LncRNA CISTR-ACT regulates cell size in human and mouse by guiding FOSL2

Abstract

Organisms regulate cell size and shape to function efficiently. Aberrant cell morphogenesis is commonly associated with disease, yet gene-regulatory mechanisms remain unknown. CISTR-ACT was the first lncRNA involved in inter-chromosomal proximities and Mendelian disease, and it is associated with mean corpuscular volume (red blood cell size). Here, functional dissection of CISTR-ACT’s DNA- and RNA-encoded mechanisms by in vitro and in vivo perturbations reveals that CISTR-ACT regulates cell size across cell types and species. CISTR-ACT’s locus is embedded in a stable inter-chromosomal environment which contains cell size genes that are regulated by CISTR-ACT in trans. CISTR-ACT’s RNA also has function and directly interacts with transcription factor FOSL2 to guide its regulation of cell morphogenesis and cell-cell adhesion genes. In the absence of CISTR-ACT, the FOSL2-chromatin binding is perturbed. Our study exemplifies how a functionally conserved lncRNA regulates cell size with multiple modes of action and ultimately contributes to clinically relevant phenotypes.

Establishment and maintenance of appropriate cell size is a prerequisite for cells to function efficiently. Here, Kiriakopulos et al. reveal that the lncRNA CISTR-ACT maintains cell size across cell types in humans and mice by regulating cell morphogenesis genes in trans via guidance of the transcription factor FOSL2.

Introduction

The human genome contains more long non-coding RNAs (lncRNAs) than protein-coding genes (GENCODE v49)¹ which regulate genes and chromatin scaffolding². LncRNAs function at the DNA³, transcriptional^4,5, and/or RNA levels^6–8. Adding to their complexity, lncRNA loci are often not conserved^9,10, although their functions can be conserved^4,5,9,11,12. Thus, mechanistically dissecting lncRNA function is paramount, yet challenging, and requires a combination of in vitro and in vivo models to fully elucidate operating principles at cellular and organismal scales^11,13.

CISTR-ACT (CISTR) was the first lncRNA gene associated with autosomal dominant Mendelian disease and chondrogenesis¹⁴. Specifically, independent familial translocations led to positional effects that disrupted CISTR’s spatial proximity in cis with PTHLH (chromosome 12) and in trans with the transcription factor (TF) SOX9 (chromosome 17), causing brachydactyly¹⁴. CRISPR live-cell imaging validated stable CISTR-SOX9 contacts across spatiotemporal dimensions^15,16, and a recent compendium of Hi-C contacts determined frequent CISTR-SOX9 trans-chromosomal contacts across cell types¹⁷. Moreover, a computational screen validated CISTR as a top-ranked activator of gene expression and confirmed that it can regulate genes via chromatin looping¹⁸.

Here, we investigated CISTR’s molecular mechanism to understand how it impacts gene regulation. We find evidence that CISTR provides the genetic basis as a causal lncRNA gene for regulating cell size by guiding FOSL2, which broadens the general understanding of clinically relevant cell size phenotypes and lncRNA modes of action.

Results

CISTR functions at the DNA and RNA levels

To uncover CISTR’s function, we first explored genetic associations and find that lead variant rs938729214 [A] 11.64 kb 5′ of CISTR strongly associates with mean sphered corpuscular volume (MSCV [red blood cell size], rank 13/821, beta = 0.4843, p = 2×10⁻⁹) in a UK biobank study¹⁹ (Fig. 1a, b). This indicates that CISTR may be associated with cell size control. Considering increased MSCV and mean corpuscular volume (MCV) in routine blood counts are indicative of various pathologies (i.e., anemias, vitamin B9/12 deficiencies, acute myeloid leukemia, liver & kidney dysfunction, myelodysplastic syndrome, unknown etiology: 35-41%) and underlying mechanisms are unknown^20–27, we started to dissect CISTR’s mechanism.

Fig. 1. CISTR’s DNA- and RNA-encoded function regulates cell size.

a Variant effect size of rs938729214 at CISTR’s locus is highlighted for the MSCV trait¹⁹. b CISTR locus with flanking genes, ENCODE cCREs, isoforms, conservation, ReMap TF binding, and CpG island. c CRISPR approaches dissect DNA, transcription, and RNA function of CISTR. Schematics of (d) CRISPR/Cas9 (red, WT n = 5 [2 independent clones, 3x or 2x passages], KO n = 5 [2 independent clones, 3x or 2x passages], intron KO n = 2 [2 independent clones]), (e) CRISPR/CasRx (cyan, control n = 3, CasRx n = 3 independent replicates), and (f) CRISPR-Display (purple, KO n = 3, ectopic n = 2, overexpression n = 2 [2 independent experiments]) with corresponding CISTR RPKM box plots. For all box plots: central boxplot line represents the median. Box limits represent upper and lower quartiles. g. Overlap of RNAseq results between CRISPR experiments (q < 0.01, log₂FC ≥ 1 and ≤ −1). h Pearson’s correlations of gene expression changes. Black datapoints indicate lower log₂FC in ΔCISTR cells, whereas red indicates higher log₂FC values in ΔCISTR cells. Heat maps of the log₂FC values of overlapping DEGs between (i) ΔCISTR and ΔCISTR_intron experiments, depicting DNA-related DEGs, and (j) ΔCISTR and overexpression experiments, depicting RNA-rescued genes. k Flow cytometry cell size measurements by forward scatter areas (FSC-A). Two-sided Welch’s t-testing determined significance. RPMI 2650 (n > 5482 cells, 2–3 replicates, p < 1×10⁻¹⁵), C-28/I2 (n > 873 cells, 3 replicates, p < 8.543×10⁻⁷). Two-way ANOVA was used for K562 cells (p < 1.8×10⁻¹², n > 6571 cells across 2 replicates). Quartiles with medians are shown. l Flow cytometry cell size measurements across the cell cycle in WT and ΔCISTR cells. Asterisks depict significance determined by two-sided Welch’s t-testing (p < 1×10⁻¹⁵, n > 13271 cells across 2 replicates). m Cell cycle analysis of WT and ΔCISTR RPMI 2650 cell populations in G₀/G₁, S, or G₂/M phase (cell numbers are identical to l). Asterisks depict significance (p < 2.2×10⁻¹⁶) determined by one-sided χ²-test. n Difference in proliferation rates between WT and ΔCISTR RPMI 2650 cells over time. Source data are provided as a Source Data file.

We chose human epithelial RPMI 2650 cells²⁸ for perturbation experiments because they robustly express CISTR (RPKM = 3.229), and we focused on isoform NR_104333.1 due to its link to brachydactyly¹⁴ and its high expression in human brain and heart (Fig. 1b, Supplementary Fig. 1a). To assess both DNA- and RNA-encoded mechanisms known for lncRNAs¹³, we deleted CISTR’s locus (ΔCISTR; Fig. 1c, d, Supplementary Fig. 1b-e), and the most conserved intronic region with highly dense ChIPseq peaks (ReMap)²⁹ and a CpG island (Fig. 1b), which could indicate CISTR’s core gene-regulatory element (ΔCISTR_intron, Fig. 1c, d, Supplementary Fig. 1b-d). To examine whether CISTR’s RNA has function, we applied two strategies: we degraded its transcripts by CasRx³⁰ (Δ: 87%; Fig. 1c, e, Supplementary Fig. 1d, f), and we overexpressed CISTR in the ΔCISTR background and ectopically localized it to regions flanking ΔCISTR by CRISPR-Display³¹ (Fig. 1c, f, Supplementary Fig. 1d, g). We then applied RNAseq on all samples.

Expression of genes directly flanking CISTR (CALCOCO1, HOXC13-AS) is unaffected in the CRISPR perturbation experiments (Supplementary Fig. 1e, Supplementary Data 1–5). However, we find thousands of strongly differentially expressed genes (DEGs) upon CISTR’s modulation when compared to controls ( > 1000-fold, q < 0.01, log₂FC ≥ 1 and ≤ −1; ΔCISTR: 2908 DEGs, CasRx: 3179 DEGs, ectopic CISTR: 1758 DEGs, CISTR overexpression: 1705 DEGs; Fig. 1g; Supplementary Fig. 1h, i; Supplementary Data 1–5). Since these effect sizes are surprising for perturbing a lncRNA gene, we compared independent isogenic ΔCISTR clones and find high inter-replicate correlation (Pearson’s r = 0.869, Supplementary Fig. 1h). Across all perturbation conditions, DEGs significantly overlap (hypergeometric tests p < 1.667×10⁻²¹³), and enrichment analysis of DEGs³² repeatedly points to cell-cell adhesion (GO:0098609), cell junction organization (GO:0034330), supramolecular fiber organization (GO:0097435), and cell morphogenesis (GO:0000902), and highlights bone marrow and midbrain cell type signatures (Fig. 1g and Supplementary Fig. 2).

Notably, the overlapping DEGs from ΔCISTR and ΔCISTR_intron experiments suggest function for CISTR’s DNA, whilst DEGs from CasRx, CISTR overexpression, and CRISPR-Display experiments propose a functional role for CISTR’s RNA (Fig. 1g, Supplementary Fig. 1i). We find 1316 overlapping DEGs between ΔCISTR and ΔCISTR_intron with highly correlated expression changes (log₂FC, Pearson’s r = 0.72, p < 0.001, Fig. 1h, i, Supplementary Fig. 1j), identifying CISTR’s conserved intron as a core regulatory DNA element (Fig. 1h). The marginal correlation of expression changes among overlapping DEGs between ΔCISTR and CasRx (754 genes, r = 0.04) indicates that CISTR’s DNA is the major driver of gene expression (Fig. 1h).

Our functional dissection strategy allowed us to determine the rescue potential of CISTR’s RNA by comparing between ΔCISTR and overexpression conditions. Log₂ fold-changes of overlapping DEGs between these perturbations negatively correlate (r = −0.36, p < 0.001, Fig. 1h), suggesting an RNA-driven rescue effect of gene expression by CISTR transcripts. Remarkably, 319 DEGs are oppositely regulated between the ΔCISTR and overexpression approaches (Fig. 1j, q < 0.01, log₂FC ≥ 1 and ≤ −1), thereby assigning rescue function to CISTR’s RNA. Thirty-seven percent of these RNA-rescued DEGs (118/319) overlap with DNA-related DEGs from the deletion experiments (Supplementary Fig. 1k), indicating that CISTR’s DNA and RNA synergistically regulate these genes. The ectopic CISTR and CISTR overexpression conditions yield highly correlated results (Pearson’s r = 0.98, p < 0.001, 63% overlap, Supplementary Fig. 1l, m), suggesting CISTR’s RNA acts independently of its localization. Collectively, we find that CISTR strongly activates many cell-cell adhesion and cell morphogenesis genes by DNA- and RNA-driven effects, which may align with the MSCV trait.

CISTR perturbations alter cell size across cell types

CISTR provides an interesting opportunity to address whether it regulates cell size because cytoskeletal organization and cell adhesion have been associated with cell structure and size changes^33–35. Understanding gene-regulatory mechanisms for cell size is paramount to reveal how deregulation of cell growth control affects a variety of diseases³⁶. To do this, we measured cell size in ΔCISTR, ΔCISTR_intron, and CasRx knockdown cells and find that they are all significantly larger than wildtype cells (WT, Welch’s t-tests p < 1×10⁻¹⁵; Fig. 1k). CISTR’s overexpression in WT decreases cell size, whilst its overexpression in ΔCISTR cells partially rescues cell size (Fig. 1k). This aligns with the RNA-mediated rescue of cell junction and morphogenesis genes upon CISTR’s overexpression (Supplementary Fig. 2). Considering that both ΔCISTR_intron cells (DNA function) and CasRx-perturbed cells (RNA function) are larger, cell size seems to be dependent on both CISTR’s DNA and RNA levels.

CISTR’s MSCV association motivated us to functionally test cell size in K562 cells, which can be differentiated into erythroid precursors with hemin³⁷ (Supplementary Fig. 1d, n). Undifferentiated and hemin-differentiated ΔCISTR K562 have significantly increased cell sizes in comparison to WT (ANOVA p < 2×10^-16, n = 3; Fig. 1k), whilst CISTR’s overexpression decreases cell size in WT. Upon CISTR’s overexpression, ΔCISTR cells show a partial rescue of cell size (ANOVA p < 2 × 10⁻¹⁶, n = 3; Fig. 1k). The DEGs from ΔCISTR K562 cells point to GO term actin filament-based process (i.e., FLNB, FSCN1, GPC6, SEMA3A, SEMA3F, SPTBN1, TUBB2A, TUBB2B, and VCAN, Supplementary Data 6, 7). Of note, CISTR is neither expressed in K562 cells nor blood³⁸, thus, the MCV phenotype is likely DNA-specific. However, overexpression of CISTR in ΔCISTR K562 cells partially rescues cell size enlargements (ANOVA p = 1.8×10⁻¹², n = 3; Fig. 1k), which again shows that both DNA and RNA regulate cell size synergistically in cell types that normally express CISTR.

We also addressed if CISTR regulates size of human C-28/I2 chondrocytes³⁹, due to its involvement in chondrogenesis¹⁴. ΔCISTR C-28/I2 cells and CISTR’s overexpression both significantly alter size (Fig. 1k, Welch’s t-tests p < 8.543×10⁻⁷), although CISTR is not expressed in C-28/I2 (Supplementary Fig. 1n-p). RNAseq determines a small, but significant overlap of 18 DEGs between ΔCISTR C-28/I2 and RPMI 2650 cells (hypergeometric test p = 1.495×10⁻⁸). These DEGs include SEMA3A⁴⁰, ARHGAP28⁴¹, and ITGA4⁴² that function in the context of the cell morphogenesis phenotypes that we observe across ΔCISTR cells. Downregulation of SOX9 and other chondrogenesis and osteogenesis genes (i.e., BMP4, MMP2, MMP16, VCAN^43–45) validates previous results¹⁴ (Supplementary Fig. 1d, n-p, Supplementary Data 8). These findings replicate K562 results and reiterate that CISTR’s genomic locus functions as a gene-regulatory element.

Cell size and cell cycle regulation are interconnected processes⁴⁶. Thus, we conducted cell cycle and proliferation assays and find that ΔCISTR cells are significantly larger across the cell cycle with largest differences in G₁ (Fig. 1l, Welch’s t-testing p < 1×10⁻¹⁵), and 10.9% more ΔCISTR cells are in G₁ when compared to WT (Fig. 1m, Supplementary Fig. 1q). This is supported by a decreased proliferation rate of ΔCISTR cells (Fig. 1n), and higher CISTR expression in G₁ (Supplementary Fig. 1r). Collectively, our findings indicate that CISTR DNA and RNA levels regulate cell size and cell morphogenesis genes together, and that CISTR’s effect on cell size may be linked to regulation of cell cycle progression at the G₁/S phase transition and proliferation rate.

Cistr^−/− mice have enlarged erythrocytes

Since CISTR’s locus is conserved in mice, we next used genetic ablation to examine Cistr’s physiological relevance on cell size at the organismal scale. Notably, blood counts of Cistr^+/- and Cistr^−/− mice show significantly increased MCV and lower hemoglobin levels in both sexes when compared to WT (Mann Whitney tests, range p < 0.0313; Fig. 2a). This anemic condition²⁶ coincides with various shapes of erythrocytes (i.e., anisocytosis [different sizes] and poikilocytosis [different shapes]; Fig. 2b), but normal hematopoietic cell populations (Supplementary Fig. 3, 4a). DEGs in blood and bone marrow of Cistr^−/− mice relate to hematopoiesis, erythrocyte homeostasis, and again cytoskeletal organization (i.e., Acta1, Actg1, Aqp1, Capg, Flna, Fscn1, Krt80, Rhog, Shank3, Sptbn2, Tln1, Tpm2, Ttn, Vasp, Was, etc.; Supplementary Fig. 4b, c, Supplementary Data 9–12). These results align with the human MSCV trait association, replicate Cistr’s DNA-encoded function in K562 cells, and propose a conserved mechanism of action across species.

Fig. 2. Cistr mice have organismal cell size phenotypes in blood and brain.

a Hematology of Cistr mice in comparison to WT, separated by genotype and sex. MCV – mean corpuscular volume, MCH—mean corpuscular hemoglobin / concentration [MCHC]. Statistical significance determined by two-tailed Mann Whitney tests (p < 0.0313; [*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001]), n = 8. b Blood films of WT and Cistr^−/− mice. Arrowheads exemplify abnormally shaped eythrocytes. Scale bars = 2 µm. c Whole mount in situ hybridizations of Cistr expression in mouse and rat embryos. d Absolute brain volume analysis of MRI exemplified by four different brain slices. Color overlays on top of wildtype brain indicate regions of significant volumetric changes (p < 0.05) between Cistr^−/− female mice and WT. e Absolute brain volume changes of 19 brain structures with significant genotypic differences (p < 0.05) across both sexes, depicted in percent changes (%). PSC – primary somatosensory cortex. f RNA-FISH of Cistr in WT piriform cortex and olfactory tubercle (P7). Gad2 marks inhibitory neurons; pl—pyramidal layer, ml—molecular layer, pol—polymorph layer, scale bars = 50 µm. g Overlap of RNAseq results in female and male Cistr^−/− cerebellum, cortex, hippocampus, and hypothalamus relative to WT. h Brain immunofluorescence with NeuN, GFAP, and DAPI; pc—piriform cortex, ot—olfactory tubercle, vn—ventral nucleus. Right: layers of piriform cortex described in (e). Scale bars = 500 and 50 µm, respectively. i Neuronal cell body sizes in the polymorph layer (pol) of the piriform cortex in micrometer (µm, F WT n = 200, F KO n = 168, M WT n = 200, F KO n = 197 cells, each condition 2 animals). The central boxplot line represents the median, box limits represent upper and lower quartiles, and asterisks depict significance (**** p = 8.94×10⁻⁸, ** p = 0.00586), determined by two-tailed Mann-Whitney tests. j Flow cytometry forward scatter area (FSC-A) measurements in WT and ΔCISTR hESC-derived Neural Progenitor Cells (NPCs, WT n = 82,982, KO n = 56,460 cells, across 3 passages). Two-sided Welch’s t-tests determined significance (p < 1×10⁻¹⁵). Quartiles with medians are shown. k Orthologous DEGs across experiments. Asterisks indicate previously reported CISTR targets¹⁴. Source data are provided as a Source Data file.

Cistr^−/− mice also have enlarged cortical regions

Cistr/CISTR is strongly expressed in rodent and human brains, respectively (Fig. 2c, Supplementary Fig. 5a). Thus, we next explored potential neuroanatomical phenotypes in 9-week-old Cistr^−/− mice, after all brain structures are fully developed^47,48. In vivo manganese-enhanced magnetic resonance imaging (MEMRI)⁴⁹ displays structures (19/185, > 10% of total structures), primarily in cortical regions, with significant genotypic volume changes that were modulated by sex (p < 0.05; Fig. 2d, Supplementary Fig. 5b-d, Supplementary Data 13), indicating a female-biased function for Cistr in mouse brains. Specifically, cingulate and somatosensory cortical areas implicated in behavioral control and integration of sensory and motor signals^50,51 are up to 12.5% larger in female Cistr^−/− mice compared to wildtype, whilst equivalent male regions are up to 2.45% smaller (10 animals each condition, p < 0.05; Fig. 2d, e, Supplementary Data 13). However, neurobehavior is normal among the tested genotypes (Supplementary Figs. 6–8; Methods), suggesting that the volumetric changes do not impair these functions. smRNA-FISH determined Cistr expression early in brain development (P7 and P28 WT mice) in regions of morphological changes (i.e., piriform cortex, olfactory bulb) observed in Cistr^−/− mice (Fig. 2f, Supplementary Figs. 9, 10a).

Cistr^−/− mouse cortex, cerebellum, hippocampus, and hypothalamus show DEGs enriched in brain development, cytoskeletal regulation, and cell projection organization in RNAseq (i.e., Actg1, Actn1-3, Fhod3, Fscn1, Rhof, Rnd1-3, Sema5a, Sptan1, Sptbn1-2 & 4, Synpo2, Tuba4a, Tuba1b, Tubb4b, Vasp, etc.; Fig. 2g, Supplementary Fig. 10b-e, Supplementary Data 14–21). Female Cistr^−/− brain samples have more DEGs than males in comparison to WT, where female-specific upregulation of neurogenic genes (i.e., Lhx8, Drd1-3) may relate to the female-biased phenotypes (Fig. 2g, Supplementary Data 14). Allen Brain Atlas data (P56)⁴⁷ confirmed that DE cytoskeletal regulators and cortical genes in Cistr^−/− mice of both sexes are expressed in the piriform cortex and olfactory tubercle (i.e., Rnd2⁵², Esam⁵³, and Stk26 [Mst4]⁵⁴, Foxp2⁵⁵, Ctip2 [Bcl11b]⁵⁶, Supplemental Fig. 11a–g). To elucidate reasons for the enlarged Cistr^−/− cortices, we measured neuronal cell bodies in the piriform cortex of adult brains (Fig. 2h). NeuN⁺ neurons in the polymorph layer of the piriform cortex are significantly larger in female Cistr^−/− mice compared to WT (n = 400, four animals, Mann-Whitney tests range p < 0.00586; Fig. 2i, Supplemental Fig. 11h–k).

This motivated us to delete CISTR in CISTR-expressing hESC-derived neural progenitor cells (NPCs) to further determine if CISTR/Cistr has conserved function. ΔCISTR NPCs are significantly larger in comparison to WT (Welch’s t-test p < 1×10⁻¹⁵, Fig. 2j), with upregulated HOXA-D genes and DEGs related to cell morphogenesis (Supplementary Fig. 11l, m, Supplementary Data 22). Despite CISTR being involved in human chondrogenesis¹⁴ and SOX9 regulation, Cistr^−/− mouse skeletons and metacarpals are normal (Supplementary Fig. 12). CISTR’s strong cardiac expression (Supplementary Fig. 5a) translated into a mild heart phenotype in Cistr^−/− mice with significant differences in the QT interval in female animals, and cardiac output in males (Supplementary Fig. 13).

Collectively, brain regions expressing Cistr (P7, P28) show developmental phenotypes in adult Cistr^−/− mice with dysregulated cytoskeletal and cortical genes. Our in vivo results replicate previous findings that lncRNA phenotypes can be subtle, highly variable, and context-dependent in different sexes and species^57,58.

CISTR’s function is conserved

All in vivo and in vitro experiments repeatedly pointed to CISTR/Cistr-mediated regulation of cell size and morphogenesis genes. Thus, we compared orthologous DEGs related to cytoskeletal organization across mouse and human RNAseq data, including DEGs previously identified in C-28/I2 chondrocytes¹⁴. We find dozens of orthologous DEGs responsible for regulating cytoskeletal processes, cell junctions, and cell-cell adhesion. For example, filamin B (FLNB)⁵⁹, semaphorins (SEMA3A, SEMA6B, SEMA6D)⁶⁰, keratins (KRT6A, KRT75)⁶¹, claudins (CLDN1, CLDN2)⁶², cadherins (CDH2, CDH7) and protocadherins (PCDH1, PCDH7)⁶³, cadherin-like desmocollins (DSC2, DSC3)⁶⁴, integrins (IGTA1, IGTA8, IGTB3)⁴², myosins (MYH4, MYL1)⁶⁵, and Rho Family GTPases 2 and 3 (RND2, RND3)⁶⁶ are all differentially regulated in human and mouse experiments (Fig. 2k). Also, we observe DE paralogs between samples and mouse sexes (i.e., Ckap2 in female blood and Ckap4 in male bone marrow, or Epb41l2 in male cerebellum and Epb41l4b in female cerebellum; Supplementary Data 9, 12, 16, 17). The orthologous DEGs and consistent cell size phenotypes on cellular and organismal scale are indicative of CISTR’s conserved function.

CISTR coalesces with cell morphology genes and proteins in spatial proximity

Recent concepts suggest that lncRNA transcripts often stay proximal to their loci and enrich for otherwise diffusible molecules to control gene regulation in trans-chromosomal hubs^67–69. Since genome structure can be conserved across cell types and species^17,70,71, and due to CISTR’s trans-chromosomal proximities^15–17,72, we reasoned that CISTR’s DNA-encoded mechanism may regulate genes in trans. This is supported by 95.06% ΔCISTR/Cistr DEGs (q < 0.01, log₂FC ≥ 1 and ≤ −1) that are in trans to CISTR, which is significantly higher than expected (i.e., ΔCISTR RPMI2650: χ²-test, p = 0.002825), and specifies CISTR as a predominant trans-activator. To determine if consistent genome topology may support this trans-regulatory role, we queried trans-chromosomal proximities using our recent machine learning-empowered Hi-C analysis tool Signature¹⁷. Remarkably, CISTR has 152 common significant trans-chromosomal contacts in ≥ 31 of 62 Hi-C datasets (Fig. 3a, p < 0.05; Supplementary Fig. 14a, Supplementary Data 23), which is significantly higher than random chance (hypergeometric testing p = 0.0284). We also find 137 common trans-contacts for CISTR’s upstream flanking bin, whilst the region downstream to the locus has only four contacts (Supplementary Fig. 14b), thereby proposing that CISTR's locus is part of a stable topological environment with many trans-contacts. The common trans-contacts of CISTR’s bin harbor hundreds of DEGs determined in ΔCISTR (204 DEGs, Fig. 3b) and CasRx (366 DEGs, Fig. 3b) experiments, suggesting that CISTR can indeed facilitate trans-regulation of gene expression via spatial proximity, possibly as a trans-acting enhancer. Even in > 90% (56/62) of Hi-C datasets, the bin harboring CISTR’s locus consistently interacts with 42 DEGs from Cas9 and CasRx experiments (Fig. 3c). Notably, 19.05% (8/42) of these are orthologous DEGs in Cistr^−/− mice (p < 0.05), including genes responsible for cell shape and adhesion (TESK1⁷³, SH2D3C⁷⁴, SLC3A2⁷⁵, Fig. 3d). Synteny mapping further shows that DEGs of both Cas9 and CasRx experiments cluster locally in the same genomic neighborhoods in human and mouse (Fig. 3e). These findings indicate that conserved genome topology may support trans-regulation of genes and conserved lncRNA function. CISTR’s locus is consistently embedded in a topological genome structure (transcription hub) where gene loci of different chromosomes that are required for cell morphology coalesce and facilitate CISTR’s trans-activating function.

Fig. 3. Inter-chromosomal genome topology of CISTR’s locus.

a Circos plot showing inter-chromosomal contacts of CISTR’s locus in ≥ 31 Hi-C datasets (p < 0.05), unplotted chromosomes have no contacts. b Common trans-contacts overlapping with DEGs of ΔCISTR and CasRx experiments. c Same as panel b, but query in ≥ 56 Hi-C datasets. d Orthologous DEGs found in human and mouse perturbations. e Syntenic mapping of orthologous DEGs in human and mouse genomes. Red depicts DEGs from Cas9, whilst blue shows DEGs in CasRx approaches. f Schematic of dCas9-mediated proximity labeling. g Dot plot of proteins enriched in dCas9-mediated proximity labelling proteomics of CISTR and TERT over negative control (BFDR ≤ 0.01, n = 2 replicates). h MTA2 ChIP-qPCR showing fold enrichment of MTA2 bound to two positions of CISTR’s locus relative to IgG control, n = 3 replicates, standard deviation is depicted. Asterisks depict significance (**p < 0.01), determined by two-sided paired t-tests. Source data are provided as a Source Data file.

To better understand how CISTR’s DNA regulates the diverse DEGs, we developed CRISPR-mediated proximity labeling proteomics with a dCas9-miniTurbo fusion protein to identify proteins proximal to CISTR’s locus and to TERT^76,77, as a control with a comparable locus size (Fig. 3f, Supplementary Fig. 15a–d; Methods)⁷⁷. High inter-replicate similarity (Pearson’s correlation, r > 0.984) and high-confidence cutoff (Bayesian false discovery rate [BFDR] ≤ 0.01; Supplementary Fig. 15e) validated specific protein enrichment relative to controls, similar to other approaches⁷⁸. CISTR’s locus associates with 11 proteins, including actin polymerization and cell size regulation proteins FLNB and ACTR3^79–82 (Fig. 3g). These proteins are involved in pre-hypertrophic differentiation⁸³ and chondrocyte proliferation⁸⁴, and in the control of transcription⁸⁵, respectively, which are consistent with CISTR’s role in cell size, cell cycle, and transcriptional regulation. Moreover, CISTR was proximal to MTA2, a component of the nucleosome remodeling and deacetylase (NuRD) complex⁸⁶, and CHD2, an autism-associated chromatin regulator that interacts with tubulin⁸⁷ and impacts neurogenesis⁸⁸ (Fig. 3g; Supplementary Data 24, 25). We validated CISTR-MTA2 interactions with ChIP-qPCR (Fig. 3h, paired t-tests p < 0.0009). The CISTR locus-based protein enrichment suggest interactions with cytoskeletal components and potentially a novel axis of NuRD-related gene regulation.

CISTR’s RNA interacts with cytoskeletal components and FOSL2

We also examined CISTR RNA binding proteins that may explain the RNA-mediated function for cell size gene regulation. To accomplish this, we established miniTurbo-dCasRx proximity labelling and targeted CISTR transcripts and TERC⁷⁶ as a control due to its similar length and expression levels to CISTR (Fig. 4a, inter-replicate correlation, r > 0.94; Supplementary Fig. 15f–i).

Fig. 4. CISTR binds and guides FOSL2.

a Schematic of dCasRx-mediated proximity labeling. b Dot plot of proteins enriched in dCasRx-mediated proteomics of CISTR and TERC over negative control (BFDR ≤ 0.01, n = 2 replicates). c FOSL2 RIP showing fold enrichment of FOSL2 bound to CISTR transcript relative to IgG control, n = 3 replicates, standard deviation is depicted. Asterisk depicts significance determined by two-sided paired t-test. d Overlap of FOSL2-bound genes in wildtype and ΔCISTR RPMI 2650 cells. HOMER analysis of wildtype-exclusive peaks shows FOSL2 consensus motif as the most significant hit (p = 1×10⁻⁶). e Overlap of DEGs from Cas9 and CasRx experiments, and of RNA-regulated genes in RPMI2650 cells with FOSL2 ChIPseq peaks in blood and neural-derived cell types¹⁴². Asterisks show significance, determined by one-sided Fisher’s exact test. f Overlap between FOSL2-bound genes exclusively in wildtype cells, CISTR’s RNA-rescued genes, and DNA-related DEGs. g Examples of FOSL2 binding profiles in WT and ΔCISTR RPMI 2650 cells at genes that are bound by FOSL2 and regulated by CISTR’s DNA- and RNA function. h Flow cytometry cell size measurements upon CISTR and FOSL2 overexpression with two doses (black arrows: low and high plasmid concentrations) in WT and ΔCISTR cells (n > 60,634 cells, 2 replicates). Asterisks depict significance determined by 2-way ANOVA t-testing (p < 2.2×10⁻¹⁶). i Model of inter-chromosomal genome topology coalescing cell size gene loci at CISTR’s locus, thereby facilitating CISTR-FOSL2-mediated gene regulation. j Perturbing CISTR regulates cell size. Source data are provided as a Source Data file.

CISTR’s RNA associates with 14 proteins (BFDR ≤ 0.01), some of which are known cytoskeletal proteins. Namely, we find TUBB4A as brain-specific tubulin subtype associated with neuronal defects^89,90 and the microtubule-associated chromatin scaffold and cytoskeletal component UBR4 (p600)⁹¹ with critical functions in neurogenesis⁹² and myofiber hypertrophy⁹³ (Fig. 4b, Supplementary Data 26, 27). Notably, CISTR also partners with FOSL2 (Fig. 4b), which facilitates bone marrow- and brain-associated gene expression^94,95. RNA-IPs validated significant CISTR-FOSL2 interactions (Fig. 4c, paired t-test p = 0.0462), which may allow tissue-specific regulation of cell size genes.

CISTR-FOSL2 interactions regulate cell size

RNA-TF interactions can be required for TF-chromatin binding⁹⁶, representing a new idea of how TF-guided gene regulation can be influenced by RNA genes. FOSL2 is an activator protein 1 (AP-1) TF family member which regulates genes involved in cell growth, cell proliferation, cytoskeletal organization, and cell adhesion^97,98. Considering that CISTR’s RNA physically interacts with FOSL2, we investigated if CISTR and FOSL2 cooperatively regulate cytoskeletal genes.

To do this, we conducted FOSL2 ChIPseq in wildtype and ΔCISTR RPMI 2650 cells to determine whether FOSL2-chromatin binding is perturbed in the absence of CISTR (Supplementary Fig. 15j). We find that FOSL2 binding at 4638 genes in wildtype cells is absent in ΔCISTR cells (Fig. 4d, Supplementary Data 28, 29). Remarkably, HOMER motif analysis⁹⁹ shows a predominant enrichment of the FOSL2 consensus motif in ChIPseq peaks that are exclusive to WT (top hit, p = 1×10⁻⁶), whilst none of the highly similar AP-1 binding motifs are detectable in ΔCISTR cells (Fig. 4d). This proposes that CISTR is required for FOSL2 to bind to its motif. Re-analyzing ChIPseq datasets from blood and neurons confirmed high overlaps with significant FOSL2 binding in the promoter regions of hundreds of DEGs from our Cas9 and CasRx experiments, and of the 319 CISTR’s RNA-driven genes (Fisher’s exact tests p value range 0.043–1.76×10⁻¹⁰, Fig. 4e).

FOSL2 is downregulated in ΔCISTR hemin-differentiated K562 cells and in male Cistr^−/− hypothalamus and blood (Supplementary Data 7, 10, 21). This motivated us to ask if CISTR contacts FOSL2 on the DNA-DNA level as well. Revisiting our Hi-C analysis, we find significant CISTR-FOSL2 contacts in 27.4% (17/62, p < 0.05) of Hi-C datasets (Supplementary Fig. 15k), including mammary epithelial cells and epidermal keratinocytes that resemble epithelial RPMI 2650 cells that we used here.

To further address whether CISTR-FOSL2 DNA-DNA and RNA-protein interactions denote its molecular mechanism, we intersected the 1316 DNA-related DEGs (Fig. 1i), the 319 RNA-rescued DEGs (Fig. 1j), and the 4638 unique FOSL2-bound genes exclusive to WT (Fig. 4d and Supplemental Fig. 15l). Remarkably, a total of 231 DEGs are FOSL2-bound genes, which significantly overlap with the DNA-related and RNA-driven genes that CISTR’s locus and RNA regulate (hypergeometric tests p = 2.124×10⁻¹⁸ and p = 2.927×10⁻¹⁵, respectively; Fig. 4f). FOSL2 binding at RNA-rescued gene loci highlights CISTR’s importance in guiding FOSL2 (Fig. 4f). CISTR’s absence in ΔCISTR cells abolishes FOSL2 binding at the regulatory regions of genes involved in cell size and cell-cell adhesion that are either regulated by CISTR’s DNA and/or its RNA (Fig. 4g). CISTR-FOSL2-regulated genes are enriched for terms related to cell size control and the cytoskeleton (Supplementary Fig. 15m, n). To determine if CISTR and/or FOSL2 can rescue cell size, we overexpressed both with low and high doses of plasmid concentrations in WT and ΔCISTR cells and assessed cell size changes (Methods, Supplementary Fig. 15o). We find a dose-dependent cell size decrease for either CISTR or FOSL2 by comparing WT vs. KO and overexpression levels (2-way ANOVA testing p < 2.2×10⁻¹⁶, pairwise comparisons Supplementary Data 33), whilst CISTR and FOSL2 did not additively rescue enlarged cell size (Fig. 4h). This finding reveals that FOSL2 is also involved in cell size control and that it can compensate for function when CISTR is depleted.

Collectively, we find evidence that CISTR cooperates with FOSL2 at the DNA and RNA levels which is facilitated by stable genome topology and the coalescence of cell size genes on different chromosomes (Fig. 4i). CISTR’s RNA-FOSL2 interaction regulates cell morphogenesis genes and maintains cell size (Fig. 4j).

Discussion

Organisms must establish and maintain appropriate cell size to function efficiently^100,101. Although hundreds of genes responsible for cytoskeletal organization have been identified, defined gene-regulatory mechanisms controlling cell size genes remain elusive^36,102. To our knowledge, this is the first study describing the genetic basis for cell size control by a causal lncRNA. Many genes of the gene network involved in CISTR’s cell size regulation are largely unrelated to cell cycle progression (i.e., cell morphogenesis, cell-cell adhesions, cell junction organization, etc.). Thus, the CISTR-FOSL2 interaction may regulate cell size through multiple courses of action.

Our findings provide an example of how a lowly expressed and highly tissue-specific lncRNA impacts cellular function and gene networks in a spatial biomolecular hub. The overexpression of CISTR and FOSL2 together do not generate additive cell size effects, which may suggest that the stoichiometry of the components is important. Low copy numbers of CISTR are sufficient and effective to finetune the abundant FOSL2-mediated transcriptional regulation of its targets, which confirms that unabundant lncRNAs can overcome stochiometric imbalances to execute their functionality⁶⁹. Of note, FOSL2 is abundant in blood and across human tissues, but much lower expressed in brain (GTEx catalogue¹⁰³). The balance between high CISTR expression and low FOSL2 levels may support their stoichiometry and complement their functionality. Recently, TF-RNA binding has been shown to be necessary for TF-chromatin binding⁹⁶, which aligns with our findings that CISTR is required for FOSL2 binding at many cell size gene loci. CISTR’s overexpression may lead to FOSL2 misguidance and/or persistent binding. However, it remains to be determined how the CISTR-FOSL2 interaction guides FOSL2 to facilitate its binding and to drive gene regulation. Our findings further suggest that the consistent genome topology of CISTR’s locus, its trans-chromosomal proximities, and the CISTR-FOSL2 association facilitates the coalescence of cell morphogenesis genes and of the FOSL2 locus, which represents a new aspect of gene regulation (Fig. 4i). The spatial proximity between chromatin strands may support epigenetic regulators, such as NuRD complex components, and TFs (i.e., FOSL2) to form stable DNA-DNA contacts to drive trans regulation of genes, similar to the “olfactosome” components¹⁰⁴, and the Pantr1 RNA orchestrating CTCF-RNA binding protein interactions¹⁰⁵.

The NuRD complex directs diverse transcriptional regulatory mechanisms by leveraging paralog switching in combination with paralog-specific interactors, such as TFs, which unlocks tissue-specific function¹⁰⁶. This may directly relate to CISTR’s tissue-specific function, explain the paralogous DEGs among the studied cell and tissue types, and even account for sex-biased phenotypes observed in Cistr^−/− mice. The CHD2–lncRNA-[CISTR]–MTA2 axis that we found could represent a novel architectural feature of how NuRD and other chromatin-regulatory complexes exhibit tissue-specific functions which deserves further investigation.

Our study may help to control aberrant cell size seen in cancer^36,102, cortical malformations associated with cytoskeletal re-organization, epilepsy, and autism-spectrum disorders^107,108. VCAN, which was dysregulated in human and mouse ΔCISTR/Cistr experiments, including ΔCISTR C-28/I2 chondrocytes, modulates chondrocyte morphology and potentially differentiation¹⁰⁹ and thus may implicate CISTR in regulating cell size in chondrogenesis and brachydactyly¹⁴. In addition to CISTR’s high cardiac expression and the cardiac phenotype we observe in Cistr^−/− mice, we found dysregulated cardiomyopathy-associated genes (i.e., TPM1¹¹⁰). Our results suggest promising future studies of CISTR’s regulatory function and its potential to affect disease-related cell size changes.

Methods

Approval details

All studies were performed under the regulation of the SickKids Research Ethics Board and Canadian Institutes of Health Research Stem Cell Oversight Committee. The local regulatory animal authority (‘Landesamt fur Gesundheit und Soziales Berlin’ — LAGeSo) approved mouse-related studies (ZH120, i.e., whole mount in situ hybridizations). The generation of the Cistr knockout mice, their genotypization and clinical phenotyping was performed by The Centre for Phenogenomics (TCP, Toronto, ON, Canada). All procedures on animals at The Centre for Phenogenomics (TCP) were reviewed and approved by TCP’s Animal Care Committee. TCP is certified by the Canadian Council on Animal Care and registered under the Animals for Research Act of Ontario. Neural progenitor cells¹¹¹ were derived from hESCs (WIBR1)¹¹² and related work was approved by the Stem Cell Oversight Committee of the Canadian Institutes of Health Research, and the Research Ethics Board of the Hospital for Sick Children.

Generation of plasmid constructs and gRNA cloning

ΔCISTR Cas9 knockout

We cloned CISTR sgRNAs into PX459 (Addgene #62988) by BbsI restriction cloning. To generate a plasmid encoding both sgRNAs under separate promoters, we amplified the sequence encoding the U6 promoter and downstream sgRNA and inserted it downstream of the sequence encoding the U6 promoter and upstream sgRNA in PX459 using KpnI and XbaI restriction cloning. For knockout in neural progenitor cells, we cloned the same CISTR sgRNAs into PX458 (Addgene #48138). We conducted all PCRs using Phusion Hot Start II DNA Polymerase (Thermo Scientific).

ΔCISTR_intron Cas9 knockout

As above, we designed CISTR_intron sgRNAs and cloned them into PX459 together.

CasRx knockdown

To generate pXR003_puroR (Addgene #219819), we amplified the puromycin resistance cassette from BPK1520_puroR (Addgene #173901) and added to pXR003 (Addgene #109053) by NheI-HF restriction cloning. To generate a multi-gRNA array plasmid which also encodes CasRx, we added a second direct repeat to pLentiRNACRISPR_005-hU6-DR_BsmBI-EFS-RfxCas13d-NLS-2A-Puro-WPRE (Addgene #138147) to convert the sgRNA backbone to a pre-gRNA backbone. We annealed, phosphorylated (T4 Polynucleotide Kinase, NEB), and BsmBI restriction-cloned oligonucleotides encoding the 36-nucleotide direct repeat sequence and new BsmBI restriction sites into pLentiRNACRISPR_005-hU6-DR_BsmBI-EFS-RfxCas13d-NLS-2A-Puro-WPRE. To generate the CISTR gRNA array, we used an annealed and phosphorylated 200-base double-stranded ultramer (IDT), and we added on the remaining sequence by two overhang PCRs. We inserted the final array into the hU6-DR_BsmBI_DR-EFS-RfxCas13d-NLS-2A-Puro-WPRE CasRx pre-gRNA backbone (Addgene #219823) using BsmBI restriction cloning. We conducted all PCRs using Phusion Hot Start II DNA Polymerase (Thermo Scientific).

CRISPR-display

We designed eight sgRNAs manually outside of repetitive sequences and flanking the CISTR locus. These regions were used as input for https://portals.broadinstitute.org/gppx/crispick/public¹¹³. We selected sgRNAs with scores around 1 and low off-target scores and aligned them back to the genome to test specificity. sgRNAs were BbsI restriction cloned into pCMV/MASC_(GLuc)_INT (Addgene #68440) and CISTR’s cDNA sequence was inserted. To generate the CMV-dCas9_puroR vector (Addgene #219825), we first removed the KRAB and MeCP2 inhibition domains from dCas9-KRAB-MeCP2 (Addgene 110821) by inverse PCR. We circularized the PCR product using KLD Enzyme Mix (NEB) to generate a new backbone, and we inserted the puromycin resistance cassette which was amplified from BPK1520_puroR (Addgene 173901) using NheI restriction cloning. We conducted PCRs using Phusion Hot Start II DNA Polymerase (Thermo Scientific) and Q5 High-Fidelity DNA Polymerase (NEB).

dCasRx-mediated proximity labeling

To generate miniTurbo-dCasRx (Addgene #219821), we omitted the BASU biotin ligase from CARPID BASU-dCasRx (Addgene 153209) by inverse PCR, and we amplified and inserted the miniTurbo biotin ligase from pcDEST-pcDNA5-miniTurboID-3xFLAG-N-term (from Dr. Gingras Lab) into the modified dCasRx vector using the NEBuilder HiFi DNA Assembly Master Mix (NEB). To generate pXR004_puroR (Addgene #219820), we added the puromycin resistance cassette from BPK1520_puroR (Addgene #173901) to pXR004 (Addgene #109054) using restriction cloning, as was done with pXR003 (Addgene 109053). We used BsmBI restriction cloning to insert the CasRx gRNA array into pXR004_puroR. We conducted all PCRs using Phusion Hot Start II DNA Polymerase (Thermo Scientific).

dCas9-mediated proximity labeling

Similar to a recent approach¹¹⁴, we designed even and odd sgRNAs pools to minimize the potential effects of steric hindrance from dCas9-miniTurbo on protein interaction and capture. We designed nine sgRNAs^113,115 using Benchling, ensuring an on- and off-target score greater than 70%, and BbsI restriction cloned them into BPK1520_puroR (Addgene #173901). To generate dCas9-miniTurbo (Addgene #219822), we omitted the dCasRx in miniTurbo-dCasRx (Addgene #219823) by inverse PCR and circularized the product using KLD Enzyme Mix (NEB). We amplified dCas9 from the Inducible Caspex expression vector (Addgene #97421) and inserted upstream of miniTurbo in the modified backbone using NEBuilder HiFi DNA Assembly Master Mix. We conducted all PCRs using Phusion Hot Start II DNA Polymerase (Thermo Scientific) and Q5 High-Fidelity DNA Polymerase (NEB).

CISTR overexpression

We assembled the overexpression plasmids using the EF-1α promoter from CARPID BASU-dCasRx (Addgene #153209) or the CMV promoter from pCMV/MASC_(GLuc)_INT (Addgene #68440), the SV40 polyA from pCMV/MASC_(GLuc)_INT (Addgene #68440), and pUC19 (NEB) as the backbone with the NEBuilder HiFi DNA Assembly Master Mix (NEB). We added BbsI restriction sites downstream of the promoter with an overhang PCR and KLD Enzyme Mix (NEB) to circularize the product. We inserted a GFP expression cassette and the puromycin resistance cassette from BPK1520_puroR (Addgene 173901) into the overexpression vector using BamHI and XbaI restriction cloning. CISTR was inserted into the plasmid using BbsI restriction cloning. We conducted PCRs using Phusion Hot Start II DNA Polymerase (Thermo Scientific) and Q5 High-Fidelity DNA Polymerase (NEB).

Cell culture

We differentiated neural progenitor cells (NPCs) as previously described from hESCs¹¹¹. We plated NPCs on Matrigel-coated culture plates, replaced neural glial (NGD) media¹¹¹, which was supplemented every 24 h with 40 µg/mL human insulin and 10 ng/mL FGF-basic, and sub-cultured with accutase (Gibco) once per week. We validated NP markers SOX2 and Nestin using qRT-PCR in comparison to RPMI 2650 cells as a negative control.

We cultured RPMI 2650 in Minimal Essential Medium (Gibco), K562 cells in Iscove’s Modified Dulbecco’s media (Sigma), and C-28/I2 cells in Dulbecco’s modified Eagle media (Sigma)/Nutrient Mixture F12 (Gibco). We supplemented all media with 10% fetal bovine serum (Gibco) and penicillin/streptomycin (Gibco). We sub-cultured cells with TrypLE Express Enzyme (Gibco) twice per week, apart from RPMI 2650 cells, which we sub-cultured once per week.

CRISPR methodology: gRNA optimization and transfection

Cas9 sgRNA efficiency test

We designed five upstream and four downstream sgRNAs with an on- and off-target score greater than 70% to delete CISTR (RefSeq NR_104333) using Benchling^113,115. To experimentally determine individual sgRNA efficiency, we seeded 1.0 × 10⁶ RPMI 2650 cells in a 6-well plate and transfected with 4 µg PX459 containing CISTR sgRNAs using 6.67 µL Lipofectamine 3000 transfection reagent (Sigma) 24 h later. 48 h after transfection, we selected transfected cells with 4 µg puromycin (Gibco) for 72 h, then harvested and extracted gDNA from the samples using the DNeasy Blood & Tissue kit (QIAGEN). We conducted genotyping PCRs using OneTaq 2X Master Mix with Standard Buffer (NEB), 1 M betaine (Millipore Sigma), and 40 µM 7-deaza-2’-dGTP (Roche). We column-purified PCR products using the Monarch PCR & DNA Cleanup Kit (NEB), performed Sanger sequencing, and TIDE analysis¹¹⁶ using default parameters to determine genome editing efficiency. We cloned the most efficient sgRNAs, into the same PX459 vector.

RPMI 2650 knockout

We seeded 1.0 × 10⁶ RPMI 2650 cells in a 6-well plate and transfected with 4 µg PX459 containing CISTR sgRNAs using 6.67 µL Lipofectamine 3000 transfection reagent (Sigma) 24 h later. 48 h after transfection, we selected transfected cells with 4 µg puromycin (Gibco) for 72 h. To generate isogenic cell clones, we first determined successful knockout in the cell population by PCR and then distributed single cells from the population using limited dilution methodology. We harvested these isogenic cell lines 18-21 days after seeding, keeping 50% in culture, and extracted gDNA using DNeasy Blood & Tissue kit (QIAGEN). We conducted genotyping PCRs using OneTaq 2X Master Mix with Standard Buffer (NEB), 1 M betaine (Millipore Sigma), and 40 µM 7-deaza-2’-dGTP (Roche). Amplicons were purified using the Monarch PCR & DNA Cleanup Kit (NEB) and sent for Sanger sequencing. We generated two independent isogenic cell lines for ∆CISTR and ∆CISTR_intron (biological replicates) and compared them to two isogenic WT clones and to independent cell passages as technical replicates.

C-28/I2 knockout

We seeded 5.0 × 10⁵ C-28/I2 cells in a 10 cm² and transfected with 12.5 µg PX459 with CISTR sgRNAs using 40 µL Lipofectamine 3000 transfection reagent (Sigma) 24 h later. 48 h after transfection, we selected cells that received the plasmid with 20 µg puromycin (Gibco) for 72 h, and RNA sequenced three passages of the knockout population.

K562 knockout

We seeded 2.0 × 10⁵ cells in a 6-well plate. 24 h later, we transfected with 2 µg PX459 containing hU and hD CISTR sgRNAs using 6 µL X-tremeGENE HP DNA transfection reagent (Sigma). 24 h post transfection, we selected transfected cells with 4 µg puromycin (Gibco) for 48 h. We replaced this media with fresh media and puromycin for an additional 72 h of selection, after which we performed genotyping on cell populations, and analyzed three passages of the populations by RNA sequencing.

NPC knockout

After sub-culturing, we prepared ~15 × 10⁶ NPs for electroporations. We seeded 2 × 10⁶ cells in a Matrigel-coated 24-well plate to serve as the negative control. Upon resuspension in Buffer R (Thermo Fisher Scientific) and combining with 6 µg of PX458_CMV-GFP plasmid (Addgene #219824) containing CISTR sgRNAs, we electroporated ~13 × 10⁶ NPs using the Neon Transfection System (Thermo Fisher Scientific) with 100 µl Neon tips cells (1100 V, 40 ms, and 1 pulse). We enriched for GFP-positive cells by FACS 48 h post-transfection, using the Aria Fusion flow cytometer. Upon expanding electroporated and FACSorted NPCs (˜3 weeks), we performed genotypizations, determined the editing efficiency of deleting CISTR, and extracted RNA from three independent experiments.

CasRx knockdown sgRNA test

We designed 30-nucleotide sgRNAs manually outside of repeat regions and checked for predicted off-target binding in BLAST (only sgRNAs with off-target binding at ≤ 18 nucleotides were accepted). sgRNAs in pX003_puroR (Addgene #219819) targeting CISTR were transfected into RPMI 2650 cells with pXR001 (Addgene #109049), and RNA was harvested with TRIzol (Invitrogen) and tested for CISTR knockdown efficiency with RT-qPCR using hCISTR_RTqPCR primers. We designed an array with the four most efficient sgRNAs.

CasRx knockdown

In a 6-well plate, we seeded 1.0 × 10⁶ RPMI 2650 cells and transfected 1 µg CISTR array in pLentiRNACRISPR_005—hU6-DR_BsmBI_DR-EFS-RfxCas13d-NLS-2A-Puro-WPRE CasRx pre-gRNA backbone (Addgene #219823) or the vector with no sgRNAs as ‘empty vector’ control using 6.67 µL Lipofectamine 3000 transfection reagent (Sigma) 24 h later. Twenty-four hours post transfection, we selected for cells that received the plasmid using 4 µg puromycin (Gibco) for 48 h and analyzed three independent experiments.

CRISPR-display

In a 6-well plate, we seeded 1.0 × 10⁶ RPMI 2650 cells, and we transfected 1.5 µg equimolar sgRNA pool (8 sgRNAs), 2.5 µg CMV-dCas9_puroR (Addgene #219825), and 6.67 µL Lipofectamine 3000. 24 h after transfection, we selected for cells that received the plasmid using 4 µg puromycin (Gibco) for 48 h. We performed two independent experiments.

dCas9-mediated proximity labeling sgRNA test

Similar to a recent approach¹¹⁴, we designed even and odd sgRNAs pools to minimize the potential effects of steric hindrance from dCas9-miniTurbo on protein interaction and capture. We seeded 2.0×10⁷ cells in 3 15 cm² plates. 24 h later, we transfected each plate with 48 µg dCas9-miniTurbo (Addgene #219822), 4.8 µg sgRNA even or odd plasmid pool in BPK1520_puroR or BPK1520_puroR no sgRNA control (Addgene #173901) and 100 µL Lipofectamine 3000 transfection reagent. 24 h after transfection, we selected for the cells that received the plasmid with 50 µg puromycin for 24 h. Genomic DNA was harvested and ChIP-qPCR was carried out using the ChIP-seq protocol below, except using anti-FLAG antibody. Six pairs of qPCR primers (hCISTR_dCas9ChIP) corresponding to sgRNA loci were used to quantify dCas9 binding at the region of interest.

dCas9-mediated proximity labeling

We seeded 2.0 × 10⁷ RPMI 2650 cells in 18 15 cm² plates (3 for experimental sample even sgRNA pool, 3 for experimental sample off sgRNA pool, 6 for TERT control sample, 6 for no sgRNA control). 24 h later, we transfected each plate with 48 µg dCas9-miniTurbo (Addgene #219822), 4.8 µg CISTR or TERC array in pXR004_puroR (Addgene #219820) no sgRNA control, and 100 µL Lipofectamine 3000 transfection reagent. 24 h after transfection, we selected for the cells that received the plasmid with 50 µg puromycin for 24 h. We performed two independent experiments.

dCasRx-mediated proximity labeling

We seeded 2.0 × 10⁷ RPMI 2650 cells in 18 15 cm² plates (6 for experimental sample, 6 for TERC control sample, 6 for no sgRNA control). 24 h later, we transfected each plate with 48 µg miniTurbo-dCasRx (Addgene #219821), 4.8 µg CISTR or TERC array in pXR004_puroR or pXR004_puroR (Addgene #219820) no sgRNA control, and 100 µL Lipofectamine 3000 transfection reagent. 24 h after transfection, we selected for the cells that received the plasmid with 50 µg puromycin for 24 h. We performed two independent experiments.

Determining knockout allelic frequency in CRISPR/Cas9 knockout populations

In cell lines that either do not express CISTR (K562 and C-28/I2), we used an approach adapted from determining allelic frequencies of SNPs¹¹⁷. We designed qPCR primers targeting 1) the region within the knocked-out locus to only amplify the wildtype allele (hCISTR_int_geno), 2) regions flanking the locus to only amplify the knockout allele (hCISTR_ext_geno), and 3) primers outside the locus to quantify the total number of alleles (hCISTR_all_geno). The frequency of the knockout allele in the sample was calculated using the following equation:

$$f_{knockout\_allele}=\frac{1}{2^{-\left({Ct}_{hCISTR\_\mathrm{int}\_grno-}{Ct}_{hCISTR\_ext\_geno}\right)}+1}$$ 1

Ct is the cycle threshold in quantitative PCR.

RNA extraction, cDNA preparation, RT-qPCR

Total RNA was extracted from cells using the phenol-chloroform extraction method according to standard protocols. Residual genomic DNA was removed using DNAse I digestion (Invitrogen) according to the manufacturer’s instructions. cDNA was synthesized using SuperScript III First Strand Synthesis System (Invitrogen). qRT-PCRs were performed using PowerUp SYBR Master Mix (Applied Biosystems) with the primers listed in the supplements and analyzed by using the 2^(−ΔΔCt) method. Relative expression was calculated using GAPDH as housekeeping gene.

RNAseq C-28/I2, RPMI 2650, K562, NPCs, mouse tissues

We performed paired-end (2×150 bp) RNA sequencing of rRNA-depleted total RNA (Illumina Stranded Total RNA Prep Ligation with Ribo-Zero Plus) using the Illumina NovaSeq 6000 platform. Upon determining RNA integrity using the BioAnalyzer 2100 (Agilent), we performed stranded total RNA prep ligation (Illumina) with Ribo-Zero Plus library preparations for total RNA of cell lines (see “CRISPR Methodology: gRNA optimization and transfection” section for information about replicate numbers in each experiment) and of mouse tissues of two to five animals (12 months of age) per sex of WT and Cistr^−/− mice. We aligned reads to hg38 using STAR aligner, version 2.7.0f¹¹⁸. We used FeatureCounts to count reads aligning to genes in GENCODE v32 (human), vM33 (mouse), and removed genes with <10 reads from the analysis. Data from samples split across multiple sequencing lanes were pooled before differential gene expression analysis. We quantified gene expression in each sample using DESeq2, version 1.44.0¹¹⁹. All log2FC of human in vitro models can be found in Supplementary Data 32.

Enrichment analysis

Functional enrichment of genes was accomplished with Metascape³².

CISTR overexpression

RPMI 2650

We seeded 1.0 × 10⁶ ΔCISTR and WT RPMI 2650 cells in a 6-well plate and transfected with 4 µg CISTR overexpression EF1α-BbsI-SV40polyA_EGFP and 6.67 µL Lipofectamine 3000 transfection reagent 24 h later. Cells were analyzed on flow cytometer 48 h post transfection.

C-28/I2

We seeded 8.0 × 10⁴ ΔCISTR and WT C-28/I2 cells in a 6-well plate and transfected with 2 µg CISTR overexpression EF1α-BbsI-SV40polyA_EGFP and 6.77 µL Lipofectamine 3000 transfection reagent 24 h later. Cells were analyzed on flow cytometer 48 h post transfection.

K562

We seeded 2.0 × 10⁵ ΔCISTR and WT K562 cells in a 6-well plate and transfected with 2 µg CISTR overexpression EF1α-BbsI-SV40polyA_EGFP and 6 µL X-tremeGENE HP DNA transfection reagent (Sigma) 24 h later. Cells were treated with 50 µM hemin dissolved in media. Hemin media was replenished after 24 h.

Flow cytometry for relative cell size measurement

We analyzed ~5,000-100,000 live cells across two to three replicates for forward scatter area values on the BD Biosciences FACSymphony A3 Cell Analyzer (running FACSDiva v9.3.1 software) using propidium iodide as a viability dye. For CISTR overexpression experiments where we transfected cells with CISTR overexpression EF1α_BbsI_SV40 polyA_EGFP (Addgene #219828), we only considered cells that expressed GFP as CISTR-overexpressing cells.

Cell cycle assay

We centrifuged 2.0 × 10⁶ RPMI 2650 cells at 400 x g for 5 min at 4 °C, aspirated supernatant, and re-suspended the pellet in 50 µL staining media (1X Hank’s balanced salt solution [Gibco], 10 mM HEPES NaOH pH 7.2 [Gibco], 2% fetal bovine serum [Gibco]). We added the resuspended cells dropwise to 1 mL of ice-cold 80% ethanol in a conical polypropylene tube while vortexing, and fixed at 4 °C overnight. We then centrifuged fixed cells at 400 x g for 5 min at 4 °C, re-suspended the cell pellet in 500 µL 2 mg/mL RNase A (Invitrogen) in 1X Hank’s balanced salt solution, and incubated for 5 min at room temperature. We then added 500 µL 0.1 mg/mL propidium iodide solution (0.1 mg/mL propidium iodide [Invitrogen], 0.6% NP-40, 1X Hank’s balanced salt solution), vortexed to mix, and incubated overnight at room temperature, protected from light. We analyzed > 13,000 cells across two replicates by flow cytometry and determined cell cycle stages by the Watson Pragmatic algorithm¹²⁰.

Cell proliferation assay

We stained 1.0 × 10⁶ RPMI 2650 cells with 2.5 µM CellTrace Violet (Invitrogen) according to the manufacturer’s protocol. We analyzed cells 48, 72, and 96 h after staining, and we analyzed > 13,000 cells across two replicates at each time point. To determine Δ proliferation:

$$\mathrm{\Delta }proliferation=\frac{{\mu }_{WTfluorescenceintensity}-{\mu }_{\mathrm{\Delta }CISTRfluorescenceintensity}}{{\mu }_{\mathrm{\Delta }CISTRfluorescenceintensity}}\times 100\%$$ 2

Cell synchronization

G₀/G₁ synchronization

We seeded 4.0 × 10⁵ WT RPMI 2650 cells and treated with 2 mM thymidine (Sigma-Aldrich) 24 h later for 19 h. After the first treatment, cells were washed twice with PBS and incubated in complete media for nine hours. Cells were then treated with 2 mM thymidine again for 16 h. Cells were washed twice with PBS before RNA was extracted.

G₂/M synchronization

We seeded 4.0 × 10⁵ WT RPMI 2650 cells and treated with 2 mM thymidine (Sigma-Aldrich) 24 h later for 24 h. After treatment, cells were washed twice with PBS and incubated in complete media for three hours. Cells were then treated with 100 ng/mL nocodazole (Millipore-Sigma) for 12 or 16 h. Cells were washed twice with PBS before RNA was extracted.

Whole mount in-situ hybridization

We performed in vitro transcription (DIG RNA labelling, Roche) of the subcloned Cistr in pGEM-T Easy (Promega). Mouse (E13.5) and rat (E14.0) embryos were fixed in PFA (4% in PBST/DEPC pH 7.5) over night. We used an Anti-Digoxigenin-AP antibody (Roche) for labelling and analyzed 12 embryos with standard whole mount lacZ staining.

Cistr knockout mouse

The Center for Phenogenomics (Toronto, Canada) generated the Cistr mouse model. Upon assessing cognate gRNA sequences to identify off-targets adjacent to canonical PAM motifs, specificity was scored as previously described^115,121. We used gRNA_U (CCCTCTCTGAGGTCGTGACA, Chr15:102746421, score 91) and gRNA_D (CAGTTTCCGAAATCGCCCGG, Chr15:102746884, score 100) for targeting Cistr (MGI:5621321, formerly Gm38436) and electroporated mouse zygotes of the C57BL/6NCrl strain (Charles River Laboratories) with Cas9 ribonucleoprotein complexes to reduce off-target mutagenesis¹²². Upon germline transmission, we screened N1 progeny of founder mice for the desired genotype and did three backcrossings. We then performed clinical phenotyping of at least seven wildtype, Cistr^+/- and Cistr^−/− animals of both sexes (weight curve [weeks 4-16], open field [week 9], combined SHIRPA and dysmorphology [week 9], grip strength [week 9], Rotarod [week 9], tail suspension [week 9], acoustic startle and prepulse inhibition [week 10], anxiety [week 10], social interaction [week 10], brain MRIs [week 9], fear conditioning [week 11], anesthetized electrocardiography and echocardiography [week 12], intraperitoneal glucose tolerance test [week 13], acoustic brainstem response [week 14], whole body X-ray with Isoflurane anesthesia and X-ray of digits [week 14], body composition [week 14], slit lamp [week 15], fundoscopy [week 15], saphenous vein blood collection with clinical chemistry [week 15], hematology and immunophenotyping [week 16], and gross pathology with necropsy which included cardiovascular, metabolic, and renal blood biochemistry, and spleen and heart weights). Euthanasia was performed by CO₂ in distress-free environment for adult mice.

Neurobehavioural phenotypization

We acclimated the animals prior to any behavioral assessment. We wiped equipment clean between subjects using Clidox (1:5:1), and tested males before females where possible. Replicates of the same experiment type were performed at a similar period of the light cycle.

Open field test

We performed an Open Field test to measure anxiety and exploratory activity of the Cistr knockout mice as previously described¹²³. We prepared an Open Field arena (43.5 cm²) equipped with infrared sensors within a sound attenuating chamber with a 275 lux LED light. We left the mice undisturbed in an anteroom for 30 min before moving them to the centre of the Open Field arena. We then allowed them to explore freely for 20 min. We used Activity Monitor software to track the time spent in either the periphery (outer 8 cm) or centre (remaining 40%) of the arena.

Pre-pulse inhibition test

We used a Pre-pulse inhibition (PPI) to measure sensory gating as a way to determine their ability to integrate new sensory information¹²⁴. Prior to testing, we left the mice in an anteroom for 30 min undisturbed, then allowed them to habituate for 5 min in the SR-LAB™ startle response system. We presented mice with each of three trial types: a) Different pre-pulse trials of 20 ms duration of white noise stimuli (70, 75, 80, or 85 dB) with or without startle pulse (110 dB/40 ms), b) Startle pulse presented alone, c) No stimulus (NOSTIM), where background noise (65 dB) is presented to measure baseline movement of the animal in the chamber. We presented each trial type to the subject six times in pseudorandom order, with the inter-trial interval varying randomly between 20 and 30 seconds. In each trial type, we measured the startle response for 65 ms after the initiation of the trial. Normal mice are expected to have a suppressed startle response when the pre-pulse is present whereas affected mice would experience lower attenuation of the startle response.

Anxiety (light/dark box assessment)

We measured anxiety using a light/dark box assessment¹²⁵, where more time spent in the light compartment is indicative of higher anxiety. We prepared an Open Field arena (43.5 cm²) equipped with infrared sensors within a sound attenuating chamber with a 275 lux LED light and a dark box insert on one side. We placed mice in the centre of the light side of the arena facing the dark box and allowed them to freely explore for 10 min. We acquired data using the Activity Monitor software.

Tail suspension test

We used a tail suspension to determine relative behavioral despair or depression¹²⁶. We suspended mice in a tail suspension box (Bioseb) by wrapping the tail in Transpore tape and placing a hook through the tape. We assessed movement for 360 seconds using Suspension software (Bioseb).

Fear conditioning test

We assessed learning and memory using a fear conditioning test¹²⁷. We left the mice in their home cages on a day rack at least 30 min before testing, then we transferred them to the NIR Video Fear Conditioning System (Med Associates Inc.). On the first day of testing, we conditioned mice with a 120 s baseline period without stimuli, 30 s of an audible tone (2800 Hz, 85 dB), a 2 s foot shock (0.75 mA), and another 150 seconds with no stimuli. 18–24 h later, we recorded the context where no sounds or shocks are delivered for 300 s. Finally, 2 h after the context protocol was completed, we delivered the cue to each subject. During cue delivery, no stimulus was presented for the first 120 s, followed by a 180 s delivery of the tone. We used Video Freeze software to determine the number of freezing episodes, percent freezing time, and motion index.

Social interaction

We used a social interaction test to measure sociability and preference for social novelty¹²⁸. We prepared a three-chamber box (59 cm × 39 cm x 22 cm), beginning with the left and right compartments closed off from the center compartment. We placed each mouse in the center compartment to habituate for 5 min. We then introduced a control mouse (same age, weight, and sex but no prior contact) in one of the side chambers and removed all dividers to allow the subject free access to all chambers for 10 min. We then introduced a second control mouse to the other side chamber and allowed free access for 10 additional minutes. We recorded all tests by video tracking EthoVision XT (Noldus).

Motor coordination test

We evaluated mouse motor coordination using an accelerating Rota-Rod (LE8200; Harvard Apparatus Canada)¹²⁹. We placed five mice on the Rota-Rod as it spun at 4 rpm. Once the mice began walking in their respective lanes, the speed was gradually increased to 40 rpm over the course of 5 min. We recorded the latency at which the mice fell off the rod across three trials with a 15 min rest period in between each one.

Hematology

We evaluated the production of blood and its components in 15-week-old conscious mice using a Hemavet Multispecies Hematology Analyzer (HV950FS; Drew Scientific, U.S.A.). These evaluations were all performed before noon on the day of analysis. We used a 25-gauge needle to puncture the saphenous vein and drew 50 µL blood into an EDTA-coated capillary tube. We dispensed 25 µL blood into a 0.6 mL microtube and mixed thoroughly to avoid clot formation and stored it at room temperature for 15 min prior to hematological analysis.

Immunophenotyping

We performed immunophenotyping according to Hyrenius-Wittsten and colleagues¹³⁰ with antibodies detecting CD5-BV421, CD62L-APC-Cy7, CD8a-CF594, CD25-PE-Cy7, CD161-APC, CD4-FITC, CD44-PE, CD5-BV421, CD23-BV786, Ly6G-BV421, CD19-BV650, Ly6C-FITC, CD21/35-PE, CD161-APC, CD11b-PE-CF594, CD11c-PE-CY7 (all BD), and MHCII-APC-Cy7 (ThermoFisher) on a BD LSR Fortessa cytometer.

Blood films

We prepared blood films according to standard protocols. We used the 3DHistech Pannoramic Flash II Slide Scanner equipped with a 40x objective for high-resolution scans, and Caseviewer software (3DHistech) for analysis.

Echos and electrocardiograms

We placed the mouse in a supine position on the imaging platform and anesthetized it with isoflurane in pure medical oxygen (5% for induction administered by induction chamber, 1.5–2% for maintenance administered by nose cone) using an Isoflurane vaporizer (Benson). We gently secured the nose and each of the animal’s limbs with adhesive tape, ensuring forefeet and hindfeet were lying flat on the ECG sensors of the platform, and placed a rectal thermometer probe. We checked the heart rate and body temperature and ensured they remained between 600 and 700 bpm and 36.5 and 38 °C, respectively. We removed the hair from the chest with an electric clipper followed by chemical hair remover and performed ultrasound biomicroscopy with a Vevo 2100 (VisualSonics). We obtained a short-axis view of the left ventricle (LV) in B-mode with the transducer in the parasternal position at the level of the papillary muscles to observe any wall motion abnormalities. From the same view, using M-mode acquisition, we measured LV cavity dimensions and wall thickness at end-diastole and end-systole. We assessed LV diastolic function using pulse-waved Doppler of blood flow across the mitral valve in the apical long-axis view. We moved the sample to the mitral valve leaflets and measured the peak mitral inflow velocity from the spectral display of pulsed wave Doppler velocities across the mitral valve. We observed mitral and aortic valve closing and opening signals and measured the isovolumic relaxation time, isovolumic contraction time, and left ventricular ejection time by positioning the sample between the LV inflow and outflow.

Electrocardiogram (ECG)

We observed the presence or absence of arrhythmogenic disturbance in mice using anesthetized electrocardiogram. We placed mice in a supine position on a heating pad set to 40 °C to maintain normothermia. We anesthetized mice as described above. Using the Lead II surface ECG (ADInstruments), we placed needle electrodes in the right arm and right and left legs. We recorded ECG traces for 2 min and acquired images of any arrhythmias or abnormal findings. We recoded data and performed analysis using LabChart 8.

X-rays

We acquired X-ray images using an UltraFocusDXA ultra-high-resolution imaging and dual-energy X-ray absorptiometry system (Faxitron Bioptics, LLC) to observe any skeletal abnormalities. We weighed the mice, anesthetized them as described above, and placed them on the X-ray tray in a dorsal position. We set the X-ray system to a voltage of 32 kV with an integration time of 4100 ms and acquired images at 1X and 1.5X magnification. We then repositioned the mouse in a lateral position with limbs non-overlapping and repeated the imaging. Prior to replacing the mouse in its cage, we removed the mouse from the X-ray tray and measured the body length.

RNA FISH

We followed the Molecular Instruments protocol for HCR RNA FISH using the B2-647 amplifier for Cistr and the B3-488 amplifier for Gad2. We acquired images using a Leica SP8 upright laser scanning confocal microscope. We used the Leica 40x HC PL APO oil objective for all acquisitions. For validation of sgRNAs for CRISPR-Display experiments, we used 15x RNA FISH probes (fluorophore CAL Fluor Red 610) from Stellaris. Probes hybridization and imaging were accomplished as recently described¹³¹.

Immunofluorescence

Mouse hemi-brains (WT and Cistr^−/−, 2 animals per sex, 2 animals per genotype, age: 12 months) were dissected, fixed in 4% paraformaldehyde, and processed for histology and immunohistochemistry at The Centre for Phenogenomics Pathology Core (Toronto). The samples were processed in a Tissue Tek VIP 6 tissue processor (Sakura Finetek, USA), paraffin-embedded, then 4 µm-thick sections were collected onto charged slides (Assure, Epic Scientific, USA). For immunohistochemistry, sections were deparaffinized, submitted to heat-induced epitope retrieval with citrate buffer, pH 6.0, in a pressure cooker, then incubated in Agilent Protein Block, Serum Free (Dako, cat # X0909). Primary antibodies (Rabbit mAb to Neurofilament, Abcam, cat # 207176, 1:1000; Rabbit mAb to NeuN, Abcam, cat # ab177487, 1:1500; Rat anti GFAP, Invitrogen, cat # 13-0300, 1:1000) were diluted in Dako Antibody Diluent solution (Agilent, cat # S3022) and incubated as cocktail overnight at 4 °C. On the following day, the tissue sections were washed in 1× Tris-buffered saline with 0.1% v/v Tween-20 (1x TBST), then incubated in secondary antibody cocktail mixture containing goat anti-rat IgG conjugated with Alexa Fluor 555 (Invitrogen, cat # A31434) and donkey anti-rabbit IgG conjugated with Alexa Fluor 647 (Invitrogen, cat # A32795) diluted at 1:200 in Antibody Diluent for 1 h at room temperature. Sections were rinsed in 1x TBST, then counterstained with DAPI (Sigma, Cat # D9542) for 5 min. Sections were treated with Sudan black B saturated solution in 70% ethanol (Sigma, cat # 199664) for 20 min at room temperature to quench autofluorescence, then Vectashield Vibrance anti-fade mounting medium (Vector; cat # H-1700) was used to mount coverslips. Slides were scanned using a Zeiss Axioscan 7, equipped with a Plan-Apochromat 40x/0.95 objective and an Axiocam 712 m.

X-rays

We analyzed skeletal morphology and measured metacarpals and proximal phalanges of 14 weeks old mice (four WT and Cistr^−/− animals of each sex) by measuring lengths of metacarpals and proximal phalanges using the MicroDicom DICOM viewer.

Brain MRIs

We performed MR imaging using a 7-Tesla magnet (Biospec 70/30 USR, Bruker, Ettlingen, Germany) equipped with Avance III HD electronics, a 200 mm gradient set (300mT/m maximum amplitude), and a console running ParaVision 6.0.1 software. The system was customized to enable scanning of multiple samples simultaneously as previously described^132–134. We prepared specimens by perfusion fixation with fixative and gadolinium contrast agent (ProHance, Bracco Imaging Canada) as described in detail previously^135,136. We performed scans using a three-dimensional T2-weighted sequence with parameters: TR = 350 ms, 12 ms echo spacing, 6 echoes, TEeff=30 ms, 630x504x504 matrix, 40 µm isotropic resolution, and 4 effective averages (total scan time, ~13 h)¹³⁷.

We registered brains of ten male and female WT mice vs. 11 male and ten female Cistr^−/− mice (9 weeks of age) together through iterative and nonlinear steps to provide an unbiased average representative of the population. An existing brain atlas with 185 individual structures⁴⁹ was registered to the resulting average image and used to compute the volumes of each structure, which were then used to calculate whole brain volume. We modeled whole brain volume results for statistical significance using the following linear model (in R notation):

$$\mathrm{Vol}~\mathrm{Genotype}+\mathrm{Sex}+\mathrm{Genotype}:\mathrm{Sex}$$ 3

Overall, whole brain volume was not different between genotypes (p = 0.150, ~1% difference). We computed the volumes of 185 total structures (combining left and right volumes) for all animals and conducted the analysis using two different methods. The output of linear regression models was transformed into p-values. First, we used the absolute volumes to assess group differences (i.e., measured volume in mm³). The same linear model used for total brain volume was used for all structures, with correction for multiple statistical comparisons through the false discovery rate (q). In the second method, we normalized volumes to total brain volume to yield each structure as a percentage of total brain volume. This can, to a degree, account for biological variability between mice. In the absolute analysis, we identified a total of 19 structures as significant using a lenient threshold of p < 0.05 (uncorrected) between WT and HOM; additionally, a total of seven structures were identified in the interaction between genotype and sex using the same threshold. In the relative analysis, we found a total of six structures as significant at p < 0.05 (uncorrected) between WT and HOM; none was significant in the interaction between genotype and sex. The medial amygdala were significantly different between sexes at q < 0.1 in the relative structure-wise analysis, which is consistent with previous findings on sexually dimorphic structures in mice⁴⁹. This structure is expected to be larger in males than in females.

Cell body size measurements of neurons and quantifications

First, we automatically optimized pixel intensity display settings with a “linear best-fit analysis” applied to the entire field of view of each channel of immunofluorescent images of 12 months old mice (WT, Cistr^−/−, each genotype two animals per sex) in Zen blue (Zeiss). We then localized the piriform cortex by using the Allen Mouse Brain Atlas⁴⁷ and measured the longest distances (x – y dimension) of 200 hundred neurons stained positive for NeuN in the pyramidal and polymorph layers of the piriform cortex. To quantify neurons, we selected the midpoint of the pyramidal layer and set a gate of ˜65,000 µm² that we used once across all animals in the pyramidal layer with high neuronal density, and twice in the polymorph layer because of lower neuron density. To measure neurofilament intensity, we applied the same ‘linear best-fit analysis’ to all samples and automatically quantified signal intensities in a region (˜780,000 µm²) overlapping molecular, pyramidal, and polymorph layers of the piriform cortex.

Hi-C analysis

We analyzed significant trans genomic contacts of CISTR’s locus (p < 0.05) in 62 Hi-C datasets of 2n genomes as recently described¹⁷. We then queried either all common contacts in > 50% (31/62) or in > 90% (56/62) of the Hi-C datasets and intersected DEGs from the Cas9 and Cas13 RNAseq experiments to determine the number of DEGs in the significantly interacting bins.

Proximity labeling

After transfection and puromycin selection, we performed proximity labeling in two independent replicates by adding 200 μM biotin to the cells, followed by incubating 2 h for dCasRx-mediated proximity labeling, or 3 h for dCas9-mediated proximity labeling, at 37 °C with 5% CO₂. Cells were washed three times with 1X PBS, scraped in 1X PBS, spun down at 300 g for 5 min, and pellets were snap frozen until ready for biotin immunoprecipitation. Experiments were performed in duplicate.

Enrichment of biotinylated proteins for mass spectrometry

We resuspended frozen cell pellets in 1.8 mL (300 µL per 15 cm² plate) ice-cold modified RIPA lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EGTA, 0.5 mM EDTA, 1 mM MgCl₂, 1% NP40, 0.1% SDS, 0.4% fresh sodium deoxycholate, 1 mM fresh PMSF, 1X fresh Protease Inhibitor cocktail) and incubated on ice for 15 min. We sonicated lysate on medium intensity using the Bioruptor Plus UCD300 (Diagenode) for 12 cycles of 30 s on and 30 s off, then added 1500 U of Benzonase (Millipore Sigma) and 120 µg of RNase A (Invitrogen) and end-over-end rotated at 4 °C for 15 min. We increased SDS concentration to 0.4% and rotated at 4 °C for 5 min. We centrifuged the lysate at 20,817 g for 20 min at 4 °C to clear debris and transferred the supernatant to a new centrifuge tube. We washed 35 μL of streptavidin agarose bead slurry per sample 3 times with 1 mL lysis buffer and centrifuged 400 g for 1 min. Following the last wash, we resuspended the beads as a 50% slurry, and we added 40 μL of the 50% slurry of streptavidin beads to the clarified supernatant and rotated using gentle end-over-end rotation for 3 h at 4 °C. We pelleted the beads by centrifugation at 400 g for 1 min at 4 °C and transferred to a new microcentrifuge tube in 1 mL fresh RIPA-wash buffer. We washed the beads once with SDS-Wash buffer (25 mM Tris-HCl pH 7.4, 2% SDS), twice with RIPA-wash buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP40, 0.1% SDS, 0.4% sodium deoxycholate), once with TNNE buffer (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% NP40, 1 mM EDTA), and three times with 50 mM ammonium bicarbonate (ABC) pH 8.0 buffer. We resuspended the beads in 200 µL of 50 mM ABC buffer containing 3 µg trypsin and incubated at 37 °C overnight with agitation. We added an additional 1.5 µg of trypsin and incubated for a further 3 h. We centrifuged the beads at 400 g for 2 min, collected the supernatant in a new 1.5 mL tube, washed the beads twice with 100 µL mass spectrometry grade H₂O (pelleting beads in between), and pooled the wash supernatant with the peptide supernatant. We centrifuged the supernatant at 16,100 g for 10 min and transferred it to a new tube. Finally, we lyophilized the pooled supernatant using vacuum centrifugation without heat for mass spectrometry.

Mass spectrometry

One-quarter of the affinity purified samples digested with trypsin were analyzed using data-dependent acquisition (DDA) LC-MS/MS. We generated nano-spray emitters from fused silica capillary tubing, with 100 µm internal diameter, 365 µm outer diameter and 5–8 µm tip opening, using a laser puller (Sutter Instrument Co., model P-2000, with parameters set as heat: 280, FIL = 0, VEL = 18, DEL = 2000). We packed nano-spray emitters with C18 reversed-phase material (Reprosil-Pur 120 C18-AQ, 3 µm) resuspended in methanol using a pressure injection cell. We directly loaded the sample in 5% formic acid at 800 nl/min for 20 min onto a 100 µm × 15 cm nano-spray emitter. We eluted peptides from the column with an acetonitrile gradient generated by an Eksigent ekspert™ nanoLC 425, and analyzed on a TripleTOF^™ 6600 instrument (AB SCIEX, Concord, Ontario, Canada). We delivered the gradient at 400 nl/min from 2% acetonitrile with 0.1% formic acid to 35% acetonitrile with 0.1% formic acid using a linear gradient of 90 min. This was followed by a 15 min wash with 80% acetonitrile with 0.1% formic acid, and equilibration for another 15 min to 2% acetonitrile with 0.1% formic acid. The total DDA protocol is 120 min. The first DDA scan had an accumulation time of 250 ms within a mass range of 400–1800 Da. This was followed by 10 MS/MS scans of the top 10 peptides identified in the first DDA scan, with accumulation time of 100 ms for each MS/MS scan. We required each candidate ion to have a charge state from 2 to 5 and a minimum threshold of 300 counts per second, isolated using a window of 50 mDa. We dynamically excluded previously analyzed candidate ions for 7 s.

Mass spectrometry data analysis

We analyzed mass spectrometry data from two independent experiments using ProHits. We converted WIFF files to an MGF format using the WIFF2MGF converter and to an mzML format using ProteoWizard (V3.0.10702) and the AB SCIEX MS Data Converter (V1.3 beta). We searched the data using Mascot (V2.3.02) and Comet (V2018.01 rev.4). We searched the spectra with the human and adenovirus sequences in the RefSeq database (version 57, January 30th, 2013) acquired from NCBI, supplemented with “common contaminants” from the Max Planck Institute (http://www.coxdocs.org/doku.php?id=maxquant:start_downloads.htm) and the Global Proteome Machine (GPM; ftp://ftp.thegpm.org/fasta/cRAP), forward and reverse sequences (labeled “gi|9999” or “DECOY”), sequence tags (miniTurbo) and streptavidin, for a total of 72,481 entries. We set database parameters to search for tryptic cleavages, allowing up to 2 missed cleavages sites per peptide with a mass tolerance of 35 ppm for precursors with charges of 2+ to 4+ and a tolerance of 0.15 amu for fragment ions. We selected variable modifications for deamidated asparagine and glutamine and oxidized methionine. We analyzed results from each search engine through TPP (the Trans-Proteomic Pipeline, v.4.7 POLAR VORTEX rev 1) via the iProphet pipeline.

SAINT analysis

We used SAINT version 3.6.3 as a statistical tool to calculate the probability of potential protein-protein associations compared to background contaminants using default parameters from two independent biological replicates⁷⁸. We used a 95% FDR iProphet filter and unique peptides ≥ 2 before running SAINT. We considered proteins with AvgP scores of ≥ 0.9 and BFDR ≤ 0.01 as high-confidence proximal interactions. All non-human proteins (did not start with “NP” in Prey column) were removed from the SAINT analysis, except for the miniTurbo fusion proteins. Dot plots were generated in ProHits-viz^78,138 using SAINTexpress file generated from ProHits, with a minimum Spectral count sum (of two replicates) cutoff of 5, Pearson correlation, and Euclidean distance settings.

Chromatin immunoprecipitation

We washed 2.0 × 10⁷ cells with PBS, crosslinked with 1% formaldehyde for 10 min, then quenched the reaction by adding 125 mM glycine to the media and incubated with shaking for 5 min. We washed cells twice with ice cold PBS, scraped the cells in PBS, and centrifuged at 300 g for 5 min at 4 °C. We resuspended the cell pellet in 5 mL lysis buffer (50 mM Tris, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton X, 1X fresh protease inhibitor), incubated on ice for 10 min, and pelleted nuclei at 600 g 4 °C for 5 min. We resuspended the pellet in 1 mL nuclear lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris, fresh protease inhibitor), incubated for 10 min on ice, and added 0.6 mL dilution buffer (0.01% SDS, 1.1% Triton-X, 1.2 mM EDTA, 16.7 mM Tris, 167 mM NaCl, fresh protease inhibitor) to the nuclear lysate. We sonicated the lysate at medium intensity for 4 cycles of 30 seconds on and 30 seconds off using the Bioruptor Plus UCD300 (Diagenode). We centrifuged the lysate at 20,817 g for 20 min at 4 °C to pellet debris. We moved the supernatant to a new tube, set aside 10% as input, and added 6 µg of MTA2 antibody (Abcam cat #ab8106) or 1 µg of FOSL2 antibody (Cell Signaling Technologies cat #19967) to 20 µg sonicated lysate and incubated on a rotator at 4 °C overnight. We washed 100 µL of protein A/G magnetic beads (Pierce) per sample 3X in 1 mL nuclear lysis buffer, then we added the lysate with bound antibody to the beads and rotated at 4 °C for 5 h. After bead incubation, we carried out 5-minute washes with rotation with each of the following buffers: low salt (0.1% SDS, 1% Triton-X, 2 mM EDTA, 20 mM Tris, 150 mM NaCl), high salt (0.1% SDS, 1% Triton-X, 2 mM EDTA, 20 mM Tris, 500 mM NaCl), LiCl (1% NP-40, 1% sodium deoxycholate, 2 mM EDTA, 20 mM Tris, 0.25 M LiCl), and twice with TE buffer. We eluted with 250 µL elution buffer (1% SDS, 10 mM Tris, 1 mM EDTA) and vortexed every 2 min for a total of 20 min at 65 °C. We transferred the supernatant to a new tube, added 1 µl RNase A (10 mg/mL), and incubated overnight at 65 °C. To reverse crosslink, we added 6 µl Proteinase K (10 mg/mL) and 1.5 µL glycogen (10 mg/mL) and incubated for 2 h at 45 °C. We purified the sample using the NEB Monarch DNA purification kit, using 7 volumes of binding buffer. FOSL2 ChIPseq experiments were performed in duplicate, MTA2 ChIP-qPCR experiments were performed in triplicate. For MTA2 ChIP-qPCR, hCISTR_dCas9ChIP_2 (position 1) and hCISTR_dCas9ChIP_4 (position 2) primer pairs were used to quantify CISTR.

RNA immunoprecipitation

We washed 2.0 × 10⁷ cells with PBS, crosslinked with 1% formaldehyde for 10 min, then quenched the reaction by adding 125 mM glycine to the media and incubated with shaking for 5 min. We washed cells twice with ice cold PBS, scraped the cells in PBS, and centrifuged at 300 g for 5 min at 4 °C. We resuspended the cell pellet in 1.5 mL RIPA lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 5 mM EDTA pH 8.0, 0.5 mM DTT, 1X fresh protease inhibitor, 100 U/mL fresh RNAseOUT) and incubated on ice for 15 min. We sonicated the lysate at medium intensity for 4 cycles of 30 seconds on and 30 seconds off using the Bioruptor Plus UCD300 (Diagenode). We centrifuged the lysate at 20,817 g for 20 min at 4 °C to pellet debris. We moved the supernatant to a new tube, set aside 10% as input, and added 1 µg of FOSL2 antibody (Cell Signaling Technologies #19967) to sonicated lysate and incubated on a rotator at 4 °C overnight. We washed 100 µL of Protein A Mag Sepharose beads (Cytiva) per sample 3X in 1 mL RIPA lysis buffer, then we added the lysate with bound antibody to the beads and rotated at 4 °C for 4 h. After bead incubation, we carried out 4 × 5 minute washes with rotation with 1 mL RIPA lysis buffer. We reverse crosslinked in 250 µL elution buffer (20 mM Tris-HCl pH 8.0, 10 mM EDTA pH 8.0, 1% SDS, 200 mM NaCl, 100 U/mL RNAseOUT) and 6 µL Proteinase K (10 mg/mL) for 1 h at 45 °C and 1 h at 65 °C. FOSL2 RIP-qPCR experiments were performed in triplicate. hCISTR_RIP_qPCR primers were used to quantify CISTR.

FOSL2 ChIPseq analysis

Before sequencing, 10% input and IgG control samples were pooled. Illumina adaptors from paired-end reads (2 × 150 bp) were removed using Trimmomatic¹³⁹. Reads were mapped to the hg38 reference genome using the Burrows-Wheeler alignment tool with default parameters, and reads with low quality and those overlapping with the ENCODE blacklist regions (https://www.encodeproject.org/annotations/ENCSR636HFF/) were filtered out. We kept duplicate reads during peak calling (--keep-dup all), and identified peaks using MACS2¹⁴⁰ version 2.2.9.1, with a q < 0.01 cutoff, and the “-f BAMPE”, “--call-summits” options. We performed differential peak analysis with DESeq2, version 1.44.0¹¹⁹, and we only considered peaks ± 2.5 kb away from annotated genes. We saved bedgraph files using the “--bdg” option and subsequently converted them to bigwig files for visualization in the UCSC GenomeBrowser.

FOSL2 ChIPseq analysis from ChIP-seq Atlas

We downloaded TF ChIP-seq datasets of blood and neural origins from the ChIP-seq Atlas Peak Browser and intersected them with the promoter regions (−2.5 kb of TSS) of our DEGs using bedtools¹⁴¹. We statistically tested the uniquely overlapping genes by Hypergeometric testing in comparison to GENCODE’s annotation.

CISTR and FOSL2 overexpression

We seeded 1.0 × 10⁶ ΔCISTR and WT RPMI 2650 cells in a 6-well plate and transfected a total of 4 µg of DNA and 6.67 µL Lipofectamine 3000 transfection reagent 24 h later. We transfected 160 ng and 1.6 µg of CISTR overexpression EF1α-BbsI-SV40polyA_EGFP for CISTR low and high concentrations, respectively, 240 ng and 2.4 µg of TFORF3083 (Addgene #144559) for FOSL2 low and high plasmid concentrations, respectively, and 4 µg EF1α-BbsI-SV40polyA_EGFP for the control. To compensate for equal molarities of the DNA input, we used pUC19. Cells were analyzed on flow cytometer 48 h post transfection.

Statistics and reproducibility

All statistical tests conducted were two-sided, unless stated otherwise. If multiple tests were carried out on the same data, error rates were corrected for multiple testing using Bonferroni correction or as stated in the results. No statistical method was used to predetermine sample size. For statistical analysis, we used R versions 4.2.1 and 4.4.0 (or as stated in the methods), Python version 3.8, and GraphPad Prism version 10. FACS outliers were omitted by the ROUT method in GraphPad Prism.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization and Funding Acquisition P.G.M.; Methodology and Formal Analysis K.K., K.S., B.J.M., B.I.M., J.D.N., T.D.Y., K.D., C.J.W., P.G.M.; Software K.S., J.J.C., M.M.; Investigation K.K., P.G.M.; Resources M.A., S.S., Y.L., B.J.N., A.C.G., M.J.J., J.L.L., P.G.M.; Writing K.K. and P.G.M.; Review and Editing all authors; Supervision P.G.M.

Peer review

Peer review information

Nature Communications thanks Yangming Wang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available in the Source Data file, the Supplementary Information, public databases or referenced studies. All raw sequence data generated in this study have been deposited at GEO (GSE273476 [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE273476] and GSE273480). Mass-spec data are deposited at the MassIVE repository with the accession number MSV000095145 [https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=93593cfa0ab446fead50314a05bb45e0]. The ProteomeXchange accession is PXD053402. All software used and the parameters for analyzing each type of sequencing data used in this study, are publicly available as described in the Methods. Additional materials can be made available by contacting the corresponding author. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67591-x.