A microfluidic platform to isolate avian erythrocytes infected with ...

3 downloads 0 Views 1MB Size Report
Yu-Hsiang Hsu & Peiran Lu & Judith L. Coleman &. William C. ... P. falciparum, human RBCs start to lose their biconcave ... vessels like the one infected by P. falciparum. ...... thank Dr. Anthony A. James for supplying avian blood samples, Dr.
Biomed Microdevices DOI 10.1007/s10544-011-9569-8

A microfluidic platform to isolate avian erythrocytes infected with Plasmodium gallinaceum malaria parasites based on surface morphological changes Yu-Hsiang Hsu & Peiran Lu & Judith L. Coleman & William C. Tang

# Springer Science+Business Media, LLC 2011

Abstract This paper reports on a microfluidic platform to isolate and study avian red blood cells (RBCs) infected to various degrees by the malaria parasite Plasmodium gallinaceum. The experimental findings point to the feasibility of using the morphological changes on the surface of the malaria infected avian RBC (miaRBCs) as biomarkers for diagnosis. A glass substrate with a controlled surface roughness was used as part of a polydimethylsiloxane (PDMS) microfluidic channels. When wholeblood samples were introduced into the channels, the miaRBCs would be preferentially slowed and eventually become immobilized on the roughened surface. The surface lesions and furrow-like structures on the miaRBC surfaces offered a markedly higher probability to interact with the roughened substrate and allowed the cells to become imobilized on the surface. The captured miaRBCs were from blood samples at various degrees of infection at 3.2%, 3.9%, 9.1%, 13.4%, 20.1%, 28%, and 37%. It was observed that the miaRBCs could be selectively captured under a wall shear rate between 2.1 to 3.2 s−1, which was directly proportional to the flow rate through the channels. This capture rate could be improved by increasing the channel length and finer flow control. It was also found that a

Electronic supplementary material The online version of this article (doi:10.1007/s10544-011-9569-8) contains supplementary material, which is available to authorized users. Y.-H. Hsu : P. Lu : W. C. Tang (*) Department of Biomedical Engineering, University of California, Irvine, CA 92697-2715, USA e-mail: [email protected] J. L. Coleman Department of Molecular Biology & Biochemistry, University of California, Irvine, CA 92697-3900, USA

roughened glass substrate with ten-point-height larger than the depth of surface lesions and furrow-like structures of miaRBCs showed a substantial enhancement on the number of immobilized infected RBCs. These findings indicated that surface morphologies, including surface lesions and furrow-like structures, can serve as an alternative biomarker for malaria diagnosis. Keywords Malaria biomarker . Malaria separation . Cellular diagnostics . Cell biomechanics . Plasmodium gallinaceum

1 Introduction Recently, advances in cell mechanics research tools have enabled the study of mechanical differences between normal red blood cells (RBCs) and malaria-infected ones (miRBCs) (Suresh et al. 2005). P. falciparum, in particular, is the most extensively studied due to the sequestration behavior of its infected RBCs compared to other three common strains (Hänscheid et al. 2001). Upon infection of P. falciparum, human RBCs start to lose their biconcave shape and become more spherical. During asexual erythrocytic stages of the parasitic life cycle, the stiffness of the cell body is increased by more than ten times, and knoblike protrusions are formed on the cell surfaces at the early trophozoite stage (Bannister and Mitchell 2003). These protrusions mediate the cytoadhesion behavior of miRBCs to vascular endothelium, causing miRBC stickiness (Cooke et al. 2005; Nash et al. 1992). The roles of wall shear stress and adhesive strength between miRBCs and RBCs were studied by culturing vascular endothelium or coating purified receptors in microfluidic channels (Antia et al. 2007; Nash et al. 1992). These findings suggested that at least three mechanical biomarkers could potentially be used

Biomed Microdevices

to diagnose malaria: (a) elevated stiffness of the cell body (Hou et al. 2010; Shelby et al. 2003), (b) cytoadhesive characteristics of cell surface (Antia et al. 2007; Nash et al. 1992), and (c) altered cell morphology. Similar to human malaria, the Plasmodium gallinaceum infected avian RBCs loose their oval shape and have modified surface morphology (Nagao et al. 2007). P. gallinaceum parasite leaves a visible lesion on the cell surface after invasion, and it remains throughout the asexual erythrocytic stages. Slender furrow-like structures start to form on the cell surface at the early trophozoite stage. These features are uniformly distributed on the cell surface and persist in the remaining cycle with identical density and distribution. The dimensions taken from AFM of this furrow-like structure is around 57.3 nm in width, 7.6 nm in depth, and 225 nm to 750 nm in length (Nagao et al. 2007). Studies also showed that malaria-infected avian RBCs (miaRBCs) did not sequester efficiently to microvessels like the one infected by P. falciparum. We hypothesized that the surface morphological change of RBCs due to malaria infection could also be a biomarker for diagnosing malaria. To verify this hypothesis, a microfluidic device with controlled wall shear rate and predefined total screen volume were developed. Experimental findings indicated that the surface lesions and furrow-like structures of infected avian RBCs could interact with a glass substrate with measured roughness under a controlled wall shear rate. This result suggested that surface morphological change of infected RBCs can serve as alternative biomarkers for cellular diagnosis in addition to the elevated stiffness and increased adhesiveness.

2 Materials and methods 2.1 Microplatform design A microfluidic platform with a long and narrow channel was designed and fabricated to study the feasibility of using the surface morphology of miaRBCs as the biomarker for eventual use in a malaria diagnostic system. The basic configuration of the platform is illustrated in Figs. 1(a) and (b). This microfluidic platform was made from standard polydimethylsiloxane (PDMS) molded from an SU-8 mold and bonded onto a 170 μm thick glass cover slide with controlled surface roughness. The diagnostic channel is 11 mm long, 340 μm wide, and 50 μm deep, and is flanked by two perfusion channels with 2 μm shallow gaps (Hung et al. 2005; Lee et al. 2006). The diagnostic channel and perfusion channels are shown in Figs. 1(c) and (d), which were filled with yellow and green dyes, respectively. The total screen volume in the diagnostic channel is 0.19 μl, and the square cross section of each of the perfusion channels is 50 μm on a side.

(a)

Diagnostic channel

170 µm glass slide

(b)

Controlled wall shear rate Diagnostic channel

2 µm geometrical barriers

Perfusion channels

(c)

(d)

Fig. 1 Conceptual illustrations of the (a) longitudinal side view and (b) lateral side view of the microfluidic platform designed for avian malaria diagnosis, and (c) (d) pictures of the microfluidic platform, where white ovals and gray ovals represent normal and infected RBCs

The diagnostic channel was designed to be 50 μm in depth, which was shallow enough to achieve reasonable probability of interaction between the miaRBCs and the roughened glass substrate while still providing enough space for normal cells to flow away from the channel. This is based on the dimension of a normal avian RBCs (naRBCs), which is around 13.23 μm in length, 6.81 μm in width, and 3.82 μm in thickness (Gaehtgens et al. 1981). The corresponding dimensions of miaRBCs are bigger than normal ones, depending on the stages of infection and how many parasites have grown inside them. The channel depth of 50 μm would then result in size-to-channel-depth ratios ranging from 1:3 to 1:15, depending on the orientation of the RBCs. With the assumption that miaRBCs and naRBCs are randomly distributed inside the blood stream, shallower channels, and thus lower size-to-channel-depth ratios, would exhibit higher probabilities of contacts between the RBCs (both normal and infected) and the substrate. It is hypothesized that miaRBCs will exhibit higher friction and stronger adhesiveness to the roughened substrate than naRBCs because the former possess lesion and furrowlike structures on their roughened cell membranes. As a result, miaRBCs will slow down and eventually become immobilized on the substrate surface while a majority of naRBCs will not. To further increase the probability of interaction between miaRBCs and the substrate, the channel should be made as long as possible and the flow rate to be as slow as possible while not excessively long or slow that the unintended immobilization of naRBCs becomes a concern. In the present work, an 11-mm long diagnostic channel was chosen to facilitate capturing a high ratio of miaRBCs to naRBCsthroughout the channel [Fig. 1(a)]. In order to preserve the unique surface morphology of miaRBCs, all experiments were performed on fresh whole blood drawn directly from avian blood vessels and loaded

Biomed Microdevices

into the channel within 30 min. A steady fluid flow mimicking microcirculation was maintained in the diagnostic channel during blood sample loading, and thus minimizing rouleaux formation and blood clotting. To verify that the captured cells were miaRBCs. DAPI fluorescent stain was used to stain the nuclei of the parasites. DAPI was introduced through the two perfusion channels, which were connected to the diagnostic channel with 2 μm shallow openings [Fig. 1(b)]. Before introducing the stain, the residual anticoagulant agent and blood plasma in the diagnostic channel must first be removed. A small amount of the anticoagulent agent, heparin, was introduced into the diagnostic channel from the heparin-coated capillary tube used for collecting blood sample. This was done to suppress blood coagulation before loading the blood into the diagnostic channel (Adams et al. 2006; Xiao et al. 1996). Removing heparin from the sample was necessary because it could alter the surface morphology of the cell membranes (Freidlin 1985). Instead of injecting through the inlet port of the diagnostic channel, fresh media was introduced through the two perfusion channels to gradually dilute and replace the heparin in the sample. This design provided high flow resistance into the diagnostic channel while allowing diffusion to become the dominant mass transport mechanism (Hung et al. 2005; Lee et al. 2006). This geometrical barrier provided a means to effectively remove heparin with minimal unwanted flow perturbations during the capturing process. Subsequently, DAPI fluorescent stain was similarly introduced through these two perfusion channels with minimal flow perturbations in the diagnostic channel. Finally, the captured cells were visually verified with a 100× oil lens from an inverted microscope through the 170 μm thick glass cover slide. 2.2 Device fabrication Figure 2 shows the baseline microfabrication steps of the microfluidic platform for avian malaria diagnosis. The PDMS microchannels was microfabricated based on standard PDMS micromolding (Xia and Whitesides 1998). The SU-8 mold was made using a technique similar to that developed by Hung et al. (2005). A (100) silicon wafer was cleaned with standard RCA-1 cleaning and a 5-minute dip in 2% HF, followed by 30 min of oven dehydration at 120°C. A two-step exposure technique was used to create the SU8 mold. The first layer of 2 μm-thick SU-8 2002 negative photoresist (Micro Chem) was spun onto the cleaned wafer. The standard photolithography process was used to define the 2 μm shallow openings between the perfusion channels and the diagnostic channel [Fig. 2(a)]. After developing this first layer, the second layer of 50 μm-thick SU-8 50 negative photoresist (Micro Chem) was spun over the 2 μm-tall features, and perfusion channels and diagnostic channel were

(a)

(b)

(d)

(e)

(c) 1st SU-8

Glass

2nd SU-8

PDMS

Si

Fig. 2 Cross-sectional drawing of the process flow of the microfluidic platform

then defined with the second photolithography step [Fig. 2 (b)]. After developing the features, this SU-8 mold was finalized with hard baking at 175°C for 30 min. Before casting PDMS, the SU-8 mold and silicon wafer surfaces were silanized with trichlorosilane (C8H4Cl3F13Si) in a vacuum chamber to ease subsequent demolding. A 10-to-1 ratio PDMS prepolymer and curing agent (Sylgard 184, Dow Corning) was then casted on the SU-8 mold to create a 2-mm thick PDMS layer [Fig. 2(c)]. After degassing under vacuum, the PDMS was cured in a 65°C oven overnight and demolded in a laminar flow hood [Fig. 2(d)]. An 18G needle was used to punch 1-mm-diameter holes for inlets and outlets. This PDMS layer and a clean 170 μm thick cover glass slide were plasma treated in an atmospheric plasma (Harrick Scientific, NY) at 200 mTorr and 200 W for 5 min to promote adhesion. These components were then quickly brought into contact and placed into a 65°C oven for 5 min to complete the final assembly of the microfluidic platform [Fig. 2(e)]. Prior to bonding to the PDMS structures, the surface of the glass substrate (Corning No. 1 glass cover slides) was roughened to various controlled degree by dipping in different concentration of hydrofluoric acid (HF). Figure 3 (a), (b) and (c) show the surface topologies of glass substrates in three conditions: (1) the untreated condition, (2) treated with 2% HF for 1 min, and (3) treated with 10% HF for 2 min, respectively. AFM measurements showed that the root-mean-squared (rms) surface roughness quantities were 0.1892 nm, 0.3988 nm, and 2.3286 nm, and the ten-point-heights were 2.0991 nm, 3.1988 nm, and 14.7238 nm, respectively. These glass substrates offered nano scale protrusions close to the dimensions of surface lesions and furrow-like structures of the miaRBCs. 2.3 Experimental To verify the hypothesis of using the unique surface morphology of miaRBCs as a biomarker for avian malaria diagnosis, the selectivity of roughened substrate to naRBCs

Biomed Microdevices

(a)

(b)

(c)

2.5 nm

3.4 nm

14.95 nm

0.0 nm

0.0 nm

0.0 nm

Fig. 3 3-D AFM images of glass substrates at (a) untreated condition, and treated with (b) 2% HF for 1 min and (c) 10% HF for 2 min, where scan area was 2 μm by 2 μm

and miaRBCs was first investigated under different ranges of wall shear rates. A fully developed fluid flow was created and maintained in the diagnostic channel with a well-controlled fluid introduction sequence. This was designed to prevent cells from clotting inside the diagnostic channel while maximizing the probability of interaction between miaRBCs and the substrate. Under a range of controlled flow fields, it was observed that the greate majority of naRBCs readily flowed away from the diagnostic channel. On the contrary, the miaRBCs carried by the fluid flow rolled and reduced their speeds when they interacted with the roughened surface, and were finally immobilized at locations throughout the channel. After all the normal cells flowed away from the diagnostic channel, the number of captured miaRBCs was visually identified and the stage of infection was determined by calculating the projected ratio of miaRBCs to naRBCs. The low affinity of smooth naRBCs to roughened substrate was verified by flowing normal RBCs onto the platform with similar substrate under identical wall shear rates. The immobilization efficiency and sensitivity were quantified by total number of naRBCs or miaRBCs captured versus total number of cells loaded. Figure 4 illustrates the loading sequence of the diagnostic channel, where the arrows below each figure indicate the direction of fluid flow, and the lengths of the arrows represent relative flow rates. This protocol was developed to control the total volume and concentration of blood sample loaded and to prevent blood clotting throughout the process. It also significantly minimized loss of RBCs due to gradual precipitation and coagulation in the dead volume of the fluid pipeline, such as the syringe in the syringe pump, where there is no agitation source. Four 200 μl pipette tips were first snuggly inserted into the inlets and outlets of the diagnostic channel and the perfusion channels (not shown). The perfusion channels and diagnostic channel were then filled with 80 μl fresh media through the inlet of the perfusion channels. The injection was performed until the media level of all four pipette tips reached similar heights and all air bubbles were flushed out (not shown). This

platform was then left on the microscope stage to allow the media levels to equalize, at which point each tip would have 20 μl of media [Fig. 4(a)]. This process took about 10 min. The media was prepared with CO2-independent culture media (GibcoTM, Invitrogen Co.) consisting of 1% penicillin/streptomycin and 4 mM L-glutamine. Heparin coated capillary tube was used to collect blood from chicken infected with the P. gallinaceum parasites. The infection percentage of the blood samples was verified with standard blood film method (Bain et al. 2005). A measured volume of 10 μl of blood sample was carefully mixed with 70 μl CO2-independent culture media. Seventy-five μl of this mixture was gently and slowly mixed into the 20 μl media in the pipette tip positioned in the inlet of the diagnostic channel with gel loading tips. This step diluted the blood sample to 10% v/v and allowed RBCs to spread evenly in the diagnostic channel into a monolayer, maximizing the likelihood for each cell to interact with the substrate. Further, the dependency of shear rate on blood viscosity was also minimized in the diluted blood sample. The concentration was calculated based on the normal cell count of avian blood, which is about 2.58 million RBCs per microliter of blood (Gaehtgens et al. 1981). Figure 5 shows the loaded blood sample at this concentration in the diagnostic channel. The pipette tip at the diagnostic channel inlet, which at this point contained 95 μl of diluted blood sample, resulted in a hydrostatic pressure difference in reference to the 20 μl media in the pipette tip at the outlet. The loaded RBCs then flowed into the diagnostic channel under this differential pressure [Fig. 4(b)]. At the same time, another 75 μl culture media was added to the inlet of the perfusion channels to suppress any pressure gradient between the perfusion channels and the diagnostic channel. The purpose of this step was to prevent the loss of normal RBCs, which could be pushed out to the perfusion channels through the 2 μm shallow openings with their deformable cell bodies. In addition, this created a continuous and steady fluidic flow in the perfusion channels, replacing plasma and anticoagulant agent in the diagnostic channel with fresh media

Biomed Microdevices Fig. 4 Schematic illustrations of blood sample loading sequence, where starts from (a) fill diagnostic channel with fresh media, (b) add blood sample, (c) flow blood sample into channel, (d) wait until blood sample filling up the channel, (e) add media to reverse the flow and create a slow flow rate, and (f) inspect captured miaRBCs. Inlet and outlet are located on the left and right hand sides, respectively

(a)

(d)

(b)

(e)

(c)

(f)

through diffusion from the perfusion channels. The loaded RBCs flowing into diagnostic channel were then monitored and followed under the microscope [Fig. 4(c)]. Once the RBCs reached the outlet of the diagnostic channel, different volumes of media were added to the pipette tip at the outlet to achieve controlled reverse flow. It was during the reverse flow that the relationship between the immobilization efficiency and the shear rates was studied [Fig. 4(d)]. An identical amount of DAPI fluorescent stain at 1 mg/ml in PBS was simultaneously added to the outlet of perfusion channels to both balance the pressure and also stain the captured miaRBCs. This step created a controlled hydrostatic potential difference between the inlet and outlet of the diagnostic channel and reversed the flow with a controlled wall shear rate. The optimized pressure difference was 1.0 to 1.5 mmH2O. The final volume of media and DAPI solution added to the outlets was 85 μl. Since the volume of the diagnostic channel was only 0.19 μl, the wall shear rate can be controlled within a reasonably tight tolerance to effectively flush the naRBCs from the channel while allowing miaRBCs to interact with the roughened substrate [Fig. 4(e)]. The total number of cells loaded was estimated to be around 51,000. Finite-element simulation with COMSOL Multiphysics 3.5a was used to estimate the

induced shear rate. Based on the simulation results, the wall shear rates were calculated to be between 2.1 to 3.2 s–1. It was observed that, under low flow rates, blood clotting could gradually develop at locations close to the diagnostic channel inlet if the blood sample was insufficiently diluted to the point that the resulting concentration was higher than 15% v/v. This condition would decrease and eventually impede fluidic flow. In this case, the developed wall shear rate could not be maintained for a period long enough for all the naRBCs to flow out of the diagnostic channel. This condition was avoided in our experiment with 10% v/v dilution of the blood sample. As the naRBCs flowed out of the diagnostic channel, the fluid inside the channel was flushed with the remaining media until the hydrostatic potential difference decreased to zero [Fig. 4(f)]. The overall process was performed under continuous flow to prevent clotting and was completed within 30 min, which was short enough to preserve the natural RBC characteristics. Once the captured miaRBCs were stained with DAPI fluorescent stain, they were then visually identified and quantified with a 100× oil lens under 1000× total magnification. The whole experiment was performed on an Olympus IX51 inverted microscope with a Hamamatsu high resolution gray scale CCD camera at room temperature. Identical procedures were conducted with naRBCs to investigate their immobilization rate on roughened substrate as the control.

3 Results and discussion 3.1 Wall shear rate and capture efficiency

Fig. 5 Micrographs of diluted blood sample loaded in the diagnostic channel. Scale bar is 50 μm

To investigate the influence of the magnitude of wall shear rates on the degrees of immobilizations of both miaRBCs and naRBCs, different flow rates were applied while cell speeds were traced after steady-state flow was established. Figures 6 and 7 show the flow speeds of miaRBCs and naRBCs under two different wall shear rates. The data

Biomed Microdevices

(a) miaRBCs (Counts) 10 8 6 4 2 0

0

10

20

30 Speed (µm·s–1)

40

50

60

20

30 Speed (µm·s–1)

40

50

60

(b) naRBCs (Counts) 10 8 6 4 2

selection of 10 miaRBCs and 10 naRBCs from each field were traced. Figure 6(a) and (b) show the distributions of measured cell speeds under 4.70 s−1 wall shear rate at 20 min after loading. The flow speeds of the 50 miaRBCs were all below 15 μm·s−1, while that of the 50 naRBCs spread over the range from 9 to 60 μm·s−1. Worthy of note is that one of the 50 miaRBCs became immobilized on the substrate. It was also found that the percentage of infected cells, determined to be 7.8% with standard blood film method prior to this experiment, was observed to be 8.82% per image field. Figure 7(a) and (b) show the measured flow speeds of miaRBCs and naRBCs under 2.14 s−1 wall shear rate at 15 min after loading. Figure 7(a) shows that 13 out of 50 miaRBCs were already immobilized and the speed of others are reduced to below 5 μm·s−1. At the same time, the flow speeds of naRBCs, again, spread over a much broader range, from 2 to 27 μm·s−1 [Fig. 7(b)], and no naRBCs were imobilized even at this slower flow rate. 3.2 Immobilization rate of naRBCs and miaRBCs

0 0

10

−1

Fig. 6 Distributions of cell speed under 4.70 s wall shear rate, where (a) and (b) are 50 traced miaRBCs and naRBCs

points were taken from five 400× fields at different locations along the diagnostic channel, and a random (a) miaRBCs (Counts) 15 12 9 6 3 00

6

12 18 Speed (µm·s–1)

24

30

12 18 Speed (µm·s–1)

24

30

(b) naRBCs (Counts) 15 12 9 6 3 0 0

6

Fig. 7 Distributions of cell speed under 2.14 s−1 wall shear rate, where (a) and (b) are 50 traced miaRBCs and naRBCs

Table 1 presents a comparison of the immobilization rates of miaRBCs and naRBCs from fresh blood samples at various infection stages. Healthy blood samples with no infection were included as the control. The flow rates were controlled to be around 2.14 s−1, and untreated glass substrates were used. The infection stage was identified by standard blood film method prior to the experiments and was listed as % infection. For the control experiments with fresh, healthy blood samples, 25 to 37 naRBCs were immobilized on the glass substrate, particularly towards the end of the fluid flow process. Similar numbers of immobilized naRBCs were observed throughout the entire set of infected blood samples at different infection stages. This could be due to the decreased wall shear rate near the end of the process when a small portion of remaining naRBCs were still lingering in the diagnostic channel. However, a more careful experiment with a fine control to achieve constant flow speed is needed to verify this speculation. The number of immobilized naRBCs could potentially be minimized by maintaining a constant flow speed. To study the relationship between the numbers of captured miaRBCs to actual infection stage, three parameters were defined and listed in Table 1. The sensitivity to miaRBCs is defined by the ratio of the total number of immobilized miaRBCs divided by the estimated total number of cells loaded, which is about 51,000 cells. The efficiency of capturing miaRBCs is defined by the ratio of total number of immobilized miaRBCs divided by the total number of miaRBCs loaded, which was estimated by multiplying % infection with the total number of loaded cells. Similarly, the sensitivity and efficiency of naRBCs are also listed. The specificity is defined by the total number of

Biomed Microdevices Table 1 Comparison of immobilization sensitivity, efficiency, and specificity between miaRBCs and naRBCs

% Infection # of miaRBCs Sensitivitya Efficiencyb # of naRBCs Sensitivitya Efficiencyb Specificityc a

Host A

Host B

Host C

Host D

Host E

Host F

Host G

0.0% 0 0.000% 0.000% 25 0.049% 0.049% NA

0.0% 0 0.000% 0.000% 37 0.073% 0.073% NA

3.2% 55 0.108% 3.370% 20 0.039% 0.041% 73.33%

3.9% 76 0.149% 3.821% 28 0.055% 0.057% 73.08%

9.1% 363 0.712% 7.822% 7 0.014% 0.015% 98.11%

13.4% 977 1.916% 14.296% 16 0.031% 0.036% 98.39%

20.1% 1093 35.719% 17.771% 9 0.029% 0.037% 99.18%

Sensitivity = the number of immobilized cells / total number of loaded cells

b

Efficiency = the number of immobilized cells / total number of specific cell type loaded, i.e., miaRBCs or naRBCs

c

Specificity = the number of immobilized miaRBCs / total immobilized cells

immobilized miaRBCs divided by the total number of captured cells. Among these samples, the blood sample from Host G was under a serious infection of up to 30% 12 h before the experiments. It suffered a large amount of RBC loss and the total cell count was only 60% compared to other samples. After studying different blood samples at different infection stages from different host chickens, it was observed that both sensitivity and efficiency of capturing miaRBCs increased with higher % infection. This implies that more miaRBCs reached early trophozoite stage. The specificity is also increased to 99.18% at 20.1% infection, as shown in Fig. 8(a). The relatively low sensitivity and specificity of immobilized miaRBCs at low infection stages could be enhanced by increasing the length of the diagnostic channel to maximize the probability of interaction between the miaRBCs and the substrate. Furthermore, a roughly linear relationship was observed between the % infection and the miaRBC immobilization efficiency, as shown in Fig. 8(b). The slope of the best-fit line is 0.919, intersecting the y-axis almost at the origin. This is consistent with the assumption that each cell in the diagnostic channel has equal probability to interact with the substrate and is independent of the presence or density of other cells. Likewise, each miaRBC has equal probability to be immobilized on the substrate. As the % infection increases, the total number of miaRBCs loaded into the diagnostic channel was also increased proportionally. Since each miaRBC has equal immobilization probability, the total number of captured miaRBCs under identical conditions should also hold a linear relationship to the total number of miaRBCs loaded. Thus, it follows that the relationship between % infection and miaRBCs capturing efficiency is linear. This leads to an important practical guideline that it is not necessary to capture all the miaRBCs loaded in the diagnostic channel to achieve an accurate

(a) Specificity (%) 100

75

50

25

0

0

5

10

15

20

25

20

25

Infection %

(b) Efficiency (%) 25

20

15

10

5

0 0

5

10

15

Infection % Fig. 8 The relationship between the % infection and immobilization (a) specificity and (b) efficiency of miaRBCs

Biomed Microdevices

Fig. 9 A Micrograph of immobilized miaRBCs stained with DAPI in the diagnostic channel under 200× magnification. The % infection was 20.1% and the scale bar (lower right) is 10 μm

diagnosis. By controlling the concentration of blood sample and the length of the diagnostic channel, a linear relationship could be identified and be used as the reference to perform diagnosis on unknown blood samples. The clinical implication is that a small portion of captured miaRBCs could be used to determine the stage of infection without the need to identify miaRBCs from a large pool of naRBCs. Note also that the immobilization sensitivity and efficiency

Fig. 10 Micrographs of immobilized miaRBCs stained with DAPI in the diagnostic channel under 1000× magnification, where (a) to (h) are miaRBCs at early infection stages, (i) to (l) are miaRBCs at trophozoite stage, and (m) to (p) are miaRBCs at schizont stage. The scale bar (lower right) is 5 μm

of naRBCs are below 0.05% independent of the infection stages. This implies a very low probability of immobilizing naRBCs on the substrate across all samples. Figure 9 shows a micrograph of a typical image field of the diagnostic channel after the capturing process. This image was taken from the experimental result of Host G blood sample listed in Table 1, and was a pseudo-color image of combining the DAPI-fluorescent and bright field images. Both the invading parasites and the nuclei of RBCs (blue) could be identified in the cell bodies of the RBCs (gray). This image also shows that miaRBCs can be readily identified at 200× magnification after naRBCs were flushed away by fluid shear flow. However, interference from DAPI stained nuclei could still confound the identification of captured cells. Raising or lowering the focal plane was also necessary to identify some parasites that lied outside the sample focal plane. Figure 10 shows the micrographs of immobilized miaRBCs at different asexual erythrocytic stages identified in the diagnostic channel at 1000× magnification. All of these images were taken by combining DAPIfluorescent and bright-field light sources. Figure 10(a) to (d) are images of miaRBCs at the early stages of infections, and Fig. 10(e) to (h) are miaRBCs with multiple infections. These results indicate that the surface lesion caused by P. gallinaceum parasite invasion could significantly promote

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)

(m)

(n)

(o)

(p)

Biomed Microdevices 300

3.3 Influence of surface roughness

250 200 % 150 100 50 0 20 %

28 % Infection %

37%

Fig. 11 A Comparison chart of increased percentages of captured miaRBCs by using glass substrates at as is condition and treated with 10% HF with three different blood samples

interaction with roughened substrates and eventually be immobilized even at early stages of infection. Figure 10(i) and (j) show the images at two different focal planes of a captured miaRBC at trophozoite stage, and Figs. 10(k) and (l) show another at two different focal planes. Figure 10(m) to (o) show one of the captured miaRBCs at schizont stage focused at different focal planes, and Fig. 10(p) shows another miaRBC also at similar stage of infection. These results indicate that the changes in biomechanical properties of RBCs caused by malaria infection could serve as biomarkers for P. gallinaceum parasite diagnosis.

Fig. 12 Time-lapse micrographs of (a–d) an immobilized miaRBC and two moving naRBCs at 15 min after loading, (e–h) and (i–l) are two moving miaRBCs interacted with roughened substrates and been immobilized at 10 and 15 min after loading. See complete time-lapse sequences in the supplementary videos. Scale bars are 10 μm

To verify that the capturing mechanism was influenced by the protrusions of the roughened substrates, identical experiments were conducted with HF treated glass substrates. The experimental results of using 2% HF treated substrate with ten-point-height of 3.1988 nm [Fig. 3(b)] did not show statistical improvement in immobilization sensitivity over untreated substrates with ten-point height of 2.0991 nm [Fig. 3(a)], showing variations between −5.5% to 5.5%. For reference, the depth of the furrow-like structures on miaRBCs were around 7.6 nm, more than twice the tenpoint heights of both untreated and 2% HF treated glass substrates. However, significant increase in capturing sensitivity was observedby using 10% HF treated glass substrate [Fig. 3(c)], which had ten-point-height around 14.7238 nm, roughly twice the depth of the furrow-like structures. Figure 11 compares the percentage of increased capturing efficiency of 10% HF treated glass substrate to untreated one for three different blood samples with different % infections. The shear rate was controlled to be around 2.14 s−1. To accommodate the high % infection, the RBC concentrations were diluted to 5% v/v. The total captured miaRBCs increased by 34.4%, 37.9%, and 194.5% with 20%, 28% and 37% blood samples, respectively. This demonstrated that a roughened substrate with nano-scale protrusions twice the feature size of the surface lesions and furrow-like structures could provide a separation mechanism to capture miaRBCs.

(a) 0s

(e) 0s

(i) 0s

(b) 5s

(f) 0.9s

(j) 1s

(c) 10s

(g) 2.0s

(k) 2s

(d) 15s

(h) 4.9s

(l) 7s

Biomed Microdevices

To further investigate the immobilization behaviors of naRBCs and miaRBCs, several normal and infected RBCs were traced by using fluorescent inverted microscope at 1000× magnification, where malaria parasites and nuclei of RBCs were stained with DAPI fluorescent dye. The 2% HF treated glass substrates were used in these studies. Figure 12 (a) to (d) are time-lapse micrographs of an immobilized miaRBCs and two naRBCs at 15 min after loading (supplemental video S1.mpg). It was observed that naRBCs in this experiment did not interact with the substrate and flowed along with the shear flow. Figure 12 (e) to (h) are time-lapse micrographs of an miaRBC interacting with the substrate and became immobilized. In the complete time-lapse video of this experiment, supplemental video S2.mpg, it was shown that miaRBCs could be captured once it interacted with the substrate. Figure 12(i) to (l) constitute a different set of time-lapse micrographs of another miaRBC interacting and immobilizing on the substrate. Similar to the other experiment, the miaRBC was captured once it interacted with the substrate. It was observed that this miaRBC wiggled slightly before it became completely immobilized, as shown in supplimental video S3.mpg.

4 Conclusions The experimental results from studying avian RBCs infected with P. gallinaceum parasites validated the hypothesis that the unique surface morphology of miaRBCs could potentially be used as an alternative biomarker to perform malaria diagnosis. The surface lesions and furrow-like structures could interact with a surface with roughness exhibiting comparably-sized nano-scale protrusions, resulting in a drag force that slowed down and eventually immobilized miaRBCs. It was also found that the immobilizing efficiency of miaRBCs exhibited a roughly linear relationship with the degrees of infection. Immobilized miaRBCs were found to consist of cells infected with P. gallinaceum parasites at different asexual erythrocytic stages, indicating that the ability to capture miaRBCs, at least to the first degree, was not dependent on the life-cycle stages of the parasites. The experiments with different degrees of surface roughness showed that immobilization sensitivity can be enhanced by more than 30% by using substrates with nano-scale protrusions higher than the depth of the furrow-like structures. This further validated the

hypothesis that changes in biomechanical properties as a result of malaria infection could be used as a biomarker for malaria diagnosis. This approach could potentially be generalized and applied to differentiate other cell types or infected cells that exhibit surface morphological changes. Future work includes a more careful study to correlate the immobilizing efficiency and sensitivity to wall shear rate under a continuous and constant shear flow. Also, the influences of the degree of surface roughness to immobilizing efficiency and sensitivity will be investigated. Acknowledgments This work was partially funded by the Defense Advanced Research Projects Agency (DARPA) N/MEMS S&T Fundamentals program under grant no. HR001-06-1-0500 issued to the Micro/nano Fluidics Fundamentals Focus (MF3) Center and the Undergraduate Research Opportunity Program. The authors wish to thank Dr. Anthony A. James for supplying avian blood samples, Dr. Abraham Lee for providing the inverted microscope, and Dr. Jian-Guo Zheng for AFM measurements. Thanks also go to Chih-Jen Kuan for valuable discussions and suggestions. We also acknowledge the assistances from H. Wong, G. Eslamian, J. T. N. Nguyen, S. Ahrar, and K. Cho of the Microbiomechanics Laboratory.

References Y. Adams, C. Freeman, R. Schwartz-Albiez, V. Ferro, C.R. Parish, K.T. Andrews, Antimicrob Agents Ch. 50, 2850 (2006) M. Antia, T. Herricks, P.K. Rathod, PLoS Pathog. 3, 939 (2007) B.J. Bain, F.R.A.C.P., F.R.C.Path, New Engl. J. Med. 353, 498 (2005) L. Bannister, G. Mitchell, Trends Parasitol. 19, 209 (2003) B.M. Cooke, N. Mohandas, A.F. Cowman, R.L. Coppel, Vet. Parasitol. 132, 273 (2005) P.J. Freidlin, Avian Pathol. 14, 531 (1985) P. Gaehtgens, F. Schmidt, G. Will, Pflugers Arch. 390, 278 (1981) T. Hänscheid, J. Melo-Cristino, B.D. Pinto, Am. J. Trop. Med. Hyg. 64, 290 (2001) H.W. Hou, A.A.S. Bhagat, A.G.L. Chong, P. Mao, K.S.W. Tan, J. Han, C.T. Lim, Lab Chip 10, 2605 (2010) P.J. Hung, P.J. Lee, P. Sabounchi, N. Aghdam, R. Lin, L.P. Lee, Lab Chip 5, 44 (2005) P.J. Lee, P.J. Hung, V.M. Rao, L.P. Lee, Biotechnol. Bioeng. 94, 5 (2006) E. Nagao, T. Arie, D.W. Dorward, R.M. Fairhurst, J.A. Dvorak, J. Struct. Biol. 162, 460 (2007) G.B. Nash, B.M. Cooke, K. Marsh, A. Berendt, C. Newbold, J. Stuart, Blood 79, 798 (1992) J.P. Shelby, J. White, K. Ganesan, P.K. Rathod, D.T. Chiu, PNAS 100, 14618 (2003) S. Suresh, J. Spatz, J.P. Mills, A. Micoulet, M. Dao, C.T. Lim, M. Beil, T. Seufferlein, Acta Biomater. 5, 15 (2005) Y. Xia, G.M. Whitesides, Annu. Rev. Mater. Sci. 28, 153 (1998) L. Xiao, C. Yang, P.S. Patterson, V. Udhayakumar, A.A. Lal, Infect. Immun. 64, 1373 (1996)