Zfp580 inactivation as a new therapeutic target to enhance recovery after stroke in mice

Abstract

Paracrine cerebral Interleukin 6 (Il6) is relevant for stroke recovery, but systemic Il6 elevation may worsen outcome. Hence, paracrine Il6 response modulation within the neurovascular unit has emerged as an attractive therapeutic approach. Lithium modulates Il6 responses and improves stroke outcome. However, lithium may cause serious adverse effects. Here, we report that Zincfinger protein 580 (Zfp580) mediates the effects of lithium on Il6 signaling. In contrast to lithium, Zfp580 inactivation had no neurotoxic effects, and Zfp580 knock out mice showed no phenotypic changes in cognitive and motor function behavioral tests. We discovered that lithium and hypoxia disinhibited Il6 via Zfp580 suppression and post-translational modification by small ubiquitin-like modifier (SUMO). After transient middle cerebral artery occlusion, loss of Zfp580 reduced paracrine Il6 and increased Il6 trans-signaling. Aside from modulating Il6 signaling, Zfp580 loss improved endothelial resilience to ischemia, was highly neuroprotective resulting in smaller infarcts and enhanced use-dependent neuroplasticity, all of which led to improved functional outcome. In conclusion, inactivation of Zfp580 exerts positive effects on multiple key mechanisms without exhibiting relevant adverse side effects, making it potentially a more specific and effective treatment target for stroke recovery than lithium. To fully assess its potential, Zfp580 inhibitors must be developed.

Keywords: Il6 signaling; Zfp580; stroke; stroke recovery; use-dependent plasticity.

Figure 1.

The induction of Il6 by lithium or Gsk3 inhibition is mediated by Zfp580, which is an inhibitor of Il6. (a) Expression levels of Il6 in microvascular cerebral endothelial cells (bEnd.3) or neurons determined by real-time RT-PCR. Presented is the Il6 CT-value derived from 1 µg total RNA. Neurons express Zfp580 at a higher level than endothelial cells. (two-tailed unpaired Student's t-test: t(10) = 9.733, p < .001). (b) Immunofluorescence staining of Zfp580 in bEnd.3 and neurons. Zfp580 is mostly found in the nucleus. Endothelial cells with increased expression ((c) and (d)) or knockdown of Zfp580 due to the use of two different miRNA-embedded shRNAs ((e) and (f)) were generated by lentiviral transduction. The mRNA expression of Il6 was determined by real-time RT-PCR. Il6 protein levels in the cell culture supernatant were determined by ELISA. Increased expression of Zfp580 led to suppression of Il6 mRNA (c) (Mann-Whitney test: p = .002) and secretion of the Il6 protein (d) (two-tailed unpaired Student's t-test: t(10) = 4.816, p < .001). Knockdown of Zfp580 induced Il6 mRNA (e) (One-way ANOVA with Sidak's multiple comparison test: F(2,15) = 140.9, p < .001) and Il6 secretion (f) (One-way ANOVA with Tukey's multiple comparison test: F(2,12) = 94.02, p < .001). We analyzed the time and dose dependent effect of lithium chloride (LiCl) on Zfp580 and Il6 expression levels compared to cells exposed to sodium chloride (NaCl) using a two-way RM ANOVA with Sidak's multiple comparison test and Geisser-Greenhouse correction. In endothelial cells, LiCl suppressed Zfp580 mRNA expression in a time-dependent manner (g) depending on time point, LiCl treatment and the interaction between both factors (F(2.636, 26.36) = 24.63, p < .001; F(1, 10) = 21.88, p < 0.001; F(3, 30) = 6.924, p = .001), and dose-dependent manner (h) depending on the interaction between dosage and LiCl treatment (F(2, 20) = 7.881, p = .003). LiCl increased Il6 mRNA expression in a time-dependent manner (i) depending on time point, LiCl treatment and the interaction between both factors (F(1.118, 11.18) = 43.16, p < .001; F(1, 10) = 2098, p < 0.001; F(3, 30) = 43.54, p < .001), and dose-dependent manner (j) depending on dosage, LiCl treatment and interaction between both factors (F(1.218, 12.18) = 77.65, p < .001; F(1, 10) = 109.1, Continued.p < 0.001; F(2,20) = 91.59, p < .001). To determine whether the effect of LiCl on Zfp580 and Il6 expression is dependent on Gsk3, we inactivated Gsk3 prior to LiCl treatment by incubating the cells with the specific Gsk3 inhibitor CHIR99021. CHIR99021 had no effect on Zfp580 expression (k), but induced Il6 more than lithium treatment (p < .001) (l). LiCl showed no additional Il6 inductive effect after CHIR99021 treatment. Instead, CHIR99021's Il6 inductive effect was significantly diminished when combined with lithium. (p < .001) (One-way ANOVA with Tukey's multiple comparison test: F(3, 20) = 115.7, p < .001). To determine whether Zfp580 mediates the effects of LiCl and CHIR99021 on Il6 expression, we knocked down Zfp580 prior to lithium (m) or CHIR99021 (n) treatment. Presented is the fold increase in Il6 mRNA compared to the respective NaCl or DMSO control. Compared to cells expressing LacZ-shRNA, knock down of Zfp580 strongly reduced the inductive effect of LiCl (two-tailed unpaired Student's t-test: t(10) = 7.380, p < .001) and CHIR99021 (Mann-Whitney test: p = .002). In neurons as second principal cells of the neurovascular unit, we observed similar effects: LiCl suppressed Zfp580 mRNA expression in a dose-dependent manner (o) depending on dosage, LiCl treatment and the interaction between both factors (F(1.810, 18.10) = 6.287, p = .002; F(1, 10) = 41.97, p < 0.001; F(2, 20) = 6.287, p = .008). LiCl increased Il6 mRNA expression in a dose-dependent manner (p) depending on dosage, LiCl treatment and interaction between both factors (F(1.051, 10.51) = 11.90, p = .005; F(1, 10) = 15.23, p = 0.003; F(2,20) = 11.70, p < .001). Incubating neurons with 10 µM CHIR99021 for 24 h had no effect on Zfp580 expression (q), but strongly induced Il6 and (r) (two-tailed unpaired Student's t-test: t(10) = 7.004, p < .001). Graphs show scatter dot plots of independent experiments and means ± standard deviation except (c) showing median ± interquartile range.

Figure 2.

Lithium inactivates Zfp580 by inducing its degradation, SUMOylation and nuclear exclusion. (a) In bEnd.3 cells, we identified additional Zfp580 bands in the immunoblot with molecular weights between approximately 18 and 130 kD which were reduced by Zfp580-RNAi. (b) Tandem affinity purification using FPLC of equimolar SUMO1-3 from a single vector. Each SUMO variant was fused to a His-tag and an S-tag. First, purification was based on the His-tag, after which a second purification step for the S-tag was Continued.carried out, and subsequent immunoblotting for Zfp580 was performed. Six bands corresponding to molecular weights of 24 kD, 35 kD, 50 kD, 70 kD, 100 kD and 130 kD were detectable, indicating the covalent SUMOylation of Zfp580. (c) Zfp580 immunoblot of cells treated with 0.2 mM, 0.7 mM or 20 mM NaCl or LiCl for 24 h. Both, NaCl and LiCl suppressed Zfp580 protein level, but LiCl to a higher extend. (d) Zfp580 immunoblot of cells treated with 10 µM CHIR99021 or DMSO for 24 h. CHIR99021 suppressed Zfp580 protein levels. (e) To determine whether the effect of LiCl on Il6 depends on the SUMOylation of Zfp580, SUMOylation was inhibited 4 h prior to LiCl treatment using 500 mM TAK-981, and Il6 expression was measured using real-time RT-PCR. TAK-981 diminished the inductive effect of LiCL on Il6 but did not eliminate it (p < .001) (Brown-Fortsythe ANOVA with Tukey`s multiple comparison test: F*(3, 11.32) = 286.3, p < .001). (f) Fractionation of cytosolic and nuclear proteins from wild-type bEnd.3 cells and Zfp580K31R-expressing bEnd.3 cells. The 130-kD and 55-kD variants of Zfp580 are detectable in the cytosol, and the 35-kD, 24-kD and unmodified 18-kD variants were detectable in the nucleus. Mutation of the predicted SUMOylation motif at K31 led to reductions in the higher molecular-weight variants. (g) Immunofluorescence staining for Zfp580 in bEnd.3 cells treated with 0.2 mM, 0.7 mM or 20 mM NaCl or LiCl for 24 h. NaCl treatment showed no effect whereas Zfp580 was dose-dependently excluded from the nucleus after LiCl treatment (One-way ANOVA with Tukey`s multiple comparison test: F(5, 54) = 40.54, p < .001). (h) Immunofluorescence staining for Zfp580 in increased wtZfp580 expressing cells treated with 0.2 mM, 0.7 mM or 20 mM NaCl or LiCl for 24 h. Exogenous wtZfp580 expression resulted in increased nuclear Zfp580; however, NaCl treatment had no effect, whereas LiCl treatment dose-dependently excluded Zfp580 from the nucleus (One-way ANOVA with Tukey`s multiple comparison test: F(5, 54) = 30.14, p < .001). (i) Zfp580 immunoblot of native bEnd.3, wtZfp580 expressing cells or Zfp580K31R expressing cells treated either with 20 mM NaCl or 20 mM LiCl for 24 h. LiCl induced the nuclear mono-SUMOylated 24-kD band and the cytosolic >70-kD variants in native bEnd.3 and wtZfp580 expressing cells. Zfp580 is drastically reduced in Zfp580K31R cells, and there is no discernible difference between cells treated with NaCl or LiCl. Notably, the 24-kD nuclear mono-SUMOylated band is totally absent. (j) - (m) Comparision of Il6 expression (real-time RT-PCR) and secretion (ELISA) in native bEnd.3, wtZfp580 expressing, and Zfp580K31R expressing cells following LiCl or CHIR99021 treatment using a two-way ANOVA with Tukey`s multiple comparison test. Il6 disinhibition is increased in wtZfp580 cells and is most pronounced in Zfp580K31R cells. (j) LiCl disinhibited Il6 mRNA expression depending on the Zfp580 variant, LiCl treatment and the interaction between both factors (F(2, 30) = 32.8, p < .001; F(1, 30) = 199.9, p < 0.001; F(2, 30) = 22.72, p < .001) and (k) secreted Il6 protein depending on the Zfp580 variant, LiCl treatment and the interaction between both factors (F(2, 30) = 203.2, p < .001; F(1, 30) = 1055, p < 0.001; F(2, 30) = 158.6, p < .001). (l) CHIR99021 disinhibited Il6 mRNA expression depending on the Zfp580 variant, CHIR99021 treatment and the interaction between both factors (F(2, 30) = 39.92 p < .001; F(1, 30) = 529.4, p < 0.001; F(2, 30) = 38.46, p < .001) and (m) secreted Il6 protein depending on the Zfp580 variant, CHIR99021 treatment and the interaction between both factors (F(2, 30) = 819.1, p < .001; F(1, 30) = 4762, p < 0.001; F(2, 30) = 802.1, p < .001) and (n) SUMO2/3 immunoblot of bEnd.3 exposed to 20 mM NaCl or LiCl for 24 h. LiCl did not induce global SUMOylation of proteins. The Graph shows a scatter dot plot of the mean of independent experiments ± standard deviations.

Figure 3.

Effects of lithium and Zfp580 on endothelial cell migration and proliferation. Endothelial cell migration was determined using the scratch wound assay. Shown is the % of the remaining gap width after 24 h compared to the initial gap size. Increased expression of Zfp580 increased migration compared to that of native bEnd.3 cells (a) (two-tailed unpaired Student`s t-test: t(10) = 14.18, p < .001), and knockdown of Zfp580 reduced migration compared to that of bEnd.3 cells transfected with the LacZ control (b) (One-way ANOVA with Tukey`s multiple comparison test: F(2, 15) = 24.5, p < .001). Exposure to 20 mM LiCl reduced migration compared to that of control cells treated with 20 mM NaCl (c) (two-tailed unpaired Student`s t-test: t(10) = 3.872, p = .003). Blockade of Il6 trans-signaling by the addition of 0.5 µg/ml, 1 µg/ml or 2 µg/ml sgp130 dose-dependently reduced the inhibitory effect of Zfp580 knockdown on endothelial cell migration (d) (One-way ANOVA with Tukey`s multiple comparison test: F(5, 30) = 25.42, p < .001) and reduced the inhibitory effect of 20 mM LiCl (e) (One-way ANOVA with Tukey`s multiple comparison test: F(5, 30) = 18.94, p < .001). This was not the case for shRNA LacZ-expressing or with 20 mM NaCl treated cells. Knockdown of Zfp580 completely abolished the effect of LiCl on endothelial cell migration (f) depending on knockdown, LiCl treatment and interaction between both factors (F(2, 15) = 19.22, p >.001; F(1, 15) = 13.69, p = 0.002; F(2, 15) = 8.971, p = .003). The inhibitory effect of LiCl on endothelial cell migration was unchanged by increased expression of wild-type Zfp580, but this effect was completely blocked by expression of Zfp580K31R (g) depending on Zfp580 variant, LiCl treatment and interaction between both factors (F(2, 15) = 161.8, p < .001; F(1, 15) = 249.49, p < 0.001; F(2, 15) = 44.82, p < .001). In Zfp580K31R, Il6 trans-signaling inhibition by sgp130 showed no further effect on endothelial cell migration (h). (i) Illustration of the signaling. Without lithium, Zfp580 suppresses Il6 expression. In the presence of lithium, Zfp580 is increasingly suppressed, SUMOylated and excluded from the nucleus. Therefore, Il6 expression is disinhibited. In the presence of sIl6r, Il6 inhibits endothelial cell migration by Il6 trans-signaling. After the addition of sgp130, Il6 trans-signaling is inhibited, and classical Il6 signaling enhances endothelial cell migration. (j) and (k) Cells were exposed to 4 h OGD directly after applying a scratch to the monolayer and incubated with BrdU. Migration (j) and proliferation (k) was determined 48 h later. Each, OGD and Zfp580 knock down reduced endothelial cell migration and proliferation. Combination of OGD and Zfp580 knock down almost completely prevented migration (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 30) = 298.6, p < .001; shRNA F(2, 30) = 22.73, p < 0.001; interaction F(2, 30) = 0.1718, p = .843) and proliferation (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 30) = 397.0, p < .001; shRNA F(2, 30) = 86.14, p < 0.001; interaction F(2, 30) = 27.27, p < .001). Graphs show scatter dot plots of the means of independent experiments ± standard deviations.

Figure 4.

OGD induces Il6 via Zfp580 suppression and inactivation, and Zfp580 loss improves endothelial and neuronal survival after OGD. Stroke was simulated in vitro by oxygen and glucose deprivation (OGD) for 4 h in endothelial cells ((a) – (e) and for 2.5 h in neurons ((f) – (n). Cells cultured under normal oxygen and glucose levels served as negative controls. Samples were taken after reoxygenation for several periods after OGD, as indicated in the figure. (a) Real-time RT-PCR for Zfp580 mRNA. In endothelial cells, mRNA expression of Zfp580 was immediately suppressed after OGD, increased after 6 h of reoxygenation and restored to baseline thereafter (descriptive results, n = 3). (b) Immunoblot for the Zfp580 protein. Zfp580 protein expression was regulated in a manner similar to that of Zfp580 mRNA expression. (c) Il6 mRNA was inversely regulated after OGD (descriptive results, n = 3). (d) Increased expression of wtZfp580 failed to block Il6 mRNA induction after OGD, whereas expression of SUMO conjugation-resistant Zfp580K31R significantly reduced Il6 mRNA induction compared to wtZfp580 (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 20) = 24.82, p < .001; Zfp580 variant F(1, 20) = 10.03, p = 0.005; interaction F(1, 20) = 0.833, p = .372). (e) Number of surviving endothelial cells 48 h after 4 h OGD. OGD caused cell loss and knock down of Zfp580 significantly increased the number of surviving cells (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 30) = 97.14, p < .001; shRNA F(2, 30) = 20.27, p < 0.001; interaction F(2, 30) = 15.42, p < .001). (f) and (g). In neurons, quantification of immunofluorescence staining for Zfp580 showed an immediately loss of Zfp580 protein levels after OGD (Two-way RM ANOVA with Sidak`s multiple comparison test: OGD F(1, 10) = 113.9, p < .001; time F(2.321, 23.21) = 3.022, p = 0.061; interaction F(3, 30) = 4.565, p = .009). (h) The mRNA expression of Continued.Zfp580 was immediately suppressed up to 24 h after OGD (descriptive results, n = 3). ((i) – (k) Neurons were subjected to 2.5 h OGD. Images were taken before and 24 h after OGD at identical positions (i). The number of surviving cells was counted (j) and the released LDH was determined from the cell culture supernatants (k). Knockdown of Zfp580 prior OGD increased the number of surviving cells (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 12) = 56.51, p < .001; shRNA F(2, 12) = 10.49, p = 0.002; interaction F(2, 12) = 4.791, p = .03) and reduced the LDH release (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 18) = 30.32, p < .001; shRNA F(2, 18) = 9.804, p = 0.001; interaction F(2, 18) = 8.931, p = .002). (l) Expression of the non-SUMOylatable Zfp580K31R variant increased the LDH release after OGD compared to the wild-type variant (Two-way ANOVA with Tukey`s multiple comparison test: OGD F(1, 20) = 61.22, p < .001; shRNA F(1, 20) = 7.245, p = 0.014; interaction F(1, 20) = 5.307, p = .032). (m) Treatment with 0.7 mM LiCl had no effect on neuronal LDH release. (n) Treatment with 20 mM LiCl significantly reduced LDH release, but LiCl was toxic at control conditions (Two-way RM ANOVA with Sidak`s multiple comparison test: OGD F(1, 6) = 155.1, p < .001; LiCl F(1, 6) = 5.28, p = 0.061; interaction F(1, 6) = 48.58, p < .001). Graphs show scatter dot plots of the means of independent experiments ± standard deviation.

Figure 5.

Zfp580 is ubiquitously expressed throughout the brain in neurons and vessels, and genomic ablation has no effect on cognitive or motor function. (a) Zfp580, NeuN, and DAPI immunofluorescence staining of C57BL/6 mouse brain slices. Zfp580 expression was found throughout the brain, particularly in neurons and blood vessels. (b) In Zfp580tm1a mice, no Zfp580 signal was detectable by immunofluorescence. Genomic ablation of Zfp580 had no biasing effect on the performance of mice in the Novel object recognition test (c)–(e), Rotarod test (f), Y-Maze test (g) and (h) or Staircase test (i) and (j). Graphs show scatter dot plots of the means of independent experiments ± standard deviations ((d)–(f), (i) and (j)), median ± interquartile range ((c), (g) and (h)).

Figure 6.

Genomic ablation of Zfp580 reduces ischemic lesion size, reduces endothelial proliferation, increases vessel resilience to ischemia, and modulates paracrine Il6 responses. Mice underwent transient filamentous occlusion of the middle cerebral artery for 45 min. (a) and (b) Lesion size was determined by MRI one day after ischemia and correlated to anatomical regions. The following mice were excluded from further analysis: mice without stroke, mice with stroke beyond the territory of the middle cerebral artery, and mice that reached predetermined humane endpoints and were therefore euthanized. There was no difference in lesion size one day after ischemia (Mann-Whitney test: p = .531). (c) Histological quantification of NeuN-DAB staining 21 days after ischemia revealed smaller lesion size in the Zfp580 knockout mice (two-tailed unpaired Student`s t-test with Welch`s correction: t(16.80) = 4.120, p < .001). (d) Caveolin-1, BrdU and DAPI staining of endothelial cells in the ischemic area of Zfp580 knockout mice and wild-type littermates 21 days after stroke. BrdU was administered from day two to five. (e) The amount of proliferating endothelial cells (Caveolin+/BrdU+ cells) was reduced in Zfp580tm1a in the ischemic area (Mann-Whitney test: p = .017). (f) Vessel density (two-tailed unpaired Student`s t-test: t(25) = 2.501, p < .019) and (g) Caveolin-1-positive vessel area (Mann-Whitney test: p = .009) were increased in Zfp580 knockout mice compared to wild-type littermates. (h) Real-time RT-PCR was performed from ipsilesional and contralesional brain slices taken from within the infarcted area 2 days after stroke. Il6 was strongly induced in the ipsilesional hemisphere in wild-type littermates, while Il6 induction was less pronounced in Zfp580 knockout mice (Two-way RM ANOVA with Sidak`s multiple comparison test: hemisphere F(1, 20) = 9.959, p = .005; treatment F(3, 20) = 5.706, p = 0.005; interaction F(3, 20) = 4.807, p = .011). ELISA for Il6 (i) and sIl6r (j) was performed from blood samples 2 days after stroke. (i) There was no difference in Il6 levels between genotypes (Mann-Whitney test: p = .063). (j) In Zfp580 knockout mice, sIl6r blood levels were higher than those in wild-type littermates (two-tailed unpaired Student`s t-test: t(16) = 2.444, p = .026) and (k) Calculated Il6:sIl6r binary complex. In Zfp580 knockout mice, Il6 trans-signaling was increased (two-tailed unpaired Student`s t-test: t(16) = 2.565, p = .021). Graphs show scatter dot plots of data from independent mice or samples derived thereof ± standard deviations ((c), (f), (h), (j) and (k)) or median ± interquartile range ((b), (e), (g) and (i)).

Figure 7.

Zfp580 knockout modulates cerebral connectivity and improves functional recovery after stroke. (a) and (b) Connectivity was determined 21 days after MCAo using diffusion tensor imaging. The connection strength of Zfp580 knockout mice compared to wild-type littermates is displayed using T-statistics with p < .001 chosen as relevant. (a) An increase of connection strength was observed in 11 connections with a focus on the contralesional external segment of the globus pallidus and the dorsal part of the contralesional retrosplenial area. (b) A decrease of connection strength was observed in 12 connections with a focus on the contralesional posterolateral visual area and contralesional dentate gyrus. Functional recovery of these mice was determined by behavioral tests. (c) There was no difference in the modified DeSimoni Neuroscore (Two-way RM ANOVA with Sidak`s multiple comparison test: genotype F(1, 27) = 0.1516, p = .7; time F(1.716, 46.32) = 14.41, p < 0.001; interaction F(2, 54) = 0.075, p = 928). Zfp580 knockout mice performed better in the Rotarod test (d) (Two-way RM ANOVA with Sidak`s multiple comparison test: genotype F(1, 25) = 4.807, p = .038; time F(1.605, 40.11) = 28.33, p < 0.001; interaction F(2, 50) = 1.849, p = .168). (e) There was no Continued.difference in performance of the left forepaw in the staircase test between genotypes (Two-way RM ANOVA with Sidak`s multiple comparison test: genotype F(1, 27) = 0.001, p = .97; time F(9.616, 259.6) = 5.144, p < 0.001; interaction F(33, 891) = 0.6454, p = .94). (f) The performance of the right forepaw recovered better in Zfp580 knockout mice (Two-way RM ANOVA with Sidak`s multiple comparison test: genotype F(1, 27) = 5.825, p = .023; time F(8.761, 236.5) = 13.87, p < 0.001; interaction F(33, 891) = 2.285, p < .001) and (g) Zfp850-knockout mice used the right forepaw more than wild-type littermates (Two-way RM ANOVA with Sidak`s multiple comparison test: genotype F(1, 27) = 7.499, p = .011; time F(10.80, 291.7) = 2.947, p = 0.001; interaction F(33, 891) = 1.536, p = .028). Graphs show scatter dot plots of data from independent mice or samples derived thereof ± standard deviations.