Using Amino Acid Analysis to Determine Absorptivity Constraints: A Validation Case Study Using Bovine Serum Albumin

Amino acid analysis of well-recovered residues offers an easy way to calculate the absorptivity constant for a known protein. The method provides an absolute measure of protein concentration, free from interference from water, excipients, and bound salts. This article demonstrates a qualified method for determining protein content by AAA.

By John C. Anders, Benne F. Parten, Glenn E. Petrie,
Robert L. Marlowe, and John E. McEntire

With protein biologics, it is critical to accurately determine an absorptivity constant. The constant is then used when routinely measuring protein concentration by ultraviolet (UV) spectrophotometry. The yield of product during manufacture and the product content and potency on the label are dependent on this constant. There have been several approaches in the past to determining the constant, but all have drawbacks — primarily poor accuracy, or interference from excipient salts, or overly tedious methods. Our study sought to demonstrate a method for determining specific absorptivity constants (as) by amino acid analysis (AAA) that provides accurate and precise quantitation of protein concentrations. (The full name of amino acids abbreviated in this article can be found in the "Amino Acid Code" sidebar.)

BASICS AND HISTORY
Although AAA techniques used today have been around since the late 1950s, remarkable progress has made those methods more sensitive, automated, and available for use in any biochemical laboratory. In AAA, protein and peptide bonds are hydrolyzed in hot hydrochloric acid leaving free amino acid residues. Amino acids are then analyzed by quantitative ion-exchange chromatography (1).

Determination of as for a small organic - molecule is often a minor undertaking. The purified and dried molecule is quantitated by gravimetric means, followed by dissolution in an appropriate buffer or solvent and simple UV–vis spectrophotometric analysis at its _ maximum. However, for protein-based biopharmaceuticals this is not as easily accomplished. The limiting step is quantifying the protein, which may contain contaminating proteins, bound salts, metals, or excessive amounts of bound water. Quantification of purified proteins to determine the as in a defined solution has been accomplished in the past by gravimetric, Kjeldahl, and colorimetric techniques (2,3) and by amino acid analysis (4). As an alternative, statistically derived approaches for predicting as of proteins are also widely used. These predictive models are based on the UV absorptivities of tryptophan, tyrosine, and cystine residues at 280 nm. With the exception of the predictive spectrophotometric methods and AAA, the above methods are not commonly used in biopharmaceutical applications today because of the large quantities of protein needed or the large degrees of error associated with their use.

History of as for proteins and peptides. Determining as for synthetic or recombinant proteins or peptides is a necessary component of biopharmaceutical production and formulation. Once an as value is determined for a particular protein in a well-defined formulation matrix, the concentration of that protein in solution can be rapidly determined from its measured absorbance and the specific absorptivity constant using a derivative of Beer's law (see Equation 1).

The as is expressed in mL mg_1 cm_1 units and can be converted to the molar absorptivity (am, formerly known as _) using the relationship am _ as _ MW, where MW is the molecular weight of the protein. Commonly used estimates of as are based on predictive ultraviolet spectrophotometric models. Gill and von Hippel developed such a predictive model for determining the am (or _) of denatured proteins in 6.0 M guanidine hydrochloride (Gdn) (5), based on the earlier work of Edelhoch (6) (see Equation 2).

Using the model in Equation 2, Gill and von Hippel calculated the am of 18 globular proteins and compared the results to measured values found in the published literature (5). Reasonably good agreement was found between predicted am values for proteins in Gdn and those values listed for Gdn in its native aqueous solutions. The average difference between predicted and actual values was 4.9%, with a maximum difference of _14.9%.

Pace et al. further refined the Gill and von Hippel equation by testing the predictive capability of a modified equation on 80 different proteins in aqueous solutions (7). From that body of data, they refined the average am values for Trp, Tyr, and Cys by statistical means (Equation 3). Examination of their data revealed that for 47 out of 80 proteins, the predicted am agreed with the literature values within 5%. The 33 other proteins were found to agree between 5% and 17%.

This complete article is available from our Download Library in PDF format. Click here to download.