1

Solution structure of Gaussia Luciferase with five disulfide bonds 1

and identification of a putative coelenterazine binding cavity by 2

heteronuclear NMR 3

4

Nan Wu1,$, Naohiro Kobayashi2,$, Kengo Tsuda3, Satoru Unzai4, Tomonori Saotome5, 5

Yutaka Kuroda5,* and Toshio Yamazaki2,* 6

7

1 College of Food and Biological Engineering, Zhengzhou University of Light Industry, 136 8

Kexue Road, Zhengzhou 450002, P. R. China. 9

2 NMR Science and Development Division, RSC, RIKEN, 1-7-22 Suehiro-cho, Tsurumi-ku, 10

Yokohama City, Kanagawa 230-0045, Japan. 11

3 Division of Structural and Synthetic Biology, Center for Life Science Technologies, RIKEN, 12

1-7-22 Suehiro-cho, Tsurumi-ku, Yokohama City, Kanagawa 230-0045, Japan. 13

4 Department of Frontier Bioscience, Faculty of Bioscience and Applied Chemistry, Hosei 14

University, 3-7-2, Kajino-cho, Koganei-shi, Tokyo, 184-8584, Japan. 15

5 Department of Biotechnology and Life Science, Graduate School of Engineering, 16

Tokyo University of Agriculture and Technology, 2-24-16 Nakamachi, Koganei-shi, Tokyo 17

184-8588, Japan. 18

$ Equal contribution 19

* Correspondence: YK: ykuroda@cc.tuat.ac.jp (Tel: +81-42-388-7794) and TY : 20

toshio.yamazaki@riken.jp (Tel: +81-45-503-9262) 21

2

Keywords: SEP tag, in vitro refolding, disulfide bonds, helical protein, hydrophobic cavity, 22

intrinsically disordered region (IDR) 23

24

Database: The chemical shifts have been deposited in the Biological Magnetic Resonance 25

Bank (BMRB) under the accession No.36288, and the atomic coordinates are deposited in the 26

Protein Data Bank under accession number PDB-ID: 6KYN. The expression vector for 27

GLuc-TG (p21GLucTG) is deposited in Addgene (ID:124660). 28

29

Conflicts of interests: The authors declare no conflicts of interests. 30

31

3

Abstract 32

Gaussia luciferase (GLuc) is the smallest luciferase (18.2kDa; 168 residues) reported so 33

far and is thus attracting much attention as a reporter protein, but the lack of structural 34

information is hampering further application. Here, we report the first solution structure of a 35

fully active, recombinant GLuc determined by heteronuclear multidimensional NMR. We 36

obtained a natively folded GLuc by bacterial expression and efficient refolding using a 37

solubility tag. Almost perfect assignments of GLuc’s 1H, 13C and 15N backbone signals were 38

obtained. GLuc structure was determined using CYANA, which automatically identified over 39

2500 NOEs of which > 570 were long-range. GLuc is an all-alpha-helix protein made of nine 40

helices. The region spanning residues 10–18, 36-81, 96-145 and containing eight out of the 41

nine helices was determined with a Cα-atom RMSD of 1.39 ű 0.39 Å. The structure of GLuc 42

is novel and unique. Two homologous sequential repeats form two anti-parallel bundles made 43

by 4 helices and tied together by three disulfide bonds. The N-terminal helix 1 is grabbed by 44

these 4 helices. Further, we found a hydrophobic cavity where several residues responsible 45

for bioluminescence were identified in previous mutational studies, and we thus hypothesize 46

that this is a catalytic cavity, where the hydrophobic coelenterazine binds and the 47

bioluminescence reaction takes place. 48

4

Introduction 49

Luciferase (Luc) is a generic term for bioluminescent enzymes that catalyze the 50

oxidation of a substrate, often termed luciferin [1]. Together with GFP, Luc is widely 51

employed as a reporter protein [2–4]. Gaussia Luciferase (GLuc) is a luciferase isolated from 52

the marine Gaussia princeps [5], which catalyzes a bright blue light by oxidizing 53

coelenterazine. GLuc is the smallest luciferase reported so far with a molecular mass of 18.2 54

kDa (excluding the secretion tag). Nonetheless, its bioluminescence intensity is strong (200 55

fold higher than Firefly Luciferase and Renilla Luciferase, the two most widely used 56

luciferase), and it is thus considered as a potential ideal reporter protein [6]. Attempts to 57

improve or redesign GLuc’s bioluminescence characteristics included the lengthening of its 58

half-life luminescence [7–9], and the redshift of its light emission peak at 480 nm [10,11], 59

which is absorbed by tissues during in vivo applications [12]. However, structural information 60

at atomic resolution is still not available, making the redesign process tedious. 61

GLuc contains 10 cysteines, and previous studies demonstrated that the natively folded 62

GLuc contains five disulfide bonds. The presence of 5 disulfide bonds increases the risks of 63

misfolding when GLuc is bacterially produced, resulting in a low yield [13]. In order to 64

overcome this misfolding problem, several methods including fusion with pelB leader 65

sequence [7,14], cell-free systems [15], low-temperature expression [16] were reported, but 66

the yield of natively folded GLuc remained insufficient for high-resolution structural studies. 67

We previously developed a Solubility Enhancement Peptide tag (SEP tag [17–19]). We 68

showed that by attaching a SEP tag containing nine aspartic acids to GLuc’s C-terminus, we 69

5

could increase the solubility of GLuc, resulting in a spontaneous refolding and the formation 70

of native SS-bonds. Indeed, we obtained nearly 1mg of soluble and functional GLuc from a 71

200 ml of E.coli cultured in Luria-Bertani (LB) [11,13,20]. 72

Here, we used the SEP-tag fused GLuc construct to produce a sufficient amount of 15N 73

and 13C uniformly labeled GLuc for NMR studies. Heteronuclear multidimensional NMR 74

spectroscopy enabled over 99% backbone 1H, 13C, and 15N chemical shifts of GLuc to be 75

assigned. Flexible regions and highly stable regions were identified by 1H-15N heteronuclear 76

NOE [21] and H/D exchange experiments [22]. The three-dimensional structure calculated by 77

using CYANA (ver 3.98 [23]) were determined with a backbone (Cα ) RMSD of 78

1.39ű0.39Š(excluding residues in the flexible regions). 79

6

Results 80

Expression and purification of GLuc 81

The natively folded GLuc possesses ten cysteines that form five disulfide bonds, which 82

can be easily misformed when the protein is expressed in E.coli, and the cysteines are 83

air-oxidized in vitro. Here, we used a SEP-Tag, C9D, which solubilizes the protein during 84

air-oxidization and refolding, thereby increasing the yield of natively folded and active GLuc 85

[20]. The final yield of GLuc after tags cleavage and two times HPLC purification (Fig. S1) 86

was 1.5 mg per liter of M9 minimal medium culture, which was sufficient for NMR analysis. 87

GLuc’s identity was confirmed by MALDI-TOF mass (15N labeled GLuc, calculated＝88

19055.8 Da, experimental=19062.5 Da, Fig. S2). To date, the yield of natively folded active 89

GLuc is almost nil when expressed without the C9D tag [20], and the solubilization tag was 90

thus essential to achieve the present amount of protein, though it was removed once the 91

protein was folded into its native conformation. 92

93

NMR analysis 94

The 1H-15N HSQC spectrum exhibited dispersed and sharp peaks (Fig. 1), indicating a 95

stable and well-folded structure. Almost all backbone chemical shifts were visible in the 96

heteronuclear NMR experiments, and over 99% of backbone 1H, 13C and 15N resonances of 97

non-proline residues were unambiguously assigned. C136 was the only un-assigned backbone 98

H-N chemical shifts. The broadened signals around residue C136 suggested that the region 99

encompassing the C136/C148 SS-bond was subjected to structural exchange, as suggested by 100

7

the 15N relaxation dispersion of D138 and L140 (Table 1), and thus the C136 H-N pair was 101

undetected. 102

103

104Fig. 1. 2D 1H-15N HSQC spectrum of GLuc. The peak assignments are shown using the 105one-letter code followed by the residue number. Resonance assignments are numbered starting at the 106first residue (lysine) behind the secretion tag, which was removed without affecting the 107bioluminescence activity. Mutations at E100A and G103R do not affect activity and are described in 108our prior paper [13]. The inset in the right bottom (marked AUC) shows the result of sedimentation 109equilibrium experiments of GLuc protein (concentration at 0.3 mg/mL). Scans from three different 110rotor speeds (●: 12,000 rpm; ■: 22,000 rpm; ▼: 37,000 rpm) monitored at 280nm. The lines represent 111the fit to a single species model. The determined molecular weight was 22 kDa, corresponding to a 112monomer. 113

114

115

116

117

118

8

Residue R2(50 Hz)-R2(1 kHz)

1/s

V12 6.66

S16 4.84 T21 2.47

D26 2.12

G28 5.60

L37 2.88

A47 2.61

S61 6.93

M69 4.87

K70 2.07

G75 4.78

T79 6.43

T125 2.26

D138 4.46 L140 4.07

T150 4.46

A152 10.85

119Table 1. R2-dispersion experiments on GLuc. Differences of effective 15N transverse relaxation 120rates at CPMG rates of 50 Hz and 1 kHz are listed only for peaks that are resolved and their rate 121difference > 2 1/s. Larger difference was observed for residues in the C-terminal flexible region 122(indicated by bold letters) indicative of its structural dynamics. The two-state model analysis of CPMG 123rates of 50, 100, 150, 200, 250, 300, 400, 500, 600, 800, 1000 Hz showed that the estimated exchange 124rate was 2500 +/- 800 1/s. Because the residues showing R2 dispersion spread around several blocks, 125we judged farther residue-specific analysis using single exchange rate is unreliable. 126 127

128

129

130

131

132

133

9

The side-chain atoms were automatically assigned by FLYA [24] (a function of 134

CYANA) using the aliphatic atoms identified in the 3D HCCH-TOCSY, 15N- and 13C-edited 135

NOESY spectra. The assignments were confirmed by visual inspection and when necessary 136

corrected manually using the NMR spectra viewer and analyzer MagRO [25,26].We assigned 137

over 82.4% of 1H, 13C and 15N atoms of entire GLuc molecule. 138

The secondary structure elements were analyzed by TALOS+ using the 1H, 13C, 15N 139

chemical shifts (Fig. 2A). TALOS+ indicated that GLuc contains 36.9% helix and 4.7% 140

sheets, in reasonable agreement with our previous prediction based on the consensus of seven 141

publicly available secondary structure predictors (30% helix and 4% sheets) as well as with 142

the results of our Circular Dichroism (CD) analysis (30% helix and 12% sheets) [13]. In 143

addition, the location of helices calculated by TALOS+ and the secondary structure prediction 144

mostly overlapped (Fig. S3). 145

146

147

148

149

150

151

152

153

10

154Fig. 2. Residue-resolved structural and dynamics features of GLuc. (A) GLuc’s Secondary 155structure predicted by TALOS+: α-helix and β-sheet tendency are shown with solid and open bars, 156respectively. (B) H/D exchange experiments. Residues that retained resonance signal after incubation 157in D2O after 20 minutes and 18 hours were marked with open bars and solid bars, respectively (1H-15N 158HSQC figures are shown in Fig. S5). (C) 1H-15N heteronuclear NOE experiment data used to assess 159GLuc backbone flexibility: 1H-15N heteronuclear NOE are shown with solid bars. The NOE values of 160residues that were not identified were assumed using the average value of the preceding and following 161residues are shown with open bars. Flexible regions of GLuc were identified with the threshold value 162of 0.5. (D) Backbone Cα displacement from the representative structures is calculated using nineteen 163NMR-derived structures, and the error bars show standard deviations. (E) GLuc’s amino acid sequence 164and secondary structure that identified from the representative structure. 165

166

167

168

11

Structure calculation and disulfide bond determination 169

Since GLuc was a monomer as demonstrated by AUC, all NOEs were used as 170

intramolecular NOEs (Fig. 1). The statistics of NOEs assigned during the 19 CYANA runs 171

are shown in Table 2. Even though distance constraints for hydrogen bonds, disulfide bonds, 172

and some manually assigned NOEs were included in the CYANA calculations in addition to 173

the standard automatically assigned NOEs, the target functions were reasonably small 174

(4.07+/-0.59), indicating that the resulting structures were consistent with the experimental 175

data. 176

Ellman’s assay indicated that all ten cysteines (C52, C56, C59, C65, C77, C120, C123, 177

C127, C136, and C148) are oxidized in the active GLuc, and thus that they should form five 178

disulfide bonds. Three disulfide bonds C59/C120, C65/C77, and C136/C148 were 179

unambiguously visible in the NMR structures, but the pairing of the remaining four cysteines 180

(C52, C56, C123, and C127), which were close to each other, was less straightforward to 181

determine. In order to determine the remaining two disulfide bonds, we set the distance 182

between the gamma sulfur (Sγ) to > 2Å so that any cysteine could freely combine with any of 183

the remaining three cysteines. We then identified cysteine pairs with Sγ distance < 3Å in the 184

380 structures obtained from 19 rounds. As a result, C52/C127 and C56/C123 were the most 185

favored pairs and were observed in 92.4% and 56.1% of the calculated structures, respectively. 186

On the other hand, C52/C56 and C123/C127 were observed in only 13.7% and 19.5% of the 187

structures. 188

189

12

190 191 192

193Table 2. Structural statistics for the nineteen best NMR-derived GLuc structures. 19419 rounds of CYANA calculation with different random seed [36]. * The averaged numbers of NOEs 195and their standard deviations are calculated over the 19 rounds of CYANA calculations. ** Averaged 196over the 18 structures (except for the representative structure). 197

198

199

200

201

202

NOE distance restraints*

All 2573.4 +/- 42.4 Intra residue (|i－j|＝0) 565.9 +/- 14.7 Sequential (|i－j|＝1) 728.6 +/- 6.6

Medium-range (2≤|i－j|≤4) 700.5 +/- 18.1 Long-range (|i－j|>5) 579.7 +/- 32.2

Dihedral angle 183 Hydrogen bonds 25 Disulfide bonds 3

RMSD to the representative structure**

Backbone Cα (residues 10–18, 36-81, 96-145) 1.39 Å +/- 0.39 Å Heavy atoms (residues 10–18, 36-81, 96-145) 1.84 Å +/- 0.20 Å

Backbone Cα (helices without α2) 1.30 Å +/- 0.36 Å Heavy atoms (helices without α2) 1.74 Å +/- 0.45 Å

Ramachandran plot (representative structure)

Residues in most favored regions 71.8 % Residues in additionally allowed regions 20.4 % Residues in generously allowed regions 4.9 %

Residues in disallowed regions 2.8 %

13

Overall fold and dynamic features of GLuc 203

The structure with the lowest overall target function among the 20 NMR-derived 204

structures in each round was selected. Nineteen structures from 19 rounds were used for 205

further analysis. Among the nineteen structures, seven structures formed the putatively 206

correct disulfide bonds (C52/C127, C56/C123, C59/C120, C65/C77, C136/C148). We finally 207

selected the structure with the lowest average pairwise RMSD (against all other eighteen 208

structures) as the representative structure. This structure also forms the putatively correct 209

disulfide bonds. 210

The nineteen superimposed NMR-derived structures with the lowest overall target 211

function show that GLuc has nine helices (α1-α9, Fig. 2E and Fig. 3), and the location of all 212

helices essentially corroborate the TALOS+ prediction except for α2 (Fig. 2A and Fig. 2E). 213

The N- (residues 1-9) and C-terminus (residues 146-168) of GLuc are highly disordered. 214

GLuc’s main structure is formed by residues 10-145, in which the structure of residues10-18, 215

35-81 and 97-145 were well-defined with an average backbone RMSD to the representative 216

structure for all other eighteen structures of 1.39Å (Table 2). Residues 19-34 and residues 217

82-96 are highly disordered and can be considered as intrinsically disordered regions (IDR 218

[27], Fig. 2D, and Fig. 3). The structures of helices α1 and α3-α9 were well-defined with an 219

average backbone RMSD of 1.30Å (Table 2). It has been reported that α3-loop-α4-loop-α5 220

(α3-α5, residues 37-72) and α7-loop-α8-loop-α9 (α7-α9, residues 109-143) are repeat 221

sequences [13,28]. The structure analysis shows that GLuc’s two repeat sequences are 222

connected by the second IDR (residues 82-96) and form an anti-parallel bundle (α3+α8 pair 223

14

and α4+α7 pair) that surrounds the N-terminal α1 helix. The anti-parallel bundles are firmly 224

tied by three disulfide bonds (C52/C127, C56/C123, and C59/C120), resulting in a high local 225

stability (Fig. 3). Moreover, all residues in the well-defined region exhibited 1H-15N 226

heteronuclear NOE values larger and more uniform than residues in the N- and C- terminus or 227

in the two IDRs, confirming that the well-defined regions obtained by calculation were 228

consistent with the rigid regions determined by HN NOE values (Fig. 2C and Fig. 2D). 229

230

15

231Fig. 3. Overall fold of GLuc (residues 10-148) determined by NMR. (A) Wire model of 232nineteen superimposed NMR-derived structures with the lowest target function. Helices were shown in 233red and loops in black. (B) Ribbon model of the representative structure. The nine helices of GLuc are 234marked from α1 to α9. (C) Ribbon model of the representative structure with the five disulfide bonds 235colored in green. Two IDRs are in cyan. (D) Ribbon model of the representative structure with its two 236moieties. The tightly packed moiety (residues 52-123) is shown in orange, whereas the loosely packed 237moiety (residues 19-51, 124-151) is shown in yellow. The central helix, α1 (residues 1-18), is in white. 238Residues R76, Q112 and Q116 are in blue, F104 and F113 are in magenta, and T66 and T124 are in 239purple. 240

16

Discussion 241

The structure of GLuc is novel, as we detected no similar structures in the Protein Data 242

Bank using DALI [29] (Fig. S4). It is even quite different from the structures of Renilla 243

luciferase (RLuc) [30], Oplophorus Luciferase (OLuc) [31] and apoaequorin [32], which like 244

GLuc uses coelenterazine as a substrate and are ATP independent luciferases. The 245

anti-parallel bundle of helices, which exhibits pseudo 2-fold symmetry, in the GLuc fold can 246

be divided into two moieties. Though both showed well-defined backbone structures, the 247

experimental data indicated differences in the side chain packing stability. The side chains of 248

residues 52-123 are tightly packed whereas those of residues19-51 and 124-151 are loosely 249

packed (Fig. 3D). The high stability of the former one reveals good agreement with the 250

residues showing extreme low H/D exchange rates, whereas residues in the latter one 251

exhibited high H/D exchange rate indicative of a low stability (Fig. 2B and Fig. S5). In the 252

tightly packed moiety, we found several hydrophilic residues with well-determined side-chain 253

structures. For instance, the chemical shifts of R76-Hε, Q112-Hε 1/2, and Q116-Hε 1/2 were 254

clearly different from averaged values observed in a flexible side chain. Furthermore, many 255

NOEs were assigned to these atoms corroborating the fact that these side chains are involved 256

in hydrogen bonds stabilizing the tightly packed moiety. Interestingly the side-chains of R76 257

and Q112 are stacked to the aromatic rings of F113 and F104, respectively, apparently 258

shifting NMR signals of these protons from ring current effect (Fig. 3D). Finally, the 259

hydroxyl protons of T66 and T124 were also clearly visible, suggesting that they are involved 260

17

in hydrogen bonds and thus in the N-terminal capping of helices α5 and α8, respectively (Fig. 261

3D). 262

Surface accessible analysis of the representative structure indicated a noticeable cavity 263

located among the central α1, α4 and α7 (Fig. 4A, Fig. 4B and Fig. S6). The cavity was made 264

by 19 residues: N10, V12, A13, V14, S16, N17, F18, L60, S61, I63, K64, C65, R76, C77, 265

H78, T79, F113, I114, V117 (For reader’s convenience, we underlined the hydrophobic 266

residues; Fig. 4C). Similar cavities formed by these 19 residues were identified in all other 267

eighteen NMR-derived structures, though the sizes and shapes of their cavities showed some 268

variation because of the limited resolution of the NMR structures. The 19 residues are 269

distributed on three structural segments: α4+α7 most rigid block, R76-T79 short loop, and the 270

central α1. The α4+α7 most rigid block was stabilized by the three disulfide bonds as 271

mentioned above (C52/C127, C56/C123, and C59/C120, Fig. 3C). In addition to the rigid 272

structure of the cavity, we observed a structural exchange suggested by 15N relaxation 273

dispersion (Table 1). S61 was close to both V12 and T79 (Fig. 4C), which exhibited the 274

second-largest dispersions. We hypothesized that the structural exchange is related to the 275

opened and closed form of this cavity. 276

277

278

279

280

281

18

282

283

Fig. 4. The cavity shown using the representative structure (residues 10-148). (A) Surface 284representation of GLuc: positive residues (Arg, Lys and His) are colored in blue; negative 285residues (Glu and Asp) are colored in red; and hydrophobic residues are colored in yellow. The 286entrance to cavity is indicated by an arrow. (B) The cavity representation (colored in transparent light 287blue) of GLuc shown from the same direction as in (A). Residues that retained 1H-15N HSQC signals 288after 20 minutes and 18 hours H/D exchanging are shown in pink and purple, respectively; residues 289located in the activity-related loop R76-T79 are in orange; two IDRs are shown in cyan. (C) Residue 290composition around the interior cavity. The cavity wall and its contributing residues are colored using 291the same color code as in (B). The insets show ribbon models of GLuc with the cavity colored in light 292blue and viewed from the same direction as in the main panel. The entrance to the cavity is indicated 293by an arrow. 294

295

19

A flexible docking simulation indicated that the cavity was large enough to 296

accommodate coelenterazine; and this was verified for all seven models that formed five 297

disulfide bonds (Fig. S7). H/D exchange indicated that the amide protons of N17, L60, S61, 298

F113, I114 and V117, which are located around the cavity, were visible after 20min 299

incubation in D2O, and among them, L60, S61, I114, and V117 signals were visible even after 300

18hrs (Fig. 2B and Fig. S5). The four residues are located in the α4+α7 most rigid block (Fig. 301

2B and 2E), indicating that the cavity wall is rigid. The hydrophobic character of the cavity’s 302

interior suggests a putative role in recruiting coelenterazine, a small poorly soluble molecule. 303

Furthermore, three activity-related residues R76, C77, H78 [10] are located in the short 304

R76-T79 loop that is near the α4+α7 most rigid block and stabilized through the C65/C77 305

disulfide bond. C65 and C77 are also associated with bioluminescence activity. In particular, 306

the mutation of C77 resulted in a vanishing luminescence [10]. This can be rationalized by 307

hypothesizing that the destruction of C65/C77 disulfide bond ruins the entire cavity structure 308

and inactivate GLuc. 309

Sequence alignment also points to the role of the cavity as a binding pocket for 310

coelentarzine. First, the seven cavity forming residues: L60, S61, V117 (in the hydrophobic 311

region); R76, H78 (in the activity-related loop); and C65, C77 (the disulfide bond, see Fig. 312

4C) are highly conserved in 12 luciferases (MoLuc, MpLuc, etc. see Fig. S8A). C65, R76, 313

C77, V117 were fully conserved, and L60, S61, H78 had a 92% conservation ratio. 314

Furthermore, the structures of OLuc, RLuc and apoaequorin also contain a similar 315

hydrophobic cavity. Altogether, these observations strongly suggest that the cavity constitutes 316

20

the coelenterazine’s binding site and is thus essential for GLuc’s bioluminescence activity. 317

Additionally, we noticed that several residues in the C-terminal region (K141-D168) are 318

remarkably conserved (Fig. S8A) despite their high flexibility as assessed by heteronuclear 319

NOE analysis (Fig. 2), which may suggest that they are functionally or perhaps structurally 320

important. The sequence alignment of residues 27-97 with residues 98-168 indicates that 321

K141-F151 in the C-terminal region has a high similarity with K70-Y80 where the 322

aforementioned activity-related loop (R76-T79) is located (Fig. S8B). Furthermore, our 323

previous mutational analysis demonstrated that W143, L144 and F151 also play an important 324

role in GLuc’s activity [11], and it is of interest to note that these conserved residues are 325

disordered in our NMR structure and can be defined as IDRs. 326

Finally, let us note that several lines of evidence suggested that these residues are not 327

completely disordered. First, residues around F151 exhibited low 1H -15N NOE values (0.0 ~ 328

-0.2, Fig. 2C), suggesting a disordered state in the nano- or pico-second time scale, and the 329

R2-dispersion experiments indicated that these residues experience a micro- or milli-second 330

time scale exchange between the folded and unfolded states rather than in a perfectly flexible 331

state (Table 1). This exchange between a folded and a less folded state was further 332

corroborated by the observation of strong intra-residues and sequential NOEs in the 3D 333

15N-edited NOESY, and the relatively broad line shapes of the peaks in the 2D 1H-15N HSQC 334

(data not shown). Taken together, our results raise the possibility of an active participation of 335

flexible regions (that can be considered as IDRs) in the coelenterazine oxidation reaction. 336

337

21

Conclusion 338

We produced a recombinant 13C, 15N labeled GLuc in E.coli, and assigned nearly all of 339

the backbone and most of the side chain chemical shifts. The N- and C-termini, as well as the 340

segment located between α1 and α3 (encompassing α2) and the loop between α5 and α6 were 341

flexible. GLuc’s structure is unique and is made of nine helices, constituting two anti-parallel 342

bundles, which are formed by, respectively, helices α3-α4 and α7-α8 of parts of homologous 343

sequential repeats. The helices are tied together by disulfide bonds to form a 4-finger 344

structure with a pseudo-2-fold symmetry surrounding the N-terminal helix 1. Finally, we 345

identified a hydrophobic cavity where coelenterazine is most likely to bind and the catalytic 346

reaction occurs. The fold of GLuc is novel, and we believe that the above reported 347

structural/dynamic information will open an avenue for redesigning the bioluminescence 348

activity of GLuc and thereby widen its scope of application. 349

22

Materials and methods 350

Expression system 351

A DNA sequence encoding the wild-type GLuc gene (UniProtKB ID: Q9BLZ2) without 352

the 17 residues secretion tag and with an E100A and G103R mutations that increased protein 353

expression was synthesized as reported previously [13]. The GLuc sequence was flanked with 354

an N terminal His-tag and a C terminal SEP-tag (Solubility Enhancement Peptide tag, C9D) 355

to facilitate protein expression, refolding, and purification [20]. Two Factor Xa cleavage sites 356

were inserted between GLuc and His-tag/SEP-Tag. The GLuc gene named GLuc-TG [13] 357

was inserted into pET21c (Novagen) at the NdeI/BamHI site to construct p21GLucTG with 358

ampicillin resistance. 359

360

Protein expression and purification 361

p21GLucTG was transformed into BL21(DE3), and pre-cultured in 1 L Luria-Bertani 362

(LB) medium at 37°C and 250 rpm shaking. When OD590nm reached 1.0, E.coli cells were 363

collected by soft centrifugation and transferred to a 1 L M9 medium containing 13C-glucose 364

and 15NH4Cl. Isopropyl β-D-Thiogalactoside (IPTG) was added at 1 mM final concentration 365

for inducing protein expression, and the temperature was lowered to 25°C for minimizing the 366

formation of inclusion bodies. After 4 hours with shaking at 250 rpm, the cells were harvested 367

by centrifugation and sonicated. GLuc was purified from the supernatant fraction using a 368

Nickel Nitrilotriacetic Acid (NTA) column followed with overnight dialysis at 4°C against 50 369

mM Tris-HCl, pH 8.0. GLuc was then air-oxidized for three days at the same conditions in 370

23

order to form the five disulfide bridges. Residual misfolded GLuc was removed using a 371

reversed phase High-Performance Liquid Chromatography (HPLC). The protein 372

concentration was determined using a Bradford assay [33], and Factor Xa was added to GLuc 373

dissolved in 50 mM Tris-HCl, 100 mM NaCl, and 5 mM CaCl2 at a ratio of 1:100 (w/w), and 374

the sample was again incubated for 8 hours at 37°C, 100 rpm for enzymatic cleavage of the 375

His- and the SEP-Tags. Uncleaved GLuc was removed using, again, reversed phase HPLC. 376

GLuc identity was confirmed by MALDI-TOF mass spectroscopy on an ABI SCIEX 377

TOF/TOF 5800 (Thermo Fisher Scientific Inc., Massachusetts, USA). GLuc was freeze-dried 378

and kept as a powder at -30°C until use. 379

380

Analytical Ultracentrifugation (AUC) 381

Sedimentation equilibrium experiments were carried out using an Optima XL-A 382

analytical ultracentrifuge (Beckman-Coulter, Inc., Brea, California, USA) with a four-hole 383

An60Ti rotor at 20°C. Before centrifugation, GLuc samples were dialyzed overnight against 384

50 mM MES and 100 mM NaCl at pH 4.7. The solvent density (1.006983 g/cm3) was 385

determined using DMA 5000 (Anton Paar). Each sample was then transferred into a cell with 386

a six-channel centerpiece. The sample concentrations were 1.2, 0.6, and 0.3 mg/mL. Data 387

were obtained at 12,000, 22,000, and 37,000 rpm. A total equilibration time of 24 hours was 388

used for each speed, with absorbance scans at 280 nm taken every 4 hours to ensure that 389

equilibrium had been reached. Data analysis was performed by global analysis of all of the 390

data sets obtained at different concentrations and rotor speeds using SEDPHAT [34]. 391

24

NMR analysis and structure calculation 392

NMR experiments for resonance assignments and 1H-15N heteronuclear NOE experiment 393

were conducted using 0.2 mM 15N single or 15N, 13C double labeled GLuc protein dissolved in 394

50 mM MES buffer pH 6.0 and 2 mM NaN3, at 293 K with 8%(v/v) D2O in a 5 mm Shigemi 395

microtube (Shigemi co., Ltd, Tokyo, Japan). NMR spectra were acquired on a Bruker 396

Avance-III 700 MHz spectrometer, equipped with a 5 mm CPTXI cryoprobe. 397

Two-dimensional and three-dimensional NMR experiments (1H-15N HSQC, HNCACB, 398

CBCA(CO)NH, HNCA, HNCO, HN(CA)CO) were performed for the backbone 15N and 13C 399

assignments. 15N-TOCSY-HSQC, 15N-NOESY-HSQC, and HCCH-TOCSY were used for 400

backbone and side-chain signal assignments. H/D exchange 1H-15N-HSQC experiment was 401

performed under the above-described conditions but by dissolving GLuc’s freeze-dried 402

powder in D2O instead of H2O. The transverse relaxation rate (R2) dispersion experiments for 403

backbone 15N atoms were performed on a Bruker Avance-III 900 MHz spectrometer, using 404

pulse scheme including constant time relaxation compensated CPMG pulse sequences [35]. A 405

series of 2D experiments were acquired with various rates of CPMG pulse (50, 100, 150, 200, 406

250, 300, 400, 500, 600, 800, 1000 Hz) in the two of 20 ms CPMG blocks. The 15N CPMG 407

irradiation intensity was set to 3125 Hz (~35 ppm). To avoid the error in peak intensity arisen 408

from the offset effect, two sets of relaxation experiments with different 15N irradiation centers 409

(111 ppm and 125 ppm) were carried out. For each signal, we read the series of peak 410

intensities using the spectrum giving the smaller errors by the offset effect. 411

412

25

Three-dimensional structure determination 413

Automated NOE assignments and structure calculations were performed using CYANA 414

(ver. 3.98) on a PC-cluster equipped with 20-core Intel Xeon E5-4627v3 (3.0 GHz) and using 415

the manually assigned chemical shifts and a list of NOE chemical shifts derived from the 3D 416

15N-, 13C-edited NOESY spectra of the aliphatic and aromatic regions. For each cycle of 417

CYANA calculation, 20 out of 100 structures were selected after 10,000 steps of simulated 418

annealing using distance constraints derived from automatically assigned NOEs. The detailed 419

algorithm and strategy are described [23]. For the automated NOE assignments of the NOEs 420

peaks, the tolerances were set to 0.04, 0.4 and 0.4 ppm for 1H, 15N and 13C signals, 421

respectively. 19 rounds of CYANA calculations with different random seeds were performed. 422

It should be noted that, for a minor number of the NOEs, the automated assignment was 423

ambiguous and depended on the random seed. We thus calculated the structures using 424

restraints that were slightly different from set to set, and we selected the best structure from 425

the structures generated using each of the sets and reported the ensemble of structures. The 426

treatment using many random seeds was previously discussed and the ensemble of structures 427

calculated using ambiguous assignments becomes more diverse and safer. We can reduce the 428

risk to make structure ensemble affected by wrong NOE assignments [36]. According to 429

experimentally determined slow-exchanging backbone amide protons and threonine (Thr) 430

side chain OH atoms, we applied 24 sets of distance constraints, two of them for hydrogen 431

bonds including Thr-OH related to N-terminal capping of α-helices and 21 of them for 432

backbone amide protons to the 19 rounds of CYANA calculations. 433

26

Other softwares 434

The secondary structure elements of GLuc were predicted by submitting the backbone 435

chemical shifts into TALOS+ (https://spin.niddk.nih.gov/bax/nmrserver/talos/) [37]. 436

Root-Mean-Square Deviation (RMSD) was calculated using a Biopython.PDB module 437

(https://biopython.org) [38]. Three-dimensional images of the NMR structures were generated 438

using PyMOL. The flexible docking simulation was calculated using AutoDock (Ver. 4.2.6) 439

and AutoDockTools (Ver.1.5.6) [39] 440

441

442

27

Acknowledgments 443

We thank members of the Kuroda Laboratory for discussion, and help and advice with 444

the experimental operations. We are especially thankful to Dr. Tetsuya Kamioka and Mr. 445

Fumiya Suzuki for advice and kind help with protein expression and purification. We are 446

grateful to Prof. Russel Hopcroft, University of Alaska–Fairbanks (UAF), for kind permission 447

to reproduce his photograph of Gaussia princeps (used in the graphical abstract; R. Hopcroft 448

copyright), and Prof. Yanhong Bai, Zhengzhou University of Light Industry, for her generous 449

support of this international joint research. 450

451

Funding sources 452

The work has financially supported by a grant-in-aid from the Japan Society for the 453

Promotion of Science (JSPS) KAKENHI- 23651213 and 26560432, the institute for Global 454

Innovation Research at TUAT, and the Doctoral Scientific Research Foundation of 455

Zhengzhou University of Light Industry (No. 2018BSJJ020). 456

457

Authors’ contribution 458

W.N, Y.K. and Y.T. conceived the project, analyzed the structure, and wrote the manuscript. 459

W.N., K.T., and T.Y. performed the NMR experiments and analyzed the data, N.K and T.Y. 460

performed and analyzed the structure calculation with W.N’s assistance, S.U. and T.S. 461

performed the AUC experiments. All authors contributed to finalize the manuscript and 462

approved it. 463

28

Competing Interests 464

The authors declare no conflicts of interests. 465

466

References 467

1 Shimomura O (2006) Bioluminescence: chemical principles and methods. World Scientific 468

Publishing, Singapore. 469

2 Oyama H, Morita I, Kiguchi Y, Miyake S, Moriuchi A, Akisada T, Niwa T & Kobayashi N 470

(2015) Gaussia Luciferase as a Genetic Fusion Partner with Antibody Fragments for 471

Sensitive Immunoassay Monitoring of Clinical Biomarkers. Anal Chem 87, 12387–472

12395. 473

3 Kaskova ZM, Tsarkova AS & Yampolsky I V (2016) 1001 lights: luciferins, luciferases, 474

their mechanisms of action and applications in chemical analysis, biology and medicine. 475

Chem Soc Rev 45, 6048–6077. 476

4 Yi S, Liu N-N, Hu L, Wang H & Sahni N (2017) Base-resolution stratification of cancer 477

mutations using functional variomics. Nat Protoc 12, 2323. 478

5 C.S. Sent-Gyorgyi BJB (2001) Luciferases, fluorescent proteins, nucleic acids encoding the 479

luciferases and fluorescent proteins and the use thereof in diagnostics, high throughput 480

screening and novelty items. U. S. Patent 6232107-B. 481

6 Tannous BA, Kim DE, Fernandez JL, Weissleder R & Breakefield XO (2005) 482

Codon-optimized Gaussia luciferase cDNA for mammalian gene expression in culture 483

and in vivo. Mol Ther 11, 435–443. 484

7 Maguire CA, Deliolanis NC, Pike L, Niers JM, Tjon-Kon-Fat LA, Sena-Esteves M & 485

Tannous BA (2009) Gaussia luciferase variant for high-throughput functional screening 486

applications. Anal Chem 81, 7102–7106. 487

29

8 Welsh JP, Patel KG, Manthiram K & Swartz JR (2009) Multiply mutated Gaussia 488

luciferases provide prolonged and intense bioluminescence. Biochem Biophys Res 489

Commun 389, 563–568. 490

9 Degeling MH, Bovenberg MS, Lewandrowski GK, de Gooijer MC, Vleggeert-Lankamp CL, 491

Tannous M, Maguire CA & Tannous BA (2013) Directed molecular evolution reveals 492

Gaussia luciferase variants with enhanced light output stability. Anal Chem 85, 3006–493

3012. 494

10 Kim SB, Suzuki H, Sato M & Tao H (2011) Superluminescent variants of marine 495

luciferases for bioassays. Anal Chem 83, 8732–8740. 496

11 Wu N, Kamioka T & Kuroda Y (2016) A novel screening system based on VanX‐497

mediated autolysis-Application to Gaussia luciferase. Biotechnol Bioeng 113, 1413–498

1420. 499

12 Gheysens O & Mottaghy FM (2009) Method of bioluminescence imaging for molecular 500

imaging of physiological and pathological processes. Methods 48, 139–145. 501

13 Wu N, Rathnayaka T & Kuroda Y (2015) Bacterial expression and re-engineering of 502

Gaussia princeps luciferase and its use as a reporter protein. BBA-Proteins Proteom 503

1854, 1392–1399. 504

14 Tannous BA (2009) Gaussia luciferase reporter assay for monitoring biological processes 505

in culture and in vivo. Nat Protoc 4, 582–591. 506

15 Goerke AR, Loening AM, Gambhir SS & Swartz JR (2008) Cell-free metabolic 507

engineering promotes high-level production of bioactive Gaussia princeps luciferase. 508

Metab Eng 10, 187–200. 509

16 Rathnayaka T, Tawa M, Sohya S, Yohda M & Kuroda Y (2010) Biophysical 510

characterization of highly active recombinant Gaussia luciferase expressed in 511

Escherichia coli. BBA-Proteins Proteom 1804, 1902–1907. 512

30

17 Islam MM, Khan MA & Kuroda Y (2012) Analysis of amino acid contributions to protein 513

solubility using short peptide tags fused to a simplified BPTI variant. BBA-Proteins 514

Proteom 1824, 1144–1150. 515

18 Khan MA, Islam MM & Kuroda Y (2013) Analysis of protein aggregation kinetics using 516

short amino acid peptide tags. BBA-Proteins Proteom 1834, 2107–2115. 517

19 Nautiyal K & Kuroda Y (2018) A SEP tag enhances the expression, solubility and yield of 518

recombinant TEV protease without altering its activity. N Biotechnol. 42, 77–84. 519

20 Rathnayaka T, Tawa M, Nakamura T, Sohya S, Kuwajima K, Yohda M & Kuroda Y 520

(2011) Solubilization and folding of a fully active recombinant Gaussia luciferase with 521

native disulfide bonds by using a SEP-Tag. BBA-Proteins Proteom 1814, 1775–1778. 522

21 Bah A, Vernon RM, Siddiqui Z, Krzeminski M, Muhandiram R, Zhao C, Sonenberg N, 523

Kay LE & Forman-Kay JD (2014) Folding of an intrinsically disordered protein by 524

phosphorylation as a regulatory switch. Nature 519, 106. 525

22 Yi Q & Baker D (2008) Direct evidence for a two‐state protein unfolding transition from 526

hydrogen‐deuterium exchange, mass spectrometry, and NMR. Protein Sci 5, 1060–527

1066. 528

23 Güntert P & Buchner L (2015) Combined automated NOE assignment and structure 529

calculation with CYANA. J Biomol NMR 62, 453–471. 530

24 Schmidt E & Güntert P (2012) A New Algorithm for Reliable and General NMR 531

Resonance Assignment. J Am Chem Soc 134, 12817–12829. 532

25 Johnson BA & Blevins RA (1994) NMR View: A computer program for the visualization 533

and analysis of NMR data. J Biomol NMR 4, 603–614. 534

26 Kobayashi N, Iwahara J, Koshiba S, Tomizawa T, Tochio N, Güntert P, Kigawa T & 535

Yokoyama S (2007) KUJIRA, a package of integrated modules for systematic and 536

interactive analysis of NMR data directed to high-throughput NMR structure studies. J 537

31

Biomol NMR 39, 31–52. 538

27 Ota M, Koike R, Amemiya T, Tenno T, Romero PR, Hiroaki H, Dunker AK & Fukuchi S 539

(2013) An assignment of intrinsically disordered regions of proteins based on NMR 540

structures. J Struct Biol 181, 29–36. 541

28 Inouye S & Sahara Y (2008) Identification of two catalytic domains in a luciferase 542

secreted by the copepod Gaussia princeps. Biochem Biophys Res Commun 365, 96–101. 543

29 Hasegawa H & Holm L (2009) Advances and pitfalls of protein structural alignment. Curr 544

Opin Struc. Biol 19, 341–348. 545

30 Loening AM, Fenn TD & Gambhir SS (2007) Crystal structures of the luciferase and 546

green fluorescent protein from Renilla reniformis. J Mol Biol 374, 1017–1028. 547

31 Tomabechi Y, Hosoya T, Ehara H, Sekine S, Shirouzu M & Inouye S (2016) Crystal 548

structure of nanoKAZ: The mutated 19 kDa component of Oplophorus luciferase 549

catalyzing the bioluminescent reaction with coelenterazine. Biochem Biophys Res 550

Commun 470, 88–93. 551

32 Head JF, Inouye S, Teranishi K & Shimomura O (2000) The crystal structure of the 552

photoprotein aequorin at 2.3 A resolution. Nature 405, 372–376. 553

33 Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram 554

quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72, 555

248–254. 556

34 Vistica J, Dam J, Balbo A, Yikilmaz E, Mariuzza RA, Rouault TA & Schuck P (2004) 557

Sedimentation equilibrium analysis of protein interactions with global implicit mass 558

conservation constraints and systematic noise decomposition. Anal Biochem 326, 234–559

256. 560

35 Tollinger M, Skrynnikov NR, Mulder FAA, Forman-Kay JD & Kay LE (2001) Slow 561

Dynamics in Folded and Unfolded States of an SH3 Domain. J Am Chem Soc 123, 562

32

11341–11352. 563

36 Buchner L & Güntert P (2015) Increased Reliability of Nuclear Magnetic Resonance 564

Protein Structures by Consensus Structure Bundles. Structure 23, 425–434. 565

37 Shen Y, Delaglio F, Cornilescu G & Bax A (2009) TALOS+: a hybrid method for 566

predicting protein backbone torsion angles from NMR chemical shifts. J Biomol NMR 567

44, 213–223. 568

38 Manderick B & Hamelryck T (2003) PDB file parser and structure class implemented in 569

Python. Bioinformatics 19, 2308–2310. 570

39 Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS & Olson AJ 571

(2009) AutoDock4 and AutoDockTools4: Automated docking with selective receptor 572

flexibility. J Comput Chem 30, 2785–2791. 573

574

33

575

576