We are not aware of limits of HCP contamination established by the FDA or any other international regulatory bodies. You should contact your regulatory body for advice and current points of view. The lack of data conclusively showing HCPs to cause safety or efficacy problems together with the fact that absolute quantitation of HCPs is rarely possible, make setting of limits a somewhat arbitrary task. We do not presume to speak for current regulatory positions on HCP detection and limits, but we can relate our considerable experience with products expressed in a variety of biological expression systems. In addition, we are pleased to offer what we feel is a rational approach to the questions of HCP contamination.
Each product submission should and will likely be considered on its own merits regarding the question of HCP contamination. Issues such as the route of administration, quantity of drug given, and frequency of dosage are all factors that could influence safety problems from HCPs. The approach should be to get HCP levels as low as reasonably possible. With advances in antibody generation and immunoassay development such as those implemented in our kits, the sensitivity and specificity of HCP analysis is significantly improved over Western blot and PAGE analysis. When the purification process is developed with feedback from a sensitive HCP ELISA, the levels of HCPs are typically less than 10 ppm. When a sensitive ELISA is not used during process development, and instead methods like PAGE and Western Blot are employed, it is common to see HCP levels greater than 100 ppm and often in the ppt range. We believe the most rational and cost effective strategy is to employ a sensitive generic ELISA methodology early in product development rather than waiting until Phase 3. Such an assay will provide very valuable feedback to purification process developers and may help insure a more positive outcome to early clinical trials.
Many of our customers have expressed an interest in automating our ELISAs to improve throughput and assay performance. Cygnus Technologies has much experience with various automation platforms. These systems include automated microtiter plate instruments as well as several of the so-called “biosensor” platforms. Using our proven and well-qualified antibodies and reagents in a biosensor platform still presents your lab with significant challenges in re-qualifying and optimizing the reagents. Before considering adapting our reagents to an automated system, you are encouraged to contact our Technical Service Department for guidance on how to approach assay development, qualification, and validation.
The promises of automation are higher throughput, faster turnaround of results, and improved performance in terms of parameters like sensitivity and precision. The practical ability of the many systems now being marketed to meet these promises is variable. Before embarking on a quest for automation we urge you to contact our Technical Services Department. After evaluating your testing needs we can guide you to the system that best meets your requirements.
The absolute quantitation of HCP assays is exceedingly difficult for a number of reasons. First of all there is no recognized reference preparation of HCP against which we can calibrate our assays. While ELISA is inherently a quantitative method when applied to a single analyte, ELISAs that attempt to measure simultaneously all of the hundreds of potential HCP contaminants in the same well using a single reporter/detector system are at best semi-quantitative assays. Many arbitrary choices and assumptions are used when preparing anti-HCP antibodies and later the choice of material to use when preparing “standards”.
♦How to obtain the HCP material for immunogen is the first choice that must be made. Do we use a null cell line or a mock transfected cell line or one with the actual product plasmid and what differences in HCP might we see from each?
♦At what point in the purification process do we obtain the HCPs for the immunogen or standards? In almost every case the array and relative concentrations of HCPs in final product are different from the HCP in the ELISA standards.
♦How does one assign a total HCP concentration to an indeterminant mixture of many proteins with different molecular weights?
♦What effect will the different affinities of antibodies to various HCPs as well as the different concentrations of those antibodies have on the ability of the assay to be quantitative?
♦How do we know that we have antibody to all the HCPs present in a given sample type?
All of these questions and arbitrary choices mean HCP assays will at best be semi-quantitative methods capable of making only relative estimates of HCP concentrations from one sample to the next. We estimate the potential error in reporting an HCP level in a sample that has a different array of HCPs from the standards to be as high as 4 fold even if you have a very good broadly reactive antibody. The commonly expressed theoretical concern is that HCP assays might under-estimate results due to incomplete coverage of all HCPs. In reality, the quantitative errors in HCP assays due to the arbitrary choices and fundamental limitations of the method as discussed above can result in either over-estimation or under-estimation and tend to do so to about the same degree. In our experience, a well generated and affinity purified antibody will react to more than 70% of individual HCPs as demonstrated by traditional 2D Western blot correlated to silver stain. What is more important than the number of individual proteins with antibody reactivity, is the total mass of those proteins that are detected. In our experience and as you might expect from theoretical considerations, the proteins in highest concentration have the best chance of generating an antibody. As such, an antibody with about 70% reactivity to an individual HCP will react with those proteins that account for more than 95% of the total mass of HCP. Given this level of uncertainty in the absolute HCP concentration, the most important criteria by which we can judge any HCP assay is on other objective parameters by which any analytical method should be qualified or validated. Those criteria include specificity and accuracy as demonstrated by sample dilutional linearity and spike recovery experiments, precision, robustness, and sensitivity. Provided the assay has sensitivity to detect at least a relative portion of the HCP in your final drug substance and the assay also meets the other objective analytical parameter specifications, such an assay is a valuable analytical method capable of demonstrating process control and reporting relative levels of HCP contamination from lot to lot as a release test.
Due to the impracticality of obtaining real final product HCPs that co-purify with product down to the final step, most of our HCP assays will utilize a source of HCPs from very upstream in the purification process. Those HCPs typically do not come from the actual product cell line but rather a null cell or mock transfected cell line. Once we have decided on the source material for the kit standards we first perform some partial purifications to remove non-HCP materials such as growth media additives, buffer salts and extraction reagents. Our initial approach to calibrate this processed material is to simply perform a BCA total protein assay using bovine serum albumin (BSA) as an arbitrary standard. When standards made with the HCP material calibrated by BCA give reasonable “stoichiometric agreement” with the amounts of antibody used in the ELISA then we feel the calibration by BCA is good enough. What is meant by “stoichiometric agreement” is that we know how much antibody is used in the ELISA. It is the quantity of antibody that actually dictates the analytical range and dose response curve of the ELISA. If we assume an average molecular weight for the total HCPs then we can reasonably estimate HCP concentrations across the valid analytical range of the assay. When the BCA assay concentrations do not reasonably approximate the ELISA stoichiometry, we process the HCP material further. Such processing involves various purification steps to remove components registering in the BCA assay but that are in fact not HCP or at least not immunoreactive HCPs. This processing might also involve affinity purification against the anti-HCP antibody. The purification stops as soon as the BCA concentration gives a realistic stoichiometric agreement in the ELISA.
Please review our technical paper in the Publication Section for our website for a very detailed discussion on the definitions, limitations, misunderstandings and myths of Process Specific versus Multi-Use assays.
Most so called Process Specific Assays are not purification process specific but are rather ambiguously termed Process Specific because they use a very upstream, null cell line preparation of antigen to generate the antibody. This arbitrary selection of an upstream source of null cell HCP such as conditioned media or lysate does not guarantee that the antibody or assay development from this antibody will detect the different and more limited array of HCPs that persist through a purification process. The vast majority of HCPs are conserved among all the strains of a given expression system. For example, of the hundreds to few thousand HCPs in the CHO proteome, there are only a very small number that differ antigenically from strain to strain or growth process to growth process. There are four key factors in making a broadly reactive antibody that can have a much greater effect on the relevant activity of the antibody to downstream HCP than the very minor antigenic differences that might be seen from strain to strain:
1. Selection of an antigen that is truly representative of HCPs found at that step in the purification process.
2. The processing of the antigen to make a potent immunogen.
3. Methods for immunizing animals that insure broad reactivity and high antibody titer.
4. Subsequent affinity purification of that antibody.
If one fails to understand how to make a good HCP antibody, then simply calling it Process Specific, based on the use of upstream, null cell antigen material does not insure it is a good antibody. The proof of the validity of an antibody is found in the validation of the assay for detection of HCP in downstream and final drug substance. All assays, whether they are arbitrarily termed Process Specific, Platform, Generic, Commercial or Multi-Use must be validated the same way using real downstream and final drug substance samples while following conventional, objective parameters for validation of analytical methods.
While we would like to detect all HCPs this is a difficult task not only to achieve, but to conclusively demonstrate given the limitations of the methods. One and two dimensional Western blot have traditionally been used as orthogonal methods in an attempt to characterize antibody reactivity to individual upstream HCPs. We find Western blot too insensitive and non-specific to be of any practical value in demonstrating antibody to individual downstream HCPs. While Western blot can detect many upstream HCPs it has very little predictive value for ELISA performance. The absence of a western blot band or spot to an individual HCP does not necessarily mean the ELISA will fail to detect that HCP. For those who desire to demonstrate antibody reactivity to individual downstream HCPs we have developed a method far superior in both sensitivity and specificity to Western blot. This method is termed 2D HPLC-ELISA. It can detect individual HCPs in most final drug substances. In the final analysis it is those HCPs that are the most important. See our technical paper for a detailed discussion of the 2D HPLC-ELISA method.
There have been some erroneous reports in the literature of cross-reactivity of anti-HCP antibodies to recombinant drug substances. These mistaken conclusions are usually based on the misuse and misinterpretation of Western blot data. Loading a large amount of product protein in the range of micrograms, in an effort to see trace HCP contaminants in downstream samples, will often result in a "ghost" WB band to the product protein. This is rarely true immunological cross-reactivity but rather a non-specific protein-to-protein interaction due to overloading of the product. WB is not a good method to assess immunological cross-reactivity, as it is often non-specific and rarely has the sensitivity to detect downstream HCPs. Furthermore, WB is of very little value in predicting how cross-reactivity would manifest itself in the more sensitive ELISA. Without the appropriate negative controls, WB should not be used to conclude immunological cross-reactivity. If you want to use WB to determine cross-reactivity, contact our technical services department for advice on how to design your WB experiments with the proper controls. Also, review our technical article on Western Blot vs ELISA: Sensitivity and Specificity Differences.
Given that most HCP antibodies are made to null cell strains, cross-reactivity to the product would not be expected unless there is a significant conservation of epitopes on the product with those in the natural proteome of the host organism. Even if there are some cross-reacting epitopes on the product that manifest as a specific band by Western blot, it should be understood that in most cases this cross-reactivity will not result in a false increase in HCP concentration by ELISA or similar sandwich assay method used to quantititate HCP in final drug substance. To understand why this is the case, one must have a comprehensive appreciation for the stoichiometric limitations of ELISA. Any cross-reacting antibody is a small subset of the total anti-HCP activity. Downstream samples contain mostly the product itself, often in the mg/mL range. This high concentration of product will greatly exceed the amount of cross-reacting antibody in the ELISA, usually by up to a million fold. At such an excess of antigen, the dose response curve for drug substance will be well past the "high dose hook" in the negatively sloped region, very near to the background signal for the assay. In this region, any ELISA reactivity to your drug substance would be observed either to not change with dilution of drug substance sample, or to actually increase with decreasing drug substance concentration. The dilutional linearity experiment we recommend as critical to validation of any HCP method can indicate either an excess of certain HCPs relative to the amount of kit antibody against them or less likely, cross-reactivity to your drug substance. If you have a lack of dilutional linearity, a stoichiometric analysis of yor dilutional data will usually allow for the discrimination of true drug substance cross-reactivity from the more common lack of antibody excess for certain HCPs that persist through your purification process.
We have developed over 100 anti-HCP antibodies to virtually all the recombinant and transgenic expression systems in use today. The only cases where we have seen true immunological cross-reactivity that interfered in the accurate detection of HCP by ELISA has been with Pichia pastoris and two transgenic plant expression systems. In all 3 cases, it was determined that the cross-reacting epitopes were due to glycosylation of the drug substances. We have developed various methods to overcome cross-reactivity. If you have confirmed cross-reactivity that interferes with accurate measurement of HCP please contact our Technical Service Department for advice on how to proceed.
Cygnus offers 3 ELISA kits to measure CHO HCPs. Kit #F015 , the first kit developed more than 12 years ago utilized a mild lysate of washed CHO cells for the immunogen as well as the for affinity purification of the antibody and for kit standards. This mild lysis procedure generated an array of HCPs similar to those found in culture media after cells have been grown to densities and viabilities similar to product harvest.
Kit #CM015 was developed about 2 years later with the rationale that since most CHO expressed products are released into the growth media (conditioned media) an antibody and/or assay made to those HCPs found in conditioned media should be more specific than a cell lysate assay.
We now offer a 3rd generation CHO kit, #F550 that uses an even more reactive antibody as well as the latest improvements in the detection of HCPs. This kit, like the #CM015 kit, used conditioned media to generate the antibody. However, we have shown that the #F550 media contained an array of HCPs more typically found in most CHO expressed products. As such we believe the #F550 kit represents more accurate results for most samples.
Testing of both the lysate and conditioned media antibodies by western blot showed that the vast majority of proteins found in cell lysate could also be found in conditioned media and that all three antibodies have very similar reactivity. While the antisera are qualitatively similar, the assays themselves may not give the same values for some sample types. There are many reasons for this but perhaps the most significant is that the relative abundance of certain HCPs will be different in lysate versus conditioned media. Since different HCP preparations are used to make the standards, generate the antibody and affinity purify the antibody, then it can be expected that various samples types which contain yet another array of purification process specific HCPs, will be differentially detected between the three assays. Any assay which attempts to simultaneously detect multiple antigens that can vary in their relative proportions from the calibrators used, will be at best a semi-quantitative estimate of HCP content. Such assays may underestimate or overestimate the true HCP content and thus it is important to appreciate the uncertainty in any value reported for a sample.
Cygnus cannot predict which assay will be best suited for a given product or sample type. Because the 3rd generation kit, #F550, utilizes our latest advances in antibody generation and assay development, we recommend that all new customers try this kit first. It is up to the user to validate whatever method they select to determine that it provides adequate sensitivity, accuracy, and specificity. That validation study should first determine the specificity of the assay. Nothing in the samples such as additives like detergent, buffer salts, pH or the presence of product protein itself should cross react with the antibody or otherwise cause an increase in non-specific binding that could result in an overestimation of HCP. With the specificity established, the user must then determine that nothing in the samples inhibits the detection of true HCP. This is determined by performing spike and recovery experiments in the various samples types to be tested as well as dilutional recovery/linearity experiments in samples containing elevated levels of HCPs such as those typically found in upstream purification process samples. With the results of these validation studies in hand it is usually possible to select one assay over the other. Assuming the assays are equally specific and accurate we normally recommend that assay which yields the best sensitivity for HCP detection in your specific downstream or final product samples. Cygnus appreciates the challenges in detection of HCPs and is pleased to offer close customer support in validating and troubleshooting product and sample specific problems. Please contact our highly experienced Technical Services Department if you have any questions.
As a recognized leader in bioprocess contaminant analysis, Cygnus Technologies is pleased to offer its expertise to develop custom antibodies and assays. If you have a unique analytical need or if our generic kits cannot be satisfactorily validated for your product and samples types, contact our Technical Service Department for a draft project plan and quotation to develop reagents and methods specific to your needs.
We do not recommend the use of linear regression to fit ELISA data particularly HCP assays. We specifically warn against the use of linear regression methods because most HCP immunoassays are not linear nor is linearity a requirement for a good immunoassay. To force data from an assay to fit the best straight line when the inherent dose response is not a straight line is a certain way to introduce inaccuracy into your results. Similarly, other more sophisticated regression methods can also introduce mathematical assumptions and enforce arbitrary and inappropriate rules that actually reduce the inherent accuracy and precision of the assay. Inaccuracies due to a poor curve fit method are most significant at the extremes of the standard curve, most often in the low end but sometimes in the high end as well. For this reason we strongly urge our clients to use Point-to-Point, Cubic Spline or 4 Parameter as the curve fitting routines since these will yield the most accurate results.
Curve fit routines other than the three we recommend may sometimes work satisfactorily for a given immunoassay method. If you decide to use another method for interpolation of sample values it will be important for you to perform a careful analytical evaluation of your proposed algorithm. If after considering the points below, your method meets or exceeds the interpolation accuracy and precision of the other 3 methods across multiple kit lots then by all means use it. We caution that just because your method works well in one immunoassay or one lot there is no guarantee it will be optimal for another and therefore each assay should be validated on a case by case basis. We recommend the 3 methods above because they are the most robust and the most accurate for immunoassay and have worked for our assays. An easy way to determine the optimal curve fit routine is by "backfitting" the signals of your standards as unknowns. If the standards, when backfit as unknowns do not give back their nominal values you may have artifacts introduced by inappropriate assumptions or restrictions in your curve fit algorithm. Finally, the most direct and objective way to assess the accuracy of an immunoassay is to assay controls with known levels of analyte, across the important analytical range of the assay. The method that yields average values closest to the nominal levels of those controls together with the best "run to run" precision is the method you should chose. Do not rely on arbitrary and indirect parameters like R square, slope, y intercept, or asymptotes as QC specifications. These parameters are often too insensitive to be useful in flagging a bad assay run.
All sample types with apparent levels of analyte greater than the LOQ of the assay should initially be evaluated for dilutional linearity as part of assay validation. This experiment involves performing a series of doubling dilutions within the analytical range of the assay using an approved assay diluent. These dilutions are then assayed and a dilution corrected contaminant concentration is determined at each dilution. This dilutional linearity study establishes freedom of sample matrix interference and also demonstrates the important condition of antibody excess for the array of contaminants in your samples. Please refer to the secton on "Hook Effect". If you will be routinely testing in-process samples in addition to final product, you should validate dilutional linearity of each sample type. This analysis is critical for HCP assays because very high concentrations of certain HCPs may approach saturation of the antibody against that particular HCP. When this happens there is a risk of significant under-quantitation for that HCP. By performing dilutional analysis one can verify if the antibody is in excess and that the sample matrix itself does not interfere. If the antibody is in a limiting concentration or the sample matrix causes a negative interference what will be observed is that the apparent HCP concentration for a sample increases with increasing dilution.
In most cases a dilution will be reached where the dilution corrected value remains essentially constant. We term this the Minimum Required Dilution or MRD. The table below shows example data where a sample did not yield good dilutional linearity at high concentration, but with further dilution an MRD was determined at which acceptable dilutional linearity was obtained. In this example we conclude that the MRD for this in-process sample is 1:8 and that the concentration of HCP to be reported is 361ng/mL. Once an MRD is established for a particular sample type, your SOP should reflect that these samples need to be diluted before assaying. We suggest defining acceptable dilutional linearity as dilution corrected analyte concentrations that vary no more than 80% to 120% between doubling dilutions. Due to the statistical limitations in the low end of the assay range you should avoid consideration of dilutional data where the assay value before dilution correction falls below two times the LOQ of the assay. Acceptable diluents may vary from assay to assay and you are encouraged to verify with Cygnus Technologies that your sample diluent is acceptable.
In general, the best diluent is the same one used to prepare the kit standards. Assay specific diluents can be purchased from Cygnus in 100ml, 500ml or 1000mL bottles. Contact Cygnus for information on acceptable diluents. Should you determine that there is significant product or matrix interference and simple dilution is not an option, it may be necessary to further process the sample by methods such as buffer exchange to render it into a more assay compatible buffer. In other cases, modification of the assay protocol can affect improved accuracy in some sample types. Users of our kits are encouraged to contact the Technical Services Department for advice on how best to solve sample accuracy issues. Example of Dilutional Linearity Data is shown below:
|Dilution CorrectedValue (ng/mL)||% change in concentration from previous dilution|
|1:64||ND < LOQ of assay||NA|
The protocols in our ELISA kits recommend the use of dual wavelength analysis for microtiter plates readers appropriately equipped. For alkaline phosphatase – PNPP based assays the two wavelengths are 405nm and 492nm. For horse radish peroxidase – TMB based assays the wavelengths are 450nm and 630nm. In both assay types the first wavelength listed, also termed the "test wavelength", is critical and thus your microtiter plate reader should be equipped with the appropriate 405 and 450 nm filters. It is not absolutely necessary to use dual wavelength analysis with any of our assays and thus the second wavelength (492nm and 630nm - also called the "reference wavelengths") are optional. In theory, the use of dual wavelength analysis should provide better precision. Microtiter plate readers with the appropriate software will automatically subtract the reference wavelength absorbance from the test wavelength absorbance. In theory dual wavelength analysis can help to overcome any non-wavelength specific imperfections in the plate. In practice, because the optical quality of our microtiter plates is excellent, there is usually no significant statistical improvement in the data by subtracting out the reference absorbance from the test absorbance. Most plate readers are equipped to perform dual wavelength analysis. Since the reference wavelength is not critical, microtiter plate readers with reference filters close to those we recommend can be substituted. For example you may use a 490nm or a 500nm filter in place of the 492nm or a 600nm or 650nm filter in place of the 630 nm filter.
For any ELISA to give accurate results there must be an excess of antibodies, both capture and enzyme conjugated, relative to the analyte being detected. It is only under the conditions of antibody excess that the dose response curve is positively sloped and the assay provides accurate quantitation. As the concentration of analyte begins to exceed the amount of antibody the dose response curve will flatten (plateau) and with further increase may paradoxically become negatively sloped in a phenomenon termed "High Dose Hook Effect". Because the possibility exists that some samples may have analyte concentrations in excess of the antibody it is necessary to validate all sample types by dilutional linearity analysis to establish if they are on the valid, positively sloped region of the curve or on the negatively sloped hook region of the curve. Failure to validate the potential for Hook Effect can result in severe under-estimation of true contaminant concentrations! The issue of hook effect in multiple antigen assays such as HCP ELISA can be more complex. The dose response curve for an HCP assay should be thought of as the cumulative dose responses of all HCPs individually with each HCP having its own hook region determined by the concentration of antibody to that particular HCP. We are practically and fundamentally limited in the amount of antibody that can be used in an HCP ELISA. It is not uncommon in HCP assays for some samples to have certain HCPs in concentrations exceeding the amount of antibody for that particular HCP. In such cases the absorbance of the undiluted sample may be lower than the highest standard in the kit however these samples will fail to show acceptable dilutional recovery/linearity as evidenced by an apparent and significant increase in HCP concentration with increasing dilution. This lack of dilutional linearity is actually the result of the hook effect for the subset of analytes in excess over their respective antibodies. Poor dilutional linearity (Hook Effect) is most likely to be encountered in samples early in the purification process. However if the purification process is selective for certain HCPs, it may also be seen in downstream and final product samples. Thus the establishment of dilutional linearity is a most critical experiment in the development and validation of HCP assays. Dilutional linearity studies are performed at a series of dilutions to establish what we term the "minimum required dilution" (MRD) for a given sample type. The MRD is the first dilution at which the dilution adjusted value for the sample in question and all subsequent dilutions remains essentially constant. The HCP value to be reported for such samples is the dilution corrected value at or greater than the established MRD. Once an MRD is established for a particular sample type, your SOP should reflect that this sample requires this dilution prior to assay.
Table 1, below, shows data where a sample did not yield good dilutional linearity at high concentration but with further dilution an MRD was determined at which acceptable dilutional linearity was obtained. In this example, we conclude that the MRD for this in-process sample is 1:8 and that the concentration of HCP to be reported is 361ng/mL. Once an MRD is established for a particular sample type, your SOP should reflect that this sample needs to be diluted before assay. We suggest defining acceptable dilutional linearity as "dilution corrected analyte concentrations that vary no more than 80% to 120% between doubling dilutions". Due to the statistical limitations in the low end of the assay range you should avoid consideration of dilutional data where the assay value before dilution correction, falls below two times the LOQ of the assay. Acceptable diluents may vary from assay to assay and you are encouraged to verify with Cygnus that your sample diluent is acceptable. In general, the best diluent is the same one used to prepare the kit standards. Assay specific diluents can be purchased from Cygnus in 100ml, 500ml or 1000mL bottles.
|Sample Dilution||Dilution CorrectedValue (ng/mL)||% change in concentrationfrom previous dilution|
|1:64||Not calculated (<2 times LOQ)||NA|
Cygnus offers two ELISA kits to measure Insulin. Both kits use the same antibodies so immunological specificity is comparable. The major difference in the kits is sensitivity and the assay protocol to achieve that sensitivity. Insulin kit #F040 has a limit of quantitation of approximately 250pg/mL, while kit #F280 has a limit of quantitation of less than 25pg/mL. Both kits have been validated to detect insulin as a trace contaminate in products produced by cell culture where insulin is a growth media additive. The sensitivity of Catalog #F280 kit should be sufficient to allow for detection of fasting insulin levels in bovine, porcine or human samples. These kits have not been FDA approved or fully validated for use in animal or human blood samples. Both kits are labeled as “For Research and Manufacturing Use Only, Not Approved for Diagnostic use in Humans or Animals".
Assays attempting to measure trace contaminants such as HCPs often in the presence of more than a million fold excess of the product, can be prone to analytical interference by the product. Similarly, samples from various points in the purification process may also contain components in their matrices that will interfere in ELISA methods. The interference can manifest itself as either a false increase or false decrease in true analyte levels. Factors such as extremes in pH, detergents, organic solvents, high product protein concentration, and high buffer salt concentrations are known interference components. When the levels of analyte are well above the Limit of Quantitation (LOQ) of the assay, dilution of the sample is often the easiest way overcome interference. In other cases, buffer exchange of the sample into an assay compatible buffer will solve the problem. The assay protocol can also be manipulated to overcome some types of sample interference. Users of our kits are encouraged to contact our Technical Services Department for advice on how best to solve sample interference problems.
The critical experiments to evaluate product and sample matrix interference are Dilutional Linearity and Spike & Recovery. A detailed description of these experiments can be found at "Poor Dilutional Linearity" and "Poor Spike & Recovery".
The assay protocol recommended in the kits' product insert have been developed with consideration given to typical limits of detection required in bioprocess contaminant analysis. The assays are very robust with respect to protocol options and thus it is possible to modify the protocol to achieve performance parameters more optimal for your analytical needs. Sample volume, incubation times, and use of various sequential (forward or reverse sequential) schemes can lead to significant changes in sensitivity, reduction of non-specific sample matrix affects, or greater upper analytical range. If you change the protocol, it will be necessary to validate that those changes achieve acceptable accuracy, specificity, and precision. Please refer to the section below on "Validation of ELISA." For advice on how best to modify an assay to meet your objectives you are encouraged to contact our Technical Services Department.
Cygnus Technologies now offers 4 ELISA kits for detection of Protein A. Our original kits, Catalog #F050 and #F050H have proven to be valid for use in detecting both natural and structurally conserved recombinant Protein A. The newest kits Catalog #F400 and #F400Z can better detect unnatural recombinant constructs of Protein A such as GE Healthcare’s MabSelect SuReTM ligand. Prior to the availability of this unique construct, all Protein A used for antibody purification was either natural Protein A derived from Staphylococcus aureus or recombinant Protein A (E.coli expressed) as patented and sold by Repligen. Repligen recombinant Protein A is very similar to natural Protein A with over 99% homology. Both Protein A preparations contain essentially the same 5 subunit binding domains (A, B, C, D, & E) and both have molecular weights of ~ 44.7kDa.The new GE Healthcare MabSelect SuReTM construct is a structurally unique molecule. MabSelect SuReTM is derived only from the B binding domain of natural Protein A. It is a multimerized, tetramer of the B domain with a C-terminal cysteine introduced for site-directed immobilization. The molecular weight of MabSelect SuReTM is 26.7kDa.
The Cygnus Protein A Kits, Catalog #F050 and #F050H use antibodies made to natural Protein A with validation performed against only the natural and the highly conserved recombinant Protein A. With the introduction of MabSelect SuReTM we have evaluated its reactivity in these kits. Our data has shown that it has only about a 20% weight to weight reactivity relative to the Protein A standards in these kits. This 20% reactivity may be greater or less depending upon other factors such as the product antibody being purified, which may contribute to further interference or non-specificity.
Cat. #F400 and #F400Z assays address the low reactivity to MabSelect SuReTM seen in our other kits, while incorporating several other improvements in the methodology. The biotinylated antibody used in our original kits has been replaced with a different antibody directly labeled with HRP. Improvements include more sensitivity, simplified assay protocol, more standards, and less probability of product matrix interference. The #F400 and #F400Z kits detect natural, conserved recombinant, and MabSelect SuReTM Protein A ligands essentially 1:1. Similar to the F050H assay, these assays use a sample denaturation step to overcome sample/product antibody interferences. Because so many laboratories have successfully validated and are routinely using our earlier kits #F050 & #F050H, we will continue to offer these kits indefinitely. First time evaluators of Protein A detection methods are encouraged to evaluate the #F400 or #F400Z kits regardless of which type of Protein A they use.
The #F400Z kit usese the same antibodies, standards and sample treament procedure as the #F400 kit. The only difference is that #F400Z has a "recovery enhancer" reagent that improves analytical recovery in antibody:fusion proteins and antibody fragments that are not completely removed during the sample treatment step.
A new high throughput sample treatment procedure has been developed to reduce tedious pipetting while increasing throughput of the sample processing step. This new method has demonstrated superior spike recovery in problematic sample matrices and also overcomes 'front to back' drift reported by some labs. Labs equipped with microplate adaptable centrifuges can now process up to 96 tests simultaneously. Instead of using individual microfuge tubes, the new method uses a 96 well "PCR" Sample Treatment Plate.
The basic procedure is outlined below:
► Pipet your samples into the wells of the Sample Treatment Plate. Additional sample dilutions can be made directly in the plate.
► Add Sample Denaturing Buffer and seal the plate with the foil provided.
► Heat the Sample Treatment Plate at 80oC in a 96 well heat block for 15 minutes.
► Centrifuge the Sample Treatment Plate to pellet the denatured antibody.
► Samples can now be quickly transferred to the antibody coated microtiter plate using a multichannel pipet.
This procedure is not only less tedious but can minimize problems when handling microfuge tubes such as inadvertent disruption of the denatured protein pellet.
No changes have been made to any of the kit reagents' or the assay protocol. Both the #F400 and #F400Z Protein A ELISA kits contain a Sample Treatment Plate and an adhesive foil for those choosing to use this more efficient procedure. The product inserts have been amended to include both sample treatment procedures. If you are using Protein A kit #F050H the Sample Treatment Plate, Cat #F402 can be purchased separately.
The routine run-to-run quality control of ELISA is best accomplished by assaying control samples across the important analytical range of the assay. We recommend 3 controls, a low control in the range of 1 to 2 times the assay LOQ, a medium control, and a high control. These controls should be made using your source of analyte (e.g. HCPs from your cell line or growth media). Furthermore, the controls should ideally be in the same matrix as your critical samples. By using your source of analyte in your sample matrices, you will be best able to identify any problems within the run or between runs and kit lots. We manufacture our kits for lot-to-lot consistency and try to avoid changes in any components or procedures that could impact accuracy in the customers' laboratory. However, as generic kits, we can only quality control them by a limited range of parameters that may or may not be sensitive to your product specific issues.
Use of laboratory specific controls is the only way to assure total quality control of the assay for your needs. Controls should be made in bulk, aliquoted for single use, and frozen at -80°C until stability studies indicate some other storage conditions are adequate. Once you have statistically established a range for these samples, they will become your most sensitive and specific tool to assure quality control of the assay. Do not rely on curve fit parameters as quality control specifications in the absence of true analyte controls. Curve fit parameters such as R square, slope, y-intercept, and upper and lower asymptotes are not sensitive or specific enough to reliably detect assay problems. Use of such parameters in the absence of true analyte controls, will frequently fail a perfectly good run and worse cause you to pass a run that would have been flagged had analyte controls been used. Contact our Technical Services Department for advice on how to make and establish controls specific to your needs.
Numbers of Replicates - When precision is very good (average replicate %CV on ODs are less than 5%) we feel duplicate analysis is adequate and the most cost effective approach. In the case of duplicate analysis we do not allow for editing of any apparent outliers since there is no statistical basis for establishing which of the duplicates is inappropriate. Thus, in duplicate analysis we suggest repeating any sample that yields a %CV greater than 20%. Alternatively, performance of the assay in triplicate or even quadruplicate may allow for editing of data points, such that it is unnecessary to perform a repeat assay. Criteria for deleting certain data points are somewhat subjective but should take into consideration the impact of your error limits on product safety or allowable levels of contaminant.
Some samples, particularly those from upstream in your purification process will have contaminant analyte concentrations above the analytical range of our very sensitive ELISA kits. Such samples may require very large dilutions in order to overcome "Hook Effect" and to achieve acceptable "sample dilutional linearity". In addition to the "Hook Effect," the matrix of some samples may interfere non-specifically with the assay and also result in under recovery of the true analyte levels. Simple dilution of those samples is often adequate to buffer out such interference provided the dilution does not reduce analyte concentrations below the limit of quantitation of the assay. In cases where dilution of your samples is not an option contact our Technical Service Department for advice on how best to overcome sample matrix interference.
Cygnus offers assay "specific diluents" for each of its kits. The catalog numbers for these diluents can be found in the kit product insert or by contacting Customer Service. We strongly recommend use of those diluents because they are the same formulation as the matrix used for the kit standards. Thus, as you "dilute your samples" in our diluent, your sample matrix begins to approach that of the standards and in this way greatly minimizes any dilutional artifacts that could occur if you where to use another diluent.
DO NOT use PBS or TBS without a carrier protein. These reagents can be problematic because the analyte diluted in the range of the assay (ng/mL) can very significantly adsorb to the dilution tube resulting in low recovery!
In some cases your product itself or certain components in the product formulation buffer may interfere (either positive or negative interference) in the ability of the assay to detect HCPs or other contaminants. Similarly, samples from upstream in the purification process may also contain material in their matrices that can interfere in ELISA methods. Factors such as extremes in pH, detergents, organic solvents, high protein concentration, and high buffer salt concentrations are known interfering components. For these reasons it is necessary to validate by universally recognized experimental procedures (i.e. ICH & FDA guidelines) that the assay will yield accurate results. The two critical experiments in assessing assay accuracy and specificity are 1) Spike & Recovery and 2) Sample "Dilutional Linearity." Should the end user of this kit determine that there is significant product or matrix interference, it may be necessary to further process the sample by methods such as dilution or buffer exchange to render it into a more assay compatible buffer. The same diluent used to prepare the kit standards is ideally the preferred material for dilution or buffer exchange of your samples. In other cases, modification of the assay protocol can improve accuracy in some sample types. For each sample type to be tested, be it final product or in-process samples, you should demonstrate that the assay can recover added HCP or other contaminants spiked into that sample matrix. This can be performed by spiking the highest standard provided with the kit into your sample types and then testing in the assay. Using the E. coli HCP kit, Cat # F410 as an example, we suggest spiking 1 part of the 100ng/mL standard into 4 parts of your sample (e.g. spike 100μL of 100ng/mL standard into 400μL of sample) as shown in Table 1. The spiked concentration into the sample in this case is 20ng/mL. A control dilution of 1 part of assay diluent (zero standard) to 4 parts of sample is also performed, to determine the contribution of endogenous HCP in the sample prior to spiking. Both the spiked and diluted, unspiked sample are assayed. Percent added recovery is determined by subtracting the endogenous contribution of HCP, from the total HCP measured in the spiked sample. We suggest acceptable recovery should be within 80% to 120% of the spiked HCP. The table below shows example data. If you desire spike and recovery at more than one concentration we recommend that the lowest spike levels should be at least 2 times the Limit of Quantitation (LOQ) of the assay and that the contribution of the endogenous HCP in the sample prior to spiking not exceed two times the spike level to be tested. These two conditions will insure better statistical accuracy.
|ng/mL||Spike Conc. ng/mL||Total HCP Measured||% Spike Recovery|
|4 parts final product + 1 part “zero standard”||0||6||NA|
|4 parts final product + 1 part 100ng/mL standard||20||25||95% [(25-6)/20]|
Cygnus kits and reagents have been formulated to tolerate shipment under ambient temperature conditions for several weeks without any significant deterioration. Orders within North America are shipped next day delivery service without cold packs. Upon receipt of the products, we recommend storage at 2-8ºC. Our reagents are validated for prolonged storage at elevated temperatures and thus assure that problems such as summertime temperatures or delays in shipping or refrigeration problems at the customer location will not cause any significant deterioration. Orders shipped outside of the United States are packed in an insulated box with cold packs or wet ice to avoid concerns for extended transport time and customs delays.
In the final analysis it is results obtained on your control samples and historical specifications that determine the integrity of the kit. If the curve absorbances, sensitivity, and controls are within range, then the kit is acceptable for use.
This study protocol is suggested as an objective method to validate that the Cygnus Technologies Host Cell Protein (HCP) ELISA kits and other bioprocess contaminate kits will yield accurate, specific and reproducible results for a given product and sample type. In addition to the ELISA validation protocol detailed below, it may also prove useful to determine the reactivity of the ELISA antibody to individual HCPs which can be resolved by protein fractionation methods such as PAGE and HPLC. Despite many acknowledged limitations Western blot, both 1 and 2 Dimensional have historically been used to characterize reactivity of polyclonal antibodies to individual HCPs. Unfortunately the analysis of anti-HCP antibodies by Western blot is of very limited predictive value due to the lack sensitivity and specificity in detecting HCP in final product and other downstream samples. It is for this reason, ELISA is proven to be the method of choice for determination of total HCP in downstream samples. To better answer the question of antibody reactivity to individual downstream HCPs we recommend a method termed 2D HPLC-ELISA. The procedure involves a chromatofocusing fractionation of proteins according to their isoelectric points in the first dimension followed by an HPLC Reverse Phase gradient fractionation in the second dimension. This method yields highly purified individual, liquid phase HCPs in a more native, un-denatured configuration. The resulting fractions can be automatically collected in microtiter wells and easily analyzed in the much more sensitive, specific and semi-quantitative ELISA with greater objectivity. For a more detailed discussion of the limitations of Western blot and the advantages of 2D HPLC fractionation please refer to the technical paper found on our web site.
In some cases your product itself or certain components in product formulation buffer may interfere (either positive or negative interference) in the ability of the assay to detect HCPs or other contaminants. Similarly, samples from upstream in the purification process may also contain material in their matrices that can interfere in ELISA methods. Factors such as extremes in pH, detergents, organic solvents, high protein concentration, and high buffer salt concentrations are known interference components. For these reasons, it is necessary to validate by universally recognized experimental procedures (i.e. ICH & FDA guidelines) that the assay will yield accurate results. Should the end user of this kit determine that there is significant product or matrix interference it may be necessary to further process the sample by methods such as sample dilution or buffer exchange to render it into a more assay compatible buffer. The same diluent used to prepare the kit standards is ideally the preferred material for dilution or buffer exchange of your samples. Please refer to the kit product insert for recommended diluents. In other cases, modification of the assay protocol can effectively improve accuracy in some sample types. Users of our kits are encouraged to contact our Technical Services Department for advice on how best to solve sample accuracy issues.
1. Dilutional linearity/parallelism experiments – All sample types to be tested that contain levels of contaminant greater than the LOQ of the assay, should initially be evaluated for dilutional linearity as part of assay validation. This experiment involves performing a number of serial dilutions using an approved assay sample diluent. These dilutions are then assayed and a dilution corrected contaminant concentration is determined at each dilution. This dilutional linearity study establishes freedom of sample matrix interference as well as the important condition of antibody excess in the ELISA for the array of contaminants in your samples. If you will be routinely testing in-process samples in addition to final product, you should validate dilutional linearity of each sample type. This analysis is particularly critical for HCP assays because very high concentrations of certain individual HCPs may approach saturation of the antibody against that particular HCP. When this happens there is a risk of under-quantitation for that HCP. By performing dilutional analysis one can verify if the antibody is in excess and that the sample matrix itself does not interfere. If the antibody is in a limiting concentration or the sample matrix causes a negative interference what will be observed is that the apparent HCP concentration for a sample increases with increasing dilution. In most cases a dilution will be reached where the dilution corrected value remains essentially constant. This dilution is what we term the Minimum Required Dilution or MRD. Table 1 below shows example data where an in-process sample did not yield good dilutional linearity at high concentration, but with further dilution, an MRD was determined at which acceptable dilutional linearity was obtained. In this example we conclude that the MRD for this in-process sample is 1:8 and that the concentration of HCP to be reported is 361ng/mL. Once an MRD is established for a particular sample type, your SOP should reflect that such sample needs to be diluted before assay. We suggest defining acceptable dilutional linearity as “dilution corrected analyte concentrations that vary no more than 80% to 120% between doubling dilutions”. Due to the statistical limitations in the low end of the assay range you should avoid consideration of dilutional data where the assay value before dilution correction falls below two times the LOQ of the assay. Acceptable diluents may vary from assay to assay and you are encouraged to verify with Cygnus that your sample diluent is acceptable. In general, the best diluent is the same one used to prepare the kit standards. See your kit product insert for recommended diluents. Assay specific diluents can be purchased from Cygnus in 100ml, 500ml or 1000mL bottles. Contact Cygnus for information on acceptable diluents.
Table 1: Example Dilutional Linearity Data for an In-Process sample
|Sample Dilution||Dilution CorrectedValue (ng/mL)||% change in concentration from previous dilution|
|1:64||Not calculated (<2 times LOQ)||NA|
2. Spike and recovery experiments - For each sample type to be tested, be it final product or in-process samples, you should demonstrate that the assay can recover added HCP or other contaminant spiked into that sample matrix. This can be simply performed by spiking some of the highest standard provided with the kit into your sample types and then testing in the assay. Spiking studies are only relevant at or below the MRD as established in Section 1 above. Using our E. coli HCP kit Cat # F410 as an example, we suggest spiking 1 part of the 100ng/mL standard into 4 parts of your sample (e.g. spike 100mL of 100ng/mL standard into 400mL of sample). The spiked concentration into the sample in this case is 20ng/mL. A control dilution of 1 part of assay diluent (zero standard) to 4 parts of sample is also performed to determine the contribution of endogenous HCP in the sample prior to spiking. Both the spiked and diluted-unspiked sample are assayed. Percent total recovery is determined by dividing the total HCP measured (from endogenous and spike) by the sum of the spike level + the endogenous contribution of HCP. We suggest acceptable recovery should be within 80% to 120% of the spiked HCP. Table 2 shows example data. If you desire spike and recovery at more than one concentration we recommend that the lowest spike levels should be at least 2 times the Limit of Quantitation (LOQ) of the assay and that the contribution of the endogenous HCP in the sample prior to spiking not exceed two times the spike level to be tested. These two conditions will insure better statistical accuracy.
Table 2: Example Spike and Recovery Data
|Sample||Spike Conc.(ng/mL)||Total HCP measured(ng/mL)||% Spike Recovery|
|4 parts final product + 1 part zero standard||0||22||NA|
|4 parts final product+ 1 part 250ng/mL std.||50||70||97.2% [70/(50+22)]|
3. Precision Experiments - We also recommend that each laboratory and perhaps each technician perform a precision study to demonstrate that they can achieve acceptable precision both within and between runs. Such precision would be best accomplished on controls prepared in your laboratory. Those controls should be made in your product matrix and from HCP derived from your cell line and process. Once prepared at the desired levels these controls can be aliquoted and stored frozen to insure stability. Once statistical values have been established on your controls you will then have an important QC tool to assure that the accuracy is acceptable from kit to kit and assay run to assay run. Beyond the above critical validation studies you may want to perform other experiments that statistically establish the LOQ and LOD in your laboratories, robustness, stability, etc. as listed in FDA and ICH guidelines. However, you may want to consider using our in-house generic validation study for this less critical data rather than perform it yourself since we view such non-critical data as inherent in the method and not something each laboratory needs to do.
Western blot is very rarely acceptable for detection of HCPs in your drug substance or drug product samples. Samples downstream in your purification process typically contain HCPs below the sensitivity of Western blot. For Western blot, you are limited in the amount of total protein you can load and still get good PAGE resolution. When you load final product or samples from downstream in the purification process the vast majority of protein will be the product itself. For example, the maximal load of protein for a PAGE run on a mini gel is on the order of 10µg/lane. If HCP contamination is 100ppm, a level typical of many final drug products, then the amount of total HCP in that 10µg of drug would be 1ng. With the sensitivity of western blot on the order of 1ng/band it could, in theory, detect HCP contamination down to 100ppm if the 100ppm were a single HCP and not a mixture of several different HCPs. As it turns out, there are usually several HCPs that contaminate final product and for this reason Western blot is almost always negative for HCP on downstream and final product samples. ELISA demonstrates less interference from drug product and shows sensitivity more than 100 fold lower than Western blot. As such, ELISA will typically allow for the detection of total HCP contamination to less than 1ppm. There are many other fundamental reasons why the sensitivity of Western blot is inferior to ELISA. For example, Western blot often requires that the PAGE step be carried out under reducing conditions (DTT or BME followed by boiling) and in the presence of high concentrations of SDS detergent. These procedural components may actually denature or block some of the native HCP epitopes that would be detectable in an ELISA. Incomplete transfer of the proteins out of the PAGE and onto the membrane and adsorption on the membrane at or near antigenic sites will also limit the amount of binding seen by Western blot.
As you try to increase the sensitivity of Western blot it is very common that the specificity of the method is also compromised. What is typically seen is that a non-immunoreactive protein present in very high concentration (e.g. your drug substance) will invariably adsorb some of the excess anti-HCP antibody non-specifically leading to the erroneous conclusion that the anti-HCP antibody seems to "cross-react" with your product. The way to confirm this non-specific binding to your product is to use a non-immune immunoglobulin of the same species and at the same concentration as the anti-HCP antibody. If the intensity of the drug substance band is the same with both the normal goat IgG and the anti-HCP antibody, you can conclude the band is non-specific. Specificity dan also be confirmed by loading much smaller quantities of your drug substance in the range of 1-4ng/lane. If there is true cross-reacting antibody that might manifest as false HCP levels, you should still see a strong product band. Beyond these experiments it should be understood that the specificity of the ELISA method is typically orders of magnitude better than Western blot owing in large part to the fact that any protein must be bound simultaneously by both the capture antibody and the detection antibody. For this reason most artifactual product bands in the Western will not yield apparent HCP activity in the ELISA method.
While Western blot is of little value for detecting HCP in all but very upstream samples, it has continually been used for demonstrating that the antibodies in the ELISA kit react with the majority of the HCPs from your cell line. This is typically a one time experiment where you lyse some cells or use conditioned media from your cell line in a Western blot and compare the blot to a protein stain from PAGE such as colloidal gold or silver stain. If the homology is adequate one can conclude that the antibody has adequate reactivity in the ELISA kit, although it is only through validation of the ELISA that one can conclude that the antibody is adequate.