SHED-Dependent Oncogenic Signaling of the PEAK3 Pseudo-Kinase

Simple Summary The human kinome is composed of about 50 pseudo-kinases with unclear function, because they are predicted to be catalytically inactive; however, they are shown to play an important role in cancer, similar to active kinases. Understanding how these pseudo-kinases promote tumor formation despite their catalytic inactivity is a great challenge, which may lead to innovative anti-cancer therapies. The PEAK1 and 2 pseudo-kinases have emerged as important components of the protein tyrosine kinase pathway implicated in cancer progression. They can signal using a scaffolding mechanism via a conserved split helical dimerization (SHED) module. In this study, we uncovered a similar SHED-dependent oncogenic activity for PEAK3, a recently discovered new member of this family. We also show that this new signaling mechanism may be implicated in acute myeloid leukemia. Abstract The PEAK1 and Pragmin/PEAK2 pseudo-kinases have emerged as important components of the protein tyrosine kinase pathway implicated in cancer progression. They can signal using a scaffolding mechanism that involves a conserved split helical dimerization (SHED) module. We recently identified PEAK3 as a novel member of this family based on structural homology; however, its signaling mechanism remains unclear. In this study, we found that, although it can self-associate, PEAK3 shows higher evolutionary divergence than PEAK1/2. Moreover, the PEAK3 protein is strongly expressed in human hematopoietic cells and is upregulated in acute myeloid leukemia. Functionally, PEAK3 overexpression in U2OS sarcoma cells enhanced their growth and migratory properties, while its silencing in THP1 leukemic cells reduced these effects. Importantly, an intact SHED module was required for these PEAK3 oncogenic activities. Mechanistically, through a phosphokinase survey, we identified PEAK3 as a novel inducer of AKT signaling, independent of growth-factor stimulation. Then, proteomic analyses revealed that PEAK3 interacts with the signaling proteins GRB2 and ASAP1/2 and the protein kinase PYK2, and that these interactions require the SHED domain. Moreover, PEAK3 activated PYK2, which promoted PEAK3 tyrosine phosphorylation, its association with GRB2 and ASAP1, and AKT signaling. Thus, the PEAK1-3 pseudo-kinases may use a conserved SHED-dependent mechanism to activate specific signaling proteins to promote oncogenesis.


Introduction
The human kinome includes >50 pseudo-kinases that are predicted to be catalytically inactive due to the lack of important residues required for full enzymatic activity [1]. Their mechanistic role in cell signaling remains unclear, but recent structural analyses suggest a scaffolding or allosteric activity by docking additional kinases for efficient protein phosphorylation [1][2][3]. Moreover, some pseudo-kinases have retained active kinase activity through an unconventional mechanism of protein phosphorylation [1][2][3]. These atypical kinases have gained recent interest because they play important roles as active protein kinases in human cancer. For instance, many pseudo-kinases, such as HER3, TRIBL, and JAK2 (named JH2), are overexpressed or mutated in various cancer types and contribute to tumor progression [1][2][3].
Crystallographic studies revealed an important mechanism by which PEAK1/2 pseudo-kinases may regulate oncogenic pathways [8,18,19]. In addition to a classical protein kinase fold harboring an occluded ATP binding site, three crystal structures revealed an original dimerization domain, i.e., split helical dimerization (SHED) [8,18,19]. This conserved module comprises a long helix N-terminal to the pseudo-kinase domain and three additional helices from the C-terminal extension that form an "XL"-shaped helical bundle. The SHED-based dimerization mechanism regulates PEAK1/2 homo-and hetero-dimerization that might influence PEAK-signaling specificity [20]. However, it is not known whether the SHED domain defines an essential mechanism of PEAK signaling.
PEAK3 was identified as a member of the PEAK family, due to the fact that its peptide sequence predicts a similar degraded ATP-binding site in the TK flanked by the SHED module [8] (Figure 1a). Moreover, PEAK3 self-associates in a SHED-dependent manner [21], but its biological activity is poorly characterized. Recently, Lopez et al. reported that PEAK3 expression can prevent CRKII-dependent membrane ruffling [21], suggesting a negative function in cell migration, dissimilar to PEAK1/2. In this study, we investigated PEAK3's functional role in cancer. We found that PEAK3 promotes cancer cell growth and migration. Moreover, the SHED module has an important function in these effects through a unique PEAK3/PYK2/AKT signaling cascade. Our findings support the existence of a common SHED-dependent mechanism of PEAK1-3's oncogenic activity, mediated by the recruitment of a unique set of signaling proteins. The split helical dimerization (SHED) module (made of αN, αJ, αK, and αN helices) (purple), the pseudo-kinase (φ-Kinase) (green), and the main known tyrosine phosphorylation sites are highlighted. (b) Phylogenetic cladogram of PEAK proteins. The tree was deduced from a multiple amino acid sequence alignment of kinase and pseudo-kinase domains and processed by PhyML and MrBayes analysis. PTEN-induced kinases (PINK) were included as an external group to root the tree. Only supports for key phylogenetic nodes are indicated (bootstrap proportion/posterior probability). Species and sequences (accession numbers) used for the alignment are listed in Table S1. (c) Analysis of the non-synonymous to silent substitution ratios (ω) of the SHED and pseudo-kinase (ΨK) domains in primate PEAKs. The nucleic sequences of the domains were aligned based on translation and processed with codeml to estimate the ω values. The phylogenetic tree on the left was generated by PhyML analysis of a multiple alignment of the ΨK amino acid sequences. Species and sequences (accession numbers) used for the alignment are listed in Table S1. The split helical dimerization (SHED) module (made of αN, αJ, αK, and αN helices) (purple), the pseudo-kinase (ϕ-Kinase) (green), and the main known tyrosine phosphorylation sites are highlighted. (b) Phylogenetic cladogram of PEAK proteins. The tree was deduced from a multiple amino acid sequence alignment of kinase and pseudo-kinase domains and processed by PhyML and MrBayes analysis. PTEN-induced kinases (PINK) were included as an external group to root the tree. Only supports for key phylogenetic nodes are indicated (bootstrap proportion/posterior probability). Species and sequences (accession numbers) used for the alignment are listed in Table S1. (c) Analysis of the non-synonymous to silent substitution ratios (ω) of the SHED and pseudo-kinase (ΨK) domains in primate PEAKs. The nucleic sequences of the domains were aligned based on translation and processed with codeml to estimate the ω values. The phylogenetic tree on the left was generated by PhyML analysis of a multiple alignment of the ΨK amino acid sequences. Species and sequences (accession numbers) used for the alignment are listed in Table S1.

PEAK3 Evolutionary Conservation
Unlike the human genome (three PEAK genes) (Figure 1a), we found a single PEAK gene in the genomes of early metazoan species, from placozoans (Trichoplax adherens) to early chordates (the lancelet Branchiostoma) [21] (Figure 1b) [8]. To better understand the PEAK family ontology, we searched for homologous sequences in key evolutionary species. We identified two PEAK members in cyclostomes (the lamprey, Petromyzon marinus, and the hagfish, Eptatretus burgeri) and three members in cartilaginous fishes (e.g., the elephant shark, Callorhinchus milii), bony fishes (e.g., Takifugu rubripes), and the coelacanth, Latimeria chalumnae. We then extracted the three homologous sequences in additional Tetrapoda species and performed a clustering analysis of MSAs using two probabilistic algorithms (maximum-likelihood and Bayesian). We used the related PTEN-induced kinases (PINK) as the external group to root the tree. The clustering tree globally followed the species phylogeny and showed the presence of a well-supported node leading to three PEAK1-3 clusters in vertebrates (Figure 1b). The branching of the three nodes suggests that a first duplication led to PEAK1/2 and PEAK3, followed by the PEAK1/2 duplication into PEAK1 and PEAK2. This shows that PEAK1-3 emerged at the onset of vertebrates and not in Sauropsida as previously proposed [21]. The presence of only PEAK1-like proteins in cyclostomes suggests that they lost PEAK2 and PEAK3. However, the lower conservation of PEAK3, as indicated by the longer internal branches, may have blurred the branching of the three groups to such an extent that the exact duplication order can no longer be deduced.
To firmly establish the lower conservation of PEAK3 compared with PEAK1/2, we examined the selective constraints exerted on the SHED and pseudo-kinase (ΨK) domains. We calculated the dN/dS ω ratio from a MSA of seven primate species that covered 74 MY of evolution. Neutral selection is associated with ω values close to 1, while values below or above 1 indicate purifying and diversifying selection, respectively. The selection exerted on the SHED and ΨK domains was lower for PEAK3 than for PEAK1 and PEAK2 (Figure 1c), which is in agreement with the difference in branch length. The lowest PEAK3 conservation is also consistent with with the observation that the corresponding gene was lost in several taxa, for instance Myomorpha in rodents (Table S1). The maintenance of duplicated genes is often associated with neo-or sub-functionalization (i.e., the acquisition of new functions or differential tissue expression, respectively) [22]. Then, we investigated PEAK1-3 mRNA expression using the Human Protein Atlas [23] and found that PEAK1 was ubiquitously expressed, particularly in spleen; PEAK2 expression was more restricted, with the highest expression in cerebellum; and PEAK3 was nearly exclusively expressed in lymphoid tissues (spleen, lymph nodes) and granulocytes ( Figure S1). Overall, these data indicate that, although PEAK3 is much less evolutionary conserved than its paralogues PEAK1 and PEAK2, it underwent purifying selection indicative of a physiological function, possibly in blood cells.

PEAK3 Overexpression in Acute Myeloid Leukemia
Next, to investigate PEAK3 protein expression in human cells, we generated a polyclonal antibody against full-length recombinant PEAK3. This antibody could detect PEAK3 in lysates from PEAK3-overexpressing U2OS and HeLa cells in western blotting ( Figure  S2a,b). It also recognized an endogenous protein of about 50 KDa in lysates of THP1 (AML) cells ( Figure S2c). We confirmed that this protein was PEAK3, because the 50 KDa band disappeared upon siRNA-mediated PEAK3 knockdown ( Figure S2c). We then used this antibody to assess PEAK3 expression in various human cell lines. We did not detect endogenous PEAK3 in the mesenchymal and epithelia cell lines tested ( Figure S2c), but only in hematopoietic cell lines, specifically in AML cell lines (Figure 2a and Figure S2c). Consistent with this finding, we observed elevated PEAK3 protein expression in 40% of the tested AML patient samples (8/20) (Figure 2b and Table S2). Analysis of the PEAK3 transcript level from TCGA transcriptomic data (gepia.cancer-pku.cn accessed on 2 January 2021) confirmed that PEAK3 was upregulated in AML (Figure 1c), particularly in the M4 and M5 subtypes ( Figure S3a). PEAK3 aberrant expression was preferentially associated with specific oncogenic mutations, i.e., DNTM3A mutations, but not with FLT3-TD mutations (Figure 2b and Figure S3b). Our findings indicate that PEAK3 is expressed in hematopoietic cells and that its expression can be deregulated in AML, suggesting a pro-tumor function.
Cancers 2021, 13, x 5 of 20 transcript level from TCGA transcriptomic data (gepia.cancer-pku.cn accessed on 2 January 2021) confirmed that PEAK3 was upregulated in AML (Figure 1c), particularly in the M4 and M5 subtypes ( Figure S3a). PEAK3 aberrant expression was preferentially associated with specific oncogenic mutations, i.e., DNTM3A mutations, but not with FLT3-TD mutations (Figures 2b and S3b). Our findings indicate that PEAK3 is expressed in hematopoietic cells and that its expression can be deregulated in AML, suggesting a pro-tumor function.

SHED-Dependent Oncogenic Activity of PEAK3
We then investigated whether PEAK3 had pro-tumor functions. PEAK3 overexpression in U2OS cells ( Figure 3a) and in THP1 cells ( Figure 3c) increased cell growth and invasion in Matrigel-coated Boyden chambers (Figure 3b). These effects were abrogated in both cell lines by overexpression of PEAK3 harboring the A436E mutation in the C-

SHED-Dependent Oncogenic Activity of PEAK3
We then investigated whether PEAK3 had pro-tumor functions.  [21]. Similarly, siRNA-mediated PEAK3 silencing in THP1 cells reduced their growth and migration in Boyden chambers coated with an endothelial cell monolayer (Figure 3d and S5). These findings indicate that, in human cancer cells, PEAK3 displays oncogenic activity that requires an intact SHED module. As PEAK1/2 localize at focal Cancers 2021, 13, 6344 6 of 18 adhesions (FAs) [7,13], we investigated PEAK3 localization by direct fluorescence analysis of U2OS cells that stably express GFP-PEAK3 and demonstrated PEAK3 localization at FAs ( Figure S6a) and partial co-localization with paxillin ( Figure S6a). We confirmed PEAK3 localization at FAs by indirect immunofluorescence using an anti-HA antibody in U2OS cells that stably express HA-ST-PEAK3 ( Figure S6b). Additionally, both imaging methods showed a strong PEAK3 nuclear localization ( Figure S6). Conversely, the PEAK3 A436E mutant did not display FA localization, but was still localized in the nucleus ( Figure S6a,b). These observations indicate that PEAK3 localizes at FAs such as PEAK1/2, and that this requires a SHED dimeric module.

PEAK3 Activates AKT Signaling
Next, to determine PEAK3 oncogenic signaling, we probed a phospho-kinase antibody array from THP1 cells transfected with a control siRNA or a siRNA targeting PEAK3 (Figure 4a). We found that PEAK3 downregulation affected the activating phosphorylation of adhesive TKs of the Src (SFK) and FAK families ( Figure S7a) and also reduced MAPK and AKT activating phosphorylations (Figure 4a and Figure S7a). We confirmed these results using western blotting (Figure 4a, right panel). PEAK3 silencing also reduced phosphorylation of the probed AKT substrates (i.e., GSK-3 alpha/beta, PRAS40, and CREB), suggesting that this pseudo-kinase is a novel upstream regulator of this pathway ( Figure S7a). Consistently, AKT activity (i.e., pS473-AKT level) increased approximately 2-fold upon PEAK3 overexpression and reduced 3-fold upon PEAK3 knockdown (Figure 4b). We obtained similar results in PEAK3-overexpressing U2OS cells, suggesting that PEAK3 might be a general inducer of AKT signaling (Figure 4c). PEAK3 overexpression induced a modest increase in AKT signaling (i.e., AKT activity and AKT substrate phosphorylation). Conversely, PEAK3 strongly activated AKT in serum-starved conditions (Figure 4c). We did not observe this effect with the PEAK3 A436E mutant, further supporting the essential role of the SHED module in PEAK3 signaling (Figure 4c). PEAK3-AKT signaling was abrogated by pharmacological inhibition of PI3K activity ( Figure S7b), suggesting that PEAK3 is upstream to PI3K. Collectively, these data indicate that PEAK3 induces PI3K/AKT signaling, even in the absence of growth factors.

SHED-Dependent PEAK3 Scaffolding Activity
To further elucidate the mechanism by which PEAK3 induces AKT signaling, we performed PEAK3 interactomic analyses of U2OS, HeLa, and THP1 cells that stably express ST-PEAK3 (Figure 5a and Figure S4b and Table S3). Label-free LC-MS/MS analysis of ST-PEAK3 complexes purified on streptavidin beads identified many signaling proteins in the three PEAK3 interactomes, including GRB2, ADP-ribosylation factor (ARF) GTPaseactivating proteins ASAP1 and 2, and another member of the PEAK family (i.e., PEAK1 in the HeLa and U2OS cell interactomes and PEAK2 in the THP1 cell interactome). This proteomic analysis also identified many 14-3-3 proteins, which regulate phospho-dependent signaling activity [24] and voltage-dependent anion-selective channel proteins (i.e., VDAC1-3), which regulate cell volume and apoptosis (Table S3) [25]. Additionally, we detected the FAK-like TK PYK2 in the PEAK3 interactome of THP1 cells (Table S3). We next confirmed some of these findings using western blotting (Figure 5b,c). Although the CRKII cytoskeletal proteins were previously reported as PEAK3 binders [21], we could not detect any endogenous CRKII protein associated with PEAK3 in our proteomic and biochemical analyses (Figure 5c and Table S3). All these interactions required an intact SHED module (Figure 5c), suggesting that PEAK3 dimerization is essential for interaction with these signaling proteins. As ASAP1/2, VDACs, and PYK2 have not been detected in the previously reported PEAK1 and PEAK2 interactomes [8,20], PEAK1-3 may recruit a unique set of signaling proteins to induce intracellular signaling.

PEAK3 Activates PYK2 to Promote AKT Signaling
The fact that PEAK3 interacts with PYK2 suggests that, like PEAK2 [4,8], PEAK3 may activate specific TKs to mediate phospho-tyrosine signaling. We tested this hypothesis by co-expressing PEAK3 and PYK2 in HEK293T cells. While PEAK3 alone did not clearly affect protein tyrosine phosphorylation, PEAK3 robustly activated ectopic PYK2, as measured on the phosphorylation level of regulatory tyrosine 402 and 881 (Figure 6a,c) [26].

PEAK3 Activates PYK2 to Promote AKT Signaling
The fact that PEAK3 interacts with PYK2 suggests that, like PEAK2 [4,8], PEAK3 may activate specific TKs to mediate phospho-tyrosine signaling. We tested this hypothesis by co-expressing PEAK3 and PYK2 in HEK293T cells. While PEAK3 alone did not clearly affect protein tyrosine phosphorylation, PEAK3 robustly activated ectopic PYK2, as measured on the phosphorylation level of regulatory tyrosine 402 and 881 (Figure 6a,c) [26]. Interestingly, PYK2 activation was accompanied by an increase in cellular protein tyrosine phosphorylation, including a 50 KD and 120 KDa band (Figure 6a). Pull-down experi-ments next identified PEAK3 and PYK2 some of these tyrosine phosphorylated proteins (Figure 6b). Similar results were obtained in THP1cells where tyrosine phosphorylation of overexpressed PEAK3 was reduced upon acute pharmacological inhibition of PYK2 (PYK2i) (Figure 6c,d). Additionally, the pharmacological inhibition of protein tyrosine phosphatases (pervanadate) largely increased PEAK3 tyrosine phosphorylation (Figure 6c), suggesting the existence of regulatory protein tyrosine phosphatases. Consequently, PEAK3 tyrosine phosphorylation by PYK2 increased its association with GRB2 and ASAP1 (Figure 6b), while PYK2i reduced this effect (Figure 6c,d). ASAP1 was previously identified as a PYK2 substrate [27] and, consistent with our findings, overexpressed PEAK3 increased ASAP1 tyrosine phosphorylation in a PYK2-dependent manner (Figure 6e). Finally, this molecular process required an intact SHED module ( Figure S8), suggesting that PEAK3 dimerization is essential for PEAK3 phospho-tyrosine signaling. Overall, these findings support a tyrosine phosphorylation-dependent PEAK3 scaffolding activity, which implicates the TK PYK2.
Interestingly, PYK2 activation was accompanied by an increase in cellular protein tyrosine phosphorylation, including a 50 KD and 120 KDa band (Figure 6a). Pull-down experiments next identified PEAK3 and PYK2 some of these tyrosine phosphorylated proteins (Figure 6b). Similar results were obtained in THP1cells where tyrosine phosphorylation of overexpressed PEAK3 was reduced upon acute pharmacological inhibition of PYK2 (PYK2i) (Figure 6c,d). Additionally, the pharmacological inhibition of protein tyrosine phosphatases (pervanadate) largely increased PEAK3 tyrosine phosphorylation ( Figure  6c), suggesting the existence of regulatory protein tyrosine phosphatases. Consequently, PEAK3 tyrosine phosphorylation by PYK2 increased its association with GRB2 and ASAP1 (Figure 6b), while PYK2i reduced this effect (Figure 6c,d). ASAP1 was previously identified as a PYK2 substrate [27] and, consistent with our findings, overexpressed PEAK3 increased ASAP1 tyrosine phosphorylation in a PYK2-dependent manner ( Figure  6e). Finally, this molecular process required an intact SHED module ( Figure S8), suggesting that PEAK3 dimerization is essential for PEAK3 phospho-tyrosine signaling. Overall, these findings support a tyrosine phosphorylation-dependent PEAK3 scaffolding activity, which implicates the TK PYK2.  This PEAK3 activating mechanism on PYK2 was then confirmed in PEAK3 overexpressing U2OS cells (Figure 7a). Importantly, this molecular effect was further enhanced in serum-starved conditions (Figure 7a), also supporting a growth factor-independent mechanism of PEAK3 activation of PYK2. PEAK3 knockdown reduced PYK2 activity in THP1 cells (Figure 7b), supporting the existence of a similar endogenous PEAK3-PYK2 signaling in leukemic cells. Having demonstrated the existence of PEAK3-AKT and PEAK3-PYK2 signaling, we then asked whether PYK2 would mediate the growth-independent AKT activation by PEAK3. Acute pharmacological inhibition of PYK2 (PYK2i) in U2OS reduced the stimulating effect of PEAK3 on AKT 3-fold, while this inhibitor had no effect on AKT activity in control cells (Figure 7c). Overall, these findings demonstrate the fact that that PEAK3 activates PYK2 as well as the support role of PYK2 activity in PEAK3-AKT signaling.
by PEAK3 expression in a PYK2-dependent manner. Immunoprecipitated ASAP1 tyrosine phosphorylation and its association with ST-PEAK3 from indicated THP1 cells treated as described in Figure 6c (n = 2). This PEAK3 activating mechanism on PYK2 was then confirmed in PEAK3 overexpressing U2OS cells (Figure 7a). Importantly, this molecular effect was further enhanced in serum-starved conditions (Figure 7a), also supporting a growth factor-independent mechanism of PEAK3 activation of PYK2. PEAK3 knockdown reduced PYK2 activity in THP1 cells (Figure 7b), supporting the existence of a similar endogenous PEAK3-PYK2 signaling in leukemic cells. Having demonstrated the existence of PEAK3-AKT and PEAK3-PYK2 signaling, we then asked whether PYK2 would mediate the growth-independent AKT activation by PEAK3. Acute pharmacological inhibition of PYK2 (PYK2i) in U2OS reduced the stimulating effect of PEAK3 on AKT 3-fold, while this inhibitor had no effect on AKT activity in control cells (Figure 7c). Overall, these findings demonstrate the fact that that PEAK3 activates PYK2 as well as the support role of PYK2 activity in PEAK3-AKT signaling.

Discussion
In this study, we show that the pseudo-kinase PEAK3 has an oncogenic function, as previously reported for PEAK1 and 2. PEAK3 overexpression enhanced cancer cell growth and migration, while its silencing reduced these pro-tumor effects. The PEAK3 expression profile suggested a restricted function in myeloid cells, such as granulocytes, and our results reveal a pro-tumoral function in AML. This finding was supported by the aberrant PEAK3 expression observed in AML patient samples and by the finding that endogenous PEAK3 is implicated in the regulation of THP1 leukemic cell features. As PEAK1/2 can mediate TK signaling induced by growth factor receptors [4,5], the absence of association between PEAK3 and the oncogenic receptor TK FLT3-TD in AML [28] is surprising. This suggests that PEAK3 may not be a critical component of the receptor TK signaling or that it displays redundant signaling functions. This hypothesis is corroborated by the finding that PEAK3 activates AKT signaling in the absence of growth factors. Moreover, the heterogeneous PEAK3 expression in AML patient samples suggests a PEAK3 function in a specific AML subtype that needs to be thoroughly characterized. The association between PEAK3 expression and DNTM3A [28] is consistent with this hypothesis.
This study also uncovers an essential role of the SHED module in PEAK3 biological activities. Specifically, PEAK3 dimerization was necessary for PEAK3 pro-tumoral activity, intracellular signaling, and localization to FAs. PEAK3 dimerization was also essential for interaction with important signaling proteins, uncovering a general scaffolding mechanism regulated by this dimeric domain to allow protein interactions. As this domain is conserved among PEAK family members, we might predict a similar function in PEAK1 and PEAK2 signaling that needs to be confirmed. Liu et al. reported a hetero-dimerization mechanism involved in PEAK1/2 signaling specificity [20]. We did not address whether this mechanism contributes also to PEAK3 signaling in this study. Nevertheless, the presence of another PEAK member in our PEAK3 interactomes suggests that heterodimers may contribute to PEAK3 signaling. However, as PEAK3 was the main PEAK member expressed in blood cells ( Figure S1), PEAK3 signaling may occur primarily via PEAK3 homo-dimerization. Our imaging analysis revealed that nuclear PEAK3 localization is not affected by protein dimerization. This suggests that the SHED module does not regulate all PEAK3 functions, and that PEAK3 has an additional nuclear function, as described for PEAK2 [29], which remains to be characterized.
One important finding from our molecular study is the identification of PEAK3 as a novel upstream activator of AKT signaling, which plays an essential role in AML development [30]. This result corroborates a previous report showing that PEAK3 is a potential regulator of AKT signaling using a shRNA screening approach [31], although this observation was not further investigated. Our results also suggest that PEAK3 might define an important mechanism of AKT activation in the absence of growth factors. This mechanism could be essential for malignant cell survival and dissemination during disease progression. Our interactomic analysis also showed that PYK2 is an important mediator of this signaling cascade. This is consistent with FAK-like functions described in AML, where high FAK expression level is associated with enhanced blast migration, increased cellularity, and poor prognosis [32]. Mechanistically, our findings support a tyrosine phosphorylation-dependent PEAK3 signaling implicating the TK PYK2 ( Figure S9). For instance, PEAK3 may act a scaffold to bring PYK2 and ASAP1 together for efficient ASAP1 tyrosine phosphorylation, enabling AKT signaling [33][34][35]. Moreover, our results suggest that PEAK3 activates PYK2 through a SHED-dependent mechanism, as previously reported for PEAK2 activation of CSK [8]. Therefore, we propose that, in addition to its scaffolding function, PEAK3 and PEAK2 use an allosteric mechanism to activate associated TKs for signaling. Since PYK2 is regulated by dimerization [26], its binding to PEAK3 might facilitate this activation mechanism. During the preparation of this manuscript, the Daly group reported a similar PEAK3 oncogenic function in breast epithelial cells and showed that PEAK3 tyrosine phosphorylation promotes GBR2 and ASAP1 binding [35], which is highly consistent with our proposed mechanism of PEAK3 oncogenic signaling.

Phylogenetic Analyses and Data Mining
Nucleic and protein PEAK sequences were retrieved from the NCBI annotated nonredundant database (http://www.ncbi.nlm.nih.gov accessed on 15 September 2020) using NCBI PHI-BLAST, BLAST, and the Annotation search tools available in the Geneious 11.1.5 software package (Biomatters, http://www.geneious.com accessed on 15 September 2020). Accession numbers are listed in Table S1. Protein sequences were aligned using MAAFT v7.450 [41]. Protein multiple sequence alignments (MSA) were processed using BMGE (Block Mapping and Gathering with Entropy) [42] and a cut-off value = 0.6. Phylogenetic trees were estimated using PhyML [43], using the General Time Reversible model with invariant and gamma decorations, and MrBayes [44] with two independent runs of four Markov chains for 1,000,000 generation samples for every 200 generations. Sample trees from the first 100,000 generations were discarded as "burn-in". At the end of the two runs, the split frequency standard deviations were~1.5 × 10 −3 , and the effective sample sizes were >2600. The non-synonymous to silent substitution ratios (ω = dN/dS) were calculated using PAML 4.4 [45]. Nucleic MSAs were based on protein alignments. The "one-ratio" (i.e., a single ω ratio for the entire tree) and the two "two-ratio" models (i.e., distinct ω values for two or three PEAK branches) were compared using the likelihood-ratio test. Transcriptomic signatures and data mining from two PEAK3 signatures were generated from the transcriptomes of the acute myeloid leukemia (AML) patient samples, high versus low PEAK3 expression from TCGA transcriptomic data (gepia.cancer-pku.cn), and the BEATAML database (GSE6891) and correlated with well-characterized oncogenic somatic mutations.

Human Samples
Primary AML samples from patients (with annotated clinical and biological characteristics, treatments, and outcome data) were obtained from the Biobank of the Hematology Clinical Department, Toulouse University Hospital, France, (HIMIP collection, BB-0033-00060; INSERM-U1037) headed by Prof C. Récher. The recording in the database of clinical and biological data has been declared to the CNIL (Comité National Informatique et Libertés; the French Committee for the protection of personal data). Samples are routinely characterized by morphologic, immunophenotypic, cytogenetic, and molecular (FLT3-ITD, NPM1, CEBPA, DNMT3A, KIT, IDH1, and IDH2 mutations) analyses.
The Proteome Profiler Human Phospho-Kinase Array Kit was purchased from R&D Systems. THP1 cells were lysed, and 300 µg of protein lysates was used for western blotting according to the manufacturer's protocol. Signals were quantified with the Amersham Imager 600 (GE Healthcare, Chicago, IL, USA) from two independent biological replicates.
Interactomic analyses were performed as described in [48]. Cell lysates were incubated with streptavidin beads (Thermo Fisher Scientific, Waltham, MA, USA) overnight and washed three times with a lysis buffer and once with a washing buffer (50 mM Tris pH 7.5 and 50 mM NaCl). PEAK3 complexes were eluted with 2.5 mM biotin in 50 mM Tris pH 8 and 0.5 mM EDTA and processed for mass spectrometry (MS) analysis. Peptides obtained after digestion were analyzed using a Qexactive-HFX system coupled with a RSLC-U3000 nano HPLC. Samples were desalted and pre-concentrated online on a Pepmap ® precolumn (0.3 mm × 10 mm). A gradient consisting of 6-25% B for 100 min, 25-40% B for 20 min, and 40-90% B for 2 min (A = 0.1% formic acid; B = 0.1% formic acid in 80% acetonitrile) at 300 nl/min was used to elute peptides from the capillary reversephase column (0.075 mm × 250 mm; Pepmap ® , Thermo Fisher Scientific) fitted with a stainless steel emitter (Thermo Scientific). Spectra were acquired with the instrument operating in data-dependent acquisition mode throughout the HPLC gradient. MS scans were acquired with resolution set at 60,000. Up to twelve of the most intense ions per cycle were fragmented and analyzed using a resolution of 300,000. Peptide fragmentation was performed using nitrogen gas on the most abundant and at least doubly-charged ions detected in the initial MS scan and a dynamic exclusion time of 20 s. Analysis was performed using the MaxQuant software (version 1.5.5.1, Max Planck Institute of Biochemistry, Munich, Germany). All MS/MS spectra were searched using Andromeda against a decoy database consisting of a combination of the Homo sapiens reference proteome (release 4 February 2019 www.uniprot.org) and of 250 classical contaminants, containing forward and reverse entities. A maximum of two missed cleavages was allowed. The search was performed with oxidation (M) and acetyl (protein N-term) as variable modifications and carbamidomethyl (C) as a fixed modification. FDR was set at 0.01 for peptides and proteins, and the minimum peptide length was 7. The t-test and graphical representation were performed with Perseus (v 1.6.10.43, Max Planck Institute of Biochemistry, Munich, Germany) and standard parameters.

Cell Imaging
Cells seeded on microscope coverslips were fixed with 4% paraformaldehyde in a Triton-X100-BRB80 buffer (0.5% Triton-X100, 80 mM PIPES pH 6.8, 1 mM MgCl 2 , and 1 mM EGTA) containing 5% FCS at 37 • C for 20 min. Cells were incubated with the indicated primary antibodies and the corresponding secondary antibodies labelled with Alexa Fluor dyes (488 or 594 nm) and DAPI diluted in PBS buffer. After mounting with a ProLong Antidafe Mountant (Thermo Fisher), cells were observed with a DMR A microscope and a PL APO 63× oil objective (1.32 NA). Images were captured using a piezzo stepper (E662 LVPTZ amplifier; Servo) and a cooled CCD Micromax camera (1300 × 1030 pixels, RS; Princeton Instruments Inc. Trenton, NJ, USA) driven by MetaMorph (v 4.5; Molecular Devices San Jose, CA, USA). Images were processed with ImageJ (https://imagej.nih.gov/ ij/ accessed on 6 June 2021).

Statistical Analysis
All statistical analyses were performed on data from at least three independent experiments with GraphPad Prism (Version 9, GraphPad Software, San Diego, CA, USA). Data are presented as the mean ± standard deviation (SD). When distribution was normal (assessed with the Shapiro-Wilk test), the two-tailed t test was used for between-group comparisons. Otherwise, the Mann-Whitney test was used.

Conclusions
Our study provides a unifying model on how PEAK pseudo-kinases regulate oncogenic signaling. GRB2 and possibly CRK proteins are common PEAK1-3 binders [8,20,21], through a conserved proline-rich sequence at their N-terminus. Moreover, our interactomic analyses identified a unique set of signaling proteins recruited by PEAK3 that were not detected in previous PEAK1/2 interactomes [8,20]. Altogether, these findings support the existence of a common SHED-dependent mechanism of PEAK1-3 oncogenic activity mediated through the recruitment of a unique set of signaling proteins to induce specific oncogenic functions. Lopez et al. found that the conserved DFG motif in the PEAK3 pseudo-kinase domain is implicated in PEAK3 dimerization [21]. This raises an interesting idea that, besides the SHED module, the pseudo-kinase domain might also participate in PEAK3 signaling. Consistent with this idea, recent molecular studies have shown that, despite the inaccessible ATP site found in crystallography, pseudo-kinases display higher kinase flexibility than previously expected to enable signaling [2,49,50]. Therefore, small molecules that affect pseudo-kinase dynamics may inhibit their function [49,50]. Whether PEAK1-3 signaling is regulated by a similar mechanism deserves investigation. In this case, these pseudo-kinases could represent attractive therapeutic targets in human cancers, including leukemia.