Expression and purification of adalimumab-Fab
The amino acid sequence used for the production of samples of labelled adalimumab-scFab has been described in details previously (Gagné et al., 2023). The heavy chain of adalimumab Fab domain and a portion of the hinge region (see (Hodgson et al., 2019) for complete amino acid sequence), residues Glu1 to Pro234 where Cys233 has been mutated to Ala233, was linked to residues Asp1 to Cys214 of the light chain via a linker made of five (GGGGS) elements plus SSGLVPRGS. The last residues of the linker contain a thrombin recognition site (LVPRGS). A poly-histidine tag (MGSSHHHHHH HHHHSSGHMLVPRGS) was fused to the amino terminal of this polypeptide. Papain cleavage, normally takes place between His2289 and Thr229, completely removed the fusion tag and the linker while leaving the thrombin recognition (LVPRGS) site at the N-terminal end of the light chain at the expense of great sample loss (80–90%). Therefore, we initially carried out the backbone and side-chain assignment of histag-adalimumab-scFab fragment, then produced a cleaved sample for final resonance assignment with the following sequence, where the extra residues (LVPRGS) were numbered starting at (-5):
Heavy chain:
1-EVQLVESGGG LVQPGRSLRL SCAASGFTFD DYAMHWVRQA PGKGLEWVSA ITWNSGHIDY
61-ADSVEGRFTI SRDNAKNSLY LQMNSLRAED TAVYYCAKVS YLSTASSLDY WGQGTLVTVS 121-SASTKGPSVF PLAPSSKSTS GGTAALGCLV KDYFPEPVTV SWNSGALTSG VHTFPAVLQS 181-SGLYSLSSVV TVPSSSLGTQ TYICNVNHKP SNTKVDKKVE PKSCDKTH
Light chain:
-5-LVPRGS-0
1-DIQMTQSPSS LSASVGDRVT ITCRASQGIR NYLAWYQQKP GKAPKLLIYA ASTLQSGVPS
61-RFSGSGSGTD FTLTISSLQP EDVATYYCQR YNRAPYTFGQ GTKVEIKRTV AAPSVFIFPP 121-SDEQLKSGTA SVVCLLNNFY PREAKVQWKV DNALQSGNSQ ESVTEQDSKD STYSLSSTLT 181-LSKADYEKHK VYACEVTHQG LSSPVTKSFN RGEC
Expression of labeled 2H-13C-15N-adalimumab-scFab was carried out by incubating Escherichia coli BL21(DE3) harboring the Histag-adalimumab-scFab construct (Gagné et al., 2023) at 37℃ (225 rpm) in M9/D2O minimal media supplemented with 2 g/L 2H-13C -glucose and 3 g/L 15N-ammonium chloride as sole source of carbon and nitrogen. Briefly, one colony was transferred to 4 mL of Luria Broth Miller (LB) and incubated for 2h at 37℃ (225 rpm). A 500 µL aliquot of the LB pre-culture was transferred to 50 mL of M9/H2O for another 6h of incubation after which, 2 mL was transferred into 200 mL of M9/D2O for an overnight pre-culture. The following day, the content was transferred to 2 liters of M9/D2O and returned to the incubator. Induction with 1 mM thio-D-galactopyranoside (IPTG) was conducted when the OD600 was 0.67. After 18h of expression, cells were recovered by centrifugation at 3,011 x g and stored at -80℃ until used.
Production of labeled 2H-13C-15N-1H-methyl-(Ile, Leu, Val)-adalimumab-Fab was conducted as described above, with the exception that 25 mg of α-ketobutyric acid-13C-3,3-d2 and 50 mg of α-ketoisovaleric-U-13C5 acid-3-d1 (dry powder) were added directly into the culture at an OD600 of 0.38, while the induction was conducted with 1 mM IPTG at an OD600 of 0.77.
Protein purifications were conducted using a slow dilution approach at pH 9.0 and with 1 M L-arginine, as described in previously (Gagné et al., 2023). Protein yields of 2H-13C-15N-histag-adalimumab-scFab and 2H-13C-15N-adalimumab-Fab were 11 and 1.3 mg/L of culture, respectively. Sample for resonance assignment contained 145 µM 2H-13C-15N-adalimumab-Fab (7 mg/mL) in 20 mM sodium acetate-d3 at pH 5.0 with 5% v/v deuterium oxide for lock frequency purposes in 50 µL and transferred in a 1.7 mm tube. Sample for side-chain methyl groups, prepared as described above with the addition of labelled intermediates as described by Goto and coworkers (Goto et al., 1999), contained 400 µM 1H-I(d1)LVmethyl-histag-2H-13C-15N-histag-adalimumab-scFab in 300 µL (5 mm Shigemi tube) in the same buffer.
NMR experiments
Data were collected at 40℃ (313 K) on Bruker AVANCE NEO 600 MHz (side-chain methyls assignment), AVANCE III-HD 700 MHz (backbone assignment) and AVANCE II 900 MHz NMR spectrometers equipped with 5mm, 1.7mm, and 5mm, respectively, TCI cryogenically cooled triple resonance inverse probeheads fitted with z-axis gradients. Chemical shift resonances were referenced with sodium 2,2-dimethyl-2-silapentane-5-sulfonate (DSS). Data collection for the assignment of the backbone resonances used the TROSY-based version(Eletsky et al., 2001) of the standard pulse sequences with deuterium decoupling during carbon evolution from the Bruker library: 2D-15NHSQC (trosyetf3gpsi), 3D-HNCO (trhncogp2h3d) (Salzmann et al., 1998), 3D-HN(CA)CO(Clubb and Wagner, 1992), 3D-HNCA (trhncagp2h3d2), 3D-HN(CO)CA (trhncocagp2h3d) (Eletsky et al., 2001), 3D-HNCACB (trhncacbgp2h3d), 3D-HN(CO)CACB (trhncocacbgp2h3d) (Grzesiek and Bax, 1993; Eletsky et al., 2001) on the 700 MHz NMR spectrometer fitted with the 1.7 mm NMR cryogenic probehead. All proton-nitrogen planes were collected using a spectral width (SW) of 18 ppm with 2048 real points (1H) and 40 ppm with 64 real points (15N). The 13C indirect dimensions were collected with a SW of 14 ppm, and 128 real points for HNCO/HNCACO, a SW of 30 ppm with 128 real points for HNCA/NHCOCA, and a SW of 80 ppm with 128 real points for HNCACB/HNCOCACB. All 3D datasets were collected with 32 transients per FID.
Side-chain methyl resonance of isoleucines (d1), leucines and valines were assigned with pulse sequences based on a carbon TOCSY element to transfer the methyl carbon magnetization down to either the alpha carbon or carbonyl prior to transfer to the bonded nitrogen and then proton for detection. These experiments are most efficient at fields of 600 MHz or less. Data were collected on sample of 1H-I(d1)LVmethyl-2H-13C-15N-histag-adalimumab-scFab at 400 µM (21 mg/mL) in 20 mM sodium acetate-d3 at pH 5.0 with 5% v/v deuterium oxide, in a 5 mm Shigemi tube. NMR pulse sequence codes were graciously provided by Prof. Lewis Kay (University of Toronto). In total a series of four 3D experiments 3D-CCC(CO)NH, 3D-HCC(CO)NH, 3D-CCC(CA)NH, and 3D-HCC(CA)NH were acquired (Tugarinov and Kay, 2003). Data were collected with a SW of 16 ppm with 2048 real points in the proton direct dimension and a SW of 40 ppm with 64 real points in the nitrogen dimension centered at 120 ppm. The indirect proton dimensions (HCC-) were collected with a SW of 3 ppm with 64 real points and the indirect carbon dimensions (CCC-) with a SW of 22 ppm with 54 real points for a total acquisition time of 54 h and 64 h, respectively.
Data analysis and resonance assignment and validation
NMR data were processed using nmrPipe (Delaglio et al., 1995). Sequential assignment was carried out manually using POKY (Lee et al., 2021).
Validation of NMR assignment
The web server of I-PINE (http://i-pine.nmrfam.wisc.edu/index.html) (Lee et al., 2019) was used initially through the POKY interface to verify and validate the adalimumab-Fab assignments. All peak lists from TROSY-based NMR experiments, namely 2D-15N-HSQC, 3D-HNCO, HN(CA)CO, HNCA, HN(CO)CA, HNCACB, and HN(CO)CACB, were used as input files. Analysis of the I-PINE output showed near-complete correctness of the manual assignments and identified a very few new assignments. Upon completion of manual assignments, we tested the new assignment protocol BARASA (Bishop et al., 2023) running under NMRBox using the same peak list and assignments as input data used for I-PINE. A total of 3 BARASA rounds were conducted, using 80 concurrent threads with 0.99 convergence p-value, and a stepwise energy drop of -500. Ca, Cb, and CO zero points were all set to 0.20, with a chemical shift energy range of -100 to 100.
Extent of assignments and data deposition
Resonance assignment of backbone atoms
We compared the 2D-1H-15N HSQC of histag-adalimumab-scFab with the fully cleaved adalimumab-Fab. The extra resonances belonging to the histag and linker were well resolved from any backbone resonances of the Fab while all backbone resonances of both samples were overlapping with each other, indicating that the assignment of the histag-scFab can be directly used for the Fab. Considering that the 1.7 mm cryoprobe only require a very small sample size (40 µL), we carried out resonance assignment of the backbone resonances on a sample of 2H-13C-15N-adalimumab-Fab. On the other hand, side-chain methyl assignment was carried out using an uncleaved sample. The Fab fragment contained a total of 442 amino acids (47.7 kDa) with the heavy chain having 228 residues (11 prolines) and the light chain having 214 residues (11 prolines).
The two-dimensional TROSY-HSQC spectrum of 2H-13C-15N-adalimumab-Fab shows well-dispersed resonances, typical of well-folded Fab (Fig. 1). A total of 381 (90.7%) 1H-15N backbone peaks were assigned, with 183 (84.3%) and 198 (97.5%) in the heavy and light chains, respectively. Assigned carbons include 407 (92.1%) 13C-O, 407 (92.1%) 13Cα, and 376 (91.7%) non-glycine 13Cβ. The Fab fragment is composed of four immunoglobulin domains that are each stabilized by one disulfide bond: Cys22-Cys96 (heavy chain, VH), Cys148-Cys204 (heavy chain, CH1), Cys23-Cys88 (light chain, VL), Cys134-Cys194 (light chain, CL), and one bond that links the heavy to the light chain Cys224-Cys214. All cysteine Cβ chemical shifts are higher than 35 ppm, which is indicative of properly formed disulfide bond, while reduced cysteine would have chemical shifts less than 35 ppm (Schulte et al., 2020).
Resonance assignment of isoleucine delta-1, leucine and valine methyl groups
Using the carbon tocsy versions of experiments, a total of 11 isoleucines (100%), 27 leucines (88%), and 30 valines (81%) were assigned (Fig. 2). Only 4 leucines have not been assigned, (Leu4, Leu108, Leu132 of the heavy chain, and Leu46 of the light chain), and 7 valines (Val5, Val37, Val 99, Val129, Val192 heavy chain and Val58 of light chain).
Comparison with NISTmAb assignment
Adalimumab, trastuzumab and the NIST-mAb are three monoclonal antibodies of the IgG1 class with light chain kappa. They all share identical primary sequences in their constant heavy 1 (CH1) and constant light domain (CL). In order to further validate our assignment, we compared the assignment of adalimumab-Fab CH1 and CL with the assignment of the corresponding residues of trastuzumab-scFab (submitted to this journal), as well as the yNISTmAb-Fab (Solomon et al., 2023). It is expected that resonances arising from amides with the same local magnetic environment will have the same chemical shifts while others that have similar or slightly different environments will produce slightly or significantly different chemical shifts. Indeed, comparison of assigned amide groups within the CH1 and CL from the three mAbs yielded the same assignment.
Validation of the backbone assignment
I-PINE was first used to validate the manual assignment and to help in identifying non-assigned residues. Out of the total 442 residues of adalimumab Fab, we initially manually assigned 395 residues corresponding to a coverage of 89.4%. I-PINE assigned 401 residues (i.e., only a few extra) and all residues except one of our 395 manual assignments matched the I-PINE results.
We also used BARASA for our manual assignment validation and identification of non-assigned residues. BARASA predicted assignments of 417 residues and we were able to assign additional residues, namely Asn92, Arg93, Ala94 of the light chain and Thr52, Trp53, Asn54, Ser55, Gly56, Ser78, Ser194 of the heavy chain. We also corrected assignments of Val58 of light chain using BARASA. The third round of BARASA confirmed correctness of all assignments.
Outlook
The resonance assignment of adalimumab will provide a powerful tool for the characterization of Humira® and its biosimilars. In addition, it will facilitate the application of NMR spectroscopy techniques to probe protein dynamics, epitope binding and drug excipients interactions at atomic level.