**Development, Optimization and Applications of Thin Film Solid Phase Microextraction (TF-SPME) Devices for Thermal Desorption: A Comprehensive Review**

#### **Ronald V. Emmons 1, Ramin Tajali <sup>1</sup> and Emanuela Gionfriddo 1,2,3,\***


Received: 5 July 2019; Accepted: 30 July 2019; Published: 5 August 2019

**Abstract:** Through the development of solid phase microextraction (SPME) technologies, thin film solid phase microextraction (TF-SPME) has been repeatedly validated as a novel sampling device well suited for various applications. These applications, encompassing a wide range of sampling methods such as onsite, in vivo and routine analysis, benefit greatly from the convenience and sensitivity TF-SPME offers. TF-SPME, having both an increased extraction phase volume and surface area to volume ratio compared to conventional microextraction techniques, allows high extraction rates and enhanced capacity, making it a convenient and ideal sampling tool for ultra-trace level analysis. This review provides a comprehensive discussion on the development of TF-SPME and the applications it has provided thus far. Emphasis is given on its application to thermal desorption, with method development and optimization for this desorption method discussed in detail. Moreover, a detailed outlook on the current progress of TF-SPME development and its future is also discussed with emphasis on its applications to environmental, food and fragrance analysis.

**Keywords:** TF-SPME; microextraction; thermal desorption; environmental analysis; flavor and fragrance; onsite sampling; in vivo analysis; ultra-trace analysis

#### **1. Introduction**

As the need for more sensitive and greener alternatives in analytical chemistry continues to grow [1–5], it is necessary for the further development of robust sample preparation technologies to meet these modern demands. Sample preparation, being the first step in any analytical procedure, is far-reaching since any and all steps in the workflow are consequently affected by the sampling and extraction method used. The sampling and clean-up of compounds before introduction into an instrument plays a critical role in the achieved sensitivity and quantitative capabilities of the method. Consequently, novel extraction techniques must be developed to enhance analytical performance, while meeting the newfound call for greener sample preparation methods. Microextraction, being characterized by a small amount of extraction phase compared to the volume of the sample [6], affords the opportunity of substantially reducing the amount of organic solvent used while still achieving similar or better results compared to more traditional extraction techniques such as solid phase extraction (SPE) [7,8] and liquid–liquid extraction (LLE) [9]. The volume of the extraction phase, being inconsequential to the overall volume of the sample [6], allows rapid non-exhaustive extraction, in some cases non-depletive, that can easily be quantitated using a variety of calibration methods [10]. Moreover, analytes from the sample matrix are extracted in their "free-form" (non-bound

or free-concentration), giving the opportunity for the analysis of bio-available analytes in various matrices. Among existing microextraction techniques, solid phase microextraction (SPME) is the most widely adopted as it allows solvent-less extraction that can be easily automated and adapted for in vivo and onsite applications [11]. The conventional configuration of SPME consists of an extraction phase coated on a solid, fiber-like support composed of fused silica, stainless steel or flexible metal alloys; this geometry allows ease of use and automated extraction and analysis [11]. As a sample preparation method, the use of SPME enables sampling and pre-concentration to be performed in one simple step, making the technique more versatile in its use and able to achieve better throughput compared to more laborious exhaustive methods such as SPE and LLE.

#### **2. From Fiber to Thin Film Format: Pros and Cons**

In common with all non-exhaustive extraction techniques, the mechanism of extraction for SPME is based on the equilibrium-driven diffusion of analytes between the sample matrix and the extraction phase [6]. The amount extracted at equilibrium between these phases is described in Equation (1) and explained in Section 4, "Fundamentals of TF-SPME". This equation implies that for most SPME-based extractions, the only parameters that are consequential when optimizing extraction efficiency for a non-exhaustive extraction are the distribution coefficient between the sample and the extraction phase (Kes), and the volume of the extraction phase (Ve). During method development and optimization, Kes is maximized through changes in different physical parameters of the extraction, such as temperature, agitation, ionic strength, the amount of organic solvent in the sample (if any) and most notably extraction phase chemistry [6]. As a result, during the development of an SPME device the physiochemical properties of the coating must be carefully selected as they affect both the extraction efficiency and specificity for the targeted analyte. Additionally, the extraction phase must also be able to be efficiently desorbed, by either thermal desorption (TD) (the phase then needing to be thermally stable) or by desorption in an organic solvent (the phase not swelling in organics).

Beyond the previously mentioned parameters, an increase in extraction phase volume also contributes to the enhanced efficiency of the sampling device [12]. This phase volume allows improved capacity for the analyte, which in turn enables a more sensitive extraction, applicable to ultra-trace level analysis, doing so, however, poses practical challenges in both engineering of the device and mass transfer phenomena. For example, when optimizing the volume of the extraction phase (Ve) for fiber SPME, a simple increase in the diameter of the coating, as seen in Equation (2), drastically prolongs the equilibration time and negatively affects the desorption efficiency. An increase in phase volume for a fiber SPME also requires a redesign of the whole device assembly, as in the case of the recently introduced Arrow-SPME [13,14]. This device, while important to the overall development of SPME due to its enhanced capacity, will not be further discussed as it does not seek to maximize the device's surface-to-volume ratio as other TF-SPME devices do.

There have been multiple developments throughout the years to increase the extraction phase volume of microextraction devices to enhance their sensitivity with varying success. Still commonly used today, stir bar sorptive extraction (SBSE) seeks to increase the volume of extraction phase by coating a magnetic stir bar with polydimethylsiloxane (PDMS) for the immersive extraction of aqueous analytes, and has proven suitable for both direct immersion and headspace extraction [15]. This geometry allows higher capacities than conventional SPME fiber due to the larger extraction volume, however, it still has yet to overcome the difficulties discussed previously in timely extraction [16] (especially in regards to large sample volumes) [17] and efficient desorption. While the large capacity can grant greater sensitivity, the extraction time can take 24 h [18] and desorption can be long as well depending on the molecular weight and volatility of the analytes [17]. Thin film solid phase microextraction (TF-SPME), a novel SPME device first developed in 2003 [12], overcomes these limitations, extraction efficiency and capacity, by the use of an alternative geometry. TF-SPME consists of a large-volume thin layer of extraction phase (originally pure PDMS) for the pre-concentration of analytes. The development of TF-SPME devices differs from previous attempts as its geometry is simply a flat planar surface, effectively increasing the surface area-to-volume ratio and thus avoiding the usual caveats of increased phase volume [12]. With enhanced capacity and faster equilibration rates compared to other microextractive methods, the practicality of the first iteration of TF-SPME devices was still limited. The geometry of the earlier developments of the technique was cumbersome, its large volume required specially suited large-volume injectors which not all labs were equipped for [12]. On a more fundamental note, an increase of extraction phase volume consequently increases the amount of background and bleed from the extraction phase itself and this critical drawback has only recently been addressed by newer generations of TF-SPME devices [19]. Although the desorption of TF-SPME devices can be fully automated, their geometry still poses a barrier for online extraction and analysis, and currently an auto-sampler that can both perform extraction and desorption for conventional TF-SPME devices has yet to be developed. The suitability of TF-SPME devices, however, for ultra-trace level analysis, along with its convenience of onsite sampling, makes it a suitable and robust alternative SPME application.

#### **3. Types of Desorption Modes for TF-SPME**

There are various different approaches that can be employed toward the desorption of TF-SPME devices, and these techniques are chosen based on the characteristics of the compounds of interest and the composition of the TF-SPME device itself. Aside from TD, the second most common desorption method used for TF-SPME is liquid desorption (LD), commonly used in conjunction with liquid chromatography but also with various separation platforms [20–23]. LD utilizes an organic solvent (or a mixture of water and multiple organic solvents) to re-extract all compounds from the extraction phase before introduction of the now-analyte enriched liquid phase into an analytical instrument. In doing so, this allows the TF-SPME device to still carry out sampling and sample clean up before desorbing into a much smaller amount of organic solvent that would usually be required for LLE. This desorption mode is a necessity for liquid-phase separations; most commonly used for non-volatiles, thermally labile compounds and biomolecules [24]. Previous studies have shown promise for the direct coupling of SPME to nanoelectrospray ionization [25,26], the nanospray solvent effectively desorbing the SPME device. More recently, nanomaterial-based TF-SPME devices have been simultaneously desorbed and analyzed by total reflection X-ray fluorescence spectrometry (TXRF) [27,28] which can be applied to both the analysis of organic compounds and metals.

#### *Desorption by Thermal Desorption Unit (TDU)*

Since the advent of thermally stable extractive phases and binders, the TD of sorbents has been an attractive method for sample introduction to analytical instrumentation, as it requires no additional organic solvent as many other methods do [7–9,21]. As the science and engineering behind these thermally stable phases progress, the inherent background of newly developed phases (solid or liquid) is reduced, thus allowing the TD of appropriate sorbents to be applicable to ultra-trace level analysis [19,29,30]. Fundamentally, TD operates by heating a sorbent with hot gas to release all volatile analytes adsorbed onto the extraction phase, the increase in temperature driving the partition coefficient of the analytes to favor the gas phase thus releasing them from the sorbent. To its advantage, TD enables the introduction of all analytes to a gas chromatograph (GC), with the exception of any non-volatiles that either remain on the TF-SPME device or are desorbed and deposited in the liner/column (as can be the circumstance with liquid injection). This characteristic of TD prolongs the life of the analytical column, reducing the amount of maintenance needed due to the reduction of particulate matter introduced to the column. The ease-of-use and efficiency of TD, along with the inherent greenness of the method (when applicable to solid sorbents), allows TD to be the desorption method of choice for most volatile and semi-volatile compounds.

The efficiency of TD coupled with the convenient geometry of fiber SPME has allowed it to be easily adapted for online analysis. Taking advantage of quick extraction times and a lack of organic solvent, the TD of fiber SPME has long been used for routine analysis. As in other SPME geometries, TD is the most efficient and green desorption method for the analysis of volatiles by TF-SPME, directly desorbing into the instrument and removing the need for an organic solvent. However, as TF-SPME devices boast a larger extraction phase volume compared to the conventional fiber SPME geometry, an adapter is needed for the GC inlet to accommodate its larger size. In the development of TD adapters for larger extraction phase volumes, Wilcockson and colleagues [31] first made a custom thermal desorption unit (TDU) that employed an external heating element to accommodate a 22 mm glass disk coated with extraction phase. Only a couple of years later, Bruheim et al. [12] used a glass insert to introduce a sheet of monophasic PDMS to a commercial programmed temperature vaporizer (PTV) injector. Since then, as there is an ever-growing need for large-volume PTV injectors for various extraction techniques, the manufacturing of automated TDUs suitable for TF-SPME have become commonplace with units from companies such as GERSTEL, Inc. (Figure 1). The geometry of the TDU itself plays a large role in the development of TF-SPME technology, as the volume and overall shape of the TF-SPME device must be developed so the intended desorption unit can accommodate the device and effectively desorb it.

**Figure 1.** Two-stage thermal desorption unit for thin film solid phase microextraction (TF-SPME) (photo courtesy of GERSTEL Inc.). For the purpose of clarity, the CIS (cooled injection system) is referred to as a cryo-trap in the text. TDU, thermal desorption unit; GC, gas chromatography.

The thermal desorption of TF-SPME devices is typically carried out by a two-stage device, as opposed to the simpler one-stage TD often used for small-volume phases, such as is the case of traditional SPME fiber devices. In two-stage TD, the first stage is responsible for the release of analytes from the extraction phase into the desorption unit. This first stage is in many ways equivalent to a one-stage TDU, however, there are more parameters to carefully optimize compared to traditional one-stage desorption (an example being a typical GC inlet). Opposed to only transferring the analyte directly into the column, with a two-stages desorption the analyte must be efficiently desorbed and transferred into the cryo-trap first, and then passed into the column. This requires optimization of the different split settings and temperature programming for each stage. The first stage, the TDU, is most often run at a constant temperature to desorb all compounds enriched on the TF-SPME device, however, a temperature ramp program is possible if desired. This temperature must then be optimized depending on the volatility of the analyte (typically recommended to be 50 ◦C below a compound boiling point) and the thermal stability of the extraction phase, the latter usually taking priority as different extraction phases require different thermal thresholds to desorb efficiently. In most instances, the lowest operating temperature that can be used for the desorption of enriched analytes with minimal carryover is optimal, as this approach prolongs the health of the TF-SPME device and minimizes the amount of bleed from the extraction phase. As an example, the operating temperatures for the two commercially available TF-SPME devices, Carboxen®/PDMS (Car/PDMS) and divinylbenzene/PDMS (DVB/PDMS), go up to 250 ◦C. This parameter, like many others, needs to be optimized by trial and error at the beginning of the analytical procedure. Additionally, the split ratio must be optimized like any other injector, however, as the TDU does not directly transfer analyte into the column but the

cryo-trap, the split-mode of the TDU does not necessarily reflect the overall behavior of the injector. During desorption by way of the TDU, the analyte is transferred through a heated capillary to a cryo-trap. After desorption, the TDU is cooled and the TF-SPME device removed to ensure that no residual analyte is erroneously transferred from the TDU to the cryo-trap during the second stage. Furthermore, the heated transfer capillary connecting the TDU and cryo-trap should always be at a greater temperature than the highest temperature of the TDU to ensure complete transfer of all analyte into the cryo-trap, preferably 20–30 ◦C higher.

Contrary to one-stage desorption, in two-stage TD, after the first stage desorbs all volatile and semi-volatile compounds from the extraction phase, the second stage is used to pre-focus analytes before their introduction into the analytical column. This pre-focusing step, while a boon to any TD method, is especially crucial when desorbing large-volume phases due to the amount of analyte desorbed and the longer desorption times typically required. During the desorption of such large-volumes, as is the case of TF-SPME, the loss of highly-volatile compounds and ultra-volatiles during a single-stage desorption method would be unavoidable. The pre-focusing step is usually achieved by the inclusion of either a sorbent tube or temperature-controlled unit (here referred to as a cryo-trap). While a sorbent-based second-stage can exhibit suitable trapping capacities of analyte, the addition of a second sorbent can complicate method optimization and thus cryo-trap stages are generally preferred. Depending on the cryo-trap device used, the trap can be cooled by liquid gas (such as nitrogen and carbon dioxide) or a solid-state device. In the use of a cryo-trapping device, while the analytes are desorbed at high temperature in the TDU, the cryo-trap is held at a low temperature, −40 to 0 ◦C being appropriate for most volatiles [16,32,33] with highly volatile substances requiring lower temperatures trending toward −150 ◦C to properly pre-focus [8,19,29,34–36]. Only after all compounds are thermally desorbed from the TF-SPME device and condensed in the cryo-trap, the temperature of the cryo-trap is increased at a high rate (usually 12 ◦C/s), achieving discrimination-free transfer of analytes into the analytical column with minimal sample loss. Proper method development and TF-SPME workflow (Figure 2), with the use of a cryo-trap for the focusing of volatiles, results in sharper chromatographic peaks and ensures that all compounds extracted are introduced into the analytical column, increasing the reproducibility and sensitivity of the analytical method. In case of direct immersion extraction, it is advisable to gently wipe the TF-SPME device before introduction into the desorption liner, as to avoid potential matrix contaminants such as water and oil entering the TDU.

**Figure 2.** Optimized TF-SPME workflow (photos courtesy of GERSTEL Inc.) for extraction and thermal desorption.

Care must be taken when deciding the type of split for the cryo-trap as it is synergistic with the TDU (Table 1). It should be mentioned that the gas flow of a two-stage unit is restricted by the cryo-trap during both stages. As a result, if splitless injection is desired a solvent-venting cryo-trap is still necessary to maximize gas flow, thus allowing efficient desorption. This solvent-vent must then be closed when the cryo-trap begins to increase in temperature, ensuring a splitless injection with maximum desorption efficiency. In short, for maximum sensitivity both stages should be run in splitless mode, the cryo-trap being operated under solvent-vent conditions before desorption into the column. In the presence of solvent, the cryo-trap can typically run in a mode that vents the solvent if the boiling point differs greatly from the analyte (approximately a 150 ◦C difference). This allows better chromatography with no loss of analyte, provided the boiling point temperature difference is sufficient and the cryo-trap is heated after the solvent vent closes, effectively making this method a splitless injection onto the column. For unknown analysis, it is typical to run both stages in split mode, and with concentrations in the ppm level, the TDU must have an appropriate split ratio with the cryo-trap being performed in either a splitless or solvent venting fashion.

**Table 1.** Common parameters of two-stage thermal desorption and their applications.


<sup>1</sup> Solvent-venting ends at the same time the cryo-trap begins to heat, allowing maximum desorption flow while still achieving a splitless injection.

It should be noted that alternative forms of TD for TF-SPME devices have been recently reported. These methods use "transmission mode" devices consisting of a coated mesh-like surface to enrich analytes and subsequently both desorb and ionize in one step using ambient ionization methods [37–39]. Direct analysis in real time (DART), an ambient ion source that has been shown to be well suited for onsite analysis [40], can be simply coupled to TF-SPME (coated mesh-like geometry) by positioning the device in-between the ion source and the mass spectrometer (MS). In this way, a stream of heated plasma (most often helium) desorbs the TF-S4PME transmission device while also ionizing analytes at the same time [37,38]. Another ambient ionization source, dielectric barrier discharge ionization (DBDI), has also been proven to be well suited for the analysis of illicit drugs using TF-SPME devices [39]. However, as the DBDI source does not thermally desorb compounds as the DART source can, a separate TD chamber was constructed using an aluminum body and temperature controller. This chamber was then filled with pre-humidified nitrogen to facilitate desorption, the flow of this now-analyte enriched gas entering the DBDI source [39].

#### **4. Fundamentals of TF-SPME**

SPME is one of the most attractive extraction techniques commonly used today due to its high throughput, low-cost and solvent-less extraction. In recent years, TF-SPME has been developed to better meet the demands of onsite and ultra-trace level analysis, as its high surface area-to-volume ratio enables it to achieve enhanced sensitivity and greater extraction efficiency [12]. In common with other extraction technologies, sampling using TF-SPME devices is characterized by partition equilibria between the free form of the analyte in a sample and the extraction phase constituting the extraction device. Consequently, the properties of the extraction phase play a significant role in the efficiency of the extraction process. The total amount of extracted analyte in direct immersion SPME, where the analyte equilibrates between only two phases, is described by Equation (1) [12]:

$$m^{eq} = \frac{K\_{\rm cs} V\_{\rm c} V\_{\rm s}}{K\_{\rm cs} V\_{\rm c} + V\_{\rm s}} \mathcal{C}\_{\rm s} \; , \tag{1}$$

where *neq* is the total amount of extracted analyte at equilibrium, *Kes* is the distribution constant of the analyte between the matrix and extraction phase, *Ve* is the volume of extraction phase, *Vs* is the volume of the sample and *Cs* is the initial concentration of analyte in the sample. As can be seen in Equation (1), the greater the volume of the extraction phase (*Ve*), the larger amount of *neq* is extracted.

In the following equation describing the equilibrium time for the analyte between the two phases (Equation (2)) [12],

$$t\_c = t\_{\mathfrak{G}\%} = \frac{3\delta(b-a)}{D\_s},\tag{2}$$

*te* is the required time for the analyte to reach equilibrium with the extraction phase, *t*95% is the time needed to extract 95% of the equilibrium amount of an analyte on the device, δ is the thickness of the boundary layer, (*b* − *a*) represents the thickness of the extraction phase and *Ds* is the diffusion constant of the analytes into the sample matrix. While an increase in *Ve* allows higher extraction efficiency and thus greater sensitivity according to Equation (1), Equation (2) demonstrates that the corresponding increase in coating thickness results in a longer equilibrium time. Hence, it is important to optimize these parameters to ensure the most efficient and practical mode of extraction [41].

Considering Equation (3) [12],

$$\frac{d\mathbf{n}}{dt} = \left(\frac{D\_s A}{\delta\_s}\right) \mathbf{C}\_{s\prime} \tag{3}$$

A thin film geometry can also enhance the sampling rate due to its high surface area-to-volume ratio, reducing the time it takes to reach equilibrium and enhancing the capacity of the extraction device. In this equation, n is the amount analyte extracted over the extraction time t, A is the area of the extraction phase, *Ds* is the diffusion constant of the analyte into the sample matrix and δ*<sup>s</sup>* is the thickness of the boundary layer [42]. In other words, employing TF-SPME devices allows rapid sampling with high extraction capacity, suitable for ultra-trace level analysis.

#### **5. Development of the First TF-SPME Device and Improvements up to 2019**

As discussed previously, the simultaneous increase of extraction phase volume and surface area for TF-SPME devices allows enhanced sensitivity with as good or better extraction rates compared to traditional fiber SPME. The development of the underlying theory of this phenomena took root in 2000 when Semenov and colleagues described the kinetics of a thin layer of extraction phase, predicting what would be the driving force for both TF-SPME and passive sampler development [43]. While not a traditional TF-SPME device, the first technique to exploit a higher surface area-to-volume ratio to increase both extraction capacity and efficiency was developed in 2001 by Wilcockson and colleagues [31]. The procedure utilized a thin film (0.05 and 0.33 μm) of ethylene-vinyl acetate as the extraction phase which was coated onto 22 mm diameter glass disks serving as the support. This passive sampling method, exhibiting a surface area to volume ratio over 1000 times higher than comparable 100 μm SPME fibers, demonstrated faster equilibration times but with the caveat of having less phase volume. Due to the decreased phase volume, sensitivity did not exceed traditional SPME fibers, however, the method has still been adopted throughout the years with success as environmental passive samplers [44–48]. While important to the development of TF-SPME, this technique does not share the same geometry or sensitivity as traditional TF-SPME devices and thus is only mentioned due to its importance in the development of parallel sampling technologies based on thin adsorbent layers.

First developed in 2003 by Bruheim and colleagues [12], the first TF-SPME device consisted of a pre-manufactured 25.4 μm sheet of polydimethylsiloxane (PDMS) as the extraction phase. This thin sheet of PDMS was attached to a stainless steel rod as support, being affixed in a "flag-like" manner during extraction and wrapped around the rod prior to manual TD in a PTV GC inlet (Figure 3). Using polycyclic aromatic hydrocarbons (PAHs) as model analytes, the study demonstrated the practical use of TF-SPME as an alternative geometry to fiber SPME in both direct immersion extraction and headspace extraction. As the extraction efficiency was up to 20 times higher when using a 1 cm × 1 cm sheet of PDMS compared to a 100 μm PDMS fiber, with the extraction rate exceeding the already-developed SBSE [15], TF-SPME devices were further developed and optimized for better sensitivity and a more convenient sampling approach.

**Figure 3.** A timeline of pivotalmomentsin the evolution of TF-SPME devices. PDMS, polydimethylsiloxane; DVB, divinylbenzene; Car, Carboxen®; HLB, hydrophilic-lipophilic balance.

Being limited to PDMS as an extraction phase, due to the availability of pre-made sorbents, priority was given in the geometric optimization of TF-SPME devices. In 2006, Bragg and colleagues modified a PDMS sheet into a 127 μm thick house-like shape (Figure 3) supported by a stainless steel wire [49]. This TF-SPME device, dimensions of 2 cm × 2 cm with the triangular portion of the device being 1 cm in height, boasted an increased phase volume of 0.0635 cm<sup>3</sup> compared to the 0.00255 cm3 phase volume achieved by the previous developed TF-SPME device [12]. An increase of over 20 times the volume of the extraction phase, the achievable amount of extracted analyte was greatly increased, and thus greater sensitivity was attained. Moreover, the house-like geometry of the TF-SPME device permitted an increase in surface area while still allowing the device to be easily wrapped around the support, ensuring ease of injection into the GC inlet. Additionally, the same study demonstrated the efficacy of using TF-SPME for field analysis of aqueous media, establishing TF-SPME as a convenient onsite extraction tool [12].

A distinct departure from previous developments in both chemistry and design, Rodil and colleagues [32] demonstrated the use of glass wool fabric as solid support for TF-SPME. Opposed to previous attempts, which utilized pre-manufactured sheets of PDMS supported by steel wire, the SPME devices designed consisted of a polyacrylate (PA) extraction phase bound to glass wool. During the development of these devices, an amount of glass wool fabric was saturated with a solution of PA, being cured while sandwiched between two sheets of polyethylene foil to ensure homogeneity of the applied extraction phase. Being the first composite TF-SPME device developed, the sampling device demonstrated increased mechanical stability and the final device was 6 cm × 0.3 cm after the sheets of foil were removed. These TF-SPME devices, named "PA strips", were compared to SBSE by performing their extraction with the same parameters and desorbing them into a TDU, very similar to the modern-day TD of TF-SPME devices. It was found, however, that these devices undergo thermal decomposition after multiple uses, resulting in high amounts of bleed into the GC and a lack of robustness of the device.

Incorporating the house-like geometry [49] and the glass fiber support [32] of earlier TF-SPME devices, Riazi et al. [29] developed the first mixed-mode extraction phase for TF-SPME using Car/PDMS and PDMS/divinylbenzene (PDMS/DVB). Adsorptive particles (Carboxen® or DVB) were suspended in a solution of the binding agent (PDMS) before being applied to a thin sheet of glass wool fabric. Instead of allowing the polymer solution to absorb into the supporting material as previous methods did [32], the coating procedure was performed using the spin coating method due to its ease of use and greater control over the phase thickness. After curing and cutting the material to a 2 cm × 2 cm

square with a 1 cm triangle (same geometry developed by Bragg et al. [49]), the TF-SPME device was held by a cotter pin during extraction of analytes. This newly developed TF-SPME device was then desorbed into a large volume inlet (TDU-2, Gerstel GmbH, Mulheim, Germany) with a CIS-4 (Gerstel GmbH, Mulheim, Germany) cryo-trap. Accordingly, results showed a marked increase in the mechanical stability of the device during extraction, as well as improved thermal stability during the desorption process.

Building on the prior success of mixed-mode TF-SPME devices, the first self-supported particle loaded TF-SPME device was developed using DVB particles loaded onto PDMS [36]. This device was made with the intent of air-sampling using high amounts of DVB embedded into a PDMS base. It was found that when increasing the amount of DVB particles onto the PDMS membrane, up to 30% (w:w) DVB allows the membrane to achieve better mechanical stability compared to non-particle load membranes, 20% being found to be optimal for mechanical stability and extraction efficiency. As a result, the TF-SPME device boasted better sensitivity for the target analyte, benzene, than previous methods using fiber SPME or monophasic PDMS TF-SPME devices. This device, however, was still unable to appropriately be used in direct immersion extraction at high agitation rates, as the mechanical stability was not as great as previous glass-coated TF-SPME devices [29].

In 2016, Grandy and colleagues developed a sampler exhibiting far less siloxane bleed and greater robustness due to the carbon mesh support used therefore creating the first TF-SPME device well suited for untargeted analysis and onsite sampling by direct immersion extraction [19]. To achieve this, higher density PDMS was used to reduce bleed along with a slight reduction in phase volume. A mixture of two components, DVB particles embedded into PDMS, was then spread out onto a carbon mesh which primarily provided support along with some affinitive properties. In spite of the reduced phase volume of this device, the new design still afforded a highly sensitive extraction with now far lower siloxane bleed. Moreover, the carbon mesh support granted much better mechanical stability compared to previous iterations of TF-SPME, allowing the device to undergo more rigorous agitation compared to previously developed TF-SPME devices. After curing the extraction phase, the devices were then cut to different sizes well suited for TD, allowing more practical device introduction compared to previous methods. As a result of these recent developments, this version of TF-SPME is now the first commercially available TF-SPME device, currently distributed by GERSTEL Inc. (Gerstel GmbH, Mulheim, Germany). Since then, much of the development of TF-SPME devices has followed the same convenient format. Different extraction phases have been tested but not commercialized yet, including hydrophilic-lipophilic balance (HLB) phases that offer a wider range of extraction [30].

With the development of new mixed-mode extractive membranes that were both thermally and mechanically stable, new TF-SPME devices were able to overcome the limitations of previous variants of TF-SPME technology [12,16,32,49]. Initial devices, being made of pure monophasic extraction phase, exhibited poor structural rigidity making their practical application cumbersome in immersive extraction [12,49]. As a consequence of the large volumes of PDMS utilized in these devices, significant siloxane bleed was found, resulting in unacceptably high backgrounds. This was at times circumvented by use of single ion monitoring (SIM), avoiding detector saturation but nonetheless a technique not well suited for untargeted analysis. With the development of glass fiber supported TF-SPME devices [29,32], better structural rigidity was met, with the first device [32] still lacking suitable mechanical stability and thermal stability for repeated analysis. The second glass fiber supported device [29], however, not only achieved better mechanical and thermal stability compared to its earlier counterparts but also demonstrated a wider range of extraction than found previously by the use of mixed-mode extraction phases. While there is certainly a benefit of using glass fiber as a structural base for TF-SPME devices, the large amount of extraction phase and binder used is still cause of concern in terms of high siloxane backgrounds.

Further developments of TF-SPME devices have been focused on self-supported membranes with the incorporation of extractive particles, allowing them to achieve efficient extraction of a wide range of analytes. The incorporation of these extractive particles provided newfound mechanical stability [29], permitting the devices to be well suited for direct immersion extraction. Given that the qualities indispensable to the sampling method and characteristics important for the ruggedness of any GC-amenable extraction device—thermal stability, mechanical stability and extraction efficiency—were finally met, a refinement of TF-SPME as a whole was essential and further development was needed. Although it is true that the developed phases are thermally stable, a large volume of extraction phase will still cause more background and bleed compared to a smaller volume. In light of this, the newest iteration of TF-SPME devices [19], applying the extraction phase to a carbon mesh, solves this issue by retaining comparable sensitivity but with drastically reduced siloxane bleed, all the while being more mechanically stable than previous TF-SPME devices. As much as the geometry is far more convenient than previous iterations, the planar surface still poses a practical obstacle in the engineering of automated sampling and desorption. The major caveat in the new design is that currently only two phases currently are offered (DVB/PDMS and Car/PDMS).

#### **6. TF-SPME Coating Methods**

The various extraction phases and coating methods used in the development of TF-SPME devices (Figure 4) are chosen based on the chemical properties of the analyte, the surrounding matrix and the adopted desorption method. There are many techniques for the coating of TF-SPME devices, including dip coating [39,50], spin coating [51] bar coating [34,36,52], electrospinning [53] and spray coating [54]. Among these different coating methods, bar and spin coating are the most common coating methods for the production of TF-SPME devices amendable for TD. The most common coating method discussed in this review is bar coating in which the liquid extraction phase is set on a substrate and then this extraction phase is spread by a bar to develop the device [24]. The first TF-SPME device made by bar coating was prepared using DVB particles impregnated on PDMS and this device was later used for air sampling with good results [36]. Moving forward, Grandy et al. prepared bar coated DVB/PDMS onto a carbon-based mesh support coupled with a portable GC-MS for the quantitation of volatile and semi-volatile organic compounds [19]. The bar coating procedure is needed to be repeated on both sides of the carbon mesh support to ensure an even coating for the final device [9]. In 2017, Piri-Moghadam et al. used bar coating for the preparation of DVB/PDMS TF-SPME devices for the analysis of common pesticides in surface water samples. They found that the use of TF-SPME, in comparison to LLE, provided enhanced selectivity, reproducibility and faster rates of extraction [9]. In another study by Piri-Moghadam and colleagues in 2018, bar coating was utilized in the development of novel TF-SPME devices in an effort to analyze various pesticides found in river water, demonstrating a greener and more sensitive alternative than LLE [55]. The use of these developed DVB/PDMS TF-SPME devices for the quantification of pesticides demonstrated enhanced sensitivity and extraction efficiency for both onsite and bench-top analysis. In the spin coating method, in similar fashion to bar coating, a layer of extraction phase is placed on the substrate, after which by spinning the substrate, a thin layer of extraction phase is homogenously produced [24]. In both bar coating and spin coating, the thickness of the extraction phase can easily be controlled by the pressure of the bar onto the substrate and the intensity of spinning, respectively. Spray coating, one of the simplest methods for the preparation of TF-SPME devices, utilizes a dissolved mixture of extractive phase in a suitable solvent to be sprayed onto a stage until the formation of a uniform film [24]. An example of this technique, Mirnaghi et al. prepared polyacrylonitrile-polystyrene (PAN-PS)-DVB and polyacrylonitrile–phenylboronic acid (PAN-PBA) TF-SPME devices for the analysis of a variety of pharmaceuticals from human plasma [56]. Through the development of these two new TF-SPME devices for consequent analysis by LC-MS/MS, the analysis of a wide spectrum of polar compounds in human plasma with high efficiency and rapid throughput was achieved. Finally, another coating method which is often used for TD is electrospinning or electrospray coating. In this method, a mixture containing a polymer is sprayed by electrical energy on the surface of substrate [24]. In 2015, TF-SPME devices were prepared by the electrospinning method using polyimide nanofibers for the investigation and quantification of phenol compounds in environmental water using GC-MS [53]. Extraction devices were first activated by acetone, increasing

hydrophilicity, resulting in greatly enhanced extraction efficiency with results demonstrating limits of quantification (LOQs) in the ppt level.

**Figure 4.** The different coating methods for preparation of TF-SPME devices.

#### **7. Applications**

#### *7.1. Environmental Analysis*

The applications of TF-SPME have been traditionally environmental in nature, as much of their development has been expedited by the need for rugged and high capacity samplers, both passive and active. Initial progress toward highly efficient samplers with greater surface-to-volume ratios were developed to meet the need for the trace level extraction of contaminants found in complex environmental matrices, but these matrices proved to be challenging due to a variety of parameters such as large volumes for air analysis or the incredible sensitivity needed for persistent contaminants in aqueous media. Under these circumstances, the enhanced sensitivity of TF-SPME coupled with its convenient geometry for both extraction and introduction to onsite and benchtop instrumentation affords it the opportunity to outperform other SPME technologies in trace level environmental analysis.

Initial developments of membrane-based TF-SPME were evaluated by the extraction of polycyclic aromatic hydrocarbons (PAHs) from aqueous samples [12]. PAHs, a class of hydrophobic semivolatiles that are commonly released into the environment through the combustion of hydrocarbons and other organic matter, continue to rise in their environmental significance as they are readily distributed throughout biosystems and are a known cause of cancer among other mutagenic effects [57]. Since then, many studies have analyzed PAHs from aqueous samples using TF-SPME in various different modes, utilizing PDMS film [16,49,58,59], TF-SPME membrane [20] and an alternative thin film-based sampler using polymer-coated aluminum [45]. In addition, glass fiber reinforced TF-SPME has also been used for the analysis of PAHs in aqueous samples, using polyacrylate coatings to extract PAHs along with

organochlorous and organophosphorus pesticides, demonstrating greatly enhanced efficiency with the partition coefficients for the TF-SPME device being up to 15 times higher than the SBSE device [32]. Less conventional extraction phases have also been introduced, as the recent success of carbonaceous nanomaterials being suitable extraction phases for SPME have led the path toward the development of mixed-mode carbonaceous TF-SPME devices. These carbonaceous TF-SPME devices have in turn been proven to be effective in the extraction of different organic compound classes, an example again being the analysis of PAHs in aqueous samples [20]. Furthermore, the analysis of PAHs have also been performed using other "parallel" thin film technologies, that is, extraction which utilizes a large surface area but does not exemplify the large phase volume needed for enhanced sensitivity. This is accomplished using PDMS-coated vials for the determination of PAHs in soil [46]. Similar alternative geometries have also been tested for their uses as passive samplers, such as ethylene-vinyl acetate (EVA) coated glass fibers for the aqueous extraction of pesticides [48] and EVA coated glass cylinders for the air sampling of volatile PCBs [44].

TF-SPME devices made of pure monophasic extraction phase, such as the first generation of TF-SPME devices [12], are still used in some studies as passive samplers. In the case of pyrethroids, a class of significantly hydrophobic insecticides that are known to cause damage to beneficial insects and fish, the suitability of TF-SPME has been just recently studied using thin films (25–500 μm) of different materials (silicon, polyethylene, polymethylmethacrylate, polyoxymethylene and polyurethane) as passive samplers [60]. Another alternative geometry of passive sampling, sorptive tape extraction (STE) [61], has also been effectively used for the direct sampling of plant volatiles [33,62] by both headspace and direct application to the plant surface. This geometry uses a tape-like PDMS thin film as a sampling device, providing an easy method of application for the non-invasive sampling of environmental and biological [61] matrices, recently being demonstrated by Boggia et al. [33] to be suitable for the analysis of herbivory-induced plant volatiles. While viewed currently as a passive sampling device, similar technology could be implemented in the development of new, more convenient TF-SPME devices.

During the development of particle-loaded TF-SPME devices, in this instance DVB particles in PDMS, Jiang et al. demonstrated the greatly enhanced sensitivity TF-SPME offers in trace air sampling and monitoring [36]. This particle loaded TF-SPME device was able to achieve extraction of a wide range of analytes with differing volatilities at high capacity. Furthermore, the study quantitatively samples benzene as a model analyte, a known carcinogen that is found commonly in fuels and smoking devices or from polluted air near a high-traffic road. Other more recent studies have confirmed the use of TF-SPME as a validated onsite sampling tool by its use in the analysis of biocides and UV blockers in sunscreen found in rivers, utilizing HLB/PAN and octadecyl (C18)/PAN TF-SPME devices [54]. Moreover, the efficacy of TF-SPME for the analysis of trace and ultra-trace level analysis in environmental matrices has been repeatedly tested and validated since its inception. In more recent times, since the development of carbon mesh-based TF-SPME devices, TF-SPME has been proven to be even more of a convenient sampling tool for onsite analysis due to the structural robustness of the device and its ease of introduction to portable instrumentation [19]. Since then, an interlaboratory study comparing these newly improved TF-SPME devices (DVB/PDMS on carbon mesh) to EPA-validated LLE methods demonstrated the efficacy of TF-SPME as a greener and more sensitive technique for the routine analysis of pesticides in water [55]. This comparative study by Piri-Moghadam and colleagues compared the suitability of different TF-SPME approaches to LLE. These approaches comprised of an in-bottle TF-SPME method using benchtop GC/MS, an onsite drill-assisted sampling that later used benchtop GC/MS and a procedure that used the same drill-assisted sampling as the previous but instead used a portable GC/MS to achieve both onsite sampling and onsite analysis. Results demonstrated (Table 2) the robustness of all TF-SPME-based methods, with the onsite extraction and analysis being the most environmentally friendly of all methods in the study [55]. In a similar fashion, the performance of TF-SPME has further been compared to other extraction methods, namely fiber SPME and SPE, in the analysis of harmful coal frothing agents that have been reportedly released into

environmental waters and, by consequence, contaminated drinking water supplies. This study, the first to use a Car/PDMS TF-SPME device with carbon mesh support, was able to reliably extract multiple components of the coal frothing agent, crude (4-methylcyclohexyl)methanol (MCHM), along with a tentative metabolite with minimal sample manipulation. Results again showed TF-SPME devices to be more efficient at trace-level analysis than other methods (Table 2) [8]. Furthermore, with the development of novel HLB TF-SPME devices, Grandy and colleagues demonstrated the wide range of analytes that can be extracted by the analysis of chlorination byproducts in residential hot tubs [30].

Applications of TF-SPME devices using other modes of desorption have also been applied in the environmental sector. Recently, de la Calle and colleagues have utilized graphene TF-SPME devices coupled with chelating agents to sample various different metals in aqueous samples before analysis using total reflection X-ray fluorescence (TXRF) [28]. Similar methods have also employed similar TF-SPME-TXRF protocols with the use of nanomaterial TF-SPME devices [27]. In recent years, nanomaterial-based TF-SPME devices have been successfully applied to TD [63]. Mohammadi et al. demonstrated the use of a self-supported TF-SPME device composed of a zeolitic imidazolate framework, which extracts the organophosphorus pesticide ethion with subsequent TD for the analysis of environmental water samples [63]. In other respects, there are examples of more conventional TF-SPME being used in conjunction with LD for environmental analysis [64], ranging from the analysis of fluorinated benzoic acids in aqueous samples [21] to the analysis of polychlorinated biphenyls (PCBs) by direct application of a TF-SPME device into fish tissue [65].

#### *7.2. Flavors and Fragrance Analysis*

TF-SPME has various applications for the analysis of flavors and fragrances in a variety of different matrices. In this regard, Stuff and colleagues applied carbon mesh-supported DVB/PDMS as an extraction phase for the sampling of various volatile compounds, such as alcohols and ethyl esters [66]. The improved sensitivity for these various classes of compounds, afforded by TF-SPME, is essential for the rigorous demands of quality control in the beverage industry. A similar study by Vernarelli et al. investigated the efficacy of TF-SPME technology for the analysis of foodstuffs using DVB/PDMS TF-SPME devices and analyzed dark chocolate, cheeses and Caesar dressing [34]. These studies [34,66] compared the performance of TF-SPME to fiber SPME with the same stationary phase, revealing enhanced extraction efficiency and capacity for the TF-SPME devices. In 2016, another study investigated the fragrance of various essential volatile compounds from grapes, including linalool and 3-isobutyl-2-methoxypyrazine (IBMP), which were measured using PDMS-based TF-SPME devices [38]. This novel solid-phase mesh-enhanced sorption from headspace (SPMESH) method developed by Jastrzembski and colleagues resulted in greater throughput and enhanced limits of quantitation (ppb) during the analysis of volatile compounds (such as odorants) compared to other traditional methods. Sol-gel coating of a stainless steel mesh substrate with a thin film of PDMS provided great thermal stability and high sensitivity for the direct analysis with SPME-DART [38]. In 2007, Bicchi and coworkers compared the results of headspace and direct contact sorptive tape extraction (STE), all the while employing fiber HS-SPME as a reference standard [62]. The analytes extracted, volatile compounds found in various solid biological matrices such as apple, perfume on human skin, rosemary and spearmint, all demonstrated increased sensitivity and extraction efficiency compared to the more conventional fiber HS-SPME method.

#### *7.3. Other Applications of TF-SPME*

In addition to the diverse applications TF-SPME provides in the environmental and flavor/fragrance fields, many other procedures have been validated with a variety of matrices. Of these applications, one of the most striking uses of TF-SPME has been the analysis of sebum, developed by Sisalli and colleagues for the non-invasive detection of sebum in vivo [61]. This novel TF-SPME device, utilizing an adhesive thin layer of PDMS, demonstrated good performance in the extraction of sebum and other constituents of skin by a simple placement of the device on human skin, later being thermally desorbed in a TDU/cryo-trap system. This rapid and non-invasive sampling of human skin is crucial for the further development of sampling methods for the cosmetic and pharmaceutical industries. In a similar fashion, Jiang et al. introduced a novel in vivo sampling method for the analysis of human skin constituents using a thin layer of PDMS [35]. To prevent saturation of the device from sebum and other common skin oils, the PDMS-TF was emplaced between two pieces of stainless steel mesh and then placed on the surface of skin for sampling (Figure 5). As a result, this developed approach was able to demonstrate great promise for applications in clinical settings due to its high reproducibility and non-invasive sampling procedure. In another example of TF-SPME being well suited for both in vivo and ex vivo sampling in clinical settings, Bessonneau et al. prepared HLB/PDMS and (C18)/PDMS TF-SPME devices for the investigation of prohibited substances in saliva, analyzed with both LC-MS/MS and GC-MS [52]. Furthermore, this study demonstrated increased analytical precision compared to their similarly developed ex vivo method, confirming the need for more rugged in vivo sampling devices for clinical settings. In 2019, Shigeyama and colleagues reported another application of TF-SPME by use of zeolite-based devices for the extraction of volatile organic compounds (VOCs) in saliva, a class of compounds that is used to determine if a patient has oral cancer [67]. This study again demonstrated the robustness of TF-SPME devices as a non-invasive method. In a more recent study, Mirabelli et al. [39] utilized self-supported TF-SPME prepared according to the method proposed by Jiang et al. [36] to extract illicit drugs in both beverages and biofluids. In this study, for the first time, it was demonstrated that the DVB/PDMS TF-SPME devices were suitable for ultrasound-assisted extraction, with a consequent drastic reduction of extraction time prior to direct coupling to a DBDI source. Moreover, this approach allowed rapid quantitative desorption, reducing the likelihood of the thermal degradation of sensitive analytes. As a result, this newly developed method demonstrated several advantageous aspects, including rapid analytical throughput and enhanced sensitivity. With the simplicity for TF-SPME devices to be used as onsite samplers, along with their incredible sensitivity, they have been proven to be suitable tools for environmental analysis, while their fast and non-invasive sampling affords them great potential for both the pharmaceutical and clinical industries (Table 3).

**Figure 5.** The development of polydimethylsiloxane (PDMS) TF-SPME devices for the analysis of skin volatiles. Reproduced from [35], with permission from Elsevier, 2013.


*Separations* **2019** , *6*, 39





EVA, ethylene-vinyl acetate; LDPE, low-density polyethylene; HLB, hydrophobic lipophilic balance; SAX, strong anion exchange; PAHs, polycyclic aromatic hydrocarbons; PCBs,polychlorinated biphenyls; PS-DVB-WAX, polystyrene divinylbenzene weak anion exchange; TXRF, total reflection X-ray fluorescence; IMS, ion mobility spectrometry.

#### **8. Concluding Remarks and Future Directions**

According to the principles of green analytical chemistry (GAC), the environmental impact of analytical methodologies should be minimized by reducing the amount of solvents used for sample pre-treatment and the use of toxic reagents, as well as developing alternative methodologies not requiring solvents and reagents [70]. As broadly discussed in this review article, TF-SPME, as an alternative geometry of SPME, is able to satisfy the requirements for greener sampling strategies yet providing ease of use, high-throughput workflows, extraction capability for trace analysis, robustness for onsite sampling, suitability for in vivo analysis and easy coupling to various separation platforms and direct MS analysis. When an analytical eco-scale for assessing the greenness of analytical procedures is performed [2,9] on conventional LLE methods, currently in use by regulatory agencies, versus newly developed TF-SPME-based approaches, the minimized use of organic solvents and production of laboratory waste allows TF-SPME to collect at least 50% less penalty points (based on parameters of an analytical process that are not in agreement with the ideal green analysis) compared to LLE. The versatility of the TF-SPME geometry both in the self-supported or supported devices enable this technique to fit various analytical needs for environmental, food and bioanalysis. As an example, TF-SPME can be used as wearable devices for passive sampling of skin emissions or as extractive probes for remote sampling by use of drones, when sampling sites are not easily reachable or their contamination levels could pose hazards to the analyst.

Due to the recent commercialization of carbon mesh supported TF-SPME devices, we envision that their use in academic and industrial premises will increase quickly in the upcoming years. Further work can be envisioned to test the ruggedness of these devices in complex fluids and evaluate their capabilities of performing multiple extraction/desorption cycles in these matrixes without significant loss of extraction efficiency. Moreover, further development of automation strategies for TF-SPME could aim for the complete automation of extraction/desorption cycles when needed, achieving the same throughput capabilities of fiber SPME.

**Author Contributions:** Conceptualization, E.G.; writing—original draft preparation, R.V.E. and R.T.; writing—review and editing, E.G. and R.V.E.; visualization, E.G., R.V.E. and R.T.; supervision, E.G.; funding acquisition, E.G.

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

**Acknowledgments:** E.G., R.V.E. and R.T. thank The University of Toledo for funding. E.G., R.V.E. and R.T. thank John Stuff from GERSTEL Inc. for providing some of the graphics for Figures 1 and 2.

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

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Advancements in Non-Invasive Biological Surface Sampling and Emerging Applications**

#### **Atakan Arda Nalbant and Ezel Boyacı \***

Department of Chemistry, Middle East Technical University, 06800 Ankara, Turkey; nalbant.atakan@metu.edu.tr **\*** Correspondence: ezel@metu.edu.tr; Tel.: +90-312-210-3208

Received: 8 July 2019; Accepted: 11 October 2019; Published: 4 November 2019

**Abstract:** Biological surfaces such as skin and ocular surface provide a plethora of information about the underlying biological activity of living organisms. However, they pose unique problems arising from their innate complexity, constant exposure of the surface to the surrounding elements, and the general requirement of any sampling method to be as minimally invasive as possible. Therefore, it is challenging but also rewarding to develop novel analytical tools that are suitable for in vivo and in situ sampling from biological surfaces. In this context, wearable extraction devices including passive samplers, extractive patches, and different microextraction technologies come forward as versatile, low-invasive, fast, and reliable sampling and sample preparation tools that are applicable for in vivo and in situ sampling. This review aims to address recent developments in non-invasive in vivo and in situ sampling methods from biological surfaces that introduce new ways and improve upon existing ones. Directions for the development of future technology and potential areas of applications such as clinical, bioanalytical, and doping analyses will also be discussed. These advancements include various types of passive samplers, hydrogels, and polydimethylsiloxane (PDMS) patches/microarrays, and other wearable extraction devices used mainly in skin sampling, among other novel techniques developed for ocular surface and oral tissue/fluid sampling.

**Keywords:** non-invasive sampling; wearable devices; extractive patches; skin sampling; passive sampling; green sampling technologies

#### **1. Introduction**

Biological surfaces provide easy-to-access information about the underlying metabolism and workings of the human body. In broad terms, there are two approaches that can be followed to gather information from biological surfaces. The first approach is the development of wearable sensors, where a direct sensing mechanism is used as a wearable device. Wearable sensors have complex designs that can track a relatively small number of compounds from various biological surfaces, or monitor the specific features of the medical condition of the wearer in different ways [1–5]. While these wearable sensors are quite intricate in their design and mechanisms, they usually provide very basic information in terms of their ability to detect only a limited number of biologically relevant information. Another approach is the development of wearable extractive sampling devices examined in this review, which can be described as being on the opposite side of the coin in the sense that they are quite simple in their design but with the help of advanced detection methods, i.e., high sensitivity mass spectrometry (MS) coupled with the myriad of ionization techniques available, provide a vast amount of information.

#### **2. Skin Sampling**

Sampling and profiling skin metabolites has been a topic of interest over the last several years [6–9]. The majority of the biological information is gathered from sweat, which is secreted from eccrine and apocrine glands, and sebum which is secreted by sebaceous glands. The sebum contains many

distinctive compounds that result from the breakdown of proteins and enzymes in the outer layer of the skin, known as the epidermis [10]. The underlying biological pathways that produce these compounds in sweat, sebum, and epidermis are quite varied and the compounds of interest have a wide range of polarity, size, and structure as a result. These diverging compounds are of interest to clinical research fields such as forensics and toxicology [11–13], not to mention the obvious applications in medicinal diagnosis [14–16], doping tests [17,18], and environmental/occupational exposomics [19–21].

Some of the major difficulties in sampling skin secretions are the challenges of collecting large volumes of sweat and the low concentrations [22,23] of metabolites and other compounds of interest. Therefore, initial studies in the mid-20th century required patients to sweat copious amounts into plastic skin bags, a rather exhaustive method of sample collection [24]. A more commonly employed method nowadays is to use small vessels filled with organic solvents [25] to extract metabolites from a relatively small area, but this method leaves the skin irritated and many of the organic solvents are not biocompatible nor environmentally friendly. Less invasive methods such as Macroduct® [26] and cotton pads [27] for sampling from skin surfaces as well as micro dialysis [28] and subcutaneous solid phase microextraction (SPME) [29] for sampling from inner skin layers have been developed but have some limitations. For instance, some of these methods can only be applied for one type of sample collection, e.g., only for sweat collection or only for sebum sampling. Some methods require additional sample preparation steps to be performed which are time consuming and may introduce additional errors into the analysis. There are substantial efforts to address the limits of the methods listed above and to pave the way for non-invasive, biocompatible in vivo and in situ skin sampling techniques and to integrate them with state-of-art analysis instruments.

#### *2.1. Direct Contact Type Sampling*

In this type of sampling approach, the sampling device, or patch, is placed directly on the skin without leaving any gap between the sampler and the surface. The advantage to this approach is that it provides good extraction capability for both volatile and non-volatile analytes. The main drawback is that many unwanted compounds can adhere to the sampling device by smearing through the skin surface resulting in a considerable amount of matrix contamination. These contaminations might affect sampler performance, resulting in absolute matrix effects, including ionization suppression/enhancement in electrospray ionization (ESI), and may cause frequent instrumental maintenance.

Patch type sampling employs sorbent materials embedded on scaffold or directly placed onto the sampling area, removed after a satisfactory sampling duration, and can either be analyzed by various analytical instrumentations allowing direct analysis (i.e., direct mass spectrometry) or be subjected to a desorption step before analysis by suitable instrumentation. Materials used in patch type sampling should be elastic enough to conform to the shape of the sampling area but resilient enough to sustain the expected wear and tear of normal daily activity if the patch is required to be left on the skin of the subject for an extended period of time. The biocompatibility of the material is another important criterion to prevent irritation of the skin for the same reasons as above.

#### 2.1.1. Polydimethylsiloxane (PDMS)

PDMS has been commonly used as an extractive phase for thin film microextraction (TFME) [30,31] as well as solid phase microextraction (SPME) [32,33] for various different analytes, matrices, and applications. Controlled polymerization of PDMS allows for adjustable elasticity and molding with relatively inert and biocompatible features which are useful for extended sampling periods. Thermal and chemical stability of PDMS, particularly against commonly used cleaning and surface activation solvents, further reinforces its applicability as an extractive phase for patch type sampling purposes.

PDMS can be used as an extractive phase in two different modes: first is the headspace sampling method where the material is not in direct contact with the sampling area and volatile compounds that emanate from skin are captured. The second mode is the direct contact mode, where the PDMS patch is directly placed on top of the sampling area. In this mode, both volatile and semi-volatile compounds can be extracted. However, the skin is usually contaminated by dust particles and many other chemicals, such as cosmetic products; therefore, the patch might convey these unwanted matrix components to the instrument. These matrix components could potentially contaminate the analytical instrument, which would require clean-up before and after each use. In an effort to prevent such compounds from adhering to the PDMS surface, Jiang et al. developed a sampling method which allowed the patch to be used either as a headspace sampling device or for direct contact type sampling from skin [34]. Jiang and colleagues employed two layers of stainless-steel mesh to separate the PDMS film from coming into direct contact with the skin, as shown in Figure 1a. In order to evaluate the reproducibility of the so-called "membrane sandwich" method, an in-vial experiment was conducted in parallel, and intra-membrane irreproducibility was found to be 9.8% (*n* = 6) and inter-membrane reproducibility was determined to be 8.2%. Inter-membrane reproducibility is particularly important for potential clinical applications, where membranes would be used only once. When detected quantities of volatile organic compounds (VOCs) were investigated, such as octanal, nonanal, and decanal, similar intensities were observed for both headspace and direct contact methods. However, for semi- and low-volatile compounds, such as 1-tetradecanol and 1-octadecanaol, higher peak intensities were observed for direct contact sampling. Some of the heavier compounds such as squalene were not detected at all by headspace sampling. These results indicate that, although the chromatograms for the headspace sampling were less contaminated by matrix related background peaks, direct sampling from the surface could be considered as a compromise between detection sensitivity and instrumental contamination. The same study also reported monitoring of food metabolites and alcohol intake directly from skin could be a potential application for such patches.

**Figure 1.** (**a**) Schematic diagram of the "membrane sandwich" headspace sampling for skin volatiles sampling. (**b**) Polydimethylsiloxane (PDMS) patches placed over the ventral ear surface for direct sampling. These were then covered with PTFE (polytetrafluoroethylene) and Tegaderm® (not shown). Reproduced from [34,35], with permission from Elsevier and IOP Publishing, 2019.

Direct sampling mode for PDMS were employed in many other studies showing the potential of PDMS patches for clinical applications, all highlighting the simplicity of using the patches in direct sampling mode. For instance, a rabbit model study was conducted by Schivo et al. that employed PDMS patches (Figure 1b) placed inside the ears of subjects to investigate histological evidence of early-stage ulcer formation by metabolomic screening [35]. The study results identified that the patches made of PDMS can be used for monitoring the differences between healthy and diseased skin metabolites. Approximately 150 unique skin related compounds were characterized, and 12 biologically relevant compounds were detected abundantly in the ulcer group. Most of the skin metabolites observed in the study were waxy compounds, although a large number of VOCs were detected as well. Results of the study underline that the abundance of non-volatile compounds makes direct contact sampling with PDMS a good fit and emphasize that the distribution of volatile and non-volatile compounds can be compared to human skin profile. In another study, Martin et al. employed PDMS patches placed onto the axilla area of female participants for VOC sampling. The

study aimed to investigate skin metabolites responsible for body odor with an emphasis on the effect of a single nucleotide polymorphism 538G → A in the ABCC11 gene on the concentrations of apocrine derived axillary odor molecules, especially 3-methy-2-hexenoic acid [36]. The investigation was concluded with four volatile fatty acid (VFA) target compounds being identified with limits of detection approximately around 100 s pg cm−2, with authors noting that a typical skin patch has an area of 0.5 cm2. It was also noted that the ABCC11 gene, other than controlling body odor, is studied for its role in breast cancer and drug resistance in cancer cells [37]. Based on the study results, combining the sampling patches with a thermal desorption unit coupled to a secondary electrospray ionization and mass spectrometry (TD–SESI–MS) setup is a promising approach for unattended automation of analysis and could be employed in non-invasive fast screening of metabolites for personalized medicine. In a pilot study [38], Martin et al. investigated changes in VOC profiles for stress-related biomarkers using PDMS devices placed on the foreheads of volunteer subjects for sample collection. The effect of stress on VOCs exhaled in breath was investigated previously [39], identifying six potential marker compounds. Similarly, the results obtained by Martin et al. suggested that some of the compounds identified in breath exhale in the previous study [39] also show altered levels in skin, namely terpenes (3-carene is suggested as a likely candidate), benzoic acid, and N-decanoic acid seem to be involved in the metabolic response to stress, which the authors believe a potential increase in oxidative metabolic pathways could be the cause of.

Moreover, a clinical study by Stevens et al. have explored the spatial distribution of bacterial populations and related VOCs using various sampling techniques, with PDMS patch type sampling being one of them [40]. In this study, PDMS patches were used for VOC sampling. Different areas of the foot, including toe clefts, and several dorsal and plantar surfaces, were sampled for microbiological and VOC analyses. Then, a spatial map of the foot was constructed for microbiological and VOC concentrations which showed significant differences between different regions. Specific volatiles such as acetic, butyric, valeric, and isovaleric acids were found in considerably high concentrations in the sole region of the foot compared to the dorsal region, and, in the case of isovaleric acid, it was never found in dorsal regions while it was readily detected on the plantar surface. Their findings indicate that key volatiles responsible for malodors are abundantly found on the sole of the foot rather than in other regions sampled in the study. This study clearly indicates that sampling from different regions of the body might give different levels of particular metabolite which could be simply related to differences in body temperature, excretion rates, or skin thickness or might be indicative of biologically relevant information. Sampling conditions of such results has to be carefully designed in order to avoid any misinterpretation.

As it has been remarked above, PDMS based patches are versatile tools and easy to adopt for numerous investigations. In addition, PDMS films can be purchased or prepared in the laboratory using well defined protocols and can be easily shaped in any size. In addition, the thermal stability of PDMS enables the thermal desorption of extracted analytes from patches using large thermal desorption units connected to gas chromatography mass spectrometry (GC–MS) [35,36,38]. However, it is worth mentioning that before use PDMS patches require extensive cleaning and activation steps that might involve a considerable amount of preparation before sampling (4–24 h). Therefore, it is crucial to prevent contamination at each step of the workflow including cleaning/pre-treatment and sampling. Moreover, during the sampling, PDMS patches should be protected from environmental wear and tear and possible contaminations by a placing a durable cover over them, as shown by Schivo et al. in their pilot study [35].

#### 2.1.2. Agarose Hydrogel

Agarose hydrogel is another commonly used material for direct contact type sampling. Agarose hydrogel lacks the robustness of PDMS patches, but it is more biocompatible than PDMS, as it is a naturally occurring homopolymer derived from red algae [41]. Agarose has completely different extractive properties suitable for hydrophilic skin metabolites owing to its complex polar structure and adjustable water content. These properties lead to agarose patches being mostly used for extraction of relatively polar metabolites as can be seen from the examples below.

For instance, Dutkiewicz et al. have developed agarose micropatches to detect several low-molecular-weight skin metabolites as a proof-of-concept study [42]. In this study, the lack of robustness of agarose hydrogels was overcome by embedding them inside cavities on a PTFE chip. Various prevalent skin metabolites were detected, and some were identified, as given in Table 1. The authors stated that high signal-to-noise ratios were achieved for clinically relevant analytes. In another study, Dutkiewicz et al. also employed micropatch-arrayed pads (MAPAs) to profile topically applied drugs. In this study, five rows of five cavities, a total of 25, containing agarose hydrogels were applied to both in vivo human skin and ex vivo porcine skin by self-adhesive 3-D printed PTFE scaffolds [43]. In a later study by the same group, MAPAs were implemented to investigate psoriasis-related skin metabolites and an automated diagnostic method was developed with satisfactory results that imply clinical use of MAPAs is a reasonable possibility [44]. A number of metabolites, namely choline, citrulline, glutamic acid, urocanic acid, lactic acid, and phenylalanine, were discovered to have altered concentrations in diseased skin areas, indicating they are related to psoriasis metabolism. One of the key advantages of MAPAs is the direct coupling with nano desorption electrospray ionization (nanoDESI). In this approach, a solvent mixture can be continuously pumped through a capillary to one of the hydrogel micropatches while another capillary delivers the solvent to the nanoDESI loading dock. This approach eliminates further steps (e.g., desorption, pre-concentration, etc.) and provides a quick and simple online method for direct mass spectrometric analysis.


**Table 1.** Identification of peaks in the mass spectra of sweat [44].

<sup>a</sup> Putative formula and name of the metabolite. <sup>b</sup> Values calculated for [M − H]<sup>−</sup> ions (mass of an electron is included). \**m*/*z*: mass-to-charge ratio. IT: ion trap; FT–ICR: Fourier transform ion cyclotron resonance. Reproduced from [44], with permission from ACS Publications, 2019.

As it has been pointed out several times by now in the studies shown above, agarose hydrogels are limited for extraction of relatively polar analytes. This limitation narrows the biologically relevant information that can be obtained from the studied system. Other biocompatible hydrogels, such as gelatin and polyvinyl pyrrolidone, could potentially offer the same, or even greater, advantage for skin sampling and further functionalization of agarose or other hydrogels could very well be a prospective area of research in the future. Selective sampling can be also be achieved by using copolymers of different hydrogels with different functional groups to target compounds with varying polarities.

#### 2.1.3. Microneedle Arrays

In recent years there has been considerable effort to employ low-invasive microneedle structures (<1-mm long, hollow structures) in drug delivery through skin [45,46]. More recently, this technique was modified to be employed in capturing circulating biomarkers [47–52], combining microneedle structures with biomarker-based immunoassay approach. This was achieved by functionalizing the tips of the microneedles with specific antibodies. Although functionalized microarrays provide high selectivity and could benefit many clinical diagnostic applications, only a limited number of studies showing the potential of the approach have been conducted. In one of the studies, Ng et al. developed a multiplex microneedle device coupled with a blotting method for fast and selective detection of multiple selected biomarkers from skin surface [53]. Microneedle devices were produced by micromoulding: polylactic acid (PLA) was melted in vacuo into the template micromould which was produced from the Sylgard® 184 (PDMS) elastomer and surface activated using chemical methods. Surface characterization of the PLA microneedle device was performed using a scanning electron microscope (SEM) at various production stages, which can be seen in Figure 2. For immobilization of different antibodies on the multiplex microneedle device, each microneedle array was individually dipped into the desired antibody solution 10 times. Microneedle arrays were then placed for 1 h on mice skin doped with mouse (Interleukin 6 and 1) as well as human (Tumor Necrosis Factor alpha) antigens. After successful antigen capture, two detection methods were employed: UV/Vis spectrophotometry for quantitative detection and scanning the images of blotting patterns left by the microneedles for qualitative visualization. Blotting technique employed by the authors should be further underlined as it requires no specialized equipment and allows rapid and selective visualization of specific antigens, albeit qualitatively. It should be kept in mind that employing the detection of antigens with this method has its disadvantages such as having a limited number of antibodies that can be loaded into the microneedle arrays. Therefore, scanning for a large number of different antigens would require several runs which would increase the invasiveness of the method since different areas have to be sampled.

**Figure 2.** (**a**) A photograph of the template microneedle array. (**b**–**d**) Light micrographs showing: (**b**) a template microneedle, (**c**) a polylactic acid (PLA) microneedle prior to surface activation, and (**d**) a surface-activated PLA microneedle. (**e**) A surface-activated PLA microneedle array under SEM; (**f**) a single microneedle on this array is enlarged and shown. (**g**) Perforation marks on hairless mouse skin, visualized by methylene blue staining. Reproduced from [53], with permission from Springer, 2019.

Microneedle arrays, as indicated by this study, show promise as a low-invasive method that can be employed in diagnosis of skin diseases. The highly selective nature of using antibodies further promotes their diagnostic potential since extensively studied diseases can be targeted by specifically targeting selected antigens. Creative approaches to the geometric structures of the arrays may allow the number of microneedles in the same area to be increased, enabling more antigens to be detected.

#### *2.2. Headspace Sampling*

#### 2.2.1. Conventional SPME Fibers

Patch type or other contact-based sampling methods [6,34–36,38,54–56] may not always be preferable for skin VOC analysis as contamination and introduction of non-volatile compounds is always a possibility. Headspace SPME is a commonly used sampling approach preferred for extracting volatiles present in various samples including complex matrices [57–60]. Several studies have scrutinized their applicability for analysis of skin VOCs [61,62] albeit not without its limitations due to their fragility, which is one of the main reasons why in situ headspace sampling is practiced in controlled environments. A simple solution for this problem is employing a housing for the SPME fibers during the sampling. This approach creates a controlled environment and ensures that extracted analytes on the fiber are only associated with the sampling area. Similar to how PDMS and agarose patches can be protected by durable covers or by being placed on PTFE scaffolds, Duffy et al. have shown that headspace SPME fibers can be protected by using wearable housing vials, seen in Figure 3a [63]. In this study, researchers compared skin VOCs before and after the acute barrier disruption of the sampling area by tape stripping to simulate impaired skin. After tape stripping, commercial divinylbenzene/carboxen/polydimethylsiloxane Stableflex fibers were used to sample skin VOCs. A total of 37 compounds were identified that were significantly altered after barrier disruption, mainly consisting of aldehydes (hexanal, nonanal, decanal), acids (nonanoic, decanoic, dodecanoic, tetradecanoic and pentadecanoic acids), and hydrocarbons (squalene). Duffy et al. have also utilized their wearable headspace SPME device in another study to investigate fragrance longevity and scent profiles of participants. In this study, skin VOCs of volunteers were analyzed before and after fragrance-derived compounds were applied to their skin to compare the difference and fragrance permanence [64]. Following the instrumental analysis, 32 fragrance-derived and 19 endogenous compounds were identified. Several endogenous VOCs were found to be suppressed by fragrance application, most apparent ones being 2-decanal, 2-undecenal, 1-dodecanol, pentadecanal, and octadecanoic acid. The most noticeable decrease in endogenous VOCs was observed immediately after fragrance application where several skin gland secretions and their oxidation products were not detected including acids, aldehydes, ketones, and hydrocarbons. Temporal and spatial profiling of fragrance-derived and endogenous compounds can prove unique insights to improve fragrance longevity and can lead to niche personalized fragrance production in the cosmetics industry.

As can be seen from the outlined studies, although such housings are not as practical as wearable patches described in previous section, they provide unique advantage for creating a well-controlled environment during sampling. Moreover, various commercial SPME fibers are readily available for purchase with different extractive properties and homemade fibers can be coated with materials that can be specialized for unique sampling purposes. Another important advantage of the SPME fibers is that they can be directly coupled with ambient ionization techniques such as nanoelectrospray ionization (nanoESI) [65], dielectric barrier discharge ionization (DBDI) [66], and open port probe (OPP) interfaces [67]. Direct coupling to MS instruments eliminates desorption and chromatographic separation steps from the analytical workflow, speeding up the overall process.

#### 2.2.2. Passive Flux Samplers

Passive flux samplers (PFSs) are commonly employed in environmental studies to measure toxic or greenhouse gases emitted from animal slaughterhouses [68–70], greenhouses [71], and building materials [72], which was later adapted to sample VOCs emanating from skin [62]. So-called passive sampling devices (PSDs) consist of a small stainless-steel plate, a trapping filter that contains an extractive phase to extract relatively volatile analytes, a PTFE O-ring creating a headspace between the extractive phase and skin surface, and a back-up plate. Its schematics can be seen in Figure 3b. PFSs can be considered more practical and robust than wearable SPME fibers with integrated housing in terms of ease-of-use. PFSs, however, are more susceptible to the environment and there is always a possibility of environmental background to be considered when using this technique in sampling unless completely isolated from the environment. Due to their ease-of-use feature these devices can be adapted for sampling from various parts of the body and track metabolic changes associated with particular conditions. For example, a recent study by Kimura et al. investigated the causes of the characteristic "elderly body odor" by sampling from different areas of the skin from participants in three different age groups (young-, middle- and old-aged) using a passive sampling device [73]. MonoTrap® DCC18 containing octadecylsilane functionalized monolithic silicate and activated carbon as extractive media providing with a large surface area and high trapping capacity (O.D. 10 mm × 1 mm thick, >150 m<sup>2</sup> g<sup>−</sup>1) was chosen for the study. Two compounds, 2-nonenal and diacetyl, were chosen as likely candidates to study as to the cause of the elderly body odor phenomena. The emission fluxes of the compounds with respect to sampling position (left forearm, left thigh, left calf, forehead, nape of the neck and abdomen), the effect of diffusion length on the dermal emission flux, and distribution of the emission fluxes, and changes by age and sex were investigated. Analysis of the collected samples illustrated that emission flux of 2-nonenal increases significantly with age while diacetyl was found to have the highest emission flux for middle aged participants. Moreover, male participants were found to have higher emission fluxes compared to female participants of similar age. Interestingly, out of all sampling areas, nape of the neck was found to be the most reliable sampling spot, since the emission fluxes of both compounds altered significantly in other sampling areas but remained relatively stable in this area. Due to the fact that eccrine and sebaceous glands are believed to be potential sources for both compounds, their abundant presence in the area is reasonable.

**Figure 3.** (**a**) Headspace sampling volatiles on the volar forearm using a wearable housing integrating solid phase microextraction (SPME) fiber, affixed to the skin with surgical tape. (**b**) Schematic view of the passive flux sampler for human skin gas. (**c**) Sampling of 2-nonenal and diacetyl at the nape of the neck. The PFS (passive flux sampler) was fixed by a piece of medical tape. Reproduced from [63,73], with permission from Wiley and Elsevier, 2019.

In another study, a homemade PSD (Figure 3c) was employed by Furukawa et al. to determine the whole body dermal emission rate of ammonia by taking simultaneous measurements from a total of 13 chosen spots around the body in an attempt to cover all regions of the body [74]. In order to prepare homemade PSD for ammonia, a commercial cellulose filter paper first was dipped in methanol solution containing 2% phosphoric acid and 1% glycerol and then dried under vacuum. After sampling from selected body parts, the trapped ammonia was back extracted with pure water and its concentration was determined by ion chromatography. Results revealed that the total rates of ammonia emission ranged from 2.9 to 12 mg h−<sup>1</sup> with an average emission rate of 5.9 <sup>±</sup> 3.2 mg h−<sup>1</sup> for each person. Partial emission rates obtained in the study per body part can be found below in Figure 4. Higher emission flux of ammonia was found at the feet, back, and lumbar, where large sweat glands are located, and lower flux was observed at upper arms, buttock, thighs, and legs. Based on these results the authors speculated that there are two routes for ammonia emission in the body. The first route

follows blood ammonia, produced by metabolic reactions that involve proteins, which can directly rise to the skin surface from blood capillaries. The second proposed route is mixing of blood ammonia into sweat, which explains the increased emission flux levels in areas where sweat glands are concentrated. As can be inferred, such ease-of-use devices are very useful for tracking metabolic pathways in the human body.

**Figure 4.** Comparison of emission flux of ammonia emanating from 13 sampling positions of ten volunteers. Error bars show standard deviation of measured values of tested volunteers (*n* = 10 for all, *n* = 5 for male, and *n* = 5 for female volunteers). Reproduced from [74], with permission from Elsevier, 2019.

In addition to metabolic profiling of natural skin metabolites summarized above, there are also some preliminary studies performed using PSDs for exposomic monitoring showing the potential of wearable devices for exposure monitoring. For instance, Sekine et al. have investigated VOCs emanating from the skin of smokers and volunteers who were exposed to second hand smoke using PSDs [75]. The device employed was identical to the ones used in the group's previous studies (MonoTrap® DCC18 trapping medium). Six volunteers (four smokers and two non-smokers) participated in three different experiments to investigate the effects of cigarette smoking (for smokers) and second-hand smoke (for non-smokers) on dermal VOC emissions. PSDs were placed for 1 h on the forearm and the back of the hand (Figure 5a) of the volunteers. In the first test, a smoker and non-smoker volunteer stayed in the same room for 15 min while the smoking volunteer smoked a single cigarette. Based on the results of the first test, acetaldehyde, toluene, 3-methyl furan (3-MF), 2,5-dimethyl furan(2,5-DMF), 3-ethenyl pyridine(3-EP), and nicotine were chosen as the dermal VOCs to be investigated in further analyses. Following tests investigated the concentrations of these compounds in smokers and non-smokers (through second-hand smoking) in different conditions. Results revealed that for smokers, there is an initial spike in nicotine and its derivative 3-EP immediately after smoking which decreases to its initial value after 2 h. In the case of secondhand smokers, the maximum dermal emission for the same compound was observed at the 2-h mark indicating that cigarette smoke has different routes of entrance to body. Another interesting result obtained in this study revealed that there is higher skin emanation of 3-MF, 2,5-DMF, 3-EP, and nicotine when a non-smoking volunteer was exposed to direct smoke emanating from a burning cigarette compared to skin emanation when the same volunteer was exposed to environmental second hand smoke. Another recent study investigating the biotic and abiotic exposome was able to successfully use a home-made wearable passive sampling device containing 3D printed cartridge filled with molecular sieves (zeolite), capable of adsorbing

both relatively polar and non-polar compounds [76]. The cartridge was attached to the particle free air flow apparatus producing a portable device collecting both biotic and abiotic species from the exposed environment with a constant rate of sampling at 0.5 L min−1. In the study, 15 participants were monitored for a time span covering up to two years in over more than 60 different geographical locations. The sampling was performed by wearing the device when possible or keeping it in two-meter proximity to the participant when it was not possible to wear the device. The study results revealed that the human exposome is varied and dynamic and closely effected by parameters including lifestyle, environment, geographic location, as well as that it can be unique for each person even for people living under similar environmental conditions. Both studies clearly highlight the importance of wearable extractive devices for monitoring public health, microbiome, and environmental exposure to chemicals just to name a few.

**Figure 5.** (**a**) Sampling of volatile chemicals on the forearm and back of the hand before and after cigarette smoking or exposure to second hand smoke (SHS). (**b**) Wrist (left) and ankle (right) sampling using a PDMS loop passive sampling device. (**c**) c.1–2 bags used for transport that were attached to track participant ID and exposure time in the occupational deployments; c.3 single wristband deployment. Reproduced from [75,77,78], with permission from Elsevier and ACS Publishing, 2019.

#### 2.2.3. Other Wearable Headspace Extractive Samplers

Unconventional headspace extractive samplers including PDMS tubing sampling loops and silicone wristbands have been suggested in several studies for qualitative monitoring purposes. For instance, Roodt et al. used PDMS tubing in a study investigating the relationship between skin microbiome and anthropophilic mosquito disease vectors [77]. PDMS tubing was cut (180 and 240 mm tubing length, 0.25 mm inner diameter) to manufacture sampler loops (Figure 5b). The flexible nature of PDMS tubing allowed samplers to be employed in non-conventional sampling areas, in the wrists and ankles of the participants. Among the wide range of extracted analytes, 88 were identified in total, several of which, most notably cyclic ketones, were not previously reported in skin volatile literature. As an exemplary application area, the correlation between human skin microbiome and the attractiveness of participants to anthropophilic blood host seeking mosquitoes were investigated. Identification of previously unreported skin VOCs indicates that such passive sampling devices have not yet reached their full potential and can be explored to sample from unconventional skin areas. Another recent study using silicone wristbands was able to demonstrate the applicability of wearable extractive devices for monitoring of occupational exposure of roofers to chemical compounds released during hot asphalt application [78]. The obtained results indicated that the silicone wristbands were able to absorb 25 polycyclic aromatic hydrocarbons (PAHs) during 8 h of exposure under working

conditions as well as differentiate the variations in the amounts of the PAH collected in divergent environmental condition, suggesting its applicability for sensitive monitoring. The silicone bands can be seen above in Figure 5c. PDMS tubing sampling loops and silicone wristbands represent the quintessential wearable sampling devices in the sense that they are simple, practical, robust devices that have the capability to be fine-tuned by surface or bulk material functionalization. The possibility of functionalization could allow these types of devices to either be used to target a specific range of analytes in terms of polarity, size, and volatility or to increase their extractive range to be used in untargeted studies. However, it should be kept in mind that these devices are completely open to the environment; therefore, a high background signal should be expected.

A summary of recent non-invasive skin sampling techniques, devices, and their applications can be found below in Table 2.



#### **3. Oral Fluid and Ocular Surface Sampling**

As summarized above, most of the research including wearable extractive devices has focused on detection of skin related analytes. Studies discussed above were mostly focused on implementing already existing tools/extractive materials directly for skin sampling, neglecting other possible matrices that can be used to gather biological information. In fact, saliva and ocular surfaces can be sampled easily and could provide unique information about a biological system. However, as it has been shown below, the potential of wearable devices for those matrices have not been thoroughly explored yet.

#### *3.1. Saliva Sampling*

Saliva sampling has been gathering interest as an alternative sampling matrix to blood and urine for forensic applications, disease biomarkers, drug and doping control, and flavor studies amongst other niche areas. One of the main reasons why saliva is an attractive alternative is that the samples can be collected relatively easy without privacy concerns which are especially important in doping tests. Saliva is commonly sampled through draining, spitting, and contact sampling with commercial products such as Salivette® or Drugwipe®, and by chewing inert materials [79–82]. Saliva sampling is inherently less invasive compared to some other biological specimens like blood or urine, and there is still potential for developing alternative methods applicable for in vivo sampling.

One advancement towards in vivo saliva analysis was realized by Bessonneau et al. employing thin film microextraction (TFME) with GC–MS and LC–MS techniques to evaluate and compare the ability of in vivo and ex vivo sampling, as well as the ability to determine prohibited substances in saliva [83]. In order to obtain maximum metabolite coverage two different type of extractive phases were used for salivary sampling. Hydrophilic lipophilic balanced (HLB) particles were embedded in PDMS to be used as the extractive phase for GC–MS analysis and HLB particles were embedded in polyacrylonitrile (PAN) to be used as extractive phase for LC–MS analysis. For in vivo sampling, TFME devices were placed under the tongue of participants for 5 min and then extracted analytes on the HLB–PDMS phase were desorbed directly to GC–MS using a high volume thermal desorption unit while extracted analytes on HLB–PAN phase were first desorbed into a suitable solvent prior to their LC–MS analysis. The same desorption protocol was followed for 5 min ex vivo sampling from 1 mL of collected saliva. Comparison of ex vivo to in vivo studies revealed similar results for hydrophilic compounds while higher peak intensities were observed for in vivo sampling for hydrophobic compounds. This is a reasonable result since hydrophobic compounds tend to have secondary interactions with labware. This study highlights the advantage of employing in vivo sampling if hydrophobic compounds are of interest and secondary interactions can potentially disrupt sampling. Moreover, in vivo sampling may provide the unique advantage of capturing short lived metabolites that are susceptible to decomposition which can be difficult to be determined with other methods.

#### *3.2. Oral Tissue Sampling*

Chen et al. developed a different approach for oral sampling by using a moving string sampling probe for in situ endoscopic MS of a living mouse and to take samples from oral surfaces [84]. The device consists of a disposable cotton sampling string, which moves through the sampling area. The string is smeared and carries a small amount of tissue samples that adhere to its surface directly to an ionization source to be subsequently analyzed by MS. Figure 6 depicts the schematic of the sampling device. In the study, the sampling probe was attached to an industrial endoscope with a camera and miniature super-bright LED. A metallic tube was bent into a V-shape and partially cut to expose the sampling strip, with the authors noting that the camera allows the sampling process to be monitored in real-time which can be recorded if desired. In this study, the samples were taken from the surface of the tongue of a volunteer who had consumed a caffeinated beverage. Both atmospheric pressure chemical ionization (APCI) and ESI mass spectra showed a strong peak at m/z 195.09 ([Caffeine + H]+) and two peaks at m/z values 116.07 ([C5H9NO2 + H]<sup>+</sup>) and 118.09 ([C5H11NO2 + H]+) which were tentatively identified as proline and valine respectively. This further affirms the capability of this method to be used for sampling small molecules such as amino acids. The proof-of-concept setup designed by the investigators employed a cotton string and APCI/ESI–MS. The moving string approach can be realized with any type string, as the authors mention, and with any type of ionization setup after necessary changes are made. Although not used in the study for such purpose, employing a camera to record the sampling process makes it possible to generate a spatial metabolite map of the sampling area to provide an additional layer of information.

**Figure 6.** (**a**) Simplified schematic of the endoscopic mass spectrometry system that uses a moving string as the sampling and transportation material. (**b**) Configuration of the sampling probe. The sample is wiped off from the surface by the moving string and carried to the extraction and ionization region near the mass spectrometry (MS) inlet. (**c**) Photograph of the probe tip. Reproduced from [84], with permission from RSC Publishing, 2019.

#### *3.3. Ocular Surface Sampling*

The human eye is an extremely intricate and complex organ that consists of many distinct parts with distinct characteristics and qualities that are unique in the human body. Seemingly small disruptions to this delicate system can cause visual impairment and discomfort. It has been an ongoing effort to identify and investigate the non-redundant proteins in the human eye, the most notable effort being The Human Eye Proteome Project launched in 2013 [85]. Diagnosis of ocular infections caused by bacteria, fungi, parasites, or viruses is another area of research where there is an interest in developing new clinical tools [86]. A recent study in 2018 conducted by the association for Research in Vision and Ophthalmology (AVRO) reported that 43% of researchers find it "difficult" or "very difficult" to obtain human eye tissue, and 43% of the participants stated that they regularly limit the scope of their work due to inability of finding eye tissue that meets the needs of their research [87]. Therefore, it is of paramount importance to develop new tools to sample compounds from the human eye to tackle the problems researchers face in this area, which can greatly benefit from the advancements in low- and non-invasive sampling techniques.

Such an attempt was made by Liou et al. in a study to sample from the eyelids of volunteers using a watercolor brush to profile caffeine and its metabolites [88]. The ethanol-soaked tip of the watercolor brush was simply rubbed against eyelids of 4 volunteers who had consumed three cups of coffee with 3 h interval between each cup. After sampling, the brush was moved in front of the inlet of the MS instrument. The analytes evaporated from the brush surface by the aid of the ethanol and then subsequently ionized by charged ESI plum before finally being detected by the instrument. The intensity of the ions and the distance of the brush to the ESI needle were investigated, with the findings summarized in Figure 7. Although the authors chose to sample from the eyelid because of the high concentration of blood vessels in the area and the natural trapping properties of eyelashes and not because of any intention to sample eye-related metabolites or compounds, the sampling concept can still be applied to detect eye related metabolites and compounds as the detection of caffeine metabolites indicates. It should also be noted that ethanol which was used to soak the tip of the brush could extract many unwanted skin matrix components and environmental contaminants resulting in high matrix effect during such direct-to-MS approaches.

**Figure 7.** Photo (**A**), a picture of the sampling brush–spray ionization/mass spectrometry set up used in this study. The sampling brush was located between the mass inlet and the ESI needle (not shown in the picture). (**B**) relationship between ion intensities and the location of the brush. In this case, caffeine was used as the test sample (concentration levels, 10 μg mL<sup>−</sup>1) and the ESI voltage was +4.5 kV. The position, indicated in red, denotes the optimized location. Reproduced from [88], with permission from Elsevier, 2019.

An interesting study, which could act as a lighthouse to steer future research, was conducted by Lopez et al. where a hydrogel contact lens was developed to slow the progression of corneal blindness caused by overexpression of zinc-dependent matrix metalloproteinases (MMPs) [89]. Monomer of poly(2-hydroxyetyl methacrylate) (pHEMA) were modified before polymerization with dipicolylamine (DPA) to synthesize the pDPA–HEMA hydrogel. DPA has a selective binding affinity toward zinc ions (up to *Kd* = 1 <sup>×</sup> 10−<sup>11</sup> M [90]) which is essential for MMPs to function. Removal of zinc by the pDPA–HEMA hydrogel contact lens deactivates MMPs and slows down the degradation of the cornea. As summarized in Figure 8, ex vivo studies using porcine cornea to simulate human cornea were conducted by the authors which demonstrated advantageous results compared to conventional treatments. Because there is no systemic circulation of zinc targeting drugs in the body, the treatment

is targeted and limited to the cornea only. In addition, DPA only selectively binds to zinc out of all biologically active metal ions excluding the possibility of side effects. Although this study is not directly related to the ocular surface sampling, it is vital to show the potential of using such biocompatible functionalized materials for extracting biological information directly in vivo from the ocular surface.

**Figure 8.** Schema of deactivation of matrix metalloproteinases (MMPs) by the dipicolylamine– poly(2-hydroxyetyl methacrylate (pDPA–HEMA)-based contact lens. Reproduced from [89], with permission from ACS Publications, 2019.

#### **4. Extractive Patches for Imaging Applications**

Recently, ambient mass spectrometric (AMS) imaging techniques have gained popularity in many areas of research providing insights for metabolomics and lipidomic differences between samples, the discovery of biomarkers creating these differences, and associated clinical applications. However, as can be predicted, these techniques when applied directly to the samples may experience high background, or ionization suppression which decreases the sensitivity of the analysis. Therefore, involvement of suitable sample preparation methods for surface imaging using direct-to-MS approaches are becoming more and more crucial. With this intention, Hemalatha et al. imprinted patterns of printing inks, plant parts, and fungal growth on electrospun nanofiber mats and employed desorption electrospray ionization mass spectrometry (DESI–MS) to rapidly analyze and capture images of the imprinted patterns and analyte droplets [91]. Nanofiber mats were electrospun from a solution of Nylon-6 dissolved in formic acid and characterized by SEM. Patterns were produced on the electrospun mats by either imprinting in the case of plant slices and ink-printed patterns or by dispersing a single drop of selected analyte directly on top of the mat. Imprinted mats were then "scanned" using a 2D moving stage with a 250 μm step distance. In a typical experiment, a spray of charged solvent was pointed to the surface of the imprinted mat. The charged solvent dissolves the compounds and ionizes them similar to electrospray ionization (ESI). Subsequently, the generated molecular ions were transferred to a MS for detection. The images of the samples were obtained by raster scanning the mats under the DESI spray and detection of the molecular ions at each spot. Major metabolites of several plants, analytes, and constituents of dyes were detected and their imprints were imaged, some of the results can be seen in Figure 9.

**Figure 9.** (**a**) Desorption electrospray ionization mass spectrometry (DESI–MS) spectrum from a turmeric rhizome slice of an imprinted nylon nanofiber mat. (**a**,**b**) Optical images of a whole and unskinned turmeric rhizome. (**c**) Imprinted slice on a nylon nanofiber mat. (**d**) DESI–MS image at m/z 219 due to α-turmerone shown. Reproduced from [91], with permission from ACS Publications, 2019.

Li et al. utilized a different imprinting (extractive) media using a similar approach by developing a solid-phase extraction device employing micro-funnels to scan the image of an imprinted strawberry as a proof of concept [92]. In order to functionalize the micro-funnel membrane Parafilm M® and Teflon tape with silicone adhesive was used to fix a mask with a chosen pattern on the surface, and C18 functionalized silicate powder was used to create grooves on the membrane with extractive micro-funnels. Figure 10 shows the workflow of the preparation and analysis of the micro-funnel based SPE device as well as the mapping of major metabolites found in the imprinted strawberry slice. Although both approaches are good candidates to decrease background signal in direct-to-MS studies, the effect of imprinting in chemical image resolution is not discussed thoroughly. In fact, limited image resolution should be expected, mainly due to diffusion phenomenon. The diffusion of analytes in the extractive phase during the chemical imprinting and further diffusion of analyte during the desorption process will be the main reasons for limited resolution. Also, it should be kept in mind that none of these imaging studies were performed directly in situ on living systems. However, they show the potential of the current progress in imprint imaging with MS towards the applicability of wearable extractive devices for non-destructive in vivo and in situ small molecule imaging.

**Figure 10.** (**a**) Flow chart of C18-mounted solid phase extraction (SPE) micro-funnel-based spray for Ion-mobility spectrometry (IMS). (**b**) Chemical images of major components of a strawberry using C18-mounted SPE micro-funnel based spray MS with 0–3 washing steps. Reproduced from [92], with permission from SAGE, 2019.

#### **5. Perspectives in Future Directions and Concluding Remarks**

The goal of this review was to summarize the current technology that has been paving the way for novel applications for non-invasive in situ sampling devices and highlighting potential methods that can be used for wearable extraction-based devices. There are two routes of progress in this area. The first is focused on collection methods of the representative specimen in a non-invasive way from the investigated system, with very little technological improvement in collection devices. The second and more important aspect is the development of wearable extractive devices which are designed to allow free movement of the object during the sampling period with the advantage of integrating the sampling and sample preparation into a single step. This approach in general allows short sampling times with information relevant to the current state of the system, and long-term monitoring providing insightful information about time weighted average (TWA) concentration of the species in the studied system. In this review we focused only on non-invasive extractive sampling tools and wearable devices that can be used in various areas, including clinical application, exposure monitoring, food quality control, and cosmetics. A brief overview of the methods and their advantages as well as their shortcomings is shown in Table 3. The current progress in the area is promising; however, only a fraction of the full potential of such devices is currently being developed and investigated. It is clear that with current direction of the development in noninvasive wearable technologies, soon many different areas will start to use the advantage of these technologies.

**Table 3.** Comparison of reviewed sampling methods (biocompatibility, non-invasiveness, and ease-of-use: ✩ low, ✩✩ medium, ✩✩✩ high).


յ յ յյյ յ

յյ յյյ յյյ


**Table 3.** *Cont*.

<sup>a</sup> Depends on the extractive coating. <sup>b</sup> Ethanol used as solvent reduces biocompatibility. <sup>c</sup> PDMS sheets are commercially available. <sup>d</sup> SPME fibers are commercially sold, however the housing is home-made. <sup>e</sup> There are some commercially available. <sup>f</sup> Can be coupled to GC via large volume thermal desorbers.

For instance, doping control could be one of the areas that could benefit from the wearable extractive patches. Such patches will protect athletes' privacy during the sample collection process as it does not require any assistance for the sampling. Currently, there are some wearable patches that are being considered for sweat sampling; however, their main target is to collect the sweat itself rather than to focus on extraction of analytes that may be present in the sweat, giving the information about drug abuse or other type of doping.

Another area that such patches will benefit is roadside sampling. Particularly if it is combined with a portable device it might benefit opioid and cannabinoid analysis. Taking in to account that marijuana has been legalized in some countries, it is necessary to have tools that can be employed by law enforcement to perform fast roadside tests compatible with portable detection systems.

As can be seen from this review all current sampling devices are focused on skin and sweat sampling. However, alternative matrices such as ocular surface and oral fluids/surface have high potential for information as many sensor-based wearable devices have been developed.

Exposomics is another area that has a great potential to use such devises, especially for evaluation of hazards arising from occupational and environmental exposure to various chemicals that are detrimental to human health. There are already some early studies in this area using extractive wearable devices to investigate human exposure to environmental chemicals. However, more research is needed, and wearable extractive devices will be one of the main tools in the analytical toolbox for exposomic studies.

Moreover, such devices will facilitate the development of protocols for hazard analysis at critical control points. One of these areas would be pesticide control in agricultural products. Because of their frequent uncontrolled use in agricultural products, pesticide residues accumulate in food which often is associated with various health problems. There are many methods developed for pesticide analysis; however, there is still a high demand for novel and reliable methods applicable to multi-residue analyses on-site. Such extractive devices can be used for pesticide profiling on fruits and vegetables during the growth which will allow regulation of the safe amount of pesticides used in the production.

There is also a reasonable demand for wearable devices with different selectivity targeting specific needs. In the near future, the current developments in nanotechnology and nanostructured materials and the synthesis of novel extractive phase will allow the properties of materials to be tuned towards enhancing the selectivity towards targeted analytes. Potentially promising smart selective materials can be summarized as, but not limited to, molecularly imprinted polymeric materials, aptamer-based surface modifications, and metal-organic frameworks.

Finally, more work will be done in future for coupling these extractive wearable devices with portable instrumentation for direct sampling and monitoring on site.

**Author Contributions:** A.A.N.: Literature review, writing and revisions. E.B.: Editing, review and revisions.

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

**Acknowledgments:** The authors would like to thank Daniel Rickert for his help in the proofreading and editing of the manuscript.

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

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Modern Approaches to Preparation of Body Fluids for Determination of Bioactive Compounds**

#### **Katarzyna Madej 1,\* and Wojciech Piekoszewski 1,2**


Received: 3 July 2019; Accepted: 11 October 2019; Published: 5 November 2019

**Abstract:** The current clinical and forensic toxicological analysis of body fluids requires a modern approach to sample preparation characterized by high selectivity and enrichment capability, suitability for micro-samples, simplicity and speed, and the possibility of automation and miniaturization, as well as the use of small amounts of reagents, especially toxic solvents. Most of the abovementioned features may be realized using so-called microextraction techniques which cover liquid-phase techniques (e.g., single-drop microextraction, SDME; dispersive liquid–liquid microextraction, DLLME; hollow-fiber liquid-phase microextraction, HF-LPME) and solid-phase extraction techniques (solid-phase microextraction, SPME; microextraction in packed syringes, MEPS; disposable pipette tip extraction, DPX; stir bar sorption extraction, SBSE). Some other extraction methodologies like dispersive solid-phase extraction (d-SPE) or magnetic solid-phase extraction (MSPE) can also be easily miniaturized. This review briefly describes and characterizes the abovementioned extraction methods, and then presents their current applications to the preparation of body fluids analyzed for bioactive compounds in combination with appropriate analytical methods, mainly chromatographic and related techniques. The perspectives of the analytical area we are interested in are also indicated.

**Keywords:** microextraction techniques; body fluids; bioactive compounds; clinical and forensic analysis

#### **1. Introduction**

Body fluids belong to the most often analyzed samples in clinical investigations, and in related fields such as forensic toxicology. Plasma/serum, whole blood, urine, and saliva constitute biological materials of special interest.

Due to complexity of the matrix, body fluids are a challenge for those who are interested in bioanalysis. In most cases of such analyses, the selection of an appropriate sample preparation method is required because of main factors such as the chemical nature of the analyte and type of sample matrix, the analytical interferences of the target compounds with their metabolites and/or with constituents of the matrix, the low concentration levels of analytes, which are not detectable by analytical instruments, and the contamination and subsequent shortening of life of the used instrument or its accessories (e.g., due to interaction of matrix constituents with sorbents of chromatographic columns).

At present, conventional extractions techniques, i.e., liquid–liquid extraction (LLE) and solid-phase extraction (SPE), are still extensively applied to the analysis of body fluids for determination of bioactive compounds [1]. Although they demonstrate many advantageous features, they also possess some drawbacks, e.g., time consumption, large consumption of toxic organic solvents and possible formation of emulsion (LLE), or the requirement of relatively expensive extraction columns and multi-step processes (SPEs) [1]. Therefore, the development of alternative sample preparation methods is desired. The development of novel sample preparation approaches is strongly stimulated by the following

requirements of modern clinical or forensic analysis: (a) high selectivity and enrichment capability, (b) simplicity and rapidity, (c) suitability for micro-amount biological samples, (d) automation, and (e) miniaturization [2]. The other key factor affecting the dynamic development of extraction methods is the introduction of new materials, which may be used as potential sorbents in bioanalysis. There are MOFs (metal–organic frameworks), [3] ionic liquids [4], fabric phases [5], graphene oxide tablets [6], cork [7], bract [8], or coated papers [9].

This review focuses on modern approaches to the preparation of body fluid samples. The principles of microextraction techniques are briefly described and then exemplified using selected applications for the preparation of body fluids analyzed for medicines, drugs of abuse, and other bioactive compounds.

#### **2. Modern and Novel Sample Preparation Techniques**

There are several sample treatment methodologies that may be applied to the preparation of body fluids analyzed for bioactive compounds. The selection of the below sample preparation approaches was mainly based on the consideration of efficient and cost-effective miniaturized techniques, so-called microextraction methods. Generally, microextraction methods may be divided into two groups: Solid-phase extractions and Liquid-phase extractions.

#### *2.1. Solid-Phase Extractions*

#### 2.1.1. Solid-Phase Microextraction and Derived Techniques

A solvent-free solid-phase microextraction (SPME) was proposed by Arthur and Pawliszyn in 1990 [10], and then, in 1993, Supelco introduced the first commercial version of an SPME device [11]. The SPME technique, like conventional SPE, involves the partitioning of the target compound between an organic phase coating a fiber (usually made of fused silica) and the sample matrix. This extraction methodology may be conducted into two ways: directly by placing a small-diameter fiber coated with a stationary phase film in an aqueous sample or in headspace mode, by placing the fiber on the headspace of the sample. The adsorbed/absorbed analytes are then thermally desorbed from the stationary phase in the injector of a gas chromatograph. However, a new trend in developing faster analytical procedures involving SPME, i.e., direct coupling of this extraction approach with mass spectrometry, was noted [12]. Although, the SPME technique is known to have some drawbacks (fragile coating layers, degradation of the fibers with multi-use, carryover problems, batch-to-batch variations of fiber coatings, and relatively high cost), currently, it is dynamically developing, mainly through the introduction of new SPME fiber materials and selective coatings [2].

The SPME approach may be also performed "in-tube", where an open tubular fused-silica capillary column is exploited as the SPME device instead of an SPME fiber [13]. A comparison of fiber and in-tube SPME techniques in combination with liquid chromatography was also provided [13]. In-tube SPME is superior to fiber SPME mainly due to the shorter equilibrium time and better suitability for automation, as well as a wider diversity of stationary phases coating commercial capillary columns.

Thin-film microextraction (TFME) constitutes another geometry for SPME, which was proposed by Jiang and Pawliszyn [14]. In TFME, a sheet of flat film, employed as the extraction phase, is reinforced by an extra support such as a stainless-steel rod, stainless-steel mesh, or blade-shaped substrate [15]. Due to the large surface area-to-volume ratio, TFME is superior to SPME in terms of adsorption capability and extraction rate. This technique, in combination with novel biocompatible matrix membranes, may be especially useful in the medicine field as a non-invasive diagnostic tool in the analysis of breath, skin, and saliva.

#### 2.1.2. Microextraction in Packed Syringe

Microextraction in a packed syringe (MEPS) is a relatively new technique because it was first applied in a fully automated procedure by Abdel-Rehim in 2004 [16] to extract local anesthetics from human plasma. The MEPS technique is a miniaturized version of conventional solid-phase extraction (SPE), which can be connected online to GC or LC, without modification [16]. In MEPS, 1 mg of the sorbent material is usually placed into a syringe (100–250 μL) as a plug. A sample is drawn through the syringe (e.g., by an autosampler) before passing through the solid material where analytes are adsorbed. The sorbent is washed to remove the biological matrix, and then the analytes are eluted with an appropriate organic solvent (e.g., directly into the instrument's injector). In comparison with conventional SPE, the main advantages of MEPS are congruence for full automation (which minimizes the sample preparation time to a minute), as well as the possibility to reuse a packed syringe (more than 100 times with body fluid samples), whereas an SPE column usually can only be used once. Compared with SPME, the MEPS technique is more powerful for the preparation of biological samples with complex matrices. Moreover, much higher extraction yields can be obtained (60–90%) compared to standard SPME (1–10%), and smaller sample volumes can be handled (10 μL) compared to SPME (>1000 μL) [16].

#### 2.1.3. Disposable Pipette Extraction

Disposable pipette extraction (DPX, TIPS, PT-SPE) is a miniaturization of conventional SPE, which was developed by Brewer [17]. In DPX, the sorbent is placed in a pipette tip and it is mixed well with the sample. This reduces the sorption material needed to retain analytes, and leads to faster and more efficient extraction process [18]. Similarly to conventional SPE or MEPS, a four-step extraction process takes place: (1) conditioning for activation of the sorbent sites, (2) aspiration of the sample, (3) removing sample matrix interferences, and (4) elution of analytes. In the steps 1, 3, and 4, an appropriate solvent (or solvents) is aspirated with air and removed one or more times. After conditioning (step 1), the sample is suctioned with air (step 2) for mixing with the adsorbent. The mixing/contact time must be controlled to reach a dynamic equilibrium during the interaction of analytes with the sorbent, and then the sample is removed from the tip. DPX is a simple and fast sample preparation approach, which minimizes the consumption of the sample and organic solvents. Recently, some miniaturized concepts for chromatographic analysis involving a pipette tip and spin column were reviewed [19]. However, until now, their application to routine analysis is limited mainly due to the small number of commercially available extracting materials and the higher cost compared to traditional SPE columns [20].

#### 2.1.4. Dispersive Micro-Solid-Phase Extraction and Magnetic Solid-Phase Extraction

Dispersive micro-solid-phase extraction (D-μ-SPE) is a miniaturized dispersive solid-phase extraction (d-SPE) consisting of the dispersion of micro- or nanosorbents in the sample solution, followed by separation of the solid sorbent from the extracted analytes by centrifugation and filtration [21]. Among the advantageous features in favor of the D-μ-SPE technique in relation to conventional SPE are rapidity (due to the dispersion phenomenon which increases the sorbent surface for interaction with analyte molecules), possibility to use a large spectrum of sorbents (which are able to disperse in the sample solution), simplicity, and reducing costs in terms of the used sorbent amounts and laboratory equipment. One of the d-SPE variants is magnetic solid-phase extraction (MSPE), where magnetic nanoparticles (MNPs) are used as sorbents in the sample preparation process. MNPs attract the consideration of many analytical chemists because they considerably simplify and accelerate the extraction process due to their capacity to be easily isolated from the sample solution through an external magnet [2]. Regarding the trend of MNPs to form agglomerates and the loss of magnetism due to their chemical activity, these materials are modified into core–shell composites by coating them with an organic (e.g., surfactant) or an inorganic (e.g., graphene or carbon nanotubes) layer. The magnetic composites created, which are used as sorbents, may also have a higher extraction ability than simple magnetic nanoparticles (e.g., Fe3O4) when samples with a complex matrix are analyzed. A recent general review of D-μ-SPE [22] focused mainly on the dispersion strategies and sorbents used, and also indicated several trends in the near future for the development of this extraction technique.

#### 2.1.5. Stir Bar Sorption Extraction

Stir bar sorption extraction (SBSE) was introduced by Baltussen with his co-workers in 1999 [23]. In this approach, a stir bar (magnetic element) is coated with a sorbent and immersed in a sample solution. The sample is stirred with appropriate speed for a specified time to reach equilibrium. After adsorption of the analyte on the sorption material, the magnetic element is transferred to a small amount of a selected organic solvent to desorb the analyte into it. Although SBSE is theoretically similar to SPME, its capacity is greater than SPME. This results from the fact that more sorbent mass is usually present in SBSE than in SPME and more analyte is transferred to the sorbent in SBSE. Currently, polydimethylsiloxane (PDME) is mainly employed as the sorbent coating, and the SBSE technique is not used very often. However, the current trend is to employ this technique in combination with many novel materials, including restricted access materials, carbon adsorbents, molecularly imprinted polymers, ionic liquids, microporous monoliths, sol–gel prepared coatings, and dual-phase materials [24]. The main advances in various "extraction/stirring integrated techniques", including the source SBSE technique, emphasizing their analytical potential, were also previously reviewed and compared [25].

#### *2.2. Liquid-Phase Extractions*

#### 2.2.1. Single-Drop Microextraction

The term "liquid-phase microextraction (LPME)" means that the extraction technique consists of a microsyringe playing two roles: a funnel for extraction and a syringe for injection into a GC port. Single-drop microextraction (SDME), the simplest mode of LPME, was introduced by Liu and Dasgupta [26], as well as by Jeannot and Cantwell in 1996 [27]. In the SDME technique, analytes are isolated from an aqueous sample (being stirred) into a small (ca. one μL) drop of a water-immiscible organic solvent hanging on the needle of a microsyringe [28]. After the extraction process, the drop is withdrawn into the syringe and usually injected directly into a GC instrument. The SDME technique may be realized via three ways: through direct immersion, in the headspace mode, and as three-phase SDME. In contrast to the direct immersion mode, where the extracting solvent drop is submerged into the aqueous sample, in the headspace mode, the drop hangs in the headspace of the sample. In three-phase SDME, the droplet contains two immiscible solvents, i.e., the polar acceptor solution and the polar donor solution. These two solvents are separated by a non-polar solvent. The major problem of the SDME technique is drop instability, but attempts are being made to solve this issue via the creation of new devices for SDME. Current trends in developments in SDME, such as the use of non-conventional solvents, novel materials (for combining sorbent and liquid-phase extractions), and the creation of more suitable SDME devices, as well as their automation and implementation in microfluidic chip technologies, were presented in a recent review paper [29].

#### 2.2.2. Dispersive Liquid–Liquid Microextraction

Dispersive liquid–liquid microextraction (DLLME) was proposed by Rezaee and co-workers in 2006 [30]. This technique is based on a ternary-component solvent system consisting of an extraction solvent, a dispenser, and an aqueous sample. In DLLME, a mixture of extraction and dispenser solvents is rapidly injected into an aqueous sample via a syringe. The extraction solvent is dispersed into the aqueous phase and forms a cloudy solution of fine droplets, which interact with an analyte. The centrifugation allows the separation of two phases, whereby the sediment phase is enriched with the analyte. The extraction solvent must be a high-density water-immiscible solvent, whereas the dispenser solvent must be a polar water-miscible one. Carbon tetrachloride, tetrachloroethylene, chlorobenzene, and ionic liquids may be used as the extraction solvent, while acetone, methanol, ethanol, acetonitrile, or tetrahydrofuran may play the role of dispensers. One of the more interesting varieties of the DLLME technique involves the solidification of a floating organic drop (DLLME-SFO) [31]. In DLLME-SFO, dodecanol or 2-dodecanol are usually used as extraction solvents. After centrifugation, the floating

organic phase is solidified quickly by cooling (e.g., in an ice bath). The solidified extraction solvent with the isolated analytes is separated, melted at room temperature, and then subjected to analysis.

#### 2.2.3. Hollow-Fiber Liquid-Phase Microextraction

The idea of a supported liquid membrane (SLM) was for first time integrated with single-use extraction units for liquid–liquid–liquid microextraction (LLLME) by Pedersen-Bjergaard and his co-workers in 1999 [32]. The authors employed a polypropylene hollow fiber as the membrane for extraction of the model compound (methamphetamine) from body fluids. In SLM, an organic solvent (e.g., 1-octanol) is usually impregnated in the small pores of a hollow fiber, which protects the extracting solvent, thus permitting extraction only on the surface of the solvent immobilized in the membrane pores. Hollow-fiber liquid-phase microextraction (HF-LPME), as a mode of liquid-phase microextraction, was firstly published by Shen and Lee in 2002 [33]. To impregnate the pores of the fiber wall with an appropriate solvent, the needle tip inserted into the hollow fiber is immersed in an organic solvent for several minutes. In order to remove the excess organic solvent from the inside of the fiber, water is injected to flush it before being removed from the solvent. For the extraction, the prepared fiber is immersed in an aqueous sample while the organic solvent in the syringe is injected completely into the hollow fiber. During the extraction, the solution is agitated using a magnetic stirrer. After the extraction process, the solvent with the isolated analyte is withdrawn into the syringe and subsequently injected into a GC instrument. The disadvantages of HF-LPME include the relatively low repeatability of the extraction process due to the air bubbles forming on the hollow-fiber surface (reduction of transport speed), and the formation of a membrane barrier between the sample and the acceptor solvent (usually an organic solvent), reducing the rate of extraction [30]. Electromembrane extraction (EME) is an extended concept of HF-LPME, which was introduced in 2006 [34]. In the EME approach, a charged analyte is extracted from a sample solution through the SLM into an acceptor solution, where its transfer is facilitated by an external electrical field applied across the SLM [35]. A review article concerning the EME technique, covering its principles, new SLMs, support materials for the SLM, new sample additives, and novel technical configurations, as well as the applications of EME in pharmaceutical analysis, was published [35].

#### **3. Applications**

Data concerning exemplary applications of solid-phase extractions and liquid-phase extractions for clinical and forensic analysis are gathered in Tables 1 and 2, respectively.





*Separations* **2019**, *6*, 53





*Separations* **2019**, *6*, 53

dispersive liquid–liquid

microextraction.

#### *3.1. Solid-Phase Extractions*

The SPME process may be fully automated and effectively accelerated using the 96-well plate format of thin-film solid-phase microextraction (TFME). A PPy–CH2–COOH (nanostructured α-carboxy polypyrrol) coating for SPME fiber, used in headspace mode (HS-SPME), was synthetized and applied to the extraction of methadone at its trace level from plasma and urine samples [36]. The carboxy end-capped polypyrrole film was electrochemically deposited on a platinum wire, and a nano-fibrous structure with a diameter of 120 nm was obtained. The nanostructure of the film provided a high surface area that allowed for the high extraction efficiency of methadone. The extraction efficiency was 95–97%. Advantageous features of the used nanostructured fiber include high mechanical stability, strong adhesion of the coating to the substrate, fast extraction equilibrium and desorption, and relatively low cost.

Eight beta-blockers and bronchodilators were isolated from body fluids (plasma and urine) employing the 96-well plate format of the TFME system and an extraction phase made of hydrophilic–lipophilic balance particles (HLB) [37]. Methanol–acetonitrile (80:20 *v*/*v*) with 0.1% formic acid was used as a desorption solvent, which was compatible with the used mobile phase of LC–MS/MS method. The developed extraction method described in the form of a protocol is fast (time of full preparation of one sample less than 2 min), automated, and efficient. The authors also emphasized that the protocol could be modified through compromising sensitivity; finally, the proposed extraction approach could be shortened to 10–15 min for all 96 samples.

The combination of LC–MS/MS with SPME, involving polymeric sorption coatings with molecular imprints (MIP), was used for the simultaneous determination of 10 antibiotic drugs in-whole blood samples [38]. Three conducting polymers, polypyrrole, polythiophene, and poly(3-methylthiophene), used as coatings in SPME, were selected for the extraction of the target compounds. Optimization of the extraction conditions included parameters such as the extraction time, kind of desorption solution, and pH of organic extraction solvent. The obtained results showed that the tested antibiotics could be divided into three groups due to sorption capacity and selectivity to SPME coatings used and, thus, the appropriate sorbent was chosen for each group of analytes. According to the authors, the proposed MIP–SPME–LC–MS/MS method is suitable for therapeutic drug monitoring (TDM) of antibiotics in clinical laboratories, as well as in forensic laboratories for assays at higher concentration levels in body fluids.

Two modes of solid-phase microextraction (in-tube and thin-film SPME) were applied to the estimation of hormone (or similar substances) levels in saliva [38] and urine [39]. Testosterone, cortisol, and dehydroepiandrosterone were assayed by online in-tube SPME in combination with LC–MS/MS [39], and estrogens( estrone, 17β-estradiol, 17α-ethinylestradiol, and estriol) were determined employing a combination of thin-film SPME (TF-SPME) with a 96-well plate system and HPLC–FLD [8]. In the second case [39], a biosorbent—bract—as a novel extraction phase for TF-SPME, was used. In both cases, the methods were fast, with a high throughput and low sample consumption (100 μL of saliva and 37 μL of urine), and only ultrafiltration of samples was required prior to analysis.

Microextraction in packed sorbent (MEPS) is a new sample preparation technology, used in combination with liquid chromatography–mass spectrometry [40] or mass spectrometry alone [41]. An automated sample work-up and quantification of four immunosuppressive drugs in whole blood, using combined MEPS with LC–MS/MS, was presented [40]. The proposed method required 50 μL of whole-blood sample. The obtained quantitative results were in good agreement with a reference LC–MS/MS method involving protein precipitation. The analytical parameters of the developed method revealed that it can be useful for TDM of the studied immunosuppressive drugs.

The MEPS technology was also used in a fast screening of cocaine and its metabolites in human urine samples examined by direct analysis in a real-time source coupled to time-of-flight mass spectrometry (DART–TOF) [41]. As the type of sorbet is one of the most important parameters in solid-phase extraction, four various adsorbent materials (C8, ENV+, Oasis MCX, Clean Screen DAU) were studied. The last material (Clean Screen DAU) worked best, showing satisfactory extraction efficiency for all studied analytes. In the described sample preparation method, a few microliters of the sample were used, and the extraction time was less than 2 min. Finally, the authors stated that coupling

MEPS technology to an autosampler and connecting it with DART/TOF mass spectrometer would allow for its full automation, making it a very useful tool for screening drugs of abuse in biological matrices.

Developing a new approach for antibiotic drug (linezolid and amoxicillin) determination in human plasma by LC/UV–MS/MS, three sample preparation methods were compared [42]. The pretreatments of samples were performed using protein precipitation (PP), solid-phase extraction (SPE), and microextraction in a packed syringe (MEPS). Among three studied extraction methods, MEPS appeared to be superior to PP and SPE, in terms of accuracy and precision.

A bio-inspired sponge, an amino-functionalized metal–organic framework (Zr–MOF–NH2), was successfully applied as a sorbent in PT-SPE to extract carbamazepine from urine samples [43]. This extraction method was combined with the HPLC–UV method. The best extraction conditions were achieved when the sample volume was 100 μL, with sample pH adjusted to 7.5, and 5 mg of the sorbent and 10 μL of methanol as eluent solvent were used. The total time of analysis, including the sample preparation step, was less than 12 min. The sorbent was used for at least eight extractions without significant loss of its capacity and extraction repeatability.

Two automated clean-up methodologies, based on monolithic packed 96-tip sets and combined with LC–MS/MS, were employed for determination of the beta-blockers (pindolol and metoprolol) in plasma [44] and anti-cancer drugs (cyclophosphamide and busulfan) in whole-blood samples [45]. In the case of the beta-blocker analysis, a 100-μL sample volume was handled, and, for the determination of the anti-cancer drugs, 1 mL of a sample was needed. In both cases, sample preparation was performed in only about 2 min for 96 samples. However, in the case of whole-blood analysis, the monolithic packed 96 tips can be re-used up to only five times, inferior to the 100 times achieved with MEPS methodology [45]. Comparing the determination of the beta-blockers in plasma, using monolithic packed 96 tips, with the protein precipitation method (using 0.1% formic acid in acetonitrile), the tips method resulted in a higher (2–3 times) S/N ratio. Generally, the tips methodology provided better results in terms of selectivity, accuracy, and precision.

A d-SPE protocol based on the modified QuEChERS procedure (using 50 mg of PSA and 150 mg of anhydrous MgSO4), followed by large volume injection programmed temperature vaporization (LVI–PTV)–gas chromatography–mass spectrometry (GC-MS) analysis, was developed for the simultaneous determination of 16 drugs and pesticides in postmortem blood samples without derivatization [46]. The validated d-SPE/LVI–PTV/GC–MS method is suitable for routine analysis in a forensic laboratory. The usability of the proposed method was confirmed by the analysis of 10 postmortem blood samples. Six samples contained cocaine, two contained MDMA, and two contained carbamazepine. Other found analytes were carbofuran, the metabolite 7-aminoflunitrazepam, amitriptyline, and diazepam.

Magnetically modified nanomaterials based on conductive polymers [47] and molecularly imprinted polymers (MIPs) [48] can be used as examples of new sorbents widely employed in MSPE for preconcentration and determination of pharmaceuticals in a complex biological matrix. Polypyrrole (PPy) with magnetic nanoparticles (MNPs)—Fe3O4—doped by sodium perchlorate (NaClO4), exhibited high extraction efficiency: 93.4–99% and 94–98.4% for citalopram (CIT) and sertraline (STR), respectively. The isolation of the medicines was performed in the optimized extraction conditions: sample pH—9.0, sorbent amount—10 mg, sorption time—7 min, elution solvent (0.06 mol/L HCl in methanol) volume—120 μL, elution time—2 min. The performance of the developed extraction method was estimated by spiking CIT and STR at trace levels in urine and plasma samples [47].

Molecularly imprinted magnetic carbon nanotubes (MCNTs@MIPs) were applied to the MSPE of levofloxacin in serum samples [48]. The prepared sorbent was characterized by high specific surface area and high selectivity to template molecules of levofloxacin. Investigations indicated that the sorbent was very selective (exhibited excellent recognition) toward levofloaxacin. Under optimal extraction conditions (sorbent amount—15–20 mg, adsorption and desorption time—60 min, eluent—methanol/acetic (6:4, *v*/*v*), eluent volume—2 mL), extraction recoveries ranged from 78.7 ± 4.8% to 83 ± 4.1%. The sorbent stability was assessed as a reduction in average recovery after five cycles and was estimated at less than 7.6%.

A validated high-performance liquid chromatography with fluorescence detection (HPLC–FLD) method combined with stir bar sorption extraction (SBSE) was developed for determination of fluoxetine in human plasma [49]. The extractions of this antidepressant drug from plasma samples were performed using laboratory-made polydimethylsiloxane (PDMS) stir bars. Several factors such as temperature and time of sorption, stirring speed, and two modes of desorption (ultrasonic and magnetic stirring) were considered in the optimization of extraction conditions. The method was successfully applied to the analysis of real plasma samples originating from depressed patients treated with fluoxetine. Taking into account the time of sample preparation (3 h), low cost of each coated bar, and possibility to reuse it 50 times, the authors recommended the proposed method as a reliable tool for routine analysis. Novel coatings for bars in the SBSE approach were introduced for body fluid preparation for the determination of antihypertensive drugs [50] and illicit drugs [51]. A monolithic vinylpyrroli–doneethylene glycol dimethacrylate (VPD-EDMA) polymer was applied to the extraction of losartan and valsartan from plasma samples [50]. Compared to commercially available PDMS (polydimethylsiloxane) and PA (polyacrylamide) stir bars, the proposed coated stir bar exhibited higher physical stability and, therefore, it was suitable for the use of an ultrasonic stirring mode, which was profitable in terms of requiring less time and less solvent for the performed extraction. A nano graphene oxide sol–gel composite (NGO/sol–gel), used as the coating of a capillary glass tube stir bar, was employed for the isolation of amphetamine and methamphetamine from urine samples [51]. In both cases, the coated bars were coupled with HPLC in combination with relatively low-sensitivity UV detection, but the obtained results indicated the usefulness of both mentioned methods for the analysis of real samples in clinical and forensic laboratories.

#### *3.2. Liquid-Phase Extractions*

Jahan and co-workers [52] proposed a coupling microextraction device combining a microsyringe with a very short capillary. The designed device enabled manually controlling the shape and size of the aqueous organic droplet during the extraction process, as well as ensuring droplet stability even under vigorous stirring conditions. The performance of the developed method was checked using human serum samples spiked with five statins (lovastatin, simvastatin, mevastatin, Fluvastatin, and atorvastatin). Using 1.2 μL of a toluene–aqueous (0.2:1) droplet, a 350–1712-fold enrichment of the statins was achieved within four minutes.

A combination of SDME and open tubular capillary electrochromatography (OT–CEC) for the simultaneous determination of seven illicit drugs (caffeine, cocaine, ephedrine, morphine, piroxicam, strychnine, and theophylline) in horse urine was presented [53]. The extraction procedure enabled good compound recovery with good precision and appropriate reduction of interferences. The obtained enrichment factors were between 38 and 102 depending on the studied compound. The proposed SDME–OT–CEC method showed potential for application in toxicology investigations, as well as being a cheaper and more environmentally friendly method compared to LC–MS methods.

SDME was also used on-line coupled with sweeping micellar electrokinetic chromatography (MECK) for the isolation and preconcentration of three alkaloids: berberine, palmatine, and tetrahydropalmatine, present in human urine at trace level [54]. In this method, analytes were firstly extracted from a basic aqueous sample solution (donor phase) into *n*-octanol, and then back-extracted into the acidified aqueous solution (acceptor phase), which formed a droplet at the tip of a capillary. The acceptor phase was introduced into the capillary using the hydrodynamic injection mode and analyzed by sweeping MEKC. The extraction method was optimized considering a number of experimental factors: extraction solvent type, time of drop formation, stirring rate, duration of pre- and back-extraction, sample temperature, addition of NaCl, and composition of donor and acceptor phase. Under the optimized conditions, the developed method gave a 1583–3556-fold improvement in sensitivity for the studied analytes within 20 min.

Organic solvent dispersive liquid–liquid microextraction (OS–DLLME) and ionic liquid dispersive liquid–liquid microextraction (IL–DLLME) were applied to determine emodin and its six metabolites in urine samples. These two extraction approaches were evaluated and compared [55]. The analytes were assayed using HPLC–UV and HPLC–MS methods. In order to optimize DLLME, several parameters were studied and optimized, including kind of extraction solvent, volumes of the extraction solvent and the disperser solvent, extraction and centrifugation time, sample pH, and concentration of added NaCl. At optimal extraction conditions, the enrichment factors for emodin and its metabolites ranged from 90–295 for the OS–DLLME method and from 63–192 for the IL–DLLME method. Comparing the abovementioned extraction methods, it was concluded that IL–DLLME was more rapid and simple, as well as more repeatable, than OS–DLLME. However, OS–DLLME exhibited higher enrichment factors, a wider linear range, and lower limits of detection for the target compounds.

A one-step in-syringe set-up for DLLME with the use of an ionic liquid was proposed, and pros and cons of the novel approach were discussed [56]. Among the advantages may be listed the use of a simple extraction unit (only a conventional plastic syringe), avoiding the centrifugation step, reducing the extraction time, and being more suitable for the automation of the whole extraction procedure than in the conventional approach. The novel approach broadens the used solvent range (not only solvents denser than water can be exploited) by simply changing the orientation of the syringe during the phase separation step [56]. Among the limitations (at the present state), low extraction recoveries and low precision of the procedure may be mentioned. The suitability of the proposed approach was evaluated via the determination of four non-steroidal anti-inflammatory drugs (ketoprofen, naproxen, flurbiprofen, and indomethacin) spiked with urine samples. The obtained wide concentration range (0.02–10 μg/mL) of the target drugs allowed assaying them at therapeutic and toxic concentration levels.

DLLME in combination with HPLC–UV was applied to the determination of five antiarrhythmic drugs, metoprolol, propranolol, diltiazem, carvedilol, and verapamil, present in human plasma [57]. The method required 660 μL of plasma sample, and the complete separation of all the analytes was achieved within 7 min. In the extraction process, acetonitrile, used as a disperser solvent (resulting from the protein precipitation), was mixed with the extracting medium (dichloromethane) and then quickly injected into an aqueous basic solution. After centrifugation, the sedimented phase with the concentrated drugs was taken and evaporated to dryness. The dried residue was dissolved in 50 μL of deionized acidified water and subjected to HPLC analysis. The developed method was successfully applied to the analysis of the selected antiarrhythmic drugs in real plasma samples at ng/mL. According to the authors and considering the determined analytical parameters, the proposed method seems to be appropriate for routine analysis in drug analysis laboratories for pharmacokinetic and pharmacodynamics studies, as well as for therapeutic drug monitoring.

In hollow-fiber liquid-phase microextraction (HF-LPME), a porous membrane of the polypropylene hollow fiber with the pore size of 0.2 μm prevents the entry of macromolecular compounds in the sample solution into the acceptor phase present inside the fiber [58]. The HF-LPME technique may be realized using two modes: two-phase and three-phase. In two-phase HF-LPME, the target compound is enriched into the organic solvent as the acceptor phase and then determined by an appropriate chromatographic method. In three-phase HF-LPME, the analyte is extracted from the aqueous donor phase to organic phase, followed by back-extraction into the aqueous donor phase which is directly subjected to analysis. The applications of two-phase and three-phase HF-LPME for the isolation and preconcentration of hydrochlorothiazide (HYD) and triamterene (TRM) in urine samples were presented, respectively [58]. Under optimal conditions, enrichment factors of 128 and 239 were achieved for HYD and TRM, respectively. Finally, the proposed method was successfully applied to the analysis of TRM and HYD in a urine sample obtained from a volunteer who received 500 mg of triameterene-H. The performed analysis showed that the concentrations of HYD and TRM were 18.2 and 7.3 μg/L, respectively.

Using HF-LPME in the three-phase mode combined with GC–MS, ketamine (KT) and its main metabolites, norketamine (NK) and dehydronorketamine (DHNK), were determined in urine samples [59]. A "green chemistry" approach to the sample extraction involved using eucalyptus essential oil as a supported liquid membrane in HF-LPME. After drying, the acceptor phase with the isolated analytes was derivatized with trifluoroacetic anhydride before analysis by GC–MS. The usefulness of the developed method was confirmed by the analysis of real urine samples originating from two patients who were suspected of taking KT. KT, NK, and DHNK were determined in the concentrations 0.0873, 5.805, and 8.760 μg per 1 mL of urine, respectively (in one case) and 7.3, 5.3, and 6.8 ng per 1 mL of urine, respectively (in the second case).

A novel generation of deep eutectic solvent (DES) as an acceptor phase in three-phase HF-LPME was introduced for the extraction of two steroidal hormones, dydrogesterone (DYD) and cyproterone acetate (CPA), from urine and plasma samples [60]. The following factors influencing extraction efficiency were studied: nature and composition of DES, composition of supported liquid membrane, salt addition, length of hollow fiber, stirring rate, and duration of extraction. Under the optimized conditions, preconcentration factors ranged from 187 to 428. The method was validated via the analysis of real urine samples collected from two patients. In the case where patient 1 consumed one tablet (containing 2 mg of cyproterone acetate +0.035 mg of ethylene estradiol), the analysis showed that the medicine concentration was below the detection limit of the method. However, positive results of the analysis were achieved in the case of patient 2, who was subjected to hormonal therapy for a long time. The usefulness of the developed method was also confirmed via analysis of plasma samples spiked with 5 and 100 ppb of DYD and 10 and 200 ppb of CPA.

#### **4. Summary and Conclusions**

Microextraction methods constitute modern approaches for sample preparation in terms of efficiency, selectivity, simplicity, speed, enrichment capability, requirement of small sample volume, and congruence for automation and miniaturization. Due to low or very low (usually on the order of several hundred microliters) or no consumption (SPME combined with thermal desorption of analytes in GC injection port) of organic solvents, these techniques may be also classified as "green" approaches. The eight microextraction techniques belonging to solid-phase (SPME, MEPS, DPX, D-μ-SPE/MSPE, SBSE) and liquid-phase (SDME, DLLME, HF-LPME) extractions were briefly described and characterized. The usefulness of these techniques in the preparation of biological samples was supported by selected examples of their applications for the determination of medicines, drugs of abuse, and other bioactive compounds in body fluids. In the most reported cases, extraction recoveries were over 80% with an acceptable RSD lower than 10–15%. To quantify the studied bioactive compounds, the proposed microextraction techniques were usually combined with an appropriate chromatographic or capillary electrophoretic method. Using such combinations, limits of detection and quantification were at concentration levels equal to ppb or ppm. Considering sample consumption, solid-phase extractions generally require lower volumes, especially MEPS and DPX techniques, where sample volume consumption range from 5 to 600 μL. It is also worth noting that some proposed microextraction methods, at the optimized conditions, may be effective enough (they have a high enough enrichment capacity) to achieve LODs at low ppm levels of the studied compounds, even in combination with a relatively low-sensitivity UV detection method.

Microextraction techniques began being introduced during the last decade of the last century, and, based on the growing number of publications associated with them, it seems that they represent the current direction of development of sample preparation methods. Considering their numerous advantages, including "green" characteristics, without loss of analytical performance, it seems that these approaches will be extensively developed and employed in many important analytical areas, such as clinical and forensic investigations, as well as in environmental and food analysis.

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

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

#### **References**

1. Niu, Z.; Zhang, W.; Yu, C.; Zhang, J.; Wen, Y. Recent advances in biological sample preparation methods coupled with chromatography, spectrometry and electrochemistry analysis techniques. *TrAC Trends Anal. Chem.* **2018**, *102*, 123–146. [CrossRef]


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Evolution of Environmentally Friendly Strategies for Metal Extraction**

**Govind Sharma Shyam Sunder 1, Sandhya Adhikari 1, Ahmad Rohanifar 1, Abiral Poudel <sup>1</sup> and Jon R. Kirchho**ff **1,2,3,\***


Received: 9 October 2019; Accepted: 3 December 2019; Published: 6 January 2020

**Abstract:** The demand for the recovery of valuable metals and the need to understand the impact of heavy metals in the environment on human and aquatic life has led to the development of new methods for the extraction, recovery, and analysis of metal ions. With special emphasis on environmentally friendly approaches, efforts have been made to consider strategies that minimize the use of organic solvents, apply micromethodology, limit waste, reduce costs, are safe, and utilize benign or reusable materials. This review discusses recent developments in liquid- and solid-phase extraction techniques. Liquid-based methods include advances in the application of aqueous twoand three-phase systems, liquid membranes, and cloud point extraction. Recent progress in exploiting new sorbent materials for solid-phase extraction (SPE), solid-phase microextraction (SPME), and bulk extractions will also be discussed.

**Keywords:** metal extraction; liquid–liquid extraction; solid-phase extraction; solid-phase microextraction; green extraction methods

#### **1. Introduction**

Metals are ubiquitous in nature serving as essential elements for human health and critical materials for modern industrialization and urbanization. While some metals such as iron are necessary for human health, many metals are toxic, and can cause physical problems such as diarrhea, nausea, asthma, kidney malfunction, different cancers, and even death [1]. Arsenic, cadmium, chromium, mercury, and lead are commonly known as heavy metals—or metalloids in the case of arsenic—and have the greatest toxicity. The maximum limits in drinking water for these metal ions according to the World Health Organization (WHO) are 10, 3, 50, 6, and 10 μg L<sup>−</sup>1, respectively [2]. The harmful effect of arsenic can mostly affect skin, respiratory, and cardiovascular systems. Elevated risk of skin and lung cancers has been reported among people who were exposed to arsenic from working in mining and smelting areas where inorganic arsenic was inhaled [3]. Cadmium and lead are harmful for the nervous system. Mercury used in electrical devices, dental fillings, Hg vapor lamps, solders, and X-ray tubes has a strong attraction to biological tissues and is carcinogenic, mutagenic, and teratogenic [4]. The Flint water crisis in 2014 affected about 100,000 people when lead from aging pipes leached into the water supply and contaminated the drinking water. This poignant example illustrates the importance of careful monitoring of heavy metals (HMs) in water systems and investigating new technologies to extract and remove them [5,6].

Other metals such as cobalt, copper, iron, and zinc have higher threshold limits. The maximum limit for copper in drinking water is 2 mg L−<sup>1</sup> according to the WHO [2]. No guideline values are provided for iron and zinc in drinking water, however, high concentrations of these elements may still cause adverse health effects or, at a minimum, an unacceptable taste for consumers [2]. The recovery, removal, and recycling of valuable metals, including gold, platinum, and rare earth elements, from natural and secondary sources such as industrial wastes is also important for their economic, strategic, and national security value. These critical elements have important applications in metallurgy and the biomedical and electronics industries [7–12].

Several methods have been used for extraction and removal of metals from different sources of water, including microfiltration [13], chemical precipitation [14], coagulation and flocculation [15], electrochemical removal [16], liquid–liquid extraction [17,18], osmosis [19], crystallization and distillation [20], photocatalysis [21], and adsorption. In this review, we focus on several techniques for extraction, determination, and removal of metals, including heavy and valuable metals, from water samples. In particular, extraction methods that aim to provide environmentally friendly, simpler and faster techniques are discussed. Approaches include recent advances in primarily liquid–liquid and solid-phase extraction. Comparison of their advantages and disadvantages will be made to illustrate efforts to develop more environmentally friendly methods.

#### **2. Liquid-Based Extraction**

Numerous liquid-based techniques like liquid–liquid extraction (LLE) [9,22–24], chemical precipitation [14,25], and cloud point extraction [26] have been utilized for extraction of metal ions from aqueous media. Among the listed methods, LLE is based on analyte partitioning between two immiscible phases. Conventional LLE is widely used for separations and preconcentration, including extraction and recovery of metal species from aqueous media by the addition of organic solvents [24]. This technique has significant advantages. These include rapid extraction kinetics, the ability to choose selective solvents, amenability to large-scale separation, and easy and flexible implementation. In spite of these advantages, traditional LLE has several drawbacks, including the extensive use of volatile and flammable organic solvents, which are potential health and environmental hazards [27]. Moreover, from the economic point of view, LLE is quite expensive because of the cost of organic extractants and their disposal [28]. These drawbacks can be circumvented by embracing new methods that provide simple, low-cost, fast, sensitive, and accurate analyses in a more environmentally friendly manner. Various advancements in liquid-based extraction for metal ions, such as aqueous biphasic and triphasic extraction, cloud point extraction, and liquid membrane extraction, are discussed herein.

#### *2.1. Aqueous Biphasic Systems*

Aqueous biphasic systems (ABS, Figure 1) forms when two immiscible aqueous-based solutions are mixed together at a certain temperature [29]. ABS have gained more attention for metal extraction since 1984 when Zvarova and co-workers successfully extracted copper, zinc, cobalt, iron, indium, and molybdenum using a polyethylene glycol (PEG) 2000–ammonium sulfate–water system in the presence of ammonium thiocyanate and sulfuric acid [30]. Because organic solvents are not required in ABS, it has several advantages over traditional solvent extraction. ABS are less toxic, more economical, biocompatible, and have a reduced environmental risk [31]. Furthermore, numerous inorganic anions can be used as water-soluble extractants resulting in metal ion partitioning between two immiscible aqueous phases, which reduces dehydration effects [32].

ABS can be formed by various mechanisms and thus are tunable to the desired extraction. Biphasic systems composed of polymer–polymer [33], polymer–salt [34], salt–salt [35], ionic liquid–salt [36], and surfactant-based systems [37] have been reported. In addition to these, other phase-forming elements are amino acids, alcohols, and carbohydrates. In ABS, factors governing the metal ion extraction include molecular weight and polymer type [38], Gibbs free energy of hydration [32], medium pH [39], presence and absence of an extracting agent [32,40], and temperature [41]. Examples of metal ion extractions using different ABS are given in Table 1.

**Figure 1.** Schematic representations of a two- versus three-phase system for metal ion extraction.




**Table 1.** *Cont.*

<sup>a</sup> octylphenolpolyethoxylene, <sup>b</sup> polyethylene glycol (average molecular mass 6000), <sup>c</sup> (ethylene oxide)11 (propylene oxide)16 (ethylene oxide)11, <sup>d</sup> poly(ethylene oxide), <sup>e</sup> (ethylene oxide)13-(propylene oxide)30-(ethylene oxide)13, <sup>f</sup> tetrabutylammonium bromide, <sup>g</sup> tributyl(tetradecyl)phosphonium, <sup>h</sup> tri(hexyl)tetradecylphosphonium chloride, <sup>i</sup> tri-*n*-butyl(carboxymethyl)phosphonium chloride, <sup>j</sup> 1-hexyl-3-methylimidazolium tetrafluoroborate, <sup>k</sup> tricaprylmethylammonium nitrate, <sup>l</sup> tetrabutylphosphonate, <sup>m</sup> 1-hexyl-3-methylimidazole dodecyl sulfonate, <sup>n</sup> 1-(2-pyridylazo)-2-naphthol, <sup>o</sup> 1-nitroso-2-naphthol, <sup>p</sup> tie-line length, <sup>q</sup> ammonium pyrrolidine dithiocarbamate, <sup>r</sup> inductively coupled plasma atomic emission spectrometry, <sup>s</sup> atomic absorption spectrophotometry, <sup>t</sup> fourier transform infrared spectrophotometry, <sup>u</sup> flame atomic absorption spectrophotometry, <sup>v</sup> inductively coupled plasma optical emission spectrometry, <sup>w</sup> total reflection X-ray fluorescence.

It is important to note that to extract a single target metal in each extraction step with a biphasic system, the extraction process for a specific metal from a mixture of metal ions must be highly selective, resulting in a potentially lengthy and costly method. A well-designed three-liquid-phase extraction system may overcome this disadvantage by selective separation and extraction of two or more targeted metals during a single extraction step.

#### *2.2. Three-Liquid-Phase Extraction*

Three-liquid-phase extraction (TLP, Figure 1) has been used for the isolation of organic macromolecules such as cellulose, enzymes, proteins, and metals [65,66]. This approach is based on the use of three immiscible liquid phases composed of different organic solvents, polymers, inorganic salts, water, or ionic liquids [67,68]. As the number of non-miscible phases is increased from two (biphasic) to three (triphasic), the steps required for separation decrease. Therefore, three metal cations can be separated simultaneously in a single step as shown in Figure 1. For example, in the case of a biphasic system, a mixture of five metals may require four steps for the separation, whereas for TLP, two steps may be sufficient. Different approaches have been considered to design a TLP system for metal extraction. These include one aqueous and two organic phases [69], one organic and two aqueous phases [70], and ionic liquid-based systems [71]. One recent study showed an improved extraction efficiency for Co2<sup>+</sup> with a TLP system when directly compared to an ionic liquid ABS approach [61]. However, in TLP, the challenges associated with the use of organic phases are reintroduced. Examples of metal ion extraction using different TLP systems are tabulated in Table 2.


**Table 2.** TLP systems for metal ion extraction.

<sup>a</sup> trialkylphosphine oxide, <sup>b</sup> diisoamyl sulfide/nonane, <sup>c</sup> di(2-ethylhexyl)phosphoric acid, <sup>d</sup> bis(2,4,4-trimethylpentyl)phosphinic acid, <sup>e</sup> 2-ethylhexylphosphoric acid mono(2-ethylhexyl)ester, <sup>f</sup> primary amine, <sup>g</sup> tri-n-octylphosphine oxide, <sup>h</sup> tri-n-butyl phosphate, <sup>i</sup> trihexyl(tetradecyl)phosphonium bis(trifluoromethylsulfonyl)imide, <sup>j</sup> polyethylene oxide-polypropylene oxide, <sup>k</sup> 1-hexyl-3-methylimidazolium tetrafluoroborate, <sup>l</sup> 1-butyl-3-methylimidazolium hexafluorophosphate, <sup>m</sup> betainium bis(trifluoromethylsulfonyl) imide.

#### *2.3. Cloud Point Extraction (CPE)*

The cloud point is the point where a solution mixture turns cloudy due to diminished solubility of one component after changes to experimental conditions such as pressure, temperature, and inclusion of additives [83]. For example, this clouding process can result in the formation of two distinct phases of nonionic and zwitterionic surfactants in which one is a surfactant-rich phase and the other has a concentration close to the critical micelle concentration [84]. The surfactant-rich phase obtained at the cloud phase condition functions to extract and preconcentrate various inorganics [85]. This phase extracts metal cations and is dispersed in the aqueous phase formed after phase separation. Detection of the cloud point occurs by various techniques (e.g., light scattering or particle counting, turbidimetry, refractometry, thermo-optical methods, and viscometry) [86]. CPE shows great promise as a more environmentally friendly method for heavy metal extractions [87]. Kazi et al. have studied extraction of Al3<sup>+</sup> by the cloud point technique where 8-hydroxyquinone was added to coordinate Al3<sup>+</sup> while the surfactant octylphenoxypolyethoxyethanol (Triton X-114) was added to extract and entrap the complex [88]. Similarly, Zhao et al. studied the extraction of Cd2<sup>+</sup>, Co2<sup>+</sup>, Ni2<sup>+</sup>, Pb2<sup>+</sup>, Zn2<sup>+</sup>, and Cu2<sup>+</sup> using a dual-CPE technique [89]. The main advantage of CPE over other techniques is the use of water

instead of organic solvents [90]. CPE is also easy to manipulate, is fast, requires minimal expense, and offers high analyte recovery [85,91].

#### *2.4. Liquid Membrane Extraction*

Membrane-based extraction is a non-equilibrium process that has been developed as an important green strategy for recovery of rare earth elements [92]. Different types of liquid membranes (LM) have been reported, such as bulk liquid membrane (BLM) [93], emulsion liquid membrane (ELM) [94], supported liquid membrane (SLM) [95], and hollow fiber-supported liquid membrane (HFSLM) [96]. Their advantages and disadvantages are summarized in Table 3. Various metal ions from common metals (copper, nickel, and cobalt) [97] and valuable metals (platinum, gold) [98,99] to radioactive species (uranium) [100] have been extracted using LM techniques. As noted in Table 3, there are several concerns regarding membrane stability when organic solvents are used.


**Table 3.** Liquid membrane systems for metal ion extraction.

#### *2.5. Summary*

In summary, LLE methods often require several extractions for complete recovery of targeted metals. Thus, LLE is often replaced by solid-phase extraction (SPE) methods to achieve higher efficiency and recovery. SPE is advantageous because consumption of organic solvent can be minimized [102]. Additionally, errors from inaccurately measured extraction volumes, especially when multiple extraction steps are required with LLE, are minimized with SPE as it does not require phase separation [103].

#### **3. Solid-Phase Extraction**

Solid-phase extraction (SPE, Figure 2) is one of the most popular sample pretreatment and separation techniques because of its simplicity, low cost, high preconcentration factors, selectivity, and versatility. Furthermore, the availability of a wide variety of sorbent materials and the ability to use only minimal amounts, or in some cases, no organic solvents, makes SPE a very environmentally friendly technique [102,104,105]. Most of the benefits of SPE methods are governed by the physical and chemical nature of the sorbent [104,106]. Recent development and applications of a number of new sorbent materials for metal extraction, such as nanosorbent materials, polymers, metal oxides, magnetic materials, metal organic frameworks (MOFs), and bioadsorbents, are discussed herein.

**Figure 2.** Schematic representation of solid-phase extraction (SPE), solid-phase microextraction (SPME, direct immersion only), and dispersive solid-phase extraction (D-SPE).

#### *3.1. Nanosorbent Materials*

Nanosorbent materials such as carbon nanotubes (CNTs) [107], graphene oxide (GO), silica [108], chitosan [109], and activated carbon [110,111] are particularly useful due to their large surface areas compared to their particle volume. Thus, they are excellent candidates as sorbent materials for metals since the high surface area provides a greater number of active sites leading to enhanced extraction efficiency. Recently, Gouda et al. developed a sorbent material based on multiwalled carbon nanotubes impregnated with 2-(2-benzothiazolylazo)orcinal (BTAO) for preconcentration of cadmium, copper, nickel, lead, and zinc from food and water samples prior to determination by flame atomic absorption [112]. Similarly, carbon nanotubes impregnated with tartrazine [113], polyaniline [114], and di-(2-ethyl hexyl phosphoric acid) [115] have been utilized as sorbent materials for preconcentration, separation, and determination of metals. Moreover, Awual et al. synthesized ligand-impregnated conjugate nanomaterials for the extraction of mercury from aqueous solution [116]. Metal oxides such as Al2O3 [117], TiO2 [118], and SiO2 [119] have been used for metal extraction due to their physical stability, cost-effectiveness, and high surface area [118]. Other examples are shown in Table 4. The utilization of nanosorbent materials is attributed to their high surface area, ease of modification, and nonspecific adsorption with metals [120,121]. However, limitations include low selectivity and, in some cases, low stability and limited reusability of the material.


**Table 4.** Nanomaterial-based solid sorbents.

<sup>a</sup> poly(methyl methacrylate) grafted agarose, <sup>b</sup> multiwalled carbon nanotubes, <sup>c</sup> diethylenetriamine, <sup>d</sup> flame atomic absorption spectrometry, <sup>e</sup> dispersive magnetic SPE, <sup>f</sup> sorption capacity, NR: not reported.

#### *3.2. Polymer-Based Materials*

Some of the limitations found with nanosorbent materials have been addressed by employing specially designed sorbent materials based on chelating resins [129–132], polymers with chelating units [133,134], ion imprinted polymers [135–138], and polymeric ionic liquids [135,139,140].

Polymeric chelating materials, unlike the inorganic nanosorbents, have the advantage of tunability in functionalization using unique chelating groups to obtain enhanced selectivity and extraction efficiencies for metals. Recently, Nunes et al. developed a greener SPE approach for the extraction of Zn and Ni by employing nylon-6 nanofibers modified with di-(2-ethylhexyl) phosphoric acid [141]. The experimental results suggested that these polymeric nanofibers were cost-effective because of their reusability even after ten cycles of extraction in addition to being ecofriendly due to the absence of organic solvents. The same polymeric material was also used for SPE of indium from LCD screens [142]. Furthermore, polymeric materials based on ionic liquids also were utilized as effective sorbent materials for extraction of metals. For example, a polymeric ionic liquid containing 3-(1-ethyl imidazolium-3-yl)propyl-methacrylamido bromide and ethylene dimethacrylate was specifically developed by Zhang et al. for extraction of antimony employing a stir cake sorptive extraction method [143]. Table 5 summarizes several additional examples of polymer sorbents including ion-imprinted polymer (IIP) materials for SPE of metals.


**Table 5.** Polymer-based sorbent materials for metal ion extraction.

<sup>a</sup> SiO2-coated magnetic graphene oxide modified with polypyrrole–polythiophene, <sup>b</sup> magnetic solid-phase extraction, <sup>c</sup> solid-phase extraction, <sup>d</sup> ethyleneglycoldimethacrylate-methacryloylhistidinedihydrate nickel(II), <sup>e</sup> glycidyl methacrylate, <sup>f</sup> ethylene dimethacrylate, <sup>g</sup> iminodiacetate, <sup>h</sup> di-(2-ethyl)phosphoric acid.

#### *3.3. Metal–Organic Frameworks*

Metal–organic frameworks (MOFs) consist of metal ions and organic linkers that are strongly bonded together. These materials have been used as effective sorbents in various applications due to their highly porous structure and the ability to be synthesized in various shapes and sizes [152,153]. Recently, Tadjarodi et al. designed a magnetic nanocomposite sorbent from HKUST-1 MOF combined with Fe3O4@4-(5)-imidazoledithiocarboxylic acid (Fe3O4@DTIM) for SPE of Hg<sup>2</sup><sup>+</sup> in canned tuna and fish samples [154]. The sorbent selectivity towards Hg2<sup>+</sup> was due to the presence of sulfur atoms in DTIM. Also, the magnetic Fe3O4 nanoparticles facilitated separation from samples by simply applying an external magnetic field while the MOF prevented aggregation of Fe3O4 nanoparticles by acting as spacers and a support matrix with the MOF cavities providing increased surface area to enhance sorption capacity. Similarly, Esmaeilzadeh developed a MOF with iron-based magnetic nanoparticles decorated with tetraethyl orthosilicate to create a silica layer on the surface [155]. The nanoparticles were subsequently functionalized with morin (2-(2,4-dihyroxyphenyl)-3,5,7-trihydroxychromen-4-one) as a chelating agent to develop a MIL-101(Fe)/Fe3O4@morin nanocomposite for the selective extraction and speciation of V4<sup>+</sup> and V5<sup>+</sup>. In this case, the silica layer provided stability for the Fe3O4 nanoparticles in acidic conditions as well as allowed for further functionalization. MIL-101(Fe) also prevented aggregation of the nanoparticles by acting as a spacer and support. In addition, Nasir et al. developed a two dimensional leaf shaped zeolite imidazolate frame work (2D ZIF-L) for arsenite adsorption [156]. Table 6 shows recently reported MOFs as effective sorbents for the SPE of metals.


**Table 6.** MOF sorbent materials for metal extraction.

<sup>a</sup> zirconium-based, <sup>b</sup> potassium nickel hexacyanoferrate, <sup>c</sup> zeolitic imidazolate framework-8, <sup>d</sup> Co(II) and 2,4,6-tri(1-imidazolyl)-1,3,5-triazine, <sup>e</sup> iso-reticular MOFs, <sup>f</sup> sorption capacity.

#### *3.4. Magnetic-Based Materials*

In the process of developing more environmentally friendly methods, incorporation of magnetic materials such as iron oxide nanoparticles into sorbent composites has increased in recent years. Magnetic materials are utilized to readily extract target metal ions from complex matrices followed by sorbent separation from samples by an external magnetic field. Following desorption of the metals, the sorbent can be recovered and effectively recycled. Magnetic nanoparticles have been combined with carbon-based [164], ionic liquid [165,166], MOF [167], and polymer [168] materials for magnetic SPE of metals. Several such examples are given in Tables 4–6, while other unique recent studies using magnetic-based materials are described below and in Table 7.

Shirani et al. developed a magnetic sorbent based on an ionic liquid linked to magnetic multiwalled carbon nanotubes for simultaneous separation and determination of cadmium and arsenic in food samples using electrothermal atomic absorption spectrometry [169]. Habila et al. synthesized a sorbent material based on Fe3O4@SiO2@TiO2, which shows unique magnetic, photocatalytic and acid resistant properties, and was used for the preconcentration of copper, zinc, cadmium, and lead prior to ICP–MS analysis [170]. The advantage of this sorbent material was it not only allowed extraction of toxic heavy metals from complex matrices, but also assisted the simultaneous degradation of the organic matrix to aid preconcentration. Additionally, Molaei et al. utilized a copolymer based on polypyrrole and polythiophene (PPy–PTh) layered on the surface of SiO2-coated magnetic graphene oxide for the extraction of trace amounts of copper, lead, chromium, zinc, and cadmium from water and agricultural samples [145].


**Table 7.** Magnetic-based sorbent materials for metal ion extraction.

<sup>a</sup> carbon-encapsulated magnetic nanoparticles, <sup>b</sup> magnetic cobalt nanoparticles functionalized with iminodiacetic acid, <sup>c</sup> magnetic phosphorous-containing polymer, <sup>d</sup> thiazolylazo resorcinol, <sup>e</sup> mercapto groups modified with benzene tricarboxylic acid, <sup>f</sup> working solution pH, <sup>g</sup> sorption capacity.

#### *3.5. Ion Exchange*

Ion exchange is another technique that can be used for the removal of metals, though it depends on the solution composition [178]. Moreover, other factors like the capacity and selectivity of sorbent material, pH, temperature, and solution salinity also play important roles in the ion exchange process [179]. Recently Murray et al. studied the removal of Pb2<sup>+</sup>, Cu2<sup>+</sup>, Zn2<sup>+</sup>, and Ni2<sup>+</sup> from natural water with polymeric submicron ion exchange resins [180]. Similarly, Vergili et al. found good extraction properties with a weak acid cation resin for the sorption of Pb2<sup>+</sup> from industrial wastewater [181].

#### *3.6. Ligand Binding*

Simple coordination chemistry, where a ligand with affinity for a metal binds and forms a complex, is a useful method to selectively isolate a metal from aqueous solution. There are numerous organic chelating agents for heavy and precious metal extraction. The overall challenge is achieving selectivity for a single metal or class of metals. Depending on the strength of binding, recovery of the isolated metal ion can also be difficult. Recent studies show dithiocarbamate ligands as one of the most useful materials to coordinate and extract transition metals from aqueous solution [182]. Because of the presence of various hybridized states of nitrogen and sulfur and the tendency to share electrons between the nitrogen and sulfur with metal ions, the removal of heavy metals by these ligands has been demonstrated [183–186]. They also are known to form colored metal complexes, which makes detection and analysis relatively easy [187]. Table 8 provides examples of ion exchange and ligand binding techniques for metal extraction.


**Table 8.** Ion exchange and ligand binding techniques for metal ion extraction.

<sup>a</sup> poly(acrylic acid), <sup>b</sup> poly(glycidylmethacrylate), <sup>c</sup> polyvinyl alcohol, <sup>d</sup> polyacrylonitrile, <sup>e</sup> graphene oxide, <sup>f</sup> poly(methylhydrosiloxane), <sup>g</sup> pyridine–pyrazole, <sup>h</sup> sorption capacity.

#### *3.7. Solid-Phase Microextraction*

Although SPE has advantages over LLE, more progress is needed in the development of more ecofriendly and cost-effective approaches to further reduce the amount of organic solvents and sorbent material, as well as to minimize cost, analysis time, and disposal of waste chemicals. Such considerations have led to the development of greener alternatives such as SPME (Figure 2), which was developed and introduced by Pawliszyn in 1990 [194]. SPME is a fiber-based version of SPE that has benefits over other extraction techniques because the sample and solvent amounts are reduced, liquid, solid and gas samples can be analyzed with higher sensitivity and cost-effectiveness, and the use of organic solvents is minimized. Briefly, a fiber-based material is used as the sorbent to extract molecules by direct immersion of the fiber into the sample solution (Figure 2) or into the headspace above the solution. Once analytes partition into the sorbent, the fiber is removed for desorption and analysis. Direct coupling to analytical instrumentation is then possible to achieve simultaneous preconcentration and determination of target species, thus reducing the analysis time [195–199].

There are limited reports on the use of SPME for metal ion detection and analysis using HPLC and GC [200,201]. For SPME–HPLC, determination of metal ions is limited to commercial adsorbents [200]. Derivatization is required to obtain a hydrophobic organometallic compound to achieve adsorption onto the fibers and desorption after injection into a SPME–HPLC chamber. Difficulties with slow analyte diffusion in HPLC complicate the analysis of metals. One notable example of SPME–HPLC was reported by Kaur et al., in which a complex of thiophenaldehyde-3-thiosemicarbazone with cobalt, nickel, copper, and palladium was followed by UV detection [202]. SPME coupled to GC is limited to volatile species, which also often requires derivatization prior to detection [203]. Apart from the need for derivatization, there are other challenges including fiber-to-fiber variation, carry over problems, relatively high cost, reusability and recycling of the coating material, instrumental compatibility and, most importantly, delicate fibers or fragile coatings [152,199,204].

A recent goal is the desire to use SPME for the direct extraction and analysis of metal ions without the need for derivatization or complicated procedures. Rahmi et al. developed a novel SPME approach for trace metal analysis by modifying the inner wall of a syringe filter tip with a monolithic chelating moiety [205]. Twenty-two elements, including titanium, iron, cobalt, nickel, copper, gallium, cadmium, tin, and rare earth elements, were extracted prior to ICP–MS analysis with extraction efficiencies higher than 80%. Rohanifar et al. developed a versatile, easily tunable, cost-effective, greener approach for SPME of heavy metals from natural waters [133]. In this study, pencil lead was used as a substrate as an

alternative to a commercially available SPME fiber or a metal wire, which significantly reduced the cost. The pencil lead was coated by electropolymerization with a sorbent composite containing polypyrrole, carbon nanotubes, and different metal chelating ligands. The resultant fiber was then used for direct immersion SPME of heavy metals followed by determination by ICP–MS (Figure 3). The chelating ligand was trapped inside the polymer matrix, which effectively captured the metal from the solution. Metals were therefore preconcentrated onto the fiber and then released in an analysis solution by treatment with acid. A composite containing polypyrrole/carbon nanotubes/1,10-phenanthroline demonstrated exceptional extraction efficiencies for silver, cadmium, cobalt, iron, nickel, lead, and zinc in several sample matrices. The accuracy of the method was validated by the analysis of a certified reference standard. Analyses were accomplished in a minimum amount of aqueous solution and were thus very environmentally friendly.

**Figure 3.** Schematic representation of the creation of an SPME fiber by electropolymerization and its application for metal extraction. Reprinted with permission from [133].

#### *3.8. Dispersive Solid-Phase Extraction*

Dispersive solid-phase extraction (D-SPE, Figure 2) is another variation of solid-phase extraction where a micron-sized sorbent is dispersed in the sample solution. This approach eliminates the need to optimize the flow rate and potential backpressure issues with a packed SPE cartridge, especially with newer nano-based materials. Enhanced contact between the analytes and sorbent results in very efficient extractions [206]. New sorbents for D-SPE for metals are beginning to be reported that utilize materials that effectively and selectively capture metal ions by chelation. Sitko et al. described the synthesis of a graphene oxide sorbent modified with (3-mercaptopropyl)-trimethoxysilane for determination of Co2+, Ni2<sup>+</sup>, Cu2<sup>+</sup> As3<sup>+</sup>, Cd2<sup>+</sup>, and Pb2<sup>+</sup> by total reflection X-ray fluorescence [207]. Preconcentration and metal capture is quite straightforward, while the analysis step is solvent free. Similarly, dithiocarbamate functionalized Al(OH)3–polyacrylamide was prepared and characterized for extraction of Cu2<sup>+</sup> and Pb2<sup>+</sup> [208]. As with SPME, the goal for D-SPE applications is to enhance selectivity for metal analysis with new selective sorbent materials. Recently, pyrrole was derivatized with carbon disulfide and chemically polymerized to obtain an air stable, water-insoluble, chelating polymer for extraction of soft metal ions [209]. Application of this new sorbent for D-SPE of Co2<sup>+</sup>, Ni2<sup>+</sup>, Cu2<sup>+</sup>, Zn2<sup>+</sup>, Cd2<sup>+</sup>, and Pb2<sup>+</sup> demonstrated excellent removal and recovery of these ions. The chelating polymer is reversible, releasing the captured metals after acid treatment for preconcentration prior to analysis by ICP–MS. D-SPE is also amenable to magnetic sorbent particles as demonstrated by the references in Table 4. Therefore, D-SPE shows tremendous promise for developing simple environmentally friendly methods to extract metals.

#### **4. Bulk Sorbent Methods**

#### *4.1. Chemical Precipitation*

Wastewater is a common medium that regularly is contaminated with heavy metal ions. To ensure safe re-entry into the environment, treated water must contain metal concentrations below an accepted level called the maximum contaminant level (MCL) for each metal ion [210,211]. Chemical precipitation is a useful approach to remove large amounts of heavy metals from inorganic waste materials and prevent contamination of the environment [211]. This technique removes ionic metal components after adding counter-ions to reduce their solubility in aqueous solution [212]. Dissolved metals are turned into insoluble components by a precipitating agent under favorable pH conditions [212]. Much research on chemical precipitation for metal extraction has been conducted because of the low cost and ease of implementation for large volumes of wastewater. However, disadvantages such as the inability to maintain pH for optimum precipitation, high volume of sludge production [213], and low selectivity of metal extraction [214] limits widespread use. The treatment method should not produce toxic chemical sludge such that disposal remains ecofriendly and cost-effective [215]. Several examples on the use of precipitating agents to extract various metals have been reported [216–219].

#### *4.2. Biosorbent Extraction*

Biosorbent extraction is particularly important for the removal of heavy metals from industrial effluents as this process utilizes readily available and inexpensive dead biomass compared to conventional sorbents [220]. Aquatic organisms like yeast, algae, and bacteria adsorb dissolved heavy metals and even radioactive elements found in their surroundings [221]. Dead fungal material, for example, does not result in increased toxicity with the extracted metal or adverse operating conditions. Furthermore, no nutrients are needed for dead mass and relatively simple non-destructive treatments are used for the recovery of bound metals, which are often in their anionic forms [220,222]. Natural biosorbents can be valuable low-cost alternatives for metal removal and cleanup, especially for developing countries with limited financial resources. In addition, recent review articles have discussed progress related to the development of ecofriendly phytoremediation and phytoextraction approaches for the removal of metals from contaminated environmental sites [223–225].

Kratochvil et al. studied the removal of molybdate (MoO4 <sup>2</sup>−) with chitosan beads for up to 700 mg g−<sup>1</sup> of molybdate [220]. Similarly, removal of Cr6<sup>+</sup> by peat moss [226] and corncobs [227] was achieved with excellent results. Marine green algae, due to presence of different proteins, lipids, or polysaccharides on the cell wall surface, show good metal binding strength [228]. Hence, for effective removal of heavy metals even at low levels, biosorbents are considered as an emerging technology [229]. However, despite the availability of large quantities of biomass, selection of the most suitable type of biomass is still a challenge. Slight variations in biomass properties can result in considerably different affinities for various metals, which also offers an opportunity to alter biomass properties to design new biosorbent materials. For example, Mallakpour et al. developed a new hydrogel nanocomposite biosorbent by embedding calcium carbonate nanoparticles into tragacanth gum for the removal of Pb2<sup>+</sup> ions from water samples [230]. Similarly, pine (*Pinus sylvestris*) sawdust was modified with thiourea groups and utilized for the extraction of precious metals from industrial solutions [231]. Table 9 shows additional examples of recently reported natural biosorbent materials for extraction of metals.


**Table 9.** Biosorbent materials for metal ion extraction.

<sup>a</sup> sorption capacity.

#### **5. Conclusions**

Recovery of metals often requires extraction from complicated matrices in large quantities, while metal analysis is routinely sought at the trace level. In either case, strategies that are considered greener and minimize their impact on the environment drive development of emerging methods for metal extraction and analysis, many of which are described in this review. Much of the evolution of metal extraction and sample preparation has benefitted from the development and use of new materials. Aqueous two- and three-phase systems reduce the amount of organic solvents needed in LLE and include the use of ionic liquids, which offer the advantageous properties of low flammability and volatility, excellent solvating ability, and high thermal stability. Solid-phase extraction further reduces the need for organic solvents and utilizes novel materials based on adsorption, biosorption, ligand binding, and ion exchange. Extension of SPE into the micro-regime shows exciting promise for effective and selective SPME of metals. Initially, limited by the derivatization of metal ions to generate volatile or hydrophobic organometallic species for gas and liquid chromatographic analysis, new SPME coatings and materials take advantage of classical coordination chemistry to permit direct analysis of metal ions. Development of unique coordination type polymers, magnetic materials, and thin-film coatings for SPE and SPME shows great promise for highly selective and ecofriendly extraction methods for the recovery of valuable metals and for efficient sample preparation and preconcentration of a range of metals from complex matrices.

**Author Contributions:** Conceptualization, G.S.S.S., S.A., A.R., A.P. and J.R.K.; writing—original draft preparation, G.S.S.S., S.A., A.R., A.P. and J.R.K.; writing—review and editing, G.S.S.S., S.A., A.R., A.P. and J.R.K.; project administration, J.R.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This manuscript was prepared with no external funding.

**Acknowledgments:** The University of Toledo is acknowledged for providing support for the student co-authors of this review.

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

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
