Have you wondered how your gas chromatography (GC) system sets and controls gas pressures, flows, and carrier gas velocities electronically? Here, we describe requirements for and the operation of electronic gas control systems for GC columns and detectors.

Computerized or electronic pneumatic control (EPC) systems for carrier gas, split flow control, and detector gases abound in modern gas chromatographs (GC), as well as in headspace samplers and column switching systems. The accuracy and repeatability of EPC are superior to that of manual adjustment, and the improved control of instrument parameters greatly reduces the possibility for making gas-related mistakes. Computerized pneumatics excel at controlling column pressure drop or detector gas flow rates. An EPC system generally relieves operators from having to make repetitive adjustments and measurements with a flow meter and stopwatch. However, running an EPC system blindfolded, so to speak, by never cross-checking actual gas behavior with selected set-points, only invites trouble. Like any computer system, the results can only be as good as the column and gas parameters that a user enters. Thus, a good working understanding of how an EPC system works and what goals are to be accomplished is essential for obtaining the best possible results. 

In an EPC system, gas flows from the gas supply input, through a metering valve, into a pressure or flow transducer, and then out to the device — inlet, detector, or other GC component — that consumes the gas. The GC system sends a set-point value to the EPC controller, which returns the measured flow or pressure value from its transducer. The set-point and actual values are compared in the EPC system, which adjusts the metering valve as required to maintain the desired set point. The EPC controller incorporates column and carrier gas characteristics to determine the necessary pressure drop at any given moment. Atmospheric and gas supply pressures plus controller temperatures are included as required to compensate for drift and instabilities. 

In the simplest configurations an EPC channel acts as a carrier gas flow controller for a packed column or controls a detector gas, such as air or hydrogen, for flame ionization detection (FID). Two controllers — one pressure and one flow — can provide the split flow and inlet pressure for a capillary column inlet splitter. Other more complex applications include pressure or flow controllers for auxiliary devices such as purge-and-trap or headspace samplers, or pressure-switching controllers for multidimensional column systems. 

Computer-controlled pneumatics cannot prevent operators from selecting inappropriate column pressures or split flow rates. An operator may easily establish incorrect conditions and become misled as to the reasons for a problem. Because of their complexity and flexibility, computerized pneumatic systems offer analysts more opportunities for errors. 

One of the most common applications for EPC is the control of capillary (open-tubular) column carrier gas. The column flow rate and the carrier gas linear velocity are complex functions of the column dimensions, oven temperature, and type of carrier gas. The mathematical relationships between GC column dimensions, temperature, and flow, pressure, and linear velocity are well understood. Fortunately for today's GC users, all of this is incorporated into the EPC system, which acts as a kind of gas calculator. 

A wide-bore 30 m × 530 µm column makes a good example to help understand the ins and outs of EPC control. At 50 °C with helium carrier gas this column will require about 4.1 psig of inlet pressure to maintain a flow of 5.6 mL/min or an average carrier gas linear velocity of 40.0 cm/s. I know this because I used the EPC "calculator" in my GC system to compute the pressure, flow, and velocity by first entering the column dimensions, carrier gas type, oven temperature, and desired velocity of 40 cm/s. Because I am using a split inlet, I entered the split ratio as well. With this column, a desired split ratio of 20:1 results in a split-flow set point of 112 mL/min. 

An alert reader will realize at this point that they could have entered an incorrect column length or internal diameter, or perhaps failed to correct an existing set of parameters from the previously installed column. This type of error would, for the most part, go unnoticed by the GC system itself unless a required pressure or flow could not be attained by the controller. Suppose that the column internal diameter was mistakenly set at 320 µm. Entering a velocity of 40 cm/s would result in a pressure drop of 20.7 psig. This is achievable by the EPC controller, which upon increasing the pressure to that level would cause the actual 530-µm column flow to increase to 41.9 mL/min with an average linear velocity of 190 cm/s! The split ratio of 20:1 would still call for a split flow rate of 112 mL/min, and so the actual split ratio would be more like 2.67:1.  

These erroneous operating parameters are within the capabilities of the split inlet pressure and flow controllers, so the GC system would not report an error even though the column was operating very far from its desired set point. A subsequent injection might alert the operator to a problem; the peaks would be eluted too rapidly and they would be much larger than expected. If not identified immediately this type of problem could cause serious difficulties later on. Of course, if the retention times were known under the correct conditions then this situation would be evident after one injection.  

Rather than wait until after a run is recorded to discover such errors, the operator can double-check flow rates or unretained peak times after a column has been installed and the preliminary setup completed. Time an unretained peak or measure the flow at the column outlet and compare that to what the EPC system says they should be. Bear in mind that direct flow measurement is more difficult with smaller internal diameter columns, and if necessary use an electronic flow meter that is rated for low flow rates. 

Small deviations in the actual column internal diameter or length will give rise to relatively small errors in actual velocities or flows, as discussed in more detail in another "GC Connections" installment (1). For the best accuracy, it's a good idea to include the stationary phase film thickness, if greater than about 1 µm, with the column dimensions entered into the GC system. 

The pressure drop that's required for a particular column flow depends on the oven temperature as well as the column dimensions and carrier gas, so the operator must specify the temperature at which the desired flow is to be achieved — this is nearly always the same as the initial oven temperature. So far so good, but what happens when the column temperature increases during oven temperature programming? The column flow rate depends on the oven temperature, as well as many other factors, so the flow and velocity will change during temperature programming. The EPC split inlet controller maintains a set pressure level and does not directly control flow or velocity — it must instead calculate the pressure required to establish a desired flow or velocity. What are the effects of choosing these different pneumatic control modes? 

As temperatures increase so does the carrier gas viscosity, which causes the flow and velocity to decrease at higher temperatures if the pressure is kept constant. Figure 1 shows the relationships of temperature and gas viscosity for three common GC carrier gases. EPC systems use these relationships to calculate carrier-gas viscosity dynamically as the oven temperature changes. Figure 1 gives a good idea of how large this viscosity effect can be. Going from 50 °C to 250 °C, for example, causes helium viscosity to increase by about 40%. The other carrier gases undergo viscosity changes of a similar magnitude. 

Figure 1: Carrier gas viscosity as a function of temperature (adapted from reference 2). The dots correspond to published measurements; the solid lines represent interpolated data and the dashed lines extrapolated data. 			

With a constant inlet pressure drop, the column flow decreases during temperature programming. Figure 2 illustrates the effect on carrier gas flow and average linear velocity of changing the oven temperature while holding the pressure drop constant, using our example wide-bore column with a constant 4.1 psig of helium carrier gas. The column flow rate of 5.6 mL/min at 50 °C decreases to 2.5 mL/min at 250 °C, a 55 % loss. Across the same temperature range, the average linear velocity decreases from 40 cm/s at 50 °C to 28 cm/s at 250 °C, a 30% loss.  

Figure 2: Theoretical effect of column temperature on (a) corrected carrier gas flow and (b) average linear velocity with constant pressure drop. Column dimensions: 30 m Ã 0.530 mm; pressure drop: 4.1 psig; carrier gas: helium; column outlet pressure: 1 atm.			

When using a constant inlet pressure with temperature programming, the reduction in flow rate at elevated temperatures can cause unwanted peak broadening as the carrier gas velocity departs from optimum, and it also can extend elution times unnecessarily. Usually in this situation, choosing a different pneumatic operating mode will produce better results. 

An EPC system can be programmed to increase the column pressure drop sufficiently to maintain a constant carrier-gas flow rate as the temperature increases. Figure 3 shows how this choice affects the pressure drop and the average linear velocity. The EPC controller maintains a constant column flow of 6.0 mL/min by increasing the pressure drop from 4.5 psig at 50 °C to 9.1 psig at 250 °C. Note that the average linear velocity increases as the oven temperature goes up, which perhaps runs counter to intuition. But with a constant flow rate the linear velocity can remain in a more efficient region somewhat higher than optimum, which largely will avoid column efficiency losses. Peaks will be eluted sooner and at lower temperatures, reducing the total run time compared to constant pressure operation. 

Figure 3: Theoretical effect of column temperature on (a) carrier gas pressure drop and (b) average linear velocity while maintaining constant corrected column flow. Column dimensions: 30 m Ã 0.530 mm; column flow: 6.0 mL/min; carrier gas: helium; column outlet pressure: 1 atm.			

Large shifts in column flow rate can affect detector operation. Constant column flow is desirable for consistent detector function during temperature programming. Mass spectrometry (MS) detection solute ionization fragmentation patterns depend somewhat on the source pressure, which in turn depends on the incoming flow rate, especially with higher column flows. Granted, the 530-µm column example discussed here is not the best choice for direct interfacing to a benchtop mass spectrometer, but the benefits can be significant in terms of consistent spectra and library searches for those narrower internal diameter columns that can be connected directly.  

When FID is used with hydrogen carrier, it's a good idea to keep the total hydrogen flow through the detector constant. This can be accomplished in two different ways, either by maintaining a constant column flow rate or by programming the detector hydrogen flow to compensate for changes in the column flow, that is, by keeping the sum of the column and detector hydrogen flows constant. 

Ultimately, the most important consideration may be peak resolution and not necessarily detector performance, especially in situations where the separation is marginal. Deciding whether constant pressure or constant flow will yield a better temperature-programmed separation, let alone the effects of changing the temperature program itself, is a more complex consideration that doesn't give itself over very well to purely theoretical modeling. Relative peak positions in a chromatogram depend heavily on the thermodynamic relationships between the stationary phase and a solute's chemical characteristics. As a result, individual peaks move in different ways relative to each other as the temperature changes. 

In general, the more the chemical natures of a pair of peaks differ, the greater the effect of changing the column temperature or temperature program will be. At higher flows, peaks will be eluted earlier in a temperature program and, therefore, at somewhat lower column temperatures. But this change of elution temperature may or may not improve the separation of a pair of peaks if both are not affected in the same way. This effect is illustrated in the series of test chromatograms shown in Figure 4. Here, a test mixture was injected with either constant pressure or constant flow programmed pneumatics, with the initial pressure drop the same in either case. The program rate was doubled in each of four consecutive runs, from 3 °C/min to 6 °C/min, 12 °C/min, and finally 24 °C/min. 

Figure 4: Series of chromatograms at increasing oven temperature program rates with constant pressure or constant flow controlled pneumatics. Temperature programming from 50 Â°C, hold 2 min, then to 250 Â°C at: (a) 3 Â°C/min; (b) 6 Â°C/min; (c) 12 Â°C/min; (d) 24 Â°C/min. Constant pressure at 4.1 psig (P) or constant flow at 6.0 mL/min (F). Column: 30 m Ã 0.530 mm, 1.3-Âµm df EC-Wax. 1.0 ÂµL manual injections split 20:1 at 200 Â°C. Peaks: 1 = n-C10, 2 = n-C11, 3 = n-C12, 4 = n-C13, 5 = 2-octanone, 6 = 1-octanol, 7 = 2,6-dimethylaniline, 8 =  2,4-dimethylphenol. 			

The overall effects of constant flow compared to constant pressure operation are clear in these chromatograms: The peaks all are eluted sooner with constant flow. The last peak in the run, 2,4-dimethylphenol (DMP), is eluted at 24.85 min with constant pressure and at 22.79 min with constant flow at 6 °C/min. Similar reductions in retention times hold for the other peaks when the flow rate is kept constant, across all of the selected programming rates. 

The effects on retention times of increasing the temperature program rate far outweigh the differences observed between constant pressure and constant flow pneumatic modes. The DMP peak moves in from 40.66 min at 3 °C/min (elution temperature = 166 °C) to 11.1 min at 24 °C/min (elution temperature = 250 °C). Again, there are no surprises here — the temperature effects are quite strong. 

The overall effect on peak resolution, however, is somewhat unexpected. The two closest-eluted peak pairs, 

-decane–2-octanone and 2,6-dimethlyaniline (DMA)–DMP, show opposite trends. Resolution increases with temperature program rate for the first pair but decreases for the second pair, regardless of the pneumatic programming mode. Interestingly, the constant pressure mode gives slightly better resolution than the constant flow mode for the first pair, but makes little difference for the later peak pair. Figure 5 plots the resolution of these two peak pairs as a function of the temperature programming rate and the pneumatic mode. 

Figure 5: Peak resolution (RS) as a function of temperature programming rate and pneumatic mode for (a) n-C13 and 2-octanone, and (b) 2,6-dimethylaniline and 2,4-dimethylphenol. P = constant pressure and F = constant flow. Conditions and chromatograms from Figure 4. 			

Although the resolution of DMA–DMP is high in all cases — there is no reason to be concerned with this peak pair itself — if there were other peaks between these two in a "real" sample, then it would be interesting to find an optimum set of conditions for the overall chromatogram instead of choosing the best case for the first peak pair. A temperature programming rate of 12 °C/min might be a good choice because it keeps the resolution of the first pair while still utilizing a good portion of the available resolving power between the second pair.  

A better choice might be to use a two-ramp temperature program instead of a compromise single-ramp program. At the time that the first peak pair is being eluted, the later-eluted DMA–DMP pair has not started to move along the column much, so cutting the temperature programming rate back to 6 °C/min or even 3 °C/min halfway through the run would make some sense. That way, the later-eluted pair would behave in much the same way as found by running the entire chromatogram at the lower rate, while the earlier pair would get the full advantage of the maximum available resolution. Of course, other solutions might include changing to a narrow-bore column or investigating the effects of a different stationary phase composition.  

Overall, these examples demonstrate that there can be significant differences in how peaks behave relative to each other as the column pneumatic and thermal conditions are varied. Clearly, such changes can affect resolution in ways that may not be anticipated. Chromatographers always should validate and confirm that required peak resolution and quantitation performance levels are met after making any changes to the chromatography. 

Computerized pneumatic control adds a very capable multipurpose tool to the chromatographer's tool belt. Such systems make setting up a GC system easier by automating some tasks. They can give more repeatable results from run to run and laboratory to laboratory. Some detector performance improvements can be obtained with constant carrier-gas flow control. And computerized pneumatics excel in facilitating advanced separations techniques. But when it comes to chromatographic performance gains in solute resolution, the choice of constant flow or constant pressure control can produce significant differences for multiple peak pairs in the same analysis. 

Introduction to Open-Tubular Column Gas Chromatography

"GC Connections" editor John V. Hinshaw is a Senior Scientist at Serveron Corporation in Beaverton, Oregon, and a member of 

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

Articles

GC Connections

How does a GC system set and control gas pressures, flows, and carrier gas velocities electronically?

Electronic Control of Carrier Gas Pressure, Flow, and Velocity

Both the volume of the sample injection and the solvent in which the sample is dissolved can affect the appearance of the chromatogram. 

At first glance, injection of sample onto a liquid chromatography (LC) column seems pretty simple — just put the sample in a vial, select the desired injection volume in the autosampler settings, and push the start button. However, before starting an LC method, we should check a few variables to be sure that we are not creating a potential problem. How does the injection volume affect the chromatogram? What about the injection solvent? Does the column size or packing particle diameter make any difference? What symptoms should we watch for that might indicate we have an injection problem? In this month's "LC Troubleshooting" installment, we'll look at a few of these factors so that we can understand what is going on and how to avoid problems. 

Let's start with a conceptual look at the effect of injection volume on the appearance of a peak in the chromatogram. Throughout this discussion I'll assume ideal, Gaussian shaped peaks, and apply distortion to them. I will also assume that the sample concentration is sufficiently small that neither the column or the detector experiences overload. The results will vary a bit with real peaks, which tend to tail somewhat, and are influenced by some LC system characteristics external to the column. Finally, in the present discussion, I will consider only isocratic separations, where the mobile-phase composition is constant; next month we'll look at what happens in a gradient. 

Figure 1: Simulated chromatograms showing the effect of increased injection volume on peak shape. Purple, green, and blue peaks for injection volume ratios of 1, 2, and 4, respectively, when no injection-related peak broadening is present. Red and black peaks represent mild and severe peak broadening because of excessive injection volume. Arbitrary time scale. See the text for details.			

If we start with a small injection volume of sample dissolved in the mobile phase, we might expect a peak such as the purple one in Figure 1. As we double the injection volume, then double it again, we would expect the peak to grow in height and area, as with the green and blue peaks, respectively. As the peak grows under nonoverload conditions, the column plate number, 

 is the baseline width of the peak between tangents drawn to the sides of the peak. If the peak is Gaussian in shape, the present assumption, the width will be 4 standard deviations, 4σ. If the plate number stays constant as the peak grows, and the retention time does not change, this means that the peak width will also stay constant. This can be seen as the peaks of Figure 1 grow from the purple to green to blue tracings. 

With a sufficiently small injection volume, we can assume that all the sample molecules arrive at the head of the column at the same time, and as they migrate down the column, the peak spreads into the familiar Gaussian shape. If the peak is too large, on the other hand, the sample molecules on the leading edge of the peak arrive at the column before those at the trailing edge. Because we're assuming that the injection solvent is identical to the mobile-phase composition, we can imagine that those first-on molecules get a head start in the chromatographic process and move down the column a bit before the last-on molecules get started. The sample molecules on the leading edge of the large injection behave exactly as they would in a smaller injection, so the front edge of the peak appears unchanged. You can see this in Figure 1, where the leading (left-hand) edge of the blue, red, and black peaks overlap perfectly. With a large injection, once the peak height reaches its maximum, the sample concentration is constant, and will remain so until the tail of the injection band is reached. Thus, when too large a sample volume is injected, the peak rises normally, then is flat-topped, then drops back to the baseline with the normal Gaussian tail. I have shown two different large-volume injections in Figure 1, red and black, where the tail (right edge) looks identical to the blue one, but offset to a later time. You can think of this as inserting a constant-concentration band in the middle of the peak, pushing the back edge to a later retention time. For the red peak, the injection volume is just a bit too large and the peak broadens and appears rounded on the top, but the black peak is quite a bit too large and the flat top on the peak is obvious. You can imagine that if the injection volume is big enough, the molecules on the leading edge of the peak could reach the detector before those on the trailing edge reach the inlet of the column. 

If we have two components in our sample, each one behaves independently. In Figure 2, I have shown what happens under the conditions of Figure 1 when two peaks are present. At small injection volumes, the peak sizes grow with injection volume for the purple, green, and blue peaks of Figure 2. Because there is no overload, the separation stays the same, so the valley between the peaks does not move. (The slight offset in the valley for the purple, green, and blue traces is an aberration in the way I smoothed the data.) In the present example, I have set the resolution (

) to 1.5, where the valley just touches the baseline (read more about resolution in last month's installment [1]). When the injection volume is too large, as we saw above, the front edge of the peak behaves normally, but the back edge is shifted to a later time; this happens for each component of the sample. So in Figure 2, for the red and black cases of too large an injection, the front edges of each peak look just as they did for the blue peak. The molecules at the back edge of the first peak of the red and black examples are still at a sufficiently large concentration when the second peak begins to be eluted so that the valley between the peaks is raised and shifted to a later time, thus reducing resolution. This loss in resolution is small, but noticeable for the red case, but for the black tracing, the injection volume is much too large, and the loss of resolution is severe. Note also that the apparent peak center (retention time) and valley shift to longer times as the magnitude of the excess injection volume increases. 

Figure 2: Simulated chromatograms showing the effect of increased injection volume on resolution. Same peak coding and conditions as in Figure 1. See the text for details.			

So to recapitulate, as we increase the injection volume, the peak height and area grow in proportion to the injection volume up to the point when the injection volume is too large. At that point, the peak broadens, and this peak broadening can reduce resolution between two adjacent peaks. The questions now are: How much can we inject before we inject too much? How much loss in resolution can we tolerate? 

Above we saw conceptually what happens when the injection volume becomes too large. Now we need to find out what happens quantitatively. (If you don't care about the calculations, skip to the next section.) You'll recall that resolution (

 are the retention times of the two peaks, respectively, and 

 are the corresponding baseline peak widths. If the peaks are closely eluted, they will have approximately the same peak width, so we can simplify the denominator of equation 2 to the average peak width, 

. For the present discussion, where we are interested in injection volume, it is more convenient to consider resolution in terms of volumes: 

 are the retention volumes of peak 1 and peak 2, respectively and 

The peak volume observed in a chromatogram has two primary contributions, the volume (width) resulting from normal Gaussian band-broadening processes within the column, and the volume resulting from injecting too large a sample (as in the red and black samples of Figures 1 and 2): 

 is the volume of the peak for a very small volume sample (for example, the purple or green peaks of Figures 1 and 2). You can see that the resulting peak volume is simply the square root of the sum of squares of each contribution, with a little extra broadening for the injection volume because of factors (2) that we don't need to go into here. 

 from the retention volume (time) of a peak, 

The retention volume is calculated from the retention factor, 

And the dead volume (in milliliters) can be estimated from the column length, 

We then can combine equations 5, 6, and 7 to allow us to estimate 

 from the column dimensions, plate number, and sample 

As has been discussed in "LC Troubleshooting" many times in the past, it is best to have 

 ≥ 1 to take maximum advantage of the column and to avoid interferences with the solvent front. For the present discussion, we'll assume 

 = 1; the effects generally will be more severe with 

Now, we have all the tools we need to determine the effect of the injection volume on resolution using equations 3, 4, and 8. We can measure retention volumes, 

 with equation 8. The observed peak volume, 

, can then be estimated for different sample volumes, 

Next, let's look at the effect of injection volume on five popular column sizes, three commonly used for conventional LC (<6000 psi [400 bar]) with lengths of 150 or 100 mm, diameters of 4.6 or 2.1 mm, and packed with 5- or 3-µm diameter, 

, particles; one for LC with mass spectrometric detection (LC–MS), 50 mm × 2.1 mm with 3-µm particles; and one for ultrahigh-pressure LC (UHPLC, >6000 psi [400 bar]), a 50 mm × 2.1 mm column containing 1.7-µm particles. These are summarized in Table I. I'll refer to these by their identification numbers 1–5 in the left-hand column of Table I. I have included estimates of realistic plate numbers, column dead volumes, retention volumes (for 

 = 1), and peak volumes for small peaks, all determined from the equations referenced in the footnote to the table. (I'll issue my standard disclaimer here: The numbers in the various tables have been rounded or truncated for display purposes, so if you repeat these calculations, your results may vary slightly from mine.) 

Table I: Column characteristics of selected columns			

In Table II, I've shown the calculated percent loss in resolution for injection volumes of 1–1000 µL for the columns of Table I. We can see some general trends in the data. If we pick some arbitrary value for the acceptable loss in resolution, such as 5%, we can see that the most popular column for conventional LC, column 1, can tolerate between 20 and 50 µL of sample injected in mobile phase. Anything that is done to reduce the peak volume generated by the column will also reduce the acceptable injection volume. For example, column 2 has approximately the same plate number as column 1, so it should give the same separation, but its smaller volume will require a reduction in injection volume to something closer to 20 µL. If we try to conserve solvent and move to column 3, which is just a smaller diameter version of column 2, the column will tolerate only 5 µL of sample. For UHPLC, column 5 will require that we reduce the volume even further. 

Table II: Percent resolution loss for various columns and injection volumes			

Table II is useful if we want to know how much of an impact a given injection volume will have for a particular column configuration. An alternative way to look at the same data is shown in Table III, where I've shown the maximum volume of sample (in microliters) that can be injected in the mobile phase for a given loss in resolution (1%, 5%, or 10%) with each column configuration. For the example above, where we considered a 5% loss in resolution, column 1 can tolerate 37 µL of sample, whereas column 5 needs to be restricted to 2.6 µL. 

Table III: Maximum injection volume for a specified loss in resolution			

Although it is nice to have tables to use for data lookup, I find that it is even nicer to have a general rule of thumb that I can apply. If such rules are simple, they are easy to remember and apply even when the table isn't handy. We can create such rules by comparing the allowable sample volumes, 

, in Table III with the small-injection peak volumes, 

, for the corresponding conditions in Table I, as summarized at the bottom of Table III. From this, we can see that for a 1% loss in resolution, 

. That is, if we inject using mobile phase as the injection solvent, we can inject up to ~15% of the volume of the first peak of interest, which I call the 15% Rule. Similarly, if we can tolerate a 5% loss in resolution, the factor is ~30% of 

. I like to work on the conservative side of things, and I've often used the 15% Rule in "LC Troubleshooting" when recommending injection volumes. By using these rules, you don't have to worry about the 

-value or retention times — just look at the volume of the first peak of interest by converting a small injection peak width from time to volume (

In this discussion, we've looked in detail at what happens to the sample peak characteristics as we inject different volumes of sample dissolved in mobile phase onto an LC column under isocratic conditions. We've seen that the appearance of the resulting peaks is dependent on the peak volume (width) of a small-injection sample, which in turn is dependent on the length and diameter of the column, as well as the packing particle size and peak retention (

-value). As a general rule, any system change that reduces the volume of the peak generated by the column (shorter column, narrower column, smaller packing particles, or shorter retention) will also reduce the volume of sample we can inject before resolution is degraded. Based on the data presented here, we were able to generate some practical rules of thumb, such as the 15% Rule, which states that for a 1% loss in resolution, we can inject up to 15% of the volume of the first peak of interest if we inject in mobile phase. 

The discussion assumed that we were using the mobile phase as the injection solvent. If a stronger (higher percent organic in reversed-phase LC) or weaker (higher percent aqueous) injection solvent is used, or the method uses gradient conditions, the results will vary. We'll consider these in next month's installment. We've also assumed that neither the column nor the detector was overloaded, either of which can change the appearance of the peaks. Finally, we've assumed that the injection volume is the only added factor contributing to peak broadening, which is rarely the case. From a practical standpoint, the injection band can broaden by as much as 50% (3) as it is washed from the sample injection valve. Additionally, extracolumn band broadening because of the system plumbing, detector cell design, and other factors can play a role in the observed peak widths.  

When all the above factors are taken into play, the recommended injection volumes in Tables II and III, or the rules of thumb should be interpreted as guidelines for a good place to start. I recommend doing a simple empirical test to see if the selected conditions are acceptable. Pick a target injection volume for the maximum desired loss in resolution, such as from Table III, then inject this volume plus an injection twice as large and half as large. If the results are acceptable for all three injections, the target volume should be safe. If you see significant changes in resolution or peak shape between the injections, you may need to further evaluate if the injection volume is sufficiently robust for your purposes. 

(3) L.R. Snyder, J.J. Kirkland, and J.W. Dolan, 

Introduction to Modern Liquid Chromatography, 3rd ed.

 (Wiley, Hoboken, New Jersey, 2010), p. 70. 

"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 in Walnut Creek, California. He is also a member of 

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

LC Troubleshooting

What symptoms might indicate we have an injection problem?

How Much Can I Inject? Part I: Injecting in Mobile Phase

Monoclonal antibodies (mAbs) are highly complex recombinant proteins used successfully as therapeutics in many disease indications. Myriad analytical techniques are required to characterize and control the quality of these molecules during their clinical development and commercial production. This installment describes recent advances and best practices in high performance liquid chromatography (HPLC) analysis of intact mAbs and their variants and fragments. Application examples, supported by key references, are included to illustrate the utility of diversified HPLC modes to characterize these heterogeneous molecules.

Monoclonal antibodies (mAbs) are established therapeutics with multiple blockbuster drugs in oncology and inflammatory diseases (1–3). According to 

 published on-line on March, 3, 2014, there were five antibody drugs (Humira, Remicade, Rituxan, Avastin, and Herceptin) and one fusion protein (Enbrel) among the top 10 drugs in 2013, and these accounted for nearly $50 billion in global sales. Hundreds of mAbs are currently under clinical development, targeting diseases in autoimmunity, inflammation, cancer, infection, ophthalmology, and other indications (4). Beyond traditional mAb formats, bispecific mAbs and antibody-drug conjugates (ADCs) are major platforms of active research and development (4–7). 

Figure 1: Structure of a typical antibody molecule with two identical heavy chains and light chains and antigen-binding sites. VH and VL are the variable domains in the heavy and light chains. The molecules can be cleaved enzymatically into a lower Fc and two Fab fragments. Adapted with permission from reference 8.			

MAbs are immunoglobulins (150 kDa) with a typical structure shown in Figure 1 (8). Important structural characteristics and physicochemical properties of mAbs are summarized in Table I and have been reviewed (9,10). Biological activity (that is, therapeutic efficacy) of the protein depends on the exact three-dimensional (3D) structure, which is likely affected by post-translational modifications or degradations (11). In contrast with small-molecule drugs, which are chemically synthesized by a variety of processing routes to achieve a high purity product, recombinant human mAbs (rhMAbs) are derived from a fermentation process using host cells such as Chinese hamster ovarian (CHO) cells, followed by extensive purification. Because of their inherent heterogeneity, the quality of mAb therapeutics is controlled by the production cell line and the bioprocessing parameters used for each batch (7,12,13).  

Table I: Important structural characteristics and physicochemical properties of mAbs			

Critical quality attributes (CQAs) of biopharmaceuticals are described by the International Conference on Harmonization Guideline Q8 (R2). CQAs are a set of physical, chemical, biological, or microbiological properties or characteristics that should be kept within an appropriate limit, range, or distribution to ensure the desired product quality, safety, and efficacy. CQAs are monitored closely during each bioprocessing step, and can be incorporated into the release specifications for drug substances and drug products as exemplified by Table II (14). Important characteristics such as identity, purity, strength, potency, composition, and safety are obtained from multiple complementary analytical techniques, including various modes of high performance liquid chromatography (HPLC). 

Table II: Typical release tests used for mAb therapeutic products. Reprinted with permission from reference 14.   			

Principles of quality risk management and assessment and quality by design (QbD) serve as the foundation for current regulatory expectations for process development (15,16). As an integral part of QbD, assessment of CQAs could be based on their impacts in four categories: bioactivity, pharmacokinetics–pharmacodynamics (PK-PD), immunogenicity, and safety. Typical mAb CQAs include size, charge, oxidation, aggregation, glycosylation, and cysteine variants. Other CQAs include process impurities such as residual levels of host-cell DNAs and proteins, antibiotics, as well as contaminants such as mycoplasma, viruses, or proteases (13,15). 

Analytical methodologies used for quality control (QC) and characterization are the cornerstones of mAb drug development processes. A well-designed quality control system includes procedure controls (in-process monitoring or controls), raw material or excipient controls, characterization testing, stability testing, and final lot release testing and shipping. Regulatory guidances define the requirements of QC practices, which are applied to characterization and method validation (17). An effective QC strategy should provide assurance that products are safe, have consistent quality, and include standard procedures to guard against unexpected incidents that impact drug quality. In the end, the QC-released mAbs must meet predefined specifications acceptable to regulatory agencies (14).  

To monitor the CQAs of heterogeneous mAb products, multiple analytical techniques are used. These include chromatography (HPLC), electrophoresis (capillary gel electrophoresis, capillary isoelectric focusing, sodium dodecyl sulfate polyacrylamide gel electrophoresis [SDS PAGE]), spectroscopy (mass spectrometry [MS], UV–vis spectrophotometry, nuclear magnetic resonance [NMR], infrared [IR] spectroscopy, circular dichroism), and other techniques (multiple-angle laser scattering, differential scanning calorimetry, enzyme-linked immunosorbent assay [ELISA], microbial testing, sterility testing, and binding and functional bioassays) (13,14,18,19). The use of HPLC (11,20–23) and ultrahigh-pressure liquid chromatography (UHPLC) (24–30) for the analysis of proteins and peptides has been reviewed in many books and journal articles (11,20,22–24,26,30,31). In this installment, we focus our discussion on the various modes of HPLC used for extended characterization, lot release testing, and stability testing of intact therapeutic mAbs and their variant and degradant forms. Selected examples and reviews of key references are used to illustrate the best practices and recent developments in mAb analyses.  

Size-exclusion chromatography (SEC) separates molecules based on molecule size by selectively excluding them from an inert porous matrix of controlled pore sizes (32). SEC is typically performed under nondenaturing operating conditions at room temperatures with isocratic mobile phases containing 0.2 M to 0.5 M sodium chloride or potassium chloride and buffered to approximately pH 6. Typical stationary phases are silica- or polymer-based supports with a hydrophilic coating or bonded layer. The Tosoh TSKgel SEC column is widely used in the biopharmaceutical industry for mAbs analyses and has been on the market since 1987 (33).  

The recent introduction of ultrahigh-performance SEC columns packed with small particles provides an opportunity for shortened analysis time and better peak resolution. For example, Jeong and colleagues (31) demonstrated a 5-min separation of an Fab dimer and monomer on a 1.7-µm bridged-ethyl hybrid (BEH) column with improved peak resolution in comparison to an existing 30-min standard HPLC separation (31). This fast SEC method has been qualified for robustness, accuracy, and precision. Diederich and colleagues (34) demonstrated that an interlaced injection technique that could further speed up the decreased SEC analysis time. By comparison of three SEC columns (TSKgel 3000 SWxl, Zenix SEC-250 [Sepax Technologies], and Acquity UPLC BEH200 [Waters]) in terms of throughput, resolution, and precision, they indicated that similar profiles were obtained (Figure 2) and that the Acquity UPLC BEH200 column has the highest resolution, whereas the Zenix column offers an economical alternative to the TSKgel column. This UHPLC method has a 6-min cycle time for single injections versus <2-min cycle time in the interlace format (staggered injections). 

Figure 2: (a) Overlay of single-injection chromatograms of the mAb sample (1.0 g/L) analyzed on three different SEC columns. (b) For comparability, elution volumes were normalized to column void volumes. Columns: 150 mm Ã 4.6 mm, 200-Ã 1.7-Âµm Acquity UPLC BEH200 SEC: 250 mm Ã 4.6 mm, 300-Ã 3-Âµm Zenix SEC-250; 300 mm Ã 7.8 mm, 250-Ã 5-Âµm TSKgel 3000 SWxl. Mobile phase: 0.2 M potassium phosphate buffer at pH 6.2, containing 0.25 M potassium chloride. Adapted with permission from reference 34.			

High-molecular-weight species (HMWS) such as aggregates in mAbs have been correlated with immunogenicity, a high-impact safety risk (35). SEC is a key technique to monitor HMWS as well as other low-molecular-weight species (LMWS). Arakawa and colleagues (36) identified several caveats when implementing SEC, such as its low resolution and the potential of nonspecific adsorption on the column, which leads to inaccurate molecular mass estimation. A careful selection of SEC method parameters is therefore needed to obtain useful data that correlate well with results from analytical ultracentrifugation, multiple-angle light scattering (MALS), or MS.  

In a critical evaluation of ultrahigh-performance SEC separations for mAbs aggregates, Fekete and colleagues (24)observed high plate counts using sub-2-µm columns, but comparatively higher HMWS values of 7–10% versus approximately 4% with conventional HPLC. Although routine SEC analysis does not require high temperatures, these results suggested the possibility of aggregation artifacts formed during the UHPLC analysis at high temperatures and pressure. Our laboratories are conducting further investigation of this aspect of ultrahigh-performance SEC, and results will be reported elsewhere. 

Charge-based separation analyses are typically included in the release and stability specifications for QC and comparability studies to support post-approval process changes. Ion-exchange chromatography (IEC) is a key purity assay for mAb charge variants, that are often functionally active species (37). Polymer-based columns, the most widely used for IEC analysis, are coated with a hydrophilic layer functionalized with ion-exchange groups such as carboxylate, sulfonate, amino, and quaternary amine, and elution is accomplished with salt or pH gradients. Although their numbers are still limited, the increasing availability of sub-3- and sub-2-µm IEC columns is beginning to impact the performance of these separations (38). 

Approximately 20% of the amino acid residues in a typical human IgG1 platform molecule are charged in nature. Post-translational modifications or degradations of these charged residues create variant forms, where either a charged site becomes removed (that is, Lys epsilon amine loss by glycation, amino-terminal free primary amine loss by cyclization, carboxyl-terminal Lys cleaved), or a new charged site is produced (for example, Asn deamidation into Asp, glycan sialylation). Alterations in charge-based content of mAb variants impact interaction with IEC ligands, resulting in different retention profiles. An IEC charge variant method can be used analytically for monitoring changes to the product profile, or preparatively as an isolation technique of new variants for further characterization or identification.   

An IEC charge variant profile for mAbs is typically generated from a negatively charged carboxylate or sulfonate cation-exchange column. Note that immunoglobulin G (IgG) molecules generally are basic in nature, with an isoelectric point at pI > 7 in a near-neutral pH buffer mobile phase system. An ion-exchange profile ideally presents a predominant main peak with minor flanking regions of acidic and basic variants eluting before or after the main peak, respectively (39). IEC profiles of mAb charge variant forms are typically heterogeneous because of multiple modifications on many sites throughout the molecule, often resulting in a series of unresolved peaks, particularly on the acidic side. In contrast, modifications to the charged termini residues are often well resolved from each other and the main peak. 

A comprehensive overview of charge variant modifications and their impact to IEC profile shifts with an extended focus on deamidation and isomerization degradation were documented by Vlasak and Ionescu (40). In a 2001 paper, Harris and colleagues (19) demonstrated the IEC separation, isolation, and subsequent characterization for six different charge variants of the rhuMAb HER2 IgG1 molecule. Strikingly well resolved variants, which differed from the main-peak intact IgG1 native form by a single Asn deamidation (negative charge added, material shifted to acidic) or Asp isomerization (charge preserved, material shifted to basic), illustrated the power of an IEC separation. Chumsae and colleagues (41) profiled a typical mAb IEC separation from which an unusual arginine modification by methylglyoxal was identified by an acidic shift. Removal of positive charges resulted in molecules that were significantly shifted into the acidic region, and seven charge-modified arginine residues, which were located in accessible, flexible complementarity determining region (CDR) sequences, were identified from isolated fractions.   

A frequently used strategy to address the poorly resolved charged variants utilizes a digestion of the mAb molecule with papain before IEC profiling. Creation of smaller Fab and Fc portions reduces the complexity generated by multiple modifications across the intact heterodimeric molecule, and often results in cleaner profiles with more distinct peaks. An interesting example from Moorhouse and colleagues (42) illustrated this approach, where Fc C-terminal lysine variants can be separately assessed from Fab amino-terminal glutamine cyclized variant forms. Individual terminal species with charge differences were fully resolved and sequentially eluted in order of increasing pI: Fc + 0 C-terminal K, Fc + 1K, Fc + 2K, Fab + 2 N-terminal pyroglutamates (pE), Fab + 1pE + 1 glutamine (Q), Fab + 2Q. 

Pace and colleagues (43) applied this same fragmentation IEC strategy to generate independent Fab and Fc deamidation rates in stability studies to assess and improve product shelf life. Kinetic degradation rates were determined based on the level of acidic charge variants in each fragment in response to varying buffer components, pH, and column temperature. Susceptible deamidation residues in the CDR-containing FAb portion could thus be specifically stabilized to extend shelf life. Zhang and colleagues (44) used IEC to monitor a previously unidentified acidic variant form that occurs in the Fc domain under thermal stress conditions (Figure 3). Characterization analyses of the acidic modification identified Asn-330 as an unexpected site of deamidation and isomerization, in the Val-Ser-Asn-Lys sequence. IEC appears to be able to effectively monitor Fc Asn-330 deamidation and isomerization, a major degradation pathway under thermal stress conditions. 

Figure 3: Cation-exchange analysis of mAb A samples under various conditions: (a) unstressed control, (b) thermally stressed at 40 Â°C for four weeks at pH 5.3, and (c) mAb A treated with pH 8.5. The arrow in (b) points to a new degraded peak unique to thermally stressed samples and termed "AV2B" in the paper. Adapted with permission from reference 44.			

				Reversed-phase liquid chromatography (LC) analysis of mAbs has multiple applications across all structural levels, from full-sized intact molecules to Fab and Fc fragment molecules and reduced heavy- and light-chain species, and peptide maps generated by proteolytic digestion (11,21,28,45). The broad utility of reversed-phase LC is because of its well-studied hydrophobic separation mechanism, the availability of very efficient reversed-phase LC columns packed with small-particle, nonporous, fully porous, or superficially porous (SPP) materials, and the use of mobile phases compatible with MS for peak identification and structure elucidation (20,45,46). 			

Many researchers have used reversed-phase LC to exploit these hydrophobic differences in mAb variant forms at the intact or large-fragment levels to provide a quick assessment of process and formulation development samples. Identified variant forms that exhibited altered hydrophobicity (which controls reversed-phase LC retention) include methionine oxidation, glycosylation, disulfide-free thiol status, and amino-terminal cyclization. Peptide mapping by reversed-phase LC–MS is the basis for de novo sequencing, identity assay for lot release testing, and detailed characterization or identification of variant forms (45).  

Reversed-phase LC of proteins and peptides has been an area of intensive research for several decades. The fundamentals and separation mechanism are generally understood and well-documented (11,20,23). Recent column developments specifically for mAbs have used approaches such as bonded phases with low silanophilic silica for better peak shape, short alkyl chain length ligands to reduce retention and adsorption, nonporous, superficially porous, or wide-pore materials to improve mass transfer, and sub-3-µm or sub-2-µm particles for enhanced efficiencies (20,23,26,31,46). 

Fekete and colleagues (47) explored the impact of mobile phase and column temperature on two new wide-pore packing materials (1.7-µm fully porous and 3.6-µm core–shell particles) on the stability and recovery of commercially manufactured IgG molecules (intact IgGs and fragments: heavy chain, Fc, and Fab). To mitigate strong secondary ionic interactions, the authors operated the column at high temperatures (80–90 °C), along with incorporating small amounts (1–3%) of a protic solvent in the mobile phase (such as 

-butanol) to enhance peak shape and recovery. In a separate paper, the same research group demonstrated the use of a coupled-column approach to enhance resolution of mAb separations (30).  

An earlier study published by Dillon and colleagues (48) described the use of reversed-phase LC–MS to screen intact mAb for heterogeneity associated with post-translational modifications (that is, glycosylation) and stability degradations (clipped species). In addition, conformational isoforms were detected and attributed to disulfide linkage heterogeneity for an IgG2 (49). Rehder and colleagues (50) applied similar optimization of reversed-phase LC–MS methods to analyze reduced and alkylated IgG light- and heavy-chain (LC and HC) molecules and for monitoring intact glycan distribution (G0, G1, G2, Man5) and amino-terminal cyclizations. 

Reversed-phase LC analysis of intact mAb also provided an unexpected separation of free thiol variant forms as demonstrated by Zhang and colleagues (51). Under the denaturing conditions of the reversed-phase LC environment (high temperature, low pH, and organic solvent), variants containing unpaired intrachain disulfides were completely separated from the expected, disulfide-bonded paired cysteine mAb (Figure 4). Free thiol variant forms presented as more hydrophobic than completely disulfide-bonded material, and were observed for both the intact mAb and the FAb fragment. The separated minor peaks, which could be completely converted back into the main peak by CuSO

 treatment, were characterized as the unbonded free thiol form of the Cys22–Cys96 disulfide pair located in the V

Figure 4: Reversed-phase LC profiles for the intact mAb A with buried unpaired cysteines using three different columns: (a) 150 mm Ã 4.6 mm, 3-Âµm Agilent (Varian) diphenyl; (b) Zorbax SB-C8; and (c) Agilent (Varian) PLRP-S column; mobile-phase A: 0.1% trifluoroacetic acid in water; mobile-phase B: 0.1% trifluoroacetic acid in acetonitrile; gradient: 36â45% B in 18 min; flow rate: 1 mL/min; temperature: 75 Â°C; detection: UV absorbance at 280 nm. Adapted with permission from reference 51.			

Affinity chromatography (52) utilizes a highly specific, often biologically based binding interaction to capture specific molecules. Antibody molecules themselves frequently serve as highly selective immobilized ligands for isolation of analytes of interest from complex matrices. Common stationary phases in affinity chromatography are protein A or protein G for IgG and boronate affinity for glycated proteins. Affinity chromatography is routinely used during the purification process for therapeutic mAb production. IgG affinity interactions with protein A or protein G are exploited both at large-scale for an initial high-capacity chromatographic separation of the product from host cell materials, and at small-scale for fast, high-throughput analysis for product titer (12). This high-affinity interaction occurs within the Fc C

3 regions, and after extraneous, nonretained host cell materials are washed through. Isolated IgG is typically eluted with an acidic pH step-gradient wash.  

Interestingly, affinity chromatography has been used to further separate Fc region oxidation variants by applying linear gradients for elution. Loew and colleagues (53) explored the impact of oxidation at two highly conserved methionine residues, Met-252 and Met-428, which are located in the Fc C

3 regions, and are directly involved in the binding interaction with protein A. Schlothauer and colleagues (54) exploited IgG affinity for the neonatal Fc receptor protein (FcRn), a physiologically functional interaction with Fc C

3 regions. FcRn affinity columns, prepared in-laboratory via biotinylated receptor material coupled to streptavidin-sepharose, were operated under physiological pH gradient conditions to elute bound IgG or Fc material.  

Another important application of affinity chromatography is boronate chromatography for antibody glycation characterization. Quan and colleagues (55) used an optimized separation that contained hydroxylated shielding reagents in the mobile phase when assaying basic IgG molecules. Shielded boronate affinity chromatography can be effectively used to quantify, isolate, and characterize glycated IgG1. 

Hydrophobic interaction chromatography (HIC) (56) is a separation mode in which the molecules in high-salt mobile phase environments interact hydrophobically with nonpolar stationary phases (for example, hydrophobic ligand bonded to a polymer or a silica support with a hydrophilic outer layer). Common mobile phases are 1–2 M ammonium sulfate buffered at ~pH 6. HIC is an "orthogonal" technique (with very different selectivity) to reversed-phase LC, IEC, or SEC and is useful for characterization of intact and large fragment IgG molecules (56). HIC uses an initial matrix of high-concentration water-structuring salt to expose IgG hydrophobic regions, which can then interact with an alkyl- or phenyl-functionalized stationary phase. A decreasing salt concentration gradient is used to elute the various IgG forms based on relative hydrophobicity. HIC is generally applicable for initial broad fractionation of complex protein solutions (for example, whole cell broths and plasma). However, analytical scientists can apply this technique to effect separations on highly purified IgG product with very subtle modifications. HIC is also a reference technique for ADC drugs for the determination of antibody–drug ratio (57). 

HIC methodologies have provided very effective separations for monitoring mAb aspartic acid isomerization and thiol variants. Cacia and colleagues (58) used an analytical TSK-Butyl (Tosoh) column to separate seven hydrophobically distinct species, related to aspartic acid isomerization variants of an intact IgG1 in a 14-min method. Wakankar and colleagues (59) further utilized an optimized HIC method to generate stability data and degradation rates for Fab aspartic acid isomerization on the same IgG1 molecule. The resulting profile had additional hydrophobic species separated, which were identified as the Fab intact intrachain disulfide and free thiol forms at heavy chain Cys22–Cys29. Harris and colleagues (18) documented the final HIC separation of six Fab-associated peaks: in order of increasing retention and hydrophobicity, LC-Asp32/HC-Cys22-Cys96 disulfide (native material), LC-Asp32/HC-Cys22-Cys96 free thiols, LC-isoAsp32/HC-Cys22-Cys96 disulfide, LC-isoAsp32/HC-Cys22-Cys96 free thiols, LC-Succinimide32/HC-Cys22-Cys96 disulfide, LC-succinimide32/HC-Cys22-Cys96 free thiols. 

Likewise, Dick and colleagues (60) used HIC analysis of an intact mAb for accelerated stability studies by temperature-induced degradation. Isomerization at a heavy-chain CDR2 aspartic acid was quantitatively monitored to establish degradation rates and determine the impact of formulation buffer components. Valliere-Douglass and colleagues (61) used a modified HIC methodology to improve recovery of an IgG2 molecule used in stability studies, by adding acetonitrile to a low pH (5.2) mobile phase and replacing ammonium sulfate with ammonium acetate as the elution salt.  

HIC is an important method to determine an overall drug-to-antibody ratio. Wakankar and colleagues used HIC with low concentrations of acetonitrile added to the low-salt mobile phase to help ensure full elution of antibody-drug conjugates (ADCs) loaded with higher-order drug levels. Using a butyl-nonporous stationary phase, five species of an intact IgG linked with an increasing number of a small-molecule drugs (zero, two, four, six, and eight) were baseline resolved in less than 15 min (Figure 5) (57).  

Figure 5: Hydrophobic interaction chromatography (HIC) analysis of a mAb-vc-MMAE on a Tosoh Bioscience Butyl-NPR column yields five predominant peaks that correspond to mAb containing zero, two, four, six, and eight small-molecule drugs. Inset: An overlay of the UV spectra of the starting mAb and the HIC peaks, normalized to the 280 nm absorbance, showing the increase in the 248 nm absorbance as the level of conjugated linker-drugs increases. Adapted with permission from reference 57.			

The wide applications of HIC for antibody analysis were further demonstrated by Valliere-Douglass and colleagues (62), who developed a variety of finely tuned, optimized HIC separations for intact IgG, Fab, and Fc fragment species. Multiple aspartate isomerization-based intermediates and isoforms in CDRs were fully separated for different IgG molecules. After peaks were characterized, HIC analysis proved to be a superior monitoring method for these degradation changes, as well as for isolating material for functional assessment. Fc fragment variant species were also well separated, including oxidized methionine forms and processed C-terminal forms such as lysine removal and proline amidation. Finally, HIC provided an efficient separation of size-based variant forms, particularly for isolating and enriching clipped species for further characterization. 

Mixed-mode chromatography uses multiple forms of interactions between analytes and the stationary phase (for example, IEC, HIC, SEC) (63). Mixed-mode chromatography offers unique selectivity for some variants that are difficult to separate by other modes of chromatography, and has been successfully utilized for large-scale protein purification in biopharmaceutical manufacturing (63). Unlike reversed-phase LC that denatures proteins, mixed-mode chromatography can be applied to isolate mAbs for further biological characterization. 

The mixed-mode phenomenon in SEC is a nuance that has been exploited for the separation of oxidized antibodies. By coupling two TSK-Gel (Tosoh) columns in series, Yang and colleagues (64) observed a pre-peak in SEC. Further characterization of the pre-peak revealed the oxidation of trypotophan residues (Trp-103) on the heavy chain of the IgG2. Wong and colleagues (65) exploited to separate oxidation variants by running ultrahigh-performance SEC columns under moderate HIC conditions, which separated an IgG2 into size-related variants, and additionally resolved a monomer pre-peak from the monomer main peak. The pre-peak was confirmed by orthogonal characterization techniques as an oxidized Trp-104 IgG monomer, which was 40–50% less active because of the location of Trp-104 in the CDR. The IgG tryptophan oxidation that was generated under the stressed condition can be clearly monitored by this mixed-mode method.  

Hydrophilic-interaction chromatography (HILIC) (56,66,67) provides enhanced retention and selectivity for separations of polar components. The stationary phases used include bare silica, and silica bonded with amide, amine, cyano, and diol polar functional groups. More-complex silica derivatizations produce anion-exchange, cation-exchange, zwitterionic, and mixed-mode stationary phases that can be operated in HILIC mode with appropriate mobile phases. Typical mobile phases are initial high concentrations of polar organic solvents (mostly aprotic but sometimes protic) that are water-miscible (for example, acetonitrile, alcohols) with aqueous buffers for pH and ionic strength control. 

HILIC has been applied to mAb carbohydrate analysis, separating these highly polar glycans compounds in a MS-compatible mobile phase. Ahn and colleagues (68) developed a HILIC separation of fluorescently labeled N-linked glycan structures released from human IgG by using UHPLC. Baseline separation was achieved within a reasonable time (<1 h) for the expected neutral and charged glycans, including many structural isomers.  

Peptide mapping by HILIC provides complementary information to typical reversed-phase LC maps. Zhu and colleagues (69) applied HILIC peptide mapping with unbonded silica to a highly glycosylated therapeutic protein (erythropoietin [EPO]). Although EPO is not a mAb, this work demonstrated the general applicability of orthogonal selectivity, both for very hydrophilic peptides that are not well retained by reversed-phase LC.  

Perspectives in Modern HPLC

The latest approaches to antibody characterization go beyond ion-exchange, size-exclusion, and reversed-phase modes to make use of affinity, hydrophobic interaction, mixed-mode, and HILIC techniques.

HPLC for Characterization and Quality Control of Therapeutic Monoclonal Antibodies

Although it has been happening to some degree for centuries, food adulteration is increasingly becoming a worldwide epidemic, as evidenced by the melamine scandal of 2008 and recent meat and fish substitutions at major food chains. Analysts are applying more sophisticated chromatographic, spectroscopic, and enzymatic analytical techniques to monitor and measure food adulterants, which are often economically motivated. In this installment, guest authors Jeff Hurst and Kendra Pfeifer from Hershey Foods explore high performance liquid chromatography (HPLC), ultrahigh-pressure liquid chromatography (UHPLC), and mass spectrometry (MS) approaches being adopted to keep ahead of the food adulteration game. 

Food adulteration, sometimes called food contamination, is an intriguing topic that can sometimes seem daunting because it is in everyday commerce and not just the analytical laboratory. First, let's discuss a few general thoughts on food adulteration.  

The focus of this installment is on detecting food adulteration with high performance liquid chromatography (HPLC), but a veritable arsenal of techniques are currently being used. These include HPLC coupled with a variety of detection methods ranging from UV to mass spectrometry (MS) and other nonchromatographic techniques such as immunoassay and mid-infrared (IR), near infrared (NIR), and Raman spectroscopy.  

Another interesting point of contention is that there are some who perceive scientists that are involved in food adulteration from the contaminations side as second-rate scientists. However, that seems to be far from the truth because those scientists exhibit not only a certain level of scientific expertise, but also ingenuity. One obviously doesn't condone these food adulteration activities, but we should be aware.  

Food fraud is seemingly large and growing, with articles on the topic appearing in publications ranging from 

 published an article titled "A la Cartel," indicating that organized crime is diversifying into food and alcoholic beverages (1). Several examples are given in the article, including the horsemeat scandal in 2013 and another case in which nearly 2500 jars of honey were filled with sugar syrup. There was also a report on the seizure of 17,000 L of fake vodka worth $1.7 million. Finally, this article contained information from Europol that in the United Kingdom crooks have switched from drugs to food since everyone buys food and drink. 

Despite the recent news coverage, food fraud is nothing new (2). The Romans had laws that focused on the adulteration of wines because wine back then tended to become bad rather rapidly and a number of items were added to improve the flavor. One of the compounds happened to be lead salts, which sweetened the wine and likely added to the lead load in the Roman population. In the early 19th century, Frederick Accum wrote a book titled 

Treatise on the Adulteration of Food and Culinary Poisons Exhibiting the Fraudulent Sophistications of Bread, Beer, Wine, Spirituous Liquors, Tea, Coffee, Crème. Confectionary, Vinegar, Mustard, Pepper, Cheese, Olive Oils, Pickles and other Articles Employed in Human Commerce

 (3). In that time period, used tea and coffee grounds could be purchased inexpensively. The tea grounds were then boiled with sheep dung and ferrous sulfate and colored with a mixture of tannin, copper acetate, and ferrrocyanide while coffee grounds were mixed with other roasted bean, gravel, sand, and chicory. Burnt sugar was used to add color to coffee. In the case of confectionary, lead, copper, and mercury salts were used to make bright colors that were eye catching for children, but toxic. Green vitrol, alum, and salt were added to give beer a good head because beer was sometimes diluted.  

The most likely event that brought the topic of food adulteration to the forefront today was the melamine incident in 2008, which killed six Chinese infants and sickened more than 30,000. Another recent incident occurred in the past month, in which meats used by two very visible restaurant chains was found to be tainted — meat more than a year old that was mixed with fresh meat (5,6).  

, the top five foods targeted for adulteration are milk, olive oil, honey, saffron, and seafood (4). The adulterants can be divided into three categories: targeted, nontargeted, and economically motivated (EMA). This division recognizes that there is a crossover effect and that HPLC plays a key role in the detection of adulterants. Although there is no exact definition for economically motivated adulteration, the United States Food and Drug Administration (FDA) adopted a working definition for EMA as the "Fraudulent, intentional substitution or addition of a substance in a product for the purpose of increasing the apparent value of the product or reducing the cost of its production, that is, for economic gain" (7). Common types of EMAs include substitution or dilution of an authentic ingredient with a cheaper product (for example, replacing extra virgin olive oil with a cheaper oil), flavor or color enhancement using illicit or unapproved substances (such as unapproved dyes), and substitution of one species with another (such as fish species fraud). 

Figure 1: Chromatogram of the separation of cyanuric acid and melamine and their C13 analogs on a 50 mm Ã 2.1 mm, 2.6-Âµm dp 100-Ã Kinetex HILIC LC column.			

In the case of milk, many individuals tend to focus on melamine as the "poster compound" for milk adulteration. Figure 1 shows a chromatogram of a melamine-contaminated milk with the compounds cyanuric acid and melamine identified by MS using an isocratic mobile phase consisting of acetonitrile and 100 mM ammonium acetate. Peaks 1 and 2 are cyanuric acid and its 

C analog and peaks 3 and 4 are melamine and its 

C analog. While melamine is a high visibility target, it is being monitored and anecdotal information indicates that it is now being replaced by urea and even amino acids because they are more difficult to identify. In addition to melamine, soy protein, corn syrup, whey, leather, and even shampoo have been reported as potential adulterants in milk. When leather is added to milk, it can be hydrolyzed to improve its solubility. The use of this material can be detected by the determination of the amino acid hydroxyproline from the hydrolysis of leather protein that is not seen in milk protein (8). There is an active group at the United States Pharmacopeia called the Skim Milk Advisory Group that is investigating potential adulterants in skim milk, one of which is soy. Figure 2 shows an example chromatogram of various samples ranging from skim milk powder to soy (9). Other potential contaminants could be different types of milk such as goat's milk, which can be detected at the 1% level by the determination of beta-lactoglobulin. 

Figure 2: UHPLC chromatograms in Masslynx (Waters) format: red = pure skim milk powder (SMP), black = 99:1 (w/w) SMPâsoy protein isolate (SPI), purple = 90:10 (w/w) SMPâSPI, and green = pure SPI. Detection: UV absorbance at 215 nm.			

The contamination of olive oil with other oils including corn, sunflower, safflower, and sesame requires continual testing to verify it is pure olive oil with both polyphenols and triglycerides used as marker compounds. Furthermore, olive oil contains a higher concentration of oleic acid than other oils, but less linoleic and linolenic.  

In a similar vein, wines are being adulterated by the addition of polyphenols. Other things that have been added to wine include pigments and glycerol to give a wine "body." In a paper published in 

 (10), the authors described an HPLC method for the anthocyanins in red wine, in which elderberry extracts were added to improve the color. The method determined that wine adulterated with the elderberry contained an extra peak attributed to cyanidin-3-bubioside-5-glucoside. 

The third food on the top five list from 

 was honey. The dilution of honey with less expensive materials, such as corn syrup, has been around for decades. The initial work on this topic was done by White of the United States Department of Agriculture using nuclear magnetic resonance (NMR) spectroscopy (11), but as technology evolved there have been a number of HPLC applications to monitor this phenomenon. In addition to the EMA activities, there is also concern about honey being contaminated with the antibiotic chloramphenicol used by beekeepers to treat their hives against the crippling foulbrood disease. Figure 3 provides a chromatogram from a liquid chromatography–mass spectrometry (LC–MS) method developed for this application (12) with either a 100 mm × 4.6 mm RP-18e column (EMD Millipore) or a 250 mm × 4.6 mm, 5-µm 

 Zorbax XDB C18 (Agilent Technologies). For both columns, the mobile phase was 45:55 (v/v) methanol–0.2% aqueous ammonia acetate at a flow rate of 1 mL/min. 

Figure 3: (a) Total ion chromatogram (TIC) and (b) extracted ion chromatogram (EIC) obtained from chloramphenicol with detection by MS in negative ESI mode.			

Because of its cost, saffron is fourth on the list (according to 

, but other literature seems to be more inclusive by indicating that spices, in general, are targets for adulteration. 

 indicated seven spices of concern: cayenne pepper, cumin, coriander, pepper, saffron, turmeric, and salt (13). Unscrupulous suppliers can adulterate cayenne pepper with sawdust and colors, cumin with sawdust, turmeric with sawdust and yellow colors, and saffron with corn silk. Figure 4 provides a chromatogram of Sudan Red I, II, III, IIB, and IV dyes corresponding to peaks 1–5, respectively, which can be used to adulterate spices by enhancing color. The separation was performed on a 150 mm × 4.6 mm, 4-µm Synergi Polar-RP 80A LC column (Phenomenex) using a 65:20:15 (v/v/v) methanol–acetonitrile–water mobile phase. Detection was UV absorbance at 480 nm. 

Figure 4: Chromatogram of sudan dyes (see text for details).			

As was indicated in the introduction, just this month there was an incident in China where tainted meat was mixed with fresh meat and provided to unsuspecting organizations including KFC and McDonald's (6). Last year in Europe, it was determined that horsemeat was mixed with beef. In addition, there is a lot of data indicating that a large percentage of fish is mislabeled, leading to a cheaper fish being sold as a more expensive one (14). This does not include the issue with farmed and wild caught salmon. Although DNA-based techniques have been widely used, a number of HPLC techniques including proteomics, or "foodomics" as it is now called, have been used to help solve this quandary. Furthermore, information about fish fraud indicates that a substantial amount of fish sold in outlets ranging from sushi bars to the local fish market is not as advertised. A recent report found that fish samples purchased at grocery stores, restaurants, and sushi bars in major cities were often mislabeled, including red snapper (actually tilefish); white tuna and butterfish (actually escolar); wild Alaskan salmon (actually farmed Atlantic salmon); caviar (actually catfish roe); and monkfish (actually puffer fish) (14). 

A paper by Giaretta (15) described an ultrahigh-pressure liquid chromatography (UHPLC) method using myoglobin as a marker for meat adulteration with an example given on detecting pork in beef. Figure 5 is an example chromatogram of a variety of meat types. This method used a Protein-Pak Hi Res Q column (Waters) with photodiode-array detection and a mobile-phase system consisting of three buffers in a discontinuous gradient. A final example can be seen in a recent study in which Chou and coworkers (16) used HPLC with electrochemical detection using copper nanoparticle–plated electrodes to differentiate meat from 15 animal species. 

Figure 5: Chromatogram of myoglobin from different sources.			

This column installment has provided a modest snapshot on the use of HPLC techniques to address challenges in EMA of the food supply. This issue will continue to be a moving target with a need to be vigilant to monitor developments of EMA in parallel with the food industry, instrument vendors, and government labs. There is still the need for simpler sample preparation protocols and it would seem that an expansion of the various "ambient" LC–MS techniques in this area could be helpful. One of the techniques that is being touted by HPLC–MS vendors is the application of exact-mass LC–MS, but, in our opinion, one must ensure that peak identifications are appropriate since sometimes standards are not available. 

Finally, as the adulterers become more sophisticated, it seems like we will need more sophisticated HPLC-based techniques such as foodomics paired with other instrumental techniques like immunoassay and Fourier transform infrared (FTIR), NIR, and Raman spectroscopy. 

Swindled: The Dark History of Food Fraud, from Poisoned Candy to Counterfeit Coffee

 (Princeton University Press, Princeton, New Jersey, 2008). 

Treatise on the Adulteration of Food and Culinary Poisons Exhibiting the Fraudulent Sophistications of Bread, Beer, Wine, Spirituous Liquors, Tea, Coffee, Crème. Confectionary, Vinegar, Mustard, Pepper, Cheese, Olive Oils, Pickles and other Articles Employed in Human Commerce 

 (Longman, Hurst, Rees, Orme and Brown, London, UK, 1820). 

http://www.nbcnews.com/id/28787126/ns/world_news-asia_pacific/t/face-execution-over-china-poison-milk-scandal/

http://money.cnn.com/2014/07/21/news/companies/kfc-mcdonalds-china/

http://foodfraud.msu.edu/wp-content/uploads/2014/01/CRS-Food-Fraud-and-EMA-2014-R43358.pdf

(8) "Dionex Solutions: Methods for Detecting Leather Protein Adulteration in Milk," Dionex Corporation, 

http://www.dionex.com/en-us/markets/food-beverage/news-articles/lp-110606.html

(9) J.E. Jablonski, C. Pardo, L.S. Jackson, B. Rohrback, J. Moore, and M. Han, "Chemometrics and UPLC-UV to Detect Adulteration of Skim Milk Powder with Soy Protein Isolate," poster from United States Pharmacopeia Skim Milk Powder Advisory Group. 

is a manager of regulatory affairs in the quality and regulatory affairs department at the Hershey Company. She has a bachelor's degree in chemistry and a master's degree in food science. Kendra has a wide variety of experiences within the corporation with a tenure in the analytical research and services organization before joining the regulatory affairs group. She was active in the Association of Analytical Communities (AOAC) methods committee and implemented robotics in the lab environment. 

is a principal scientist with the Hershey Company. He is the author of a substantial number of papers on HPLC and food analysis and is a member of numerous scientific organizations including the American Chemical Society (ACS), the Institute of Food Technologists (IFT), the American Society for Mass Spectrometry (ASMS), and the American Association for Integrative Medicine (AAIM). He indicates that Hershey was his first "real job." 

Column Watch

The top five adulterated foods, the latest contaminants found in them, and the methods to detect those compounds

When Bad Things Happen to Good Food: Applications of HPLC to Detect Food Adulteration

	Gas chromatography (GC) method specifications can tell us a lot about how much thought went into developing a method and how problematic it may be to implement. Further, the parameters and settings that are not included in the specification can tell us a lot about the quality of the method, and how many "assumptions" we need to make to acquire data. A recent ChromAcademy webcast considered some of the more common issues with GC methods and how one might use the method specification to pre-empt issues. 

	Does the method specify flow rate, head pressure, or carrier gas linear velocity? Linear velocity is arguably the best way to work because it is a function of the column geometry and one can quickly identify if the flow rate is at or near the optimum or optimum practical linear velocity for that particular gas type. Flow rate may vary depending on the local pressure conditions, and head pressure settings are meaningless if the incorrect column has been installed. Remember that most systems measure pressure - from which flow and linear velocity are calculated by the instrument based on column dimensions - so it's important to specify the column dimensions correctly. 

	While it isn't absolutely necessary to specify the septum purge flow (some instruments don't allow you to control this parameter), it does remind you to check it manually once in a while. If you don't have an electronic flow meter to check that your instrument digital read out isn't fooling you, then get one as soon as possible! 

	What is the expansion volume of the sample vapor created on injection? At some point this will have been evaluated to check that the sample solvent isn't "backflashing" out of the liner and causing carryover and peak area reproducibility problems. It's difficult to check this without knowing the inlet temperature, pressure, carrier type, and injection solvent. It's also good to specify the type of liner - "straight through, deactivated quartz glass liner 900 mm × 4 mm." The linear geometry just makes it easier to check the available volume (π

 × length, all in millimeters) to assess the possibility of the sample vapor overfilling the liner. Don't forget that as the injection is typically made halfway into the liner, then it is usual to halve the calculated liner volume to assess the available volume. 

	It's pretty common to see splitless methods without purge times or with purge times that are just not sensible. The split valve should be turned on as soon as all analytes have been transferred to the column to avoid tailing solvent peaks or raised baselines during temperature-programmed analysis. 

How to spot weaknesses in methods before problems occur

LCGC North America-10-01-2014

Electronic Control of Carrier Gas Pressure, Flow, and Velocity

How Much Can I Inject? Part I: Injecting in Mobile Phase

HPLC for Characterization and Quality Control of Therapeutic Monoclonal Antibodies

When Bad Things Happen to Good Food: Applications of HPLC to Detect Food Adulteration

Critical Evaluation of GC Methods

A Single-Method Approach for the Analysis of Volatile and Semivolatile Organic Compounds in Air Using Thermal Desorption Coupled with GC–MS

Vol 32 No 10 LCGC North America October 2014 Regular Issue PDF