Quantitative Protein Analysis Technical Guide

Table of Contents

Letter from the Editor
Index of Experts
Q1: What is your current method of choice for quantitative protein analysis, and why?
Q2: How do you work to optimize the protocol, and how does that vary based on the sample type and/or research question?
Q3: You've run your experiment, but detect no sample spot (nor peak, et cetera). Now what?
Q4: How do you work to identify potential interference in your quantitative analysis?
Q5: Which statistical approaches do you use to validate your quantitative experiments?
Q6: What are some of the recurring issues you run into when performing quantitative protein analyses?
List of Resources

Download the PDF version here

Letter from the Editor

Top-down or bottom-up. Relative versus absolute. Labeled or unlabeled. Whether discovery-stage or targeted, "the nature of quantitative proteomics analysis is such that tailored experimental approaches are needed for each sample type or research question," says Manimalha Balasubramani at the University of Pittsburgh.

And with so many choices, it can be tough to know which approach to take. From iTRAQ to SILAC, choices for quantitative protein analysis methods abound. (See Glossary, page 13.)

But as researchers look to forge a global picture of the molecular mechanisms behind biological processes, quantitative proteomics analyses are increasingly important — supplying critical perspective as to the functions and interactions of certain gene products within a sample.

To help guide your quantitative protein analyses, Genome Technology sought out experts who deal with all sorts of sample types and research questions on a day-to-day basis — proteomics core lab leaders. Here, an invited panel dishes insider advice on protocol optimization, error-correction, statistical challenges, and more.

If after poring over our contributors' responses you still need more, be sure to check the handy reference list at the back of this guide for a look at what's new in the quantitative protein analysis literature.

— Tracy Vence

Index of Experts

Many thanks to our experts for taking the time to contribute to this technical guide, which would not be possible without them.

Manimalha Balasubramani
University of Pittsburgh
Genomics and Proteomics Core Laboratories

Shixia Huang
Baylor College of Medicine
Proteomics core Laboratory

David Smalley
University of North Carolina, Chapel Hill
Proteomics Core Facility

Martin Steffen
Boston University School of Medicine
Proteomics Core Facility

Q1: What is your current method of choice for quantitative protein analysis, and why?

Label-assisted and label-free quantitative proteomics techniques are equally popular today. We frequently use chemical labeling with iTRAQ reagents for measuring relative changes in protein abundances. The iTRAQ technology is applied to investigator-initiated projects that require systems-wide comparisons; these are typically high-complexity samples that have been manipulated to some extent in order to enrich or fractionate a subset of proteins the multiplexed analysis affords higher throughput, precision, and lower cost, especially for pilot-scale experiments. Low-complexity samples generated by immuno-pulldown or other methods can be anticipated to have high technical variance from sample manipulation. In such cases we encourage the investigator's lab to consider metabolic labeling (SILAC) approaches frequently, we perform qualitative ms analyses during the protocol development stages. We apply label-free techniques such as emPAI and iBAQ to measure protein abundances in within-run comparisons and, for example, to infer stoichiometry of protein complexes. We are also moving toward targeted ms methods with heavy isotope labeled synthetic peptides in the sample digest for absolute quantification.

— Manimalha Balasubramani

Our core facility uses antibody-based technologies including antibody arrays, reverse phase protein microarrays, and Luminex xMAP technology — a multiplex, bead-based immunoassay these technologies label the samples/antibodies either directly with fluorescent dye or with biotin and uses streptavidin-conjugated dye as a detection system. The advantage of these technologies is the ability to assay low-abundance proteins with the protein Ids readily available. My favorite platform so far for accurate quantitation is the Luminex xMAP technology because of its sensitivity, reproducibility, and wide dynamic range. However, based on the specific project, it could be a choice of any of these methods though antibody arrays lack quantitative accuracy, it has higher throughput and the potential to assay tens of thousands of proteins at once to make up for it. When budget allows, I recommend combining all the platforms since that allows us to obtain the maximal amount of data for a particular project.

— Shixia Huang

The method of choice is highly dependent on the experiment itself. If we are trying to examine larger differences (e.g. more than five-fold), label-free spectral counting works well for smaller differences, we use isobaric labels (iTRAQ or TMT). These are robust and work on a wide range of samples. Under certain circumstances, SILAC is preferred for example, if we want to examine differences in phosphorylation from samples in culture, we try to start out with several milligram of protein per sample. Labeling with an isobaric tag is cost prohibitive and therefore SILAC is preferred. We also use super-SILAC for clinical samples if we want to examine specific sub-proteomes.

— David Smalley

We have an older ThermoFinnigan LTQ [Thermo Scientific] and are not able to utilize low-mass reporter ions for quantification. However, we are able to utilize the mTRAQ reagents [Sciex], performing an initial screen of quantitative changes using MS1 data, and then have the ability to confirm them using MRM on the same labeled sample.

— Martin Steffen

Q2: How do you optimize the protocol, and how does that vary based on the sample type and/or research question?

We use cell lysates from mouse or human to evaluate the performance of the quantitative method and to further optimize it the nature of quantitative proteomics analysis is such that tailored experimental approaches are needed for each sample type or research question. In this respect, quantitative analyses are unlike the more routine protein identification services. We start with an initial exploratory experiment by GeLC-MS or gel-free LC-MS methods to assess potential biases for each sample type during the experimental planning stage, we look at ways to minimize biases from technical variation either from sample processing, mass spectrometry analysis, or data interpretation. At this point, core lab staff may be closely working with the researcher's team in developing new methodology for the proposed experiments, and some of these projects may become collaborative in nature.

— Manimalha Balasubramani

Sample preparation is the key first step for protocol optimization, and sample preparation protocol differs for different sample types in each proteomic platform for antibody-based platforms, especially glass slides arrays (antibody or reverse phase protein arrays), we optimize the assay protocol to obtain high signal-to-noise ratio, lower assay background, and higher dynamic range. For example, we test different labeling methods with various dyes and adjust our laser scanner settings; and we include proper positive and negative controls and quality controls for each of the assays. We develop for each sample type often, specific research projects ask unique questions and we need to adjust our protocol for that particular project. Some investigators have knowledge about the samples and targets to be assayed so we have an idea on which protocol works the best. Others are purely discovery based and we usually test the samples and protocol on a smaller scale to make sure the protocol works for that particular project. To minimize inconsistency and increase the chance to obtain useful data, we meet with investigators first to understand the project, take time to design the experiment, guide them step-by-step to prepare, store, and deliver the samples, and to understand the data at the end. We have written protocols for each project/experiment and quick checklists on what to do with each type of samples.

— Shixia Huang

Since my laboratory obtains a wide variety of samples, it is essential that we standardize the protocols as much as possible and design the workflows so we can accept a wide variety of samples. While in-depth literature searches are important, there are no shortcuts to optimizing protocols. Comparing experimental conditions using identical samples is the only way this can be done effectively.

— David Smalley

Most of the optimization efforts are directed toward making the sample prep method highly reproducible. We take pains to ensure that the final isolated proteins are free of any additional amino group (buffers, nucleic acids, et cetera). We also take care to make sure that the amount of protein is accurately quantified. Lastly, any method is validated by comparing a control sample with itself.

— Martin Steffen

Q3: You've run your experiment, but detect no sample spot (nor peak, et cetera). Now what?

It is critical to have quality controls at each stage. We would look at their performance first to investigate a failed sample run. At present, core laboratories are missing uniformly adopted quality-control and quality-assurance standards. However, each laboratory has its own set of well-characterized quality controls that are valuable in investigating system performance issues and troubleshooting sample runs for example, we have quality control standards to assess digestion, labeling, LC-MS runs, and other parts of the iTRAQ workflow.

— Manimalha Balasubramani

We take the following steps to check the cause of error but the orders might be different:

1) We would check on the instrument setting and run some previously tested samples to check the instrument performance

2) We would check to make sure that we used the kit/antibodies against the species where the samples were from

3) We would check to make sure that we added/loaded the samples correctly and followed the protocols properly

4) We would check with those who prepared the samples to make sure they followed the protocol, such as used the appropriate buffer and delivered the samples to us at the concentration as it was claimed

— Shixia Huang

We typically establish all protocols before processing any real samples and attempt to include quality controls into all steps. For example, we spike in exogenous peptides to verify our LC separation and MS analysis is working properly. In addition, we try to have some type of quality control before we accept any sample for mass spectrometry analysis for protein characterization, we strongly recommend all new users submit their sample in a stained gel this gives us an opportunity to assess the amount and purity. In addition, we don't have to deal with interfering substances, such as detergents, glycerol, or other substances that may affect our workflow. Even if individuals are submitting samples in solution, we recommend that they examine a small portion by PAGE so they have a better understanding of exactly what they are submitting. This is particularly true for individuals trying to perform co-immunoprecipitations to identify binding partners differences between samples and controls should be obvious on gels. Proteomic analysis of these samples is relatively straightforward, but the co-immunoprecipitations are highly antigen/antibody specific and can take a lot of work to optimize once we are confident we have a good sample, our methods are usually routine enough to provide good results. Accepting samples prior to having clearly established a protocol usually leads to more trouble than it's worth.

— David Smalley

With mTRAQ labeling, the problem we would first suspect is a partial, or complete absence, of labeling of one or all of our samples monitoring the relative labeling of peaks from the autolysis of trypsin can help estimate the efficiency of labeling

— Martin Steffen

Q4: How do you work to identify potential interference in your quantitative analysis?

For a long time, the magnitude of ratio underestimation with iTRAQ was not appreciated. It was known that background interference and other factors resulted in ratio compression effects. Recent work from Steve Gygi's lab [at Harvard Medical School] and others report technical limitations with TMT — another isobaric tagging technology — and how these can be overcome with use of MS3 experiments or others these fixes are not yet in a form that can be seamlessly implemented in other research laboratories, but nevertheless offer a lifeline for multiplexing technologies such as iTRAQ and TMT.

— Manimalha Balasubramani

We make sure to include quality-control samples and standards (when possible) and develop criteria to assess the quality of the data measured. We always include buffer controls in our immunoassays, negative control slides for reverse phase arrays, and control groups for antibody arrays. When we do large-scale projects, we test it on a smaller scale first before applying a large number of samples to that specific platform.

— Shixia Huang

There are two issues in this question the first is separation artifacts, which is a huge problem when we are dealing with processed samples. For example, for individuals examining the proteomes of subcellular fractions, the reproducibility of their isolation process must be assessed. This is normally done by performing analytical replicates. For example, if we are looking at changes in a membrane fraction following a particular treatment, we perform the isolation in duplicate — both control and treated — and label all four samples with various iTRAQ tags. This allows us to assess the variability associated with the isolation procedure and the differences between the control and treated group. The TMT 6-plex or the iTRAQ 8-plex is used when we would like to compare three or four sets of samples, respectively.

The second issue is interfering peaks in the mass spec analysis. While this has been well documented, we believe this problem is overstated, particularly if we are examining less complex samples. Our goal is normally to identify proteins that are differentially expressed, and we err on the side of underestimating differences rather than finding "false positives " on specific samples. We do use MS3 analysis as described by Ting et al., but this is not routine and doesn't tend to alter our overall results

— David Smalley

Since we measure the relative amounts of two or more samples simultaneously, the interferences in chromatographic separation should be minimal.

We gain confidence for a calculated abundance ratio for a particular protein when we see that the ratio calculated by each peptide pair is roughly similar. Finally, for those proteins critical to our biological argument, they are checked by the orthogonal quantification method of MRM.

Finally, for those proteins critical to our biological argument, they are checked by the orthogonal quantification method of MRM.

— Martin Steffen

Q5: Which statistical approaches do you use to validate your quantitative experiments?

Quantification experiments aim to arrive at protein abundance levels in the sample under consideration whereas shotgun proteomics delivers peptide-level data from which protein identifications and abundances need to be inferred. This challenge has been met by better protein inferences from database searching by target-decoy strategy and use of stringent false-discovery rates to report peptide and protein identifications.

We now report peptide/ protein identifications with 1 percent local false-discovery rate and quantifications based on unique peptides using commercial database search programs. Many of the projects we get are proof-of-principle or pilot scale with a limited number of samples for analyses. Here, the researcher plans a quantitative proteomics experiment as an initial screen with a limited sample size and later follows up with orthogonal approaches using larger sample sizes to independently validate the quantitative experiment. Clinical samples have far more complex requirements.

For statistical analysis since biological variability has to be controlled for. We recommend the researcher to work with statisticians from the initial planning stage of the project.

— Manimalha Balasubramani

As a core lab, we work with the investigators and the statistician for the data analyses. Usually, we have frequent communications with both sides to help the statistician understand the experiment before analyzing the data and help the investigator understand the data and interpret it. One thing I emphasize the most is the quality of the statistical data, which not only includes the P value and fold changes, but also the raw data, including the signal intensity, signal-to-noise ratio, local background et cetera. Statistical approaches used by our collaborating statistician include ANOVA, two sample T-test or paired T-test, AUC analysis, et cetera.

— Shixia Huang

Statistical analysis of proteomics data is a challenge and one can spend considerable time and effort trying to prove the results obtained from mass spectrometry purely on a statistical basis.

While a few recent publications attempt to tackle this issue with some success, none do so in way that provides ease and confidence to individuals who are not statisticians. A better alternative is view mass spectrometry as a discovery tool and use an independent technique, such as immunoblots, to validate a portion of the results this has the added benefit of greatly increasing the number of samples that may be analyzed and generally satisfies reviewers regardless of their scientific background.

— David Smalley

We attempt to do this empirically by observing the variation in control versus control samples, and spiked peptides.

— Martin Steffen

Q6: What are some of the recurring issues you run into when performing quantitative protein analyses?

An area that the researcher's lab can focus on up front to help reduce variability in the analyses is by paying attention to sample preparation. Reproducibility and quality of the preparation needs to be assessed well in advance of a quantitative proteomic experiment. For example, running a couple of different sample preparations by 1D gels is a simple and efficient method to check quality and reproducibility. From a core lab standpoint, quantitative proteomics has multifaceted challenges. Although we can predict the time and effort needed for performing proteomic analyses, since the researcher may be initiating a first proteomics experiment, the time spent in data interpretation and hand-off is unquantifiable. We may not face this challenge with targeted proteomics. However, for the science to move forward in challenging areas such as quantitative proteomics, it is necessary for the researcher to arm themselves with basic bioinformatics skill sets in interpretation of tandem mass spectrometry data. The most successful projects we performed are those that had a well-posed research question, tailored sample preparation, and mass spectrometry analyses with a strategy for a publishable result or manuscript

— Manimalha Balasubramani

Despite the fast development of proteomic technologies in recent years, there is still no single platform as the platform to cover all. The biggest challenge for us is access to high-quality and high-throughput arrays or multiplex assays to cover a wide range of proteins in the proteome. It is still challenging to obtain quality antibody for a single protein in the laboratory and to optimize assays for that one protein. To generate antibodies against tens of thousands of proteins and optimize assays to ensure its functionality for each of them surely poses great challenges to the scientific community. I believe working together and sharing expertise and data are the keys to solve these problems. The Human Protein Atlas is a great example of that global effort and has made great progress. Since initiated in 2003, the Human Protein Atlas has released more than 15,000 antibodies, targeting proteins from more than 12,000 genes. My hope is that our community works together to make great effort to use those antibodies to make arrays and immunoassays to cover all the proteins with antibodies available. There are efforts being made in this area, but I would like to see expedited progress on a larger scale and with wider access.

— Shixia Huang

The two major obstacles that we face are that we normally only perform relative quantitation and differences between samples must still be relatively large (more than 40 percent for lower abundant proteins in complex mixtures). More quantitative approaches are not technically difficult, but the costs and labor associated with them are high. This is particularly true if you are not doing routine analysis of the one or a few proteins over months or years. More global approaches, such as the Human Proteome Detection and Quantitation Project, which is trying to create a complete suite of assays for approximately 20,500 human gene products, show promise, but are years away from realization. This is not to understate the tremendous advances in both qualitative and quantitative analysis that have occurred in the field of proteomics in the last several years.

— David Smalley

Because of the large dynamic range of protein abundance within a cell, and our instrumentation, we have difficulties going deep into the proteome. We deal with this by sub-fractionation of the labeled peptides using HPLC prior to loading peptides onto the C-18 column for mass spec analysis

Another method which we use occasionally is isolating sub-cellular fractions of the proteome prior to labeling (nucleus, mitochondria, et cetera). However, this can be challenging because these methods are typically less reproducible than whole cell protein preparations.

— Martin Steffen

List of resources

A compendium of terms and papers to help answer all your quantitative protein analysis questions.

Glossary

emPAI: exponentially modified protein abundance index

geLC-MS: 1S SDS-polyacrylamide gel electrophoresis, followed by liquid chromatography-mass spectrometry

iBAQ: intensity-based absolute quantification

iTRAQ: isobaric tags for relative and absolute quantitation

MS1, 3: first device in tandem mass spectrometry (MS-MS), third device in tandem mass spectrometry

PAGE: polyacrylamide gel electrophoresis

SILAC: stable isotope labeling with amino acids in cell culture

TMT: tandem mass tag

Glossary

Bettmer J. (2012). Application of isotope dilution ICP-MS techniques to quantitative proteomics. Analytical and Bioanalytical Chemistry. 397(8):3495-3502.

Blein-Nicolas M, Xu H, de Vienne D, Giraud C, Huet S, Zivy M. (2012). Including shared peptides for estimating protein abundances: a significant improvement for quantitative proteomics. Proteomics. Epub: doi 10 1002/pmic 201100660.

English JA, Manadas B, Scaife C, Cotter Dr, Dunn MJ. (2012). Partitioning the proteome: phase separation for targeted analysis of membrane proteins in human post-mortem brain. PLOS One. 7(6): e39509.

Fan J,Mohareb F,Jones Am,Bessant C. (2012). MRMaid: The SRM assay design tool for Arabidopsis and other species. Frontiers in Plant Science. 3: 164.

Gautier V, Mouton-Barbosa E, Bouyssie D, Delcourt N, Beau M, Girard JP, Cayrol C, Burlet-Schiltz O, Monsarrat B, Gonzalez De Peredo A. (2012). Label-free quantification and shotgun analysis of complex proteomes by 1D SDS-PAGE/nanoLC-MS: Evaluation for the large-scale analysis of inflammatory human endothelial cells. Molecular & Cellular Proteomics. Epub: doi 10 1074/mcp m111 015230.

Geromanos SJ, Hughes C, Ciavarini S, Vissers JP, Langridge JI. (2012). Using ion purity scores for enhancing quantitative accuracy and precision in complex proteomics samples. Analytical and Bioanalytical Chemistry. Epub: doi 10 1007/s00216012-6197-y.

Peterson AC, Russell JD, Bailey DJ, Westphall MS, Coon JJ. (2012). Parallel reaction monitoring for high resolution and high mass accuracy quantitative, targeted proteomics. Molecular & Cellular Proteomics. Epub: doi 10 1074/mcp o112 020131.

Podwojski K, Stephan C, Eisenacher M. (2012). Important issues in planning a proteomics experiment: statistical considerations of quantitative proteomic data. Methods in Molecular Biology. 893: 3-21.

Rodríguez-Suárez E, Whetton AD. (2012). The application of quantification techniques in proteomics for biomedical research. Mass Spectrometry Reviews. Epub: doi 10 1002/mas 21347.

Tang W. (2012). Quantitative analysis of plasma membrane proteome using two-dimensional difference gel electrophoresis. Methods in Molecular Biology. 876: 67-82.

van den Ouweland JM, Kema IP. (2012). The role of liquid chromatography-tandem mass spectrometry in the clinical laboratory. Journal of Chromatography. b883-884: 18-32.

Vaudel M, Burkhart JM, Zahedi RP, Martens L, Sickmann A. (2012). iTRAQ data interpretation. Methods in Molecular Biology. 893: 501-509.

Wang M, You J. (2012). Mass spectrometry for protein quantification in biomarker discovery. Methods in Molecular Biology. 815: 199-225.

Yang SJ, Nie AY, Zhang L, Yan GQ, Yao J, Xie LQ, Lu HJ, Yang PY. (2012). A novel quantitative proteomics workflow by isobaric terminal labeling. Journal of Proteomics. Epub: doi 10 1016/j jprot 2012 07 011.