New potency assays for anti-VEGF antibodies

Written by Dr. Janet Thode Posted in Method validation

Introduction and Background

The group of vascular endothelial growth factors (VEGF) plays a crucial role in the formation of new blood vessels (angiogenesis), since they stimulate the growth, i.e. the division and migration of endothelial cells, as the name already suggests. Since some tumor types can better supply themselves with nutrients via increased angiogenesis and thus grow faster, anti-VEGF antibodies represent significant cancer therapeutics. However, the various types of cancer are not the only indication; enhanced angiogenesis is also important in certain eye diseases.

Anti-VEGF antibodies or fragments include bevacizumab (Avastin®), ranibizumab (Lucentis®) or aflibercept (Eylea®) and an increasing number of biosimilars already on the market and in late-stage clinical trials, such as Mvasi®, Zirabev® or BevaciRel®.

For potency analysis of anti-VEGF antibodies during release and stability testing, different potency assays, so-called bioassays, can be used. The classical potency assay for anti-VEGF antibodies is the so-called HUVEC proliferation assay [1]. In this bioassay, primary HUVECs (human umbilical vein endothelial cells) are seeded into a 96-well plate considering a specific cell number after appropriate pre-culture and mixed with the anti-VEGF antibody in different concentrations, which was previously pre-incubated with VEGF. After an incubation period of 3 - 4 days, a cell viability reagent (e.g. Alarmar blue® or CCK-8®) is added and further incubated for a certain time. Subsequently, fluorescence or absorbance is measured using a microplate reader to determine the number of live cells and generate a dose-response curve. The bioassay is based on the principle that endothelial cells naturally want to divide by the addition of VEGF but are prevented from doing so to varying degrees by the concentration-dependent binding of the anti-VEGF antibody to VEGF. Therefore, I would rather speak of a cell proliferation inhibition assay ????

If you take a closer look at this bioassay, you will notice that it has some disadvantages. The pure execution time of the bioassay is not quite short, and the use of primary cells provides an increased variability in the results of different bioassays due to different donors and number of passages. Moreover, the culture of primary cells can also be a bit challenging... Considering that the cells have to be continuously cultured for a certain time prior to their use in the bioassay and that this takes a tremendous amount of time, it is understandable that recent trends are e.g. the use of cryopreserved, ready-to-use cells. In addition to conventional ELISAs, a 2018 application note about the use of Perkin Elmer’s AlphaLISA® technology written in collaboration with the Swiss contract laboratory IBR Inc. also shows a possibility for an alternative, cell-free and thus much faster potency assay [2].

Two other alternative potency assays, a reporter gene assay [3] and an enzyme-fragment complementation (EFC) assay, are presented below with a focus on their method validations.

 

Test principles

A new cell line (NFAT-RE-Luc2P/KDR HEK293) was generated for the reporter gene assay. For this purpose, HEK293 cells were stably transfected with the genes for VEGF receptor 2 (VEGFR-2) and the VEGF165 response element nuclear factor of activated T cells (NFAT-RE) as well as the reporter gene for luciferase. Thus, the HEK293 cells carry the VEGFR-2 on their surface. In case of VEGF binding, the two receptor molecules dimerize and a subsequent signaling cascade leads to the binding of transcription factor NFAT to its response element NFAT-RE in the nucleus, and thus initiating the expression of luciferase. If an anti-VEGF antibody comes into play at different concentrations, this means that correspondingly less luciferase is produced and thus the light signal catalyzed by the luciferase is weaker, since the anti-VEGF antibody "traps away" the available VEGF in a concentration-dependent manner, or, to put it more elegantly, "neutralizes" it. This assay including the cell line is commercially available from Promega.

The enzyme fragment complementation assay (PathHunter® Bevacizumab Bioassay by Eurofins DiscoverX) is also based on a stably transfected HEK293 cell line. Here, "only" the genes for the two monomers of VEGFR-2 were introduced into the HEK293 cells. The special feature, however, is that the two monomers were fused inside the cell with one fragment each of β-galactosidase (EA and ED). The two enzyme fragments EA and ED are inactive when separated. In case of VEGF binding, dimerization of the receptor monomers occurs resulting into close spatial proximity, and thus generating the active enzyme, which in turn hydrolyzes an added substrate, thereby generating chemiluminescence. In case of anti-VEGF antibodies present, dimerization is prevented, and the signal remains absent.

 

Comparison of the two bioassays, focus: method validation

It should be said at the beginning that for the enzyme fragment complementation assay, the information presented was compiled from various sources [4-6] and the information basis was not always entirely clear and complete (more on this also in the discussion). Moreover, since I noticed some errors, I recalculated all the data shown by the sources, so that in this blog post, in addition to the flawless information, the values corrected according to my calculations are shown.

The table below compares the results of the individual parameters required to validate a potency assay, as well as other useful information of the two bioassays.

 
   Reporter gene assaya  Enzyme fragment complementation assay
 2016 [4]  2017 [5]  2018 [6]
 Cells  Cryopreserved, ready-to-use cellsb
 Time for assay execution  < 8 h  < 24 h  
 „Kind“ of validation and experimental set up

 „preliminary validation“ (acc. to ICH Q2R1)

 1 linearity experiment with 4 replicates (number of analysts unknown, probably 1)

 Qualification (acc. to ICH Q2R1)

 1 linearity experiment with 1 measurement each at 3 days (number of analysts unknown, probably 1)

 Qualification

 1 linearity experiment with 1 measurement each at 4 days (by 1 analyst)

 Validation acc. to USP <1033>

 1 linearity experiment, 2 analysts, 6 independent runsc

  Linearityd  5 Concentrations (50, 75, 100, 125, 150%); R2 = 0.9968  4 Concentrations (50, 75, 125, 150%), R2 = 0.9879e  6 Concentrations (50, 71, 90, 100, 121, 150%), R2 = 0.9935  6 Concentrations (50, 71, 90, 100, 121, 150%), R2 = 0.9899f
  Trueness

 Use of the linearity data; recoveries between 85 - 120% (only graphically displayed)

  Use of the linearity data; mean recovery: 95.9%  Use of the linearity data; mean recovery: 98.0%  Use of the linearity data; mean recovery: 99.9%g
 Repeatability  5 analyses at the same day; RSD = 7.0% (basis of data not shown)  RSD = 7.3% (basis of data not shown)  Not evaluated or shown  Use of the linearity data; RSD from 6.4% to 14.2%, mean RSD 10.1%h
  Intermediate Precision  1 analysis at 5 days each; RSD = 7.6% (basis of data not shown)  Use of the linearity data; RSD from 2.7% to 5.0%, mean RSD 3.5%i  Use of the linearity data; RSD from 2.7% to 8.5%, mean RSD 5.5%j  Use of the linearity data; RSD from 6.6% to 14.2%, mean RSD 10.2%h
  Specificity  No inhibition in case of denaturated BVZ or anti-CD25, -EGFR, -TNFα or -IL-6 antibodies  Not evaluated or shown
  Robustness  After generation of the cell line, use of the cells frozen at different passages (3, 15, 33): comparable curves  Use of 3 different cell lots: comparable VEGF curves with an RSD of 11% related to the EC50 of the curves of the 3 cell lots  Not evaluated or shown

 a This table shows the results of the publication, further validation data - not shown here - can be found in the Promega’s user manual [7].
 b In the reporter gene assay, ready-to-use cells are part of Promega's kit; in Wang's publication, the cell line was established.
 c unclear wording of the source, see also discussion
 d This is not the plot of the signal (RLU) against the (logarithmized) BVZ concentration (i.e. linearity in its true sense), but the plot of the measured relative potency against the expected / theoretical relative potency. It therefore provides information about the trueness.
 e Own calculation based on the 4 concentrations (the R2 in the published figure considers a 100 = 100% value that is not shown)
 f Individual data not shown, own calculation using the data given for trueness
 g Own calculation, but the individual data shown for the percentage error are conform
 h No presentation of the underlying individual values. In the table shown for repeatability and intermediate precision, in two cases (100% and 150%), the values are exactly the same (including all 4 decimal places) for both types of precision, which suggests typing errors, therefore overall data are questionable
 i Data of own calculations, due to two calculation errors in the published individual RSDs
 j Data of own calculations since all published RSDs were different

 BVZ = Bevacizumab       EC50 = Half maximal effective concentration       RSD = relative standard deviation 

 

Discussion

Prior to addressing technical points, it should be mentioned that there were errors in each of the 3 sources [4-6] of the enzyme fragment complementation assay. These ranged from incorrect page number details of the references listed, to incorrectly calculated values (see above), to incorrect labeling in the reaction principle (in 2017 it is correctly shown, in 2018 PK would have to be replaced by ED), to (possibly) incorrect labeling. For example, in 2018, the accuracy data speak of the "geometric mean", but the calculations of 2016 and 2017 showed the arithmetic one... and for the comparison of the PathHunter® assay against the classical HUVEC assay, the axis of the PathHunter® assay seems to be mislabeled, as probably not the VEGF concentration, but rather the BVZ concentration is shown, which can be deduced from the shape of the curve compared to the figures of the other publications...

Besides that, some wording was a bit unclear... Does the information on the 2018 execution ("2 analysts, 2 plates, 2 replicates/plate; 6 Independent runs per analyst to achieve n=8 per target potency") mean that each analyst has 1 plate with 1 replicate each? And how do you get n = 8 with 6 runs? Should it be n = 6 or did each analyst really use 2 plates corresponding to n = 12? And when did these 6 independent runs take place (2 staggered during one day or at 6 different days)?

But in order not to chalk up too much "non-technical" information, we should stop here and take a closer look at the method validations and the two potency assays themselves.

First, let's have a look at the linearity data. Apart from the 2016 assay, 5 or more concentrations are examined in each case, as required by ICH Q2(R1). But do these data really tell us anything about the linearity of the method? They each show a super correlation between measured and theoretically expected value when different concentrations are applied, but do not tell us how the signal changes as a function of concentration. However, we should keep in mind that in such assays no single values are measured, but the samples and the standard are applied in a variety of dilutions, resulting in 4-parameter logistic fit (4PL) curves, and the curves of the samples are subsequently compared with that of the standard for similarity. Thus, we must move away from the simple picture of a straight line where a halved concentration results in a signal half as strong, as this is not applicable here. If we use different concentrations of the standard to evaluate linearity or trueness and dilute them further, this results in laterally shifted curves. In this respect, a good correlation between measured and expected value at different concentrations tells us something about the trueness of the method, but indirectly also something about linearity. If the obtained results agree well with the theoretical ones despite applied dilutions, the dilutions do not seem to be a problem. Or put the other way round: good results can be obtained in the range of applied dilutions.

And speaking of working range, yes, at the first glance we might think that all 4 validation experiments shown cover the same range since they all span 50 - 150%. However, since we don't have any information about the standard, we don't know if the same standard was used for all experiments and if not, how its concentration and activity might differ. Comparability of assays may become easier in the future by using an international standard developed by the WHO [8].

However, both assays seem to be superior to the classical HUVEC assay in terms of working range. Wang speaks about a dose-response curve for the reporter gene assay in the range of 4.6 - 370 ng/mL (about 80x window) and of 100 - 500 ng/mL (about 5x window) for the HUVEC assay. However, in the graphs shown, the window varies only between about 4x - 10x (visual estimation by me) [3]. In the 2018 presentation of the enzyme fragment complementation assay [6], this assay is said to have a 3x window and the HUVEC assay a 1.5x to max. 3x.

As already evident by the super correlations of the linearity data, the evaluation of trueness via recovery in the enzyme fragment complementation assay also shows excellent results with averaged recoveries of 96 - 100%. For the reporter gene assay, the variation in recovery was somewhat greater with values between 85 - 120%, although it should be noted that this is a bit of an apples to oranges comparison. The recoveries mentioned for the enzyme fragment complementation assay are averaged, the variation for reporter gene assay reflects the broadness of the individual results.

I am reluctant to go on at length about repeatability because either the data basis is not shown, or it has not been studied or it is not clear how many analyses were really done by one analyst.

When looking at the intermediate precision of the enzyme fragment complementation assay over the 3 years, it is noticeable that it increases as the data set is enlarged (from 4 to 6 concentrations) and a second analyst is included, as expected. This can be nicely seen here. The same can be observed for the reporter gene assay considering the data of a further method validation [7].

Specificity has been adequately investigated for the reporter gene assay, but no data are available for the enzyme fragment complementation assays based on the sources consulted, which is a pity since investigation of specificity is required for potency assays according to ICH Q2(R1) as well as USP <1033>.

Using different passages or batches of the cell line to evaluate the robustness of the assays is reasonable and reflects subsequent conditions of use. However, for the enzyme fragment complementation assay, it would have been nicer to include bevacizumab curves instead of VEGF curves for evaluation. If we consider - apart from these experiments - that both assays have been repeatedly qualified / validated in different years each and have consistently yielded good results, both assays seem to be very robust.

Validation of a bioassay does not necessarily have to be performed according to ICH Q2(R1). An alternative is the USP chapter <1033> "Biological Assay Validation", as applied here to the enzyme fragment complementation assay in 2018. Therefore, relative trueness as bias, specificity, intermediate precision, and working range must be examined, and a more extended statistical evaluation may be applied. The decision according to which regulatory requirements the method validation should be performed is left to the user.

Furthermore, it should be mentioned, that both the reporter gene assay and the enzyme fragment complementation assay have been evaluated with other anti-VEGF therapeutics, such as aflibercept and ranibizumab [5, 7].

And if I would be asked now which of the two potency assays, I would prefer to establish in my laboratory (if I would have one), I first like to say that both potency assays are on solid footing from my point of view. Taking into account the time needed to execute the respective assay and the fact that in the lab where I have recently had the privilege to support transfers and validations of bioassays, other reporter gene assays with the same luciferase substrate options are already established and familiar to the analysts, I would opt for the reporter gene assay (but this is just my very personal, subjective opinion).

 

References

[1] Wang Y., Fei D., Vanderlaan M., Song A. (2004) Biological activity of bevacizumab, a humanized anti-VEGF antibody in vitro. Angiogenesis. 7(4):335-345

[2] Morra L., Moser R. (2018) Alpha Technology: A Fast and Sensitive Orthogonal Approach to Cell-based Potency Assays, www.perkinelmer.com/de/libraries/APP_AlphaLISA_Bevacizumab, accessed 01/27/2023

[3] Wang L. et al. (2016) Development of a robust reporter-based assay for the bioactivity determination of anti-VEGF therapeutic antibodies. J. Pharm. Biomed. Anal. 125:212-218

[4] Lamerdin J., Daino-Laizure H., Saharia A., Charter N.W. (2016) Accelerating Biologic and Biosimilar Drug Development: Ready-to-Use, Cell-Based Assays for Potency and Lot-Release Testing. BioProcess Inter. 14:36-44.

[5] Lamerdin J., Caldwell P., Daino-Laizure H., Paranjpe G., Saharia A. (2017) Ready-to-Use Potency Assays for Bevacizumab, Aflibercept, & Ranibizumab, www.discoverx.com/tools-resources/document-resource-library/documents/bebpa-usa-2017-ready-to-use-potency-assays-for-bevacizumab,-aflibercept,-ranibizumab, accessed 01/27/2023

[6] Urquhart P., Lamerdin J. (2018) Accelerating Biologics and Biosimilars Development:Case Studies on Implementing Robust and Simple Potency Bioassays, www.discoverx.com/CMSPages/GetAmazonFile.aspx?path=~%5Cdiscoverx%5Cfiles%5C80%5C80188177-deda-4aee-9035-e05a77fa3e9b.pdf&hash=8314548a675bebfaf40d2ce953615441cff230a9f85882ad1cbbac255adb25a8, accessed 01/27/2023

[7] Promega Technical Manual VEGF Bioassay (TM544, revised 12/22)

[8] Jia H., Harikumar P., Atkinson E., Rigsby P., Wadhwa M. (2021) The First WHO International Standard for Harmonizing the Biological Activity of Bevacizumab. Biomolecules. 11(11):1610

 

Disclaimer

At the end of this hopefully interesting blog article a perhaps not completely unimportant last information: Mentioning the brand names in this article was necessary for the comprehensibility and the brand names are protected by their owners. At no time have we received any financial support by the companies mentioned in this post or by the companies of the brands mentioned. This blog article reflects solely my opinion.