Phosphorothioate backbone modification of oligodeoxynucleotides induces TLR9-independent lung epithelial antimicrobial responses

Abstract

Pneumonias remain a leading cause of death worldwide. Seeking novel strategies to protect susceptible patients, we have reported that inhaled delivery of a diacylated lipopeptide and a synthetic CpG oligodeoxynucleotide (ODN) protects animals against a broad range of infectious pneumonias by stimulating antimicrobial responses from the lung epithelium. Toll-like receptor 9 (TLR9) is well-established as the primary cellular receptor for CpG ODNs. However, we recently reported that ODNs also stimulate TLR9-independent generation of antimicrobial mitochondrial reactive oxygen species. By testing a variety of synthetic ODN molecules, we found that ODNs containing a phosphorothioate backbone, but not those with a phosphodiester backbone, induce TLR9-independent pathogen killing in lungs and improve mouse survival. Phosphorothioate-backboned ODN binds mitochondrial protein voltage-dependent anion channel 1 (VDAC1) at its N terminus, initiating pneumonia-protective metabolic reprogramming in lung epithelial cells that yield the protective antimicrobial effect. Thus, the phosphorothioate backbone of ODN is a critical structural pattern that activates TLR9-independent, metabolically-modulated innate immune protection that may be harnessed to protect vulnerable patients against pneumonia.

Introduction

Pneumonia has long been recognized as a leading cause of death among healthy and immunosuppressed people worldwide [1–3]. Increasing antimicrobial resistance, proliferation of emerging and weaponized pathogens, aging populations, and expanding categories of immunocompromised patients all contribute to a heightened risk of pneumonia in vulnerable populations [4–6]. Innate and intrinsic immune responses protect the host by recognizing broad classes of pathogens, detecting common pathogen-associated molecular patterns (PAMPs) by germline-encoded pattern recognition receptors (PRRs). Activation of evolutionarily-conserved PRRs rapidly promotes signaling and antimicrobial effector events that protect the host against invading pathogens without direct reliance on immunologic memory or epitope-level specificity [7]. Thus, therapeutic manipulation of PRRs and stimulation of innate immunity has the potential to improve survival of patients who would have otherwise died from lethal lung infections.

Our laboratory investigates strategies to manipulate the intrinsic antimicrobial capacity of the lung epithelium to prevent and treat respiratory infections in vulnerable patients. In addition to conferring a physical barrier, the airway and alveolar epithelia recognize invading pathogens, modulate lung leukocyte responses and release microbicidal molecules, including antimicrobial polypeptides (AMPs) and reactive oxygen species (ROS) [8–10]. Harnessing these immune defense mechanisms, we found that inhaled treatment with certain pattern recognition receptor (PRR) agonists targeting lung epithelial cells promotes intrapulmonary pathogen killing and host survival [11]. The lead agent we investigated includes a Toll-like receptor TLR 2/6 ligand, Pam2CSK4 (Pam2), and a Toll-like receptor TLR9 ligand, ODN M362 (ODN). A single nebulized treatment with combined Pam2-ODN provides significant protection against pneumonia caused by all tested bacterial, fungal and viral pathogens [12–16].

ODN M362, a synthetic CpG oligodeoxynuceotide, contains CpG motifs and a phosphorothioate backbone. The CpG motif is frequently found in bacterial and viral genomes and as well as in mammalian mitochondrial DNAs [17, 18]. In addition to synthetic ODNs, the phosphorothioate DNA modification is found in various bacterial genomes, including human pathogenic and commensal bacteria [19, 20]. CpG ODNs are potent immunostimulators that can activate TLR9, triggering MyD88-mediated and NF-κB or IRF-dependent antimicrobial responses [21–23].

Investigating the mechanisms of ODN-induced epithelial antimicrobial effects in the lungs, we recently discovered that the interaction of intracellular CpG ODN with mitochondrial protein voltage-dependent anion channel 1 (VDAC1) initiates immunometabolic modulation in the lung epithelium. VDAC1 and adenine nucleotide translocase 1 (ANT1) are localized in the mitochondrial outer and inner membrane, respectively. The VDAC1-ANT1-mitochondrial creatine kinase (mCK) complex regulates the exchange of metabolites between mitochondria and cytosol, including nucleotide transportation across the mitochondrial membrane [25]. ODN-VDAC1 interaction perturbs cellular nucleotide distribution, activating the AMP-activated protein kinase (AMPK) and acetyl-CoA carboxylase (ACC) pathways, and promoting fatty acid β oxidation which augments flavin adenine dinucleotide (FADH2)-carried electron flux to the electron transport chain [24]. We here examined the biochemical interactions between VDAC1 and ODN to explore the mechanisms of antimicrobial protection. We report that the phosphorothioate backbone of the ODN modulates mitochondrial energy metabolism and induces TLR9-independent antimicrobial activity via direct binding of ODN M362 to the N terminus of VDAC1.

Results

Phosphorothioate-backboned ODNs stimulate TLR9-independent mtROS generation

We have previously reported that ODN M362 induces mitochondrial ROS (mtROS) generation in lung epithelial cells [24, 26]. ODN M362 is a short, synthetic and single-stranded DNA-like molecule that has CpG dinucleotide sequences and a phosphorothioate modification to the native phosphodiester backbone. To determine which structural component of ODN M362 leads to mtROS induction, a variety of ODNs belonging to different classes of CpG ODNs and with different nucleotide sequences or secondary structures (Table 1) were tested in HBEC-3kt cells using fluorescent mtROS-detecting dye, mitoSOX. CpG ODNs containing any phosphorothioate backbone (P=S ODN) induced mtROS formation in HBEC-3kt cells while ODN with only phosphodiester backbone (P=O ODN) did not (Figure 1A), suggesting that mtROS induction was principally determined by the ODN backbone linkage and not by deoxynucleotide sequence or class. As CpG ODNs are TLR9 agonists, we tested whether TLR9 or its downstream TIR adaptor MyD88 were required for mtROS induction. Similar patterns were observed in primary mouse tracheal epithelial cells (mTECs) isolated from either wild type mice, Tlr9 knockout mice or MyD88 knockout mice upon treatment with the different CpG ODNs (Figure 1B–D), indicating that mtROS induction was mediated by P=S ODN but did not require TLR9. We further synthesized three typical CpG ODNs from classes A, B, or C, testing paired ODNs that shared the same respective sequences varied by either P=S or P=O backbone linkages. Only P=S backboned ODNs induced mtROS production, and they did so independent of sequence or class (Figure 1E). In contrast, both P=S and P=O ODNs triggered activation of NF-kB (Figure 1F, Figure S1), reflecting sequence-dependent but not backbone-dependent activation of TLR9 signaling events. Thus, in addition to the canonical TLR9 signaling pathway, P=S ODNs activate TLR9-independent mtROS generation.

Table 1.

Synthetic CpG oligonucleotides (ODNs)

Name	Class	Backbone	Structure	Sequence	Vendor
ODN M362	C	P=S	Single Stem-loop	tcgtcgtcgttc:gaacgacgttgat	Invivogen
ODN 2395	C	P=S	Single Stem-loop	tcgtcgttttcggcgc:gcgccg	Invivogen
ODN 2006	B	P=S	Single Linear	tcgtcgttttgtcgttttgtcgtt	Invivogen
ODN 2006G5	B	P=O	Single Linear	TCGTCGTTTTGTCGTTTTGTCGTTGGGGG	Invivogen
ODN 2216	A	P=S & P=O	Single Stem-loop	ggGGGACGA:TCGTCgggggg	Invivogen
ODN 1585	A	P=S & P=O	Single Stem-loop	ggGGTCAAC:GTTGAgggggg	Invivogen
ODN SL01	B	P=S	Double Stem-loop	tcgcgacgttcgcccgacgttcggta	Invivogen
ODN SL03	C	P=S	Double Stem-loop	tcgcgaacgttcgccgcgttcgaacgcgg	Invivogen

Figure 1. Phosphorothioate-backboned ODNs stimulate TLR9-independent mtROS generation.

(A) mtROS production after treatment with the indicated ODNs in HBEC-3kt cells. (B) mtROS production after treatment with the indicated ODNs in wild type (B), Tlr9^−/− (C) and MyD88^−/− (D) mTECs. (E) mtROS after treatment with ODNs of identical sequence but different backbones in HBEC-3kt cells. (F) NF-κB nuclear translocation measured by imaging flow cytometry after treatment with the indicated ODNs. * p≤0.0002 vs. ODN M362 by one-way ANOVA using Holm-Sidak method; † p<0.0001 vs. PBS by one-way ANOVA using Holm-Sidak method. ODN, oligodeoxynucleotide; P=O, phosphodiester backbone; P=S, phosphorothiorate backbone; mTEC, primary mouse tracheal epithelial cells.

The phosphorothioate backbone modification of ODN promotes mtROS formation

The phosphorothioate backbone modification is the most widely used chemical modification in oligonucleotide-based therapeutics [27]. Phosphorothioate ODN modification replaces a non-bridging oxygen molecule in the native phosphodiester (P=O) backbone with a sulfur atom to make a phosphorothioate (P=S) backbone (Figure 2A). The resulting P=S ODN confers increased nuclease resistance [28–30]. To examine whether the enhanced biostability of P=S ODNs is responsible for the increased mtROS generation, we synthesized a 5-methylcytosine modified ODN M362 (Figure 2A) in which the methyl group on the cytosine base comparably increases intracellular ODN stability to that achieved by P=S modification [31]. When treated with P=S, P=O or methyl ODN M362, we found that P=S ODN induced significantly more mtROS in HBEC-3kt cells, mTECs, and primary normal human bronchial epithelial cells (NHBE) than did P=O or methylated molecules (Figure 2B–D). Treatment with the endonuclease inhibitor aurintricarboxylic acid (ATA) [32] had no effect on mtROS induction for P=S or P=O backboned ODNs (Figure 2E), further demonstrating that nuclease resistance or increased bioavailability was not responsible for the P=S backbone effect. Taken together, these results indicate that the phosphorothioate backbone modification of CpG ODNs induces epithelial mtROS production.

Figure 2. The phosphorothioate backbone modification of ODN promotes mtROS formation.

(A) Chemical structure of synthetic CpG ODNs, shown are ODNs with phosphodiester backbone, phosphorothioate backbone and 5-methylation at cytosine base. mtROS production from HBEC-3kt cells (B) and primary mouse (C) and primary human (D) lung epithelial cells after treatment with ODNs of identical sequence but different in backbone or 5-methylcytosine modifications. (E) mtROS production in the presence or absence of a DNA endonuclease inhibitor in HBEC-3kt cells. * p<0.001 vs. PBS by one-way ANOVA using Holm-Sidak method. ** p<0.0001 vs. all other groups by one-way ANOVA using Holm-Sidak method. P=O, phosphodiester backbone; P=S, phosphorothiorate backbone; mTEC, primary mouse tracheal epithelial cells; NHBE, primary normal human bronchial epithelial cells; ATA, aurintricarboxylic acid.

Phosphorothioate-backboned ODN but not phosphodiester-backboned ODN alters mitochondrial energy metabolism

We have reported that ODN M362 interacts with VDAC1, resulting in a series of metabolic shifts in mitochondrial energy production [24]. ODN M362 alters cellular ATP/ADP/AMP localization, increases delivery of electrons to the electron transport chain (ETC), increases mitochondrial membrane potential (Δ_Ψm), differentially modulates ETC complex activities and consequently results in mtROS formation at ETC complex III [24]. We hypothesized that P=S ODN but not P=O ODN causes such lung epithelial metabolic modulation. As in Figure 3A, mitochondria isolated from epithelial cells treated with fluorescently-labeled P=S ODN demonstrate significant increases in fluorescence intensity compared to mitochondria isolated from cells treated with unlabeled ODN, while mitochondria isolated from cells treated with fluorescently-labeled P=O ODN did not. We further found that direct treatment with P=S ODN but not P=O ODN in isolated live mitochondria recapitulated the inducible mtROS generation in cells (Figure 3B). In targeted pulldown assays with mouse and human cells, more VDAC1 and mitochondrial ADP/ATP carrier protein 1 (ANT1) were detected in immunoprecipitation of P=S ODN treated cells than that of P=O ODN treated cells (Figure 3C–E, Figure S2).

Figure 3. Phosphorothioate-backboned ODN but not phosphodiester-backboned ODN alters mitochondrial energy metabolism.

(A) Fluorescence intensity of mitochondria isolated from HBEC3-KT cells treated with FITC-labeled or unlabeled P=S or P=O ODNs. (B) mtROS production in mitochondria isolated from mouse lungs and treated with P=S ODN or P=O ODN. Mitochondria were isolated from the biotinylated-ODN-treated HBEC3-KT cells, lysed, incubated with streptavidin beads. The streptavidin precipitants were resolved by polyacrylamide gel electrophoresis and then probed for VDAC1 in (C) human or (D) mouse mitochondrial lysates following treatment with biotinylated-P=S ODN or P=O ODN or probed for ANT11 in (E) human mitochondrial lysates following treatment with the indicated biotinylated-ODNs. Measurements of cytosolic (F) and mitochondrial (G) levels of ATP in P=S ODN or P=O ODN-treated HBEC3-KT cells. (H) Phospho-AMPKa ELISA in P=S ODN or P=O ODN-treated HBEC3-KT cells. Measurements of fatty acid oxidation (I), acetyl-CoA levels (J), ratio of reduced:oxidized CoQ in mitochondria (K), complex II (L), III (M) and V (N) activities and Δ_Ψm (O) in HBEC3-KT cells treated with P=S ODN or P=O ODN. (P) Oxygen consumption following the indicated treatments, shown as mean ± SEM. * p<0.0001 vs unlabeled ODN (P=S) by Student’s t test; ** p<0.001 vs PBS by one-way ANOVA using Holm-Sidak method; † p=0.003 vs PBS by one-way ANOVA using Kruskal-Wallis method; ‡ p≤0.003 vs PBS by one-way ANOVA using Holm-Sidak method. P=O, phosphodiester backbone; P=S, phosphorothiorate backbone; FAO, fatty acid oxidation; CII, complex II; CIII, complex III; OCR, oxygen consumption rate.

We found that P=S ODN treatment caused mitochondrial ATP levels to increase and cytosolic ATP levels to decrease inside cells, while P=O ODN had much less effect (Figure 3F & 3G). Congruently, only P=S ODN induced phosphorylation of AMPKα (Figure 3H). Together, these P=S ODN-induced and AMPK-regulated metabolic shifts promoted fatty acid β-oxidation (Figure 3I), increased acetyl-CoA production (Figure 3J), transferred additional electrons to ETC via the TCA cycle (Figure 3K) and elevated ETC complex II activity (Figure 3L). Additionally, P=S ODN treatment but not P=O ODN treatment caused decreases in ETC complex III and complex V activity (Figure 3M & 3N). The P=S ODN induced mitochondrial ATP accumulation is consistent with the increased mitochondrial membrane potential Δ_Ψm observed in P=S ODN treated cells (Figure 3O). By inhibiting oxidative phosphorylation with oligomycin, we observed increases of oxygen consumption which is caused by mtROS formation following P=S ODN treatment but not after P=O ODN treatment (Figure 3P). In summary, our data indicate that phosphorothioate-backboned ODN but not phosphodiester-backboned ODN drives the protective mitochondrial energy metabolism modulation. While modest binding of P=O ODN to VDAC1 is noted by immunoprecipitation assays, this effect is too small to be detected above background fluorescence. Nonetheless, this modest binding likely explains the intermediate phenotype observed in some of the metabolic readouts.

The phosphorothioate backbone of ODN elicits TLR9-independent antimicrobial activity

We have shown that ODN M362-induced metabolic modulation is associated with improved host and lung epithelial survival of bacterial infections [24]. We next examined whether P=S ODN provides greater antimicrobial effects than P=O ODN by assessing the bacterial burden in cells. Delivering Pam2 in combination with either P=S ODN or P=O ODN, we tested cultured HBEC3-KT cells (Figure 4A), MLE15 cells (Figure 4B), wild type mTECs (Figure 4C & 4D) and wild type mouse alveolar epithelial cells (AEC) (Figure 4E) as either submerged or air-liquid interface cultures. In combination with Pam2, P=S ODN induced greater bacterial killing than P=O ODN in all tested cells. Results of in vitro cell viability assay after Pam2ODN treatment in HBEC3-KT and MLE15 cells confirmed no cellular damage imposed by Pam2ODN (Figure S3). Notably, a similarly greater protective effect of P=S ODN over P=O ODN was observed in Tlr9 knockout mTECs (Figure 4F & 4G), reinforcing the finding that the P=S backbone-induced protection is not TLR9 dependent. Consistent with the in vitro data, P=S ODN protected both wild type and Tlr9 knockout mice against bacterial pneumonia better than P=O ODN (Figure 4H & 4I), further supporting the observation that the phosphorothioate backbone of ODN elicits TLR9-independent antimicrobial activity.

Figure 4. The phosphorothioate backbone of ODN elicits TLR9-independent antimicrobial activity.

Bacterial burden of HBEC3-KT cells (A) or MLE15 cells (B) treated with the indicated ligands. Bacterial burden of wild type primary mouse tracheal epithelial cells growing submerged (C) or at air-liquid interface (D) or wild type primary mouse alveolar cells growing at air-liquid interface (E) after treatment with Pam2 and P=S or P=O ODNs. Bacterial burden of Tlr9^−/− primary mouse tracheal epithelial cells growing submerged (F) or at air-liquid interface (G) after treatment with Pam2 and P=S or P=O ODNs. (D) Survival of wild type mice challenged with P. aeruginosa one day after nebulized treatment with Pam2 and P=S or P=O ODNs (n=10 mice/group). (E) Survival of Tlr9^−/− mice challenged with P. aeruginosa one day after nebulized treatment with Pam2 and P=S or P=O ODNs (n=25 mice/group). * p<0.003 vs PBS by one-way ANOVA using Holm-Sidak method; * p<0.01 vs PBS by one-way ANOVA using Kruskal-Wallis method; † p<0.05 vs Pam2ODN (P=O) by one-way ANOVA using Holm-Sidak method; ‡ p<0.008 vs PBS by log ranks using the Mantel-Cox method. P=O, phosphodiester backbone; P=S, phosphorothiorate backbone; mTEC, primary mouse tracheal epithelial cells; AEC, primary mouse alveolar cells.

Phosphorothioate ODN directly binds to mitochondrial membrane protein VDAC1

Our metabolic data indicates that interaction between P=S ODN M362 and VDAC1 plays a crucial role in initiating TLR9-independent antimicrobial responses. The P=S ODN perturbed intracellular ATP distribution suggests that P=S ODN can directly bind VDAC1. To test this hypothesis, we adopted an in silico ligand-protein docking and molecular simulation approach. In the absence of a published 3D structure of ODN M362, we extracted 14 3D structures of single strand phosphorothioate DNAs deposited in the RCSB Protein Data Bank (RCSB PDB), ranging from 5 to 22 nucleotides in length (Table S1). These phosphorothioate DNAs were first tested with the mitoSOX assay to validate their capacity of mtROS induction. As shown in Figure S4, phosphorothioate DNAs induced significant mtROS generation in HBEC3-KT cells but at a level less than that the 25 nucleotide-long ODN M362 did. The 3D structure of ODN M362 was predicted using RNA COMPOSER and PyMOL (Figure S5). Molecular docking was performed between structures of phosphorothioate DNA ligands and human VDAC1 (PDB: 2jk4) (Table S1, Figure S5). Predicted receptor-ligand interfaces were listed in Table S2 and were summarized in Figure S6 where common receptor interface residues were visualized. The N-terminal amino acid residues from 8 to 45 (ADLGKSARDVFTKGYGFGLIKLDLKTKSENGLEFTSSG) was identified as the fragment most likely harboring phosphorothioate DNA binding sites (Figure 5A). We synthesized the 38-amino acid peptide to carry out ligand-receptor binding assays and found that the VDAC1 N-terminal peptide is directly bound by P=S ODN M362 but not by P=O ODN M362, as demonstrated by electrophoretic mobility shift assay (Figure 5B, Figure S7). Fluorescence polarization assays measure the kinetics of ligand-receptor reactions [33]. Fluorescence polarization assays demonstrated that the binding affinity of VDAC1 N-terminal peptide for P=S ODN is much higher than that for P=O ODN (Figure 5C). As VDAC1 and P=S ODN interaction is mainly determined by electrostatic force established between negatively charged sulfur groups in P=S ODN molecules and positively charged ammino acid residues in VDAC1, we replaced positively charged amino acid residues in VDAC1 N-terminal peptide with either negatively charged or neutral amino acids (Figure 5D). The 38-amino acid mutated peptide (ADDGESADEVFTEGYGFGLIELDLETESENGLEFTSSG) cannot bind P=S ODN M362 as demonstrated by electrophoretic mobility shift assay (Figure 5E, Figure S7) and fluorescence polarization assay (Figure 5F). These biochemical data support P=S ODN M362 binding VDAC1 at its N-terminal domain.

Figure 5. Phosphorothioate ODN binds mitochondrial membrane protein VDAC1 N terminus.

(A) AlphaFold-predicted and PyMOL-generated 3D structure of the 38-amino acid VDAC1 N-terminal peptide. (B) Electrophoresis mobility shift assay of VDAC1 N-terminal peptide incubated with P=S ODN or P=O ODN. (C) Fluorescence polarization assay of VDAC1 N-terminal peptide incubated with fluorescein (FAM) labeled P=S ODN or P=O ODN. (D) AlphaFold-predicted and PyMOL-generated 3D structure of the 38-amino acid mutated peptide in which positively charged amino acids were replaced with either negatively charged or neutral amino acids. (E) Electrophoresis mobility shift assay of VDAC1 N-terminal peptide or the mutated peptide incubated with P=S ODN. (F) Fluorescence polarization assay of VDAC1 N-terminal peptide or the mutated peptide incubated with FAM labeled P=S ODN.

VDAC1 binding phosphorothioate ODN contributes to ODN protection against bacterial infection

In addition to phosphorothioate ODN, other chemical inhibitors such as erastin, cyclosporin A or VBIT-4 have been used to inhibit VDAC1 by different mechanisms yet yielded parallel effects on blocking nucleotide transportation, mtROS production and pathogen killing [24]. These inhibition data suggest that impairment of VDAC1 function leads to metabolic reprogramming and generates antimicrobial responses. It has been reported that knockout of Vdac1 alters mitochondrial respiration, promotes reactive oxygen species generation and activates hypoxia-inducible factor pathway in mouse cells [34], indicating that genetic deletion of Vdac1 results in a similar deficiency of VDAC1 function as the chemical inhibition does. We hypothesized that ODN M362 binding to VDAC1 N terminus interrupts VDAC1 function and induces TLR9-independent antimicrobial activity. We investigated whether preventing ODN-VDAC1 binding reduces mtROS generation and inhibits ODN-induced antimicrobial responses. We used a competitive peptide binding strategy to inhibit ODN-VDAC1 interaction. Briefly, a library of peptide sequences was designed with random variations in the VDAC1 N-terminal sequences. After adding a cell-penetrating sequence to the N-terminus of the peptides [35], we used AlphaFold and PyMOL to predict the 3D structure of each candidate in the competitive peptide library. Next, molecular docking was performed to predict interactions between these predicted structures of the competitive peptides and 3D structure of ODN M362. PyMOL was used to visualize peptide-ODN M362 interactions and obtain receptor-ligand interface residues (Figure S8). Free energy of each peptide-ODN M362 pair was calculated with a PRODIGY (PROtein binDIng enerGY prediction) algorithm and the top 7 candidates with the highest predicted binding affinity were listed in Table 2 and synthesized. These 7 competitive inhibitory peptide candidates were added individually to HBEC-3kt cells prior to ODN M362 treatment. The Lys31Ala substitution of peptide 61 and the Thr33Ala substitution of peptide 63 both showed decreased ODN-induced mtROS formation (Figure 6A). In contrast, the Lys31Met substitution of peptide 88 showed no inhibition of mtROS generation in ODN-treated cells (Figure 6A). Fluorescence polarization assay was applied to compare binding of the P=S ODN ligand with VDAC1 N terminal peptide and the peptides 61, 63 or 88, and the difference in fluorescence polarization (Δ mP) between mixed protein/ligand samples and control ligand only sample were measured. In Figure 6B, peptide 61 & 63 presented with greater binding ability than the wild type VDAC1 peptide at a higher concentration (> 5 μM) and peptide 88 showed the least binding. Additionally, competitive peptide 61 displayed both dose-dependent and incubation time-dependent inhibitory effects on mtROS formation (Figure 6C & 6D).

Table 2.

Sequences of inhibitory peptide design and simulation of ODN M362 binding

#	Peptide sequence*	Predicted free energy of peptide binding to ODN M362
	GRKKRRQRRRPPQ LGKSARDVFTKGYGFGLI (10–27)	ΔG= −87.54 kcal mol−1
19	GRKKRRQRRRPPQ LGKSARDVFTKGYGFGLA	ΔG= −112.85 kcal mol−1
22	GRKKRRQRRRPPQ DGKSARDVFTKGYGFGLI	ΔG= −125.94 kcal mol−1
40	GRKKRRQRRRPPQ LGKSARDVFTKGYGFGAI	ΔG= −115.06 kcal mol−1
54	GRKKRRQRRRPPQ LGKSARDVFTKGYGFGRI	ΔG= −117.52 kcal mol−1
	GRKKRRQRRRPPQ LKTKSENGLEFTSSGSA (31–47)	ΔG= −87.43 kcal mol−1
61	GRKKRRQRRRPPQ AKTKSENGLEFTSSGSA	ΔG= −119.66 kcal mol−1
63	GRKKRRQRRRPPQ LKAKSENGLEFTSSGSA	ΔG= −117.50 kcal mol−1
88	GRKKRRQRRRPPQ MKTKSENGLEFTSSGSA	ΔG= −111.31 kcal mol−1

^*Italic sequences are TAT cell penetrating peptide sequences, bode sequences are with or without variations of amino acid residue in the VDAC1 N-terminal peptide sequences.

Figure 6. VDAC1 binding phosphorothioate ODN contributes to ODN protection against bacterial infection.

(A) mtROS reading after incubation with candidate competitive peptides and then treated with ODN in HBEC-3kt cells. (B) Fluorescence polarization assay of VDAC1 N-terminal peptide or competitive peptides incubated with fluorescein (FAM) labeled P=S ODN. (C) mtROS reading after incubation with competitive peptide 61 at indicated concentrations and then treated with ODN. (D) Relative mtROS reading after incubation with competitive peptide 61 at indicated times and with ODN treatment. Measurements of Phospho-AMPKa ELISA (E), Δ_Ψm (F), Complex II (G) and Complex III (H) activities in HBEC3-KT cells incubated with competitive peptide 61 and control peptide 88 and with ODN treatment. Bacterial burden of HBEC3-KT cells (I) or MLE15 cells (J) incubated with competitive peptide 61 and control peptide 88 and treated with ODN. * p<0.005 vs ODN by one-way ANOVA using Holm-Sidak method; ** p≤0.002 vs ODN by one-way ANOVA using Kruskal-Wallis method.

We next tested whether ODN binding by competitive peptide 61 abolished P=S ODN-induced metabolic shifts and protection phenotype, using peptide 88 as a negative control. Competitive peptide 61, but not peptide 88, abrogated ODN-induced phosphorylation of AMPKα (Figure 6E) and ODN-enhanced Δ_Ψm (Figure 6F) and restored the basal level of Complex II and Complex III activity in ETC (Figure 6G & 6H). Further, adding competitive peptide 61, but not control peptide 88, abrogated the protective effect by ODN against bacterial infection in both HBEC3-KT cells (Figure 6I) and MLE15 cells (Figure 6J).

In summary, our findings indicate that the phosphorothioate backbone modification is a molecular pattern that modulates mitochondrial energy metabolism and induces TLR9-independent antimicrobial activity. We demonstrated that P=S ODN binds mitochondrial membrane proteins VDAC1 at the N terminus, blocks mitochondrial nucleotide transport and eventually leads to metabolic reprogramming that generates downstream signals for activation of protective antimicrobial responses.

Discussion

Synthetic CpG ODNs usually have two structural modifications, CpG motifs and phosphorothioate linkages, which mimic naturally occurring pathogen-associated molecular patterns recognized by PRRs. After cellular uptake, CpG ODNs are initially localized in early or late endosomes [30, 36], where endosomal TLR9 recognizes CpG ODNs via the CpG motif [17, 18, 37]. However, TLR9 binding is not the only potential outcome for internalized ODNs. Internalized ODNs traffic through the major endocytic process, of which many are released from late endosomes and some are cleared in lysosomes by degradation [38, 39]. Intracellular ODNs can enter the cytosol, the nucleus or other organelles, where proteins are rich and allow for ODN binding. Interestingly, the phosphorothioate backbone modification predominantly determines which protein is preferentially bound by intracellular ODNs [30, 40]. We have shown that mitochondrial protein complex VDAC1/ANT1 can be pulled down together with P=S ODN and that P=S ODN is colocalized with VDAC1 in lung epithelia cells [24]. Here we provide biochemical evidence that P=S ODN binds VDAC1 at its N terminus.

VDAC1 protein has a unique 3D structure including a pore-forming β-barrel with 19 transmembrane β-strands and a horizontally oriented α-helix inside the pore harboring nucleotide binding sites [41]. The N-terminal region of VDAC1 constitutes a mobile fragment consisting of the α-helix (amino acids 6–20) and a linker sequence that connects the α-helix to β-strand 1 of the β-barrel [42]. To facilitate nucleotide transportation, the N-terminal α-helix is proposed to be involved in voltage gating and the conserved linker sequence is to move the N-terminal domain out of the pore and expose it to the cytosolic side of the β-barrel [43]. The motion of the N-terminal region is essential for ATP transport because ATP passes through the VDAC1 pore by interacting sequentially with a series of lysine or arginine residues on the α-helix, the linker, and β-barrel wall [44]. Thus, when the VDAC1 N-terminal domain is occupied by P=S ODN M362 mitochondrial nucleotide transport is barricaded.

Phosphorothioate linkage enhances nuclease resistance, improves cellular uptake and sub-cellular distribution and prolongs the in vivo half-life of P=S ODNs. Furthermore, phosphorothioate modification allows P=S ODNs to adopt flexible conformations around the P=S backbone and facilitates the more anionic sulfur atoms of the P=S linkages interacting with positively charged amino acids on target proteins [40, 45]. These features explain why P=S ODN can bind much more tightly with VDAC1 inside cells than can P=O ODN. Electrostatic interactions determine the VDAC1 and P=S ODN binding as substitution of positively charged amino acids with negative or neutral ones in the VDAC1 N-terminal peptide completely abolishes formation of the peptide and P=S ODN complex. We leveraged these findings to show that P=S ODN-VDAC1 binding activates protective metabolic signaling pathways against bacterial infections in the lungs. The protective effect of P=S ODN vanishes when the P=S ODN-VDAC1 binding is competitively inhibited. Therefore, the P=S ODN protection phenotype is contingent upon the presence of a phosphonothioate modification in the ODN backbone. We thus report that the unique structural pattern presented by P=S ODN has previously unpublished, TLR9 independent antimicrobial functions.

Phosphorothioate modification can enhance ODN bioavailability, biodistribution and biophysical interactions of CpG ODNs in animals and humans. Congruent with our published data and indicative of effects beyond simple nuclease resistance, it has been reported that treatment with G3139, a P=S ODN, inhibits mitochondrial metabolite flux [46, 47]. However, how a modified backbone structure could impact ODN immunostimulatory functions has not been previously explored. Here we provide biochemical evidence that P=S ODN M362 but not P=O ODN M362 binds VDAC1 and disrupts the ATP distribution across mitochondrial membranes in lung epithelia cells. We identify an immunometabolism-modulated mechanism of pathogen clearance that is triggered by the P=S backbone binding to a mitochondrial protein. Thus, both the P=S modification and the CpG motif of CpG ODNs contribute to distinct therapeutic capabilities to stimulate innate immune defense through parallel mitochondrial VDAC1 or endosomal TLR9 pathways. This new understanding of the biochemical mechanisms of ODN interaction supports P=S CpG ODN as a useful intervention to prevent or treat respiratory infectious diseases in vulnerable patients.

Materials and methods

Tissue cultures and cell lines

To isolate mouse tracheal epithelial cells (mTECs), mice were anesthetized and tracheas were excised and digested in 1.5 mg/ml Pronase overnight at 4 °C. mTECs were harvested and cultured on collagen coated tissue culture plates or transwells in Ham’s F12 media supplemented with differentiation growth factors and hormones as previously described [48]. Mouse alveolar epithelial cells were harvested and cultured on transwells as previously described [24]. Immortalized Human bronchial epithelial (HBEC3-KT) cells and murine lung epithelial (MLE-15) cells were cultured as previously described [24]. All human cell experiments were performed in accordance with Institutional Review Board of The University of Texas MD Anderson Cancer Center (MDACC).

Mice

Wild type C57BL/6J mice and MyD88^−/− mice were purchased from The Jackson Laboratory (Bar Harbor, ME). TLR9^−/− mice were provided by Dr. Shizuo Akira [49]. All mouse experiments were performed in accordance with the MDACC Institutional Animal Care and Use Committee.

In vivo Pam2ODN nebulization

10 ml of combined 4 μM Pam2CSK4 (Invivogen) and 1 μM ODN M362 (Invivogen) in 1× phosphate-buffered saline (PBS) was placed in an Aerotech II nebulizer (Biodex, Shirley, NY) and delivered to unrestrained mice in an exposure chamber via an influx polyethylene tube. Nebulization was driven by 10 L/min air supplemented with 5% CO₂. The exposure chamber connects with an identical efflux polyethylene tube with a low resistance microbial filter (BB50T, Pall, East Hills, NY) at its end vented to a biosafety hood [11].

In vivo bacterial infection

As previously described [24], 1 ml of Pseudomonas aeruginosa strain PA103 frozen stock (1×10⁸ colony-forming units [CFU]/ml) was incubated overnight in 100 ml of Tryptic Soy Broth (TSB) at 37°C with 5% CO₂, then expanded in 1 liter of fresh LB media at 37°C to OD 600 of 0.52. Bacterial suspensions were centrifuged, washed, re-suspended in 1× PBS, and aerosolized using the same nebulization system described for Pam2ODN treatment. A nebulized inoculum of 10 ml of 2×10¹⁰ CFU/ml was delivered. The infected mice were closely monitored for at least 8 days. The relevant euthanasia-triggering criteria consist of any evidence of distressed behaviors including hypothermia, impaired mobility, respiratory distress, or inability to access food or water. When mice were identified to meet the criteria, they were subjected to euthanasia. At least 8 mice per group were evaluated for survival analysis. Challenges were performed a minimum of 3 times.

In vitro pathogen killing assay

HBEC3-KT cells or MLE-15 cells were cultured on 6-well plates in complete media until cell growth reached ~80% confluence. Cells were replaced with fresh, antibiotic-free media containing PBS, Pam2, ODN or Pam2-ODN. The final concentrations of Pam2 or ODN in media were 2.4 μM or 0.6 μM, respectively. 4 h after the treatment, 20 μl of P. aeruginosa PA103 (1×10⁵ CFUs/ml) were added to each culture well. 4 h after bacteria inoculation, 20 μl of supernatant from each well was aspirated, serially diluted, plated on a TSB agar plate and incubated for 16 h at 37 °C. Bacterial CFUs were counted after the incubation. Studies were performed a minimum of 3 times with 4 biological replicates per condition.

Metabolic assays

Metabolic analysis was conducted with HBEC3-KT cells. Metabolic assays were performed as previously described [24], including mitoSOX assay, ATP measurement, fatty acid β-oxidation assay, acetyl-CoA measurement, ETC complex activities, JC-1 measurement of Δ_Ψm, Seahorse measurement of OCR and etc.

Nuclear translocation analysis by imaging flow cytometry or immunofluorescence assay

Single cell suspensions of PBS or ODN-treated HBEC3-KT cells were prepared for imaging flow cytometry analysis using the ImagestreamX Mark II imaging flow cytometer (Luminex Corporation). Cells were stained with a 1:100 dilution of NF-kB p65 (F-6) Alexa Fluor 488 conjugated antibody (Santa Cruz Biotechnology) overnight at 4° C and then DAPI for 15 minutes. Stained cells were re-suspended in sterile PBS for image acquisition at 40–60x. 500–1000 events of single-color control files and 5 distinct populations of about 15,000 images per sample of single-focused cells were acquired. Compensated image analysis with IDEAS^® 6.1 software’s wizards included cell populations hierarchically gated for single cells, cells in focus, and double positive for DAPI and NF-kB p65. Active nuclear translocation was measured with a score >1 using the ‘Similarity’ feature. PBS or ODN-treated mTECs on 96-well plate were fixed with 2% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked with 2% goat serum in 1× PBS. Cells were incubated with rabbit against p65 (E379, Abcam) antibody at a dilution of 1:500 overnight at 4°C, then with anti-rabbit Alexa Fluor 594 secondary antibody (Life Technologies, Carlsbad, CA) at a dilution of 1:500, and counterstained with 4′,6-diamidino-2-phenylindole (DAPI) at 1:1000. Cells were visualized using a Keyence BZX-810 microscope using DAPI and Texas Red Filters. Images were corrected for background fluorescence and overlaid using ImageJ before quantification of nuclear translocation of p65. Fields of view were counted individually per group.

Western blotting and immunoprecipitation

Mitochondria lysates were prepared using isolated mitochondria from biotinylated ODN-treated HBEC3-KT cells. To precipitate proteins bound by biotinylated ODN, streptavidin beads (Pierce) were incubated with mitochondria lysates containing 300 μg protein overnight at 4° C under constant gentle rotating. After incubation, streptavidin beads were centrifuged, washed with 1× PBS containing 0.05% Tween-20, resuspended in 50 μl of 2× SDS loading buffer, and boiled for 10 minutes. Elutes from the streptavidin beads were loaded onto SDS PAGE gel (Bio-Rad) and immunoblotted with VDAC1 or ANT1 antibody. Western blotting was performed as previously described [24].

Proteomics analysis

The biotinylated ODN-bound proteins were precipitated with streptavidin beads from mitochondria lysates and resolved by PAGE (Bio-rad). The PAGE gels were stained using a silver staining kit (Pierce). Silver-stained gel pieces were excised, washed, destained and digested in-gel with 200 ng modified trypsin (sequencing grade, Promega) and Rapigest (TM, Waters Corp.) for 18 h at 37° C. In-solution samples were precipitated with 5:1 v/v of cold acetone at −20° C for 18 h, then centrifuged and the acetone was removed prior to treatment with Rapigest (100 °C for 10 min), followed by addition of trypsin. The resulting peptides were extracted and analyzed by high-sensitivity LC-MS/MS on an Orbitrap Fusion mass spectrometer (Thermo Scientific, Waltham MA). Proteins were identified by database searching of the fragment spectra against the SwissProt (EBI) protein database using Mascot (v 2.6, Matrix Science, London, UK) and Proteome Discoverer (v 2.2, Thermo Scientific). Typical search settings were: mass tolerances, 10 ppm precursor, 0.8d fragments; variable modifications, methionine sulfoxide, pyro-glutamate formation; enzyme, trypsin, up to 2 missed cleavages. Peptides were subject to 1% FDR using reverse-database searching.

Structure prediction and molecular docking

3D structures of 14 single strand phosphorothioate DNAs were extracted from the RCSB Protein Data Bank (RCSB PDB). 3D structure of Human VDAC1 was obtained from the RCSB PDB (PDB: 2jk4). The protein-DNA docking was performed using HDOCK server. The resulting docking structures were visualized and analysed using PyMol (Molecular Graphics Systems, version 2.0, Schrodinger, LLC.). Individual and common receptor interface residues between VDAC1 and DNAs were visualized using ggplots package in R studio. 3D structure of ODN M362 were predicted using RNA COMPOSER and PyMOL. 3D structures of competitive peptides were predicted using AlphaFold and PyMOL. Free energy of the competitive peptide binding to ODN M362 was predicted using the PRODIGY-LIGAND server.

Peptide design and synthesis

Competitive peptides with conjugated cell penetrating peptide sequences (TAT sequences: GRKKRRQRRRPQ) at the N terminus [35] were synthesized and purified by GenScript (Piscataway, NJ). All the peptides were purified to >95%. Subsequently, the peptides were lyophilized and shipped. The VDAC1 N-terminal peptide and its correspondent mutated peptide were synthesized by Biosynth (formerly New England Peptide, Gardner, MA) with >95% purity. The purified peptides were lyophilized for shipping. Peptides were dissolved in water and stored in −80°C. Peptide concentrations were determined using bicinchoninic acid protein assay (Pierce).

Electrophoretic mobility shift assay

As previously described [50], 2 μL of phosphorothioate or phosphodiester ODN M362 (25 nM) were mixed with 2 μL of VDAC1 N-terminal peptides (8–45) from 0–32 μM and incubated 30 min on ice. 1 μL of 5 × SDS loading dye in TBE was added to the mix for a total volume of 5 μL. Half of the sample (2.5 μL) was loaded on a 16.5% acrylamide native gel. The gel was pre-warmed at 150 V in 0.5x TBE buffer before samples loading and then run at 90 V for 2 hours. The gel was stained with SYBR Gold in 0.5 × TBE for 10 min. Bio-Rad ChemiDoc Imaging System with blue light was used to detect fluorescence of peptide-bound ODN M362.

Fluorescence polarization assay

Fluorescence polarization assay was performed and the binding data was analysed as previously described [50]. Fluorescence polarization (mP) was measured at 25°C using a BioTek Synergy Neo microplate reader with an excitation wavelength of 485nm and emission wavelength of 528 nm. The difference in fluorescence polarization (Δ mP) between mixed protein/ligand samples and ligand only sample were calculated. A larger Δ mP indicates a greater degree of binding between the protein and the ligand.

Quantification and statistical analysis

Statistical analyses were performed using SigmaPlot 14.0 (Systat Software, San Jose, CA) and GraphPad Prism 8 (GraphPad Software, San Diego, CA). One-way ANOVA was used to compare the means of multiple treatment conditions or multiple time points. The Holm-Sidak method was used, unless normality testing failed, in which case Kruskall-Wallis method was used. Means of two groups were compared using two-way Student’s t-test. Survival comparisons were performed using logrank testing by the Mantel-Cox approach.

Supplementary Material

Tables S1 to S2

Figures S1 to S8

Data and materials availability:

Further data information is available upon reasonable requests and should be directed to and will be fulfilled by the corresponding authors.