Regulation of Lrp6 phosphorylation

Introduction

Wnt genes encode a family of evolutionarily conserved secreted glycoproteins that regulate many aspects of embryonic development as well as the homeostasis of adult tissues [1–3]. Wnts initiate several signaling cascades via diverse cell-surface receptors [4]. These cascades are grossly subdivided as canonical Wnt/β-catenin signaling and non-canonical (β-catenin-independent) Wnt signaling [4–6].

Gsk3 mediates PPSP phosphorylation in response to acute Wnt signaling

Gsk3 is a serine/threonine protein kinase, which is encoded by shaggy in Drosophila and by Gsk3α and Gsk3β in vertebrates. Gsk3 has a plethora of substrates and participates in numerous cell signaling cascades [50–52]. In Wnt/β-catenin signaling, Gsk3 is a master negative regulator of β-catenin. In spite of this negative role, Gsk3 phosphorylates Lrp6 and thereby positively regulates Wnt/β-catenin signaling [26]. In vitro, Gsk3 directly binds Lrp6 ICD and phosphorylates PPSP repeats [26, 53]. In cultured cells, overexpression of Gsk3 promotes PPSP phosphorylation at S1490 [26]. When artificially anchored to the plasma membrane, Gsk3 activates Wnt/β-catenin signaling in cultured cells and Xenopus embryos [26]. Wnt induces Lrp6 PPSP phosphorylation at repeats A, C, and E (S1490, T1572, S1607) and most likely also at repeats B and D (T1530, S1590). Wnt-induced PPSP phosphorylation is inhibited by pharmacological Gsk3 inhibitors or genetic ablation of Gsk3 [26, 45, 47]. Gsk3 substrates share the unique phosphorylation motif [S/T]-X-X-X-p[S/T], which typically must first be primed by phosphorylation at the +4 S/T site [51]. Gsk3 phosphorylation of Lrp6 PPSP repeats is unusual, as there is no priming residue at +4 positions in the PPSPXS repeats. The results indicate that Gsk3 mediates PPSP phosphorylation in response to acute Wnt signaling. However, its physiological relevance in vivo remains to be demonstrated.

Protein kinase A mediates Lrp6 PPSP phosphorylation in response to parathyroid hormone

Protein kinase A (PKA) is a cyclic AMP (cAMP)-dependent serine/threonine protein kinase. PKA phosphorylates transcription factor CREB and hundreds of other targets [54]. A recent report showed that parathyroid hormone (PTH) can signal through Lrp6 by forming a ternary complex with its receptor PTH1R and Lrp6 to promote osteoblast differentiation [48]. PTH treatment induces rapid Lrp6 S1490 phosphorylation and activates PTH and β-catenin signaling. A pharmacological PKA inhibitor blocks PTH-induced S1490 phosphorylation, suggesting that this kinase is involved in Lrp6 phosphorylation. In vitro, PKA can directly phosphorylate Lrp6, but the sites phosphorylated have not been analyzed [48]. PKA also mediates a Wnt-mediated myogenic pathway during mouse development [55], suggesting that PKA may be more widely involved in Lrp6 phosphorylation. In the osteoblastic cell line Saos-2, PTH/PKA signaling can also induce Ser9 phosphorylation and inhibition of Gsk3, thereby promoting canonical Wnt signaling [56], raising the possibility that PTH/PKA effect phosphorylation at two different levels which synergize to activate canonical Wnt signaling. Like Lrp6, PKA has also been implicated in regulating Wnt/Planar Cell Polarity (PCP) signaling during convergent extension movements [57, 58]. It is unknown if the role of Lrp6 in PCP signaling involves PKA.

Grk5/6 mediate PPSP phosphorylation in response to acute Wnt signaling

Grk5/6 are membrane-associated protein kinases belonging to the G protein-coupled receptor kinase (Grk) family that specifically recognize and phosphorylate agonist-activated G protein-coupled receptors (GPCR). GPCR phosphorylation leads to GPCR desensitization, endocytosis, intracellular trafficking, resensitization, and modulation of intracellular GPCR signaling [59]. However, Grks also phosphorylate non-GPCR substrates, such as Synuclein and Tubulin [60–64]. In Wnt/β-catenin signaling, Grk2 binds Apc and inhibits the signaling [65]. Conversely, Grk5/6 can promote both Lrp6 S1490 phosphorylation and Wnt signaling [66]. Grk5 is necessary for S1490 phosphorylation in response to acute Wnt stimulation. Grk5/6 knock down phenocopies loss of Wnt signaling in zebrafish. In vitro, Grk5/6 can directly phosphorylate PPSP repeats A–E. Therefore, Grk5/6 functions as an Lrp6 PPSP kinase in response to acute Wnt signaling. Interestingly, parathyroid hormone, whose receptor itself is phosphorylated by Grk5 [67], also regulates Lrp6 PPSP phosphorylation [48].

Pftk1/cyclin Y mediate PPSP phosphorylation during mitosis

The cyclin-dependent protein kinase (Cdk) superfamily of serine/threonine protein kinases plays important roles in eukaryotic cell cycle regulation. Eukaryotic cell division is an ordered process that consists of four distinct phases: G1 phase, S phase, G2 phase (collectively known as interphase), and M phase (mitosis). Cdks are dependent on associations with their activating subunits, the cyclins, to be enzymatically active [68]. The levels of regulatory cyclins typically oscillate throughout cell cycle so that Cdks can initiate phase-specific events and control transitions of specific phases [68]. Apart from directly activating Cdks, cyclins contain regions that can restrict Cdks to specific subcellular locations or targets [69, 70].

Recently, a Cdk member was identified as a PPSP kinase, namely Drosophila L63 and its vertebrate homologue PFTAIRE protein kinase 1 (Pftk1) [71]. Homologues of Pftk1/L63 are PFTAIRE protein kinase 2 (Pftk2) and PCTAIRE protein kinase 1-3 (Pctk1-3) [72, 73]. Cyclin Y (Ccny) is a partner of L63/Pftk1 [71, 74, 75] and regulates Pftk1 activity at multiple levels [71]. In the presence of Ccny, Pftk1 phosphorylates Lrp6 PPSP repeat A both in vivo and in vitro, while the kinase on its own is inactive towards Lrp6. Ccny contains an N-terminal myristoylation signal and is predominantly enriched at the inner surface of plasma membrane [71, 75], where it binds the receptor Lrp6 and recruits Pftk1 [71]. Importantly, overexpressed Ccny promotes Pftk1 phosphorylation of Lrp6, while disruption of Ccny membrane association abolishes this ability [71]. Ccny is regulated by ubiquitin-mediated proteolysis and its level along with Pftk1 peaks during G2/M phase of cell cycle [71]. Remarkably, Lrp6 PPSP phosphorylation and Wnt/β-catenin signaling are also under cell cycle control and peak during G2/M phase in cultured cells and in Xenopus embryos [71]. Ccny inhibition in Xenopus embryos induces antero-posterior patterning defects, which phenocopy loss of Wnt/β-catenin signaling. Intriguingly, a reduction of Lrp6 phosphorylation and Wnt signaling provoked by Lrp6 knock down is partially rescued by knock down of cdc25 [71]. Cdc25 is a protein tyrosine phosphatase, which plays a key role in regulating entry into mitosis, by dephosphorylating cyclin B-bound Cdk1 [76]. Cdc25 knock down arrests cells in G2 phase, thereby promoting Lrp6 phosphorylation and downstream signaling. The surprising finding that Lrp6 phosphorylation and Wnt/β-catenin signaling peak during mitosis is also supported by previous findings that β-catenin levels peak between G2/M [77, 78] and that Cdc25 mutants in mice and in Drosophila show enhanced Wnt signaling [79, 80]. Taken together, Pftk1/Ccny represents a mitotic Cdk, which functions as Lrp6 PPSP kinase.

Mitotic Wnt signaling suggests that Wnt signaling may play a broader role beyond transcriptional regulation. In fact, Apc is localized to kinetochores, centrosomes, and the +end of microtubules [81, 82] and regulates spindle orientation, cell fate specification, chromosome alignment, chromosome stability, and the mitotic checkpoint [81–86]. Likewise, β-catenin binds to centrososmes and regulates their separation and normal spindle formation [87, 88]. Similarly, Gsk3 and Axin also have direct functions in the mitotic apparatus [89–91]. Another prominent example for mitotic Wnt/β-catenin signaling is Caenorhabditis elegans, where it is essential for orienting the mitotic spindle during asymmetric cell division in embryonic development [92–94]. Taken together, this suggests that Pftk1/Ccny phosphorylates Lrp6 at G2/M to induce Wnt/β-catenin signaling for orchestrating a mitotic program beyond transcriptional regulation.

Casein kinase 1 family

While all other kinases mentioned are proline-directed kinases involved in PPSP phosphorylation, members of the CK1 family are responsible for phosphorylation of PPSPXS sites as well as the S/T cluster. The CK1 family of serine/threonine protein kinases is highly conserved from yeast to humans [95]. The family comprises CK1α,δ,ε (discs overgrown/dco in Drosophila) and γ1-3 isoforms (gilgamesh/gish in Drosophila). Consensus CK1 recognition motifs are (1) p[S/T]-X_1-2-[S/T] and (2) [D/E]-X_1-2-[S/T], where the S/T is the CK1 phosphorylation site. Phosphorylated S/T or the negatively charged D/E residue primes the phosphorylation [96, 97]. Additional consensus recognition sites have also been documented [98]. In Wnt/β-catenin signaling, CK1 members phosphorylate multiple pathway components, including β-catenin, Dishevelled (Dvl/Dsh), Apc, Axin and Tcf/Lef to negatively and/or positively regulate signaling [99, 100]. In Lrp6 phosphorylation, multiple CK1 members act to positively and negatively regulate the receptor as discussed below.

Hierarchy of Lrp6 phosphorylation at S/T cluster and PPSPXS motif A

With multiple Lrp6 kinases, their effects may be hierarchical, additive, or mutually exclusive [103]. Certain Lrp6 protein kinases including Gsk3 and CK1 family members are known to be primed by pre-phosphorylation at adjacent sites [104]. Since Lrp6 contains multiple Gsk3 and CK1 phosphorylation sites, this suggests that there may be a hierarchy of phosphorylation events. Analysis of this hierarchy faces two main difficulties. First, each Lrp6 phosphorylation site is phosphorylated by more than one kinase. This may explain why, using the same phospho-specific antibodies, different researchers arrive at different conclusions [25, 26], because they may in fact monitor distinct phosphorylation events. For example, S1490 is phosphorylated Wnt-dependently by Gsk3 and Grk5/6, and Wnt-independently by Pftk1 and PKA. Second, while in vitro kinase assays with partially purified or recombinant kinases are a clean and informative way to reduce complexity to study Lrp6 phosphorylation, the physiological relevance of such studies may be compromised. This is because, in in vitro assays, kinases tend to be more promiscuous than in in vivo.

With these caveats, three hierarchies of Lrp6 phosphorylation have been proposed which, however, are not mutually exclusive (Fig. 3). There is good evidence that T1493 phosphorylation is dependent on S1490 phosphorylation in cultured cells and in vitro [25, 26]. This would be consistent with the established CK1 consensus phosphorylation site p[S/T]-X_1–2-[S/T] [25, 71, 105]. It seems reasonable to extrapolate this hierarchy to all PPSPXS sites (Fig. 2b). However, while pre-phosphorylation is necessary for CK1 phosphorylation, it seems not sufficient in vivo. For example, in various cell lines, Lrp6 shows constitutive phosphorylation at S1490 but not at CK1 sites [25, 71]. Instead, CK1 site phosphorylation requires Wnt-mediated receptor clustering and signalosome formation [24]. For another example, Pftk1/Ccny-mediated S1490 phosphorylation triggers CK1 phosphorylation only in the presence of Wnt [71]. Furthermore, Wnt-dependent CK1 phosphorylation of the S/T cluster (including T1479) is independent of S1490 phosphorylation [25, 106]. It is unclear if S/T cluster phosphorylation involves another priming event or not, since CK1 can also phosphorylate non-primed sites [96]. There is evidence that S/T cluster phosphorylation stimulates PPSP phosphorylation [106] (Fig. 3b). This is consistent with the direct binding of Gsk3 to the phosphorylated S/T cluster [107].

Taken together, this suggests a sequential priming model, which starts from the S/T cluster and progresses C-terminally (Fig. 3c). If this model is correct, it would imply that CK1 and Gsk3 act in an alternating fashion (“ping-pong” mode). It will be important to systematically address the hierarchy of Lrp6 phosphorylation, for example, by using phospho-specific antibodies against the relevant sites, by determining the kinetics of phosphorylation following Wnt stimulation, and by combining this with specific knock down of relevant protein kinases.

Processive or distributive phosphorylation of multiple PPSPXS sites

Phosphorylation is referred to as processive if a protein kinase binds once to a substrate and phosphorylates all of the available sites before dissociating. Alternatively, phosphorylation may be distributive, where each phosphorylation requires a separate binding event between the kinase and substrate [103]. With multiple phosphorylation sites, mixed forms of processive and distributive phosphorylation may also occur.

As an example of processive phosphorylation, Gsk3 phosphorylates sites that contain a phospho-Ser at the +4 position, and once Ser656 of glycogen synthase has been phosphorylated by CK2, Gsk3β sequentially phosphorylates S652, S648, S644, and S640. It has been shown that each phosphorylation depends on prior phosphorylation at the +4 position [108, 109]. The modular Lrp6 structure initially may suggest a processive phosphorylation mechanism, where the kinases scan along the cytoplasmic domain to phosphorylate Lrp6. For Gsk3, however, the PPSP sites are spaced too far apart to be priming in a conventional processive mode. Furthermore, Gsk3 is inhibited via end product inhibition by phosphorylated PPSP sites [53, 107, 110, 111] (see below). This suggests that, upon Wnt stimulation, at least Gsk3 may phosphorylate Lrp6 in a distributive mode. For any of the other Lrp6 kinases, we do not know whether they scan Lrp6 in a processive mode.

Multiple PPSPXS motifs function cooperatively to transduce Wnt signaling

Each PPSPXS motif on its own, for example, when artificially anchored to the plasma membrane, is able to activate Wnt/β-catenin signaling and recruit Axin [25–27, 45]. Progressive deletion of Lrp6 PPSPXS motifs incrementally reduces their ability to promote Wnt signaling [25, 45, 46]. However, in the context of full length Lrp6, a single PPSPXS is insufficient to transduce Wnt signaling [45, 46]. Instead, multiple PPSPXS motifs function cooperatively to transduce Wnt signaling. Consistently, in response to Wnt stimulation, Lrp6 is phosphorylated at multiple PPSPXS motifs [25–27, 45]. A higher stoichiometry of Lrp6 phosphorylation may promote higher affinity binding to Axin and Gsk3 and hence promote downstream signaling. This highlights that it will be important in the future to determine the contribution of individual kinases in regard to their contribution to stoichiometric Lrp6 phosphorylation, which has not been analyzed either in vivo or in vitro.

Lrp6 phosphorylation and signaling in signalosomes

Real-time imaging showed that within minutes of Wnt treatment, cytoplasmic Axin colocalizes with Lrp6 in punctate structures below the plasma membrane. Further analysis showed that these structures represent Lrp6 aggregates, or signalosomes, which are phosphorylated, ribosome-sized multiprotein complexes containing Fzd, Dvl, Gsk3, and Axin. Signalosomes presumably represent some sort of endocytic vesicles, which, however, are negative for most endocytic markers, except for occasional colocalisation with caveolin. This would be consistent with the fact that Wnt stimulation requires caveolin mediated endocytosis of Lrp6 [43]. The scaffold protein Dvl is required for Lrp6 aggregation and Lrp6 phosphorylation at T1479 [24, 42] and S1490 [24, 42, 47]. Dvl itself forms cytoplasmic polymers, which can be visualized as microscopic punctae and which are recruited to the plasma membrane upon Wnt stimulation, presumably via binding to activated Fzd [112]. DIX domain mediates Dvl polymerization, which probably increases the avidity for Axin [113, 114]. The ability of Dvl to polymerize is essential for Lrp6 phosphorylation and Wnt signaling [24, 114]. Lrp6 signalosome formation also requires the action of phospholipids and lipid kinases. Wnt stimulates activities of phosphatidylinositol 4-kinase type II α (PI4KIIα) and phosphatidyl-4-phosphate 5-kinase type I (PIP5KI) through Fzd and Dvl. This in turn induces formation of PtdIns(4,5)P2 [42], which stimulates Lrp6 phosphorylation through an unknown mechanism. Phospholipid is known to be necessary for Wnt-induced Dvl phosphorylation and Wnt signaling [115, 116].

Taken together, these results suggest a signalosome model for Lrp6 phosphorylation and signaling (Fig. 1b), where Wnt induces the formation of a ligand–receptor complex which recruits Dvl, Axin, and Gsk3. Dvl polymers promote Lrp6 aggregation, which triggers the phosphorylation by CK1γ and Gsk3, presumably providing a high density of phosphorylation sites (Fig. 1c). The phosphorylation is a consequence but not the cause of aggregation, since Lrp6 signalosomes also form in the presence of dominant-negative CK1γ, which blocks phosphorylation [24]. The key prediction of this model is that any kind of Lrp6 polymerization should induce phosphorylation and Wnt signaling. Results obtained with Lrp6 fused to artificial oligomerization domains are consistent with this [23]. The net result of Lrp6 signalosome formation is the recruitment and inhibition of Gsk3 by phosphorylated Lrp6 (see below), which derepresses β-catenin [24].

According to an initiation–amplification model, Lrp6 phosphorylation happens in two phases [47]. In the initiation phase, Wnt induces an Lrp6–Fzd complex, which recruits the Axin–Gsk3 complex through Fzd-bound Dvl to initiate Lrp6 phosphorylation. In the amplification phase, initial phospho-Lrp6 serves as high affinity docking site for more Axin–Gsk3, so that Gsk3 acts in cis and possibly in trans to phosphorylate Lrp6 [47]. During the amplification phase, there is a positive feedback loop: phospho-Lrp6 will recruit more Axin–Gsk3 complex, which may phosphorylate more Lrp6 and inhibit more Gsk3 because of product inhibition, thereby amplifying Wnt signaling. The signalosome model and dynamic polymerization-phosphorylation are consistent with this proposed amplification. In contrast, it appears there is neither a need nor compelling evidence for signaling happening in a distinct initiation phase.

Downstream of Lrp6 phosphorylation: Axin–Gsk3 recruitment and Gsk3 inhibition

Conclusions

In recent years, significant advances have been made in understanding how the Wnt co-receptors Lrp6 and Fzd transduce Wnt stimuli. Wnt signaling occurs on Lrp6 signalsomes, which contain receptor complexes bound to Dvl polymers. In response to signalosome formation, Lrp6 phosphorylation is triggered on an S/T cluster and on multiple PPSPXS sites by Gsk3 and multiple CK1 members, notably CK1γ. Phospho-Lrp6 recruits Axin and additional Gsk3, whose activity is inhibited by product inhibition. This derepresses β-catenin and promotes Wnt signaling. In addition to Gsk3, other kinases have been implicated in PPSP phosphorylation. Grk5/6 phosphorylate Lrp6 to mediate acute Wnt signaling. PKA phosphorylates Lrp6 to mediate parathyroid hormone signaling via β-catenin. A cyclin-dependent protein kinase complex, Pftk1/Ccny, phosphorylates Lrp6, which explains why Wnt signaling is under mitotic control. Thus, there is context-dependent PPSP phosphorylation by different kinases (Wnt-, PTH-, or mitosis-dependent PPSP phosphorylation).

The multitude of kinases involved indicates that Lrp6 phosphorylation represents an important regulatory node and is an opportunity for signaling cross-talk, suggesting that there may be additional kinases yet to be identified. Open questions are how different kinases cooperate in Lrp6 phosphorylation, and whether they act additively, sequentially, or mutually exclusively. Furthermore, the stoichiometry of Lrp6 phosphorylation by individual PPSP kinases and their combinations is unclear. Lastly, the physiological significance for most kinases in vivo needs to be analyzed. However, the tools and conceptual framework are now at hand to address these questions.