Pnictogen-Centered Cascade Exchangers for Thiol-Mediated Uptake: As(III)-, Sb(III)-, and Bi(III)-Expanded Cyclic Disulfides as Inhibitors of Cytosolic Delivery and Viral Entry.

Dynamic covalent exchange cascades with cellular thiols are of interest to deliver substrates to the cytosol and to inhibit the entry of viruses. The best transporters and inhibitors known today are cyclic cascade exchangers (CAXs), producing a new exchanger with every exchange, mostly cyclic oligochalcogenides, particularly disulfides. The objective of this study was to expand the dynamic covalent chalcogen exchange cascades in thiol-mediated uptake by inserting pnictogen relays. A family of pnictogen-expanded cyclic disulfides covering As(III), Sb(III), and Bi(III) is introduced. Their ability to inhibit thiol-mediated cytosolic delivery is explored with fluorescently labeled CAXs as transporters. The promise of inhibiting viral entry is assessed with SARS-CoV-2 lentiviral vectors. Oxygen-bridged seven-membered 1,3,2-dithiabismepane rings are identified as privileged scaffolds. The same holds for six-membered 1,3,2-dithiarsinane rings made from asparagusic acid and para-aminophenylarsine oxide, which are inactive or toxic when used alone. These chemically complementary Bi(III) and As(III) cascade exchangers inhibit both thiol-mediated cytosolic delivery and SARS-CoV-2 lentivector uptake at concentrations of 10 μM or lower. Crystal structures, computational models, and exchange kinetics support that lentivector entry inhibition of the contracted dithiarsinane and the expanded dithiabismepane rings coincides with exchange cascades that occur without the release of the pnictogen relay and benefit from noncovalent pnictogen bonds. The identified leads open perspectives regarding drug delivery as well as unorthodox approaches toward dynamic covalent inhibition of cellular entry.

Introduction

Thiol-mediated uptake is emerging as the method of choice to directly deliver substrates into the cytosol of cells.[1−10] The inhibition of thiol-mediated uptake is of interest because many viruses use the same mechanism to enter cells.[11] Currently unfolding understanding is centered around dynamic covalent exchange cascades with thiols and disulfides from possibly coupled cellular partners (Figure B).[1] The receptors known to participate in oligonucleotide phosphorothioate uptake[12−14] and viral entry[1,14−19] provide an impression of the scope and complexity of the network involved. For thiol-mediated cytosolic delivery, cyclic oligochalcogenides I such as asparagusic acid 1 have received most attention as dynamic covalent cascade exchangers (CAXs, defined here as exchangers that produce a new, or offer another, exchanger upon exchange; Figure A).[1,17] Upon opening by cell surface thiol/ates, they provide a tethered thiol/ate to launch dynamic covalent exchange cascades (Figure B). More recently, cyclic oligochalcogenides have been identified as unorthodox candidates to also inhibit viral entry.[11] Here, we introduce pnictogen (Pn)-expanded cascade exchangers (p-CAXs) II such as aminophenyl-dithiarsinane 2 and dithiabismepane 3 to integrate pnictogen relays into chalcogen exchange cascades (Figure ) and show that they inhibit the thiol-mediated cytosolic delivery of fluorescent transporters on the one hand and the entry of SARS-CoV-2 lentivectors on the other.

Figure 1

(A) Ring expansion of cyclic disulfides (I) with pnictogen relays (II), with AspA 1, phenyl-dithiarsinane 2, and dithiabismepane 3 as privileged scaffolds (blue ellipses: σ holes; 2 shown as zwitterion with intramolecular As···O pnictogen bond, 3 with intramolecular oxygen ligand). (B) Initiation of dynamic covalent chalcogen exchange cascades between cyclic disulfides and cellular thiols and disulfides. (C) Selected conceivable dynamic covalent chalcogen exchange cascades on pnictogen relay 2 with and (D) pnictogen relay 3 without a permanent substituent include analogous ring-opening cascades (IV–VI), hopping along thiol/ates (VII), and noncovalent pnictogen bonding and hypervalent relays via transient or permanent oxidation (VIII).

The combination of pnictogen and chalcogen exchange chemistry is not unusual in medicinal chemistry.[20−29] Arsphenamine, also known as Ehrlich’s magic bullet against syphilis, the first systematically developed drug, highlights beautifully how the emergence of pnictogen exchanging dynamic covalent networks resembles the adaptive chalcogen exchanging networks operating in thiol-mediated uptake.[24] British Anti-Lewisite (BAL, 4) operates with the same dynamic covalent chemistry.[25,26] Pnictogen relays are commonly used as antimicrobials and in nanomedicine.[20−23] Ranitidine bismuth citrate, an antiulcer drug, and related bismuth complexes have been shown recently to suppress SARS-CoV-2 replication, possibly acting as helicase or protease inhibitors.[27−29]

While ligand exchange, in general, has a long tradition as an integral part of dynamic covalent chemistry,[30,31] the combination of chalcogen and pnictogen exchange has received little attention. Pioneering work by Johnson and co-workers confirmed operational thiol/ate exchange on arsenic, antimony, and bismuth relays in dynamic covalent macrocyclic architectures.[32,33] In chemical biology, FlAsH probes[34,35] and molecular walkers exchanging along thiol tracks[36] are among the highlights. With regard to cellular uptake, facilitated, glutathione (GSH)-dependent penetration of bismuth has been proposed.[20] The cellular entry of HIV has been inhibited with para-amino-phenylarsine oxide (aPAO, 5).[37,38] In addition to this special virology example related to membrane fusion, combined pnictogen and chalcogen exchange chemistry has rarely been considered explicitly for thiol-mediated uptake and its inhibition.

Results and Discussion

Concepts

The p-CAX II appeared attractive for thiol-mediated uptake because of the promise to expand the dynamic covalent chalcogen exchange cascades offered by CAX I (III, Figure B).[1] With a permanent phenyl substituent as in 2, dynamic covalent exchange with cell surface thiol/ates on pnictogen relays has to proceed by ring opening, which liberates a thiolate that can exchange with cellular disulfides (IV, Figure C). From the resulting, doubly bridged conjugate, cascades similar to the ones known from III can evolve except that the exchanges integrate the specific advantages of pnictogen relays (V).

Without permanent phenyl substituent as in 3, chalcogen exchange with an exofacial thiol/ates can occur also without ring opening (Figure D).[36] From the resulting conjugate, ring-opening thiol/ate release for exchange with disulfides (VI) is conceivable analogue to III and IV.[25,26,36] Alternatively, p-CAX II could move along thiol/ate tracks (VII) in a manner similar to the reported walkers,[36] except that exchange cascades in thiol-mediated uptake, in general, do not occur along rigid tracks but involve significant conformational changes, as exemplified by gp120 in HIV entry,[38] possibly SARS-CoV-2 spike protein,[39] recent surprises with the HaloTag,[40] and so on. Promising new perspectives with pnictogen relays include the involvement of pnictogen bonds[41−43] and hypervalent oxidized intermediates VIII, both known to provide new mechanisms in transport and catalysis.[44−50] Already in chalcogen-centered cascades, chalcogen bonding[42,51−55] accounts for exchange without activation energy.[56] The integration of pnictogen bonding increases the strength and versatility (Figure A, vide infra). Independent of the type of bonding, these pnictogen-thiol interactions could temporarily block exofacial thiols to result in the inhibition of the thiol-mediated uptake.

Synthesis

Inhibitor candidates (2–18), reporters (19–20), and transporters (21) used in this study were, if not commercially available (4, 16, 17), prepared in a few steps following or adapting the reported procedures (Figure and Schemes S1–S7). Consistent with the literature,[25,26] and best exemplified by Ehrlich’s magic bullet, the spectroscopic data on the structure of the final pnictogen complexes were not always conclusive due to their dynamic nature. In these cases, the drawn structures should be considered as representatives for the entire dynamic covalent network involved. To facilitate reading, p-CAX structures are abbreviated, indicating pnictogen relay, ring size, and substituents, e.g., As-[6]APA for p-CAX 2, with AP standing for para-aminophenyl and A for acid.

Figure 2

Structures of p-CAXs and dithiols.

Structures

The crystal structure of As-[6]APA2 showed the cis-1,3,2-dithiarsinane in chair conformation 2a with the carboxylate in equatorial and the phenyl in axial position (Figures A and S33). Also most stable in M06-2X/6-311++G**/aug-cc-pVTZ-PP models in water (2c, Figures E,G and S12–S16), this axial phenyl might indicate a strong anomeric effect that weakens the σ hole and thus pnictogen bonding opposite to the phenyl substituent. In the solid state, dithiarsinanes 2a dimerized into supramolecular macrocycles stabilized by hydrogen-bonded ion pairs (Figure B). These dimers assembled into bilayers with the arsenic relays on their surface (Figure C). The closest As(III) contacts are two sulfurs from different chairs. Their position elongating the covalent S–As bonds was consistent with relatively weak intermolecular pnictogen bonds (3.55, 3.71 Å; the sum of the van der Waals radii (VdW) = 3.65 Å, Figure D).[57]

Figure 3

(A–D) Crystal structure of As-[6]APA2. (E) Computational models of 2, with pertinent As–X distances d and pnictogen-bond angles, compared to analogues. (F) Same for Bi-[7] 3, with Bi–X distances d compared to covalent bonds dc. (G) Representative computed structures (dVdW: As–S = 3.65 Å, As–Cl = 3.60 Å, As–O = 3.37 Å, Bi–S = 3.87 Å, Bi–Cl = 3.82 Å, Bi–O = 3.59 Å).

Pnictogen bonds in the solid state supported pnictogen-bonding contributions to exchange cascades in thiol-mediated uptake (Figure , VIII). Computed chloride complexes 2c revealed the existence of pnictogen bonds with nearly perfect angles that increased upon protonation of the arylamine (Figures E, 3.49, 3.42 Å; VdW = 3.60 Å). 1H NMR spectroscopy titrations indicated pKa = 4.0 ± 0.1 for carboxylic acid and pKa = 8.1 ± 0.1 for ammonium deprotonation (Figures S5 and S6), suggesting that in neutral water, As-[6]APA2 exists as zwitterion with deepened σ holes for strong pnictogen-bonding contributions.

Zwitterionic bicyclic trans boat 2b and trans chair with two axial substituents were less stable than cis chair 2a (+6.6, +0.8 kcal mol–1, Figure S13). The As–O distance revealed intramolecular pnictogen bonding in bicyclic 2b (3.18 Å; VdW = 3.37 Å), which weakened upon deprotonation (3.25 Å) and with p-nitro and cyanophenyl (3.20 Å) but not pentafluorophenyl substituents (3.07 Å, Figures E and S14).

For the second target Bi-[7] 3, a known compound,[25] formation of the corresponding bicyclic structure occurs by ligand exchange rather than pnictogen bonding because the permanent phenyl in 2 is replaced by a releasable ligand (Bi-[3.2.1] 3a: dc = 2.11 Å, Figure F). Pnictogen bonding of water to caged Bi-[3.2.1] 3a was strong (2.65 Å) and did not weaken upon bond switching by proton transfer (3b: 2.67 Å). Pnictogen bonding of chloride was tighter in Bi-[3.2.1] 3a (2.84 Å, 162.7°, Figure G) compared to uncaged 3c (2.95 Å, 172.7°). Thiolate ligands in Bi-[7] 3b with powerful (2.71 Å) and less favored hypervalent Bi(V) complexes 3d with weaker pnictogen bonds (3.25 Å, VdW = 3.59 Å; Figure G) were of particular interest as possible intermediates in exchange cascades during thiol-mediated uptake (Figure D, VI–VIII).

Exchange Cascades

Dynamic covalent exchange with p-CAXs was validated in an aqueous buffer solution using the environment-sensitive NBD reporter 19 (Figure A). Upon addition of p-CAXs in stoichiometric amounts, fluorescence quenching of varying degrees and rates was observed, indicating the initial formation of Pn complex 22 with two pnictogen-bonding thiols by ligand exchange (Figures A,B, S7, and S9). If ligand X is releasable, entropy gains should favor the direct formation of 23 with three thiolate ligands and only one intramolecular pnictogen-bonding thiol. From 22 and 23, ligand and pnictogen-bonding donor exchange could continue in different ways, including possible relay transfer and polymerization. Similar Pn-generated quenching was absent when the oxidized reporter 20 was used instead of 19 (Figures and S8). Consistent with computational simulations (3d, +40.3 kcal mol–1 vs 3b, Figure G), experimental support for permanent contributions from hypervalent complex VIII could thus not be secured (Figure ).

Figure 4

(A) Pnictogen exchange between the reporter 19 and various inhibitors (X = exchangeable ligand or permanent substituent (R in (B, C))). (B) Fluorescence kinetics of 19 (0.1 μM) upon addition of inhibitors at t = 0 (0.1 μM, 2: filled squares; 5: empty diamonds; 14: filled circles; 15: filled triangles; or none: gray dashed line) in buffer (5 μM TCEP, 10 mM HEPES, 0.1 M NaCl, pH 7.4), λex = 465 nm, λem = 555 nm. In parenthesis: As–Cl distances in computed complexes analogue to 2c, in Å (from Figures E and S12 and S16). (C) Same as (B) upon addition of 2 (0.1 μM) at t = 0, followed by thiols (0.4 μM, 4: orange; 16: blue; 17: green; 0.8 μM GSH: purple; or none: gray) at t = 10 min. (D) as (B) upon addition of 3 (0.1 μM) at t = 0, followed by thiols (0.4 μM, 4: orange; 16: blue; 17: green; 0.8 μM GSH: purple) at t = 100 s. (E) as (B) upon addition of 3 (0.1 μM) at t = 0, followed by GSH (0, 1, 10, 100, 500, 1000 μM) at t = 80 s.

Since the extent of quenching should depend on the pnictogens, intensity changes do not report binding strength. However, the difference in rate found within the arsenic series As-[7]AP14 > As-[6]AP15 > As-[6]APA2 > As-[-]AP5 should reflect the propensity to exchange (Figure B). This trend matched the length of the As–Cl pnictogen bonds in chloride complexes analogous to 2c except for the acyclic As-[-]AP5 without thiolate ligands, thus suggesting that computational models are predictive (Figures E, 4B, and S16). With permanent phenyl substituents necessarily present in the As(III) series and excluding possible oxidation of the pnictogen, the observed rates have to report on ring-opening thiolate exchange to initially afford 22 with two pnictogen bonded thiols. From 22, exchange of ligands and pnictogen-bond donors can continue up to the possible release of the original dithiol, excluding, however, trithiolates like 23. The observed rates supported that ring-opening exchange with the contracted 1,3,2-dithiarsinane is slower than with the expanded 1,3,2-dithiarsepane rings (As-[7]AP14 > As-[6]AP15, As-[6]APA2), and that γ carboxylates further disfavor ring opening by σ-hole inactivation and ammonium stabilization (As-[6]AP15 > As-[6]APA2; Figure E, 2b).

The reversibility of the process was confirmed by fluorescence recovery upon the addition of excess thiols (4, 16, 17, or GSH) to complex 22 and the follow-up exchange products (Figures C–E and S10). With arsenic complexes, exchange was most efficient with BAL (4), followed by lipoic acid (16). Consistent with the arsenic detoxification mechanism by BAL,[26] this trend supported that further ring contraction from formal six-membered dithiarsinane to five-membered dithiarsolane rings is favorable. The inability of DTT 17 to displace the reporter from complexes 22 and beyond (Figure C, green) confirmed that expansion from arsinanes to seven-membered arsepane rings is not favored. This finding was consistent with the fast exchange of reporter 19 with arsepane 14 (Figure B, filled circles). These differences implied that As(III) relays are easily extracted from dithiarsepanes but not from dithiarsinanes, particularly As-[6]APA2. The high toxicity of As-[7]AP14 and As-[6]AP15 further supported that the lack of toxicity and thus the inhibitory activity of 2 in living cells (vide infra) originate from the preservation of the pnictogen relays in the dithiarsinane scaffold, thus validating operational exchange cascades like IV and V that involve intermediates like 22 (Figures C and 4A) and important contributions from pnictogen bonding.

Trends opposite to As(III) were found with Bi(III), i.e, Bi-[7] 3. This complementarity originated partially from the presence of an exchangeable ligand X, which should undergo direct initial exchange with reporter 19 to trithiolate complex 23 rather than 22, also for entropic reasons. Contrary to the inactivity with As(III), the dynamic covalent exchange of DTT 17 with the larger Bi(III) complex 23 and beyond became the best. This suggested that ring-expanded seven-membered 1,3,2-dithiabismepane rings were preferred (Figure D, green). Even BAL (4) was less effective, exchanging fast but continuing to exchange further, maybe polymerize, suggesting that also the contracted five-membered 1,3,2-dithiabismolane rings were inferior to the privileged bismepanes. Also contrary to As(III), exchange with lipoic acid (16) into six-membered dithiabismane rings was clearly less favored, thus confirming the preference for bismepanes. As with the contracted As(III) ring in As-[6]APA2, these trends supported that the expanded Bi(III) ring in Bi-[7] 3 did not lose the pnictogen relay during biological activity, a conclusion that was consistent with the formation of complexes like 23 during exchange cascades, the inactivity of the contracted Bi-[5] 7, and the toxicity of the phenyl-substituted Bi-[7]P6 (vide infra).

Operational intramolecular pnictogen bonding was further supported by the inability of glutathione (GSH) to exchange with complexes 22 and 23 under the same conditions (Figure C,D, purple). However, higher GSH concentrations as in the cytosol led to exchange with complex 23 and beyond (EC50 ≈ 1 mM, Figure E). Compared to GSH, exchange with reduced BSA (bovine serum albumin) was naturally more favored considering the preorganized “vicinal” thiols obtained from disulfide reduction. Since reduced BSA contains 17 vicinal thiols, the obtained EC50 values were in the stoichiometric range for both pnictogen complexes (EC50 ≈ 6 nM for reporter 19 with 3, EC50 ≈ 50 nM for 19 with 2, Figure S11). Extrapolated to biological systems, these results implied that (1) different pnictogen relays prefer to exchange with differently spaced thiols in proteins and (2) extracellular GSH at μM concentrations would not interfere with this process, while intracellular GSH at mM concentrations could take over pnictogens from the thiols in proteins.

Inhibition of Cytosolic Delivery by Thiol-Mediated Uptake

The fluorescently labeled epidithiodiketopiperazine (ETP) 21 is the transporter of choice to report on the inhibition of thiol-mediated cytosolic delivery because it is very bright, and dynamic covalent chemistry at the disulfide does not significantly affect fluorescence intensity (Figure ).[11,58] Transporter 21 rapidly penetrates HeLa Kyoto (HK) cells and stains cytosol and nucleus.[58] High-content high-throughput (HCHT) imaging was used to simultaneously determine inhibition and cytotoxicity.[11,59] Inhibitor candidates were added to HK cells in multiwell plates and incubated first for 1 h. In a protocol referred to as pre-incubation, the inhibitors were then removed to avoid exchange with transporter 21. For co-incubation, inhibitors were kept in the media to allow for exchange also with transporter 21. After 30 min allowed for transporter 21 to penetrate the cells, Hoechst 33342 and propidium iodide were added to stain the nucleus of all cells and to label necrotic and apoptotic cells, respectively. Their ratio reported relative cell viability (RV, Figures B–E and S2–S4 and Tables and S1–S3). Only intact, propidium iodide negative cells were maintained to determine the fluorescence intensity of transporter 21. This fully automated procedure records uptake independent from toxicity, thus affording correct uptake data even at the onset of toxicity. Uptake inhibition is monitored as decreasing fluorescence with increasing inhibitor concentration and reported as minimal inhibitory concentration (MIC) at 15% inhibition and, if accessible, IC50 at 50% inhibition (Figures B–E and S2–S4 and Tables and S1–S3).

Figure 5

(A) Structure of ETP transporter 21. (B–D) Representative dose–response curves for the transporter uptake inhibition after incubation of HK cells for 1 h with inhibitors Bi-[7] 3 (B), Sb-[7]P9 (C), and As-[6]APA2 (D), followed by 21 (10 μM). Relative average fluorescence intensity (magenta circles; empty circles: insignificant due to high toxicity) and cell viability (gray empty circles), ± standard deviation (SD), fit with the Hill equation. (E) Comparison of RV50 and MIC of pnictogen inhibitors 2, 3, 5, and 8–15 for the inhibition of uptake of 21 (red circle: Bi, blue diamonds: As, green triangles: Sb). (F) Comparison of SARS-CoV-2 lentiviral vector entry (VE, %) with 10 μM inhibitor and MIC for the inhibition of the uptake of 21.

Table 1

Biological Activities of p-CAX Inhibitorsa

	Cb	Pn	MIC/IC50 (μM)c	RV50 (μM)d	vector entry (%)e	cell viability (%)f
1	2	As-[6]APA	6/29	>50	54	101
2	3	Bi-[7]	2/30	∼50	47	89
3	5	As-[-]AP	2/6	7.5
4	6	Bi-[7]P				3
5	7	Bi-[5]			>100	89
6	8	Sb-[7]	7/12	>50	>100	26
7	9	Sb-[7]P	0.7/1.7	3.2		3
8	10	Sb-[5]	5/23	>50	>100	90
9	11	Sb-[6]	9/26	>50	>100	54
10	12	Sb-[2.2.2]	10/36	>50
11	13	As-[7]P	1/6	3.9
12	14	As-[7]AP	5/9	5.6
13	15	As-[6]AP	0.6/4	19

Inhibitory activities of pnictogen-thiol compounds against the uptake of ETP transporter 21 and lentivectors.

Compounds, with the indication of pnictogen (Pn), ring size ([x]), and substituents (P, phenyl; AP, para-aminophenyl; A, acid); see Figure for structures.

The concentrations needed to inhibit the uptake of ETP transporter 21 (10 μM) by 15/50% into HeLa Kyoto cells under co-incubation conditions.

The concentrations needed to reduce the cell viability by 50% under the conditions in c.

Relative entry of SARS-CoV-2 spike pseudo-lentiviral vector into A549 cells overexpressing ACE2 and TMPRSS2 in the presence of 10 μM inhibitors.

Viability of cells under the conditions in e.

1,3,2-Dithiabismepane Bi-[7] 3 was envisioned as a lead structure (Figure F,G). An MIC = 2 μM and an IC50 ≈ 30 μM were obtained under co-incubation conditions (Figure B, red ● and Table , entry 2). Both were below a sharp onset of toxicity at ≈50 μM (gray ○). Results with and without inhibitor removal before the addition of transporter 21 were similar, indicating that p-CAX 3 exchanged mostly with cellular thiols and disulfides (Figure S2). The homologous dithiastibepane Sb-[7] 8 had an about 4 times weaker MIC = 7 μM, while the steeper dose response with co-incubation translated into a twice stronger IC50 = 12 μM (Table and Figure S3). Toxicity was low up to 50 μM. The complementary dithiarsepane was not accessible because of the toxicity of the starting material.

Responding to the search of lentivector entry inhibition in the antimony series (vide infra), an unexchangeable phenyl in dithiastibepane Sb-[7]P9 was considered to limit intramolecular O–Sb contacts to noncovalent pnictogen bonds and enforce ring-opening exchange cascades (Figure C). With MIC = 700 nM and IC50 = 1.7 μM, the phenylated dithiastibepane 9 was among the most potent inhibitors of the whole series but showed high toxicity (Figure C). Ring contraction to the minimalist dithiastibolane Sb-[5] 10 kept the MIC = 5 μM below the lead structure Sb-[7] 8. The intermediate dithiastibinane Sb-[6] 11 was less active, possibly due to the inactivation of the antimony relay by intramolecular pnictogen bonding. The corresponding trithiastibabicyclooctane Sb-[2.2.2] 12, a well-explored supramolecular chemistry motif operating with pnictogen bonds,[43] had almost the same, comparably weak MIC = 10 μM.

Permanent phenyl substituents also opened the door to arsenic relays. However, dithiarsepane As-[7]P13, the homologue of the most active Sb(III) 9, was too toxic for meaningful inhibition experiments (Figure E). High toxicity found for As-[-]AP5 was more surprising because it has been described early on as an inhibitor of HIV uptake.[37,38] Toward less toxic As(III) p-CAXs, the fusion of the aPAO and the DTT motif in As-[7]AP14 was disappointing. With record MIC = 600 nM, ring contraction from arsepane As-[7]AP14 to arsinane As-[6]AP15 was more promising, but with RV50 = 19 μM, toxicity remained rather high.

The solution was to replace the alcohol in ring-contracted As-[6]AP15 with an acid, that is to build the arsinane ring from asparagusic acid 1. The resulting aPAO-AspA hybrid As-[6]APA2 excelled with good activity (MIC = 6 μM, IC50 = 29 μM) and low toxicity up to 50 μM (Figure D,E). This finding was remarkable because the individual components were toxic (5) and inactive as inhibitors (1), respectively.[11] In As-[6]APA2, intra- and intermolecular pnictogen bonding and variable amine protonation were available to modulate the reactivity (Figure ). This could, for instance, trigger thiolate exchange but prevent dithiol release from the arsenic relay with deepened σ holes and temporary covalent As–O bonds, respectively (Figure B), or account for selectivity (Figure C,D) during exchange cascades (Figure A,C, IV–V).

Inhibition of Lentivector Entry

The potential of new p-CAXs to inhibit viral entry was explored with a lentivirus expressing the SARS-CoV-2 spike protein with D614G mutation and coding for a luciferase reporter.[11] A549 human lung alveolar basal epithelium cells overexpressing ACE2 and TMPRSS2 were treated for 1 h with p-CAXs before incubation for 6 h with the lentivirus and a 3 day period for luciferase expression. Dithiabismepane Bi-[7] 3 at a concentration of 10 μM inhibited the expression by more than 50% (Figures A and 5F and Table ). Even at 2 μM, 25% inhibition of SARS-CoV-2 lentivector uptake was observed (Figure A).

Figure 6

Normalized luminescence intensity (black) and viability (gray) for A549 human lung alveolar basal epithelium cells overexpressing ACE2 and TMPRSS2 after incubation with (A) Bi(III), As(III), and (B) Sb(III) inhibitor candidates and controls at concentrations c for 1 h, then with a lentivirus expressing the D614G SARS-CoV-2 spike protein and coding for luciferase for 6 h, followed by 3 days for luciferase expression (upward arrows: increase of uptake due, presumably, to the onset of toxicity). Experiments were performed in triplicate, error bars represent standard deviation.

The bismuth control Bi-[7]P6 was toxic, and Bi-[5] 7 promoted rather than inhibited lentivector uptake (Figure A and Table ). Uptake activation was commonly observed at the onset of membrane damage, i.e., cytotoxicity, and not further pursued, although potential use for gene transfection is understood. The different behavior of controls 6 and 7 supported that the oxygen-bridged dithiabismepane scaffold of Bi-[7] 3 is essential to inhibit uptake; bismuth extraction by complete exchange with cellular thiols would equal out the activity of Bi-[7] 3 and Bi-[5] 7. In 3, the vicinal alcohols could serve to prevent such extraction by contributing temporarily covalent Bi–O bonds.

Without bismuth, the DTT or DTE precursors fail to inhibit thiol-mediated uptake.[11] Their activation as cascade exchangers by oxidation to thiosulfonate 18 gave an IC50 ≈ 50 μM and inactivity at 10 μM (Figure A).[11,60] The inhibition of SARS-CoV-2 lentivector uptake by CAX 18 was thus five times weaker compared to p-CAX 3 with IC50 ≈ 10 μM. This comparison supported that the activation of cyclic disulfides by ring expansion with pnictogen relays is more powerful than activation by oxidation.

Moving from bismuth to arsenic, the only nontoxic inhibitor of the cytosolic delivery with transporter 21 was dithiarsinane 2 (Figure ). Inhibition of SARS-CoV-2 lentivector uptake by this nontoxic arsenic relay As-[6]APA2 was in the range of the bismuth relay Bi-[7] 3, characterized by an IC50 ∼ 10 μM (Figure A). This finding suggested that organocyclic As(III) relays are as promising as Bi(III) relays as long as toxicity is under control. We repeat that in As-[6]APA2, quite remarkably, this is realized by combining components that are either toxic (As-[-]AP5) or inactive (1) when separated, with coupled pnictogen bonding and amine protonation available to avoid excessive and achieve selective exchange cascades. It could also be argued that it is the carboxylate in As-[6]APA2 that detoxifies As-[-]AP5 because the negative charge hinders access to intracellular targets. This explanation would be supported by the persistently high toxicity of less hydrophilic analogues 13–15. However, in the context of thiol-mediated uptake, the penetration of cells and the inhibition of cellular entry are considered as different expressions of the same dynamic covalent exchange cascades (Figure ). The existence of good inhibitors that do not penetrate cells is thus unlikely, and the origin of the low toxicity of As-[6]APA2 is related to unique selectivity and reactivity (Figures and 4) rather than reduced cellular uptake. Efforts to use analogues of 2 for cytosolic delivery are ongoing and would be reported in due course.

Unlike bismuth and arsenic, antimony relays failed to inhibit lentivirus entry so far (Figure B). Sb-[7] 8, the homologue of the best bismuth relay Bi-[7] 3, was not much less potent as an inhibitor of the transporter 21 (Figure E) but only enhanced lentivirus uptake (Figure B). Presumably toxicity-related, uptake activation did not disappear even at 1 μM. Like the bismuth homologue Bi-[5] 7, ring-contracted Sb-[6] 11 and Sb-[5] 10 were increasingly toxic activators (Figure ).

Trends as for Sb-[7] 8 were found with the phenylated analogue Sb-[7]P9. This p-CAX was the most potent inhibitor of transporter 21 (Figure C,E) but as toxic as the bismuth homologue Bi-[7]P6 in the lentivector uptake assay (Figure ). The previously reported Sb(III) and Bi(III) pnictogen-bonding catalysts with three permanent 3,4,5-trifluorophenyl substituents[47] were also tested and found to be nontoxic but inactive as inhibitors of cytosolic uptake of transporter 21, also in the presence of up to five equivalents of reduced asparagusic acid 1 as pnictogen-bonding counterpart of the activating motif in As-[6]APA2 (not shown).

Conclusions

In summary, the insertion of pnictogen relays to activate cyclic disulfide cascade exchangers has afforded a family of pnictogen-expanded cascade exchangers (p-CAXs). Two conceptually new lead motifs stand out. The bicyclic bismuth relay 3 and the phenylated arsenic relay 2 exhibit rich structural diversity (Figure ) and dynamic covalent exchange cascades (Figure ), and they inhibit both the thiol-mediated cytosolic delivery with a classical transporter (Figure ) and the entry of SARS-CoV-2 lentivectors (Figure ). Their potency exceeds the activity of the previous best, that is the cyclic thiosulfonate 18, by nearly 1 order of magnitude. They are also much better than the popular ebselen,[61−63] which inhibits also both thiol-mediated cytosolic delivery with transporter 21 and the entry of SARS-CoV-2 lentivectors, but the latter only at high IC50 with high toxicity.[60] The availability of privileged scaffolds with and without permanent phenyl substituent on the pnictogen relay is important because they provide access to different cascade exchange chemistries (Figures C,D, 3, and 4). The availability of privileged scaffolds with As(III) and Bi(III) relays is attractive because their coordination chemistry covers two extremes that provide complementary characteristics in the dynamic covalent exchange cascades related to thiol-mediated uptake (Figure C,D). Namely, As(III) prefers the six-membered 1,3,2-dithiarsinane and Bi(III) the seven-membered 1,3,2-dithiabismepane rings. Exchange kinetics of these most stable rings supports the fact that lentivector entry inhibition coincides with exchange cascades that occur without the release of the pnictogen relay and involve important contributions from pnictogen bonds.

Future will tell whether the inhibition of the cytosolic delivery by thiol-mediated uptake and lentivector entry by one and the same dynamic covalent CAX is more than a coincidence. So far, bismuth- and antimony-centered antiviral candidates are mostly expected to inhibit proteases and other cysteine-rich proteins, often exhibiting multitarget activity but mostly without obvious involvement in the cellular entry.[27−29] However, the number of known CAXs capable to inhibit both the cytosolic delivery by thiol-mediated uptake and viral entry continues to increase with this study. Well-known for other viruses, particularly HIV,[1,38] the possibility of thiol-mediated uptake of SARS-CoV-2 has so far received little support. However, thiol-mediated processes have been considered,[39] and the number of receptors possibly involved in both continues to increase (transferrin receptor,[1,16,17] SCARB1,[1,15,17] PDIs,[38] integrins, etc). Thiol-mediated uptake thus increasingly emerges as complex circuitry that possibly encodes more generally for the entry into cells. Elucidation of this dynamic covalent multitarget network and the possible role of the here introduced pnictogen relays in this more general context is promising with regard to drug delivery and drug discovery, but exceptionally challenging.

Experimental Methods

Materials

Pnictogen exchangers 3,[25]5,[64]8,[25]10,[65]12,[43]13,[25] and 15(66) and transporter 21(58) were prepared by the reported procedures. Synthetic procedures and characterization of 2, 6, 7, 9, 11, 14, and 20 are given in the Supporting Information.

Computational Studies

Calculations were performed using the Gaussian 09 program (Rev. D01).[67] All structures were optimized using M06-2X/6-311++G** for light atoms (H, C, N, O, F, S, Cl) and aug-cc-pVTZ-PP for heavy atoms (As, Bi) in water solution, using the SMD model.[68] Frequency calculations were performed at the same level to confirm minima (no negative frequencies).

Pnictogen-Exchange Kinetics

Kinetics of fluorescence intensity change (λex = 465 nm, λem = 555 nm) were monitored during the addition of 20 (0.1 μM), pnictogen inhibitors (2, 3, 5, 8, 9, 13, 14, 15; 0.1 μM), and thiols (4, 16, 17, GSH, reduced BSA; various concentrations) in aqueous buffer (10 mM HEPES, 0.1 M NaCl, pH 7.4) containing TCEP (5 μM) at 25 °C.

High-Content High-Throughput Inhibitor Screening

Pre-Incubation Method

HeLa Kyoto cells were incubated with inhibitor solutions (various concentrations, duplicated) in FluoroBrite Dulbecco’s modified Eagle’s medium (DMEM) on a 96-well plate at 37 °C with 5% CO2 for 1 h. The cells were then washed and incubated with transporter 21 (10 μM) in FluoroBrite DMEM for 30 min at 37 °C with 5% CO2. The cells were then washed again and incubated with Hoechst 33342 (10 μg/L) and PI (1 μg/mL) in FluoroBrite DMEM for 15 min at 37 °C with 5% CO2. The cells were washed and kept in clean FluoroBrite DMEM at 37 °C with 5% CO2 for live cell imaging.

Co-Incubation Method

Similarly, HeLa Kyoto cells were incubated with inhibitors for 1 h and transporter 21 was added without the washing process. The rest of the procedures are identical to those of the pre-incubation method.

Imaging and Data Analyses

SDCM images of cells were acquired using an IXM-C automated microscope at 377/50 nm excitation filter and 477/60 nm emission filter for Hoechst 33342 (blue), 475/34 nm excitation filter, and 536/40 nm emission filter for transporter 21 (green), and 531/40 nm excitation filter and 593/40 nm emission filter for PI (red). The blue and red channel images were used to segment the cell bodies and detect dead cells. Then, integrated fluorescence intensity values of 21 per cell were extracted only of live cells, normalized, and analyzed using the Hill equation to give MIC, IC50, and the Hill coefficient.[11]

Lentivector Entry Inhibition

Freshly prepared stock solutions of inhibitors in dimethyl sulfoxide (DMSO) (10, 50, or 500 mM) were diluted 1000 times with culture media to give each desired solution with 0.1% of DMSO. A549 cells were treated with these solutions for 1 h and then the lentivirus (see the Supporting Information) was added. After 6 h, the culture media containing both compounds and lentivirus was discarded and fresh culture media added. After 72 h, 10 μL of the culture media was sampled and mixed with 50 μL of PBS containing 4 μM of Coelentrazine (Apollo Scientific) and luminescence generated by the Gaussia luciferase reporter was measured. Cell viability was measured in parallel with cell counting kit WST8 (Sigma Aldrich).[11] Details for all procedures are given in the Supporting Information.