Skip to main content
European Commission logo
English English
CORDIS - EU research results
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary
Content archived on 2024-06-18

Biosensor nanoarrays for environmental monitoring

Final Report Summary - BIOMONAR (Biosensor nanoarrays for environmental monitoring)

Executive Summary:
BIOMONAR aimed to develop a suite of dynamic nanoarray biosensors for monitoring of environmental pollutants and pathogens, comprising three complementary platforms incorporating members of a single family of selector proteins, namely bacterial periplasmic binding proteins (PBPs). The project takes advantage of two of the main characteristics of this group of proteins: their demonstrated sensitivity to the immediate environment and their proven amenability to be "tailored" to the detection of specific target chemicals or groups of chemicals. The PBP protein family currently holds supreme promise to achieve predictable and controllable nanobiotechnological interfaces for biosensing. Using state-of-the-art "test-tube evolution" approaches, combined with a high throughput screening, BIOMONAR worked towards generation of a library of engineered PBPs, with differing detection specificities and sensitivity ranges, that could be integrated into three classes of sensor platforms: functionalised solid surfaces, PBP-containing liposomes, and live bacterial strains. A number of elements in this process (detection, culture, biosensor strain innovations) are supportive of commercialisation.

Using several known PBPs, including those for maltose (MBP), ribose (RBP) dipeptides (DppA), and nickel (NikA) the functioning of the three types of sensors was established. At functionalised surfaces, the sensitivity of detection by MBP, DppA, and NikA was significantly greater at nanostructured surfaces as compared to flat ones. Nanomolar concentrations of all targets were detectable. For liposomal systems, a robust coupling protocol was established which maintains PBP functionality and liposome signaling abilities. It was determined that an MBP coverage on the liposomal surface of 0.1 mol% was optimal and lead to a limit of detection for maltose of as little as 10 nM, with a dynamic range of 10 nM to 50 uM. Such a low limit of detection has not been achieved using antibody approaches, thus demonstrating the importance of PBPs for certain analytes. A highly specific, sensitive high-throughput bioassay was developed using sulforhodamine B-encapsulating liposomes for fluorescence signal generation. The platform can be used for the detection of other analytes taking advantage of other recognition elements of the PBP protein class. Existing whole-cell bioreporters for a range of targets were characterised. A PAO-based flow-chip was designed and constructed to enable the simultaneous exposure to the sample of multiple bioreporters, each able to detect a different chemical or a group of chemicals. The performance of the microfluidic flow cell was tested in environmental samples. It was established that bacterial growth could be sustained in natural environmental water and that test compounds could be detected in a natural context. Furthermore, any natural particulates did not interfere with the flow cell system.

Engineering of new PBPs proved problematic. It was concluded that previous literature reports were erroneous. In developing new strategies, significant progress was made in sorting out the fundamental structure/function relationships in PBP functioning.

Significant advances were made in understanding the chain of detection events that is involved in determining the response of the sensors to environmental targets. The dynamic speciation of target compounds (metal ions, organics, and nanoparticles) was quantified for a range of environmentally relevant conditions. Key knowledge gaps addressed were the chemodynamics of target interactions with nanoparticles and speciation within biogels having properties relevant for biological interfaces.

All together, the developed theoretical framework enables computation of the flux of target species at a given biosensor surface. This knowledge allows the biosensor signals to be translated into exposure conditions in the environmental media, according to the myriad of processes involved in the full detection chain.
Project Context and Objectives:
Environmental samples are complex and it is unrealistic to aim for a ‘one size fits all’ biosensor. A non-systematic approach to biosensor design, resulting in a suite of very diverse sensors with few common features, hinders conceptual understanding of environmental processes and design of sensors for new targets. Macroscale effects of target chemicals and biologicals on organisms or the ecosystem represent the integration of many smaller-scale events over given spatial scales and timescales. This implies that progress in quantitative characterisation, modelling, and prediction of macroscale impacts can only be made if the underpinning molecular level processes are understood. Characterisation of environmental molecular scale processes is a challenging task that encompasses a wide range of issues. For pollutants in aqueous environments, for example, the current condition for good water status is to comply with a quality standard. However, it is well established that, by itself, the total concentration of a pollutant is a poor predictor of its potential impact. The same holds for biological targets, and in environmental matrices other than water bodies (e.g. atmosphere, soil, tissue).
BIOMONAR proposes that deployment of a battery of nanoarray biosensors with a common but highly versatile selective element, for measurement of a broad spectrum of chemical and biological parameters, complemented by well-defined chemical partitioning based sensors, is the best solution for environmental monitoring. In such a way BIOMONAR aims to achieve a sufficiently detailed level of physicochemical and biological understanding that allows interpolation and extrapolation to the particular conditions of interest for the given organism and target analytes. It has been shown that a common dynamic interpretation framework is applicable for a suite of speciation sensors that span a range of time scales. Interpretation from a kinetic spectrum of speciation data is a practical strategy for addressing questions of bioavailability of inorganics and organics, and of dynamic features of water bodies. Understanding the relative contributions from the diffusive and kinetic routes is very significant for relating speciation of compounds and metals to biouptake. For example, in the case of metals, such information provides an indication of the extent to which complex species are utilised by an organism. The kinetic spectrum information allows the computation of compound bioavailability for any set of biouptake flux conditions.
Whereas the construction of more sophisticated nanobiotechnological sensors is the core of this project, embedding their signal output in an interpretable framework is an additional must for full exploitation of their potential. In other words, measurements from chemical and biological analysis cannot stand alone, but have to be placed in the context of a full detection chain. Scientifically well-founded criteria for acceptable measurement, threshold or intake levels must be based on a quantitative understanding of the link between physicochemical reactivity of the given compounds on one hand and their biological effect(s) on the other. BIOMONAR tackles this challenge by coupling the information provided by chemo- and bio- sensor datasets and by investigating the complete detection chain.
The three sensor platforms constructed in BIOMONAR (functionalised solid surfaces, liposomes, and live bacteria) probe different aspects in the ‘exposure to effect’ chain of processes: each responds to a certain proportion of the total target concentration and has a characteristic dynamic window.
Detailed studies of both the natural and engineered bio/bio (b, c) and bio/nonbio (a, b) interfaces are carried out, allowing the design of both biological (cell surfaces) and non-biological (liposomes, solid surfaces) structures that will interact in a predictable manner with the selective PBPs and with the target compounds. Aspects characterised included the fundamental properties of biointerfaces (e.g. permeability towards various chemical species, Donnan potential profiles), dynamics of processes occurring at nonbio/bio and bio/bio interfaces, and the interactions of PBPs with surfaces and target molecules at the nanoscale (conformational changes, binding affinity, association/dissociation rates). A key element is analysis of the dynamic nature of the processes that generate the sensor response.

One of the major strengths of BIOMONAR’s approach is to use an array of sensing units, containing a multitude of selective elements and hence exhibiting a practically unlimited sensing capability. This implies a major step forward in the development, control and performance of a new generation of nano-engineered biosensors and biochips. The BIOMONAR approach will allow the rapid accumulation of rich and diverse information on sample composition, as well as on the nature of the surface interactions between the target chemicals and the sensor elements at the nano- and micro-scales. Data furnished by the three sensor platforms is quantitatively interpreted on the basis of the fluxes and reactivities of target compounds and the rates of the involved processes. The results provide insights into the dynamic relationships between the composition of the exposure medium and the ensuing biological impact.

Why focus on periplasmic binding proteins?
Periplasmic binding proteins form a large superfamily of interception proteins from bacteria, archaea and eukaryotes, located outside the cytoplasmic membrane (or, in case of Gram negative bacteria, in the periplasmic space) of the cell. They play a role in scavenging and sensing diverse compounds in the environment and act as initial receptors for active transport and chemotaxis. Ligand-bound complexes typically dock to an integral membrane transport protein complex allowing transport of the ligand to the cell’s interior. In other cases, ligand-bound PBPs interact with a chemotaxis membrane sensory complex which relays a signal transduction chain propelling movement of the bacterial cell. Essential for all PBPs is that they have an ‘open’ and a ‘closed’ configuration after binding their ligand molecule, which involves a distinct bending or swift movement around a central ‘hinge’. PBPs are found in mM concentration in the periplasm of bacterial cells, leading to high specific substrate affinities by the cell. Binding affinities of PBPs are in the range of nM to µM, i.e. in the order of what is expected for the number of hydrogen bonds formed between the PBP and its target.
The PBP superfamily is currently comprised of some 50 distinct subfamilies with similar domain organisation. Ligands bound by the different PBPs include a variety of compounds (thus representing a range of potential targets), among which are sugars such as galactose, maltose, ribose and arabinose, or short sugar chains (maltotriose, cyclodextrins). The most well-characterised PBP of this family is MalE, the maltose binding protein from Escherichia coli, for which crystal structures with and without a bound target have been published. Other PBPs are involved in the recognition of amino acids such as glutamine, glutamate/aspartate and histidine, or di- and tripeptides. Metal ions such as Fe(III), Ni(II), Hg(II), and molybdate are also intercepted by PBPs, in some cases in concert with specific siderophore molecules. Other PBPs have been documented for phosphate and sulphate. PBPs, therefore, show great potential for exploitation in biosensing due to their specific and sensitive recognition of a myriad of target molecules and the concomitant target-induced conformational rearrangements. At the outset of the project there was published data indicating that PBPs:
• are amenable to fluorescence resonance energy transfer, FRET (exhibit intermolecular movement upon binding)
• have relatively high binding constants for chemical effectors
• are subject to successful engineering of binding pockets for new effectors, and
• can be functionalised in bacterial bioreporters via an artificial signalling chain.
The above features identify PBPs as excellent candidates for a versatile universal selection element for diverse chemical and biological targets. BIOMONAR paid attention to designing the local sensor environment to protect the PBP activity. A number of existing PBPs are initially re-examined in the three BIOMONAR platforms (see below). The work then progresses to molecular engineering of new PBP specificities for detection of novel target molecules.
Scientific and Technological Objectives
Current biosensor and biochip designs are based on several basic configurations, differing in both the biological sensing entity and the detector/transducer elements. Whereas each system design has its own specific merits, important inherent characteristics may limit further development or optimal usage. BIOMONAR combines the advantages of three different biosensing concepts, each containing an overarching new selective element.
For example, many solid-state biosensor designs depend on purified proteins (e.g. antibodies) that are bound to a surface, and the direct conformational change upon target binding, or some other association parameter, is detected. Selection and manufacturing of the antibodies, however, requires a labor-intensive process. Similarly, enzyme activity-based biosensors use purified proteins that are subsequently embedded in suitable materials or immobilised on surfaces; the enzymatic reaction elicited in response to the presence of a target compound typically leads to signal amplification, thus lowering the detection threshold. Both types of protein-based biosensor design have the advantage of reacting ‘as is’, meaning that no a priori synthesis of the sensing or reporting elements is needed.
At the other end of the biosensor scale, live cells and microorganisms form the basis of the sensor. Living systems have the non-negligible advantage that no protein purification is needed for the sensor to function, and the selective element is self-replicating. This last point can be a major cost advantage. In addition, cell- and microorganism-based sensors inherently provide biologically relevant information. The output of the cell allows conclusions to be drawn about the biological impact of the target molecule. Only living systems, for example, can provide direct information as to the toxicity, mutagenicity, or bioavailability of a toxic pollutant. However, a consequence of this feature is that whole cell biosensors are often less specific and require different maintenance than a solid-state device. As outlined below, sensing elements in whole cell biosensors often consist of transcription factors, which elicit gene expression. This amplifies the signal to a higher detection level, but often at the cost of increased response time (i.e. the time needed for gene transcription and protein synthesis).
BIOMONAR proposes the development of a new generation of biosensors that will combine the advantages of all existing sensor systems, while maintaining a flexibility for easy exchange of the selective PBP components and/or sensing platform according to the target(s) and processes of interest. This new generation of biosensors allows rational and facile design of the most optimal biological sensing element. Ideally, a protein class should be used, which will allow a versatile design of recognition specificities for molecules that are currently on the ‘most wanted’ list, as well as for potential targets. Based on extensive available data and accumulated experience, BIOMONAR believes that the class of bacterial periplasmic binding proteins (PBPs) is ideal for such multi-biosensor system purpose and for rational design. To specify this further, PBPs such as ribose-binding protein can be easily purified and maintained in solid-state sensor configurations, after which their intramolecular conformational change upon sugar binding can be detected by e.g. FRET. PBPs with completely different substrate recognition properties have been designed, for example, to serotonin. Finally, PBP target compound binding and signal transduction to elicit reporter gene expression has been achieved in engineered Escherichia coli cells by a hybrid signal transduction pathway. Despite these proofs of principle, however, the proposed class of PBPs has not been mined extensively as next generation bio- and nanosensor tools.
The most pertinent and overarching scientific and technological objectives of BIOMONAR are:
1) Generate the next class of biosensors based on PBP as sensing element, in three different system platforms: (a) solid-state nano-patterned sensor, (b) liposomal sensor, (c) whole cell living bacterial biosensor.
2) Demonstrate the design flexibility principle, by creating a suite of previously non-existing sensing specificities, for a target list including a heavy metal (cadmium), a pharmaceutical (triclosan), an endocrine disrupting compound (tributyltin), an engineered nanoparticle (Ag), and at least one biological pathogen.
3) Investigate the physicochemical basis of the biological sensor-target interaction, involving bio/bio and nonbio/bio interfacial processes. Apply this knowledge to optimisation of the three different platforms, with relevance to the expected biosensor performance (detection limits, precision, accuracy) and to biological and ecotoxicological questions related to the target molecule.
4) Promote and instruct in the use of this new generation of biosensors as alternative and complementary methods for existing chemical and toxicological analysis in environmental and biosafety analysis.
Project Results:
The results are summarised below according to the main sections of work.

ELUCIDATION OF BIOINTERFACIAL MECHANISMS AND EXPOSURE CONDITIONS AT THE BIOINTERFACE

Aim: to understand the nature of the interaction between target molecules and the biological sensing elements (periplasmic binding proteins, PPBs).

Characterisation of PBP binding of target molecules in solution: how strong is the interaction?
The binding of target molecules to selected periplasmic binding proteins (PBPs) was characterised by a range of techniques. The PBPs included the binding proteins for maltose (MBP), nickel (NikA), and dipeptides (DppA). The techniques employed included fluorescence spectroscopy, solid phase microextraction (SPME), stripping chronopotentiometry at scanned deposition potential (SSCP) and liposomal assays. Results for PBPs immobilised on surfaces and within liposomes are presented in later sections. Using SSCP in solution, NikA was found to strongly bind both Ni(II) and Pb(II). For comparison with results from the bacterial reporter cells, measurements were also performed in culture media (M9 minimal medium and in LB medium). The extent of metal ion binding to NikA was greatly reduced in M9 medium (due to competition from phosphate) and no binding was seen in LB medium.

For organic targets, to establish the interpretation basis and kinetic window for speciation analysis by SPME, the interaction of DNT and diclofenac with the model protein, bovine serum albumin (BSA) was studied. The shape of the uptake curve and time necessary to reach partition equilibrium between the sample and SPME fibre reflects the dynamics of the solution speciation, whilst the amount of analyte accumulated at equilibrium reflects the free (unbound) portion in the sample solution. BSA was found to form associates with both DNT and diclofenac that are labile on the timescale of SPME. Studies were also performed in the presence of sorbing oxidic nanoparticles, and for the case where the nanoparticles themselves enter the solid phase. This information provides a sound basis for interpretation of measurements with the PBP-based sensors: PBPs may behave as soft nanoparticles, and in environmental media the parameters of interaction between the target organic and the PBP will be conditional with respect to the type and concentration of the nanoparticles and other competing binding agents in the sample medium.

Characterisation of bio/bio and nonbio/bio interfacial processes: the significance of the charge characteristics and permeability of the interphase
To achieve a comprehensive understanding of the dynamic interactions between the sensing platforms (functionalised surfaces, liposomes, sensing bacteria) and the target analytes, major progress was achieved on the theoretical evaluation of:
(i) the relevant physico-chemical (electrohydrodynamic) properties of the platforms, whether they are macroscopic planar sensing surfaces or sensing (bio)colloids,
(ii) the interfacial mechanisms governing the association/dissociation of the analytes with the reactive sites located either over the surface or volume of the sensing platforms. Theory now allows for defining the relationships between the physicochemical characteristics of the platforms as determined in (i) and their dynamic reactivity versus the analytes of interest.
A general theory was developed for derivation of the key electrostatic and hydrodynamic permeability properties of sensing planar platforms that are coated with gel layers. The model has been extended to account for the often encountered chemical stratification present e.g. in sensing multilayered hydrogels. The theory was applied to interpretation of experimental data collected on bilayered polycationic/polyanionic platforms as well as on a phospholipidic bilayer supported by polycationic cushion. It is now possible to extract the physicochemical properties of soft (bio or non-bio) patterned surfaces from their measured electrokinetic profiles (i.e. streaming current versus ionic strength).

Determination of physicochemical properties of biogels and target speciation within biogels: the distribution of compounds within a gel can be different from that in the surrounding medium
The dynamic behaviour of natural metal complexes in their partitioning between aquatic media and model biogels was characterised. The partitioning of e.g. cadmium(II)/humic acid species between water and aqueous alginate gel was measured by voltammetric and spectroscopic techniques. The results are quite intriguing with huge accumulation of cadmium ions by the gel, up to orders of a factor of 100 for freshwater ionic strength levels. To some extent, the accumulation of cadmium ions is caused by electrostatic effects due to the negative charge of the gel phase. But the strongest drive for the cadmium(II) accumulation derives from the specific chemical interaction between cadmium ions and binding sites on the alginate structure. Surprisingly enough, the negatively charged polyions of humic acid also accumulate in the alginate gel phase, the more so at higher ionic strengths. This result was also found with a synthetic polyacrylamide gel, PAA, that is employed in a widely used speciation sensor. Indeed, a striking concentration profile is observed by scanning confocal laser microscopy, with the large humic acid entities accumulating at very high levels at the water/gel interface. The outcome is that the concentrations and kinetic properties of metal species within the gel may differ drastically from those in the surrounding aqueous medium. The results are significant for interpretation of sensor signals, for predictions of bioavailability, and for understanding a range of interfacial processes in aqueous systems, e.g. as occurring at/within biofilms.

Conclusions:
Some periplasmic binding proteins are shown to have strong interaction with target molecules. The nature of the interface in which receptor molecules are embedded can influence the sensor response.

BIOSENSOR PLATFORM ENGINEERING

Aim: to integrate biological sensing elements (periplasmic binding proteins) within three different types of sensors: functionalized surfaces, liposomes, and bacterial reporters.

Microfluidic flow cell and platform engineering: combining optical sensing with microfluidics
A Mach-Zehnder Interferometry (MZI) evaluation platform for evanescent wave based characterisation of PBP target interactions was developed and optimised. In terms of performance and manufacturability, an asymmetric MZI and the Micro ring resonator (MRR) were found to be the preferred platforms. The interferometric sensor platform is based on TriPleX™ waveguide technology. The ring resonators and the asymmetric MZI operate in the near infrared (850 nm) enabling the use of very cost effective VCSELs as a light source. The experimentally obtained response of the ring resonators is in good agreement with theory, while the measured through and drop responses show very low on-chip losses. The chips show good coupling efficiencies to external fibers due to integrated spotsize convertors. The corresponding signal-to-noise ratio allows measurements of changes in refractive index smaller than 1×10-6 RIU. The ring resonators are combined with an 850 nm VCSEL as light source and prototype electronic equipment for signal processing.

For lab-on-chip-applications the integrated optical sensor has to be combined with microfluidics: so-called opto-fluidics. The boundary conditions on the required microfluidics fabrication technology are quite often difficult, if not impossible to combine with those of the integrated-optics circuitry. Therefore, it is preferable to fabricate the micro-fluidic structures on one substrate (fused silica or borofloat), and the integrated optical structures on a different substrate (silicon or fused silica), and subsequently attach both structures to each other. This enables flow-through applications. With the produced Micro Ring resonator and Asymmetric-MZI, thermal bonding was used for substrate attachment (a bonding process at a high temperature close to the weakening point of (one of) the substrate(s)).

For final version of the MRR and MZI sensor chips includes a reference sensor integrated on the same chip, and both sensors share the laser light which is split 50/50 on chip. The reference ring sensor compensates for drift induced by temperature changes.

The performance of the chips was tested for monitoring of a biochemical surface reaction, namely growth of a layer of bovine serum albumin on the ring resonator system. The results indicated an average BSA monolayer thickness of slightly more than 3 nm, perfectly in line with other results. In view of the measurement resolution of 0.1 pm peak shift, this biochemical surface sensitivity allows for the detection of an average BSA-layer growth smaller than 1 pm.

For measurements with live bacterial reporter cells, porous aluminium oxide based flow cells were developed which allow microfluidics to supply either a strip of porous aluminium oxide (PAO) or a more highly subdivided culture chip (the so-called MDCC). Biosensor strains cultivated on the porous support can be grown and exposed to compounds to be sensed. The flow cell is a highly flexible device, test compounds can be added or removed to examine the kinetics of biosensor response; alternatively microorganisms expressing libraries of periplasmic binding proteins can be screened. Chips with 32 and 64 wells, together with user-friendly chip holders, have been produced and successfully tested with reporter cells (see below).

Functionalised surfaces
A biosensing platform was developed by modifying nanostructured gold surfaces with sugar via a thiol-linker, and using commercial Maltose Binding Protein as biological intermediate. Scanning electron microscopy indicated a nanopattern consisting of lines, with a pitch between lines of 500 nm, comprising 48% of active area. The atomic force microscopy images display an increase of the roughness of the functionalised surface of almost 2 nm after thiol-glucose surface modification and of 4 nm after MBP immobilisation. The thio-glucose nanopattern was tested with surface plasmon fesonance (SPR): MBP was injected at different concentrations (1, 5, 10, 25 and 50 μg/ml) and the SPR signal increased after every injection of protein.

The sensitivity of detection of the MBP on the surface as well as the affinity for the immobilised sugar was measured. The sensitivity of detection of protein on sugar surface had similar values for the non-structured area (0.030 a.u.*ml/μg) and for the nano structured areas (0.032 a.u.*ml/μg). However, the sensitivity normalised to the active area is more than double in the nanostructured areas.

The affinity of MBP for the thio-glucose immobilised on gold nanopatterned surface was measured. The dissociation constant was estimated as 301.91pM confirming the strong affinity of the protein also for the immobilised sugar; the affinity of MBP for the thio-glucose surface decrease in presence of maltose: the dissociation constant of protein in presence of maltose was estimated as 13.3069 nM. The interaction of bialaphos with immobilised DppA and of Ni(II) with immobilised NikA at nanopatterned and flat surfaces was characterised. Nanomolar concentrations of both targets were detectable.

Liposomes
A robust coupling protocol is established which maintains PBP functionality and liposome signaling abilities. Using MBP, recognition of maltose was achieved in a liposome-based microtiter plate competitive assay. A variety of anti-MBP antibodies were immobilised on a microtiter plate surface, then MBP-liposomes were incubated with varying concentrations of free MBP in the wells. A concentration dependent decrease in fluorescence signal from the SRB dye within MBP-liposomes was obtained with increasing concentrations of MBP, indicating successful recognition and conjugation of MBP on the liposome surface.

Thorough studies on the competitive maltose-surface were carried out. It was found that surface-bound small molecules did not provide a successful binding surface and that high molecular weight polymers such as amylose or amylopectin were necessary to exhibit good surface binding. Such polymer surfaces provided appropriate competition and low limits of detection for maltose detection. Various conditions were optimised to obtain the best signal-to-noise ratios and dose-response curves. It was determined that an MBP coverage on the liposomal surface of 0.1 mol% was optimal and lead to a limit of detection for maltose of as little as 10 nM, with a dynamic range of 10 nM to 50 M. Such a low limit of detection has not been achieved using antibody approaches, thus demonstrating the importance of PBPs for certain analytes.

Specificity testing was also performed. Here, other sugar components were tested up to concentrations of 6.5 mM. Dextrose, galactose, glycerol, fructose, raffinose, sorbitol, sucrose, thioglucose, trehalose and xylose were tested, and no non-specific signals were obtained. Cross-reactivity towards glucose and lactose only occurred at 10,000-fold greater concentrations of these sugars versus maltose.

Reporter cells
The previously published RBP sensor in E. coli was successfully reconstructed and optimized. The sensor has excellent detection of ribose and a relatively high specificity for this sugar. By contrast, and despite several lines of experiments both in vitro and in vivo, the published mutant RbsB sensor for TNT in an E. coli host was not functional, and induction of a reporter gene upon addition of DNT or TNT was not observed. It was concluded that the original literature must have been erroneous, and thus fundamental studies were initiated to establish better predictive rules for PBP re-engineering (see below). A likely problem with the RbsB-TNT mutant protein is its stability and false localization in the cell.

In parallel, approaches were developed to improve bacterial sensor performance by increasing the fluorescent output signal. The original bacterial system consists of three plasmids, each of which expresses a different element of the system, i.e.: ribose binding protein (rbsB), trz1 fusion protein and ompC::gfp fusion reporting element. To improve bacterial sensor performance, three different approaches were explored:
1. Integration of the fluorescent element into the bacterial chromosome.
2. Increasing fluorescence signal intensity by duplication of the ompC::gfp fusion reporting
element.
3. Simplifying the system by expressing ompC::gfp and trz1 from the same vector (two
plasmids merging into one plasmid) that has a high copy number in the bacteria.
By merging two plasmids the response is 5-fold higher than that of the original system.
”On-chip” exposure of immobilised reporter bacteria was tested using the bacterial microfluidic flow cells described above. Various growth conditions and bacterial strains were tested. The design of the PAO-based flow-chip enables the simultaneous exposure to the sample of multiple bioreporters, each designed to detect a different chemical or a group of chemicals. This concept was demonstrated using a 6-member array tailored to detect genotoxicants, heavy metals, 2,4-DNT and related aromatics and oxidative stress. A constitutively fluorescent strain served as a positive control for the viability of the bacteria. The toxicant-containing sample was flowed through one channel, while the other channel was continuously exposed to a control medium. As expected, exposure of the array to HQ, CdCl2 or paraquat led to selective induction. Paraquat, a strong oxidant and a redox-cycling agent, also induced the genotoxicity reporter. On-chip exposure of a 4-member array to a mixture of model chemicals (NA, paraquat, CdCl2 ¬and quinol) caused the simultaneous induction of all strains, as expected.

To demonstrate the biochip potential as a simple minimal maintenance “plug-in” sensor cartridge, bacteria bioreporters were deposited on-chip, freeze dried and kept for up to 12 weeks at -20 °C. The bacterial activity was retained after 12 weeks storage of freeze-dried biochips. The storage of freeze-dried reporter strains was demonstrated on both PAO and MDCC. This reduces the need for microbiology facilities to set up living cell biosensor systems.

Conclusion: the principle of PBPs as sensing elements was shown to function in all three types of sensing platforms and an opto-fluidics system was developed for detection.

DIRECTED EVOLUTION AND SCREENING FOR NEW PBP EFFECTOR MUTANTS

Aim: to identify PBPs for new targets

Two RBP mutant libraries were produced. The first one contained alanine-substitutions at any of the known amino acid positions of RBP, resulting in 235 mutants. The goal of using this library was to better understand the role of the different amino acids for the correct functioning of RBP and, consequently, improve the design rules for making new mutants. The library was reconstructed into an E. coli expression host in order to be able to screen for functional RBPs. This library was kept in an organised format of individual clones, which each were individually (and in triplicate) tested for their potential to be induced with ribose. Essentially three categories of behaviour were recorded: (i) those with unchanged ribose induction potential, (ii) those with partly diminished induction, and (iii) those with loss of ribose induction potential. Some 20 positions were found that clearly affected inductibility by ribose. In addition, 3 positions were found which cause a constitutive phenotype.

The second library consisted of some 2 million mutations that would change individual or combinations of 11 different residues that in silico simulations suggested to be necessary for changing substrate binding from ribose to 1,3-cyclohexane-diol. The library was screened for positively reacting mutants using plate assays with 1,3-cyclohexanediol and by fluorimetry. In addition, the library was screened by flow cytometry and fluorescence assisted cell sorting. Various mutations were detected causing total disruption of the signal, or a constitutive expression of gfp. Some 200 mutants have been recovered from the libraries with potentially significant increase of responsiveness to 1,3-cyclohexanediol that are now being tested individually and in independent triplicate assays with a variety of inducers to confirm or refute the change in inducibility.

In addition, a random mutagenesis approach was undertaken to change the specificity of the ribose-binding periplasmic protein (RBP) from ribose to other target chemicals. In this procedure, an error-prone PCR procedure of the rbsB gene (encoding for the RBP) was applied, which introduced mutations at random positions through the gene. The rbsB-mutants library was screened for induction by 6 different chemicals using a robotic system. In total, about 1800 mutants (19 x 96-wells plates) were screened for induction by one concentration of carbamazepine, clyndamaycin, acetominophen, acetylsalisic acid, ibuprofen and diclofenac. Mutations that displayed an induction by one of the tested chemicals were isolated and tested with a range of concentrations. Unfortunately, none of the isolated chemicals produced a reproducible dose-dependent response to any of the tested chemicals.

Construction of a nickel-responsive bacterial reporter based on the periplasmic binding protein NikA was attempted. Although some difference in induction was observed in the absence of NikA periplasmic protein, the performance of the Ni-inducible sensor is not satisfactory. Since the solution studies showed significant binding of Pb(II) by NikA, it was tested whether the system could act as a Pb-inducible sensor. No induction was observed.

To increase screening capacity of the mutant libraries, a flow cytometry analysis and fluorescence assisted cell sorting approach was developed. This permits analysis of batch libraries of up to 10 million cells (mutants) in a few hours and recovery of cells from the library mixture with particular fluorescence characteristics (e.g. increased response to a new effector).

Molecular modeling was used to calculate the binding energies for new substrates in the binding cavity of RBP, and to predict amino acid modifications with lower overall free energy of binding. As these calculations are limited to the binding pocket and its immediate surroundings, methods were sought to better understand the potential implication of amino acid residues in RBP for the process of binding the substrate and the chemoreceptor. Two methods were used to analyze contributions of "conserved" amino acids on RBP functioning. In the first, the probabilities of co-occurrence are calculated of amino acid neighbours in a range of 200 known RBP, with the hypothesis that co-evolving residues play an important role for the protein function. This is calculated using a procedure of "Mutual Information Index" or "Direct Information Index". The second method, probes the cause of the difference between ribose and galactose binding protein (GBP), which both bind naturally to the same chemoreceptor, yet only have some 30-35% conserved amino acid sequence. The question here is whether true new substrate binding can only arise if the protein has a completely "restructured" binding cavity, which is not possible to engineer with a few amino acid substitutions within each one of them. A large number of both RBP and GBP sequences are retrieved in the procedure, which then aligns all those sequences as good as possible for comparison. Subsequently, the program estimates what the amino acid sequence might have been of the last common ancestor between the two protein groups, which might have been a protein capable of both ribose and galactose binding. This sequence can then be synthesised and tested as a platform for new substrate evolution. Alignments have been made which are now being statistically refined to a phylogenetic tree using maximum likelihood analysis. A number of ancestor nodes have been inferred. The analysis indicated that there is a tendency for residues on one side of the protein to strongly co-evolve, for example, because of the importance to bind to the receptor interface. Regions on the outside of the proteins tend to have more variability, probably because there is less constraint on substrate or receptor binding.

Conclusion: further understanding was gained about the functionality of PBPs. Engineering of new PBPs is not straightforward.

APPLICATION TO ENVIRONMENTAL MONITORING

Aim: to establish the link between the signals from PBP-based sensors and target speciation in environmental matrices.

The response of PBPs to targets in real sample solutions will depend on the different physicochemical forms of the target in the sample matrix, i.e. its speciation. To lay the foundations for work with target-PBP systems, and in real systems, theory was developed for the formation and dissociation rate constants for binding of metal ions and organics by soft nanoparticulate complexants. For metal ions, the developed conceptual framework provides a basis for estimation of the kinetic parameters for inner sphere metal complex formation by dispersed soft nanoparticles, with appropriate accounting for the electrostatic properties, the binding site density, and the particle radius. The approach enables intrinsic rate parameters for nanoparticulate complexants to be related to experimentally observed kinetic behaviour of an entire dispersion, and facilitates comparison of rate constants for nanoparticulate complexants with those for simple ligand types. The theory has been successfully applied to description of the kinetics of humic complexation of rapidly dehydrating (copper(II)) and slowly dehydrating (nickel(II)) metal ions. For rapidly dehydrating metal ions, transport limitations and electric effects involved in the formation of the precursor outer-sphere associate appear to be important overall rate-limiting factors. The theoretical concepts are also applicable to metal accumulation at biological interfaces, pending a reconsideration of the boundary conditions imposed at the consuming interphase and an account of soft structure that hinders the transport of metals before actual uptake and/or sorption. The latter points are currently being developed in theoretical works.

Significant advances were also made in the interpretation of dynamic speciation of organic targets as measured by SPME. Samples were considered in which the organic target molecule can be freely dissolved, in various protonated forms, and can undergo a physicochemical association reaction with a site, located either on the surface or within the body of a nanoparticle, or a PBP, to yield a complex. For the example case of the pharmaceutical diclofenac in dispersions of impermeable (silica, SiO2) and permeable (bovine serum albumin, BSA) NPs, it was shown that only the protonated neutral form of diclofenac is accumulated in the solid phase, and hence this species governs the eventual partition equilibrium. On the other hand, the rate of the solid/water partition equilibration is enhanced in the presence of the sorbing nanoparticles of silica and BSA. This feature demonstrates that the NPs themselves do not enter the solid phase to any appreciable extent. The enhanced rate of attainment of equilibrium is due to a shuttle-type of contribution from the NP-species to the diffusive supply of diclofenac to the water/solid interface. For both types of nanoparticulate complexes, the rate constant for desorption (kdes) of bound diclofenac was derived from the measured thermodynamic affinity constant and a diffusion-limited rate of adsorption. The computed kdes values were found to be sufficiently high to render the NP-bound species labile on the effective time scale of SPME. In agreement with theoretical prediction, the experimental results are quantitatively described by fully labile behaviour of the diclofenac/nanoparticle system and an ensuing accumulation rate controlled by the coupled diffusion of neutral, deprotonated and NP-bound diclofenac species.

In addition, strategies were explored to interpret speciation measurements by SPME in the presence of sorbing nanoparticles that penetrate the solid polymer phase. Two ways were found to overcome these interference problems, i.e. the application of a dialysis membrane to prevent nanoparticulate size entities to reach the sensor surface, or the chemical modification of the sensor surface with a negatively charged polyelectrolyte brush that repels the negative silica nanoparticles.

The performance of the microfluidic flow cell was tested in environmental samples using test bioreporters, namely two E. coli strains, E. coli CP38 and E. coli RMF1, expressing the constitutive and the inducible gfp gene, respectively. Both strains were grown using the flow cell using water sampled from a Dutch canal (Delfgauw) as the growth medium, supplemented with ampicillin (50 ug/L) to maintain plasmid and nalidixic acid (10 mg/L) as the compound to be detected by RMF1, the biosensor strain. These experiments explored the possibility of bacterial growth using natural environmental water and the ability to detect the test compound in a natural context. These results showed that E. coli used as a biosensor platform is able to grow on environmental water and that the particulates do not interfere with the flow cell system. Additionally, this experiment supports that the environmental biosensing is feasible.

The liposome-MBP assay was subjected to real-world sample testing including a variety of beer samples and over the counter drugs. The stability of liposomes within all of these complex matrices was confirmed, i.e. no liposome lysis occurred with the exception of a vitamin D3 product that contained soybean oil, which lead to minimal lysis of liposomes. When spiked into these samples, maltose could readily be detected, even in the Vitamin D3 sample. Accurate maltose quantification was possible in a pharmaceutical where maltose was the active ingredient. In other OTC medications and beer samples, the assay response was saturated without further spiking with maltose. This is likely due to the presence of maltodextrins used as an additive during beer manufacturing and in certain OTC drug formulations. Polymers of maltose are known to also be recognised by MBP and would at present be interferences in such applications. These results demonstrated that (a) the liposome-MBP assay is highly accurate, specific and robust, (b) the liposome technology is sufficiently mature for the use with real-world samples, and (c) like any assay, analysis of sample matrix effects is mandatory to define its appropriate analytical applications. Thus, a highly specific, sensitive high-throughput bioassay was developed using sulforhodamine B-encapsulating liposomes for fluorescence signal generation. The platform can now be used for the detection of other analytes taking advantage of other recognition elements of the PBP protein class.

Conclusion: significant advances were made in understanding and predicting the behavior of metal ions and organic pollutants in environmental systems. Notably, the unique reactivity of soft nanoparticles, such as periplasmic binding proteins, towards ionic reactants was identified.
Potential Impact:
The BIOMONAR results have significant potential impact. The various areas of strategic impact are elaborated below.

PROVIDING SOLUTIONS WELL BEYOND STATE OF THE ART IN ENVIRONMENTAL MONITORING
The common PBP biorecognition elements provide a coherent and transparent means to compare and integrate data obtained for the different sensor platforms. The three sensor platforms (surface-immobilised, liposomal, reporter cell) each provide detailed information on various steps of the chain of events ranging from transport of the target from the exposure medium to the biological interface to the ultimate biological impact. That is, the suite of sensors enables the link between the target speciation in the exposure medium and the biological impact to be established on a sound scientific footing via analysis of the complete detection chain. This level of sophistication in the sensing system and in the dynamic approach to data interpretation goes far beyond the state-of-the-art in environmental monitoring. BIOMONAR determines the reactivity and fluxes of target compounds and their mode of interaction with biological and nonbiological interfaces. The approach is flexible and readily adapted to as yet unidentified targets.
A particular focus is placed on strategies for monitoring and data analysis in mixtures of compounds: this is the usual situation in environmental samples, but is currently poorly accounted for. It cannot be assumed that a target has equal biological impact in isolation and in a mixture. As well as developing practical tools for calibration in mixed target systems, BIOMONAR develops rigorous models to describe the impact of mixed targets on the sensor signals.
Furthermore, much greater understanding of PBP function has been established. This knowledge has implications for a wide range of applications from fundamental biology to engineering of PBPs for sensing applications.

EUROPEAN ADDED VALUE BY SUPPORTING EC ENVIRONMENTAL POLICY
The Water Framework Directive aims to protect aquatic systems by setting of appropriate environmental quality standards. Implementation of this policy requires formulation of appropriate parameters to define these standards, and provision of the tools with which they can be measured routinely by relevant end-users. The biosensors developed by BIOMONAR offer a set of tools for in situ monitoring of environmental pollutants. The quantitative data interpretation protocols developed in BIOMONAR help to identify the parameters which are most relevant for predictions of biological impact.

NEW TECHNICAL OPPORTUNITIES FOR THE IMPLEMENTATION BY THE EUROPEAN BIOSENSORS INDUSTRY
BIOMONAR develops a set of biosensor platforms based on a common detection element. The design is flexible and easily adapted for new targets, thus allowing industry to rapidly meet demand for sensors for new targets. The PBPs are stable in the long-term, and can be easily transported. The commercial beneficiaries helped to make early design decisions, adding the importance of eventual considerations for large-scale manufacture, robustness, quality control, and cost (coupled with basic market analysis).
BIOMONAR has enabled collaborative development of a flow cell for biosensing (SME partners LioniX and MicroDish). The BIOMONAR outcomes are considered successful with future applications in the water industry and in threat or pollutant detection which aligns well with the project aims. The MDCC culture system combined with a flow cell has advantages in this respect: it facilitates exposure of the microorganisms to water for sensing, removal of waste products and avoidance of contaminating/confounding materials such as agar (with contaminants that may give false positives or confound the assay).
MicroDish is also pursuing exploitation of flow cells for the sensing of nutrients within other areas, particularly in the context of regrowth (fouling) potential within drinking water pipes. Whilst this specific area connects most strongly with local (Dutch) markets, MicroDish foresees an expansion in water quality testing and biosensors more generally within other applications, e.g. antibiotic or mutagen detection. The flow cell has been developed as a product and is offered on the MicroDish web site (www.microdish.nl).
The work with the published PBP for Ni may find opportunities for exploitation by partner Applikon in the mining industry in which the measurement of Ni concentrations in effluent streams (waste water and process water) is of high importance.
The platforms are translatable to markets outside the environmental domain, e.g. life sciences (pharmaceutics, medical diagnostics). This latter aspect is especially facilitated by the role of SME beneficiary LioniX as a high technology enabler. The access to US end-users and markets (see following paragraph) via beneficiaries Cornell University, LioniX, and Applikon will raise the profile of the European-led research and place the European biosensor industry in a favorable position on an international level.

PARTICIPATION FROM UNITED STATES TO ADD TO STD EXCELLENCE AND INCREASED IMPACT
Via direct involvement as a beneficiary (Cornell University, CU) BIOMONAR has active US participation. CU brings expertise in liposomal sensor platforms which add to the scientific and technological excellence of the project. CU is well established amongst the network of US nanotechnology researchers via National Science Foundation supported research centers. CU’s dissemination activities in the US increase the international impact of BIOMONAR’s results.
Beneficiary Applikon, as a member of the Metrohm group, has access to the US market in environmental analysis. Its US headquarters have laboratories for dedicated method development and applications support. Metrohm also runs regular training sessions for clients and is a frequent exhibitor at international trade shows to maximise the exploitation of market opportunities worldwide. Accordingly, the technology developed by BIOMONAR shall be directly integrated into Applikon’s international network. Beneficiary LioniX also has a relatively large customer base in the US, which provides opportunities for commercialisation in the US market.
List of Websites:
http://www.sdu.dk/Om_SDU/Institutter_centre/Ifk_fysik_og_kemi/Forskning/Forskningsgrupper/Raewyn_M_Town/BIOMONAR.aspx?sc_lang=en

Coordinator: Raewyn M. Town, email: raewyn.town@sdu.dk