Advancements in Automation for Plasma Proteomics Sample Preparation

Abstract

Automation is necessary to increase sample processing throughput for large-scale clinical analyses. Replacement of manual pipettes with robotic liquid handler systems is especially helpful in processing blood-based samples, such as plasma and serum. These samples are very heterogenous, and protein expression can vary greatly from sample-to-sample, even for healthy controls. Detection of true biological changes requires that variation from sample preparation steps and downstream analytical detection methods, such as mass spectrometry, remains low. In this mini-review, we discuss plasma proteomics protocols and the benefits of automation towards enabling detection of low abundant proteins and providing low sample error and increased sample throughput. This discussion includes considerations for automation of major sample depletion and/or enrichment strategies for plasma toward mass spectrometry detection.

Graphical Abstract

2. Introduction

Plasma proteomics^1-9 has helped identify biomarkers to detect the presence of a variety of diseases, including cancer¹⁰, Alzheimer’s disease^11-15, and sepsis.^16-24 Plasma is minimally invasive, relatively simple to obtain (in comparison to organs or cerebrospinal fluid), and rich in information. Plasma is mostly comprised of water (~90%), but includes other components, such as tissue leakage and interleukin proteins, mineral salts and ions, carbohydrates, nucleic acids, and metabolites.^{1, 25, 26} Plasma is collected by adding anti-coagulants (e.g., ethylenediaminetetraacetic acid (EDTA) and lithium-heparin) to whole blood followed by centrifugation.²⁷ The complexity of plasma presents significant analytical challenges. For example, the dynamic range of plasma protein concentration is greater than ten orders of magnitude, ranging from serum albumin at mg/mL levels to interleukin 6 at pg/mL levels (Figure 1), which requires front-end sample preparation strategies prior to mass spectrometry (MS) analysis.¹

Figure 1:

Dynamic range of plasma. Plasma proteins, excluding hemoglobin that is a component of red blood cells, have a range of ~ 10 orders of magnitude, with the most abundant protein (i.e., albumin) containing ~55% of the proteome. Reproduced with permission from ref. 7.⁷

Bottom-up proteomics techniques^11-24 are the most widely adapted analyses for plasma. Scaling bottom-up proteomics experiments to analyze hundreds of plasma samples, can, however, become tedious, time-consuming, and error-prone, as there are many liquid-transfer steps.²⁸ In bottom-up proteomics workflows, plasma samples are depleted or enriched, and protein concentration is determined by using an assay, such as bicinchoninic acid (BCA), Bradford, or Lowry assay. Protein concentration assays are used to determine the total protein mass in a sample so that 1) a specific mass can be digested, and 2) the corresponding amount of trypsin can be added (typically added in a specific protein mass:trypsin mass ratio). In our experience, protein content varies significantly from patient to patient with plasma samples and offers a better normalization than volume alone.

After determining the protein concentration, proteins are digested using the enzyme trypsin or a mixture of trypsin/Lys-C, producing lysine and arginine-terminated peptides at the C-terminus. Peptides generated from protein digestion are desalted to remove interfering species/chaotropic agents prior to reversed-phase high-performance liquid chromatography – tandem mass spectrometry (rpHPLC – MS/MS) analysis (Figure 2). Quantitative options can rely on either label-free or peptide chemical tagging strategies [e.g., Isobaric Tag for Relative and Absolute Quantitation (iTRAQ)²⁹ or Tandem Mass Tags (TMT)^30-33], which enable multiplexing at the expense of adding additional desalting steps and sample loss. Enhancements to proteome depth can also be incorporated to the workflows through the addition of peptide fractionation steps, including strong cation exchange (SCX) or high pH reversed-phase fractionation. The totality of steps in these workflows can result in higher sample variation than is desirable, especially in the context of clinical diagnostics which must provide accurate, reliable, reproducible, and precise information.^34-37

Figure 2:

Overview of bottom-up plasma proteomics protocol used to process plasma samples. Plasma is either pre-processed by depletion or enrichment or maintained as a crude sample. Typically, protein concentration is determined via an assay such as BCA, Bradford or Lowry assay so that a certain protein mass is digested with trypsin and then subjected to instrumental analysis. In some instances, peptide samples are labeled by chemical reagents and/or subjected to further separation by using fractionation techniques. Figure was created with Biorender.com. Abbreviations: HAP = high abundant proteins; LAP = low abundant proteins; iTRAQ = isobaric tags for relative and absolute quantification; TMT = tandem mass tags; SCX = strong cation exchange; RP = reverse-phase; LC-MS/MS = liquid chromatography-tandem mass spectrometry.

There are several well-written reviews that discuss basics of plasma^1-3 , plasma processing with proteomics techniques^2-7, and challenges of plasma analysis.^7-9 However, there is limited discussion of the benefits and complexities of incorporating automation for bottom-up proteomics workflows of plasma samples. Automation of many steps in the typical bottom-up proteomics workflow has great potential to advance relevant biological or chemical problems which require processing hundreds or thousands of samples. In this mini-review, we detail factors that need to be considered in order to successfully prepare plasma samples for MS analysis using automation including plasma prefractionation, protein concentration assays, sample multiplexing, and fractionation.

3. Automated Platforms and Protocols for Plasma Proteomics

3.1. Automated LC Systems

LC systems have been implemented to automate plasma proteomics workflows. Guryča et. al. successfully developed an online depletion platform using a LC system containing a quaternary pump, two six-port valves, an autosampler, diode array detector, and two column ovens.³⁸ The Top 14 most abundant proteins were depleted on a Seppro IgY14 immuno-depletion column (now discontinued) and depleted plasma flowed from the first to the second six port valve, where protein was immobilized onto a trap column. Once captured, protein was reduced, alkylated, and desalted. Desalted protein was eluted from the trap system using an acetonitrile gradient and a separation column was used to generate 10 fractions. Fractions were then digested off-line with trypsin and analyzed using LC – MS/MS. This approach, while highly accessible to labs with LC systems, has not been widely applied.

3.2. Challenges Associated with Using Automated LC Systems

Automation is extremely beneficial in increasing sample throughput while decreasing overall error and sample-to-sample variation, however there are some challenges in using an automated LC system. After sample depletion, the volume of the unbound fraction dilutes the amount of detectable protein using a protein concentration assay. Typically, plasma samples are concentrated to smaller volumes offline using centrifugation and automation of this step can facilitate larger number of samples being handled simultaneously. In addition, although the sensitivity improves by increasing the number of fractions in the strategy explained in this section, a major caveat is that sample-throughput is reduced since it can take months to process small numbers of fractions without automation.

3.3. Robotic Liquid Handlers

Over the past two decades, there have been tremendous advances to automate proteomics experiments. Several biotechnology companies have introduced a variety of sophisticated robotic liquid handler platforms (Table 1). These systems can be integrated with 3^rd party devices/components, including temperature controllers, vacuum manifolds, positive-pressure apparati, plate sealers, plate readers, incubators, and storage systems. New technologies have also been introduced for contact-free liquid transfer from 384-well and 1536-well microplates using sound energy.^{39, 40} These transfer strategies utilize acoustic droplet ejection technology (Beckman Coulter) to enable as low as 2.5 nL of volume to be accurately transferred. Implementing robotic liquid handlers into proteomics workflows have helped streamline experiments, increase throughput, and reduce sample-to-sample variation among large sample-sets.^{4, 27, 41-48} The major limitation with the use of the liquid handlers is the cost of the instrument and their integrations. In addition, these robotic liquid handlers use tips and calibration solutions that are vendor specific and not compatible with consumables from other vendors. This makes the user dependent on the vendor and may lead to challenges obtaining consumables during supply shortages. The benefits of robustness, reproducibility, and time consumption for large-scale efforts can justify the costs ($5,000 – $300,000) of robotic platforms. However, these instruments require highly sophisticated and skilled users, so there will likely be start-up time to teach users to develop and implement protocols, perform routine maintenance, and troubleshoot potential issues that may arise during operation.

Table 1.

Liquid handlers for laboratory automation available for research.

Vendor	Instrument	Integrated apparatus	Throughput (# samples)	Pipetting volume	Desk capacity
Agilent Technologies	Assay Map Bravo	BenchCel, centrifuge, plateLoc, barcode labeler, miniHub, vacuum filtration station, peltier, orbital shaker and magnetic bead station	96-384	0.003 - 250μL	9
Beckman Coulter	Biomek i7, Biomek i5	Barcoding, vacuum, positive pressure, heating and cooling, shaking and incubating, plate reader, plate stacking, liquid level sensing	96-384	0.5 - 5000μL	45
Eppendorf	epMotion 5075vt	Thermal module, thermomixer, vacuum unit	8	0.2μL – 1mL	12
Gilson	Pipetmax®	Orbital Shaker	8	1 - 200 μL	9
Hamilton	Microlab Vantage, Microlab STAR, Microlab NIMBUS, Microlab Prep	Barcode reader, Internal plate gripper, heater shaker, positive pressure module, plate sealing, vacuum station, incubator shaker, heater cooler, thermal cycler, h motion	96-384	0.5 - 1000μL	72
Hudson Robotics	SOLO Liquid Handler, Micro 10x Robotic Benchtop 12-channel dispenser	Heating and cooling nests, incubators, magnetic bead nests, automatable centrifuges, microplate readers, plate crane robotic arm, and microplate stacker	12	500nL – 10mL	12
Opentrons	OT-2	Thermocycler module, temperature module, Magnetic module	96-384	1-1000μL	11
PerkinElmer	Janus, SciClone G3, Zephyr G3	PlateStak, plate readers and shakers	96-384	0.5 - 5000μL	32
Sirus Automation	OmniTasker	Vortex, sonicate, barcode scanning, serial dilution, weigh, cap/uncap, vial sorting, powder dispensing	8	2 -50mL	36
Tecan Freedom EVO	Fluent, Freedom EVO, Tecan D300e Digital Dispenser	Barcoding, Carousel, Tip storage, Heating and cooling, plate reader, mixing and shaking, vacuum SPE, Magnetic separation, Thermal Cycler, Sample fractionation, Colony picking, Centrifuge, Positive pressure	8-384/1536	0.0025 -125μL	72

Deck Capacity: Indicates the number of positions available.

Several groups have contributed to developing automated protocols for plasma proteomics projects.^{4, 27, 41-52} Few protocols (e.g., O-link, Simoa, and Roche) use fluorescent-based PCR detection methods^{52, 53}, while most protocols are intended for samples that will be analyzed via LC – MS/MS.

Automated liquid handlers have also been implemented to prepare plasma samples for instrumental analysis. For example, the microlab star liquid handler (Hamilton) has been used to test the scalable automated proteomics pipeline (ASAP², Table 1), developed by Dayon and coworkers.^{27, 41, 43, 44} Plasma is depleted offline (e.g., multiple affinity removal system, MARS-14), and online buffer exchange is performed using polymeric reversed phase cartridges and a vacuum manifold. Protein digestion, peptide desalting, TMT tagging, pooling, C18 desalting, and SCX purification are performed using a 96 well-plate format. Samples are then analyzed using LC – MS/MS. This protocol has successfully been applied to test EDTA vs. heparin anti-coagulant plasma²⁷ and analyze samples (N=1005) from the Pan-European human dietary intervention study Diet, Obesity, and Genes (DiOGenes).^{44, 54} More recently, the Biomek i7 (Beckman Coulter)⁴⁸ and Hamilton Vantage⁴² systems have also been used to automate plasma protocols (Table 1). Plasma samples were processed using a protocol which optimized pipetting and liquid transfer steps on the Biomek i7.⁴⁸ In six steps, 96 samples can be prepared for LC – MS/MS analysis in ~ 5 hours.⁴⁸ Single-pot solid phase enhanced sample preparation (SP3) and in-Stage Tip (iST) preparation methods were employed to prepare naked mole rat plasma samples.⁴² Unlike other protocols, crude samples were processed. After cell lysis and BCA protein assay, samples were digested, desalted, and peptides were labeled using TMT-pro reagents. Peptides were fractionated online using LC – High-field Asymmetric Ion Mobility Spectrometry (FAIMS) – MS/MS and LC – FAIMS-MS/MS with real-time search (RTS). For small volumes (~ 1 μL), Mann and coworkers have developed an automated protocol that uses undepleted plasma.⁴ The Plasma Proteome Profiling protocol is executed by obtaining whole blood samples (via a finger prick), transferring 5 μL of blood to a pipette-tip-based centrifugal device, isolating 1 μL of plasma (via centrifugation), and employing in-stage tips to perform protein digestion and desalting before LC–MS/MS analysis.⁴ This protocol was initially tested on samples obtained for diagnostic analyses in a clinical setting⁴, and has since been successfully applied to study the effects of sustained weight loss⁴⁶, biases in plasma biomarker studies⁴⁵, and non-alcoholic fatty liver disease.⁴⁷ Downstream in the workflow, high pH fractionation and concatenation were automated by the construction of the “spider fractionator,”⁵⁵ an eight-rotor valve fraction collector coupled to a nano-HPLC system. As peptides are separated under high pH conditions, the rotor valve switches every ~90 seconds to concatenate fractions.⁵⁵ Using the spider fractionator allows for smaller sample load (~20 – 60 μg) and reduced sample loss, as samples are collected in a “loss-less” manner.⁵⁵

3.4. Considerations Associated with Automation of Plasma Proteomics Workflows

There are many factors that must be considered when applying automated protocols to plasma samples. First, the goal of the project has to be very clear. One has to consider balancing sample preparation time with increasing sample throughput. A comprehensive literature search performed from 2012 to 2016 by Geyer et. al. discovered that most plasma proteomics studies (i.e., 80%) do not employ a chemical labeling strategy⁴; however, peptide fractionation occurs in over 50% of experiments (i.e., 64%).⁴ Chemical labeling and peptide fractionation can improve sample throughput and allow for deeper coverage of the proteome, respectively. However, chemical labeling can result in sample loss, and peptide fractionation will increase sample preparation and analysis time. Therefore, the improvement of the quality of the dataset by reducing sample complexity has to be weighed against this time loss. In addition, it is important to be mindful of available volumes for plasma experiments. Core facilities may have limited sample availability per researcher (~100 μL), therefore, it is imperative to optimize experiments using small volumes. In terms of automation, if well-plates are being employed for experiments, then the user will also be limited in the volume available to perform experiments.

Each part of the bottom-up proteomics workflow is essential to obtain high-quality data. Thus, several quality control (QC) checks should be implemented within plasma proteomics workflows to ensure reliable results. ^{56, 57} Examples of QC checks include, but are not limited to, ensuring prefractionation strategies (particularly important with column-based strategies), digestion efficiency, labeling efficiency (e.g., 95% or better), and LC-MS/MS acquisition are reproducible. These checks will allow for the quick removal of contaminants and ill-suited samples, reagents, and/or buffers before spending time and costs on future steps. They also help to minimize the number of repeat runs needed to generate good replicate data. Additionally, all samples should be treated identically to reduce variation. Ideally, this would entail using reagents from the same source and lot.

4. Automation of Plasma Prefractionation

A primary consideration for proteomic analysis of plasma samples is whether prefractionation is necessary for the assay. Ideally, crude plasma would be analyzed to avoid any drawbacks from prefractionation steps and to save time due to less sample processing steps; however, this may require longer gradients and sophisticated data analysis platforms. Various prefractionation strategies can be incorporated into sample preparation workflows to reduce the complexity and dynamic range of plasma samples, such as immunodepletion, affinity enrichment, and physicochemical-based separations, which have been discussed previously.^{2, 9, 58-60}

Immunodepletion, the most common prefractionation strategy, removes high abundant proteins (HAP) from the plasma sample to increase the proportion of low abundant proteins (LAP).⁶⁰ A certain number of HAP are captured on the stationary phase using antibodies specific to HAP while moderate abundant proteins (MAP) and LAP flow through the column and are collected. Currently available commercial products target varying numbers of HAP for removal, enabling the depletion of 50-99% of HAP, but the benefits plateau at removing 12 or 14 HAP.^{5, 60-63} Subsequently, the ProteoPrep20 Plasma Immunodepletion Kit and the SuperMix System, the latter of which removes MAP following an initial immunodepletion step, have been discontinued.^64-66 Currently commercially available products that have been integrated into automated workflows, including HAP targets, are summarized in Table 2. Other depletion strategies, such as SEER, Inc.’s Proteograph, are discussed below. Immunodepletion techniques are simple and reproducible but suffer from nonspecific binding, resulting in the removal of more proteins than expected and/or varying efficiencies of targeted protein removal. Another drawback is that additional concentration steps are typically required after depletion.

Table 2.

List of commercial immunodepletion products used in semi-automated protocols.

Proteins	MARS HSA;1 Pure- Proteome Magnetic Beads3	Pure- Proteome Protein A & G Magnetic Beads3	MARS1 & High Select2 HSA/IgG	MARS-61	MARS-71	MARS-141	High Select 142
Albumin	✔		✔	✔	✔	✔	✔
α-1-acid-glycoprotein						✔	✔
α-1-antitrypsin				✔	✔	✔	✔
α-2-macroglobulin						✔	✔
Apolipoprotein AI						✔	✔
Apolipoprotein AII						✔
Complement C3						✔
Fibrinogen					✔	✔	✔
Haptoglobin				✔	✔	✔	✔
IgA				✔	✔	✔	✔
IgD							✔
IgE							✔
Ig G		✔	✔	✔	✔	✔	✔
IgG (Light Chains)							✔
IgM						✔	✔
Transferrin				✔	✔	✔	✔
Transthyretin						✔

Vendors:

¹Agilent Technologies, Inc. (Santa Clara, CA).

²ThermoFisher Scientific (Waltham, MA).

³Millipore Sigma (St. Louis, MO).

An alternative approach to prefractionation is affinity enrichment, wherein certain classes of proteins are trapped on a solid support to equalize LAP and HAP concentrations. This technique is useful for studying post-translational modifications such as glycosylation and phosphorylation;^{67, 68} strategies for automating these analyses have been described.^69-72

Other prefractionation strategies are based on physicochemical principles such as molecular weight, charge, hydrophobicity, and isoelectric point.^{58, 60, 61} These techniques tend to be simpler and less expensive than others because they do not require antibodies, but they can present challenges with reproducibility and standardization. Recently, strategies using nanotechnology for HAP removal have been developed.⁷³

In this section, “fully automated” refers to those prefractionation strategies that require minimal to no human input (i.e., humans only set up the robotic system’s method and deck). Strategies that require some manual input or preparation have been classified as “semi-automated.”

4.1. Semi-Automated Prefractionation Strategies

Immunodepletion strategies using columns, such as Multiple Affinity Removal Systems (MARS) columns, greatly decrease the need for manual input when using a HPLC system with a refrigerated autosampler compartment, a refrigerated fraction collector, and a 96-well plate for collection.^{41, 43} This is an example of a semi-automated strategy. Using a single HPLC system with a MARS-6 or MARS-14 column would allow 126 samples to be depleted per week. The LC gradient lasts 38 minutes and generally follows the manufacturer’s recommendations.^{41, 43, 74} However, manual input is still needed to prepare samples for depletion, as MARS columns require the sample to be diluted 4-fold and filtered before injection. An automated centrifuge, positive pressure apparatus, or vacuum manifold integrated to a robotic platform could automate these steps. In most cases, the HPLC is offline from the robotic system, requiring a lab member to transfer the 96-well collection plate to the robotic system. Alternatively, there are robotic arms that can transfer the plates to the liquid handler. Therefore, while multiple HPLC systems can increase throughput, human intervention would be necessary to assist with the increased throughput.

Using superparamagnetic beads coated with PureProteome Protein A/G Mix Magnetic Beads and Albumin Magnetic Beads, a digital microfluidic platform can immunodeplete four samples simultaneously in ten minutes, which is much faster than column-based methods.⁷⁵ The sample and magnetic beads must be manually added to the device, unless a robotic pipetting system could be programmed to perform this task. The magnetic beads are separated from their supernatant through magnetic separation, and then mixed and incubated with plasma samples. Following incubation, a magnetic field immobilizes the beads, and the beads are finally separated from the plasma sample, which is now depleted. Custom Microdrop software controls droplet movement through the device and the magnetic lens. Overall, this method is semi-automated in that it requires little manual input, and additional electrodes could be added to increase throughput.

Protocols for adapting Thermo Scientific’s High Select Top 14 protein depletion resin in a 96-well plate format have recently been developed in our laboratory.^{76, 77} Briefly, resin containing the antibody for Top 14 abundant proteins are incubated at a 1:20 ratio (plasma: resin) and applied pressure using a positive pressure apparatus to collect the flowthrough with the low abundant proteins for downstream digestion.^{76, 77} Prior to adding the sample, the resin slurry should be vortexed thoroughly or stirred with a magnetic bar to ensure the resin slurry is homogenous across the wells. The manufacturer recommendation is to perform end-over-end mixing to achieve efficient binding of the HAP with the antibody, which is executed by pipette mixing the slurry with the plasma. Additional options for mimicking end-over-end mixing, such as the Alligator Vertical Tumble Stirrer,⁷⁸ can be tested to optimize the incubation method. Lastly, optimizing the volume ratio of sample to resin slurry may be necessary depending on the type and concentration of samples. The automated protocol can increase the throughput of sample preparation in a shorter amount of time to process 96 samples in parallel in approximately thirty minutes, as opposed to depleting one sample every 38 minutes with a MARS-14 column.⁷⁷ The sample preparation can also be simplified since no additional concentration steps are required after the HAP removal. Automation of the resin mixing step could facilitate automation of the entire sample preparation workflow without the need to transfer depleted samples from an offline HPLC fraction collector to the robotic liquid handler. Centrifugal filtration rapidly and inexpensively separates HAP from LAP based on molecular weight but can separate proteins arbitrarily and remove less HAP than other techniques.^{5, 79} Boccardi et. al. used 30 kDa molecular weight cut off (MWCO) filters as a first step in prefractionation, and as a second step, the cut-off fractions were processed using the parallel Multi Dimensional-Liquid Chromatography (pMD-LC).⁷⁹ PMD-LC uses solid phase extraction containing three stationary phase (anionic, cationic, and lipophilic) arrays to fractionate samples and has been automated using a Liquid Handler Biomek NX^P workstation. Fractions were digested in a 96-well plate with trypsin on the workstation and then moved to the HPLC-autosampler coupled to a MALDI-TOF/TOF MS. This two-step prefractionation method proved reproducible (>85% Pearson correlation coefficient) and effective at removing large proteins.⁷⁹ The authors noted that 16 samples could be prepared in two days; however, their use of a 96-well plate suggests that throughput could be increased.⁷⁹ Except for the initial MWCO filtration step, which could be automated on the Biomek Platform or others, the entire workflow is automated.

Chromatographic separations allow for proteins to be separated by properties, such as charge or molecular weight, over a period of time. Anion-exchange chromatography (AEC) can be used to first partition HAP into fractions, which can be profiled on weak cation, immobilized metal affinity chromatography (IMAC), and reverse phase (RP) ProteinChip array chemistries. Both steps of this workflow were automated on a Biomek 2000 Laboratory Automation Workstation with ProteinChip Biomarker Integration Package and designed for 96 samples to be processed simultaneously.⁸⁰ Only minimal manual input is required—to set up the Biomek workstation and to centrifuge plasma samples before analysis, the latter of which could be automated.

Rhode et. al. designed a multidimensional workflow featuring size exclusion (SEC), AEC, and optionally, lectin affinity chromatography as prefractionation steps.⁸¹ All steps except SEC were automated using an original workstation and micro-column arrays designed for a 384-well plate format. SEC was performed with a Highload Superdex 200 column and an Akta purifier. Additionally, tryptic digestion, desalting, and LC-MS/MS application were automated with the original workstation.^{82, 83} Because this workflow produces 96 SEC and 4,128 AEC fractions, only four samples could be processed through SEC and AEC steps in three days. Although low throughput, this workflow is precise and allows deep proteome profiling.⁸¹ Overall, setting up the SEC step, transferring those fractions (in a 96-well plate) to the workstation, and setting up the other steps are the only times manual input is required.

Integrating nanoparticle (NP) protein coronas into LC-MS workflows is an alternative to antibody-based immunodepletion or affinity enrichment. A protein corona consists of a layer of proteins that coat a NP, at the nano-bio interface, upon contact with a biological fluid such as plasma.⁷³ By altering the physicochemical properties of the NP, the identity and/or quantity of proteins that form the corona changes; in this application, integrating multiple NPs with a protein corona strategy allows for separation prior to LC-MS/MS analysis.⁷³ Proteograph is an automated platform developed by SEER Inc. that uses an SP100 Automation Instrument and ten magnetic NPs to sample the plasma proteome over seven orders of magnitude.⁷³ Proprietary NPs, which represent a variety of physiochemical properties, selectively and reproducibly bind to proteins. This increases the range of detectable low-abundant proteins in plasma samples. In a cohort of 141 patients from a non-small-cell lung cancer study, Proteograph required two weeks for analysis and identified more proteins than MARS-14 depleted plasma from the same cohort.⁷³ MARS-14 depletion was accomplished with one HPLC setup (similar to above description).

4.2. Fully-Automated Prefractionation Strategies

Stable Isotope Standards and Capture by Anti-Peptide Antibodies (SISCAPA) is a patented method for multiplexed enrichment and quantitation of peptide surrogates using magnetic beads coated with anti-peptide rabbit monoclonal antibodies.⁸⁴ SISCAPA is an “addition-only” method, meaning the sample preparation steps consist exclusively of liquid additions before a final magnetic bead extraction.⁸⁵ It is designed to be fully-automated through the affinity enrichment prefractionation step, digestion, and MS analysis, and 96 samples can be enriched and digested in four hours.⁸⁵ SISCAPA can precisely measure peptides from mesothelin (ng/mL levels) to albumin (mg/mL or g/mL levels), with an overall workflow coefficient of variation (CV) of <6% for all peptides.⁸⁶ Across laboratories, SISCAPA has also been shown to be highly precise (11% CV) across laboratories.⁸⁷ Although most commonly automated with the Agilent Bravo system, SISCAPA has also been automated with a Biomek NX^P workstation. The Biomek system can process 192 samples in 4.5 hours with workflow CVs <7%.⁸⁸

4.3. Other Strategies

A popular commercial product for affinity enrichment is ProteoMiner beads, which utilize a combinatorial library of hexapeptides coupled to a chromatographic support.^{89, 90} When plasma is applied, HAP saturate the beads and excess HAP are washed out, thereby diluting HAP concentrations and equalizing LAP and HAP concentrations.^{89, 90} ProteoMiner beads are robust, reproducible, and simple to use. In regard to automation/high throughput analysis, however, there are no studies using ProteoMiner beads.

Spin columns allow for immunodepletion to be performed using a centrifuge instead of an HPLC setup. Agilent sells a spin column equivalent to each of their MARS columns, and Thermo also sells HighSelect Top 14 spin columns and Pierce Top 12 spin columns.^{91, 92} Sample is added to a column containing a resin of the antibodies targeting certain HAP for removal and end-over-end mixing or vortexing is performed during an incubation period. The column is then placed in a collection tube and centrifuged to collect the depleted sample. Spin columns offer a rapid method to process samples and, depending on the size of centrifuges available, can greatly increase throughput.

Computational, “in silico” strategies have also been explored as alternatives to immunodepletion and affinity enrichment.⁶⁰ As a part of their plasma proteome profiling pipeline, Geyer et al. built a library from depleted plasma of one individual (using a MARS-6 column followed by ProteoPrep 20) and undepleted plasma from ten individuals.⁴ Data was searched against this library and improved protein identifications by 39%.⁴ Similarly, “in silico” enrichment can be carried out by finding a list of peptides from target proteins via Protein Atlas (www.proteinatlas.org) and SRMAtlas (www.srmatlas.org) and then analyzed using targeted proteomics.⁶⁰ While promising alternatives to traditional sample preparation methods, these “in silico” depletion/enrichment strategies need further evaluation in large-scale studies.

To automate these strategies, laboratories could integrate a centrifuge and rotation system to a robotic platform. The addition of these systems to robotic platforms would help automate semi-automated strategies discussed previously. BioNex Solutions microplate centrifuges, Hettic Rotanta 460 centrifuges, and Sigma 6-16K centrifuges (the latter two of which can be adapted for microplates, deep well plates, and conical tubes) are examples of centrifuges that automated liquid handling workstations, such as Biomek systems, can incorporate.⁹³ Various mixing systems can also be integrated into Biomek workstations, but researchers would want to pick one that mimics end-over-end mixing as opposed to shaking.⁷⁸ Capper/de-capper applications can also be integrated into Biomek workstations that could further facilitate the automation of spin columns.⁹³

5. Conclusions

Bottom-up proteomics techniques are continuously applied to analyze plasma samples. A PubMed search (08/14/2022) using keywords “proteomics” and “plasma” results in 10,819 hits. Automation increases sample throughput and reduces the chance of experimental error. In addition, advances in automation allow 100s of samples to be processed simultaneously and up to thousands potentially in a single day. Analyzing large-sample sets provide the statistical power necessary to answer clinically-based biological questions or test and validate biomarker candidates. Various options are now available for researchers to automate prefractionation and downstream plasma proteomics experiments. These protocols have been developed on several types of robotic liquid handlers that are commercially available as discussed herein. Platform flexibility, costs, upkeep, tips and labware compatibility are important considerations towards investment in an automated system. For those without access to robotic liquid handlers, it is still possible to adopt procedures that will allow for semi-automated sample preparation protocols. For example, using a manual multichannel pipette or manual high-throughput system can enable 96-364 well plates to be processed efficiently. Automation in plasma proteomics protocols will greatly help advance the goals of clinically-based studies.