFA-1-deficient T cells (68) or dendritic cells lacking integrins altogether (69), suggest that interstitial and intranodal motility of T cells and DCs in the absence of antigen is much less dependent on integrins than is the entry into and egress out of the LNs.

Following screening of antigen within the LNs, T cells will egress through the efferent lymphatics in a process regulated by LFA-1/ICAM-1-interactions (67). Egress is guided by the lipid sphingosine-1-phosphate (S1P). S1P is found in low concentrations inside the LNs, but high concentrations in lymph and blood, creating a gradient which attracts the T cells through binding of the GPCR S1P receptor 1 (S1PR1) expressed by naïve T cells (70). Following egress through efferent lymphatics, the lymphatics connect LNs in series, but eventually merge with the thoracic duct allowing the T cells to recirculate through the blood. Interestingly, S1P-mediated egress from inflamed tissues is also partially dependent on interactions between LFA-1/ICAM-1 and VLA-4/VCAM-1. Here LFA-1/ICAM-1 are required for effective migration of memory T cells through afferent lymphatics into LNs (71–73).

Antigen-stimulated T cells downregulate L-selectin and glycosylate P-selectin glycoprotein ligand-1 (PSGL-1) resulting in functional PSGL-1 ligand which allows for binding of L-, P-, and E-selectins that are upregulated on inflamed endothelial tissues reviewed in (4). Other integrins are also involved in migration along, and adhesion to vascular endothelium as well as transendothelial migration of antigen-stimulated T cells. Of particular importance are VLA-4/VCAM-1, α4β7/MAdCAM-1, and αV integrin-mediated (74) interactions that facilitate migration to distinct inflammatory sites reviewed in (75, 76).

Within the T cell follicles of the LNs, T cells recognise their cognate antigen-MHC complex on the surface of antigen presenting cells (APCs) or target cells. This induces TCR signalling and triggers the formation of an immunological synapse (IS) at the contact area between the T cell and the APC/target cell (77). The IS is a highly specialised and dynamic cell-cell interface that allows for fine-tuning of signalling events leading to T cell activation (78, 79). Integrins and especially LFA-1 are key components in IS formation. In the immature IS, PI3K-dependent chemokine-mediated LFA-1 activation initiates the adhesive contact between T cells and APCs/target cells allowing the T cell to scan its interaction partner for cognate antigens (80). Concomitantly, LFA-1 triggers recruitment of organelles such as mitochondria to the IS thereby preparing the T cell for optimal TCR-induced activation and Ca²⁺ signalling during later activation stages (80).

In the immature IS, LFA-1 is found at the centre of the synapse with TCRs and downstream kinases clustered at the periphery (81, 82). During maturation of the IS, the synapse is reorganised into annular supramolecular clusters (SMACs) allowing for spatiotemporal clustering of receptors, adhesion molecules, and signalling effector proteins ( Figure 3A ) (83). In the mature IS, centripetal movement relocates TCR/pMHC complexes to the central SMAC (cSMAC) together with their co-stimulatory molecules, intracellular kinases, and adaptor proteins. Simultaneously, LFA-1/ICAM-1 complexes are redistributed to the integrin-enriched peripheral SMAC (pSMAC) surrounding the cSMAC (81). In the pSMAC, LFA-1 both stimulates T cell activation by increasing the accumulation of TCR/pMHC complexes in the cSMAC, and recruiting signalling molecules to the pSMAC; LFA-1 may also help segregate the phosphatase CD45 to the distal SMAC (dSMAC) (84, 85). In addition to LFA-1, the α4β1 integrin, VLA-4, is also enriched in the pSMAC and is involved in T cell activation by regulating the mobility of SLP-76, an essential adaptor protein functioning downstream the TCR (86). It is thought that VLA-4 can restrain SLP-76 in the pSMAC, so that SLP-76 both remains in closer contact with its upstream activators and avoids the cSMAC, where signalling complexes will eventually be internalised and degraded to terminate TCR signalling (86, 87). Collectively, an important function of integrins in the IS appears to be regulating the localisation of both inhibitory and stimulatory signalling molecules. The phosphatidylinositol (PIP) composition of the IS might also contribute to this spatial regulation of signalling proteins within the IS ( Figure 3B ). Early studies confirmed the accumulation of PIP₃ inside and outside the IS between APCs and T cells (88–90). However, PIP₃ seemed less concentrated at the cSMAC, and later studies using transgenic CD8⁺ T cells have found that PI3K-generated PIP₃ seems specifically accumulated in the periphery of the IS (91–93). Further, both PIP₂ and PIP₃ are cleared from the cSMAC during conjugate formation, and sustained PI3K activity is necessary for proper T cell activation possibly through the regulation of PIP₃ binding proteins (91–93). It has further been suggested that PI3K-dependent actin remodelling in the periphery of the IS can mediate synaptic force on the target cell thereby potentiating target cell killing by CD8⁺ T cells (94).

Inside-out signalling in T cells is a complex process mediated by a range of proteins that collectively result in increased LFA-1 affinity and avidity. Following TCR stimulation, multiple proteins are recruited to phosphorylated CD3 ITAMs, including the tyrosine kinase ZAP-70. ZAP-70 consequently phosphorylates tyrosine residues of the scaffolding protein LAT. These residues act as anchors for a range of T cell signalling proteins, including SLP76, which is also phosphorylated by ZAP-70. The pSLP76/LAT-complex functions as a scaffold for downstream effector proteins, including PLCγ (95). PLCs catalyse the hydrolysis of PIP₂ to generate the second messenger signalling molecules diacylglycerol (DAG) and inositol 3-phosphate (IP₃). In turn, IP₃ stimulates the release of Ca²⁺ from the endoplasmic reticulum (ER). Ca²⁺ and DAG stimulate activation of the RAP1 guanine exchange factor (GEF), CalDAG-GEFI (also known as RASGRP2) which activates RAP proteins by exchanging GDP for GTP (96). In T cells, RAP1 is a dominant isoform, with both Rap1a and Rap1b being expressed. Of note however, CalDAG-GEFI is not expressed in mouse lymphocytes, suggesting other RAP GEFs are involved in the regulation. Another pathway leading to RAP1 activation is recruitment of the CRKL-C3G complex by the WAVE2-Arp2/3-Abl complex (97, 98). This results in activation of the RAP-GEF, C3G (Also known as RAPGEFI), thus further activating RAP1. Active GTP-bound RAP1 is critical for the process of LFA-1 activation (99–103).

Chemokine receptors are GPCRs that following chemokine binding induce a multitude of signals, some which converge in activation of the small GTPase RAP1 via the activation of phospholipase Cβ (PLCβ) which also hydrolyses PIP₂ to DAG and IP₃ (104–106). Besides activating the PLC-dependent signalling-cascade, chemokine receptors also induce activation of PI3Kγ resulting in initiation of PI3K-mediated signals discussed further below.

GTP-bound RAP1 interacts with RIAM (107–109) and RAPL (100). In turn, this complex mediates activation (109), and plasma-membrane binding of TALIN1 (108). The FERM3 (F3) domain of TALIN1 in turn binds the β chain of LFA-1 thereby mediating conformational activation of LFA-1 from low to intermediate affinity, as well as mediating downstream cytoskeletal remodelling (110, 111). RIAM has a PH domain that preferentially interacts with PIP₂. By binding PIP₂ RIAM is thought to act as a proximity detector mediating binding of activated RAP1 and TALIN1 to the membrane (112). Another PH-domain containing protein involved in the process is SKAP1 (also known as SKAP55). SKAP1 is constitutively associated with ADAP (also known as FYB) and has been shown to also mediate binding of RAP1 to the plasma membrane through its PH domain (113, 114). Together SKAP1/ADAP integrates with the RIAM-RAPL-RAP1 complex during TCR-induced LFA-1 activation, and likely stabilises the complex (115). In parallel during LFA-1 activation, KINDLIN-3 binds the cytoplasmic tail of β integrins and is required for stabilisation of the high affinity conformation of LFA-1 (116–118). TALIN1 thus mediates conformational maturation to an intermediate affinity of LFA-1, whereas binding of both KINDLIN-3 and TALIN1 to the β-chain results in the high affinity conformation of LFA-1 (116).

Several negative regulators of LFA-1 activation exist. RhoH is required to keep LFA-1 in a non-adhesive state (119). RhoH also contributes to TCR signalling by interacting with ZAP70 and LCK (120, 121). Interestingly, chemokine-induced LFA1 activation is suppressed by RhoH, whereas RhoH enhances TCR-induced LFA-1 activation, suggesting that RhoH can divert T cells from chemotactic towards antigen-dependent response (122). CBL-B is an E3 is ubiquitinase that can suppress LFA-1 activation by interfering with the capacity of CRK-L to recruit and activate C3G (123).

Together these intricate regulatory mechanisms integrate migratory signals, such as chemokines, and TCR engagement with integrin activation. Consequently, LFA-1 affinity is turned on and off in a highly regulated manner by multiple microenvironmental cues.

Multiple proteins involved in the process of LFA-1 activation, including CYTOHESIN-1, SKAP1, and KINDLIN-3, have PH domains that bind PIP₃ and may hence regulate LFA-1 affinity in a PI3K dependent manner ( Figures 4A, B ).

It is intriguing that PI3K signalling regulates multiple processes involved in T cell migration. PI3K-mediated CD62L-shedding and reduced CD62L, S1P1, and CCR7 expression results in decreased LN entry and is an important step in T cell differentiation to effector subsets. Concurrently, as PI3K-signaling increases integrin affinity, PI3K signals can regulate migration and adhesion, including transendothelial migration into LNs. Consequently, inhibition of PI3K or disruption of PI3K signalling will affect all of these rheostats of migration, but not always in predictable ways. Hence, PTEN-deficient T cells with high PIP₃ levels are excluded from LNs after adoptive transfer (190). Nevertheless, APDS patients suffer from lymphadenopathy and this is revered upon treatment with a PI3Kδ inhibitor (36).

Studies of migration of PI3K-deficient T cells as well as use of inhibitors in mice provide some indication of how inhibition affects T cell distribution in vivo. p110γ-deficient T cells show reduced migration towards chemokines, whereas p110δ-deficient T cells respond to chemokines similarly to WT cells (191). Similarly, p110γ selective inhibitors affect responses to chemokines, whereas p110α/β/δ selective inhibitors do not affect responses to chemokines, except for at very high concentrations, likely as a result of off-target effects (192). Following LN entry, p110γ-deficient T cells seem to migrate interstitially similarly to WT T cells, and chemokine-induced interstitial migration seems independent of PI3K signalling (193). However, treatment with Wortmannin as well as disruption of regulatory p85 subunits of class IA PI3K showed that these cells migrated at lower velocities than WT cells, although T cell location within the LN did not seem altered (194). PI3Kδ under steady state does not contribute to T cell migration or chemokine-dependent migration per se, as p110δ^D910A T cells migrated like WT T cells in endothelial cell-coated transwell assays as well as following adoptive transfer (195). However, following antigenic challenge, p110δ was required for efficient migration to the site of inflammation and presence of antigen, consistent with a key role for PI3Kδ in regulating integrin affinity (195). Disruption of p110δ results in increased track velocities of OT-II CD4⁺ T cells in LN slices with OVA-pulsed DCs, as a result of decreased interaction times with the peptide-presenting DCs in the slices (128). Similar results have been observed for p110γ-deficient T cells that are defective for antigen-dependent and chemokine-dependent migration of effector CD4⁺ and CD8⁺ T cells (196, 197). Interestingly, CD28 seems to also be important for homing of antigen-stimulated T cells to non-lymphoid tissues, whereas CD28 (Y173F) that is uncoupled from PI3Kδ was defective. This suggests that CD28-mediated activation of PI3K is involved in migration of activated T cells to non-lymphoid sites (198). Consequently, when inhibiting PI3Kδ, homeostatic migration of naïve T cells (T_N) seems unperturbed (as these have low PI3K activity in the first place), whereas activated T cells show decreased antigen-dependent migration into non-lymphoid tissues.

Central memory T cells (T_CM) that are CD62L⁺CCR7⁺LFA-1⁺ are consequently supported by PI3Kδ inhibition, whereas effector T cells (T_eff) and effector memory T cells (T_EM) (CD62L^-CCR7^-LFA-1⁺) are inhibited ( Figure 6A ). This is largely supported by the fact that APDS patients have reduced T_CM cells, and increased T_eff cells (32), whereas p110δ^D910A mice have normal memory T cells, but reduced T_eff cells (22) ( Figure 6B ). These PI3K-dependent alterations of T cell memory responses are possibly affected by altered expression of migratory receptors, however, differentiation of p110δ^D910A T cells to T_eff is largely defective, implicating PI3K more broadly in differentiation and migration.