A supramolecular cucurbit[8]uril-based rotaxane chemosensor for the optical tryptophan detection in human serum and urine

Sensing small biomolecules in biofluids remains challenging for many optical chemosensors based on supramolecular host-guest interactions due to adverse interplays with salts, proteins, and other biofluid components. Instead of following the established strategy of developing alternative synthetic binders with improved affinities and selectivity, we report a molecular engineering approach that addresses this biofluid challenge. Here we introduce a cucurbit[8]uril-based rotaxane chemosensor feasible for sensing the health-relevant biomarker tryptophan at physiologically relevant concentrations, even in protein- and lipid-containing human blood serum and urine. Moreover, this chemosensor enables emission-based high-throughput screening in a microwell plate format and can be used for label-free enzymatic reaction monitoring and chirality sensing. Printed sensor chips with surface-immobilized rotaxane-microarrays are used for fluorescence microscopy imaging of tryptophan. Our system overcomes the limitations of current supramolecular host-guest chemosensors and will foster future applications of supramolecular sensors for molecular diagnostics.

phenanthroline-1,3,6,8(2H,7H)-tetraone (2) (17.0 mmol, 1.00 eq), the solution turned red and was heated to reflux. The reaction mixture changed its color to green after 4 h and was cooled to room temperature. The quenching with 400 mL of ice water caused precipitation of a brown-grey solid, which was filtered off and dried in vacuo. The solid was transferred into a soxhlet extractor and extracted with 1.50 L chloroform for one day. The extract was evaporated and recrystallized from 100 mL pyridine, yielding 3.10 g of a yellow solid (12.9 mmol, 76.0%). were stirred at 265 °C for 4 h. The black viscous mixture was heated to 300 °C for 1 h and then cooled to room temperature. Next, 100 mL of 1 M aqueous HCl was added and stirred under reflux for 30 minutes. This procedure was repeated three times in total, and after each boiling, the black solid was filtered off. The filtrates were combined, and aqueous NaOH (5 M) was added until pH 12.0 was reached. The formed yellow  [3,8]phenanthroline-2,7-diium dibromide (5) (178 µmol, 1.00 eq) was dissolved in 10.0 mL ultrapure water and 60.9 mg of potassium thioacetate (533 µmol, 3.00 eq) dissolved in 5.00 mL ultrapure water was added. The reaction solution was stirred for 2 days at room temperature under the exclusion of light. The solvent was removed under reduced pressure, and the product was obtained as a brown solid with a yield of ≥ 99% due to residual potassium thioacetate and used as such in the following synthetic      Table 3. . Supplementary Fig. 4. Test of the analyte-sensing functionality of rotaxane 1 as a chemosensor by an emission-based analyte assay with indole. The response of rotaxane 1 towards indole remained similarly strong in 1X PBS compared to water. In contrast, due to competing salt cations, the bimolecular chemosensor CB8•MDAP completely lost its functionality in 1X PBS. a Schematic representation of analyte binding of rotaxane 1 towards the electron-rich aromatic analyte indole and for the bulky (not fitting next to the dye due to its size) analyte 1-AdOH. b Emission quenching of rotaxane 1 (black line) was observed after the addition of indole (dotted red line) due to the dye-analyte interaction. The subsequent addition of 1-AdOH (dotted blue line) did not cause any significant signal change, since 1-AdOH is too bulky to bind next to the DAP dye and cannot displace the dye due to the interlocked system with the installed stoppers as anchor groups. c Schematic representation of analyte binding to a self-assembled CB8•MDAP chemosensor. Bulky and strongly binding analytes such as 1-AdOH displace the bound dye and disassemble the chemosensor. d Emission response of the CB8•MDAP chemosensor (black line) in the presence of indole (dotted red line) and 1-AdOH (dotted blue line). The higher emission signal compared to the starting signal after 1-AdOH addition indicated the disassembly of the CB8•MDAP complex (dye emission is slightly quenched in the host cavity). All emission spectra were recorded in 1X PBS (black) at 25 °C (λex = 393 nm, λem = 450 nm).

Binding affinity determination in a direct binding assay (DBA)
with Ka = binding constant,   Supplementary Fig. 10. Bar graphs of a the emission intensity of the steroid-depleted human serum sample prior (yellow) and after the addition of 10 µM of rotaxane 1 (light green), after stepwise spiking with Trp (green to blue), and after indole addition (dark violet). b Bar graphs of emission intensities of rotaxane 1 upon the Trp addition corrected for the auto emission of the serum sample and c emission quenching of rotaxane 1 in serum. The excitation wavelength λex = 393 nm and the emission wavelength λem = 450 nm were used. All measured data points are shown; bar graph height indicates the mean (middle line), and error bars represent the standard deviation, with n = 5 independent replicates (coefficient 1.5).

HPLC-based quantification of L-Trp in urine samples
Supplementary Fig. 11. Plot of the linear calibration curve (black dots and red line) and the intensity height of the fluorescence signal (λex = 295 nm, λem = 340 nm) of the unknown urine samples, measured in n = 2 independent replicates for urine sample 1 and n = 3 independent replicates for urine sample 2 and for the Trp standard solutions. BSA cannot bind to rotaxane 1 due to its compact structure in which hydrophobic amino acids such as Phe and Trp are buried inside the protein core. However, the endopeptidase pepsin digests the protein into smaller peptides carrying exposed amino acids on their surface, which can be bound and thus detected by the rotaxane chemosensor. The expected increase in the hydrolysis rate of BSA with increasing pepsin concentration was observed.     Table 1 in the main text.

Selection of commonly used non-covalent supramolecular chemosensors in several aqueous media and biomedia
The following section briefly summarizes the stability and sensing abilities of a selection of commonly used non-covalent supramolecular chemosensors. In these experiments, we wanted to compare their functionality for sensing in aqueous media, saline buffer, and biomedia such as human blood serum. The chemical structures of the selected chemosensor systems are shown in Supplementary Fig. 20. Each system was examined in a 96-well plate format by measuring its emission intensity at 25 °C. The concentration of each of the host components was adjusted to 10 µM to use comparable concentrations to rotaxane 1 in human blood serum (see Supplementary Fig. 23).
Supplementary Fig. 20. Chemical structure of host (top) and dye (bottom) molecules of the selected chemosensor systems.
All tested dyes show the commonly found emission turn-on when complexed by one of the macrocycles. The supramolecular complexes CB7•berberine chloride and sCx4•lucigenin both show a strong emission signal in water (see black spectra in Supplementary Fig. 21a and b), verifying that they are stable systems in water.
However, their transfer to a saline buffer system, e.g., 1X PBS, results in a strongly decreased emission intensity. Their emission intensity is similarly quenched in human blood serum and urine. These results indicate that both supramolecular complexes dissociate in the presence of salts due to the competitive binding of cations (CBn portals and calix cavity). The quenching of lucigenin by chloride may also play a role. 6,7 Be this as it may, both supramolecular systems cannot be used as chemosensors in saline media or even biofluids.
A different problematic behavior was observed for the mixture of β-cyclodextrin as host and 2-p-toluidino-6naphthalenesulfonic acid (2,6-TNS) as dye. While the addition of 2,6-TNS to β-CD and its derivatives usually causes an emission increase in water, 8 we observed in our experiments in the micromolar concentration range no emission enhancement in water, buffers, or urine due to the low binding affinity (e.g., log Ka ~ 3). 9 Only in human blood serum an emission enhancement was observed, which can be attributed to dye binding to the protein human serum albumin. 10,11 These results show that both systems are unsuitable for sensing in aqueous media or biomedia. In addition, the interference of the dye with proteins in human serum further limits its application in biofluids. Another CBn-based supramolecular complex, CB7•MDAP, is less affected by salt cations since CB7 has a larger binding affinity towards the dye 2,7-dimethyldiazapyridinium (MDAP, log Ka = 9.4) 9 compared to berberine chloride (log Ka = 7.2). 9 This system is therefore to a certain extent stable in saline buffers and even human blood serum (see Supplementary Fig. 22), although it displays a reduced emission intensity compared to water as a solvent. Nevertheless, the system is not suitable for sensing low micromolar concentrations of tryptophan ( log Ka ~ 3 for CB7 in water), especially not in biomedia due to the presence of many competitive binders such as biogenic amines, steroids, and proteins, and due to the high binding affinity of the indicator dye for the host. For example, Supplementary Fig. 22 shows that the addition of 66-times excess of tryptophan to CB7•MDAP in human blood serum does not significantly alter the emission intensity. Unlike CB7, CB8 can bind two analytes at the same time as a result of its bigger cavity size. 12 This unique property allows the formation of 1:1:1 complexes consisting of the macrocycle CB8, a dye, and an analyte. In this case, the interaction of the dye with the analyte inside the cavity of the CB8 leads to a quenching of the dye's emission. We compared rotaxane 1 with the non-covalent chemosensing system CB8•MDAP as both contain the same dye moiety. In short, the emission intensity of CB8•MDAP and rotaxane 1 were measured upon the stepwise addition of tryptophan (c = 0 -425 µM) in water, 1X PBS, and human serum at λem = 450 nm (λex = 393 nm) (see Supplementary Fig. 23). Both systems show a significant emission response upon the addition of tryptophan in water (black dots) and 1X PBS (blue triangles).
In human blood serum, the initial emission intensity of CB8•MDAP is more than 20% reduced compared to rotaxane 1. Moreover, the observed emission quenching upon the addition of an excess Trp (> 400 µM) is significantly larger for rotaxane 1 than for CB8•MDAP, see Supplementary Fig. 23. Therefore, rotaxane 1 offers a larger useable detection window, i.e., the partial emission quenching per tryptophan unit, compared to CB8•MDAP, which makes the rotaxane more sensitive for smaller amounts of tryptophan in a more extended concentration range. In addition, the salt influence ("matrix effect") on the tryptophan detection of both chemosensor systems was evaluated by measuring the emission intensity quenching of a 25 µM tryptophan solution in the presence of different amounts of sodium chloride. The results are shown in Supplementary Fig.   23. The colored boxes illustrate the disturbing influence of an unknown salt concentration on the sensing abilities of the corresponding chemosensors. The results demonstrate that rotaxane 1 is less affected by salts than CB8•MDAP. In addition, the integrity of CB8•MDAP, but not of rotaxane 1, is destroyed in the presence of hydrophobic interferents (see also Supplementary Fig. 4).

Structured immobilization by microcontact printing of rotaxane 1
The immobilization of rotaxane 1 was conducted using microchannel cantilever spotting (μCS). For a detailed description of the experimental setup, see the methods section in the main part. As control experiments, a CB8•MDAP chemosensor microarray was set up according to previous work. 13 CB8 ink containing 1 mg/mL of propargyl-functionalized CB8 (in DMSO containing 40% TCEP dissolved in water (3 mg/mL) and 20% glycerol) was micropatterned on thiol surfaces (SiO2-SH) via μCS to obtain the CB8 microarray. After printing, the substrate was irradiated with UV light (254 nm) for 10 min, washed with water and ethanol, and dried by a N2 stream. The remaining free thiols of the surface were blocked through incubation with an aqueous N-ethylmaleimide (10mg/mL, blocking agent) solution at pH 7.0 for 2 h. Then, the substrate was covered with 20 μL of a 50 μM MDAP solution for 1 min, washed with water and dried with an N2 stream to obtain the CB8•MDAP chemosensor array.

Initial test of analyte detection
The response of the CB8•MDAP microarray towards the presence of an analyte is shown in comparison to the response of rotaxane 1 microarrays with different analytes in HEPES buffer (see Supplementary Fig. 14). The incubation of the rotaxane 1 array with pure HEPES buffer as a negative control shows no significant decrease in fluorescence. As expected, the emission intensity of rotaxane 1 decreased for the incubation with indole showing that indole binds to the immobilized rotaxane 1 and thereby quenches the emission of the DAP dye.
In contrast, the incubation of the rotaxane 1 microarray with memantine, an analyte with a larger binding affinity towards CB8 than DAP, shows only an insignificant emission change, indicating that the analyte is not bound. In contrast, the CB8•MDAP microarray shows an almost 100% emission quenching upon incubation with memantine (Mem), indicating that the MDAP dye is displaced by the bigger and stronger binding analyte memantine. 13

Sensitivity tests
The sensitivity of the rotaxane 1 microarrays for analyte detection was first examined by incubation with L-tryptophan solutions at different concentrations in 1X PBS. The emission intensities of the spots of the microarrays before and after the incubation with the analyte and the corresponding fluorescence images are shown in Supplementary Fig. 15. It is possible to detect tryptophan concentrations down to 10 −8 M, which is indicated by the greater emission decrease of the rotaxane 1 microarray after the incubation with 10 −8 M compared to the control with 1X PBS. However, a concentration of 10 −9 M of L-tryptophan is no longer discernible from pure 1X PBS.

Measurements in serum
Besides that, the Trp detection in spiked serum was established. A human blood serum sample with a Trp concentration lower than 1 µM (quantified via HPLC) was spiked with 1, 5, 10 and 25 µM of Trp. The detection was done by incubation of the rotaxane 1 microarrays with the Trp-spiked serum solutions, and the obtained emission intensities are shown in Supplementary Fig. 16. The plotted emission intensities in Supplementary Fig. 16a show a quantitative emission switch-off with increasing Trp concentration and thereby confirm the applicability of the rotaxane 1 microarray.
The applicability of the rotaxane 1 microarrays was then examined by incubation with untreated human blood serum in its untreated form and was diluted with 1X PBS. The untreated, pure serum sample contained around 58 µM Trp, quantified by analytical HPLC (see section 3). The emission intensities of the spots of the microarrays before and after the incubation with the corresponding analyte and the corresponding fluorescence images are shown in Supplementary Fig. 17. The bigger emission decrease of the rotaxane 1 microarray after the incubation with serum and diluted serum compared to the control with 1X PBS shows that it is possible even to detect small amounts of Trp containing serum in 1X PBS. Even undiluted serum (with the highest concentration of potential interfering small molecules/proteins) is not hindering Trp detection.