Introduction

The 60-amino-acid-long homeodomain (HD) is one of the most important eukaryotic DNA-binding motifs, highly conserved in sequence, structure and mechanism of DNA binding1. The fold adopted by the HD has been studied in a number of high-resolution HD-DNA complex high-resolution structures solved by X-ray crystallography and NMR spectroscopy (for examples, see2,3,4,5). Despite the similarity in the structure of the DNA-binding motif, HD-containing proteins regulate many distinct biological processes. In particular, cell identities are controlled by the large family of HOX transcription factors6,7,8,9, while early embryonic development and tumorigenesis are controlled by the PBC and MEIS families10,11. PBX1 (PBC family), MEIS1 and PREP1 (MEIS family) belong to a class of HDs characterized by a three-amino-acid-loop-extension (TALE) between the first and second α-helix of the HD12. PREP1 (also known as PKNOX1) or MEIS1 form with PBX1, a stable DNA-independent heterodimer, through the PBC-A domain of PBX1 and the conserved N-terminal MEIS-A and MEIS-B domains of PREP1 or MEIS113,14. PBX1:PREP1 and PBX1:MEIS1 pairs are necessary for subcellular localization and PBX1 stability15,16,17, while the presence of both HDs is required for DNA binding, as the proteins on their own show limited affinity for DNA18,19. In contrast to that of PREP/MEIS:PBX, HOX:PBX complex formation is DNA-dependent. HOX and PBX1 bind DNA cooperatively and their dimerization requires a highly conserved hexapeptide motif located N-terminal to the HOX HD, which contacts a pocket created by the TALE motif of PBX3,9,20,21,22,23. However, due to DNA conformability, Extradenticle (EDX; the Drosophila homolog of PBX1) retains some cooperativity with Ultrabitorax (Ubx) even after the removal of the Hox hexapeptide23. Recent evidence show an additional hexapeptide-independent mode of PBX recruitment through an additional motif, UbdA, located immediately C-terminal to the HOX HD24,25.

As shown in many crystal and NMR structures3,21,26, HD transcription factors contact DNA recognition sites through residues N51 and R55 in the third α-helix of the HD. N51 mediates the bidentate hydrogen bond with an adenine base, while R55 forms two hydrogen bonds with a guanine and both contact DNA in the major groove3. In addition to these hydrogen bonds, van der Waals interactions and water-mediated hydrogen bonds with the DNA sugar–phosphate backbone occur. Common ways to increase a transcription factor specificity for DNA involve interactions with the minor groove of DNA and/or cooperative partnerships with other transcription factors. For example, the binding of HD protein MATa1 with MATα2 depends on the 21-residues C-terminal tail of MATα2, located immediately after the HD2,27. The C-terminal tail of MATα2 undergoes a conformational change upon binding DNA in the presence of MATa1, thus becoming ordered, allowing contact to the MATa1 HD at a surface that does not participate in DNA binding27. It was also observed for HOX proteins that even non-specific interactions with DNA can increase DNA-binding affinity21.

PREP1 does not have a hexapeptide-like motif and strongly associates with PBX1 in the absence of DNA; therefore, the DNA-binding mechanisms of PREP1-PBX1 and PBX-HOX must be different. We have recently reported that DNA binding induces a conformational change in the full-length PBX1:PREP1 heterodimer19. Moreover, PBX1 and PREP1 can bind DNA independently when uncoupled, but require a specific functional interaction at the HD level for high affinity, thereby being sensitive to the DNA-binding state of the partner19. However, no further information is available on the dimerization and DNA binding activity of the PBX1-PREP1 combination at the HD level.

In this study, we have explored the functional and physical-chemical properties of isolated HDs and compared the affinity to DNA of different PREP1 HD constructs. We have quantified the ability of each of these HDs to bind DNA on their own and in combination. PREP1 HD, while specifically discriminating between different DNA motifs, binds DNA with low affinity. The presence of a preformed PBX1 HD:DNA complex significantly increases the DNA affinity of PREP1. Moreover, the binding properties of the PREP1 HD are different depending on the presence of C- and N-terminal flanking regions. The DNA interactions of two PREP1 HD constructs that differ in the amino- and carboxy-terminal extensions were studied by NMR spectroscopy. Indeed, the data show that the PREP1 HD carries important determinants for specificity and its flanking sequences are essential for formation of a high affinity complex with DNA. Most importantly, our data demonstrate that also PBX1:PREP1 HDs cooperate on DNA binding.

Results

Recombinant HDs

Purification of high quality individual PBX1 or PREP1 C-terminally truncated or full-length proteins was unsuccessful, as the single proteins, if not co-expressed, aggregate during the first purification step28. Isolated HDs of PREP1 and PBX1 were used and their interactions with DNA analyzed. Four different PREP1 HD constructs were cloned (Fig. 1A), the shortest (PREP1hd) consisted almost only of the HD (68 amino acids, from G257 to S325), and the longest (PREP1HD, from Q240 to F344) included residues from both N- and C-terminal sides. In addition, also two intermediate lengths, PREP1257–344 (PREP1hd-C) and PREP1240–325 (PREP1hd-N), were expressed. The 240-QLQLQL-245 stretch in the N-terminal extension is computationally predicted by I-Tasser29 to adopt α-helical conformation; the C-terminal extends to residue 344, which corresponds to the break point of protein stability, as previously determined28. The PBX1HD (residues 227–317) construct was identical to that previously used for X-ray crystallography studies3.

Figure 1
figure 1

Panel A: Homeodomain (HD) protein constructs used for characterizations described in this study, in comparison with full-length proteins. PBX1HD and HOXB1HD constructs are the same used in the X-ray crystallographic study3. PREP1 HD is present in four lengths, PREP1257-325 (PREP1hd), PREP1240-344 (PREP1HD) and two intermediate lengths (PREP1257-344 - PREP1hd-C - and PREP1240-325 -PREP1hd-N-), where respectively the N- or C-terminal extensions of the HD are omitted. Production of the proteins is described in the Materials and Methods section. PBC-A and PBC-B domains of PBX1 are those required for its dimerization with PREP1. TALE lies between helix 1 and helix 2 of the HD. MEIS-A and MEIS-B domains are PREP1 motifs required for heterodomerization with PBX1. Panel B: Consensus sequences for DNA binding by PREP1-PBX1: the very frequent decameric (PMH) and the less frequent octameric (PH) oligonucleotides30,31,32, are both highlighted in red in the sequence. Control probes correspond to two sequences that do not contain any PBX1 orPREP1 binding site. Oligonucleotides used for FP were 5′ labelled with 6-carboxyfluorescein (6-FAM) (see the Materials and Methods).

Full size image

The recombinant proteins were purified by glutathione S-transferase (GST) affinity chromatography as previously described28, and the proteins used for electrophoretic mobility shift assay (EMSA), circular dichroism (CD), and NMR spectroscopy studies were further purified as described (see the Materials and Methods section). An SDS PAGE (polyacrylamide gel electrophoresis) of the four PREP1 HDs, after the first chromatographic step (GST affinity), is shown in Supplementary Figure S1A, while gel-filtration profiles and SDS PAGEs of purified PREP1HD and PREP1hd are shown in Figure S1B and C and show the degree of purity of the preparations.

PREP1 HD C-terminal extension contains important determinant for DNA binding

The DNA sequence that was employed to measure binding to the HD, was based on ChIP-seq data on whole embryo trunk and several murine and human cell lines30,31,32,33. These studies have provided consensus sequences for DNA binding by PREP1-PBX1: the very frequent decameric (TGATTGACAG) and the less frequent octameric (TGATTGAT) oligonucleotides conform to the above consensus sequences. 5′-fluorescently labeled oligonucleotides (PMH, and PH) were used to measure the affinity of the individual HDs to DNA by fluorescence polarization (FP) and EMSA. A second form of the latter (PH*) corresponds to the sequence used to determine the X-ray structure of PBX1-HOXB1 HDs3,30, and aspecific sequences were used as control probes (Fig. 1B).

We first compared the affinities of the four PREP1-isolated HD constructs for the PREP1-PBX1 specific PMH, as determined by FP (see below). PREP1HD has a binding affinity (KD) (15.4 μM) two-fold lower than that of PREP1hd (31.8 μM) (Table 1). The C-terminal tail seems to contribute more significantly than the N-terminal tail, as the KD of PREP1hd-C (15.5 μM) is the same as that of PREP1HD (Table 1), whereas the KD of PREP1hd-N (23.7 μM) is reduced. These results suggest that the presence of PREP1 HD C-terminal extension residues is important for high affinity DNA binding. Through FP, the PREP1HD mutant was tested, in which K331, K333, K334, and K335, located in the C-terminal extension, were all mutated to alanine (PREP1HDKKKK→AAAA); lysine residues are positively charged and might be responsible for the higher affinity of PREP1HD and PREP1hd-C to the negatively charged DNA. Indeed, mutation of the four lysine residues increased the KD to 21 μM, thereby suggesting that the contribution of the C-terminal tail is based in part on an electrostatic effect (Table 1).

Table 1 KD values for individual HDs with different DNA sequences, measured by FP.
Full size table

How isolated PREP1 and PBX1 HDs bind to different sequences

A comparison of the affinities reveals that PREP1HD discriminates between binding sites (Table 1), exhibiting a preference for PMH (KD 15.4 μM) with respect to PH (KD 23.5 μM), and control probe (KD 158.7 μM). PREP1hd consistently shows a higher KD with any of the DNA motifs compared to PREP1HD; this indicates a generally lower affinity for DNA.

PBX1HD has a KD for PMH (0.36 μM) 6-fold lower than for PH (2.1 μM) and the control oligo (2.5 μM), indicating that PBX1HD has higher affinity but lower specificity for DNA. The same affinity (KD 3.3 μM) is obtained with PH* the sequence of which is absolutely identical to that used in the crystal structure study3. The high affinity of PBX1HD for the control oligo was confirmed also for a second control (control*, KD 2.1 μM). The data are summarized in Table 1.

In EMSA, DNA probes (4 μM) were incubated with increasing amounts of HD, using protein:DNA ratios of 0.5 (protein at 2 μM), 1 (protein at 4 μM), and 2 (protein at 8 μM). To visualize DNA and proteins, the electrophoresis gels were stained with ethidium bromide and subsequently with coomassie blue.

As shown in Fig. 2A, PREP1HD forms two different complexes with PMH, the first is visible with all protein:DNA ratios (lanes 1–3 left panel of Fig. 2A), and the second, slower migrating diffused band, at a protein:DNA ratio of 2 (lanes 2, 3, left panel of Fig. 2A). The latter band likely represents binding of two PREP1HD to DNA. In the cases of PH (central panel) and control probes (right panel), the bands tend to be diffused as if the complexes were less stable, dissociating while the gel is running. Nevertheless, the migration of the protein (coomassie staining) perfectly correlates with that of DNA (ethidium bromide staining).

Figure 2: EMSA with single HDs.
figure 2

DNA probes were incubated with different amounts of HD. Protein:DNA ratios were 0.5 (DNA excess), 1 (equimolar ratio), or 2 (protein excess). Gels were stained both with ethidium bromide and coomassie blue to visualize DNA or proteins, respectively. Panel A: PREP1HD forms two different complexes with PMH: the first is visible, at protein:DNA ratio of 0.5 (lanes 1 and 2, left panel), and the second as a smeared band, at protein:DNA ratio of 2 (lane 3, left panel). This slower migrating band might represent binding of two PREP1HD proteins to DNA. In the case of PH (central panel) and control probes (right panel), a specific DNA complex is not formed. Panel B: PREP1hd forms a single complex with PMH; whereas binding to control and PH oligos is weak and no specific bands are formed. Panel C: PBX1HD forms a single complex with PMH and PH oligos at protein:DNA ratio of 0.5 (lane 1); by increasing the protein concentration, a slower migrating band becomes visible, possibly due to binding of a second monomer to DNA. PBX1 binds also the control oligo.

Full size image

PREP1hd shows (Fig. 2B) the same behavior as PREP1HD for PMH, but binding to the control oligo is absent, as no bands are evident.

In the case of PBX1HD (Fig. 2C), a single complex with both PMH and PH is seen only at a protein:DNA ratio of 0.5 (lane 1), but at higher ratios, a slower migrating band becomes visible. As for PREP1, this extra band probably corresponds to binding of a second monomer to DNA. Such behavior in EMSA has been previously reported also for other DNA-binding proteins34. While PREP1HD and PREP1hd do not bind to the control probe, in the case of the PBX1HD clear bands are formed, indicating little discrimination of PBX1HD between sequences. This agrees with the results of the ChIP-seq analysis30,31, which failed to identify a real consensus motif for DNA sites bound uniquely by PBX1. In the literature examples of PBX1 consensus sequences are reported35,36. However, in those studies it has not been investigated whether PBX1 was binding in heterodimeric form.

Also PREP1hd-C and PREP1hd-N form a single complex with PMH oligo, as shown in the Supplementary Figures 2A and B.

PREP1 and PBX1 HDs synergize on binding DNA

We determined by FP if PREP1HD and PREP1hd bind to a preformed PBX1HD:DNA complex. PBX1HD (at the concentration of its KD for PMH, 0.3 μM) and PMH (at 9 nM) were pre-mixed and then titrated with increasing concentrations of PREP1HD or PREP1hd. In this case, the KD of PREP1HD and PREP1hd were 1.9 μM and 5.1 μM, respectively (Table 2), both significantly lower than for the single titrations of PREP1HD and PREP1hd with PMH. As a negative control, we titrated PREP1HD with the preformed HOXB1HD:PMH complex; in this case the KD value remained within the range of affinity of PREP1HD alone (Supplementary Figure S3C), in agreement with literature data showing that PREP1 does not bind HOXB137. Thus, the combination with PBX1HD specifically increases the DNA affinity for PMH of both PREP1HD and PREPhd.

Table 2 FP KD measurement of PREP1 HDs in the presence of a preformed PBX1HD: DNA complex.
Full size table

Through EMSA, the interactions of HD heterodimers with DNA were also investigated. The experiments were performed using a preformed HD:PMH complex (fixed molar ratio, ensuring that half of the DNA was free and half bound to the HD), titrated with increasing concentration of the second HD.

The PBX1HD:PMH complex (protein at 2 μM, DNA at 4 μM) was titrated with PREP1HD (2, 4, and 8 μM). Binding of PREP1HD to the preformed PBX1HD:DNA complex is cooperative, as PREP1HD preferentially binds the PBX1HD:DNA complex rather than free DNA; however, no PREPHD:DNA complex is visible. Already at low PREP1HD concentrations (Fig. 3A, lanes 3 and 4) we observe a slower migrating band that runs above monomeric PREP1HD:DNA and PBX1HD:DNA complexes (lanes 1 and 2, respectively). Mass spectrometry analysis of this band (lane 5, indicated with an asterisk) confirmed the presence of both PREP1HD and PBX1HD (see Supplementary Figure S2C for the sequence coverage, while protein correspondence in National Center for Biotechnology Information (NCBI) database are shown in Table 3).

Figure 3: Evaluation of the cooperative binding between pairs of HDs.
figure 3

Panel A: A sample containing a preformed PBX1HD:PMH complex (at 4 and 8 μM, respectively) was titrated with increasing amounts of PREP1HD (2, 4 and 8 μM). PREP1HD binds preferentially to the PBX1HD:PMH complex rather than to the free DNA; the DNA pool decrease appreciably upon an increase in PREP1HD concentration. At the lower PREP1HD concentrations (lane 3) a slower migrating band is observed above the monomeric PREP1HD:DNA and PBX1HD:DNA complexes (identified in lanes 1 and 2, respectively). The band marked with a star in lane 5 was analyzed by mass spectrometry. Panel B: Titration of PREP1HD: PMH complex with PBX1HD. PREP1HD (4 μM) and PMH (4 μM) were titrated with increasing concentrations of PBX1hHD (2, 4 and 8 μM). Upon increase of PBX1HD a slower migrating band appears, corresponding to PREP1HD:DNA:PBX1HD complex. Panel C: Titration of PBX1HD: PMH complex with PREP1hd. PBX1HD (4 μM) and PMH (4 μM) were titrated with increasing concentrations of PREP1hd (2, 4 and 8 μM). Upon increase of PREP1hd only a band corresponding to PREP1hd:DNA is visible, therefore, PREP1hd in EMSA does not show clear binding to the preformed PBX1HD:DNA complex. Panel D: Titration of PMH:PBX1HD with PREP1hd-N. PMH and PBX1HD at a fixed concentration (8 μM and 4 μM, respectively) were titrated with increasing amounts of PREP1hd-N (2, 4, 8, and 16 μM). Upon an increase in PREP1hd-N no retarded band above the individual PMH:PREP1hd-N and PMH:PBX11HD complexes (lanes 3-5) are observed Panel E: Titration of PBX1HD:PMH with PREP1hd-C. PMH oligo and PBX1HD at a fixed concentration (8 μM and 4 μM, respectively) were titrated with increasing concentrations of PREP1hd-C (2, 4 and 8 μM). Upon an increase in PREP1hd-C, a slower migrating band corresponding to PREP1hd-C :DNA: PBX1 HD complex is visible.

Full size image
Table 3 Mass spectrometry analysis of EMSA titrations bands from Fig. 3.
Full size table

Likewise, in the converse experiment a PREP1HD:PBX1HD:DNA complex appears upon titration of the PREP1HD-DNA complex with PBX1HD (Fig. 3B).

As a control, we titrated the HOXB1HD-PMH preformed complex with increasing amounts of PREP1HD. Again, a PREP1HD:HOXB1HD:DNA complex was not observed, as we could not visualize any retarded band (Supplementary Figure S3D). In conclusion, EMSA experiments, in agreement with FP, confirm that binding of PREP1HD to the DNA site is relatively weak; it is clearly cooperative with PBX1 but not with HOXB1.

PREP1hd does not show a clear cooperative behavior with the PBX1HD:PMH preformed complex in EMSA (Fig. 3C). Upon increase of PREP1hd, we observe only bands that correspond to PREP1hd:DNA or to PBX1 HD:DNA complexes. A weak slow-migrating band, more intense than the control no-PREP1hd (lane 1), may be visible only at high concentration of PREP1hd (lane 5). This result agrees with the FP data, where we observed a weak cooperative effect.

However, a clear cooperative effect is observed with PREP1hd-C and not with PREP1hd-N, indicating that the sequences responsible for cooperative binding are located in the C-terminal extension (Fig. 3D,E).

PREPHD N-terminal and C-terminal extensions are mainly unstructured

Structurally, PBX1 HD is well characterized, both in the presence and in the absence of DNA, and in complex or not with Hox proteins3,38,39,40. In order to have a structural rationale for the increased DNA binding affinity of PREP1HD, we investigated whether the N- and the C- terminal extensions of PREP1HD were structured and if their presence influences/alters the HD structural core. Circular dichroism (CD) spectra (see Supplementary Figure S4) indicate that PREP1HD does not have a higher content of α-helix with respect to PREP1hd, thereby suggesting that the N- and C- terminal extensions of PREP1HD are unstructured. Moreover, superposition of the two-dimensional (2D) 1H-15N HSQC (heteronuclear single quantum correlation) NMR spectra of 15N-labeled PREP1hd and PREP1HD (Fig. 4A) reveals a high similarity between the two constructs. The majority of PREP1HD resonances (magenta) almost coincide with those of PREP1hd (blue), whereas amide resonances belonging to the N- and C-terminal extensions of PREP1HD all cluster in the random coil region between 8.0 and 8.5 ppm on the 1H NMR axis. Residues located at the N-terminus (residues 242-QLQL-245) have backbone chemical shift values suggestive of α-helical conformation (CSI = 1) (see Supplementary Figure S5). Accordingly, a chemical shift (13Cα, 13CO, 13Cβ, 15N, 1Hα, 1HN)-based three-dimensional (3D) model of PREP1HD generated by CS23D2.041 reveals that the N- and C-terminal extensions of PREP1HD are unstructured, with the exception of a small helical turn involving residues 242-QLQL-245 (see Supplementary Figure S5), in agreement with I-Tasser predictions. The systematic comparison of amide chemical shifts between PREP1HD and PREP1hd shows that the major differences are located not only around residues G256-S257 and S324-S325 (that correspond to N- and C-terminal residues of PREP1hd, but are internal in PREP1HD), but also in more internal regions (K260-R263, H269 and A270, I317-D323). This suggests that the presence of weak intramolecular interactions between the N- and C-terminal extensions and the HD core (Fig. 4B,C). In agreement with this hypothesis, we observe by CD an approximate 4 °C increase in the melting temperature (Tm = 58.24 °C) of PREP1HD compared to PREP1hd (Tm = 54.86 °C), as shown in Fig. 4D. In conclusion, a comparison of the CD and 1H-15N HSQC spectra of both constructs indicates on the one hand that the N- and C-terminal extensions of PREP1HD are mainly unstructured, but on the other hand suggests that the N- and C- terminal extensions weakly interact with the HD structural core. This was assessed by the increased PREP1HD Tm and by the chemical shift differences observed within the HD structured region.

Figure 4
figure 4

Panel A: Superposition of 1H-15N HSQC spectra of PREP1HD (magenta) and PREP1hd (blue). Panel B: Combined amide chemical shift difference (Δδ) between PREP1HD and PREP1hd corresponding residues. Panel C: Cartoon representation of the PREP1HD model: the residues (in blue) shared with PREP1hd, the N- and C- terminal extensions (in cyan) of PREP1HD. Spheres indicate PREP1HD residues showing significant (>average) amide Δδ with respect to PREP1hd. Panel D: Normalized CD melting curves for PREP1HD (magenta) and PREP1hd (blue) at 222 nm.

Full size image

NMR titration of 15N PREP1HD and 15N PREP1hd with DNA

We next investigated the interaction of DNA oligo PMH with 15N-labeled PREP1HD and PREP1hd using chemical shift perturbation, a very sensitive tool to detect residues that directly interact with a ligand or are indirectly affected by binding. Upon addition of substoichiometric quantities of unlabelled PMH to both 15N-labeled PREP1HD and PREP1hd we observe small chemical shift displacements (CSD) and significant disappearance and/or intensity reduction of several amide peaks in the 1H-15N HSQC spectra, mainly affecting the three helices of the HD (Fig. 5B and Figures S5, S6, and S7). This behavior is indicative of binding in the intermediate exchange regime on the NMR chemical shift time scale, compatible with dissociation constants in the low micromolar range, as measured by FP (Table 1).

Figure 5: NMR titration of 15N-labeled PREP1hd and PREP1HD with PMH DNA oligo.
figure 5

Data reported correspond to a protein:PMH molar ratio of 1:0.5. Panel A: Superposition of 1H-15N HSQC spectra of free (black) and PMH-bound (red) PREP1HD. (left) and PREP1hd (right) Panel B: Plot showing the peaks intensity reduction observed upon PMH binding to PREP1hd and PREP1HD. Panel C: Residues disappearing or showing significant amide chemical shift displacement (as reported in the CSD plot in Supplementary Figure S6) are highlighted on PREP1HD structural model (top, generated using CS23D2.0 web server) and PREP1hd NMR structure (bottom, 1 × 2 N.pdb,).

Full size image

Already at substoichiometric ratio, in both constructs peak disappearance is mostly evident for the third α-helix of the HD, involving residues 312-NARRRIL-318 that are fundamental for DNA interaction19,21,42. Comparison of the peak intensity reduction and/or disappearance occurring in PREP1HD and PREP1hd, clearly shows that the broadening effect is much more pronounced in PREP1HD than in PREP1hd (Fig. 5B), in agreement with its higher affinity for DNA. The N- and C-terminal extensions of PREP1HD (i.e., residues preceding G257 and following S325, respectively) also experience a clear intensity reduction, thereby suggesting their involvement in binding. As a matter of fact, significant amide CSD were also observed for residues located on the N- and C-terminal tails of PREP1HD before the first α-helix (L245, Q254, K262, R263) and just after the third α-helix (S324, S325) (see Supplementary Figure S8).

In order to have a 3D visualization of the DNA binding pocket, we have highlighted on the NMR 3D models of PREP1hd and PREP1HD, those residues experiencing peak disappearance or significant CSD upon DNA interaction (CSD > average + standard deviation) (Fig. 5C).

Collectively these data confirm that the third α-helix is important for DNA binding and suggest that the N- and C-terminal extensions of PREPHD also contribute to the interaction, possibly increasing the affinity.

Discussion

Transcription factor activity depends on specific interactions between amino acids and critical DNA bases, as well as on interactions with other protein partners. PREP1 binds DNA through its highly-conserved HD, well separated from the PBX1-dimerization domain. Until now, the high-affinity binding to DNA has been solely attributed to the dimerization of PBX1 and PREP1 at the N-terminus. Here, we have shown that the isolated PREP1 HD binding to DNA involves also residues outside of the HD core. Although the measured affinities do not apply in vivo, where the full-length proteins and not just the HDs are present, this work represents the first biochemical and biophysical characterization of the isolated PREP1 HD.

1. As already demonstrated by several high-resolution structures3,21,26. mutational analysis19 and on the basis of studies of other HDs43, the positively charged N312-L318 stretch in the third helix of the HD contacts directly DNA. In agreement with these findings, the backbone amide peaks that correspond to these residues undergo substantial peak intensity reduction upon addition of DNA at substoichiometric concentrations in 2D 1H15N HSQC experiments.

2. Both the N- and C-terminal regions adjacent to the PREP1 HD core are unstructured (except for a small helical turn involving residues 242-QLQL-245). However, they stabilize the core HD structure increasing the Tm by 4 °C and promote PREP1 HD binding to DNA increasing affinity.

3. Indeed, the 2D 1H15N HSQC peak intensity reduction upon DNA binding is more pronounced in PREP1HD compared to that of PREP1hd, and this suggests that the N- and C-terminal extensions increase DNA affinity, in agreement with EMSA (Fig. 3) and FP (Table 1) experiments. The NMR peaks corresponding to residues located in the N- and C- terminal extensions (such as Q254, K262, R263 before the first α-helix, and S324, S325 after the third α-helix) decrease their intensity in the presence of DNA, supporting this interpretation. Furthermore, mutation of four lysine residues (K331, K333, K334, and K335) (Table 1) in the electropositive patch of the C-terminal extension decreases DNA affinity, suggesting that a charge component contribution to the binding. Conceivably, the residues flanking the PREP1 HD core may not be directly involved in DNA binding, but can stabilize the three DNA-bound α-helices. In this model, the regions flanking PREP1 HD may contribute to DNA recognition, restricting the possible conformations of the DNA-binding domain thus favoring a unique interaction with DNA.

Previous studies have reported that the affinity for DNA of the full-length complex is in the low-nanomolar range19. This suggests that the interaction between the N-terminal regions is responsible for the profound changes in the DNA-binding domain activity. In addition, conformational changes and flexibility of the PBX1:PREP1 complex favor the robustness of the protein-DNA interaction, as deletion of the poly-alanine linker of PBX1 decreases 150-fold the affinity of the whole complex19. Now we demonstrate that the presence of PBX1HD enhances the ability of PREP1HD to bind DNA, with an 8-fold decrease in the KD. The preference for a preformed PBX1HD:DNA complex over free DNA is well visible in EMSA assays, where already at very low molar ratios PREP1HD binds to PBX1HD:DNA, rather than to free DNA, with no detectable intermediates. Mass spectrometry analysis of the slower migrating band (Fig. 3A) confirms the presence of both PBX1HD and PREP1HD with high sequence coverage. Cooperativity is specific for PBX1HD as HOXB1HD does not cooperate with PREP1 (data shown in Supplementary Information).

When PBX1 and PREP1 HD contact DNA, the binding state of one determines the DNA preference of the other, as already observed by mutational studies of key residues in the third α-helix of PBX1 and PREP1 HDs19.

Our fluorescence and EMSA binding data suggest that the most likely mechanism mediating cooperativity is the presence of an uncharacterized motif, namely, the hexapeptide in Hox, which is involved in the dimerization of the HDs. In this respect, a region flanking the PREP1 HD at the C terminus, although mainly unstructured, appears to be important to enhance the cooperativity.

However, in the absence of an X-ray crystal structure, we cannot exclude other mechanisms: for example, cooperativity may be mediated by a conformational change in the DNA or by overlapping contacts of PBX1 and PREP1 with DNA23,34,44,45,46.

PREP1 on its own binds DNA weakly, but its affinity for the specific site increases when combined with PBX1. This in vitro cooperativity of PREP1 and PBX1 might result in a precise organization of transcriptional regulation, in which PREP1 contributes to selectivity and PBX1 to affinity.

Materials and Methods

Cloning of recombinant proteins

HDs proteins, listed in Fig. 1A (PREP1, PBX1) were all subcloned in a pGEX-6P vector (GenBank: KM817768) using BamHI/XhoI cloning sites. Primers used for cloning are listed in Table S1. The DNA template of PREP1 HD in which the four residues K231, K233, K234, and K235 were mutated to alanine was purchased from Genscript Corporation (Piscataway, NJ) and showed the following sequence:

5′CAGCTTCAGTTACAGTTAAACCAAGATCTCAGCATCTTGCATCAAGATGATGGTTCATCTAAGAACAAGAGGGGCGTCCTGCCAAAGCATGCCACGAACGTGATGCGGTCCTGGCTCTTCCAGCACATCGGGCATCCCTACCCAACAGAGGATGAGAAAAAACAGATTGCTGCTCAGACAAATTTGACACTACTCCAAGTCAACAACTGGTTCATCAATGCCAGAAGACGAATTCTTCAGCCAATGTTGGATTCAAGTTGTTCAGAGACCCCCGCAACAGCGGCAGCAACTGCTCAGAACCGGCCAGTTCAGAGG 3′.

Protein expression and purification

Recombinant proteins were expressed and purified as described in the literature28. Expression was performed in BL21(DE3)pLysS E. coli strain (Promega, Madison, WI). Uniformly 15N- and 13C-15N-labelled PREP1HD and PREP1hd were expressed by growing E. coli BL21(DE3)pLysS cells (Promega) in minimal bacterial medium containing 15NH4Cl, with or without 13C-D-glucose (both from CortecNet, Voisins-le-Bretonneux, France). Protein expression was induced with 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 16–20 h at 16 °C. Cells were harvested by centrifugation and resuspended in lysis buffer (20 mM Tris-HCl pH 7.4, 1 M NaCl, 10% glycerol, 0.5 mM EDTA (Ethylenediaminetetraacetic acid) and 1 mM DTT (Dithiothreitol)) supplemented with Protease Inhibitor Cocktail Set III Calbiochem (Billerica, MA). Sonication was done with a Bandelin Sonopuls (Berlin, Germany) sonicator for 3 × 45 seconds with 5 pulses at 40% of max power. After sonication, bacterial lysates were cleared by centrifugation at 40,000 × g for 1 hour. Proteins were purified using glutathione-sepharose 4B beads (GE Healthcare, Milano, Italy) according to manufacturer’s instructions. GST was cleaved off with 10 μg/ml of 3C- preScission protease (GE Healthcare) for 16 hours at 4 °C. GST-free proteins were diluted in buffer (20 mM Tris pH 7.4, 10% glycerol, 0,5 mM EDTA, 0,5 mM EGTA and 1 mM Dithiotheitol) to a final 0.1 M NaCl concentration and purified on a Resource S (GE Healthcare) cation exchange column using a 0.1–1.0 M NaCl gradient for elution. The recombinant proteins used for EMSA, CD, or NMR spectroscopic experiments (PBX1, PREP1HD and PREP1hd HDs) were further purified by size-exclusion chromatography on a Superose 6 10/300 column (GE Healthcare) equilibrated in 20 mM Na2HPO4/NaH2PO4 pH 7.2, 150 mM NaCl, 5% glycerol, and 1 mM DTT at a flow rate of 0.3 ml/min. Protein markers used for size-exclusion chromatography were the gel filtration standards from Bio-Rad (Hercules, CA). Protein concentrations were determined by the UV absorption at 280 nm. The extinction coefficients of the proteins are reported in Supplementary Table S2, and they were calculated using the online tool ProtParam47.

DNA oligonucleotides

DNA oligonucleotides for EMSA, NMR and CD experiments were unlabeled, however, those used in FP were 5′-labeled with 6-FAM (6-Carboxyfluorescein) dye. They were purchased from Sigma and purified by HPLC by the manufacturer. 0.1–1 mM double-strand DNA oligonucleotides were prepared by annealing equimolar concentration of each strand in 10 mM Tris-HCl pH 8.0, 50 mM NaCl, 1 mM EDTA. This mixture was heated to 95 °C for 5–10 minutes and slowly cooled to room temperature. The sequences of the DNA motifs are shown in Fig. 1B; for NMR spectroscopic characterization we used the DNA oligo PMH.

Fluorescence polarization

Serial dilutions of the protein were performed in 20 mM Na/K phosphate pH 7.2, 150 mM NaCl, 5% glycerol and 1 mM TCEP (Tris(2-carboxyethyl) phosphine). 20 μl of the protein solutions were transferred in a microplate and mixed with a fixed volume of 6-FAM-DNA (9 nM as final concentration). Binding reactions were incubated to reach steady state equilibrium at room temperature, in the dark, for 30 minutes. Fluorescence polarization assays were performed in a 20 μl final volume in flat bottom, black plates (Corning® Low Volume 384 Well Black Flat Bottom Polystyrene NBS™ Microplate). Saturation binding isotherms were generated at a fixed concentration of DNA with increasing concentrations of the proteins (from 0.1 to 60.0 μM) in 20 mM Na/K phosphate pH 7.2, 150 mM NaCl, 5% glycerol and 1 mM TCEP. KD values were calculated using non-linear Michaelis-Menten fittings. Readings were acquired on a Tecan Infinite F200 fluorimeter, with an excitation filter at 485 nm (20 nm bandwidth) and emission filter at 535 nm (25 nm bandwidth). Data were fitted with GraphPad Prism (http://www.graphpad.com/scientific-software/prism). One-site Michaelis-Menten binding model best fitted the experimental curves. All the experiments were performed in triplicate and the final KD values reported in the tables correspond to the average of three independent experiments.

EMSA

Non-denaturing gels were prepared in a final volume of 15 ml with 10% Acrylamide:bisacrylamide solution (37.5:1), 0.8% glycerol, 0.5 × TBE (Tris/Borate/EDTA, 45 mM Tris base, 45 mM boric acid, 1 mM EDTA pH 8.3), 0.1% APS (Ammonium Persulfate) and 6 μl TEMED (Tetramethylethylenediamine). Binding reactions were assembled at room temperature in a total volume of 15 μl (in 20 mM Na/K Phosphate pH 7.2, 150 mM NaCl, 5% glycerol, 1 mM TCEP) and incubated 10–15 minutes before loading on non-denaturing 10% acrylamide gel in running buffer 0.5 × TBE. The gel was pre-electrophoresed, at 4 °C, for 20 minutes at 90 V. Electrophoretic running continued for 50 minutes at 4 °C. The gel was stained for 10 minutes in ethidium bromide (diluted 1:10000) in 50 ml of 0.5 × TBE, for DNA detection. Then the gel was stained in coomassie staining for protein detection. Poly (dI/dC) were purchased from Roche Diagnostics S.p.A. (Monza, Italy).

Mass spectrometry

The bands of interest were cut from gels and trypsinized as described48. Peptides were desalted as previously described49, dried in a Speed- Vac and resuspended in 7 μl of 0.1% formic acid reversed-phase capillary chromatography/electrospray ionization-mass spectrometry (LC-ESI-MS/MS) of 5 μL of each sample was performed on a Fourier transformed-LTQ mass spectrometer (FT-LTQ) (Thermo Electron, San Jose, CA). Peptide separation was achieved through a linear gradient from 100% solvent A (5% acetonitrile, 0.1% formic acid) to 20% solvent B (60% acetonitrile, 0.1% formic acid) over 20 min and from 20% to 80% solvent B in 5 min at a constant 0.3 μL/min flow rate on Agilent chromatographic separation system 1100 (Agilent Technologies, Waldbronn, Germany). The liquid chromatography system was connected to a 10.5 cm fused-silica emitter of 100 μm inner diameter (New Objective, Inc. Woburn, MA USA), packed in-house with ReproSil-Pur C18-AQ 3 μm beads (Dr. A. Maisch Gmbh, Ammerbuch, Germany) using a high-pressure bomb loader (Proxeon, Odense, Denmark).

Data acquisition mode was set to obtain one MS scan followed by five MS/MS scans of the five most intense ions in each MS scan. MS/MS spectra were limited to one scan per precursor ion followed by 1 min of exclusion. The mascot generic format (MGF) files were extracted using DATA SuperCharge (v.1.19, www.cebi.sdu.dk) while Database search was performed using Mascot Daemon v.2.3.2 set up with the following parameters: Database NCBInr, Taxonomy Homo Sapiens, enzyme Trypsin, Max missing cleavage 2, fixed modification carbamidomethyl (C), variable modification oxidation (M), peptide tolerance 10 ppm, MS/MS tolerance 0.5 Da, Instrument ESI-TRAP.

Circular Dichroism spectroscopy

CD spectra were acquired on a Jasco J-815 CD spectrometer at 20 °C, from 200 to 260 nm. Typical protein concentrations were 10 μM in 20 mM Na2HPO4/NaH2PO4 pH 7.2 and 150 mM NaF. Spectra were averaged over 4 scans and corrected by subtracting the buffer spectrum and smoothed. The observed ellipticity θ (mdeg) was converted into molar residue ellipticity (MRE, deg cm2 dmol−1). The same samples were used for thermal denaturation experiments, performed recording the θ values at 222 nm, from 10 °C to 96 °C, with 1 °C intervals, 1 °C/min rate, average time 0.5 sec and bandwidth 2 nm. The Tm was calculated with non-linear curve fitting of the equation:

where A0 and A1 are θ at the unfolded and folded states, respectively, A2 is ΔH/R (set to 105 as initial guess) and A3 is Tm.

NMR spectroscopy

NMR experiments were performed at 28 °C on a Bruker Avance 600 MHz spectrometer equipped with TCI cryoprobe and pulsed field gradients. Data were processed with TopSpin 3.2 (Bruker) and analyzed using CcpNmr Analysis 2.1.550. All the samples were dissolved in the same NMR buffer (20 mM Na/K phosphate buffer pH 6.0, 150 mM NaCl, 1 mM DTT, 5% glycerol, 5% D2O, 0.3 mM 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS)). Backbone (1HN, 15N, 13Ca,13CO) and 13C side chain resonances of PREP1HD recombinant protein were assigned through 2D and 3D 1H-15N HSQC, 1H-13C HSQC, HNCA, CBCA(CO)NH and CC(CO)NH NMR experiments, acquired at 28 °C on a 0.43 mM 15N-13C- labeled PREP1HD sample. As the 1H-15N HSQC spectra of 15N-labeled PREP1hd and PREP1HD constructs are almost super-imposable, PREP1hd 1HN and 15N chemical shifts were assigned based on PREP1HD assignment. NMR titrations were performed with stepwise additions of the DNA oligo PMH (8.1 mM) into 15N-labeled PREP1HD or PREP1hd (0.1 mM) up to a 1.5 molar excess. At each titration point, 1D 1H and 2D 1H-15N HSQC spectra were acquired. The weighted average of the 1HN and 15N chemical shifts displacements (CSD) upon binding was calculated as CSD = [(ΔHN2+ΔN2/25)/2]1/2 51. We considered residues showing CSD > average+σ0 as significantly affected upon DNA interaction. The σ0 value was calculated excluding any residue for which the CSD value was bigger than 3σ; recalculating σ and iterating these calculations until no further residues were excluded. This procedure has been used to avoid biasing the distribution by including the small number of residues with large CSDs51.

For each titration point and for each protein residue, we also calculated the decrease in peak intensity (I) as I/Io, where Io is the intensity of the peak in the free protein.

Secondary Structure and 3D Model of PREP1HD

The secondary structure elements were identified with CSI 3.0 server (http://csi3.wishartlab.com/cgi-bin/index.php)52 based on PREP1HD 13Cα, 13CO, 13Cβ, 15N, 1Hα, 1HN chemical shifts. Chemical shifts were also used as input to generate a 3D model of PREP1HD using CS23D2.0 web server (http://www.cs23d.ca/index.php)41.

Additional Information

How to cite this article: Zucchelli, C. et al. New Insights into Cooperative Binding of Homeodomain Transcription Factors PREP1 and PBX1 to DNA. Sci. Rep. 7, 40665; doi: 10.1038/srep40665 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.