Na Li, Ph.D., Technical Account Manager, Sino Biological, and Yue Liu, Ph.D., President, Ab Studio
Multispecifics are challenging to design, develop, and manufacture, owing to their high complexity. Computer-aided antibody design and artificial intelligence combined with more traditional antibody development technologies can accelerate the development of multispecific candidates with good safety, efficacy, and manufacturability. Support from a contract research organization with the ability to produce and test complex multispecifics in a high-throughput manner further facilitates the rapid discovery and development of successful multispecific therapeutics.
Bispecific Antibodies Offer Improved Targeting
Monoclonal antibodies (mAbs) have been very successful treatments for a wide range of diseases, particularly cancers and autoimmune disorders. However, mAbs can bind to only one target. Bispecific (bsAbs) and trispecific antibodies overcome this issue by providing two or three binding sites, respectively, increasing specificity and efficacy while reducing side effects.
bsAbs recognize and target two different epitopes, enabling the inhibition and/or activation of two different signaling pathways. Alternatively, they can bring two different molecules or cells together so that they can interact and achieve a greater effect. These abilities open numerous doors with respect to potential applications for the treatment of not only cancer and immune disorders but infectious diseases and other ailments.
Examples include bsAbs that target a cancer cell and a T cell receptor (T cell engagers) and bsAbs that target two different sites on a single cancer cell. Five bsAbs have received marketing authorization — four in the United States and three in the EU. It is estimated that bsAbs account for close to 20% of the clinical antibody pipeline for immuno-oncology, immune-mediated and autoimmune disorders, diabetes, asthma, and rheumatoid arthritis.1 As of March 2022, 180 bispecific antibodies were in preclinical development, and more than 50 were being investigated in clinical trials.2,3
More Complexity Means More Development Challenges
Multispecific antibodies, including bsAbs, trispecifics, and tetraspecifics, comprise different antibody chains with different structures and binding properties. As such, they are much more complex than mAbs with respect to their design, development, and manufacture. Many different aspects must be balanced during the discovery and developmental process.
The selection of the specific chains is dictated by the targeted indication and the anticipated mechanism of action (MoA). One of the first hurdles is to choose an appropriate MoA for a given indication, which can include well-established targets (e.g., PD-L1) or new targets discovered as the understanding of the disease biology progresses. For the former, the goal is to develop a best-in-class drug, while the latter will be a first-in-class therapeutic.
Whether to use patented antibody chains or non-patented structures is another important decision. The former requires a license, while the latter often has manufacturability issues, such as a propensity toward aggregation or poor thermostability. Indeed, the design of a multispecific antibody must take into consideration the structures of the different chains in order to achieve an optimal molecule with good stability, function, safety, and efficacy that is also manufacturable.
Manufacturing issues range from maximizing the production of the desired heterodimer over homodimers and other variants with minor differences in post-translational modifications (PTMs) to avoiding aggregation and improving the thermostability sufficiently to ensure longer shelf lives, which requires sequence optimization.
Given the high degree of complexity and numerous challenges, early development efforts regarding bsAbs and trisAbs progressed slowly with a high rate of failure. As new computer modeling and artificial intelligence (AI) technologies have been developed for antibody design, significant advances have been achieved in recent years. As a result, the multispecific space looks promising for the future.
Computer-Aided Design Making a Difference
Computer-aided antibody design (CAAD) is becoming increasingly important in the development of multispecific antibodies. It enables the selection of publicly available sequences that avoid aggregation and stability issues that also have not been used in other therapeutic applications. Starting with the disease indication and MoA and aided by CAAD and AI, it is possible to enhance and accelerate the discovery process.
As an example, in one bsAb project, Ab Studio was able to dramatically improve the selectivity for heterodimer formation. Using CAAD with in silico modeling and the application of AI algorithms, it was determined that the introduction of a specific mutation in a specific part of the sequence could increase heterodimer formation from a low 62% to greater than 90%.
Ongoing Improvements to Computer Modeling
Much of the computer modeling used for multispecifics is built on the capabilities established for mAbs, which is quite accurate due to the high numbers of crystal structures that have been obtained for different mAbs. Access to this data increases the accuracy of mAb models.
For multispecifics, there are two or more antibody chains involved that can potentially interact with one another. In addition, there are two or more antibody–antigen complexes involved in the binding of multispecifics. To achieve optimum efficacy and safety, therefore, it is necessary to optimize not only the manufacturability of these complex molecules but also the multiple binding interactions that lead to their therapeutic effects.
Even though there are multiple chains, optimization of manufacturability for multispecifics is still highly effective using mAb data. In addition, as more multispecifics are produced and evaluated, the data generated for them is also incorporated, leading to software revisions that improve model performance. It is, in fact, a cyclic process. Using one strategy, in silico design is followed by production of physical multispecific candidates that are subjected to numerous physicochemical and biological assays. The data generated is fed back into the model, generating an improved version that is applied to a next round of design.
At the same time, AI, machine learning, and other computation tools are advancing at a rapid pace. These advances are also being incorporated into CAAD systems, further enhancing their ability to predict multispecific behaviors. In addition, they are enabling more personalized designs that better balance safety, efficacy, and manufacturability. As a result, the design and development of complex multispecifics will be increasing at an accelerated rate.
One limitation to the development of next-generation therapies of all kinds, including multispecifics, is a lack of sufficient talent, particularly people with a combination of biology and computer science skills. Within the current education system, training of students along these lines is limited. If more bench scientists involved in drug discovery had deeper knowledge of computer modeling and AI, the industry would be much better positioned to realize the advantages offered by these tools and technologies.
CAAD at Ab Studio
Ab Studio is a California-based startup combining AI and CAAD with traditional antibody discovery and engineering technologies to develop next-generation therapeutic antibodies (e.g., bsAbs, trisAbs) with the best balance of safety, efficacy, and manufacturability. Three unique CAAD-based antibody discovery platforms are used as part of a quality-by-design approach to developing complex antibodies for the treatment of cancer, infectious diseases, and central nervous system disorders, with several candidates in preclinical and clinical stages.
The strategy for multispecific development begins with a toolbox of more than 10 antibody structures. Depending on the indication and MoA, three or four of these structures are selected to begin the development work. Initial bench-scale investigation enables the selection of the best structure from a safety and efficacy perspective. Proprietary AI and CAAD technologies are then used to optimize the sequence for safety, efficacy, and manufacturability. With this approach, it is possible not only to increase the accuracy of multispecific design with respect to matching the MoA and clinical need but also to significantly reduce the development time.
One recent example is the development of bispecific and trispecific antibodies based on llama single-domain antibodies (VHHs) that bind to SARS-CoV-2 spike (S) protein and block its binding to the human ACE-2 receptor. CAAD was used to design bsAbs and trisAbs with the VHHs fused to human IgG1 Fc domains that target multiple epitopes and therefore remain effective against emerging variants of the SARS-CoV-2 virus. In addition, the multispecific antibodies have favorable biophysical characteristics for manufacturability, including minimal aggregation and good thermostability, even at 45 °C.
Strategies to Improve Production and Quality
In addition to the design process, production of bsAbs has proven to be challenging as well. With the rapid development of recombinant DNA technology and a deep understanding of antibody engineering, diverse bsAb formats are emerging, which are usually classified into two distinct types: IgG-like bsAbs and non-IgG-like bsAbs. IgG-like bsAbs have a conserved immunoglobulin constant domain, thus retaining Fc-mediated effector functions, such as complement-dependent cytotoxicity (CDC) and antibody-dependent cellular cytotoxicity (ADCC). Non-IgG-like bsAbs have a smaller size because of the lack of Fc fragments.
Choosing proper expression systems is vital to efficient expression and production. For some non-IgG bsAbs, such as BiTE and tandem bispecific scFv, they can be expressed in Escherichia coli, yeast, or mammalian cells, including CHO and HEK293 cells. E. coli is a common choice for scFv expression because bacteria can grow rapidly on cheap media and produce heterologous proteins in large amounts. However, expressed scFv molecules lack the formation of intra-domain disulfide bonds, which is essential to the "immunoglobulin fold” structure. Additional protein refolding and recovery steps are usually required to generate soluble scFv molecules. To overcome these problems, mammalian cells can be used to express bispecific scFvs, owing to their ability to perform complex posttranslational modifications.
Like conventional mAbs, IgG-like bsAbs are expressed predominantly in mammalian cells. However, the production of bsAbs is more challenging due to the double number of heavy and light chain genes. Typically, it requires co-transfection of at least two expression plasmids into the mammalian cells, and the ratio of the two plasmids may also influence both the quality and quantity of the expressed bsAbs. During the early stages of bsAb development, HEK293 cells are often used for transient expression. Sometimes, it is difficult to scale up transient expression of IgG, and the titers are relatively low compared with stable CHO cells. Due to various structural similarities between mAbs and bsAbs, many established purification processes for conventional mAbs are compatible with bsAbs. A variety of purification methods, such as affinity, charge, size, hydrophobicity, and mixed-mode-based separation techniques, are employed for the purification of bsAbs.
As an example, in a featured case of bsAb production at Sino Biological, the bsAb construct has a knobs-in-holes (KiH) design in its heavy chain. After the three-step purification process, protein A affinity chromatography, ion exchange, and gel filtration, we were able to achieve greater than 95% monomer purity.
Supporting High-Throughput Development at Sino Biological
High-throughput production of complex multispecific antibodies is essential for improving in silico modeling approaches to the development of custom-designed candidates. AI-driven development platforms require large quantities of data obtained in the wet lab for physical samples in order to continually improve.
Sino Biological is an internationally recognized reagent supplier and contract research organization. As the global leader in recombinant technology, Sino Biological specializes in recombinant protein production and antibody development. On the basis of its extensive expertise and experience in mammalian cell expression and proprietary technology platforms that are specifically optimized, Sino Biological provides fast and efficient bispecific antibody expression services. Starting from the antibody sequence, Sino Biological can deliver multiple bsAb formats, such as BiTE, Diabody, CrossMab, and DVD-IgG. Numerous bsAb projects have been completed in the past few years with greater than 90% overall success rates, and yields of 250 mg/L or more have been achieved. Furthermore, in addition to traditional CHO and HEK293 cell lines, Sino Biological has access to the FucoFree cell line for the production of bsAbs with enhanced antibody-dependent cellular cytotoxicity (ADCC) or complement-dependent cytotoxicity (CDC) effector functions.
Investment in high-throughput capabilities has made it possible to support the production of up to 1,000 antibody-based products at a time, with lead times as short as three to four weeks for bsAb production. Sino Biological is currently investing in robotics technology to enhance the cloning process to further reduce lead times and potential costs as well.
This high-throughput capability helps support projects at companies like Ab Studio that are searching for optimal multispecific candidates by pursuing multiple rounds of in silico modeling and AI computation followed by physical candidate testing.
Sino Biological rapidly produces large numbers of potential multispecific candidates that have been identified by Ab Studio using its CAAD and AI platforms. As a one-stop contract research organization, Sino Biological then performs testing and analysis of those candidates, evaluating physicochemical properties and performing functional assays to assess biological activity. The resulting data is fed back into Ab Studio’s platform, and further modeling is conducted to identify a smaller set of more optimized candidates. The cycle is repeated until the best solution is clearly identified.
Sino Biological and Ab Studio are not partner companies or co-marketing partners.
References
1. “Bispecific Antibodies - Current Status and Prospects.” 9 Mar. 2022.
2. Mullard, A. “FDA approves 100th monoclonal antibody product.” Nat. Rev. Drug Discov. 20:491-495 (2021).
3. Labrijn AF, ML Janmaat, JM Reichert, PWHI Parren. “Bispecific antibodies: a mechanistic review of the pipeline.” Nat. Rev. Drug Discov. 18:585–608 (2019).
