Chemical Proteomics Reveals Human Off-Targets of Fluoroquinolone Induced Mitochondrial Toxicity

Abstract

Fluoroquinolones (FQs) are an important class of potent broad-spectrum antibiotics. However, their general use is more and more limited by adverse side effects. While general mechanisms for the fluoroquinolone-associated disability (FQAD) have been identified, the underlying molecular targets of toxicity remain elusive. In this study, focusing on the most commonly prescribed FQs Ciprofloxacin and Levofloxacin, whole proteome analyses revealed prominent mitochondrial dysfunction in human cells, specifically of the complexes I and IV of the electron transport chain (ETC). Furthermore, global untargeted chemo-proteomic methodologies such as photo-affinity profiling with FQ-derived probes, as well as derivatization-free thermal proteome profiling, were applied to elucidate human protein off-targets of FQs in living cells. Accordingly, the interactions of FQs with mitochondrial AIFM1 and IDH2 have been identified and biochemically validated for their contribution to mitochondrial dysfunction. Of note, the FQ induced ETC dysfunction via AIFM1 activates the reverse carboxylation pathway of IDH2 for rescue, however, its simultaneous inhibition further enhances mitochondrial toxicity. This off-target discovery study provides unique insights into FQ toxicity enabling the utilization of identified molecular principles for the design of a safer FQ generation.

Keywords: Antibiotics; Chemical proteomics; Fluoroquinolone; Human off-targets; Mitochondrial toxicity.

Figure 1

Whole proteome analysis of Ciprofloxacin‐treated human cells. (A) Structures of the FQs Ciprofloxacin and Levofloxacin. (B) Time‐dependent whole proteome analysis of Ciprofloxacin‐treated HEK‐293 cells (75 μM) versus control. Threshold lines represent a log₂ regulation of ±1 and – log₁₀ (P‐value) of 1.3 (two‐sided two‐sample t‐test, n=3 replicates per group). Subunits of the electron transport chain (ETC) complexes I and IV are depicted in blue and green, respectively. A pronounced and time‐dependent down‐regulation of both complexes was detected. Similar effects were observed for the reported plasma concentration (7.5 μM), however, shifted to later onset (see Figure S2). (C) Volcano plot depicting the proteome regulation of periodontal ligament cell line (PDL‐hTERT) treated with Ciprofloxacin (75 μM) for 3 days versus vehicle control‐treated cells featuring significant down‐regulation of complex I (blue) and complex IV (green) subunits of the ETC. Threshold lines represent a log₂ regulation of ±1 and – log₁₀ (P‐value) of 1.3 (two‐sided two‐sample t‐test, n=4 replicates per group). (D) Alphabetically sorted table of top up‐regulated proteins (threshold: log₂ regulation≥1.5). (E) Alphabetically sorted table of top down‐regulated proteins (threshold: log₂ regulation≤−1.5). (F) Top 5 most strongly regulated pathways (detailed pathway analysis in Figure S6).

Figure 2

Fluoroquinolone off‐target discovery via affinity‐based protein profiling (AfBPP). (A) Schematic workflow of MS‐based AfBPP. Cells are incubated with affinity probe or vehicle control and subsequently irradiated with UV light to covalently link potential target proteins. After lysis, biotin is appended to the alkyne handle via CuAAC chemistry and engaged proteins enriched on avidin beads. Following tryptic digest, the resulting peptides are measured via LC–MS/MS and raw data analyzed to infer potential protein targets of the probe. (B) Set of FQ‐derived affinity probes based on Ciprofloxacin and Levofloxacin. (C) Representative AfBPP volcano plot of Cipro P1 (2.5 μM, 1 h, 37 °C) in PDL‐hTERT cells versus the DMSO control. Threshold lines indicate log₂ enrichment≥2 and the statistical significance –log10(P value)≥1.3 (two‐sided two‐sample t‐test, n=4 replicates per group). Literature‐known unspecific photome hits are depicted in red. Proteins of interest are highlighted in blue. (D) Corresponding table of potential off‐target proteins of Cipro P1 in PDL cells. (E) Venn‐diagram illustrating the overlapping enriched proteins of Cipro P1 in HEK293, A549 and PDL‐hTERT cell lines. (F) Venn‐diagram depicting the overlapping protein hits in PDL‐hTERT cells across the four FQ‐derived probes. Within the P1 and P2 probe sets putative protein (off−)targets are strongly conserved (8 shared hits with P1 probes, 9 shared hits with P2 probes). The overlap between Levo P1 and both P2 probes is 5 proteins. No protein is adhering to the standard thresholds across all probes.

Figure 3

Thermal protein profiling (TPP) of HEK293 cells with Ciprofloxacin. (A) General TPP workflow. Cells were treated either with vehicle control or Ciprofloxacin (70 μM) in duplicates for 1 h and cell aliquots were subjected to a temperature gradient with 10 increments. After lysis and ultra‐centrifugation to remove aggregated proteins, the soluble fractions were digested by trypsin and the resulting peptides isotopically labelled with TMT 10‐plex with one distinct channel per temperature. After combination of all 10 TMT channels per sample and offline HILIC fractionation, the samples were measured by LC–MS/MS and the raw data analyzed to infer protein melting curves. (B) Scatter plot of the melting point differences of both Ciprofloxacin‐treated replicates in relation to their vehicle‐treated replicates. Stabilized proteins adhering to all filters are labelled in black. Mito‐ribosomal subunits are depicted in blue and were found to be strongly destabilized. (C) Table of stabilized proteins adhering to all filters. (D) Inferred melting curve of the most strongly stabilized protein ADAL. (E) Inferred melting curve of the protein NUDT1 that was found stabilized in the TPP experiment and also has been enriched in the AfBPP experiments.

Figure 4

Target validation studies for NUDT1 and ADAL1. (A) Activity of NUDT1 purine nucleoside triphosphate hydrolase converting reactive oxygen species (ROS)‐damaged oxidized purine nucleotides to their corresponding monophosphates as exemplary depicted for 8‐oxo‐dGTP. Also N⁶‐methyl‐(d)ATP nucleotides are substrates of NUDT1 resulting in N⁶‐methyl‐(d)AMP, that is substrate of the adenosine deaminase‐like protein ADAL1 converting the methylated nucleoside monophosphate to (d)IMP. (B) NUDT1 in vitro activity assay in presence of Ciprofloxacin. The enzymatic activity is partially, but significantly and concentration‐dependently inhibited by physiologically relevant concentrations of Ciprofloxacin. The mean and SD of 3 replicates per condition is plotted. The graph is representative of three independent experiments. (C) Gel‐based AfBPP of recombinant NUDT1 spiked into a human cell lysate background by Cipro P2 and Levo P2. The labelling is abolished in the heat denatured NUDT1 control (ΔT) and partially outcompeted by the parent FQ further validating NUDT1 engagement by the two FQs. The corresponding Coomassie‐stain is depicted as loading control. The complete gel is depicted in Figure S20C. (D) In vitro activity assay of ADAL1 in presence of Ciprofloxacin showing significant inhibition only at concentrations >300 μM, which is likely not relevant in physiological conditions. The mean and SD of 3 replicates per condition is plotted. The graph is representative of two independent experiments. In the bar plots the statistical relevance based on one‐way ANOVA with Dunnett's multiple comparison test is depicted (ns meaning P‐value >0.05, * P‐value≤0.05, ** P‐value≤0.01, *** P‐value≤0.001, **** P‐value≤0.0001).

Figure 5

Effect of FQs on AIFM1. (A) Schematic overview of AIFM1 and MIA40 interaction. (1) NADH‐dependent dimerization of AIFM1 enables binding and import of MIA40 forming a stable active trimer. MIA40 mediates the import of specific nuclear‐encoded proteins into the intermembrane space (IMS) as key component of the mitochondrial disulfide relay system and further facilitates the oxidative folding of imported substrates. Various MIA40‐AIFM1 substrates are structural components of ETC complex I and IV or are involved in their biogenesis as assembly factors. (2) AIFM1 knockout or perturbation of the AIFM1‐MIA40 interaction results in reduced MIA40 and consequently substrate protein concentrations in the IMS. Figure based on Salscheider et al. AIFM1 knockout cells were generated using CRISPR/Cas9, and individual clones were verified by sequencing, immunoblotting and proteomics. (B) Volcano plot depicting the proteome regulation of AIFM1 knockout HEK293 cells treated with Ciprofloxacin (75 μM) for 3 days versus vehicle control‐treated AIFM1 knockout cells featuring only minor additional alterations of ETC subunits. Threshold lines represent a log₂ regulation of ±1 and – log₁₀ (P‐value) of 1.3 (two‐sided two‐sample t‐test, n=4 replicates per group). The volcano plot of the wild‐type (WT) control experiment is depicted in Figure S23A. (C) Boxplot showing the comparison of respiratory chain complex protein levels in HEK293 WT or AIFM1 KO cells treated with ciprofloxacin or left untreated. Changes of protein levels of all significantly affected subunits of the respective respiratory chain complexes upon ciprofloxacin treatment were plotted. In WT cells, levels of complex I and IV subunits decrease upon ciprofloxacin treatment. This effect is attenuated in AIFM1 KO cells indicating that ciprofloxacin acts on AIFM1.

Figure 6

Effect of FQs on the mitochondrial isocitrate dehydrogenase IDH2. (A) IDH2 catalyzes the conversion of isocitrate and α‐ketoglutarate (αKG) bidirectionally and is therefore important for metabolic rewiring in hypoxic conditions and in case of mitochondrial dysfunction when reductive glutaminolysis (red pathway) is the prominent carbon source for citrate formation. (B) IDH2 in vitro activity assay in presence of Ciprofloxacin. Partial, but concentration‐dependent inhibition in a physiologically relevant range was detected. The mean and SD of 4 replicates per condition is plotted. The graph is representative of three independent experiments. (C) Mitochondrial NADPH levels are decreased by approx. 25 % after incubation of PDL‐hTERT cells with Ciprofloxacin (7.5 μM, 6 h) in line with the in vitro inhibition of the forward oxidative reaction of IDH2. To prevent the onset of ETC dysregulation in line with our proteomic data (compare Figure 1, S1,2,4) the lower concentration was deliberately chosen for this time point (average and mean of 3 independent experiments plotted, unpaired two‐tailed t‐test, P‐value <0.001). (D) Prolonged incubation, especially with the higher Ciprofloxacin concentration (75 μM, 3 days) resulted in elevated mitochondrial NADPH levels in PDL‐hTERT cells, potentially as a result of the inhibition of the reverse reductive carboxylation reaction of IDH2 that is reported to be prominent in settings of mitochondrial dysfunction (average and mean of 3 independent experiments plotted). In the bar plots in B, C and D the statistical relevance based on one‐way ANOVA with Dunnett's multiple comparison test is depicted (ns meaning P‐value >0.05, * P‐value≤0.05, ** P‐value≤0.01, *** P‐value≤0.001, **** P‐value≤0.0001).

Figure 7

Schematic overview of identified Ciprofloxacin adverse effects regarding mitochondrial toxicity. AIFM1 is involved in the MIA40‐mediated IMS import and oxidative folding of specific nuclear encoded proteins including subunits of the ETC complex I and IV (compare Figure 5A). Ciprofloxacin impairs the IMS import machinery leading to ETC impairment. Dysfunction of the ETC, especially of complex I results in a decrease of the NAD⁺/NADH ratio. The change in redox‐homeostasis and shortage of NAD⁺, a key cofactor for the TCA cycle, can lead to the metabolic rewiring involving IDH2 as discussed in Figure 6A. Importantly, IDH2 function was also shown to be impaired by Ciprofloxacin, therefore also this compensation mechanism is impaired by FQs.