Introduction

Small molecule drug discovery projects often rely heavily on the application of biophysical techniques to characterize protein–ligand interactions [1]. It is frequently the case that multiple methods are required to adequately profile the interaction for any single target of interest, with many of these experiments yielding complementary data. Hydrogen/deuterium exchange coupled with mass spectrometry (HDX-MS) [2,3,4] is one such biophysical method that provides an information rich fingerprint of protein–ligand interactions [5]. We have applied the method extensively to profile many different classes of ligands, including fragments [6], elaborated small molecules [7,8,9,10,11,12], and macrocyles [13].

To define the HDX behavior of a protein of interest, measures of deuterium content for the protein are made at a number of pre-determined time points. The number of time points included in the experimental design, along with the span of the experimental time window, will dictate how accurately the rate of exchange can be measured. Following incubation in the D2O buffer for the appropriate time period, each exchange reaction is quenched with reduced temperature and pH, and the protein is digested with acid stable protease prior to LC-MS. The deuterium incorporation is measured for discrete regions of the protein by following the HDX at the peptide level in the mass spectrometer. Therefore, the number and size of the proteolytic peptides measured will define the spatial (amino acid) resolution of the data. The HDX-MS experimental design is therefore a compromise between (a) the number of time points and technical replicates required to accurately cover the experimental time window, and (b) the amount of instrument time, reagents, and data processing effort required to complete the experiment.

Our current experimental workflow is a pairwise comparison between two samples. This differential experiment includes six periods of exchange (n = 3) for each sample, plus a minimum and maximum exchange control, yielding 40 individual HDX-MS measurements. For small proteins (< 50 kDa), it is not uncommon to have upwards of 100 peptides in the experiment. Therefore, we often require the calculation of over 4000 data points to complete the analysis. To facilitate this time-consuming data processing step, we have developed software specifically for the analysis of differential HDX-MS experiments [14,15,16].

In order to reduce the time burden on the experimentalist, and to minimize random and gross errors associated with manual sample preparation, we developed a fully automated platform for HDX-MS experiments [17]. An updated dual-column version of this platform was subjected recently to a stringent two-site validation study [18]. We demonstrated for the first time that two remotely located HDX-MS platforms could generate equivalent protein-ligand interaction HDX data across multi-day studies [18]. When combined with our software, we have a created a robust, well-validated, HDX platform that is widely utilized across our small molecule discovery projects. Many laboratories, and a number of commercial vendors, have subsequently implemented similar automation systems [19].

As discussed above, our current automation platform is well suited for differential HDX-MS experiments. However, we considered that a move away from an integrated design, where samples are prepared and delivered to the MS system in real time (serial preparation and injection), to a decoupled configuration made up of two separate pieces of automation (parallel preparation, serial injection) would offer a number of benefits. The four main advantages identified were (1) the ability to prepare large numbers of samples in parallel using plate-based liquid handling and disposable labware. (2) A reduction in time to prepare samples. (3) Improved flexibility for both preparation and analysis experimental designs. (4) A three-fold increase in the utilization of the mass spectrometer.

In this work, we describe a unique automation platform that is comprised of two novel pieces of automation. Combined, this new HDX platform offers a number of advantages over existing integrated automation platforms, while maintaining the quality of the data generated.

Sample Preparation System

The sample preparation module (Figure 1a) is based around a Freedom EVO 150 (TECAN, Männedorf, Switzerland). Temperature regulation is provided by two Minichillers (Hubler, Offenburg, Germany) and a cryostat (Julabo USA Inc., Allentown, PA). Microplates for reagent storage, the on-exchange reaction and the low pH quench, are placed on carriers to keep them under temperature control during the experiment. Final samples (50 μL) are frozen directly into individual sample vials (Figure 1b) cooled to < − 70 °C within a custom plate holder. The TECAN scheduling algorithm enables the system to prepare up to 24 HDX samples in parallel across the three 96-well plates (each of which can be prepared with up to 12 time points of on-exchange). Samples are typically prepared with the following nine exchange times: Dmin, 10 s, 30 s, 90 s, 270 s, 810 s, 2430 s, 7290 s, and Dmax. Incubations are overlapped, such that a complete set of 24 samples with nine time points can be prepared in under 3 h.

Figure 1
figure 1

The decoupled automation HDX-MS platform. (a) The TECAN sample preparation system capable of preparing up to 288 on-exchange samples in a single experiment. (b) Following on-exchange samples are quenched, mixed, and then frozen within individual sample vials arrayed in a 96-well SBS format. (c) The sample injection system is comprised of a UR3 robot, a LEAP autosampler, and a HPLC system interfaced with a Q-Exactive mass spectrometer (Thermo Electron, Bremen, Germany)

Full size image

Sample Injection System

The injection module (Figure 1c) is built around a UR3 Robot arm (Universal Robots, Odense S, Denmark) programed to present samples in series to a CTC HTS-PAL autosampler (LEAP Technologies, Durham, NC) located in a refrigerated chromatography cabinet. During analysis, samples are kept frozen to <− 70 °C in a pair of custom cooled racks with capacity for 192 individual samples (Figure 2a). For long-term storage, samples are held in a – 80 °C freezer. For each injection, the UR3 robot picks up a single matrix tube from a rack and places it into the shuttle which is then guided inside the refrigerated cabinet held at 4 °C. Inside the cabinet, the sample vial is accessible to the CTC PAL autosampler (Figure 2b). To initiate the thawing of the sample, 50 μL of cooled quench solution is then added to the vial, followed by a 2-min hold. Once thawed, the sample (100 μL) is aspirated from the sample vial and injected into the HPLC-MS instrument configured for online enzyme digestion (Figure 2c). The shuttle is then removed from the chromatography cabinet and the sample vial is discarded by the robot. Injection of the next sample begins in parallel with the HPLC elution of the preceding sample, in order to minimize the injection to injection cycle time. Employing a 5-min gradient elution method, the 192 samples can be injected within 24 h.

Figure 2
figure 2

Detail views of the sample injection system. (a) The UR3 robot arm is configured within reach of a custom plate rack capable of holding two 96-well format plates at <- 70 °C (black boxes, gray covers on the left side of the UR3). Condensation is mitigated by addition of a pair of independently actuated plate covers that are heated to 25 °C. The plate cover is only raised for removal of a sample vial. The compressed air actuated shuttle presents the sample vial into the chromatography cabinet. (b) Once docked, the vial is accessible to the CTC-PAL autosampler. (c) Online digestion and LC-MS are performed with a standard HDX-MS configuration (LC valves, pepsin column and heater shown). Trap column and LC column are out of this picture

Full size image

Results and Discussion

Having designed and constructed the new decoupled platform for HDX-MS, we undertook a series of experiments to compare the performance of the new system against data generated with our previous system. Using vitamin D receptor (VDR) ± 25-OH-VD3 as a model system, we concluded that the data were broadly similar. As expected, there were minimal differences in measurement precision, or back exchange between the systems (see the Supplementary Material).

In order to evaluate the performance of the system over time, and between scientists, we recently performed a small study with a model protein (cytochrome C) (Figure 3a). Over a period of nine months, we measured the HDX profile for cytochrome C six times (each experiment contained nine time points that were performed in triplicate). As detailed in Figure 3b and c, we did not observe any large changes in the data obtained between run days, or between operators.

Figure 3
figure 3

Performance evaluation for the decoupled automation platform. Panel (a) provides a summary of the experimental design. Over a 9-month period of time, six HDX experiments were performed by three scientists. Each scientist measured the HDX behavior of a model protein, cytochrome C. As shown in (b) and (c), the data obtained were largely identical and demonstrate the suitability of the system for performing HDX-MS studies over a period of many months, which is common within our laboratory

Full size image

Since the completion of the system in 2017, we have completed over 400 differential HDX-MS experiments with the integrated platform with minimal downtime. Therefore, we have demonstrated the robustness of the platform over an extended period of time.

Data Analysis

Our current integrated HDX-MS automation was designed and optimized to perform pairwise comparisons between a test and reference sample. Therefore, our internal software for data analysis (HDX Workbench [16]) was optimized to process data acquired in a pairwise fashion. The decoupled automation platform described here allows us to prepare up to 24 samples in parallel. Therefore, we are no longer limited to pairwise comparisons. Accordingly, we have updated HDX Workbench to accommodate a multiple comparison experimental design (see the Supplementary Material).

Conclusions

We have designed and implemented a new decoupled automation platform for HDX-MS experiments. The performance of the system with respect to back exchange and measurement precision is essentially equivalent to our previous systems. However, the predicted increase in flexibility of experimental design and improved capacity have been realized. The system has been demonstrated to be robust and reliable, having generated over 400 differential datasets within the last 30 months.