Letter to the Editor

The following letter came in response to recent articles on lot-release testing and host-cell proteins:

Members of our peer-review board recommended publishing this longer than average communication.

A Rational Approach to Process-Specific Host-Cell Protein Detection

So-called generic host-cell protein (HCP) antibodies and assay kits have proven to be valuable for guiding purification process development. With comprehensive analytical validation, many of these assays demonstrate relevant specificities and sensitivities for use as lot-release tests. But concerns remain: Could they miss certain HCPs? Rather than require a comprehensive qualification/validation to demonstrate adequate reactivity to individual HCPs, European regulators suggest that applicants develop “process-specific assays” under the arguable assumption that the antibodies would be superior regardless how they are generated, purified, and incorporated into an assay.

Concerns about reactivity to individual HCPs are valid, but specific guidance is lacking on how to demonstrate adequate reactivity. Review articles have been published, and industry stakeholders have convened in ad hoc groups to propose guidance, but such efforts fall short of offering scientifically effective ways to address theoretical concerns. I’d like to suggest an approach to qualification of all assays, generic or process specific.

My review of various publications suggests much ambiguity in related terminology. So I first propose the following definitions:

A generic assay is intended to detect all HCPs that might contaminate a product. These assays are generic in that they detect HCPs independent of downstream purification processes. Generic assays use HCPs obtained from far upstream (e.g., conditioned media) or after some minimally selective purification step (e.g., clarification or filtration). Commercial assays are a type of generic assay useful when most HCPs are conserved among related strains and processes. As with any generic assay, they must be qualified by end users following sound analytical criteria.

Cell-line–specific assays use HCP from null or mock-transfected cells of the same cell line used to express a product protein. These assays are still generic in that they are based on upstream HCPs. As with commercial assays, it is assumed that HCP from cells expressing the product are substantially the same as those in a null cell line. Some authors have used the term multiproduct assay as if this represents a superior approach to HCP assay development. But multiproduct assays are nothing more than generic assays used for products expressed in a particular strain. Cell-line–specific assays must be characterized like other generic assays.

A growth-process–specific assay uses HCPs derived from a given cell line under a defined growth process. Using upstream HCPs, these assays are also generic. Because different growth processes can produce significant differences in the array of total HCPs; such assays could in theory offer somewhat better detection than other types of generic assays. But their utility must be confirmed by comprehensive characterization.

Purification-process–specific assays are derived from an immunogen specific to a purification process. These assays are difficult to develop for a number of reasons. Critical issues include the choice of purification step (how far downstream) and obtaining adequate quantities of antigen truly specific to that process. A common approach is to use null product purification runs, leading to the arguable assumption that HCPs recovered in the absence of product are the same as those found when product is present. In addition to difficulties in obtaining relevant HCPs, purification-process–specific assays can be problematic by virtue of their being too specific. I have seen cases in which they fail to detect atypical HCPs that occur due to a change or deviation in a purification process. For this reason, my company recommends that users of downstream-process–specific assays also apply to detection of atypical HCPs.

Assay Development: It is important to acknowledge the limitations of HCP detection methods such as enzyme-linked immunosorbent assay (ELISA) or equivalent platforms. Generation of antibodies and development of assays involves arbitrary choices. The first choice is the source of antigen: from null or mock cells (under what growth conditions), further processed or not, and from which purification step. Equally important are how the immunogen is prepared, immunization methods, preparation of the affinity support, stringency of affinity purification, and how to develop/optimize the ELISA.

For several reasons, HCP assays are seldom truly quantitative. The most fundamental limitation is that HCPs used for immunogen and assay standards are usually different in both array and relative concentrations from those downstream and in a final drug substance. After the arbitrary choice of antigen, a developer must determine how to quantitate the HCPs. A common approach is to assign HCP concentration using “total protein” assays such as the bicinchoninic acid assay (BCA) or a Coomassie-blue assay using bovine serum albumin (BSA) as the “standard.” Doing so can provide reasonable estimates of total HCP under ideal circumstances, in which interfering and non-HCP components have been removed or subtracted from true HCPs. But in many cases, a crude source of upstream HCPs such as clarified conditioned media “calibrated” by means of a total protein assay overestimates HCP.

When a poorly quantitated HCP preparation is assayed in another generic assay with more accurately quantitated standards, that preparation can yield values substantially lower than indicated by the protein assay. Paradoxically, this can lead to an erroneous conclusion that the generic assay is underestimating HCP. Many reports of generic assays that fail to detect HCP from another cell line or growth process have been based on inaccurate quantitation of crude upstream HCP preparations by BCA or Coomassie. Reliance on Western blotting, with its limited sensitivity and specificity, also can give false conclusions that an antibody is inadequately reactive. Such incomplete experiments will “beg the question” for a more process-specific assay.

Assay Performance: It matters less how assays detect upstream HCPs than how well they detect downstream HCPs. Ultimately the reason we perform HCP assays is to look for HCP in a final drug substance. Theoretical and specious arguments must give way to objective science using appropriately sensitive and specific methods to assess reactivity to those HCPs that persist through purification. Assumptions that any antibody generated to a particular strain or growth process will be superior to another generated to an almost identical strain is an unnecessary leap of faith. Generation of broadly reactive and relevant HCP antibody is rarely as simple as injecting a few rabbits with some upstream source of HCP in Freund’s complete adjuvant. People successful in this field have developed critical methods to improve antibody quality.

The proof of an antibody or assay should reside in qualification/validation data. Certainly different strains can produce a few different proteins. However, the vast majority of the proteome (representing many hundreds of proteins) will be antigenically conserved among all strains of a given species. Similarly, although cell culture conditions can up- or down-regulate relative quantities of certain HCPs, the majority found in media are qualitatively similar across all growth conditions. This observat
ion is substantiated by the fact that generic assays have been validated to work with many different products using different cell lines and growth processes.

Within the limits of HCP detection, HCP assays must qualitatively detect a representative portion of HCPs in a final drug substance. If an assay is sensitive to HCPs in final drug substance and can be objectively validated by conventional criteria (e.g., dilutional linearity, spike recovery, specificity, and precision), then that assay can be deemed adequate as a release test. A real analytical and safety need to determine reactivity to individual HCPs requires better analytical methods than are used traditionally. Although a Western blot test may yield a qualitative indication of reactivity to upstream HCPs, it is too orthogonal, nonspecific, and insensitive to be of predictive value for how an ELISA detects those HCPs in a final drug substance. So we find very little utility in one- and two-dimensional Western blot analyses. If questions of reactivity to individual HCPs are critical, then a user needs more sensitive and specific analytical methods.

My company has developed one such method: 2D HPLC-ELISA (www.eprogen.com/proteomics/documents/CygnusTechnologyPaper.pdf). This involves fractionation of either upstream or final drug substance by a first-dimension chromatofocusing step separating proteins into isoelectric point (pI) ranges. The pI-range fractions are then subjected to gradient reverse-phase HPLC to further resolve individual proteins. The two-dimensional fractionation results in 1,000–2,000 individual fractions, each of which is tested in an ELISA for reactivity. Unlike in polyacrylamide gel electrophoresis (PAGE) and Western blotting (which require denaturation by boiling with reducing agents and strong detergents followed by problematic transfer to a membrane), these liquid-phase fractions remain in more native configurations similar to a final drug substance. Using this method, we can detect individual HCPs in downstream and final drug substance long after Western blot results are negative for HCP. Positive fractions can then be subjected to mass spectrometry for identity.

Recommendations: Do not rely on PAGE or Western blot to drive your early phase purification process development. Both methods are too insensitive to optimize HCP clearance. When generic assays are available, attempt to qualify them for use in purification process development. If the associated antibody is inadequate, then have one custom made by someone experienced in the field.

Seek consensus from your regulatory agency and reviewer comparing what they require and what you propose for HCP analysis and validation. Agree on terminology. Defend your approach if it is superior to conventional methods.

The proof of any antibody and assay comes from characterization and qualification studies. So-called “cell-line specific,” “growth-process specific,” and “purification-process–specific” antibodies may not be superior to a well generated, highly purified, broadly reactive generic antibody. If you can objectively qualify an existing assay for use as a release test, then developing another antibody and assay will be redundant at best.

Characterize your chosen assay to individual HCPs using downstream or final drug substance samples. This requires methods analogous to the 2D HPLC-ELISA method discussed above. Do not rely on Western blot data of upstream HCPs as your method for determining antibody reactivity. How well an antibody detects HCPs in upstream, null-cell preparations is of little relevance to the question of whether it detects HCP in a final drug substance.

About the Author

Author Details
Ken Hoffman is president of Cygnus Technologies Inc., 4701 Southport Supply Road, SE, Suite 7, Southport, NC 28461; 1-910-454-9442, fax 1-910-454-9443; [email protected].