High-Throughput Screening to Obtain Crystal Hits for Protein Crystallography

Getting a crystal to perform diffraction experiments remains a challenge because it's difficult to predict what parameters will influence crystallization. We favor the experimenter by screening 1536 crystallization conditions in a single plate and use advanced imaging technologies to identify even the smallest crystal hits. Structural biology techniques are developing rapidly, and the field has been revolutionized by computational structure prediction.

This has caused the field to become more integrative with several experimental or computational approaches more commonly being combined to generate a more detailed and accurate representation of biological mechanisms. Generating crystal hits is a key step to performing single-crystal X-ray diffraction experiments and to developing techniques, like MicroED and serial crystallography. Using advanced imaging methods, UV-TPF and SHG, along with the power of the MARCO algorithm, helps us identify useful crystals at all size scales.

Our high-throughput crystallization methods have generated a large amount of data for probing questions regarding efficiency of crystallization cocktail components. The specialized imaging we do reveal how non-linear optical methods can be used to detect vanishingly small crystals. Finding crystallization conditions is critical for crystal-based structural methods, which account for 90%of all structural models in the protein data bank.

In 2021, close to 2 million files were downloaded per day from the PDB, emphasizing the impact structures have in paving the way for further scientific investigations. To begin, use a robotic liquid handling system for delivering five microliters of paraffin oil to the wells of a 1, 536-well plate. Invert the thawed cocktail containing 384 well plates to mix the solutions.

Incubate the plates at 30 degrees Celsius to dissolve any persistent precipitates. Dispense equal measures of the buffered PEG 3350 and the additive screen components from the 96-well plate into the 384-well plate to a final volume of 50 microliters. Stamp four 384-well plates into a 1, 536-well plate, filling all four quadrants.

Using a liquid handling robot equipped with a 384 syringe head, deliver 200 nanoliters of cocktail solution into each well of the 1, 536-well plate. Centrifuge the unpacked thawed sample at 10, 000 g for two minutes at room temperature. Observe the sample to identify any precipitation, color, and condition.

Warm the prepared 1, 536-well plate to 23 degrees Celsius, and then centrifuge it at 150 g for five minutes. Using a brightfield microscope, take pictures of the plate containing the cocktail only to be used as a negative control. Dispense 200 nanoliters of the sample to each well using a liquid handling robot.

After centrifuging the plate at 150 g, incubate at the desired temperature for six weeks. Capture brightfield images of the plates with samples at weekly intervals from day one to week six. Perform sonic imaging with the SHG and UV-TPEF.

Automated plate imaging throughout the experiment duration allowed for the brightfield identification of rapidly and slow-growing crystals. The UV-TPEF imaging indicated the proteinaceous nature, while the SHG imaging confirmed the crystalline nature of the sample. Sonic imaging enabled the identification of crystals obscured by precipitation or film, which do not produce an SHG signal or microcrystals that may be mistaken as precipitate.

Non-protein Crystals lacking UV-TPEF signal, but displaying strong brightfield and SHG signals were also identified. To analyze the crystallization images, first open the AI-enabled, open source graphical user interface. Next, click on Import, select Images from the dropdown menu and then choose From Raw Archive/Directory.

Click on Browse For Folder in the pop=up window, and then navigate to the folder containing the images. Choose the desired files and import them into the interface by clicking Open. Once the selected files appear in the Selected Paths window, select one or more to download into the interface and then click on Import Runs.

To view the image of the first well, click on the greater than symbol to the left of the sample name on the window of the Slideshow Viewer. Then double-click on the appropriate read. Resize the whole window to enlarge the image.

The Image Details box includes information about the image and scoring. The Cocktail Details box contains metadata about the cocktail components. To move to the next well, click on the Next button in the navigation panel or press the right arrow key on the keyboard.

A specific well can be navigated by entering the well number in the By Well Number window. All reads can be viewed by checking the Show All Dates box. All spectra can be viewed by checking the Show All Spectra box.

Each individual spectrum image can be viewed by clicking on the Swap Spectrum button. To score the crystal images using the MARCO algorithm, first select a specific run from the list on the left side of the window, then click on Classify Selected Run and view the MARCO scoring information in the Image Details window once all 1, 536 wells have been scored. To view a subset of the scored images, for example, the images classified as crystals by MARCO, in the Image Filtering panel, tick the Crystals and MARCO boxes and then click on the Submit Filters button.

To generate a manual human scored set, assign scores to each well by clicking the appropriate button located in the Classification panel at the bottom of the window, or use the keyboard number pad to assign a score.