Biosensors for immune cell analysis—A perspective

Detection of cell secreted molecules in the cell array

One of the more powerful microtechnologies for high-throughput single cell secretome analysis was developed by Love and co-workers who called their approach microengraving.^{29, 30, 41, 42} It employs microwells fabricated in PDMS to be large enough to house single cells. The PDMS mold is filled with cells and is pressed against a glass slide pre-coated with antibodies against secreted molecules. This way the immune cells become enclosed inside and secrete signaling molecules (antibodies or cytokines) into picoliter volumes. To determine secreting cells, the cover slip is removed and analyzed using a laser microarray scanner. Another variant of a platform for analyzing cell-secreted proteins was reported by Kishi and co-workers who developed microwells so as to place a single cell inside a well and then detect secreted proteins (antibodies) on its side walls.⁴³ This approach was used to identify production of influenza and hepatitis B specific antibodies.

The Heath lab had a different take on analyzing the immune cell secretome. This group micropatterned anti-cytokine antibodies on glass surfaces, enclosed these surfaces inside a microfluidic device, then captured immune cells within the microfluidic channels, and finally employed pneumatic actuation to segment channels into smaller, nl volume compartments.⁴⁴ Based on density of cell seeding into channels, one could ensure that a fraction of chambers contained single cells. When challenged with antigens, immune cells trapped in the microfluidic device released cytokines, which were captured in the same small volume next to secreting cells. This method allowed them to identify polyfunctional T cells capable of producing multiple cytokines.⁴⁴

Our lab has taken a slightly different angle of attack in developing leukocyte cytometry platforms. Our original interest was in being able to capture specific leukocyte subsets (i.e., CD4 T-helper cells or CD8 cytotoxic T cells) from complex samples such as blood containing multiple extraneous cell types. Therefore, we began by developing surfaces and washing protocols for capturing CD4 or CD8 T cells⁴⁵ and demonstrating that the proportions of cells captured on the surface reflected those in the blood.⁴⁶ Subsequently, we turned our attention to analyzing the function of the captured cells. The logic being that in addition to cell numbers and subset proportions, cell function could be used for diagnostic purposes. Thus, we proposed two types of immune cell capture and cytokine detection surfaces (shown in Figure ​Figure2):2): (1) arrays of antibody spots with cell capture spots printed next to cytokine detection spots^{47, 48} and (2) microwell arrays containing Abs for cell capture and cytokine detection.^{49, 50}

Open in a separate window

Figure 2

Co-localizing leukocytes and secreted cytokine signals. (a) Arrays of Ab spots for capturing leukocytes and detecting secreted cytokines. Reproduced by permission from Zhu et al., Lab on a Chip 8, 2197–2205 (2008). Copyright © 2008 by The Royal Society of Chemistry. (b) Arrays of hydrogel microwells functionalized with Abs to enable capture of single cells and detection of secreted cytokine molecules. Reprinted with permission from Zhu et al., Analytical Chemistry 81, 8150–8156 (2009). Copyright © 2009 by American Chemical Society. (c) Integration of microarrays into microfluidic devices for analysis small blood volume.

These surfaces were integrated into microfluidic devices (Figure ​(Figure2c)2c) and were used to capture CD4 or CD8 T cells from minimally processed blood and then detect secreted cytokines (IFN-γ, IL-2, and TNF) from the cells. Unlike approaches taken by other groups that require off-chip purification of T or B cells prior to analysis, we employ either whole blood or RBC-lysed blood and perform both cell capture and cytokine detection in the same device.^{47, 48} This may be advantageous for applications where only small amounts of blood are available, such as pediatric immunology or small animal research or in point of care applications where blood processing needs to be minimized. Figure ​Figure3a3a provides a typical result of immune cell analysis using antibody array (strategy 1 described above) where T-cells, stained for CD3, are juxtaposed with secreted signal for IFN-γ and TNF-α. Figures ​Figures3b3b and ​and3c3c show an example of microwell arrays (strategy 2 above) being used for capturing single T-cells and detecting secreted IFN-γ from the bound cells.

Open in a separate window

Figure 3

(a) Mixed array of cell and cytokine specific Abs. T cells are captured on anti-CD4 Ab spots. Reproduced by permission from Zhu et al., Lab on a Chip 8, 2197–2205 (2008). Copyright © 2008 by The Royal Society of Chemistry; (b) single Tcells captured in Ab-functionalized microwells. (c) IFN-γ signal co-localized with single cells. Green fluorescence is due to immunostaining for secreted IFN-γ whereas red fluorescence is for CD4 surface antigen. (b) and (c) Reprinted with permission from Zhu et al., Analytical Chemistry 81, 8150–8156 (2009). Copyright © 2009 American Chemical Society.

Novel strategies for sensor regeneration and signal amplification

Ideally, biosensors should be responsive and regenerable on the time scales relevant for cellular communication and should be able to select an analyte of interest in a complex biological fluid out of a plethora of interfering substances. Sensor regeneration is particularly important. In Sec. 3A of this perspective, we highlighted the need to develop biosensors for dynamic and continuous monitoring of signaling molecules secreted by the immune cells. One question that has yet to be effectively addressed is how to design affinity biosensors based either on antibodies or aptamers that would be sensitive down to nM or pM analyte concentration while simultaneously allowing for detection of fluctuations in the level of cell-secreted molecules. Sensitivity implies high affinity for the analyte of interest and slow dissociation kinetics. This means that while increase in concentration of cell-secreted molecules is detectable, downward trends in concentrations are difficult to monitor. We see one solution, whereby affinity detection will be combined with regeneration of the sensor in a sequence of detect-regenerate-detect steps.

What would these regenerable sensors look like? While optical, particularly SERS, sensing has matured significantly, its μM detection limit⁶⁷ does not allow it to detect signal proteins from single cells or small colonies. Its complex peripheral instrumentation also discourages integration with small wells like those in Figure ​Figure1.1. Finally, the nanocrystals may interact with the cells and trigger unwanted responses. Certain fluorescent detection methods have shown unprecedented resolution (single receptor binding event) and sensitivity (single cytokine)⁶⁸ but are not suitable for a practical assay platform. In contrast, electrochemical sensing represents a more promising platform from the stand point of low cost, simplicity, and practicality. Some of the widely used biosensors are based on field effect transistors and electrochemical impedance spectroscopy (EIS). While promising, these approaches suffer from multiple drawbacks including signal drifts, irreproducible signals, ionic strength sensitivity, and interaction of probes with counterions, all of which are common problems for these sensors.^{69, 70, 71}

Cheng laboratory has developed a microfluidic regeneration strategy for any sensing platform involving probes.⁷² Bipolar membranes with cation and anion-selective features are synthesized on silica chips using UV photosynthesis methods. The bipolar membrane behaves like a semi-conductor diode, with the electrons and holes replaced by anions and cations. Like a diode, when a reverse DC bias is applied across the bipolar membrane, mobile charges are depleted at the junction. Since both membranes are highly ion-selective, whose mobile ions are all counterions at equilibrium, the application of a DC field can deplete all mobile ions at the junction. This ion depletion effect reduces the electrolyte conductivity at the junction and, hence, amplifies the local electric field by orders of magnitude. In a recent paper,⁷² we showed that the electric field can exceed 1 MV/cm at the junction such that water molecules at the junction can be dissociated into proton and hydroxyl ions by the so-called Wien effect.⁷³ The resulting proton and hydroxyl ions migrate through their respective ion-selective membranes and are therefore, separated to two sides of the bipolar membrane. By using two such bipolar membranes, we are able to produce two controllable proton and hydroxyl ion actuators on the chip and, with a down-stream static mixer, the generated ions are mixed into a solution of a specific pH between 2 and 11 with high precision as shown in Figure ​Figure7.7. This chip generated a high pH spike that can be used to dehybridize the captured molecules and regenerate the probe.

Open in a separate window

Figure 7

An on-chip bipolar membrane pH actuator. The image in (a) shows how a DC field across a bipolar membrane can dissociate water and generate protons and hydroxyl ions at a controllable rate. Two such bipolar membranes are fabricated up stream of a flow channel and the protons/hydroxyl ions are mixed by a static mixer into a buffer of a specific pH between 2 and 10 in (b). The mixed streams are shown in (c) with a color chart for the universal pH dye.

The quantification of molecules secreted is often necessary for immune cell analysis. However, signals from electrochemical sensors based on Faradaic reactions (or field-effect transistor sensors based on charge sensing⁷⁴) are often sensitive to local pH and ionic strengths. The above pH actuation device can be used to control the pH near the sensor for a large culture. To control the local ionic strength, we employ other features of ion-selective membranes—ion concentration and depletion.^{75, 76, 77, 78} As shown in Figure ​Figure8,8, we are able to lower the local ionic strength to deionized water level and to increase it from DI water level to mM. The same technique can be used to enhance the sensitivity of the molecular sensor by increasing the local concentration of the target molecules.

Open in a separate window

Figure 8

(a) Optical microscopic image of a charge-selective membrane based preconcentrator. (b) Concentration of fluorescently labeled molecules taking place 10 s after applying a voltage bias of 10 V. The scale bars represent 50 μm.

Another intriguing strategy to enhance the sensitivity of molecular a sensor is to mimic excitable potential dynamics of a neuron cell—its response is much larger than the input when the input amplitude exceeds a threshold. Their IV curves do not possess linear Ohmic characteristics and in fact, often demonstrate negative differential resistance. Just as excitability amplifies input action potential signals, nonlinear oscillatory cardiac neuron dynamics acts like a clock and bistable dynamics in brain neurons allows for a binary memory device. Intra-cell sensors have confirmed the highly nonlinear nature of action potential dynamics during neuron signaling and cardiac oscillations.⁷⁴ Such “neuron” sensors can be utilized to detect antibody-antigen docking in our immune cell cultures, as a more direct sensor of antigen docking onto a single cell. In fact, one can imagine an artificial membrane unit that can communicate with a real immune cell by converting cytokine signals to electronic signals and vice-versa.

Analogously, a promising strategy for external signal protein detection is to replace the linear electrochemical sensing techniques with a nonlinear one based on ion currents instead of Faradaic currents from electron-transfer reactions. The usual paradigm of linear EIS spectroscopy must be modified in the process. For example, the phase-lag between input and output is locked at a frequency-independent value for nonlinear sensors. In Chang’s lab, several on-chip nanoporous membrane and nanoslot real-time nucleic acid and signal protein sensors have been created with this design.^{75, 76, 77, 78} With molecular probes functionalized onto the membranes to mimic receptors, the membrane is designed such that its charge is opposite to the target molecules (see Figure ​Figure9).9). As such, target molecule binding produces a PN membrane for ions, akin to a PN semi-conductor diode for holes and electrons. This strategy produces a very specific sensor that generates an amplified ion conductance signal that does not rely on electron transfer. The ion current is picked up by electrodes in isolated chambers uncontaminated by the culture. Insofar as the nonlinear I-V features of a semi-conductor diode and cell membrane ion channels can be reproduced by the ion-selective membrane components, other features such as oscillatory and bistability should also be possible. If molecule capturing can induce oscillations of a specific frequency at one sensor, electrical signals from a large array of sensors can be monitored by a single electrical detector thus, simplifying the sensor network design. This is yet another promising future direction.

Open in a separate window

Figure 9

Nonlinear PNP and PN membrane sensors for cytokine and mRNA. Molecular docking produces a surface membrane with an opposite charge. This nonlinear ion dynamics due to molecule docking produces hysteretic IV with nonlinear amplification of the ion current signal for the docking events. (Note that the signal is on the order of several volts instead of the usual mV in electrochemical sensing.) These membrane components can be integrated into the culture scaffold in Fig. ​Fig.11.

Cellular communication via secreted signals

It should be noted that in addition to T and B cells there are other important cellular participants of the immune system including granulocytes, monocytes, dendritic cells, etc. These cells release cytokines, reactive oxygen species, or other signaling molecules in response to pathologies. However, because these cells are not designed to mount pathogen-specific immune responses, they are more challenging to use in disease diagnosis. This said, inflammation—general state of immune cell activation—has emerged as an important factor in cancer, tissue injury, and regeneration. Therefore, sensitive, temporal analysis of inflammatory markers will likely become more and more significant as understanding of biological mechanisms evolves.

In fact, we see biosensors and microfluidic platforms as being a key to improving mechanistic understanding of immuno-inflammatory responses. Given the key role of paracrine signals and the involvement of multiple cell types in generating these signals, there is a clear need to develop platforms for monitoring heterotypic cellular signaling.⁸⁷ While the development of first microfabrication approaches for heterotypic cell cultures dates back to over a decade ago,^{88, 89} the use of reconfigurable microfluidic chambers now allows to sequester individual cell types and to actively control how and when communication happens.^{90, 91} Integration of such cell culture systems with biosensors for local and continuous monitoring of secreted signals will help to elucidate which cell type the signals originate from and how reciprocal communication propagates.