Author: Gjalt Huisman, Vice President and Head of Biotherapeutics
Lead optimization for discovery of small molecule drug candidates is well established, but less so for biological drug candidates. While biologics are much easier to modify than non-natural small molecules via introduction of changes in their amino acid sequence through genetic engineering technologies, this approach is frequently met with trepidation. Generally, the risk-benefit balance of small improvements resulting from a few mutations tilts towards potential safety concerns. However, large improvements in biologics function can drive significant benefits in efficacy and safety and are thus highly relevant to patients and should be extremely attractive to the pharmaceutical industry. Up until now, such improvements have been difficult to generate.
Lead optimization in pharmaceutical discovery: small molecules vs. biologics
Biological lead compounds, as soon as they have been identified as having desired therapeutic activity, generally undergo little further modification to improve potency and physical characteristics. In contrast, small molecule drug leads typically see substantial optimization in order to optimize potency, bioavailability, and minimization of off-target effects. For example, atorvastatin was the fifth HMG-CoA reductase inhibitor to enter the market nine years after lovastatin and became the best-selling drug in the world in 2003. Statins all have a common 3,5-dihydroxyhexanoate (DHH) core, with remarkable chemical structure variation at the 6-position. Many variants of the DHH-core have been generated and tested, and, ultimately, eight statins reached the market. Ideally, such highly efficacious compounds would have been identified earlier.
Such a compound-rich historic timeline is common for small molecule drugs, but very rare for protein drugs, with the exception of antibodies. Antibody optimization is routinely practiced, relying on the ability to sort out (select) antibodies with desired properties (affinity, specificity, developability, stability) via ultra-high-throughput screening protocols. The key principle that supports this workflow is the fact that the antibody and the DNA encoding it can be linked throughout the procedure using display (virus, yeast) technologies. In contrast, optimization of non-antibody biological candidates is largely restricted to post-production modification (for example, by PEGylation or genetic introduction of additional sequences such as PASylation or XTEN technologies) to increase half-life and mask potential immunogenic epitopes. Otherwise, improvements are sought in formulation development to minimize stability liabilities or post-production aggregation, etc.
Antibody engineering approaches have clearly been successful: In 2016, 27 of the top 200 drugs contained an antibody, either as stand-alone, fusion, or drug-conjugate (MedAdNews 2017), while small molecule drugs numbered 136. Other biologics in this list include protein drugs (n=20), vaccines (n=8), and peptide drugs (n=10) including seven insulins. While antibody sequences are typically optimized for function, other biopharmaceuticals, therapeutic enzymes, and other therapeutic proteins are not. Do such un-optimized biopharmaceuticals provide sufficient efficacy to patients and completely resolve the symptoms of disease? We suspect, by and large, they do not.
The reason biologics are not as effective as one would like is related to the process that generated these molecules: Darwinian evolution. Darwinian evolution provides a beautiful interplay between the internal and external environments of a living organisms to arrive at a state where the organism can thrive. However, Darwinian evolution in man is relatively slow and, with the exception of the immune response, no mechanisms have evolved naturally to treat modern disease. Insulin is naturally secreted by the pancreas to regulate glucose levels; the molecule did not naturally evolve to be administered as an injectable to diabetes patients. Similarly, lysosomal enzymes mature intracellularly as they are transported from the endoplasmic reticulum, via the Golgi apparatus, to the lysosome; they did not naturally evolve to find the lysosomes in all relevant tissues from an infusion bag every other week for the treatment of inborn errors of metabolism such as Gaucher, Fabry, or Pompe Disease.
Novel technologies are required to provide biotherapeutic candidates that optimally meet the need of the patient. Directed evolution of proteins is a well-established engine to discover enzyme catalysts for small molecule drug manufacturing (see Nature 2010 for a review). This body of work has established that enzymes can be modified at >15% of their primary sequence, resulting in many orders of magnitude performance improvement from a combination of increased activity, specificity (affinity), stability, and others. For small molecule manufacturing, lead optimization is focused on variables related to pharmacokinetics and pharmacodynamics such as bioavailability, half-life, activity, and selectivity. These characteristics are critically important for biologics as well, and directed evolution using relevant high-throughput assay technologies can provide greatly improved biological lead candidates.
Directed evolution comprises the accumulation of beneficial mutations in the protein of interest to generate a protein variant with the desired attributes for a specific application. As in natural evolution, directed evolution is a continuous, iterative process where different mutations accumulate combinatorially when screened for desired therapeutic function. In its most effective forms, directed evolution combines high-throughput (HTP) molecular biology, HTP screening, and HTP sequencing, supported by a laboratory information system and bioinformatics infrastructure to orchestrate and coordinate a highly efficient workflow. As a result, greatly improved proteins can be generated in short periods of time. This is illustrated in the example below.
Biologic lead optimization for function in the upper intestines
The environment of the upper intestines, specifically the duodenum and jejunum, provides a harsh environment in which polymeric food components are efficiently degraded to monomers and oligomers by the action of hydrolytic enzymes. Proteins are degraded to short peptide fragments and free amino acids, polysaccharides to individual sugars, and fats to fatty acids and glycerol by the action of proteases, amylases, and lipases. While this digestive process has naturally evolved to be highly efficient, in some disease settings it would be desirable to augment it to remove toxic metabolites from within the confines of the GI-tract, thereby minimizing exposure to the patient.
Almost 40 years ago, Hoskins et al. explored the use of an encapsulated enzyme to remove phenylalanine in the GI-tract for the treatment of hyperphenylalaninemia (also known as phenylketonuria or PKU). Several studies followed, in which the enzyme was protected from proteolysis in the duodenum and jejunum by various immobilization, encapsulation, and chemical modification (PEGylation) technologies, and rational protein engineering, ultimately demonstrating that the lack of a readily obtainable, GI-stable enzyme prevented the realization of this approach. However, using our technologies and screening of more than 50,000 proteins over eight rounds of directed evolution led to the identification of an enzyme that is sufficiently stable in the GI tract of dogs and monkeys to remove clinically relevant amounts of phenylalanine (U.S. Patent 9,611,468).
This GI-stable, phenylalanine degrading enzyme was obtained in about nine months using a combination of structure-guided library design, HTP molecular biology, HTP screening, and HTP sequencing. The rationale that was followed included site-saturation mutagenesis of the surface of the protein to identify new variants in which proteolytic sites had either been removed or made inaccessible for the GI proteases, trypsin and chymotrypsin. In addition, more stable variants were identified by subjecting libraries to heat treatment, ultimately culminating in an enzyme that is readily manufactured at large scale. The final enzyme that is slated to enter the clinic in 2018 contains >20 mutations. Exactly why this highly evolved enzyme is so protease stable is unknown and the total number of predicted trypsin and chymotrypsin cleavage sites barely changed. Clearly, and not unexpectedly, it is not just the presence of such sites at the surface that determines the enzyme’s propensity to be proteolytically degraded, accessibility of such sites is a key component as well.
Advanced directed evolution technologies hold tremendous promise to deliver new biologic lead candidates. Biologics are seldom a panacea for disease treatments and moreover always difficult to manufacture due to their natural instability. Biological leads can now readily be optimized for increased efficacy and manufacturing. Increased stability of biologics whether to serum, to the intracellular environment, to conditions in the lysosome, or to the GI-tract, ultimately have a significant impact on the efficacy of the drug.
Immunogenicity is generally a first concern when considering the potential consequences of protein sequence changes. However, let us not forget that immunogenicity is the result of a process in which proteins are degraded to smaller fragments, some of which have a high affinity to the major histocompatibility complex II (MHC-II) and are presented to the T-cells. The dynamics of this essentially chemical process are fully determined by the primary sequence of the protein and hence should be addressable with the right technology. Advanced directed evolution has arrived at the doorsteps of making this happen.