Repurposing host-guest chemistry to sequester virulence and eradicate biofilms in multidrug resistant Pseudomonas aeruginosa and Acinetobacter baumannii

The limited diversity in targets of available antibiotic therapies has put tremendous pressure on the treatment of bacterial pathogens, where numerous resistance mechanisms that counteract their function are becoming increasingly prevalent. Here, we utilize an unconventional anti-virulence screen of host-guest interacting macrocycles, and identify a water-soluble synthetic macrocycle, Pillar[5]arene, that is non-bactericidal/bacteriostatic and has a mechanism of action that involves binding to both homoserine lactones and lipopolysaccharides, key virulence factors in Gram-negative pathogens. Pillar[5]arene is active against Top Priority carbapenem- and third/fourth-generation cephalosporin-resistant Pseudomonas aeruginosa and Acinetobacter baumannii, suppressing toxins and biofilms and increasing the penetration and efficacy of standard-of-care antibiotics in combined administrations. The binding of homoserine lactones and lipopolysaccharides also sequesters their direct effects as toxins on eukaryotic membranes, neutralizing key tools that promote bacterial colonization and impede immune defenses, both in vitro and in vivo. Pillar[5]arene evades both existing antibiotic resistance mechanisms, as well as the build-up of rapid tolerance/resistance. The versatility of macrocyclic host-guest chemistry provides ample strategies for tailored targeting of virulence in a wide range of Gram-negative infectious diseases.


Supplementary
Attached as separate excel. aeruginosa. An increase in OD600 is observed in lower concentrations of P [5]a (1-100 µM) and the PBS control, which is due to the production of the toxin pyocyanin, affecting the OD600 measurement. In the 1000 µM P [5]a concentration, the toxin production is suppresses. As such, the 1000 µM concentration shows a standard growth curve. The data represent average values of biological replicates ± s.d (n=3 per group). The QTAIM molecular graph showing bond critical points and bond paths were obtained from the M062X/6-311G**electron density for the favored and 3-oxo-C12 complexes calculated using AIM2000; e) The estimated distortion and interaction energies, the binding site areas, the hydrophobic surface areas and the calculated total electron density at the bond critical points of the key vWd interactions in the binding cavity for the M062X/6-311G**electron density for the P [5]a and 3-oxo-C12, and P [5]a and 3-oxo-C6 complexes.

RNA-sequencing
RNA was extracted from the A549 cells incubated under different conditions using an RNeasy Mini Kit (Qiagen) following manufacturer's guidelines. Cells were homogenized with QIAshredder columns (Qiagen). After extraction, RNA was quantified using a Qubit 2 fluorometer (Thermo Fisher Scientific) with the Qubit RNA HS Assay Kit (Thermo Fisher Scientific), and stored at -70°C. The RNA sequencing method was designed based on the Drop-seq protocol described in 9 .
Briefly, 10 ng of RNA was mixed with Indexing Oligonucleotides (Integrated DNA Technologies were pooled together and the total RNA was isolated from three samples using an RNeasy Mini Kit (Qiagen) following manufacturer's guidelines. The rRNA was removed using a Ribo-ZeroTM rRNA Removal kit for Bacteria (Epicentre). The first strand cDNA was constructed using the SuperScript II Reverse Transcriptase (Thermo Fisher Scientific), while the second strand was constructed using random hexamers (Thermo Fisher Scientific) and HiFi HotStart Readymix (Kapa Biosystems). The sequencing library was prepared using the Nextera XT (Illumina) tagmentation reaction with 1 ng of RNA serving as an input, following manufacturer's instructions. Samples were purified twice using 0.6× and 0.9× Agencourt AMPure Beads (Beckman Coulter), and eluted in 10 µL of ultra-pure water.
The concentration of both libraries was measured using a Qubit 2 fluorometer (Thermo Fisher Scientific) and a Qubit DNA HS Assay Kit (Thermo Fisher Scientific). The quality of the sequencing libraries was assessed using the LabChip GXII Touch HT electrophoresis system (PerkinElmer), with the DNA High Sensitivity Assay (PerkinElmer) and DNA 5K / RNA / Charge Variant Assay LabChip (Perkin Elmer). Samples were stored at -70°C. The libraries were sequenced on an Illumina NextSeq 500, with the addition of the custom primer, producing read 1 of 20 bp and read 2 (paired end) of 55 bp. Sequencing was performed at the Functional Genomics Unit of the University of Helsinki, Finland. RNA sequencing data has been deposited to NCBI Gene Expression Omnibus (Accession ID: GEO Submission GSE182853).

Read alignment and RNA-seq data analysis
A549 cells. Raw sequence data was filtered to remove reads shorter than 20 bp. The original pipeline suggested for processing drop-seq data was used. Briefly, reads were additionally filtered to remove polyA tails 6 bp or longer, then aligned to the human (GRCh38) genome using STAR aligner 17 with default settings. Uniquely mapped reads were grouped according to the 1-9 barcode, and gene transcripts were counted by their Unique Molecular Identifiers (UMIs) to reduce bias emerging from PCR amplification. Digital expression matrices (DGE) reported the number of transcripts per gene in a given sample (according to the distinct UMI sequences counted). Differentially expressed genes were identified using DESeq2 18 with the cut-off for the adjusted p-value set to 0.001. A heatmap of gene expression levels was created using Heatmapper (http://www1.heatmapper.ca/). Venn diagrams were created to evaluate the distribution of differentially expressed genes between specified groups using Venny (http://bioinfogp.cnb.csic.es/tools/venny/).

P. aeruginosa.
The bacterial sequencing reads were filtered for quality and aligned against the P. aeruginosa PAO1 genome (accession number NC_002516.2) using the BowTie2 read aligner 19 . DESeq2 20 was then used to obtain the list of differentially expressed genes. Genes were considered differentially expressed if the log2 fold change was > ±2 and the adjusted P value was < 0.001. RNA sequence data has been deposited to Gene Expression Omnibus (Accession ID: GEO Submission GSE182847).

Supplementary methods Computational Discussion
Density Functional Theory (DFT) was used to explore the interactions of 3-oxo-C12 and 3-oxo-C6 with the P[5]a receptor. The M062X/6-311G** level of theory, appropriate for supramolecular interactions due to its handling of weak non-covalent interactions, was employed 10,11,21 . This choice was validated as dispersion-uncorrected functionals (i.e. B3LYP) failed to provide meaningful data. The HSLs show good surface-size complementarity with P [5]a; as the total binding surface area is generally proportional to the strength of a binding interaction, this is consistent with our expectations and the long hydrophobic chains all reside in the internal cavity which adjusts to the size of the encapsulated guest. The lactone and the neighbouring hydrophilic functionalities interact with the rim-functionalities of the P[5]a molecule. The 3-Oxo-C12 with P[5]a complexes show far higher affinity than any other pair examined in the study.

Total binding-surface areas
Electron density isosurface values. The electron density best approximates the size and shape of the molecule, thus larger molecules with more electrons feature larger isosurface values than smaller molecules with fewer electrons. As shown by the calculated density isosurfaces (Supplementary Fig.  7), the pillararene's cylinder closes around 3-oxo-C12 with P Hydrophobic/philic surface areas Hydrophilic regions are shown by contouring the "hydrophilic grid" at a negative isosurface value of -6 kcal/mol (blue surfaces) while the hydrophobic regions are shown by contouring the associated grid at a suitably negative threshold of −0.5 kcal/mol (orange surfaces). These surfaces include both the interactions between the two components, and their interaction with solvent (especially in the case of the host). Note that 3-oxo-C12 shows an extensive series of hydrophobic interaction within the cavity, and hydrophilic interactions at the mouth of the cavity due to the interaction between the lactone and the nearby functional groups with the ammonium rim groups of the pillararene. 3-oxo-C6, on the other hand, has the entire HSL inserted into the tunnel removing these beneficial interactions meaning the guest has very few interactions with the host ( Supplementary  Fig. 8). This is accompanied by a disruption of the external interactions for the host: the surface is far more disrupted meaning that there is less possible interaction between host and solvent than in either of the other two systems. These distortions and weak interactions help explain the comparably low affinity of 3-oxo-C6 for the host compared to the 3-oxo-C12.

The structural parameters
Binding is facilitated when the incorporation of the guest does not induce a significant distortion in the preferred conformation of the host. This can be quantified by looking at the relative difference in shape between the complexes and the isolated systems. As will be discussed below in 6.3, electronic distortions follow a similar principle.
For these systems, we can quantify the disruption most clearly by calculating the centroid distances in the cavity (the distances between the aromatic rings and the centre of the cavity), and torsional strain induced on the pillararene (the angle between adjacent aromatic rings; Supplementary Fig. 8d). In the isolated cavity, prepared using our computational model, the distances tend to be 4.0-4.1 Å, and the bond angle 110º. This is identical to the values observed in the crystal structure.
Along with the non-specific hydrophobic interactions, various strong, defined intramolecular interactions are responsible for the high affinity. These can be quantified as observed by the plotted topological molecular graph showing bond critical points and bond paths using quantum theory of atoms in molecules (QTAIM). The analysis is provided at Supplementary Fig. 8c. An extended table of the electronic charge density (ρBCP) and its Laplacian (∇ 2 ρ) at the bond critical point (BCP) representing both the intrahost and host-guest non-covalent interactions is provided below as Supplementary

The electronic parameters
Absolute, and shifts in, the electrostatic charge distributions To evaluate electrostatic interactions, we employed differential electrostatic potential (dESP) maps overlaid on electron isodensity contours to visualize both absolute and shift in the electrostatic charge distribution upon complexation ( Supplementary Fig. 7). In isolated P [5]a, the positive charge potential (blue) is located mostly on the rim ammonium functionalities, and the negative charge potential (red) is distributed on the outer and inner face of the pillararene cavity as well as on the counterions; however the degree of charge separation is minimal. The isolated 3-oxo-C12 and 3-oxo-C6, are largely neutral (green) with only slight charge separation on the headgroup. Upon complexation the system changes significantly, inducing shifts in electron density to maximize complex stability. The differences between the systems can be readily identified. Association with 3-oxo-C12 adds electron density to the ammonium functionalities, making the terminals of the cavity less polarized than in the isolated system. In contrast, 3-oxo-C6 dramatically distorts the electronics of the system as demonstrated by the less organized, more numerous, and more intense blue and red surfaces: the host is not stabilized by the presence of the guest. This is best illustrated by the difference maps. In these, 3-oxo-C12 induces a relative negative shift in the ammonium groups (as electron density from the guest is transferred to these charged elements). For 3-oxo-C6 the surface is far more compromised with multiple different areas shifting density demonstrating the large structural distortions to the preferred electronic distribution that occur to accommodate 3-oxo-C6; this relatively greater distortion is yet another reason for the lower affinity of this system.

Molecular Orbital Theory analysis:
To determine the change transfer and the electronic contributions of complexation process, the frontier molecular orbitals were calculated for all complexes. We find that the energy differences between the complexes are minor in all cases, the frontier molecular orbital component of the binding energy is favourable as the HOMO-LUMO gap decreases upon complexation, but the differences between the systems are minimal for both host (ΔΔEHOMO-LUMO(Host)= -0.09, and -0.17 eV for P [5]a upon complexation with 3-