Insight into the interaction between the RNA helicase CGH-1 and EDC-3 and its implications

Previous studies indicated that the P-body components, CGH-1 and EDC-3 may play a crucial role in the regulation of lifespan in Caenorhabditis elegans. Homo sapiens DDX6 or Saccharomyces cerevisiae Dhh1p (CGH-1 in C. elegans) could form complexes with EDC3 (Edc3p in yeast), respectively, which is significant for translation inhibition and mRNA decay. However, it is currently unclear how CGH-1 can be recognized by EDC-3 in C. elegans. Here, we provided structural and biochemical insights into the interaction between CGH-1 and EDC-3. Combined with homology modeling, mutation, and ITC assays, we uncovered an interface between CGH-1 RecA2 domain and EDC-3 FDF-FEK. Additionally, GST-pulldown and co-localization experiments confirmed the interaction between CGH-1 and EDC-3 in vitro and in vivo. We also analyzed PATR-1-binding interface on CGH-1 RecA2 by ITC assays. Moreover, we unveiled the similarity and differences of the binding mode between EDC-3 and CAR-1 or PATR-1. Taken together, these findings provide insights into the recognition of DEAD-box protein CGH-1 by EDC-3 FDF-FEK motif, suggesting important functional implications.

www.nature.com/scientificreports/ Currently, we still don't know in detail how EDC-3 recognizes CGH-1 in vitro and in C. elegans? A yeast two-hybrid analysis had previously found that CGH-1 and EDC-3 interact and that the LSm domain of EDC-3 is not required for this interaction 16 . Moreover, in C. elegans, a substantial increase was observed in the lifespan of edc-3(ok1427) mutants harboring a deletion in the edc-3 locus. The corresponding mutant protein lacks 202 amino acids, including the conserved FDF domain of EDC-3 15 . This line of evidence implies that EDC-3 FDF mediated recruitment of CGH-1 might also be involved in the regulation of lifespan in C. elegans.
To dissect the structural and biological function relationship, the first step is to elucidate the interaction mechanism between CGH-1 and EDC-3 at the molecular or structural level. In this study, we aimed to investigate the interaction of EDC-3 with CGH-1 in vitro. Here, we expressed and purified recombinant CGH-1 248-420 and purchased EDC-3 peptide from company as described in the section "Materials and methods" (Fig. 1A,C), measured their binding affinity by isothermal titration calorimetry (ITC) assay. We found that EDC-3 235-271 binds to CGH-1 248-420 with a dissociation equilibrium constant (K D ) of approximate 0.34 μM. Based on homology modeling [we are not able to crystalize the complex of CGH-1 248-420 /EDC-3  ], mutation and biochemical analyses, we demonstrated that EDC-3 FDF-FEK is anchored to CGH-1 RecA2 domain through a conserved hydrophobic surface. Additionally, GST-pulldown assays and in vivo colocalization experiment confirmed the physical interaction between CGH-1 and EDC-3. Intriguingly, the binding mode of EDC-3 with CGH-1 is different from that of CAR-1 or PATR-1 (Pat1 ortholog in C. elegans) with CGH-1. Altogether, these findings provide insights into the recognition of DEAD-box protein CGH-1 by EDC-3 FDF-FEK motif, suggesting valuable implications for the further research of mRNA decay or/and lifespan in C. elegans.

Materials and methods
Plasmids and constructs. Expression plasmids regarding CGH-1 and CAR-1 were constructed as previously described 17 . In brief, the cDNA fragments encoding CGH-1 RecA2 (residues 248-420), CAR-1 FDF-TFG (residues 184-268) were amplified by PCR from C. elegans genome library and cloned into the NdeI and XhoI site of a modified pET-28a (Novagen) vector (p28a), in which a thrombin protease cleavage site was removed. Mutants of CGH-1 248-420 were individually generated through the MutanBEST kit (TaKaRa), and then confirmed by DNA sequencing.
The cDNA encoding a fragment of C. elegans EDC-3 (residues 230-566) plus a C-terminal His 6 tag was synthesized by Sangon Biotech (Shanghai, China) Co., Ltd. The DNA fragment was then cloned into a pGEX-4T-1 expression vector using restriction enzymes BamHI and XhoI, and finally confirmed by DNA sequencing.
Protein expression and purification. CGH-1 protein expression and purification were performed described as in our most recent paper 17 . In brief, proteins were expressed in E. coli Gold (DE3) strain (Novagen). E. coli cultures with kanamycin were incubated at 37 °C till A 600 reached about 1.0, and subsequently induced by 0.5 mM IPTG (isopropyl β-D-1-thiogalactopyranoside) at 16 °C overnight. Cells were resuspended in Buffer A (20 mM Tris-HCl, pH 7.5, 1 M NaCl), and then lysed using high pressure at 4 °C. Proteins were initially purified by Ni-chelating resin (Qiagen), and further purified by size exclusion chromatography (SEC) on a Superdex 75 column (GE healthcare) in Buffer A. The purified proteins were dialyzed into Buffer B (20 mM Tris-HCl, pH 7.5, 150 mM NaCl).
Both GST-tagged and His 6 -tagged fragment of EDC-3 protein was induced at 22 °C overnight in E. coli BL21 (DE3), initially purified by Ni-chelating resin (Qiagen) and further purified by SEC using a Hiload 16/60 Superdex 75 in a buffer containing 25 mM Tris-HCl, pH 7.5, and 500 mM NaCl, and followed by dialysis into phosphate buffered saline (PBS).

Peptide preparation.
A peptide containing the C. elegans EDC-3 FDF fragment (residues 235-271) and PATR-1 fragment (residues 30-67) were synthesized from GL Biochem (Shanghai, China) Co., Ltd, respectively. Stock solutions of peptide were prepared in Buffer B described above, and were adjusted to pH 7.5 ± 0.02 for subsequent experiments. The amino acid sequences of EDC-3 FDF peptide and PATR-1 peptide were shown as below: Homology modeling. The amino acid sequences of CGH-1 and EDC-3 were obtained from UniProt database (http:// www. unipr ot. org/), individually aligned with the orthologs using Clustal Omega 18 and subsequently analyzed using ESPript 3.0 19 . The complex structure of H. sapiens DDX6 and EDC-3 (PDB code 2WAX) was employed as a template for homology modeling. Homology modeling was performed using the alignment and PDB file to run the script in Modeller program 20 . All structural figures were displayed using PyMOL (http:// www. pymol. org/). Isothermal titration calorimetry. ITC assays were performed as previously described 17 . In brief, it was carried out on a MicroCal PEAQ-ITC (Malvern) at 20 °C with the following settings: reference power, 5 μcal/s; initial delay, 60 s; stir speed, 750 rpm; spacing time, 120 s. Proteins (or its mutants) and peptides were adjusted to 0.05 mM and 0.5 mM, respectively. Experiments were initialed by injection of 1 μl each of peptide into the sample cell (200 μl) filled with protein (or its mutants) solution, followed by 18 injections of 2 μl. The ITC data were fitted with a one-binding-site model using the Origin analysis software.

GST-pulldown assays.
To test if GST-tagged EDC-3 can pulldown CGH-1 protein, about 100 μg protein, GST alone or GST-EDC-3 230-566 , was incubated with GST resin (GE Healthcare) in PBS for 1 h. The resin was then washed three times with the same buffer. Subsequently, the resin was incubated with ~ 50 μg recombinant CGH-1 248-420 protein or its 4A mutant in the same buffer for 2 h. The resin was then washed five times using the same buffer. The captured proteins in resin were finally heated and analyzed using 12% SDS-PAGE.
To investigate whether CAR-1 184-268 can affect CGH-1 248-420 binding to EDC-3 230-566 , approximate 100 μg GST-tagged EDC-3 230-566 was incubated with GST resin, followed by three times of wash in PBS. The resin was then incubated with 50 μg CGH-1 248-420 in the absence or presence of the same amount of CAR-1 184-268 for 2 h, followed by five times of PBS wash, and captured proteins were finally analyzed as described above.
To investigate whether C. elegans PATR-1 peptide will affect the CGH-1 binding to GST-tagged EDC-3 230-566 protein, approximate 100 μg GST-tagged EDC-3 230-566 was incubated with GST resin, followed by three times of wash in PBS. The resin was then incubated with 50 μg CGH-1 248-420 without PATR-1 peptide (0 μM) or with varied concentration of PATR-1 peptide (12.5, 25, or 50 μM) for 2 h, and was washed five times in PBS. The captured proteins in resin were finally heated and analyzed using 12% SDS-PAGE.  Supplementary Table S1. Table 1. Thermodynamic parameters of the CGH-1 248-420 (or its mutants) bound to its ligand EDC-3 235-271 by ITC assays. Screening for transgene by PCR. Three days after injection, F1 animals expressing GFP marker were transferred to individual NGM plates and allowed to produce progeny for 2 or 3 days. Progeny of F1 were collected with 50 ml DNA lysis buffer (500 µg/ml Proteinase K, 100 mM NaCl, 50 mM Tris, 20 mM EDTA, and 1% SDS), and screened by PCR amplification with designed primers. Primer sequences are listed in Supplementary  Table S1.
Imaging. Images were collected using Leica DM4B and M165 FC microscopes.

CGH-1 RecA2 interacts with EDC-3 FDF-FEK domain in vitro. Previous studies in H. sapiens and
S. cerevisiae revealed that DEAD-box RNA helicases DDX6 and Dhh1p mediate protein-protein interactions, depending on the recruitment of specific interacting partners (reviewed in 21 ). CGH-1 is similar to its orthologs DDX6/Dhh1p, which contains two RecA-like domains connected by a short liker (Fig. 1A). In our most recent study, we have demonstrated in vitro that the ATPase activity of CGH-1 can be robustly stimulated by the MIF4G domain of NTL-1a (Not1 in yeast and CNOT1 in humans) in the presence of poly(U) RNA and ATP 17 . The Ccr4-Not deadenylase complex is responsible for the main ploy(A) removal activity in C. elegans 22 . We have also investigated in vitro that C. elegans CGH-1/CAR-1 (Scd6p in yeast and LSm14A in humans) interface 17 .
In addition to NTL-1a and CAR-1 and their orthologs, the complex formed by DDX6/Dhh1p and EDC-3 is vital for the translational repression, mRNA decay, and P-body assembly and localization [23][24][25] . In one recent study, it has been implicated that CGH-1 and EDC-3 might be involved in the regulation of lifespan during aging in C. elegans 15 . However, how EDC-3 interacts with CGH-1 in detail is unclear. Based on the studies of EDC-3 orthologs, we assumed that CGH-1 could also bind to EDC-3 FDF motif via its RecA2 domain. To this end, we expressed and purified the recombinant CGH-1 248-420 containing the RecA2 domain (Fig. 1A). As shown by the results of SEC and SDS-PAGE analysis, the quality for proteins is pure and uniform ( Supplementary Fig. S1). To confirm this hypothesis, we next tested the EDC-3 FDF peptide for its ability to bind to CGH-1 248-420 by performing the isothermal titration calorimetry (ITC) assays. Remarkably, the results showed that the EDC-3 FDF peptide directly binds to CGH-1 248-420 with a dissociation equilibrium constant (K D ) of ~ 0.34 μM and an N value of ~ 1 (Figs. 2C and 3F, and Table 1). Additionally, a GST-tagged fragment, which includes the central FDF and C-terminal YjeF-N domain of EDC-3 (EDC-3 230-566 ), but not GST alone, can interact with CGH-1 248-420 in a GST-pulldown assay (Fig. 4A). Overall, we conclude that CGH-1 RecA2 domain interacts with EDC-3 FDF domain in vitro.

A model of CGH-1 RecA2 in complex with EDC-3 FDF-FEK.
To elucidate the recognition mechanism of CGH-1 RecA2 domain by EDC-3 FDF, we tried to co-crystallize CGH-1 248-420 with EDC-3 FDF peptide, but didn't result in co-crystals. We then performed homology modeling by the Modeller program 20 . The model was produced using the complex crystal structure of human DDX6 and EDC3 (PDB code 2WAX) as the template due to the high sequence similarity (Fig. 1B,D,E).
We superimposed the completed model to the crystal structure of apo CGH-1 RecA2 domain, which has been recently determined by us (PDB code 7DTJ 17 ), giving a root-mean-square-deviation (RMSD) of ~ 0.75 Å over the backbone Cα atoms, suggesting that the resulting model is well consistent with our determined crystal structure of CGH-1 RecA2. In the model, CGH-1 RecA2 domain consists of the typical RecA-like folds in topology similar to other DEAD-box RNA helicases (Fig. 1E). EDC-3 FDF is embedded into the shallow groove formed on the helices α10 and α14 of CGH-1 RecA2 domain, and folded into the helices H1 and H2 (Fig. 1E).
To further confirm our structural model, we performed ITC experiment to determine the binding affinity of EDC-3 FDF peptide for CGH-1 248-420 WT or its mutants. The substitution of each of CGH-1 residues such as Phe 261 , Gln 266 , His 269 , with Ala weakened the binding affinity by a factor of ~ 3.4-7.5 (Fig. 2C,D, Table 1, and Supplementary Fig. S2) www.nature.com/scientificreports/ designated as 4A) in the hydrophobic pocket of CGH-1 severely weakened the binding affinity by a factor of ~ 25.7 (Fig. 2C,D, Table 1, and Supplementary Fig. S2), indicating that the 4A mutation of CGH-1 dramatically disrupted the interaction between CGH-1 and EDC-3 peptide. In line with data, less amount of the 4A mutant of CGH-1 248-420 binds to GST-tagged EDC-3 230-566 than wild-type CGH-1 248-420 does in a GST-pulldown assay (Fig. 4A). These results clearly suggested that the hydrophobic interactions mediated by CGH-1 hydrophobic pocket and EDC-3 FDF motif play a very important role in the recognition of CGH-1 by EDC-3 ( Fig. 2A).
In addition, the substitution of Lys 277 with Ala (K277A) or Glu (K277E) in CGH-1 weakened the binding affinity by factors of ~ 8.1 and ~ 17.5, respectively (Fig. 2C,D, Table 1, and Supplementary Fig. S2), supporting our structural model that Lys 277 of CGH-1 mainly mediates the electrostatic interaction with Asp 246 of EDC-3   www.nature.com/scientificreports/ for stabilization between CGH-1 and EDC-3 ( Fig. 2A). Moreover, in line with our model that Arg 393 of CGH-1 may electrostatically interact with Asp 262 of EDC-3, the substitution of Arg 393 with Ala (R393A) decreased the binding affinity by a factor of ~ 6.1 (Fig. 2B-D). The mutation of Glu 396 to Ala (E396A) also weakened the binding affinity, by a factor of ~ 2.5 (Fig. 2B,D). Taken together, these ITC titration results are well consistent with the interfaces provided by the model, indicating that EDC-3 FDF-FEK is anchored to CGH-1 in a specific manner. 16 . In line with this result, in vitro pulldown assays also found that the EDC-3 230-566 fragment excluding the LSm domain binds to CGH-1 RecA2 domain (Fig. 4A). Moreover, current models based on in vitro ITC assays suggest that EDC-3 should have a high affinity for CGH-1 in vivo.

Co-localization of CGH-1 and EDC-3 in C. elegans. A previous yeast two-hybrid analysis indicated that CGH-1 and EDC-3 interact and that the LSm domain of EDC-3 is not required for this interaction
To investigate the interaction of CGH-1 and EDC-3 in C. elegans, GFP::CGH-1 and 3xFLAG::mCherry::EDC-3 in situ were constructed via CRISPR/Cas9 technology 26 . In adult germ cells, GFP::CGH-1 and 3xFLAG::mCherry::EDC-3 were concentrated in foci which were distributed in a perinuclear pattern around nuclei. We crossed GFP::CGH-1 with 3xFLAG::mCherry::EDC-3 and found that a large portion of CGH-1 colocalized with EDC-3, supporting a physical interaction between the two proteins in vivo (Fig. 4D).
The interaction between C. elegans PATR-1 and CGH-1. Structural analysis of the yeast Pat1-Dhh1 complex reveals how Pat1 recognizes Dhh1 via its N-terminal Phe-Asp-Phe (FDF) motif 23 . Sequence alignment indicates that the yeast Pat1 DFDF motif is not conserved, and the corresponding residues are DDDW (H. sapiens and X. laevis), NGDW (D. melanogaster), and NAEL (C. elegans), respectively ( Supplementary Fig. S3A). It has been previously demonstrated by Co-immunoprecipitation assays combined with mutation that the hydrophobic Tryptophan (Trp 46 ) in the Asp-Trp (DW) motif of human PATL1 (Hs Pat1b) is target to the FDF-binding site on DDX6 23 . A substitution of Trp 46 to Ala (W46A) or Asp (W46D) in the YFP-tagged PATL1 impaired the interaction with human HA-tagged DDX6 23 .
In C. elegans, Boag et al. 8 demonstrated by cell biology that CGH-1 associates with PATR-1 in patr-1-dependent somatic P bodies. In contrast, patr-1 is not required for the formation of storage bodies in developing oocytes 8 . During oogenesis, CGH-1 associates primarily with CAR-1 and other regulators, and with and protects particular translationally regulated maternal mRNAs to form functional storage bodies or P granules 8,27 .
In the structural model of CGH-1 and EDC-3, the two phenylalanine residues (Phe 247 and Phe 249 ) of EDC-3 reside the hydrophobic patch which consists of His 269 , Cys 270 , Thr 273 and Leu 274 of CGH-1 ( Fig. 2A). In line with the structural model, the ITC data indicated that the 4A mutant dramatically impaired the binding affinity between EDC-3 and CGH-1 (Fig. 2C,D). Since the DFDF motif is absent in the amino acid sequence of C. elegans PATR-1, we hypothesize that the hydrophobic patch which consists of four CGH-1 residues His 269 , Cys 270 , Thr 273 and Leu 274 is not important for its interaction with PATR-1. To test this hypothesis, we prepared the four-alanine mutant (4A) and performed ITC experiment. To the end, we demonstrated that the 4A mutant almost has the www.nature.com/scientificreports/ same binding affinity with the wildtype CGH-1 protein ( Table 2, Fig. 5B, Supplementary Fig. S4). In other words, the four residues (His 269 , Cys 270 , Thr 273 and Leu 274 ) are not involved in the interaction with CePATR-1 30-67 peptide. Additionally, the mutation of each CGH-1 residue Gln 266 , His 269, or Glu 396 to Ala showed slight decrease in the binding affinity when compared to that of wildtype (Table 2, Fig. 5B, Supplementary Fig. S4). Instead, the substitution of each CGH-1 residue of Phe 261 , Val 262 , and Tyr 386 to alanine, reduced the binding affinity by factors of 2-3, indicating that Phe 261 , Val 262 and Tyr 386 may be involved in the hydrophobic interaction with CePATR-1. Additionally, the mutation of Lys 277 to Ala (K277A) or Glu (K277E) decreased the binding affinity by factors of approximate 2.1 and 3.9, respectively. Moreover, substitution of Arg 393 to alanine (R393A) weakened the binding affinity by a factor of 2.0. These results indicated that Lys 277 and Arg 393 may mediate electrostatic interaction with CePATR-1 peptide. Similarly, Lys 277 and Arg 393 have also been found important to interact with EDC-3 (this work) and CAR-1 17 .
We recently identified a highly conserved FDF (Phe-Asp-Phe) motif between beta-strand 12 and alpha-helix 13 in CGH-1 RecA2 domain but the function is currently unclear (Fig. 1B) 17 . Mechanistically, Phe 355 in CGH-1 may mediate hydrophobic interaction with Thr 266 in the TFG motif of CAR-1 and a mutation of Phe 355 to Ala (F355A) in CGH-1 decreased the binding affinity by a factor of approximate 5 as measured by ITC assay 17 . Since PATR-1N-terminus has also a TFG motif, we hypothesize that PATR-1 may also interact with CGH-1 via similar binding mechanism (Supplementary Fig. S3B and C). Strikingly, the F355A mutant of CGH-1 shows undetectable binding to PATR-1 peptide in our ITC assay (Fig. 5A, Supplementary Fig. S4). These data indicate that Phe 355 play a very important role in the recognition of PATR-1 by CGH-1, and the F355A mutation in the FDF motif of CGH-1 almost abolishes the binding of CePATR-1 30-67 peptide to CGH-1 248-420 in vitro assay. Taken together, these results imply that PATR-1 TFG-binding site is similar to CAR-1 TFG-binding site on the CGH-1 RecA2 domain.
Finally, we ask whether CAR-1 or PATR-1 peptide can affect the binding of CGH-1 to EDC-3? Co-incubation of the same amount of CGH-1 248-420 and CAR-1 184-268 with GST-EDC-3 230-566 showed little or no effect in the binding of CGH-1 248-420 to EDC-3 230-566 in a GST-pulldown assay (Fig. 4B). Moreover, co-incubation of increased amount of CePATR-1 30-67 peptide with the solution of CGH-1 248-420 and GST-EDC-3 230-566 does not impair the interaction of CGH-1 with GST-tagged EDC-3 230-566 in a GST-pulldown assay (Fig. 4C). These data may reflect a fact that EDC-3 has a higher affinity than CAR-1 or PATR-1 when they are binding to CGH-1 in vitro, as confirmed by our ITC assays.

Discussion
The similarities and differences of the binding mode between EDC-3 and CAR-1 for CGH-1. In C. elegans, CAR-1 is a germline specific cytokinesis, apoptosis, RNA-binding protein 27,28 , and contains three conserved domains: N-terminal Sm-like (Lsm) domain, central domain with FDF, FFD, TFG motifs, and C-terminal RGG box 28 . Our most recent works have delineated the recognition mechanism of CGH-1 by CAR-1 17 . By ITC assays, we found that the binding affinity of EDC-3 235-271 is approximately ninefold stronger than that of CAR-1 184-268 when they are bound to CGH-1 248-420 (indicated by K D , 0.34 μM vs 3.03 μM; Fig. 3F) in the same buffer conditions. It has also been demonstrated that a CAR-1 peptide (184-214) containing only the FDFEK motif binds to CGH-1 248-420 with a K D of ~ 43 μM by ITC assay in the same buffer conditions 17 . The binding affinity of EDC-3 FDF-FEK motif is more 126 times than that of the corresponding region of CAR-1 (0.34 μM (this work) vs 43 μM 17 ).
To understand the molecular basis by which cause so big differences in the binding affinity for EDC-3 and CAR-1, both containing the FDF binding motif that is anchored to CGH-1 hydrophobic pocket, we analyzed the binding mode between EDC-3 and CAR-1 when they are bound to CGH-1, respectively. From the comparison of structural models, we found that both EDC-3 and CAR-1 utilize the FDF motif to bind to the hydrophobic pocket of the CGH-1 RecA2 domain (Patch 1; Fig. 3A-D). For the recognition of EDC-3 FDF motif, CGH-1 residues His 269 , Cys 270 , Leu 271 , Asn 272 , Thr 273 , and Leu 274 may involve ( Fig. 2A). For the recognition of CAR-1 FDF motif, CGH-1 residues Ala 260 , Val 262 , His 269 , Cys 270 and Leu 274 might be involved 17 . The patterns of both Patch 1 are similar though the details are not exactly same (Fig. 3A,B). In other words, Patch 1 should not be the basic reason that make the big differences in the binding affinity of EDC-3 and CAR-1 when each of them is bound to the C-terminal RecA-like domain (RecA2) of CGH-1. There are some other reasons those might be critical to affect the binding affinity between EDC-3 and CAR-1when each is bound to CGH-1 RecA2.
We also investigated the differences in the binding modes between EDC-3 and CAR-1 for CGH-1, which is mainly reflected in the Patch 2 of the EDC-3/CGH-1 complex and Patch 3 in the CAR-1/CGH-1 complex (Fig. 3A-D). In EDC-3, the two CGH-1 binding sites FDF and FEK, which are respectively involved in Patches 1 and 2, are coupled by the helix H1 (Fig. 3A,C,E). In contrast to EDC-3, the FEK motif immediately follows the FDF motif, and they share the common phenylalanine Phe 192 in CAR-1 (Fig. 3E). In CAR-1, only the FDF motif is involved in the interaction with CGH-1 (Patch 1), the Glu 193 and Lys 194 of CAR-1, in which the sidechains are opposite to CGH-1, are not able to interact with CGH-1 (Fig. 3D,E). Instead, the ETFG motif (residues 265-268) in the αC helix of CAR-1 docks to CGH-1 and forms the second binding site (Patch 3 17 ). Taken all into together, we uncover the similarities and differences in the binding modes between EDC-3 and CAR-1 for CGH-1.
Coincidentally, the structural basis for the interactions of the P body components EDC3 and Tral (CAR-1 in C. elegans) with the DEAD-box RNA helicase Me31B (CGH-1 in C. elegans) are mutually exclusive in D. melanogaster, has been elucidated by structural biology combined with mutational and competition studies 25 . Similar binding mode between CAR-1/CGH-1 and ScPat1/ScDhh1p. Here, we have also discussed the binding modes in the complexes of CAR-1/CGH-1 and ScPat1/ScDhh1p. Intriguingly, by structural comparison of ScPat1/ScDDh1p and CAR-1/CGH-1, we found that ScPat1 show the similar binding mode to CAR-1  Supplementary Fig. S3C), the big difference is that the (E)TFG motif is located at the N-terminus of (D)FDF motif in the amino acid sequence of ScPat1, however, in the amino acid sequence of CAR-1, the (E)TFG motif is located at the C-terminus of αC helix (Patch 3). CePATR-1 is a homologue of ScPat1 ( Supplementary Fig. S3A and B). Though the (E/D)TFG motif is highly conserved, the (D)FDF motif is highly varied in the species from C. elegans to human ( Supplementary Fig. S3A). In C. elegans, the corresponding residues are NAEL in CePATR-1 protein. On one side, in the CGH-1/CePATR-1 complex, the ETFG motif of CePATR-1 may also dock to CGH-1 to form Patch 3. The molecular interaction might be mediated by the hydrophobic interaction between Phe 355 in the FDF motif of CGH-1 and PATR-1 TFG motif. This hypothesis was confirmed in our ITC assays where shows that the F355A mutation in CGH-1 almost abolished the binding of PATR-1 30-67 peptide to CGH-1 248-420 mutant (Fig. 5A). On the other side, CGH-1 residues His 269 , Cys 270 , Thr 273 and Leu 274, which has been confirmed important for the recognition of EDC-3, are not involved in the binding of PATR-1 peptide to CGH-1 because of the absence of a typical FDF motif in PATR-1. This hypothesis was further confirmed by our ITC assays where shows the 4A mutant has the same binding affinity as the wild-type CGH-1 (Fig. 5B, Supplementary Fig. S4). However, a small hydrophobic patch including Phe 261 , Val 262 and Tyr 386 in CGH-1 may be involved in binding of CGH-1 to PATR-1 (this work), and these three residues also interact with CAR-1 17 . Whether PATR-1 NAEL motif can recognize this small hydrophobic patch need be further studied in our future work.
In summary, EDC-3 mainly employed its short linear motifs, including FDF motif and FEK motif, to interact with CGH-1 RecA2 domain. The results of sequence and structural alignments indicated that the recognition mechanism is conserved. Furthermore, we uncovered the similarity and differences in binding of EDC-3, CAR-1, or PATR-1 to CGH-1.