The Q Exactive HF, a Benchtop Mass Spectrometer with a Pre-filter, High-performance Quadrupole and an Ultra-high-field Orbitrap Analyzer*

The quadrupole Orbitrap mass spectrometer (Q Exactive) made a powerful proteomics instrument available in a benchtop format. It significantly boosted the number of proteins analyzable per hour and has now evolved into a proteomics analysis workhorse for many laboratories. Here we describe the Q Exactive Plus and Q Exactive HF mass spectrometers, which feature several innovations in comparison to the original Q Exactive instrument. A low-resolution pre-filter has been implemented within the injection flatapole, preventing unwanted ions from entering deep into the system, and thereby increasing its robustness. A new segmented quadrupole, with higher fidelity of isolation efficiency over a wide range of isolation windows, provides an almost 2-fold improvement of transmission at narrow isolation widths. Additionally, the Q Exactive HF has a compact Orbitrap analyzer, leading to higher field strength and almost doubling the resolution at the same transient times. With its very fast isolation and fragmentation capabilities, the instrument achieves overall cycle times of 1 s for a top 15 to 20 higher energy collisional dissociation method. We demonstrate the identification of 5000 proteins in standard 90-min gradients of tryptic digests of mammalian cell lysate, an increase of over 40% for detected peptides and over 20% for detected proteins. Additionally, we tested the instrument on peptide phosphorylation enriched samples, for which an improvement of up to 60% class I sites was observed.

Suppl. Figure 11 20140116_EXQ00_RiSc_SA_IGORPH_02_lowres Mass spectrometry related settings used for the instrument comparison. We varied the values for parameters relevant for the best performance of each instrument. These include: resolution (due to the compromise of slightly lower resolution for the same transient time), maximum injection time (due to the faster transient), NCE (due to a different HCD energy scaling function) and peptide match (due to the instrument not achieving full topN).
Performance of the Q Exactive HF on a single cycle with identification rate of over 70% at a speed of 17 Hz. Top panel represents the full scan preceding the fragmentation scans, where the isotope used for each fragmentation scan is marked with an arrow and number indicating its position in the cycle. In the inset the isotopes are ranked on intensity with the location of the sequenced ones marked, showing the instrument is sequencing over an order of magnitude within a single cycle. The annotated fragmentation spectra in this cycle are presented below. The first two sequencing events are of lower intensity, which can be attributed to the high sequencing speed and the software catching up with the most up-to-date information. The pie-charts indicate the Andromeda Score associated to the identification on log2 scale and the Precursor Ion Fraction (PIF; also isolation purity) of the fragmentation spectrum.

Supplementary Figure 2
Normalized Collision Energy (NCE) titration series. Data collected with the standard settings (Xcalibur was configured to exclude charge-state unassigned, 1, and 6 and up), while varying the NCE, on a HeLa whole cell lysate (see Supplementary Table 2 for further description). The range was chosen based on experience with the Q Exactive, where the optimal value was found to be 25 (data not shown), and the expectation a higher value was required due to a change in the NCE to CE scaling function. A. Optimal overall performance, within a small variation, was found at NCE level of 27 (setting used for further experiments). B. Success-rate within charge-states (0 denotes a single peak for which no charge state could be determined). Even though both unassigned and charge-state 1 are excluded, these are still sequenced as the control software has to decide the charge state based on not yet complete data. The identification success-rate of charge-state 5 is generally low, while charge-states 2, 3, and 4 have over 60% success-rate. C. Higher charge-states are excluded, given the minimal contribution from chargestate 5 to the overall identifications (here at NCE 27), potentially creating a better focus for the mass spectrometer.

Ion Injection Time (IIT) titration series.
Data collected with the standard settings, while varying the maximum allowed IIT, on a HeLa whole cell lysate (see Supplementary Table 2 for further description). The range was chosen to encompass the speed of the Orbitrap ultra-high field analyzer running at 15k resolution for the MS2 scans. A. With longer maximum IIT's the instrument is achieving less MS2 scans (bars in the graph), as the MS2 scan fill time does not fit in the transient time of the previous MS2 scan (i.e. parallel mode). Below 25 ms there is only a slight increase in performed MS2 scans. With the longer IIT's the sequencing success-rate is however increasing (line in the graph). B. That the success-rate is increasing can be attributed to the longer allowed IIT, causing less MS2 scans actually to reach the maximum IIT value (i.e. max-out). C. This ensures that the full ion population of 1e5 is reached in more cases. D. The optimal IIT for the MS2 scans is however found around 25 ms (plus the reported median 6 to 14 ms scan time overhead makes this value in a lot of cases fully parallel with the previous MS2 scan). Even with the almost 70% max-out rate and consequently lowered sequencing success-rate at this IIT, the parallel operation ensures that many more MS2 scans can be performed making up for the difference.
The lowered performance at 20 ms is expected, as at 25 ms most of the MS2 scans were already in parallel mode and the further lowered sequencing success rate is not making up for the extra MS2 scans being performed.

Supplementary Figure 4
S-lens RF-level titration series. Data collected with the standard settings, while varying the S-lens RFlevel, on a HeLa whole cell lysate (see Supplementary Table 2 for further description). The chosen value for the S-lens RF-level affects the transmission through the S-lens. A. The increase in transmission for the lower m/z region at higher S-lens RF-levels has the effect that more isotopes will pass the minimal ion threshold required, resulting in more fragmentation scans being performed. B. The increased ion transmission also results in lower IIT's for MS2 scans (only MS2 scans presented not reaching the maximum IIT). C. We find optimal performance at the S-lens RF-level of 60, where most unique peptide sequences are reported. Of concern with the S-lens RF-levels is that source fragmentation is more likely to occur at higher values, resulting in isotope patterns for non-tryptic peptides that cannot be identified. This most probably happens for S-lens RF-level 65 and 70, where fewer unique peptide sequences are reported. At the used S-lens levels we found however that any fragmentation results in hard to detect levels, as an unspecific search of the fragmentation spectra did not yield any non-tryptic peptides.

Supplementary Figure 5
Underfil ratio titration series. Data collected with the standard settings, while varying the underfil ratio, on a HeLa whole cell lysate (see Supplementary Table 2 for further description). The chosen value for the underfil ratio affects the ion threshold for selecting an isotope for fragmentation. Data for underfil ratio 80 was excluded due to inconsistent behavior. A. The higher underfil ratios cause less MS2 scans to be recorded, however the identification rate increases. B. We find that the best sequencing success is at an underfil ratio of 40. Given the low differences as a result of differing underfil ratios, the method was set up with an underil ratio of 20 to maximize the number of MS2 scans being performed. The low difference between the best and worst setting can be explained from the high complexity of the sample, providing many highly abundant candidates at any given time during the gradient.

Supplementary Figure 7
Peak-depth calculations. The peak-depth is calculated by first calculating the number of visible isotopes for each cycle. After sorting this list by descending intensity, the position of the isotope associated to each of the MS2 scans in the cycle is determined. A. The number of visible isotopes with charge state 2 and up at any given point in the gradient fluctuates between 400 and 500. This value is slightly elevated for the Q Exactive with the normal Orbitrap analyzer. A likely explanation for this is the marginally reduced resolution for the Orbitrap ultra-high field analyzer running in high speed mode (from 70k to 60k), resulting in a slight decrease of detectable isotope patterns. B. The high speed of the Orbitrap ultra-high field analyzer ensures that a higher percentage of available isotopes are sequenced.
the suggested design (which is sparse in the sense that not every combination is required to be measured) we ran 17 one hour gradients with the standard HeLa described in the manuscript. We utilized a combination of Quadratic Modeling and Multiple Linear Regression to uncover and quantify any pairwise interactions. The output of the analysis is expressed as coefficients, where error-bars (standard error) smaller than the coefficient is better and can be visualized in a surface plot making it human readable. For the coefficients, a single term shows the linear and quadratic effect (e.g. iso and iso*iso, respectively) which that parameter has on the performance; two separate terms multiplied is the pairwise interaction between those terms. Model fitting was performed on the data processed with and without the MaxQuant option 'second peptide', which attempts to identify a second isotope pattern lying in the isolation window for already identified MS2 scans. A. We found that by far the largest effect (and the only statistically significant one as indicated by the '*') is induced by the NCE value, which controls the energy applied during fragmentation. When this energy is too high, the peptide is fragmented into many smaller, internal fragments which cannot be meaningfully interpreted. Although the other factors are not significant within the 95% confidence interval, the trend of the coefficients is as expected. The next two largest effects are induced by S-Lens RF-level and isolation window size (both quadratic effects). We find a single pair-wise interaction for isolation window and NCE, which seems to be stronger for the 'second peptide' results. This is as expected because a wider isolation window will lead to more second peptides in the fragmentation spectrum. Co-fragmentation of peptides could require a different energy to result in the most information rich fragmentation spectrum for all peptides involved. B. Surface plot visualizations of the results when processing the data without 'second peptide'. On the right-hand side the 4D surface plot is displayed, from which it can be read out that the S-Lens RF-level of 60 supplies the best performance. The data imply that for the interacting parameters NCE and isolation window an inverse relationship holds (i.e. larger isolation windows require lower fragmentation energy). To verify whether this effect is significant the parameter range for the NCE would need to be decreased. The black dot in this 4D surface plot indicates the previously individually optimized values for the chosen parameters (NCE=27, S-Lens=60, isolation window=1.4), which falls inside the optimal area. The 3D surface plot on the left-hand side for the S-Lens RF-level of 60 is provided as a reading guide for the 4D surface plot. C. Surface plot visualizations of the results when processing the data with the MaxQuant option 'second peptide'. While the overall impact of the parameters is similar, the performance appears to benefit from a slightly wider isolation window.
Based on the above results we conclude that the previously found values (NCE=27, S-Lens=60, isolation window=1.4) are optimal, as they fall within the optimal area for both with as well as without 'second peptide' indicated by the design of experiment evaluation. As there is a clear fit to the previously independently optimized values and we detected only a rather weak set of interacting parameters we conclude that significantly increased performance cannot be achieved by varying this sub-set of parameters. However, we note that due to its ability to handle sparse data, the DOE approach can be highly efficient for experimental parameter optimization for different sample and chromatography types, given that large ranges can be investigated without accumulating many mass spectrometry measurements.