Rapid screening methods for yeast sub‐metabolome analysis with a high‐resolution ion mobility quadrupole time‐of‐flight mass spectrometer

Rationale The wide chemical diversity and complex matrices inherent to metabolomics still pose a challenge to current analytical approaches for metabolite screening. Although dedicated front‐end separation techniques combined with high‐resolution mass spectrometry set the benchmark from an analytical point of view, the increasing number of samples and sample complexity demand for a compromise in terms of selectivity, sensitivity and high‐throughput analyses. Methods Prior to low‐field drift tube ion mobility (IM) separation and quadrupole time‐of‐flight mass spectrometry (QTOFMS) detection, rapid ultrahigh‐performance liquid chromatography separation was used for analysis of different concentration levels of dansylated metabolites present in a yeast cell extract. For identity confirmation of metabolites at the MS2 level, an alternating frame approach was chosen and two different strategies were tested: a data‐independent all‐ions acquisition and a quadrupole broad band isolation (Q‐BBI) directed by IM drift separation. Results For Q‐BBI analysis, the broad mass range isolation was successfully optimized in accordance with the distinctive drift time to m/z correlation of the dansyl derivatives. To guarantee comprehensive sampling, a broad mass isolation window of 70 Da was employed. Fragmentation was performed via collision‐induced dissociation, applying a collision energy ramp optimized for the dansyl derivatives. Both approaches were studied in terms of linear dynamic range and repeatability employing ethanolic extracts of Pichia pastoris spiked with 1 μM metabolite mixture. Example data obtained for histidine and glycine showed that drift time precision (<0.01 to 0.3% RSD, n = 5) compared very well with the data reported in an earlier IM‐TOFMS‐based study. Conclusions Chimeric mass spectra, inherent to data‐independent analysis approaches, are reduced when using a drift time directed Q‐BBI approach. Additionally, an improved linear dynamic working range was observed, representing, together with a rapid front‐end separation, a powerful approach for metabolite screening.


| INTRODUCTION
Studies involving metabolomics rely strongly on the use of mass spectrometry technology in order to both confirm metabolite identity and quantify the small molecules in complex matrices ideally with minimal user intervention. To this end, instrumental and software advances within the last decade have made high-resolution mass spectrometry (HRMS) an essential tool for supporting various metabolomics strategies including metabolite discovery and differential analysis. Nevertheless, obtaining a high-resolution mass spectrum for an unknown metabolite is known not to be sufficient to enable confirmation of metabolite identity since, even if the chemical sum formula determination according to accurate mass and isotope pattern matching is accurate, the number of potential structure matches remains too high. [1][2][3] Because of these major challenges faced for non-targeted metabolome assessment, employment of additional analytical selectivity via the use of high-resolution fragment (HR MS/MS) spectra can be used to support metabolite identification as a complementary and chemically informative descriptor. 4,5 In the case of HR MS/MS, data-dependent acquisition strategies rely on isolation and fragmentation of precursor ions which have exceeded a user-defined intensity threshold or any other measurable criteria such as isotopologue pattern, mass defect or the presence of a diagnostic ion. 6 However, some limitations for non-targeted assessment are also apparent. For example, some important precursor ions may not be selected by acquisition software for fragmentation particularly for closely eluting species with large differences in abundance, while difficulties in relative quantification (differential analysis) are often encountered. As an alternative, data-independent algorithms (DIAs) allow multi-event acquisitions to be carried out with fragmentation of all precursors regardless of abundance or spectral characteristics typically performed. 7 In the crudest implementation alternating between low-energy (0 V) and high-energy collision cell events, this acquisition workflow generates a complex HR MS/MS dataset making data mining and correct association of fragments with precursor ions extremely challenging. In fact, DIA strategies are only made feasible due to innovative developments such as shifting of the m/z isolation window of the quadrupole to effectively constrain the range of precursors undergoing fragmentation at a given time point.
Most successfully applied in the field of proteomics, strategies such as SWATH (sequential window acquisition of all theoretical spectra) 8 and MS E9 allow rapid collection of HR MS/MS spectra corresponding to one of the constrained windows of precursor ions. Importantly, these windows can be defined by the user according to both retention time and m/z considerations, which has proven extremely effective for improving proteome coverage despite challenges arising from chimeric MS/MS spectra.
With the same goal of providing a complementary and chemically informative descriptor for metabolites, the combination of gas phase ion mobility separation with HRMS (i.e. IM-HRMS) is now of emerging interest for non-targeted metabolomics studies. [10][11][12][13] Of particular importance in this area is the use of mobility-derived collision cross section (CCS) to support metabolite annotation, and also the possibility of harnessing the IM separation to support HR MS/MS workflows. Nesting of ion mobility separation prior to isolation and fragmentation (commercially available as IM-QTOFMS instrumentation) may therefore provide a path for development of advanced DIA strategies that use a combination of mobility separation and variable quadrupole isolation to provide higher quality MS/MS spectra. In the case of metabolomics, the potential of this marriage of analytical approaches is well recognized, particularly as high-quality CCS libraries are emerging, 14 but can also be practically limited by the high structural diversity, low molecular masses and CCS of many metabolites. Therefore, rather than aiming for a "full picture" metabolomics workflow, strategies involving targeting of a large subset of the metabolome such as the examples elaborated in chemical derivatization workflows developed by the Li group 15-18 might provide a better means to exploit the selectivity of novel IM-supported DIA approaches for metabolomics. In particular, differential 12 C-/ 13 C-isotope dansylation labeling of the metabolomic subset containing primary or secondary amines, or phenolic hydroxyl groups, has revealed more than 600 metabolites in human urine samples.
In this paper, we demonstrate that using advanced IM-QTOFMS instrumentation equipped with a prototype continuous band quadrupole driver, the transmission settings and/or collision energy applied to dansylated precursors can be successfully directed by IM separation (i.e. the quadrupole transmission can be programmed to suit the drift times of dansylated precursors; see Figure 1 for more information). In this way, transmission and fragmentation of dansylated metabolites can be driven strongly toward only signals arising from precursors of interest based on their more predictable conformational ordering in the m/z versus drift time space.
Moreover, employment of IM-driven quadrupole isolation was expected to allow the speed of the liquid chromatography (LC) separation of dansylated metabolites to be increased substantially. It is worth noting that initial work on this type of instrumental setup was performed in the field of proteomics using a similar prototype of an Agilent 6560 IM-QTOFMS. 19 A similar concept was also realized using other vendor instrumentation, namely Waters Vion IM QTofMS and its SONAR™ technology, and successfully tested in the field of drug metabolism and pharmacokinetics. 20,21 The present work provides a detailed performance comparison of a fast LC separation (<5 min cycle time) in combination with an IM-QTOFMS platform using two IM-DIA modes for the analysis of dansylated metabolites present in yeast cell extracts.

| Sample preparation and derivatization
An equimolar mixture of the aforementioned metabolites was prepared from a single standard solution and evaporated under reduced pressure using a GeneVac EZ2 solvent evaporation system to a final amount of  Precursors isolated by the combination of IM-Q-BBI are then subject to collision energy, allowing collection of cleaner fragment spectra using TOFMS volume was mixed with 100 μL of 0.5 M Na 2 CO 3 /NaHCO 3 buffer (pH 9.50) in order to maintain an alkaline pH of 9.5. After vortexing, 100 μL of dansylation reagent (20 mg mL −1 dansyl chloride in 90:10 acetonitrile:acetone solution) was added and vortexed again. A 2 mL amber glass vial was used as reaction vessel and in the following the mixture was incubated for 1 h at 60°C in an agitator (250 rpm). The derivatization reagent was prepared freshly, whereas the buffer solution was kept at 4°C and pH was always controlled before usage.
To remove the excess of non-reacted derivatization reagent, an aliquot of 250 μL was transferred into another 2 mL amber glass vial, where approximately 25-30 mg of PSA (primary secondary amine) SPE bulk material was weighed in beforehand, and was incubated for another 15 min at 60°C in an agitator (250 rpm). Finally, aliquots of 200 μL were carefully transferred into HPLC vials with inserts and were stored for a maximum of 24 h at 6°C in a cooled tray until analysis. The mass spectrometer was mass-calibrated before every measurement using the Agilent ESI Tune solution. This tune mix was also employed for single field CCS calibration as described previously. 23

| Optimization of collision energy ramps
To optimize the collision energy ramps, the equation typically used (CE = (slope × precursor m/z)/100 + offset) was varied in its values for slope and offset (1, 3, 5 and 5, 10, 15, 20, respectively). If the precursor ion was present at approximately 10% of the base peak, a condition was considered as optimal. Decisions on optimal parameters were made via visual inspection of MS2 spectra of eight dansylated compounds (1DNS-glycine, 1DNS-serine, 1DNS-uracil, 1DNS-proline, 1DNS-aspartate, 1DNS-adenine, 1DNS-phenylalanine and 2DNS-glutamine) using MassHunter Qualitative Analysis (B08.00, Agilent Technologies). An offset of 15 and a slope value of 1 yielded overall acceptable fragmentation patterns and these settings were thus chosen for all acquisition modes to calculate the collision energy ramp.

| Optimization of drift time directed Q-BBI
As pointed out previously, DT-IM and MS are not considered as orthogonal techniques, though they will deliver complementary information, since there is a clear correlation between mass-to-charge ratio and drift times of particular molecular classes in low-field drift time IM instrumentation. When dealing with research areas where analytes are characterized by similar building blocks, e.g. lipidomics, the relationship between mobility and mass-to-charge ratio can   Table 1.    23 Regarding the peak area, a good repeatability for both precursor and fragment ions was achieved. However, it is worth mentioning that for glycine a clear difference in peak area repeatability between precursor and fragment ions was observed.
This can be explained by the rather low peak area of the fragment ion and hence ion counting statistics.

| DISCUSSION
The presented approach aims for a significant enhancement of analytical selectivity in non-targeted analysis by combining dansylation of metabolites and rapid UHPLC with prototype Q-BBI directed by drift-tube IM mass spectrometry. Apart from the selectivity enhancement by derivatization, one of our working hypotheses relies on the chimeric spectra obtained when using an AI approach, whereas, when employing a BBI using this prototype quadrupole, the analytes of interest can be selected before the collision cell leading to less overlap with matrix constituents. As a consequence, this also leads to cleaner fragmentation spectra. This can be clearly observed in the fragment spectra obtained by the two approaches (see Figure S3, supporting information). A minor advantage can be also seen in data storage, since data files using the IM-Q-BBI approach are roughly 40% smaller compared to the IM-AI approach In terms of sensitivity, the precursor ion counts achieved with the two approaches were similar; however, the IM-AI mode revealed a factor of 2-4 higher sensitivity for the fragment ions.
Transmission of fragment ions will be improved in future work by optimizing the m/z versus drift time ramp for quadrupole isolation in a high-energy step.
In Figure 3