Insight

The downstream processing of biologic manufacturing is divided into three core stages: capture, purification, and polishing. Among these procedures, endotoxin removal remains one of the most challenging tasks. As a component of the cell wall of Gram-negative bacteria, endotoxin is an inevitable contaminant throughout pharmaceutical production, and it can induce pyrogenic reactions that seriously endanger medication safety.

In the downstream purification of biopharmaceuticals, endotoxins readily form stable complexes with target products and exist in diverse morphological states, posing substantial obstacles to their elimination. A single purification step is often insufficient to reduce endotoxin levels below regulatory limits. Therefore, establishing an efficient, reliable, and cost-effective endotoxin clearance strategy is critically important. This paper systematically reviews and evaluates the clearance mechanisms and application progress of membrane separation, chromatographic separation, and other physicochemical approaches, aiming to provide process development with a systematic technical solution for endotoxin control.

In aqueous solutions, endotoxins exist either as monomers with a molecular weight of 10–30 kDa or aggregate into micelles and vesicles exceeding 1000 kDa with a particle diameter up to 0.1 μm. Leveraging this structural characteristic, ultrafiltration achieves endotoxin separation via size exclusion: large endotoxin aggregates are retained by the membrane, while water, salts, and small-molecule drug substances permeate through into the filtrate.

The endotoxin removal efficiency of ultrafiltration is affected by endotoxin aggregation state, target protein concentration, and detergent addition. Studies have demonstrated that when filtering endotoxin-contaminated protein solutions, ultrafiltration membranes with a molecular weight cutoff (MWCO) of 100 kDa allow partial endotoxin monomers to pass through, with overall removal efficiency ranging from 28.9% to 99.8%. Lower protein concentration and diluted solution systems yield better removal performance, as endotoxins tend to exist as monomers in dilute solutions and more easily penetrate the filter membrane.

Furthermore, the addition of detergents such as Tween 20 promotes the dissociation of endotoxin aggregates into monomers and weakens the interaction between endotoxins and proteins, facilitating endotoxin permeation while retaining and recovering target proteins. For instance, in the endotoxin removal of the small-molecule drug BMS-753493 (molecular weight: 1.57 kDa), both 3 kDa and 10 kDa ultrafiltration membranes can reduce endotoxin levels to below 0.03 EU/mL. The 10 kDa membrane achieves approximately 95% drug recovery, markedly superior to the 45% drug loss observed with the 3 kDa membrane.

Nevertheless, ultrafiltration has inherent application limitations. It is particularly suitable for removing endotoxin aggregates from water, salt solutions, and small-molecule drugs with low endotoxin binding affinity, while exhibiting limited efficiency for target molecules with molecular sizes close to or smaller than endotoxin aggregates.

Membrane adsorption integrates the separation mechanisms of affinity and ion exchange chromatography. By immobilizing ligands or functional media onto membrane substrates, it significantly enhances mass transfer efficiency while reducing process duration and capital investment. Common carrier materials include cellulose, nylon, and polyvinylidene fluoride (PVDF). This technology features high flow rates and minimal diffusion limitations, and single-use application eliminates elution, cleaning, and regeneration procedures, thereby lowering product contamination risks and buffer consumption. However, membrane adsorbers generally exhibit lower binding capacity than traditional chromatographic media, and frequent replacement may increase long-term operational costs.

Early research indicated that nylon membranes immobilized with histidine tags only achieved a 65% endotoxin removal rate at an initial endotoxin concentration of 387 EU/mL, with efficiency declining further as endotoxin loading increased, revealing inherent limitations in binding capacity.

In recent years, the development of novel membrane adsorbers has substantially improved removal performance. Incorporating amphipathic carbonaceous particles into PVDF substrates achieves over 99.8% endotoxin removal and more than 90% protein recovery. Doping PolyBall nanoparticles into cellulose acetate membranes delivers an endotoxin binding capacity of approximately 2.7×10⁶ EU/mg, outperforming the binding capacity of the same nanoparticles in suspension (1.5×10⁶ EU/mg). With progressive breakthroughs in mitigating binding capacity constraints while maintaining high product recovery, membrane adsorption demonstrates promising potential for industrial bioprocessing.

Chromatographic technologies, especially ion exchange and affinity chromatography, serve as core approaches for endotoxin removal in downstream purification workflows.

Anion exchange chromatography separates endotoxins from target molecules based on charge differences under controlled pH conditions. Endotoxins have an isoelectric point (pI) of approximately 2 and remain negatively charged under conventional chromatographic conditions (pH > 2), enabling adsorption onto positively charged stationary phases.

The charge state of target proteins is determined by the solution pH relative to their isoelectric point: proteins carry negative charges when pH > pI, positive charges when pH < pI, and are electrically neutral at pH = pI. For effective separation of positively charged proteins and endotoxins, the system pH is adjusted above the protein pI to render the protein strongly positively charged and repelled by the stationary phase, resulting in earlier elution. Endotoxins are retained via electrostatic interaction with the stationary phase.

Studies show that increasing resin bed volume and extending residence time reduce endotoxin residuals in flow-through fractions, with residual levels controlled at approximately 400 EU/mL and protein recovery generally exceeding 80%. Separation failure or product loss may occur when strong electrostatic attraction exists between proteins and endotoxins, or when target proteins bind excessively to chromatographic resins.

As a typical case, during the purification of Melan-A antigen (pI = 8.7), strong intermolecular interactions caused endotoxins to co-elute with the target protein. Gradually raising the pH to 9.2 effectively weakened such interactions, lowering flow-through endotoxin levels from 1400 EU/mL to 500 EU/mL without compromising protein recovery.

Overall, anion exchange chromatography can reduce high initial endotoxin levels (>1000 EU/mL) by up to five orders of magnitude, and moderate endotoxin loads (<100 EU/mL) by three to four orders of magnitude. Spent resins can be regenerated and restored to original adsorption performance via detergent cleaning protocols. The primary limitation lies in its incompatibility with inherently negatively charged targets, such as plasmid DNA (pDNA) and acidic proteins.

Solvent extraction separates endotoxins from target molecules based on differential partition coefficients in immiscible two-phase systems. Endotoxins preferentially partition into the organic phase, while hydrophilic therapeutic molecules remain in the aqueous phase. For example, 1-octanol extraction achieves 64%–99.9% endotoxin removal from various bacteriophages, yet causes 30%–60% product loss. Residual trace 1-octanol in the aqueous phase may also interfere with subsequent Limulus Amebocyte Lysate (LAL) endotoxin detection, requiring additional post-treatment.

As an optimized alternative, temperature-induced phase separation systems constructed with the nonionic surfactant Triton X-114 have been widely adopted. The system is fully miscible with water at 0 °C and undergoes phase separation above 23 °C, partitioning endotoxins into the detergent-rich phase while retaining target proteins in the aqueous phase. This approach delivers 45%–99.9% endotoxin removal with product recovery typically above 80%. Combined with isothermal extraction using sodium dodecyl sulfate (SDS), this strategy is highly effective for pDNA endotoxin removal, reducing residuals to approximately 16 EU/mL with over 80% pDNA recovery. However, SDS induces protein denaturation, restricting its application to protein-based biotherapeutics.

Despite advantages including simple operation, easy scale-up, and high efficiency under heavy initial contamination loads, solvent extraction is limited by potential product degradation during temperature cycling. Additionally, residual endotoxin levels in the aqueous phase after extraction often exceed formulation specifications, necessitating combination with subsequent polishing steps to meet quality regulatory requirements.

Depth filtration and adsorption achieve efficient endotoxin removal via porous media integrating size sieving and adsorptive capture mechanisms. Beyond physical size exclusion of large endotoxin aggregates, functional additives including quaternary ammonium groups, poly-ε-lysine, and PolyBall nanoparticles are embedded into matrix materials such as cellulose and diatomaceous earth. These additives capture permeating endotoxin monomers and micro-vesicles through electrostatic and hydrophobic interactions.

This synergistic filtration-adsorption mechanism enables effective capture of endotoxins in all aggregation states, making depth filtration particularly suitable for processing high-concentration and large-volume feedstocks to drastically reduce initial endotoxin loads at the upstream stage of downstream workflows. As single-use pretreatment units, depth filters eliminate resin regeneration requirements, significantly reducing cleaning validation risks and process cycle time. Optimized depth filter cartridges achieve over 99% endotoxin removal while maintaining target product recovery above 95%, providing robust support for downstream purification workflows.

Downstream purification is the critical link to ensuring endotoxin compliance of final biopharmaceutical products. The technologies reviewed in this paper, ranging from traditional separation based on molecular size and charge to novel adsorption relying on specific intermolecular affinity, collectively form a multi-dimensional and comprehensive endotoxin clearance system.

Future development will focus not only on breakthroughs in individual technologies but also on rational combination and process optimization tailored to the physicochemical properties of target products. Meanwhile, the development of novel affinity ligands and depth filter media featuring high binding capacity, superior selectivity, and low biotoxicity, as well as the integration of continuous biomanufacturing workflows, will serve as key pathways to enhance product safety and production efficiency. The ultimate goal is to establish a more robust, controllable quality assurance system for pharmaceutical products.