Affinity Chromatography — from Textile Dyes to Synthetic Ligands by Design, Part I

At its widest definition, affinity chromatography encompasses techniques from immobilized metal chelate affinity to molecular imprinting and technologies from affinity capillary electrophoresis to affinity precipitation, affinity partitioning, and affinity membranes.¹ A very narrow definition would focus on specific or selective, reversible interactions between the immobilized ligand and target, dependent on a unique topological relationship involving orientation and molecular reactions. These include biological as well as biological mimic ligands. This is an area that has seen very significant growth and research development in recent years. The advent of affinity chromatography is usually attributed to Cuatrecasas, Wilchek, and Anfinsen.² In 1968 — a year after the introduction of cyanogen bromide activation — they described the purification of certain enzymes on "inhibitor gels" synthesized by coupling the inhibitor to CNBr-activated agarose. They introduced the notion of protein purification based on biologically functional pairs, the molecular recognition between a target protein and an immobilized partner. The technique was rapidly assimilated into the protein purification armory. Three years later, Cuatrecasas' review cited 100 papers and applications, ranging from immunoaffinity to the isolation of nucleic acids and the separation of complex cellular structures and cells.³ By 1984, the body of literature had swollen to 1,800 papers, and a 2004 Medline search revealed 35,000 affinity chromatography citations.⁴

Perhaps affinity chromatography's greatest success at scale has been monoclonal antibody purification. The demand for Protein A resins is more than 10,000 liters annually, increasing at 50% per year and representing a market in excess of $50 million in 2002.⁵ The use of immunoaffinity chromatography enables the production of both plasma-derived⁶ and recombinant⁷ coagulation factors VIII and IX, as well as other plasma proteins and biopharmaceuticals from natural and recombinant sources. Process development is constantly seeking high efficiency processing with the minimum number of steps and maximum output at the required purity. Larsson discussed customizing ligands to the separation task-at-hand instead of relying on "off-the-shelf," group-specific methodologies.⁸ Indeed the concept of "designer" adsorbents was introduced ten years earlier at the peak of expectations in biopharmaceutical product development, but at a time (1992) when FDA had approved only 12 recombinant products.⁹

Figure 1. Triazine Structure with Two Substitution Positions and a Spacer Arm to the Matrix.

SERENDIPITOUS ORIGINS

Subramanian described the serendipitous origins of dye affinity chromatography.

¹⁰

Blue dextran, a 2,000 kDa soluble polymer with the covalently bound dye Cibacron Blue F3GA, was developed to measure the void volume in gel filtration columns introduced in the early 1960s. It had been noted that some proteins, when co-chromatographed with Blue dextran, interacted with the marker. It was established in 1968 that the dye chromophore was responsible for binding.

^11,12

Thus, the evolution of dye affinity is concurrent with the general development of modern separation techniques. The introduction of the terms mimetic or biomimetic derives from the proposed (and now discarded) mechanism of dye affinity in which the structures of the chromophores mimic the naturally occurring heterocyclic nucleotides on which many proteins and enzymes depend.

¹³

The synthetic nature of the ligand also led to the introduction of the term pseudo-affinity, referring to the ligand's lack of biological function.

Undoubtedly the earliest (1973) and most studied ligand is Cibacron Blue F3GA (Ciba-Geigy), which is also available as Procion Blue H-B (Imperial Chemical Industries) and C.I. Reactive Blue 2 (the Color Index name). These textile dyes from different manufacturers are no longer available, but single synthetic analogue adsorbents are manufactured by at least two suppliers of chromatographic media. The first reported separations predate plasma proteome studies by 30 years and concern the depletion of serum albumin from human plasma to enable identification and purification of low concentration proteins.¹⁴ This was carried out using a Procion Blue dextran-Sepharose conjugate. At the time, Travis and Pannell were interested in a1-antitrypsin and described the difficult separation of this protein from albumin at high ionic strength where any non-specific ion exchange binding is at a minimum.¹⁵ This paper identified an initial problem with dye affinity chromatography, namely the leakage of the dye into the eluate (properties of synthetic adsorbents will be discussed next month in Part II of this article). In later studies, the authors used the dye directly conjugated with agarose, thus reducing the leakage problem, improving the binding capacity to 40 mg/mL gel, and enabling elution of the albumin under non-denaturing conditions.¹⁶

The 1970s academic focus on protein purification is particularly evident in a 1979 review listing over 100 different proteins and enzymes purified using immobilized triazine dyes.¹⁷ This also heralds a migration from a dye descriptor to a chemical identification of the core triazine. Additionally, it is notable that the dyes under investigation were triazine, substituted at two positions and coupled to the matrix through the third.

At this time it became evident that it was necessary to screen dye-ligand columns to find the most selective adsorbent.^18,19 Atkinson et al. noted that "this dye (C.I. Reactive Blue 2), however, is only one of a large family of triazine dyes, most of which bind proteins," indicating the need for a systematic approach to selection.²⁰ It was also believed at this time that these ligand adsorbents would be highly suited to downstream processing.²¹

Stellwagen (1990) noted that "since the color, and hence the structure, of each reactive dye is different, each reactive dye will have a somewhat different affinity for a bifunctional site on a given protein. Unfortunately, the affinity of a particular reactive dye for a bifunctional site cannot be predicted with any confidence, necessitating an empirical screening procedure to optimize chromatography."²² Only two years later, Lowe et al. wrote, "A fundamental advantage of affinity techniques is their predictive and rational character, since the ligand selected is designed to interact specifically with the protein to be purified."²³

At this pivotal point, dye affinity technology began to shift to de novo synthesis and rational ligand design.⁹ However, because some triazinyl derivatives are colored — although most are not — the concept of dye ligand chromatography, in which the ligand is chosen from a random battery of commercial dyestuffs, has lived on past its expiration date. Two reviews have summarized the field in the context of bioprocessing applications.^{24, 25}

GENESIS OF SYNTHETIC LIGAND ALTERNATIVES

The shortcomings of dyes as affinity ligands has been recognized for some time. They are rarely single chemical entities and — produced as bulk chemicals for a consumer product industry — they do not fulfill the rigorous requirements of biopharmaceutical development and manufacturing. It had been noted that textile dyes contain isomers of the main product, stabilizing and diluent agents, as well as anti-dusting agents such as dodecylbenzene. Of course, these impurities and contaminants must be removed.

²⁶

In one study of Cibacron Blue F3GA, up to 15 different colored components were identified. Furthermore, different preparations did not necessarily contain the same components.

²⁷

In 1988, Burton et al. reported, for the first time, the synthesis of single isomer variants of C.I. Reactive Blue 2, thus marking the link between dye structure and protein binding.

²⁸

Additionally, the interactions between target proteins and dye structures present many different alternatives for highly specific binding of proteins at their active sites and less specific interactions at other sites. These reactions may include a complex combination of electrostatic, hydrophobic, hydrogen bonding, and charge transfer reactions. Thus, ligand engineering or design is a way of targeting specific binding.

Work on biomimetic ligands, as defined by Clonis, was initiated in 1984 and is exemplified by two development projects.²⁹ The first biomimetic dye ligand was developed for trypsin by linking benzamidine to the reactive chlorotriazine through a diaminomethylbenzene group.²⁹ The ligand was designed on the basis of the unusual cationic substrate preference of trypsin-like enzymes. In a second example, Reactive Blue 2 was specifically redesigned to confer specificity for calf intestinal alkaline phosphatase.³⁰ From there, it was a short step to generate well-defined dye adsorbents in order to develop a series of adsorbents for which the ligand structures and the ligand density were known and could be controlled.³¹ However, there are many more variables in the performance of affinity chromatography. Greater degrees of sophistication are required, both in the design of the ligand and the ligand-adsorbent conjugate and in the execution of the technique. In a second step toward designed ligands (but still restricted to the concept of "designer dyes"), the interaction between the "parent" dye (Cibacron Blue) and analogs with horse liver alcohol dehydrogenase was established using X-ray crystallography.³² A new set of terminal-ring (opposite to the anthraquinone) analogs were synthesized with favorable affinity for ADH. These ligands belong to a second generation of adsorbents, the results of rational molecular design techniques.

Lowe et al. reviewed the advent of computer-aided ligand design, which has subsequently developed to include design based on X-ray crystallographic data, NMR protein structures, and homology data from suitable databases.³³ This has been made possible by concomitant software developments.³³ , design may start from peptide ligand models as mimics of protein-protein binding interactions. Peptides themselves can be used as ligands, but present an expensive and less stable choice for bioprocessing. Instead, the peptide template can be used to model a synthetic alternative.

There are several different approaches to ligand design and synthesis that form the basis of the technology used today for commercial development of third-generation adsorbents, including:

• optimization of existing ligands (analog synthesis);

• rational design (computer modeling of ligand structures);

• systematic screening of ligand arrays;

• rational design combined with library construction and screening.

Often, insufficient structural information is available for a target protein, and, in some cases, even the target may be unknown. If a biological ligand is known or a peptide ligand has already been developed, those models may be used. It is not surprising, therefore, that most commercially developed ligands and adsorbents have been derived from targeted construction and screening of solid phase libraries. Lowe recently reviewed combinatorial approaches to affinity chromatography.³⁴ Although virtual libraries can be constructed from, for example, the Available Chemicals Directory's 2.5 million entities, and these libraries can be reduced to real and manageable sub-libraries, this route has not been commercially successful for bioprocessing.³⁵ It is important to note that earlier screening of solution phase libraries has also failed to produce any commercial adsorbents, since the orientation of the ligand to its target in solution and as an immobilized entity may present different aspects to the protein. In this article, only solid phase libraries and screening will be discussed.

Figure 2. Basic Chemistry of Ligand Synthesis from Dichlorotriazine-Agarose

It is possible to identify two major steps in the development of "customized" ligands: 1) construction and screening of suitable libraries, known as intelligent combinatorial chemistry and 2) development of the ligand adsorbent conjugate.

LIBRARY CONSTRUCTION

Customized robotics systems are available (for example, see

www.chemspeed.com

) that enable the parallel synthesis of reasonably large libraries. A 96-well plate format is convenient. A successful approach employed by ProMetic BioSciences is to construct libraries in micro-column arrays in which each well contains 250 L of packed adsorbent (PuraBead). Each unique adsorbent is based on a 100m agarose bead media, epoxy-activated and coupled to a triazine, with or without a spacer of one to six carbons. Triazine is selected because of its stability, ease of derivation, and safety. The resultant dichlorotriazine can be derived either symmetrically (with the same ligand at each of the two positions) or with different moieties at each of the substitution sites.

³⁶

The reaction scheme, where R1 and R2 are amine substituents, is shown in Figure 2. One thousand amines provide a diversity of 500,000 different ligands based on a single triazine scaffold.

Where v is the total number of possible combinations and n is the number of amines.

In addition, triazine lends itself to dendrimetic expansion, and therefore larger, branched ligands may be developed with even greater diversity than those available from single triazine derivation.³⁷ Such ligands offer the possibility of improved specificity and selectivity for "difficult-to-isolate" molecular species.

Diversity-optimized combinatorial libraries can be constructed in silico using Cerius2 (www.accelrys.com) software. Diverse "R" groups are used as input, "combichem" molecular descriptors are calculated for all fragments, and a principal component analysis (PCA) is performed. A coverage-based diversity selection is used to produce a subset of the PCA space, and an analogue builder tool is used to combine R groups with a triazine core. In this way, a diversity-optimized virtual library is constructed. R group subsetting and proportional sampling functions are used to select sets of R groups to generate a 16 x 16 combinatorial library with maximum diversity.³⁸ These libraries, or subsets of them, may then be synthesized to contain anionic, neutral, cationic, hydrophilic, and hydrophobic R groups in a 96-well format.

LIBRARY SCREENING

General libraries are screened — the primary screen — under the desired separation conditions (conditions of the feed stream, stability of the protein, etc.) of defined pH, ionic strength, and buffer composition. The screen is carried out in a straightforward load, wash, elute, sanitize sequence using, for example, a Tecan Genesis (

www.tecan.com

) robotic liquid handing system. Eluates are assayed for the target protein. The initial screen allows the identification of lead ligands with promising binding and dissociation characteristics. These ligands are selected, and sublibraries are constructed using analogues of the general library's primary amines. The sublibrary is screened - the secondary screen - using the conditions of the primary screen. Further iterations may be necessary to home in on the target ligand. A typical screening result is shown in Figure 3.

Figure 3. Secondary Screen of a Symmetrical Library (# XY) (8 x 8 array) for a Recombinant Albumin. The figures represent protein concentration using a simple BCP assay. The ligand B2 demonstrate a development opportunity, since near neighbors do not exhibit similar binding and elution properties.

A lead ligand is selected for development. The development program involves three stages, starting with discovery, in which the selectivity of the target ligand is established to meet a desired purity (activity) specification. In the second stage, adsorbent and chromatography development lead to the establishment and optimization of capacity, recovery, binding, and elution conditions and the alkali resistance (stability) of the adsorbent. In the third stage, the adsorbent is scaled up for manufacturing, stability and leakage studies are initiated, and safety (toxicity) data are generated in preparation for submitting a Drug Master File to FDA. This development process is summarized in Figure 4 for a typical customized ligand and adsorbent development program from a client-oriented vendor.

Figure 4. Development Timeline and Major Activities for Customized, Synthetic Ligand Adsorbents. The commercialization stage entails preparation for routine manufacturing, consistency lot production, and initiation of long-term studies. Drug master file information also is generated at this stage.

Part II of this article will conclude in the August 2004 issue and include properties and applications of synthetic ligand adsorbents, as well as a look into the future.

REFERENCES

1. An overview of affinity chromatography. Wilchek M, Chaiken I.

Methods Mol Biol.

2000;147:1-6.

2. Selective enzyme purification by affinity chromatography. Cuatrecasas P, Wilchek M, Anfinsen CB. Proc Natl Acad Sci U S A. 1968 61(2):636-43.

3. Affinity chromatography. Cuatrecasas P, Anfinsen CB. Annu Rev Biochem. 1971;40:259-78.

4. Affinity chromatography. Wilchek M, Miron T, Kohn J. Methods Enzymol. 1984;104:3-55.

5. Meeting the increasing demand for Protein A chromatography media. Ramberg M. Antibody Production and Downstream Processing. IBC Conference 13-15 February 2002, San Diego, CA, USA.

6. Use of monoclonal antibodies for purification of coagulation factors. Wah P-S. IBC Conference: Biological Safety and Production 99. 19-23 April 1999, Boston, MA, USA.

7. Production processes of licensed recombinant factor VIII preparations. Boedeker BG. Semin Thromb Hemost. 2001 Aug;27(4):385-94.

8 Customized ligands optimise affinity chromatography procedures. Larsson L-J. Biopharm Europe September 2000: 69-71.

9. Designer dyes: "biomimetic" ligands for the purification of pharmaceutical proteins by affinity chromatography. Lowe CR, Burton SJ, Burton NP, Alderton WK, Pitts JM, Thomas JA. Tibtech 1992; 10: 442-448.

10. Dye-ligand affinity chromatography: the interaction of Cibacron Blue F3GA with proteins and enzymes. Subramanian S. CRC Crit Rev Biochem. 1984; 16(2): 169-205.

11. Some kinetic and molecular properties of yeast phosphofructokinase. Kopperschlager G, Freyer R, Diezel W, Hofmann E. FEBS Lett. 1968 Aug;1(3):137-141.

12. Purification and allosteric properties of yeast pyruvate kinase. Haeckel R, Hess B, Lauterborn W, Wuster KH. Hoppe Seylers Z Physiol Chem. 1968 May;349(5):699-714.

13. Spectroscopic studies of Cibacron Blue and Congo Red bound to dehydrogenases and kinases. Evaluation of dyes as probes of the dinucleotide fold. Edwards RA, Woody RW. Biochemistry. 1979 Nov 13;18(23):5197-204.

14. Selective removal of albumin from plasma by affinity chromatography, Travis, J., Pannell, R. Clin. Chim. Acta 49 (1973) 49-52.

15. A method for preparing albumin-free plasma. Travis, J., Pannell, R. Behring Inst. Mitt. 54 (1974) 30-32.

16. Isolation of albumin from whole human plasma and fractionation of albumin-depleted plasma. Travis, J., Bowen, J., Tewksbury, D., Johnson, D., Pannell, R. Biochem. J. 157 (1976) 301-306.

17. Protein purification using immobilised triazine dyes. Dean PD, Watson DH. J Chromatogr. 1979 Oct 1; 165(3): 301-19.

18. Use of triazine dyes in the affinity chromatographic purification of alkaline phosphatase from calf intestine. Bouriotis V, Dean PD. J Chromatogr. 1981 Feb 27; 206(3): 521-30.

19. Affinity chromatography on immobilised triazine dyes. Studies on the interaction with multinucleotide-dependent enzymes. Clonis YD, Lowe CR. Biochim Biophys Acta. 1981 May 14; 659(1): 86-98.

20. Triazine-dye affinity; chromatography. Atkinson T, Hammond PM, Hartwell RD, Hughes P, Scawen MD, Sherwood RF,Small DA, Bruton CJ, Harvey MJ, Lowe CR. Biochem Soc Trans. 1981 Aug; 9(4): 290-3.

21. The applications of reactive dyes in enzyme and protein downstream processing. Clonis YD. Crit Rev Biotechnol. 1988; 7(4): 263-79

22. Chromatography on immobilized reactive dyes. Stellwagen E. Methods Enzymol. 1990; 182: 343-57.

23. Designer dyes: "biomimetic" ligands for the purification of pharmaceutical proteins by affinity chromatography. Lowe CR, Burton SJ, Burton NP, Alderton WK, Pitts JM, Thomas JA. Tibtech 1992; 10: 442-448.

24. Dye-affinity techniques for bioprocessing: recent developments. Garg N, Galaev IYu, Mattiasson B. J Mol Recognit. 1996 Jul-Aug; 9(4): 259-74.

25. Dye-ligand affinity systems. Denizli A, Piskin E. J Biochem Biophys Methods. 2001 Oct 30; 49(1-3): 391-416.

26. Theory and practice in affinity techniques. Haff LA, Easterday RL. In: Sundaram PV, Eckstein F (eds) Academic Press, London 1978.

27. Analytical evaluation of the purity of commercial preparations of Cibacron Blue F3GA and related dyes. Hanggi D, Carr P. Anal Biochem. 1985 Aug 15; 149(1): 91-104.

28. Design and applications of biomimetic anthraquinone dyes. I. Synthesis and characterisation of terminal ring isomers of C.I. Reactive Blue 2. Burton, SJ, McLoughlin, SB, Stead CV, Lowe CR. J. Chromatogr. A 435; 1988: 127-137.

29. Biomimetic dyes as affinity chromatography tools in enzyme purification. Clonis YD, Labrou NE, Kotsira VP, Mazitsos C, Melissis S, Gogolas G. J Chromatogr A. 2000 Sep 1; 891(1): 33-44.

30. Design and applications of biomimetic anthraquinone dyes. Purification of calf intestinal alkaline phosphatase with immobilised terminal ring analogues of C.I. reactive blue 2. Lindner NM, Jeffcoat R, Lowe CR. J Chromatogr. 1989 Jun 28;473(1):227-40.

31. Well defined dye adsorbents for protein purification. Scoble J, Scopes R. J Mol Recognit. 1996 Sep-Dec; 9(5-6): 728-32.

32. Design and applications of biomimetic anthraquinone dyes. III. Anthraquinone-immobilised C.I. reactive blue 2 analogues and their interaction with horse liver alcohol dehydrogenase and other adenine nucleotide-binding proteins. Burton SJ, Stead CV, Lowe CR. J Chromatogr. 1990 May 25;508(1):109-25.

33. New developments in affinity chromatography with potential application in the production of biopharmaceuticals. Lowe CR, Lowe AR, Gupta G. J Biochem Biophys Methods. 2001 Oct 30;49(1-3):561-74.

34. Combinatorial approaches to affinity chromatography. Lowe CR. Curr Opin Chem Biol. 2001 Jun;5(3):248-56.

35. Rational design, synthesis and verification of affinity ligands to a conserved antibody pocket. Ersoy O, Carredano E, Gronberg A, Steensma E, Zou J. Recovery of Biological Products XI, Banff, Canada. 14-19 September 2003.

36. International Patent Application PCT/DK96/00399

37. Orthogonal, convergent syntheses of dendrimers based on melamine with one or two unique surface sites for manipulation. Zhang W, Nowlan DT 3rd, Thomson LM, Lackowski WM, Simanek EE. J Am Chem Soc. 2001 Sep 19;123(37):8914-22.

38. Application of affinity ligand technology to plasma proteomics. Burton S, Baines D, Tatton H, Pearson J. Presentation at Plasma Product Biotechnology Meeting, 22-26 April, 2003. Curacao, Netherlands Antilles.