Molecularly Imprinted Micelles for Fluorescent Sensing of Nonsteroidal Anti-Inflammatory Drugs (NSAIDs)

Abstract

Nonsteroidal anti-inflammatory drugs (NSAIDs) are widely used over-the-counter drugs and their uncontrolled disposal is a significant environmental concern. Although their fluorescent sensing is a desirable method of detection for its sensitivity and simplicity, the structural similarity of the drugs makes the design of selective sensors highly challenging. A thiourea-based fluorescent functional monomer was identified in this work to enable highly efficient synthesis of molecularly imprinted nanoparticle (MINP) sensors for NSAIDs such as Indomethacin or Tolmetin. Micromolar binding affinities were obtained in aqueous solution, with binding selectivities comparable to those reported for polyclonal antibodies. The detection limit was ~50 ng/mL in aqueous solution, and common carboxylic acids such as acetic acid, benzoic acid, and citric acid showed negligible interference.

1. Introduction

Nonsteroidal anti-inflammatory drugs (NSAIDs) are the most commonly administered over-the-counter drugs to reduce pain, decrease fever, and control inflammation [1]. Due to their potential effects on different organisms, their wide usage and uncontrolled disposal are a significant environmental concern [2, 3]. Although many NSAIDs are degraded by microorganisms in soil, rate of degradation varies greatly depending on the types of soil and other environmental conditions [4].

The important biological activity of NSAIDs and their environmental risks have motivated many researchers to develop sensitive methods for their detection and monitoring. Traditional analyses include electrophoresis [5] and liquid chromatography–mass spectrometry [6] but the sophisticated instruments utilized are inconvenient for rapid in-field monitoring. Recognition-based fluorescent sensing is a method of choice for its simplicity, high sensitivity, and ease of operation [7–11]. However, NSAIDs have very similar structures, with a carboxylic acid on a similarly sized hydrophobic aromatic moiety (Chart 1). The structural similarity represents a difficult challenge in the selective detection of these drugs by fluorescence [12–14], and traditional macrocyclic supramolecular hosts such as cyclodextrins tend to bind NSAIDs indiscriminately [15]. The difficulty prompted researchers to explore alternative formats of sensing, such as arrays of sensors, which showed good promise [16]. Other choices include enzyme-linked immunosorbent assays (ELISA) [17, 18], but the structural similarity even challenges antibodies which are known for their highly specific binding [17].

Chart 1.

Structures of common NSAIDs.

One way to prepare a receptor for a target analyte is through molecular imprinting [19, 20]. The technique uses the analyte (or a surrogate) as the template and, through templated polymerization, creates analyte-complementary binding sites in a highly cross-linked polymer network. Molecularly imprinted polymers (MIPs) have been referred to as “plastic antibodies” and found numerous applications [21–30]. With appropriate functional monomers (FMs), MIPs can have excellent molecular recognition for molecules of many different sizes, making them extremely useful in biomedical applications [31–33]. In fact, MIPs generated for NSAIDs displayed selective adoption of these drugs [34–37], but their conversion into selective fluorescent sensors is hampered by the insolubility, heterogeneous distribution of binding sites, and other challenges associated with traditional MIPs.

Our group developed a method of molecular imprinting in doubly cross-linked micelles (Scheme 1) [38]. The method involves first surface-crosslinking of micelle of 6 with diazide 7 using the click reaction and then core-cross-linking with divinylbenzene (DVB) using free-radical polymerization photolytically initialed by DMPA (2,2-dimethoxy-2-phenylacetophenone). The cross-linked micelles are also functionalized with a layer of hydrophilic ligand (i.e., 8) for increased hydrophilicity and facile purification. The method can be applied to a wide range of small-molecule drugs [38, 39], peptides [40], and carbohydrates [41, 42]. The resulting molecularly imprinted nanoparticles (MINPs) showed strong abilities to distinguish closely related structures including leucine and isoleucine in peptides [40] and inversion of a single hydroxyl in oligosaccharides [41, 42]. MINPs are ~5 nm in diameter, and mimic proteins in their nanosize, hydrophilic exterior, and hydrophobic core. The number of binding sites per MINP can be conveniently controlled by the surfactant/template ratio. For example, when this ratio was kept the same as the surfactant aggregation number of the micelle (~50), MINP obtained was found to have an average of one binding site per nanoparticle [38].

Scheme 1.

Preparation of MINP by surface–core double cross-linking of template-containing micelle of 6. The template molecule is color-coded with the blue shape representing a hydrophobic moiety and the violet sphere a hydrophilic group to anchor the template near the surface of the micelle.

In this work, we report the design and synthesis of several fluorescent FMs for selective NSAID detection. Structure of fluorophore was found to impact the sensing strongly, with minor variation in substitution pattern totally changing the behavior of the resulting sensor. One FM, containing a thiourea group bonded to a sulfonated 1,5-naphthalene derivative, allowed highly selective binding of the drugs. Strong binding between the thiourea group of the FM and the carboxylic acid of NSAID helped the fluorescent probe reside near the imprinted binding site so that the guest binding was readily detected by fluorescent change.

2. Results and Discussion

2.1. Design and Synthesis of Fluorescent Functional Monomers to Bind Carboxylates

Molecular imprinting can be very effective at creating template-specific binding sites in a polymer network. To create a fluorescent sensor for NSAIDs, however, we need to not only have a strong and selective binding for the drug but also convert the binding into an easy-to-detect fluorescent signal [34–37, 39]. Our strategy is to employ a functional monomer containing an environmentally sensitive fluorophore and a strong carboxylate-binding moiety in close proximity [43]. In addition, highly specific, strong binding must exist between the FM and the NSAID drug so that the fluorophore will stay near the imprinted site.

Compounds 9–12 fit the above criteria [44]. The FMs generally have a push–pull fluorophore known to be highly sensitive to its microenvironment, based on either the amino-nitrobenzoxadiazole (amino-NBD) or aminonaphalenesulfonate (ANS) framework. All compounds have at least one polymerizable vinyl group from a methacrylate, methacrylamide, or styrene. All have a reasonable level of hydrophobicity for their facile incorporation into micelle for the MINP preparation. FM 9 has a guanidinium cation to bind a carboxylate through a hydrogen-bond-reinforced salt bridge. FMs 10–12 employ either a urea or thiourea group to form double hydrogen bonds with a carboxylate. In all cases, the carboxylate-binding moiety is adjacent to the fluorophore, in the hope that the binding would trigger a noticeable change in the latter’s emission. Although hydrogen-bonds are weakened by strong solvent competition in aqueous solution, they are known to become much stronger in the hydrophobic core of micelles [45, 46] and at the surfactant/water interface [47].

To test the suitability of these compounds as fluorescent FMs for NSAID sensing, we first studied their binding of simple carboxylates in a micellar solution of CTAB (cetyltrimethylammonium bromide) (Table 1). The cationic surfactant is commercially available and mimics 6 without any concerns of polymerization. We evaluated the binding with sodium acetate, butyrate, and octanoate, respectively, with the expectation that a more hydrophobic carboxylate would have a stronger driving force to enter the micelle and display an enhanced binding. As mentioned above, a successful fluorescent sensor could only be obtained if the fluorescent FM stayed near the template during polymerization/cross-linking. A strong binding between the FM and the template in the aqueous micellar solution thus is a prerequisite.

Table 1.

Binding constants between carboxylate guests and FM 9–12in CTAB solution.^a

entry	FM	guest	Ka (103 M−1)
1	9	sodium acetate	-b
2	9	sodium butyrate	-b
3	9	sodium octanoate	-b
4	10	sodium acetate	0.016 ± 0.3
5	10	sodium butyrate	0.34 ± 0.05
6	10	sodium octanoate	3.01 ± 0.26
7	11	sodium acetate	0.17 ± 0.06
8	11	sodium butyrate	0.49 ± 0.04
9	11	sodium octanoate	1.51 ± 0.14
10	12	sodium acetate	0.18 ± 0.06
11	12	sodium butyrate	0.78 ± 0.23
12	12	sodium octanoate	5.18 ± 0.12

^aThe binding constants were obtained from fluorescence titrations performed in duplicates in 2 mM CTAB solution. [FM] = 2.0 μM.

^bBinding constant could not be obtained because of weak emission of the fluorophore. See Figures S7–S15 for the titration curves.

In our hands, the emission of 9 in CTAB solution was very weak and showed little response to the addition of the carboxylate salts. Meanwhile, even though the emission of 10 was quenched strongly by the carboxylates and good binding properties were obtained (Table 1, entries 4–6), the solubility of the molecule in micellar solutions, whether of CTAB or 6, was very low. The poor solubility made 10 unsuitable as a FM because our previous studies showed that MINP prepared with 6 as the cross-linkable surfactant contained ~50 cross-linked surfactants. For typical MINPs, we want to reach at least 1:50 for the FM/surfactant ratio, so that the final MINP will have an average of one binding site per nanoparticle [38].

We suspected that the poor solubility of 10 was caused by strong intermolecular interactions among the molecules as a result of the neutral, relatively rigid structure. In addition, a urea group has an excellent hydrogen-bond acceptor (carbonyl) and two good hydrogen-bond donors (NH), making self-association of the FM very strong. In view of these challenges, we designed and synthesized 11 and 12, which had a more flexible structure, an anionic sulfonate group, and a thiourea instead. Without a strong hydrogen-bond acceptor, thioureas tend to self-associate less strongly than ureas.

Both 11 and 12, to our delight, were easily incorporated into micelles and were actually soluble in water themselves. In addition, they both bound the carboxylates, exhibiting larger bonding constants (K_a) with an increase in the hydrocarbon chain length of the guest (Table 1, entries 7–12). The submillimolar binding affinities suggest, under typical MINP preparation conditions (i.e., 10 mM of 6 and 0.2 mM of FM), a substantial amount of the carboxylate guest will be complexed with the FM [48]. Upon further screening, compound 11 was rejected because its resulting MINP displayed very little fluorescent response to the template. Fortunately, compound 12, with a simple change of substitution pattern from 2,6- to 1,5 on the naphthalene ring, showed strong binding for hydrophobic carboxylates in the CTAB micelle (Table 1, entries 10–12) and the resulting MINP also worked well (vide infra).

2.2. Fluorescent MINPs for Indomethacin

Our initial NSAID target was Indomethacin (1). Not only is this drug commonly used to treat osteoarthritis and rheumatoid arthritis, the COX-2 inhibitor has been used to construct fluorescent probes to target the Golgi apparatus of cancer cells [49]. Preparation of MINP(1), i.e., MINP prepared with Indomethacin (1) as the template, follows previously reported procedures and is described in the Experimental Section. The surface- and core-cross-linking of the micelle was monitored by ¹H NMR spectroscopy (Figure S1). The surface-cross-linking chemistry had been confirmed by mass spectrometry previously [50]. The micelles underwent characteristic changes in size during the surface-cross-linking, surface-functionalization, and core-cross-linking, and could be monitored by dynamic light scattering (DLS, Figures S2–3). The DLS size had been confirmed by transmission electron microscopy (TEM) [51, 52].

Figure 1a shows the emission spectra of 2.0 μM MINP(1) in 50 mM Tris buffer (pH = 7.4) upon the addition of 0–10 μM Indomethacin. The MINP was prepared with a 2:1 ratio of FM/template. The emission maximum (λ_em) of the MINP occurred at 414 nm and shifted gradually to the red upon the addition of Indomethacin (1). The number was significantly shifted to the blue than that of the FM in CTAB solution, 430 nm (Figure S15). This is in agreement with the common behavior of ANS-based fluorophores that tend to emit at a shorter wavelength in a less polar environment [53]. A CTAB micelle is highly dynamic. FM 12 in the uncross-linked micelle thus had a much higher chance to be exposed to water than in the doubly cross-linked MINP [54].

Figure 1.

(a) Emission spectra of MINP(1) in 50 mM Tris buffer (pH = 7.4) upon the addition of different concentrations of 1. The MINP was prepared with a 2:1 FM/template ratio. λ_ex = 310 nm. [MINP(1)] = 2.0 μM. The concentration of MINP was calculated based on a M.W. of 49300 g/mol determined by DLS. (b) Nonlinear least squares fitting of the emission intensity of MINP(1) at 414 nm to a 1:1 binding isotherm.

The addition of 0–10 μM Indomethacin caused significant quenching of the MINP’s emission. Meanwhile, a small red shift was observed during the titration. The red shift was expected from the hydrogen bonds formed between the carboxylate of 1 and the thiourea of the FM in the binding site, which increased the electron-donating ability of the thiourea [43]. As shown in Figure 1b, the emission intensity at 414 nm fit nearly perfectly to a 1:1 binding isotherm [55], affording a binding constant of K_a = (1.68 ± 0.20) × 10⁵ M⁻¹. This value was substantially higher than those observed between FM 12 and octanoate in CTAB micelles (5.18 × 10³ M⁻¹). Although the two guests have a different hydrophobe, the nearly 30-times stronger binding in MINP suggests that an imprinted micelle did provide a better binding environment to the guest than the generic nonpolar core of a CTAB micelle.

An important parameter to optimize is the ratio of FM/template in the MINP preparation. Too small an amount of the FM would leave many templates uncomplexed during cross-linking and polymerization. Although imprinted sites would form under such a situation, the binding pocket would not have a nearby fluorescent group to report the binding, neither the thiourea group to hydrogen-bond with the guest—the former is expected to decrease the sensitivity and the latter to decrease the binding affinity. Too large an amount of the FM during the MINP preparation, on the other hand, would incorporate many polymerized fluorophores in the MINPs without a nearby binding site for the guest. Neither situation is desirable.

To identify the optimal FM/T ratio, we prepared MINPs with 1 as the template and 1, 2, and 3 equivalents of FM 12. We then measured the apparent K_a values determined by fluorescence titration for not only the template itself, but also two other NSAIDs, Tolmetin (2) and Ibuprofen (3). We wanted to know the best FM/T ratio for both the binding affinity and selectivity.

As shown by entries 1 and 4 of Table 2, the binding constant of MINP(1) determined by the fluorescence titration for the template more than doubled from 79 × 10³ to 168 × 10³ M⁻¹ when the FM/T ratio increased from 1:1 to 2:1. However, a further increase of the FM in the preparation reversed the trend and afforded a binding constant of 120 × 10³ M⁻¹ (entry 14). Compound 12 served three roles in the MINP preparation and sensing: as a FM to bind the carboxylate by its thiourea during MINP preparation, as a binding group to bind compound 1 when the guest was added during the titration, and as a fluorescent reporter to signal the binding. A higher FM/T ratio can increase the percentage of the template in the T–FM complex, which will help the imprinting and also the re-binding of the template by the MINP.

Table 2.

Binding constants of MINP(1) prepared with different FM/T ratios for NSAIDs and selected small-molecule acids.^a

entry	FM/T ratio	guest	Ka (103 M−1)	CRRb
1	1:1	Indomethacin (1)	79 ± 0.3	1
2	1:1	Tolmetin (2)	14 ± 2	0.18
3	1:1	Ibuprofen (3)	0.7 ± 0.3	0.01
4	2:1	Indomethacin (1)	168 ± 20	1
5	2:1	Tolmetin (2)	22 ± 3	0.13
6	2:1	Ibuprofen (3)	1.1 ± 0.3	0.01
7	2:1	Diclofenac (4)	5.1 ± 0.5	0.03
8	2:1	Ketoprofen (5)	15 ± 0.1	0.12
9	2:1	Mefenamic acid (6)	4.7 ± 0.6	0.03
10	2:1	acetic acid	0.3 ± 0.1	0.002
11	2:1	butyric acid	0.1 ± 0.1	0.0006
12	2:1	benzoic acid	0.3 ± 0.2	0.002
13	2:1	citric acid	0.2 ± 0.1	0.001
14	3:1	Indomethacin (1)	120 ± 27	1
15	3:1	Tolmetin (2)	25 ± 4	0.21
16	3:1	Ibuprofen (3)	1.2 ± 0.1	0.01

^aThe binding constants were obtained from fluorescence titrations performed in duplicates in in 50 mM Tris buffer (pH = 7.4). λ_ex = 310 nm. [MINP(1)] = 2.0 μM.

^bCRR is the cross-reactivity ratio, defined as the binding constant of a guest relative to that of the template for a particular MINP. See Figures S16–30 for the titration curves.

Meanwhile, the binding selectivity displayed a different trend, as measured by the cross-reactivity ratio (CRR), defined as the binding constant of a guest relative to that of the template for a particular MINP. Among the two analogues studied, Tolmetin (2) is more similar to Indomethacin (1) than Ibuprofen (3), in terms of both functionality and size. Indeed, the CRR for Tolmetin was 0.18, 0.13, and 0.21, respectively as the FM/T increased from 1 to 3. The absolute binding constant for Tolmetin was 14, 22, and 25 × 10³ M⁻¹ for these MINPs (entries 2, 5, and 15). The data, thus, suggests that the 2:1 FM/T ratio gave the best selectivity for the template as a result of increasing the binding for the template more than it did for its structural analogue.

For the best MINP(1), we also measured its binding constant for all the other NSAIDs shown in Chart 1, as well as several carboxylic acids (acetic acid, butyric acid, and citric acid). Among the structural analogues, Tolmetin (2) and Ketoprofen (5) showed the highest cross reactivity (CRR = 0.13 and 0.12, respectively). The other three NSAIDs (Ibuprofen, Diclofenac, and Mefenamic acid) showed very little binding (CRR = 0.01–0.03).

Polyclonal antibodies have been generated and used in competitive ELISA assays for Indomethacin [56]. Tolmetin gave about 0.09 corss reactivity in the assay, slightly better than our 0.13. However, Dichlofenac displayed 0.09 cross-reactivity with the natural antibody but only 0.03 in our case. Thus, the selectivity of our synthetic antibody compares very favorably overall with that of natural polyclonal antibodies which require live animal and a much longer time to produce. As cross-linked polymeric nanoparticles, MINPs also have the benefit of tolerating high temperature [38, 57], organic solvent [57], and extreme pH [58].

For MINP binding in water, hydrophobic interactions are known to contribute strongly [38]. The different NSAIDs have different inherent driving force to enter a hydrophobic environment. Among the drugs tested, compounds 1, 2, and 5 seemed to have the highest hydrophobicity, as shown by their strongest binding to the nonimprinted control materials (vide infra, Figures 2–3). Meanwhile, the imprinted hydrophobic binding site can accommodate or reject a drug based on the size and shape of the latter’s hydrophobe. Since we have a strong carboxyl-binding thiourea near the imprinted site, complementarity in hydrogen bonds is also expected to be important. The binding selectivity of the MINP, at the end, is a composite term reflecting all the above. Importantly, none of the other potentially interfering carboxylic acids showed any significant binding, including acetic acid, butyric acid, benzoic acid and citric acid (Table 2, entries 10–13).

Figure 2.

Change of fluorescence emission intensity of MINP(1) caused by different carboxylic acids in 50 mM Tris buffer (pH = 7.4). [MINP(1)] = 2.0 μM. [Acid] = 10 μM. λ_ex = 310 nm.

Figure 3.

Change of fluorescence emission intensity of MINP(2) caused by different carboxylic acids in 50 mM Tris buffer (pH = 7.4). [MINP(1)] = 2.0 μM. [Acid] = 10 μM. λ_ex = 310 nm.

To demonstrate the generality of our method, we also prepared MINP with Tolmetin (2) as the template (Figures S4–6). Table 3 summarizes its binding constants obtained for various guests. Consistent with successful imprinting, the MINP bound its template most strongly among the analogues. The binding constant (84 × 10³ M⁻¹) was about half of that observed for 1 by MINP(1). This trend was reasonable, given the larger size of Indomethacin. It is well known that the strength of hydrophobic interactions is directly proportional to the area of hydrophobic surface being buried upon binding [59]. As long as the imprinted sites match their templates well, a larger template will have a stronger driving force to enter its imprinted site than a small template (to its own).

Table 3.

Binding constants of MINP(2) prepared with different FM/T ratios for NSAIDs and selected small-molecule acids.^a

Entry	guest	Ka (103 M−1)	CRRb
1	Indomethacin (1)	8.9 ± 1.8	0.11
2	Tolmetin (2)	84 ± 17	1
3	Ibuprofen (3)	1.3 ± 0.3	0.02
4	Diclofenac (4)	5.4 ± 1.1	0.06
5	Ketoprofen (5)	19 ± 3	0.23
6	Mefenamic acid (6)	4.3 ± 0.5	0.05
7	acetic acid	0.3 ± 0.1	0.004
8	butyric acid	0.7 ± 0.2	0.008
9	benzoic acid	1.3 ± 0.4	0.02
10	citric acid	0.6 ± 0.1	0.007

^aThe binding constants were obtained from fluorescence titrations performed in duplicates in 50 mM Tris buffer (pH = 7.4). λ_ex = 310 nm. [MINP(2)] = 2.0 μM.

^bCRR is the cross-reactivity ratio, defined as the binding constant of a guest relative to that of the template for a particular MINP. See Figures S31–40 for the titration curves.

Table 2 shows that NSAID 2 and 5 showed the strongest cross-reactivities in the binding of MINP(1), suggesting these two drugs were more similar to each other than other NSAIDs examined in our work. Table 3 shows that, when 2 was used as the template, the drug that showed the strongest cross-reactivity was 5, with a CRR of 0.23. Thus, both binding studies were able to classify NSAIDs according to their structural similarity.

2.3. Fluorescent Sensing of NSAIDs

Figure 2 shows the change of fluorescence intensity of MINP(1) by 10 μM of different acids in 50 mM Tris buffer (blue columns). The red columns are the responses of the nonimprinted nanoparticle (NINP), prepared with FM 12 but without any template. The imprinting effect was very strong in our MINP, as shown by the large difference in the responses of the imprinted versus nonimprinted sensor. The other acids all had a smaller effect on the emission intensity of MINP(1) and the responses seemed to largely follow the binding affinities. Table 2, for example, shows that Tolmetin (2) and Ketoprofen (5) gave the highest CRR for MINP(1) among the non-templating NSAIDs. In Figure 2, these two drugs also showed the largest responses, after the template itself. Fluorescent sensing was also performed with MINP(2). As shown by Figure 3, excellent selectivity was observed again for the templating drug, Tolmetin.

Compounds 1 (Indomethecin), 2 (Tolmetin), and 5 (Ketoprofen) interacted with the NINP more strongly than other molecules tested (Figures 2 and 3, red columns), most likely as a result of their stronger hydrophobicity. It is worth noting that when Tolmetin was used as the template, Ketoprofen still displayed a sizable MINP/NINP difference. This is reasonable given their similar size/shape (Chart 1). Indomethecin, on the other hand, is clearly larger than the template. It is interesting that there was almost no difference in the MINP vs NINP binding in this case. The result suggests that the interference in the latter case should be derived from nonspecific binding, different from that of Ketoprofen, which came from its similarity to the template in size and shape.

The strong binding of MINP(1) and MINP(2) for their targeted drugs suggests the detection for Indomethacin and Tolmetin should be quite sensitive. To calculate the detection limit, we measured the emission intensity of 2.0 μM MINP(1) in the presence of 1–5 μM Indomethacin in 50 mM Tris buffer. The detection limit, calculated from 3δ/slope, was about 140 nM (Figure S41). The concentration translates to 50 ng/mL, a little over three times of what was reported for the antibody-based ELISA assay (15 ng/mL) [56]. For Tolmetin, our detection limit was 200 nM or 51 ng/mL by MINP(2) (Figure S42).

3. Conclusions

The high structural similarity among NSAIDs makes it difficult even for natural antibodies to distinguish the drugs. The molecularly imprinted cross-linked micelles, nonetheless, displayed excellent abilities to bind and distinguish these drugs. The choice of fluorescent functional monomer is highly important, with compound 12 being the only one useful among the four synthesized. Once the correct FM and all the other ingredients of MINPs are available, MINP-based fluorescent sensors can be prepared and purified in less than 2 days for different drugs. The binding studies showed that these sensors could bind the desired drug selectively among analogues and also classified the analogues based on their structural similarity to the original template. Fluorescent sensing could easily detect 100–200 nM or ~50 ng/mL of the drugs in water.

4. Experimental Section

4.1. Materials

A typical procedure for the preparation of MINPs and NINPs is as follows [38]. A solution of compound 1 in methanol (10 μL of 18.5 mg/mL, 0.0004 mmol) was mixed with a solution of compound 12 in methanol (20 μL of 16.7 mg/mL, 0.0008 mmol) in a vial containing methanol (1 mL). After the mixture was stirred for 1 h at room temperature, methanol was removed in vacuo. A micellar solution of compound 6 (9.3 mg, 0.02 mmol) in H₂O (2.0 mL), divinyl benzene (DVB, 2.8 μL, 0.02 mmol), and 2,2-dimethoxy-2-phenylacetophenone (DMPA, 10 μL of a 12.8 mg/mL solution in DMSO, 0.0005 mmol) were added to the FM–T complex. The mixture was subjected to ultrasonication for 10 min before compound 7 (4.13 mg, 0.024 mmol), CuCl₂ (10 μL of a 6.7 mg/mL solution in H₂O, 0.0005 mmol), and sodium ascorbate (10 μL of a 99 mg/mL solution in H₂O, 0.005 mmol) were added. After the reaction mixture was stirred slowly at room temperature for 12 h, compound 8 (10.6 mg, 0.04 mmol), CuCl₂ (10 μL of a 6.7 mg/mL solution in H₂O, 0.0005 mmol l), and sodium ascorbate (10 μL of a 99 mg/mL solution in H₂O, 0.005 mmol) were added. After being stirred for another 6 h at room temperature, the reaction mixture was transferred to a glass vial, purged with nitrogen for 15 min, sealed with a rubber stopper, and irradiated in a Rayonet reactor for 12 h. The reaction mixture was poured into acetone (8 mL). The precipitate was collected by centrifugation and washed with a mixture of acetone/water (5 mL/1 mL) three times, methanol/acetic acid (5 mL/0.1 mL) three times until the emission peak at 448 nm (for the dansyl) disappeared, and then with excess methanol. The off-white powder was dried in air to afford the final MINPs (16 mg, 80%). NINPs were prepared following similar procedures except no template was used.

4.2. Physico-chemical characterization

Routine ¹H and ¹³C NMR spectra were recorded on a 400 and 600 MHz NMR spectrometer. ESI-MS mass was recorded on Shimadzu LCMS-2010 mass spectrometer. Dynamic light scattering (DLS) data were recorded at 25 °C using PDDLS/CoolBatch 90T with PD2000DLS instrument. Isothermal titration calorimetry (ITC) was performed using a MicroCal VP-ITC Microcalorimeter with Origin 7 software and VPViewer2000 (GE Healthcare, Northampton, MA). Syntheses of compounds 6–8 were reported previously [38]. Syntheses of the fluorescent functional monomers (9–12) are reported in the Supplementary Information (SI).

For the fluorescence titration, a stock solution of the acid (200 μM) was prepared in 50 mM Tris buffer (pH = 7.4). Aliquots (2 μL) of the acid stock solution were added to 2.00 mL of MINP(1) solution (2.0 μM) in the same buffer. After each addition, the sample was allowed to sit for 1 min at room temperature before the fluorescence spectrum was collected. The excitation wavelength (λ_ex) was 310 nm. The excitation slit width was 10 nm, and the emission slit width was 10 nm. The binding constant was obtained by nonlinear least squares fitting of the emission intensity at 414 nm to 1:1 binding isotherm.

Highlights:

• Molecular imprinting of nonsteroidal anti-inflammatory drugs (NSAIDs) in cross-linked micelles

• Fluorescent sensing of the bound NSAIDs

• Binding selectivities comparable to those of polyclonal antibodies

• Detection limit of ~50 ng/mL in aqueous solution

• Minimal interference from common carboxylic acids