Fouling of an anion exchange chromatography operation in a monoclonal antibody process: Visualization and kinetic studies

Fouling of chromatographic resins over their operational lifetimes can be a significant problem for commercial bioseparations. In this article, scanning electron microscopy (SEM), batch uptake experiments, confocal laser scanning microscopy (CLSM) and small‐scale column studies were applied to characterize a case study where fouling had been observed during process development. The fouling was found to occur on an anion exchange (AEX) polishing step following a protein A affinity capture step in a process for the purification of a monoclonal antibody. Fouled resin samples analyzed by SEM and batch uptake experiments indicated that after successive batch cycles, significant blockage of the pores at the resin surface occurred, thereby decreasing the protein uptake rate. Further studies were performed using CLSM to allow temporal and spatial measurements of protein adsorption within the resin, for clean, partially fouled and extensively fouled resin samples. These samples were packed within a miniaturized flowcell and challenged with fluorescently labeled albumin that enabled in situ measurements. The results indicated that the foulant has a significant impact on the kinetics of adsorption, severely decreasing the protein uptake rate, but only results in a minimal decrease in saturation capacity. The impact of the foulant on the kinetics of adsorption was further investigated by loading BSA onto fouled resin over an extended range of flow rates. By decreasing the flow rate during BSA loading, the capacity of the resin was recovered. These data support the hypothesis that the foulant is located on the particle surface, only penetrating the particle to a limited degree. The increased understanding into the nature of the fouling can help in the continued process development of this industrial example. Biotechnol. Bioeng. 2013; 110:2425–2435. © 2013 Wiley Periodicals, Inc.     

Characterization of product capture resin during microbial cultivations

Various bioactive small molecules produced by microbial cultivation are degraded in the culture broth or may repress the formation of additional product. The inclusion of hydrophobic adsorber resin beads to capture these products in situ and remove them from the culture broth can reduce or prevent this degradation and repression. These product capture beads are often subjected to a dynamic and stressful microenvironment for a long cultivation time, affecting their physical structure and performance. Impact and collision forces can result in the fracturing of these beads into smaller pieces, which are difficult to recover at the end of a cultivation run. Various contaminating compounds may also bind in a non-specific manner to these beads, reducing the binding capacity of the resin for the product of interest (fouling). This study characterizes resin bead binding capacity (to monitor bead fouling), and resin bead volume distributions (to monitor bead fracture) for an XAD-16 adsorber resin used to capture epothilone produced during myxobacterial cultivations. Resin fouling was found to reduce the product binding capacity of the adsorber resin by 25–50%. Additionally, the degree of resin bead fracture was found to be dependent on the cultivation length and the impeller rotation rate. Microbial cultivations and harvesting processes should be designed in such a way to minimize bead fragmentation and fouling during cultivation to maximize the amount of resin and associated product harvested at the end of a run.     

A tool for increasing the lifetime of chromatography resins

There is a steadily increasing demand for speed, cost efficiency and process understanding within biopharmaceutical process development. To match this, a high-throughput method for screening of cleaning-in-place (CIP) conditions for chromatography resins has been developed. The methodology includes fouling of MabSelect SuRe chromatography resin in 96-well filter plates, cleaning of the fouled resin by incubation in different CIP agents, and finally, analysis of the residual impurities on the resin after cleaning. This article describes the improvements that transformed the method from low throughput and significant manual interference to a totally automated method with high throughput and good reproducibility.Key words: bioprocess, cleaning-in-place, chromatography, high-throughput, monoclonal antibody, process development, protein A, screening     

Membrane fouling by cell-protein mixtures: in situ characterisation using multi-photon microscopy

Fouling of the membrane by cell and protein mixtures can result in severe flux declines, leading to the eventual need to clean or replace the membrane. In this study multi-photon microscopy, a fluorescence-based technique is used to 3-D image in situ the fouling of microfiltration membranes by suspensions containing combinations of washed yeast, bovine serum albumin (BSA) and ovalbumin. Appropriate fluorescent labelling allows the three foulant species to be clearly identified. Images correlate well with filtration data and clearly show the cake of yeast cells capturing protein aggregates. The proteins exhibited very different filtration behaviour. When filtering washed yeast together with ovalbumin and/or a 50:50 mixture by mass of BSA and ovalbumin, the ovalbumin fouling dominates the system. Capture of aggregates by the cake did not reduce fouling of the membrane by the protein and increased the resistance of the cake. For mixtures of BSA and washed yeast, the presence of a cake of yeast cells did reduce fouling of the membrane by the protein, however, the extra resistance due to the cake resulted in a flux lower than that when filtering BSA alone.     

Bioprocess considerations for expanded-bed chromatography of crude canola extract: sample preparation and adsorbent reuse

Compared to the conventional microbial and mammalian systems, transgenic plants produce proteins in a different matrix. This provides opportunities and challenges for downstream processing. In the context of the plant host Brassica napus (canola), this work addresses the bioprocessing challenges of solid fractionation, resin fouling by native plant components (e.g., oil, phenolics, etc.), hydrodynamic stability, and resin reuse for expanded bed adsorption for product capture. Plant tissue processing and subsequent protein extraction typically result in an extract with a high content of solids containing a wide particle-size distribution. Without removal of larger particles, the column inlet distributor plugged. The larger particles (> 50 microm) were easily removed through centrifugal settling comparable to that attainable with a scroll decanter. The remaining solids did not affect the column performance. Less than 4% of the lipids and phenolics in the fed extract bound to STREAMLINE trade mark DEAE resin, and this small proportion could be satisfactorily removed using recommended clean-in-place (CIP) procedures. Hydrodynamic expansion and adsorption kinetics of the STREAMLINE trade mark DEAE resin were maintained throughout 10 cycles of reuse, as was the structural integrity of the resin beads. No significant accumulation of N-rich (e.g., proteins) and C/O-rich components (e.g., oil and phenolics) occurred over the same period.     

Optimisation of glass surfaces for optical immunosensors

The surfaces of glass sensor chips were modified with dextran to generate a layer protecting the sensor surface from unspecific protein binding and also serving as a matrix for covalent protein immobilisation. Dextran was coupled to the glass surface in different concentrations either covalently on amino-functionalised glass chips or via biotin-avidin binding. Unspecific binding of BSA was monitored with the grating coupler system, and was increasingly suppressed with increasing dextran concentrations. Using a solution with 100 mg/ml carboxymethylated dextran decreased the signals to approximately 2% of those obtained at an untreated glass chip. Antibodies were successfully immobilised in the dextran and binding to the corresponding Cy5-labelled antigen was repeatedly monitored using a fluorescence sensor system (total internal reflection fluorescence (TIRF)).     

Enhancing protein capacity of rigid macroporous polymeric adsorbent.

A macroporous poly(glycidyl methacrylate-triallyl isocyanurate-divinylbenzene) resin was synthesized and modified with diethylamine to yield an anion-exchange resin suitable for protein adsorption. Efforts were made to enhance protein ion exchange capacity of the resin by investigating the copolymer composition. Different synthesis recipes were attempted, and the resultant resins were characterized by measuring the specific surface area and the adsorption ability using bovine serum albumin (BSA) as a model protein. The intraparticle pore size distribution measured by mercury porosimetry showed that the pores in the range of 40-120 nm took 75% of the total pore volume, indicating that the ion exchanger was favorable for protein adsorption. BSA capacity obtained with an appropriate recipe was as high as 78.6 mg/g wet resin or 50 mg/mL packed volume, which was higher than the capacities of some commercially available ion exchangers. Moreover, by using a pore diffusion model, the effective pore diffusivity of BSA was found to be 5.5 x 10(-12) m(2)/s, similar to those in the commercial ion exchangers.     

Effect of fouling on the capacity and breakthrough characteristics of a packed bed ion exchange chromatography column

This study examined the impact of fouling with yeast homogenate on capacity and breakthrough performance of an ion exchange packed bed column. Column performance was assessed by analysis of breakthrough curves obtained with BSA as a test protein. The overall impact of fouling on breakthrough performance depended heavily on the level of clarification of the feed stream. Challenging the column with particulate-free homogenate caused no change in column performance. Loading successive small volumes of poorly clarified homogenate, interspersed with frequent column salt washes, did not alter significantly the column capacity. By contrast, when the column was challenged with an equivalent cumulative volume of poorly clarified homogenate, dynamic binding capacity decreased significantly and changes in breakthrough curves suggested increased intraparticle and external mass transfer limitations. These changes were ascribed to deposition of solid particulates in void spaces in the bed and colloidal contaminants in the bead pores.     

A mechanistic study of in situ chemical cleaning-in-place for a PTFE flat sheet membrane: fouling mitigation and membrane characterization

This study aimed at unfolding the role and mechanisms of chemically enhanced cleaning-in-place (CIP) regimes in fouling control of polytetrafluoroethylene (PTFE) made flat sheet (FS) membrane bio-reactors (MBRs). The trans-membrane pressure (TMP) was successfully maintained below 10 kPa using a daily CIP regime consisting of 100 to 600 mg l^?1 of NaOCl and cake layer resistance control was shown to be critical for effective high-flux MBR operation. In contrast, in the control unit without the CIP, the TMP exceeded 35 kPa at a flux of 40 LMH. The extracellular polymeric substances associated with proteins (EPS_protein) were also controlled effectively with a daily application of the CIP to the fouled membrane. Moreover, the CIP prompted a thinner and looser bio-cake layer on the membrane surface, suggesting that in situ CIP can be a favorable method to control FS membrane fouling at high-flux MBR operation.     

Characterization of intermolecular interaction between cyanidin‐3‐glucoside and bovine serum albumin: Spectroscopic and molecular docking methods

The intermolecular interaction between cyanidin‐3‐glucoside (Cy‐3‐G) and bovine serum albumin (BSA) was investigated using fluorescence, circular dichroism and molecular docking methods. The experimental results revealed that the fluorescence quenching of BSA at 338 nm by Cy‐3‐G resulted from the formation of Cy‐3‐G–BSA complex. The number of binding sites (n) for Cy‐3‐G binding on BSA was approximately equal to 1. The experimental and molecular docking results revealed that after binding Cy‐3‐G to BSA, Cy‐3‐G is closer to the Tyr residue than the Trp residue, the secondary structure of BSA almost not change, the binding process of Cy‐3‐G with BSA is spontaneous, and Cy‐3‐G can be inserted into the hydrophobic cavity of BSA (site II′) in the binding process of Cy‐3‐G with BSA. Moreover, based on the sign and magnitude of the enthalpy and entropy changes (ΔH⁰ = – 29.64 kcal/mol and ΔS⁰ = – 69.51 cal/mol K) and the molecular docking results, it can be suggested that the main interaction forces of Cy‐3‐G with BSA are Van der Waals and hydrogen bonding interactions. Copyright © 2013 John Wiley & Sons, Ltd.     

The Bead blot: a method for identifying ligand-protein and protein-protein interactions using combinatorial libraries of peptide ligands

Small molecules that bind proteins can be used as ligands for protein purification and for investigating protein-protein and protein-drug interactions. Unfortunately, many methods used to identify new ligands to desired proteins suffer from common shortcomings, including the requirement that the target protein be purified and/or the requirement that the ligands be selected under conditions different from those under which it will be used. We have developed a new method called the Bead blot that can (i) select ligands to unpurified proteins, including trace proteins, present in complex materials (e.g., unfractionated plasma); (ii) select ligands to multiple proteins under a variety of conditions in a single experiment; and (iii) be used with libraries of different types of ligands. In the Bead blot, a library of ligands, synthesized on chromatography resin beads, is incubated with a starting material containing a target protein for which a ligand is sought. The proteins in the material bind to their complementary ligands according to specific affinity interactions. Then the protein-loaded beads are immobilized in a porous matrix, and the proteins are directionally eluted from the beads and captured on a membrane superimposed on the beads. The location of the target protein on the membrane is determined, and because the position of the protein(s) on the membrane reflects the position of the bead(s) in the matrix, the bead that originally bound the protein is identified, with subsequent elucidation of the ligand sequence. Ligands to several targets can be identified in one experiment. Here we demonstrate the broad utility of this method by the selection of ligands that purify plasma protein complexes or that remove pathogens from whole blood with very high affinity constants. We also select ligands to a protein based on competitive elution.     

Structural and performance characteristics of representative anion exchange resins used for weak partitioning chromatography

Weak partitioning chromatography (WPC) has been proposed for the purification of monoclonal antibodies using an anion exchange (AEX) resin to simultaneously remove both acidic and basic protein impurities. Despite potential advantages, the relationship between resin structure and WPC performance has not been evaluated systematically. In this work, we determine the structure of representative AEX resins (Fractogel® EMD TMAE HiCap, Q Sepharose FF, and POROS 50 HQ) using transmission electron microscopy and inverse size exclusion chromatography and characterize protein interactions while operating these resins under WPC conditions using two mAb monomers, a mAb dimer, mAb multimers, and BSA as model products and impurities. We determine the isocratic elution behavior of the weakly bound monomer and dimer species and the adsorptive and mass transfer properties of the strongly bound multimers and BSA by confocal laser scanning microscopy. The results show that for each resin, using the product K_p value as guidance, salt, and pH conditions can be found where mAb multimers and BSA are simultaneously removed. Isocratic elution and adsorption mechanisms are, however, different for each resin and for the different components. Under WPC conditions, the Fractogel resin exhibited very slow diffusion of both mAb monomer and dimer species but fast adsorption for both mAb multimers and BSA with high capacity for BSA, while the Sepharose resin, because of its small pore size, was unable to effectively remove mAb multimers. The POROS resin was instead able to bind both multimers and BSA effectively, while exhibiting a greater resolution of mAb monomer and dimer species. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:425–434, 2017     

Role of integrins in regulation of collagen phagocytosis by human fibroblasts

Phagocytosis of collagen fibrils by fibroblasts is an important pathway for degradation of extracellular matrix in mature connective tissues. To study regulatory mechanisms in phagocytosis, 2-μm fluorescent beads coated with either collagen (COL) or bovine serum albumin (BSA) were incubated with human gingival fibroblasts in vitro. For these studies single cell suspensions were prepared by trypsinization, and bead internalization and collagen receptor expression were assessed by flow cytometry. After 3-h incubations, up to 8-fold more cells internalized COL beads than BSA-coated beads. Increased collagen coating concentration was associated with elevated proportions of cells that internalized COL beads, and was observed also in the presence of competing fibronectin-coated beads. The number of beads per cell and the percent of phagocytic cells increased proportionally with higher bead loadings. At > 4 beads per cell a maximum of ∼︁80% of cells were phagocytic. Cells reacted with mAbs against the α1, α2, and α3 integrin subunits were, respectively, 5%, 98% and 93% positively stained above background controls. All cells that internalized COL beads exhibited α2 staining but there were large proportions of phagocytic cells that were not stained for α1. In unfixed cells, bead internalization caused an immediate reduction of surface staining of membrane-bound α2 by ∼︁55% which returned to control levels within 3 h, indicating that cell-surface α2 was internalized by phagocytosis. Preincubation of cells with up to 8 COL beads per cell reduced the proportion of phagocytic cells and the number of internalized beads after a second COL bead incubation 4 h later. To assess the relationship between the percent of phagocytic cells and α2 integrin levels, serum starvation and cycloheximide experiments were conducted. Compared to controls, serum starvation for 24 h induced a 3.2-fold increase of cells internalizing COL beads but did not alter α2 staining levels. In contrast, 3 h cycloheximide treatment reduced α2 staining to 60% of control levels and this treatment also inhibited COL bead internalization. GRGDTP peptide as well as mAbs against the α1 and α2 subunits significantly reduced internalization of COL beads by 1.8 to 2.6-fold, whereas GRGESP peptide and α3 mAb exerted no effect. Internalization of BSA beads was not affected by any of these treatments. Collectively, these data indicate that the α2 integrin, along with other, as yet unidentified components, is likely involved in COL bead internalization. The α2 integrin subunit is rapidly recycled or synthesized following a phagocytic load. In contrast, the α1 integrin is not directly required for phagocytosis but may regulate the internalization step. © 1996 Wiley-Liss, Inc.     

In vivo NIRF imaging-guided delivery of a novel NGR–VEGI fusion protein for targeting tumor vasculature

Pathological angiogenesis is crucial in tumor growth, invasion and metastasis. Previous studies demonstrated that the vascular endothelial growth inhibitor (VEGI), a member of the tumor necrosis factor superfamily, can be used as a potent endogenous inhibitor of tumor angiogenesis. Molecular probes containing the asparagine–glycine–arginine (NGR) sequence can specifically bind to CD13 receptor which is overexpressed on neovasculature and several tumor cells. Near-infrared fluorescence (NIRF) optical imaging for targeting tumor vasculature offers a noninvasive method for early detection of tumor angiogenesis and efficient monitoring of response to anti-tumor vasculature therapy. The aim of this study was to develop a new NIRF imaging probe on the basis of an NGR–VEGI protein for the visualization of tumor vasculature. The NGR–VEGI fusion protein was prepared from prokaryotic expression, and its function was characterized in vitro. The NGR–VEGI protein was then labeled with a Cy5.5 fluorophore to afford Cy5.5-NGR–VEGI probe. Using the NIRF imaging technique, we visualized and quantified the specific delivery of Cy5.5-NGR–VEGI protein to subcutaneous HT-1080 fibrosarcoma tumors in mouse xenografts. The Cy5.5-NGR–VEGI probe exhibited rapid HT-1080 tumor targeting, and highest tumor-to-background contrast at 8 h post-injection (pi). Tumor specificity of Cy5.5-NGR–VEGI was confirmed by effective blocking of tumor uptake in the presence of unlabeled NGR–VEGI (20 mg/kg). Ex vivo NIRF imaging further confirmed in vivo imaging findings, demonstrating that Cy5.5-NGR–VEGI displayed an excellent tumor-to-muscle ratio (18.93 ± 2.88) at 8 h pi for the non-blocking group and significantly reduced ratio (4.92 ± 0.75) for the blocking group. In conclusion, Cy5.5-NGR–VEGI provided highly sensitive, target-specific, and longitudinal imaging of HT-1080 tumors. As a novel theranostic protein, Cy5.5-NGR–VEGI has the potential to improve cancer treatment by targeting tumor vasculature.     

Visualizing two-component protein diffusion in porous adsorbents by confocal scanning laser microscopy.

The use of confocal scanning laser microscopy (CSLM) has recently been described for the visualization of intraparticle protein profiles during single-protein finite bath uptake experiments. By coupling of fluorescent molecules to proteins the penetration of porous media by labeled macromolecules could be detected by scanning single adsorbent particles for fluorescence emission after laser excitation. Thus the internal protein distribution profile, which is a central element in modeling of protein transport in porous adsorbents, became experimentally accessible. Results from the simultaneous visualization of two proteins by this technology are shown here. The use of two different fluorescent dyes for protein labeling and two independent detectors in the CSLM allowed for the first time ever the direct observation of a two-component diffusion process within a porous stationary phase. The finite bath uptake of human immunoglobulin G (hIgG) and bovine serum albumin (BSA) to two different ion exchange adsorbents (SP Sepharose Fast Flow and Source 30S) and to an affinity adsorbent (Protein A Sepharose) was measured using Cy5 and Oregon Green as labels. Single adsorbent particles were scanned for intensity distribution of fluorescence emission from the two fluorophors. The intraparticle profiles obtained from the confocal images were translated into a relative protein concentration thus allowing the calculation of protein uptake kinetics from direct measurement in the stationary phase. The confocal technique may prove to be a very powerful means of data generation for modeling of multi-component mass transfer phenomena in protein adsorption.     

Calcium alginate beads coated with chitosan: Effect of the structure of encapsulated materials on their release

Bovine serum albumin, human haemoglobin and dextran (with different molecular weights) were encapsulated in calcium alginate beads coated with chitosan. Their release from these modified alginate beads was studied to determine what parameters related to the encapsulated materials govern their release during bead formation and storage. By comparing release of albumin (BSA) and haemoglobin (Hb) that have about the same molecular weight (67000 for BSA and 64500 for Hb), it was found that pH played an important role during both bead formation and storage. pH influences the degree of ionisation of proteins and thus the interactions between proteins and alginate; it also has an influence on the Ca²⁺-alginate and alginate-chitosan interactions. With neutral molecules such as dextran, release is directly connected to the chain molecular weights, although the flexibility of the encapsulated molecules favours their diffusion through the bead alginate-Ca²⁺ core and through the polyelectrolyte chitosan-alginate membrane.     

Direct observation of the reversible unwinding of a single DNA molecule caused by the intercalation of ethidium bromide

Ethidium bromide (EtBr) is the conventional intercalator for visualizing DNA. Previous studies suggested that EtBr lengthens and unwinds double-stranded DNA (dsDNA). However, no one has observed the unwinding of a single dsDNA molecule during intercalation. We developed a simple method to observe the twisting motions of a single dsDNA molecule under an optical microscope. A short dsDNA was attached to a glass surface of a flow chamber at one end and to a doublet bead as a rotation marker at the other end. After the addition and removal of EtBr, the bead revolved in opposite directions that corresponded to the unwinding and rewinding of a dsDNA, respectively. The amount of intercalating EtBr was estimated from the revolutions of the bead. EtBr occupied 57% of base pairs on a single dsDNA at 1 mM of EtBr, indicating that EtBr molecules could bind at contiguous sites to each other. The isotherm of intercalation showed that negative cooperativity existed between adjoining EtBr molecules. The association constant of EtBr and dsDNA (1.9 (±0.1) × 10⁵ M⁻¹) was consistent with that of previous results. Our system is useful to investigate the twisting of a single dsDNA interacting with various chemicals and biomolecules.     

Characterization of very dense mineral oxide–gel composites for fluidized-bed adsorption of biomolecules

Efficient design of fluidized-bed biomolecule adsorption from crude feed stock requires particles with elevated density, large adsorption capacity and broad chemical stability. Moreover, combinations of small particle diameters with high densities allow for high fluidization velocities while preserving a rapid mass transfer.This approach has been implemented by combining stable porous mineral oxide of high density (2.2, 4.7, 5.7, 9.4 g/ml) with functionalized hydrogels. The cross-linked hydrogel derivative fills the internal porosity of the beads and provides a high equilibrium binding capacity.Various porous mineral oxides (silica, titania, zirconia and hafnia) have been characterized in term of fluidization behavior, surface reactivity and chemical resistance to harsh CIP procedures. Porous zirconia particles were also modified into ion-exchangers by suitable surface modification and intraparticle polymerization of functionalized stable derivatives of acrylic monomers. Back-mixings in fluidized bed columns were analyzed by residence time distribution analysis of inert tracers. 328 and 218 mixing plates per meter were found for respectively, bed expansions of 1.7 and 2.9. The dynamic protein adsorption behaviors of zirconia-based polymeric anion-exchange sorbents were obtained in fluidized-bed, using BSA as model protein. A dynamic binding capacity of 62 mg/ml was observed at a fluidizing velocity of 320 cm/h. These investigations substantiate the favorable physical and chemical characteristics anticipated for dense composite beads for use as fluidized bed adsorbents.     

Transport limitation of chlorine disinfection of Pseudomonas aeruginosa entrapped in alginate beads

An artificial biofilm system consisting of Pseudomonas aeruginosa entrapped in alginate and agarose beads was used to demonstrate transport limitation of the rate of disinfection of entrapped bacteria by chlorine. Alginate gel beads with or without entrapped bacteria consumed chlorine. The specific rate of chlorine consumption increased with increasing cell loading in the gel beads and decreased with increasing bead radius. The value of an observable modulus comparing the rates of reaction and diffusion ranged from less than 0.1 to 8 depending on the bead radius and cell density. The observable modulus was largest for large (3-mm-diameter) beads with high cell loading (1.8 x 10(9) cfu/cm(3)) and smallest for small beads (0.5 mm diameter) with no cells added. A chlorine microelectrode was used to measure chlorine concentration profiles in agarose beads (3.0 mm diameter). Chlorine fully penetrated cell-free agarose beads rapidly; the concentration of chlorine at the bead center reached 50% of the bulk concentration within approximately 10 min after immersion in chlorine solution. When alginate and bacteria were incorporated into an agarose bead, pronounced chlorine concentration gradients persisted within the gel bead. Chlorine did gradually penetrate the bead, but at a greatly retarded rate; the time to reach 50% of the bulk concentration at the bead center was approximately 46 h. The overall rate of disinfection of entrapped bacteria was strongly dependent on cell density and bead radius. Small beads with low initial cell loading (0.5 mm diameter, 1.1 x 10(7) cfu/cm(3)) experienced rapid killing; viable cells could not be detected (<1.6 x 10(5) cfu/cm(3)) after 15 min of treatment in 2.5 mg/L chlorine. In contrast, the number of viable cells in larger beads with a higher initial cell density (3.0 mm diameter, 2.2 x 10(9) cfu/cm(3)) decreased only about 20% after 6 h of treatment in the same solution. Spatially nonuniform killing of bacteria within the beads was demonstrated by measuring the transient release of viable cells during dissolution of the beads. Bacteria were killed preferentially near the bead surface. Experimental results were consistent with transport limitation of the penetration of chlorine into the artificial biofilm arising from a reaction-diffusion interaction. The methods reported here provide tools for diagnosing the mechanism of biofilm resistance to reactive antimicrobial agents in such applications as the treatment of drinking and cooling waters. (c) 1996 John Wiley & Sons, Inc.