The 5th International Symposium on Higher- Order Structure of Protein Therapeutics (HOS 2016) attracted 112 scientists and vendors to the Renaissance Hotel in Long, Beach, Calif., April 10–13. HOS meetings focus on developing and applying a range of analytical technologies to characterize the structures related to the activity of a protein. Protein activity may vary, depending on its structure and that of its interaction partners. As a protein approaches its natural state, experimental difficulty increases dramatically. Scientists clearly need better tools, some of which were demonstrated at HOS 2016, reviewed here.
Background
Protein structure is organized into five levels. The first level, primary structure, is the configuration of amino acids as they are linked together. Secondary structure refers to the conformation of individual proteins in space, and includes sheets, helixes, tubes and bends. Tertiary structure is the detailed arrangement of proteins in three-dimensional space. Proteins often associate to form larger structures that are responsible for immunogenicity and other signaling roles; this is known as the quaternary structure. The fifth, or quinary level, refers to the closeness of other molecules, particularly within cells or concentrated solutions such as liquid formulations. These short-range interactions are weak and are particularly difficult to study with current technology.
Quinary structure describes the protein structure in a confined, concentrated matrix typical of living cells or injectable formulations. This contrasts with studies using traditional proteomics in use over the last 20 years, which focused on detecting and assaying rare biomarkers at ppb, ppt or even single-molecule concentration. A typical quinary matrix has solids ranging from 10 to 30 wt%. This matrix is experimentally difficult since much chemistry is carried out in dilute solutions. Weak interactions in concentrated liquids have received less attention.
Debye-Hückel theory, developed 100 years ago, does not extend in a useful way. One poster at the meeting presented evidence that ionic interactions were important when ions are present, as they usually are.
Student grant recipient Rachel D. Cohen, University of North Carolina at Chapel Hill, described how nuclear magnetic resonance (NMR) studies of amide proteins are affected by the intracellular pH of cells from E. coli. As pH decreases from neutral, the interaction strength between the B domain of protein G and other intracellular proteins increases, which adds to the effective weight of protein G. NMR was able to measure the free energy for unfolding as a function of pH. Cohen concluded that quinary interactions have an electrostatic component.
The technical program illustrated the shortcomings of the best-available analytical technologies for characterizing quinary structure and activity. Specifically, current technology provides only fleeting glimpses where a detailed view is needed.
Characterization of binding epitopes
Proteins interact with each other at specific locations, often with distinct lock-and-key structures. Binding characteristics can change with differing starting materials, for example, cell lines, and with changes in process and storage conditions. Biosimilars present a particularly difficult problem, since follow-on drugs are developed without help from the original product’s innovators.
Hydrogen-deuterium exchange probes structural similarity
Regulators want to make sure that proposed changes in biosimilars do not alter the drug’s mechanism of action, including its efficacy. One of the earliest techniques used to develop biosimilars was the exchange of hydrogen ions with deuterium ions on amide amino acids (HDX). If two sample proteins react differently, they probably have a different structure, but if they react similarly, they may have a comparable structure. Each test that shows similar results for the original product and the biosimilar add to the evidence that the old and new are not significantly different. A poster by Peter Li (Bioanalytix Inc., Cambridge, Mass.) described HDX-MS as an orthogonal approach to the characterization of the critical quality attributes of product variants such as deamidation, oxidation, disulfide scrambling and other post-translational modifications, including glycosylation.
A more focused study from Medimmune (Gaithersburg, Md.) used ion mobility separation (IMS) and HDX-MS for stress-testing an IgG4 protein. The protein was held at a range of temperatures for up to 35 days; resulting data was consistent with several degradation pathways. IMS plus HDX-MS provided key information leading to the identification of degradation products and associated pathways. In several cases, a drug product was modified to improve stability or optimize process conditions to reduce degradation pathways.
Waters (Milford, Mass.) automated HDX-MS by linking an ACQUITY UPLC M-Class System with HDX technology with a mass spectrometer such as the Xevo G2-S QTof and DynamX application-specific software. Waters has also developed an in-line immobilized trypsin column for peptide digestion.
Oxidative footprinting probes structure
HDX is effective for amide hydrogens, but proteins have additional atoms. Yining Huang of Washington University (St. Louis, Mo.) described epitope mapping with a process called fast photochemical oxidation of proteins (FPOP). With FPOP, a small quantity of hydrogen peroxide is added to a protein solution. Upon irradiation with a laser, hydroxide radicals (OH•) are generated. The OH• is very reactive and replaces bound protons to form an alcohol. MS can locate the vulnerable sites in a manner similar to that of HDX, except OH• substitution is faster and more general.
Following tryptic digest of the protein, LC/MS of the derivatized peptides is identified by a Byonic search engine (Protein Metrics, San Carlos, Calif.), which feeds the information to Protein Metrics Byologic software to record the degree of hydroxylation for each amino acid. Software performance was compared with manual processing; the profiles are consistent, but the software is much faster. Results are generally consistent with other mapping studies with HDX and X-ray crystallography.
With a synchrotron light source, it is possible to use a short pulse of X-rays to generate the hydroxyl radical. This provides even better discriminating power than NMR, electron microscopy and light scattering, according to Corie Ralston of Lawrence Berkeley National Laboratory (Berkeley, Calif.). Preliminary work with the advanced light source showed that microsecond pulses are ideal for studying the structural dynamics of complex biological macromolecules.
Protein conformation array ELISA
Characterization of binding epitopes is important to show equivalence in biosimilars and bridge studies involving process changes, including cell lines. Array Bridge Inc. (St. Louis, Mo.) presented a poster describing the Protein Conformation Array (PCA) ELISA. Starting with the amino acid sequence of the biologic, overlapping peptides of the sequence are created, and polyclonal antibodies are raised to the short length of peptide. Combining all the polyclonals in an ELISA is a highly discriminating test when performed over the entire length of the target protein. The poster described how the PCA ELISA can be adapted to xMAP technology (Luminex, Austin, Texas) for multiplex workflows with comparable results. The poster stated: “The PCA-ELISA offers a simple, robust and sensitive methodology for comparative analysis of higher-order structure of biosimilar and innovator mAB molecules in all stages of development.”
Spectroscopic techniques for comparing structure
NMR
The high specificity of NMR makes it ideal for characterizing proteins. A report by Brad Jordan (Amgen, Thousand Oaks, Calif.) showed that NMR of antibody fragments often remains unchanged when the fragments are conjugated with linkers and drugs. Similarly, the drug actives seem to have conserved NMR patterns. Thus, the signals for the building elements can be recorded and then subtracted from the antibody–drug conjugate (ADC), making the entire process much more tractable.
A poster by Mats Wikstrom and colleagues at Amgen used NMR to study the similarity of higher-order structures. A 1-D profile provides a fingerprint. Line-shape analysis can show structural differences that are less obvious when obtained by circular dichroism (CD), FTIR and differential scanning calorimetry (DSC). Two-dimensional NMR is even more effective for revealing information-rich details.
Recent advances in NMR technology, particularly 900- and 600-MHz instruments, provide useful signals for naturally abundant 1H-13C and 1H-15N. Robert Brinson, of NIST’s Institute for Bioscience and Biotechnology Research, showed how changes in structure can be seen clearly by focusing on methyl groups. Amide protons are also useful, but the signals from methyl groups are cleaner. NMR is particularly useful in establishing consistency in production and process bridging, including work with biosimilars. A 900-MHz instrument provides better signal quality and detection; with research done and production started, 600-MHz instruments are generally suitable.
Circular dichroism
Circular dichroism is used to characterize the optical activity of substances. The measurable spectrum ranges from the far-ultraviolet to the low-infrared, and includes Raman. With the appropriate spectral region, CD seems to be the preferred technology for monitoring changes such as accelerated degradation, stability and hold times. The technique is especially useful for studying secondary protein structure such as a helix. For example, temperature studies of proteins show large changes in CD spectra with heating, as the secondary structure is lost.
JASCO (Easton, Md.) featured the entry-level J-1100 CD spectrometer for QA/QC applications and the J-1500, which offers accessories that enable rapid scanning to support research applications. The company also offers the J-1700 CD spectrometer, which covers the spectral range from vacuum-ultraviolet to the near-infrared.
The wavelength of light in CD is a key differentiator. A report from Applied Photophysics Ltd. (Surrey, U.K.) focused on degradation studies of IgG1. Ultraviolet spectra from degraded and control samples were measured with a Chirascan-auto qCD spectrometer. The spectra appeared similar but, when working with a 2-sigma confidence interval, the degradation products were clearly differentiated in the near-UV spectra in all but one case. No significant changes could be demonstrated for the far-UV CD spectra. The authors attributed this to small changes in the local environment of the chromophores, as opposed to major changes in the peptide backbone, and said that the ability to “generate quantitative CD data will substantially strengthen the role of CD analysis throughout biotherapeutic development programs.”
Super-resolution electron microscopy
NanoImaging Services (San Diego, Calif.) displayed images of therapeutic aggregates, viruses, antibodies, protein complexes and engineered biotherapeutics that revealed size, shape and morphology. Resolution was sub-3Å, which is sufficient to show antibody–antigen interactions as well as the single-particle structure of viruses. With cryo-EM, seeing is believing. Sizing and counting experiments can be performed, including tests for aggregation. In at least one instance, the key step is to form trimers, which can dimerize to give a hexamer.
Microcalorimetry and other technologies
Several presenters cited differential scanning microcalorimetry for comparing structures in physical-stability studies. The plots provided are empirical, and can be useful in revealing differences.
Conclusion
Scientists are struggling to classify materials as being “the same” or “different.” As more definitive technologies are developed, they will be reported at subsequent HOS meetings (www.casss.org). Special thanks to Ms. Rene Olsen for leading the CASSS team in organizing HOS 2016.
Robert L. Stevenson, Ph.D., is Editor Emeritus, American Laboratory/Labcompare; e-mail: [email protected].