Senescence-associated Barley NAC (NAM, ATAF1,2, CUC) Transcription Factor Interacts with Radical-induced Cell Death 1 through a Disordered Regulatory Domain*

Background: Plant NAC transcription factors (TF) are important regulators of senescence. Results: Senescence-associated barley HvNAC013 uses intrinsic disorder in transcriptional activation and interactions. Conclusion: Radical induced cell death 1 exploits the intrinsic disorder of HvNAC013 and other TFs for interactions without structure induction in HvNAC013. Significance: This first structural characterization of NAC intrinsic disorder may reveal general features of important TF regulatory interactions. Senescence in plants involves massive nutrient relocation and age-related cell death. Characterization of the molecular components, such as transcription factors (TFs), involved in these processes is required to understand senescence. We found that HvNAC005 and HvNAC013 of the plant-specific NAC (NAM, ATAF1,2, CUC) TF family are up-regulated during senescence in barley (Hordeum vulgare). Both HvNAC005 and HvNAC013 bound the conserved NAC DNA target sequence. Computational and biophysical analyses showed that both proteins are intrinsically disordered in their large C-terminal domains, which are transcription regulatory domains (TRDs) in many NAC TFs. Using motif searches and interaction studies in yeast we identified an evolutionarily conserved sequence, the LP motif, in the TRD of HvNAC013. This motif was sufficient for transcriptional activity. In contrast, HvNAC005 did not function as a transcriptional activator suggesting that an involvement of HvNAC013 and HvNAC005 in senescence will be different. HvNAC013 interacted with barley radical-induced cell death 1 (RCD1) via the very C-terminal part of its TRD, outside of the region containing the LP motif. No significant secondary structure was induced in the HvNAC013 TRD upon interaction with RCD1. RCD1 also interacted with regions dominated by intrinsic disorder in TFs of the MYB and basic helix-loop-helix families. We propose that RCD1 is a regulatory protein capable of interacting with many different TFs by exploiting their intrinsic disorder. In addition, we present the first structural characterization of NAC C-terminal domains and relate intrinsic disorder and sequence motifs to activity and protein-protein interactions.


Senescence in plants involves massive nutrient relocation and age-related cell death. Characterization of the molecular components, such as transcription factors (TFs), involved in these processes is required to understand senescence. We found that
HvNAC005 and HvNAC013 of the plant-specific NAC (NAM, ATAF1,2, CUC) TF family are up-regulated during senescence in barley (Hordeum vulgare). Both HvNAC005 and HvNAC013 bound the conserved NAC DNA target sequence. Computational and biophysical analyses showed that both proteins are intrinsically disordered in their large C-terminal domains, which are transcription regulatory domains (TRDs) in many NAC TFs. Using motif searches and interaction studies in yeast we identified an evolutionarily conserved sequence, the LP motif, in the TRD of HvNAC013. This motif was sufficient for transcriptional activity. In contrast, HvNAC005 did not function as a transcriptional activator suggesting that an involvement of HvNAC013 and HvNAC005 in senescence will be different. HvNAC013 interacted with barley radical-induced cell death 1 (RCD1) via the very C-terminal part of its TRD, outside of the region containing the LP motif. No significant secondary structure was induced in the HvNAC013 TRD upon interaction with RCD1. RCD1 also interacted with regions dominated by intrinsic disorder in TFs of the MYB and basic helix-loop-helix families. We propose that RCD1 is a regulatory protein capable of interacting with many different TFs by exploiting their intrinsic disorder. In addition, we present the first structural characterization of NAC C-terminal domains and relate intrinsic disorder and sequence motifs to activity and protein-protein interactions.
Transcription in eukaryotes is complex and involves a large number of transcription factors (TFs). 2 Gene-specific TFs bind to regulatory regions of target genes and thereby enhance or impede the progression of RNA polymerase. The DNA-binding domain (DBD) is the defining feature of TFs and is used for gene family assignment (1). TFs also contain a transcription regulatory domain (TRD). TRDs do not have well defined sequence motifs and structures and are often classified according to regions enriched in specific amino acids, e.g. glutamine (2). These characteristics have hampered prediction of protein interacting regions in TRDs (3).
TRDs are often characterized by intrinsic disorder (ID), referring to an ensemble of protein conformers without a single stable tertiary structure and with fluctuating secondary structures of low population (4). The function of disordered regions in TFs relates to molecular recognition, which is often, but not always, linked to folding upon binding (4). The flexibility associated with ID allows TFs to interact efficiently with a number of different target proteins (5). Binding by regions with ID can be mediated by molecular recognition features (MoRFs), which are short sequence stretches with propensities to undergo disorder to order transition upon binding (6). Binding regions can also be predicted from linear motifs (LMs), referring to sequence patterns in proteins binding a common target (7). Although MoRFs are not associated with a specific sequence motif an overlap often exists between a LM and a MoRF. LMs typically have a few specificity determining residues favoring structural order in a highly flexible carrier region (7). Recently, we predicted that members of the NAC (NAM (no apical meristem), ATAF, CUC (cup-shaped cotyledon)) TF family have largely disordered TRDs with varying sequences containing numerous subgroup-specific sequence motifs (8) making further characterization of great relevance to both TF function and ID.
The plant-specific NAC TFs are involved in controlling many aspects of plant life (9 -12). They are characterized by a conserved N-terminal DBD, the NAC domain, with a unique mainly ␤-sheet fold (13). In vitro DNA binding studies showed that several NAC proteins bind the NAC DNA-binding site (NACBS) with the core sequence CGT(GA) (8,14). NACBSs have been identified for example, in the ANAC072 target promoter of ERD1 (early responsive to dehydration stress 1) (11) and in the NTL6 target promoters of pathogenesis-related (PR) genes (12). NAC TFs also contain a TRD that can function as an independent protein module in stimulating transcriptional activation. Thus, overexpression of the Arabidopsis thaliana ANAC019 TRD mimicked hypersensitivity by the plant stress hormone abscisic acid (8).
Protein-protein interactions are also essential for proper regulation by TFs. NAC proteins can homo-and heterodimerize through a conserved surface in the NAC domain (8,13), which is essential for DNA binding in vitro (14). Several reports have demonstrated interactions between a NAC TF and a virus protein, in a manner dependent on the structural integrity of the NAC domain (10,15). The NAC domain was also responsible for regulatory interactions between NAC proteins and E3 protein ubiquitin ligases (16,17), and additional interactions of NAC proteins with unrelated proteins have been identified (10,18) suggestive of a broad functional and structural repertoire of NAC-protein interactions. Recently, RCD1 (RADICAL-IN-DUCED CELL DEATH 1), a regulator of stress and development in Arabidopsis (19), was shown to interact with a large number of TFs belonging to different TF families, including NAC (20). This may suggest a common feature characterizing interactions between RCD1 and these structurally different TFs.
NAC TFs are also involved in senescence. Thus, the T-DNA knock-out mutation of Arabidopsis NAP significantly delayed leaf senescence, and overexpression caused precocious senescence (21). The ore1 (oresara1) mutant was identified by its delayed leaf senescence phenotype (22), and later ORE1 was implicated in ethylene-dependent regulation of senescence (23). Research of crop species has also contributed significantly to the understanding of NAC TF function in senescence. RNA interference studies of wheat Gpc-B1, encoding the NAC TF NAM-B1, resulted in delayed flag leaf senescence and significant reduction in grain zinc, iron, and protein content (24). This was associated with increased residual nitrogen, zinc, and iron in the flag leaf, thereby demonstrating a role for NAC genes in nutrient redistribution to the developing grain during leaf senescence (24). We have provided an inventory of 48 barley NAC genes, of which several transcripts accumulated in senescing tissue (25).
To improve understanding of the NAC TF-associated regulatory mechanisms of senescence we present extensive insights in structure-function relationships of two senescence-induced NAC genes, HvNAC005 and HvNAC013, from barley. The NAC domain of both proteins bound the NACBS, but only HvNAC013 activated transcription. To understand this at a molecular level we present the first structural characterization of a NAC TRD. The C-terminal domains of both proteins were largely intrinsically disordered. However, only the domain of HvNAC013 interacted with HvRCD1. Other types of TFs also interacted with RCD1 through regions with ID. We discuss the implications of ID in NAC and other TFs for regulatory interactions with RCD1 and different functionality of modularly similar NAC TFs involved in senescence.
Bioinformatics Tools-ID was predicted using PONDR VL3 or Disopred2 (26,27). MEME (multiple EM (expectation maximus) for motif elicitation) was used to identify sequence motifs. Multiple alignments were made by Clustal X (28), and BLAST at the National Center for Biotechnology Information was used to analyze sequences.
cDNA Synthesis and Quantitative Real-time PCR (qRT-PCR)-RNA was isolated from ϳ100 g of frozen homogenized material from Golden Promise cultivar using the Spectrum TM Plant Total RNA kit (Sigma) and recommendations. First strand cDNA synthesis was performed as described (25). qRT-PCR was performed using the ABI Prism 7900HT Sequence Detection System with the Power SYBR Green PCR master mix (Applied Biosystems). HvNAC005, HvNAC013, and HvRCD1 were cloned with primers flanking the coding regions (supplemental Table S1). The cDNA template used in the PCR was synthesized from senescent or green leaf tissue.
Assays in Yeast-Bait and prey fragments were fused to GAL4 DBD or transactivation domain in pGBKT7/pGADT7 (Clontech) (supplemental Table S1). HvNAC013 mutants were constructed using the QuikChange mutagenesis kit (Stratagene). Plasmids were transformed into yeast strain pJ694A and assayed as described (8). Some plates contained 3-amino-1,2,4triazole as indicated to inhibit self-activation. ␤-Galatosidase activity was measured using a liquid culture assay and orthonitrophenyl-␤-galactoside substrate. Fragments of AtRCD1, amplified from cDNA from the Arabidopsis Biological Resource Center, and bHLH11 or MYB91, amplified from REGIA TF cDNA, were recombined into pDEST32/pDEST22 (Invitrogen). Protein extracted as described (16) was detected by Western blotting using GAL4-DBD monoclonal antibodies (Clontech) or anti-tubulin antibodies as loading control.
In Vitro Pulldown Assay-Sonicated bacterial lysate (1 or 0.2 ml) with expressed GST-HvRCD1(485-579) or GST was immobilized on 50 l of glutathione-Sepharose 4B with 100 g of His 6 -NAC. The complex was eluted with 50 l of Tris-HCl, pH 8.0, 10 mM reduced glutathione. Interaction was analyzed by Western blotting using anti-His 6 (Qiagen) or anti-GST (GE Healthcare) antibodies.
In Vitro Interaction Studies-HvNAC013 and HvRCD1 interaction were analyzed by intrinsic fluorescence spectroscopy using a PerkinElmer LS50B and 3 M protein in 10 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 7.0. Excitation was at 280 or 296 nm, RT, averaging five scans, and subtracting buffer backgrounds. Theoretical emission spectra for noninteracting proteins were generated by adding spectra of individual proteins. A similar series of CD spectra of proteins alone and in complex was recorded using the same settings.

Expression of Barley HvNAC005 and HvNAC013
Is Induced by Senescence-Senescence in plants involves nutrient relocation and age-related cell death (24,32), and identification and characterization of molecular components of these processes are required to understand senescence. Recently, barley HvNAC005 and HvNAC013 were identified as genes up-regulated during senescence (25). To investigate their expression pattern in more detail, a time course experiment was performed. Three stages of flag leaf development were investigated, starting with young green leaves and ending with fully senescent, although still turgid, leaves ( Fig. 1). We included two marker genes of senescence, encoding a papain-like cysteine peptidase (AM941122) and the Rubisco small subunit (ABK79421), respectively (32). This showed that the expression levels of HvNAC005 and HvNAC013 were up-regulated as senescence progressed. Interestingly, HvNAC005 and HvNAC013 were expressed at staggered intervals. The expression of HvNAC005 was increased more strongly than that of HvNAC013 and showed an 8-fold increase 15 days after ear emergence. HvNAC005 and HvNAC013 were induced ϳ20and 13-fold, respectively, in fully senescent leaves compared with nonsenescent leaves.
HvNAC005 and HvNAC013 Belong to Phylogenetically Different Subgroups-To investigate the phylogenetic clustering of HvNAC005 and HvNAC013, an unrooted tree of predicted barley NAC proteins (25) was constructed based on their conserved NAC domains (Fig. 2). This showed that HvNAC005 and HvNAC013 belong to distant subgroups. To further access subgroup specificity, Arabidopsis, wheat, and rice NAC proteins implicated in senescence (21,22,24,(33)(34)(35) were included in the tree. This together with BLAST searches revealed that HvNAC005 is closely related to NAP from Arabidopsis (21). HvNAC013 belongs to a group, which in addition to its closest Arabidopsis homologue, ANAC046, contains ORE1 and the CUC and vascular-related NAC domain (VND) proteins (8, 36).

HvNAC005 and HvNAC013 Have Classic DNA-binding NAC Domains and C-terminal Domains with Intrinsic Disorder-
The HvNAC013 and HvNAC005 cDNAs were obtained from senescent barley leaves and shown to encode proteins of 346 and 355 amino acid residues, respectively (supplemental Fig.  S1). The NAC domain of both HvNAC005 and HvNAC013 showed a high degree of sequence similarity to NAC domains from typical NAC proteins (supplemental Fig. S1) (8). In a previous study, ANAC019, representing the NAC subgroup of HvNAC005, and NAC2/ORE, representing the NAC subgroup of HvNAC013, were used as probes to select a NACBS containing CGT[GA] as core sequence (14). A palindrome version of NACBS (palNACBS), which was used previously to examine NAC DNA binding (8,14), was used to examine the in vitro DNA-binding properties of HvNAC005 and HvNAC013 (Fig.  3A). Purified GST-tagged recombinant versions of the NAC domains, GST-HvNAC005(1-172) and GST-HvNAC013(1-176), were used to examine the abilities of HvNAC005 and HvNAC013 to bind palNACBS in titration series using from 10 to 500 ng of protein. Both HvNAC005 and HvNAC013 bound palNACBS, in contrast to GST. For HvNAC013, binding could be detected using 10 ng of protein (Fig. 3C), whereas HvNAC005 showed weaker affinity with binding observed using 100 ng of protein (Fig. 3B), which is similar to affinities observed for other NAC proteins (8). The contribution of specific bases and amino acid residues to the interaction between the NACBS and NAC TFs has not been characterized in molecular details. To ensure the importance of the NACBS for binding, EMSAs were, therefore, also performed using an oligonucleotide where the two NACBSs were replaced by simple nucleotide repeats (Fig. 3A) (14). No binding was detected with substituted palNACBS (Fig. 3, B and C). Fig. 3, D and E, shows that binding of both HvNAC005 and HvNAC013 to palNACBS was outcompeted by unlabeled palNACBS (lanes 3-5) but not by substituted palNACBS (lanes 6 -8).
We previously predicted the structurally divergent C-terminal domains of Arabidopsis NAC proteins to be intrinsically disordered (8). To investigate if HvNAC005 and HvNAC013 share this characteristic they were examined for ID using the predictor PONDR VL3 (26). This analysis suggested that HvNAC005 and HvNAC013 have a large degree of disorder within their C-terminal domains, although they show different distributions of disorder (Fig. 4). HvNAC013 displayed an ϳ100 amino acid residues stretch with ID within the C terminus (Fig. 4B). Conversely, HvNAC005 showed ID immediately following the NAC domain followed by a stretch of order before another stretch with ID (Fig. 4A).
HvNAC013 Is a Transcriptional Activator with Several Regions Contributing to Transcriptional Activity-Many NAC C termini are TRDs (8,37). These domains are diverse and contain subgroup-specific sequence motifs (8). The MEME motif search tool was used to identify motifs shared between HvNAC013 and its closest homologues from barley and rice (Fig. 2). Four motifs with conservation of both polar and nonpolar amino acid residues were identified (Fig. 5, A and B) and named either after motifs reported previously (LP motif) (37) or after prominent amino acid residues characteristic of the motif (DV, YF, RR). Whereas the LP motif is frequent in NAC TRDs (8), the other three motifs are specific for a small subgroup of crop NAC TFs. Similar analyses also revealed sequence motifs in HvNAC005 (supplemental Fig. S1).
Coding regions of HvNAC005 and HvNAC013 were fused to yeast GAL4 DBD (Fig. 5A) and analyzed for their ability to activate transcription and promote yeast growth in the absence of histidine and adenine. Murine tumor suppressor p53 (pVA3-1) and SV40 large antigen (pTD1-1) were included as positive controls for transformation and activity dependent selective growth (38). Full-length HvNAC013 was able to activate transcription, and the activation potential resided in the C terminus (residues 177-346), whereas the N-terminal NAC domain (residues 1-176) was unable to activate reporter genes (Fig. 5C). Surprisingly, neither truncated nor full-length HvNAC005 had transactivation activity in yeast (Fig. 5C).
C-terminal deletions were made to examine the importance for transcriptional activity of the identified motifs (Fig. 5A). The analysis showed that the region containing the LP motif is sufficient for transcriptional activity (Fig. 5C). Mutations were introduced in the LP motif to change the highly conserved leucine and aspartate, which could participate in proteinprotein interactions. Examination of HvNAC013(L212A) and HvNAC013(D217A) showed that both mutant proteins were active in the yeast growth assay (Fig. 5C).

JOURNAL OF BIOLOGICAL CHEMISTRY 35421
cations removing identified sequence motifs, and activity was completely abolished by removal of the LP motif region. Both L212A and D217A mutant versions resulted in a significant decrease in activity compared with wild type HvNAC013 (Fig.  5D). The level of expression of the GAL4 DBD fusion proteins was examined by Western blotting. This showed expression of all proteins with the expected molecular mass values and no obvious correlation between protein expression levels and activity; i.e. the variance in protein level did not explain the lack of activation activity of, e.g. HvNAC005 and HvNAC13(1-200), and the decreased level of activity of HvNAC013(L212A) and HvNAC013(D217A) (Fig. 5E). In conclusion, the TRD of HvNAC013 has a complex structure in which distinct sequence motifs or regions in concert contribute to activity. Only deletion of all motifs abolished activity.
The HvNAC005 and HvNAC013 C Termini Are Intrinsically Disordered but Can Adopt ␣-Helix Structure-To examine the structure of the C-terminal domains of HvNAC005 and HvNAC013 experimentally, these domains were produced as recombinant proteins with an N-terminal histidine (His 6 ) tag. The proteins, which were produced in E. coli BL21(DE3)pLysS, were purified by metal affinity chromatography using standard procedures (29). Although the His 6 tag had no significant influ-  ence on the structure as revealed by far-UV CD analysis (Figs. 6B and 7B), His 6 -HvNAC013(177-346) was cleaved using tobacco etch virus protease before biophysical analysis to produce HvNAC013(177-346), except in the experiment shown in Fig. 6B, where it was kept for comparison. However, it was impossible to produce sufficient amounts of untagged HvNAC005(173-355) for structure analysis. Therefore, the histidine-tagged version of HvNAC005(173-355) was used for all analyses. The mass values of His 6 -HvNAC005(173-355) and His 6 -HvNAC013(177-346) were verified by mass spectrometry to be 22.8 and 21.7 kDa, respectively (data not shown). Both His 6 -HvNAC005(173-355) and His 6 -HvNAC013(177-346) displayed slow migration during SDS-PAGE compared with standard globular proteins and had a migration rate corresponding to a molecular mass of ϳ30 kDa (Fig. 6, A and B), suggestive of ID (39). The structure of both proteins was also analyzed by CD. The far-UV CD spectra showed a strong negative absorbance at 200 nm for His 6 -HvNAC005(173-355), His 6 -HvNAC013(177-346), and HvNAC013(177-346) with little ellipticity around 220 nm indicating lack of a significant amount of secondary structure and highly characteristic of unstructured proteins (Figs. 6, A and B, and 7B). Slightly negative ellipticities were observed at 222 nm, which is characteristic for the ␣-helical structure suggesting in all cases a maximum content of helical structure below 20% (40). The existence of a dimeric state, which could contribute to slow migration by SDS-PAGE was excluded as the titration series showed linear dependence of the signal on concentration in far-UV spectra (data not shown).
Different protein conformational classes (native, molten globule, pre-molten globule, and denaturant-unfolded) have characteristic hydrodynamic dimensions and molecular weight correlations (41). This aspect of HvNAC013(177-346) was analyzed using size exclusion chromatography. The calculated molecular mass of HvNAC013(177-346) is 18.0 kDa, corresponding to a Stokes radius of 20.5 Å for a globular protein. However, HvNAC013(177-346) eluted in a volume corresponding to a molecular mass of 41.5 kDa and a Stokes radius of 28.01 Å (Fig. 6C). This suggested that HvNAC013(177-346) exists in a pre-molten globule-like state (Fig. 6C). NMR experiments using 15 N-labeled HvNAC013(177-346) were also performed to examine for both the secondary structure and oligomeric state of HvNAC013(177-346). The HSQC spectrum WebLogo indicates its relative frequency at the given position (x axis). C, yeast transactivation assay. Fusion proteins of GAL4 DBD and HvNAC013 or HvNAC005 fragments were expressed in yeast and screened for their transactivation activity on the HIS3 and ADE2 reporter genes. Empty pGBKT7 served as negative control and pVA3-1 and pTD1-1 as positive controls. D, quantitative ␤-galactosidase assays. The cells were assayed for ␤-galactosidase activity using ortho-nitrophenyl-␤-galactoside as substrate. Activity was mea- showed a clustering of signals with low dispersion in the proton dimension (Fig. 6D) suggesting that HvNAC013(177-346) is highly unfolded. Furthermore, NMR experiments in the interval from 20 to 200 M were identical and confirmed that the C-terminal domain of HvNAC013 exists predominantly as a monomer in solution (data not shown).
Structure formation in proteins with ID is typically studied as a function of solvent conditions (4), and TFE can be used to analyze secondary structure propensity (42). This additive was used to investigate if His 6 -HvNAC005(173-355) and HvNAC013(177-346) were prone to inducible structure changes using far-UV CD. For both proteins addition of TFE changed the global minimum from 200 nm, typical of an unfolded state, to 222 and 208 nm, characteristic of ␣-helix structure (Fig. 7, A and B). In both cases, the ␣-helix structure was adopted through intermediate structural stages. Whereas HvNAC013(177-346) changed the structure in the range of TFE concentrations used, His 6 -HvNAC005(173-355) reached equilibrium at a TFE concentration of 30% and did not gain additional structure by a further increase of the TFE concentration (Fig. 7, A and B).
Altogether, and in agreement with in silico analyses, the biophysical studies of the C-terminal domains of HvNAC013 and HvNAC005 showed that these are largely unfolded. The HvNAC013 TRD displayed a pre-molten globular state, yet both HvNAC005 and HvNAC013 C termini were able to adopt an inducible ␣-helix structure by addition of TFE.
HvNAC013 TRD Interacts with the Promiscuous Plant-specific RST Domain of HvRCD1-None of the characterized NAC protein interactions involves the NAC TRD. However, it was recently reported that the Arabidopsis protein RCD1, which interacted with TFs from many different families, also interacted with the closest Arabidopsis homolog of HvNAC013, ANAC046 (20). The mutant rcd1 allele in Arabidopsis caused radical-induced cellular superoxide accumulation and cell lesion spreading (43). The promiscuity of RCD1 suggested that this protein interacts with structurally flexible TF regions such as TRDs making dissection of the interactions of great functional and structural interest. It was, therefore, obvious to test if HvNAC005 and HvNAC013, and in particular their flexible C termini, showed affinity for promiscuous RCD1.
A BLAST search using AtRCD1 as query revealed an orthologue in barley, named HvRCD1. From expressed sequence tag data, HvRCD1 was predicted to consist of 547 amino acid residues. However, the cDNA sequence isolated from barley leaves encoded a 579-amino acid residue protein corresponding in structure to AtRCD1 (supplemental Fig. S2). The proteins contain an N-terminal WWE protein-protein interaction domain followed by a poly(ADP-ribose) polymerase domain, and a C-terminal plant-specific RCD1, SRO, TAF4 (RST) domain (Fig. 8A). The RST domain was identified from RCD1-like proteins and TAF4, a component of the transcription initiation factor complex TFIID (20). The global sequence similarity between HvRCD1 and AtRCD1 is 34% and an even higher degree of similarity was found for the conserved domains.
A directed yeast two-hybrid assay using full-length HvNAC013 and HvRCD1 as bait and prey, respectively, suggested that the proteins interacted (Fig. 8B). Similar results were obtained when HvRCD1 acted as bait and HvNAC013 as prey (Fig. 8C). Analysis of the ability of the isolated domains of HvNAC013 to interact with HvRCD1 showed that the interaction was mediated by the TRD of HvNAC013 (Fig. 8D). It was previously suggested that the RST domain was involved in interactions between RCD1 and target TFs (20). Therefore, the TRD of HvNAC013 and the RST domain of HvRCD1 (residues 485-579) were analyzed for the ability to interact. This revealed that the interaction was mediated by the unstructured TRD of HvNAC013 and the RST domain of RCD1 (Fig. 8D).
Recombinant versions of the HvNAC013 TRD and HvRCD1 RST domains, His 6 -HvNAC013(177-346) and GST-HvRCD1(485-579), were produced and used for in vitro pulldown assays (Fig. 8E). Western blot analysis showed that His 6 -HvNAC013(177-346) bound to immobilized GST-HvRCD1(485-579), but not to GST. This in vitro experiment confirmed the interaction between HvNAC013 and HvRCD1 and further confirmed that the TRD of HvNAC013 and the RST domain of HvRCD1 were sufficient for the interaction. HvNAC005 and HvRCD1 were also analyzed for interactions in yeast two-hybrid assays and in vitro pulldown assays. However, no interaction was detected for these proteins (Fig. 8, F and G), which cannot be explained by a lack of DBD-HvNAC005(1-355) expression in yeast (Fig. 5E).
To extend our examination of structure-function relationships of the HvNAC013 TRD, we analyzed the truncations used for transcriptional analysis (Fig. 5) for the ability to interact with HvRCD1 (Fig. 8H). In contrast to full-length HvNAC013 and its TRD , HvNAC013(1-250), lacking the 96 C-terminal amino acid residues of HvNAC013, did not interact with HvRCD1. In contrast, mutation of the conserved aspartate (D217A) and leucine (L212A) of the LP motif had no effect on the interaction. In conclusion, the ability of the HvNAC013 TRD to interact with RCD1 resides within the 96 C-terminal amino acid residues of HvNAC013, and no interaction was detected between the LP motif, conferring transcriptional activity to HvNAC013, and HvRCD1 (Fig. 8H).
Fluorescence spectroscopy was also used to analyze the interaction between HvNAC013(177-346) and His 6 -HvRCD1(485-579). Protein-protein interactions can be followed by fluorescence spectroscopy because changes in tryptophan and tyrosine fluorescence indicate a change in the environment around the fluorophore. Only one tryptophan residue is present in the HvNAC013 and HvRCD1 domains at positions 344 and 492, respectively. HvRCD1(485-579) has one tyrosine residue at position 510, whereas six tyrosine residues are present in HvNAC013(177-346), of which five are among the last 96 amino acid residues. The proteins were excited at either 280 or 296 nm and emission scans between 310 and 400 nm were recorded. Spectra of individual proteins and mixed HvNAC013(177-346) and His 6 -HvRCD1(485-579) were recorded (Fig. 9, A and B). The sum of the spectra of individual proteins was used as reference for noninteracting proteins. ⌬ max was calculated and a minor change of 4.5 and 3.0 nm in fluorescence max upon mixing of the two proteins was observed from the excitation wavelengths. This suggested that the environment around the aromatic residues changed to more hydrophobic upon interaction (Fig. 9B). This can be explained by a direct involvement of one or several tryptophan and tyrosine residues. Three of the tyrosine residues are located within the YF and RR motifs of HvNAC013 (Fig. 9B).
It is the general notion that many ID proteins fold upon binding to their protein interaction partners (4). However, more transient and flexible interactions may also occur that are not accompanied by structural changes of either partner (44). CD was used to analyze the structural changes occurring upon interaction further (Fig. 9, C and D). As the structure of the RST domain was unknown, a CD spectrum of His 6 -HvRCD1(485-579) was recorded. The pronounced minimum at 222 and 208 nm suggested that the domain is composed mainly of ␣-helical structures (Fig. 9C), in accordance with secondary structure predictions (supplemental Fig. S2). The CD spectrum of interacting HvNAC013(177-346) and His 6 -HvRCD1(485-579) seemed to be dominated by the HvNAC013 structure (Fig. 9D). A change in absorbance at both 222 and 208 nm was observed for mixed compared with individual proteins suggesting a shift toward a more unfolded state or more extended structure within the complex (Fig. 9D). This is a strong indication that HvNAC013 and HvRCD1 do not interact through a coupled binding and folding mechanism, but rather through a slight reorientation of the conformational ensemble of either or both proteins, possibly involving extended structure formation.
HvRCD1 Interacts with HvNAC013 and Other TFs through Their ID Regions-To test the hypothesis that RCD1 can interact with many structurally different TFs because of their structural flexibility, the interaction between AtRCD1 and previously reported interaction partners AtbHLH011 and AtMYB91 (20) was analyzed further using two-hybrid assays. These proteins represent major plant TF families and are characterized by a basic helix-loop-helix or SANT (Swi3, Ada2, N-CoR. TFIIIB) DBD and contain regions with predicted ID (Fig. 10A and sup-plemental Fig. S3). Both full-length proteins and the regions with ID were examined for interaction with AtRCD1. This confirmed the interaction between AtRCD1 and the two proteins (Fig. 10B). Furthermore, the ID region of AtMYB91 and the C-terminal ID region of AtbHLH011 was sufficient for detectable interaction. In contrast, the N-terminal ID region of AtbHLH011 did not interact with AtRCD1 (Fig. 10B). In conclusion, structurally verified and predicted ID regions of three plant TFs, HvNAC013, AtMYB91, and AtbHLH011, representing three major, structurally unrelated plant TF families, were sufficient for interactions with RCD1. However, ID as a single parameter is not sufficient for interaction with RCD1 as demonstrated by the inability of the N-terminal ID region of AtbHLH011 and the C-terminal ID region of HvNAC005 to interact with HvRCD1 suggesting that specificity may reside in additional details. were expressed in E. coli and probed for interactions with purified His 6 -HvNAC013(177-346) using antihistidine and GST antibodies. Glutathione-Sepharose beads incubated with His 6 -HvNAC013 and GST and His 6 -HvNAC013 served as negative controls. Input, total protein added to the glutathione beads; last wash, the volume was equal to the elution volume. GST and GST-HvRCD1(485-579) are indicated with arrows. F, GST pulldown assay of the HvRCD1 RST domain and the HvNAC005 C-terminal domain (residues 173-355). Conditions as described in E. G, directed yeast two-hybrid assay between HvNAC005 and HvRCD1. Conditions as described in B-D, but without 3-amino-1,2,4-triazole. H, directed yeast two-hybrid assay between HvNAC013 and HvRCD1. Conditions are as described under B-D, but using 90 mM 3-amino-1,2,4-triazole to inhibit self-activation. PARP, poly(ADP-ribose) polymerase.

DISCUSSION
In this study we present a comparative structure-function analysis of two distant NAC TFs associated with senescence. Both HvNAC013 and HvNAC005 have a prototypical NAC DBD and bound the NACBS with affinities typical of NAC TFs (8) suggesting that they mediate function through binding to NACBSs in target genes. They also both have C-terminal domains dominated by ID, typical of TRDs (5), but only HvNAC013 activated transcription. Thus, despite structural similarities, HvNAC013 and HvNAC005 could have different functionality in the same physiological context. Although most NAC TFs are transcriptional activators (8,37) some, e.g. Arabidopsis CALMODULIN BINDING NAC and VND-INTER-ACTING2 (VNI2), function as repressors (9,45). VNI2 inhibited transcriptional activation by VND TFs involved in xylem vessel formation, possibly by forming inactive heterodimers with the VND proteins, and had intrinsic repression activity typical of an active repressor (9). It is not possible to predict if HvNAC005 is a repressor. It will be interesting to test if HvNAC005 functions as a repressor modulating activity of NAC TFs involved in senescence through heterodimerization, in analogy to the mechanism suggested for VNI2 (9), or hinders activity of NAC TFs by competitive binding to NACBSs.
As shown in this study several motif-containing regions contribute to transcriptional activity of the TRD of HvNAC013. The LP motif was sufficient for activity, and mutational changes of single amino acid residues in this motif resulted in decreased activity. Removal of the complete motif from the full-length protein may have an even larger effect on activity, because it  contains several highly conserved amino acid residues likely to be of functional importance. The motif is frequent in Arabidopsis NAC TFs (identified in 28 of 109 NAC TFs) and positioned at the border between order and disorder in both HvNAC013 and the Arabidopsis NAC TFs (8). Transcriptional activity, evolutionary conservation and proliferation, and conservation of the order-disorder prediction for the LP motif region suggest that it plays an important functional role in NAC TFs. In ANAC012, involved in xylem fiber development, this motif was not able to activate transcription. This activity was instead mediated by the WQ motif, not found in HvNAC013 (37), demonstrating complexity of the NAC TRDs. Analysis of both HvNAC013 and ANAC102 (37) transcriptional activity was performed in yeast. An even more complex in planta picture, involving specific and general TFs and cofactors, can be expected. Based on knowledge from well studied TFs, such as p53, for which many interaction partners binding to ID regions have been identified (39), the NAC C-terminal domains can be expected to participate in interactions with numerous protein partners. Here, we have initiated a systematic analysis of one of the complex NAC TRDs and show that a specific sequence motif is functionally important, possibly by interacting with a specific protein partner.
The biophysical analysis of HvNAC013 and HvNAC005 showed that ID dominates their C-terminal domains, although both predictions and CD analyses suggested local and highly limited formation of secondary structures (Figs. 4 and 6, A and  B). Although RCD1 was identified as a potential regulator of NAC TFs (20), interactions between RCD1 and NAC TFs have not been characterized in details. The indispensability of the 96 C-terminal amino acid residues of HvNAC013 for the interaction with HvRCD1 together with the predicted structure for the last 24 residues of HvNAC013 (Fig. 4) may suggest that a MoRF is present in this part of HvNAC013 (6,7). However, analysis of the HvNAC013-HvRCD1 interaction by fluorescence spectroscopy and CD did not suggest significant folding upon interaction. In contrast, the small structural changes observed may suggest redistribution of the conformational ensemble suggestive of formation of a more extended conformation upon interaction. Lack of folding upon binding is also the case for protein interactions involving, e.g. the intrinsically disordered chain of the T-cell receptor (46). Thus, fuzziness, defined as disorder in the partner-bound state (44), may also characterize the interaction of the HvNAC013 TRD with RCD1, although atomic resolution structural information is needed to conclude this.
Arabidopsis RCD1 can interact with such different proteins as viral movement protein from turnip crinkle virus (47), the salt tolerance pathway component Salt Overly Sensitive (48), and different TFs (20). As shown here, the barley RST domain is sufficient for the interactions with TFs. This domain was identified from RCD1-like proteins and TAF4, which is involved in assembly of the general TF complex TFIID (49). This suggests that RCD1 is a regulator of TFs. Very interestingly, HvNAC013, but not HvNAC005 for which no activation activity was assigned, interacted with HvRCD1. The promiscuity of RCD1 interactions has not been explained previously. Here we show that barley and Arabidopsis RCD1 interact with TF regions dominated by ID, which may allow for flexibility in the interac-tions. Secondary structure predictions (supplemental Fig. S2) and CD analysis (Fig. 9C) suggested that the RST domain forms an ␣-helix fold that would make a solid platform for multiple, regulatory interactions with partner proteins. Thus, RCD1 could possibly be a novel hub. Hub proteins are divided into party hubs, which have multiple, simultaneous interactions, and date hubs with multiple, sequential interactions (50). Date hubs have been associated with both ID and transient binding (51). In the case of RCD1 and the TF interactions, ID is associated with the TFs, and the RST domain could be a folded date hub domain.
14-3-3 proteins, involved in crucial processes such as mitogenic signal transduction, cell cycle control, and apoptosis, are also folded hubs (52). More than 200 proteins, representing a diverse array of signaling proteins and receptors, interact with 14-3-3 proteins. Structure analysis of complexes between a 14-3-3 protein and peptides from different interaction partners showed that the broad specificity could be explained by a broad binding site allowing binding of different peptides. 14-3-3 recognition most likely involves coupled binding and folding of the recognition region (52). Like the RST domain of RCD1, 14-3-3 hub proteins have an ␣-helix structure, and interactions between the RST RCD1 domain and different TFs may have other similarities to interactions involving 14-3-3 proteins.
Further analyses should address if different TFs use common LMs (7) in binding to RCD1. Interestingly, HvNAC013 shares a 23-amino acid residue motif (supplemental Fig. S1, residues 296 -318), positioned at the border of the predicted disorder to order transition in the binding region (Fig. 4), with Arabidopsis ANAC046, ANAC087 ANAC079/80, and ANAC100. This motif, and the RR motif, which displays significant ␣ helix propensity (data not shown) suggestive of a MoRF (6), should be analyzed further for a functional role in interaction with RST domains. AtRCD1 was more promiscuous in its interactions with TFs than its paralogue SRO1, which also contains an RST domain (20). It will also be interesting to analyze HvNAC013, HvNAC005, and other NAC TFs for interactions with the RST domain protein TAF4 as a direct link to transcriptional core components. Another TFIID component, TAF9, interacts functionally with just a 5-amino acid residue motif (FSDLW) of p53 (3) making even small motifs of interest to binding.
The physiological relevance of the interaction between the TFs and RCD1 remains elusive. In Arabidopsis, RCD1 plays multiple roles in both stress and reactive oxygen species responses, and development (19,20), and the rcd1 mutant plant displayed several pleiotrophic phenotypes including senescence. Despite having a poly(ADP-ribose) polymerase domain RCD1 does not have poly(ADP-ribose) polymerase enzymatic activity (53), and the molecular function of RCD1 remains unclear. Although HvRCD1 was expressed during senescence, no significant induction of the expression was seen during senescence (data not shown), and AtRCD1 also exhibited only subtle regulation in response to stress exposure (53), suggesting that the physiological functions of RCD1 is not reflected at the transcriptional level. In summary, RCD1 and its relatives are likely to play broad physiological roles through regulation of different TFs. ANAC046, the Arabidopsis NAC most closely related to HvNAC013, was one of only three NAC TFs that interacted with RCD1 (21). This together with conservation of the interaction suggest that the biochemical interaction between HvNAC013 and HvRCD1, here described in mechanistic details, is of relevance to HvNAC013 and a role in plant senescence.