# On-chip background noise reduction for cell-based assays in droplets

Written by Kevin Tian, AP 225, Fall 2011

--Ktian 17:05, 9 November 2011 (UTC)

Title: On-chip background noise reduction for cell-based assays in droplets

Authors: Pascaline Mary, Angela Chen, Irwin Chen, Adam R. Abate and David A. Weitz

Journal: Lab on a Chip (2011), Vol 11, Pages 2066-2070

## Paper Summary

One area that has frequently become a problem in droplet-based microfluidics is that the technology is limited to homogenous assays. The primary reason for this limitation is that it is difficult to wash out reagents from the reaction vessels. Multi-step processes or simultaneous reaction and detection steps are made extremely difficult due to this inability to effectively remove excess reagents. Previous washing methods utilized magnetic beads to isolate the reagents of interest. The method is applicable to cells by coating the surfaces of these cells with magnetic particles, and miniaturization of magnetic actuators is possible due to high field gradients. However the major drawback is the appropriate control systems are required to manipulate these magnetic fields.

This paper proposes a high-throughput method capable of 14 times dilution. The framework involves using dielectrophoresis (DEP) to inject specific quantities of reagents into droplets. The system splits the larger droplets into eight smaller droplets via three consecutive T-junctions. This process can be repeated for greater dilution factor. One great advantage of the technique is that it avoids the dangers of off-chip handling procedures, which can result in reagent loss, longer response times and cross-contamination of samples.

In order to demonstrate the effectiveness of the method, the authors apply it to detect an enzyme mediated, site-specific, protein labeling reaction on the surface of yeast. Repetition of the dilution process for a total of two times allows for a reduction of background noise of up to factor of 100 within the droplet.

## Materials and Methods

Microfluidic Device Fabrication

Fabrication is done entirely with Polydimethylsiloxane (A.K.A. PDMS) using standard soft-lithography techniques. The droplet-dilution module requires electron for triggering the dilution process. The electrodes are made by fabricating microchannels of desired shape, gently heating the PDMS device and injecting a low mpt solder into the channels (when cooled we get solid solder electrodes). Electrical connections are made using eight-pin terminal blocks (from Digikey) that are glued to the device surface for strain relief. The glue used was Loctite UV cured.

On the electrical side, the voltages applied come from a a function generator producing 20kHz pules that are amplified 1000x by a Trek high-voltage amplifier.

Microfluidic Device Operation

Figure 1.

Three separate devices were fabrication for this paper. These are all shown in Figure 1 (a,d and g, which are the drop maker, dilution/splitting and detection devices respectively). All fluid flows are controlled by syringe pumps (PHD 22/2000, Harvard Apparatus). The essential process is as follows:

• 1) Droplet Formation Module (Figure 1a)
• The module uses flow-focsing junctions with a 25x25$\mu m$ nozzle (Figure 1b). Droplets are made in a fluorinated oil (HFE 7500, 3M, St Paul, Minnesota) containing 1.8% (wt/wt) of EA surfactant (Raindance Technologies).
• The resulting emulsion is collected in a 500$\mu L$ glass syringe (Hamilton gastight) which is used as input to the 2nd module.
• Figure 1c shows $40 \mu m$ droplets flowing in the channels.
• 2) Dilution/Splitting Module (Figure 1d)
• Droplets are spaced out by an oil flow and enters a flow-focusing channel (50$\mu m$ x 40$\mu m$, height x width) that causes droplets to flow single-file.
• [Dilution] Near the electrodes, a T-junction connects the droplet channel to a secondary injection channel (which flows larger droplets with no 'reagent'). Via DEP the two droplets are fused together, thus 'diluting' the original droplet(see Figure 1e)
• [Splitting] After dilution three successive T-junctions symmetrically break drops into smaller drops (See Figure 1f)
• 3) Detection Module (Figure 1g and 1h)
• The smaller drops resulting from the second module are re-injected into this module
• The drops are spaced out by an additional oil flow and the intensity of their fluorescence is measured.

Droplet Detection

The detection module does not in of itself detect the fluorescence signal, as the module is placed on an inverted microscope and the fluorescence is detected with a photomultiplicator (PMT) attached to the epifluorescent port. A 20mW cyan Laser (Picarro) is used for excitation, which is aligned with the microscope's optical axis and focused onto the sample by a 40x objective lens. Droplets flow through a 15$\mu m$ wide x 20$\mu m$high channel, where the fluorescence is detected.

Data is colected via a National Instruments DAQ card, controlled using Labview. Statis images captured by an EM-CCD Camera (Qimaging Rolera MGI).

Reagents in droplets

This section describes the experiment used to demonstrate the functionality of the above apparatus.

Primary droplet solution contains fluorescein at 1mM is buffered with 1 $\times$ tris buffered saline (TBS). Drops are diluted with pure 1$\times$ TBS. The droplet enzymatic assay is performed by using a droplet maker with two inlet channels, one for a yeast suspension and the other with substrate and enzymes.

Yeast cells:

• Grown in Yeast-extract Peptone Dextrose (YPD) medium.
• Cell density is measured, followed by twice centrifuging and resuspension at appropriate concentrations in TBS containing 1$mg~mL^{-1}$ of bovine serum albumin(BSA).
• In order to prevent sedimentation and to match the surrounding medium's density with yeast density, 35% (v/v) Optiprep (Axis-Shield) is added to the suspension.

Reagent solution is composed of:

• 40mM 4'-phosphopantetheinyl transferase (SFP synthase)
• 10mM CoA-488 (New Englands Biolabs)
• 10mM $MgCl_2$
• 780 $\mu g ~ mL^{-1}$ BSA

## Results

Dilution Evaluation

Figure 2.

The droplet maker produces droplets containing fluoroscein at high-volume fraction, which are then injected into the dilution/splitting module via tubing connecting the two devices. Flow rates are adjusted accordingly to achieve synchronization between reagent drop reinjection and dilution buffer injection. Some other minor details regarding the adjustments necessary to maintain continuous flow and generation of droplets is discussed.

Image analysis is performed to determine drop radii, which can be used to compute dilution ratio (simply a ratio of volumes). If we define quantities, $Q_{drop},~~Q_{inject}$, representing the flow rate of regent drops and injected buffer respectively, then we know can make the claim that the dilution ratio is equal to the flow rate over droplet rate which is:

$flowrate~ratio = {(Q_{inject}+Q_{drop}) \over Q_{drop}}$

This is assuming a continuous droplet formation rate (which can be adjusted for optimal rates). This is verified in Figure 2a, where ratios of volume before/after injection and dilution ratio are plotted. As one can see, it is essentially an x=y plot, to within error.

Thus to control the dilution ratio one needs only to adjust the flow rate ratio. Thus the concentration of fluorescein in the droplets before and after dilution ($C_{initial}, C_{final}$ respectively)) is given by:

${C_{final} \over C_{final}} = {Q_{drop} \over {Q_{drop}+Q_{inject}} }$

Each T-junction reduces the droplet volume by 1/2 each time, leading to a final reduction of 1/8 droplet volume size at the end of 3 consecutive T-junctions. However at high reagent injection rates secondary breakup events occur at the T-junction, limiting the drop injection flow rate to $100~\mu L h^{-1}$

Fluorescence measurement results are depicted in Figure 2b, where distributions are given of the normalized intensities emitted by drops. This was given for an initial drop fluorescein concentration of 1$\mu M$, and drops after 6, 8, and 10 times dilution. As one can see there is a definite decrease in the standard deviation after successive dilutions. It is claimed that standard deviations of diluted solutions is ~5 times larger than that of drops containing the same fluorescein concentration formed without dilution [Paraphrased from article. I believe this is a mistyped statement and it should be the other way round, since lower intensity peaks indicate higher dilution, yet very sharp normalized distributions. Plus the method would be pointless if the statement was true.].

Enzymatic Assays: SFP Labeling Reaction

Figure 3.

The purpose of using fluoroscein above was due to the prevalence of fluorometric assays to quantify enzymatic reactions. Fluorescence Resonance Energy Transfer (FRET) avoids the washing steps before measurement, however has only a limited dynamic range of detection (relative to simple fluorescence measurements). Additionally FRET requires pre-labeling cell surfaces with fluorophores (not always possible). Thus the authors present a fluorescent labeling reaction on the cell surface.

The expression of the S6 peptide sequence on the cell surface is targeted in this reaction. In this reaction, the Alexa Fluor488-substituted phophopantetheine group of CoA-488 substrate is covalently transferred to the serine side chain within the S6 sequence by the enzyme SFP synthase (Figure 3).

Without washing, one immediately notes the problem; there is a bulk fluorescence and surface fluorescence that must be distinguished from one another. This is only possible by adjusting the fluorescent substrate to be smaller than the number of ligated molecules. However this restricts one to non-ideal concentrations for the reaction. Dilution instead allows one to have high fluorescent substrate initially, only to remove significant fractions of those in the bulk to leave mostly those that had bound to the cell surface in the desired reaction.

Figure 4.

The enzymatic droplet assay was performed by coflowing a suspension containing yeast cells engineered to display the S6 peptide with a second stream of SFP synthase and the CoA-488 substrate. The enzymatic reaction begins after droplet formation (when the streams are mixed). The drops are incubated over night before dilution in the dilution/splitting module. The drops are twice diluted which should yield 100 times reduction of unreaction fluorophore concentration.

An illustration of how effective this technique is is shown in Figure 4a-f.

• Figures 4a-c depict bright field images of the cell before dilution, 10 times and 100 times dilution.
• Figures 4d-f depict fluorescence images of the cell before dilution, 10 times and 100 times dilution.

As one can clearly see the difference is quite profound, even by eye. An observation of Figure 4g yields the quantitative picture after dilution. The first peak centered on I=0.142 is the fluorescent background. The brightest peak, centered on I=0.256 corresponds to cell-surface fluorescence. It is quite easy to separate the two intensities as being from one versus the other.

[Note: What would have been really nice would have been a comparison of this same graph for an undiluted set of cells. The graph looks nice and all, and does demonstrate what it needs to. However I can't tell how big of an improvement it is data-wise without a direct comparison to what it was like before, after 10x and 100x dilution. Quite the shortcoming for the final conclusion of the paper.]

## Discussion & Conclusions

The authors have presented a method that provides means to significantly reduce background noise in the performance of cell-based droplet assays. The washing step is performed through electrocoalescence followed by a breakup to reduce drop size post-dilution. It is also possible to screen cells at the same high throughput without external washing. External washing would break the emulsion and risk losing a significant fraction of cells.

The improvement in the signal to noise ratio demonstrated in the paper is certainly has many practical applications.

On a more personal note, the paper highlights how microfluidics takes advantages of the various properties of emulsions in the encapsulation of reagents in droplets suspended in an oil medium, thus allowing for separation of each droplet, yet allowing for high throughput by simple fluid flow. Although some details describing the process are lacking, it is not quite unexpected of a paper to omit these important details.