Cyclic Ion Mobility-Mass Spectrometry and Tandem Collision Induced Unfolding for Quantification of Elusive Protein Biomarkers

Abstract

Sensitive analytical techniques that are capable of detecting and quantifying disease-associated biomolecules are indispensable in our efforts to understand disease mechanisms and guide therapeutic intervention through the early detection, accurate diagnosis, and effective monitoring of disease. Parkinson’s Disease (PD), for example, is one of the most prominent neurodegenerative disorders in the world, but the diagnosis of PD has primarily been based on the observation of clinical symptoms. The protein α-synuclein (α-syn) has emerged as a promising biomarker candidate for PD, but a lack of analytical methods to measure complex disease-associated variants of α-syn has prevented its widespread use as a biomarker. Antibody-based methods such as immunoassays and mass spectrometry-based approaches have been used to measure a limited number of α-syn forms; however, these methods fail to differentiate variants of α-syn that only display subtle differences in sequence and structure. In this work, we have developed a cyclic ion mobility-MS method that combines multiple stages of activation and IM selection to quantify α-syn variants using both mass- and structure-based measurements. This method can allow for the quantification of several α-syn variants present at physiological levels in biological fluid. Taken together, this approach can be used to galvanize future efforts aimed at understanding the underlying mechanisms of PD and serves as a starting point for the development of future protein structure-based diagnostics and therapeutic interventions.

Graphical Abstract

INTRODUCTION

Recent reports have indicated strong and varied links between cellular proteoform expression levels and human disease.^1–9 Such proteoforms are now routinely analyzed using a variety of high-resolution tandem mass spectrometry (MS) techniques, including those incorporating a wide range of separation technologies and MS-based imaging modalities.^10–20 While most proteoform populations are defined based on variations in post-translational modification or sequence,¹ there is a growing recognition that structural proteoforms can be similarly linked to discrete biological functions and disease states.^21,22 Examples of diseases where such structural proteoforms play central roles include many forms of cancer^23,24 and protein misfolding disorders,²⁵ including Alzheimer’s Disease.²⁶

The etiology of Parkinson’s Disease (PD), one of the most prominent neurodegenerative disorders in the world, also relies on structural proteoforms.²⁷ Specifically, PD pathology is characterized by the loss of dopaminergic neurons and the accumulation of Lewy bodies, which are primarily comprised of the protein α-synuclein (α-syn).^28,29 However, the diagnosis of PD has primarily been based on clinical symptoms that are only present after an irreversible loss of dopaminergic neurons.^30,31 Several mutations of the α-syn gene (e.g., A53T and E46K) have been linked to severe forms of familial PD through protein misfolding.^29,32–38 Such point mutations are heterozygous in nature – both wild-type (WT) and mutant α-syn are co-expressed within an organism exhibiting the altered α-syn gene. Post-translationally modified (PTM) forms of α-syn, such as α-syn phosphorylated at S129 (pS129), N-terminally acetylated, or C-terminally truncated, and α-syn oligomers have also been associated with nongenetic forms of PD.^39–42 While these prior experimental observations indicate α-syn as a promising PD biomarker and therapeutic target, they also highlight that the molecular etiology of PD, and the broader disease associations of a-syn, remains poorly understood. For example, α-syn is also associated with other neurodegenerative diseases such as dementia with Lewy bodies (DLB) and multiple system atrophy (MSA).⁴³ In addition to the complexity of α-syn at the proteoform level, structural differences between α-syn strains that originated from the brains of patients with PD, DLB, and MSA have been observed.⁴⁴ In order to uncover the molecular mechanisms connecting expression levels of individual α-syn proteoforms to the pathology of PD and other disorders, as well as to leverage α-syn as a biomarker for neurodegenerative disease, there is a need to develop analytical methods that are capable of quantifying individual α-syn proteoforms in a conformationally-sensitive manner.²⁹

Antibody-based methods have commonly been used to measure α-syn concentrations because of their sensitivity and limits of detection, but commercially available antibodies are limited and make it challenging to differentiate many of α-syn variants implicated in disease.^45,46 Alternatively, MS-based methods offer sensitive, label-free assessments of proteins in complex mixtures.^47–49 MS methods have been used to quantify α-syn at the intact and enzymatically digested levels.^50–52 Though, current MS methods often fail to detect variants that are linked to disease through alterations in their 3D structure or are masked by iso-mass protein interferences which are challenging to differentiate even with high resolution MS.

Ion mobility (IM), a gas-phase technique that separates gas-phase ions based on their size and charge, has been widely used combination with mass spectrometry (IM-MS) for applications in structural biology.^53,54 Ions are typically trapped and released into a drift region filled with inert gas, where the ions are separated based on differences in mobility under a weak electric field. Native IM-MS experiments allow native-like protein ions to be transferred to the gas phase to while retaining aspects of higher-order protein structure.^54,55 Collision induced unfolding (CIU) is an extension of native IM-MS where native-like protein ions are activated through energetic collisions with a background gas, causing the ions to increase in internal energy and unfold in the gas phase.⁵⁶ As the protein unfolds, changes in its collision cross-section can be detected using IM. Prior work has demonstrated that CIU can uncover partially folded conformational intermediates which are unique to isoforms and structural states that are overlooked by other MS methods.⁵⁶ For instance, CIU has been used to capture subtle differences in protein structure due to disease-associated sequence variation,^57–59 disulfide bonding,⁶⁰ posttranslational modifications,⁶¹ and ligand binding.⁶² More recently, CIU has been used as a tool to quantify large (66–150 kDa) proteins exhibiting similar molecular weights and sequences.^57,63 Despite the ability of CIU to differentiate and quantify subtly different protein variants, CIU has yet to be used to quantify disease-associated small (<15 kDa) biomarkers present in biological matrices.

Herein, we utilize cyclic ion mobility-mass spectrometry (cIM-MS) to quantify elusive α-syn proteoforms associated with PD.⁶⁴ We show that CIU can differentiate WT and E46K α-syn and that calibration curves can be generated by plotting IM peak area versus the ratio of WT to mutant present in several known mixtures. For the first time, we leverage cIM-MS enabled tandem CIU (CIU²) experiments for such quantitative experiments, which produces more differentiated unfolding patterns when compared to conventional CIU (CIU¹) experiments. We demonstrate that the relative quantification of WT and E46K α-syn based on data from CIU² experiments produces higher sensitivity and lower limits of detection compared to CIU¹ experiments. Finally, we develop a fully automated quantitative CIU² workflow where samples are desalted online using size exclusion chromatography (SEC) and CIU² data are collected as the protein elutes from the column. Our quantitative CIU² method can determine the relative amounts three disease-associated variants of α-syn directly from artificial cerebrospinal fluid (aCSF). Overall, this work demonstrates that CIU can be used as a quantitative MS technique to measure proteoforms which are difficult to differentiate using conventional quantitative MS techniques. We envision that the method described here could be used in clinical research to provide insight into the role of different α-syn proteoforms in the progression of PD and could pave the way for structure-based diagnostics and therapeutic interventions.

EXPERIMENTAL SECTION

Materials and Reagents

Phospho-S129 (pS129) α-syn was purchased from Proteos (Kalamazoo, MI). Lysogeny agar, lysogeny broth (LB), and iso-propyl-ß-D-1-thiogalactopyranoside (IPTG) were purchased from Fischer Scientific (Fair Lawn, NJ). Chloramphenicol was purchased from Millipore (Billerica, MA). Water was purified in-house using a Milli-Q water purification system from Millipore. Micro Bio-Spin P-6 columns were purchased from Bio-Rad (Hercules, CA). Vivaspin 6 centrifugal concentrators with a 3 kDa molecular weight cutoff were purchased from GE Healthcare (Chicago, IL). All other materials and reagents were purchased from Sigma-Aldrich (St. Louis, MO).

Protein Expression and Purification

Wild-type (WT), E46K mutated (E46K), and A53T mutated (A53T) α-syn were expressed and purified using a previously described method, with slight modifications.⁶⁵ Plasmids (Addgene, Watertown, MA) were transformed into Escherichia coli BL21 (DE3) cells. A single colony was used to inoculate 25 mL of LB media supplemented with 100 μg/mL carbenicillin for incubation at 37 °C with continuous shaking at 250 rpm for ~6 h. The pre-culture was used to inoculate 0.5 L of LB media with 100 μg/mL carbenicillin, which was then incubated under the same conditions until an optical density at 600 nm of 0.6 was reached. Protein expression was induced using 1 mM IPTG and the culture was left to grow at 37 °C for 4 h with shaking at 250 rpm. Following incubation, cells were pelleted via centrifugation at 5000 × g for 20 min at 4 °C. Pellets were flash frozen using liquid nitrogen and stored at −80 °C.

Cell pellets were resuspended in 50 mL of the starting anion exchange chromatography (AEX) mobile phase (100 mM AmAc) and incubated over a boiling water bath for 15 min. The lysate was centrifuged at 20000 × g for 45 min at 4 °C. The supernatant was loaded onto a 5 mL HiTrap Q HP AEX column (Cytiva, Marlborough, MA) which was equilibrated with 100 mM AmAc. The protein was eluted using a gradient from 0–0–100–100 % B over 0–10–30–40 column volumes (mobile phase A was 100 mM AmAc and mobile phase B was 1 M AmAc). Fractions containing α-syn were pooled, loaded onto a 16 mm i.d. × 600 mm length HiPrep Sephacryl S-200 HR SEC column (Cytiva), and eluted with 100 mM AmAc. Fractions containing α-syn were pooled, diluted to 40 μM (as determined by a bicinchoninic acid assay) with 100 mM AmAc, flash frozen with liquid nitrogen, and stored at −80 °C.

Sample Preparation

For analyses of α-syn in AmAc, the stock solutions were further purified using P-6 spin columns equilibrated with 100 mM AmAc and diluted to a working concentration of 5 μM. Biological samples were mimicked by spiking the stock protein solutions into artificial cerebrospinal fluid (aCSF). aCSF was prepared to contain 145 mM sodium chloride, 2.68 mM potassium chloride, 1.40 mM calcium chloride, 1.01 mM magnesium sulphate, 1.55 mM dibasic sodium phosphate, and 0.45 mM monobasic sodium phosphate, and adjusted to pH 7.4 using sodium hydroxide. Biological samples were prepared by diluting the α-syn stock solutions in aCSF, which were either injected as is or preconcentrated using a Vivaspin 6 concentrator that had been passivated overnight with a 1 mg/mL solution of bovine serum albumin for online SEC experiments.

Ion Mobility-Mass Spectrometry

All IM-MS experiments were performed using a Select Series Cyclic Ion Mobility-Mass Spectrometer from Waters (Milford, MA).⁶⁴ The backing and source pressures were 2.35 mbar and 8.56 × 10⁻³ mbar, respectively. The trap traveling-wave ion guide was pressurized to 3.20 × 10^–2 mbar of nitrogen gas. The helium cell flow was set to 120 mL/min of helium and pressurized to 2.05 mbar. The cyclic separation region was operated at 40 mL/min of nitrogen and pressurized to 1.76 mbar. The transfer cell was pressurized to 3.20 × 10^–2 mbar of nitrogen gas. The time-of-flight (TOF) mass analyzer was operated in V-mode over 50–8000 m/z at a pressure of 4.54 × 10^–7 mbar. IM separation was achieved with a traveling wave at 30 V wave height traveling at 1000 m/s. The racetrack bias was 70 V and the repeller voltage was 100 V. IM data was acquired over 200 bins at a rate of 3 TOF pushes per bin. The scan rate was 1 s. Static spray nanoelectrospray ionization (nESI) experiments were performed by transferring 3 μL of sample to a borosilicate capillary emitter pulled in-house using a P97 micropipette puller (Sutter, Novato, CA) to an approximate tip inner diameter of 5–10 μm. Pulled emitters were coated in gold using a SC7620 mini sputter coater (Quorum Technologies, Laughton, UK). Ions were generated using a NanoLockSpray source operated in positive mode at 1.0 kV. The cone voltage was 5 V, the source offset was 0 V, and the source temperature was 20 °C.

CIU¹ and CIU²

Conventional CIU¹ experiments were performed using static spray nESI. Ions corresponding to a single charge state were selected using the quadrupole and subjected to collisions in the traveling wave ion trap prior to one pass of separation in the cyclic ion guide. The collision voltage in the trap was ramped from 5 to 100 V in 5 V increments with a 30 second dwell time at each voltage to construct the CIU¹ data. Initial CIU² experiments were performed using static spray nESI. Ions corresponding to a single charge state were selected using the quadrupole. The collision energy of the traveling wave ion trap was set to the voltage leading to the desired amount of CIU¹ activation. Ions were first separated by one pass around the cyclic ion guide. Using a “slicing” sequence set in the instrument control software, ions with higher mobility than those corresponding to the feature of interest were ejected to the TOF without acquisition. Ions to be subjected to CIU² were then ejected and stored in the stacked ring ion guide directly prior to the array of the cyclic ion guide (i.e., pre-array storage). Unwanted ions with lower mobility relative to the ions of interest were ejected to the TOF without acquisition. The stored ions were then subjected to a second round of CIU, where activation was achieved by increasing the reinjection voltage from storage to the cyclic ion guide from 5 to 100 V in 5 V increments with a 30 second dwell time at each voltage.

Automated CIU² Analysis Using Online SEC

Automated CIU² experiments were performed using a Waters ACQUITY M-Class UPLC system. One μL of sample was injected in triplicate onto a prototype 250 Å microflow size exclusion chromatography column (1 × 50 mm, 1.7 μm) provided by Waters.⁶⁶ The mobile phase was 100 μM AmAc at 30 μL/min. The column was directly interfaced with the cIM-MS via the LockSpray ESI source equipped with a low-flow ESI probe. The ion source conditions were tuned to provide the highest signal intensity while avoiding unintentional in-source unfolding: the capillary voltage was 2.0 kV, the cone voltage was 30 V, the source offset was 0 V, the source temperature was 50 °C, the cone gas flow rate was 0 L/h, the desolvation gas flow rate was 750 L/h, the desolvation gas temperature was 200 °C, and the nebulizer pressure was 6 bar. The quadrupole was set to simultaneously isolate the 7+ charge state of WT and variant α-syn present in a mixture. The trap collision voltage was set to 40 V. Ions from 64–66 ms were stored in the pre-array storage and reinjected into the cyclic ion guide with 55 V of additional reinjection energy (example Cyclic sequence table is found in Table S1). CIU² data were acquired under these conditions as the α-syn peak eluted from the SEC column. In this case, data was only acquired for a single set of CIU² conditions (i.e., the reinjection voltage was not ramped as described in the experiments above).

Data Processing

IM spectra were extracted from .raw data files using TWIMExtract 1.6.⁶⁷ Data was extracted over an m/z range that encompassed a single charge state. The extracted data was processed further using CIUSuite 2.⁶⁸ Data was smoothed using a Savitzky-Golay filter with a window size of 5 and 1 smooth iteration (a 1D Savitzky-Golay filter was used for individual IM spectra and a 2D Savitzky-Golay filter was used for full CIU datasets). Three-dimensional CIU fingerprints were generated using the plot function and compared using root-mean-square deviation (RMSD) analysis. Gaussian fitting of the smoothed data was performed using protein only mode. The maximum number of protein components was 3, the expected protein peak width was 5 ± 2 ms, and the peak overlap penalty was set to relaxed. Areas of the Gaussian fits were used to generate calibration curves for quantification. Calibrations curves were fit to a linear regression using Prism 9.5.1 (GraphPad, San Diego, CA).

RESULTS AND DISCUSSION

Overview of CIU² Analysis

The general workflow used for CIU² experiments with α-syn is shown in Figure 1. Ions were generated using nESI conditions which produced a native-like charge state distribution. Charge states that were sufficiently intense and presented a compact initial structure without additional activation were selected for CIU¹ analysis. CIU¹ fingerprints representative of the 6+, 7+, and 8+ charge states of WT α-syn are shown in Figure 1A. Additional signals associated with higher charge states were present, but either dissociated or did not unfold upon activation in the trap region. We proceeded with using the 7+ charge state for comparison between WT and variant α-syn. The 6+ ion only underwent one CIU¹ unfolding transition, limiting the number of states to be compared between proteins, and the 8+ ion unfolded via transient CIU¹ intermediates, which might affect the reproducibility of later quantitative analysis.

Figure 1.

Workflow for CIU² experiments. A) CIU¹ fingerprints for the 6+, 7+, and 8+ charge states of WT α-syn. B) Example RMSD plot used to compare CIU data. C) Regions of the cIM-MS instrument that were important for CIU² experiments. D) Feature detection for a 7+ CIU¹ fingerprint of WT α-syn. E) Feature detection for a WT α-syn CIU² fingerprint that resulted from selection of 7+ ions from 56–58 ms at 20 V CIU¹ collision energy. The arrival time axes in Figure 1D and Figure 1E were aligned (as demonstrated in Figure S1) for direct comparison between CIU¹ and CIU² data.

An example of a plot that was used to compare CIU data is shown in Figure 1B. The unique ability to select ions based on mobility and perform multiple stages of activation with cIM-MS was utilized to perform CIU² experiments and further probe differences in the gas-phase unfolding of α-syn proteoforms. Figure 1C highlights regions of the cIM-MS instrument that were important for CIU² analyses. The initial steps of a CIU² experiment are performed by activating ions in the trap, separating the unfolded species with one pass around the cyclic ion guide, and storing unfolded intermediates in the pre-array storage ion guide. The second stage of activation was performed by increasing the reinjection voltage that transferred ions from the pre-array storage to the cyclic ion guide. Finally, the reinjected ions underwent one pass of separation around the cyclic ion guide and were sent to the TOF for detection. Any region of a CIU¹ fingerprint could be selected for CIU² analysis, including stable features that were present over multiple levels of activation or more short-lived unfolding intermediates. We used comparison plots such as the one shown in Figure 1B to identify conditions that appeared to be the most successful at differentiating WT and variant α-syn with a single stage of activation and used CIU² to further probe these differences.

Figure 1D shows feature detection for the 7+ WT α-syn CIU¹ fingerprint and Figure 1E shows feature detection for CIU² fingerprint resulting from selection of 7+ ions between 56–58 ms with 20 V of CIU¹ activation using the trap. The timing of injection and reinjection from the pre-array storage for CIU¹ and CIU² was kept constant between experiments, which allowed us to align the arrival time axes of CIU¹ and CIU² data (Figure S1). Interestingly, substantial differences were observed between the CIU¹ and CIU² data. The median arrival time of Feature 1 (56.58 ms) in the CIU² fingerprint falls within the mobility range that was selected from the population of CIU¹-activated ions (56–58 ms). The median arrival time of Feature 2 in the CIU² fingerprint was ~3 ms longer than the median of Feature 2 in the CIU¹ fingerprint. Features 3 and 4 were prominent in the CIU² fingerprint but were not sufficiently intense for detection in the CIU¹ fingerprint. These results demonstrated that low-level intermediates which were suppressed by other CIU¹ features could be selected and amplified using CIU².

Quantification of WT and E46K α-Syn Using CIU¹

WT and E46K α-syn only differ in molecular weight by 1 Da and were difficult to distinguish by MS alone (Figure 2A). However, notable differences were observed when the two proteoforms were intentionally activated and subjected to CIU¹. Figures 2B and 2C show CIU¹ fingerprints for WT and E46K α-syn, respectively. Figure 2D shows an RMSD plot comparing the WT and E46K α-syn CIU¹ fingerprints. An RMSD of 8.79 % was calculated between the two fingerprints (a baseline RMSD of 1–2 % was observed between technical replicates). The transition region at 40 V of activation resulted in the most substantial differences between WT and E46K IM spectra (Figure 2E).

Figure 2.

Quantifying WT and E46K α-syn using CIU¹. A) Overlaid mass spectra for the7+ charge state of WT and E46K α-syn. B) CIU¹ fingerprint for the 7+ charge state of WT α-syn. C) CIU¹ fingerprint for the 7+ charge state of E46K α-syn. D) Plot comparing CIU¹ fingerprints in panels B and C. E) Overlaid IM spectra extracted from the CIU¹ data in panels B and C at 40 V of activation. F) Calibration curve generated by plotting the area of the second peak in panel E as a function of the amount of WT present in mixtures containing both WT and E46K α-syn. Error bars indicate the standard deviation of three technical replicates (some error bars are too small to be visible).

While we did not observe baseline resolution between WT and E46K α-syn, the observed peak intensities differed between variants and could be used to determine the relative ratio of WT to E46K α-syn in a mixture containing the two analytes. Relative calibration curves were generated by plotting the area of each peak or the relative ratio between the areas of different peaks in the spectra as a function of the relative amount of WT α-syn present in mixtures containing WT and E46K α-syn at different ratios. The calibration curve in Figure 2F demonstrates the correlation between amount of WT present in solution and the area of the second peak in Figure 2D. This calibration curve displayed the highest degree of linearity and sensitivity – plotting the areas of other peaks or ratios between peaks observed in the data were less quantitative (Figure S2).

CIU² Experiments Improve CIU-Based Quantification of WT and E46K α-Syn

While quantitative information was obtained from CIU¹ data, we hoped to utilize the additional features observed in α-syn CIU² data (Figure 1E) to further differentiate WT and E46K α-syn and improve quantitative analysis. The regions of the RMSD plot in Figure 2C that showed the greatest difference between WT and E46K α-syn were selected for CIU² analysis (the same regions that were used for quantification by CIU¹ in Figure 2). Several regions along the arrival time distribution observed with 40 V of CIU¹ activation were mobility selected and subjected to CIU² analysis. Figures 3A and 3B show CIU² fingerprints for WT and E46K α-syn, respectively, after selection of ions from 64–66 ms. Figure 3C shows an RMSD plot comparing the WT and E46K α-syn CIU² fingerprints in Figures 3A and 3B. The most differentiating conditions were observed at the transition occurring with 55 V of additional reinjection energy (Figure 3D). Improved signal-to-noise was observed in Figure 3D when compared to Figure 2E. This improvement could be due to the removal of additional chemical noise during the multiple stages of mobility selection and activation that took place during CIU². Additionally, only the intensity of one peak in Figure 3D differed between WT and E46K α-syn, where the intensities of all three peaks differed slightly in Figure 2E. In comparison to the calibration curve in Figure 2F that was generated from CIU¹ data, a two-fold increase in sensitivity was achieved during quantitative CIU² analysis that was performed by integrating Peak 1 in Figure 3D. In terms of limits of detection (LODs), the CIU¹ method could detect a minimum of 13.5% (675 nM) of one variant, where the CIU² approach could detect as low as 3.9% (195 nM) – a 3.5-fold improvement. Calibration linearity was also improved with CIU² – likely due to the improved signal-to-noise in the raw data. Overall, this highlights the utility of CIU² analysis to further differentiate closely-related proteins and improve conformation-based quantitative analysis.

Figure 3.

Quantifying WT and E46K α-syn using CIU². A) CIU² fingerprint for the 7+ charge state of WT α-syn resulting from selections of ions from 64–66 that were activated with 40 V in the ion trap. B) CIU² fingerprint for the 7+ charge state of E46K α-syn resulting from selections of ions from 64–66 ms that were activated with 40 V in the ion trap. C) Plot comparing CIU² fingerprints in panels A and B. D) Overlaid IM spectra extracted from the fingerprints in panels A and B at 55 V of additional activation upon reinjection. E) Calibration curve generated by plotting the area of the first peak in panel D as a function of the amount of E46K in mixtures containing both WT and E46K α-syn. Error bars indicate the standard deviation of three technical replicates (some error bars are too small to be visible).

Automated Quantification of Multiple α-Syn Variants in Biological Matrices

To further test the quantitative CIU² assay, we quantified WT and several disease-associated variants of α-syn directly from aCSF. Previously reported physiological levels of α-syn have varied considerably. A recent review indicated that previous studies have measured α-syn levels ranging from 4 pM to 5 nM in CSF and 250 pM to 370 nM in blood plasma.²⁹ We prepared samples at 20 nM (total α-syn concentration), which was well within the range of the reported physiological levels. However, 20 nM α-syn was undetectable by the instrument. We therefore used centrifugal preconcentrators to enrich 5 mL volumes of sample to levels that were detectable by static spray nESI (1–5 μM final concentration). A comparison of MS data generated for a 20 nM sample of WT α-syn in 100 mM AmAc with and without preconcentration is shown in Figure 4A, demonstrating that preconcentration is a viable approach to enable detection of α-syn at physiological levels from clinically relevant volumes of sample.

Figure 4.

Quantifying α-syn variants in aCSF using CIU². A) Overlaid mass spectra for 20 nM α-syn with and without pre-concentration. The y-axes were offset to assist with visualization. B) Total ion chromatogram from the SEC separation of WT α-syn in aCSF. C) Overlaid IM spectra for WT and E46K α-syn that were acquired using online SEC with the CIU² conditions used for quantification in Figure 3. D) Calibration curve generated by plotting the area of the first peak in panel C as a function of the amount of E46K in mixtures containing both WT and E46K α-syn. E) Calibration curve generated by plotting the area of the first peak in Figure S3 as a function of the amount of A53T in mixtures containing both WT and A53T α-syn. F) Calibration curve generated by plotting the ratio of the area of the first peak to the area of the second peak in Figure S3 as a function of the amount of pS129 WT in mixtures containing both WT and pS129 WT α-syn. Red points in panels D–F represent the measurement of unknown samples prepared in aCSF at 20 nM 30 % and 50 % variant α-syn (these points are overlaid on top of the calibration and were not used in the linear regression). Error bars indicate the standard deviation of three technical replicates (some error bars are too small to be visible).

While we have demonstrated that quantitative information could be obtained for α-syn variants from MS-compatible AmAc solutions, biological matrices such as aCSF contain non-volatile salts that would suppress signals for α-syn. We therefore used an online SEC separation coupled with a conventional ESI source equipped with a 50 μm i.d. ESI probe to automate both sample introduction and desalting processes prior to CIU. A chromatogram from an injection of WT α-syn diluted in aCSF onto a prototype 1 × 50 mm, 250 Å SEC column is shown in Figure 4B. Baseline resolution was not achieved between α-syn and salts present in the sample matrix using the prototype column. However, we chose to use this column because it allowed for rapid separations at low flow rates and minimized secondary interactions with the column.⁶⁶

IM spectra for 5 mL 20 nM WT and E46K α-syn that was preconcentrated and desalted with online SEC are shown in Figure 4C. The same IM conditions were used as in Figure 3. Next, the quantitative accuracy of the CIU² method with sample preconcentration and online desalting was assessed. A set of standards were prepared by mixing WT and E46K α-syn in aCSF at different ratios (the total concentration of α-syn in each mixture was 5 μM), and the standards were directly injected for online SEC-CIU² analysis. Two “unknowns” were prepared in the same manner, but with 20 nM of total α-syn at ratios of 30 % and 50 % E46K. The unknown samples served to mimic samples from organisms carrying the E46K mutation, where both WT and E46K α-syn coexist.³⁸ The unknowns were concentrated to approximately 50 μL from a 5 mL initial volume and then analyzed by SEC-CIU² in the same way as the standards. Figure 4D shows the calibration curve for WT and E46K α-syn that was generated from online SEC-CIU² experiments. The results from the analysis of the unknown samples are overlaid on the calibration curve in red. The online SEC-CIU² method was able to measure the relative abundance of the E46K variant within 3 % of the prepared ratio. The online method yielded an LOD of 4.4% E46K mutant.

Finally, we applied the same conditions that were used to quantify the E46K variant to two other α-syn variants that have been implicated in disease: the A53T mutant and WT α-syn phosphorylated at the 129^th serine. While these variants exhibit greater differences in molecular weight compared to the WT than the E46K mutant, adducts with salts present in biological samples could still interfere with unadducted protein signals and make quantification challenging, even with desalting steps. Standards and unknowns for A53T and pS129 α-syn were prepared and analyzed as described above for E46K α-syn. Figures 4E and 4F show the results of the analyses of A53T and pS129 α-syn. A53T α-syn could be quantified using the first peak in the IM spectrum, as with E46K. In the case of pS129 α-syn, the peak with maximum abundance changed with the ratio of WT to variant, so the ratio of the areas of the first two peaks were used for quantification. In this case, the CIU² method in Figure 3 that was developed to quantify WT and E46K CIU data also yielded quantitative information for A53T and pS129 α-syn, and again measured the relative amount of variant present in the unknown mixtures within 3% of the prepared ratio. The LODs for A53T and pS129 were 10.2% and 10.1%, respectively. Overall, these results demonstrate that CIU² analysis can be used to quantify disease-relevant variants of α-syn that are difficult to differentiate using other MS-based approaches.

CONCLUSIONS

Our cIM-MS enabled CIU² experiments, enabled through multiple stages of mobility selection and activation, can probe subtle differences in protein structure and leverage those differences for the purposes of structural proteoform quantitation. Here, CIU was used to obtain quantitative information regarding the relative abundance of disease-associated a-syn proteoforms present in solution which are difficult to differentiate using standard methods. While conventional CIU¹ experiments, which use just one stage of activation and no mobility selection, provided usable a-syn quantitative data, quantification with CIU² data improved calibration linearity, sensitivity, and limit of detection. Ultimately, these results demonstrate the utility of CIU as a quantitative MS technique for protein biomarker analysis that can be deployed to interrogate analytes that exhibit subtle differences in higher order structure.