In biotherapeutic production, delivery of the therapeutic agent is a significant factor for achieving successful therapeutic effect. With this consideration in mind, the delivery mechanism requires forethought. Among the more innovative drug delivery systems being developed today are those employing nanoparticle technologies.

The development of nanoparticles in recent years has expanded into a range of clinical applications. Nanoparticles were developed to address the limitations posed by free therapeutics as well as to overcome biologics barriers, including systemic, environmental, and cellular barriers. Work in nanoparticle development has focused on ways to optimize drug delivery systems so that these systems can be applied across patient populations and across different diseases (1).

“Safe and effective delivery of therapeutic cargo to its target tissue at a clinically relevant dose is crucial for developing successful drugs regardless of the route of administration,” says John Lewis, PhD, CEO of Entos Pharmaceuticals. “Nanoparticle engineering must be purpose-driven and designed around the physical and biochemical properties of the therapeutic cargo and the disease it is designed to treat. The cargo size, charge, and ability to tolerate chemical modification greatly impact the type of nanoparticle used in a drug development program.”

“Additionally,” Lewis states, “the complexity and cost of manufacturing must also be considered when developing novel therapies. All of these factors need to be considered at the start of the drug development process to maximize both clinical and commercial success.”

The biggest benefits and challenges of using nanoparticles come with designing a nanoparticle delivery system that can deliver a molecule to the right target tissue safely and efficiently, Lewis explains. What’s more, it is challenging to design nanoparticles that do not require complex or costly manufacturing.

“For example, viral vectors transduce target cells with high efficiency but can only be used to deliver DNA, have cargo capacity limits, can be costly to manufacture, and cannot be used for repeat dosing due to their immunogenicity,” Lewis points out.

Currently marketed lipid nanoparticles (LNPs) also have limitations, Lewis adds, such as poor biodistribution that preferentially targets the liver, poor tolerability, especially at high doses, and lack of utility for delivering DNA. To address such limitations, Entos developed a technology platform (Fusogenix PLV) that combines tiny fusion proteins with well-tolerated lipids. The company’s approach combines the best features of viral vectors and LNPs to deliver RNA, DNA, or novel gene-editing technologies safely, effectively, and repeatedly by direct fusion with target cells, Lewis states.

“The nature of your engineered nanoparticle will determine if its benefits outweigh its challenges,” says Lewis.

With much discussion in the biopharma industry surrounding nanoparticle technology, specifically for its use in bio-drug delivery, LNPs are at the forefront. “Among the myriad of nanoparticles, LNPs are the leading candidates for delivery of many different biological drugs, which includes nucleic acids such as mRNA [messenger RNA], siRNA [small interfering RNA], pDNA [plasmid DNA], gene editing tools such as CRISPR [clustered regularly interspaced palindromic repeats]/cas9 and oligonucleotides,” says Rajiv Kumar, PhD, principal scientist, Applied Research and Clinical Team, West Pharmaceutical Services (2,3).

Kumar notes that some current key focus areas in the LNP space revolves around designing new novel lipids, such as ionizable lipids to improve the efficiency of delivery and the release of encapsulated payloads from nanoparticles; targetability of LNPs; and improving the toxicity profile of LNPs (4,5). Despite LNPs accounting for the majority of nanoparticles for biologics drug delivery in clinical trials today, preliminary reports have begun to shine light on a new generation of hybrid nanoparticles, such as the combination of lipids and polymers, says Kumar. These reports have been showing high potential in these hybrid nanoparticles of successful biologic drug delivery (6), he adds.

“Polymer nanoparticles and microparticles have shown some early promise in delivery of proteins (peptides and antibodies), however, more work is required to fully utilize the potential of polymer nanoparticles for biologics delivery,” (7) Kumar states.

As LNPs are currently the most well-known and used nanoparticle technology today for biologic-based drug delivery, a closer look at their technical advantages can suggest further applications in the future.

According to Kumar, the types of advantages (3) that LNP systems convey over conventional drug delivery systems include:

Encapsulation: in the case of nucleic acid-based therapies, the encapsulation of nucleic acids in the core of nanoparticles protect them from degradation by nucleases, thus providing higher stability and protection.

Immunogenicity: LNPs minimize adverse immunogenic responses.

Size advantage: the size and compositions of LNPs enable the therapeutic payload, such as nucleic acids, to efficiently cross negatively charged cell membranes and be released into the cell’s cytoplasm.

Target-specific: targeting of specific disease sites or organs can be achieved by conjugating targeting molecules on the surface of LNPs.

Safety: the biodegradable nature of lipids used in LNPs provides a safe and non-toxic platform for delivering biological drugs.

However, despite the advantages that a nanoparticle technology-based drug delivery system may have, what are the risks inherent in their manufacture?

Kumar elaborates that LNPs are the most complex of manufactured nanoparticles in terms of composition and activity. For instance, each individual lipid component (ionizable lipid, helper lipid, cholesterol, and polyethylene glycol [PEG]-lipid) has a specific role—from encapsulation to formation of nanoparticles and from stabilization to successful delivery and therapeutic efficacy. “Although standard lipid components are in place, a precise tuning and optimization is required for designing a LNPs system for a specific biologic drug payload delivery,” (2) he explains.

In addition, LNP production requires an extremely precise process of mixing different components into specific ratios, says Kumar. The reason for this strict precision is to obtain monodisperse nanoparticles that have high encapsulation efficiency. Following initial production, the nanoparticles must then be purified by ultrafiltration/diafiltration processes to conduct buffer exchange and remove any impurities or solvents from nanoparticle suspension. Finally, the purified nanoparticles are sterile filtered before fill/finish (8).

Some of the key challenges in the manufacturing of LNPs are the quality of nanoparticles, which must be monodisperse and with a high payload content; they require highly precise and uniform mixing; and complex downstream processing, especially purification using ultrafiltration, Kumar highlights. Moreover, containment is a challenge. “One of the main risks is the selection of the right type of containment system (e.g., types of vials and stoppers), which still needs to be evaluated as some of the coatings can have an adverse impact on the quality/stability of nanoparticles during storage and transportation,” he explains.

Should the challenges and risks in nanoparticle production be mitigated, then the biopharm industry would have a powerful drug delivery system that, in the end, may alleviate patient concern over safety. Some of the key benefits for the biopharma industry include the availability of a flexible platform for encapsulating a variety of drugs, including hydrophobic drugs (i.e., most small-molecule drugs), hydrophilic drugs (i.e., biologics, protein therapeutics, antisense oligonucleotides), and highly negatively charged drugs (i.e., nucleic acids), emphasizes Kumar.

Furthermore, new disease targets would be accessible through the expedient of designing targeted nanoparticles systems, and combination therapies could be more feasible by means of designing nanoparticles systems that can encapsulate two drugs together for synergistic/combination effects, Kumar adds.

Meanwhile, the benefits to patients include better therapeutic outcomes since nanoparticle-based delivery of drugs results in higher therapeutic benefits compared to conventional drugs. In addition, because nanoparticles-based systems are designed to accumulate in a specific organ or disease site, the undesired side off-target effects typically seen with conventional drug delivery are minimized in patients. This benefit is especially attractive for some of the highly toxic oncology drugs, Kumar says.

A new area being explored in the field of drug delivery is the use of exosomes as delivery vehicles. Exosomes are naturally occurring nanovesicles derived from cells that play an integral role in intercellular transport of materials. It has been shown that exosomes can incorporatetherapeutics, such as small molecules or nucleic acid drugs, and deliver them to specific types of cells or tissues; that is to say, exosomes can achieve targeted drug delivery (9).

As above, the benefit of targeted drug delivery is that it increases local concentration of a therapeutic and can thus minimizes side effects. Exosome engineering, however, is still in its infancy. However, early research has shown that exosome-mediated drug delivery demonstrates low toxicity, low immunogenicity, and high engineerability (9).

Kumar notes that exosomes have been studied in biomarker applications and therapeutic agents in several degenerative diseases and oncology applications (10), but that they have already been deployed in several clinical trials and have shown satisfactory results in not only biomarker applications, but also as drug delivery systems, in vaccine applications, and as therapeutic agents themselves (11). However, although exosomes hold potential, their heterogenous nature pose manufacturing challenges, especially related to their isolation and purification. Manufacturing of exosomes requires further exploration and improved standardization, Kumar states.

The use of nanoparticle engineering may be agood solution for current challenges in the delivery of therapeutics, but overcoming manufacturing issues will be required to standardize production and ensure safety, efficacy, and efficiency.

1. Mitchell, M. J.; Billingsley, M. M., Haley, R. M.; et al. Engineering Precision Nanoparticles for Drug Delivery. 

, 101–124. DOI: 10.1038/s41573-020-0090-8
2. Kulkarni, J. A.; Witzigmann, D.; Thomson, S. B.; et al. The Current Landscape of Nucleic Acid Therapeutics. 

, 630–643. DOI: 10.1038/s41565-021-00898-0
3. Cullis, P. R.; Hope, M. J. Lipid Nanoparticle Systems for Enabling Gene Therapies. 

 (7), 1467–1475. DOI: 10.1016/j.ymthe.2017.03.013
4. Sabnis, S.; Kumarasinghe, E. S.; Salerno, T.; et al. A Novel Amino Lipid Series for mRNA Delivery: Improved Endosomal Escape and Sustained Pharmacology and Safety in Non-human Primates. 

 (6), 1509–1519. DOI: 10.1016/j.ymthe.2018.03.010
5. Herrera-Barrera, M.; Ryals, R. C.; Gautam, M.; et al. Peptide-Guided Lipid Nanoparticles Deliver mRNA to the Neural Retina of Rodents and Nonhuman Primates. 

 (2), eadd4623. DOI: 10.1126/sciadv.add4623
6. Mukherjee, A.; Waters, A. K.; Kalyan, P.; et al. Lipid-Polymer Hybrid Nanoparticles as a Next-Generation Drug Delivery Platform: State of the Art, Emerging Technologies, and Perspectives. 

, 1937–1952. DOI: 10.2147/IJN.S198353
7. Pagels, R. F.; Prud’homme, R. K. Polymeric Nanoparticles and Microparticles for the Delivery of Peptides, Biologics, and Soluble Therapeutics. 

, 519–535. DOI: 10.1016/j.jconrel.2015.09.001
8. Shepherd, S. J.; Warzecha, C. C.; Yadavali, S. et al. Scalable mRNA and siRNA Lipid Nanoparticle Production Using a Parallelized Microfluidic Device. 

 (13), 5671–5680. DOI: 10.1021/acs.nanolett.1c01353
9. Liang, Y.; Duan, L.; Lu, J.; Xia, J. Engineering Exosomes for Targeted Drug Delivery. 

 (7), 3183–3195. DOI: 10.7150/thno.52570
10. Muthu, S.; Bapat, A.; Jain, R.; Jeyaraman, N.; Jeyaraman, M. Exosomal Therapy—A New Frontier in Regenerative Medicine. 

, 7. DOI: 10.21037/sci-2020-037
11. Rezaie, J.; Feghhi, M.; Etemadi, T. A Review on Exosomes Application in Clinical Trials: Perspective, Questions, and Challenges. 

 (1), 145. DOI: 10.1186/s12964-022-00959-4.

Feliza Mirasol is the science editor for 


Volume 36, No. 9
September 2023
Pages: 8–10

When referring to this article, please cite it as Mirasol, F. Nanoparticle Engineering in Drug Delivery Under the Microscope. 

Articles

Cover Story

Nanoparticles offer the potential for a safer, more effective method of drug delivery to the patient.

Nanoparticle Engineering in Drug Delivery Under the Microscope

Monoclonal antibodies (mAbs) are an important class of therapeutic proteins used for treating various diseases such as cancers, autoimmune diseases, and genetic disorders (1,2). Therapeutic mAb products are known to have different variants, such as conformational isomers, charge variants, glycan variants, and size variants (3). Size variants, such as aggregates and fragments, are formed when samples containing mAbs undergo stresses, such as those encountered during the various stages of the product life cycle, including cell culture, purification, storage, formulation, transportation, administration, and handling. The formation of aggregates as the result of various stress factors can lead to the reduction of process yield and product shelf life. The administration of samples containing mAb aggregates to patients may also result in adverse immune responses such as anaphylaxis and neutralization of endogenous proteins. Hence, monitoring, analysis, and control of mAb stability is a critical challenge, and researchers and manufacturers continue to show interest in it as well (3–5).

Monoclonal antibody aggregation is influenced by a variety of factors, such as buffer composition, ionic strength, protein concentration, mechanical stress, temperature, interfacial surfaces, and chemical contaminants. Aggregates could vary in size, structure, and morphology (2). They can be soluble oligomers (less than 100 nm size) or particles of sub-visible (0.1–100 µm) and visible (more than 100 µm) size range (6). Therapeutic protein aggregates of different size ranges are analyzed using an array of techniques such as size exclusion chromatography (SEC), dynamic light scattering (DLS), asymmetrical flow field-flow fractionation, electrophoresis, analytical ultracentrifugation, light obscuration, nanoparticle tracking analysis, microscopes, and flow-based imaging techniques (7,8). Microscopes and flow-based imaging techniques are widely used and play a crucial role in aggregate analysis. The application of different microscopes to analyze aggregates of mAb has also been mentioned in the 

Optical, electron, and atomic force microscopy are some of the commonly used microscopes to visualize aggregates of therapeutic proteins (8). Electron microscopes, such as transmission electron microscopy (TEM) and scanning electron microscopy (SEM) can be used in the visualization of specimens as small as 1 nm to a few microns. TEM and SEM can further be used to determine the size, shape, and morphology of aggregates. Optical microscopes can be easily used to visualize particles having sizes of more than 1 µm. Brightfield and fluorescence microscopes are commonly available optical microscopes. In brightfield imaging, aggregates can be visualized as dark spots against a bright background, whereas in a fluorescence microscope, aggregates can be visualized by labeling them with fluorescent dyes. However, the use of microscopes to analyze therapeutic protein aggregates has a general disadvantage in that only a small fraction of the sample can be analyzed at a time, and, hence, the output might not be a true representation of the entire sample. Further, manual identification of subtle changes in the size, shape, and distribution of protein aggregates from a large number of microscopic images would be cumbersome, and would also be susceptible to human error, variability, and biases (2,8–11).

For this reason, image analysis algorithms are being increasingly used to characterize aggregates. These algorithms are able to analyze a large number of microscopy images via automatic quantitation for various attributes such as size, count, shape, and morphology. Image analysis algorithms offer outputs with high accuracy and reproducibility that would be otherwise challenging to achieve manually, thereby facilitating high throughput screening of protein aggregates (9–11). These algorithms can be used either as a complementary tool to already established techniques used for therapeutic protein aggregate characterization, or to address research applications where other existing techniques cannot be used (9). The advantages of using image analysis algorithms with microscopic techniques are summarized in 

A fluorescence microscope is a type of optical microscope that enables selective determination of specimens due to tagging with a fluorescent dye. The microscope works on the principle of fluorescence where the analyte is irradiated using light from an excitation source followed by the emission of light having a higher wavelength, which is selectively detected (9). Apart from tagging the sample with fluorescent dye, this technique does not require any other complex sample preparation steps (8,11). A simple widefield fluorescence microscope is cheap and easy to maintain and use. Wider illumination areas combined with lesser operational complications are also its distinctive features. Moreover, images in a fluorescence microscope can be acquired with reduced exposure times. However, as explained in the previous section, the use of fluorescence microscopes, in general, has a disadvantage in that only a small fraction of the sample can be analyzed at a time (9–11). The fluorescence microscope has an additional drawback wherein its images are corrupted by noise and blurred particles from out of the focal plane. The noise results in images with very low signal-to-noise ratio (9).

A novel semi-automated image analysis algorithm that removed different noises in the fluorescence microscope image and enabled the analysis of therapeutic mAb aggregates has been proposed recently. In this approach, fluorescent microscope images of aggregates captured using a widefield fluorescence microscope were used (9). The images were captured at x4 magnification and had a resolution of 1536 x 2048 and a scale-to-pixel ratio of 133:50 (m/pixels). In the algorithm (illustrated in 

), the first step was importing the microscope images containing fluorescent dye-tagged mAb aggregates into MATLAB (MathWorks) as a 3D matrix. The two dimensions of the matrix corresponded to the number of pixels along the length and breadth of the image. The third dimension of the matrix was designated to correspond to red, green, and blue channels. The contribution of the green channel was taken forward for processing the images. Median filtering was then applied, followed by the application of a total variation algorithm and fast Fourier transform. The image obtained after denoising was used to normalize the background. After denoising, either weighted median or modified 2D Otsu thresholding was done. Thresholding was followed by standard binary morphological processing steps such as closing, filling, and cleaning, which were used to label individual active regions in the image for subsequent analysis. The scale-to-pixel ratio of the image was used to normalize the data from pixel to metric format. The details of aggregate particles obtained from the images in the form of circularity, diameter, perimeter, and area were arranged in ascending order and sorted into various groups for plotting size distribution graphs. A flowchart showing different steps used in fluorescence microscope image processing is depicted (

), and a representative fluorescence microscope image, binary image obtained after its processing, and the size distribution of aggregates obtained from a set of images are also shown (

). A detailed description of all the steps used in fluorescence image processing has already been published (9). 

Brightfield microscopy is a widely used imaging technique (8,12). It is the simplest of all microscopy techniques, which makes it widely popular. In this technique, the sample is illuminated with white light, which ultimately results in showing off specimens as dark images in the sample. Brightfield microscopy can be used to visualize and analyze sub-visible and visible aggregates. Generally, brightfield images of protein aggregates appear as dark spots depending on the refractive index of the protein aggregates and the surrounding medium (12). However, brightfield microscopy may not be suitable for detecting aggregates with low refractive index or with low contrast between the aggregate and the background. Monoclonal antibody aggregates are usually translucent in nature, and hence their effective visualization in brightfield images requires filtering the sample containing the aggregates onto a membrane followed by staining with a dye. The translucent nature, random irregular shapes of the aggregates, low contrast between background and aggregate, and inconsistent stray light can make aggregate analysis tedious (8,12–14). Hence, as per the authors’ knowledge, there have been severely limited efforts to automatically quantify information from brightfield images of therapeutic mAb aggregates that have not been stained with any dye.

A brightfield image analysis algorithm that can analyze mAb aggregates of more than 1-µm size was recently published using images of aggregates captured by an imaging system (Cytell Cell Imaging System, GE Healthcare) (10). The acquired images at x4 magnification had a resolution of 1920 x 1440 and a scale-to-pixel ratio of 11:10 (m/pixels). The images were called in for image analysis on MATLAB, and the processing algorithm involved several steps, starting from converting the brightfield image to grayscale image. This conversion step was followed by determining the number of pixels corresponding to each color code from 0 to 255, where 0 represents black and 255 represents white. The pixel-color distribution was plotted and was used to segregate the aggregates, background, and hemocytometer lines using thresholding. The aggregates are dark colored and cover the left portion of the peak while the background and particle boundary were included in the peak and on the right side of it (

). Following thresholding, morphological operations were performed to improve the images. The particles in the output binary images were then labeled, and the area corresponding to each label was determined. The scale-to-pixel ratio was used to normalize the data from pixel to metric format. From the extracted data, aggregate size distribution was obtained. A flowchart showing overall steps used in the brightfield image processing is depicted in 

, and a detailed description of all the steps used for image processing has been published elsewhere (10). 

An electron microscope is used to obtain high-resolution images of biological specimens utilizing electrons used as the source of illumination. Transmission electron microscopy (TEM) is a type of electron microscope used to view thin specimens through which electrons can pass to form the image. TEM is used to analyze the interior of cells, structure of proteins, viral structures, and nanoparticles. Electron microscopes can provide visual information about mAb aggregates in the submicron size range. However, the application of output micrographs is usually limited to visual assessment of aggregates, and, thus, there have been limited efforts to quantify the information from electron micrographs. An image analysis algorithm for automated detection and numerical size distribution of therapeutic mAb aggregates was therefore developed. The acquired micrographs were subjected to de-noising, low-pass filtering, background normalization, conversion to binary images, and automatic determination of heterogeneous aggregate sizes and shapes to obtain aggregate distribution. A detailed description of the TEM image processing has been published elsewhere (15).

Image analysis algorithms have been applied to different projects to analyze mAb aggregates. Three case studies are presented in this article. In the first case study, application of the brightfield image analysis algorithm to characterize mAb aggregates obtained by subjecting the mAb samples to rapid interfacial agitation through bubbling stress is shown. Application of the fluorescence microscope image processing algorithm to analyze mAb aggregates in serum and phosphate solubilizing buffer (PBS) is demonstrated in the second case study. Third case study depicts the aggregation of therapeutic mAb in serum having the combined presence of ferrous ion (Fe

mAb aggregation by air/liquid interfacial agitation stress through air bubbling

The exposure of mAb-containing samples to various interfacial surfaces is known to impact the product’s stability. Interfacial surfaces encountered by mAbs during various stages of manufacturing, transportation, and handling occur in the form of solid/liquid, liquid/liquid, and air/liquid surfaces. Air/liquid interface is commonly observed in the form of air bubbles and headspaces. The combination of agitation and interfacial stress is highly detrimental to the stability of mAbs and could occur during various operations including sparging, mixing, pumping, shaking, filling, transportation, and handling (16). In this study, samples containing 1 mg/mL of mAb in PBS at 37 °C were subjected to rapid air/liquid interfaces and agitation stress using a peristaltic pump (

) followed by analysis using an array of techniques, including brightfield microscope imaging at x4 magnification and image processing to analyze aggregate size in the sub-visible and visible ranges. All the images of samples that were subjected to stress for different time points were analyzed. Brightfield images of samples subjected to stress resulted in dark-colored aggregates. The output images obtained after image analysis showed white particles against a black background. In the case of samples containing mAb subjected to stress for 60 minutes, image processing resulted in aggregate sizes up to 20 µm. Aggregates having sizes of up to 60 µm were obtained from samples containing mAb subjected to stress for more than 60 minutes. A representative image of a sample containing mAb that was stressed for 120 minutes, the output binary image obtained after image processing, and the size distribution of aggregates obtained from several images are shown in 

. The average diameter of aggregates obtained by the algorithm at different time points ranged from 5–9 m. 

Comparing aggregate size distribution in serum vs. in PBS

The aggregate size distribution in the sample containing mAb after administration into the human body could vary and requires deeper investigation (17,18). Buffers such as PBS and 

 models such as serum, plasma, cerebrospinal fluid, and synovial fluid can be used to mimic physiologically relevant conditions to determine the stability of therapeutic proteins (19). Serum is a complex biological matrix that contains many components, including proteins, lipids, and small molecules, and it interferes with the analysis of therapeutic mAb when using techniques such as size exclusion chromatography (SEC) and dynamic light scattering (DLS) (19,20). Hence, tagging the therapeutic protein with a suitable fluorescent dye and analyzing it using a fluorescence-based technique are required to analyze mAb aggregates in serum (19). This case study explored the comparison of aggregate size distribution of mAb in PBS and in serum using the fluorescence microscope image processing algorithm (

). The mAb was tagged by fluorescein isothiocyanate (FITC), a fluorescent dye. Among various samples containing mAb subjected to different stresses, the samples subjected to stirring and interfacial agitation stress resulted in green-colored FITC-tagged aggregates at all the analysis time points in both PBS and serum. Further, serum samples containing mAb aggregates obtained by stirring and interfacial agitation resulted in a higher percentage of particles having more than 10 µm size compared to PBS samples at all the analysis time points. A representative image of serum and PBS containing FITC-tagged mAb subjected to stirring stress that was captured at 48 hours and the size distribution of aggregates obtained by compiling several images captured at this time point are shown in 

. Overall, the case study showed that the aggregate size distribution of mAb in different fluids, such as PBS and serum, that are used to show physiologically relevant conditions would be different. 

Aggregation of therapeutic IgG in the combined presence of Fe

 are known to cause degradation of proteins. H

 leads to Fenton reaction, resulting in the formation of hydroxyl radicals (•OH) that are known to cause oxidative damage to biomolecules such as DNA, proteins, and cells (21). The degradation of biomolecules in the combined presence of H

 is known to be higher than the degradation in the presence of only H

 (22). This case study explored the aggregation of mAbs in the combined presence of H

 in saline, artificially prepared extracellular fluid, macromolecule-free serum, and serum (23). In the case of human serum, the aggregation of FITC-tagged mAb in the presence of both H

 were explored using the fluorescence microscope analysis algorithm (

). It was observed in fluorescence microscope that serum samples containing mAb having the presence of both 0.8 mM Fe

 resulted in the presence of green-colored aggregates (

), whereas serum samples containing mAb that did not have both H

 did not result in aggregates that were in the analysis range of fluorescence microscope image analysis. The processing of fluorescence microscope images of serum containing H

 showed that aggregates having sizes up to 60 µm were observed, with average aggregate size being 12.5 µm. 

Apart from the application of established analytical techniques, mAb aggregates can also be analyzed using image analysis algorithms coupled to microscopy images. The application of algorithms to process microscopic images can be utilized either as a complementary tool to already established techniques, or to address research applications where existing tools cannot be applied.

1. Bansal, R.; Dash, R.; Rathore, A. S. Impact of mAb Aggregation on Its Biological Activity: Rituximab as a Case Study. 

 (9), 2684–2698. DOI: 10.1016/j.xphs.2020.05.015
2. Sreenivasan, S.; Jiskoot, W.; Rathore, A. S. Rapid Aggregation of Therapeutic Monoclonal Antibodies by Bubbling Induced Air/Liquid Interfacial and Agitation Stress at Different Conditions. 

, 97–109 . DOI: 10.1016/j.ejpb.2021.08.010
3. Sreenivasan, S.; Kumar, D.; Malani, H.; Rathore, A. S. Does Interaction of Monoclonal Antibody Charge Variants with VEGF-A and ELISA Reagents Affect its Quantification? 

, 113513. DOI: 10.1016/j.ab.2019.113513.
4. Joubert, M. K.; Luo, Q.; Nashed-Samuel, Y.; Wypych, J.; Narhi, L.O. Classification and Characterization of Therapeutic Antibody Aggregates. 

 (28), 25118–25133. DOI: 10.1074/jbc.M110.160457
5. Vázquez-Rey, M.; Lang, D. A. Aggregates in Monoclonal Antibody Manufacturing Processes. 

 (7), 1494–1508. DOI: 10.1002/bit.23155
6. Kannan, A.; Shieh, I. C.; Hristov, P.; Fuller, G. G. In-Use Interfacial Stability of Monoclonal Antibody Formulations Diluted in Saline I.V. Bags. 

 (4), 1687–1692. DOI: 10.1016/j.xphs.2020.10.036
7. Bansal, R.; Gupta, S.; Rathore, A.S. Analytical Platform for Monitoring Aggregation of Monoclonal Antibody Therapeutics. 

, 152. DOI: 10.1007/s11095-019-2690-8
8. Zölls, S.; Tantipolphan, R.; Wiggenhorn, M. Particles in Therapeutic Protein Formulations, Part 1: Overview of Analytical Methods. 

 (3), 914–935. DOI: 10.1002/jps.23001
9. Sreenivasan, S.; Sonawat, D.; Mandal, S.; Khare, K.; Rathore, A. S.Novel Semi-Automated Fluorescence Microscope Imaging Algorithm for Monitoring IgG Aggregates in Serum. 

, 11375. DOI: 10.1038/s41598-021-90623-7
10. Sreenivasan, S.; Sonawat, D.; Rathore, A. S. Image Analysis Algorithm-Based Platform for Determining Micron and Higher Aggregate Size Distribution of Therapeutic IgG Using Brightfield and Fluorescence Microscope Images. 

, 1747–1763. DOI: 10.1007/s11095-021-03108-7
11. Demeule, B.; Gurny, R.; Arvinte, T. Detection and Characterization of Protein Aggregates by Fluorescence Microscopy. 

 (1–2), 37–45. DOI: 10.1016/j.ijpharm.2006.08.024
12. Selinummi, J.; Ruusuvuori, P.; Podolsky, I.; et al. Bright Field Microscopy as an Alternative to Whole Cell Fluorescence in Automated Analysis of Macrophage Images. 

 (10), e7497. DOI: 10.1371/journal.pone.0007497
13. Buggenthin, F.; Marr, C.; Schwarzfischer, M.; et al. An Automatic Method for Robust and Fast Cell Detection in Bright Field Images from High-Throughput Microscopy. 

, 297. DOI: 10.1186/1471-2105-14-297
14. Filipe, V.; Jiskoot, W.; Basmeleh, A. H.; et al. Immunogenicity of Different Stressed IgG Monoclonal Antibody Formulations in Immune Tolerant Transgenic Mice. 

 (6), 740–752. DOI: 10.4161/mabs.22066
15. Kumar, M.; Pant, A.; Bansal, R.; et al. Electron Microscopy-Based Semi-Automated Characterization of Aggregation in Monoclonal Antibody Products. 

, 1458–1465. DOI: 10.1016/j.csbj.2020.06.009
16. Li, J.; Krause, M. E.; Chen, X.; et al. Interfacial Stress in the Development of Biologics: Fundamental Understanding, Current Practice, and Future Perspective. 

, 44. DOI: 10.1208/s12248-019-0312-3
17. Schuster, J.; Koulov, A.; Mahler, H. C.; et al. 

Anurag S. Rathore, asrathore@biotechcmz.com, is a professor in the Department of Chemical Engineering at the Indian Institute of Technology Delhi and a member of 

®’s Editorial Advisory Board, and Shravan Sreenivasan is a graduate student working under the supervision of Prof. Rathore.


Volume 36, No. 9
September 2023
Pages: 12–21

When referring to this article, please cite it as Rathore, A. S.; Sreenivasan, S. Image Analysis Algorithm for Therapeutic mAb Aggregate Analysis. 

Development

Image analysis algorithms coupled with microscopy techniques can be used to characterize aggregates of therapeutic proteins.

Image Analysis Algorithm for Therapeutic mAb Aggregate Analysis

Understanding of cell physiology, health, and metabolism during cell culture and microbial fermentation enables more efficient, cost-effective production of high-quality biologic drugs. This knowledge guides decision-making about process development and enables improvement of bioreactor designs. Both activities, therefore, impact process and bioreactor performance.

Inline monitoring provides access to process information in real time, dramatically increasing bioprocess understanding, from fundamental process behaviors to more advanced attributes such as cellular functioning. As such, it can lead to significant improvements in productivity, quality, and process stability and better performing culture runs, often resulting in savings of time, materials, and human resources.

Recognition of the importance of real-time process monitoring is resulting in the development of solutions for tracking an increasing range of process parameters and other cellular attributes. Some of these solutions leverage existing technologies in new ways, while others are novel techniques developed specifically for inline monitoring within bioreactors.

In an ideal world, it would be possible to monitor all metabolic pathways involved in cell culture and microbial systems. That is not practically feasible, of course. Historically, according to Melissa Semple, senior product manager for automated perfusion at Cytiva, the focus has been on monitoring three areas: the controlled environment, molecular proxies for media consumption, and cell density.

The controlled environment, Semple explains, is represented by sensors on bioreactor equipment, such as mass-flow-controller gas flow rates, agitation speed, pump speeds, weigh-scale changes, vessel pressure, etc.“The data gathered with these instruments provides assurance that we have control over the equipment and can generate the desired/intended conditions,” he says.

The environment is also monitored for features that have critical impact on the cells such as pH, dissolved oxygen (DO), and when possible or the sensor is suitable, dissolved carbon dioxide (CO2), Semple comments. During microbial fermentation, the cell line may also need to be monitored for reactive oxidative species via continuous redox sensing.

These process parameters are tracked because they affect cell growth, productivity, and product quality, according to Shunsuke Shiina, R&D scientist at AGC Biologics. The data are thus used to control the airflow rate, agitation speed, media feed rate, and pH. “This multitude of sensors is required to manage the most basic life support of the culture or fermentation process, and they are treated as critical to the control of the process because they monitor critical process parameters (CPPs),” Semple adds.

Semple also notes that the importance of these sensors has led to their continuous improvement, leading to the development of true in situ or in-line sensors that are continuously in contact with the process solution. The readings obtained from these sensors are used for fast control of bioreactor devices and ultimately the culture/fermentation environment for the cells, with management of the process environment generally achieved via automated closed-loop controllers.

Molecular proxies, Semple continues, represent the carbon sources that the cells rely on commonly for growth and energy (e.g., glucose, glutamine), as well as metabolized products for these molecules (e.g., lactate, glutamate) and cell wastes, with ammonia historically the proxy molecule.“These molecules are really the most basic and minimal indicators roughly representing the supply and consumption of media components. They give process operators a partial picture of the culture/fermentation metabolism and have been workhorse readings relied on for process optimization.”

Metabolite readings have typically, however, required collection of samples for offline analysis. There is therefore a need for sensors that can measure these—and ideally a wide array of other metabolites—in situ to the culture so that immediate and more continuous readings of the culture health can be obtained, according to Semple.

Cell density is monitored closely because the density must be determined precisely to know when to perform certain key process steps, such as transfection or infection, which can have a large impact on the final yield of a process, observes Clément Dumont, product manager with Univercells Technologies.

In situ measurement of viable cell density and total cell density is ideal because the data can be used to create control loops for feeding and enable quick responses to changes in cell densities in real-time, Semple says. Unfortunately, this determination is most often made via automated microscopy analysis of off-line samples. She does note, though, that more continuous permittivity readings from in situ sensors made for bioprocessing can in most cases be correlated to viable cell density.

Monitoring of common parameters such as DO, pH, temperature, sparge, agitation, and CO2 overlay predate the emergence of modern process analytical technologies (PATs) and thus, says Pramthesh Patel, vice-president of process development and manufacturing, science, and technology for Avid Bioservices, have relied on earlier analytical technologies such as dedicated probes and sensors.

In recent years, Patel notes that the emergence of innovative PATs has allowed for the real-time monitoring of a host of additional parameters, offering a clearer view into what is happening to, and in some cases inside, the cell. “This latest collection of areas of monitoring have focused on cell viability, nutrient and metabolite levels, protein quality/integrity, glycosylation, and the redox state of the cell,” he explains. The PATs that have allowed for this expanded monitoring ability include capacitance, Raman spectroscopy, fluorescence lifetime imaging (FLIM), and near-infrared (NIR) and mid-wavelength infrared (MWIR) spectroscopy.

“PAT tools are allowing process attributes that were historically measured offline into the realm of real-time monitoring,” agrees Christopher Kistler, fellow scientist, biologics product development with Catalent Biologics. He observes that capacitance technology is used to build biomass prediction models from electrical signals, and Raman/NIR spectroscopy solutions for inline monitoring of metabolites are in development. David Ede, process technology manager at Sartorius, highlights new enzymatic probes for determination of nutrient and waste metabolite levels and emphasizes that Raman spectroscopy has been adapted for the measurement of concentrations of many different analytes, including glutamine, ammonium, amino acids, and even proteins. Dumont cautions, however, that these inline solutions for metabolites have yet to be seamlessly integrated with all single-use bioreactors.

Different methods for different biomolecules

To be truly useful, inline monitoring strategies must be selected according to the needs of each specific process. “Users need to be clear on what they are trying to achieve first to select the suitable PAT systems,” states Dumont. “While the technologies used for basic cell-culture monitoring are often quite standard, the parameters monitored to determine the quality and attributes of the final product depend on the type of product, and currently no standard solutions for inline assessment of those parameters exist,” he explains.

Different biologic drug substance classes may require different methods of bioprocess monitoring, Ede agrees. Chemical synthesis processes often use chromatographic methods such as high-performance liquid chromatography, while microbial processes rely on infrared technology to measure total biomass. For protein-based therapies, which are mostly produced by fed-batch processes, the key parameters are metabolite levels and cell density. These can be measured by enzymatic and capacitance probes, as well as by Raman spectroscopy. The process volumes can also be large, so CO2 levels may need to be monitored as well.

Novel modalities have additional process-monitoring requirements. Processes for the manufacture of virus-based therapies, Ede observes, require measurement of product titer and the ratio of full to empty capsids, which can be determined offline by various technologies, such as multiangle-light scattering following a size-exclusion chromatography step. For cell-based therapies, meanwhile, the correct cellular activation or differentiation status is essential because the cells are the product. Online sensors can measure cell size, number, and aggregation, but they may not be able to detect the correct differentiation of cells. Raman spectroscopy is under investigation for this application. “Multifrequency capacitance measurements could potentially provide more advanced cell parameters in the future, but there is little evidence of their application in cell therapy yet,” Ede remarks.

The application of PAT tools alongside prediction models and the comparison of analytical techniques enables customization of inline monitoring to the drug substance, Kistler concludes. He also notes that the range of PAT tools now available can be customized within a specific class of biologic drug substance, for example, an antibody/cell line may require a specific Raman model.

The nature of the drug substance can also influence the sensor design that is most appropriate for a specific process-monitoring application, according to Semple. “The shape and nature of the molecules being detected determine the sensor design,” she contends.She also notes that often early versions of sensor technologies are available in offline form, allowing for detection of critical molecules before the sensors are ready for use under aseptic conditions.

In fact, sensors are often redesigned to be used seamlessly in an aseptic environment, Semple says. “This means they either need to maintain accurate function after being exposed to a steam-in-place cycle used to prepare a stainless-steel bioreactor or fermenter for use or a gamma irradiation or microwave cycle used to prepare a single-use bioreactor consumable.”

Another important aspect of in situ sensor design is continuous accuracy throughout the duration of a run. Semple also notes that some sensor designs will not translate to future in situ designs for continuous monitoring, such as those in which materials used in the sensing mechanism are consumed in a reaction. “These sensors,” he says, “will only be adaptable to offline forms that will accept discrete samples pulled one at a time from a bioreactor or atline solutions where culture is continuously distributed to the sensor.”

“The growing body of available information garnered with the latest PATs offers a wealth of benefits to drug manufacturers, both at the outset of process development when experimentation is a critical and often complex activity, and when manufacturing is taking place based on an established process,” Patel states.

One advantage is increased freedom and versatility when designing experiments for identifying optimal manufacturing processes, observes Patel. “Without real-time monitoring, each experiment would have to be carefully conducted individually and then followed by another independent experiment to build on the knowledge gained during the previous experiment. However, the power of the latest PATs shows what is happening to the cell in real-time. Together, these technologies enable continuous experimentation involving dynamic challenges to the environment within the bioreactor, driving the identification of the process development sweet spot much more rapidly and efficiently.”

In addition, PAT tools can provide various enhancements over the life cycle of a process, according to Ede. They can enable process optimization by testing different parameters, collecting data on process robustness, and leveraging computer modeling to improve performance.

Once a process is optimized and characterized, PAT technologies can enable real-time control by setting tolerance zones for each parameter and applying a control strategy to adjust them within the limits, Ede continues. For instance, using a program such as Sartorius’ SIMCA-online to monitor the process in real time can provide early warning of any anomalies that may affect product quality or yield and allow implementation of an automated control strategy that can reduce operator errors or sterility issues. “This approach leads to lower product variation and fewer process deviations,” he states.

Bioreactor design can also increase flexibility, according to Dumont. He points to fixed-bed bioreactors such as the Univercells scale-x bioreactor, which is composed of a structured fixed bed that entraps cells.“This design makes the implementation of perfusion processes easier than in stirred-tank reactor for which advanced cell retention devices need to be used,” he explains. He adds that when this feature is combined with a PAT tool for glucose measurement, perfusion processes can be run for extended periods of time with constant glucose concentrations, affording a well-controlled media environment for the cells.

Separately, Univercells is implementing a new approach for cell-density determination using a correlation model fed by glucose and lactate data, which Dumont notes highlights the important of using PAT tools for data collection in combination with advanced data analysis techniques.

Real-time data from inline tools can provide data that are linked directly or indirectly to specific process parameters. In either case, the data can be used to improve manufacturing processes, according to Shiina. “The information obtained from in-line sensor can indicate if the feed amount and timing of media addition is appropriate for the cell’s activity. In addition, sensor data can be fed back to operators and/or bioreactor devices in real time to control important process elements, such as the supply of oxygen and air, the agitation number, and pH,” he explains.In the future, Shiina believes such feedback loops may ultimately be controlled by artificial-intelligence-based software leveraging both accumulated data and inline measurements.

Semple adds that continuous sensor data allows operators to control processes in a more consistent manner, potentially improving control of the environment on a cell-specific basis so that the final process will be robust from batch-to-batch. “The end result is greater confidence in the ability to develop processes that will be reliable from a quality and yield perspective,” he observes.

Kistler notes that PAT tools also bring more data into processing decisions at the time that decisions are being made, which increases confidence in those decisions. As an example, he highlights how PAT tools facilitate the precise control of feed delivery rates and provide instant feedback on residual metabolite concentrations, allowing process performance to be evaluated more holistically. “PAT tools could enable a future state where having access to product quality data in real-time will allow for evaluation of process and product data together; potentially improving quality, while minimizing the number of experiments required. Decisions can be made with more information and in an appropriate timeframe for course correction if necessary. Most importantly,” Kisstler emphasizes, “processing errors can be detected as they happen, and mitigated before they have the opportunity to become catastrophic.”

Contributing to improved bioreactor designs

Data collected with inline sensors have applications beyond process improvement and control. They can also be used, Patel observes, to inform strategies for enhancing the design, structure, or functionality of bioreactors. As an example, he notes that existing data can reveal additional data that would be good to collect, which in some cases may create opportunities to incorporate new probes/technologies directly into bioreactor designs rather than have them supplied as add-ons.

In fact, Semple notes that some of the most important improvements in bioreactor design in the course of history have come from the industry identifying the most critical molecules or cell features to the health and control of a cell-driven process.

“Making single-use interfaces available in bioreactor bag designs is key to fully utilizing the benefit these sensors can bring for a process while also reducing the risk of contamination in the long run,” Ede agrees.

There is at least one caveat, however, according to Shiina. “Using this information for long-term bioreactor design is dependent on a few factors, including the scale and shape of the bioreactor and the optimal locations and numbers of inline sensors being continuously improved.”

Semple expands on that thought: “As the biology of cell metabolism becomes better understood, more indicators of culture performance are becoming critical indicators for improved consistency in cell-culture and fermentation outcomes. The importance of a molecule to the health or productivity of the target cell type will determine where the biopharma industry will focus.The ability of the sensor to survive the environment of a bioreactor or fermenter will determine the placement of the sensor as a bench-top or in situ sensing device.If continuous culture sampling can be provided that works with the atline sensor solution, the continuous sampling feature becomes important for certain cell-culture or fermentation processes. These solutions will be created when the new sensed value is critical to the function or consistency of the process controlled.”

PAT tools designed for real-time monitoring can be based on novel technologies, but often they leverage existing analytical techniques adapted to perform in an inline, continuous manner. “It is not always necessary to develop new technology, as existing technology can sometimes be leveraged to address these new applications, with the application of Raman spectroscopy for real-time monitoring of product quality attributes being one such example,” Kistler says.

Capacitance and FlIM are two other primary PATs recently introduced that do not represent fundamentally new technologies, adds Patel.“They are each long-established technologies based on the fundamentals of physics. However, it is only recently that these approaches have been applied to advancing the monitoring capabilities within biopharmaceutical processes,” he notes.

As biopharmaceutical solutions have expanded past protein-based therapies to include gene and cell therapies, customers are looking to quantifying more molecular species, cell species, and cell forms (live/dead) in their pursuit of a controlled production process. Fourier-transform infrared spectroscopy (FTIR) and field-effect transistor (FET) biosensor solutions are two technologies attracting interest for inline, real-time monitoring as a result, according to Semple.

Potential for inline sterility monitoring

One important aspect of bioprocess testing that has not yet benefited from inline techniques is sterility monitoring. Kistler notes that several working groups across the industry are investigating rapid microbial testing capabilities. “Once the testing is fast and accurate enough, the next step would be to move the testing away from an offline paradigm.”

“If successful,” Kistler notes, “such inline sterility testing would have the potential to reduce the existing significant turnaround time between the end of a manufacturing batch and the return of sterility test results, not only reducing release times but affording manufacturers the opportunity to terminate problematic batches prior to committing additional valuable, human, material, or other resources.”

New solutions needed for novel modalities

As more molecule types are created through cell-culture and fermentation processes, Semple expects the drive to monitor various molecules (e.g., proteins, viruses, DNA, and RNA) and be able to distinguish between various forms of those molecules (e.g., glycosylation states of target proteins, plasmid DNA, messenger RNA, and so on) will create a need for additional real-time monitoring solutions.

Indeed, both cell and gene therapies require relevant inline monitoring solutions that are currently lacking, according to Ede. “For cell therapies, sensors are needed that can measure cell density, aggregation, activity, and differentiation to guide cells into a favorable state. For gene therapies, sensors that can quantify product titer in real time are needed to accelerate process development and reduce the risk of GMP [good manufacturing practice] production,” he says.

Use of fixed-bed bioreactors for adherent cell culture of viral vectors and other virus-based therapies is, according to Dumont, creating the need for new real-time monitoring solutions as well. Univercells has several projects underway, including one that is focused on using inline glucose and lactate measurement combined with an advanced correlation model for biomass prediction in the bioreactor.

“The adoption of soft sensors for cell density monitoring provides some key benefits, especially when used in fixed-bed bioreactors,” Dumont says. He adds that the model being developed by Univercells provides data that are representative of the entire fixed bed, while some other measurement methods provide a more local measurement and can be less representative of the whole cell culture. The model is also, according to Dumont, particularly beneficial when used with a range of bioreactors that enable direct scalability, because the models created at small scale can be applied directly at larger scale.

Technical, validation, and other challenges

While inline sensing for real-time monitoring of bioprocesses is leading to improved bioreactor and process performance, the development of these solutions is not a simple task. Introduction of new sensors and PAT tools within single-use systems, in particular, is challenging with respect to optimizing sensor location, resistance to sterilization procedures (e.g., gamma or x-ray irradiation), and aseptic installation, notes Shiina. He adds that confirmation and validation of the comparability of new inline sensors with offline equipment is important, particularly if offline assays must initially be used as new technologies are implemented.

Patel points out that many of the latest sensors/PAT tools have the potential to be vulnerable to light interference. This attribute creates a challenge because plastic single-use bioreactors are not particularly well-protected from light. Another issue, he says, is the fact that disposable bioreactors can only accommodate so many probes, potentially limiting how many PATs are utilized.

Validation of the performance of new sensors/PAT tools within the matrix environment of the bioreactor must also be considered, according to Patel. “The latest tools are developed and tested within relatively pristine environments allowing for specific focus on the target being analyzed. However, the bioreactor is populated with a diverse mixture of materials, any of which may impact the performance of a sensor or probe. Understanding this challenge and having processes by which to ensure these devices are performing accurately is of upmost importance,” he explains.

In addition, Dumont stresses the need for PAT systems to be integrated seamlessly in the customer environment. “A single process often combines different equipment from different providers. Standardization in the new PAT system being introduced will be of critical importance,” he observes. New systems must also provide proper measurement in a wide range of conditions so they can be used in different processes and combined with different equipment in a flexible manner.

Other issues include the need for scalability and management of sensor life and calibration. In single-use systems, Kistler points out that there is no way to manipulate a sensor as an end user. “The industry is moving towards the integration of sensors with bag film, enabling sensors to be calibrated at the point of assembly, only. Sensors then need to be stable through the product lifecycle, from sterilization through packaging, shipment, and ultimately, use in a production batch,” he says.

Compliance needs can create obstacles as well. As an example, Semple comments that in situ sensor design must consider regulatory requirements around impact on viable cells, chemical compatibility, particle generation, and leachable/extractable profiles when in contact with process liquids, even if those liquids are to be later purified. Ede adds that single-use sensors are part of regulated products and GMP processes, that require rigorous testing and qualification before each launch, which can extend development timelines.

PAT tools must be supported by appropriate data management and analysis solutions to ensure maximum benefit of the collected data. “For all of the information from a PAT tool to be useful, the infrastructure must be in place to analyze and handle the dataset, and this consideration is often overlooked,” notes Kistler.

Ede agrees, “To make the most of the data generated, you need to feed it back into the process.” That includes tools that allow performance of multivariate data analysis (MVDA), model creation, and detection of process anomalies through real-time batch monitoring. Control software is also needed that connects various sources of process knowledge, sets process strategies, and then links those strategies to the necessary actuators, such as pumps, stirrer motors, and valves,” he adds.

Other factors that are required to turn the data into useful information, Kistler says, include the ability to access information from the PAT tools, integrating the process and PAT data together, the networking required to move data off the production floor, and the skill sets required to develop models.

AGC Biologics, in fact, has placed high importance on developing a process information system to collect data from inline monitoring and link it to process control. The company is also, according to Shiina, looking to deploy MVDA of wavelengths originating from many substances in the culture media obtained by Raman spectroscopy sensors. “In particular, we anticipate that the information obtained from MVDA will be used to control key components,” he says.

The entire goal of real-time monitoring within biopharmaceutical processes, stresses Patel, is to unlock what has previously been a black box to scientists—what is happening to cells and the environment inside a bioreactor in real-time. “While we have made great strides in building windows into the black box, blind spots still remain. In order to address those remaining blind spots, we need to create additional windows (probes/PATs) that offer even more precision and sensitivity,” he contends.

For Patel, “the most important advance we can make is enabling the real-time monitoring and analysis of the physiology of a cell inside the bioreactor, as well as getting a pretty good handle on product quality real time. This will allow us to peek inside of the cell and view transcription/translation activities, as well as its overall energy state of the cell. Ultimately, this will help us understand the conditions that best make the cell ‘happy’ and allow us to recreate those conditions to optimize biopharmaceutical processing.”

“The stronger our understanding of the biological and molecular processes during the development of biopharmaceuticals, the stronger our control of our processes and the greater the process consistency. Ultimately our understanding of biological processes will reduce lost batches, slim the number of runs needed to develop new processes based on past experimentation, and lead to a more responsive and sustainable industry. Such achievements can only come through a strong understanding of the cause-effect nature of the biological processes under control,” Semple concludes.

Cynthia A. Challener is a contributing editor to 


Vol. 36, No. 9
September 2023
Pages: 22–28, 38

When referring to this article, please cite it as Challener, C.A. Advances in Inline Monitoring for Improved Bioreactor Performance. 

Upstream Processing

Process and bioreactor performance are directly impacted by real-time monitoring capabilities.

Advances in Inline Monitoring for Improved Bioreactor Performance

Downstream processing in the biomanufacturing industry has been under relentless pressure for several decades. The source of this pressure is the much greater scalability of upstream production, which can be scaled up by increasing the productivity of cells without increasing the size of equipment or the volume of media. In contrast, the scale of downstream processing depends on the mass of product, and there is a strictly linear relationship between this mass and the size of equipment, the volume of buffers, the area of filters, and the quantity of chromatography resins required. In short, there is no economy of scale in downstream processing. Accordingly, it costs more to accommodate a larger-scale upstream process because the only solution is the addition of further downstream process trains to cope with the extra demand, which can be termed “numbering up” (1). Previously, the authors explored the need for innovation in downstream processing as a means to overcome this bottleneck and increase process efficiency and sustainability (2,3). Such innovations include the streamlining of existing processes (4), the integration of simpler and less expensive technologies that are used in other industries (5), the replacement of fixed equipment with disposable modules (6,7), and the development of high-tech solutions that are game-changers in terms of process redesign. This article explores recent progress in downstream processing, emphasizing innovative purification approaches and the growing interest in mechanistic modeling and machine learning methods for the development of precise predictive models. It also considers the incorporation of process analytical technology (PAT) and green chemistry concepts to reduce water and energy use, limit the amount of waste, and establish more environmentally conscious purification processes.

Many downstream process trains begin with a capture chromatography step because this is often the most efficient way to separate the target product from most impurities. Protein A, a cell wall protein from 

, is widely used for antibody purification due to its strong and selective binding to the Fc region of immunoglobulins. However, protein A chromatography has limitations, including high costs in early clinical manufacturing, ligand leaching, and low affinity for certain antibody variants. The demand for novel affinity ligands has increased along with the prevalence of diverse therapeutic modalities, such as antibody–drug conjugates (ADCs) and bispecific antibodies. Protein A mimetics have been developed to address these challenges because they are more stable than protein A under harsh cleaning conditions, thus reducing the impact of leaching while also saving costs (8).

As an alternative to peptides, aptamers are single-stranded DNA or RNA oligonucleotides that bind to target molecules with high specificity and affinity. Aptamers are easily synthesized and modified to enhance stability, specificity and binding capacity (9). They have been developed for the purification of coagulation factors, interferons and ADCs, showcasing their versatility and potential for diverse therapeutic modalities.

Molecularly imprinted polymers (MIPs) are synthetic polymers containing molecular recognition sites that complement target molecules in terms of shape, size and functional groups. MIPs are created by polymerizing functional monomers around a template molecule, then removing the template to leave behind specific binding sites (10). MIPs offer advantages over natural affinity ligands, including robustness, stability and lower costs. Researchers have investigated the use of MIPs to purify therapeutic proteins such as lysozyme (11) and interferons (12), although further development and optimization are needed.

Scale-down modeling has traditionally been used during process development to road-test new process variants at a smaller volume than the normal process, reducing space requirements and costs (13). A more recent innovation is ultra-scale-down (USD) bioprocessing, a powerful technique for rapid process development and optimization in biopharmaceutical manufacturing that enables experiments at much smaller volumes and with accelerated timescales, minimizing the use of time and resources and allowing multiple options to be tested in parallel (14). Miniaturized high-throughput systems, such as microscale bioreactors and microfluidic systems, are used to facilitate the efficient screening of numerous process parameters (15,16).

Microscale bioreactors closely mimic the performance of larger-scale systems while requiring less sample material, whereas microfluidic systems can precisely control the flow of small volumes of liquid (17). High-throughput experimentation can be integrated with microscale bioreactors and microfluidic systems to streamline process development even further, thus reducing the time to clinic for new biopharmaceuticals (18).

New filtration technologies have been developed to increase the efficiency and selectivity of clarification following the harvesting of cell cultures (19). Examples include single-use, high-capacity depth filters that improve clarification while reducing fouling and preventing large pressure drops (20), and advanced membrane chromatography modules, such as multimodal membranes, that facilitate the removal of aggregates and other impurities with higher selectivity and capacity (21). Fiber-based purification technologies have emerged as another alternative to traditional packed-bed chromatography, offering advantages such as shorter processing times and greater scalability (22). These technologies use hollow-fiber membranes or monolithic fibers as stationary phases.

Hollow-fiber membrane chromatography has been used to capture and polish monoclonal antibodies (mAbs) and virus-like particles, whereas monolithic fibers have been used to purify therapeutic proteins and nucleic acids with high resolution and throughput (23). Modular bed-supported cassette chromatography allows the flexible and scalable design of purification systems using interchangeable cassettes containing different chromatography media. This approach improves the performance of affinity, ion exchange and hydrophobic interaction chromatography by enhancing mass transfer, flow distribution, and media utilization. The modularity of this technology allows for the rapid screening and optimization of chromatography conditions, as well as the seamless scale-up of purification processes.

Continuous chromatography uses less resin than traditional batch chromatography but is more productive. It enables steady-state operations with a higher throughput, increasing both the purity and yield of the product (24). Examples of continuous chromatography include simulated moving bed chromatography (25) and periodic counter-current chromatography (26). These technologies have been used to purify mAbs and other therapeutic proteins (27).

As the biopharmaceutical industry moves toward the manufacturing of more complex molecules and personalized medicines, there is a growing need for the fast and reliable development of robust downstream processes. This can be challenging and expensive if there are many different parameters that need to be tested and varied simultaneously. One promising solution is the development of digital twins, which are models of real-world processes. Digital twins are constructed using mechanistic models, which are based on natural physiochemical laws that describe a process mathematically. Advanced software is now available to develop such models, which incorporate relevant effects such as adsorption and fluid dynamics. A digital twin of a chromatography process based on mechanistic models allows the industry to perform thousands of experiments 

—even outside of the calibration range—while avoiding the need for similar numbers of laboratory experiments (28).

As more data become available, machine learning has been integrated with mechanistic modeling to create more accurate predictive models (29,30). Machine learning algorithms have therefore been used to optimize chromatography process parameters (28). Dynamic end-to-end bioprocess digital twins use mechanistic models of each unit operation and integrate them into a holistic model encompassing the entire bioprocess (31). Digital twins can be used to study the impact of process changes, identify bottlenecks, and optimize overall process performance. For example, digital twins have been used to simulate the performance of continuous and integrated bioprocessing systems, providing valuable insights into process dynamics, equipment sizing, and scheduling issues (32,33).

PAT is a set of methods that are used to analyze and control pharmaceutical manufacturing processes by measuring critical process parameters (CPPs) that affect the critical quality attributes (CQAs) of a product (34). The application of PAT in biopharmaceutical manufacturing has been shown to enhance process understanding, reduce variability, and increase overall efficiency. PAT tools detect deviations as they occur and allow them to be controlled, thus ensuring product quality and consistency (35). Examples of PAT tools include Raman spectroscopy, near-infrared spectroscopy, and at-line high-performance liquid chromatography (HPLC) (36–38). Advanced analytical techniques such as mass spectrometry and capillary electrophoresis can also be used to achieve the rapid and accurate characterization of complex biomolecules (39). These methods can also detect impurities and contaminants, improving product quality and safety. Integrating these advanced analytical techniques with automation and digitalization can enable real-time monitoring and control, further enhancing overall process efficiency (40).

The biopharmaceutical industry is increasingly focused on reducing its environmental impact (41). This includes the development of more sustainable manufacturing approaches by reducing water and energy use and implementing green chemistry principles during purification (42). Water and energy consumption are significant concerns due to the extensive use of water for cleaning and cooling and the energy-intensive nature of many purification methods. To minimize water and energy use, companies are investing in more efficient equipment and processes, including water reuse strategies and the optimization of chromatography to reduce buffer consumption. Green chemistry principles are also being implemented to develop more environmentally friendly purification processes, such as the use of safer solvents and auxiliaries, minimizing waste generation, and the design of energy-efficient processes (43). For example, aqueous two-phase systems can be used as alternatives to traditional solvent-based extraction methods (44).

The rapid evolution of downstream processing in the biopharmaceutical industry is marked by continuous innovation and technological development (45). The authors have described novel affinity ligands, purification technologies, mechanistic modeling, processing modes, automation, digitalization, and sustainability as key innovation areas to address the challenges of purifying increasingly complex therapeutic modalities. As the industry continues to expand and adapt, these advances will be integrated to improve the efficiency, productivity, and quality of biopharmaceutical manufacturing processes, leading to safer and more efficacious biotherapeutics for patients, including personalized medicines. Historically, the regulators saw innovation as a threat to well-established, conservative processes but gradually came to see it in a positive light, especially following the launch of the quality by design initiative (46). By embracing the latest trends and technologies, the biopharmaceutical industry can resist the pressure on the scalability of downstream processing and can continue to revolutionize the manufacturing of life-saving therapeutics, as it has for several decades (47).

1. Aldridge, S. Downstream Processing Needs a Boost. Genet. Eng. News 2006, 26 (1).
2. Gottschalk, U.; Brorson, K.; Shukla, A. A. The Need for Innovation in Biomanufacturing. 

 (6), 489–492. DOI: 10.1038/nbt.2263
3. Gottschalk, U. The Future of Downstream Processing. 

 Supplement May 2011, s36–s41.
4. Gottschalk, U. New and Unknown Challenges Facing Biomanufacturing. 

 (3), 24–28.
5. Thömmes, J.; Twyman, R. M.; Gottschalk, U. Alternatives to Packed-bed Chromatography for Antibody Extraction and Purification. In 

, 2nd ed.; Gottschalk, U., Ed.; John Wiley & Sons, 2017; pp. 215–232.
6. Shukla, A. A., Gottschalk, U. Single-use Disposable Technologies for Biopharmaceutical Manufacturing. 

 (3), 147–154. DOI: 10.1016/j.tibtech.2012.10.004
7. Gottschalk, U. Disposables in Downstream Processing. 

, 171–183. DOI: 10.1007/10_2008_22
8. Hober, S.; Nord, K.; Linhult, M. Protein A Chromatography for Antibody Purification. 

Innovations in downstream processing accelerate development, enhance efficiency, and promote sustainability.

BioPharm International, September 2023

Nanoparticle Engineering in Drug Delivery Under the Microscope

Image Analysis Algorithm for Therapeutic mAb Aggregate Analysis

Advances in Inline Monitoring for Improved Bioreactor Performance

Innovations in Downstream Processing

Examining Next-Gen Technology for Adventitious Agent Testing

Outsourcing Formulation Services for Biologics

Clinical Trials and their Impact on the Pharma Industry

Clearing Up Confusion in Inspection Terminology

Billions of Reasons to be Buoyant