Insights From Aggregation Intermediates

Margaret Speed Ricci*
*Amgen Inc
One Amgen Center Drive
Thousand Oaks, CA 91320

My story begins more than half a lifetime ago in the 1991-1996 timeframe. As a Course X student of the BioProcess Engineering Center (BPEC), I started my thesis work focused on thermodynamic modeling of aggregation with Professor Daniel Wang (RIP). After a semester or so of enthalpy and entropy calculations, I pivoted my attention from in silico thermodynamic modeling to studying real world kinetic intermediates and found myself in Professor Jonathan King’s lab. Jon had a history of adopting stray engineering students like me and immersed us in the art of discussing experimental approaches and philosophy/politics over coffee prepared 3X Greek style with my mentors and collaborators, Anna Mitraki, Cammie Hasse-Pettingell, Carol Teschke, Peter Privilege, Barrie Greene, Susan Sather, Carl Gordon, Anne Robinson, Scott Betts, and Cindy Woolley (who kept Jon and us in line).

Discovery of aggregation intermediates

My approach to my graduate research was to have 3 or 4 experimental strategies on the burners in hopes that one might pan out, and if/when that happened, to run with it. In exploring folding and aggregation pathways of P22 tailspike protein, I had the opportunity to study both in vivo and in vitro folding, assembly, and off-pathway mechanisms aggregation. Along the way, I utilized a plethora of analytical techniques (as a learning opportunity and out of desperation), including traditional denaturing SDS polyacrylamide gel electrophoresis, non-denaturing gel electrophoresis, Western blotting, dynamic light scattering, and size exclusion chromatography. Most importantly, I poured literally hundreds of gels by hand and gowned double layers of lab coats to load samples into wells and run those gels in the cold room. This was a significant character-building exercise in my rite-of-passage as a graduate student.

Then one day, I vividly remember silver staining a nondenaturing gel and seeing aggregation intermediates appear like rungs on a ladder with Anna anxiously looking over my shoulder. This advancement was possible by standing on the shoulder of my predecessors, who had done extensive research on the intermediates for productive folding and assembly. For this new chapter in our folding/aggregation storybook, this result was the pivotal moment that proved that we could trap a sequential series of multimeric intermediates on the aggregation pathway.

Conformational insights

This ability to monitor aggregation intermediates provided us a way to test interesting hypotheses on the mechanism of folding and aggregation. We could interrogate all sorts of conformational questions using monoclonal antibodies against tailspike chains that discriminated between folding intermediates and native states, as characterized by previous researchers in the P22 community.

By nondenaturing Western blot, we found that the aggregation intermediates displayed nonnative epitopes in common with productive folding intermediates but were not recognized by antibodies against native epitopes. The nonnative epitope on the folding and aggregation intermediates was located on the partially folded N-terminus, indicating that the N-terminus remained accessible and nonnative in the aggregated state.

Furthermore, we were able to block productive folding by adding monoclonal antibodies against the native epitopes to the in vitro refolding solution. In contrast, antibodies directed against the N-terminal nonnative epitope did not block folding, thereby indicating that the conformation of the N-terminus was not a key determinant of the productive folding and chain association pathway.

We had proved that the aggregation intermediates represented discrete species formed from noncovalent association of partially folded intermediates rather than aggregation of native-like trimeric species.

Specificity of aggregation

The general scientific consensus of that era regarded aggregation as nonspecific coagulation of incompletely folded polypeptide chains to form a disordered precipitate. However, various observations of amyloidosis and inclusion body formation suggested that aggregation of partially folded chains may actually proceed through specific interactions.

It had been difficult to determine experimentally for populations of insoluble aggregates whether each agglomerate was composed predominantly of one particular protein or of heterogeneous species. We circumvented this difficulty by characterizing the early soluble aggregation intermediates along the pathway and interrogated the mechanism by which self-assembly occurred for both the P22 tailspike as well as the P22 coat protein with Carol Teschke.

We demonstrated that for a mixture of proteins refolding in vitro, folding intermediates did not coaggregate with each other but only with themselves. Western blot analysis was used to characterize in vitro aggregation samples electrophoresed through a nondenaturing gel and probed by tailspike monoclonal antibodies or coat polyclonal antibodies. The ladders of aggregation intermediates that formed in the mixed refolding samples were remarkably similar to the controls rather than to a smear of various combinations of heterogeneous coaggregates. This suggested that aggregation occurred by specific interactions of certain conformations of folding intermediates rather than by nonspecific coaggregation. Therefore, we had proven the specificity of aggregation of soluble multimeric intermediates.

Mechanism of aggregation

As a chemical engineer, I was obligated to probe the mechanism of polymerization. We wondered whether the polymerization of misfolded chains occurred by sequential addition of monomeric subunits to the growing aggregate, as observed for various systems of biological self-assembly. Alternatively aggregation could potentially occur through multimeric cluster-cluster polymerization reactions, in which two multimers of any size could associate to form a larger aggregate.

Examination of the distributions of aggregation intermediates resolved by nondenaturing gel electrophoresis revealed that the polymerization reaction continued after the monomeric folding intermediates had been depleted. Aggregation could proceed without the presence of monomers by the association of multimers of various sizes via a cluster-cluster mechanism. Kinetic modeling of the gel densitometry data confirmed that aggregation occurred via the multimeric polymerization rather than the sequential polymerization mechanism.

To confirm this modeling by direct measurement, we utilize two-dimensional nondenaturing gel electrophoresis, in which a lane of multimers resolved on a nondenaturing gel was allowed to aggregate within the gel before electrophoresing in the second dimension. The 2D gel showed a prominent diagonal pattern of bands representing monomer, dimer, trimer aggregate, tetramer, and larger multimers. A series of higher ordered multimers were observed above the species on the diagonal in the gel. This pattern indicated that the monomers associated to form higher aggregates, dimers formed tetramers, and trimer aggregates formed hexamers. This was direct experimental evidence that aggregation occurred by multimeric polymerization. If sequential polymerization was required for aggregation, then only the monomers would have self-associated.

Reflections on MIT experience

As my graduate work at MIT came to a close, I felt fortunate to have Professors Danny Wang and Jonathan King as co-advisors. Having said that, I don’t know of any two people who were more dissimilar in terms of personality, politics, or philosophy. Jon had a penchant for exploring wherever the science meandered for whatever undefined period of time, whereas Danny Wang took an engineering approach to execute a directed line of experiments with the publication in mind. To have them as co-advisors meant that sometimes I had to negotiate between the two of them and beg for a 1 month allowance to wrap up one particular experimental line that Jon wanted to explore and then commit to toggling to another line of experiments for which Danny had conviction. I think this gave me life skills in diplomacy and project management and prepared me to work for any boss on the planet.

Post-MIT experience

I got my dream job by accident. When I was finishing up my thesis, I had the opportunity to attend a FASEB conference on Copper Mountain. I was a lowly graduate student presenting a poster, and all the keynote speakers were eminent professors in the field of protein folding and aggregation. Jon was scheduled as a featured speaker, but he got into a car accident on the way to Logan airport. He was OK but his doctor advised him against flying due to a mild concussion, so I was on point to give his keynote talk. This was in the day of 35mm slides, which Cindy shipped to me overnight to present the next day.

You never know who is in your audience. David Brems approached me after the presentation and expressed interest in bringing me to Amgen Thousand Oaks to interview for a scientist position. I had already had an offer in hand from Merck (MIT of the South) and had no intention of moving to Southern California – the land of earthquakes, mudslides, fires, and riots. Nevertheless I took the leap outside of my comfort zone to join Amgen.

My first assignment was to solve a precipitation issue with a therapeutic protein, leptin. This was not a formulation issue, as the protein was soluble at protein concentrations of 5-20 mg/mL in the pH 4 range, but the protein had limited solubility at physiological pH and precipitated upon subcutaneous injection. Since we could not reformulate the body, we reengineered the molecule for improved solubility. Our roadmap for designing solubility analogs was the observation that murine leptin was 20X more soluble than human leptin at neutral pH. Therefore, we set out to improve neutral pH solubility of human leptin by introducing various combinations of mutations based on the murine leptin sequence, which differed from the human leptin sequence at 22 sites.

The results were fascinating. Certain murine-like mutations were found to be critical in modulating solubility. Not all murine-like mutations improved solubility, and in fact, some mutations resulted in decreased solubility. Eliminating the residues that decreased solubility and optimizing combinations of mutations that included the key substitutions at 2 tryptophan sites resulted in solubility that exceeded that of murine leptin by 3-fold. This was not simply a pI-precipitation or net hydrophobicity issue, but rather irreversible self-association involving specific residues. This project was a perfect opportunity to apply the learnings from the King lab to identify the specific sites for self-association.


Twenty-five years later, I’m still at Amgen transforming molecules into medicines. As Vice President of Drug Product Technologies, I’m accountable for our end-to-end drug product portfolio from molecule assessment through commercial lifecycle management for biologics and synthetics modalities. Along the way, I had the opportunity to hold numerous roles within R&D and the Operations functions, starting with formulation and analytical responsibilities and then led several functions accountable for the early and late stage portfolio across Attribute Sciences and Drug Product organizations. In each role, I felt like I was majoring in a different subject matter every month, whether it be antibody structural variants, glass chemistry, silicone oil interactions, surfactant degradation, microbial in-use stability, etc. In a way, my start as an engineer transplant in the King lab trained me for tackling each of these challenges with scientific curiosity.

There have been many touchpoints in my career where I experienced echoes of the King lab. I was discussing the specificity of aggregation with a job candidate in an interview, and all of sudden the candidate stopped talking and exclaimed, “Wait! You’re ‘Speed’ of ‘Speed, Wang, and King’!!!???!” Then when I was showing our rental house to a potential tenant, Agi Hamburger, she mentioned that she used my old lab notebooks as a guide for her work in the King lab. And whenever any of my staff generate a confounding lab result, I go back to the time when I can hear Jon’s voice telling us (over coffee) that we really need to control for residual detergent on the glassware and trace metals in the water. And you know, he is right!


  1. Betts S, Speed M, King J. 1999. Detection of Early Aggregation Intermediates by Native Gel Electrophoresis and Native Western Blotting. In: The Process and Products of Protein Misassembly, Methods in Enzymology. 309:333-349.
  2. Ricci MS, Pallitto MM, Narhi LO, Boone T, Brems DN. 2006. Mutational approach to improve physical stability of protein therapeutics susceptible to aggregation: Role of altered conformation in irreversible precipitation. Book Chapter. In: Misbehaving Proteins: Protein (Mis)Folding, Aggregation, and Stability. Murphy RM, Tsai AM, Eds. New York. Springer. pp. 331-350.
  3. Ricci MS and Brems DN. 2004. Common Structural Stability Properties of 4-Helical Bundle Cytokines: Possible Physiological and Pharmaceutical Consequences. 2004. Review article. In: Current Pharmaceutical Design. 10:3901-3911.
  4. Speed MA, Wang DIC, King J. 1995. Multimeric Intermediates in the Pathway to the Aggregated Inclusion Body State for P22 Tailspike Polypeptide Chains. Protein Science. 4:900-908.
  5. Speed MA, Wang DIC, King J. 1996. Specific Aggregation of Partially Folded Polypeptide Chains: The Molecular Basis of Inclusion Body Composition. Nature Biotech. 14:1283-1287.
  6. Speed MA, Wang DIC, King J. 1997. Conformation of P22 Tailspike Folding and Aggregation Intermediates Probed by Monoclonal Antibodies. Protein Science. 6:99-108.
  7. Speed MA, King J, Wang DIC. 1997. Polymerization Mechanism of Aggregation. Biotech. & BioEng. 54:333-343.