Identification of heteromolecular binding sites in transcription factors Sp1 and TAF4 using high‐resolution nuclear magnetic resonance spectroscopy

Abstract

The expression of eukaryotic genes is precisely controlled by interactions between general transcriptional factors and promoter‐specific transcriptional activators. The fourth element of TATA‐box binding protein‐associated factor (TAF4), an essential subunit of the general transcription factor TFIID, serves as a coactivator for various promoter‐specific transcriptional regulators. Interactions between TAF4 and site‐specific transcriptional activators, such as Sp1, are important for regulating the expression levels of genes of interest. However, only limited information is available on the molecular mechanisms underlying the interactions between these transcriptional regulatory proteins. We herein analyzed the interaction between the transcriptional factors Sp1 and TAF4 using high‐resolution solution nuclear magnetic resonance spectroscopy. We found that four glutamine‐rich (Q‐rich) regions in TAF4 were largely disordered under nearly physiological conditions. Among them, the first Q‐rich region in TAF4 was essential for the interaction with another Q‐rich region in the Sp1 molecule, most of which was largely disordered. The residues responsible for this interaction were specific and highly localized in a defined region within a range of 20–30 residues. Nevertheless, a detailed analysis of ¹³C‐chemical shift values suggested that no significant conformational change occurred upon binding. These results indicate a prominent and exceptional binding mode for intrinsically disordered proteins other than the well‐accepted concept of “coupled folding and binding.”

Introduction

The expression of genes is regulated by interactions between ribonucleic acid (RNA) polymerase II and a number of regulatory proteins. This process is initiated by the formation of the multicomponent preinitiation complex (PIC), which assembles over the core promoter of a gene.1, 2 The PIC is composed of RNA polymerase II and a number of general transcription factors including TFIIA, ‐B, ‐D, ‐E, ‐F, and ‐H.3, 4 Among these general transcriptional factors, TFIID is a key element because the first step in the formation of the PIC is the binding of TFIID to the core promoter, and it serves as the scaffold for the assembly of the remainder of the transcription complex.

TFIID is a large multiprotein complex that is composed of the TATA‐box binding protein (TBP) and at least 14 TBP‐associated factors (TAFs). TAF4 (also known as hTAF_II130/hTAF_II135 in humans), the fourth subunit of TFIID, is involved in interactions with cellular transcriptional activators, and these interactions partially participate in increasing transcriptional levels in a site‐specific manner. This subunit also plays a critical role in maintaining the stability of the TFIID complex.5 Sequence analyses revealed that TAF4 possesses two highly conserved domains, CI and CII, as well as four glutamine‐rich domains (Q‐domains) in the central region of molecule [Fig. 1(B)].6 Of these domains, the structures of the CI and CII domains were analyzed by X‐ray crystallography.7, 8 In contrast, only a limited information is available on the structures of the four Q‐domains located in the central regions of the molecule although these Q‐domains are involved in interactions with a number of gene‐specific transcriptional activators including specificity protein 1 (Sp1) and the cyclic AMP (adenosine 3′,5′‐cyclic monophosphate) (cAMP)‐response element‐binding protein (CREB).9, 10 Early biochemical studies indicated that one or more of the four Q‐domains were involved in these interactions, and the central region of human TAF4 including all four Q‐domains was sufficient for interactions with the activation domain of human Sp1.11

Figure 1.

(A) Schematic representation of the primary sequence of the transcription factor Sp1. Two Q‐rich domains, Sp1‐QA and Sp1‐QB, and three zinc finger domains are indicated in gray and black, respectively. The QB region divided into two fragments, QBn and QBc, is also indicated. (B) Schematic drawing of the primary structure of human TAF4. Four Q‐domains present in the molecule, Q1, Q2, Q3, and Q4, are indicated in gray. The positions of the highly conserved regions CI between Q2 and Q3, and CII at the C‐terminus are also indicated. In addition, the fragment proteins studied here, TAF4N/C, Q1, Q2, Q12, CI, and Q34, are designated.

The promoter‐specific transcription factor Sp1 plays a primary role in the regulation of more than 100 genes.12, 13, 14 Sp1 has been shown to recognize and bind to the GC‐rich consensus sequence (GC‐box), and the binding of Sp1 to these promoter regions increases the transcriptional levels of target genes. Furthermore, Drosophila TAF4 (dTAF_II110) has been shown to directly interact with the human transcriptional activator Sp1, resulting in enhancements in the transcriptional levels of genes located downstream.15, 16 Therefore, the interaction between Sp1 and TAF4 in the general transcriptional factor is important for the recruitment of RNA polymerase II and the activation of a target gene. Sp1 contains two Q‐domains (Sp1‐QA and Sp1‐QB) in the middle of the molecule, and three C₂H₂ type zinc fingers at the C‐terminus [Fig. 1(A)].17, 18, 19 Although the structures of C‐terminal Zn‐finger motifs were revealed by high‐resolution nuclear magnetic resonance (NMR) spectroscopy, structural details on other regions including the two Q‐domains have yet to be elucidated.20 Q‐rich domains have been detected in many transcription‐activating proteins, and are considered to represent one of the common motifs involved in molecular interactions.21, 22 The two Q‐domains in Sp1 and four in TAF4 are also important for the interaction between these proteins.

Recent advances in bioinformatics and genome sequencing suggest that an increasing number of eukaryotic proteins are identified to be intrinsically disordered.23, 24 Many of these intrinsically disordered proteins (IDPs) and intrinsically disordered regions (IDRs) in proteins are involved in interactions with other biomolecules such as deoxyribonucleic acid (DNA), RNA, carbohydrates, and proteins.25, 26, 27 While these IDPs are largely unstructured by themselves, marked conformational changes often occur upon binding to an interacting partner, which is known as the “coupled folding and binding mechanism.”28, 29 We recently analyzed the structural details of Q‐domains in Sp1 and TAF4, and found that these domains were intrinsically disordered.30 We analyzed the ¹H‐¹⁵N heteronuclear single quantum coherence (HSQC) spectra of ¹⁵N‐Sp1‐QB in the absence and presence of unlabeled TAF4N/C, and found that the peak intensity of several residues decreased by the addition of TAF4N/C.30 Interactions between proteins were also analyzed using circular dichroism (CD), but any significant conformational change was not detected. Therefore, we considered that these proteins interact without any concomitant conformational changes. However, this was somewhat ambiguous because NMR peaks decreased in intensity upon complex formation, which precluded from obtaining any information on the structure of the complex. We herein reexamined this interaction using optimized conditions for the detection of NMR signals derived from the complex. We unambiguously identified the heteromolecular interaction site in Sp1 and TAF4, and also confirmed that no significant conformational change occurred upon the interaction between these proteins. The results represent a prominent and exceptional binding mode for IDPs that is not categorized in the well‐accepted concept of “coupled folding and binding.”

Results

Optimization of experimental conditions and NMR signal assignment

In the first step of the analysis, we need to assign the NMR signals of TAF4N/C, the molecular weight of which was approximately 45 kDa, and its assignment is challenging. We found that TAF4N/C was consisted of IDRs, and it was possible to divide into three fragment proteins (Q12, CI, and Q34; Fig. 1) without significant perturbations to the structure. This is demonstrated by the observation that the superposition of the NMR spectra of these fragment proteins, which were recorded separately, perfectly matched to the spectrum obtained by the entire protein (see Fig. S2 in Ref. 30).

We attempted to analyze a short fragment protein corresponding to the Q1 region of TAF4 under the same conditions used in the previous study (140 mM NaCl, 20 mM sodium phosphate, pH 7.3 and 4°C). However, the analysis under these conditions was unsuccessful, because most NMR peaks disappeared in the ¹H‐¹⁵N HSQC spectra (only approximately 60 peaks were visible, while the expected number was 96; Supporting Information, Fig. S1). After a comprehensive screening of experimental conditions, we found that pH 5.0 and 37°C gave the most promising spectra, while an increase in pH by one unit (pH 6) decreased the signal‐to‐noise ratio [Fig. S2(A)]. Therefore, we selected these experimental conditions (pH 5.0 and 37°C) in this study. Although it is slightly more acidic than physiological conditions, the appearance of ¹H‐¹⁵N HSQC spectra did not change significantly at different pH conditions (pH 6.5, 6.0, 5.5, and 5.0; data not shown). Furthermore, the CD spectra recorded at pH 7.3 and 5.0 were indistinguishable from each other [Fig. S2(B)]. We also performed a small angle X‐ray scattering experiment to estimate the values of the radius of gyration (R _g), and confirmed that the overall structures of Sp1‐QB and TAF4N/C were not significantly affected under the different pH conditions [Fig. S2(C)].

Figure 2(A–C) shows the ¹H‐¹⁵N HSQC spectra of TAF4‐Q12, TAF4‐CI, and TAF4‐Q34 fragment proteins. All spectra appeared to be poorly dispersed, particularly along the ¹H chemical shift axis, on which most of the peaks appeared within the range between 7.7 and 8.7 ppm. This result suggests that proteins have neither secondary nor tertiary structures that are stabilized by regular hydrogen bonds and the rigid packing of side chains, indicating that most part of these proteins are intrinsically disordered. A careful examination of the spectrum of TAF4‐Q12 revealed that several peaks appeared in a lower field region (with higher chemical shift values) in the ¹H axis, suggesting that weak hydrogen bonds might be formed. According to the NMR, the CD spectra also suggested that most part of TAF4‐Q12 and TAF4‐Q34 were disordered [Fig. 2(D)]. On the other hand, the central fragment, TAF4‐CI (conserved region I), showed large double minima at 222 and 208 nm, suggesting a predominantly α‐helical structure. This was consistent with the X‐ray crystallographic study, which indicated that the protein consisted of a five‐helix “wedge” structure.7

Figure 2.

The ¹H‐¹⁵N HSQC spectra of ¹⁵N‐TAF4‐Q12 (A), ¹⁵N‐TAF4‐CI (B), and ¹⁵N‐TAF4‐Q34 (C) measured at pH 5.0 and 37°C. In each panel, two spectra are superimposed, which are recorded in the absence (red) and presence (blue) of unlabeled Sp1‐QB. Peaks that showed a significant change upon the addition of Sp1‐QB, as well as those located outside the “disordered region” (between 7.7 and 8.7 ppm), are labeled by the residue name and number. On the other hand, the peaks of which chemical shifts and intensities did not change are seen in black, because of the superposition of red and blue. (D) Far UV‐CD spectra of TAF4‐Q12 (red), TAF4‐CI (black), and TAF4‐Q34 (blue) at pH 5.0 and 37°C. (E) The apparent chemical shift change of TAF4‐Q12, $\Delta \delta =\sqrt{{(\Delta {\delta }_{1\mathrm{H}})}^2+{(\Delta {\delta }_{15\mathrm{N}}/5)}^2}$, by the addition of unlabeled Sp1‐QB.

Identification of the Sp1‐interaction site in the TAF4 molecule

On the basis of the results of backbone chemical shift assignments, we attempted to elucidate the binding region of TAF4N/C with Sp1‐QB by comparing the ¹H‐¹⁵N HSQC spectra recorded in the absence and presence of an excess amount of unlabeled Sp1‐QBs. Although Sp1 has two Q‐rich regions (Sp1‐QA and Sp1‐QB), a careful examination using NMR and surface plasmon resonance revealed that Sp1‐QA did not contribute significantly to the interaction with TAF4N/C.30 Therefore, we examined only for Sp1‐QB in this study. We measured the ¹H‐¹⁵N HSQC spectra for TAF4‐CI and TAF4‐Q34 in the presence of more than a twofold amount of unlabeled Sp1‐QB, but neither the chemical shift values nor the peak intensity were changed [Fig. 2(B,C)]. On the other hand, the ¹H‐¹⁵N HSQC spectra of TAF4‐Q12 showed marked differences in the chemical shift values of several residues upon addition of an equimolar amount of Sp1‐QB, suggesting a specific interaction between TAF4‐Q12 and Sp1‐QB [Fig. 2(A)]. The apparent chemical shift change plotted against residue numbers showed that only a stretch of residues (456–480) was perturbed, suggesting a possible binding site for Sp1‐QB [Fig. 2(E)].

The above results indicated that only the Q1 domain in TAF4‐Q12 might contribute to the interaction with Sp1‐QB. In order to clarify this, we prepared two fragment proteins, TAF4‐Q1 and TAF4‐Q2 (see Fig. 1), and recorded their ¹H‐¹⁵N HSQC spectra separately in the absence and presence of unlabeled Sp1‐QB [Fig. 3(A,B)]. Apparent chemical shift changes by Sp1‐QB were plotted against residue numbers in Figure 3(C,D). While chemical shift values for the residues in TAF4‐Q1 changed markedly in the presence of an equimolar amount of Sp1‐QB, no significant changes were observed for TAF4‐Q2 even in the presence of a fivefold excess amount of Sp1‐QB. Furthermore, the distribution of the apparent chemical shift change showed almost the same pattern as that obtained from the interaction between TAF4‐Q12 and Sp1‐QB. This result suggests that the effects of the addition of Sp1‐QB were quantitatively similar in these experiments [Figs. 2(E) and 3(C,D)], confirming that the residues responsible for binding are solely located in the region of TAF4‐Q1.

Figure 3.

(A) ¹H‐¹⁵N HSQC spectra of ¹⁵N‐TAF4‐Q1 in the absence (red) and presence (blue) of an equimolar amount of unlabeled Sp1‐QB. (B) ¹H‐¹⁵N HSQC spectra of ¹⁵N‐TAF4‐Q2 in the absence (red) and the presence (blue) of a fivefold excess amount of Sp1‐QB. Peaks that showed significant differences in chemical shift values are labeled by the residue name and number. The peaks observed in black indicate that neither the chemical shifts nor intensities of the peaks were changed. (C,D) The apparent chemical shift change of ¹⁵N‐TAF4‐Q1 and ¹⁵N‐TAF4‐Q2 upon the addition of unlabeled Sp1‐QB.

We also performed a similar experiment with the opposite combinations of isotopic labeling, in which ¹H‐¹⁵N HSQC spectra were recorded for ¹⁵N‐Sp1‐QB in the presence of an excess amount of unlabeled TAF4‐Q12, TAF4‐CI, and TAF4‐Q34. In consistent with the above observations, we found that the chemical shift values of several peaks changed only when unlabeled TAF4‐Q12 was added, suggesting a specific interaction between TAF4‐Q12 and Sp1‐QB (Fig. S3).

Revisiting the interaction between Sp1‐QB and TAF4

We also attempted to identify the residues in Sp1‐QB that are responsible for the binding with TAF4. The results of chemical shift perturbation experiment suggested that the residues located in the C‐terminal half of Sp1‐QB were important for the interaction with TAF4‐Q12 [Fig. S3(D)]. In order to confirm this, we prepared two fragment proteins in which the N‐ and C‐terminal halves of Sp1‐QB were included (Sp1‐QBn and Sp1‐QBc, see Fig. 1), and their ¹H‐¹⁵N HSQC spectra were measured in the presence of unlabeled TAF4N/C or TAF4‐Q12. No significant differences were observed in the spectra of Sp1‐QBn, even in the presence of a fivefold molar excess amount of unlabeled TAF4N/C [Fig. S4(A)]. In contrast, a large displacement for several peaks were observed in the spectra of Sp1‐QBc with increasing concentration of unlabeled TAF4‐Q12 added [Fig. S4(B)]. The plot of the apparent chemical shift change against residue number showed that the most affected residues were located exclusively in the C‐terminal half of Sp1‐QB (QBc) [Fig. S4(C,D)].

The titration experiments shown in Figure S4(B) demonstrated that the chemical shift values of ¹⁵N‐Sp1‐QBc were dependent on the concentration of unlabeled TAF4‐Q12 present in the solution. In order to analyze these results in more detail, we made two assumptions: (1) the formation of a 1:1 complex between Sp1‐QBc and TAF4‐Q12, and (2) “a fast exchange regime,” namely, the apparent chemical shift change was proportional to the concentration of the resultant complex. Based on these assumptions, we analyzed the apparent chemical shift change as a function of the concentration of TAF4‐Q12 added [Fig. S4(E)]. The results of curve fitting were converged well, and binding constants for several representative residues were estimated to be 1.4–2.8 × 10⁴ M ⁻¹.

We performed a similar experiment using the opposite combination of isotopic labeling, that is, unlabeled Sp1‐QBn or Sp1‐QBc was added separately to ¹⁵N‐TAF4‐Q12. The results obtained were the same as those expected, namely, the change in the chemical shift values of ¹⁵N‐TAF4‐Q12 was only observed when unlabeled Sp1‐QBc was added (Fig. 4). Collectively, these results suggest that the interaction between TAF4 and Sp1 is specific and highly localized. In Sp1, only the C‐terminal half (Sp1‐QBc) was needed for the interaction with TAF4. On the other hand, in TAF4, only the first Q‐rich domain (TAF4‐Q1) was exclusively important for the interaction with Sp1. We summarized interactions between several fragment proteins in Table 1.

Figure 4.

(A) The ¹H‐¹⁵N HSQC spectra of ¹⁵N‐TAF4‐Q12 measured in the absence (red) and presence (blue) of unlabeled Sp1‐QBn. (B) The ¹H‐¹⁵N HSQC spectra of ¹⁵N‐TAF4‐Q12 measured at various concentrations of the unlabeled Sp1‐QBn. Spectra in the absence and the presence of 0.5, 1.0, 1.5, 2.0, 2.5 molar ratio of unlabeled Sp1‐QBc to ¹⁵N‐TAF4‐Q12 are colored in red, orange, green, blue, purple, and black, respectively. The region indicated by the dashed box in the spectrum is expanded in the right panel. The chemical shifts of several peaks change linearly. (C) The apparent chemical shift change of ¹⁵N‐TAF4‐Q12 on the addition of an equimolar amount of unlabeled Sp1‐QBc.

Table 1.

Summary for Intermolecular Interactions Investigated in the Present Study

Observed protein	Added protein	Molar ratioa	Change in spectra	Corresponding figure
15N‐TAF4‐Q12	Sp1‐QB	1:1.5	Movedb	Fig. 2(A,E)
15N‐TAF4‐CI	Sp1‐QB	1:3	n.d.c	Fig. 2(B)
15N‐TAF4‐Q34	Sp1‐QB	1:2	n.d.	Fig. 2(C)
15N‐TAF4‐Q1	Sp1‐QB	1:0.8	Moved	Fig. 3(AC)
15N‐TAF4‐Q2	Sp1‐QB	1:5	n.d.	Fig. 3(B,D)
15N‐TAF4‐Q12	Sp1‐QBn	1:5	n.d.	Fig. 4(A)
15N‐TAF4‐Q12	Sp1‐QBc	1:0.5–2.5	Moved	Fig. 4(B,C)
15N,13C‐TAF4‐Q1	Sp1‐QBc	1:0.8	n.d.	Fig. S5(A,B)
15N‐Sp1‐QB	TAF4‐Q12	1:1	Moved	Fig. S3(A,D)
15N‐Sp1‐QB	TAF4‐CI	1:2.5	n.d.	Fig. S3(B)
15N‐Sp1‐QB	TAF4‐Q34	1:2.5	n.d.	Fig. S3(C)
15N‐Sp1‐QBn	TAF4N/C	1:5	n.d.	Fig. S4(A,C)
15N‐Sp1‐QBc	TAF4‐Q12	1:0.5–2.5	Moved	Fig. S4(B,D)
15N,13C‐Sp1‐QBc	TAF4‐Q12	1:1	n.d.	Fig. S5(C,D)
15N‐Sp1‐QB	TAF4N/C	1:1	Disappearedd	Fig. 3(B) in Ref. 30 e

^aThe ratio of [observed protein]:[added protein].

^bRemarkable change in chemical shift values for several residues.

^cNot detected.

^dDisappearance of several peaks in a concentration‐dependent manner.

Under the conditions of pH 7.3 and 4°C.

No significant conformational change upon binding

The chemical shift values of NMR signals are very sensitive indicators that reflect any change in the electron density around the nuclear spin, and very subtle changes in the local environment may result in a large difference in the chemical shift value. Therefore, slight changes in chemical shift value in ¹H‐¹⁵N HSQC spectra are not necessarily linked to conformational changes in a protein. On the other hand, the chemical shift values of ¹³C^α, ¹³C^β, and ¹³C′ are known to be influenced sensitively by the ϕ and ψ dihedral angles of the residue of interest, and have been used to analyze secondary structures at residual resolution.31 Therefore, we obtained these ¹³C‐chemical shift values of ¹⁵N,¹³C‐Sp1‐QBc and ¹⁵N,¹³C‐TAF4‐Q1 by analyzing three‐dimensional (3D)‐HNCACB and 3D‐HNCO spectra. We also measured the same spectra of these ¹⁵N,¹³C‐proteins in the presence of unlabeled TAF4‐Q12 or Sp1‐QB [Fig. 5(A,B)]. The slices of 3D‐HNCACB and 3D‐HNCO spectra for several representative residues are also shown in Figure S5. Notably, no significant differences in ¹³C chemical shift values were found throughout the molecules in both proteins, even for residues that showed a marked change in the chemical shift values in ¹H‐¹⁵N HSQC spectra (residues 456–480 for TAF4‐Q1, and 435–445 and 454–465 for Sp1‐QBc). These results suggest that the interaction between Sp1 and TAF4 is not accompanied by any significant conformational changes in either protein, at least at the level of the secondary structure.

Figure 5.

(A) Change in chemical shift values of ¹³C^α, ¹³C^β, and ¹³C′ obtained from 3D‐HNCACB and 3D‐HNCO spectra of ¹⁵N,¹³C‐TAF4‐Q1 upon the addition of an equimolar amount of unlabeled Sp1‐QB. (B) Changes in the chemical shift values of ¹³C^α, ¹³C^β, and ¹³C′ of ¹⁵N,¹³C‐Sp1‐QBc upon the addition an equimolar amount of unlabeled TAF4‐Q1. (C) Far UV‐CD spectra of TAF4‐Q1 (red), Sp1‐QBc (blue), and their equimolar mixture (purple). The black line represents the simple sum spectrum of TAF4‐Q1 and Sp1‐QBc that was recorded separately. The intensity of CD was expressed as an apparent ellipticity value with the unit of mdeg because two proteins consisted of different numbers of amino acid residues.

In order to confirm this, we also measured the CD spectra of Sp1‐QBc (93 amino acid [a.a.]) and TAF4‐Q1 (80 a.a.). Since all residues in a protein contribute to the signal of the CD spectrum, we used these minimal fragment proteins of interest in order to facilitate the detection of any conformational changes. Otherwise, the presence of many residues that do not contribute to the interaction may obscure potentially small changes in the spectra. We measured the CD spectra of Sp1‐QBc, TAF4‐Q1, and their equimolar mixture, and compared the spectrum of the mixture with the sum of the two separately measured spectra [Fig. 5(C)]. The two spectra, obtained from the mixed sample and calculated sum of two separately measured spectra, were indistinguishable from each other. These results also suggest that there were no significant conformational changes in Sp1 or TAF4.

Discussion

TAF4N/C was partially structured, but largely disordered

We previously reported that the ¹H‐¹⁵N spectrum of TAF4N/C at pH 7.3 and 4°C showed very poor dispersion, particularly along the ¹H chemical shift axis.30 In addition, the overlay of the spectra of three fragment proteins (TAF4‐Q12, ‐CI, and ‐Q34) recorded separately was almost identical to that of the entire protein, TAF4N/C, suggesting that there are no long‐range interactions between these fragment proteins. Based on these findings, we considered TAF4N/C to be largely disordered under the physiological conditions.

On the other hand, other observations suggest the presence of partially structured regions in TAF4N/C. One of them was the CD spectrum of TAF4N/C, which showed a small, but significant shoulder around 222 nm, suggesting the presence of some amount of α‐helical structure. Another argument was based the small‐angle X‐ray scattering (SAXS) measurements. From the analysis of Guinier plot, the R _g value of TAF4N/C was estimated to be 49.0 ± 7.4 Å, which was smaller than that expected for the urea‐denatured state of the protein with 431 a.a. residues (approximately 70 Å).32

These discrepancies may now be rationalized by measuring the CD spectra of three fragment proteins separately [Fig. 2(D)]. Two fragment proteins that are dominated by the Q‐rich sequences, TAF4‐Q12 and TAF4‐Q34, showed a deep minimum at approximately 200 nm, suggesting that most parts of the molecules are disordered. On the other hand, the fragment protein corresponding to the highly conserved region, TAF4‐CI, showed deep minima at 222 and 208 nm, which are characteristics α‐helical structure. This result was in good agreement with the X‐ray crystallographic study, which indicated that the protein consists of a five‐helix wedge structure.7 Therefore, we considered the structure of TAF4N/C to be “bipartite,” consisting of two highly disordered segments of the Q‐rich sequence connected by a highly conserved α‐helical structure.

TAF4‐Q1 is responsible for the interaction with Sp1‐QB

One or more of four Q‐domains in the central region of TAF4 have been considered to be involved in the interaction with various transcriptional activators such as Sp1 and CREB.9, 10 As summarized in Table 1, we herein revealed that the region corresponding to the Q1‐domain in TAF4 was exclusively important for the interaction with Sp1‐QB domain. A detailed analysis of chemical shift perturbation experiments suggested that the interaction with Sp1‐QB was specific and localized within a narrow range of 456–480 in TAF4N/C [Figs. 2(E) and 3(C,D)].

A careful examination of the ¹H‐¹⁵N HSQC spectrum of TAF4‐Q12 revealed that several peaks appeared in the lower field region (with higher chemical shift values) in the ¹H chemical shift axis, suggesting the formation of weak hydrogen bonds [Figs. 2(A), 3(A), and 4(B)]. Furthermore, most of these dispersed peaks were markedly affected by the addition of unlabeled Sp1‐QB fragment proteins. In order to elucidate the presence of any secondary structural elements, we analyzed ¹H, ¹⁵N, ¹³C^α, ¹³C^β, and ¹³C′ chemical shift values using the CSI 2.0, CSI 3.0, and TALOS+ programs.33, 34, 35 The results of these programs were consistent with each other [Fig. S6(A)]; all predicted the presence of a short α‐helical structure preceded by two β‐strands connected by a loop (a possible strand‐turn‐strand motif) in the region of TAF4‐Q1. We also analyzed the primary sequence of TAF4, and found that not only the highly conserved region (CI), but also the Q1 region was conserved very well between vertebrates including fish and amphibia [Fig. S6(B)], suggesting the importance of this region in the function of TAF4.

In contrast to Q1 domain, the other three Q‐domains (Q2, Q3, and Q4) as well as the central highly conserved region, CI, did not show any evidence for an interaction with Sp1‐QB. TAF4 is known to interact with a number of transcriptional activators, which are considered to recognize different parts of the TAF4 molecule.36 For example, the highly conserved region CI has been reported to bind to the short hydrophobic motif DΨΨζζΨΦ present in transcriptional regulators.7 Therefore, regions other than the Q1 domain may be important for interactions with a various transcriptional regulators in the versatile regulatory protein TAF4.

C‐terminal half of the Sp1‐QB domain is responsible for the interaction with TAF4

We recently found that the peaks corresponding from the center to C‐terminus of Sp1‐QB decreased in intensity in the presence of unlabeled TAF4N/C at pH 7.3 and 4°C.30 We herein reexamined the interaction between Sp1‐QB and TAF4 using smaller fragment proteins under slightly acidic conditions (pH 5.0 and 37°C). Although the effects of the addition of unlabeled TAF4 proteins were different from each other (decreases in intensity vs. changes in chemical shift), the two sets of results corresponded very well. Only the residues within the range from the central to C‐terminal part of Sp1‐QB (Sp1‐QBc) were affected by the addition of unlabeled TAF4 proteins, suggesting the importance of these regions for the interaction. As mentioned in the previous report, these regions contain a stretch of hydrophobic residues (⁴⁶⁴WATLQLQNL⁴⁷²), which were indicated to be important for the interaction by the mutational studies. It should be noted that most of the peaks in the ¹H‐¹⁵N HSQC spectra are derived from the backbone amide moieties, the spectrum also provides the information about tryptophan side chain. However, we could not observe any significant changes in the signal from indole‐ring of W464. A rapid motion in the side chain might obscure the effect of the binding.

The present results corresponded quantitatively well to the previous study. The analysis of the titration experiments between ¹⁵N‐Sp1‐QBc and unlabeled TAF4‐Q12 [Fig. S4(E)] provided the binding constant of 1.4–2.8 × 10⁴ M ⁻¹. These values consistent with the previous estimation by surface plasmon resonance experiments between full‐length Sp1‐QB and TAF4N/C at pH 7.3 and 15°C (K _a = 1.45 × 10⁴ M ⁻¹ by Ref. 30). From these estimations for the binding constants (1.5 × 10⁴ M ⁻¹) and the experimental conditions used in this study (50 μM of protein), about 33% of the molecules were forming the complex. Although the ratio of the complex is not very high, a careful examination by CD spectra and ¹³C‐chemical shift values would reveal the secondary structural changes if present.

Although we revealed that the interaction between Sp1‐QBc and TAF4‐Q1 domains was very specific and highly localized, the binding constant between them were not very high. In this context, we should mention the “superactivation” of Sp1, in which the promoter activity of Sp1 is synergetically enhanced by the presence of repeated binding sequences for Sp1 (GC‐boxes) located upstream of the DNA molecule.9, 18 From these biochemical observations, Sp1 molecules have been considered act as homooligomers in the cell to enhance the affinity for other transcription factors, such as TAF4. In addition, we should be aware that both Sp1 and TAF4 are DNA‐binding proteins, and the effective concentration of them along the DNA molecule could be much higher than the overall average concentration.

A novel interaction mode of the disordered region

In contrast to the results obtained for ¹H‐¹⁵N HSQC spectra, no significant changes were observed in the chemical shift values of ¹³C^α, ¹³C^β, and ¹³C′, which are sensitive indicators of backbone ϕ and ψ dihedral angles of the residue, suggesting that the structures of Sp1‐QB and TAF4N/C do not significantly change when they interact with each other. Among these proteins, the interacting residues in TAF4‐Q1 were suggested to partially fold into strand‐turn‐strand followed by an α‐helix based on analyses using CSI 2.0/3.0 and TALOS+ [Fig. S6(A)]. On the other hand, the residues in Sp1‐QBc did not show any preference for the formation of the secondary structure (data not shown), suggesting that the protein is intrinsically disordered. One of the important structural features for these IDPs is “flexibility,” which has been suggested to enable them to access a broad conformational space to interact with a wide array of biomolecular targets. Many IDPs undergo a disorder‐to‐order transition to form well‐defined structures upon binding to their cellular targets. This process is called coupled folding and binding, and is suggested to be a common mechanism for IDPs/IDRs to interact with their target molecules. We herein revealed the binding regions in TAF4N/C and Sp1‐QB, and found that Sp1‐QBc did not change its conformation even after binding to its target molecule, TAF4‐Q1. The results of the present study suggest a prominent and novel binding mode for IDPs/IDRs, which are not categorized by the well‐accepted concept of the coupled binding and folding mechanism.

Materials and Methods

Materials

¹⁵N‐ammonium chloride, ¹³C₆‐d‐glucose, and deuterium oxide were purchased from SI Science (Saitama, Japan). Other reagents were purchased from Nacalai Tesque (Kyoto, Japan).

Protein preparation

The expression and purification of the glutamine‐rich domains of Sp1 (QB [342–488], Bn [342–424], Bc [421–488]), and TAF4N/C (TAF4 [408–838]) were performed as described previously.30

The expression plasmids for TAF4‐Q12 (TAF4 [408–557]), TAF4‐CI (TAF4 [558–665]), and TAF4‐Q34 (TAF4 [666–838]) were constructed as a fusion protein with a hexahistidine and ubiquitin tag (His‐Ub) at the N‐terminus.30 The encoding region for each protein was amplified by polymerase chain reaction to introduce recognition sites for Bsp TI at the 5′‐end and Eco RI at the 3′‐end. The amplified DNA fragment was inserted into the pET‐His6‐Ub expression vector between the Bsp TI and Eco RI sites. The obtained vectors were introduced into the Escherichia coli strain Rosetta2(DE3)pLysS (Novagen).

Protein concentration measurements

Protein concentrations were measured by absorption at 277–280 nm on the assumption that the total extinction coefficients of proteins were the same as the linear combination of the extinction coefficients of tryptophan and tyrosine residues present in the proteins.37 Since several fragment proteins (TAF4‐Q12 and TAF4‐Q34) did not contain any tryptophan or tyrosine residues, their concentrations were measured as follows. First, tyrosine‐containing mutant proteins were prepared. In the case of TAF4‐Q12, a tyrosine residue was added at the N‐terminus (called TAF4‐YQ12). In the case of TAF4‐Q34, F828 located between Q3 and Q4 was replaced by tyrosine (named TAF4‐Q34‐F828Y). Next, the concentrations of the purified tyrosine‐containing mutants TAF4‐YQ12 and TAF4‐Q34‐F828Y were determined by the absorption at 277 nm. Finally, the concentrations of wild‐type tyrosine‐less proteins were estimated by comparing the relative peak intensity of 1D‐NMR spectra to those recorded for tyrosine‐containing mutants with known concentrations.

CD spectroscopy

CD spectra were measured on a Jasco J‐820 spectropolarimeter. An assembling cell composed of a pair of quartz plates with a 0.1‐mm path length, which had been calibrated by measuring the absorption of a protein solution with known concentration, was used to record spectra between 250 and 190 nm at a protein concentration of 50 μM at 25°C.

NMR experiments

NMR experiments were performed on a Bruker DMX600 spectrometer with a triple‐axis gradient and triple‐resonance probe. A typical ¹H‐¹⁵N HSQC experiments were performed at protein concentration of 50 μM and 37°C. In the titration experiments, the concentration of ¹⁵N‐labeled protein was fixed at 50 μM and a various concentration of unlabeled protein was added. Sequence‐specific resonance assignments were obtained by analyzing the 3D‐HNCACB, 3D‐CBCA(CO)NH, 3D‐HNCO, 3D‐HN(CA)CO, and 3D‐hNcocaNH spectra recorded at protein concentration of 100 μM and 37°C.38, 39, 40, 41, 42 The solvent conditions used were 20 mM sodium acetate (pH 5.0), and 10% D₂O. The chemical shift value was referenced to 4,4‐dimethyl‐4‐silapentane‐1‐sulfonic acid (DSS). Spectra were processed with nmrPipe,43 and analyzed with XIPP, a new version of PIPP,44 which was kindly gifted by Dr. Garrett.

Small‐angle X‐ray scattering

SAXS measurements were performed with the spectrometer installed at BL‐10C of Photon Factory, a synchrotron radiation facility of Institute of Materials Structure Science, High Energy Accelerator Research Organization (Tsukuba, Japan). The scattering intensity in the q‐range between 0.02 and 0.2 Å⁻¹ was recorded with a PILATUS 2M detector. The q‐range was estimated by diffraction from the standard sample, silver behenate. Here, q = 4πsinθ/λ, where 2θ and λ (= 0.1488 nm) are the scattering angle and X‐ray wavelength, respectively. The two‐dimensional data measured were converted into a one‐dimensional curve by a circulation average and the scattering profile, I(q), of the protein was then obtained using the standard procedure of data corrections for cell and background scattering, transmission, and beam intensity.

During X‐ray exposure, protein solutions were continuously flowed in a 1 mm thickness cell at a rate of 0.3 μL s⁻¹ to avoid the radiation damage by X‐ray beam. All samples were confirmed to be free of radiation damage at an exposure time of 120 s.

Supporting information

Supporting Information Figures.