Comparative studies on manual and automatic backbone chemical shift assignments of ²H/¹³C/¹⁵N-labeled Ube2g1

Abstract

Background

Ube2g1, one of human E2 enzymes, possesses an additional acidic loop in the vicinity of the active Cys90 residue. The presence of the backbone chemical shifts (bCSs) is essential for various NMR studies of Ube2g1. Triple resonance experiments for the bCSs assignment of ¹³C/¹⁵N-labeled Ube2g1 were not efficient due to the relatively high molecular size (more than 20 kDa).

Methods

²H/¹³C/¹⁵N-labeled Ube2g1 was prepared to increase the T2 relaxation time, and then the TROSY-based triple resonance spectra were recorded with ²H decoupling. After the bCSs of Ube2g1 were assigned manually, the resulting bCSs were compared with those from two different automatic assignment programs. The AutoAssign program that utilizes only the peak lists of the NMR spectra accomplished 69 % assignment of the bCSs. On the other hand, the RASPnmr program requires the additional reference CSs that can be predicted from a relevant PDB coordinate. The homologous models of Ube2g1 were generated by various modeling programs, and then their CSs were predicted by using the SHIFTX2 program.

Results

The implementations of the predicted bCSs into the RASPnmr analysis resulted in complete and accurate assignment of the bCSs of ²H/¹³C/¹⁵N-labeled Ube2g1.

Conclusions

The reference bCSs calculated from various homologous models of Ube2g1 enabled the automatic bCSs assignment process by the RASPnmr program, and a similar application will be possible for the bCS assignments of other proteins.

Background

Ubiquitylation is one of important cellular modifications and regulates varieties of biological functions, such as cell cycle control, transcriptional regulation, and DNA repair (Hershko and Ciechanover 1998). During the ubiquitylation, E2 enzyme accepts a ubiquitin from E1 enzyme and then transfers the ubiquitin to a substrate with the assistance of E3 enzyme. More than several dozens of E2 enzyme have been identified in humans, and E2 enzymes have structural diversity for their own specific ubiquitylation activities. Human ubiquitin-conjugating enzyme E2 G1 (Ube2g1) is mainly expressed in skeletal muscle and testis and is moderately expressed in 15 other tissues (Watanabe et al. 1996; Lin and Wing 1999). Interestingly, Ube2g1 has a distinct acidic loop (residues 97–109) in the vicinity to the active Cys90 and the mutation of the acidic loop resulted in the impaired K48 ubiquitylation activity (Choi et al. 2015). Ube2r1 (human Cdc34) and Ube2g2 are another E2 enzymes that have a similar acidic loop. It has been reported that the acidic loops of Cdc34 and Ube2g2 also play an important role during their specific K48 ubiquitylation reactions (Petroski and Deshaies 2005; Li et al. 2007). Especially, Ube2g2 has a particular K48 ubiquitylation mechanism in the presence of gp78 (E3 enzyme), in which poly-ubiquitin chain is preassembled at the catalytic Cys residue (Li et al. 2007).

The X-ray crystallography is already a well-established technique and has many advantages compared to the NMR technique for determining the three-dimensional (3-D) structure of a protein. However, most protein NMR experiments are performed in aqueous solution and thus are suitable to study the protein–protein interaction and detailed molecular dynamics of a target protein. The presence of backbone chemical shifts (bCSs) is a prerequisite of various NMR experiments to study the nature of a target protein. Here, we determined the bCSs of ²H/¹³C/¹⁵N-labeled Ube2g1 and compared the efficiency of two different automatic bCSs assignment programs, the AutoAssign (Moseley et al. 2001) and the RASPnmr (MacRaild and Norton 2014) programs.

Methods

Protein expression and purification

Human ube2g1 gene was cloned into pGEX-4T-1 vector using BamH 1/Xho I restriction enzymes. The ²H/¹³C/¹⁵N-labeled Ube2g1 was obtained by growing E. coli Rosetta DE3 strain in 99 % D₂O M9 minimal media supplemented with 1 g of ¹⁵N-NH₄Cl, 2 g of ²H/¹³C-glucose, and 0.5 g of ²H/¹³C/¹⁵N-Celtone base powder (Cambridge Isotope Laboratories, Inc.). One milliliter of 2 M MgSO₄ and 0.1 ml of 1 M CaCl₂ that were prepared in 99 % D₂O solutions were appended for 1-l M9 culture. MEM vitamin (100X, SIGMA) and the trace metal D₂O solutions were prepared by re-dissolving the freeze-drying powders using 99 % D₂O solutions. The trace metal solution (200 μl) that consisted of 2 mM CoCl₂, 2 mM CuSO₄, 10 mM FeCl₂, 10 mM H₃BO₃, 10 mM MnSO₄, 5 mM Na₂MoO₄, 2 mM Na₂SeO₃, and 5 mM ZnSO₄ was added into 1-l M9 culture.

The GST-tagged Ube2g1 protein was purified by using Hitrap-GST affinity column (GE Healthcare). The elution fractions were concentrated, and then the GST tag was cleaved by thrombin digestion. Ube2g1 was finally purified by gel filtration column chromatography (GFC) using Superdex 75 (GE Healthcare) in buffer (pH 7.0, 50 mM HEPES, 100 mM NaCl, 1 mM DTT), and the remained GST tag in the purified Ube2g1 protein was completely removed by passing into the small volume of GST-affinity column.

NMR experiments and data analysis

Three hundred microliters of 0.8 mM ²H/¹³C/¹⁵N-Ube2g1 was prepared in a buffer (pH 7.0, 50 mM HEPES, 100 mM NaCl, 5 mM DTT, and 5 % D₂O) and then was transferred into a Shigemi tube. For the backbone chemical shift (bCS) assignment of Ube2g1, TROSY version NMR data of ¹H,¹⁵N-Trosy, trHNCO, trHN(CA)CO, trHNCA, trHN(CO)CA, trHN(CO)CACB, and trHNCACB was recorded at 25 °C using Bruker 900 MHz NMR spectrometer equipped with a cryogenic probehead. All data were processed using the NMRPipe program (Delaglio et al. 1995), and the resulting NMR spectra were analyzed using the SPARKY program (Goddard and Kneller).

The automatic assignments of bCSs were done using two different programs, the AutoAssign (Moseley et al. 2001) and the RASPnmr (MacRaild and Norton 2014). The second one required an additional input of the reference CS values that can be predicted from the Protein Data Bank (PDB) coordinate. The four different homologous model structures of Ube2g1 were generated via the Web-based analysis programs (I-TASSER (Yang et al. 2015), IntFOLD2 (McGuffin et al. 2015), Phyre2 (Kelley et al. 2015), RaptorX (Kallberg et al. 2014)). The SHIFTX2 program (Han et al. 2011) calculated the reference CSs of ²H/¹³C/¹⁵N-Ube2g1 using these four different PDB structures. All visualization of the PDB structures was done using the Chimera program (Pettersen et al. 2004).

Results and discussion

Although Ube2g1 consists of 170 amino acids (molecular weight, ~20 kDa), its apparent molecular size that includes the effect of protein shape seemed to be much higher. The molecular size of Ube2g1 was roughly estimated to be ~25 kDa during the final GPC purification. The triple resonance experiments using ¹³C/¹⁵N-labeled sample were not efficient due to the short T2 relaxation time, and thus we prepared per-deuterated Ube2g1 protein by growing E. coli cell in D₂O-M9 media containing ²H/¹³C-glucose and ¹⁵N-NH₄Cl. The deuterium-labeled protein can increase the T2 relaxation time of the amide protons, due to the smaller gyromagnetic ratio of deuterium compared to that of proton (4.065 × 10⁷ vs. 2.675 × 10⁸ rad s⁻¹ T⁻¹). Three pairs of NMR spectra that showed a sequential connectivity between two residues were recorded with ²H decoupling, (i) trHNCO and trHN(CA)CO, (ii) trHNCA and trHN(CO)CA, and (iii) trHN(CO)CACB and trHNCACB. The assignment of backbone chemical shifts (bCSs) of ²H/¹³C/¹⁵N-Ube2g1 (H_N, N, C_α, C_β, and C_CO) was manually performed. The bCSs of most residues were completely assigned except for those of Met1. The C_α, C_β, and C_CO of Pro45 and Pro69 could not be assigned because of the presence of consecutive Pro46 and Pro70, respectively (Table 1). The automatic bCS assignment can provide a great convenience to manual assignment that is a time-consuming and tedious process. We first tried to assign the bCS of Ube2g1 using the AutoAssign program that utilizes the peak lists of all pairs of NMR spectra (Moseley et al. 2001). During the analysis, the error values were varied for each different types of CSs (H_N, N, C_α, C_β, and C_CO). Although the variation of error values caused different assignments, any clear improvement in the assignment results was not identified for the different analyses (Fig. 1). Unfortunately, this automatic assignment using the AutoAssign program was not complete for the bCSs of Ube2g1. Only 58 and 11 % of the residues were correctly and wrongly assigned, respectively, and 31 % of the residues still remained to be unassigned.

Table 1 The backbone chemical shift values of ²H/¹³C/¹⁵N-labeled Ube2g1

Fig. 1

Automatic bCSs assignment of ²H/¹³C/¹⁵N-labeled Ube2g1 using the AutoAssign program. The correctly and wrong assigned residues are indicated with black and red bars, respectively. The analyses using varied error values resulted in somehow different assignments. However, there was no clear variation of the assignment accuracy among the results. The analyses using the AutoAssign program accomplished overall 58 and 11 % of correct and wrong assignments compared to the manually assigned bCSs of ²H/¹³C/¹⁵N-labeled Ube2g1

To increase the efficiency of automatic assignment, we tried to use the RASPnmr program that requires additional information of the CS reference that can be calculated from the PDB coordinate of a target protein (MacRaild and Norton 2014). The RASPnmr program is capable of auto-assigning only for the residues of which the reference bCSs are present. The X-ray structure of Ube2g1 has not yet been available, and the 3-D coordinate of the truncated Ube2g1 (2AWF) is only deposited to the Protein Data Bank. There are many PDB coordinates of various E2 proteins, and thus the model structures of Ube2g1 could be readily prepared by using various homologous modeling programs. Therefore, we generated four different homologous models of Ube2g1 using the Web-based modeling programs, I-TASSER (Yang et al. 2015), IntFOLD2 (McGuffin et al. 2015), Phyre2 (Kelley et al. 2015), RaptorX (Kallberg et al. 2014), to take account of the effects from wrong CS references during the RASPnmr analyses. These programs can produce the coordinate also for the unstructured regions of Ube2g1 (the N/C-terminal parts and the acidic loop). The model structures of Ube2g1 could be well superimposed, except for the flexible regions including the N/C-terminal parts and the acidic loop (Fig. 2). The presence of the flexible parts in these homologous models of Ube2g1 was also confirmed by the measured ¹⁵N-relaxation data (T1, T2, and ¹H,¹⁵N-heteronuclear NOE values) (Fig. 3). The estimated rotational correlation times using the T1/T2 ratios (Kay et al. 1989) were roughly 12 to 14 ns for the rigid part of Ube2g1, which indicates the molecular size of Ube2g1 is larger than 20 kDa. The theoretical CS values of the model structures were calculated by using the SHIFTX2 program (Han et al. 2011), and then were implemented to the automatic assignments of the RASPnmr program as the bCSs reference. Indeed, the automatic assignments using the RASPnmr program accomplished almost complete and accurate assignment of the Ube2g1 bCSs (Fig. 4). Although four different sets of the reference bCSs resulted in slightly different and wrong assignments during the RASPnmr analyses, the analyses resulted in almost identical and correct auto-assignments. The probabilities of the correct assignments using the Phyre2 model were lower in the acidic loop region than those using the other models, but the values were still higher than 70 %. There are many programs to predict the homologous model of protein structure, and they utilize their own specific algorism. The homologous models of Ube2g1 seem to be accurate for the E2-core domain, but those of the unstructured parts vary with the algorisms of the modeling program (Fig. 2). Nevertheless, the presence of homologous 3-D structure enables an efficient performance of the RASPnmr program and thus likely accelerates the tedious bCS assignment that is very slow especially for a larger protein.

Fig. 2

Four homologous structures of Ube2g1 are overlaid. The homologous models were calculated using the Web-based modeling programs; I-TASSER (red), IntFOLD2 (green), Phyre2 (blue), RaptorX (purple). The structured regions of Ube2g1 are well superposed, but the unstructured regions (the N/C-terminal and the acidic loop parts) are not

Fig. 3

The R1, R2, and ¹H, ¹⁵N heteronuclear NOE (NOE) values of Ube2g1. The ¹⁵N-relaxation data of ²H/¹³C/¹⁵N-labeled Ube2g1 were measured using 800 MHz NMR at 30 °C. The ¹⁵N-relaxation data of the amide bonds show that the acidic loop region (i.e., residues 97–109) is more flexible compared to the other structured regions of Ube2g1

Fig. 4

The summaries of the automatic assignment results from the RASPnmr analysis. The reference CSs were calculated for the model structures (I-TASSER, IntFOLD2, Phyre2, and RaptorX) using the SHIFTX2 program, and then the predicted CSs values were implemented to the RASPnmr analyses. The probabilities of the correctly assigned residues by the RASPnmr program are shown for the different reference CS values

Conclusions

The perdeuteration of Ube2g1 protein increased the T2 relaxation time and made it possible to record various pairs of triple resonance NMR spectra that are used for the bCS assignment. The calculation of the reference bCSs using the SHIFTX2 program based on the homologous models of Ube2g1 enabled the highly efficient automatic bCS assignment by using the RASPnmr program.