Introduction

The glycine-rich RNA-binding protein, AtGrp7, was first determined as a component of a negative feedback loop in the circadian clock regulation of Arabidopsis thaliana^[1]. This protein can bind to a set of transcripts such as 5′ UTR, 3′ UTR, and intron RNA in Arabidopsis plants, including its own transcript^[1-3]. Through interaction with the RNA targets, AtGrp7 can promote or affect alternating splicing, pri-miRNA processing, flowering, and stomata opening and closing, besides influencing circadian oscillations at the post-transcriptional level^[4-7]. AtGrp7 is also involved in RNA-based pathogen defense, cold and drought tolerance, and other stress responses^[8-14]. AtGrp7 is composed of an N-terminal canonical RNA recognition motif (RRM) and a C-terminal unstructured glycine-rich domain. The RRM domain functions as an RNA chaperone in RNA metabolism, while the glycine-rich domain acts as a shuttle for RNA transport between the nucleus and cytoplasm, and contributes to RNA binding^[15]. To deeply understand the biological functions of AtGrp7 in plants, high-resolution structures of AtGrp7 RRM domain, especially of its complexes with variant RNAs are necessary.

Protein preparation of RRM domains originally from eukaryotes was usually performed via a regular purification procedure. Supernatant from cell extracts was passed through an Ni-NTA column for His-tagged protein, followed by a tobacco etch virus (TEV, or other proteases) cleavage to remove the tag, and then subjected to size-exclusion chromatography^{[16, 17]}. In some cases, an ion-exchange purification step was applied between the Ni-NTA process and size-exclusion chromatography. However, in our initial trial about the AtGrp7 RRM domain using this procedure, we observed obvious UV signals of mixed proteins and contaminants for the TEV-cleaved AtGrp7 RRM domain; the contaminants decreased following purification by size-exclusion chromatography. For this reason, other purification protocols should be tried in order to get pure AtGrp7 RRM domain for structural and biological researches.

In the present study, we aimed to obtain AtGrp7 RRM domain free of contaminants by using a two-step purification protocol of denaturing and refolding. The preliminary structure and function of the refolded AtGrp7 RRM domain were confirmed by nuclear magnetic resonance (NMR) spectroscopy and isothermal titration calorimetry (ITC).

1 Materials and methods 1.1 Cloning and overexpression of AtGrp7 RRM domain (AtGrp7_1-90)

DNA encoding the AtGrp7_1-90 sequence from Arabidopsis thaliana was synthesized by Genscript, Nanjing, China. The recombinant plasmid AtGrp7_RRM was obtained by inserting the synthesized DNA fragment into a modified pET16b vector at the XhoI/BamHI restriction sites. The modified pET16b was generated by insertion of the DNA fragment encoding a B1 domain of streptococcal protein G (GB1) and a TEV protease digestion site into the original pET16b vector after the N-terminal His₁₀ tag, resulting in an extra glycine at the N-terminus of the expressed protein after TEV cleavage. The recombinant plasmid AtGrp7_RRM was verified by DNA sequencing.

AtGrp7_1-90 was overexpressed in the Escherichia coli BL21 (DE3) strain after transformation of AtGrp7_RRM. The cells were first grown in 10 mL LB broth with ampicillin (100 μg/mL) at 37 ℃ (220 rpm) until an OD₆₀₀ of 0.6 was reached. For unlabeled AtGrp7_1-90, the cells were transferred to 1 L LB broth with ampicillin (100 μg/mL) and grown at 37 ℃ (220 rpm) until an OD₆₀₀ of 0.4 was attained, and were subsequently grown at 18 ℃ and 220 rpm for 0.5~1 h. For ¹³C, ¹⁵N- or ¹⁵N -labeled AtGrp7_1-90, cells were harvested when an OD₆₀₀ of 0.6 was reached, and the resulting pellet was resuspended in M9 minimum media, with 1 g/L ¹⁵N-labeled NH₄Cl (Cambridge Isotope Laboratories, Andover MA) as the sole nitrogen source, and uniformly labeled ¹³C-glucose (2 g/L; Cambridge Isotope Laboratories, Andover MA) or unlabeled glucose (4 g/L) as the sole carbon source. Cells were grown in M9 minimum media at 37 ℃ (220 rpm), until an OD₆₀₀ of 0.3 was reached, and subsequently grown at 18 ℃ and 220 rpm for 0.5~1 h. When the OD₆₀₀ reached 0.6 (or 0.4) for cells grown in LB broth (or M9 minimum media) at 18 ℃, cells were induced with 0.1 mmol/L isopropyl-β-D-thiogalactoside (IPTG) at 18 ℃ and 220 rpm for 16 h. Then, the cells were centrifuged at 5 000 g for 15 min at 4 ℃, and the cell pellets were frozen in liquid nitrogen and stored at –80 ℃.

1.2 Protein refolding and purification of AtGrp7_1-90

The cell pellets harvested from 1 L of the media were resuspended in 30 mL freshly prepared lysis buffer [50 mmol/L Tris-HCl (pH 8.0), 500 mmol/L NaCl, 8 mol/L urea, 0.5 mmol/L ethylene diamine tetraacetic acid (EDTA), 2 mmol/L dithiothreitol (DTT), 2 mmol/L phenylmethylsulfonyl fluoride (PMSF)]. The cells were sonicated at a 30% amplitude (Ningbo Scientz Biotechnology company; 15 min per cycle of 6 s on, followed by 6 s off intervals) for 5 cycles at 25 ℃, followed by shaking on an orbital shaker at 20 rpm for 1 h before centrifugation (15 000 g at 25 ℃ for 40 min). The supernatant was purified by passing through a column of Ni-NTA agarose resin pre-equilibrated with lysis buffer. After washing with high salt buffer [50 mmol/L Tris-HCl (pH 8.0), 1 mol/L NaCl, 8 mol/L urea, 0.5 mmol/L EDTA, 2 mmol/L DTT, 2 mmol/L PMSF, 20 mmol/L imidazole] and equilibration with low salt buffer (100 mmol/L NaCl; other components same as the high salt wash buffer), 40 mL elution buffer [50 mmol/L Tris-HCl (pH 8.0), 100 mmol/L NaCl, 8 mol/L urea, 0.5 mmol/L EDTA, 2 mmol/L DTT, 500 mmol/L imidazole] was used for elution, and the first 20 mL eluted fraction of the target protein was concentrated (to a final volume of 2~3 mL). The concentrated proteins were slowly added into a 40-fold volume of ice-cold refolding buffer [50 mmol/L Tris-HCl (pH 8.0), 1 mol/L NaCl, 0.5 mmol/L EDTA, 2 mmol/L DTT] drop by drop with stirring, at an interval of 30 s. The resulting solution was left overnight at 4 ℃ with stirring. The refolded proteins were concentrated to a volume of 2~3 mL again by centrifugation (2 000 g; Amicon Ultra-15, MWCO: 3 000, Millipore, USA), and then diluted with TEV buffer [50 mmol/L Tris-HCl, 100 mmol/L NaCl, 0.5 mmol/L EDTA, 5 mmol/L β-mercaptoethanol (BME), pH 8.0] to a low salt concentration (less than 200 mmol/L). Next, 18 μg/mL of TEV protease was added, and the proteins were left at 4 ℃ overnight for dialysis in TEV buffer. The TEV-cleaved protein mixture was passed through a column of Ni-NTA resin to obtain His-tagged and GB1-free samples. Then, AtGrp7_1-90 was loaded into a HiLoad 16/600 Superdex 75 preparation-grade column pre-equilibrated with NMR buffer [20 mmol/L NaPi (pH 6.0), 50 mmol/L NaCl, 0.5 mmol/L EDTA, 5 mmol/L BME], and purified at a flow rate of 0.5 mL/min. Finally, the AtGrp7_1-90 protein was exchanged with ddH₂O (18.2 MΩ·cm) by dialysis, lyophilized, and stored at -20 ℃. Protein samples were dissolved in the NMR buffer for subsequent ITC and NMR experiments.

1.3 Isothermal titration calorimeter measurements

6-nt (5′-TTCTGG-3′) and 32-nt (5′-ATTTTGTTCTGGTTCTGCTTTAGATTTGATGT-3′) DNA, which were DNA counterparts of the RNA from the 3′ UTR of AtGrp7^[18], were synthesized by Sangon Biotech (Shanghai, China) and used without further purification. ITC experiments were performed at 25 ℃ using MicroCal ITC200 for the AtGrp7_1-90 protein and the 6-nt or 32-nt DNAs dissolved in the NMR buffer. The concentration of AtGrp7_1-90 was in a range of 0.01~0.03 mmol/L, and those of DNAs were in a range of 0.1~0.3 mmol/L. All solutions were degassed for 5 min at 25 ℃, and titrations were performed at temperatures of (25±0.1) ℃. The cell volume was 200 μL and the injection volume was 40 μL, with an interval of 120 s between 20 consecutive injections. The ITC data was analyzed by the origin 7.0 software, which was supplied by MicroCal.

1.4 NMR spectroscopy

All NMR experiments were performed on a Bruker Avance Ⅲ HD 700 MHz spectrometer equipped with a QCI cryogenic probe. Three-dimensional HNCACB, CBCA(CO)NNH, HNCA, HN(CO)CA, and HNCO spectra were recorded at 20 ℃ for the ¹³C, ¹⁵N-labeled proteins in free form or as complexes with 6-nt 5′-UUCUGG-3′ from 3′ UTR of AtGrp7 (synthesized by Genscript, Nanjing, China). NMR titration experiments of ¹⁵N-labeled AtGrp7_1-90 and 6-nt (or 32-nt) DNA were performed with protein to DNA molar ratios of 0, 1:1, and 1:2. Chemical shift differences were calculated according to the following equation:

$ \mathit{\Delta }\delta =\sqrt{\mathit{\Delta }\delta _{\text{H}}^{2}+0.2\mathit{\Delta }\delta _{\text{N}}^{2}} $

(1)

All NMR data was processed using nmrPipe^[19] and analyzed using nmrViewJ^[20].

1.5 CS-Rosetta model generation

CS-Rosetta^[21-23] structural models were generated using the web server at the University of Wisconsin (https://csrosetta.bmrb.wisc.edu/csrosetta). In general, a TALOS table containing the protein sequence information and chemical shifts of ¹⁵N, ¹HN, ¹H_α, ¹³C_α, ¹³C_β, and ¹³C′ from the NMR backbone assignment were used as the inputs, and 3 000 models were calculated. The ten lowest energy structures calculated by CS-Rosetta was chosen for structural analysis.

2 Results and discussion 2.1 Construct design, expression optimization and initial purification of AtGrp7_1-90

The RRM domain of AtGrp7 from Arabidopsis thaliana was predicted to be 10~80 aa long, according to the Pfam database (http://pfam.xfam.org)^[24]. The first 100 residues, including the predicted RRM domain, were then used to perform protein Blast from the Uniprot database (http://www.uniprot.org/blast), resulting in more than 250 RRM-containing proteins with ≥70% sequence identities, in which a three-dimensional structure of the RRM domain of Nicotiana tabacum glycine-rich RBP1 (PDB code 4C7Q; 84% sequence identity) was already resolved^[25]. By sequence comparison and structure analysis, we predicted that the functional RRM domain of AtGrp7 is encompassed in the range of 7~86 aa. Finally, 1~90 N-terminal residues of AtGrp7 were selected for the present study. AtGrp7_1-90 was cloned into a modified pET16b vector with N-terminal His₁₀ and GB1 (B1 domain of streptococcal protein G)-tagged, which facilitated protein purification and was propitious to the yield, solubility, and stability of target proteins^[26]. A TEV cleavage site was inserted in between the DNAs encoding GB1 and AtGrp7_1-90 for the purpose of removing the His₁₀-tag and GB1 in the overexpressed protein.

Protein expression of AtGrp7_1-90 was tested in BL21 (DE3) at 18 ℃ and 37 ℃, by induction with 1 mmol/L IPTG for 4 h, and with 0.1 mmol/L IPTG for 16 h. A little higher protein expression level was observed at 18 ℃, as indicated by SDS-PAGE [Fig. 1(a)], and this temperature was then chosen as the optimum condition for protein overexpression. However, although Dnase I was added during the lysis step, our initial purification trial of AtGrp7_1-90 using the regular protocol that was used for Pin1 WW domains^[27] revealed that the TEV-cleaved proteins contained some contaminants, for which the UV peak centered at 269 nm indicated a mixture of proteins (~280 nm) and some contaminants like DNA or imidazole (~260 nm) [Fig. S1(b)]. The contaminant signal was not obvious after the first Ni-NTA purification, and it became clear after the second Ni-NTA step [Fig. S1(a) and (b)]. After size-exclusion chromatography, the amount of contaminants decreased. However, a shoulder peak could be implicated from the shape of the UV peak [Fig. S1(c)]. Considering the large extinction coefficients of contaminants like nucleotides or imidazole with respect to those of proteins, we explained the above phenomena as follows. Firstly, contaminants had a relatively weak binding affinity for the AtGrp7 RRM domain, and their amounts relative to those of the RRM protein were low. Secondly, the UV signal of contaminants was masked by that of overexpressed proteins with extinction coefficients ~20 000 L·mol^-1·cm^-1 (for GB1-fused AtGrp7_1-90) in the first Ni-NTA. Thirdly, after TEV cleavage, the absence of GB1 (with extinction coefficient 11 500 L·mol^-1·cm^-1) resulted in a greater contribution of the contaminants to the UV signal than in the first Ni-NTA step. Finally, most of the contaminants were separated from AtGrp7_1-90 by size-exclusion chromatography. However, a part of the contaminants remained due to their relatively weak binding to the RRM protein. Although an ion exchange before size-exclusion chromatography could improve the purity of AtGrp7_1-90 to some extent, the shape of the UV peak indicated that contaminants were still present in the final purified protein (data not shown).

2.2 Purification of AtGrp7_1-90 by denaturing-refolding

To solve the problem of impurities by regular purification, a two-step denaturing-refolding protocol was tried for AtGrp7_1-90. In the first step, 8 mol/L urea was included in all buffers during the first Ni-NTA purification step, to denature and purify proteins from cell extracts, of which the eluted fraction had a normal UV peak center at 277 nm [Fig. S1(a)]. The eluted proteins were relatively pure in the denatured condition, and only weak bands were seen in the SDS-PAGE gel [Fig. 1(b)]. In the second step, a quick-dilution method was performed to refold AtGrp7_1-90. The denatured AtGrp7_1-90 was slowly added to a 40-fold volume of ice-cold refolding buffer with high (1 mol/L) salt concentration. No precipitate was observed, and the refolding buffer remained clear during the refolding process. For AtGrp7_1-90, such a refolding process by quick-dilution showed priority to the refolding process by dialysis using stepwise decreasing urea concentrations, the latter resulted in significant loss of the AtGrp7 RRM domain protein. As calculated on the unit of moles for proteins in the all purification steps, the yield of protein recovery by refolding in the quick-dilution refolding process was 81% (43.1 out of 52.9 mg) and 73% (21.6 out of 29.4 mg) (Table 1) for unlabeled and ¹³C, ¹⁵N-labeled AtGrp7_1-90, respectively. Effective TEV cleavage was observed after dialysis of TEV-included proteins overnight [Fig. 1(c)]. The resulting AtGrp7_1-90 from the second Ni-NTA step showed a typical UV peak for proteins, which was obviously different from that for the regular purification process [Fig. S1(b)], and suggested the effective elimination of contaminants by denaturing-refolding. Final purification by size exclusion chromatography yielded pure AtGrp7_1-90 with an estimated molecular weight of 12.6 kDa [Fig. 1(c)~(d)], of which the expected molecular mass is 9.8 kDa. Undoubtedly, more experiments should be performed to identify the composition of the contaminants.

An optional ion-exchange purification step in the denatured conditions could be applied between the first Ni-NTA and the refolding steps to further remove impurities. Eluted proteins from the first Ni-NTA step were exchanged with a low salt buffer [20 mmol/L NaPi (pH 8.0), 50 mmol/L NaCl, 0.5 mmol/L EDTA, 6 mol/L urea], and then passed through a high-performance Q column with gradient NaCl concentrations from 50 mmol/L to 2 mol/L NaCl. His₁₀ and GB1-tagged AtGrp7_1-90 proteins (pI 5.68) flowed through without binding to the high-performance Q column, while most of the impurities with negative charges in the denatured condition were eluted out with increasing salt concentrations. However, the unlabeled sample with ion-exchange resulted in significant (44%) protein loss (47.0 mg loss out of 105.8 mg). Therefore, no ion-exchange was applied during the denaturing-refolding of ¹³C, ¹⁵N-labeled AtGrp7_1-90 for NMR structure determination. On the other hand, when highly pure AtGrp7_1-90 is required, ion exchange is still a good choice, especially if the protein amount is not the first consideration.

2.3 Verification of refolding and binding of AtGrp7_1-90

To verify the effectiveness of AtGrp7_1-90 refolding, a ¹H-¹⁵N HSQC spectrum was recorded for the refolded ¹⁵N-labeled protein. Peaks in ¹H-¹⁵N HSQC of AtGrp7_1-90 were well-dispersed in a range of δ 6.0~10.4 in the ¹H dimension (Fig. 2), indicating a globally folded structure of the refolded AtGrp7_1-90. In contrast, residue peaks of unstructured proteins were usually within δ 7.5~8.5 along the ¹H dimension in the ¹H-¹⁵N HSQC spectra. Success in protein refolding was further confirmed by almost fully overlapped ¹H-¹⁵N HSQC spectra of the refolded AtGrp7_1-90 and the proteins from regular purification [Fig. S1(d)]. On the other hand, such almost same spectra lead to the question about the importance of the little amount of contaminants, implying that purification protocol of AtGrp7_1-90 should be further optimized.

To check whether the folded AtGrp7_1-90 has a proper function in substrate binding, NMR titration experiments were performed between this protein and a 6-nt DNA counterpart of the RNA from the 3′ UTR of AtGrp7^[18]. Significant peak shifts in ¹H-¹⁵N HSQC spectra of the complex in a molar ratio of 1:1 were observed in contrast to that of the free AtGrp7_1-90 (Fig. 2), indicating that the refolded AtGrp7_1-90 could bind to its known target. Increasing the protein to DNA molar ratio to 1:2 resulted in no change in the ¹H-¹⁵N HSQC spectrum.

Quantitative analysis of the binding between the AtGrp7_1-90 and DNA counterparts from the 3′ UTR RNA of AtGrp7 was performed by ITC. The dissociation constant (K_d) of AtGrp7_1-90 and the 32-nt DNA was determined to be 0.8 μmol/L, compared to 0.1 μmol/L, as determined by fluorescence correlation spectroscopy for the full-length AtGrp7^[18]. The difference in dissociation constants could be mainly ascribed to the different sizes of AtGrp7 proteins used in the two studies, and partly attributed to the different techniques. The dissociation constant decreased to 9.8 μmol/L when the critical 5′-TTCTTG -3′ sequence^[18] was used (Fig. 3), suggesting that nucleotides other than the 6-nt DNA also contributed to AtGrp7_1-90 binding. Correspondingly, binding of AtGrp7_1-90 and 32-nt DNA in the molar ratio of 1:1 resulted in many disappeared or weakened peaks in the ¹H-¹⁵N HSQC spectrum (Fig. S2), implying that a much stronger binding was seen in the case of AtGrp7_1-90 and 32-nt DNA. Our results supported the opinion that more nucleotides should be considered in the studies of DNA-protein interaction, instead of the shortest form alone, as used in structure determination.

2.4 Structure models and binding analysis of AtGrp7_1-90

Structure models of AtGrp7_1-90 were constructed using chemical shifts derived from the NMR backbone assignment as input parameters by CS-Rosetta, which applied a SPARTA-based procedure to select protein fragments from the PDB database and assemble them using the Rosetta protocol. The structure models of AtGrp7_1-90 (Fig. 4) showed a typical RRM fold of a twisted five-stranded β-sheet together with two α-helices, in an order of βαββαββ. Analysis of ten superimposed models with the lowest energies yielded an average C_α-RMSDs of (0.07±0.01) nm against the lowest energy model. The representative structure model of AtGrp7_1-90 showed a high structural similarity to an RRM domain from Nicotiana tabacum with PDB code 4C7Q (Fig. S3), in accordance with their high sequence identity of 84% (and sequence similarity of 95%). Because the calculated models of AtGrp7_1-90 were generated by CS-Rosetta in a fragment basis, their high structural similarity to 4C7Q suggested that the models were reasonable and could be used for further binding analysis. The representative CS-Rosetta model of AtGrp7_1-90 was deposited at https://modelarchive.org.

We then directly applied the 6-nt 5′-UUCUUG-3′ from 3′UTR RNA of AtGrp7 to check its binding mode on AtGrp7_1-90. Chemical shift perturbation mapping was obtained based on NMR backbone assignments of AtGrp7_1-90 with and without the 6-nt RNA. The residues on AtGrp7_1-90 majorly influenced by the 6-nt RNA were located at three distinct regions of the primary sequence, which were V12-A18, I39-F56, and T80-R87 in a discontinuous manner [Fig. 5(a)]. To gain a clear view of the binding mode of AtGrp7_1-90 with the 6-nt RNA, we generated a complex model by referring to a RRM domain (sequence identity of 44%) of hnRNP G protein in the complex with a 6-nt RNA (PDB code 2MB0). The AtGrp7_1-90 structure model has a relatively high structural similarity to the RRM domain of hnRNP G in complex form, with a Dali z-score^[28] of 11.9 [Fig. S4(a)]. Meanwhile, its structural similarity to the aforementioned RRM domain from the Nicotiana tabacum glycine-rich RBP1 (PDB code 4C7Q) is 13.2. In general, the complex model of AtGrp7_1-90 and its 6-nt RNA substrate matched well with the chemical shift perturbation data [top and bottom of Fig. 5(b)]. A couple of residues (A18, T19, T45) experienced relatively large chemical shift changes while they are far away from the RNA binding surface on the model structure, which were attributed to local environment changes of the corresponding residues in loops. The main residues of AtGrp7_1-90 involved in RNA binding (with a chemical shift change of δ ≥ 0.4) were V12, T80, N82, and S86. Surprisingly, the residues around T80-R87 had large chemical shift changes due to RNA binding, which were unstructured in the free AtGrp7_1-90, as indicated by the structure model calculation. Structure analysis of the similar RRM domain of hnRNP G in complex form suggested that the residues around T80-R87 of AtGrp7_1-90 may experience large conformational changes after RNA binding [Fig. S4(b) and (c)], which were missed in the structure models. On the other hand, R49 in AtGrp7_1-90, which was a key residue for RNA binding^[29], did not show significant chemical shift changes after RNA binding [Fig. 5(a)]. Since our ITC data showed that the dissociation constant of the AtGrp7_1-90 R49Q mutant and the corresponding 6-nt DNA decreased by one order of magnitude compared to that of the wild-type AtGrp7_1-90 (data not shown), it is more likely that the backbone amides of R49 are less sensitive to RNA/DNA binding, and a high-resolution structure is required to elucidate the molecular details of AtGrp7_1-90/RNA binding.

3 Conclusions

AtGrp7_RRM, encoding the RRM domain AtGrp7_1-90 of Arabidopsis thaliana, was cloned into a modified pET16b vector in which GB1-fusion could improve the yield, solubility, and stability of AtGrp7_1-90. By using a quick-dilution refolding procedure, a satisfactory yield of protein was obtained for NMR structure determination; the residual contaminants in the purified protein were also avoided. But still, current results suggest that more experiments have to be performed to identify the composition of the contaminants and to optimize the purification protocol of AtGrp7_1-90. The structure of the AtGrp7 RRM domain was fully recovered by refolding, as supported by its finger-print ¹H-¹⁵N HSQC spectrum and CS-Rosetta model structures. ITC and NMR titration experiments further confirmed that the RRM domain of AtGrp7_1-90 had the right function of RNA/DNA binding. In future studies, we aim to obtain atomic level three-dimensional NMR structures of the AtGrp7 RRM domain and its complex with target RNA to understand its function as an RNA/DNA chaperone.