High-Throughput Chromatography Technical Guide

Table of Contents

Letter from the Editors
Index of Experts
Q1: How does your choice of separation materials impact your HPLC system?
Q2: What sample preparation or purification procedures do you use?
Q3: How do you ensure reproducibility of the gradient elution?
Q4: What detection methods do you use, and how do they help optimize specificity?
Q5: How do you optimize the detection of low-abundance proteins, or other hard-to-identify proteins?
Q6: What do you consider to be the most important during analysis: detection levels, speed, or quality of detection? Why?
List of Resources

Download the PDF version here

Letter from the Editor

Genome Technology is proud to present this latest technical reference guide that tackles the ins and outs of high-throughput chromatography. It'll take you through choosing your separation materials, detecting your sample, and even analyzing the data that you produce.

Chromatography, though, wasn't always this complicated. In the early 1900s when Russian botanist Mikhail Tsvet developed chromatography, he separated plant pigments using powdered chalk and alumina. Then, the early days of centimeters-wide glass columns gave way in the 1970s when pressures of 500 psi, then 6,000 psi, were applied to chromatography columns. Today, pressures reaching 15,000 psi and columns nearing 1.7 microns ushered in ultra-high-performance liquid chromatography, now a thoroughly modern tool.

UPLC and MudPIT, or multidimensional protein identification technology, are just two methods used today that are expanding the field of proteomics and proteome analyses. Hard-todefine, low-abundance proteins, such as transcription factors and protein kinases, as well as metabolomic proteins that can serve as useful biomarkers, are now more readily characterizable with these advanced separations techniques.

Whether used for biomarker analysis, drug discovery and development, or quality assurance testing, capillary to nanoscale separations help achieve the highest sensitivity, resolution, and reproducibility possible when dealing with complex protein mixtures. To help ease your workload, we've assembled here advice and opinions from experts all over the world. They'll tell you how they cope with detecting minuscule amounts of protein and how they optimize reproducibility versus speed of detection while keeping the quality of their data intact.

Ciara Curtin and Jeanene Swanson

Index of Experts

Genome Technology would like to thank the following contributors for taking the time to respond to the questions in this tech guide.

Emily Hilder
Lecturer
Australian Centre for Research on Separation Science
University of Tasmania

Rebecca Lew
Lecturer
Department of Biochemistry & Molecular Biology, Monash University

Edouard Nice
Head, Protein Biosensing Laboratory
Ludwig Institute for Cancer Research

Vincent T. Remcho
Professor
Oregon State University

Nabil Saad
Postdoctoral Researcher
Metabolomics Laboratory
University of California, Davis

Q1: How does your choice of separation materials impact your HPLC system?

The separation column is the most important component of the HPLC system and consequently there have been many recent developments in column technology, in particular for proteomics analysis. Most separations are still performed using reversed phase (RP) columns but the nature of the chromatographic stationary phase is of extreme importance for the enrichment and separation of post-translational modified peptides or proteins. Immobilized metal ion affinity chromatography (IMAC) or titania-based stationary phases can be used for enrichment of phosphopeptides and will increase the number of detected peptides. Monolithic columns (e. g. ProSwift from Dionex) where the column is made of a single, porous piece of polymer are extremely useful to increase sample throughput. Monolithic columns offer easy preparation and functionalization, enhanced mass transfer, high column efficiency, and they are more robust compared to pellicular HPLC columns. However, monolithic columns suffer from one very important limitation: they are very often and very easily overloaded which can be problematic for proteomics analysis. An alternative is ultra performance chromatography (UPLC) using sub 2m particles for the stationary phase at elevated pressures and enables fast chromatographic separations of proteomic samples. Faster scanning mass spectrometers would enable the routine use of UPLC, which will enable higher sample throughput.

More critically, the complexity of proteomic samples that often contain thousands of peptides means that one-dimensional separation of peptides or proteins is not sufficient and additional separation dimensions must be introduced in order to reduce the sample's complexity. The most widely used combination is ion exchange chromatography in both anion and cation form, prior to RP separation. This approach, termed multidimensional protein identification technology (MudPIT) and introduced by Yates et al. (Wolters DA, Washburn MP, Yates JR, III, 2001), usually involves a single biphasic column packed with strong cation exchange (SCX) stationary phase, and C18 RP and has been very successful.

— Emily Hilder

Our work focuses on peptides, for which we mostly use reverse-phase materials, although we do the occasional ion exchange or size exclusion chromatography as well. We work generally at the analytical level (microliter/min flow rates), but also delve down into the nanoLC level (with a new instrument) and up to the preparative level. Because of these different requirements, we have found the Agilent 1100 series to be the most flexible, although it does limit what we can do. We have recently found funding for a nanoLC for analysis of low abundance proteins and peptides. This feeds directly into a mass spectrometer for identification.

— Rebecca Lew

The most important things dictating our choice of HPLC system are the sample size and the dimensions of the columns we are running. Many of our separations are done at the low or sub-mg level and probably more than 80 percent of all our separations are done of columns of 2.1mm ID or less. Because of our requirement to recover samples for multidimensional purification protocols, or for sensitive and specific off-line analysis, a reliable microfraction collector is required. I have never been convinced of the requirements for systems which are "bio-compatible" and typically run any column in any of our systems by the use of suitable adaptors.

— Edouard Nice

The composition and properties of eluents and samples factor largely in selection of appropriate HPLC components and systems. For reverse-phase separations, which are the most prevalent, there are a number of good options. Pumps generally consist of 316 stainless steel pump heads, synthetic sapphire pistons, and check valves with fluoropolymer seats and ruby spheres. Good tubing options include stainless steel and PEEK (polyetheretherketone) for high-pressure portions of the system and Teflon for low-pressure applications. Injection valves will often have stainless steel or PEEK flow paths and fittings. Detectors will have solvent-tolerant flow cells and windows. For ion exchange separations, corrosionresistant materials are preferable. Here, PEEK, polycarbonate, and fluoropolymers are often used in construction of the LC system. Certainly, HPLC systems used for reverse-phase LC can be used for ion exchange, but it is important to ensure that any high-salt content buffers, acidic or alkaline eluents, etc., are thoroughly flushed from the system prior to shut-down to prevent serious problems such as corrosion and precipitation/clogging. With size exclusion separations, materials selections are driven by the same kinds of factors — solvent compatibility, corrosive effects, analyte stability. It is amazing how robust modern HPLC systems are if they are cared for properly.

— Vincent T. Remcho

With regards to the column stationary phase:

For metabolomic applications, the choice of separation material used, including particle size, is dictated by the HPLC instrumentation hardware available for the user. In the case where an ultra-high pressure LC (U-HPLC) system is available (pressure limit can vary from 600bar to 1000bar depending on the manufacturer), then we opt to use either a 1.8um or 1.7um particle size columns, depending on availability. The choice of the specific type of stationary phase depends on the exact application at hand. For regular conventional HPLC, we opt to use monolithic silica-based columns due to their high porosity and high-throughput capability (shallow Cterm in the Van-deemter plot). I believe more should be said on the HPLC instrument per se affecting the quality of separation. The dwell volume and extracolumn effect of the instrument are of prime concern to eliminate/minimize any ensuing bandbroadening. Different HPLC injectors can adversely contribute to band-broadening, depending on their configuration.

— Nabil Saad

Q2: What sample preparation or purification procedures do you use?

Most often I use solid phase extraction (SPE) for sample preparation or purification. This includes both commercial SPE devices as well as custom SPE materials developed within our laboratory for selective extraction using, for example, affinity interactions. Trap columns that have high capacity and low void volumes can be used for efficient injections of large volumes of highly diluted samples. SPE can also be used to combine a reversed phase (RP) separation column with columns such as ion exchange (IEX) or immobilized metal-ion affinity chromatography (IMAC) to increase the amount and quality of analytical information that can be collected. That is, peptides which have been enriched, trapped, pre-fractionated, or desalted on other column types, are eluted and separated on an RP column prior to detection by MS.

— Emily Hilder

Sample preparation obviously varies with the sample. We use LC to answer different kinds of questions; for example, some of what we do is test the effects of enzyme inhibitors on the cleavage of a synthetic substrate by a purified recombinant enzyme. These samples are obviously very clean, and the only preparation needed is to stop the enzyme reaction and usually reduce the sample volume. For these purposes, we add four volumes of 1 percent trifluoroacetic acid in methanol to the samples, spin them in a microfuge to remove any precipitated material, and dry them using a centrifugal evaporator (Speed-Vac). The samples are then reconstituted in aqueous LC buffer at the desired volume prior to analysis.

For analysis of peptides in complex biological materials, we would normally extract the sample with disposable reverse-phase columns (e.g., Seppak cartridges). This step removes the larger proteins and also desalts the sample, which is useful for ion exchange chromatography.

— Rebecca Lew

It has always been my hypothesis that one should process bulk biological samples as quickly as possible through the initial stages of purification. This, coupled with the use of protease inhibitors, reduces the possibility of sample degradation. The choice of protocol is very much dependent on the nature of the sample, but for soluble factors it usually involves trace enrichment so that both concentration and purification can be achieved. Anion exchange, RP, and ligand dye matrices have all proved effective. The use of large pore size bulk materials allows columns of suitable dimensions to be packed and sample loading to be achieved at high flow rates (up to 50mls/min allowing liter volumes to be processed). For interactive supports recovery can be achieved by step-wise elution allowing the procedure to be performed with a single pump. Samples can be loaded directly through the pump. However, the filters should be removed to avoid fouling. We also have a number of protocols where circulating or exfoliated cancer cells are recovered from biological samples such as blood, feces, or urine. In these cases we use magnetic beads functionalized with specific antibodies to recover cells prior to lysis in situ. Magnetic particles have the advantage that they can be easily recovered from crude, particulate samples and that extensive washing to reduce residual background contamination is facile. We have also observed that these supports show reduced non-specific binding compared with conventional chromatographic supports. Finally, for the preparation of serum samples for proteomics analysis the use of depletion columns to remove the most abundant proteins, coupled with downstream multidimensional purification protocols, allows low level components to be identified.

— Edouard Nice

We routinely employ a variety of sample preparation methods, most often solid phase extraction (SPE), but also solid phase micro extraction (SPME), filtration, precipitation, and occasionally accelerated solvent extraction. We have in the past used an SPE approach to derivatization as well, in which SPE cartridges are preloaded with the derivatization reagent (in our case it was dinitrophenylhydrazine, DNPH) and the analyte is selectively functionalized to facilitate separation and detection.

— Vincent T. Remcho

In our metabolomic applications, we use misciblesolvent extraction, namely (3:3:2 Isopropanol : Acetonitrile : Water) from lyophilized plant material or fresh one.

— Nabil Saad

Q3: How do you ensure reproducibility of the gradient elution?

To ensure reproducibility of gradient elution in liquid chromatography the best approach is to use a good system suitability check. This involves using a standard protein mixture of known composition and sufficient complexity to routinely check the system performance. As a minimum, the system suitability check should be run after every five samples. This is far more reliable than relying on indirect indicators such as backpressure profiles, which are particularly difficult to use in capillary or nanoscale HPLC systems. Using a system suitability check will result in more rugged and reliable data.

— Emily Hilder

Unfortunately, even with the best LC system and the utmost care, slight day-today variations in gradient elution occur. These are generally minor — anything major requires some serious investigation before proceeding. To get around the variations, we routinely include a series of control samples in any given LC run. These include blank runs (solvent only) and repeated samples of synthetic peptide standards, if available. These would normally be run at the start and again at the end of a group of unknown samples, to account for any shift in retention time during the automated set of LC runs.

— Rebecca Lew

Reproducibility for gradient elution is achieved by careful attention to re-equilibration times. Indeed I suggest building the re-equilibration into the overall elution program. Many labs have reported a "first run of the day" effect when non-reproducible retention times are observed, especially if the column has been left overnight in primary buffer. To overcome this, the column should be exposed to Buffer B and then the controlled re-equilibration performed.

— Edouard Nice

Modern gradient systems, both high pressure and low pressure, are remarkably reproducible if used properly. High-pressure gradient formers use multiple pumps, one per solvent, and dispense directly to a mixing chamber at high pressure. Low-pressure gradient formers use a single pump. Prior to the pump, a set of proportioning valves regulate the quantities of various solvents that are dispensed to a mixing chamber at low pressure. The two approaches have different merits and limitations, and both are widely applied. In both cases, it is good practice to use a broad range in the gradient to ensure reproducibility. That is, if a gradient from 10 percent methanol to 25 percent methanol is desired, it is better to use two mixed solvents (solvent A being 10 percent methanol and solvent B being 25 percent methanol) and run a gradient from 100 percent solvent A to 100 percent solvent B, than to require the system to deliver very small volume increments from neat solvents in vessels A and B. Most instrument vendors offer good advice in their user manuals, along with specific examples of expected performance under differing conditions.

— Vincent T. Remcho

We start off by gravimetrically preparing our mobile phase, then by allowing enough time for column reequilibration (10 column volumes for particle columns, 5 volumes for monolithic columns), and last by the use of a QC mixture, before and after the batch of samples analyzed.

— Nabil Saad

Q4: What detection methods do you use, and how do they help optimize specificity?

All detector types used for conventional HPLC are applicable for proteomics analysis, and most often UV detection is used, in particular for quality-controlling of the separation (void volume, impurities, base line, and gradient stability) and for tracing fractions when the sample is being fractionated. However, it is not possible to differentiate by UV spectra alone whether two or more peptides are co-eluting. While the UV detector is mainly used for accurate characterization and quantitation, mass spectrometric detection is used and can be considered the workhorse of proteomics. In our laboratory we primarily use an ion trap mass spectrometer with electrospray ionization to perform
MSn. An Orbitrap MS can also be used for more accurate and more sensitive results.

— Emily Hilder

Our standard detection method is simply UV absorbance, most commonly at 214 nm. We have a diode array detector that will also record the absorbance along a set UV spectrum, so we can also check at alternative wavelengths (e.g., 280 nm) as an aid in specificity. As mentioned, we have also just acquired a mass spectrometer that can help identify components as they emerge from the LC column. I suppose a quality mass spec (and well-trained personnel) represents the ultimate check for specificity!

— Rebecca Lew

We routinely use UV absorbance as our primary method of on-line detection. To optimize specificity, samples are routinely recovered using a built-in microfraction collector and samples taken for downstream analysis using biosensors (eg BIAcore,
IAsys), mass spectrometry (both MALDI and LCMS/MS), bioassay, SDS-PAGE, and western blotting.

— Edouard Nice

We primarily utilize UV-absorbance detection, though we also employ fluorescence,
electrochemical, refractive index (RI) and mass spectrometric detectors. RI is useful for some of our sugar analysis applications, though it is quite temperature-sensitive and notoriously insensitive. It does offer the advantage of universal applicability when modest detection limits are acceptable. UV is much more sensitive and a bit more selective. Many analytes do have UV chromophores and so it is a workhorse detector for us. For high selectivity applications we turn to fluorescence or electrochemical detectors, with a preference for fluorescence detection largely attributable to its good fit for biological applications (there are a number of excellent fluorophore derivatization reagents available today). MS detection is a useful option, though method development is more time consuming and in general a much higher level of user experience is required for success.

— Vincent T. Remcho

Our detector of choice is mass spectrometry. Mass spectrometers detect compounds by mass-to-charge ratio. Further structural information about analyzed compounds can be obtained by tandem MS, i.e. MSn.

— Nabil Saad

Q5: How do you optimize the detection of low-abundance proteins, or other hard-to-identify proteins?

The detection of low-abundance proteins is best optimized by depletion of the most abundant proteins prior to analysis. High-abundant proteins such as albumin, IgG, transferrin, haptoglobin, IgA, antitrypsin and fibrinogen comprise up to 90 percent of the total protein mass in human plasma and will interfere with identification and characterization of important low-abundant proteins by limiting the dynamic range of mass spectral analyses. Commercial solutions such as the Multiple Affinity Removal System (MARS) available from Agilent Technologies are available and research efforts in this area are also focused on the development of miniaturized and alternative systems. For the specific determination of known low-abundant proteins and peptides an alternative is the use of affinity columns that can be used to selectively trap and enrich the target proteins for improved detection.

— Emily Hilder

A terrific question — I wish I had a terrific answer! Low-abundance proteins are the bane of many a researcher's existence, but of course, these same proteins are the most interesting (plus all the easy proteins have already been studied). I suppose the best methods involve high-quality antibodies and some sort of immunological identification step, such as an ELISA or immunoblot of LC fractions. We have used these methods to identify peptides and peptide fragments. We are lucky in that we also study proteolytic enzymes, where we either are looking at the cleavage of synthetic substrates, which we have in abundance, or isolating the enzymes themselves and use catalytic activity as a measure of the enzyme. Of course, the latter is only compatible with nondenaturing LC methods, such as ion exchange or size exclusion.

— Rebecca Lew

As I outlined in the sample workup procedures, the ability to detect low-abundance proteins resides in effective depletion and purification strategies. Prior to the development of what we now call proteomics, low-abundance proteins were obtained by purification to homogeneity using multidimensional purification protocols followed by N-terminal sequence analysis using Edman degredation. Frequently more than five chromatographic steps were required and purification factors of up to 106 were achieved. The ability of mass spectrometry to analyze complex samples means that purification to homogeneity is no longer required. However, because of the large concentration range found in many samples (probably up to 1010 in samples such as blood) it is becoming obvious that both depletion of the more abundant housekeeping proteins followed by trace enrichment and further fractionation is required. The current literature clearly shows the benefits of mutidimensional purification, in terms of number of proteins analyzed and the detection of low-abundance proteins, at both the protein and peptide levels in either top-down or bottom-up proteomics protocols.

— Edouard Nice

Low-abundance proteins call for preconcentration (using SPE, SPME, etc.) and/or high sensitivity detection strategies for successful analysis. Isolation and preconcentration of the target protein fraction simplifies the detection step. Careful selection of a fluorophore label is necessary. We have used kits from Molecular Probes (now part of Invitrogen) for some time, as the kits are designed specifically to work with small volume samples and are convenient. While it is possible to purchase the tag separately, rather than as part of a kit, it is then necessary to divide the bulk reagent prior to use. Given the reactivity/stability of the reagent and the hassle involved in dividing it, we seldom opt for this approach.

— Vincent T. Remcho

Q6: What do you consider to be the most important during analysis: detection levels, speed, or quality of detection? Why?

The most important factor during analysis depends entirely on the sample type and the desired outcome of the analysis, which will vary from sample to sample. Sample throughput is often the bottleneck for proteomics analysis but the speed of the separation is still not often the most important factor as a fast separation usually means that the detection levels or quality of detection such as MS data, is compromised. For example, in the analysis and characterization of low-abundant proteins the detection levels are far paramount followed by quality of detection for accurate identification.

— Emily Hilder

For us, speed would be the least important, perhaps because we are not processing hundreds of samples at a time, and it always saves time (and money) to analyze the samples properly the first time and avoid having to repeat the experiment. Detection levels are important, but they generally do not change once established. Thus, once you know your limits, you can plan accordingly. So by a process of elimination, if nothing else, I would say that the quality of detection, by which I assume you mean the reliability, reproducibility, and confidence in positive identification, would be the most important parameter.

— Rebecca Lew

It is impossible to differentiate between the requirement for sensitivity and quality of detection. Both are essential. For us, at present, speed is of lower priority, the current bottleneck being the downstream proteomics analysis.

— Edouard Nice

I'm reluctant to single out any of these qualities as being most important. In reality, all are important in any given application. The extent of importance of each figure of merit changes based on the application. If searching for a protein biomarker in a complex mixture, sensitivity and selectivity are critical, and speed of analysis is of secondary importance. If assessing production efficacy for a recombinant protein produced using a well-characterized and well-understood approach, speed is paramount and selectivity and sensitivity are less important. For any given application, the critical parameters are different. Clearly, then, it is important to fully understand the application, the analytical methodology applied, and the desired data in order to successfully apply a separations method.

— Vincent T. Remcho

In order of importance:

• Quality of detection
• Detection levels
• Speed

To elaborate further and to qualify: quality of detection will affect detection levels. Speed or highthroughput can be reached provided quality of detection is not compromised. However, in case we're dealing with targeted analysis (single compound, for instance), then speed supersedes the quality of detection of the remaining compounds in the sample (not the targeted compound).

— Nabil Saad

List of Resources

We've mined the literature, the Internet, and even bookstores to unearth a variety of publications and online tools that may be able to help you wrap your head around your chromatography questions. Whether you're an old hand or if you're a newcomer to chromatography, something in here just may help you solve that pesky problem.

Publications

Choi BK, Ayer MB, Siciliano S, Martin J, Schwartz R, Springer JP. (2005) Interpretation of High-throughput Liquid Chromatography Mass Spectrometry Data for Quality Control Analysis and Analytical Method Development. Combinatorial Chemistry & High Throughput Screening. 8(6): 467-476

Qian WJ, Jacobs JM, Liu T, Camp DG, Smith, RD. (2006) Advances and Challenges in Liquid Chromatography-Mass Spectrometry Based Proteomic Profiling for Clinical Applications. Mol Cell Proteomics. 5(10): 1727-1744

Washburn MP, Wolters D, Yates, JR III. (2001) Large-scale analysis of the yeast proteome by multidimensional protein identification technology. Nature Biotechnology. 19: 242-247

Millea KM, Krull IS, Cohen, SA, Gebler, JC, Berger, SJ. (2006) Integration of Multidimensional Chromatographic Protein Separations with a Combined "Top-Down" and "Bottom-Up" Proteomic Strategy. Journal of Proteome Research. 5(1): 135-146

Desmet G, Cabooter D, Gzil P, Verelst H, Mangelings D, Heyden YV, Clicq D. (2006)
Future of high pressure liquid chromatography: Do we need porosity or do we need pressure? J Chromatogr A. 1130(1): 158-66.

Cabooter D, Heinisch S, Rocca JL, Clicq D, Desmet G. (2007) Use of the kinetic plot method to analyze commercial high-temperature liquid chromatography systems. I: Intrinsic performance comparison. J Chromatogr A. 1143(1-2): 121-33.

Clicq D, Heinisch S, Rocca JL, Cabooter D, Gzil P, Desmet G. (2007) Use of the kinetic plot method to analyze commercial high-temperature liquid chromatography systems II. Practically constrained performance comparison. J Chromatogr A. 1146(2): 193-201.

Websites

A Guide to HPLC:
http://www.pharm.uky.edu/ASRG/HPLC/hplcmytry.html

31st International Symposium on High Performance Liquid Phase Separations
http://www.Hplc2007.org

separationsNOW.com
http://www.separationsnow.com/coi/cda/home. cda?chId=4

Books

Practical HPLC Method Development.
Snyder LR, Kirkland JJ, Glajch JL.
Wiley-Interscience; 2nd edition (March 3, 1997)

Troubleshooting HPLC Systems: A Bench Manual.
Sadek PC.
John Wiley & Sons, Inc. New York, NY (2000).

Acknowledgments

We are grateful to Erol Gulcicek (Yale University), Ira Krull (Ben Gurion University), and John Yates (Scripps Research Institute), for their input and advice in formulating the questions for this guide.