Selective Separation and Preconcentration of Caffeine from Natural and Pharmaceutical Products using New Polyurethane Foams

Two polyurethane foam-based sorbents (PUF) were synthesized by imprinting and grafting techniques and examined for selective separation and preconcentration of caffeine (CAF) in some pharmaceutical products and in black tea. Molecularly imprinted PUF was synthesized based on hydrogen-bonding interactions between CAF and alizarin yellow G (AYG) and subsequent polymerization into PUF. The static experiments indicated optimum sorption conditions at pH=6.5 and 5.5 for imprinted PUF (AY-IPUF) and grafted PUF (AY-GPUF), respectively. In the online experiments, the suitable preconcentration time was found to be 40 and 20s for (AY-IPUF) and (AY-GPUF), respectively, at a flow rate of 1.75 mL.min -1 . Desorption of CAF has been affected by passing 500 μL of 0.05, 0.01 mol.L −1 HCl eluent onto (AY-IPUF) and (AY-GPUF), respectively. The online methods have provided satisfactory enrichment factors of 8.4 and 10.5 for (AY-IPUF) and (AY-GPUF), respectively. The time consumed for preconcentartion, elution and determination steps was 1.48 and 1.05 min, thus, the throughput was 42 and 57 h -1 , for (AY-IPUF) and (AY-GPUF), respectively. The developed sorbents were studied for the determination of CAF in pharmaceutical samples which will be helpful to minimize caffeinism. Finally, in silico bioactivity, ADMET and drug-likeness predictive computational studies of caffeine were also carried out.

Since there are several assorted factors that may regulate an individual's retaliation to CAF, it's then difficult to sketch a broad inference about its effects on the general population caused by the varied CAF's intake levels. For instance, pharmacokinetic factors might influence how rapidly the caffeine is absorbed, distributed, metabolized and eliminated (ADME) after being ingested. Meanwhile, this can also be governed by pharmacodynamic factors such as the interaction as well as the consequences of these interactions between CAF and its sites of action in the body (Turnbull, Rodricks, Mariano, 2016). CAF is a psychoactive drug widely usedand perhaps abused. Also, it can enter to the body from many different foods such as coffee and tea (Santos et al., 2010).
The simplest and most cost-effective technique for the trace and ultra-trace determination of drugs is separation and preconcentration using solid-phase extraction (SPE) (Talio et al., 2014). The application of polyurethane foam (PUF) produces smaller resistance for fluid passage. Thus, it results in a low overpressure in the system, reducing the risk of leakage. PUF is safely disposable, not costly and needs only simple preparation procedures. Moreover, this sorbent is highly resistant to rough changes in pH values, despite of swelling in the presence of some concentrated organic solvents, such as ethanol (Marchisio et al., 2005).
SPE is a fast method used for selective extraction, concentration and purification of target analytes prior to analysis. In SPE, solutes are extracted from a liquid-phase into a solid-phase, consisting of small and porous particles of silica or a cross-linked polymer. The extraction might be performed either in batch or column mode (Poole, 2003). Thousands of SPE applications are now available for the extraction of key analytes in the pharmaceutical, clinical/toxicological, environmental and biological fields, and new methods continue to be rapidly developed (Bojdi et al., 2014;Ge et al., 2013). The sorbents that permit faster reaction rates are preferred for the achievement of faster extraction and higher loading capacities (Camel, 2003). Amongst the various sorbents employed in SPE systems, the polyurethane foams (PUF) had a significant importance in the last few years. The PUF is an excellent sorbent material due to its high available surface area, as well as its comparatively low cost. In addition, it is stable in acids and bases (except for the concentrated nitric acid), it will not also change its structure when heated up to 180 o C (Gawargious, Abbas, Hassan, 1988). Thus, it is a very suitable material for both on-line and off-line preconcentration column packing. The on-line methods have the advantage of avoiding the sample manipulation between preconcentration and analysis steps, thereby minimizes the analyte loss and contamination probabilities, allowing higher precision of the analysis (Abdel Azeem, Mohamed Attaf, El-Shahat, 2013).
PUF can be directly used without previous pretreatment (Burham, Abdel-Azeem, El-Shahat, 2008). On the other hand, several chelating agents and liquid ion exchangers have been immobilized on PUF and used as a co-adjuvant in separate procedures (Wanyika et al., 2010).
In our previous work (Azeem, Ali, El-Shahat, 2011), the untreated-PUF was employed for the sorption of CAF.
But due to the lack of selectivity of PUF, it can easily interfere with other co-existing substances. Herein, we attempted to graft PUF with Alizarine yellow G (AY) to increase the sensitivity and selectivity of the sorbent towards CAF. This shall improve the accuracy and precision of the new method.
In the present study, selective AY imprinted PUF (AY-IPUF) and AY grafted PUF (AY-GPUF) were prepared and used for clean-up, and the approach of preconcentration is applied prior to spectrophotometric measurement of CAF to introduce a simple, rapid, costeffective and sensitive method for the evaluation of CAF in complex matrices. The sorbents were examined both in static batch and column dynamic modes to evaluate a suitable phase for continuous analysis. The proposed procedures will be validated by the fast determination of CAF in a representative pharmaceutical product as well as in black tea samples.

Chemicals and solutions
All chemical reagents were of analytical grade. The pH adjustment was done using 0.1 mol.L -1 HCl, acetic/ acetate buffer and ammonia/ammonium buffer in pH ranges of 1-3, 4-6 and 7-9, respectively. A standard solution of CAF (0.25 mg.mL -1 ) was prepared using a high purity CAF supplied by The Egyptian Drug Authority (EDA). All solutions were prepared on daily basis by appropriate dilution of specified aliquots from the standard solutions using double-distilled water. AY was purchased from Aldrich, USA. Polyol (polyethylene oxide-polyethylene glycol), stannous octoate (C 16 H 30 O 4 Sn) and toluene diisocyanate (TDI) [CH 3 C 6 H 3 (NCO) 2 ] were obtained from Safa Foam Company (Egypt) and were used without any further purification. The commercial pharmaceutical sample (Pronto-Plus, tablets) was purchased from October Pharma S.A.E. (Egypt) and was assayed.

Apparatus
Absorbance measurements were carried out using a UV/Vis double-beam spectrophotometer (model 1601 Shimadzu, Kyoto, Japan). The pH measurements were carried out using a calibrated EDT (Dover Kent, UK) and pH-mV meter (model GP353) equipped with an EDT combined glass electrode (± 0.01 accuracy). An automatic shaker with up to 200 rpm with a thermostatic water bath was used for stirring the samples. A Thermo-Nicolet Avatar FT-IR spectrometer (Thermo scientific, USA) was employed to record IR spectra. A homemade peristaltic pump was used to propel the samples into the mini-column to overcome backpressure and to minimize pulsation in the system and the connections were made by using Tygon tubes with 1.52 mm internal diameter. Two similar polyethylene tubes (3 mm internal diameter and 4 cm length) were used as mini-columns and were packed with 150 mg of AY-GPUF or AY-IPUF. The column was successively rinsed with 20 mL H 2 O then 5 mL of a 0.1 mol.L -1 HCl solution and finally with distilled water until the effluent was neutral. The mini-columns were pre-conditioned individually prior to their use by passing 5 mL ammonia buffer (pH = 9.0) to activate the adsorption sites and solvate the functional groups in AYG. Afterwards, 20 mL of doubly distilled water was pumped through the mini-column, at a flow rate of 1.75 mL.min -1 to eliminate the remaining ammonia buffer. This step was daily performed before starting the system operation.

Method of computation: In silico predictive study
In silico studies of caffeine were conducted because the pharmacokinetic factors might influence an individual's response to CAF including how rapidly caffeine is absorbed, distributed, metabolized and eliminated (ADME) after being ingested. It also can be governed by pharmacodynamics factors such as the interaction as well as the consequences of these interactions between caffeine and its site(s) of action in the body. Differences in ADME [comprising retarded gastric emptying which can hinder absorption], and metabolism [explicitly differences in action of cytochrome P450 1A2 (CYP1A2)] can play a significant role in caffeine metabolism, escorting to differences in CAF levels in plasma and time course from the same dose.

The pharmaceutical sample
Six tablets (Pronto-Plus product) were weighed individually to obtain the average mass. Tablets were then ground into fine powder and mixed well. The mass corresponding to one tablet for each formulation was weighed and dissolved in 100 mL (ethanol/water-20/80 v/v) in a volumetric flask, in an ultrasonic bath for 15 min.

Sample preparation
Determination of caffeine in black tea was carried out by UV/Vis Spectrophotometry using a 0.25 g tea sample dissolved in boiling distilled water. Fresh distilled water was then added to reach a net volume of 20 mL. 20 mL sample solution was pipetted to 250 mL volumetric flask. A 10 mL 0.01 mol.L -1 hydrochloric acid and 2 mL basic lead acetate solution were then added, and the contents were completed to the flask's mark with distilled water. The mixture was shaken up and filtered. 50 mL filtrate were pipette and added to a 100 ml flask, 0.2 mL 4.5 mol sulfuric acid was added and again distilled water was added up the flask's mark, shaken up and finally filtered (Wang, Li, Xu, 2006).

Synthesis of AY-GPUF and AY-IPUF
Alizarine yellow G was recommended as a reagent in the chemical modification of PUF due to its chemical structure which contains hydroxyl (OH) and carboxylic (COOH) groups (El-Shahat, Burham, Abdel Azeem, 2010). The OH groups can be grafted into the backbone of PUF by competing with a polyol to the toluene diisocyante groups during polymerization. The remaining COOH groups are easily ionized to form negative adsorption sites suitable for the adsorption of caffeine.
The preparation of AY-GPUF was based on a previously published work (El-Shahat, Burham, Abdel Azeem, 2010). 20.0 g liquid polyol were mixed with 0.5 g silicon oil, 0.02 g triethanolamine, 1.0 g doubledistilled water, and 0.02 g (6.46 × 10 -5 mmole) of AY. Then, 0.02 g stannous octoate was added to the mixture and vigorously stirred for about 3 min. 13 g TDI were added at once under stirring for 30 s. Foaming reaction started in the liquid mixture by the evolution of carbon dioxide gas. Curing into solid polymeric material takes place afterward into a yellow-colored foam.
AY-IPUF was synthesized by mixing CAF and AY by the ratio 1:4 mmole to form a non-covalent complex. One milliliter aliquot of the complex solution was added to liquid polyol (20 g) and continued the same synthesis protocol as mentioned for AY-GPUF.
After the removal of the CAF template species, the cavity can then be used to selectively rebind the template from a mixture of chemical species. Desorption of CAF from the imprinted foam was examined by HCl solution at various concentrations and different soaking periods. The amount of recovered CAF was measured using spectrophotometry. The results shown in Figure 2 indicate that absorbance increase rapidly at the beginning; suggesting that most CAF was desorbed followed by a slow release. Also, the best concentration of HCl for complete desorption of CAF from AY-IPUF was found to be 0.1 mol.L -1 in a volume of 15 mL. The soaking time was studied within the range of 0.5 -24 hours, which is necessary to strip out the bound CAF without shaking the sorbent. Results indicated that the extraction of all CAF takes at least 24 hours. The HCl was recommended as it does not affect the base line of absorbance. Desorption under shaking or flow conditions was found to occur at shorter times as will be discussed later. The IR spectrum was used for characterizing CAF-free AY-IPUF and untreated PUF.
Selective Separation and Preconcentration of Caffeine from Natural and Pharmaceutical Products using New Polyurethane Foams

Off-line experiments
The optimum pH for CAF uptake was determined by using the batch equilibration technique. 20 millimeters of solutions, containing 5.0 μg.mL -1 CAF were adjusted to pHs in the range 1-10, then shaken with 0.1 g of foam for 30 min. After equilibration, the remained CAF was determined by the recommended method. The effect of shaking time on the extraction of CAF was studied using the batch experiment. For this propose, 20 mL solutions containing 0.25 μg.mL -1 CAF at optimum pH were automatically shaken with 0.1 g sorbent for time intervals 2, 5, 10, 15, 20, 25 or 30 min. The total capacity of each sorbent was then investigated. Typically, 0.1 g of foam was added to 20 mL of CAF containing 5.0 μg.mL -1 at optimum pH and shaken for 30 min at ambient temperature. After equilibration, the remaining CAF in the supernatant solution was spectrophotometrically determined at λ max. = 274 nm, as shown in Figure 3.

On-line preconcentration
In order to evaluate the performance of both AY-GPUF and AY-IPUF foams under dynamic conditions, adsorption studies were carried out using mini-columns integrated into the on-line preconcentration system, so 0.15 g foam was packed in a cylindrical polyethylene tube. The on-line system was operated in a time-based mode. Preconcentration of CAF was performed by pumping the CAF sample (0.25 μg.mL -1 ) adjusted to pH 5.5 and pH 6.5 through the mini-column at a flow rate of 1.7 mL.min -1 with 20 s and 40 s preconcentration time for AY-GPUF and AY-IPUF, respectively. Then, a double-distilled water carrier is passed through the mini-column till the absorbance baseline has been restored. Elution of CAF was achieved by injecting the carrier stream by 500 μL from 0.05 mol.L -1 and 0.025 mol.L -1 hydrochloric acid solutions, respectively, at a rate of 1.75 mL.min -1 . The displaced analyte from the sorbent was measured using UV spectrophotometer based on absorbance peak height at 274 nm. This procedure was applied for both standard and pharmaceutical samples.

In silico predictive screenings
The work investigated here was carried out with the intent to forecast the bioactivity, drug-likeness and ADMET traits of caffeine utilizing in silico computational tools. Pre-ADMET and Swiss ADME were employed to foretell in-vivo ADMET attributes of CAF. Physicochemical properties of CAF were also predicted where the molecular formula is C 8 H 10 N 4 O 2 , molecular weight is 194.19 g.mol -1 , the number of heavy atoms is 14, fraction Csp3 (the ratio of sp 3 hybridized carbons over the total carbon count of the molecule) being 0.38, with zero number of rotatable bonds, the number of H-bond acceptors are 3, number of H-bond donors being 0, molar refractivity is 52.04, TPSA (topological polar surface area) is 61.82 Å². In silico % absorption of CAF was calculated to be 87.67% by the reported equation (%ABS = 109 -(0.345 X TPSA) (Al-blewi et al., 2019).
CAF is a non-BBB (blood-brain barrier) permeant but it has a high gastro intestinal absorption with an HIA (human intestinal absorption) percentage of 93.82%. It is a P-gp (p-glycoprotein) non-inhibitor. The PPB (plasma protein binding) emerged to be 14.08. The predicted passive HIA (human gastrointestinal absorption), permeation through BBB and P-gp substrates are ordered together in an instinctive graphical categorization model i.e. BOILED-Egg diagram as shown in Figure 5. Molinspiration software was utilized for the prediction of bioactivity outcome of CAF for six different protein structures i.e. GPCR ligand (-0.53), ion channel modulator (-0.98), kinase inhibitor (-1.07), nuclear receptor ligand (-2.10), protease inhibitor (-1.23), and enzyme inhibitor (-0.22). All the predicted values of the bioactivity score are negative, displaying a low affinity of CAF towards the selected protein structures.
Molsoft was used to predict drug-likeness score which is 1.03 in case of CAF as shown in Figure 4. CAF does not inhibit any of the studied cytochrome P450 isomers. It displayed moderate Caco-2 cell permeability (21.26%) and low MDCK cell permeability (2.95%). Toxicological attributes of mutagenicity (Ames Test) and carcinogenicity (mouse and rat) for caffeine were also predicted. It is predicted as mutagenic by Ames test, carcinogenicity in mouse is negative, i.e. there is evidence of carcinogenic activities in mouse and in rats, the prediction is positive i.e. it is not carcinogenic in nature. Different drug-likeness rules in (Lipinski et al., 2001), (Ghose, Viswanadhan, Wendoloski, 1998), (Veber et al., 2002), (Egan, Lauri, 2002) and (Muegge, Heald, Brittelli, 2001) were appraised to find out the number of violations if any by CAF. According to the drug-likeness rules of Lipinski, Veber and Egan, caffeine has drug likeliness as it does not violate any of these rules (0 violations). But predictions in terms of Ghose and Muegge rule flaunted that caffeine does not have drug-likeliness as it has 1 violation in each set of rules. The bioavailability score of CAF is 0.55 and its bioavailability radar is shown in Figure 6.

Fourier transfer infra-red (FTIR)
The IR spectra of the untreated PUF exhibited the absorption bands correlated to νNCO, νOH, νCO and νNH 2 groups at 2100, 3509, 1655, 3111 and 3299 cm -1 , respectively. However, the spectra of AY-GPUF and AY-IPUF showed disappearance of the bands corresponding to νNCO, νOH groups and appearance of two bands corresponding to νCO and νNH 2 at 1700-1663 and 3231-3346 cm -1 , respectively, as shown in Figure 7. FIGURE 6 -Bioavailability radar of caffeine: the pink area represents the optimal range for each property (lipophilicity: XLOGP3 between -0.7 and +5.0, size: MW between 150 and 500 g.mol-1, polarity: TPSA between 20 and 130 (Å2), solubility: log S not higher than 6, saturation: fraction of carbons in the sp3 hybridization not less than 0.25, and flexibility: no more than 9 rotatable bonds. In this example, the compound is predicted not orally bioavailable, because too flexible and too polar.
Selective Separation and Preconcentration of Caffeine from Natural and Pharmaceutical Products using New Polyurethane Foams

Effect of pH on adsorption of CAF
The effect of solution's pH on the retention of CAF was studied in the range of pH = 2.5-9.5. For AY-IPUF sorbent, the sorption % has increased in the range of pH = 2.5-6.5 which might be attributed to the presence of CAF on it is protonated form. While at pH range of 5.5-9.5, the sorption % has slightly decreased, which might be due to reduction in the concentration of the positively charged protonated CAF which is available for ion-pair formation and less extent of a negatively charged ionized form of AY and independence of extraction on pH due to hydrophobic sorption by π-π interaction rather than electrostatic interaction. In case of AY-GPUF, the maximum sorption % was observed at pH = 5.5 which reveals the optimum pH value for ionization of COOH in AY reagent. Before pH 5.5, the sorption % is lower due to the persistence un-dissociated form of AY, while beyond this value the proportion of protonated CAF decreases as can be seen in Figure 9. Therefore, the optimum pH values for adsorption of CAF are 6.5 and 5.5 for AY-IPUF and AY-GPUF, selected for the present study. The percentage of sorption was determined using Eq. 1 (Azeem, Ali, El-Shahat, 2011): where C o and C are initial and remaining concentration (μg.L -1 ) of CAF, respectively. Evidently, CA F-f ree AY-I PU F showed characteristic absorption bands at 3467, 1714, 1663 and 3330 cm -1 corresponds to νOH, νCO and νNH 2 , which indicate the incorporation of these groups during complex formation to CAF., At pH ≥ 4.0, the COOH of alizarin will be ionized into the negatively charged form (COO -) making the polymer surface negatively charged in AY-GPUF. Meanwhile, positive charges are formed in the protonated nitrogen atoms in CAF causing attraction forces between the protonated CAF and the polymer surface. While in the case of AY-IPUF, similar adsorption mechanism takes place but the sorbent forms a cyclo-polymeric chain around the CAF molecule as shown in Figure 8.

Influence of shaking time
The influence of shaking time on the sorption of CAF was investigated by the batch off-line procedure in the range 5-30 min at optimum pH. The results are shown in Figure 10. Obviously, both sorbents have shown fast kinetics for the sorption of CAF. The equilibrium state has been reached for both AY-GPUF and AY-IPUF after a 10 min shaking period. So, a stirring interval of 10 min would be recommended to reach the equilibrium state. The faster sorption of CAF ions on the modified PUFs probably reflects a good accessibility of the studied analytes ions to the new chelating sites in the sorbents and ensure surface phenomenon and that readily accessible to CAF ions solutions. Kinetics of sorption was found to verify the firstorder rate equation represented by Lagragren relation (Eq. 2) (Salam, 2008): where q e and q t are the amount of CAF (µg.g -1 ) extracted at equilibrium and at any shaking time t (min), K is the first-order rate constant (min -1 ). The data presented in Figure 11 reveals a negative slope for the imprinted foam while a positive slope was obtained for the grafted one. The values for the rate constant (K) were found to be 0.028 and 0.0087 min -1 and half-life times are 33.06 and 106.2 min for AY-GPUF and AY-IPUF, respectively.

Sorption capacity
The maximum exchange capacity of the sorbent (maximum amount extracted per gram of the sorbent) was determined by shaking 100 mg of the sorbent with excess CAF concentration (5.0 mg.L -1 ) for 30 min at ambient temperature. The remaining amount of CAF in the supernatant solution was determined as described in the recommended procedure. The sorption capacity Q (μg.g -1 ) was calculated from (Eq. 3) (Huo et al., 2019): Modeling of the adsorption isotherm by Langmiur and Freundlich equations is represented in Figure 13. The Langmuir isotherm takes into consideration that sorption occurs on uniform surfaces containing sites with an equivalent adsorbing capacity in relation to the solute. Also, it considers adsorption occurs on a monolayer (de Almeida et al., 2007). Therefore, for a solute-solvent system in which only the solute is significantly adsorbed onto the solid surface, the Langmuir isotherm can be expressed as in Eq. 4: where C e is the concentration of the solute after equilibrium is achieved (mg.L −1 ), q ads is the amount of solute retained on the adsorbent at equilibrium (mg. where V is the sample volume (L) and m is the weight of sorbent (g). The maximum capacity for AY-GPUF and AY-IPUF was found 6500 and 1400 μg.g -1 , respectively. Variation in the sorption capacity with the sorbent type is expected to be parallel to the stability of the CAF-AY complex. The increased capacity of the grafted foam could be due to non-selective binding of CAF and extraction is mainly occurred on the polymer surface. However, the imprinted foam is highly selective and the capacity is mainly dependent on the number of template sites previously formed in the polymer backbone.

Equilibrium adsorption isotherms
The sorption isotherm of CAF was carried out by determination of the sorbent uptake (μg.g -1 ) at varying of the CAF concentration. Typically, 0.1 g foam sorbent was equilibrated with CAF solutions in the concentration range 1-600and 1-40 μg.mL -1 in case of AY-GPUF and AY-IPUF, respectively.
The isotherm of AY-GPUF shows an "L" isotherm as shown in Figure 12(A). It shows one increasing section in the range 1-300 μg.mL -1 followed by plateau indicating saturation of sorbent by CAF.
The isotherm of AY-IPUF shows an "S" isotherm, as shown in Figure 12(B). The first section occurred in the concentration range of 1-20 μg.mL -1 and exhibited lower slope which may be due to the slow sorption of CAF at early stages by the highly selective imprinted sites. Further increase in CAF concentration leads to a saturation of the selective adsorption sites and the analyte becomes adsorbed by electrostatic forces at the surface which has a quite faster rate as indicated by the higher slope within the concentration range 30-40 μg.mL -1 then the sorbent total capacity was reached. Obviously, the extracted amount by grafted foam is much greater than that for the imprinted foam which emphasis the conclusion mentioned in the section: Sorption capacity. g −1 ), Q is the maximum adsorption capacity of the monolayer, and the term b can be assumed as apparent adsorption equilibrium constant. The application of the Langmuir isotherm to the present results yielded a linear relationship and the observed intercept is positive. Similarly, both grafted and imprinted foams have shown similar straight-line plots with positive slopes. The linear correlation coefficients for the grafted and imprinted sorbents are 0.990, and 0.101, respectively. The values for Q and b are 4.2 mg.g -1 and 0.006 L.mg -1 for AY-IPUF and 7.75 mg.g -1 and 0.006 L.mg -1 for AY-GPUF, respectively.
Freundlich isotherm assumes that multiplayer adsorption of analyte takes place on a heterogeneous surface and the extent of solute adsorption increases infinitely with increasing the analyte concentration (Ababneh et al., 2019;Avila, 2005). The empirical equation of Freundlich isotherm can be represented by Eq. 5 (Kumar et al., 2012): where C ads is the amount of solute retained on the adsorbent, C e is the solute concentration in solution after equilibrium was reached, and 1/n and K F are the isotherm parameters that measure of adsorption intensity and adsorption capacity, respectively. The equation of Freundlich isotherms was linearized by applying a logarithm as in Eq. 6 (Pragathiswaran, Sibi, Sivanesan, 2013): The values of K F and n could be obtained by plotting log C ads against log C e.
The calculated constants of Langmuir and Freundlich parameters are shown in Table I. Sorption by AY-GPUF is better described by Langmiur model (R 2 =0.990) than the Freundlich model (R 2 =0.981). The Freundlich isotherm is usually employed for the modeling of heterogeneous systems, being successfully used for the description of adsorption processes in solution (Hameed, Din, Ahmad, 2007;Li et al., 2009). It can be observed that all the values of n are higher than 1.0, which can be attributed to the molecular interaction between adsorbent and adsorbate.
In the controversy, adsorption by AY-IPUF showed a sigmoidal curve (S isotherm) and has a point of inflection which may be a result of at least two types of opposing mechanisms leading to a phenomenon of cooperative adsorption. This can be observed from the nonlinear behavior of the two isotherms described by Langmiur and Freundlich models. The correlation coefficient was 0.101 and 0.875 for Langmiur and Freundlich models, respectively, which are considered insignificant.
Moreover, the value of n for the adsorption of CAF onto AY-IPUF is less than that for AY-GPUF, which is probably ascribable to the two adsorption cavities in AY-IPUF (De Boer, 1968). This may be due to adsorption to the imprinted cavity necessitates a particular orientation of the CAF molecule at the solid-liquid boundary phase as well as one CAF molecule may need more than one AY i.e the ratio adsorption stiochiometry of CAF:AY is 1:2, 1:3, or higher. The equation of the Freundlich isotherms commonly fits solute adsorption onto rough surfaces better than Langmuir isotherms (Ferreira, Andrade, dos Santos, 2004), once it takes into account the heterogeneity of the solid surface and the variable energy distribution of the adsorption sites. However, it can be applied only to a narrow range of solute concentrations, which, in turn, limits its application in the prediction of adsorptive capacities (Banerjee, Chattopadhyaya, 2017). The AY-GPUF showed higher adsorption capacity and affinity for CAF, as the sorbent surface is readily available for attracting CAF since the AY reagent is anchored to the outer functional groups causing the adsorption process to suffer less steric hindrance. Also, the ratio of CAF:AY is expected to be 1:1 due to the linear grafted polymeric chain of AY-GPUF (See Figure 8).

Selectivity
The effect of foreign ions which might interfere with the sorption of CAF was examined. Model solutions containing 5 μg.mL -1 of CAF and interfering substance at various concentrations were prepared and mixed to the sorbent for batch extraction. The recovery of CAF was compared to the presence of the tested substance to a reference value obtained in the absence of the foreign compound. The results are shown in Table II. The tolerance limit is defined as the amount of interference causing an error in the recovery by ≥ ±5%. Starch, serine, histidine, citrate, could be tolerated at up to 200 mg.L -1 , at least. Alkali and alkaline earth elements do not form stable complexes and are not retained on the sorbents under the working conditions used for the system (Anthemidis, Ioannou, 2006).

Performance under flow conditions
The sample's flow rate is a very important parameter which affects the sorption's efficiency as well as the analysis time ( Figure 14). In investigating the flow rate between 0.5 and 2.6 mL.min -1 , there was a maximum and constant absorbance in the range 0.5-1.8 in case of both sorbents followed by a decrease in absorbance at higher flow rates. Therefore, a 1.75 mL.min -1 flow rate was recommended for further experiments. The dynamic capacities of the AY-GPUF and AY-IPUF sorbents were also determined from breakthrough point. In this experiment, a model solution of 5 μg.mL -1 CAF was adjusted to pH = 6.5 and pH = 5.5 for AY-IPUF and AY-GPUF, respectively. The column dynamic capacities were calculated from the breakthrough point and were found to be 32.88 mg.g -1 and 42.02 mg.g -1 for AY-GPUF and AY-IPUF, respectively. These values are well above the content of CAF in the analyzed pharmaceutical sample. Preconcentration time was investigated in the range of 5-45 s using 5 μg.mL -1 CAF solution at a flow rate of 1.75 mL.min -1 as presented in Figure 15. Linear relationships were observed within the preconcentration time range of 1-20 and 1-35 s for AY-IPUF and AY-GPUF, respectively, which reflect quantitative sorption of CAF. At extended preconcentration periods, the recovery becomes nonlinear, perhaps due to the washing effect of the sample (Wells, 2003). The calculated enrichment factors were found 8 and 10 for AY-IPUF and AY-GPUF, respectively. The time consumed for preconcentartion, elution and determination steps was 1.48 and 1.05 min, thus, the throughput was 42 and 57 h -1 for (AY-IPUF) and (AY-GPUF), respectively.

Desorption of CAF
Since CAF extraction occurred via electrostatic attraction forces between positively charged protonated CAF and negatively charged COOin AYG, consequently, the desorption should be performed by the suitable electrolyte which neutralizes the negative charge on the ligand. The use of acids as eluents for desorption was found to be highly effective because acids can rapidly neutralize the charge by immediate adjustment of pH. Different acids, including HNO 3 , CH 3 COOH and HCl, were tested. The best results were obtained by using HCl as it gives a very weak background signal at the wavelength of measurements. Subsequently, various concentrations from HCl were studied and the best recovery was obtained at 0.01 and 0.05 mol.L -1 for AY-GPUF and AY-IPUF, respectively. Furthermore, the elution volume was optimized by studying the recovery of CAF at eluent volumes varying between 200 and 1200 μL. It was found that quantitative desorption was achieved by 500 μL HCl in both sorbents, therefore it was chosen for adequate desorption.

Analytical performance
The limit of detection, defined as three times the standard defined deviation of the blank (3σ) (Thomsen, Schatzlein, Mercuro, 2003), was found to be 2.26 and 0.98 μg.mL -1 for AY-IPUF and AY-GPUF, respectively. Similarly, the limit of quantification (10σ) was found to be 7.54 and 3.25 μg.mL -1 for AY-IPUF and AY-GPUF, respectively. Also, the precision for 5 replicate determinations expressed as RSD was 0.3 and 1.2%, respectively, at the concentration of 5 μg.mL -1 CAF. The linear range obtained by using the on-line preconcentration system is 1-80 μg.mL -1 with a correlation coefficient of 0.982, and 0.993 for AY-IPUF, and AY-GPUF, respectively. Table III compromises the analytical performance data for the on-line preconcentration procedure using AY-IPUF and AY-GPUF sorbents. The accuracy of the proposed preconcentration methods was investigated by the addfound test (Abdel Azeem, Mohamed Attaf, El-Shahat, 2013). The recovery was found to be in the range of 99-104% and the corresponding RSD varied from 0.05 to 7.4% as indicated in Table IV.

Application to real samples
Extraction of CAF from black tea was achieved by using AY-GPUF, AY-IPUF as an extracting media. The RSD of extraction of CAF from black tea solution was 0.33 and 0.257% for AY-IPUF and AY-GPUF, respectively. The results are summarized in Table V. The validity of the proposed procedures was also tested by determining CAF in pharmaceutical preparations. The recovery was found to be 100 and 108% and RSD 0.9 and 0.4% for AY-GPUF and AY-IPUF, respectively (Table VI). Therefore, the proposed sorbents are highly sensitive and could be used successfully for routine analysis of natural samples as well as pharmaceutical products.

Comparison with other methods
A comparison of the analytical performance data of our developed AY-IPUF and AY-GPUF method to other reported procedures for preconcentration/separation of CAF is presented in Table VII. The developed AY-GPUF showed a higher capacity than the untreated-PUF. The detection limit for AY-GPUF was slightly higher than the hypercrosslinked polystyrene method (Andreeva, Dmitrienko, Zolotov, 2010). The analytical range of the methods under investigation was wider than all reported methods which conveys the validity for analysis of samples with varying concentrations. Also, the enrichment factor was found comparable to the homogeneous liquid-liquid microextraction method (Amini, Hashemi, 2018). Finally, the developed procedure provided a successful alternative for the preconcentration/ separation and determination of CAF with a sufficient accuracy and precision.

CONCLUSION
CAF was explored for its physicochemical, pharmacokinetic, and pharmacodynamic traits utilizing computational tools along with in silico predictive bioactivity and drug-likeness scores. Apart from these predictive studies, a novel, rapid, yet sensitive, method has been developed for quantification of caffeine in pharmaceutical and natural samples. The proposed imprinted and grafted sorbents have shown fast sorption/desorption characteristics and allowed the on-line determination of the CAF. The extensive sample preparation is not needed which allows easy and rapid method development. The run-time for each sample was 5 min. Adequate sensitivity and selectivity were achieved with a traditional UV detector. The attained LOD in this study allows for the quantification of CAF in pharmaceutical preparations as well as in pure samples. Overall, the proposed method is highly sensitive and selective for the determination of CAF in complex matrices.