**Silk Fibroin Nanoparticles for Drug Delivery: Effect of Bovine Serum Albumin and Magnetic Nanoparticles Addition on Drug Encapsulation and Release**

#### **Olga Gianak 1, Eleni Pavlidou 2, Charalambos Sarafidis 2, Vassilis Karageorgiou <sup>3</sup> and Eleni Deliyanni 4,\***


Received: 4 February 2018; Accepted: 9 April 2018; Published: 23 April 2018

**Abstract:** Silk fibroin nanoparticles were prepared in the present study based on phase separation between silk fibroin and polyvinyl alcohol. The drug encapsulation efficiency of the prepared nanoparticles was examined at a range of concentrations from 10 ppm to 500 ppm for pramipexole, curcumin, and propranolol hydrochloride. Silk fibroin nanoparticles encapsulated with propranolol presented the highest drug release profile. In order to improve the drug encapsulation efficiency and drug release performance, a modification of silk fibroin nanoparticles with bovine serum albumin and magnetic nanoparticles was tried. The modification was found to improve the drug encapsulation and release of the modified nanoparticles. Bovine-serum-modified nanoparticles presented the best improvement.

**Keywords:** silk fibroin; drug delivery; magnetic silk fibroin; bovine serum albumin

#### **1. Introduction**

Drug delivery aims to bring the compound with pharmaceutical activity to the exact point of need in the organism with the right concentrations in order to increase the efficiency of action [1]. The improvement of cellular uptake, the reduction of the side effects, the control of drug release, the enhancement of drug bioavailability, and the reduction of the drug degradation rate are the purposes of drug delivery systems [2]. Recently, nanoparticles are considered to be suitable for drug delivery due to their ability to act as modifiable platforms, their tunable size, and their high surface-to-volume ratio [2]. In addition, they can be used to deliver hydrophilic and hydrophobic drug molecules as well [3].

Silk fibroin, from the silk worm Bombyx mori, consists of hydrophobic and hydrophilic regions [2]. The hydrophobic domains, i.e., protein crystals and beta sheets, are dominated by repeats of alanine, glycine-alanine, and glycine-alanine-serine. Recently, silk fibroin, due to its excellent properties, such as biocompatibility, biodegradability, and low immunogenicity, has been extensively tested for a drug delivery application since silk materials exhibit high encapsulation efficiency and controllable drug release kinetics [4–6].

Bovine serum albumin (BSA) is a water-soluble protein with a well-defined structure. Positively or negatively charged drug molecules can be non-covalently bound to the charged amino acids of albumin. Silk fibroin contains hydrophobic amino acids that could create strong electrostatic interactions through the amino acids' carboxyl groups and the amino groups of bovine serum albumin. These interactions prevent leakage of hydrophobic drugs from bovine serum albumin; thereby, it is expected to improve drug encapsulation, delivery to affected body areas, and drug release rate [7]. Albumin was found to enable the localization of drug carriers in specific tissues (i.e., liver and heart tissues), regulating in this way biodistribution and drug release [8]. Besides this, it was shown that readily available and inexpensive serum albumin can overcome many drugs' shortcomings, such as insolubility, instability in biological environments, poor uptake into cells and tissues, sub-optimal selectivity for targets, and unwanted side effects [9].

Additionally, magnetic carriers and external magnetic fields that focalize on tumors have emerged as a hopeful strategy to enhance drug accumulation at tumor sites. Magnetic targeting has the advantage of not requiring complex chemical modification of targeting ligands on the surface of nanocarriers when compared with conventional tumor targeting. Exploiting a magnetic field as a driving force represents a noninvasive therapeutic approach [10,11].

In the present study, we report aqueous-based preparation methods for silk fibroin nanoparticles, based on phase separation between silk fibroin and polyvinyl alcohol (PVA), with a simple, inexpensive, and appropriate method for pharmaceutical and biomedical applications of drug delivery. Curcumin, pramipexole, and propranolol hydrochloride were chosen as model drugs in this study for their different molecular weights (MWs), surface chemistries, and disease targets. Curcumin, a natural compound with diphenolic groups, is extracted from the rhizome of turmeric. It has been used in clinical trials for cancer therapy [12,13]. Propranolol is a non-selective beta adrenergic blocking agent and has been used in the treatment of hypertension, angina pectoris, and many other cardiovascular disorders [14]. Pramipexole is a potent dopamine D2 agonist with a preference for D3 receptors and is approved worldwide for the treatment of Parkinson's disease symptoms [15].

Moreover, nanoparticles of silk fibroin were modified with bovine serum albumin for the enrichment of silk fibroin with the above-mentioned properties of albumin [7–9]. Silk fibroin-bovine serum albumin nanoparticles were prepared since bovine serum albumin has been found to be biocompatible, biodegradable, non-toxic, and non-immunogenic and has interesting results in drug encapsulation and release [16]. Finally, magnetic silk fibroin nanoparticles were also prepared and examined to identify the contribution of magnetic nanoparticles to the encapsulation efficiency and in vitro drug release of the model drugs.

#### **2. Materials and Methods**

#### *2.1. Materials*

Polyvinyl alcohol high molecular weight solid, (PVA 98-99 hydrolized) and curcumin (95% total curcuminoid content) from Tumeric rhizome were purchased from A Johnson Company (New Brunswick, NJ, USA). Ethanol, Lithium Bromide 99%, and Bovine Serum Albumin (BSA) (lyophilized powder, 66.000 kDa), were purchased from Sigma-Aldrich (St. Louis, MO, USA). Propranolol hydrochloride and Pramipexole were purchased from Fagron Hellas (Trikala, Greece). FeCl3•6H2O, FeCl2•4H2O, and NH4OH of analytical grade were purchased from Sigma-Aldrich.

#### *2.2. Methods*

#### 2.2.1. Preparation of Silk Fibroin Solution

Silk fibroin aqueous solution was prepared according to described protocols [17]. Briefly, 5 g cocoons of Bombyx mori were boiled in 2 L of an aqueous solution of sodium carbonate (0.02 M) for 30 min and the degummed silk fibroin was thoroughly rinsed with deionized water, air dried

overnight, and then dissolved in a 9.3 M LiBr solution at 60 ◦C. The solution was dialyzed against deionized water using Slide-a-Lyzer dialysis cassettes (MWCO 3.500, Pierce, Thermo-Fischer Scientific, Waltham, MA, USA) for 2 days for salt removal. In order to clean the solution from impurities, silk aggregates, and debris of cocoons, the silk fibroin solution was centrifuged twice at 9000 rpm for 30 min. The prepared silk fibroin solution was stored at 4 ◦C for further use.

#### 2.2.2. Preparation of Silk Fibroin (SF) Nanoparticles

Silk fibroin particles were prepared according to the following method [18]: 5 mL of silk fibroin solution was mixed with 2 mL ethanol with a sample pipette and then vortexed for 10 s. Subsequently, 50 mL of 5% PVA solution were added to the fibroin/ethanol mixture and vortexed for 10 s. The fibroin/ethanol/PVA solution mixture was refrigerated for 24 h and centrifuged. In order to remove PVA and ethanol from the nanoparticles, the mixture was washed with deionized water three times and centrifuged after each wash [18].

#### 2.2.3. Preparation of Silk Fibroin-Bovine Serum Albumin (SF-BSA) Nanoparticles

For the preparation of the Silk Fibroin-Bovine Serum Albumin nanoparticles (SF-BSA-NPs), 2.5 mL of the silk fibroin solution were mixed with 2.5 mL of 2% BSA solution and then the mixture was stirred for 15–20 min. The SF-BSA-NPs were prepared following the method used to prepare the silk fibroin nanoparticles.

#### 2.2.4. Synthesis of Fe3O4 Nanoparticles

For the preparation of the magnetic silk fibroin nanoparticles (SFm-NPs), the magnetic material used was magnetite nanoparticles (Fe3O4) prepared in the laboratory. The Fe3O4 nanoparticles were prepared according to the modified Massart method [19] via the co-precipitation of FeCl3•6H2O and FeCl2•4H2O [18]; briefly, 11.2 mmol FeCl3•6H2O and 5.6 mmol FeCl2•4H2O were dissolved in 150 mL of deionized water, heated to 60 ◦C under agitation in an inert atmosphere, and then aqueous ammonia solution was added dropwise until the solution's pH became 10. The precipitate that was formed was magnetically collected, washed with deionized water and ethanol, and freeze-dried.

#### 2.2.5. Preparation of Silk Fibroin-Fe3O4 (SFm) Nanoparticles

For the preparation of the magnetic silk fibroin nanoparticles (SFm-NPs), 0.1 g Fe3O4 was added to 5 mL of silk fibroin solution followed by the preparation of nanoparticles as above described.

#### 2.2.6. Preparation of Drug-Encapsulated Silk Fibroin Nanoparticles

Curcumin (MW = 368.39 g/mol), propranolol hydrochloride (MW = 259.34 g/mol), and pramipexole (MW = 211.32 g/mol) were the model drugs tested for the encapsulation in the silk fibroin nanoparticles; the chemical formulas of the drugs are presented in Table 1.

Amounts of the pramipexole and propranolol were dissolved in deionized water and mixed with 5 mL of the silk fibroin solution; curcumin was dissolved in ethanol and then mixed with 5 mL of the silk fibroin solution. Different amounts of the drugs were estimated for the preparation of solutions with concentrations of 10, 35, 50, 100, 200, and 500 ppm for each drug. Finally, the drug-encapsulated silk fibroin nanoparticles were prepared as previously described.

The encapsulation efficiency in the silk fibroin nanoparticles was determined by measuring the UV-vis absorbance (HITACHI U-2000) of the supernatant after the centrifugation that follows the formation of silk fibroin nanoparticles at 430, 265, and 235 nm for curcumin, pramipexole, and propranolol, respectively. Standard calibration curves of model drugs were used for drug quantification (r2 ≥ 0.999). All experiments were performed in triplicate and the mean value is presented. The difference between the total amount of drug used in the experiment and the amount

that remained in the supernatants was expressed as encapsulation efficiency (EE) in silk fibroin nanoparticles [20] and determined by the following equation:

$$\text{Encapsulation Efficiency} = \frac{\text{amount of drug that remained in the particles}}{\text{total amount of drug used}} \times 100\tag{1}$$


**Table 1.** Model drugs encapsulated in silk fibroin nanoparticles.

MW = molecular weight.

#### 2.2.7. Silk Fibroin Nanoparticles Characterization

For the identification of the crystalline phase of silk fibroin nanoparticles, X-ray powder diffraction (XRD) patterns were recorded on a Philips PW 1820 diffractometer (Amsterdam, The Netherlands) with Cu Kα radiation from 20◦ to 60◦. The morphology of silk fibroin nanoparticles as well as that of drug-encapsulated silk fibroin nanoparticles was imaged using a JEOL JMS-840A scanning electron microscope (JEOL, Tokyo, Japan). The control samples of silk fibroin nanoparticles suspensions in water, the samples of SF-BSA-NPs and SFm-NPs, and their respective drug-encapsulated nanoparticles suspensions in water were lyophilized and the powder obtained was subjected to Fourier Transform Infrared measurement using a Perkin-Elmer FTIR spectrophotometer (model Spectrum 1000, Rodgau, Germany). Nanoparticle size, size distribution, and surface charges of the nanoparticles, and those of their respective drug-encapsulated nanoparticles, were determined by Dynamic Light Scattering (DLS) with a NanoBrook ZetaPALS Brookhaven Instruments (Holtsville, NY, USA) equipped with a diode laser (wavelength, λ = 532 nm). Hysteresis loops for the SF-magnetite nanoparticles were recorded with an Oxford 1.2 H/CF/HT vibrating sample magnetometer (VSM).

#### 2.2.8. In Vitro Drug Release

The amount of drug released in vitro from the drug-encapsulated silk fibroin nanoparticles, at specific time intervals, was calculated as follows: 4 mL of the drug-encapsulated nanoparticles were added to 20 mL of phosphate buffer saline (PBS) at pH 7.4 and shaken at a rate of 120 rpm at 37 ◦C. At specified time intervals, 2 mL were sampled and centrifuged at 9000 rpm for 10 min. The concentration of the drug in the supernatant was measured by UV-vis spectrometry at 430, 265, and 235 nm for curcumin, pramipexole, and propranolol, respectively. The percentage release was determined as the ratio of the measured amount of the drug released at different time intervals to the initial amount of drug encapsulated in the nanoparticles [12]. All experiments were performed in triplicate and the mean value is presented.

#### **3. Results and Discussion**

Weighting the residual solid of a certain known volume of silk solution after drying at 60 ◦C, the concentration of the silk fibroin aqueous solution prepared was found to be approximately 4.6% (*w*/*v*).

#### *3.1. XRD Characterization, Morphology, and FTIR Characterization of the Silk Fibroin Nanoparticles*

XRD peaks of silk fibroin are associated with its crystalline structure; Figure 1 presents the XRD results of the silk fibroin nanoparticles prepared in the current study. The 2θ diffraction peak presented in the XRD pattern of the silk fibroin nanoparticles at 2θ = 19.2◦ could be attributed to the silk II crystal structure of fibroin, indicating an increased crystallization degree of silk fibroin in the nanoparticles [21,22].

**Figure 1.** XRD patterns of silk fibroin nanoparticles (SFNPs).

The morphology of silk fibroin nanoparticles was illustrated by SEM images and is presented in Figure 2a. The nanoparticles observed presented a spherical and/or cylindrical shape with an average mean diameter of about 290 nm. Silk fibroin nanoparticles previously prepared with the same method by Shi et al. [18] presented an average diameter of about 686.5 nm as estimated by SEM image. This difference in size could be attributed to the lower concentration (4%) of silk fibroin solution that was used in this study to prepare the silk fibroin nanoparticles compared to the study by Shi et al. (6%). According to Zhao et al., the increase of the starting silk fibroin solution concentration has an effect on the particle size increase [23].

**Figure 2.** *Cont*.

**Figure 2.** SEM images of SF nanoparticles (**a**); SF-propanolol NPs (**b**); SF-pramipexole NPs (**c**); and SF-curcumin NPs (**d**).

In Figure 3, the FTIR spectrum of the prepared silk fibroin nanoparticles is presented. The characteristic peak at 1632 cm−<sup>1</sup> attributed to an amide I β-sheet and the peak at 1260 cm−<sup>1</sup> attributed to amide III are due to random coil. So, when the nanoparticles of silk fibroin were formed both the silk I crystal structure and the silk II crystal structure of silk fibroin were observed [24,25]. The assignments of the FTIR bands of silk fibroin nanoparticles are presented in detail in Table 2.

**Figure 3.** FTIR spectra of silk fibroin nanoparticles and drug-encapsulated silk fibroin nanoparticles.


**Table 2.** Assignments of the FTIR bands of silk fibroin nanoparticles and drug-encapsulated silk fibroin nanoparticles.

(A)*n*, polyalanine; (AG)*n*, polyalanine glycine [26].

#### *3.2. Encapsulation Efficiency*

The encapsulation efficiency (EE) of the silk fibroin nanoparticles for the three model drugs was studied at a range of concentrations from 10 to 500 ppm. For the curcumin-encapsulated silk fibroin nanoparticles the highest EE was found to be 98 ± 2% and was achieved at 500 ppm, for propranolol the highest EE was 69 ± 3% at 50 ppm and for pramipexole the highest EE was 68 ± 2% at 35 ppm (data not shown).

#### *3.3. Size, ζ-Potential, Morphology, and FTIR Characterization of Silk Fibroin Nanoparticles and Drug-Encapsulated Silk Fibroin Nanoparticles*

The size of silk fibroin nanoparticles measured by DLS is presented in Table 3. As seen in the Table, the average diameter of the silk fibroin nanoparticles was found to be 301 ± 11 nm, which is consistent with the size estimated by SEM; the average size was found to increase after the encapsulation of the model drugs (curcumin, pramipexole, and propranolol) in the silk fibroin nanoparticles. The polydispersity index, ranging from 0.10 to 0.25, revealed a uniform particle distribution.


**Table 3.** Diameter of silk fibroin nanoparticles measured by dynamic light scattering.

The surface charge of the silk fibroin solution, of the drug solutions, of the silk fibroin nanoparticles, and of the silk fibroin nanoparticles encapsulated with the model drugs was determined by ζ-potential and the results are presented in Table 4. For all samples, the ζ-potential was negative; it was observed that the ζ-potential of propranolol-encapsulated silk fibroin nanoparticles (SF-propranolol NPs) was higher than that of curcumin (SF-curcumin NPs) and pramipexole-encapsulated silk fibroin nanoparticles (SF-pramipexole NPs), leading to the conclusion that propranolol was encapsulated in the center of the silk fibroin nanoparticles while curcumin and pramipexol were adsorbed on the surface of the nanoparticles.


**Table 4.** ζ-potential of model drug solutions, silk fibroin nanoparticles, and drug-encapsulated silk fibroin nanoparticles.

The morphology of drug-encapsulated silk nanoparticles, and that of the silk nanoparticles for the sake of comparison, was illustrated by SEM images (Figure 2). The nanoparticles observed presented a spherical and/or cylindrical shape with an average mean diameter of 290 nm. As expected, the size of silk fibroin nanoparticles increased after drug encapsulation. The mean diameter of the drug-encapsulated silk fibroin nanoparticles, as estimated from the SEM images, was 490, 392, and 294 nm for propranolol-, pramipexole-, and curcumin-encapsulated silk fibroin nanoparticles, respectively. It is worth noting that these results for nanoparticle sizes are relatively smaller than those measured with DLS; this can be attributed to the dry state of the nanoparticles when measured with SEM.

In Figure 3, the FTIR spectra of silk fibroin nanoparticles and of drug-encapsulated silk fibroin nanoparticles are presented. In the spectra of the curcumin-encapsulated silk fibroin nanoparticles, it was observed that certain characteristic peaks of the original curcumin were either not evident or exhibited a slight shift or alteration, which demonstrated that certain structural changes and chemical reactions may have occurred between the curcumin and the silk fibroin [12,27]. Curcumin showed its characteristic group absorption peaks in sharp absorption bands at 1605 cm−1, 1502 cm−<sup>1</sup> (–C=O and –C–C vibrations), 1435 cm−<sup>1</sup> (an olefinic –C–H bending vibration), 1285 cm−<sup>1</sup> (an aromatic –C–O stretching vibration), and 833 cm−<sup>1</sup> (a C–H bond of alkene group) [12,27]. In addition, propranolol exhibited characteristic peaks in sharp absorption bands at 1680–1620 cm−<sup>1</sup> (C=C stretching) and 1260–1000 cm−<sup>1</sup> (C–O stretching) [28]. On the contrary, the FTIR spectrum of pramipexole-encapsulated fibroin nanoparticles presented no difference compared to pure fibroin nanoparticles; the characteristic peaks of pramipexol were not evident, leading to the conclusion that after encapsulation with pramipexole the chemical bonding remained unchanged [29]. The assignments of the FTIR bands of silk fibroin nanoparticles as well as those of the drug-encapsulated silk fibroin nanoparticles are presented in detail in Table 4.

#### *3.4. In Vitro Drug Release*

The drug release profile, presented in Figure 4, indicated that the encapsulated curcumin and pramipexole silk fibroin nanoparticles had a short and low-level release percentage, about 1.2% and 0.25%, respectively (presented also in the insets of the Figure 4) while the propranolol-encapsulated silk fibroin nanoparticles exhibited a release of about 65% within 6 days. This could be due to the lack of interactions of propranolol with the silk fibroin as well as to the hydrophilic properties of propranolol. Curcumin, as a hydrophobic drug, was possibly attached to silk fibroin via hydrophobic interactions and π-π stacking, and presented a reduced release rate compared with that of the hydrophilic drug propranolol [4,12,30]. Pramipexole was expected to be attached to silk fibroin through strong electrostatic interactions attributed to the –NH2 groups, which could have resulted in the decreased release rate.

**Figure 4.** Drug release rate of model drugs from nanoparticles of silk fibroin.

#### *3.5. Modification of SF with Bovine Serum Albumin and Magnetic Nanoparticles*

In order to enhance the properties of silk fibroin nanoparticles, SFm-NPs and silk fibroin–bovine serum albumin nanoparticles (SF-BSA-NPs) were prepared via the introduction of magnetic nanoparticles (Fe3O4) and BSA during the silk fibroin nanoparticle formulation, and the drug encapsulation and release efficiency of these silk fibroin nanoparticles were examined in detail. Since propranolol presented the highest encapsulation and release performance, it was used as the model drug for the performance of the two modified types of silk fibroin nanoparticles.

#### 3.5.1. Characterization of Modified Nanoparticles

DLS measurements for silk fibroin-bovine serum albumin nanoparticles and magnetic silk fibroin nanoparticles are presented in Table 5. Both samples indicated a small polydispersity index of 0.30 and 0.36, respectively, indicating a uniform diameter distribution of nanoparticles.


**Table 5.** Diameter of magnetic and fibroin-albumin nanoparticles measured by Dynamic Light Scattering.

SFm-NPs = magnetic silk fibroin nanoparticles; SF-BSA-NPs = silk fibroin-bovine serum albumin nanoparticles.

The encapsulation efficiency of magnetic silk fibroin and silk fibroin–bovine serum albumin nanoparticles is presented in Table 6. From the Table, it can be seen that the modification of silk fibroin with magnetic nanoparticles and bovine serum albumin increased the drug encapsulation compared to the pure silk fibroin nanoparticles. Propranolol is a hydrophilic drug and the presence of bovine serum albumin makes the silk fibroin–bovine serum albumin nanoparticles more hydrophilic; for this reason the encapsulation efficiency was increased. Also, the increased encapsulation efficiency of magnetic silk fibroin loaded with propranolol may be attributed to the adsorption of propranolol on Fe3O4.


**Table 6.** Encapsulation Efficiency (EE) of magnetic silk fibroin and silk fibroin-bovine serum albumin nanoparticles encapsulated with propranolol.

The surface morphology of magnetic silk fibroin particles and the silk fibroin-bovine serum albumin nanoparticles as well as their encapsulated-with-propranolol counterparts, illustrated by SEM images, is presented in Figure 5. The magnetic silk fibroin (SFm) and silk fibroin-bovine serum albumin nanoparticles (SF-BSA) exhibited a spherical shape with a diameter of about 288 nm, while the propranolol-encapsulated silk fibroin-bovine serum albumin nanoparticles (SF-BSA + propranolol) and propranolol-encapsulated magnetic silk fibroin nanoparticles (SFm + propranolol) exhibited a particle size of about 384 nm and 490 nm, respectively (Figure 5).

**Figure 5.** SEM images of SFm nanoparticles (**a**); SFm + propranolol NPs (**b**); SF-BSA-NPs (**c**); and SF-BSA + propranolol NPs (**d**).

The 2θ diffraction peak presented in the XRD pattern of the magnetic silk fibroin nanoparticles (Figure 6a), that appeared at 2θ = 19.2◦, could be attributed to silk I, a random coil of silk fibroin, while for the encapsulated-with-propranolol magnetic silk fibroin nanoparticles, the characteristic peak at 2θ = 20.7◦ could be attributed to silk II, a β-sheet of fibroin. In addition, diffraction peaks corresponding to the (3 1 1), (4 0 0), and (4 4 0) planes of the cubic crystal structure (fcc) of Fe3O4, show the formation of magnetic silk fibroin nanoparticles [19]. The aforementioned XRD results are in agreement with the FTIR results.

Prior to the measurements, the instrument was calibrated against a NIST-certified Ni standard. Each sample had approximately the same dimensions and an additional calibration procedure was performed in order to minimize shape effects. All measurements were performed in a controlled room temperature of 22 ◦C and consisted of a gradual increase of the applied external magnetic field up to about +1.9 T (μ0H) then down to −1.9 T and again up to +1.9 T while measuring the magnetic moment of the sample. The magnetic properties are attributed to the Fe3O4 particles and the overall appearance of the loop is typical for nanoparticles of that composition and size. All measurements have exactly the same appearance, i.e., initial slope, curvature, and saturation plateau, indicating that there are no significant differences in the Fe3O4 particles between the samples and that the procedure for the preparation of the magnetic silk fibroin nanoparticles (SFm-NPs) does not affect significantly the magnetic properties of the Fe3O4 particles. Remanence, Mr, the Mass Magnetization which corresponds to a zero field after the first application of the maximum external field equals to 5–7% of the saturation magnetization and coercivity, μ0Hc, the reverse external field which zeroes the magnetic moment of the sample is about 0.01 T. Both the remanence and the coercivity are very low in all cases as expected for a soft ferromagnetic material. Saturation magnetization, the maximum Mass Magnetization which could be observed, was obtained for the sample via extrapolation from data of Figure 6b,c and was found to be σ<sup>s</sup> = 27.0 (8b) and 27.9 (8c) emu/g. These results correspond to about 50% wt. Fe3O4 content, since the saturation magnetization depends on the Fe3O4 content which is diluted in the silk fibroin, confirmed that Fe3O4 particles are successfully incorporated into the silk fibroin and that their magnetic properties are preserved after drug encapsulation.

**Figure 6.** (**a**) XRD patterns of the magnetic silk fibroin nanoparticles and the propranolol-encapsulated magnetic silk fibroin nanoparticles; (**b**) hysteresis loop of the magnetic silk fibroin nanoparticles; and (**c**) hysteresis loop of the propranolol-encapsulated magnetic silk fibroin nanoparticles.

FTIR results of silk fibroin-bovine serum albumin nanoparticles and propranolol-encapsulated silk fibroin-bovine serum albumin nanoparticles are presented in Figure 7a,b, respectively, and the characteristic peaks are presented in Table 7. The bands at 1645 cm−<sup>1</sup> and 1491 cm−<sup>1</sup> are due to the BSA modification and can be attributed to amide I (C=O stretching) and amide II (C–N stretching and N–H bending) vibrations of BSA, respectively (Table 7). The bands of BSA at 1392 cm−<sup>1</sup> (–CH2 bending) and ~1260 cm−<sup>1</sup> (amide III, C–N stretching, and N–H bending) may be overlapped by the bands

attributed to silk fibroin [7]. A slight shift in the spectra for the propranolol-loaded nanoparticles could be due to drug binding.

**Figure 7.** FTIR spectra of: (**a**) silk fibroin-bovine serum albumin nanoparticles; (**b**) propranolol-encapsulated silk fibroin-bovine serum albumin nanoparticles; (**c**) magnetic silk nanoparticles; and (**d**) propranolol-encapsulated magnetic silk fibroin nanoparticles.

In Figure 7c,d, the FTIR results of fibroin magnetic nanoparticles (SFm) and propranolol-encapsulated fibroin magnetic nanoparticles (SFm + propranolol) are presented, while the characteristic peaks are also presented in detail in Table 7. The spectra presented characteristic bands at 471 and 588 cm−<sup>1</sup> and 506 and 595 cm−1, respectively, attributed to Fe–O bonds [4], which confirmed the successful preparation of magnetic silk fibroin nanoparticles (Figure 7c). From the spectra, the existence of both β-sheets and a-helices for magnetic fibroin nanoparticles could be observed, while for the propranolol-encapsulated silk fibroin magnetic nanoparticles β-sheets could be also observed.


**Table 7.** Assignment of the main bands present in SF-BSA NPs, SF-BSA + propranolol NPs, and SFm and SFm + propranolol NPs.

(A)*n*, polyalanine; (AG)*n*, polyalanine glycine [28].

#### 3.5.2. In Vitro Drug Release

In Figure 8, the in vitro drug release profiles for the BSA- and magnetic-nanoparticles-modified silk fibroin nanoparticles are presented compared to the initial silk fibroin nanoparticles. Due to their higher encapsulation efficiency, the propranolol-encapsulated SF-BSA nanoparticles showed a higher drug release rate as compared to the propranolol-encapsulated silk fibroin nanoparticles [13]. In contrast, propranolol-encapsulated magnetic silk fibroin nanoparticles showed less release, which may be attributed to the strong electrostatic interactions of magnetite with the model drug. Another reason for the lower release could be the higher content of β-sheets in the propranolol-loaded silk fibroin magnetic nanoparticles causing a slow degradation of the silk fibroin and finally reducing the drug release rate [26].

**Figure 8.** Drug release rate of fibroin nanoparticles, magnetic fibroin nanoparticles, and fibroin–bovine serum albumin nanoparticles encapsulated with propranolol.

#### **4. Conclusions**

Silk fibroin nanoparticles were prepared by an easy and safe-to-manipulate method that is amenable to a wide range of drugs and thus useful for silk-based drug delivery systems. Silk fibroin nanoparticles facilitated the entrapment of model drugs (curcumin, pramipexole, and propranolol) with different molecular weights and hydrophobicities and made drug release controllable. Silk fibroin particles loaded with propranolol exhibited a higher release percentage than those loaded with curcumin or pramipexole. Moreover, modified silk fibroin nanoparticles with magnetic nanoparticles as well as with bovine serum albumin were examined for the enhancement of the propranolol encapsulation and release from these nanoparticles. Magnetic silk fibroin nanoparticles increased the propranolol encapsulation efficiency. The silk fibroin-bovine serum albumin particles increased the encapsulation efficiency as well as the release rate of propranolol and were found to be a promising drug carrier.

**Author Contributions:** Olga Gianak and Eleni Deliyanni conceived and designed the experiments; Olga Gianak performed the experiments; Eleni Pavlidou performed the SEM measurements and analyzed the relative data; Charalampos Sarafidis performed the magnetic experiments and analyzed the relative data; Vassilis Karageorgiou performed the z-dynamic and dynamic light scattering measurements, analyzed the relative data, and contributed to the manuscript's writing; Eleni Deliyanni and Olga Gianak wrote the paper.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Technical Note* **Method Development of Phosphorus and Boron Determination in Fertilizers by ICP-AES**

#### **Emanouela Viso and George Zachariadis \***

Laboratory of Analytical Chemistry, Department of Chemistry, Aristotle University, 54124 Thessaloniki, Greece; emmyviso@gmail.com

**\*** Correspodence: zacharia@chem.auth.gr; Tel.: +30-231-099-7707

Received: 14 May 2018; Accepted: 3 July 2018; Published: 9 July 2018

**Abstract:** Simultaneous determination of phosphorus and boron in fertilizers was performed by Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES). Three different samples were analyzed, of which two were inorganic and one was of organic composition. Analysis of the samples was performed after heated acidic digestion to completely dissolve them, using two different acid mixtures. A solution of HCl + HNO3 was used to digest the inorganic fertilizers, and a solution of H2SO4 + HNO3 for the organic fertilizer. The spectral emission lines used were 213.617 nm and 214.917 nm for P and 249.772 nm, 249.677 nm and 208.957 nm for B. The detection and quantification limits for P were between 10–20 mg/kg and 40–80 mg/kg respectively, while for B they ranged between 10–30 mg/kg and 40–100 mg/kg respectively. The repeatability of the technique was found to be within the range 0.9–17.0% for P and 1.7–23.4% for B, expressed as relative standard deviation (RSD). The concentrations found by the proposed method are in good agreement with those reported on their package labels.

**Keywords:** phosphorus; boron; inductively coupled plasma; atomic emission spectrometry; fertilizers; acid dissolution; wet digestion

#### **1. Introduction**

Fertilizers contain many elements which act as nutrients for plants, but also elements which may be toxic, and which are therefore responsible for contamination of soils. This results in the accumulation of chemicals in food [1,2]. As the development of agricultural productivity is directly related to the use of fertilizers, it is necessary to analyze them with sensitive, multi-elemental technical analyzers to monitor their quality [3]. Common fertilizers are either of inorganic or organic composition, natural or synthetic. Chemical analysis and quality control of fertilizers result in improved agricultural production [4], and methods have been developed. In this context, the determination of minerals in fertilizers becomes necessary, and for this purpose, several techniques have been reported in the literature [1,2,5]. Some techniques for the determination of heavy metals are based on Graphite Furnace atomic absorption spectroscopy (GF-AAS), Continuous Source Flame Absorption Spectroscopy (HR-CS-FAAS) [1,6–9], or Cold Vapor Atomic Absorption Spectroscopy (CV-AAS) for Hg [5]. In addition, methods have been developed to determine nutrients in fertilizers by LIBS [10,11], or by total X-ray fluorescence (TXRF) techniques [12]. Inductively coupled plasma atomic emission spectroscopy (ICP-AES) allows rapid analysis and simultaneous determination of primary and secondary nutrients as trace elements in fertilizers [2,13,14]. ICP-AES is a well-known analytical technique suitable for simultaneous determination of almost 70 chemical elements, as it provides high sensitivity and satisfactory detectability, has a wide range of applications, and does not suffer from important spectral interferences [15–17], except in case of solutions with high TDS. The great advantage of simultaneous determination of various elements in fertilizers, like P and K, has already been highlighted by other researchers [18]; this becomes more attractive considering the capability of measuring also semi-metals, like boron [19], or non-metals, like phosphorus and sulfur. However, in many cases for total element fraction, it requires complete digestion of solid samples before quantitative analysis, which means a sample treatment step, which can potentially affect accuracy and overall analytical performance of the method. In the present study, a method for the simultaneous determination of total phosphorus and boron by ICP-AES in fertilizers was developed, after digestion of the samples with two different acidic solutions. The examined acid mixtures were selected according to the nature of the fertilizer samples. For determination of the total fraction of an element, inorganic fertilizers usually do not require strong oxidative media, while organic fertilizers may need strong oxidative acids and heating or incineration followed by acidic dissolution for complete decomposition or other pretreatment techniques [19,20].

#### **2. Experimental**

#### *2.1. Instrumentation*

An inductively coupled plasma atomic emission spectrophotometer, model OPTIMA 3100 XL (Perkin Elmer, San Francisco, CA, USA), was employed; the operating conditions are listed in Table 1. The instrument was equipped with a liquid sample introduction system, including a peristaltic pump to aspirate solutions into the nebulizer at variable flow rates. Tygon-type pump tubes were used for sample delivery, and the nebulization system consisted of a cross-flow nebulizer and a Scott double-pass spraying chamber. A quartz torch is mounted horizontally on the same axis to the spectrophotometer window (axial viewing of emission). The torch inner injector tube was composed of alumina, which is sufficiently resistant to highly acidic solutions. The radio frequency generator (RF) that maintains the plasma is 40.68 MHz; this was adjusted to an incident power of 1350 watts for this method. The spectrophotometer was equipped with a polychromator in which an echelle-type diffraction grating was installed, and the detector was a segmented charged-coupled device. The ICP-AES instrument was supplied by analytical grade argon as a plasma gas. Three phosphorous and three boron atomic spectral lines were examined, as given in Table 2. The phosphorous line at 178.221 nm was finally rejected for reasons described below.


**Table 1.** Operating conditions and settings of the ICP-AES.

**Table 2.** Spectral lines employed for ICP-AES measurements.


#### *2.2. Reagents and Solutions*

For the preparation of the working standard solutions, KH2PO4 (99.5%) and H3BO3 (99.5%) were obtained from Merck (Darmstadt, Germany). For acid digestion of fertilizer samples, concentrated HNO3 (65%), HCl (37%), and H2SO4 (95–97%) of analytical grade were obtained from Merck (Darmstadt, Germany). All dilutions and solutions were carried out using ultrapure water of Milli-Q quality (18.2 MΩ, Millipore, Bedford, MA, USA).

#### *2.3. Fertilizer Samples*

Inorganic and organic commercially available fertilizers from local companies in Greece were analyzed in the present study to develop a fast analytical method. The inorganic samples were crystalline, water-soluble, general-purpose fertilizers, while the organic one was a composite fertilizer for vegetables. Their indicative nutritional content, as given on their packaging, is provided in Table 3. Phosphorus and boron contents of the samples were calculated, and are presented in Table 4.



\* In 3rd sample (organic) organic matter is 15% *w*/*w*.



#### *2.4. Preparation of Working Standard Solutions*

A mixed standard aqueous solution containing 50 mg/L each of phosphorus and boron was prepared from standard solutions of 100 mg/L P (KH2PO4, 439.4 mg/L) and 100 mg/L B (H3BO3, 571.97 mg/L). Finally, five working standard solutions containing 0.00, 1.00, 2.50, 10.0, 25.0 mg/L of phosphorus and boron, respectively, were prepared by proper dilutions of the above stock standard solution (50 mg/L).

#### *2.5. Acid Digestion of Samples*

For the selection of the appropriate acid mixtures, dissolution tests were performed by acid digestion of all samples. The most efficient mixtures were selected for the liquid digestion process based on the solubilization effect. Thus, for the 1st and the 2nd sample (inorganic), the acid digestion procedure was performed with the addition of 5 mL of 37% HCl + 1.5 mL of 65% HNO3, and then heating for 2 min on a hot plate in high fume hood (ca 80 fpm face velocity) for nitrogen oxides fumes. For the 3rd sample (organic), the acid digestion procedure was performed by adding 5 mL H2SO4 95–97% + 1 mL HNO3 65%, followed by heating for 3 min on a hot plate in a fume hood (ca 80 fpm). For the acid digestion procedure, accurately weighed amounts of the samples were placed in 100 mL open conical flasks and heated to 130 ◦C. When digestion was complete, the mixture was allowed

to cool to room temperature. The mixtures were then transferred to 100 mL volumetric flasks and diluted with Milli-Q water. The obtained diluted solution was further diluted at a ratio of 1:10 and 1:100 successively. The acid digestion process was repeated three times for each sample. Preparation of acidic blank mixtures was performed three times.

Standard addition samples were also prepared by the addition of standard solutions of P and B to 0.500 g of the first sample. This process was carried out to study the recovery of the analytical method. To this sample, 228 g/kg P (435 mg/L) of KH2PO4 and 174 g/kg B (0.75 mg/L) of H3BO3 were added. The sample was then subjected to acidic digestion with 5 mL HCl + 1.5 mL HNO3 and warmed up. The sample after acid digestion was diluted to 100 mL, followed by successive dilutions of 1:10 and 1:100. The whole procedure was performed three times.

#### **3. Results and Discussion**

#### *3.1. Regression Analysis*

The settings for the spectral lines and the operating conditions of ICP-AES were defined in the computer software. For quantitative analyses of the samples, the required calibration curves were constructed using the series of mixed working standard solutions.

The results of the regression analysis for each element at each tested wavelength are given in Table 5. The emission spectra of P and B obtained from the mixed working standard solutions are given in Figures 1 and 2 respectively. As shown in Table 5, at the examined spectral lines, both P and B showed good correlation coefficients. The correlation coefficient R is greater than 0.999 for all elements, except at the third spectral line of P (178.221 nm), where the correlation was much lower because of baseline instability. Apparently, no sensitivity is observed also, and for this reason, this spectral line P (178.221 nm) was excluded from further measurements.

**Figure 1.** *Cont*.

**Figure 1.** Phosphorus emission spectra at 213.617 nm and 214.914 nm, respectively. Superimposed traces refer to working standard solutions of increasing concentration, as described in Section 2.4.


**Table 5.** Results of regression analysis using the mixed standard aqueous solutions of boron and phosphorus.

**Figure 2.** *Cont*.

**Figure 2.** Boron emission spectra at 249.772 nm and 249.677 nm respectively. Superimposed traces refer to standard solutions of increasing concentration, as described in Section 2.4.

#### *3.2. Repeatability, Detectability and Recovery of Method*

Acceptable repeatability of a method is expressed by low values of the relative standard deviation, RSD%. Repeatability testing was done by measuring the working standard three times on the same day. The solutions used were aqueous mixed solutions of P and B on all spectral lines. The concentrations used to control the repeatability were 1.00 mg/L, 2.50 mg/L, 10.0 mg/L, and 25.0 mg/L. The results are listed in Table 6. Based on these results, good repeatability is observed at concentrations of 10.0 mg/L and 25.0 mg/L, but less good at 1.00 mg/L concentration. The repeatability for P in the examined atomic lines ranged between 0.9–17%, and for B between 1.7–23.4%, respectively.

Limit of detection (LOD) and limit of quantification (LOQ) are defined as the minimum concentration or quantity of the analyte that can be detected or quantified respectively with reasonable certainty, according to IUPAC recommendations [21]. The detection and quantification limits of the developed method were calculated using the 3s criterion (i.e., at a 99.6% confidence level). The calculated detection limits for the examined spectral lines of phosphorus were found to be between 10–20 mg/kg, and those for B between 10–30 mg/kg. This variation is due to variability of the baseline signal values in different spectral lines and with the different acid mixtures. The corresponding limits of quantification were found to be between 40–80 mg/kg for P, and between 40–100 mg/ kg for B. The results are presented in Tables 7 and 8. In these tables, the results are given as typical instrumental LODs and LOQs (expressed in mg/L), and also as method equivalent LODs and LOQs (expressed in mg/kg), considering a typical sample mass of 0.500 g for the analysis and initial dilution to 100 mL, as described in paragraph 2.5.

Because no reference material was available, the recovery for P and B was calculated after adding an amount of the standard solution to the first sample of fertilizer. Table 9 gives the calculated recoveries for P and B in each spectral line, which are satisfactory regarding the spectral lines 213.616 nm for P and 249.772 nm for B, correspondingly. The added amount of the standard was 228 g/kg (or 435 mg/L) for P (as KH2PO4), and 176 g/kg (or 0.75 mg/L) for B (as H3BO3).


**Table 6.** Method repeatability.

**Table 7.** Detection and quantification limits using a constitutive blank of 3 mL HCl + 1 mL HNO3. LODs and LOQs were calculated as described in Section 3.2.


**Table 8.** Detection and quantification limits using a constitutive blank of 5 mL H2SO4 + 1 mL HNO3. LODs and LOQs were calculated as described in Section 3.2.



**Table 9.** Recovery results.

#### **4. Application of the Method to Fertilizers Samples**

The results of the analysis of elements P and B in three samples of inorganic and organic fertilizer samples are presented in Table 10, while the corresponding label compositions are given in Tables 3 and 4. Acid digestion was performed three times for each sample under the same conditions and with the same amounts of samples and solvents. Then, each of the three sets of samples was diluted three times, i.e., initial dilution to 100 mL, which was used for boron determination, and subsequent dilutions 1:10 and 1:100, which were used for phosphorus determination.

Regarding the 1st inorganic fertilizer, the indicated content for P in the form P2O5 is 20% *w*/*w*. Thus, the concentration expressed in elemental P is 87 g/kg. The concentration using either of the two spectral lines after the analysis is in good accordance with that on the label. On the other hand,

the concentration of B as given on the label is 0.015% *w/w*, which means 0.15 g/kg. The concentration as found after analysis i of the second and third spectral lines agrees with the indicated value.

Regarding the 2nd inorganic fertilizer, the indicated content for P given in the form of P2O5 is 10% *w/w*, so the concentration of elemental P is 43 g/kg. This concentration is also in agreement with the concentration resulting from sample analysis using the first spectral line of P. The B content given on the label is 100 ppm, which means 0.10 g/kg. The concentrations found from the first spectral line of boron are consistent with the reported content.

Finally, the reported content of the 3rd organic fertilizer sample is 30% *w/w* P2O5; therefore, the elemental P concentration is expected to be 131 g/kg, which is in accordance with the concentrations resulting from ICP-AES analysis. For B, its content is not indicated on the package. The concentration from the ICP-AES analysis was found to be at 0.12 ± 0.03 g/kg using the first emission line.


**Table 10.** Concentration of P and B in fertilizer samples found by ICP-AES analysis after digestion.

#### **5. Conclusions**

In this study, the ICP-AES technique was applied to the development of a multi-elemental fertilizer analytical method. The sample solution after heated acid wet digestion was introduced into the inductively coupled plasma atomizer. From the conducted research, the following conclusions regarding the performance of the developed method were drawn. The spectral line of P (178.221 nm) showed low correlation coefficient and poor sensitivity. The spectral lines P (213.16 nm) and P (214.914 nm) showed very good correlation, R > 0.999, with spectral line 214.914 nm being preferable in terms of sensitivity. The three spectral lines of B showed similar correlation coefficients, with a spectral line at 249.677 nm being preferable to the others in terms of sensitivity. The repeatability at high concentration levels of the working curve is better compared to the lower concentrations, and the calculated recoveries were satisfactory. Wet acid digestion of samples was completed in short pretreatment time using mixtures of HCl, HNO3, and H2SO4 acids. The developed method was applied to three different fertilizers of inorganic and organic natures. The P and B results are in accordance with those reported on the product packages. According to the results, it has been concluded that the developed method is a reliable, simple, and fast method for the simultaneous determination of P and B in fertilizers. The method can be applied to rapid examination during fertilizers production, as well as in the analysis of commercial products.

**Author Contributions:** Conceptualization and Investigation G.Z.; Methodology E.V.; Writing-Original Draft Preparation, E.V.; Writing-Review & Editing, G.Z.; Supervision, G.Z.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

1. Borges, A.R.; Becker, E.M.; Lequeux, C.; Vale, M.G.R.; Ferreira, S.L.C.; Welz, B. Method development for the determination of cadmium in fertilizers samples using high-resolution continuum source graphite furnace atomic absorption spectrometry and slurry sampling. *Spectrochim. Acta B* **2011**, *66*, 529–535. [CrossRef]


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **NSAIDs Determination in Human Serum by GC-MS**

**Adamantios Krokos 1,2, Elisavet Tsakelidou 2, Eleni Michopoulou 1, Nikolaos Raikos 2, Georgios Theodoridis <sup>1</sup> and Helen Gika 2,\***


Received: 30 April 2018; Accepted: 3 July 2018; Published: 16 July 2018

**Abstract:** Non-steroidal anti-inflammatory drugs (NSAIDs) are being widely consumed without medical prescription and are often the cause of intoxication, usually in young children. For this, there is a special need in their determination in routine toxicology analysis. As screening methods mainly focus on drugs of abuse (DOA) that are alkaline compounds in their majority, they are not optimized for acidic drugs, such as NSAIDs. Thus, more specific methods are needed for the detection and quantification of this class of drugs. In this study, the efficient extraction of NSAIDs from blood serum and their accurate determination is studied. Optimum pH extraction conditions were studied and thereafter different derivatization procedures for their detection. From the derivatization reagents used, *N*,*O*-Bis(trimethylsilyl)trifluoroacetamide (BSTFA) with 1% Trimethylchlorosilane (TMCS) was found to be the optimum choice for the majority of the examined NSAIDs; pH of 3.7 was selected as the most efficient for the extraction step. Herein the formation of the lactam of diclofenac was also thoroughly investigated. The developed Gas Chromatography-Mass Spectrometry (GC-MS) method had a run time of 15 min with the mass spectrometer operating in Electron Impact (EI) within the mass range of 40 to 500 amu. The method was linear with *R*<sup>2</sup> above 0.991 and limits of quantitation (LOQ) ranging from 6 to 414 ng/mL. The intra-day accuracy and precision were found between 1.03%–9.79% and 88%–110%, respectively, and the inter-day accuracy and precision were between 1.87%–10.79% and 91%–113%. The optimum protocol was successfully applied to real clinical samples, where intoxication of NSAIDs was suspected.

**Keywords:** NSAIDs; derivatization; GC-MS; serum

#### **1. Introduction**

Non-steroidal anti-inflammatory drugs (NSAIDs) are acidic compounds with anti-inflammatory properties at high concentrations and several other properties at low concentrations (i.e., salicylic acid is used as anticoagulant drug) [1]. These drugs exhibit toxicity in the upper concentration levels or after a long time of intake [2,3]. Some of the toxic side effects are related with gastrointestinal disorders, intestinal ulceration, aplastic anemia, myocardial infarction, cerebrovascular events, inhibition of platelet aggregation, and renal dysfunction [4]. Especially, acetaminophen and nimesulide exhibit significant hepatotoxicity [2,5–7]. In addition, there are several cases of suicide attempts or crime commissions which are related with NSAIDs [8].

There is a number of analytical methods which report NSAIDs' determination in biological fluids, most of them related to pharmacokinetic studies or to the support of animal studies for the estimation of exposure and the investigation of potential risk after consumption [4,9].

There are reports of determinations of these analytes by immunoassays, Gas Chromatography-Flame Ionization Detection (GC-FID), High Performance Liquid Chromatography-UltraViolet/Diode

Array Detection (HPLC-UV/DAD)/fluorimetric detector, spectrofluorimetry, thin layer chromatography-UV/fluorimetric detector, GC-MS, GC-MS/MS, capillary electrophoresis, LC-MS, and LC-MS/MS [10–14]. Immunoassays suffer from low selectivity due to cross-reactions, while all other methods except LC-MS/MS include time-consuming sample pretreatment. There is a good number of published multi-analyte LC-MS/MS methods for measuring NSAIDs and their metabolites that provide sensitive and reliable concentration data from biological matrices [5]. In particular, those methods are the most suitable for low concentrated salicylates originating from nutrition. However, GC-MS still represents an integral tool of a clinical and/or toxicological laboratory due to the fact that mass spectra databases are available aiding in the identification of unknowns, At the same time, latest instruments and materials hold the promise for better chromatographic separations and sensitivity, critically important for complex biological samples. Thus, GC-MS analysis provides advantages clearly important in such applications.

For thedetection of NSAID by GC-MS, various sample pretreatment protocols have been applied. Acidic liquid-liquid extraction is often used, however some researchers have also developed simple SPE protocols instead [3,11]. Furthermore, it has been shown that the selection of a suitable derivatization agent is a key factor for the sensitivity of their detection. Apart from this, there are several issues which are related with the stability of these compounds in the GC-MS conditions.

In this paper, a multi-analyte GC-MS method was developed for the simultaneous quantification of eight NSAIDs, based on a specific pro-analysis procedure. The optimum conditions were selected based on experiments focusing on the sample treatment, to obtain a method able to address the needs of a toxicological analysis for these challenging acidic analytes. The method provides a valuable tool for the determination of eight different NSAIDS by GC-MS, with the pros that the GC-MS technique offers, as stressed above, the additional potential of identification of metabolites, degradation products, or others.

#### **2. Materials and Methods**

#### *2.1. Chemicals and Reagents*

All solvents were of analytical or LC-MS grade. Acetonitrile (ACN) LC-MS was purchased from Chem Lab NV, (Zedelgem, Belgium). Nordiazepam-d5 solution 1 mg/mL in methanol reference, material reference standards (>99%) of acetaminophen (APAP), acetyl salicylic acid (ASA), salicylic acid (SA), ibuprofen (IBP), diclofenac (DCF), nimesulide (NI), niflumic acid (NFA), mefenamic acid (MFA), and naproxen (NAP) were purchased from Sigma-Aldrich (Saint Louis, MO, USA). Ethyl acetate (99%) was supplied from Penta (Livingston, NJ, USA). All derivatization reagents; *N*,*O*-Bis(trimethylsilyl)trifluoroacetamide with 1% trimethylchlorosilane (BSTFA & 1% TMCS), *N*-tert-Butyldimethylsilyl-*N*-methyltrifluoroacetamide (MTBSTFA), Pentafluoropropionic anhydride (PFPA), 2,2,3,3,3-Pentafluoro-1-propanol (PFPOH), Trifluoroacetic anhydride (TFAA), Heptafluorobutyric anhydride (HFBA), were for GC derivatization ≥ 99% grade and were purchased from Sigma-Aldrich (Saint Louis, MO, USA). Serum samples were obtained from 4 cases that arrived in AHEPA University General Hospital and were suspected for intoxication. Drug-free human serum was obtained from healthy donors and before its use it was screened by GC/MS for the presence of the NSAIDs.

#### *2.2. Preparation of the Standard Solutions*

Stock solutions of all compounds were prepared in ACN at 10,000 μg/mL. From these, a mix solution containing all nine drugs was prepared and diluted to the following concentrations: 1000 μg/mL, 200 μg/mL, 100 μg/mL, 40 μg/mL, 20 μg/mL, 4 μg/mL, and 2 μg/mL.

#### *2.3. Sample Preparation*

For the selection of the optimum pH conditions, 200 μL of spiked human serum at 200 μg/mL was adjusted at pH of 3.7, 4.7, and 5.7 with the addition of 100 μL of HCOOH/HCOONa buffer solutions prepared at these pH values (by mixing appropriate values of 0.2 M solutions of HCOOH and HCOONa). Then 1 mL of ethyl acetate was added and the mixture was shaken for 10 min. After centrifugation for 5 min at 6300× *g* the organic phase was collected and dried under gentle nitrogen stream at room temperature. The obtained residue was redissolved either in 50 μL of ethyl acetate or in 50 μL of derivatization reagents, as described in Section 2.5. In any case, 1 μL of the extract was injected in the GC-MS system.

#### *2.4. Derivatization Procedure*

For the selection of the optimum derivatization procedure, five different reagents were used: (a) BSTFA with 1% TMCS, (b) MTBSTFA, (c) HFBA, (d) PFPA with PFPOH, and (e)TFAA. The procedure followed for silylation was as follows; addition of 50 μL BSTFA, 1% TMCS, or 50 μL MTBSTFA in the dry residue and after vortexing the mixture was heated for 20 min at 70 ◦C, finally 1 μL was injected in the system. Acetylation was performed with the addition of either 50 μL of HFBA, or of 30 μL PFPA with 20 μL PFPOH, or of 50 μL of TFAA. The mixture was heated for 20 min at 70 ◦C and then after cooling down was evaporated and the residue was dissolved in 50 μL of ethyl acetate. From this extract, 1 μL was injected to GC-MS.

#### *2.5. GC/MS Analysis*

GC-MS analysis was performed on an Agilent Technologies 7890A GC, equipped with a CTC autosampler and combined with a 5975C inert XL EI/CI MSD with Triple-Axis Detector (Agilent Technologies, Santa Clara, CA, USA). GC separations were performed on a 30 m Agilent J&W HP-5ms UI capillary column, with a film thickness of 0.25 μm and an i.d. of 0.25 μm. Back-flash was performed with a 1.5 m deactivated Agilent column with a film thickness of 0.18 mm. The method had a duration of 15 min with the following temperature program: Initial oven temperature at 120 ◦C, hold for 1 min, and then increase to 300 ◦C with a 15 ◦C/min rate. A back-flash step followed, at 300 ◦C for 10 min. Injection of 1 μL of sample was made on a PTV injector operating from 200 ◦C to 320 ◦C. The mass spectrometer (MS) was operated at electron impact ionization mode (EI, 70 eV) and the mass scan range was from 40 to 500 amu.

#### **3. Results**

The majority of the studied drugs are categorized in the acidic class of compounds as they contain carboxyl moieties, except from APAP and NI. Their chemical structures, together with their pKa values, can be seen in Table 1. These are expected to be extracted more efficiently by the organic solvent under acidic conditions, where the ionization of their carboxyl group is suppressed. However, as their physicochemical properties are varying due to their structure and the presence of different substitution groups, the optimum pH for their extraction is needed to be examined.

**Table 1.** Chemical structures and pKa values of the studied non-steroidal anti-inflammatory drugs (NSAIDs).


**Table 1.** *Cont.*

#### *3.1. Extraction pH*

Under three different pH conditions, extraction of the eight NSAIDs from a serum sample spiked at 200 μg/mL was performed, and the extracts were thereafter analyzed without any derivatization. As expected, detection sensitivity of the underivatized drugs is low, however it was acceptable for comparative purposes. Apart from the eight drugs, two more peaks were considered: (1) The peak which corresponds to salicylic acid, a product occurring from ASA in the sample, and (2) the diclofenac-lactam peak, which is obtained by DCF partial conversion during GC analysis. Table 2 provided the peak areas of the studied drugs at three different pHs tested for their extraction. As can be seen, more acidic pH conditions favor the extraction of the majority of the drugs resulting in higher detected peak areas. NI, however, provides higher signal under pH 4.7. For the case of diclofenac, its lactam is detected in higher proportion under all three pH conditions. Apart from that, the methylester of dichlofenac (DCF-ME) is also detected with relatively high signal. Salicylic acid is detected with higher amount under the more acidic pHs compared to acetyl salicylic acid.


**Table 2.** Absolute area of each NSAID at different pH extraction conditions.

#### *3.2. Analysis Without Derivatization*

When the extracted NSAIDs were analyzed directly without prior derivatization, the detection sensitivity was not satisfactory for the majority of the eight drugs. The chromatographic peaks were broad with excessive tailing, which was attributed to intramolecular interactions and the interaction of polar carboxylic groups of drugs with the column's supporting material. Detection of such compounds can be challenging as several transformation products of the drugs were also detected. During GC/MS analysis, NFA, MFA, and NAP were not found stable without derivatization, as their decarboxylated compounds (-CO2) were also detected. This has also been reported in previous studies [15]. As an example, together with NAP detected at 9.86 min, its NAP-CO2 degradation product is also detected at 7.04 min with characteristic ions at *m*/*z* 184 and 141. Another issue is that ASA is converted to SA over time in both aqueous and organic solutions, thus it can also be detected due to that reason. This can mislead its determination, due to the fact that SA is always detected in serum after administration as it is the active metabolite of ASA. Because of that, freshly prepared solutions of ASA should always be used, and SA should also be determined when ASA is present in serum.

In addition, for the cases of ASA, SA, MFA, NFA, IBP, NAP, and DCF, their methyl ester products were also detected in the spiked extract.

As it concerns DCF, it was observed that when the analysis was performed directly in the serum extract without prior derivatization, only a small amount of DCF could be detected, whereas the highest amount was detected as diclofenac lactam. This was firstly reported by El Haj et al. in 1999, who studied the methanolic solutions of DCF [15]. The authors attributed the formation of lactam to the high temperatures applied in the inlet during the GC-MS analysis. Here, in order to investigate this phenomenon, various analyses were conducted and the findings are discussed below.

#### *3.3. Derivatization Study*

In order to enhance the chromatographic peak characteristics of the compounds and the detection sensitivity, the derivatization procedure was performed with different reagents. The aim was to select the optimum procedure and facilitate the simultaneous determination of the eight NSAIDs. In spiked serum samples at 200 μg/mL the derivatization procedures were conducted as described in Section 2.4, and the obtained chromatographic traces with the five reagents are summarized in Table 3. The chromatographic peaks were assigned based on the GC-MS libraries used [NIST v11 and Mass Spectral Library of Drugs, Poisons, Pesticides, Pollutants, and Their Metabolites, 3th Edition (PMW\_Tox3.l)]. All the obtained peaks could be identified, apart from the acetyl derivatives, the spectra of which didn't exist in any of the used MS libraries, web-based libraries, or in the literature. In Table 3, the retention time and the characteristic ions of each drug or drug derivative are juxtaposed for the five derivatization procedures and that without any derivatization. As can be seen, AS and ASA, could not be detected at all with HFBA, TFA, or PFPA-PFPOH derivatization. In some cases, more than one derivative was detected, for example for APAP with MTBSTFA, while for others the underivatized drug

was also detected, such as for NI with silylation reagents. In the case of PFPA-PFPOH, derivatization of NFA gives two peaks which cannot be assigned based on their spectra, as acetyl-derivatives are not registered in the MS spectral libraries used. The major peak is that at 6.9 min, which was attributed to NFA-pfp, while the other peak at 8 min can be another derivative of NFA. Ideally, the optimum procedure should lead to a sole derivative, whereas when the underivatized compound or more than one derivative is detected, the complexity of the detection is increased and the reproducibility of the obtained results is hindered. Based on the obtained results, HFBA, TFA, and PFPA-PFPOH are not considered to be the most appropriate for the simultaneous detection and determination of the eight NSAIDs for the above mentioned reasons.

As it concerns the detection sensitivity of the tested derivatization protocols, silylation provided better results in comparison to the other reagents for all NSAIDs except from IBP. The latter formed a derivative with PFPA/PFPOH which exhibited higher peak area when compared to the BSTFA and MTBSTFA derivatives. Between the two silylation reagents, BSTFA was selected as the best one. In Figure 1, a characteristic total ion chromatogram of serum sample spiked with the studied NSAIDs is presented. The selection was based on the fact that the majority of the peaks had higher peak areas and that MTBSTFA, in the case of APAP, forms two derivatives at the same peak height. Only NI derivatization with BSTFA seems to have low yield, as it gives a small peak of the derivative and the largest peak as NI. However, this is observed for the other silylation reagent as well, whereas NI does not derivatized at all with HFBA, PFPA/PFPOH, and TFA. This means that NI will finally be determined by considering the peak of the underivatized drug. In Table 4, the obtained peak areas are given for the eight drugs, where the higher peak areas can be seen for the silyl-derivatives. In the cases where the peak of the underivatized drug is detected this is noted by an asterisk.

**Figure 1.** TIC of a serum sample spiked with the studied non-steroidal anti-inflammatory drugs (NSAIDs) at 5 μg/mL, derivatized with BSTFA & 1% TMCS.




**Table 4.** Peak areas of the obtained derivatives with the five derivatization protocols applied in the eight NSAIDs.

\* Underivatized compound.

#### *3.4. Application of the Method*

With the aim to apply the method to real clinical human samples, recovery of the optimum procedure was examined and linearity and limits of detection and quantification of the method was assessed. Recovery (R%) was experimentally calculated and expressed as the percentage ratio of the peak areas of the serum spiked before extraction at 10 μg/mL to the peak areas of the serum extract spiked after extraction. The R% ranged from 51.50% for ASA to 85.81% for NAP. The R% for all the studied drugs are presented in Table 5.


**Table 5.** Recovery % of the studied NSAIDs.

Linearity of the method was evaluated by analyzing human serum samples spiked with a mixture of NSAIDs at concentrations of 0.5, 1, 5, 10, 25, and 50 μg/mL using as internal standard 20 μL nordiazepm-D5 (C = 5 μg/mL). BSTFA derivatization was applied with 1% TMCS as described in Sections 2.3 and 2.4, and the peak areas ratios were considered for quantitation. For the case of NI, the peak of underivatized drug was considered, as in low concentrations the derivative is not detected at all. The limits of quantitation (LOQ) were experimentally calculated as signal to noise ratio 10:1 and the limits of detection, and limits of detection (LOD) as signal to noise 3:1. The equations of calibration curves based on linear regression, LOQs, and LODs for all drugs are given in Table 6.

Selectivity was determined on the drug-free blood serum samples obtained from six different healthy volunteers from the laboratory staff. No traces of the studied NSAIDs or other interferences could be detected.

Here it should be noted that freshly prepared ASA solutions in ACN were used for the determination of ASA, as it was observed that ASA transforms to SA. This has been observed to proceed faster in methanol than in ACN [16]. As SA is also present in real samples, its co-determination is required for both reasons.


**Table 6.** Linearity features of the method such as *R*2, limits of detection and quantitation, and the linear equation for all drugs.

\* Nimesulide was determined underivatized.

The accuracy and precision of the method was evaluated within a day (*n* = 4) and over a period of a week (*n* = 3) at low, medium, and high concentration levels (1, 10, 50 μg/mL). The intra-day accuracy and precision were found to be between 1.03%–9.79% and 88%–110%, respectively, while the inter-day accuracy and precision were between 1.87%–10.79% and 91%–113%. In Table 7, the data from accuracy and precision assays are presented.


**Table 7.** Intra- and inter-assay data of the method.

The method was then applied to four clinical cases where exposure to NSAIDs was suspected. The findings are presented in Table 8. S1 and S2 correspond to serum samples which were taken 12 h after APAP ingestion, and they were found below toxic levels (below 50 μg/mL, Rumack-Matthew nomogram). S3 corresponds to a case of a 13 year old girl claiming a suicide attempt with DCF,

however the concentration after 3 h at 2.31 μg/mL was at the therapeutic levels. S4 corresponds to a patient treated with ASA (the sample was taken 40 min after ingestion) and the concentration was found to be below the therapeutic levels. The extracted ion chromatograms of S2 and S3 are presented in Figure 2a,b.


**Table 8.** Concentrations found with the applied method in the serum cases positive to NSAIDs.

**Figure 2.** Extracted ion chromatograms of real serum sample (**a**) S3 at 349 *m*/*z* found positive for diclofenac determined at 2.31 μg/mL and (**b**) S2 at 206 *m*/*z* found positive for APAP determined at 7.5 μg/mL.

#### **4. Discussion**

According to our findings, more efficient extraction of the studied drugs was performed under an acidic pH of 3.7. NI showed also high extraction recovery in higher pH as well.

For most of the studied drugs, derivatization was required for their sensitive detection, however NI provided high, sharp peak in the initial extract and can be best determined without any derivatization.

Some of our findings lead us to the conclusion that determination of these drugs needs cautious sample treatment and data interpretation. We have observed, like previous findings [16], that ASA is not stable in solution and it is hydrolysed into SA, and finally over time both compounds are in equilibrium in the solution. Degradation of ASA has been studied thoroughly [16–18], however there are recent studies where the authors overlook this parameter [1,19]. In order to overcome this transformation, freshly prepared solution should be used for ASA.

Without derivatization, the majority of the NSAIDs seem to be unstable under the high temperature conditions in GC-MS, and a variety of different derivatives are also detected.

The decarboxylated products of NAP, MFA, and NFA are detected to elute some minutes earlier than the parent compound peaks. The formation of these compounds have been reported by others [15] and most probably is due to the fact that the underivatized compounds are labile compared to their derivatives under the same high temperature conditions in GC-MS.

In addition, for almost all the studied drugs, their methylesters were also detected. The esterification process seems to take place at the high temperature conditions in GC-MS, especially in methanolic solutions, whereas this doesn't seem to happen for the derivatives. For this, ACN was used as a solvent.

What is most interesting is that DCF transformed to its dehydrated product, the lactam. To investigate this, a series of experiments was conducted. First, a solution of DCF was analyzed directly with GC-MS, and it was observed that almost all of the diclofenac was transformed to its lactam form. When the same solution was analyzed after derivatization with BSTFA + 1% TMCS only diclofenac-tms was determined, indicating stability of the molecule with derivatization. Then a real human serum sample positive to DCF and a human serum sample spiked with DCF were analyzed after derivatization with BSTFA + 1% TMCS. In both cases, and in contrast to the previous finding, DCF-lactam-tms was mainly detected. Contrastingly, in an aqueous solution of DCF, which was analyzed after extraction and derivatization, similarly to a real sample, only DCF-tms could be detected. This could mean that diclofenac-tms is more stable than the underivatized drug, which is dehydrated under high temperatures in the GC-MS. However, it seems that the serum matrix under acidic conditions enhances the full conversion of DCF to lactam, as DCF-lactam-tms is the sole peak detected in this case, whereas in the absence of serum matrix DCF-tms is detected. The transformation of DCF to DCF-lactam in water samples, pharmaceutical dosage forms, and urine was reported previously but it hasn't been thoroughly studied [11,16]. In a study, DCF-lactam was falsely reported as DCF [20], and in another study DCF-lactam-tms wasn't detected at all [21].

#### **5. Conclusions**

Based on our findings, the most efficient sample preparation protocol for the accurate determination of the studied commonly prescribed NSAIDs is derivatization, more specifically with BSTFA, except from NI where no derivatization step is needed, after acidic extraction.

For some of these drugs, cautious handling is needed, as ASA hydrolyses quickly in solution and DCF is converted to its lactam form in the serum matrix under acidic pH. This conversion is enhanced at high temperatures when its carboxyl group is not protected.

The method needs only 200 μL of blood serum and can determine even trace amounts of the studied compounds.

**Author Contributions:** Conceptualization, A.K. and H.G.; Methodology, A.K. and N.R.; Validation, A.K. and E.T.; Formal Analysis, A.K. and E.M.; Resources, N.R.; Data Curation, A.K. and E.M.; Writing-Original Draft Preparation, A.K. and H.G.; Writing-Review & Editing, G.T.; Supervision, G.T. and H.G.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **Separation Optimization of a Mixture of Ionized and Non-Ionized Solutes under Isocratic and Gradient Conditions in Reversed-Phase HPLC by Means of Microsoft Excel Spreadsheets**

#### **Chrysostomi Zisi, Athina Maria Mangipa, Eleftheria Boutou and Adriani Pappa-Louisi \***

Department of Chemistry, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece; chrizisi@hotmail.com (C.Z.); amangipa@chem.auth.gr (A.M.M.); empoutou@chem.auth.gr (E.B.)

**\*** Correspondence: apappa@chem.auth.gr; Tel.: +30-2310-997765

Received: 2 February 2018; Accepted: 12 March 2018; Published: 18 March 2018

**Abstract:** The crucial role of mobile phase pH for optimizing the separation of a mixture of ionized and non-ionized compounds on a Phenomenex extended pH-range reversed-phase column (Kinetex 5 μm EVO C18) was examined. A previously developed Excel-spreadsheet-based software was used for the whole separation optimization procedure of the sample of interest under isocratic conditions as well as under single linear organic modifier-gradients in different eluent pHs. The importance and the advantages of performing a computer-aided separation optimization compared with a trial-and-error optimization method were realized. Additionally, this study showed that the optimized separation conditions for a given stationary phase may be used to achieve successful separations on new columns of the same type and size. In general, the results of this work could give chromatographers a feel of confidence to establish desired separations of a mixture of ionizable and neutral compounds in reversed-phase columns.

**Keywords:** reversed-phase liquid chromatography; ionizable and non-ionizable analytes; isocratic and gradient elution in different eluent pHs; computer-assisted separation optimization; visualization of predicted chromatograms

#### **1. Introduction**

Reversed-phase liquid chromatography (RPLC) is one of the most widely used chromatographic techniques. It is a consequence of its universality, relatively low costs, and general simplicity of analytical procedures. Still, the development of any method with the desired RPLC separation might be long due to the large number of chromatographic settings that might be adjusted (mobile phase composition, pH, temperature, etc.). The most popular trial-and-error approach has several disadvantages since it is often time-consuming, usually requires a large number of preliminary experiments, and might not be fully efficient. Model-based techniques (fully or semi-automated software programs) can be used in the process of searching for desired RPLC separations [1–5]. These methods usually provide very successful separations based on a series of preliminary experiments. We believe that the number of experimental data may be reduced by utilizing the optimal separation conditions predicted for a specific column by an optimization software to other new columns of the same type and size.

In the present contribution, we report on the optimization of reversed-phase separations of mixture of ionized and non-ionized solutes under isocratic conditions as well as under single linear organic modifier-gradients in different eluent pHs using the Excel-spreadsheet-based software previously developed for simulating and optimizing liquid chromatographic separations [6]. Furthermore, the importance and the advantages of performing a computer-aided separation optimization compared with a trial-and-error optimization method will be confirmed as well as the crucial role of mobile phase pH for optimizing the separation of ionizable compounds [7,8]. This study also provides an example that the optimized separation conditions predicted for a certain column by an optimization software result in almost baseline separation of the test analytes in different columns of the same type and size.

#### **2. Materials and Methods**

#### *2.1. Materials and Reagents*

All chemicals were used as received from commercial sources. The solutes tested are: four monoprotic acids, 2-bromo-4-nitrophenol (2B-4NP), 4-bromo-2-nitrophenol (4B-2NP), 3-bromophenol (3-BP), and 2,4-dibromophenol (2,4-DBP); two monoprotic bases, p-chloro aniline (p-CA) and p-bromoaniline (p-BA); and two non-ionized compounds, benzene (B) and toluene (T). The orthophosphate system (85% H3PO4, KH2PO4, Na2HPO4) was employed for the preparation of buffer solutions in HPLC studies. Acetonitrile (ACN) or methanol (MeOH) of HPLC grade was used as organic modifiers.

#### *2.2. Buffers and Standard Sample Solutions*

Aqueous phosphate buffers with a total ionic strength of 0.2 M were used for preparing the mobile phases with different pH values. The composition of the different buffers employed was found in [9]. The working solutions with single solute or solute mixtures were prepared at a concentration of 360 μg mL−<sup>1</sup> for benzenes, 240 μg mL−<sup>1</sup> for 3-BP and 2,4-DBP, 24 μg mL−<sup>1</sup> for 2B-4NP and 4B-2NP and 8 μg mL−<sup>1</sup> for anilines.

#### *2.3. HPLC System and Conditions*

The liquid chromatography system consisted of a Shimadzu LC-20AD pump, a Shimadzu DGU-20A3 degasser, a model 7125 syringe loading sample injector fitted with a 5 μL loop and a Shimadzu UV–visible spectrophotometric detector (Model SPD-10A, Kyoto, Japan) operating at 254 nm. The column was thermostatted at 25 ◦C by a CTO-10AS Shimadzu column oven.

Four different Phenomenex reversed phase columns of the same size (150 × 4.6 mm) were used. Three of them (Kinetex 5 μm EVO C18, Kinetex 5 μm XB-C18 and Kinetex 2.6 μm XB-C18) were of core–shell technology i.e.**,** with core–shell silica of different particle sizes. The Kinetex 5 μm EVO C18 column exhibits high pH stability from 1–12 similar with that of fully porous organo-silica column (Gemini 5 μm NX-C18), which was also used in this study.

The systematic chromatographic behavior of solutes was investigated on the Kinetex 5 μm EVO C18 column and in mobile phases consisting of diluted aqueous phosphate buffers with a total ionic strength of 0.02 M and a fixed pH value at 2, 3, 5, 7, or 9 modified with ACN. The mobile phase pHs were measured in aqueous buffers before the addition of the organic solvent. Three isocratic runs were performed in different eluent pHs with different ACN volume fraction, (*φ* = 0.3, 0.4, and 0.5) and three *φ*-gradient runs were performed by linearly increasing the ACN content in the mobile phase from an initial value of volume fraction *φ*<sup>0</sup> = 0.3 to a final one *φ<sup>f</sup>* = 0.5. In all gradients, a linear elution program was applied with the same starting time (*tin* = 0 min) but with different gradient duration, *tG*. Moreover, the effect brought on retention of test solutes the use of MeOH as organic modifier instead of ACN was investigated at a fixed eluent *pH* = 3 performing three isocratic runs with *φMeOH* = 0.4, 0.5 and 0.6 as well as three simple linear *φMeOH*-gradient runs from *φ*<sup>0</sup> = 0.4 to *φ<sup>f</sup>* = 0.6 with different gradient duration. The retention data obtained under the above chromatographic conditions, as well as under optimal conditions determined by the optimization procedure adopted in this study, are given in Table S1 in the Supporting Information.

The hold-up time of the Kinetex 5 μm EVO C18 column was estimated to be *t*<sup>0</sup> = 0.983 min, whereas the dwell time *tD* = 0.73 min, at the flow rate set at 1.0 mL min<sup>−</sup>1.

#### *2.4. Excel-Spreadsheet-Based Optimization Software*

The separation optimization procedure under isocratic and simple gradient conditions in different eluent pHs modified with ACN or MeOH was performed using the Excel-spreadsheet-based program developed for simulating and optimizing liquid chromatographic separations [6]. The Excel-spreadsheet -based software 'Isocr&GradSeparationOptimization', with detailed instructions, available on the ACS Publications website at doi:10.1021/acs.jchemed.7b00108 was modified in order to be used for the separation optimization of the mixture of solutes under consideration. An Excel file with the name "Optimization in eluent pH 3 modified with ACN" is provided in the Supporting Information as an example for the whole computer-aided separation optimization procedure adopted in this study.

#### **3. Results**

#### *3.1. Effect of the Eluent pH on the Retention of Ionizable Solutes*

The effect of eluent pH on the retention factor, *<sup>k</sup>*, of a monoprotic acid ( *HA* <sup>↔</sup> *<sup>H</sup>*<sup>+</sup> <sup>+</sup> *<sup>A</sup>*<sup>−</sup> ) or base (*BH*<sup>+</sup> <sup>↔</sup> *<sup>H</sup>*<sup>+</sup> <sup>+</sup> *<sup>B</sup>*) may be expressed as [10–12]

$$k = \frac{k\_0 + k\_1 \mathbf{10^{j(pH - pK)}}}{\mathbf{1} + \mathbf{10^{j(pH - pK)}}} \tag{1}$$

where *k*<sup>0</sup> and *k*<sup>1</sup> are the retention factors of the neutral and fully ionized species of these ionogenic analytes, *j* is an indicator parameter, which is equal to 1 for acids and −1 for bases and *pK* = −*logK*, *K* being the equilibrium constant of the appropriate acid/base equilibrium in the eluent, given by *<sup>K</sup>* <sup>=</sup> [*H*+][*A*<sup>−</sup>] [*HA*] for a monoprotic acid and by *<sup>K</sup>* <sup>=</sup> [*H*+][*B*] [*BH*+] for a monoprotic base. Note that the values of *k*0, *k*1, and *pK* depends on the organic content of mobile phase and can be determined by fitting to Equation (1) either isocratic data obtained in different eluent pHs modified with the same organic content or by fitting gradient data of a fixed change of organic content with a fixed gradient duration in different eluent pHs [13].

The influence of eluent pH on the retention of each of the examined solutes is shown in Figure 1 created by fitting to Equation (1) the experimental retention data obtained under isocratic conditions in different eluent pHs with *φACN* = 0.3, which are given in Table S1. From this figure it is clear the influence of pH on the different types of ionogenic analytes, as well as the superiority of eluent pH 3 in the separation of the test mixture of ionized and non-ionized solutes tested.

**Figure 1.** Variation of *tR* as a function of mobile phase pH for each of the examined solutes. Points are experimental data taken from Table S1 for isocratic runs performed in different eluent pHs with *φACN* = 0.3. Lines are obtained by fitting experimental data of ionized solutes to Equation (1).

#### *3.2. Computer-Aided Separation Optimization in Different Eluent pHs*

A computer-assisted separation optimization of a mixture of solutes comprising a visualization of predicted chromatograms involves the following steps [6]:


The whole separation optimization procedure of the test analytes under isocratic conditions as well as under single linear *φ*-gradients in different eluent pHs was implemented on different MS Excel spreadsheets. The optimization procedure followed for separation optimization in eluent pH 3 is given as an example.

The isocratic retention data of solutes, *tR*(*exp*), given in Table S1 and obtained on the Kinetex 5 μm EVO C18 column with different ACN volume fraction (*φ* = 0.3, 0.4, and 0.5) in eluent pH 3 were fitted to the quadratic retention model *lnk* <sup>=</sup> *<sup>c</sup>***<sup>0</sup>** <sup>+</sup> *<sup>c</sup>***1***ϕ*<sup>+</sup> *<sup>c</sup>***2***ϕ***<sup>2</sup>** [14] (where *<sup>k</sup>* is the solute retention factor, *<sup>k</sup>* <sup>=</sup> (*tR* <sup>−</sup> *<sup>t</sup>***0**)/*t***0**, and *<sup>φ</sup>* is the volume fraction of ACN in the mobile phase) using the spreadsheet 'retention fit' of the file "Optimization in eluent pH 3 modified with ACN" provided in the Supporting Information. Note that, although the spreadsheet 'retention fit' is designed for fitting isocratic retention data to the quadratic retention model, the linear retention model, *lnk* <sup>=</sup> *<sup>c</sup>***<sup>0</sup>** <sup>+</sup> *<sup>c</sup>***1***ϕ*, could be used instead, if the chromatographic behavior of solutes was studied in a very narrow range of *φ.*

After correction of the baseline of the experimental chromatograms recorded in eluents with *φACN* = 0.3 and 0.4 by pressing *Ctrl+q* on the spreadsheet with the name 'BLcor.', available in "Optimization in eluent pH 3 modified with ACN" file, modeling of the peak shapes is straightforward. For fitting the peak shapes of each solute recorded in different chromatograms with *φACN* = 0.3 and 0.4 to the model

$$y = h(t\_R) \exp\left(\frac{-\left(t - t\_R\right)^2}{D\left(t\_R\right)^2}\right).$$

the spreadsheet 'peak shape fit' is used. This spreadsheet is designed to estimate the peak shape parameters (*h*0, *h*1, *h*2, *D*0, *D*1, and *D*2) of a quadratic dependence of both peak height, *h*, and peak width, *<sup>D</sup>*, on *tR*, given by *<sup>h</sup>*(*tR*) <sup>=</sup> *<sup>h</sup>***<sup>0</sup>** <sup>+</sup> *<sup>h</sup>***1***tR* <sup>+</sup> *<sup>h</sup>***2***t***<sup>2</sup>** *R* and *<sup>D</sup>*(*tR*) <sup>=</sup> *<sup>D</sup>***<sup>0</sup>** <sup>+</sup> *<sup>D</sup>***1***tR* <sup>+</sup> *<sup>D</sup>***2***t***<sup>2</sup>** *R*. It should be noted that the data for two experimental chromatograms are enough for the above procedure since the peaks experimentally recorded in this study permit a linear dependence of both the peak height and peak width on *tR* instead of the quadratic one initially assumed.

After the retention times and peak shape parameters for all of the solutes studied under isocratic conditions have been estimated, the values of these parameters (i.e., *c*0, *c*1, *c*2, *h*0, *h*1, *h*2, *D*0, *D*1, and *D*2) are transferred into the spreadsheet 'isocr.optim.' in "Optimization in eluent pH 3 modified with ACN" file. A screenshot of this spreadsheet is displayed in Figure 2. The isocratic separation optimization of the sample mixture is easily automated by pressing *Ctrl+w*. Then, the minimum resolution, *Rs*, and the maximum of *tR* values, *tR*(*max*), of all solutes separated under isocratic conditions in eluent pH 3 are recorded as a function of the organic content *φ* on columns A, B, and C, where *φ* is altered between two values (*φ*(*min*) *=* 0.3 and *φ*(*max*) *=* 0.5, preset in cells B15 and B16) with a selected interval *δφ* = 0.005 placed in cell B17. Simultaneously, the values of *Rs* and *tR*(*max*) vs. *φ* are plotted in a graph, see the inset

Graph A of the layout of this worksheet, and simulated chromatograms are generated for each mobile phase strength *φ* in the inset Graph B of the worksheet. The execution of the macro is accomplished by finding the optimal eluent, *φACN* = 0.365, which leads to the best separation of the sample—i.e., the separation with a desirable value of resolution—*Rs* = 1.5 (preset in cell D11), in the shortest separation time, which in this case is only 10.36 min. The inset Graph B of Figure 2 depicts a perfect similarity between the simulated chromatogram created for the optimal eluent with *φACN* = 0.365 (plotted as the red solid line) and the original experimental one (plotted as the blue dashed line).


**Figure 2.** Screenshot of the MS Excel supplementary spreadsheet 'isocr. optim.' used for isocratic separation optimization of solutes in eluent pH 3 modified with ACN. See the text for details.

A procedure similar with that described above for isocratic separation optimization and simulation is also followed for optimizing single linear gradient conditions and simulating chromatograms obtained under selected different gradient profiles. A screenshot of the spreadsheet 'grad. optim.' is depicted in Figure 3.

The values of *φ<sup>0</sup>* and *φf*—i.e., *φ<sup>0</sup>* = 0.3 and *φ<sup>f</sup>* = 0.5 for the data set analyzed—are placed in cells G2 and G3, respectively, the features of chromatographic system—i.e., the values of *tD* = 0.73 min and *t*<sup>0</sup> = 0.983 min in cells B9 and B10—whereas the estimated retention and peak shape parameters of all solutes are transferred in cells I2:Q10. Note that, in this procedure the retention adjustable parameters, *c*0, *c*1, and *c*2, were determined in the worksheet 'retention fit' from initial isocratic conditions. In contrast, the peak shape parameters, *h*0, *h*1, *h*2, *D*0, *D*1, and *D*2, were obtained from gradient runs between *φ*<sup>0</sup> = 0.3 to *φ<sup>f</sup>* = 0.5 with different gradient durations, *tG* = 5 and 20 min, in the worksheet 'peak shape fit', since the peak widths in gradient elution are normally compressed compared to those obtained under isocratic conditions. By pressing *Ctrl+e*, the minimum resolution, *Rs*, and the maximum of *tR* values, *tR*(*max*), of all solutes separated under gradient conditions with different gradient durations, *tG*, are calculated in columns A, B, and C, where *tG*, varied between two values, *tG*(*min*) = 5 min and *tG*(*max*) = 20 min, defined in cells B15 and B16 with a selected interval *δtG* = 0.5 min (placed in cell B17). Moreover, a plot is simultaneously created with these values of *Rs* and *tR*(*max*) vs. *tG*, as well as a graph is generated for simulated chromatograms obtained under different gradient profiles. Again, the execution of the macro is accomplished by finding the optimal gradient duration, *tG* = 6.5 min, which leads to the best separation of the sample—i.e., the separation with a satisfactory resolution, i.e., *Rs* = 1.5 (preset in cell D10)—in the shortest separation time, which in this case is only 8.15 min. The inset Graph B of Figure 3 depicts a perfect similarity between the predicted/simulated chromatogram in the optimal gradient elution with a duration *tG* = 6.5 min (plotted as the red solid line) and the original experimental one (plotted as the blue dashed line).

**Figure 3.** Screenshot of the MS Excel supplementary spreadsheet 'grad. optim.' used for separation optimization under single linear gradient conditions of solutes in eluent pH 3 modified with ACN. See the text for details.

The same optimization approach is applied to chromatographic data obtained under isocratic conditions as well as under single linear *φ*-gradients in other examined eluent pHs—i.e., at pH = 2, 5 and 7, respectively—depicted in Table S1. The retention data recorded in eluent pH = 9 were not analyzed by means of the above spreadsheet optimization program since peak shape distortions appeared for some solutes at that mobile phase pH. The optimal conditions found for the separation of test mixture of solutes on the Kinetex 5 μm EVO C18 column in different eluent pHs are summarized in Table 1. The chromatograms recorded under the optimal gradient conditions found by the optimization algorithm are depicted in Figure 4. In the same Figure, the influence of mobile phase pH on the elution order as well as on the peak shape of ionizable compounds is also illustrated.

**Table 1.** Optimal conditions found for the separation of the mixture of solutes using an Excel spreadsheet-based optimization program.

**Figure 4.** UV detected chromatograms of the mixture of 8 ionized and non-ionized solutes obtained on Kinetex 5 μm EVO column under optimal gradient conditions in different eluent pHs. The elution order of solutes is shown in the Figure. See Table 1 for details of optimal gradient conditions.

#### *3.3. Utility of Computer-Aided Separation Optimization*

Once the optimal separation conditions of the sample of interest were found for the Kinetex 5 μm EVO C18 column by the proposed Excel-spreadsheet-based software, the effectiveness of the same optimal conditions into separation of solutes in different reversed-phase-type columns of the same size was tested. Indeed, a perfect resolution of the sample of interest is achieved in the chromatograms recorded in different columns under the optimal conditions determined for the Kinetex 5 μm EVO C18 column; see, as an example, Figure 5 for the application of optimal gradient conditions determined at eluent pH 3. Consequently, the optimal conditions derived for the separation of a sample on a certain column by the optimization algorithm could be successfully applied to other columns of the same type and size. Moreover, in Figure 5, the superiority of the core–shell technology columns and especially of the Kinetex EVO column is illustrated, since it is clear that a complete separation of test compounds was achieved within the minimum run time.

**Figure 5.** UV detected chromatograms of the mixture of 8 ionized and non-ionized solutes obtained on different columns under optimal gradient conditions found for eluent pH = 3 and Kinetex 5 μm EVO column. The elution order of solutes is shown in the Figure. See Table 1 for details of optimal gradient conditions.

The importance and the advantages of performing a computer-aided separation optimization are clearly shown in Figures 6 and 7. Figure 6 is a screenshot of the MS Excel spreadsheet 'isocr. optim.' used for isocratic separation optimization of test solutes in eluent pH 3 modified with MeOH. The retention data recorded in this eluent, depicted in Table S1, were analyzed by means of the Excel-spreadsheet optimization program following a procedure similar to that described above for separation optimization in eluent pH 3 modified with ACN. The optimal eluent, *φMeOH*= 0.575 was automatically found by pressing *Ctrl+w*. However, the selection of the optimal separation conditions is also possible from a good appreciation of the inset Graph A of Figure 6, which is also automatically created as described above. As shown in this figure, *tR*(*max*), decreases with increasing organic content *φ* in the mobile phase (purple circle markers), as is expected for a reversed-phase-type elution. However, the dependence of *Rs* (the resolution of the least resolved pair of adjacent solutes) on *φ* (depicted by the green diamond markers) is rather peculiar. For example, the resolution in mobile phases with either *φMeOH* = 0.46 or *φMeOH* = 0.52 is almost zero, which means that at least two solutes co-elute under the above isocratic conditions even though the run times, i.e., the values of *tR*(*max*) correspond to these eluent concentrations are longer than that in the optimal eluent with *φMeOH* = 0.575. Consequently, the foreknowledge of the precise dependence of *Rs* and *tR*(*max*) upon *φ*, provided by Excel-spreadsheet-based software adopted in this study and not by a trial-and-error method gives chromatographers a feel of confidence for the selection of the optimal conditions for a

desired separation. Indeed, Figure 7 depicts a perfect resolution of the test solutes in the chromatogram recorded on the Kinetex 5 μm EVO C18 column and in the optimal eluent with *φMeOH* = 0.575. In contrast, the resolution of the same sample in a mobile phase with *φMeOH* = 0.5 is worse than that in an eluent *φMeOH* = 0.575 even though the separation time is longer, see also Figure 7.

**Figure 6.** Screenshot of the MS Excel spreadsheet 'isocr. optim.' used for isocratic separation optimization of solutes in eluent pH 3 modified with MeOH. See the text for details.

**Figure 7.** UV detected chromatograms of the mixture of eight ionized and non-ionized solutes obtained on Kinetex 5 μm EVO column under different isocratic conditions in eluent pH = 3 modified with MeOH. The elution order of solutes is shown in the Figure.

#### **4. Conclusions**

In this study, the whole separation optimization procedure of the test analytes under isocratic conditions as well as under single linear *φ*-gradients in different eluent pHs was successfully implemented by using an Excel-spreadsheet-based software, a user-friendly and widespread software platform, based on a few initial experiments for each eluent pH: three isocratic runs and two single linear *φ*-gradient runs performed in the studied range of mobile phase strength, *φ.* In the adopted optimization process, for computational simplicity, the solute retention parameters were obtained from the analysis of isocratic data, whereas a Gaussian function was used to fit peak shapes. The importance and the advantages of performing a computer-aided separation optimization compared with a trial-and-error optimization method were realized. The optimal separation conditions derived by the optimization algorithm for the separation of the sample of interest on a certain reversed-phase column were successfully applied to other same-type columns of the same size. The superiority of the core–shell technology columns and especially of the Kinetex EVO column was illustrated. In general, we consider that the results of this study could give chromatographers a feel of confidence for the selection of the optimal separation conditions for a sample of ionizable and neutral compounds in reversed-phase columns.

**Supplementary Materials:** The following are available online at www.mdpi.com/2297-8739/5/1/19/s1, Table S1: Experimental retention data, *tR*(*exp*) in min, of test solutes obtained on the Kinetex 5 μm EVO C18 column and under isocratic and linear gradient conditions in different eluent pHs modified with ACN or MeOH (PDF), Excel-spreadsheet-based software: "Optimization in eluent pH 3 modified with ACN" (ZIP).

**Author Contributions:** The experimental design was constructed and supervised by A.P.-L. The experiments were performed by A.M.M. and E.B. The data were analyzed by C.Z. The manuscript was drafted and written by A.P.-L. and C.Z.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Separations* Editorial Office E-mail: separations@mdpi.com www.mdpi.com/journal/separations

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18