Stroke Triggers Dynamic m⁶A Reprogramming of Cerebral Circular RNAs

Abstract

We previously showed that stroke alters circular RNA (circRNA) expression profiles. Many circRNAs undergo epitranscriptomic modifications, particularly methylation of adenosine to form N⁶-methyladenosine (m⁶A). This modification significantly influences the circRNA metabolism and functionality. Hence, we currently evaluated if transient focal ischemia in adult C57BL/6J mice alters the m⁶A methylation of circRNAs. Changes in m⁶A were profiled in the peri-infarct cortex following immunoprecipitation coupled with microarrays. Correlation and gene ontology analyses were performed to understand the association of m⁶A changes with circRNA regulation and functional implications after stroke. Many circRNAs showed differential regulation (up or down) after stroke, and this change was highest at 24h of reperfusion. Notably, most circRNAs differentially regulated after stroke also exhibited temporal changes in m⁶A modification patterns. The majority of circRNAs that showed post-stroke differential m⁶A modifications were derived from protein-coding genes. Hyper- than hypomethylation of circRNAs was most prevalent after stroke. Gene ontology analysis of the host genes suggested that m⁶A-modified circRNAs might regulate functions such as synapse-related processes, indicating that m⁶A epitranscriptomic modification in circRNAs could potentially influence post-stroke synaptic pathophysiology.

INTRODUCTION

Stroke triggers rapid changes in the expression profiles of several classes of noncoding RNAs, including microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and circular RNAs (circRNAs) (Chokkalla et al., 2019; Dharap et al., 2009; Dharap et al., 2012; Kim et al., 2018; Mehta et al., 2020; Mehta et al., 2017; Morris-Blanco et al., 2021). These genomic alterations shape the secondary brain damage and influence functional recovery after stroke (Dharap et al., 2009; Dharap et al., 2012; Mehta et al., 2023a; Mehta et al., 2020; Mehta et al., 2017). Interestingly, among all noncoding transcripts, circRNAs are highly stable due to their resistance to exonuclease-mediated degradation and half-lives that are >2.5 times longer than their linear counterparts (Enuka et al., 2016; Jeck et al., 2013). In addition, defined by high conservation, dynamic expression patterns, and abundant enrichment in the brain, circRNAs are believed to specify neuronal identity and define neuronal functions (Dong et al., 2023; Rybak-Wolf et al., 2015).

In most RNAs, ~0.4–0.6% of adenosines undergo methylation at the N⁶ position to form N⁶ methyladenosine (m⁶A) (Wei et al., 1975; Zhang et al., 2019). The m⁶A modification is a dynamic and reversible process controlled by specific methyltransferases (writers), demethylases (erasers), and readers (Chokkalla et al., 2020). The m⁶A modification influences the structure, stability, and function of RNAs (Chokkalla et al., 2019; Chokkalla et al., 2020). The m⁶A modifications usually occur in the coding regions of long exons, with higher concentrations observed near the stop codons and the 3′ untranslated regions (UTRs) and a lower enrichment in the 5′ UTRs (Chokkalla et al., 2020; Meyer et al., 2012).

We recently reported that stroke significantly increases m⁶A methylation of mRNAs in the ischemic brain (Chokkalla et al., 2019). We also showed that elevation in m⁶A levels resulted from decreased demethylase FTO in the post-stroke brain, while the m⁶A writer complex (METTL3, METTL14, and WTAP) remained unchanged (Chokkalla et al., 2019). Recent evidence indicates that adenines in circRNAs also undergo m⁶A modification and most m⁶A-tagged circRNAs originate from protein-coding genes (Li et al., 2020; Zhou et al., 2017). This is because m⁶A sites in mRNAs are predominantly located in the last exon, whereas circularization of the last exon of genes is rare (Zhang et al., 2014). We previously showed that stroke significantly alters the cerebral circRNA expression profiles. As the m⁶A modification modulates the biogenesis, stability, and function of circRNAs (Huang et al., 2022), we presently evaluated the effect of transient focal ischemia on m⁶A methylation of circRNAs.

METHODS

Focal ischemia:

The Research Animal Resources and Care Committee of the University of Wisconsin-Madison approved all animal procedures. Animals were randomly assigned to study groups and cared for in accordance with the Guide for the Care and Use of Laboratory Animals, U.S. Department of Health and Human Services Publication Number. 86–23 (revised). Adult male C57BL/6J mice (12 weeks, 26±2 g, Charles River, USA) were subjected to 1h of transient focal ischemia by intraluminal middle cerebral artery occlusion (MCAO) using a 6–0 silicon-coated monofilament (Doccol Corporation, USA) under isoflurane anesthesia, as described earlier (Chokkalla et al., 2023; Chokkalla et al., 2019; Kim et al., 2018; Mehta et al., 2023a; Mehta et al., 2017; Morris-Blanco et al., 2021). The body temperature was maintained at 37.0 ± 0.5°C during the surgical procedure and 90 min of reperfusion. The regional cerebral blood flow and physiological parameters (pH, PaO2, PaCO2, hemoglobin, and blood glucose) were monitored. Sham-operated mice underwent a similar surgical procedure except for MCAO. Mice were euthanized at 6h, 12h, and 24h of reperfusion. All efforts were made to minimize animal suffering and to reduce the number of animals used.

m⁶A Immunoprecipitation:

From each mouse, total RNA was extracted from the peri-lesional cortex by acid phenol-chloroform and precipitated with ethanol. The RNAs tagged with m⁶A were immunoprecipitated with affinity-purified anti-m⁶A rabbit polyclonal antibody (Synaptic Systems). Briefly, 3–5 μg total RNA and m⁶A spike-in control were added to IP buffer (50 mM Tris-HCl, pH7.4, 150mM NaCl, 0.1% NP40, 40U/μL RNase Inhibitor) containing 2 μg anti-m⁶A antibody and incubated Dynabeads^™ M-280 Sheep Anti-Rabbit IgG at 4°C for 2h. The beads were washed (3X) with IP buffer and 500 μL Wash buffer (50 mM Tris-HCl, pH 7.4, 50 mM NaCl, 0.1% NP40, 40 U/μL RNase Inhibitor). The enriched RNA was extracted with Elution buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 0.05% SDS, 40U Proteinase K, 1 μL RNase inhibitor).

RNase R Treatment, Labeling and Hybridization:

The immunoprecipitated and supernatant RNAs were digested with RNase R (Epicentre, Inc.) to enrich circRNAs and remove linear RNAs and spiked in with spike-in control RNA, amplified and labeled with Cy3 (supernatant) and Cy5 (immunoprecipitated) using Super RNA Labeling Kit (Arraystar). The circRNAs were purified with RNeasy Mini Kit, fragmented, and hybridized to the m⁶A-circRNA Epitranscriptomic Microarray slide. The slides were washed, fixed, and scanned using an Agilent Scanner G2505C.

circRNA m⁶A microarray data analysis:

The data from each spot on a microarray was extracted using Agilent Feature Extraction software (version 11.0.1.1). Raw intensities of immunoprecipitated, Cy5-labelled, and supernatant, Cy3-labelled, were normalized with an average of log2-scaled Spike-in RNA intensities. The probe signals having Present (P) or Marginal (M) QC flags were retained for m⁶A quantity analyses. Differentially m⁶A-methylated circRNAs between two comparison groups were identified by filtering with the fold change and statistical significance (p-value) thresholds. The significantly differentially expressed circRNAs (fold change of ≥1.5 and ≤0.75; p<0.05) with methylation level of ±5% were identified at each time point of reperfusion compared to sham for further analysis. The gene ontology (GO) analysis was performed based on host genes of differentially m⁶A methylated circRNAs (Huang da et al., 2009; Zhou et al., 2016). The correlation between the m⁶A % difference and the m⁶A IP signal was calculated using Spearman correlation with R programming for circRNA regulation and circRNA m⁶A modification (at each time point of reperfusion) and m⁶A circRNAs and m⁶A mRNAs (at 12h). The Bonferroni method was applied to adjust for multiple testing, resulting in adjusted correlation p-values.

RESULTS

Transient focal ischemia altered m⁶A tagging to circRNAs:

We previously showed that transient MCAO significantly altered circRNA profiles in the peri-ischemic cortex (Mehta et al., 2017). To understand if ischemia-altered circRNAs underwent differential m⁶A modification, we profiled the immunoprecipitated m⁶A-methylated RNAs isolated from the peri-infarct cortex of mice subjected to transient MCAO at 6h, 12h, and 24h of reperfusion. Immunoprecipitated m⁶A-methylated RNAs from sham-operated mice served as control. Out of a total of 13,842 circRNAs profiled using microarray, 1,138 (8.2%) were differentially regulated at 6h (60 up and 1078 down), 749 (5.4%) were differentially regulated at 12h (245 up and 504 down) and 1,768 (12.7%) were differentially regulated at 24h (911 up and 857 down) of reperfusion following 1h of transient MCAO compared with sham (Fig. 1A and Supplementary Table 1). Table 1 shows the top 10 (5- up/hyper and 5- down/hypo) differentially regulated m⁶A-modified circRNA at 6h, 12h and 24h of reperfusion following transient MCAO. Interestingly, among these, 146 circRNAs exhibited differential regulation across all time points studied (Fig. 1B).

Fig. 1: circRNAs hypermethylation increased with reperfusion time.

Differential regulation of circRNAs between 6h, 12h, and 24h of reperfusion following transient MCAO (A). Differentially regulated circRNAs (both up and down) between the comparison groups were examined based on normalized IP intensities of microarray data and identified through filtering using fold change (≥1.5 Up and ≤0.75 down) and statistical significance thresholds (p≤0.05 of n=5/group). 6h, 12h, and 24h are the reperfusion time points after transient MCAO. The distribution of common and unique circRNAs among the three groups is illustrated in Venn diagram (B). Volcano plots display the upregulated (in red) and downregulated (in blue) circRNAs at 6h, 12h, and 24h post-reperfusion compared to sham (C). Solid-colored dots represent circRNAs that are both differentially regulated and differentially m⁶A methylated, while empty circles represent circRNAs that are differentially regulated but not differentially methylated compared to sham (C). The number of hyper-m⁶A methylated circRNAs increased temporally from 6h to 24h of reperfusion compared to sham (D).

Table 1:

Top 10 differentially expressed and methylated circRNAs after transient MCAO

circRNA	Expression(Δ fold)	Diff % of m6A	Genomic Location	Host Gene
6h reperfusion
circRNA_22217	2.71	15%	Antisense	Rsph14
circRNA_40569	3.70	13%	Exonic	Adamts9
circRNA_29814	2.03	13%	Sense overlap	Nfkbiz
circRNA_013120	1.77	11%	Intronic	Trrap
circRNA_19388	2.39	9%	Sense overlap	Scarb1
circRNA_29934	0.23	−25%	Intergenic	XX
circRNA_45972	0.44	−27%	Sense overlap	Frmpd4
circRNA_36350	0.32	−27%	Intronic	Negr1
circRNA_41021	0.18	−32%	Exonic	Gm5724
circRNA_36340	0.26	−33%	Exonic	Erich3
12h reperfusion
circRNA_40569	6.75	22%	Exonic	Adamts9
circRNA_45269	3.31	19%	Exonic	Ulk4
circRNA_23642	3.10	18%	Exonic	Smtnl2
circRNA_26154	2.90	18%	Exonic	Dcdc2a
circRNA_30653	2.06	17%	Exonic	Uhrf1
circRNA_013555	0.41	−19%	Antisense	Exoc6b
circRNA_35596	0.52	−19%	Exonic	Zbbx
circRNA_21597	0.30	−20%	Exonic	Cd46
circRNA_39168	0.35	−22%	Intronic	XX
circRNA_34685	0.41	−23%	Sense overlap	Macrod2
24h reperfusion
circRNA_32433	2.88	40%	Exonic	A1cf
circRNA_007078	4.67	33%	Exonic	Asxl2
circRNA_29625	6.25	30%	Exonic	Pdia5
circRNA_009418	3.11	29%	Antisense	Zfp711
circRNA_19052	5.84	29%	Sense overlap	Asxl2
circRNA_35595	0.41	−18%	Sense overlap	Bche
circRNA_008701	0.48	−19%	Exonic	Kcnn2
circRNA_29934	0.44	−19%	Intergenic	XX
circRNA_36698	0.73	−20%	Exonic	Reck
circRNA_21833	0.28	−20%	Intergenic	XX

All the circRNAs shown are mouse (mmu) isoforms. The Δ fold and Diff % of m⁶A are in comparison with sham. Values are mean of n =5/group with <15% SD in each case. XX indicates that the host gene for this circRNA has not yet been identified.

m⁶A methylation pattern also changed following stroke. A significant portion of the differentially regulated circRNAs displayed m⁶A methylation between 6h to 24h reperfusion (Fig. 1C). Specifically, at 6h post-reperfusion, ~59% of circRNAs showed altered m⁶A methylation. Among these, 24 hypermethylated and 2 hypomethylated were upregulated, while 5 hypermethylated and 637 hypomethylated were downregulated. At 12h post-reperfusion, ~29% of circRNAs displayed m⁶A modification, with137 hypermethylated and 3 hypomethylated among the upregulated group and 20 hypermethylated and 54 hypomethylated among the downregulated. Similarly, at 24h post-reperfusion, ~59% of circRNAs exhibited m⁶A modification, with 719 hypermethylated and 21 hypomethylated among the upregulated and 81 hypermethylated and 221 hypomethylated among the downregulated (Fig. 1D). Interestingly, circRNA m⁶A hypermethylation pattern increased with increasing reperfusion time following transient MCAO. Moreover, a substantial portion of circRNAs exhibited differential m⁶A methylation following transient MCAO, independent of alterations in their steady-state expression levels (Supplementary Fig. 1).

m⁶A-modified circRNAs are majorly formed from the exonic region of the protein-coding genes and are linked to synaptic function after stroke

Bioinformatics analysis showed that ~75% of the circRNAs that showed both altered expression and altered m⁶A methylation after stroke predominantly originate from the exonic location of protein-coding genes between 6h and 24h of reperfusion (Fig. 2A). Whereas, 13–15% of the post-stroke m⁶A-modified circRNAs showed sense overlap (Fig. 2A). The circRNAs originating from antisense, intronic and intergenic regions of the genome showed minimal (2–3%) post-stroke alterations in m⁶A methylation at any of the reperfusion periods evaluated (Fig. 2A). Interestingly, exonic m⁶A hypermethylation increased progressively from 6h to 24h reperfusion (2% to 57%) following transient MCAO (Fig. 2A) indicating that stroke influences protein-coding genes significantly. The common and unique distribution of m⁶A-modified circRNAs between the 3 reperfusion time points studied is shown in Fig. 2B.

Fig. 2: Majority of circRNAs methylated differentially after stroke are exonic in origin and linked to synaptic function.

Differentially m⁶A methylated circRNAs of different genomic origins showed progressive hypermethylation with time of reperfusion compared with sham (A). Only three circRNAs exhibited both differential expression and m⁶A modification at all reperfusion time points following transient MCAO (B). Spearman’s correlation analysis identified a strong positive correlation between m⁶A methylation (%) and expression (IP signal) at all time points (C). Overall differential (%) circRNA m⁶A methylation pattern at different time points of reperfusion compared to sham (D). GO analysis for the biological process, and cellular component of post-stroke differentially m⁶A-modified circRNAs based on the host genes (E).

Most m⁶A-modified circRNAs showed a positive correlation with their expression (showed either increased levels and methylation or decreased levels and methylation) at all 3-time points of reperfusion (Fig. 2C). However, the correlation analysis between m⁶A-modified circRNAs and their corresponding m⁶A-modified mRNAs (host genes) analyzed at 12h of reperfusion (Chokkalla et al., 2019) did not show any significant correlation (Data not shown). Overall, the data indicated an extensive change in circRNA m⁶A methylation compared to the sham control, reaching its peak at 24h of reperfusion (Fig. 2D). The Gene Ontology (GO) enrichment analysis of the host/overlapping genes associated with differentially post-stroke m⁶A methylated circRNAs showed a predominance of synapse-associated functions indicating their role in shaping synaptic responses following a stroke (Fig. 2E and Supplementary Fig. 2).

DISCUSSION

We presently show that stroke induces m⁶A methylation of circRNAs that are predominantly derived from the exonic regions of protein-coding genes. We further observed that m⁶A-modified circRNAs increased temporally with reperfusion time, correlating with the deregulation of synaptic function in the post-stroke brain. As an abundant modification of RNAs that modulates all aspects of post-transcriptional RNA metabolism, m⁶A methylation of mRNAs is thought to mediate the neuronal response to post-stroke secondary brain damage (Chokkalla et al., 2023; Chokkalla et al., 2019). We previously reported that post-stroke m⁶A methylation of mRNAs inversely correlates with the expression and/or activity of FTO which is the major m⁶A demethylase in brain (Chokkalla et al., 2023; Chokkalla et al., 2019).

Interestingly, similar to mRNAs, m⁶A is the most abundant epitranscriptomic modification on circRNAs as well (Zhou et al., 2017). Stroke is known to alter the expression of circRNAs, thus potentially impacting the regulatory framework of these molecules (Mehta et al., 2017). As an example, circRNA CDR1as, which was decreased in the brain after stroke, was shown to modulate the abundance and function of the microRNA miR-7 (Mehta et al., 2023a; Mehta et al., 2017; Piwecka et al., 2017). The miR-7 targets α-synuclein (α-Syn) mRNA and prevents its translation. We recently showed that stroke down-regulates miR-7, leading to derepression of α-Syn protein expression that promotes ischemic brain damage (Kim et al., 2018; Mehta et al., 2023b). We further showed that CDR1as reconstitution promotes the survival of miR-7 that, in turn, suppresses α-Syn and thus facilitates functional recovery after stroke (Mehta et al., 2023a).

Our present study shows that circRNAs exhibit varying levels of m⁶A methylation during the reperfusion periods from 6h to 24h following transient MCAO. The relationship between the altered circRNA methylation to post-stroke outcomes is currently not clear. However, given their high abundance, conservation, stability, and dynamic spatiotemporal expression in tissues, circRNAs were thought to participate in the etiology of CNS disorders such as neuropsychiatric disease and Alzheimer’s disease (Dong et al., 2023; Mehta et al., 2020; Puri et al., 2023). Furthermore, the m⁶A is known to promote circRNA biogenesis as well as their specificity to diseases (Rybak-Wolf et al., 2015; Tang et al., 2020). Our data showed a positive correlation between the differential regulation of circRNAs and their m⁶A methylation levels. This association was most prominent at the 24h of reperfusion following transient MCAO. Intriguingly, only 3 out of 146 circRNAs differentially expressed at all 3 reperfusion time points following transient MCAO exhibited altered m⁶A methylation. This suggests a temporally distinct regulation of circRNA biogenesis likely due to back splicing occurring at m⁶A-enriched sites situated around the start and stop codons of linear mRNA genes (Tang et al., 2020). We observed that circRNAs that were differentially regulated and modified with m⁶A after stroke predominantly (~75%) originated from the exonic regions of the protein-coding genes. Notably, the number of circRNAs exhibiting hypermethylation increased, while those displaying hypomethylation declined with progressing reperfusion time.

The circRNAs are known to be formed from the same genes as mRNAs by non-canonical backsplicing events (Kristensen et al., 2019). Hence, we analyzed if the mRNAs formed by the host genes also show altered m⁶A methylation after stroke. However, bioinformatics analysis between m⁶A-modified circRNAs and their corresponding m⁶A-modified mRNAs (formed from the same genes) at 12h of reperfusion following transient MCAO showed no significant correlation (data not shown). This is understandable as the m⁶A sites are predominantly found in the last exons of mRNAs, which rarely undergo circularization (Meyer et al., 2012; Zhang et al., 2014). Furthermore, the majority of circRNAs with m⁶A sites originate from genes encoding mRNAs with m⁶A sites on different exons compared to mRNAs that lack those. This suggests that while m⁶A modification is essential for mRNA function, it does not always act as a driving force for their circularization (Zhou et al., 2017). Furthermore, in neurons m⁶A levels did not affect m⁶A mRNA or m⁶A circRNA expression (Zhang et al., 2023). Our observations indicate that numerous circRNAs, despite not being changed in their levels, undergo significant differential m⁶A methylation following transient MCAO, which could potentially act by sequestering m⁶A reader proteins and thereby modulating the availability of these proteins for the regulation of other mRNA targets (Mehta et al., 2020). In support, it was suggested that m⁶A reader YTHDC1 interaction promotes the production of a common subset of circRNAs in rhabdomyosarcoma, potentially leading to cascading effects on cellular processes (Dattilo et al., 2023).

Secondary brain damage after transient MCAO in adult mice progresses with increasing time and matures by 24h of reperfusion (Buscemi et al., 2019). Intriguingly, we noted that overall m⁶A methylation changes exhibited the highest change at 24h of reperfusion. This observation suggests a potential mechanistic link between m⁶A methylation dynamics and the pathophysiology of ischemic brain injury. Among other vital functions, m⁶A-modified circRNAs derived from protein-coding genes strongly associate with synaptic regulation. Evidence suggests that even a mild or a moderate cerebral ischemia can cause synaptic dysregulation (Hofmeijer and van Putten, 2012).

The earliest consequence of cerebral ischemia appears to be the dysfunction of synaptic activity, leading to a breakdown in synaptic transmission and, ultimately, electrical silencing within the penumbra (Hofmeijer and van Putten, 2012). It is intriguing that neuronal circRNA production tends to cluster around synaptic function, with many circRNAs exhibiting sudden changes in abundance coinciding with synaptogenesis (Dong et al., 2023; You et al., 2015). These findings raise the possibility that m⁶A modifications on differentially regulated circRNAs might influence their translation potential, thereby impacting synaptic function after focal ischemia. For example, evidence indicates that circZNF609 could be translated by IRES driven pathway, while a point mutation in circZNF609 m⁶A sites could reduce its translation efficiency by ~50% (Di Timoteo et al., 2020; Legnini et al., 2017).

CONCLUSIONS

The present study shows that ischemic stroke leads to a dynamic temporal reprogramming of the cerebral circRNA epitranscriptome, which has the potential to affect post-stroke pathophysiology. Further investigation into the specific functional roles of these differentially methylated circRNAs is warranted to elucidate their potential as therapeutic targets to improve post-stroke outcomes.

HIGHLIGHTS.

• Stroke induces differential regulation of cerebral circRNAs

• Most circRNAs with altered regulation exhibit temporal changes in m⁶A modification

• m⁶A modified circRNAs primarily originate from protein-coding genes and display a higher level of hypermethylation compared to hypomethylation

• circRNA m⁶A modification might be crucial in regulating synapse-related processes after stroke