Multicolor fluorescence detection in droplet microfluidics typically involves bulky and complex epifluorescence microscope-based detection systems. Here we describe a compact and modular multicolor detection scheme that utilizes an array of optical fibers to temporally encode multicolor data collected by a single photodetector.
Fluorescence assays are the most common readouts used in droplet microfluidics due to their bright signals and fast time response. Applications such as multiplex assays, enzyme evolution, and molecular biology enhanced cell sorting require the detection of two or more colors of fluorescence. Standard multicolor detection systems that couple free space lasers to epifluorescence microscopes are bulky, expensive, and difficult to maintain. In this paper, we describe a scheme to perform multicolor detection by exciting discrete regions of a microfluidic channel with lasers coupled to optical fibers. Emitted light is collected by an optical fiber coupled to a single photodetector. Because the excitation occurs at different spatial locations, the identity of emitted light can be encoded as a temporal shift, eliminating the need for more complicated light filtering schemes. The system has been used to detect droplet populations containing four unique combinations of dyes and to detect sub-nanomolar concentrations of fluorescein.
Droplet microfluidics provide a platform for high throughput biology by compartmentalizing experiments in a large number of aqueous droplets suspended in a carrier oil 1. Droplets have been used for applications as varied as single cell analysis 2, digital polymerase chain reaction (PCR) 3, and enzyme evolution 4. Fluorescent assays are the standard mode of detection for droplet microfluidics, as their bright signals and fast time response are compatible with detecting sub-nanoliter droplet volumes at kilohertz rates. Many applications require fluorescence detection for at least two colors simultaneously. For instance, our lab commonly performs PCR-activated droplet sorting experiments that use one detection channel for the result of an assay, and uses a secondary background dye to make assay-negative droplet countable 5.
Typical detection stations for droplet microfluidics are based on epifluorescence microscopes, and require complicated light manipulations schemes to introduce excitation light from free space lasers into microscope to be focused on the sample. After fluorescence is emitted from a droplet, the emitted fluoresced light is filtered so that each detection channel utilizes one photomultiplier tube (PMT) centered on a wavelength band. Epifluorescence microscope-based optical detection systems provide a barrier to entry due to their expense, complexity, and required maintenance. Optical fibers provide the means to construct a simplified and robust detection scheme, since fibers can be manually inserted into microfluidic devices, removing the need for mirror-based light routing, and allowing light paths to be interfaced using optical fiber connectors.
In this paper, we describe the assembly and validation of a compact and modular scheme to perform multicolor fluorescence detection by utilizing an array of optical fibers and a single photodetector 6. Optical fibers are coupled to individual lasers and are inserted normal to an L shaped flow channel at regular spatial offsets. A fluorescence collection fiber is oriented parallel to the excitation regions and is connected to a single PMT. Because a droplet passes through the laser beams at different times, data recorded by the PMT shows a temporal offset that allows the user to distinguish between the fluorescence emitted after the droplet is excited by each distinct laser beam. This temporal shift eliminates the need to separate emitted light to separate PMTs using a series of dichroic mirrors and bandpass filters. To validate the efficacy of the detector, we quantitate fluorescence in droplet populations encapsulating dyes of different color and concentration. The sensitivity of the system is investigated for single color fluorescein detection, and shows the ability to detect droplets with concentrations down to 0.1 nM, a 200x sensitivity improvement as compared to recent fiber based approaches reported in the literature 7.
1. SU8 Master Fabrication
2. PDMS Device Fabrication
3. Preparation of Optical Components
4. Offline Mixed Emulsion Generation
5. Optical Fiber Insertion
6. Fluorescence Detection of Mixed Emulsions
Fabrication of a PDMS device that allows for the insertion of optical fibers requires a multistep photolithography procedure to create channels of varying height (Figure 1). First, an 80 µm tall layer of SU-8 is spun onto a silicon wafer and patterned using a mask to create the fluid handling geometry. Next, an additional 40 µm layer of SU-8 is spun onto the wafer, and patterned using a second mask to create features that will form 120 µm tall laser fiber insertion channels. Last, 100 µm more SU-8 is spun onto the wafer, and patterned to give 220 µm tall detection fiber insertion channels. The master is developed and used to mold a PDMS device, which is in turn bonded to a glass slide. The final fabrication contains 80 µm tall flow geometry, accessed via holes punched in the top of the device and 120 µm and 220 µm tall optical fiber channels accessed through the side of the device.
The geometry of the device used for detection is shown in Figure 2. Externally generated 80 µm emulsions are re-injected into a drop reinjection port and spaced by spacer oil before they flow into a L shape detection channel. Two laser-coupled fibers are inserted into 120 µm tall channels provide excitation at discrete spatial locations in the flow channel. Because the multimode fibers inserted into these channels have a 125 µm external diameter and a 105 µm core diameter, the entire cross section of the flow channel is illuminated by laser light. In our device the detection flow channel makes an abrupt 90 degree turn after the last laser excitation region to create close optical access for a 225 µm OD fiber used to detect emitted fluorescence. The larger fiber is used because of its high numerical aperture (NA = 0.39), which is similar to a standard microscope objective.
A droplet flowing through the detection channel encounters discrete excitation regions in front of each of the laser-coupled optical fibers (Figure 3A). The single PMT coupled to the detection fiber records emitted fluorescence with a temporal shift the maps to the excitation source. For a droplet that contains dyes that excite at 473 nm (region "1" in Figure 3A) and at 405 nm (region "2" in Figure 3B), the resulting PMT signal contains a signal doublet with a peak separation that correlates with the time it takes a droplet to move from one excitation region to the next. By temporally encoding spectral data in this manner, the need for further light filtering and separate PMTs for each fluorophore color is eliminated. Figure 3B shows signal doublets for a train of 4 droplets with different dye combinations. The first peak in the double corresponds to the fluorescence emitted after excitation by the 473 nm laser in region "1" and the second peak corresponds to fluorescence emitted after excitation by the 405 nm laser in region "2".
The efficacy of the detection scheme was tested by detecting a mixed emulsion of droplets containing dyes that excite at 405 nm (dextran-conjugated blue, CB) and at 473 nm (fluorescein, FITC). Droplets were flowed through the detection regions at roughly 50 Hz and data was recorded for 60 sec, then post-processed with a computing program script. The data is plotted in Figure 4 and shows clear separation between the four re-injected droplet types. Because the detection system does not differentiate between different colors of emitted light, each droplet is described by its maximum total fluorescence after being excited in region "1" (peak 1) and in region 2 (peak 2).
The dynamic range and absolute sensitivity of the detector is investigated by turning off the 405 nm laser and measuring the fluorescence of an emulsion containing fluorescein-only droplets (Figure 5). The data shows the ability to detect dye concentration that range from 0.1 to 100 nM FITC on a single detector with the same detection settings.
Figure 1. Multi-layer photolithography. The microfluidic device is fabricated by creating a master with three different feature heights. First, a 80 µm tall layer of SU-8 is spun on and patterned using a mask for the fluid channels. Next, 40 µm more SU-8 is spun on and 120 µm tall features are patterned with a 2nd mask through both of the SU-8 layers to yield geometry to accommodate 125 µm tall optical fibers. After this, 100 µm more SU-8 is spun on and patterned with a 3rd mask to yield 225 µm tall features. After developing, PDMS molding, and plasma bonding to glass, the resultant device contains large channels accessible from the side for fiber insertion, in addition to standard enclosed fluid channels. Please click here to view a larger version of this figure.
Figure 2. Device design and optical layout. The fluidic portions of the device are comprised of a droplet reinjector and an oil spacer, coupled to an L-shaped channel. Fibers coupled to individual lasers are inserted normal to the direction of flow. A detection fiber is inserted perpendicular to the laser fibers in order to collect fluorescent light. This light is filtered through a bandpass filter, before being detected by a single PMT. Please click here to view a larger version of this figure.
Figure 3. Temporal encoding of multicolor information. (A) Droplets flow through excitation regions "1" and "2" at different times, providing time shift that allows the identification of the emitted light. (B) Data from a train of 4 different droplets flowing through the detection channel. The first peak of each doublet corresponds to the fluorescence emitted after excitation in region "1" and the second peak in each double corresponds to the fluorescence emitted after excitation in region "2". The height of droplet peaks are proportional the amount of dye excited at a given wavelength, except when spectral bleed occurs due to the use of dyes with overlapping excitation spectra. The width of the peaks are determined by the amount of time that it takes a droplet to traverse an excitation region. Please click here to view a larger version of this figure.
Figure 4. Multicolor detection of a mixed droplet population. Scatter plot showing maximum droplet intensities for a mixed droplet population after excitation and emission by a 473 nm laser (peak 1) and a 405 nm laser (peak 2). The droplets contain combinations of fluorescein (FITC) and cascade blue (CB) at nM concentrations. Please click here to view a larger version of this figure.
Figure 5. Single color detection on highly varied droplet population. Histogram of emissions by a mixed droplet population containing 0.1-100 nM of fluorescein after excitation by a 473 nm laser. Please click here to view a larger version of this figure.
Fiber optic detection requires the alignment of optical fibers with respect to fluid channels. Because our device utilizes guide channels fabricated with multilayer photolithography, placement of masks with respect to each other is of great importance. If the fiber guide channels are too close to the fluid channel, there is a potential for fluid leakage; if the guide channels are located too far away or misaligned, the fluorescence signal gathered by the detection fiber may be significantly diminished. Proper alignment can be aided by designing alignment marks, such as concentric circles into masks to be co-located during photo patterning. Additionally, manual insertion of fiber optics into the device is a delicate process with the potential to break the tips of fibers within the device, rendering them unusable. Restraining fiber against the glass slide bonded to the PDMS device and inserting the fibers slowly while observing with the microscope are keys to successful fiber insertion. Last, reinjection of pre-formed emulsions should occur with a minimum of handling. Ideally, once the emulsion is made and deposited directly into an empty syringe, it should not be transferred until it is used for droplet reinjection.
Multicolor detection requires that the peaks in a signal doublet be distinct from one another and that signals from consecutive droplets do not overlap. The detection setup that we use here uses a pair of 125 µm optical fibers separated at a center-to-center distance of 300 µm, providing a non-illuminated region of ~175 µm between excitation regions. This allows an 80 µm droplet to be excited by only one light source at a time and ensures that each of the peaks in a doublet is solely a result of one type of laser excitation. Designs where the separation between the excitation regions is similar to the droplet diameter give problematic signals with overlapping emission peak doublets. The width of a signal doublet correlates to the time that it takes for the leading edge of a droplet to enter the first excitation regions to the time the trailing edge of the droplet to leave the final excitation region, a distance of ~500 µm in the device we describe. In order to maintain distinctness of adjacent detected droplets, signal doublets need to be spaced by a low signal region at least as wide as the signal doublet. For the configuration described here, 80 µm droplets are spaced at least 1 mm apart.
Although multi-color detection with a single photodetector allows for significant savings in terms of cost and complexity compared to standard techniques, there is some loss of robustness because emitted light is not filtered by wavelength. This can be problematic when detecting droplets that contain multiple fluorophores that excite at the same wavelengths, because the fluorescence source is indistinguishable. These limitations can be overcome by performing experiments with nonoverlapping excitation profiles, or by designing experiments where spectral bleed over does not occur into a "channel" where a sensitive assay is being performed. This is compatible with many droplet microfluidic detection applications, where a bright background dye is used to indicate the presence of a droplet, and a second fluorophore is used to report the result of a molecular biology assay 10,11.
Compared to standard techniques, our approach has two main advantages: 1) spatially registered excitation regions eliminate the need for complicated light filtering, and 2) use of fiber optics removes the needs for a microscope and optical hardware to direct light. Because we address common technical concerns with microfluidic detection systems, previous investigators have developed methods to address similar concerns. For instance, Martini et al. frequency-encode multicolor fluorescence recorded by a single detector by projecting an excitation source through uniquely patterned "bar code" filters 12. Our use of an optical fiber array allows the utilization of a similar temporal encoding scheme using standard, off the shelf components. The literature shows optical numerous optical coupling techniques that are more controlled that the simple fiber insertion that we describe. Bliss et al. use liquid filled optical wave guides for the carefully controlled delivery and collection of light from specific microchannel locations 13, Martinez Vasquez et al. use waveguides laser etched into fused silica substrates to deliver excitation light 14, and Vishnubhatla et al. are able to precisely fabricate fiber insertion channels using a combination of laser irradiation and chemical etching 15. Each of these techniques require additional steps beyond those used to fabricate the base fluidic architecture of a microfluidic chip. While these fabrication procedures may be necessary for highly sensitive detection applications, we have found that fiber insertion into a molded PDMS channel is adequate for most of the everyday applications in our lab. The detection system was able to detect FITC concentrations down to 100 pM in 80 µm droplets, roughly equivalent to the detection of ~17,000 fluorophore molecules. This sensitivity is adequate for most detection applications in our laboratory currently using an epifluorescence-based system, the most common of which being PCR-activated cell sorting of assay positive droplets containing the equivalent of 106 fluorophore molecules 11. The detection sensitivity also compares favorably to recently reported sensitivities of other optical fiber based systems — for instance, it is 200x more sensitive than the detection limit of 20 nM of Alexa Fluor 488 reported in > 100 µm droplets7.
We have presented a compact and modular droplet microfluidic multi-color detection scheme as an alternative to more costly, complex, and bulky epifluorescence microscope based systems. The necessary detection components used here (lasers, fibers, fiber adapters, filter, and PMT) can be purchased for <$10,000, making the multi-color droplet detection accessible to a great number of investigators, and allowing individual investigators to afford multiple detection stations in a laboratory. The system also remove some expertise-based barriers to entry, as a degree of familiarity is required to align free space optical systems in conjunction with an epifluorescence microscope. Additionally the small footprint of the detections setup makes it ideally suited for portable and diagnostic applications.
The authors have nothing to disclose.
This work was supported by DARPA grant number 84389.01.44908, an NSF CAREER award (DBI-1253293), an NIH exploratory/developmental research grant (CA195709), and NIH New Innovator Awards (HD080351, DP2-AR068129-01), and a New Directions grant from the UCSF resource allocation program.
Photomasks | CadArt Servcies | ||
3" silicon wafers, P type, virgin test grade | University Wafers | 447 | |
SU-8 3035 | Microchem | Y311074 | |
SU-8 2050 | Microchem | Y111072 | |
Sylgard 184 silicone elastomer kit | Krayden | 4019862 | |
1 ml syringes | BD | 309628 | |
10 ml syringes | BD | 309604 | |
27 gaugue needles | BD | 305109 | |
PE 2 polyethylene tubing | Scientific Commodities, Inc. | B31695-PE/2 | |
Novec 7500 | Fisher Scientific | 98-0212-2928-5 | Commonly knowns as HFE 7500 |
Ionic Krytox Surfactant | Synthesis instructions in ref #10 | ||
Dextran- conjugated cascade blue dye | Life Technologies | D-1976 | |
Fluorescein sodium salt | Sigma | 28803 | |
Quad bandpass filter | Semrock | FF01-446/510/581/703-25 | |
PMT | Thorlabs | PMM02 | |
Fiber port | Thorlabs | PAFA-X-4-A | |
lens tube | Thorlabs | SM1L05 | |
Patch cable with 200 um core / 225 um cladding optical fiber with one stripped end and one FC/PC connector | Thorlabs | Custom | |
Patch cable with 105 um core / 125 um cladding optical fiber with one stripped end and one FC/PC connector | Thorlabs | Custom | |
125 um fiber stripping tool | Thorlabs | T08S13 | |
225 um fiber stripping tool | Thorlabs | T10S13 | |
laser fiber adapter | OptoEngine | FC/PC Adapter | |
405 nm CW laser at 50 mW | OptoEngine | MDL-III-405 | Distributor for CNI lasers |
473 nm CW laser at 50 mW | OptoEngine | MLL-FN-473-50 |