Hexavalent chromium constitutes an important water pollutant due to its cancerogenic properties. In countries such as Colombia it is widely used as a preservative in the leather industry. Bacillus sphaericus S-layer exhibits high metal-binding capacity hence the use of this microorganism to remove metals such as chromium from polluted regions based on immobilised cells systems is suggested to be an interesting proposal. The paper describes thermodynamic adsorption process of Cr(VI) on immobilised biomass of B. sphaericus cells on polyurethane foam, sol–gel, and sawdust. The adsorption isotherms describe different immobilisation mechanisms regarding the support utilised. For example, polyurethane and sawdust immobilisation process were described with the classical Langmuir model while sol–gel followed Generalised Freundlich–Kiselev behaviour. Moreover, through determining breakthrough curves in packed columns with each medium we found surface diffusion and mass bulk transfer rate as crucial factors when scaling-up the process. Le chrome hexavalent est un polluant important de l'eau en raison de ses propriétés cancérigènes. Dans des pays tels que la Colombie, le chrome est largement utilisé comme conservateur dans l'industrie du cuir. La couche de surface de Bacillus sphaericus montre de fortes capacités de fixation des métaux, d'où la suggestion de l'intérêt potentiel de l'utilisation de ce micro-organisme pour éliminer des métaux comme le chrome de régions polluées en se basant sur des systèmes cellulaires immobilisés. L'article décrit le processus d'adsorption thermodynamique du Cr(VI) dans une biomasse immobilisée de cellules de B. sphaericus, sur mousse de polyuréthane, avec la méthode sol–gel et sur de la sciure de bois. Les isothermes d'adsorption décrivent différents mécanismes d'immobilisation en fonction du support utilisé. Par exemple, les processus d'immobilisation avec la mousse de polyuréthane et la sciure étaient décrits avec le modèle classique de Langmuir tandis que la méthode sol–gel suivait un comportement Freundlich–Kiselev généralisé. De plus, avec la détermation des courbes d'interférence des colonnes à remplissage pour chaque milieu, nous avons conclu que la diffusion en surface et que le taux de transfert de masse étaient des facteurs cruciaux pour faire passer le processus à une plus grande échelle. Water contamination has increasingly conveyed the attention of the society because industries constantly waste recalcitrant pollutants due to its difficult and expensive technologies to remove them. Among those, poisonous heavy metals cause environmental and human health impact, some of them irreversible such as nausea, renal dysfunction, cancer, etc. (Wyatt et al., 1998). Tannery industry is one of the major water polluters in Colombia and worldwide (Téllez et al., 2004; Baral et al., 2006), since these factories utilise high concentration of hexavalent chromium (Cr(VI)) solutions (Kathiravan et al., 2010) for leather preservation. Once these heavy metals are disposed in the water reservoirs it is necessary to implement technologies aiming to remove or reduce it to Cr(III), which is believed to be less perilous since it does not have Cr(VI) cancerogenic capacity. S-layers are composed of self-assembling protein monomers and from paracrystalline arrays that are one of the most common surface structures on archea and bacteria. S-layer lattices display interesting geometries such as oblique (p1, p2), square (p4), or hexagonal symmetry. Gram-positive bacteria Bacillus sphaericus are reported to accumulate heavy metals such as uranium, arsenic, cobalt, and chromium by adsorption to S-layers and cell wall components. This ability has been proposed to be used in the synthesis of nanoclusters that could coordinate metals such as palladium enable them to have a role as nanocatalysts, or remove toxic metals from water trough utilising columns packed with immobilised cells or the purified protein (Sleytr and Beveridge, 1999). Several strains have been reported with interesting metal-binding features. For example, B. sphaericus JG-A12, isolated from an uranium mining waste pile accumulates high amounts of metals such as U, Cu, Pd(II), Pt(II), and Au(III) (Pollmann et al., 2006). Likewise, we recently reported B. sphaericus OT4b31 with the ability of trapping As, Co, Fe, and Cr(VI) up to 30 mg/L and was isolated from Beetles larvae (Velásquez and Dussan, 2009). Field applications for removing metals in polluted regions demand the implementation of immobilisation protocols that requires the use of packed columns where cells, once attached to the surface, grab the metals until saturating while flowing the polluted water (Kathiravan et al., 2010). In that regard, Soltmann et al. (2003a) evaluated the performance during the uranium and copper bioadsorption process with B. sphaericus cells, spores, and proteins immobilised on sol–gel. Cells display the highest binding capacity compared to matrices with spores and the purified protein. Nevertheless, they did not carry out any thermodynamical evaluation which is typically performed to determine the relation between concentration of the metal in the liquid phase with the amount attached to the surface in equilibrium (commonly known as adsorption isotherms). We believe that this is crucial in order to scale-up the process because this allows to mathematical model the process and elucidated the limiting transfer stages. Alternative biosorption strategies are based on active or passive immobilisation approaches. Active immobilisation requires the entrapment of the biomass based on chemical or physical forces so it does not demand live cells as long as the proteins conserve its functionality. On one hand, sol–gel composites are based on inorganic oxide compounds that exhibit interesting features for immobilising cells such as mechanical, thermal, photochemical stability and more importantly, they are inert to microorganisms (Gill and Ballesteros, 2000). Specifically, B. sphaericus JG-A12 biomass and purified S-layer immobilised in sol–gel is reported to remove up to 50% of uranium and copper. On the other hand, cells growing on surfaces forming biofilms support passive immobilisation and this can be utilising either inert or biological active media. Biofilm appearance demands a dissimilar phenotype compared to its planktonic counterpart since it is necessary to transcribe several genes related to expopolysaccharide production and other appendages that require to maintain living cells (Gonzalez Barrios et al., 2006). Low-cost supports have been used to passively immobilise cells such as gelatin and polyurethane. However, we are not awared of some reports carrying out thermodynamical or scale-up studies for field applications of S-layer for trapping heavy metals. Here, in this work we thermodynamically characterise the performance of living and dead B. sphaericus biomass in polyurethane and sol–gel by determining the isotherms and evaluating the breakthrough curves at different particle size and flow to identify the limiting transfer rate stage. B. sphaericus OT4b31 was cultivated in nutrient broth for 18 h at 30°C and 150 rpm (Kathiravan et al., 2010). Then, 30 mL of biomass were centrifuged at 3700 rpm during 30 min at 4°C, the precipitate was re-suspended in 8 mL of pH = 7 phosphate buffer, and mixed through vigorously stirring with 15 g of sodium silicate and 32 mL of distilled water. After homogenisation, 10 mL of chlorhydric acid (9.6% v/v) were added to onset the gel maturation process that took around 17 h. The mature gel was washed with 100 mL of distilled water and oven dried during 24 h at 40°C. Forty milliliters of overnight culture at the conditions previously described were transferred to a 200 mL of fresh nutrient broth and incubated with 0.8 g of polyurethane (1 × 1 × 1 cm3and 0.5 × 0.5 × 0.5 cm3) for 5 days at 30°C and 150 rpm. The same procedure was carried out with 3 g of small (0. 2 cm) and large size (0.6 cm) sawdust. Each model was statistical assessed with the experimental results of each support in order to find the model the best describes relates both the amount immobilised metal and liquid concentration. Three different PVC columns were utilised for determining breakthrough curves (12 cm high and 2 cm diameter, 45 cm high and 2.5 cm diameter, 20 cm high and diameter 2.5 cm) were packed with 81.2, 0.8, and 4 g of sol–gel, polyurethane (big size and small size) and sawdust (medium size and small size), respectively. 1 and 3.75 mL/min of a Cr(VI) solution flowed through the column and samples were taken every 5 min in order to determine the saturation point. First, cells were grown to exponential phase and harvested by centrifugation at 1200 rpm and 4°C during 10 min. These were three times washed with buffer Tris–HCl 50 mM (pH 7.4). Then, cells were analysed through sonication with three pulses and the cell membrane was collected by centrifugation during 1.5 h at 19 000 rpm and 4°C. The pellets were resuspended in TritonX 100 0.5% and washed four times with the same buffer These were resuspended in guanidine hydrochloride 5 M without agitation during 2 h and the solution was centrifugated at 19 000 rpm and 4°C during 20 min. Finally, the supernatant was dialysed with distilled water in a sigma® D9277 membrane at 4°C and during 20 h. The membrane is centrifugated at 19 000 rpm, 4°C and 20 min, and the bottom resuspended in 1 M phosphate buffer (pH 7) (Pum and Sleytr, 1995; Soltmann et al., 2003b; Völlenkle et al., 2004; Hu et al., 2008). Sodium silicate 1.98 M was progressively added to a HCl 1.31 M solution until reaching a pH of 4. Then, a solution with phosphate buffer pH 7 and 1 M with the isolated protein was added. The gel was overnight incubated, washed three times with distilled water, and dried at 30°C during 60 h. Thirty-five milligrams of the sol–gel with the protein were added to 50 mL of Cr(VI) solution in constant agitation (100 rpm). Negative control consist of sol–gel at the same conditions without the presence of the protein. Adsorption isotherms were obtained to understand the mechanisms behind the process of cell entrapment when reaching the equilibrium for all supports. Interestingly, even though the immobilised biomass presents the same nature (B. sphaericus cells), different supports display a dissimilar mechanism when adsorbing because the adsorption models for polyurethane, sawdust and sol–gel suggest dissimilar processes of Cr(IV) immobilisation (Figure 1). Substrate capability of adsorbing Cr(VI) was also evaluated in order to assess the performance of the biomass and S-layer proteins trapping the metal. Sawdust and sol–gel exhibit 13% of metal removal capacity while polyurethane was not detected. These results indicate that sol–gel seems to have a big affinity for the metal and the contribution of the protein and biomass should be subject to further test. Adsorption isotherms for the immobilisation of hexavalent chromium in Bacillus sphaericus biomass for polyurethane foam (A), sol–gel (B), and sawdust (C). On one hand, polyurethane exhibits Langmuir kinetics with two clear zones detected: a positive correlation between the equilibrium concentration and the amount adsorbed and the saturation region. Nevertheless the positive correlation region was hardly found because this trend is present for narrow interval concentration. We obtained the isotherms for two different particle sizes (0.5 × 0.5 × 0.5 cm3 and 1 × 1 × 1 cm3) and two different temperatures (results not shown) aiming to uncover any important effect of the immobilisation area on the equilibrium considering that this support requires the formation of a biofilm (passive immobilisation) and that this information is also relevant for scaling-up purposes. Both particle sizes exhibit the Langmuir trend but the smaller size displays a bigger adsorption capacity, hence these results suggest that the available area that controls the amount of immobilised biomass plays an important role in the maximum amount of pollutant adsorbed but the monolayer mechanism remains (Figure 1A). Moreover, this support enables the system to show both high affinity and chromium adsorption rates, both reflected on Q and b (Table 1). On the other hand sawdust also supports a linear correlation between the equilibrium concentration and the amount adsorbed per unit area for the interval of concentrations analysed and the slope is not highly influenced by the temperature or particle size (Table 1). Even though both polyurethane and sawdust are interesting cheap supports due to its adsorption capacity the polyurethane presents more adsorption capacity. The adsorption process in the nanostructure oxide matrices such as aqueous silica nanosols seems to be radically different from other supports. Even though the increase in the equilibrium concentration at liquid phase leads to an augment in the amount of adsorbed metal on the surface, this trend is not linear and the relationship is strongly affected with the temperature (Figure 1A). Soltmann et al. (2003a) performed a similar immobilisation strategy for uranium and copper trapping. Nevertheless they did not carry out any thermodynamical approach to analyse the process of adsorption. Nevertheless our results fit theirs in terms of adsorption capacity as our bound metal percentage is similar for different metals (uranium and copper). This particular behaviour was only possible to be described with the BET conduct (Marczewski and Szymula, 2002; Mendoza de la Cruza et al., 2009). We believe that at low temperatures (18°C) two different mechanisms are taking place depending of the stage of the process. Firstly, monolayer adsorption appears to govern the situation and once the monolayer is constructed the formation of multilayers (Mendoza de la Cruza et al., 2009) rules the process whereas at high temperatures (30°C) the biomass is not uniformly distributed on the active sites causing a multilayer formation for the whole concentration range (Mendoza de la Cruza et al., 2009). BET isotherms are easily identified through the identification of an inflection point which seems to be our case (Figure 2A, 18°C isotherm) and can be mathematically described with the G-FK model (Mendoza de la Cruza et al., 2009): We evaluated the isotherms at two different drying temperatures in order to corroborate the results obtained for the polyurethane regarding particle size. It is known that this temperature process plays a big role during the formation of the particles because it affects the particle size, and, possibly, particle rugosity. Our results demonstrate that the immobilisation process on the sol–gel support is based on multilayer deposition regardless of the particle size and temperature because G-FK model describes adequately the results obtained (Table 1). Different m values possibly indicate that the adsorption temperature is important for the deposition mechanism (low and high m represent monolayer and multilayer behaviour, respectively; Marczewski and Szymula, 2002). Breakthrough curves in polyvinyl chloride columns for living and dead Bacillus sphaericus cells in polyurethane (A) (for 1 and 2 mL/min in polyurethane and 1 and 3.75 mL/min in sawdust), at different inflows for polyurethane and sol–gel (B), and particles sizes in polyurethane and sawdust (C). A deeper comparison between matrices could be reached if determining the adsorbed chromium relative to the amount of biomass. In this case, sol–gel constitutes an excellent candidate as it displays an adsorption capacity around 1025 mg chromium per gram of biomass while saw dust and polyurethane 130 and 580, respectively. The scalability of these supports is basically settled on both cost and adsorption capacity. In that regard, a comparison among them allows to select sol–gel because it shows the highest adsorption capacity yet it is broadly available and easy to handle for field applications. Scaling-up constitutes one of the big challenges when applying biosorption technologies on field. Along this comes the need of finding the limiting transfer rate, saturation time, and the effect of important design parameters such as diameter, flow, and particle size. We then determine breakthrough curves in order to find the underpinnings behind the mass transfer process in pilot scale given the fact that this information can be applied on field. First we assessed the significance of utilising live cells by determine curves with either living or non-living cells (results not shown). Both scenarios exhibit a similar behaviour as they take similar saturation times. The cells state during the biosorption is a significant information when using passive immobilisation because this fact resolves the need of providing nutrients with the fluid that contains the pollutant to maintain the cells alive (Figure 2A). Then, our results suggest that the cost for field applications could be lowered because no nutrients have to be added. These results corroborate previously findings reported by us in this journal where the same strain exhibits similar Cr(VI) removal capacity when it is not immobilised on a specific support (Velásquez and Dussan, 2009). Fluid flow effects were also analysed to uncover the limiting transfer stage. It is well known that there exist several mass transfer stages that the metal must overcome to reach the surface of the support so it can be trapped. First we weighed the importance of the bulk-surface transfer through determining breakthrough curves at different flows. It is expected that if this stage plays a big role the saturation time should have to present high sensitivity when modifying the flow. Polyurethane and sol–gel columns displayed similar sensitivity and this sensitivity turned out to be considerable (Figure 2B). These results opposed those obtained for sawdust columns where no significant effect of the flow on the saturation time was obtained (results not shown). Once the particle reaches the surface it is necessary to diffuse inside the particle to reach the active sites where the S-layer is located. This process is in part governed by the size of the particle since it is indirectly correlated with the average length of the pathway the pollutant has to go through. Breakthrough curves were evaluated at two different sizes for sawdust and polyurethane. In this case particle size seems to have a strong influence on polyurethane (Figure 2C), even stronger compared to fluid flow. These results suggest that intraparticular diffusion and bulk transfer are the limit the transport in polyurethane and sawdust respectively. In order to demonstrate the ability of the protein to bind the metal, we carried out a concentration of the protein and tested its capability to bind Cr(VI). S-layer purification was corroborated through sodium dodecyl sulfate electrophoresis which showed the presence of bands the molecular weight expected (120–125 kDa, results not shown). The S-layer protein was then immobilised on the sol–gel and exposed to a 2.5 mg/L solution of Cr(VI). We found that the Cr(VI) removal was 80% with the presence of the protein (S-layer). Negative control was carried out with the sol–gel exposed at the same conditions finding a 5% of removal. This results demonstrate that the main role of the S-layer during the adsorption process. Here in this work we carried out a comprehensive comparison between three different supports for S-layer biomass immobilisation from the scaling-up perspective. Overall the results indicate that all sol–gel, polyurethane, and sawdust are suitable for immobilising B. sphaericus biomass and remove Cr(VI). Nevertheless, our study also points out some differences among these supports that should be considered for scaling-up this technology. First, we found that even though the biomass presents the same source the thermodynamical equilibrium between the liquid concentration and the amount of Cr(VI) immobilised on the surface depends mainly on the support. We also found that sol–gel equilibrium was only possible to be described by the G-FK model which suggests multilayer deposition, contrary to sawdust and polyurethane which exhibit the classical monolayer behaviour. Regarding the mass transfer, it was found that intraparticular diffusion governs the metal transfer for polyurethane and bulk transfer rate for sawdust possibly indicating that different parameters should be the pivot for scaling-up based on the support. We recommend polyurethane for Cr(VI) using B. sphaericus due to its high adsorption capacity, low price, and availability. This work was funded by Vicerrectoria de Investigaciones at Universidad de Andes.