2D-NMR, an established analytical tool akin to magnetic resonance imaging, reliably assessed the atomic structures of biologically similar products, yielding the equivalent of a fingerprint for each. This was the first ever inter-laboratory study of four versions of a therapeutic protein drug all manufactured from living cells.
The findings demonstrate that two-dimensional nuclear magnetic resonance spectroscopy, or 2D-NMR can be a robust and powerful complementary technique for companies and regulators when assessing these biosimilars, according to Robert Brinson, a research chemist at the National Institute of Standards and Technology (NIST). This type of assessment is part of a set of comparisons required to determine whether a follow-on biological product is highly similar to an existing product, so that there is no “clinically meaningful” difference between the two.
“Other analytical methods provide useful information, but 2D-NMR is one of the few approaches that can yield complete assignment of three-dimensional structure across the entire molecule in solution at atomic-level resolution,” Brinson explains. “Our study indicates that 2D-NMR data can yield a precise and unique ‘fingerprint’ of structural information in a biological product.”
Results were reported for measurements of four independently manufactured versions of filgrastim, a biological drug used to help ward off infection and anemia in cancer patients. At four laboratories, researchers used 2D-NMR to map the atomic structures of the original filgrastim product licensed in the U.S. and three unapproved biosimilar versions.
A biosimilar, according to the Food and Drug Administration (FDA), is a biological product shown to be “highly similar to an FDA-approved biological product, and has no clinically meaningful differences in terms of safety and effectiveness.” Only minor differences in clinically inactive components are allowable in biosimilar products.
Biosimilar versions of approved biological drugs at the end of their patent life are expected to cost less but be as safe and effective for licensed clinical uses. To date, the FDA has approved one biosimilar (a version of filgrastim), while the European Union has approved about 20 biosimilars over the last 10 years.
Unlike chemically synthesized drugs – aspirin, for example – biological drugs usually are composed of large, complex protein molecules and are produced by living systems. This makes producing exact duplicates impossible, even from batch to batch in the same biomanufacturing process.
For specified health conditions and symptoms, the nearly exact copies that result must be shown to achieve the same clinical effects as the already-licensed biological product.
In addition to reporting on the utility of 2D-NMR for high-precision measurement of the detailed atomic structure of biosimilars, the new paper describes statistical methods used to assess biosimilarity. They include one for rapid analysis of many datasets, which can be generated, for example, when monitoring batch-to-batch variation during production.
In the next phase of the work, Brinson says, NIST and collaborators will compare 2D-NMR measurements of a monoclonal antibody – molecules able to bind to specific targets such as cancer cells – that NIST is developing as a reference material.
Monoclonal antibodies are the largest class of approved protein therapeutics in the world, and the ability to extend 2D-NMR methods to this class of therapeutic would represent an important landmark in their analytical characterization. Thirty laboratories on five continents will participate in the upcoming project.
Beyond ascertaining the precision of 2D-NMR across a large network of laboratories, the effort is expected to yield a catalog of best practices to ensure the reliability and repeatability of results.
Houman Ghasriani, Derek J Hodgson, Robert G Brinson, Ian McEwen, Lucinda F Buhse, Steven Kozlowski, John P Marino, Yves Aubin, David A Keire
Precision and robustness of 2D-NMR for structure assessment of filgrastim biosimilars
Nature Biotechnology, 2016; 34 (2): 139 DOI: 10.1038/nbt.3474
Image: Protein Data Bank