Advertisement

A hydrophobic residue stabilizes dimers of regulatory ACT-like domains in plant basic helix–loop–helix transcription factors

Open AccessPublished:April 23, 2021DOI:https://doi.org/10.1016/j.jbc.2021.100708
      About a third of the plant basic helix–loop–helix (bHLH) transcription factors harbor a C-terminal aspartate kinase, chorismate mutase, and TyrA (ACT)-like domain, which was originally identified in the maize R regulator of anthocyanin biosynthesis, where it modulates the ability of the bHLH to dimerize and bind DNA. Characterization of other bHLH ACT-like domains, such as the one in the Arabidopsis R ortholog, GL3, has not definitively confirmed dimerization, raising the question of the overall role of this potential regulatory domain. To learn more, we compared the dimerization of the ACT-like domains of R (RACT) and GL3 (GL3ACT). We show that RACT dimerizes with a dissociation constant around 100 nM, over an order of magnitude stronger than GL3ACT. Structural predictions combined with mutational analyses demonstrated that V568, located in a hydrophobic pocket in RACT, is important: when mutated to the Ser residue present in GL3ACT, dimerization affinity dropped by almost an order of magnitude. The converse S595V mutation in GL3ACT significantly increased the dimerization strength. We cloned and assayed dimerization for all identified maize ACT-like domains and determined that 12 of 42 formed heterodimers in yeast two-hybrid assays, irrespective of whether they harbored V568, which was often replaced by other aliphatic amino acids. Moreover, we determined that the presence of polar residues at that position occurs only in a small subset of anthocyanin regulators. The combined results provide new insights into possibly regulatory mechanisms and suggest that many of the other plant ACT-like domains associate to modulate fundamental cellular processes.

      Keywords

      Abbreviations:

      ACT (aspartate kinase, chorismate mutase, and TyrA), ALPHA (amplified luminescent proximity homogeneous assay), 3-AT (3-amino-1,2,4-triazole), bHLH (basic helix–loop–helix), GL3ACT (ACT-like domains of GL3), GST (glutathione-S-transferase), pAD (GAL4 activation domain), pBD (GAL4 DNA-binding domain), PDB (Protein Data Bank), PheH (phenylalanine hydroxylase), PPI (protein–protein interaction), RACT (ACT-like domains of R), SD (synthetically defined), TCEP (Tris(2-carboxyethyl)phosphine), TF (transcription factor), Y2H (yeast-two hybrid)
      Control of gene expression relies on the proper organization of transcription factors (TFs) and other proteins on gene regulatory regions. This is largely accomplished through protein–DNA and protein–protein interactions (PPIs). Thus, in addition to harboring DNA-binding domains that interpret the cis-regulatory code in the genome, TFs are characterized by the presence of one or more PPI domains. The same TF can regulate different sets of genes in different cell types or conditions, a consequence of their ability to form different complexes as part of what is known as combinatorial control (
      • Brkljacic J.
      • Grotewold E.
      Combinatorial control of plant gene expression.
      ). Thus, determining how PPI domains participate in TF assembly is fundamental to understand gene regulation.
      The basic helix–loop–helix (bHLH) family of TFs is among the largest in animals (
      • Lambert S.A.
      • Jolma A.
      • Campitelli L.F.
      • Das P.K.
      • Yin Y.
      • Albu M.
      • Chen X.
      • Taipale J.
      • Hughes T.R.
      • Weirauch M.T.
      The human transcription factors.
      ) and plants (
      • Feller A.
      • Machemer K.
      • Braun E.L.
      • Grotewold E.
      Evolutionary and comparative analysis of MYB and bHLH plant transcription factors.
      ). The bHLH domain is structurally conserved, and it is about 60 amino acids long organized into two functionally distinct regions (
      • Ferré-D'Amaré A.R.
      • Prendergast G.C.
      • Ziff E.B.
      • Burley S.K.
      Recognition by Max of its cognate DNA through a dimeric b/HLH/Z domain.
      ,
      • Ma P.C.M.
      • Rould M.A.
      • Weintraub H.
      • Pabo C.O.
      Crystal structure of MyoD bHLH domain-DNA complex: Perspectives on DNA recognition and implications for transcriptional activation.
      ). The N-terminal basic region is responsible for binding to the canonical E-box DNA motif (CANNTG), but DNA recognition requires the formation of homodimers or heterodimers with other bHLH proteins, and such PPIs are mediated by the HLH region (
      • Massari M.E.
      • Murre C.
      Helix-loop-helix proteins: Regulators of transcription in eucaryotic organisms.
      ). In addition to the bHLH domain, members of this TF family often harbor other conserved motifs that participate in PPIs and that have contributed to the classification of the family (
      • Feller A.
      • Machemer K.
      • Braun E.L.
      • Grotewold E.
      Evolutionary and comparative analysis of MYB and bHLH plant transcription factors.
      ,
      • Pires N.
      • Dolan L.
      Origin and diversification of basic-helix-loop-helix proteins in plants.
      ,
      • Stevens J.D.
      • Roalson E.H.
      • Skinner M.K.
      Phylogenetic and expression analysis of the basic helix-loop-helix transcription factor gene family: Genomic approach to cellular differentiation.
      ). Some of these motifs are shared between plant and animal bHLH TFs, such as the leucine zipper motif that is often present immediately C terminal to the second helix of the HLH motif, and are important in stabilizing bHLH-mediated dimer formation and providing DNA-binding specificity to the homodimers or heterodimers (
      • Blackwood E.M.
      • Eisenman R.N.
      Max: A helix-loop-helix zipper protein that forms a sequence-specific DNA-binding compelx with Myc.
      ,
      • Bresnick E.H.
      • Felsenfeld G.
      The leucine zipper is necessary for stabilizing a dimer of the helix-loop-helix transcription factor USF but not for maintenance of an elongated conformation.
      ,
      • Shively C.A.
      • Liu J.
      • Chen X.
      • Loell K.
      • Mitra R.D.
      Homotypic cooperativity and collective binding are determinants of bHLH specificity and function.
      ). Other motifs appear to be specific to the plant kingdom and include the ACT-like domain, a fold first identified in the aspartate kinase, chorismate mutase, and TyrA enzymes (hence the name, ACT) (
      • Aravind L.
      • Koonin E.V.
      Eukaryotic transcription regulators derive from ancient enzymatic domains.
      ) and later shown to be present in about a third of the plant bHLH TFs (
      • Feller A.
      • Hernandez J.M.
      • Grotewold E.
      An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
      ). Similar to the structurally related bHLH protein interaction and function domain (
      • Feller A.
      • Yuan L.
      • Grotewold E.
      The BIF domain in plant bHLH proteins is an ACT-like domain.
      ,
      • Chang F.
      • Cui J.
      • Wang L.
      • Ma H.
      REPLY: The BIF domain is structurally and functionally distinct from other types of ACT-like domains.
      ), the ACT-like domain can mediate PPIs (
      • Feller A.
      • Hernandez J.M.
      • Grotewold E.
      An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
      ,
      • Cui J.
      • You C.
      • Zhu E.
      • Huang Q.
      • Ma H.
      • Chang F.
      Feedback regulation of DYT1 by interactions with downstream bHLH factors promotes DYT1 nuclear localization and anther development.
      ).
      ACT domains are 70 to 80 amino acids long and have primarily been found in proteins involved in the regulation or biosynthesis of amino acids. When part of enzymes, they can participate in allosteric regulation by pathway intermediates, frequently involving the formation of homodimers, or higher order structures (
      • Cho Y.
      • Sharma V.
      • Sacchettini J.C.
      Crystal structure of ATP phosphoribosyltransferase from Mycobacterium tuberculosis.
      ,
      • Grant G.A.
      The ACT domain: A small molecule binding domain and its role as a common regulatory element.
      ). Several ACT structures have been solved, some bound to ligands (
      • Grant G.A.
      The ACT domain: A small molecule binding domain and its role as a common regulatory element.
      ), and those ACT domains that are part of enzymes generally show a βαββαβ topology (
      • Curien G.
      • Biou V.
      • Mas-Droux C.
      • Robert-Genthon M.
      • Ferrer J.L.
      • Dumas R.
      Amino acid biosynthesis: New architectures in allosteric enzymes.
      ), although flexibility in the structure of the domain is becoming evident as more structures are solved (
      • Gai Z.
      • Wang Q.
      • Yang C.
      • Wang L.
      • Deng W.
      • Wu G.
      Structural mechanism for the arginine sensing and regulation of CASTOR1 in the mTORC1 signaling pathway.
      ,
      • Xia J.
      • Wang R.
      • Zhang T.
      • Ding J.
      Structural insight into the arginine-binding specificity of CASTOR1 in amino acid-dependent mTORC1 signaling.
      ,
      • Saxton R.A.
      • Chantranupong L.
      • Knockenhauer K.E.
      • Schwartz T.U.
      • Sabatini D.M.
      Mechanism of arginine sensing by CASTOR1 upstream of mTORC1.
      ).
      Maize R was the first plant bHLH regulator identified (
      • Ludwig R.
      • Habera L.F.
      • Dellaporta S.L.
      • Wessler S.R.
      Lc, a member of the maize R gene family responsible for tissue-specific anthocyanin production, encodes a protein similar to transcriptional activators and contains the myc-homology region.
      ), and its function is to regulate the accumulation of anthocyanin pigments, by physically interacting with the R2R3–MYB domain (MYB domain harboring two MYB repeats most similar to the second and third MYB repeats of the product of the c-myb proto-oncogene) of C1 (
      • Goff S.A.
      • Cone K.C.
      • Chandler V.L.
      Functional analysis of the transcriptional activator encoded by the maize B gene: Evidence for a direct functional interaction between two classes of regulatory proteins.
      ,
      • Grotewold E.
      • Sainz M.B.
      • Tagliani L.
      • Hernandez J.M.
      • Bowen B.
      • Chandler V.L.
      Identification of the residues in the Myb domain of maize C1 that specify the interaction with the bHLH cofactor R.
      ). R homodimer formation and DNA binding require an extended bHLH that includes a short leucine zipper motif but is inhibited by dimerization of the C-terminal ACT-like domain, which has a ββαββα organization (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ). Thus, the R ACT-like domain functions as a regulatory switch that dictates whether R-containing regulatory complexes are tethered to target genes through the bHLH or through the R2R3–MYB partner (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ). R is a member of subgroup IIIf of plant bHLH proteins (
      • Feller A.
      • Machemer K.
      • Braun E.L.
      • Grotewold E.
      Evolutionary and comparative analysis of MYB and bHLH plant transcription factors.
      ,
      • Pires N.
      • Dolan L.
      Origin and diversification of basic-helix-loop-helix proteins in plants.
      ). This subgroup also includes the partially redundant Arabidopsis GL3 (GLABRA3) and EGL3 (ENHANCER OF GL3) bHLH proteins, which participate in the specification of root epidermal cell fate (
      • Bernhardt C.
      • Lee M.M.
      • Gonzalez A.
      • Zhang F.
      • Lloyd A.
      • Schiefelbein J.
      The bHLH genes GLABRA3 (GL3) and ENHANCER OF GLABRA3 (EGL3) specify epidermal cell fate in the Arabidopsis root.
      ,
      • Bernhardt C.
      • Zhao M.
      • Gonzalez A.
      • Lloyd A.
      • Schiefelbein J.
      The bHLH genes GL3 and EGL3 participate in an intercellular regulatory circuit that controls cell patterning in the Arabidopsis root epidermis.
      ), in the control of trichome (leaf hair) formation (
      • Morohashi K.
      • Grotewold E.
      A systems approach reveals regulatory circuitry for Arabidopsis trichome initiation by the GL3 and GL1 selectors.
      ,
      • Morohashi K.
      • Zhao M.
      • Yang M.
      • Read B.
      • Lloyd A.
      • Lamb R.
      • Grotewold E.
      Participation of the Arabidopsis bHLH factor GL3 in trichome initiation regulatory events.
      ,
      • Payne C.
      • Zhang F.
      • Lloyd A.
      GL3 encodes a bHLH protein that regulate trichome development in Arabidopsis through interaction with GL1 and TTG1.
      ), and in the control of anthocyanin accumulation (
      • Gonzalez A.
      • Zhao M.
      • Leavitt J.M.
      • Lloyd A.M.
      Regulation of the anthocyanin biosynthetic pathway by the TTG1/bHLH/Myb transcriptional complex in Arabidopsis seedlings.
      ), by interacting with different R2R3–MYB proteins, participating in the formation of different MYB, bHLH, and WD (tryptophan–aspartic acid repeat-containing proteins) complexes (
      • Lloyd A.
      • Brockman A.
      • Aguirre L.
      • Campbell A.
      • Bean A.
      • Cantero A.
      • Gonzalez A.
      Advances in the MYB–bHLH–WD repeat (MBW) pigment regulatory model: Addition of a WRKY factor and co-option of an Anthocyanin MYB for betalain regulation.
      ,
      • Xu W.
      • Dubos C.
      • Lepiniec L.
      Transcriptional control of flavonoid biosynthesis by MYB–bHLH–WDR complexes.
      ). Notwithstanding the very similar domain structure of GL3/EGL3 and R, and the presence of an ACT-like domain (Fig. 1A), results are inconsistent on whether the C-terminal region of GL3 is capable of forming homodimers (
      • Patra B.
      • Pattanaik S.
      • Yuan L.
      Ubiquitin protein ligase 3 mediates the proteasomal degradation of GLABROUS 3 and ENHANCER OF GLABROUS 3, regulators of trichome development and flavonoid biosynthesis in Arabidopsis.
      ,
      • Wen J.
      • Li Y.
      • Qi T.
      • Gao H.
      • Liu B.
      • Zhang M.
      • Huang H.
      • Song S.
      The C-terminal domains of Arabidopsis GL3/EGL3/TT8 interact with JAZ proteins and mediate dimeric interactions.
      ).
      Figure thumbnail gr1
      Figure 1Homodimerization of the R and GL3 ACT-like domains. A, amino acid alignment between the ACT-like domains of R (residues 525–610) and GL3 (residues 552–637). The alignment was generated using ClustalW in BioEdit, version 7 (
      • Hall T.A.
      BioEdit: A user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT.
      ). Secondary structures were predicted from the ACT domain of phenylalanine hydroxylase (Protein Data Bank: 5FII). The resultant α-helices and β-strands are indicated by spirals and horizontal arrows, respectively. The numbers over colored arrowheads represent the amino acids that were substituted in the mutation assays, where the residues marked in same color were switched with each other. B, yeast two-hybrid assays probing interaction of R525–610 and GL3552–637 fused to the GAL4 activation domain (pAD) or GAL4 DNA-binding domain (pBD) in the yeast strain PJ69.4a (
      • James P.
      • Halladay J.
      • Craig E.A.
      Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast.
      ) containing the HIS3 and ADE2 genes under the control of GAL4-binding sites. Interaction is manifested by growth on SD media deficient in Leu/Trp/His and Leu/Trp/His/Ade (–L–T–H, –L–T–H–A) for 5 days, whereas SD media–deficient Leu/Trp (–L–T) selects for the bait and pray plasmids. Three independent transformants were analyzed for each plasmid combination corresponding to three biological replicates. ACT, aspartate kinase, chorismate mutase, and TyrA; SD, synthetically defined.
      Here, using a combination of yeast two-hybrid (Y2H) assays with the amplified luminescent proximity homogeneous assay (ALPHA), we demonstrate that similar to R, the ACT-like domain of GL3 can homodimerize, yet with significantly lower affinity. We used structural predictions combined with the limited sequence homology between these ACT-like domains to identify a key residue (Val 568 in R) that, when replaced with the corresponding residue in GL3 (Ser), significantly impairs dimerization strength. Conversely, the replacement of this Ser residue with Val in GL3 is sufficient to enhance dimerization by more than fivefold. The analysis of the distribution of these two residues across plant ACT-like domains showed that GL3 is likely the exception, since most ACT-harboring bHLH factors contain Val or other aliphatic residues at the equivalent 568 position. The evaluation of apparent equilibrium dissociation constants (Kd) with ALPHA for the various wildtype and mutant ACT-like domains allowed us to correlate strength of interaction with Y2H assay results, providing insights regarding the meaning of Y2H results as well as a possible explanation for prior inconsistencies (
      • Patra B.
      • Pattanaik S.
      • Yuan L.
      Ubiquitin protein ligase 3 mediates the proteasomal degradation of GLABROUS 3 and ENHANCER OF GLABROUS 3, regulators of trichome development and flavonoid biosynthesis in Arabidopsis.
      ,
      • Wen J.
      • Li Y.
      • Qi T.
      • Gao H.
      • Liu B.
      • Zhang M.
      • Huang H.
      • Song S.
      The C-terminal domains of Arabidopsis GL3/EGL3/TT8 interact with JAZ proteins and mediate dimeric interactions.
      ). Taken together, our studies provide significant insights regarding the dimerization of ACT-like domains.

      Results

      The GL3 and R ACT-like domains dimerize with very different affinities

      GL3 has a very similar domain organization as R, including a C-terminal region (residues 552–637) that shows the ββαββα secondary structure that characterizes the R ACT-like domain (RACT, residues 525–610; Fig. 1A). We used the Y2H assay to investigate dimerization of GL3552–637, in conditions in which R525–610 shows robust dimerization (
      • Feller A.
      • Hernandez J.M.
      • Grotewold E.
      An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
      ). For this, we fused GL3552–637 to the GAL4 activation domain (in plasmid pAD–GL3552–637; harboring the LEU2 selectable marker) and to the GAL4 DNA-binding domain (in plasmid pBD–GL3552–637; harboring the TRP1 selectable marker). We transformed pAD–GL3552–637 and pBD–GL3552–637 into the PJ69.4 yeast strain (
      • James P.
      • Halladay J.
      • Craig E.A.
      Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast.
      ) and assayed growth in synthetically defined (SD) media lacking leucine and tryptophan (–L–T); leucine, tryptophan, and histidine (–L–T–H); and leucine, tryptophan, histidine, and adenine (–L–T–H–A). In contrast to cells harboring pAD–R525–610 and pBD–R525–610, which grew robustly in –L–T–H–A media, no growth in –L–T–H (usually considered adequate for weak interactions) or –L–T–H–A (usually considered adequate for strong interactions) was observed for cells with pAD–GL3552–637 and pBD–GL3552–637 (Fig. 1B), which is consistent with published results (
      • Patra B.
      • Pattanaik S.
      • Yuan L.
      Ubiquitin protein ligase 3 mediates the proteasomal degradation of GLABROUS 3 and ENHANCER OF GLABROUS 3, regulators of trichome development and flavonoid biosynthesis in Arabidopsis.
      ). No growth in –L–T–H or –L–T–H–A media was observed for either R525–610 or GL3552–637 fused to pAD or pBD, in the presence of the corresponding empty plasmids (Fig. S1). The growth of the various yeast strains in selective media as an indication of PPI was complemented by β-galactosidase assays, taking advantage that the PJ69.4 yeast strain harbors the LacZ gene under a GAL4-controlled promoter (
      • James P.
      • Halladay J.
      • Craig E.A.
      Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast.
      ). β-Galactosidase assays confirmed the results obtained in selective media (Fig. S2). We also compared the growth of the various strains shown in Figure S3 in –L–T–H media supplemented with increasing concentrations (10–50 mM) of the HIS3 enzyme inhibitor, 3-amino-1,2,4-triazole (3-AT). The results show almost a perfect correlation between the ability of cells to grow in the presence of 3-AT and the enhanced selection provided by omitting adenine from the media (Fig. S3, compare with Fig. S2).
      Western blots of extracts of yeast cells expressing several proteins fused to the GAL4–AD and showing distinct interaction strengths in yeast showed similar levels of accumulation when using commercial antibodies to GAL4–AD (Fig. S4). These results indicate that the different strengths of interaction observed in yeast are unlikely a consequence of different stability of the proteins.
      To compare the dimerization binding affinities of GL3552–637 and R525–610, we expressed and affinity purified each protein fused to glutathione-S-transferase (GST) or the N6His tag (Fig. S5, AD) and determined the respective Kd values using the ALPHA, by performing competition and saturation binding assays (Fig. 2A). For the competition binding assay, purified untagged (i.e., without N6His or GST tag) versions of R525–610 and GL3552–637 (Fig. S5E) were used as competitors (Fig. 2A and Table 1). In agreement with R525–610 dimerizing very robustly (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ), it showed a Kd of 100 nM (Fig. 2B and Table 1). In contrast, in identical conditions, GL3552–637 dimerization is characterized by a much higher Kd of 1.3 to 1.5 μM (Fig. 2C and Table 1). Based on these results, we conclude that GL3552–637 is capable of forming homodimers, but this interaction is too weak to be detected by Y2H.
      Figure thumbnail gr2
      Figure 2Different binding strengths characterize the ACT-like domains of R and GL3. A, application of amplified luminescent proximity homogeneous assay (ALPHA) to determine dissociation constants (Kd) of different combinations of ACT-like domains. When N6His-tagged and GST-tagged ACT-like domains interact, this brings in close proximity (<200 nm) the donor and acceptor beads, resulting in high-energy emission at 620 nm (
      • Yasgar A.
      • Jadhav A.
      • Simeonov A.
      • Coussens N.P.
      AlphaScreen-based assays: Ultra-high-throughput screening for small-molecule inhibitors of challenging enzymes and protein-protein interactions.
      ). Apparent Kd values are determined either by saturation (left) using excess of one protein over the other, or competition (right), using proteins that lack the N6His or GST tags. B, determination of the apparent Kd for RACT and C, GL3ACT by competitive binding assays using ALPHA. Various concentrations of untagged RACT (0–8 μM) and GL3ACT (0–10 μM) were incubated with a mixture of 100 nM of N6His-RACT and N6His-GL3ACT and 100 nM of GST-RACT and GST-GL3ACT, respectively. The Kd values were calculated by one-site fit model. Each of the lines corresponds to one biological replicate, and each experiment was done in triplicate. The Kd values shown correspond to the average ± standard deviation. ACT, aspartate kinase, chorismate mutase, and TyrA; GST, glutathione-S-transferase.
      Table 1Summary of the dimerization strength of the ACT domains of R and GL3 evaluated by saturation binding using ALPHA and Y2H assays
      ProteinMutationY2HApparent Kd value (μM)
      The Kd values represent the average ± standard deviation of three biological replicates.
      RWT–L–T–H–A0.10 ± 0.01
      V568SNo growth1.72 ± 0.20
      L569H–L–T–H–A0.16 ± 0.03
      A573SNo growth0.43 ± 0.06
      V568S/L569HNo growth1.59 ± 0.23
      V568S/A573SNo growth0.99 ± 0.17
      L569H/A573S–L–T–H0.11 ± 0.02
      V568S/L569H/A573SNo growth1.51 ± 0.10
      GL3WTNo growth1.54 ± 0.12
      S595V–L–T–H0.26 ± 0.01
      H596LNo growth1.49 ± 0.24
      S600ANo growth1.59 ± 0.52
      S595V/H596L–L–T–H0.27 ± 0.05
      S595V/S600A–L–T–H0.36 ± 0.04
      H596L/S600ANo growth2.23 ± 0.33
      S595V/H596L/S600ANo growth1.77 ± 0.19
      a The Kd values represent the average ± standard deviation of three biological replicates.

      Identification of key dimer interface residues that specify interaction strength of ACT domains

      To identify the residues that are potentially responsible for the different dimerization strengths of RACT and GL3ACT, we took advantage of homology modeling using phenylalanine hydroxylase (PheH; Protein Data Bank [PDB]: 5FII) as a template, as it showed the highest secondary structure similarity with RACT and GL3ACT. The RACT and GL3ACT predicted secondary structures showed overall very similar organization, despite having only 33% amino acid identity (Fig. 1A). The homology-modeled monomeric RACT and GL3ACT were then subjected to homodimer predictions using GalaxyHomomer (
      • Baek M.
      • Park T.
      • Heo L.
      • Park C.
      • Seok C.
      GalaxyHomomer: A web server for protein homo-oligomer structure prediction from a monomer sequence or structure.
      ) and GalaxyRefineComplex (
      • Heo L.
      • Lee H.
      • Seok C.
      GalaxyRefineComplex: Refinement of protein-protein complex model structures driven by interface repacking.
      ). We considered two possible configurations for the respective dimeric forms, side-by-side (Fig. 3) and face-to-face (Fig. S6) arrangements based on what is known on how other ACT domains dimerize (
      • Grant G.A.
      The ACT domain: A small molecule binding domain and its role as a common regulatory element.
      ). Previously, we showed that substitution mutations of S570A, Q572A, and S574A in RACT abolished its dimerization in Y2H assays (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ). The effect of these mutations, which are located on the β2-strand (Figs. 3 and S6), is more consistent with the side-by-side configuration, as these three amino acids would be right at the dimer interface. Based on this potential configuration for the RACT and GL3ACT dimers, the β2-strand in each is predicted to be important for the interaction. There are three amino acid differences in this region: V568 in RACT corresponding to S595 in GL3ACT, L569 in RACT corresponding to H596 in GL3ACT, and A573 in RACT corresponding to S600 in GL3ACT (Figs. 1A and 3). Significantly, the three amino acids in RACT are hydrophobic, whereas the corresponding ones in GL3ACT are polar.
      Figure thumbnail gr3
      Figure 3Predicted structures of RACT and GL3ACT homodimers in side-by-side configuration. The structures of RACT and GL3ACT were predicted based on the structure of the ACT domain of phenylalanine hydroxylase (Protein Data Bank: 5FII). The red, orange, and yellow colors represent the amino acids substituted in our mutational assays. The amino acids (S570, Q572, and S574) of which mutations significantly abolish the ability of the ACT-like domain to dimerize in previous study (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ) were indicated in gray. ACT, aspartate kinase, chorismate mutase, and TyrA.
      To determine the contribution of each of these three residues to the different dimerization strengths of RACT and GL3ACT, we made the corresponding single, double, and triple amino acid substitutions in each of the two ACT domains and tested their ability to interact in Y2H assays. When we substituted V568 by S in R525–610 (R525–610;V568S), dimerization was completely abolished, reflected in no growth in –L–T–H or –L–T–H–A and very low β-galactosidase activity, compared with the robust dimerization of R525–610 (Fig. 4A, compare #2 and #3; and Fig. S2A). A similar dimerization abatement was observed for the A573S substitution (Fig. 4A, compare #2 and #5; and Fig. S2A). In contrast, the L569H mutation had no effect on the dimerization activity of R525–610, as evidenced by the similar growth and β-galactosidase activity with yeast strain harboring the wildtype versions (Fig. 4A, compare #2 and #4; and Fig. S2A). The double substitutions followed the expected trend, with the V568S/L569H and V568S/A573S showing no R525–610 dimerization (Fig. 4A, #6 and #7; and Fig. S2A). Unexpected was the observation that the yeast strain harboring the L569H/A573S double mutant displayed significantly better interaction than the strain with the single A573S mutant (Fig. 4A, compare #5 and #8; and Fig. S2A).
      Figure thumbnail gr4
      Figure 4Effect of mutations on the ability of RACT and GL3ACT to homodimerize in yeast two-hybrid assays. A, homodimerization of RACT mutants and GL3ACT mutants in yeast. B, homodimerization of GL3ACT and GL3ACT mutants in yeast. The RACT and GL3ACT harboring single, double, or triple substitutions were fused to either the GAL4 DNA-activation domain (pAD) or GAL4 DNA-binding domain (pBD). The dimeric interactions of the RACT and GL3ACT mutants were examined on the SD media (–L–T, –L–T–H, –L–T–H–A) for 5 days. Autoactivation controls are provided in . The empty vector combinations (pAD and pBD represented as #1 in A and B) were used as negative control. Three independent transformants were analyzed for each plasmid combination corresponding to three biological replicates. SD, synthetically defined.
      The reciprocal residue changes were incorporated into GL3552–637 and tested for dimerization using the Y2H assay. Compared with GL3552–637, which showed no interaction, yeast cells harboring pAD–GL3552–637;S595V and pAD–GL3552–637;S595V showed growth in –L–T–H, but not in –L–T–H–A (compare #2 and #3 in Fig. 4B), consistent with increased β-galactosidase activity (Fig. S2B), indicating that the S595V substitution is sufficient to confer the ability of GL3 to dimerize to levels that can now be detected in Y2H assays. No significant interaction was observed for any of the other single or multiple mutations, with the exception of S595V/H596L and S595V/S600A, which showed growth in –L–T–H and moderate β-galactosidase activity (Figs. 4B and S2B). While the proposed structural models can accurately predict the effect of the other mutations, it fails to explain the ability of the S595V/H596L and S595V/S600A mutants in GL3 to dimerize.
      To quantitatively evaluate and compare the interaction strengths of the various R and GL3 ACT mutants, Kd values were determined by saturation binding assays using ALPHA and the respective proteins tagged with GST or N6His (Fig. 2A). The apparent Kd values determined by saturation binding for R525–610 were very similar to those determined by competition binding and corresponded to 100 ± 27 nM (competition binding; Fig. 2B) and 100 ± 10 nM (saturation binding; Table 1, Figs. 5A and S7). Also consistent between both methods, the Kd values for GL3552–637 were 1.28 ± 0.17 μM estimated by the competition assay (Fig. 2C) and 1.54 ± 0.12 μM by the saturation assay (Table 1, Figs. 5B and S7). We then evaluated apparent Kd values for all the mutants analyzed by Y2H. ACT-like domain variants that showed no interaction by Y2H had Kd values that were 400 nM or higher, whereas those that interacted in yeast displayed Kd values in the range of 100 to 400 nM (Fig. 4 and Table 1). Our results indicate that the homodimerization of RACT and GL3ACT most likely is a consequence of a side-by-side association as found in other ACT domains with a βαββαβ structure (
      • Schuller D.J.
      • Grant G.A.
      • Banaszak L.J.
      The allosteric ligand site in the Vmax-type cooperative enzyme phosphoglycerate dehydrogenase.
      ,
      • Patel D.
      • Kopec J.
      • Fitzpatrick F.
      • McCorvie T.J.
      • Yue W.W.
      Structural basis for ligand-dependent dimerization of phenylalanine hydroxylase regulatory domain.
      ,
      • Zhang S.
      • Huang T.
      • Ilangovan U.
      • Hinck A.P.
      • Fitzpatrick P.F.
      The solution structure of the regulatory domain of tyrosine hydroxylase.
      ), in which are the α1-helix and the β2-strand at the interface of the two subunits (Fig. 3).
      Figure thumbnail gr5
      Figure 5Equilibrium dissociation constant (Kd) evaluation for RACT, GL3ACT, and respective mutants. A, saturation binding assays with RACT mutants. B, saturation binding assay with GL3ACT mutants. The 100 nM N6His-RACT and 500 nM N6His-GL3ACT were incubated with different concentrations of GST-RACT (0–4 μM) and GST-GL3ACT (0–5 μM) proteins at room temperature for 2 h, before collecting amplified luminescent proximity homogeneous assay responses. The saturation binding curves of RACT and GL3ACT were obtained by one-site fit model. We show the mean of three biological repeats with two technical repeats each with error bars denoting the standard deviation of the mean. All the data used for the saturation binding curves are provided in , and the mean Kd values are shown in .

      Heterodimer formation of subunit interface mutants

      The analyses so far involved the formation of homodimers, in which both subunits were identical (i.e., either wild type or mutant). To determine the role of the amino acid residues in the subunit interface identified as important for dimerization, we investigated their role in the formation of heterodimers, in which the subunits harbored different mutations. For this, we combined the various mutants for RACT and GL3ACT in Y2H assays.
      As we showed before (#3 in Figs. 4A and S2A), V568S abolishes RACT dimerization, yet pAD–R525–610;V568S robustly interacts with pBD–R525–610 (Fig. 6A, compare #2 and #5). A similar situation was observed for R525–610;A573S; it is unable to form homodimers (Fig. 4A, #5; and Fig. S2A), yet it forms robust heterodimers with pBD–R525–610 (Fig. 6A, compare #4 and #9). When the V568S and A573S mutations are combined in each subunit, heterodimerization continues to happen, yet is somewhat impaired (reflected in heterodimers growing in –L–T–H but not in –L–T–H–A, Fig. 6A, #7). The V568S/L569H double mutant of RACT also formed a heterodimer (Fig. 6A, #6), and the Kd value (0.33 ± 0.02 μM) was comparable to that of V568S/A573S (0.27 ± 0.02 μM) in saturation binding assays (Fig. S8). A rather different situation was observed with GL3ACT, as none of the heterodimers tested, including those that contained one of the subunits harboring the S595V mutation that allowed dimerization (Fig. 4B, #3; and Fig. S2B), showed growth (Fig. 6B). Taken together, these results highlight the importance of V568 in R525–610 and S595 in GL3552–637 in establishing the strength of the dimerization. The heterodimer results are consistent with an antiparallel orientation of the dimers, as shown in Figure 3.
      Figure thumbnail gr6
      Figure 6Effect of mutations on the ability of RACT and GL3ACT to form heterodimers in yeast two-hybrid assays. A, heterodimerization test of RACT and its mutants in yeast. B, heterodimerization test of GL3ACT and its mutants in yeast. The effect of single and multiple amino acid mutations of RACT and GL3ACT on heterodimer formation was evaluated on selective SD media (–L–T, –L–T–H, –L–T–H–A) for 5 days. SD, synthetically defined.

      Analysis of other plant ACT domains

      Given our results showing the importance of V568 for the dimerization of RACT, we asked how often this residue if present in ACT-like domains associated with bHLH TFs, and whether the presence of this (or similar residue) was correlated with robust homodimer formation.
      For this, we first retrieved the sequence of all 175 maize bHLH factors from GRASSIUS (grassisus.org) (
      • Yilmaz A.
      • Nishiyama M.Y.
      • Garcia-Fuentes B.
      • Souza G.M.
      • Janies D.
      • Gray J.
      • Grotewold E.
      GRASSIUS: A platform for comparative regulatory genomics across the grasses.
      ). Because of the low amino acid sequence conservation between ACT-like domains (
      • Feller A.
      • Hernandez J.M.
      • Grotewold E.
      An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
      ,
      • Feller A.
      • Yuan L.
      • Grotewold E.
      The BIF domain in plant bHLH proteins is an ACT-like domain.
      ), we determined whether they harbored an ACT domain by predicting secondary structures using PSIPRED (
      • Buchan D.W.A.
      • Minneci F.
      • Nugent T.C.O.
      • Bryson K.
      • Jones D.T.
      Scalable web services for the PSIPRED protein analysis workbench.
      ). We found that 44 of 175 bHLH factors had an ACT-like domain, a frequency similar to what was previously determined for Arabidopsis (
      • Feller A.
      • Hernandez J.M.
      • Grotewold E.
      An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
      ), and this domain is always located at the C-terminus, as is the case for R and GL3 (Fig. 1A). ACT-like domains characterize members of bHLH subfamilies I, II, III, IV, and Vb (Table S1 and Fig. 7A), as was previously determined for Arabidopsis (
      • Feller A.
      • Hernandez J.M.
      • Grotewold E.
      An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
      ). These results demonstrate that the origin and distribution of ACT-like domains in bHLH TFs precedes the divergence of monocot and dicot.
      Figure thumbnail gr7
      Figure 7Analyses of other maize ACT-like domains. A, phylogenetic relationship of 175 maize and 137 Arabidopsis bHLH TFs. The phylogenetic tree was constructed using the entire amino acid sequence of each of the proteins using maximum likelihood methods with 1000 bootstraps with MEGA7 (
      • Kumar S.
      • Stecher G.
      • Tamura K.
      MEGA7: Molecular evolutionary genetics analysis version 7.0 for bigger datasets.
      ). The maize bHLH subfamilies were assigned according to what was done in Arabidopsis TFs of the same clade (
      • Pires N.
      • Dolan L.
      Origin and diversification of basic-helix-loop-helix proteins in plants.
      ). The subfamilies are indicated by the lines in different colors with the respective numbers. The bootstrap support values and taxon names were omitted in the interest of making the illustration legible. The number of ACT-like domains in each subfamily is indicated by the colors in the outer circle, and the numbers in red indicate how many members of each subfamily were found to homodimerize in yeast two-hybrid (Y2H) experiments. B, examples of homodimerization analyses of maize ACT-like domains by Y2H assays. A total 42 of 44 maize bHLH TFs were subjected to the Y2H assay. The interactions were determined on selective SD media (–L–T, –L–T–H, –L–T–H–A). The #1 and #2 indicate the homodimer of RACT as a positive control and empty vector combination as a negative control, respectively. Autoactivation tests are presented in . All the relevant information for the 42 constructs and all Y2H results are presented in . ACT, aspartate kinase, chorismate mutase, and TyrA; bHLH, basic helix–loop–helix; SD, synthetically defined; TFs, transcription factors.
      To determine how many of these maize ACT-like domains formed homodimers detectable by Y2H assays, we cloned 42 of 44 in the pAD and pBD Y2H vectors and analyzed their ability to form homodimers. Of the 42 ACT-like domains tested, 12 (including R) were capable of homodimerizing (Figs. 7B and S9). All the ACT-like domains displaying robust dimerization in the Y2H assays belonged to bHLH subfamilies III(d+e), IIIf (to which R and GL3 belong), and IVa (Fig. 7A, dimerization indicated by numbers in red). The remaining 30 ACT-like domains that showed no yeast growth in selective media are either incapable of dimerizing, or the dimerization is too weak (Kd >400 nM) to be determined by Y2H assays, as we established for GL3ACT.
      We then investigated whether the presence of Val (or similar aliphatic amino acid) at the equivalent position of V568 in RACT was associated with the strength of the dimerization. The alignment of the 42 maize ACT-like domains showed that this position is occupied by Val, Ile, Leu, or Ala (and an exceptional Pro). However, the presence of a Val was not indicative of whether the respective ACT-like domain would dimerize in Y2H or not (Fig. S10). Taken together, our results demonstrate that while V568 is clearly essential for robust dimerization in R, the presence of this residue is not an indication of whether an ACT-like domain will dimerize in Y2H assays or not. Moreover, our results suggest that strong dimerization is a property of only a subset of maize ACT-like domains associated with bHLH TFs.

      Discussion

      We used homology modeling to predict the structure of the ACT-like domains of R and GL3. We show that both ACT-like domains can dimerize, albeit with very different affinities, evidenced by over an order of magnitude difference in their Kd values. This knowledge was then applied to identify key subunit interface residues responsible for the different interaction strengths. The RACT β-strand that provides the subunit interface is characterized by more aliphatic residues, whereas the GL3ACT corresponding region contains polar residues at the same locations. Indeed, the V568S and A573S substitutions completely abolish RACT dimerization in Y2H assays (Figs. 4 and S2) reflected in a significant reduction in the affinity of the interaction (Figs. 4 and S2, Table 1), underscoring the importance of this β2-strand in subunit interaction. Strikingly, the S595V substitution in GL3ACT (the corresponding substitution to V568S in RACT), but not the S600A (corresponding to A573S in RACT), confers GL3ACT the ability to dimerize with an affinity comparable of that of RACT (Figs. 4 and S2, Table 1). The position of V568 in our models coincides with the position of I65 in the structure of the ACT domain of PheH (PDB: 2PHM), residues that were determined to be important in maintaining the hydrophobic packing of the protein (Fig. S11) (
      • Carluccio C.
      • Fraternali F.
      • Salvatore F.
      • Fornili A.
      • Zagari A.
      Structural features of the regulatory ACT domain of phenylalanine hydroxylase.
      ).
      These results are significant from several perspectives. Dimerization of the ACT-like domain in R was proposed to inhibit dimerization and DNA binding of the adjacent bHLH motif (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ). Thus, the presence of strong ACT dimerization could be used as an indication of whether the DNA-binding activity of the bHLH is regulated by the ACT-like domain or not. GL3 is an Arabidopsis R ortholog and also controls anthocyanin accumulation (
      • Gonzalez A.
      • Zhao M.
      • Leavitt J.M.
      • Lloyd A.M.
      Regulation of the anthocyanin biosynthetic pathway by the TTG1/bHLH/Myb transcriptional complex in Arabidopsis seedlings.
      ). However, GL3 is one of three Arabidopsis bHLH factors that can control anthocyanin accumulation, the other two being TT8 and EGL3 (
      • Gonzalez A.
      • Zhao M.
      • Leavitt J.M.
      • Lloyd A.M.
      Regulation of the anthocyanin biosynthetic pathway by the TTG1/bHLH/Myb transcriptional complex in Arabidopsis seedlings.
      ). Both GL3 and EGL3 have Ser at the V568 position, whereas TT8 has Thr. This contrasts dramatically with all the maize ACTs that have an aliphatic residue in this position (Fig. S10). To determine if this is a particularity of the Arabidopsis anthocyanin regulators, we retrieved the sequence of the ACT domain from the predicted anthocyanin regulators from many plant species (Fig. S12). Interestingly, the identity of the residue at the V568 position can clearly distinguish between different subclades, including a MYC1/R subclade with mainly Val/Ala at this position, a GL3 subclade with Ser/Cys at this position, and the TT8 group with occasionally Val/Ile but largely a Thr for Brassicaceae family (Fig. S12). To what extent these variations reflect differences in the regulatory mechanisms in which these TFs participate remains to be investigated.
      The possibility to modulate the dimerization strength of ACT-like domains, for example, by replacing the residue at the equivalent V568 position with Ser/Cys, opens interesting opportunities to alter the regulatory activity of bHLH factors harboring these structurally conserved domains. It remains to be determined if the dimerization of ACT-like domains is affected by the interaction with small molecules, as is the case for many ACT domains (
      • Grant G.A.
      The ACT domain: A small molecule binding domain and its role as a common regulatory element.
      ). But if these were the case, it is possible to envision a situation in which the regulation of specific metabolic or developmental pathways is made more or less responsive to particular ligands, with potential in agriculture.
      There are many different methods to assay PPIs, each with its strengths and limitations (
      • Rao V.S.
      • Srinivas K.
      • Sujini G.N.
      • Kumar G.N.
      Protein-protein interaction detection: Methods and analysis.
      ,
      • Sharma V.
      • Ranjan T.
      • Kumar P.
      • Pal A.K.
      • Jha V.K.
      • Sahni S.
      • Prasad B.D.
      Protein–protein interaction detection: Methods and analysis.
      ,
      • Phizicky E.M.
      • Fields S.
      Protein-protein interactions: Methods for detection and analysis.
      ). By and large, it is not known what is the limit in the interaction strength that permits to detect PPIs by the Y2H method. Studies conducted on the interaction of retinoblastoma to other proteins suggested that the weakest binding constant that would give a positive by the Y2H assay is around 1 μM (
      • Phizicky E.M.
      • Fields S.
      Protein-protein interactions: Methods for detection and analysis.
      ). While this certainly depends on many factors, including the specific proteins tested and their level of expression in yeast, our results are largely in agreement and indicate that interactions involving the R and GL3 ACT domains with Kd values above 400 nM will be difficult to assess using Y2H. Moreover, the difference in Kd value ranges between hybrids that support growth in –L–T–H (∼110–340 nM) and those that support growth in –L–T–H–A (<160 nM) is not very different. These results indicate that, while for extreme cases, the additional selection furnished by omitting adenine from the media or adding 3-AT can help discriminate strong versus weak interactions (
      • Stynen B.
      • Tournu H.
      • Tavernier J.
      • Van Dijck P.
      Diversity in genetic in vivo methods for protein-protein interaction studies: From the yeast two-hybrid system to the mammalian split-luciferase system.
      ,
      • Ou B.
      • Yin K.Q.
      • Liu S.N.
      • Yang Y.
      • Gu T.
      • Wing Hui J.M.
      • Zhang L.
      • Miao J.
      • Kondou Y.
      • Matsui M.
      • Gu H.Y.
      • Qu L.J.
      A high-throughput screening system for Arabidopsis transcription factors and its application to Med25-dependent transcriptional regulation.
      ), there is a range of interaction strengths for which the distinction is less clear.
      The overall results provide novel functional insights into how structurally conserved, yet sequence divergent, ACT-like domains present in one of the largest families of plant TFs dimerize. Our studies bring forward a model for the dimerization of these ACT-like domains, models that are consistent with the mutational data. The possibility to alter the strength of the interactions by switching single conserved residues provides a powerful tool to elucidate the function of these domains in vivo.

      Experimental procedures

      Plasmids

      The gateway entry clones harboring ACT-like domains were synthesized by GeneArt Gene Synthesis (ThermoFisher Scientific) or generated by PCR from full-length open reading frame clones generated by the Maize TFome project (
      • Burdo B.
      • Gray J.
      • Goetting-Minesky M.P.
      • Wittler B.
      • Hunt M.
      • Li T.
      • Velliquette D.
      • Thomas J.
      • Gentzel I.
      • Brito M.d.S.
      The maize TFome–development of a transcription factor open reading frame collection for functional genomics.
      ). The pENT constructs were recombined into the pDEST17 and pDEST15 (ThermoFisher Scientific) for recombinant protein expression and purification. The pAD–GAL4–GWC1 and pBD–GAL4–GWC1 vectors (
      • Machemer K.
      • Shaiman O.
      • Salts Y.
      • Shabtai S.
      • Sobolev I.
      • Belausov E.
      • Grotewold E.
      • Barg R.
      Interplay of MYB factors in differential cell expansion, and consequences for tomato fruit development.
      ), referred to as pAD and pBD, were used for Y2H assays. For untagged ACT-like domains of R and GL3, the corresponding regions were inserted into the BamHI/HindIII sites of pET28–SUMO (
      • Mossessova E.
      • Lima C.D.
      Ulp1-SUMO crystal structure and genetic analysis reveal conserved interactions and a regulatory element essential for cell growth in yeast.
      ) using the following primer sets. R: forward 5′-CACCGGATCC GACGCCGGCACCAGCAACGTCA-3′, reverse 5′-AAGCTTTCACCGCTTCCCTATAGCTTTGCGA-3′; GL3: forward 5′-ATGGATCCTTTACTGGTTTAACCGATAA-3′ reverse 5′-GCAAGCTT TCAACAGATCCATGCAACCC-3′.

      Y2H assays

      We used the PJ69.4a yeast strain that harbors integrated in the genome the HIS3 and ADE2 selectable genes controlled by GAL4-binding sites (
      • James P.
      • Halladay J.
      • Craig E.A.
      Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast.
      ). The pAD and pBD constructs were transformed into PJ69.4a by the lithium chloride method with a slight modification (
      • Gietz R.D.
      • Schiestl R.H.
      Quick and easy yeast transformation using the LiAc/SS carrier DNA/PEG method.
      ). Briefly, yeast cells (absorbance at 600 nm ≈ 0.4) were resuspended in 100 mM lithium acetate, 50% PEG, and 30 μg/ml salmon sperm DNA (Invitrogen), and incubated with 500 ng of pAD and pBD plasmid at 30 °C for 30 min followed by 42 °C for 20 min and incubated 30 °C for 1 h. To select yeast colonies, the transformed yeast cells were cultured on an SD medium lacking Leu and Trp (–L–T) at 30 °C for 3 days. The positive colonies were subcultured on SD–Leu/–Trp (SD–L–T), SD–His/–Leu/–Trp (SD–H–L–T), and SD–Ade/His/–Leu/–Trp (SD–A–H–L–T) to test physical interactions. The strength of homodimerizations in the yeast cells was quantitatively measured using the β-galactosidase assay (
      • Rose M.
      • Botstein D.
      Construction and use of gene fusions to lacZ (β-galactosidase) that are expressed in yeast.
      ) The crude extracts of the yeast cells were incubated with 0.8 mg/ml o-nitrophenyl-β-d-galactopyranoside (Sigma), and the resultant absorbance values at 420 nm were normalized to protein concentration, evaluated by Bradford assays (
      • Bradford M.M.
      A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding.
      ). Arbitrary β-galactosidase units were calculated as follows: (absorbance at 420 nm × 1.7)/(0.0045 × time of incubation × volume of extract × protein concentration in mg/ml). Yeast growth and β-galactosidase assays were conducted in three biological replicates, each obtained from independent transformants.

      Recombinant protein purification

      The ACT-like domains of R and GL3 were fused to the C terminus of His6 or GST and transformed into the BL21(DE3) strain. The cells were cultured in 100 ml (absorbance at 600 nm ≈ 0.4) at 37 °C, and induced by 0.6 mM IPTG (final concentration). After incubation at 37 °C for 2 h, the cells were harvested and resuspended in 5 ml of modified PBS buffer (500 mM NaCl,10 mM Na2HPO4, 2 mM KH2PO4, and 0.05% Triton X-100) followed by sonication (Misonix Ultrasonic Liquid Processors S-4000). The cell lysate was centrifuged at 3500g for 20 min and filtered through Miracloth (Calbiochem).
      For poly-His-tagged protein purification, the cell filtrate was incubated with 200 μl of a 50% (w/v) slurry of nickel–nitrilotriacetic acid agarose (ThermoFisher Scientific), equilibrated with PBS buffer at 4 °C for 1 h, and washed with 5 ml of PBS buffer containing 50 mM imidazole. The resin was eluted by three subsequent washes of 500 μl each of 400 mM imidazole in PBS buffer. For GST-tagged protein purification, 200 μl of glutathione–agarose beads (Roche) in PBS buffer were mixed with the cell filtrate and incubated at 4 °C for 1 h. The resin-bound protein was rinsed with 5 ml PBS buffer and eluted three times with 500 μl each of PBS buffer containing 10 mM reduced glutathione (pH 8.0). To purify the untagged ACT-like domains of R and GL3, the pET28–SUMO constructs were transformed into BL21(DE3). The bacterial cells were cultured in 500 ml until an absorbance at 600 nm ≈ 0.4 and further incubated in 0.6 mM IPTG at 16 °C for 12 h. The harvested cells were resuspended in buffer containing 10 mM Tris–HCl, 150 mM NaCl, 0.5 mM Tris(2-carboxyethyl)phosphine (TCEP)–HCl, and 5 mM imidazole followed by sonication and centrifugation. Cleared supernatants were bound to the columns by gravity flow and washed two times with 50 ml buffer (50 mM Tris–HCl and 350 mM NaCl plus 10, 30, or 50 mM of imidazole, pH 8.0). The elution was carried out four times with 10 ml buffer (50 mM Tris–HCl, 350 mM NaCl, and 250 mM imidazole, pH 8.0). The SUMO tag was removed from the eluted protein during the dialysis in the buffer supplemented with 350 mM NaCl, 50 mM Tris–HCl, 0.5 mM TCEP–HCl (pH 8.0), and in-house produced N6His-SUMO protease (protease:protein ratio = 1:16) for 2 h at room temperature. The suspension was then transferred to the second nickel–nitrilotriacetic acid column, and the cleaved protein was eluted with 2× column volume of buffer containing 350 mM NaCl, 50 mM Tris–HCl, 0.5 mM TCEP–HCl, and 10 mM imidazole (pH 8.0). All purified proteins were analyzed on SDS-PAGE (15%, 37.5:1 acrylamide:bisacrylamide; BioRad) after Coomassie brilliant blue (G-250; Thermo Scientific) staining.

      Western blot analysis

      Total proteins were extracted from the yeast cells by the urea/SDS method according to the manufacturer's protocol (Clontech) without modifications. The extracted proteins were quantified by the Bradford assay (Bio-Rad), and 20 μg of each extract were loaded onto 15% SDS-PAGE gel (37.5:1 acrylamide/bisacrylamide) followed by transfer to a polyvinylidene fluoride membrane at 100 V for 75 min. Blocking was done with 5% fat-free milk in 1× Tris-buffered saline with Tween-20 (10 mM Tris–Cl [pH 8.0], 150 mM NaCl, and 0.01% Tween-20) at 4 °C overnight. Membranes were probed with GAL4 AD monoclonal antibody in a dilution of 1:2500 (630402; Clontech) at room temperature for 1 h and rinsed three times with 5% fat-free milk in 1× Tris-buffered saline with Tween-20 at room temperature, each for 10 min. For the secondary antibody, we used a 1:10,000 dilution of a antimouse antibody coupled to IRDye 800CW (LI-COR Biosciences), which was added to the blot and washed for 1 h after incubation, as described for the primary antibody. Western blots were visualized using a Sapphire Biomolecular imager (Azure Biosystems).

      ALPHA

      Dimerization kinetics were determined by measuring Kd values using the ALPHA assay on a Synergy Neo2 Hybrid Multi-Mode Reader (BioTek) according to the manufacturer's PerkinElmer protocol. For the competition binding assay, 100 nM of N6His-RACT and GST-RACT and 100 nM of N6His-GL3ACT and GST-GL3ACT were combined with different concentrations of untagged RACT (0–8 μM) and GL3ACT (0–10 μM) and incubated for 2 h at room temperature. Saturation binding assays were performed by incubating different concentration of His-tagged RACT (0–4 μM) and GL3ACT (0–5 μM) proteins with 100 nM GST-tagged RACT or 500 nM GL3ACT. All protein mixtures were further incubated with Nickel Chelate Alpha Donor beads (20 μg/ml; PerkinElmer) and anti-GST AlphaLISA Acceptor beads (20 μg/ml; PerkinElmer) at room temperature for 2 h. A total of 40 μl of final mixtures were transferred into white 384-well OptiPlate (PerkinElmer), and the signal was read in the Alpha-compatible reader (Biotek). Dissociation constants obtained from both assays were calculated by fitting with one site fit model in GraphPad Prism, version 6.0 (GraphPad Software, Inc).

      Protein structure prediction

      Protein monomer secondary structures of GL3552–637 and R525–610 were predicted using I-TASSER (
      • Yang J.
      • Yan R.
      • Roy A.
      • Xu D.
      • Poisson J.
      • Zhang Y.
      The I-TASSER suite: Protein structure and function prediction.
      ). The homodimer structure was estimated by GalaxyHomomer (
      • Baek M.
      • Park T.
      • Heo L.
      • Park C.
      • Seok C.
      GalaxyHomomer: A web server for protein homo-oligomer structure prediction from a monomer sequence or structure.
      ) and GalaxyRefineComplex (
      • Heo L.
      • Lee H.
      • Seok C.
      GalaxyRefineComplex: Refinement of protein-protein complex model structures driven by interface repacking.
      ). The secondary structures of R and GL3 were obtained based on the ACT domain of PheH (PDB: 5FII).

      Identification of bHLH TFs containing ACT-like domains in maize and phylogenetic reconstructions

      The sequences for all maize bHLH TFs were retrieved from GRASSIUS (
      • Yilmaz A.
      • Nishiyama M.Y.
      • Garcia-Fuentes B.
      • Souza G.M.
      • Janies D.
      • Gray J.
      • Grotewold E.
      GRASSIUS: A platform for comparative regulatory genomics across the grasses.
      ). The maize bHLH TFs were aligned with Arabidopsis bHLH TFs by MUSCLE (
      • Edgar R.C.
      Muscle: a multiple sequence alignment method with reduced time and space complexity.
      ) to determine subgroup. The phylogenetic tree including all bHLH TFs of Arabidopsis and maize was constructed by maximum likelihood method with 1000 bootstrap replicates in MEGA7 (
      • Kumar S.
      • Stecher G.
      • Tamura K.
      MEGA7: Molecular evolutionary genetics analysis version 7.0 for bigger datasets.
      ).
      The bHLH TFs containing ACT-like domain were selected on the basis of the presence of ββαββα by PSIPRED (
      • Buchan D.W.A.
      • Minneci F.
      • Nugent T.C.O.
      • Bryson K.
      • Jones D.T.
      Scalable web services for the PSIPRED protein analysis workbench.
      ). The amino acid sequences and secondary structures were aligned to each other by T-Coffee (http://tcoffee.crg.cat/). The amino acid variation was determined by WebLogo, version 3 (https://weblogo.berkeley.edu/logo.cgi). The sequences used in this study are provided in Table S1.
      GL3- and TT8-like proteins in 15 different species were isolated from Pytozome (version 12.1; https://phytozome.jgi.doe.gov/pz/portal.html) and Phylogenes (http://www.phylogenes.org/). The sequences were aligned with MUSCLE, and the phylogenetic tree was built by the neighbor joining method with 1000 bootstraps in MEGA7 (
      • Kumar S.
      • Stecher G.
      • Tamura K.
      MEGA7: Molecular evolutionary genetics analysis version 7.0 for bigger datasets.
      ). Secondary structures of the proteins were evaluated as described previously.

      Data availability

      All data are contained within the article.

      Supporting information

      This article contains supporting information (
      • Kong Q.
      • Pattanaik S.
      • Feller A.
      • Werkman J.R.
      • Chai C.
      • Wang Y.
      • Grotewold E.
      • Yuan L.
      Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
      ,
      • James P.
      • Halladay J.
      • Craig E.A.
      Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast.
      ).

      Conflict of interest

      The authors declare that they have no conflicts of interest with the contents of this article.

      Acknowledgments

      We acknowledge the assistance of Dr Lim Heo in the Feig laboratory for assistance with modeling and advice. We thank members of the Grotewold laboratory, Kaillathe (Pappan) Padmanabhan, and Hervé Bègue for assistance and helpful discussions. This research was supported by National Science Foundation grants MCB-1513807 and MCB-1822343 .

      Author contributions

      E. G. and Y. S. L. conceived the concept and designed the experiments. Y. S. L. performed most of the experiments with assistance from A. H.-T. and J. S. Y. S. L. and E. G. wrote the article with valuable comments provided by A. H.-T., J. S., and J. H. G. All authors read and approved the final article.

      References

        • Brkljacic J.
        • Grotewold E.
        Combinatorial control of plant gene expression.
        Biochim. Biophys. Acta Gene Regul. Mech. 2017; 1860: 31-40
        • Lambert S.A.
        • Jolma A.
        • Campitelli L.F.
        • Das P.K.
        • Yin Y.
        • Albu M.
        • Chen X.
        • Taipale J.
        • Hughes T.R.
        • Weirauch M.T.
        The human transcription factors.
        Cell. 2018; 172: 650-665
        • Feller A.
        • Machemer K.
        • Braun E.L.
        • Grotewold E.
        Evolutionary and comparative analysis of MYB and bHLH plant transcription factors.
        Plant J. 2011; 66: 94-116
        • Ferré-D'Amaré A.R.
        • Prendergast G.C.
        • Ziff E.B.
        • Burley S.K.
        Recognition by Max of its cognate DNA through a dimeric b/HLH/Z domain.
        Nature. 1993; 363: 38-45
        • Ma P.C.M.
        • Rould M.A.
        • Weintraub H.
        • Pabo C.O.
        Crystal structure of MyoD bHLH domain-DNA complex: Perspectives on DNA recognition and implications for transcriptional activation.
        Cell. 1994; 77: 451-459
        • Massari M.E.
        • Murre C.
        Helix-loop-helix proteins: Regulators of transcription in eucaryotic organisms.
        Mol. Cell. Biol. 2000; 20: 429-440
        • Pires N.
        • Dolan L.
        Origin and diversification of basic-helix-loop-helix proteins in plants.
        Mol. Biol. Evol. 2010; 27: 862-874
        • Stevens J.D.
        • Roalson E.H.
        • Skinner M.K.
        Phylogenetic and expression analysis of the basic helix-loop-helix transcription factor gene family: Genomic approach to cellular differentiation.
        Differentiation. 2008; 76: 1006-1022
        • Blackwood E.M.
        • Eisenman R.N.
        Max: A helix-loop-helix zipper protein that forms a sequence-specific DNA-binding compelx with Myc.
        Science. 1991; 251: 1211-1217
        • Bresnick E.H.
        • Felsenfeld G.
        The leucine zipper is necessary for stabilizing a dimer of the helix-loop-helix transcription factor USF but not for maintenance of an elongated conformation.
        J. Biol. Chem. 1994; 269: 21110-21116
        • Shively C.A.
        • Liu J.
        • Chen X.
        • Loell K.
        • Mitra R.D.
        Homotypic cooperativity and collective binding are determinants of bHLH specificity and function.
        Proc. Natl. Acad. Sci. U. S. A. 2019; 116: 16143-16152
        • Aravind L.
        • Koonin E.V.
        Eukaryotic transcription regulators derive from ancient enzymatic domains.
        Curr. Biol. 1998; 8: R111-R113
        • Feller A.
        • Hernandez J.M.
        • Grotewold E.
        An ACT-like domain participates in the dimerization of several plant bHLH transcription factors.
        J. Biol. Chem. 2006; 281: 28964-28974
        • Feller A.
        • Yuan L.
        • Grotewold E.
        The BIF domain in plant bHLH proteins is an ACT-like domain.
        Plant Cell. 2017; 29: 1800-1802
        • Chang F.
        • Cui J.
        • Wang L.
        • Ma H.
        REPLY: The BIF domain is structurally and functionally distinct from other types of ACT-like domains.
        Plant Cell. 2017; 29: 1803-1805
        • Cui J.
        • You C.
        • Zhu E.
        • Huang Q.
        • Ma H.
        • Chang F.
        Feedback regulation of DYT1 by interactions with downstream bHLH factors promotes DYT1 nuclear localization and anther development.
        Plant Cell. 2016; 28: 1078-1093
        • Cho Y.
        • Sharma V.
        • Sacchettini J.C.
        Crystal structure of ATP phosphoribosyltransferase from Mycobacterium tuberculosis.
        J. Biol. Chem. 2003; 278: 8333-8339
        • Grant G.A.
        The ACT domain: A small molecule binding domain and its role as a common regulatory element.
        J. Biol. Chem. 2006; 281: 33825-33829
        • Curien G.
        • Biou V.
        • Mas-Droux C.
        • Robert-Genthon M.
        • Ferrer J.L.
        • Dumas R.
        Amino acid biosynthesis: New architectures in allosteric enzymes.
        Plant Physiol. Biochem. 2008; 46: 325-339
        • Gai Z.
        • Wang Q.
        • Yang C.
        • Wang L.
        • Deng W.
        • Wu G.
        Structural mechanism for the arginine sensing and regulation of CASTOR1 in the mTORC1 signaling pathway.
        Cell Discov. 2016; 2: 16051
        • Xia J.
        • Wang R.
        • Zhang T.
        • Ding J.
        Structural insight into the arginine-binding specificity of CASTOR1 in amino acid-dependent mTORC1 signaling.
        Cell Discov. 2016; 2: 16035
        • Saxton R.A.
        • Chantranupong L.
        • Knockenhauer K.E.
        • Schwartz T.U.
        • Sabatini D.M.
        Mechanism of arginine sensing by CASTOR1 upstream of mTORC1.
        Nature. 2016; 536: 229-233
        • Ludwig R.
        • Habera L.F.
        • Dellaporta S.L.
        • Wessler S.R.
        Lc, a member of the maize R gene family responsible for tissue-specific anthocyanin production, encodes a protein similar to transcriptional activators and contains the myc-homology region.
        Proc. Natl. Acad. Sci. U. S. A. 1989; 86: 7092-7096
        • Goff S.A.
        • Cone K.C.
        • Chandler V.L.
        Functional analysis of the transcriptional activator encoded by the maize B gene: Evidence for a direct functional interaction between two classes of regulatory proteins.
        Genes Dev. 1992; 6: 864-875
        • Grotewold E.
        • Sainz M.B.
        • Tagliani L.
        • Hernandez J.M.
        • Bowen B.
        • Chandler V.L.
        Identification of the residues in the Myb domain of maize C1 that specify the interaction with the bHLH cofactor R.
        Proc. Natl. Acad. Sci. U. S. A. 2000; 97: 13579-13584
        • Kong Q.
        • Pattanaik S.
        • Feller A.
        • Werkman J.R.
        • Chai C.
        • Wang Y.
        • Grotewold E.
        • Yuan L.
        Regulatory switch enforced by basic helix-loop-helix and ACT-domain mediated dimerizations of the maize transcription factor R.
        Proc. Natl. Acad. Sci. U. S. A. 2012; 109: E2091-2097
        • Bernhardt C.
        • Lee M.M.
        • Gonzalez A.
        • Zhang F.
        • Lloyd A.
        • Schiefelbein J.
        The bHLH genes GLABRA3 (GL3) and ENHANCER OF GLABRA3 (EGL3) specify epidermal cell fate in the Arabidopsis root.
        Development. 2003; 130: 6431-6439
        • Bernhardt C.
        • Zhao M.
        • Gonzalez A.
        • Lloyd A.
        • Schiefelbein J.
        The bHLH genes GL3 and EGL3 participate in an intercellular regulatory circuit that controls cell patterning in the Arabidopsis root epidermis.
        Development. 2005; 132: 291-298
        • Morohashi K.
        • Grotewold E.
        A systems approach reveals regulatory circuitry for Arabidopsis trichome initiation by the GL3 and GL1 selectors.
        PLoS Genet. 2009; 5e1000396
        • Morohashi K.
        • Zhao M.
        • Yang M.
        • Read B.
        • Lloyd A.
        • Lamb R.
        • Grotewold E.
        Participation of the Arabidopsis bHLH factor GL3 in trichome initiation regulatory events.
        Plant Physiol. 2007; 145: 736-746
        • Payne C.
        • Zhang F.
        • Lloyd A.
        GL3 encodes a bHLH protein that regulate trichome development in Arabidopsis through interaction with GL1 and TTG1.
        Genetics. 2000; 156: 1349-1362
        • Gonzalez A.
        • Zhao M.
        • Leavitt J.M.
        • Lloyd A.M.
        Regulation of the anthocyanin biosynthetic pathway by the TTG1/bHLH/Myb transcriptional complex in Arabidopsis seedlings.
        Plant J. 2008; 53: 814-827
        • Lloyd A.
        • Brockman A.
        • Aguirre L.
        • Campbell A.
        • Bean A.
        • Cantero A.
        • Gonzalez A.
        Advances in the MYB–bHLH–WD repeat (MBW) pigment regulatory model: Addition of a WRKY factor and co-option of an Anthocyanin MYB for betalain regulation.
        Plant Cell Physiol. 2017; 58: 1431-1441
        • Xu W.
        • Dubos C.
        • Lepiniec L.
        Transcriptional control of flavonoid biosynthesis by MYB–bHLH–WDR complexes.
        Trends Plant Sci. 2015; 20: 176-185
        • Patra B.
        • Pattanaik S.
        • Yuan L.
        Ubiquitin protein ligase 3 mediates the proteasomal degradation of GLABROUS 3 and ENHANCER OF GLABROUS 3, regulators of trichome development and flavonoid biosynthesis in Arabidopsis.
        Plant J. 2013; 74: 435-447
        • Wen J.
        • Li Y.
        • Qi T.
        • Gao H.
        • Liu B.
        • Zhang M.
        • Huang H.
        • Song S.
        The C-terminal domains of Arabidopsis GL3/EGL3/TT8 interact with JAZ proteins and mediate dimeric interactions.
        Plant Signal. Behav. 2018; 13: 1795-1814
        • James P.
        • Halladay J.
        • Craig E.A.
        Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast.
        Genetics. 1996; 144: 1425-1436
        • Baek M.
        • Park T.
        • Heo L.
        • Park C.
        • Seok C.
        GalaxyHomomer: A web server for protein homo-oligomer structure prediction from a monomer sequence or structure.
        Nucleic Acids Res. 2017; 45: W320-W324
        • Heo L.
        • Lee H.
        • Seok C.
        GalaxyRefineComplex: Refinement of protein-protein complex model structures driven by interface repacking.
        Sci. Rep. 2016; 6: 32153
        • Schuller D.J.
        • Grant G.A.
        • Banaszak L.J.
        The allosteric ligand site in the Vmax-type cooperative enzyme phosphoglycerate dehydrogenase.
        Nat. Struct. Biol. 1995; 2: 69-76
        • Patel D.
        • Kopec J.
        • Fitzpatrick F.
        • McCorvie T.J.
        • Yue W.W.
        Structural basis for ligand-dependent dimerization of phenylalanine hydroxylase regulatory domain.
        Sci. Rep. 2016; 6: 23748
        • Zhang S.
        • Huang T.
        • Ilangovan U.
        • Hinck A.P.
        • Fitzpatrick P.F.
        The solution structure of the regulatory domain of tyrosine hydroxylase.
        J. Mol. Biol. 2014; 426: 1483-1497
        • Yilmaz A.
        • Nishiyama M.Y.
        • Garcia-Fuentes B.
        • Souza G.M.
        • Janies D.
        • Gray J.
        • Grotewold E.
        GRASSIUS: A platform for comparative regulatory genomics across the grasses.
        Plant Physiol. 2009; 149: 171-180
        • Buchan D.W.A.
        • Minneci F.
        • Nugent T.C.O.
        • Bryson K.
        • Jones D.T.
        Scalable web services for the PSIPRED protein analysis workbench.
        Nucleic Acids Res. 2013; 41: W349-W357
        • Carluccio C.
        • Fraternali F.
        • Salvatore F.
        • Fornili A.
        • Zagari A.
        Structural features of the regulatory ACT domain of phenylalanine hydroxylase.
        PLoS One. 2013; 8e79482
        • Rao V.S.
        • Srinivas K.
        • Sujini G.N.
        • Kumar G.N.
        Protein-protein interaction detection: Methods and analysis.
        Int. J. Proteomics. 2014; 2014147648
        • Sharma V.
        • Ranjan T.
        • Kumar P.
        • Pal A.K.
        • Jha V.K.
        • Sahni S.
        • Prasad B.D.
        Protein–protein interaction detection: Methods and analysis.
        Plant Biotechnol. 2017; 1: 391-411
        • Phizicky E.M.
        • Fields S.
        Protein-protein interactions: Methods for detection and analysis.
        Micro. Rev. 1995; 59: 94-123
        • Stynen B.
        • Tournu H.
        • Tavernier J.
        • Van Dijck P.
        Diversity in genetic in vivo methods for protein-protein interaction studies: From the yeast two-hybrid system to the mammalian split-luciferase system.
        Microbiol. Mol. Biol. Rev. 2012; 76: 331-382
        • Ou B.
        • Yin K.Q.
        • Liu S.N.
        • Yang Y.
        • Gu T.
        • Wing Hui J.M.
        • Zhang L.
        • Miao J.
        • Kondou Y.
        • Matsui M.
        • Gu H.Y.
        • Qu L.J.
        A high-throughput screening system for Arabidopsis transcription factors and its application to Med25-dependent transcriptional regulation.
        Mol. Plant. 2011; 4: 546-555
        • Burdo B.
        • Gray J.
        • Goetting-Minesky M.P.
        • Wittler B.
        • Hunt M.
        • Li T.
        • Velliquette D.
        • Thomas J.
        • Gentzel I.
        • Brito M.d.S.
        The maize TFome–development of a transcription factor open reading frame collection for functional genomics.
        Plant J. 2014; 80: 356-366
        • Machemer K.
        • Shaiman O.
        • Salts Y.
        • Shabtai S.
        • Sobolev I.
        • Belausov E.
        • Grotewold E.
        • Barg R.
        Interplay of MYB factors in differential cell expansion, and consequences for tomato fruit development.
        Plant J. 2011; 68: 337-350
        • Mossessova E.
        • Lima C.D.
        Ulp1-SUMO crystal structure and genetic analysis reveal conserved interactions and a regulatory element essential for cell growth in yeast.
        Mol. Cell. 2000; 5: 865-876
        • Gietz R.D.
        • Schiestl R.H.
        Quick and easy yeast transformation using the LiAc/SS carrier DNA/PEG method.
        Nat. Protoc. 2007; 2: 35-37
        • Rose M.
        • Botstein D.
        Construction and use of gene fusions to lacZ (β-galactosidase) that are expressed in yeast.
        Methods Enzymol. 1983; 101: 167-180
        • Bradford M.M.
        A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding.
        Anal. Biochem. 1976; 72: 248-254
        • Yang J.
        • Yan R.
        • Roy A.
        • Xu D.
        • Poisson J.
        • Zhang Y.
        The I-TASSER suite: Protein structure and function prediction.
        Nat. Methods. 2015; 12: 7
        • Edgar R.C.
        Muscle: a multiple sequence alignment method with reduced time and space complexity.
        BMC Bioinformatics. 2004; 5: 1-19
        • Kumar S.
        • Stecher G.
        • Tamura K.
        MEGA7: Molecular evolutionary genetics analysis version 7.0 for bigger datasets.
        Mol. Biol. Evol. 2016; 33
        • Hall T.A.
        BioEdit: A user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT.
        Nucleic Acids Symp. 1999; 41: 95-98
        • Yasgar A.
        • Jadhav A.
        • Simeonov A.
        • Coussens N.P.
        AlphaScreen-based assays: Ultra-high-throughput screening for small-molecule inhibitors of challenging enzymes and protein-protein interactions.
        Methods Mol. Biol. 2016; 1439: 77-98