EDC3 phosphorylation regulates growth and invasion through controlling P‐body formation and dynamics

Abstract

Regulation of mRNA stability and translation plays a critical role in determining protein abundance within cells. Processing bodies (P‐bodies) are critical regulators of these processes. Here, we report that the Pim1 and 3 protein kinases bind to the P‐body protein enhancer of mRNA decapping 3 (EDC3) and phosphorylate EDC3 on serine (S)161, thereby modifying P‐body assembly. EDC3 phosphorylation is highly elevated in many tumor types, is reduced upon treatment of cells with kinase inhibitors, and blocks the localization of EDC3 to P‐bodies. Prostate cancer cells harboring an EDC3 S161A mutation show markedly decreased growth, migration, and invasion in tissue culture and in xenograft models. Consistent with these phenotypic changes, the expression of integrin β1 and α6 mRNA and protein is reduced in these mutated cells. These results demonstrate that EDC3 phosphorylation regulates multiple cancer‐relevant functions and suggest that modulation of P‐body activity may represent a new paradigm for cancer treatment.

EDC3 is highly phosphorylated in tumors, thereby preventing P‐body assembly. Inhibiting EDC3 phosphorylation decreases tumor growth and invasion, suggesting key roles for the regulation of RNA decapping and destruction by EDC3 in cancer progression.

Introduction

Processing bodies (P‐bodies) are conserved, membrane‐less cytoplasmic assemblies of non‐translating mRNA protein complexes (mRNPs) (Anderson & Kedersha, 2009; Jain & Parker, 2013). P‐bodies harbor key proteins involved in mRNA decay and translational control and are strongly implicated in regulating both processes (Luo et al, 2018; Standart & Weil, 2018). mRNAs can be both stored and degraded in P‐bodies (Sheth & Parker, 2003; Teixeira et al, 2005; Hubstenberger et al, 2017; Wang et al, 2018). Initial studies suggested that the activity of P‐bodies in storing and degrading RNAs could account for 40% of the variation in the level of proteins when compared to mRNA levels. However, more recent studies have indicated that this variation is actually highly context‐dependent (Vogel et al, 2010; Vogel & Marcotte, 2012; Liu et al, 2016). P‐body assembly and their protein/mRNA composition also change rapidly in response to changing cellular environments, thus suggesting P‐bodies as a key post‐transcriptional regulators of gene expression (Teixeira et al, 2005; Eulalio et al, 2007b; Standart & Weil, 2018). Interestingly, P‐bodies are potently induced by certain chemotherapeutic agents (Rzeczkowski et al, 2011; Kaehler et al, 2014), and while recent studies have linked tumorigenicity to alterations in stress granules, related cytoplasmic mRNP granules; (Anderson et al, 2015; Somasekharan et al, 2015), the role of P‐bodies and regulation of P‐body mRNA content in cancer progression remains an intriguing but unstudied area. P‐body mRNA content has been recently analyzed in human and yeast cells (Hubstenberger et al, 2017; Di Stefano et al, 2019; Matheny et al, 2019). To date, a clear consensus on the types of mRNAs and how they are regulated has not emerged. Nonetheless, P‐body‐enriched mRNAs have been argued to encode various regulatory proteins that control cell division, chromatin state, morphogenesis, differentiation, and stem cell pluripotency (Hubstenberger et al, 2017; Di Stefano et al, 2019).

P‐bodies contain most of the core 5′–3′ mRNA decay pathway including the decapping complex (DCP2, DCP1a/b), many decapping enhancers/translation repressors (DDX6/ME31B, PATL1, 4E‐T EDC3/LSM16, EDC4/GE‐1), factors regulating mRNA stability (SYNCRIP, HNRNPU/Q, DHX9), and the 5′–3′ exonuclease XRN1 (Sheth & Parker, 2003; Anderson & Kedersha, 2006, 2009; Eulalio et al, 2007a; Balagopal & Parker, 2009; Jain & Parker, 2013; Aizer et al, 2014; Luo et al, 2018). Numerous other mRNA‐binding proteins involved in translational control and mRNA stability, including nonsense‐mediated decay (UPF1, SMG7) and miRNA/siRNA components (AGO2, GW182), also localize in P‐bodies; whether P‐bodies regulate these processes remains controversial (Sheth & Parker, 2003; Eulalio et al, 2007b; Anderson & Kedersha, 2009). Other proteins, including signaling factors, chaperones, and motor proteins, are also enriched in P‐bodies, the latter two of which likely relate to P‐body assembly, disassembly, and cellular mobility (Hubstenberger et al, 2017; Luo et al, 2018; Standart & Weil, 2018). Important issues to decipher in helping to understand the function of these organelles are (i) how P‐body assembly is regulated, (ii) how targeting of specific mRNAs to P‐bodies occurs, and (iii) how mRNA fate (i.e., storage in a repressed state versus decay) within P‐bodies is determined. The core P‐body protein enhancer of mRNA decapping 3 (EDC3), is at the intersection of these questions.

In Saccharomyces cerevisiae, EDC3 is a critical scaffolding protein for P‐body formation by virtue of its ability to dimerize and interact with numerous other P‐body proteins (Decker et al, 2007). EDC3 also stimulates mRNA decapping by relieving the auto‐inhibition of the DCP1/2 decapping complex (Paquette et al, 2018). However, while recognized for an unusual role in promoting deadenylation‐independent decapping and decay of YRA1 pre‐mRNA and mature RPS28B mRNA (Badis et al, 2004; Dong et al, 2010), its role in deadenylation‐dependent decay of other S. cerevisiae mRNAs is less clear. In Schizosaccharomyces pombe, EDC3 also drives P‐body assembly, at least in part via liquid–liquid phase separation (LLPS); this likely occurs in S. cerevisiae as well (Damman et al, 2019). LLPS describes the mechanism of assembly of liquid‐like bodies in cells, such as P‐bodies, whose structure is typically driven by proteins harboring intrinsically disordered regions (IDRs) (Chang et al, 2014), which EDC3 also possesses. This observation suggests that EDC3's IDR is potentially critical to the regulation of P‐body formation.

EDC3 protein structure is highly conserved throughout eukaryotes and consists of an N‐terminal LSm domain, an IDR, a central FDF domain, and a C‐terminal YjeF‐N domain (Tritschler et al, 2007; Ling et al, 2008). The LSm domain is thought to modulate DCP1a and DCP2 binding and Edc3’s P‐body localization (Fromm et al, 2012), whereas the FDF domain can interact with and is sufficient to bind the RNA helicase Dhh1/DDX6/Me31B (Decker et al, 2007; Tritschler et al, 2009a; Fromm et al, 2012), which plays a critical role in repressing translation and also stimulates yeast and mammalian P‐body assembly (Tritschler et al, 2009a; Rao & Parker, 2017). Finally, the C‐terminal YjeF‐N domain controls EDC3 dimerization (Decker et al, 2007; Ling et al, 2008), while the IDR of EDC3 was shown to be vital for EDC3‐RNA interactions that aid in LLPS assembly (Schutz et al, 2017). The IDR of EDC3 interacts with its YjeF_N dimerization domain and leads to an irreversible LLPS maturation process (Schutz et al, 2017). Despite such knowledge, the extent to which EDC3 functions are regulated, possibly by post‐translational modifications, is poorly understood.

Here we report that the IDR of EDC3 binds tightly to the Pim1 and 3 protein kinases which in turn phosphorylate EDC3 on S161. This phosphorylation prevents the localization of EDC3 to P‐bodies. Prostate cancer cells harboring the S161A mutation have diminished tumor cell migration, invasion, and growth in mouse xenograft models, pointing to the regulatory importance of this post‐translational modification. Treatment of tumor cells with protein kinase inhibitors blocks EDC3 phosphorylation and alters the number of P‐bodies. Thus, EDC3 phosphorylation is a crucial regulator of multiple cellular functions likely through the modulation of P‐bodies.

Results

EDC3 physically interacts with the Pim1 and Pim3 protein kinases

The mechanism by which EDC3 activity is regulated and whether this modification plays a role in controlling P‐body function is unknown. Because protein interactions could affect the activity of EDC3, we evaluated current databases (e.g., BioGRID) and identified the Pim1 protein kinase as a potential EDC3 binding partner (Huttlin et al, 2015; Huttlin et al, 2017). Because the family of Pim kinases have been implicated in stimulating the growth and progression of multiple malignancies, including leukemia, prostate, and triple‐negative breast cancer (TNBC) (Chen et al, 2005; Cen et al, 2013; Horiuchi et al, 2016; Braso‐Maristany et al, 2017; Padi et al, 2017), we evaluated whether Pim1 could modulate EDC3 function and by so doing regulate tumor growth and metastasis.

To confirm if Pim and EDC3 physically interact, we performed in vitro pulldown assays using biotin‐labeled purified Pim1 and recombinant EDC3. The results demonstrate that streptavidin beads co‐precipitated Pim1 and EDC3, while EDC3 alone did not bind to the beads (Fig 1A). To examine this interaction in cells, the three Pim family members, Pim1, 2, and 3, were transfected individually into HEK‐293T cells that stably expressed GFP‐tagged EDC3. GFP‐tagged EDC3 was then immunoprecipitated (IP) using GFP‐nanobody coated beads (GFP‐Trap). EDC3 readily co‐IP'd with Pim1 and Pim3, but not Pim2, suggesting that EDC3 has a different affinity for each Pim isoform (Fig 1B). Pim1 and 3 share a higher degree of sequence homology, 71%, compared to Pim1 and Pim 2 which are only 61% homologous (Nawijn et al, 2011; Alvarado et al, 2012; Saurabh et al, 2014). Pim2, when compared to Pim1 and 3, has several disordered loops in the N‐terminal kinase lobe and, when compared to the other two family members, shows differences in the kinase hinge and P‐loop residues that could be responsible for the EDC3 interaction (Bullock et al, 2005; Qian et al, 2005; Bullock et al, 2009). Thus, Pim1 and 3 protein kinases, but not Pim2, bind EDC3 and can potentially regulate its function.

Figure 1. EDC3 interacts strongly with the Pim protein kinase.

1. Bacterially purified recombinant EDC3 (0.2 μg) was incubated in the presence and absence of biotinylated recombinant Pim1 (0.5 μg) protein with streptavidin‐coated magnetic beads. The beads were precipitated and subjected to Western blotting with EDC3 and Pim1 antibodies (Abs).
2. Pim1, 2, or 3 and GFP‐EDC3 expression vectors were transfected into HEK‐293T. GFP‐Trap Nanobody beads (GFP‐Trap) were used to immunoprecipitate (IP) GFP‐EDC3, and the associated proteins were analyzed by Western blotting with the indicated Abs.
3. Lysate of MDA MB‐231 cells was incubated with Protein G Mag Sepharose with either control IgG or EDC3 primary antibody. The IP products were subjected to Western blotting with the indicated Abs.
4. Pim1 kinase dead (KD) and GFP‐EDC3 plasmids were co‐transfected into HEK‐293T cells. The lysates were precipitated with GFP‐Trap beads. The IP products were subjected to Western blotting with the indicated Abs.

Source data are available online for this figure.

Next we investigated whether EDC3 and Pim1 interact under physiological conditions using the TNBC cell line MDA‐MB‐231 which expresses high levels of Pim1 (Horiuchi et al, 2016; Braso‐Maristany et al, 2017), and EDC3 was able to co‐IP endogenous Pim1, but not AKT (Fig 1C). As a positive control, DDX6, known to interact with EDC3, was also co‐precipitated with EDC3 (Fenger‐Gron et al, 2005; Ayache et al, 2015; Fig 1C). The binding of Pim1 and EDC3 is independent of Pim’s kinase activity as a kinase‐dead (KD) Pim1 and EDC3 co‐IP in HEK‐293T cells (Fig 1D). Similarly, after treatment with a small‐molecule pan‐Pim inhibitor (PIMi) Pim447, Pim1 or Pim3 co‐IP’d with EDC3 in HEK293T cells (Fig EV1A). Unlike EDC3, DCP1a and DDX6 failed to bind Pim1, suggesting Pim1 interacts directly with EDC3 not through a complex with DCP1a and DDX6, two known EDC3 interacting proteins (Fig EV1B). Together these data demonstrate a specific and robust interaction between Pim1/3 and EDC3 independent of kinase activity.

Figure EV1. EDC3 interaction and phosphorylation by Pim and AKT protein kinases.

1. Pim1 or Pim3 plasmids were transfected into HEK‐293T cells stably expressing GFP‐EDC3, and 48 h later, PIMi (PIM447, 3 μM) was added to the cells for 6 h followed by an immunoprecipitation (IP) with GFP‐Trap beads. The whole‐cell lysates and IP products were subjected to Western blotting with the indicated Abs.
2. GFP, EDC3‐GFP, DDX6‐GFP, or DCP1a‐GFP plasmids were transfected into HEK‐293T cells which constitutively overexpress Pim1. After IP with GFP‐Trap, samples were analyzed by Western blotting. Input represents the whole‐cell lysate, while IP denotes the samples analyzed by immunoprecipitation.
3. The Pim3 and each of the EDC3 plasmids encoding various protein domains used in Fig 2B were transfected into HEK‐293T cells. The individual fragments were immunoprecipitated with GFP‐Trap beads and analyzed by Western blotting with the indicated Abs. EDC3 wild type (W), F1, F2, F3, and F4 fragments were fused to GFP. C1 cells contained the GFP Flap vector which expresses GFP Flag and an S‐tag as a negative control. The C2 cells expressed EDC3 wild type and empty vector control (pcDNA3).
4. In vitro kinase assays was performed as outlined in Materials and Methods using purified recombinant EDC3 (0.2 μg) and Pim3 (0.2 μg) proteins, with and without PIMi (Pim447, 3 μM) for 6 h. Samples were analyzed by Western blotting with the indicated Abs.
5. In vitro kinase assays were performed as in (D). Samples were analyzed by Western blotting with the indicated Abs.
6. HSB‐2 leukemia cells were treated with PIMi (PIM447, 1 μM) for the indicated times, and protein extracts were subjected to Western blotting with the indicated Abs.
7. PC3‐LN4 cells were treated with PIMi (PIM447, 3 μM), AKTi (GSK690693, 5 μM), Torin1 (0.2 μM), or the combination of PIMi and AKTi for 24 h. Protein extracts were subjected to Phos‐tag SDS:PAGE with EDC3 Abs.

Source data are available online for this figure.


Pim1 and Pim3 bind to peptide sequences in the intrinsically disordered region of EDC3

To define which region of EDC3 is required to interact with Pim1 and Pim3, we mapped the EDC3 domain that binds to Pim1 and Pim3. EDC3 has four functional domains: an N‐terminal LSm domain, a central FDF domain, the C‐terminal YjeF‐N domain, and a low complexity intrinsically disordered linker region (IDR) (Fig 2A; Tritschler et al, 2007; Ling et al, 2008). We generated four separate GFP‐tagged EDC3 expression vectors, each representing a unique functional domain of the protein (F1, F2, F3, and F4) along with full‐length EDC3 vector (W), and expressed them in HEK‐293T cells along with Pim1 or Pim3. Vector controls of Pim1 (C1) or GFP‐EDC3 (C2) alone were also used. Binding assays performed with GFP‐Trap beads demonstrate that only the low complexity linker region (F2; amino acids 104–198) showed high affinity for Pim1 and Pim3 (Figs 2B and EV1C). As reported, DCP1a and DDX6 were detected in cells expressing F1 and F3 region of EDC3, respectively (Tritschler et al, 2008; Tritschler et al, 2009a). These results suggested that Pim1 interacts closely with amino acids 104–198 of EDC3.

Figure 2. Pim1 binds to peptide sequences in the low complexity region of EDC3.

1. Schematic of the functional domains of EDC3 and the fragments used to construct GFP fusion proteins. The phosphorylation site on EDC3 serine 161 is shown.
2. HEK‐293T cells were transfected with Pim1 and each EDC3 domain (Fig 2A) fused to the C‐terminus of the GFP‐tagged vector. The individual fragments were immunoprecipitated with GFP‐Trap beads and analyzed by Western blotting with the indicated Abs. EDC3 wild type (W), F1, F2, F3, and F4 were EDC3 fragments vector fused to GFP. The C1 lane was only transfected with the GFP Flag vector, while C2 contained EDC3 wild type fused to GFP and empty vector (pcDNA3).
3. 2D [¹H,¹⁵N] HSQC spectrum of EDC3_104–197 (100 µM EDC3_104–197; 20 mM HEPES pH 6.2, 500 mM NaCl, 0.5 mM TCEP; 298 K; 14.1 T). HSQC spectrum shows a single cross‐peak for each N‐HN pair, i.e., one per amino acid (other than prolines). The lack of chemical shift dispersion in the ¹H chemical shift dimension is typical for intrinsically disordered protein/region (IDR), as they lack the hydrogen bonding network in secondary structure elements of folded proteins.
4. 2D [¹H,¹⁵N] HSQC spectra of EDC3_104–197 with (red; 1:2 molar ratio) and without (blue) Pim1. We titrated Pim1 into ¹⁵N‐labeled EDC3, and the 2D [¹H,¹⁵N] HSQC spectrum of EDC3 (residues 104–197) when bound to Pim1 was measured. Direct comparison with the unbound 2D [¹H,¹⁵N] HSQC spectrum of EDC3 showed that while the majority of EDC3 peaks are unchanged, ~ 30 peaks become either broadened beyond detectability demonstrating they directly bind to Pim1 or show small chemical shift perturbations. Upon completion of the sequence‐specific backbone assignment of the EDC3 IDR, we identified that EDC3 residues 137–167 are necessary for the interaction with Pim1.

Source data are available online for this figure.

To further define the interacting domain of EDC3, we used biomolecular NMR spectroscopy. We recorded a 2D [¹H,¹⁵N] HSQC spectrum of EDC3 (residues 104–197, EDC3 IDR). This spectrum shows a single cross‐peak for each N‐HN pair, i.e., one per amino acid (other than prolines). Thus, the 2D [¹H,¹⁵N] HSQC spectrum functions as a “fingerprint” of the protein (Fig 2C). The lack of chemical shift dispersion in the ¹H chemical shift dimension is typical for intrinsically disordered protein/region (IDR), as they lack the hydrogen bonding network in secondary structure elements of folded proteins. Next, we titrated Pim1 into ¹⁵N‐labeled EDC3, and the 2D [¹H,¹⁵N] HSQC spectrum of EDC3 (residues 104–197) when bound to Pim1 (Fig 2C and D). Direct comparison with the unbound 2D [¹H,¹⁵N] HSQC spectrum of EDC3 showed that while the majority of EDC3 peaks are unchanged, ~ 30 peaks become either broadened beyond detectability (demonstrating they directly bind to Pim1) or show small chemical shift perturbations. Upon completion of the sequence‐specific backbone assignment of the EDC3 IDR (i.e., this step assigns each cross‐peak to a specific amino acid of EDC3), we identified that EDC3 residues 137–167 are necessary for the interaction with Pim1. These residues include the Pim1 phosphorylation site recognition motif ¹⁵⁶RRRHNS¹⁶¹. This domain of EDC3 is unique to the Pim1 kinase interaction and plays no known role in binding other P‐body proteins including DCP1a, DCP2, and DDX6 (Decker et al, 2007; Tritschler et al, 2007; Ling et al, 2008; Tritschler et al, 2009a; Fromm et al, 2012). Thus, EDC3 residues 137–167 are necessary to interact with Pim1.

EDC3 is a Pim and AKT protein kinase substrate

EDC3 contains a classic consensus phosphorylation site for the Pim protein kinase R‐X‐R‐H‐X‐(S/T) at serine 161 (S161) (Bullock et al, 2005; Peng et al, 2007). The AKT protein kinase is capable of phosphorylating this sequence but prefers R‐X‐R‐X‐X‐(S/T) (Obata et al, 2000). The Pim protein kinases, as well as AKT, play a regulatory role in the control of mRNA translation by modulating the activity of mTORC1 via inducing phosphorylation of proteins regulating this kinase (Zhang et al, 2009; Beharry et al, 2011; Cen et al, 2013; Warfel & Kraft, 2015; Song et al, 2016; Song et al, 2018; Padi et al, 2019) and by phosphorylating directly eIF4B (Cen et al, 2013; Song et al, 2018). A potential alternate mechanism for controlling translation is by phosphorylating EDC3 and regulating mRNA storage and degradation in P‐bodies. As demonstrated by an EDC3 S161 phospho‐antibody using recombinant EDC3 in the presence of ATP and magnesium, Pim1, Pim3, and AKT (Larance et al, 2010) are all shown to phosphorylate EDC3 (Figs 3A and B, and EV1D). The addition of Pim447 to the in vitro assay blocked the Pim‐induced phosphorylation of EDC3 (Fig EV1E). To further confirm the Pim1 kinase phosphorylation of EDC3, Pim1 was induced in PC3‐LN4 cell line with a doxycycline (DOX)‐inducible promoter. Inducing Pim1 with DOX in this tumor cell line caused an increase in EDC3 phosphorylation (Fig 3C). This increase was similar to that of the ribosomal protein S6, which is a downstream target of Pim regulation of mTORC1 (Zhang et al, 2009; Padi et al, 2019; Luszczak et al, 2020). The addition of Pim447 to these cells reversed the effect of Pim1 overexpression, eliminating phosphorylation. These results indicate EDC3 is a direct substrate of the Pim and AKT and points to the potential importance of this post‐translational modification in controlling the function of this P‐body protein.

Figure 3. EDC3 is a substrate of the Pim and AKT protein kinases.

1. In vitro kinase assays were performed as outlined in Materials and Methods with purified recombinant EDC3 (0.2 μg) and Pim1 (0.2 μg). Samples were analyzed by Western blotting with the indicated Abs.
2. In vitro kinase assay with EDC3 (0.2 μg) and AKT (0.1 μg). Samples were analyzed by Western blotting with the indicated Abs.
3. PC3‐LN4 cells containing a doxycycline (DOX)‐inducible Tripz‐Pim1 vector were treated with 50 or 100 ng/ml DOX for 48 h and subsequently incubated with DMSO or PIMi (PIM447, 3 µM) for 24 h. Western blots were probed with the indicated Abs.
4. PC3‐LN4 cells were treated with PIMi (PIM447) at increasing doses (0, 3, 5, and 10 μM) for 24 h. Extracts were subjected to Western blotting with the indicated Abs. Relative values of phospho‐signals were indicated below p‐EDC3 blot by quantification of intensity for p‐EDC3 S161 levels using ImageJ (NIH). p‐EDC3 S161 levels were normalized to levels of total EDC3 within the respective samples and compared to the p‐EDC3 S161 levels in the untreated samples.
5. PC3‐LN4 was treated with PIMi (PIM447, 3 μM), AKTi (GSK690693, 5 μM), or the combination for 24 h. Extracts were subjected to Western blotting with the indicated Abs.
6. PC3‐LN4 CRISPR‐Ctr and Pim1 KO cells were treated with PIMi (PIM447, 3 μM), AKTi (GSK690693, 5 μM), or the combination for 24 h. Extracts were subjected to Western blotting with the indicated Abs.

Source data are available online for this figure.

Next, we examined the contribution of Pim and AKT to EDC3 S161 phosphorylation using prostate and leukemia cell lines. Prostate tumors have been reported to contain increased levels of Pim1 (Dhanasekaran et al, 2001; Nawijn et al, 2011). To examine whether EDC3 S161 phosphorylation depends on the Pim protein kinase, EDC3 phosphorylation was monitored in PC3‐LN4 cells treated with increasing doses of the Pim inhibitor (PIMi), Pim447 for 24 h (Fig 3D). Changes in IRS1 phosphorylation, a known Pim specific substrate, were analyzed as a control. PIMi treatment (PIM447, 10 µM) was sufficient to cause approximately a 70% loss of EDC3 phosphorylation, suggesting that this phosphorylation is modified by Pim. There was no effect of this PIMi treatment on AKT activity as confirmed by the lack of change in the phosphorylation level of TSC2 S939, a specific AKT phosphorylation substrate (Cai et al, 2006) (Fig 3D). The difference in the response of insulin receptor substrate 1 (IRS1), a known Pim substrate (Song et al, 2016), and EDC3 to PIMi treatment might be accounted for by the ability of AKT to phosphorylate the EDC3 site as well. Similarly, PIMi treatment of T‐cell acute lymphoblastic leukemia (T‐ALL) cell line HSB‐2 (Padi et al, 2017), a cell line that does not contain activated AKT, completely blocked EDC3 phosphorylation (Fig EV1F) suggesting that Pim kinases control the phosphorylation of EDC3 in a context‐dependent manner. To further understand the activity of AKT on the phosphorylation status of EDC3 in the PC3‐LN4 cells, we treated these cells with an AKT inhibitor (AKTi) either alone or in combination with PIMi. Our results demonstrated that the addition of the PIMi reduced the level of phosphorylation (lane 2, Fig 3E), while the AKTi alone showed faint or no effect (lane 3, Fig 3E). As noted, Pim1 was co‐immunoprecipitated with EDC3 in Fig 1C, while AKT was not, further supporting the notion that Pim is the primary kinase of EDC3 phosphorylation. Since these kinases share overlapping phosphorylation sites, AKT might regulate EDC3 when Pim kinase activity is depressed or when growth factors highly stimulate AKT activity. Indeed, the combination of both inhibitors induced an almost complete loss of EDC3 phosphorylation (lane 4, Fig 3E), suggesting that an AKT effect could be seen when Pim kinases are inhibited. Phos‐tag analysis of PC3‐LN4 cells showed that more than 80% of EDC3 is in the phosphorylated form in tumor cells, and most importantly, the combination of PIMi and AKTi treatment largely dephosphorylated the EDC3 allowing for the further migration on SDS:PAGE (Fig EV1G). In Pim1 KO PC3‐LN4 cells, pan‐PIMi treatment reduced EDC3 phosphorylation, suggesting that in addition to Pim1, Pim 3 controls EDC3 phosphorylation (Fig 3F). Inhibition of AKT in the Pim1 KO cells attenuated EDC3 phosphorylation, suggesting that Pim plays the predominant role in the cells treated with AKTi. This finding is consistent with our previous observation that AKT inhibition upregulated Pim kinase levels (Cen et al, 2013). Together these results demonstrate that the Pim kinases phosphorylate EDC3 and function as the dominant EDC3 kinase.

EDC3 phosphorylation inhibits its localization in P‐bodies

To determine if EDC3 phosphorylation by Pim and AKT signaling affects its localization in P‐bodies, PC3‐LN4 cells were treated with Pim447, the AKT inhibitor (AZD5363), or a combination of both drugs and subjected to immunofluorescence (IF). Strikingly, we observed a significant increase in EDC3 foci following individual drug treatment with the most significant increase in the combination‐treated cells (Fig 4A and B). Importantly, EDC3 foci colocalized 100% with DCP1a foci demonstrating that we were observing new P‐bodies. Hank’s balanced salt solution (HBSS) incubation starves the cell of nutrients inducing a stress response and P‐body formation (Aizer et al, 2014). Cells starved with HBSS showed induced P‐bodies as shown in control (Fig EV2A and B). This treatment decreased phosphorylation of EDC3 and IRS1 phosphorylation but had no effect on the phosphorylation of AKT substrate GSK3β, suggesting that the activity of Pim kinases was attenuated under nutrient starved conditions (Fig EV2C). Even in these stressed cells, combination inhibitor treatment caused a further increase in the number of EDC3 foci per cell (Fig EV2A and B). This induction could be due to complete suppression of both Pim and AKT activity. As previously observed in Drosophila cells, EDC3 knockdown cells were still able to form P‐bodies (Eulalio et al, 2007b). As expected, the number of P‐bodies containing DCP1a in EDC3 KO cells was not sensitive to Pim or AKT inhibitor treatment (Fig 4C and D). PC3‐LN4 EDC3 KO cells without any treatment contained more P‐bodies than the parental cell line. This EDC3 independent P‐body formation might be secondary to the induction of a stress signal, or alternatively, the availability of proteins, e.g., DDX6, that would normally be sequestered by EDC3. It is possible that phosphorylated EDC3 acts as a “dominant negative”, impairing EDC3 dimerization and/or sequestering other PB proteins (DCP1a, DDX6). Knockout of EDC3 alone may not negatively influence P‐body formation. Thus, the increased P‐body number after PIM and AKT inhibitor treatment is dependent on dephosphorylation of EDC3. Taken together, these results suggest the importance of phosphorylation by these protein kinases in the regulation of P‐body formation.

Figure 4. EDC3 phosphorylation reduces its P‐body localization.

1. Immunofluorescence (IF) staining of PC3‐LN4 cells following 6‐h treatment with DMSO (Control), PIMi (PIM447, 3 µM), AKTi (GSK690693, 5 μM), and both inhibitors (AKTi + PIMi) and stained with anti‐EDC3 and anti‐DCP1a antibodies as well as DAPI.
2. Quantification of EDC3/DCP1a foci number from Control (Ctrl), PIMi, AKTi, and the AKTi + PIMi combination in (A).
3. Immunofluorescence (IF) staining of PC3‐LN4 EDC3 knockout cells following 6‐h treatment with DMSO (Control), PIMi (PIM447, 3 µM), AKTi (GSK690693, 5 μM), and both inhibitors (AKTi + PIMi) and stained with anti‐EDC3, and anti‐DCP1a antibodies (Abs) as well as DAPI.
4. Quantification of DCP1a foci number for Control (Ctrl), PIMi, AKTi, and AKTi + PIMi combination in (C).

Data information: IF Images were subsequently merged using FIJI image analysis software. The yellow box shows a zoom of the merged image. Images were captured under 40× magnification; scale bar: 10 µm. P‐body number was counted by FIJI. The quantification data of (B and D) are presented as mean ± SEM (n = 30 cells/treatment group). P values < 0.05, calculated using unpaired Student’s t‐test, are considered significant. P values > 0.05 not significant (n.s.). The analyses done included multiple independent fields.

Figure EV2. EDC3 phosphorylation reduces its P‐body localization.

1. Immunofluorescence (IF) staining of PC3‐LN4 cells following treatment with DMSO (Control), PIMi (Pim447, 3 µM), and AKTi (AKT inhibitor, AZD5363 3 µM) in the presence of Hank’s balanced salt solution (HBSS) for 2 h with anti‐EDC3 Ab and DAPI. Images were captured under 40× magnification; scale bar: 10 µm.
2. Quantification of P‐body number in (A) using FIJI image analysis software. The experiments were done in three independent fields. Data are presented as the mean ± SEM (n = 30 cells/treatment group). P values < 0.05 were calculated using unpaired Student’s t‐test.
3. PC3‐LN4 cells were incubated in HBSS for 2 h, and lysates were subjected to Western blotting with the indicated Abs.
4. Quantification of P‐body number per cells for PC3‐LN4 parental (Ctrl), EDC3 SA, and EDC3 SD using FIJI image analysis software. The experiments were quantified in three independent fields. Data are presented as the mean ± SEM (n = 30 cells/treatment group). P values < 0.05 were calculated using unpaired Student’s t‐test.
5. Immunofluorescence (IF) staining of PC3‐LN4‐EDC3 S161D (SD) cells following 6‐h treatment with DMSO (Control) and PIMi + AKTi (Pim447, 3 µM; AZD5363, 3 µM) combination with anti‐EDC3 Ab, anti‐DCP1a Ab, and DAPI. Images were captured under 40× magnification; scale bar: 10 µm.

Source data are available online for this figure.


To test if EDC3 in P‐bodies is phosphorylated, a unique method was developed for the affinity purification of P‐bodies through the expression of GFP or RFP tagged P‐body proteins (Fig EV3A). In U2OS cells, P‐bodies isolated with RFP DCP1a exhibit the correct size (150–500 nm), spheroid shape, and positive staining for RNA (Fig EV3, EV4, EV5), validating the P‐body isolation method. Utilizing PC3‐LN4 cells expressing GFP‐DCP1a as the P‐body tag, we demonstrate that EDC3 and its interactors, DDX6 and DCP1a, were all present within purified P‐bodies as well as other known P‐bodies components, PUM1, TTP1, YTHDF2, LSM4, and LSM14 but not cytosolic IRS1, indicating that these P‐bodies were not contaminated with other proteins not known to be in P‐bodies (Fig 5A). Importantly, p‐EDC3 was not detected in these purified P‐bodies (Fig 5B) nor was it found colocalized by IF with P‐bodies (Fig 5C). These observations suggest that phosphorylation of EDC3 inhibits its ability to associate with P‐bodies.

Figure EV3. P‐body purification.

1. Diagram detailing the P‐body purification method.
2. RFP‐DCP1a protein was immunoprecipitated with RFP‐Trap beads and imaged under fluorescence microscopy. Representative images are of enriched P‐bodies (PB) are shown.
3. The samples from Fig 5B were stained with an RNA dye RNA‐SYTO‐select. Representative images are of enriched P‐bodies (PB) as shown.
4. Overlay of images with RFP and RNA‐SYTO from Fig 5B and C.
5. DCP1a P‐bodies bound to RFP‐Trap nanobody beads following isolation.

Data information. In (B–E), images are acquired by fluorescence microscopy. Magnification: 100×, scale bar: 1 µm.

Figure EV4. Analysis of P‐body RNA composition in DMSO and kinase inhibitor‐treated cells.

1. Graphical representation of principal component (PC1 and PC2) analyses comparing the variation between triplicate groups of P‐body (PB) RNA purified samples or whole‐cell (WC) RNAs from drug‐treated [PIMi (PIM447, 3 µM), AKTi (AZD5363, 5 µM)] and untreated GFP‐EDC3 PC3‐LN4 cells.
2. Hierarchical distance‐based clustering of normalized expression data. Data of RNA sequences from total RNA in each condition show DMSO: WC and Drug: WC. DMSO: PB and Drug: PB are form PB RNA sequences.
3. Venn diagram comparing the P‐body purified RNA dataset from DMSO‐treated cells from the present study to the P‐body RNA dataset (Mock) from Hubstenberger et al.
4. RNA‐seq analysis of enriched (orange) or depleted (green) RNA within P‐bodies in DMSO treatment.
5. Specific RNA subtypes found in enriched and depleted P‐body RNAs purified from kinase inhibitor‐treated GFP‐EDC3 PC3‐LN4 cells.

Source data are available online for this figure.


Figure EV5. EDC3 S161 influences cellular growth and migration.

1. Western blot of lysates prepared from CRISPR/Cas9 PC3‐LN4 EDC3 KO (#1, 2) cells, EDC3 S161A (#1, 2), EDC3 S161D (#1, 2) mutants, and parental cells with the indicated Abs.
2. PC3‐LN4 EDC3 KO cells containing doxycycline (DOX)‐inducible EDC3 WT, S161A, or vector control (Mock) were treated for 48 h with 0, 75, and 150 ng/ml DOX. All extracts were subjected to Western blotting with the indicated Abs.
3. Representative images of transwell migration assays performed in PC3‐LN4 EDC3 KO cells containing doxycycline (DOX)‐inducible EDC3‐S161A with and without DOX (100 ng/ml) induction (Fig 7C). The number of cells that migrated (40 h) was counted as outlined in Materials and Methods. Scale bar: 400 µm.
4. Representative images of transwell invasion assay of PC3‐LN4 EDC3 wild‐type (naive), EDC3 S161D, EDC3 KO, and EDC3 S161A cell lines (Fig 7D). Cells were plated in transwell dishes coated with Matrigel and incubated for 72 h. Invading cells were stained as outlined in Materials and Methods and imaged. Scale bar: 400 µm.
5. Relative mRNA expression of indicated genes in PC3‐LN4 WT and EDC3 S161A mutant cells.
6. Relative mRNA expression of indicated genes PC3‐LN4 EDC3 KO cells expressing EDC3 WT or EDC3 S161A mutant plasmids.
7. Relative mRNA expression of indicated genes in PC3‐LN4 WT and EDC3 S161A mutant cells.
8. Relative mRNA expression of indicated genes PC3‐LN4 EDC3 KO cells expressing EDC3 WT or EDC3 S161A mutant plasmids.

Data information: Real‐time PCR data (E–H) are presented as mean ± SEM (n = 3/group). Unpaired Student's t‐test was used for statistical analysis. *P < 0.05, **P < 0.01.

Source data are available online for this figure.


Figure 5. Phospho‐EDC3 (S161) status in P‐bodies.

• A, B
  Western blot of whole cell lysates (WCL) and P‐body (PB) purified from PC3‐LN4 cells stably expressing GFP or GFP‐DCP1a used as a P‐body tag. The P‐bodies (PB) were isolated with GFP‐Trap beads and analyzed by Western blotting with the indicated Abs.
• C
  Immunofluorescence (IF) of PC3LN4 cells under normal culture condition with anti‐phospho‐EDC3, anti‐EDC3 Abs, and DAPI. Images captured under 40× magnification; scale bar: 10 µm.
• D
  Immunofluorescence of PC3‐LN4‐EDC3 S161A (SA) cells under normal culture condition with anti‐DCP1a and anti‐EDC3 Abs as well as DAPI. Images captured under 40× magnification; scale bar: 10 µm.

Source data are available online for this figure.

To exclude the indirect effects of Pim and PI3K/AKT inhibitors on EDC3 P‐body localization, a PC3‐LN4 cell line harboring a phospho‐dead EDC3 S161A mutation or phospho‐mimetic EDC3 S161D mutation was generated by CRISPR/Cas9. The number of EDC3 foci in PC3‐LN4 cells with the S161A mutation increased significantly (Figs 5D and EV2D). In contrast, kinase inhibitor treatment of PC3‐LN4 EDC3 S161D cell line failed to induce the formation of any EDC3 containing P‐bodies (Fig EV2D and E). Together these data reinforce the finding that EDC3 S161 phosphorylation controls EDC3 localization in P‐bodies.

P‐body purification and enriched EDC3‐dependent RNAs in prostate cancer cells

Pim and AKT phosphorylation of EDC3 may alter RNA transit and transcript fate in P‐bodies. Unphosphorylated EDC3 interacts with other P‐body components and localized into P‐bodies. Pim and AKT inhibition might also affect this interaction and alter P‐body target RNA populations. To understand whether there is a change in RNA populations, following treatment with Pim447 and AZD5363 for 6 h or DMSO, P‐bodies were isolated using GFP‐EDC3 in biological triplicates, and RNA‐seq was performed. The isolated RNA was sequenced at a minimum depth of 20 million reads, mapped, and transcripts assembled using the RMTA workflow in CyVerse's Discovery Environment (Merchant et al, 2016; Peri et al, 2020).

Principal component analyses and sample distance matrices revealed expected clustering between replicates, with the majority of variance driven by differences in RNA populations between P‐bodies and total RNA (Fig EV4A). Our P‐body RNA dataset from DMSO‐treated cells showed extensive overlap (~ 50%) with previously published P‐body RNA populations despite differences in isolation methods and source material (Fig EV4C; Hubstenberger et al, 2017). Fisher's exact test using GeneOverlap (R) demonstrated that the overlap between the two P‐body datasets, Hubstenberger and ours (http://nemates.org/MA/progs/overlap_stats.html), was highly statistically significant at P < 3.5 e⁻¹¹⁴, with an representation factor of 1.9–3.2. Small differences were also seen in the P‐body RNA content between the kinase inhibitor‐treated and DMSO‐treated groups of PC3‐LN4 GFP‐EDC3 (Fig EV4B). We then used anota2seq (Oertlin et al, 2019) to search for populations of RNAs that were enriched in P‐bodies under kinase inhibitor conditions relative to DMSO independent of changes in whole‐cell RNA populations. However, this analysis did not disclose a marked difference in the mRNAs that were enriched in P‐bodies purified from kinase inhibitor‐treated cells versus DMSO‐treated cells (Fig 6A). This result appears logical since P‐bodies from treated and untreated samples were both isolated by using a nanobodies directed at isolating GFP‐EDC3. Dephosphorylation of EDC3 with kinase inhibitors markedly increased the number of P‐bodies in the treated cells yielding a great deal more RNA found in P‐bodies than in control cells.

Figure 6. Enriched EDC3‐dependent RNAs in P‐bodies purified from prostate cancer cells.

1. Comparison of P‐body RNA and total RNA levels with and without AKT and PIM inhibitor treatment by anota2Seq.
2. Specific RNA subtypes found in P‐body RNAs purified from DMSO‐treated GFP‐EDC3 containing PC3‐LN4 cells.

To delineate what RNAs are found in EDC3 containing P‐bodies, the expression profiles of RNAs that were enriched in P‐bodies relative to the total RNA in these control tumor cells were examined using DESeq2 (Materials and Methods). 7,542 P‐body transcripts were identified with the level of 3,711 mRNAs enriched and 3,831 depleted in P‐bodies when compared to total cellular RNA (adjusted P‐value < 0.001; Fig EV4D). Analysis further demonstrated that while the total cell lysate after rRNA deletion included a mixture of mRNAs, lncRNA, sn/snc/snoRNA, and pseudogenes, in control tumor cells 80% of EDC3 containing P‐body RNAs were either mRNAs or lncRNAs (Fig 6B). Similar results were obtained in kinase inhibitor‐treated cells both in those RNAs that were either enriched or depleted in EDC3‐positive P‐bodies when compared to the total cell RNA (Fig EV4E).

S161A mutation in EDC3 suppresses tumor cell growth, migration, and invasion

We next examined the biological impact of EDC3 phosphorylation by testing whether EDC3 Ser161 phosphorylation affects cell growth using the IncuCyte real‐time imaging analysis platform. Tumor cells with S161A mutation grew much more slowly in culture versus WT, EDC3 S161D, and EDC3 KO PC3‐LN4 cell lines (Fig 7A). These data suggested that EDC3 phosphorylation is involved in controlling cell growth. Given the effects on tumor growth, we next investigated the outcome of EDC3 S161 phosphorylation on migration and invasion of cells. Transwell assays were performed using EDC3 S161A, EDC3 S161D knock‐in, and EDC3 KO cells. Both the S161D and KO clones demonstrated increased migration and invasion compared to parental cells, while the EDC3 S161A mutant cells had significantly diminished migratory and invasive capability (Figs 7B and D, and EV5D). To further demonstrate the activity of the EDC3 S161A mutant on migration and invasion, DOX‐inducible EDC3 S161A expression vectors were transduced into EDC3 KO cells (Fig EV5B). The induction of EDC3 S161A with DOX treatment caused a decrease in both the migratory and invasive ability of these cells (Figs 7C and E, and EV5C), again suggesting the importance of this phosphorylation. Together these results demonstrate that the phosphorylation state of EDC3 has a profound impact on the biologic activity of these cancer cells.

Figure 7. S161A mutation in EDC3 suppresses tumor cell growth, migration, and invasion.

1. PC3‐LN4 naïve and CRISPR/Cas9 genome‐edited EDC3 S161A (SA), S161D, and EDC3 knockout (KO) cell lines were plated (4,000 cells/well in 24 well plates) and imaged every 12 h using phase contrast in the IncuCyte ZOOM platform. The graph shown was generated with IncuCyte Basic Software following 5 days of growth. The experiments were done in triplicates. Values are mean ± SEM. P values < 0.05, calculated using unpaired Student’s t‐test.
2. Quantification of transwell migration assays performed for PC3‐LN4 EDC3 wild‐type, EDC3 S161A, EDC3 KO, and EDC3 S161D cell lines. The number of migrated cells (40 h) was counted as outlined in Materials and Methods.
3. Quantification of transwell migration assays performed for PC3‐LN4 EDC3 KO cells containing Doxycycline (DOX)‐inducible EDC3‐S161A with and without DOX (100 ng/ml) induction. The number of migrated cells (40 h) was counted as outlined in Materials and Methods.
4. Quantification of transwell invasion assay comparing PC3‐LN4 EDC3 wild‐type, EDC3 S161A, EDC3 KO, and EDC3 S161D cell lines. Cells were plated in transwells coated with Matrigel and incubated for 72 h. The cells that invaded the matrix were counted as outlined in Materials and Methods.
5. Quantification of transwell invasion assay performed for PC3‐LN4 EDC3 KO cells containing DOX‐inducible EDC3‐S161A were either treated with or without DOX (100 ng/ml). The number of cells that invaded was counted as outlined in Materials and Methods.
6. Relative mRNA expression of ITGA6, ITGB1, and KLF4 in PC3‐LN4 WT and EDC3 S161A mutant cell lines.
7. Western blot analysis of proteins extracted from PC3‐LN4 EDC3 wild‐type, EDC3 S161A, EDC3 knockout (KO, C60 cell line), and EDC3 S161D cell lines using the specified antibodies.
8. Relative mRNA expression of ITGA6, ITGB1, and KLF4 in PC3‐LN4 EDC3 KO‐C60 cells, containing reexpressed EDC3 WT or EDC3 S161A mutant cDNAs.
9. Western blot analysis of proteins extracted from PC3‐LN4‐EDC3‐KO cells expressing GFP, EDC3 WT, or the EDC3 S161A mutant using the specified antibodies.
10. Western blot analysis of proteins extracted from PC3‐LN4 cells treated with DMSO or the combination of Pim447 and AZD5363 for 96 h using the specified antibodies.

Data information: All invasion and migration assays were performed in triplicates and 6 different fields at 10× magnification were counted in 3 separate inserts. The data are presented as the mean ± SD. P values < 0.05, calculated using unpaired Student’s t‐test. Real‐time PCR data (F & H) are presented as mean ± SEM (n = 3/group). Unpaired Student's t‐test was used for statistical analysis. *P < 0.05, **P < 0.01 and NS, not significant.

Source data are available online for this figure.

To further understand the mechanism that drove decreased migration and invasion of PC3‐LN4 EDC3 S161A, we compared the expression of Integrins (ITG), which play an important role in cell–cell attachment, in WT and EDC3 S161A mutant PC3‐LN4 cells. Integrins are heterodimeric transmembrane protein receptors expressed on the cell surface that bind to surrounding extracellular matrix (ECM). We observed a decrease in the mRNA expression of integrins A2, A5, A6, and B1 in PC3 LN4 cells with EDC3 S161A knock‐in mutant and in PC3 LN4‐EDC3KO cells complemented with GFP EDC3 S161A when compared to their respective controls (Fig EV5E–H). Because integrins α6 and β1 have an important role in cancer progression (Sroka et al, 2016) and are a potential therapeutic target for anti‐cancer chemotherapy (Toth et al, 2019), we further focused on the expression of these two integrins. The mRNA and protein levels of integrin α6 and β1 were both reduced in PC3‐LN4 EDC3 S161A cells and in EDC3 KO cells complemented with GFP EDC3 S161A, and treatment of wild‐type cells with PIM and AKT inhibitors decreased these integrins as well (Fig 7F–J). In addition, we investigated Kruppel‐like factor 4 (KLF4) mRNA and protein expression, since knockdown of EDC3 has been shown to cause an increase in the expression of KLF4 in human epidermal keratinocytes and regulate invasion in breast cancers (Badis et al, 2004; Yu et al, 2011; Wang et al, 2015). Indeed, the protein and mRNA of KLF4 level were reduced in PC3‐LN4 EDC3 S161A cells and PC3‐LN4 cells treated with both kinase inhibitors (Fig 7F–H and J). These results suggest that the expression of KLF4, ITGB1, and ITGA6 is regulated in an EDC3 phosphorylation‐dependent manner and thus could play an important role in regulating cell growth and motility.

EDC3 phosphorylation levels are elevated in tumors and affect growth

To test whether an EDC3 S161A mutation would affect tumor growth in immunosuppressed mice, male SCID mice were injected subcutaneously with equal numbers of WT or EDC3 S161A mutant PC3‐LN4 cell lines, and their growth was measured over time (Fig 8A). The two cell lines containing the mutant protein showed a dramatic reduction in their ability to form tumors and had significantly slower tumor growth. The observation that the EDC3 S161A has a profound effect on tumor growth suggested that Pim and AKT kinases could, through EDC3 phosphorylation, have a significant role in modulating tumor growth.

Figure 8. P‐EDC3 levels regulate tumor growth and are elevated in multiple breast cancer cell lines.

1. Two independently derived CRISPR/Cas9 PC3‐LN4 EDC3 S161A mutants (EDC3 SA#1, #2) and wild‐type cells were injected subcutaneously in the flank (5 × 10⁶) of SCID mice (n = 8/group), and tumor growth was monitored until the control group reached 1,000 mm³. The subcutaneous tumor volume measurement (mm³) was done as outlined in Materials and Methods. Data are presented as mean ± SEM (n = 8/group).
2. Phosphorylated EDC3 levels were examined in normal immortalized breast cell lines, MCF10A and MCF12A, human breast cancer cell lines HCC38, HCC1954, MDA‐MB‐231, MDA‐MD‐468, BT‐549, MRKNU1, and mouse 4T1 by Western blotting with the indicated Abs.
3. Dual immunohistochemistry staining of human breast tumor with p‐EDC3 Ab and total EDC3 (brown). Scale bar: 50 µm.
4. Dual immunohistochemistry staining of human normal resected breast tissue sections with p‐EDC3 Ab and total EDC3 (brown). Scale bar: 50 µm.

Source data are available online for this figure.

To further examine the phosphorylation of EDC3 in tumor samples, we analyzed a panel of breast cancer cell lines with Western blotting (Fig 8B). EDC3 was highly phosphorylated across multiple breast cancer cell lines compared to the non‐tumorigenic epithelial cell lines, MCF10A, and MCF12A. To test whether increased EDC3 phosphorylation was seen in human tumors, immunohistochemistry (IHC) staining of a resected patient breast tumor was analyzed with antibodies against both phospho‐EDC3 S161 (red) and total EDC3 protein (brown). The patient tumor sample showed increased red staining consistent with EDC3 phosphorylation (Fig 8C), while the surrounding normal glandular breast tissue (Fig 8D) stained brown demonstrating predominantly unphosphorylated EDC3.

Collectively, these data demonstrate that EDC3 phosphorylation is both elevated in human tumor cell lines and can be visualized in human tumor samples. The inhibition of EDC3 phosphorylation by targeted genetic manipulation demonstrates that this phosphorylation site can have a profound effect on tumor cell growth both in culture and animals, as well as migration and invasion. Kinase inhibitors that target this pathway appear to function in part by dephosphorylating EDC3, a result that has the potential to change the RNA content of P‐bodies. Changes in P‐body RNA could have significant effects on tumor growth.

Discussion

We demonstrate that the Pim protein kinases phosphorylate the P‐body protein EDC3 in multiple tumor types, including T‐ALL, breast, and prostate cancer. This phosphorylation blocks the ability of EDC3 to associate with P‐bodies which correlates with profound cellular consequences, including regulation of tumor cell growth, migration, and invasion. Inhibition of phosphorylation by a combination of small‐molecule kinase inhibitors suppressed these biological outcomes and increased P‐body number. Together these findings suggest a unique mechanism by which activation of protein kinases in tumor cells can control RNA storage and potentially the translation of specific mRNAs modulating critical cellular functions.

Overexpression of Pim1 has been linked to increasing invasiveness and/or poor prognosis in patients with a number of cancers, e.g., breast, prostate, and lung cancers (Chen & Tang, 2019). Pim1 levels are elevated in TNBC and prostate cancer cell lines (Dhanasekaran et al, 2001; Horiuchi et al, 2016; Braso‐Maristany et al, 2017), consistent with the observed increase in phosphorylation of EDC3 in these cell lines. Consistent with these results, IHC staining demonstrated increased phosphorylation of EDC3 in a pathologic breast cancer samples versus normal. These results suggest that the level of EDC3 phosphorylation may be regulated in specific breast cancer subtypes and could modulate EDC3 activity to control mRNA storage and destruction in P‐bodies.

Using NMR spectroscopy, Pim1 is shown to interact with residues 137–167 in the central IDR domain of EDC3 (residues 104–197). In yeast, this region is thought to play a role in mRNA binding and LLPS, both of which contribute to P‐body formation, suggesting the possibility that Pim binding or the phosphorylation of S161 decreases the ability of this protein to interact with specific mRNAs, and/or alter P‐body assembly. The effect of EDC3’s phosphorylation on P‐body formation is supported by our observation that the phosphorylated EDC3 was not detected in P‐bodies (Fig 5B and C) and EDC3 S161A mutants increased P‐body formation (Fig 5D). EDC3 functions as a scaffold for multiple proteins that play a role in mRNA degradation, e.g., DDX6, DCP1a, and DCP2, and phosphorylation could inhibit P‐body formation by altering EDC3’s dimerization and/or interaction of other proteins, and thus its ability to form multiprotein complex.

Our P‐body purified RNA dataset shows a significant overlap with previously published data examining P‐body RNA content (Hubstenberger et al, 2017) generated using an alternative method of P‐body isolation. The differences between the datasets could reflect differences between prostate cancer and HEK‐293T cells or the methods of P‐body isolation (Fig EV4B). Importantly, Hubstenberger et al used Lsm14 for P‐body purification. LSM14, which is a component of both P‐bodies and stress granules, is only found in P‐bodies with 4E‐T. Although the LSm domains of EDC3 and LSM14 are structurally quite similar, they interact with distinct binding partners except DDX6 (Brandmann et al, 2018). This observation is also reported in Drosophila (Tritschler et al, 2008), since EDC3 and LSm14A compete for binding to DDX6 (Jonas & Izaurralde, 2013). While EDC3 interacts with DCP1–DCP2 decapping machinery (Kshirsagar & Parker, 2004; Tritschler et al, 2008), the LSM14 complexes in eukaryotic organisms are involved in repressing mRNA translation via 4E‐T (Rajyaguru et al, 2012; Jonas & Izaurralde, 2013). In yeast, the decapping activators, including EDC3 and Scd6 (LSm14A), function both to repress translational initiation and/or stimulate DCP1/2 (Nissan et al, 2010). Thus, the difference in P‐body RNA content observed by these two isolation methods could also be dependent on the abundance, modification status, and function of EDC3 versus LSM14 proteins within P‐bodies. Further experiments will be needed to compare and understand the differences between EDC3 and LSm 14a containing P‐bodies.

The role of P‐bodies in regulating cancer growth or tumorigenesis is not well studied. It has been shown that overexpression of the metastatic lymph node 51 protein, which is upregulated in some breast and other cancer types, leads to P‐body disassembly and, eventually, cancer progression (Degot et al, 2002; Cougot et al, 2014). Another study shows that spleen tyrosine kinase induces autophagy‐mediated clearance of P‐bodies that occurred throughout mesenchymal–epithelial transition and the P‐body clearance required tumor metastasis. Consistent with these findings, our results demonstrate a strong anti‐proliferative and diminished migratory ability in cancer cells with increased P‐bodies due to EDC3 dephosphorylation. This increase in P‐body number could lead to the increased sequestration or decay of a specific subset of mRNAs, limiting the translation of mRNAs related to the cell attachment, transcription, migration, and cell growth. ITGB1 and ITGA6 could be an example of two such proteins found to be decreased when EDC3‐containing P‐bodies are present. KLF4, a transcription factor associated with differentiation, is another example. Interestingly, these genes are involved in migration and invasion in prostate and breast cancers. The hetero‐complex of ITGB1 and ITGA6 proteins contributes to prostate tumor metastasis to bone (Sroka et al, 2016). KLF4, which is highly expressed in about 70% of breast cancers, could regulate cancer stem cell formation and invasion in breast cancer cells (Yu et al, 2011). Thus, P‐body formation may represent an important component to regulate multiple pathways via attenuating mRNA translation during malignant transformation.

It is possible that PIMi and AKTi might affect the activity of other factors which influence P‐body formation independent of EDC3. However, EDC3 KO cells are insensitive to inhibitor treatment, suggesting that EDC3 is centrally important in these effects (Fig 4C and D). Additionally, the protein and mRNA levels of ITGB1, ITGA6, and KLF4 are all decreased both in EDC3 S161A cells (Fig 7F–I) and by inhibitor treatment, suggesting that EDC3 phosphorylation plays a major role in regulating these events (Fig 7I). However, it is currently unknown whether PIMi and AKTi could also affect other factors which could regulate the formation or activity of P‐bodies to dephosphorylate EDC3.

Importantly, in the clinic, Pim kinase levels are shown to be increased in breast cancer patients that are treated with inhibitors of the PI3 kinase/AKT pathway and the resistance to PI3K inhibitor can be mediated by increases in the Pim1 kinase in cancer treatment (Song et al, 2018). Pim and AKT kinase inhibitors synergistically inhibit EDC3 phosphorylation and increase the number of P‐bodies. This induction of EDC3‐containing P‐bodies could be part of the mechanism of action of these kinase inhibitors and may control a set of mRNAs and their translation. We have previously shown that Pim and AKT upregulate translational initiation via eIF4B and DEDPC5 phosphorylation, and addition of kinase inhibitors blocks translational initiation (Cen et al, 2014; Padi et al, 2019). Translational repression also increases P‐body number and RNA storage, which may also impact on this process (Eulalio et al, 2007b; Rao & Parker, 2017). Thus, oncogenic kinases, i.e., Pim and AKT, could be potential master regulators of translation through orchestrating multiple steps spanning mRNA repression, decay, and translational initiation.

In summary, we show that S161 phosphorylation plays a significant role in regulating the P‐body localization of EDC3 protein and the growth, invasion, and migration of tumor cells. These results suggest that EDC3 phosphorylation may regulate a subset of P‐bodies or mRNAs that play a crucial role in controlling essential biologic processes within these tumor cells.

Materials and Methods

Antibodies and reagents

The following antibodies were purchased from Cell Signaling Technology: anti‐Pim1 (Cat#2907), anti‐Pim2 (Cat#4730), anti‐Pim3 (Cat#4165), anti‐AKT (Cat#9272), anti‐phospho‐AKT (S473, Cat#4058), anti‐HIS (Cat#2365), anti‐EDC3 (Cat #14495), anti‐DDX6 (Cat #9407), anti‐GFP (Cat#2555), anti‐phospho‐IRS1 (s1101, Cat#2385), anti‐eIF4B (Cat#3592), anti‐phospho‐eIF4B (Ser406, Cat#5399), anti‐GSK3B (Cat #9315), anti‐phospho‐GSK3B (Ser9, Cat #9336), anti‐S6 (Cat #9202), anti‐phospho‐S6 (Ser235/236, Cat#4856), and anti‐DCP1a (Cat#15365). Antibodies bought from Santa Cruz Biotechnology included anti‐EDC3 (sc‐365024), anti‐DCP1a (sc‐100706), and anti‐DDX6 (sc‐376433). Anti‐IRS1 (Cat #06‐248) antibody was obtained from Merck Millipore. The anti‐phospho‐EDC3 (S161, Cat#600‐401‐J38) was purchased from Rockland antibodies. Anti‐DCP1a (ab47811) was from Abcam. Anti‐DDX6 (GTX102795), anti‐Lsm14a (GTX120902), and anti‐TTP (GTX130974) were from GeneTex. Anti‐PUM1 (15982‐1‐AP), anti‐Lsm4 (10834‐1‐AP), and anti‐YTHDF2 (24744‐1‐AP) were purchased from Proteintech. HRP‐conjugated anti‐β‐actin (A3854) was from Sigma, and HRP‐linked mouse IgG and rabbit IgG were purchased from GE Healthcare Life Sciences. Insulin was obtained from Sigma. The kinase inhibitors BKM 120 (A11016), AZD5363 (A11759), and AZD1208 (A13203) were purchased from Adooq Bioscience. PIM447 and GSK690693 were kind gifts from Novartis Oncology and GlaxoSmithKline, respectively.

Cell culture

The breast cancer cell lines 4T1, MDA‐MB‐231, MDA‐MB‐468, HCC38, HCC1954, and BT‐549, were purchased from the American Type Culture Collection (ATCC). MRK‐nu‐1 (MRKNU1) cells were purchased from the Japanese Collection of Research Bioresources Cell Bank. All of these cell lines, including the PC3‐LN4 and PC3‐LN4 mutant cell lines, were cultured in RPMI with 10% fetal bovine serum (FBS) and 1% penicillin and streptomycin at 37°C in 5% CO₂. T‐cell acute lymphoblastic leukemia cell line HSB‐2, a kind gift from Jon C Aster (Harvard Medical School, Boston, MA), was also cultured in RPMI with 10% fetal bovine serum (FBS) and 1% penicillin and streptomycin at 37°C in 5% CO₂. BT‐549 were grown in the above media, plus 0.023 U/ml insulin. The MCF10A and MCF12A cell lines were cultured in DMEM/F12 medium with horse serum (5%), EGF (20 ng/ml), hydrocortisone (0.5 μg/ml), cholera toxin (100 ng/ml), insulin (10 μg/ml), and 1% penicillin and streptomycin at 37°C in 5% CO₂. The U2OS and HEK‐293T cells were cultured in DMEM with 10% FBS and 1% penicillin and streptomycin at 37°C in 5% CO₂. Regular testing of mycoplasma contamination was performed in these cell lines using MycoAlert™ Mycoplasma Detection Kit (LT07‐118, Lonza), and only mycoplasma free cells were used for experimentation. Cell lines were authenticated by short tandem repeats single nucleotide polymorphism at the University of Arizona Genetics Core Facility.

Transient transfection and DNA plasmids

HEK‐293T cell lines were transfected with 10 µg of DNA plasmids using Lipofectamine 3000 transfection reagent (Life Technologies) or Xfect transfection reagent (Takara Bio) following the manufacturer's instructions. EDC3 and EDC3 S161A mutants were cloned into the pLJM vector (GFP fusion) and TRIPZ (full‐length cDNA), and mutated with site‐directed mutagenesis by pfu ultra‐polymerase using PCR. Plasmids purchased from Addgene include pT7‐EGFP‐C1‐HsEDC3 (Addgene plasmid #25032; http://n2t.net/addgene:25032; RRID:Addgene_25032), pT7‐EGFP‐C1‐HsDCP1a (Addgene plasmid #25030; http://n2t.net/addgene:25030; RRID:Addgene_25030), and pT7‐EGFP‐C1‐HsRCK (Addgene plasmid #25033; http://n2t.net/addgene:25033; RRID:Addgene_25033) (Tritschler et al, 2009b). The pcDNA3‐Pim1 plasmid has been described elsewhere (Song et al, 2018). PIM1 CRISPR Guide RNA2 in pLentiCRISPR v2 was purchased from GenScript.

Establishment of stable cell lines with lentivirus infection

Lentiviruses were produced by co‐transfection of the lentivirus vectors with the packaging vectors, psPAX2 and VSV‐G, into HEK‐293T cells. Forty‐eight hours post‐transfection, the medium was concentrated using “Lenti Concentrator” (Origene # TR30025) and used to infect recipient cell lines along with polybrene (8 µg/ml). The infected tumor cells were selected in 1 μg/ml puromycin containing cell culture medium.

Mouse xenografts

SCID mice were purchased from Taconic Laboratories (Germantown, NY). The mice were housed in micro‐isolator cages (Allentown Caging Equipment Company, Allentown, New Jersey) and maintained under specific pathogen‐free conditions. Tumor cell injections in Matrigel (Becton Dickinson) were done subcutaneously (SC) into the left flank of male mice 6–7 weeks of age in a total volume of 100 µl. The tumor volume estimation (mm³) was carried out using the formula (a ² × b/2), where “a” is the smallest diameter and “b” is the largest diameter. All procedures were completed under the University of Arizona Institutional Animal Care and Use Committee (IACUC) protocol.

Immunoblotting

Western blots were performed as described previously (Padi et al, 2019). Briefly, protein lysates (20–40 μg) were subjected to SDS–PAGE and transferred to nitrocellulose membranes. After blocking with 5% milk in TBS‐T buffer (50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 0.05% Tween 20) for 30 min at room temperature, the membranes were incubated overnight at 4°C with the indicated primary antibody in 3% BSA in TBS‐T. Specific proteins were detected with the corresponding HRP‐conjugated secondary antibody by enhanced chemiluminescence (GE Lifesciences, Piscataway, NJ). The signals were detected by X‐ray film and G:BOX Chemi XT4 (SYNGENE).

Phos‐tag SDS–PAGE

The Phos‐tag gel was prepared by following manufacturer’s protocol with Phos‐tag Acrylamide (Wako, Japan). The separation gel was prepared with 12.5 µM of Phos‐tag, 25 µM of ZnCl₂, and 6.5% acrylamide. Bis‐Tris buffer (357 mM, pH 6.8) was used in the resolving and stacking gels. After separation of phospho‐proteins in the running buffer (100 mM Tris, 100 mM MOPS, 0.1% SDS, and 5.0 mM sodium bisulfite), proteins were transferred to membrane with transfer buffer containing 0.1% SDS using the wet method.

Cell viability

PC3‐LN4 cell lines were seeded into 96‐well plates at a density of 5,000 or 10,000 cells per well and pan‐Pim kinase inhibitors (AZD1208 or PIM447), PI3K/AKT inhibitors (BKM‐120, AZD5363, or GSK690693), or combinations of these agents at the indicated doses for 72 h using DMSO as a control. Cell viability was measured using the XTT cell proliferation assay (Trevigen Cat # 4891‐025‐K) following the manufacturer's protocol. After incubation of cell culture with XTT reagent, the absorbance of the colored formazan product was measured at 450 nm.

Cell growth assay

PC3‐LN4 cell lines were seeded into 24‐well plates at a density of 4,000 cells per well for 6 days. The IncuCyte Real‐time imaging system (IncuCyte™) was used to measure cell proliferation using a non‐label cell monolayer confluence approach. The images were captured every 12 h.

Cell invasion and migration assay

Migration assays were performed using 24‐well Corning FluoroBlok cell culture inserts (Corning, Cat# 351152) fitted with an 8 μm pore size light‐blocking PET membrane. For invasion assays, a Corning BioCoat 24‐well plates (Corning, Cat# 354165) fitted with an 8 μm pore size light‐blocking PET membrane coated with Matrigel was used. To study migration (3 × 10⁴ cells/insert) and invasion (5 × 10⁴ cells/insert) of tumor cells, PC3‐LN4 EDC3 KO cells containing a doxycycline (DOX)‐inducible EDC3 wild type or mutant were seeded at a concentration of 3 × 10⁴ in serum‐free RPMI into the upper chamber of the insert. The bottom chamber contained RPMI with 10% FBS. Plates were incubated for 72–96 h at 37°C in a 5% CO₂‐humidified incubator. The vector‐containing cells were preincubated with 100 ng/ml of DOX for 48 h before plating, and the concentration of DOX was maintained in the migration medium. After the termination of the experiment, NucBlue^® Live (Molecular Probes) was added to the bottom chamber, and the chamber incubated for 30 min. Invasion and migration were evaluated by counting the number of cells on the underside of the insert that had migrated/invaded through the membrane following visualization with NucBlue^® Live or GFP. Ten 10× fields were counted per each insert, and the experiments were performed in triplicate. An unpaired t‐test was used to evaluate statistical significance.

P‐body purification

PC3‐LN4 GFP‐EDC3 cells were grown in two 15‐cm dishes per sample to allow for sufficient RNA to be isolated. The cells are washed with 5 ml of cold PBS, then 1.5 ml of cold PBS was added to the dish, and the cells scraped off. Cells were then pelleted at 100 g for 1 min and either analyzed immediately or frozen in liquid nitrogen and stored at −80°C. The following steps were conducted at 4°C on ice. Pellets are suspended in 900 μl of lysis buffer (50 mM Tris, pH 7.4, 1 mM EDTA, 150 mM NaCl, 0.2% Triton X‐100) in the presence of 65 U/ml RNaseOut ribonuclease inhibitor (Promega) and EDTA‐free protease inhibitor cocktail (Roche). Cells were then passed through a needle (25Gx5/8) 20 times and spun for 5 min at 1,000 g to remove nuclear components. The supernatant was then collected and spun at 17,000 g for 10 min. Following centrifugation, the supernatant was removed, and the remaining small pellet resuspended in 500 μl lysis buffer. The resuspended fraction was then spun at 17,000 g for 10 min. The P‐body‐enriched pellet was resuspended in 75 μl lysis buffer and 225 μl of dilution buffer (10 mM Tris, pH 7.4, 0.5 mM EDTA, 150 mM NaCl). The P‐body‐enriched fraction was added to GFP‐Trap Beads^® (Chromotek), which had been equilibrated in these buffers and placed on a nutator for 2 h. The samples were then washed 4× with dilution buffer, and the RNA was extracted overnight from the beads using TRIzol (Invitrogen, 15596–018). The RNA was further purified on a Quick‐DNA/RNA Microprep Plus column (Zymo Research) per the manufacturer's protocol.

RNA extraction

Total RNA was isolated from cells and P‐body samples using TRIzol reagent (Invitrogen, 15596–018) followed by column extraction using Quick‐DNA/RNA Microprep Plus Kit (Zymo Research) per the manufacturer’s protocol. Briefly, cold TRIzol was added directly to the culture dish or GFP‐Trap beads following P‐body purification. The samples were stored at −20°C overnight, thawed on ice, and chloroform was added to the mixture and shaken. This mixture was incubated for 2–3 min and centrifuged at 4°C for 15 min. Following centrifugation, the upper aqueous phase was collected and added to the Quick‐DNA/RNA Microprep Plus Kit column, and the standard protocol was followed.

Quantitative real‐time PCR

RNA was isolated using Quick‐RNA MiniPrep Kit (Zymo Research Cat #R1055), and cDNA was synthesized from 0.1 to 1 μg RNA using the iScript cDNA synthesis kit (Bio‐Rad, Cat # 1708890) following the manufacturer’s protocol. The quantification of real‐time PCR products was performed using SsoAdvanced™ Universal SYBR^® Green Supermix (Bio‐RAD, Cat#1725271) on a CFX96 Real‐Time System. Samples were assayed in triplicate, and RNA levels were normalized to 18S expression levels. Primer sequences for real‐time PCR are provided in Table EV1.

CRISPR‐Cas9 genome editing

The Alt‐R^®CRISPR‐Cas9 system was used as previously described to generate EDC3 knockout (KO) and knock‐in (KI) phospho‐mutants (S161A and S161D) in PC3‐LN4 cells.

For KO, we used dual cutting to remove exon 6, guide crRNA sequence 5′‐ATACTCCTGTTGAATGAGCG‐3′ and 5′‐ATCAACTTGACCCTACTTAC‐3′.

For KI, we generated a single cut dsDNA break by using crRNA guide 5′‐TCGGCACAACTCCTGTAAGT‐3′ and a 50 bp homology arm on each side to enhance the repair.

Protein purification and NMR spectroscopy

DNA coding Pim1 (residues 29–313; ~ 33 kDa) and the central IDR domain of EDC3 (residues 104–197) were sub‐cloned into RP1B (Thio₆His₆‐TEV‐) (Peti & Page, 2007). For protein expression, plasmid DNAs were transformed into Escherichia coli BL21 (DE3) RIL cells (Agilent). Pim1 expression was achieved by growing cells in Luria Broth in the presence of selective antibiotics at 37°C to an OD₆₀₀ of ~ 0.8, and the addition of 1 mM IPTG induced expression. Induction proceeded overnight at 18°C prior to harvesting by centrifugation at 8,000 × g. Cell pellets were stored at −80°C until purification.

For NMR measurements, expression of uniformly (¹H,¹⁵N)‐ or (¹H,¹⁵N,¹³C)‐labeled EDC3 was achieved by growing cells in H₂O based M9 minimal media containing 1 g/l ¹⁵NH₄Cl and/or 4 g/l [¹H,¹³C]‐D‐glucose (CIL or Isotec) as the sole nitrogen and carbon sources, respectively, using established protocols (Peti & Page, 2016). EDC3 expression was achieved by growing cells at 37°C to an OD₆₀₀ of ~ 1.0, and the addition of 1 mM IPTG induced expression. Induction proceeded for ~ 4 h at 37°C. Cell pellets were lysed in lysis buffer (25 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole, 0.1% Triton X‐100) containing EDTA‐free protease inhibitor cocktail (Roche) using high‐pressure homogenization (Avestin C3). The lysate was clarified by centrifugation at 42,000 × g and filtered through a 0.22‐μm PES filter before loading onto a His‐Trap HP column (GE Healthcare). Bound proteins were washed with Buffer A (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole) and eluted with increasing amounts of Buffer B (50 mM Tris pH 8.0, 500 mM NaCl, 500 mM imidazole) using a 5–500 mM imidazole gradient. Peak fractions were pooled and dialyzed overnight at 4°C in dialysis buffer (50 mM Tris pH 7.5, 500 mM NaCl, 0.5 mM TCEP) with 5:1 volume ratio of TEV protease overnight. The next day, a “subtraction” His₆‐tag purification was performed to remove TEV and the cleaved His₆‐tag. EDC3 was concentrated to ~ 6 ml, 5 mM (final concentration) DTT added, and further heat purified (80°C; 10 min). Pim1 and EDC3 were further purified using Size Exclusion Chromatography (SEC, Superdex 75 26/60) in NMR Buffer (20 mM Na₂HPO₄ pH 6.2, 500 mM NaCl, 0.5 mM TCEP). Purified protein was either used immediately or flash‐frozen in liquid nitrogen for storage at −80°C.

NMR data were collected on Bruker Avance NEO 600 MHz, and 800 MHz spectrometers equipped with TCI HCN z‐gradient cryoprobes at 298 K. NMR measurements of EDC3 were recorded using (¹H,¹⁵N)‐ or (¹H,¹⁵N,¹³C)‐labeled protein at a final concentration of 0.7 mM in NMR buffer and 90% H₂O/10% D₂O. The sequence‐specific backbone assignments of EDC3 were achieved using 3D triple resonance experiments including 2D [¹H,¹⁵N] HSQC, 3D HNCA, 3D HN(CO)CA, 3D HN(CO)CACB, 3D HNCACB, 3D HNCO, 3D HN(CA)CO, and 3D HNH. All NMR data were processed using TopSpin 4.05 (Delaglio et al, 1995) and analyzed using CARA (http://www.cara.nmr.ch).

In vitro kinase assay

EDC3 and Pim1, Pim3, or AKT proteins were purified as described above and used in an in vitro kinase assay. 200 ng of Pim1 and Pim3 or 100 ng of AKT was added to reaction buffer (20 mM Tris pH 8.0, 200 mM NaCl, 2 mM CaCl₂, 25 mM MgCl₂, 5 mM DTT, 0.05% Tween 20, and 0.1% BSA) followed by 200 ng of purified EDC3 full‐length protein and 2 μM ATP and incubated for 1 h at 30°C. The addition of 2× SDS–PAGE buffer terminated the reaction followed by boiling for 10 min.

Immunofluorescence

Immunofluorescence (IF) staining was performed on PC3‐LN4 cells. 1 × 10⁴ cells were seeded in each chamber of an 8‐well chamber slide (Nunc Lab‐Tek II Chamber Slide System, Life Technologies) and cultured for 48 h. The cells were incubated with DMSO, PIM447, AKTi, or a combination for 6 h in normal culture conditions or 2 h in Hank’s balanced salt solution (HBSS, Life Technologies). Cells were fixed with 4% paraformaldehyde in PBS for 15 min and subsequently permeabilized with chilled methanol in the freezer (−20°C) for 15 min, followed by blocking with 2% normal goat serum in PBS for 15 min. Samples were incubated with the primary antibody in 1% normal goat serum/PBS; 1:200 for EDC3 (F‐9: sc‐271806, Santa Cruz), and DCP1a (ab47811; abcam), 1:300 P‐EDC3 (Rockland) overnight at 4°C and then incubated with secondary antibody (1:2,000) goat anti‐Mouse IgG1, Alexa Fluor 568, Goat anti‐Rabbit IgG (H + L) PE‐Alexa Fluor 647 (Life Technologies), and DAPI (5 ng/ml) in 2% goat serum for 90 min at room temperature. Specimens were mounted with ProLong Diamond Antifade Mountant (Molecular probes). The images were acquired using upright and inverted fluorescence microscopes (BX50 or IX71, Olympus), which were equipped with UPlan F1 40× and UPlanApo 60× objective lens and a camera (DP72; Olympus). The images were acquired by the cellSens software (Olympus). The images (4,140 × 3,096 pixels) were taken and analyzed by ImageJ software. All images are applied pseudocolor by the cellSens and ImageJ software.

Immunohistochemistry

Immunohistochemistry was performed on paraffin‐embedded tissues. Briefly, the tissue was baked and dewaxed for 30 min at 55°C. Following this, the sample was treated with tri‐EDTA at 98°C for 15 min to allow for epitope retrieval and blocked with endogenous peroxide for five min at room temp (RT). The antibody was applied for 15 min at RT. An HRP‐conjugated polymer was applied for 8 min at RT, followed by DAB for 10 min at RT. A second‐round Tris‐EDTA and peroxide block was applied for the secondary antibody (rabbit or mouse) for 20 min at RT. AP polymer is added for 30 min at RT, and DAB is reapplied as before. Hematoxylin counterstain was applied for 5 min at RT, and the slides were removed from the Leica Bond RX, dehydrated in ethanol, cleared with xylene, and coverslipped. The images were acquired using an upright fluorescence microscopy (BX50, Olympus), which was equipped with UPlan 20× objective lens and a camera (DP72; Olympus). The images (4,140 × 3,096 pixels) were acquired by the cellSens software (Olympus).

Co‐immunoprecipitation

Cells were transfected with plasmids using lipofectamine 3000 or Xfect transfection reagent in HEK293T cells per the manufacturer's instructions. GFP‐tagged proteins were co‐immunoprecipitated using GFP‐Trap MA beads (Chromotek) following manufacturer’s protocols. Cells were harvested using lysis buffer (10 mM Tris/HCl pH 7.5; 150 mM NaCl; 0.5 mM EDTA; 0.5% NP‐40). The lysates were then incubated with 25 µl of GFP‐Trap beads, rotated for 1 h at 4°C, and magnetically separated. The beads were washed 3× with lysis buffer and boiled in a 2× SDS‐sample buffer for 10 min. Co‐precipitation with biotin‐labeled Pim1 was performed with Dynabeads™ M‐280 Streptavidin (Invitrogen). 2 μl of recombinant EDC3 and 0.5 μg Pim protein were incubated in 300 μl of binding buffer (1% Triton X, 50 mM HEPES pH 7.5; 150 mM NaCl) and incubated at RT for 10 min. 20 μl of washed bead slurry in the binding buffer was added to each sample and rotated for 2 h at 4°C. Beads were washed 4× with binding buffer and boiled in 2× SDS–PAGE loading buffer for 10 min at 95°C. Co‐IP with cell lysates for endogenous proteins was performed with anti‐EDC3 antibody and Protein G Mag Sepharose (GE Healthcare) beads following the manufacturer's “classic” protocol. 25 μl of bead slurry is washed per sample and equilibrated in buffer, and 10 μl of antibody is added to beads and rotated for at least 15 min. Beads are washed once in binding buffer, and protein lysate is added. The sample was rotated for 1 h at 4°C, followed by washing three times with the binding buffer and elution with 2× SDS–PAGE loading buffer with boiling for 10 min at 95°C.

RNA‐sequencing and bioinformatic analyses

Strand‐specific mRNA‐enriched RNA‐seq libraries were prepared (BGI, San Jose). Paired‐end Illumina RNA‐sequencing was performed in triplicate on a BGISEQ‐500. Reads were processed and mapped to the human reference genome (GRCh38.97) with accompanying reference annotation using the RMTA workflow (Peri et al, 2020) with default parameters. RNA‐sequencing data have been uploaded to NCBI's SRA database under project ID PRJNA615885. The percentage of mapped reads is shown in Table EV2. Read counts per gene were determined using feature Counts in Subread v 1.6.3 (Liao et al, 2014). Differentially expressed genes were identified using DESeq2 in R. Differentially expressed genes in pairwise comparisons with an adj. P‐value of 0.001 or less were examined for GO‐enrichment using Gorilla's online tool (Eden et al, 2009). 5′UTR sequences of differentially expressed transcripts were extracted from the reference genome using gffread and examined for motif enrichment using the Discriminative Regular Expression Motif Elicitation (DREME) software v5.1.1 with both shuffled and non‐enriched sequences as background controls (Bailey et al, 2009). Changes in p‐body RNA composition between samples relative to whole‐cell RNA populations were performed using anota2seq package in R, substituting p‐body mRNAs for translated mRNA and utilizing a max adjusted P‐value of 0.1 (Oertlin et al, 2019). The ATtRACT online database (https://attract.cnic.es/) was utilized to identify RNA binding proteins that recognize our RNA‐seq associated motifs; identified proteins were then compared to significantly enriched proteins in the P‐body proteome of Hubstenberger et al, 2017. Biogrid (https://thebiogrid.org/) was used to verify if these proteins were known EDC3 interactors.

Statistics

Values reported and shown in graphical displays are the mean ± standard deviation or standard error of the mean. As indicated. P‐body images were analyzed using ImageJ software. All images were analyzed in triplicate. After counting P‐bodies, they were then expressed as “P‐body number” per cell. Comparisons of the mean expression across groups were made using an unpaired 2‐tailed Student's t‐test. For all comparisons, P‐values less than 0.05 were considered statistically significant.

Author contribution

WP, ADLN, JRB, SKRP, ASK, and KO conceptualized the study. WP, ADLN, JRB, AEC, ASK, and KO acquired the funding. JJB, SKRP, NS, MC‐V, JHS, GM, NF, YL, MRH, and JMCG were involved in the investigation. WP, JRB, AEC, ASK, and KO supervised the study. WP, ASK, and KO performed validation. JB, SKRP, NS, and MC‐V wrote the original draft. WP, JRB, ADLN, SKRP, ASK, and KO reviewed and edited the draft.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 5

Source Data for Figure 7

Source Data for Figure 8

EMBO Reports (2021) 22: e50835.

Data availability

All NMR chemical shifts have been deposited in the BioMagResBank (BMRB: 28100; http://www.bmrb.wisc.edu/data_library/summary/index.php?bmrbId=28100). RNA‐sequencing data have been uploaded to NCBI's SRA database under project ID: PRJNA615885.