Molecular Imprinting of Benzylpiperazine: A Comparison of the Self-Assembly and Semi-Covalent Approaches

Molecularly imprinted polymers (MIPs) for benzylpiperazine (BZP, 1), an illicit designer drug, were developed by using both self-assembly and semi-covalent approaches. From an array of potential functional monomers (FMs) and using a combination of pre-synthetic interaction studies (by molecular modelling and NMR analysis) and binding assays, the highest performing self-assembly 1-MIPs were confirmed to result from methacrylic acid (7) as FM, ethylene glycol dimethacrylate (EGDMA) or trimethylolpropane trimethacrylate (TRIM) as crosslinkers and chloroform as the porogen and rebinding solvent at template (T): FM ratios of 1:1 and 1:2, giving imprinting factors (IF) 3 to 7. The semi-covalent 1-MIPs were designed using benzylpiperazine (4-vinylphenyl) carbamate (16) as the template–monomer adduct in combination with either EDGMA or TRIM. Our comparative analysis showed the semi-covalent polymers to have a stronger affinity for 1 (significantly lower Kd values and higher IFs) and faster uptake than the self-assembly systems. Both approaches have comparable cross-reactivity: marginal to low against cocaine (17) and morphine (18) and high against ephedrine (19) and phenylpiperazine (20). They also have comparable selectivity: highly selective towards 1 against 17, moderate against 18 and non-selective against 19. EGDMA-based self-assembly MIPs displayed a greater imprinting effect (higher IFs and NIP-to-MIP Kd ratios) than TRIM-based MIPs, while the TRIM-based semi-covalent MIP outperformed its EGDMA-based equivalent. By virtue of its modest selectivity against the test illicit drugs, 1-MIPs could potentially be used as a dummy MIP for the broad-based capture and enrichment of illicit drug blends for subsequent laboratory analysis.


Introduction
Benzylpiperazine (BZP, 1) is one of the dominant bioactive compounds in a relatively new class of piperazine-based illicit designer drugs in circulation [1]. This compound family includes 1-(3,4-methylenedioxyphenyl)piperazine (2), 1-(3-trifluoro-methylphenyl)piperazine (3), 1-(3-chlorophenyl)piperazine (4) and 1-(4-methoxyphenyl)piperazine (5) (Figure 1). The 1-hydrochloride salt (its most widely distributed form) is a white powder that is usually sold as tablets or capsules. It acts as a stimulant, increasing blood pressure, auditory vigilance and heart rate [2]. The biological activity of 1 is thought to be mediated through 5-HT-uptake inhibition and 5-HT1 antagonism effects [2][3][4][5]. Co-administration of 1 with 3 results in psychoactive effects, including hallucinations, similar to those associated with 3,4-methylenedioxymethamphetamine (MDMA), and is 10 times more active than amphetamine [6,7]. Ingestion of 1 can have lethal outcomes [8,9]. Increasingly, countries have listed 1 as an illicit substance, but up until 2008, 1 was legally available to people over the age of 18 in United Kingdom and New Zealand [10,11]. A laboratory-based analyses of 1 could easily be conducted by using spectroscopic and chromatographic techniques [12,13], but reports on potential on-site detection have been limited to electrochemical methods [14,15] and presumptive colour tests [12,16]. While presumptive colour tests are rapid and cheap, these are prone to false positives, some tests are pH dependent and reagents have limited stability even with low temperature storage. Electrochemical methods rely on nonportable instrumentation with lower resolution than the chromatographic techniques. Spectroscopic on-site analysis is now possible with the emergence of portable instruments, e.g., FTIR; however, this would still require specialised training, as with all other forms of laboratory-based analytical instrumentation.
Molecular imprinted polymers (MIPs) have been used by our group [17,18] and others [19] for the capture and detection of illicit drugs. Nevertheless, there has been no report in the literature on MIPs designed for the capture of 1 to date. Thus, in this study, we aimed to develop 1-selective MIPs, initially as an extraction and enrichment material and potentially as an on-site detection system. MIPs are generated from a functional monomer (FM), an analyte used as a template (T)-1 in this study, a crosslinker (XL) for structural rigidity and a porogen (solvent) usually by radical polymerisation. Post-synthesis template extraction develops a complementary binding site within the polymer matrix, allowing highly selective analyte [20] or related analogue rebinding, i.e., dummy templating [21]. Due to their ease of synthesis, high stability and low cost, the use of MIPs as detection elements for several sensing platforms [22][23][24][25] is gaining momentum. At their simplest, MIPs are used as highly specific solid-phase extraction systems and afford easy identification and quantification by traditional analytical means [26][27][28][29][30][31].
A laboratory-based analyses of 1 could easily be conducted by using spectroscopic and chromatographic techniques [12,13], but reports on potential on-site detection have been limited to electrochemical methods [14,15] and presumptive colour tests [12,16]. While presumptive colour tests are rapid and cheap, these are prone to false positives, some tests are pH dependent and reagents have limited stability even with low temperature storage. Electrochemical methods rely on nonportable instrumentation with lower resolution than the chromatographic techniques. Spectroscopic on-site analysis is now possible with the emergence of portable instruments, e.g., FTIR; however, this would still require specialised training, as with all other forms of laboratory-based analytical instrumentation.
Molecular imprinted polymers (MIPs) have been used by our group [17,18] and others [19] for the capture and detection of illicit drugs. Nevertheless, there has been no report in the literature on MIPs designed for the capture of 1 to date. Thus, in this study, we aimed to develop 1-selective MIPs, initially as an extraction and enrichment material and potentially as an on-site detection system. MIPs are generated from a functional monomer (FM), an analyte used as a template (T)-1 in this study, a crosslinker (XL) for structural rigidity and a porogen (solvent) usually by radical polymerisation. Postsynthesis template extraction develops a complementary binding site within the polymer matrix, allowing highly selective analyte [20] or related analogue rebinding, i.e., dummy templating [21]. Due to their ease of synthesis, high stability and low cost, the use of MIPs as detection elements for several sensing platforms [22][23][24][25] is gaining momentum. At their simplest, MIPs are used as highly specific solid-phase extraction systems and afford easy identification and quantification by traditional analytical means [26][27][28][29][30][31].
Here we report on our efforts towards the development of a 1-selective MIP via selfassembly and semi-covalent approaches. Being synthetically simpler, a majority of MIPs fall into the former classification. With self-assembly MIPs, the FM, T, XL and porogen are pre-mixed and allowed to self-associate prior to polymerisation. To optimise the formulation, pre-synthetic approaches such as virtual imprinting [32][33][34], semi-empirical calculations [11], thermodynamic studies [35,36], spectroscopic (NMR [37,38], UV-VIS [39,40] and FTIR [41]) analyses, chemometric methods [42] and combinatorial screening [33,43,44] have been extensively employed. Commercially available FMs are commonly utilised, except in cases where specialty monomers are required. Covalent and semi-covalent MIPs, on the other hand, require the formation a T-FM adduct, which, at times, involves numerous synthetic steps. Regardless, the covalent (T attached covalently to FM moiety during binding) and semi-covalent (T-FM interaction during rebinding is non-covalent, as with the selfassembly systems) approaches have been highly successful, especially in cases where the parent template was poorly functionalised and thus offered limited possibilities for strong non-covalent interactions at the pre-polymerisation stage of MIP synthesis [45][46][47]. Our self-assembly MIP development process commenced with molecular modelling interaction and NMR titration (MM-NMR) studies to screen potential FMs and XLs ( Figure 2) and determine the optimal FM-T interactions [17,18]. In the case of the semi-covalent approach, we synthesised and employed O-4-vinyl 4-benzylpiperazine-1-carbothioate (benzylpiperazine (4-vinylphenyl) carbamate, 16, Figure 5) as a covalent surrogate for 1.
We successfully prepared high-performing self-assembly 1-MIPs from methacrylic acid (7, Figure 2) as an FM, at T:FM ratios of 1:1 and 1:2, using chloroform as a porogen (IF's 3 to 7), with the EGDMA-based MIPs outperforming the TRIM MIPs. Nevertheless, the semi-covalent MIPs, particularly the TRIM-based polymers, have a stronger affinity for 1 (significantly lower Kd values and higher IFs) and faster uptake than the self-assembly systems. Both approaches exhibited comparable cross-reactivity-marginal to low against cocaine (17) and morphine (18) and high against ephedrine (19) and phenylpiperazine (20); and selectivity-highly selective towards 1 against 17, moderate against 18 and non-selective against 19 (see Figure 7 for structures of test illicit substances).
We successfully prepared high-performing self-assembly 1-MIPs from methacrylic acid (7, Figure 2) as an FM, at T:FM ratios of 1:1 and 1:2, using chloroform as a porogen (IF's 3 to 7), with the EGDMA-based MIPs outperforming the TRIM MIPs. Nevertheless, the semi-covalent MIPs, particularly the TRIM-based polymers, have a stronger affinity for 1 (significantly lower K d values and higher IFs) and faster uptake than the self-assembly systems. Both approaches exhibited comparable cross-reactivity-marginal to low against cocaine (17) and morphine (18) and high against ephedrine (19) and phenylpiperazine (20); and selectivity-highly selective towards 1 against 17, moderate against 18 and nonselective against 19 (see Figure 7 for structures of test illicit substances).
An initial analysis might suggest that these high levels of cross-reactivity and modest selectivity against the tested illicit drugs were unfavourable; however, these 1-templated MIPs are useful for broad-based capture and enrichment of illicit substances usually blended with, rather than specifically for, 1, analogous to dummy-templated MIPs. These interactions were demonstrated by significant upfield shifts in the resonance signals (2.5-6 ppm) of the carbons in the piperazine ring of 1 in the presence of FMs 6, 7 and 12.
The resonance shifts could be attributed to the hydrogen-bond interactions between the adjacent nitrogens and the acidic moieties of the FMs. Conversely, the carbonyl carbons of FMs 6, 7 and 12 also exhibited a downfield movement (deshielding) of their signals, an expected effect of the hydrogen-bond interactions. The Job plots generated for 1 and FMs 6, 7 and 12 also showed well-defined 1:1 stoichiometries (Supplementary Figures S9-S11), further confirming favourable 1-FM interactions.

Selection of Crosslinker
XL makes up about 80% of a MIP formulation and potentially has the greatest influence on non-specific binding in MIPs. Thus, to further our molecular modelling and NMR studies, the affinity of 1 to EGDMA, TRIM and DVB polymers was measured. These crosslinked polymers were prepared in the absence of 1 and any FM in acetonitrile (AN) and CHCl 3 , the chosen porogens for 1-MIPs. Although 14 has been successfully used both as a functional monomer and crosslinker in OmniMIPs [49] and our MM and NMR data suggest that it could potentially work as a crosslinker (though not as a functional monomer) for 1-MIP, we excluded 14 in this study. High levels of non-specific binding of 1 were observed in DVB polymers, two to four times higher than in EDGMA and TRIM polymers. Of the three crosslinkers, only DVB is capable of π-π stacking interaction with 1, and this is most likely the reason for the observed high level of superficial binding of 1.

Preparation of 1-MIPs
Having identified 6, 7 and 12 as suitable FMs from MM and NMR studies, the corresponding 1-MIPs were prepared in T:FM ratios of 1:1, 1:2 and 1:4 in CH 3 CN and CHCl 3 as porogens and EGDMA and TRIM as crosslinkers. The T:FM ratios of 1:1 and 1:2 were chosen as per our MM and NMR results, showing the most favourable interaction between 1 and FMs 6, 7 and 12 at these ratios. The 1:4 T:FM ratio is the most common ratio used in the literature to promote T-FM interactions and complex formation. The 1-MIPs were prepared in acetonitrile and chloroform to determine the effect of porogen polarity on the binding capabilities of the MIPs. Similarly, the influence of the level of crosslinking on the binding efficiency of 1-MIPs was also evaluated by employing di-(EGDMA) and tri-(TRIM) XLs.
All 36 formulations of 1-MIPs were prepared using 7 mL of porogen (~1 mmol FM + XL per 1 mL), with AIBN as the initiator, at 60 • C for 12 h. The resulting MIPs were ground with the fraction 32-63 µm collected and Soxhlet extracted with 10% acetic acid in methanol, followed by pure methanol. Extraction was repeated until 1 could no longer be detected in the HPLC trace of the methanol extract.
Subsequent reference to these 1-MIP formulations follows the following codes: E and T for EGDMA and TRIM XLs, respectively; 6, 7 and 12 for the corresponding FM, 1, 2 or 4 for 1:1, 1:2 and 1:4 T:FM ratios, respectively; and CH 3 CN or CHCl 3 for the porogens. Thus, E7 1-MIPCHCl3 refers to a 1-MIP prepared using EGDMA as XL and 7 as FM in a 1:1 T:FM ratio, with chloroform as the porogen.

Physical Characterisation
Scanning electron microscopy was used to examine the surface morphology of both MIPs and NIPs (non-imprinted polymer). No significant difference in surface morphology was observed between MIPs and their corresponding NIPs. MIP surface morphology was unaffected by variations in the FM or T:FM ratio, but differences based on XL were apparent (Supplementary Figures S16 and S17). The TRIM-based MIPs appeared to be more compact, more dense and less porous than the surface of the EGDMA crosslinked polymers (Supplementary Figure S18). This was consistent with previous reports, which indicate that the nature of the crosslinker can affect polymer surface morphology [50]. It is also clear from Figure 3 that using CH 3 CN as the porogen produced a macroporous surface morphology, while CHCl 3 produced smoother, more dense surfaces with fewer visible pores. This most likely resulted from a delayed phase separation from CHCl 3 ; that is, the growing polymer was more soluble in CHCl 3 than in CH 3 CN, a product of solvent-polymer polarity mismatch [51]. The porogen effect on morphology was evident in swelling experiments, wherein all CHCl 3 polymers displayed an enhanced swelling capacity relative to the equivalent CH 3 CN polymers. E71-MIPCHCl3 refers to a 1-MIP prepared using EGDMA as XL and 7 as FM in a 1:1 T:FM ratio, with chloroform as the porogen.

Physical Characterisation
Scanning electron microscopy was used to examine the surface morphology of both MIPs and NIPs (non-imprinted polymer). No significant difference in surface morphology was observed between MIPs and their corresponding NIPs. MIP surface morphology was unaffected by variations in the FM or T:FM ratio, but differences based on XL were apparent (Supplementary Figures S16 and S17). The TRIM-based MIPs appeared to be more compact, more dense and less porous than the surface of the EGDMA crosslinked polymers (Supplementary Figure S18). This was consistent with previous reports, which indicate that the nature of the crosslinker can affect polymer surface morphology [50]. It is also clear from Figure 3 that using CH3CN as the porogen produced a macroporous surface morphology, while CHCl3 produced smoother, more dense surfaces with fewer visible pores. This most likely resulted from a delayed phase separation from CHCl3; that is, the growing polymer was more soluble in CHCl3 than in CH3CN, a product of solventpolymer polarity mismatch [51]. The porogen effect on morphology was evident in swelling experiments, wherein all CHCl3 polymers displayed an enhanced swelling capacity relative to the equivalent CH3CN polymers.

Evaluation of the Imprinting Effect
The rebinding of 1 was initially evaluated by batch adsorption experiments, using 0.8 mM solution of 1 in CH3CN or CHCl3 with 30 min exposure to polymers of various masses (5, 10, 20 and 30 mg). As expected, the rebinding capacity was a function of the polymer mass used, typically plateauing at 20-30 mg polymer loading and increasing with increasing proportion of FM for both MIPs and NIPs. Our data suggest that the rebinding of 1 was not affected by the nature of the crosslinker; instead, it was porogen and FM dependent. In all instances, higher rebinding was observed with the CH3CN MIPs, but the imprinting effect was enhanced in CHCl3 MIPs, as a consequence of lower NIP rebinding of 1.
This initial rebinding assays allowed us to screen MIPs and select those with an imprinting efficacy or imprinting factor IF (MIP binding/NIP binding) of ≥2 for all mass loadings for further binding characterisation. This imprinting efficacy criterion was obtained with MIPs E71-MIPCHCl3, E72-MIPCHCl3, T71-MIPCHCl3 and T72-MIPCHCl3, with CHCl3 as binding solvent; and E71-CH3CN, bound in CH3CN ( Figure 4). Template uptake by these polymers

Evaluation of the Imprinting Effect
The rebinding of 1 was initially evaluated by batch adsorption experiments, using 0.8 mM solution of 1 in CH 3 CN or CHCl 3 with 30 min exposure to polymers of various masses (5, 10, 20 and 30 mg). As expected, the rebinding capacity was a function of the polymer mass used, typically plateauing at 20-30 mg polymer loading and increasing with increasing proportion of FM for both MIPs and NIPs. Our data suggest that the rebinding of 1 was not affected by the nature of the crosslinker; instead, it was porogen and FM dependent. In all instances, higher rebinding was observed with the CH 3 CN MIPs, but the imprinting effect was enhanced in CHCl 3 MIPs, as a consequence of lower NIP rebinding of 1.
This initial rebinding assays allowed us to screen MIPs and select those with an imprinting efficacy or imprinting factor IF (MIP binding/NIP binding) of ≥2 for all mass loadings for further binding characterisation. This imprinting efficacy criterion was obtained with MIPs E7 1-MIPCHCl3 , E7 2-MIPCHCl3 , T7 1-MIPCHCl3 and T7 2-MIPCHCl3 , with CHCl 3 as binding solvent; and E7 1-CH3CN , bound in CH 3 CN (Figure 4). Template uptake by these polymers varied from 17 to 75% and from 3 to 26% across the MIPs and their corresponding NIPs, respectively.

Binding Isotherms
The minimum time required to reach optimal template binding for MIPs E7 1-MIPCHCl3 , E7 2-MIPCHCl3 , T7 1-MIPCHCl3 and T7 2-MIPCHCl3 was determined prior to undertaking any other binding assays. No marked difference in the rebinding of 1 was observed for incubation times between 30 min and 24 h. Thus, we opted to use 60 min for subsequent binding tests. As only one MIP from the CH 3 CN series (E7 1-CH3CN ) exhibited noteworthy imprinting, no further analysis of this series was undertaken. varied from 17 to 75% and from 3 to 26% across the MIPs and their corresponding NIPs, respectively.

Binding Isotherms
The minimum time required to reach optimal template binding for MIPs E71-MIPCHCl3, E72-MIPCHCl3, T71-MIPCHCl3 and T72-MIPCHCl3 was determined prior to undertaking any other binding assays. No marked difference in the rebinding of 1 was observed for incubation times between 30 min and 24 h. Thus, we opted to use 60 min for subsequent binding tests. As only one MIP from the CH3CN series (E71-CH3CN) exhibited noteworthy imprinting, no further analysis of this series was undertaken.
The binding isotherms were generated by varying the rebinding solution concentration of 1 between 0.5 and 10 mM, using a constant polymer loading of 20 mg and 60 min incubation (Supplementary Figure S19A-D). Maximum binding (Bmax) and affinity constants (Kd) were calculated from non-linear regression of the Langmuir binding isotherms, using GraphPad Prism 9.4.1; the best-fit values obtained from the one-site-total-binding equation are presented in Table 1. These data show MIPs bound twice the amount of 1 than their corresponding NIPs, and this is attributable to the imprinting effect. E72 and T72 polymers (MIPs and NIPs) bound twice the amount of 1 than E71 and T71 polymers, a result that is unsurprising given that the E72 and T72 feed formulations contain twice the amount of functional monomer 7.
The Kd values for the MIPs (0.22-0.62 µM) were consistently lower than the corresponding NIPs (0.71-1.52 µM) because of the higher affinity template binding arising from the imprinting process. Furthermore, the Kd values were higher for the 1:1 (E71 and T71) than for the 1:2 (E72 and T72) formulations, suggesting that the affinity for 1 was stronger at a higher functional monomer content. The NIP-to-MIP Kd ratios were higher for the EGDMA-based polymers (∼5) than for the TRIM-based polymers (∼2) and were consistent with the rebinding results and imprinting factors presented in Figure 4. The binding isotherms were generated by varying the rebinding solution concentration of 1 between 0.5 and 10 mM, using a constant polymer loading of 20 mg and 60 min incubation (Supplementary Figure S19A-D). Maximum binding (B max ) and affinity constants (K d ) were calculated from non-linear regression of the Langmuir binding isotherms, using GraphPad Prism 9.4.1; the best-fit values obtained from the one-site-total-binding equation are presented in Table 1. These data show MIPs bound twice the amount of 1 than their corresponding NIPs, and this is attributable to the imprinting effect. E7 2 and T7 2 polymers (MIPs and NIPs) bound twice the amount of 1 than E7 1 and T7 1 polymers, a result that is unsurprising given that the E7 2 and T7 2 feed formulations contain twice the amount of functional monomer 7.  Figure S19), using the one-site-total-binding equation generated using GraphPad Prism 9.4.1.
The K d values for the MIPs (0.22-0.62 µM) were consistently lower than the corresponding NIPs (0.71-1.52 µM) because of the higher affinity template binding arising from the imprinting process. Furthermore, the K d values were higher for the 1:1 (E7 1 and T7 1 ) than for the 1:2 (E7 2 and T7 2 ) formulations, suggesting that the affinity for 1 was stronger at a higher functional monomer content. The NIP-to-MIP K d ratios were higher for the EGDMA-based polymers (~5) than for the TRIM-based polymers (~2) and were consistent with the rebinding results and imprinting factors presented in Figure 4.

Evaluation of Imprinting Effect
E16 MIPCHCl3 and T16 MIPCHCl3 were evaluated for their ability to selectively rebind 1 by batch adsorption experiments, using 1 mL of 0.8 mM solution of 1 in CHCl 3 and an incubation time of 30 min (as with the self-assembly MIPs). Both MIPs displayed a steady increase in template rebinding as a function of increasing polymer loading (from 5 mg to 30 mg), as is consistent with binding-site homogeneity [60]. The NIPs, on the other hand, displayed minimal binding, and this only slightly increased across all polymer loadings evaluated ( Figure 6).
sponding NIPs were prepared under identical conditions, using the XLs only.
SEM micrographs of the MIPs displayed clear evidence of a macroporous surface structure (Supplementary Figure S20) consistent with the surface structure expected from low-polarity solvents such as CHCl3 [18,51,[56][57][58][59]. Both E16MIPCHCl3 and T16MIPCHCl3 exhibited low levels of swelling, with volume increases of 3-4% only upon exposure to the porogen CHCl3.

Evaluation of Imprinting Effect
E16MIPCHCl3 and T16MIPCHCl3 were evaluated for their ability to selectively rebind 1 by batch adsorption experiments, using 1 mL of 0.8 mM solution of 1 in CHCl3 and an incubation time of 30 min (as with the self-assembly MIPs). Both MIPs displayed a steady increase in template rebinding as a function of increasing polymer loading (from 5 mg to 30 mg), as is consistent with binding-site homogeneity [60]. The NIPs, on the other hand, displayed minimal binding, and this only slightly increased across all polymer loadings evaluated ( Figure 6).  T16 MIPCHCl3 showed a greater affinity for 1, binding at least twice the amount absorbed by E16 MIPCHCl3 , with maximum binding at 30 mg polymer loading, resulting in 49% and 26% uptake of 1, respectively. This result is in keeping with previous reports suggesting that XLs with more than two polymerizable groups, such as TRIM, result in a more porous macrostructure with improved mass transfer properties (better access to the imprinted cavities) and higher capacities [61,62].
The imprinting effect, measured by the imprinting factor (IF; amount bound by MIP/amount bound NIP) ranged from 5.1 to 21, results consistent with the presence of well-defined 1-specific cavities ( Figure 6). It should be noted, however, that the NIPs lacked the FM component usually associated with NIP synthesis, as such, producing a poorer mimic on a non-imprinting effect, but nonetheless one that has been used in semi-covalent imprinting previously [52,63].

Binding Isotherms
Uptake of 1 was observed to be rapid, with optimal binding noted after 10 min for both MIPs. Binding isotherms (Supplementary Figure S19E) were generated across a 0.3-10 mM concentration range (of 1) in CHCl 3 , using 30 mg of polymer and a 60 min rebinding time in order to be consistent with the self-assembly systems. The binding constant, K d , and B max were calculated using non-linear regression of the Langmuir binding isotherms generated from GraphPad Prism, with the best-fit values obtained from the one-site-total-binding equation presented in Table 1. Since the NIPs showed negligible binding of 1, only the MIPs were analysed. The K d value for T16 MIPCHCl3 (0.02 µM) is 5 times lower than E16 MIPCHCl3 (0.09 µM), as is consistent with earlier observations (Figure 6) that the former has a greater affinity for 1. These K d values are also 3 and 33 times lower than the equivalent self-assembly MIPs (E7 1-MIPCHCl3 and T7 1-MIPCHCl3 ), respectively. The binding capacities (B max ) of these semi-covalent MIPs are close in values and comparable to T7 1-MIPCHCl3 but 3 times lower than E7 1-MIPCHCl3 .
ing time in order to be consistent with the self-assembly systems. The binding constant, Kd, and Bmax were calculated using non-linear regression of the Langmuir binding isotherms generated from GraphPad Prism, with the best-fit values obtained from the onesite-total-binding equation presented in Table 1. Since the NIPs showed negligible binding of 1, only the MIPs were analysed. The Kd value for T16MIPCHCl3 (0.02 μM) is 5 times lower than E16MIPCHCl3 (0.09 μM), as is consistent with earlier observations (Figure 6) that the former has a greater affinity for 1. These Kd values are also 3 and 33 times lower than the equivalent self-assembly MIPs (E71-MIPCHCl3 and T71-MIPCHCl3), respectively. The binding capacities (Bmax) of these semi-covalent MIPs are close in values and comparable to T71-MIPCHCl3 but 3 times lower than E71-MIPCHCl3.

Selectivity Studies
Both self-assembly (E71-MIPCHCl3, E72-MIPCHCl3, T71-MIPCHCl3 and T72-MIPCHCl3) and semi-covalent (E16MIPCHCl3 and T16MIPCHCl3) MIP systems were subjected to selectivity tests-single component (cross-reactivity) and binary competitive rebinding of 1 against cocaine (17), morphine (18), ephedrine (19) and phenylpiperazine (20) (Figure 7). These tests were conducted using 20 mg of polymer in 0.8 mM solutions of analyte in CHCl3, with an incubation time of 60 min. A range of illicit drugs and precursors containing similar functionalities were chosen as competing analytes to simulate in-field capability for illicit detection.

Cross-Reactivity Tests
The results for the non-competitive cross-reactivity assays are given in Supplementary Figure S21, while Table 2 shows the cross-reactivity indices (XRF) to enable a direct comparison of the uptake of each competing analyte by the MIP with respect to 1. The

Cross-Reactivity Tests
The results for the non-competitive cross-reactivity assays are given in Supplementary Figure S21, while Table 2 shows the cross-reactivity indices (XRF) to enable a direct comparison of the uptake of each competing analyte by the MIP with respect to 1. The XRF for an analyte is defined as the ratio of the analyte (17, 18, 19 or 20) to 1 MIP binding within the MIP. Table 2. Non-competitive cross-reactivity results for 1-MIPs against cocaine (17), morphine (18), ephedrine (19) and phenylpiperazine (20 The four self-assembly MIPs (E7 1-MIPCHCl3 , E7 2-MIPCHCl3 , T7 1-MIPCHCl3 and T7 2-MIPCHCl3 ) displayed high levels of cross-reactivity with 19 and 20, with the XRFs ranging from 0.82 to 1.44, low to moderate with 18 (XRF = 0.23 to 0.43) and negligible with 17 (XRF = 0.08 to 0.15). Selective binding (∆B = MIP binding − NIP binding) followed the same trend. These results reflect the effect of functional group character (electronic, size), relative proximity and orientation on each analyte with respect to 1. Analytes 17 and 18 are both large molecules that differ in shape, potential and available functional groups to 1, leading to a poor 'fit' within the imprinted cavities (Supplementary Figure S22). Consequently, they displayed low sorption and low affinity to the MIPs. In contrast, 19 and 20 are similar in size and chemical character to 1, consistent with the observed high XRFs (Supplementary Figure S23). The very high uptake of 19, equivalent or higher than 1, could be attributed to multipoint H-bonding interactions with the MAA FM by virtue of its NH and OH groups, which are H donor/acceptor species, leading to potentially greater binding affinity. Furthermore, analyte 19 is the only test target containing an amine unit that is not part of a ring system, which could mean greater fluxionality, resulting in better orientation within the binding cavity.
The semi-covalent MIPs (E16 MIPCHCl3 and T16 MIPCHCl3 ) showed a cross-reactivity trend that is similar to the self-assembly MIPs, except for analyte 17, which recorded high levels of cross-reactivity (XRF's 0.64 and 0.85 for E16 MIPCHCl3 and T16 MIPCHCl3 , respectively) in contrast to its behaviour towards the self-assembly MIPs. Note that the post-polymerisation cleavage of the carbothioate linker of 16 ( Figure 5) to release the template moiety leaves a phenol functionality within the cavity capable of interacting with 1 or other complementary analytes. The XRFs for 19 and 20 are slightly lower than those observed with the self-assembly MIPs, thus suggesting a less favourable affinity to the imprinted cavities containing the phenol functionality than with the MAA carboxyl unit. The unexpected high cross-reactivity with 17 could possibly be due to its benzoyl group, which is similar in structure and size to the benzyl group of 1, that could easily fit the imprinted cavities and interact with the phenol group by pi-pi stacking.

Binary Competitive Assays
The selectivity of the MIPs was further tested using binary mixtures of 1 and competing analytes 17, 18 and 19. Analyte 20 was not pursued in this assay due to HPLC separation issues with 1.
The uptake of all analytes was observed to be higher for MIPs than in their corresponding NIPs (Supplementary Figure S24), thus suggesting that they have access to and an affinity for the imprinted cavities. To facilitate binding selectivity analysis, we introduced the quantitative values summarized in Table 3. First, the uptake of 1 as the sole analyte, a measure of the MIP binding capacity, is taken as a reference; the binding of binary analytes is then normalised against this reference. Thus, if the normalised binding of an analyte is greater than 1, i.e., greater than 100%, then this means that its uptake is higher than the expected binding capacity. The selectivity index (SI) of an analyte is defined here as the ratio of normalised analyte binding to normalised 1 binding.
Although selectivity assessment is more essential for MIPs than NIPs, and despite the fact that NIP binding could be variable due to its superficial nature, the results of this study showed analyte 19 to be competitive with 1 in all NIP systems, giving normalised binding values as high as 3.15. The binding of 17 and 18, however, was always lower or equivalent to that of 1. Uptake of 1 by the semi-covalent NIPs is lower than the reference in all cases, while self-assembly NIPs registered 1 binding higher than the reference in some binary mixtures. Table 3. Binary competition results for 1-MIPs against cocaine (17), morphine (18) and ephedrine (19). In the case of MIPs, 19 was also observed to be highly competitive, reducing the uptake of 1 to a low of 0.42, i.e., 42% (T16 MIPCHCL3 ), with respect to the reference and displayed normalised binding and SIs higher than 1 (as high as 1.73 and 1.85, respectively, with E7 1-MIPCHCL3 ) in all cases. Analyte 18 displayed low levels of competition with 1 in the self-assembly MIPs (SI = 0.16 to 0.37), slightly reducing the binding of 1 in some cases. However, the presence of 18 brought about a significant reduction in 1 binding (below 50%), resulting in a moderate SI of 0.46 (E16 MIPCHCL3 ) and 0.61 (T16 MIPCHCL3 ) in the case of the semi-covalent MIPs. Analyte 17 did not have a significant impact on the binding of 1, except in T16 MIPCHCL3 , and displayed low-to-minimal competition tendency against 1, with normalized binding and SI values all being <0.2. In some instances (also observed with the NIPs), the presence of 17 and 18 enhanced the uptake of 1, which could be attributed to its interaction with these analytes by hydrogen bonding and/or pi-pi stacking. On the other hand, in the presence of 19, the uptake of 1 in all polymers tested was reduced, thus confirming its observed competitive tendency against 1.
Consistent with the results of the non-competitive cross-reactivity assays, the results of the competitive (binary mixtures) binding studies showed all MIPs to be highly selective towards 1 in the presence of 17 (SI < 0.2), moderately selective in the presence of 18 (SI > 0.2 to~0.6) and non-selective in the presence of 19 (SI > 1).

Selectivity of 1-MIPs: Implication on Their Applications
The poor-to-moderate selectivity of 1-MIPs against 18, 19 and 20 could initially be construed as unfavourable; however, these results indicate that 1-MIPs are capable of broadbased capture of blends of illicit substances with, rather than specifically for, 1, comparable to the 'dummy' MIP approach. The high cross-reactivity against 20 means that 1-MIPs could potentially extract 1-(3-trifluoromethyl-phenyl)piperazine (TFMPP, 3), a derivative of 20 most commonly blended with 1, and other known substituted phenylpiperazine blend ingredients such as 4, 5 ( Figure 1) and 4-fluorophenylpiperazine [64]. Other pills are also known to be mixed with cocaine (17) [16,65] and ephedrine (19) [64], which, according to our selectivity studies, could also be extracted by 1-MIPs.
Illicit drugs are never pure, and conducting an analysis of their composition, including minor ingredients and adulterants, is essential in drug profiling studies [66]. The competitive uptake of 18, 19 and 20 by 1-MIPs means that, when present in trace amounts, they could be pre-concentrated within the MIP, which could enhance their detection. We speculate that other minor non-piperazine derivatives blended with 1 could also be captured and pre-concentrated with 1-MIPs. Unfortunately, due to the nature of these analytes, we could not easily obtain real samples to test.
Our results suggest no significant difference in the selectivity between self-assembly and semi-covalent 1-MIPs. Thus, the MIP materials could be prepared by the synthetically simpler self-assembly imprinting using cheap commercially available monomers (monomer 7 and crosslinkers), thus making these materials very competitive over other analytical extraction methods.
Our 1-MIP materials are suited to laboratory test setting, with potential as the recognition element for in-field sensing devices. Sample preparation would be simple, and with their enhanced enrichment capability, due to the presence of imprinted sites, they would be useful for the capture of illicit drugs, particularly in biological samples. MIPbound drug analytes could be re-extracted and differentiated by a number of analytical procedures-notably, chromatographic methods-preferably with MS detection and capillary electrophoresis, among others.
HPLC-grade acetonitrile and chloroform were obtained from Merck and were used as received. All other solvents were distilled prior to use, unless otherwise stated.

Preparation of N,O-Bismethacryloyl Ethanolamine (NOBE, 14)
NOBE (14) was prepared as per the method utilised by Sibrian-Vazquez and Spivak [49]. Ethanolamine (0.976 g, 16 mmol) was mixed with 15 mL of dichloromethane. TEA (3.74 g, 5.15 mL, 37 mmol) was added in small portions to the initial mixture, with stirring, and the reaction mixture was cooled to 0 • C. Methacryloyl chloride (3.867 g, 3.6 mL, 37 mmol) was added dropwise, with vigorous stirring, while keeping the temperature at 0 • C. After the complete addition of methacryloyl chloride, the temperature was increased to 40 • C and allowed to react for 24 h at this temperature. The reaction mixture was filtered, and the precipitate (Et3NHCl) was then discarded. The filtrate was extracted with 0.5 M NaHCO3 (3 × 15 mL) and 0.5 M sodium citrate (3 × 15 mL). The solvent was evaporated under vacuum, the compound was isolated by column chromatography (EtOAc/hexanes 50:50, EtOAc 100%), and 14 was isolated as a pale yellow oil. Yield: 59%. 1

Preparation of 7-Hydroxy-4-methylcoumarin Acrylate (15)
First, 15 was prepared by the literature method [67]. TEA (12.1 g, 120 mmol) and 7hydroxy-4-methylcoumarin (8.809 g, 50 mmol) were dissolved in chloroform (150 mL) and cooled to 0 • C. Acryloyl chloride (10.9 g, 120 mmol) was added dropwise, with vigorous stirring. The reaction mixture was then returned to room temperature and stirred for a further 12 h. The solvent was evaporated under vacuum, and the product was purified by dissolving the precipitate in methanol (200 mL). The methanol solution was then poured into water (1 L), the precipitate was collected by filtration, and 15 was isolated as a white flake. 1

Molecular Modelling
Template-monomer molecular interactions were modelled using Spartan '04 software, using the AM1 force field. This molecular orbital computational method predicts the stable configuration of the template (T), functional monomer (FM), FM-FM clusters and T-FM clusters and calculates their standard heats of formation (∆H f ). The molecules were randomly positioned, and the T-FM clusters were modelled with respect to increasing the template-monomer ratio from 1 to 4. To account for the FM-FM interaction, the FM-FM clusters of up to five molecules were also surveyed. The energies of interaction of the T-M clusters, ∆E • (cluster), at different molecular ratios were then calculated using the following equation: 3.6. NMR Spectroscopic Analysis 1 H and 13 C Nuclear Magnetic Resonance (NMR) spectra were recorded at 300.13 and 75.47 MHz, respectively, using a Brüker Advance 300 MHz Spectrometer in conjunction with Brüker Topspin v1.3 software. Experiments involving 6, 7 and 14 were performed in CDCl 3 , while experiments with 12 were performed in DMSO-d 6 at a temperature of 301 K. For the NMR titration, molar aliquots of monomer were sequentially added to the template 1 (0.1 mmol) up to a maximum of 16 equivalents. After each aliquot addition, the sample was mixed and allowed to spin for five minutes before spectrum acquisition. The experiment was repeated in the absence of 1. For the Job experiments, 11 samples were prepared with varying 1 and monomer molar ratios, ranging from 0 to 1, using 0.2 mM solutions. The total volume was constant at 0.5 mL.

Self-Assembly MIPs
The required amounts of functional monomer (0.34 mmol, 0.68 mmol or 1.36 mmol) and crosslinker (6.64 mmol, equivalent to 20 × T) were added to a solution of 1 (60 mg, 0.34 mmol) in 7 mL porogen (CH 3 CN or CHCl 3 ). Based on our MM-NMR analysis, 6, 7 and 12 were selected as functional monomers and EGDMA and TRIM as crosslinkers. MIPs were prepared using 1:1, 1:2 and 1:4 T:M ratios in chloroform and acetonitrile. The reaction mixture was degassed with N 2 before AIBN (50 mg) was added. The mixture was heated to 60 • C in an oven (Thermoline). NIPs were prepared by using the same method but without the addition of 1.
Polymers were ground wet in methanol and sieved with the fraction between 32 and 65 µm collected. Template removal was by Soxhlet extraction, using a 10% acetic acid-methanol mix for 48 h, followed by 100% methanol for 12 h. The polymers were dried at 40 • C for 24 h.
Polymers were ground wet in methanol and sieved with the fraction between 32 and 65 µm collected. The template adduct was cleaved by heating the polymer at reflux over 1 M NaOH for 12 h and then neutralised with dilute HCl. The polymers were then washed with methanol for 12 h, using a Soxhlet extractor. Finally, the polymers were dried at 40 • C for 24 h [48,49,55].

Batch-Binding Tests for 1 3.8.1. HPLC Analytical Method
Batch rebinding experiments were carried out using a known concentration of 1 stock solution in either acetonitrile or chloroform. The required mass of polymer was left in contact with the 1 solution for the required time. The quantification of 1 was achieved by HPLC, using a Shimadzu High Performance Liquid Chromatograph (HPLC) (LC-20AD) fitted with an EconosphereTM C18, 5 µm column (Grace ® ).
For 1 binding in CH 3 CN, the mobile phase comprised 50% CH 3 CN and 50% buffer solution (25 mM K 2 HPO 4 ; 30 mM KCl; 7 mM TEA; adjusted to pH 3 with H 3 PO 4 ). A 10 µL injection volume was used with a run time of 10 min, flow rate of 2 mL min −1 and detection wavelength of 254 nm.
For binding in CHCl 3 , the mobile phase consisted of 70% CH 3 CN and 30% buffer solution (25 mM K 2 HPO 4 ; 30 mM KCl; 7 mM TEA; adjusted to pH3 with H 3 PO 4 ). A 10 µL injection volume was used with a run time of 15 min, flow rate of 0.95 mL.min-1 and detection wavelength of 254 nm. A calibration curve was generated using six solutions in the range of 0.1 to 0.8 mM.

Sorption Tests: Evaluation of Imprinting Effect
A sorption study to evaluate the imprinting efficiency was performed. Various polymer masses from 10.0 to 30.0 mg were placed into 5 mL vials to which was added 1.00 mL of 0.0800 mM 1 in CH 3 CN or chloroform. The mixture was shaken for 30 min, filtered and the filtrate analysed directly by HPLC. The amount of free 1 was subtracted from the initial binding solution concentration to obtain the amount of 1 bound in the polymer. All binding experiments for this study were performed in triplicate to ensure reproducibility.
Results of the sorption tests were favourable for only 4 of the self-assembly systems tested: E7 1-MIPCHCl3 , E7 2-MIPCHCl3 , T7 1-MIPCHCl3 and T7 2-MIPCHCl3 MIPs. The other systems were no longer pursued in subsequent studies.

Time-Binding Study
To a set of triplicates of 30.0 mg of self-assembly systems E7 1-MIPCHCl3 , E7 2-MIPCHCl3 , T7 1-MIPCHCl3 and T7 2-MIPCHCl3 , and semi-covalent MIPs E16 MIPCHCl3 and T16 MIPCHCl3 polymers, 1.00 mL of 0.0800 mM 1 was added and the mixture shaken for a designated time of contact. The binding times investigated were 0.5, 2.0, 4.0, 7.0 and 18 h. After binding, the mixtures were filtered, and the filtrates were analysed by HPLC. The amount of bound 1 was then obtained by subtracting the amount of 1 left in solution from the initial concentration.

Saturation Binding
A series of 20.0 mg of self-assembly polymers E7 1-MIPCHCl3 , E7 2-MIPCHCl3 , T7 1-MIPCHCl3 and T7 2-MIPCHCl3 were incubated in different concentrations of 1 for 1 h, after which, the mixtures were filtered, and the filtrates were analysed directly by HPLC. The amount of bound 1 was then obtained by subtracting the amount of 1 left in solution from the initial concentration. Binding isotherms were generated from GraphPad Prism 9.4.1, and the best-fit values from non-linear regression, using the one-site-total-binding equation, were obtained.
The same procedure was followed for the semi-covalent MIPs E16 MIPCHCl3 and T16 MIPCHCl3 , except that 30.0 mg of polymers was used for all binding measurements. The HPLC method for 1, 18 and 20 is outlined in Section 3.8.1. Analyte 17 was analysed using a Shimadzu 2010 gas chromatograph coupled to a Shimadzu QP2010 mass spectrometer and a Shimadzu AOC-20s auto sampler. The column was a ZB-5MS capillary column (30 m × 0.25 mm) coated with 0.25 µm of stationary phase. High-purity helium was used as the carrier gas at 71 kPa, with a column flow rate of 1 mL/min, a total flow rate of 9 mL/min and a split ratio of 15. Samples (1 µL) were injected and run using the following program: initial column temperature was 100 • C, which was held for 1 min before increasing to 300 • C at a rate of 10 • C/min. Analyte 19 was analysed using the following HPLC method: The mobile phase consisted of 75% aqueous buffer solution (50 mM K 2 HPO 4 adjusted to pH 3.5 with H 3 PO 4 ) and 25% 3:7 H 2 O:CH 3 CN (with 10 mM TEA). A 10 µL injection volume was used with a run time of 10 min, flow rate of 0.8 mL min −1 and detection wavelength of 190 nm. A calibration curve was generated using solutions in a range from 0.1 to 1 mM.

Scanning Electron Microscopy
Polymer morphology was examined using a Phillips XL30 scanning electron microscope. The sample was deposited on a sticky carbon tab and coated with gold, using a SPI gold spotter coating unit. SE micrographs of the polymers were obtained at 20,000× magnification at 15.0 kV.

Swelling Measurements
A total of 30 (3) mg of each polymer was packed into an NMR tube, and the height of the dry polymer measured. A solution of 1 (1.00 mL of 0.0800 mM) in acetonitrile or chloroform was added and allowed to soak for 24 h. Polymers were allowed to settle, and the bed height of the swollen polymers was measured. The swelling factor was calculated from the ratio of the bed height of the swollen polymer to the dry polymer.

Conclusions
Using self-assembly (non-covalent) and semi-covalent methods, we designed 1-specific MIPs. In the case of the self-assembly MIPs, a range of potential functional monomers (FM) were screened using a combination of pre-synthetic interaction studies (by molecular modelling and NMR analysis) and binding assays. The best performing self-assembly 1-MIPs gave IFs of 3 to 7 and were formulated from FM 7 with ethylene glycol dimethacrylate (EGDMA) and trimethylolpropane trimethacrylate (TRIM) crosslinkers, using chloroform as porogen and rebinding solvent at T:FM ratios of 1:1 (E7 1-MIPCHCl3 and T7 1-MIPCHCl3 ) and 1:2 (E7 2-MIPCHCl3 and T7 2-MIPCHCl3 ). The binding parameters K d and B max were consistent with the MIPs exhibiting a stronger affinity towards 1 (lower K d ), resulting in greater number of binding sites (higher B max ) than their corresponding NIPs and observed to be higher for the 1:2 than the 1:1 T:FM formulations. The imprinting effect, as per IFs and NIP/MIP K d ratios, was observed to be higher for the EGDMA-based MIPs than for the TRIM-based MIPs. The semi-covalent 1-MIPs were designed using O-4-vinyl 4benzylpiperazine-1-carbothioate or benzylpiperazine (4-vinylphenyl) carbamate (16) as the template-monomer (TM) adduct copolymerized with either EDGMA (E16 MIPCHCl3 ) or TRIM (T16 MIPCHCl3 ). We found that 1 could be cleaved from 16 post-polymerisation, leaving a phenol moiety within the imprinted sites capable of hydrogen bonding with 1 upon re-exposure. T16 MIPCHCl3 exhibited a greater affinity for 1-higher IFs, higher binding capacities and lower K d than E16 MIPCHCl3 . The K d values for semi-covalent MIPs are significantly lower than their self-assembly equivalents, while their B max are, at the most, comparable.
The self-assembly MIPs displayed high levels of cross-reactivity with 19 and 20, low to moderate with 18 and marginal with 17. The analytes 17 and 18 are both large molecules that differ in shape, potential and available functional groups to 1 and are therefore a poor fit for the imprinted cavities. In contrast, 19 and 20 are similar in size and chemical character to 1, consistent with the observed high cross-reactivity. The semi-covalent MIPs showed similar cross-reactivity trend to the self-assembly MIPs, except for analyte 17, which recorded high levels of cross-reactivity. The unexpected high cross-reactivity with 17 could possibly be due to its benzoyl group, which is similar in structure and size to the benzyl group of 1, and could easily fit the imprinted cavities and interact with the phenol moiety. Consistent with the results of the non-competitive cross-reactivity assays, the results of the competitive (binary mixtures) binding studies showed both self-assembly and semi-covalent MIPs to be highly selective towards 1 in the presence of 17, moderately selective in the presence of 18 but non-selective in the presence of 19. Competition of 1 with 20 was not studied due to separation problem.
The semi-covalent MIPs were observed to have a stronger affinity for 1 and faster uptake than the self-assembly systems. Both approaches gave MIPs of comparable binding selectivity and cross-reactivity. Overall, the observed IF values were significantly higher with the semi-covalent MIPs than with their self-assembly equivalents; however, it is worth noting that the semi-covalent NIP reference does not contain any FM, thus resulting in minimal 1 binding.
By virtue of its modest selectivity against the test for illicit drugs, 1-MIP, could potentially be used as a dummy MIP for the broad-based capture and enrichment of illicit substances blended with 1, e.g., 19 and substituted 20, for subsequent laboratory analysis. Our preliminary data also demonstrated high affinity for 1 in an aqueous environment, thus raising the possibility for its use in illicit-drug testing.