How neocarcerand Octacid4 self-assembles with guests into irreversible noncovalent complexes and what accelerates the assembly

Cram’s supramolecular capsule Octacid4 can irreversibly and noncovalently self-assemble with small-molecule guests at room temperature, but how they self-assemble and what accelerates their

assembly remain poorly understood. This article reports 81 distinct Octacid4•guest self-assembly pathways captured in unrestricted, unbiased molecular dynamics simulations. These pathways

reveal that the self-assembly was initiated by the guest interaction with the cavity portal exterior of Octacid4 to increase the portal collisions that led to the portal expansion for guest

ingress, and completed by the portal contraction caused by the guest docking inside the cavity to impede guest egress. The pathways also reveal that the self-assembly was accelerated by

engaging populated host and guest conformations for the exterior interaction to increase the portal collision frequency. These revelations may help explain why the presence of an exterior

binding site at the rim of the enzyme active site is a fundamental feature of fast enzymes such as acetylcholinesterase and why small molecules adopt local minimum conformations when binding

to proteins. Further, these revelations suggest that irreversible noncovalent complexes with fast assembly rates could be developed—by engaging populated host and guest conformations for

the exterior interactions—for materials technology, data storage and processing, molecular sensing and tagging, and drug therapy.

Irreversible bimolecular complexes are highly desirable because (1) the residence of one molecule inside the cavity of another molecule is permanent until the breakdown of the

cavity-containing molecule, (2) the complex formation is one-to-one stoichiometric, and (3) for drug therapy, the permanent residence and 1:1 stoichiometry confer the cavity-residing

molecule with desired high metabolic stability, high potency-to-mass ratio, low-acting dose, and low off-target activity1,2,3,4. These complexes generally refer to irreversible covalent

complexes that are self-assembled by two molecules between which there is a covalent bond resulting from the self-assembly process. The desirability of irreversible covalent complexes is

apparent from the drug action mechanisms of aspirin5, penicillin6, and sotorasib7, all of which involve an irreversible covalent complex. Notably, sotorasib was approved in May 2021

(https://www.fda.gov/news-events/press-announcements/fda-approves-first-targeted-therapy-lung-cancer-mutation-previously-considered-resistant-drug) as the first-in-class personalized

treatment for a lung-cancer mutation previously considered resistant to drug therapy due to sotorasib’s unique capability to clinically block the function of KRASG12C (an enzyme mutant

responsible for ~13% non-small-cell lung cancers) via irreversible complexation. However, the irreversible covalent complexes from in situ cysteine conjugation are limited by the infrequent

presence of the noncatalytic cysteine in a protein cavity, and inhibiting KRASG12C offers treatment for only a subset of cancer patients. A paradigm shift is needed for the design of

irreversible bimolecular complexes.

In terms of both intrinsic binding from the thermodynamic perspective and constrictive binding from the kinetic perspective8,9,10, the irreversible complexes include irreversible noncovalent

complexes that are self-assembled by two molecules between which there is no covalent bond. Here, the intrinsic binding is the complexation governed by intermolecular interactions between

the two molecules and between the solvent and each of the two molecules, while the constrictive binding is the complexation controlled by the thermal energy required for the guest to

overcome the steric hindrance from the host during the assembly or disassembly process, as exemplified below by Cram’s supramolecular capsules known as carcerand11, hemicarcerand12, and

neocarcerand13, 14.

The hallmark of the constrictive binding is the heightened energy barrier for the assembly or disassembly of a host•guest complex. This barrier can make the dissembled or assembled molecules

kinetically stable, namely, it takes a long time to assemble the two disassembled molecules or disassemble the two assembled molecules if the barrier for the conversion is heightened. When

the barrier is extremely high, assembly or disassembly requires covalent-bond making or breaking, respectively. For example, a guest can be noncovalently trapped in the cavity of the two

bowl-shaped fragments that are rim-to-rim tethered by linkers (of a host known as carcerand) when the guest is present in the reaction medium for the tethering reaction; the disassembly of

the resulting carcerand complex is not allowed unless the host is broken by excessive heat11. When the barrier is moderately high, assembly or disassembly requires annealing. For example, a

guest can enter or exit the cavity of a host known as hemicarcerand once the bimolecular system is heated to expand the cavity portal, and the guest remains inside the cavity once the system

is cooled to contract the portal; the disassembly of the hemicarcerand complex is disallowed without heating12, 15. When the barrier is slightly high, assembly or disassembly can occur

spontaneously and slowly at an ambient temperature. For some host•guest complexes, the disassembly can be disabled at the ambient temperature by the complexation that subsequently heightens

the barrier for disassembly. For example, a guest can enter the cavity of a host known as neocarcerand at room temperature and remain in the cavity at room temperature unless the portal is

opened by heating because the docking of the guest at the cavity induces a host conformational change that consequently closes all cavity portals14.

The applications of carcerand and hemicarcerand are however limited because their complexes cannot be formed in situ and adiabatically. Although neocarcerand can form its complexes in situ

and adiabatically13, 14, 16, the applications of neocarcerand complexes are also limited due to their slow complexation rates. This underscores the need to understand how two molecules

self-assemble into an irreversible noncovalent complex and what accelerates their assembly as these high-level questions hold the key to designing irreversible noncovalent bimolecular

complexes with fast complexation rates for broad applications.

To promote irreversible noncovalent bimolecular complex design, this article reports 81 distinct pathways of the irreversible noncovalent self-assembly of neocarcerand Octacid4 with three

known guests13—1,4-dioxane (dioxane), p-xylene (xylene), and naphthalene (Fig. 1 and Table S1). These pathways were captured in multiple distinct, independent, unrestricted, unbiased, and

classical isobaric–isothermal molecular dynamics (MD) simulations at a high time resolution with an aggregated simulation time exceeding 3.761664 milliseconds at 298–363 K, rendering the

structural and kinetic information needed to answer the two high-level questions and guide the design of irreversible noncovalent complexes that can be formed in situ and adiabatically with

desired kinetics for materials technology, data storage and processing, molecular sensing and tagging, and drug therapy.

a Octacid4•p-xylene. b Octacid4•naphthalene. c Octacid4•1,4-dioxane. Carbon and oxygen are in blue and red, respectively. The axial or equatorial portal of Octacid4 comprises the C3 and C19

atoms or the O7, O15, C6b, and C17a atoms, respectively. Hydrogen and counter ion are not displayed for clarity.

In view of the current state of computational work on guest/ligand-binding pathways17,18,19,20, unrestricted and unbiased MD simulations of the self-assembly of Octacid4 with its known

guests are challenging to perform because the complexation times of Octacid4•guest were estimated to be a few minutes (or hours for the bulkiest guest naphthalene) according to NMR

experiments13. These complexation times are many orders of magnitude longer than current MD simulation times that are on the order of milliseconds.

To demonstrate the challenge of capturing the Octacid4•guest self-assembly pathways, multiple distinct, independent, unrestricted, unbiased, and classical isobaric–isothermal MD simulations

(hereafter shortened to simulations) were performed using the fully deprotonated, apo Octacid4 that was surrounded by eight neutralizing sodium cations, 10 xylenes to mimic the use of the

guest in 10 equivalent excess in the NMR experiments13, 60 NaCl molecules to approximate the ionic strength of the experimental conditions13, and 2359 TIP3P water molecules21 to mimic the

experimental aqueous solution13. Indeed, no autonomous complexation was observed in any set of 20 14,251.6-ns simulations at 298, 340, 363, or 370 K (Table S1) because the simulation time

was many orders of magnitude shorter than the experimentally estimated complexation time for xylene.

According to the reported NMR experiments13, the experimentally observed self-assembly of Octacid4 with dioxane, xylene, or naphthalene in a sodium borate buffer at pH 9 is mechanistically

driven by the sodium cation as a phase-transfer catalyst22 that chelates with the carboxylates on the Octacid4 surface and consequently accumulates the guest on the host surface via the

cation–π interaction23 or the sodium chelation. The guest accumulation on the host surface is similar to immersing the sodium-chelated Octacid4 in a neat guest solution. Simulating the

latter can substantially accelerate the self-assembly according to the law of mass action, enabling determination of relative complexation times of different guests with Octacid4 for

mechanistic insights into the self-assembly. Simulating the latter also allows the use of linear-regression analysis to examine the convergency and internal consistency of the MD

simulations. A goodness of fit (r2) of 8 Å, (iv) Δt = 1.00 fs of the standard-mass time40, 43, (v) the SHAKE-bond-length constraint applied to all bonds involving hydrogen, (vi) a protocol

to save the image closest to the middle of the “primary box” to the restart and trajectory files, (vii) a formatted restart file, (viii) the revised alkali-ion parameters44, (ix) a cutoff of

8.0 Å for noncovalent interactions, (x) a uniform 10-fold reduction in the atomic masses of the entire simulation system (both solute and solvent)40, 43, (xi) NTWX = 100 steps for

coordinates’ output, and (xii) default values of all other inputs of PMEMD.

Available in the Supporting Information of Ref. 43, FF12MClm is a revised AMBER protein forcefield with no parameterization for any Octacid4•guest complexes40. This forcefield is able to (1)

capture the experimentally observed exponential decay of the non-native state population of fast-folding proteins over simulation time with r2 > 0.90 and (2) fold these proteins with

agreements between simulated and experimental folding times within factors of 0.6–1.445. FF12MClm was used in this study to investigate the noncovalent self-assembly of small-molecule guests

with Octacid4 whose aromatic linkers can flip between left‐ and right‐handed configurations and usher the guest into the host cavity. This was because of the effectiveness of FF12MClm in

simulating the experimentally observed flipping between left‐ and right‐handed configurations for C14–C38 of bovine pancreatic trypsin inhibitor in solution40 and because of the need to

compress the simulation time (viz., speed up simulations) by a factor of 101/2 through 10-fold uniform reduction of the system mass. While the hydrogen mass repartitioning scheme can also

speed up simulations, it was not used in this study because it would affect dynamic properties of the system46. The forcefield parameters for the fully deprotonated Octacid4, dioxane,

xylene, and naphthalene are available in the Supplementary Information of Ref. 14. The forcefield parameters for the neutral Octacid4 were developed using a published procedure14 and

provided in Data S1. The ab initio calculations for developing the fully protonated Octacid4 forcefield parameters were performed using Gaussian 98 (Revision A.7; Gaussian, Inc. Wallingford,

CT). The chemical structures of HC2 and HCD2 and the assembly of HC2 into the fully protonated Octacid4 are shown in Fig. S1.

All MD simulations were performed using a dedicated cluster of 100 12-core Apple Mac Pros with Intel Westmere (2.40/2.93 GHz) and computers at the University of Minnesota Supercomputing

Institute and the Mayo Clinic high-performance computing facility at the University of Illinois Urbana-Champaign National Center for Supercomputing Applications.

The mean complexation time and its 95% confidence interval for the Octacid4•xylene self-assembly at 298 K was obtained from the 40 individual complexation times of xylene listed in Table S2

using the parametric survival function [the Surreg() function] implemented in the R survival package Version 3.2.047.

All noncovalent interaction gradient isosurfaces were generated using the NCIPLOT program (Version 4)48 with keyword RANGE (3, –0.1 to –0.02, –0.02 to 0.02, and 0.02–0.1 au) for Fig. 5 or

keywords LIGAND (4.0 Å) and RANGE (3, –0.1 to –0.02, –0.02 to 0.02, and 0.02–0.1 au) for Figs. 2–4 and 6.

All energy minimization and frequency calculations were performed using Gaussian 16 (Revision C.01; Gaussian, Inc. Wallingford, CT) and HF/6-31 G* or B3LYP/6-31 G* on computers from the Mayo

Clinic high-performance computing facility at the University of Illinois Urbana-Champaign National Center for Supercomputing Applications. Each frequency calculation was done using %mem = 

40 Gb and %nproc = 16. The Cartesian coordinates of the Octacid4 conformations (with one or two pairs of bidendate linkers) before and after the minimization are provided in Data S3.

The time series of the Cartesian coordinates of all heavy atoms of Octacid4 and the guest (that was initially outside of the Octacid4 cavity and then entered and remained inside the cavity)

was extracted from the output file of an MD simulation and saved to a concatenated PDB file using the CPPTRAJ of the AMBER 14/16/18 package. The guest involved in the complexation was

identified through the analysis of eight intermolecular distances between a guest atom (C3 for xylene, O1 for dioxane, and C1 for naphthalene) and the eight diphenoxymethane carbon atoms of

Octacid4. The eight distances for each of 150 guests were calculated from the output file using the CPPTRAJ, and the guest was considered (confirmed by visual inspection) to be inside the

Octacid4 cavity if all eight distances were