Using 19F NMR and two‐level factorial design to explore thiol‐fluoride substitution in hexafluorobenzene and its application in peptide stapling and cyclisation

Hexafluorobenzene undergoes 1,4‐selective thiol‐fluoride disubstitution and is an attractive disulfide crosslinking reagent for peptide cyclisation and stapling. Little attention has been directed toward understanding the scope of this reaction. Traditional reaction optimisation relies on a one‐variable‐at‐a‐time approach, which can exclude important combined effects of reaction variables. This study initially explored base and solvent effects to inform a subsequent two‐level factorial design approach to understand how to control the reactivity and product selectivity in a model reaction of hexafluorobenzene. We describe new conditions that selectively afford higher order substitution products for example, 1,2,4,5‐tetrasubstitution, making hexafluorobenzene a possible suitable scaffold for future branched or multicyclic peptide systems. Moreover, our new conditions provide an improved rapid (<1 minute) and selective peptide disulfide stapling and cyclisation approach under peptide‐compatible conditions.

exclude important combined effects of reaction variables. This study initially explored base and solvent effects to inform a subsequent two-level factorial design approach to understand how to control the reactivity and product selectivity in a model reaction of hexafluorobenzene. We describe new conditions that selectively afford higher order substitution products for example, 1,2,4,5-tetrasubstitution, making system, prepared by reaction of hexachlorobenzene at high temperature and pressure with potassium fluoride. [1] It has been applied to investigate tissue oxygenation in vivo, [2] but it can also be used as a solvent in organic reactions, [3] in 1 H/ 13 C NMR, UV and IR spectroscopy and as a reference standard for 19 F NMR. The presence of six inductively electron withdrawing fluorine atoms (the F atom has the highest Pauling scale electronegativity value in the periodic table), makes HFB highly electrophilic and promotes the ring reactivity to be dominated by nucleophilic aromatic substitution (S N Ar). S N Ar on poly-fluorinated arenes is a valuable method for constructing highly functionalised heterocyclic aromatic molecules. Fluoride displacement takes place with a range of nucleophiles, [4] including thiols (R-SH). The S N Ar reaction on electron-withdrawing group activated rings, usually performed in a basic medium, generally proceeds via an addition-elimination mechanism with the intermediate formation of a reactive negatively charged adduct between the arene and the nucleophile (Meisenheimer complex). [5] When HFB reacts with 2 M equivalents of a thiol nucleophile, the anion stabilization of the first thiol increases the rate of the second substitution affording exclusively disubstitution products with a 1,4-regiosubstitution pattern. Subsequent substitutions, generally observed with higher amounts of thiols and increasingly vigorous reaction conditions, follow the same regioselectivity, resulting in 1,2,4,5-tetrasubstitution and possible 1,2,3,4,5,6-hexasubstitution. Mono-, tri-and penta-substitution products are rarely isolated in significant quantities. In principle, the extent of substitution can be ordered by careful control of the reagents nucleophilicity and the reaction conditions and is summarized in Figure 1. [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] In general, there are few approaches reported to selectively afford mono-substitution of HFB, while disubstitution is generally obtained under mild conditions at room temperature. Stronger bases, longer reaction times, higher temperatures and a greater concentration of thiols can orient the reaction toward tetrasubstitution while hexa-substitution can be achieved with even harsher conditions and a larger excess of thiols. Aromatic thiolates (i. e., thiophenol or derivatives) react more readily than aliphatic thiolates and this difference influences the reaction conditions. HFB has broad applications as a molecular scaffold for cysteinecontaining peptide stapling and multicyclisation. In the presence of a deprotected cysteine-containing peptide, HFB undergoes a 1,4-disubstitution with no observable higher order substitution products (Scheme 1). Such transformations are typically performed employing DMF as solvent and either TRIS-base, Et 3 N or DIPEA as base (Table 1). [15,[21][22][23][24][25][26][27][28] In the work from Zhang et al, [22] the reaction is promoted by a glutathione S-transferase (GST) enzyme that, after TCEPmediated deprotection of the Cys (StBu), catalyzes the macrocyclisation. However, this is only applicable when the peptide sequence contains a γ-Glu-Cys-Gly (GSH) motif. Examples of the same reaction with different base-solvent combinations or alternative methodologies are not reported.
The remaining four fluorine atoms are conceptually also available for substitution, making this an excellent scaffold for stepwise F I G U R E 1 General approaches to selectively access mono-, di-, tetra-and hexa-substituted fluorobenzenes multicyclisation or poststapling introduction of additional functionality. The literature indicates that further substitution is possible using nonpeptide thiols and typically requires either higher temperature, longer reaction times and/or stronger inorganic bases.
Factorial design (FD) is a powerful experimental design approach that, with a small number of experiments, provides high-quality information in the whole experimental domain. The alternative One-Variable-At-a-Time (OVAT) approach gives very specific information only for the conditions in which the experiments have been performed.
However, a FD reflects the interactions between controllable variables (factors), while the OVAT would only be valid if the variables are totally independent from each other. [29] Therefore, FD is advantageous with respect to time, money and relevancy of the results and can be usefully applied in the chemistry field. Time, temperature, volume of solvent(s) and amount of reagent(s) are just a few variables that play a role within, or may affect the outcome of, a standard laboratory reaction. Consequently, they are factors frequently considered during optimisation processes and their mutual interaction is highly likely. FD allows these to be studied individually, in addition to the interactions between them to be observed.
The fluorine atoms of fluorobenzenes provide a diagnostic reporter for product formation by 19 F NMR. 19

| 19 F NMR analysis
All proton-decoupled 19 F NMR spectra where recorded on a Bruker Avance 600 MHz NMR, operating at 564.7 MHz at 300 K S C H E M E 1 General reaction scheme for a S N Ar on HFB with two-cysteine-containing peptides T A B L E 1 General stapling and cyclisation approaches that exploits the reaction between di-cysteine peptides and HFB

Conditions
Peptides Ref TRIS-base, TCEP, DMF, rt, 2 hours WGKGCGKGUGKGCW [28] in high-resolution NMR tubes with 0.7 mL of each sample. Signals were observed as singlets for the products of interest, which were assigned based on known 19 F NMR resonance data. [15,16] Data analysis (integration, peak-picking) was carried-out using TopSpin 3.1, Bruker UK Ltd (Coventry, UK). All spectra are provided in

| Liquid chromatography-mass spectrometry (high resolution) analysis
Characterization of crude peptides and reaction products were conducted using an Agilent 1260 Infinity II LC system with Agilent 6530 Accurate-Mass QToF spectrometer, equipped with an Agilent ZORBAX SB-C18 Stable Bond Analytical (5 μm particle size,
The reaction outcomes were analyzed with 19 Figure S3).

| Solid phase peptide synthesis
Each linear di-cysteine peptide sequence was prepared using auto-

| General procedure for peptide stapling in solution
The crude peptide (1 eq.) and any other solid reagent (e.g., Cs 2 CO 3 , 20 eq. or TCEP 2 eq.) were weighed and solubilized in small vials (from 1.5 to 3 mL, 1.5 mL of solvent for 50 mg of peptide) with magnetic stirring.
Any other liquid reagent (e.g., DBU, 20 eq.) and, last, HFB (10 eq. or 0.5 eq. as required) were added to a sealed vial that was stirred at 21   These chemical shifts are in accordance with previously reported literature assignments. [15,16] and analyzed following the procedure described above. Peptide (6)

| Initial condition scoping-OVAT
The initial aim of this study was to explore how, by applying specific combinations of reaction conditions, it is possible to selectively obtain higher conversion to either di-or tetra-thiol-substituted fluorobenzenes, under mild and peptide-friendly conditions. A wide range of procedures have been reported to afford similar products (Figure 1), however, many of these are not well suited to peptides and amino acids. For example, the use of temperatures higher than 50 C, metal catalysis and a large excess of thiols should ideally be avoided. In order to narrow the scope of conditions that could be applied to a FD space, the model reaction between N-acetyl cysteine and HFB was studied using an initial OVAT approach (Scheme 2, A). As HFB and the two major products both contain all equivalent fluorine nuclei, each exhibited a characteristic singlet at approximately −165, −134 and −96 ppm for unreacted HFB, disubstitution and tetrasubstitution products, respectively (Scheme 2B). [15,16] This allowed relative quantification of the main reaction products by signal integration and dividing by the number of fluorine nuclei, Figure S2.
Preliminary studies (see Figure S4) showed that disubstitution was obtained as the predominant product using the organic base DIPEA, even with increasing equivalents of thiol and time. However, the inorganic base K 2 CO 3 generally encouraged a larger proportion of higher-order substitution products, particularly when using a slightly  took place. Gentle heating (40 C) afforded lower conversion when compared to the same reaction at RT, likely due to the relative volatility of HFB.
We have previously reported that careful selection of solvents can increase chemoselectivity in thiol-fluoride substitutions in reactions of peptide amines/thiols with pentafluoropyridine. [26,31] Subsequently, additional bases were studied in a range of polar aprotic solvents to try to gauge the importance of properties such as dielectric constant and polarizability (Table 3)  has recently shown promise as a "green" alternative to DMF for peptide synthesis. [32] For each solvent, the bases DIPEA, DBU and cesium carbonate were probed. The reactions were performed in the required solvent (5 mL) using the required base (20 mol. eq.) and HFB (34.6 mM, 1 mol eq.), and the mixtures were stirred for 4 hours at 21 C prior to 19  precipitate as a mixture of disubstituted (major) and tetrasubstituted (minor) products. When DMSO was used as the solvent, all products remained in solution and afforded a similar product distribution as was the case for DBU in DMSO, with around 40% tetrasubstituted product.
In summary, there appears to be an overall trend for increasing solvent dielectric and polarisability to afford higher conversion to disubstitution and in some cases (DMSO as solvent) the tetrasubstituted product.
Therefore, the higher polarity solvent may be a better insulator of the charges of the thiolate nucleophile, and the Meisenheimer complex formed in the reaction. [33] The outlier to this trend is propylene carbonate which has the highest polarisability and dielectric properties of the solvents investigated, yet seems to be less efficient than DMSO in promoting substitution. This may be due to its higher viscosity affecting efficient mixing or perhaps another property not considered here. In propylene carbonate, the fluoride leaving group, which is normally a very poor nucleophile in solution, [34] may be now strong enough to perform the reverse reaction or at least compete with the thiolate. In general, we observed that selective high conversion to disubstitution or tetrasubstitution could be controlled by the choice of solvent-base combinations that is, 1,4-disubstitution can be obtained cleanly using DMSO/DIPEA or THF/Cs 2 CO 3 or propylene carbonate/DBU. Interestingly, DMF (used in perfluoroaryl peptide stapling, Table 1) does not appear to be optimal for this transformation, especially if in combination with DIPEA; and MeCN/DBU could be a valuable alternative.

| Two-level FD-screening experiment
The above analyses indicated that the reaction outcome may be sensitive to thiol concentration and the nature of the base and solvent. A more detailed FD screening experiment was used to further explore the combined roles of reaction time, temperature, concentration and reagent molar equivalence on product outcome and that may be used to control di-and tetra-fluoride-thiol substitution. This employed a two-level, five factorial full design space ( Table 2)  168 hours), samples were filtered, removing cesium salts and unreacted cesium carbonate and the reaction outcomes were measured using 19 F NMR (Figure 2A). In each case, only the major products or starting material were integrated, yet in some cases, some other minor 19 F resonances were observed (mostly mono-substitution but some others unidentified were generally observed with higher temperature and longer reaction time for example, reactions 9 and 31-see Supporting Information for raw data (Table S1) and 19 F NMR spectra ( Figures S5-S35).
Analysis of variance (ANOVA, Figure 2B (Table 1), our exploration indicated that MeCN and DMSO may provide the best combination of solubility (peptide and base) and potential for clean and rapid reaction ( Table 4).
As such, the combinations MeCN/DBU and DMSO/Cs 2 CO 3 were applied to a model di-cysteine peptide system to explore their applica-  Figure 3D). These reactions appear to happen faster than those of the model system with N-acetyl cysteine and this is possibly a proximity effect (second substitution step is intramolecular).
In each case ( Figure 3B,D), the presumed degradation products could not be ascribed to any identifiable products and no higher order substitution (i.e., tetrasubstitution -"double stapling") products were observed after prolonged reaction times. We hypothesized that a larger amount of peptide might favor the formation of such multicyclic systems. Therefore, the fast and clean reaction in DMSO/Cs 2 CO 3 was repeated with a lower amount of HFB (around 0.5 eq., Figure 3C) and TCEP. We could verify that the combination of these factors afforded the stapled product rapidly and cleanly as before, but in this case, 6 remained detectable/stable for at least 2 hours and no tetra-substitution products were observed. The reaction was complete even with a stoichiometric or sub-stoichiometric amount of HFB and TCEP may play and important role in maintaining the stability of the product.
There are many examples of single-component stapling/ cyclisation (e.g., alkene metathesis, lactamisation) carried out on solidsupported peptides, whereas similar two-component reactions are generally performed in solution because of potential site-isolation and by-products formation on resin. [39] We also wanted to determine whether the perfluoroaryl-stapling of a di-cysteine peptide was F I G U R E 4 UV chromatogram (λ = 280 nm) and mass spectrum (ESI) of the crude cleavage product following onresin perfluoroaryl-stapling reaction with DMF/DIPEA possible on resin. Solid phase reactions offer undeniable advantages such as user-friendly handling, no laborious workup, easier purification and possible automation, resulting in a greener, faster, cheaper and possibly higher yielding process. Unfortunately, our optimal stapling conditions were deemed unsuitable for polystyrene resins, therefore, we investigated the efficiency of the standard solution-phase procedure (DMF, DIPEA) for stapling a 2-Cys peptide on-resin. [26,27] After selective on-resin trityl deprotection of cysteine residues, the stapled product (6) was successfully obtained after overnight shaking of the resin with a reaction mixture made of HFB and DIPEA in DMF (Figure 4).
Finally, to demonstrate that the HFB-mediated cyclisation on solid-phase could be applied to a longer and more complex peptide sequence, we successfully cyclised the Skin Penetrating and Cell Entering (SPACE) disulfide peptide [40,41] (peptide 7) on-resin using DMF/DIPEA. The cyclised product (peptide 8) was obtained with clean conversion and only a small amount of starting material remaining after 18 hours reaction time and following cleavage from the resin (Scheme S1 and Figures S43,   S44). In addition to the procedures reported in Table 1, this is further evidence that this peptide stapling technique is applicable to changes in the peptide sequence components, length or interthiol spacing.

| CONCLUSIONS
In conclusion, the application of a two-level FD study in combination with 19 F NMR, has provided a more detailed understanding of the reactivity of HFB toward N-acetylcysteine and how this is particularly sensitive to base and solvent effects. This work also provided new conditions that afford the selective preparation of either 1,4-di or 1,2,4,5-tetra thiol-fluoride substitution products, principally controlled by the above factors. It is envisaged that these approaches can be exploited in future for the synthesis of branched or multicyclic peptide systems and poststapling modifications. Finally, new conditions (DMSO/Cs 2 CO 3 and MeCN/DBU) that permit rapid (<1 minute and < 1 hour, respectively), clean and selective peptide stapling under peptide-compatible conditions were introduced. The products obtained using our procedure (without purification) are of equal if not better crude purity than previously reported peptide stapling approaches. [15,26,27] We also demonstrate that the 2-component on-resin perfluoroaryl stapling is achievable with high crude conversion using conditions that were previously only used in solution.