Genetic Encoding of a Non-Canonical Amino Acid for the Generation of Antibody-Drug Conjugates Through a Fast Bioorthogonal Reaction

Antibody-drug conjugates (ADCs) used nowadays in clinical practice are mixtures of antibody molecules linked to a varying number of toxins at different positions. Preclinical studies have shown that the therapeutic index of these traditional ADCs can be improved by the site-specific linkage of toxins. However, current approaches to produce homogeneous ADCs have several limitations, such as low protein expression and slow reaction kinetics. In this protocol we describe how to set up an expression system to incorporate a cyclopropene derivative of lysine (CypK) into antibodies using genetic code expansion. This minimal bioorthogonal handle allows rapid conjugation of tetrazine derivatives through an inverse-demand Diels-Alder cycloaddition. The expression system here reported enables the facile production and purification of trastuzumab bearing CypK in each of the heavy chains. We explain how to link the antibody to the toxin monomethyl auristatin E and characterize the immunoconjugate by hydrophobic interaction chromatography and mass spectrometry. Finally, we describe assays to assess the stability in human serum of the dihydropyridazine linkage resulting from the conjugation and to test the selective cytotoxicity of the ADC for breast cancer cells with high levels of HER2 receptor.


Introduction
Antibody-drug conjugates (ADCs) combine the selectivity of biotherapeutics and the potency of small cytotoxic molecules. Most ADCs aim to decrease the side effects of traditional chemotherapy by targeting drugs that affect DNA or microtubule polymerization to cancer cells 1 . Firstgeneration ADCs approved by the Food and Drug Administration (FDA) rely on the modification of lysines and cysteines, which generates mixtures of molecules modified at different positions with decreased pharmacokinetic properties 2 . By contrast, site-specific conjugation of drugs to the antibodies can generate compounds with improved therapeutic indeces 3,4 . Seeking to address the challenge of producing homogeneous ADCs, several selective chemical and enzymatic modifications have been reported 1,5 . However, current methods can target only certain position on the antibody, suffer from low protein expression, provide linkers with low stability, or rely on slow and low-yielding reactions.
Incorporation of non-canonical amino acids (ncAA) through genetic code expansion enables the site-specific installation of a plethora of bioorthogonal reactive groups into proteins, potentially overcoming the limitations of other methods used to generate ADCs. Encoding ncAAs in response to a target (stop) codon relies on aminoacyl-tRNA synthetase/tRNA pairs that are orthogonal to the endogenous pairs that incorporate canonical amino acids 6 . Several ncAAs have been incorporated into antibodies to generate ADCs. However, most suffer from various liabilities for applications in therapeutic drug conjugation. p-acetylphenylalanine (pAcF) 7,8 is not fully bioorthogonal, requires low pH (4.5) and long reaction times (> 60 h), while azides such as p-azidophenylalanine (pAzF) 7,9,10 , p-azidomethylphenylalanine (pAMF) 11 , and an azide derivative of lysine (AzK) 12,13 may be reduced in the cell 14 , and the copper used to catalyze Huisgen cycloadditions can induce oxidative damage 15 .
Although alternative ncAAs based on trans-cyclooctene (TCO), cyclooctyne (SCO) and bicyclo[6.1.0]nonene (BCN) have recently been encoded in an antibody for bio-imaging purposes, the expression system suffers from very low yields (0.5 mg/L) 16 . Moreover, cyclooctenes and cyclooctynes are large and hydrophobic handles that may increase the susceptibility of the ADC for aggregation -ADC payloads are traditionally hydrophobic and the physicochemical properties of the linker have been shown to greatly impact pharmacokinetics and therapeutic index 17 . By contrast, 1,3-disubstitued cyclopropenes are small reactive groups that should cause minimal alteration in the protein structure and physichochemical properties 18 . Cyclopropenes selectively and rapidly react with tetrazines via an inverse electron-demand Diels-Alder cycloaddition 19 . In this protocol we make use of a derivative of lysine (CypK, Figure 1b) bearing a methyl-cyclopropene that is less affected by steric hindrance than larger strained unsaturated cycles and has a reaction rate constant in the order of 1-30 M -1 s -1 in aqueous media 18,20 .
We recently reported how to incorporate CypK into antibodies to generate ADCs by reacting this minimal bioorthogonal handle with tetrazinebearing molecules 21 .
The aforementioned expression systems can produce the therapeutic anti-HER2 immunoglobulin 1 (IgG1) trastuzumab with CypK at similar levels to the wild type antibody. We selected the first position of the CH1 domain on the heavy chain to encode the ncAA (HC-118TAG). This is the most commonly modified site in ADCs 23 and is known as HC-118 (EU numbering) but has also been referred to as HC-121 (sequence position) and HC-114 (Kabat numbering) 24 . Since this position is conserved throughout all IgG1s, these expression systems should be amenable to most therapeutic antibodies.
We show trastuzumab(CypK) 2 can be easily purified by protein A followed by fast protein liquid chromatography with a hydrophobic interaction column (FPLC-HIC). Subsequently the antibody is covalently linked within 3 h to the microtubule polymerization inhibitor monomethyl auristatin E (MMAE), which is used in the FDA-approved ADC Adcetris. Here we use a benzyl-tetrazine derivative of MMAE (tetrazine-vcMMAE) with a linker comprising a glutarate spacer and a valine-citrulline protease-labile component followed by a p-aminobenzylalcohol self-immolative unit; this linker is cleaved by Cathepsin B in the lysosome upon internalization of the ADC resulting in the traceless release of the toxin 25 . In order to show the broad scope of the reaction, the antibody is also linked to the fluorophore tetramethylrhodamine (TAMRA). We explain how to verify the identity of the conjugate by liquid chromatography coupled to mass spectrometry (LC-MS) and to calculate the drug-to-antibody ratio (DAR) using high performance liquid chromatography with a hydrophobic interaction column (HPLC-HIC).
As part of the characterization of the antibody performance, we describe how to test the stability of the dihydropyridazine linkage in human serum. This parameter is more easily assessed in trastuzumab-TAMRA because it can be quantified by a simple ELISA and the interpretation of the results is not complicated by the protease labile component of trastuzumab(MMAE) 2 . Finally, the selectivity and potency of trastuzumab(MMAE) 2 is assessed by comparing the cytoxicity of the ADC across cell lines expressing different levels of HER2. This assay also provides a functional proof of the ADC stability when performed after incubating the immunoconjugate in human serum.  21 to 2.5 mL with reduced serum medium. In a separate tube, dilute 135 µL of transfection reagent to 2.5 mL with reduced serum medium. 5. Five minutes after preparing the solutions, mix the plasmids and the transfection reagent solution and incubate for 20 min to allow the formation of complexes between the DNA and the transfection reagent. 6. In the meantime, centrifuge 125 million cells at the target density for 5 min at 500 x g, resuspend with the expression medium containing CypK and add the DNA-transfection reagent mixture. CypK can be purchased or synthesized as reported previously 18 . 7. After incubating cells for 20 h, add 250 µL of transfection reagent enhancers included in the kit. 8. Harvest antibodies from the supernatant 6-7 days after addition of CypK (no change of medium is required during expression).

Produce and Characterize the Antibody
*Alternatively, to obtain higher and more consistent yields, a stable cell-line can be generated as described in Oller-Salvia et al. 2018 21 . In this case, trastuzumab(CypK) 2 is expressed simply by addition of CypK 5 mM in the expression medium.

Representative Results
The reported transient expression system (Figure 1a) yields 22 ± 2 mg of trastuzumab(CypK) 2 per liter of culture medium, which represents 2/3 of the wild type antibody produced under the same conditions (Figure 1c). The stable cell line can increase this yield up to 31 ± 2 mg/L 21 .
Trastuzumab(CypK) 2 can be conjugated with tetrazine-vcMMAE, which yields quasi homogeneous trastuzumab(MMAE) 2 within 3 h at 25 °C (Figure 2). The high hydrophobicity of this cytotoxin requires addition of 10% acetonitrile when 5 or more molar equivalents of toxin per CypK are used. Alternatively, the cycloaddition is also completed within 20 h using 2 equivalents of tetrazine-vcMMAE without acetonitrile (Figure 2c). Trastuzumab(CypK) 2 reacts with tetrazine-TAMRA within 2 h at 25 °C and 3-6 h are required when the temperature is decreased to 4 °C ( Figure  3c).
The expected DAR for trastuzumab(MMAE) 2 measured by HPLC-HIC is 1.9 (Figure 2b). The peak initially observed in the chromatogram at 8.0 min represents the unconjugated antibody (DAR 0) and should have completely disappeared when the reaction is completed. The species with DAR 1 elutes at 9.1-9.6 min and should have an area < 10% after 3 h; and the target product with DAR 2 has a retention time of 10.5-11.0 with an expected area > 90%. The mobility shift and fluorescence in SDS-PAGE gels confirms the incorporation of TAMRA (Figure 3b) and the identity of the immunoconjugates is verified by LC-MS (Figure 2d-e and Figure 3d).
Incubation of trastuzumab(TAMRA) 2 for 5 days in human serum and subsequent analysis by ELISA confirms that the payload remains attached to the antibody (Figure 4b). Regarding the cytotoxicity assay, trastuzumab(MMAE) 2 shows high potency in SK-BR-3 (HER2 high) breast cancer cells, with a half maximal effective concentration (EC 50 ) of 55 ± 10 pM (Figure 4d). Trastuzumab(MMAE) 2 maintains the cytotoxicity after 5 days of incubation in human serum (Figure 4c). Conversely, when the ADC is assayed on MCF-7 (HER2 low) the EC 50 is 200-fold lower (Figure 4d). The wild type antibody, trastuzumab(CypK) 2 and tetrazine-vcMMAE show extremely low toxicity (Figures 4d and 4e), whereas MMAE displays high non-selective cytotoxicity in both cell lines (Figure 4e).

Discussion
The transient expression procedure to produce trastuzumab(CypK) 2 described in this protocol is simple and allows for high modularity. The yields obtained are within the ones expected in an academic setting 27 and stable cell lines can be generated to further boost the production yield 21 . During expression, concentrations of CypK lower than 5 mM may result into lower ncAA incorporation, and higher amounts may affect cell growth and decrease antibody yields. CypK as a free amino acid has low water solubility, thus it should be first dissolved at 100 mM in 0.1 M NaOH and then added to the culture medium. After diluting CypK in the medium and before adding it to cells, it is critical to neutralize the medium with HCl and filter to sterilize. Subsequently, using the transfection reagents specified in this protocol and following the incubation times recommended by the manufacturer is important for a high-yielding expression. For further details on transient expression of human antibodies, the reader is referred to other published protocols 31,32 . When the antibody is purified, a high excess of protein A resin is required as indicated to ensure full antibody pull down from the supernatant. In order to prevent the precipitation of trastuzumab during elution, it is recommendable to use a solution with high buffering capacity, dilute immediately with PBS and exchange the buffer avoiding excessive concentration. Always keep the antibody <5 mg/mL. The conjugation of trastuzumab(CypK) 2 with tetrazine-vcMMAE is faster than most reactions reported with other bioorthogonal handles for ADCs. Moreover, this cycladdition occurs under very mild conditions: room temperature or lower and physiological pH. It is important to dilute the DMSO stock solutions of the reactants with acetonitrile prior to addition of PBS and the antibody; otherwise the tetrazine derivatives will precipitate. Acetonitrile is required only due to the high hydrophobicity of MMAE and TAMRA, but less hydrophobic molecules may not need the addition of a co-solvent. Alternatively, tetrazine-vcMMAE can be conjugated without acetonitrile and only 2 molar equivalents of tetrazine-vcMMAE within 20 h. This little amount of toxin could involve a substantial decrease in the manufacturing cost of ADC when compared to current ncAA-based technologies. Trastuzumab(CypK) 2 is fully reactive for at least 4 months when preserved at 4 °C.
HPLC-HIC enables an accurate determination of DAR since MMAE is highly hydrophobic and provides an excellent resolution of the peaks corresponding to antibody conjugates with 0, 1 and 2 toxins. Unreacted tetrazine-vcMMAE elutes around 13.7 min and is detected at 280 nm. This technique requires a starting material with high purity. Moreover, it is not recommendable to quench the reaction with other tetrazine-reactive molecules such as BCN-OH since they can alter the retention times and the shape of the peaks. It is essential that the salt concentration of the samples matches the one in the mobile phase at the start of the gradient in order to obtain a good separation, especially if more than 10-20 µL are injected.
Regarding the LC-MS analysis, deglycosylation of the antibody samples is required to obtain a single peak upon deconvolution of the raw spectrum. The accuracy of the total antibody and ADC masses may vary depending on the callibration of the instrument. Hence, in order to calculate the mass for the modification, subtract the mass obtained for the unmodified antibodies from the one obtained for the ADC. Modern high-resolution mass spectrometers should provide a relative error below 1:10000. Although LC-MS can also be used to calculate the ratio between the different species, this value is usually an overestimation because the modification may affect the ionization capacity of the species generated and low amounts of impurities may not be detected.
The stability of the linker in ADCs is critical because the premature release of the drug results in higher toxicity and lower efficacy; the free cytotoxin damages healthy tissues and the naked antibody competes with the armed one for the target binding sites on diseased cells. A release below 5%, which is within the variability of the stability assay, should be expected.
Finally, the selectivity of an ADC targeting HER2 such as trastuzumab(MMAE) 2 can be assessed by comparing the cytotoxicity in SK-BR-3 cells (HER2 high) and MCF-7 cells (HER2 low) since the latter express 15-fold less HER2 receptors than the former 28 . The immunoconjugate is expected to result into a cell viability at least 2 orders of magnitude lower in SK-BR-3 when compared to MCF-7. The EC 50 in SK-BR-3 should be in the two-digit nanomolar range reflecting the high potency of this ADC 29,30 . The unmodified antibody, either trastuzumab(MMAE) 2 or trastuzumab, should show no toxicity in this assay. Tetrazine-vcMMAE should have an effect 3 orders of magnitude lower than the ADC since the linker removes the activity of the peptidomimetic toxin. Conversely, because MMAE is able to permeate the cell membrane