MULTIDIMENSIONAL APPROACHES FOR THE ANALYSIS OF COMPLEX SAMPLES USING HPLC

Transcription

1 MULTIDIMENSIONAL APPROACHES FOR THE ANALYSIS OF COMPLEX SAMPLES USING HPLC By: Sercan PRAVADALI-CEKIC BSc. (Advanced Science), BSc. (Honours) A thesis submitted in accord with the requisites of the degree of Doctor of Philosophy School of Science and Health University of Western Sydney Parramatta, New South Wales, Australia May 2014 S. Pravadali-Cekic

2 Statement of Authentication I, Sercan Pravadali-Cekic, declare that the work presented in this thesis is, to the best of my knowledge and belief, original except as referenced in the text of this thesis. I hereby declare that I have not submitted this material, either in full or in part, for a degree at this or any other institution. Signed: Date:

3 Acknowledgements I would like to thank my supervisors Prof. Andrew Shalliker and Dr. Xavier Conlan for their help, support, guidance and patience throughout my candidature; in particular a thank you to Prof. Andrew Shalliker as my primary supervisor for the expert guidance, assistance and advice, who also always encouraged me to do my best. I would like acknowledge The University of Western Sydney Postgraduate Research Award for providing me the opportunity and necessary financial assistance. I would like to acknowledge Dr. Mariam Mnatsakanyan for her previous work in the analysis of coffee samples for antioxidants via 2DHPLC relating to the project and Dr. Paul G. Stevenson for his previous work in home-built algorithms for peak detection, recognition and illustration. Thank you to Danijela Kocic for her expertise in Mass Spectrometry and Michelle Camenzuli, both for their help, support and most of all their friendship for the duration of my PhD. Finally, as an expression of gratitude, thank you to my family and husband, who have always supported and encouraged me throughout my candidature. I am indebted to them for their love and care.

4 Table of Contents Publications arising from this thesis... vii Publications to be published from this thesis... vii Conference Presentations... viii List of Tables... ix List of Figures... x List of Abbreviations and Symbols... xiii Abstract... xvii Preface... xviii CHAPTER 1 Introduction Uni-Dimensional HPLC Limitations of Uni-Dimensional HPLC Multidimensional HPLC Multiple Separation Steps - Two-Dimensional HPLC... 5 Orthogonality... 6 Geometric Approach to Factor Analysis (GAFA)... 6 Modes of Two-Dimensional HPLC Two-Dimensional System Selectivity Stationary Phase Mobile Phase Multi-Detection Column Technology Active Flow Technology (AFT) Parallel Segmented Flow (PSF) Curtain Flow (CF) i

5 1.4 Complex Samples of Natural Origin Sample Dimensionality Research Problems Project Aim Project Objectives CHAPTER 2 Comprehensive Sample Analysis using High Performance Liquid Chromatography with Multi-Detection Introduction Experimental Chemicals and Reagents DPPH Acidic Potassium Permanganate Manganese(IV) Tris(2,2 -bipyridine)ruthenium(iii) ([Ru(bipy) 3 ] 3+ ) Sample Preparation Instrumentation and Chromatographic Conditions Chromatographic Analysis Chemiluminescence Detection DPPH Detection UV-Vis Absorbance Detection Mass Spectrometry Results and Discussion Conclusion ii

6 CHAPTER 3 Multiplexed Detection: Fast Comprehensive Sample Analysis using HPLC with AFT Columns Introduction Experimental Chemicals and Reagents Sample Preparation Instrumentation and Chromatographic Conditions Instrumentation Chromatographic Analysis UV-Vis Detection DPPH Detection Mass Spectrometry Data Analysis Results and Discussion Conventional HPLC Multi-Detection Multiplexed HPLC Tobacco Leaf Extract Coffee Extract Conclusion CHAPTER 4 Selectivity in 2DHPLC: A Comparison Between Columns from Different Manufacturers Introduction Experimental Chemicals iii

7 4.2.2 Sample Preparation Instrumentation and Chromatographic Conditions Instrumentation Chromatographic Columns and Separation Data Analysis Results and Discussion Qualitative Assessment of the Selectivity Changes Overall System Comparison Cyano Phase Pentafluorophenyl Phase Conclusion CHAPTER 5 Optimisation of a 2DHPLC System for Enhancing the Separation and Subsequent Analysis of Tobacco Introduction Experimental Chemicals Sample Preparation Instrumentation and Chromatographic Conditions Instrumentation Chromatographic Columns Chromatographic Separation Data Analysis Results and Discussion Qualitative Assessment of the Selectivity Changes Overall System Comparison iv

8 Cyano Phase in First Dimension PESC18 Phase in First Dimension Pentafluorophenyl Phase in First Dimension Summary: Quadrant System Analysis Conclusion CHAPTER 6 A Preliminary Investigation into Two-Dimensional HPLC Incorporating Active Flow Technology (AFT) Columns PART A: 2DHPLC - CONVENTIONAL, PARALLEL SEGMENTED FLOW AND CURTAIN FLOW CHROMATOGRAPHIES 6.1 Introduction Experimental Chemicals Sample Preparation Instrumentation and Chromatographic Conditions Instrumentation Chromatographic Columns Chromatographic Separation Data Analysis Results and Discussion Uni-Dimensional HPLC Conventional versus PSF Columns Conventional versus CF Columns Two-Dimensional HPLC Conventional Columns PSF Columns v

9 CF Columns Relative Performance Summary PART B: AN EVALUATION OF THE 2DHPLC SYSTEM 6.4 Experimental Chemicals and Samples Chromatographic Columns Instrumentation and Chromatographic Conditions Data Analysis Results and Discussion Conclusion CHAPTER 7 General Conclusions and Future Directions Conclusions Future Directions References vi

10 Publications arising from this thesis 1. Pravadali, S., Bassanese, D.N., Conlan, X.A., Francis, P.S., Smith, Z.M., Terry, J.M., and Shalliker R.A., Comprehensive sample analysis using high performance liquid chromatography with multi-detection. Analytica Chimica Acta, (0): p Publications to be published from this thesis 1. Pravadali-Cekic, S., Kocic, D., and Shalliker, R.A., Multiplexed detection: Fast comprehensive sample analysis using HPLC with AFT columns Pravadali-Cekic, S., Stevenson, P.G., and Shalliker, R.A., Selectivity in 2DHPLC: A comparative study between columns obtained from different manufacturers Pravadali-Cekic, S., Stevenson, P.G., and Shalliker, R.A., Optimisation of a 2DHPLC system for enhancing the separation and subsequent analysis of tobacco Pravadali-Cekic, S., and Shalliker, R.A., The effects of HPLC instrumentation technology on the performance of curtain flow columns vii

11 Conference Presentations Nov th International Symposium on High Performance Liquid Phase Separations and Related Techniques Conference 2013 The Hotel Grand Chancellor, Hobart, Tasmania, Australia A Preliminary Investigation into Two-Dimensional HPLC coupled with Active Flow Technology (AFT) Columns Oral presentation Dec th Annual RACI Research and Development Topics 2012 Deakin University, Waurn Ponds, Victoria, Australia Selectivity in Detection: The Analysis of Alkaloids and Phenols using HPLC with Mass Spectral, UV, and Chemiluminescence Detection Poster presentation June th International Symposium on High Performance Liquid Phase Separations and Related Techniques 2012 Anaheim Marriott, Anaheim, California, United States of America Selectivity in Detection: The analysis of Alkaloids and Phenols using HPLC with Mass Spectral, UV, and Chemiluminescence Detection Poster presentation Parallel Segmented Flow Chromatography Columns for Efficient Two-Dimensional HPLC Poster presentation Dec th Annual RACI Research and Development Topics 2011 La Trobe University, Melbourne, Victoria, Australia Analysis of Complex Samples using Parallel Segmented Flow Two-Dimensional HPLC Oral presentation Dec th Annual RACI Research and Development Topics 2010 University of Tasmania, Hobart, Tasmania, Australia Defining the Reliability of Chemical Signatures for Complex Samples Derived from Analysis using Two-Dimensional HPLC Poster presentation viii

12 List of Tables Table 2.1 Table 3.1 Table 4.1 Table 5.1 Table 5.2 Table 6.1 Table 6.2 Table 6.3 Table 6.4 Table 6.5 Table 6.6 Identified known compounds via MS detection in tobacco leaf extract and Peak Height Ratios for each mode of detection. Detected DPPH peaks response to UV-Vis and MS. GAFA calculations for the 2DHPLC separation and in each of the quadrants. Two-dimensional HPLC selectivity systems. GAFA calculations for the 2DHPLC separation of tobacco leaf extract and in each of the quadrants. 2D System Parameters. 2D System Modes. 1DHPLC Sensitivity and efficiency measures for conventional, PSF and CF column formats. 2DHPLC - Sensitivity and efficiency measures of second dimension separation of a cut (maximum peak height) for conventional, PSF and CF 2DHPLC systems. 2DHPLC - Maximum sensitivity and efficiency measures obtained of second dimension separations for conventional, PSF and CF 2DHPLC systems. LC instrument comparison - Column format efficiency and sensitivity. ix

13 List of Figures Figure 1.1 Figure 1.2 Figure 1.3 Figure 1.4 Figure 1.5 Figure 1.6 Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 2.5 Figure 3.1 Figure 3.2 Illustration of a non-orthogonal two-dimensional separation space, where β represents the spreading angle and practical peak capacity is represented by the grid. Column band profile in parabolic distribution. Illustration of Active Flow Technology column design. AFT column - PSF column format. Column band profile; AFT takes the central region from the parabolic shaped band to detector and wall region to waste. AFT column - CF column format. UV-Vis absorbance detector chromatogram. Chemiluminescence detector - Acidic potassium permanganate chromatogram. Chemiluminescence detector Manganese (IV) chromatogram. Chemiluminescence detector Tris(2,2 -bipyridine)ruthenium(iii) chromatogram. DPPH detector chromatogram. Illustration of conventional HPLC setup with: (a) MS detector, (b) UV-Vis detector and (c) DPPH detector. Illustration of multiplexed HPLC setup with DPPH, UV-Vis and MS detectors. Figure 3.3 Conventional HPLC multi-detection chromatograms (UV-Vis, DPPH, and MS TIC): (a) Tobacco and (b) Coffee. Figure 3.4 Figure 3.5 Figure 3.6 Figure 3.7 Tobacco sample multiplexed chromatograms (a) UV-Vis 280 nm, (b) DPPH 520 nm, (c) MS (TIC). Coffee sample multiplex chromatograms (a) UV-Vis 280 nm, (b) DPPH 520 nm, (c) MS (TIC). Decaffeinated coffee sample multiplexed chromatograms (a) UV-Vis 280 nm, (b) DPPH 520 nm, (c) MS (TIC). Caffeine compound UV response: (a) Coffee (b) Decaffeinated coffee; Caffeine compound DPPH response (c) Coffee (d) Decaffeinated coffee (e) Caffeine standard. Figure 4.1 Uni-dimensional separations of Ristretto café espresso on ThermoFisher Scientific columns: (a) CN and (b) PFP; Unidimensional separations of Ristretto café espresso on Phenomenex columns: (c) CN and (d) PFP. All chromatographic conditions are 20 %min -1 gradient, from water to methanol, then held for 4 minutes at 100% methanol and returned to initial conditions in 1 minute (total run time 10 minutes). ( indicates caffeine peak) x

14 Figure 4.2 Uni-dimensional separations of Ristretto café espresso on Phenomenex columns at a 10 %min -1 gradient, from water to methanol, held for 8 minutes at 100% methanol and returned to initial conditions in 2 minute (total run time 20 minutes): (a) CN and (b) PFP. Figures from Mnatsakanyan et al. ( indicates caffeine peak) Figure 4.3 Two-dimensional separations of Ristretto café espresso. (a) ThermoFisher Scientific - 1DCN/2DC18, (b) ThermoFisher Scientific - 1DPFP/2DC18, in both dimensions mobile phase ran at a 20 %min -1 gradient, from aqueous to organic solvent at 2 mlmin -1. ( indicates caffeine peak) Figure 4.4 Two-dimensional separations of Ristretto café espresso by Mnatsakanyan et al. (a) Phenomenex - 1DCN/2DC18, (b) Phenomenex - 1DPFP/2DC18, in both dimensions mobile phase ran at a 10 %min -1 gradient, from aqueous to organic solvent at 1 mlmin -1. ( indicates caffeine peak) Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9 Figure 4.10 Figure 5.1 Figure 5.2 Figure 5.3 Figure 5.4 Scatter plots for the 2DHPLC separations with ThermoFisher Scientific columns (a) CN, (b) PFP; Phenomenex columns by Mnatsakanyan et al. (c) CN, (d) PFP as the first dimension. The quadrants are defined by the dashed lines. Heart-cut segments from CN phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. Heart-cut segment from CN phase on C18 separation of Ristretto café espresso: (a) ThermoFisher Scientific columns Cut time between minutes and (b) Phenomenex columns Cut time between minutes (image from M. Mnatsakanyan et al.). ( indicates caffeine peak) Heart-cut segment from CN phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. Heart-cut segment from PFP phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. Heart-cut segment from PFP phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. Molecular structure of a silica bonded with (a) C18, (b) CN and (c) PFP. Surface comparison of (a) C18 and (b) Hypercarb column. Uni-dimensional separations of White Ox Tobacco extract on (a) CN, Hypercarb, PESC18 and PFP at a 20 %min -1 gradient, from water to methanol, (b) CN, Hypercarb, PESC18 and PFP in gradient mobile phase at a 20 %min -1 gradient, from water to acetonitrile, then held for 4 minutes. Two-dimensional separations of White Ox Tobacco extract. xi

15 Figure 5.5 Figure 5.6 Figure 6.1 Figure 6.2 Figure 6.3 Figure 6.4 Figure 6.5 Figure 6.6 Figure 6.7 Figure 6.8 Figure 6.9 Figure 6.10 Figure 6.11 Figure 6.12 Figure 6.13 Chromatographic overlay of second dimension separation (C18) of all incremental heart cuts for System 1-1DCN/2DC18 (H 2 O:MeOH). Separation of a tobacco leaf extract heart-cut segment ( min) from the CN first dimension on a C18 column in the second dimension utilising System 1 (1DCN/2DC18 (H 2 0:MeOH)). Two-dimensional system setup with an autosampler, two pumps, two degassers, two system controllers, two UV-Vis detectors, a 10-port two position switch valve, and two columns in (a) Conventional mode, (b) PSF mode and (c) CF split-flow mode. Caffeine peak Conventional HPLC chromatograms overlayed with PSF chromatograms in uni-dimensional HPLC. Caffeine Peak Conventional HPLC chromatograms overlayed with CF chromatograms in uni-dimensional HPLC. Second dimension maximum caffeine peak height Conventional columns comparison. Theoretical plate count versus increasing injection volume. Second dimension maximum caffeine peak height PSF columns comparison. Second dimension maximum caffeine peak height CF columns comparison. HPLC autosampler with CF column in (a) split-flow system setup and (b) 2 pump system setup. UHPLC autosampler with CF column in (a) split-flow system setup and (b) 2 pump system setup. Switch valve manual injection with CF column in (a) split-flow system setup and (b) 2 pump system setup. Overlay of conventional and CF in split-flow and 2 pump column formats chromatograms of butylbenzene peak with a (a) HPLC autosampler, (b) UHPLC autosampler and (c) switch valve manual injector. Schematic of HPLC sample injection method: (a) HPLC autosampler subsystem and (b) Waters U6K injector. Laminar flow profile of a liquid in an open tube. xii

16 List of Abbreviations and Symbols (i) (o) 1D 2D ACN AFT AU B C C C18 CAD CF CL (D) CN ELSD FLD FTIR g GAFA H HILIC Inlet ratio Outlet ratio Uni-Dimensional/One-Dimensional Two-Dimensional Acetonitrile Active Flow Technology Absorbance Units Slope of linear relationship Correlation Correlation matrix Octadecyl Corona-discharge Detector Curtain Flow Chemiluminescence (Detector) Cyano Evaporative Light Scattering Technique Fluorescence Detector Infrared Detection Grams Geometric Approach to Factor Analysis Height equivalent to theoretical plate Hydrophilic Interaction Liquid Chromatography xiii

17 HPLC i.d. IEX k k' High Performance Liquid Chromatography Internal Diameter Ion Exchange Chromatography Retention factor Scaled retention matrix k T Transposed scaled matrix K 0 L LC mau mg MeOH min MS mv ml MΩ N n NMR NP n p n T Specific permeability Column length Liquid Chromatography Milli-Absorbance Units Milligrams Methanol Minute Mass Spectrometry Milli-Volts Millilitres Megaohm Theoretical Plate Count Number of peaks Nuclear Magnetic Resonance Spectroscopy Normal Phase Peak capacity Two-dimensional peak capacity xiv

18 n Tƒ P d PDA PFP PSF Q R RI RP RPHPLC RSD S SD SEC SFC t 0 t g TIC t M TOF t R u F UHPLC Two-dimensional practical peak capacity Particle size Photodiode Array Pentafluro-Phenyl Parallel Segmented Flow Quadrant Resolution Refractive Index Reversed Phase Reversed Phase High Performance Liquid Chromatography Relative Standard Deviation Sensitivity Standard Deviation Size Exclusion Chromatography Supercritical Fluid Chromatography Column dead time Gradient duration Total Ion Count Dead volume time Time of Flight Retention Time Flow rate Ultra High Performance Liquid Chromatography xv

19 vs. UV UV-Vis Å α β β Δc ΔP η µau Versus Ultraviolet Ultraviolet Visible Absorbance Detector Angstrom Selectivity factor Spreading angle Spreading angle matrix Concentration difference Pressure Viscosity Micro-Absorbance Units µl Micro-litres xvi

20 Abstract Complex samples of natural origin, such as, tobacco leaf extracts and coffee bean extracts, comprise a vast number and variety of compounds. The characterisation of such complex samples can be arduous and time consuming. Multidimensional analytical techniques have the potential to provide detailed and informative characterisation data, especially if these techniques exploit separation selectivity and detection selectivity. This thesis explores multidimensional approaches for the analysis of complex samples using High Performance Liquid Chromatography (HPLC) and selective detection. Uni-dimensional HPLC (1DHPLC) multi-detection and multiplexed detection through the use of Active Flow Technology (AFT) columns were used to characterise tobacco leaf extracts to determine key chemical entities by systematically reducing the data complexity of the sample whilst obtaining a greater degree of molecule-specific information. Even though substantial chemical data was obtained, there were limitations; in particular the separation performance of 1DHPLC. This lead to studies directed towards two-dimensional HPLC (2DHPLC), where separation selectivity could be exploited by optimising stationary phase and mobile phase combinations; tailored to the characteristics of the sample. In this work, various combinations of stationary and mobile phases were explored to determine the optimal 2DHPLC selectivity conditions for the separation and analysis of tobacco leaf extracts. However, the performance of 2DHPLC is compromised by the interface between each dimension, namely the compatibility of the solvents and the volume of fluid required in the transfer from the first to the second dimension. This incompatibility often results in loss of separation performance. A preliminary investigation was therefore undertaken to assess the performance of AFT columns in 2DHPLC. The outcomes from this work showed that efficiency and sensitivity could both be improved with the use of AFT columns, however, the results were highly dependent on the construction of the HPLC system. xvii

21 Preface The ever increasing need for regulation and testing drives the demand for faster and more efficient analytical methods. The focus of this thesis is selectivity in analysis, albeit, detection selectivity or physical separation by way of chromatography. For that purpose, the content of this thesis is largely orientated towards the analysis of complex samples. Chapter 1 is an introduction to the thesis topic detailing the basics of unidimensional HPLC and multidimensional HPLC. It also introduces the limitations and issues associated with current chromatography columns and a briefing on new column technology developments, i.e., Active Flow Technology (AFT) that may overcome these issues. Two AFT column formats are described, Parallel Segmented Flow (PSF) and Curtain Flow (CF). This chapter also details the research problems, project aim and objectives. Chapter 2 is a comprehensive analysis of tobacco leaf extract using uni-dimensional HPLC with multiple detection. In total, six different methods of detection were employed. The body of work in Chapter 2 investigates the types of compounds found in tobacco leaf extract and outlines the practical limitations of the detection methods for the characterisation of chemical signatures. Chapter 3 is an extension of Chapter 2, but employs PSF columns to achieve comprehensive characterisation of the tobacco leaf extract within a fraction of the time it took using the multi-detection process used in Chapter 2. The PSF column provided the platform required for simultaneous multiplexed detection of biomolecules, allowing the combination of both destructive and non-destructive detectors, without additional dead volume tubing. In Chapter 2 a Phenomenex C18 column was used and in Chapter 3 a ThermoFisher Scientific C18 column was employed, since AFT columns are prepared only by ThermoFisher Scientific. Chapter 4 explores the selectivity differences between column brand manufacturers. Specifically, how columns from different manufacturers performed for complex samples, not just simple laboratory selectivity testing. To evaluate this broader scoping selectivity testing, 2DHPLC was used, such that the second dimension of the 2D system served as a selectivity detector. It was found that the general xviii

22 performance of columns of the same phase from different manufacturers showed significantly different retention behaviour for certain analytes. Chapter 5 is a 2DHPLC selectivity study on the compounds found in a tobacco leaf extract. Four different stationary phases were employed in the first dimension and a C18 column was used in the second dimension. Each of the stationary phases was also subjected to two different solvent systems. In total eight systems with different selectivity properties were analysed and compared. One of the main issues related to choosing divergent retention behaviour between both dimensions is solvent incompatibility during volume transfer from the first dimension to the second dimension, leading to Chapter 6 of this thesis. The work in Chapter 6 explored aspects of coupling each of the two dimensions in 2DHPLC. This work tested the performance of AFT columns in 2DHPLC. This study was a preliminary investigation to test the concept of AFT columns with 2DHPLC and more specifically the hydro-dynamic flow between dimensions. During this study, unexpected results were obtained as the AFT columns did not perform as anticipated. This caused a further investigation to the underperformance of AFT columns. The performance of a CF column was compared to a conventional column format on a HPLC system, a UHPLC system and a system with a manual injector. The results showed that the HPLC system had large system dead volume and had pressure fluctuations during the injection process, diminishing the performance efficiency and sensitivity of AFT columns. Therefore, an important conclusion was that the user must be aware of the limitations of the HPLC/UHPLC system so as to gain the greatest benefit from AFT columns. Chapter 7 is a general conclusion of the work undertaken here, and future directions for 2DHPLC-AFT. xix

23 CHAPTER ONE Introduction 1

24 1.1 Uni-Dimensional HPLC High Performance Liquid Chromatography (HPLC) is an analytical chromatographic technique for the separation and analysis of compounds and is used in a range of industries such as pharma and food, clinical, toxicology, forensic and environmental sciences. The fundamental aspect of a liquid chromatographic (LC) technique is the physical separation of analytes through distribution and conditional interactions between a flowing liquid known as the mobile phase and a solid particulate, porous bed known as the stationary phase; the liquid moving in a definite direction. It is capable of fast and accurate quantitative analysis in automated operation modes whilst being a flexible technique for the sample of interest, with respect to selectivity in these specific interactions [1]. Uni-dimensional HPLC is the use of a single stationary phase (separation mechanism) where the compounds of a mixture are displaced along a single separation axis providing one separation dimension of a sample mixture. There are many modes of HPLC, reversed-phase (RP), normal-phase (NP), size exclusion chromatography (SEC), ion exchange chromatography (IEX), and chiral chromatography just to name a few. RPHPLC is the most popular method of choice and is the mode in which this study is focused. RPHPLC employs a non-polar stationary phase with a polar mobile phase, typically aqueous/organic mixtures [1]. Operation can be either isocratic elution or gradient elution when more complex samples are to be analysed, since in gradient elution the peak capacity is greatly increased. The time at which an analyte elutes from the column and is detected by the detector is known as retention time (t R ), which is dependent on the selective interactions between solute and stationary and mobile phases [1]. Analyte separation efficiency is based on a number of factors including, but not limited to, mobile phase stationary phase selectivity, temperature, flow rate, and solvent ph [1]. The choice of stationary phases is vast and in combination with a variety of solvent systems, selectivity can be tuned in order to optimise resolution. Resolution (R) is a term used to define how well one component is separated from another, according to equation 1.1 [1,2]. R = N 4 (α 1) α k (k+1) (1.1) 2

25 Where N is the efficiency in terms of the number of theoretical plates (see equation 1.2), k is the degree of retention (see equation 1.3), known as the retention factor, and is the selectivity factor (see equation 1.4) [1]. Each of these variables can be more or less independently adjusted to bring about separation. N = L H = (t R σ )2 (1.2) k = t R t M t M (1.3) α = k B k A (1.4) Where, L is the length of the column, H is the height equivalent to a theoretical plate, t R is the retention time of the peak, t M is the dead volume time, and σ is the standard deviation of distribution. Efficiency is a measure of separation power and performance afforded by the column, either expressed as the number of theoretical plates (N), or by the height equivalent to a theoretical plate (H). The greater the efficiency, the narrower the band width, the greater the peak capacity (n P ), as shown in equation 1.5 [2]. According to equation 1.1, resolution is limited by the number of theoretical plates (N) available, which is proportional to the length of the column (L) and related to the particle size (P d ). n P = 1 + N 4 B c B t 0 t g +1 (1.5) Where B is the slope of the linear relationship between the log of retention factor and solvent composition, t 0 is the column dead time, t g is the gradient duration, and Δc is the difference in concentration at the beginning and end of the organic modifier [3]. Thus, to achieve a good degree of resolution and a high number of theoretical plates, a longer column would be required. However, the longer the column, the greater the required pressure to push the mobile phase through the column (see equation 1.6) [1-3]. P = u F η L K 0 πr 2 P d 2 (1.6) Where u F is the flow rate, η is the viscosity, L is the column length, K 0 is specific permeability, πr 2 is the column radius area and P d 2 is the particle diameter. As a 3

26 consequence, the peak capacity (separation power) of uni-dimensional separation is also limited by the pressure and hence restricted to relatively simple samples. Otherwise, the peak capacity is exceeded and the information derived from such chromatographic separations provides limited information [1-3] Limitations of Uni-Dimensional HPLC The limitations of uni-dimensional HPLC has been discussed in detail by Guiochon [4] and essentially can be narrowed to three main factors; (i) insufficient theoretical plates available for separation - restricted by the pressure limitations of equipment and columns, (ii) heterogeneity of the column bed, and (iii) heterogeneous migration of analytes as a result of viscous friction generated by forcing a fluid through a finely divided bed at high pressure. This causes non-isothermal conditions within the column and hence solutes in warmer regions within the column bed migrate faster than solutes in the cooler regions of the column bed [4]. A chromatographer may increase the length of a column to increase the number of theoretical plates for greater resolution, however, this increases the pressure as a linear function (equation 1.6), eventually reaching the pressure limitation of the system [1-3]. Thus, operational flow rates must then be lowered and separations eventually become inefficient and very time consuming. Almost anything can be achieved, but the currency to pay the price is time. These limitations severely restrict the analysis of complex samples, whose components have a randomness to their structures and hence their retention properties, effectively just 18% of peak capacity being utilised [5]. This factor further limits the resolving power of the column, as discussed by Felinger [6] and places restrictions on the operation of uni-dimensional separations. Thus gradient elution must often be employed for the separation of analytes that make up a complex mixture so as to maximise the peak capacity. Therefore, the separation power and resolution of a uni-dimensional system is impractical for complex samples, essentially due to the insufficient peak capacity [4,7-10] and the need for more time efficient techniques. Thus, for the analysis of complex samples a move to multidimensional separation strategies is required. 4

27 1.2 Multidimensional HPLC Multidimensional HPLC can be defined in two ways, (i) multiple separation steps and/or (ii) multiple detectors Multiple Separation Steps - Two-Dimensional HPLC Two-dimensional HPLC (2DHPLC) is the employment of two or more separation steps (columns) operated simultaneously. The sample is injected onto the first column or first dimension and then through a switch valve interface between the two columns a volume fraction from the first dimension is transferred to the second dimension for further separation [7,11]. 2DHPLC systems provide significant peak capacity improvements as the theoretical peak capacity of a 2DHPLC system (n T ) is equal to the product of the peak capacity (equation 1.5) of the first (n P(1D) ) and second (n P (2D) ) dimensions (see equation 1.7), resulting in greater separation power and thus greater resolution [7,11]. Equation 1.7 is valid for truly orthogonal separation steps, but in reality this never occurs, and separation power is lost by selectivity correlation (f coverage ) (see equation 1.7a) [12,13]. n T = n P(1D) n P (2D) (1.7) n Tf = n P(1D) n P (2D) f coverage (1.7a) Because of the expansion in the separation power, 2DHPLC has the potential to improve resolving power to provide chemical signatures for complex samples, potentially providing unique peak displacement. The probability of two or more components having the exact same retention displacement in the first and second dimension is significantly low and decreases as the correlation between dimensions decreases [12,13]. Thus, as the two dimensions become less correlated, the lesser the chance that two components will have the same retention time in both separation dimensions. As each uni-dimensional component of the 2D system provides numerous combinations of mobile and stationary phases, the combination possibilities are multiplied for 2DHPLC. The more different the selectivity conditions of each dimension, the greater the separation power and efficiency, providing ideally orthogonal retention behaviour [12,13]. 5

28 Orthogonality Orthogonality of two-dimensional separations can be achieved in two ways, (i) combination of two different separation techniques (i.e. LC x SFC) and (ii) two of the same separation techniques each with different separation conditions (i.e. 2DLC normal and reversed phase for example). Orthogonality is by definition the statistical independence of one dimension to the other [13-16]. The closer the two dimensions are to being orthogonal to one another, peak capacity and separation space and power increases. Although orthogonality is presumed to be dependent on the independence of the separation mechanisms of the two separation dimensions utilised, it is also the combination of separation conditions and solute properties [13-16]. If the first and second dimensions have exactly the same sets of selectivity conditions, the orthogonality of the 2DHPLC system is diminished and essentially resulting in the same retention behaviour of a uni-dimensional HPLC system. However, it is impractical for two dimensions to be perfectly orthogonal as solute retention between the first and second dimension will have some degree of correlation, reducing the available separation space [13,16]. Thus, the objective of dimension selectivity of a 2DHPLC system would be ensuring that both dimensions are as divergent as possible with respect to solvent compatibility between the two dimensions. Orthogonality can be measured using various data processing methods, such as a Geometric Approach to Factor Analysis (GAFA) [16,17], which has been the predominant descriptor in this study. A simple overview of GAFA is therefore discussed. Geometric Approach to Factor Analysis (GAFA) GAFA is a processing method that provides information about the separation space utilised in two-dimensional separations by the statistical measure of orthogonality of the system. It also provides a measure of the number of peaks detected (n), spreading angle (β), correlation, practical peak capacity (n Tƒ ) and percentage usage of separation space [16]. GAFA is based on a series of equations by utilising the peak capacities of each dimension and retention time on the first and second dimension of each peak to calculate correlation (equation 1.8) and derive a correlation matrix (equation 1.9), by which the spreading angle is calculated (equation 1.10) and the spreading angle matrix is established (equation 1.11) [16]. 6

29 C = ( 1 n 1 ) k T k (1.8) Where n is the number of peaks in each data set, k is the scaled retention matrix and k T is the transposed retention matrix. 1 C 1 2 C 1 j C C = (1.9) C i 1 C i 2 1 Where C ixj = C jxi is the correlation calculated from equation 1.8 of the two sets of retention data. It is through this series of equations that the degree of correlation between two dimensions is calculated. The correlation coefficient value obtained is between -1 and 1, where 0 is orthogonal and -1 or 1 is 100% correlated. β ik = cos 1 C ik (1.10) 0 β β = 1 2 β (1.11) Where the spreading angle β ik is the inverse cosine of the calculated correlation value C (equation 1.8) between two retention time data points from which the spreading angle matrix can then be calculated in the case of a two-dimensional separation space (see equation 1.11). GAFA also allows for the calculation of the practical peak capacity that is available between two separation dimensions. Theoretical peak capacity is the product of peak capacities of both dimensions (see equation 1.7), however due to overlapping of zone areas the practical peak capacity is less than theoretical peak capacity, which is the product of peak capacity of both dimensions minus the unavailable separation space due to correlation between dimensions (f coverage ) (see equation 1.7a) [16]. Figure 1.1 illustrates the separation space of the two-dimensional retention space, detailing the practical peak capacity as indicated by the gridded section, spreading angle (β), and unavailable separation space (f coverage ) shown by areas A and C. Figure 1.1 also shows the relationship between spreading angle, peak capacity and thus measures the orthogonality, where the greater the spreading angle the greater the practical capacity and thus less correlated the dimensions are to one another, giving a close to orthogonal system and vice versa [16]. 7

30 Separation Step 2 2 nd Dimension A β C Separation Step 1 1 st Dimension Figure 1.1 Illustration of a non-orthogonal two-dimensional separation space, where β represents the spreading angle and practical peak capacity is represented by the grid. The analytical performance of the selectivity of a 2DHPLC system can be analysed and measured through GAFA and allows the chromatographer to optimise a 2DHPLC system for the analysis of a complex sample accordingly [13,15,16]. Through the measure of orthogonality, the 2DHPLC system of choice would be the one that firstly resolves the target component(s) and then the highest number of peaks within the objectives of the analytical demand, irrespective of the measure of correlation. Having said that it is common for the analyst to choose a system that offers the greatest measure of separation divergence since it offers the greatest probability of component separation. Modes of Two-Dimensional HPLC 2DHPLC can be achieved in either offline or online mode. Offline 2DHPLC is the manual collection of discrete eluent fraction(s) from the first dimension into vial(s) and then re-injection onto the second dimension, one injection per fraction. Offline 2DHPLC is quite a simple and time efficient approach, however, the method is subject to sample contamination during collections, sample degradations during waiting period of re-injection, and sample dilution as the entire collected fraction is difficult to completely re-inject into the second dimension, resulting in decreased sensitivity [7,11]. 8

31 Online 2DHPLC involves the use of a volume transport medium between the two dimensions, typically an automated two-position switch valve [10]. The online approach has three methods of analysis: (i) Heart cutting, (ii) Online comprehensive and (iii) Comprehensive incremental heart cutting. Heart cutting 2DHPLC technique is when a fraction of eluent volume from the first dimension at a particular time is cut and transferred by the automated switch valve into the second dimension separation step. Online comprehensive 2DHPLC is the transportation of the entire first dimension into the second dimension for complete analysis. The switch valve(s), usually employing two identical sampling loops, fractionates the first dimension in real time by a load and inject process, where a fraction is loaded onto one loop another fraction is injected into the second dimension [11]. Online comprehensive 2DHPLC is a time efficient and automatic method of analysis, allowing complete fraction transfer without loss or contamination. However, online comprehensive 2DHPLC has a complex separation process, in which the second dimension is required to be fast enough to compensate for the constant re-injection of the first dimension fractions, whilst maintaining sufficient resolution. The second dimension must not only complete its separation step but also be re-equilibrated, ready for the next fraction, otherwise the sampling frequency can cause the peaks in the second dimension to overlap, this is known as the wrap around effect [7,11,18]. Lastly, comprehensive incremental heart cutting 2DHPLC is operation mode (i) i.e., heart-cutting, but in a comprehensive manner. It involves a series of heart cuts across the first dimension at every period of time, followed by injection into the second dimension for analysis. Thus, for every fraction cut from the first dimension and transferred into the second dimension there is also a single injection process into the first dimension [7,11,18]. For example, a 10 minute separation on the first dimension is cut at every 0.1 of a minute for 0.1 minutes and is transferred into the second dimension, giving us 100 fractions across the 10 minutes and therefore 100 single first dimension injections. Through this approach the entire sample across the first dimension is also analysed in the second dimension, resulting in a comprehensive 2D analysis of the sample, without time constraints or wrap around effects [7,11]. Consequently, a higher peak capacity column can be used in the second dimension in order to maximise the separation power, but at the cost of time. Comprehensive incremental heart cutting 2DHPLC is potentially a useful technique 9

32 for the characterisation of chemical signatures of complex samples, provided the cost in analysis time can be afforded, and this depends on the purpose of the analysis Two-Dimensional System Selectivity Stationary Phase There are countless stationary phases available to chromatographers from many column manufacturers. In 2DHPLC the type of stationary phase used in both dimensions is a very important factor in maximising the separation power. As separation power is define by the peak capacity as a result of chromatographic separation performance, different types of stationary phases provide different selectivity properties and thus peak capacity. In the case of 2DHPLC, separation power is expressed as the product of each dimension peak capacity (see equation 1.7 and 1.7a). The selectivity of each dimension (including stationary phase) must be as different as possible to attain maximum peak capacity, which is the essence of separation power. The difference in stationary phase selectivity can be defined through the degree of correlation of a two-dimensional system as discussed in Multiple Separation Steps - Two-Dimensional HPLC; the lower the degree of correlation the greater the separation power [7,19]. The selectivity properties of a stationary phase are dependent on the chemical and physical nature of the surface, and how analytes interact at a molecular level. Even in reversed phase systems, the selectivity choices are substantial. Some of the key factors that influence selectivity in RPHPLC include the length of alkyl chains, the presence of an aromatic ring for resonance π-π interactions with aromatic solutes, double or triple bonds within the stationary phase for non-resonance π-π interactions and polar embedded groups [2,20,21]. In reversed phase systems some of the more popular phases are the Octadecyl (C18) [21,22], Cyano (CN) [21,23], Pentafluorophenyl (PFP) [15], and carbonaceous adsorption-type phases such as the Hypercarb and carbon clad zirconia stationary phases [24,25]. Mobile Phase Another factor in the selectivity of the dimensions in 2DHPLC is the mobile phase. The mobile phase also is important, but this depends on the nature of the stationary phase. In reversed phase HPLC for example, the mobile phase plays a rather passive 10

33 role in attaining selectivity, however, in normal phase roles the mobile phase is far more dominant [7,19]. Mobile phase carries the sample through the column and allows the analytes to interact with the stationary phase [1]. In RPHPLC, non-polar columns are used in combination with polar solvent mixtures of water and an organic solvent. As stated earlier the mobile phase can be run either in isocratic elution, where the solvent composition remains constant throughout the separation, or gradient elution, where solvent composition changes during the separation. Various aspects of the solvent can significantly affect the retention of the anlaytes, such as the polarity of the solvent, solvent ph, additives and temperature [1,2]. Through the choice of solvent used as the mobile phase, the separation and selectivity differences between dimensions can be optimised. However, care must be taken when choosing mobile phases for each dimension to ensure solvent environment compatibility between the dimensions. Solvent incompatibility is one of the most serious limiting factors in efficient 2DHPLC [10,26]. Factors, such as, solvent immiscibility [27], solvation phenomena [19], ph mismatch [28,29] and viscous contrasts [30] can lead to serious loss of performance, with the effects becoming worse as the transfer volume between dimensions increases Multi-Detection The role of a detector in HPLC is to enable visualisation of the chromatographic separation and perhaps also to provide insight into the chemical or physical properties of the analytes within the sample [2]. There are many detectors suited to liquid chromatography, such as, the ultraviolet-visible detector (UV-Vis) or a photodiode array, which is the detector of choice for simplicity and versatility, fluorescence detectors (FLD), refractive index (RI) evaporative light-scattering detector (ELSD), and corona-discharge detector (CAD), also finding mainstream usage [1,2]. Bio-type detectors are also useful since they can provide chemical information about the sample, such as the antioxidant detector - 2,2-diphenyl-1- picrylhydrazyl free radical (DPPH ) [31-33], or even the chemiluminescence detector (CL) [34], which, depending on the type of chemiluminescence reagent can assist in evaluating the properties of the analyte components, and increase the selectivity of detection. For example, the permanganate reagent responds to morphine, while not heroin, and tris(2,2 -bipyridine)ruthenium(iii) reagent provides exactly the reversed response [35], hence these two chemiluminescence reagents provide great 11

34 discrimination between heroin and morphine. Detectors, such as, circular dichroism, mass spectrometry (MS), nuclear magnetic resonance (NMR), and infrared detection (FTIR) provide insight into the physical nature of the sample [2]. Since each of these detectors listed above respond to different sample attributes, selectivity in detection can be generated, which in itself can be termed as multidimensional HPLC, especially if a multiplexed approach is implemented [36-38]. The combination of sample specific detectors can provide two dimensions to the sample, for example, the use of DPPH for the detection of antioxidants [36] and CL detection method for the analysis of opium poppy alkaloids [38]. Thus, multidimensional detection systems can provide information of the types of compounds and their chemical attributes. Chapter 2 details the mechanisms of detection of the UV-Vis, CL, MS and DPPH detectors. 1.3 Column Technology Active Flow Technology (AFT) In the recent years column technology for HPLC has advanced greatly; peak capacities have increased considerably through the introduction of ever decreasing particle sizes and the more efficient core shell particles. Since the separation is more efficient, a flow-on effect to increased sensitivity is also apparent [38-45]. Chromatographers have known for years that column beds are not homogeneous, but rather the bed density varies systematically across the radial cross section of the column [46-49], and along the column axis [47,49-52]. The wall effect is an important contributor to the loss of separation performance [44,53-55]. Shalliker and Ritchie [44] recently reviewed aspects of column bed heterogeneity and hence need not be discussed here further, except to say, that the variation in column bed packing density and the wall effects lead to a distortion of the solute plug, such that bands elute through columns in profiles that resemble partially filled soup bowls rather than thin flat solid discs [44]. Figure 1.2 shows this as a two-dimensional profile [44]. The authors claim that in real-time experiments looking from above the profile reveals the hollow nature of the trailing section of the plug. The end result is that it takes many more plates to separate these partially hollow plugs than would be required if the discs were solid and flat [49,51,54]. 12

35 To overcome the band broadening issues associated with wall effects and the variation in radial packing density a new form of column technology known as Active Flow Technology (AFT) was designed [44]. The purpose of this design was to remove wall effects through the physical separation of solvent eluting along the wall region, from that of mobile phase eluting in the radial central region of the column [56]. There are two main types of AFT columns; (i) Parallel Segmented Flow (PSF) columns and (ii) Curtain Flow (CF) columns [44]. Figure 1.2 Column band profile in parabolic distribution Parallel Segmented Flow (PSF) AFT columns utilise a multi-channel end-fitting with a special purpose-built frit that allows the separation of flow between the centre and wall regions of the column [56]. Figure 1.3 illustrates the design of the multi-channel end fitting. PSF is a column format using the AFT end-fitting on the outlet of the column only (see Figure 1.4). The flow from a PSF column elutes from either of two separated radial zones: The central flow region of the bed, which is separated from the peripheral or wall flow region. This is achieved by using an annular frit design, and a multi-channel end fitting. An impervious ring divides the outlet frit into two parts; an inner portion of frit channels flow from the central region of the bed out a central exit port on the outlet fitting, while an outer ring of frit channels solvent that migrates down the wall region out the peripheral ports on the outlet fitting (Figure 1.5) [56]. 13

36 Figure 1.3 Illustration of Active Flow Technology column design [41]. Flow Inlet Outlet Figure 1.4 AFT column - PSF column format. These AFT columns work by taking the concentrated, more uniform volume portion of the band, which is the central region of the total flow, to the detector and removing the tailing sections of the band profile through the peripheral ports to either a waste vessel, or to additional detectors. Figure 1.5 illustrates the segmentation of the band profile. Figure 1.5 Column band profile; AFT takes the central region from the parabolic shaped band to detector and wall region to waste. 14

37 The exiting flows are adjusted to the desired ratio through pressure management. In essence, this design effectively establishes within the larger format column, a virtual column having a narrower diameter, the dimensions of which are related to the volumetric ratio of flow exiting the column through the centre, relative to the flow exiting through the peripheral zones. For example, if 43% of the solvent exits the column through the central zone then the radius of this virtual column is effectively 3.0 mm i.d., while a 2.1 mm i.d. column would be obtained if 21% of the solvent left via the central port and 79% from the peripheral ports [43,55]. Fine tuning of the ratio between the central and wall flow showed improvements in the number of theoretical plates and increased sensitivity of detection [40,53]. It has been reported that PSF columns have shown significant increases in efficiency (N) by up to 43% and sensitivity of up to 21% compared to the column s conventional format. This was achieved at a ratio of 45:55 (central flow:wall flow) [40]. Although, the wall region is taken to waste, it can also be utilised for analysis by connecting the peripheral port to a detector. Camenzuli et al. [37] had investigated the PSF column format as a platform for multiplexed detection. Their investigation showed that although the peripheral ports (wall flow) were the tailings of the band profile having decreased efficiency compared to a convention column; detection in UV-Vis showed sensitivity to be approximately the same as a conventional column. Thus, as a spinoff advantage of AFT columns, PSF columns are appropriate for multiplexed detection [37] Curtain Flow (CF) Curtain Flow (CF) column formats utilise the AFT end-fittings at both the inlet and the outlet of the column. Figure 1.6 illustrates the format of the CF column. In this mode of operation the sample is injected into the central port of the inlet, whilst additional mobile phase is introduced through the peripheral ports of the inlet to curtain the migration of solutes through the central region of the column. The AFT outlet of the CF column is managed in the same way as the PSF column. In this manner a wall-less, infinite-diameter or virtual column is established [43,47,55]. The purpose of CF columns is to actively manage the migration of sample through the column to prevent the sample from reaching the wall region. Thus, the solute concentration upon exit to the detector would be maximised increasing sensitivity of 15

38 around 2.5 times greater than the conventional column format when using UV detection [53]. There are various ways to set up the solvent delivery system to the central and peripheral ports of the inlet; (i) 2 pump system [44] and (ii) split-flow system [43]. Details about theses set-ups are further explained in Chapter 6 - Part B. CF columns are highly suitable for low concentration samples and to achieve high sensitivity in detectors that are limited in high flow rates, such as MS [43]. Figure 1.6 AFT column - CF column format. 1.4 Complex Samples of Natural Origin Natural origin samples, such as, plant extracts, foods and drugs contain hundreds and even thousands of complex compounds with vast sample dimensionality [57,58]. Characterisation or profiling of such complex matrix samples is arduous and time consuming, where consideration to certain factors, such as, the nature of the compounds, volatility and sample dimensionality must be taken. There is always a need and push for simpler and more straightforward techniques of obtaining as much information as possible with analytical efficiency, of complex natural origin samples, especially industries, such as, food and pharmaceuticals. This study focuses on the samples - tobacco leaf extract and espresso coffee because they are complex samples having a rich variety of compounds. They are also samples that are easily prepared with minimal sample prep, an important consideration in this study, since the work in this thesis is more concerned with enabling the process of discovery, rather than the discovery of the plant value. Tobacco leaf is a complicated chemical system containing compounds such as carbohydrates [59,60] alkaloids, (e.g. nicotine, nornicotine, anatabine and anabasine [61,62]), amino acids [61,63], alcohols [64] and antioxidants [65]. The constituents of coffee have also 16

39 been extensively researched ranging from carbohydrates, nitrogenous components (i.e. alkaloids, trigonelline, nicotinic acid, proteins and free amino acids), chlorogenic acids, lipids, volatile compounds (i.e. tocopherol, triglycerides, sterols and fatty acids), and carboxylic acids [66]. Coffee is also well known for its antioxidant properties, with various methods of detection hyphenated with HPLC [13,15,32,67-70]. Both samples make an excellent candidate for the analysis of chemical profiling using multidimensional approaches in HPLC Sample Dimensionality Coffee and tobacco samples contain hundreds of different types of compounds with a range of functional groups. Such samples are known to have multiple (n) dimensions with respect to the number of unique features they possess. Sample dimensionality can be expressed through dimensionality from a detector (i.e. HPLC-MS or HPLC- DPPH) specific to compounds of interest, technique dimensionality (HPLC/TLC or HPLC/GC) involving different separation methods or dimensionality from selectivity (2DHPLC) which is based on primarily the chromatographic forces (i.e. ionic, polar, π π, van der Waals, and aromatic forces as well as size and shape) between sample analytes and stationary/mobile phase. To analyse a multidimensional sample, a multidimensional system would be required, for example in the case of 2DHPLC, low molecular weight polystyrenes can be separated based on three unique features/properties or dimensions that is molecular weight, tacticity and enantiomers under three different chromatographic conditions (i.e. 3 dimensions) that is specific for each property [71,72]. 1.5 Research Problems To characterise and profile complex samples can be arduous and time consuming, especially samples of natural origin that vary vastly in compound types. For specific information about a sample, for example bio information, such as, antioxidants, there are countless methods developed for their detection. However, when the nature of the sample is unknown and the analyst requires as much information about the sample as possible, various issues come into play. Firstly, sample dimensionality and selectivity must be taken into account, where one must choose the appropriate selectivity conditions of mobile and stationary phases that would ensure optimal 17

40 separation with respect to sample dimensionality. To choose the correct selectivity conditions the nature of the sample must be known, which can be determined through various detection methods. This brings us to our second issue and that is the correct choice of detection to provide the information for characterisation in-line with the analytical requirements. There are many methods of detection, such as, UV- Vis, MS, FLD, CL, etc. that can be hyphenated with HPLC or 2DHPLC, however, some of these detection methods are destructive to the sample. The third issue is sample amount, if there is only a small amount of sample then a destructive method of detection would not be suitable for sample characterisation, limiting the analyst to non-destructive detection methods, such as UV-Vis, FLD and RID, to conserve the sample. Lastly, separation power and efficiency must also be considered to adequately characterise and profile a complex sample. Without proper separation power and efficiency other issues arise such as peak overlapping, co-elution of peaks and poor peak shape, making it difficult to differentiate between positive and negative detections. As mentioned previously, uni-dimensional HPLC is limited in separation power and efficiency, thus a two-dimensional HPLC would be sought. 1.6 Project Aim The aim of this study is to develop a fast and efficient multidimensional HPLC method that will enable the characterisation of chemical signatures of complex samples. 1.7 Project Objectives To characterise a complex sample through various methods of detection simultaneously. To optimise selectivity conditions of a 2DHPLC system for the analysis of complex samples. To improve the efficiency and sensitivity of a 2DHPLC system through the use of AFT columns. Ultimately, to combine 2DHPLC analysis with multiplex detection through the use of AFT columns within a single analysis period. 18

41 CHAPTER TWO Comprehensive Sample Analysis using High Performance Liquid Chromatography with Multi-Detection 19

42 2.1 Introduction Tobacco is a highly complex substance containing thousands of different compounds; at least 3800 constituents were reported by Dube and Green [73]. Tobacco products with varying strengths and flavours are produced by blending different types of unprocessed and processed tobacco, with flavours, solvents, preservatives, binders, strengtheners, and fillers added for taste, consistency, moisture-holding capacity, dilution, preservation, stability and to increase the volume of the end product [61]. Due to the complexity and thousands of compounds with a wide range of compound types, tobacco extracts are an excellent candidate to rigorously test the suitability of a multiple detection system in a uni-dimensional separation environment. UV-Vis absorbance detectors are frequently used with HPLC systems due their versatility, sensitivity and relatively low cost. UV-Vis spectral characteristic selectivity is inherent, based on the absorption spectrum of the chromophores of interest and the ability to monitor at particular wavelengths. However, due to the complex nature of natural extracts that contain many groups of compounds with similar chromophores, the selectivity gains do not necessarily help to distinguish between key components of interest [74]. Detection systems involving post-column reactions, such as chemiluminescence systems and the diphenylpicrylhydrazyl (DPPH ) free radical assay afford more exclusive selectivity, which may aid in simplifying the sample data complexity, without increasing the physical separation of sample components. Chemiluminescence has been shown to offer detection selectivity for target analytes in plant derived materials such as opium poppy extracts, wine, coffee and fruit juice [68,75-77]. The selectivity of these systems is dependent on the type of chemiluminescence reagent in use [34]. Acidic potassium permanganate is particularly sensitive to readily oxidisable phenols and related compounds [78], and has been used both for quantitative analysis and to screen for the antioxidant capacity of sample components, which adds further capacity to the detection selectivity when investigating complex chemical systems [79]. Chemiluminescence reactions with ruthenium complexes, such as, tris(2,2 -bipyridine)ruthenium(iii), have predominantly been used to detect tertiary amines, but less substituted amines and 20

43 various organic acids also produce some light with this reagent [80,81]. Phenolic functionality, however, quenches the emission [82], and therefore tris(2,2 bipyridine)ruthenium(iii) offers somewhat complementary selectivity to that of acidic potassium permanganate. In contrast to these highly selective chemiluminescence reagents, manganese(iv) produces light with a much wider range of compounds, including many that can be detected by tris(2,2 bipyridine)ruthenium(iii) and acidic potassium permanganate [83]. The combination of these detection modes, including the chemiluminescence and the DPPH detectors used in this study has not been previously utilised for tobacco samples. The information that is derived from these detectors provides a new dimension to the characterisation of tobacco compounds. The on-line DPPH free radical assay is based on a chemical reaction between the separated sample components in the eluent and the stable 1,1-diphenyl-2- picrylhydrazyl free radical [32,68], which provides an assessment of the antioxidant capacity of compounds within the sample. The radical form of DPPH in methanol exhibits a deep purple colour, and when reaction occurs, the pale yellow reduced form (2,2-diphenyl-1-picrylhydrazine) is produced. Detection therefore involves monitoring this decolourisation at 520 nm [33]. Mass spectrometry can be employed to expand the chemical information gained from the detection, allowing for the analysis of species without chromophores and to elucidate the chemical structure based on the molecular weight and fragmentation patterns of the mass spectrum. The sensitivity offered by mass spectrometry and the ability to monitor specific ions allows for a very flexible detections system that adds significantly to the information gained from the other detections systems [84,85]. In this chapter the complexity of tobacco leaf extract is investigated using a multiple detection approach in order to outline the practical limitations of detection systems for uni-dimensional liquid chromatography separations. 21

44 2.2 Experimental Chemicals and Reagents Mobile phase solvents were HPLC grade. Methanol was purchased from Merck (Kilsyth, Victoria, Australia) and Ultrapure Milli-Q water (18.2 MΩ) was prepared in-house and filtered through a 0.2 µm filter. DPPH 2,2-Diphenyl-1-picrylhydrazyl free radical (DPPH ) was purchased from Merck. The DPPH reagent (0.1 mm) was prepared in methanol and degassed by helium sparging. Acidic Potassium Permanganate The permanganate chemiluminescence reagent was prepared by dissolving potassium permanganate (1 mm; Chem-Supply, Gilman, SA, Australia) in an aqueous sodium polyphosphate solution (1% m/v; +80 mesh, Sigma-Aldrich, Castle Hill, NSW, Australia) and adjusting the ph to 2.5 with sulfuric acid (Merck). Manganese(IV) Potassium permanganate was reduced in aqueous solution by sodium formate (Ajax, Melbourne, Victoria, Australia) to form a manganese dioxide precipitate, which was collected by vacuum filtration (GF/A micro fibre filter paper, Whatman, Maidstone, England) and rinsed with deionised water [86]. A portion of the freshly precipitated wet material (0.6 g) was then dissolved in 500 ml of 3 M orthophosphoric acid (Ajax) and sonicated for 30 min. The colloid was then heated for 1 hour (80 C) with stirring, cooled to room temperature and the concentration determined by titration [86]. The stock reagent was diluted to the required concentration (0.5 mm) with 3 M orthophosphoric acid. Tris(2,2 -bipyridine)ruthenium(iii) ([Ru(bipy) 3 ] 3+ ) Firstly, tris(2,2 -bipyridine)ruthenium(ii) perchlorate was prepared via aqueous metathesis reaction between tris(2,2 -bipyridine)ruthenium(ii) dichloride hexahydrate (Strem, Newburyport, Massachusetts USA) and sodium perchlorate. The 22

45 orange-red precipitate was collected by vacuum filtration, washed with minimal ice water (~2 ml) and dried over phosphorus pentoxide for 24 hours [87]. The chemiluminescence reagent solution was then prepared by dissolving the crystals (1 mm) in acetonitrile containing 0.05 M perchloric acid (Merck). Lead dioxide (0.2 g per 100 ml) was then added, followed by 1 minute of vigorous stirring to allow formation of [Ru(bipy) 3 ] 3+. The resultant suspension was allowed to settle prior to filtration through an in-line filter (consisting of a small Pasteur pipette packed tightly with glass wool) Sample Preparation An extract was prepared using 5 g of a commercially available, chopped, loose-leaf tobacco (White Ox Tobacco, CTC Tobacconist, Rooty Hill NSW Australia) in 50 ml Milli-Q water. The sample was sonicated for 1 hour and then filtered through 0.45 µm pore filter prior to injection onto the HPLC Instrumentation and Chromatographic Conditions Chromatographic Analysis Separation was conducted using a 250 x 4.6 mm C18 reversed phase column (Aqua, 5µm P d, 125Å) from Phenomenex Australia (Lane Cove, NSW, Australia). All chromatographic analyses were undertaken using gradient elution with the initial mobile phase of 100% water running to a final mobile phase of 100% methanol, at a rate of 1 %min -1, then held for 20 minutes at 100% methanol. Flow-rate was set at 0.8 mlmin -1 and injection volumes were of 40 µl. Chemiluminescence Detection Experiments were conducted on an Agilent Technologies 1200 series liquid chromatography system, equipped with a quaternary pump, solvent degasser system and autosampler (Agilent Technologies, Forest Hill, Victoria, Australia). Postcolumn acidic potassium permanganate and tris(2,2 -bipyridine)ruthenium(iii) chemiluminescence detection was performed by merging the column eluate (0.8 mlmin -1 ) with the reagent stream (1.0 mlmin -1 ) at a T-piece located immediately prior to the entrance of a coiled-tubing flow-cell (0.8 mm i.d. PTFE tubing) mounted flush against the window of a photomultiplier tube (Electron Tubes model 9828SB, 23

46 ETP, NSW, Australia) set at a constant voltage of 900 V from a stable power supply (PM28BN, ETP), via a voltage divider (C611, ETP). No wavelength discrimination was used in the chemiluminescence detection. The flow-cell, photomultipler tube and voltage divider were encased in a light-tight housing. For manganese(iv) chemiluminescence measurements, the column eluate (0.8 mlmin -1 ) and a formaldehyde solution (2.0 M; 1.0 mlmin -1 ) were merged at a T-piece located 22 cm from the entrance of a commercially available chemiluminescence detector (GloCel, GlobalFIA, Washington, USA) with a dual-inlet serpentine flow-cell [88] and extended range photomultiplier module (Electron Tubes model P30A-05. ETP, NSW, Australia). This stream was then combined with the manganese(iv) reagent (1.0 mlmin -1 ) within the flow-cell. A peristaltic pump was used to deliver each of the reagent solutions. DPPH Detection The DPPH chromatographic experiments were conducted on a Shimadzu analytical HPLC system, comprising a Shimadzu LC-20ADvp quaternary pump, Shimadzu SIL-10ADvp auto injector, Shimadzu SPD-M10Avp PDA detector and a Degassex model DG-440 inline degasser unit (Phenomenex, Lane Cove NSW Australia). The eluent stream was combined with the DPPH reagent at a T-piece, with a flow rate of 1.0 mlmin -1, using a second stand-alone Shimadzu pump. The combined eluent stream then entered a reaction coil (100 µl), which was maintained at a temperature of 60.0ºC within a column heater. The eluent then entered the Shimadzu PDA detector set at a wavelength of 520 nm to detect the radical scavenging compounds within the eluent. UV-Vis Absorbance Detection The UV-Vis absorbance chromatographic data was obtained using the same LC system as stated in Section Instrumentation and Chromatographic Conditions - DPPH Detection, with the PDA detector set at a wavelength of 280 nm. Mass Spectrometry Characterisation of the prominent tobacco constituents was achieved with the aid of high-resolution mass spectrometry. Experiments were conducted on an Agilent 1200 series analytical HPLC system, comprising of a binary pump, solvent degasser 24

47 Signal Intensity (mau) system and autosampler (Agilent Technologies, Forest Hill, Victoria, Australia). A 6210 MSDTOF mass spectrometer (Agilent Technologies) was used with the following conditions: drying gas, nitrogen (7 Lmin 1, 350 C); nebulizer gas, nitrogen (16 psi); capillary voltage, 4.4 kv; vaporizer temperature, 350 C; and cone voltage, 60 V. 2.3 Results and Discussion The water extract from the tobacco leaf is a highly complex sample; this is clearly indicated in the chromatogram in Figure 2.1 obtained using UV-absorbance detection at 280 nm. The majority of the components elute early in the separation (before 40 min) consistent with molecules that have an affinity for water. This is important as it gives a very general indication on the water solubility of the molecules (indirectly based on the methylene selectivity of the C18 column). Secondly, as the gradient profile changes, the sensitivity of the absorbance-type detection is not altered, providing that appropriately treated solvents are used for the separation. This gradient may, however, have an influence on the chemiluminescence and DPPH reactions, which can be more sensitive to changes in solvent composition. Time (min) Figure 2.1 UV-Vis absorbance detector chromatogram. 25

48 Figures 2.2 to 2.4 show chromatograms obtained using acidic potassium permanganate, manganese(iv) and tris(2,2 -bipyridine)ruthenium(iii) chemiluminescence detection systems, respectively. The difference in detection selectivity between these systems and the UV-Vis absorbance detector can be clearly observed. In particular, considerably fewer peaks are observed with the tris(2,2 bipyridine)ruthenium(iii) reagent (Figure 2.4). Interestingly, more peaks were observed using acidic potassium permanganate (Figure 2.2) than manganese(iv) (Figure 2.3). In general, manganese(iv) is a less selective reagent than acidic potassium permanganate [78,83] and the apparent contrary observation in this study is likely to be due to the composition of the water extract from the tobacco leaf. The nature of the tobacco extract is determined by the solvent used for extraction process. Thus, the water will extract compounds that have an affinity for that solvent. For example, polyphenol compounds range in polarity and water will extract the more polar polyphenols. In general, water is particularly efficient at extracting polyphenols from plant matrices [57], and as polyphenolic compounds are known to react readily with acidic potassium permanganate the sample extraction process may favour this mode of detection. Acidic potassium permanganate has previously been shown to have some correlation to DPPH detection [89]. From Figure 2.5 it can be seen that a few of the major peaks observed with DPPH detection come from the same regions as those observed with permanganate chemiluminescence detection. For example, peaks U-V and C in the minute region and peak D in the minute region. It is noteworthy that in this instance more peaks were observed with the permanganate detection system, and the relative signal intensity of the peak clusters is different for the two modes of detection. For permanganate chemiluminescence, the peaks between 40 and 80 minutes elicit a greater signal intensity that those between minutes, with the opposite observed for the DPPH assay. 26

49 KMnO4 Figure 2.2 Chemiluminescence detector - Acidic potassium permanganate chromatogram. MnO2 Figure 2.3 Chemiluminescence detector Manganese (IV) chromatogram. 27

50 Signal Intensity (µau) [Ru(bipy)3] Figure 2.4 Chemiluminescence detector Tris(2,2 -bipyridine)ruthenium(iii) chromatogram. DPPH Figure 2.5 DPPH detector chromatogram. 28

51 Table 2.1 summarises a tentative assignment of sample components, based on our mass spectrometry data and previous literature reports of these molecules in tobacco [62,90-95], which includes a wide range of chemical classes, including alkaloids, organic acids and polyphenols. The tentatively assigned compounds are listed in order of retention time, and have been assigned letter codes to identify them across the range of detection modes in Figures 2.1 to 2.5. Their peak height ratio (peak vs. tallest peak within each chromatogram) was also calculated to easily observe their signal intensity across the detection modes. The components detected by the MS were then cross referenced across the chromatograms based on their retention times. Due to the use of different LC systems for each detector, as well as post-column detection arrangements there were a slight systematic shift in retention times, which was taken into consideration and adjusted for when sample components were tentatively assigned. Upon close and cautious inspection of Table 2.1, however, it is clear that there are limitations with the multiple detection system in use here. For example, quercetin (component X) is known to generate strong chemiluminescence response with acidic potassium permanganate [78], but none is observed here, in spite of a small signal for quercetin (a known antioxidant) using the on-line DPPH assay. The use of the two detectors indicates that quercetin may not be present in high enough concentrations remembering that mass spectrometry is generally much more sensitive than any of the detectors in use in this study [96]. From this we may conclude that it may not actually be quercetin but another co-eluting species may be eliciting the DPPH reactivity and this molecule must not be sensitive to acidic potassium permanganate chemiluminescence. Similar issues arise in the chromatogram obtained using tris(2,2 bipyridine)ruthenium(iii) chemiluminescence detection. Whilst several of the major peaks that can be aligned with those in the mass spectrometry data were identified as molecules that could be expected to produce light with this reagent (such as the organic acids, trigonelline and malic acid), several others were ascribed to phenolic constituents (such as rutin and neochlorogenic acid) that would not be likely to elicit an intense response with tris(2,2 -bipyridine)ruthenium(iii) [82]. We attribute this latter case to coincident peaks from co-eluting sample components, where those that generated significant chemiluminescence responses were not identified in the mass 29

52 spectrometry data, leading to potential misinterpretation of the chemiluminescence data. Upon close inspection of the chromatograms, while quite good for a uni-dimensional separation of a complex sample, it is clear that there are many unresolved peaks. In order to improve the problems associated with this detection protocol a significant enhancement in separation space is required, such as that offered by multidimensional separation. The use of multidimensional chromatography with comprehensive cutting across the first dimension would also bring a significant increase in the total analysis time, but in the case of particularly complex samples, the need for comprehensive analysis may outweigh the desire for fast separations. For instance, with the tobacco leaf extract presented here, using a multiple detector protocol enables a large amount of chemical information to be gained. With appropriate experimental design, all of these detection protocols can be employed in conjunction with multidimensional separation. This will enable a much clearer understanding of the chemical nature of the sample to be achieved while removing the issues outlined here with a uni-dimensional separation system. Since several of these detectors were destructive, only one could be used in any single detection train, unless post column flow stream splitting were to be employed. In this work, the analysis was repeated for each detection mode, which greatly increases the time required to comprehensively review the sample. If comprehensive detection processes were to be employed with two-dimensional separations, the complexity of the analysis would perhaps limit the scope for application. An alternative approach would be to split the flow after elution from the column, however, at the cost of sensitivity, since the amount of sample reaching each detector would be decreased. With this in mind, we consider that there is a need to develop detection processes that permit a more efficient work flow in order to maximise sample information. 30

53 Table 2.1 Identified known compounds via MS detection in tobacco leaf extract and Peak Height Ratios for each mode of detection. Compound ID Compound Name MS (RT min) UV KMnO4 MnO2 [Ru(bipy) 3 ] 3+ DPPH Key Peak Height Ratio A nicotine B nicotine 1 n-oxide C n-nitrosonornicotinine D norcotinine E nornicotine 18.5 F 4-(n-methyl-n-nitrosoamino)-1-3-pyridyl-butanone 5.25 G n-nitrosoanatabasine H n-nitrosoanatabine I anabasine J anatabine K/L 2,2'-bipyridyl M 2,3'-dipyridyl N 3-hydroxycotinine 75.8 O n-ethynorcotinine P cotinine 31.3 Q chlorogenic acid R rutin S capsidiol T caffeoylputrescine U cryptochlorogenic acid V neochlorogenic acid W scopoletin 75.8 X quercetin Y cinnamic acid Z solanesol AA malic acid AB neophytadiene 3.4 AC farnesol No Detection AD nonacosan AE alpha-tocopherol (vitamin e) No Detection AF 2-acetylpyrrolidine 99.7 AG trigonelline

54 2.4 Conclusion Limitations were observed when using multiple detectors to gain chemical information from a uni-dimensional separation of an exceedingly complex sample. Detection limits and co-eluting species were the main drawbacks associated with this system and the separation process is a key component of the development of multidetectors systems. Multidimensional chromatography, while more time consuming, is seen as the primary way in which these limitations can be overcome. This combination of multiple detectors and multidimensional chromatography will afford a clearer understanding of complex chemical systems; however, more time efficient multi-detection protocols must also be developed. 32

55 CHAPTER THREE Multiplexed Detection: Fast Comprehensive Sample Analysis using HPLC with AFT Columns 33

56 3.1 Introduction Natural products, such as those derived from plant extracts, are exceedingly complex, both with respect to the number of components that make up the sample, but also in the varied nature of these constituents. Obtaining a chemical profile can be an arduous process, particularly if screening for a number of different types of compounds. In Chapter two, a comprehensive study of a tobacco sample using HPLC with six different methods of detection was undertaken. The detectors employed were; UV-Vis, DPPH (for antioxidants), three types of chemiluminescence reaction detection, and mass spectrometry. From this study a significant amount of information about the sample was obtained, albeit, rather slowly since the separation itself was in the order two-hours for each of the detectors [97]. Having said that multiple detectors can be used in a series to reduce lengthy run-times, however, this will lead to band broadening, which can be substantial and diminish the HPLC efficiency. Furthermore, only one destructive detector can be used in any single series of detectors, and in the prior study for example, five of the detectors were destructive. Thus, a method of analysis that is time efficient and allows simultaneous destructive and non-destructive detection versatility, whist maintaining HPLC would be desirable. Active Flow Technology (AFT) has been recently employed as a platform for multiplexed detection [37,38]. The column technology was initially developed to overcome the effects of column flow heterogeneity and to increase separation efficiency and sensitivity [40-42,53,56] by establishing a wall-less virtual column within an actual column [55]. The background behind AFT columns has been detailed in numerous publications prior and in Chapter 1 here, and need not be discussed further, suffice to say, that the technology improves separation efficiency because it is able to separate the flow eluting in the radial centre of the column, from the flow near the wall. It does this using a three piece frit and multi-port exit fitting (see Figure 1.3) [56]. A spin-off advantage of this design is that it also subsequently enables more than one detector to be used in a single analysis, with each of the detectors functioning in parallel. This allows simultaneous destructive and nondestructive detectors to be employed. The first reported use of AFT with multiplexed detection was by Camenzuli et al., where they used DPPH and Fluorescence (FLD) detectors [36] and in another study two chemiluminescence detectors and a UV 34

57 detector [38]. This study will extend the concept of multiplexed detection further, showing how DPPH, UV-Vis and MS detectors can be used in a multiplexed arrangement with AFT columns. 3.2 Experimental Chemicals and Reagents Mobile phase solvents were HPLC grade. Methanol was purchased from Merck (Kilsyth, Victoria, Australia) and Ultrapure Milli-Q water (18.2 MΩ) was prepared in-house and filtered through a 0.2 µm filter. 2, 2-Diphenyl-1-picrylhydrazyl free radical (DPPH ) was purchased from Merck. In conventional HPLC, the DPPH reagent (0.25 mm) was prepared in methanol. In multiplexed mode, 0.1% formic acid was added to the mobile phases and DPPH reagent to cater for the mass spectral component of the multiplexed analysis Sample Preparation Samples of tobacco were prepared according to Section Sample Preparation. The espresso coffee (Nestlé Nespresso Ristretto and Decaffeinated, Sydney, NSW Australia) samples were prepared fresh prior to analysis by extraction using an espresso coffee machine (Dè Longhi - Model EN 95.S), diluted four fold, then filtered through 0.45 µm pore filter Instrumentation and Chromatographic Conditions Instrumentation All chromatographic experiments were carried out using a Thermo UHPLC system coupled with TSQ Vantage mass spectrometer equipped with HESI II source (Thermo Scientific, San Jose, USA). The LC component was a Dionex Ultimate 3000 equipped with a quaternary pump, auto injector with an in-line degassing unit and RS diode array detector. The Thermo Vantage TSQ was operated as supplied from the manufacturer. 35

58 Chromatographic Analysis Conventional Mode Separation was conducted using a 250 x 4.6 mm C18 reversed phase column (Hypersil Gold, 5 µm P d, 175Å) from Thermo Scientific (Runcorn, Cheshire, United Kingdom). All chromatographic analyses were undertaken using gradient elution with the initial mobile phase of 100% water running to a final mobile phase of 100% methanol, at a rate of 1 %min -1 then held for 20 minutes at 100% methanol. Flowrate was set at 0.8 mlmin -1 ) and injection volumes were 80 µl. A diagram illustrating the setup of the conventional HPLC system with each of the detectors is illustrated in Figure 3.1. Multiplexed Mode The column used in Conventional Mode was converted to an AFT (PSF) column format and then used for multiplexed detection separations. This PSF column was equipped with four exit ports, enabling up to four separate modes of detection to be employed (Figure 3.2). In this study, the flow ratios through of the four detectors was set at: 18.5% of the total flow rate from the central port, which was connected to the MS detector, 22.5 % of the flow from peripheral port 1 was connected to a UV-Vis detector and 59% of the flow from peripheral port 2 was connected to a DPPH detector via a T-piece. Peripheral port 4 was not used. UV-Vis Detection The UV-Vis absorbance chromatographic data was obtained using the instrument stated in Chapter 2, Section Instrumentation and Chromatographic Conditions - DPPH Detection, with two UV wavelengths set at 280 nm and 254 nm. DPPH Detection Conventional Mode The conventional chromatographic DPPH experiments were conducted on the Thermo UPLC system. The post-column eluent stream was combined with the DPPH reagent at a T-piece, with a flow rate of 0.8 mlmin -1 (1:1, eluent 36

59 stream:dpph reagent), using an additional stand-alone Shimadzu Prominence LC- 20AD pump and a Degassex model DG-440 inline degasser unit to deliver the DPPH reagent. The combined eluent stream then entered a reaction coil (100 µl), which was maintained at a temperature of 60ºC within a column heater. The eluent then entered the Thermo UPLC UV-Vis detector set at a wavelength of 520 nm to detect the radical scavenging compounds within the eluent. Figure 3.1 (c) illustrates this setup. Multiplexed Mode of Operation The same Shimadzu HPLC system used for the conventional DPPH detection process was used in the AFT multiplex mode of operation. Since the flow rate through the DPPH reagent detector (peripheral port 3) was reduced, the volumetric flow of the DPPH reagent was reduced accordingly. Specifically, the eluent stream from peripheral port 2 had a flow rate of 0.47 mlmin -1, so the DPPH reagent set at 0.47 mlmin -1 also to maintain the 1:1, eluent stream:dpph reagent ratio. The combined eluent stream then entered a reaction coil (100 µl), which was maintained at a temperature of 60ºC within a column heater, and then entered the Shimadzu PDA detector set at a wavelength of 520 nm to detect the radical scavenging compounds within the eluent. Figure 3.2 illustrates this setup. Mass Spectrometry Mass spectrometry (MS) detection was conducted using electrospray ionisation in positive mode. Using Full Scan detection method, Total Ion Count (TIC) analysis was carried out on the TSQ Vantage mass spectrometer equipped with HESI II source under the following conditions: Vaporiser temperature 500 C, capillary temperature, 350 C; sheath gas set at a rate of 60 unites; auxiliary gas flow 40; sweep gas flow at 5 units, and spray voltage, 3.5 kv [43] Data Analysis Data analysis was undertaken using Origin and Microsoft Excel. The MS data was processed using LC Quan software (Thermo Fisher Scientific, San Jose, USA). 37

60 Figure 3.1 Illustration of conventional HPLC setup with: (a) MS detector, (b) UV- Vis detector and (c) DPPH detector. 38

61 Figure 3.2 Illustration of multiplexed HPLC setup with DPPH, UV-Vis and MS detectors. 39

62 3.3 Results and Discussion Conventional HPLC Multi-Detection The tobacco leaf extract and the Ristretto coffee samples were analysed using conventional HPLC with each of the following detectors; UV-Vis, DPPH and MS TIC. The chromatograms from each detector are shown in Figure 3.3. A new sample was prepared for each analysis mode. Each of the samples showed positive response to the DPPH reagent, but determining which compounds having the DPPH signal corresponded to the respective components in UV-Vis and MS chromatograms required that each be aligned manually a source of potential uncertainty Multiplexed HPLC A multiplexed HPLC analysis was carried out using an AFT column in PSF mode and set up as illustrated in Figure 3.2. This type of setup allowed the compounds that responded to DPPH to be easily matched up to the UV-Vis response and MS based on retention time and where a positive response was seen from the MS detector, the molecular mass of the peak was recorded. Table 3.1 lists the retention time of the DPPH peaks, and the response of such peaks in the UV-Vis and/or the MS detector, which thus provided the molecular mass. Tobacco Leaf Extract Figure 3.4 shows the chromatograms from each detector in the multiplexed detection assay of the tobacco leaf extract. From this data it is immediately apparent than one advantage of this approach to bio-discovery is that within the time required for a single injection three detection responses were obtained that yield information on the sample complexity. Using a conventional approach the three detectors would have to be connected in series, with the destructive detector last. However, in this instance two destructive detectors would be employed, which means at least two injections would be required. Post column flow stream splitting could have been used as an alternative, however, at the sacrifice of separation performance due to the additional dead volume, and the reduction in sensitivity that follows. AFT has been shown to improve separation performance and improve sensitivity, hence offering a significant advantage over that of the post column flow stream split [40,42]. 40

63 Absorbance (mau)/ Relative Abundance* Absorbance (mau)/ Relative Abundance* (a) Conventional HPLC - Tobacco UV (280 nm) DPPH (517 nm) MS - TIC (*Height Normalised) Time (min) (b) Conventional HPLC - Coffee UV (280 nm) DPPH (517 nm) MS - TIC (*Height Normalised) Time (min) Figure 3.3 Conventional HPLC multi-detection chromatograms (UV-Vis, DPPH, and MS TIC): (a) Tobacco and (b) Coffee. 41

64 Table 3.1 Detected DPPH peaks response to UV-Vis and MS. Tobacco DPPH Peak Response and Mass DPPH Peak Retention Time (min) DPPH Response UV-Vis Response MS Response Mass Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes No Yes No Yes Yes No No Yes No Yes Yes Yes Yes Yes No Yes Yes No Yes Coffee DPPH Peak Response and Mass DPPH Peak Retention Time (min) DPPH Response UV-Vis Response MS Response Mass Yes Yes Yes Yes Yes Yes Yes No No Yes No Yes Yes No No Yes No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Decaffeinated Coffee DPPH Peak Response and Mass DPPH Peak Retention Time (min) DPPH Response UV-Vis Response MS Response Mass Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes Yes Yes Yes

65 The chromatographic profile for the DPPH antioxidant detection assay of tobacco leaf extract, showed 13 well resolved peaks, 5 of which also gave a response in both the UV-Vis and MS detectors. Importantly, 8 peaks did not respond to either the UV- Vis at the wavelengths tested, or the MS when operated in positive ion mode. Two peaks gave no response to either the UV-Vis or the MS. Hence this clearly indicates the advantage of multiplexed detection processes for complete sample characterisation, where if it were not for the simultaneous use of the DPPH detector in multiplexed mode, such peaks would go undetected with the sole use of MS or UV-Vis. For clarity, peaks that were detected by all three detectors are indicated by the dotted red lines in Figure 3.4. The DPPH peaks that are detected by either only UV-Vis or MS are indicated by the blue dotted line and the green dotted line represents DPPH peaks that gave no UV-Vis and MS response. The majority of DPPH peaks, particularly those after 40 minutes showed a very strong response to the compounds reacting with the DPPH reagent, potentially indicating high antioxidant activity. However, these compounds gave little to no response in the UV- Vis and MS chromatograms. As the MS was operated in positive ion mode the DPPH peaks that gave no MS response may not ionize under the MS conditions in this study. Coffee Extract Coffee is known to have a high antioxidant content, with numerous components exhibiting response to the DPPH reagent [32,36,67]. In this study, the coffee analysis with the DPPH reagent resulted in 20 well resolved peaks, 13 of which also showed a response in the UV-Vis and MS detectors as indicated by red dotted lines in Figure 3.5. There are four regions which have the same chromatographic profile within each chromatogram, indicated by the red boxes. In these boxes, where UV- Vis and MS showed little response to these peaks, there was a strong response from the DPPH detector. Four components that responded to the DPPH reagent were not detected either by the UV-Vis or MS detectors (represented by the blue dotted lines) and three of the components that responded to the DPPH reagent gave no UV-Vis or MS response at all (represented by the green dotted line). The molecular mass of the components that responded to the MS are recorded in Table 3.1. The component that elutes at the 10 minute mark according to the DPPH assay resulted in a strong 43

66 response, with no response from the UV-Vis and with a very small response from the MS detector. Figure 3.4 Tobacco sample multiplexed chromatograms (a) UV-Vis 280 nm, (b) DPPH 520 nm, (c) MS - TIC. 44

67 Figure 3.5 Coffee sample multiplex chromatograms (a) UV-Vis 280 nm, (b) DPPH 520 nm, (c) MS - TIC. 45

68 An interesting and reoccurring facet of the analysis of coffee with DPPH antioxidant detection is that peak B in Figure 3.5(a), which corresponds almost perfectly with the elution of caffeine shows a very strong antioxidant response. Caffeine, however, is not an antioxidant and shows no response to the DPPH reagent. A casual analysis of this assay may result in an analyst mistakenly assigning this antioxidant response to caffeine as the retention time is almost perfect coincident with caffeine, rather, this antioxidant is a minor component that co-elutes with caffeine. The analysis of decaffeinated coffee, for example, illustrates this interesting facet of the separation. The decaffeinated coffee sample was analysed using exactly the same methodology that was employed for the caffeinated coffee, with multiplexed detection. The resulting chromatographic responses for each of the three detectors are shown in Figure 3.6. In total 18 DPPH peaks were recorded, 16 of which were also detected by the UV-Vis and MS detectors. One of the DPPH peaks also only gave a UV-Vis response and one DPPH peak gave no UV-Vis or MS response. Aside from a decrease in the signal intensities between the caffeinated and decaffeinated samples there is a very similar retention profile across all three detectors, except, however, the absence of the caffeine band in the UV-Vis and MS detection responses. The antioxidant responses at around the 10 minute mark in the DPPH detector, however, still shows the presence of components A and B that co-elute with the caffeine, showing clearly that caffeine was not responsible for the antioxidant behaviour, rather a co-eluting component at much lower concentration than caffeine. This was verified further by the injection of pure caffeine, which has strong response in the UV-Vis and MS detectors, but no response in the DPPH detector (see Figure 3.7). This nice example of sample and detection selectivity, and the confusion that co-eluting species can produce in real complex samples shows the importance of sample selectivity with multiple detectors and indicates that for complete sample analysis more powerful separations are required in order to resolve more components, some of which may have important biological function. In this case, a minor component in the coffee sample with strong antioxidant activity co-eluted with the major component (caffeine). Extraction of this low concentration component from the coffee sample would therefore require better separation, perhaps multidimensional. 46

69 Figure 3.6 Decaffeinated sample multiplexed chromatograms (a) UV-Vis 280 nm, (b) DPPH 520 nm, (c) MS - TIC. 47

70 Coffee Coffee Decaffeinated Coffee Decaffeinated Coffee Caffeine Figure 3.7 Caffeine compound UV response: (a) Coffee (b) Decaffeinated coffee; Caffeine compound DPPH response: (c) Coffee (d) Decaffeinated coffee (e) Caffeine standard. 3.4 Conclusion AFT columns were designed to overcome band broadening issues associated with column bed heterogeneity, especially the wall effect, and to increase efficiency in terms of theoretical plate count and increase sensitivity. The multi-port end fitting design of the AFT column offers the added benefit of providing opportunities for multiplexing detection processes, yielding detailed sample information and absolute reliability in the assignment of components between each detection mode. Multiplexed detection with AFT columns provided a large amount of sample information, three detectors operated simultaneously within the run time of one injection and allowed for the exact match of retention time of peaks within each detection mode. Furthermore, two of these detectors were destructive detectors. 48

71 CHAPTER FOUR Selectivity in 2DHPLC: A Comparison Between Columns from Different Manufacturers 49

72 4.1 Introduction HPLC is an analytical separation technique based on the conditional interactions between molecules of a sample, the stationary phase, and the mobile phase [2]. Selectivity in separation can be achieved through application on various combinations of different stationary phases and solvents. It is because of this that the chromatographer has a great deal of control in gaining resolution, even more so when implementing multidimensional protocols, since two different types of selective interactions are utilised in the one separation process [2]. Along with the selection of the most appropriate combination of stationary phases that yield the best separation performance for the target sample analytes, the analyst is faced also with the selection of phases from different manufacturers. Many manufacturers produce effectively the same phase, usually on different base material, yet for all intents and purposes, providing the same retention behaviour. The question remains, however, as to how similar these same phases are in the separation of complex samples, where the constituent sample diversity is broad and perhaps largely unknown? The aim of this study is to investigate selectivity differences for a selection of the same phases from two different manufacturers when applied to the analysis of complex samples derived from natural origin. In this thesis, only one set of columns was tested, those sourced from ThermoFisher Scientific, and since in prior work, not undertaken as part of this doctoral study, columns manufactured by Phenomenex were tested [15]. Those columns were tested using the exact 2DHPLC system as used in the present study. The data here is compared to that prior work to account for any selective differences between column manufacturers as the following chapters shift into the utilisation of AFT columns from the only available source, ThermoFisher Scientific. 50

73 4.2 Experimental Chemicals Mobile phase solvents were HPLC grade. Methanol was purchased from Merck (Kilsyth, Victoria, Australia) and Ultrapure Milli-Q water (18.2 MΩ) was prepared in-house and filtered through a 0.2 µm filter Sample Preparation A sample of espresso coffee (Nestlé Nespresso Ristretto, Sydney, NSW Australia) was prepared fresh prior to analysis by extraction using an espresso coffee machine (Dè Longhi - Model EN 95.S), diluted four fold, then filtered through 0.45 µm pore filter Instrumentation and Chromatographic Conditions Instrumentation All sample analyses were undertaken on a Waters 600E Multi Solvent Delivery LC System, equipped with a Waters 717 plus auto injector, two - Waters 600E pumps, two - Waters 2487 series UV-Vis detectors (set at 280 nm) and two - Waters 600E system controllers. Separation dimensions were interfaced via a 10-port, two position switching valve that was electrically actuated. Chromatographic Columns and Separation Uni-Dimensional HPLC Two types of stationary phases were employed for uni-dimensional separations; these were Cyano (CN) and Pentafluorophenyl (PFP) phases. Two sets of these columns were utilised, one from ThermoFisher Scientific (Runcorn, United Kingdom), the other from Phenomenex (Lane Cove, NSW Australia). The columns sourced from ThermoFisher Scientific were the Hypersil GOLD Cyano (CN) and Hypersil GOLD Pentafluorophenyl (PFP) both in 100 x 4.6 mm, format, packed with 5 µm particles. The columns sourced from Phenomenex were the Luna CN and the Luna PFP, both in 150 x 4.6 mm format and packed with 5 µm particles. Two differences are apparent between the columns, the first being the column length. The 51

74 ThermoFisher Scientific columns were shorter, so as to speed the analytical testing protocol. To compensate for this separation conditions across the two-dimensional domain were adjusted accordingly. This should be of minor consequence since selectivity is dependent on the nature of the surface, not so on the column dimensions. The second difference is the pore size. The particles in the Hypersil columns had 175 Å pores, while those in the Luna columns were 100 Å. This difference is more important and potentially could influence selectivity. However, the purpose of the study was to evaluate if separation conditions did change as a function of the manufacturer supplying the column. However, in that regard, the test was not to determine which performance was the best, rather whether or not there were differences, and if so, how substantial. Regardless of the column, uni-dimensional separations were undertaken in gradient elution with an initial mobile phase composition of 100% water running to a final mobile phase of 100% methanol at a rate of 20 %min -1, then held at 100% methanol for 4 minutes and returned to initial conditions in 1 minute, giving a total run time of 10 minutes. Flow-rate was set at 2.0 mlmin -1 and injection volume was 100 µl. Columns were maintained at 45 C using a column heater. Two-Dimensional HPLC In two-dimensional separations the analysis of the coffee was undertaken only on the ThermoFisher Scientific columns; the data was compared to prior studies on Phenomenex columns undertaken by Mnatsakanyan et al. [15]. The columns utilized in the first dimension were limited to the Hypersil GOLD CN and PFP columns, in formats 100 x 4.6 mm, 5 µm particles. The second dimension consisted of a ThermoFisher Scientific Hypersil GOLD C Å (100 x 4.6 mm, 5 µm P d ) column. Gradient elution was undertaken in both dimensions. In the first dimension, gradient elution began with an initial mobile phase of 100% water running to a final mobile phase of 100% methanol, at a rate of 20 %min -1, then held at 100% methanol for 4 minutes and returned to initial conditions in 1 minute. Flow-rate was set at 2.0 mlmin -1 and injection volume was 100 µl into the first dimension. The 2D analyses of the samples were undertaken using a comprehensive incremental heart-cutting approach [13,68] involving the transfer of a 200 µl aliquot sampled from the first 52

75 dimension, into the second dimension, followed by elution in the second dimension. The second dimension, consisting of a C18 column, also operated in gradient elution mode running from 100% water to 100% methanol at a rate of 20 %min -1, held for 4 minutes at 100% methanol and returned to initial conditions in 1 minute at a flow rate of 2.0 mlmin -1. These elution conditions, in both dimensions were repeated (following reinjection of the sample). Each subsequent injection process sampled the next 0.4 minute interval in the first dimension, starting at 0.8 minutes in the first dimension, being completed after a total of 16 cuts. Both dimensions were temperature regulated at 45 C using a column heater Data Analysis The two-dimensional chromatographic data was analysed using Mathematica (with home-built programs) [98]. Two-dimensional chromatographic surface plots were produced using this software. To assess and gain a statistical measure of the selectivity differences between systems a geometric approach to factor analysis (GAFA) was used on the 2D data for the measure of the number of detected peaks (n), the spreading angle (β), the correlation, the practical peak capacity (n Tƒ ) and the per cent usage of separation space [16]. The orthogonality measured in this study was based on data collected with UV detection at 280 nm, and thus compounds without appreciable UV response at that wavelength were not included in the orthogonality aspect. 4.3 Results and Discussion Prior work undertaken by Mnatsakanyan et al. [15] tested several stationary and mobile phase combinations for the two-dimensional analysis of coffee, which included propyl Cyano (CN) and the Pentafluorophenyl (PFP) phases, in the first dimension and a C18 column in the second dimension. Each of these stationary phases employed the same Luna 5 m support, from Phenomenex [15]. In the present study similar stationary phases were employed, except purchased from ThermoFisher Scientific and these were based on the Hypersil range, also 5 m particles. The purpose of comparing these different manufacturers in this study is simply because AFT columns, which are to be utilised in Chapter 6 of this thesis are only available from ThermoFisher Scientific. So an additional focus of this chapter is 53

76 to contrast the selectivity differences with respect to the manufacture with the aim of determining whether users could expect to obtain similar separation outcomes irrespective of the column type, if having at least the same chemistry. In order to assess differences in selectivity between different manufacturers the twodimensional retention behaviour of the components in an espresso coffee extract were evaluated. This sample was chosen because of its complexity and diverse range of components [66]. Prior to assessing the two-dimensional separation performance for each set of columns, uni-dimensional separations were undertaken. The chromatograms in Figure 4.1 illustrate these uni-dimensional separations for each of the CN and PFP columns from both column manufacturers. Both sets of columns were operated under the same chromatographic conditions (see Section Chromatographic Columns and Separation Uni-Dimensional HPLC). The selectivity difference between the CN and PFP is apparent, as expected there were substantial differences in the selectivity between the CN and PFP phases for each respective manufacturer; specifically, there was greater retentivity on the PFP phases than the CN phases. Exact differences were difficult to quantify since the peak capacity of all columns was exceeded, leading to a large degree of co-elution. 54

77 Absorbance (AU) Absorbance (AU) Absorbance (AU) Absorbance (AU) (a) ThermoFisher Scientific CN CN (b) ThermoFisher Scientific Scientific PFP PFP Time (min) (c) Phenomenex CN CN Time (min) (d) 4.0 Phenomenex 3.5 PFP PFP Time (min) Time (min) Figure 4.1 Uni-dimensional separations of Ristretto café espresso on ThermoFisher Scientific columns: (a) CN and (b) PFP; Uni-dimensional separations of Ristretto café espresso on Phenomenex columns: (c) CN and (d) PFP. All chromatographic conditions are 20 %min -1 gradient, from water to methanol, then held for 4 minutes at 100% methanol and returned to initial conditions in 1 minute (total run time 10 minutes). ( indicates caffeine peak) Although there is a general similarity between sample profiles on each respective phase between the two manufacturers, i.e., similar bimodal distributions of compounds can be seen, separated by the dotted lines in Figure 4.1, there were also important differences. In particular, caffeine was more strongly retained on the Phenomenex columns than the ThermoFisher Scientific columns, especially in the case of the CN column. Also, there was a greater apparent degree of separation power on the Phenomenex columns compared to the ThermoFisher Scientific column, likely due to Phenomenex columns being 30% longer. The 1DHPLC analysis of coffee was also carried out by Mnatsakanyan et al. [15] on the Phenomenex Luna CN and PFP columns (Figure 4.2) under a slower gradient of 10 %min -1 from 100% water to 100% methanol at a flow rate of 1.0 mlmin -1. Their 1DHPLC work on the Phenomenex columns was carried out more than 3 years ago 55

78 and showed similar retention behaviour compared to the 1DHPLC analysis carried out on the Phenomenex columns in this study, indicating that the retention behaviour, at least on the Phenomenex columns has been sustained over a reasonable period. Mnatsakanyan et al. also achieved greater resolution due to slower gradient conditions. (a) Phenomenex CN (b) Phenomenex PFP Figure 4.2 Uni-dimensional separations of Ristretto café espresso on Phenomenex columns at a 10 %min -1 gradient, from water to methanol, held for 8 minutes at 100% methanol and returned to initial conditions in 2 minute (total run time 20 minutes): (a) CN and (b) PFP. Figures from Mnatsakanyan et al. [15]. ( indicates caffeine peak) In order to gain a greater understanding of the retention behaviour for these sets of columns in analyses of complex samples of great diversity we undertook more detailed tests using 2DHPLC. In these studies, the CN and PFP columns were used in the first dimension, with a C18 column in the second dimension. The surface plots in Figure 4.3 show the distribution of the components across the respective separation spaces when the ThermoFisher Scientific columns were used, compared to the surface plots when the Phenomenex columns were used by Mnatsakanyan et al. [15] (Figure 4.4). Since a common C18 column was used in the second dimension in each respective test (ThermoFisher Scientific C18 in this study and a Phenomenex C18 in Mnatsakanyan et al. s study) the peak displacement is reflected by the nature of the first dimension. The detected peak location is represented by the dots, based on the peak maxima as illustrated in Figure 4.5. These 2D chromatograms show significant differences in retention behaviour of the extracted compounds between the different stationary phases (i.e., CN v PFP) for the coffee sample. However, it is also interesting to note that the 2D chromatographic profiles on the same phase from 56

79 C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) different manufacturers (i.e., CN v CN and PFP v PFP) are also quite different, in particular the PFP phase from both manufacturers. This kind of column/sample analysis may give a two-dimensional characterisation on the nature of the compounds present in the sample. (a) (b) CN Dimension Retention Time (min) PFP Dimension Retention Time (min) Figure 4.3 Two-dimensional separations of Ristretto café espresso. (a) ThermoFisher Scientific - 1DCN/2DC18, (b) ThermoFisher Scientific - 1DPFP/2DC18, in both dimensions mobile phase ran at a 20 %min -1 gradient, from aqueous to organic solvent at 2 mlmin -1. ( indicates caffeine peak) (a) (b) CN Dimension Retention Time (min) PFP Dimension Retention Time (min) Figure 4.4 Two-dimensional separations of Ristretto café espresso by Mnatsakanyan et al. [15] (a) Phenomenex - 1DCN/2DC18, (b) Phenomenex - 1DPFP/2DC18, in both dimensions mobile phase ran at a 10 %min -1 gradient, from aqueous to organic solvent at 1 mlmin -1. ( indicates caffeine peak) 57

80 C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) (a) (b) Q2 Q3 Q2 Q3 Q1 Q4 Q1 Q4 CN Dimension Retention Time (min) PFP Dimension Retention Time (min) (c) (d) CN Dimension Retention Time (min) PFP Dimension Retention Time (min) Figure 4.5 Scatter plots for the 2DHPLC separations with ThermoFisher Scientific columns (a) CN, (b) PFP; Phenomenex columns by Mnatsakanyan et al. [15] (c) CN, (d) PFP as the first dimension. The quadrants are defined by the dashed lines. To determine the correlation between the dimensions in a 2D system a geometric approach to factor analysis (GAFA) was carried out, the results of which are presented in Table 4.1. The GAFA data collected in the study by Mnatsakanyan et al. [15] has also been tabulated for comparison. The two-dimensional retention data was assessed qualitatively and quantitatively to determine the separation power afforded by each system for each sample. GAFA was applied to measure the correlation between each dimension by determining the number of peaks (n), the spreading angle (β), the practical peak capacity (n Tƒ ) and per cent usage of the separation space [16]. Each two-dimensional surface plot has been sectioned into quadrants (Q) as indicated by the dotted lines on the 2D chromatograms. Table 4.1 details the selectivity changes that have occurred between each 2DHPLC system. Overall, some systems show a lower correlation to the C18 phase than other systems, but these are dependent on which quadrant of the separation is compared. In some quadrants correlation is high, others low. Hence depending on the type of 58

81 compounds being separated some systems would out-perform regions of other systems or the overall system. Thus, each quadrant of each system must be individually assessed by GAFA to detail the selectivity changes and differences between each phase and manufacturers. Table 4.1 GAFA calculations for the 2DHPLC separation and in each of the quadrants. System Detected Peaks (n) Correlation Spreading Angle Peak Capacity Usage (%) ThermoFisherScientific Columns 1. 1DCN/2DC18 (H 2 O:MeOH) Q Q Q Q4 N/A N/A N/A N/A 2. 1DPFP/2DC18 (H 2 O:MeOH) Q Q Q Q4 N/A N/A N/A N/A Phenomenex Columns* 1. 1DCN/2DC18 (H 2 O:MeOH) Q Q Q Q4 N/A N/A N/A N/A 2. 1DPFP/2DC18 (H 2 O:MeOH) Q Q Q Q4 N/A N/A N/A N/A *Mnatsakanyan, et al. [15] 59

82 4.3.1 Qualitative Assessment of the Selectivity Changes Overall System Comparison A Ristretto café espresso coffee sample was analysed on two 2DHPLC systems by two different column manufacturers. The system that resolved the most number of peaks for the coffee sample was the CN/C18 system for both column manufacturers. However, the number of overall peaks detected by CN/C18 columns supplied by ThermoFisher Scientific is about 20% greater than the Phenomenex CN/C18. The peak counts obtained on the PFP/C18 2D system were similar irrespective of the manufacturer. Regardless of peak count the GAFA results for the CN/C18 columns from both manufacturers are slightly different, where the Phenomenex CN column had a higher correlation to the C18 phase than the ThermoFisher Scientific CN column. The CN/C18 2D system that incorporated the Phenomenex columns utilised 56% of the 2D peak capacity as opposed to the ThermoFisher Scientific CN/C18 2D system that utilised 71% of the peak capacity. The PFP/C18 systems from each manufacturer resulted in quite different twodimensional retention behaviour compared to the CN/C18 systems. The degree of correlation between the ThermoFisher Scientific PFP and C18 column was higher than the CN/C18 system; the correlation value was 0.62, spreading angle of 52 and peak capacity usage 65%. Likewise, the degree of correlation between the PFP and C18 Phenomenex columns was higher than between the CN and C18 Phenomenex columns. The correlation was 0.95, resulting in a very small spreading angle - just 19, and the peak capacity usage was only 29%. This was a very different outcome compared to the ThermoFisher Scientific PFP/C18 2D system. Overall, based on the GAFA results the comparison between manufacturers showed that the ThermoFisher Scientific columns (CN and PFP) offered more divergent retention behaviour relative to the C18 column than the Phenomenex CN and PFP columns. The scatter plots in Figure 4.5 show the similarity between chromatographic profiles of the CN/C18 systems from both manufacturers, while a significant difference is seen between manufacturers for the PFP/C18 systems, 60

83 C18 Dimension Retention Time (min) Absorbance (AU) visually supporting the GAFA data results. More detailed analysis of the retention behaviour, based on retention in each of the four quadrants is discussed below. Cyano Phase Quadrant 1: In this quadrant there is little retention on the CN phase indicative of compounds that have minimal non-resonance π-π bonding or dipole-dipole interactions with the CN phase. However, these compounds when transported to the second dimension showed a substantial range in retention across the C18 phase. This behaviour indicates that these compounds have a range in hydrophobicity, but limited functionality to be able to take advantage of π-π bonding. This type of separation and selectivity has also been observed in the study of M. Mnatsakanyan et al. [15] where further mass spectral (MS) analysis was carried to confirm the nature of the compounds. The MS analysis showed that compounds in this region were hydroxylated, including compounds that were of low molecular weight carboxylic or phenolic acids. The MS analysis also verified the presence of monomeric flavan-3- ols. In this quadrant the same number of peaks was detected for both manufacturers, although in this quadrant, the Phenomenex columns showed a higher degree of separation divergence than the ThermoFisher Scientific columns. Figure 4.6 shows the second dimension separations from 2 heart cut sections from the CN phase in the first dimension for the ThermoFisher Scientific system. Note the time axis is limited to the period defined by the dimensions of Q ThermoFisher Scientific 2nd Dimension (C18) Cut Cut Time (min) CN Dimension Retention Time (min) Figure 4.6 Heart-cut segments from CN phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. 61

84 Quadrant 2: In this quadrant from both manufacturers, a greater range in retention across the CN and C18 phase is seen. For the ThermoFisher Scientific system the correlation in Q2 was 0.25, the spreading angle was 75 and the peak capacity usage was 87%. Although, the number of peaks detected in this quadrant for the Phenomenex columns is half that observed on the ThermoFisher Scientific columns, it too produced fairly impressive selectivity differences between the CN and C18 phases with a correlation of 0.46, spreading angle of 63 and peak capacity usage of 76%. The range in retention on the CN phase suggests these compounds have selective interactions, including non-resonance π-π bonding with the CN phase; and the second dimension separation shows that these compounds also range in hydrophobicity and have non-polar substituents [15]. Overall, in Q2 the ThermoFisher Scientific column set showed superior performance to the Phenomenex column set. Some examples of the separation power afforded by 2DHPLC are illustrated in Figure 4.7, which shows the second dimension separation (i.e., C18 phase) following heart cuts from the first dimension (CN phase) at time intervals between minutes from the ThermoFisher Scientific system and minutes from the Phenomenex system [15] (Figure 4.7). The minute cut (200 µl from the first dimension at 1.0 mlmin -1 ) from Mnatsakanyan et al. s [15] study is approximately equivalent to the region of the ThermoFisher Scientific system at the minute cut (200 µl from the first dimension at 2.0 mlmin -1 ), at least with respect to the solvent composition. Here, the chromatographic profiles show similar selective behaviour with respect to each system conditions, although the profiles in the second dimension are significantly different. Nevertheless, the wide retention time distribution across the C18 dimensions on both systems (ThermoFisher Scientific and Phenomenex) shows the multitude of components that co-elute in the 200 µl region on the CN phases in the first dimension. Thus, illustrating the power of twodimensional separations. One very important difference between the different brand CN columns is that caffeine elutes in Q2, on the Thermo system, and Q3 on the Phenomenex system. Quadrant 3: Compounds in this quadrant are late eluting compounds in both dimensions, which stipulate that these compounds have a higher affinity to the CN phase with strong π-π bonds and dipole-dipole interactions. They are also non-polar 62

85 compounds, since they are strongly retained on C18. In Q3, when using the Phenomenex column set, it was this region that resulted in the separation of the most number of components, whilst this was not the case for the ThermoFisher Scientific system, since in general, retention was less substantial on the ThermoFisher Scientific columns, perhaps as a result of the larger pore size, hence a lower column surface area. Figure 4.8 illustrates a second dimension separation for a heart cut made at minutes from the first (CN) dimension. The power of 2DHPLC is evident here, where across the CN dimension in the space of 200 µl only 1 large peak is apparent, however across the C18 dimension we can see at least 5 peaks that were co-eluting within the 200 µl region from the first dimension. In Q3, 49 peaks were detected for the ThermoFisher Scientific system, with a correlation of 0.45, spreading angle of 63 and peak capacity usage of 71%. The separations undertaken by Mnatsakanyan et al. [15] showed similar 2D profiles in Q3 to the separations reported here, however, in their study more peaks were detected (62), although the correlation was higher (0.80), which actually resulted in a lower peak capacity utilisation (50%, compared to 71% for the ThermoFisher Scientific column set). The major component - caffeine was detected in Q3 when Phenomenex columns were used, while when ThermoFisher Scientific columns were used, caffeine was observed to elute in Q2. Interestingly, the retention of caffeine on the C18 columns was similar, irrespective of the manufacturer: elution occurring at approximately the same solvent composition. 63

86 C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) Absorbance (AU) (a) Time (min) CN Dimension Retention Time (min) (b) CN Dimension Retention Time (min) Figure 4.7 Heart-cut segment from CN phase on C18 separation of Ristretto café espresso: (a) ThermoFisher Scientific columns Cut time between minutes and (b) Phenomenex columns Cut time between minutes (image from M. Mnatsakanyan et al. [15]). ( indicates caffeine peak) 64

87 C18 Dimension Retention Time (min) Absorbance (AU) Time (min) CN Dimension Retention Time (min) Figure 4.8 Heart-cut segment from CN phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. Quadrant 4: No peaks were detected for this quadrant from both manufacturers. Pentafluorophenyl Phase Quadrant 1: In this quadrant for both manufacturers there was a greater retention range of compounds across the PFP phase compared to the CN phase, indicating perhaps the greater significance of resonance π-π bonding compared to the nonresonance π-π bonding on the CN phase. Hydrogen bonding may also have played a more substantial role on the PFP phase. Figure 4.9 illustrates the second dimension (C18) separation for the cut made at 0.8 minutes on the first dimension (PFP). Perhaps compounds in this region are low molecular weight aromatic species, which is supported by the mass spectral data derived by Mnatsakanyan et al. [15]. They found that the compounds eluting in Q1 were essentially low molecular weight hydrophobic alkaloids, moderately polar, that contained aromatic rings, such as, low molecular weight carboxcylic acids, which interact with the PFP surface via π-π bonding, less so with the non-resonance nitrile functionality of the CN phase. The GAFA data shows retention behaviour in this quadrant was quite divergent, effectively orthogonal irrespective of the manufacturer The Phenomenex system yielded a correlation of 0.08, spreading angle of 85 and peak capacity usage of 96%. The ThermoFisher Scientific system obtained a correlation value of 0.18, spreading 65

88 C18 Dimension Retention Time (min) Absorbance (AU) angle of 79 and peak capacity of 91%. In general, very similar outcomes in terms of the separation power, irrespective of the manufacturer Time (min) PFP Dimension Retention Time (min) Figure 4.9 Heart-cut segment from PFP phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut time between minutes. Quadrant 2: Approximately 45% of the total number of peaks separated on the ThermoFisher Scientific system eluted in this quadrant, compared to just 5% on the Phenomenex system. The GAFA outputs for the ThermoFisher Scientific system showed that the PFP and C18 phases displayed almost orthogonal retention behaviour for the compounds that eluted in this region of the 2D separation space (too few peaks eluted in this quadrant when using the Phenomenex system to provide a meaningful comparison). This retention behaviour is quite beneficial since the separation space is maximised. This retention behaviour is indicative of compounds that interact via resonance π-π bonding and are generally quite non-polar since they are moderate to strongly retained on the C18 phase in the second dimension. Quadrant 3: In this quadrant, for both PFP systems the components were displaced along the main diagonal and consequently there was high correlation between the PFP and C18 phases, with less resolving power than in Quadrants 1 and 2. Mnatsakanyan et al. [15] deduced that these compounds were generally high molecular weight sugars. Figure 4.10 illustrates the overlay of a series of separations in the second dimension (C18 phase) in Q3 for heart-cut sections from the first 66

89 C18 Dimension Retention Time (min) Absorbance (AU) dimension between minutes, a number of peaks can be seen in each separation that were co-eluting in the first dimension ThermoFisher Scientific 2nd Dimension (C18) Cut Cut Cut Cut Cut Time (min) PFP Dimension Retention Time (min) Figure 4.10 Heart-cut segment from PFP phase on C18 separation of Ristretto café espresso on ThermoFisher Scientific columns Cut times between minutes. Quadrant 4: No peaks were detected in this quadrant from either manufacturer. Some solvent breakthrough peaks were observed that resulted from the transport of solutes in large volume plugs that had strong solvating power. Hence, it was difficult for these compounds to be retained in the second dimension. This data was not analysed as it is likely to provide erroneous information. 4.4 Conclusion A detailed study comparing the selectivity differences on column sets obtained from two different manufacturers was undertaken. The purpose of doing so was to gauge whether the separations of Mnatsakanyan et al. [15] in prior work undertaken on Phenomenex columns would be valid on ThermoFisher Scientific columns of the same phase. This study was required since in later chapters in this thesis a new column design was tested in two-dimensional HPLC, and these columns were only available from ThermoFisher Scientific. The results showed that even columns of the same phase displayed substantial differences in retention behaviour and hence, caution should be paid to pre-determining the separation outcomes if users are to change manufacturer supplies of the same phase. 67

90 CHAPTER FIVE Optimisation of a 2DHPLC System for Enhancing the Separation and Subsequent Analysis of Tobacco 68

91 5.1 Introduction Plant derived organic samples such as tobacco leaves are comprised of literally hundreds of diverse compounds [57,58]. Finding a suitable characterization method for analysis of these types of complex samples can be arduous because of the complexity of the sample, hence multidimensional separation strategies are often invoked. The separation power and theoretical peak capacity afforded by twodimensional High Performance Liquid Chromatography (2DHPLC) has the potential to separate and characterise these complex samples of natural origin, since the separation space is expanded across multiple separation steps [7,11]. Effectively the practical peak capacity is increased according to product rule (Equation 1.7) taking into consideration the unavailable separation space due to correlation between the dimensions (see Equation 1.7a). Therefore, in order to maximise the separation power, each of these dimensions should provide effectively orthogonal separation behaviour, ensuring that the 2D space is utilised as much as possible [11,13,14,16]. Under the conditions whereby the peak capacity has been expanded due to the separation being undertaken across two dimensions the probability of two components having exactly the same retention time is very much diminished, becoming less probable as the correlation between each dimension decreases. Ideally, the separation could yield a chemical signature. Hence, it is a worthwhile task to undertake appropriate optimisation of the separation performance of a 2DHPLC separation system, paying particular attention to the differences in selectivity between each dimension. A geometrical approach to factor analysis (GAFA) is a useful technique to assess the separation power, or, perhaps more to the point, the effective utilisation of the separation space and the degree of dimension correlation in two-dimensional systems [16]. Thus, for an efficient separation power the two dimensions must be minimally correlated, achievable when selectivity between the two dimensions is most different, i.e., different stationary and mobile phase combinations that can exploit different sample attributes [11]. A more detailed discussion concerning GAFA was presented in Chapter 1. The complex nature of the sample being subjected to analysis using 2DHPLC makes it difficult to assess accurately the separation power and separation differences 69

92 between the dimensions: choosing a suitable set of standard reference compounds demands a detailed understanding of the types of compounds found within the sample. Then these compounds, or model sets of these compounds, need to be available commercially. Often a very limited set of compounds can be found and this offers limited applicability to the actual sample [13]. Perhaps the best approach to measuring the power of the two dimensional separation is to employ the sample itself, and use effectively selective detection processes to monitor changes in component displacements observed using different two-dimensional systems. That is, changes made in the selectivity of the first dimension can be accessed using a constant second dimension. Effectively, the second dimension serves as a selectivity detector [15]. This provides a convenient methodology to track the changes in the separation power in a two-dimensional system, albeit a system in which the second dimension has been defined from the outset. Defining the nature of the second dimension could be construed as a limiting factor in the design of the two-dimensional system, however, there are key requirements of the second dimension that must be meet in order for the process to function effectively: These include, (1) solvent compatibility between dimensions, and (2) the second dimension should be more retentive (in general) than the first dimension. Point (2) is particularly important as the stationary phase in the second dimension must be able to effectively entice solutes from what may be a thermodynamically strong first dimension solvent environment. Hence retention would be more problematic if the second dimension was less retentive than the first. There is a large list of stationary phases from various column manufacturers that broaden the selectivity choice for selective sample analysis. In this study a number of stationary phases were employed, Octadecyl (C18) was utilised in the second dimension and Cyano (CN), Pentafluorophenyl (PFP), PESC18 1 and Hypercarb were employed in the first dimension. The C18 stationary phase is a commonly used column in reversed-phase chromatography consisting of a hydrophilic chain of 18 1 The PESC18 column was not commercially available at the time of production of this thesis and there was no literature available on the nature of its functionality. The column was donated to us specifically for this project from ThermoFisher Scientific (Manor Park, Tudor Rd, Runcorn, Cheshire, United Kingdom). 70

93 hydrocarbons bonded to silica and is a member of a series of stationary phases known as alkyl bonded phases in which there are variations in the length of the alkyl chain. Another popular member is, for example, the Octyl (C8) phase, which is less retentive than the C18. Their chemistry involves very strong dispersive forces (hydrophobic) and moderate ionic bonding, providing no dipole, π-π bonding and limited shape selectivity [21,22]. Cyano (CN) columns are packed with nitrile bonded silica particles. The nitrile group is typically tethered to the silica surface using a short (often 3-carbon member) alkyl chain. This phase is strongly dipole and ionic, and has moderate to low hydrophobicity, allowing the rapid elution of hydrophobic compounds that have limited functional groups. Because of the carbon nitrogen triple bond, the CN column also enables weak non-resonance π-π bonding. Cyano columns can also be used in both reversed phase or normal phase chromatography [21,23] and nowadays fall into the category of hydrophilic interaction liquid chromatography (HILIC) phases. Pentafluorophenyl (PFP) stationary phases consist of the PFP moiety attached to the silica surface by a short alkyl chain, typically 3-member carbon. This type of surface is rather complex; highly polar bonds are present between the fluorine atoms and the aromatic ring, the ring structure itself is largely non-polar, but offers the opportunity for resonance π-π bonding, which is further influenced by the orientation of the ring to the surface of the silica [99-101]. The combination of the PFP moiety on the silica base provides selectivity for the analysis of difficult to separate halogenated compounds and also non-halogenated polar compounds, including compounds that have hydroxyl, carboxyl and/or nitro groups [102]. Figure 5.1 is a cartoon drawing that illustrates the phases of C18, CN and PFP. 71

94 (a) (b) (c) Figure 5.1 Molecular Structure of a silica bonded with (a) C18, (b) CN and (c) PFP. The Hypercarb stationary phase offers quite different retention properties to the C18 phase. While being extremely non-polar, the surface is far more two-dimensional, offering little opportunity for partitioning type retention behaviour that can be obtained on bonded phase stationary phases. The Hypercarb stationary phase is therefore an adsorption phase, which because of its design also provides strong electrostatic retention mechanisms, compared to other conventional liquid chromatography columns like the CN or C18 columns. The Hypercarb stationary phase is prepared from porous graphite, hence the strong electrostatic interactions and the non-polar nature of the surface [25]. On a molecular level, the porous graphite is made up of carbon atoms, hexagonally arranged as sheets on the surface. It is also highly crystalline and has no functional groups present. It is a surface that offers strong retentivity for non-polar analytes, similar to a strongly retentive alkylbonded silica. However, the chemical surface properties of the Hypercarb column is stereo selective and has the capability of separating geometric isomers, diastreomers and other similar compounds with strong π-π bonding, providing retention and separation uniqueness for very polar and closely related compounds [24,25]. Retention is also dependent on the solute molecular area in contact with the graphitic surface. Flatter, larger compounds are able to align closer to the surface of the graphite thereby increasing the bonding area and thus retention. Another chromatographic benefit of the Hypercarb stationary phase is that it possesses great chemical stability [24]. 72

95 (a) (b) Figure 5.2 Surface comparison of (a) C18 and (b) Hypercarb column [25]. Each of these stationary phases were chosen as a first dimension because their retention mechanisms would potentially exploit different sample attributes, while at the same time offering different retention behaviour to the C18 phase. The solvent systems used in this study were water/methanol and water/acetonitrile mixtures in gradient elution; application was therefore restricted to RP-RP 2DHPLC. Methanol and acetonitrile were chosen as the mobile phases because these solvents also offer scope for selectivity. Both solvents have approximately the same polarity, but the carbon-nitrogen triple bond present in methyl cyanide (acetonitrile) could suppress the non-resonance π-π bonding of stationary phases, or even enhance the separation of π-π bonding solutes on stationary phases that are otherwise inert to these sample attributes. 5.2 Experimental Chemicals Mobile phase solvents were HPLC grade. Methanol and acetonitrile were purchased from Merck (Kilsyth, Victoria, Australia) and Ultrapure Milli-Q water (18.2 MΩ) was prepared in-house and filtered through a 0.2 µm filter. 73

96 5.2.2 Sample Preparation Tobacco leaf extract was prepared from the same stock of loose leaf tobacco (White Ox) using the same extraction method used in Chapter 2, Section Sample Preparation Instrumentation and Chromatographic Conditions Instrumentation All sample analyses were undertaken on the same Waters 600E Multi Solvent Delivery LC System used in Chapter 4, Section Instrumentation and Chromatographic Conditions - Instrumentation. Chromatographic Columns The 2DHPLC analysis consisted of different stationary and mobile combinations. The columns utilized in the first dimension were: 1. Hypersil GOLD CN 175 Å (100 x 4.6 mm, 5 µm P d ) 2. PESC Å (150 x 4.6 mm, 5 µm P d ) 3. Hypersil GOLD PFP 175 Å (100 x 4.6 mm, 5 µm P d ) 4. Hypercarb 250 Å (100 x 4.6 mm, 5 µm P d ) The second dimension comprised a Hypersil GOLD C Å (100 x 4.6 mm, 5 µm P d ) column. All columns were purchased from ThermoFisher Scientific (Manor Park, Tudor Rd, Runcorn, Cheshire, United Kingdom). Chromatographic Separation Gradient elution was undertaken in both dimensions. When columns 1, 2 and 3 were used as the first dimension, gradient elution began with initial mobile phase of 100% water running to a final mobile phase of 100% organic solvent (methanol or acetonitrile), at a rate of 20 %min -1, then held at 100% organic solvent for 4 minutes and returned to initial conditions in 1 minute. Flow-rate was set at 2.0 mlmin -1 and injection volume was 100 µl into the first dimension. The 2D analyses of the samples were undertaken using a comprehensive incremental heart-cutting approach 74

97 [13,68] involving the transfer of a 200 µl aliquot sampled from the first dimension, into the second dimension, followed by elution in the second dimension. The second dimension consisted of a C18 column, also undertaking gradient elution from 100% water to 100% organic solvent (methanol or acetonitrile) at a rate of 20 %min -1 with a flow rate of 2.0 mlmin -1, where then mobile phase was held at 100% organic solvent for 4 minutes and returned to initial conditions in 1 minute. This process was repeated (following reinjection of the sample) at every 0.2 minute interval across the first dimension, starting at 0.6 minutes in the first dimension. When the Hypercarb column (column 4) was used as the first dimension in the 2DHPLC analysis, the initial gradient condition on the first dimension began at 80/20 and ended at 0/100 (water/organic solvent). The second dimension conditions (C18) remained the same as when columns 1, 2 and 3 where employed. Hence in all experiments the second dimension functioned in exactly the same manner. In that way, changes in the two-dimensional separation reflect the change in selectivity of the first dimension. In all instances both dimensions were temperature regulated at 45 C. In total eight systems were analysed (see Table 5.1). Table 5.1 Two-dimensional HPLC selectivity systems. First dimension Second dimension Solvent system in both dimensions 1. Cyano Carbon 18 Water:Methannol 2. Cyano Carbon 18 Water:Acetonitrile 3. PESC18 Carbon 18 Water:Methannol 4. PESC18 Carbon 18 Water:Acetonitrile 5. Pentafluorophenyl Carbon 18 Water:Methannol 6. Pentafluorophenyl Carbon 18 Water:Acetonitrile 7. Hypercarb Carbon 18 Water:Methannol 8. Hypercarb Carbon 18 Water:Acetonitrile 75

98 5.2.4 Data Analysis The two-dimensional data was analysed using the same software and parameters in Chapter 4, Section Data Analysis. 5.3 Results and Discussion This study focuses on the comparison of selective separation performance of a number of stationary phases, CN, PFP, Hypercarb and PESC18 phases, in combination with two different mobile phases (methanol or acetonitrile) as the first dimension in 2DHPLC analysis of tobacco samples. Figure 5.3 illustrates the uni-dimensional chromatogram for the first dimension for each of the eight systems for the tobacco leaf extract. The selectivity difference for each column under each solvent system is apparent in the first dimension. In each of the chromatograms there was a bimodal distribution of compounds, except for the separation achieved on the Hypercarb column. In the CN phase most of the compounds eluted in the first three minutes, however on the Hypercarb retention was greater with elution occurring continuously throughout the gradient, although the signal response was low, perhaps indicating strong and irreversible component adsorption onto the Hypercarb surface. In comparison to the Hypercarb stationary phase, the signal response on the CN column reached an absorbance value maximum at approximately 3.5 AU (water/methanol system), due to the co-elution of the many components, whereas on the Hypercarb column the absorbance did not exceed 0.5 AU when the methanol mobile phase system was employed. However, when the organic component was changed to acetonitrile, less retention was apparent on the Hypercarb column and the signal intensity increased to around 2.2 AU for some early eluting components. Here, it is likely that the π-π bonding capacity of the acetonitrile mobile phase suppressed the π-π bonding capacity of the Hypercarb stationary phase, hence reducing retention, and perhaps also reducing selectivity. Likewise, the increased elution strength of the solvent system with acetonitrile for separations on the PESC18 and PFP columns appears to have sharpened the elution of the later eluting components, resulting in an increase in the signal intensity. On both the PESC18 and PFP 1D 76

99 Absorbance (AU) Absorbance (AU) separations the majority of compounds eluted after the 3.6 minute mark, representing the opposite behaviour compared to the CN phase. However, because peak capacity in the 1D mode had been exceeded, specific changes are difficult to determine and thus a 2DHPLC system is required for an effective separation. (a) Tobacco (MeOH) CN HyperCarb PESC18 PFP (b) Tobacco (ACN) CN HyperCarb PESC18 PFP Time (min) Figure 5.3 Uni-dimensional separations of White Ox Tobacco extract on (a) CN, Hypercarb, PESC18 and PFP at a 20 %min -1 gradient, from water to methanol and held for 4 minutes, (b) CN, Hypercarb, PESC18 and PFP in gradient mobile phase at a 20 %min -1 gradient, from water to acetonitrile, then held for 4 minutes. 0.0 Time (min) Consequently, 2DHPLC was employed to further evaluate the nature of the selectivity changes occurring on each of the columns and solvent systems, as illustrated in the two-dimensional surface plots shown in Figure 5.4, where the C18 column was used as the second dimension for each system. With the use of a common C18 column as the second dimension, the peak displacement is reflected by the nature of the first dimension. The detected peak location is represented by the dots, based on the peak maxima. These 2D chromatograms show significant differences of retention behaviour of the extracted compounds between the different stationary phases and mobile phases for the tobacco sample (Figure 5.4). To quantify the degree of system correlation in each of the 2D systems a geometric approach to factor analysis (GAFA) was carried out, the results of which are reported in Table 5.2. The two-dimensional data was assessed qualitatively and quantitatively to determine the separation power afforded by each system for each sample. GAFA was applied to the data sets to determine the number of peaks isolated (n), the spreading angle (β), the practical peak capacity (n Tƒ ) and per cent usage of the separation space [16]. The measure of separation difference is represented by the correlation value. Each chromatogram has been sectioned into quadrants (Q) as 77

100 indicated by the dotted lines on the 2D chromatograms. Each quadrant of each chromatogram was analysed using GAFA, as well the entire two-dimensional separation space. This approach was applied in Chapter 4 also. The results in Table 5.1 indicate that overall, a number of the separation systems in the first dimension have high correlation to the C18 phase. The results also show that acetonitrile, generally being a stronger solvent than methanol, hence less retention and greater component co-elution, yields a lower peak capacity in almost all of the 2D systems tested. In order to fully explore the nature of the selectivity changes for such a diverse range of compounds as found in the tobacco samples, the data was evaluated in each of the four quadrants, separately, and across the entire separation space, wholly; thus, providing a more detailed assessment of differences in retention behaviours. 78

101 C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) (a) (b) Q2 Q3 Q2 Q3 Q1 Q4 Q1 Q4 CN Dimension Retention Time (min) CN Dimension Retention Time (min) (c) Q2 Q3 (d) Q2 Q3 Q1 Q4 Q1 Q4 PESC18 Dimension Retention Time (min) PESC18 Dimension Retention Time (min) (e) Q2 Q3 (f) Q2 Q3 Q1 Q4 Q1 Q4 PFP Dimension Retention Time (min) PFP Dimension Retention Time (min) 79

102 C18 Dimension Retention Time (min) C18 Dimension Retention Time (min) (g) (h) Q2 Q3 Q2 Q3 Q1 Q4 Q1 Q4 Hypercarb Dimension Retention Time (min) Hypercarb Dimension Retention Time (min) Figure 5.4 Two-dimensional separations of White Ox Tobacco extract: (a) System 1 1DCN/2DC18 (H 2 O:MeOH) (z-axis 0.3 AU), (b) System 2 1DCN/2DC18 (H 2 O:ACN) (z-axis 0.3 AU), (c) System 5 1DPESC18/2DC18 (H 2 O:MeOH) (zaxis 0.1 AU), (d) System 6 1DPESC18/2DC18 (H 2 O:ACN) (z-axis 0.1 AU), (e) System 7 1DPFP/2DC18 (H 2 O:MeOH) (z-axis 0.2 AU), (f) System 8 1DPFP/2DC18 (H 2 O:ACN) (z-axis 0.5 AU), (g) System 3 1DHypercarb/2DC18 (H 2 O:MeOH) (z-axis 0.07 AU), (h) System 4 1DHypercarb/2DC18 (H 2 O:ACN) (zaxis 0.07 AU). In both dimensions mobile phase ran at a 20 %min -1 gradient, from aqueous to organic solvent and was kept at 100% organic solvent for 4 minutes. 80

103 Table 5.2 GAFA calculations for the 2DHPLC separation of tobacco leaf extract and in each of the quadrants. System Detected Peaks (n) Correlation Spreading Angle Peak Capacity Usage (%) 1. 1DCN/2DC18 (H 2 O:MeOH) Q Q Q Q4 N/A N/A N/A N/A 2. 1DCN/2DC18 (H 2 O:ACN) Q Q Q Q4 N/A N/A N/A N/A 3. 1DPESC18/2DC18 (H 2 O:MeOH) Q Q Q Q DPESC18/2DC18 (H 2 O:ACN) Q Q2 N/A N/A N/A N/A Q Q DPFP/2DC18 (H 2 O:MeOH) Q Q Q Q4 N/A N/A N/A N/A 6. 1DPFP/2DC18 (H 2 O:ACN) Q Q Q Q4 N/A N/A N/A N/A 7. 1DHypercarb/2DC18 (H 2 O:MeOH) Q Q2 N/A N/A N/A N/A Q3 N/A N/A N/A N/A Q DHypercarb/2DC18 (H 2 O:ACN) Q Q Q Q

104 5.3.1 Qualitative Assessment of the Selectivity Changes Overall System Comparison The tobacco leaf extract was analysed using eight 2DHPLC systems. Overall, System 1 (CN/C18, with water/methanol) yielded the highest number of detected peaks, with a correlation of 0.72, spreading angle of 44 and peak capacity usage of 58%. Although, System 1 obtained the highest peak count, correlation between the two dimensions was high, albeit, overall this system produced the best two-dimensional separation performance of the eight systems tested. Figure 5.5 represents the chromatographic overlay of the second dimension separations of all the incremental heart cuts for System 1. In comparison, Systems 3 to 6 yielded very high correlations between the first and second dimensions (up to for System 4 (PESC18/C18 with water/acetonitrile)). Although, the GAFA analysis of Systems 7 and 8 (Hypercarb/C18) indicated highly divergent retention behaviour between the first and second dimensions, visual inspection of the two-dimensional surface plots indicate poor separation and resolution, i.e. high solute crowding [103]. In effect, very little of the two-dimensional separation space was effectively utilised. Furthermore, there was a high concentration of solute that eluted within the void space of the C18 column in the second dimension, indicating that the solvent transfer volume was too strong to enable interactions with the second dimension stationary phase. As a consequence, the 2D systems that utilised the Hypercarb phase were not investigated further. The surface plots in Figure 5.4 show higher use of separation space in certain regions of the two-dimensional separation space than other areas. For example in Figure 5.4 (a) and (b), the majority of peaks are displaced in quadrants 1, 2 and 3, but no peaks are detected in quadrant 4. Therefore, for a more detailed discussion on the twodimensional performance, the separation space was divided into four equal quadrants, consistent with the study undertaken in Chapter 4. At any rate, the displacement of solutes into quadrant 4 would likely result in very poor chromatographic performance since these compounds would be eluting from the first dimension in a strong solvent environment and then be eluting from the second dimension with little retention. This is the perfect scenario for solvent mismatch phenomena to interfere with elution behaviour, especially given the transport volume 82

105 Absorbance (AU) Absorbance (AU) between the two dimensions is in the order of 10-fold greater than a regular injection plug. Thus, resulting in a dilute mixture of solute dispersed in a large volume plug of strong solvent. Under these conditions solvation effects [104] and viscous fingering could be important [26,30,105,106] Time (min) Time (min) Figure 5.5 Chromatographic overlay of second dimension separation (C18) of all incremental heart cuts for System 1-1DCN/2DC18 (H2O:MeOH). Cyano Phase in First Dimension. Quadrant 1: In Q1, components in the tobacco sample were weakly retained on the CN phase. This is indicative of components that have very little affinity for nonresonance π-π bonding. Although, retention of these compounds was limited on the CN phase their distribution across the C18 phase in the second dimension was substantial, thus indicating that these compounds had a significant diversity in polarity, likely to be related to the amount, or structure of a hydrocarbonaceous backbone. The variation in the component distribution in the C18 phase is shown in the 1D chromatogram of a segment cut at 0.8 minutes in the first dimension using System 1 (Figure 5.6). It is important to note that the chromatogram shown in Figure 5.6 results from compounds that were retained in just 200 µl of the first dimension, i.e. between 0.8 and 0.9 minutes at a flow rate of 2.0 mlmin -1. This is an excellent 83

106 Absorbance (AU) example of the power of selectivity in the optimisation of resolution; even so, there is still incomplete separation in second dimension, as evidenced by the shouldering and skewing of peaks. This type of separation and selectivity was also observed in a selectivity study of espresso coffee by Mnatsakanyan et al. [15]. The use of methanol in the solvent system (System 1) yielded 29 peaks whereas when acetonitrile (System 2) was utilised only 18 peaks were detected. System 1 also obtained a lower correlation value of 0.29 than System 2 (0.46) and a spreading angle of 73 and practical peak capacity of 86%, whereas System 2 obtained a spreading angle of 63 and practical peak capacity of only 76% nd Dimension (C18) Chromatogram of 0.8 min cut Time (min) Figure 5.6 Separation of White Ox Tobacco extract heart-cut segment ( min.) from the CN first dimension on a C18 column in the second dimension utilising System 1 (1DCN/2DC18 (H 2 0:MeOH)). Quadrant 2: Components in the tobacco sample were distributed widely across the separation space of the CN dimension, and moderately so across the C18 dimension. The majority of peaks were detected in this quadrant for both systems that employed the CN and C18 phases (water/methanol and water/acetonitrile solvent systems) and there were similar numbers of components in this quadrant for both Systems 1 and 2. However, System 1 was a little more correlated (0.34) than System 2 (-0.21). System 1 had a spreading angle of 70 and practical peak capacity of 82%, where System 2 resulted in a slightly higher spreading angle (78 ) and practical peak capacity of 97%. The GAFA data for this quadrant is indicative of highly divergent retention 84

107 behaviour. In the analysis undertaken by Mnatsakanyan et al. [15] a similar separation behaviour was observed for coffee samples, and shown here in Chapter 4. In contrast to Q1, the peak displacement in Q2 was latitudinal and the number of peaks resolved in Q2 was about doubled that of Q1. The range of retention displacement on the CN phase suggests a range of selective interactions perhaps based largely on non-resonance π-π bonding while in the second dimension separation was more limited and hence these compounds may have similar polarities. Quadrant 3: Compounds in this quadrant are late eluting compounds in both dimensions, which indicate that these components are less polar, yet are likely to be able to exhibit non resonance π-π bonding with the functionality of the CN phase. Systems 1 and 2 for Q3 had the same number of peaks detected (33 peaks), with similar GAFA outcomes, where System 1 (water/methanol solvent system) had a correlation of 0.14, spreading angle of 82 and peak capacity usage of 93%. System 2 (water/acetonitrile solvent system) had a correlation of (inverse), spreading angle of 87 and peak capacity usage of 97%, albeit, with a greater degree of solute crowding in the lower left hand corner of the quadrant, i.e., towards the overall centre of the two-dimensional separation space. Mnatsakanyan et al. [15] obtained a much higher correlation for the components in coffee in Q3, albeit using the Phenomenex column set tested in Chapter 4 of this thesis. Quadrant 4: No peaks were present in System 1 and 2. PESC18 Phase in First Dimension The PESC18 column as the first dimension showed a high correlation to the C18 column as the second dimension. In System 1 (water/methanol solvent system), 111 peaks were detected with a correlation value of 0.81, spreading angle of 36 and practical peak capacity usage of 49%. On the other hand, in System 2 (water/acetonitrile solvent system), only 37 peaks were detected and the correlation of the PESC18 column to the C18 column was much higher of 0.94, a spreading angle as low as 20 and peak capacity usage of only 31% (lowest one of all the systems). The high correlation can also be observed in the two-dimensional surface plots (Figure 5.4 (c) and (d)) in which the retention profile is largely aligned along the main diagonal. Having said that there were regions of the two-dimensional 85

108 separation space that displaced very divergent retention behaviour. For example, in System 3, the central region of the two-dimensional space (highlighted by the yellow rectangle) a lower degree of correlation is observed as the peak displacement showed effectively orthogonal retention behaviour. GAFA was carried out for this region and resulted in the detection of 27 peaks, a correlation of only 0.097, spreading angle of 84.4 and a high practical peak capacity usage of 95.1%. A similar peak displacement is observed in System 4 at the top border regions of Q1 and Q4 (indicated by the yellow rectangle). However, only 6 peaks were detected for this region and a GAFA analysis resulted in a correlation value of 0.637, spreading angle of 50.4 and practical peak capacity of only 64.1%. Regardless of this region displaying orthogonal retention behaviour, the overall performance of the PESC18 and C18 combination in 2DHPLC for the sample as a whole was poor, indicative of the likeness in the chemical nature of the PESC18 and C18 columns. Pentafluorophenyl Phase in First Dimension System 5 (PFP/C18 water/methanol solvent system) resulted in the detection of 108 peaks, with a correlation of 0.81, spreading angle of 36 and a practical peak capacity usage of only 49%. System 6 (PFP/C18 water/acetonitrile solvent system) resulted with similar GAFA results to System 5, with a correlation of 0.78, spreading angle of 39 and peak capacity usage of 52%; however, the number of peaks detected by System 6 were about half that of System 5 (57 peaks).the overall performance of PFP/C18 under both solvent conditions resulted in highly correlated systems. Thus, quadrant analysis was carried within each system to observe the degree of orthogonality within the specified regions. Quadrant 1: A greater number peaks were resolved in Q1 using System 1 (water/methanol solvent system) than System 2 (water/acetonitrile solvent system), with a greater degree of retention of compounds across the PFP phase. System 1 obtained a correlation of 0.61, spreading angle of 53 and peak capacity of 66%. Although the separation performance in System 2 resulted in a lower peak count than System 1, the other GAFA performance metrics were similar. Quadrant 2: When methanol was as the mobile phase (System 5) there was more separation across the second dimension (C18) than there was the case when acetonitrile (System 6) was used. This is supported by the GAFA results in which the 86

109 number of peaks detected for System 5 was greater than the number of peaks in System 6. This quadrant on both systems also obtained a high peak capacity usage of up to 97%, which supports the visual displacement of the peaks in the twodimensional surface plot (Figure 5.4). The high number of components eluting in this quadrant reflects components that have a low affinity for the PFP phase, while at the same time having a high affinity for the C18 phase [15]. Effectively, these two stationary phases (PFP and C18) for the types of compounds that elute in this region of the 2D space, provide near orthogonal retention behaviour, Thus if the components of interest to the analyst elute in this region, coupling these two columns provides for a very powerful separation; convenient also because the solvent environments that transport solute from the first dimension to the second are reasonably compatible. Quadrant 3: In this quadrant, a similar separation pattern to Q3 of the PESC18/C18 system can be seen, as the peaks are displaced primarily along the main diagonal, indicating high correlation between both dimensions for the types of compounds that elute here. Having said that while components were aligned along the main diagonal, there was significant scatter that resulted in a reasonably high degree of space utilisation; up to 76% for the PFP/C18 in combination with water/methanol as the mobile phase. The compounds in this quadrant are likely to be higher in molecular weight compared to those in Q1, and quite hydrophobic. Quadrant 4: No solute peaks were observed to elute in this region. Summary: Quadrant System Analysis Quadrant 1: Systems 1, 7 and 8, yielded almost orthogonal retention behaviour for the components in tobacco that eluted in this quadrant. Correlation values were between 0.03 (System 8 Hypercarb/C18 with water/acetonitrile) and 0.30 (System 7 Hypercarb/C18 with water/methanol). Whereas Systems 3 and 4 (PESC18/C18) were highly correlation in this quadrant (correlation values were 0.91 (System 3) and 0.85 (System 4)). The lowest utilisation of separation space was obtained for System 3, with a spreading angle of 25 and peak capacity usage of just 36%. Quadrant 2: System 2 (CN/C18 water/acetonitrile) gave the highest number of detected peaks in this quadrant and correlation in this region was also low (

110 (inverse)), effectively producing near orthogonal retention behaviour. Overall, Q2 within the majority of systems for the tobacco leaf extract performed with almost near orthogonal retention behaviour, with high spreading angles and large peak capacity usage. Quadrant 3: The peaks in this quadrant are those compounds that have higher molecular weights, are complex molecules, and generally hydrophobic. They are compounds that were strongly retained on both dimensions. The majority of systems showed a low correlation between dimensions in Q3. However, Systems 4 (PESC18/C18 water/acetonitrile) and 6 (PFP/C18 water/acetonitrile) showed moderate correlation (up to 0.69 (System 6)). In general, despite the low to moderate correlation observed through visual inspection of the surface plots, components tended to be aligned loosely along the main diagonal, although well separated, i.e., low solute crowding. Quadrant 4: The majority peaks in this quadrant for all systems, except those containing the Hypercarb as the first dimension, were classified as solvent peaks, or components that were unable to be retained onto the C18 phase in the second dimension because the solvent transport volume was too strong to allow solute interaction with the second dimension stationary phase. Generally speaking it is difficult to utilise this region of the two-dimensional separation space in RP-RP 2DHPLC due to solvent incompatibility between dimensions. 5.4 Conclusion A detailed study into the selectivity changes of stationary phase on the first dimension of 2DHPLC on complex samples was conducted. Depending on the phase of the first dimension and solvent system in both dimensions, a sample containing the same components can produce very different 2D chromatographic profiles. The overall best system for a sample will depend on the objective of the separation analysis. For example, if the number of components is the primary interest, then System 1 (CN/C18 water/methanol) would be the most suitable for the tobacco leaf extract, as it had the highest overall peak count. However, if the objective was to maximise selectivity difference, with respect to peak count then System 2 (CN/C18 88

111 water/acetonitrile) would be most suitable as it obtained the lowest overall system correlation, with a high separation angle and greater 2D peak capacity usage. The analysis of regional quadrants of the 2D systems showed that depending upon the nature of the components of interest, two-dimensional performance could be quite different than the overall performance of a system. Thus, to achieve perfectly uncorrelated selectivity between the two dimensions, with utilisation of all the separation space for complex samples may be impossible [13]. However, regional quadrant 2D analysis may be best for complex compounds of interest, as orthogonality is established on the nature of the sample with respect to separation environment [13]. 89

112 CHAPTER SIX A Preliminary Investigation into Two- Dimensional HPLC Incorporating Active Flow Technology (AFT) Columns 90

113 6.1 Introduction To ensure optimal peak capacity in 2DHPLC, each of the separation dimensions should be as different as possible. The greater the selectivity difference between the dimensions the greater the separation power [107]. However, a limitation in achieving perfectly orthogonal retention behaviour comes about because it is very difficult to obtain compatible solvent environments in systems that utilise stationary phases that offer such divergent retention behaviour [10,26]. Thus, the incompatibility of the solute transport between dimensions often results in poor peak shape, loss of efficiency and subsequently a decrease in total peak capacity [104]. This is very often worsened by using transfer volumes much larger than the ordinary injection volumes (often more than 10-fold larger) employed in uni-dimensional HPLC. Solvent mismatch can occur due to solubility effects or differences in viscosity of the sample solvent (solute) and the mobile phase. The solubility between the solute and mobile phase is determined by their relative thermodynamic and kinetic properties. Solubility is a result of a phenomenon known as solvation, which occurs when the properties of the solute and solvent are similar; the solute molecules spread out and become surrounded by solvent molecules. When the properties of the solute and solvent are divergent, the degree of solubility is reduced and thus a solvent mismatch can occur, leading to distorted peak shapes and band broadening [19,108]. Solvation phenomena are particularly important when the molecular weight of the solute increases [109]. The viscosity contrast between the solvents inside a column bed can also lead to loss in separation efficiency, through a phenomenon known as viscous fingering (VF). VF is caused by the hydrodynamic fluid instability due to the difference of viscosities between the solute and mobile phase, where the one solvent pushes the other through the column bed. If the solvent with the lower viscosity pushes the higher viscosity solvent, then the interface between these solvents becomes unstable, and the lower viscosity solvent penetrates the higher viscosity solvent in a complex manner that resembles fingers hence the name [26,105,106, ]. In chromatography, the effect is worsened as the column internal diameter and the injection plug volume increases. Since large transfer volumes are employed in 91

114 2DHPLC, the solvent plug is susceptible more so to these hydrodynamic instabilities, reducing the potential separation power that is afforded by a 2DHPLC system. Thus, chromatographers must be aware of the effects of solvent mismatch and take due care when utilising two different solvents, even if fully miscible in all proportions [26]. Practitioners have used a variety of 2DHPLC system setups to overcome solvent mismatch issues, such as, solvent evaporation interfaces [113], reverse osmosis [114], peak trapping recycle chromatography [115], and on-column focusing at the second dimension inlet prior to second dimension analysis [ ]. Such techniques aim to reduce the solute volume of the transfer fraction onto the second dimension to minimise the effects of solvent incompatibility. Other 2DHPLC techniques to minimise sample transfer volume from the first dimension involve the use of narrow internal diameter (i.d.) columns on the first dimension such as a 3.0 mm or 2.1 mm i.d. coupled with a wider bore column, such as, 4.6 mm i.d. on the second dimension. Although such techniques minimise the transfer volume allowing the chromatographer to use highly divergent dimensions, other issues arise, such as, efficiency loss on the narrower bore columns and lower detection sensitivity. The effects of column bed heterogeneity and associated issues, such as the wall effect, are more drastic on narrow-bore columns [43,55], like the 2.1 mm i.d. columns and thus having a much lower theoretical plate count (efficiency) and separation performance compared to a 4.6 mm i.d. column. Sensitivity is also affected, as a lower injection volume in the first dimension is required with the use of 2.1 mm i.d. column, thus although a smaller transfer volume is made on the second dimension, a smaller concentration of sample is also transfer into a larger i.d. column, resulting in substantial dilution. Hence, a technique that caters for smaller transfer volumes between dimensions, whilst maintaining optimal efficiency and sensitivity would be ideal. The new Active Flow Technology (AFT) chromatography columns that were used for the multiplexed detection study in Chapter 3 provide an opportunity to couple the two dimensions in a manner that reduces the volume transfer between dimensions, without losing sampling efficiency, since the solute plug is uniformly sampled from the homogeneous radial central region of the column. Furthermore, prior studies in uni-dimensional HPLC have shown that even though up to 85% of the sample plug 92

115 was directed to waste, the sensitivity remained the same as the same injection on a conventional column [55]. Furthermore, the modulation rate of sampling between dimensions remains constant even though the volume transferred can be reduced. Potentially this will reduce the effects of incompatible solvent mixing in the second dimension. The purpose of this study is to evaluate whether AFT columns, when used in 2DHPLC provide a more efficient means of volume transfer between dimensions, and also whether these two-dimensional separations are more efficient. 93

116 PART A: 2DHPLC CONVENTIONAL, PARALLEL SEGMENTED FLOW AND CURTAIN FLOW CHROMATOGRAPHIES 6.2 Experimental Chemicals Mobile phase solvents were HPLC grade. Methanol was purchased from Merck (Kilsyth, Victoria, Australia) and Ultrapure Milli-Q water (18.2 MΩ) was prepared in-house and filtered through a 0.2 µm filter Sample Preparation A 0.5 mg/ml caffeine standard was prepared in 60:40 (water:methanol) solution. The caffeine was purchased from Sigma-Aldrich PTY Ltd. (Castle Hill, NSW Australia) Instrumentation and Chromatographic Conditions Instrumentation All sample analyses were undertaken on a Waters 600E Multi Solvent Delivery LC System, equipped with a Waters 717 plus auto injector, two - Waters 600E pumps, two - Waters 2487 series UV-Vis detectors (set at 280 nm) and two - Waters 600E system controllers. Separation dimensions were interfaced via a 10-port, two position switching valve that was electrically actuated. Sample loops between dimensions ranged from µl depending on the column format (see Table 6.1 for details). When Parallel Segmented Flow (PSF) columns were used, the central flow was taken to the UV-Vis detector, two peripheral ports were blocked and the third peripheral port flow was taken to waste. When Curtain Flow (CF) columns were in use, the inlet flow was set up in split-flow mode, in which 40% of the flow was connected to the column radial central port and 60% by-passed the injector and was connected directly to a peripheral port, while the other two peripheral ports were blocked. The outlet of the CF column was set up in the same way as the PSF column. Figure 6.1 illustrates the system setup of 2DHPLC in conventional, PSF and CF mode. In total fifteen 2DHPLC system modes were tested and are listed in Table

117 (a) (b) (c) Figure 6.1 Two-dimensional system setup with an autosampler, two pumps, two degassers, two system controllers, two UV-Vis detectors, a 10-port two position switch valve, and two columns in (a) Conventional mode, (b) PSF mode and (c) CF split-flow mode. 95

118 Table 6.1 2D System Parameters. Column Internal Injection Volume Loop Size (µl) Flow rate Diameter (mm) (µl) (mlmin -1 ) (21%)* (43%)* * 21% and 43% refer to the segmentation ratio through the central port of the 4.6 mm i.d. column in AFT column format. Table 6.2 2D System Modes. System First Dimension (mm i.d.) Second dimension (mm i.d.) 1. Conventional C18 (4.6) Conventional C18 (4.6) 2. Conventional C18 (3.0) Conventional C18 (4.6) 3. Conventional C18 (2.1) Conventional C18 (4.6) 4. Conventional C18 (2.1) Conventional C18 (3.0) 5. Conventional C18 (3.0) Conventional C18 (3.0) 6. Conventional C18 (2.1) Conventional C18 (2.1) 7. PSF43% C18 Conventional C18 (4.6) 8. PSF21% C18 Conventional C18 (4.6) 9. PSF43% C18 PSF43% C PSF43% C18 PSF21% C PSF21% C18 PSF21% C PSF21% C18 PSF43% C CF (i)*40% (o)*43% C18 Conventional C18 (4.6) 14. CF (i)40% (o)43% C18 CF (i)40% (o)43% C CF (i)40% (o)21% C18 CF (i)40% (o)43% C18 *(i) = inlet, (o) = outlet Chromatographic Columns The first and second dimensions consisted of Hypersil GOLD C18 columns Å (100 x 4.6 mm, 5 µm P d ). The columns were purchased from ThermoFisher Scientific (Manor Park, Tudor Rd, Runcorn, Cheshire, United Kingdom). 96

119 Chromatographic Separation Isocratic elution was undertaken in both dimensions with premixed 80:20 (water:methanol) solvent. Injection volume and flow rate on the first and second dimension are dependent on the 2D system mode (see Table 6.1 for details). The 2D analysis of the sample was undertaken using a comprehensive incremental heartcutting approach [13,68] involving the transfer of an aliquot sampled from the first dimension, into the second dimension, followed by elution in the second dimension. This process was repeated (following reinjection of the sample) at 0.1 minute intervals across the entire caffeine peak as it eluted from the first dimension; the entire caffeine peak was transferred to the second dimension throughout the course of injections in the first dimension, depending on the column format Data Analysis Data analysis was conducted using Variance [119] and the graphs were created using Origin 7.0 Pro (OriginLab). 6.3 Results and Discussion Uni-Dimensional HPLC Conventional versus PSF Columns PSF columns were designed to increase the efficiency of a separation by taking the most concentrated radial central section of the eluent band to the detector. This was achieved using a purpose built, multiport end fitting incorporating an annulus frit. The central to peripheral flow ratios can be varied by adjusting the relative pressure drop between these flow regions at the column outlet. Effectively the AFT column functions as a wall-less column, with virtual diameters that are dependent on the flow proportions. For example, if 21% of the flow through exits a 4.6 mm i.d. column from the radial central port a virtual 2.1 mm i.d. column is established. In this study, the separation performance of AFT columns in uni-dimensional and two-dimensional HPLC was evaluated in comparison to conventional columns of various internal diameters. The chromatograms in Figure 6.2 are elution profiles in 1D chromatography of caffeine, eluting from various HPLC columns, both conventional 97

120 Absorbance (AU) columns with internal diameters of 2.1, 3.0 and 4.6 mm and AFT columns operating in PSF mode with a physical 4.6 mm i.d., but operating as virtual 2.1 and 3.0 mm i.d. columns. Sensitivity was highest for the PSF columns, with less peak tailing than any of the conventional formats. More specifically, the sensitivity recorded for the virtual 2.1 mm i.d. PSF column (i.e. 21% of flow from the radial central exit port) was improved by 9% compared to the conventional 4.6 mm i.d. column and 57% compared to the 2.1 mm i.d. column. While the virtual 3.0 mm i.d. PSF column had gains of 19% and 33% compared to the conventional 4.6 and 3.0 mm i.d. columns respectively. The theoretical plate count for all systems was measured using the second moment method, see equation 1.2. To show the relative performance of each column format in terms of theoretical plates, the results were assigned a metric value, with the conventional 4.6 mm i.d. column as a reference. The relative performance metrics are summarised in Table 6.3. The gain in theoretical plates also obtained from the PSF columns was impressive; for the virtual 2.1 mm i.d. column, 54% and 96% increase compared to the 4.6 and 2.1 mm i.d. conventional columns respectively, and for the virtual 3.0 mm i.d. column, 76% and 67% compared to the 4.6 and 3.0 mm i.d. conventional columns respectively DC mm i.d. 3.0 mm i.d. 4.6 mm i.d. PSF21% PSF43% Normalised Time (min) Figure 6.2 Caffeine peak Conventional HPLC chromatograms overlayed with PSF chromatograms in uni-dimensional HPLC. 98

121 Absorbance (AU) Conventional versus CF Columns CF column formats were designed to improve sensitivity. The design of these columns confines the sample to the radial central region of the bed and a curtain flow of mobile phase limits radial dispersion to the wall. A PSF outlet section on the column maintains the requirements of the infinite diameter column producing a virtual wall-less column, even if the sample eventually reaches the wall. By confining the sample in the radial central region of the column throughout the migration process, detection sensitivity and separation efficiency have both been shown to increase [43,44,53,56]. The chromatographic data presented in Figure 6.4, however, is at variance to prior reported works [44,53,120]. The overlay of the chromatograms in Figure 6.3 show the elution profiles of caffeine, comparing conventional 2.1, 3.0 and 4.6 mm i.d. columns to CF columns functioning as virtual 2.1 and virtual 3.0 mm i.d. columns. Table 6.3 summarises the separation metrics - efficiency and sensitivity derived from these profiles. A minor increase in sensitivity is apparent in the CF modes, but far less than reported in prior works. Peak tailing on the CF columns was also substantial and this resulted in very poor efficiency. This unexpected result will be discussed later in this thesis DC mm i.d. 3.0 mm i.d. 4.6 mm i.d. CF (i)40% (o)21% CF (i)40% (o)43% Normalised Time (min) Figure 6.3 Caffeine Peak Conventional HPLC chromatograms overlayed with CF chromatograms in uni-dimensional HPLC. 99

122 Table 6.3 1DHPLC Sensitivity and efficiency measures for conventional PSF and CF column formats. 1D Column Format Efficiency (Performance Metric) Sensitivity (AU) 2.1 mm i.d mm i.d mm i.d PSF 21% PSF 43% CF (i)*40% (o)*21% CF (i)40% (o)43% * (i) = inlet, (o) = outlet Two-Dimensional HPLC Fifteen 2DHPLC systems were tested consisting of various combinations of conventional, PSF and CF column formats. The sample was cut in comprehensive incremental heart cutting mode across the first dimension and transferred into the second dimension to observe the fluid transfer process. A two-dimensional separation was conducted on a single component standard (caffeine standard) to investigate the fluid transport performance to the second dimension rather than the separation power afforded by a multidimensional HPLC system. The sensitivity and efficiency of the fraction that gave the highest absorbance response in the second dimension within each of systems was analysed and are summarised in Table 6.4; where all measurements were assigned a metric value to show the relative performance of each system, with the reference being a 2D system containing a 4.6 mm i.d. conventional columns in both dimensions. Inclusive of all fractions within each of the best performing systems, the maximum sensitivity and efficiency that was obtained was also summarised in Table

123 Table 6.4 2DHPLC - Sensitivity and efficiency measures of second dimension separation of a cut (maximum peak height) of caffeine for conventional, PSF and CF 2DHPLC systems. System Transfer Volume (µl) 1) 4.6 x 4.6 mm i.d ) 3.0 x 4.6 mm i.d. 43 3) 2.1 x 4.6 mm i.d. 21 4) 2.1 x 3.0 mm i.d. 21 5) 3.0 x 3.0 mm i.d. 43 6) 2.1 x 2.1 mm i.d. 43 7) PSF43% x Conventional 4.6 mm i.d. 8) PSF21% x Conventional 4.6 mm i.d ) PSF43% x PSF43% 43 10) PSF43% x PSF21% 43 11) PSF21% x PSF21% 21 12) PSF21% x PSF43% 21 13) CF (i)40% (o)43% x Conventional 4.6 mm i.d. 14) CF (i)40% (o)43% x CF (i)40% (o)43% 15) CF (i)40% (o)21% x CF (i)40% (o)43% Performance Metric Sensitivity 1 Efficiency 1 Sensitivity 0.38 Efficiency 1.31 Sensitivity 0.17 Efficiency 1.28 Sensitivity 0.31 Efficiency 0.96 Sensitivity 0.79 Efficiency 0.86 Sensitivity 0.47 Efficiency 0.61 Sensitivity 0.49 Efficiency 1.37 Sensitivity 0.21 Efficiency 0.54 Sensitivity 0.55 Efficiency 1.37 Sensitivity 0.43 Efficiency 1.11 Sensitivity 0.24 Efficiency 1.02 Sensitivity 0.27 Efficiency 1.35 Sensitivity 0.59 Efficiency 1.37 Sensitivity 0.74 Efficiency 0.91 Sensitivity 0.53 Efficiency

124 Table 6.5 2DHPLC - Maximum sensitivity and efficiency measures obtained of second dimension separations of caffeine for conventional, PSF and CF 2DHPLC systems. System Transfer Volume (µl) Max Performance Metric 1) 4.6 x 4.6 mm i.d. 100 Sensitivity 1 Efficiency 1 2) 3.0 x 4.6 mm i.d. 43 3) 2.1 x 4.6 mm i.d. 21 4) 2.1 x 3.0 mm i.d. 21 5) 3.0 x 3.0 mm i.d. 43 6) 2.1 x 2.1 mm i.d. 43 7) PSF43% x Conventional 4.6 mm i.d. 8) PSF21% x Conventional 4.6 mm i.d ) PSF43% x PSF43% 43 10) PSF43% x PSF21% 43 11) PSF21% x PSF21% 21 12) PSF21% x PSF43% 21 13) CF (i)40% (o)43% x Conventional 4.6 mm i.d. 14) CF (i)40% (o)43% x CF (i)40% (o)43% 15) CF (i)40% (o)21% x CF (i)40% (o)43% Sensitivity 0.38 Efficiency 1.20 Sensitivity 0.17 Efficiency 1.02 Sensitivity 0.31 Efficiency 0.73 Sensitivity 0.79 Efficiency 0.72 Sensitivity 0.47 Efficiency 0.44 Sensitivity 0.49 Efficiency 1.32 Sensitivity 0.21 Efficiency 0.42 Sensitivity 0.55 Efficiency 1.41 Sensitivity 0.43 Efficiency 0.97 Sensitivity 0.24 Efficiency 0.86 Sensitivity 0.27 Efficiency 1.37 Sensitivity 0.59 Efficiency 1.32 Sensitivity 0.74 Efficiency 0.86 Sensitivity 0.53 Efficiency

125 Absorbance (AU) Conventional Columns A series of combinations of coupled conventional C18 columns having various internal diameters were tested in 2DHPLC. The second dimension chromatographic profiles of the caffeine band originating from the highest concentration region of caffeine in the first dimension for each of the respective combinations are shown in Figure 6.4. System x 4.6mm i.d. gave the highest sensitivity, while System x 4.6 mm i.d. gave the highest efficiency. The gain in efficiency in System 2 compared to System 1 was as a direct result of the smaller injection plug or cut transfer volume from the first to the second dimension, i.e., 43 compare to 100 L (see Figure 6.5). In contrast the combination of 2.1 x 4.6 mm i.d. columns (System 3) respectively, was less efficient than the 3.0 x 4.6 mm i.d. combination (System 2) and relatively the same in efficiency to the 4.6 x 4.6 mm combination (System 1) (see Table 6.5). System x 4.6 mm i.d., had a transfer volume of 21 µl and according to the relationships of injection volume and theoretical plate count (see Figure 6.5), the smaller injection volume should have obtained a much greater efficiency performance than the 3.0 x 4.6 mm i.d. (System 2) and 4.6 x 4.6 mm (System 1), however due to the extra column dead volume in the system, the potential performance benefit had decayed quickly Conventional 2D Systems: 1) 4.6 x 4.6 mm i.d. 2) 3.0 x 4.6 mm i.d. 3) 2.1 x 4.6 mm i.d. 4) 2.1 x 3.0 mm i.d. 5) 3.0 x 3.0 mm i.d. 6) 2.1 x 2.1 mm i.d Normalised Time (min) Figure 6.4 Second dimension maximum caffeine peak height Conventional columns comparison. 103

126 Figure 6.5 Theoretical plate count versus increasing injection volume. PSF Columns While there is the ability to tune PSF columns in an endless array of combinations, there are in fact just four combinations of merit in this study; those that can be compared to conventional column formats. That is, the establishment of virtual column combinations in 2.1 and 3.0 mm i.d. equivalents. Hence, the PSF column was operated in both dimensions with central flow streams of either 21% or 43% of the total flow, and these were coupled accordingly to yield virtual 2.1 x 2.1, 2.1 x 3.0, 3.0 x 2.1 and 3.0 x 3.0 mm i.d. combinations. The second dimension chromatographic profiles of the caffeine band originating from the highest concentration region of caffeine in the first dimension for each of the respective combinations of these virtual columns are shown in Figure 6.6, and for reference, the conventional 4.6 x 4.6 mm i.d. (System 1), and 3.0 x 3.0 mm i.d. (System 5) systems are included. System 9 PSF43% x PSF43% gave the highest sensitivity within the PSF 2D systems, but sensitivity was substantially less than the corresponding conventional systems. It thus appears that removing almost 67% of the solvent flow from each of the dimensions has as serious consequence to the sensitivity in detection. However, sensitivity is just one important metric, the other is efficiency. The most efficient of these 2D systems was also System 9 PSF43% X PSF43%, the virtual 104