Affinity Chormatography Removes Endotoxins

January 1, 2005
Masayo Sakata, PhD

Yoshihisa Yamaguchi

Chuichi Hirayama, PhD

Ivars Bemberis

Masami Todokoro, PhD

Masashi Kunitake, PhD

Minoru Nakayama

BioPharm International, BioPharm International-01-01-2005, Volume 18, Issue 1

Protein solutions used for research, vaccines, or therapeutics need to be free of contaminants. One of the chief concerns is the presence of endotoxins (lipopolysaccharides) because their removal from protein solutions is a challenge. Typically, removal techniques utilize adsorption onto surfaces of beads in batch reactions, onto beads packed in columns, or onto membrane surfaces.

Protein solutions used for research, vaccines, or therapeutics need to be free of contaminants. One of the chief concerns is the presence of endotoxins (lipopolysaccharides) because their removal from protein solutions is a challenge. Typically, removal techniques utilize adsorption onto surfaces of beads in batch reactions, onto beads packed in columns, or onto membrane surfaces.

When poly(ε-lysine), currently used as a food preservative, is impregnated into cellulose beads, it can be safely used as a ligand for endotoxin removal using affinity chromatography. Poly(ε-lysine) contains the cationic binding sites necessary for endotoxin adsorption. Attaching the ligand to cellulose base beads achieves greater selectivity for endotoxins.


Pyrogens are a class of substances that can raise body temperature. They commonly originate from constituents of gram-negative cell walls but can also leach from some chemicals and materials. Figure 1 illustrates a gram-negative bacterial cell wall and a detail of a lipopolysaccharide.


Gram-negative bacteria shed portions of their cell walls and lipopolysaccarides into their environment, flooding it with endotoxins. Endotoxins are potent stimulants of mammalian immune systems, causing pyrogenic and shock reactions. Their removal in parenterals is mandated by FDA, which has established critical limits of pyrogen load for parenteral products.


Figure 1. Schematics of (a) Gram-Negative Bacterial Cell Wall and (b) Lipopolysaccharide1

Pyrogens in parenterals and medical device products can come from a variety of sources. They often are contributed by the raw materials, process water, excipients, cell culture and fermentation additives, chromatography media, process equipment, and packaging components. Rigorous control of microbiological contamination will effectively minimize levels of endotoxin contamination.

The biological activity of endotoxin is associated with the lipopolysaccharide (LPS), which is composed of a non-polar lipid (lipid A), a core polysaccharide, and heteropolysaccharide (O-antigen). The lipid A component imbues endotoxins with toxicity; immunogenicity is associated with the polysaccharide component.3 LPS is an amphipathic substance4 that possesses both anionic regions (the phosphoric acid groups) and hydrophobic regions (the lipophilic groups).

A small amount of endotoxin, about 0.1 ng per kg of body weight, can cause a pyrogenic reaction. The standard reporting unit for endotoxin data is one endotoxin unit (EU), the equivalent of 0.1 ng. A typical gram-negative bacterium contains 10-15 g of LPS, which means that at least 105 bacterial cells are required to contribute 0.1 ng.5 This is the basis of the endotoxin level requirements listed in Table 1.

Table 1. Endotoxin Limits for Various Products6

In physiological solutions, LPS aggregates form supramolecular assemblies (MW up to 1 x 106), with phosphate groups as the head group, and exhibit a negative net charge because of their phosphate groups. Aggregation of endotoxins is ascribed to the O-antigen end of the molecule, which gives it detergent-like abilities enabling micelle formation. Divalent cations have a role in stabilizing the aggregated structure of LPS while detergents destabilize the structure. Aggregation impacts not only the size of the endotoxins but also their chemical nature.

When E. coli and other gram-negative bacteria are used to produce recombinant proteins, it is particularly important to ensure that bacterial LPSs are removed from the final product. Until we adapted affinity chromatography to this task, no general method was available for the removal of endotoxins, particularly from protein solutions. Early techniques, usually adapted to specific products, can be found in the literature. These include ultrafiltration, solvent extraction, heat sterilization, solid phase adsorption, size exclusion, ion exchange, hydrophobic interaction, and reversed phase chromatography.


Selective adsorption has proven to be the most effective technique for the removal of LPS from final solutions of bioproducts. Chromatography-based endotoxin-removal techniques require strong selectivity to achieve low residual endotoxin levels without affecting protein recovery. Since LPS has both anionic regions and hydrophobic regions, a LPS-selective ligand should have a combination of cationic properties and hydrophobic properties. Figure 2 shows the modes of LPS interaction with the matrix. Equally important, the chromatography matrices must have low non-specific binding of endotoxins and proteins.

Figure 2. Modes of Interaction with the Matrix

There are two chromatographic methods for removal of endotoxins — binding the endotoxins to positively charged surfaces and allowing protein solutions to flow through; or binding the proteins to negatively charged surfaces and allowing endotoxins to flow through. Endotoxins form complexes with proteins and peptides, which can present a serious concern in terms of protein recovery. Mobile-phase conditions must be carefully adjusted to destabilize these complexes. Therefore, considerable effort has been put into developing adsorbents capable of retaining high LPS selectivity under physiological conditions (ionic strength of µ = 0.05-0.2, neutral pH).

For anion exchange chromatography media to effectively bind endotoxins, pH-mediated destabilization of the endotoxin's secondary structure is needed. Strong anion exchangers, principally quaternary amines (Q-chemistries) are effective at pH 8. Recently, a special chromatography application was developed in which ligands are attached to membranes in a flow-through device. These membrane adsorbers are effective when contaminant concentrations are low. Mostly by employing Q-chemistry, highly efficient polishing to remove trace amounts of endotoxins and other contaminants is possible. By configuring charged membranes into stacked-disc layers, useful, disposable, flow-through devices can accelerate final purification steps in biopharmaceutical production.

For proteins that bind on cation exchangers, loading at ~ pH 4 will bind the protein, but the negatively charged LPS will not interact with the matrix and will flow through. While the protein is bound to the matrix, it can be efficiently washed with buffers or detergents to improve endotoxin removal.

Gradient or step elution can be used to elute proteins at low salt concentration (0.3M NaCl). Endotoxin can then be eluted with high salt (0.5M NaCl).7 If proteins are in high salt buffers during the purification process, hydrophobic interaction chromatography may be used for endotoxin removal, but the protein of interest must elute from the matrix before the endotoxin.7 The endotoxin must bind strongly to the matrix to ensure high protein recovery.

Some researchers have tried reversed-phase (RP) chromatography. However, RP matrices tend to quickly load with lipids, which are difficult to clean. A saturated matrix will bleed endotoxin if not correctly cleaned and regenerated.


While ion exchange chromatography has long been employed for endotoxin removal from protein solutions, greater success can be realized with affinity ligands. Researchers at Chisso and Kumamoto University developed a novel endotoxin affinity removal media. Cellufine ETclean has high LPS selectivity under physiological conditions (ionic strength of µ = 0.05 -0.2, neutral pH). It combines porous cellulose beads for the base matrix and an FDA-approved food preservative, poly(ε-lysine), for the affinity ligand. Two matrices were developed. ETclean-S has a small pore structure to exclude protein and other macromolecules (molecular exclusion limit, M


= 2 x 10


). ETclean-L has a large pore structure (M


= 2 x 10


) that permits proteins and large endotoxin aggregates access to the matrix surface as well and the internal pores.

Cellulose beads are an appropriate base matrix for chromatographic endotoxin removal media because cellulose is a clean and environmentally friendly natural polymer. The matrix is mechanically stable and it can be pumped and stirred, which makes it easier to pack in chromatography columns since the resin will not significantly shrink or swell during buffer exchange (typically less than 3%).

Poly(ε-lysine) is a natural substance produced by aerobic fermentation of Streptomyces albulus. It is a long-chain polymer, and research has determined that LPS-adsorbing activity rises with the molecular weight (degree of polymerization) of the ligand. FDA recognizes poly(ε-lysine) as a safe food preservative that inhibits the growth of gram- positive and gram-negative bacteria and yeast. The ligand can be tracked by HPLC-octadecyl-silane methods and stability is excellent — leaching less than 0.4 ppb/mL after 7-day exposure to 0.2 M NaOH.


High ligand density alone is not sufficient for selective endotoxin adsorption. Several base beads were evaluated to determine the influence of pore size on endotoxin removal. Ligand adsorbing capacity increased as the pore size of the matrix increased, as seen in Table 2. Pore size optimization is important because, in aqueous solutions, LPS often forms supramolecular aggregates. Therefore, ETclean matrices were developed with an amino-group content of 0.6-1.4 meq/g and with matrix-pore sizes of 2 x 10


to >2 x 10


as M


of polysaccharides. Very high capacity was obtained using beads with exclusion limits M


~ 2 x 10


. Larger bead porosity allowed the LPS to fully access all ligand sites. In the table we also list K


, the apparent disassociation constant because adsorption capacity depends on both M


and K



Table 2. Effect of Adsorbent Pore Size on Endotoxin Adsorption4


Adjusting mobile phase conditions such as ionic strength and pH controls the selectivity of chromatography resins. The four parts of Figure 3 show the effects of ionic strength on the selective adsorption of endotoxin by various adsorbents. Boxes in figures 3b and 3c are regions of optimal performance.

Figure 3. Effect of Ionic Strength on Selective Adsorption of Endotoxin from a BSA Solution Containing Endotoxin

The non-derivatized cellulose beads (Cellufine-GC15) used for ETclean-S displayed poor endotoxin-adsorbing activity (17-38%) at all ionic strengths, although the beads adsorbed endotoxin without adsorption of bovine serum albumin (BSA) (Figure 3a). ETclean-S selectively adsorbed endotoxin in the solution at a wide ionic strength (µ range 0.05-0.4) and a pH of 7.0, without adsorption of BSA (Figure 3b). ETclean-L showed a high adsorption for both LPS and BSA when ionic strength is low, but its BSA adsorption sharply decreased with increasing ionic strength of the buffer. However, at µ ≥0.4 it selectively adsorbs LPS without adsorbing BSA (Figure 3c). By comparison, polymixin-immobilized Sepharose showed adsorbing activities for both endotoxin and BSA at a low ionic strength of µ = 0.05 to 0.1, and adsorption decreased with an increase in the ionic strength (Figure 3d). The adsorbent therefore could not selectively adsorb LPS from the BSA solution at any ionic strength.

It is essential to reduce endotoxin to 100 pg/mL or less in fluids used for intravenous injections to avoid eliciting pyrogenic reactions in mammals.8 The endotoxin-removing activity of ETclean-S was compared with that ETclean-L, and the results are shown in Table 3. Various protein solutions, which were naturally contaminated with LPS at concentrations from 1,500 to 32,000 pg/mL, were used as samples.

Table 3. Selective Removal of Endotoxin from Protein Solutions Using Cellufine ETclean Matrices4

The data in Table 3 confirm high endotoxin removal rates across a variety of protein solutions. The table also shows that ETclean matrices were highly selective for LPS in solutions with highly differing protein pIs. The large pore matrix consistently reduces endotoxin to <10 pg/mL levels. With either matrix, endotoxin clearance to <100 pg/mL was readily achieved with consistently high protein recoveries. ETclean-S, with a smaller pore size, showed higher protein recovery (99%) with all proteins.

The performance of ETclean is strongly affected by the amino group density and ionic strength of the buffer solution. Adsorption can be optimally controlled via ionic strength. At low amino-group content, LPS adsorption only occurs at very low ionic strength. When the amino content is high (1.4 meq/g), some protein binding is induced. Optimal performance is obtained with amino-group content of 0.6 meq/g. LPS, DNA, and RNA, which are anionic biopolymers with phosphoric acid groups, are adsorbed very well by all the adsorbents. By contrast, the adsorption of protein was more dependent on Mlim of the adsorbent than its amino-group content.9

Test data prove that adsorption of BSA was caused mainly by the entry of BSA into the pores of the adsorbent. This finding indicates that BSA (MW ~6.9 x 104) can penetrate readily into the larger pores of ETclean-L but cannot penetrate into the pores of ETclean-S. On the other hand, endotoxin cannot enter the pores of ETclean-S because with a MW near 1 x 106 (as aggregates) it 's much larger than the Mlim of ETclean-S. Much of the standard endotoxin, however, was well adsorbed even by ETclean-S.


Subtle differences in conformation can enhance chromatographic protein separation. The strong LPS-binding activity of ETclean matrices is due to the simultaneous effects of the cationic properties originating from ligands and from the matrix's hydrophobic or other properties. Since endotoxins have both anionic and hydrophobic regions, selectivity for LPS by these adsorbents is due to both of these simultaneous effects.


Table 3 illustrates that polycation-immobilized cellulose adsorbents bind LPS more strongly than protein. This is because the pKa (disassociation constant) of the phosphate residues of LPS is lower than the pI of protein (pI: 4.6-11.0) and probably because the LPS is adsorbed by multipoint attachment onto the polycation chain of the adsorbent surface.

The high LPS selectivity of the beads with small pore size is due to the size-exclusion effects. By contrast, selectivity of the particles with large pore sizes is due to the decreases in ionic interactions for net-negative charged proteins, which arise when the buffer's ionic strength is adjusted to 0.2 or stronger. Buffer conditions can be adjusted to aid in capturing LPS. Detergents also aid in destabilizing aggregates as will the exclusion of divalent ions.

Column runs were performed to demonstrate the efficacy of the ETclean matrix. Figure 4 demonstrates the post-binding elution of protein and bound endotoxin using a salt gradient. Again, high protein recovery with simultaneously high LPS removal was achieved. For successful column chromatography, it is necessary to clean and regenerate the column for subsequent use. ETclean can be completely regenerated by washing with a 0.2 M NaOH solution, followed by a 2 M NaCl solution. This cleaning protocol allows multiple reuses of these matrices.

Figure 4. Removal of Endotoxin from Lysozyme


Affinity chromatography techniques using Cellufine ETclean matrices can reduce concentrations of endotoxin to 100 pg/mL or lower in drugs and fluids used for intravenous injection at physiological pH and ionic strength. These processes do not affect the recovery of even acidic proteins such as BSA. ETclean beads provide high removal capacities, from 180 to 480 µg/mL (or up to 4.8 x 10


EU/mL). Cellufine ETclean offers the optimal chromatographic affinity to selectively bind endotoxins employing mixed-mode anion exchanger plus hydrophobic-interaction chromatography mechanisms as displayed in Figure 5.

Figure 5. Chromatographic Interactions of the Cellufine ETclean-S and ETclean-L-matrices

To use these matrices in therapeutic manufacturing requires selection of appropriate process parameters. In most cases, the operation should be at physiological conditions employing bind and elute strategies for endotoxin removal. This ensures the highest protein recovery. Selection of pore size will be governed by the need for high capacity versus the need to achieve low residual endotoxin levels. ETclean-S is appropriate for where the goal is very high removal while ETclean-L offers high surface area for high capacity. Ionic strength should be low for ETclean-S and high for ETclean-L.


This work was supported by an Industrial Technology Research Grant Program in 2003 from New Energy and Industrial Technology Development Organization (NEDO) of Japan.


1. Vaara M, Nikaido H. Outer membrane organization. In: Rietschel ET, editor, Handbook of Endotoxins Vol. 1. Amsterdam: Elsevier; 1984. p. 1-45. Reprinted in European Pharmacopoeia, 3rd ed. Strasbourg France: EDQM: 1997.

2. FDA. 21 CFR 211.94(c) and (d). Drug product containers and closures.

Available at

3. Todar K. Todar's Online Textbook of Bacteriology. Available at

4. Sakata M, Todokoro C, Hirayama C. Removal of endotoxin from protein solution using poly(ε-lysine)-immobilized cellulose beads. American Biotechol. Lab. 2002; 20:36.

5. Shands JW Jr, Chun PW. The dispersion of gram-negative lipopolysaccharide by deoxycholate. Subunit molecular weight. J. Biol. Chem. 1980; 255(3):1221-1226.

6. FDA. CDER Guideline on validation of the limulus amebocyte lysate test as an end-product endotoxin test for human and animal parenteral drugs, biological products, and medical devices. Available at

7. Carter D. Endotoxin removal from proteins. Available at

8. Matsumae H, Minobe S, Kindan K, Watanabe T, Tosa T. Specific removal of endotoxin from protein solutions by immobilized histidine. Biotechnol. Appl. Biochem. 1990; 12:129-140.

9. Sakata M, Hirayama C. HPLC and other chromatographic topics. Biopolymer separations by chromatographic techniques. Encyclopedia of Chromatography Online. Available at


1. Li L, Luo RG. Protein concentration effect on protein-lipopolysaccharide (LPS) binding and endotoxin removal.

Biotechnol. Lett

. 1997; 19:135-138.

2. Petsch D, Anspach FB. Endotoxin removal from protein solutions. J. Biotechnol. 2000; 79:97-119.

3. Minobe S, Watanabe T, Sato T, Tosa T, Chibata I. Preparation of adsorbents for pyrogen adsorption. J. Chromatogr. 1982; 248:401-408.

4. Issekutz AC. Removal of gram-negative endotoxin from solutions by affinity chromatography. J. Immunol. Methods 1983; 61:275-281.

5. Anspach FB, Kilbeck O. Removal of endotoxins by affinity sorbents. J. Chromatogr. A 1995; 711:81-92.

6. Morimoto S, Sakata M, Iwata T, Esaki A, Hirayama C. Preparations and applications of polyethyleneimine-immobilized cellulose fibers for endotoxin removal. Polymer J. 1995; 27: 831-839.

7. Todokoro M, Sakata M, Matama S, Kunitake M, Ohkuma K, Hirayama C. Pore-size controlled and poly(ε-lysine)-immobilized cellulose spherical particles for removal of lipopolysaccharides. J. Liq. Chrom. & Rel. Technol. 2002; 25(4):601-614.

8. Shima S, Sakaki H. Poly-L-lysine produced by streptomyces. Part III chemical studies. Agric. Biol. Chem. 1981; 45(11):2503-2508.

9. Hirayama C, Ihara H, Nagaoka S, Furusawa H, Tsuruta S. Regulation of pore-size distribution of poly(γ-methyl L-glutamate) spheres as a gel permeation chromatography packings. Polymer. J. 1990; 22(7):614-619.

10. Hirayama C, Sakata M, Ihara H, Ohkuma K, Iwatsuki M. Effect of pore size of an aminated poly(γ-methyl L-glutamate) adsorbent on the selective removal of endotoxin. Anal. Sci. 1992; 8:805-810.

11. Hou KC, Zaniewski R, Depyrogenation by endotoxin removal with positively charged depth filter cartridge. J. Parenter. Sci. Technol. 1990; 44:204-209.

12. Hou KC, Zaniewski R. The effect of hydrophobic interaction on endotoxin adsorption by polymeric affinity matrix. Biochem. Biophys. Acta. 1991; 1073:149-154.

13. Gagnon P. Why anion exchange chromatography works so well for endotoxin removal ... sometimes. Validated Biosystems Quarterly Resources Guide for Downstream Processing. 1998. Available at

14. Sakata, M, Todokoro M, Kai T, Kunitake M, Hirayama C. Effect of cationic polymer adsorbent pKa on the selective removal of endotoxin from an albumin solution. Chromatographia 2001; 53:619-623.

15. Boratynski J, Syper D, Weber-Dabrowska B, Lusiak-Szelachowska G, Pozniak G, Gorski A. Preparation of endotoxin-free bacteriophages. Cell Mol. Biol. Lett. 2004; 9(2):253-9.

Ivars Bemberis is director of cellufine chromatography at Chisso America Inc., 6100 Fox Haven Court, Midlothian VA 23112, 804.683.7867, fax 804.739.7422,

Masayo Sakata Ph.D. is assistant professor of applied chemistry and biochemistry in the faculty of engineering at Kumamoto University

Chuichi Hirayama Ph.D. is professor of applied chemistry and biochemistry in the faculty of engineering at Kumamoto University

Masashi Kunitake Ph.D. Kunitake is associate professor of applied chemistry and biochemistry in the faculty of engineering at Kumamoto University.

Yoshihisa Yamaguchi is a researcher at Kumamoto Technology and Industry Foundation

Minoru Nakayama is a graduate student at the School of Science and Technology of Kumamoto University

Masami Todokoro Ph.D. is assistant manager of the strategic business development office of the healthcare products development team at Chisso