Circular RNA circZFPM2 regulates cardiomyocyte hypertrophy and survival

Heart tissue samples for RNA sequencing were obtained from 5 HCM patients after myectomy and 5 healthy donors (Supplemental Table 1). For qPCR validation, heart tissue samples from 18 HCM patients and 18 healthy donors were utilized (Supplemental Table 2). This study was conducted with the approval of the institutional ethics committee of the Hannover Medical School, Germany, and in accordance with the guidelines from the declaration of Helsinki and its amendments or comparable ethical standards. For details, see Supplementary material online.

Neonatal rat cardiomyocytes (NRCMs), mouse atrial-like cardiomyocytes (HL-1 cells), human induced pluripotent stem cells (hiPSCs) (hHSC_Iso4_ADCF_SeV-iPS2, alternative name: MHHi001-A) [18], HCM patient-derived iPSCs [32] and human embryonic kidney cells (HEK293) were cultured according to standard protocols. HiPSCs were differentiated into cardiomyocytes by modulating the Wnt pathway as previously described [27]. Cell treatments, transfections, and cellular assays were conducted as indicated in the Supplementary material online.

Mouse, rat and human circZFPM2 were knocked down using a siRNA-mediated strategy. For overexpression experiments in NRCMs, the rodent circZFPM2 sequence with flanking circularization elements was cloned into the AAV MCS 1.3 plasmid using the restriction enzymes EcoRI and XbaI. For human overexpression studies, circZFPM2 was in vitro transcribed using the RiboMax large RNA production system (Promega) with a modified protocol. For details, see Supplementary material online.

Total RNA was isolated with QIAzol (Qiagen) according to the manufacturer’s protocol, of which 500–1000 ng were reverse transcribed using the Biozym cDNA Synthesis Kit with random hexamer primers according to the manufacturer’s protocol and assessed by quantitative real-time PCR (qRT-PCR) (Absolute Blue qPCR SYBR Green Mix, Thermo Scientific). For details, see Supplementary material online.

To identify circRNAs that participate in the pathogenesis of HCM, we performed global circRNA profiling in cardiac tissue from healthy donors and HCM patients (Supplemental Table 1). 1020 circRNAs were detected, of which 110 (Supplemental Table 3) were significantly differentially expressed (p < 0.05 and fold change > 1.5) between the two groups (Fig. 1A). By conducting in silico and experimental validation of the 30 most differentially expressed and most abundant circRNAs, we identified circZFPM2 as the most promising candidate for further validation. In detail, in silico validation was performed using the integrative genomics viewer (Supplementary Figure S1A). CircRNAs for which the sequencing reads did not cover the backsplice site, were excluded. Furthermore, circZFPM2 was the only candidate, which was not detectable in hiPSCs, but which became abundantly expressed during the differentiation into hiPSC-CMs, suggesting CM relevant functions (Supplementary Figure S1B). CircZFPM2 derives from exon 2 and 3 of the ZFPM2 locus, which codes for a zinc finger protein that is essential for cardiogenesis and hematopoiesis in mammals [8, 53]. CircZFPM2 is annotated in the human and mouse genome (has_circ_0003380 and mmu_circ_0000597; circBase.org). By performing sequence alignment using NCBI standard nucleotide BLAST, we further detected a high degree of sequence homology in the rat and pig genome (Fig. 1B). We experimentally validated the correct backsplice regions in these four species by RT-qPCR using specific divergent primers and Sanger sequencing (Fig. 1C). Increased resistance towards RNase R treatment of RNA isolated from murine left ventricular tissue revealed circular RNA configuration (Fig. 1D). This was corroborated by prolonged stability of circZFPM2 compared to linear ZFPM2 after treatment of NRCMs with the transcription inhibitor actinomycin D (Fig. 1E).

For further characterization of circZFPM2, we evaluated its expression profile in a mouse organ panel revealing highest expression in the heart (Fig. 2A), which gradually increases with age, independent of the linear ZFPM2 (Fig. 2B). In a collection of different human cell types, circZFPM2 is most abundant in hiPSC-CMs (Fig. 2C), while on the subcellular level, it is primarily localized in the cytoplasm (Fig. 2D).

Next, we aimed to validate the initial RNA sequencing data revealing a 2.5-fold upregulation of circZFPM2 in human HCM heart tissue, in further HCM models. First, we utilized HCM patient-derived iPSC-CMs carrying the disease causing point mutation R723G in the MYH7 gene [47]. Consistently, circZFPM2 was significantly upregulated in the diseased cells compared to the wild type control hiPSC-CMs, which is in line with increased expression of the hypertrophy marker genes BNP and ANP (Fig. 2E). Importantly, we confirmed the significantly increased expression of circZFPM2 in independent biopsy material from HCM patients compared to healthy donor heart tissue (Fig. 2F). Interestingly, the expression of circZFPM2 in other in vivo and in vitro cardiac disease models such as myocardial infarction (MI), hypoxia exposed NRCMs or TGFβ treated human cardiac fibroblasts, remained unchanged, suggesting HCM-specific regulation of this circRNA candidate (Supplementary Figure S2 A–C).

In conclusion, circZFPM2 is a cardiomyocyte-enriched, cytoplasmic circRNA that is upregulated in HCM.

To elucidate whether circZFPM2 (dys)regulation has functional consequences, we first performed loss-of-function experiments in NRCMs. We designed a specific siRNA targeting the backsplice region of circZFPM2 resulting in efficient knockdown without affecting the expression of the linear host gene (Fig. 3A, B).

Knockdown of circZFPM2 led to significantly increased expression of the hypertrophy marker genes BNP and ANP (not significant for MCIP) (Fig. 3C). In line with these findings, lack of circZFPM2 induced cellular hypertrophy in NRCMs compared to the scrambled siRNA treated control cells (Fig. 3D).

Since HCM is associated with altered cell metabolism, we conducted a Seahorse XF Cell Mito Stress assay after circZFPM2 knockdown to test for a potential mitochondrial functional impairment [42]. Indeed, the oxygen consumption rate (OCR) of NRCMs was reduced after circZFPM2 silencing, most pronounced was the spare respiratory capacity (Fig. 3E, F). Interestingly, either as a cause of or as a consequence for mitochondrial impairment, we detected an elevated production of reactive oxygen species (ROS) after circZFPM2 knockdown (Fig. 3G). Consequently, circZFPM2 knockdown increased cytotoxicity (Fig. 3H), caspase activity (Fig. 3I) and the number of apoptotic cells (TUNEL⁺; Fig. 3J), collectively leading to decreased cell viability (Fig. 3K) of NRCMs.

To test whether circZFPM2 overexpression might exert opposite, i.e. beneficial effects compared to the knockdown, we designed a circZFPM2 overexpression cassette in an AAV backbone, harboring the circRNA sequence flanked by intronic sequences containing inverted repeats (ALU elements) to facilitate the circularization (Fig. 4A) [31]. The transfection of this plasmid into the cardiomyocyte cell line HL-1 was already sufficient to induce a significant upregulation of circZFPM2, without influencing the linear counterpart (Fig. 4B). We next produced and transduced AAV6-circZFPM2 viral particles which led to a robust and stable overexpression of circZFPM2 in NRCMs, while the host gene expression remained unchanged (Fig. 4C).

In order to test circZFPM2 effects in combination with a hypertrophic stimulus, we treated NRCMs with leukemia inhibitory factor (LIF). Simultaneous overexpression of circZFPM2 blunted the expression of hypertrophy marker genes (significant only for MCIP) (Fig. 4D). Nevertheless, circZFPM2 overexpression fully rescued the LIF induced cellular hypertrophy to control levels (Fig. 4E).

In contrast to the previous knockdown experiments, circZFPM2 overexpression reduced cellular ROS production (Fig. 4F). Already at baseline, AAV6-circZFPM2 treatment reduced cytotoxicity and caspase activity. When NRCMs were additionally exposed to the cardiotoxic anti-cancer agent doxorubicin (0.25 µM), circZFPM2 partially rescued the cytotoxic effects and cell viability (Fig. 4G–I) [35].

In conclusion, overexpression of circZFPM2 rescues NRCM function including reduced hypertrophy, improved mitochondrial performance, and cell viability.

To extend the clinical value of our studies, we conducted loss- and gain-of-function experiments in human iPSC-CMs [34]. Similar to the experiments performed in NRCMs, we designed species-specific siRNAs targeting the backsplice site to knock down the human circZFPM2, without affecting host gene expression (Fig. 5A). To induce hypertrophy in hiPSC-CMs, cells were treated with ET-1 instead of PE or LIF, as performed for NRCMs. The pro-hypertrophic stimulation with ET-1 caused the most consistent hypertrophic response in hiPSC-CMs compared to PE or LIF, as evident by increased expression levels of several hypertrophic marker genes. CircZFPM2 silencing increased hypertrophic marker gene expression at baseline and showed additive effects when cells were treated with pro-hypertrophic stimuli (Fig. 5B). Metabolic impairment after circZFPM2 knockdown, as determined by Seahorse XF Cell Mito Stress assay, was even more pronounced in the human cardiomyocytes compared to rodents (Fig. 5C–E). Congruent with the effects seen in NRCMs, ROS production was enhanced after siRNA transfection (Fig. 5F) and, in turn, cytotoxicity and caspase activity were both significantly increased in hiPSC-CMs (Fig. 5G, H).

In summary, these experiments suggest that circZFPM2 is not only sequence, but also functionally conserved across species including in humans.

To provide further translational evidence, we next aimed to investigate therapeutic effects of circZFPM2 by means of in vitro transcribed (IVT) circRNA treatment. We transcribed the linear human circZFPM2 sequence in vitro and circularized the RNA with the help of a backsplice site-complementary DNA splint and the T4 DNA ligase (Fig. 6A). This yielded IVT circZFPM2 of the correct size and with the correct backsplice sequence, which consequently showed robust, stable, and dose-dependent expression levels after transfection into hiPSC-CMs (Supplementary Figure S3A–D).

Functionally, IVT circZFPM2 significantly reduced hypertrophy marker gene expression at baseline and after hypertrophy induction (Fig. 6B, C). For the assessment of metabolic alterations and cell fitness, we stressed hiPSC-CMs with doxorubicin (1 µM) 24 h after IVT circZFPM2 transfection and incubated for additional 48 h before cell assays were performed (Fig. 6D). Similar to viral overexpression in rodents, ROS production was decreased through IVT circZFPM2 treatment (Fig. 6E). While cytotoxicity was significantly rescued, caspase activity remained unchanged (Fig. 6F, G).

These results suggest that treatment with IVT circZFPM2 confers comparable cardioprotective effects in hiPSC-CMs compared with AAV6-mediated overexpression in NRCMs.

First, we performed an siRNA-mediated knockdown of circZFPM2, which did not lead to increased hypertrophic marker gene expression, neither at baseline, nor in combination with pro-hypertrophic stimuli (Fig. 7A). Nevertheless, circZFPM2 silencing elevated caspase activity under basal condition (Fig. 7B) and further increased the cytotoxic effects of doxorubicin (Fig. 7C).

Strikingly, treatment of HCM-CMs with IVT circZFPM2 led to a drastic reduction of hypertrophic marker genes (Fig. 7D) concomitant with partially rescued cytotoxicity and caspase activity (Fig. 7E, F).

In line with our AAV6 in vitro overexpression model in non-diseased CMs, transfection of IVT circZFPM2 additionally reduced both apoptosis and cellular hypertrophy, while increasing cell survival in HCM-derived hiPSC-CMs.

Furthermore, in order to demonstrate a more translational therapeutic approach by overexpressing circZFPM2 in a multi cellular 3D cardiac model, HCM-CMs were transduced with AAV6-circZFPM2 or a AAV6-GFP control. Together with ADSCs, HCFs, and HUVECs, the transduced HCM-CMs were assembled into HCM human cardiac organoids (HCM-hCOs) to investigate whether overexpression could protect against cardiac dysfunction using both a structural and functional miniature 3D cardiac replication of HCM.

HCM-hCOs transduced with a GFP control virus exhibited homogenous distribution of GFP-positive HCM-CMs, illustrating a high transduction efficiency for our 3D AAV6 overexpression model (Fig. 8A). Morphologically, HCM-hCOs transduced with circZFPM2 exhibited improved circularity as well as reduced size, as evident by a reduced perimeter and diameter compared to control organoids (Fig. 8B–F). Functionally, HCM-hCOs overexpressing circZFPM2 showed improvement of several contractile parameters; reduced time-to-peak (Fig. 8G), relaxation time (Fig. 8H, N), 90-to-90 transient (Fig. 8I, N), 50-to-50 transient (Fig. 8J, N) and 10-to-10 transient (Fig. 8K, N). However, neither the contraction amplitude (Fig. 8L, N), nor the peak-to-peak time (Fig. 8M, N), were effected by elevated circZFPM2 levels in HCM-hCOs.

Finally, to decipher the role of circZFPM2 signaling, we performed RNA sequencing of total RNA from hiPSC-CMs, in which we silenced circZFPM2 with our backsplice site-specific siRNA.

We observed 72 dysregulated genes (P_adj ≤ 0.05, FC ≥ 1.5) in circZFPM2-silenced hiPSC-CMs compared to scrambled siRNA (Fig. 9A, B). Gene set enrichment analysis using the KEGG database resulted in the enrichment of the “Cardiac muscle contraction” signaling pathway (Fig. 9C), and additionally, the GO terms, “Positive Regulation of Mitochondrial Membrane Potential” (GO:0010918), “Mitochondrial Inner Membrane” (GO:0005743) as well as “Mitochondrial Membrane” (GO:0031966) (Fig. 9D, E) were enriched in our RNA sequencing dataset. In line with our Seahorse data, which indicate impaired mitochondrial respiration (Fig. 3E, F and Fig. 5C–E), these results suggest a potential mode of action of circZFPM2 in HCM pathology via the modulation of oxidative phosphorylation in CMs. Furthermore, we aimed to identify potential interaction partners of circZFPM2 via RNA pulldown experiments. To do so, a probe complementary to the backsplice site of circZFPM2 and a scrambled probe as a control was designed and RNA pulldowns performed from hiPSC-CM cell lysates. Next, the samples were analyzed via mass spectrometry and peptide profiles were searched against human entries of uniprot database. From this dataset, we identified 4001 protein groups, of which 472 were quantified in circZFPM2 probe samples. Further analysis revealed 10 significantly enriched proteins (p_adj ≤ 0.05, FC ≥ 2) with PYGM, UTS2 and IMMT being the most meaningful candidate as potential interaction partners for circZFPM2 in CMs (Fig. 9F). PYGM is a key enzyme in glycogenolysis and thereby essential in providing sufficient energy for the contraction of muscle cells [33]. UTS2 has already been linked to several diseases, including cardiac fibrosis, cardiac hypertrophy, and heart failure [39]. It has been shown to effect cardiac contractility, cardiac hemodynamics, as well as induce cardiomyocyte hypertrophy [20, 39, 43]. IMMT is part of the MICOS-complex and located at mitochondrial cristae junctions [54]. Knockdown of IMMT results in structural damage of mitochondria, increased ROS production, and compromised oxidative phosphorylation [54, 60]. Moreover, despite not meeting the criteria of a fold change ≥ 2, other proteins involved in key mitochondrial or sarcomeric functions such as DNM1L and LDB3 were significantly enriched by our specific circZFPM2 probe. DNM1L contributes to mitochondrial fission, and muscle-specific loss-of-function experiments demonstrated functionally abnormal mitochondria, muscle wasting, and weakness [13, 50]. Further, LDB3 stabilizes sarcomere structures in cardiac muscle by binding α-actinin, and mutations within LDB3 have already been associated with various types of cardiomyopathies [4, 29, 55, 58].

Taken together, these data further suggest an effect of circZFPM2 on cardiac muscle contraction and mitochondrial function within cardiomyocytes in the context of HCM.

Despite the demonstration of clear functional and therapeutic effects of circZFPM2 in cardiac hypertrophy, some limitations of the present study should be noted. Firstly, the precise underlying molecular mechanisms of circZFPM2 in cardiomyocytes remain unknown and therefore these beneficial therapeutic effects warrant further investigation. So far, our proteomics analysis from RNA pulldown experiments identified binding partners for circZFPM2 that are likely to mediate circZFPM2 beneficial effects, which have yet been verified in upcoming studies. Secondly, it was highly interesting that we observed stronger therapeutic effects of circZFPM2 in HCM patient-derived CMs. Although ideally, these experiments need to be performed with isogenic corrected HCM lines instead of in wild type hiPSC-CMs as the control group. Thirdly, despite using state of the art 2D in vitro cardiomyocyte models by utilizing NRCMs and hiPSC-CMs, as well as implementing a 3D model of HCM-derived human cardiac organoids, our study lacks information about the role of circZFPM2 in an HCM in vivo model. However, as the cardiac organoids are comprised of human cells, they add a translational aspect lacking in an in vivo mouse model, while maintaining the 3D structure and functional aspects of a whole heart organ system, with the added benefits of an observable HCM phenotype seen through the contraction measurements. Owing to the high sequence and functional conservation of circZFPM2, in vivo studies in large animals such as pigs are also eventually warranted for any future pre-clinical testing.