June 15, 2007 (Vol. 27, No. 12)

Trevor Stokes

How to Choose the Best Target, Optimize Delivery, and Reduce Immunogenicity

One challenge for protein engineers is to produce biopharmaceuticals that can withstand the proteolytic conditions of the gastrointestinal tract and serum. This was one of many challenges discussed at Cambridge Healthtech’s “PEGS Summit” held last month in Boston. To meet that demand, Manuel Vega, CEO of Nautilus Biotech (www.nautilusbiotech.com), described a 2-D scanning approach that identified a single residue change that could confer proteolysis resistance in plasma and tissues in Nautilus’ two lead products—interferon alpha (Belerofon) and human growth hormone (HGH; Vitatropin).

The approach includes experimental scans of proteolytic sites across the entire linear peptide sequence. “In the end, we have been able to identify all of the sites for proteolysis that matter,” said Vega, “These are sites that have an impact on the in vivo half life.” Researchers at Nautilus validated Belerofon, Vitatropin, and a few other proteins in vivo in multiple animal models, including rat, monkey, and mouse for pharmacokinetics and biological activity.

Though Nautilus reduced proteolysis of interferon alpha, HGH, and other proteins with single point mutations, Vega admitted that some proteins may need more than one mutation at proteolysis sites. Vega said Nautilus researchers successfully tested this approach on proteins that ranged in size from 80 amino acids up to 450. Proteolysis-resistant variants also showed improved PK profiles. “That means that proteolysis plays a role in the process of clearance of circulating proteins,” said Vega.

Proteolysis-resistant proteins show extended half-lives in the gastrointestinal tract as well, thus favoring absorption through the blood barrier. This, Vega said, opens the door for oral administration of therapeutic proteins.

Therapeutic Protease Engineering

Protease activity is not always a problem to be avoided. Human proteases themselves can be exploited as biologics. The major problems with many applications of so-called therapeutic proteases, however, are that they are not active enough, not specific enough, or are too rapidly inactivated by the highly active serine protease inhibitors, SERPINs, naturally present in human serum.

Wayne M. Coco, vp of protein engineering at Direvo Biotech (www.direvo.com), presented a strategy that uses combinatorial libraries and confocal fluorimetry screening to engineer a nonspecific human protease, trypsin, for high specificity against anti-TNFalpha, a cytokine involved in systemic inflammation. The Direvo scientists accomplished this goal while creating a trypsin variant with an 80-fold increase in specific activity and with several thousand-fold less inhibition in 100% human serum (IC50), according to Coco.

“Our unique combination of library strategies with confocal screening in 100% human serum represents a general platform for the comprehensive optimization of therapeutic proteases,” said Coco. “We can do highly miniaturized, high-throughput screens in formats that are predictive of ultimate application.”

Combinatorial libraries for variant testing included whole gene random mutagenesis and libraries focused on either individual residues or protein domains. “From the improvements that we get in those initial libraries, we recombine the best mutations with proprietary DNA-shuffling techniques,” said Coco.

The Direvo screening technique uses confocal laser technology developed by Evotec (www.evotec.com) and exclusively owned by Direvo for protein engineering. The confocal laser not only detects changes in fluorophore-labeled protein populations, but changes at the individual molecule level, thus contributing to miniaturization, throughput, and unique assay formats.

One goal for protein-based pharmaceutical developers is to reduce the risk of immunogenicity. As part of the central human histocompatibility system, HLA proteins and their binding motifs offer a potential point of prediction of how an individual will immunologically react to a foreign protein.

To address immunogenicity risk, Umesh Muchhal, protein engineering group leader at Xencor (www.xencor.com), presented ImmunoFilter®, an experimentally derived database of HLA-binding prediction matrices that the company says covers at least 90% of the U.S. population.

To generate ImmunoFilter, researchers characterized all peptide sequence combinations of synthetic nine-mers, the minimum peptide length that binds to the MHC II class of molecules. The matrix excluded cysteine to avoid disulphide-bond complications, but the group extrapolated results from threonine, valine, and alanine.

After binding to 171 randomized libraries, researchers used a competition-binding experiment with a reference-labeled peptide, and in the process generated a matrix of binding propensities. “That is the beauty of these propensities, even though it takes a lot of work, the final tool is a software package where you plug in your protein, hit the button, and instantaneously you have a full report of HLA epitope propensities across the whole protein,” said John Desjarlais, vp of research.

One of Xencor’s core platforms, XmAb™ Fc technology, increases antibody potency through the Fc domain. “A nice application of ImmunoFilter has been the demonstration that in spite of the fact that we have engineered the Fc domain, according to ImmunoFilter, we have not put in immunogenic substitutions,” said Desjarlais. “The other aspect of our analysis has been to prioritize, according to publicly available data, which alleles to characterize. That is based on population penetrance.

“By characterizing about 32 HLA-DRs we can cover over 90 percent of the U.S. population,” said Desjarlais. “It’s almost impossible to cover all of them, but 90 percent is not so bad.” Centocor and Eli Lilly currently have licensing agreements with Xencor for ImmunoFilter.

Lysosomal Storage Diseases

Therapeutic improvements are a driving force behind protein engineering, but it remains important to understand the underlying biology. Tim Edmunds, vp of therapeutic protein research at Genzyme (www.genzyme.com), focused on lysosomal protein targeting, in which enzyme replacement therapy is already available for several lysosomal storage diseases such as Gaucher disease.

Gaucher disease is an inherited metabolic disease where people accumulate fatty glucocerebroside, which prevents cells and organs from functioning properly. Blood transfusions with glucocerebrosidase can reverse several disease phenotypes such as increased liver size and abnormal blood counts. These transfusions of glucocerebrosidase at high doses (15–60 units/kg) need to be done every two weeks. Cerazyme is taken up by macrophages in a matter of minutes.

Unlike cell-surface proteins, the cell must take up lysosomal storage proteins. “That generates some significant challenges in terms of delivery to target cells. For lysosomal storage disease proteins, carbohydrate receptor density varies from cell type to cell type, which poses another challenge,” according to Edmunds.

Another challenge is immunogenicity, since patients who genetically lack glucocerebrosidase will detect the transfused enzyme as foreign. “We have used a variety of techniques now to look at modifying lysosomal storage disease proteins, mainly focusing on glucocerebrosidase.

“In order to be efficiently delivered to the macrophages, the carbohydrates need to be remodeled,” Edmunds said. Genzyme says it accomplished this with Cerazyme, the company’s recombinant product, where sugar moieties on the protein were cut back to expose terminal mannose residues, increasing macrophage uptake twofold.

To increase mannose-binding sites, researchers expressed recombinant glucocerebrosidase in baculovirus cell lines and in the presence of glycosylation inhibitors. The result was a series of mannose chain lengths from three to nine, which surprisingly had little impact on uptake by distribution of protein.

The researchers also examined potential amino acid changes to lessen the dosage regiment. To increase the half-life, researchers have PEGylated the protein variants that are currently being studied in animal model systems. “We are trying to understand the biology and the mechanism. This is by no means an imminent product,” said Edmunds.

Tyrosine Kinase Receptors

For target identification, high throughput is a must. MedImmune (www.medimmune.com) says that its new filter-based approach improves throughput since each 10-cm filter contains about 30 million antibody phage-expression library clones. “We can screen an entire antibody library within a few days,” said Herren Wu, vp of antibody discovery and protein engineering.

As clone densities increase, so does the chance of confluence, which makes individual clones impossible to distinguish. By using high-affinity tagged antibody fusion proteins, researchers can overcome this hurdle and increase the screen’s sensitivity. MedImmune focused on ephrin A4 tyrosine kinase receptor, implicated in cancer cells as a model system, and found a target within one week. Normally the process can take up to a month from substrate to target. They identified an antibody that binds ephrin, phosphorylates the receptor, and causes its degradation in cell-based assays.