High-Throughput Absorbance-Activated Droplet Sorting for Engineering Aldehyde Dehydrogenases

Abstract

Recent decades have seen a dramatic increase in the commercial use of biocatalysts, transitioning from energy-intensive traditional chemistries to more sustainable methods. Current enzyme engineering techniques, such as directed evolution, require the generation and testing of large mutant libraries to identify optimized variants. Unfortunately, conventional screening methods are unable to screen such large libraries in a robust and timely manner. Droplet-based microfluidic systems have emerged as a powerful high-throughput tool for library screening at kilohertz rates. Unfortunately, almost all reported systems are based on fluorescence detection, restricting their use to a limited number of enzyme types that naturally convert fluorogenic substrates or require the use of surrogate substrates. To expand the range of enzymes amenable to evolution using droplet-based microfluidic systems, we present an absorbance-activated droplet sorter that allows droplet sorting at kilohertz rates without the need for optical monitoring of the microfluidic system. To demonstrate the utility of the sorter, we rapidly screen a 10⁵-member aldehyde dehydrogenase library towards D-glyceraldehyde using a NADH mediated coupled assay that generates WST-1 formazan as the colorimetric product. We successfully identify a variant with a 51 % improvement in catalytic efficiency and a significant increase in overall activity across a broad substrate spectrum.

Figure 1

Operation of the iAADS platform. a) The optical path from source to detector. Collimated light from the microscope condenser passes through a 35 μm×25 μm transparent window within an opaque lithographic mask attached to the bottom of the glass substrate of the microfluidic device. The mask window positioned directly underneath the channel constitutes the detection region. Transmitted light is collected by the microscope objective and directed to a PMT. b) Illustration of the working principle of the iAADS platform. When droplets flow through the detection volume, the light transmitted through a single droplet is recorded by the PMT. Signals are continuously monitored by a FPGA, which based upon a user‐defined threshold can issue a sorting pulse. If a sorting pulse is generated, an actuating dielectric force moves the droplet towards the positive outlet, where it is detected by measuring an impedance variation as it passes across electrodes connected to a lock‐in amplifier. c) Left: Representative time traces managed by the FPGA. Right: Representative time traces displayed on the lock‐in amplifier software.

Figure 2

Characterization of the iAADS platform. a) Intensity histogram obtained from a mixture of droplets containing 0 μM, 25 μM, 50 μM and 100 μM fluorescein solutions. b) Average intensities of the component populations as a function of concentration indicate a concentration limit of detection of 7.1 μM and an analytical sensitivity of 0.001 VμM⁻¹. The means and standard deviations were extracted by fitting the histogram in panel (a) to a sum of four Gaussian components. 300,000 droplets were processed to generate these data. c) Representative time trace obtained when the sub‐population of 100 μM droplets shown in (a) are sorted at a rate of 950 Hz and with a sorting efficiency of 97 %. The auxiliary input signal is shown in red and facilitates the monitoring of sorted droplets (shown in blue).

Figure 3

Evolution of VpALDH. a) The screening assay used for the evolution of VpALDH. b) Histogram of absorbance signals from 180,000 individual droplets containing 0 μM, 75 μM, 150 μM or 300 μM WST‐1 formazan solutions. c) Extracted means and standard deviations, obtained by fitting the data in panel (b) to a sum of four Gaussian components. Such an analysis yields a concentration detection limit of approximately 15 μM for WST‐1 formazan. 100,000 droplets were processed to generate these data. d) A representative histogram for approximately 900,000 droplets observed during library screening.

Figure 4

Characterization of VpALDH variants. a) The quaternary (right) and tertiary structures (lower left) of VpALDH were predicted by AlphaFold2. The oligomerization interface between two monomeric subunits forming a larger ß‐sheet is shown in the upper left. The chains in the homo‐tetramer (right) are color‐coded and the mutations identified by re‐screening are highlighted in red. The catalytic cysteine at position 286 is shown as blue sticks (upper left, dotted circle). b) Correlation of half‐lives, t _1/2, against melting temperatures, T _m, is shown for the wild‐type VpALDH and the obtained variants. c) Percentage changes in specific activity of VpALDH variants compared to the wild‐type. Variants were tested on several C1 to C6 aldehydes and on one aromatic aldehyde at a concentration of 5 mM. Measurements were performed in triplicates (n=3) and standard deviations are shown as error bars.