In-Loop Analyte Degradation in Two-Dimensional Liquid Chromatography: Example and Solutions

The use of two-dimensional liquid chromatography (2D-LC) to assess peak purity is becoming common in the pharmaceutical industry. In this application space, it is critically valuable to demonstrate that all compounds eluted from a column during an LC assay are accounted for, and that no impurities are “hiding” under the peaks of known compounds. Although the addition of the second dimension (2D) separation makes 2D-LC an exquisitely powerful tool, it also warrants additional care when interpreting the results to avoid misidentification of apparently new peaks as coeluted impurities, when in fact they are analytical artifacts. In this installment of “LC Troubleshooting,” we describe one such possible artifact that can arise because of the degradation of compounds during the transfer of fractions of the first dimension (1D) column effluent to the 2D separation. We suggest simple experiments to determine whether new peaks observed in 2D chromatograms result from degradation, ultimately increasing our confidence in the interpretation of 2D-LC results.

In recent years, two-dimensional liquid chromatography (2D-LC) has gained popularity in pharmaceutical analysis. Although more work is still needed before this technique can be used routinely in Good Manufacturing Practice (GMP) laboratories, it has enabled many powerful methodologies in the drug development space across pharmaceutical modalities ranging from small synthetic compounds to large biomolecules. The applications include complex sample profiling, coupling separation modes in a single method, peak purity assessment, online desalting, trace analysis, and others (1). Among these, 2D-LC peak purity assessment is especially attractive in the development of synthetic drugs to help reduce risks arising from potential unknown organic impurities that could be coeluted with an active pharmaceutical ingredient (API) in conventional one-dimensional LC (1D-LC) analyses, and thus remain “hidden” to the analyst. The International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use guidelines (ICH) require peak purity assessment during LC method validation per ICH Q2 to ensure chromatographic peak homogeneity (often referred to as “peak purity”) (2). However, confirming the absence of coelution using traditional ultraviolet (UV) absorbance or mass spectrometric (MS) techniques can be challenging when impurities are present at low levels (for example, at a 0.05% reporting limit for unknown organic impurities per ICH Q3A [3]), or when dealing with compounds exhibiting high structural similarity, if not impossible (for example, stereoisomers having indistinguishable UV absorption and mass spectra). In comparison, 2D-LC provides unique advantages over conventional 1D-LC in this space by introducing a complementary dimension of separation. The addition of this highly selective separation stage can improve the likelihood of discovering an impurity coeluted from a 1D separation, especially when coupled with MS for characterization.

Challenges Encountered in the Development of a 2D-LC Method for Peak Purity Assessment

There are many possible ways to carry out 2D-LC separations, involving a variety of different interfaces and means of transferring analytes from the outlet of the first column to the inlet of the second one (4). The transfer approach most used in practice involves one or more valves fitted with one or more open capillaries that are typically referred to as sample loops. In this case, the 1D column effluent flows into a loop, and it is held there until the 2D separation is ready to accept the introduction of this fraction of effluent into the 2D column (that is, the injection). This transfer mode is compatible with both heart-cutting (LC-LC) and comprehensive (LC×LC) 2D-LC methods with either gradient or isocratic elution in one or both dimensions. Heart-cutting 2D-LC (reversed-phase [RP-RP]) with loop transfer has been routinely used at Bristol Myers Squibb to assess peak homogeneity for both APIs and synthetic intermediates as part of method development (5,6). In such applications, the method conditions in the first dimension are treated as fixed and based solely on the existing method for the compound at hand. These 1D methods are often diverse in column selection (both chemistry and dimensions) and mobile phase (MP) composition (organic modifier, additives, and pH). As a result, the volume of a 1D peak targeted for analysis can range from 50 µL for an ultrahigh-performance liquid chromatography (UHPLC) method to 600 µL for some high-performance liquid chromatography (HPLC) methods, depending on the physicochemical properties of a specific compound and the method of choice.

To achieve a quantitative 2D-LC peak purity assessment, two actions must be taken:

1. Choosing 2D method conditions that complement those of the 1D separation
2. Transferring the entire target peak to the second dimension for quantitation

A shallow 2D gradient (with analysis time on the order of tens of minutes) is often used. Moreover, stationary and mobile phase chemistries [for example, methanol (MeOH) to complement acetonitrile (ACN), and use of pHs differing by several units to affect the ionization states of target analytes] are chosen to complement the selectivity of conditions frequently used in the first dimension. All these parameters are selected to maximize the likelihood that the 2D separation can resolve compounds that are coeluted from the 1D column.

The quantity of an impurity that is well-resolved by the second dimension could be underestimated or overestimated if the target 1D peak is not quantitatively transferred to the second dimension. For example, an impurity eluted in the tail of the target 1D peak could be missed or underestimated if only the middle portion of the peak is sampled. Conversely, an impurity with an actual level below the 0.05% reporting limit could appear to be present at more than 0.05% (and thus reportable) if it is eluted in the tail of the 1D target peak and only the tail is sampled. In such instances, the total mass of the main peak in the first dimension is underrepresented. Therefore, it is important to find a way to quantitatively transfer the entire target 1D peak. One solution involves splitting the 1D peak into multiple fractions that are sampled into separate loops. These separate fractions can either be analyzed separately using the second dimension (as in Figure 1b), or they can all be injected serially and then eluted once together from the 2D column (this approach is referred to as “multi-inject” and was used to acquire the data shown in Figure 3b). For example, if the interface is fitted with six 40 µL loops and five are filled to 32 µL (80% filling to avoid analyte loss; one loop is frequently used as a bypass in case the fractions cannot be analyzed as soon as they are sampled) (7), the largest 1D peak that can be quantitatively transferred is 160 µL. Of course, wider 1D peaks can be accommodated by changing to a set of sampling loops with larger volumes. In cases where multiple fractions of a single target 1D peak are analyzed separately, collected fractions may be held in the sampling loops for times up to several hours.

Figure 1: “Coeluting peak” observed by 2D-LC separation carried out with multiple fractions analyzed separately: (a) 1D chromatogram obtained using the stability indicating method; and (b) 2D chromatograms of five high-resolution cuts revealing a potential “coeluting” impurity. 1D conditions: Injection volume - 5 µL; Column - Waters XSelect CSH C18, 150 mm x 4.6 mm i.d., 3.5 µm particle size; MPs (mobile phases) - 0.05% trifluoroacetic acid (TFA) in water: MeCN (acetonitrile) (A-95:5, B-5:95); Gradient - 0%B hold for 2 min, 0-100%B in 20 min, 100%B hold for 3 min, equilibration 0% B for 5 min; Flow rate - 1.0 mL/min; Column temperature - 30 °C; Detection – UV (ultraviolet) absorbance at 222 nm. 2D conditions: Column - Agilent RRHD Zorbax Bonus RP, 100 mm x 4.6 mm i.d., 1.8 µm particle size; MPs - 10 mM NH4OAc (ammonium acetate) in water: MeOH (methanol): MeCN (A-95:5:0, B-5:75:20); Gradient - 0%B hold for 1 min, 0-100%B in 15 min, 100%B hold for 1 min, equilibration 0% B for 3 min; Flow rate - 1.0 mL/min; Column temperature - 40 °C; Detection – UV absorbance at 222 nm. Interface: Loop size - 40 µL; 80% fill; Active solvent modulation enabled. API sample: 0.6 mg/mL in water: MeCN (50:50).

In-Loop Analyte Degradation Is a Possibility That Should Be Considered During Method Development

When fractions are held in the sampling loops for long durations, it creates an opportunity for undesired in-loop degradation of analytes captured from the 1D separation. A peak resulting from the detection of a degradant compound can be mistaken as a “coeluted impurity” when the main compound is unstable in the 1D elution solvent or buffer on the timescale of tens of minutes, even if such degradation is not observed during 1D-LC analysis. For example, during the peak purity assessment for an HPLC stability indicating method (Figure 1a), the API peak (~150 µL peak volume) was divided into five cuts for 2D-LC analysis. An additional impurity peak was observed with a resolution better than ten relative to the API in the second dimension (Figure 1b). The impurity level was estimated at 0.09% area, above the 0.05% reporting limit, in the combined cuts. However, the relative peak areas (impurity or API) in individual cuts showed an increasing trend as a function of storage duration in-loop. The fractions that were analyzed last had higher levels (0.19% area in Cut 1 and 0.11% area in Cut 2) than those of the cuts analyzed earlier (not detected in Cuts 4 and 5). Note that the order of analysis for the fractions is dictated by the instrument software and is, in this example, the opposite of the order in which they are collected to avoid cross contamination of the contents of each of the loops. The MS spectrum of this “coeluted impurity” indicated an 18 Da mass increase relative to the API, which coincided with the mass of a known acid-induced hydrolysis degradant that is ordinarily well resolved from the API by the 1D method.

During the development of this stability indicating method (1D-LC), MPs at different pHs were evaluated extensively. Acidic MPs with 0.05% trifluoroacetic acid (TFA) were deemed suitable to separate all specified impurities (Impurity 1-5) without causing significant on-column hydrolysis (Figure 2a). Neutral MPs, in contrast, did not provide sufficient resolution of Impurities 4 and 5 from the API regardless of stationary phase chemistry. The risk of on-column degradation because of the acidic MPs is well controlled under normal use of the 1D-LC method since the level of this hydrolysis degradant, namely Impurity 3, was always under 0.02% area throughout the robustness study.

Figure 2: In-loop analyte degradation confirmed by 1D-LC. (a) chromatograms obtained by 1D-LC using the stability indicating method: (top) API as-is, (middle) API spiked with Impurity 3, and (bottom) API degraded in water: MeCN (acetonitrile): TFA (trifluoroacetic acid) (65:35:0.05, v/v/v) at 30 °C for 30 min.; (b) chromatograms obtained by 1D-LC, but using the 2D conditions from the 2D-LC experiment: (top) chromatogram for Cut 2 from Figure 1b, (middle) API as-is, and (bottom) API spiked with Impurity 3. Other 1D-LC conditions were the same as those in Figure 1, unless stated otherwise.

Recommended Experiments to Check for In-Loop Analyte Degradation

Two experiments can be used to determine whether an apparent impurity peak observed in the second dimension is the result of in-loop analyte degradation. The first approach (Test 1) is to evaluate the in-solution stability of the compound that is the main component of the target 1D peak, using the 1D MP composition (estimated at the point of elution from the 1D column) as the diluent, at the 1D column temperature. In our case, the API was diluted in a water:MeCN:TFA (65:35:0.05, v/v/v) solution and maintained for 30 min at 30 °C before analyzing it using the 1D method (Figure 2a). A degradation peak corresponding to Impurity 3 was observed at 0.16% area, a clear increase over the API as-is sample (<0.02% area). In this approach, Test 1 provided a positive control to confirm the possibility of in-loop analyte degradation, which is also consistent with the observation that the apparent “coeluted impurity” observed in the second dimension had the same MS profile as the known hydrolysis product of the API. The second approach (Test 2) is to carry out a 1D-LC analysis of the API as-is sample using the 2D chromatographic conditions from the 2D-LC experiment. If the apparent impurity resolved by the 2D separation is actually present in the as-is sample, it should also be resolved by the 2D separation alone. Figure 2b shows that no peaks were observed in the API as-is sample in the retention time window expected for the apparent impurity based on the 2D-LC result (about 10.0–10.3 min.; see Figure 1b, Cut 2). The absence of the “coeluted impurity” in the as-is sample in Test 2 served as a negative control, suggesting that in fact the apparent impurity peak observed in Figure 1b was an artifact generated during the 2D-LC analysis. Results from these two experiments (positive for Test 1, negative for Test 2) can be used to assess whether an apparent impurity peak observed in the second dimension is because of in-loop degradation. On-column stability of the analytes of interest in 2D MPs should be also considered before running the peak purity assessment by 2D-LC.

Suggestions to Mitigate the Risk of In-Loop Analyte Degradation

For compounds that are demonstrated to be stable on-column but unstable in a solution prepared using MP as the diluent, a practical solution to mitigate in-loop degradation is to minimize the fraction storage time by immediately analyzing a fraction in the second dimension once it has been collected in the loop. The simplest approach without modifying the 2D-LC configuration is to build a series of 2D-LC methods where each method involves a single cut but taken at different times across the width of the target 1D peak, so that the entire ensemble of 2D-LC analysis yields data similar to what would be obtained from a single 2D-LC involving multiple cuts across a single 1D target (for example, as in Figure 1). However, the extent to which these two results are similar will depend on the retention time repeatability of the 1D method. A slight retention time shift in the first dimension might cause a partial transfer of the main peak or a repeat analysis of some parts of the peak, and these variations could result in inaccurate quantitation of low-level impurities. Here, we describe two solutions that are accessible using commercially available hardware and software.

The first solution is to use larger loops for transferring the entire 1D peak in one cut. For the instrumentation used in this study, the largest commercially available loop is 180 µL. This corresponds to a maximum 1D peak volume of 144 µL (assuming 80% filling), which can cover main peaks encountered in most UHPLC methods, and some HPLC methods. To demonstrate this, the same API peak targeted in Figure 1a was transferred, captured using one loop, and analyzed using the same elution conditions as in Figure 1b. Figure 3a showed that the apparent impurity peak observed in Figure 1b was no longer observed above 0.02% area, again supporting the idea that the peak observed in Figure 1b was in fact the product of in-loop degradation.The second solution is to implement the multi-inject feature described previously, which enables serial injection of up to five fractions into the 2D column at once before eluting the material from the column using a single gradient elution program. To show the impact of this approach, five cuts of the target 1D API peak (32 µL per cut) were all injected into the 2D column following by a single gradient elution program (Figure 3b). In this case, the peak we attribute to in-loop degradation was not observed at more than 0.02% area in the second dimension (with compound identity confirmed by MS). The multi-inject function enables quantitative transfer of larger volume 1D peaks even without the need to change physical loops, which is convenient. Our experience with use of this mode has been that loops need to be washed prior to the next 2D-LC run to avoid carryover, so this is a tradeoff to consider.

Figure 3: Solutions to mitigate undesired in-loop degradation. (a) use a single cut with a larger loop (180 µL) followed by immediate 2D separation; or (b) apply the multi-inject approach to serially inject all 1D fractions into the 2D column before eluting the material with a single gradient elution program. 2D-LC conditions the same as those in Figure 1, unless mentioned otherwise.

Of course, it is also possible to combine the two approaches—that is, use larger loops and the multi-inject feature to accommodate all HPLC methods while mitigating potential in-loop analyte degradation. For instance, a five-cut multi-inject analysis using 180 µL loops could transfer a total of 720 µL peak volume to the second dimension at once. A practical concern is the potential for mobile phase mismatch to compromise the performance of the 2D separation when such a large volume is transferred, even if an active solvent modulation approach is used to mitigate this risk (4).

Summary

Peak purity assessment is emerging as an important application of 2D-LC in the pharmaceutical industry. In-loop analyte degradation can occur in 2D-LC applications, which can lead to false positives. In the context of peak purity assessment, a false positive would lead one to conclude that the 2D separation has separated an impurity from the main compound present in the target 1D peak, when in fact the additional peak observed in the second dimension is an analytical artifact. Obviously, it is essential to understand when this might occur, carry out experiments to check for in-loop degradation when it is suspected, and make decisions when developing 2D-LC methods that minimize the occurrence of such artifacts in the first place.

In this installment of “LC Troubleshooting,” we have discussed a case study that used a real pharmaceutical API to demonstrate what in-loop degradation looks like when it occurs, and tests that can be performed to assess whether or not in-loop degradation is occurring. In cases where the extent of in-loop degradation depends on the time a fraction of 1D effluent is held in a sampling loop, minimizing this holding time is critical. Two accessible ways to do this are to: 1) use large sampling loops so that entire 1D target peaks can be quantitatively sampled using just one or two loops; or 2) use the multi-inject approach that enables serial injection of multiple fractions of 1D effluent into the 2D column at once following by elution using a single gradient program. In the case study discussed here, we have shown that both approaches are viable, however they both have advantages and disadvantages that users should consider prior to implementation.

Acknowledgment

We would like to thank Dr. Yehia Baghdady for his work in implementing large transfer loops.

References

(1) D’Atri, V.; Fekete, S.; Clarke, A.; Veuthey, J. L.; Guillarme, D. Recent Advances in Chromatography for Pharmaceutical Analysis. Anal. Chem. 2019, 91, 210–239. DOI: 10.1021/acs.analchem.8b05026

(2) “Guideline Q2 (R2) Validation of Analytical Procedures”.” International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use. https://database.ich.org/sites/default/files/ICH_Q2%28R2%29_Guideline_2023_1130.pdf (accessed 2024-07-24)

(3) “Guideline Q3A (R2) Impurities in New Drug Substances.” International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use. https://database.ich.org/sites/default/files/Q3A%28R2%29%20Guideline.pdf (accessed 2024-07-24)

(4) Stoll, D. R.; Leme, G. M. Instrumentation for Two-Dimensional Liquid Chromatography. In Multi-Dimensional Liquid Chromatography: Principles, Practice, and Applications, CRC Press, 2022.

(5) Shackman, J. G.; Kleintop, B. L. Peak Purity Assessment in a Triple‐Active Fixed‐Dose Combination Drug Product Related Substances Method Using a Commercial Two‐Dimensional Liquid Chromatography System. J. Sep. Sci. 2014, 37, 2688–2695. DOI: 10.1002/jssc.201400515

(6) Wang, Q.; He, B. L.; Shackman, J. G. Measuring Atropisomers of BMS-986142 Using 2DLC as an Enabling Technology. J. Pharm. Biomed. Anal. 2021, 193, 113730. DOI: 10.1016/j.jpba.2020.113730

(7) Moussa, A.; Lauer, T.; Stoll, D. R.; Desmet, G.; Broeckhoven, K. Numerical and Experimental Investigation of Analyte Breakthrough from Sampling Loops Used for Multi-Dimensional Liquid Chromatography. J. Chromatogr. A 2020, 1626, 461283. DOI: 10.1016/j.chroma.2020.461283

About the Authors

Ziqing Lin is a Senior Principal Scientist in the Chemical Process Development department at Bristol Myers Squibb. His primary focus is on analytical method invention for synthetic drug substances and antibody-drug conjugates. He holds a Ph.D. in Biomedical Engineering from Purdue University.Ziqing Lin is a Senior Principal Scientist in the Chemical Process Development department at Bristol Myers Squibb. His primary focus is on analytical method invention for synthetic drug substances and antibody-drug conjugates. He holds a Ph.D. in Biomedical Engineering from Purdue University.

Qinggang Wang is a Scientific Director in Chemical Process Development in Bristol-Myers Squibb Company. He received his B.S. in Chemistry and Ph. D. in Analytical Chemistry from Tsinghua University, Beijing, China. His work has been focused on CMC analytical supports for various projects within BMS portfolio, inclding small molecules, synthetic peptides and oligonuleotides. He has authored or co-authored 18 peer-reviewed publications and three book chapters in analytical science. His research focuses on application of novel analytical techniques to pharmaceutical analysis. He is the recipient of BMS BMSIARC award for outstanding achievement in analytical science in 2019.

About the Column Editor

Dwight R. Stoll is the editor of “LC Troubleshooting”. Stoll is a professor and the co-chair of chemistry at Gustavus Adolphus College in St. Peter, Minnesota, USA. His primary research focus is on the development of 2D-LC for both targeted and untargeted analyses. He has authored or coauthored more than 75 peer-reviewed publications and four book chapters in separation science and more than 100 conference presentations. He is also a member of LCGC’s editorial advisory board. Direct correspondence to: amatheson@mjhlifesciences.com