Developing versatile, rapid, and accurate analytical techniques for the detection and quantification of water in a variety of materials remains an important and ubiquitous analytical problem. This article describes an effective and sensitive ionic liquid (IL)-based capillary gas chromatography (GC) method with thermal conductivity detection (TCD) to determine the water content in liquid samples. The open tubular capillary columns, coated with specific ILs developed in this study, increased the sensitivity and significantly increased the ruggedness of this technique. The lower limit of detection (LOD) of this method is ~2 ng water. Samples containing high levels of water are also readily analyzed without pretreatment. Conversely, organic solvents can be measured in water by the same approach using either TCD or flame ionization detection. Standardization was performed with National Institute of Standards and Technology (NIST) reference materials and the accuracy was compared with another independent method, Karl Fischer titration (KFT). The developed method is highly sensitive, fast, and is not affected by interferences and side reactions common with existing KFT methods.

The determination of water content in solvents and consumer products, including foods, pharmaceuticals, and industrial materials is of great importance. Analytical testing for the presence and concentration of water is one of the most frequent, important, and ubiquitous measurements made in modern industrial society. Thus, a versatile, simple, and efficient analytical technique for the accurate quantification of water is imperative. The essentially universal presence of water requires accurate, easy, and sensitive techniques to quantify it. Techniques such as gravimetry (1), Karl Fischer titration (KFT) (2,3), gas chromatography (GC) (4–13), near-infrared (NIR) spectroscopy (4,14–16), solvatochromic sensing (17,19), fluorine nuclear magnetic resonance spectroscopy (F-NMR) spectroscopy (18), isotope ratio mass spectrometry (IRMS) (19), and others have been reported in the literature, but only a few methods are widely accepted and used. 

Currently, the most commonly used method for water analysis is KFT, which was first reported in 1935 (2). In this titrametric method, I

 is reduced to HI in the presence of water (9). There are four components in the Karl Fischer reagent, consisting of: iodine, sulfur dioxide, a suitable base (RN) (originally pyridine was used, but now imidazole is more common); and a suitable solvent (methanol, ethanol, diethylene glycol monomethyl ether, and so on) (20). 

The accepted mechanism of this two-step reaction is as follows (20,21): 

 The end point is determined potentiometrically. Two types of KFT methods are used: coulometric titration and volumetric titration. The former is used to detect trace amounts of water, ranging from 10 µg to 99 µg (1 ppm–5%), and it requires ~5 g or more of sample (22). Volumetric titration is used to detect water quantities greater than 1 mg (10 ppm–100%) and the amount of sample required varies from 0.1 mg to 500 mg (23). Therefore, prior knowledge of the approximate amount of water present in the sample is required to choose the correct Karl Fischer method of analysis.  

Although this is a well established method, interference of side reactions (9), reagent instability (21), sample insolubility (24), and pH issues (4) prevent it from being accepted as a universal method. Variations on the basic KFT methodologies have been developed in an attempt to overcome these problems (3,9,24). However, many issues still remain, not the least of which is that the reagents degrade with time and there is residual water in all KFT reagents.  

Another applied method for water detection is based on GC. Early attempts using GC were mainly based on packed (molecular sieve) columns, involving both direct detection by thermal conductivity detection (TCD) (7–9,11,13) and indirect detection (that is, reacting water with calcium carbide to convert to acetylene) with flame ionization detection (FID) (5). Peak asymmetry (4), poor sensitivity, poor efficiency, strong adsorption of water and many solvents by the stationary phase (7), overlapping of the water peak by other larger peaks, and the inability to detect higher amounts of water (4) limited its application in many cases. Attempts to eliminate peak asymmetry (4,25) using wide-diameter open-tubular columns (6) and capillary columns (10) showed some improvement. Additionally, most conventional capillary column GC stationary phases are degraded by water.  

A truly useful, broadly effective capillary GC method for water should meet several criteria including the following:   

Water should not alter or degrade the stationary phase, thereby altering retention times and peak shapes.

There must be a considerable difference in the retention of water and most or all organic solvents especially when the solvent peak is very large relative to the water peak. 

The water peaks should show good efficiency and symmetry. 

The water and solvent chromatogram should have sufficient separation space for an appropriate, baseline-separated internal standard.

In this study, we examine an IL-based capillary GC method with TCD for the direct determination of water content in liquid samples. The unique characteristics of ILs, including high thermal stabilities (26), variable polarities (27,28), and exceptional stability to water and oxygen, make them excellent choices as stationary phases for this methodology. As will be shown, open-tubular capillary columns coated with specific IL stationary phases (developed for water analysis) can increase the sensitivity and greatly enhance the applicability and reliability of this technique. Furthermore, analysis times can be decreased to 6 min (or less than 3 min in many cases), and samples with virtually any concentration of water can be analyzed. The efficacy of this approach is demonstrated with 50 different solvent samples.  

The analysis was performed using a model 6890N gas chromatograph (Agilent Technologies Inc., Wilmington, Delaware), equipped with a 7683B series autoinjector, thermal conductivity detector, and Chemstation plus software (Rev. B.01.03). An Agilent Technologies 10-µL syringe (5181-1267), was used with the autosampler, and a Hamilton 10-µL syringe (Reno, Nevada) was used for all manual injections. The fused-silica capillary columns were coated with IL stationary phases synthesized as previously reported (26,29). The columns were 30 m long with a 0.25-mm internal diameter and a 0.2-µm film thickness. KFT was performed using an Aquastar V1B volumetric titrator (EM Science, a division of EM industries, Inc., Cherry Hill, New Jersey). 

The water reference material: 8509, moisture in methanol (97 ± 13 ppm water) was obtained from the National Institute of Standards and Technology (NIST, Gaithersburg, Maryland). The 4-Å molecular sieves, Hydranal water standard 1.0, tetrahydrofuran, and both the internal standards (acetone and acetonitrile) were purchased from Sigma-Aldrich (St Louis, Missouri). Aquastar combititrant 2 and Aquastar methanol were purchased from EMD Chemicals (Gibbstowm, New Jersey). The high-purity water was obtained by filtering the deionized water with a Synergy 185 water system (Millipore, Billerica, Massachusetts). The testing solvents were from Sigma-Aldrich, Mallinckrodt, EMD, Fisher, Omni Solvent, and Acros Organics (see "Supplementary Information" for more details).  

The accurate quantification of water was achieved using one of two internal standards, either acetone or acetonitrile. Two different internal standards were available in case one was coeluted with the analyte solvent under the conditions of the experiment. Internal standards were dried to less than 10 mg/kg by storing over 30% w/v 3-Å molecular sieves for seven days before use. A typical sample preparation involved drying a 2-mL autosampler vial overnight at 130 °C followed by the addition of approximately 1 mL of solvent. The mass of the solvent was recorded using an analytical balance. 5.0 mg of IS was added before analysis. 

For all analyses, helium was used as a carrier gas at 1 mL/min. Inlet and detector temperatures were set at 250 °C. Splitless injections were made with a purge flow of 20 mL/min. Similar results are obtained using split mode with 20:1 split ratio. Oven temperatures varied from 40 °C to 120 °C (See "Supplemental Table sI", which is available on the online version of the article). For initial screening, the columns were conditioned at 120 °C for 2 h. The sample of interest was injected in the splitless mode (0.2–5 µL) to examine the separation between water and the bulk solvent. This preliminary experiment helped to determine which IS should be used.

Water quantification was achieved by integration of the IS peak and the water peak. The concentration of IS in milligrams per kilogram was multiplied by the ratio of water peak to IS peak and the result was divided by the response factor (RF). The resulting number represents the concentration of water in the sample in milligrams per kilogram. Each solvent was prepared in triplicate and each individual sample was injected in triplicate for a total of 

 = 9 individual integrations for each solvent analyzed.  

The limit of detection (LOD) and the limit of quantification (LOQ) were calculated according to the guidelines of the US Food and Drug Administration (FDA) (30,31), using the following equations: 

where σ = the standard deviation (SD) of the response and 

The σ is normally obtained from the SD of the blank sample (31). As it is impossible to obtain a sample without water, the first point of the calibration plot, where no added water is used as the blank sample and its SD was used as σ. The slope was obtained from the regression analysis of the plot; peak area ratio of water and IS versus the amount of water. 

Figure 1 shows the structures of the three ILs that were used as stationary phases in this study. Generally, IL stationary phases containing trifluoromethylsulphonate (TfO

 ) anions resulted in more symmetric water peak shapes than those that contained PF

  or bis[(trifluoromethyl)sulphonyl]imide (NTf

 ), with the same cation (data not shown). In addition to the IL-based GC columns, a commercially available polyethylene glycol (PEG) column was also evaluated for comparison.

Figure 2 shows the GC separations of small amounts of water impurities from two of the organic solvents that were analyzed. Note that the water peak can be eluted before or after the large organic solvent peak, depending on the relative elution order of the two on the IL-based stationary phases. In most cases, the "separation window" between the water and solvent peaks is sufficiently large (Figure 2a) that a thermal gradient can be used to further narrow the peak width of water and reduce analysis times (Figure 2b). A comparison of the separation of water from methylene chloride on an IL column versus a commercial PEG column (see "Experimental") is shown in Figure 3. Note that the IL-based column (Figure 3a) produced better peak shapes and selectivity even though the separation conditions were optimized for the PEG column. In all cases (for all solvent samples) the IL-based separations were substantially better in terms of selectivity and efficiency.

Table I compares the amount of water measured in 10 solvents and a NIST methanol standard using four different capillary GC methods and the KFT. The samples were chosen so that the KFT analysis could be done without the use of special reagents (for example, as are needed to quantify water in aldehydes and ketones) (3). Also, most of these samples could be analyzed on the commercial PEG column. As can be seen from the data, the IL-based columns usually produced the most precise and accurate results (as indicated from the relative standard deviations [RSDs] and NIST standard respectively, Table I). It should be noted that no more than 1 µL of sample was used for any capillary GC determination. Conversely, the KFTs used 5–15 g of sample depending on the water content of the solvents (larger samples were required for samples containing the least water) (Table I). Figure 4 shows a standard addition plot for the determination of water in SRM 8509. As can be seen, standard addition is a viable alternative to determine water content if a response factor is not available. High levels of accuracy and precision are possible when using a calibrated syringe.

The LODs and LOQs for the IL columns were better than those found for the commercial PEG column, even when analyzing samples and using conditions that are necessary for favorable separations on the PEG column. The limit of detection (LOD) of coulometric KFT is 10 µg, and it requires at least 5 g of sample. The IL-based GC method required only 0.2 µL of sample to obtain a much lower detection limit (~2.0 ng or ~5000×greater sensitivity).

Table II lists the water content of 50 different solvents as determined with the three IL columns evaluated in this study. Figure 5 indicates the number of these solvents that could be successfully analyzed on each column. To be considered a successful separation, the water (analyte) peak had to be separated from both the solvent and at least one of the internal standardss (that is, acetone or acetonitrile). All three IL columns could be used to quantify water in 38 out of the 50 solvents (Figure 5). Overall, the HAIM-PEG-TFO column was the most successful because it could be used to quantify water in 47 of the solvent samples, including two samples that were not amenable to separation on the other tested columns.

  column was used to quantify water in 42 solvents, including pyridine and 1,2-dichlorobenzene, which were not possible using the other two columns that were investigated. The DMIM-PEG TfO

  column often gave the best separation window between the water peak and solvent peaks, compared with the other columns. Hence all the separations that are possible on this column can be done at even higher temperatures than those used in this report (Table sI) or using temperature gradients (Figure 2b) if desired. In most cases, this means that many analysis times can be less than 3 min (results not shown).

As capillary columns containing IL stationary phases are effective in the measurement of small amounts of water in organic solvents, it would not be surprising if they also can be used to determine trace organic solvents in water. Such a determination is shown in Figure 6. Note that in Figure 6a, TCD is used to show the elution of the water peak relative to trace organic solvents. In Figure 6b, even lower concentrations of the organic contaminants are seen when using FID. Of course, water cannot be seen with this detector. It should be noted that the injection of water samples is not recommended on ordinary commercial columns that are not based on ionic liquids. Virtually all traditional commercial columns show appreciable degradation and continuously changing chromatograms when analyzing water samples. Analyzing water for organic solvents tends to be much more damaging for these columns than analyzing organic solvents for trace amounts of water. For example, three successive commercial PEG columns had to be used for this study while the separations and conditions of all IL columns used remained unchanged throughout the study (after more than 1600 injections).

A rapid, easy, ionic-liquid (IL)–based GC method has been developed for the quantification of water in extremely diverse solvent samples. LODs using this technique are superior when compared with KFT. Furthermore, the IL-GC methodology requires less sample and is free from other complications associated with KFT. The IL-GC method can be used regardless of the chemical nature of the solvent and produces no additional waste products. When compared to other GC methods using commercially available PEG columns, the IL-based columns possessed superior selectivity for water in all solvents tested. Furthermore, they show no degradation or chromatographic changes with time. Typical analysis times ranged from less than 3 min to 7 min. It is expected that this approach could greatly simplify the analysis of water in a variety of applications. The extension of this approach to gases and solid samples is under investigation and will be reported subsequently.  

This work was financially supported by the Robert A. Welch Foundation (Y-0026). We would like to thank Professor Zoltan A. Schelly of the University of Texas at Arlington for providing the KFT device. 

Supplementary information for this article that did not fit in the print edition is available here. That information includes Table sI.

Articles

Developing versatile, rapid, and accurate analytical techniques for the detection and quantification of water in a variety of materials remains an important and ubiquitous analytical problem. This article describes an effective and sensitive ionic liquid (IL)-based capillary gas chromatography (GC) method with thermal conductivity detection (TCD) to determine the water content in liquid samples.

Rapid, Efficient Quantification of Water in Solvents and Solvents in Water Using an Ionic Liquid–Based GC Column

 In gas chromatography, selectivity is commonly thought about exclusively in the column stationary phase and sometimes extended to the detector. This installment of "Column Watch" explains how to incorporate selectivity throughout the sample analysis cycle, from the sampling stage through data analysis and report generation. 

Many samples to be analyzed using gas chromatography (GC) are complex. There may be many analytes of interest to be separated, possibly at very low concentrations. Some of the analytes may be difficult compounds, meaning they may be thermally labile, active, or of too high molecular weight for analysis by GC. There also may be a high loading of matrix and the presence of matrix interferents. From sample collection to data analysis, there are many steps in the process. Of course, the final step is the reporting of the required information about a sample. To get reproducible and accurate results, it is important to think about and understand the sample first, and then to identify the steps to select the best techniques for the analysis. 

In chromatography, when selectivity or being selective is mentioned, first thoughts are always directed toward the analytical column's stationary phase, especially in GC where this parameter has the biggest effect on the separation of analytes. In GC, the mobile phase, a gas, has little interaction with analytes, the matrix, or the stationary phase, and mainly serves to transport compounds down the column. However, each of the steps identified in the collection, analysis, and reporting of a sample can have selectivity applied to it. Figure 1 identifies some of these key steps. Although some of the steps involved in selectivity enhancement are often the same for gas and liquid chromatography, I will focus this discussion on considerations in GC. 

 Figure 1: Steps in sample collection, preparation, and analysis. 

The first and most important step in any separation is to determine how to collect a representative sample, enough for the analysis. One must consider the nature of the sample (gas, liquid, solid, or something in-between); the analytes (volatility, stability); and the location of the sample. It may be relatively easy to collect and return some samples to the laboratory for analysis, while in other cases, sample changes (such as oxidation, chemical breakdown, or chemisorption) can occur during the transport or storage process. It may be the case that the analysis needs to be performed in situ.  

Selectivity can be considered in this first step — to selectively collect only the portion of the sample of interest with as little matrix as possible. Selectivity can be based on volatility, where only analytes within a certain volatility range are sampled, or on the nature of the analyte, where only target analytes with certain characteristics like polarity are collected. In a case where there is a high level of matrix interferents, a technique may be selected to avoid these;  for example, a reversed-phase solid-phase extraction (SPE) sorbent may be used to concentrate organics and to avoid trapping water. 

One technique for selectively sampling on site is thermal desorption; the thermal desorption tube packing material is selected to retain the target analytes and not the matrix. Solid-phase microextraction (SPME) and stir-bar sorptive extraction (SBSE) are other techniques where the phase can be selected for the analytes and not the matrix, and can be used on site. 

The aims of sample preparation can be to make the analytes of interest amenable to GC; to concentrate the analytes; and to remove matrix (if possible). Selectivity also can be considered at this step, to extract only the analytes and not the matrix. Again, selectivity can be based on volatility or the nature of the analytes.  

For example, when using SPE, analysts can choose a phase that is selective for either the analytes or the matrix by thinking about the chemistry of the analytes and selecting a sorbent stationary phase that provides the best interaction to enable extraction of the analytes from the sample. Likewise, in SPME, SBSE, and liquid–liquid extraction (LLE), the solvent can be selective. Techniques like headspace and purge-and-trap sampling use volatility to selectively extract the analytes from the sample. Whereas parameters such as temperature are important to ensure the best extraction efficiency of target analytes, their effect on the coextraction of matrix components also must be considered. With dynamic headspace and purge-and-trap sampling, the selectivity of the trap also can be chosen to reduce the trapping of matrix coextractives.  

Unfortunately, during sample preparation often there is the potential for the loss of analytes and the concentration of unwanted matrix. Given that there are many steps in the analysis of the sample where we can be selective for one over the other, it is important not to lose analytes at this step for the purpose of removing matrix; the latter can be examined again later on in the process. One important factor to consider in sample preparation is that the more individual steps used to isolate or concentrate the analytes of interest, the bigger the chance for analyte loss.  

Although not as popular as it once was, chemical derivatization of analytes (or perhaps the matrix) can be used to achieve additional selectivity. By adding new functional groups or eliminating "old" functional groups, chromatographic selectivity changes can be dramatic and may sometimes allow analytes to be completely resolved from interferences. Derivatization can enhance selectivity by introducing functional groups that allow the use of a selective detector (as discussed later). In addition, derivatization can be used to enhance volatility such that nonvolatile high-molecular-weight compounds that ordinarily couldn't be passed through a GC column can be separated by GC. The downside to analyte derivatization is that it adds an extra step, sometimes lengthy, to the overall analysis. 

Steps 1–3 may be automated or manual, but with step 4, we are now entering the instrumentation, and most modern GC instruments utilize automation exclusively.  

The aim of sample introduction is to repeatably and reproducibly introduce a representative portion of the sample onto the analytical column in a narrow sample band with no chemical change. Usually, the analytes are dissolved into a liquid solvent that requires evaporation in the GC inlet; this liquid expands into a large vapor volume that then must be squeezed into a narrow column. Analytes already in the vapor phase have to be quickly flushed onto the column to ensure a narrow sample band and hence produce sharp peaks. Sample introduction is often the trickiest part of the process, with problems such as mass discrimination, activity, and breakdown frequently occurring in the inlet. However, discrimination doesn't have to be all bad; selective discrimination based on Grob-type principles uses the volatility of the sample components to selectively transfer analytes into the analytical column from the GC inlet. There are two main areas where this principle can be applied. At the front-end of the chromatogram, it can be used to eliminate major components. An example of this is large volume injection (LVI), where the solvent is evaporated out of the split exit before it is closed and the concentrated analytes are transferred onto the column. The principle also can be applied to the back-end of the chromatogram, to leave nonvolatile material within the inlet liner, protecting both the column and detector as well as reducing run times. 

Frequently, the temperature of a hot GC inlet is not optimized, leading to unnecessary problems for sample analysis results and maintenance of the instrument. When selecting the inlet temperature, the default seems to be 250 °C, which may be too hot or too cold for certain analyses. Frequently, a temperature is chosen that is too high, because the analyst considered the boiling points of the components rather than thinking about their volatility. Figure 2 shows a plot of the literature boiling point of a hydrocarbon mixture against the effective boiling point — the temperature at which 100% of the analyte is vaporized and transferred onto the column. To vaporize the analytes, the inlet temperature is important, but so is the flow through the liner, which is dependent on column flow and split flow. As Figure 2 shows, the inlet temperature for this particular inlet type, an Atas Optic 2 (Atas GL, Eindhoven, The Netherlands) and liner flow, only needs to be a maximum of half of the boiling point for total transfer of the analyte onto the column. 

Figure 2: Literature boiling point vs. effective boiling point. 

Figure 3 shows an example, for a hydrocarbon mixture, of profiling the inlet temperature required to transfer components out of the inlet while retaining others within the inlet liner. For example, at an inlet temperature of only 80 °C, the C

 hydrocarbon and those above are retained within the inlet liner, while 80% of C

 is vaporized and swept out and 100% of C

 and below are vaporized. To vaporize 100% of C

, an inlet temperature of 130 °C is required, but this is much lower than its boiling point of around 320 °C. 

Figure 3: C14âC36 hydrocarbon vaporization efficiency plotted as a function of temperature. 

As mentioned, this information is very useful when considering selectivity within the sample introduction step of the analysis. Here, selectivity is based on the volatility of the analytes; the control of the inlet temperature can be used to selectively transfer or exclude sample components from the analytical column. Figure 4 shows the discrimination envelope, where again for this inlet the temperature can be selected to either transfer an analyte or retain it within the inlet liner, with knowledge of the literature boiling point.  

Figure 4: Temperature discrimination envelope: upper curve = the total transfer of analyte; lower curve = total exclusion of analyte. 

For example, the least volatile analyte of interest in a sample has a literature boiling point of around 350 °C, which means the inlet temperature must be at least 140 °C (upper curve) for total transfer from the inlet onto the column.  However, a large matrix peak with a literature boiling point of around 430 °C is preferred to be retained within the liner rather than being transferred onto the column, where it will potentially damage the stationary phase, dirty the detector, and lengthen the run time to elute it from the column. An inlet temperature of 170 °C (lower curve) or higher would mean transfer of this component onto the column. Therefore, an inlet temperature of at least 140 °C would mean all of the compounds of interest would be transferred to the column, but a higher inlet temperature, especially around 170 °C, would also mean transfer of less-volatile matrix components, an undesirable result. 

Selectivity within the inlet also can be used for the concentration of analytes with LVI. The volume of solvent injected into a GC inlet is limited by its vapor volume; if the volume is too large, the capacity of the liner can be exceeded, causing contamination of the inlet, septum purge, split, and even the carrier gas inlet lines. Transferring too large a volume of solvent onto the column can result in distorted peaks because of the long solvent band. An LVI approach with a programmable temperature vaporizer (PTV) enables the evaporation of the majority of the solvent before the analytes are vaporized and transferred onto the column. The inlet temperature is chosen to enable the evaporation of the solvent while retaining the analytes within the inlet liner; usually this is 30 °C below the boiling point of the solvent. The discrimination envelope can be used here, but this time in reverse. The solvent boiling point is put into the upper curve for total transfer out of the split exit, although a slightly lower temperature is preferable to increase the solvent evaporation time so that it is more controlled. The boiling point of the most volatile analyte of interest is put into the lower curve. This temperature shouldn't be exceeded; otherwise loss out of the split exit can occur. This procedure can be used to eliminate any major component at the front end of the chromatogram, not just solvents. 

Figure 5 shows the power of selectivity within the GC inlet to selectively take a volatility cut of a petrochemical sample and transfer it onto the column for separation while flushing the unwanted parts out of the split exit. The upper trace (Figure 5a) shows the entire sample transferred and separated on the column. If the less-volatile components were not of interest, as in the example discussed above, up to 100% of C

 can be transferred and separated on the column, as shown in Figure 5b, which shows the front end of the chromatogram. If only the less volatile components were of interest, only 100% of C

 and above can be transferred to the column, as shown in Figure 5c. Figure 5d shows a volatility cut, where only 100% of C

 is transferred with the minimum of more- and less-volatile analytes. In all of these traces, the unwanted components are flushed out of the split exit rather than being retained in the liner. Hence, traces of unwanted components are seen, the amount depending on the split ratio. Figure 5b can be achieved with any GC inlet type, selecting the correct inlet temperature to get 100% transfer of all analytes of interest; retaining unwanted less-volatile matrix components in the inlet liner is important for reproducibility of the target compounds while minimizing column and detector maintenance and GC run times. This approach also has the advantage of vaporizing the sample at the lowest possible temperature, hence causing fewer problems for thermally labile analytes. Figures 5c and 5d can only be achieved in cases where the inlet temperature can be varied during the sample introduction, with inlets such as PTV inlets. 

Figure 5: GC analysis of a petrochemical sample: (a) entire sample transferred and separated; (b) transfer of only 100% of C20 and earlier; (c) transfer of only 100% of C40 and later; (d) transfer of only 100% of C30. 

The separation of analytes and any transferred matrix on the analytical column is the main area where selectivity is considered in the GC analysis of a sample. By thinking about selectivity in steps 1–4, and finding ways to collect, extract, and transfer only the analytes of interest and not matrix, we can have fewer peaks to separate in the fifth step. In addition, less instrument and column maintenance may be required.  

When selecting the GC column stationary phase, phrases such as "like separates like" or "polar separates polar" are useful. It is important to consider the chemical composition of the whole sample — both analytes and matrix — in separating them all from each other. Volatility as well as the functional groups present also must be considered when choosing the stationary phase, because very polar columns often are not able to reach the temperatures required to elute high-molecular-weight analytes. Stationary phase film thickness, column length, and internal diameter should be just enough to separate the components while minimizing band broadening and run times. 

Sometimes, however, even if an analyst is very selective in the choice of analytical column as well as in steps 1–4, there just aren't enough theoretical plates for acceptable separation of the large number of analytes and matrix peaks transferred onto the column. Or, the components of the sample might have mixed characteristics in terms of polarity to make selection of the proper stationary phase difficult, resulting in good separation of some compound classes but not others. Introducing further selectivity approaches can help with these problems.  

Heart cutting transfers a "cut" of the remaining coeluted peaks from the separation on one stationary phase onto a secondary column containing a different, more selective stationary phase. Several cuts can be performed across a chromatogram. 

Comprehensive two-dimensional gas chromatography (GC×GC).

 This method transfers many smaller cuts, usually 1–10 s long, onto a second column with a different stationary phase throughout the whole analysis; therefore every peak is separated on two different stationary phases. This separation enhancement is achieved by having the separation on the second column take no longer than the cut from the first column; for example, if the cuts from the first column are 4-s long, the separation on the second column takes no longer than 4 s. 

 Another method of introducing more selectivity is to hyphenate a different chromatographic technique to GC. For example, the sample might be first analyzed by liquid chromatography (LC), where selectivity in the mobile and stationary phases are very different, followed by analysis using GC. This is sometimes referred to as 

There are more than 20 detectors for GC on the market, many using quite different chemical or physical principles. When choosing a detector for a GC method, the aim is to select one that has adequate sensitivity with a linear response over the concentration range and is stable, reproducible, and gives the required information to confirm the presence or absence of the target analytes, along with a response for quantitation if required. By introducing selectivity in this detection step, we can improve sensitivity or make it much easier to observe the target analyte among coeluted analytes and matrix components. 

Certain detectors respond only to certain compound classes; for example, an electron-capture detector only responds to analytes capable of capturing an electron. Other detectors only respond to certain atoms; for example, a sulfur chemiluminescence detector is used for sulfur-containing molecules, and a nitrogen–phosphorus detector is used for nitrogen- or phosphorus-containing molecules. A mass-selective detector adds another dimension of information and selectivity by providing a full mass spectrum, where a single ion or ions can be chosen that are unique to that molecule but not the coeluted peaks; the accurate mass of a molecule; or an MS-MS product ion spectrum. A detailed discussion of GC detectors is beyond the scope of this paper. It suffices to say that detection selectivity can be another addition to our "spectrum" of selectivity enhancements. 

Unfortunately, data analysis sometimes takes just as long if not longer than the whole sample collection, preparation, and analysis. Properly understanding the problem and the questions to be answered is the starting point. Then, introducing selectivity at each of the steps discussed above to obtain high quality data showing the analytes of interest with as little interference as possible helps reduce the data analysis time as well as improve the quality of the results. Unfortunately, even after carefully optimizing each of the previous steps to be as selective as possible, there can still be some coelutions between analytes and matrix that cause problems, especially if very low-concentration analytes are being analyzed in dirty or complex samples.  

If the sample has been analyzed with a detector that generates another dimension of data, more selectivity can be introduced at this step, too. With full scan data from a mass spectrometer, ions can be selected for qualitative and quantitative analysis that are unique to that analyte, which means the ions are not present in coeluted peaks, ion ratios show little variation, and integration is much easier baseline-to-baseline, reducing manual data analysis time. 

Failing that, advanced techniques like deconvolution automatically unknit coeluted mass spectra for confirmation or identification of an analyte. The deconvoluted ions can then be used for accurate quantitation. Statistical analyses of the data using chemometric techniques like principal component analysis (PCA) are also helpful when analyzing complex data sets. PCA can assist with method development, the identification of outliers, determining patterns of association and variability, applying confidence intervals to determine the reproducibility of data, and determining mass spectral match coefficients. Even simple analyses can benefit from this automated approach. 

Yet however good these advanced data analysis techniques are, they will always work best on high quality data, generated from a fully optimized sample collection, preparation, and analysis method where the techniques are selected to fit the analysis. 

Like all parts of a method, optimization of the data analysis step takes time but is crucial to getting accurate results with as little manual input as possible, resulting in less data analysis time overall. 

Before determining which techniques to use at each step of the analysis, here are some factors that should be considered:  

The sample phase. Is it a gas, a liquid, a solid, or something in between?

The analytes, including their volatility, nature, and concentration. Are very low detection limits needed? Are they thermally labile, active, or high-molecular-weight compounds?

The matrix (volatility and nature). What proportion of the total sample is the matrix?

The sample location. Can a representative portion be taken back to the laboratory easily without the sample changing? Is sampling in situ needed? Does the analysis need to be performed on site?

Can the number of sample preparation steps be minimized?

How many samples are there? Is it necessary or even possible to automate the sampling or extraction technique?

Column Watch

Learn how to use selectivity throughout the sample analysis cycle, from the sampling stage through data analysis and report generation.

It's All About Selectivity

A lively exchange of perspectives emerged recently on the LinkedIn group for 

 as well as the group for our sister magazine, 

 article, was about what it takes to become a scientist and why so many students, particularly in the United States, become discouraged along the way. The article noted that about 40% of students planning to major in science or engineering end up switching to other subjects or even failing to get a degree. When premedical students are included, the number increases to as high as 60%. That is twice the attrition rate of all other majors combined.  			

A number of participants in the LinkedIn communities said it is normal for many science students to drop out and that this phenomenon is not new. Science is challenging, and requires hard work, diligence, and passion to obtain even an undergraduate degree. One noted that her roommate thought she was a graduate student, because she spent so much time in the laboratory and studying instead of socializing like the students majoring in business and liberal arts. If you don't have what it takes, or the love of it, many say, it's fine to move on.  

That may be. But another important factor is the preparation students receive in these areas earlier in their education. There are too many American students who finish high school poorly prepared in math and science. And that's not just a problem for those who want to go on to earn a PhD in chemistry. In the current national conversation about the economy, and particularly over the question of which sectors can provide middle-class jobs for the future, many companies say they have difficulty finding people with the right education and skill sets in the United States. To cite just one example, the factories that remain in this country will produce high-end products that must be made with complex robotics rather than with large staffs of uneducated laborers. That means that even on factory floors, employees will need a stronger background in math and science than previous generations of American workers.  

Improving the U.S. educational system is not easy. But student motivation is an important factor that should not be dismissed. As many people pointed out in the forum, our culture worships actors and athletes, not scientists or other intellectuals. (Just think of how scientists are portrayed in the television series, "The Big Bang Theory.") 

Yet I have always believed the microculture trumps macroculture. We live in a larger society, but the greatest influence on our lives comes from the environment immediately surrounding us — the people we interact with every day. So even in an educational system with many gaps, an individual chemistry teacher can motivate students to continue with studies that may at first seem dry or difficult. One participant in the forum is a teacher who understands this well. In particular, he knows the value of paying attention to the average students, not just to the best and brightest. He writes:   

"I once sat in front of a GC–MS system with a bunch of 'problem students' — the back-row gang. I talked of gasoline, additives, and octane. You could hear a pin drop. One of those went on to get a BS in chemistry because he saw what he could do. Little things matter." 

Moving, isn't it? The good news is that it's not necessary to be employed as a teacher to contribute to this cause. Any scientist can visit a school and serve as a role model. You don't have to be famous, or a Nobel prize winner. You can just share what you do, what it took to get there, and why it was worth it.  

In this online discussion, many participants expressed pride and satisfaction in their choice of profession. If some small percentage of today's generation of scientists visited a school this year to talk about why their work is rewarding, that could lead to countless students making the decision to stick with their preparation for future employment in the field.  

Think about it. You, too, can make a difference. Would you like to be an ambassador for science? 

Departments: From the Editor

Inspiring Science Students

We recently experienced some mysterious and sporadic liquid chromatography (LC) results using an ultrahigh-pressure LC (UHPLC) system. This incident took us a week of intensive investigation involving many hypotheses and numerous blind alleys until we found the root cause. We believe this issue has not been previously reported and could easily be missed by many LC users (1–4). Here is our story. 

Our problems began when we performed potency and impurity assays of two newly developed clinical drug products (capsules and tablets). We followed a validated 42-min procedure that had been used successfully for two years and modified it into a faster and equivalent 15-min UHPLC method (5). The sample preparation procedure in the modified method uses 20 mM ammonium formate, pH 3.7, as the extraction solvent. We used an Agilent model 1290 UHPLC system (Santa Clara, California) with 3-µL injections and performed many area-percent analyses without any issues. However, this time the potency assay failed system suitability tests because the response factors for duplicate standard (calibrator) preparations differed by ~20%. Curiously, peak area repeatability from the same vial was excellent (<0.3% RSD), indicating that the autosampler was not malfunctioning. Duplicate standard preparations were then made by two analysts. Highly variable results and high responses for the calibrator were observed, which led to artificially low potency values of the drug products. We decided to launch a more systematic investigation.  

We started our investigation by interviewing the primary analyst and checking all the calculations for the method. We noted higher than expected responses from some calibrator vials but not others — even though all calibrator solutions (~0.5 mg/mL) should have provided nearly identical responses. It seemed quite inconceivable that we obtained responses from two duplicate calibrator preparations that varied as much as 20–30%. Peak areas in chromatograms of drug product extracts typically were quite consistent. Next, we scrutinized each analyst's weighing techniques and the quantitative transfer procedure of the weighed powder to the volumetric flask. We verified the analytical balance calibration and visually checked the homogeneity of the reference standard. At that point, all practices appeared to be in order, and we found no clues for the disparate responses. Next, we turned our attention to the existing sample preparation procedure and designed an experiment using stronger extraction solvents. For quicker problem diagnosis, we switched to a 2-min fast LC, nonstability indicating method (Waters XBridge C18 column [Milford, Massachusetts], isocratic mobile phase consisting of 25% acetonitrile in 0.1% trifluoroacetic acid, with a 1-mL/min flow rate and a 2-µL injection volume) (6). The results of the extraction study are summarized in Table I, which shows encouraging results in study 1 with all three extraction solvents (ammonium formate, 0.1 N hydrochloric acid and 20% acetonitrile in 0.1 N hydrochloric acid) yielding the expected potency of ~100%. However, reinjecting the very same sample vials on a different model 1290 UHPLC system yielded data with 30% higher peak areas for the standards and low potency results of ~74–78% (study 2, Table I). A repeat of the same sample set in study 3 using the 15-min UHPLC method on the second system also yielded the low potency results, but at ~8–11% higher potency than those in study 2 (fast LC). A review of historical response factors showed higher responses for some of the calibrator solutions as the key issue. The high calibrator responses were consistent with an extra injection volume of ~0.6 µL; that is, 30% for 2 µL, 20% for 3 µL, and 6% for 10 µL. We were puzzled with these consistent observations that could be caused by an extra 0.6 µL that was injected at some times, but not at others. 

 Table I: Drug product potency assay results with different extraction solvents 

We ruled out instrumental problems because the same observations were made with two UHPLC systems — both well maintained and almost brand new. We considered other hypotheses such as software issues or corrupted sample sequences, contamination of the sampling needle with surfactant from the drug product excipients, overfilling the sample vials, and over-tightening of the septa that might cause pressurization of the vials (3,4). We designed experiments to test each hypothesis using the 2-min isocratic method, which allowed us to watch each injection closely and check the resulting peak area quickly. Nevertheless, none of these hypotheses could be verified. Finally, during close scrutiny of the one of the "offending" calibrator vials, we saw a "shimmering" reflection caused by an extra layer of liquid adhering to the neck of the vial (Figure 1). This was a "eureka" moment, and we knew we had found the root cause. 

 Figure 1: A picture and a drawing of an HPLC vial containing an extra layer of liquid near the neck of vial. This extra layer of liquid is easily formed by inverting the filled vial and is physically stable. 

As it turns out, our analyst is very conscientious and often shakes or inverts LC vials to homogenize sample extracts before loading sample trays. Because of the high surface tension of the aqueous buffer, an extra layer of solution adheres to the septum or forms around the inside of the neck of the vial. This layer of liquid somehow causes an extra ~0.6 µL to be injected. Drug product extracts have surfactants from the formulation that reduce surface tension and tend not to form this extra layer. As a result, the calibrator solution injections are larger, leading to biased results for the samples. To verify this explanation, we conducted an experiment on two calibrator vials without and with the extra liquid layer. Figure 2 shows that while both responses are linear and have the same slope, the 

-intercepts are 0.00 µL and -0.65 µL respectively, indicating a bias of 0.65 µL from the vial with the extra layer. 

Figure 2: Two response curves obtained by injecting the same standard solution from two vials: (a) without and (b) with the extra liquid layer. 

Why Doesn't This Problem Occur More Often?

The next question was obvious: Why hasn't this problem been observed previously or reported in the literature? The extra layer appears to form easily using all types of LC sample vials. We subsequently tried to reproduce this problem with other HPLC systems in our laboratory (Waters Alliance, Waters Acquity UPLC, and Agilent 1200) without any success. The Waters Alliance system has a side-port sampling needle and the Waters Acquity UPLC system has a double-needle design; those systems do not appear to be susceptible to over-volume injection from the extra layer. However, the Agilent 1200 autosampler appears to have a similar injection mechanism to the 1290, so why wouldn't it be affected?  

Response from Agilent indicated that the model 1290 UHPLC autosampler uses an injection technique that pulls an air gap of ~0.6 µL into the tip of the sampling needle inside the vial near the septum before the needle leaves the vial. The use of an air gap, sometimes referred to as the "leading bubble technique" (7), has long been known (8) as a way to decrease band-broadening and dispersion during sample injection. Some Agilent autosamplers use this approach to increase the precision for samples of small volumes. In the present case, instead of an air gap, the needle will contain ~0.6 µL more sample if a plug of liquid is trapped in the vial neck after vial inversion. According to Agilent Technologies, a new firmware revision is under evaluation that will allow reproducible injections of very small volumes without this issue. 

Another useful observation is that this extra layer of liquid can be formed only for sample-solvent strengths less than 20% acetonitrile or 25% methanol. Liquids with higher organic-solvent concentrations have lower surface tension, which will not allow the extra layer to form. As an alternative to inversion, vortexing sample vials provides any necessary mixing but lessens the chance of forming this layer because of the rapid vibration of the vortexing action. 

We learned several valuable lessons from this investigation and troubleshooting episode.   

Don't invert LC vials before loading them into the sample tray. Most sample solutions are homogenous, and further agitation should not be required. In cases where agitation may be helpful (for example, frozen samples, emulsions, or suspensions), use vortexing and check afterward to confirm the absence of any extra layer of liquid. 

Use stronger extraction solvents (if possible) to ensure robust and quantitative extractions from all sample matrices. Our revised drug product method now uses a stronger extraction solvent of 20% acetonitrile in 0.1 N hydrochloric acid and a larger injection volume of more-diluted sample solutions (5 µL for UHPLC or 20 µL for HPLC). The increased organic reduces the solution surface tension, thus minimizing the chance of forming the extra liquid layer in the vial neck; the larger injection volume reduces the percent error. 

Effective LC troubleshooting may include looking for common symptoms and verifying corrective actions (for example, instrument, column, baseline, pressure, data, and sample preparation) using a step-by-step procedure to isolate and diagnose the cause (1–4). A good strategy is to propose hypotheses and then conduct experiments to verify or eliminate each potential causative factor. The use of short run times can speed up such problem diagnosis.

Although toll-free technical service hotlines and service engineers typically are helpful in troubleshooting, there may be circumstances where personal vigilance and keen observations are required by analysts. Here, keeping our eyes open and focused on the key issue provided us with the solution to this mystery of sporadic potency results.

The authors thank Jack Pellett, Derrick Yazzie, Tania Ng, Max Zhong, Kelly Zhang, and Nik Chetwyn from Genentech (San Francisco, California); Tom Waeghe of Mac-Mod (Chadds Ford, Pennsylvania); Tom Jupille and John Dolan from LC Resources (Walnut Creek, California); and Christian Gotenfels of Agilent Technologies for their helpful discussions and suggestions. 

is a research assistant at Genentech, Small Molecule Analytical Chemistry and QC, in South San Francisco, California. 

is senior scientist responsible for multiple early phase drug development projects and new analytical technologies at Genentech, Small Molecule Analytical Chemistry and QC, in South San Francisco, California. He is also a member of LCGC's editorial advisory board. 

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

 for more than 25 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 

 (Humana Press, Clifton, New Jersey, 1989).  

HPLC Maintenance and Troubleshooting Guide, in Modern HPLC for Practicing Scientists

 (Wiley-Interscience, Hoboken, New Jersey, 2006), Chapter 10. 

Handbook of Pharmaceutical Analysis by HPLC

, S. Ahuja and M.W. Dong, Eds. (Elsevier/Academic Press, Amsterdam, 2005), Chapter 9.   

LC Troubleshooting

A mysterious autosampler problem is solved.

A Case of Sporadic LC Assay Results

Kevin Schug, of the University of Texas at Arlingto n, interviews Harold McNair, professor emeritus of Virginia Tech's chemistry department, about his illustrious career in chromatography. 

 Who introduced you to gas chromatography (GC)?

 Dr. A.J.P. Martin, Dr. Steve Dal Nogare, and Professor A.I.M. Keulemans. You could not have wished for a better group of mentors. In 1957, I performed my first GC injections. I was a graduate student at Purdue University (West Lafayette, Indiana) and had just finished a Masters degree in electrochemistry. I wanted an exciting new topic for my PhD work. I took a summer job at Amoco's Research Labs in Whiting, Indiana. My job was to screen around 80 new liquid phases for selectivity to separate butane-1 and isobutylene as well as some C

 unsaturated methyl esters. During that summer job, Nobel Laureate A.J.P. Martin came to the labs to install his new gas density balance detector. 

I was assigned as his "gofor" for the day. It was a memorable experience; he really turned me on to GC with his knowledge, enthusiasm, and tips on making tea (heat the cup!). The next summer, I worked with Steve Dal Nogare from DuPont in Wilmington, Delaware, and was indoctrinated to temperature-programmed GC. Steve was a great mentor and very enthusiastic about GC. Finally, my Fulbright fellowship with Professor A.I.M. Keulemans in 1959 convinced me I would work on GC for as long as I could. So many of the pioneers in GC came to his laboratory and I was fortunate to work with many of them — and this fired my enthusiasm for GC. 

 What do you see as your greatest achievement in the field?

 My greatest achievement is collaborating with my students — undergraduates, graduates, and postdoctoral fellows — and visiting professors (easily more than 200 total). This group of intelligent, humorous, hard-working people made my research and teaching both interesting and challenging — most of the time! It would be unfair to mention any single individual as there were too many major breakthroughs from a wide range of people. Highlights would definitely include reporting the first capillary gas chromatography–mass spectrometry (GC–MS) results in 1961; introducing temperature programmed liquid chromatography (LC) in 1981; using mobile-phase modifiers to stabilize retention times on silica gel; developing the first directly coupled LC–GC experiment using two independent computers in 1981; identifying the role of pH in capillary zone electrophoresis (CZE) to add to the electrostatic mobility of highly hydrated protons; and also early work on analyzing steroids in urine by gas chromatography time-of-flight mass spectrometry (GC–TOF-MS). The latter study was one of the earliest demonstrations of the use of TOF-MS coupled to GC for determining traces of steroids in urine and this work was honored with an "outstanding PhD thesis award" from Virginia Tech (Blacksburg, Virginia). 

A research project in 1976, which correlated p

's of weak aromatic amines with retention time on acidic silica gel, was recognized by the National Aeronautics and Space Administration (NASA) as the second most cited scientific publication for that year. The Viking Mars Landing was number one! Our work was one of the first publications on the LC analysis of aromatic diamines. Not only did this class of compounds hold high interest for NASA in relation to polymer research, but they also held high biological significance; thus, many researchers in the biomedical and pharmaceutical communities also paid close attention to this work. 

 Figure 1: Red Lion Inn, Blacksburg, Virginia. McNair and his students versus HPLC short course students in a game of volleyball (1997). 

But equally as important as the quality of the scientific publications was the work ethic, the group effort, and strong friendships I established while at Virginia Tech. Some of the best memories were the long van rides to Pittcon each year. I was always so proud when one of my students gave a great lecture at some conference. I am still in contact with almost all of my former students. 

 Did you consider entering academia after your postdoctoral fellowship?

 Not really. I was focused on industry. I had spent five summers working in industry at Phillips Petroleum Co., Amoco Central Research, American Cyananid, DuPont, and Esso Analytical R&D. I knew industry had better equipment, strong support staff and services, real significant problems, and, of course, much better pay. In fact, when I did come to Virginia Tech in 1968, my plan was to stay only a few years then go back into international marketing, hopefully in Europe, in the scientific instrument business. But after three years, I fell in love with teaching students, choosing my own research projects, and being able to play tennis almost any time and day. Blacksburg was a great place to raise a family.

Incidentally, I did actually get a tennis and a scholastic scholarship for the University of Arizona and played there for two years until the number of physics and chemistry labs every week left no time for tennis. I still enjoy hitting and try to play two or three times a week. 

 You told me that two of your favorite activities, teaching short courses and writing books, came from your industrial experience. Why is this?

 After one year with Esso R&D in 1961, I returned to Europe to set up operations for F&M Scientific (later H.P., Avondale, Pennsylvania). Even though Europe had introduced and pioneered the art of GC, there were few progressive instrument manufacturers. F&M Scientific and later Aerograph (Varian) realized the importance of temperature-programmed GC; both became major leaders in the field based on the quality of their instrumentation. I lived in Amsterdam from 1962 to 1964 and European customers were begging for seminars, hands-on training courses, booklets — anything to help them learn about GC. Part of my job was then, and still is today, telling everyone how great GC really is. My book 

 originated from a lecture at the University of Athens, Greece, in 1963. After one lecture, the faculty begged me to come back the next day. After four lectures in four days, they begged me to write up the lectures, which became the first draft of 

. It is written in straightforward English for Europeans with English as their second language. It was a simple book, easy-to-read, and was eventually translated into eight different languages with more than 130,000 copies sold. 

After joining Varian in 1965, I set up a one-person unit, GC training, which offered "hands-on" GC training, across the United States initially, but later in Europe, Canada, Australia, and Mexico. This was a huge success. The unit not only showed a small profit each year, but also, as I recall, about 20% of all attendees bought or had already bought a Varian GC system. It was a great marketing success. 

Thus, logically, I carried the model for short courses and GC books with me to Blacksburg. The short courses helped train my students, generate money in my earlier academic career, recruit new graduate students, and several companies left their best chromatographic systems in my labs. 

 What do you see as the most exciting development in separation science at the moment or in the future?

 Liquid chromatography–mass spectrometry (LC–MS) applied to biological compounds. So many people who are not trained as chemists or chromatographers are buying and using LC–MS regularly to study important problems that involve the analysis of complex mixtures of biomolecules. 

This article was first published in the November 2009 issue of 

. The interview has been edited for length and clarity.

received his PhD in Chemistry from Virginia Tech under McNair in 2002, performed postdoctoral research with Professor Wolfgang Lindner at the University of Vienna (2003–2005), and is currently associate professor in the Department of Chemistry and Biochemistry at the University of Texas at Arlington. 

    Research in the Schug lab spans the fundamentals and applications of liquid-phase separations, in conjunction with electrospray ionization mass spectrometry (ESI-MS). Additionally, ESI-MS is studied as a means for preserving and analyzing noncovalent complex formation between small molecules, while MALDI-MS is used for fingerprinting complex mixtures.

    Schug was the recipient of the 2009 LCGC Emerging Leader in Chromatography Award, the 2009 Eli Lilly and Company Young Investigator Award in Analytical Chemistry, and an NSF CAREER award. Schug credits his research and teaching principles to his previous mentors: Harold McNair and Wolfgang Lindner. He currently manages a group of 21 graduate and undergraduates at the University of Texas at Arlington.

History of Chromatography

An interview with Harold McNair reveals his greatest accomplishments and proudest moments.

Harold McNair: An Innovator in Gas Chromatography Research and Teaching

This two-part column explains the evolution of lipids analysis, current approaches such as targeted and untargeted lipidomics, and the variety of techniques involved, including sample preparation, separations, and, of course, mass spectrometry.

The study of lipid biology is undergoing a remarkable, technology-driven transformation that most notably involves mass spectrometry (MS) and its ancillary techniques, such as liquid chromatography (LC) and ionization sources. Greater investment inflows and a surge of activity, especially in the field of drug and biomarker discovery, is fueling the growth of global lipid analysis — lipidomics — making it a standard research tool in academic, pharmaceutical, and biotechnology sectors. Additional areas of interest in lipidomics include plant, microbial, and nutritional research. 

Lipids are absolutely essential for life, playing diverse and important roles in nutrition and health. Alterations in lipid metabolites are associated with various human diseases including obesity, heart disease, and diabetes mellitus (1,2). The ability to profile the lipid composition of biological samples is important in disease diagnosis and drug discovery, attracting strong social and economic interests. For example, the discovery that cholesterol and triglycerides are linked to heart disease affected clinical testing, as well as drug, food, and lifestyle enterprises. Indeed, for the last 50 years, we experienced a lipid phobia, the emblematic manifestations of which are readily apparent in the many food labels advertising "low fat" and "low cholesterol." Cholesterol-lowering agents, such as statins like atorvastatin (Lipitor, from Pfizer, New York, New York), are among the most successful drug classes, accounting for $21.5 billion a year market in the United States. Other commonplace drugs, such as nonsteroidal anti-inflammatory drugs including acetylsalicylic acid (aspirin) and celecoxib (Celebrex, Pfizer), are also directed against lipid-metabolizing enzymes. However, not all lipids are bad for us. Actually, some lipids promote health, such as omega-3 fatty acids and vitamins A, D, and E, and those benefits explain their exponential growth as food supplements and nutraceuticals.  

New research in lipid biology suggests that, especially in the areas of drug discovery and disease diagnostics, we may have formerly adopted a myopic view of lipid metabolism. For years, biomedical research focused on analyzing only a handful of lipids. From the extrapolated results of those narrowly focused analyses, we sought to understand more than we could reasonably expect to know: the multifaceted mechanisms of action and the effects of drugs, the causes of complex diseases, and the intricate biological alterations associated with those diseases. 

As we move into a new era of lipid analysis, the potential to accurately and rapidly measure hundreds of individual molecular species provides the opportunity to use more complex lipid profiles for drug discovery and disease diagnostics. Because lipids are present in all living organisms, other areas of applications such as plant, microbial, and nutritional research also could benefit from improvements in lipid analysis and a better understanding of lipid metabolism. 

 Figure 1: Representative structures for major lipid categories and examples of core structures in red.			

Lipids constitute one of the largest classes of biological macromolecules. Together with nucleic acids, proteins, and carbohydrates, they are present in living organisms that span the spectrum of biological complexity: animals, plants, fungi, protists, bacteria, archaea, and viruses. Chemically, lipids are hydrophobic or amphipathic small molecules (<1500 Da) of biosynthetic origin, which can be counted on the order of tens of thousands. Lipids have enormously diverse chemical structures (Figure 1) and are classified into eight main categories (3): 					

Each lipid heads its own subclassification hierarchy, according to the classification system proposed by the Lipid Metabolites and Pathways Strategy consortium (LIPID MAPS, La Jolla, California; 

Several websites provide useful overviews of lipid structure and function, as well as analytical procedures for lipid analysis: Lipid Library (

), Lipid Bank (lipidbank.jp), and the Cyberlipid Center (

Parallel to the development of new technologies is how our understanding of the biological role of lipids has changed with time. Lipids were known to serve as the structural backbone of cell membranes and as storage for metabolic energy. Recently, the biomedical community learned that lipids play pivotal roles in regulating a wide variety of cellular processes in all organisms. Within each cell exist thousands of types of lipids whose composition, or lipid profile, changes in response to chemical signals from the cell's environment. Studying lipid profiles can provide insight to certain health and disease processes. 

Comprehensive analysis of a wide array of lipids in biological samples is a challenge primarily for analytical chemistry. These complex mixtures of lipids have a large variety of chemical structures and a large, dynamic range of concentrations. Consequently, interest in adapting novel technologies for lipids analysis continues undiminished. 

Our knowledge of lipid biology has been hampered by analytical limitations. At the beginning of the 20th century, lipid analysis was based largely on gravimetric measurements. The use of radioactive and colorimetric assays allowed the initial characterization of lipid metabolism and the understanding that one lipid species could be interconverted into another one. In the 1950s, lipids were used as molecule tools to probe the basic mechanisms of ion fragmentation following electron ionization (EI) (4,5). As such, they provided a glimpse of the potential of MS in biochemical applications.  

By the 1960s, the application of thin-layer chromatography (TLC) made the separation of the complex mixture of lipids present in biological samples possible.  

Identifying the specific fatty-acyl substituents in glycerophospholipids involved separating lipid classes by TLC, scraping the TLC spots, extracting and hydrolyzing the fatty-acyl group, and derivatizing for gas chromatography–mass spectrometry (GC–MS). Although still used today, this approach often requires the extraction of a fairly large amount of lipids. Moreover, the derivatization process renders analyses highly labor-intensive and insufficiently sensitive to target lipids present at low concentrations in biological tissues. 

In the 1980s, with the development of fast-atom-bombardment mass spectrometry (FAB-MS), we could directly analyze nonvolatile lipids, such as glycerophospholipids, as intact molecules, preserving the information inherent in their chemical structure. We also could determine the position of the fatty acyl substituents on the glycerol backbone of glycerophospholipids and even define the position of double bonds present within the fatty-acyl chain through remote-site fragmentation mechanisms. Nevertheless, fragmentation of the lipid analyte during ionization and the high chemical backgrounds observed in early studies with FAB were disadvantages, in particular for the analysis of complex lipid mixtures, making the interpretation of the lipid spectra a tedious task. 

The development, in the late 1980s, of electrospray ionization (ESI) (see sidebar, "Solvent-Based Ionization Techniques"), for which John Fenn shared the 2002 Nobel Prize in Chemistry, revolutionized the study of lipid biology. ESI enabled the introduction of nonvolatile lipid species from the liquid phase into the gas phase directly, without the need for prior derivatization. Furthermore, ESI greatly simplifies lipid analysis. It is a "soft" ionization technique, which, compared with FAB-MS, results in decreased molecular ion decomposition, better reproducibility, and lower detection limits. Because this soft ionization technique does not fragment lipids, we can correlate comprehensive detection of an entire range of lipids within a complex mixture to experimental conditions or disease state. 

A particular advantage of ESI lies in its ability for coupling to an LC system, making the analysis highly sensitive and quantitative. Also, because of its enclosed design, LC minimizes the exposure of lipids to atmospheric oxygen, reducing sample oxidation relative to that occurring during TLC separations.  

The parallel development of MS techniques lends itself to lipid analysis. Those techniques are tandem MS (MS-MS) and the high-mass-resolution instruments, which include triple-quadrupole, ion-trap, time-of-flight (TOF), Fourier-transform ion cyclotron resonance (FT-ICR), and orbital trap (Orbitrap, Thermo Fisher Scientific, Waltham, Massachusetts) mass spectrometry. MS proved itself a sensitive tool for identifying, characterizing, and quantifying lipid species. Therefore, it replaced some of the immunological, radioactive, fluorometric, and colorimetric techniques used in traditional laboratory analysis. 

Today, mass spectrometry, a term whose meaning encompasses sample preparation, chromatography, ionization and mass analyzers, provides the tools to simultaneously measure the biological levels of thousands of lipids. 

Figure 2: Lipidomics, the systems-level scale analysis of lipids and their interacting partners, can be viewed as a subdiscipline of "metabolomics," under the umbrella of systems biology. Abbreviations: aa, amino acids.			

In the 2000s, the development of new approaches for MS-based lipid analysis led to the emerging field of lipidomics (6–8) (Figures 2 and 3). Lipidomics is more than just the decades-old lipids analysis performed with new technological tools. It is a powerful new way to observe and describe lipid biology. We define 

 as the discipline that studies the large-scale changes in lipids accompanying perturbations of biological systems or specific phenotypes. Because lipids constitute a subset of metabolites (including water-soluble components such as amino acids, carbohydrates, and nucleotides), lipidomics is considered part of metabolomics (Figure 2). Nevertheless, the distinct solubility properties of most lipids (water-insoluble) dictate their separate analysis in metabolomic experiments. Notably, the lipid composition of a biological sample reflects both genetic and environmental components (Figure 2), which differentiates lipidomics from the other "–omics" disciplines (genomics, transcriptomics, and proteomics).  

Figure 3: Exponential growth in the number of publications containing the term lipidomics.			

A typical lipidomic experiment often begins with extracting lipids from tissues or cells (Figure 4). The complex lipid mixture is then analyzed with an MS system. A lipid profile results, yielding information about the lipid composition of the starting material and the abundance of individual lipids it contains. Typically, such experimental readouts are referenced to a control condition, to deduce distinct lipid species whose levels change upon perturbation of the biological system under investigation, which may include diseases, nutritional or pharmacological treatments, environmental stresses, and genetic manipulations.  

Figure 4: Flow chart of a lipidomic analysis.			

Generating comprehensive profiles of the lipids present in biological samples presents a major challenge for lipidomics. The samples vary in concentration (from attomolar to nanomolar) and chemical complexity (thousands of components). Complex analytical strategies study lipid phenotypes as well as perform comparative analysis of a 

, the entire set of lipids present in an organism. Currently, we are pursuing complementary strategies for lipidomic investigation: targeted lipidomics and untargeted lipidomics. 

 is a hypothesis-driven approach that focuses on analyzing a selected group of lipids. These lipids can be related to a specific class or metabolic pathway, a particular set of biomarkers, or substrate and product of lipid enzymatic reactions (9–13). The specific lipids that will undergo analysis are selected according to the questions asked, and ad hoc analytical methods are developed for the lipids' absolute quantification.  

Technological advancements, however, help increase the number of lipids that we can simultaneously quantify in a single analysis. This increase enables us to measure lipids from different biochemical pathways, our objective being to quantify the entire lipidome in a more classical "-omic" style. Today, using selected retention times with ultrahigh-pressure liquid chromatography (UHPLC) together with multiple-reaction monitoring (MRM) transitions and triple-quadrupole instruments, we can monitor several hundred lipid species in a single LC–MS analysis (9–13).  

, the second approach, is a typical "-omic" approach. As such, it is hypothesis-generating and exploratory in nature; its purpose is not to identify each observed lipid but rather to perform an unbiased screening of lipids, obtaining profiles or "fingerprints" of as many lipids as possible in biological samples. By comparing lipid profiles, we can thus determine patterns of variations between different groups: healthy versus diseased, control versus treated, and wild-type versus genetically modified (14–17). Dealing with variations in thousands of molecular species, an untargeted lipidomic strategy applies statistical tools such as computational sorting, multivariate analysis, and pattern-recognition analysis to group the observed changes in lipid species. 

For untargeted lipidomics, mass spectrometers that offer high resolution and high accuracy are generally used, including TOF, FT-ICR, and orbital trap systems (14–17). Interestingly, because untargeted lipidomics lead to the detection of unexpected or unknown lipids in biological samples, elucidating their chemical structures often relies on MS-MS or multiple fragmentation (MS

Part I of this series explains how lipids analysis has evolved.

LCGC North America-02-01-2012

Rapid, Efficient Quantification of Water in Solvents and Solvents in Water Using an Ionic Liquid–Based GC Column

It's All About Selectivity

Inspiring Science Students

A Case of Sporadic LC Assay Results

Harold McNair: An Innovator in Gas Chromatography Research and Teaching

New Frontiers for Mass Spectrometry in Lipidomics, Part I

Electrospray Ionization for Mass Spectrometry

Vol 30 No 2 LCGC North America February 2012 Regular Issue PDF