Gld2 activity and RNA specificity is dynamically regulated by phosphorylation and interaction with QKI-7

ABSTRACT

In the cell, RNA abundance is dynamically controlled by transcription and decay rates. Posttranscriptional nucleotide addition at the RNA 3ʹ end is a means of regulating mRNA and RNA stability and activity, as well as marking RNAs for degradation. The human nucleotidyltransferase Gld2 polyadenylates mRNAs and monoadenylates microRNAs, leading to an increase in RNA stability. The broad substrate range of Gld2 and its role in controlling RNA stability make the regulation of Gld2 activity itself imperative. Gld2 activity can be regulated by post-translational phosphorylation via the oncogenic kinase Akt1 and other kinases, leading to either increased or almost abolished enzymatic activity, and here we confirm that Akt1 phosphorylates Gld2 in a cellular context. Another means to control Gld2 RNA specificity and activity is the interaction with RNA binding proteins. Known interactors are QKI-7 and CPEB, which recruit Gld2 to specific miRNAs and mRNAs. We investigate the interplay between five phosphorylation sites in the N-terminal domain of Gld2 and three RNA binding proteins. We found that the activity and RNA specificity of Gld2 is dynamically regulated by this network. Binding of QKI-7 or phosphorylation at S62 relieves the autoinhibitory function of the Gld2 N-terminal domain. Binding of QKI-7 to a short peptide sequence within the N-terminal domain can also override the deactivation caused by Akt1 phosphorylation at S116. Our data revealed that Gld2 substrate specificity and activity can be dynamically regulated to match the cellular need of RNA stabilization and turnover.

1. Introduction

The terminal nucleotidyltransferase Gld2 (germline development 2, PAPD4, TUT2 or TENT2) is a non-canonical poly (A) polymerase (PAP) first identified in Caenorhabditis elegans, where its main function is the extension of short mRNA poly(A) tails to prolong RNA lifespan[1] and regulation of meiosis and germline development [2]. Gld2 homologs were subsequently identified in other eukaryotes, including mammals [3–6], where the activity of Gld2 extends to the 3ʹ adenylation of microRNAs (miRNAs) [7]. Gld2 encodes a PAP-associated and nucleotide transferase domain (NTD) with catalytic carboxylates and a nucleotide binding pocket[1]. ATP specificity is determined by the active site composition, and a single histidine insertion, which is conserved in uridylyltransferases [8–10], converts Gld2 from an adenylyltransferase to an uridylyltransferase [11]. Similar to its uridylyltransferase homolog Cid1 from fission yeast, C. elegans Gld2 binds to poly(A) tails of mRNAs via a positively charged groove or surface area to add nucleotides to the poly(A) tail [8–10,12,13]. The recently solved crystal structure of rodent Gld2 indicates that the expanded catalytic activity towards RNAs other than poly(A) tails can be attributed to a change in the surface charge of the RNA binding groove, leading to a reduced specificity towards A-residues and allowing for the adenylation of more diverse substrates [14].

Recent studies revealed a role for mammalian Gld2 catalysed monoadenylation as a determinant of miRNA stability, where the addition of a mono(A) to the 3ʹ end of a mature miRNA extends miRNA lifespan and stability [15]. In mouse liver cells, 3′ adenylation catalysed by Gld2 protects the miRNA miR-122 from exonuclease activity [16]. Similarly, in human fibroblasts, Gld2 induced monoadenylation prohibits decay of miR-122 [17]. In humans, miR-122 is the most abundant liver-specific small RNA and is essential for Hepatitis C virus (HCV) infection [18]. During HCV infection, miR-122 is recruited to the 5′ untranslated region of the viral RNA and stimulates viral replication [19]. The HCV core protein binds to Gld2 and reduces miR-122 adenylation, allowing the virus to tightly control miR-122 levels and viral replication [20]. Human Gld2 can adenylate a variety of RNA substrates in vitro, ranging from pre-miRNA to mature miRNA and polyadenylated RNA [11]. In mammalian cells, Gld2 catalyzes both miRNA monoadenylation and mRNA polyadenylation to dynamically control RNA stability, raising the question of how RNA substrates are recognized and Gld2 activity is regulated. Our lab recently showed that Gld2 activity is regulated by post-translational phosphorylation in the disordered N-terminal domain [21]. The oncogenic kinase Akt1 phosphorylates Gld2 at S116 to drastically reduce enzyme activity, while an unknown kinase phosphorylates S62 in cell extracts, leading to increased Gld2 activity [21]. Other phosphorylation sites shifted substrate affinity to mRNA or miRNA [21].

Substrate specificity of human Gld2 can be directed by interacting with the RNA-binding proteins QKI-7 (Quaking-7) [22,23] and CPEB (cytoplasmic polyadenylation element-binding protein) [17]. QKI-7 is a member of the signal transduction and activation RNA (STAR) protein family. The three major QKI isoforms, QKI-5, QKI-6, and QKI-7 [24,25], mainly differ in their C-terminal sequence [26]. p53 controls QKI gene expression and, in turn, QKI acts as glioblastoma multiforme tumour suppressor gene. Depletion of all QKI isoforms reduces the level of miR-20a in human cells, leading to elevated levels of the miR-20a target TGFBR2, inducing oncogenesis [27]. The QKI-7 isoform is predominantly localized in the cytoplasm and was shown to recognize target mRNAs via a small recognition motif on the target RNA known as the QKI response element (QRE) [28]. Recent studies showed that QKI-7 binds to the N-terminal domain of Gld2 via its QKI-7 isoform specific C-terminal domain and acts as a linker to connect Gld2 to Argonaut-2 (Ago2) in the RNA-induced silencing complex for adenylation and stabilization of miR-122^22, [23]. A second RNA binding protein, CPEB, was shown to recruit Gld2 and another non-canonical PAP, Gld4, to control polyadenylation-dependent translation of the tumour suppressor protein p53¹⁷. Here, Gld2 controls miR-122 stability, modulating CPEB expression in turn. CPEB then recruits Gld4 to the p53 mRNA for polyadenylation and translation [17]. It is now apparent that Gld2 activity and substrate specificity is regulated by both post-translational phosphorylation and by interactions with RNA binding proteins.

To investigate the interplay between Gld2 post-translational phosphorylation and the effect of RNA binding proteins on substrate specificity and activity, we examine how a combination of these two factors influences activity. We probed three previously identified RNA binding proteins (QKI-7, CPEB, and Lin28A) that interact with Gld2 or the nucleotidyltransferase homologs TENT3a/b. While CPEB and Lin28A mostly decreased Gld2 activity, QKI-7 generally increased the activity of the Gld2 variants. We further show that QKI-7 can shift the ability of Gld2 to recognize specific RNA substrates, depending on the phosphorylation status of Gld2, and that the combination of QKI-7 and phosphorylation status can differentially alter Gld2 activity and RNA substrate specificity. Finally, we demonstrate that QKI-7 binds to specific regions in the disordered N-terminal domain of Gld2. Our results show that QKI-7 and post-translational phosphorylation work together to finetune the adenylation activity and RNA substrate specificity of Gld2.

2. Results

2.1. Cell extracts from HEK 293 cells phosphorylate a Gld2 derived peptide

We have previously shown that Gld2 activity is regulated by phosphorylation of its N-terminal domain [21]. While multiple phosphorylation sites were reported in the N-terminal domain (Figure 1(a)), we most notably identified S116 as a target of the oncogenic kinase Akt1, where phosphorylation abolishes Gld2 activity towards both mRNAs and miRNAs [21]. We now investigated whether this activity can be observed in cell extracts. We first tested Akt1 activity in vitro by incubating recombinant, doubly phosphorylated active Akt1 with [γ-]³²P]-ATP and a peptide corresponding to amino acids 106–125 of Gld2 (peptide P8, Figure. 1(a), Table 1) or a GSK-3b positive control peptide (Figure 1(b)). Successful phosphorylation is signified by the transfer of the radiolabeled γ-phosphate from ATP to the peptide as described before [29]. Recombinant active Akt1 was produced as described before [29]. As expected, Akt1 phosphorylated P8, albeit with less efficiency than the control GSK-3b peptide. Next, we incubated Akt1 overexpressing cell (Figure S1) extracts with the GSK-3b, P8, and a negative control peptide (P4, corresponding to amino acids 46–65 of Gld2, Table 1), which is not an Akt1 substrate, as well as a cell extract only/no peptide control and an Akt1 enzyme only control (Figure 1(c,d)). We have previously demonstrated that pS62 (site found in peptide P4) was only detected when Gld2 was incubated with EGF-stimulated cell lysates, but not in unstimulated cell lysates. We also showed that HEK 293 cell extracts do not phosphorylate Gld2 at S116 in the absence of an Akt1 overexpressing plasmid [21]. As expected, P4 was not phosphorylated above background levels, but both P8 and the GSK-3b peptides were phosphorylated by the extracts from Akt1 overexpressing cells, which showed consistently high expression of mCherry-Akt1 (Figure 1(c), S1). These data confirm that Akt1 indeed phosphorylates Gld2 in a cellular context, making post-translational phosphorylation a powerful determinant of Gld2 activity, and thus RNA stability.

Figure 1.

HEK 293T cell extracts phosphorylate a Gld2 peptide. A) Gld2 encodes a nucleotidyltransferase (NTD), a poly(A) polymerase-associated domain (PAP), and a disordered N-terminal domain. The N-terminal domain (148 amino acids (aa)) was divided into 20 amino acid long peptides that overlap each other by 5 amino and are termed peptides P1–P9 (see Table 1). Serine phosphorylation sites are noted. B) Fully activated recombinant Akt1 was incubated with Gld2 peptide P8 and a control peptide. Akt1 phosphorylates a GSK-3b peptide (positive control) and the Gld2-P8 peptide in vitro to form radiolabeled [³²P]-peptide. C) Cell extract from HEK 293 T cells expressing Akt1 phosphorylate a GSK-3b peptide (positive control) and the Gld2-P8 peptide containing the Akt1 phosphorylation site S116, but not the Gld2-P4 peptide. D) 15-minute time points from Figure 1c were plotted showing that Gld2-P4 is not phosphorylated significantly above background (cell extract), and that Gld2-8 and GSK-3b (positive control) are efficiently phosphorylated. A no peptide control using recombinant Akt1 (enzyme) showed no activity. Error bars represent one standard error calculated from triplicate reactions. Significant changes calculated using a two-tailed t-test are indicated by asterisks. p ≤ 0.05 (*); p ≤ 0.01 (**); p ≤ 0.001 (***); p ≤ 0.0001 (****)

Table 1.

Gld2 peptides used in this study

Peptide Name	Amino acid range	Amino acid sequence
P1	1–20	MFPNSILGRPPFTPNHQQHN
P2	16–35	HQQHNNFFTLSPTVYSHQQL
P3	31–50	SHQQLIDAQFNFQNADLSRA
P3/4	39–57	QFNFQNADLSRAVSLQQLT
P4	46–65	DLSRAVSLQQLTYGNVSPIQ
P5	61–80	VSPIQTSASPLFRGRKRLSD
P6	76–95	KRLSDEKNLPLDGKRQRFHS
P7	91–110	QRFHSPHQEPTVVNQIVPLS
P8	106–125	IVPLSGERRYSMPPLFHTHY
P9	121–140	FHTHYVPDIVRCVPPFREIA

2.2. RNA-binding proteins affect Gld2 catalytic activity

Human Gld2 binds RNA via positively charged surface areas [14], but RNA substrate specificity in large relies on interactions with RNA-binding proteins. While we previously showed that Gld2 affinity towards RNA substrates can be altered by phosphorylation in the N-terminal domain, these post-translational modifications mostly affect the enzymatic activity and have less effect on RNA specificity [21]. Several RNA-binding proteins have been reported to interact with Gld2 or its homologs: CPEB and QKI-7 were shown to interact with Gld2 [5,22,30] and Lin28A interacts with the human nucleotidyltransferase homologs TENT3a and TENT3b [31,32]. Most interestingly, QKI-7 was shown to enhance Gld2 specificity towards miR-122 stabilization [23]. In an initial screen to identify general trends in the interplay between Gld2 phosphorylation and RNA binding proteins on Gld2 activity, we tested the ability of three RNA binding proteins on Gld2 wildtype (WT) and phosphomimetic variant activity (Figure 2, S4). Previous reports showed increased Gld2 activity in the presence of the RNA binding proteins QKI-7 [23] and CPEB [17], and we expect that this increased activity will be observed in activity assays with WT Gld2 enzyme and RNA binding proteins. Considering that phosphorylation of Gld2 can either reduce (S116) or enhance (S62) Gld2 activity [21], it is unclear what the cumulative effect of post-translational modification and RNA binding proteins have on enzymatic activity. We hypothesize that in general, an RNA binding protein that increases Gld2 WT activity would also increase the activity of the phosphomimetic variants. The presence of the phosphomimetic residue may however influence the ability of the RNA binding proteins to interact with Gld2 and/or recognize the RNA substrate. Thus, the presence of the RNA binding proteins could well result in an increase, decrease, or no change in phosphomimetic variant activity. All proteins were purified to apparent homogeneity (Figure S2, and as shown before [21]) and assayed in equimolar concentrations to Gld2. The phosphomimetic Gld2 variants represent previously identified Gld2 phosphorylation sites in the N-terminal domain. Each Gld2 variant (WT, S62E, S69E, S95E, S110E, S116E, and a S62E/S116E double mutant) was tested in combination with three individual RNA binding proteins (CPEB, QKI-7, and Lin28A) and four different RNA substrates (Table 2) to identify general trends in changes of adenylation activity (Figure 2, S4). The miRNA miR-122 and the poly(A) tail mimic of 15 adenine residues (15A RNA) were chosen based on previous studies showing them to be suitable Gld2 substrates [11,21]. The QKI-7 RNA response element (QRE) and CPEB cytoplasmic polyadenylation element (CPE) were designed based on previous studies on CPEB and QKI-7 binding sites [33,34].

Figure 2.

Interacting proteins modulate the catalytic activity of wildtype Gld2 and phosphomimetic variants. Activity assay gels of wildtype Gld2 and glutamic acid variants with an interacting protein. A) Gld2 WT, or B) and C) glutamic acid variants of known Gld2 phosphorylation sites were incubated with [α-³²P]-ATP in the absence or presence of an interacting protein (CPEB, QKI-7, or Lin28A) with one of the following RNA substrates: CPE (45 nts), QRE (11 nts), miR-122 (22 nts), or 15A RNA (15 nts). Reactions were incubated at 37°C for 15 minutes and analysed via gel electrophoresis and phosphorimaging. Reaction products were quantified by exposing a Whatman filter strip dotted with known concentrations of [α-³²P]-ATP to the same phosphoscreen as the gel. The fold change in product compared to product formed in the absence of any interacting protein is stated below the gels. Representative full-sized gels can be found in Figure S4.

Table 2.

RNA substrates used in activity assays

Name	Sequence
miR-122	5ʹ(p)-CAAACACCAUUGUCACACUCCA-3’
15A	5ʹ(p)-AAAAAAAAAAAAAAA-3’
QRE	5ʹ(p)-UUCACUAACAA-3’
CPE	5ʹ(p)-CUUUUUAUAUCCCAUUUUUAUAUCGAUCUCUUAUUUUACAAUAAA-3’

We first surveyed the Gld2 WT protein (Figure 2(a)). Various RNA substrates were incubated with Gld2 in the presence or absence of the RNA-binding proteins. Adenylation of RNA substrates was monitored by using [α-³²P]-ATP, where transfer of the radiolabeled ATP to the RNA 3ʹ end was visualized by subsequent gel analysis and phosphorimaging. Addition of CPEB did not alter adenylation activity with miR-122 and 15A substrates but led to a decrease in activity by 2-fold with CPE or 10-fold with QRE. Addition of QKI-7 resulted in either unchanged or 2-fold reduced activity (QRE), while addition of Lin28A reduced activity towards three substrates (miR-122, QRE, and CPE), but not 15A RNA. The activity of Gld2 on 15A RNA is generally lower than the activity with miR-122, as we showed previously [21], and remained largely unchanged regardless of the RNA binding protein added.

Next, we assayed Gld2 protein variants S62E and S116E (Figure 2(b)). We previously showed that S62 is phosphorylated in cell extracts by an unknown kinase [21]. Gld2 S116 is phosphorylated by recombinant Akt1 [21] and in cell extracts overexpressing Akt1 (Figure 1). The phosphomimetic variants encode glutamic acid in the reported serine phosphorylation sites and we have previously shown that these phosphomimetic variants have altered catalytic activity compared to the wildtype and are comparable in activity to the phosphorylated protein variants [21]. We previously showed that S62 phosphorylation increases Gld2 activity by 4- to 6-fold, while S116 phosphorylation, and a phosphomimetic of the doubly phosphorylated S62/S116 almost abolishes Gld2 activity [21]. Here, we show that the Gld2 S62E/S116E double variant incubated with CPE in the presence of QKI-7 displayed the greatest activity increase with a roughly 3-fold increase and QKI-7 addition increases the activity of most other Gld2 variants by up to 2-fold (Figure 2(b)). Addition of CPEB or Lin28A in most cases had little overall effect on Gld2 activity or led to a reduction in activity, which might be attributed to the RNA binding proteins’ ability to sequester the RNAs without promoting adenylation.

Finally, we surveyed the phosphomimetic variants S69E, S95E, and S110E for their activity with the RNA binding proteins and RNA substrates (Figure 2(c)). These specific mutants showed only minimal changes in activity compared to WT Gld2 [21]. While overall activity remained very similar, we observed the general trend that QKI-7 addition increases the activity in most substrate/Gld2 variant combinations.

A general trend in this survey was that the addition of QKI-7 to Gld2 variants lead to an increase in product formation, the highest increase being a 3-fold increase of the S62E/S116E on CPE RNA, restoring this catalytically impaired variant to some extent (Figure 2(b)). This indicates that the interaction between QKI-7 and the various phosphorylated states of Gld2 can dynamically regulate Gld2 activity and direct Gld2 towards certain substrates. Addition of Lin28A in most cases led to a reduction in activity towards almost all RNA substrates, except for S95E activity on mRNA, which resulted in a moderate 2.5-fold increase (Figure 2(c)). We previously showed that the S95E mutant has a relatively low affinity towards mRNA [21] and Lin28A may have a compensating effect. It is possible that Lin28A sequesters the substrate RNAs without interacting with Gld2, leading to generally reduced substrate availability. Addition of CPEB to enzymatic reactions led to a moderate increase in activity, up to 1.7-fold with Gld2 S62E/S116E and CPE (Figure 2(b)). However, CPEB mostly decreased the activity of the Gld2 variants with the greatest reduction to 0.1-fold for Gld2 WT and QRE (Figure 2(a)).

In summary, incubation of CPEB with any Gld2 variant led to mixed results with CPE and miR-122, while a decrease in Gld2 activity was always observed with QRE and 15A RNA (Figure 2). The observed changes in activity in the presence of interacting proteins are dependent on the Gld2 variant and RNA substrate, indicating a complex interplay of Gld2 phosphorylation status, RNA binding protein, and RNA substrates. Since QKI-7 mostly had an enhancing effect on Gld2 activity, we decided to investigate the interplay between QKI-7 binding and Gld2 phosphorylation status on Gld2 adenylation activity in more detail.

2.3. QKI-7 shifts Gld2 substrate preference depending on the phosphorylation status of Gld2

QKI-7 was previously reported to bind to the N-terminal disordered domain of Gld2 [22], thus we constructed a Gld2 WT variant lacking this region (amino acids 1–147, Gld2 ΔN). This Gld2 ΔN variant was used as a negative control in the assays below. Here, we performed a competitive RNA assay to determine if Gld2 prefers certain RNA substrates over others when incubated with a mixture of RNAs, and whether the presence of QKI-7 will alter Gld2 substrate specificity or enhance Gld2 activity in general (Figure 3). In the assay, all RNA substrates (CPE, QRE, miR-122, and 15A) were present in a reaction with QKI-7 and a Gld2 variant. While Gld2 displayed activity towards all four substrates in single-substrate assays (Figure 2(a)), the main product bands in the competitive assay correspond to CPE and miR-122 adenylated products in the absence of QKI-7 in competition assays, indicating a preference for these RNAs over 15A and QRE (Figure 3(a)). Upon the addition of QKI-7, the activity of Gld2 increased between 1.2 and 5.3-fold on these two substrates, depending on the Gld2 variant. Interestingly, we did not observe any adenylation products for the QRE or 15A RNA substrates. In the competitive assay, addition of QKI-7 increased Gld2 adenylation activity with both CPE and miR-122 substrates (Figure 3(b,c)). Gld2 WT displayed the greatest increase in activity for both RNAs with the greatest change observed for CPE. S62E, S69E, S95E, and S110E adenylation activity was also enhanced by the addition of QKI-7. However, the activity of these variants slightly shifted towards the adenylation of CPE, while S116E remained largely unchanged. The double mutant S62E/S116E displayed an equal increase in activity towards both substrates, miR-122 and CPE. It is notable that the activity of Gld2 ΔN is enhanced compared to the other Gld2 variants and independent of QKI-7.

Figure 3.

Gld2 variants are more active on specific RNA substrates. Competitive RNA activity assay gels of wildtype Gld2 and glutamic acid variants with an interacting protein. In a competitive RNA assay where various RNA substrates would compete for Gld2 binding, only RNA substrates that Gld2 favourably binds to will be adenylated and visualized. Gld2 variant and/or QKI-7 was incubated with [α-³²P]-ATP and 4 different RNAs (CPE, QRE, miR-122, and 15A RNA) present in the reaction. Reactions were incubated at 37°C for 15 min and analysed via gel electrophoresis and phosphorimaging, shown with A) a representative gel of Gld2 WT activity. Reaction products were quantified by exposing a Whatman filter strip dotted with different known concentrations of [α-³²P]-ATP to the same phosphoscreen as the gel. The fold change in product compared to the absence of QKI-7 is stated below the gels for B) CPE (45 nts) and C) miR-122 (22 nts)

In summary, these data confirm that Gld2 activity generally increases in the presence of QKI-7, and that the phosphorylation status of Gld2 fine tunes this change in activity. Furthermore, our data indicates that QKI-7 does indeed require the Gld2 N-terminal domain to foster the protein-protein interaction, and that the disordered N-terminal may have an inhibitory effect on Gld2 activity.

2.4. The interplay of QKI-7 binding and Gld2 phosphorylation status dictates adenylation activity and RNA specificity

To fully understand the effects of QKI-7 on the activity of Gld2 with different RNA substrates, QKI-7 and different Gld2 variants were incubated with a single RNA substrate (Figure 4, gels shown in Fig S3, S4). Product formation was measured as an end-timepoint reaction and the fold-change in the activity was calculated from triplicate reactions. The focus was initially on CPE and miR-122 as the competitive RNA assay indicated that all Gld2 variants preferred those two RNA substrates in the absence or presence of QKI-7.

Figure 4.

QKI-7 increases the catalytic activity of Gld2 variants. Bar graphs showing the relative activity (%) of Gld2 variants in the absence and presence of QKI-7 with A) miR-122 (22 nts), B) 15A RNA (15 nt), C) CPE (45 nts), or D) QRE (11 nts). Error bars represent one standard error calculated from triplicate reactions. Significant changes calculated using a two-tailed t-test are indicated by asterisks. p ≤ 0.05 (*); p≤ 0.01 (**). Fold changes were calculated using activity assay gels from Figure S3.

We first assayed Gld2 variants with and without QKI-7 with miR-122 (Figure 4). Previous studies showed that Gld2 adenylates miR-122 in vivo and that QKI-7 directs Gld2 towards miR-122 in collaboration with the Ago2 protein to increase miR-122 stability [20,23]. Indeed, we observed an increase in miR-122 adenylation for almost all Gld2 variants (Figure 4(a)). Gld2 WT displayed a 2.9-fold increased activity towards miR-122 in the presence of QKI-7. Interestingly, the S62E/S116E mutants, which is almost inactive in the absence of QKI-7, showed somewhat restored enzymatic activity, with a 4.7-fold increased activity. The S69E, S110E, and S116E mutants displayed between 2.2 to 2.8-fold increased activity, while S62E and S95E showed no significant changes in activity. As expected, the Gld2 ΔN mutant was not affected by QKI-7 addition. These data show that while QKI-7 can increase the adenylation activity of Gld2 towards miR-122, this effect can be reduced by Gld2 phosphorylation at S62 or S95.

Next, we assayed the other preferred substrate from the competition assay, CPE RNA (Figure 4(b)). CPE RNA is not generally known to be a specific motif recognized by QKI-7 but was a preferred substrate in our competition assay (Figure 3). Again, we observed an increase in adenylation activity for some variants (Figure 4(b)). Gld2 WT, S62E, S95E, and S116E had over 1.7-fold increase in activity with CPE. It is notable that S62E and S95E did not change in activity towards miR-122 but displayed enhanced activity towards CPE upon QKI-7 addition. Gld2 S116E showed the greatest increase in activity with CPE of approximately 5-fold. The double mutant S116E/S62E, which showed enhanced activity towards miR-122, did not display a significant change in activity on CPE RNA. Gld2 ΔN activity was again unchanged.

The assay was repeated with 15A RNA using only Gld2 WT, S62E, S116E, S62E/S116E, and ΔN (Figure 4(c)), as we showed that S116 is an Akt1 phosphorylation site (Figure 1 and [21]) and S62E is phosphorylated in cell extracts [21]. While Gld2 activity on 15A RNA is generally low and 15A RNA is not significantly adenylated in competition assays (Figure 3), S62E displayed significantly increased activity while S116E and S62E/S116E showed significantly decreased activity [21]. In the presence of QKI-7, only Gld2 S116E displayed a significant increase in activity towards 15A RNA (1.4-fold, Figure 4(c)), indicating that Gld2 activity towards mRNAs is in general not impacted by QKI-7 binding, unless Gld2 is phosphorylated by Akt1, in which case QKI-7 can restore some activity towards mRNAs.

Finally, we assayed the QRE RNA (Figure 4(d)), a small RNA motif previously reported to be recognized by QKI-7 [22], leading to the recruitment of Gld2 to adenylate and stabilize mRNAs. In the presence of QKI-7, Gld2 WT, S62E, S116E, and S62E/S116E displayed significantly increased activity of up to 2.5-fold with QRE. A slight, but significant decrease in activity was observed for Gld2 ΔN (86%), indicating that the N-terminal domain is required for QRE recognition, and QKI-7 possibly sequesters QRE RNA, making it unavailable for adenylation. Overall, while Gld2 prefers to adenylate miR-122 over QRE (Figure 3), QKI-7 increases Gld2 activity towards QRE substrates regardless of Gld2 phosphorylation status but is instead dependent on the N-terminal domain of Gld2 (Figure 4(d)).

In summary, our data confirms that both Gld2 phosphorylation and the interaction with QKI-7 determine enzymatic activity on the four RNA substrates, allowing the cell to fine tune activity and specificity dependent on QKI-7 binding and kinase activity.

2.5. QKI-7 binds to specific regions in the disordered N-terminal domain of Gld2

QKI-7 has been reported to bind in the disordered N-terminal region of Gld2 [22]. To identify the binding region(s) of QKI-7, the N-terminal region of Gld2 (amino acid 1–140) was divided into 9 peptides (P1 - P9), each 20 amino acids long with a 5 amino acid overlap between adjacent peptides (Table 1). The peptides were labelled on the 5ʹ-end with 6- carboxyfluorescein (6-FAM) and fluorescence anisotropy was used to determine the binding of QKI-7 to each of the 9 peptides (Figure 5). A decrease in depolarized emission indicates binding between the Gld2 peptide and QKI-7. Enhanced binding was observed with P3 (amino acids 31–50) and P4 (amino acids 46–65), and minor binding to P2 (amino acids 16–35), P7 (amino acids 91–110), and P9 (amino acids 121–140). As P3 and P4 overlap by 5 amino acids, P3/4 (amino acids 39–57) was synthesized and assayed. This peptide also displayed significant binding to QKI-7, indicating that the QKI-7 binds most tightly to residues 31 to 65 (SHQQLIDAQFNFQNADLSRAVSLQQLTYGNVSPIQ), and that residues 31 to 38 (SHQQLIDA) are likely a core motif, as it is the common element in all peptides with the highest affinity (P3 and P3/4). A secondary binding site is found towards the end of the N-terminal domain at peptides P7 and P9, near the Akt1 phosphorylation site at S116. In summary, we narrowed down the QKI-7 binding site in the Gld2 N-terminal domain to a region spanning approximately 30 amino acids, which interestingly is distinct from the Gld2 phosphorylation sites.

Figure 5.

QKI-7 binds to specific regions in the disordered N-terminal domain of Gld2. Bar graph showing the relative depolarized emission (%) of 5 nM Gld2 N-terminal peptides in the absence and presence of 10 nM QKI-7. Fluorescence anisotropy was used to determine the relative depolarized emission (%). Each Gld2 peptide (P) was labelled on the 5ʹ-end with 6-FAM and incubated at room temperature for 20 min. Fluorescence polarization was measured at excitation 492 nm and emission 535/20 nm. Error bars represent one standard error calculated from triplicate reactions. Significant changes calculated using a two-tailed t-test are indicated by asterisks. p ≤ 0.05 (*); p ≤ 0.01 (**); p ≤ 0.001 (***)

3. Discussion

3.1. RNA binding proteins Lin28A and CPEB do not modulate Gld2 activity in vitro

Gld2 regulates the stability of both mRNAs and miRNAs, specifically the liver miRNA miR-122. There is little indication that Gld2 expression itself is regulated, but recent studies showed that Gld2 activity can be reduced by either Akt1-dependent phosphorylation [21], or by binding to the Hepatitis C virus core protein [20], leading to a decrease in miR-122 stability. On the other hand, phosphorylation by a yet unknown kinase can also enhance catalytic activity [21]. Gld2 can be directed to specifically adenylate and stabilize miR-122 through interactions with the RNA binding protein QKI-7 [23]. Here, we showed that the activity of Gld2 can be dynamically regulated in an interplay between site-specific phosphorylation and protein–protein interactions. CPEB has previously been shown to interact with both Gld2 and Gld4 to collaborate in the regulation of p53 mRNA adenylation, in this case Gld4 was the adenylyltransferase recruited for p53 mRNA adenylation [17,35]. It is thus not overly surprising that the addition of CPEB, in general, did not lead to an increase in RNA adenylation, but most often the opposite – a decrease in adenylation. In all cases, the addition of CPEB reduces Gld2 variant activity on QRE and 15A RNA. It is possible that CPEB sequesters these substrate RNAs, but does not directly interact with Gld2 in vitro, leading to reduced RNA adenylation and product formation (Figure 2). Surprisingly little change was seen in the CPE RNA reactions, but it is possible that CPEB does not recognize CPE without a larger RNA context. Similarly, Lin28A addition did not show a clear trend in activating Gld2. Lin28A was not previously shown to be associated to Gld2 but interacts with the uridylyltransferase homologs TENT3a/b [31,32]. With RNA substrates QRE and miR-122, all variants displayed either unaltered or reduced activity. Lin28A usually binds to pre-miRNAs, and not mature miRNAs, requiring the pre-miRNA-specific stem-loop structure for efficient binding and recruitment of TENT3a/b [36]. The only RNA substrate in our pool that is likely to form a stem-loop secondary structure is CPE (as calculated by mfold [37]). However, CPEB in the presence of CPE also led to no change in enzymatic activity in almost all cases, except for a 1.7-fold increase in the case of the already highly active S62E variant, and the catalytically incompetent S62E/S116E double mutant. Whether these observed increases in activity hold true in more detailed kinetic analysis remains to be elucidated.

3.2. QKI-7 and serine phosphorylation dynamically regulate Gld2 activity and substrate specificity

Since QKI-7 increased the adenylation activity of many Gld2 variant/substrate combinations, we analysed this interaction in more detail. It is notable that miR-122 and CPE are favoured substrates over 15A RNA and the QKI-7 recognition element QRE. We mapped the effect of both phosphorylation and addition of QKI-7 on enzymatic activity to visualize this dynamic range (Figure 6(a)). Normalizing to WT Gld2 activity with miR-122 allowed us to combine both our previously published kinetic analysis of Gld2 and Gld2 mutant activity to visualize the broad range of regulation. As an example, we previously published that S116E causes a reduction to 1% activity compared to WT Gld2 [21]. Our current data show, that addition of QKI-7 increases S116E activity by 2-fold (Figure 4). This leads to the calculation of 2% activity of S116E with QKI-7 compared to WT without QKI-7 (top left). Notably, deletion of the disordered N-terminal domain significantly increased Gld2 activity by at least 28-fold, suggesting an auto-inhibitory or regulatory function. Gld2 ΔN was more stable and expressed at higher levels than the full-length Gld2 variants. This increased stability was also observed when the N-terminal region of Cid1, the fission yeast homolog of Gld2, was deleted [38]. Disordered regions have been reported to be more unstable than folded regions and more sensitive to proteolytic degradation [39]. In addition, those regions can have autoinhibitory functions or be part of a protein complex to inhibit other proteins [40]. This was previously shown for the oncogenic kinase Akt1, where the PH domain serves an auto-inhibitory function [41].

Figure 6.

Gld2 phosphorylation and protein–protein interactions dynamically regulate Gld2 activity. A) Heat map depicting the fold change in activity due to phosphorylation (phosphomimetic variants encoding glutamate) or the addition of QKI-7 to the Gld2 activity assay, or a combination of both. Data in the left column was calculated from our detailed kinetic study on Gld2 activity [21]. All other data was obtained in this study and is normalized to WT without QKI-7 and with a miR-122 substrate. B) Schematic of the interplay between Gld2 phosphorylation and QKI-7 binding that dynamically control Gld2 activity. The N-terminal domain of Gld2 is autoinhibitory, S116 phosphorylation by Akt1 locks the N-terminal domain in a catalytically incompetent conformation. Protein–protein interaction with QKI-7, as well as S62 phosphorylation by an unknown kinase [21] increase Gld2 activity

The dynamic range of the combination of phosphorylation and QKI-7 binding allowed for a 1000-fold regulation of Gld2 activity, from almost inactive S116E and S62E/S116E enzymes (1% activity remaining) to a QKI-7 bound S62E variant with 10-fold increase in activity (QRE substrate) over the WT enzyme (Figure 6(a,b)). While QKI-7 binding can restore some of the activity of the catalytically incompetent S116E variant, it can only regain up to 5% of the WT activity. This indicates that the Gld2 N-terminal domain is locked in a conformation that inhibits adenylation while still allowing for QKI-7 binding, in addition to RNA binding as we showed previously [21]. While phosphorylation determines the overall activity of the Gld2 variants, QKI-7 fine tunes the activity depending on the RNA substrate (Figure 6(a)).

The interplay between phosphorylation and protein–protein interactions as a means of regulating activity has previously been shown in RNA polymerases. The phosphorylation status of the C-terminal domain (CTD) of RNA polymerase II changes during the progression of transcription to dictate which proteins are bound to the CTD [42,43]. The recruitment of specific proteins in response to the phosphorylation status of the CTD is important for transcription, chromatin remodelling, and RNA processing [43]. While RNA binding proteins may recruit Gld2 to specific RNAs as in the case of QKI-7 and miR-122 in vivo [23], the phosphorylation status of Gld2 might also play a role in fine-tuning the ability of Gld2 to interact with other proteins in response to cellular/environmental stimuli. This is evident during certain cellular events such as oocyte maturation and the cell cycle M phase where nuclear PAPs have long been shown to be inhibited due to phosphorylation events [44–46]. During those events, non-canonical PAPs such as Gld2 remain active for cytoplasmic polyadenylation to occur [35]. The need to ensure polyadenylation of the proper targets during those crucial times could potentially be enforced through phosphorylation and specific protein-protein interactions. In fact, phosphorylation of CPEB at S174 in Xenopus oocytes is crucial for Gld2 to associate with the cytoplasmic adenylation complex and polyadenylate target mRNAs [35]. Thus, phosphorylation status of the RNA binding proteins may be another mode of regulating the Gld2-RNA binding protein interaction, adding another layer of complexity to this regulatory network. Taken together, these studies and our results indicate a potential mechanism of regulating Gld2 catalytic activity where Gld2 relies on RNA binding proteins to target certain RNAs with interactions and activity further controlled by the phosphorylation status of Gld2.

3.3. QKI-7 binds Gld2 in the disordered domain

QKI-7 has been previously shown to bind to residues 1–141 in the disordered N-terminal domain of Gld2 [22]. We were able to narrow down the region(s) where QKI-7 binds by dividing the N-terminal domain into overlapping sections and using fluorescence anisotropy to determine binding (Figure 5). Our results show enhanced binding to three peptides (P3, P3/4, and P4) and less tight but significant binding to three other peptides (P2, P7, and P9). This indicated that Gld2 binds to two specific regions within the disordered peptide, near the P3/4 peptide and near the end of the disordered domain. However, as the assay utilizes 20 amino acid long peptides, it is possible that QKI-7 simultaneously binds to other residues the N-terminal domain of Gld2 or tertiary structures. A notable common feature of the peptides with significant binding is their enrichment of glutamine residues. Up to two glutamines are found in the non-significant peptides (P1, P5, P6, and P8) but 3–4 glutamines for peptides with significant QKI-7 binding. The peptides most tightly bound to QKI-7 (P2, P3, P3/4, and P4) encode two consecutive glutamines (Table 1). The only exception is P9 which was considered significant and has no glutamines. Further work will be focused on deleting specific regions and mutating the glutamine residues in full-length Gld2 and performing pulldowns and activity assays with QKI-7. Interestingly, most phosphorylation sites of the Gld2 N-terminal domain [21] do not fall within the areas identified as QKI-7 binding (Figure 5), indicating that QKI-7 binding to Gld2 is independent of phosphorylation, and that phosphorylation sites can be accessed regardless of QKI-7 binding. Phosphorylation sites S110 and S116 are in peptide P8, which does not bind to QKI-7. Similarly, S62 and S69 are found in peptide P5, which also does not bind to QKI-7. The only exception is S95, located in peptide P7. As we previously showed, Gld2 S95E on its own displays decreases in affinity (k_D) to 15A RNA compared to WT [21]. Here, we show, that binding of QKI-7 leads to increased activity with miR-122 and QRE, but not 15A RNA, further shifting activity towards these substrates.

In summary, our data shows that Gld2 enzyme activity can be regulated by both phosphorylation and binding of QKI-7, and that dephosphorylation and phosphorylation can dynamically alter enzymatic activity while bound to QKI-7.

4. Materials and methods

4.1. Kinase assay with cell extracts

Kinase assays were conducted according to the previously established methods [47] except for the following changes. HEK 293 T cells were transiently transfected with a plasmid encoding mCherry tagged Akt1 (2.5 ug/ml) to monitor transfection efficiency, and stimulated with epidermal growths factors (50 ng/ml) at room temperature for 15 min. The cells were lysed in 5X kinase assay buffer (25 mM MOPS [pH 7.0], 12.5 mM β-glycerolphosphate, 25 mM MgCl₂, 5 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA) [pH 8.0], 2 mM ethylenediaminetetraacetic acid (EDTA), 20 μM ATP, and 0.4 μCi of 33 nM [γ-³²P]-ATP) using brief sonication (20% amplitude, 1 sec pulses for 30 s, repeated 6-times) and soluble proteins were obtained by centrifugation at 64,000 × g for 15 min. Soluble proteins (200 ug of total protein per reaction) were used for the kinase assay with 200 0 μM peptide.

4.2. Plasmid construct and protein purification

Gld2 WT and the six glutamic acid variant plasmids were constructed as previously described [21]. Gld2 ΔN was constructed using the same method. Briefly, amino acids 148–480 were amplified from Gld2 WT and inserted into pGEX-6P-2 using BamHI and XhoI restriction sites. All Gld2 variants were purified as previously described [21].

Human QKI-7 was ordered codon optimized for Escherichia coli expression and inserted in pET-15b at the NdeI and BamHI restriction sites from Genewiz. The plasmid was transformed into E. coli BL21(DE3) cells (Invitrogen) and grown to an OD600 of 0.6 at 37°C. Protein production was induced by 500 μM isopropyl β-D-1-thiogalactopyranoside (IPTG) and grown at 37°C for 3 h. Cells were harvested in Nickel Buffer A (50 mM Tris-HCl [pH 8.0], 300 mM NaCl, 1 mM phenylmethylsulfonyl fluoride (PMSF)) and lysed with a French pressure cell press. The cell lysate was centrifuged at 64,000 × g for 1 hr at 4°C and the supernatant was loaded onto HisPur™ Ni-NTA Resin (ThermoScientific). The resin was washed with Nickel Buffer B (50 mM Tris-HCl [pH 8.0], 300 mM NaCl, 20 mM imidazole) and protein was eluted with Nickel Buffer C (50 mM Tris-HCl [pH 8.0], 300 mM NaCl, 250 mM imidazole). Eluted fractions containing only QKI-7 were pooled and dialysed into storage buffer (50 mM Tris-HCl [pH 8.0], 200 mM NaCl, 10% glycerol) at 4°C overnight. The proteins were aliquoted and stored at −80°C until further use.

Human Lin28A was ordered from Addgene (# 43,797) and inserted into pET-15b at the XhoI and BamHI restriction sites. The plasmid was transformed into E. coli BL21(DE3) cells (Invitrogen) carrying the pRARE plasmid (Novagen) and grown to an OD₆₀₀ of 0.6 at 37°C. Protein production was induced by 500 μM IPTG and grown at 37°C for 2 h. Cell lysis and nickel affinity chromatography were performed as described above with QKI-7. All elutions were pooled, concentrated, and dialysed into Q Buffer A (50 mM Tris-HCl [pH 8.0], 100 mM NaCl) at 4°C overnight. The protein sample was loaded onto a HiTrap Q FF 1 mL (GE Healthcare). Protein purification was automated on the ÄKTA Pure (GE Healthcare). Protein was eluted with an increasing gradient of Q Buffer B (50 mM Tris-HCl [pH 8.0], 1 M NaCl). Eluted fractions containing only Lin28A were pooled, concentrated, and dialysed into storage buffer at 4°C overnight. The proteins were aliquoted and stored at −80°C until further use.

Human CPEB was ordered from PlasmID (HsCD00339059) and inserted into pET-15b at the NdeI and XhoI restriction sites. The plasmid was transformed into E. coli BL21(DE3) cells (Invitrogen) and grown to an OD600 of 0.6 at 37°C. Protein production was induced by 500 μM IPTG and grown at 16°C for 18 h. CPEB was found to be insoluble, and purification was carried out under denaturing conditions. Cells were suspended in 6 M urea and lysed with a French pressure cell press. The cell lysate was centrifuged at 64,000 × g for 1 hr at 4°C and the supernatant was loaded onto HisPur™ Ni-NTA Resin (ThermoScientific). The resin was washed with 6 M urea, 20 mM imidazole and protein eluted with 6 M urea, 250 mM imidazole. All elutions were pooled and concentrated to 500 µL. The protein sample was loaded onto a Superdex 20,010/300 GL (GE Healthcare). Protein purification was automated on the ÄKTA Pure (GE Healthcare), and the column was washed with 6 M urea at 0.1 mL/min to separate the proteins. Eluted fractions containing only CPEB were pooled, concentrated, and dialysed into storage buffer (50 mM Tris-HCl [pH 8.0], 200 mM NaCl, 250 mM L-Arginine, 10% glycerol) at 4°C overnight. The soluble protein obtained after centrifugation of dialysed samples was aliquoted and stored at −80°C until further use.

4.3. Nucleotide addition assays

Nucleotide addition reactions contained 3.2 mM MgCl₂, 1 mM dithiothreitol (DTT), 0.5 µM ATP (0.445 μM unlabelled ATP and 0.055 μM [α-³²P]-ATP (Perkin Elmer)), 1 μM RNA, 100 nM Gld2 variant, and 100 nM interacting protein (QKI-7, CPEB, or Lin28A). Reactions with CPEB or Lin28A included 5 µM ZnCl₂ and the competitive assay had 400 nM of each RNA. Reactions were incubated at 37°C for 15 minutes and stopped with the addition of 2x RNA loading dye (95% formamide, 0.1% w/v xylene xyanol, 0.1% bromophenol blue, 10 mM EDTA). Reactions were analysed via gel electrophoresis and phosphorimaged on a Storm 860 Molecular Imager. A Whatman strip dotted with different known concentrations of [α-³²P]-ATP was exposed to the same phosphorsceen as the gel and used for product quantification. Microsoft Excel was used to generate the plot and SigmaPlot (Systat Software) was used to calculate the standard errors and p-values from triplicate reactions. All RNA substrates were bought from SigmaAldrich with a 5ʹ-phosphate group to mimic natural RNA substrates.

4.4. Fluorescence anisotropy

Gld2 peptides used for fluorescence anisotropy were bought from the lab of Dr. Shawn Shun-Cheng Li (Western University). All peptides were synthesized with 6-FAM on the 5ʹ-end and a spacer molecule (6-aminohexanoic acid) between 6-FAM and the first amino acid to ensure minimum disturbance to the peptide. A 96-well black plate was used with 100 μL reaction per well containing 3.2 mM MgCl2, 1 mM DTT, 5 nM labelled peptide, and 0 or 10 nM QKI-7. Reactions were incubated for 20 minutes at room temperature in the dark. Fluorescence polarization was measured on a Victor3V (PerkinElmer) with an excitation of 492 nm and emission of 535/20 nm. The relative depolarized emission was normalized to the 0 nM QKI-7 reactions for each Gld2 peptide. Microsoft Excel was used to generate the plot and SigmaPlot (Systat Software) was used to calculate the standard errors and p-values from triplicate reactions.

Funding Statement

This research was funded by an Alexander Graham Bell Canada Graduate Scholarship (Doctoral) from the Natural Sciences and Engineering Research Council of Canada to C.Z.C., the Natural Sciences and Engineering Research Council of Canada to IUH (RGPIN 04776-2014), and the Ontario Ministry of Research and Innovation (ER-18-14-183). Canadian Network for Research and Innovation in Machining Technology.

Notes on contributors

C.Z.C, M.I.F and N.B. designed and executed the experiments, I.U.H aided in experimental design, and C.Z.C., T.S. and I.U.H wrote and C.Z.C. and I.U.H edited the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed here.