Affinity Separation

AFFINITY SEPARATION

K. Jones, Affinity Chromatography Ltd, Freeport,Ballsalla, Isle of Man, UK

Copyright ^ 2000 Academic Press

Introduction

Of the collection of separation technologies knownas ‘afRnity’, afRnity chromatography is by far themost widely used variant. AfRnity chromatography isbecoming increasingly important as the speed of therevolution taking place in biotechnology processingincreases. The concept of an ‘afRnity’ separation re-sults from a naturally occurring phenomenon existingwithin all biological macromolecules. Each biologicalmacromolecule contains a unique set of intermolecu-lar binding forces, existing throughout its internaland external structure. When alignment occurs be-tween a speciRc site of these forces in one moleculewith the site of a set of forces existing in another(different) molecule, an interaction can take placebetween them. This recognition is highly speciRc tothe pair of molecules involved. The interactive mech-anism can be converted into a universal mutual bind-ing system, where one of the binding pair is attachedto an inert matrix, packed into a column and usedexclusively to capture the other matching molecule.When used in this (afRnity) mode, the technique isprobably the simplest of all chromatographicmethods. It is, however, restricted almost exclusivelyto the separation and puriRcation of biological mac-romolecules, and is unsuitable for small molecules.

AfRnity chromatography or bioselective adsorp-tion chromatography was Rrst used in 1910, but it wasonly in the 1960s that afRnity chromatography aspractised today was developed as a puriRcation tech-nique. By the late 1970s the emergence of recombinantDNA technology for the manufacture of protein phar-maceuticals provided a new impetus for this highlyspeciRc chromatographic method, implemented by thedemand for ever-increasing product purity implicit inregulatory frameworks devised by (amongst others)the USA’s Food and Drug Administration (FDA). Fi-nally, the need to reduce the cost of drugs is underconstant scrutiny by many Governments, particularlythose with controlled health schemes funded by rev-enue raised by taxation. These mutually incompatiblepressures indicate the need for more efRcient sep-aration systems; the afRnity technique provides thepromise of meeting all necessary requirements.

Separation and puriRcation methods for biologicalmacromolecules vary from the very simple to theesoteric. The type of technique adopted is basicallya function of source, the fragility of the molecule andthe purity required. Traditionally, high purity proteinpharmaceuticals have used multistage processing, butthis is very inefRcient as measured by the well-documented fact that 50}80% of total productioncosts are incurred at the separation/puriRcation stage.In contrast, the highly selective indigenous propertiesof the afRnity method offer the alternative ofvery elegant single-step puriRcation strategies. Theinherent simplicity and universality of the method hasalready generated a wide range of separation tech-nologies, mostly based upon immobilized naturallyoccurring proteinaceous ligands. By comparing the‘old’ technologies of ‘natural’ ligands or multistageprocessing with the ‘new’, exempliRed by syntheticdesigned ligands, the most recent advances in af-Rnity processing can be described.

Biological Recognition

As nature evolved, life forms had to develop a protec-tive mechanism against invading microorganisms ifthey were to survive. Thus there is a constant battlebetween the cell’s defence mechanism and the attack-ing microorganisms, a battle resolved by the cellsgenerating antibodies (the immunoglobulins) able torecognize the protein coat of attacking microorgan-isms and signal killer cells to destroy the invadersbefore they cause harm to the host. Equally, if micro-organisms were to survive, they had continually tomutate and change their protein coats to avoid detec-tion by existing antibodies. The ‘attack and destroy’process is a function of changes in the molecularstructure in a speciRc part of the protein, with onlythe most minute of changes occurring at the surface ofthe protein. Evolution has thus designed a systemwhere every protein has a very precise structure, butone which will always be recognized by another. Oneelement of the interacting pair can be covalentlybonded onto an inert matrix. The resulting chromato-graphic medium can then be packed into a column,and used to separate exclusively its matching partnerfrom an impure mixture when added as a solution tothe top of the column. This fact can be stated asfollows } for every protein separation problem thereis always an afTnity solution. The process ofproducing a satisfactory medium is quite difRcult.

Sepsci*1*TSK*Venkatachala=BGI / AFFINITY SEPARATION 3

Table 1 Affinity ligands and purified proteins

Immobilized ligand Purified protein

Divalent and trivalent metal ion Proteins with an abudance of his, tryp and cys residuesLectins Glycoproteins, cellsCarbohydrates LectinsReactive dyes Most proteins, particularly nucleotide-binding proteinsNucleic acids Exo and endonucleases, polymerases, other nucleic acid-binding proteinsAmino acids (e.g. lys, arg) ProteasesNucleotides, cofactors substrates and inhibitors EnzymesProteins A and G ImmunoglobulinsHormones, drugs ReceptorsAntibodies AntigensAntigens Antibodies

The matching pair must be identiRed, and oneof them isolated in a pure form. Covalent bond-ing onto an inert matrix in a stable manner mustalways allow the ‘docking’ surface of the protein tobe positioned to make it available to the target pro-tein. The whole also has to be achieved at an accept-able cost.

This technique has resulted in many successful ap-plications, often using antibodies as the afRnitymedium (immunoafRnity chromatography), butlarge scale separations using these ‘natural’ ligandsare largely restricted by cost and regulatory reasons.Although immunoafRnity chromatography is stillwidely practised, in recent years the evolution ofdesign technologies has provided powerful new ap-proaches to mimic protein structures, resulting in thedevelopment of synthetic ligands able to work inharsh operational environments and at low cost.

The Af\nity Process

The afRnity method is critically dependent uponthe ‘biological recognition’ existing between species.By permanently bonding onto an inert matrix a mol-ecule (the ligand) that speciRcally recognizes the mol-ecule of interest, the target molecule (the ligate) canbe separated. The technique can be applied to anybiological entity capable of forming a dissociablecomplex with another species. The dissociation con-stant (Kd) for the interaction reSects the comp-lementarity between ligand and ligate. The optimalrange of Kd for afRnity chromatography lies be-tween 10�4 and 10�8 mol L�1. Most biologicalligands can be used for afRnity purposes provid-ing they can be immobilized, and once immobilizedcontinue to interact successfully with their respectiveligates. The ligand can be naturally occurring, anengineered macromolecule or a synthetic molecule.Table 1 provides some examples of immobili-zed ligands used to purify classiRed proteins. The

afRnity method is not restricted to protein separ-ations; nucleic acids and whole cells can also beseparated.

The simplicity of the chromatographic process isshown in Figure 1. The ligand of interest, covalentlybonded onto the inert matrix, is contained in thecolumn, and a solution containing the target (theligate) is passed through the bed. The ligand recog-nizes the ligate to the exclusion of all other molecules,with the unwanted materials passing through the col-umn packing while the ligate is retained. Once thebed is saturated with the target molecule (as mea-sured by the breakthrough point), contaminating spe-cies are washed through, followed by collection of thetarget molecule as a very pure fraction using aneluting buffer solution. Finally, the column iscleansed from any strongly adsorbed trace materials,usually by regeneration with a strong alkali or acid,making it available for many more repeat runs. Anoutstanding advantage of the afRnity process isan ability to concentrate very dilute solutions whilestabilizing the captured protein once adsorbed ontothe column. Many of the in-demand proteins manu-factured by genetically engineered microorganismsare labile, allowing only minute quantities to be pres-ent in the fermentation mix before they begin todeteriorate. An ability to capture these very smallquantities while stabilizing them in the adsorbantphase results in maximization of yield, making mass-ive savings in total production costs.

Although the technical processing advantages areclear there is a major difRculty in the appli-cation of afRnity chromatography as understoodby most practitioners today. Most ligands describedin Table 1 suffer from two primary disadvan-tages: a lack of stability during use; and highcost. Fortunately these problems have now been over-come, and afRnity chromatography is now accep-ted as the major separations technology forproteins.

4 I / AFFINITY SEPARATION / Derivatization

Figure 1 Schematic diagram of affinity chromatography.

Table 2 Support matrices

Support matrix OperationalpH range

Agarose 2}14Cellulose 1}14Dextran 2}14Silica (8Glass (8Polyacrylamides 3}10Polyhydroxymethacrylates 2}12Oxirane}acrylic copolymers 0}12Styrene}divinylbenzene copolymers 1}13Polyvinyl alcohols 1}14N-Acryloyl-2-amino-2-hydroxy-1, 2-propane 1}11PTFE Unaffected

PTFE, polytetrafluoroethylene.

Matrices

By deRnition matrices must be inert and play no partin the separation. In practice most play a (usually)negative role in the separation process. To minimizethese disadvantages matrices have to be selected withgreat care. There is a theoretically perfect matrix,deRned as consisting of monodispersed perfectlyshaped spheres ranging from 5 to 500 �m in dia-meter, of high mechanical strength, zero nonspeciRcadsorption and with a range of selectable pore sizesfrom 10}500 nm, a very narrow pore size distributionand low cost. This idealized matrix would then pro-vide the most efRcient separation under all ex-perimental conditions. As always, a compromise hasto be reached, the usual approach being to accentuatethe most attractive characteristics while minimizingthe limitations, usually by manipulating the experi-mental conditions most likely to provide the optimumresult.

The relative molecular masses of proteins varyfrom the low thousands to tens of millions, makingpore size the most important single characteristic ofthe selected matrix. Very large molecules need veryopen and highly porous networks to allow rapid andeasy penetration into the core of the particle. Struc-tures of this type must therefore have very large pores,but this in turn indicates low surface areas per unitvolume, suggesting relatively low numbers of surfacegroups to which ligands can be covalently attached.The matrix must also be biologically and chemicallyinert. A special characteristic demanded from biolo-gical macromolecular separations media is an abilityto be sanitized on a routine basis without damage.This requires resistance to attack by cleansing re-agents such as molar concentrations of strong alkali,acids and chaotropes. In contrast to analytical separ-ations, where silica-based supports are inevitablyused, silica cannot meet these requirements and isgenerally not favoured for protein separations.Table 2 contains examples of support matrices usedin afRnity separations.

The beaded agaroses have captured over 85% ofthe total market for biological macromolecule separ-ations, and are regarded as the industry standard towhich all other supports are compared. They haveachieved this position by providing many of the desir-able characteristics needed, and are also relativelyinexpensive. Beaded agaroses do have one severe lim-itation } poor mechanical stability. For analyticalapplications speed and sensitivity are essential, de-manding mechanically strong, very small particles.Beaded agaroses are thus of limited use analytically,a gap Rlled by high performance liquid chromatogra-phy (HPLC) using silica matrices. For preparative andlarge scale operations other factors are more impor-tant than speed and sensitivity. For example, masstransfer between stationary phase and mobile phase ismuch less important when compared to the contri-bution from the chemical kinetics of the bindingreaction between stationary phase and protein.Band spreading is also not a serious problem. Whencombined with the highly selective nature of theafRnity mechanism, these factors favour the com-mon use of large sized, low mechanical strength par-ticles.

In recent years synthetic polymeric matrices havebeen marketed as alternatives. Although nonbiodeg-radable, physically and chemically stable, with goodpermeabilities up to molecular weights greater than107 Da, the advantages provided are generally off-set by other quite serious disadvantages, exempliRedby high nonspeciRc adsorption. Inorganic matriceshave also been used for large scale protein separ-ations, notably reversed-phase silica for large scalerecombinant human insulin manufacture (molecularweight approximately 6000 Da), but are generallynot preferred for larger molecular weight pro-ducts. A very slow adoption of synthetic matrices is


Table 3 Activation materials

Activating reagent Bonding group on ligand

Cyanogen bromide Primary aminesTresyl chloride Primary amines, thiolsTosyl chloride Primary amines, thiolsEpichlorohydrin Primary amines, hydroxyls, thiols1,4-Butanediol diglycidyl ether Primary amines, hydroxyls, thiols1,1�-Carbonyldiimidazole Primary amines, hydroxylsCyanuric chloride Primary amines, hydroxylsDivinylsulfone Primary amines, hydroxyls2-Fluro-1-methylpyriinium-toluene-4-sulfonate Primary amines, thiolsSodium periodate Primary aminesGlutaraldehyde Primary amines

indicated as improvements are made to current ma-terials and the prices of synthetics begin to approachthose of agarose beads. Other factors resist any signif-icant movement towards synthetic matrices. Mostinstalled processing units are designed for low perfor-mance applications. Higher performance matriceswould need reinstallation of new, much higher costhigh performance plant; plant operators would needretraining; operating manuals would need rewriting;and plant and factory would need reregistration withthe FDA. In combination, the implication is thatpenetration of high performance systems for largescale applications will be slow, and agarose beads willcontinue to dominate the market for protein separ-ations.

Covalent Bonding

A basic requirement of all chromatographic media isthe need for absolute stability under all operationalconditions through many cycles of use. Consequentlyall ligands must be covalently bonded onto thematrix, and various chemistries are available toachieve this.

A number of factors are involved:

1. The performance of both ligand and matrix arenot impaired as a result of the coupling process.

2. Most of the coupled ligand is easily accessible tothe ligate.

3. Charged or hydrophobic groups are not generatedon the matrix, so reducing nonspeciRc adsorption.

4. The immobilized ligand concentration is optimalfor ligate bonding.

5. There is no leakage of immobilized ligand fromthe matrix.

Some ligands are intrinsically reactive (or can bedesigned to be so) and contain groups that can be

coupled directly to the matrix, but most require coup-ling via a previously activated matrix. The afRn-ity matrix selected must have an adequate number ofappropriate surface groups onto which the ligand canbe bonded. The most common surface group is hy-droxyl. The majority of coupling methods involve theactivation of this group by reacting with entities con-taining halogens, epoxy or carbonyl functionalgroups. These surface residues are then coupled toligands through primary amines, hydroxyls or thiolgroups, listed in Table 3.

Polysaccharides, represented by agarose, havea high density of surface hydroxyl groups. Traditionstill dictates that this surface is activated by cyanogenbromide, but it is well established that this reagentforms pH-unstable iso-urea linkages, resulting ina poorly performing product. Furthermore CNBr-activated agarose needs harsh coupling conditionsif high yields of Rnal media are to be obtained,suggesting high wastage of often expensive ligands.This factor is particularly evident with fragile entitiessuch as the very-expensive-to-produce antibodies,and yet many workers simply read previous literatureand make no attempt to examine alternative farsuperior coupling methods. The advantages of mildcoupling regimes are demonstrated in Figure 2, wherethe use of a triazine-activated agarose is compared toCNBr-activated agarose. Yield is signiRcantly in-creased, largely by coupling under acidic rather thanalkaline conditions.

Intermolecular Binding Forces

Almost all chromatographic separations rely uponthe interaction of the target molecule with eithera liquid phase or a covalently bonded molecule on thesolid phase, the exceptions being those relying uponmolecular size, e.g. molecular sieves and gel Rltra-tion. In afRnity separations ligates are inevitably


Figure 2 Triazine coupling. (A) Coupling of human serum albumin (HSA) to ready-acitivated supports as a funciton of pH. (B) Timecourse of coupling of human IgG to ready-activated supports at 43C. , CNBr-activated agarose 4XL; , triazine-activated agarose4XL.

complex biological macromolecules or assemblies,mostly or exclusively consisting of amino acids enti-ties linked together in a speciRc manner. This com-plexity of structure provides many opportunities toexploit the physicochemical differences betweenthe target molecule and the ligand to be used. Eachstructure contains the four basic intermolecular bind-ing forces } electrostatic, hydrogen bonding, hydro-phobic and van der Waals interactions } spreadthroughout the structure in an exactly deRned spatialmanner. The degree of accessibility and spatial pre-sentation within the pore of the medium, and thestrength of each force relative to each other, dictatewhether these forces are utilized to effect theseparation. The biological recognition between spe-cies is a reSection of the sum of the various mo-lecular interactions existing between them, andthis summation is Rxed for the ligate. However,various ligands may be found that emulate someor all of the available binding forces to variousdegrees.

AfRnity adsorbents are therefore assigned toone of three broad ligand categories: nonspeciRc,group speciRc or highly speciRc. NonspeciRc ligandshave only a superRcial likeness to biological ligandsand binding is usually effected by just one of thefour binding processes described above. Although ionexchange materials can be used in a similar manner toafRnity adsorbents, they only exhibit the singleforce of electrostatic binding. They are thus limited torelatively indiscriminate binding. In this case the onlycriterion for binding is that of an overall charge.

Fortunately there are a vast number of biologicalligands that can interact with more than one macro-molecule and consequently group-speciRc ligands arecommonplace. Since group-speciRc adsorbents retaina range of ligates with similar binding requirements,a single adsorbent may be used to purify a number ofligates. Group-speciRc ligates can be used when thedesired ligate is present in high concentration, but thisimplies that some preprocessing has taken place anda concentration step interposed. The use of non/group-speciRc adsorbents can only offer partialseparations. This results in having to apply severalstages in series, each only capable of removing a pro-portion of the impurities. In contrast a highly selec-tive ligand can exclusively remove the target in onestep, but often the resulting complexes are very tight-ly bound, have low binding capacities, are easilydenatured and are expensive to produce. Until re-cently these adsorbents were restricted to technicallydifRcult isolations. Today the use of computer-assisted molecular modelling systems provides oppor-tunities to investigate relationships between designedligands and relevant protein structures. For the Rrsttime logical design approaches can be applied andconsequently stable inexpensive ligands have nowbecome available.

Analytical Scale-Up

Modern biotechnology uses two different typesof chromatography. Analytical separations requirethat run time is minimized, while resolution and


sensitivity are maximized. In contrast, for preparativeand process applications, the objective is to maximizepurity, yield and economy. These techniques havedeveloped separately, simply because biological mac-romolecules pose unique difRculties, makingthem unsuitable for ‘standardized’ analysis. A majorinSuence on this division has been that scale-up usu-ally occurs very much earlier in the development ofa process, causing biochemists to turn to the tradi-tional low performance methods of ion exchange(IE), hydrophobic interaction (HI) and gel per-meation chromatography (GPC). The highly efR-cient afRnity chromatography method was gener-ally ignored, primarily because of the difRculty ofhaving to develop a unique ligand for each separationrather than having ‘off-the shelf’ column pack-ings immediately available from external suppliers.For analytical purposes high performance afRnityliquid chromatography (HPALC) is a rarity, a func-tion of the limited availability of suitable matrices(Table 2) and the afRnity process itself.

Where quantiRable high speed chromatography isrequired, reversed-phase HPLC (RP-HPLC) has noequal. Unfortunately there is no general purposemethod for biomolecules to parallel the inherentpower of RP-HPLC. The success of RP-HPLC foranalysis can be judged from the large number ofpublished applications developed for the ‘Rrst wave’of protein pharmaceuticals manufactured by geneti-cally engineered microorganisms. These extracellular(relatively) low molecular weight proteins includehuman insulin, human growth hormone and the in-terferons. However, as molecular size and fragilityincrease, so difRculties in using HPLC increase,a primary reason why much analysis is still conductedon low performance systems. Low performance sys-tems are easily scaled up; RP-HPLC is not.

Biological separation systems must be asepticallyclean throughout the process. The mixtures are inevi-tably complex and usually contain many contamina-ting similarly structured species. Such species canadsorb very strongly onto the medium, demandingpost-use washing with very powerful reagents to ster-ilize and simultaneously clean the column. Silica-based matrices cannot survive this type of treatment,hence scale-up of analytical procedures is generallyprecluded. The Rrst wave of commercial proteinpharmaceuticals have generally proved to be relative-ly stable under high stress conditions. On the otherhand intracellular proteins, often of high molecularweight, are unstable. Analysis by high performanceRP-HPLC methods then becomes problematic. De-mand for fast, high resolution, analytical methodswill continue to increase for on-line monitoring andprocess validation. Such techniques have already been

used to determine degradation of the target protein(for example deamidated and oxidized elements); toidentify previously unidentiRed components; to estab-lish the chromatographic identity between recom-binant and natural materials; to develop orthogonalmethods for the identiRcation of unresolved impu-rities; and for many other demanding analyticalapproaches.

Af\nity versus Traditional Media

When projects are transferred from research to devel-opment two sets of chromatographic techniques arecarried forward: analysis, usually based on RP-HPLC; and larger scale serialized separation steps,often incorporating traditional methods of ion ex-change, hydrophobic interaction and gel permeationchromatography. Major decisions have to be taken atthis juncture } to scale up the separation processesdeveloped during the research phase or to investigatealternatives. Regulatory demand and shortened pat-ent lifetimes compel managements to ‘fast track’ newproducts. Commercial pressure is at a maximum.Being Rrst to market has the highest priority in termsof technical and commercial reward. Very little timeis left to explore other separation strategies. It isknown that serial application of IE, HI and GPCinevitably leads to very high manufacturing costs,but which comes Rrst? Most often the decision istaken to begin manufacture using unoptimized separ-ations as deRned in research reports. It is only inretrospect that very high production costs becomeapparent. By then it is too late } regulatory systemsare Rrmly in place.

There is an alternative. If researchers were moreaware of process economics and the consequences ofregulatory demand, selection of superior separationprocesses could then result. Although most resear-chers are fully aware of the advantages of single-stepafRnity methods, paradoxically the high selectiv-ity advantage of afRnity chromatography is also aweakness. Suitable off-the-shelf afRnity adsor-bents are often unavailable, in which case an adsor-bent has to be custom synthesized. Since the majorityof biochemists have no desire (or time) to undertakeelaborate chemical synthesis, antibody-based adsor-bents are commonly used. However, raising suitableantibodies and purifying them before immobilizationonto a preactivated support matrix is an extremelylaborious procedure. In addition, proteins are so of-ten tightly bound to the antibody that subsequentelution involves some degree of denaturation and/orloss of acitivity. Ideal media require the incorporationof elements of both nonselective and selective adsor-bents to provide adsorbents with a general applicabil-


Figure 3 Comparison of multistep versus affinity separation.

ity. If stable, highly selective and inexpensive af-Rnity ligands were available, then opportunities wouldexist for researchers to develop efRcient high yieldseparations even in the earliest phases of investigation.These systems could then be passed forward to pro-duction with the knowledge that optimally efR-cient separations are immediately achievable.

Production costs of any pure material reSect theabsolute purity level required and the difRculty ofachieving it. Therapeutic proteins have high purityrequirements and the larger the administered dose,the purer it has to be. Since many protein pharma-ceuticals will be used at high dose levels, purities needto exceed 99%, occasionally up to 99.999%. Thatthese purities can be met by traditional methods ispossible, but it is widely documented that the applica-tion of such methods massively increases productioncosts. Between 50 and 80% of total production costsof therapeutic proteins are incurred at the puriRcationstage. The manufacturing cost of a product is directlyrelated to its concentration in the mother liquor;the more dilute it is, the higher the cost of recov-ery. Since traditional puriRcation processes on theirown cannot selectively concentrate a target protein tothe exclusion of all others, they have to be used inseries. The number of stages required can vary be-tween four and 15. Each step represents a yield loss,and incurs a processing cost. Yields of less than 20%are not uncommon. Figure 3 shows an enzyme puriR-ed in multiple stages and by a one-step afRnityprocess.

It was these limitations that caused biochemists toexamine highly selective ligands. Almost any com-pound can be used as an afRnity ligand providedit can be chemically bonded onto a support matrixand, once immobilized, it retains its ability to interactwith the protein to be puriRed. The ligand can be

a simple synthesized entity or a high molecular weightprotein. The afRnity technique is theoretically ofuniversal application and any protein can be separ-ated whatever its structure and origin. As always,there are major limitations. The most effectiveafRnity ligands are other proteins. Unfortunatelysuch proteins are difRcult to Rnd, identify, isolateand purify. This results in high costs. An even greaterdeterrent is that most proteins are chemically, cata-lytically and enzymically unstable, a particularly un-attractive feature if they are to be used for the manu-facture of therapeutic substances; and regulatoryauthorities generally reject applications using pro-teinaceous ligands.

In anticipating that one day stable inexpensive af-Rnity media would be in demand, a team led by C.R.Lowe began an investigation into which syntheticligand structures offered the greatest possibilityof developing inexpensive stable ligands. It was con-cluded that structures that could be manipulated intospeciRc spatial geometries and to which intermolecu-lar binding forces could easily be added offeredthe highest chance of success. Model compoundswere already available; the textile dyes.

Synthetic Ligands

Textile dyes had already proved to be suitable ligandsfor protein separations. Blood proteins, dehydrogen-ases, kinases, oxidases, proteases, nucleases, transfer-ases and ligases can be puriRed by a wide variety ofdyes. However, they did not prove to be the break-through so eagerly awaited. An essential feature of allchromatographic processes is exact repeatabilityfrom column to column, year after year. Textile dyesare bulk chemicals, most of which contain manyby-products, co-produced at every stage of the dye


Figure 4 Leakage of blue dye from various commercial products. , 0.1 mol L�1 NaOH; , 0.25 mol L�1 NaOH; , 1 mol L�1

NaOH. Key: A, Mimetic Blue 1 A6XL (affinity chromatography); B, Affi-Gel Blue (Bio-Rad); C, Blue Trisacryl-M (IBF); D, Fractogel TSKAF-Blue (Merck); E, C.I. Reactive Blue 2 polyvinyl alcohol-coated perfluropolymer support; F, Blue Sepharose CL-6B (Pharmacia); G,immobilized Cibacron Blue F3G-A (Pierce); H, Cibacron Blue F3G-A"Si500 (Serva); I, Reactive Blue 2-Sepharose CL-6B (Sigma).

Figure 5 Schematic representtion of ligand}protein interaction.W, electrostatic interaction; X, hydrogen bonding; Y, van derWaals interaction; Z, hydrophobic interaction. ***, originalbackbone; - - -, new structure added; 2, original backbonemove; , fields of interaction.

manufacturing process. This fact alone makes repro-ducibility problematic. Furthermore, the bondingprocess between dye and matrix was poorly re-searched. This resulted in extensive leakage. All com-mercially available textile dye products leak exten-sively, especially under depyrogenating conditions(Figure 4). Despite these limitations, it was recog-nized that dye-like structures had a powerful underly-ing ability to separate a very diverse range of proteins.Their relatively complex chemical structures allowspatial manipulation of their basic skeletons intoan inRnite variety of shapes and conRgurations. Pro-teins are complex three-dimensional (3-D) structuresand folds are present throughout all protein struc-tures. An effective ligand needs to be shaped insuch a manner that it allows deep insertion intoa suitable surface Rssure existing within the 3-Dstructure (Figure 5). In contrast, if the ligand onlyinteracts with groups existing on external surfaces,then nonspeciRc binding results and proteins otherthan the target are also adsorbed. A much moreselective approach is to attempt to insert a ligand intoan appropriate fold of the protein, and add bindinggroups to correspond with those present in a fold ofthe protein. If all four of the basic intermolecularforces (Figure 5: W, electrostatic; X, hydrogen bond-

ing; Y, van der Waals; Z, hydrophobic) align with thebinding areas in the protein fold, idealized afRn-ity reagents result. The use of spacer arms minimizessteric hindrance between the carrying matrix andprotein.


Figure 6 Comparison of ligand leakage from mimetic ligandaffinity adsorbent A6XL ( ) and conventional textile dye agarose( ).

The Rnal step is to design appropriate bondingtechnologies to minimize potential leakage. Until re-cently this type of modelling was a purely theoreticalexercise. It was only the introduction of computer-assisted molecular modelling techniques that allowedthe theory to be tested. Before the arrival of logicalmodellling the discovery of selective ligands was en-tirely based upon empirical observation, later fol-lowed by a combination of observation, experienceand limited assistance from early computer generatedmodels. Although several novel structures evolvedduring this period, a general approach to the design ofnew structures remained elusive. At this time onlyvery few 3-D protein structures were available, againgreatly restricting application of rational design ap-proaches. As more sophisticated programmes, simu-lation techniques, protein fragment data and manymore protein structures were released, logical designmethods were revolutionized. However, many mil-lions of proteins are involved in life processes, and itis clear that many years will elapse before the major-ity of these will be fully described by accurate models.Consequently intuition and experience will continueto play a major role in the design of suitable ligands.Of available rationally designed synthetic molecules,the Mimetic�� range can currently separate over 50%of a randomly selected range of proteins. Stabilityunder depyrogenating conditions has been demon-strated for these products (Figure 6). This results inminimal contamination from ligand and matrix im-

purities, substantial increases in column lifetime, andimprovements in batch-to-batch reproducibility.

Rational Design of Af\nity Ligands

Modi\cation of Existing Structures

The Rrst example of a rational design of new bio-mimetic dyes used the interaction between horseliver alcohol dehydrogenase (ADH) and analogues ofthe textile dye Cibacron Blue F3G-A (Figure 7). Ithad been established that the parent dye binds in theNAD#-binding site of the enzyme, with the an-thraquinone, diaminobenzene sulfonate and triazinerings (rings A, B and C, respectively, in Figure 7)apparently adopting similar positions to those of theadenine adenosine ribose and pyrophosphate groupsof NAD�. The anthraquinone ring (A) binds ina wide apolar fold that constitutes, at one end, theadenine bridging site, while the bridging ring (B) ispositioned such that its sulfonate group interacts withthe guanidinium side chain of Arg271 (Figure 8).Ring C binds close to where the pyrophosphatebridge of the coenzyme binds with the reactivetriazinyl chlorine adjacent to the nicotinamide ribose-binding site. The terminal ring (D) appears to bebound in a fold between the catalytic and coenzymebinding domains, with a possible interaction of thesulfonate with the side chain of Arg369. The bindingof dye to horse liver ADH resembles ADP binding butdiffers signiRcantly at the nicotinamide end ofthe molecule with the mid-point position of ring D dis-placed from the mid-point position of the nicotinamidering of NAD# by about 1 nm. Consequently a numberof terminal-ring analogues of the dye were synthesizedand characterized in an attempt to improve the speciR-city of dye binding to the enzyme. Table 4 lists some ofthe analogues made by substituting }R in the D ring(Figure 7), together with their dissociation constants.These data show that small substituents bind moretightly than bulkier groups, especially if substituted inthe o- or m-positions with a neutral or anionic group.Further inspection of the computer model given asFigure 8 showed that the dye analogues were too shortand rigid to bind to horse liver ADH in an identicalmanner to the natural coenzyme, NAD#. Conse-quently analogues of the parent dye were designed andsynthesized with central spacer functionalities toincrease the length and Sexibility of the molecule(Figure 9). This product proved to be some 10 timessuperior to any previously synthesized compound.This work provided the Rrst proof that rationallydesigned molecules could be converted into stable,inexpensive, chromatographic media, while providingthe most remarkable separations.


Figure 7 Principal structural elements of the anthraquinone dye, Cibacron Blue F3G-A.

Figure 8 Putative binding pocket for the terminal-ring analogue (m-COO�) of Cibacron Blue F3G-A in the coenzyme binding site ofhorse liver alcohol dehydrogenase (ADH). The site lies lateral to the main coenzyme binding site and comprises the side chains of twojuxtaposed cationic residues Arg47 and His51.

De Novo Design

Most early efforts in proving the rational designtechnology was based upon dye structures. To dateall dyes considered have been anionic, presumablybecause the charged chromophores of these ligands

mimic the binding of naturally occurring anionic het-erocycles such as NAD#, NAPD#, ATP, coenzymeA, folate, pyridoxal phosphate, oligonucleotidesand polynucleotides. However, some proteins,particularly proteolytic enzymes, interact with cationicsubstrates. The trypsin-like family of enzymes forms


Table 4 Apparent affinities of terminal-ring analogues of an-thraquinone dyes for horse liver alcohol dehydrogenase (ADH)

R Apparent Kd (�mol L�1)

m-COO� 0.06H 0.2o-COO� 0.2o-SO�3 0.4m-SO�3 1.6m-CH2OH 4.5m-CONH2 5.7p-COO� 5.9p-SO�3 9.3p-PO2�3 10.5p-N#(CH3)3 172.0

one of the largest groups of enzymes requiringcationic substrates and includes enzymes involved indigestion (trypsin); blood clotting (kallikrein, throm-bin, Factor Xa); Rbrinolysis (urokinase, tissue plas-minogen activator) and complement Rxation. Theseenzymes possess similar catalytic mechanisms andbind the side chains of lysine or arginine in a primarypocket proximal to the reactive serine (Ser195), withspeciRcity being determined partly by the side chainof Asp189 lying at the bottom of the pocket, andpartly by the ability of the individual enzymes to formsecondary interactions with the side chains of othernearnearby substrate amino acids. For example, tis-sue kallikrein differs from pancreatic trypsin inthat it displays a marked preference for phenylalaninein the secondary site, probably because the phenylring on the phenylalanine residue neatly slips intoa hydrophobic wedge-shaped pocket between thearomatic side chains of residues Trp215 and Tyr99(Figure 10). SpeciRcity for the secondary amino acidresidue is less stringent in trypsin since Tyr99 is re-placed by Ala99 and the hydrophobic pocket cannotbe formed. By designing a mimic for the Ph}Argdipeptide should result in a speciRcity for kallikrein.Figure 11 uses p-aminobenzamidine and phene-thylamine functions substituted on a monochloro-triazine moiety. However, the active site of pancreatickallikrein lies in a depression in the surface of theenzyme. The expected steric hindrance is eliminated

by insertion of a hexamethylene spacer arm betweenthe designed ligand and the matrix. After synthesis ofthis medium it was demonstrated that puriRed pan-creatic kallikrein was strongly bound, with over 90%of activity being recovered on elution with 4-aminobenzamidine, whereas trypsin appeared largelyin the void of the column. This medium was able topurify kallikrein 110-fold from a crude pancreaticacetone powder in a single step.

There is an alternative to the rational design ap-proach } the use of combinatorial libraries.

Combinatorial Libraries

The driving force behind the development of combi-natorial libraries has been the many failed attemptsto design therapeutic substances using theoreticalknowledge allied to rational design; very few suchapproaches succeeded. In contrast, combinatorial li-brary design is now thought by some to provide thebest opportunity of discovering new novel peptidesand small molecule structures for pharmaceutical ap-plication. A quite natural extension of the conceptis to use combinatorial libraries to discover ligandscapable of achieving highly efRcient protein sep-arations. When directed at drug discovery the earliestworkers built libraries from peptides. For ligandslibraries will generally utilize simple chemical mol-ecules and occasionally smaller peptides. To distin-guish this subsection from the earlier methods aconvenient designation is the term Chemical Combi-natorial Library (CCL)��.

Although procedures for rationally designedligands are well established, the newness of CCLsuggests that CCL design of afRnity ligandsshould be regarded as embryonic rather than immedi-ately available for commercial application. There arethus two diametrically opposed systems } the rationaldesign process, based on logic, experience and know-ledge; and CCL which is illogical and completelyrandom. One description of CCL is ‘a method ofincreasing the size of the haystack in which to Rndyour needle’. A very recent approach is to combineboth rational and CCL techniques, a process termed‘intelligent’ combinational design. At the time ofwriting there are no published examples of ligandsderived from CCL, although patents have been Rledin this area.

Regulations and Drug Master Files

For researchers the relevance of regulations oftenseems remote, and yet the decisions taken in even theearliest stages of research can take on a great signiR-cance if the target product becomes a commercial


Figure 9 Horse liver alcohol dehydrogenase separation. Using the modified Cibacron Blue F3G-A (Figure 7) structure given above,the selectivity is greatly enhanced making it possible to separate the isoenzymes.

reality. It is regrettable that many researchers slavish-ly follow previously published data on a given separ-ation problem without giving thought to the longer-term implications of their decisions. Sections abovedescribe the adverse economic effects of multi-stage processing, but an equally important factor isregulatory issues. The most widely used regulationsare those deRned by the USA’s Food and Drug Ad-ministration (FDA). Any company wishing to importproducts relevant to regulations existing in the USAmust conform exactly to FDA requirements; drugs inparticular are very strictly controlled. Detailed de-scriptions of any plant and process used in drugmanufacture have to be lodged in documented formwith the FDA, wherein every aspect of process de-scription is given. This must include raw material

deRnition and suppliers, stability data for every stepof the process, formulation methods, packaging,labelling, toxicity data and so on. The documentationhas to be revised annually and any changes notiRed.Furthermore plant and process is open to inspectionat all times for full audit of procedure. There is onelarge anomaly within the regulations. The largestvolume of material in contact with a drug duringmanufacture is water, solvents and salts, all of whichare exactly deRned in terms of their physicochemicalcharacteristics. The next largest is chromatographicmedia; ambiguously media do not have to be de-scribed in the same detail.

The outstanding stability of the syntheticchromatography media provides an excellent oppor-tunity to develop and register Drug Master Files


Figure 10 Model of the Phe}Arg dipeptidyl substrate bound in the acitve site of porcine pancreatic kallikrein. The illustration showsAsp189 at the bottom of the primary binding pocket as well as the side chains of Tyr99 and Try215, which form the secondary bindingpocket, with the phenyl ring of the Phe residue sandwiched between the hydrophobic side chains of these residues.

Figure 11 Comparison of the structures of (A) the Phe}Argdipeptide, and (B) the ‘biomimetic’ ligand designed to bind at theactive site of the porcine pancreatic kallikrein.

(DMFs) with the FDA. DMFs allows companies touse synthesized ligands for very high purity proteinpharmaceuticals with total conRdence. New DrugApplications (NDA) and Investigational New Drugs(IND) documents incorporating such stable afRn-ity media can now be submitted to the FDA, safe inthe knowledge that all appropriate information is onRle. The effective guarantee of minimum quality

standards and Good Manufacturing Practice (GMP)is an integral part of a DMF. Few researchers select-ing a speciRc medium consider the long-term implica-tions of stability under depyrogenating conditions,the number of cycles that can be achieved (lifetime inuse), its availability in bulk, whether it is manufac-tured under aseptic conditions and the price whensupplied in bulk. If the researcher makes a goodinitial selection, the research data produced can beutilized in development phases with conRdence. ‘Fasttracking’ is facilitated, with minimum aggravation,maximum efRciency and minimum puriRcationcosts.

Alternative Af\nity Approaches

Aqueous two-phase systems have been extensivelyapplied to bimolecular puriRcations, by attaching af-Rnity ligands to one of a pair of phase-forming poly-mers, a method known as afTnity partitioning.Although a substantial body of research literature isavailable, few systems appear to have been adoptedfor commercial purposes. Reactive dyes, with theirsimple and well-deRned coupling chemistries, havegenerally been favoured as the active ligand. Theadvantage of afRnity partitioning is that the pro-cess is less diffusion-controlled, binding capacities


are high and the recovery of bound proteins iseasier, created by the process operating with fewertheoretical plates than those generated by chromatog-raphy columns. This technique has also been com-bined with afTnity precipitation, where a homo-bifunctional ligand composed of two ligand entitiesconnected by a spacer (for example a bis-dye) is used.However, even in combination this approach suf-fers from considerable nonspeciRc binding and rela-tively low puriRcation factors. A review of this com-bination suggests that it is more suited to large scale,low purity products. In contrast, perSuorocarbonemulsion chemistry utilizing mixer-settlers may of-fer more promise. By using a series of mixer-settlersconnected in a loop a continuous process has beendeveloped. A ligand (usually a reactive dye) iscovalently bonded to a high density perSuorocarbonemulsion and contacted with the crude protein solu-tion. After settling in the Rrst tank the emulsion ispumped to a second settler and washed before passingto the third settler for elution. The emulsion is regen-erated in the fourth settler. The supernatants fromeach settler, still containing some unbound targetprotein, are normally discarded. Although reasonablerecoveries and yields are obtained, signiRcant devel-opment is needed for this type of system to becomecompetitive with conventional chromatography col-umn methods.

Another favoured research approach to improvingefRciency is to use expanded beds. Various tech-niques have been tried, with the primary objective ofeliminating the ‘solid bed’ effect, where the bedacts as a Rlter, trapping insolubles and creating signif-icant back-pressure. By partially removing the normalconstraints of upper and lower retaining frits, whichpack the particles tightly in the bed, the particles canexpand, thereby releasing trapped solid impurities.Consequently longer operational cycles and higherSows result. One limitation of the expanded bedsystem is that adsorption can only be carried out inone stage, resulting in a less efRcient process.

Expanded beds are only an intermediate stage to-wards Uuidized beds. Several variations of Suidizedbed technology have been adopted to evaluate themfor afRnity processing. One example is the use ofperSuorocarbon emulsions in a countercurrent con-tactor. The afRnity perSuorocarbon emulsion isloaded with crude source material into the base ofa column in a similar manner to that of an expandedbed. The adsorbent is then removed from the base ofthe bed and carried forward through four identicalcontactors where washing, elution and regenerationare carried out successively. This process is claimed toimprove signiRcantly removal of target proteins com-pared to an expanded bed system.

Af\nity Membranes

UltraRltration membranes are commonly employedas a ‘polishing’ stage of multistage separation pro-cesses for several commercially important proteins.Consequently attaching standard afRnity ligandsto create afRnity membranes has become an ac-tively researched area. The most obvious advantageof a membrane structure is the high rate of transportof the medium through the porous structure by Rltra-tion, thus minimizing the normally encountered dif-fusion limitations of mass transfer. High adsorptionrates are achieved, especially if long distance electro-static interactions are involved in the binding mech-anism. However, in contrast to ion exchangemembranes, similar high transport effects are notobserved when used in the afRnity mode, elimin-ating much of the initial attraction of this form ofdevice. Other theoretically attractive features in-cluded: an inherent ability to control pore size acrossa very wide range, offering an opportunity toincrease capacity of a given system; and ability tooperate in either batch mode or Rltration mode. Inboth cases experimental data have not conRrmed theseassumed advantages; a 10-fold change in the pore sizeresulted in only a two-fold capacity increase and whenin Rltration mode, although adsorption is fast, severepeak broadening on elution is experienced.

The chemistry relevant to particulate media is iden-tical to that required for membranes, in effectmaking the systems compatible and allowing an easytechnology interchange. The covalent bonding of af-Rnity ligands to the surface of a membrane followsexactly the same chemistry as that applied to partic-ulate media, and the same adsorption/desorptionprinciples apply to both. Consequently the only dif-ference between membrane systems and those of con-ventional chromatography is the exploitation of thecharacteristics of the membrane matrix comparedwith a particulate bed. Although the high mechanicalstrength of membranes is one major advantage, plusthe scale-up is claimed to be very easy by stackingmembranes (although scaling afRnity columns isalso very straightforward), it has been discovered thatif the pressure drop across the membrane is too highsealing problems occur; the mobile phase then Sowsbeyond the edges and past the membranes. Further-more afRnity membranes should be capable ofuse with unclariRed extracts, but is has been generallyobserved that membrane capacity and lifetime areprogressively reduced with time of use. Even withclariRed broths, membrane fouling regularly occurs.This is almost certainly the reason why afRnitymembranes have not found favour in large scaleprocessing.


Conclusion

Protein separations can be achieved by a variety ofafRnity techniques, but separations in thechromatography mode are by far the most widelyused. Nature deRned an appropriate pathway to highlyefRcient separation } utilization of the phenom-enon of the automatic recognition mechanism existingbetween a given protein and at least one other. Bycovalently bonding one of the pair onto an inert matrixa theoretically simple separation process can be de-vised. Although these immunoafRnity separationsare widely practised today, severe limitations exist, notleast of which are cost and instability of the afRn-ity medium when in use. As modern design aids havebecome commonplace, in conjunction with newertechniques such as the development of combinatoriallibrary arrays, it has proved possible to mimic natureand replace immunoafRnity matrices by speciR-cally designed synthetic ligands. These new ligands notonly accurately emulate the exquisite precision of thenatural protein}protein interaction mechanisms, butalso provide the opportunity to manipulate the ligandstructures, thus offering far more efRcientseparations than any previously achieved. For a givenprotein, from whatever source and at any dilution, it isnow possible virtually to guarantee that a highly cost-effective and highly efRcient separation pro-cess can be developed for eventual commercial use.

Designed ligand processes have already been ad-opted for several very large biotechnology projectsscheduled to manufacture bulk protein pharmaceut-

icals. A mandatory part of any new protein pharma-ceutical process is the acceptance by regulatoryauthorities of the separation process involved. Thatsynthesized afRnity ligand separation processeshave now been fully accepted by the foremost regula-tory authority, the USA’s Food and Drug Administra-tion, conRrms a worldwide acceptance of the powerof ligand design technologies.

See Colour Plate 1.

Further Reading

Briefs K-G and Kula M-R (1900) Fast protein chromatogra-phy on analytical and preparative scale using modiRedmicro-porous membranes. Chemical EngineeringScience 47: 141}149.

Burton SJ, Stead CV and Lowe CR (1988) Design andapplications of biomimetic anthraquinone dyes. Journalof Chromatography 455: 201}216.

Chase HA (1994) PuriRcation of proteins using expandedbeds. Trends in Biotechnology 12: 296}305.

Dean PDG, Johnson WS and Middle FA (1985) AfTnityChromatography: A Practical Approach. Oxford: IRLPress.

Jones K (1990) A review of afRnity chromatography.Chromatographia 32: 469}480.

Kenny A and Fowell S (1990) Methods in Molecular Biol-ogy: Practical Protein Chemistry. New York: HumanaPress.

Kopperschlager G (1994) AfRnity extraction with dye-ligands. Methods in Enzymology 228: 121}129.

Walker JM and Gaastra W (1987) Techniques in MolecularBiology. London: Croom Helm.

CENTRIFUGATION

D. N. Taulbee and M. Mercedes Maroto-Valer,University of Kentucky-Center for Applied EnergyResearch, Lexington, KY, USA

Copyright^ 2000 Academic Press

Introduction

Centrifugation is a mechanical process that utilizes anapplied centrifugal force Reld to separate the compo-nents of a mixture according to density and/orparticle size. The principles that govern particle be-haviour during centrifugation are intuitively compre-hensible. This may, in part, explain why centrifu-gation is seldom a part of post-secondary sciencecurricula despite the broad range of scientiRc, medicaland industrial applications in which this technique

has been employed for well over 100 years. Applica-tions that range from the mundane, industrial-scaledewatering of coal Rnes to the provision of an invalu-able tool for biomedical research.

The Rrst scientiRc studies conducted by Knight in1806 reported the differences in orientation ofroots and stems of seedlings when placed in a rotatingwheel. However, it was not until some 60 years laterthat centrifuges were Rrst used in industrial applica-tions. The Rrst continuous centrifuge, designed in1878 by the Swedish inventor De Laval to separatecream from milk, opened the door to a broad range ofindustrial applications. About this same time, the Rrstcentrifuges containing small test tubes appeared.These were modest, hand-operated units that attainedspeeds up to 3000 rpm. The Rrst electrically driven

Sepsci*11*TSK*Venkatachala=BGI / CENTRIFUGATION 17

Documents

Affinity Separation