Comparability of Steroid Collision Cross Sections Using Three Different IM-HRMS Technologies: An Interplatform Study

Steroids play key roles in various biological processes and are characterized by many isomeric variants, which makes their unambiguous identification challenging. Ion mobility-mass spectrometry (IM-MS) has been proposed as a suitable platform for this application, particularly using collision cross section (CCS) databases obtained from different commercial IM-MS instruments. CCS is seen as an ideal additional identification parameter for steroids as long-term repeatability and interlaboratory reproducibility of this measurand are excellent and matrix effects are negligible. While excellent results were demonstrated for individual IM-MS technologies, a systematic comparison of CCS derived from all major commercial IM-MS technologies has not been performed. To address this gap, a comprehensive interlaboratory comparison of 142 CCS values derived from drift tube (DTIM-MS), traveling wave (TWIM-MS), and trapped ion mobility (TIM-MS) platforms using a set of 87 steroids was undertaken. Besides delivering three instrument-specific CCS databases, systematic comparisons revealed excellent interlaboratory performance for 95% of the ions with CCS biases within ±1% for TIM-MS and within ±2% for TWIM-MS with respect to DTIM-MS values. However, a small fraction of ions (<1.5%) showed larger biases of up to 7% indicating that differences in the ion conformation sampled on different instrument types need to be further investigated. Systematic differences between CCS derived from different IM-MS analyzers and implications on the applicability for nontargeted analysis are critically discussed. To the best of our knowledge, this is the most comprehensive interlaboratory study comparing CCS from three different IM-MS technologies for analysis of steroids and small molecules in general.


External Calibration for IM-MS Drift tube ion mobility mass spectrometry (DTIM-MS)
In DTIMS, ion mobility (K, m 2 s -1 V -1 ) depends on the steady state drift velocity (vd, ms -1 ) and the strength of the applied electric field. In the case of a linear drift tube of length (L, m), an ion's mobility (Eq. 1) can be calculated from the drift time (td, s) and is often reported as a reduced mobility K0 (Eq. 2) by taking into account normalized pressure and temperature with N representing the gas number density and N0 the Loschmidt constant, p0 and T0 are NIST standard temperature and pressure, respectively. 1 In practice, the external calibration of DTIM-MS is achieved using an infusion of ions with known CCS to construct an external calibration curve with slope β and intercept tfix at a single field strength (Eq. 5). Here, β and tfix depend on applied conditions (e.g., drift gas, E, field strength, ion optics), mg represents mass of the buffer gas and mi the mass of the analyte ion 3 .
DTIMS or DTIM-MS operating as a primary method is used to establish reference values for external calibration. Excellent interlaboratory reproducibility (0.34%) and low interlaboratory bias (0.54%) compared to a reference instrument have been reported for DTIM-MS 3 .

Traveling wave ion mobility-mass spectrometry (TWIM-MS)
TWIM transmission and separation relies on application of non-uniform fields and travelling potential waves which drag ions through the drift gas [4][5][6] . Typically, such TWIM-MS platforms provide IM resolving powers below ~40-60 depending on the instrument generation 6 . TWIM separation is a complex process in which ions of lower mobility can be overtaken by a potential wave resulting in a back-and-forth movement along the separation path. Due to the complex ion movement and related non-linear contributions to ion mobility, TW CCSN2 relies exclusively on external calibration which is (analogous to DTIM-MS; see Eq. 5) corrected by charge state and reduced mass to yield a modified CCS (CCS', Eq. 6). 6-8 Experimentally determined arrival times are corrected to td ' to account for non-IM-related contributions using a correction via the instrument dependent coefficient C and the ions' m/z. 6 Calibration coefficients A and B are then determined by fitting of logarithmic plot (Eq. 8) and TW CCSN2 of unknown ions is calculated using a power function (Eq. 9). 6,7 With TWIM, ions can experience high electric field strengths resulting in ion heating during ion injection and TWIM separation. This in turn leads to increasing vibrational energies and subsequent changes in ion conformation including isomerization or fragmentation 9 . The interlaboratory comparability for different types of TWIM-MS instruments is reported to be better than ±1.5% for the majority of investigated compounds in two interlaboratory comparisons 10,11 . However, this is only true when the same external calibration method and calibration ions are used. In contrast to DTIM-MS, the choice of calibrant ions is reported to have an impact impact in TWIM-MS calibration 7 and different calibration strategies including various calibrant ion mixtures have been reported 5,7,12 . Recent progress in development of universal calibration approaches aim at replacement of the currently used empirical power law function, but solving equations for TWIM ion motion is necessary for this purpose 12 . However, such calibration procedures are not currently used on a routine basis and a detailed discussion is out of scope for the present work.

Trapped ion mobility-mass spectrometry (TIM-MS)
In contrast to time-dispersive IM technologies, TIM-MS involves the use of a gas flow and a counteracting electrical field allowing accumulation (trapping) and selective release of ions.
Ions are dragged through the TIM device using a gas flow, while an electromagnetic field gradient (EFG) is used to hold ions in a stationary position along with radio frequency (RF) potentials applied to achieve for radial confinement 13 . The dragging force of the gas flow with a defined gas velocity vgas and the applied EFG counteract each other and ions are trapped along the EFG depending on their mobility 14  Vm is the applied voltage at time of release from TIM and a and b are empirical constants 14 : With this approach, DT CCSN2 of Agilent tune mix ions were reproduced with excellent agreement (95 th percentile ~1%). However, a systematic error can be observed when high TIM ramp rates are used as the constants a and b depend on the operation settings, but the exact relationship remains unknown. Alternatively, a calibration based on solving the Boltzmann transport equation was recently reported 14 . However, this is not currently applied on a routine basis and detailed discussion is out of scope of the present work.

Data acquisition
Chromatographic separation was performed using an Acquity UPLC BEH C18 column (2.1 mm × 100 mm, 1.7 μm particle size, Waters) equipped with an in-line filter kit (0.2 μm, Waters).  Table S1. Additional measurements for assessment of long-term stability of DT CCSN2 were made using a Phenomenex Kinetex C18 column (4.6 mm × 50 mm, 2.6 µm particle size) due to unavailability of the initially used column.
For DTIM-MS measurements, the following ESI settings were used on an Agilent 6560 IM-QTOFMS (Agilent Technologies, Santa Clara, CA) equipped with a Dual AJS ESI Ion Source: gas temperature was set to 225 °C and a gas flow of 8 L/min was used. Nebulizer was operated at 30 psi, sheath gas temperature and flow were set to 350°C and 12 L/min, the capillary voltage was 3500 V, and the nozzle voltage was set to 500 V. DT CCSN2 was determined using stepped-field and single-field methods in both polarities. The mass range was set to 50-1700 m/z and measurements were performed using the 2 GHz Extended Dynamic Range setting.
This instrument platform can be used with maximum IM resolving powers of ~70 19 and was achieved a resolving power of 50-60 in the present work.

S7
For direct infusion measurements, a syringe pump was used with flow rates of 1.8 mL/h and 1.2 mL/h for analyzing standard mixtures using stepped-field and single-field methods, respectively.
The DTIM-MS was operated with drift gas pressure of 3.95 Torr and a temperature of 299.5 K.
For stepped-field acquisition, the total runtime was 3. To increase the apparent resolving power of DTIM-MS data, HRdm 2.0 software was additionally used to process the 4-bit multiplexed datafiles. Interpolated files obtained from PNNL pre-processor were used as input files and data was processed using following settings: HR processing level was set to "low", m/z width multiplier was set to 6 and IF multiplier setting of 1 was used resulting in an IM resolving power of ~110 for boldenone undecylenate [M+H] + .
The following settings were used for feature finding (peak picking) in Mass Profiler 10.0: abundance was measured as maximum ion volume and charge state was limited to 1-2. For alignment and normalization, the retention time tolerance was set to ± 0.3 min, drift time tolerance was ±1%, the acceptable mass tolerance was set to ± (15.0 ppm + 2.0 mDa) and all S9 features with a Q-score ≤ 70.0 were rejected. Feature results were exported as .csv files and were further used for data analysis using Microsoft Excel and R.
For stepped-field data evaluation, the fourth time segment was selected in IM-MS Browser 10.0, and frames were summed. The isotopic envelope of a target ion (at least two ions were required) was selected and "Calculate CCS (Multi field)" was used. Result tables were exported as .csv files and further statistical analysis was performed.
All TIM-MS files were analyzed using Bruker TASQ (2021) software. After automatic recalibration, features were detected (peak picking) based on accurate mass and retention time. TIM CCSN2 was determined automatically in the TASQ software after automated integration of high-resolution EICs using a trace width of ±3 mDa and a retention time window of 0.5 min for peak detection. Results were averaged across three replicates. Ion quality was determined based on the principal ion only. The classic chromatogram peak finder mode was used, and ion signals were determined based on the signal height using 250 counts as minimum peak area and 50 counts as minimum intensity threshold. LC peaks were smoothed and de-noised using a Gaussian peak model and the minimum signal-to-noise ratio was 3 along with minimum moving average filter with a box width of 3, the minimum peak height was set to 3000 counts and minimum required peak width was set to 5 datapoints. The retention time used for peak picking in the IM-dimension was set to 0.2 min and an alignment tolerance of 0.01 Vs·cm -2 was used for TIM-MS data. The same settings were used for TWIM-MS data, but 0.02 ms was used as mobility tolerance. IM spectra were extracted as .csv from single files and data was replotted in Excel.

Additional Data Analysis and Visualization Data analysis and visualization
In addition to Microsoft Excel, R (4.1.2) 21 language together with RStudio (2021.9.1.372) 22 was used for creating graphs and analyzing data. The following packages were used for this purpose: dplyr 23 , ggplot2 24 , ggally 25 , ggpubr 26 and ggbreak 27 . R base and dplyr were used to re-arrange data tables. Functions from R base were used for basic data analysis (e.g., creating linear models). Ggplot2, ggally and ggpubr were used to create raw plots, (e.g., correlation matrix, violin plots or residual plots and ggbreak was used to insert axis breaks into violin plots).

Analytical performance of the individual IM-MS platforms
Precision under repeatability conditions for measurement of DT CCSN2 was excellent for triplicate measurements. An average RSD of 0.14%±0.12% was obtained using the stepped- steroids and 6 ions of steroid phase II metabolites that were measured in these previous studies were also analyzed in our work. The average absolute difference for single field-calibrated DT CCSN2 was 0.23% for positive mode data and 0.33% for deprotonated ions of phase II steroid metabolites (max. = 0.66%) illustrating the excellent interlaboratory agreement using DTIM-MS instruments of the same class (see Table S2).   Figure S4 shows the absolute bias of secondary calibrated IM-MS methods against steppedfield DT CCSN2 as reference:

Precision under repeatability conditions for measurement of
The corresponding bias for each ion sorted by ion species and CCS' is shown in Figure S3.
Absolute bias was smallest between single-field calibrated DTIM-MS recorded on the same instrument with an average absolute bias of 0.31±0.22% and a narrow bias distribution (95 th percentile 0.67%). This was followed by TIM-MS, which had a higher average absolute bias of 0.71±0.80% and 95 th percentile of 1.22%. A similar average absolute bias was observed for single laboratory and interlaboratory TW CCSN2 with average absolute bias of 0.70±0.87% and 0.81±0.93%, respectively. 95 th percentiles were 1.64% and 2.01%, respectively. Bias for each ion sorted by ion species and CCS' is shown in Figure S5. Figure S6 illustrates the absolute bias of secondary TIM-MS and TWIM-MS using single field DT CCSN2 as reference values. Absolute bias was smallest for TIM-MS with an absolute bias of 0.47±0.70% and 95% of the values within 1.03%. The average absolute bias observed for single laboratory and interlaboratory TW CCSN2 were 0.82±0.76% and 0.68±0.79% with 95 th percentiles being 1.92% and 1.87%, respectively.

Comparison to published DT CCSN2 data
Comparisons of DT CCSN2 data determined on the same class of DTIM-MS instrument in two recent studies revealed good agreement for a small set of both steroids and steroid phase II metabolites. A small systematic bias (≤0.5%) for [M-H]of 6 steroid phase II metabolites in one study 28