**Characterization of the Dielectrophoretic Response of Di**ff**erent Candida Strains Using 3D Carbon Microelectrodes**

### **Monsur Islam 1,2, Devin Keck 1, Jordon Gilmore 1,3 and Rodrigo Martinez-Duarte 1,\***


Received: 3 February 2020; Accepted: 26 February 2020; Published: 28 February 2020

**Abstract:** Bloodstream infection with *Candida* fungal cells remains one of the most life-threatening complications among hospitalized patients around the world. Although most of the cases are still due to *Candida albicans*, the rising incidence of infections caused by other *Candida* strains that may not respond to traditional anti-fungal treatments merits the development of a method for species-specific isolation of *Candida*. To this end, here we present the characterization of the dielectrophoresis (DEP) response of *Candida albicans*, *Candida tropicalis* and *Candida parapsilosis*. We complement such characterization with a study of the *Candida* cells morphology. The *Candida* strains exhibited subtle differences in their morphology and dimensions. All the *Candida* strains exhibited positive DEP in the range 10–500 kHz, although the strength of the DEP response was different for each *Candida* strain at different frequencies. Only *Candida tropicalis* showed positive DEP at 750 kHz. The current results show potential for manipulation and enrichment of a specific *Candida* strain at specific DEP conditions towards aiding in the rapid identification of *Candida* strains to enable the effective and timely treatment of *Candida* infections.

**Keywords:** characterization; dielectrophoresis; carbon electrodes; three-dimensional (3D); diagnostics; *Candidiasis*

### **1. Introduction**

*Candida* species are one of the most prevalent fungal pathogens in hospitals around the world. In the United States alone, 5%–10% of hospitalized patients will acquire a nosocomial infection and 80% of such infections are caused by *Candida* species [1]. As early as 1995, *Candida* species became recognized as the fourth most common cause of nosocomial bloodstream infections in the United States, and most recently reported as the 3rd most common cause of nosocomial bloodstream infections in the intensive care unit (ICU) [2]. Concerningly, nosocomial bloodstream infections from *Candida* have a crude mortality rate of 39% overall, and this figure can be as high as 47% for patients infected in the ICU [2]. More than 17 different *Candida* species have been identified as responsible for invasive candidiasis (IC), an umbrella term referring to various severe diseases resulting from *Candida* infection [3]. While *Candida albicans* remains the most frequently isolated *Candida* strains from infected blood [4], the incidence of the infections caused by other species has increased significantly worldwide. For example, a survey in European countries showed that around 50% infection was caused by *Candida albicans*, whereas incidence rates were 14% for each *Candida glabrata* and *Candida parapsilosis*, 7% for *Candida tropicalis*

and 2% for *Candida krusei* [5]. In Chile, the most frequently isolated non-*albicans* species was *Candida parapsilosis*, followed by *Candida tropicalis* and *Candida glabrata* [6]. The emergence of non-*albicans* species as pathogens is concerning because many of them do not respond to conventional anti-fungal therapy, which are generally targeted for *Candida albicans*. For example, *Candida tropicalis* is less susceptible to fluconazole, a common anti-fungal medication, when compared to *Candida albicans* [7]. Hence, with an increased incidence of infections with different *Candida* species, there is a need for a method that allows for rapid identification of the *Candida* species, so that timely measures can be taken towards species-specific treatment of *Candida* infections.

Dielectrophoresis (DEP) is a technique that offers the potential for sorting different *Candida* species in a label-free fashion towards a rapid and affordable assay. DEP is a relatively simple procedure that works by exploiting the specific response of different cells to an electric field gradient [8–12], and has been used for the manipulation, separation, and enrichment of many bioparticles that include bacteria and other bloodborne pathogens [13–20] including *Candida albicans* [21–27]. The fact that DEP has been demonstrated in the sorting of cells featuring minor observable differences between them [9,28–30] encourages the study of DEP to isolate specific *Candida* strains. However, till date, no DEP characterization of *Candida* strains other than *Candida albicans* is available. Hence, there is a knowledge gap preventing the wider use of DEP as a method to sort *Candida* strains. Methodical characterization of the DEP response of *Candida* strains can enable the use of different DEP platforms towards a more rapid way to identify the type of *Candida* causing an infection and an informed approach to combat it. For example, specific *Candida* strains can be isolated and enriched from a dilute sample in a timely manner in a DEP-based sample preparation protocol previous published by the authors [8], which can increase sensitivity of common detection techniques [31].

In this work, we present the morphological characteristics and a first study on the DEP response of three different *Candida* strains: *Candida albicans*, *Candida parapsilosis*, and *Candida tropicalis*; which are three of the most frequently isolated *Candida* strains from infected samples. We used 3D carbon microelectrode arrays to obtain the results presented here due to their improved performance over more traditional planar electrodes [13,32,33].

### **2. Materials and Methods**

### *2.1. Cell Culture and Sample Preparation*

*Candida albicans* (ATCC 18804), *Candida parapsilosis* (ATCC22019), and *Candida tropicalis* (ATCC750) were cultured in dynamic conditions at 37 ◦C and 215 rpm in yeast malt broth (YMB) and passed regularly to maintain a healthy culture. To prepare the sample for DEP experiments, 100 μL of 4-day old cell culture were mixed with 2.5 mL of an optimized DEP buffer solution composed of 8.6 wt% sucrose, 0.3 wt% dextrose and 0.1 wt% bovine serum albumin to achieve a concentration of around 106 cells/mL. The electrical conductivity of this DEP buffer solution was 20 μS/cm. Cells were then pelleted through centrifugation at 5000 rpm for 5 min and then resuspended into fresh DEP buffer solution. This centrifugation and re-suspension protocol were repeated three times to ensure complete removal of any remaining YMB culture media.

### *2.2. Device Fabrication*

The microfluidic DEP device used in this study featured 3D carbon microelectrode arrays. The fabrication of the carbon microelectrodes has been reported several times in our previous work [8–12,31,34–37]. Briefly, the fabrication process included two-step photolithography of SU-8 (Gersteltec, Switzerland), a negative tone photoresist, on a Si/SiO2 substrate. The SU-8 microstructures were carbonized at 900 ◦C in a nitrogen environment using a heating rate of 5 ◦C/min to obtain carbon microstructures. The resultant carbon electrode array (3161 electrodes total) featured intercalated 3D electrodes as shown in Figure 1a; each carbon microelectrode had a height of 100 μm and diameter of 50 μm while the spacing between them was around 58 μm in all directions. A thin layer of SU-8 was

then patterned around the 3D carbon electrodes to insulate the connecting planar leads and planarize the channel bottom. On a parallel process, a 1.8 mm-wide and 32 mm-long channel was patterned from 127 μm-thick sheet of double sided pressure sensitive adhesive, or PSA (Switchmark 212R, Flexcon, Spencer, MA, USA), using xurography and adhered to a previously machined polycarbonate (PC) piece. The details of this method are detailed in our previous publication [38]. The DEP chip was then assembled by manually positioning the PC/PSA arrangement around the carbon microelectrode array, followed by sealing using a rolling press. The cross-section of the assembly of the microfluidic device is illustrated in Figure 1b.

**Figure 1.** Experimental set up for the characterization of DEP response of the *Candida* strains using 3D carbon microelectrodes. (**a**) Scanning Electron Microscope (SEM) image of the 3D carbon microelectrodes. (**b**) Cross section of the microchannel showing the different elements of the DEP device. The polarity of the 3D carbon electrodes to induce the non-uniform electric field for DEP is also illustrated. The region of interest (ROI) during experiments was immediately after the last column of the electrode array. The ROI for (**c**) the control experiment (no field applied); (**d**) during the trapping stage when cells displayed a strong DEP trapping behavior; and (**e**) immediately after turning the field off to release any previously trapped cells. The black dots in these images are *Candida* cells. (**f**) An indicative plot of the normalized intensity obtained after computational analysis of the ROI throughout an experiment. The blue dashed line denotes the time when the electric field was turned off, and marks the transition between the wash and release stages. The area under the curve in the "Release" section is identified by the green area, which is reported here as the DEP trapping response of the cells in each experiment. At least three experiments were conducted for each data point, i.e., a given *Candida* strain and frequency. See text for further details.

### *2.3. Experimental Protocol*

Experiments revolved around characterizing: (1) cell morphology and (2) DEP response. Few studies are available regarding the morphological characterization of few *Candida* strains [39,40] and a morphology study was performed here to better understand how the unique morphology of the

cells from each species may contribute to a difference in their response to an electric field. To this end, 4-day cell cultures were observed under an optical microscope (Nikon Eclipse LV100, Tokyo, Japan) to measure the dimensions of at least 30 cells per strain. Images were recorded through an Andor Zyla CMOS camera.

The experimental set up for characterization of the DEP response of the cells at different frequencies is illustrated in Figure 1. The experimental protocol followed can be separated into 3 stages: (1) cell trapping, (2) washing, and (3) cell release. Of note, cell release only occurred in the frequencies that lead to cell trapping due to positive DEP: this is when the cells were attracted to the regions of high field gradient that are around the carbon electrodes in this work. The desired flow rate in the experimental device was implemented using a syringe pump (FusionTouch 200, Chemyx, Stafford, TX, USA). The electrode array was polarized as illustrated in Figure 1a using a function generator (BK Precision 4052, Yorba Linda, CA, USA). During the first experimental stage, trapping, 20 μL of the sample containing the cells was flowed at 10 μL/min through the electrode array polarized at specific frequency (10 kHz–1 MHz) and magnitude of 20 Vpp. In the second stage, washing, a cell-free DEP buffer solution was flowed with a flow rate of 10 μL/min through the still polarized electrode array for 5 min to wash any non-trapped cells. In the last stage, release, the polarizing signal was turned off while maintaining the same flow rate. This last stage lasted for 110 s. The entire experiment was monitored through a 10× objective lens in a Nikon Eclipse LV100 microscope. However, only the release stage and the last 10 s of the wash stage were recorded using an Andor Zyla CMOS camera running at a frame rate of 5 frames per second. Hence, each of the video recordings used for data analysis was 120 s, or 600 frames, long. The electric field was turned off at frame 50.

### *2.4. Data Analysis*

Images of cell cultures were manually analyzed in the NIS Element Basic software native to the microscope to measure the major (*x*) and minor axes (*y*) of at least 30 cells per strain. The major and minor axes were identical for a perfect circle. The average values and standard deviation for all measurement were calculated using built-in mathematical functions in Microsoft Excel.

The videos obtained during DEP experiments were analyzed with ImageJ (National Institutes of Health, Bethesda, MD, USA) to plot the average pixel intensity at a region of interest (ROI), with area 830 μm × 700 μm and established immediately after the electrode array, throughout each experiment done for a particular cell strain and frequency of interest (Figure 1c–e). The analysis was designed such that a difference in the intensity of the ROI before and after turning the field off can be directly correlated to the strength of the trapping DEP force acting on a given sample at that specific frequency. To this end, the average intensity in the ROI was measured for a total of 600 frames, where frames 1–50 were for the frames recorded before turning the field off and frame 51–600 were recorded after the field was turned off. In order to properly isolate the DEP response from each experiment, all the intensity values after turning the field off were normalized against the average intensity before the field was off. These normalized values were then plotted using Origin Pro software (OriginPro 2016, Northampton, MA, USA). An example of such plot is shown in Figure 1f for an experiment that exhibited cell trapping and release. The larger the curve would denote a larger number of cells trapped, and released, and thus a stronger DEP trapping response. No DEP trapping resulted in no curve, i.e., a flat line after turning the field off. Here, we report the area under the curve in frames 51–600 to represent the DEP trapping response of the *Candida* strains.

### **3. Results**

### *3.1. Morphology of the Candida Strains*

Candida albicans (Figure 2a) displayed an average spherical morphology, which means the major axis diameter (*x*) of Candida albicans is identical to the minor axis diameter (*y*). Candida albicans featured an average diameter of *x* = *y* = 5.12 ± 0.75 μm, as depicted in Figure 2d. In addition to their largely spherical morphology, Candida albicans were commonly found in their budding phase of reproduction offering an alternate morphology of two or more attached spheres, as indicated by the dashed circles in Figure 2a. Such morphologies of the Candida albicans cells are in agreement with previous reports [39,40].

**Figure 2.** Morphology of (**a**) Candida albicans, (**b**) Candida tropicalis, and (**c**) Candida parapsilosis. The dashed circles indicate the budding behavior of the cells. (**d**) Plot of the major (*x*) vs minor diameter (*y*) of the different Candida strains. The diameters in the major and minor axes of a cell are illustrated in the inset. For a spherical cell, *x* is identical to y as denoted by the red dashed line. Each data point represents the average from at least 30 cells for each strain. The error bar represents the standard deviation from all measurements.

Candida tropicalis exhibited multiple morphologies. The first was a spherical morphology similar to that of Candida albicans but with larger diameter of *x* = *y* = 5.98 ± 0.75 μm, which is consistent with previous findings by other authors [39]. As in the case for Candida albicans, Candida tropicalis also displayed spherical morphology in its budding phase of reproduction. The second shape is referred to as pseudohyphae [39,41,42] and resulted when cells began to bud but instead of separating the membranes of the cell merged to become one elongated cell. This happened multiple times in our observations, leading the pseudohyphae to have elongated ellipsoidal morphology and even beginning to resemble a tree when multiple branches of elongated ellipsoids formed. The length of the pseudohyphae ranged from 7 μm to 27 μm with an average width of 1.89 ± 0.4 μm as seen in Figure 2b.

Candida parapsilosis shown in Figure 2c displayed an ellipsoidal morphology. The diameters of each of the two axes were found to be *x* = 6.29 ± 0.83 μm and *y* = 3.8 ± 0.84 μm, and these values are in the range of the dimensions reported elsewhere for this strain [39]. Although less frequent than in *Candida albicans, Candida parapsilosis* was also found in its budding stage but with a featuring morphology that resembled two or more attached ellipsoids instead of spheres.

### *3.2. Trap, Wash and Release of Candida Cells*

We were able to trap, wash and release different *Candida* strains at will. Upon applying the electric field at specific frequencies, the *Candida* cells experienced positiveDEP force and got trapped in the regions of high field gradient around the carbon microelectrodes (Figure 3). Our DEP device has a capacity of trapping around 4000 cells as previously reported [8]. In our DEP experiments, the microchannel was loaded with an experimental sample featuring a cell concentration of 106 cells/mL, which translates to a total number of cells ~7000 present in the microchannel to start with. Hence, upon applying the electric field, our device could trap the *Candida* cells to its capacity, and the wash protocol ensured carrying away the untrapped cells, enabling purification of the trapped cells. Due to the cell trapping, no cells were seen in the flow in the recording area at the end of the electrode array before the cell release (Figure 1d). Upon turning off the electric field, the trapped cells on the electrodes were eluted through the recording area (Figure 1e). Of note, a small number of cells was observed to be non-specifically adhered to the electrodes after turning off the polarizing signal. The number of such cells was significantly smaller than those released. Furthermore, this non-specific adhesion was observed for all strains studied in this work. Hence, the effect of such adhesion on the characterization of the DEP response for the different strains, and the differences between them, was deemed not significant. This non-specific adhesion may be detrimental to future assays where the recovery of targeted cells would be required but this out of the scope of the work presented here.

**Figure 3.** Examples of trapping of (**a**) *Candida albicans*, (**b**) *Candida tropicalis*, and (**c**) *Candida parapsilosis* cells on the carbon microelectrodes (the dark circles and connecting lines) due to positive dielectrophoresis. These specific examples are when the frequency of the applied electric field is 100 kHz. Note that for *Candida tropicalis*, few short length pseudohyphae cells were also trapped and indicated with the red dotted ellipse in (b).

### *3.3. DEP Response of the Candida Strains*

The DEP trapping response for the different *Candida* species characterized in this work is plotted in Figure 4. *Candida albicans* and *Candida parapsilosis* exhibited a trapping DEP response in the frequency range 10–500 kHz. *Candida tropicalis* also showed a trapping positive DEP response at 750 kHz. No positive DEP was observed beyond 750 kHz for any of the *Candida* strains. Although all the three strains showed positive DEP response in the range 10–500 kHz, the strength of the DEP response was different for each strain. The highest DEP trapping response for *Candida tropicalis* and *Candida albicans* was at 50 kHz, whereas *Candida parapsilosis* showed the peak DEP response at 100 kHz. *Candida albicans* exhibited relatively weak DEP trapping compared to the other two strains of *Candida* in the frequency

range 50–500 kHz. However, at 10 kHz, the positive DEP response was strongest for *Candida albicans* among the three *Candida* strains.

**Figure 4.** The DEP trapping response of *Candida albicans*, *Candida tropicalis*, and *Candida parapsilosis* at frequencies ranging from 10 kHz to 1 MHz. At least three experiments were carried out for each data point, bars denote the average values while the bars represent standard deviation. The upward arrow indicates that the higher value represents higher cell trapping due to DEP.

### **4. Discussion**

The *Candida* strains studied here showed subtle differences in cell morphologies. Figure 2d shows how the cells dimensions overlap for all the strains, making it difficult to sort *Candida* strains solely based on their sizes. The dimensional overlap is more prevalent when comparing *Candida albicans* and *Candida tropicalis*. Moreover, both *Candida albicans* and *Candida tropicalis* exhibited spherical morphology, further complicating differentiation based on shape. Although pseudohyphae forms of *Candida tropicalis* exhibits different shape than the spherical cells, *Candida albicans* is also known to transform to hyphal state during infectious process [43–45] and this must be taken into account even when we did not observe hyphal forms in our handling of *C. albicans*. Although size and shape are not enough for cell separation, these can be complemented with the DEP response of specific strains.

All the *Candida* strains exhibited positive DEP response in the frequency range 10–500 kHz. However, *Candida tropicalis* and *Candida parapsilosis* consistently exhibited higher positive DEP response than *Candida albicans* in the frequency range 50–250 kHz. Such behavior might be a result of the budding behavior of *Candida albicans*. *Candida albicans* were more commonly found in their budding stage as a group of multiple attached spheres. The drag force created by a conglomeration of attached spherical entities would have a higher magnitude than the drag force of a smaller spherical entity [46]. If the drag force becomes larger than the DEP force created from the electrical field gradient, the cells would not attract to the electrodes [30]. However, at 10 kHz, different DEP behavior was observed, where *Candida albicans* exhibited strongest DEP response. Our hypothesis is that different *Candida* strains might feature different cell membrane potential, which resulted in different behaviors in different frequency range. However, the cell membrane potentials for *Candida* strains are unknown at present. A separate, more extensive study to determine the cell membrane potential of these *Candida* strains is ongoing and will be reported in a future work.

*Candida tropicalis* shows both spherical and pseudohyphae morphology in the media studied here as shown in Figure 2b. However, mostly spherical *Candida tropicalis* cells were observed on the carbon microelectrodes during trapping (Figure 3b). A small amount of short ranged pseudohyphae cells with length ranging from 7 μm to 12 μm were also trapped on the carbon electrodes. No long range pseudohyphae cells were trapped on the carbon microelectrodes. We speculate that as the *Candida tropicalis* cells transforms to pseudohyphae cells, the cell membrane potential might also change. The difference in the cell membrane potential of the spherical and pseudohyphae cells might be in direct proportion of the length of the pseudohyphae cells. For example, the short pseudohyphae cells might feature a cell membrane potential close to that of the spherical cells. This may lead to trapping of short length pseudohyphae cells in the current DEP conditions along with the spherical cells, whereas the long pseudohyphae cells did not experience any positive DEP force and flow with media during the washing step.

The current results indicate a potential for DEP to be utilized to distinguish between different types of *Candida* strains from an already purified blood sample and help diagnose the specific strain causing disease. One of the important results in this direction is that *Candida tropicalis* was the only *Candida* strain to show a positive response at 750 kHz. In terms of strain identification, a sample could be subjected to an electric field with frequency of 750 kHz to only trap *C. tropicalis* while eluting *Candida albicans* and *Candida parapsilosis*. Further separation between *C. albicans* and *C. parapsilosis* could be done in a second stage, i.e., polarized at 10 kHz to emphasize trapping of *C. albicans*, in a multi-stage carbon-electrode DEP device as previously reported [47]. Furthermore, each *Candida* strain exhibited a difference in the strength of the positive DEP response, which can be also utilized for diagnosis purposes. This characteristic could be used in multiple ways. One potential method would be to tailor the strength of the electric field within the DEP device to be strong enough to attract one type of cell, but too weak to trap another *Candida* strains allowing these *Candida* strains to be eliminated as the cause of infection. The trap and wash protocol demonstrated here can be used to enrich desired *Candida* cells in a small sample volume for further analysis. Using the current DEP set up, it is possible to enrich a cell sample up to 150 folds within a few hours as we previously reported [8]. This rich enrichment can lead to a timely detection of the *Candida* cells by enabling concentrated and purified samples to improve the sensitivity of common detection protocols such as PCR [31]. Another approach might be using streaming DEP for rapid cell sorting, where cells are focused into specific streams of elution instead of trapping cells on the electrodes. Streaming DEP can enable focusing of different *Candida* cells into different streams utilizing the different strength of DEP on the different *Candida* strains [11,30]. The streams can be collected separately at the outlet of the microfluidic system and used for cell detection. Such timely detection of the *Candida* cells can enable timely initiation of medical treatment specific to the responsible *Candida* strains.

It should be noted that the value of DEP in a practical solution for rapid diagnosis of *Candida* infection is envisioned to be the isolation of different strains of *Candida* from each other, but not directly from blood. A multi-stage protocol is preferred when attempting to isolate potential targets of interest from a blood sample in clinical diagnostics. For example, centrifugation can enable a first rapid coarse separation of serum, buffy coat and red blood cells (RBC) [48], with the *Candida* cells expected to be in the buffy coat [49]. Buffer exchange protocols common in clinical diagnostics can then be implemented to re-suspend the cellular content of the buffy coat in a buffer optimized for DEP and any other downstream processing. Further stages in the process can include size exclusion to further isolate *Candida* cells from other blood cells until only particles that may resemble *Candida* are present in the sample. At this stage, a DEP assay for fine separation could be used to perform isolation and purification of specific strains to increase the performance of detection assays such as PCR as detailed above. If necessary, the DEP properties of blood cells with similar sizes than *Candida* have been characterized [9,50,51] and such knowledge could be used to aid in their separation. Such a multi-stage process can be readily implemented in a clinical setting. However, the details of such integrated assay are out of the scope of this paper. Of note, the integration of DEP with centrifugal microfluidics towards enabling such integrated assay has been reported by one of us [13,35].

### **5. Conclusions**

The characterization of the morphology and DEP response of three frequently isolated *Candida* strains from infected samples: *Candida albicans*, *Candida tropicalis*, and *Candida parapsilosis* was presented here. The studied *Candida* strains only exhibited subtle differences in morphology that makes direct observation an incomplete method to sort them. 3D carbon electrode DEP was implemented for characterizing the DEP response of the *Candida* cells. All three *Candida* strains showed a strong positive DEP response in the frequency range 10–500 kHz. However, positive DEP at 750 kHz was only observed for *Candida tropicalis*. Furthermore, the *Candida* strains exhibited the positive DEP at different strength. *Candida tropicalis* and *Candida parapsilosis* showed relative high DEP response than *Candida albicans* in the frequency range 50–250 kHz, whereas at 10 kHz, the DEP response was strongest for *Candida albicans*. Together, the morphological and DEP differences among the different strains could provide a framework to enable sorting different strains.

This is to the best of our knowledge the first study reporting the DEP responses of different *Candida* strains. The current results show promise towards using DEP as a tool to enable separation of different *Candida* species. Ongoing work focuses on characterizing the membrane capacitance of the strains presented here and expanding this study to other relevant strains such as *Candida glabrata* and *Candida krusei*. The results presented here indicate that one can potentially manipulate and enrich a specific *Candida* strain at specific DEP conditions and encourages further work towards rapid identification to the enable effective and timely treatment of candidiasis.

**Author Contributions:** Conceptualization, M.I. and R.M.-D.; Data curation, M.I. and J.G.; Formal analysis, M.I., D.K., J.G. and R.M.-D.; Investigation, M.I. and J.G.; Methodology, M.I. and R.M.-D.; Project administration, R.M.-D.; Supervision, R.M.-D.; Validation, M.I.; Writing—original draft, M.I., D.K. and J.G.; Writing—review & editing, M.I. and R.M.-D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors thank Mark Blenner and his laboratory from the Department of Chemical & Biomolecular Engineering at Clemson University for facilitating the culture of Candida cells.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Adriana Coll De Peña 1,2, Nurul Humaira Mohd Redzuan 2, Milky K. Abajorga 2, Nicole Hill 1, Julie A. Thomas 2,\* and Blanca H. Lapizco-Encinas 1,\***


Received: 23 May 2019; Accepted: 30 June 2019; Published: 4 July 2019

**Abstract:** Bacterial viruses or phages have great potential in the medical and agricultural fields as alternatives to antibiotics to control nuisance populations of pathogenic bacteria. However, current analysis and purification protocols for phages tend to be resource intensive and have numbers of limitations, such as impacting phage viability. The present study explores the potential of employing the electrokinetic technique of insulator-based dielectrophoresis (iDEP) for virus assessment, separation and enrichment. In particular, the application of the parameter "trapping value" (*Tv*) is explored as a standardized iDEP signature for each phage species. The present study includes mathematical modeling with COMSOL Multiphysics and extensive experimentation. Three related, but genetically and structurally distinct, phages were studied: *Salmonella enterica* phage SPN3US, *Pseudomonas aeruginosa* phage φKZ and *P. chlororaphis* phage 201φ2-1. This is the first iDEP study on bacteriophages with large and complex virions and the results illustrate their virions can be successfully enriched with iDEP systems and still retain infectivity. In addition, our results indicate that characterization of the negative dielectrophoretic response of a phage in terms of *Tv* could be used for predicting individual virus behavior in iDEP systems. The findings reported here can contribute to the establishment of protocols to analyze, purify and/or enrich samples of known and unknown phages.

**Keywords:** bacteriophage; dielectrophoresis; electric field; electrophoresis; electrokinetics; virus

### **1. Introduction**

Bacteriophages, estimated to have a total population of 1031, are possibly the most abundant and genetically diverse biological entities on earth [1,2]. Over the past decade, there has been an increase in observations of antibiotic resistance leading to the need for alternative treatments for bacterial infections. Phage therapy possesses great potential to control multi-drug resistant organisms, such as in the medical and agricultural fields [2]. To employ phages safely for such purposes it is important to have an in-depth knowledge of a representative for each phage group. Our research focuses on understanding the biology of unusually large, so-called "giant" phages, with >200 kb dsDNA genomes, such as *Salmonella enterica* phage SPN3US (240 kb). Increasing numbers of phages that share a core set of genes with SPN3US have recently been isolated, most for the goal of using them for phage therapy purposes. Despite their obvious potential for biocontrol applications, in reality little is known about the biology of these phages. For instance, the virions of giant phages related to SPN3US are comprised of many (>70) different proteins ranging in copy numbers from just a few to 1560 copies per virion, a large proportion of which (~80%) have no known specific function. These characteristics have led us to conduct genetic studies on SPN3US as a model for related "giant" phages.

To conduct characterization studies on SPN3US, or any phage, requires the use of techniques to purify and enrich particles from a non-homogenous sample containing bacterial cells, cellular debris, and virions, as illustrated in Figure 1. Traditional bacteriophage purification methods, developed for model phages, such as T4 and T7, have numbers of limitations, including being time and labor intensive and often involving a complex series of steps (e.g., CsCl gradient ultracentrifugation) [3]. However, due to the great variability in virion composition between different phages, traditional procedures are frequently not suitable for the purification of many of the newer "environmental" phages, damaging their virions and causing loss of viability. For instance, CsCl gradient purification causes the virions of *Bacillus* virus, phageG to completely disintegrate, and those of *B. thuringiensis* phage 0305φ8-36 to lose infectivity by several orders of magnitude [4]. Similarly, even environmental phages that are related to model phages, such as T4, can respond very differently to the protocols employed for the model phage (e.g., [5]). To further complicate matters, standard purification protocols often do not completely remove all bacterial debris, such as cell wall proteins and endotoxins, which can impede downstream analyses (e.g., mass spectrometry) and represents a major problem for therapeutic preparations of phages [6,7]. Given these limitations of traditional phage purification techniques, alternative separation and enrichment processes are being explored [6–8].

**Figure 1.** Transmission electron microscopy (TEM) of SPN3US-infected *Salmonella* and purified SPN3US. (**A**) Negatively stained thin section showing two *Salmonella* cells in the process of being lysed at the end of infection by SPN3US. Regions of the cell walls undergoing rupture due to phage enzymes are indicated with black arrowheads. Particles of SPN3US progeny are indicated with white arrowheads. Note the extensive amount of cell debris in the sample. (**B**) Negatively stained image of a single SPN3US virion from a preparation that has undergone purification via CsCl gradient ultracentrifugation to remove cellular debris from the sample. SPN3US virions consist of a head (which contains the dsDNA genome) and tail which ends in a complex baseplate that attaches to a *Salmonella* cell to initiate infection.

Microfluidics has revolutionized the manner in which many bioanalytical assessments are performed. It has opened the doors to perform high resolution and sensitivity purification assays [9]. Electrokinetics (EK), electric field-driven techniques, is one of the main pillars of microfluidics due to its great flexibility and simplicity of application. Dielectrophoresis (DEP) has proven to be a robust platform for the separation, sorting and enrichment of a wide array of biological particles ranging from macromolecules to parasites [10–14]. Dielectrophoresis is the migration of particles under the influence of a non-uniform electric field. Unlike electrophoresis (EP), DEP exploits particle polarization effects, not the electrical charge, leading to a greater flexibility since it works with both DC and AC electric potentials [15]. Insulator-based DEP (iDEP) is a technique where non-uniform electric fields are produced employing insulating structures, usually embedded in a microchannel, creating a truly 3-dimensional dielectrophoretic effect [16]. Is it important to note that iDEP systems can suffer from

electrolysis and Joule heating effects [17] due to the requirement of high voltages. Therefore, operating conditions need to be carefully selected.

Microorganisms have been extensively studied in dielectrophoretic-based systems, including both electrode-based DEP (eDEP) and iDEP systems [18]. A challenge in the dielectrophoretic manipulation of viral particles is the inherently small size as larger applied electric potentials are required to generate sufficient dielectrophoretic forces [19]. Some of the first studies were focused on the assessment of viruses that are pathogenic to humans. In 1996 the Furh research group demonstrated the enrichment and stable trapping of influenza and Sendai viruses in an eDEP system with two sets of planar electrodes that allowed for the creation of 3D field cages [20,21]. This work was later extended by Grom et al. demonstrating the ability to transport and accumulate hepatitis A virus in a field cage consisting of eight microelectrodes. [19]. Hughes et al. reported a series of studies on the characterization of herpes simplex virus with DEP [22–24]. Akin et al. reported an iDEP system with an interdigitated electrode array for real-time trapping and imaging of vaccinia virus [25]. Masuda et al. presented a 3-dimensional iDEP system that allowed the filtration and selective transportation of a single influenza to promote single-virus cell infection [8]. Prakash et al. employed a droplet-based system for the detection of influenza viruses using PCR; illustrating the potential of DEP for diagnostics [26]. Ding et al. utilized gradient iDEP to concentrate Sindbis virus to increase the concentration of the virus from two to six times within the channel using voltages as low as 70 V [12]. Other recent reports have focused on the development of sensors. Singh et al. created a sensor for influenza virus employing carbon nanotubes that were electrodeposited by means of DEP [27]. Madiyar et al. reported the capture and detection of vaccinia virus with DEP by using carbon nanoelectrode arrays [28]. Some earlier studies involved plant viruses. Morgan and Green demonstrated the first application of eDEP for the manipulation of tobacco mosaic virus (TMV) using AC electric fields [29]. Ermolina et al. characterized the dielectric properties of cow pea mosaic virus and TMV in a system with castellated electrodes [30,31]. Lapizco-Encinas reported the enrichment of TMV in an iDEP system with cylindrical insulating posts [32].

In contrast to human and plant viruses, there have been few studies on the suitability of DEP for phage enrichment. Sonnenberg et al developed an eDEP system for the isolation detection of T7 bacteriophage from whole blood [33], illustrating the potential of DEP for clinical applications. Madiyar et al. demonstrated single virus and large ensemble trapping of T4r and T1 bacteriophages from a dilute solution under conditions with a nanoelectrode array made of carbon nanofibers [34].

The contributions mentioned above are excellent examples of some of the latest advancements in the dielectrophoretic manipulation of viral particles. However, it is evident that systems capable of purifying newer "environmental" phages, including giant phages, are still an unexplored area. Similarly, unexplored are systems capable of handling a larger throughput containing several viral species which would be of potential value in phage therapy as typically cocktails or mixtures of different types of phages are employed. In this contribution, we present the first report on the assessment and enrichment of *Salmonella* phage SPN3US, and for comparison purposes, two related giant *Pseudomonas* phages: φKZ and 201φ2-1, in two distinct iDEP systems. In particular, the application of the parameter "trapping value" (*Tv*) is explored as a standardized iDEP signature for each virus species. This work includes mathematical modeling with COMSOL Multiphysics®(version 4.4, COMSOL Inc., Stockholm, Sweden) and experimentation with iDEP devices containing an array of circular or oval-shaped insulating posts. For model information, please see the supplementary material. The dielectrophoretic trapping of viral particles under the influence of DC electric potentials was fully characterized in order to discern the specific trapping conditions ("sufficient" trapping) for each one of the distinct viral species. All viruses in this study exhibited negative dielectrophoretic behavior. The results illustrating virus trapping and enriching allowed the identification of the specific Trapping value (*Tv*, Equation (3)) for each type of phage [35,36]. This is the first iDEP study on large bacteriophages and these findings could be used for the design of new iDEP systems aimed to separate and enrich samples of both known and unknown phages.

### **2. Theory**

Particles can exhibit either positive or negative DEP, depending on their relative polarizability with respect to the suspending media [37]. Positive DEP (pDEP) occurs when the particle is more polarizable than the medium, resulting in particle attraction to the regions with higher electric field gradient. Negative DEP (nDEP) is the opposite effect. In our iDEP channels (Figure 2A,B), the constrictions between posts are the areas of high field gradients. Under nDEP, all virus species in this study could be trapped in the constriction regions. In iDEP systems particles are captured when the effects of DEP and linear EK forces, which are opposite, are balanced [38]. For a particle to become trapped the following condition has been identified [39,40]:

$$\frac{\mu\_{DEP} \nabla E^2 \cdot \overrightarrow{E}}{\mu\_{EK} \, E \cdot \overrightarrow{E}} \le -1,\tag{1}$$

Separating the above expression into system-dependent and particle-dependent parameters:

$$\frac{\nabla E^2 \cdot \stackrel{\rightarrow}{E}}{E^2} \le -\frac{\mu\_{EK}}{\mu\_{DEP}},\tag{2}$$

where the left-hand side of the equation is the Trapping value which is independent of particle properties, as it only depends on the electric field magnitude (<sup>→</sup> *E*) and gradient of the electric field squared (∇*E*2). This parameter, identified by the Casals-Terr<sup>é</sup> [35] and Hayes [36] groups, characterizes the condition required to trap a specific type of particle:

$$Tv = \frac{\nabla E^2 \cdot \overrightarrow{E}}{E^2}.\tag{3}$$

**Figure 2.** Schematic representation of one insulator-based dielectrophoresis (iDEP) channel employed in this study. (**A**) Top view of a full channel for design Circle-200-220. (**B**) 3D representation of the channel. For the two designs analyzed in this study, an illustration of four insulating posts with dimensions is included: (**C**) Circle-200-220, (**D**) Oval-200-220&80-170. Design names illustrate post size and post spacing.

### **3. Materials and Methods**

#### *3.1. Microdevices, Viral Samples and Suspending Medium*

Experiments were conducted in two distinct microchannel designs made from PDMS employing standard soft lithography techniques; microfabrication information is included here [40]. The microchannels were 10.16 mm long, 40 μm deep and 880 μm wide, specific post dimensions are included in Figure 2C,D. This study employed high titer stocks (1010–1012 pfu (plaque-forming units)/mL) of three related viruses: *Salmonella* Typhimurium phage SPN3US [41], *Pseudomonas aeruginosa*

phage φKZ [42] and *P. chlororaphis* phage 201φ2-1 [43]. These phage stocks underwent a low speed clarification spin (~8000 g, 10 min, 4 ◦C) to remove large bacterial debris. All virus samples were fluorescently labeled as follows: 1 mL of a phage stock was spun down at 13,000 rpm for 10 min, after discarding the supernatant the pellet was resuspended in 0.5 mL of distilled water. Then, 2 μL of SYTO 11 dye (Invitrogen, Carlsbad, CA, USA) was added to the sample and incubated for 20 min. After removing the excess dye, the sample was resuspended in 0.5 mL of the suspending medium. The suspending medium was sterilized deionized water with a conductivity of 14 μS/cm and a pH of 7.07; under these conditions the zeta potential of the PDMS channel was approximately −108.57 mV.

Phage samples were assayed for viability via plaque assays in triplicate using the standard double overlay technique using LB agar bottom plates and overlays made from LB broth and 0.34% agar. Briefly, each phage sample underwent a 10-fold dilution series in SM buffer and these were spotted onto overlays made containing 100 μL of a fresh overnight culture of the appropriate bacterial strain. Plaques were enumerated after overnight incubation at 30 ◦C.

### *3.2. Equipment and Experimental Procedure*

Phage response was observed and recorded as videos with a Leica DMi8 inverted microscope (Wetzlar, Germany). Direct current (DC) electric potentials were applied with a high voltage supply (Model HVS6000D, LabSmith, Livermore, CA, USA). COMSOL Multiphysics®4.4 was used to predict the magnitude of the trapping value (*Tv*, Equation (3)). Each experiment started with a clean channel to which a 5–10 μL sample of the corresponding labeled virus was added, followed by the application of DC electric potentials. For the purpose of this study, a "sufficient" trapping voltage was determined as the required voltage to obtain a visually observable band or cluster of trapped viral particles.

### **4. Results and Discussion**

### *4.1. Experimental Characterization of the Dielectrophoretic Trapping of Phage Virions*

A series of experiments were carried out to characterize the required voltage to trap and enrich each type of phages in both iDEP devices with nDEP. After a sample of the fluorescently labelled virus was introduced into the channel, the applied voltage was manually increased until "sufficient" trapping of the viral species was observed. Each experiment was repeated at least five times to ensure reproducibility, a summary of these results is included in Table S1 (supplementary material). Figure 3A,C illustrate images of the trapping of all three phage species in the circle-shaped iDEP channel at applied potentials between 1100 and 1200 V. Lower voltages, in the range of 750–800 V, were required with the oval-shaped posts, as depicted in Figure 3D,F. Figure 3G shows a plot of the required trapping voltage necessary to achieve "sufficient" trapping. As expected, for all viral species, the required voltages are lower with the oval-shaped posts, since narrower posts generate higher electric field gradients (∇*E*2), producing greater dielectrophoretic forces [40]. The characteristic trapping voltage and *Tv* for each viral species is a strong function of the size, shape and polarizability of the viral species. As demonstrated by Hughes et al. [44] the total conductivity of a particle depends on the conductivity of the bulk material, and the individual conductances of the compact and diffuse layers of the electrical double layer (EDL). This group successfully extended this analysis with the dielectrophoretic characterization of simplex virus-1 capsids [45]. In a later contribution, Ermolina et al. [30] illustrated that surface conductance, which is directly related to polarizability, is a dominant parameter in the EP response of submicron particles, such as viruses.

### *4.2. Modeling Predictions for the Trapping of Phage Virus*

The trapping value (*Tv*), which characterizes the conditions required to trap a specific type of particle [35,36], was determined using COMSOL Multiphysics®software (Table S1). The geometries of interest were imported into COMSOL along with trapping voltage (Figure 3G) associated for each species in order to predict the parameters <sup>∇</sup>*E*2, → *E* and *E*2. These values were estimated across a cutline located at the centerline of one constriction between two posts. Images depicting the cutlines used in these estimations are illustrated in Figure S1.

**Figure 3.** Results of the dielectrophoretic trapping of all three phages. Circle-shaped posts: (**A**) SPN3US at 1200 V, (**B**) φKZ at 1100 V and (**C**) 201φ2-1 at 1100 V. Oval-shaped posts: (**D**) SPN3US at 800 V, (**E**) φKZ at 750 V and (**F**) 201φ2-1 at 750 V. (**G**) Experimental characterization of the trapping voltage, and (**H**) Estimation of the trapping value (*Tv*) in both iDEP channel designs for the three types of bacteriophages in this study. Table S1 in the Supplementary Material includes a summary of the trapping voltage and *Tv* estimations.

As defined by Casals-Terré [35] and Hayes [36] groups, the *Tv* parameter normalizes the required conditions for trapping a specific type of particles for any type of iDEP design. The results in Figure 3H confirm the applicability of *Tv*. It can be observed for each one of the three viral species that the *Tv* values for both iDEP designs are quite similar, a finding which is consistent with our previous analyses that each of these three phages has a large virion composed of mostly homologous proteins and very similar dimensions [46]. Notably, the trapping voltage (Figure 3G) for each phage was unique, which indicates that if, in future studies, the separation of a mixture of similarly related viruses was required, we should focus on the trapping voltages, not the trapping values. In addition, our results indicate that devices with wider posts might be more suitable for such separation purposes (i.e., wider posts (circles) produced trapping voltages with a larger distribution between the three virus species than obtained with narrow posts (ovals)). Furthermore, these findings open the exciting

possibility of using *Tv* for the designing of iDEP devices with distinct post geometries; and also for predicting the trapping conditions of different types of particles (from viruses to cells) in a given iDEP device. Consequently, these findings will be relevant for future studies on mixtures of phages, even those including related phages, as found in phage therapy cocktails.

### *4.3. Viability Assessments after Dielectrophoretic Trapping*

To evaluate the potential of iDEP for bacteriophage purification and enrichment, the phage samples were assessed for viability after exposing them to high voltages. To do this, the stained input samples that had been previously fluorescently stained for the trapping voltage experiments were run on a circle post design in triplicates, and exposed to the following sequence of voltages: 400 V for 30 s, 800 V for 20 s and 400 V for 10 s. The voltages were chosen to represent the experimental conditions with an initial voltage to move the sample to the post array, followed by a trapping voltage and a release voltage to move the sample to the outlet reservoir prior to extraction. While some sample did reach the outlet reservoir without experiencing the total magnitude of the electric field gradient within the constrictions at 800 V, all viruses retrieved were still exposed to high voltages for at least one minute. Upon extraction of the samples from their respective channels, the samples of each phage were plated on their respective bacterial host. The clearings in the bacterial lawns observed in Figure 4A represent bacterial cell lysis generated by the presence of viable bacteriophages. Remarkably, the two *Pseudomonas* phages, φKZ and 201φ2-1, had titers in the high range after trapping (Figure 4B, Table S2), a finding which supports that iDEP can indeed be used as a purification technique and not only as an analytical tool for bacteriophages.

**Figure 4.** Qualitative viability assessments for all three phages studied here. (**A**) Three samples of phages SPN3US, φKZ and 201φ2-1 that had been fluorescently labelled and treated were spotted onto the lawns of their respective bacterial hosts in three replicate experiments. Volumes of phage spotted are indicated at the bottom of the image. (**B**) Enumeration of viable particles (plaque-forming units, pfu) of the samples in (A) for phages SPN3US, φKZ and 201φ2-1, with the exception of SPN3US sample 1 which was not able to be titered.

The titers of SPN3US after iDEP were more variable despite comparable trapping magnitudes of its virions in the iDEP channels relative to φKZ and 201φ2-1 (Figure 4). Potential causes for the SPN3US titer variability include that certain parts of its virion may be more susceptible to damage during trapping after virions have been treated with SYTO 11 than similar structures in φKZ and 201φ2-1. This seems highly likely as our analyses indicate that each phage has a reduced viability immediately after SYTO 11 treatment (Table S3). It is also feasible that some of the variation in the viability of the SPN3US sample extracted from iDEP may have been the consequence of variation within the channel to the outlet reservoir. Potential future research directions could focus on increasing the yield of enriched phages, and quantification of iDEP relative to existing phage purification techniques with regard the amount of bacterial contaminants removed.

### **5. Conclusions**

Presented here is the assessment of three phages: SPN3US, φKZ and 201φ2-1 in two distinct iDEP devices, one with circle-shaped and one with oval-shaped insulator posts. Experimental work demonstrated the successful trapping of all three phage species, where the voltage requirement to achieve trapping of virions was lower in devices with circular insulating posts, since these produce lower dielectrophoretic forces than the oval-shaped posts. A mathematical model created with COMSOL was then employed to estimate the trapping value (*Tv*) for each phage type. This parameter, as identified by other research groups, normalizes the required conditions, in terms of electric field and electric field gradient, for trapping a specific type of particle. The results demonstrated that the *Tv* for a specific species is reasonably constant within the two distinct designs studied here, opening the exciting possibility of using *Tv* for the designing of iDEP devices targeting specific viral species, and also for predicting the required voltage for trapping a specific type of particle, including viruses, in distinct iDEP devices. In addition, these findings suggest that iDEP has potential for analyses of mixtures of phages, even those including related phages, such as found in phage therapy cocktails.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-666X/10/7/450/s1. Figure S1: Representation of the cutline employed in COMSOL for the determination of *Tv*. Table S1: Comparison between trapping voltages and *Tv* of SPN3US, φKZ, and 201φ2-1 in the circle and oval post channel designs. Table S2: Enumeration of viable particles of phages SPN3US, φKZ, and 201φ2-1 after trapping in the circle designs. Table S3: Enumeration of viable particles of phages SPN3US, φKZ, and 201φ2-1 before and immediately after staining with SYTO 11.

**Author Contributions:** A.C.D.P, J.A.T and B.H.L.-E conceived and designed the experiments, and contributed reagents/materials; A.C.D.P performed the microfluidics experiments; N.H.M.R and M.K.A. prepared virus stocks and performed viability experiments; A.C.D.P and N.H. performed mathematical model analysis; A.C.D.P, J.A.T, and B.H.L.-E. analyzed the data and wrote the paper; A.C.D.P, J.A.T, N.H. and B.H.L.-E. reviewed and edited the paper.

**Funding:** A.C.D.P., N.H. and B.H.L.-E would like to acknowledge the financial support provided by the National Science Foundation (CBET-1705895). J.A.T., N.H.M.R. and M.K.A. acknowledge support by the National Institutes of Health (UA5GM126533).

**Acknowledgments:** We thank Ru-ching Hsia for transmission electron microscopy (TEM) of SPN3US and SPN3US-infected *Salmonella*. TEM was performed at the UMB Electron Microscopy Core Imaging Facility. We thank Roberto Gallo-Villanueva for providing Figure 2B.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Multi-Stage Particle Separation based on Microstructure Filtration and Dielectrophoresis**

### **Danfen Yin, Xiaoling Zhang, Xianwei Han, Jun Yang and Ning Hu \***

Key Laboratory of Biorheological Science and Technology, Chongqing University, Ministry of Education, Bioengineering College, Chongqing University, Chongqing 400030, China; yindf@cqu.edu.cn (D.Y.); zhangxiaoling@cqu.edu.cn (X.Z.); 20161901019@cqu.edu.cn (X.H.); bioyangjun@cqu.edu.cn (J.Y.) **\*** Correspondence: huning@cqu.edu.cn; Tel.: +86-23-6510-2291

Received: 30 December 2018; Accepted: 29 January 2019; Published: 31 January 2019

**Abstract:** Particle separation is important in chemical and biomedical analysis. Among all particle separation approaches, microstructure filtration which based particles size difference has turned into one of the most commonly methods. By controlling the movement of particles, dielectrophoresis has also been widely adopted in particle separation. This work presents a microfluidic device which combines the advantages of microfilters and dielectrophoresis to separate micro-particles and cells. A three-dimensional (3D) model was developed to calculate the distributions of the electric field gradient at the two filter stages. Polystyrene particles with three different sizes were separated by micropillar array structure by applying a 35-Vpp AC voltage at 10 KHz. The blocked particles were pushed off the filters under the negative dielectrophoretic force and drag force. A mixture of Haematococcus pluvialis cells and Bracteacoccus engadinensis cells with different sizes were also successfully separated by this device, which proved that the device can separate both biological samples and polystyrene particles.

**Keywords:** Microfilter; Dielectrophoresis; Particle separation, micropillar

### **1. Introduction**

Microfluidic technology involves the control and manipulation of small amount of fluid confined in micron-sized geometry [1,2]. Microfluidic operations have many advantages, including faster analyses, minimum consumption of samples and reagents, shorter reaction times, and high-throughput screenings [3]. Moreover, miniaturization makes it possible to develop portable devices, which means that miniaturized laboratory equipment can be taken where it is needed. The influence of microfluidic technology in the academic community has rapidly increased over the last ten years because it has a number of potential applications in such areas as biological analysis [4], clinical examination [5], and food safety inspection [6,7].

Particle separation is important in chemical and biomedical analysis [7], and microfluidic techniques have been effective at separating particles. Many microfluidic methods have been developed to separate particles [8] using the flow field, microstructure, or forces created by electricity [9], optics [10–13], acoustics [14,15], magnetics [16–18], hydrodynamics [9,19,20], or gravity [20,21]. Some of these methods require fluorescent immunolabeling or magnetic labeling of the targeted or non-targeted particles, which is not only complicated but also possibly pollutes the reactants [22]. Other methods do not require label and separate particles by their intrinsic qualities, such as dielectric properties, deformability, size, deterministic lateral displacement, etc.

Microfilters, which are frequently used in many fields for tasks such as particle capture, enrichment and separation, have a controllable pore size and distribution. Microfilters do not need sophisticated injection systems, which make them efficient, fast, and simple.

Wilding et al. [23] demonstrated a micropillar array structure to separate blood cells and particles from nanoliter samples. However, the use of micropillar array structures to separate cells can lead to blockages in the flow path, limiting the development of the structure. Recently, different kinds of microfluidic chips were developed to avoid blockages, such as independent micropillars based filter, etc. Mohamed et al. [24] designed a sorting chip consisting of four channel segments. The channel's width sequentially contracts from 20 μm to 15 μm, then to 10 μm, and ends at 5 μm, while the height remains constant. Cells injected into the chip were captured at different segments according to their sizes. Tan et al. [25] proposed another filtering structure to separate cancer cells. On the microfluidic device, each filter element consists of three independent micropillars arranged in a circular arc with a 5 μm gap, which ensured that only larger cancer cells cannot pass and other cells can bypass from the side due to the increased flow resistance. McFaul et al. [26] developed a funnel-shaped micropillar array with a large gap in upstream and a small gap in downstream to reduce the influence of blockages. These methods can alleviate the degree of clogging to a certain extent, but they cannot completely prevent blockages.

Dielectrophoresis (DEP) is also a useful and simple technology for particle separation [27]. Positive dielectrophoresis (pDEP) and negative dielectrophoresis (nDEP) are both used to separate blood cells, tumor cells (including CTCs), algae cells, etc. The DEP force is a net force caused by the non-uniform electric field around the particles, which could be generated by the geometry of the electrodes or insulators [28]. The most frequently used DEP approaches are electrode-based DEP and insulator-based DEP [8,29,30]. The electric field gradient in iDEP is produced by using insulator micropillars, which in eDEP is produced by complex shaped electrodes. Particle separation using iDEP can avoid many problems that may occur in eDEP. Mohammadi et al. [31] developed an efficient micropost array to capture particles using insulator-based dielectrophoretic generated by a DC voltage source, and did numerical simulation to find the most efficient design of the post array, which demonstrated the effectiveness of the combination of micropost array and dielectrophoresis

However, the blockage problem frequently occurs in particle separation by using microfilter devices. Recently, our group developed a new microfluidic device made of a series of filters to separate particles using AC voltages. AC voltages were chosen to generate iDEP because the electrodes in AC voltages were less likely to be electrolyzed than in the DC voltages. Compared with the proposed methods, our device combined the advantages of microfilters and dielectrophoresis. Under the manipulation of dielectrophoretic force, the blocked particles were pushed off from filters to ensure that the particles can be separated continuously. This device can separate particles of three or more different sizes simultaneously, by adding more separation stages. More importantly, all filter stages can work at a constant frequency and voltage, by adjusting the geometry parameters of micropillars and ITO electrodes.

### **2. Materials and Methods**

### *2.1. Related Theories*

The DEP technology is an electrokinetic transport mechanism driven by polarization [32] that could be a useful tool to control the motion of particles. When surrounded by an electric field, homogeneous dielectric particles will be polarized [33,34] and the time-averaged DEP force can be expressed as [35,36]:

$$F\_{DEP} = 2\pi\varepsilon\_m a^3 \text{Re}[K(\omega)] \nabla |E\_{rms}|^2 \tag{1}$$

where *ε<sup>m</sup>* is the permittivity of the medium, *a* is the particle radius, ∇|*Erms*| <sup>2</sup> is the gradient of the square of the RMS electric field, and *K*(*ω*) is the Clausius–Mossotti factor. For a particular sphere, the real part of *K*(*ω*) ranges from −0.5 to 1, and is determined by the frequency of the applied field and the complex permittivity of the medium [37]. *K*(*ω*) can be calculated by:

$$K(\omega) = \frac{\widetilde{\mathfrak{E}}\_p - \widetilde{\mathfrak{E}}\_m}{\widetilde{\mathfrak{E}}\_p + 2\widetilde{\mathfrak{E}}\_m} \tag{2}$$

where *<sup>ω</sup>* is the angular field frequency, &*<sup>ε</sup> <sup>p</sup>* is the complex permittivity of the particle and &*ε<sup>m</sup>* is the complex permittivity of the medium. For isotropic homogeneous dielectrics [38], the complex permittivity can be expressed as:

$$
\widetilde{\varepsilon} = \varepsilon - j\frac{\sigma}{\omega} \tag{3}
$$

with *<sup>j</sup>* <sup>=</sup> √−1, *<sup>ε</sup>* and <sup>σ</sup> being the is permittivity and conductivity, respectively.

*F*DEP is balanced with the drag force of particles in the fluid [39]. For a homogeneous spherical particle in a laminar flow regime, the Stokes drag force can be expressed as:

$$F\_{dra\%} = 6\pi\eta av$$

where *η* is the fluid viscosity, *a* is the radius of the particle, and *v* is the velocity of the particle relative to the fluid.

### *2.2. Experiment*

### 2.2.1. Microfabrication

The microfilter device in the experiment was fabricated using the standard soft lithography technique [22,40,41]. To produce a 50 μm thickness structure, a positive mold was fabricated by using SU-8 3050 (Microchem, Westborough, MA, USA) spin-coated on a 3-inch silicon wafer (ePAK, Austin, TX, USA) at 3000 rpm for 30 s [42]. A mold was realized by using a photolithography aligner device (URE-2000/25, Institute of Optics and Electronics, Chengdu, China). Replicas of the mold were made in polydimethylsiloxane (PDMS, Dow Corning, Midland, MI, USA). As shown in Figure 1A, the microfluidic filter device had two reservoirs with a diameter of 4-mm and a long micro-channel with a length of 2.3 cm, a width of 6 mm, and a depth of 50 μm. Considering the convenience and reliability of connection between the signal generator and ITO electrodes, a 30 mm × 6 mm ITO glass was chosen to fabricate ITO electrodes. In addition, to integrate more separation stages and induced a high strength DEP force under a lower voltage, the distance between electrodes should be miniaturized. Finally, the distance between two ITO electrodes was 200 μm, which was wider than the micropillars (shown in Figure 1A). The filter consisted of a series of hexagonal micropillars (Figure 1C,D), which had two interval sizes inside the microchannel (25 μm and 14 μm). The injection port was fabricated using a 3-mm diameter puncher and the port to plug outlet tubing was fabricated using a 1-mm diameter puncher.

For better observation, indium-tin-oxide (ITO) [43] (220-nm ITO film thickness, 7 Ω/sq.) was used to fabricate the electrodes due to its good transparency. A bare ITO electrode should be carefully washed by using acetone, isopropanol, ethanol, and ultrapure water in turns before use. Then the ITO electrode was fabricated with a SU-8 2000 series negative resin using the standard soft lithography technology. Finally, the PDMS replica was sealed onto the ITO glass with electrodes via O2-plasma activation (PDC-MG, Chengdu, China) of both surfaces.

**Figure 1.** (**A**) Schematic illustration of the microfilter device, the distance of the two ITO electrodes is 200 μm; (**B**) a picture of the microfilter device; (**C**,**D**) scanning electron micrograph of micropillar structures with a height of 50 μm, and the gap of the micropillars are 25 μm and 14 μm, respectively.

### 2.2.2. Experimental Solutions

Polystyrene particles were suspended in a 1-mM phosphate-buffered saline (PBS) buffer with a low electrical conductivity of 0.17 S/m to minimize Joule heating in the filter region [8,44,45]. Polystyrene particles of 37-μm, 16.3-μm, and 9.7-μm diameters were suspended in the PBS buffer at a concentration of 104−105 particles per milliliter. To avoid the adhesion of particles to the channel walls and minimize the interactions between particles, Tween 20 (TP1379, Bomeibio, Hefei, China) was added to the mixture solution at a concentration of 0.1% *v/v* [41].

Haematococcus pluvialis (FACHB, Wuhan, China) and Bracteacoccus engadinensis (FACHB, Wuhan, China) were cultured in Blue-Green Medium (BG11, FACHB, Wuhan, China). A mixture of Haematococcus pluvialis and Bracteacoccus engadinensis were diluted in the culture medium at a concentration of 104−10<sup>5</sup> cells per milliliter, and 0.4 g/mL sorbitol was added to suspend the cells.

### 2.2.3. Experimental Manipulation and Visualization

A pipe connecting the outlet and the peristaltic pump was inserted into the outlet reservoir. The particle solution was added in the inlet reservoir by a pipette and introduced into the micro-channel by suction provided by the peristaltic pump. Serious leakage may occur during solution injection, but suction can help avoid this disadvantage because of the excessive pressure. Before the experiment, the microfluidic filter device was washed with 1-mM PBS buffer without particles for 5 min [46]. The inlet reservoir was brimmed with the particle mixture solution using a 100-μL pipette when the experiment began. As shown in the Figure 2, the sinusoidal signal used in the experiment was supplied by a function generator (SDG1020, Siglent, Solon, OH, USA) and a high-voltage amplifier (ATA-2042, Agitek, Xi'an, China) [41]. The AC electric field was fixed at 10 KHz and 35 V peak-to-peak value during polystyrene particles experiments. For algal cells, the AC electric field was fixed at 8 KHz and 100 V peak-to-peak value [47,48]. A microscope (IX73, Olympus, Tokyo, Japan) was used to monitor

particle motion, and a digital single lens reflex (Canon, Tokyo, Japan) was used to record videos and images in the microfluidic filter device through the microscope.

**Figure 2.** Experimental setup including the microfilter device, microscope, peristaltic pump, signal generator, and high-voltage amplifier.

### **3. Results and Discussion**

### *3.1. Electric Field Gradient Distribution*

Clausius-Mossotti factor value will determine whether the particles were subjected to positive dielectrophoresis (pDEP) or negative dielectrophoresis (nDEP). It depends on the parameters of particles, buffer solution, and applied electric signal. In our separation system, 1mM phosphate buffer was used as a buffer solution. The conductivity σm and permittivity of buffer solution are 0.17 S/m and 7.04 × <sup>10</sup>−<sup>10</sup> F/m, respectively. And the conductivity of polystyrene particles could be calculated by *σ<sup>p</sup>* = 2*Ks*/*r* (r: the radius of the particle). The recommended value of *Ks* for surface conductance is 10−<sup>9</sup> S. The permittivity of polystyrene particles is 2.04 × <sup>10</sup>−<sup>10</sup> F/m. Clausius–Mossotti factor could be calculated by Equation (2) and Equation (3). The results showed that the value was always negative in the frequency range of 0–10<sup>7</sup> Hz. In addition, considering the ITO electrodes are easier for electrolysis at a frequency less than 10 KHz and the same DEP force can be induced by a lower voltage and higher frequency electric signal, 10 KHz was selected as the operating frequency.

To find a proper structure of the filter to separate particles of three sizes in two stages of filters by applying the same voltage, we developed a three-dimensional (3D) model in COMSOL Multiphysics 5.0 (COMSOL, Newton, MA, USA) to investigate the distributions of the electric field gradient and intensity. Figure 3 illustrates the distributions of the electric field gradient at the two stages. The strongest intensity is 9.46 × 1016 V2/m3 in Figure 3A and 1.65 × 1017 V2/m3 in Figure 3B. The electric field gradient at the second stage must be higher than that at the first stage because the DEP force is proportional to the cube of the particle radius (*a*3).

The non-uniform electric field is generated by applying an AC electric field of 35 Vpp and 10 KHz using ITO electrodes placed as shown in Figure 1A [49,50]. The strongest intensity and gradients of the electric field exist near the edge of filter. The particle mixture was injected into the inlet well and flew through the two-stages filter in sequence due to pressure-driven flow. When entering into the non-uniform electric field region, the trajectories of some particles change because of the nDEP force. When the particles passed the first stage filter, 37-μm particles moved to the opposite direction of the flow because of the strong DEP force, while 16.3-μm and 9.7-μm particles passed through the first stage filter. In the second stage filter, 16.3-μm particles were trapped before the second stage filter, while 9.7-μm particles passed through the second stage filter.

**Figure 3.** Distributions of ∇|*Erms*| <sup>2</sup> near the two stages microfilters (**A**,**B**), when 35 Vpp at 10 KHz is applied.

### *3.2. Separation of Three Different Particles*

For particular particles and media, changing the AC frequency influenced the direction of the DEP force. A frequency of 10 KHz was chosen to ensure the particles experienced a negative DEP force. During the experiment, we found that if the flow velocity was fixed, the filters were easily clogged at low AC voltages; however, a strong DEP force generated at high AC voltages, which prevented all particles from passing through the first stage filter. Similarly, if the AC voltage was fixed, a high velocity caused congestion, while a low velocity influenced the separation efficiency. Thus, there is a balance between voltage and fluid velocity [51]. To achieve better continuous microfiltration, voltage and velocity should be optimized. Experiments were conducted with AC voltages ranging from 30−50 Vpp and flow velocities ranging from 0.5−2 μL/min [51]. The best separation condition for the particles was found to be 35 Vpp and 1 μL/min, where in the first stage the 16.3-μm and 9.7-μm particles can pass through and in the second stage the 9.7-μm particles can pass through the filter; none of them got trapped when the voltage was above 35 Vpp.

Figure 4 shows a continuous separation of 37-μm, 16.3-μm, and 9.7-μm polystyrene particles in the microchannel (The complete separation process is recorded in Video S1 and Video S2). The AC voltage and frequency imposed were 35 Vpp and 10 KHz, respectively, and the flow velocity was 1 μL/min. As seen in Figure 4A1–A4, in the first stage, only the 16.3-μm and 9.7-μm particles could pass through the filter, while 37-μm particles were stopped by the filter (25-μm interval). Figure 4B1–B4 show results near the second stage; the filter (14 μm interval) stopped the 16.3-μm particles and pushed them to the roof of hexagon and only allows the 9.7-μm particles to pass through. It can be seen from Figure 4B1–B4 that not all 9.7-μm particles can pass through the second stage filter at once. Some particles were bunched to form a pearl chain and suddenly pass through the filter as a group.

The microfilter would be blocked right after the 50-μL particle mixture is pumped into the micro-channel without applying AC field [51], which limits the popularization and application of microstructure filtration methods. Figure 4 demonstrate that the device kept working when the AC field (35 Vpp, 10 KHz) was applied, indicating the microfluidic device solved the blockage problem successfully, which makes continuous separation possible.

**Figure 4.** The separation process of 37-μm, 16.3-μm and 9.7-μm particles. (**A1**–**A4**) The first stage. (**B1**–**B4**) The second stage.

### *3.3. Separation of Algae Cells*

Haematococcus pluvialis and Bracteacoccus engadinensis were also used to verify the feasibility of the device. Haematococcus pluvialis is famous for its high content of astaxanthin, which is the strongest antioxidant in nature and plays an important role in aquaculture, health care, and cosmetics industries [52]. Haematococcus pluvialis cells are often mixed with other algae cells in nature, thus, it is necessary to sort and purify Haematococcus pluvialis. Bracteacoccus engadinensis cells were mixed with Haematococcus pluvialis cells to mimic the natural condition.

The sizes of Haematococcus pluvialis cells are between 15 and 30 μm, and the sizes of Bracteacoccus engadinensis cells are 10 to 15 μm. In this study, we applied an AC signal with a voltage of 100 V and a frequency of 8 KHz, and the flow velocity is the same as the particle separation. As shown in Figure 5A1–A4, all of the Bracteacoccus engadinensis cells and some of the smaller Haematococcus pluvialis cells can pass through the filter unimpededly in first stage with a pore size of 25 μm, but larger Haematococcus pluvialis cells larger than 25 μm cannot pass. Like the 37-um particles, large Haematococcus pluvialis cells were pushed away by the negative dielectrophoretic force near the entrance of the first stage filter. The results of the second stage are depicted in Figure 5B1–B4. Smaller Bracteacoccus engadinensis cells can pass through the filter with the size of 14 μm, while Haematococcus pluvialis cells and large Bracteacoccus engadinensis cells were trapped before the filter under dielectrophoretic force. The complete separation process is recorded in Video S3 and Video S4.

Due to large size, only a few Haematococcus pluvialis cells can pass the first stage since more Haematococcus pluvialis cells can been seen before the entrance of the first stage filter (Figure 5A1–A4), while only a few Haematococcus pluvialis cells can be seen in the second stage (Figure 5B1–B4). Accordingly, because of the small size, all of the Bracteacoccus engadinensis cells can pass through the first stage, and most of them can pass through the second stage and reach the outlet directly. After separation, there are only Haematococcus pluvialis cells left in the channel before the first stage, Bracteacoccus engadinensis cells in the outlet, and merely a small amount of these two algae cells with similar sizes in the channel between the first stage and the second stage. Thus different sizes of Haematococcus pluvialis cells and Bracteacoccus engadinensis cells were separated.

**Figure 5.** Separation process of Haematococcus pluvialis cells and Bracteacoccus engadinensis cells at a voltage amplitude of 100 V and a frequency of 8 KHz. (**A1**–**A4**) The first stage. (**B1**–**B4**) The second stage.

### **4. Conclusions**

We presented a microfluidic filter device that combines the advantages of negative dielectrophoretic force and microfilters, to separate particles of different sizes. Microfilter is one of the most widely used particle/cell separation methods due to simple operation. However, blockages limit its popularization and applications. We were committed to solving the blockage problem of microfilter and purify Haematococcus pluvialis cells using the simplest filter structure in this study. A 3D model was developed to analyze electric field distribution. Based on the simulation results and particles size, micropillars based separation device was designed and fabricated. In addition, the geometry parameters of micropillars were optimized to ensure two separation stages work by using a generator. The feasibility of this method was demonstrated by the continuous flow separation of polystyrene particles with three different sizes. Haematococcus pluvialis cells and Bracteacoccus engadinensis cells were also separated in this device without blockages phenomenon.

Considering our device is based on the filtration theory, this microfluidic device could separate complex sample, which contains many kinds of (bio)particles with different size, by integrating several separation stage with appropriate micropillars and applying appropriate AC signal. By optimizing the geometry parameter of micropillars, and adjusting the distance between two adjacent micropillars, circulating tumor cells, white blood cells, red blood cells, or blood plasma could be separated from whole blood samples [53–55].

The developed microfluidic filter device can be conveniently fabricated and generates a strong dielectrophoretic force near the filter of each stage. The device can separate particles of different sizes efficiently with minimal Joule heating due to the use of pressure-driven flow and AC electric field. However, some particles, which should theoretically pass through the filter, were trapped due to the too strong DEP force and adhesion between particles. It reduces the separation efficiency [6]. Additionally, (bio) particles collection structures should be integrated, to avoid too much (bio) particles block the flow path.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-666X/10/2/103/ s1, Video S1: The separation process of particles in the first stage, Video S2: The separation process of particles

in the second stage, Video S3: The separation process of Haematococcus pluvialis cells and Bracteacoccus engadinensis cells in the first stage, Video S4: The separation process of Haematococcus pluvialis cells and Bracteacoccus engadinensis cells in the second stage.

**Author Contributions:** N.H. conceived and designed the experiments; D.Y. built up the device and performed the experiments; X.Z. simulated the electric field gradient; X.H. contributed microchip fabrication; D.Y. wrote the paper; N.H. and J.Y. revised the paper.

**Funding:** National Natural Science Foundation of China: 21827812, 31571005, 81871450; Natural Science Foundation of Chongqing: cstc2018jcyjAX0389.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Micromachines* Editorial Office E-mail: micromachines@mdpi.com www.mdpi.com/journal/micromachines

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18