Identification of pharmacological agents that indu...
A screening system for measuring nuclear HMGB1 release
To measure HMGB1 release from the nucleus in an optimal fashion, we took advantage of the so called “retention using selective hooks” (RUSH) system (Fig. 1A)27. In this system, the protein of interest (here HMGB1) is fused to a streptavidin-binding peptide (SBP) as well as a green fluorescent protein (GFP) to facilitate monitoring of its subcellular localization by fluorescence videomicroscopy. Such a construct is stably expressed in cells (here human osteosarcoma U2OS cells) together with streptavidin that is targeted towards a specific subcellular compartment, here to the nucleus by means of three nuclear localization sequences (Str-NLS3) motif (Fig. S1). Driven by the interaction between SBP and streptavidin, the protein of interest together with its GFP tag is retained by the nuclear-targeted streptavidin protein, used as a hook. We observed that Str-NLS3 localized in punctiform structures within the nucleus together with the HMGB1-SBP-GFP fusion protein (Fig. 1B). Biotin has a subnanomolar affinity to streptavidin and can outcompete its binding to SBP, causing rapid release of the HMGB1-SBP-GFP fusion protein from its interaction with streptavidin-NLS3. However, HMGB1-SBP-GFP remained retained in the nucleus, where it was diffusely distributed, similarly to what is observed for endogenous HMGB121. It is worth noting that the localization of the Str-NLS3 was not changed upon addition of biotin (Fig. 1C). The signal distribution changed when an agent that readily releases HMGB1 from the nucleus, namely the anthracycline mitoxantrone (MTX) was added to the cells. When MTX was combined with biotin, HMGB1-SBP-GFP appeared in the cytoplasm. However, MTX alone, without biotin, failed to cause such a redistribution of HMGB1-SBP-GFP (Fig. 1D,E). Fluorescence videomicroscopy allowed for real time observations of this process (Fig. S2, Videos S1 and S2). We reasoned that this kind of experimental design would introduce a marked degree of specificity into the system, meaning that non-specific destruction of the cells or fluorescence quenching would cause a loss of the nuclear HMGB1-SBP-GFP signal (Fig. S3) that would not depend on biotin addition. In contrast, specific inducers of HMGB1 release would cause the loss of the nuclear GFP signal only in the presence of biotin.
An optimized fluorescent biosensor–based cell line for the identification of HMGB1 releasing agents. (A) Scheme of the RUSH (Retention Using Selective Hooks)-HMGB1 cell assay. U2OS cells stably expressing Streptavidin-NLS3 (Str-NLS3) as a hook that localizes in the nuclear dots and can be detected with an anti-streptavidin antibody ((B) red color). In the absence of biotin, HMGB1-SBP-GFP is retained in the nucleus due to the streptavidin-SBP interaction and co-localizes with Str-NLS3. Upon biotin addition, the HMGB1-SBP-GFP reporter is released from the hook and diffuses in the nucleus, while the hook remains punctiform. Granularity of HMGB1-GFP and streptavidin-AF568 was quantified after the cells were exposed to biotin (C). (D) HMGB1 releasing agent mitoxantrone (MTX, 2 µM) induced the exodus of HMGB1 only in the presence of biotin. Cytoplasmic HMGB1 was quantified after the cells were exposed to MTX either in the absence or presence of biotin (E). Data are reported as means ± SEM at 24 h post treatment (n = 4; *P < 0.05, **P < 0.01, and ***P < 0.001, two-tailed Student’s t test, compared to control cells). Scale bar = 10 µm.
Identification of FDA-approved drugs that release HMGB1 from the nucleus
In the next step, we used the aforementioned system to screen the 1200 drugs contained in the Prestwick chemical library for their capacity to induce an increase in cytoplasmic HMGB1-SBP-GFP levels that would exclusively be detectable in the presence of biotin. Among the top-20 agents inducing this phenomenon, we found two epigenetic modifiers (azacitidine and suberoylanilide hydroxamic acid, SAHA), several microtubule inhibitors (docetaxel, paclitaxel and nocodazole) as well as several anthelmintic agent (albendazole, fenbendazole, flubendazole, mebendazole, oxibendazole) that exert their mode of action by binding to tubulin and thereby inhibit microtubule formation (Fig. 2A). We performed an additional screen on the collection of all small molecules that are FDA-approved anticancer agents and confirmed the capacity of epigenetic modifiers (azacitidine, decitabine, SAHA) and inhibitors of microtublular dynamics (docetaxel, paclitaxel) to efficiently induce the cytoplasmic accumulation of HMGB1-SBP-GFP levels (Fig. 2B). As expected, immunogenic cell death inducers such as the cardiac glycoside digoxigenin and the platinum-based antineoplastic oxaliplatin also induced the release of HMGB1-SBP-GFP from the nucleus21,28. Representative images of the biotin-dependent HMGB1-SBP-GFP-releasing activity of SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole are shown in Fig. 3. The dose and time dependency of such effects are documented in Fig. 4.
Identification of HMGB1 releasing agents from chemical libraries. (A) U2OS cells stably co-expressing streptavidin-NLS3 and HMGB1-SBP-GFP were seeded in 384-well plates either in the absence or in the presence of biotin before treatment with 1200 small molecules from the Prestwick chemical library (most of which are approved by FDA, EMA and other agencies) at a final concentration of 10 µM for 48 h. Following HMGB1-GFP fluorescence was quantified in the nucleus as well as the cytoplasm and, based on Hoechst 33342 staining, nuclear pyknosis was assessed as an indicator for cell death. Quantitative data were normalized by z-scoring (mean, n = 4) hierarchically clustered and depicted as heat map, with red and blue values indicating positive and negative effects, respectively. The top 20 compounds are shown. (B) U2OS cells stably co-expressing streptavidin-NLS3 and HMGB1-SBP-GFP were pre-incubated or not with biotin before treatment with a collection of FDA-approved anticancer agents at different concentrations for 48 h. Following cytoplasmic HMGB1-SBP-GFP fluorescence was quantified. Obtained results were hierarchically clustered (average linkage*, similarity metric using pearson distance*) and depicted as a heat map. Agents from the custom library have been used at the following concentrations azacitidine (1, 5, 10, 20 µM); paclitaxel (0.1, 0.5, 1, 2 µM); docetaxel, (0.1, 0.5, n 1, 2 µM); oxaliplatin (10, 50, 100, 200 µM); decitabine (1, 5, 10, 20 µM) and suberoylanilide hydroxamic acid (SAHA; 1, 5, 10, 20 µM).
Nucleo-cytoplamic translocation of HMGB1-SBP-GFP. U2OS cells stably co-expressing Streptavidin-NLS3, HMGB1-SBP-GFP were pre-incubated or not with biotin before treatment with suberoylanilide hydroxamic acid (SAHA), fenbendazole (FENB) and oxibendazole (OXB), azacitidine (AZA), decitabine (DECI) all at 10 µM, and oxaliplatin (OXA) at 100 µM for 48 h. Hoechst 33342 and CellTracker Orange CMTMR dye were used to visualize the nucleus and cytoplasmic region, respectively. The chemical staining allowed for the segmentation of nuclear and cytoplasmic areas by image analysis for subcellular HMGB1-GFP intensity using the MetaXpress software. Scale bar = 10 µm.
SAHA, Azacitidine, Decitabine, Oxaliplatin, Fenbendazole and Oxibendazole induce time/dose-dependent HMGB1 release. U2OS-SBP-HMGB1, Streptavidin-NLS3 co-expressing cells were pre-incubated or not with biotin before treatment with suberoylanilide hydroxamic acid (SAHA), fenbendazole (FENB) and oxibendazole (OXB) at 5, 10 and 20 µM, azacitidine (AZA) at 7.5, 15 and 30 µM, decitabine (DECI) at 10, 20 and 40 µM; and oxaliplatin (OXA) at 50, 100 and 200 µM for the indicated time points. Cells were then stained with Hoechst 33342 and CellTracker Orange CMTMR before fixation with 4% paraformaldehyde. Cytoplasmic GFP intensity was quantified and normalized to untreated controls. Data are shown as means ± SEM (n = 4; *P < 0.05, **P < 0.01, and ***P < 0.001, two-tailed Student’s t test, compared to control cells).
In vitro and in vivo validation of HMGB1 releasing effects
In the next step, we determined whether the aforementioned compounds may also release a GFP-HMGB1 fusion protein (without SBP), as well as the endogenous HMGB1 protein (without GFP). For this, U2OS cells that were stably transduced with GFP-HMGB1 were exposed to SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole, and the increase in the cytoplasmic GFP-HMGB1 signal was confirmed (Fig. 5A–C). Alternatively, parental U2OS cells (devoid of GFP) were cultured with SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole, then fixed, permeabilized and subjected to immunofluorescence staining of HMGB1, again confirming its nucleo-cytoplasmic translocation (Fig. 5D–F). This result was further corroborated using subcellular fractionation to separate the nuclei from the cytoplasm, followed by immunoblot detection of HMGB1 (Fig. 6A–F). Moreover, an HMGB1-specific enzyme-linked immunosorbent assay (ELISA) revealed a significant, dose-dependent release of HMGB1 into the supernatant of the cells cultured in the presence of SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole or oxibendazole. This result was obtained with U2OS cells (Fig. 6G), as well as with mouse methylcholantrene-induced fibrosarcoma MCA205 cells (Fig. 6H). Moreover, intraperitoneal injection of SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole into mice caused a time-dependent increase in the plasma concentration of ELISA-detectable HMGB1 (Fig. 7). Altogether, these results confirm that SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole act as potent inducers of HMGB1 release.
SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole induce HMGB1 release in regular biosensor cells and wild-type cells. U2OS cells stably expressing a GFP-HMGB1 fusion (A–C) or wild-type U2OS cells (D–E) were maintained in control conditions (Ctrl); mitoxantrone (MTX) at 2 µM; suberoylanilide hydroxamic acid (SAHA), fenbendazole (FENB) and oxibendazole (OXB) at 5 (l), 10 (h) µM; azacitidine (AZA) at 15 (l), 30 (h) µM; decitabine (DECI) at 20 (l), 40 (h) µM; and oxaliplatin (OXA) at 50 (l), 100 (h) µM for 24 h or 48 h, followed by assessment of cytoplasmic GFP-HMGB1 fluorescence intensity (A,B) or endogenous HMGB1 level after immunostaining with an anti-HMGB1 antibody and Alexa-fluor 488 conjugated 2nd antibody and quantification of relative enrichment in the cytosol (D,E). Representative images of biosensor cell line (C) and immunofluorescence (F) are reported (scale bar = 10 µm). Data are shown as means ± SEM (n = 4; *P < 0.05, **P < 0.01, and ***P < 0.001, two-tailed Student’s t test, compared to control cells).
Biochemical detection of HMGB1 release. Human osteosarcoma U2OS cells or murine fibrosarcoma MCA205 cells were maintained in control conditions (Ctrl), suberoylanilide hydroxamic acid (SAHA), azacitidine (AZA), decitabine (DECI), oxaliplatin (OXA), fenbendazole (FENB) and oxibendazole (OXB), at indicated concentrations for 48 h, followed by subcellular fractionation and detection of nuclear and cytoplasmic HMGB1 expression by western blotting. Beta-actin and Histone H3 were used as loading controls of cytoplasmic and nuclear proteins respectively. Representative immunoblots (A,B) and densitometry data (C–F) are depicted. Densitometry data are represented as mean ± SEM of three independent experiments. (G,H) Cells were treated with staurosporine (STS) at 1 µM; mitoxantrone (MTX) at 3 µM; suberoylanilide hydroxamic acid (SAHA), fenbendazole (FENB) and oxibendazole (OXB) at 5, 10, 20 µM, azacitidine (AZA) at 7.5, 15, 30 µM, decitabine (DECI) at 10, 20, 40 µM; and oxaliplatin (OXA) at 50, 100, 200 µM for 48 h, followed by the assessment of HMGB1 (EctoHMGB1) release into cell culture supernatants by means of an HMGB1-specific ELISA. Data are reported as means ± SEM (n = 3; ***P < 0.001, two-tailed Student’s t test, compared to Ctrl cells) Additional data on MTX and STS-induced HMGB1 release can be found in Fig. S6).
HMGB1 increases in mice plasma post IP injection of HMGB1 releasing agents. Female C57BL/6 mice (4 mice/group) were intraperitoneally administrated with suberoylanilide hydroxamic acid (SAHA) (100 mg/Kg), azacitidine (AZA) (50 mg/Kg), decitabine (DECI) (100 mg/Kg), oxaliplatin (OXA) (10 mg/Kg), fenbendazole (FENB) (200 mg/Kg) and oxibendazole (OXIB) (200 mg/Kg). Blood samples were collected at the indicated time points post injection (p.i.) and plasma was prepared as described in material and methods. Plasma HMGB1 levels were measured by means of a HMGB1-specific ELISA kit according to the manufacture protocol. Boxplots report the lower and upper quartile plus the median value. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (two-way ANOVA analysis).
Mode of action of HMGB1 release
We investigated possible communalities in the mode of action of SAHA, azacitidine, decitabine, oxaliplatin, fenbendazole and oxibendazole. Previously, histone deacetylase inhibitors such as trichostatin A (TSA) have been shown to stimulate the hyperacetylation of HMGB1 as well as histones which facilitates the release of HMGB1 from chromatin2,5,29. Nonetheless, only SAHA was able to cause an increase in histone acetylation, while none of the other agents did so (Fig. 8). One physiological cellular state in which HMGB1 is released from the nucleus is mitosis where HMGB1-SBP-GFP can be found in a cytoplasmic location, provided that biotin has been added to the cells to unlink reporter and hook (Fig. S4, Videos S3 and S4). However, only the histone deacetylase (HDAC) inhibitors trichostatin A (TSA, as a positive control), SAHA and anthelmintic agents fenbendazole and oxibendazole caused a major blockade of the cell cycle in mitosis, contrasting with azacitidine, decitabine as well as oxaliplatin, which arrested the cells in the S-phase (Fig. S5). Hence, the drugs characterized here do not have a common cell cycle-blocking activity that would explain their effects on HMGB1 nuclear release. We also investigated the possibility that azacitidine and decitabine would act by inhibiting their known pharmacological targets (which are the DNA methyl transferases 1, 3a and 3b; DNMT1, DNMT3a, DNMT3b)30,31 or rather through off-target effects. The knockdown of DNMT3a or DNMT3b, was sufficient to induce the nucleo-cytoplasmic translocation of GFP-HMGB1, and this effect was not further enhanced by azacitidine and decitabine (Fig. 9). These finding plead in favor of an on-target effect of azacitidine and decitabine.
Effects of SAHA, Azacitidine, Decitabine, Oxaliplatin, Fenbendazole and Oxibendazole on the acetylation of histones and HMGB1. U2OS cells were maintained in control conditions (Ctrl), or were treated with suberoylanilide hydroxamic acid (SAHA), azacitidine (AZA), decitabine (DECI), oxaliplatin (OXA), fenbendazole (FENB) and oxibendazole (OXIB) at the indicated concentrations for 6 (A) or 12 (B) h, followed by total protein extraction and detection of acetylated/deacetylated histones and HMGB1 by western blotting.
Effect of DNMT knockdown on azacitidine- and decitabine-induced HMGB1 release. U2OS cells expressing GFP-HMGB1 were transfected with small interfering RNAs (siRNA) targeting DNA methyl transferases 1 (DNMT1), 3a (DNMT3a) and 3b (DNMT3b). Forty-eight hours post transfection, the cells were either cultured in control conditions or exposed to azacitidine (AZA; 15, 30 µM) or decitabine (DEZI; 15, 30 µM) for additional 24 or 48 hours followed by assessment of cytoplasmic GFP-HMGB1 fluorescence (A,B). Cytoplasmic GFP intensity was quantified and normalized to untreated controls of respective siRNAs transfection. Data are shown as means ± SEM (n = 4; *P < 0.05, **P < 0.01, and ***P < 0.001, two-tailed Student’s t test, compared to control wells). Knockdown efficacy was validated by western blot (C). SiRNAs targeting DNMT1, DNMT3a and DNMT3b were utilized in U2OS GFP-HMGB1 expressing cells following the same protocol as described above. Cells were treated with AZA (15, 30 µM) and DECI (15, 30 µM) for 24 or 48 h and normalized cytoplasmic GFP intensity was reported as heat map (D).
Concluding remarks
In the present paper, we developed a robust assay for identifying pharmacological agents that cause the nuclear release of HMGB1. Supporting the specificity of the screening system that we developed, we identified several drugs that fell into two categories (epigenetic modifiers, microtubule inhibitors) as bona fide inducers of HMGB1 translocation. Indeed, these agents as well as some classical cytotoxicants (mitoxantrone, oxaliplatin) were highly efficient in stimulating HMGB1 release, both in vitro, on cultured human or mouse cells, as well as in vivo, in mice. Multiple distinct technologies to detect HMGB1 release yielded coherent results, as documented for the RUSH assay, the subcellular localization of an GFP-HMGB1 fusion protein (independent from RUSH), the immunofluorescence detection of endogenous HMGB1 protein, immunoblot determinations after subcellular fractionation, as well as ELISA-based detection of extracellular HMGB1. Importantly, the mode of action of these agents with respect to HMGB1 release are probably heterogeneous, because they were not linked to a common mechanism of protein hyperacetylation or similar cell cycle effects. Interestingly, DNA hypomethylating agents induced nuclear HMGB1 exodus via on-target effects, suggesting that the methylation status of DNA determines chromatin interaction with HMGB1. However, further studies are necessary to understand such an effect, given that so far HMGB1 binding to chromatin only has been related to histone H3K9 dimethylation, not to DNA methylation32. Moreover, an increase (not a decrease) of the methylation of HMGB1 itself has been linked to its release33.
In the present study microtubule targeting