Automated Immunoprecipitation, Sample Preparation, and Individual Ion Mass Spectrometry Platform for Proteoforms

Abstract

Charge detection mass spectrometry (CDMS) is a well-established technique that provides direct mass spectral outputs regardless of analyte heterogeneity or molecular weight. Over the past few years, it has been demonstrated that CDMS can be multiplexed on Orbitrap analyzers utilizing an integrated approach termed individual ion mass spectrometry (I²MS). To further increase adaptability, robustness, and throughput of this technique, here, we present a method that utilizes numerous integrated equipment components including a Kingfisher system, SampleStream platform, and Q Exactive mass spectrometer to provide a fully automated workflow for immunoprecipitation, sample preparation, injection, and subsequent I²MS acquisition. This automated workflow has been applied to a cohort of 58 test subjects to determine individualized patient antibody responses to SARS-CoV-2 antigens. Results from a range of serum donors include 37 subject I²MS spectra that contained a positive COVID-19 antibody response and 21 I²MS spectra that contained a negative COVID-19 antibody response. This high-throughput automated I²MS workflow can currently process over 100 samples per week and is general for making immunoprecipitation-MS workflows achieve proteoform resolution.

Graphical Abstract

For decades, conventional mass spectrometers have produced spectral outputs containing the mass-to-charge (m/z) ratio of detected ions.¹ In cases where defined charge state peak distributions are visible, the mass of the analyte or multiple analytes in question can be reliably determined.^2–5 However, as sample complexity has increased the need to comprehend complex proteoform mixtures in mass spectrometric analysis without well-defined charge states has become necessary. To alleviate sample complexity and charge state congestion, separation techniques are often used such as liquid chromatography,⁶ capillary electrophoresis,⁷ and ion mobility.⁸ These approaches distinctly separate analytes in the time domain, but remain limited by the potential for analyte populations emerging at a given time point being overly complex in m/z space. Addressing this problem, direct charge assignment on each ion in m/z space can be determined with charge detection mass spectrometry (CDMS).^9–11

To simplify analysis, CDMS produces a spectral output directly in the mass domain, eliminating the need for discernible charge states to calculate observed mass values and enabling the generation of spectral data from samples that are otherwise prohibitively complex.^12–15 In the past, CDMS has only been available on specialized instruments simultaneously measuring the m/z and charge (z) of individual ions within an electrostatic linear ion trap. Recently this was adapted to commercial mass spectrometer configurations, with the Orbitrap analyzer shown capable of CDMS and published as individual ion mass spectrometry (I²MS) analysis.⁹ To date, I²MS has been deployed for applications including the intact analysis of virus-like particles (VLPs),¹⁶ detection and identification of proteoforms from a ~1000 proteoform mixture,¹⁰ and imaging of proteoforms in human tissues.^10,17 The I²MS technique also set the standing record for isotopic resolution at high mass (466 kDa).¹⁸

Previously, we used I²MS to visualize and characterize anti-SARS-CoV-2 antibody repertoires in the plasma of COVID-19 convalescent and vaccinated subjects using a new serology platform named immunoglobulin-mass spectrometry (Ig-MS).^19,20 Ig-MS is a method for isolating, visualizing, and characterizing patient antibody repertoires against a defined antigen of interest. This is achieved by covalently immobilizing purified antigen on a solid substrate (e.g., magnetic beads) to be used as bait to selectively bind responsive immunoglobulins in patient plasma and/or serum samples. Following binding, unbound and nonspecifically bound species can be washed away, after which antigen-bound antibodies are retained for subsequent elution. Eluted material can be fully reduced to separate isolated antibodies into their light- and heavy chains, which can be prepared and infused into an electrospray source for I²MS data acquisition. This approach is uniquely capable of capturing the relative abundances of individual antibody clones in a subject’s immune response to an antigen, improving upon other established technologies that only establish the presence of antibodies against an antigen (e.g., lateral flow assays) or quantify total antibody titers (e.g., ELISA).^21–24

Published workflows for Ig-MS are semiautomated, with the antibody pulldown automated via the KingFisher platform and I²MS acquisitions automated by a PAL3 robot via the Xcalibur software suite.^19,20,25 Conversely, sample reduction and buffer exchange into electrospray-compatible buffer are performed manually in existing protocols. Notably, the methanol-chloroform-water precipitation approach used for buffer exchange is both laborious and prone to high variability, limiting both the throughput and fidelity of the entire workflow.²⁶ In this work, we integrated automated IP, sample handling, injection, and automatic ion control (AIC, the equivalent of automated gain control for I²MS)²⁷ to create Auto-Ig-MS, the first workflow for unattended sample preparation and CDMS analysis (Figure 1). We demonstrate how numerous technologies connect through the Xcalibur sequence creation software to produce mass domain spectra for a sample cohort with minimal intervention. We apply Auto-Ig-MS to rapidly characterize anti-SARS-CoV-2 antibody repertoires in a cohort of 58 non-symptomatic individuals, 27 of which tested positive for COVID-19 prior to sample collection.

Figure 1.

Schematic of Auto-Ig-MS, a full automation workflow that joins immunoprecipitation with the KingFisher, sample preparation and injection with the SampleStream platform, ion attenuation with AIC and subsequent charge detection analysis of individual ions via I²MS into a single fully automated pipeline. This technique was applied to a cohort of 58 possible COVID-19 positive subjects.

EXPERIMENTAL SECTION

Human Subject Authorizations and Plasma Sampling.

Serum samples that were obtained from the Northwestern Medicine Healthcare Worker SARS-CoV-2 Serology Study (Northwestern Serology Study) were used for this study. The Northwestern Serology Study is an epidemiological study designed to explore the predictors of infection form SARS-CoV-2 in healthcare workers.²⁸ At enrollment, participants were administered electronic surveys to assess demographics (age, sex, race), health history, occupational category, and work- and nonwork-related exposure risk factors for SARS-CoV-2. The study also collected serum from participants between May 28, 2020 and July 8, 2020 by clinical teams located throughout the Northwestern healthcare system (IRB #STU00212515). The study used the high-throughput ARCHITECT i2000SR Immunoassay System from Abbott Laboratories (Abbott Park, IL, USA), to identify serum samples with IgG antibodies to SARS-CoV-2 nucleocapsid antigen. Remnant samples were stored at −80 °C for further use. Of the 6510 who enrolled in the cohort, 317 were positive for antinucleocapsid IgG at baseline. From this group of individuals who were positive for anti-IgG for nucleocapsid antigen at baseline, 37 were included in this study.

Additionally, a commercially obtained plasma sample from a convalescent individual collected at the beginning of the COVID-19 pandemic was used as a positive control (All-Cells). Pooled serum collected prior to the emergence of SARS-CoV-2 was used as a negative control/null background (Thermo Fisher Scientific, BP2657100 UNSPSC 12352207).

Expression and Purification of Spike-In Standard RBD Protein.

Wuhan SARS-CoV-2 spike protein receptor binding domain (RBD, residues 319–541) was expressed off of plasmid pCAGGS SARS-CoV-2 RBD²⁹ (BEI Resources, BEI NR-52309) in the Expi293 expression system followed by purification on AKTAxpress (GE Healthcare Life Science) FPLC purification system, as previously described.^19,20

Fabrication of Magnetic Beads Conjugated with SARS-CoV-2 Spike RBD.

Production of magnetic RBD bait beads was described previously.^19,20 Briefly, for each lot of beads 0.5 mg of recombinant Wuhan RBD were covalently bound to 500 μL of MyOne carboxylic acid Dynabeads (Thermo Fisher Scientific) following manufacturer instructions. Beads were washed twice with 0.5 mL of 25 mM MES, pH 6.0 for 10 min at room temperature and activated by sequentially adding 500 μL of 50 mg/mL N-hydroxysuccinimide (NHS) and 500 μL of 50 mg/mL 1-ethyl-3-(3-(dimethylamino)propyl) carbodiimide (EDC) followed by a 30 min incubation at room temperature. Following activation beads were washed twice with 0.5 mL 25 mM MES, pH 6.0 for 10 min at room temperature and resuspended into 0.5 mL 25 mM MES, pH 6.0. 0.5 mg of purified RBD protein was added to the bead suspension and the mixture was incubated for 30 min at room temperature. Excess RBD was removed and discarded and the beads were quenched via incubation with 1 mL of 50 mM Tris, pH 6.8 for 15 min at room temperature. Beads were washed 4 times with 0.5 mL 1× PBS + 0.1% human serum albumin (HSA) and finally resuspended into 1 mL 1× PBS + 0.1% HSA for storage at 4 °C.

Isolation and Enrichment of SARS-CoV-2-Specific Antibodies from Patient Samples.

The preparation and completion of patient SARS-CoV-2 antibody pulldowns have been described in detail previously.^19,20 Briefly, recombinant Wuhan RBD was covalently loaded onto commercialized magnetic beads following manufacturer protocols. Each pulldown was assembled by combining 100 μL of patient plasma with 35 μL of RBD-loaded bead suspension and diluting to a volume of 1 mL with 1× TBS. Assembled pulldowns were incubated overnight at 4 °C with end-over end mixing, after which beads were pulled down, supernatants were removed, and beads were resuspended into 1 mL wash buffer (1× TBS + 0.1% Tween +1% NP-40 + 1% NP-40 substitute). Suspensions were transferred into a 96-deep-well plate and loaded into a KingFisher Flex (Thermo Fisher Scientific) which automatically performed an additional 4 washes in 1 mL wash buffer followed by 2 washes in 1 mL 1× TBS and a 30 min incubation in 100 μL 100 mM glycine, pH 11.5 at 37 °C to elute antibodies bound to bead-bound RBD.

Automated Plate Transfer and Sample Preparation for Mass Spectrometry.

Reduction Plate Assembly.

To achieve complete denaturation and reduction of isolated antibodies for MS analysis, prior to the antibody pulldown a reduction plate was prepared by dispensing 160 μL 8 M urea, 25 μL 1 M TCEP, and 100 ng of spike-in mAb (CR3022) into wells of a standard 96-well plate corresponding exactly to the wells being used for the pulldowns on the KingFisher Flex. To minimize volume loss due to evaporation, the reduction plate was covered using a pierceable adhesive seal (Research Products International, Item #202515).

First Plate Rearrangement.

Following antibody isolation and enrichment by the KingFisher, a PAL3 robot enabled automatic plate manipulation to facilitate further preparatory steps. First, the robot arm released the LCP syringe tool typically equipped for liquid pipetting. Next, the arm picked up a Grabber D885-SL1 tool, allowing it to interface with and pick up Grabber DW-96. With the grabber equipped, the PAL3 robot arm then lowered into the KingFisher, extracted the elution plate containing patient-derived SARS-CoV-2 antibodies, and deposited it onto a 96-well magnet in a refrigerated Peltier stack drawer. After transferring the plate, the robot arm dropped off the grabber, released the Grabber D885-SL1 and re-equipped the LCP syringe tool.

Sample Reduction.

To perform the denaturation and reduction of all samples, the PAL3 robot next used the LCP syringe tool to individually extract the entirety of each elution fraction (~70 uL) and deposit it into the corresponding well in the reduction plate, washing the needle and syringe between each sample to prevent cross-contamination. After all samples had been transferred in this way, they were incubated in the reduction cocktail for 1 h at room temperature to allow denaturation and reduction to proceed to completion.

Second Plate Rearrangement.

As before, the PAL3 robot arm released the LCP syringe tool and picked up Grabber D885-SL1 allowing it to interface with Grabber DW-96. Using the grabber, the arm picked up the now-empty elution plate and deposited it on an external rack. Then, it picked up the reduction plate and deposited it onto the 96-well magnet in the refrigerated Peltier stack drawer.

Automated Sample Cleanup and Buffer Exchange.

Samples were buffer exchanged into electrospray-compatible system buffer (70% water/30% acetonitrile/0.2% formic acid) using the SampleStream platform, as described previously.^20,25 Briefly, for each sample 70 μL of material was extracted from the reduction plate by the onboard PAL3 robot equipped with an LCP syringe tool and injected into the SampleStream flowcell. Samples were focused onto a 5 kDa MWCO regenerated cellulose membrane twice with 250 μL of system buffer at a flow rate of 100 μL/min, and eluted from the flowcell in a final volume of 80 μL at a flow rate of 80 μL/min. Eluted material was drawn back into the LCP syringe tool and deposited into a clean vial.

I²MS Data Acquisition.

Direct injection and I²MS data acquisition was automated using SampleStream. In brief, 70 μL of buffer exchanged sample was picked up by the PAL3 robot, delivered to a Q Exactive HF mass spectrometer (Thermo Fisher Scientific) at a flow rate of 1.2 μL/min, and sprayed through an Ion Max Source fitted with a HESI II probe. Upon sample injection, a process coined Automated Ion Control (AIC) attenuated ensemble ion signals down to an ion injection time (1–400 ms) corresponding to optimal individual ion collection on a per-sample basis to maximize the number of individual ion signals without producing a large number of multiple ion events at singular m/z values.²⁷ AIC has been reported previously, which operates on a continuous, per-scan basis similarly to Automated Gain Control (AGC).²⁷ Other instrument parameters included: eFT, off; Orbitrap central electrode voltage, 1 kV; trapping gas pressure, 0.3; spray voltage, 2.9 kV; sheath gas, 0 L/min; in-source CID, 15 eV; source temperature, 320 °C; m/z acquisition range, 650–2500 m/z; resolution 240,000 @ 200 m/z (2 s acquisition times, ~1638 transient scans collected over 55 min.).

Automation Software Compatibility.

Automation routines were interfaced directly through Xcalibur (Thermo Fisher) software analogous to traditional LC/MS workflows. To allow for direct control of both KingFisher (ThermoFisher) and SampleStream (Integrated Protein Technologies Inc.) module functionalities to be seamlessly linked to Xcalibur, a custom PAL plugin interface (Integrated Protein Technologies Inc.) was created as shown in Figure S1. Via the plugin, custom KingFisher scripts were uploaded and executed step-by-step in a continual fashion. Similarly, SampleStream membrane channel temperature control and pressure reading monitoring during protein focusing and subsequent elution out of the chamber were performed by the plugin.

Data Processing.

Selective temporal overview of resonant ions (STORI³⁰) plot analysis coupled with a voting charge assignment algorithm was conducted using STORIBoard (Thermo Fisher Scientific) to determine the charge for each individual ion collected, as previously described.^9,30 In brief, for each ion the slope of individual ion signal accumulation as a function of its specific frequency was calculated. Individual ion slopes were used to determine the charge for every individual ion signal detected via the STORIBoard voting charge assignment algorithm.

Calculation of Ig-MS Ion Titer, Degree of Clonality, and Antibody Diversity.

Ion titer (IT), degree of clonality (DoC), and antibody diversity (AD) were calculated as previously reported.^19,20

Data and Statistical Analysis.

Average response values across replicates for ELISA, IT, DoC, and AD (n = 1–11) were analyzed alongside the Serology Test Index and Wilkins Titer as analytical values for each patient (“statistical data”). Additionally, each patient reported body mass index (BMI), age, sex, high blood pressure status, immunocompromised status, and obesity status (“physical data”). Binary patient traits were converted to a binary numerical form: statuses for various indications (1 for yes and 0 for no) and sex (1 for female and 0 for male). All values were normalized on a scale from 0 to 1 before analysis, and all analytical values were normalized after being converted to log10. For the purposes of this study, a chemiluminescent ELISA against SARS-CoV-2 nucleocapsid²⁸ was interpreted as the “ground truth” for the presence/absence of α-SARS-CoV-2 Igs.

To visualize higher correlations in the data, we conducted a principal component analysis (PCA) using the singular value decomposition function (svd) in MATLAB 2021b and plotted the data along the first two principal components. PCA was conducted on the following batches of data: all data, only the physical data, and only the statistical data. These three batches were analyzed for both all patients and only patients deemed positive for COVID-19 infection. For the all-patient analysis, data points were manually formatted to denote positive or negative COVID-19 infection, and for the positive-only analysis, data points were manually formatted to denote the patients with above-median ion titer and below-median ion titer.

Data and Software Availability.

RAW files and processed data sets can be found on the MassiVE repository, MSV000094517. Total ion current plots and m/z spectra were taken from Thermo Xcalibur Qual Brower (v. 4.1.31.9), and mass spectra were visualized in mMass (v. 5.5.0). Data analysis plots were created with either MATLAB R2021b or Microsoft Excel, and figures were created using Adobe Illustrator and Microsoft PowerPoint. Additional desired software and data that support the findings of this study are available from the corresponding authors upon request.

RESULTS AND DISCUSSION

We set out to develop an automated platform for CDMS sample preparation and data acquisition. Conceptually, this required automating and coordinating the activities of several pieces of instrumentation, including a KingFisher Flex (for automating target enrichment protocols), a PAL robot equipped with a SampleStream system (for automating sample buffer exchange and infusion into a mass spectrometer), and a Q Exactive HF mass spectrometer (for individual ion detection). To achieve this, we equipped the PAL3 robot with a Grabber DW-96 tool to allow the robot arm to grab and pick up standard 96-well plates, enabling automation of the plate rearrangements necessary to transfer sample plates between the other instruments involved in the pipeline. Facile deployment of automation routines was enabled by seamlessly linking all instrument functionalities to Xcalibur via a custom PAL plugin interface (Figure S1). These software advancements worked in concert with the instrumentation modifications described above to yield Auto-Ig-MS, which is to our knowledge the first fully automated platform for CDMS analysis (Figure 1). Using this platform, were able to submit plates of plasma samples to a KingFisher instrument, execute a sequence in Xcalibur (Figure S2), and the instrumentation processed samples through all preparatory steps and MS data acquisition without the need for human intervention (Video S1). In this way, current implementations operating continuously can process over 100 samples per week.

We first applied the automated platform to CR3022, a single IgG clone known to bind to SARS-CoV-2 spike RBD.²⁹ These positive control samples were assembled by spiking 50 ug of purified antibody into null human plasma collected prior to the emergence of SARS-CoV-2. Figure 2a shows representative I²MS spectra generated when these samples were subjected to Ig-MS. The spectra show that CR3022 was reproducibly captured from the plasma samples by the antigen beads, as expected, and can be clearly observed in the spectra alongside low levels of nonspecifically bound species at slightly lower mass values (22,500–24 100 Da). To monitor longitudinal performance and stability of the assay, these positive control samples were run intermittently throughout subject samples during this study. Ion titer data from the nonspecifically bound background material present in these positive control samples were calculated as previously described and applied toward the determination of thresholds by which positive (i.e., SARS-CoV-2 antibodies present) versus negative (i.e., SARS-CoV-2 antibodies absent) responses could be unambiguously identified among patient cohort spectra to make the assay diagnostic for infection with SARS-CoV-2. Figure 2b shows the ion titer values corresponding to the background material present in each of the positive control samples run over the course of this study. Looking at these data in aggregate, we determined the average value (Figure 2b, yellow line) as well as the 95% confidence threshold (Figure 2b, red line). In this way, we defined as positive any assay result with an ion titer higher than the 95% confidence threshold, while results with ion titer values less than this threshold were considered negative.

Figure 2.

Examples of (a) 3 positive control sample LC mass distributions containing intense spiked-in CR3022 standard and low level nonspecific binding distributions from pre-COVID 19 patient plasma. (b) Nonspecific binding titers for all 25 positive controls acquired throughout the COVID 19 patient cohort to determine average nonspecific binding titer (yellow solid line) and 95% confidence (red line) values creating the titer boundary between a positive vs negative COVID 19 patient LC antibody response.

Next, we applied the platform to fully automate preparation and I²MS data acquisition from plasma samples derived from a cohort of 58 healthcare workers with routine occupational exposure to the virus. Figure 3 depicts results from a single representative acquisition. I²MS acquisitions were standardized to ~55 min in duration and comprised 1638 transient scans collected every 2 s. A typical total ion current profile trace is shown in Figure 3a. Ensuring collection of individual ions requires vigilant attenuation of ion density entering the orbitrap during each scan. This is typically achieved by diluting samples down to sub-nM concentrations and restricting orbitrap ion injecting time. Here, our approach emphasized the latter strategy, and employed an algorithm called automated ion control (AIC) to restrict ion injection time and attenuate ion density down to the individual ion regime.²⁷ Figure 3b shows a scan collected prior to AIC attenuation, at an ion injection time typical of ensemble measurements. At this unregulated injection time, ensembles of enriched patient light- and heavy chain repertoires are collected. Figure 3c shows a scan from later in the same acquisition after correction by AIC. As a result of AIC attenuation, ion inject time has been significantly reduced, and as a result the raw spectrum now features easily recognizable individual ion “shelves” corresponding to populations of antibody light- (Figure 3c, blue line) and heavy (Figure 3c, green line) chains from the sample. Figure 3d shows the final composite I²MS spectrum from the acquisition. As previously demonstrated, isotopic distributions are present from ~23–25 kDa that represent distinct light chain clones; similarly, signal corresponding to heavy chains is also observed at approximately 50 kDa.^19,20 In this way, I²MS spectra were generated characterizing the anti-SARS-CoV-2 antibody repertoires from all 58 individuals being investigated.

Figure 3.

Example run of a patient sample including (a) total ion current profile including AIC procedure time followed by data collection at the optimized individual ion injection time, (b) pre-AIC acquisition scan with an ensemble of patient light chain (LC) and heavy chain (HC) antibody responses, (c) post-AIC acquisition scan revealing individual ion distributions of LC (blue solid line) and HC (green solid line) antibody responses, and (d) composite I²MS spectrum of densely populated LC and HC regions.

Next, antibody ion titers and clonality metrics were calculated from the light chain region of each subject’s I²MS spectrum as previously described.^19,20 We applied the titer threshold established using the positive control samples to each result, and by this metric determined that 37 of the 58 individuals investigated were positive for the presence of anti-SARS-CoV-2 immunoglobulins while the remaining 21 were negative. Figure 4 shows exemplary representative examples of positive and negative subject spectra. The I²MS spectrum for the subject in Figure 4a is visibly sparse outside of the spiked-in CR3022 standard. Unsurprisingly, the ion titer calculated for this was quite low making it a clear negative result. In Figure 4b, the I²MS spectrum features a highly abundant signal outside of the spiked-in standard. The calculated ion titer value is high enough to qualify this as a positive response, and the low antibody diversity indicates that this subject’s immune response to SARS-CoV-2 infection was monoclonal. Finally, the I²MS spectrum in Figure 4c is highly populated suggesting a robust and complex immune response to SARS-CoV-2. Supporting this notion, the ion titer and antibody diversity metrics calculated for this spectrum are both highly elevated. Importantly, the averaged raw spectra accompanying the I²MS spectra in Figure 4 are far too complex to extract mass information from using established deconvolution algorithms, highlighting a main strength of our individual ion approach. Overall, the group of positive subjects identified by Auto-Ig-MS included 19 out of the 27 individuals that tested positive for COVID-19 prior to sample collection. We attribute this discrepancy to timing considerations. A subject in the very early stages of COVID-19 can test positive for infection but not yet have enough time to mount a robust immune response. Conversely, a convalescent subject can possess α-SARS-CoV-2 Igs while no longer testing positive for the presence of the virus. Correlating serological data to infection status requires careful consideration of disease progression.

Figure 4.

Ensemble (left) and corresponding I²MS light chain (LC) region spectrum (right) for three different subjects. Possible COVID-19 LC antibody responses encompass a range of results including (a) a negative titer and monoclonal response, (b) a positive titer and monoclonal response, and (c) a positive titer and polyclonal response. The spiked-in CR3022 antibody standard utilized as a reference to calculate titer values is boxed in red.

We next sought to establish correlations between our Ig-MS metrics and subject health metrics such as BMI, age, and sex. We first investigated such correlations using results from the entire subject cohort used for this study. Figures S3–S5 show Ig-MS metrics plotted against subject BMI, age, and sex, respectively. No strong correlations (|R²| ≥ 0.7) were observed in any of these comparisons. We also plotted Ig-MS metrics from the entire cohort against one another in a pairwise fashion to see if any of them correlated with one another in a meaningful way (Figure S6). Here, we did observe correlation between ion titer and antibody diversity. This correlation makes sense, as an increase in the number of observable antibody clones in a sample will necessarily contribute to an increase in the titer of that sample. Finally, we performed principal component analysis on data from the entire cohort as shown in Figure S7. By this analysis, samples did not cluster together in any significant way.

We reasoned that stronger correlations between Ig-MS metrics and subject health metrics may be observed from subjects that qualified as positive for infection with SARS-CoV-2 according to our assay, so we repeated our correlation analysis including data only from the 37 positive subjects, with results presented in Figures S8–S12. When including only data from positive subjects, no correlations between Ig-MS metrics and health metrics were observed in this epidemiological study. Figure S11 shows a modest correlation between Ig-MS ion titer and antibody diversity in the positive subject cohort, though interestingly it is weaker than the correlation observed when data from the entire cohort are included. In the future, a study with larger scale should be performed to uncover correlations that can be appropriately applied to a wide population.

Finally, we further explored the correlation between ion titer and antibody diversity. Figure 5a shows a plot of antibody diversity versus ion titer using the raw, unmodified values for each metric calculated for each subject in this study. When the data were plotted in this way, a small number of samples with extremely high ion titers distorted the correlation, negatively impacting the R² value for the resulting trendline. When the log10 of these metrics are plotted against each other instead (as in Figure 5b), the correlation between them improves considerably; furthermore, when the data are presented in this way the graph is effectively divided into 4 quadrants describing the positivity and clonality of subject antibody repertoires against SARS-CoV-2 (according to the metrics established above) that enables the data to be readily interpreted at a glance. When arranged this way, subject immune responses generally sort into “Negative/Monoclonal” or “Positive/Polyclonal”, which makes sense intuitively. Additionally, this presentation of the data enables facile detection of monoclonal responses, which could accelerate discovery of efficacious candidates for therapeutic antibody drug development. The pipeline presented here is uniquely capable of performing this sort of analysis, as Ig-MS is the only approach capable of quantifying the clonality of an individual’s response against a given pathogen at the level of circulating antibodies.

Figure 5.

(a) Plot of antibody clonality vs titer for the 58 patient cohort processed through our automated workflow. The log plot (b) of the same points shows a correlation between these responses and a wide range of negative/positive and monoclonal/polyclonal antibody response values.

The results presented here serve as proof-of-principle and provide output examples of fully automated sample preparation and data acquisition via Auto-Ig-MS. This pipeline takes advantage of the many accessories compatible with the PAL robot platform in combination with a custom plugin to enable coordinated operation of several instruments using Xcalibur, improving on previous workflows by replacing variable and error-prone human handling with the robustness and precision of robotics. This platform also enables increased productivity, as the robotic instruments can continue to work around the clock to yield continuous data generation.

Importantly, while SARS-CoV-2 was the pathogen of interest in this study, Auto-Ig-MS is highly adaptable. By simply exchanging the antigen loaded onto the magnetic beads for use in the enrichment step, the platform can be direct to a different target; indeed, the assay can be target to virtually any pathogen provided that its associated antigens can be produced recombinantly in a sufficiently pure form. Researchers now have the capacity to very quickly obtain genetic sequence information for emerging pathogens – SARS-CoV-2, for instance, was fully sequenced within weeks of the virus’ first appearance.³¹ Rapidly acquired genomic data enables rapid synthesis of pathogen-derived proteins. Thus, Auto-Ig-MS could be deployed promptly in response to future novel pathogens to elucidate patient immune responses in real-time, positioning the approach as a powerful tool for combating the next pandemic and a valuable companion assay to established technologies such as ELISA and lateral flow assays. It is also worth noting that the pipeline can be easily adapted for analysis of material enriched via a normal immunoprecipitation by simply immobilizing antibodies against a target protein of interest in the magnetic beads for use in the automated KingFisher pulldown step. In such a case, the enriched target protein of interest will be eluted from the beads and subjected to buffer exchange and I²MS analysis to generate a proteoform landscape for the target. By clever design of the immunoprecipitation step, the platform described here can be readily adapted for many applications.

CONCLUSIONS

We present Auto-Ig-MS, a fully automated platform for the isolation, buffer exchange, and mass spectrometric data acquisition of patient-derived antibodies against SARS-CoV-2 that is, to our knowledge, the very first fully automated platform for target enrichment and subsequent CDMS analysis to be described. A single installation can prepare and process over 100 samples a week with no requirement for human monitoring or other intervention. As proof-of-principle, this platform was applied to the automatic preparation and I²MS analysis of antibody repertoires generated against SARS-CoV-2 in a cohort of 58 subjects. Importantly, this approach is entirely modular, and can be easily applied to other pathogens of interest or slightly modified for use with regular immunoprecipitations.