**Polycaprolactone Composite Micro/Nanofibrous Material as an Alternative to Restricted Access Media for Direct Extraction and Separation of Non-Steroidal Anti-Inflammatory Drugs from Human Serum Using Column-Switching Chromatography**

**Hedvika Raabová 1 , Lucie Chocholoušová Havlíková 1 , Jakub Erben <sup>2</sup> , Jiˇrí Chvojka <sup>2</sup> , František Švec <sup>1</sup> and Dalibor Šatínský 1,\***


**Abstract:** Application of the poly-E-caprolactone composite sorbent consisting of the micro- and nanometer fibers for the on-line extraction of non-steroidal anti-inflammatory drugs from a biological matrix has been introduced. A 100 µL human serum sample spiked with ketoprofen, naproxen, sodium diclofenac, and indomethacin was directly injected in the extraction cartridge filled with the poly-E-caprolactone composite sorbent. This cartridge was coupled with a chromatographic instrument via a six-port switching valve allowing the analyte extraction and separation within a single analytical run. The 1.5 min long extraction step isolated the analytes from the proteinaceous matrix was followed by their 13 min HPLC separation using Ascentis Express RP-Amide (100 × 4.6 mm, 5 µm) column. The recovery of all analytes from human serum tested at three concentration levels ranged from 70.1% to 118.7%. The matrix calibrations were carried out in the range 50 to 20,000 ng mL−<sup>1</sup> with correlation coefficients exceeding 0.996. The detection limit was 15 ng mL−<sup>1</sup> and the limit of quantification corresponded to 50 ng mL−<sup>1</sup> . The developed method was validated and successfully applied for the sodium diclofenac determination in real patient serum. Our study confirmed the ability of the poly-E-caprolactone composite sorbent to remove the proteins from the biological matrix, thus serving as an alternative to the application of restricted-access media.

**Keywords:** restricted access media; nanofibers; microfibers; on-line extraction; biological samples; column-switching chromatography

,

### **1. Introduction**

Non-steroidal anti-inflammatory drugs (NSAID) are widely used in pain, osteoarthritis, and rheumatoid arthritis treatment. A mechanism of their action lies in the inhibition of cyclo-oxygenase enzyme, an enzyme serving prostaglandin biosynthesis. These drugs play a crucial role in the suppression of the inflammatory response of an organism. The reduction in prostaglandin levels results in antipyretics, analgesics, and anti-inflammatory activity of NSAID. The most widely used NSAIDs in clinical practice are acetylsalicylic acid, paracetamol, indomethacin, ibuprofen, ketoprofen, sodium diclofenac, flurbiprofen, mefenamic acid, piroxicam, and nabumeton. Monitoring of NSAID in body fluid is essential for toxicological and pharmacokinetics studies. Their determination as a part of rational pharmacotherapy that aims to reduce risks and achieve disease treatment goals is less common. The levels are specifically monitored either in patients with an impaired renal function when the insufficient excretion occurs or when non-compliance is suspected [1,2]. The

**Citation:** Raabová, H.; Havlíková, L.C.; Erben, J.; Chvojka, J.; Švec, F.; Šatínský, D. Polycaprolactone Composite Micro/Nanofibrous Material as an Alternative to Restricted Access Media for Direct Extraction and Separation of Non-Steroidal Anti-Inflammatory Drugs from Human Serum Using Column-Switching Chromatography. *Nanomaterials* **2021**, *11*, 2669. https://doi.org/10.3390/ nano11102669

Academic Editor: Angelo Ferraro

Received: 3 September 2021 Accepted: 6 October 2021 Published: 12 October 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

pharmaceutical industry is another sector where NSAID are determined in pharmaceutical formulations to ensure safety and effectiveness of drugs [3].

Numerous methods for NSAID determination in biological fluids including gas chromatography [4–6], capillary electrophoresis [7–9], and liquid chromatography [10,11] were developed over the past years. Although the analytical instruments were designed to provide increasingly faster and more sensitive analysis, the treatment of the biological matrixes prior to the instrumental analysis remains the bottleneck in the drug level monitoring [12]. The NSAID are usually determined in plasma, where the direct analysis of untreated matrix can lead to lower sensitivity, unprecise results, and in the worst case to irreversible damage of the chromatography column caused by protein precipitation. Thus, the sample pretreatment step cannot be omitted. This step is mainly based on timeconsuming, laborious, and error-prone methods, such as protein precipitation, solid-phase extraction, and liquid–liquid extraction that can be potentially hazardous for laboratory staff [13]. Several approaches have been developed to overcome these shortcomings. The emphasis has been primarily placed on higher speed, automation, reduced consumption of organic solvents, smaller sample volumes, and increased selectivity [13,14]. Restricted access media (RAM) represent one of the applicable methods [15].

RAM sorbents separate the low molecular analytes from macromolecular interferences mostly via size exclusion effect. This analyte extraction and macromolecules removal occurs simultaneously after the injection of the untreated biological fluids [15]. The sample preparation is simplified and the analysis time is shortened while maintaining the desired extraction efficiency and avoiding column clogging [16]. The extraction mechanism of the original RAM is primarily based on their two surfaces. The outer surface is covered by hydrophilic groups avoiding access of macromolecules in the material, and the hydrophobic functional groups on the inner pore surface that are responsible for analyte retention. Additionally, the pores in the sorbent material contribute to the matrix removal by the size-exclusion mechanism [17].

The potential of RAM can be adequately utilized in on-line liquid chromatography extractions. The RAM cartridges are most often coupled with a column-switching chromatography system [16]. This setup was reported several times for NSAID analysis [18,19]. The main component of this system is a double position six-port selection valve that redirects the mobile phases between both extraction cartridge and analytical column. The extraction begins after the sample is injected in the cartridge. The undesired matrix components are removed from the analytes using the washing mobile phase. The columnswitching valve then changes its position to complete the extraction step, and the analytes are eluted in the analytical column where they are separated using the separation mobile phase. The washing mobile is meanwhile transferred in the waste container. This process can be carried out either in a straight-flush or in a back-flush mode differing in the way the analytes are eluted from the extraction cartridge. They are injected and eluted from the extraction cartridge in the same direction as washing mobile phase flow in the former approach while the mobile phase elutes the analytes in the opposite direction to the sample injection in the latter [20].

Advanced approaches combine the RAM concept with the molecularly imprinted polymers, magnetism, or supramolecular solvents to enhance selectivity and specificity of the analyte extraction [21–23]. The use of nanomaterials is also included in these developments. For example, application of the restricted access nanotubes was reported [24]. We demonstrated first nanofibers as an alternative to RAM elsewhere [25]. We confirmed that the composite material produced from poly-ε-caprolactone exhibited satisfactory efficiency in extraction of parabens from human serum and bovine milk. The idea of using nanofibers as an alternative to RAM emerged as a result of their characteristics. The nanofibers have a large surface area to volume ratio enabling them to capture considerable quantities of low molecular weight analytes. In contrast, we speculated that the macromolecules are not retained because of the curvature of the fibers. The macromolecules appeared not to be flexible enough to be attached at multiple points to the highly curved nanofiber for their

sufficient retention. This is not an issue with the low-molecular weight analytes [25]. Additionally, the large spaces between the fibers support the macromolecule passage through the sorbent to the waste.

We described the advantages of material combining poly-E-caprolactone micro- and nanofibers (micro/nano PCL) for the on-line extraction of analytes from milk and serum matrixes in our previous study [25]. We confirmed the protein removal capability, good mechanical stability in the extraction/HPLC system, and re-usability of this sorbent for more than 300 analyses. The current study extends exploration of the potential of micro/nano PCL material in therapeutic drug monitoring and a method for NSAID determination in human serum samples was developed. This method was validated according to the International Council for Harmonization guideline and used for the determination of diclofenac content in a real patient serum.

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

### *2.1. Reagents and Materials*

Ketoprofen (≥98.0%, KET), naproxen (≥98.0%, NAP), sodium diclofenac (≥98.5%, DCF), and indomethacin (≥98.5%, IND) were purchased from Sigma-Aldrich (Darmstadt, Germany) and used as model analytes. Phosphoric acid (Honeywell, Morris Plains, NJ, USA), acetonitrile (ACN), and methanol (MeOH) (VWR, Paris, France) served as mobile phase modifiers. The standard solutions were acidified by acetic acid from Penta (Prague, Czech Republic). The proteinaceous matrix was prepared by dissolution of lyophilized human serum Lyo hum N (Erba Lachema, Brno, Czech Republic). Water was purified by Millipore Milli-Q Direct Water Purification System from Merck (Darmstadt, Germany). The composite poly-ε-caprolactone sorbent was fabricated at the Technical University of Liberec from polyε-caprolactone (Mw 43.000) provided by Polysciences (Heddesheim, Germany), and chloroform (≥97%) and ethanol (99.97%) bought from Penta Chemicals (Pardubice, Czech Republic).

### *2.2. Instrumentation*

### 2.2.1. Column-Switching Chromatography System

A Shimadzu Prominence instrument (Shimadzu Corporation, Kyoto, Japan) equipped with three LC-20AD Prominence pumps, a DGU-20A Prominence on-line degasser, a SIL 20AC Prominence autosampler, a CTO-20AC Prominence column oven, an SPD-M20A Prominence UV/VIS photodiode array detector, and a CBM-20A Prominence system controller was used for the separation and analysis. Switching between the analytical column and the extraction cartridge was provided via the six-port high-pressure flow line switching valve FCV-12AH. LC Solution software (version 5.97, Shimadzu Corporation, Kyoto, Japan) controlled the instrument. The separation of NSAID extracted from the human serum was carried out on the Ascentis Express RP-Amide (100 × 4.6 mm, 5 µm) analytical column combined with the RP-Amide (5 × 4.6 mm, 2.7 µm) guard column (Sigma-Aldrich, Darmstadt, Germany). The system configuration in extraction and separation mode is depicted in the Figure 1.

### 2.2.2. Meltblown and Electrospun Fibers

Poly-E-caprolactone nanofibers and microfibers were fabricated using a process described in Supporting Information and was firstly reported elsewhere [26,27]. The meltblown equipment (Laboratory equipment J&M Laboratories, Ashland, OH, USA) in combination with DC electrospinning lab-made system based on a multi-needle spinner was used for the preparation of the fibers. The production equipment consisting of both electrospinning and meltblown systems is depicted in Figure S1 in Supporting Information section. A scanning electron microscope VEGA 3 (Tescan, Brno, Czech Republic) was used for the imaging of the fibrous sorbent. The morphology of micro/nanofibrous material is shown in Supporting Information section in Figure S2.

### *2.3. Preparation of Extraction Cartridge*

A commercially available PEEK cartridge 4.6 × 10 mm (Merck, Darmstadt, Germany) was manually filled with 44 mg of the composite micro/nanofibrous PCL. The details of extraction cartridge preparation and schematic of filling are presented in Supporting Information section and in Figure S3. The cartridge in a plastic holder was then placed in the chromatographic system and washed with the mobile phase containing 10% ACN in 0.085% (*v*/*v*) aqueous phosphoric acid for 15 min and neat ACN was then injected six times. This washing was carried out to ensure complete removal of impurities that could remain in fibers after their production. *Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 4 of 12

**Figure 1.** Column-switching chromatographic system in (**A**) on-line extraction and (**B**) on-line ex-**Figure 1.** Column-switching chromatographic system in (**A**) on-line extraction and (**B**) on-line extraction-separation mode.

### *2.4. Preparation of Standard and Matrix Solutions*

traction-separation mode.

remain in fibers after their production.

*2.4. Preparation of Standard and Matrix Solutions*

2.2.2. Meltblown and Electrospun Fibers Poly-ɛ-caprolactone nanofibers and microfibers were fabricated using a process described in Supporting Information and was firstly reported elsewhere [26,27]. The meltblown equipment (Laboratory equipment J&M Laboratories, Ashland, OH, USA) in combination with DC electrospinning lab-made system based on a multi-needle spinner was used for the preparation of the fibers. The production equipment consisting of both electrospinning and meltblown systems is depicted in Figure S1 in Supporting Information section. A scanning electron microscope VEGA 3 (Tescan, Brno, Czech Republic) was used for the imaging of the fibrous sorbent. The morphology of micro/nanofibrous material is shown in Supporting Information section in Figure S2. *2.3. Preparation of Extraction Cartridge* A commercially available PEEK cartridge 4.6 × 10 mm (Merck, Darmstadt, Germany) was manually filled with 44 mg of the composite micro/nanofibrous PCL. The details of The stock KET, NAP, DCF, and IND solutions were prepared at a concentration of 1 mg mL−<sup>1</sup> by dissolving the individual standard substances in MeOH. These solutions were stored in the dark at 4 ◦C. The mix stock solution containing 0.2 mg mL−<sup>1</sup> NSAID was prepared by mixing the stock solutions. An appropriate volume of the mix stock solution was diluted with a water–acetonitrile–acetic acid solution (50/48/2, *v*/*v*/*v*) to obtain standards solutions at seven concentration levels of 50; 100; 500; 1000; 5000; 10,000; and 20,000 ng mL−<sup>1</sup> . These solutions were used for the plotting the calibration curve. Commercially available lyophilized human serum was reconstituted according to manufacturer instructions and then diluted ten times with 20% aqueous ACN. This serum solution was spiked with the mix stock solution to obtain matrix solutions at concentrations of 50; 100; 500; 1000; 5000; 10,000; and 20,000 ng mL−<sup>1</sup> and these solutions were centrifuged at 14,000× *g* rpm (21,578× *g*) for 15 min. The supernatant was then injected in the chromatographic system and the matrix calibration curve plotted. Matrix solutions were prepared fresh daily.

The stock KET, NAP, DCF, and IND solutions were prepared at a concentration of 1 mg mL−<sup>1</sup> by dissolving the individual standard substances in MeOH. These solutions were stored in the dark at 4 °C. The mix stock solution containing 0.2 mg mL−<sup>1</sup> NSAID was prepared by mixing the stock solutions. An appropriate volume of the mix stock solution

extraction cartridge preparation and schematic of filling are presented in Supporting Information section and in Figure S3. The cartridge in a plastic holder was then placed in the chromatographic system and washed with the mobile phase containing 10% ACN in

### *2.5. Real Sample*

A real human serum was obtained from a patient administered with 250 mL continual intravenous infusion containing 75 mg of sodium diclofenac. Blood sampling was carried out in the Department of Clinical Biochemistry and Diagnostics of the University Hospital Hradec Králové. This real life serum was handled the same way as the matrix working solutions. First, it was diluted 10 times with 20% aqueous ACN and then centrifuged at 14,000× *g* rpm (21,578× *g*) for 15 min.

### *2.6. Analytical Method*

The analytical run started with the extraction step when 100 µL sample, standard or matrix solution, was injected in the extraction cartridge filled with micro/nano PCL. The NSAID were captured by the sorbent while the proteins and other potentially interfering macromolecular substances were removed by the washing with the mobile phase composed of 10% ACN in 0.085% aqueous H3PO4. Simultaneously, the analytical column was conditioned using the 30% ACN in 0.085% aqueous H3PO4. The column-switching valve redirected this separation mobile phase after 1.5 min in the extraction cartridge. Hereby, the analytes were eluted in the analytical column and separated using the ACN gradient increasing the ACN percentage to 45% in 2.5 min and this mobile phase was pumped through the column for another 1.5 min. Then, the percentage was ramped to 55% in 4 min and to 75% in 0.5 min. Finally, the ACN percentage decreased to the initial concentration in 1 min at which the analytical column was re-equilibrated for 3 min. The flow rate of both mobile phases was held on 1 mL min−<sup>1</sup> . The total analysis time was 15 min. The separated analytes were detected using the diode array detector. KET, DCF, and IND were monitored at 270 nm while NAP at 232 nm. The column was held at 20 ◦C since the micro/nano PCL dissolve in ACN at higher temperatures [26].

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

### *3.1. Optimization of Chromatographic Separation*

We tested ACN and MeOH as organic components of the mobile phase with ACN produced peaks with a better symmetry. The NSAID drugs are weak acids. Therefore, 0.085% aqueous H3PO<sup>4</sup> solution pH 2.6 was used as the aqueous part of the mobile phase to increase the retention in the column. We tested two analytical columns Ascentis Express RP-Amide (100 × 4.6 mm, 5 µm) and Ascentis Express F5 (100 × 4.6 mm, 5 µm) for the separations. Both columns use the core-shell particles technology improving the efficiency and reducing the flow resistance in column switching system. However, they are more susceptible to impurities from proteinaceous samples resulting in an increase in back pressure and eventually column clogging. Therefore, the Ascentis Express (5 × 4.6 mm, 2.7 µm) guard columns packed with the same stationary phase were inserted in the system to protect the analytical column. The Ascentis Express RP-Amide analytical column enabled a better separation of the analytes. Ascentis Express F5 stationary phase was not well suited because the extensive peak broadening and coelution of ketoprofen-naproxen and sodium diclofenac-indomethacin pairs as presented in Figure 2.

### *3.2. Optimization of On-Line Extraction*

We reported elsewhere [25] that the removal of proteins from nanofibrous sorbent using an aqueous mobile phase occurs within the first minute. This finding was confirmed by monitoring elution at a wavelength of 280 nm that is the absorption maximum for proteins. Human serum contains in addition to proteins also more lipophilic ballast substances, such as, for example, lipoproteins and vitamins, that are difficult to remove from the system since aqueous washing is insufficient. In contrast, a higher concentration of organic solvents in the washing mobile phase and excessive washing lead to undesirable losses of desired analytes. Therefore, the composition of the washing mobile phase and the duration of extraction step had to be optimized to remove most of the lipophilic ballast molecules without any loss of the analytes. Removal of these compounds using the mobile

phases containing ACN concentration of 5%, 10%, 15%, and 20% were tested. The recovery of adsorbed analytes on nanofibers decreased with the rising ACN concentration. However, the effect of this process on the removal of ballast substances was negligible. This is why we used a mixture comprising 10% ACN in 0.085% aqueous phosphoric acid. This washing mobile phase was the best compromise between analyte loss and ballast removal. The duration of the extraction step was studied applying the similar approach. The peak of matrix impurities was observed in the chromatogram at 270 nm after 1 min washing. Extension of the washing time by 30 s resulted in a reduction in the area of the matrix ballast peaks without reducing the peak areas of analytes. No further improvement was observed after further increasing the extraction time to 2 min. Thus, the extraction time was finally held at 1.5 min. Chromatogram of standard solution and spiked serum under the optimized conditions of extraction and separation is shown in Figure 3. *Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 6 of 12 **Figure 2.** Chromatograms of standard solution spiked with 5000 ng mL<sup>−</sup><sup>1</sup> NSAID separated on (**A**) Ascentis Express RP-Amide column and (**B**) Ascentis Express F5 column at a wavelength 270 nm. 1—ketoprofen, 2—naproxen, 3—sodium diclofenac, 4—indomethacin*. 3.2. Optimization of On-Line Extraction* We reported elsewhere [25] that the removal of proteins from nanofibrous sorbent using an aqueous mobile phase occurs within the first minute. This finding was confirmed by monitoring elution at a wavelength of 280 nm that is the absorption maximum for proteins. Human serum contains in addition to proteins also more lipophilic ballast substances, such as, for example, lipoproteins and vitamins, that are difficult to remove from the system since aqueous washing is insufficient. In contrast, a higher concentration of

*Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 6 of 12

**Figure 2.** Chromatograms of standard solution spiked with 5000 ng mL<sup>−</sup><sup>1</sup> NSAID separated on (**A**) Ascentis Express RP-Amide column and (**B**) Ascentis Express F5 column at a wavelength 270 nm. 1—ketoprofen, 2—naproxen, 3—sodium diclofenac, 4—indomethacin*.* **Figure 2.** Chromatograms of standard solution spiked with 5000 ng mL−<sup>1</sup> NSAID separated on (**A**) Ascentis Express RP-Amide column and (**B**) Ascentis Express F5 column at a wavelength 270 nm. 1—ketoprofen, 2—naproxen, 3—sodium diclofenac, 4—indomethacin. was observed after further increasing the extraction time to 2 min. Thus, the extraction time was finally held at 1.5 min. Chromatogram of standard solution and spiked serum under the optimized conditions of extraction and separation is shown in Figure 3.

ery of adsorbed analytes on nanofibers decreased with the rising ACN concentration. However, the effect of this process on the removal of ballast substances was negligible. **Figure 3.** Chromatograms of (**A**) standard solution and (**B**) matrix solution spiked with 1000 ng mL−<sup>1</sup> NSAID at a wavelength 270 nm. 1—ketoprofen, 2—naproxen, 3—sodium diclofenac, 4—indomethacin.

#### This washing mobile phase was the best compromise between analyte loss and ballast *3.3. Extraction Efficiency*

removal. The duration of the extraction step was studied applying the similar approach. The peak of matrix impurities was observed in the chromatogram at 270 nm after 1 min washing. Extension of the washing time by 30 s resulted in a reduction in the area of the matrix ballast peaks without reducing the peak areas of analytes. No further improvement was observed after further increasing the extraction time to 2 min. Thus, the extraction time was finally held at 1.5 min. Chromatogram of standard solution and spiked serum under the optimized conditions of extraction and separation is shown in Figure 3. The analyte recovery from the human serum matrix was compared to the recovery obtained using extraction from standard solutions and expressed as a percentage value considering the standard solution being 100%. Figure 4 demonstrates that at levels up to 100 ng mL−<sup>1</sup> , the analytes are extracted from the serum with an efficiency comparable to that found for the standard solutions. KET is an exception since its recovery exceeded 100% at almost all concentration levels. It can be caused by coelution with a matrix interference that increased the total peak area. This finding reduces the accuracy of the method for KET. However, the RSD values ranging from 1.16% to 0.017% and the precision is not affected.

This is why we used a mixture comprising 10% ACN in 0.085% aqueous phosphoric acid.

**Figure 3.** Chromatograms of (**A**) standard solution and (**B**) matrix solution spiked with 1000 ng mL−1 NSAID at a wave-

The analyte recovery from the human serum matrix was compared to the recovery obtained using extraction from standard solutions and expressed as a percentage value considering the standard solution being 100%. Figure 4 demonstrates that at levels up to

that found for the standard solutions. KET is an exception since its recovery exceeded 100% at almost all concentration levels. It can be caused by coelution with a matrix interference that increased the total peak area. This finding reduces the accuracy of the method for KET. However, the RSD values ranging from 1.16% to 0.017% and the precision is not

, the analytes are extracted from the serum with an efficiency comparable to

length 270 nm. 1—ketoprofen, 2—naproxen, 3—sodium diclofenac, 4—indomethacin.

*3.3. Extraction Efficiency*

100 ng mL−<sup>1</sup>

affected.

#### *3.4. Optimization of Injected Sample Volume 3.4. Optimization of Injected Sample Volume*

Different injection volumes were tested to achieve better sensitivity using UV detection and to simultaneously determine the protein removing capacity of the extraction cartridge. The proteins were completely removed in less than 1 min after the injection volumes 10, 25, and 50 µL. The cleaning of up of 100 µL centrifuged serum was achieved in 1.5 min. A further increase in the injection volume and simultaneous extension of the washing phase resulted in larger analyte loss. Therefore, 100 µL serum was designated as the maximal injection volume for the 10 × 4.6 mm i.d. extraction cartridge. Different injection volumes were tested to achieve better sensitivity using UV detection and to simultaneously determine the protein removing capacity of the extraction cartridge. The proteins were completely removed in less than 1 min after the injection volumes 10, 25, and 50 µL. The cleaning of up of 100 µL centrifuged serum was achieved in 1.5 min. A further increase in the injection volume and simultaneous extension of the washing phase resulted in larger analyte loss. Therefore, 100 µL serum was designated as the maximal injection volume for the 10 × 4.6 mm i.d. extraction cartridge.

### *3.5. Validation*

*3.5. Validation* The method was validated with respect to the linearity, precision, accuracy, selectivity, and sensitivity following the ICH Q2 R1 guideline to evaluate the reliability of the results [28]. The validation parameters were chosen to primarily confirm micro/nanofibrous PCL applicability for the NSAID extraction since this sorbent is not standardized and commonly used. The ICH Q2 R1 protocol meets these requirements better than the M10 guideline that is usually used for the validation of bioanalytical methods. The chromatographic system suitability test was carried out confirm the suitability of the HPLC instrument for NSAIDs analysis. Mean values and standard deviations of the retention The method was validated with respect to the linearity, precision, accuracy, selectivity, and sensitivity following the ICH Q2 R1 guideline to evaluate the reliability of the results [28]. The validation parameters were chosen to primarily confirm micro/nanofibrous PCL applicability for the NSAID extraction since this sorbent is not standardized and commonly used. The ICH Q2 R1 protocol meets these requirements better than the M10 guideline that is usually used for the validation of bioanalytical methods. The chromatographic system suitability test was carried out confirm the suitability of the HPLC instrument for NSAIDs analysis. Mean values and standard deviations of the retention time, peak capacity, symmetry factor, resolution, and repeatability of the analytical run were calculated from results of six injections of standard solutions and evaluated according to the European Pharmacopoeia recommendations. The results are presented in the Table 1.

### Linearity, Accuracy, Precision, and Limits of Detection and Quantification

All samples were measured in triplicate using the optimized conditions. The calibration curves were established for standard and serum matrix solutions at concentration levels of 50; 100; 500; 1000; 2000; 10,000; and 20,000 ng mL−<sup>1</sup> . The linear relationship between the NSAID quantity and peak area was confirmed for KET, DCF, and IND in the concentration range 50 to 20,000 ng mL−<sup>1</sup> . The calibration curve for NAP was linear for a range 50 to 10,000 ng mL−<sup>1</sup> . The correlation coefficient for each drug was exceeded 0.996. Standard and matrix matched calibration curves including regression equations are demonstrated in Supporting Information section in Figure S4.

The accuracy was determined via the recovery study carried out for all the drugs in the human serum matrix. The recovery was calculated as a comparison of peak areas of NSAID in standard and serum matrix solutions at three concentration levels of 100; 1000 and 10,000 ng mL−<sup>1</sup> measured in triplicate. The results of recovery also determined the extraction efficiency for each drug.



The serum solutions were measured six times at three concentration levels of 100; 1000; and 10,000 ng mL−<sup>1</sup> for intra-day precision determination. The results expressed as RSD (%) were determined in a range of 7.89% to 10.24% for a concentration level of 100 ng mL−<sup>1</sup> , 0.17% to 0.90% for a concentration level of 1000 ng mL−<sup>1</sup> , and 0.09% to 0.46% for a concentration level of 10,000 ng mL−<sup>1</sup> . The inter-day precision was calculated for three consecutive measurements of matrix spiked with 1000 ng mL−<sup>1</sup> NSAID in three days. The results expressed as RSD (%) were in a value range of 3.62 and 7.88%. The lowest concentration of the calibration curve equaled to 50 ng mL−<sup>1</sup> (10σ) was established as a limit of quantification (LOQ). The limit of detection 15 ng mL−<sup>1</sup> (LOD) was calculated from the LOQ value as a tree-folds (3σ) variation. The results of the validation are summarized in the Table 2. These results are comparable to those obtained using other methods described for diclofenac determination, including column-switching [18], protein precipitation [29], and SPE via commercial sorbent [30] applied as sample preparation procedures.

### *3.6. Reusability of the Extraction Cartridge*

The manually packed extraction cartridge was used during the entire experiments corresponding to the analysis of more than 100 serum samples. Neither extraction efficiency nor the back pressure of PCL composite sorbent had changed during these experiments. Thus, our composite micro/nanofibrous sorbent is better suited for the extensive use in high-pressure liquid chromatography systems than nanofibers reported elsewhere [31]. The microfibers represent a stable scaffold resistant to the high pressure while, simultaneously, the large surface area to volume ratio of the nanofibers contributes to the high extraction efficiency. Moreover, the cotton-like texture of micro/nano composite material is easier to manually fill into the cartridge. The durability of our cartridge paralleled our previous study [20] where we applied 200 injections of 200 µL serum in commercial RAM LiChrospher RP-18 ADS column without observing a decrease in extraction efficiency.

### *3.7. Analysis of Real Sample*

The real patient serum samples after continual intravenous infusion containing 75 mg sodium diclofenac were handled as a serum matrix and the content of diclofenac determined using our method. The diclofenac peak was recognized based on its retention time and UV spectrum. The calculated concentration of diclofenac was 32.57 ng mL−<sup>1</sup> in ten times diluted serum corresponding to 325.7 ng mL−<sup>1</sup> in the original patient serum. This result confirmed that the extraction using nanofibrous sorbent is enough sensitive to handle the real samples. The chromatogram of the patient serum is shown in Figure 5.


**Table 2.** HPLC method validation results.

<sup>1</sup> Each concentration of calibration standard was measured in triplicate; <sup>2</sup> Each concentration of matrix calibration standard was measured in triplicate. Calibration curves were linear for (1) ketoprofen, diclofenac sodium, and indomethacin at seven concentration levels, (2) naproxen at six concentration levels; <sup>3</sup> Relative standard deviation (RSD) was calculated from six injections of the matrix solutions spiked with analytes at concentration levels c<sup>1</sup> = 100 ng mL−<sup>1</sup> , c<sup>2</sup> = 1000 ng mL−<sup>1</sup> and c<sup>3</sup> = 10,000 ng mL−<sup>1</sup> ; <sup>4</sup> Relative standard deviation (RSD) was calculated from the average of three injection of matrix solutions spiked with analytes at concentration level c = 1000 ng mL−<sup>1</sup> for three days; <sup>5</sup> Accuracy was determined as a method recovery of matrix solutions at concentration levels c<sup>1</sup> = 100 ng mL−<sup>1</sup> , c<sup>2</sup> = 1000 ng mL−<sup>1</sup> and c<sup>3</sup> = 10 000 ng mL−<sup>1</sup> measured in triplicate.; <sup>6</sup> Limit of detection (LOD) was calculated from the signal-to-noise ratio in a 3-fold (3σ) variation; <sup>7</sup> Limit of quantification (LOQ) was calculated from the signal-to-noise ratio in a 10-fold (10σ) variation. *Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 10 of 12

**Figure 5.** Standard solution spiked with 50 ng mL−1 diclofenac and real patient human serum containing the same drug. The concentration of diclofenac in real sample was equal to 32.57 ng mL−1 in ten times diluted serum. **Figure 5.** Standard solution spiked with 50 ng mL−<sup>1</sup> diclofenac and real patient human serum containing the same drug. The concentration of diclofenac in real sample was equal to 32.57 ng mL−<sup>1</sup> in ten times diluted serum.

#### **4. Conclusions 4. Conclusions**

Our study built on previous extraction experiments with micro/nano PCL that had promising properties enabling the direct extraction of analytes from proteinaceous matrix. We tested the micro/nano PCL fibers as the extraction sorbent for on-line solid-phase extraction of NSAID. After the injection of 100 μL human serum matrix, the sorbent enabled removal of the proteins and the majority of other macromolecular interferences within 1.5 min. The analyte recovery from the matrix was comparable to that obtained with the extraction from solutions of standards. This demonstrated that micro/nano PCL is a promising sorbent for KET, NAP, DCF, and IND even from the proteinaceous matrixes. The applicability of micro/nano PCL sorbent for the therapeutic drug monitoring was ex-Our study built on previous extraction experiments with micro/nano PCL that had promising properties enabling the direct extraction of analytes from proteinaceous matrix. We tested the micro/nano PCL fibers as the extraction sorbent for on-line solid-phase extraction of NSAID. After the injection of 100 µL human serum matrix, the sorbent enabled removal of the proteins and the majority of other macromolecular interferences within 1.5 min. The analyte recovery from the matrix was comparable to that obtained with the extraction from solutions of standards. This demonstrated that micro/nano PCL is a promising sorbent for KET, NAP, DCF, and IND even from the proteinaceous matrixes. The applicability of micro/nano PCL sorbent for the therapeutic drug monitoring was explored

plored using a real sample containing diclofenac. The extraction provided good sensitivity even at low concentrations typical of biological samples and purified them sufficiently.

to improve the bioanalytical methods for sample pretreatment in terms of higher effectiveness at a lower costs compared to commercial RAM. This makes it a valuable tool for the on-line extraction/chromatography methods developed for the therapeutic drug mon-

**Supplementary Materials:** The following are available online at www.mdpi.com/xxx/s1. Figure S1. The combination of meltblown and electrospinning: 1—drum collector, 2—composite micro-nanofiber layer, 3—airstream with fibers, 4—multi needle spinner, 5—needle counter electrode, 6—meltblown die, 7—pumps, 8—extruder, 9—transmission, 10—high voltage sources, 11—engine; Figure S2. The morphology of micro/nanofibrous PCL material before and after using in high pressure chromatography system; Figure S3. The filling process of the extraction cartridge; Figure S4. Cali-

**Author Contributions:** Conceptualization, H.R. and D.Š.; methodology, H.R., L.C.H. and D.Š.; software, H.R.; validation, H.R., L.C.H., D.Š. and F.Š.; formal analysis, J.E. and J.C.; investigation, D.Š. and H.R.; resources, J.E. and J.C.; data curation, H.R.; writing—original draft preparation, H.R. and L.C.H.; writing—review and editing, H.R., F.Š. and D.Š.; visualization, H.R.; supervision, D.Š.; project administration, J.C. and D.Š.; funding acquisition, D.Š. and F.Š. All authors have read and

**Funding:** The authors gratefully acknowledge the financial support of the GAČR project No. 20- 19297S, STARSS project (Reg. No. CZ.02.1.01/0.0/0.0/15\_003/0000465) co-funded by the ERDF, and

itoring enabling fast, safe, and precise analysis.

bration curves—matrix (green), standard (orange).

agreed to the published version of the manuscript.

GAUK project No. 1134119 and SVV 260 548.

using a real sample containing diclofenac. The extraction provided good sensitivity even at low concentrations typical of biological samples and purified them sufficiently. Based on our results, we expect that the nano/micro fibrous PCL sorbent has a potential to improve the bioanalytical methods for sample pretreatment in terms of higher effectiveness at a lower costs compared to commercial RAM. This makes it a valuable tool for the online extraction/chromatography methods developed for the therapeutic drug monitoring enabling fast, safe, and precise analysis.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/nano11102669/s1. Figure S1. The combination of meltblown and electrospinning: 1—drum collector, 2—composite micro-nanofiber layer, 3—airstream with fibers, 4—multi needle spinner, 5—needle counter electrode, 6—meltblown die, 7—pumps, 8—extruder, 9—transmission, 10—high voltage sources, 11—engine; Figure S2. The morphology of micro/nanofibrous PCL material before and after using in high pressure chromatography system; Figure S3. The filling process of the extraction cartridge; Figure S4. Calibration curves—matrix (green), standard (orange).

**Author Contributions:** Conceptualization, H.R. and D.Š.; methodology, H.R., L.C.H. and D.Š.; software, H.R.; validation, H.R., L.C.H., D.Š. and F.Š.; formal analysis, J.E. and J.C.; investigation, D.Š. and H.R.; resources, J.E. and J.C.; data curation, H.R.; writing—original draft preparation, H.R. and L.C.H.; writing—review and editing, H.R., F.Š. and D.Š.; visualization, H.R.; supervision, D.Š.; project administration, J.C. and D.Š.; funding acquisition, D.Š. and F.Š. All authors have read and agreed to the published version of the manuscript.

**Funding:** The authors gratefully acknowledge the financial support of the GACR project ˇ No. 20-19297S, STARSS project (Reg. No. CZ.02.1.01/0.0/0.0/15\_003/0000465) co-funded by the ERDF, and GAUK project No. 1134119 and SVV 260 548.

**Acknowledgments:** The authors gratefully acknowledge the cooperation with Department of Clinical Biochemistry and Diagnostics at the Faculty Hospital in Hradec Králové for providing the patient serum samples.

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

### **Abbreviations**


### **References**


## *Article* **Optimization of Microalga** *Chlorella vulgaris* **Magnetic Harvesting**

**Maria G. Savvidou <sup>1</sup> , Maria Myrto Dardavila 2,\*, Ioulia Georgiopoulou <sup>2</sup> , Vasiliki Louli <sup>2</sup> , Haralambos Stamatis <sup>3</sup> , Dimitris Kekos <sup>1</sup> and Epaminondas Voutsas <sup>2</sup>**


**Abstract:** Harvesting of microalgae is a crucial step in microalgae-based mass production of different high value-added products. In the present work, magnetic harvesting of *Chlorella vulgaris* was investigated using microwave-synthesized naked magnetite (Fe3O<sup>4</sup> ) particles with an average crystallite diameter of 20 nm. Optimization of the most important parameters of the magnetic harvesting process, namely pH, mass ratio (mr) of magnetite particles to biomass (*g/g*), and agitation speed (rpm) of the *C. vulgaris* biomass–Fe3O<sup>4</sup> particles mixture, was performed using the response surface methodology (RSM) statistical tool. Harvesting efficiencies higher than 99% were obtained for pH 3.0 and mixing speed greater or equal to 350 rpm. Recovery of magnetic particles via detachment was shown to be feasible and the recovery particles could be reused at least five times with high harvesting efficiency. Consequently, the described harvesting approach of *C. vulgaris* cells leads to an efficient, simple, and quick process, that does not impair the quality of the harvested biomass.

**Keywords:** *Chlorella vulgaris*; microwave-synthesized magnetite particles; response surface methodology; harvesting process optimization

### **1. Introduction**

Microalgae have been extensively investigated over the past decades as a source for biofuel production due to their high lipid and carbohydrate yields, as well as being a natural source of high value-added bioactive compounds such as polyphenols, carotenoids, fatty acids, and antibiotics [1–3]. Bioactive chemicals derived from natural sources present higher biological activity and acceptance by consumers compared to the synthetic alternatives [4].

Nowadays, microalgae are of significant importance in the fields of human health, cosmetics, food, and animal feed. In comparison to terrestrial crop plants, microalgae can provide higher productivity and photosynthetic performance, and since they can be cultivated on infertile land, they do not compete with existing food production methods [5,6]. Of the several major production steps of microalgae components, harvesting is both energy and time demanding. It is estimated that microalgae biomass harvesting is responsible for 20–30% of the total biomass production cost [7,8]. Moreover, the harvesting step is crucial for the downstream process, since it leads to a slurry of highly concentrated solid matter [9].

Various harvesting techniques have been introduced over the years such as centrifugation, sedimentation, flocculation, filtration, flotation, and their various combinations. These methods provide (variable) sufficient harvesting efficiencies, however, they are relatively energy and time consuming. Furthermore, none of them can be used as an efficient method for all microalgal cell types, and their application depends on the desired end-product

**Citation:** Savvidou, M.G.; Dardavila, M.M.; Georgiopoulou, I.; Louli, V.; Stamatis, H.; Kekos, D.; Voutsas, E. Optimization of Microalga *Chlorella vulgaris* Magnetic Harvesting. *Nanomaterials* **2021**, *11*, 1614. https://doi.org/10.3390/nano11061614

Academic Editor: Jose L. Arias

Received: 22 May 2021 Accepted: 17 June 2021 Published: 20 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

of microalgae cells [10,11]. The reduced lipid levels regarding flocculation, as well as clogging and fouling in filtration, are also considerable disadvantages [12,13]. Moreover, the added cost of the chemicals used in chemical-flocculation or the microorganisms in bio-flocculation, along with the potential contamination of the harvested biomass, are further drawbacks [13–16].

Magnetic separation, which has been extensively used in the food industry, wastewater treatment, and the steel industry, Refs. [3,17,18] has gathered popularity throughout the past decade as a method for harvesting microalgal cells. Magnetic biomass harvesting, which uses micro- or nano-sized magnetic particles and an applied external magnetic field, is a highly efficient process for cell separation; the method is fast and efficient [17,19–22]. The process is based on the interactions between the magnetic particles and the surface of the microalgae cells. The magnetic particles can attach to the cells via electrostatic forces, hydrogen bonds, acid-base interactions or van der Waals forces, forming magnetic particle–cell bonding [23]. Key factors affecting the magnetic separation efficiency include cell type and their growth stage, magnetic particle dosage, pH, ions in the cultivation medium, temperature, and gradient of the magnetic field [3]. Both naked and surface-functionalized (using polyacrylamides, chitosan, poly diallyldimethylammonium chloride (PDDA), aminoclay, polyethylenimine (PEI), 3-aminopropyl triethoxysilane (APTES)) magnetic nanoparticles have been applied for the harvesting of *C. reinhardtii*, *C. vulgaris*, *N. salina*, *N. maritime*, *S. dimorfus*, and *S. ovalternus* microalgae, among others [2,3,24–29]. Nonetheless, successful upscaling of the magnetic harvesting method depends on some important characteristics of the magnetic particles i.e., their biocompatibility, reusability, and ease of use, as well as the optimization of the parameters affecting the harvesting efficiency [30].

*Chlorella sp.* is a microalga rich in vitamins, polysaccharides, minerals, lipids, and other high-value products [31]. Lutein, a compound with proven anti-cataract properties, is also found in *Chlorella sp.*, while its extracts exhibit anti-tumor, anti-inflammatory, antioxidant, and anti-microbial properties, and reduce blood pressure and cholesterol levels [32–34]. As a result, *Chlorella sp.* is one of the most extensively used microalga in cosmetics, pharmaceuticals, food, and animal feed industries [31,35].

Although *Chlorella sp.* holds second place in the global "algal products" market [36], its harvesting process is still a major challenge. That is why the harvesting of freshwater microalgae *Chlorella vulgaris* (using microwave-synthesized Fe3O<sup>4</sup> magnetic particles) was examined in this work. The purpose of this study is to develop an efficient *C. vulgaris* harvesting method by examining and optimizing the most important parameters. Specifically, three harvesting parameters have been investigated, namely pH, mass ratio of magnetite particles to biomass, and agitation speed; these were characterized using response surface methodology (RSM). Reusability of the magnetic particles, as well as integrity of the microalgae cells, were also examined. Scanning electron microscopy (SEM) was used to explore the interactions between the microalgae cells and the magnetic particles, while the adsorption mechanism of the *C. vulgaris* cells on the magnetite particles was deduced by measuring the adsorption isotherm at ambient conditions. Finally, zeta potential measurements were performed in order to examine the separation mechanism of *C. vulgaris* microalgae with the aid of the synthesized iron oxide magnetic particles.

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

### *2.1. Chemical Reagents*

For the synthesis of the magnetite (Fe3O4) particles, iron (II) sulfate (FeSO4·7H2O) salt (Chem-Lab NV, Zedelgem, Belgium) and sodium hydroxide pellets (NaOH) (Panreac Quimica, SA) were employed. The cultivation medium of the *Chlorella vulgaris* strain was composed of sodium nitrate (NaNO3), calcium chloride (CaCl2·2H2O), magnesium sulfate (MgSO4·7H2O), potassium hydrogen phosphate (K2HPO4·3H2O), sodium chloride (NaCl), EDTA disodium (Na2EDTA), iron (III) chloride (FeCl3), manganese (II) chloride (MnCl2) and zinc chloride (ZnCl2) and cobalt chloride (CoCl2) (Sigma–Aldrich, St. Louis, MO, USA). Potassium dihydrogen phosphate (KH2PO4), vitamin B1, and vitamin B12

provided from Merck (KGaA, Darmstadt, Germany). Deionized water was used for the solubilization of the chemical reagents. Magnetite nano powder (50–100 nm) was purchased from Sigma-Aldrich (St. Louis, MO, USA) for comparison with the microwave-synthesized Fe3O<sup>4</sup> particles used in this work. A trypan blue assay was performed for examining cell membrane integrity (Merck, KGaA, Darmstadt, Germany). All the reagents used were of analytical grade.

### *2.2. Microwave Synthesis of Magnetite (Fe3O4) Particles*

Magnetic particles were prepared using a simple, rapid, and low-cost precipitation method that has been previously reported [20,27]. Briefly, 1 g of FeSO4·7H2O was diluted in 100 mL of deionized water at ambient temperature. Gradual addition of 1 M NaOH (with continuous magnetic stirring) was used to adjust the pH of the solution to 12. As a result, a black precipitate of Fe(OH)<sup>2</sup> was formed. By adding deionized water, the final volume of the solution was fixed at 200 mL. The solution was then placed in a common microwave oven where it was radiated for 10 min at 700 W. As a consequence, the Fe(OH)<sup>2</sup> precipitate was oxidized to Fe3O4. Subsequently, the magnetite particles were collected by employing a NdFeB magnet (50.8 mm × 50.8 mm × 25.4 mm, Supermagnete, Gottmadingen, Germany) with 12,600–12,900 G (N40) magnetic induction intensity. Thereafter, the Fe3O<sup>4</sup> particles were washed several times with deionized water in order to remove any residual ions, dried in an oven at 60 ◦C under vacuum, pulverized to a fine powder using a mortar and pestle, and stored in dry conditions for further use.

### *2.3. Microalgal Strain and Cultivation*

*Chlorella vulgaris* strain (UTEX 1809) was purchased from Culture Collection of Algae at the University of Texas (Austin, TX, USA). The strain was grown in a medium with the following composition (per liter): NaNO3, 250 mg; CaCl2·2H2O, 25 mg; MgSO4·7H2O, 75 mg; K2HPO4·3H2O; 75 mg; KH2PO4, 175 mg; NaCl, 25 mg; trace element solution, 6 mL from stock solution; vitamin B1, 0.12 µg; vitamin B12, 0.10 µg. One liter of trace element solution contained Na2EDTA, 750 mg; FeCl3, 97.0 mg; MnCl2, 41.0 mg; ZnCl2, 5.0 mg; CoCl2, 2.0 mg; and Na2MoO4·2H2O, 4.0 mg. The pH was adjusted to 6.8–7.0 and the temperature was controlled at 24 ◦C. Cell cultures of *C. vulgaris* were grown in 1000 mL Erlenmeyer flasks containing a culture volume of 600 mL at 150 rpm. The injected concentration was adapted to an optical density of 0.1 at 600 nm (OD 600nm). Periodic purity assessment was performed by microscopic examination. The photobioreactor was illuminated at 50 µmol photon m−<sup>2</sup> s <sup>−</sup><sup>1</sup> with 24/7 white LED lamps at the cell culture surface. *C. vulgaris* growth was monitored by measuring the OD at 600 nm with a UV-VIS S-22 BOECO (Shimadzu, Kyoto, Japan) for up to 15 days. All experiments were conducted in duplicate.

### *2.4. Characterization of Fe3O<sup>4</sup> Particles and Microalgae*

### 2.4.1. X-ray Diffraction of Magnetite Particles and Magnetic Properties

In order to determine the crystallographic structure and phase composition of the synthesized magnetic particles, X-ray diffraction was employed by using a Brooker X-ray D8 advance diffractometer equipped with a Cu-Kα radiation source. The radiation of the specimen was generated at 40 kV and 40 mA at room temperature, with a scanning rate of 0.1◦ per minute, from 2θ 28◦ up to 89◦ . The phenomenon mean diameter size of the magnetite crystallites was calculated via the Debye–Scherrer formula method [37] by exploiting the obtained XRD raw data using the Diffrac. Suite Eva software. Furthermore, the same XRD study was performed for the high purity Fe3O4 nanoparticles purchased from Sigma-Aldrich, in order to compare the two spectra and deduce on the purity and microstructural characteristics of the microwave-synthesized magnetic particles. The magnetic properties of Fe3O<sup>4</sup> were assessed at room temperature using a vibrating sample magnetometer.

2.4.2. Observation with Scanning Electron Microscopy Coupled with Energy Dispersive X-ray Spectroscopy (SEM-EDS)

The morphological characteristics of treated and untreated *C. vulgaris* algal cells, were studied via scanning electron microscopy (SEM) using an FEI Quanta 200 scanning electron microscope.

Using the same technique, the *C. vulgaris*—Fe3O<sup>4</sup> interaction was also evaluated. For this purpose, magnetically harvested *C. vulgaris* cells were fixed in 2.5% (*v*/*v*) paraformaldehyde at 4 ◦C for 1 h, washed in 0.05 M phosphate buffer (pH 7.4), and consecutively dehydrated in 20%, 35%, 50%, 75%, 90%, and 100% (*v*/*v*) ethanol solution [22].

All the aforementioned samples were sputter-coated with Au prior to SEM examination. Moreover, EDS technique was employed (using an EDAX analyzer) in order to obtain the elemental compositional analysis of the microalgal cells and the magnetite particles.

### 2.4.3. Microalgae Analysis

Algae concentration (g L−<sup>1</sup> ) was calculated using a calibration curve of known optical density at 600 nm and respective dry weights that were determined gravimetrically after drying the algae cells at 60 ◦C. The optical density was measured with a UV-VIS S-22 BOECO spectrophotometer.

### 2.4.4. Zeta Potential Measurements

The zeta potential values of the prepared Fe3O<sup>4</sup> particles and the *C. vulgaris* cells were measured with the Zetasizer Nano-ZS (Malvern, UK) at ambient temperature and calculated according to the Smoluchowski equation using the Zetasizer software. In order to perform these measurements, Milli-Q water was used as the dispersion medium for the magnetite particles (3 mg L−<sup>1</sup> ) and the microalgae cells (50 mg L−<sup>1</sup> ). The pH of the mixtures was fixed to 3.0, 5.0, and 7.0 by adding small quantities of aqueous NaOH and HCl. In each case, at least three zeta potential measurements were recorded, and the reported values correspond to their arithmetic mean.

### *2.5. C. vulgaris Magnetic Harvesting Using Fe3O<sup>4</sup> Particles*

### 2.5.1. Magnetic Harvesting Experimental Procedure

The magnetic harvesting experiments of *C. vulgaris* biomass were carried out under ambient conditions in batch mode. The experimental procedure included the following steps: (a) The pH of the microalgal cultivation was fixed at the desired value through the addition of aqueous NaOH or HCl solutions, (b) in order to facilitate the flocculation between the algal cells and the particles, 10 mL of the microalgae cultivation was mixed with a given amount of magnetic particles for 10 min using a mechanical stirrer operating at desired rpm, and (c) the flocs were separated from the cultivation medium with the application of a strong magnetic field, imposed using the same permanent magnet which was employed during the synthesis of the magnetite particles, for 3 min.

Finally, the harvesting efficiency (*HE*%) of the process was calculated from Equation (1):

$$HE\% = \frac{OD\_0 - OD\_1}{OD\_0} \cdot 100\tag{1}$$

where, OD<sup>0</sup> is the initial absorbance of the microalgae cultivation at a wavelength of 600 nm, and OD<sup>1</sup> is the absorbance at the same wavelength of the supernatant liquid that separates from the microalgae-particles flocs after the application of the magnetic field. The OD<sup>0</sup> and OD<sup>1</sup> values were measured with the UV-VIS S-22 BOECO spectrophotometer.

2.5.2. Effect of Harvesting Process Parameters on Microalgae Magnetic Harvesting Using Fe3O<sup>4</sup> Particles

In order to maximize the harvesting efficiency, the effect of the pH of the mixture, the mass ratio (mr) of magnetite particles (g)/dry biomass (g) in the algal broth, and the agitation speed (rpm) of the *C. vulgaris* cultivation-Fe3O<sup>4</sup> particles mixture, were

examined. Three values of pH (3.0, 5.0, and 7.0), three m<sup>r</sup> ratios (10:1, 12:1, and 14:1) and three agitation speeds (250, 350, and 450 rpm) were compared. Each separate magnetic harvesting experiment was performed three times, and the reported *HE*% results are the mean values.

Preliminary harvesting experiments aiming to evaluate the effect of the duration of flocs' magnetic separation were also performed. According to the results, the strong attachment of Fe3O<sup>4</sup> on the cell membrane, along with the application of a powerful magnetic field, results in the rapid harvesting of the flocculated biomass, therefore the study of this parameter was excluded.

2.5.3. Experimental Design for Optimization of *C. vulgaris* Harvesting Using the Response Surface Methodology

The optimum conditions for magnetic harvesting with Fe3O<sup>4</sup> particles were established using central composite design (CCD) under the application of response surface methodology (RSM). The RSM is a combination of mathematical and statistical techniques applicable when several interactive parameters are examined and is considered useful in order to evaluate the relative significance of the operating factors on magnetic harvesting. In the present work, based on preliminary experiments, the independent parameters affecting magnetic harvesting were found to be pH, mass ratio, and agitation speed. Thus, three-level full factorial design (3<sup>k</sup> ) was used, considering these operational parameters. The experimental design was devised based on the central level (0) between the minimum (−1) and the maximum levels (+1) of the normalized values. Therefore, twenty experiments were carried out using different values of the examined factors (Table 1): pH (3.0–7.0), mass ratio (10:1–14:1), and agitation speed (250–450 rpm). For a complex system, such as the one under consideration, the response cannot be mathematically expressed through a first-order or a second-order polynomial equation. Therefore, the results of the harvesting experiments were fitted to a modified model, and the quality of the fitted model was quantitatively assessed by the ANOVA method to characterize the interaction effect between independent factors and the microalgae-harvesting efficiency.


**Table 1.** Independent variables and the distribution of their level.

Regression analysis and estimation of the coefficients were performed using Design Expert® trial version 12 software.

### 2.5.4. Adsorption Isotherms

The elucidation of the adsorption capacity and mechanism of the microwave-synthesized magnetite particles on the *C. vulgaris* cells is provided by the adsorption isotherms. In the present study, two different models—Langmuir and Freundlich—were employed in order to analyze the experimental data. These models are frequently reported in related literature as sufficiently accurate to describe the adsorption of magnetic particles on microalgae cells [19,21].

According to the Langmuir model, the amount of dry algae cells adsorbed per unit weight of magnetic particles (*Q<sup>e</sup>* [g g−<sup>1</sup> ]), is given by Equation (2):

$$Q\_{\varepsilon} = Q\_m \cdot \frac{\mathbb{C}\_{\varepsilon} \cdot \mathbb{K}\_l}{1 + \mathbb{C}\_{\varepsilon} \cdot \mathbb{K}\_l} \tag{2}$$

which can be transformed into the following linear form:

$$\frac{\mathcal{C}\_{\varepsilon}}{Q\_{\varepsilon}} = \frac{1}{Q\_{m} \cdot \mathcal{K}\_{l}} + \frac{\mathcal{C}\_{\varepsilon}}{Q\_{m}} \tag{3}$$

According to the Freundlich model, *Q<sup>e</sup>* is given by Equation (4):

$$Q\_{\varepsilon} = \mathsf{K}\_{f} \cdot \mathsf{C}\_{\varepsilon}^{\frac{1}{n\_f}} \tag{4}$$

which can be transformed into the following linear form:

$$
\ln \mathcal{Q}\_{\varepsilon} = \ln \mathcal{K}\_f + \frac{1}{n\_f} \cdot \ln \mathcal{C}\_{\varepsilon} \tag{5}
$$

In Equations (2)–(5), *Q<sup>m</sup>* [g g−<sup>1</sup> ] is the maximum adsorption capacity, *C<sup>e</sup>* [g L−<sup>1</sup> ] is the concentration of the microalgae cells in the supernatant after completion of the harvesting process, *K<sup>l</sup>* [L g−<sup>1</sup> ] is the Langmuir adsorption constant, *K<sup>f</sup>* [g g−<sup>1</sup> ] is the Freundlich constant (which is related to the adsorption capacity), and *n<sup>f</sup>* [dimensionless] is the Freundlich factor of the heterogeneity of the adsorption sites.

The adsorption experiments were conducted at ambient temperature in a batch mode in duplicates. For each experiment, an amount of Fe3O<sup>4</sup> was added to 10 mL of algae cultivation of known initial concentration (*C*<sup>0</sup> [g L−<sup>1</sup> ]) in order to realize the magnetic harvesting procedure that is described in Section 2.5.1.

Consequently, the amount of dry algae adsorbed per unit weight of magnetite particles (*Q<sup>e</sup>* [g g−<sup>1</sup> ]) was calculated according to Equation (6).

$$Q\_{\varepsilon} = \frac{\mathcal{C}\_0 - \mathcal{C}\_{\varepsilon}}{m} \cdot V \tag{6}$$

where, *m* [g] is the mass of the magnetic particles used and *V* [L] is the volume of microalgae culture used, i.e., 0.01 L.

### 2.5.5. Reusability of Magnetic Particles

The study on the reusability of the magnetic particles was realized by carrying out tests with the same (used) particles for cycles, in accordance with the procedure outlined by Markeb et al. [38]. The magnetic particles and microalgae cells obtained from magnetic harvesting were mixed with 10 mL of NaOH (0.5 M) and agitated at 200 rpm for 10 min. The suspension was then sonicated for 10 min. Afterwards, 3 mL of methanol and 3 mL of chloroform were added, and the solution was sonicated for another 10 min. Following sonication, the magnetite particles were collected with a permanent magnet, washed three times, and dried overnight. The detached magnetic particles were used to evaluate the harvesting efficiency tests under the same conditions in order to investigate their reusability.

### *2.6. Statistical Analysis*

Tukey's method, based on one factor ANOVA at the 5% confidence level, was used for the statistical analysis, which was performed with SPSS 15.0.1 software (SPSS Inc., Chicago, IL, USA). Statistically significant differences were reported when the probability of the results (*p*) value is less than 0.05 assuming the null hypothesis.

### *2.7. Cell Membrane Integrity of C. vulgaris Cells*

The cell integrity of *C. vulgaris* was determined by the trypan blue staining method. Total of 10 uL of cells were harvested and after addition of 10 uL 1% trypan blue solution the cells were incubated for 10 min at room temperature. The intact cells (viable) remained green (no penetration of the trypan blue solution) while the broken cells appeared blue (stain diffused in the protoplasm).

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

*2.6. Statistical Analysis*

#### *3.1. Characterization of Fe3O<sup>4</sup> Particles 3.1. Characterization of Fe3O<sup>4</sup> Particles*

(stain diffused in the protoplasm).

*2.7. Cell Membrane Integrity of C. vulgaris Cells*

Figure 1 demonstrates the X-ray diffraction spectra of the magnetite particles purchased from Sigma-Aldrich (Figure 1a) and the magnetic particles synthesized in this work (Figure 1b). According to the analysis conducted via XRD software Diffrac. Suite Eva, one predominant phase was identified, that of magnetite (PDF 02-1035 Fe3O<sup>4</sup> Magnetite), for both materials. Figure 1demonstrates the X-ray diffraction spectra of the magnetite particles purchased from Sigma-Aldrich (Figure 1a) and the magnetic particles synthesized in thiswork (Figure 1b). According to the analysis conducted via XRD software Diffrac. Suite Eva, one predominant phase was identified, that of magnetite (PDF 02-1035 Fe3O<sup>4</sup> Magnetite), for both materials.

Tukey's method, based on one factor ANOVA at the 5% confidence level, was used for the statistical analysis, which was performed with SPSS 15.0.1 software (SPSS Inc., Chicago, IL, USA). Statistically significant differences were reported when the probability

The cell integrity of *C. vulgaris* was determined by the trypan blue staining method. Total of 10 uL of cells were harvested and after addition of 10 uL 1% trypan blue solution the cells were incubated for 10 min atroom temperature. The intact cells (viable) remained green (no penetration of the trypan blue solution) while the broken cells appeared blue

*Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 7 of 18

of the results (p) value is less than 0.05 assuming the null hypothesis.

**Figure 1.** X-ray diffraction spectra of (**a**) commercially available magnetite particles and (**b**) the microwave-synthesized magnetite particles. **Figure 1.** X-ray diffraction spectra of (**a**) commercially available magnetite particles and (**b**) the microwave-synthesized magnetite particles.

The noteworthy similarity between the two spectra suggests that they have close microstructural characteristics. Particularly, regarding the microwave-synthesized magnetite (Figure 1b), eight different relatively strong diffraction peaks are displayed in the XRD spectrum at 2 *θ* 30.10°, 35.48°, 42.96°, 53.71°, 57,17°, 62.77°, 70.80°, and 74.27°which correspond to the crystallographic planes (220) (311), (400), (422), (511), (440), (620), and (533) The noteworthy similarity between the two spectra suggests that they have close microstructural characteristics. Particularly, regarding the microwave-synthesized magnetite (Figure 1b), eight different relatively strong diffraction peaks are displayed in the XRD spectrum at 2 *θ* 30.10◦ , 35.48◦ , 42.96◦ , 53.71◦ , 57,17◦ , 62.77◦ , 70.80◦ , and 74.27◦ which correspond to the crystallographic planes (220) (311), (400), (422), (511), (440), (620), and (533) respectively, indicating a cubic inverse spinel structure. Likewise, the same crystallographic planes, can be observed for the commercial magnetite powder (Figure 1a) at 2 *θ* 30.18◦ , 35.55◦ , 43.20◦ , 53.64◦ , 57,20◦ , 62.82◦ , 71.20◦ , and 74.34◦ . One more crystallographic plane is present in the spectrum of commercial magnetite at 2 *θ* 87.36◦ , namely the (642). This last plane cannot be identified with certainty in the XRD pattern of the synthesized magnetite, due to the relatively higher noise encountered. In both cases, it is clearly observed that five diffraction peaks, namely (220), (311), (400), (511), (440) are the main ones. The aforementioned analysis of the Bragg reflections for both magnetite particles was realized by identifying each diffraction peak from the raw XRD data using the Diffrac. Suite Eva software.

Using the tools provided by the same software, the phenomenon mean diameter size of magnetite crystallites was calculated according to the Debye–Scherrer equation [37] for each diffraction peak that was identified. So, an average crystallite diameter size equal to 20 ± 4 nm and 20 ± 3 nm was calculated for the synthesized magnetite and the commercial one respectively.

For the determination of the magnetic properties of the microwave-synthesized magnetite particles, the saturated mass magnetization was determined by vibrating the sample in a magnetometer. The magnetization loop is illustrated in Figure 2. From these results, it is evident that the particles follow a unhysteretic loop, illustrating superparamagnetic behavior at room temperature, thus indicating that their size is below 35 nm [39], which is verified in our case as the particles demonstrate an average diameter of 20 nm. The synthesized magnetic particles (Fe3O4) exhibit saturation magnetization of 60 emu g−<sup>1</sup> .

**Figure 2.** Magnetization curve for the synthesized Fe3O<sup>4</sup> .

Thus, it is concluded that the simple, rapid, and energy-efficient synthesis route followed in this work leads to the formation of high purity nanocrystalline magnetite, which is similar in microstructure to the commercially available that was employed as a means of comparison.

### *3.2. Algae Harvesting Efficiency Optimization*

1

### 3.2.1. Characterization of *C. vulgaris* and Magnetic Particles Interaction Using SEM

Evaluation of the interactions between the microalgae and the (synthesized) magnetic particles was tested by characterizing *C. vulgaris* before (control) and after mixing with the magnetic particles using the scanning electron microscopy (SEM). The morphology of *C. vulgaris* cells alone and with magnetic particles, are illustrated in Figure 3. As it can be seen in Figure 3a, the surface of the untreated cells is smooth and spherical, whereas after the magnetic harvesting at pH 3.0 (Figure 3c), it can be observed that all the cell surfaces are fully covered with magnetic particles. The cell surfaces appear rougher and the patterns that are visible in Figure 3a are not discerned in Figure 3c. This confirms that the amount of magnetic particles we chose to use is enough to fully cover every single cell. Moreover, energy dispersive X-ray spectroscopy (EDX) reports the iron peak in Figure 3d, verifying the presence of magnetite on the cells in contrast to the untreated cells (Figure 3b).

with the magnetic particles using the scanning electron microscopy (SEM). The morphology of *C. vulgaris* cells alone and with magnetic particles, are illustrated in Figure 3. As it can be seen in Figure 3a, the surface of the untreated cells is smooth and spherical, whereas after the magnetic harvesting at pH 3.0 (Figure 3c), it can be observed that all the cell surfaces are fully covered with magnetic particles. The cell surfaces appear rougher and the patterns that are visible in Figure 3a are not discerned in Figure 3c. This confirms that the amount of magnetic particles we chose to use is enough to fully cover every single cell. Moreover, energy dispersive X-ray spectroscopy (EDX) reports the iron peak in Figure 3d, verifying the presence of magnetite on the cells in contrast to the untreated cells (Figure

**Figure 3.** Scanning electron microscopy (SEM) images performed on (**a**) *Chlorella vulgaris* cells before magnetic harvesting, **Figure 3.** Scanning electron microscopy (SEM) images performed on (**a**) *Chlorella vulgaris* cells before magnetic harvesting, (**b**) EDX analysis for the elemental composition of the cells where the absence of iron peak is reported, (**c**) harvested *Chlorella vulgaris* cells at pH 3.0, (**d**) EDX analysis for the elemental composition of the harvested *Chlorella vulgaris* cells where the iron peak (arrow) is shown.

#### (**b**) EDX analysis for the elemental composition of the cells where the absence of iron peak is reported, (**c**) harvested *Chlorella vulgaris* cells at pH 3.0, (**d**) EDX analysis for the elemental composition of the harvested *Chlorella vulgaris* cells where 3.2.2. Magnetic Harvesting of *C. vulgaris*

the iron peak (arrow) is shown. 3.2.2. Magnetic Harvesting of *C. vulgaris* A face-centered (distance of each axial point from the center: alpha = 1) composite design was developed for modelling and optimizing (statistically) the magnetic harvesting of *C. vulgaris*. Twenty experiments were carried out examining the effect of pH, mass A face-centered (distance of each axial point from the center: alpha = 1) composite design was developed for modelling and optimizing (statistically) the magnetic harvesting of *C. vulgaris*. Twenty experiments were carried out examining the effect of pH, mass ratio (mr), and agitation speed on harvesting efficiency. The range of the examined parameters was chosen based on preliminary experiments and corresponding literature. Specifically, studies on *Chlorella vulgaris* and related species [2,12,40,41] have been conducted in a pH range between 2 and 12, while maximum harvesting efficiency was achieved in most cases at acidic or neutral pH. The *Chlorella vulgaris* cell walls are very robust and remain resistant at acidic pH, in comparison to other microalgal species that do not attain such a resilient cell wall and might be affected. Although harsh conditions can affect the viability of the cells, the harvesting process followed in this study is very quick, without significantly affecting the biomass composition. Regarding stirring, low agitation speed values [19,42] of around 250 rpm [43], or even higher values at 800 rpm [44] have been used for harvesting *Chlorella* species. This data led us to choose a pH range from 3 to 7 and an agitation speed range between 250 and 450 rpm.

> The experimental results are presented in Table 2. It is observed that the decrease of pH, and the increase of the other two parameters, improve the harvesting efficiency.

Thus, the best experimental harvesting efficiencies higher than 99% were obtained for pH 3.0 and mixing speed greater or equal to 350 rpm. However, even at a pH similar to that of the cell culture (pH 7.0), the harvesting efficiency is proved to be satisfactory (>85%) for m<sup>r</sup> = 14:1 and an agitation speed of 450 rpm. These results are in agreement with those reported by Zhu et al. [9] and Bharte and Desai [12], for different *Chlorella* species that demonstrated maximum harvesting efficiency in acidic pH, thus verifying our findings. Analogous studies on *C. vulgaris* harvesting have shown that harvesting based on a flocculation method can lead to high harvesting efficiency, but under different optimum conditions compared to this study. For example, Tork et al. [42] reported a 90% efficiency at basic pH (pH = 11.7) values using cationic starch nanoparticles. Similarly, in the study of Leite et al. [40] for *Chlorella Sorokiniana*, a harvesting efficiency greater than 97% (combined with the appropriate velocity gradient and agitation time) was achieved under highly alkaline conditions (pH = 12). Moreover, a lower optimal agitation rate of 150 rpm was reported by Almomani [19] and Razack et al. [45] (using as a flocculant iron oxide nanoparticles and seed powder of clearing nut respectively). However, the above differences concerning optimal harvesting conditions may be due to the different biomass and magnetic materials, absence or presence of surface modification etc.

**Table 2.** Design table formed by RSM—CCD, presenting the experimental conditions and the experimental values of harvesting efficiency.


Analysis of variance (ANOVA) was used to determine the relationship between the dependent (harvesting efficiency %) and independent variables (pH, mass ratio, and agitation). Transformation of the response was considered necessary in order to bound harvest efficiency within reasonable limits. The transformation chosen is presented below:

$$y' = \ln\left(\frac{y - lower}{upper - y}\right) \tag{7}$$

where, *y* stands for the harvesting efficiency (%), lower and upper for the boundaries (40 and 100 respectively), and *y* 0 for the transformed harvesting efficiency.

Table 3 shows the statistical results of harvesting efficiency for RSM using ANOVA, while a linear model with respect to the examined factors and their interactions is obtained as follows:

$$y' = -2.473 - 2.022 \ast A + 0.299 \ast B + 0.027 \ast C - 0.003 \ast A \ast C + 0.229 \ast A^2 \tag{8}$$

where, *y* 0 stands for the transformed harvesting efficiency, *A* for pH, *B* for the mass ratio (gmagn. par./gbiomass), and *C* for the agitation speed (rpm). It is clarified that through Equation (8) the relation between the independent variables and the transformed dependent variable (*y* 0 ) is expressed. The calculated harvest efficiency (*y*) is finally determined by solving Equation (7) with respect to *y*:

$$y = \frac{100 \ast e^{y'} + 40}{1 + e^{y'}} \tag{9}$$

**Table 3.** Analysis of variance results for RSM model regarding harvesting process, where *A* is pH, *B* is mass ratio, and *C* is agitation speed.


The model was evaluated based on the F- and *p*-values. For an F-value > 1 and *p*-value < 0.01, the model under consideration is deemed valid. In the present work, both the high F-value (47.58) and low *p*-value (<0.0001) of the model prove its significance and accuracy. All the factors considered were significant (high F-value and *p*-value < 0.05). Specifically, the most important variant is shown to be pH (higher F-value). The obtained value of R<sup>2</sup> (0.94) implies the high correlation between the dependent and independent variables, while the adjusted R<sup>2</sup> is only slightly lower (0.93) and in reasonable agreement with the predicted R<sup>2</sup> (0.86). Finally, the adequate precision (26.60) is higher than 4 indicating that the model can be used to navigate the design space.

The juxtaposition of the predicted against the experimental values of harvesting efficiency (shown in Figure 4) proves that most of the data points are close to the 45-degree line, proving the reliability of the model.

### *3.3. Separation Mechanism*

In order to study the possible electrostatic interactions between the algal biomass and the microwave-synthesized magnetic particles, their zeta potential (ζ) values were measured within the pH range of the magnetic harvesting experiments (pH = 3 to 7). It was found that both the magnetic particles and the biomass cells attain a negative charge. In particular, measurements show ζ values in the range of −17.4 to −33.2, and of −16.6 to −24.9, for the Fe3O<sup>4</sup> particles and the *C. vulgaris* cells respectively, in the specific dispersing medium (Milli-Q water).

*Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 12 of 18

**Figure 4.** Predicted versus experimental values of harvesting efficiency. **Figure 4.** Predicted versus experimental values of harvesting efficiency.

*3.3. Separation Mechanism* In orderto study the possible electrostatic interactions between the algal biomass and the microwave-synthesized magnetic particles, their zeta potential (ζ) values were measured within the pH range of the magnetic harvesting experiments (pH = 3 to 7). It was found that both the magnetic particles and the biomass cells attain a negative charge. In particular, measurements show ζ values in the range of −17.4 to −33.2, and of −16.6 to As shown in Figure 5, the zeta potential of the magnetite particles increases with decreasing pH, whereas the *C. vulgaris* cell ζ values remain practically the same as the pH decreases from 7 to 5, and then increases with a further decrease of the pH. Negative surface charge of microalgae biomass is expected in acidic solutions due to the presence of negatively charged groups on the surface of the cells. Concerning the bare Fe3O<sup>4</sup> particles, a net negative surface charge has been previously reported [20,46].

−24.9, for the Fe3O<sup>4</sup> particles and the *C. vulgaris* cells respectively, in the specific dispersing medium (Milli-Q water). As shown in Figure 5, the zeta potential of the magnetite particles increases with decreasing pH, whereas the *C. vulgaris* cell ζ values remain practically the same as the pH Since in the present study the magnetic particles and the *C. vulgaris* cells are both negatively charged, it could be assumed that the interactions between them are not dependent on a charge neutralization mechanism. Nevertheless, in any aqueous solution, metal oxide (MeOx) particle surfaces can undergo the following protonation/deprotonation reaction:

$$\text{MeOxO}^{-} + \text{H}\_{2}\text{O} \overset{\text{H}\_{2}\text{O}}{\leftrightarrow} \text{MeOxOH} \overset{\text{H}\_{2}\text{O}}{\leftrightarrow} \text{MeOxOH}\_{2}^{+} \tag{10}$$

negatively charged groups on the surface of the cells. Concerning the bare Fe3O<sup>4</sup> particles, a net negative surface charge has been previously reported [20,46]. Since in the present study the magnetic particles and the *C. vulgaris* cells are both negatively charged, it could be assumed that the interactions between them are not dependent on a charge neutralization mechanism. Nevertheless, in any aqueous solution, metal oxide (MeOx) particle surfaces can undergo the following protonation/deprotonation reaction: MeOx—O- + H2O 2 ↔ MeOx—OH 2 ↔ MeOx—OH<sup>2</sup> <sup>+</sup> (10) Consequently, under acidic conditions and as the pH decreases, an increase of the Consequently, under acidic conditions and as the pH decreases, an increase of the positively charged sites on the surface of the metal oxide particles is expected, resulting in an increase in the zeta potential values [28] without necessarily altering the net charge of the particles. This is in agreement with the ζ values obtained here for the microwavesynthesized Fe3O<sup>4</sup> particles. The maximum harvesting efficiency values measured appear at the lowest applied pH values, i.e., at pH = 3. Hence, this behavior may be attributed to local electrostatic interactions between positively charged sites on the surface of the magnetite particles and the negatively charged surfaces of the *C. vulgaris* cells. This assumption is further confirmed by the fact that the harvesting efficiency increases with decreasing pH.

positively charged sites on the surface of the metal oxide particles is expected, resulting in an increase in the zeta potential values [28] without necessarily altering the net charge of the particles. This is in agreement with the ζ values obtained here for the microwavesynthesized Fe3O<sup>4</sup> particles. The maximum harvesting efficiency values measured appear at the lowest applied pH values, i.e., at pH = 3. Hence, this behavior may be attributed to local electrostatic interactions between positively charged sites on the surface of the magnetite particles and the negatively charged surfaces of the *C. vulgaris* cells. This assumption is further confirmed by the fact that the harvesting efficiency increases with decreas-Nevertheless, electrostatic interactions may not be the primary binding mechanism of the synthesized magnetite nanoparticles with *C. vulgaris* cells. Particularly, Fe3O<sup>4</sup> particles can be attached to the microalgal cells through hydrogen bonding; due to the protonation of the magnetite particles under acidic conditions, hydrogen bond donor chemical species OH<sup>2</sup> + formed on Fe3O<sup>4</sup> can interact with hydrogen bond acceptor groups present on *C. vulgaris* cells, such as amino or carboxy groups [47,48]. Furthermore, due to the nano-size of the Fe3O<sup>4</sup> particles, a large specific surface area and high surface energy is expected, characteristics that imply strong adsorption of the particles to the cells [49,50].

ing pH.

**Figure 5.** Zeta Potential values of microwave-synthesized Fe3O<sup>4</sup> particles and *C. vulgaris* cells at different pH values of the dispersing medium. **Figure 5.** Zeta Potential values of microwave-synthesized Fe3O<sup>4</sup> particles and *C. vulgaris* cells at different pH values of the dispersing medium.

#### Nevertheless, electrostatic interactions may not be the primary binding mechanism *3.4. Adsorption Isotherms*

alized magnetite particles [19,21].

of the synthesized magnetite nanoparticles with *C. vulgaris* cells. Particularly, Fe3O<sup>4</sup> particles can be attached to the microalgal cells through hydrogen bonding; due to the protonation of the magnetite particles under acidic conditions, hydrogen bond donor chemical species OH<sup>2</sup> + formed on Fe3O<sup>4</sup> can interact with hydrogen bond acceptor groups present on *C. vulgaris* cells, such as amino or carboxy groups [47,48]. Furthermore, due to the nano-The adsorption isotherms of the synthesized magnetite particles on the *C. vulgaris* cells are illustrated in Figure 6a,b along with the fitted curves using the Langmuir and Freundlich models respectively. The parameters of both models were obtained using the least squares linear fitting and are presented in Table 4.

size of the Fe3O<sup>4</sup> particles, a large specific surface area and high surface energy is expected, **Table 4.** Parameters estimated using the Langmuir and Freundlich models.


Freundlich models respectively. The parameters of both models were obtained using the least squares linear fitting and are presented in Table 4. The results indicate that the adsorption isotherm of Fe3O<sup>4</sup> particles on the *C. vulgaris* cells is better described by the Langmuir model of adsorption, according to which, a full coverage of the microalgae cells by the magnetic particles occurs. The maximum adsorption capacity, *Qm*, is predicted to be equal to 22.95 g of dry biomass per gram of magnetic particles. Moreover, the high *K<sup>l</sup>* value given by the Langmuir model denotes a strong attraction between the *C. vulgaris* cells and the microwave-synthesized magnetic particles [28]. The adsorption mechanism of Fe3O<sup>4</sup> particles on *C. vulgaris* presented here has been previously reported for the same microalgae species both on naked and surface function-The results indicate that the adsorption isotherm of Fe3O<sup>4</sup> particles on the *C. vulgaris* cells is better described by the Langmuir model of adsorption, according to which, a full coverage of the microalgae cells by the magnetic particles occurs. The maximum adsorption capacity, *Qm*, is predicted to be equal to 22.95 g of dry biomass per gram of magnetic particles. Moreover, the high *K<sup>l</sup>* value given by the Langmuir model denotes a strong attraction between the *C. vulgaris* cells and the microwave-synthesized magnetic particles [28]. The adsorption mechanism of Fe3O<sup>4</sup> particles on *C. vulgaris* presented here has been previously reported for the same microalgae species both on naked and surface functionalized magnetite particles [19,21].

**Figure 6.** Experimental adsorption isotherm of microwave-synthesized magnetite particles on the *C. vulgaris* cells at 25 °C, fitted with (**a**) Langmuir model and (**b**) Freundlich model; pH = 7, rpm = 450, *C. vulgaris* cell concentration 0.22 g L−1 . **Figure 6.** Experimental adsorption isotherm of microwave-synthesized magnetite particles on the *C. vulgaris* cells at 25 ◦C, fitted with (**a**) Langmuir model and (**b**) Freundlich model; pH = 7, rpm = 450, *C. vulgaris* cell concentration 0.22 g L−<sup>1</sup> .

**Table 4.** Parameters estimated using the Langmuir and Freundlich models.

### *3.5. Regeneration and Reusability of Fe3O<sup>4</sup> Particles*

**Langmuir Model Freundlich Model** *Q<sup>m</sup>* [g g−<sup>1</sup> ] *K<sup>l</sup>* [L g−<sup>1</sup> ] R<sup>2</sup> *K<sup>f</sup>* [g g−<sup>1</sup> ] *n<sup>f</sup>* R<sup>2</sup> 22.95 95.10 0.99 41.30 3.81 0.94 Both microalgae harvesting and the downstream processing of harvested biomass (e.g., to obtain biofuels or bio-products) call for recovery and regeneration of the magnetic particles, alongside testing the cell integrity to ensure the intracellular abidance of bioactive compounds and thereby their efficient and easy extraction. Our data indicate that the magnetic particles can be reused for at least five cycles, with less than 20% decrease of the algae harvesting efficiency—even during the 5th cycle, (Figure 7). This is similar to the results reported by Almomani [19]. These findings indicate that the regeneration procedure followed is an acceptable method, since an efficient reusability of the magnetic particles is observed.

*3.5. Regeneration and Reusability of Fe3O<sup>4</sup> Particles*

ticles is observed.

has also been reported [54].

Both microalgae harvesting and the downstream processing of harvested biomass (e.g., to obtain biofuels or bio-products) call for recovery and regeneration of the magnetic particles, alongside testing the cell integrity to ensure the intracellular abidance of bioactive compounds and thereby their efficient and easy extraction. Our data indicate that the magnetic particles can be reused for at least five cycles, with less than 20% decrease of the algae harvesting efficiency—even during the 5th cycle, (Figure 7). This is similar to the results reported by Almomani [19]. These findings indicate that the regeneration procedure followed is an acceptable method, since an efficient reusability of the magnetic par-

Analogous studies demonstrated a reusability efficiency of 80% after three cycles with chloroform: methanol treatments and ultrasonication [51], and at least 85% after five cycles of acid–base treatment combined with ultrasonication [52]. Markeb et al. [38], using NaOH, methanol, and chloroform as organic solvents, and ultrasonication, reported a minimum 80% recovery of the magnetic nanoparticles. Treatment with NaOH at pH = 12 resulted in 95% recovery for polypropylene/iron oxide nanoparticles, with the recovered magnetic particles retained almost the same microalgae biomass harvesting efficiency as per the newly synthesized ones [53]. An 80% reusability efficiency of cationic surfactantdecorated iron oxide nanoparticles after microalgae detachment using SDS and sonication

**Figure 7.** *Chlorella vulgaris* harvesting efficiency (initial concentration of 0.22 g L−1) at five different consecutive cycles. **Figure 7.** *Chlorella vulgaris* harvesting efficiency (initial concentration of 0.22 g L−<sup>1</sup> ) at five different consecutive cycles.

Staining with trypan blue (confirming compromised plasma membranes) verified the integrity of the bounded cells on the magnetic particles (Figure 8) and confirmed that no precious intracellular macromolecules were released in the harvesting medium. The trypan blue staining was performed at the most efficient pH for harvesting efficiency (pH = 3). Thus, the cells maintain their ability to produce high value-added products suitable for further use in various biotechnological applications and/or biofuels production. Analogous studies demonstrated a reusability efficiency of 80% after three cycles with chloroform: methanol treatments and ultrasonication [51], and at least 85% after five cycles of acid–base treatment combined with ultrasonication [52]. Markeb et al. [38], using NaOH, methanol, and chloroform as organic solvents, and ultrasonication, reported a minimum 80% recovery of the magnetic nanoparticles. Treatment with NaOH at pH = 12 resulted in 95% recovery for polypropylene/iron oxide nanoparticles, with the recovered magnetic particles retained almost the same microalgae biomass harvesting efficiency as per the newly synthesized ones [53]. An 80% reusability efficiency of cationic surfactant-decorated iron oxide nanoparticles after microalgae detachment using SDS and sonication has also been reported [54].

Staining with trypan blue (confirming compromised plasma membranes) verified the integrity of the bounded cells on the magnetic particles (Figure 8) and confirmed that no precious intracellular macromolecules were released in the harvesting medium. The trypan blue staining was performed at the most efficient pH for harvesting efficiency (pH = 3). Thus, the cells maintain their ability to produce high value-added products suitable for further use in various biotechnological applications and/or biofuels production. *Nanomaterials* **2021**, *11*, x FOR PEER REVIEW 16 of 18

> cles were characterized by X-ray diffraction, identifying one predominant phase, namely that of magnetite, with a mean diameter size of 20 nm. The simple and rapid production route followed in this work leads to the formation of high purity nanocrystalline magnetite with a superparamagnetic behavior at room temperature. The adsorption isotherm of the Fe3O<sup>4</sup> particles on the *C. vulgaris* cells at ambient temperature, demonstrated a full coverage of the microalgae cells by the magnetic particles in accordance with the Langmuir model. Furthermore, the magnetite particles do not impose loss of microalgae cells' integrity. Response surface methodology verified the experimental optimum operational harvesting conditions, namely pH = 3, mass ratio = 14:1, and agitation speed = 450 rpm, under which a harvesting efficiency equal to 99.6% was achieved. pH was proved as the most crucial variant. The magnetic particles were successfully detached from the microalgal cells and reused for five cycles maintaining at least 80% of their initial harvesting efficiency. The separation mechanism was primarily attributed to the formation of hydrogen bonds between the magnetite particles and the microalgae cells under acidic conditions and to the expression of nano-size effects related to the high surface energy and large specific surface area of the particles. The high magnetic harvesting efficiencies (greater than 99%) obtained at pH 3.0 and mixing speed greater or equal to 350 rpm using microwave-synthesized naked magnetite (Fe3O4) particles, the cell integrity after the harvesting procedure and the ability to reuse the synthesized magnetic particles for at least five cycles of harvesting, contribute to the novelty of the present work and indicate that the proposed

> **Author Contributions:** Conceptualization, E.V.; methodology, M.G.S., M.M.D.; validation, M.G.S., M.M.D., V.L., E.V.; software, I.G.; investigation, M.G.S., M.M.D.; resources, M.G.S., M.M.D., E.V.; data curation, M.G.S., M.M.D., V.L., H.S., D.K., E.V.; writing—original draft preparation, M.G.S., M.M.D.; writing—review and editing, V.L, E.V.; supervision, H.S., E.V.; project administration, V.L.,

> **Funding:** This research has been co-financed by the European Union and Greek national funds through the Operational Program Competitiveness, Entrepreneurship and Innovation, under the call, "Special Actions, Aquaculture-Industrial Materials-Open Innovation in Culture" (project code:

E.V. All authors have read and agreed to the published version of the manuscript.

**Figure 8.** Trypan blue staining (**a**, **b**) on *Chlorella vulgaris* cells bound with magnetic particles (at pH 3.0) indicates the integrity of the cell membrane. Intact cells (viable) remained green, without the penetration of the trypan blue solution, while the broken cells appeared blue as stain diffused in the protoplasm. **Figure 8.** Trypan blue staining (**a**,**b**) on *Chlorella vulgaris* cells bound with magnetic particles (at pH 3.0) indicates the integrity of the cell membrane. Intact cells (viable) remained green, without the penetration of the trypan blue solution, while the broken cells appeared blue as stain diffused in the protoplasm.

process is efficient and very promising.

**Institutional Review Board Statement:** Not applicable

**Informed Consent Statement:** Not applicable

T6YBP-00033).

**4. Conclusions**

### **4. Conclusions**

*Chlorella vulgaris* harvesting was assessed using iron oxide magnetic particles synthesized using a Fe (II) precursor with the aid of microwave irradiation. The magnetic particles were characterized by X-ray diffraction, identifying one predominant phase, namely that of magnetite, with a mean diameter size of 20 nm. The simple and rapid production route followed in this work leads to the formation of high purity nanocrystalline magnetite with a superparamagnetic behavior at room temperature. The adsorption isotherm of the Fe3O<sup>4</sup> particles on the *C. vulgaris* cells at ambient temperature, demonstrated a full coverage of the microalgae cells by the magnetic particles in accordance with the Langmuir model. Furthermore, the magnetite particles do not impose loss of microalgae cells' integrity. Response surface methodology verified the experimental optimum operational harvesting conditions, namely pH = 3, mass ratio = 14:1, and agitation speed = 450 rpm, under which a harvesting efficiency equal to 99.6% was achieved. pH was proved as the most crucial variant. The magnetic particles were successfully detached from the microalgal cells and reused for five cycles maintaining at least 80% of their initial harvesting efficiency. The separation mechanism was primarily attributed to the formation of hydrogen bonds between the magnetite particles and the microalgae cells under acidic conditions and to the expression of nano-size effects related to the high surface energy and large specific surface area of the particles. The high magnetic harvesting efficiencies (greater than 99%) obtained at pH 3.0 and mixing speed greater or equal to 350 rpm using microwave-synthesized naked magnetite (Fe3O4) particles, the cell integrity after the harvesting procedure and the ability to reuse the synthesized magnetic particles for at least five cycles of harvesting, contribute to the novelty of the present work and indicate that the proposed process is efficient and very promising.

**Author Contributions:** Conceptualization, E.V.; methodology, M.G.S., M.M.D.; validation, M.G.S., M.M.D., V.L., E.V.; software, I.G.; investigation, M.G.S., M.M.D.; resources, M.G.S., M.M.D., E.V.; data curation, M.G.S., M.M.D., V.L., H.S., D.K., E.V.; writing—original draft preparation, M.G.S., M.M.D.; writing—review and editing, V.L., E.V.; supervision, H.S., E.V.; project administration, V.L., E.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research has been co-financed by the European Union and Greek national funds through the Operational Program Competitiveness, Entrepreneurship and Innovation, under the call, "Special Actions, Aquaculture-Industrial Materials-Open Innovation in Culture" (project code: T6YBP-00033).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Additional data for this study are not available on public database, the corresponding author can provide them upon request.

**Acknowledgments:** The authors would like to thank Petros Schinas and Nikolaos Panagiotou for the assistance in SEM analysis and XRD, as well as Evangelos Hristoforou for the assistance in VSM analysis.

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

### **References**

