Production of antibodies and antibody fragments containing non-natural amino acids in Escherichia coli

ABSTRACT

Therapeutic bioconjugates are emerging as an essential tool to combat human disease. Site-specific conjugation technologies are widely recognized as the optimal approach for producing homogeneous drug products. Non-natural amino acid (nnAA) incorporation allows the introduction of bioconjugation handles at genetically defined locations. Escherichia coli (E. coli) is a facile host for therapeutic nnAA protein synthesis because it can stably replicate plasmids encoding genes for product and nnAA incorporation. Here, we demonstrate that by engineering E. coli to incorporate high levels of nnAAs, it is feasible to produce nnAA-containing antibody fragments and full-length immunoglobulin Gs (IgGs) in the cytoplasm of E. coli. Using high-density fermentation, it was possible to produce both of these types of molecules with site-specifically incorporated nnAAs at titers > 1 g/L. We anticipate this strategy will help simplify the production and manufacture of promising antibody therapeutics.

KEYWORDS:

• Antibody–drug conjugate
• bioconjugation
• e. coli expression
• non-natural amino acid
• site-specific
• therapeutic protein

Introduction

Therapeutic proteins have unprecedented potential to improve human health; the diversity and specificity in protein functions impart the ability to treat a variety of conditions such as cancers, autoimmune diseases, metabolic disorders, and infectious diseases. Antibodies and antibody fragments have emerged as a particularly exciting therapeutic modality due to their high specificity for a given target, which is thought to reduce the adverse effects associated with treatment.¹ Because of this, the market for monoclonal antibody (mAb)-derived drug products has grown rapidly in the past few decades; the FDA recently approved the 100^th mAb product for use in the United States, and nearly one-fifth of its yearly new drug approvals belong to this class.² mAbs and their derivatives have been especially impactful in the field of cancer treatment, with dozens of products currently approved and many more in clinical development.³

It has long been recognized that the therapeutic properties of mAbs can be enhanced through covalent fusion to small molecules. For example, the antitumor properties of these drugs can be significantly improved by tethering a cytotoxic small molecule to the protein to form a derivative called an antibody–drug conjugate (ADC).^4,5 Traditionally, ADCs were generated using reactive side groups on naturally occurring amino acids such as the thiol of cysteine or the ε-amine of lysine as the bioconjugation sites.^6,7 Cysteine conjugations have several drawbacks, however. Cysteine conjugation can result in heterogeneous ADC populations with varying drug-to-antibody ratio (DARs). This heterogeneity can affect the stability, pharmacokinetics, and efficacy of the ADC. The heterogeneity also makes characterization and optimization more challenging. A more advanced technique referred to as THIOMAB utilizes engineered cysteines that are introduced as unpaired cysteines on the surface of the protein for thiol conjugation.⁸ Surface-exposed cysteines can lead to protein dimerization or be capped by glutathione, requiring reduction prior to conjugation. Also, cysteine conjugations typically require prior reduction, increasing the number of process steps.⁷ In an alternative approach, conjugation is performed by reacting N-hydroxysuccinimide (NHS) esters with surface-exposed lysines.⁷ Amine-based conjugation technology similarly suffers due to partial conjugation of surface-exposed lysines that results in heterogeneous mixtures. This approach has been used to improve the therapeutic properties compared to mAbs in their native form; however, traditional stochastic conjugation leads to heterogeneities in the number and location of modifications, which can affect the biophysical properties of the protein and result in undesirable outcomes such as compromised activity, stability and pharmacokinetics.^5,6,9

Site-specific protein bioconjugation approaches allow researchers to build highly tunable protein constructs capable of improving the specificity and efficacy of targeted drug delivery while minimizing negative effects on therapeutic activity.^6,10 Producing an ADC this way has the potential to widen the therapeutic window due to the selection of the single species that optimizes physicochemical properties, efficacy, and safety. Non-natural amino acid (nnAA) incorporation is a particularly exciting and well-studied bioconjugation method based on the site-specific genetic insertion of amino acids modified with bio-orthogonal functional groups. nnAAs are most commonly incorporated co-translationally using an engineered, orthogonal transfer RNA (tRNA) that recognizes the amber stop codon, ochre stop codon, or other nonsense codons, as well as an aminoacyl-tRNA synthetase (RS) that catalyzes the aminoacylation of the orthogonal tRNA with the nnAA.¹¹ The charged tRNA then inserts the nnAA at a genetically specific location(s) during protein translation. However, nnAA incorporation has been difficult to implement in traditional mAb production hosts such as mammalian cells due to limited amber suppression efficiency, which adversely impacts production titers and product homogeneity.^11,12

The production of nnAA-containing antibodies in stable Chinese hamster ovary (CHO) cell lines has recently been reported.¹² However, incorporation of nnAAs into antibodies produced in CHO is limited by low amber suppression efficiencies of ~ 40%, which results in low nnAA-containing IgG titers. In addition, truncated IgGs (resulting from translation termination at the TAG codon) were co-purified with the nnAA-IgGs, necessitating additional chromatography steps to remove truncated product. In an earlier report, the co-purification of truncated and nnAA-containing species was prevented by placing the nnAA incorporation site closer to the N-terminus so that truncated species would not bind to protein A resin during downstream purification.¹¹ This approach, however, is limiting because nnAAs are constrained to sites within the Fab region of the IgG. On the other hand, organisms more amenable to robust nnAA incorporation, such as the common prokaryotic protein expression host Escherichia coli (E. coli), typically fail to produce quality mAbs at commercially relevant titers.^13–17

In this work, we describe a method for incorporating nnAAs into therapeutically relevant mAbs and mAb fragments expressed in the cytoplasm of an E. coli strain engineered for cytoplasmic disulfide bond production. In an effort to maximize conjugation kinetics, an azide-containing nnAA optimized for bioconjugation, para-azidomethyl-l-phenylalanine (pAMF), was used.^18,19 This nnAA has been used for strain-promoted azide-alkyne cycloaddition-based (SPAAC) copper-free click conjugation chemistry that enabled the production of homogeneous ADCs with DAR values approaching complete conjugation.¹⁸ E. coli has unique features compared to most other production hosts for producing nnAA-containing proteins at high titers, rates, and yields. Specifically, the nnAA tRNA and RS can be encoded on a multicopy plasmid and expressed from high strength promoters, which facilitates nnAA incorporation at efficiencies approaching that of canonical amino acid incorporation. Nevertheless, the production of mAbs and mAb fragments in E. coli has proven challenging thus far, limiting its usefulness as a production platform for ADCs and related biologics.

While recombinant proteins are usually secreted into the culture media in other hosts (e.g., mammalian or fungal systems),^12,20 in E. coli, they can be expressed in the cytoplasm or secreted to the periplasmic space.¹³ Each cellular compartment has unique properties and the expression location can be chosen based on the protein to be produced. The cytoplasm is naturally a reducing environment and is ideal for proteins lacking disulfide bonds. Proteins that require disulfide bond formation are often produced as inclusion bodies that can be resolubilized and refolded into functional forms, or secreted into the oxidizing periplasmic compartment in their soluble and active forms.^15,21 Both methods for oxidized protein synthesis are challenging, and titers are often low (<100 mg/L). Furthermore, protein refolding requires extensive process optimization and a series of buffer exchanges that can be difficult to scale, resulting in low yields and time-consuming processes.^22–24

Periplasmic expression in E. coli is complicated by the need to secrete proteins across the cytoplasmic membrane. Efficient secretion typically requires screening multiple parameters, such as induction strength, expression conditions, and signal peptide sequence. This makes it especially challenging for periplasmic expression of hetero-multimeric proteins such as IgGs that require secretion of two separate proteins in the correct ratio for optimal hetero-tetramer assembly.^15,25 Secreting proteins into the periplasm imparts a significant metabolic burden and can negatively affect cell health. Additionally, the periplasm lacks nucleotide triphosphates necessary for the function of chaperones systems such as GroEL-ES or the HSP70 system, which could assist the folding of recombinant proteins.¹³ Finally, periplasmic expression is limited by the smaller volume of the periplasmic compartment in relation to the cytoplasmic space, restricting the titer of periplasm-targeted proteins.¹⁴ Due to these challenges with periplasmic expression, we focused on the synthesis of mAbs and mAb fragments in the oxidizing cytoplasm of an E. coli strain that was redox optimized for high-level expression of antibody light chain (LC) lacking nnAAs in the cytoplasm.²⁶ Production of this LC is relatively straightforward because it requires overexpression of only a single protein in the proper redox environment. Notably, it was also shown that the same LC will express poorly in a strain background exhibiting a wild type, reducing cytoplasm. High-level expression of IgGs is substantially more challenging because it requires overexpressing multiple proteins at the correct ratio for optimal IgG assembly and titers. While cytoplasmic expression of IgGs has been achieved previously in a different strain with an oxidizing cytoplasm, the titers were too low for commercial antibody production.¹³ This prompted us to test whether this newly engineered E. coli strain could achieve high level soluble expression of other antibody therapeutic scaffolds beyond LCs.²⁶

In this report, we demonstrate soluble expression of a single-chain variable fragment (scFv), an IgG antigen-binding (Fab) fragment, and a full-length IgG mAb complex at high levels in an oxidizing E. coli cytoplasm. Of these scaffold proteins, a LC from our previous report²⁶ and the full-length IgG were selected for nnAA incorporation with pAMF.¹⁸ Both products were expressed in a bioreactor using high-density fermentation in cells transformed with plasmids bearing the genes for the nnAA tRNA and RS and the product of interest,^27,28 resulting in high yields of >1 g/L and pAMF incorporation efficiency > 99% for each protein. The nnAA-containing LC was used as a pre-fabricated LC (PFLC) to produce an IgG using cell-free protein synthesis (CFPS). This IgG and a nnAA-containing anti-CD74 IgG wholly synthesized in E. coli were conjugated with cytotoxic warheads to produce ADCs, both of which showed potent cell killing in vitro.

This study demonstrates that it is possible to produce a variety of key mAb-derived therapeutic molecules in the cytoplasm of an E. coli strain that has been engineered to have the appropriate redox potential. We demonstrate that this strain produces antibody LCs and full-length IgG complexes at the highest ever reported titers, enabling the creation of nnAA variants from these two protein scaffolds, both of which expressed at high titers in high-density fermentation. Highly efficient homogenous bioconjugation was demonstrated with both pAMF-containing products. This result demonstrates the utility of E. coli cytoplasmic expression as a platform for making protein bioconjugates for oncology. The use of E. coli for production should accelerate the design and development of these types of molecules, providing the industry with a new tool for creating targeted cancer therapeutics.

Results

Shake flask expression screen of antibody scaffolds

Previously, we demonstrated that our E. coli strain SBDG419, which has been engineered to have an oxidizing cytoplasm, could express a variety of LCs that could be incorporated into antibodies produced in CFPS reactions.²⁶ The high titer and productivity observed in the SBDG419 strain suggested that it could be a promising platform for producing antibody fragments and full-length antibodies potentially containing nnAAs. Therefore, an expression screen of several therapeutically relevant proteins without nnAAs was performed in this strain to find potential scaffolds for nnAA incorporation, including a scFv,²⁹ an anti-CD3 antibody Fab,³⁰ and a full IgG complex.

Initially, the expression of a trastuzumab scFv was tested, as scFvs have successfully been produced in the E. coli cytoplasm previously,^17,31 and they are relatively less complex proteins. scFvs consist of a single polypeptide with two intrachain disulfide bonds. An expression plasmid was constructed in which the V_H domain was fused to the N-terminus of the V_L domain via a 5X GGGGS amino acid linker (V_H-linker-V_L) and cloned into a high copy (pUC origin) pJ411 plasmid. Expression from this plasmid was tested in SBDG419 in shake flasks. After cell harvest and lysis, sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis of soluble lysates revealed that the trastuzumab V_H-linker-V_L scFv could successfully be expressed in the cytoplasm of E. coli SBDG419 (Figure 1). The scFv was purified using protein L resin, resulting in a purified protein titer of 40 mg/L of shake flask culture (Table 1). Analysis of the protein by SDS-PAGE (Figure 1A) and intact liquid chromatography-mass spectrometry (LC-MS) analysis (Figure S1A) under reducing and non-reducing conditions confirmed the identity of the proper scFv and that the purified protein had two pairs of disulfides formed.

Figure 1. Expression of therapeutically relevant antibody fragments and full-length IgGs in E. coli strain SBDG419.

(a–c) show SDS-PAGE gels of soluble lysates and purified proteins from each test expression in shake flasks. MW = molecular weight of protein standards in ladder. (a) Trastuzumab scFv (lane 1: soluble lysate (reducing), lane 2: purified (reducing), lane 3: purified (non-reducing). (b) anti-CD3 Fab (lane 1: soluble lysate (non-reducing), lane 2: purified (non-reducing), lane 3: purified (reducing)). (c) anti-CD74 IgG (lane 1: soluble lysate (reducing), lane 2: purified (non-reducing), lane 3: purified (reducing)).

Table 1. Summary of titers and amber suppression efficiency of pAMF incorporation into nnAA-LC and nnAA-IgG produced in E. coli. For titer calculation method descriptions, refer to the materials and methods section within the supplementary information. Amber suppression efficiencies were calculated by dividing the titer (*from bioreactors, **from shake flasks) of nnAA-containing proteins by the titer of the proteins without nnAA incorporation.

Protein	Shake flask titer (g/L)	Bioreactor titer (g/L)	Titer calculation method	Amber suppression efficiency
Trastuzumab scFv	0.04	–	Protein L capture	–
Anti-CD3 Fab	0.03	–	KappaSelect CH1-XL capture	–
Trastuzumab LC	0.72^Citation26	6.2^Citation26	KappaXL HPLC	–
Anti-CD74 IgG	0.23	–	Phytip capture	–
nnAA-LC	0.35	3.9	KappaXL HPLC	63%*
nnAA-IgG	0.12	1.1	Phytip capture	52%**

To express the anti-CD3 Fab, the coding sequences for the LC and V_H and C_H1 domains of the heavy chain (HC) components of an anti-CD3 antibody were cloned as a single operon in the order HC-LC into pJ411 with a strong ribosome-binding site preceding both open reading frames. The Fab was tested using the protocol described above for scFv expression. Analysis of soluble cell lysates from cultures by non-reducing SDS-PAGE revealed successful expression and assembly of the Fab (Figure 1(b)). The Fab was purified using CH1-XL resin, and eluates were analyzed by SDS-PAGE (Figure 1(b)) and intact LC-MS (Figure S1B), which confirmed proper assembly of the protein. The captured protein titer for the Fab produced in shake flasks was similar to the scFv, at 30 mg/L (Table 1).

The feasibility of expressing antibody LCs (Table 1 and ref. 26), an scFv, and a Fab in SBDG419 suggested that this strain could possibly serve as a platform to produce full IgG complexes. Successful cytoplasmic expression of IgGs in E. coli has previously been described, but titers are typically very low (<20 mg/L in shake flasks).^13,16,25,32 To produce IgGs in SBDG419, a pJ411-based expression plasmid was designed that contained the HC and LC sequences of an anti-CD74 IgG in a bicistronic operon, with the HC preceding the LC. To ensure sufficient expression of the LC, an additional ribosome-binding site 5’ to the LC start codon was included. Expression of the IgG was tested in shake flasks according to the same protocol used for the LC, scFv, and Fab proteins. Cell lysates were assayed for IgG expression using a small-scale protein A-based purification assay (Phytip purification), which showed that the IgG was expressed at titers of 230 mg/L of shake flask culture volume (Figure 1(c)). Analysis of IgG assembly by capillary electrophoresis (Caliper assay) showed that the IgG was well assembled, with 70% fully assembled species (with complete intermolecular disulfide formation) present after crude purification (Figure S2), demonstrating that antibodies could feasibly be produced in the cytoplasm of E. coli strain SBDG419.

The successful production of antibody-based therapeutic molecules using SBDG419 suggests that this strain may be a suitable host for the incorporation of nnAAs into several of these proteins. While the scFv and Fab titers were lower, indicating the need for more optimization with these modalities, the LC and IgG titers were both >100 mg/L, so these proteins were subsequently chosen as scaffolds for nnAA incorporation.

Production of a nnAA-containing LC (nnAA-LC) for high DAR ADC synthesis

We have previously shown that pre-fabrication of IgG LCs leads to a ~ 2X boost in titers of antibodies produced in CFPS reactions.²⁶ Therefore, by incorporating a nnAA into a LC, ADCs with greater DAR could be generated at higher titers than in a CFPS reaction in which both the LC and HC were co-expressed. High DAR ADCs allow for more concentrated delivery of active warhead to an intended tissue without increasing therapeutic dosage, making an antibody intermediate with a high number of site-specific conjugation handles an attractive molecular scaffold for next-generation ADCs.

In order to incorporate nnAAs into recombinantly-expressed proteins in E. coli via amber suppression, three genetic elements are required: (1) a coding sequence for a protein of interest containing a TAG amber codon at the desired nnAA incorporation site, (2) an orthogonal RS that will recognize the nnAA of interest and load it onto a cognate tRNA,¹⁸ and (3) a cognate amber suppressor tRNA (AS tRNA) not recognized by any of the native RSs that will be charged with the nnAA and associate with the UAG codon of the mRNA of interest,^27,33 causing the nnAA to be incorporated into a protein as it is translated by the ribosome. To achieve amber suppression in our SBDG419 strain, a second plasmid (deemed the RS plasmid) was designed that encoded pAMF RS and its cognate AS tRNA. Initially, their DNA was cloned into a medium copy p15a plasmid behind an inducible T7 promoter. On the LC plasmid (Figure 2(a)), the coding sequence for trastuzumab LC was mutated to encode a single TAG amber codon replacing residue K42, a surface-exposed residue that would be accessible to downstream conjugation of a cytotoxic payload. Expression of the LC containing one copy of pAMF in place of K42 (nnAA-LC) was tested in the same way as the native proteins described above, with several modifications. Strains were grown in the presence of both kanamycin and carbenicillin to ensure retention of both plasmids, and, at the time of induction, the nnAA pAMF was added to the culture.

Figure 2. Expression and activity of nnAA-LC, nnAA-IgG-X and DAR8 ADC-X.

(a) Depiction of two plasmids used for nnAA incorporation in E. coli strain SBDG419 or SBDG175. Left: pJ411 product plasmid bearing the sequence of the protein of interest (POI) behind a T7 promoter (T7 pr.) and followed by a T7 terminator (T7 term.). The product plasmid is high copy (pUC origin of replication) and bears a kanamycin resistance gene (KanR). Right: pJ434 RS plasmid bearing the sequence of the pAMF RS behind a T7 promoter (T7 pr.) and constitutive Pc0 promoter (Pc0 pr.). The pAMF RS sequence is followed by one copy of the AS tRNA and a T7 terminator (T7 term.). The RS plasmid is a medium copy (p15a origin of replication) and bears an ampicillin resistance gene (AmpR). (b) SDS-PAGE analysis of lysates expressing nnAA-LC using different promoter combinations driving pAMF RS and AS tRNA expression. MW = molecular weight of protein standards in ladder. Lane 1: Trastuzumab light chain (Tras. LC) control (no nnAAs incorporated), lane 2: nnAA-LC expression using RS plasmid with a T7pr., no pAMF nnAA added, lane 3: nnAA-LC expression using RS plasmid with a T7pr., pAMF added, lane 4: nnAA-LC expression using RS plasmid with a combined T7pr. and Pc0 promoter, no pAMF added, lane 5: nnAA-LC expression using RS plasmid with a combined T7pr. and Pc0 promoter, pAMF added. The arrow indicates the LC band. (c) Cell killing activity of ADC-X against a cell line expressing its target antigen (left panel, “Antigen (+)”), or with no expression of its target antigen (right panel, “Antigen (-)”). (d) Table describing titer and quality attributes of nnAA-LC and the DAR8 ADC-X. Purity % was assessed by analytical SEC, assembly % was calculated with denaturing Caliper capillary electrophoresis assay, and DAR was calculated by intact LC-MS analysis.

After harvest and lysis of the test expressions, the soluble lysate was analyzed for the presence of nnAA-LC via SDS-PAGE, where the presence of a band at ~ 25 kDa suggested the expression of the protein in shake flasks (Figure 2(b)). The intensity of the LC band (Figure 2(b), lane 3), however, was much less intense than that of the trastuzumab LC without nnAAs (Figure 2(b), lane 1). We hypothesized that pAMF RS mRNA and AS tRNA levels could be boosted and nnAA-LC titers improved by adding a constitutive promoter to the RS plasmid that was active before induction.³⁴ The pAMF RS mRNA and AS tRNA were therefore cloned again into a medium copy plasmid, but with both a T7 promoter followed by a constitutive Pc0 promoter (Figure 2(a)). Test expression of the nnAA-LC in the presence of RS plasmid with a Pc0 promoter resulted in higher titers of nnAA-LC as revealed by SDS-PAGE analysis of the lysate (Figure 2(b)). This was further confirmed by purification of nnAA-LC via small scale Phytip purification using tips packed with CaptureSelect KappaXL resin. The elution fractions were analyzed by SDS-PAGE, revealing an isolated band corresponding to nnAA-LC that shifted slightly higher upon reduction, suggesting that nnAA-LC exhibits proper disulfide formation (Figure S3A). The incorporation of one pAMF residue was confirmed by intact LC-MS (Figure S3B): the observed mass of the purified nnAA-LC matched the expected mass, and conjugation with a small molecule dibenzylcyclooctyne (DBCO)-amine gave the expected mass shift corresponding to the addition of a single DBCO-amine. Analysis of the soluble lysate by KappaXL HPLC showed that nnAA-LC was expressed at 350 mg/L in shake flasks. This titer was approximately half of that observed for the equivalent LC without nnAAs reported previously, which would still likely result in gram per L titers during fermentation (Table 1).²⁶ Therefore, the production process was transferred to a high-density fermenter to verify process scalability for manufacturing.

Initially, nnAA-LC production feasibility was demonstrated in small scale (500 mL) fermenters using the fed-batch process developed for the trastuzumab LC²⁶ with the addition of 100 µg/mL carbenicillin in seed flasks and 2 mM pAMF at the time of induction. After nnAA-LC production was confirmed at small scale (data not shown), the process was scaled up to a 30 L bioreactor. The nnAA-LC from the soluble lysate of the 30 L fermentation was purified over protein L resin, resulting in a final volumetric titer of 3.87 g/L. To ensure the quality of the nnAA-LC protein when produced at scale, a test conjugation was performed with the DBCO-exatecan payload to be used in the final ADC and analyzed using LC-MS. The deconvoluted intact LC-MS spectra showed that 96% of the protein had been conjugated (Figure S3C), confirming that nnAA-LC produced via high-density fermentation will be a good candidate PFLC for making high DAR ADCs.

To test the feasibility of using nnAA-LC for CFPS of an ADC, the purified nnAA-LC was used as the PFLC in a CFPS reaction synthesizing the HC of an antibody (nnAA-IgG-X) containing three additional TAG amber codons (Figure 3). This CFPS reaction should result in a final antibody containing pAMF at 8 sites (3 in HC, 1 in LC). The nnAA-IgG-X produced from the CFPS reactions was purified over protein A resin, then polished by size-exclusion chromatography (SEC), after which analytical SEC analysis revealed that the protein had been purified to 97% homogeneity (Figures 2(d) and S4(a)).

Figure 3. Schematic of cell free protein synthesis-based production of nnAA-IgG-X and ADC-X using nnAA-LC as pre-fabricated light chain.

The nnAA-LC protein, produced in E. coli strain SBDG419, is purified and added as a reagent to a cell free protein synthesis reaction producing the heavy chain of nnAA-IgG-X. The nnAA-IgG-X is purified from the reaction and conjugated with a DBCO-exatecan payload to produce a homogenously-conjugated DAR8 ADC: ADC-X. Modified from reference 26.

Next, the nnAA-IgG-X was conjugated to a cytotoxic payload containing an exatecan linker payload connected to a DBCO moiety by a tumor-specific protease-cleavable linker, producing ADC-X. After overnight conjugation, the protein was cleaned up by SEC and analyzed using intact LC-MS. The final DAR of the conjugated ADC-X was 7.9 (Figure 2(d) and S4(b)), with payload conjugation at 98.8% of pAMF residues in the molecule. The ADC-X was then tested for cell killing activity in vitro. ADC-X displayed potent cell killing against an antigen-expressing cell line and no activity against a negative control cell line with no antigen expression (Figure 2(c)). This demonstrates that even with high drug loading (DAR8), ADC-X exhibited no off-target killing, which is a desirable safety feature for an ADC.

In summary, our results show that we developed a strain of E. coli with an oxidizing cytoplasm that is capable of producing the trastuzumab LC containing a single nnAA (nnAA-LC), which was accomplished through the use of a two-plasmid system, one for the expression of the product gene and one for the expression of the AS tRNA and RS. The nnAA-LC production was successfully scaled, resulting in high protein titers produced in a high-density fermentation at 30 L scale. Pre-fabrication of the nnAA-LC enabled the high titer production of a high DAR ADC in a CFPS expression system that exhibited potent tumor killing activity in vitro. The successful production of a nnAA-LC at high titers suggested that this strain could be used to express more complex nnAA-containing therapeutic proteins in the E. coli cytoplasm.

Production and characterization of a full length nnAA-containing IgG (nnAA-IgG)

Our successful demonstration of full IgG production at high titers in the cytoplasm of E. coli SBDG419 suggested that production of a nnAA-containing antibody may also be feasible in this strain. A nnAA-containing IgG would serve as a scaffold for an ADC that could be site-specifically labeled with a cytotoxic payload. To test the production of a nnAA-containing IgG in E. coli, a series of six expression plasmids derived from the anti-CD74 IgG expression construct expressed above were designed with a single codon at the F404 site replaced with a TAG amber codon for the incorporation of pAMF. To balance the expression levels of the HC and LC for optimal IgG assembly, the ribosome-binding sites (RBSs) of each protein coding sequence were individually mutated to vary the expression levels of each protein (Table S1). Each expression plasmid was co-transformed into SBDG419 with the RS plasmid used above (containing both a T7 and Pc0 promoter) for nnAA-LC production. Expression of the IgGs was then tested in shake flasks using the same protocol discussed above for nnAA-LC.

The expression of nnAA-IgGs was quantified by Phytip purification of the nnAA-containing anti-CD74 antibodies. The six expression constructs resulted in varying nnAA-IgG concentrations in soluble lysate, ranging from 0.1 to 0.72 g/L (Figure 4(a)). The assembly of each nnAA-IgG was also evaluated using a Caliper gel electrophoresis assay, with the assemblies of nnAA-IgGs varying from 40% to 77% (Figure 4(a)). From these analyses, construct H1L2 was chosen as the optimal expression plasmid for the nnAA-IgG because it resulted in both a high in-lysate nnAA-IgG titer of 0.6 g/L and a high assembly level of 60%. Given the volume of lysate produced, this construct resulted in nnAA-containing anti-CD74 antibody (nnAA-IgG) expression at titers of 120 mg/L of shake flask culture, approximately half of the titer observed during production of the anti-CD74 IgG without nnAAs (Table 1). To determine if this productivity could also be achieved during fermentation and to generate enough material for biological assays, the process was scaled in bioreactors.

Figure 4. Expression, purification, and biological activity of the anti-CD74 nnAA-IgG and its derivative ADC.

(a) Graph showing in-lysate nnAA-IgG concentrations (orange bars, left axis) and nnAA-IgG assembly (black squares, right axis) produced from each expression plasmid construct. In-lysate nnAA-IgG concentrations were quantified by Phytip purification assay and nnAA-IgG assembly was analyzed by a Caliper gel electrophoresis assay. Plasmids are named based on the RBS used for either the heavy chain or light chain, e.g., the H2L1 construct bears heavy chain RBS 2 (H2) and light chain RBS 1 (L1). The RBS sequences are numbered from strongest (RBS #1) to weakest (RBS #4) predicted strength. (b) SDS-PAGE gel of anti-CD74 nnAA-IgG samples. MW = molecular weight of protein standards in ladder. Lane 1: soluble lysate from 5 L fermentation to produce the anti-CD74 nnAA-IgG (reducing), lane 2: purified nnAA-IgG (non-reducing), and lane 3: purified nnAA-IgG (reducing). (c) Cell killing activity of the anti-CD74 ADC (blue squares) compared to the anti-CD74 nnAA-IgG (magenta triangles) observed in cell lines exhibiting CD74 antigen expression (left, “CD74 (+)”) or without CD74 antigen expression (right, “CD74 (-)”). (d) Table summarizing titer, product quality, and biological activity of the anti-CD74 nnAA-IgG and ADC. The % purity was assessed by analytical SEC, % assembly was determined by Caliper electrophoresis assay, DAR was calculated using intact LC-MS, and EC₅₀ cell killing activity was determined by cell viability measurements (see panel C). n.d. = not detected

The nnAA-IgG production was scaled in a 5 L high-density fermentation using the same fed-batch protocol that was successfully used to produce nnAA-LC. Initial analysis of the IgG concentration in lysates generated from the fermentation revealed that 1.1 g/L of nnAA-IgG had been produced. The nnAA-IgG was captured from the soluble lysate over protein A resin that included a 1-hour wash step in the presence of 2 mM oxidized glutathione to promote disulfide bond formation.³⁵ After a size-exclusion polishing step, the quality of the nnAA-IgG was analyzed by SDS-PAGE (Figure 4(b)), analytical SEC, and Caliper assembly assay. The presence of some partially assembled species could be observed in the polished protein sample. However, analytical SEC analysis revealed that the protein was of high purity at 97% monomer (Figures 4(d) and S5(a)), and the Caliper assay showed that interchain disulfides had properly formed, as the protein was 96% assembled (Figures 4(d) and S5(b)).

Next, the nnAA-IgG was conjugated with a cytotoxic maytansinoid linker payload.³⁶ The conjugation reaction was then analyzed by intact protein LC-MS. The deconvoluted spectra revealed that the reaction had gone to completion, as observed by an appropriate mass shift of the conjugated molecule compared to the unconjugated parent (Figure S6). The calculated DAR of the conjugated species was 2.0, confirming that a single pAMF nnAA had successfully been incorporated into each HC within the IgG (Figure S6). The conjugated nnAA-IgG bearing the maytansinoid payload was cleaned via SEC to remove free drug-linker prior to assaying its cell killing activity.

The nnAA-IgG and its derivative ADC were tested for cell killing activity against cells with or without CD74 expression. While the parent nnAA-IgG showed no cell killing activity against CD74 (+) cells, the ADC showed potent cell killing activity against the same cell line, with an EC₅₀ of 0.54 nM. Neither the nnAA-IgG or the ADC had any observable cell killing activity against a CD74 (-) cell line, demonstrating targeted cell killing dependent on CD74 expression. This result confirms that cytoplasmic expression of nnAA-IgGs in E. coli is a feasible method to produce ADCs that have potent biological activity.

Discussion

In previous work, we described an engineered strain of E. coli, SBDG419, optimized for high-level cytoplasmic expression of LC.²⁶ In this report, we show this strain is capable of expressing other disulfide containing antibody-derived therapeutic proteins in the cytoplasm, including a scFv, a Fab fragment, and a full-length IgG. E. coli strain SBDG419 was able to achieve high titers for the proteins tested in shake flasks. Furthermore, proper disulfide formation and complex assembly, both crucial attributes for biological activity, were confirmed. While others have reported the successful production of antibodies and antibody fragments in the E. coli cytoplasm, the reported titers are typically low (<200 mg/L).^13,16 Strain SBDG419 is similar to strains used in previous reports in which the cytoplasmic glutathione reductase (gor) and thiodoxin reductase (trxB) are knocked out and the disulfide bond isomerase, dsbC, is expressed in the cytoplasm.³⁷ However, SBDG419 was constructed with an additional knock out to the thioredoxin trxA, which was previously shown to result in higher yields of cytoplasmically-expressed proteins that contain disulfide bonds.^26,38 Removal of trxA, in combination with the use of high copy expression plasmids, was key to achieving the high titers of antibodies and antibody fragments demonstrated herein. Furthermore, the modifications made to this strain may facilitate E. coli in producing additional disulfide containing molecules, such as cytokines, hormones, and enzymes.

The incorporation of nnAAs into therapeutic protein scaffolds provides a robust mechanism for their site-specific covalent modification with cytotoxic payloads or half-life extending PEG molecules.¹¹ In particular, nnAAs containing azide moieties, such as pAMF, allow for the use of copper-free SPAAC conjugation chemistry, providing bio-compatible conjugation handles that are not cross-reactive with native amino acids.¹⁸ This technology allows the production of bioconjugates with > 95% homogeneity in conjugation sites. However, nnAA incorporation is challenging when using typical biologics production platforms such as mammalian cells.¹²

Herein, we developed a platform for efficient nnAA incorporation into mAb-based protein scaffolds using our engineered E. coli strain SBDG419. To achieve efficient amber suppression, an RS and tRNA pair were expressed from a medium copy plasmid. Using this approach, there was minimal misincorporation or truncation at the site of nnAA incorporation for both proteins tested, and the disulfide bond formation of nnAA-containing proteins was comparable to the starting IgG and LC. Notably, strain SBDG419 was able to express nnAA-containing proteins at high titers in high-density fermentation, suggesting that this method could be useful for other disulfide containing therapeutic candidates requiring site-specific conjugation handles.

Previous work showed that pre-fabricated antibody LCs could be used to increase titers of IgGs produced in CFPS.²⁶ In this work, we demonstrated the production of an antibody LC containing a nnAA at high titers of 3.87 g/L from high density fermentation. This is roughly half the titer for the LC protein, suggesting that amber suppression is a bottleneck when expressing the nnAA-LC. In demonstration of the higher DARs that nnAA-LC enables, it was used to produce a nnAA-IgG containing 8 pAMF residues in a CFPS reaction. Generation of homogenous site-specifically labeled high DAR IgGs is challenging, but pre-fabrication of a nnAA-LC enables us to achieve higher drug loading on an ADC with the 2-fold increase in CFPS titer from using PFLC. The successful scale up of nnAA-LC production suggests that the fermentation process developed for the LC could be applied to the production of other proteins containing nnAAs in E. coli.

LCs containing nnAAs could also enable the production of more complex therapeutic ADCs using site-specific conjugations in a two-step process. First, a nnAA-LC would be produced in E. coli and purified. Next, the purified LC would be added to a CFPS IgG expression reaction where a different nnAA is incorporated into the HC. The resulting IgG would contain two different nnAAs that could be used for two orthogonal conjugation chemistries. This would allow for two different payloads to be conjugated to the IgG in a one-pot reaction, generating a homogeneous ADC with dual payloads. One exciting class of this type of molecule is immunostimulatory ADCs, which are ADCs bearing both a traditional cytotoxic payload and an immune-stimulating molecule. Conceptually, this class of conjugates could be used as a dual cancer treatment and tumor vaccine modality that could potentially result in durable oncolytic immunity. Similarly, the nnAA-LC technology described herein would enable the production of ADC therapeutics that carry two cytotoxic payloads with different mechanisms of action that work synergistically to improve tumor cell killing, such as combining a DNA-damaging reagent and a DNA repair inhibitor.

As a final proof-of-concept, we demonstrated that well-behaved antibodies could be produced at high titers in the cytoplasm of SBDG419. Proper assembly of antibodies expressed in bacterial systems is a significant challenge, even in strains that have been engineered to have a more oxidizing cytoplasm.³⁹ Accordingly, commercial therapeutic antibodies are most often produced in CHO cells, which typically yield high quality, well-assembled IgGs.⁴⁰ In this work, we demonstrated that E. coli strain SBDG419 can produce a full-length IgG that is > 95% assembled by tuning the ratio of HC to LC expression. IgGs produced in E. coli lack the native glycosylation present on CHO-produced IgGs. Previous studies have shown that aglycosylated antibodies have pharmacokinetic properties similar to their glycosylated, CHO-produced version.³⁹ The glycosylation process results in a heterogeneous mixture of multiple glycoforms.⁴¹ Thus aglycosylated antibodies are a good alternative for applications that do not require glycosylation for therapeutic activity.³⁹ Furthermore, the ADC produced from the nnAA-IgG showed potent cell killing activity (EC₅₀ = 0.54 nM) in vitro, demonstrating that the antibodies produced in this E. coli strain retain their biological activity. Subsequent experiments to further validate the use of E. coli as a platform for nnAA-IgG production will include an assessment of product stability, process consistency, and product homogeneity and an evaluation of ADC activity and safety in vivo.

This nnAA-IgG was produced at titers of 1.1 g/L in a high-density fermentation. This titer surpasses the previous highest reported titers for cytoplasmic IgG expression by 2 orders of magnitude. In addition, this is the first report of nnAA-containing IgG production in E. coli. An E. coli-based antibody production process could be executed on a timescale of around 2.5 days, reducing process development and commercial production costs. With these titers, the per day productivity of IgG synthesis in E. coli would rival or surpass those of established CHO-based IgG production processes, which typically take at least 14 days, and could enable E. coli to become an alternative host for the production of non-glycosylated antibodies.¹²

E. coli strain SBDG419 successfully expressed full-length anti-CD74 nnAA-IgG containing a nnAA in the C_H3 domain of the antibody, which required the coordinated expression of genes for LC, HC, pAMF RS and AS tRNA. No truncated product was co-purified as observed by intact protein LC-MS and SDS-PAGE. This result suggests that IgGs with varying nnAA placement could feasibly be expressed in the cytoplasm of our E. coli strain, enabling production of nnAA-IgGs with conjugation handles at sites optimized for conjugation or biological activity. Furthermore, the high levels of amber suppression achieved in this strain led us to believe that it could also feasibly produce nnAA-IgGs containing multiple nnAAs per chain at reasonable titers, which would be extremely challenging with CHO-produced nnAA-containing IgGs due to the low amber suppression efficiency. This promising result demonstrates that high titer production of a nnAA-IgG containing a single pAMF residue in the HC is feasible in the cytoplasm of E. coli. Future studies will include testing the incorporation of multiple nnAAs per chain to evaluate the feasibility of producing high DAR ADCs from nnAA-IgGs produced in E. coli.

In summary, we have shown that a redox optimized E. coli strain has the potential to be a powerful platform for the cytoplasmic expression of antibody-related protein scaffolds, including scFvs, Fabs, LCs, and full-length antibodies, with proper disulfide formation and assembly. With the addition of a tuned nnAA RS and tRNA system, this E. coli strain is capable of expressing complex antibody fragments and antibodies containing nnAAs for site-specific conjugation at high titers of >1 g/L in high density fermentation. Most critically, the final conjugates produced from nnAA-containing proteins exhibited desired biological activities such as antigen-specific cell killing. We anticipate this platform could enable the production of a variety of nnAA-containing drugs in the cytoplasm of E. coli and potentially accelerate progress in healthcare by enabling the low-cost and rapid production of advanced bioconjugate antibody therapeutics that were previously hard or impossible to make.

Materials and methods

Plasmids & strains

To produce protein expression plasmids, each product gene was synthesized and then cloned into pJ411 by ATUM. This vector has a kanamycin resistance marker, pUC high copy origin of replication, and the expression cassette has a T7 promoter for high-level transcription. For the Fab and IgG, which are composed of two separate HC and LC proteins, a bicistronic operon was synthesized with the HC gene followed by the LC and then cloned into the pJ411 vector by ATUM. Both the HC and LC had independent ribosomal-binding sites. Plasmid sequences were verified by sequencing.

For the incorporation of nnAAs into proteins of interest, the selected codons in product genes where nnAAs would be incorporated were substituted with the amber codon “TAG.” The gene for nnAA-LC was mutated to contain 1 TAG codon at the position corresponding to amino acid K42, while the nnAA-IgG contained one TAG codon at position F404 on the HC. Both positions K42 and F404 were labeled according to the Kabat numbering scheme for immunoglobulins.⁴² The coding sequence for the pAMF RS B03¹⁸ was cloned into a medium copy pJ434 plasmid behind a constitutive Pc0 promoter.⁴³ One copy of the amber suppressor tRNA, pAzPhe1(AS tRNA),^18,33 was included on the pJ434 plasmid 3’ to the pAMF RS coding sequence after a 20 base pair spacer sequence.

All plasmids were transformed into E. coli SBDG419²⁶ and plated on selective media (Teknova, L1025 or L1288). For proteins with no nnAA incorporation, only the high copy product plasmid was transformed. For nnAA-containing proteins, each product plasmid was co-transformed with the RS-tRNA plasmid.

Recombinant protein expression in shake flasks

For initial screening experiments, protein expressions were performed in shake flasks. For each experiment, a single colony was picked from transformation plates into 2–3 mL of Terrific Broth (Teknova, T7060) containing appropriate antibiotic(s) (Teknova, K2125 or C2130). After overnight incubation at 37°C in a shaking incubator at 250 RPM, cultures were inoculated into 50 mL of fresh Terrific Broth + antibiotics (50 μg/mL kanamycin and, when applicable, 100 μg/mL carbenicillin) in a shake flask. Cultures were incubated at 37°C in a shaking incubator at 250 RPM until the optical density at 595 nm (OD595) reached 1.0–2.0. At this time, 0.2% arabinose (Teknova, A2010) was added to induce expression of the protein(s) of interest. For cultures expressing nnAA proteins, pAMF was added at 2 mM. Cultures were transferred to 25°C for expression of all proteins. After expression for 16–18 hours, cells were harvested by centrifugation for 5 minutes at ~ 7,000 g. Cells were resuspended in 10 mL per gram of wet cells in phosphate-buffered saline (PBS, ThermoFisher 14,190,250) containing 0.1 mg/mL lysozyme (Sigma-Aldrich, L6876) and benzonase (Sigma-Aldrich 70,746–3). After incubation on ice for 30 minutes, cells were lysed by sonication. Soluble lysates were isolated by centrifugation at > 20,000 g for 30 minutes.

Recombinant protein expression in bioreactors

Trastuzumab LC was expressed as previously described.²⁶ The nnAA-LC and nnAA-IgG were expressed as described previously for pre-fabricated trastuzumab LC expressed in E. coli with the following modifications.²⁶ Carbenicillin (Teknova, C2130) was included in seed cultures to help maintain RS-tRNA plasmid, and 2 mM final concentration of pAMF was added to the bioreactor immediately after induction with arabinose.

CFPS production of nnAA-IgG-X using pre-fabricated LC

CFPS reactions producing the 8× nnAA-IgG-X were performed at 0.25 L scale in bioreactors (DASgip, Eppendorf) as previously described,²⁶ with the following modifications. Reactions contained a plasmid encoding the IgG-X HC containing 3 TAG codons for the incorporation of three pAMF residues per HC. The nnAA-LC was prefabricated and added as a purified protein to a final concentration of 0.5 g/L.

Analytical SEC analysis

Product quality was assessed by injection of 30–50 μg of protein onto an analytical SEC column (Zenix-C SEC-300, 3 μM, 7.8 × 300 mm, Sepax Technologies 233300–7820) with a mobile phase of 50 mM sodium phosphate, 140 mM NaCl, 5% isopropanol, pH 6.5.

ADC cell killing assay

Cytotoxicity effects of the ADCs and unconjugated IgGs were measured with a cell proliferation assay. A total of 12,500 suspension cells (anti-CD74 nnAA-IgG and anti-CD74 ADC) or 625 adherent cells (nnAA-IgG-X and DAR8 ADC-X) in a volume of 25 μL were seeded in a 384-well flat bottom white polystyrene plate on the day of assay for suspension cells, or the day before for adherent cells. Antibodies and ADCs were formulated at 2× starting concentration in cell culture medium and filtered through SpinX 0.22 μm filter columns (Corning Costar, 8160). Filter sterilized samples were serial diluted (1:4) under sterile conditions and added onto cells in quadruplicates. Plates were cultured at 37°C in a CO₂ incubator for 72 hours (suspension cells) or 120 hours (adherent cells). For cell viability measurement, 30 μL of Cell Titer-Glo® reagent (Promega Corp, G7570) was added into each well, and plates processed as per product instructions. Relative luminescence was measured on an ENVISION® plate reader (Perkin-Elmer). Data were fitted with non-linear regression analysis, using log(inhibitor) vs. response, variable slope, 4-parameter fit equation using GraphPad Prism. Data were expressed as percent relative cell viability vs. dose of ADC in nM with error bars indicating the Standard Deviation (SD) of the quadruplicates.

Abbreviations

AS tRNA	=	amber suppressor tRNA
ADC	=	antibody-drug conjugate
CFPS	=	Cell-free protein synthesis
CHO	=	Chinese hamster ovary
DAR	=	Drug-to-antibody ratio
DBCO	=	dibenzocyclooctyne
E. coli	=	Escherichia coli
Fab	=	fragment antigen-binding of an antibody
HC	=	heavy chain
IgG	=	immunoglobulin G
LC	=	light chain
LC-MS	=	Liquid chromatography-mass spectrometry
mAb	=	monoclonal antibody
NHS	=	N-hydroxysuccinimide
nnAA	=	non-natural amino acid
nnAA-IgG	=	nnAA-containing IgG
nnAA-LC	=	nnAA-containing light chain
pAMF	=	para-azidomethyl-L-phenylalanine
PBS	=	phosphate-buffered saline
PFLS	=	pre-fabricated light chain
RBS	=	Ribosome binding site
RS	=	aminoacyl-tRNA synthetase
SDS-PAGE	=	Sodium dodecyl sulfate polyacrylamide gel electrophoresis
SEC	=	Size exclusion chromatography
SPAAC	=	strain-promoted azide-alkyne cycloaddition
scFv	=	single chain variable fragment
T7 pr.	=	T7 promoter
T7 term.	=	T7 terminator
tRNA	=	transfer RNA

Disclosure statement

E.C., J.L., and T.H. were employees and shareholders of Sutro Biopharma, Inc. during the work on this publication. All other authors are employees and shareholders of Sutro Biopharma, Inc.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2024.2316872

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.