IGF2R circular RNA hsa_circ_0131235 expression in the middle temporal cortex is associated with AD pathology

Abstract Objective To identify circular RNAs as candidates for differential expression in the middle temporal (MT) cortex in a well‐characterized cohort with contrasting Alzheimer disease (AD) pathology and cognition. Top screen candidates were assessed for proof of circularity and then quantified by qPCR in a larger number of samples. Methods An initial RNA sequencing screen was performed on n = 20 frozen human tissue samples. Filters were applied to select candidate circular RNAs for further investigation. Frozen human tissue samples were selected for global AD pathology burden and global cognition scores (n = 100). Linear and divergent primers were used to assess circularity using RNaseR digestion. RT‐qPCR was performed to quantify relative hsa_circ_0131235 abundance. Results Eleven circular RNAs were selected for further investigation. Four candidates produced circular RNA primers with appropriate efficiencies for qPCR. RNaseR treatment and analysis by both basic PCR and qPCR confirmed hsa_circ_0131235 circularity. There was a significant main effect of AD pathology on hsa_circ_0131235 expression. Conclusions Elevated hsa_circ_0131235 expression in the MT cortex was significantly associated with AD pathology.

TA B L E 1 Gene targets and associated control and divergent primers. Primer sequences, efficiencies, and predicted product size listed for each primer pair   (Piwecka et al., 2017). Due to their high stability, circRNAs may also serve as synaptic tags to maintain molecular memory (Rybak-Wolf et al., 2015). A number of circRNAs have been previously associated with AD pathology and may explain more of the variation in clinical dementia rating of AD cases than ApoE4 risk gene load (Dube et al., 2019).
In order to identify circRNAs that may be involved in AD pathophysiology, we performed a small RNA sequencing screen of potential circRNAs in frozen tissue from the human middle temporal (MT) cortex. Using a series of filters, we selected 11 candidate circRNAs for follow-up. Following assessment of circularity, 1 target, IGFR2, was quantified by RT-qPCR in a larger sample set. This cohort included well-matched individuals with contrasting presentations of AD pathology and cognition, to allow for somewhat independent evaluation of circRNA association with cognition or pathology.
IGFR2 hsa_circ_0131235 was established as clearly circular, and expression was found to be associated with AD pathology.

| Human tissue
Frozen human tissue was used according to human research protocols as approved by the Partners Institutional Review Board. 100 cases from the Rush Religious Orders Study (ROS) and Memory and Aging Project (MAP) cohorts (Bennett et al., 2013) were selected according to global AD pathology burden and the presence of cognitive impairment. Subjects were classified into four contrasting groups according to two variables. The first variable was Braak Score, with those individuals with a Braak score of 4 or below deemed low AD pathology burden, and those with a Braak score of 5 or 6 deemed high AD pathology burden. The second variable was clinical consensus as to the presence of cognitive impairment at the point of death.

| RNA sequencing
Exploratory RNA sequencing was performed on frozen human MT tissue (n = 20, 5 samples per diagnostic group) to identify candidate circRNA targets. Reads were aligned to the hg19 genome using STAR aligner, and read counts were normalized using DESeq2 (Love et al., 2014). High confidence (KNIFE score > 0.9) circRNAs were identified by the presence of head-to-tail splice junctions using the KNIFE algorithm (Hansen, 2018). The table of circRNA counts per sample (Table S1) was filtered as follows: (a) CircRNA detected in at least 4 out of 20 samples, (b) circRNA detected in at least two diagnostic groups, and (c) the sum of reads in at least one group should be >10.

| Primer design
Primers for reference genes (RGs) and target genes were designed using Primer-BLAST (NCBI). Eight RGs were evaluated based on published stability in human brain tissue (Rydbirk et al., 2016). Primers were subjected to a 10-fold dilution series and efficiency calculated using the formula: (10 −1/slope)−1 ). UBE2D2 and RPL13 demonstrated good efficiency in these samples and were used as reference genes (Table 1).

| cDNA synthesis
cDNA was synthesized using Maxima Reverse Transcriptase (Thermo Fisher). 12 µl RNaseR-treated or vehicle-treated RNA was used to assess circularity, and 5 µg Total RNA was used for quantification.

| Agarose gel electrophoresis
cDNA was used in basic PCR reactions for visualization on an agarose gel. Reactions were performed using PCR Master Mix (Thermo Fisher) on a C1000 Touch Thermal Cycler (Bio-rad) system using
Reactions were performed in triplicate using RT 2 SYBR Green MasterMix (QIAGEN) on a CFX96 (Bio-rad) system using standard cycling conditions, (10 min at 95°C followed by 40 cycles of 15 s at 95°C, 90 s at 60°C), 0.4 µmol/L primers, and 1 µl cDNA. Control and divergent target gene primers were used for circularity confirmation. Only divergent target gene primers were used for differential expression quantification.
RT-qPCR data were analyzed using efficiency adjusted ΔΔCT ([2*efficiency] ΔCT ) and normalized to RG expression. Independent t tests were used to analyze cohort demographics and RNaseRtreated RNA enrichment. Two-way ANOVA was performed to examine the effect of pathology and cognition on circRNA expression.

| RNA sequencing
RNA sequencing was performed on RNA from 20 samples from the human MT cortex (Table S2), to a read depth of between 35 million to 52 millions reads per sample in all but one sample. 9,670 circRNAs were identified by the KNIFE algorithm as high confidence circRNAs in at least one sample, but the vast majority were identified in fewer than 3 samples, and at very low read counts (Table S1).

| PCR primer testing
To ensure accurate quantification of linear and circRNA targets, efficiency testing was first performed on each primer pair. This is particularly important for the circRNAs, where the primer must cross the head-to-tail splice junction and thus there is little choice in location for primer design. Primer pairs with efficiencies of <0.8 or >1.2 were eliminated from further experiments (Table 1). IGF2R: hsa_circ_0131235, FAM126A: hsa_circ_0008951, FGD6: hsa_circ_0099549, and TBCK: hsa_circ_0007540 were taken forwards for confirmation of circularity.

| Circular RNA validation
To confirm RNA circularity was not an artifact of cDNA synthe-

| Demographics of validation cohort
There were no significant differences in age, education, gender, or postmortem interval between groups (Figure 3). Group were divided based on their cognitive status (unimpaired vs. impaired) and AD pathology burden (Braak score; Figure 3, Table S3). As expected, there was a main effect of cognitive status on global cognition, a Z-score normalized summary score of 19 cognitive tests across multiple cognitive domains, and minimental state examination (MMSE) score. For the purpose of this study, a Braak score of IV or below was considered low pathology, and V or above considered high, as this transition point is when the MT cortex begins to exhibit significant hyperphosphorylated tau pathology (Braak et al., 2006). As expected, there was a significant main effect of pathology on global pathology score.

| Differential expression of IGF2R circRNA in MT cortex
hsa_circ_0131235 was quantified in non-RNaseR-treated RNA from the validation cohort. RT-qPCR was performed on 100 samples, and hsa_circ_0131235 was detected between 27 and 32 cycles. There was a significant main effect of pathology on hsa_circ_0131235 expression (F(1, 95) = 7.994). Mean hsa_circ_0131235 expression was increased in the MT cortex of individuals with high AD pathology relative to those with low AD pathology (Figure 4). No main effect of cognition or interactions between AD pathology and cognition was observed.

| D ISCUSS I ON
Noncoding RNAs have been shown to play a fundamental role in AD pathogenesis (Akhter, 2018), with dysregulated expression profiles of circRNA, microRNA, and mRNA identified in AD rat model hippocampi (Wang et al., 2018). A recent study published in Nature Neuroscience found that 31 circRNAs were significantly associated with Clinical Dementia Rating (CDR), nine with Braak Score, and nine with AD versus control status. Strikingly, circRNA expression accounted for more variation in CDR performance than ApoE4 status (Dube et al., 2019). hsa_circ_0131235 was not highlighted as significant in this sequencing based study, but it was not well powered to address the AD versus control pathology contrast. In our study, we confirmed hsa_circ_0131235 circularity and showed increased expression in the MT cortex in individuals with high AD pathology.
As we did not find a significant correlation with cognitive ability, hsa_circ_0131235 may have potential as a marker of AD pathology burden. It is also important to note that while we did not find clear evidence of circularity post-RNaseR digestion from the other three targets selected, this does not necessarily discount their being circular; there are reports of circRNAs that are degraded due to the presence of nicks in the sequence. Alternative approaches such as northern blotting and RNase H may be used to address this limitation of our study (Barrett & Salzman, 2016).
The "microRNA sponge" hypothesis of circRNA function suggests they inactivate corresponding microRNAs, indirectly upregulating protein expression (Lukiw, 2013;Piwecka et al., 2017). We can hypothesize a similar relationship between hsa_circ_0131235 and IGF2R. IGF2R facilitates Aß peptide clearance (Mellott et al., 2014), and expression levels have been correlated with altered functioning of lysosomal enzymes (Kar et al., 2006). Increased hsa_circ_0131235 may be part of a biological mechanism to prevent damage from Aß aggregation. IGF2/IGF2R signaling has been implicated in metabolic regulation and diabetes risk (Cianfarani, 2012) with circulating IGF2R associated with type two diabetes mellitus (T2DM) (Chanprasertyothin et al., 2015). Given diabetes is one of the strongest risk factors for the development of AD and a common comorbidity, dysregulation of IGF2 signaling through IGF2R may contribute to the neuropathology of AD.
CircRNA dynamic expression is generally distinct from linear isoforms of the same gene (Hansen et al., 2013;Piwecka et al., 2017).
While we did not quantify IGF2R linear transcript expression in MT cortex in this experiment, RNA sequencing of over 500 samples from the same ROSMAP cohort shows no differences in linear IGFR2 mRNA expression between low and high Braak stage tissue from the transcriptionally similar dorsolateral prefrontal cortex (Bihlmeyer et al., 2019;Jager et al., 2018;Kang et al., 2011). It may therefore be likely that hsa_circ_0131235 is independently regulated to the linear transcript, but our current findings cannot give us information about potential mechanisms linking the two.
As our findings are limited to the middle temporal cortex, further studies quantifying IGF2R circRNA in additional cortical regions, such as the parietal cortex are necessary to evaluate its use as a biomarker for AD and to fully understand the mechanisms of hsa_circ_0131235 in AD pathogenesis. It would also be of benefit to assess the localization of hsa_circ_0131235 with regard to both cell types and subcellular domains using in situ hybridization or Cognitively Impaired ** advanced versions of these techniques such as RNAscope (Wang et al., 2012). This would begin to address whether there is direct physical involvement of this circRNA with plaque structures or whether the mechanism of association may be more indirect.
Additional investigations into IGF2R circRNA expression in other neurodegenerative pathologies are warranted to determine the disease specificity of this finding, as well as their potential for diagnosis and therapeutics.

ACK N OWLED G M ENTS
This project was funded by SEA through institutional start-up funds.

D I SCLOS U R E S
All authors report no disclosures.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.2048.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon request.