Characterization of Gas-Phase Native(-Like) Proteins Using Structures for Lossless Ion Manipulations

Abstract

High resolution mobility-based ion separations in Structures for Lossless Ion Manipulations (SLIM) have been useful for ion mobility separations for a variety of molecular classes in the gas phase. Here, we present multi-pass SLIM separations for gas-phase proteins in their near-native state exhibiting charge state dependent arrival time distributions using carbonic anhydrase (29 kDa), alcohol dehydrogenase (148 kDa), and apo-transferrin (79 kDa). The experimental CCS values were obtained from calibration curves for the arrival times of Agilent Tune Mix ions. For multi-pass separations, the ATDs were converted to CCS values by deconvoluting the multi-pass arrival times into accurate single-pass values amenable to the single-pass calibration curves. Mass spectra of carbonic anhydrase (CA) showed three different charge states (z = 9+ to 11+). Their corresponding mobility peaks were baseline-separated using 8-m single-pass separations. When compared to corresponding drift tube ion mobility (DTIMS) measurements, the CCS values obtained from DTIMS and SLIM were in agreement within experimental error. Single-pass analysis of alcohol dehydrogenase (ADH) exhibits three predominant charge states (z = 23+ to 25+) with mobility overlap between adjacent charge states. The mobility peak resolution for ADH improved with multi-pass separations (up to 24-m path length). In addition, CCS distributions obtained for charge states z = 16+ to 18+ of apo-transferrin reveal a transition from a compact unimodal form (z = 18+ and 19+) to broader multi-modal CCS distributions for z = 16+. For apo-transferrin, 40-m multi-pass separations were performed allowing for complete isolation of the selected mobility range corresponding to z = 17+ leading to selective isolation of a narrow arrival time window. The extended mobility separations provided minimal alterations to the structure of the proteins, and the experimentally derived CCS values showed minimal change as a function of separation time or number of passes. Mobility-based ion separations for native-like proteins, using SLIM, open opportunities for native-IMS applications as well as other manipulations enabled by SLIM like mobility selective isolation and collection.

Graphical Abstract

Introduction:

In structural biology, understanding the intricate details of protein conformations and interactions is crucial for elucidating their biological functions. Native mass spectrometry (MS) has emerged as a powerful technique to study proteins in their near-native state. Improved ability to characterize higher-order structures^{1, 2} and complex formation^{3, 4} has benefited from the development of high-resolution mass analyzers (i.e., Fourier-transform ion cyclotron resonance, orbitrap) with extended m/z ranges. More recently, native MS has been complemented by ion mobility spectrometry (IMS) allowing for elucidation of functional roles of protein assemblies and their subunits^{5, 6} and dynamic conformational changes.^{7, 8}

Ion mobility spectrometry (IMS), in combination with MS, enables separation of ions based on their mass, charge, and shape, thereby providing additional structural information characterized by collision cross section (CCS) measurements. For instance, a traveling wave (TW) based IMS was used to study the effect of charge reduction (via electron transfer) on protein compaction and rearrangement in order to uncover charge-state dependent conformational flexibility.⁷ In another example, the effects of buffer and desolvation efficiency were investigated providing guidance for CCS calibration and measurements for native-like proteins with a broad range of mobility.⁹ Moreover, advancements in IMS technology have enabled high resolution mobility separations and high precision CCS measurements showing potential for separation and isolation of molecules in targeted conformational space from protein standards or protein complexes in native-like states.

Among these, Structures for Lossless Ion Manipulations (SLIM) technology offers high-resolution mobility-based separations afforded by extended ion path lengths ^{10, 11} and exceptional control over ion motion ^{12, 13} with high ion utilization efficiency.¹⁴ Prior work also targeted native-MS separations using constant field SLIM devices.¹⁵ In this work, we demonstrate the use of the traveling wave SLIM platform capable of high resolution multi-pass separations ^{10, 16} to determine conformational distributions of three native-like proteins: carbonic anhydrase (CA, MW: ~29 kDa), alcohol dehydrogenase (ADH, MW: ~148 kDa), and apo-transferrin (MW: ~79 kDa). SLIM base-line separations for three charge states of carbonic anhydrase indicate distinct structural differences without overlap in conformational space. ADH and apo-transferrin with greater molecular weight required multi-pass separations with extended path lengths (~24 m) to resolve slight conformational differences between adjacent charge states. For complex molecules such proteins charge state-based mobility separations allow for improved structural understanding of the select charge state population.^{7, 17} In addition, a mobility range of interest was selected for multi-pass separations allowing for complete isolation of ions in the chosen CCS range. Numerical simulations were performed to calculate the CCS values for the selected charge states to show agreement with experiment within ~6% error. The multi-pass SLIM separations applied for analysis of native-like proteins show promise for selective collection of molecules in targeted conformational space for providing insights into high-order structures when complemented with multimodal molecular characterization techniques (i.e., simulations, high-resolution imaging).

Methods

Protein preparation:

Soluble proteins, bovine carbonic anhydrase (C2624), alcohol dehydrogenase from baker’s yeast (A7011-15KU), and human apo-transferrin (T2036) were purchased from Sigma-Aldrich and used without further purification. Ammonium acetate and buffer components for the reconstitution of proteins were also purchased from Sigma-Aldrich. Lyophilized powders of bovine carbonic anhydrase, alcohol dehydrogenase, and apo-transferrin were resuspended in 200 mM ammonium acetate (pH 6.9).

All proteins were desalted by passing through P6 buffer exchange columns (7326221, Bio-Rad), equilibrated with 200 mM ammonium acetate. All concentrations were calculated with respect to the most abundant oligomers, and they were diluted in 200 mM ammonium acetate to reach the concentration used for native MS: 30 μM (CA), 5 μM (ADH), 12 μM (apo-transferrin).

NanoESI-SLIM:

Nano-electrospray emitters used for this study were prepared in-house using pulled borosilicate glass capillaries (BF100-78-10, Sutter Instrument, Table S1 for the recipe). The dimensions of the glass capillaries were 0.78 mm (ID) and 1 mm (OD). The glass capillaries were pulled using a micropipette puller (P-1000, Sutter Instrument) to produce nano-ESI emitters of ~5 μm o.d. (and ~2 μm i.d.) To supply the electric field required for ionization, a platinum wire was used, and a voltage varying between 800 V to 1500 V was applied. The nano-ESI tip was positioned ~5 mm away from the inlet to the SLIM instrument. The voltages and distances varied slightly between experiments depending on experimental specifics for optimal signal intensity.

SLIM instrumentation description:

Figure 1 displays the schematic diagram of the SLIM TWIMS-MS system used for analysis of native-like proteins. Ions generated by nano-ESI tips were transmitted into a curved high-pressure ion funnel (HIF) at ~10 Torr through a multi-capillary inlet (5x250 μm i.d.) heated to 50 °C to retain proteins in near-native state (per literature citing <100°C as ideal ¹⁸). Subsequently, ions traveled across the low-pressure ion funnel (LIF, ~2.6 Torr nitrogen) to enter the SLIM chamber held at ~2.5 Torr nitrogen. Generally, when the pressure decreases from LIF and moving towards SLIM (i.e. a negative pressure gradient), we can expect some neutrals to traverse from upstream pressure regions into the SLIM chamber. Typically, a positive pressure gradient between LIF and SLIM chamber (i.e. when the SLIM is at a slightly higher pressure than the LIF) while beneficial to keep neutrals from entering the SLIM chamber, also mandates higher field gradient to drive proteins of lower mobility across the pressure barrier between the two regions. Hence, we use nearly identical pressures between LIF and SLIM, but marginally negative pressure gradient between the two regions to provide a gentle electric field driving the proteins into the SLIM chamber. The upstream HIF provided a curved path for the ions and mitigating a line of sight for the neutrals between source and the SLIM. The last electrode of the LIF served as a conductance limit for pressure regulation as well as a DC electrode that could be dynamically controlled. The SLIM consisted of two planar electrode surfaces on printed circuit boards (PCB) mirrored with about 3 mm spacing. The standard symmetric electrode configurations were used,¹⁹ i.e. the top and bottom boards were mirror reflections of each other. A TW generated as ramp waveform profiles at a speed of 112 – 128 m/s and an amplitude of 37 V (i.e. a profile creating a linear drop of 37 V across 8 electrodes) was used. Ion accumulation was performed using in-SLIM ion accumulation between 20 and 50 ms, depending on experiment specifics. In “accumulation” mode, or during accumulation time (t_acumulation), the conductance limit electrode (CL) was supplied with a DC voltage that provided uniform electric field consistent with LIF region. During this time, the TW voltage in the SLIM accumulation region (colored green in the SLIM schematic show in Figure 1) was low, and typically less than 5V; and the ions migrate towards a blocking electrode and accumulate against it. During accumulation time the blocking voltage (red electrode representation in Figure 1) was set to a high value of 80 V relative to the TW (5V). After accumulation, the CL electrode voltage was set to a value of 90 V (see Table S3 in Supporting information showing voltage on all ion optic elements in the mode where ions pass continuously) to block incoming upstream ion beam post-accumulation. The accumulation of ions against the high voltage barrier can lead to activation of proteins at extremely large accumulation values, which was not the case in this work.

Figure 1.

Schematic diagram of the SLIM-MS instrumentation used for multi-pass IMS separations of native-like proteins. The spatial profile of the Ramp rise traveling waveform profile is displayed in the red box. The various regions in the SLIM device are highlighted. Inset shows a representative voltage timing diagram on the SLIM for injecting ion packets into the separation region.

Subsequently the ions in the SLIM accumulation region were injected into the SLIM separation region (black color region in SLIM schematic of Figure 1) by raising the voltage of the TW in the accumulation region to 37 V for both the TW electrode and the red blocking electrodes. Figure 1 (inset) shows qualitative timing diagrams to perform ion accumulation and ion injection. As shown, the accumulation of ions for an IMS/MS acquisition is performed at the later part of the separation time corresponding to the previous IMS/MS acquisition event. The t_injection time was set to an arbitrary value to ensure all ions exit the accumulation region. The typical ion injected packet widths were dependent on the time ions accumulated against the blocking voltage, the TW amplitude (37 V) and TW speed after switching to the injection mode.

Subsequently, ion separations were performed by applying ramp rise waveforms (displayed in Figure 1 and elsewhere ²⁰) with an amplitude of 37 V and speeds used were between 112 m/s and 128 m/s. The ramp waveform was selected since it provided the most uniform electric field from among other waveforms investigated on the SLIM, and also for its lower electric field intensities in comparison to other waveforms (i.e., square, sine) and hence surmised as having minimal ion heating effects during separation. ²⁰ The nominal separation path length of the system was 8-m serpentine path. In addition, RF (0.57 – 0.62 MHz, 300 V_p-p) and a DC potential applied to the guard electrodes (7V above the midpoint of the TW) were used for ion confinement. The separated ions exited the SLIM region to enter an Agilent 6538 qTOF-MS via an ion funnel and quadrupole ion guide. Additionally, the SLIM system used here had the capability of selectively routing a chosen mobility window to a path that ended at an electrode (typically a glassy carbon substrate) connected to a current amplifier.

The IMS peaks were parsed into the MS analyzer using a “TOF pusher transient time” that is 0.288 ms, in the extended mass range of 10,000 Th. A pre-defined frame length was used to acquire IMS/MS spectrum covering the full separation time for different proteins. The generated data was exported from the MS via a digitizer to Falkor software (developed in-house) which enabled conversion to a format compatible to display IMS and MS 2D data. In each IMS/MS separation experiment “frame” was made up of 10,000 “TOF cycles”, and a total of 200 to 500 frames were recorded and summed for data analysis.

For landing of native-like carbonic anhydrase described in Figure S6, selected ions were routed to a glassy carbon substrate. The details of the deposition capabilities of the instrument are as previously described. ²¹ The conductive deposition substrate was connected to a current amplifier (Stanford Research Systems, model SR570) to measure the ion beam current to the deposition substrate. The ion current was recorded using an oscilloscope (Tektronix model TDS5104B) connected to the current amplifier.

CCS calibration and data analysis:

A mixture of hexakis(fluoroalkoxy)phosphazines (HFAP; ESI-L low concentration tuning mixture, Agilent Technologies) was infused for 8-m single pass separations before and after acquiring data for native proteins. Data were visualized, and the arrival times of the calibrants and analytes were extracted using the in-house developed software tool (https://github.com/PNNL-Comp-Mass-Spec/UIMF-Viewer). The centroids of the mobility peaks were determined by fitting data to a gaussian distribution using the Origin Pro software. The average of the peak centroids obtained before and after infusing native proteins was used to create a calibration curve. The calibration procedure is provided in detail elsewhere. ²² Briefly, the reduced ^DTCCS values, ${\Omega }^{\prime }$, of the phosphazenes ions available in Supporting Information Table S2²³ were plotted against the averaged centroids of the mobility peaks. The data were fitted to a power function with an intercept of the form $y=a+b{x}^c$ where $y$ is the reduced ^DTCCS of the calibrant, and $x$ is the averaged peak centroid. The fitted curve was then used to calculate the reduced ^DTCCS of native proteins for given arrival time distributions. The corresponding CCS values, $\Omega$, were determined using eq (1)

$$\Omega ={\Omega }^{\prime }\cdot z\sqrt{\frac{1}{\mu }}$$ (1)

where $\mathrm{z}$ is the ion charge state and $\mu$ is the reduced mass of the ion-neutral pair defined as $\mu =\frac{M_{ion}.M_{gas}}{M_{ion}+M_{gas}}$. The CCS values from multi-pass separations were obtained using eq (2)

$$\Omega ={\Omega }_{ref}+z\sqrt{\frac{1}{\mu }}\cdot b\cdot c\cdot {\left(\frac{x_n}{n}\right)}^{c-1}\cdot \frac{(x_n-x_{ref,n})}{n}$$ (2)

where ${\Omega }_{ref}$ is the reference CCS selected from single-pass separations (i.e., CCS corresponding to the mobility peak centroid), $b$ and $c$ are the parameters acquired by fitting the single-pass separation data to a calibration curve of the form $y=a+b{x}^c$, $x_n$ is the arrival time obtained from multi-pass experiments, $x_{ref,n}$ is the arrival time corresponding to ${\Omega }_{ref}$, and $n$ is the total number of passes (i.e., $n=1$ for single-pass separations). The long initial path length (single pass length) of 8m in the SLIM provides sufficiently long arrival times in relation to instrument ${\mathrm{t}}_0$ values. This in conjunction with deconvoluting the multi-pass arrival time differences to single-pass equivalents enabled deriving accurate CCS values in multi-pass cases by using single-pass calibration curves. The CCS distributions determined using eq (2) were validated against those obtained directly using calibration curves from 2-pass separations of calibrants (Supporting Information Figure S1). Subsequently the relationship in eq (2) was applied for processing all n-pass separation data discussed below. The method used here helps extract CCS values from multi-pass experiments using single-pass calibration data, and without having to generate calibration curves for each pass. This method above is complimentary to similar multi-pass calibration approaches reported recently^24-26, which use the ion velocity derived explicitly from the single pass experiments and the 1m single-pass path length in a cyclic IMS system.

Molecular Dynamics (MD) Simulations and Collisional-Cross-Section (CCS) calculations:

A detailed method for the computation of protein structure using NAMD and the CHARMM36m force field was outlined previously. ^27-31 Briefly, proteins were initially simulated in an aqueous environment (with starting x-ray crystallographic structures obtained from PDB database) before transitioning to gas-phase conditions. Initially, the protein-water system’s energy was minimized. Then, the system’s temperature was gradually increased from 0 to 300 K. After performing NVT (Number-Volume-Temperature) equilibrium to stabilize temperature and NPT (Number-Pressure-Temperature) equilibrium to stabilize both temperature and pressure, the MD simulation was continued until protein’s structure was at an equilibrium, which was tested by calculating its CCS value using IMoS ^{32, 33} with Lennard-Jones Trajectory Method (LJTM). The criterion for having reached equilibrium was that the RMSD (root-mean-square deviation) of the protein structure evolving within the aqueous volume reached and stabilized at a value less than 2%. Protein was then extracted from its aqueous ensemble for gas-phase simulation. Before advancing, protein’s residues were protonated for a specific charge state of interest. One selected charge state (from those seen in the SLIM experiment) was chosen. Charge distributions (or charge sites) were determined using Collidoscope and the residues were protonated using CHARMM software.^34-37 While multiple charge state distributions are possible, we only considered Collidoscope reported most probable charge state distribution for CCS computations in this manuscript. Then, the energy minimization was repeated for the charged gas-phase proteins, and the simulation continued until the CCS values converged (where convergence criterion was that the CCS change was <1%). Subsequently, to avoid local minima and potential trapping of ions at larger CCS values (as reported previously ^38-41), harmonic biasing potentials were applied to guide the protein’s alpha carbons toward a reduced radius of gyration, thus decreasing its CCS. ^{42, 43} The process, described in detail elsewhere, ²⁷ involved stepwise compression with harmonic biasing potentials for 1.5 ns, followed by 1.5 ns relaxation intervals, repeated for 5 cycles. This approach, using incrementally compressions and relaxations of the protein, was continued till the change in CCS between two compression cycles was less than 1% indicating arrival at the most relaxed structure and its corresponding CCS value.

Results

The ESI-MS spectrum showed the mobility peaks corresponding to z=9+ and 10+ charge states of native-like state of carbonic anhydrase (Figure 2a). The arrival time distributions (ATD) in Figure 2a show the baseline separation between CA z = 9+ and 10+ acquired by a single-pass experiment (~8 m path length). The two charge state peaks are clearly distinguished, almost close to baseline. An additional separation passes through the SLIM system yielded a minor increase in the separation of the two peaks (Supporting Information Figure S2). Separation of adjacent charge states without ATD overlap allows for mobility isolation of the charge state of interest and can be useful for applications that typically rely on separation in the m/z domain. The inset in Figure 2a shows the deconvoluted mass spectra with the highest peak at ~29 kDa in agreement with literature. ⁴⁴ To determine the collisional cross-section (CCS) of the proteins in their near-native state, the CCS calibration curve (shown in Figure S3a Supporting Information) was created using the averaged arrival times of phosphazine ions. Subsequently, we calculated CCS values for the charge states using the relationship obtained from the calibration curve. Under certain experimental conditions (Figure S4, where sine waveforms with a higher TW amplitude of ~45 V, i.e. greater than 37 V were used which may have resulted in a marginal rise in ion temperatures) we were able to see 11+ charge state for the CA molecule, whose CCS was also assigned using the calibration curve.. The measured CCS values for z = 9+, 10+, and 11+ of CA, extracted from corresponding single-pass calibration curves collected during experiments, are shown in Figure 2b. In addition, Figure 2b reports the obtained CCS values for the corresponding charge states of CA using a commercial drift tube IMS instrument, with the protein suspended in two different buffers used for improved CCS calibration of native-proteins: ammonium acetate (AmAc) and ethylenediamine diacetate (EDDA). ⁹ The experimental CCS values obtained from SLIM and DTIMS are falling within ~0.6% experimental precision (^DTCCS_z=9+/z=10+ = 2480/2520 Å). The percent error in CCS values is comparable to that using the calibrant in same molecular class and solution/charge conditions which afford precise CCS measurements. ⁹ A point to note is that the standard deviations of the CCS determination (Figure 2b) are lower in the SLIM system compared to DTIMS. The longer nominal path length in the SLIM (8 m), compared to 1 m in DTIMS, provides longer arrival times. Any arrival time shifts (that contribute to the error bars in CCS) are therefore a smaller percentage of the arrival time in the SLIM measurements, leading to lower error bars when the ATDs and converted to CCS. Thus, the longer path lengths, or effectively the higher separation power offered by SLIM increased the precision in the CCS assignment. The CCS measured in SLIM for z = 10+ is compared to that of the simulated molecular structure (as briefly described in Methods section above). The CCS of the obtained structure was calculated using the IMoS software, and the calculated ${}^{\text{MD}}\text{CCS}_{\mathrm{z}=10+}$ is 2700 Ų, an agreement with 6% difference compared to the average CCS (2536Ų) measured in SLIM. To compare the native-MS separations using SLIM, we obtained the MS spectrum of denatured CA (Supporting Information Figure S5a) where a series of partially unfolded peaks are revealed, in addition to several potential fragment peaks. The deconvoluted mass spectrum shown in Figure S5b also shows evidence of loss of Zn.

Figure 2.

(a) Arrival time distributions and mass spectra corresponding to z = +9 and +10 of native-like carbonic anhydrase in a single-pass experiment and the deconvoluted mass spectrum inside the red box (b) Calibrated Collisional Cross-Sections (CCS) and comparison with literature derived (Stiving et al ⁹) CCS values of proteins suspended in ammonium acetate (AmAc) and ethylenediamine diacetate (EDDA). Inset structure of bovine-CA z = 10+ and corresponding computed CCS value are noted.

Figure 3a shows the ATDs and mass spectra of tetramers of native-like alcohol dehydrogenase tetramers (ADH, ~148 kDa) analyzed by conducting ~8 m single-pass separations in the same platform. The plot in Figure 3a reveals three different charge states z = +23, +24 and +25. The observed charge states correlate with previously reported native-MS spectra for ADH tetramers using custom Thermo Fisher QE-EMR system⁴⁵, albeit the observed charge states being lower in the current setup. This could arise from multiple factors like lower capillary temperature used here (50°C), dual funnels providing low voltage gradients to transfer ions into the SLIM, and a small negative pressure gradient between LIF and SLIM allowing a smaller electric field to drive low mobility proteins gently into the SLIM system (Table S3 provides voltages used in the overall ion optics used here).

Figure 3.

(a) Extracted ATDs (top) and mass spectra (bottom) of native ADH tetramer z = 23+, 24+, and 25+ from a single-pass separation (b) Extracted ATDs for multi-pass experiments with two passes (top), three passes (middle) and four passes (bottom) identical to the ATD in 3a.

The ATDs of the observed charge states in the single-pass experiment are overlapping and do not show clear separation between adjacent charge states. Multi-pass separations were implemented to obtain improved peak resolution. Figure 3b presents the extracted ATDs for ADH tetramers with up to four passes respectively. We observe the mobility peaks from the single-pass separation experiment are distributed in a narrower range of arrival time (top panel in Figure 3b). When greater number of passes are executed, the ATDs provide improved separation between peaks. The labels in Figure 3b also indicate the TIC values for each pass. Within experimental variation of signal intensity between experiments, the TIC loss as the number of passes increase appears to be low. However, clearly there is a downward trend in the TIC also contributed by a gradual increase in the peak width and progressive erosion in S/N as peak tails fall below noise due to broadening. Specifically, the TIC (area under the curve) was found to be 2.78x10⁶, 2.03x10⁶ and 1.86x10⁶ for passes two three and four respectively, showing a reduction in signal intensity. We limited the number of passes attempted for ADH in this experiment due to broadening of the peaks, lapping of peaks and drop in TIC values.

For the single pass experiment, the calibration curve calculated using a mixture of phosphazine standard calibrant molecules (detailed in Figure S3b) provided the collisional cross section ($\Omega$) values given by the equation $\frac{\Omega }{\mathrm{z}\frac{1}{\sqrt{\mu }}}=\mathrm{A}+\mathrm{B}{\mathrm{t}}^{\mathrm{c}}=\text{-}990.77469+222.15288{\mathrm{t}}^{0.36194}$. Using Eq (2) we were able to obtain CCS values from the ATDs of multi-pass separations which provided peaks with improved separation. Using the two expressions above, the CCS distributions for z = 23+, 24+, and 25+ were calculated and are shown in Figure 4a, 4b, 4c respectively for each charge state. The associated theoretical CCS for the z = 24+ charge state for the structure calculated for the molecule is reported in Figure 4b inset. The experimental value obtained from SLIM for the CCS of ADH 24+ was 7204 Ų , while the theoretical calculation of 7585 Ų correlated closely with prior literature value of 7540 Ų when using DTIMS⁹, with a precision within 6% when the values are compared. Also, progressive passes seem to result in progressively thinner (albeit mildly so) CCS distribution providing better precision in CCS assignment for the ATDs. An interesting observation is seen in Figure 4a for charge state +23, where with increasing number of passes, a pre-arriving shoulder feature becomes more clearly resolved. When comparing the ATDs of +23 and +24 (Figure 3a) one can notice co-arrival of the left-side shoulder of +23 and the apex of +24 indicative of potential contribution of proton transfer post-separation from +24 to formation of the pre-arriving feature of +23. It is noteworthy that the CCS distributions for each charge state do not appear to be affected up to four passes (corresponding to an IM separation time >2 s) indicative of minimal conformational changes during the experiment. This may be attributed to that effective ion temperatures (for given drift velocity) are comparable to the near-ambient temperature under typical operating conditions in SLIM leading to long lifetimes of the initial native-like structures.^{15 46}

Figure 4.

Collision cross section distributions for individual charge states of ADH, (a) +23 charge state (b) +24 charge state and (c) +25 charge state. Figure 4b shows inset with calculated structure of ADH and computed CCS value of 7585 Ų.

It is interesting that the time-based resolving power for z = 24+ gradually increased from ~20 to ~24 as the number of passes (n) increased from n = 1 to 4. In addition, the resolution between two peaks corresponding to z = 23+ and 24+ improved from 0.62, 0.80, 0.90, and 0.90 for n = 1, 2, 3, and 4 respectively. While notable improvement in resolution was achieved up to three passes, only less than 0.2% increase in two-peak resolution was observed after three passes. Analysis of native proteins using the multi-pass platform shows that the percent increase in resolving power (6% increase for additional pass) is lower than the theoretical increase scaled to $\surd \mathrm{n}$ which was accomplished for small molecules.^{10, 47} While prior literature does point to challenges in resolving conformation distributions,^{48 49} a broader sweep of the parameter phase-space (i.e. waveform shape, amplitude and speed) may be likely investigations for a future work.

Figure 5 shows CCS distributions for four different charge states of apo-transferrin (~79 kDa) which were obtained from ~8-m single-pass separation. The mobiligram in the inset (top) shows the presence of four distinct peaks corresponding to z = 16+, 17+, 18+, and 19+ charge states, with their corresponding arrival time distribution. Using a calibration curve (see Supporting Information Figure S3c) and the arrival time distributions for each of the charge states leads to the corresponding CCS distributions for all the charge states which are overlaid on top of each other in the figure.

Figure 5.

(a) CCS distributions of apo-transferrin of z = 16+, 17+, 18+, and 19+. The inset displays corresponding mass spectra of native apo-transferrin obtained from single-pass separations. The simulated CCS for z = 17+ is 4934 Ų within ~2% error compared to the intensity-weighted average CCS (4840 Ų). (b) ATDs of the charge state 16+, 17+, 18+, and 19+ of apo-transferrin.

The theoretical CCS for 17+ charge state of apo-transferrin was computed, and the structure and the CCS value (4934 Ų) are noted as an inset in Figure 5a, with the average CCS for apo-transferrin from the SLIM experiment determined to be 4840 Ų. This constitutes a ~2% difference between the two values. Interestingly, 16+ charge state presents a multimodal distribution, which when observing from ATDs of individual charge states (Figure 5b), arises from charge stripping from 17+ and 18+ charge states. The higher charge states (18+ and 19+) appear to largely retain their Gaussian distributions. The charge transfer to 16+ is occurring post-separation, evidenced by alignment of the additional features of 16+ ATD with arrival time of the apex of 17+. The 16+ charge state however is a minor peak in terms of its relative abundance (see MS spectrum in Figure 5a inset) and hence the relative proportion of 17+ signal that is subjected to the charge transfer is marginal.

Several native-MS platforms have advanced the field significantly, with high resolution IMS with TIMS-TOF^{49, 50} with acquisition times of few hundred milli-seconds as well as cyclic-TWIMS systems^{46, 48, 51} using multi-pass methods. The preservation of protein’s native structure during these separations is achieved through varieties of instrumental and operational paradigms. SLIM has previously demonstrated, for a range of small molecules, manipulations including trapping⁵², accumulation⁵³, enrichment, isolation and deposition²¹ of ions in a mobility selective fashion. This feature of SLIM presents a particularly exciting opportunity for the characterization of native proteins if they could be isolated with mobility selection for the purposes of enrichment to enable secondary structural characterization. The caveat is the need to retain the protein structures in their native-like state without unfolding or loss. Previous work has observed conformational changes occurring with increase in experiment time scales in IM devices. For small native molecules (i.e., cytochrome-c, ubiquitin) experiment time to observe conformation transitions was on the order of tens of milliseconds. The time scales appropriate for larger native proteins were an or two order(s) of magnitude longer than smaller molecules. In a recent study performed by the Bush group¹³ it was discovered that the time scale for structural changes to occur was an order of magnitude longer using SLIM-based IM instrument, due to molecular internal energy (ion temperature) in SLIM comparable to the ambient-like temperatures.

To demonstrate the isolation of native-like proteins using SLIM, we targeted the ATDs of apo-transferrin to select and isolate a small mobility window corresponding to z = 17+ (red dashed box in Figure 6a). The ATDs for each of the charge states overlap, hence the chosen selection window would have contamination from other peaks in the m/z and ATD domain. However, through progressive selection over multiple passes, the 17+ charge state can be isolated with no other peak being present both in the m/z and mobility (arrival time) domain. Figure 6a shows the charge states z = 16+ and 18+ arrive before t = 500 ms and a portion of z = 17+ ions in the selected arrival time window appear (orange dashed box) after completing four additional passes. Ions in the selected time window initially included a small population of adjacent charge states due to overlapping of their arrival time distributions, i.e. during the single-pass separation, the tail of the z = 18+ mobility peak overlapped with the shoulder of the z = 17+ peak which appeared as the first sharp mobility peak at t ~ 600 ms. The arrival time window associated with the targeted charge state was re-selected for additional passes of separations. At t~1400 ms, a mobility peak appears only containing a portion of z = 17+ ions which illustrates that repeated selections of the arrival time window in additional multi-pass separations allowed for complete isolation of the targeted portion of the z = 17+. This arrival time distribution also contained no other species in the mass domain, providing both mass-and-mobility selective isolation from the protein spectrum. The resulting CCS distribution of the selected z = 17+ ions is displayed as Figure 6c. When compared to the distribution without mobility selection the CCS values are identical (Figure 6c red line vs orange line). Thus, a desired range of mobilities can be selected to isolate a specific population through sequential selections (if-needed) to arrive at mass and mobility selected peaks. And these selections can clearly be performed while maintaining the structure of the native protein.

Figure 6.

(a) A portion of the mobility peak for z = 17+ present in the mobiligram was selected (red dashed box) and routed to complete four additional laps (a total of 40-m path length) for isolation of the z = 17+ from other charge states. (b) Shows a zoomed in area of the 2D graph of m/z and arrival time. (c) The resulting CCS distribution of the selected mobility range for z = 17+ (orange) in comparison to that of the full CCS distribution prior to mobility selection and multi-pass separations.

Additionally, such selection can be performed using SLIM at a reasonable speed and ion flux. Several picoamps (per individual IMS peak) of current are transmitted through the SLIM (see Supporting Information Figure S6 which shows ~30pA for Carbonic Anhydrase 10+ charge state). This value while lower than ion collection currents in continuous mass selective soft-landing experiments reported previously, it arises from the smaller kinetic energy of ions in SLIM operating at 2.6 torr. However, it still constitutes a collection on the order of ~1 million particles on the timescale of a minute (see Supporting Information for details). Also, across protein sizes (e.g. while using Carbonic Anhydrase and Apo-transferrin) the kinetic energies of ions within the SLIM (and thus as they exit SLIM towards a collection electrode at a constant electric field ~20 V/cm) was less than 5eV enabling soft landing conditions (see Supporting Information for details). These will be subjects for further exploration to enable comprehensive characterization of proteins by using SLIM as a structural separation as well as a preparative platform for secondary characterization (e.g. cryoEM).

Conclusion:

In this work, we report the initial application of multi-pass SLIM IMS separations for gas-phase proteins in their near-native state. The mobility peaks of native(-like) proteins was collected using an 8-m nominal path length single-pass SLIM separation. Carbonic Anhydrase presented baseline separation of three adjacent charge states (z = 9+ to 11+). On the other hand, single-pass separations did not fully resolve mobility peaks for charge states z = 23+ to 25+ of alcohol dehydrogenase. Observable improvement in mobility peak resolution was observed by performing multi-pass separations. The CCS values obtained from SLIM experiments were independent of the ion path length, showing protein structure is stable over the extended timescale of analysis. The analysis of charge states z = 16+ to 19+ for apo-transferrin show separation of the charge states and unimodal distributions (for 18+ and 19+) in the time domain, and some evidence of charge stripping post separation to create a multimodal distribution for low abundance 16+ charge state. The capability has the utility for selection and isolation of a mobility (or CCS) range of interest using SLIM selection, re-routing and multi-pass capability. From overlapping peaks with 18+, 17+ and 16+ charge states, we selected a mobility window which initially included a major portion of the z = 17+ mobility peak and some portion (shoulder and tail) of the z = 16+ and 18+ peaks. With additional mobility selection and separations over the path length of ~40 m the mobility peak for the single charge state (z = 17+) was completely isolated without change in its CCS value. The multi-pass SLIM separations demonstrated here for analysis of native-like proteins show potential for selective isolation of molecules within a targeted conformation/mobility/arrival-time space achieving mass-and-mobility selection, opening avenues for secondary characterization of SLIM isolated macromolecules.

Supplementary Material

A Supporting Information document accompanies this manuscript where we provide arrival time distribution data, comparison and validation of calibration strategies used, mobilograms, calibration curves, deconvoluted mass spectra, and current measurement made on selected protein arrival times.