For solution-phase samples, the world of mass spectrometry defaults to electrospray ionization (ESI). ESI is used for the analysis of a broad variety of compounds, ranging from polar to moderately nonpolar. However, ESI possesses limitations that prevent the ionization of certain analytes—particularly nonpolar compounds. This study aims to compare the ionization efficiency of complementary ionization techniques, and to demonstrate that multiple methods can improve the analytical results with respect to limits of detection and matrix tolerance. Atmospheric pressure photoionization (APPI) is an ionization method that complements ESI, excelling in the analysis of nonpolar and moderately polar analytes. In this study, we optimized methods using APPI and ESI for the detection and quantitation of pharmaceuticals frequently detected in the environment, including antibiotics, beta blockers, and selective-serotonin reuptake inhibitors, and tested their matrix tolerance relative to artificial wastewater. While most of these compounds ionized preferentially by ESI, some performed significantly better using APPI.

Every day, in the world of analytical chemistry, we strive to discover better ways to analyze, to achieve goals such as lower limits of detection, broader dynamic ranges, and techniques that tolerate greater interferences. Along with the need for improved methods of quantitation, there is pressure on the field of mass spectrometry (MS) to be able to successfully detect multitudes of analytes of interest from single injections. The question becomes, “How do we improve upon the detection of analytes?” However, sometimes using a single injection leads to a reduction in performance because it limits us to using a single ionization method. When you can optimize not only your separation but also your ionization method, your method may truly improve.
The basic principle of MS is to ionize molecules under study into gaseous ions, separate these ions in accordance with their mass-to-charge (

) ratio, and detect them (1). Today’s commercial instruments are capable of transferring 97–99% of ions successfully from the source to the detector, so the greatest improvements in detection and methods today are focused on the source. The ultimate question is, “When ionizing a sample, will the use of complementary ionization techniques improve the figures of merit of the individual analyses, or not?” 

Note: All tables and figures for this article are posted online at https://www.chromatographyonline.com.

ESI and APPI: Complementary Ionization Techniques

By choosing the appropriate ionization source, liquid chromatography with tandem mass spectrometry (LC–MS/MS) can be used for the detection of trace levels of contaminants, such as antibiotics and endocrine disrupting compounds (EDCs) from environmental samples. The electrospray ionization (ESI) source has been used as a powerful soft ionization technique for the analysis by MS of a wide array of sample types, ranging from polar to nonpolar, and has also been used for the analysis of thermolabile molecules of high molecular weight (MW) (2). Although ESI is popularly used for the analysis of environmental pollutants, it may not be able to ionize all contaminants efficiently. Certain contaminants are either poorly ionized, or not ionized at all. ESI is limited to analytes that are of low to high polarity and moderate to high MW (3).

Atmospheric pressure photoionization (APPI), introduced in 2000, is also a soft ionization technique. APPI was found to have success with the analysis of compounds with low to no polarity, and compounds of low to moderate MW, but cannot be used on thermolabile compounds. These parameters are what make APPI and ESI complementary to each other (4). This complementarity opens new doors for studies already utilizing the ESI method, because APPI can be used for complementary analysis of compounds that may not be detected by ESI. Additionally, APPI has shown tolerance to matrix components beyond what ESI has, due to its ionization pathway (5). APPI can be used for the analysis of a wide range of compounds, including drugs, human endogenous compounds, lipids, natural compounds, pesticides, synthetic organics, and petroleum derivatives (6). There are very few research papers reporting comparative studies of ionization efficiencies in MS for the detection of antibiotics and EDCs (7–9).

A triple-quadrupole mass analyzer was used in this study, utilizing both the full-scan mode optimization and multiple-reaction monitoring (MRM) quantitation. Full-scan mode can give qualitative analysis of a sample’s composition under study, and MRM mode is a highly selective mass monitoring mode with a wider linearity dynamic range, improved limit of quantitation (LOQ), increased sensitivity, and superior accuracy. One of the advantages of the MRM scan mode is an increase in signal‑to‑noise ratio because of removal of nonanalyte ions and isobaric precursors by monitoring fragments.

There are different MS acquisition parameters that affect the signal intensity of ions. The MassHunter Data Acquisition software sets a default value for all acquisition parameters for each ionization source (

). There is a sheath gas flow chamber in the electrospray ionization source that is absent in the atmospheric pressure photoionization source. As a result, the sheath gas temperature and sheath gas flow rate parameters are present only for ESI while APPI has an additional vaporizer parameter that is not present in ESI. Fragmentor voltage, collision energy, cell accelerator voltage, gas temperature, vaporizer, gas flow (L/min), nebulizer (psi), sheath gas temperature, and sheath gas flow rate were all optimized for each analyte prior to data acquisition in this study.

The current global population is growing at the annual rate of 1.09%, and pharmaceuticals are continuing to be prescribed and consumed at an alarming rate. In 76 countries across the globe, antibiotic consumption as described in defined daily doses (DDD) increased by 65%—from 21.1 billion doses in 2000 to 34.8 billion doses in 2015—and the overall antibiotic consumption rate has increased by 39% (10). In addition to antibiotics, beta blockers and antidepressants are two classes of pharmaceuticals gaining popularity. Beta blockers are a class of drugs frequently used to treat hypertension, heart disease, and other cardiovascular events. Although the true nature of their efficacy remains to be questioned in certain studies, beta blockers are still highly prescribed due to the diverse range of clinical symptoms they can successfully treat (11,12). In addition, according to 2017 data from the National Center for Health Statistics (NCHS), the rate of antidepressant use in America has increased by 65% since 1999 (13).

Unfortunately, this increase in pharmaceutical use means more pharmaceutical waste is likely to end up in the environment. While there is an urgency to know the exact harm this excess will cause, the priority is to increase our ability to detect as many pharmaceuticals in environmental samples as possible. This will then allow for proper removal techniques to be employed before the harmful substances have a chance to contaminate the environment (14).

Numerous studies have shown that some pharmaceuticals are not completely removed during wastewater treatment, and ultimately enter the environment in low concentrations. Important effects of pharmaceuticals entering the environment in low concentrations include antibiotic resistance, genotoxicity, acute or chronic toxicity, and endocrine disruption (15). Antibiotics and EDCs are emerging pollutants detected throughout the world, yet they remain unregulated by the U.S. Environmental Protection Agency (EPA) (16).

Sir Alexander Fleming, the British bacteriologist, discovered penicillin in 1928 from the fungus 

, spurring a novel era of antibiotics derived from microorganisms and antibiotic synthesis (17). The ability antibiotics have to eradicate a wide range of bacterial infections led to their increased use over time. Unfortunately, these bacteria have developed mechanisms to combat the actions of antibiotics. Thus, the over-prescribing of antibiotics, along with a lack of patient knowledge regarding the importance of correct antibiotic administration, has become an insidious issue that is known as 

. Antibiotic resistance arises as microorganisms develop the ability to survive against the action of antibiotics, meaning that when antibiotic‑resistant bacteria infect animals and humans, the antibiotic regimen that would normally eradicate the bacteria becomes useless. This is the reason being able to successfully detect antibiotics from wastewater samples is so important. In this study, five different classes of antibiotics were used: beta-lactams, macrolides, nitroimidazoles, sulfonamides, and tetracyclines.

Antibiotics can either be bacteriostatic, meaning they prevent the growth of bacteria, or they can be bactericidal, meaning that they destroy the bacteria entirely. However, an antibiotic’s mechanism of action is more important when considering treatment options. Beta‑lactams inhibit the biosynthesis of bacterial cell walls by making penicillin-binding proteins unavailable for new peptidoglycan synthesis, which causes the lysing of bacteria. The beta-lactams used in this study are ampicillin, ceftriaxone, cephalexin, and penicillin G. Macrolides inhibit protein synthesis during translocation in bacteria by dissociating peptidyl-tRNA from the middle of the 23S rRNA of the ribosome’s 50S subunit, causing early detachment of unfinished peptide chains. The macrolides used in this study are erythromycin and tylosin, an antibiotic popularly used in farm animals. Tetracyclines prevent the attachment of aminoacyl t-RNA to the A site in bacterial ribosomes by acting on the 16S rRNA of the 30S subunit inhibiting protein synthesis. Oxytetracycline and tetracycline were used in this study. Sulfonamides prevent the multiplication and growth of bacteria by inhibiting certain steps in the metabolism of folic acid. Sulfamethoxazole and trimethoprim were the sulfonamides used in this study. Nitroimidazole antibiotics inhibit nucleic acid synthesis that occurs in bacterial cells by disruption of the DNA in microorganisms. Metronidazole and 1,2 dimethyl-5-nitroimidazole were the nitroimidazoles used in this study (18).

EDCs are natural compounds or synthetic chemicals that mimic natural hormones in the body and interfere with the action of the natural hormones (19). These compounds most profoundly cause adverse effects on reproduction, developmental, neural, and immune systems of human beings and animals. Research suggests that EDCs reduce fertility and increase the risk of cancer, diabetes, obesity, and endometriosis (20). Among various EDCs, beta blockers (acebutolol, atenolol, metoprolol, and propranolol) and SSRI antidepressants (citalopram, paroxetine, and venlafaxine) were used in this study.

Make or Break for Successful Analysis: Matrix Effects and Wastewater

Properly dealing with impurities is a necessary complication in every field of research. In MS, matrix effects variability in ionization efficiency of analytes of interest as coeluted species serve to either enhance or inhibit the ionization process for an analyte. This issue becomes increasingly problematic when trying to discern analytes of interest from wastewater. Water that is obtained as a byproduct of agricultural, industrial, domestic, and commercial activity is termed 

. Wastewater contains nutrients such as calcium, iron, nitrogen, phosphorus, and potassium, and components such as fats, sugars, and proteins. In this study, synthetic wastewater was made to mimic the wastewater from the influent of a “typical” wastewater treatment plant, with its composition designed to imitate the dissolved inorganic solids and dissolved organic solids of real wastewater. The synthetic wastewater prepared in this study followed the protocol described by H.E. Gray, 2012 (21).

APPI has been found to be less susceptible to matrix effects compared to ESI. This is likely due to the fact that APPI is more selective in ionization. The photon emitted from the krypton lamp, at 10.6 eV, can ionize analytes, but potentially not the matrix components; the difference in how a sample is ionized can be the difference between more or less matrix interference (22). Using APPI involves the ejection of an electron from the analyte molecule to produce the gaseous radical cation (23). It is also possible, however, that the matrix component can act as a dopant and ionize sample components with high ionization energy through electron transfer, leading to signal enhancement.

Although matrix effects cannot be removed completely, they can be minimized by optimizing the sample preparation procedure and LC–MS parameters. Solid‑phase extraction (SPE) with an appropriate sorbent can reduce matrix effects by eliminating interfering matrices. The formula for calculation of the matrix effect is:

Physicochemical characteristics of antibiotics and EDCs help to determine the environmental fate of these compounds. Table 1 contains important physicochemical characteristics of antibiotics and EDCs under study, including solubility, p

less than 2.5 means the compound is hydrophilic and readily found in the aqueous phase.

A total of 12 antibiotics and 7 EDCs were analyzed in this study. Ceftriaxone sodium salt hemi(heptahydrate) and erythromycin with a purity of >98% were purchased from Acros Organics. Propranolol hydrochloride (99%), metronidazole (99%), acebutolol hydrochloride, metoprolol tartrate (98%), tetracycline hydrochloride (96%), and oxytetracycline hydrochloride were purchased from Alfa Aesar. TCI Co. was the main supplier of chemicals: cephalexin monohydrate (>98%), sulfamethoxazole (>98%), penicillin G potassium salt (>98%), atenolol (98%), venlafaxine hydrochloride (>98%), trimethoprim (>98%), citalopram hydrobromide (>98%), and 1,2 dimethyl-5-nitroimidazole. Ampicillin sodium salt was procured from Affymetrix Inc. Tylosin tartrate (95+%) and paroxetine hydrochloride (98+%) were bought from Ark Pharm Inc. All antibiotics and EDCs were used without further purification. All the solvents used in the analysis, such as acetonitrile, methanol, and toluene, are of HPLC-grade and purchased from Fisher Chemical. Potassium phosphate monobasic (99.8%) was purchased from EK Industries Inc. Sodium acetate trihydrate (100.7%), magnesium sulfate heptahydrate (99.9%), ammonium chloride (99.7%), and calcium chloride dihydrate (99.9%) were bought from Fisher Scientific. All solvents and chemicals were used without further purification.

 The MS parameters were optimized as follows: 1.00 ppm sample of each analyte was analyzed for the selection of the precursor ion, optimization of fragmentor voltage, optimization of cell accelerator voltage, optimization of gas temperature, optimization of gas flow rate, and optimization of collision energy. All the ions formed were analyzed for intensity. Ions with 

 equal to and greater than the MW of each analyte were considered to determine the precursor ion. See Figure 1 for the flowchart of optimization strategy.

A calibration curve for each analyte was obtained by the internal calibration method using the optimized MS parameters. Calibration was performed in the range of 1.00 ppt to 10.0 ppm for each analyte under study. Each of the standard solutions for antibiotics was spiked with the mixture of internal standards of antibiotics (azithromycin d

) to produce a final concentration of 100 ppb of each internal standard. Each of the standard solutions for EDCs was spiked with the mixture of internal standards of EDCs (metoprolol d

) to produce a final concentration of 10.0 ppb of each internal standard.

A synthetic wastewater matrix solution was prepared by dissolving potassium phosphate monobasic, sodium acetate trihydrate, magnesium sulfate heptahydrate, ammonium chloride, and calcium chloride dihydrate in MilliQ water. The concentration and quantity of reagents used for synthetic wastewater matrix preparation is given in 

Analysis of the antibiotics and EDCs was performed in triplicate for each parameter using an Agilent Technologies 1290-6460 Triple Quadrupole LC–MS/MS instrument using two ionization sources: a Jet Spray ESI source and an APPI source operated in positive mode. Full-scan mode was used for the optimization of MS parameters and MRM mode was used for calibration and analysis of wastewater. Data interpretation was performed using the instrument’s MassHunter Workstation Software. HPLC parameters for analysis are given in 

Calibration was performed in the range of 1.00 ppt to 10.0 ppm in matrix for all the analytes under study. Internal standards were added as described previously. Limits of quantitation were calculated based on the quadratic regression lines and were calculated in the instrument software. LOQs lower than 1.00 ppt were reported as <1.00 ppt as they were below the lowest calibrator.

A setup of Waters Oasis Prime HLB cartridges and a SPE vacuum manifold was used for offline SPE to extract antibiotics and EDCs from the synthetic wastewater matrix calibration sample. Waters Oasis PRIME HLB cartridges, 1 mL barrel syringe with 30 mg universal polymeric reversed-phase sorbent, were employed. SPE pretreatment was performed by washing the column with 2 mL of HPLC-grade methanol, 2 mL of Millipore deionized water, and 2 mL of Millipore deionized water at pH 2 under gravity. Then the samples were loaded on the column under vacuum at 10 to 20 mL/min rate.

Washing and Elution Step for Antibiotics Sample:

After loading the sample on the column, the cartridge was washed with 2 mL Millipore deionized water for the antibiotic sample. The column was then eluted first with 2 mL of methanol, and then with 1 mL of 1:1 methanol–acetone under gravity and collected and combined in test tubes.

Washing and Elution Step for EDCs Sample:

EDCs sample cartridges were washed with 1 mL of 5:95 methanol–water solvent. After the first washing step, the column was dried for 15 to 30 min under a vacuum. The column was eluted first with 1 mL of ethyl 9:1 acetate–methanol under gravity (this eluate was collected in a test tube labelled as fraction 1), then washed with 1 mL of 5% methanol/2% acetic acid in water, then with 1 mL of 5% methanol/2% NH

OH in water under a vacuum and dried for 10 to 15 min under a vacuum. After drying, the column was eluted with 1 mL of 2% NH

OH in methanol and combined with eluate present in the test tubes labelled as fraction 1.

 Wastewater samples were collected from the influent of the aeration treatment at the Environmental Resources Training Center (ERTC), a training center for drinking water and wastewater treatment at Southern Illinois University–Edwardsville, USA. The samples were analyzed for the presence of antibiotics and EDCs to demonstrate the effectiveness of the method development on real wastewater samples.

The wastewater samples were sequentially filtered through VWR 417 (40 µm) filter paper, then through VWR 696 (1.2 µm) glass microfibre filter paper, and then through Ahlstrom 193 (0.7 µm) microfibre glass filter. The filtrate was separated into six bottles each with 250 mL of filtrate, three samples for analysis of antibiotics and three for EDCs. All samples were spiked with appropriate internal standards as described previously.

Samples were adjusted to pH 3.0 using 6.0 M sulfuric acid before performing the SPE. Waters Oasis Prime HLB cartridges (6 mL, 200 mg universal polymeric reverse-phase sorbent) were used for wastewater sample analyte extraction. The SPE method for wastewater was identical to the synthetic matrix sample preparation, except the quantity of reagent solvents used was five times greater, due to the increased volume and cartridge bed mass.

The optimized MS parameters for the precursor ion of antibiotics for ESI as an ionization source can be found in 

, while parameters for APPI are shown in 

. Ions with the highest intensity peak with a 

equal to or greater than the MW of the analyte were selected as potential precursor ions and MS parameters were optimized using these ions.

The optimized MS parameters for the product ion of antibiotics for ESI as an ionization source can be found in 

 less than the MW of the analyte and with the highest ion abundance were optimized to determine the optimized collision energy of the product ions.

The regression equations for antibiotics and EDCs without matrix were selected such that they were equivalent for the calibration curve performed with and without matrix. The calibration curve correlation coefficient (

) criteria were established as higher than 0.99 for all the antibiotics and EDCs without matrix using ESI, shown in 

. Ampicillin, ceftriaxone, cephalexin, sulfamethoxazole, oxytetracycline, tetracycline, and metronidazole have limits of detection lower than 1 ppt using ESI and sulfamethoxazole has the highest LOQ among the antibiotics analyzed (226.1 ppb). All standard calibration curves are shown in the Supplemental Information (See 

). Paroxetine, propranolol, and venlafaxine have a limit of detection lower than 1 ppt using ESI, and acebutolol has the highest LOQ (14.81 ppb).

Ampicillin, cephalexin, penicillin G, trimethoprim, and tetracycline have limits of detection lower than 1 ppt using APPI as the ionization source, and tylosin has the highest LOQ among the antibiotics analyzed (3401 ppb), as shown in 

. Citalopram, paroxetine, and propranolol have limits of detection lower than 1 ppt using APPI as the ionization source and venlafaxine has the highest LOQ (3,788 ppb) as shown in 

The efficiency of both ionization sources was determined by comparing the LOQ for each of the pharmaceuticals obtained using ESI and APPI. LOQ less than 1 ppt in both the ionization sources means the most efficient ionization source could not be determined. Many compounds that are thermolabile will degrade using APPI, meaning the compound will not be detected by APPI and ESI was the ionization source that was preferred here. Erythromycin, tylosin and metronidazole ionized efficiently by ESI based on the comparative LOQ result. This is hypothesized to be due to the p

 of each compound (Table 1) being greater than the pH of the mobile phase (3.80) used in the analysis of antibiotics, allowing it to protonate easily. Acebutolol, atenolol, metoprolol, and venlafaxine were ionized efficiently by ESI based on the comparative LOQ results obtained from the calibration curve performed with and without matrix. A complete breakdown of this analysis can be found in Table 2 and Table 3.

Penicillin G, sulfamethoxazole, and 1,2 dimethyl-5-nitroimidazole have lower LOQs when ionized by APPI; APPI is the preferred ionization source for these analytes. The p

 values of these compounds (Table 1) are less than the pH of the mobile phase (3.80), leading them not to be protonated in solution. In addition, sulfamethoxazole and 1,2 dimethyl-5-nitroimidazole each have a high degree of conjugation in their structures, facilitating the absorption of photons and molecular radical ion formation (M

) (24). This specific trend was not observed in the case of trimethoprim, indicating that some other preferred ion formation pathway must be present. Citalopram also has a higher degree of conjugation in its structure, facilitating the absorption of photons and M

, making the APPI source highly efficient for the analysis of these compounds.

-tests were conducted at a significance level (α) of 0.05 on the data of the calibration curve performed with an artificial matrix and without a matrix. 

-values of the test for both the ESI and APPI ionization sources for analysis of antibiotics and shows that all population means are equal; therefore, there is no significant difference between data obtained with or without a matrix. 

 shows the population means are also equal between the data obtained with and without artificial matrix for EDCs.

The matrix effects on antibiotics and EDCs using ESI and APPI were calculated using the LOQ obtained from calibration curves performed with and without matrix. The trends can be seen in Tables 

Real wastewater samples from the ERTC were analyzed for the detection of antibiotics and EDCs. Table 4 shows the concentration of antibiotics in real wastewater samples and Table 5 shows the EDCs calculated using the calibration curve equation obtained from the calibration curve performed with and without synthetic matrix using ESI and APPI.

Paired t-tests were used to compare the concentration of antibiotics and EDCs in sample, calculated using the calibration curve performed with artificial matrix and without matrix. Paired t-tests were conducted at a significance level (α) of 0.05 to determine if there was a significant difference between the concentration of antibiotics and EDCs in the real wastewater samples collected from the ERTC wastewater treatment plant using the calibration equations obtained with artificial matrix and without matrix. Only three of the 

, demonstrated a statistical difference between using the matrix calibration curve versus the water curve shown in 

, illustrating that while important for some analytes, overall it had insignificant effect in this specific study.

There are few studies comparing the efficiencies of ionization sources for the MS analysis of low MW pharmaceutical analytes. While many researchers are limited to using ESI as their only ionization source, the use of complementary ionization techniques produces better results for the quantitation of analytes and should thus be considered for future studies. When purchasing an instrument costing $250,000 and up, the addition of a $25,000 additional ion source to improve analyte coverage in the analysis should be viewed as a reasonable added cost, considering that it nearly doubles the analytical capabilities of the instrument in terms of analyte coverage.

It was found that ESI is preferable for the analysis of pharmaceuticals, such as antibiotics, beta blockers, and SSRI antidepressants. However, ESI is not suitable for the ionization of all the pharmaceuticals with high sensitivity. APPI is an excellent complement to ESI, as it is highly efficient in the ionization of analytes that ESI is unable to ionize.

There was no significant difference observed in the presence of matrix effects at very low analyte concentrations. With higher concentrations of analytes, however, matrix effects should be taken into consideration when using these methods due to the significant difference observed. Since the LOQ of most of the analytes was less than 1 ppt, further study is needed to determine the ionization efficiency of ESI and APPI for these compounds by calibrating at lower concentrations.

By determining the ionization energy of an analyte using the appropriate software, it can be predicted which compounds will ionize by APPI or ESI preferentially. This determination will aid in the analysis of other classes of environmental pollutants, including other groups of pharmaceuticals and pesticides.

Using ESI and APPI as complementary ionization techniques yields a more complete picture of what compounds are present when analyzing in full scan mode, as well as better quantitation of analytes when appropriately optimized. This information holds value because employing multiple ionization techniques is an easy fix that creates a cost-effective method for analyte detection that can improve the outcome of future research studies.

C.S. Ho, C.W.K. Lam, M.H.M. Chan, R.C.K. Cheung, L.K. Law, L.C.W. Lit, and H.L. Tai, 

A. Leinonen, T. Kuuranne, and R. Kostiainen, 

A. Garcia-Ac, P.A. Segura, L. Viglino, C. Gagnon, and S. Sauvé, 

A. Yamamoto, N. Kakutani, K. Yamamoto, T. Kamiura, and H. Miyakoda, 

E.Y. Klein, T.P. Van Boeckel, E.M. Martinez, S. Pant, S. Gandra, S.A. Levin, and R. Laxminarayan, 

C.S. Wiysonge, H.A. Bradley, J. Volmink, B.M. Mayosi, and L.H. Opie, 

A.B. Boxall, M.A. Rudd, B.W. Brooks, D.J. Caldwell, K. Choi, S. Hickmann, and G.T. Ankley, 

M. Street, S. Angelini, S. Bernasconi, E. Burgio, A. Cassio, C. Catellani, F. Cirillo, A. Deodati, E. Fabbrizi, V. Fanos, and G. Gargano, 

H.E. Gray, “Laboratory Methods for the Advancement of Wastewater Treatment Modeling. M.S Theses and Dissertations,” (Comprehensive) [Online], Wilfrid Laurier University, Waterloo, Ontario, Canada, 2012.

L. Silvestro, I. Tarcomnicu, S. Rizea Savu, “Tandem Mass Spectrometry- Molecular Characterization, Chapter: Matrix Effects in Mass Spectrometry Combined with Separation Methods — Comparison HPLC, GC and Discussion on Methods to Control these Effects,” IntechOpen, London, United Kingdom, 2013.

T.J. Kauppila, H. Kersten, and T. Benter, 

 are with the Department of Chemistry at Southern Illinois University Edwardsville, in Edwardsville, Illinois, USA. Direct correspondence to: kevtuck@siue.edu

Articles

Atmospheric pressure photoionization (APPI) and electrospray ionization (ESI) are assessed for the mass spectrometry (MS) quantification of pharmaceuticals frequently detected in environmental waters, including antibiotics, beta blockers, and selective-serotonin reuptake inhibitors.

Ionization Efficiency for Environmentally Relevant Compounds Using APPI Versus EI

Per- and polyfluoralkyl substances (PFAS) are compounds that contain perfluorinated or polyfluorinated carbon chain moieties such as F(CF

. PFAS are widely used in consumer and industrial applications. There has been increased concern over the levels of PFAS, such as perfluorosulfonate (PFOS) and perfluoro-octanoic acid (PFOA), in the global environment and their potential adverse effects on the environment. In the United States, recent EPA methods 533 and 537.1 were released for analyzing PFAS in drinking water samples. A semi-automated system that combines elements of fully manual and automated procedures was developed and tested. Both methods use 250 mL water samples that were spiked with 2–32 ng/L PFAS standards. Styrenedivinylbenzene (SDVB) cartridges were installed on the system and the vacuum was turned on. The cartridges were conditioned, and then the samples were loaded across the cartridges (25–50 min). The sample bottles were automatically rinsed with water and the rinses were also loaded. Cartridges were dried and eluted for collection. Samples were reduced in volume using an automated concentrator. A total of ~30 PFAS were analyzed with recoveries well within the method acceptance windows (70–130%). Data showed good reproducibility with native background contributions from the system (<0.20 ng/L). Total processing time was between 50 min (537.1 method) and 75 min (533 method).

Per- and polyfluoralkyl substances (PFAS) are compounds that contain perfluorinated or polyfluorinated carbon chain moieties, such as F(CF

. Recent years have seen a great increase in the need for fast and reliable PFAS analyses in various types of environmental samples, especially in drinking water.

PFAS have been widely used in various consumer and industrial applications. These applications include firefighting foams, floor polishes, grease-proof coatings for paper products approved for food contact, insecticide formulations, mining and oil well surfactants, stain-resistant coatings for textiles, leather, and carpets. Some of the PFAS, such as perfluorosulfonate (PFOS) and perfluoro-octanoicacid (PFOA), have been studied extensively for their increasing levels in the environment. In animal studies, adverse effects on multiple organs (liver and pancreas), developmental problems in rodent offspring exposed in the womb, endocrine disruption, and reduced immune function, have been found (1). Possible effects on human health are still being investigated (2).

PFAS have been added in 2009 and 2019 to the 2004 Stockholm Convention to which the majority of countries in the world have acceded (3). The U.S. Environmental Protection Agency (EPA) has established an advisory level of drinking water not to exceed 70 ng/L of PFAS (4). Recent years have seen the releases of EPA methods 533 and 537.1 for PFAS analyses in drinking water using solid-phase extraction (SPE), which are being followed by commercial and governmental environmental laboratories in North America (5,6).

Manual methods for PFAS extraction generally involves using a vacuum manifold with cartridges placed on top to load the samples and then elute them into collection vessels. Because this system is open to the laboratory background, it is prone to lead to sample contamination. Water present in the extracts needs to be removed after elution of the cartridges in a separate step. The extractions typically require the attention of a laboratory technician throughout the entire procedure. Fully automated methods have the advantage of a closed system, which reduces the chances of background contamination. Solvent use for cleaning the system is likely to be less than with the manual method since the automated systems run cleaning solvents through the sample lines and connections only. Water removal from the extracts can be done with inline drying cartridges (sodium sulfate). Automated systems are typically run with a built-in computer which allows for unattended extractions freeing up workers for other tasks (7).

Combining the best features of both the manual and automated methods can lead to a semi-automated system that is relatively inexpensive and able to extract drinking water samples for PFAS reliably and with good reproducibility. Many laboratories do not have sufficient funds available to purchase fully automated systems. However, the semi-automated method described here is a good solution for laboratories that do not have as much financial capital.

The system can extract six or 12 samples of 250 mL drinking water in polypropylene bottles with Delrin fittings (Fluid Management Systems). The tubing is made of polyethylene so that background concentration of PFAS (as present in Teflon) is greatly reduced. The system operates under the vacuum, which stays on during the entire procedure. Various features of a semi-automated SPE system are shown in Figure 1. The system has a stage 1 manifold with stopcocks for cartridge conditioning, sample loading, and automated bottle rinsing. Stage 2 is the manifold used for elution, and it contains a glass tank with six or 12 positions for collection vessels. Each of the manifolds are comprised of a vacuum regulator and a vacuum gauge. On the front of the base of the system, there is a nitrogen pressure gauge, a nitrogen gas regulator and valve (for drying and automated bottle rinsing), a valve that switches the vacuum suction between stages 1 and 2, and a valve that sends the waste to either aqueous waste or organic waste.

FIGURE 1: An illustration of the various features of the semi-automated solid-phase extraction (SPE) system.

Samples were prepared in deionized water (250 mL) containing 1 g/L ammonium acetate. Acetic acid was used to adjust the pH from 6 to 8. The sample bottles were then spiked with relevant native 533 PFAS standards. Next, 500 mg (6 mL volume) styrene-divinylbenzene (SDVB) cartridges (Fluid Management Systems) were put in each of the manifold one positions. The vacuum was turned on, and the cartridges were conditioned with 10 mL methanol, 10 mL 0.1 M phosphate buffer, and 3 mL phosphate buffer with 2 mL of water. The cartridges were kept wet during the entire conditioning process. The samples were then loaded across the cartridges at 5 mL/min. Sample bottles were then rinsed automatically with 10 mL of 1 g/L ammonium acetate in water, which was loaded across the cartridges, followed by rinsing bottles with 1 mL methanol that was then also loaded. Cartridges were then dried for 5 min under nitrogen. The cartridges were then transferred to the stage 2 side for elution. The rinse bottles were filled with 5 mL 2% ammonium hydroxide in methanol, and this solution was sprayed across the walls of the samples bottles automatically using nitrogen gas. The rinses were then pulled across the cartridges under vacuum. The procedure was then repeated resulting in a total of 2 x 5 mL of elution across the cartridges. The eluents were collected for further concentration and analysis.

For the analysis, 250 mL water samples were prepared and spiked with 537.1 PFAS standards. Sample bottles were loaded onto the system and rinse bottles were filled with 7.5 mL water. Next, 6 mL 500 mg SDVB cartridges were put in each of the manifold one positions, and the vacuum was turned on. The cartridges were conditioned with 15 mL methanol (soak 1 min) and then with 18 mL water. Samples were loaded across the cartridges under vacuum (20–25 min, ~8 in. mercury). The sample bottles were then rinsed automatically with nitrogen with 7.5 mL water twice (total 15 mL water) and loaded across cartridges (dry for 5 min under vacuum). The cartridges were then transferred to the stage 2 side for elution. Rinse bottles with 4 mL methanol were used to rinse sample bottles automatically and rinses were loaded across the cartridges (twice, 1 min soak, 8 mL methanol total). Eluent was collected for analysis into collection vessels.

Extracts were concentrated to dryness in an automated concentrator (Fluid Management Systems). The concentrator has 24 positions for concentration at 65 °C under a nitrogen flow of 9 psi. The polypropylene collections vials are graduated for 15 mL and are conical in shape. The system uses a dry bath heating element with a time-based endpoint, and the temperature and time log were saved.

A Waters Acquity ultrahigh-pressure liquid chromatography (UHPLC) system coupled to a quadrupole time-of-flight (QTOF) Xevo G2-XS mass spectrometer and high resolution mass spectrometry (HRMS) was used for analyzing PFAS. Separation was performed with a Waters Acquity HSS T3 column (2.1 mm x 100 mm, 1.8 μm) and samples were analyzed by electrospray in negative ionization mode. Samples were reconstituted in 1 mL methanol (1:3 ratio) and then sonicated prior to injection (20 μL).

A gradient solvent program was used with 0.1% for- mic acid in liquid chromatography–mass spectrometry (LC–MS) grade methanol (solution A) and 0.1% formic acid in LC–MS grade water (solution B). The details of the gradient method are seen in Table I.

Samples were analyzed in negative electrospray ionization (ESI) mode. Desolvation gas (nitrogen) flow rate was kept at 750 L/h, capillary voltage at 0.5 kV, and the source and desolvation temperatures at 120 

C, respectively. Mass range was set from 50 to 950 Da with a scan rate of 0.055/s.

The concentration and recovery of each PFAS was determined by comparing the spiked sample response to that of a reference standard at the same concentration (0.5 ng/mL). A labeled PFAS (d

-N-MeFOSA-M) was spiked to each sample and reference standard in order to obtain a normalized response for accuracy (8).

Figure 2 shows recoveries for PFAS analyzed in accordance with EPA method 533. This method is comprised of perfluorocarboxylic acids that have a linear chain of carbon atoms all substituted with fluor and a terminal carboxylic acid group, perfluorosulfonic acids that are similar in structure but have a terminal sulfonic acid functional group, and three other compounds (4:2 FTS, 6:2 FTS, and 8:2 FTS) that are sulfonic acids in which the two carbon atoms nearest to the sulfonic functionality have no fluor atoms substituted. Average recoveries were between 80–105% for a set of samples with relative standard deviations (RSDs) shown as error bars. The recoveries were all well within the acceptance window of the method, which is 70–130% in this instance.

FIGURE 2: Recoveries of native PFAS compounds in EPA method 533 extracts at 25 ng/L. The abscissa label is native PFAS compounds.

Table II shows recoveries for method 537.1 PFAS run the on the system. Samples were run at 2, 4, 24, and 32 ng/L. Data are between 80–120% and again all within the 70–130% acceptance window of the method. The data also shows that good recoveries can be obtained for both lower and higher PFAS concentrations.

Figure 3 has background concentrations for method blank runs done on the system. The contributions from the system are low: the highest blank concentration found is that of PFOA at 0.2 ng/L. The other PFAS shown are present at lower concentrations, mostly <0.05 ng/L. The system is clearly suitable for trace level determination of PFAS in water. The use of polyethylene and polypropylene materials is clearly very effective in keeping native background of PFAS to a minimum.

FIGURE 3: Native background of 537.1 PFAS compounds using the semi-automated system. The abscissa label is native PFAS compounds.

Sample data shows high recoveries for all spiked native PFAS for both 533 and 537.1 methods. All native recoveries are well within the 70–130% acceptance windows for both methods. RSDs are mostly low showing that the semi-automated system gives very good reproducibility. Native background contributions from the system are very low. The automated sample concentration step (to dryness) accounts for little loss of these compounds. The semi-automated system provides an inexpensive alternative to both manual and automated methods, with fast processing times between 50 and 75 min (depending on the different loading times described in the methods). Samples can be taken from collection bottle to a LC vial in one quick, consistent, reproducible process.

Wastewater samples can be run on the semi-automated system too, using larger 25 mL volume cartridges—still with 500 mg SDVB—that make it possible to fill up the cartridge barrel with Delrin plastic filtration wool. Using a semi-automated system to run wastewater samples will slow down movement of the particulate matter in the wastewater and prevents plugging of the cartridge.

(1) L. Cui, Q.F. Zhou, C.Y. Liao, J.J. Fu, and G.B. Jiang, 

(2) K.E. Pelch, A. Reade, T.A.M. Wolffe, and C.F. Kwiatkowski, 

(4) Fact Sheet PFOA & PFOS Drinking Water Health Advisories (U.S. Environmental Protection Agency, Washington, D.C., 2016).

Method 533: Determination of Per- and Polyfluoroalkyl Substances in Drinking Water by Isotope Dilution Anion Exchange Solid Phase Extraction and Liquid Chromatography/ Tandem Mass Spectrometry

 (U.S. Environmental Protection Agency, Washington, D.C., 2019).

(6) J.A. Shoemaker, D.R. Tettenhorst, P.E. Grimmett, and B.K. Boutin, 

Method 537.1: Determination of Selected Per- and Polyfluorinated Alkyl Substances in Drinking Water by Solid Phase Extraction and Liquid Chromatography/Tandem Mass Spectrometry

 (LC/MS/MS) (U.S. Environmental Protection Agency, Washington, D.C., 2018).

(7) P.M. Germansderfer, R. Addink, T.G. Hall, and H.R. Shirkhan, 

 is Technical Director of Toxic Report Laboratories in Billerica, Massachusetts. Direct correspondence to: 

 is Manager of Toxic Report Laboratories and Vice-President of Sales at Fluid Management Systems in Billerica, Massachusetts.

There is a need for fast and reliable PFAS analyses to monitor drinking water quality. We report on the use of semi-automated sample preparation with EPA methods 533 and 537.1, assessing performance and processing time.

Analyzing Per- and Polyfluoroalkyl Substances in Drinking Water Using EPA Methods 533 and 537.1 with Semi-Automated Solid-Phase Extraction (SPE)

I grew up professionally in a laboratory with two HPLC systems, along with capillary and packed-column gas chromatography (GC) instruments. I moved on to be a technical support chemist for J&W Scientific, a pioneering company in capillary GC columns (now part of the Agilent family). My graduate studies were centered in capillary zone electrophoresis, and I have since contributed to microfluidic separations. I have a deep history and real appreciation for separations science—I have taught it at the graduate level for twenty-five years. I am currently the Chair of the ACS Subcommittee on Chromatography and Separations Chemistry (SCSC).

That being said, I have struggled to write this article. Why? I wasn’t sure. I basically write for a living, so 900 words should not be that hard. But I think I have figured it out, and this isn’t going to be a popular opinion: I do not believe in classic linear separations anymore; further, I think separations science is wildly underdeveloped and underappreciated.

I do not think it will be a popular opinion because many people have spent their lives learning and developing these techniques, and many people (myself included) have made a good living working within these industries. Suggesting they are not as useful as they could be and may need to be replaced is discomforting.

So where is this coming from? Well, I guess it gets back to seeing and thinking about the stunningly small injection volumes and fast total analysis times from microfluidics and high-speed GC. Couple that with multidimensional separations in general, then add in hybrid 3D fabrication techniques for microscale systems and gradient style techniques (increasing concentration while separating), and one can quickly surmise the days of long linear runs of univariate data are numbered.

I was around for the introduction of capillary GC and capillary electrophoresis, which were to replace the packed column GC immediately (it took twenty years or more) and HPLC (it never happened). Things change slowly. Acceptance of new techniques is limited. If engineers and scientists have existing working solutions, they do not really need to change.

That being said, where do I think separations science will be in 10 years? 50 years?

Extending observations about injection bandwidth, separation speed, available resolution, developing gradient techniques, orthogonality of techniques, and hybrid 3D fab capabilities, well...at a pretty amazing place.

There will be “on demand” capability, where the desired target from a complex sample is isolated, concentrated, purified to homogeneity, and delivered to a detector or to a next step. The internal workings will be a highly efficient, multidimensional programmable separations scheme set on foundational physical interaction and hard-won empirical results. All this will be held in a volumetrically small system engineered and created using hybrid 3D fabrication based on the fundamentals of the separations sciences we all know and love. It will probably look like the Borg ship from “Star Trek,” but our beloved linear separations schemes will be gone. We have witnessed similar evolutions and revolutions in various spectroscopies (nuclear magnetic resonance [NMR] is an exemplar) and especially mass spectrometry (MS). Although, admittedly, they do not resemble the Borg ship. We can get there, but it will take several axes of separation and uber-efficient sample transfer.

If we can do separations on demand, we can do what I call “whole sample information extraction”—that is, analyze everything (total information theory) in a sample with a gigantic and complex scanning mode. Imagine starting with a blood sample, separate based on size, various similar cells gathered together, then separate on some physical principle such that all cells in a fraction are identical. Then, with each of these fractions, disrupt the cells and separate all the organelles and biochemicals to pure complex structures, disrupt these and separate the resulting complexes or pure materials and delivery to the detector (or multiple detectors). The now-pure homogeneous fractions can be fed into information-rich detection systems (MS, cryoEM, NMR, whole genome sequencing, lipidomics, and so on).

Since this system can be operated in a dynamic mode, the various fractions can be queried against a variable input (diseased or healthy, polluted or pure, immune response or non) and AI or machine learning strategies can be employed to determine the most important elements.

Well, I suspect I have made myself a pariah in my own scientific field—a lot of my friends are making a very good living using and improving our current techniques. However, if my students and younger colleagues’ enthusiasm is any indicator, we will be 3D printing our way into a very interesting future based on our beloved and respected principles of separations. I look forward to a much greater impact on our society as these capabilities are realized.

 is a Professor at the School of Molecular Sciences at Arizona State University. Direct correspondence to: 

Are the days numbered for long linear runs of univariate data in separation science?

The Future of Separation Science: Goodbye Old Friends

Suhas Nawada is with the Van’t Hoff Institute for Molecular Sciences at University of Amsterdam, in Amsterdam, The Netherlands.

Tristram Budel is with Atum3D, in Gouda, The Netherlands.

For approximately a decade, three-dimensional (3D)-printed columns have been hailed as the future of liquid chromatography (LC). However, the resolution of typical printing methods has fallen short of the requirements to make an effective analytical- or preparative-scale column. In this article, we describe a new 3D-printing method that can create large-volume columns with a feature resolution of 10 µm.

During the ongoing quest for fundamental advancements in high performance liquid chromatography (HPLC) column performance in the past two decades, several methods of producing ordered stationary phases have been explored. To quote John Knox, “The overall effect of a very homogenous bed is to greatly enhance chromatographic efficiency” (1). Over the past decade, the emergence of additive manufacturing, or three-dimensional (3D)-printing, led to many in the chromatography field to consider it as the eventual method to create ordered stationary phases with tailored geometries (2).

Studies have been performed on orientations (3), element shapes (4), and novel porous geometries that are best suited for 3D-printed columns. The efficiency of such structures were demonstrated in both computational simulations and experimental tests, with reduced chromatographic plate heights as low as 1.12. Furthermore, a bifunctional resin has been developed that allows for direct printing of anion exchange columns (5,6).

However, a limitation in such studies has been the feature size of the porous beds that were created. This limitation is because of the limited resolution of most 3D-printing methods. Common printing methods, such as fused deposition modelling, stereolithography, and selective laser sintering, typically operate with nominal resolutions in the order of 20–100 µm. By operating with nominal resolutions, common printing methods generate feature sizes of porous beds typically greater than 300 µm, which is two orders of magnitude short of the feature sizes that are necessary for use in HPLC columns.

In contrast to conventional methods, the high-resolution 3D-printing method called 

is capable of creating pore sizes of under 1 µm with high fidelity. However, as detailed in a previous article by De Malsche and coworkers (7), the printing method requires impractically long production times for analytical-scale columns typically used in HPLC. Therefore, the current capabilities of 3D-printing technologies are inadequate to create analytical- or preparative‑scale columns because of limitations in both resolution and print speeds.

To address these limitations, we have developed a 3D-printing technique called 

(HSLA) (8), which combines traditional stereolithography and photolithography, and is capable of high-resolution and fast printing of high-volume structures.

 (a) Prototype hybrid stereolithography (HSLA) setup. A 40-mm i.d., 40-cm length column with a pore size of 50 µm is being printed using a commercial UV-curable resin. (b) An axial cutaway of the 50-µm printed lattice, and (c) optical profilometer scan of a radial cross-section of a 20 µm printed lattice.

Figure 1a shows a prototype HSLA setup during the printing process. A lower resolution pattern is illuminated, using an ultraviolet (UV) light source (shown in blue in Figure 1a) and a digital mirror device. To achieve higher resolution than are achieved with traditional stereolithographic methods,
a high-resolution photomask with predefined patterns is used between the resin tray and the digital pattern. After a layer is cured, the build platform lifts to peel the printed piece from the resin tray.

For example, the 40-mm i.d. column seen in Figure 1a was printed at a speed of 5 mm of column length per hour, translating to 6.5 mL/hr. The 40-mm i.d. column represents an improvement of greater than three orders of magnitude compared to the fastest two photon polymerization systems for a similar resolution (9). The high printing speed can be used to create large preparative-scale columns, or several analytical-scale columns in parallel—up to 32 columns with a 4.6-mm i.d., like the one shown in Figure 1a.

Figures 1b and 1c show axial and radial cutaways of two different lattices printed using HSLA. Figure 1b shows a 50-µm simple cubic grid printed in the setup shown in Figure 1a, with a designed porosity of 50% and a layer thickness of 25 µm. Figure 1c shows a lattice with a feature size of 20 µm with a layer thickness of 10 µm. In both cases, the ordered nature of the printed structures is evident, with the photomask features being imprinted on the structure with good fidelity.

To create a 3D structure, several patterns were defined on the photomask, with a linear stage moving the photomask after a layer is cured. A photomask-switching system with two patterns, shown in Figure 2, was used to create the lattice seen in Figure 1b. More-complex 3D structures can also be created using more photomask patterns. Because LC column microstructures typically consist of simple, repeatable geometries, a photomask-switching system is well-suited for column production.

 Example of digital and photomask patterns used in HSLA for creating a walled column with a simple cubic grid using two photomask patterns.

A commercial nonporous high-resolution photopolymer resin called “Nanoclear” (FTD Resins) was used for all tests. For this study, 1-, 2.1-, 4.6-, 10-, and 20-mm i.d. columns were produced and thoroughly flushed using isopropanol to clear the porous structure of uncured resin. An external casing with fittings and flow distributors was used to connect the column to an HPLC system.

To perform residence time distribution (RTD) tests on the printed 4.6-mm i.d. × 5 cm L structures, 1-µL injections of 1 mg/mL of uracil were performed on a Waters Acquity H-Class system across range of flow rates (0.01–4 mL/min). For the RTD tests, two flow orientations were tested—with and against the print direction of the column, which is shown in Figure 3a. In addition to the RTD tests, pycnometric measurements were performed to determine the total void volumes of the
printed columns.

Uracil injections were performed in triplicate on the 4.6-mm i.d. columns from two directions. The resulting plate heights show a classical van Deemter curve, with a clear column efficiency minimum (Qmin) of 0.25 mL/min. As seen in Figure 3b, aligning the flow direction with the print direction (that is, top to bottom, as seen in Figure 3a) is clearly advantageous compared to orienting the column against the direction of the print.


 Residence time distributions by flow orientation. (a) The orientation of the printed column with respect to the build platform, and (b) van Deemter curves with fluid flow with and against print direction for 4.6-mm i.d. × 5 cm length columns with a 50-µm body-centered cubic lattice.

 = 2.25) was demonstrated with the column oriented in the direction of the print. To the best of our knowledge, these represent the lowest absolute plate heights for 3D-printed columns, and a five-fold improvement compared to previous tests using conventional printing methods (3). In the region of the optimum flow rate, the column also exhibited low relative standard deviations of 4 and 7%, for the first moment and plate height, respectively.

Switching the flow direction of the column resulted in a near-doubling of plate heights. This difference can be explained by the minor printing artifacts that occur because of a phenomenon called 

, in which a small potion the bottom of a pore in a lattice layer cures and solidifies (10).

In addition to determining the plate heights, the scalability of the printing method was tested by measuring the void volumes of columns of different internal diameters. Columns with internal diameters of 1–20 mm × 2.5 cm L were printed in triplicate and measured. As seen in Figure 4, the volume of the printed columns was consistently between 90 and 100% of the designed volume, indicating good fidelity between the designed photomasks and the printed structures. However, larger columns (10- and 20-mm i.d.) showed greater variation in measured void volumes. A possible cause for larger columns showing a greater variation in void volumes is the post-processing method that was used to remove the uncured resin from the printed lattices. Isopropanol was flushed through the columns using an analytical-scale HPLC system for a fixed column volume. However, to achieve constant flushing velocity, the optimum flow rates for the larger i.d. columns were significantly higher than the maximum flow rate limits that HPLC systems will allow. As a result, the larger printed columns delivered incomplete removal of the uncured resin. Tests with preparative-scale systems are necessary, not just to perform separations on such larger i.d. columns, but to achieve the desired porosities and void volumes.

 Pycnometric measurements of column void volume compared to the designed volumes of printed lattices at different internal diameters, with a constant length of 2.5 cm.

This article outlines the development and operation of a novel 3D-printed method, HSLA, geared toward the production of liquid chromatography columns. The method is capable of operating on length scales of micrometers to >40 cm to produce analytical- and preparative-scale column structures in reasonable timeframes, representing a significant improvement compared to existing 3D-printing methods. The characterization of the printed structures shows good agreement between the designed and measured void volumes and reduced plate heights of 

 = 2.25. The results here show that full-scale analytical HPLC and even industrial preparative-scale 3D printed columns with an organized internal structure that is consciously designed to fit the application are soon to be a widespread reality.

L. Scorza, U. Simon, M. Wear, A. Zouliatis, S. Dimartino, and A. McCormick, 

W. De Malsche, F. Matheuse, K. Broeckhoven, G. Desmet, D. Cabooter, and S. Eeltink, 

S. Nawada, EU patent: EP 19170376.8-1022, 19 April 2019.

Three-Dimensional Microfabrication Using Two-Photon Polymerization, Micro and Nano Technologies 

(William Andrew Publishing, Norwich, New York, USA, 2nd Ed., 2019), pp. 385–408.

H. Gong, M. Beauchamp, S. Perry, A. Wooley, and G. Nordin, 

 is with the Van’t Hoff Institute for Molecular Sciences at University of Amsterdam, in Amsterdam, The Netherlands.

is with Atum3D, in Gouda, The Netherlands.

 is a director of Research and Development at Restek. He also serves on the Editorial Advisory Board for 

 and is the Editor for “Column Watch”. Over the past 20 years, he has worked directly in the chromatography industry, focusing his efforts on the design, development, and application of chromatographic stationary phases to advance gas chromatography, liquid chromatography, and related hyphenated techniques. His main objectives have been to create and promote novel separation technologies and to conduct research on molecular interactions that contribute to retention and selectivity in an array of chromatographic processes. His research results have been presented in symposia worldwide, and have resulted in numerous peer-reviewed journal and trade magazine articles. Please direct correspondence to: 

Three-dimensional (3D) printing shows promise for creating stationary phases
with tailored geometries. We present a method that overcomes limitations of
other approaches, and can produce columns with a feature resolution of 10 μm.

Novel 3D-Printing Method to Create Liquid Chromatography Columns

In a previous study, polyphyllin III, an important natural dioscin ingredient, was bonded to silica particles using the “one-pot method” for preparing a stationary phase, called SP-D. SP-D was further modified to produce two stationary phases, namely the phenyl isocyanate-dioscin (Phe-D) and 3,5-dimethyl phenyl isocyanate-dioscin bonded silica stationary phase (DMP-D), respectively. The Phe-D and DMP-D were evaluated by achiral and chiral analytes. The high performance liquid chromatography (HPLC) method was optimized and applied to the analysis of the main active ingredient, polyphyllins, that were contained in gongxuening capsules, a traditional Chinese medicine (TCM) widely used to treat gynecological diseases. The two synthesized stationary phases were also applied to the separation of the amino acid (AA) enantiomers. For this study, 10 α-AAs (lysine, leucine, cysteine, arginine, isoleucine, threonine, serine, valine, alanine, and histidine) were selected and studied for chiral separation using Phe-D and DMP-D stationary phases. In this study, 10 observed AA enantiomers were separated to different degrees on at least one stationary phase whereas four AAs were enantioseparated on both Phe-D and DMP-D. These results indicated that the two synthesized stationary phases have potential applicability in quality control (QC) of TCM and chiral separation as well as offering a new application choice for analyses of natural products.

As the pharmaceutical and natural medicine industries continue to evolve, new separation issues posed by various samples need to be resolved. As a result, the demand for more reversed-phase high performance liquid chromatography (HPLC) stationary phases is becoming significant. The ongoing research in designing, synthesizing, and testing the performance of stationary phases, is therefore crucial (1,2).

It is possible to prepare a stationary phase that possesses both support and active functions by bonding the active ingredients to the surface of the silica gel base material. Some stationary phases with ligands extracted from medicinal plants such as emodin, curcumin, saponin, magnolol, allicin, quercetin, and isatin have been prepared (3–6). This type of stationary phase can provide not only the functions of hydrophobic interaction, but also hydrogen-bonding interaction, π-π, and n-π conjugation. They are quick and highly selective for structurally similar compounds (7–9). Modified or derivative natural products can be bonded to the surface of the silica gel to prepare HPLC chiral stationary phases (CSP). These CSPs, based on products such as cyclodextrins, polysaccharides, and macrocyclic glycopeptide antibiotics and proteins, were proven to be highly efficient (10–15). The interactions, hydrogen-bonding, π-π stacking, inclusion interaction, and ion exchange, between the solute and CSP played a key role in chiral resolution (16,17).

, and several Discordance plants, including 

Dioscorea zingiberensis C.H. Wright (D. zingiberensis)

 (DNM), possess anti-inflammation, immunoregulation, hypolipidemic, antiviral, antifungal, and antiallergic effects (18). Dioscin is also used as an important ingredient for the synthesis of various steroid hormone drugs.

Figure 1 shows the structural formula of dioscin. As the figure shows, dioscin contains active hydroxyls that provide favorable conditions for binding events. The multiple hydroxyls and double bonds can participate in hydrogen-bonding interaction, dipole–dipole interaction, and π-π interactions, and the structure is large enough to shield the influence of the acidic Si-OH residue on the stationary phase and improve the separation process. The plane and rigid structure of dioscin can also improve the stereoselectivity by providing a larger contact surface for the solute molecules. In this case, we bonded dioscin to silica gel to obtain a dioscin-bonded stationary phase, called SP-D. Then, two derived stationary phases were produced, namely phenyl isocyanate-dioscin bonded silica stationary phase (Phe-D) and 3,5-dimethyl phenyl isocyanate-dioscin bonded silica stationary phase (DMP-D), respectively. In a previous study, these two derived stationary phases were evaluated, and it was determined that they had typical reversed-phase (RP) chromatographic performance, which makes both similar to the performance of octadecylsilyl (ODS) columns (19).

Quality control (QC), including qualitative and quantitative analysis, is a major bottleneck for traditional Chinese medicines (TCM) with their increased use worldwide (20). There are many ingredients, including active ingredients, auxiliary components, and inactive ingredients in TCM. Because of the sheer quantity of ingredients in TCM, it is too difficult to identify and characterize all of them. Therefore, a characteristic HPLC chromatogram is always chosen to compare and evaluate TCM (21,22). For quantitative analysis, a single component cannot represent the full efficacy of any TCM. Analysis of multiple active components, through quality markers (Q-markers) detection is therefore necessary (23,24). As a result, the combination of a characteristic chromatogram and quality markers detection is a powerful way to ensure quality of TCMs.

Chirality caused by molecular asymmetry has special effects on physical, chemical, biological, and pharmacological properties appearing at a molecular level (25). Amino acids (AA), the simplest chiral molecule, have at least one stereogenic center (with the exception of glycine). The important analytical task of the chiral separation of an amino acid is achieved mainly by chromatographic methods, especially HPLC, including direct and indirect separation (26).

In a previous study, we prepared two natural active ingredient dioscin-based stationary phases, Phe-D, and DMP- D, respectively. The bonding method, the “one-pot method,” used to produce the stationary phases directly without tedious intermediate steps, was proven to be simple and convenient. The Phe-D and DMP-D stationary phases were characterized by elemental analysis, scanning electron microscopy, thermogravimetric analysis, and infrared (IR) spectroscopy. These new stationary phases were evaluated by benzene and its homologs, polycyclic aromatic hydrocarbons, and other solutes as probes. Through these evaluations, Phe-D and DMP-D were verified to have typical RP chromatographic performance, similar to that of ODS columns, and the separation principle was shown to be related to the hydrophobic effect. In addition, these stationary phases were confirmed to have the ability to provide hydrogen bonding and π-π conjugation sites for solutes (19).

To evaluate and compare these two stationary phases, we investigated their application to achiral and chiral separation. In this study, we used HPLC characteristic chromatograms of 10 batches of gongxuening capsules and determined the content of six main polyphyllins in 10 batches of gongxuening capsules. Then, the application of Phe-D and DMP-D to chiral separation of 10 amino acids was studied and evaluated.

Polyphyllin I and II (Guizhou Dida Biochemistry Co.), polyphyllin III (Nanjin Biochemistry Co.), polyphyllin V (Chengdu Herbpurify Co., Ltd.), polyphyllin VI and VII (Sichuan Weikeqi Biotech Co., Ltd.), and gongxuening capsules (Yunnan Baiyao Group Co., Ltd.) were used. The AA enantiomers lysine, leucine, cysteine, and arginine came from Shanghai Lanji Technologies Co., Ltd. The AA enantiomers isoleucine and threonine came from the Chinese Academy of Sciences’ Institute of Biochemistry. For the remaining AA enantiomers, serine came from the Shanghai Chemical Reagents Procurement Station, valine came from Feinbiochemical Heidelberg, alanine from the Shanghai Third Reagent Factory, and histidine from the Shanghai Renfa Comprehensive Utilization Factory.

All the other reagents were analytically pure and procured through commercial channels.

Experimental Instrument and Chromatographic Conditions

The Agilent 1260 HPLC system, which included the G1322A Vento, a G1312C delivery pump, a G1329B standard automatic sampler, a G1316A column oven, a G1315D detector, and the Agilent 1200 LC chromatographic work station, was used as the instrumentation for this study. The separation was performed nate-dioscin bonded silica gel stationary phase, 4.6 × 250 mm; and DMP-D: 3,5-dimethyl phenyl isocyanate-dioscin bonded silica gel stationary phase, 4.6 × 250 mm). The mobile phase consisted of acetonitrile and 0.01% phosphoric acid for QC, and for the chiral separation, methanol and a 0.04 mol/L buffer solution were used. The flow rate was 0.4 mL/min. The injection volume was 10 μL. The column oven was maintained at 35 °C and the ultraviolet (UV) detector was set at 203 nm for quality control. The column oven was maintained at 25 °C, and the UV detector was set at 210 nm for chiral separation. Column efficiency was determined with biphenyl, the theoretical plate number of Phe-D is 4776 block/m, and the theoretical plate number of DMP-D is 5028 block/m.

Preparation of Reference and Sample Solutions

The appropriate amount of Polyphyllin I, II, III, V, VI, and VII were dissolved in methanol to prepare the reference solutions of various concentrations (0.375 mg/mL, 0.375 mg/mL, 0.400 mg/mL, 0.300 mg/mL, 0.450 mg/mL, and 0.450 mg/mL, respectively).

The content of 20 capsules from 10 different batches of gongxuening capsule was weighed and porphyrized. Next, 1 g of the substance was added into a conical flask with cover along with 25 mL of methanol. The flask was then sealed and weighed. This solution in the flask was treated ultrasonication for 40 min and then cooled to room temperature. Afterward, methanol was added to the solution to compensate any weight loss. The solution was shaken well and filtered with a 0.22 μm microfiltration membrane to remove particulates and the subsequent filtrate was kept as the sample solution.

Preparation of Sample and Buffer Solution for Chiral Separation

Improved analysis of pharmaceutical and natural medicine products requires advances in reversed-phase LC stationary phases. We examine two synthesized stationary phases with applicability in quality control and chiral separation for analysis of natural products.