Monitoring erythrocytes in a microchip channel that narrows uniformly: Towards an improved microfluidic-based mimic of the microcirculation

ATP, both in standard form and that released by RBCs, was measured using the luciferin/luciferase reaction [13]. The light intensity of the chemiluminescence produced in this reaction is proportional to the amount of ATP present. Firefly tail extract was used as the source of luciferase (FLE-50, Sigma, St. Louis, MO). Two milligrams of synthetic d-luciferin (L6882, Sigma) were added to each vial of firefly tail extract to enhance the sensitivity of the reaction. The luciferin/luciferase mixture was then prepared by dissolving these reagents in 5 ml of physiological salt solution (PSS). This buffer contained the following (in mM): 4.7 KCl, 2.0 CaCl₂, 1.2 MgSO₄, 140.5 NaCl, 21.0 tris(hydroxymethyl)aminomethane, and 11.1 dextrose with 5% bovine serum albumin. The buffer was adjusted to a pH of 7.4 and filtered three to five times. The luciferin/luciferase mixture was stirred with a magnetic stir bar for at least 1 h, and centrifuged at room temperature for 15 min at 2500 rpm to remove large particulates that can clog microfluidic channels. Smaller particulates were removed by filtering the mixture with a SFCA filter (average 0.80 μm pore size). The chemiluminescent reagents were always prepared on the day of use. ATP standards were also diluted in PSS.

RBCs were isolated and prepared on the day of use. For obtaining rabbit RBCs, male and female New Zealand White Rabbits (2.0–2.5 kg) were anesthetized with ketamine (8.0 mg kg⁻¹) and xylazine (1.0 mg kg⁻¹) followed by pentobarbital sodium (15 mg kg⁻¹ i.v.). After tracheotomy, the rabbits were mechanically ventilated (tidal volume 20 ml kg⁻¹, rate 20 breaths/min; Harvard ventilator, Holliston, MA). A catheter was placed into a carotid artery, heparin (500 units, i.v.) was administered, and after 10 min, the animals were exsanguinated. Blood was collected into vials, and the RBCs were separated from other formed elements and plasma by centrifugation at 500 × g at 4 °C for 10 min. The supernatant and buffy coat were removed by aspiration. Packed RBCs were resuspended and washed three times in PSS. The RBCs were then diluted with the buffer to the appropriate hematocrit. For all studies reported here, a hematocrit of 7% was employed since the hematocrit of RBCs in resistance vessels is generally less than 20% [19].

PDMS channel structures were produced based on previously published methods [15], [20]. To produce structures 38 μm in depth, a 4 in. silicon wafer (Silicon Inc., Boise, ID) was coated with SU-8 10 negative photoresist (MicroChem Corp., Newton, MA) using a spin coater (Brewer Science, Rolla, MO) operating with a spin program of 1000 rpm for 20 s. The photoresist was prebaked at 90 °C for 5 min prior to UV exposure with a near-UV flood source (Autoflood 1000, Optical Associates, Milpitas, CA) through a negative film (2400 dpi, Jostens, Topeka, KS), which contained the desired channel pattern. All channel designs were drawn in Freehand (PC version 10.0, Macromedia Inc., San Francisco, CA). Following this exposure, the wafer was postbaked at 90 °C for 5 min and developed in Nano SU-8 developer (MicroChem Corp.). The thickness of the photoresist was measured with a profilometer (Alpha Step-200, Tencor Instruments, Mountain View, CA) and this height corresponds to the channel depth (38 μm) in the resulting PDMS channels. To make masters with 93 μm raised structures, essentially the same process was used, except that SU-8 50 negative photoresist was used with a spin program of 1500 rpm for 20 s.

Glass plates that had microbore tubing connectors attached were made in-house. Soda lime glass plates (7.0 cm wide, 10.25 cm long, 2 mm thick) were purchased from a local glass shop. Fluid access holes were made in the glass plate using a 1.5 mm diamond drill bit and a Dremel rotary tool (Dremel, Racine, WI). The syringe connector portion of a Luer adaptor was removed with the Dremel rotary tool and accompanying cutting disk and epoxied to the opposite side of the glass with J.B. Weld (Sulfur Springs, TX). The epoxy was cured in a 75 °C oven for 2 h. PDMS microchips with the channel design in Fig. 1A were reversibly sealed to the glass base by lining up the ends of the channels with the introduction reservoirs which were drilled in the glass plate. However, due to higher pressures associated with high flow rates, the PDMS chips containing the channel that scaled down uniformly were irreversibly sealed to the glass base. This was accomplished using a plasma cleaner (PDC-32G, Harrick Plasma, Ithica, NY). The PDMS chip and the glass plate were placed in the plasma for 150 s. Afterward, the two components were quickly aligned and sealed together.

The setup for measuring ATP from both standards and RBC samples is composed of two syringe pumps, the PDMS chip, glass substrate and photomultiplier tube, and is the same as previously reported [14]. In each part of this study, microbore tubing with an inside diameter of 150 μm was used to deliver all reagents from the syringe barrels to the microchip channels. As with the 250 μm i.d. tubing used for deformation in previous studies, no ATP release from RBCs has been witnessed at practical flow rates from tubing this large in our laboratory [14]. In addition, before each determination of RBC-derived ATP, a calibration was performed using ATP standards and the exact conditions as those employed during the analysis of the real sample.

In the first part of the study (see Section 3.1), RBCs and ATP standards were pumped at a constant linear rate of 0.732 cm/s. At the mixing tee, the RBC samples are combined with the luciferin/luciferase mixture. The resultant flow rate propels the RBCs and the chemiluminescent product past the PMT within seconds. Flow rates of 2.50 μL/min are used to deliver the reagents to the channel that narrows (see Section 3.2). In order to keep the range of channel widths constant for each section, a piece of negative film such as that used to create the silicon master is affixed to the PDMS wafer over the channel. The negative film contains a transparent 4 mm × 4 mm box. This limits the amount of light recorded by the PMT to that which comes from this specific region. The measurement of channel widths in each region is detailed in the imaging section below. A steady-state signal is achieved because in both studies, the resultant chemiluminescent intensity is recorded for 60 s using a data acquisition program written in-house (LabWindows/CVI, National Instruments, Austin, TX). Amounts of ATP released from RBC samples were determined by comparison to values for standards prepared in the PSS on the day of the study. Measurements of the standards were performed in triplicate, and the results from these standards were used to generate a working curve.

The design used in our previous microchip study allowed us to test whether the microfluidic platform would be a viable tool for the determination of deformation-induced release of ATP from RBCs [14]. In that study, it was verified that a skimming layer exists between the RBCs and the side of the channel during flow, as is seen in vivo. Also, it was reported that ATP release from RBCs in microchips and microbore tubing with comparable dimensions and linear flow rates were statistically the same. In addition, RBC hematocrit was determined to have a direct relationship to ATP release. These results complement those reported in studies employing microbore tubing as a biomimic of resistance vessels [12], [13] and in vivo [21].

In the previous study, two different channel cross sections (60 μm × 38 μm and 100 μm × 54 μm; width × depth) were used to compare the effect of channel dimensions. It was assumed that the results obtained were due to the difference in the cross-sectional areas between the channels. However, this assertion could not justifiably be made due to the fact that two variables, channel depth and width, were changing. Here, a follow-up study was designed to ensure that channel cross-section was indeed the factor that determined ATP release instead of the possibility that one channel dimension alone could be a limiting factor in the process. The channels used in this study had depths of 38 and 93 μm and a constant width of 60 μm.

ATP standards were pumped through the deformation channel at a linear rate of 0.732 cm/s. At the mixing tee, the sample stream was combined with a luciferin/luciferase mixture and the resultant chemiluminescence was measured by a PMT located directly below the reaction channel. In this design, the RBCs are mechanically deformed before being mixed with the chemiluminescent reagents. Calibration curves were then constructed for each channel, using ATP standards with concentrations of 0, 2, 5 and 8 μM. A 7% hematocrit of RBCs suspended in PSS was measured in the same manner as the standards. The amount of ATP released was determined by comparing the raw luminescent intensity to those of the standards. Fig. 1B shows the ATP release from RBCs flowing through microchip channels with the same width, but different depths. The measured amount of RBC-derived ATP was 0.323 ± 0.010 μM (r² for standard curve = 0.9899) for the channel with a cross-section of 60 μm width and 93 μm depth. Subsequently, an increase in ATP release (0.641 ± 0.014 μM, r² for standard curve = 0.9915) was measured when the depth of the channel was decreased to 38 μm. These results are consistent with previous findings that RBCs release increased amounts of ATP as a function of decreased tubing diameter [12], [13]. It is also consistent with previous work by Sprague et al. [5], [6], [7] who reported that RBCs released increments of ATP when pulled through pores of decreasing diameter. Importantly, the eventual ATP measurements in Sprague's work were eventually made off line in a standard luminometer. This suggests that the increase in ATP seen in that work (and work reported here) is due to an increased amount of ATP derived from the RBCs, as opposed to an increase in ATP concentration due to the smaller volumes in the decreasing cross-sectional area of the channels. Moreover, they are also important because they prove that the quantity of ATP present due to RBC deformation is not limited by the smallest dimension, yet is dependent upon the cross-sectional area. That is, if erythrocyte deformation was a function of the smallest dimension present, the ATP release in the two channels would be the same. From this point on, comparisons of deformation-stimulated ATP will be discussed in terms of a difference in channel cross-sectional area.

When examining ATP release from RBCs in a channel that scales down (Fig. 2), two variables that play a role in the phenomenon, cross-sectional area and linear flow rate, are changing simultaneously and cannot be isolated. Although challenging, determination of RBC-derived ATP in this type of channel is important as resistance vessels in vivo display similar characteristics. Creating a more accurate biomimic using a tapered channel introduces an advantage in terms of the fabrication and detection schemes. That is, in order to measure the amount of ATP released from RBCs flowing through three different cross-sectional areas, three separate channels would have to be fabricated, requiring an increase in both time of analysis and raw materials. The design used in this study, however, allows for a myriad of biologically relevant cross-section ranges to be studied on the same chip, in the same flow stream and in a high-throughput fashion.

One important feature of the design reported here is that the mixing of the RBCs with the chemiluminescence reagents occurs before the streams enter the deformation channel. Therefore, the flow profile of RBCs must be monitored after the two introduction channels combine, as the physiological event of interest (deformation-induced release of ATP from the RBCs) occurs after the initiation of mixing with the luciferin/luciferase mixture. Also, any ATP released will have to mix with the luficerin/luciferase solution to enable the chemiluminescent reaction to occur. Originally, it was thought that even with low Reynold's numbers (the highest Reynold's number encountered in this work is still less than 3), some convection might exist in the system due to the fact that RBCs are particles. However, as micrographs A and C in Fig. 3 show, laminar flow dominates the flow profile, even as the channel scales down to the smallest cross-sections. The presence of laminar flow necessitates a different approach to introducing reagents in this chip design. That is, as ATP is released and measured, it would be unknown if ATP is derived from deformation due to a certain cross-sectional area (as is defined by the channel dimensions), or if deformation occurred from the focusing of RBCs due to the laminar flow profile. In reality, the cross-sectional area responsible for ATP release would be about half that of the channel cross-sectional area, with this decreased area coming from focusing of the RBC flow stream by the luciferin/luciferase flow stream (Fig. 3A and C). This problem was solved by preparing a 7% hematocrit of RBCs in the luciferin/luciferase mixture prior to RBC introduction to the microfluidic channels. This ensures that RBCs are mixed with the luciferin/luciferase mixture and when the two introduction channels combine, the width of the RBC stream is approximately the same as the channel width. This is seen in micrographs B and D of Fig. 3.

When mixing the RBCs with the luciferin/luciferase solution before introduction to the microfluidic channels, it is important to ensure that diffusion-based mixing does not differ between the ATP standards and RBC samples, as the presence of RBCs increases the viscosity of the solution, and viscosity effects both diffusion coefficients and the Reynolds number. An imaging study was performed in which a solution containing 2 μM fluorescein-labelled dextran was substituted for the chemiluminescent reagents. The size of the dextran molecule (70 kDa) was chosen in order to more closely replicate that of the enzyme luciferase (∼60 kDa). This is not to say though, that the extent of mixing is limited to the movement of the luciferase molecule. ATP possesses a much larger diffusion coefficient and will mix more rapidly than luciferase. As the micrographs in Fig. 4 show, the extent of mixing that occurs in standards (A and C) is very similar to that in 7% RBC samples (B and D), confirming that the calibration performed should ensure accurate results.