Practical Two-Dimensional Liquid Chromatography in Drug Metabolism Studies and Bioanalysis

Q. When did your laboratory start working with 2D-LC and when did you start using this technique routinely in your laboratory? What applications is it being used for?

A: We started with the first implementation around 2012, primarily focusing on improving sensitivity in radioactive drug metabolism studies by concentrating the samples online (1). Often we only consume a small portion of the samples we have at our disposal or that can be easily harvested. Injection of a much larger volume of sample can be a way to improve the sensitivity of your analysis.

We gradually expanded the approach with additional pumps, valves, and columns allowing online preconcentration in combination with heart-cut two‑dimensional liquid chromatography (2D-LC) (2), providing additional selectivity in cases where we encounter co-eluting radioactive metabolite peaks or to speed up the tedious purification process when structure elucidation with nuclear magnetic resonance (NMR) is required.

A downscaled micro-ultrahigh‑pressure liquid chromatography–mass spectrometry (micro-UHPLC–MS) version was built based on the experience we obtained with the systems used for radioactive and ultraviolet (UV) analyses. This system is primarily applied for very sensitive LC–MS bioanalysis (3).

Q. You developed an online preconcentration method using high-volume injections of biological samples for sensitive metabolite profiling and quantification (2). What benefits does this approach offer and could it be useful in other applications?

A: Radiolabelled mass balance studies are a critical component of drug development to provide insight in the absorption, distribution, metabolism, and excretion (ADME) of a new drug. As radioactivity can be counted in exhaled air and excreta, and “what goes in, should come out”, it allows us to make up a mass balance and know when the majority of drug-related material has left the body, and from which route. In most studies, especially for the in vivo human study, the amount of radioactivity dosed is kept to a minimum, while sample volumes are rather large. Therefore, it is common practice to concentrate the samples prior to injection. We used to perform this step in an offline fashion, but more frequently encountered recovery issues potentially related with advances in formulation science, that allow to also advance drug molecules that are “sticky” and/or have low solubility. Using an online preconcentration approach we could easily avoid losses of radioactivity, that is, compound‑related material, and save a lot on otherwise time‑consuming method development. Since the purpose of the drug metabolism studies is the identification and quantification of drug metabolites, the compounds we are analyzing are still unknown prior to analysis. Therefore, we cannot check their stability and recovery, and should avoid any related risk, such as drying steps, during offline sample concentration. Another approach we routinely apply is partial drying of a sample, for example, after protein precipitation with acetonitrile or methanol extraction of faeces. A fixed volume of dimethyl sulfoxide (DMSO) is added to the sample and dried to the DMSO content. This way, complete drying of the sample is avoided at which point the majority of degradation usually occurs and also makes reconstitution of the compounds in solution much more difficult. The DMSO solution can be injected afterwards without affecting chromatographic performance thanks to an online dilution during sample loading with a high content of water.

The approach we use for online preconcentration was, of course, already widely used in many applications. The main difference with “trap-elute” approaches routinely applied, is that we use a much larger 4.6 × 50 mm high‑performance liquid chromatography (HPLC) column for sample loading. This makes it possible to inject up to 100 mL of urine and tens of mL protein-precipitated plasma or blood, faeces extracts, and so on, without deterioration of the chromatographic separation. Moreover, LC resolution often improves as a result of the back-flushing mode we use to elute the trapped sample compounds to the analytical column. We also notice a longer life-time of our analytical columns when this approach is used. Often a replacement of the trapping HPLC column is sufficient to restore deteriorating LC performance after multiple large volume injections.

Q. You combined the method using very high‑volume injections with heart-cut 2D-UHPLC for selective drug metabolite trace analysis (2). What is novel about this approach?

A: This is an extension of the online preconcentration approach discussed previously, which allows large volume injections (1–100 mL) in combination with heart-cut 2D-LC. This setup combines ultimate sensitivity with improved selectivity. The setup uses four columns: two 3-mm-id. analytical columns for the first and second dimension separation, each preceded by a 4.6-mm-id. column to concentrate the sample or trap the compounds of interest. This way we were able to reach a quantification limit of 770 attogram/mL of a tritiated imipramine metabolite by injection of 100 mL of urine (2). For context, 770 attogram/mL is the same concentration as one 4 g sugar cube dissolved in more than 2 million Olympic swimming pools of 2.5 million litres each.

Q. What were the main challenges you encountered when developing this method?

A: These setups are not readily-available, “off-the-shelf” systems. Most UHPLC users and vendors consider a few tens of µL already a large injection volume. Therefore, ten years ago there were not many UHPLC systems available providing the flexibility to go beyond 50–200 µL injection volume. Since the ability to inject larger volumes is the only way to make a combination of UHPLC with online radioactive detection work for metabolism studies (4), we had already tackled this problem in previous years. Using a modular approach with separate autosamplers, pumps, valves, and so on, a lot of powerful combinations can be built. The most challenging part, however, is not the hardware configuration but finding the right software solution to control all the different modules in an easy and robust way. For this we heavily rely on the vendors, and especially their expertise in supporting less routine applications.

Q. What applications is this technique being used for in your laboratory and what benefits does it offer the analyst? Could this technique be useful in other applications?

A: Since the constituents of a metabolism sample are still unknown prior to analysis and often differ from sample to sample, we regularly encounter situations where new metabolites pop up that were not found earlier in other in vitro or in vivo samples. If these metabolites co‑elute with other metabolites, we can use the heart‑cut 2D-LC approach to separate these, which is the only way for accurate quantification in radiochemical detection for which baseline separation is required.

Another application is selectivity enhancement, separating analytes of interest from interfering matrix components. Despite the fact that we are dealing with relatively dirty in vitro and in vivo samples (such as urine, blood, faeces, bile, and tissue), we usually keep sample preparation to a minimum since we don’t know the metabolites we are analyzing yet and any sample preparation might introduce loss or degradation of some or all of the metabolites. The online preconcentration approach previously discussed is very good if a selective detection method such as radioactive detection or qualitative MS is used. If, however, the detection suffers from selectivity (such as UV) or matrix effects (for example, quantitative MS), heart-cut 2D-LC can be a solution. We use this, for example, to get baseline separated UV peaks of metabolites for which we don’t have authentic standards. UV detection is more reliable to estimate the relative abundance of metabolites (if the chromophore is not dramatically changing relative to the parent drug) compared to MS, for which the response is more structure dependent. The combination with an HPLC trapping column prior to the first dimension, allows us to further boost the sensitivity of our analyses. Another application is the purification of metabolites for NMR structure identification, requiring relatively large amounts of compound in high purity. The combination of online preconcentration with heart-cut 2D-LC offers a way to purify large volumes of sample with high chromatographic performance on 3-mm-id. columns packed with sub-2.5-µm particles. The purification could also be done with offline 2D-LC. The main benefits of the online approach, however, is that it reduces manipulation of the sample and, thus, reduces the time and potential errors and artefacts during sample preparation. A larger volume injection and trapping between the two dimensions is preferred over drying and reconstitution in a small volume, with the risk of degradation or incomplete recovery of the analyte of interest.

Q. The 4-column multidimensional micro-UHPLC–MS method you developed offers high sensitivity and selectivity for the quantitative analysis of drugs in biological samples (3). What is novel about this method and what benefits does it offer the analyst?

A: The micro-UHPLC-MS setup was built based on the setup and experience obtained with the system using two 4.6-mm-id. trapping columns and two 3-mm-id. analytical columns. While the previous setup was aiming for sensitive and selective LC–radioactive and LC–UV detection, this setup was targeting ultrasensitive LC–MS quantification. Since sampling efficiency in electrospray ionization (ESI) largely improves with decreasing flow rates to the ESI emitter (5,6,7), a gain in MS sensitivity can be obtained in going to columns with a smaller internal diameter. Although in this perspective, nano-LC should provide the biggest gain in sensitivity, we are making use of micro-LC in this setup because it is considered the best compromise between sensitivity and throughput, which is also very important in the bioanalytical environment in which we are working. The sensitivity gain, often advocated for nano- and micro-LC–MS, should also be put in perspective of the sample volume available. If the sample volume is not the limiting factor, the sensitivity gain in going to nano- or micro-LC will usually be negligible compared to an injection of a larger volume of the same sample on a larger bore column that has a higher sample loadability. Trap-elute is a potential solution to increase loading on smaller bore columns, but we often observe a less than proportional gain in sensitivity, for example, a 10-fold higher sample loading often results in only three‑fold higher sensitivity. This can be explained by the equally concentrated sample matrix constituents negatively affecting sensitivity by matrix effects (ion suppression). Using heart-cut 2D‑LC, the analyte of interest can not only be concentrated but also separated from the sample matrix. Therefore, we usually see a proportional gain in sensitivity when using the 4-column micro-UHPLC–MS setup. By gradually decreasing column internal diameters from 2.1 mm (first trapping column) to 1 mm (first dimension analytical column) to 0.5 mm (second trapping column) and finally to 0.15 mm (second dimension analytical column) we are able to inject much larger sample volumes onto micro-LC with a reasonable throughput (±15 min total run times). As illustrated in reference 3 the lower limit of quantification (LLOQ) we can obtain with this setup are about 250- to 500-fold better than with direct injection on micro-LC and 10–50 times better compared to trap-elute. For midazolam and its 1’-OH-metabolite we were able to reach an LLOQ of 100 fg/mL. This is comparable to a concentration of one 4 g sugar cube in 16 000 Olympic swimming pools of 2.5 million litres. In the midazolam example, we were limiting ourselves to the maximum injection volume feasible on the nano-LC system due to its maximum loop size and syringe volume. We are confident that, based on the experience we have with the larger-bore 4-column system, we can inject even higher volumes using multiple injection cycles without affecting the chromatographic performance.

Q. How is this approach used in your laboratory?

A: This 4-column micro-UHPLC–MS setup requires more expertise, is more cumbersome to use, and has a lower throughput than our standard UHPLC–MS methods applying 2.1 mm × 50 mm columns and 1–3 min run times. Therefore, we only use it for samples where the absolute sensitivity is required to get results. Typical examples where this setup can be of value in our field are quantification of drug candidates in microdose studies, extremely potent compounds, and (protein and other) biomarkers. Offline 2D-LC and different sample preparation protocols are also considered and can offer some benefits, for example, when multiple analytes are targeted. The main benefits of the online 2D-LC approach in this micro-LC setup are identical to that discussed for the system using larger bore columns, that is, increased automation and reduced sample manipulation, resulting in less chances for errors and degradation or loss of the analyte of interest.

Q. 2D-LC is often regarded mainly as a research tool and not suitable for routine analysis. Is this view changing in practice? Is there more scope to use 2D-LC routinely in pharmaceutical and biopharmaceutical analysis?

A: There is clearly a trend towards increased use of 2D-LC in pharmaceutical analyses, especially heart-cut 2D-LC (rather than comprehensive 2D-LC). While currently 2D-LC is primarily exerted in less routine, non-GXP environments. I am confident that the use in more routine analyses will grow further, especially in areas such as large molecule analyses (such as proteins and oligonucleotides) where sample preparation can be challenging and one dimensional LC does not offer the required separation power.

Recently, more and more LC vendors are offering 2D-LC instrumentation as a full package or as an optional expansion of regular LC systems including the right software to warrant better robustness (for example, communication between the different modules) as well as improved user friendliness. This proves the increasing interest of the industry in 2D-LC and significantly lowers the threshold for new users and application in more regulated and routine environments.

In order to be successful in the implementation of technology such as 2D-LC in the (bio)pharmaceutical industry, it is important to make its use as simple and straightforward as possible. There is no time to tweak the system for every new study, nor is there time to allow elaborate method development for every sample. This is also clear in the choices we made for the setups and methods we are using. The valve setup is chosen in such a way that it allows switching between the larger volume injections (large loop size) using online preconcentration and smaller direct injections (small loop size) without any replumbing. We use trap (5–10 μm particles) and analytical columns (2.2–2.5 μm particles) packed with the same stationary phase avoiding that any analyte of interest would be more retained on the trap than on the analytical column. We use stationary phases in the first and second dimension that might not be extremely orthogonal but work for most drug‑like molecules, such as, C18 and phenyl-hexyl, and always combine this with an as orthogonal as possible pH (for example acidic and basic) between both dimensions. The use of a trapping column in combination with solvent dilution between both dimensions might look more complex than a simple loop approach but allows the use of orthogonal conditions in the second dimension without having to worry about elution strength (8,9,10). The only thing that needs some, but little optimization is the solvent strength for loading the sample and the dilution ratio with water to avoid breakthrough of analytes. A more apolar analyte will generally require a higher solvent strength to avoid loss by adsorption to the loop wall. Since metabolites are usually more polar than the candidate drug, we use the candidate drug for optimization and to test potential losses or carry over. Next, we adjust the loading pump/dilution pump flow rate ratio to get the right dilution factor depending on the expected polarity of the metabolites.

Q. Do you have any advice for analysts who are using 2D-LC for the first time (or thinking about it)?

A: Every implementation of new technology should, of course, bring value. Therefore, it is important to define your current challenges and weigh the pros and cons of an offline sample preparation and/or offline 2D-LC approach versus an online approach. If, for example, instrument availability is the major bottleneck in your laboratory, an online 2D-LC approach that generally consumes more instrument time might not be the best way forward.

When building your own instrument, it helps if you can start from existing, ideally commercially-available, 2D-LC setups. It is also good to go step-by-step and get fully acquainted with every step before you go to the next step. The systems we are currently using were not configured at once, but built step-by-step over many years. Every additional step was only taken when we got fully acquainted with the previous installation of an additional valve, etc., to allow additional functionality. Only when we got full control and enough experience, were we thinking about the next potential improvement. We first optimized the UHPLC-radioactive detection coupling. Then we added two pumps, two valves and a column for the online preconcentration and expanded this approach to three other systems. Then the initial setup was complemented with some additional pumps, columns and valves to allow heart-cutting 2D-LC. The knowledge obtained with this was used to get the 4-column micro-UHPLC–MS system up and running. We added another pump providing a make‑up flow to the MS, to get a larger split ratio going into the radioactive detector. And now we are thinking about the next step where we want to add an extra UV detector to follow both the first and second dimension in UV simultaneously to improve the efficiency in metabolite purification.

Q. Are there any areas of chromatography that you think are particularly exciting or innovative at the moment?

A: The increasing number of commercially available “plug and play” nano- and micro-LC–MS systems is an interesting trend. The new low flow MS sources with integrated column technology or “dummy proof” LC connections allow (almost) trouble-free application of micro- and nano-LC–MS by non‑expert users to gain sensitivity or reduce sample consumption. Hopefully, we can also welcome some smaller footprint systems in the future, integrating nano-LC and MS in one optimized system, thereby further reducing dead volumes to allow higher throughput. This might convince more analysts to use micro- or nano-LC–MS to also reduce solvent consumption for more “environment-friendly” analyses.

If, however, peak capacity is what you are looking for this can now be better achieved in combination with 2 m columns filled with a stationary phase based on micropillar array technology (11) for high-resolution separation power, which is definitely another promising and exiting innovation in the field.

References

1. V. Koppen, R. Jones, M. Bockx, and F. Cuyckens, J. Chromatogr. A. 1372, 102–109 (2014).
2. V. Koppen, C. Van Looveren, I. François, and F. Cuyckens, J. Chromatogr. A. 1601, 164–170 (2019).
3. R. de Vries, L. Vereyken, I. François, L. Dillen, R.J. Vreeken, and F. Cuyckens, Anal. Chim. Acta. 989, 104–111 (2017).
4. F. Cuyckens, V. Koppen, R. Kembuegler, and L. Leclercq, J. Chromatogr. A. 1209, 128–135 (2008).
5. M.S. Wilm and M. Mann, Int. J. Mass Spectrom. Ion. Process. 136, 167–180 (1994).
6. J.H. Wahl, D.R. Goodlett, H.R. Udseth, and R.D. Smith, Electrophoresis 14, 448–457 (1993).
7. D.R. Goodlett, J.H. Wahl, H.R. Udseth, and R.D. Smith, J. Microcolumn. Sep. 5, 57–62 (1993).
8. C. Venkatramani and A. Patel, J. Sep. Sci. 29, 510–518 (2006).
9. R. Majors and P.J. Schoenmakers, LCGC North America 26(7), 600–608 (2008).
10. B.W.J. Pirok, A.F.G. Gargano, and P.J. Schoenmakers, J. Sep. Sci. 41, 68–98 (2018).
11. W. De Malsche, J. Op De Beeck, S. De Bruyne, H. Gardeniers, and G. Desmet, Anal. Chem. 84(3), 1214–1219 (2012).

Filip Cuyckens is Scientific Director and Fellow at Janssen R&D, Drug Metabolism and Pharmacokinetics, Beerse, Belgium.

He earned a pharmacist degree in 1998, a degree in industrial pharmacy in 2002, and a Ph.D. in pharmaceutical sciences in 2003 at the University of Antwerp, Belgium. He has more than 20 years of experience in analytical chemistry, particularly in liquid chromatography, and mass spectrometry. At Janssen R&D he is currently responsible for an Analytical Sciences team in the Drug Metabolism and Pharmacokinetics department. His team focuses on metabolite profiling and identification from discovery to late development, providing support for studies such as metabolic soft spot analysis,

reactive metabolite trapping, and animal and human mass balance studies. Another part of the team performs imaging mass spectrometry and sensitive LC–MS analysis for the quantification of drug candidates, metabolites, and biomarkers. He has authored or co‑authored 80 scientific publications.