Affinity Chromatography: the Fine Print by Pete Gagnon, Validated Biosystems

Affinity Chromatography: The Fine Print by Pete Gagnon, Validated Biosystems

Affinity chr o m a t o g ra p h y is often held up effector functions. Antibodies are a good as the ideal in chr o m a t o g ra p h y. Exquisite example. The non antigen-binding parts of sp e c i f i c i t y . Unmatched simplicity. Load it, the molecule are richly endowed with recep- wash it, elute the purified product. Wh a t tors that interact with a variety of proteins, could be better? Affinity chr o m a t o g ra p h y ca r b o hy d r ates, and cells in the immune sys- could be better for one thing. There is a ver y tem. Conformational modification of these high price tag for both the specificity and the receptors can alter their functionality. All of si m p l i c i t y , and this isn’t referring to the dollar these factors — elevated tendency towar d cost for the media itself. Affinity chr o m a t o g - aggregation, proteolysis, and alteration of rap h y invol v es complications that have effector functions — can alter product effec- immensely important ramifications in purifi- tiv i t y , safety, and pharmacokinetics — and cation process development. The only thing this is before even taking into considerat i o n more costly than the complications them- that the elution conditions may have altered se l v es, is overlooking them. a protein’s primary therapeutic function. Pr oduct denatur at i o n . One of the most One of the ways to deal with product serious and persistent concerns about affinity al t e r ation is to develop milder elution condi- chr o m a t o g ra p h y is the potential for product tions. The first objection to this suggestion is de n a t u r ation. This is especially the case with usually that “If I use a less extreme pH, then strong affinity ligands that require harsh con- the elution will be less effective. ” In fact, elu- ditions for elution. The elution method of tion pH can often be raised without any loss choice is usually exposure to low pH, typical- of effectivi t y , and it’s certainly worth eval u a t - ly in the range of pH 2.5-3.0. Detailed studies ing a given application to find out what the of protein conformation under these condi- real limits are. If modifying pH alone doesn’t tions have documented permanent conforma- bear fruit, there are other opportunities. tional changes as a result of such exposure. Biological affinity interactions are mediated Hydrophobic residues normally protected in by complex combinations of hyd r o p h o b i c the interiors of the primary structural domains in t e r actions, charge interactions, hyd r o g e n become exposed on the surface, increasing bonding, and other mechanisms. Instead of the tendency for the protein to become using just low pH to twist a protein out of in vol v ed in nonspecific hydrophobic interac - shape so severely that it can’t remain bound tions. Elevated tendency toward aggregation to an affinity ligand, it is almost always possi- is one of the results, as is easily proven by ble to target one or more of the actual bind- examining size exclusion chr o m a t o g ra p h y ing mechanisms. This alone usually won ’ t be profiles following affinity purification. sufficient to cause elution, but it almost A usually less obvious but still common al w ays ameliorates the severity of the pH side effect of such changes is an elevat e d conditions required to elute the protein. For te n d e n c y toward proteolysis of the product. example, hydrophobic interactions can be This tends not to be noticed unless it’s weakened with up to 50% ethylene glycol. se v ere, but it’s common nevertheless. The Et h ylene glycol is actually stabilizing to most most insidious side effect — and the most proteins up to this level, but at the same time overlooked — is modification of secondary it is a very effective polarity reducer. It’s also 2 nonionic and won ’ t interfere with down - studied affinity interactions consistently stream charge-based purification methods. re v eal that when binding occurs, it is accom- Hydrogen bonds can be suspended by the panied by a phenomenon called induced fit. inclusion of 1.0M urea. At this concentrat i o n Induced fit refers to a situation where after it ’ s conformational effects on a protein are coming into contact with one another, either nil, and like ethylene glycol, it is nonionic. the affinity ligand, its receptor, or both, 0.5-1.0M NaCl will help to suspend any undergo conformational changes that lock charge interactions, without significant them into place. One of the best prac t i c a l enhancement of hydrophobic interac t i o n s . indicators that induced fit is occurring is The alternative use of a chaotropic salt like when you can bind a product to an affinity sodium perchl o r ate at the same concentrat i o n ligand under mild conditions, but very harsh will likewise suspend charge interactions and conditions are required to remove it. One of ma y further weaken the interaction through the best cha r acterized examples of this is the its chaotropic properties. binding of protein A to IgG, whi c h binds in Using combinations of these mecha n i s m s the hydrophobic cleft between the Cg2 and often makes it possible to raise elution pH by Cg3 domains. X-ray crystallographic data a full pH unit; sometimes substantially more. sh o ws that the Cg3 domain is unaffected by The question invariably arises: Don’t these the contact, but the adjacent Cg2 domain is complex formulations exert the same denatu- displaced longitudinally toward the protein A rat i ve effects as pH alone? They don’t. Thi n k and Cg3. Besides altering the local confor- of it like this. Let’s say that you have a partic- mation, this destabilizes the receptor-r i c h dis- ular oak board that you would like to remove tal third of the Cg2 domain, whi c h in turn from a wall so that you can use it for some- causes a partial rotation and destabilization thing else. It is screwed to the wall, nailed to of the carbohyd r ate domains between the the wall, and glued to the wall. If you just try Cg2 domains. The effect is apparently perma- to rip it off with a crow- b a r , your chances of nent. No matter how careful you are to devel - re c o vering it intact are grim. But if you op gentle elution conditions, comparison of re m o ve the screws and nails, then weaken the purified product against a non-affinity the glue by pouring hot water around the purified control virtually always confirms edges, you can remove it with much less me a s u r able changes in key behavi o r al fea- fo r ce, and your chances of recovering it tures of the product. This is not to say that intact are improved proportionately. This rai s - de v eloping mild elution conditions is not es the second objection to this approach: it wor t h w hile. Doing so can make an immense cancels out part of the simplicity of affinity difference. But it may not be able able to chr o m a t o g ra p h y. Library research will often avert denaturation problems entirely. lighten the load by revealing the dominant Le a ch i n g . The second major concern with me c hanisms of an affinity interaction, but affinity chr o m a t o g ra p h y is leaching of bioac- there will still need to be some experimenta- ti ve ligand and contaminants that may be tion to develop the most effective while least associated with it. This issue gets under- denaturing elution buffer. pl a yed but it is very serious. The inevitability In spite of taking great pains to develop a of leaching is the reason the FDA insists that nondenaturing elution buffer, you may still an y biological affinity ligand used in the find that your product has an elevated ten- manufacture of a biological product meet the de n c y to aggregate, or that it exhibits elevat - same application requirements as the end ed vulnerability to proteolysis, or that it product itself. This extends even to how the exhibits modified primary or secondary func- affinity ligand is purified. For example, the tions. The ugly truth is — to varying degrees protein A going into many chr o m a t o g ra p h y — that this is largely inevitable. The best products is purified by affinity chr o m a t o g r a- 3 ph y on immobilized human polyclonal IgG. se c o n d a r y , like complement fixation. Thi s The IgG column is potentially contaminated le a ves you with a high probability that with virus, whi c h can potentially leach into le a c hed affinity ligand will interfere with the purified protein A, and from there into either the direct function of your product, or your final product. Even to the extent that with other functions that may modulate effi- virus contamination is adequately controlled, ca c y, alter biological clearance, or affect there is still the issue that protein A is derived other aspects of pharmacokinetic behavi o r . from pathogenic bacteria, and may contain Ex p e c t e d l y , chr o m a t o g ra p h y media manu- dangerous contaminants. Some chr o m a t o g r a- facturers go to considerable lengths to mini- ph y media manufacturers have ackn ow l - mize leaching. Some even claim to have edged these issues by obtaining their protein de v eloped immobilization chemistries that A exclusively from recombinant lines of non- eliminate it. This is naive howe ve r . pathogenic bacteria, and purifying it by non- Br e a k d o wn of the chr o m a t o g ra p h y matrix or biological methods. If you are making you r immobilization linkage has been shown to be own monoclonal antibody columns for purifi- a very minor source of bioaffinity ligand cation of a biological, pathogenicity of the le a c hing. It arises mostly from enzymatic cell line in whi c h the monoclonal is grown is de g r adation. With immobilized antibodies, not an concern, but you still have to deal proteolytic clipping of Fab or F(ab)’2 frag - with the virus issue, and with the same ments is common, the enzymes coming from media components that are scrutinized for the column feedstream, or copurified with your end product — BSA, host cell proteins, the original antibody preparation. Wit h DN A, endotoxin, etc. immobilized protein A, heavy proteolytic Unfortunately the issues discussed above de g r adation is widespread, and virtually represent the simplest aspect of leaching. The uncontrollable. This is assured by the ligand biological ligand itself is likely to have more structure. Protein A consists of 5 compact, significant effects on product safety and effi- cylindrical, protease-resistant IgG-binding ca c y than any of the contaminants that may domains, joined together by notoriously pro- be co-immobilized with it. In the case of pro- tease-labile sequences of random coil. Other tein A for example, more than 150 publica- ligands exhibit their own unique breakdown tions describe its interference with virtually patterns, but they all break down, and they ev ery immunological mechanism known . all leach. This shouldn’t be surprising. In fact it wou l d This makes removal of leachate essential. be more surprising if it were not the case; it The obstacle here is that the leached ligand is has known major effects on the functions of bound to your product. The same affinity that IgG — a molecule that can be reasonably al l o wed the column to capture your product regarded as the hub of natural immunity. It in i t i a l l y , ensures that leachate will be reasso- should be equally obvious that parallel con- ciated with your product as soon as buffer cerns apply to any affinity ligand. The who l e conditions permit. If the molecular cha ra c t e r - basis of affinity chr o m a t o g ra p h y rests on its istics of the leachate are substantially different ability to discriminate the unique features of a from those of your product, it may be possible gi ven molecule, from molecules lacki n g to discriminate product-leachate complexes those particular features. The problem arises from uncomplexed product. For example, if with the fact that any feature that makes a your product binds poorly to an anion particular molecule unique, is probably ex c hanger but the leached ligand binds ver y going to be linked directly to some aspect of st r o n g l y , complexes may exhibit intermediate its function. Using antigens and antibodies as be h a vior that allows them to be selectivel y an example, the linked functions may be pri- re m o ved. You’ll have to sacrifice some of you r ma r y , like antigen binding, or they may be product in the bargain (wha t e v er portion is 4 complexed with leachate) but this approach seem on its face. If there is leached ligand in sometimes works. More often, the behavior of the system, it is bound to your product. The the complexes is sufficiently different to sup- association may sterically occlude antigenic port a reduction of leachate, but not quantita- sites that are necessary to obtain accurat e ti ve removal. For example protein A is more measurement. This is not just a theoretical hydrophobic than IgG, but only slightly. The po s s i b i l i t y . It’s a proven problem. To obtain elution behavior of product-leachate com- ac c u r ate measurement, you need to have a plexes from a HIC column is so similar to similarly configured standard curve. Protein uncomplexed product that a 50% reduction of A affinity for IgG varies with species, class, le a c hed protein A may require sacrificing of and subclass of the antibody. The stronger the 50% of the antibody. affinity of the protein A for the particular sub- A usually more successful approach is to class, the shallower the curve. This means, use a separation method carried out under that you have to use non-protein A purified conditions in whi c h the leachate is dissociat- an t i b o d y of the same subclass as your prod- ed from the product. For example, if you uct to make your standard curve. The accura- ha ve small leachate fragments bound to a cy of the curve is also affected by conforma- fairly large product molecule, you may be tion of the protein A: single binding domains, able to separate them by size exclusion chr o - doublets, triplets...whole molecule. This is an ma t o g ra p h y under buffer conditions similar especially troublesome point because, to be to those used to elute the original affinity ac c u r ate, the conformation of the protein A in column. The chief risk with this approach is the standard curve should be the same as in that the prolonged exposure of the product the sample. Unfortunately there’s no direct to these conditions may be prohibitivel y way to know what the conformation is in the denaturing. Cation exchange has proven to sample. There are other complications as be a good alternative. Most proteins will bind well. From a practical perspective, the only to cation exchangers if the pH is low way to obtain unassailably accurate meas- enough, for example a pH known to dissoci- urements of leached protein A, is to separat e ate the leachate from the product. As long as it entirely from your product, and measure the leachate has elution cha r acteristics suffi- the amount of recovered protein A versus a ciently different from the product, you may be standard curve of purified protein A. Similar able to fractionate them from one another. difficulties apply to other affinity ligands. The reason that this scenario is usually better Co s t . Some of the other negative features than size exclusion is that when proteins are of affinity chr o m a t o g ra p h y are better recog- immobilized on cation exchangers, they are nized. The most obvious is cost. Commerci a l sterically restricted and thereby resistant to bioaffinity ligands generally range from 7-15 the conformational changes they wou l d times the cost of ion exchangers on the same undergo in the same conditions in free solu- base matrix. If you are making your own tion. Many proteins that denature rapidly in affinity media, the direct cost may be lo w pH solutions remain stable for hours reduced, but by the time you include you r when bound to a cation exchanger at the de v elopment costs, it will probably end up same pH; ample time for removing the being more expensive than commerci a l le a ch a t e . media. This cost ripples through your overa l l If for some reason you can’t quantitativel y purification costs. Documented comparisons re m o ve leached ligand, you still need to sh o w a 1-step affinity purification to be 9 quantify it and validate that the maximum times more expensive per unit of product le v els occurring in your product have no than a 2-step ion exchange process. In the ad verse affect on product safety or efficacy. case of products intended for in vivo applica- Ev en this is more complicated than it wou l d tions, the cost will be compounded by the 5 need for additional steps to remove leacha t e Do all these factors mean that you should and other trace contaminants. At a recent avoid the use of affinity? No. But they do symposium, a representative of a highly com- mean that you shouldn’t accept it’s superficial petent pharmaceutical company stated that simplicity at face value. Knowing the poten- they were able to achi e v e better than 99% tial effects on product safety and efficacy, purity with their initial bioaffinity purification kn o wing that it will require method devel o p - step...then added that three additional chr o - ment — both to control denaturation and to ma t o g ra p h y steps were required to meet the re m o ve leachates, and considering the elevat - full scope of regulatory requirements. ed validation costs because of those factors, Reduced operating life is another limita- ask yourself this question: Does the inclusion tion with biological affinity media, and of an affinity step in your process offer frac - another elevated cost factor. Most cannot tionation performance so much better than withstand the harsh conditions used for nonaffinity alternatives that it can justify the cleaning and sanitizing nonbiological media, liabilities? If the answer is yes, then it’s a can- or if they can, they can do so for a much di d a t e . lesser number of process cycles. Many are also subject to biological degradation from Se ve r al parts of this article are adapted the feedstream. Protein A, as mentioned from the book Purification Tools for ab o ve, is highly labile to proteolysis, and Monoclonal Antibodies (ISBN 0-9653515-9- especially under the neutral to slightly alka- 9). Most of the key technical points are sup- line conditions at whi c h feedstreams are nor- ported by citations in Chapter 9. mally applied. As significant as initial pro- curement and replacement costs are, they This article was downloaded from Val i d a t e d are trivial in comparison with the elevat e d Biosystems Quarterly Resource Guide validation costs that accompany the issues of for Downstream Processing product denaturation and ligand leachi n g . ht t p : / / w w w. va l i d a t e d . c o m Be y ond validation, especially if you are in a co m p e t i t i ve market, reduced product per- ©1999, Validated Biosystems, Inc. formance may be the most costly factor of all.