Macrolide antibiotics exert antileukemic effects by modulating the autophagic flux through inhibition of hERG1 potassium channels

Macrolide antibiotics (MAs) have a wide spectrum of activities against Gram-positive bacteria, but they have also been proposed as anticancer drugs for multiple tumor types. 1 On these bases, clinical trials have been started

Macrolide antibiotics (MAs) have a wide spectrum of activities against Gram-positive bacteria, but they have also been proposed as anticancer drugs for multiple tumor types. 1 On these bases, clinical trials have been started, and their results evidenced clinical benefits. 1 MAs have also been tested in hematologic malignancies, alone or in combination with chemotherapeutic drugs or tyrosine kinase inhibitors. 2,3 How MAs exert antineoplastic activity is unclear. Different mechanisms have been proposed, including modulation of autophagy. [4][5][6] We tested the activity of two MAs, clarithromycin (Cla) and erythromycin (Er), on acute leukemia (AL) cells, both myeloid and lymphoid. Both MAs induced leukemia cell death in our cell lines. Cla was efficacious in AML cells, whereas the LD 50 (dose lethal to 50% of animals tested) values of Er, although higher compared with Cla, were generally lower in ALL cells (representative IC 50 (half-maximal inhibitory concentration) values of both MAs are shown in Figure 1a, whereas all the data are reported in Supplementary Table S1).
Next, we tested whether MAs affected the autophagic process of AL cell. Cla-induced vacuoli formation (Supplementary Figure S1 Autophagy is regulated by different signaling pathways: the PI3K/ Akt pathway, which inhibits autophagy by interacting with mTORC1, and the Erk1/2 pathway, which activates autophagy and promotes cell survival. 4,7 Cla determined an early decrease in both Akt and Erk1/2 phosphorylation (Figure 1d and Supplementary Figure S2). The former effect is consistent with autophagy activation, whereas the latter points to inhibition of prosurvival signals and hence induction of apoptosis. Consistently, Cla induced the activation of caspase 3 in FLG 29.1 cells (Figure 1e and Supplementary Figure S2). Overall, MAs stimulate autophagy in AL cells through the inhibition of Akt, but they subsequently block the autophagic flux and induce autophagic cell death. In parallel, MAs also inhibit the Erk-dependent survival signals, thus triggering apoptosis. In other words, MAs activate both autophagic and apoptotic cell death in AL cells.
MAs can block hERG1 currents in reconstitued cellular models. 8,9 Conversely, blocking hERG1 has antileukemic effects both in vitro and in vivo. [10][11][12] Therefore, we hypothesized that the effects of MAs on AL cells could depend on hERG1 inhibition. We first evaluated the effects of MAs on hERG1 currents of AL cells, using Cla and the AML cell line FLG 29.1 as a model (details are given in Supplementary Figure S3). Cla inhibited hERG1 currents in a concentration-dependent manner (IC 50 = 38.5 ± 7.0 μM). It also reduced hERG1B currents (IC 50 = 66.8 ± 9.8 μM), that is, the main hERG1 isoform expressed in leukemia cells. 13,14 Therefore, Cla blocks hERG1B with high efficacy, differently from classical hERG1 blockers like E4031.
Next, we studied the effects of hERG1 blocking (treatment with E4031) and silencing on the autophagic process. For the latter purposes, the level of autophagy in hERG1-silenced FLG 29.1 cells (FLG 29.1-sh7, whose characterization is shown in Supplementary Figure Figure S2), but unchanged levels of MFI relative to the cyto-ID-stained autophagic compartments, compared with FLG pLKO cells (mean 258 ± 7 vs 266 ± 11; Figure 1i). Hence, silencing hERG1 triggers autophagy but, at difference with hERG1 blockade, does not impair the autophagy flux. Consistently, blocking hERG1 with E4031 or silencing its expression both induced a decrease in Akt and Erk 1/2 phosphorylation ( Figure 1j and Supplementary Figure S2), but only E4031 led to activation of caspase 3, which triggers apoptotic cell death ( Figure 1k and Supplementary Figure S2). Overall, these data confirm the hypothesis that hERG1 modulates the autophagic flux in AL cells, and that pharmacologically blocking the channel triggers autophagy, but blocks the autophagic flux, thus inducing both autophagic and apoptotic cell death, in analogy to MAs.
Finally, we tested whether the effects of MAs on AL cell death was related to their blocking effect on hERG1 currents. MAs were tested on hematopoietic cells that do not express the hERG1 channel: normal peripheral blood mononuclear cells (PBMNCs), and EBV infected B lymphocytes (EBV-B). 11 Cla poorly affected both PBMNCs and EBV-B cells, showing LD 50 values higher than 200 μM (Supplementary Figure S5). Cla had significantly (P o0.01) higher LD 50 values in FLG 29.1-sh7 than in FLG 29.1-plkO (Supplementary Figure S5). This is clear from Figure 1l, showing the data obtained with 52 μM Cla in silenced and control cells. Finally, we tested the combination of MAs and E4031 on AL cell death. If added in conjunction with E4031, then both Cla and Er showed a larger effect on cell death (Figure 1m). This is consistent with the fact that MAs block hERG1 currents by binding a site different from the one targeted by E4031. 8 Overall, we conclude that the antileukemic effect of MAs is totally or partially mediated by their blockade of hERG1 currents.  Figure S6).
Either MAs were then tested in combination with chemotherapeutic drugs: mice injected with HL60 AML cells were treated with Cla and Cyt, whereas mice injected with REH ALL cells (notoriously resistant to corticosteroids 11 ) were treated with Er and the corticosteroid dexamethasone (Dexa). Combination treatment of Cla with Cyt significantly increased the overall survival of HL60-luc2 injected mice compared with control group (P = 0.0024; Figure 1g). Combination treatment of Er with Dexa improved overall survival, (Figure 1h), reduced BM engraftment and roughly abolished PB leukemia burden compared with Dexa treatment (inset to Figure 1h).
In this paper, we provide evidence that (1) MAs have antileukemic activity, either in vitro or in vivo, in both AML and ALL, alone or in combination with chemotherapeutic drugs, (2) these effects depend on a complex modulation of both autophagy and intracellular signaling pathways regulating cell survival and apoptosis, and (3) are mediated by hERG1 channels. Compared with hERG1 blockers, MAs have a low risk of inducing torsade-depoints cardiac arrhythmias. 9 We thus propose to include these compounds in treatment schedules of resistant acute leukemias in combination with chemotherapeutic drugs. 15

Figure 2. Effects of combination treatment with MAs and conventional antileukemic agents in ALL and AML cell lines in vitro and in vivo.
A panel of leukemic cell lines were cultured with or without MSC (suspension) and exposed to LD 50 of doxorubicin (Doxo; 0.1 μg/ml) or cytarabine (Cyt; 45 nM) with or without the corresponding LD 50 dose of the MAs antibiotic Cla/Er for 48 h. (a) ALL cell lines (BCP-ALL: 697, REH) exposed to LD 50 of Doxo in the presence of LD 50 of Cla. (b) AML cell lines (FLG 29.1; HL60) exposed to LD 50 of Cyt in the presence of LD 50 of Cla. The percentage of Annexin V+/propidium iodide (PI) − cells was measured. Values are mean ± s.e.m. of three indipendent experiments each performed in triplicate. (c) 697 cells were cultured with or without MSC and exposed to LD 50 of prednisone (5 μM) with or without the LD 50 dose of Er for 48 h. The percentage of Annexin V+/PI − cells was measured. Values are mean ± s.e.m. of two indipendent experiments each performed in triplicate. (d) Three representative pediatric AML primary samples were cultured onto MSCs and treated with LD 50 doses relative to FLG 29.1 of either Cyt (45 nM, see Supplementary Information), MA antibiotic Cla (56 μM) and E4031 (50 μM) for 48 h. The percentage of Annexin V+/PI − cells was measured. Values are mean ± s.e.m. of one experiment performed in triplicate. Statistical analysis was carried out with the Student's t-test (AML-1: Cyt+Cla vs Cyt, Po 0.01; AML-2: Cyt+Cla vs Cyt, P o0.01; AML-3: Cyt+Cla vs Cyt, P o0.01). (e) SCID mice were injected with HL60-luc2 cell line (5 × 10 6 cells intraperitonially (i.p.)) and starting from day 5, animals were treated daily for 14 consecutive days with saline (control, n = 4), Cyt (6.25 mg/kg, i.p., n = 4); Cla (15 mg/kg, by oral gavage, n = 4). Images were acquired with Photo Acquisition software (Biospace Laboratory, Paris, France) and processed with M3 Vision software (Biospace Laboratory). Median values of counts per minutes (c.p.m.) reported for each group of treatment at different time points are shown in the right panel. (f) NOD SCID mice were inoculated with REH cells on day 0 and after one week treated for 2 weeks with saline (Con, n = 4) and Er (Er15, 15 mg/kg, n = 4) and sacrificed 3 weeks after cell injection. Leukemia BM engraftment and PB burden were evaluated by FACS analysis estimating the hCD45+/mCD45+ ratio and were reported as percentage of the control for each treatment group. (g) SCID mice were injected with HL60-luc2 cell line (5 × 10 6 cells i.p.) and starting from day 5, animals were treated daily for 14 consecutive days with saline (control, n = 4), Cyt (6.25 mg/kg, ip, n = 4); Cla (15 mg/ kg, by oral gavage, n = 4) and Cyt (6.25 mg/kg, i.p.) plus Cla (15 mg/kg, by oral gavage, n = 4). Survival curves of each experimental group, estimated by Kaplan and Meier analysis are reported (P = 0.0024). (h) NOD SCID mice were inoculated with REH cells on day 0 and treated for 14 consecutive days with saline (Con, n = 5), Dexa (15 mg/kg, n = 5) and Dexa (15 mg/kg) plus Er (15 mg/kg, n = 5). Survival curves of Control, Dexa and Dexa+Er experimental group, estimated by Kaplan and Meier analysis are reported (P = 0.208): median survival is 22 days in control group, 48.5 days in Dexa group and 108 days in Dexa+Er group. Inset. An additional group of mice (n = 3 per each group of treatment, treated as reported above) were analyzed 3 weeks after cell injection and BM and PB collected. Leukemia BM engraftment and PB burden were evaluated by FACS analysis estimating the hCD45+/mCD45+ ratio and were reported as percentage of the control for each treatment group.
Letter to the Editor