https://doi.org/10.5061/dryad.q573n5tt0

Description of the data and file structure

To better show the differences between the groups, we have highlighted the mean and standard deviation for each group in red.

The in vitro experiments were conducted on primary rat hippocampal neurons. In accordance with previous study in hippocampal CA1, a progressive decline of nuclear β-catenin was observed compared with control group, whereas no differences were found in total and cytoplasmic β-catenin at 2-24 h after oxygen-glucose deprivation/reoxygenation (OGD/R) (Fig. 1A). Next, the proteasome inhibitor MG132 was used to investigate whether OGD/R promotes nuclear β-catenin degradation through the ubiquitin-proteasome pathway. As expected, the OGD/R-induced downregulation of nuclear β-catenin was completely blocked by MG132 (Fig. 1B). Consistently, OGD/R treatment significantly increased the K48-Ub of nuclear β-catenin in primary hippocampal neurons (Fig. 1C). Flag-KDM2A, but not the Flag-KDM2A-∆F-box, reduced nuclear β-catenin expression both in HEK-293T cells and SH-SY5Y cells (Fig. 1E).

To determine whether SKP1 and CUL1 together with KDM2A participate in HPC-mediated ischemic tolerance in rats, firstly we detected the expressions of SKP1 and CUL1 in CA1 after tGCI with or without hypoxia. The number of SKP1-positive neuron-like cells increased at 26 h after tGCI, while HPC inhibited this increase of neuron-like cells at the same time points of reperfusion after tGCI (Fig. 2B). Also, western bolt analysis showed that HPC obviously attenuated the tGCI-induced upregulation of SKP1 in the early stage of reperfusion (Fig. 2C). Notably, HPC reversed the increase of SKP1-positive glia-like cells and the reduction of SKP1-positive neuron-like cells induced by tGCI in the late stage of reperfusion (Fig. 2B).

In line with SKP1, the number of CUL-positive cells increased at 26 h, while decreased markedly at 168 h after tGCI. However, when compared with tGCI groups at corresponding time points of reperfusion, the expression of CUL1 reduced in the HPC groups (Fig. 3B). The above results were further confirmed by western blot analysis (Fig. 3C). We also confirmed that the trend of SKP1 expression in cytoplasmic fraction was similar to that of whole cell lysate in CA1, whereas CUL1 was less expressed in cytoplasm. There were no significant differences in cytoplasmic CUL1 between tGCI and HPC groups (Fig. 3E). Subsequently we measured SKP1 and CUL1 levels in the nucleus. In line with the expressions of total SKP1 and CUL1, the expressions of nuclear SKP1 and CUL1 were significantly increased after tGCI compared with the Sham group. However, HPC markedly reduced the tGCI-induced elevation of nuclear SKP1 and CUL1 in CA1 (Fig. 3F). Co-immunoprecipitation assay showed that the interactions of KDM2A-SKP1 and KDM2A-CUL1 were increased within nucleus in CA1 after tGCI, which was inhibited by HPC (Fig. 3G).

Further, we observed an obvious increase of K48-Ub of nuclear β-catenin in tGCI group at 26 h and 50 h after reperfusion via co-immunoprecipitation assay. Conversely, in HPC group, the increased K48-Ub of nuclear β-catenin was reversed (Fig. 4A). To investigate the role of KDM2A in mediating the recognition of substrate β-catenin by SCFKDM2A E3 ligase complex and subsequently catalyzing the K48-Ub of nuclear β-catenin in CA1 after tGCI, KDM2A small-interfering RNA (siKDM2A) was then utilized. The validity in inducing KDM2A silence, the maintenance for nuclear β-catenin stabilization after tGCI and the safety for CA1 neurons of siKDM2A administration had been previously confirmed5. As shown in Fig. 4B, nuclear KDM2A was analyzed after siKDM2A administration and the result was consistent with our previous studies5. There was no significant difference on nuclear CUL1 with or without siKDM2A administration. Interestingly, siKDM2A administration inhibited nuclear SKP1 in tGCI rats when compared to NC group (Fig. 4B). As expected, the increase of K48-Ub level of nuclear β-catenin and the interactions among KDM2A, CUL1, SKP1 and nuclear β-catenin in CA1 induced by tGCI were counteracted by siKDM2A administration (Fig. 4C). To clarify the relation between KDM2A-mediated demethylation and K48-Ub of nuclear β-catenin, daminozide, a potent small-molecule inhibitor of demethylase activity of KDM2A, was utilized. Daminozide at 20 μM and 40 μM up-regulated total lysine methylation of Sham rats when compared to vehicle, and the effect at 40 μM on total lysine methylation was more significant (Supplementary Fig. 1). However, no effects on total KDM2A level were observed after daminozide administration. Thus, 40 μM of daminozide was applied for subsequent studies. Further experiments confirmed that daminozide treatment reversed the tGCI-induced reductions of nuclear β-catenin and methylated β-catenin, as well as the elevation of the K48-Ub of nuclear β-catenin in CA1 at 26 h of reperfusion, whereas there were no significant differences in CA1 of Sham rats between daminozide and vehicle treatments (Fig. 4D&E).

To further support the involvement of β-catenin in HPC-mediated cerebral ischemia tolerance, the adeno-associated viruses (AAV) containing Ctnnb1 small interfering RNA (AAVi-Ctnnb1) to inhibit the expression of β-catenin was injected into bilateral CA1 region. The expression of β-catenin significantly decreased after AAVi-Ctnnb1 administration at the dosage of 7.92 × 109 v.g or 15.84 × 109 v.g in Sham rats, which was more effective at the higher dosage (Fig. 5B). Thus, the dosage of 15.84 × 109 v.g was selected in the following experiments. As expected, AAVi-Ctnnb1 administration dramatically aggravated neuronal damage in CA1 with HPC, whereas AAVi-Ctnnb1 or AAVi-CON did not alter the numbers of surviving and NeuN-positive cells in Sham rats (Fig. 5C-E). Besides, compared with the AAVi-CON group, AAVi-Ctnnb1 significantly reduced nuclear β-catenin and survivin in CA1 of Sham and HPC rats (Fig. 5F&G).

Compared with sham-operated rats, the number of USP22-positive cells presented a sustained decrease during 26-168 h after reperfusion of tGCI. Conversely, HPC reversed tGCI-induced reductions of USP22-positive cells to a certain extent (Fig. 6B).  Also, the reduction of USP22 in CA1 at 26 and 50 h of reperfusion after tGCI was confirmed by western blot. Inconsistently, no statistical differences were observed in the level of total USP22 between tGCI and HPC groups (Fig. 6D). In line with total USP22, tGCI significantly down-regulated cytoplasmic and nuclear USP22 in CA1. However, HPC remarkably inhibited the reduction of nuclear USP22 induced by tGCI, but had no effect on cytoplasmic USP22 (Fig. 6E&F).

Next, we examined the interaction between USP22 and β-catenin within nucleus after tGCI with or without hypoxia by co-immunoprecipitation assay. The results showed that the interaction between USP22 and β-catenin in nucleus was significantly weakened at 50 h after tGCI. In contrast, HPC completely reversed this reduction induced by tGCI (Fig. 6G). Western blot analysis showed that USP22 expression was augmented significantly after AAV-USP22 administration in Sham rats (Fig. 7C). Similar change was observed in total β-catenin level (Fig. 7D). AAV-USP22 or AAV-Con had no impact on the neuronal number of Sham rats. Obviously, AAV-USP22 treatment markedly ameliorated neuronal damage in CA1 after tGCI, which was demonstrated by an increase in the number of surviving and NeuN-positive cells as well as a decrease in Fluoro-Jade B (F-JB)-positive cells. Additionally, an additive neuroprotective effect was observed in HPC group with AAV-USP22 administration (Fig. 7E-H).

Further, Morris water maze (MWM) test was performed to measure learning and memory functions. The swimming path tracings during the training period (learning) and probe trial (memory) in each group were shown as Fig. 8B. The escape latency and path length of rats in six groups decreased in a time-dependent manner over 5 d hidden-platform training sessions. For the rats injected with AAV-Con, the path length and escape latency in tGCI group were significantly prolonged compared with Sham group. Inversely, HPC improved the spatial learning ability of rats at 4rd to 6th day after tGCI (Fig. 8C&D). In the probe phase, the percentage of time that rats spent in the target quadrant after removing the platform was recorded to assess memory function. As shown in Fig. 8E, HPC partially reversed the reduction induced by tGCI in target quadrant occupancy. These results were consistent with our previous study in the absence of AAV intervention5. Compared with AAV-Con administration, USP22 overexpression significantly enhanced spatial learning ability and long-term memory of tGCI rats, as presented by the shorter time in the swimming path length and escape latency as well as the longer time in the target quadrant occupancy (Fig. 8B-E). There was a slight increase in cytoplasmic USP22 of CA1 after treatment with AAV-USP22, but the difference was not statistically significant. Additionally, AAV-USP22 did not alter cytoplasmic β-catenin expression. Consistent to our previous study, no survivin expression could be observed in cytoplasmic fraction of CA14. As expected, compared to AAV-Con group, AAV-*USP22 *significantly upregulated the nuclear USP22 and β-catenin in CA1 of Sham, tGCI and HPC groups. Moreover, AAV-USP22 reversed tGCI-induced reduction of survivin in CA1 (Fig. 8G). Furthermore, the elevation of K48-Ub of β-catenin and the reduction of methylated β-catenin within nucleus in CA1 after tGCI were counteracted with AAV-USP22 treatment）(Fig. 8H).

Files and variables

File: Supplementary_Data_1.pdf

Description: The numerical source data for the graphs

File: Supplementary_Figures.pdf

Description: Western blot analyses of lysine methylation level and KDM2A in  Sham rats treated with daminozide and all uncropped and unedited blot images

Code/software

Statistical analyses were performed with Statistical Package for Social Sciences Software for Windows, version 25.0 (SPSS, Inc., Chicago, IL, USA). The group size “n” simply refers to biologically independent cell/animal experiments rather than technical replications. The evaluation of sample size was based on our and others published studies and the results from our preliminary experiments in this study. All data represent at least 3 independent experiments. For comparison between two groups conforming to the normal distribution, the unpaired two-tailed Student’s t-test was applied. Multiple comparisons were conducted using one-way ANOVA when the data between Sham and tGCI groups were normally distributed. Analyses were followed by a Bonferroni or Tamhane’s T2 post hoc test according the results of homogeneity of variance test. When the data did not conform to normal distribution, nonparametric tests were used, including Mann-Whitney's U test for the comparison between two groups and Kruskal-Wallis' for multiple comparisons. As the data from path length, escape latency, and swimming speed are repeatedly measured and time-dependent variables, these data were treated with Mauchly's test of sphericity and Multivariate Analysis of Variance (MANOVA). Data graphing was performed with GraphPad Prism version 6.0. All variables were expressed as mean±standard deviation (SD) and represented by scatter plots. Values of p<0.05 were considered statistically significant.