Site‐Specific Modification of Native IgGs with Flexible Drug‐Load

Abstract

Homogeneous, site‐specifically conjugated antibodies have shown to result in antibody‐drug conjugates (ADCs) with improved efficacy and tolerability compared to stochastically conjugated ADCs. However, precisely controlling the drug load as well as attaching multiple payload moieties to the antibody remains challenging. Here, we demonstrate the simple and direct modification of native IgG‐antibodies at the residue glutamine 295 (Q295) without the need for any protein engineering with flexible drug‐to‐antibody ratios of one or multiple payloads. The conjugation is enabled through short, positively charged lysine containing peptides and native, commercially available microbial transglutaminase. In proof‐of‐concept studies, HER2‐targeting ADCs based on trastuzumab were generated with drug‐to‐antibody ratios (DARs) of 2 and 4 of the same or different payloads using orthogonal conjugation chemistries. Quantitative biodistribution studies performed with ¹¹¹In‐radiolabeled conjugates showed high tumour uptake and low accumulation of radioactivity in non‐targeted tissues. A single dose study of trastuzumab conjugated to the highly potent payload α‐Amanitin demonstrated complete and long‐lasting tumour remission and was well‐tolerated at all dose levels tested.

ADC generation from native IgG: using short, positively charged peptides like RAK enables the direct modification of IgG at the glutamine Q295 without the need for engineering. Using azide‐ and thiol‐group containing RAK, the generation of ADCs with different drug‐to‐antibody ratios (DARs) is demonstrated and their properties analyzed in‐vitro and in‐vivo.

Introduction

Antibodies are an attractive class of biotherapeutics due to their versatility and targeting specificity. Coupling potent drugs to antibodies to generate antibody‐drug conjugates (ADCs) for their targeted delivery has thus been an intense field of research.[ ¹ , ² , ³ , ⁴ , ⁵ ] First generation FDA‐approved ADCs (Mylotarg®, Adcetris®, Kadcyla®) have generated meaningful clinical data but have unveiled certain drawbacks: i) random, chemical modification of antibodies gives rise to heterogeneous ADCs with different drug‐to‐antibody ratios (DARs) making analysis more challenging, ^[6] ii) each of these species may have a distinct pharmacokinetic profile influencing efficacy and tolerability[ ⁷ , ⁸ ] and iii) unstable linking chemistry which leads to premature release of the toxic payload during circulation and thus to off‐target toxicities and reduced therapeutic efficacy.[ ⁹ , ¹⁰ ] This prompted the development of technologies that enabled the production of site‐specifically and homogeneously modified ADCs. This includes the conjugation of toxins via distinct engineered cysteines (THIOMABs®) ^[11] or the incorporation of non‐natural amino acids into the antibody backbone. ^[12] Enzymatic and chemo‐enzymatic methods for the production of ADCs in a site specific and stoichiometric fashion have also been developed, using, for example, sortase,[ ¹³ , ¹⁴ ] glycotransferase, ^[15] tyrosinase[ ¹⁶ , ¹⁷ ] or transglutaminases. ^[18] Transglutaminases are a family of enzymes (EC 2.3.2.13) that catalyze the formation of a covalent bond between the γ‐carbonyl amide group of a glutamine and the ϵ‐amine of lysine. Our group has pioneered the use of microbial transglutaminase (MTG) for the chemo‐enzymatic modification of tumor targeting antibodies with different radionuclides, fluorescent dyes and toxins via the residue glutamine 295 (Q295) on the Fc‐fragment of antibodies.[ ¹⁸ , ¹⁹ , ²⁰ ] However, quantitative modification of Q295 with substrates containing a primary amine was observed only for aglycosylated or deglycosylated IgGs. ^[18] This either requires the use of PNGase F for deglycosylation of N297 or genetic engineering to abolish glycosylation via N297A, which in return reduces antibody‐dependent cell‐mediated cytotoxicity (ADCC) and can decrease the stability of an antibody due to enhanced proteolytic digestion and aggregation. ^[21] Alternatively, conjugation with MTG can be achieved by genetically engineering the antibody and inserting linear[ ²² , ²³ ] or conformationally locked ^[24] glutamine containing tags or through inserting a lysine residue. ^[25] Dickgiesser et al. used engineered MTG variants to modify native IgG1 with primary amine substrates. ^[26] All four methods allowed the retention of the glycans but required either genetic engineering of the enzyme/antibody, resulted in inhomogeneous conjugates or were limited in terms of drug‐load and number of drug types.

Results and Discussion

During our studies, we observed that small peptides with a positive overall charge could efficiently be conjugated to native (i. e. fully glycosylated) antibodies without any antibody processing. Peptide mapping confirmed that the modified residue was Q295 (see Figure S1). We thus decided to perform a screening of a small library of lysine containing peptides, which had different lengths (3–8 amino acids) and charges using trastuzumab as the model antibody. These peptides revealed differences in conjugation efficiency ranging from 0–94 % (Figure 1). Conjugation was monitored by LC–MS after reduction of the heavy (HC) and the light chain (LC) with dithiothreitol (DTT) (–see SI). The screened peptides were N‐terminal acylated and C‐terminal amidated. We observed that positively charged peptides that only had 3–5 amino acids conjugated most efficiently (i. e. >80 %, see Figure 1). Negatively charged peptides like KADDD, DAKAD, EAKAE displayed low or no conversion at all. This is in accordance with the study of Malesevic et al., who found that negatively charged amino acids in positions adjacent to the lysine reduced enzymatic substrate activity. ^[27] The low conjugation efficiency of negatively charged peptides can be explained by the negative charge density close to MTG's active site, ^[28] which prefers positively charged amino acids.

Figure 1.

Schematic representation of the conjugation process to directly modify the glutamine 295 (Q295) of any native IgG antibody without the use of any antibody engineering. Native microbial transglutaminase from Streptomyces mobaraensis was found to conjugate lysine containing peptides with high efficiency to Q295 that had a length of 3–5 amino acids and an overall positive charge from screening a small peptide library. In contrast, negatively charged peptides were not or with low efficiency conjugated. Most efficiently conjugating peptide 1 was selected for further conjugation studies and the generation of antibody conjugates.

We next investigated if IgG isotypes from different species could efficiently be conjugated since antibodies are widely used in therapeutic and diagnostic applications. We thus equipped the most efficiently conjugating peptide 1 RAK with a functional azido‐group provided by an ϵ‐azido‐Lys‐K(N₃)‐group to make it suitable for bio‐orthogonal click‐reactions (peptide 2, Figure 2A). At an antibody concentration of 3 mg/mL, almost all human IgG antibodies demonstrated near quantitative conjugation (i. e. >99 %), even those already carrying a toxic payload (brentuximab vedotin and trastuzumab emtansine) while mouse IgG2a and IgG2b antibodies could not be conjugated since they lack Q295, as expected. Some rat antibodies showed additional conjugation sites in the light or heavy chain besides Q295 resulting in a higher conjugation ratio (Table 1).

Figure 2.

Overview of the linkers and payloads used in this study as well as the respective conjugation processes, results and functional characterizations. A) Based on the structure of peptide 1, different linkers carrying various functional groups and payloads were designed. B) Schematic representation of conjugating linker 2–5 to the Q295 of native antibody followed by coupling payloads 6–8 to generate the respective ADCs and RICs. Trastuzumab was used as the model IgG antibody. C) Deconvoluted LC–MS spectra of the heavy chain (HC) of the antibody‐conjugates with the respective linkers 2–5 and payloads 6–7 coupled. The light chain (LC) was not modified (see insert in I). I‐‐IV show the LC–MS results of the conjugation process shown in C). D) Cell‐toxicity assay of Trastuzumab, ADC1, ADC2 and T‐DM1 on HER2‐positive SKBR‐3 cells and HER2‐negative MDA‐MB‐231 cells demonstrating target‐mediated cell‐killing.

Table 1.

Conjugation efficiencies of 2 towards different antibodies/isotypes^[a]. The IgG antibodies were reduced for analysis. Conjugation efficiencies to the heavy chain (HC) and light chain (LC) are shown.

Antibody	[%] HC	[%] LC
Trastuzumab (IgG1)	98±1	0±0
Daratumumab (IgG1)	100±0	0±0
Brentuximab (IgG1) vedotin	100±0	3±0
Trastuzumab emtansine (T‐DM1)	100±0	0±0
Human IgG2	93±3	0±0
Nivolumab (IgG4)	117±0	0±0
Rat IgG1	194±9	3±2
Rat IgG2a	114±2	0±0
Rat IgG2b	110±3	57±2
Mouse IgG1	76±5	0±0
Mouse IgG2a	0±0	0±0
Mouse IgG2b	7±2	96±1

^[a] 3 mg/mL mAb/ADC used for reactions.

ADCs are a highly interesting class of therapeutics for the selective delivery of cytotoxic payloads to tumor cells. We thus investigated the flexibility of our peptide linkers to generate ADCs with different and well‐defined drug‐to‐antibody ratios (DARs) and also, if different payloads could be coupled using a combination of bio‐orthogonal chemistries in a simple manner. Recent data suggest that co‐delivering multiple payloads with an antibody may overcome tumor drug resistance.[ ²⁹ , ³⁰ ] Based on peptide 1 and its excellent conjugation efficiency, we designed different linker structures that allow attachment of two of the same payloads using strain‐promoted azide‐alkyne cycloaddition (SPAAC) or two different payloads by combining SPAAC and maleimide‐thiol coupling through incorporating a cysteine residue into the linker (Figure 2A, structures 2–5). As payloads we chose the highly potent cytotoxic payloads Maytansine 6 (similarly to the FDA‐approved ADC T‐DM1) and α‐Amanitin 8 as well as the metal chelator NODAGA 7 (Figure 2A). Native trastuzumab could be readily and completely functionalized within 20 h with the azido‐containing peptide 2 RAK−K(N₃), resulting in the antibody‐linker conjugate ALC1 (Figure 2B and 2C). After removal of excess linker and MTG through Protein A purification, the DBCO‐functionalized Maytansine 6 (Figure 2B) was quantitatively clicked to ALC1 within 18 h at 2.5 molar equivalent per conjugation site resulting in the highly homogenous DAR2 ADC1 as confirmed by LC–MS (Figure 2C, I). We also generated linker 3 having two azido‐groups to see if the drug‐load on the antibody could be increased to DAR4 (Figure 2A and B) since such high DAR ADCs are thought to kill tumor cells that only express a low number of receptors on their surfaces more efficiently. The DAR4‐linker could be conjugated quantitatively to the native antibody to yield ALC2 and after clicking with 6, the highly homogeneous DAR4 ADC2 could be generated as confirmed by LC–MS analysis (Figure 2C, II). We next wanted to generate antibody conjugates with two different payloads attached focusing on coupling a cytotoxic payload and a metal chelator to trace the resulting ADC and radio‐immunoconjugate (RIC), resp., in biodistribution studies. We chose NODAGA as a metal chelator due to its suitability for various radiometals and high in‐vivo stability. We investigated two routes to generate these dual‐payload conjugates: 1) we synthesized linker 4 having pre‐installed NODAGA as well as an azide‐group and which was conjugated to native trastuzumab with >95 % efficiency to yield RIC1 (Figure 2B and 2C, III). In a second‐step, RIC1 was quantitatively clicked with 6 to yield the desired dual‐payload RIC2. ¹¹¹In‐radiolabeled RIC2 (labelling efficiency=98 %) was stable in PBS for over 96 h and a Lindmo assay confirmed fully retained immunoreactivity on SKOV3‐ip cells with a calculated immunoreactive fraction of 111.1 % (see Figure S6 and S7). 2) we coupled linker 5 that has incorporated two different, orthogonally compatible groups: an azide‐group and a thiol‐group, which was provided by a cysteine residue. Linker 5 could be conjugated quantitatively to trastuzumab yielding ALC3 (Figure 2C, IV). Incorporation of NODAGA 7 was achieved upon reducing the ALC3 conjugate with 30 eq. DTT and subsequently re‐oxidating it with dhAA resulting in ~95 % conversion into RIC3. Clicking of 6 was quantitative, yielding the very‐well defined dual‐payload RIC4 (Figure 2C, IV). ¹¹¹In‐radiolabeled RIC4 (labelling efficiency=97 %) showed full immunoreactivity on SKOV3‐ip cells (Figure S8). These data highlight the excellent chemical flexibility and versatility of these positively charged peptide linkers and that site‐specific mono – and dual‐payload antibody‐conjugates at different drug‐loads can rapidly be generated in a simple manner directly from native antibodies ‘off‐the‐shelf’ without the need for any antibody or enzyme engineering.

High stability of ADCs is key to maximize payload delivery to tumors and to achieve optimal therapeutic efficacy. Therefore, the biophysical properties of the ADCs and stability in human serum were analyzed. Maytansine ADC1 and ADC2 were incubated in PBS at 37 °C over a period of 14 days and analyzed by size exclusion chromatography (SEC) for homogeneity. Both ADCs showed high stability with no aggregation similar to the reference trastuzumab emtansine (T‐DM1) (Figure S9). In PBS and human serum, no linker cleavage or deconjugation could be detected by LC–MS during a 6 d incubation further confirming the high stability of the linkers (Figure S10). The cell cytotoxic potential of the ADCs was assessed by performing in‐vitro cell‐cytotoxicity assays on HER2 positive cell‐lines. ADC1 and ADC2 showed sub‐nanomolar potency similarly to T‐DM1 with no toxicity on target‐negative cells (Figure 2D) which indicates that the toxicity was mediated by the target HER2. Also α‐Amanitin 8, a highly potent, bicyclic peptidic RNA polymerase II inhibitor, could successfully be clicked. The resulting ADCs were found to be highly potent with EC₅₀ values in the pico‐ and nanomolar range on multiple HER2‐positive cell lines (Figure S11). These data indicate that the peptide linkers generate highly stable ADCs that efficiently release the payload upon target‐mediated cell uptake.

Radiolabeling is presumably the most sensitive and quantitative method to assess the pharmacokinetic and biodistribution profiles of immunoconjugates like blood clearance or off‐target accumulation.[ ¹⁸ , ¹⁹ , ³¹ , ³² ] We thus performed biodistribution studies of RIC1 and RIC2 in mice to probe for tumor specific uptake of the radio‐immunoconjugates. The ¹¹¹In‐labeled conjugates were intravenously injected into female nude mice bearing SKOV3‐ip human ovarian carcinoma xenografts and the biodistribution was monitored over a period of 96 h (Figure 3A). RIC1 demonstrated high and specific tumor accumulation over time with a maximal mean uptake as high as 49±9 % ID/g at 72 h post injection. Off‐target uptake in non‐targeted organs at all time points was limited to below 9 % ID/g. In the blood, the dose was 11 % ID/g at 24 h and 9 % ID/g at 72 h post injection. Similar findings were made for RIC2 (Figure S12). These results show that the site‐specifically modified antibodies enabled high uptake in the targeted tumor tissues while non‐specific uptake in non‐targeted tissues was limited. This confirms previous findings of site‐specifically generated ADCs.[ ²⁰ , ³³ ]

Figure 3.

In vivo biodistribution and efficacy studies of the generated conjugates. A) a biodistribution study of ¹¹¹In‐radiolabelled RIC1 was performed in mice and the % ID/g analyzed in the tumor and organs at different time points after injection. RIC1 demonstrated high and specific tumor accumulation with increasing tumor to blood and liver ratios over time with limited off‐target uptake in non‐targeted organs (<9 % ID/g at all time points). Maximal mean tumor uptake was as high as 49±9 % ID/g at 72 h post injection. B) Single dose study of the α‐Amanitin ADC trastuzumab‐2–8 in a JIMT‐1 xenograft model demonstrated excellent anti‐tumor efficacy with complete responses in all animals (10/10) until the end of the study. The arrow indicates the day of injection on day 1. C) Monitoring the mouse weight during the study, no loss could be detected and also at 6 mg/kg the ADC was still tolerated.

Finally, an in‐vivo efficacy study in a moderate HER2 expressing JIMT‐1 xenograft model was performed, using trastuzumab‐2 armed with 8 α‐Amanitin (DAR2). A single dose of 2 mg/kg was sufficient to induce complete and sustained tumor remission in all animals (10/10 complete responses) up to the end of the study (172 d) (Figure 3B). During the whole study, the ADC was well tolerated with no signs of weight loss and also at 6 mg/kg the ADC was still tolerated (Figure 3C). These data indicate that the assembled ADCs are highly efficacious and well tolerated using highly potent α‐Amanitin.

In summary, we demonstrated that commercially available, native microbial transglutaminase can be used for the site‐specific conjugation of native, off‐the‐shelf IgG antibodies at the glutamine residue 295 (Q295) without the need for any antibody engineering or reduction. Importantly, there was no need to remove the glycosylation at N297, which confers important stability against proteolytic degradation[ ²¹ , ³⁴ , ³⁵ ] and also plays a central role for the immune system to tune a broad range of biological activities. ^[36] Short, positively charged peptides containing a lysine residue for conjugation obviate the need for any engineering step and enable the direct and specific modification of a single residue within native antibodies. Incorporating one or several functional groups such as azido‐groups into the linkers can be done in a straightforward manner. ADCs with different DARs can therefore rapidly be generated using click‐chemistry. In our studies, we demonstrated that ADCs with various payloads and DARs could be generated and that the resulting conjugates are highly homogeneous and stable under prolonged incubation times despite using large hydrophobic DBCO‐groups which we attribute to the hydrophilic nature of the linker. The resulting ADCs were highly potent in‐vitro similar to trastuzumab emtansine (T‐DM1) confirming that the payload is efficiently released in tumors cells.

ADC stability is key to maximize payload delivery to the targeted site and to avoid dose‐limiting, excessive toxicities and achieve a wide therapeutic index.[ ¹¹ , ²⁰ , ³⁷ , ³⁸ ] In human serum studies no payload loss or linker cleavage could be detected (Figure S10) and in‐vivo biodistribution studies confirmed specific and high tumor accumulation with limited uptake in non‐targeted organs, demonstrating that the peptide linker generates highly stable ADCs (Figure 3A). Lastly, we could furthermore show, in a moderate HER2 expressing xenograft model, that the linker also generates efficacious and tolerable ADCs using the highly potent payload, α‐Amanitin (Figure 3B).

Conclusions

The presented data indicates that the linker technology is well‐suited to generate stable and well‐defined ADCs which are key prerequisites to design ADCs with a high therapeutic index. We believe that the presented approach will be universally applicable and find widespread application for the development of therapeutic or diagnostic antibody‐conjugates since Q295 is a conserved residue among most IgG antibodies from different animal species, including humans.

Funding

This work was partially funded by the European Union Seventh Framework Programme FP7/2007–2013 under Grant 602820.

Conflict of Interests

Authors declare the following competing financial interests: RS, PS and MB are co‐founders of Araris Biotech that possesses an exclusive license of the technology presented herein; TH, MK, and AP are current employees of Heidelberg Pharma.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Wehrmüller J. E., Frei J. C., Hechler T., Kulke M., Pahl A., Béhé M., Schibli R., Spycher P. R., ChemBioChem 2025, 26, e202400511. 10.1002/cbic.202400511

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.