A Chalcogen-Bonding Cascade Switch for Planarizable Push-Pull Probes

Abstract

Planarizable push-pull probes have been introduced to image physical forces in biology. However, the donors and acceptors needed to polarize mechanically planarized probes are incompatible with their twisted resting state. The objective of this study was to overcome this “flipper dilemma” with chalcogen-bonding cascade switches that turn on donors and acceptors only in response to mechanical planarization of the probe. This concept is explored by molecular dynamics simulations as well as chemical double-mutant cycle analysis. Cascade switched flipper probes turn out to excel with chemical stability, red shifts adding up to high significance, and focused mechanosensitivity. Most important, however, is the introduction of a new, general and fundamental concept that operates with non-trivial supramolecular chemistry, solves an important practical problem and opens much chemical space.

COMMUNICATION

To image physical forces in biology, a general concept in supramolecular chemistry that focuses on chalcogen bonds at work is introduced to access the desired stable, evolvable mechanochemistry tools - even the source of inspiration of their rational design is illustrated to perfection in freeze-thaw cycles in DMSO (Figure).

Planarizable push-pull (PP) chromophores have been introduced^[1] as mechanosensitive^[2] fluorescent membrane probes^[1–3] to image membrane tension^[4] in living cells.^[5,6] The current best is constructed around twisted dithienothiophene (DTT) dithienothiophene S,S-dioxide (DTTO2) conjugates (Figure 1a, D’ = S, A’ = SO₂).^[7] The two “flippers”^[7] are twisted out of coplanarity by repulsion between the methyls (●) and the σ holes ()^[8–11] on the sulfurs next to the connecting bond (Figure 1b, 1). The PP system is prepared first with “sulfide” donors and “sulfone” acceptors in the DTT and the DTTO2 bridges, respectively (Figure 1b, 8, 9). Conjugation of DTT and DTTO2 upon mechanical co-planarization then turns on this intrinsic PP system and shifts the excitation maximum to the red (Figure 1b, right). The emission maximum is nearly mechanoinsensitive because the probes emit only from the planar form.^[12]

Figure 1.

a) Flipper dilemma and b) CB cascade switch: a) In planarizable PP probes, exocyclic donors D and acceptors A are needed in planar but incompatible with twisted form. b) Twisting of the central bond (1) turns off PP D (2, ○) and A (3, ○) because Lewis base Y (5) hardly CBs with the shallow σ holes of DTT (4, ) and gets repelled by a lone pair (), while deep σ holes of DTTO2 (6, ) CB () Lewis base X which “back-donates” electron density (7, , left side). Upon planarization (1), PP D (2, ○) and A (3, ○) turn on because the flow of electron density from DTT to DTTO2 deepens σ holes on DTT (4, ) to allow CB () with Y to attract electron density (5, ), and fills σ holes on DTTO2 (6, ) to repel and rotate X away (7, , right side). 8, 9: Endocyclic PP DTT D', DTTO2 A', not involved in CB cascade switching.

To achieve significant red shifts upon planarization in the ground state, additional PP donors D and acceptors A are required (Figure 1a). These exocyclic substituents represent a true dilemma because in the twisted resting state, the DTT donors and DTTO2 acceptors, at least partially decoupled from each other and equipped with extra D and A, could become too rich and too poor in electron density, respectively, and decompose easily (Figure 1a). Because both DTTs and DTTO2 are comparably electron-rich,^[13] this problem is more pronounced on the DTT side. Therefore, D and A should ideally turn on only in response to flipper planarization. Sulfides, previously introduced as covalent PP turn-on donors,^[12] failed to afford operational probes.^[14] Noncovalent 1,4-chalcogen bonds (1,4-CBs)^[8–11] as in 1 were more successful, also because spontaneous degradation into reactive intermediate RI–1 could be sufficiently suppressed with a proximal triazole (Figure 2).^[15]

Figure 2.

Double-mutant cycle 1–4 to probe chalcogen-bonding cascade switching, with indication of headgroup elimination followed by the addition of nucleophiles to intermediate RI–1 as origin of the instability of 1 and 2.

To truly overcome rather than just suppress the flipper dilemma, the key was to remember the original inspiration of planarizable PP probes.^[1,16] In living lobsters, their twisted, carotenoid-orange pigments are mechanically planarized by inclusion into their binding proteins. These surrounding proteins also provide the non-covalent turn-on donors and acceptors that are needed to produce giant red shifts in planar conformation without spontaneous decomposition of the protein-free twisted conformer.^[16] Non-covalent donors and acceptors also account, combined with twisting, for color finetuning in the chemistry of vision.^[17–19] To apply these lessons from nature to flipper probes, we thought to equip exocyclic PP A and D with Lewis bases X and Y, respectively, to serve as CB acceptors (Figure 1b, 2, 3). Intramolecular 1,4 O···S and N···S interactions are well appreciated as tools for conformational control in supramolecular chemistry, medicinal chemistry, materials sciences, organic synthesis, and so on.^[9] In twisted flippers (Figure 1b, left), the PP D (2, ○) is turned off because Lewis base Y (5) fails to CB with shallow σ holes of electron-rich DTT (4, ) and instead gets repelled by a sulfur lone pair (). At the other end, the PP A (3, ○) is also off because Lewis base X binds () to deep σ holes of electron-poor DTTO2 (6, ) to “back-donate” the electron density withdrawn by the PP acceptor A (7, ). Upon planarization (Figure 1b, right), electron density flows from DTT to DTTO2 to fill DTTO2 σ holes (6, ) and deepen DTT σ holes (4, ). The deep σ hole of DTT (4, ) then CBs to Y (5, ) to promote the injection of electron density from D (2, ○), whereas the filled σ holes (6, ) cause X to be repelled (7, ) to turn on A (3, ○).

To elaborate on this CB cascade switching concept, aldehydes were selected as turn-on PP A because recent studies confirmed that back-donating CBs reduce anion binding to DTTO2 monomers.^[13,9] Intriguing PP dithienodiketophosphepins from the Baumgartner group supported triazoles as promising PP D switch.^[20] Chemical “double-mutant cycles”^[21] were completed with A switched 2 and D switched 3 besides the original 1 and the cascade switched 4 (Figure 2).

Molecular dynamics simulations were conducted to probe the correlation between CB length and the depth of s holes, the latter modeled by varying the partial charge of the chalcogen-bonding sulfur in the DTT of 4. In the resulting heatmap, the rotation of the triazole nitrogen away from the increasingly negative charge of the DTT sulfur afforded a bimodal distance distribution (Figure 3a, left). With increasing positive charge, i.e., deepened s holes, intensity for the shorter distance increased, favoring a single-peaked distance distribution with the chalcogen bond shortened from 3.34 to 3.25 Å. Complementary results were obtained for the distance distribution between the DTTO2 sulfur and aldehyde oxygen (Figure 3a, right). With the increasing negative charge on the sulfur, the CB first elongates and then breaks, with the aldehyde flipping 180° away from sulfur. These results were in perfect agreement with the concept of cascade switching (Figure 1b).

Figure 3.

a) Molecular dynamics simulations of cascade switching, showing heatmaps of the length d of 1,4 S-N chalcogen bond from triazole to DTT (left, 0 to 1800 counts, +200 per increment) and the 1,4 S-O chalcogen bond from aldehyde to DTTO2 (right, 0 to 4000 counts, +500 increments) in exhaustively sampled 4’ conformations (methyl instead of propylsulfonate group on the triazole) as a function of the partial charge on the sulfur atoms. b) Reverse phase HPLC traces of 1 (bottom) and 4 (top) after 0, 12, 24, 48, 72 and 92 hours (from back to front) in 10 mM Tris, 100 mM NaCl buffer pH = 7.4 with 10% DMSO (arrow: hydrated RI–1). c) ESI-MS of 4 (top) and 1 (bottom).

The new flipper probes 2–4 were prepared by multistep synthesis from commercially available tetrabromothiophene (Schemes S1-S3, Figures S22-S46). The deletion of the fragile thenyl carbon in 1 greatly facilitated the synthesis of 3 and 4. For illustration, the ESI MS of original 1 showed mostly RI–1 at m/z = 516, cascade switch 4 only [M+H]⁺ at m/z = 695 (Figure 3c). Fully stable in DMSO, the first degradation products of original 1 (and 2) in neutral buffer could already be observed after 12 h (Figures 3b, bottom, S13). Cascade switched 4 (and 3) were stable for weeks under the same conditions (Figure 3b, top). Continuous irradiation of flipper probes 1–4 in L_o LUVs at 500 nm showed that they bleach much slower than 5,6-carboxyfluorescein (CF) and that CB switches have little influence on this intrinsically high apparent photostability (Figure S12, note that fluorescent degradation products, e.g., hydroxylated RI-1, will pass unnoticed).

Excitation and emission spectra of the complete double-mutant cycle 1–4 were recorded in dioxane (L_b, bulk liquid, Figures 4a, S1–S10, dashed), and in the liquid-disordered (L_d) membranes composed of DOPC, the L_o SM/CL 7:3 membranes, and the solid-ordered (S_o) DPPC membranes of LUVs (Figure 4a, solid).^[7,15] For accurate assignments of maxima in the excitation spectra, the vibrational finestructures appearing upon probe planarization were resolved by spectral deconvolution, and shifts were assigned to the 0–0 transition band (Figures S4a, S11; Table 1). In all environments, the excitation maximum of the cascade switched 4 was consistently the most red shifted (Figure 4b, Table 1). Without exception, all red shifts in the double-mutant cycle showed perfect additivity (Figure 4b, red and blue bars) to ultimately reach appreciable differences for 4 vs 1 in more ordered membranes (Figure 4a, solid, red vs cyan). Such substantial red shifts can be decisive, for instance, for compatibility with super-resolution microscopy.^[22]

Figure 4.

a) Normalized fluorescence excitation spectra (λ_em = 670 nm) of 1 (cyan), 2 (pink) 3 (blue) and 4 (red) in dioxane (L_b, dashed) and S_o DPPC LUVs (solid; 75 μM lipid in 10 mM Tris, 100 mM NaCl, pH 7.4, 25 °C). b) Double-mutant cycle analysis of 1–4 in dioxane (L_b, central cycle) and in L_d (DOPC), L_o (SM/CL 7:3) and S_o (DPPC) membranes (peripheral cycle). Lowest energy excitation maxima of 1 are indicated in grey, shifts caused by CN→CHO mutations in blue, ether→triazole mutations in red (in nm). 1: L_b shifts CHO > triazole, 2: L_d shifts maximal, 3: S_o shifts triazole > CHO. c) FLIM images of SM/CL 7:3 (left) and DOPC GUVs of 4 (right, 1 μM). d) FLIM images of 4 in HeLa cells under isoosmotic (left) and hyperosmotic conditions (right; 2 μM, scale bars: 5 μm). e) Excitation spectra of a non-amphiphilic derivative of 2 (0.5 μM) in a mixture of ethyleneglycol and glycerol (1:0 to 2:3; pink to purple). f) Normalized fluorescence intensities of derivatives of 1 (cyan) and 2 (pink) as a function of the viscosity of media. CL: cholesterol; DOPC: dioleoyl-sn-glycero-3-phosphocholine; DPPC: dipalmitoyl-sn-glycero-3-phosphocholine; SM: egg sphingomyelin. LUVs: large unilamellar vesicles; GUVs: giant unilamellar vesicles. FLIM: fluorescence lifetime imaging microscopy.

Table 1:

Double-mutant cycle for chalcogen-bonding cascade switching.

cpd[a]	Lb[b] λex[c] (nm)	ΦLb[d] (%)	So[e] λex[f] (nm)	ΔλSo-Lb[g]	ISo/ILd[h] (ΦSo (%))	Ld[i] λex[j] (nm)	ΔλLd-Lb[k]	ΔλSo-Ld[l]	τLd[m] (ns)	Lo[n] λ ex[o] (nm)	ΔλLo-Ld[p]	τLo[q] (ns)	τLo/τLd[r]	HeLa τISO[s] (ns)	τhyper[t] (ns)
1	421	16	519	+98	2.4 (38)	432	+11	+87	2.92	511	+79	5.54	1.90	5.08	4.46 (1.14)
2	432 (+11)	21	524 (+5)	+92	2.8 (60)	449 (+17)	+17	+75	2.77	524 (+13)	+75	5.57	2.01	5.01	4.46 (1.12)
3	427 (+6)	7	540 (+21)	+113	6.9 (48)	454 (+22)	+27	+86	1.94	528 (+17)	+74	3.88	2.00	3.24	2.94 (1.10)
4	439 (+18)	15	547 (+28)	+108	4.0 (60)	469 (+37)	+30	+78	2.17	540 (+29)	+71	4.49	2.07	3.69	3.39 (1.09)

^[a]Compounds, Figure 2.

^[b]“Bulk liquid” dioxane.

^[c]The wavelength of the lowest energy excitation band in L_b (in deconvoluted spectra, the 0–0 transition band, Fig. S11; Δλ_ex = λ_ex − λ_ex(1)0.

^[d]Fluorescence quantum yield in dioxane, measured against Nile Red as a standard.

^[e]Solid-ordered membrane: DPPC LUVs, 25 °C.

^[f]The wavelength of the lowest energy excitation band in S_o ((in deconvoluted spectra, the 0–0 transition band, Fig. S11; Δλ_ex = λ_ex − λ_ex(1)).

^[g]Δλ_ex = λ_ex (S_o) − λ_ex (L_b).

^[h]Ratio of fluorescence intensity I in S_o divided by L_d membranes (estimated Φ_fl in S_o membranes: Φ_So = Φ_fl (L_b) × (I_So/I_Ld)).

^[i]Liquid-disordered membrane: DOPC LUVs/GUVs, 25 °C.

^[j]The wavelength of the lowest energy excitation band in L_d ((in deconvoluted spectra, the 0–0 transition band, Fig. S11; Δλ_ex = λ_ex − λ_ex(1)).

^[k]Δλ_ex = λ_ex (L_d) − λ_ex (L_b).

^[l]Δλ_ex = λ_ex (S_o) − λ_ex (L_d).

^[m]Average fluorescence lifetime in L_d membranes (from FLIM of DOPC GUVs). n = 3, SD ≤ 0.06 ns.

^[n]Liquid-ordered membrane: SM/CL 7:3 LUVs/GUVs, 25 °C.

^[o]The wavelength of the lowest energy excitation band in L_o (in deconvoluted spectra, the 0–0 transition band, Fig. S11; Δλ_ex = λ_ex − λ_ex(1)).

^[p]Δλ_ex = λ_ex (L_o) − λ_ex (L_d).

^[q]Average fluorescence lifetime in L_o membranes (from FLIM of SM/CL GUVs). n = 3, SD ≤ 0.07 ns.

^[r]Ratio of fluorescence lifetime in L_o divided by that in L_d membranes.

^[s]Average fluorescence lifetime in HeLa Kyoto cells under isotonic conditions (from FLIM images, Figure 4). n = 3, SD ≤ 0.07 ns.

^[t]Same under hyperosmotic conditions.

CL: cholesterol; DOPC: dioleoyl-sn-glycero-3-phosphocholine; DPPC: dipalmitoyl-sn-glycero-3-phosphocholine; SM: egg sphingomyelin. LUVs: large unilamellar vesicles; GUVs: giant unilamellar vesicles. FLIM: fluorescence lifetime imaging microscopy.

Consistent with turned-off cascade switches in twisted probes, the red shifts in the double-mutant cycle in L_b dioxane were overall small (Figure 4a, dashed; Figure 1b). Larger red shifts for aldehydes than for triazoles (Figure 4b, 1) could indicate that the A switch is not fully turned off in L_b dioxane (Figure 1b, 3, 6, 7). This observation calls for stronger turn-on PP acceptors.

In lipid bilayer membranes, in contrast, contributions from triazole D (red bars) were generally more important than those from aldehyde A (blue bars, Figure 4b, Table 1). Larger shifts by both D and A switches were observed in L_d membranes than in the other phases, indicating that cascade switching facilitates planarization (Figure 4b, 2; Figure 1b). In L_o membranes, red shifts were correspondingly smaller. In S_o membranes, triazole D caused much larger shifts (red bars) than aldehyde A (blue bars, Figure 4b, 3). This difference could indicate that the rotation of the aldehyde might be hindered in S_o membranes (Figure 1b, , 3). This underperformance in S_o membranes supported existence and relevance of turn-on aldehyde A in L_o and particularly L_d membranes. Overall enhanced mechanosensitivity in the biologically relevant L_d and L_o membranes and losses of mechanosensitivity in the biologically irrelevant S_o membranes identified CB cascade switching as most promising for the imaging of physical forces in biological systems.

The fluorescence quantum yield of cascade-switched flipper 4 in dioxane was with Φ_Lb = 15% almost the same as the Φ_Lb = 16% of original 1 under these conditions (Table 1, Figure S14). However, the fluorescence enhancement upon planarization in ordered membranes was larger for 4 (I_So/I_Ld = 4.0) compared to original 1 (I_So/I_Ld = 2.4) (Table 1). This implied that the effective fluorescence quantum yield of the planarized cascade switch 4 should be Φ_So > 60%, higher than the Φ_So > 38% of original 1 (Table 1). According to FLIM images of L_d and L_o GUVs,^[5,6] fluorescence lifetimes of cascade switch 4 were shorter than the original 1 (Figures 4c, S17, Table 1). However, like intensity ratios, lifetime increases upon planarization from L_d to L_o membranes were slightly higher for 4 (τ_Lo/τ_Ld = 2.07) than for 1 (τ_Lo/τ_Ld = 1.90; Table 1).

Longer lifetimes and higher quantum yield of aldehyde 4 compared to nitrile 3 might illustrate that, contrary to molecular rotors,^[2] the “brightness” of flippers reports on more than the number of rotatable bonds. Indeed, the fluorescence intensities of non-amphiphilic derivatives of 1 and 2 decreased in increasingly viscous solvents (Figures 4e, f, S16). Such behavior is opposite to molecular rotors,^[2] but consistent with the viscosity dependence of the excited state untwisting into the emissive planar form.^[12] The fluorescence decrease was accompanied by only a small (<10 nm) red shift of the excitation maximum (Figure 4e). These results support that the large red shifts and the fluorescence enhancements observed with the probes 1–4 in ordered phase of membranes originate from planarization and polarization in the ground state, reporting on the spacial confinement in equilibrium.

All members of the double-mutant cycle cleanly labeled the plasma membrane of HeLa Kyoto cells (Figures 4d, S18). Several other derivatives with different headgroups failed to do so (not shown). According to FLIM images, intensity weighted average fluorescence lifetime of cascade switch 4 decreased in response to hyperosmotic stress (Figures 4d, right, S19–S21, Table 1). This decrease was as for original 1, thus correctly reporting decreasing membrane tension as tension-induced disassembly of more ordered microdomains.^[5]

In summary, these results identify all new members of the double-mutant cycle 1–4 as operational membrane tension probes, with general access to chemical stability being the most significant advance, followed by red shifts that add up perfectly to reach important values. They introduce double-mutant cycle analysis to fluorescent membrane probes and, most importantly, validate a new, general and fundamental concept that functions with non-trivial supramolecular chemistry, solves practical problems - as exemplified with the flipper dilemma -, and opens broad new perspectives.