Combined dielectrophoretic and impedance system for on‐chip controlled bacteria concentration: Application to Escherichia coli

The present paper reports a bacteria autonomous controlled concentrator prototype with a user‐friendly interface for bench‐top applications. It is based on a microfluidic lab‐on‐a‐chip and its associated custom instrumentation, which consists of a dielectrophoretic actuator, to preconcentrate the sample, and an impedance analyzer, to measure concentrated bacteria levels. The system is composed of a single microfluidic chamber with interdigitated electrodes and an instrumentation with custom electronics. The prototype is supported by a real‐time platform connected to a remote computer, which automatically controls the system and displays impedance data used to monitor the status of bacteria accumulation on‐chip. The system automates the whole concentrating operation. Performance has been studied for controlled volumes of Escherichia coli samples injected into the microfluidic chip at constant flow rate of 10 μL/min. A media conductivity correcting protocol has been developed, as the preliminary results showed distortion of the impedance analyzer measurement produced by bacterial media conductivity variations through time. With the correcting protocol, the measured impedance values were related to the quantity of bacteria concentrated with a correlation of 0.988 and a coefficient of variation of 3.1%. Feasibility of E. coli on‐chip automated concentration, using the miniaturized system, has been demonstrated. Furthermore, the impedance monitoring protocol had been adjusted and optimized, to handle changes in the electrical properties of the bacteria media over time.


Introduction
In the last few years, the electrical properties of cells and pathogens have been used to explore new methods of manipulation and characterization, such as dielectrophoresis (DEP) [1] or impedance analysis (IA) [2,3]. For instance, DEP has been recently used to control stem cells to form embryonic bodies in shorter time [4] and Fatoyinbo et al. [5] have measured biophysical parameters of cells (cytoplasmic conductivity, membrane conductivity, and cell-wall conductivity) by analyzing its cells' DEP behavior. Moreover, IA was also advantageous to detect ovarian cancer cells SKV3 [6] or to detect insulin levels in blood serum [7] so as to diagnose diabetes or trauma. We present a miniaturized and compact specific solution to concentrate bacteria in a controlled manner using a fully automated instrument combining DEP and IA.
Bacteria concentration is a time-consuming procedure in regular microbiology laboratory practices that involves cellculturing processes [8,9] to obtain a significant sample. This could be improved by using DEP as a means of concentration in tiny fluidic spaces. DEP refers to the force experienced by a particle inside a nonuniform electric field [10,11] and is a convenient, rather selective, handling method that has been applied in many biological fields and in lab-on-a-chip (LoC) devices [12][13][14]. An example of this is the work reported by Lapizco-Encinas et al. [15], where several types of bacteria in water were concentrated and separated by DEP induced by insulator-based structures (iDEP), or in the paper presented by Braff et al. [16], where bacteria were successfully DEP trapped in PMMA constructs. DEP selectivity has also been repeatedly reported as a benefit for sample preparation, since it allows isolation of the desired cell or pathogen based on their electric and geometric properties [17][18][19]. As an example, Moon et al. [19] used DEP to separate and detect circulating tumor cells, whose size and resistance to filtering shear stress presented significant differentiating properties, from regular blood cells. This also becomes an advantage in the case of environmental samples, where soil particles with the same bacteria size are also present and could not be eliminated by filtration or centrifugation. This has also been solved by using DEP [20], taking advantage of its selectivity by cell electrical properties. Hence, we used DEP here for concentration purposes.
On the other hand, current bacteria-detection protocols are expensive in terms of equipment and time, typically requiring several days to obtain results [21,22]. Techniques such as pathogenic-specific antibody-coated magnetic beads [23,24] or hybridization of DNA fragments of bacteria [25] have shown to improve the analysis time down to several hours, but they still need complex equipment. This could be improved by using IA. Impedance frequency dependence, which is related to the electrical conductivity and permittivity properties of the material, was reported as an effective solution to characterize cells and their behavior, also in LoC devices [26]. Some publications have reported the use of IA technique to control bacterial growth or to detect its presence [27]. One example of such work is the paper presented by Dweik et al. [28], where bacterial presence was rapidly detected by measuring the antibody/antigen bonding using IA in the 100 Hz to 10 MHz range. Another example is the work of Grossi et al. [29], where the quantity of bacteria during a culture process was detected by impedance measured at 200 Hz using a sinusoidal signal with a 50 mV amplitude.
The combination of DEP and IA [3,30] in a single equipment based on LoC and microfluidic technologies allows to develop a practical bench-top device. In recent years several biosensors and applications aiming for the successful combination of both techniques have been presented. Hamada et al. [3] presented a bacterial detection device combining both positive DEP and negative DEP with dielectrophoretic impedance measurement. The biosensor relied on a pair of interdigitated electrodes (IDEs) for separate DEP concentration and dielectrophoretic impedance measurement, while using commercial devices to operate the application. The cellular solution conductivity varies through time, which affects the impedance measurement, which has not been considered, and measurement instability produced by the magnitude of DEP voltage has been reported. Dastider et al. [30] have designed an impedance biosensor for the specific detection of Esherichia coli O157:H7 combining DEP and IA techniques at 2 L/min flow rate, which is relatively low. This work used different IDEs for cellular separation and detection purposes. The detection IDE was functionalized with polyclonal anti-E. coli antibodies for specific detection of E. coli O150:H7, removing versatility of the device. Moreover, the presented results for cells' concentration detection, based on impedance measurements, did not consider the solution conductivity variations, as well as the influence of DEP voltages on the impedance measurement.
Our work presents a completely customized equipment for a quick and easy way to concentrate bacteria with DEP technique at relatively high flow rates [31,32], while monitoring its concentration by means of IA technique in a real-time scenario. It addresses the issues associated with the combination of these techniques by simplifying the equipment but also by trying to solving some issues generally avoided, to the best of our knowledge in other scientific works.
The device, with its main components, is presented in Fig. 1. It is composed of a customized electronic module and an LoC. The flowing bacteria sample is preconcentrated through the generated DEP generation and concentration is measured through IA monitoring, with a four-electrode sensor topology, embedded on a single microfluidic chamber. The electronic module is supported by a real-time platform for continuous concentration monitoring, connected to a remote computer through a standard Ethernet connection, which enables the system configuration and data display. First, it allows automated functionalities, such as multiplexing signals between the DEP generator and the IA analyzer in the microfluidic chip, in order to avoid DEP voltages disturbance of IA measurement, and auto-scale of the electronic instrumentation gains when necessary, for better signal acquisition. Second, it is connected to a remote computer with a user-friendly front-end user panel, where the system user can configure the experiment variables, such as measurement time for signal multiplexing, signal operation frequency, and output gain, while displaying the impedance measurements related to actual bacteria concentration level.
The solution presented controls, in an automated way, the bacteria concentration, and monitoring process and has been validated for E. coli, which presents pathogenic variants that cause morbidity and mortality worldwide [33]; therefore being a topic of interest. E. coli is one of the main antimicrobial-resistant pathogens for healthcare-associated infections reported to the National Healthcare Safety Network [34], being the primary cause of widespread pathologies such as significant diarrheal and extraintestinal diseases [33] or urinary tract infections [35]. Furthermore, E. coli can be found as a bacterial food contamination [21] and causes avian colibacillosis, one of the major bacterial diseases in the poultry industry and the most common avian disease communicable to humans [36].
The aims of our study are (i) to prove the feasibility of DEP generator and IA analysis combination for controlled concentration using a single equipment together with a single microfluidic chip; (ii) to establish a protocol for autonomous concentration procedure; and (iii) to develop a complete electronic equipment with an electronic instrumentation, embedded software control, and user interface for a complete autonomous and reliable bacteria concentrator device, based on DEP generator and IA technique.
This novel, specific device has been proven as a robust and reliable automated system and protocol for bacteria controlled concentration. It will provide the scientific community with a rapid tool for bacteria presence detection, by avoiding previous slow preparations in preconcentration and culture processes, reducing procedure times for a faster diagnosis and treatment.

The dielectrophoretic effect
DEP [11] defines the movement of an electrically neutral particle when a nonuniform electric field is applied. If the particle is considered homogeneous and isotropic and is polarized linearly, then the dielectrophoretic force is defined by Eq. (1) [37,38]-where V is the volume of the particle; E is the electric field; and ␣ is the effective polarizability, which is defined by the expression (2): where ε 0 and ε m are the vacuum permittivity and the medium permittivity, respectively, and F CM is the Clausius-Mosotti factor. The F CM sign describes the force direction: if F CM is positive, the particle is attracted to an electrical field maximum (which is called positive DEP or p-DEP) and if negative, to an electrical field minimum (negative DEP or n-DEP). Hence, the DEP force allows control of the movement of a particle by varying the applied signal, changing the electrode shape, placing dielectric structures, or modifying media properties. Here we used a pair of interdigitated gold electrodes to preconcentrate E. coli cells. In order to define the suitable trapping frequency, an E. coli geometry model is considered. This bacterium is approximated to an ellipsoid shape with two dielectric layers [10], which modifies the Clausius-Mosotti factor expression: where ε m is the medium permittivity; ε p is the particle permittivity; and A i is the depolarization factor of an individual ellipsoid axe (i = x, y, z), where e is the eccentricity that involves the ellipsoid dimensions (where "b" is the height and "a" the width): The representation of expression (3) showed that the optimal frequency to manipulate E. coli cells by p-DEP is 1 MHz as we know from previous studies of the group [39,40]. This frequency was therefore chosen for the preconcentrating stage.

Impedance and available measurement methods
The bioimpedance [41,42] can be measured as the voltage response of a biological material to the application of a current bias signal, and is defined by the Ohm's law. The methods of impedance measurement are classified by the number of electrodes used: two-, three-, and four-electrode methods. The difference among methods resides in how bias current signal is applied and how the sensor voltage signal response is read. A two-electrode configuration is the basic topology, defined by the working electrode, where bias signal is applied, and the reference electrode, which tracks the bias current signal and provides a reference for the voltage measurement. However, as the current bias signal flows through the reference electrode, this topology entails some problematic behavior as the voltage reference is distorted due to electrode polarization. In order to avoid this effect, the three-electrode topology adds a third electrode to supply the bias current signal, while the reference electrode remains as a voltage reference.
Although this is an improvement, the impedance measurement with this topology can be distorted due to the working electrode impedance polarization, as the current bias signal is directly applied where the single-ended voltage measurement signal is read. In this paper, the four-electrode method was used ( Fig. 2A, electrodes ER1, ER2, ECI1, ECI2), which was composed of two current injection electrodes and two voltage reading electrodes, as this electrode topology avoids electrode polarization distortion in impedance measurement due to a complete differential voltage measurement [43].

Microfluidic chip design and fabrication
The designed microfluidic chip design is showed in Fig. 2B. This had two IDEs, which were shared between the DEP generator and impedance analyzer readout electronics, and two lateral electrodes, which were used to inject the necessary current so as to obtain the impedance measure. The IDEs were formed by 40 pairs of 6 mm × 50 m electrodes separated by 50 m. The lateral electrodes (6 mm x 300 m) were separated by 200 m from the interdigitated ones. These electrodes were attached to a PDMS microfluidic chamber with a volume of 4.8 L. The fabrication of the microfluidic chips followed a protocol based on three main steps: microchannel molding, electrode fabrication, and microfluidic chip bonding.
First, SU8 50 (MicroChem) masters were fabricated over glass slides (Deltalab) and PDMS replicas were created. In order to do this, the glass slide was cleaned and activated by Piranha attack for 15 min. Then a 50-m-high SU-8 50 (MicroChem) was spun over the slides. They were later exposed and developed so as to obtain the desired microchannels. Afterward, a 10:1 ratio of PDMS prepolymeric solution (Dow Corning Sylgard184) was mixed, degassed, and poured into the mould to replicate the microchannels. Finally, the PDMS was cured at 70°C for 1 h and peeled from the master.
Second, in order to fabricate the microelectrodes over a set of the LoC sealing glass slides (Deltalab), a lift-off soft lithographic process was used. AZ 1512 (AZ Electronic Materials) photoresist was chosen as a sacrificial layer in this process. First, a Piranha cleaning procedure was performed over the glass slides. Later, AZ 1512 was spun on these slides, exposed, and developed. Then, two metal layers, 20 nm of Ti and 80 nm of gold, were vapor-deposited sequentially. The electrode structures were finally obtained by removing the AZ photoresist.
As a final microfluidic chip fabrication step, once the PDMS replica and the microelectrodes were finished, both parts were assembled to create a sealed structure. First, the surfaces were cleaned using an oxygen plasma process. Hereinafter, the PDMS channels were aligned and attached to the glass substrate. Later, cables were welded to each electrode pad using conductive silver paint and mechanically strengthened using an epoxy glue mix, later cured at room temperature for 60 min. Finally, two NanoPort Assemblies were attached in order to set the inlet and outlet fluidic connections.

Dielectrophoretic signal generator electronics
The designed dielectrophoretic signal generator module is presented in Fig. 3. Four channels with different phases (0°, 90°, 180°, 270°), which could be connected in different ways to the electrodes DEP1 and DEP2, were defined so as to add versatility to the board in dielectrophoretic terms. Each channel generates a sinusoidal signal at 1 MHz with variable output voltage from 1 to 15 Vpp (peak to peak) to control the DEP force intensity, and is composed of three modules: (i) A square signal generator that provides four shifted and frequency stable signals (A); (ii) a power driver that boosts the signal from the previous module so as to activate the following stage (B); (iii) a class E amplifier, which generated the DEP sinusoidal signal (C).
The first module, the square signal generator, is based on the LTC6902 (Linear Technology). The synchronized outputs are shifted 1 = 0°, 2 = 90°, 3 = 180°, and 4 = 270°, respectively. Their output frequency is selectable by an external resistor (R SET ), following the Eq. (7) where (N = 10 is related to frequency working range and M = 4 is the number of active outputs), LTC6902 outputs have a supplying limit of 400 A. Hence, a power driver is used to increase the current capabilities. An UCC27424 (Texas Instruments) is chosen for this purpose. This device boosts the current levels of the input signal up to 4 A, which is sufficient current to drive the final module. This module is a class E amplifier that generates the necessary sinusoidal signals to apply DEP. This amplifier configuration generates high-frequency signals with stable output voltages [44][45][46] by injecting a square high current control signal.
The class E amplifier is composed of an inductor Le, a capacitor Ce, and a resonance tank formed by the inductor L and the capacitor C. The L-C tank generates a 1 MHz sinusoidal signal by using the 1 MHz square signal from the previous modules. The circuit parameters (Le, Ce, C, L) were configured in function of the necessary output frequency, the output impedance, and the equivalent resistance of the microfluidic chip. Thus, four independent channels perfectly synchronized at 1, 2, 3, and 4 are obtained.

Impedance analyzer electronics
A fully customized electronic circuit was specifically designed to carry out the IA experiments. As previously stated, the microfluidic device impedance measurement is based on the four-electrode topology. A four-electrode method is composed of two current injection (ECI1 and ECI2) electrodes and two voltage reading (ER1 and ER2) electrodes. The main advantage of this system is that electrode impedances are cancelled, obtaining a more reliable measure. The circuit specifications were defined taking into account the sample media impedance, and considering the microfluidic device characteristics and the frequency ranges where bacterium could be discriminated [47,48].
The impedance analyzer architecture consists of two modules: the current injection module (CI in Fig. 4B) that provides a frequency configurable voltage sinus signal (V RS ) that is converted to a current signal (voltage-to-current converter circuit) to bias/drive the current injection electrodes ECI1 and ECI2. An instrumentation amplifier (IA) senses the differential voltage between the reading electrodes ER1 and ER2 (V IS ).
The second module, signal digitalization and postprocessing (SDPP in Fig. 4A), calculates the impedance measurement through the voltage signals provided by the previous stage, and automatically controls the hardware configuration. This module is composed of a real-time platform sbRIO9632 (National Instruments) with an embedded software for data processing and hardware control. A signal conditioning stage converts voltage signals from a bipolar single-ended signal to a unipolar differential signal to be processed by an analog-to-digital converter (ADC).
The first module (CI), current Injection, is based on a signal generator AD9833 (Analog Devices) and a voltage-tocurrent converter. The signal generator AD9833 provides a stable voltage signal with a wide variable frequency range, 0 to 12.5 MHz, which is controlled by an SPI communication protocol. The voltage-to-current converter is a modified Howland cell based on AD8066 (Analog Devices) operational amplifiers (OA1 and OA2) that guarantee a wide bandwidth and a high slew rate while maintaining a low spectral noise and a low offset performance. The Howland cell uses R SET and the reference signal (V RS ) amplitude to define a stable current signal (I OUT ) at the output of the circuit (8) regardless of the connected load.
The differential voltage between ER1 and ER2 electrodes is acquired by means of the instrumentation amplifier (IA) INA163 (Texas Instruments), which allows a wide bandwidth with a low spectral noise and low total harmonic distortion. The measured voltage (signal V IS ) is related to the differential voltage between the reading electrodes (ER1 and ER2), G being the instrumentation amplifier gain. This V IS signal is then adapted and processed by the SDPPM module in order to extract the impedance of the media.
The second module (SDPP), signal digitalization and postprocessing, consists of a 12-bit, dual, low-power ADC ADC12D040 (Texas Instruments), capable of converting both analog input signals at 40 MSPS simultaneously. Twelvebit resolution does not represent a significant drawback in the final system resolution, as V RS is scaled to the full-range ADC analog input and the system provides a real-time gain auto-scale for the instrumentation amplifier gain G. The analog inputs are converted from single ended to differential with a differential amplifier (DA) AD8138 (Analog Devices), with a high slew rate with low distortion and input noise. The impedance measurement is carried out with a digital lock-in based on the frequency response analyzer (FRA) approach [49]. The FRA is a real-time mathematical processing system, embedded in the 400 MHz microprocessor from the real-time platform sbRIO9632, which adopts sine and cosine signals related to V RS , and by means of two multipliers and a filter stage, the real (V REAL ) and imaginary (V IM ) components values (10) of the measured signal V IS are obtained (Fig. 4C).
The key measurement in our work is the impedance magnitude (|Z CELL |) (Eq. (11)). This value is calculated based on the V REAL and V IM components.
For accurate hardware control, the real-time platform sbRIO9632 has a FPGA Spartan-3 (Xilinx), which allows us to provide steady clock signals, needed on the instrumentation, which can be automatically adjusted, allowing complete realtime control of the chip electrodes multiplexing. As stated in Section 2.1, the microfluidic chip had two IDEs, which were shared between the DEP generator and the IA readout electronics. When an IA measurement was done the DEP generator was disconnected, suspending the trapping process. If this process was not properly timed, bacteria already trapped would be lost in the process, so the real-time control allowed an optimized timing process minimizing the bacteria loss. Moreover, the disconnection of DEP voltage signals contributes to a better bacterial concentration monitoring avoiding distortion and instability on the IA measurement. The IA process had been programmed and tested to last for a period of the applied current signal, plus 1 ms for multiplexor switching times and stabilization. In addition, real-time platform allows complete parallel signal acquisition for all the frequency ranges, and the development of an embedded hardware control, such as R SET multiplexed auto-scale, instrumentation amplifier gain G auto-scale, and signal generator automatic frequency sweep. This real-time embedded hardware control represents the basic features of an automated and complete FRA approach. The real-time platform allows the system configuration and data display, with a userfriendly front-end user panel (Fig. 4C), by means of an external computer connected to the platform with a standard Ethernet connection.

Bacteria culture
A laboratory sample formed by E. coli 5K strains (genotypes: F − , hdsR, hdsM, thr, thi, leu, lacZ) was grown overnight in 10 mL of Luria-Bertani broth at 37°C. The achieved cell concentration (estimated by performing viable cell counts in LB agar) was 10 9 cells/mL. Then, the E. coli culture was pelleted by centrifugation at 5000 rpm for 5 min. Bacteria were then resuspended in 10 mL of DI water. Finally, the samples were diluted (final concentration of 2 × 10 7 cells/mL) and frozen in 1 mL collecting tubes for storage purposes.

Conductivity measurements
As E. coli concentration was measured by means of IA, bacteria samples' conductivity was monitored while in vitro a major factor in IA reliability, using a commercial bench-top conductivity meter Corning 441. Prior to the experiments, bacteria samples were diluted in DI water with a conductivity of 8.2 × 10 −5 S/m, but the conductivity of the samples at the time of the experiment, after the process of storage and thawing, was subject to variations. A sample conductivity analysis had to be done at the beginning of the experiment. The conductivity meter probe was calibrated and introduced into the 1 mL collecting tubes until it was totally covered by the bacteria sample.

Experimental setup
The microfluidic chip was placed over an inverted microscope stage (Olympus IX71) connected to a digital camera (Hamamatsu Orca R2). Moreover, the microfluidic chip was connected to a six-port manual valve (Valco). This valve was also connected to a 5 mL syringe filled with DI water (8.2 × 10 −5 S/m) and placed on an infusion micropump (Cetoni NEMESYS) so as to obtain a continuous flow rate. The microfluidic chip's gold electrodes were connected to the custom combined DEP and IA device.

Results and discussion
The designed combined device was validated by a series of E. coli concentration and impedance measurement tests. First of all, so as to validate the system as an autonomous bacteria concentrator, and study the effect of real-time monitoring by means of IA measurement, E. coli was continuously injected through the valve to the microfluidic chip at a 5 L/min flow rate, and preconcentrated by DEP by two counter-phased signals of 15 Vpp. In addition, the impedance module was programmed to proceed with a 3-ms impedance measurement every 30 s meanwhile DEP module was continuously trapping bacteria. As a first approach, the conductivity of the solution has not been corrected to study the effect its variations over time on the IA measurement. Different tests for different applied current signal frequencies were done. Taking into consideration the electronics and microfluidic chip design, impedance measurement was performed at continuous alternating current of 10 A in the 500 Hz to 5 kHz frequency range, where bacterium could be discriminated [47,48] and evaluated using 100 Hz spaced sampling intervals.
The measured bioimpedance (|Z|), depicted in Fig. 5A, clearly shows a decreased impedance as the trapped bacteria concentration increases, regardless of the frequency. This behavior was clearly explained by the conductivity changes taking place in bacteria samples over time. Measured conductivity was recorded periodically in-tube during the experiments showing a rise from 0.5 × 10 −3 to 2.5 × 10 −3 S/m until it stabilized. This conductivity change, related to the original sample prior to the trapping process, may be translated into a theoretical variation in impedance. This estimated impedance, related to measured bacteria sample in-tube conductivity, was calculated considering the microfluidic chip electrodes' geometric characteristics. In Fig. 5B impedance variation (⌬ |Z|) through time for the measured on-chip impedance, during the trapping process, and for the estimated on-tube impedance are shown.
Results show a very similar behavior through time of both measurements. Acquired data variations through time for the first 40 min, before conductivity stabilization, were −52.41 ⍀/min for measured impedance and −54.79 ⍀/min for conductivity related impedance, which confirms that the first impedance measurements are related to bacteria sample conductivity rather than trapped bacteria concentration, underlining the need for a media conductivity correcting protocol.
A 2D finite element method based study with Multiphysics software (Comsol) further shows the dominating effect of sample conductivity changes on the bioimpedance measurements when left uncontrolled. E. coli 5 K physical and electrical properties were defined for the different model layers ( wall = 0.68 S/m, ε r_wall = 74, membrane = 5 × 10 −8 S/m, ε r_membrane = 9.5, cytoplasm = 0.19 S/m, ε r_cytoplam = 49.8). Then different medium conductivities were defined, as well as the applied potential to the external lateral electrodes. Current conservation and an initial state of potential 0 were applied for all the layers. Afterward, an adaptive physical controlled and extra fine mesh was applied. Finally, a frequency domain analysis at 1.7 kHz was performed. Thus, surface current density (ec.normJ) of bacteria was obtained (Fig. 5C andD). From the analysis of the obtained simulations, we could assure that in case of a single bacteria diluted on a buffer with a conductivity which varies from 0.5 × 10 −3 to 2.5 × 10 −3 S/m, current density is 99.9% located outside the bacteria. Hence, measured impedance is totally related to sample buffer conductivity rather than bacteria concentration (Fig. 5C). Controlling buffer conductivity to be stable and at the levels of Milli-Q water, around 8.2 × 10 −5 S/m, current density is mainly located in the cell membrane (Fig. 5D) and impedance variation related to the quantity of trapped bacteria.
Hence, when the cells' media is not controlled by cleaning processes, impedance variations are strongly related to changes in the conductivity of the media due to bacteria [50,51]. To solve this issue, which is not confronted in other works to the best of our knowledge, an automated periodic cleaning process was implemented as part of the device working protocol assuring a reliable impedance measurement.
In the resulting protocol, the microfluidic chip was first filled with Milli-Q water media to obtain the threshold impedance measurement. Afterward, a 50 L sample of E. coli was injected through a controlled valve to the microfluidic chip and trapped by DEP forces while flowing continuously at 10 L/min, higher flow rate compared with other solutions for DEP and IA combination, such as 2-4 L/min [30]. After each 50 L sample of bacteria was injected into the channel, 50 L of Milli-Q water, with a specified conductivity of 8.2 × 10 −5 S/m, was automatically injected at 10 L/min to ensure a steady media conductivity for the impedance measurement. Once the Milli-Q water was injected, the impedance electronic module was activated and the DEP generator deactivated by means of multiplexor. Four contiguous impedance measurements were performed each time in order to evaluate precision. Afterward, another 50 L sample of E. coli was injected and the process repeated until all the samples were injected. So, the impedance measurement is always performed after each 50 L bacteria sample was injected, trapped, and cleaned.
The whole process was performed to scan the 500 Hz to 5 kHz IA frequency range each 100 Hz. The DEP was generated by applying two 15 Vpp counter-phased signals through the IDEs. The results of the experimental impedance measurements for three frequencies (500, 1700, and 5000 Hz) are depicted in Fig. 6.
Results are depicted as the increment (⌬ |Z| = |Z| − |Z 0 |) between the different impedance magnitude measurements for every bacteria sample injected (|Z|) and the initial media impedance magnitude measurement (|Z 0 |). Figure 6A depicts ⌬ |Z| measurements through time for the initial and final frequency value, 500 Hz and 5 kHz, respectively, as well as the 1.7 kHz frequency ⌬ |Z| measurements, which seem to be more sensitive and reliable with an accuracy error of less than 2% of bacteria concentration with a correlation of 0.988. Precision can be evaluated with the coefficient of variation, which is the SD of the four experiment repetitions divided by the mean value of the four repetitions' measurement. The mean value of the coefficient of variation is 3.1% on the whole range, although the device is more precise for lower bacteria concentration levels where the coefficient of variation is below 3%. Thus, steady and sensitive ⌬ |Z| measurement at different frequencies, which is bacteria dependent, was observed. Furthermore, bioimpedance control of the achieved sample concentration showed a reliable sensitivity for the protocol including a bacteria-cleaning step. The controlled and steady low media conductivity microenvironment solves issues regarding overall system viability.
The DEP module had a proven trapping efficiency of 85.65 ± 1.07%, for a single 50 L bacteria sample injected at continuous flow of 10 L/min, by measuring the escaped and the collected bacteria of a single load by cytometric analysis [40]. Although the whole process trapping efficiency had not been tested, each sample load was estimated to increment the bacteria concentration 2 × 10 8 bacteria/mL inside the microfluidic chip. Figure 6B depicts the ⌬ |Z| measurements for each bacteria concentration increment (bacteria/L) when 1.7 kHz frequency is applied. However, our main goal was to verify that the process of bacterial concentration while monitoring the concentration is feasible, as it has been proved. The measured impedance values were related to the quantity of bacteria concentrated with a correlation of 0.988 and a coefficient of variation of 3.1%, avoiding distortion and instability related to undesired effects such as media conductivity variations and DEP voltage interferences.

Concluding remarks
Here we describe a novel device and automated protocol, based on DEP and IA, to concentrate bacteria in benchtop setups in a controlled manner. The system consists of a microfluidic chip, with integrated electrodes, and its associated custom instrumentation electronics. It performs bacteria injection, trapping, cleaning, and continuous short-time impedance measuring while achieving the desired levels of concentration. As a proof of concept, it has been applied to concentrate E. coli and to automatically monitor its concentration. The electronic apparatus was validated using a microfluidic chip with four integrated gold electrodes specifically designed for the application. The automated system was tested by trapping and measuring samples of E. coli 5K at a concentration of 2 × 10 7 cells/mL. Concentration and realtime detection of the trapped bacteria inside the microfluidic chip were proven, working a high flow injection rate, up to 10 L/min, for different buffer conductivities [31,32]. Bacteria media conductivity, and its variability, was demonstrated to be a challenging issue when monitoring concentration by means of IA. An automated protocol integrated in the overall system was proposed to solve this problem, strengthening the system versatility and robustness. Before each measurement, the designed system cleans the bacteria samples periodically, while trapped on the microfluidic chip, with Milli-Q water at a controlled conductivity of 8.2 × 10 −5 S/m. To our best knowledge, this proposed system is a useful tool to solve some current microbiology laboratories shortcomings. Bacteria can be concentrated to given specifications while performing analytical procedures. The development of LoC-based equipment, removing the need of huge and expensive devices, is an important research field aiming for smaller systems with better functionalities, such as the integrated application specific integrated system stimulator for electrokinetically driven microfluidic devices presented by Gomez-Quiñones et al. [52]. Nowadays, electronic technology allows further miniaturization of devices such as our concentrator. A SOI technology such as XTO18 from XFAB would be suitable to combine digital instrumentation and class E amplifiers inside a unique chip. However, some drawbacks must be considered when integrating the full system into the LoC device, as it would either increase disposable cost or reduce applicability due to possible contaminations. Still, the simplicity of the presented microfluidic device and the development of the custom electronics on a single application specific integrated system, along with an automated procedure protocol, pushes toward the development of robust and reliable LoC automated bacterial concentrator relying on DEP concentration and IA monitoring.
Plan 2008-2011, Iniciativa Ingenio 2010, Consolider Program, CIBER Actions, and financed by the Instituto de Salud Carlos III with assistance from the European Regional Development Fund. This material is based upon work supported by the Botín Foundation, Santander, Spain.