Cytokine Receptor Endocytosis: New Kinase Activity-Dependent and -Independent Roles of PI3K

Type I and II cytokine receptors are cell surface sensors that bind cytokines in the extracellular environment and initiate intracellular signaling to control processes such as hematopoiesis, immune function, and cellular growth and development. One key mechanism that regulates signaling from cytokine receptors is through receptor endocytosis. In this mini-review, we describe recent advances in endocytic regulations of cytokine receptors, focusing on new paradigms by which PI3K controls receptor endocytosis through both kinase activity-dependent and -independent mechanisms. These advances underscore the notion that the p85 regulatory subunit of PI3K has functions beyond regulating PI3K kinase activity, and that PI3K plays both positive and negative roles in receptor signaling. On the one hand, the PI3K/Akt pathway controls various aspects downstream of cytokine receptors. On the other hand, it stimulates receptor endocytosis and downregulation, thus contributing to signaling attenuation.


Multiple Pathways for endocytosis of Cytokine Receptors
Receptor endocytosis is initiated at the plasma membrane and can be generally divided into clathrin-mediated endocytosis (CME) or clathrin-independent endocytosis (CIE) based on the involvement of the endocytic coat protein clathrin (31,32). In CME, activated receptors recruit clathrin adaptors such as the AP2 complex, inducing the formation of a clathrin coat that stabilizes membrane curvature and drives invagination. Subsequently, vesicles are pinched off from the plasma membrane by the dynamin GTPase (10,33). CIE is a composite of several distinct pathways, the best studied being the caveolin-mediated endocytosis (34,35). These pathways, which can be either dynamin dependent or independent (13), require actin polymerization and either Src-family kinases in the case of caveolin-mediated endocytosis (36) or small GTPases such as RhoA and Rac1 for other CIE pathways (37).
Ubiquitination plays an important role in receptor endocytosis through both CME and CIE (55). Through sequential actions of ubiquitin-activating (E1), ubiquitin-conjugating (E2), and ubiquitin-ligating (E3) enzymes, a small protein ubiquitin is covalently attached to lysine residues on target receptors. Because ubiquitin itself contains lysines that can serve as acceptor sites, target proteins can be subjected to mono-ubiquitination, multiubiquitination (mono-ubiquitination on multiple lysines), or poly-ubiquitination. Mono-ubiquitination has been shown to mediate protein trafficking and signaling (56), whereas polyubiquitination can promote protein degradation (55). Endocytic adaptor proteins and the endosomal sorting complex required for transport (ESCRT) contain ubiquitin-binding domain or ubiquitin-interacting motif (UIM), thereby facilitating their interaction with ubiquitinated receptors. This allows endocytic adaptors to target ubiquitinated receptors to the endocytic machinery and allows the ESCRT complexes to direct budding of ubiquitinated receptors into intraluminal vesicles within endosomes, thereby halting receptor signaling (57).
Class IA PI3Ks function as heterodimers with a p110 catalytic subunit (p110α, β, or δ) and a p85-like regulatory subunit (p85α, β or their splice variants p55α, p50α, or p55γ) (4). p85 stabilizes and maintains p110 in an inhibited state and directly interacts with phosphorylated cytoplasmic tyrosines in cytokine receptors upon ligand binding. Conformational changes in p85 induced by receptor binding relieve its inhibition of p110 (67). Recent evidence suggests that the association and activation of PI3K by cytokine receptors promotes receptor endocytosis in addition to the activation of downstream Akt signaling (68,69). Moreover, the contribution of the p85 regulatory subunit in these mechanisms can be PI3K kinase activity independent (48). Below, we discuss two new paradigms by which class IA PI3Ks regulate cytokine receptor endocytosis.

Pi3K and Actin-Mediated endocytosis of iL2 Receptor (iL2R)
IL2 receptor belongs to the Type I cytokine receptors and is important for T cell immune function (70,71). IL2R is composed of IL2Rα, IL2Rβ, and the common γ chain. Internalized IL2Rα recycles back to the plasma membrane, whereas IL2Rβ and the common γ chain are sorted to the lysosome and degraded (72,73). IL2Rβ was among the first cytokine receptors shown to be internalized via CIE (74). It is constitutively internalized, but internalization is augmented by IL2 binding (69,75).
Endocytosis of IL2Rβ is clathrin-and caveolin independent and relies on RhoA, dynamin, Rac1, PAK kinases (p21-activated kinases), and actin polymerization (38,49,76,77). New studies showed that two rounds of actin polymerization are enlisted for IL2Rβ internalization. The first round relies on WAVE (WASP-family verprolin homologous protein), through a WAVEinteracting sequence in the cytoplasmic tail of IL2Rβ (78). This round of actin polymerization occurs before receptor clustering and is thought to be responsible for receptor recruitment near the base of membrane protrusion to initiate pit formation. The second round occurs just before receptor internalization and involves Pak1 phosphorylation of cortactin, another activator of actin polymerization (79,80), thereby increasing its association with N-WASP (neuronal Wiskott-Aldrich syndrome protein) (77). Interestingly, dynamin, which mediates vesicle scission in the later stage of IL2Rβ internalization, also controls the transition of WAVE complex and N-WASP recruitments (78).
Sauvonnet's group showed that PI3K plays multiple roles in regulating IL2R CIE (69). First, IL2 stimulation activates PI3K, leading to the production of PI(3,4,5)P3 and the recruitment of Vav2, the guanine nucleotide exchange factor that activates Rac1 (81). Inhibitors of PI3K kinase activity, knockdown of p85 and Vav2, or overexpression of a mutant p85 devoid of p110-binding domain all inhibit IL2R endocytosis. Second, p85 binds directly to Rac1, with higher affinity for the GTP-bound active form. A model is thus proposed that IL2R activation of PI3K leads to the recruitment of both Vav2 and its substrate Rac1, which can stimulate the Rac1-Pak1-cortactin-N-WASP cascade to promote actin polymerization, driving IL2R internalization (Figure 1) (69). Because the WAVE complex is a known downstream effector of Rac GTPases (82,83) and PIP3 (84,85), PI3K may also regulate IL2Rβ CIE through WAVE.
Recently, endophilin and its interacting protein Alix (ALG-2interacting protein X) have also been implicated in CIE of IL2Rβ (86,87). Endophilin is a Bin/Amphiphysin/Rvs domain protein that is involved in vesicle endocytosis and membrane curvature generation (88,89). This pathway, termed fast endophilinmediated endocytosis (FEME) by the McMahon group, is utilized by IL2R as well as several G-protein-coupled receptors and bacterial Shiga and cholera toxins (87,90). It is characterized by endophilin-positive uptake structures after ligand-induced receptor activation. Endophilin also works together with dynamin and actin in membrane scission (90,91). As with the PI3K/Vav2 pathway described above, the FEME pathway depends on dynamin, Rac, Pak1, and actin polymerization (87), suggesting that FEME and PI3K/Vav2 mechanisms may be part of the same pathway. Importantly, PI3K kinase activity is required for FEME, because PI(3,4)P2, converted from PI(3,4,5)P3 by SHIP1/2-dependent dephosphorylation, is necessary for lamellipodin-dependent recruitment of endophilin in FEME (Figure 1) (87). The exact molecular details of this pathway, the degree to which the PI3K/ Vav2 and FEME pathways are distinct or can be employed under different context, and whether PI3K regulates other aspects await future interrogations. In addition, whether other cytokine receptors can also utilize similar endocytic pathways is currently unclear.

Cbl-DePenDenT UBiQUiTinATiOn OF p85 MeDiATeS epoR enDOCYTOSiS
The EpoR is another member of the Type I cytokine receptors and is essential to drive red blood cell production (92,93). In contrast to the IL2R, which forms heteromeric receptor complexes and associates with both JAK1 and JAK3 for signaling, EpoR forms homodimers and couples to only JAK2 for signaling. Epo-induced endocytosis is a key element in negative regulation FiGURe 2 | Clathrin-dependent endocytosis of erythropoietin receptor (epoR). Upon Epo stimulation, activated JAK2 phosphorylates EpoR cytoplasmic tyrosines to recruit p85 (step 1). Subsequently, ubiquitinated p85, mediated by c-Cbl (step 2), recruits Epsin-1 (step 3), linking EpoR to the endocytic machinery for downregulation (48,68).
of Epo signaling (48,94) and controls cellular Epo sensitivity and the level of Epo in the circulation (95,96). Studies in our laboratory have shown that Epo induces internalization of EpoR via CME, and we identified a novel function of p85 in EpoR endocytosis and downregulation (Figure 2) (48,68). Epo stimulation activates JAK2, resulting in the phosphorylation of multiple EpoR cytoplasmic tyrosine residues, including Y 429 , Y 431 , and Y 479 . These phosphotyrosines serve as mutually redundantly docking sites for binding of the p85 subunit of PI3K to EpoR (48). p85 binding activates the catalytic p110 subunit, resulting in PI (3,4,5) P3 production and Akt signaling, which is required for erythroid differentiation. Unexpectedly, Epo-induced EpoR internalization does not require PI3K kinase activity (48). Instead, Epodependent ubiquitination of p85 by the E3 ligase c-Cbl recruits the endocytic adaptor protein, Epsin-1, through its UIM. Epsin-1 then connects the EpoR/p85 complex to the clathrin-mediated endocytic machinery for internalization (68).
The physiological relevance of this pathway is highlighted by mutated EpoRs found in patients with primary familial and congenital polycythemia (PFCP), a proliferative disorder of the red cell lineage characterized by increased red blood cell mass (97,98). PFCP patients harbor mutations that delete the C-terminal cyto solic domain of the EpoR, resulting in EpoR truncations lacking all three tyrosines responsible for p85 binding. Mutated EpoRs mimicking those found in PFCP patients cannot bind p85 and are unable to recruit Epsin-1 to engage the endocytic machinery. As a result, these receptor variants do not internalize upon Epo stimulation and exhibit Epo hypersensitivity. Similarly, knockdown of Cbl also causes Epo hypersensitivity in primary erythroid progenitors. Restoring p85 binding to PFCP receptors rescues Epo-induced Epsin-1 co-localization and normalizes Epo hypersensitivity (48,68). These results elucidate the molecular mechanism underlying Epo-induced p85-mediated EpoR internalization and demonstrate that defect in this pathway may contribute to the etiology of PFCP. Although still controversial, non-canonical heterodimeric complexes consisting of EpoR and the βc receptor have been implicated in non-hematopoietic tissues (99). Whether the p85-Cbl pathway plays a role in endocytosis of these complexes is unclear.
PI3K is activated by most cytokine receptors, whereas Cbl also functions downstream of many signaling receptors. Therefore, the p85-Cbl pathway might be utilized more broadly to contribute to endocytosis of other cytokine receptors. In addition, the same molecules may be employed in different ways for receptor endocytosis and downregulation. For example, the thrombopoietin receptor activates PI3K for signaling, and utilizes Cbl for downregulation. However, instead of ubiquitinating p85 as in the case of the EpoR, the thrombopoietin receptor itself is poly-ubiquitinated by Cbl upon stimulation, leading to its degradation (63).

COnCLUSiOn AnD PeRSPeCTiveS
The two new paradigms reviewed here underscore the contribution of PI3K in CME (e.g., EpoR) as well as CIE (e.g., IL2R) of cytokine receptors. Besides class I PI3K discussed here, class II PI3K, which produces PI(3)P and PI(3,4)P2, has also been shown to participate in late stage CME (100). These broaden the roles of PI3K family kinases as fundamental and integral regulators of endocytosis in general.
The mechanisms underlying PI3K's contributions are both kinase activity dependent and -independent. PI3K kinase activity is required to recruit Vav2 and endophilin for IL2R internalization. By contrast, in a PI3K kinase activity-independent manner, p85 recruits activated Rac1 to promote IL2R endocytosis and recruits Cbl/Epsin-1 to promote EpoR internalization. Therefore, PI3K plays both positive and negative roles upon cytokine receptor activation. On the one hand, the PI3K/Akt pathway controls various aspects downstream of cytokine receptors. On the other hand, it stimulates receptor endocytosis and downregulation, thus contributing to signaling attenuation.
These advances also highlight the emerging concept that p85 has functions beyond regulating PI3K kinase activity (101)(102)(103)(104)(105). For example, cytokinesis defects observed in p85α-deficient cells are restored by expression of a p85α mutant that does not bind p110 (102). It was also shown that p85 exhibits in vitro GTPaseactivating protein (GAP) activity toward Rab5, which regulates vesicle trafficking and actin remodeling (106,107). A p85α mutant with defective GAP activities perturbed PDGF receptor trafficking and caused cellular transformation via a kinase-independent mechanism (105,108). Whether the GAP activity of p85 or Rab5 contributes to IL2Rβ or EpoR endocytosis is unclear. Moreover, p85 also interacts with dynamin (109), the contribution of this interaction is not known. Other p85-interacting proteins, such as phosphatases (e.g., SHP2) and adaptor proteins (e.g., IRS1), may also contribute to its function (110,111).
One last layer of complexity we would like to bring up has risen from recent studies concerning dynamin isoform-specific functions. Normally, vertebrates express three dynamin (Dyn) isoforms: Dyn2 is ubiquitously expressed, whereas Dyn1 and Dyn3 are most highly expressed in specific tissues (112,113). Under normal conditions, Dyn1 contributes little to CME in non-neuronal cells; however, Reis et al. recently showed that Akt, the canonical kinase downstream of PI3K, activates Dyn1 in epithelial cells to induce accelerated CME with altered dynamics (114). These results raise the interesting possibility that cytokine receptors may stimulate their endocytosis through Akt-dependent activation of Dyn1, adding to the concept that the endocytic machinery can be specifically adapted by signaling receptors to regulate their own endocytosis. Regulatory controls of endocytic components and mechanisms significantly impact physiology and human diseases. Much of what we know about the cross talk between endocytosis and signaling comes from work done with model receptors such as receptor tyrosine kinases (RTK). Many of these lessons may translate to cytokine receptors, because JAK kinases activate many pathways in common with RTKs. Also, in many cases, JAK kinases are integral partner of cytokine receptors, making receptor/JAK complexes equivalent to RTKs (115,116). However, signaling is not identical and differences are to be expected. Among the open questions are the following: First, do JAK kinases regulate endocytosis beyond receptor phosphorylation? Can they modulate the endocytic machinery directly? Second, does the PI3K/Akt signaling cascade provide a feedback loop for receptor endocytosis in general? Consistent with this notion, Akt promotes EGF receptor degradation by phosphorylating and activating the PIKfyve kinase (FYVE-containing phosphatidylinositol 3-phosphate 5-kinase), which stimulates vesicle trafficking to lysosomes (117). Third, does the GAP activity of p85 and/or other p85-interacting proteins play a role in cytokine receptor endocytosis?
Fourth, how do cytokine receptors employ the molecular toolbox of signaling and endocytic proteins in different cell types and contexts such as normal vs. disease states? More detailed mechanisms are needed to understand the reciprocal cross talk between endocytosis and signaling, which will help to improve our understanding of the physiological functions of cytokine receptors.

AUTHOR COnTRiBUTiOnS
All the authors contributed to the writing of the review.

ACKnOwLeDGMenTS
The authors are grateful to their colleagues Drs. Sandra Schmid and Peter Michaely, who generously shared their insights. They apologize to the many researchers whose work was not discussed because of space constraints.

FUnDinG
This study was supported by funding from the National Institutes of Health, USA to LJH (HL089966) and a Taiwan National Science Council Grant (103-2917-I-564-029) to PHC.