Time-resolved serial femtosecond crystallography (tr-SFX) offers significant advancements in structural biology, but its application is limited by the reliance on light-inducible systems. Most biological systems, especially enzymes, are not light-dependent, requiring alternative triggering methods for studying their catalytic mechanisms. While techniques like caged substrates and protein modification exist, these are sample-specific and not generally applicable. Studying enzyme catalysis under near-physiological conditions using tr-SFX promises deeper insights into reaction mechanisms and can guide therapeutic strategies. Triggering enzyme reactions by mixing protein crystals with substrates is a promising approach, with theoretical calculations suggesting that millisecond diffusion times are achievable in small crystals. Several mixing approaches have been developed for both XFELs and synchrotrons, but these often involve high sample flow rates, making them unsuitable for scarce samples. Existing techniques such as liquid jets and the LAMA method, although offering some improvements, still have limitations in terms of sample consumption and time resolution. This research focuses on developing a more efficient and versatile on-demand mixing strategy to overcome these challenges.

Existing methods for time-resolved serial crystallography (SFX and SSX) often rely on continuous flow systems or methods that are not readily adaptable to a wide range of enzymes and substrates. Continuous flow methods, while effective, consume substantial quantities of sample, especially when multiple time points are required to capture the entire reaction cycle. The use of light-inducible systems necessitates the use of light-sensitive proteins or the engineering of light-sensitive substrates, limiting the scope of applicable enzymes. The liquid application method for time-resolved studies (LAMA) provided an improvement, but still faces challenges in terms of sample throughput and achieving sufficient time resolution for fast enzymatic reactions. The research aims to address the limitations of existing approaches by presenting a novel technique that improves time resolution and reduces sample consumption.

This study introduces a novel drop-on-drop sample delivery system for time-resolved mixing experiments. This system adapts the drop-on-tape method, known for its fidelity, low sample consumption, and versatility. The modified setup employs two droplet dispensing heads: an acoustic droplet ejector (ADE) for delivering a larger, crystal-containing drop, and a piezoelectric injector (PEI) for dispensing bursts of smaller, picoliter-sized drops of concentrated ligand solution. The PEI can operate at frequencies of 1–30 kHz, allowing for rapid mixing. The speed of the Kapton tape carrying the drops is adjustable (600–10 mm s⁻¹), enabling the control of mixing time (0.1–6 s). The number of substrate drops added to the crystal drop can also be controlled, allowing for a variable amount of substrate delivered to the protein crystals. The researchers used numerical simulations to assess the mixing efficiency of the drop-on-drop method, comparing it with diffusion-only scenarios and analyzing the interplay of droplet collisions, hydrodynamic flow, and diffusion. Proof-of-principle mixing experiments with fluorescent dyes were performed to validate the simulations. Two enzyme systems were used to demonstrate the system's applicability to enzyme-catalyzed reactions: hen egg white lysozyme (HEWL) with the inhibitor N-acetyl-D-glucosamine (GlcNAc), and a bacterial serine β-lactamase (CTX-M-15) with the β-lactam antibiotic ertapenem. Time-resolved serial femtosecond crystallography (tr-SFX) experiments at SACLA XFEL were conducted to collect diffraction data at various time points after ligand addition. Data processing involved standard methods for serial crystallography, including data reduction, indexing, merging, and refinement. Isomorphous difference and polder OMIT electron density maps were employed to analyze ligand binding and conformational changes.

Numerical simulations showed that the collision-driven mass flow generated by the high-frequency and high-velocity impact of substrate droplets significantly accelerates mixing compared to diffusion alone. Experimental results using fluorescent dyes confirmed that mixing is much faster than predicted by diffusion models, achieving rise times of <100–150 ms. Time-resolved serial crystallography data for HEWL revealed clear evidence of GlcNAc binding within 0.6 s of ligand addition, demonstrating the system's capability to capture events on sub-second timescales. Similarly, for CTX-M-15, binding of ertapenem was observed within 2 s, providing structural evidence of the acyl-enzyme intermediate. The drop-on-drop method achieved sub-second time resolution with remarkably reduced ligand consumption, requiring only small volumes of ligand solutions and microcrystal slurries. The method is adaptable to a wide range of delay times (50 ms to 10 s) and crystal sizes (5–100 µm), providing flexibility in optimizing experimental conditions for various protein systems. The successful application to two different enzyme systems highlights the versatility of the approach.

The drop-on-drop method presented in this study significantly improves upon existing techniques for studying enzyme catalysis using serial crystallography. The sub-second time resolution achieved, combined with significantly reduced sample consumption, addresses critical limitations of previous approaches. The ability to capture ligand binding and early reaction intermediates on a timescale relevant to enzymatic reactions opens up new avenues for investigating reaction mechanisms and dynamics. The high-frequency, picoliter-droplet mixing strategy demonstrates a practical and effective way to initiate reactions within crystals, minimizing the time required for ligand diffusion to reach the active site. The adaptability of the system to a range of time delays and crystal sizes expands its potential applications. The demonstrated success with both HEWL and CTX-M-15, representing different enzyme types and substrate affinities, suggests broad applicability to various enzyme-substrate systems. Future research could explore optimizing parameters such as droplet size and collision frequency for even faster mixing.

This research successfully demonstrates a novel on-demand, drop-on-drop method for time-resolved serial crystallography that overcomes limitations associated with ligand diffusion, achieving sub-second time resolution with significantly reduced sample consumption. The method's versatility and efficiency expand the capabilities of dynamic structural biology studies, and its transferability from XFEL to synchrotron sources promises broader accessibility. Future research should explore its application to a wider range of enzyme systems and more complex biological processes.

The current setup primarily focuses on time points within a relatively narrow time window (0.1–6 s). Although adjustable, extending the achievable time delays may require additional modifications. While the system demonstrated its capability with two model systems, more extensive validation across a wider range of enzymes and substrates is required. The method is currently limited by the availability of appropriate XFEL or synchrotron beamlines with suitable pulse rates.