Patent Granted for Flash Chromatography Technology

The Australian Patent Office has granted a patent related to W. R. Grace & Company's (Columbia, Maryland) Reveleris flash chromatography system. The patent concerns the use of evaporative light scattering detection (ELSD) with flash chromatography. The technology covered in this allowed claim enables streamlined purification of new chemical and biological entities for drug discovery. 

Grace previously received similar U.S. patents related to various aspects of flash chromatography using two kinds of detectors, such as ELSD and UV, to collect sample fractions. These techniques improve new molecule purification by increasing recovery, improving reliability, and simplifying instrument operation. 

Chris Reddy Receives Award for Oil Spill Research

Chris Reddy, a marine geochemist at the Woods Hole Oceanographic Institution (WHOI) (Woods Hole, Massachusetts), has been selected to receive the 2014 Clair C. Patterson Award from the Geochemical Society. The award recognizes one scientist a year who has made an innovative breakthrough in environmental geochemistry in service to society. 

Reddy was nominated for the award because of his work on developing and applying advanced chromatographic and spectrometric techniques to identify the source, transport, and fate of petroleum hydrocarbons in the coastal and open ocean. The award will be presented at Goldschmidt2014 (Sacramento, California) in June 2014. 

"One of my overarching interests in studying oil spills is that there are compounds in oil that seem to persist a very long time while there are others that can disappear quickly. I find it fascinating that a spectrum like this exists," said Reddy. He added: "Clair C. Patterson was a tenacious trailblazer in studying lead and lead pollution. For me to be associated with him is, quite frankly, humbling." 

Jeff Seewald, chair of WHOI's Department of Marine Chemistry and Geochemistry said: "His application and development of innovative analytical approaches have resulted in a quantum leap in our ability to predict the persistence of anthropogenic and naturally occurring petroleum in the marine environment." 

Seewald added: "Chris is motivated by a strong desire to understand the consequences of human activity on the environment and has a special gift for communicating relevant information to the scientific community, policy makers, and the general public." 

Articles

Departments

The Australian Patent Office has granted a patent related to W. R. Grace & Company's (Columbia, Maryland) Reveleris flash chromatography system.

Peaks of Interest

Many parameters within liquid chromatography–mass spectrometry (LC–MS) methods are "locked and left" — that is, they are optimized using a few methods when new, then only changed if absolutely necessary because of poor analytical performance. This is fine when methods are performing as they should be and producing data which is fit for purpose, but the following question should also be asked: Do you know when your data aren't as good as they could be? Here we consider some parameters in LC–MS that can be optimized to improve sensitivity and reproducibility in LC–MS. 

The most significant choice in many LC–MS methods is the mode of ionization. The generally accepted rule is that electrospray ionization (ESI) works best for higher-molecular-weight compounds that are more polar or ionizable, and atmospheric pressure chemical ionization (APCI) is best for lower-molecular-weight, less-polar compounds. Atmospheric pressure photoionization (APPI) was originally designed to work with less-polar analytes, however the careful choice of dopant or the use of direct ionization sources can significantly extend the capability of APPI techniques. In all atmospheric pressure ionization source types there are a multitude of interrelated factors that affect the degree of analyte ionization, including analyte p

 value, electronegativity, relative proton affinity, ionization energy, and volatility. "New" analytes should be screened using all available techniques to optimize instrument response.  Furthermore, although the "polarity" of the interface may seem obvious when screening lower-molecular-weight acids or bases, one can often be surprised at which polarity mode gives the optimum signal with more-complex molecules. Be sure to screen analytes in both polarity modes to ensure optimum response, especially when the analyte is not ionizable, and avoid "guessing" which mode might be best. 

The capillary (sprayer) voltage can have a major effect on efficiency of ionization, and it will change with the analyte type, eluent system, and flow rate. This parameter is perhaps the most overlooked variable in terms of improving instrument response and one should consider optimizing to improve not only sensitivity but also reproducibility. At a higher applied potential, one should be cautious to assess quantitative reproducibility because there are various nonideal spray modes that will give rise to a signal but will result in variable ionization efficiency. 

The efficiency of ion sampling from within the electrospray will vary widely with analyte type, applied voltage, solvent system, drying gas parameters, and eluent flow rate.  The position of the sprayer relative to the sampling orifice (both axially and laterally) should be optimized when the highest sensitivity is required.   

Nebulizing gas flow rate and heating requirements will change with the nature and flow rate of the eluent, and, in general, smaller droplets are preferred within ESI experiments to improve the efficiency of the droplet charging process. When changing eluent systems (especially the amount or nature of the organic component) or flow rate, consider optimizing the nebulizing gas settings. Similarly, drying gas requirements will change with the changing nature of the eluent system (including during gradient analyses where the range of organic composition alters greatly [large ΔΦ]) and should be optimized especially when eluent systems are highly aqueous. Electrospray efficiency is often improved by adding "dopants" such as isopropanol to allow a reduction in droplet surface tension; however, note that the sprayer position, nebulizing gas, and drying gas settings may need to be altered for optimum signal response. 

In ESI, analytes may be charged through processes such as charge transfer or triboelectric charging at the capillary tip. Although this may produce a satisfactory response, one should consider adjusting the eluent system to give the ionized form of the analyte in solution (eluent pH higher than analyte p

 for acids and lower for bases). This can result in orders of magnitude improvement in instrument sensitivity; however, it may also require an adjustment in high performance liquid chromatography (HPLC) operating conditions to properly retain the ionized analyte or reoptimize the separation selectivity. The use of ion-pairing reagents such as trifluoroacetic acid should be avoided and volatile buffers chosen so that the buffer p

 is within ±1 pH unit of the eluent system pH. 

Ion suppression or enhancement effects occur when there are species that strongly compete with, or are suppressed by, the presence of your analyte. When quantitative sensitivity is poor or the analyte signal is highly irreproducible, one should investigate the effects of the eluent system and matrix on the instrument response. These effects can often be overcome by altering the eluent system to reduce the ionic strength or change the buffer type, or by more studious sample preparation, designed to isolate the analyte from the interfering matrix component. 

The vast majority of LC–MS instruments are capable of analyte declustering, especially with respect to water molecule clusters. The application of an accelerating voltage in the first stage of the mass spectrometer to cause low energy collisions with background molecules can often lead to an improvement in instrument sensitivity, primarily due to a reduction in noise around the analyte signal. 

When performing MS-MS analysis, it is important when seeking optimum sensitivity and quantitative accuracy to optimize the collision energy as well as the "dwell" time between measuring ion transitions to avoid "cross-talk" effects that can significantly affect quantitative measurements. 

The Essentials

Secondary parameters in the interface and mass analyzer can have a major impact on sensitivity and reproducibility. Here, we examine how and when to consider optimizing these parameters through a study of the working principles of LC–MS analysis.

Optimizing LC–MS and LC–MS-MS Methods

Currently available methods for the detection of matrix effects in liquid chromatography–mass spectrometry (LC–MS) are tedious and complex; therefore, a simpler method is required. Although there are no methods to completely eliminate matrix effects, the most well-recognized technique available to correct for matrix effects is that of internal standardization using stable isotope–labeled versions of the analytes. As this method can prove expensive, an alternative method of correction is likely to be useful. In this study, a simple method based on recovery is assessed for the detection of matrix effects. Two alternative methods for the rectification of matrix effects in LC–MS are also assessed: standard addition and the coeluting internal standard method. 

High performance liquid chromatography (HPLC) coupled to mass spectrometry (MS) has become the predominant analytical method for the quantitative determination of analytes in biological matrices because of its high specificity, sensitivity, and throughput (1–3). However, matrix effects have become a major concern in quantitative liquid chromatography–mass spectrometry (LC–MS) because they detrimentally affect the accuracy, reproducibility, and sensitivity (3). Matrix effects occur when compounds that are coeluted with the analyte interfere with the ionization process in the MS detector, thereby causing ionization suppression or enhancement (2–7). Compounds with high mass, polarity, and basicity are possible candidates to cause matrix effects (4–8). However, the mechanisms involved in matrix effects have not been fully explored. One of the proposed theories to explain matrix effects is that the coelution of interfering compounds, especially basic compounds, may deprotonate and neutralize the analyte ions and, thus, reduce the formation of protonated analyte ions (2,4). Another theory postulates that less-volatile compounds may affect the efficiency of droplet formation and reduce the ability of charged droplets to convert into gas-phase ions (2–4,8). In addition, matrix effects may also be caused by high viscosity interfering compounds that could possibly increase the surface tension of the charged droplets and reduce the efficiency of droplet evaporation (2,4,6).  

Several methods have been proposed for the detection and assessment of matrix effects, including postextraction spike and postcolumn infusion methods. The postextraction spike method evaluates matrix effects by comparing the signal response of an analyte in neat mobile phase with the signal response of an equivalent amount of the analyte in the blank matrix sample spiked post-extraction. The difference in response determines the extent of matrix effect (2,3,9). The major drawback of this method is that for endogenous analytes such as metabolites (for example, creatinine) blank matrix (urine or plasma) is not available. The postcolumn infusion method assesses matrix effects qualitatively. A constant flow of analyte is infused into the HPLC eluent, followed by injection of the blank sample extract. A variation in signal response of the infused analyte caused by coeluted interfering compounds indicates ionization suppression or enhancement (3,10). By identifying the ionization suppression or enhancement regions of the chromatogram, analytical methods can be developed to eliminate matrix effects by preventing the elution of the analyte peak in regions where matrix effects occur. However, the process of postcolumn infusion is time consuming and requires additional hardware, and it is not appropriate for multianalyte samples. Considering these drawbacks of existing methods, we propose a simple, fast, and reliable method to detect matrix effects that can be applied to any analyte including endogenous compounds such as creatinine and to any matrix without requiring any additional hardware. 

To obtain accurate and reliable LC–MS data, several methods have been suggested to reduce or eliminate matrix effects. Matrix effects can be reduced simply by injecting small amounts of samples or by diluting samples (11,12). However, this approach can only be feasible when the sensitivity of the assay is very high (12). Methods to reduce or eliminate matrix effects include optimizing sample preparation to remove interfering compounds from the samples (1,9,10,13), changing chromatographic parameters to avoid coelution of analytes and interfering compounds (4,14–17), and changing MS conditions to reduce the occurrence of matrix effects in the ion source. However, these methods are not without their limitations. Most of the sample cleanup methods fail to remove impurities that are similar to the analyte and, hence, likely to be coeluted with the analyte (11,18). Modifying chromatographic conditions can be time-consuming, and some of the additives used in the mobile phase to improve separation have been found to suppress the electrospray signal of the analytes (3,4,9,15). Furthermore, even when the sample is devoid of coeluted substances, trace impurities present in the mobile phase can significantly suppress the analyte peak (19). 

It is clear from the above that matrix effects in LC–MS cannot be completely eliminated. Therefore, the only option available is the rectification of data to eliminate the matrix effects. Calibration techniques such as the external-matched standards method, the echo-peak technique, and the most commonly used approach, the internal standard method, have been developed to correct the data (15,18–21). However, these calibration techniques also have their drawbacks. For example, the matrix-matching technique requires many blank matrices and appropriate blank matrices are not always available for the preparation of external standards (9,11,18,23). It is also impossible to match the matrix of the calibration standards with each of the samples exactly, as each sample has coeluting, interfering compounds that are thereby exposed to a different extent of ionization suppression (18). Echo-peak does not compensate for matrix effects completely because both standard and analyte peaks are not eluted at the exact same retention time (11). The stable isotope–labeled internal standards (SIL-IS) approach is the best available option but it is expensive and standards are not always commercially available for the analyte of interest (4,9,23).  

The standard addition method for correcting matrix effects is widely used in spectrophotometric analysis, especially in atomic spectroscopy (24–27). However, this method is less well documented with other analytical techniques and currently there is no record of its practical use in compensating matrix effects in LC–MS. Standard addition does not require a blank matrix and is therefore appropriate for compensating matrix effects for any analyte including endogenous metabolites in biological fluids (20,28,29). In this study, we investigated the possibility of using the method of standard addition in routine LC–MS analysis to compensate for matrix effects and to thereby obtain improved data. We also investigated the use of a coeluting structural analogue of the analyte as the internal standard as an alternative to the expensive and often unavailable stable isotope–labeled internal standard for correcting matrix effects. Although coeluting structural analogue compounds are used to extend the linear range of calibration curves (30), there have been no reports of their use in compensating matrix effects in routine LC–MS analysis. In addition, we report a simple and effective method for the detection as well as the correction of matrix effects in routine LC–MS. All studies were carried out using a creatinine assay applied to human urine samples. 

Separation of compounds was achieved with an Agilent 1100LC binary pump and Agilent 1100 autosampler (Agilent Technologies). The HPLC column used was a 150 mm × 2.1 mm, 4-μm 

 Cogent Diamond-Hydride 100A column (MicroSolv Technology). For detection and quantification, an API 3000 tandem mass spectrometer equipped with a turbo ion spray interface and the software program Analyst 1.5 (Applied Biosystems) was used. 

 (2-amino-1-[trideuteriomethyl]imidazolidin-4-one), and cimetidine (

'-[2[(5-methyl-1H-imidazol-4-yl)methylthio]-ethyl]guanidine) were purchased from Sigma. HPLC-grade acetonitrile and deionized water from a Milli-Q water system (Millipore) were used for mobile-phase preparation. Human urine samples were obtained from volunteers. 

Separations were carried out at an ambient temperature of approximately 25 °C. The flow rate was 200 μL/min. The injection volume was 10 μL. Mobile-phase A consisted of deionized water containing 0.1% (v/v) formic acid, and mobile-phase B comprised 0.1% (w/v) formic acid in acetonitrile. 

Detection of Matrix Effects and Standard Addition Method

Gradient elution was used for analyte separation. The mobile-phase composition was varied from 90% B to 50% B over the first 20 min, and then maintained at 50% B for 1 min, before returning to 90% B from 21 to 24 min. The percentage of B was maintained at 90% B for 10 min for re-equilibration before the next sample was injected. The total run time was 24 min. 

An isocratic elution method was used for elution. The mobile-phase composition was 55% B over 10 min. The total run time was 10 min. 

, and cimetidine were detected in a positive multiple reaction monitoring (MRM) mode in which the transitions of 

 252.8 to 95.1 were monitored, respectively. The MS and electrospray ionization parameters were optimized and the analyzer settings were as follows: ion spray voltage (IS) 5000 V; entrance potential (EP) 10 V; orifice/declustering potentials (DP) 26 V (creatinine), 36 V (creatinine-

), and 36 V (cimetidine); ring/focusing potentials (FP) 140 V (creatinine), 250 V (creatinine-d3), and 140 V (cimetidine); collision energy (CE) 29 V (creatinine, creatinine-

 and cimetidine); collision exit potential (CXP) 8 V (creatinine, creatinine-

, and cimetidine). Curtain gas (CUR), nebulizer gas (NEB), and the collision gas (CAD) flows were kept at 12, 8, and 4, respectively (in arbitrary units used in the instrument). The temperature of the ion spray was 300 °C. A dwell time of 150 ms was used for all transitions. Resolution of both Q1 and Q3 were 1 amu. 

Human urine samples were prepared by filtering a small amount of the urine through a 0.22-μm polytetrafluoroethylene (PTFE) filter (Millipore). 

The filtered urine sample was diluted 10-fold by mixing 30 μL of filtered urine and 270 μL of deionized water. This was followed by 100-fold dilution by mixing 900 μL of acetonitrile, 10 μL of internal standard (0.05 mM creatinine-

 or 0.05 mM cimetidine), 10 μL of the diluted filtered urine sample, and 80 μL of deionized water to a total volume of 1000 μL. 

Preparation of Standard Solutions and Quantification

Stock solutions (1.00 mg/mL) of creatinine, deuterated creatinine, and cimetidine were prepared in deionized water and stored at -20 °C. 

Detection of Matrix Effects Using Recovery 

Standard solutions were prepared by mixing appropriate volumes of 1.0, 0.1, and 0.01 mM creatinine, 900 μL of acetonitrile and deionized water to a final volume of 1000 μL. The concentrations of standards were 0.0001, 0.0003, 0.0005, 0.001, 0.003, 0.006, 0.01, 0.015, and 0.02 mM of creatinine in 90% (v/v) acetonitrile. A calibration curve of peak area of creatinine versus creatinine concentration was obtained. 

Creatinine was added to five filtered urine samples and then diluted 1000-fold to a final added creatinine concentration of 0.006 mM. The creatinine concentrations of spiked and unspiked urine were determined using the above calibration plot to calculate the assay recovery. 

Comparison of the Stable Isotope–Labeled Internal Standard and Standard Addition Methods 

Standard solutions for the SIL-IS method were prepared by mixing 10 μL of 0.05 mM creatinine-

, 900 μL of acetonitrile, appropriate volumes of 1.0, 0.1, and 0.01 mM creatinine, and deionized water to make the final volume up to 1000 μL. The final concentration of creatinine-

 in all standards was 0.0005 mM and the concentrations of creatinine were 0.0001, 0.0003, 0.0005, 0.001, 0.003, 0.006, 0.01, 0.015, and 0.02 mM in 90% (v/v) acetonitrile. A calibration curve of the ratio of the peak area of creatinine and the peak area of creatinine-

 versus creatinine concentration was then plotted to determine the concentration of the above five aliquots of 1000-fold diluted urine. 

For the standard addition method, two levels of additions were prepared. A 0.003 mM addition was prepared by mixing 10 μL of 10-fold diluted filtered urine, 30 μL of 0.1 mM creatinine, 900 μL of acetonitrile, and 60 μL of water to make the final volume up to 1000 μL. The 0.006 mM addition was prepared by mixing 10 μL of 10-fold diluted filtered urine, 60 μL of 0.1 mM creatinine, 900 μL of acetonitrile, and 30 μL of water to make the final volume up to 1000 μL. The zero addition was prepared in the same manner but water was added in place of the 0.1 mM creatinine solution. A calibration plot of peak area of creatinine versus the added creatinine concentration was plotted for each sample and the original concentration of creatinine in five samples were estimated by extrapolating the line of best fit to zero peak area. 

A negative control experiment was also carried out by running five samples containing no internal standard and determining their concentration using the calibration curve prepared in the section titled "Detection of matrix effects using recovery" above. 

Comparison of the Stable Isotope–Labeled Internal Standard and Coeluting Internal Standard Methods

 in all standards was 0.0005 mM and the concentrations of creatinine were 0.0001, 0.0003, 0.0005, 0.001, 0.003, and 0.006 mM in 90% (v/v) acetonitrile. 

Standard solutions for the coeluting internal standard method were prepared by using 0.05 mM cimetidine in place of creatinine-

 and by following the procedure mentioned in the previous section to obtain similar final concentrations of creatinine and the internal standard. A calibration curve of the ratio of the peak area of creatinine and peak area of the internal standard versus creatinine concentration were then plotted for each internal standard to calculate the concentrations of creatinine in seven aliquots of 1000-fold diluted urine. A negative control experiment was run in the same manner as described in the previous section. 

To eliminate matrix effects, they should be assessed during early analytical method development. In this paper, we propose a simple and reliable method, recovery value, to estimate the extent of matrix effects in biological matrices analysis. In the recovery value test, the original concentration of creatinine in each of the five urine samples was determined using a calibration plot of the peak area of creatinine versus creatinine concentration. Each of the urine samples was then spiked with a known concentration of standard, 0.006 mM creatinine, and the total creatinine concentrations of the spiked samples were calculated using the same calibration curve. The percent recovery was determined by using the following formula:  

 is the added known concentration of standard; 

 is the concentration of analyte in final solution after spiking with known concentration of standard; and 

 is the original concentration of analyte in initial solution. 

When percent recovery is closer to 100 (within normally expected error of ±3 standard deviations [SD]), it can be deduced that there is minimal or no matrix effect. A value of percent recovery less than 100 indicates that ionization suppression is present and a value greater than 100 indicates ionization enhancement.  

As presented in Table I, the mean percentage of recoveries for the five urine samples was 65.98%. This result shows that the matrix effects profoundly affected the urine samples and thus degraded the accuracy of the quantitative LC–MS analysis. A high percentage of matrix effects was anticipated because the sample cleanup for the urine samples involved only filtration and dilution in this experiment and there was no mechanism in place to compensate for the resulting matrix effects. Furthermore, the results of subsequent experiments described later in this paper confirm the presence of matrix effects. The use of the recovery value to detect matrix effects is simple and cost-effective since it requires only two LC–MS runs. Also, it is ideal for the determination of the concentration of analytes in biological fluids that have complex matrices since it does not require a blank matrix and only involves the addition of analytes into the same sample matrix. Thus, it is certainly an easier and simpler method when compared to other methods available such as post-column infusion and post-extraction spike method, as discussed above. Unlike post-column infusion, that assesses the matrix effects in a qualitative manner, the recovery value method provides quantitative data.  

Table I: The percent recoveries of creatinine in five human urine samples. C0 is the known concentration of creatinine standard added to sample, 6 mM; C1 is the concentration of creatinine measured in the unspiked urine sample; C2 is the concentration of creatinine measured after spiking (to a final spiked concentration of 6 mM). Recovery calculation procedure is explained in the results and discussion section.			

Comparison of Stable Isotope–Labeled Internal Standard and Standard Addition Methods

Because of its proven accuracy and reproducibility, the SIL-IS method was chosen as the reference (or positive control) method and was thus compared with other methods. In the comparison of SIL-IS and standard addition methods, the SIL-IS method was run and analyzed using the same chromatographic and MS conditions as that of the standard addition method. The calibration curve of the peak area ratio of creatinine to that of creatinine-

 (SIL-IS) versus creatinine concentration was linear with an 

value of 0.989 up to a concentration of 0.015 mM creatinine. 

Although standard addition is a method that is widely used in atomic spectrophotometric analysis, it has not been so used in quantitative LC–MS analysis. In this study, we investigated the possibility of using standard addition to rectify matrix effects in LC–MS. Three runs were performed for each sample (as described in the experimental section), but there was no need for calibration standards; that is, each sample has its own calibration curve with three points. A calibration plot obtained for a typical sample in this study is shown in Figure 1. The calibration line was extrapolated to zero response to estimate the original concentration of creatinine in the urine sample.  

Figure 1: Standard addition method. A three-point calibration line is extrapolated to zero response to estimate the original concentration of creatinine present in the urine sample.			

If the linear range of the calibration of analyte is already established, and the concentration of analyte in both the added and original samples fall within that range, an easier alternative approach for standard addition is to do only one addition and calculate the only unknown in the following equation, the original analyte concentration in the sample:  

 is the original concentration of analyte in initial solution, [S] is the known concentration added, 

 is the signal response from initial solution, and 

 is the signal response from the final solution after the addition of a known concentration of standard. 

Ellison and Thompson (31) highlighted that standard addition can only be precise and feasible when the analytical calibration curve is linear throughout the targeted concentration range. From our earlier experiments it has been demonstrated that the calibration curve for creatinine was linear to a concentration of 0.015 mM, and we were therefore able to estimate the original concentration of analyte in the samples by using equation 2. 

Peer-Reviewed Article

We assessed a simple method based on recovery for the detection of matrix effects and two alternative methods for the rectification of matrix effects in LC–MS: standard addition and the coeluting internal standard method.

Strategies for the Detection and Elimination of Matrix Effects in Quantitative LC–MS Analysis

Here we describe novel application areas in which hydrophilic interaction liquid chromatography (HILIC) has vast importance and utility, including peptides, glycopeptides, two-dimensional liquid chromatography in proteomics, oligonucleotides, lipids and lipidomics, glycan analysis, proteins, glycoproteins, phosphoproteins, and related biopolymers. We also describe some specific HILIC conditions that have proven useful for these classes of biologically active analytes and provide typical chromatograms from the recent literature, both scientific and commercial. 

In part I of this two-part series on modern hydrophilic interaction liquid chromatography (HILIC) applications of relevance to the biotechnology industry, we discussed the fundamental principles and mechanisms suggested from the literature for this, now very widely accepted and practiced form of modern high performance liquid chromatography (HPLC) and ultrahigh-pressure liquid chromatography (UHPLC). We also discussed just a few of the common applications already described for simple and complex glycans often derived from glycopeptides or glycoproteins or for synthetic, standard glycans. We discussed the current roles of HILIC in glycoprofiling and glycan analysis from any source, but in particular by deglycosylation of proteins and peptides. In part II, we discuss modern usage and specific applications of HILIC for two other major classes of analytes important in biotechnology today: glycopeptides and peptides, and glycoproteins and proteins. Other types of proteins, such as phosphoproteins and their applications, are also discussed. Finally, we propose some possible future applications in which HILIC can prove itself useful, including complex glycans, glycomics, oligo-drugs (for example, heparins), glycosaminoglycans, and others.  

HILIC stationary phases were developed with the aim of providing a tool for the analysis of hydrophilic compounds, which are poorly retained in reversed-phase liquid chromatography (LC) and, therefore, are not ideally resolved (1). Although the elution order of analytes in HILIC is generally reversed compared to that in reversed-phase LC (that is, the most hydrophobic compounds are eluted first in HILIC and the most hydrophilic are eluted last), it has been recognized that the retention order is not inversely proportional (2). Because of various interactions in the HILIC mode that contribute to the separation and selectivity in reversed-phase LC and HILIC, the techniques to a large degree are orthogonal (3,4). 

Besides the analysis of small molecules, many early HILIC applications investigated the separation of peptides. Some of these include those of Alpert (1), Yoshida (5), and Mant and colleagues (6). More recently, Gilar and Jaworski (3) compared three HILIC stationary phases and concluded that for the separation of charged species, such as peptides, the interactions with the charged silanols on silica significantly contributed to the retention mechanism. 

The separation orthogonality of HILIC is often used for the analysis of compounds that are difficult to resolve in reversed-phase LC. For example, Zhu and colleagues (2) have used both reversed-phase LC and HILIC for analyses of peptide maps of erythropoietin and found that the methods were complementary. HILIC was more suitable for the detection and quantification of small or hydrophilic peptides and also glycosylated peptides that were not adequately resolved in reversed-phase LC. Moreover, HILIC peptide maps contained some unique, very hydrophobic peptides that were not detected in reversed-phase LC. 

HILIC was also applied to the quantitative bioanalysis of therapeutic peptides and proteins. This class of biological therapeutics holds great potential; however, the analysis has been challenging. Biological fluids contain many peptides, especially when the analysts perform proteolysis of serum proteins before liquid chromatography–mass spectrometry (LC–MS). Two-dimensional liquid chromatography (2D LC), reversed phase × HILIC, combined with multiple reaction monitoring mass spectrometry (MS), was used for the quantitation of peptides of interest. Limits of detection below 10 ng/mL were achieved (7).  

As stated above, HILIC and reversed-phase LC separation modes are orthogonal, and they can be utilized for 2D LC. D'Attoma and colleagues (8) adopted HILIC as the second dimension of 2D LC for the separation of protein digests. They reported an effective peak capacity of 2600 in a 180-min analysis. A HILIC × reversed-phase 2D LC setup was also evaluated by Di Palma and colleagues (10) for the separation of complex proteomic samples. The authors observed a favorable separation of peptides on zwitterionic HILIC in the first dimension compared to the more typical, first dimension — strong cation-exchange chromatography. Importantly, the authors noted a potential for enrichment of phosphopeptides in HILIC × reversed-phase 2D LC (9,10). McNulty and Annan (11) also investigated the potential of HILIC × reversed-phase 2D LC for proteomic analysis (11). Similar to previous reports, they observed significant advantages of HILIC for phosphoproteomic analysis. Because of charge–charge interactions with the zwitterionic stationary phase, or simply because of the more hydrophilic nature caused by phosphate groups, the phosphopeptides were eluted in a later elution time window when compared to general peptides (9–11). Phosphopeptides also appeared to be better recovered and more detectable in the HILIC × reversed-phase 2D LC method (11). A similar observation was made by Alpert (12), who proposed that the solute's interaction with the charged, hydrophilic sorbent may affect their elution patterns. This separation mode was then termed 

electrostatic repulsion hydrophilic interaction chromatography

 (ERLIC), and it can be used for the selective isolation of phosphopeptides from a protein digest. To further this observation, Sze (13,14) has used ERLIC–weak anion-exchange approaches (HILIC on weak anion-exchange surfaces) for selectively retaining phosphopeptides, free of other, acidic peptides, by operating at pH 2, where the phosphate group retains a negative charge and peptide carboxyls are neutral. 

HILIC was recently applied to the separation of another class of therapeutic molecules: oligonucleotides. These molecules (with a hydrophilic, phosphate backbone) were separated on a prototype hydroxymethyl methacrylate based monolithic column (15), or on commercially available HILIC columns (16). The mobile phase included triethylammonium acetate (TEAA), an ion-pairing system often used in reversed-phase LC. However, it is not clear to what extent the ion pairing contributes to oligonucleotide resolution in the HILIC mode. Keeping in mind the mobile-phase compatibility for LC–MS analysis, Gong and McCullagh (17) investigated such resolutions with other buffers, such as ammonium acetate, bicarbonate, or formate. They achieved a usable separation of an 18-mer oligonucleotide from its 18-mer, single methylated version, with a 100 mM ammonium acetate buffer on a 100 mm × 2.1 mm, 3.5-μm 

 HILIC column. From these reports, it appears that HILIC can resolve oligonucleotides up to a 20-mer length (15). This was a poorer, separation performance than achieved with state-of-the-art reversed-phase LC columns (18). It remains to be seen whether HILIC can provide useful selectivity for the separation of phosphorothioate, phosphodiester, or depurinated species, because these are relevant impurities of oligonucleotide drugs that are often unresolved by reversed-phase LC. 

HILIC has also found applications in lipidomic research, particularly for the characterization of polar lipids (19–21). Because of its nature, HILIC is capable of resolving lipids (phospholipids) according to their polar head group. The lipids are resolved into classes according to their polarity (Figure 1), and the species are separated according to the hydrophobic portion of the lipids (fatty acid chain length), which is typically not achieved (21). On the other hand, the separation of fatty acids of different lengths and degree of unsaturation (number or position of double bonds), is readily achieved in reversed-phase LC. Therefore, the HILIC × reversed-phase 2D LC–MS method is currently being explored for advanced polar lipids analysis (21).  

Figure 1: Analysis of polar ceramide lipids. Two columns, HILIC and reversed phase, were connected in series. Lipids were eluted with a methanol gradient. Mobile-phase A: 95% acetonitrile, 5% methanol, 5 mM ammonium formate, and 0.2% formic acid. Mobile-phase B: methanol, 5 mM ammonium formate, and 0.2% formic acid. Ceramides were eluted with a methanol gradient at a flow rate of 250 Î¼L/min. The HILIC first dimension resolves analytes according to their hydrophobic glycan group; the reversed-phase column separates fatty acid chains according to their length while being insensitive to glycan moiety. Identification was performed by MS. (Adapted with permission from reference 21.) 			

HILIC shows significant promise in the analysis of protein glycosylation. As discussed in the first part of this series (22), HILIC is a very useful method for the analysis of released glycans, together with fluorescent detection. This approach provides both high resolution and sensitivity, but has limitations when used for site specific glycosylation analysis. Upon cleavage of glycans from proteins, one loses information about the original site heterogeneity. In other words, if two glycosylation sites contain the same glycan, these species are pooled before analysis, and the original level of glycosylation at each site remains unknown. That is why some researchers explored an alternative approach based on proteolysis of the glycoprotein followed by analysis of the resulting glycopeptides (23,24). The selection of methods used for glycoprotein analysis via glycan or glycopeptides routes is shown in Figure 2. The unique peptide "mass tag" attached to the glycans can identify the nature and degree of glycosylation at each site (4,25). A second major advantage of the glycopeptide based approach, is the ability to directly interrogate O-linked glycans, thus removing the need for specific O-linked deglycosylation enzymes, which are not readily available at this time (24,26,27).  

Figure 2: Glycoprotein analysis scheme. When applicable, HILIC technique is highlighted in red.			

Glycopeptide analysis utilizes two unique HILIC features. First, the glycosylation (addition of a highly polar moiety) causes a marked enhancement of peptide retention in HILIC (Figure 3). Consequently, the glycosylated peptides are typically resolved from nonglycopeptides present in the protein digest (4,24). Second, the glycopeptides are separated into their glycoforms, including positional isomers of glycans (Figures 3 and 4). This is useful for both ultraviolet (UV) and MS analysis, given the isobaric nature of isomers and potential for in-source fragmentation of labile glycan moieties (4).  

Figure 3: HILICâUVâMS analysis of tryptic digests of (a) mAb and (b) human 1-alpha-acid glycoprotein. Glycopeptides are more retained in HILIC (highlighted in blue). The elevated collision energy MS spectrum is shown in the bottom panel for the peak labeled with an asterisk. The glycan moiety is fragmented into characteristic ions, confirming that the selected peak is indeed a glycopeptide. For chromatographic conditions and details see reference 4. Adapted with permission from reference 4.			

Several groups have adopted HILIC for systematic studies of glycoproteins. Two distinct schools of thought can be observed. Some researchers prefer using specific enzymes for protein digestion such as trypsin, Glu-C, or Lys-C. In such an approach, the proteolysis creates highly specific glycopeptides and, desirably, one peptide per glycosylation site (4,23,24,28–32). These peptides are then analyzed by HILIC-MS and the heterogeneity and nature of glycosylation are revealed. Unfortunately, in some cases the proteolysis creates long peptides with masses of 3–6 kDa, which do not present easy targets for MS analysis. The lack of suitable cleavage sites can also create peptides with two or more glycosylation sites and compromise the method's capacity for site-specific glycosylation analysis (4).  

Figure 4: HILIC MRM analysis of tryptic digest of purified murine IgG1 antibody. EEQFNSTFR tryptic peptide glycoforms are resolved in HILIC (for conditions, see reference 4). MRM permits selective detection and quantification of all glycopeptide species, even chromatographically unresolved ones. The quantitation results were comparable with the HILIC-fluorescence method (39). (Adapted with permission from reference 39.)			

A second approach uses nonspecific proteolysis with Pronase or Proteinase K enzymes (25–27,33,34). The cleavage creates short, 2–6 amino acid long glycopeptides, which lend themselves well to MS analysis. In addition to labile glycan moiety fragments (most abundant are 204.09, 366.14, and 657.24 Da, see inset in Figure 3b), the collision-induced dissociation (CID) spectrum of a glycopeptide contains signals of the peptide or peptide+GlcNac ions (31–33,35). These both provide information about the glycosylation site. The site identification is typically confirmed by multiple peptides generated by nonspecific enzymatic cuts around the glycosylation site. However, the multitude of signals complicates the HILIC-MS data interpretation (25,26,34) and dilutes the available signal. Low-abundance glycans may not be detected (33). 

Besides analysis in the column format, HILIC solid-phase extraction (SPE) methods have been utilized for glycopeptide enrichment from complex peptide samples (Figure 2). This has become a very popular and useful overall approach for improved glycan analysis and glycoprofiling. Micro-SPE tips (30), short HILIC guard columns (29), and a batch SPE approach (28,29) have been described. Although some authors prefer multilectin specific enrichment of glycopeptides or glycoproteins (36), the simple HILIC glycopeptide enrichment has gained in popularity (30,37). One should be aware, though, that some hydrophobic glycopeptides with a short glyco moiety (O-linked peptides), may not be sufficiently resolved from hydrophilic nonglycopeptides. HILIC SPE glycopeptide enrichment is a fast and convenient method, but it does not remove all nonglycosylated peptides from the enriched glycopeptides (4,30). 

Several groups have successfully used the HILIC glycopeptide approach for analysis of singly glycosylated proteins such as monoclonal antibodies and IgGs (4,23,29,32). However, the analysis of multiple glycosylated proteins remains a difficult task. Multiple glycoforms of each glycopeptide share the same HILIC separation space, and thus challenge the separation capability of HILIC. Even when MS is used, it is difficult to perform the data interpretation (4). This is especially the case for samples like vaccines that contain multiple viral hemagglutinin isoforms as their immunogenic proteins (35,38). A combination of 2D LC and MS may provide a suitable solution. Reversed phase × HILIC 2D LC relies on the fact that reversed-phase LC resolves glycopeptides by the peptide hydrophobicity, whereas the peptide glycoforms are eluted as a single peak or a closely spaced group of peaks (29). HILIC, on the other hand, can resolve a given glycopeptide into its glycoforms (Figures 3 and 4). Off-line 2D LC with the first dimension consisting of reversed-phase LC and the second dimension of HILIC–UV–MS was evaluated (4). MS offers an alternative to UV detection, as well as a highly desirable sensitivity for the detection of glycopeptides, often present at significantly lower concentrations than other peptides in the digest. The concentrations are lower because only a certain fraction of proteins could be glycosylated and because the glycopeptide signal is split into subspecies as a result of the glycan site microheterogeneity. Minor glycoforms are often present at levels less than 10%. Sensitive triple-quadrupole MS detection, utilizing multiple reaction monitoring (MRM) of glycopeptides, was demonstrated (39), capable of the detection of minor glycoforms (Figure 4). The MRM sensitivity was comparable with the HILIC–fluorescence technique (39,40), and it permits the quantitation of glycoforms that are chromatographically unresolved. It remains to be seen whether this promising approach will be adopted as a workhorse method for glycosylation analysis. 

Finally, the complicated task of metabolomic profiling can take advantage of the 10–100× increased sensitivity of HILIC available with LC–MS detection. Also, as can be done with any set of small molecules that differ by polar group composition, one can take advantage of the multimodal mechanisms available to HILIC separations. This quality makes it particularly suited for better selectivity by amplifying these types of differences rather than using reversed-phase LC or other modes of partition chromatography. 

HILIC has been described and utilized for more than two decades and while a substantial amount of literature covers the field and there is substantial commercial interest in its applications and usage, most of this has involved, so far as biotechnology is concerned, glycan (glycoprofiling) (22) and glycopeptide analyses (vide supra). Much less of an effort has been described for glycoproteins or larger species (41–54). Very little usage has been made for perhaps the most important player in biotechnology today: recombinant antibodies (MAbs). In general, HILIC has been applied far more for low-molecular-weight species, conventional drugs, forensics, natural products, and virtually any analyte that has properties compatible with HILIC's separation strengths. This means, in general, that the analyte must be fairly hydrophilic, polar, partially ionic, or a combination of all three of these separation properties. It must also be fairly water or alcohol soluble, and less organic miscible, or have a mixture of both aqueous and organic solubilities. 

In glycan and glycopeptide separations, it is the carbohydrate portion of these molecules that makes the analytes suitable for HILIC and perhaps less suitable for reversed-phase LC. In the case of glycopeptides, they are suitable for both reversed-phase LC and HILIC approaches because their organic and aqueous solubilities and physicochemical interactions with the mobile phases are approximately equal. This may well be why 2D LC incorporating both HILIC and reversed-phase LC offers such important advantages. In the case of glycoproteins or higher order species (MAbs) this is no longer the case and their organic solubilities usually overshadow aqueous miscibility and physicochemical interactions. Glycoproteins are more organic-like than aqueous, and hence they are much more miscible with organic solvents than with water or alcohols.  

Thus, it is no surprise that the vast majority of HILIC applications in the biotechnology areas, so far at least, have been glycans and glycopeptides. This situation may well remain true until we discover how to make HILIC much more glycoprotein friendly than it is at present, perhaps by using multiphase HILIC supports. However, that approach then introduces ion-exchange separation modalities, perhaps more so than HILIC interactions, which then muddies the interpretation of whether the separation mechanism is pure HILIC or if it is mixed mode. In any event, the number of literature reports or application notes wherein HILIC has been used for glycoproteins, or proteins in general, remains far lower than for glycans and glycopeptides. However, there are some very useful and interesting reports, which we will now proceed to relate. 

Much of the initial theory and methods for HILIC of peptides or proteins originated with Alpert of PolyLC, often in collaboration with various academic colleagues. In several of their presentations and commercial literature or applications, specific protein results are attributed as arising from such colleagues. For example, in 2006 Carroll and colleagues (53,54) described a number of different types of proteins successfully separated by HILIC methods. Figure 5 provides a very nice chromatogram for a mixture of intact mitochondrial membrane proteins. The sample was an organic extract of bovine heart mitochondria, with a very large variety of different proteins eluted in about 1 h. Because of the nature of the two phases, this is a classic HILIC type separation. Figure 5 also indicates the names of some of the proteins involved, all of which were eluted in 60 min, with the gradient conditions indicated. This particular work is discussed in greater detail below.  

Figure 5: HILIC of intact mitochondrial membrane proteins (47). Column: 100 mm Ã 2.1 mm PolyHydroxyethyl A; gradient: 20 mM ammonium formate (pH 3.7) with 0.5% HFIP; (63% 2-propanol, 22.5% acetonitrile) to 30% 2-propanol. Sample: 2-propanolâacetonitrile (pH 3.7) extract of bovine heart mitochondria. (Reproduced with permission of PolyLC and John Walker, Medical Research Council Dunn Human Nutrition Unit, Cambridge, UK, 2006.)			

Another illustration of HILIC protein separations is given in Figure 6, which works well for membrane-type proteins (55). Again, a linear gradient was used. One other, early example of the capabilities of HILIC, for difficult to resolve membrane proteins, is shown in Figure 7, again from PolyLC (56). In this particular application, the same HILIC column was used with somewhat different gradient elution conditions, as given. The identified proteins are indicated in Figure 7. The total time for elution was <40 min. Several other applications of HILIC for different protein types have been described over the years in applications from PolyLC and others (for example, histone H4 acetylation and methylation variants, and cation-exchange HILIC on a PolyCat A column with a gradient of sodium perchlorate in 70% acetonitrile was able to resolve histone H1.5 phosphorylation variants) (57).  

Figure 6: HILIC works well for membrane proteins, with removal of SDS, Coomassie Blue, and salts from an electroeluted membrane protein. Column: 200 mm Ã 4.6 mm, 5-Î¼m dp PolyHydroxyethyl Aspartamide; gradient: 70â0% n-propanol in 50 mM formic acid. (Reproduced with permission from A. Alpert, PolyLC, and P. Jeno, University of Basel, Basel, CH.)			

As already suggested, PolyLC probably has the largest number of HILIC-based protein separations of any commercial vendor and these proteins are usually difficult to resolve using other HPLC approaches. Thus, on one of their website pages (58) the company describes several such valuable applications as follows. One such application involves using strong cation-exchange HILIC for lung surfactant proteins and their variants, again using the company's PolySulfoethyl A column, with a mobile-phase gradient combining methylphosphonic acid, sodium perchlorate, acetonitrile (mobile-phase A) and 100 mM sodium perchlorate as mobile-phase B. In another application, phosphorylation variants of histone H1 were separated on a PolyCAT A column using 70% acetonitrile in the mobile phases, and resolution was excellent, with a reversal in the order of elution when compared with 0% acetonitrile. Other histone H1.5 phosphorylation variants were similarly resolved (58,59). HILIC modes were superior to capillary electrophoresis for separation of these particular protein variants.  

Figure 7: Top-down proteomics of hard samples, HILIC of mouse brain membrane proteins. Column: PolyHydroxyethyl A ; gradient: 70â40% acetonitrile in Tris-HCl (pH 6.5) with 40 mM HFIP; sample: brain homogenate pellet extracted with HFIP. (Reproduced with permission from A. Alpert and S.D. Garbis, Academy of Athens, ASMS 2005 poster #ThP 563.)			

Histone H4, acetylation, and methylation variants were well resolved on a PolyCAT A column, operated in the HILIC mode (58). This particular column has both cation-exchange and HILIC properties, so that hydrophobic interactions can be superimposed on the electrostatic effects. The column was also able to separate variants with the same number of acetylated Lys residues, but differing in the number of methylated Lys residues. The identification of these variants was only possible by using this top-down separation step before digestion of individual resolved variants and bottom-up peptide mapping identification (59). If not fully resolved, they would have been eluted together with the far more abundant, less highly acetylated variants. In such a case, it would have been much more difficult to unambiguously identify both the acetylation and methylation variants (58,59).  

In one of its on-line discussions of HILIC (60), PolyLC discussed the general applicability of HILIC for intact proteins: "While water-soluble proteins such as cytoplasmic enzymes do not lend themselves well to HILIC, HILIC works well for proteins that do not normally occur free in aqueous solution, such as membrane proteins and histones." 

Let us return now to the scientific literature for some final, described applications of HILIC for proteins of various properties and sizes. In a very recent publication, the authors reviewed past efforts to utilize HILIC in any form for intact proteins (61). They again stated that, other than the original publication by Alpert (1), very little had been published on the purification of proteins by HILIC. The authors nicely review most of the prior literature, which was perhaps two or three previous publications that used HILIC or variants thereof for analyzing proteins. They further speculated as to reasons for this seeming anomaly was that many proteins would be denatured, with concomitant loss of bioactivity, in the presence of the high concentration of organic solvents needed in HILIC. However, many classes of proteins do tolerate such conditions without denaturation. In any event, most HPLC work is directed at the detection and measurement of proteins, not at determining their biological activity. Another perhaps more valid reason may have been that many proteins are not soluble in the high concentrations of organic solvents usually needed for HILIC. However, to some extent, the lack of solubility can be countered by the addition of solubilizing agents such as hexafluoroisopropanol (HFIP) or unbuffered acids to the solvents for HILIC. In any event, the number of successful applications of HILIC for intact proteins has clearly grown in recent years, both academically and commercially.  

Tetaz and colleagues (61) used the initial chromatographic conditions described by Carroll and colleagues (54) and then developed a modified method that permitted the chromatography of numerous intact soluble proteins by HILIC (Figure 8). In some instances, it was possible to recover functionally active proteins. Their work described the resolution of two different lipophilic proteins (apoA-I and a truncated version of human apoM, lacking its signal peptide), plus another mitochondrial membrane-associated protein. The work was done on five different commercially available columns. It was further demonstrated that HILIC may be used to separate glycosylated isoforms of intact human apoM (61). In this study, small aliquots of the collected HILIC fractions of interest were directly analyzed by nanoelectrospray MS using quadrupole time-of-flight (TOF) tandem MS. The specific gradient elution HILIC conditions, finally optimized on various columns, are given and are very similar to those reported earlier by Carroll and colleagues (54).  

Figure 8: HILIC separation of human apoM (1 mg) on a TSKgel Amide80 column. (Adapted with permission from reference 61.)			

These authors were able to demonstrate the successful chromatography of both apoA-I and apoM, an additional class of lipophilic proteins not described before in HILIC (61). However, a more interesting and potentially significant finding was obtained with the separation of glycosylation variants of apoM. This suggested the potential of HILIC for separations never previously attempted or described. Figure 8 shows the HILIC separation of human apoM on a TSK gel Amide80 column (Tosoh Bioscience), with conditions described (61). Identification of the various proteins present in Figure 8 was done by analyzing collected fractions. As shown by the ESI-MS data, the earliest eluted peaks at 13–15 min were nonglycosylated apoM, and the fractions from 16–18 min were different, glycosylated isoforms of apoM. This may have been one of the first reports of an HPLC technique that could separate different glycosylated isoforms of an intact glycoprotein, other than by use of affinity chromatography (lectin, antibody, or boronic acid groups). The HPLC resolution of glycosylation variants of any protein has been a very difficult proposition, although separations of glycosylation variants of peptides have been much easier to realize, as described elsewhere in this column, vide supra.  

In a separate, but quite related publication, Carroll and colleagues (54) defined the mitochondrial proteome by measurement of molecular masses of numerous, membrane proteins. As the authors stated, 

"The use of this form of chromatography for purifying membrane proteins [together with SDS-PAGE with marker proteins and ESI-MS analysis of the HILIC fractions to confirm molecular masses] is a new development that has allowed some well known, but hitherto poorly characterized proteins to be more thoroughly studied." 

Figure 9 illustrates the overall approaches used in this particular study (note that Figure 9a is the same as Figure 5). There are two HILIC chromatograms, showing elution times of the major protein peaks and the nature of the gradient used for elution of each extract. Various protein fractions from each chromatogram were manually collected, transferred to sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and run alongside known protein markers of known molecular weights (data not shown). These were then (tentatively) identified by standard Coomassie Blue staining, and the molecular weights were approximately determined by the standard, known molecular weight marker proteins, as usual. Individual SDS-PAGE bands were cut from the plates, destained, denatured, and digested by trypsin to perform peptide mapping ESI-MS and database identification of the original proteins present. Each protein eluted from the HILIC column was also subjected to ESI-MS to determine its correct absolute molecular mass. In the above manner, especially improved by the use of the HILIC protein separation step, it was then possible to identify each of the major proteins present in Figure 9a and 9b (54). This technique was similarly described in proteomics by Berger and colleagues (62,63) and is now commonly termed 

Figure 9: Fractionation of hydrophobic proteins by HILIC. Solvent extracts of bovine mitochondria and complex I were made at pH 3.7 and then fractionated by HILIC. Different elution fractions from each chromatogram were then analyzed by SDS-PAGE with known marker standards (54). Elution profiles (solid lines) were monitored by UV absorbance at 225 nm. The column was equilibrated in buffer 1 and analytes were eluted with a gradient (dotted lines) of buffer 2. Flow rates: (a) 0.1 mL/min, (b) 0.05 mL/min. (Reproduced with permission of Proceedings of [US] National Academy of Sciences [PNAS] and the National Academy of Sciences US.)			

Conclusions and Possible Future Applications for HILIC in Biotechnology Tomorrow 

In summary, when we consider both parts I and II on this subject, and the major applications of HILIC in the literature and commercial materials today, it is clear that the method has become widely accepted, applied, and improved, with more commercially available HILIC columns as well as their accompanying, application notes. This is especially true for the three major classes of applications: glycans, glycopeptides or peptides, and glycoproteins or proteins. Although the first two classes have received the main emphasis in the literature and commercially since about 1990, it is hoped that applications for glycoproteins or proteins and the other groups alluded to above will soon increase and become more widespread. In addition, it should be clear that HILIC may well be an ideal approach for many other types of mono- and oligosaccharides, especially for heparins, glucagons, and more-widespread applications in glycomics (complex oligosaccharides such as glycosaminoglycans). It is also hoped that HILIC will play an increasingly large role in the proteomics and peptidomics areas, especially via the 2D LC–MS formats described above. And finally, it is hoped that HILIC will continue to be found advantageous for glycopeptide isolation in typical, protein-derived peptide mapping, often in combination with reversed-phase LC (2D LC). It is also entirely feasible that the approach will soon show utility for improved resolutions in the HPLC analysis of complex oligonucleotides, as suggested above.  

The authors acknowledge with gratitude and appreciation the assistance and guidance provided by several colleagues. We acknowledge receipt of numerous HILIC application notes and lecture slides, and a MerckSeQuant Primer on HILIC from Dave Lentz, Knut Irgum, and Tobias Jonsson of SeQuant Merck Millipore. We are especially indebted to Dr. Jonsson for providing us with numerous literature and applications references dealing with HILIC of proteins. And, for his astute comments and observations leading to constructive revisions in some parts of the above. 

We acknowledge the use of certain figures from various presentations by Andy Alpert and his applications literature from PolyLC. We acknowledge receipt of another Primer on HILIC from Waters Corporation, as well as the use of several of their application notes on HILIC for biotechnology applications.  

Biotechnology Today

The modern usage and specific applications of HILIC for two other major classes of analytes - glycopeptides and glycoproteins - are discussed.

HILIC and Its Applications for Biotechnology, Part II

What good is that big, ugly peak at the beginning of the chromatogram?

Often considered a necessary evil, the first peak in a chromatogram can be a useful diagnostic tool for troubleshooting liquid chromatographic (LC) separations. Most people I encounter refer to this as the column dead time peak, abbreviated 

. However, it has a wide variety of other names: junk peak, garbage peak, solvent front, or hold-up time, with 

 as the most common alternative abbreviation. This represents the time it takes something to go through the LC column that does not interact with the column. A corresponding dead volume (or hold-up volume), 

, is the volume of mobile phase inside the column. This volume comprises both the volume of mobile phase between the packing particles (the interstitial volume) and the volume within the particles (the pore volume). We'll see that 

 can be a useful diagnostic tool to identify potential problems with an LC method. 

If we want to use the column dead time as a tool, we need to be able to identify it. Most LC detectors will generate a peak at 

, the most obvious exception being the mass spectrometric detector (liquid chromatography–mass spectrometry [LC–MS]). Therefore, the chromatogram usually has a peak similar to the first baseline disturbance in Figure 1. If the sample is very clean and has minimal unretained material, a small baseline disturbance as shown in Figure 1a may appear. More commonly, there is sufficient unretained material to generate a large, off-scale peak (Figure 1b). Although there are more-exact measurement techniques for 

O, most of us just use the retention time of the peak. I prefer to pick a measurement that is easy to reproduce, because most of the time an estimate of 

 is sufficient. For Figure 1a, this is the point the disturbance crosses the baseline, noted by the arrow. Because a large unretained peak usually is off scale so that the top of the peak may be inconvenient to locate, I usually pick the point where the peak rises from the baseline (arrow in Figure 1b). Of course the retention time reported by the data system is another convenient measurement of the dead time.  

Figure 1: Examples of t0 peaks (arrows). (a) Chromatogram with little unretained material; (b) large peak normally observed at t0.			

 or to determine it if there is no corresponding disturbance in the baseline, as with LC–MS, we can estimate the column dead volume, 

. If you are using a 4.6-mm i.d. column, 

 is in millimeters. Thus, for a 150 mm × 4.6 mm column, 

 = 0.01 × 150 mm = 1.5 mL. For columns of other internal diameters, you can use  

 is the column internal diameter in millimeters. For a 50 mm × 2.1 mm column, 

/1000 = 0.11 mL. Either of these estimates is good to within approximately ±10% for columns packed with totally porous particles. These estimates are based on the assumption that 

 represents ~65% of the volume of an empty column and that about half of this volume is inside the particles and half is between the particles. 

The column dead time is simply the column volume divided by the flow rate, 

 to help diagnose problems submitted to me by readers. Here are some of the ways this can be useful. 

I find that it is useful to check to be sure that the presumed 

 peak is in the right place. For this, simply calculate 

 using equation 1 or 2 and 3, then compare this to the observed peak in the chromatogram. For example, if the chromatogram of Figure 1b was obtained with a 150 mm × 4.6 mm column operated at 2 mL/min, 

 ≈ 1.5 mL/2 mL/min = 0.75 min. This agrees with the observed peak at approximately the same retention time. If the calculated and measured values of 

 differ by more than ~20%, it is advisable to try to figure out why. Several possibilities are discussed below. 

 is larger than expected, the most likely cause is a flow-related problem. For isocratic methods, the retention time of retained peaks should change by the same proportion as 

 when the flow rate is changed. If, for example, the above case had an observed 

 of 1.0 min, the retained peaks should increase by 1.0/0.75 = 1.33-fold. If this is confirmed, check for flow-related problems. Larger than expected values of 

 indicate a drop in the flow rate. Always check the most obvious case first — is the flow rate set properly? It also should be obvious that historical retention data should be consulted to be sure an abnormality in 

Assuming that the flow rate is set correctly, the most likely causes of a flow rate problem are leaks, air bubbles, and problems with the check valves or pump seals. A secondary symptom may be low pressure, depending on how far off 

 is. If leaks are not obvious, I would open the pump purge valve and run 5 mL or so of solvent to waste from each flow channel in use. Thus, if you are using a two-pump high-pressure mixing system, purge both pumps; if it is a low-pressure mixing system, purge each solvent line. This should remove any bubbles from the system. If the problem persists, carefully check each fitting in the pressurized flow stream for leaks. Sometimes the fine point of a twisted laboratory wipe or facial tissue can be used to probe the fittings for possible leaks. If leaks in the flow stream are not found, a pump problem is most likely.  

If you are using acetonitrile as one of the solvents and the pump does not have active check valves, it is possible for the inlet check valves to stick. This is from the formation of polymers on the surface of the check-valve seat, and usually can be corrected by sonicating the check valves for a few minutes in methanol. A more detailed discussion of this can be found in an earlier "LC Troubleshooting" column (1). The outlet check valves also can leak if they become contaminated. If you have a pump with outlet check valves, sonicating them may help. If you choose to sonicate the check valves, be careful that you know how they are assembled, in case they come apart in the process. Worn pump seals also can leak, resulting in a lower than expected flow rate. Check the maintenance log for the pump. If the seals haven't been replaced in the past year, I would suggest replacing them. If the seals are newer, you can inspect the pump more closely for possible signs of leaking. Most pumps have a hole or drain tube below the pump head behind the check valves, where any leakage from the seals will exit. Look for signs of leakage, such as visible liquid or white deposits of buffer residue. Replace the pump seals if there is any question of its integrity. 

 peak comes out earlier than expected, one of two possibilities exists. The easier to address is a mistake in the flow rate setting so that the flow rate is too high. Although I suppose it is possible, I have never heard of a pump or controller software failure that resulted in excessive flow rates, so operator error is the most likely source of a flow-related problem. 

With the most common forms of LC (reversed phase, normal phase, ion exchange, ion pairing, and so forth), 

 should be the first disturbance in the chromatogram. If something is eluted earlier than the expected retention of 

, the compounds may be excluded from the pores of the packing material. The exception to this is in size-exclusion chromatography, where everything should come out before 

. If a molecule is restricted from entering the packing pores, it only has access to the volume of column between the particles (the interstitial volume), which I mentioned at the beginning was approximately half of the total solvent volume, or 30–35% of the column volume if the total dead volume is ~65%. I've seen the interstitial volume quoted as 40% of the column volume (2), but in either case, a significant portion of the dead volume is inside the particles. If sample molecules are restricted from entering the pores of the packing, they will be eluted before 

. Such peaks may be confused with the true 

 peak, because we are used to assigning the first peak in the chromatogram as 

Two common causes of restricted access to the pores exist. The most obvious is that the molecules are too large to enter the pores. A rule of thumb is that the pore diameter should be 3–4 times the hydrodynamic radius of the molecule. Most analytical columns have pores in the 8–12 nm (80–120 Å) range, which will accommodate molecules of up to ~10,000 Da. Above this size, for example with proteins, larger-pore columns, with 30–40 nm pores, are used. Thus, if your sample is a pharmaceutical product, although the analyte of interest may be <1000 Da, the formulation may contain polymers or other excipients in excess of 10,000 Da that may be excluded from the pores. In some cases, dimers or other aggregations of sample molecules can result in a material that is stable enough to chromatograph, but too large to enter the pores. Any of these large molecules may be eluted before the column dead volume. 

A second cause of restricted access to the pores is that chemical repulsion between a sample molecule and the pore may exist. An example of this is illustrated in Figure 2 (3). In Figure 2a, a sample of adrenaline (base), benzyl alcohol (neutral), and naphthalene sulfonate (acid) is separated on a C18 column with a methanol–buffer mobile phase at pH 6. Adrenaline has a p

 of 8.55, so it will be fully ionized under these conditions, and naphthalene sulfonate has a p

 of <1, so it will also be ionized. Adrenaline is unretained because in its charged form, it is very polar and not retained. On the other hand, naphthalene sulfate has sufficient nonpolar nature that it is well retained, even though it is ionized. With the addition of 14 mM octane sulfate as an ion-pairing reagent, the results of Figure 2b are obtained. Here, the ion-pairing reagent is assumed to be immobilized on the surface of the C18 stationary phase, creating an in situ ion-exchange surface with a negative charge that is used to retain the positively charged adrenaline. However, although the pH of the mobile phase was not changed, you can see that the naphthalene sulfonate is now unretained. This is because the pores now contain a net negative charge that repels the negatively charged naphthalene sulfonate. You can imagine a similar situation where a pore with a net charge would repel a sample molecule of opposite charge, resulting in exclusion from the particles and elution before the column dead time. This phenomenon, called ion exclusion, occasionally is observed in ion-exchange chromatography. Any chemical change in the pore surface that repels sample molecules will have the same result.  

Figure 2: Illustration of exclusion of a charged molecule from packing pores. (a) No ion pairing reagent present, charged base is poorly retained, acid is retained; (b) with ion pairing reagent, charge on particle surface attracts base, increasing its retention, but repels the acid, excluding it from the pores. See text for details. Adapted from reference 3.			

 is to check for the quality of the separation, especially when readers submit problem chromatograms for me to diagnose. For isocratic separations (those with a constant mobile phase composition), the retention factor, 

 is the retention time of the peak of interest. The retention factor is a measure of the distribution of the sample between the stationary phase and the mobile phase. As an indicator of chromatographic quality, I like to see 1 < 

 < 10, as has been discussed in past "LC Troubleshooting" columns (for example, reference 4). If 

 < 1 is observed, the peak is likely to have poor retention-time reproducibility and more likely than strongly retained compounds to have interferences from the tail of the 

 peak. The retention factor can be estimated by using 

 as the unit of measure for retention and measuring retention beginning at the observed value of 

. For the chromatogram of Figure 1b, this is done by dividing the baseline up in units of 

 instead of minutes. The first peak is eluted a little over 1 

 value of a little more than 1. Similarly, the second and third peaks have 

 values of a bit more than 2 and 3, respectively. Although equation 4 does not apply to gradient elution, the general principle of keeping the first peak away from 

 to avoid interferences still holds. From this standpoint, I like to see the first peak of interest in a gradient come off the column at least 1 

Although at first glance, the solvent peak at the beginning of the chromatogram has no value, it can be a useful tool to help diagnose problems with the chromatogram. We can compare column dead time estimates with the observed retention time of the 

 peak and get an idea of what might be going wrong. When the first peak comes out after the expected retention time, the problem is usually flow-related and most commonly caused by a leak. Peaks that are eluted before the expected retention 

 time are most likely excluded from the pores of the column, because they are too large or are repelled from the pores. So we can see that nothing (

"LC Troubleshooting" Editor John Dolan has been writing "LC Troubleshooting" for

for more than 30 years. One of the industry's most respected professionals, John is currently the Vice President of and a principal instructor for LC Resources, Walnut Creek, California. He is also a member of 

's editorial advisory board. Direct correspondence about this column via e-mail to 

LC Troubleshoting

Column Dead Time as a Diagnostic Tool

To achieve lower detection limits and better selectivity, researchers are always exploring new ways to enhance sample preparation and separation technology. The addition of electrical potential to existing technologies or for entirely new approaches is relatively unexplored but can create another dimension of selectivity in sample preparation. In addition, speed and sensitivity are added benefits. This installment looks at the use of electroenhancement to solid-phase extraction, solid-phase microextraction, and membrane and liquid extraction.

We are all familiar with the role of electromigration in enhancing separations in capillary and gel electrophoresis and related techniques and in capillary electrochromatography. In addition, electrical potential has been used in electrochemical and other forms of chromatographic detection. The use of electrical potential to enhance and add another dimension of selectivity to sample preparation is still a relatively new, noncommercialized technique. The purpose of this installment is to discuss how electrically driven enhancement has already proven to offer some major advantages to the preparation of samples whose analytes of interest are ionized or can become ionized by the appropriate adjustment of pH to the separation media. We explore five sample preparation methods where the application of a charge across barriers provides enhanced selectivity, oftentimes leaving undesired matrix and impurities behind, thereby accomplishing the goals of sample preparation.   

A look at five sample preparation methods where the application of a charge across barriers provides enhanced selectivity

LCGC North America-01-01-2014

Peaks of Interest

Optimizing LC–MS and LC–MS-MS Methods

Strategies for the Detection and Elimination of Matrix Effects in Quantitative LC–MS Analysis

HILIC and Its Applications for Biotechnology, Part II

Column Dead Time as a Diagnostic Tool

Electrical Potential as a Driving Force in Sample Preparation

The Origins of GC Carrier Gases: Putting a Genie in the Bottle

Vol 32 No 1 LCGC North America January 2014 Regular Issue PDF