Design of Experiments for Early-Stage Process Development of Antibody-Drug Conjugates: A CDMO Example
Antibody-drug conjugates (ADCs) are a new class of biopharmaceutical drugs that use a chemical linker to bind monoclonal antibodies together with highly potent anti-cancer agents. For CDMOs, ADCs present a lucrative development opportunity, but success hinges on the ability to scale up processes effectively while maintaining Good Manufacturing Practice (GMP) and regulatory compliance. The data analytics process known as Design of Experiments (DOE) helps facilitate the development of ADC processes. By enabling identification of important process parameters and a robust Design Space, DOE helps facilitate a faster and more reliable process for scale-up.
This article is posted on our Science Snippets Blog.
A CDMO Example of ADC Process Development
Piramal Grangemouth is a world-leading bio-conjugation CDMO with more than 15 years of experience in antibody drug conjugation and bio-conjugation. The site supports proof of concept (milligram scale) through to commercial manufacturing, including bulk drug substance (BDS) offering development, manufacturing and quality control. Piramal GMP manufactures 40 distinct conjugates and two commercial products.
In a recent webinar with Sartorius, Piramal presented a case study and example of how to develop a robust process for the reactive stages of an antibody drug conjugate (ADC). The webinar covers the process development, considerations needed for DOE, and the parameters/responses used to create a robust design space.
“Antibody drug conjugates (ADCs) bring together the best features of antibodies & cytotoxic drugs” - Xavier Despinoy (Piramal)
Antibody-drug conjugates represent an innovative therapeutic application because they combine the ability to carefully target specific antigens – using tumor-specific monoclonal antibodies (mAbs) that are not sufficiently lethal to cancer cells – with the potent cancer-killing activity of highly cytotoxic small molecule drugs that would be too toxic to humans if administered alone.
Because these agents are capable of delivering highly cytotoxic payloads directly to tumor cells, they can be used to achieve high lethality toward the targeted cancer cells while leaving healthy cells unharmed.
Read more: How CDMOs Can Digitalize Their Cell and Gene Therapy Processes
Creating the Right Process
An important aspect of success with ADC development lies in getting the model right. Understanding what the data mean and developing a process that is robust enough for scale-up relies on obtaining accurate process data. The start of this journey is creating a design of experiments that answers the right questions.
The goals when developing early phase antibody-drug conjugates include:
- Develop scientifically sound analytical methods suitable to support pre-clinical and ultimately, clinical release and stability testing of ADC
- Develop process conditions to meet key quality attributes for the ADC
- Have sufficient understanding of process robustness to enable safe scale-up
- Establish control strategy
DOE can facilitate and speed both analytical and process development activities.
Analytics Development
ADC models are inherently complex, and selecting the right parameters isn’t straightforward or easy. In addition to the typical complexity of creating an antibody assay, you also have to factor in the payload. So, that means factoring in the effects (and cross-influence) of all three elements: the antibody + payload + conjugate.
You need to consider the critical quality attributes of ADCs and consequently develop appropriate analytical methods to measure the effects of parameters on product quality. These may include SEC, DAR, distribution (HIC, PLRP), icIEF, residual free drug and CE-SDS (R+NR), which need developing at the early stage of the product development process.
All of these methods are needed to develop the process appropriately. In addition, you also have data relating to the antibody functional assays such as binding (ELISA), potency (the cell kill assay) and specific effector functions. These methods are very complex and require significant development.
Significant data analysis is required to develop the antibody functional bioassays.
Read more:
5 Stages of Development to Go to GMP
At Piramal the approach to process development for ADC has been designed to speed up the program, with some activities running in parallel with others. There are typically five stages in the process:
Miniconjugations > Formulation Studies > Reactive Stages Development > Scale-up Studies > Toxicology Batch Manufacture
The analytical development starts even before process development begins, and it continues throughout the process development phase and later continues with methods qualification.
The formulation studies may start before the reactive stage (they can run in parallel). But it takes a number of weeks, or months sometimes, to get the appropriate final formulation.
The final stage before going to GMP manufacture is testing the entire process with all operating units included at an intermediate scale.
Why Use DOE?
Design of Experiments (DOE) is a proven process used by CDMOs and biopharma development companies worldwide to more quickly get statistically optimal results and define a Design Space with fewer experiments in less time.
DOE allows systematic assessment of the effect of multiple factors (and their interactions) on responses in a very efficient way.
Factor = Process Parameter
Response = Critical Quality Attribute
It allows you to define the “Design Space” – the safe operating conditions where critical quality attributes (CQAs) meet the targets/ranges.
When and Where Do You Do ADC DOE?
DOE is a development tool used at early phases to support both ADC analytical and process development. In later phases, DOE is an important tool to understand how critical specific parameters are and to define the Design Space during pre-commercial process characterization studies.
What does a typical antibody drug conjugation process look like? The diagram below illustrates it.
The antibody might not always be in the right matrix to enable conjugation, so the first step is typically a buffer exchange or a simple pH adjustment if the chemistry is compatible. Following that, the reactive stage(s) can be one or several steps, during which ultimately, you’re introducing the cytotoxic payload. Then, this is followed by a purification step which can be chromatography or TFF or a combination of both. Next is formulation, which is the stage where you ensure your product is in a stable state. Then you perform a final sterilizing filtration.
A Specific Example
Below is an example showing the four steps used by Piramal in this case. The first step is to adjust the pH – which means dilution to get the targeted concentration and appropriate pH for the chemical reaction to happen. This is followed by the reduction with TCEP. The next step is conjugation of the payload using a solvent. The last step is to quench the excess payload introduced.
Factors and Responses to Consider for DOE
Taking into account the steps involved in ADC, you will need to consider the factors and responses that will be included in your Design of Experiments. As you can see in the illustration below, in this case, there are a large number of potential factors and parameters from which to select. These include process parameters: protein concentration, pH, temperature, TCEP equivalence, payload equivalence, reactions time, percentage solvent, etc. The Responses are the ADC critical quality attributes, such as aggregation, binding, potency, charge profile, drug load/distribution, etc.
Considerations Prior to DOE
The parameters ranges you select will play a role in how effective your design is. This is impacted by your prior knowledge, scouting experiments and/or manufacture fit.
There are a number of statistical designs available for DOE so the selection should be guided by the parameters you have selected and resources at your disposal, and the information you want out of the design. For early phase, we are typically using factorial design, either full or fractional.
Once you have selected the process parameters and the statistical design, the next step is preparatory work to enable you to execute the design. For example, if pH or concentration is an important factor, you must consider how you can generate the starting antibody at the appropriate pH and concentration.
The selection of an appropriate scale down model is also important in order to avoid introducing undesired variability during execution, which would negatively impact the ability to model the true process effects.
Proposed Factors and Ranges
In the case study presented, the factors and responses selected for the DOE included:
- protein concentration from 5 to 15 mg/mL
- temperature from 16-26 °C
- pH from 6.8 to 7.8
- reduction time from 60 to 180 minutes
- ranges investigated for the drug antibody ratio (DAR) were from a minimum of 3.4 to maximum of 4.4 with the target at 3.9
Setting Up the DOE
The study was designed to check the Drug Antibody Ratio (DAR) and make sure it stayed between 3.4 and 4.4, with the ideal target at 3.9 (the “response”).
The quality attributes are hard specifications that must be fulfilled. They define the “sweet spot,” (or Design Space) and are used in robust setpoint calculations. These are imported in the MODDE Design Wizard and create the model. Keep in mind the target will influence the Design Space and Optimal Set-Point.
The experiment was set up as a full factorial design (a classical design for DOE) with 16 experiments in corners and three center-points. There were many experiments inside the specifications and a high R2 gives a high probability for a large Design Space.
It is not common to get a large Design Space from a screening design, but this design did fulfill the important criteria. Excellent chemical knowledge supported the selection of appropriate factor ranges and good recordkeeping of experiments with correct values entered into the worksheets provides clear results.
The Analysis Wizard in MODDE 13 Makes It Easy to Find the Best Model
The quality of the model is summarized with four bars:
- R2 (green) describes how well the model fits to the data
- Q2 (blue) describes how well the model can predict the data
- Reproducibility (turquoise) is very high as the replicates was very close together
- Model Validity (yellow) is a bit low but that is common when Reproducibility is very high
A Full Factorial Design Using MODDE
Watch the webinar or download the slides to see the details and data from a Design of Experiments example for ADC. The MODDE Design Wizard make finding the best model simple.
Watch the webinar (or download the slides) to see the full design of experiments setup and analysis.
Presented by Xavier Despinoy (Piramal) and Erik Johansson (Sartorius).