Microfluidic pillar array sandwich immunofluorescence assay for ocular diagnostics
- First Online:
- Cite this article as:
- Green, J.V., Sun, D., Hafezi-Moghadam, A. et al. Biomed Microdevices (2011) 13: 573. doi:10.1007/s10544-011-9528-4
- 226 Views
Uveitis and primary intraocular lymphoma (PIOL) are diseases associated with the invasion of lymphocytes into various regions of the eye, accompanied by expression of inflammatory cytokines. While these diseases are very different in terms of survivability and treatment options they have similar symptoms that make accurate diagnosis challenging. Furthermore, the diagnostic yield with state-of-the-art techniques for cell and cytokine analysis of vitreous and aqueous humor samples is under 20% due to inadequate sensitivity. This paper describes a simple sandwich immunofluorescence assay (sIFA) microfluidic device that is capable of identifying important analytes in ocular biopsies as a potential alternative to current diagnostic approaches. Detection is accomplished by capture of the target molecules on antibody-coated, vertical, oval shaped pillars in a microfluidic device followed by a biotinylated detection antibody and finally florescent avidin for target molecule quantification. Cytokine concentration measurements were carried out on aqueous humor samples from rats with endotoxin-induced uveitis as well as human cataract patients. Results correlated well with conventional protein quantification techniques and additionally, measurements from the human samples surpassed detection limits of current state-of-the-art immunoassay techniques. The single-digit femtomolar range of detection of this sIFA system provides lower limits of detection when compared to traditional techniques and allows for the mapping of the cytokine content of vitreous biopsies with detection limits that have yet to be realized using cost effective microfluidics. Furthermore, the relative simplicity of the device design, fabrication and ability to automate makes it easily translatable from the laboratory to a clinical setting.
KeywordsMicrofluidicImmunoassayUveitisPrimary intraocular lymphoma
Uveitis is the general term for inflammation within the uvea, the region of the eye between the sclera and the retina. This inflammation can result from a variety of causes such as infectious agents, toxic compounds, or autoimmune disorders, which may not necessarily originate in the eye. In many instances, however, the exact cause of uveitis cannot be determined. The inflammation associated with uveitis manifests itself in the form of lymphocyte proliferation into one or more regions of the uvea, accompanied by the secretion of inflammatory cytokines. This inflammation, if not treated properly, leads to scarring within the eye and irreversible loss of vision. Approximately 2.3 million Americans suffer from uveitis, and vision-loss due to uveitis accounts for 10%–15% of blindness in the United States (National Institute of Health Report 2007).
Primary intraocular lymphoma (PIOL) is a condition in which malignant B-lymphocytes invade the vitreous humor and subsequently the retina and optic nerve (Chan and Wallace 2004). This invasion can result in the formation of clumps or sheets within the vitreous humor, causing blurred or impaired vision. More importantly, the presence of lymphoid B-cells in the vitreous humor may be an indication of tumor formation in the brain, retina, or other areas of the central nervous system (Chan and Wallace 2004). While the incidence of PIOL is relatively low (one in 100,000), it has tripled in the last 20 years due in part to the increasing number of patients with autoimmune diseases (Schabet 1999). PIOL is a subset of primary central nervous system lymphoma (PCNSL). The different forms of PCNSL collectively account for 4%–6% of brain tumors (Hochberg and Miller 1988).
Given the manifestation of PIOL in the form of lymphocyte invasion into the vitreous humor, early indications of PIOL are similar to those of uveitis and distinguishing between these two conditions is a major challenge faced by ophthalmologists. A major aspect of this challenge is the need to identify the subpopulations of lymphocytes that have infiltrated various parts of the eye, as well as detecting and measuring the relative amounts of certain cytokines. As with most forms of cancer, the treatment and management of PIOL relies on sufficient early detection followed by chemo- and radio-therapy to the brain and eye. Nevertheless, the prognosis of PIOL is poor and the need for effective early detection strategies is critical (Chan and Wallace 2004).
The vitreous biopsy remains the mainstay for diagnosis of ocular lymphoma. These biopsy samples are obtained by removing the gel-like vitreous humor with a microcutter-suction system while simultaneously infusing with a liquid substitute to flush the eye and to maintain the hydrostatic pressure of the eye. The resulting samples are typically large in volume (10 to 500 mL) and have a very small numbers of cells, making cellular characterization difficult. Prior work has shown that cytokines may play a role in distinguishing uveitis from PIOL but literature is not in agreement (Chan and Wallace 2004). It has been hypothesized and reported by several groups that interleukin-6 (IL-6) is produced at high levels by inflammatory cells in uveitis and IL-10 is produced at high levels by malignant B-lymphocytes in PIOL, resulting in PIOL being strongly associated with an increased IL-10 to IL-6 ratio (>1) (Cassoux et al. 2001; Chan et al. 1995; Whitcup et al. 1997; Wolf et al. 2003). However, there have also been reports of patients with PIOL who had a vitreous IL-10 to IL-6 ratio <1 (Akpek et al. 1999; Buggage et al. 1999). Based on these studies it can be concluded that no clear correlation between cytokine presence and disease type has been identified. Furthermore, identifying other types of cytokines that may be present has been difficult due to detection limits of current techniques. Banerjee et al. performed a study that utilized multiplexed bead analysis, the state-of-the-art small volume enzyme linked immunosorbant assay (ELISA) technique, to analyze the vitreous humor of patients with chronic uveitis (Banerjee et al. 2007). Elevated levels of IL-6 and no detection of interleukins 10, 12, 13, 15, 17, tumor necrosis factors (TNF), interferon-γ (IFN-γ), granulocyte-macrophage-colony-stimulating factor (GM-CSF) and granulocyte-stimulating factor (G-CSF) were reported (Banerjee et al. 2007). A protein detection platform with lower limits of detection may allow one to profile cytokine levels in ocular disease. These cytokine profiles can be of direct use when distinguishing between hard to diagnose diseases such as uveitis and PIOL.
Conventional sandwich well plate immunoassay techniques have detection limits in the single digit pg/mL range whether the detection method is colorimetric (ELISA) or fluorescent (fluorescent immunoassays). These well-plate immunoassays rely on diffusion of the antigen to the capture antibodies immobilized on the well plates and antibody-antigen kinetics; however, more importantly are limited to the volume of the well in terms of sample size. Bead-based immunoassay techniques are also traditionally performed on well plates and they too are limited in terms of sample size; however, the main difference between conventional well plate and bead-based immunoassays is that well plate immunoassays have one type of antibody immobilized on the base of a microwell while bead-based immunoassays use micro-beads with several types of antibodies immobilized on the surface suspended in a microwell. The main advantage of bead-based immunoassays is that many different antigens can be tested for simultaneously within a small sample. However detection limits for both platforms are in the single digit pg/mL range. Furthermore, sample volumes are limited to the size of the microwells utilized and these systems cannot accommodate sample recycling to enhance capture efficiency for low-concentration analytes.
Microfluidic immunoassaying is a heavily researched field; however to date most systems typically require complex fabrication, expensive materials, non-user friendly protocols or do not provide sufficiently low detection limits (Ng et al. 2010). For example, Cheow et al. achieved ~100 fold enhancement in detection using a microfluidic electrokinetic concentrator in conjunction with a conventional 96-well plate ELISA system (Cheow et al. 2010). However, this approach requires relatively complex fabrication steps (nanostructure assembly within microchannels). On the other hand, when more simple devices with user-friendly protocols are utilized, such as the chemically-modified polydimethyl(siloxane) (PDMS) microfluidic straight channel ELISA device described by Yu et al. or the point-of-care fluorescent immunoassay microchip that requires no external equipment by Hosokawa et al. detection limits are equivalent to or higher than that of conventional well plate immunoassays (Yu et al. 2009; Hosokawa et al. 2006).
By contrast, the device described in the present work is not only simple (in terms of design and fabrication) and user-friendly (in requiring only a set series of flow steps prior to optical imaging) but also able to surpass conventional immunoassay detection limits by virtue of its geometrical features and design, surface modification strategy and operational protocol. Also, the ability of microfluidic devices to handle a wide range of sample volume enabled the investigation of the effect of sample-volume size as both large-volume (model vitreous biopsy ~10 mL) and small-volume (human aqueous humor ~200 μL and rat aqueous humor samples ~10 μL) were examined.
The aim of this study was to demonstrate the efficacy of our microfluidic sandwich immunofluorescence assay (sIFA) system by analyzing the aqueous humor samples from Lewis rats with endotoxin-induced uveitis (EIU) and human cataract patients. The microfluidic device was used to test for IL-6, IL-10, TNFα and IFNγ in the aqueous humor of EIU Lewis rats as measurements of these cytokines have been previously reported (De Vos et al. 1994a, b; Hoekzema et al. 1991, 1992; Ohta et al. 2000). These measurements were also compared to measurements obtained from traditional ELISA kits, as ELISA is the current gold standard for cytokine detection in ocular diagnostics. Levels of IL-5, IL-10, TNFα and IFNγ were measured in the aqueous humor of pre-operative cataract patients. The sensitivity of the microfluidic sIFA device provides lower limits of detection when compared to traditional ELISA and allows for the mapping of the cytokine content of vitreous biopsies with detection limits that have yet to be realized using cost effective microfluidics. Furthermore, the simplicity of the sIFA device and the fact that it requires no sample pre-processing makes it viable for use in clinical settings.
2.1 Device design and fabrication
The device layout was designed and drawn using AutoCAD software and printed with high resolution on a chrome mask (FineLine Imaging). The resulting chrome mask was used to generate a negative master for device fabrication at the George J. Kostas Nanoscale Technology and Manufacturing Research Center at Northeastern University. Silicon wafers were coated with SU-8 photoresist to a thickness of approximately 50 μm and then exposed to ultraviolet light (365 nm, 11 mW/cm2) with the chrome mask overlaid using a Quintel 2001 mask aligner.
To generate PDMS replicas, a mixture of silicone elastomer and curing agent (10:1 ratio) was poured over the master wafers, degassed, and allowed to cure overnight in an oven at 65°C (Xia and Whitesides 1998). The cured PDMS was then cut using a scalpel and pulled off the wafers. Inlet and outlet holes were punched on each PDMS replica and the PDMS slabs were rendered hydrophilic through a simple two-step extraction/oxidation process that has been previously reported by Vickers et al. (2006). To create the microfluidic channel the hydrophilic PDMS replicas were bonded to glass slides by exposing the PDMS replicas and the glass slides to an oxygen plasma (100 mW with 8% oxygen for 30 s) in a PX-250 plasma chamber (March Instruments) and then immediately bringing the PDMS replicas and the glass slides in contact followed by baking the device for 5 min at 65°C. This process creates an irreversible bond between the PDMS replicas and the glass slides and also prepares both the PDMS pillars and the glass slide for chemical surface modification.
2.2 Surface treatment
Solutions of silane (Gelest), n-[γ-malimidobutyryloxy]succinimide ester (GMBS; Pierce), bovine serum albumin (BSA; Fisher Scientific), and capture antibody were prepared as follows. A 4% (v/v) solution of 3-mercaptopropyl trimethoxysilane (a moisture sensitive compound) in ethanol was prepared under a nitrogen atmosphere. The coupling molecule, GMBS, was stored as a stock solution containing 50 mg GMBS in 0.5 mL dimethyl sulfoxide (DMSO; Sigma); this stock solution was diluted with ethanol to create a 0.28% (v/v GMBS stock/ethanol) solution. The BSA solution was prepared in phosphate-buffered saline (PBS; 1X pH 7.4, Fisher) to a concentration of 5 mg/mL while the capture antibody solution was prepared in PBS to a concentration of 0.02 mg/mL. Surface functionalization of the microfluidic devices was carried out in four steps. In the first step, the devices were flushed with silane solution immediately after bonding and allowed to react at room temperature for 30 min. Unreacted silane was removed by flushing with ethanol. The GMBS solution in ethanol was then introduced into the devices and allowed to react for 15 min. The devices were flushed with ethanol to remove unreacted GMBS and then with PBS to remove the ethanol. The capture antibody solution was then flowed through the devices and allowed to react for 30 min before flushing with PBS. Next, the devices were injected with a 5% BSA solution, incubated for 30 min and finally flushed with PBS. The devices were then directly used in experiments or stored at 4°C. The same capture antibody attachment protocol was employed for all device types and all solutions were injected manually.
2.3 Evaluation of the extraction of PDMS devices
Five extracted devices (as described in section 2.1) and five non-extracted devices were bonded to glass slides with oxygen plasma and functionalized with 0.01 mg/mL fluorescein rabbit IgG (Vector Laboratories, Inc.) by following the surface modification protocol in section 2.2. The effectiveness of the extraction process was then determined by acquiring a fluorescence image of a single oval post at the center point of each device using a Nikon Eclipse TE2000 inverted microscope with a charged-coupled device camera (CCD) (example shown in Fig. 1(c)). Using image analysis software (Nikon NIS Elements 2.1) the average fluorescence intensity was determined for a rectangular region around the post. This measurement was made on the same z-plane for each device, specifically the top of each device (glass being on the bottom). Measurements at different-z planes indicated no fluorescence intensity variation along the height of the pillars.
2.4 Computational Fluid Dynamics (CFD)
The CFD analysis was performed by using Fluent (ANSYS, Inc.), a pressure correction based, multi-block, multi-grid, unstructured/adaptive solver. A three dimensional (3-D) model was created using Gambit, the geometry and mesh software component of Fluent. The channel design was created by first specifying points on a plane followed by connecting the appropriate points with lines to create the desired geometry. The layout was then extruded in the z-direction to form a 3-D structure. The mesh was designed to have increased density at points of interest and mesh independence was verified by examining higher density meshes. Navier-Stokes simulations were carried out using the CFD software and iterations were performed until the residual sums for the x-velocity, y-velocity, and z-velocity components reached values <1 × 10−5. The fluid inside the channel was simulated as water and a no-slip boundary condition was utilized with laminar flow assumed. The CFD result shows velocity profiles at a z-plane of 25 μm, which represented the middle of the channel fabricated with 50 μm total height.
2.5 Endotoxin-induced uveitis and aqueous humor collection
All experiments were performed in accordance with the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Care and Use Committee of the Massachusetts Eye and Ear Infirmary. Male Lewis rats (8–10 week old) were obtained from Charles River Laboratories. Uveitis was induced in rats by injecting 100 μg of lipopolysaccharide (LPS) extracted from salmonella typhimuirium (Sigma) diluted in 0.1 mL sterile saline into one hind footpad of each animal. Rats were used because this injection causes a systemic inflammation, one of its main manifestations being cytokine accumulation in the anterior chamber, resembling the acute phase of human anterior chamber uveitis (Rosenbaum et al. 1980; Hafezi-Moghadam et al. 2007). Animals were maintained in an air-conditioned room with a 12 h light-dark cycle and were given free access to water and food until used for the experiments. The aqueous humor samples were drawn 24 h following LPS injection by anterior chamber puncture using a 30 G needle while the animals were under anesthesia.
2.6 Cytokine measurement in aqueous humor of EIU rats
Aqueous humor samples from EIU rats were analyzed using the sIFA devices and purchased conventional well plate ELISA kits (Pierce). IL-6, IL-10, TNFα, and IFNγ were measured using the protocol developed in-house for the sIFA device and by following the instructions provided by the manufacturer of the conventional well plate ELISA kits. In the microfluidic devices, paired antibodies already tested for use in sandwich immunoassays (BD Biosciences & eBioscience) were selected as they were previously tested for cross-reactivity with other cytokines.
Briefly, 14 EIU Lewis rat aqueous humor samples (~6–10 μL each) were pooled and the resulting solution was diluted 25× with PBS and run through a 0.2 μm sterile syringe filter to remove cellular debris. 200 μL of the pooled EIU sample was then loaded into a syringe and connected to the microfluidic device inlet using tygon tubing. One hundred microlitre of the sample was then run through the device at a flow rate of 10 μL/min using a syringe pump (Harvard Apparatus PHD2000) over a 10 min period. The sample was then incubated in the device for a period of 15 min at room temperature. Following incubation, another 100 μL of sample was run through the device and incubated as previously described. Next, the sample was rinsed out of the device by injecting 500 μL of PBS manually. Two hundred microlitre of the corresponding biotinylated detection antibody at a concentration of 0.02 mg/mL was then injected into the device and incubated at room temperature for a period of 1 h. The device was then flushed with 500 μL of PBS and injected with 200 μL of fluorescent NeutrAvidin (488/524 nm; excitation/emission) biotin-binding protein (Invitrogen) and finally flushed with 500 μL of PBS after a 30 min incubation. The NeutrAvidin solution was stored as a stock solution containing 1 mg NeutrAvidin in 1 mL PBS; this stock solution was diluted with PBS to create a 0.004% (v/v NeutrAvidin stock/PBS) working solution. A fluorescence image at the center (containing ~8 pillars) of each device was acquired with the exposure time of the CCD camera set to 0.25 s while using a fluorescein (480 ± 30/535 ± 40 nm) excitation/emission filter. Using image analysis software the average fluorescent intensity was determined for a rectangular region around the eight pillars. This measurement was made on the same z-plane for each device, specifically the top of each device (glass being on the bottom). Measurements at different-z planes indicated no fluorescence intensity variation along the height of the pillars. Fluorescent intensity readings were then converted to cytokine concentrations using standard calibration curves created by analyzing solutions of known recombinant rat cytokine (BD Biosciences & eBioscience) concentrations in the same manner as the pooled EIU rat samples. PBS controls were run in parallel with each experiment and this background noise was subtracted from each reading.
The cytokine measurements obtained from the sIFA device and the purchased conventional well plate ELISA kits were corrected for dilution and then compared to investigate the integrity of the sIFA system presented. The pooled EIU rat samples were tested for IL-6, IL-10, TNFα, and IFNγ a total of three times per cytokine. Therefore, three separate readings were obtained for each cytokine tested.
2.7 Aqueous humor collection of patients undergoing cataract surgery
Aqueous humor samples (~50 to 200 μL) were collected from 12 patients prior to cataract surgery (mean age, 67 years; range, 58–77 years). Sample collection followed the tenets of the Declaration of Helsinki, was approved by the Institutional Review Board of the Schepens Eye Research Institute, and written informed consent was obtained from each patient.
2.8 Cytokine measurement in aqueous humor of cataract patients
Aqueous humor samples of cataract patients were analyzed using the sIFA device. IL-5, IL-10, TNFα, and IFNγ were measured using a protocol similar to that described above, along with antibody pairs that were tested for cross-reactivity with other cytokines (BD Biosciences & eBioscience). The only change was with respect to sample incubation within the device. For the human sample experiments, the incubation step was replaced with a continuous flow step in order to utilize the larger volume of human samples (50–200 μL) and to limit the time required to run the assay. The 50 to 200 μL human aqueous humor sample was diluted to a total volume of either 1 or 3 mL and run through a 0.2 μm nylon sterile syringe filter. One millilitre of the diluted sample was then loaded into a syringe and then run through the device at a flow rate of 10 μl/min. In this set of experiments, four capture devices, for IL-5, IL-10, TNFα, and IFNγ, respectively, were linked in series by tygon tubing. After the sample was run through the devices, a syringe with 0.4 mL of PBS was run through the devices at a flow rate of 10 μL/min to ensure the entire sample was run through each device. Following this step, the devices were exposed to their appropriate detection antibodies and then fluorescent NeutrAvidin according to the protocol previously described. A fluorescence image was acquired, as previously described, with the exposure time of the CCD camera set to 1.0 s. Fluorescent intensity measurements were made and converted to cytokine concentrations using standard curves created using solutions of known recombinant human cytokine concentrations (BD Biosciences & eBioscience). PBS controls were run in parallel with each experiment and the associated background was subtracted from each sample reading.
For cytokine content comparisons with previously reported data (Curnow et al. 2005; Zenkel et al. 2010), a separate set of experiments were carried out. Three human aqueous humor samples were diluted to a volume of 3 mL to provide three readings of each cytokine tested while nine others were diluted to a volume of 1 mL to provide only one reading per cytokine tested.
2.9 Cytokine measurement in vitreous biopsy model
2.10 Statistical analysis
Detection of aqueous humor cytokines of EIU Lewis rats by purchased conventional well plate ELISA kits and the sIFA device presented
Cytokine Concentration [pg/mL]a
1417 ± 29
7170 ± 119
1269 ± 448
6913 ± 275
Detection of aqueous humor cytokines of pre-operative human cataract patients. Each sample was diluted to 3 mL and tested in three separate devices
Cytokine Concentration [pg/mL]a,b
20.5 ± 0.1
3.0 ± 0.1
26.3 ± 0.1
323 ± 2.9
11.3 ± 0.1
360.9 ± 0.1
Detection of aqueous humor cytokines of pre-operative human cataract patients. Each sample was diluted to 1 mL and tested in only one device
Cytokine Concentration [pg/mL]a,b
3 Results and discussion
3.1 PDMS pillar modification and computational fluid dynamics
The sIFA device was designed and fabricated with the goal of achieving low detection limits while keeping device operation and analysis simple. With these objectives in mind, a single channel device with antibody-coated oval shaped pillars was created in an effort to maximize the amount of antibody-coated surface area exposed to a flowing sample. In order to maximize the amount of capture antibody on the PDMS pillars available for antigen binding, the PDMS was rendered hydrophilic by following a protocol developed by Vickers et al. (2006). The PDMS was soaked in a series of organic solvents and unreacted oligomers that cause PDMS to undergo hydrophobic recovery following oxygen plasma were extracted. This improves the oxidation of the PDMS surface during oxygen plasma treatment by ensuring that a maximal number of silanol (Si-OH) groups are generated. These silanol groups serve as attachment points for the mercaptosilane used in the first step of antibody immobilization. To verify that the extraction technique increased the amount of antibody on the PDMS pillars, fluorescent IgG was immobilized onto the pillars of extracted and non-extracted PDMS devices and results are shown in Fig. 2. This image clearly shows that extracted PDMS has more fluorescent IgG immobilized on the surface relative to non-extracted PDMS. Indeed comparison of the average fluorescent intensity of the images for these surfaces, 99.71 ± 6.76 (arbitrary units; n = 5) and 46.85 ± 7.79 (n = 5) respectively, shows an enhancement by a factor of two with extracted PDMS.
As previously mentioned, the pillars were designed to have an oval cross-section in order to increase the surface area to volume ratio. Furthermore, the oval pillars were staggered in an effort to promote antibody-antigen interactions on the face of the posts by creating zones of low flow velocity behind and in front of each post. CFD modeling of the sIFA device, shown in Fig. 3, confirmed the existence of these zones (dark blue regions). It is important to note that there were regions of contraction in the device (between the two rows of pillars); therefore zones of high flow rates did exist, as shown in Fig. 3. Even though zones of high flow rates were not ideal, they were necessary in order to achieve the zones of low flow and to achieve increased residence time for binding near the pillars in a flow-through system. As shown in Fig. 1(c), equivalent levels of fluorescent intensity were observed on both the front and back sides of these pillars, suggesting that the target antigen did achieve a high level of contact because of the increased surface area provided by the oval cross-section of the pillars and the zones of low flow. Furthermore, the use of a device with pillars as opposed to a simple straight channel device with no features was justified. In Fig. 1(c), it can clearly be seen that antigen capture on the pillars is much greater than that on the bottom (glass) or the top (PDMS) surfaces of the device.
3.2 Cytokine analysis of the aqueous humor of EIU rats
The rat IFNγ cytokine fluorescence intensity calibration curve for a 200 μL sample size is shown in Fig. 4(a) and the corresponding linear natural log plot of the same data is shown as Fig. 4(b). Rat IL-6, IL-10 and TNFα calibration curves along with their corresponding linear natural log plots can be found in the supplementary information as Figs. S1 and S2. The detection ranges for each cytokine were 26 to 65,000 pg/mL; 18 to 46,250 pg/mL; 17 to 42,500 pg/mL; and 20 to 50,000 pg/mL for IL-6, IL-10, TNFα, and IFNγ, respectively. Cytokine measurements of IL-6, IL-10, TNFα, and IFNγ in the aqueous humor of EIU Lewis rats using the sIFA device and conventional well plate ELISA kits are shown in Table 1. Furthermore, it should be noted that rat aqueous humors were pooled in order to provide an adequate amount of sample to be tested using both the microfluidic sIFA device and conventional well plate ELISA kits. This series of experiments was performed to verify the accuracy of the microfluidic measurements by comparison with a conventional technique.
In the pooled EIU rat samples IL-6 and TNFα were present in detectable levels and there was no detection of IL-10 and IFNγ. The elevated levels of IL-6 and TNFα measured 24 h after LPS injection agree with previous reports (De Vos et al. 1994a, b; Hoekzema et al. 1991, 1992). However, increased concentrations of IFNγ have also been reported by De Vos et al. and these findings contradict our data (De Vos 1994a). De Vos et al. used reverse-transcription polymerase chain reactions to analyze their samples and this method provides indirect measurements of cytokine content. Briefly, RNA is extracted from cells in a sample, reverse transcribed into DNA and then the DNA is amplified for measurement using traditional polymerase chain reactions. It is therefore possible that measurement strategies, indirect vs. direct, caused this discrepancy. Overall, the EIU rat cytokine measurements in the present work proved the integrity of the sIFA device as the measurements correlated well with conventional well plate ELISA techniques. The detection limits were hindered by the large dilution (25×) required to obtain enough sample to run multiple experiments and it was evident further optimization was required.
3.3 Cytokine analysis of the aqueous humor of human cataract patients
The aqueous humor of human cataract patients was then analyzed in an effort to challenge the sIFA device with human samples and to further validate the device. The removal of a patient’s aqueous humor is part of routine cataract surgery and the sample is known to contain inflammatory cytokines that can be found in vitreous biopsies of uveitis and PIOL patients. Furthermore, larger sample sizes obtained from human cataract patients allowed for increased flexibility in the way that the sample was run through the microchannel and consequently lower detection limits were achieved. For the EIU rat samples, 200 μL was loaded into a syringe and flowed through the device and incubated (flow for 10 min followed by incubation for 15 min; cycle repeated twice). Therefore the antibody-antigen interactions took place in both a dynamic and static environment. In the case of human samples, 1 mL was loaded into a syringe and flowed through the device continuously with no incubation. This flow-through regime allowed for the entire sample to be exposed to the immobilized capture antibody within the microchannel and in turn the achievement of lower detection limits. This was concluded as variables such as flow rate, sample characteristics and device surface modification protocol were held constant. Antibody-antigen pairs were changed, human as opposed to rat, however the increased limit of detection was prevalent for all human cytokine types when tested in a 1 mL sample. Cytokine concentrations were also changed in order to avoid fluorescent signal saturation and to probe lower detection limits. Furthermore, the microfluidic devices were linked in series so multiple cytokines could be tested for in one 1 mL sample. In this arrangement, the highly selective antibody-antigen interactions showed no effects of non-specific binding (data not shown) and the sample dilution was minimized.
Aqueous humor samples obtained from patients undergoing routine cataract surgery were tested for IL-5, IL-10, TNFα and IFNγ. These cytokines were selected as two previous studies reported no detection of IL-5; detection of IL-10 below 10 pg/mL; and contradicting results for the levels of TNFα and IFNγ detected (Curnow et al. 2005; Zenkel et al. 2010). The larger human sample sizes along with an optimized flow regime in the sIFA system allowed detection limits in the tens of femtograms per mL (fg/mL). The human cytokine calibration curve for IFNγ and the corresponding linear natural log plot are shown in Fig. 4(c) and (d) respectively. Human IL-5, IL-10 and TNFα calibration curves along with their corresponding linear natural log plots can be found in the supplementary information as Figs. S3 and S4. These samples were continuously run through the device at a flow rate of 10 μL/min; higher flow rates resulted in less effective detection. Lower flow rates were not investigated as the time required to run the full 1 mL sample through the device would have been unreasonably long (>100 min). The detection ranges for each cytokine were 148 to 1.48 × 106 fg/mL, 90 to 900,000, 80 to 800,000 fg/mL and 88 to 875,000 fg/mL for IL-5, IL-10, TNFα, and IFNγ respectively.
Initially three separate human samples were diluted ~30× to obtain enough sample (3 mL) to evaluate the precision of the sIFA device when testing human samples. The results, shown in Table 2, show good correlation with literature in terms of typical values of each cytokine for Patients 1 and 3 (Curnow et al. 2005; Zenkel et al. 2010). Patient 2, however, had elevated levels of IL-5 and IL-10. Analysis of medical history indicated that Patient 2 had an extensive history that included breast cancer, a hysterectomy and several prescribed medications at the time of aqueous humor removal. It is difficult to make internal comparisons of cytokine presence in the aqueous humors of cataract patients as the patients are typically older and have extensive medical histories that may include several medications or prior surgeries. Nevertheless, for all detectable cytokines, standard deviations were small in magnitude. Therefore, the sIFA device was able to provide results with high repeatability. The remaining human cataract aqueous humor samples were diluted to a volume of 1 mL and measured only once for each cytokine in an attempt to achieve lower detection limits by minimizing sample dilution. The results from these experiments are shown in Table 3 and for this set, measurements from all patients correlated well with literature.
As a result of decreased dilution, detection limits were decreased. Indeed, the sIFA system was able to provide measurements for cytokines whose concentrations were previously reported as undetected or below-limit in the literature (Curnow et al. 2005; Zenkel et al. 2010). Specifically, with the sIFA device, measurements of IL-10 and TNFα were made below reported detection limits (Curnow et al. 2005). However, for all patients IL-5 was not detected. It may be possible that IL-5 is present in lower concentrations; however the dilutions on the order of 10× caused our detection limits to be raised by a factor of 10 when compared to the calibration curves.
3.4 Cytokine analysis of a vitreous biopsy model sample
A decrease in the limit of detection (LOD) was obtained when a 1 mL sample was run continuously as opposed to flowing through the device and incubating a 200 μL sample. This phenomenon intuitively makes sense as more antigen is available for binding with larger samples. Since the cytokine mapping of vitreous biopsies, which are 10 to 500 mL in volume, from uveitis and PIOL patients is the primary goal of the sIFA device, 10 mL samples spiked with various amounts of human IL-6 were tested. This was an appropriate model as vitreous biopsies are predominantly aqueous. Results of an IL-6 calibration curve with a 10 mL sample and the corresponding linear natural log plot are shown in Fig. 5(a) and (b) respectively. The detection range for IL-6 with a 10 mL sample was 0.021 to 205 fg/mL and this far exceeds conventional protein immunoassay detection limits, which are in the single digit pg/mL range.
The sIFA device provides a simple serial injection platform where multiple cytokines can be tested for by linking devices in series and requires no sample pre-processing or collection of antibody-coated beads. The microfluidic approach presented herein is capable of not only meeting current detection limits (200 μL sample size) but surpassing these with large sample sizes (10 mL). By virtue of its unique design and ability to utilize large sample sizes, the sIFA device provides lower limits of detection when compared to traditional ELISA, which is the current “gold standard” for vitreous biopsy analysis. In the near future, the sIFA device will be implemented in the clinical setting in order to map the full cytokine profiles of uveitis and PIOL patients to discover possible correlations between cytokine presence and disease type.
The authors gratefully acknowledge financial support from the National Science Foundation through grant CBET-0827868. The authors also thank Mehdi Abedi for his CFD contributions.