A droplet microfluidic platform for high-throughput photochemical reaction discovery

The implementation of continuous flow technology is critical towards enhancing the application of photochemical reactions for industrial process development. However, there are significant time and resource constraints associated with translating discovery scale vial-based batch reactions to continuous flow scale-up conditions. Herein we report the development of a droplet microfluidic platform, which enables high-throughput reaction discovery in flow to generate pharmaceutically relevant compound libraries. This platform allows for enhanced material efficiency, as reactions can be performed on picomole scale. Furthermore, high-throughput data collection via on-line ESI mass spectrometry facilitates the rapid analysis of individual, nanoliter-sized reaction droplets at acquisition rates of 0.3 samples/s. We envision this high-throughput screening platform to expand upon the robust capabilities and impact of photochemical reactions in drug discovery and development.


Table of Contents
Droplet generation from microwell plates (MWPs) was performed using equipment and methods described previously (1,2). Samples were drawn into either 100 µm inner diameter (i.d)

ESI-MS Analysis of Droplet Samples
Tubing containing droplets was threaded through a capillary electrophoresis (CE) ESI-MS sprayer (Agilent Technologies, Santa Clara, CA) until approximately 0.5 mm was protruding. Sheath and droplet flows were driven by Fusion 400 syringe pumps (Chemyx, Stafford, TX). Droplets were flowed into the sheath sprayer (Supplementary Figure 2) and merged with a dilution stream of 50:50 methanol:water w/ 0.5% formic acid (100 µL/min flow rate). ESI-MS analysis was performed on an Agilent 6410 triple quadrupole mass spectrometer (Agilent Technologies, Santa Clara, CA). ESI potential was set to 2500 V, nebulizer gas to 15 psi, and drying gas from MS source was flowed at 10 L/min at 325 ºC. Mass spectrometer was set to scan from 75 to 750 m/z at 73 ms per scan. Droplet responses for any given m/z value were taken as the average of 3 consecutive data points from within each droplet's observed peak.

LED Light Strip Photoreactor Setup
This photoreactor setup was used for the photoredox trifluoromethylation experiments described in Figure 2 of the main text and Supplementary Figures 3 and 9 of the SI. A 150 mm wide x 15 mm deep polystyrene petri dish was lined with aluminum foil to promote internal reflection of light. A 4.4 W blue LED strip (Creative Lighting Solutions, Columbia Station, OH) was placed around the edge of the interior of the dish (Supplementary Figure 2). A small slit was cut from the petri dish wall to run wires and tubing through. A small hole was cut out of the reactor wall to allow for the LED power cord and droplet tubing to enter. A coil of perfluoroalkoxy (PFA) tubing (100 µm inner diameter; 360 µm outer diameter) was coiled around the center of the reactor 2 cm from the light strip. Tubing with droplets runs through the middle of the sprayer. Sheath liquid flows directly around tubing (Blue arrow). Electrospray is aided by use of nebulizer gas (Black arrow) (c) Photoreactor setup. To perform in-tubing droplet reactions, a petri dish was coated with aluminum foil, with an LED array (i) lining the rim and reactor PFA tubing (ii) coiled at the center.

General Procedure A: Preparation of Trifluoromethylation Reaction Solutions
Photocatalyst (1 mol%), pyridine N-oxide (4 equiv), and acetonitrile (0.2 M) were added to a vial charged with a stir bar. The solution was sparged with a stream of nitrogen gas for 5 min. Acetic anhydride (4 equiv) was subsequently added, and the solution was stirred for 10 min to facilitate formation of the acylated species. Separate solutions of substrate in acetonitrile (0.2 M) were also prepared. 10 µL of each solution were combined in a PCR tube to yield the final reaction mixture.

Procedure for In-Droplet and Non-Droplet Control Experiments
Experiments were conducted to compare reactivity in 4 nL droplets with batch reactions run on a standard multiwell plate screen scale of 20 µL (Supplementary Figure 3). 4 nL droplet reactions were prepared and run according to the aforementioned procedures and irradiated for 10 min inside of our photoreactor. 20 µL reactions were performed in PCR tubes, with the PCR tubes placed directly in the middle of the photoreactor for 10 min of irradiation. Following irradiation, the solutions were then formed into 4 nL droplets for direct comparison to the two volume scales. The quotient of the product signal over the summed product (P) and substrate (S) signals ( + ) was used to appraise reaction progress. For the two substrates that performed the best in the 20 µL reactions (1 and 2), only slight increases in product formation were observed when run in droplet format; however, the increase was drastic for the lower performing substrates (3 and 4) (Supplementary Figure 3). The changes in reaction performance can be attributed to the narrower sample geometry. The 100 µm i.d. tubing presents a substantially narrower pathlength, lowering the amount of light absorbed and possibly promoting more uniform irradiation across the entire sample. Such an effect could be helpful in promoting the observation of product in poorly performing reactions, or in reducing reaction time requirements in screening for both flow reaction and batch reaction screening.
Supplementary Figure 3. Comparison of in-droplet reactions vs. non-droplet batch reactions. (a) General schemes for running reactions at different scales (Left) Reactions run at 20 µL were irradiated immediately after mixing in PCR tubes and then reformatted into 4 nL droplets for analysis (Right) For in-droplet reactions, premixed solution was reformatted into 4 nL droplets, which were then irradiated. (b) Evaluation of performance across 4 substrates in either 20 µL or 4 nL volume. In every case, P/(P+S) response was found to be similar or significantly higher in droplet format. N=20 droplets for each reaction. (c) Example spectra from both 20 µL (Top) or 4 nL (Bottom) volume PF15 reactions. The yellow arrow indicates substrate m/z value, while the red arrow indicates product m/z value. In the 20 µL reaction, the substrate response was over double that of the product; however, the product response was even greater than that of the substrate in the 4 nL reaction.

LED Array Photoreactor Setup
This photoreactor setup was used for the photoredox Smiles-Truce Rearrangement experiments described in Figure 3 of the main text and Supplementary Figures 4 and 7. A 25 LED array of Cree Royal Blue XTE LEDs (2 W per LED, 50 W total output) was assembled (Supplementary Figure 4). The LEDs were mounted onto a heat sink, with two fans placed below and adjacent to the heat sink, in order to provide sufficient cooling to maintain reactions at ambient temperatures. An acrylic shield positioned 5 cm above the LED array provided a mounting stage for the reactor tubing, as well as an additional layer of protection for the LEDs. A custom-built plastic amber light shield (built by Ann Arbor Plastics, Saline, MI) was placed around the setup for user eye protection.

General Procedure B: Preparation of Smiles-Truce Rearrangement Reaction Solutions
To a flame dried 1-dram vial was added tetrabutylammonium benzoate (30 mol%), and [Ir(dF(CF3)ppy)2(5,5'd(CF3)bpy)]PF6 photocatalyst (1 mol%). The vial contents were then dissolved in anhydrous acetonitrile (0.2 M). Finally, the alkene was added (1.2 equiv). This solution was sparged under argon for 15 min. Separate solutions of substrate in acetonitrile (0.2 M) were also prepared. For reactions formed directly from well-plates, 10 µL of each solution were deposited into a well to form the final reaction mixture. Three droplets were made for every reaction condition, with the average droplet response reported. 10 droplets for every sulfonamide substrate were run in which no alkene reagent was present to generate an average control value.
Droplet reactions were irradiated based on the method outlined in General Procedure C (see below).

Oscillating Flow Reactor
To allow for extended reaction times in-flow inside of our PFA tubing, an oscillatory flow scheme was employed (Supplementary Figure 5). We implemented a visible light-driven alkene aminoarylation reaction as our system of choice to provide us with a manifold for performing high throughput reaction discovery in continuous flow. Droplets were formed from substrate and reaction mixture into a 100 µm i.d. PFA tube. In Supplementary Figure 5, droplet contents alternating between containing the N-((3,4-difluorophenyl)sulfonyl)acetamide substrate and the N-((4-cyanophenyl)sulfonyl)acetamide substrate, denoted as 3,4-F and 4-CN sulfonylacetamides respectively. A PCR tube had a 400 µm hole drilled in the cap and was filled with PFD. The outlet of the tubing was threaded through the hole and submerged in PFD to avoid evaporation of samples inside of the tubing. Upon irradiation, the droplets were flowed at 200 nL/min, first withdrawing towards the syringe pump for 10 min, followed by 10 min of infusing away from the syringe. This process was performed 3 times, allowing for 1 hour of continuous flow reaction. The current setup allowed for simultaneous irradiation of 40 reaction droplets and could conceivably increase to >100 droplets with longer tubing lengths and shorter oscillation periods. As currently demonstrated systems for oscillating flow in a reactor have been limited to a single plug, such a setup would amount to over 100x more samples to be irradiated in a single incubation period (3).

General Procedure C: In-Droplet Reaction Screen Setup
Following droplet generation, reactor tubing (100 µm i.d., 360 µm o.d.) containing droplets was wrapped around a 100 x 50 mm glass recrystallization dish and placed on top of our 25 LED array light setup (Supplementary Figure 4) for irradiation. Droplet reactions were run at a flow rate of 200 nL/min in an oscillatory manner by programming a syringe pump to alternate between refill and infusion modes at 10 min intervals, yielding a total residence time of 1 h. Following irradiation, droplet samples were characterized by ESI-MS analysis. Upon ESI-MS analysis, product m/z signals from each reaction were compared to the same m/z signal from the controls (blank droplets) to generate the heatmap in Figure 3. A minimum increase of 10 5 in extracted ion count (EIC) was set as the threshold for a hit, as it represented the lowest increase required to observe droplet signal over background noise. For reagent addition experiments, droplets were formed from substrate solution, and reagents were added on-chip.

Chip Fabrication
Microfluidic chips were fabricated using standard soft lithography procedures (4,5). SU-8 2050 photoresist was spun to 100 µm depth on silicon wafers (University Wafer, Boston, MA) then developed using photolithography to form negative masters. Uncured PDMS (Curbell Plastics, Livonia, MI) was poured on top of clean masters or blank wafers and allowed to cure for 1 h at 65 °C. Patterned PDMS and blank PDMS were baked for 1 h at 150 °C, followed by 1 min of exposure to atmospheric plasma and baking for 2 h at 150 °C to create an irreversible bond. Chip channel surfaces were treated with 2% trichloro(1H,1H,2H,2H-perfluorooctyl)silane in PFD by flowing 10 µL internal volumes through over 10 min followed by 2 h of baking at 65 ºC. Chips were soaked in acetonitrile overnight to prevent solvent loss from droplets. Channels fabricated into PDMS devices were 100 µm in depth. Droplets were flowed in from a 100 µm wide channel that expanded to 200 µm wide at the point of intersection with the reagent addition channel, which was 100 µm wide. The final device was 200 µm wide at all openings to accommodate direct insertion with 360 µm o.d. tubing. Channels were wetted with PFD to help ease insertion of tubing (Supplementary Figure 8).

Reagent Addition Chip Screen
In-droplet photoreactions with online MS analysis were performed using reagent addition chips. Reagent addition PDMS chips were fabricated to enable trains of droplets to be imported from 100 µm i.d. PFA tubing, flowed through an addition region to receive reagents, and then be exported to 150 µm i.d. PFA tubing for irradiation and ESI-MS analysis. Larger 150 µm i.d. tubing was used for the droplets post-addition as the larger volume droplets were sometimes unstable in the 100 µm i.d. tubing (2). Consistent addition of reagent to 4 nL acetonitrile droplets was achieved with the employed geometry (Supplementary Figure 8). By keeping droplet flow consistent (800 nL/min), the amount of reagent added to each droplet was controllable by the flow of the reagent stream. The reagent solution for the trifluoromethylation reaction, consisting of photocatalyst, pyridine N-oxide, and TFAA, (see General Procedure A) was utilized in the demonstration of the reagent addition device, as it showed a deep yellow color. At 100 nL/min reagent flow, the final droplets were composed of 33 ± 2% added reagent, while at 200 nL/min reagent flow created droplets with 45 ± 4% added reagent, showing consistent addition to droplets at both flow rates. Also tested for this geometry was the carry-over between droplets. To test for this, droplets were made in 10 x 10 units, alternating between being composed of either blank acetonitrile or trifluoromethylation reagent solution, and blank acetonitrile used as the addition stream. As the reagent addition stream is now colorless, any material carry-over from trifluoromethylation reagent droplets into the addition stream will lead to a yellow hue in the proceeding droplets. From this approach, blank acetonitrile droplets following trifluoromethylation reagent droplets had no observable yellow coloration, showing that this geometry can be performed with minimal carry-over between droplets (Supplementary Figure 8).
The above system was applied to the Smiles-Truce rearrangement described in the main manuscript. 4 nL acetonitrile droplets containing the 4-CN sulfonylacetamide substrate and transanethole (Supplementary Figure 9) segmented by 12 nL PFD were flowed through the reagent addition device at 800 nL/min, with 200 nL/min reagent addition flow, creating 7 nL full reaction droplets for irradiation and analysis. Irradiation time was approximately 7 min, calculated from the volume of the tubing contained within the reactor and the 1000 nL/min volumetric flow rate. Analysis of droplets post-irradiation at 20 droplets/min not only showed that product had formed, but that the formation was highly consistent across all the droplet samples (Supplementary Figure  9). To confirm that signal was a result of in-droplet chemistry, premixed reaction mixture was made into droplets and analyzed without irradiation. Minimal signal was observed, with the droplet samples barely distinguishable from background noise. These results indicate the successful application of our system for performing and analyzing in-droplet photoredox reactions.
Supplementary Figure 9. In-droplet Smiles-Truce rearrangement reaction using reagent addition device. MS traces represent the m/z of the Smiles-Truce rearrangement product (m/z = 309). (a) Droplet samples processed with online flow reactor. (b) Control samples of (a) which was not exposed to blue LED irradiation. Minimal signal was observed, indicating that results from (a) are the result of in-droplet chemistry.