Separations at the Speed of Sensors - Analytical Chemistry (ACS

Feb 13, 2018 - The response time to generate an analytical signal typically varies from ∼1 to 20 s, and they are generally limited to a single analy...
0 downloads 8 Views 1MB Size
Article Cite This: Anal. Chem. XXXX, XXX, XXX−XXX

pubs.acs.org/ac

Separations at the Speed of Sensors Darshan C. Patel,†,§ M. Farooq Wahab,†,§ Thomas C. O’Haver,‡ and Daniel W. Armstrong*,† †

Department of Chemistry and Biochemistry, University of Texas at Arlington, Arlington, Texas 76019, United States Department of Chemistry and Biochemistry, University of Maryland at College Park, College Park, Maryland 20742, United States



S Supporting Information *

ABSTRACT: The virtue of chemical sensors is speed and analyte specificity. The response time to generate an analytical signal typically varies from ∼1 to 20 s, and they are generally limited to a single analyte. Chemical sensors are significantly affected by multiple interferents, matrix effects, temperature, and can vary widely in sensitivity depending on the sensor format. Separation-based analyses remove matrix effects and interferents and are compatible with multiple analytes. However, the speed of such analyses has not been commensurate with traditional sensors until now. Beds of very small size with optimal geometry, containing core−shell particles of judicious immobilized selectors, can be used in an ultrahigh-flow regime, thereby providing subsecond separations of up to 10 analytes. Short polyether ether ketone lined stainless steel columns of various geometries were evaluated to determine the optimal bed geometry for subsecond analysis. Coupling these approaches provides subsecond-based detection and quantitation of multiple chiral and achiral species, including nucleotides, plant hormones, acids, amino acid derivatives, and sedatives among a variety of other compounds. The subsecond separations were reproducible with 0.9% RSD on retention times and showed consistent performance with 0.9% RSD on reduced plate height in van Deemter curves. A new powerful signal processing algorithm is proposed that can further enhance separation outputs and optical spectra without altering band areas on more complex separations such as 10 peaks under a second.

A

spectroscopic or electrochemical chemosensors and the separating power to eliminate matrix effects and interferences and to detect/quantitate multiple species. Most real-life analyses for drug development, biomarker discovery, and many other areas are still carried out by liquid chromatography because of the aforementioned deleterious and insurmountable issues with chemosensors. To meet the sensor time scale, researchers have attempted to achieve subsecond separations in hydrodynamic chromatography, electrophoresis, and chip-based arrays.7−11 As subsecond separations become a reality, separation-based sensors will soon become feasible.12 However, this approach also will require new signal processing approaches not commonly considered in either the sensor or separation fields. Such a powerful analysis tool even would have ramifications outside the area of sensors, for example, the emerging field of 3D-LC analysis, where it has been postulated that 1 million analytes can be analyzed.13 Indeed, the importance of separation speed was recently addressed and is a subject of extensive research.14−19 Currently, there are challenges associated with the realization of high-speed separations. In the case of packed beds, the adsorbent spheres cannot be made indefinitely smaller because of pressure increase as an inverse function of particle diameter squared (d2p). Liquid fluids have a thousand times higher viscosity than gases, which is also inversely related to pressure. These limitations restrict flow rates and hence the separation speed. Moreover, there are theoretical limits on band efficiency

dvances in material science, spectroscopy, and electrochemistry have led to the development of chemosensors, which can quickly generate signals for targeted component(s) in a specific system. Molecular luminescence sensors for detecting calcium,1 enantiomeric composition, and detection of chiral functional amines in cellular systems have recently been proposed.2 Fast, in situ chiral sensing has been achieved using approaches such as multicomponent assemblies monitored by exciton-coupled circular dichroism.3−5 Semiconductor-based gas sensors have been proposed, which can take 5−20 s from full response and recovery.6 Although very promising, most analytical sensors potentially suffer from a common disadvantage of mainly responding to a single component and, to varying degrees, responding less so to interferents. Molecular luminescence is known to be extremely sensitive to the system composition, with quenching by the matrix and photobleaching as joint problems. Electrochemical sensors are typically only useful for a single component detection and are prone to interferences, e.g., like ion-selective electrodes or blood glucose sensors. Separation-based approaches have the power of resolving multiple analytes in significantly more complex and varying systems but lack the response time of sensors. For example, chromatography effectively eliminates the matrix and detects the target analytes with no concurrent interference. A paradigm shift is needed to introduce a change in the time scale of separations, which by current standards takes 5−30 min for analyzing a multicomponent mixture. Ultra efficient separations coupled with the response time of optical sensors would be an ideal union of virtues. This idea naturally leads to the concept of separation-based sensors, which combines the speed of © XXXX American Chemical Society

Received: November 28, 2017 Accepted: February 5, 2018

A

DOI: 10.1021/acs.analchem.7b04944 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

ensure high-quality particles with minimal fines. Fines were removed by suspending the modified silica in heptane and removing the supernatant. Next, using optimal density and viscosity, several potentially promising combinations of solvents were identified, and slurries were subsequently analyzed using the settling test and filament test as well as optical microscopy.21 Stationary phases were suspended in dispersing slurry solvents and packed at high pressures (8000−13 000 psi) using a pneumatically driven Haskel pump with methanol or acetonitrile as a push solvent.21 Bare SPP silica was packed in 85:15 acetone/dichloromethane slurry using various slurry concentrations (5−15% w/ v). Teicoplanin and derivatized quinine stationary phases were packed using absolute ethanol/dimethyl sulfoxide and acetone/ dichloromethane slurries, respectively. After packing, columns were exposed to a >7 mL/min flow rate (for 0.3 and 0.46 cm i.d. columns, equivalent linear velocity used for 2.1 mm i.d. columns) or at >700 bar back pressure to evaluate bed stability. Removal of Instrument Response by Fourier Transform Deconvolution. The process described below was performed using MATLAB R2017b. To separate ten peaks under a second, before applying the Segmented Derivative Peak Sharpening (SDPS), we used a Fourier transform (FT) deconvolution to the system response from a chromatogram. The use of FT deconvolution to remove extra-column effects has been demonstrated earlier for multiple peaks25 and for a single peak using MATLAB.26 Briefly, in FT deconvolution, the raw chromatographic signal Rcol (t), where R is the absorbance signal as a function of time t with the column, was transformed into the frequency domain (v). To obtain the instrument response, R(t), we removed the column, and the connection tubings were directly attached to the detector. Similarly, the R(t) is also transformed in the frequency domain. Dividing Rcol (v) with R(v) in the frequency domain results in a “true” instrument-free chromatographic signal in the frequency domain S(v); obtaining the inverse Fourier transform of S(v) results in the chromatogram free of system dispersion in the time domain S(t). (See Supporting Information for MATLAB script).

due to eddy dispersion, which can contribute up to 75% of the total signal band broadening.20 Despite these drawbacks, one can obtain reduced plate heights of 0.99). As expected, when a wavelength (220 nm) was chosen where the absorbance of the analyte exceeds 2000 mAU, the calibration curves with respect to height began to drift from linearity. These findings are not unexpected and are in line with the assumptions and limitations of Beer−Lambert’s law. Thus, all the good practices of routine HPLC quantitation do not appear to break down in subsecond chromatography. Pushing the Limits: 10 Components in a Second. The ultimate potential of ultrahigh-speed separations was explored by attempting to analyze 10 analytes within a second with the optimized conditions presented in this study. When operated at the maximum obtainable flow rate (8 mL/min), the dead time of a 1.0 × 0.3 cm i.d. column is ∼0.42 s, which leaves only 0.58 s to separate retained analytes for a subsecond separation. A significant portion of this dead time (0.12 s) is in fact the travel

Figure 3. Subsecond HILIC separations of structurally and functionally related analytes on bare 1.9 μm SPP silica packed in 1.0 × 0.3 cm i.d. columns. (A) Nucleosides, 8.0 mL/min flow rate; (B) cytokinins (plant hormones), 7.9 mL/min flow rate; (C) auxins (plant hormones), 7.9 mL/min flow rate; (D) salicylic acid and derivatives, 7.9 mL/min flow rate. Method: 90:10 ACN/100 mM NH4OAc mobile phase, 254 nm UV, 250 Hz sampling frequency, 0.0 s detector response time.

Subsecond Chiral Separations. Figure 4A−C show baseline resolved chiral separations on teicoplanin bonded supports (1.0 × 0.3 cm i.d, 2.7 μm SPP) for three drugs with successively larger resolution factors, all well within a second. No particular method optimization was performed in this case, and the use of pure methanol with high flow rates resulted in baseline resolved subsecond separations, attributed to the high selectivity and efficiency of the stationary phase. The oxazepam separation in Figure 4C shows an impressive resolution of 1.9. An even more remarkable result is shown in Figure 4D for a baseline subsecond chiral separation of 2-phenylpropionic acid with an RS of 3.1 using derivatized quinine selector bonded to E

DOI: 10.1021/acs.analchem.7b04944 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 4. Subsecond chiral separations on teicoplanin and quinine chiral selectors. The top row consists of (A) 4-methyl-5-phenyl-2-oxazolidinone, (B) lorazepam, and (C) oxazepam, which were separated on teicoplanin bonded to 2.7 μm SPP packed on a 1.0 × 0.3 cm i.d. column. Method: mobile phase: methanol, 7.5 mL/min, 220 nm. The bottom row consists of (D) 2-phenylpropionic acid, 90:10 ACN/100 mM HCOONH4 mobile phase, 7.85 mL/min; (E) DNB-phenylglycine and 2-phenylpropionic acid, 90:10 ACN/100 mM HCOONH4 pHa 7.4 mobile phase, 7.7 mL/min; (F) FMOC-Val and 2-phenylpropionic acid, 90:10 ACN/200 mM HCOONH4 pHa 6.8 mobile phase, 7.85 mL/min, 254 nm UV, 250 Hz sampling frequency, 0.0 s detector response time.

time of solutes in extra-column volume. To overcome this challenge, the instrument response for this separation was removed using Fourier transform (FT) deconvolution as described in the experimental section. Removal of instrument response further improved the resolution between species due to the elimination of extra-column band broadening. Further, when a bare silica column is operated at a neutral pH, ionized silanols would result in a negative charge on silica. The resulting effect would repel negatively charged analytes (e.g., acids) or prevents them from entering the pores (Donnan exclusion), which is advantageous in this case.12 Retention also is significantly influenced by frictional heating at high flow rates leading to a reduction in elution times (up to 15% at high flow rates). With use of this information, mixtures of polar and charged analytes were selected for optimum performance on the column at 8 mL/min using 90:10 ACN/100 mM NH4OAc eluent. Figure 5A shows the raw data for 10 analytes eluted in a second. Note that ultrahigh efficiencies of the 1.9 μm SPPbased stationary phase and a high sampling frequency (250 Hz) were crucial in obtaining this remarkable separation. This result marks the largest number of analytes separated in the subsecond regime in any liquid-based separations. However, the separations of all adjacent peaks were not baseline. Approaches such as multivariate curve resolution and the power law exist to enhance resolution.29 However, such methods require an extensive set of calculations, or they change the peak areas, which can lead to nonlinear calibration curves and merging of small peaks with taller ones. In this case, an in silico resolution enhancement method can assist in resolving species to baseline separation (see Theory section). The chromatogram shown in Figure 5A was divided into 20 identical length segments, and the second and fourth derivative of each segment were multiplied with the A1 and A2 multipliers (see Supporting Information). By optimizing these multipliers, a notably sharpened chromatogram with near baseline resolution of all 10 species was obtained (see Figure 5B). A

Figure 5. Subsecond separation of 10 analytes in HILIC mode performed on 1.9 μm bare SPP silica packed on a 1.0 × 0.3 cm i.d. column. (A) The raw data from the analysis with extra-column band broadening removed using Fourier transform deconvolution; (B) the deconvoluted chromatogram sharpened with the resolution enhancing technique of segmented derivative peak sharpening. For both chromatograms, mobile phase: 90:10 ACN/100 mM NH4OAc, 8.0 mL/min flow rate, 254 nm UV, 250 Hz sampling frequency. Analytes in order of elution: (1) anthraquinone-2-sulfonic acid, (2) (R)-1,2,3,4tetrahydro-1-naphthol, (3) thiourea, (4) maleic acid, (5) levamisole hydrochloride, (6) DL-tryptophanamide, (7) 6-amino uracil, (8) transcinnamic acid, (9) 3-indole acetic acid, (10) 2,3,4-trimethoxybenzoic acid.

virtue of this resolution enhancement method is that the sharpened chromatogram theoretically retains the peak areas of the original chromatogram and can be of use in quantitation of poorly resolved species (e.g., second dimension LC × LC or other complex separations), see Figure 1. Note that this general F

DOI: 10.1021/acs.analchem.7b04944 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry process (algorithm) also can be used for spectroscopic signals such as (FTIR) to resolve overlapped bands better (see Supporting Information). While improving the separation speed is beneficial in many ways, and further enhancements are likely in the future, it is useful to view many of the avenues where subsecond separations can be of significant and immediate use. Highthroughput screening is currently done in a slower format, where 5−30 min separations are considered acceptable even for two chiral analytes.18 We are presently using these separation sensors for monitoring reaction intermediates with a lifetime of 50−900 ms. Multidimensional liquid chromatography (nLC × LC), an emerging technique that has significant advantages for analyzing complex mixtures, has not reached its potential due to the severe lack in suitable columns for the nth dimension.41 A multitude of challenges associated with the inadequate performance of the nth dimension has led to the limited usage of comprehensive and multiple heart-cutting approaches in LC × LC. The introduction of ultrahigh-efficiency short columns that can provide subsecond separations into a second dimension in 2DLC and can bring the much-needed paradigm shift. However, there are other associated challenges with 2DLC that must be overcome in the future, which include sampling rate, dilution effect, changes in the first-dimension flow rate, column orthogonality, as well as solvent and pH mismatch. The key aspect affecting the use of short columns in the second dimension is the mass or volume overload. Comprehensive and multiple heart-cut LC × LC can be applicable in more extensive analyses without the issue of wraparound. Wrap-around arises in LC × LC when an injection is made onto the second dimension, while not all the analytes have eluted from the previous injection in second dimension.42 Subsecond liquid separations also can be of utility in online LC with rapid cycle time for process analytical technology applications.

can now potentially be monitored in real-time leading to detailed insights into the reaction mechanisms.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b04944. Instrumentation and modifications to the UHPLC for short columns; chromatographic conditions; detector linearity and calibration curves based on peak height and peak area; retention time reproducibility for subsecond separations; subsecond separation of 10 analytes with and without peak enhancements; utility of segmented derivative peak sharpening in spectroscopic techniques; MATLAB script for removal of instrumental band broadening using Fourier transform deconvolution; segmented derivative peak sharpening (PDF) Excel sheet for segmented peak sharpening (XLSX)



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; Phone: (817) 272-0632; Fax: (817) 272-0619. ORCID

Darshan C. Patel: 0000-0002-7198-1163 M. Farooq Wahab: 0000-0003-4455-2184 Thomas C. O’Haver: 0000-0001-9045-1603 Daniel W. Armstrong: 0000-0003-0501-6231 Author Contributions §

D.C.P. and M.F.W. have equally contributed to this work.

Notes

The authors declare no competing financial interest.



■ ■

CONCLUSIONS Kinetic performance evaluation of short columns packed with superficially porous particles revealed the 1.0 × 0.3 cm i.d. as the optimum bed geometry for subsecond separations. Using teicoplanin bonded, derivatized quinine bonded, and bare silica particles packed columns, as short, ultrahigh-efficiency separation sensors, the fastest LC separation, with subsecond peak capacity was obtained for chiral and achiral analytes. Additionally, several biologically relevant classes of compounds such as plant hormones, nucleosides, and salicylic acid derivatives can be separated in a subsecond regime. Fourier transform deconvolution methods can be utilized to remove the extra-column band broadening from separations with multiple analytes. More importantly, Segmented Derivative Peak Sharpening, a new and efficient resolution enhancement technique, was developed. This method improved the resolution of species significantly while retaining the peak properties such as peak area and retention time. The algorithm can be used in spectroscopy (e.g., FTIR) to reveal hidden (overlapped) bands. These very fast separations approach or exceed the speed of many conventional sensors while maintaining the advantages of resolution (e.g., matrix elimination, multiple analyte measurements, numerous detection capabilities). Unstable or highly reactive chemistries are now possible candidates for direct monitoring with the help of subsecond separations. Short lived intermediates or impurities

ACKNOWLEDGMENTS The authors acknowledge Agilent Technologies for the gift of superficially porous particles. REFERENCES

(1) Roberts, S.; Seeger, M.; Jiang, Y.; Mishra, A.; Sigmund, F.; Stelzl, A.; Lauri, A.; Symvoulidis, P.; Rolbieski, H.; Preller, M.; Deán Ben, X. L.; Razansky, D.; Orschmann, T.; Desbordes, S.; Vetschera, P.; Bach, T.; Ntziachristos, V.; Westmeyer, G. G. Calcium Sensor for Photoacoustic Imaging. J. Am. Chem. Soc., 2017; DOI: 10.1021/ jacs.7b03064. (2) Wen, K.; Yu, S.; Huang, Z.; Chen, L.; Xiao, M.; Yu, X.; Pu, L. J. Am. Chem. Soc. 2015, 137 (13), 4517−4524. (3) You, L.; Long, S. R.; Lynch, V. M.; Anslyn, E. V. Chem. - Eur. J. 2011, 17 (39), 11017−23. (4) Joyce, L. A.; Maynor, M. S.; Dragna, J. M.; da Cruz, G. M.; Lynch, V. M.; Canary, J. W.; Anslyn, E. V. J. Am. Chem. Soc. 2011, 133 (34), 13746−52. (5) Zhu, L.; Anslyn, E. V. J. Am. Chem. Soc. 2004, 126 (12), 3676−7. (6) Su, M.; Li, S.; Dravid, V. P. J. Am. Chem. Soc. 2003, 125 (33), 9930−9931. (7) Clicq, D.; Vervoort, N.; Vounckx, R.; Ottevaere, H.; Buijs, J.; Gooijer, C.; Ariese, F.; Baron, G. V.; Desmet, G. Journal of Chromatography A 2002, 979 (1), 33−42. (8) Guetschow, E. D.; Steyer, D. J.; Kennedy, R. T. Anal. Chem. 2014, 86 (20), 10373−10379. (9) Jacobson, S. C.; Culbertson, C. T.; Daler, J. E.; Ramsey, J. M. Anal. Chem. 1998, 70 (16), 3476−3480.

G

DOI: 10.1021/acs.analchem.7b04944 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

(42) Jandera, P.; Hájek, T.; Č esla, P. J. Sep. Sci. 2010, 33 (10), 1382− 1397.

(10) Piehl, N.; Ludwig, M.; Belder, D. Electrophoresis 2004, 25 (21− 22), 3848−3852. (11) Umehara, R.; Harada, M.; Okada, T. J. Sep. Sci. 2009, 32 (3), 472−478. (12) Wahab, M. F.; Wimalasinghe, R. M.; Wang, Y.; Barhate, C. L.; Patel, D. C.; Armstrong, D. W. Anal. Chem. 2016, 88 (17), 8821− 8826. (13) de Villiers, A. LCGC North America 2017, 35 (8), 525−526. (14) Welch, C. J. ACS Cent. Sci. 2017, 3 (8), 823−829. (15) Welch, C. J.; Regalado, E. L. J. Sep. Sci. 2014, 37 (18), 2552− 2558. (16) Barhate, C. L.; Breitbach, Z. S.; Pinto, E. C.; Regalado, E. L.; Welch, C. J.; Armstrong, D. W. Journal of Chromatography A 2015, 1426, 241−247. (17) Regalado, E. L.; Welch, C. J. J. Sep. Sci. 2015, 38 (16), 2826− 2832. (18) Barhate, C. L.; Joyce, L. A.; Makarov, A. A.; Zawatzky, K.; Bernardoni, F.; Schafer, W. A.; Armstrong, D. W.; Welch, C. J.; Regalado, E. L. Chem. Commun. 2017, 53 (3), 509−512. (19) Barhate, C. L.; Regalado, E. L.; Contrella, N. D.; Lee, J.; Jo, J.; Makarov, A. A.; Armstrong, D. W.; Welch, C. J. Anal. Chem. 2017, 89 (6), 3545−3553. (20) Dasgupta, P. K.; Chen, Y.; Serrano, C. A.; Guiochon, G.; Liu, H.; Fairchild, J. N.; Shalliker, R. A. Anal. Chem. 2010, 82 (24), 10143− 10150. (21) Wahab, M. F.; Patel, D. C.; Wimalasinghe, R. M.; Armstrong, D. W. Anal. Chem. 2017, 89 (16), 8177−8191. (22) Laemmerhofer, M.; Lindner, W. J. Chromatogr. A 1996, 741 (1), 33−48. (23) Patel, D. C.; Breitbach, Z. S.; Yu, J.; Nguyen, K. A.; Armstrong, D. W. Anal. Chim. Acta 2017, 963, 164−174. (24) Wahab, M. F.; Pohl, C. A.; Lucy, C. A. Journal of Chromatography A 2012, 1270, 139−146. (25) Wright, N. A.; Villalanti, D. C.; Burke, M. F. Anal. Chem. 1982, 54 (11), 1735−1738. (26) Vanderheyden, Y.; Broeckhoven, K.; Desmet, G. Journal of Chromatography A 2016, 1465, 126−142. (27) Patel, D. C.; Breitbach, Z. S.; Wahab, M. F.; Barhate, C. L.; Armstrong, D. W. Anal. Chem. 2015, 87 (18), 9137−9148. (28) Patel, D. C.; Wahab, M. F.; Armstrong, D. W.; Breitbach, Z. S. Journal of Chromatography A 2016, 1467, 2−18. (29) Wahab, M. F.; Dasgupta, P. K.; Kadjo, A. F.; Armstrong, D. W. Anal. Chim. Acta 2016, 907, 31−44. (30) Wong, V.; Shalliker, R. A.; Guiochon, G. Anal. Chem. 2004, 76 (9), 2601−2608. (31) Talus, E. S.; Witt, K. E.; Stoll, D. R. Journal of Chromatography A 2015, 1378, 50−57. (32) Jespers, S.; Broeckhoven, K.; Desmet, G. LC GC Eu 2017, 30 (6), 284−291. (33) Reising, A. E.; Schlabach, S.; Baranau, V.; Stoeckel, D.; Tallarek, U. Journal of Chromatography A 2017, 1513, 172−182. (34) Shalliker, R. A.; Broyles, B. S.; Guiochon, G. Journal of Chromatography A 2000, 888 (1−2), 1−12. (35) Wei, B.; Rogers, B. J.; Wirth, M. J. J. Am. Chem. Soc. 2012, 134 (26), 10780−10782. (36) Barhate, C. L.; Wahab, M. F.; Breitbach, Z. S.; Bell, D. S.; Armstrong, D. W. Anal. Chim. Acta 2015, 898, 128−137. (37) Wang, Y.; Wahab, M. F.; Breitbach, Z. S.; Armstrong, D. W. Anal. Methods 2016, 8 (31), 6038−6045. (38) Kotoni, D.; Ciogli, A.; Molinaro, C.; D’Acquarica, I.; Kocergin, J.; Szczerba, T.; Ritchie, H.; Villani, C.; Gasparrini, F. Anal. Chem. 2012, 84 (15), 6805−6813. (39) Cancelliere, G.; Ciogli, A.; D’Acquarica, I.; Gasparrini, F.; Kocergin, J.; Misiti, D.; Pierini, M.; Ritchie, H.; Simone, P.; Villani, C. Journal of Chromatography A 2010, 1217 (7), 990−999. (40) Ismail, O. H.; Ciogli, A.; Villani, C.; De Martino, M.; Pierini, M.; Cavazzini, A.; Bell, D. S.; Gasparrini, F. Journal of Chromatography A 2016, 1427, 55−68. (41) Stoll, D. R.; Carr, P. W. Anal. Chem. 2017, 89 (1), 519−531. H

DOI: 10.1021/acs.analchem.7b04944 Anal. Chem. XXXX, XXX, XXX−XXX