Articles

Issue PDF

Click the title above to open the LCGC Europe December 2019 regular issue, Volume 32, Number 12, in an interactive PDF format. 

Vol 32 No 12 LCGC Europe December 2019 Regular Issue PDF

If I increase the flow rate of my separation when using UV absorbance detection, should I expect peak area to change?

	In my personal experience with troubleshooting my equipment in my laboratory, and in thinking about topics for this column, I have found that effective troubleshooting skills and techniques are built on a solid foundational understanding of how the system under study (which is broken, if we are troubleshooting) is supposed to work. On a number of occasions I have found myself thinking about and discussing with students and liquid chromatography (LC) practitioners the impact of flow rate on characteristics of chromatograms and peaks. For this month’s “LC Troubleshooting” I’ve decided to dig into this basic, but very important, topic, with the intention that a deeper theoretical understanding of what should happen will help diagnose problems that may be related to flow rate when something does not look right.

	It is instructive to start a discussion of the effect of flow rate on LC separations with a kind of inventory of possible effects, along with a comparison of the predictions of simple theory and observations from real experiments (supported by more elaborate theories).

	In this article I am going to focus on the last two rows of Table 1 because I have found through discussions with a variety of people that some confusion originates from these topics. Readers interested in the topics addressed in the second and third rows are referred to the references cited there for more information.

Relevant Background on Principles of Detection by Absorption of UV-Visible Light (UV Detection)

	When thinking about the effects of flow rate on UV detection, it is critically important to recognize that we refer to UV detection as a type of “concentration-sensitive” detection. Concentration-sensitive detection is fundamentally different from “mass-sensitive” detection. Readers interested in the differences between these types of detection, and which LC detectors fall into which category, are referred to a recent educational article focused on this topic by Pavel Urban (4). Briefly, concentrationâsensitive detectors respond to changes in analyte concentration presented to the detector (that is, moles/L, or mg/mL), whereas mass-sensitive detectors usually respond to changes in the mass of analyte presented to the detector over time (for example, pg/s). In the case of UV detection in particular, the detector reports absorbance values (

) in response to changes in analyte concentration (

) arriving at the detector. These absorbance values can be related to analyte concentration using the BeerâLambert law:

where ε is a measure of the absorptivity of the analyte, and b is a measure of the length of the light path through the detector flow cell. Readers interested in more details associated with the inner workings of UV detectors are referred to a recent article by Michael Dong and Jedrzej Wysocki in LCGC (5).

Details Related to the Effect of Flow Rate on Peak Height (UV Detection)

	To understand the effects of flow rate on peak height and area, we need to start with a model of chromatographic peaks. In the simplest case, we use a Gaussian distribution as a model of the peak shape, which expresses the dependence of analyte concentration in the LC column effluent arriving at the UV detector on time.

 is the concentration of the analyte arriving at the detector at time i, “moles of A” is the number of moles of the analyte injected into the column, ti is a time point in the chromatogram, tR is the retention time of the analyte, and σv and σt are the standard deviations of the distribution (that is, a measure of the peak width) in volume and time units, respectively. At the apex of a chromatographic peak, 

 and we have exp(0) = 1. Thus, the concentration of the analyte at the peak apex, and therefore the peak height, is entirely determined by the pre-exponential term:

	Now, the moles of analyte injected are not affected by the flow rate, nor is 

. Although there is no explicit dependence of σ

 on flow rate, the flow rate will affect the peak height whenever the flow rate affects the plate height (H) of the column in use, which is almost always the case. The relationship between plate height and σ

 is the column efficiency or plate number for the column, and VR is the retention volume of the analyte (

	From a theoretical point of view, we know quite a bit about the dependence of plate height on flow rate through relationships such as the van Deemter equation (6). The general shape of this type of dependence is shown in Figure 1.

	The details of these relationships are not important here. The important fact is that, for relatively small changes in flow rate, the changes in plate height and σ

, and therefore peak height, will be relatively minor, as shown by the experimental data discussed below. Readers interested in learning more about the dependence of plate height on flow rate are referred to the literature, which is a rich source of material on this topic (7).

Details Related to the Effect of Flow Rate on Peak Area(UV Detection)

	Whereas the peak height is determined entirely by the preâexponential term in equation 1, the peak area is determined by the integral of this equation, where the limits of integration are the time points that define the “start” and “end” of the peak. Indeed, when we talk about peak area we sometimes refer to the “area under the curve”. Now, if we consider a chromatographic peak obtained with a specific set of conditions and think about what happens when we double the flow rate, we will observe that the width of the peak decreases by about a factor of two. The degree of decrease would be exactly a factor of two in a case where the plate number is not affected by flow rate because the ratio of 

 is dictated by the plate number, as in equation 5. However, in most real situations, the plate number is affected by flow rate as discussed above, and the degree of change in width will be slightly different accordingly.

	The net effect of flow rate on peak area in the case of UV detection is a consequence of two things happening at the same time: 1) the peak width changes in time units, expanding or contracting the integration window; and 2) the peak height is independent of flow rate, such that even if the peak becomes wider, time is added to the window over which the analyte is detected at a high concentration. In other words, the analyte flows through the UV detection cell at a finite velocity. The time over which the analyte can absorb photons is determined by the length of the light path the analyte travels through, and the velocity through that path. As the flow rate is reduced, the velocity through the detection path decreases, the residence time increases, and there are more opportunities for photons to be absorbed. Following this logic, we would expect to observe that peak area will increase in proportion to the inverse of the flow rate (that is, ).

	To illustrate the key points made above, I’ve made some experimental measurements of peak height and area at different flow rates, all under isocratic conditions. Figure 2 shows a series of chromatograms obtained at different flow rates in the range of 0.1 to 3.0 mL/min for the analyte acetophenone. From these chromatograms we see two clear trends: 1) the peak height varies slightly across these flow rates, but not in a simple linear way; and 2) the area under each peak obviously increases dramatically as flow rate is reduced.

	Figure 3 shows a more quantitative view of peak height (a) and area (b) results from the chromatograms shown in Figure 2. We see that the shape of the dependence of the peak height on flow rate is the inverse of the shape of the plate height versus flow rate curve shown in Figure 1. Whereas there is a minimum in the 

 curve in Figure 1, there is a maximum in the dependence of peak height on 

 around 1.0 mL/min in Figure 3(a). This is expected because of the inverse relationship between C

	On the other hand, the dependence of the peak area on flow rate (Figure 3[b]) is very different. We see that the peak area increases in direct proportion to the inverse of the flow rate. This is because each part of the peak moves through the detection flow cell more slowly at a lower flow rate, the residence time in the detection zone is longer, and each analyte molecule contributes more to the measured absorbance.

	Our effectiveness in troubleshooting problems with LC separations improves as we deepen our basic understanding of how the separations work. In this article we have examined the dependence of peak height and area on flow rate when using UV detection. Whereas peak height is only weakly dependent on flow rate, the peak area is strongly dependent on 

, and decreases significantly as flow rate is increased. The extent of the expected decrease is important to know when troubleshooting problems with quantitation. For example, a leak between the injector and detector could also lead to decreases in peak area at higher flow rates (and consequently higher pressures).

		F. Gritti, M. Martin, and G. Guiochon, 

, 3365–3384 (2009). doi.org/10.1021/ac802632x.	

		S. Fekete, J.-L. Veuthey, D.V. McCalley, and D. Guillarme, 

, 127–138 (2012). doi.org/10.1016/j.chroma.2012.10.056.	

		J. Halvorson, A.M. Lenhoff, M. Dittmann, and D.R. Stoll, 

, 185–194 (2016). doi.org/10.1016/j.chroma.2016.12.084.	

, 984–987 (2016). doi.org/10.1021/acs.jchemed.5b00986.	

, 271–289 (1956). doi.org/10.1016/0009-2509(56)80003-1.

, 8593–8602 (2015). doi.org/10.1021/ac504473p.

 is the editor of “LC Troubleshooting”. Stoll is a professor and 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 60 peerâ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: 

LC Troubleshooting

Effects of Flow Rate on UV Detection in Liquid Chromatography

	Chromatographic peak integration continues to be a major regulatory issue and was first discussed in this column in 2015. Is the approach to manual intervention and manual integration outlined still acceptable in the light of regulatory citations and guidance documents published since then?

	Peak integration is at the heart of chromatographic analysis. Understanding how data are acquired by a chromatography data system (CDS), how peaks are integrated, and how each integration parameter operates is essential to effective analysis. The key reference on chromatographic integration was written over 20 years ago by Normal Dyson (1). Chromatographic integration in regulated laboratories was discussed in this column in 2015 (2) where a structured approach to first manual intervention (change of peak windows and processing parameters with automatic baseline placement) and then manual integration (manual repositioning baselines) was presented and discussed. The aim was to take a scientifically sound approach to integration in a regulated laboratory. Manual intervention can be applied to any analysis, but should be applied consistently to all injections in a sequence because chromatography is a comparative analytical technique. Newton and McDowall also discussed peak integration in the third part of a six-part series on data integrity, which included an outline of the order of processing of files in a sequence and the contents of an integration standard operating procedure (SOP) (3). Manual integration needs to be carefully controlled and managed because this is an area where an inspector will focus. Before we look in detail at integration, let us step back and look at the applicable regulations.

Laboratory Controls and Record Requirements

	It is important to understand the regulatory requirements for laboratory controls and records as these provide a major input to any discussion. The US good manufacturing practice (GMP) regulations for laboratory controls include this requirement (4):

	21 CFR 211.160(b) Laboratory controls shall include the establishment of scientifically sound and appropriate specifications, standards, sampling plans, and test procedures designed to assure that components, drug product containers, closures, in-process materials, labeling, and drug products conform to appropriate standards of identity, strength, quality, and purity.

	This is a relatively simple regulation to understand: everything done in the laboratory including peak integration must be scientifically sound. A similar requirement for scientific soundness in the laboratory is found in section 11.12 of EU GMP Part 2 for active pharmaceutical ingredients (5).

	As chromatography is a comparative analytical technique all injections should be integrated in the same way as much as possible. This generalization is tested near the limits of quantification and where complex mixtures are separated. Next, we have:

	21 CFR 211.194(a) Laboratory records shall include complete data derived from all tests necessary to assure compliance with established specifications and standards, including examinations and assays,….. (4).

	The requirement for laboratory records is also simple: complete data has been discussed in this column earlier (6) and other articles (7,8). The EU GMP regulations are not as simple to interpret; chapter 4 refers to raw data that is not defined (9). A discussion of the meaning of raw data has been presented earlier and is equivalent to complete data in the US GMP regulations (6).

	How are these regulations interpreted by inspectors? Here are some examples involving scientifically unsound practices and uncontrolled integration from the FDA:

	Integration of chromatograms for method STM-0076 <redacted> has been performed inconsistently. The <redacted> chromatograms exhibit tailing. Prior to March 2017 <redacted> was generally integrated as an impurity. In March 2017 a recommendation was made in STM- 0076 to <redacted> at this retention time.
	The recommendation is often but not always followed and results in this area being integrated <redacted>. There is a lack of scientific justification to support if the tailing portion should be integrated. The change ... was implemented without fully evaluating the impact on previously processed data (10).

	Your test methods were not capable of demonstrating the purity of your drugs... analysts reprocessed data up to 12 times, and only included the final result in the report for review by Quality Assurance. Your Deputy Manager, Quality Control stated that it is common practice to “play with parameters” to get the proper integration (11).

	Failure to ensure that all test procedures are scientifically sound and appropriate to ensure that your API conform to established standards of quality and purity.
	You failed to establish adequate test procedures. For example, your analyst manually integrated a HPLC test for <redacted> API despite the fact that the chromatogram lacked peak resolution. …... You lacked an approved protocol for manual integration or quality oversight of the practice (12).

	Method QC/STP/I2252-04 was not followed in analysis if <redacted> tablets for <redacted> by gas chromatography. The method requires the standards be prepared with <redacted>. During the preparation of standards for sequences QC863VEN1606A, the standards were prepared with both <redacted> and <redacted>, so the same standards could be used to evaluate <redacted> tablets for <redacted> and <redacted> tablets for <redacted>. The <redacted> and <redacted> peaks coelute, potentially reducing the accuracy of the standard area count compared to the approved method (13).

	In an earlier “Questions of Quality” column on integration (2) it was noted that there was no definition of manual integration. This column implicitly defined manual integration as manual placement of the baseline by a chromatographer. The Parenteral Drug Association’s (PDA’s) Technical Report No. 80 defines manual integration as a:

	Process used by a person to modify the integration of a peak area by modifying the baseline, splitting peaks or dropping a baseline as assigned by the chromatography software to overrule the pre-established integration parameters within the chromatographic software (14).

	Is this definition acceptable? It is wordy, repetitious, and could be better phrased. The use of the word overrule is contentious. As noted above, CDS software is not perfect and an application can struggle to separate overlapping peaks obvious to a trained eye. The biggest issue with this definition is that there is no mention of scientific soundness as defined in the FDA GMP regulations presented earlier (4).

	A simpler, more concise, and better definition of manual integration could be “manual repositioning of peak baselines with scientific justification for their positioning”.

	Implicit within this definition is the use of CDS software-otherwise you’d be drawing baselines on paper. However, this also requires that the chromatographer is trained, and ideally software technical controls should prevent manual repositioning of baselines where this is not justified by the type of the analysis. In addition, the audit trail should record the actions of the analyst optimizing the peak integration, and the system should provide a means to recall the original automated integrated chromatogram.

	From the citations above, would it be reasonable to ban manual integration in regulated laboratories? Let’s think this through. Experienced analysts know that chromatographic analysis can be affected by temperature, humidity, column history, as well as mobile phase preparation, so that one day’s analysis often varies slightly from the previous day’s run. To achieve consistent output and measurement, it is critical to adapt and optimize factors such as peak detection threshold or retention time windows to ensure consistent, correct, and accurate integration. But how can reviewers, approvers, quality, or outside auditors recognize the legitimate vs. egregious use of manual integration?

	Banning the use of manual integration is a common response to avoid questions about data integrity. However, there are three outcomes to this crude action:

		Laboratories will have to accept poor and inconsistent integration.

		Analysts will find a workaround that permits them to integrate each chromatogram with a different set of integration parameters (typically involves performing quantification in a laboratory information management system [LIMS], or worse, a spreadsheet, without traceability back to the integration methods).

		Analysts will be forced to spend hours of their day developing complex and manipulative methods to address variations between chromatograms with a single processing method. Typically, this will require many “integration events” that could even include placing peak starts and ends at specific time points; in effect, performing manual integration to satisfy the “no manual integration” rule and deceive the reviewer.

	In the wrong hands, with the wrong intent, and without a robust training and review process, altering chromatographic peak processing parameters has been misused by analysts to falsify results. How can this be managed?

	Since the earlier “Questions of Quality” column on integration (2) there have been many publications on data integrity from regulatory authorities such as Medicines and Healthcare products Regulatory Agency (MHRA), World Health Organization (WHO), European Medicines Agency (EMA), Pharmaceutical Inspection Coâoperation Scheme (PIC/S), and Food and Drug Administration (FDA) (15–20), as well as industry bodies such as GAMP, PDA, and European Compliance Academy (ECA) (14,21,22). There is also the recent publication of the draft of the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) M10 on bioanalytical method validation (23) that combines the current views outlined in earlier guidance documents issued by the EMA and FDA (24,25). However, only the PDA and ICH guidance documents have specific sections on chromatographic integration (14,23).

	The PDA’s Technical Report No. 80 has a large section on chromatographic integration, and the guidance document illustrates both acceptable and unacceptable peak integration practices (14). ICH M10 section 3.3.6 outlines the current thinking for integration of chromatograms in bioanalysis (23):

		Chromatogram integration and reintegration should be described in a study plan, protocol, or SOP.

		Any deviation from the procedures described a priori should be discussed in the Bioanalytical Report.

		The list of chromatograms that required reintegration, including any manual integrations, and the reasons for reintegration should be included in the Bioanalytical Report. Original and reintegrated chromatograms and initial and repeat integration results should be kept for future reference and submitted in the Bioanalytical Report for comparative BA/BE (bioanalytical or bioequivalence) studies.

	Gone is the burdensome FDA requirement for a manager to preapprove any manual integration (25), to be replaced by a plan or SOP for controlling the integration process with the before and after chromatograms included in the study report showing the impact of the changes of manual integration.

	We need to consider why we need to integrate peaks manually and the reason is that there are situations where a CDS cannot integrate peaks correctly. Some examples are:

		Rising, falling, or excessively noisy baselines

		Slowly eluting peaks (where the CDS has difficulty identifying the peak end).

	The reasons for the inability of the integration method may be due to:

		Poor method development and validation where the analytical procedure or the integration method is not optimized or robust

		Requirements for quantitation of very small peaks, especially new, unexpected peaks resulting from impurities or excipients

		Complex sample matrices resulting in interfering peaks that may still be present after sample preparation, for example, biological samples, contrast media

		Analysis of complex mixtures may result in a heavy manual integration workload as the CDS method is not able to integrate all peaks automatically, for example, contrast media, fermentation samples.

	There must be a scientifically sound (4) justification for manual intervention and manual integration as outlined in the earlier “Questions of Quality” column (2). Remember that manual integration slows an analytical process and is inefficient (3).

	The more that peak integration is automated, the faster it is, with the bonus of lower regulatory scrutiny. Otherwise the review could take longer than the actual analysis.

	This places responsibility on the laboratory to develop robust analytical chromatographic procedures with reliable separations that are fit for use. A key component of this approach is that the resultant peak integration must be consistent, not the use of consistent parameters or settings to achieve that peak integration. This is a subtle but vital difference that is not always appreciated.

	The regulatory world is changing for analytical procedures and this should have a positive impact on peak integration. There is an ICH final concept paper (26) for the development of ICH Q14 on analytical procedure development guidelines and a revision of ICH Q2(R1) on validation. The current problem with ICH Q2(R1) (27) is that it is focused on method validation with nothing mentioned about the most important phase of the life cycle: method development. Understanding how a method works and what are the key variables is often overlooked due to time pressures. Failure to plan is planning to fail.

	A life cycle approach has been addressed by a proposed USP general chapter <1220> on analytical procedure life cycle management (28). The USP life cycle begins by defining the analytical target profile (ATP), that is, the start of the process to define and validate a design space for each analytical procedure. It ensures that critical parameters are managed and controlled, and changes within the design space are known and predictable, for example, organic modifier changes of the mobile phase. Knowing the design space of an analytical procedure should result in better peak shape and resolution and hence accurate automated peak integration. Outcomes of this analytical procedure life cycle management (APLM) or method life cycle management (MLCM) approach should be more robust methods, reproducible chromatography, scientifically consistent peak integration, and, hopefully, reduced out of specification (OOS) results.

Is the FDA Banning the Use of Inhibit Integration Events?

	The use of the inhibit integration function is a hot regulatory topic now, as can be seen from this FDA regulatory citation:

	1. Failure to ensure that test procedures are scientifically sound …. Our investigators observed that the software you use to conduct high performance liquid chromatography (HPLC) analyses of API for unknown impurities is configured to permit extensive use of the “inhibit integration” function without scientific justification.

	For example, our investigator reviewed the integration parameters you used for HPLC identification of impurities in release testing for <redacted>. These parameters demonstrated that your software was set to inhibit peak integration at four different time periods throughout the analysis.

	Inhibiting integration at various points during release testing for commercial batches is not scientifically justified. It can mask identification and quantitation of impurities in your API, which may result in releasing API that do not conform to specifications (29).

	This citation is based on 21 CFR 211.160(b) (4), which was presented earlier in this column, and the key questions to ask are if, where, and when can integrate inhibit be used? It has been said in some audits and inspections that this function cannot be used. This is an untenable situation and there is no explicit or implicit statement in the GXP regulations for this attitude. However, it comes back to scientific soundness and a laboratory must be able to justify the use of the function. Let us consider the following scenario:

		There is a baseline perturbation with a large negative peak after an injection. A peak of interest elutes shortly after the negative peak. The use of integrate inhibit is fully justified from the start of the injection until the baseline has returned to normal and before elution of the peak of interest. Otherwise there is a large probability that baseline placement of the analyte could be adversely influenced by the negative peak.

		Similar scenarios occur when extraneous peaks are washed from the column, or baseline perturbations from mobile phase changes during a gradient elution or wash at the end of a chromatogram.

	What is more problematic is the use of integrate inhibit in the middle of a run as cited above (29). If system, blank, or other nonâsample peaks occur in the middle of a chromatogram, traditionally those peaks were not integrated. Because of suspicions that the excluded peaks might be real impurities, excluding these system peaks needs to be carefully documented and justified in the method development and validation reports, otherwise they should be integrated and marked clearly as system peaks.

	Trial injections using actual samples feature in many warning letters (30) and question 13 of the FDA data integrity guidance (20) states that:

	FDA prohibits sampling and testing with the goal of achieving a specific result or to overcome an unacceptable result (e.g., testing different samples until the desired passing result is obtained).

	This is correct and should never be acceptable in a GXP laboratory SOP.

	However, consider the following situation: you are analyzing low volume samples from a nonclinical study. There is a total volume of 20 μL plasma sample that is extracted and there is only enough for a single injection from each sample. Ask yourself the question; are you going to commit an analytical run of samples without checking that the system is ready? The cost of repeating the study is a high six figure sum if a run does not work. Therefore, from a practical perspective, we need a way of checking that a system is ready for analysis, but one that does not involve testing into compliance with samples. Ah, somebody says use system suitability test (SST) injections. The problem is that you may need several replicates to determine if the system is ready, and SST injections should never be started until you are confident that the system is equilibrated.

	We propose the following approach for system evaluation, readiness injections, or equilibration checks:

Questions of Quality

Chromatographic peak integration continues to be a major regulatory issue and was first discussed in this column in 2015. Is the approach to manual intervention and manual integration outlined still acceptable in the light of regulatory citations and guidance documents published since then?

Can We Continue to Draw the Line?

The 44th International Symposium On Capillary Chromatography (ISCC) and 17th GC×GC Symposium will be held at the Palazzo dei Congressi, in Riva del Garda, Italy, from 24–29 May 2020.

	Over the years, the ISCC conference has established its reputation as a forum for microcolumn separation techniques. Since the first meeting in Hindelang in 1975, the most important developments in capillary gas chromatography (GC), microcolumn liquid chromatography (LC), and electromigration techniques have been presented in this symposium series. The format and the atmosphere of the 44th meeting will be similar to the previous meetings, with a particular emphasis on mass spectrometry (MS) this year.

	Past meetings have been held in Hindelang, Riva del Garda, Monterey, Baltimore, Gifu, Kobe, Wintergreen, Park City, Las Vegas, Dalian, Albuquerque, Portland, and San Diego. This year the “Palazzo dei Congressi” in Riva del Garda, Italy, will accommodate the 44th meeting. The six-day event will feature recent findings from leading academic and industrial experts in the form of lectures and posters. Apart from the most recent advances in the fields of pressure and electrodriven microcolumn separation techniques and comprehensive two-dimensional (2D)-GC, this year will again have particular emphasis on comprehensive separation technologies combined with capillary chromatography and 2D-GC with various forms of mass spectrometry, from unit-mass to high resolution, and from single- to hybrid analyzers.

	The conference also offers sessions on capillary GC, microcolumn LC, electromigration methods, and microfabricated analytical systems, which are expected to cover lab-on-a-chip, column technology, coupled and multidimensional techniques, comprehensive techniques, hyphenated techniques, sampling and sample preparation, trace analysis, and automation. Application sessions include environmental; energy, petrochemical, industrial; biomedical, pharmaceutical; and the analysis of natural products, food, flavours, and fragrances. Workshop seminars from instrument manufacturers and an extensive exhibition of instrumentation, accessories, and supplies will run in parallel to the scientific programme.

	At the meeting, the 2020 Marcel Golay Award, sponsored by PerkinElmer, will be presented in recognition of outstanding contributions in the field of separation science. The Leslie Ettre Award, sponsored by PerkinElmer, will be presented to a young scientist for research on capillary GC applied to environmental or food analyses.

	The Giorgio Nota Award, sponsored by Waters, will be presented to a scientist in recognition of a lifetime of achievement in capillary LC. The John Phillips Award, sponsored by LECO and Restek, will be awarded to individuals who have made outstanding contributions to the field of GC×GC analysis. The GC×GC Lifetime Achievement Award, sponsored by LECO and Restek, honours an experienced GC×GC scientist who has made significant contributions to the field. Chromaleont, ISCC and GC×GC, Fort Worth Texas 2021, the Division of Analytical Chemistry of the Italian Chemical Society (SCI) and the Interdivisional Group

	of Separation Science of the Italian Chemical Society, Elsevier, Secyta, and Separations (MDPI) will promote scholarships for young researchers. For more information, please visit: 

Events

44th International Symposium On Capillary Chromatography and the 17th GC×GC Symposium

	LCGC Europe spoke to Leon Barron and Matteo Gallidabino to discuss novel nontargeted approaches to analyze explosive materials using ion chromatography (IC) with high resolution mass spectrometry (HRMS), and the challenges and solutions analysts can encounter when developing nontargeted methods.

Q. Nontargeted analysis (NTA) is currently gaining wider acceptance. What is NTA and where is it being used in your area of research?

Most analyses are targeted in nature. That is, a number of specific compounds are selected before the analysis occurs. Nontargeted analysis (NTA) refers to applications where no specific analytes are shortlisted beforehand and the instrument captures everything it can detect, so that the data can be reviewed in a flexible way later. There are several ways to perform NTA including: (a) using all of the data generated by the instrument to classify or differentiate samples as a whole from each other by, for example, principal component analysis (PCA); (b) identifying specific “features” in the data that change significantly following exposure to a toxic substance, for example, and, whilst still not necessarily knowing its identity. As an extension of NTA, suspect screening is the identification of new compounds in the sample by matching measured data to one or more databases or by manual search using theoretical ion accurate m/z . For liquid chromatography (LC)- and gas chromatography (GC)-based techniques, NTA has been most useful when coupled to high resolution mass spectrometry (HRMS) instruments, which comprise of either time-of-flight (TOF) or orbital ion trap-based mass analyzers. The use of HRMS helps immensely to resolve significantly larger numbers of features and arguably represents the best means to most rapidly identify these afterwards too. In my area of environmental and forensic chemistry, NTA is becoming more commonly used. For example, environmental metabolomics is now emerging more to identify any endogenous metabolite features that change in aquatic species as a result of exposure to contamination or specific environmental conditions in rivers (1–4). We have also extensively used suspect screening to identify new organic contaminants, including pharmaceuticals, illicit drugs, explosives, their metabolites, precursors, and transformation products in complex samples such as wastewater and river water to monitor communityâscale activities (5–7). In forensic science, the ability to retrospectively mine such large datasets is very useful to go back and assess whether an analyte might have been present (8).

Q. Have there been any major technological breakthroughs nontargeted analysis?

Arguably the most significant technological breakthrough that has pushed NTA forwards is the increased commercial availability of HRMS instruments that can be coupled to separation techniques such as LC or GC, for example. Mass accuracies of <1 ppm are now readily achievable with resolutions up to 140,000 fullâwidth at half peak maximum (FWHM), providing elemental composition-level information in many cases. Similarly, the ability to perform different modes of data independent analysis (DIA) offers extra flexibility for NTA including “all ion” fragmentation and sequential window acquisition of all theoretical mass spectra analysis (SWATH), for example. Along with the ability to perform traditional targeted analysis too (data-dependent analysis [DDA]), it has become possible to perform targeted, NTA, and suspect screening using the same instrument and, in certain cases, simultaneously. As a result, published targeted methods are generally growing with respect to the number of analytes they include as new compounds are discovered or added continually (9,10). This has presented analysts with a new challenge scale and treatment of data. Given the amount of data these instruments acquire, data analysis is now the bottleneck. It can take significantly longer to review and interpret the results generated for a single sample than it takes to run it in the laboratory! Whilst excellent processing tools and databases exist for MS data, there is less focus on separations data in my opinion and machine learning has recently proved useful here (10,11). I see massive potential in the use of machine learning generally moving forwards, not only for suspect screening but also for NTA, for example, for prediction of changes in ‘omics datasets or linking these to effects following toxicant exposure.

Q. You recently developed a method using both targeted and nontargeted gradient ion chromatography (IC) with HRMS to profile black powder substitutes and gunpowder residues. What were the aims of this research?

The aim of this work was to develop and validate a new gradient IC-HRMS method that would be broadly applicable to quantitative determination of trace concentrations of lowâmolecular-weight inorganic and organic anions, but primarily that would be suitable for forensic casework in energetic materials analysis, including ammunition and explosives (12). We also wanted to exploit HRMS to potentially offer us more information about the sample. First, we focused on identifying a black powder substitute in fingermarks and sweat deposits from a donor using IC-HRMS. Using a mixture of targeted analysis and NTA with PCA, we investigated the time since the materials were handled, which was very exciting! Following this, we identified features in the data that drove any temporal trends to potentially serve as a new way to include or exclude similar residues of such materials found at a crime scene that were relevant to the case. We also aimed at gunshot residue to see whether we could use NTA to classify by the original ammunition used.

 Samples submitted to forensic analysis are usually characterized by a higher number of species than those usually targeted by traditional methods. This means that they also potentially contain complementary information, which may be helpful to track back the trace origin and deposition mechanism. Hence, a complementary aim of this project was to assess if enhanced intelligence could actually be extracted from forensic samples through the judicious combination of NTA with modern data analytics.

Q. What were the main obstacles you encountered in this project and how did you overcome them?

 I have been working with IC-MS since around 2001 and coupling the two techniques has often been cumbersome in comparison to LC–MS in my experience (13). Thankfully, modern IC-MS systems are now available with integrated instrument control and data analysis. However, in this particular application there were two main challenges because IC generally operates using purely aqueous eluents. First, IC eluates are not very volatile and auxiliary pumps are often needed to deliver organic solvent into the eluate to aid gas phase transfer in electrospray ionization (ESI), especially for trace analysis. Second, extraction of complex samples typically encountered in forensic casework normally use organic solvents not always compatible with IC separations. Therefore, we have been trying to find ways to circumvent these two issues by using organic solvent in the eluent itself, which presents its own challenges (14). So far, we have tried a number of additives, including methanol, acetonitrile, and ethanol, which have each removed the need for auxiliary pumps making the coupling process much simpler, and keeps the system cleaner at the same time. The trade-off is IC selectivity, which changes markedly and, in some cases, unpredictably. Furthermore, some organic solvents transform under alkaline conditions, which leads to interference; for example, acetonitrile can hydrolyze in hydroxideâbased eluents to yield acetate ions in the background signals. In this work, we aligned the IC eluent with that of the sample extraction solvent directly, that is, 50:50 (v/v) ethanol–water. We also improved the selectivity over previous IC methods by introducing a gradient separation using carbonate–bicarbonate as an eluent and this worked very well.

	As we do not usually know the identity of an energetic material in forensic science, it is necessary to perform both organic and inorganic screening. This extract would therefore normally be analyzed directly by an LC–HRMS method for a large suite of organic high-order explosives, precursors, and transformation products and then, following solvent exchange to remove ethanol, by IC-MS. By developing the separation in ethanolic eluents, our aim was that the sample extract could be analyzed directly, which could increase throughput and robustness.

Challenges to overcome during this project were numerous, and not just limited to the decision of which strategy to adopt. A priority for the IC-HRMS method was that it could easily be aligned with current forensic practices. As 50:50 (v/v) ethanol–water is often used to extract explosive samples in casework, we decided to adopt this as the eluent. Preliminary tests on ESI performance supported the choice because they showed that this mixture led to the same or better signals than conventional eluents used in liquid-based chromatographic techniques. That was promising, but the viscosity of the mixture obviously also brought some challenges with the column back pressure that had to be addressed by increasing the column temperature. Also, ESI using 50:50 (v/v) ethanol–water has been rarely investigated before, so the best conditions were essentially unknown. Therefore, the implementation of the method basically became a problem of fine-tuning all the parameters involved in the chromatographic, ionization, and detection steps! We eventually used a statistical-based, design-of-experiments (DOE) approach to deconvolute this complexity and properly investigate analyte separation and responses to find optimal conditions. The use of predictive modelling methods was therefore not just limited to the evaluation of the new approach in an operational context, but also to its optimization. Thanks to this, in any case, we were able to achieve excellent analytical performance.

Q. What is novel about your approach and what benefits does it offer over previous techniques?

The use of IC-MS is not new, but its use in forensic science is really only emerging now. The benefits of this particular method were that a larger number of anions could be detected (n = 19) than previously possible as a result of the optimization of gradient conditions. The lower limit of detection lies in the low µg/L range, making it suitable for direct trace analysis across a range of applications if needed. One of the main benefits this method offered is obviously its direct integration into standard workflows, making the analytical process far simpler and practical for the analysis of organic–solventâbased extracts. As well as keeping the system clean, it also enables elution of hydrophobic/non-charged species, which would otherwise be fully retained by IC, thereby widening its scope.

	While several applications have been reported in environmental science using IC-MS, for example (15), the use of ICâHRMS is quite rare. This technique offers obvious advantages in forensic science, especially for NTA and suspect screening. Here, our approach allowed us to show how both the targeted anions and the NTA profile of the rest of the contaminated sweat sample changed over a period of hours following contact with the black powder substitute, even after washing their hands! Similarly, for a range of different gunshot residues collected after firing a gun, we were able to link these with three original ammunition brands that were used. Using NTA, we were able to tentatively identify several new potential compounds afterwards that could provide additional linkages between different evidence types together. The ability to provide some degree of source apportionment is a major advantage of any technique in forensic science and we established this proof of principle here.

	MG: This novel approach is quite revolutionary and has all the characteristics to have a large impact on forensic practice. We proved that the combination of NTA-based techniques and advanced data analysis could actually provide enhanced intelligence for use in crime investigation. Not only does the method allow the main components in the submitted samples to be rapidly identified, but it can also extract additional information such as the time since handling and potential origin of the analyzed traces. Some supplementary work is still needed to truly implement these possibilities in actual casework, but our method is an effective step forwards. In this regard, it has the potential to unlock a range of new possibilities in forensic profiling, and also to further highlight the value of forensic science in crime investigation.

Q. Are you planning to develop this research further?

Yes, we have just started a new project that will combine LC and IC-HRMS analysis together to identify precursors and indicative reactant species related to threat agent manufacture, including explosives and drugs, for example. In 2017, we were the first laboratory to identify residues of highâorder explosives in municipal sewage using LC–HRMS, having performed drug-based wastewater epidemiology for many years (7,16). This project seems like an obvious way to integrate both techniques for application to a very complex matrix, such as wastewater, using NTA. Lastly, confirmatory analysis even of simple anions and cations is very much needed in other areas of forensic science, for example, in support of “acid attack” investigations. I also plan to extend my previous research into the analysis of disinfectants and their by-products. We have already made some progress recently using IC-HRMS for drinking water (17), but it will be good to extend our knowledge on the breadth of toxic species formed following disinfection processes in several other areas too.

	MG: We previously showed how machine learning could help to associate different gunshot traces found at the scene of a gun crime (18). The approach worked well, but it could be further improved if coupled with NTA data and, thus, integrated in an ‘omics workflow. We are working on that and going to test this hypothesis, and also for the analysis of arson accelerants. The final objective is to develop a transversal profiling approach that can be applied across different fields, and better support the criminal judicial system in the decision-making.

Q. Do you have any practical advice for chromatographers who are embarking on developing a nontargeted analysis method?

 First, try to make your NTA method as generic as possible so that it can capture a wide chemical space. This may mean developing a longer, shallower gradient to separate as many features as possible. Also, be aware that a single NTA method will not cover everything. You may need to identify species that might fall outside the scope and whether you need multiple separation modes to cover what you need, for example. This is exactly why we put LC and IC-HRMS together for NTA of explosives-related evidence. Lastly, when setting up your sequences, make sure to randomize your samples, controls, and quality controls (QCs). NTA can often produce deceptively nice trends or classifications, but it is important that groupings or observed changes in the data are actually real, and not just a product of instrument performance drift. You may also want to think about your mass analyzer because you may not always get very high mass accuracy, resolution, and data acquisition speed all in one HRMS instrument.

	In my opinion, visualizing, manipulating, and interpreting the data are actually the hardest parts, rather than the laboratory science. Vendor-licensed and open-source software is available and helps with data normalization and chromatogram alignment, and a range of online databases support new compound identification if needed later on. Also, to give you added flexibility, try to learn a coding language such as R or Python. In many cases, freely available codes have already been written for complex tasks and these can be a very useful resource, not only for NTA but also other areas such as PCA and machine learning. It is not as hard as you think!

et al., in Comprehensive Analytical Chemistry

  is a senior lecturer in forensic science at King’s College London, UK. He received his  Ph.D. in analytical chemistry from Dublin City University, Ireland,  in 2005. His expertise lies in analytical chemistry, particularly in separation science, mass spectrometry, and machine  learning for targeted and nontargeted applications in environmental, forensic, and biological systems analysis.
	
	

 is currently a senior lecturer in forensic science at Northumbria University in Newcastle, UK. He has a comprehensive background in criminalistics that he obtained at the School of Criminal Justice of the University of Lausanne, Switzerland. His work focuses on next-generation analytical techniques, successfully combining separation methods, mass spectrometry, and advanced data analytics to provide enhanced information in a forensic context and better support the court in the decision-making process.

Features

LCGC Europe spoke to Leon Barron and Matteo Gallidabino to discuss novel nontargeted approaches to analyze explosive materials using ion chromatography (IC) with high resolution mass spectrometry (HRMS), and the challenges and solutions analysts can encounter when developing nontargeted methods.

LCGC Europe-12-01-2019

Vol 32 No 12 LCGC Europe December 2019 Regular Issue PDF

Effects of Flow Rate on UV Detection in Liquid Chromatography

Can We Continue to Draw the Line?

44th International Symposium On Capillary Chromatography and the 17th GC×GC Symposium

Exploring Energetic Materials Using Nontargeted Analysis