Efficient Data Acquisition with Three-Channel Centerpieces in Sedimentation Velocity

Abstract

Three-channel 3D printed centerpieces with two sample sectors next to a joint solvent reference sector were recently described as a strategy to double the throughput of sedimentation velocity analytical ultracentrifugation experiments [Anal. Chem. 91 (2019) 5866-5873]. They are compatible with Rayleigh interference optical detection in commercial analytical ultracentrifuges, but require the rotor angles of data acquisition to be repeatedly adjusted during the experiment to record from the two sample sectors. Here we present an approach to automate this data acquisition mode through the use of a secondary, general-purpose automation software, and an accompanying data pre-processing software for scan sorting.

Nearly a century after the initial development of analytical ultracentrifugation (AUC) [1,2], this classical technique of physical biochemistry is again in rapid development [3-5]. This is rooted in the powerful and simple concept of observing particle redistribution in solution after application of a gravitational force, which can be adapted to modern problems. Taking advantage of new theoretical, computational, and experimental capabilities [6,7], the traditional areas of application have been expanded into a variety of fields such as food sciences [8], archeology [9], characterization of nanoparticles [10-14], biotechnology [15-18], and the study of extremely strong, and ultra-weak macromolecular interactions [19-22]. In the last few years, interest in instrumental developments has included new detectors [23,24], calibration and alignment procedures [25-29], and strategies for sample handling [30-32].

A long-standing limitation of AUC, however, is the relatively low throughput, which allows only 7–8 samples to be studied in a single run that may take several hours or days. This can pose a problem, in particular, for screening applications comparing many different molecules or solvents, protein interaction studies requiring titration series spanning a wide range of concentration, and studies of multi-protein interactions that require multiple control samples [7,10,15,33-37].

To increase the sample throughput, our laboratory has previously introduced an alternative strategy of using the absorbance detection system [38]. The absorbance detection system was originally designed to function like a dual-beam spectrophotometer with one reference and one sample solution [39]. However, we have shown that the logarithm of the recorded light intensity in each sector exhibits a data structure equivalent to that of interference optical data (the time-invariant radially-dependent baseline offset playing the same role as the local detector sensitivity profile), and therefore data from each sector can be analyzed separately using the same algebraic noise decomposition method [40,41]. This allows a standard double sector centerpiece to be operated with two independent samples, and thereby run SV experiments with twice the throughput [34,38]. Unfortunately, this popular strategy is limited to absorbance detection.

A key part of AUC instrumentation is the ‘centerpiece’, which holds a sector-shaped liquid volume between optical windows and is sealed against high vacuum during ultracentrifugation. Recently we have demonstrated how 3D printing technology liberates the experimenter from the constraints of commercial centerpieces that use designs going back half a century or more [30,31]. Current 3D printing allows the inexpensive fabrication of centerpieces that are watertight and provide a seal against the windows during exposure to centrifugal forces in the AUC, which in sedimentation velocity (SV) experiments can reach up to ~300,000 g at 60,000 rpm. Similarly, suitable spacers can be printed to elevate the centerpiece to the filling port of existing cell assembly barrels [31]. While they can be printed in a variety of materials [30,31], our work has mostly focused on epoxy-like photopolymer ‘MicroFine Green’ (ProtoLabs Inc., Maple Plain, MN) that can be printed by micro-stereolithography with sufficiently high resolution such that no separate gaskets are required for the vacuum seal. We have demonstrated that such centerpieces allow unhindered macromolecular sedimentation at a precision that is on par with commercial Epon epoxy centerpieces [31]. Various centerpiece designs can be created with OpenSCAD [42], a parametric scripting language for the creation of 3D objects from mathematically defined primitives. This flexibility, combined with quick turnaround of commercial 3D printing, opens new experimental possibilities in AUC. This includes centerpieces that require significantly lower sample volumes, provide short optical pathlengths for work at high concentrations, and/or provide longer observable solution columns for higher precision sedimentation velocity experiments [20,22,31].

3D printing has enabled a different strategy for increasing sample throughput, as it is possible to accommodate three sectors in the centerpieces instead of only two [31], as shown in Figure 1. Whereas previous designs featured sector angles of 1° and 1.5° to reduce sample volume, in the current work we have created sector angles of 2.2° to allow easier filling of the sectors, and to provide more tolerance for the adjustment of detection angles. The sectors have a dome-shaped ceiling [31] to raise the meniscus and extend the observable radial solution column lengths. As any other design, they can be printed with different optical pathlengths as needed.

Figure 1.

(A) Design of 3D printed three-channel centerpiece with sector angles of 2.2°, filling and venting holes, dome-shaped ceilings, and embossed rims surrounding the sectors that serve as integrated gaskets for vacuum seal. (B) Photographs of centerpieces 3D printed in an epoxy-like photopolymer “MicroFine Green” (ProtoLabs, Maple Plain, MN) using modification of a design file 3DPX-010758 of the NIH 3D Print Exchange (3dprint.nih.gov). Shown are centerpieces with 4 mm (left) and 12 mm pathlength (right), respectively. (C) Principle of illumination modes of three-channel centerpieces with Rayleigh interference optical detection. Two coherent parallel planar beams traverse either the left and the middle sector (left sketch), or the right and the middle sector (right sketch). The middle sector contains a joint reference buffer (blue), such that in the left configuration the purple sample can be observed, and in the right configuration the green sample can be observed. The two configurations have a 3-4° different rotation angle at the time of data acquisition, and due to the reversed beams they differ in the sign of the fringe shift signal.

While three sectors should, in principle, enable absorbance detection in transmitted intensity mode for three samples per cell, unfortunately this is not possible with the current commercial AUC operating software, which assumes the standard double-sector cell configuration in the synchronization of data acquisition with the rotation of the rotor. There is no obvious reason for this other than the custom of using double sector centerpieces, which has been hard-wired into the AUC operation software. Fortunately, however, the Rayleigh interference optical system in the most widely used ProteomeLab and XLA/I instruments allows the user to set the data acquisition angles. (In the most recently presented AUC model of this manufacturer this opportunity has been removed.) This detection system is based on phase differences of coherent light traversing either a sample or a reference solution, measured through shifts in their interference fringe pattern recorded in a camera [43]. Three sectors offer the possibility of measuring two samples if they share the same reference solution, as depicted in Figure 1C.

The two configurations can be selected through different rotor angles for data acquisition. In ProteomeLab and XLA/I instruments the rotor angles of data acquisition are determined by the user during the run setup [4]. They depend on the exact position of the magnet underneath the overspeed disc of the rotor, which serves as an index for the rotor angle. Optimal angles of sequential rotor positions differ by 90° or 45° for 4-hole or 8-hole rotors, respectively, and follow almost precisely α_n= α₀ + n×45 (or α_n= α₀ + n×90 for 4-hole rotors). When using the three-channel centerpieces this adjustment can proceed as usual, but must be followed by exploring adjustment of the alternate configuration approximately Δα = 3-4° distant from the first. The angular difference between the two configurations will be the same for all cells α_nA= α₀ + n×45 and α_nB= α₀ + Δα + n×45 (assuming 8-hole rotors, with n denoting the rotor hole numbers containing three-channel centerpieces), since it is determined solely by the geometry of the 3D printed centerpiece. Therefore, after initial adjustment, the two configurations can be recalled simply by (re)starting data acquisition after shifting the set of data acquisition angles alternately by +Δα and −Δα, with Δα being the initially defined and constant angle shift. Proof of principle of this data acquisition strategy was shown previously [31].

In this form, the run provides higher throughput, but is also much more labor intensive, since it requires periodic manual switching of the acquisition angles for the two samples in each cell. The motivation of the present communication is to alleviate this problem and to switch rotor angles automatically during the SV run. Even though this has not been envisioned in the AUC user interface, and the latter does currently not offer any support in this regard, it is possible to use general-purpose automation software to activate the necessary AUC user interface functions. To this end, we employed the freeware scriptable software AutoIt (version 3.3.14.5, AutoIt Consulting Ltd.; https://www.autoitscritp.com), which has been used previously in other scientific context to expand capabilities of commercial software and for laboratory automation [44-46]. AutoIt is a scripting language and scheduler for mouse and keyboard inputs, and can therefore be used effectively as meta-program using functions of commercial software as building blocks.

A sample script can be found in the Supplementary Material for an experiment with eight 3-channel cells in an 8-hole rotor. (It can be easily adapted to other rotor/cell configurations.) It consists of a timed sequence of mouse clicks at positions that correspond to menu items on the AUC user interface. Since they are specified in units of screen pixels, the positions must be recorded prior to the SV run for the specific computer screen and display resolution used. The script relies on the reproducibility of screen positioning of the AUC user interface program. Short wait times of 100 msec are included between clicks to allow execution of the invoked AUC user interface functions.

In more detail, the script will be activated after the start of the AUC run and after Rayleigh interferometric scans have been set to the angles α_nA. Execution of the script will bring the AUC operating program into the foreground and carry out a loop for changing scan settings, in this example 600 times, corresponding to running the script for approximately 16 hours. In each iteration of this loop, a first block defines the clicks needed to stop the current scan. (When the script is initiated, this step will be included but there is no method scan to stop; this will not affect the following steps.) Next are the specific clicks needed to adjust the laser settings for all eight cells, one at a time: 1) select Detail; 2) select Laser Setup; 3) clicking on the Laser delay slider, adjust the laser settings by Δα (in the present example 3°) if the loop number is even, or - Δα if it is odd; 4) after closing the details, move the slider of the cell selection down (in the present script using two clicks) to view the next cell to adjust. After all cells are adjusted, the last block moves the cell selection slider in the methods window up to the top and presses the Start Method Scan button. Finally, each iteration of the loop contains a pause (here for 85 sec) to allow data acquisition with the current settings before starting over with the adjustments.

In contrast to standard SV runs where the scan files are consecutively numbered and located in a single folder representing the start time of the scan set, after execution of the script the scan files have the same filenames but can be found in consecutive folders. They alternate between the two samples. Thus, we created a software program 3ChannelSorter that renames all scans (using different identifiers for the two samples), multiplies the scan data from the left sample with −1 to restore positive signals, and saves the scan files in a single folder for conventional data analysis. This software is available for download from our laboratory sharepoint site sedfitsedphat.nibib.nih.gov.

Figure 2A-D shows sedimentation boundary data recorded from left and right sectors that were filled with identical bovine serum albumin solutions. In contrast to the previous proof-of-principle data with sparse, uneven manual switching [31], the scans are now automatically acquired with high frequency for each sample sector, evenly spanning the entire SV experiment. A superposition of the sedimentation coefficient distributions c(s) is shown in Figure 2E. The standard deviation of the BSA monomer peak is 0.02 S, consistent with experimental error [26]. Slight variation of the BSA trimer peak s-value is observed, consistent with known limitations in the precision of measured sedimentation coefficient of trace components [18,47]. Neither the data acquisition mode nor the 3D printed centerpiece has any discernable impact on the data quality.

Figure 2.

(A-D) Sedimentation boundaries recorded in the left (A, C) and right (B, D) sample sectors of two three-channel centerpieces with 12 mm (A and B) and 4 mm (C and D) pathlength, respectively. (The 4 mm centerpiece was centered in height relative to the filling port, which allows measurement of gradients of up to ~100 fringes/cm [48].) Both sample sectors were filled with 1 mg/ml bovine serum albumin in phosphate buffered saline, with the buffer filled into the middle sector. SV experiments were carried out at 20° C and 50,000 rpm with 8 three-channel centerpieces in an 8-hole rotor following standard protocols [4,49] except for centerpieces and data acquisition script. Raw data in the left configuration (A, C) of Figure 1C are inverted. (E) Superposition of sedimentation coefficient distributions c(s) [6] calculated from the sedimentation boundaries. The s-value of the BSA monomer peak is 4.353 S (red), 4.393 S (blue), 4.402 S (cyan), 4.384 S (green).

In summary, in the present brief communication we show how one hurdle toward doubling the sample numbers in analytical ultracentrifugation using Rayleigh interference optics can be overcome. In practice, the remaining limitation is that sample loading is still laborious, and that we found three channel centerpieces in the current design not to seal as well as double sector centerpieces. However, the latter may be solved by slight changes in the embossed seals, or effectively addressed through the use of separate gaskets. The key point of the present work is to demonstrate how some restrictions in the use of AUC that are built into its current operating software may be easily lifted using freely available computer automation and simple scripting, thus motivating further developments of this technique. Exploiting similar strategy, user interface automation may be useful for the next generation of ultracentrifuges, and other computer-controlled laboratory instruments, to extend manufacturer envisaged instrument use.

• 3-channel centerpieces increase throughput in analytical ultracentrifugation

• The required new data acquisition mode can be implemented with automation software

• Computer user interface functions of instrumentation can be scripted and scheduled