All chemicals and solvents of HPLC (high performance liquid chromatography) grade were obtained from commercial suppliers and used as received. High purity water used both in reactions and chromatographic analysis was obtained using a Merck Millipore Synergy UV water purification system. Colorless Corning CoStar 0.65 ml centrifuge tubes were used for enzymatic reactions and sample preparation (dilution and centrifugation), while colorless 2 ml glass vials with PTFE-lined screw-cap septa and 0.2 ml glass inserts were used for short-term sample storage and injection on HPLC equipment. A stock solution of the enzyme CGTase derived from Bacillus macerans was received as a kind gift from Amano Enzyme, Inc., Nagoya, Japan. The stock solution was stored at 5°C and used as received.

Chromatographic analysis was performed on a Thermo Scientific Dionex Ultimate 3000 HPLC (ultra-high pressure) system equipped with a Waters Acquity UPLC BEH Amide 1.7 µm 2.1 × 150 mm column maintained at 30°C, and an autosampler module maintained at 20°C. Detection was carried out using an Agilent Technologies 1,260 Infinity ELSD (evaporative light scattering detector), operating with the evaporator at 90°C, nebulizer at 70°C, and a N₂ gas flow of 1.0 L/min. The ELSD enables the detection of the chromophore-lacking oligosaccharides. Calibration curves from 0.01 mg/ml to 10 mg/ml for α-, β- and γ-CD and linear α-1,4-glucans up to maltooctaose were used to correct for differences in the ELS detector response for different oligosaccharides. The calibrations were based on masses injected (0.018–3.66 µg) and the resulting response curves were fitted to a simple power equation M = kA ^p (where M is the injected mass of compound, A is the area under the peak in the chromatogram and k and p are fitted parameters) using non-linear curve fitting (in OriginPro 2018b from OriginLab Corp.) See recent paper for details (Larsen and Beeren, 2021). The gradient profile for HPLC runs was a linear gradient from 75% acetonitrile in water to 55% acetonitrile in water over 8 min with a flow rate of 0.6 ml/min. Both eluents contained 0.1% formic acid by volume.

Reaction mixtures with the desired concentrations of salts were prepared by mixing appropriate amounts of two types of stock solutions in buffered water (50 mM sodium phosphate at pH 7.5): 1) a solution containing α-CD (10 mg/ml) and various salts (4.0 M); and 2) a solution containing α-CD (10 mg/ml). In templated experiments, cyclohexanol or cyclohexanecarboxylic acid was dissolved in these resulting mixtures at a concentration of 10 mM. For the experiments with a higher concentration of sodium phosphate buffer, a stock solution of α-CD (10 mg/ml) in 0.45 M sodium phosphate buffer at pH 7.5 was prepared. The starting mixtures containing salt, α-CD and template (if any) in buffer were then aliquoted (165–365 µl) into reaction vessels and kept at ambient temperature. All reactions were then initiated by adding CGTase stock solution (50 µl per ml of starting mixture) to the starting mixtures followed by thorough mixing. The reactions were then monitored at various time points: Aliquots for analysis (4–5 µl) were taken out and rapidly diluted (31 fold) in a 1% trifluoroacetic acid (TFA) solution in 3:1 acetonitrile/water with 10 mM ammonium chloride to stop the enzymatic reaction. For experiments with NaCl, aliquots for analysis (20 µl) were taken out and rapidly diluted (six fold) in a 1% TFA solution in water. Samples were then centrifuged (10,000 RPM for 4 min) to prevent column blockage by insoluble species such as enzyme and salts, and the top fractions (leaving behind 20 µl) were then transferred to 2 ml glass vials with 0.2 ml glass inserts, kept at 20°C and then injected on the HPLC instrument within 48 h. Injection volumes were 10 µl or 2 µl (for 31-fold and 6-fold diluted samples, respectively). Peaks in the chromatograms corresponding to α-CD, β-CD and γ-CD and linear α-1,4-glucans up to maltooctaose were identified by comparison with authentic samples obtained from commercial suppliers.

Simulations of dynamic combinatorial libraries (DCLs) with and without the anion SCN⁻ were carried out using the DCLSim software developed in the Otto group (Corbett et al., 2004) and kindly made available to us. The program requires input of the concentration of the building block (glucopyranose units in this case), the composition of the oligomers (library members) formed in the DCL (α-CD, β-CD and γ-CD, with six, seven or eight glucose units in this case), the relative formation constants K _f of the library members (determined from the pseudo-equilibrium composition of α-, β-, and γ-CD in an untemplated library), the binding constants (K _a) of each library member to the template (SCN⁻) and the concentration of the template. Details about how the relative formation constants K _f of α-, β-, and γ-CD were calculated can be found in the Supplementary Material. The binding constants used were obtained from a study carried out by Tokunaga and coworkers, where the authors used ¹H-NMR spectroscopy to determine binding constants between inorganic anions and cyclodextrins (Matsui et al., 1997).

To explore the influence of anions on CGTase-mediated DCLs of cyclodextrins, we examined a series of DCLs prepared in the presence of different sodium salts at concentrations up to 4 M. The following series of anions was investigated, ranked according to the Hofmeister series: HPO₄ ²⁻/H₂PO₄ ⁻ > Cl⁻ > NO₃ ⁻ > Br⁻ > ClO₄ ⁻ > SCN⁻. The libraries were prepared by addition of CGTase (50 μl stock solution per ml reaction mixture) to solutions of α-CD (10 mg/ml) with the desired sodium salts at various concentrations up to 4 M concentration in phosphate buffer (50 mM, pH 7.5). Despite the lower solubility of sodium phosphates, we chose also to include phosphate buffer in this study albeit at a maximum of 0.45 M. The library compositions were monitored as the dynamic system evolved over time using hydrophilic interaction liquid chromatography (HILIC) with an evaporative light scattering detector (ELSD), which enabled the separation and quantification of the chromophore-lacking glucan mixtures. The influence of varying concentrations of different salts on the equilibrium cyclodextrin distribution and the time taken to reach this steady distribution is summarized for all anions tested in Table 1.

Figure 2 shows representative data obtained for DCLs of cyclodextrins prepared in the presence of increasing amounts of NaBr. Figure 2A depicts chromatograms showing the distributions of α-1,4-glucan products formed after pseudo-equilibrium is obtained (2–6 h) in the absence and presence of different concentrations of NaBr. It is immediately evident that the relative concentration of α-CD increased at the expense of β-CD and γ-CD in the presence of increasing concentrations of NaBr. Amplifications of α-CD were seen for all the anions tested, but the magnitude of the effect was anion-dependent (Table 1). The evolution of each DCL was monitored over time (Supplementary Figures 1–6). For all DCLs, the CD yield decreased gradually overtime, due to background hydrolysis and the consequent build-up of short linear α-1,4-glucan and glucose. For the DCL without salt, a steady distribution of α-CD, β-CD and γ-CD was obtained after approximately one hour, when a pseudo-thermodynamic equilibrium of cyclodextrins was reached. At this point >90% of the glucan material was still present as cyclodextrins. With increasing concentrations of NaBr, the evolution of the DCL became slower, and it took up to approximately four hours (with 4 M NaBr) to reach a steady distribution of cyclodextrins (Figure 2C). In all cases, addition of anions led to a slower evolution of the dynamic enzymatic system (Supplementary Figures 1–6). With the chaotropic anions ClO₄ ⁻ and SCN⁻ (Table 1, entries 15–20) this retardation was quite significant at 1–2 M concentrations, and at concentrations of 3 M and higher, the activity of the enzyme was zero or close to zero within just 30 min of being exposed to the salt solutions (Supplementary Figure 8), presumably due to denaturation of the enzyme under these conditions.

The equilibrium cyclodextrin distributions obtained in the presence of different anions at 2 M concentration are summarized in Figure 3. Following a largely systematic trend, the relative yield of α-CD increases upon moving from the most kosmotropic anions (HPO₄ ²⁻/H₂PO₄ ⁻,Cl⁻, minor changes) to the most chaotropic anions (ClO₄ ⁻, SCN⁻, large changes). These results suggest that all the tested anions have a specific, albeit weak affinity for α-CD and function as templates in the enzyme-mediated DCL of cyclodextrins. Selected binding constants for Cl⁻, NO₃ ⁻, Br⁻, SCN⁻ _, and ClO₄ ⁻ interacting with α-CD, β-CD and γ-CD have previously been reported (Table 2). These binding constants were determined using a variety of techniques: ¹H-NMR spectroscopy (Matsui et al., 1997), conductance (Wojcik and Rohrbach, 1975), potentiometry (Gelb et al., 1983), spectrophotometry (Buvári and Barcza, 1979), volatilization (Sanemasa et al., 1988) and isothermal titration calorimetry (Sullivan et al., 2018). While the numeric values of the binding constants vary somewhat depending on the method, the trends are certainly clear. In the kosmotropic end of the Hofmeister series, the binding of Cl⁻ to α-CD was found to be negligible in most cases, which corresponds well with our data, where only minor changes in the cyclodextrin distribution occur upon addition of Cl⁻ (Table 1, entries 3–6). The minor increase in the relative yield of α-CD, from 32 to 37%, does, however, indicate that there could be a very weak binding between α-CD and Cl⁻, as found in some cases in the literature (Wojcik and Rohrbach, 1975). Next in the series, both NO₃ ⁻ and Br⁻ bind to α-CD with small and similar binding constants (1–4 M⁻¹), which matches the small but significant amplification of α-CD observed with these anions (Table 1, entries 7–14). The chaotropic anions SCN⁻ and ClO₄ ⁻ have significantly higher affinities for α-CD (16–46 M⁻¹) and led to much larger changes in the cyclodextrin distribution (Table 1, entries 15–20). The fact that the amplification of α-CD is larger with SCN⁻ than ClO₄ ⁻ can be explained by the relatively higher competing affinity of ClO₄ ⁻ for β-CD, as the distribution obtained in a DCL is influenced by the binding interaction of the template with all members of a library. It is noteworthy that each of the small anions tested amplified and bound most strongly to α-CD, whereas we have previously observed the amplification of larger CDs with 8, 9, and 10 glucopyranose units in the presence of the large superchaotropic anion B₁₂I₁₂ ²⁻ (diameter 11.7 Å) (Larsen and Beeren, 2019). There is clearly, thus, a relationship between the size of the anion (Table 2 column 3) and the preferential formation of the CD(s) with a suitable size cavity. Overall, we found that the addition of anions in high concentrations to the CGTase-mediated DCLs of cyclodextrins leads to changes characteristic of template effects, and remarkably, the system remains dynamic even at 2 M concentrations of the denaturing salts NaSCN and NaClO₄.

To gain further support for our conclusion that chaotropic anions influence the production of specific cyclodextrins in CGTase-mediated dynamic systems through direct template effects, we sought to simulate the DCL generated in the presence of increasing concentrations of NaSCN. DCLSim is a software developed in the Otto group (Corbett et al., 2004) that enables the prediction of product distributions in templated DCLs operating under thermodynamic control. To simulate the DCLs, binding constants for the interaction of the template with each library member (α-CD, β-CD and γ-CD) is required, and this was available for SCN⁻ (Table 2, column I, Matsui et al., 1997). The relative formation constants K _f for α-CD, β-CD and γ-CD are also needed for the simulation and these could be calculated from the relative concentrations of α-CD, β-CD and γ-CD generated at equilibrium in the DCL without salt. We simulated DCLs of cyclodextrins with SCN⁻ at 0, 1 and 2 M concentrations. The results, summarized in Figure 4, show that the simulations correlate well with the experimental results, supporting the conclusion that chaotropic anions function as templates in this system, and at high concentrations can strongly influence the product selectivity in CGTase-mediated cyclodextrin synthesis.

As kosmotropic salts can lead to stronger binding between cyclodextrins and hydrophobic guests (Buvári and Barcza, 1979; Holm et al., 2014), we sought to investigate whether we could enhance the template effect of hydrophobic templates in this dynamic cyclodextrin system by using high concentrations of NaCl. For these experiments, we chose to employ cyclohexanol (template 1) and cyclohexanecarboxylate (template 2) as templates, as both guests have a relatively low affinity for β-CD in the absence of salts (ca. 700 and ca. 300 M⁻¹, respectively) (Gelb and Schwartz, 1989; Rekharsky et al., 1994), thus giving room for a possible improvement in templating effect upon addition of NaCl. A series of CGTase-mediated DCLs were set-up starting from α-CD (10 mg/ml) with template (10 mM) in phosphate buffer (50 mM, pH 7.5) with 0–3 M NaCl (Supplementary Figure 7). It was found that the addition of NaCl in increasing concentrations up to 3 M did in fact lead to a moderate increase in the selectivity for β-CD obtained in the presence of each template (from 80 to 87% for template 1 and from 75 to 82% for template 2) (Figure 5). It is worth noting that increasing the concentration of NaCl from 0 to 3 M in the absence of template also changes the cyclodextrin distribution, but in the opposite direction, with a decrease in the relative yield of β-CD from 57 to 53%, thus counteracting the observed increase in the template effect with 1 and 2. Accordingly, the actual kosmotrope-induced increase in the template effect is potentially larger than the effect observed here.