Rapid and Efficient Generation of Stable Antibody–Drug Conjugates via an Encoded Cyclopropene and an Inverse‐Electron‐Demand Diels–Alder Reaction

Abstract Homogeneous antibody–drug conjugates (ADCs), generated by site‐specific toxin linkage, show improved therapeutic indices with respect to traditional ADCs. However, current methods to produce site‐specific conjugates suffer from low protein expression, slow reaction kinetics, and low yields, or are limited to particular conjugation sites. Here we describe high yielding expression systems that efficiently incorporate a cyclopropene derivative of lysine (CypK) into antibodies through genetic‐code expansion. We express trastuzumab bearing CypK and conjugate tetrazine derivatives to the antibody. We show that the dihydropyridazine linkage resulting from the conjugation reaction is stable in serum, and generate an ADC bearing monomethyl auristatin E that selectively kills cells expressing a high level of HER2. Our results demonstrate that CypK is a minimal bioorthogonal handle for the rapid production of stable therapeutic protein conjugates.


Supplementary Materials and Methods
Cell culture conditions for sorting and expression. Mammalian CHO and HEK cultures were maintained at 37°C with 8% CO 2 in humidified incubators. Suspension cultures were maintained in the same conditions, but with a shaker at 125 rpm. PylS and 4 copies of PylT U25C . We PCR amplified PylS/PylT from our previously reported PiggyBac plasmids. 1 The four copies of PylT are arrayed in tandem, and are placed adjacent to the PylS expression cassette. Because the U6::PylT repeats are difficult to amplify by PCR, we used AccuPrime Pfx polymerase (ThermoFisher) and designed primers to be 300bp away from the start of the PylT cassette. 2 All other PCRs were done with Q5 or Fusion polymerases (NEB). pcDNA3.4, with a multiple cloning site from pcDNA3.1 inserted into the main open reading frame, was linearized by PCR.
The two PCR fragments were joined via Gibson assembly into NEB Stable cells and grown at 30°C to reduce recombination of the tRNA repeats. Heavy and light chains were cloned into their own plasmids using Gibson. Trastuzumab genes were ordered as gBlocks from IDT. UAG amber codons were introduced into Gibson homology regions for a multi-piece Gibson assembly.
A PiggyBac plasmid SE315 with 5 copies of PylT U25C , an EF1α controlled PylS, and an IRES followed by a blasticidin resistance gene was previously cloned. This plasmid was used to generate amber suppressing CHO-S lines. Trastuzumab HC::IRES::GFP (GFPspark) and LC::IRES::BFP (mTagBFP2) constructs were assembled via PCR and Gibson into separate DNA2.0 stable CHO plasmid pD2531. The HC plasmid was cloned with both wild type and HC-118TAG variants. These plasmids were used to generate stable antibody producing lines. For all plasmids, ncAA and UGA stop codons, often in tandem, were exclusively used at the end of genes. UAG amber codons were only used at sites chosen for incorporation of CypK or BocK.
Transient expression. For transfection, Expi293F cells were grown in Expi293 media without added antibiotics to a density of 2.5x10 6 cells/mL. HC and LC pKYM1 plasmids were mixed 1:1 and transfected using Expifectamine (Gibco) in the presence of CypK. The following day, boosters were added per the manufacturer's instructions.
Antibodies were harvested 7 days later from the supernatant.
Stable expression cell lines. Stable cell lines to produce trastuzumab with CypK were generated in two steps: insertion of the genetic code expansion machinery and integration of the heavy and light chains.
In the first step, we generated the amber suppressing lines (CHO-S HC118TAG) by using PiggyBac to insert a cassette with five copies of PylT and PylS under the U6 and EF1α promoters, respectively, into the genome ( Figure S4a). SE315, which has 5' and 3' ITRs for integration, was co-transfected with PiggyBac transposase at a 10:1 ratio using PEI into CHO-S cells in suspension culture. A blasticidin resistance marker was included in the vector in order to select for the cells that had efficiently integrated the PylS/PylT cassette ( Figure S4b). Two days after transfection, 10 µg/mL of blasticidin was added and the culture was maintained for 10 more days. The batch was grown without blasticidin for one week to recover. Following blasticidin selection, we sought to isolate a clonal line with high amber suppression efficiency. To this end, we transiently transfected a plasmid encoding super-folder green fluorescent protein (sfGFP) with an amber codon at position 150 (CMV-sfGFP150TAG), which only fluoresces when the amber codon is read through. We grew the resulting cells in the presence of 2 mM N ε -(tert-butyloxycarbonyl)-L-lysine (BocK) for 4 days with 1:100 anticlumping agent. The batch was filtered to remove clumps and cells with high amber suppression, as indicated by GFP fluorescence, were isolated into 96 well plates using fluorescence-activated cell sorting (FACS).
The efficiency of the selected amber suppressing lines that grew well was verified through rephenotyping using a transient GFP::TAG::mCherry plasmid in adherent cells grown in 2 mM BocK for 4 days. This system allows visualizing relative transfection efficiency through GFP fluorescence and amber suppression efficiency in the mCherry channel ( Figure S4c). The best amber suppressing CHO lines were transiently transfected with either the read-through control pCMV-mCherry-GFP or the amber mutant containing a pCMV-mCherry-TAG-GFP reporter. Cells were then grown in no ncAA and analysed after four days on a BD LSR II Flow Cytometer ( Figure S4d). A line with a high GFP/mCherry ratio was chosen and scaled up to suspension culture.
In stage two of the cell-line production, we introduced the antibody HC and LC genes in five steps ( Figure S4e): insertion, amplification/selection, sorting, manual picking, and small-scale expression tests. First, we transfected trastuzumab HC and LC, each embedded on a separate plasmid ( Figure S4f), into the amber suppressing line selected above. We used DNA2.0 plasmids with optimized 5' and 3' untranslated regions (UTRs) around the HC and LC for enhanced expression. pEF1a-Heavychain-ires-GFP and pEF1a-Lightchain-ires-BFP, both on plasmid pD2531 were transfected into ambersuppressing CHO lines using TransIT-PRO transfection reagent. Second, we used the glutamine synthetase marker embedded in the plasmids to amplify the gene through selection with increasing levels of an inhibitor (methionine sulfoximin, MSX) of this essential enzyme. The cells were grown in FreeStyle CHO with no glutamine in the presence of 25 µM MSX for two weeks. They were allowed to recover for one week in media with no glutamine and no MSX. Third, in order to select for cells expressing the highest levels of HC and LC, we used the two distinct fluorescent protein reporters (GFP and BFP) downstream of each antibody chain to isolate the top 2.5% doublepositive lines using FACS ( Figure S4g). These reporters were preceded by internal ribosome entry sites for co-translational expression with the antibody chains.
Two additional rounds of MSX + FACS selection were performed. Populations from the first round were selected for two more weeks in suspension media with no glutamine and 100 uM MSX, recovered for a week, and the top 1% of double positives were sorted into 96 well plates. Cells were sorted at five cells per well into Ham's F12 with dialyzed FBS (Gibco) without glutamine. Wells with fast-growing colonies that appeared to grow from a single clone (based on number of colonies in the well) were examined by eye for GFP expression under a fluorescent microscope. Ten clones from the wild type antibody and ten from the HC118TAG antibody were scaled up and tested for expression in 2 mM CypK (synthesized as previously described). 3 The best expressing lines were pooled, and underwent another round of MSX amplification at 1 mM for two weeks in glutamine free suspension media. The batch was sorted for top 1% of double positive cells again as stated before. Ten clones from each batch were tested for expression in 2 mM CypK.
Finally, the highest expressing lines that had favorable growth properties such as growth rates and lack of clumping were tested for expression in suspension media with no glutamine, 1:100 anti-clumping agent (Gibco), and 1 mM BocK (E1610.0025, Bachem). Cultures were started at 10 6 cells/mL and run for 7 days. were set with a precursor tolerance of 10 ppm and a fragment ion mass tolerance of 0.15 Da. One missed enzyme cleavage was allowed and variable modifications for oxidized methionine, carbamidomethyl cysteine, pyroglutamic acid, phosphorylated serine, threonine and tyrosine, and cypK. MS/MS data were validated using the Scaffold programme (Proteome Software Inc., USA). 5 All data were additionally interrogated manually.
All SDS-PAGE gels were run using NuPage TM 4-12% polyacrylamide gels (BioRad) and MES buffer. were interpolated in this curve.
Fluorescence quantification was carried out by running samples diluted 1:100 with PBS on SDS-PAGE gels. Fluorescence intensity was measured by gel densitometry using Fiji.
Cytotoxicity assay. This assay was performed as previously reported. 8     SK-BR-3s is not due to a different effect of the toxin on the cells. The toxicity of the MMAE peptidomimetic toxin has previously reported to be cell-independent due to its membrane permeability, underscoring the requirement of attaching it to a targeting antibody to provide selectivity. 9 Furthermore, we observed that the toxicity of MMAE was 3-4 orders of magnitude higher than the toxin with the linker (tetrazine-vcMMAE, figures 3c and 3d in the main text). The ADC was first incubated for 5 days in 90% human serum and then added to cells (+ serum dataset). Trastuzuamb(MMAE) 2 freshly diluted in PBS (-serum dataset) was added in the same experiment for comparison. As described in the experimental section for the cytotoxicity assay, these compounds were incubated 5 days with cells and viability was assessed using CellTiter Glo 2.0. Higher concentrations of the ADC could not be used in these experiments in order to maintain the normal serum levels during the treatment for the viability essay. This is why the inflection point is not observed for MCF-7 in this experiment as opposed to the plot in Figure 3d in the main text. The error bars represent the standard deviation of biological triplicates.