OR WAIT 15 SECS
Volume 31, Issue 2
The approaches for sample preparation of preclinical evaluation of safety and efficacy are addressed taking into consideration the shortcoming with the contemporary approaches.
Biotherapeutics are an emerging class of treatment that are produced by harnessing the protein synthetic machinery of living cells (1). These drugs have become the centerpiece of biotechnology industry and an integral part of modern medicine, evidenced by the annual global expenditure of $1.2 trillion for 2016 (2). These products also have demonstrated their effectiveness over other existing regimens for treating complex diseases such as oncology, cardiovascular, and other serious medical disabilities (3). Despite their success, the adoption of biotherapeutics is marred by issues around affordability as well as risks associated to the challenges in assessment of their safety and efficacy arising from their complexity.
In contrast to chemical-based drugs that are manufactured by simple addition of various pre-determined quantities of ingredients in an ordered manner, biotherapeutics present a different level of intricacies with respect to their production, and it is impossible to produce an exact replica of the drug in different batches originating even from the same manufacturer, let alone different producers. This may be attributed to the complexities related to the use of cell lines, media formulation, and other bioprocessing steps that are prone to variability to varying degrees (4).
The inherent complexity of biotherapeutics makes their disposition different from small molecules (5). To complicate matters further, even slight changes in product attributes (e.g., primary sequence, glycosylation profile) arising due to multiple factors (product, patient, and treatment) may culminate into serious clinical implications and elicit adverse immune responses that can lead to the death of a patient (6).
Another dimension that has been added to this discussion is the rise of biosimilars, driven by the spurt of patent expirations of biotherapeutic blockbusters (1). For this class of products, the fundamental paradigm is to rely on analytical comparability to the highest possible level without the need of furnishing extensive clinical trial data (7). This, however, highlights the need for a comprehensive, meticulous, and rigorous preclinical evaluation of safety and efficacy of a biosimilar product to assuage any risks associated to the abbreviated clinical evaluation.
Strategies for evaluation of safety and efficacy of a biotherapeutic are a topic of continuous discussion and assessment. Not all biologics can be subjected to a uniform panel of tests owing to their differences in their modes of action as well as their inherent biological/chemical complexities (8). In this 38th article in the “Elements of Biopharmaceutical Production” series, the authors focus on the approaches for sample preparation of preclinical evaluation of safety and efficacy taking into consideration the shortcoming with the contemporary approaches. Two case studies-one involving a microbial therapeutic product (granulocyte colony stimulating factor)-and other-a mammalian therapeutic monoclonal antibody (bevacizumab) have been used to illustrate the key aspects.
GCSF is a 18.8 kDa cytokine that is generally prescribed to boost neutrophil counts of cancer patients undergoing chemotherapy. It is expressed in Escherichia coli (E. coli) as inclusion bodies. Obtaining the commercial GCSF formulation requires a series of bioprocessing steps including refolding and chromatography. Often, these processing steps result in formation of certain molecular variants and impurities in addition to the pure GCSF. These include the oxidized, formyl methionine (f-Met), reduced, and aggregated forms of GCSF (10,11). From the processing perspective, while clearance of aggregates and the reduced GCSF impurity are quite achievable in most commercial processes, adequate clearance of the oxidized and f-Met GCSF is a challenge (12). In addition, it has been reported that GCSF has a free cysteine residue (cys-17) that can trigger aggregate formation, and so post-manufacturing aggregation of GCSF is a possibility (13). Recombinant GCSF processes can also yield other product-related species such as deamidated, N-terminal truncated, and norleucine forms of GCSF (14). The level of these species, however, is usually controllable through fermentation and purification steps. There is no consensus among the regulatory agencies on the maximum allowable limit for these species in the final product formulation; hence an evaluation needs to be performed on a case by case basis.
Figure 1 illustrates that preclinical material that is carefully manufactured to represent a single attribute at a time would allow one to parse the impact of these attributes on safety and efficacy; otherwise the attribute is masked in the usual preclinical and clinical samples due to low signal threshold (owing to the presence of multiple attributes in a single sample).
Figure 1. Illustration of sample preparation strategy and preclinical evaluation of the product variants (oxidized, reduced, aggregates, and f-Met). The key steps in the evaluation involved carefully preparing samples such that each samples differed from other in terms of a single attribute. Here, product information from the existing literature can be capitalized upon to gain information about various product variants, their risks and challenges. This is followed by routine analysis using an array of tests (In vitro binding assays, in vitro potency assessment, pharmacokinetics, pharmacodynamics, toxicity assessment). The data from these studies is together used to make overall risk assessment by assigning a severity score to each variant, in accordance with the principle laid down in quality-by-design (QbD) paradigm. (Figure courtesy of authors)
To prepare GCSF samples that can help unravel the effect of each individual specie, the first step is to understand how the refolding process affects the quality of the product. A full factorial design of experiments (DoE) was performed to evaluate the effect of refold pH and cystine/DTT (15). In this study, levels of monomer, oxidized, and reduced GCSF were taken as product quality attributes. The results of the DoE were used to identify conditions, which will result in formation of high level of one of the product related species under consideration. For example, the cystine/DTT ratio is governed by the number of cysteine residues in the protein and thus the number of disulfide bonds necessary for the protein to assume its functional native conformation. In this case, the results showed that both refold pH and cystine/DTT ratio have a positive correlation with levels of native GCSF and oxidized GCSF (16). However, the refold pH did not have a significant impact on formation of reduced GCSF. Further, while the cystine/DTT ratio had a positive correlation with native GCSF and oxidized GCSF, the effect was opposite on formation of reduced GCSF. In view of the above information, inclusion bodies (IBs) were refolded under two conditions favoring aggregate formation and reduced GCSF, respectively. The refolded samples were subsequently buffer exchanged and subjected to the downstream purification.
The choice of the purification platform is crucial for this study. The platform should be capable of resolving all the product related species so as to allow the collection of a pool that is enriched in a particular species of interest. Multimodal chromatography has been previously established as a suitable tool for this purpose as it resolves all the product variants of GCSF (10). This process was used to achieve the desired resolution and pooling.
Analytical characterization encompasses utilization of a collection of high resolution, high performance, orthogonal analytical tools that together offer the ability to fingerprint a biotherapeutic and monitor levels of all of the species mentioned previously. In this case, reversed-phase high performance liquid chromatography (RP-HPLC) and size-exclusion chromatography (SEC) were used. While the former quantifies levels of the oxidized, f-Met, and reduced impurities in the samples, the latter measures the levels of aggregates and fragments. Further, enzyme linked immunosorbent assay (ELISA) and PicoGreen assays were performed to measure levels of host cell proteins (HCPs) and host cell DNA (HCD), respectively.
Biological characterization involved use of assays that can assess the mechanism of action of GCSF. These included binding assays to determine the binding affinities of the GCSF variant to the pertinent receptor, GCSF-R, preferably using a label-free, optical-based technique such as biolayer interferometry or surface plasmon resonance spectroscopy; cell based assays on cell lines that express GCSF-R (Leukemic cell line, MNFS-60); pharmacokinetics and pharmacodynamics (PK/PD) on an animal model evaluation; and toxicity assays (biochemical, histopathological, and immunogenicity assays). Further, under the quality-by-design (QbD) paradigm, gene expression from the neutrophils isolated from the treated animal groups would offer us an understanding of any deviations from the usual signaling pathways for the product variants.
Several interesting observations were made from the study. First, the binding affinity of GCSF variant to GCSF-R followed the order: Reduced> Pure=Oxidized=f-Met> Aggregate. The enhanced binding affinity of the reduced GCSF samples was attributed to stabilization of one disulphide bond in the absence of the other (this explanation is drawn from the analogy of similar cytokine, Interleukin-6) (17). Second, the oxidized GCSF samples were seen to offer comparable safety and efficacy profile vis-à-vis the pure GCSF. Third, the PK attributes of reduced, aggregated and f-Met GCSF forms were inferior; however, all the tested GCSF variants were successful in inducing dimerization of GCSF-R with full activation of neutrophils. Fourth, no histologically evident damage to the major organs of tested animals using GCSF variants were observed. In the aggregate administered groups, however, muscle injury manifested as sluggishness and tilt in neck was observed. Fifth and final, based on the risk assessment of these variants, aggregated and reduced GCSF samples were categorized as critical quality attributes (CQAs), while oxidized and f-Met GCSF samples were found to be non-CQAs.
Although previous studies had established that both the disulphide bonds are necessary for biological activity of GCSF (18), it is not clearly evident that the Cys64-Cys74 disulphide bonds stabilizes the GCSF structure in absence of Cys 36-Cys 42 disulphide bond. In addition, a previous report has linked an immunogenic response to the presence of f-Met species in the GCSF sample (12). Data from this study, however, show that f-Met species differ in terms of its disposition in the body, but have equivalent binding affinity and biological activity as compared to the GCSF and its presence at levels <5% does not pose any safety/efficacy concern. Similarly, for the oxidized GCSF samples, the results obtained in this study are consistent with the previous studies (19).
Charge variants, namely acidic and basic variants, commonly exist in significant quantities in commercial formulations of mAb therapeutic products. Charge heterogeneity is typically not believed to affect safety and efficacy of a therapeutic product (21). As a result, the commonly followed approach involves assignment of a specification for the variants based on statistical analysis of the levels seen during commercial manufacturing. Thereafter, monitoring of product quality is performed to demonstrate consistency.
Traditional practice for assessing impact of charge variants on product safety and efficacy involves either use of fractions that contain a mixture of individual variants or use of individual variants that have been isolated. It is typically not possible, however, to isolate all of the charge variants individually, and if used as a pool, it is not possible to elucidate the effect of each individual variant. In addition, interactions may be possible among the different quality attributes and these also need to be understood. Recently, researchers have proposed a correlation between charge heterogeneity and glycosylation of monoclonal antibody (22).
There are multiple challenges that need to be overcome for performing such an evaluation. First, a rigorous method for separation of these variants is needed (23) and is non-trivial due to the fact that the physicochemical properties of these variants are nearly identical. Researchers have attempted to achieve this by using chromatography, both using salt (24) as well as pH (25) gradient. Although separation of basic variants has been achieved to a satisfactory extent, resolving acidic variants using either of these methods is partial at best. Second, isolating individual variants in amounts that are sufficient for further analytical characterization that is required for assignment of the modifications or performing a cell based assay for determination of the biological activity remains a time- and resource-intensive exercise (26). For instance, in the authors’ experience, at least 50 injections are necessary in ultra-high-pressure liquid chromatography (UHPLC) for isolating enough of a single variant fraction. This is equivalent to ~30 hours of instrument run time and costs more than $250/variant. The issue is further compounded by the fact that the collected fraction may or may not represent a pure isolated variant, and for almost all cases, the isolated fraction is deemed as the single assay peak that may or may not be a combination of one or more variant.
Figure 2 offers an interesting approach to achieve the desired purpose and uses the principles of separation, biology, and statistics. First, a downstream process is created for separating these species at the preparative scale. The selection of resin, as is known, is guided by the physicochemical properties of the mAb product under investigation. In this case study, the mAb product had the following characteristics: pI, 7.9-8.2; size, 149 kDa; origin: Mus musculus. Based on this information and prior experience, porous HS resin (Sulphopropyl -50 μm and residence time of 8 min) was used. The chromatographic method used a linear salt gradient for separation of various charges species of the mAb product. Fractions of fixed volume (e.g., 1 mL) were collected during elution.
Figure 2. Illustration of an approach for characterization of monoclonal antibody (mAb) charge variants. The key steps included collecting fractions during the mAb purification, assessing the charged species in each sample using the analytical CEX method, and performing cell potency assay on these samples. Statistical modelling using CEX content of each fraction as input and cell potency as output would help to parse the impact of each variant on cell potency without the need to isolate them in purity. (Figure courtesy of authors)
Second, an appropriate analytical method is created. The reversible nature of the modifications that result in the charge heterogeneity of mAbs (particularly that of acidic variants) makes them susceptible to changes over the course of storage. This necessitates the availability of a rapid analytical method that can help in determination of the number of individual charged species in each of the fractions collected during process chromatography. To achieve this task, a previously described non-linear sigmoidal shape salt gradient was used, involving use of a steep slope during elution of other components that elute before or after the main component. This results in a significant reduction in time of analysis compared to traditional methods (4 minutes with sigmoidal gradient versus 40 minutes for linear gradient). This rapid method enabled us to measure the levels of all charged species in each of the chromatography fractions. In addition, aggregate analysis was also performed in these fractions and those containing aggregates beyond a threshold level were not considered in the analysis to avoid possible compounding of contributions from aggregate and charge variants.
Third, bioassay was performed on the relevant model system (L929 cell line in this case) to determine the biological activity (anti-proliferation in this case). Fourth, empirical modeling of the resulting data was performed to correlate each of the charged variant to biological activity. Once charge variants that were seen to have statistically significant impact on activity had been identified, an empirical model using these variants was developed. The refined model was a quadratic fit incorporating significant parameters and their interactions.
Several interesting observations were made out of the study. First, seven acidic and seven basic peaks, in addition to the main peak, were resolved using the non-linear sigmoidal method. The ability to resolve the various species directly impacts the thoroughness of the resulting product understanding. The anti-proliferative activity of the fractions was found to be in the range 16-112%, thereby indicating that the number of fractions used in the analysis represents sufficient variability that is adequate to capture the impact of the variants on the activity. Second, keeping roughly the same amounts of acidic and basic variants in a typical process pool, variants were screened for impact on biological activity. It was observed that of the 14 variants under consideration (acidic and basic), only A4, A5, A6, A7, and B2 have statistically significant impact on cell proliferation, and hence these were identified as CQAs. Third, it was demonstrated that lots of mAb differing in terms of relative levels of each variant (identified as CQA) but having the same cumulative sum of acidic, basic, and main product would exhibit vastly different proliferative activity. Fourth, application of the proposed model as a dial-in-tool for identifying product pool with optimum proliferative activity and product yield was demonstrated.
Several researchers have attempted to evaluate the impact charge heterogeneity in mAbs on in vitro potency and in vivo PK (26,27). However, they suffer from the fact that they use product that contains a mixture of charge species and hence a species-specific evaluation is not possible. For example, charge variants like Lys-C variant and N-terminal pyroglutamate species, which have been recognized as non-CQAs (26), may not be there in the feed material in adequate quantities. The product understanding generated here can also be used when making decisions about pooling of process chromatography columns so as to achieve favorable process economics.
This article discusses the shortcomings with respect to the nature of samples that are used for preclinical and clinical evaluation of safety and efficacy of product variants of biotherapeutics using contemporary approaches. The two case studies presented in this article exemplify that appropriate sample preparation and customized testing using different orthogonal tools would allow to extract rich information about the safety and efficacy of the product, which is otherwise not feasible to achieve.
This work was funded by the Center of Excellence for Biopharmaceutical Technology grant from Department of Biotechnology, Government of India (number BT/COE/34/SP15097/2015).
1. A. S. Rathore, Trends Biotechnol., 27 (12), 698C-705 2009.
2. G. Walsh, Nat. Biotechnol. 2014, 32 (10), 992-1000.
3. C. Warnke, C. Hermanrud, M. Lundkvist, A. Fogdell-Hahn, Drugs Ther. Stud., 2 (1) 2012.
4. A.J. Chirino, A. Mire-Sluis, Nat. Biotechnol., 22 (11), 1383-1391, 2004.
5. W. Putnam, et al., Trends Biotechnol., 28 (10), 509-516, 2010.
6. A.S. Rathore, S.K. Singh, Protein Ther., 41-67, 2017.
7. A.S. Rathore, S.K. Singh, N. Nupur, G. Narula, In Biomarker Discovery in the Developing World: Dissecting the Pipeline for Meeting the Challenges, 83-97 2016.
8. A. Beck, S. Sanglier-Cianférani, A. Van Dorsselaer, Anal. Chem., 84 (11), 4637-4646, 2012.
9. S.K. Singh, D. Kumar, A. S. Rathore, AAPS J., 1-16, 2017.
10. A.S. Rathore, R. Bhambure, Anal. Bioanal. Chem., 406 (26), 6569-6576, 2014.
11. N. Nupur, et al., J. Chromatogr. B, 1032, 165-171, 2016.
12. R. Bhambure, D. Gupta, A.S. Rathore, J. Chromatogr. A, 1314, 188-198, 2013.
13. S. Raso, et al., Protein Sci., 14 (9), 2246-2257, 2005.
14. A. Hausberger, et al., J. BioDrugs, 30 (3), 233-242, 2016.
15. V. Kumar, A. Bhalla, A.S., Rathore, Biotechnol. Prog., 30 (1), 86-99, 2014.
16. P.D. Bade, S. P. Kotu, A.S. Rathore, J. Sep. Sci., 35 (22), 3160-3169, 2012.
17. J.N. Snouwaert, F.W. Leebeek, D.M. Fowlkes, J. Biol. Chem., 266 (34), 23097-23102, 1991.
18. H. S. Lu, et al., Arch. Biochem. Biophys., 268 (1), 81-92, 1989.
19. J.-W. Chu, B. R. Brooks, B.L. Trout, J. Am. Chem. Soc., 126 (50), 16601-16607, 2004.
20. S. K. Singh, G. Narula, A. S. Rathore, Electrophoresis, 37 (17-18), 2338-2346, 2016.
21. Group, C. M. C. B. W.; others. Emeryville, CA CASSS 2009.
22. J. M. Yang, et al., Anal. Biochem., 448, 82-91, 2014.
23. V. Joshi, V. Kumar, A. S. Rathore, J. Chromatogr. A, 1406, 175-185, 2015.
24. S. Fekete, et al., Pharm. Biomed. Anal., 113, 43-55, 2015.
25. S. Fekete, et al., J. Pharm. Biomed. Anal., 102, 282-289, 2015.
26. L. A. Khawli, et al., MAbs, 2 (6), 613-624, 2010.
27. Y. Y. Zhao, et al., PLoS One, 11 (3), e0151874, 2016.
Vol. 31, No. 2
When referring to this article, please cite it as A. Rathore et al., "Preclinical Evaluation of Product Related Impurities and Variants," BioPharm International 31 (2) 2018.