Generation of a Collision Cross Section Library for Multi-Dimensional Plant Metabolomics Using UHPLC-Trapped Ion Mobility-MS/MS

The utility of metabolomics is well documented; however, its full scientific promise has not yet been realized due to multiple technical challenges. These grand challenges include accurate chemical identification of all observable metabolites and the limiting depth-of-coverage of current metabolomics methods. Here, we report a combinatorial solution to aid in both grand challenges using UHPLC-trapped ion mobility spectrometry coupled to tandem mass spectrometry (UHPLC-TIMS-TOF-MS). TIMS offers additional depth-of-coverage through increased peak capacities realized with the multi-dimensional UHPLC-TIMS separations. Metabolite identification confidence is simultaneously enhanced by incorporating orthogonal collision cross section (CCS) data matching. To facilitate metabolite identifications, we created a CCS library of 146 plant natural products. This library was generated using TIMS with N2 drift gas to record the TIMSCCSN2 of plant natural products with a high degree of reproducibility; i.e., average RSD = 0.10%. The robustness of TIMSCCSN2 data matching was tested using authentic standards spiked into complex plant extracts, and the precision of CCS measurements were determined to be independent of matrix affects. The utility of the UHPLC-TIMS-TOF-MS/MS in metabolomics was then demonstrated using extracts from the model legume Medicago truncatula and metabolites were confidently identified based on retention time, accurate mass, molecular formula, and CCS.


Introduction
Metabolomics is a revolutionary systems biology tool for understanding plant metabolism and elucidating gene function [1][2][3]. The continual refinement, increasing scope and larger-scale have led to the evolution of modern plant metabolomics, which has matured as a valuable tool for advancing our understanding of plant biology and physiology [2,[4][5][6]. Although the vast utility of metabolomics is well documented in the literature, its full scientific promise has not yet been realized due to multiple technical challenges. These challenges have been reviewed by the metabolomics community and a set of grand challenges identified by the Plant, Algae, and Microbial Metabolomics Research Coordination Network [7]. At the top of these grand challenges are accurate and confident chemical identification of all observable metabolites and increasing the metabolome depth of coverage. Here, we propose a multi-dimensional, combinatorial solution for both using UHPLC-trapped ion mobility coupled to mass spectrometry (UHPLC-TIMS-TOF-MS). phase compounds [26]. Mobility is inversely related to CCS, and a small mobility represents a large CCS value. Therefore, mobility data are often presented as an inverse reduced mobility (1/K 0 ) [15]. CCS is a physical property of a compound measured using a specific drift gas, temperature and pressure. Thus, the ion mobility community has recently published recommendations for reporting ion mobility metadata [27] to help ensure that the many variables that influence K 0 and CCS measurements are reported. Trapped ion mobility spectrometry (TIMS; Figure 1) circumvents some of the previous challenges of IMS and offers higher resolutions of approximately 200 with much smaller instrumental size relative to drift tube IMS [28,29]. CCS values can also be acquired with very high reproducibility using TIMS. Thus, we have opted to use TIMS in our instrumental metabolomics approach. temperature and pressure. Thus, the ion mobility community has recently published recommendations for reporting ion mobility metadata [27] to help ensure that the many variables that influence K0 and CCS measurements are reported. Trapped ion mobility spectrometry (TIMS; Figure 1) circumvents some of the previous challenges of IMS and offers higher resolutions of approximately 200 with much smaller instrumental size relative to drift tube IMS [28,29]. CCS values can also be acquired with very high reproducibility using TIMS. Thus, we have opted to use TIMS in our instrumental metabolomics approach.
Here, we report the creation of a CCS library for improved identification of compounds in plant metabolomics by adding another dimension of chemical parameter matching. We used UHPLC-TIMS-QTOF-MS/MS in negative electrospray ionization mode to calculate CCS values for inclusion in the library. A substantial number of plant natural products include phenolic and carboxy substituents that ionize very well in the negative ESI, which also has the advantage of lower chemical noise that ultimately results in better MS S/N ratios for these compounds. Data generated were highly reproducible, and we describe additional methods to accurately annotate ions observed in the ion mobility spectrum. Reported CCS measurements were robust during coelution and also in the presence of a complex plant matrix. The utility of the CCS library for the identification of plant natural products based upon accurate mass, Rt, and CCS was demonstrated using an extract from the model legume Medicago truncatula and data evaluation using Bruker MetaboScape software. Figure 1. Trapped ion mobility spectrometry (TIMS) is temporally well aligned with UHPLC as a second dimensional separation technology. TIMS also offers higher resolution and a high duty cycle. Ions from the source are pushed through the ion tunnel by a drift gas and opposed by an electric field gradient, in effect trapping them where those forces are equal on the ion. The two stages of the TIMS Figure 1. Trapped ion mobility spectrometry (TIMS) is temporally well aligned with UHPLC as a second dimensional separation technology. TIMS also offers higher resolution and a high duty cycle. Ions from the source are pushed through the ion tunnel by a drift gas and opposed by an electric field gradient, in effect trapping them where those forces are equal on the ion. The two stages of the TIMS allow for accumulation of ions in analyzer 1 while separation and mobility measurements take place in analyzer 2. This results in a near 100% duty cycle. As the electric field gradient is lowered, the ions are released from the second analyzer towards the time-of-flight (TOF) detector.
Here, we report the creation of a CCS library for improved identification of compounds in plant metabolomics by adding another dimension of chemical parameter matching. We used UHPLC-TIMS-QTOF-MS/MS in negative electrospray ionization mode to calculate CCS values for inclusion in the library. A substantial number of plant natural products include phenolic and carboxy substituents that ionize very well in the negative ESI, which also has the advantage of lower chemical noise that ultimately results in better MS S/N ratios for these compounds. Data generated were highly reproducible, and we describe additional methods to accurately annotate ions observed in the ion mobility spectrum. Reported CCS measurements were robust during coelution and also in the presence of a complex plant matrix. The utility of the CCS library for the identification of plant natural products based upon accurate mass, R t , and CCS was demonstrated using an extract from the model legume Medicago truncatula and data evaluation using Bruker MetaboScape software.

CCS Library
A Bruker timsTOF Pro™ and concurrent gas composed of 99.4% N 2 and 0.6% O 2 from a Parker Balston Model N2-80A membrane based nitrogen generation system were used to record TIMS CCS N2 values for 146 authentic plant natural products mostly consisting of flavonoids, glycosylated flavonoids, isoflavonoids, triterpenes, and glycosylated triterpenes. The major ion species observed were [M-H] − and the data for this group were as follows. The masses of authentic standards sampled ranged from 210 to 826 Da. CCS measurements ranged from 141. 26 Tables S1 and S2 (Excel File) and CCS values plotted in Figure S1.

Statistics and Reproducibility
The CCS values, average, standard deviation, and relative standard deviation (RSD) were recorded for triplicate analyses of each authentic standard and are summarized in Figure 2. The [M-H] − was the primary ion of interest. The minimum standard devation was 0.0058 Å 2 , the average was 0.1988 Å 2 (±0.1469 Å 2 ), the maximum was 0.7401 Å 2 for all the triplicate analyses. The minimum RSD was 0.0037%, the average RSD was 0.1075% (±0.0720%), and the maximum RSD was 0.3459%, for all triplicate analyses of [M-H] − ions. Major ions are all within the strictest range of highly reproducible measurements recently published by Stow et al. [29]. The high reproducibility of the reported data illustrates the precision of the acquired CCS data and their use for compound identification.
Calibration of the TIMS using a primary standard is a necessary and crucial step for accurate CCS determinations [27,29], and we typically use the Agilent ESI Low-concentration Tune Mix. Calibration was performed initially using CCS values provided within the Bruker software. However, a list of slightly different CCS values for the Agilent ESI Low-concentration Tune Mix compounds was recently published by the ion mobility community [29]. Thus, multiple CCS values exist in the literature for the primary tune mix standards, and the community has recognized the need to extract a consensus value [27], but this has not yet happened. Thus, we calculated CCS values for our standards using both primary calibration lists and report these in Tables S1 and S2 (Excel File) until a more complete community vetting of the calibration values can be completed. The average relative standard deviation of triplicate measurements for all measured authentic standards was 0.1026%. H] -was the primary ion of interest. The minimum standard devation was 0.0058 Å 2 , the average was 0.1988 Å 2 (±0.1469 Å 2 ), the maximum was 0.7401 Å 2 for all the triplicate analyses. The minimum RSD was 0.0037%, the average RSD was 0.1075% (±0.0720%), and the maximum RSD was 0.3459%, for all triplicate analyses of [M-H] -ions. Major ions are all within the strictest range of highly reproducible measurements recently published by Stow et al. [29]. The high reproducibility of the reported data illustrates the precision of the acquired CCS data and their use for compound identification.

Comparison of CCS Values with Published Data
We compared our CCS values to a previously published library from Pacific Northwest National Laboratory (PNNL) using DTIMS [29,30]. There were 12 common entries: four fatty acids and nine phenolic acids. The fatty acid measurements had a maximum difference of 0.71%, but interestingly the fatty acids were all uniformly measured here to have a smaller CCS value compared to the PNNL database [30]. The average and relative standard deviation difference was 0.53% (±0.16%). The reported phenolic acid CCS values from PNNL were all larger with the average percent difference of 2.10% (±0.74%). The higher resolution of TIMS relative to DTIMS should result in greater precision in the CCS measurements, while the accuracy of the TIMS determination would depend on accurate calibration.

Isomer Comparisons
Isomer differentiation is a primary motivation for inclusion of IMS in our multi-dimensional metabolomics platform as it is often not achievable by UHPLC-MS alone. Comparisons of multiple groups of isomers were performed after the creation of the CCS library. The largest set of isomers analyzed was the monohydroxyflavones which included seven different isomers. These compounds were used to compare isobaric structural changes and positional variance of the hydroxyl group on CCS values. Triplicate CCS measurements were performed in a broad survey mode using 0.60-1.50 1/K 0 mobility range and 90 ms ramp time that resulted in an approximate TIMS resolution of 40. Higher TIMS resolutions are easily obtainable, and higher resolution experiments were performed used 0.68-0.76 1/K 0 mobility range and 400 ms ramp time that yielded an approximate resolution of 120 for the monohydroxyflavones. The CCS apex of the peaks were recorded, and their RSDs were determined to range from 0.09% to 0.26%.
Resolution is a measure of the separation of two components within a mixture and defined for ion mobility in Equation (2): where ∆K 0 is the difference in ion mobilities, W K0 is the average peak width in mobility units, ∆CCS is the difference in collisional cross sections and W CCS is the average peak width in CCS units. A resolution of 1 represents baseline separation at 4σ (i.e., peak width 4 times the standard deviation or ± 2σ and confidence level of 95.4%), and a resolution of 0.5 represents resolution at approximately half height or 2σ. Thus, a matrix was established, and resolution values were calculated for all possible combinatorial separations of all monohydroxyflavone pairs ( Figure 3) obtained at TIMS resolutions of 40 and 117. The calculated resolution values indicate that 10 of the 21 combinatorial monohydroxyflavone pairs (almost 60%) can be resolved at half height using broad range experimental TIMS parameters (0.60-1.50 1/K 0 ) inclusive of a rapid ramp time of 90 ms and a modest resolution of approximately 40. These parameters produce lower resolution but higher sampling rate. However, increasing the ramp time (400 ms) and narrowing the mobility scan range (0.68-0.76 1/K 0 ) increases resolution for differentiating isobaric monohydroxyflavones. All but two, i.e., 19 of 21 or 90%, of the combinatorial monohydroxy possibilities could be separated using a TIMS resolution of ≥ 117. These results emphasize the added value of TIMS in differentiating isobaric and isomeric compounds and the results would be expected to be similar or better for other larger polyhydroxy isomeric compound classes.
The relationship between CCS and m/z is well established as illustrated in Figure S2 for fatty acids. However, small changes in the spatial structure can often have significant impacts on CCS. For example, the tetrahydroxyflavones differ only by a double bond and two hydrogens in the central ring structure relative to tetrahydroxyflavanones, yet the tetrahydroxyflavanones (average CCS of 162.94 Å 2 ) on average have a CCS 2.01 Å 2 larger than the tetrahydroxyflavones (average CCS of 160.95 Å 2 ). See Figure 4. This is likely due to the planar orientation of the flavone C ring attributed to the SP2 hybridization of the C2 atom associated with the double bond. The C ring can then be in-plane with both the A and B rings reducing its overall spatial geometry with the constraining conjugated bonds between C2 and C3 with the attached B ring benzene. In contrast, the orientation of a flavanone, which has an equatorial attachment of the B ring to C2 and no constraining conjugated bonds, allows for free movement and rotation of the B ring causing an increase in average spatial conformation.
sampling rate. However, increasing the ramp time (400 ms) and narrowing the mobility scan range (0.68-0.76 1/K0) increases resolution for differentiating isobaric monohydroxyflavones. All but two, i.e., 19 of 21 or 90%, of the combinatorial monohydroxy possibilities could be separated using a TIMS resolution of ≥ 117. These results emphasize the added value of TIMS in differentiating isobaric and isomeric compounds and the results would be expected to be similar or better for other larger polyhydroxy isomeric compound classes.  SP2 hybridization of the C2 atom associated with the double bond. The C ring can then be in-plane with both the A and B rings reducing its overall spatial geometry with the constraining conjugated bonds between C2 and C3 with the attached B ring benzene. In contrast, the orientation of a flavanone, which has an equatorial attachment of the B ring to C2 and no constraining conjugated bonds, allows for free movement and rotation of the B ring causing an increase in average spatial conformation.

Ion Mobility Spectra Interpretations
Extracted ion mobility (EIM) spectra for the expected [M-H] -ions and base peak ion mobility (BPM) spectra containing all adducts and aggregates were smoothed. Mobility peak apices were used to calculate the reported average CCS values. The BPM spectra were often complex and contained several adducts and aggregates making the assignment of the [M-H] -challenging or at times

Ion Mobility Spectra Interpretations
Extracted ion mobility (EIM) spectra for the expected [M-H] − ions and base peak ion mobility (BPM) spectra containing all adducts and aggregates were smoothed. Mobility peak apices were used to calculate the reported average CCS values. The BPM spectra were often complex and contained several adducts and aggregates making the assignment of the [M-H] − challenging or at times impossible. Thus, CCS values are reported only for annotated peaks of high confidence where the m/z and isotopic pattern matched that of the annotated adduct. Many of these adducts were also found in the mass spectra of a related library [23]. Some annotation challenges could be overcome using EIM spectra or a new TIMS fragmentation method using an increased TIMS ∆6 voltage described in Section 2.3. Relying on the most intense mobility peak was not always found to be a reliable method for ion annotation.
A series of rutin ion mobility spectra provided in Figure 5 illustrate the explainable complexity of its BPM spectrum. The top BPM spectrum shows more than a single peak for rutin. The assignment of the correct structure to each peak and related CCS is not obvious. The oxidized ion formation has been previously described [23]. While adducts can be reduced, they cannot be avoided.
reliable method for ion annotation. A series of rutin ion mobility spectra provided in Figure 5 illustrate the explainable complexity of its BPM spectrum. The top BPM spectrum shows more than a single peak for rutin. The assignment of the correct structure to each peak and related CCS is not obvious. The expected [M-H] -with a m/z of 609 has a clear high intensity peak. A sodium formate adduct ion ([M-H+Na+HCO2 -] -) at 677 m/z and an oxidized ion [M-3H] -at 607 m/z were also observed. The oxidized ion formation has been previously described [23]. While adducts can be reduced, they cannot be avoided.

In-TIMS Ion Manipulation
The TIMS tunnel in the timsTOF Pro TM instrument is a two-stage device that includes ion accumulation followed by TIMS analysis. Bruker OtofControl software enables users to optimize the concentration of the ions in the analyzer, sensitivity, duty cycle, and/or resolution dependent on user defined experimental goals and applications. When ions are transferred from the accumulator to the analyzer portion of the TIMS tunnel, the ion velocity can be controlled by adjusting the transfer voltage parameter (TIMS Δ6). Increasing the TIMS Δ6 parameter results in increased ion velocities and increased collisional energies when ions collide with the drift gas. Moderate increases in the TIMS Δ6 can result in 'in-TIMS' fragmentation and can often dissociate adduct and aggregate ions to simplify the BPM while even higher energies can begin to fragment the molecular ions.
Several examples are provided that illustrate the utility of using the TIMS Δ6 parameter in spectral interpretation. These include rutin (Figures 6 and S3), naringin (Figure 7), chrysin (Figure

In-TIMS Ion Manipulation
The TIMS tunnel in the timsTOF Pro TM instrument is a two-stage device that includes ion accumulation followed by TIMS analysis. Bruker OtofControl software enables users to optimize the concentration of the ions in the analyzer, sensitivity, duty cycle, and/or resolution dependent on user defined experimental goals and applications. When ions are transferred from the accumulator to the analyzer portion of the TIMS tunnel, the ion velocity can be controlled by adjusting the transfer voltage parameter (TIMS ∆6). Increasing the TIMS ∆6 parameter results in increased ion velocities and increased collisional energies when ions collide with the drift gas. Moderate increases in the TIMS ∆6 can result in 'in-TIMS' fragmentation and can often dissociate adduct and aggregate ions to simplify the BPM while even higher energies can begin to fragment the molecular ions.
Several examples are provided that illustrate the utility of using the TIMS ∆6 parameter in spectral interpretation. These include rutin ( Figure 6 and Figure S3), naringin (Figure 7), chrysin ( Figure S4), and naringenin ( Figure S5). Rutin adducts are reduced with higher voltage transfers, and if too high, the compound fragments. The CCS of naringin increased without any m/z change when the TIMS ∆6 voltage exceeded the default setting of −100 V and going from −125 through −200 V, suggesting the opening of the central ring structure to form naringin-chalcone which increases the size of the analyzed ion in the TIMS drift region. Chrysin did not form as many adducts. Thus, high voltages do not influence the ion mobility spectra until enough energy is applied to fragment the molecule. In summary, the TIMS ∆6 voltage can be adjusted to help identify observed ion mobility spectral peaks.
high, the compound fragments. The CCS of naringin increased without any m/z change when the TIMS Δ6 voltage exceeded the default setting of −100 V and going from −125 through −200 V, suggesting the opening of the central ring structure to form naringin-chalcone which increases the size of the analyzed ion in the TIMS drift region. Chrysin did not form as many adducts. Thus, high voltages do not influence the ion mobility spectra until enough energy is applied to fragment the molecule. In summary, the TIMS Δ6 voltage can be adjusted to help identify observed ion mobility spectral peaks.  . Peak 9, shown in the green trace, is an ion that emerged with greater TIMS Δ6 voltage and was not found initially during the analyses. The new peak (peak 9) has been putatively annotated as naringin chalcone. A chain breakage of the naringin likely occurred with the higher energies from the TIMS Δ6 voltage resulting in opening of the C ring and elongated naringin to naringin chalcone. The −120 ion is likely the further fragmentation of the chalcone chain breakage forming a separate mobility peak 5.

CCS Matching with Authentic Standards
We analyzed a mixture of authentic standards by direct infusion TIMS-MS and UHPLC-TIMS-MS to evaluate the robustness of the CCS measurements and matching. The mixture included rutin, naringin, naringenin, chrysin, and 6-hydroxyflavone. These compounds include a diverse mixture of hydroxyl, methoxy, and glycosylated flavones and these are commonly used in our lab as a quality control mixture to monitor UHPLC and MS performance. We used a TIMS ion current control (ICC) of 2-5 Mio to prevent potential space charge effects that could alter the measured CCS during the TIMS experiments. The resultant CCS data are included in Table 1 and the compounds separated within the same analysis. The difference in the direct infusion measured CCS values and the library CCS values had an average percent difference of 0.16557% for all authentic compounds.  . Peak 9, shown in the green trace, is an ion that emerged with greater TIMS ∆6 voltage and was not found initially during the analyses. The new peak (peak 9) has been putatively annotated as naringin chalcone. A chain breakage of the naringin likely occurred with the higher energies from the TIMS ∆6 voltage resulting in opening of the C ring and elongated naringin to naringin chalcone. The −120 ion is likely the further fragmentation of the chalcone chain breakage forming a separate mobility peak 5.

CCS Matching with Authentic Standards
We analyzed a mixture of authentic standards by direct infusion TIMS-MS and UHPLC-TIMS-MS to evaluate the robustness of the CCS measurements and matching. The mixture included rutin, naringin, naringenin, chrysin, and 6-hydroxyflavone. These compounds include a diverse mixture of hydroxyl, methoxy, and glycosylated flavones and these are commonly used in our lab as a quality control mixture to monitor UHPLC and MS performance. We used a TIMS ion current control (ICC) of 2-5 Mio to prevent potential space charge effects that could alter the measured CCS during the TIMS experiments. The resultant CCS data are included in Table 1 and the compounds separated within the same analysis. The difference in the direct infusion measured CCS values and the library CCS values had an average percent difference of 0.16557% for all authentic compounds. Authentic standards were further spiked into a plant root extract and analyzed by UHPLC-TIMS-MS to test the robustness of the CCS measurements and matching within a representative metabolomics sample and to determine if significant matrix affects would alter the CCS measurements. The results were compared to a sample of the same authentic standards without plant matrix. The results in Table 2 reveal an even lower difference then for the direct infusion experiments, and demonstrate the robust accuracy of the measurements. The matrix had a negligible effect. The accuracy was maintained.

CCS Value Matching with Plant Extracts
To further demonstrate the utility and application of our CCS library we analyzed and processed plant root extracts. Medicago truncatula roots were extracted with a mixture of 80% methanol and 20% water and analyzed using UHPLC-TIMS-QTOF-MS. Analyzed extracts of M. truncatula roots were processed using MetaboScape. Spectral matching of the root data to library data was performed using exact mass, isotopic patterns, MS/MS, and CCS values. The matching scores are easily visualized with an annotation quality (AQ) graphic reported within the MetaboScape software. A total of 151 compounds in the M. truncatula extract were matched with the plant compound library [24] based on accurate mass, and 40 of the 151 were also matched with CCS values. The additional orthogonal CCS data matching provides increased confidence in the identification based upon the Metabolomics Standards Initiative criteria [21]. The data are provided in Table S3 and the processing results are illustrated in Figure 8.
CCS matching in the above experiment was performed using multiple thresholds; i.e., CCS matching score with a moderate confidence matching of ∆CCS ≤ 2.0% and a high confidence matching ≤0.3%. This is within the range of other IMS literature reports of <2% ∆CCS [26]. It is possible that new calibration methods or calibrant compounds could improve the reproducibility even further. These matching scores are based on the apex of the mobility peaks and do not reflect the resolving power of the peaks. We used ion charge control (ICC) to prevent oversaturation of the compounds during the mobility analysis which reduces the accumulation time, and therefore the ion concentration, when ions are detected beyond an intensity threshold. Figure 8. A representative matching report using Bruker MetaboScape software. The annotation quality (AQ) column shows a color coded matching confidence of each of the parameters used in the search (from left to right): accurate mass, retention time, isotopic pattern (mSigma), MS/MS fragmentation, and CCS. Each column has a set of customizable matching parameters. If the first matching criteria is met the square is grey, if the narrower criteria is met both squares are green. The CCS matching is a new orthogonal data feature that expands the confidence in metabolite identifications. The bottom square alone that is grey represents a matching of less than 5% variation between the library and the measured value. Both squares green represents a matching less than 2% variation between the library and the measured value.

Chemicals
Authentic standards were acquired from Sigma-Aldrich (  Each column has a set of customizable matching parameters. If the first matching criteria is met the square is grey, if the narrower criteria is met both squares are green. The CCS matching is a new orthogonal data feature that expands the confidence in metabolite identifications. The bottom square alone that is grey represents a matching of less than 5% variation between the library and the measured value. Both squares green represents a matching less than 2% variation between the library and the measured value.

Chemicals
Authentic standards were acquired from Sigma-Aldrich ( . The aqueous solvent used for UHPLC contained 0.055-0.1% formic acid. Initially 0.1% formic acid was used, but it was lowered to 0.05% to reduce the presence of adducts.

UHPLC
Authentic standards were analyzed in triplicate on different days. Authentic standards were suspended in 80% methanol at concentrations ranging from 10-50 ppm and placed in autosampler vials with 300 µL reduced volume inserts. Two microliters of authentic standards were injected onto a Waters UPLC system using a BEH C18 column, 2.1 × 150 mm, 1.7 µm particles that was operated at 60 • C. The flowrate was 0.560 mL/min and gradient elution performed. A linear gradient was performed using binary solvents (solvent A composed of 0.1-0.05% formic acid in water and solvent B composed of 100% acetonitrile). The gradient started at an initial ratio of 95:5% (A:B) and was linearly ramped to 30:70% over 30 min. The gradient was then ramped linearly to 5:95% in 0.1 min (from 36.00 min to 36.10 min) and held at 5:95% for 4 min. Afterwards, the gradient was immediately returned to 95:5 and allowed to re-equilibrate for 4 min. During this re-equilibration time a 20 µL of Agilent ESI-L low concentration tune mix was introduced into the system via a sample loop and 6-way valve for internal calibration.

ESI-TIMS-QTOF-MS
The Waters UHPLC was connected to a Bruker timsTOF Pro™ instrument equipped with trapped ion mobility coupled to a hybrid quadrupole, time-of-flight mass spectrometer; i.e., UHPLC-ESI-TIMS-QTOF-MS. Ions were generated in the negative electrospray ionization mode. The ESI source used 10 L/min of drying gas at a temp of 250 • C. The drying gas also served as the concurrent TIMS gas. The drying gas was generated by a Parker Balston Model N2-80A membrane based nitrogen generation system operated at an exit pressure of 100 p.s.i. and measured to be composed of 99.4% nitrogen and 0.6% oxygen. The ESI also used a 3500 V capillary voltage, and a 3.0 bar nebulizer pressure. Following the recommended reporting for ion mobility mass spectrometry measurements [27] all user controllable TIMS parameters are reported here. The TIMS cartridge tunnel in pressure was set to 2.50 mbar and tunnel out pressure was 0.74 mbar. The TIMS ion funnels used a fixed radiofrequency of 840 kHz with 275-350 Vpp amplitude depending on the analytes mass. A higher RF amplitude was used for larger compounds and optimized using Agilent tune mix. For maximum sensitivity, compounds were analyzed in different groups based on their m/z value: 200-300 m/z used 275 Vpp, 300-600 m/z used 300 Vpp, and 600+ m/z used 350 Vpp. For routine sample analyses a single RF amplitude is sufficient. Our metabolomics analyses used a funnel 1 RF of 250 Vpp. TIMS inverse reduced mobility (1/K 0 ) data were collected over a range of 0.4-1.8 1/K 0 . To prevent TIMS tunnel saturation and space charge effects during TIMS analyses, the ion charge control (ICC) software option was used to adjust the accumulation time of the ions in the first segment of the TIMS, and the ICC set between 2-5 million (Mio). Mass spectral data were collected on the QTOF-MS using a scan range of 100-1500 m/z.

Calibration
Ion mobility and m/z measurements were both externally calibrated using Agilent ESI-L Low Concentration Tune Mix before batch analyses of samples. Batch analyses times were at most 48 h and after this time, the instrument was recalibrated. In addition, a 20 µL aliquot of Tune Mix diluted 1:30 was injected at the end of each analysis and used for internal sample calibration during data processing. The mass and ion mobility were calibrated with calibrants and analyte values interpolated. At least three calibration points were used for a linear TIMS calibration fit. The instrument was externally calibrated to the Bruker default CCS values and the data were internally recalibrated with calibrant injected at the end of each sample analysis. CCS values are reported using both the Bruker default and more recent community calibration values which were slightly different (Tables S1 and S2) [29].

Plant Extract Preparation
Plant extracts were prepared from Medicago truncatula Jemalong A17 root. The tissues were harvested in liquid nitrogen, lyophilized, and ground into a fine powder using a mortar and pestle. Dry tissues were stored at −80 • C until extracted. 10 mg of M. truncatula dried root tissue was extracted with 1 mL of a mixture of 80% methanol and 20% water. The tissues were vortexed for 2 min, sonicated for 5 min, centrifuged at 4000× g for 40 min, and the supernatant collected.

Data Processing and CCS Determination
Data were processed using a pre-release version of Bruker's MetaboScape 5.0 and DataAnalysis 5.2 software. Mobility and m/z were calibrated using at least 3 calibration points from the Agilent Tune Mix (302, 602, 1034 m/z). UHPLC peaks for the authentic standards were visually selected from their base peak chromatograms (BPC). The expected m/z corresponding to the [M-H] − ion was then used to generate an extracted ion mobility (EIM) spectrum. The ion mobility spectrum was smoothed using a single cycle Savitzky Golay algorithm, 0.005 [V*s/cm 2 ] (10 points). Ion mobility spectral peak picking parameters used a sensitivity of 97%, relative intensity threshold of 15%, and minimum peak valley of 10% to generate a table of observed ion mobilities. The software calculated the CCS value from the maximum m/z in the ion mobility spectrum. These peaks were then investigated for the [M-H] − and the peak containing the highest "purity" or a single ion species based on the mass spectra of the peaks mobility range was selected as the true [M-H] − and the CCS value was calculated. CCS data for three analyses per compound were transferred to Microsoft Office Excel where the statistical averages, standard deviations, and relative standard deviations were calculated and recorded for each of 146 compounds detected (Tables S1 and S2).

TIMS ∆6 Adduct Investigations
Adduct formation was investigated using different transfer settings within the TIMS cartridge and more specifically the TIMS ∆6 voltage. The range in the TIMS ∆6 voltage started at relatively low voltage differences (100 V) and was gradually increased (up to 200 V) during direct infusion experiments of single compounds. Segments of the acquisition stepped through increasing voltage differences resulting in increased transfer speeds and collision energies with the drift gas. Data were processed using DataAnalysis visualizing BPM and EIM of the [M-H] − to determine relative intensity changes between different voltage differences of each segments.

Conclusions
We report here a CCS library focused on plant natural products and specialized metabolites that contains 146 compounds, 343 CCS values, and 29 isomers annotated (various hydroxylated and/or methoxylated and/or glycosylated flavonoids and isoflavonoids). This library is expected to improve metabolite identification confidence and compound identification annotation quality scores through matching of additional orthogonal CCS values included in the library. We also report a method to simplify the spectral complexity of the ion mobility spectra whereby putative adducts and aggregates are dissociated using an increased TIMS ∆6 voltage and MS analyses to verify structural changes. We also found that there were no significant matrix affects observed (less than the standard deviation of the measurements) on the measured CCS as illustrated with a group of standards analyzed by direct infusion and UHPLC-TIMS and spiked within a plant extract. TIMS provides robust, reproducible CCS data with isolated compounds, complex compound mixtures, and compounds within a biological matrix. UHPLC-TIMS-MS/MS provides a four-dimensional metabolomics platform which improves metabolomics' depth-of-coverage and metabolite identification confidence.  [29] for the Agilent Tune Mix, Table S3. UHPLC-TIMS-MS/MS profiling results of M. truncatula root extracts in which compound identifications were based upon Rt, exact mass, isotopic patterns MS/MS and TIMS CCS N2. , Figure S1. Comparison of mass, molecular weight (MW, x-axis) and average CCS (y-axis) of various compounds, Figure S2. Fatty acids m/z compared to average CCS, Figure S3. Rutin ion mobility spectra and MS spectral changes associated with increasing collision energies, Figure S4. Chrysin and naringenin do not have altered ion mobility spectra or fragmentation with the increased TIMS ∆6 voltages.