Fig 1.
Computational systems pharmacology pipeline identifies GLB as a potential repurposing candidate.
Computational workflow of eight steps (1–8) customized to assess repurposable drugs against ATD using RNAi screening data (step 1). Specifically we trained a discriminative logistic regression classifier to distinguish the modulators of ATZ accumulation (step 2–4). We compared the sequence of these genes to those of the known targets and identified three human orthologues that had significant sequence similarity to the known drug targets (steps 5–7). Of these, we picked the highest confidence gene (step 8) and identified the drug that interacts with the least promiscuous human targets based on available data, leading to GLB.
Fig 2.
Effect of GLB on steady-state levels of ATZ in HTO/Z (A) and HTO/M (B) cell lines.
Immunoblot analysis of cell lines treated with various concentrations of GLB. Carbamazepine at 30 μM was used as a positive control. After 48h incubation of GLB or CBZ, cells were harvested and separated into soluble and insoluble fractions for western blotting with anti-AT. GAPDH was used as a loading control for the soluble fraction and to validate the separation of soluble and insoluble fractions. Gel Code blue staining was used a loading control for the insoluble fraction. The relative densitometric values for ATZ level in experimental versus control conditions are shown at the top to demonstrate the magnitude of the effect.
Fig 3.
Effect of GLB on kinetics of ATZ secretion in HTO/Z cells by pulse-chase radiolabeling methods.
(A) Fluorograms of control (top) and GLB-treated cells (bottom) using the HTO/Z cell line. IC = intracellular fraction; EC = extracellular fraction. After pulse-chase labeling, cell lysates and extracellular fluid were immunoprecipitated with anti-AT. (B) Densitometric analysis of kinetics for IC in the HTO/Z cell line. Kinetics of disappearance from IC was determined by densitometric analysis of fluorograms from five independent experiments. Dashed lines show the half-time for disappearance of ATZ in IC. (C) Densitometric analysis of kinetics for EC. Statistical analysis on kinetics utilized two-way ANOVA with Bonferroni correction (Mean ± SEM, n = 5).
Fig 4.
Mechanism of ATZ degradation targeted by GLB.
(A) Effect of GLB on ATZ levels in the HTO/Z cell line in presence or absence of MG132. Cells were incubated for 48 hours with GLB (40 μM) and MG132 (30 μM) was added for the last 6 hours of this incubation period. Relative densitometric values are shown at the top. (B) Effect of GLB on ATZ levels in HTO/Z cell line and autophagy-deficient HTOZATG14KO (ATG14KO) subclone. Cells were incubated for 48 hours with varying doses of GLB (bottom) and then analyzed for ATZ and β-actin by immunoblotting. Blotting for ATG14 is shown on the right.
Fig 5.
Effect of GLB on ATZ levels in the HTO/Z cell line in the absence or presence of MRP1 or MRP3 siRNA.
HTO/Z cells were transfected with MRP1 siRNA, MRP3 siRNA or negative control siRNA at a final concentration of 30nM. After transfection, the cells were incubated in the presence or absence of GLB for 48 hours. Immunoblotting analysis for ATZ and β-actin is shown in A. Effect on LC3-II/LC3-I/β-actin ratio in control cells is shown in B and in GLB-treated cells in C, using densitometric scanning and arbitrarily setting the value in the absence of MRP siRNA at a value of 1.0.
Fig 6.
Effect of GLB analogs on ATZ levels and insulin secretion in mammalian cell line models.
(A) The structures of GLB and GLB analogs. (B) Western blotting analysis of the effect of G2, G3 and G4 on ATZ levels in HTO/Z cell line. GAPDH was included as a loading control. GLB at 40 μM was used as positive control; DMSO was used as negative control. (C) Insulin levels in Min-6 pancreatic β-cell line. The cells were incubated with GLB, G2, G3 and G4 in the indicated doses. Low and high glucose conditions were used as negative and positive controls (Mean ± SD, n = 3). The results show that insulin secretion is significantly increased by GLB at doses as low as 1 nM (p<0.0001) but not affected by analogs G2 (p = 0.8483), G3 (p = 0.9222) or G4 (p = 0.7886).
Fig 7.
In vivo effect of GLB analog G2 on hepatic ATZ load (A) and hepatic fibrosis (B) in the PiZ mouse model.
The drug was administered by orogastric gavage or subcutaneous osmotic pump predicted to deliver a dose of 10 mg/kg/day over 3 weeks to male PiZ mice at 4–6 months of age. A separate group of age-matched male PiZ littermates were treated with placebo by the same route of administration. Examples of PAS/D and Sirius red staining for livers from 2 untreated (left) and G2 treated mice (right) are shown in (A) and (B), respectively. Results of quantitative morphometric analysis from 4 separate experiments for PAS/D (n = 17, p = 0.0156) are shown on the right in (A). Results of quantitative morphometric analysis from 4 separate experiments for Sirius red (n = 23, p<0.0001) are shown on the right in (B). Results of hydroxyproline content analysis for 1 experiment (n = 7, p = 0.0397) is shown on the far right in (B). Statistical analysis used Unpaired, two-tailed Student t-test for PAS/D staining and hydroxyproline content and unpaired, two-tailed Student t-test with Welch-correction for the Sirius red staining data.