Pointed-P2, the Drosophila ortholog of human ETS1 and ETS2, is a transcription factor involved in Ras/MAP kinase-regulated gene expression. In addition to a DNA-binding ETS domain, Pointed-P2 contains a PNT (or SAM) domain that serves as a docking module to enhance phosphorylation of an adjacent phosphoacceptor threonine by the ERK2 MAP kinase Rolled. Using NMR chemical shift, 15N relaxation, and amide hydrogen exchange measurements, we demonstrate that the Pointed-P2 PNT domain contains a dynamic N-terminal helix H0 appended to a core conserved five-helix bundle diagnostic of the SAM domain fold. Neither the secondary structure nor dynamics of the PNT domain is perturbed significantly upon in vitro ERK2 phosphorylation of three threonine residues in a disordered sequence immediately preceding this domain. These data thus confirm that the Drosophila Pointed-P2 PNT domain and phosphoacceptors are highly similar to those of the well-characterized human ETS1 transcription factor. NMR-monitored titrations also revealed that the phosphoacceptors and helix H0, as well as region of the core helical bundle identified previously by mutational analyses as a kinase docking site, are selectively perturbed upon ERK2 binding by Pointed-P2. Based on a homology model derived from the ETS1 PNT domain, helix H0 is predicted to partially occlude the docking interface. Therefore, this dynamic helix must be displaced to allow both docking of the kinase, as well as binding of Mae, a Drosophila protein that negatively regulates Pointed-P2 by competing with the kinase for its docking site.
Gene expression and signal transduction require tightly controlled macromolecular interactions and post-translational modifications involving both modular domains and intrinsically disordered regions of regulatory proteins. This paradigm is well exemplified by the ETS transcription factors. In addition to a conserved DNA-binding ETS domain, ∼1/3 of all ETS factors also contain a PNT domain,1 which is an ETS-specific member of the widespread family of SAM domains. Although sharing a common core architecture of four α-helices and a fifth small α- or 310-helix, SAM domains exhibit remarkably diverse association states and function in a wide variety of protein–protein and protein–RNA interactions.2 Defining the molecular basis for this diversity remains an important challenge. In the case of the ETS factors, one source of specificity is provided by additional helices appended to the core SAM domain. The PNT domains of Yan, ETV6 (or Tel), ERG, ELF3, and FLI1 contain only the minimal helical bundle, whereas those of GABPA and SPDEF have one additional N-terminal helix3 and those of ETS1 and ETS2 have two.4 Furthermore, owing to differing surface features,5 the PNT domains of Yan and ETV6 self-associate,6, 7 whereas the remainder are monomeric in isolation. Also, in disordered regions immediately preceding the PNT domains of closely related mammalian ETS1/ETS2 and Drosophila melanogaster Pointed-P2 are phosphoacceptors for the orthologous Ras-activated MAP kinases ERK2 and Rolled, respectively.4, 8, 9
The role of the ETS1 PNT domain in Ras/MAP kinase signaling has been investigated extensively using combination of cell-based and biophysical measurements. The monomeric PNT domain is a docking module for ERK2, enhancing the efficiency of phosphorylating Thr38 and Ser41 in a disordered region preceding this structured domain.10, 11 The docking interfaces on the PNT domain and the kinase have been mapped coarsely through mutagenesis, NMR spectroscopic, chemical footprinting, and competition studies.10, 12–15 To both accommodate the proposed docking mechanism and position the phosphoacceptors in the catalytic site of the kinase, a significant conformational change in the ETS1 PNT domain, such as the unfolding of the appended N-terminal helix H0 (residues Lys42-Thr52), appears to be required.16 Indeed, NMR relaxation and amide hydrogen exchange (HX) measurement have shown that this helix is only marginally stable and structurally flexible.4 Furthermore, upon phosphorylation, this helix H0 remains folded, but adopts a broad distribution of conformations displaced from the core helical bundle (H2–H5). This in turn contributes to enhanced electrostatically driven interactions with the TAZ1 domain of the general transcriptional coactivator acetyltransferase CBP, and ultimately, increased expression of Ras-responsive ETS1 target genes.4, 17
Pointed-P2, the Drosophila ortholog of ETS1, plays a similar role in the well-characterized EGF and Sevenless receptor tyrosine kinase/Ras-mediated signal transduction pathway to control fly eye development.18–21 Monomeric Pointed-P2 is a transcriptional activator and antagonist to Yan, an ETS family transcriptional repressor that polymerizes via head-to-tail self-association of its PNT/SAM domain.7 Both control the expression of a common set of target genes, including a crucial regulator called Mae.22 Upon stimulation of the receptor tyrosine kinase signaling cascade, the MAP kinase Rolled is activated and enters the nucleus to phosphorylate Yan, thereby leading to its CRM1-mediated export and subsequent cytoplasmic degradation.23, 24 This is facilitated by the SAM domain of Mae,25 which acts as a tight-binding heterotypic partner of the Yan PNT/SAM domain to favor its depolymerization and to expose a critical phosphoacceptor site.7 In parallel, Rolled phosphorylation of Pointed-P2 leads to transcriptional activation of the genes previously repressed by Yan. The PNT domain of Pointed-P2 is docking module for the kinase to enhance phosphorylation of the adjacent phosphoacceptor Thr151.26 However, as part of a negative feedback mechanism, Mae also heterodimerizes with the PNT domain of Pointed-P2 to block kinase docking and hence down regulating phosphorylation-enhanced gene expression.26 Parenthetically, recent studies have shown that mammalian TEL2 is functionally similar to Mae and can abrogate transcriptional output from ETS1/ETS2, albeit by a currently undefined mechanism.27
In this brief communication, we have used NMR spectroscopy to investigate further the similarities between the PNT domains and the phosphoacceptor regions of ETS1 and Pointed-P2 [Fig. 1(A)]. Based on main chain chemical shifts, the PNT domain of Pointed-P2 also contains additional helices H0/H1 appended to the core SAM helical bundle. NMR relaxation and amide HX studies confirm that these appended helices are dynamic and only marginally stable. Furthermore, rat ERK2 phosphorylates in vitro a Pointed-P2 fragment at three acceptor sites (Thr145, Thr151, and Thr154) in the disordered region immediately N-terminal to the PNT domain, and these post-translational modifications have, at best, only minor effects on the secondary structure and dynamics of the PNT domain. NMR-monitored titrations of Pointed-P2 with ERK2 also reveal that both the phosphoacceptor region and the PNT domain interact with the kinase. Based on these similarities, we hypothesize that the ETS1 and Pointed-P2 PNT domains share similar mechanisms of MAP kinase docking and recruitment of transcriptional coactivators.
Pointed-P2 PNT domain contains a dynamic helix H0
The well-dispersed 15N-HSQC spectrum of PntP2142–252 confirms that the monomeric PNT domain-containing fragment of Pointed-P2 adopts an independently folded structure [Fig. 1(B)]. With the exception of Glu142, Val143, Phe166 and the N-terminal Gly-Ser-His-Met tag, almost complete 1HN, 15N, 13Cα, 13Cβ, and 13C′ resonance assignments were obtained for this construct. These chemical shifts were used to identify the secondary structural elements of PntP2142–252 with the MICS (Motif Identification by Chemical Shift) algorithm.28 Based on this analysis, PntP2142–252 contains six helical regions that closely match those identified in Ets129–138 [Figs. 2(A,B)], as well as Ets269–172 (data not shown4). This strongly suggests that the PNT domains of Pointed-P2, ETS1, and ETS2 also share a common tertiary structure of a SAM core (helices H2–H5) with appended N-terminal helices H0/H1. In the NMR-derived structural ensemble of Ets129–138, these two appended helices are essentially continuous, yet bend at Phe53 to allow extended packing against the core helices H2 and H5. The corresponding residue in PntP2142–252 is Phe166.
The dynamic properties of PntP2142–252 were examined using 15N relaxation and amide HX measurements. As shown in Figure 3(A), the heteronuclear 15N-NOE values of residues in helix H0 progressively decrease toward its N-terminus. A model-free analysis of the 15N T1, T2, and NOE values of PntP2142–252 shows a similar trend with lower S2 order parameters for amides near the start of this helix, as well as at the C-terminus of helix H5 (Supporting Information Fig. 1S). These data are indicative of enhanced backbone mobility on a nsec–psec timescale relative to the well-ordered SAM core. However, it is difficult to estimate the nature of the conformations sampled by these fast motions from 15N relaxation measurements alone.30 More importantly, rapid amide HX was detected for residues throughout helices H0/H1 [Fig. 4(A)]. The measured HX rate constants for these residues are comparable to those predicted for an unstructured polypeptide under similar conditions.31 The only other regions of the PntP2142–252 showing such behavior are the disordered termini and exposed loops, as the remaining amides in the structured PNT domain exchange too slowly to be detected by the CLEANEX magnetization transfer approach. This clearly demonstrates that helices H0/H1 are only marginally stable and must undergo substantial local unfolding to allow facile exchange with water. Very similar 15N relaxation and HX behavior was observed for helix H0 in Ets129–138, indicating that the PNT domains of these two ETS family members also exhibit common dynamic properties.4
Pointed-P2 PNT domain phosphoacceptors are disordered
The effect of phosphorylation on the properties of Pointed-P2 was examined using active ERK2 to post-translationally modify PntP2142–252in vitro. Treatment with the kinase yielded products with two or three phosphorylated residues. The sites of modification were identified unambiguously via complementary NMR spectroscopic methods. The first method exploits a weak two-bond 31P–13Cα scalar coupling to selectively detect the amide 1HN–15N signals from a phosphorylated serine/threonine (i) and its (i + 1) neighbor via a 31P-edited HNCA spectrum [Figs. 1(B,C)].32 The second relies on the observation that the 13Cβ signal of a random coil serine/threonine shifts downfield by ∼2–4 ppm upon phosphorylation [Figs. 1(D,E)].33 Based on these measurements, 2P-PntP2142–252 is clearly phosphorylated at both Thr145 and Thr151, whereas 3P-PntP2142–252 contains an additional modification at Thr154. Of these phosphoacceptors, only Thr151 is within a MAP kinase consensus sequence (Pro-x-Ser/Thr-Pro).34, 35
The phosphoacceptor threonines of PntP2142–252 are within the disordered region N-terminal to the helical PNT domain. Similar to Ets129–138,4 the conformational flexibility of these residues is evident from both random coil chemical shifts [Fig. 2(B)], low heteronuclear 15N-NOE values [Fig. 3(B)], and rapid HX [Fig. 4(A)]. Furthermore, phosphorylation does not induce any predominant secondary structure for this region of the Pointed-P2 fragment [Figs. 2(C,D)] and only slightly dampens fast nsec–psec timescale motions of pThr151 and pThr154 detectable via 15N-NOE measurements [Figs. 3(B,C)].
Phosphorylation of PntP2142–252 has also no pronounced effect on the structure or dynamics of the PNT domain. A comparison of 15N-HSQC spectra reveals that amide chemical shift perturbations owing to phosphorylation are localized to residues near the phosphoacceptors [Figs. 1(A,D,E) and Supporting Information Fig. 2S]. Thus, the structure of the PNT domain is not altered upon modification of the threonines. A small increase in the MICS–helical scores of residues 155–157 is noted in 2P-PntP2142–252, presumably owing to phosphorylation of Thr151, yet these values decrease in 3P-PntP2142–252 with the subsequent modification of Thr154 (Fig. 2). Within experimental error, the heteronuclear 15N-NOE values of the unmodified and modified forms of PntP2142–252 are similar, indicating that phosphorylation does not dampen any fast timescale motions of amides in helix H0 (Fig. 3). Importantly, CLEANEX measurements also show that several residues in helix H0/H1 still undergo rapid HX in 2P-PntP2142–252 and 3P-PntP2142–252, albeit at a slightly reduced rate relative to the unmodified protein (Fig. 4). The average ∼2.5-fold reduction in HX rate constants for corresponding residues in helix H0/H1 of 3P-PntP2142–252 relative to the unmodified protein suggests that, in particular, pThr154 might marginally stabilize these helices by acting as an N-terminal cap.36 However, chemical shift analyses by the MICS algorithm do not detect such a predominant role for either pThr151 or pThr154 (Fig. 2). Furthermore, the modest changes (particularly when viewed on a free energy scale) could be owing to electrostatic or inductive effects of a phosphothreonine on the intrinsic exchange rates of its neighboring residues,31 or simply to subtle differences in experimental conditions as other amides in loop regions and at the C-terminus of the 3P-PntP2142–252 also showed a comparable reduction in measured HX rate constants. Regardless, these experiments indicate that the structure and dynamics of PntP2142–252 are perturbed minimally by phosphorylation.
MAP kinase docking by Pointed-P2
The interaction of PntP2142–252 with ERK2 was also examined using 15N-HSQC-monitored titrations. Upon addition of the unlabeled kinase, an overall decrease in the 1NH–15N signal intensities of amides throughout PntP2142–252 was observed [Fig. 5(A,B)]. This is attributed to faster 1NH and 15N transverse relaxation owing to the formation of a high-molecular-mass complex with the kinase. Based on the kinetic studies and equilibrium-binding measurements, the Km or Kd values ERK2 kinases and PNT domains are in the range of 10–100 μM,10, 11, 26 and thus partial saturation of PntP2142–252 is expected under these experimental conditions. More importantly, residues in the phosphoacceptor region, as well as helices H0 and H5, showed substantially greater intensity changes yet no chemical shift perturbations. This suggests that these residues undergo pronounced exchange broadening30 owing either to direct contacts with the kinase, as would be expected for the phosphoacceptors, or to indirect conformational perturbations. Indeed, when mapped on a homology model of PntP2142–252, the residues in helices H0 and H5 are both in close proximity and adjacent to side chains in the PNT domain shown previously by mutagenesis to be involved in docking interactions with the ERK2 Rolled [Fig. 5(C)].26 Similar results have also been observed for the interaction of the ETS1 PNT domain with ERK2.10, 39
Using NMR spectroscopy, we have examined the secondary structural and dynamic properties of a fragment of Pointed-P2, encompassing its phosphoacceptor region and adjacent PNT domain. Based on main chain chemical shifts, 15N relaxation, and rapid amide HX measurements, the PNT domain of Pointed-P2 closely resembles that of ETS1 with a dynamic helix H0/H1 appended to a SAM core (helices H2–H5). Phosphorylation of three threonines in disordered N-terminal region of PntP2142–252 does not significantly perturb the structure of its PNT domain and only slightly increases the protection of residues in helix H0/H1 against HX. Comparable subtle effects were observed for Ets129–138 when phosphorylated at Thr38 and Ser41.4 However, as evidenced by changes in residual dipolar couplings and interproton NOE interactions, the dynamic helix H0 adopts a broad distribution of conformations in 2P-Ets129–138 that are more displaced from the PNT/SAM core than in the structural ensemble of unmodified Ets129–138.4 This increased displacement is likely owing to electrostatic repulsion between pThr38/pSer41 and several negatively charged residues in helices H2 and H5. Given the sequence similarity of the two ETS family members, we speculate that such a conformational shift may also occur when Pointed-P2 is phosphorylated. Testing this hypothesis will, of course, require more detailed tertiary structural analyses of PntP2142–252 in its unmodified and modified forms.
In addition to confirming that Thr151 is phosphorylated by ERK2, we also identified Thr145 and Thr154 as previously unrecognized phosphoacceptors adjacent to the PNT domain of Pointed-P2. Mutational studies have demonstrated that phosphorylation of Thr151 is critically required for the in vivo function of Pointed-P2,18–20 whereas the biological roles, if any, of these additional nonconsensus sites have not been examined. It is certainly possible that these modifications are an artifact resulting from using a large scale in vitro kinase system to produce milligram quantities of modified PntP2142–252 for NMR spectroscopic studies, and/or owing to differences between mammalian ERK2 and Drosophila Rolled. However, the two MAP kinases are highly related, sharing 82% (89%) sequence identity (similarity), exhibit comparable docking interactions,40 and function in homologous signaling cascades.9, 18, 20 Furthermore, Ets129–138 is also phosphorylated in vitro by ERK2 at the corresponding Thr38 and Ser41, and in vivo tests have confirmed that both these residues contribute to Ras-enhanced transactivation by ETS1.4, 17 Thus, the potential roles of Thr145 and Thr154 in the control of gene expression by Pointed-P2 remain to be evaluated.
The results of this study have several implications for understanding the role of Pointed-P2 in the Drosophila signal transduction. Based on a mutational analysis, the Bowie group26 identified several residues, centered around Phe234 at the start of helix H5, that contribute to the docking of Pointed-P2 and ERK2 Rolled [Fig. 5(C)]. Consistent with this analysis, amides within helix H5 showed pronounced spectral perturbations upon titration of PntP2142–252 with ERK2. Furthermore, by comparison with the X-ray crystallographic structure of the Mae-SAM/Yan-SAM heterodimer [Fig. 5(D)],7 these residues are also within the expected interface for Mae, thus leading to the proposal that Mae attenuates the activity of Pointed-P2 by sterically blocking kinase docking.26 A homology model of PntP2142–252, generated from the NMR spectroscopically derived structural ensemble of Ets129–138, predicts that helix H0 and the adjacent phosphoacceptors would partially occlude this interface [Fig. 5(C)]. If so, then the dynamic helix must be displaced to allow the binding of either ERK2 Rolled or Mae. This could lead to the 15N-HSQC signal losses for residues in helix H0 observed when ERK2 was added to PntP2142–252. Unfortunately, we found the isolated Mae PNT domain to be very insoluble in vitro and thus were unable to examine its predicted effect on PntP2142–252 using NMR spectroscopy.
We speculate that the conformational flexibility and marginal stability of helix H0, detected by NMR relaxation and HX studies, might facilitate the phosphorylation of Pointed-P2 in two related ways. First, when folded, helix H0 positions the phosphoacceptors near the docking surface of the PNT domain, perhaps enhancing the initial association with the MAP kinase. Subsequently, the facile unfolding of this helix could then allow simultaneous interactions of the PNT domain with a docking site on the kinase and the phosphoacceptors with its active site to enable proximity-enhanced catalysis.11 A similar proposal has been made for the interaction of ETS1 and ERK2.4, 16.
It is well established that phosphorylation of Pointed-P2 at Thr151 is necessary for the activation of its target genes.18–21 However, the molecular mechanisms underlying this process have not been defined. By analogy to ETS1,4, 17 it is plausible that phosphorylation of Pointed-P2 leads to enhanced recruitment of Nejire, the Drosophila ortholog of the mammalian coactivator CBP. Although studies have shown that Nejire functions during successive stages of Drosophila eye development,41 such a direct interaction with Pointed-P2 has not been reported. In an effort to test this hypothesis, we expressed the predicted TAZ1 domain of Nejire using a range of methods established for the TAZ1 domains of CBP and p300.42 Unfortunately, we were unsuccessful in obtaining a soluble, folded protein fragment as required for NMR-monitored binding studies with PntP2142–252. Therefore, future investigations will be required to uncover the link between Pointed-P2 phosphorylation and transcriptional activation. Our demonstration that the PNT domains and adjacent phosphoacceptors of ETS1 and Pointed-P2 share very similar secondary structural and dynamic properties should help guide this research.
Materials and Methods
The gene-encoding residues 142–252 of D. melanoganster Pointed-P2 (Genbank NM_079737.2) was cloned by PCR methods into the pET28a vector for expression in Escherichia coli BL21 (λDE3) cells as a His6-tagged construct. The single cysteine (Cys250) was mutated to serine to avoid potential oxidation. Following established protocols,4 the samples of 15N/13C-labeled protein were produced in M9 minimal media, containing 1 g/L 15NH4Cl and 3 g/L 13C6-glucose, with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) induction overnight at 30°C. Harvested cells were resuspended and homogenized in lysis buffer (20 mM sodium phosphate, 500 mM NaCl, 5 mM imidazole, 2 mM dithiothreitol [DTT], pH 7.4) and the expressed protein isolated by Ni+2-affinity chromatography, followed by thrombin digestion to remove the His6-tag. After further purification using S75 size exclusion chromatography, the protein was dialyzed against NMR sample buffer (20 mM 3-(N-morpholino)-propanesulfonic acid, 10 mM NaCl, 2 mM DTT, pH 6.7) and concentrated by ultrafiltration. The resulting construct contains a non-native N-terminal Gly-Ser-His-Met from the cleavage site and is denoted as PntP2142–252.
In vitro phosphorylation
PntP2142–252 was phosphorylated by overnight incubation at 30°C in a 40:1 molar ratio with ERK2 kinase (125 mM Tris, 5 mM DTT, 50 mM MgCl2, 100 mM NaCl, pH 7.5), as described previously for Ets129–138.4 The active rat ERK2 was prepared from E. coli BL21 (λDE3) grown in TB media with 0.8% glycerol, 0.1 mg/mL carbenicillin, 0.08% glucose, and 0.04 mM IPTG for induction, according to the previously published methods.10, 17 The resulting 2P- and 3P-PntP2142–252 were separated by anion exchange chromatography on a Mono Q column (20 mM Tris, pH 7.5, 10% glycerol, 2 mM DTT, gradient 0–1M NaCl), confirmed by MALDI-ToF mass spectrometry, and dialyzed into NMR sample buffer.
Spectra were obtained for proteins in NMR sample buffer with 10% D2O at 25°C using 600 MHz Varian Inova or Bruker Avance III spectrometers. Data were processed with NMRpipe43 and analyzed using Sparky.44 Main chain resonance assignments were obtained using standard 15N-HSQC, HNCO, HN(CA)CO, CBCA(CO)NH, HNCACB, and C(CO)TOCSY-NH experiments recorded with 1H/13C/15N cryogenic probes.45 The 1H–15N spectrum of a 31P-edited HNCA experiment32 was recorded using a room temperature 1H/13C/15N/31P QXI probe. Amide 15N relaxation parameters29 were obtained using a 600-MHz NMR spectrometer and analyzed using TENSOR2.46 CLEANEX amide HX measurements47 were recorded and analyzed as described previously.4