Skip to main content
Clinical and Translational Science logoLink to Clinical and Translational Science
. 2023 Aug 28;16(11):2123–2129. doi: 10.1111/cts.13616

Deriving protein binding‐corrected chemical concentrations for in vitro testing

Aditya R Kolli 1,
PMCID: PMC10651662  PMID: 37605430

Abstract

Extracellular chemical concentrations are considered physiologically relevant for in vitro testing and are evaluated in traditional in vitro systems using cell culture media containing 5%–10% fetal bovine serum (FBS). However, depending on the physicochemical properties, and in vitro testing conditions, cells could be exposed to variable unbound extracellular concentrations. If in vitro unbound concentrations are not calculated, it is challenging to distinguish the chemical potency and concentration‐driven responses. In this study, one‐ and two‐protein binding models were used to estimate protein binding corrected chemical concentrations of various chemicals for in vitro testing. As ceftizoxime, moxifloxacin, and nicotine have low protein binding affinity, the in vitro protein binding in 5%–10% FBS is less than 5% and can be considered negligible. However, protein binding of moderate and highly protein‐bound chemicals must be corrected for as the in vitro unbound concentrations in 5%–10% FBS containing cell culture media will vary over a range of chemical concentrations. In vitro pharmacological and toxicological assessments must incorporate protein binding‐adjusted in vitro concentrations to ensure physiologically relevant exposures. A user‐friendly Excel spreadsheet is provided to help bench scientists calculate protein binding‐corrected chemical concentrations for in vitro testing.


Study Highlights.

  • WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

In vitro evaluations are performed by mimicking unbound plasma concentrations, but even a small amount of serum alters the unbound chemical concentrations available to exert pharmacological and toxicological activity.

  • WHAT QUESTION DID THIS STUDY ADDRESS?

How to select or determine protein binding‐adjusted chemical concentrations for in vitro testing.

  • WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

Chemical‐protein binding measured in serum can be used to determine unbound chemical concentrations in 5%–10% serum‐containing media.

  • HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY AND OR TRANSLATIONAL SCIENCE?

Using physiologically relevant and protein binding‐adjusted chemical concentrations during in vitro assay enables improved translation of exposure‐responses. Protein binding correction also helps in understanding chemical potency and concentration‐driven exposure‐responses.

INTRODUCTION

In vitro pharmacological and toxicological evaluation of chemicals is performed using a wide variety of cell‐based assay systems. Different immortalized cell lines, primary cells, or a combination of multiple cell types are cultured in two‐dimensional or three‐dimensional environments to capture physiological responses in vitro. However, determining the concentration that must be tested in vitro remains challenging because several factors influence the in vitro concentrations of chemicals.

One of the major factors influencing the in vitro responses of chemicals is the amount of unbound chemicals present at the target site. The concentrations of the unbound chemicals primarily depend on the protein levels and their binding affinities. Human plasma contains 60–80 mg/mL of protein, and the predominant proteins are albumin, alpha‐1‐acid glycoprotein (AAG), lipoproteins, and globulins. 1 The concentrations of albumin and AAG range from 35 to 50 mg/mL (60% plasma protein, ~600 μM) and 0.6–1.2 mg/mL (1%–3% plasma protein, ~20.6 μM), respectively, and they have different chemical binding affinities. 1 In contrast, the total concentration of proteins in commercially obtained fetal bovine serum (FBS) is lower (30–45 mg/mL), and the amounts of albumin (~338 μM) and AAG (~13 μM) are also lower in comparison to human plasma. 1 Although the protein composition varies across species, the binding affinities (for example, to albumin site I) in bovine, rabbit, and rat to chemicals are similar to human 2 making cross‐species translation of protein binding valid. However, the binding of chemicals to proteins such as albumin and AAG is not the same. Albumin is alkaline and it is known to bind to acidic and neutral chemicals. In vivo binding to albumin is considered non‐saturable and reversible. The interaction of albumin with chemicals can be weak but has the potential to bind large amounts of chemicals. 3 Although bases bind well to albumin, they favor binding to plasma proteins like AAG, which is assumed to have high binding affinity but a low capacity. 4

In vitro binding of chemicals is never zero because they can bind to different constituents of the cell culture media. In vitro cell culture media typically contain 5%–10% FBS and other supplements. A small fraction of serum could result in large amounts of chemicals to remain protein bound. Chemicals that are greater than 99% protein bound in human plasma could still be greater than 90% protein bound in 5%–10% FBS‐containing cell culture media 3 resulting in a 10‐times higher in vitro bioavailability of chemicals that will shift the exposure‐response curve. Several methods, such as rapid equilibrium dialysis, ultrafiltration, ultracentrifugation, or high‐performance chromatography, have been developed to measure and interpret protein binding to chemicals. 5 , 6 However, it is experimentally challenging to determine protein binding for chemicals with greater than 99.5% binding affinity, so using theoretical modeling for extrapolation will be beneficial. 7 Several single‐ and multiple‐site binding models are available to describe chemical‐protein interactions. 7 , 8 , 9 Wan and Rehngren 9 applied a single protein binding model to a diverse set of chemicals with different physicochemical properties to elucidate that a single protein‐chemical binding kinetics. Myatt 1 applied a two‐protein binding site system using experimental data to calculate unbound chemical concentrations. Blakeley et al. 7 derived a cubic root‐based solution to describe chemical binding to proteins and ligands. Several other mathematical models to understand the protein binding and the fate of chemicals in the in vitro system have been developed but are not widely used. 10

Theoretically, in vitro pharmacological and toxicological assessments of chemicals using physiologically relevant unbound concentrations should mimic target site unbound concentrations. But depending on the protein and chemical levels, the in vitro unbound concentrations will differ and can influence exposure‐responses. In this study, the unbound fractions of 12 chemicals at different chemical concentrations and protein levels were calculated. The chemicals were selected because their plasma protein binding levels were determined in two or more experimental conditions (i.e., by varying protein or chemical concentrations).

METHODS

The protein binding‐corrected in vitro extracellular concentrations can be estimated using a one‐on‐one binding model described by McDonnell et al., 11 and Wan and Rehngren. 9 Briefly, the binding association constant (Ka) is given by Equation 1:

Ka=CtotCubCub2+CubPtotCubCtot (1)

where the subscripts tot and ub are the total and unbound levels of chemical (C) and protein (P). The Ka is calculated using experimental measurements performed using 100% human plasma or with FBS using maximum chemical concentrations. The percentage of unbound levels of chemical (fub%) for various protein levels at different chemical concentrations can be estimated using Equation 2:

fub%=100100×KaCtot+KaPtot+1KaCtot+KaPtot+124Ka2PtotCtot2KaCtot (2)

For chemicals with differential protein binding to more than one protein, for example, to plasma albumin and specifically to AAG, a one‐on‐one protein binding model for each protein was applied in a serial approach using physiologically relevant protein concentrations to estimate the fub%. Next, the serial protein binding model can be applied by assuming Ka1 (Ka for first protein, P1) to be higher than Ka2 (Ka for second protein, P2). The unbound concentrations remaining after binding to P1 given by Cub,P1 were calculated using Equation 3. Next, the unbound concentrations remaining after binding to P2 given by Cub,P2 were calculated by Equation 4. The percentage of unbound levels of chemical (fub%) remaining is calculated using Equation 5.

Cub,P1=Ctot1Ka1Ctot+Ka1P1+1Ka1Ctot+Ka1P1+124Ka12P1Ctot2Ka1Ctot (3)
Cub,P2=Cub,P11Ka2Cub,P1+Ka2P2+1Ka2Cub,P1+Ka2P2+124Ka22P2Cub,P12Ka2Cub,P1 (4)
fub%=Cub,P2×100Ctot (5)

The experimental data for different chemicals 1 , 3 , 12 , 13 , 14 , 15 , 16 , 17 were obtained from literature to derive the fub% at various protein levels. A user‐friendly Excel sheet for calculating protein binding‐corrected in vitro concentrations is provided in the Supplementary Material.

RESULTS

The unbound concentrations of chemicals in typical in vitro cell culture media containing 5%–10% FBS could be very different depending on that chemical's protein binding affinity. To perform reliable in vitro assessments, the in vitro extracellular concentrations should be corrected for in vitro protein binding to mimic the target unbound extracellular concentrations. The human plasma and FBS albumin protein concentrations are assumed to be 600 and 338 μM, respectively. 1 The Ka for each chemical was calculated for the experimental condition with maximum protein and chemical concentrations, and it was further used to predict the fub% and unbound in vitro concentrations. The unbound chemical levels vary depending on the total chemical concentration and amount of total protein present (Figure 1). In 100% human plasma, the experimentally measured protein binding of 10 μM nicotine is less than 29%. 18 The Ka for nicotine was determined by assuming 29% protein binding in 100% human plasma. Using the same Ka and FBS protein fractions, the bound fraction of nicotine in 5%–10% FBS is less than 3% and could be considered insignificant given other experimental factors. Like nicotine, ceftizoxime and moxifloxacin are less than 32% bound in 95% and 100% human plasma, respectively, 14 , 15 and the bound fraction for both in 5%–10% human plasma or FBS is less than 5% and could be considered insignificant for in vitro testing. The unbound chemical concentrations for the remaining chemicals at different concentrations could be very different in 5%–10% human plasma or FBS. The goodness of fit plot for predicted and measured fub% for all chemicals except lopinavir and ritonavir employing a one‐on‐one binding model are within two‐logfold difference, as shown in Figure 2. The one‐on‐one binding model calculated fub% should apply to most chemicals at physiologically relevant concentrations.

FIGURE 1.

FIGURE 1

Influence of protein and chemical concentrations on unbound concentrations. The unbound chemical concentration corrections for human plasma and FBS are compared. Experimental data (dots) for flavopiridol, 1 lopinavir, 3 , 12 ritonavir, 3 , 13 ceftriaxone, 14 cefoperazone, 14 moxalactam, 14 ceftizoxime, 14 moxifloxacin, 15 trovafloxacin, 15 RO4929097, 16 and thymoquinone 17 were obtained from the literature. FBS, fetal bovine serum.

FIGURE 2.

FIGURE 2

Predicted and experimental percent unbound chemical levels. The black solid line represent unity, the dotted line represents 1.25‐fold deviation, and the dashed line represents two‐fold deviation from model predictions. (fub%), percentage of unbound levels of chemical.

Certain chemicals, such as lopinavir and ritonavir, have rapid and higher preferential binding to AAG than other proteins. 3 Such differential binding would require a two‐step serial binding model to describe the fub%. The Ka for human plasma was determined for the high chemical concentration and total plasma albumin protein level. Next, the two‐step serial protein binding model was applied using physiological AAG concentrations and assuming a very high Ka for AAG (~100×) in comparison other proteins. First the chemical binding to high affinity protein (AAG) was calculated, then using the remaining unbound chemical concentration, the binding to the remaining protein was calculated to obtain the total unbound chemical concentrations. The model estimations for lopinavir and ritonavir at 5%, 10%, 20%, and 50% FBS are shown in Figure 1. The overall percentage of unbound concentration present in different levels of FBS varies with the total chemical concentration. The model estimated that at the physiologically relevant in vivo concentration of 10 μM, 5% FBS‐containing media would have unbound levels of 47.3% for lopinavir and 39.5% for ritonavir. The goodness of fit plot for predicted and measured percentage of unbound concentrations are shown in Figure 2 and the predictions for higher fub% are within two‐logfold but lower fub% are predicted to be much lower than experimental values.

DISCUSSION

In this work, the protein binding‐corrected concentrations for in vitro testing of different chemicals were calculated using one‐ and two‐protein binding models. Preliminary chemical screening is typically carried out using various cell lines cultured in 5%–10% FBS‐containing cell culture media. Depending on the protein binding levels and chemical concentration, the levels of unbound chemical available to elicit a response are very different (Figure 1). Several functional assays have demonstrated shifts in half‐maximal effective concentrations with varied protein levels, as the unbound chemical concentrations in the FBS‐containing cell culture media are different. 15 , 19 The dose–response for a typical functional cell‐based assay exhibits a sigmoid shape and is influenced by the chemical potency and unbound concentrations. If protein binding is not corrected, it is challenging to deconvolute a dose–response influenced by the chemical potency‐driven effect and total unbound concentration‐driven effect. This may eventually lead to poor in vitro to in vivo translation.

Protein binding is an important factor influencing the unbound concentrations and must be determined in the relevant matrix. Given the differential composition of proteins in serum, the levels of protein binding to chemical vary across species. 20 However, the chemical protein binding measured in a single matrix may be used to calculate the association constant and enable translation across species to obtain preliminary insights. Protein binding based on chemical physicochemical properties 10 and a quantitative structure–activity relationship has been proposed. 21 However, the estimated fub% still needs to consider protein‐binding kinetics to derive unbound concentrations at different total concentrations. The one‐on‐one binding model should be sufficient to describe the chemical‐protein binding of various chemicals at physiologically relevant concentrations (Figure 1). The overall amount of protein levels in FBS is 50% that of human plasma protein, and since cell culture media contains 5%–10% FBS, the protein binding of nicotine, ceftizoxime, and moxifloxacin can be considered negligible for in vitro testing (Figure 1). Chemicals with high protein binding (>98%) should be tested for binding affinities against other proteins, such as AAG, for more reliable estimations of unbound chemical concentrations. The simplified two‐protein binding model presented in this study can then be applied to determine unbound chemical concentrations. A higher prediction of fub% for highly protein binding chemicals (lopinavir and ritonavir) indicates the importance of applying computational methods to determine fub% especially considering that experimental measurements at lower concentrations could be challenging to determine 7 (Figure 2). This model formalism can potentially be extended to describe chemical binding to three proteins, assuming the association constants are significantly different and need further exploration. However, additional experimental data are needed to further validate such models.

Protein binding‐correction can potentially be used in two different approaches based on the experimental setup. For a dose–response curve performed using different concentrations, the total chemical concentrations can be corrected to obtain unbound chemical concentrations driven responses pre‐ or post‐experimentation. Alternatively, protein binding corrected total concentrations to mimic unbound physiologically relevant chemical concentrations of interest can be predetermined and evaluated in vitro. The present study has a few limitations. The level of chemical‐protein binding was not directly assessed across all chemical concentrations and was predicted based on limited experimental measurements. Further experimental data are required to enable verification of model performance across a wide range of chemicals. As several other factors influence in vitro unbound concentrations and their kinetics, mass‐balance models based on chemical partitioning are available to further characterize in vitro biokinetics and need to be evaluated separately. 10 Alternatively, in vitro cellular assays that require short‐term exposure to chemicals could potentially be evaluated in serum‐free media. Recent advances in in vitro testing led to the development of microphysiological systems containing different materials that may have higher chemical binding and needs to be evaluated. 22

In conclusion, the unbound chemical concentrations will vary depending upon the total chemical concentrations and protein levels. Chemical‐protein binding can be translated across different species and varied protein levels. In vitro cell‐based assays utilizing FBS‐containing media should be corrected for chemical‐protein binding to ensure reliable estimation of dose–response during pharmacological and toxicological assessments.

AUTHOR CONTRIBUTIONS

A.R.K. designed research, performed research, contributed new reagents/analytical tools, analyzed data, and wrote the manuscript.

FUNDING INFORMATION

Philip Morris International is the sole source of funding and sponsor of this research.

CONFLICT OF INTEREST STATEMENT

A.R.K. is an employee of Philip Morris International.

Supporting information

Data S1

ACKNOWLEDGMENTS

The author thanks Florian Calvino‐Martin for reviewing the manuscript. The author acknowledges scientific editorial, legal, and external affairs teams at Philip Morris International for reviewing the manuscript, which did not result in any scientific changes.

Kolli AR. Deriving protein binding‐corrected chemical concentrations for in vitro testing. Clin Transl Sci. 2023;16:2123‐2129. doi: 10.1111/cts.13616

DATA AVAILABILITY STATEMENT

A user‐friendly Excel spreadsheet to calculate protein binding and derive unbound concentrations for in vitro testing is included in the Supplementary Material. No experimental datasets were generated in this study. References for experimental data obtained from literature are provided in this article.

REFERENCES

  • 1. Myatt DP. The correlation of plasma proteins binding capacity and flavopiridol cellular and clinical trial studies. Biomed Spectrosc Imaging. 2017;6:59‐73. doi: 10.3233/BSI-170165 [DOI] [Google Scholar]
  • 2. Kosa T, Maruyama T, Otagiri M. Species differences of serum albumins: I. Drug binding sites. Pharm Res. 1997;14:1607‐1612. doi: 10.1023/a:1012138604016 [DOI] [PubMed] [Google Scholar]
  • 3. Hickman D, Vasavanonda S, Nequist G, et al. Estimation of serum‐free 50‐percent inhibitory concentrations for human immunodeficiency virus protease inhibitors lopinavir and ritonavir. Antimicrob Agents Chemother. 2004;48:2911‐2917. doi: 10.1128/AAC.48.8.2911-2917.2004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Smith SA, Waters NJ. Pharmacokinetic and pharmacodynamic considerations for drugs binding to alpha‐1‐acid glycoprotein. Pharm Res. 2018;36:30. doi: 10.1007/s11095-018-2551-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Pacifici GM, Viani A. Methods of determining plasma and tissue binding of drugs. Clin Pharmacokinet. 1992;23:449‐468. doi: 10.2165/00003088-199223060-00005 [DOI] [PubMed] [Google Scholar]
  • 6. Meyer‐Almes F‐J. Kinetic binding assays for the analysis of protein–ligand interactions. Drug Discov Toda Technol. 2015;17:1‐8. doi: 10.1016/j.ddtec.2015.08.004 [DOI] [PubMed] [Google Scholar]
  • 7. Blakeley D, Sykes DA, Ensor P, Bertran E, Aston PJ, Charlton SJ. Simulating the influence of plasma protein on measured receptor affinity in biochemical assays reveals the utility of Schild analysis for estimating compound affinity for plasma proteins. Br J Pharmacol. 2015;172:5037‐5049. doi: 10.1111/bph.13263 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Kalvass JC, Phipps C, Jenkins GJ, et al. Mathematical and experimental validation of flux dialysis method: an improved approach to measure unbound fraction for compounds with high protein binding and other challenging properties. Drug Metab Dispos. 2018;46:458‐469. doi: 10.1124/dmd.117.078915 [DOI] [PubMed] [Google Scholar]
  • 9. Wan H, Rehngren M. High‐throughput screening of protein binding by equilibrium dialysis combined with liquid chromatography and mass spectrometry. J Chromatogr A. 2006;1102:125‐134. doi: 10.1016/j.chroma.2005.10.030 [DOI] [PubMed] [Google Scholar]
  • 10. Proença S, Escher BI, Fischer FC, et al. Effective exposure of chemicals in in vitro cell systems: a review of chemical distribution models. Toxicol In Vitro. 2021;73:105133. doi: 10.1016/j.tiv.2021.105133 [DOI] [PubMed] [Google Scholar]
  • 11. McDonnell PA, Caldwell GW, Masucci JA. Using capillary electrophoresis/frontal analysis to screen drugs interacting with human serum proteins. Electrophoresis. 1998;19:448‐454. doi: 10.1002/elps.1150190315 [DOI] [PubMed] [Google Scholar]
  • 12. Kumar GN, Jayanti VK, Johnson MK, et al. Metabolism and disposition of the HIV‐1 protease inhibitor lopinavir (ABT‐378) given in combination with ritonavir in rats, dogs, and humans. Pharm Res. 2004;21:1622‐1630. doi: 10.1023/B:PHAM.0000041457.64638.8d [DOI] [PubMed] [Google Scholar]
  • 13. Denissen JF, Grabowski BA, Johnson MK, et al. Metabolism and disposition of the HIV‐1 protease inhibitor ritonavir (ABT‐538) in rats, dogs, and humans. Drug Metab Dispos. 1997;25:489‐501. [PubMed] [Google Scholar]
  • 14. Leggett JE, Craig WA. Enhancing effect of serum ultrafiltrate on the activity of cephalosporins against gram‐negative bacilli. Antimicrob Agents Chemother. 1989;33:35‐40. doi: 10.1128/AAC.33.1.35 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Zeitlinger M, Sauermann R, Fille M, Hausdorfer J, Leitner I, Muller M. Plasma protein binding of fluoroquinolones affects antimicrobial activity. J Antimicrob Chemother. 2008;61:561‐567. doi: 10.1093/jac/dkm524 [DOI] [PubMed] [Google Scholar]
  • 16. Wu J, LoRusso PM, Matherly LH, Li J. Implications of plasma protein binding for pharmacokinetics and pharmacodynamics of the γ‐secretase inhibitor RO4929097. Clin Cancer Res. 2012;18:2066‐2079. doi: 10.1158/1078-0432.Ccr-11-2684 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. El‐Najjar N, Ketola RA, Nissilä T, et al. Impact of protein binding on the analytical detectability and anticancer activity of thymoquinone. J Chem Biol. 2011;4:97‐107. doi: 10.1007/s12154-010-0052-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Kolli AR. Plasma protein binding and tissue retention kinetics influence the rate and extent of nicotine delivery to the brain. Toxicol Lett. 2023;380:69‐74. doi: 10.1016/j.toxlet.2023.04.006 [DOI] [PubMed] [Google Scholar]
  • 19. Schmidt S, Rock K, Sahre M, et al. Effect of protein binding on the pharmacological activity of highly bound antibiotics. Antimicrob Agents Chemother. 2008;52:3994‐4000. doi: 10.1128/AAC.00427-08 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Colclough N, Ruston L, Wood JM, MacFaul PA. Species differences in drug plasma protein binding. MedChemComm. 2014;5:963‐967. doi: 10.1039/C4MD00148F [DOI] [Google Scholar]
  • 21. Ghafourian T, Amin Z. QSAR models for the prediction of plasma protein binding. Bioimpacts. 2013;3:21‐27. doi: 10.5681/bi.2013.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Auner AW, Tasneem KM, Markov DA, McCawley LJ, Hutson MS. Chemical‐PDMS binding kinetics and implications for bioavailability in microfluidic devices. Lab Chip. 2019;19:864‐874. doi: 10.1039/c8lc00796a [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1

Data Availability Statement

A user‐friendly Excel spreadsheet to calculate protein binding and derive unbound concentrations for in vitro testing is included in the Supplementary Material. No experimental datasets were generated in this study. References for experimental data obtained from literature are provided in this article.


Articles from Clinical and Translational Science are provided here courtesy of Wiley

RESOURCES