The Ubiquitin Ligase Zinc Finger SWIM Domain‐Containing Protein 8 Regulates Oligodendrocyte Development Through the Argonaute2/MicroRNA‐7 Axis

ABSTRACT

Proteostasis of proteins with intrinsically disordered regions (IDRs) is of particular importance to the development and function of the central nervous system (CNS). The conserved ZSWIM8 ubiquitin ligase, an essential regulator of mammalian brain development, is known to target IDR proteins involved in neuronal cell migration. Here we show that ZSWIM8 is also indispensable for oligodendrocyte maturation and myelination in the CNS. Loss of ZSWIM8 in the brain causes gross accumulation of IDR‐rich proteins including many RNA‐binding proteins (RBPs). Substrate recognition by ZSWIM8 requires its own IDRs, while ZSWIM8‐mediated ubiquitination of AGO2 also depends on microRNA binding. AGO2 stabilization in ZSWIM8‐null tissues disrupts target‐directed microRNA degradation (TDMD) of MiR7, leading to altered gene expressions and myelination defects in vivo. Together, these results not only establish ZSWIM8 as a versatile regulator of IDR proteins but also highlight the crucial roles of RBP/miRNA homeostasis in oligodendrocyte development.

ZSWIM8, an E3 ubiquitin ligase that broadly targets IDR‐rich proteins, drives MiR‐7‐dependent AGO2 degradation and the turnover of many RBPs. Loss of ZSWIM8 leads to MiR‐7 accumulation in oligodendrocyte progenitor cells and myelination defects in the developing brain.

1. Introduction

The structural information of a protein is not only manifested by folded globular domains but is also embedded in intrinsically disordered regions (IDRs). These IDRs are present in most mammalian proteins and play critical roles in mediating conformational changes and protein–protein interactions (Tsang et al. 2020; Shen et al. 2021; Gallego et al. 2020). However, many IDR‐rich proteins are prone to folding errors and susceptible to aggregation, which are often hallmarks of neurodegenerative disorders (Zbinden et al. 2020). To maintain protein homeostasis in an IDR‐abundant protein pool, eukaryotic cells have developed a complex protein quality control (PQC) system, including molecular chaperones and the ubiquitin‐proteasome system (UPS), to facilitate protein (re)folding and to clear detrimental proteins (Kolhe et al. 2023; Pohl and Dikic 2019). In the UPS pathway, proteins targeted for proteasomal degradation are often marked by polyubiquitination catalyzed by E3 ubiquitin ligases. Although a plethora of studies have established the substrate‐enzyme relationship between numerous proteins and E3s (Timms et al. 2023; Zhang et al. 2023), how misfolding‐prone IDR‐containing proteins (IDPs) are specifically targeted remains poorly understood. A limited number of E3 ligases (San1 (Rosenbaum et al. 2011) and Grr1 (Stieg et al. 2018) in yeast, CHIP (Narayan et al. 2011) and WWP2 (Vamadevan et al. 2022) in mammalian cells) are known to directly utilize their IDRs for recognition of IDR‐enriched substrates. This so‐called “disorder‐targets‐misorder” mechanism sets forth a unique principle underlying IDR recognition by E3s. However, the generality of this mechanism of substrate recognition by mammalian E3s, particularly in vivo, has not been rigorously tested.

Previously we identified that EBAX‐1 (Elongin B/C‐Binding Axon Regulator), a novel substrate recognition subunit of Cullin‐RING E3 ubiquitin ligases (CRL), guards neurodevelopment accuracy through quality control of the IDR‐enriched axon guidance receptor SAX‐3 in C. elegans (Wang et al. 2013). Later, we reported that the mammalian ortholog of EBAX‐1, namely ZSWIM8, preferably targets the IDR region of DAB1, an adaptor downstream of REELIN, thereby safeguarding migration of neural progenitor cells (NPCs) in the developing mouse brain (Wang et al. 2023). As opposed to the N‐terminal BC‐box, CUL2‐box, and SWIM motif with known functions, the majority of ZSWIM8 C‐terminal sequence is disordered, with three large IDRs punctuated with ordered sequences. In C. elegans and Drosophila, genetic mutations that create truncated proteins losing all or most of the C‐terminal IDRs caused loss‐of‐function phenotypes, suggesting that these regions are indispensable for the EBAX‐1/ZSWIM8 family (Wang et al. 2013; Kingston et al. 2022). These studies converge on the notion that ZSWIM8 and its homologs may also use the “disorder‐targets‐misorder” mechanism for general substrate binding. Beyond those known targets, ZSWIM8 must have a much broader substrate specificity in mammals since we and others have reported that whole‐body and brain‐specific knockout of ZSWIM8 leads to severe developmental defects and lethality in mice (Wang et al. 2023; Shi et al. 2023; Jones et al. 2023).

Two recent studies uncovered a new role of ZSWIM8 in target‐directed microRNA degradation (TDMD) of MiR7 in various cell lines (Shi et al. 2020; Han et al. 2020) through degrading Argonaute 2 (AGO2), a core component of the RNA‐induced silencing complex (RISC) (Liu et al. 2004; Hutvagner and Simard 2008; Meister 2013). In mice, the microRNA MiR7 is encoded by Mir7a‐1, Mir7a‐2, and Mir7b genes located on three different chromosomes (Zhao et al. 2020; Horsham et al. 2015). Simultaneous suppression of three MiR7 precursors causes defective neurogenesis and microcephaly‐like brain defects at birth (Pollock et al. 2014; Zhang et al. 2018), while upregulation of MiR7 is observed during neuronal differentiation and maturation (Chen et al. 2010; Caygill and Brand 2017).

In contrast to their emerging roles in neuronal cells, the expression dynamics and in vivo function of ZSWIM8 and MiR7 in glial cells remain largely unclear. An early study shows that MiR7 downregulation is important for oligodendrocyte (OL) maturation in vitro (Zhao et al. 2012). We also noted that ZSWIM8 is expressed in cells of OL lineage (see later for details). OLs are specialized glial cells that form myelin sheath around nerve fibers in the brain and provide electrical insulation to increase axonal conduction velocity (Elbaz and Popko 2019). Their proper development is essential for neural circuit function and cognitive health. Oligodendrocyte progenitor cells (OPCs) start to appear at the late embryonic stage and differentiate into mature oligodendrocytes after birth (Hayashi and Suzuki 2019; Richardson et al. 2006; La Manno et al. 2021; Kuhn et al. 2019). Despite extensive studies (Elbaz and Popko 2019; Gaesser and Fyffe‐Maricich 2016; Zhao et al. 2010; Dugas et al. 2010), regulatory mechanisms of the OL maturation process are still incompletely understood. For example, MiR7 expression in neuron is controlled by a long non‐coding RNA (lncRNA) Cyrano and a circular RNA (circRNA) Cdr1as (Piwecka et al. 2017; Kleaveland et al. 2018; Mehta et al. 2023), which are essentially absent in OL lineage cells (Hansen et al. 2013), raising the question of how MiR7 is turned down to allow for OL maturation. How the PQC system (including ZSWIM8) functionally coordinates with OL development is also largely unknown.

Here, we provide genetic, biochemical, and multi‐omic data demonstrating that ZSWIM8, being a master regulator of IDR‐enriched proteins, is required for OL development in vivo. ZSWIM8 induces AGO2 ubiquitination and degradation in a miRNA‐binding‐dependent manner, which in turn triggers TDMD of MiR7, ensuring OL maturation and myelination. These findings have expanded our understanding of substrate recognition mechanisms of IDR‐containing E3 ubiquitin ligases and uncovered the pivotal role of such regulatory machinery in vivo.

2. Results

2.1. ZSWIM8 Deletion Impairs Oligodendrocyte Maturation and Myelination

To determine the role of ZSWIM8 in CNS myelination, we first examined myelin basic protein (MBP) expression in brain sections from wild‐type (WT) and ZSWIM8 conditional knockout (cKO) mice. Coronal brain sections from postnatal day 14 (P14) Zswim8 ^flox/flox and Zswim8 ^flox/flox ;Nestin‐Cre littermates (hereafter referred to as Z ^f/f and Z ^f/f ;N‐Cre, respectively) were prepared for immunohistochemistry (IHC) analysis. In the mutant brains, a pronounced reduction in the overall thickness of the corpus callosum was observed (Figure 1A), and a significant decrease in the MBP‐positive myelinated area within this region was revealed by imaging at higher magnification (Figure 1B–D). Similarly, defects in myelin sheath were also seen in the S1 cortex of Zswim8‐null brains (Figure S1A–C). During postnatal development, the OPC‐specific marker PDGFRA reaches the peak level at P7 in the forebrain then quickly drops (Figure 1E), concurrent with a gradual increase of markers for premyelinating (CNP) and mature OL markers (MBP) (Figure 1E). Although PDGFRA protein levels were comparable between P14 Z ^f/f and Z ^f/f ;N‐Cre forebrains, both CNP and MBP levels were significantly reduced in the mutants (Figure 1F,G). These initial findings suggest that ZSWIM8 may be required for proper oligodendrocyte maturation and function in the neonatal brain.

FIGURE 1.

ZSWIM8 is required for oligodendrocyte maturation and myelination. (A) Immunohistochemistry (IHC) of MBP in coronal brain sections from Z ^f/f and Z ^f/f ;N‐Cre littermates (P14). The corpus callosum is outlined by dotted lines based on DAPI staining (not shown). Same exposure conditions for MBP images were used for control and mutant groups. Scale bar, 100 μm (top). Magnified images of the corpus callosum in the boxed medial regions are shown at the bottom (scale bar, 20 μm). (B) High magnification images (60×) of anti‐MBP IHC showing myelinated areas in the corpus callosum. Scale bar, 20 μm. (C) Quantification of the thickness of the corpus callosum (marked by white dotted lines in D). Data are presented from three pairs of littermates. *p < 0.05 (Student's unpaired t‐test). The isthmus of the corpus callosum above hippocampus in 8–12 coronal brain sections of each brain sample was used for quantification. (D) Quantification of the myelinated area in the corpus callosum based on MBP intensities in (E). Data are presented from three pairs of littermates. **p < 0.01 (Student's unpaired t‐test). (E) Western blot analysis of PDGFRA (OPC marker), CNP (premyelinating OL marker), and MBP (mature OL marker) in the forebrain from WT mice of the indicated ages. N = 1 for each age. (F) Forebrain samples were collected from three pairs of littermates (Z ^f/f and Z ^f/f ;N‐Cre) at P14 and probed with the indicated antibodies. (G) Quantification of protein levels shown in (B). Data are presented from three pairs of littermates. **p < 0.01 and ***p < 0.001 (Student's unpaired t‐test). (H, I) Corpus callosum tissues from Z ^f/f and Z ^f/f ;N‐Cre animals at P7 (H) and P14 (I) were immunostained for OLIG2 (OL lineage marker), PDGFRA (OPC marker), and CC1 (mature OL marker). Scale bar, 20 μm. (J) The density of OLIG2‐marked OL lineage cells, OLIG2⁺ PDGFRA⁺ OPCs, and OLIG2⁺ CC1⁺ mature OLs in the corpus callosum was quantified in Z ^f/f and Z ^f/f ;N‐Cre animals at indicated stages. N = 3 pairs of littermates. *p < 0.05 and **p < 0.01 (Student's unpaired t‐test). (K) The percentage of OPCs or mature OLs among all OLIG2⁺ cells in the corpus callosum was quantified for each condition. N = 3 pairs of littermates. ***p < 0.001 (Student's unpaired t‐test). (L) Representative transmission electron microscopy (TEM) micrographs of the corpus callosum from P14 Z ^f/f and Z ^f/f ;N‐Cre animals are demonstrated. Scale bar, 10 μm. Magnified images are shown on the right. Scale bar, 1 μm. (M) The numbers of myelinated axons in random 100 μm² areas (N = 5) were quantified in (L) for each littermate. **p < 0.01 (Student's unpaired t‐test). (N) Representative transmission electron microscopy (TEM) images of cross sections of single axons from P14 Z ^f/f and Z ^f/f ;N‐Cre animals are demonstrated. Scale bar, 500 nm. (O) Quantification of the g‐ratio on individual axons (N = 20–30). **p < 0.01 (Student's unpaired t‐test).

We then investigated the effects of ZSWIM8 deletion on oligodendrocyte development in greater detail. Consistent with the profiles described above (Figure 1E–G), the number of differentiating and mature OLs (marked by OLIG2⁺ CC1⁺ double‐positive staining) increased considerably from P7 to P14 in the corpus callosum of WT mice, which did not happen in Z ^f/f ;N‐Cre mice (Figure 1H–K). During this period, total OLIG2⁺ cells remained comparable in number and density between controls and KO samples, whereas OPCs (OLIG2⁺ PDGFRA⁺) actually accumulated at a higher density at P7 in the Z ^f/f ;N‐Cre tissues than in the WT. These results strongly indicate a maturation defect in the oligodendrocyte lineage when ZSWIM8 was missing. A similar reduction in mature OLs was observed in deep layers of the S1 cortex in Z ^f/f ;N‐Cre mice (Figure S1D–F). We further examined ultrastructural changes in myelination in P14 corpus callosum using transmission electron microscopy (TEM). Both the number of myelinated axons (Figure 1L,M) and the thickness of myelin sheath (as indicated by the G‐ratios, Figure 1N,O) were significantly reduced in Z ^f/f ;N‐Cre mutant brains. Together, our results from multiple assays indicate that deletion of ZSWIM8 impedes oligodendrocyte maturation, leading to hypomyelination in neonatal brains.

2.2. Deletion of ZSWIM8 in the Nervous System Causes Broad Accumulation of IDPs

As an E3 ubiquitin ligase component, ZSWIM8 is expected to regulate myelination via turnover of key substrates in the developing brain. We therefore analyzed the whole proteomes of forebrain tissues dissected from Z ^f/f and Z ^f/f ;N‐Cre littermates at P14 (Figure 2A; Table S1). Gene ontology (GO) analyses indicated that in Z ^f/f ;N‐Cre brains, numerous proteins involved in RNA metabolism (e.g., ELAV1, AGO2, XRN2 and PRP6) were upregulated, whereas various proteins required for neuronal and glial functions (e.g., SYN1, NEUG, MBP, and CNP) were downregulated (Figure 2A–C and Figure S2A,B). The differential expression of several proteins was further confirmed by western blot (Figure 2D). Further GO analyses underscored the upregulation of RNA‐binding proteins (RBPs) and downregulation of components required for brain development in the absence of ZSWIM8 (Figure 2E,F). Moreover, endogenous ZSWIM8 could co‐immunoprecipitate with RBPs such as AGO2 and ELAV1 from mouse brain tissues when protein degradation was blocked by intracranial injection of the proteasome inhibitor, Bortezomib (Figure S2C). These results not only provide proteome‐level evidence supporting our previous findings on the critical role of ZSWIM8 in neurodevelopment, but also suggest a broader requirement for ZSWIM8 in controlling the levels of RBPs.

FIGURE 2.

Deletion of ZSWIM8 in the mouse nervous system causes accumulation of proteins with IDRs. (A) P14 forebrain homogenates from Z ^f/f and Z ^f/f ;N‐Cre animals (N = 3 pairs of littermates) were analyzed by mass spectrometry. Representatives of differentially expressed proteins in Z ^f/f ;N‐Cre mice are indicated by arrows. Examples of RNA‐binding proteins in the upregulated pool and proteins involved in neuronal and glial functions in the downregulated pool are highlighted by yellow dots and cyan dots, respectively. Mass spectrometry data of Z ^f/f and Z ^f/f ;N‐Cre samples were analyzed by the intensity‐based absolute quantification (iBAQ) method and plotted. (B, C) Gene ontology (GO) analyses of upregulated (B) and downregulated proteins (C) in Z ^f/f ;N‐Cre animals at P14. Five enriched biological processes in each group are shown. (D) Western blot verification of the indicated proteins identified from the proteomic study of Z ^f/f and Z ^f/f ;N‐Cre samples at P14. AGO2, argonaute RISC catalytic component 2; CNP, cyclic nucleotide phosphodiesterase; ELAV1, ELAV‐like RNA binding protein; MBP, myelin basic protein; PRP6, pre‐mRNA processing factor 6; XRN2, 5′–3′ exoribonuclease 2. β‐tubulin was used as loading control. (E, F) GO term analyses of proteins upregulated (E) or downregulated (F) in Z ^f/f ;N‐Cre animals at P14 stages. Seven enriched GO terms in each group are shown. CC, cellular component; MF, molecular function. Color codes stand for Benjamini–Hochberg‐adjusted p‐values. (G–J) IDR contents of proteins identified in the proteomic study were analyzed by the three indicated algorithms (VXLT, VSL2b, and PrDOS). IDR‐rich proteins (IDR > 50%, G) and IDR‐poor proteins (IDR < 5%, I) were first calculated for their respective percentages in each sample group (P14_total, P14_up, or P14‐down) then normalized against their percentages in P14‐total. Proteins with various numbers of large IDRs were predicted and analyzed in the same way (H, J). ***p < 0.001 relative to P14_total (one‐way ANOVA with multiple comparisons). ^### p < 0.001, relative to P14_up (student's paired t‐test).

The fact that RBPs are rich in IDRs (Castello et al. 2012, 2016; Hentze et al. 2018), together with our earlier studies on SAX‐3, ROBO3 and DAB1 (Wang et al. 2013, 2023), led us to hypothesize that ZSWIM8 preferentially recognizes and degrades IDPs. Indeed, among proteins that were upregulated in ZSWIM8 knockout P14 forebrains (“P14_Up”), a higher percentage of IDR‐rich proteins was observed than in the total forebrain proteome (“P14_total”), as demonstrated by three IDR predictor algorithms (VLXT, VSL2b and PrDOS) from the D²P² database (Oates et al. 2013) (Figure 2G). The P14_Up group also contained a greater proportion of proteins with ≥ 3 large IDRs (defined as disordered sequences ≥ 30 amino acids long) (Zarin et al. 2019; van der Lee et al. 2014) than the total proteome (Figure 2H). Correspondingly, IDR‐poor proteins were less represented in the P14_Up group (Figure 2I,J). On the contrary, more IDR‐poor and fewer IDR‐rich proteins were found in the downregulated proteins in ZSWIM8‐null P14 forebrain (“P14_Down”) as compared to P14_Up (Figure 2G,I). We also noticed that IDPs (IDR% > 5%) made up a greater proportion of the P14 forebrain proteome (P14_total) than of the reference mouse proteome (“mmu”) (Oates et al. 2013) (Figure S2D; see Section 4 for details). These results from global proteomic analysis suggest that IDPs in general may have important functions in neo‐/post‐natal development of the brain, and their biased accumulation in the mutant tissues supports the role of ZSWIM8 as a key regulator of IDPs in the brain.

2.3. ZSWIM8 Targets ELAV1 and AGO2 for Degradation by Different Means

Upregulation of ELAV1 and AGO2 in ZSWIM8‐null mice and their physical interaction with endogenous ZSWIM8 in mouse brain (Figure 2D and Figure S2C) suggest that they could be direct substrates of CRL^ZSWIM8, in which ZSWIM8 functions as a substrate recognition subunit (Figure 3A,B) (Wang et al. 2013). ELAV1 (also known as human antigen R, HuR) regulates pri‐miRNA processing including that of MiR7 (Choudhury et al. 2013), and is a typical IDR‐enriched protein (Figure S3A) just like known ZSWIM8 substrates such as SAX‐3, ROBO3 and DAB1. Indeed, overexpression of WT ZSWIM8 evidently enhanced ELAV1 ubiquitination and promoted its degradation in a cycloheximide (CHX) chase assay in HEK293FT cells (Figure S3B–E). By contrast, ELAV1 was considerably stabilized by the ZSWIM8‐ΔBox mutant, which lacks both the BC‐box and the CUL2‐box and cannot form a competent E3 complex (Figure 3A,B and Figure S3D,E). In addition, although ELAV1 is predominantly a nuclear protein, it can shuttle to the cytoplasm (Fan and Steitz 1998) and acute sodium arsenite treatment triggered its retention in cytoplasmic stress granules, where it nicely colocalized with WT ZSWIM8 but not the ZSWIM8 ΔIDR mutant (Figure S3F,G). Since sodium arsenite induces oxidative stress and protein misfolding (Arimoto et al. 2008; Xu et al. 2023), the above colocalization results are consistent with the established function of ZSWIM8 in the quality control of IDPs. Moreover, these results underscore the importance of C‐terminal IDRs of ZSWIM8 in ELAV1 binding, supporting the notion that ZSWIM8 uses the “disorder‐targets‐misorder” mechanism for substrate selection.

FIGURE 3.

ZSWIM8 promotes AGO2 degradation in an IDR‐ and miRNA‐dependent manner. (A) Schematic representation of the CRL^ZSWIM8 E3 ubiquitin ligase complex. Conserved BC‐box, CUL2‐box, and SWIM motif in the N‐terminus of ZSWIM8 are labeled, whereas the function of the remainder of ZSWIM8 has not been previously characterized. The BC box and CUL2 box mediate interaction with Elongin B/C and CUL2, respectively. (B) Schematic diagrams highlighting the IDRs of ZSWIM8, including their positions and sequence similarities among human, mouse, and worm orthologs (left). Truncation mutants of mouse ZSWIM8 used in this study are shown on the right. (C) Immunoblotting of endogenous ZSWIM8 and AGO2 in the forebrain of Z ^f/f and Z ^f/f ;N‐Cre animals at P14. Equal amounts of lysates were used for immunoblotting. Tubulin was used as the loading control. Representative images were shown from four pairs of littermates. (D) AGO2 levels as determined in (C) were quantified from four pairs of littermates. ***p < 0.001 (Student's unpaired t‐test). (E) IDR prediction of mouse AGO2 by five algorisms (http://www.pondr.com). The thick black line indicates an IDR only detected by the VL3‐BA algorithm, which was trained on both ordered and disordered proteins. This IDR covers 43 amino acid residues (a.a. 247–289) in the PAZ domain of AGO2. (F) HEK293FT cells were transfected with the indicated constructs and treated with 50 μg/ml cycloheximide (CHX) for 0, 1.5, and 3 h before harvest. Equal amounts of lysates were blotted with anti‐FLAG, anti‐HA, and anti‐tubulin antibodies. Representative images were shown from four independent experiments. (G) Four independent experiments as in (F) were used for quantification. *p < 0.05 relative to the indicated control (Student's unpaired t‐test). (H) HEK293FT cells were transfected with the indicated constructs and treated with bortezomib (5 μM, 4 h). FLAG‐HA‐AGO2 was immunoprecipitated by anti‐HA nanobody beads, and polyubiquitination signals were detected by anti‐Myc antibodies. Representative images were shown from four independent experiments. (I) Four independent experiments as in (H) were used for quantification. *p < 0.05 relative to the indicated control (one‐way ANOVA). (J) Structural superposition of the hAGO2‐MIR20a complex (PDB: 4F3T, pink) and hAGO2‐MIR27a bound to target RNA (PDB: 6MFN, blue) shows considerable flexibility of the PAZ domain upon mRNA binding. (K) Polyubiquitination of FLAG‐HA‐AGO2 in the presence of FLAG‐ZSWIM8 WT or FLAG‐ZSWIM8 ΔIDR was examined as in (H). Representative images were shown from four independent experiments. (L) Four independent experiments as in (K) were used for quantification. *p < 0.01 and **p < 0.01 relative to the indicated control (one‐way ANOVA). (M) Immunocytochemistry (ICC) staining of V5‐tagged ZSWIM8 (WT or ΔIDR) and HA‐tagged AGO2 in HUVEC cells. Scale bar, 10 μm. Sodium arsenite (AS, 500 μM, 30 min) was used to induce oxidative stress and formation of cytosolic granules. Zoom‐in images of the boxed regions are shown on the right. Scale bar, 2 μm. (N) Structural basis for the critical role of AGO2‐Y529 in miRNA binding (PDB: 4F3T). The highly conserved Y529 residue locates in the miRNA‐binding pocket of AGO2. The Y529E mutation that mimics phosphorylation at this site strongly inhibits loading of miRNA into AGO2. (O) HEK293FT cells were transfected with indicated constructs and treated with bortezomib (5 μM, 4 h). FLAG‐HA‐AGO2 was immunoprecipitated by anti‐HA nanobody beads. Polyubiquitination of AGO2 variants was determined by anti‐Myc antibodies. Representative images were shown from three independent experiments. KE, K525E; YE, Y529E. (P) AGO2 ubiquitination as in (O) was quantified from three independent experiments. **p < 0.01 (Student's unpaired t‐test).

ZSWIM8 has been reported to target AGO2 for degradation in various cell lines, which initiates TDMD of MiR7 (Shi et al. 2020; Han et al. 2020). We also found that AGO2 was consistently upregulated in Z ^f/f ;N‐Cre brain tissues (Figures 2D and 3C,D). However, different from IDR‐enriched ELAV1, AGO2 is organized into several well‐folded domains (Schirle and MacRae 2012) and only the PAZ domain shows a weak sign of disorder (Figure 3E). Direct evidence for AGO2 ubiquitination by CRL^ZSWIM8 and the molecular details thereof have been lacking. We found that, as seen with ELAV1, the BC‐box and CUL‐2 box of ZSWIM8 were also required for AGO2 degradation in HEK293FT cells (Figure 3F,G). However, simple overexpression of WT ZSWIM8 only marginally increased AGO2 ubiquitination (Figure 3H,I). This is distinct from the case of ELAV1 (Figure S3B,C), implying different recognition mechanisms for these two substrates.

When bound to miRNA alone, the PAZ domain of AGO2 caps and protects the 3′ end of the miRNA. Target mRNA binding causes a large movement of the PAZ domain in the ternary complex, releasing the miRNA 3′ end to facilitate base‐pairing with the mRNA (Figure 3J) (Schirle and MacRae 2012; Elkayam et al. 2012; Sheu‐Gruttadauria et al. 2019). We thus hypothesized that RNA binding‐dependent conformational change of AGO2 could create/expose cryptic structural features that can be recognized by ZSWIM8 for ubiquitination.

In agreement with this idea, overexpression of MiR7a significantly stimulated AGO2 ubiquitination by ZSWIM8 in a ZSWIM8^IDR‐dependent manner (Figure 3H,I,K,L). The IDRs were also required for ZSWIM8 to colocalize with AGO2 within cytoplasmic granules upon arsenite treatment (Figure 3M). To further examine the role of RNA binding in AGO2 ubiquitination and degradation, we took advantage of two RNA‐binding mutants of AGO2, K525E and Y529E (Figure 3N). The AGO2 Y529E mutant, which cannot bind miRNA (Quevillon Huberdeau et al. 2017; Schirle et al. 2014), failed to be ubiquitinated by ZSWIM8 beyond the background level, even in the presence of MiR7 (Figure 3O,P). The K525E mutation, which severely impairs target mRNA binding without affecting miRNA association (Golden et al. 2017), had no discernable effects on AGO2 ubiquitination by ZSWIM8 (Figure 3O,P). Therefore, we propose that AGO2 association with miRNA, but not target mRNA, is a prerequisite for its ubiquitination and degradation mediated by ZSWIM8, which in turn deprotects miRNAs (e.g., MiR7). Such miRNA dependency represents a previously unappreciated mechanism of substrate recognition by quality‐control E3s such as ZSWIM8.

2.4. ZSWIM8 Regulates miRNA Expression in the Developing Brain

The upregulation of AGO2 and ELAV1 in the absence of ZSWIM8 prompted us to examine our proteomic results for other experimentally identified miRNA‐binding proteins (miRBPs) from a previous study (Treiber et al. 2017). Notably, we found that over a third (65/180) of those known miRBPs were upregulated in the Z ^f/f ;N‐Cre brain at P14 (Figure S4A). Furthermore, small RNA sequencing (sRNA‐Seq) identified 10 miRNAs consistently upregulated in ZSWIM8 knockout brains (Figure 4A,B; Tables S2 and S3). Seven of these (MiR758‐3p, MiR376b‐3p, MiR409‐3p, MiR543‐3p, MiR154‐3p, MiR369‐3p, and MiR431‐5p) are products of genes clustered at the Chr. 12F1 locus of the mouse genome, possibly co‐transcribed in a manner negatively controlled by ZSWIM8 (Figure S4B; Tables S2 and S3). On the other hand, members of the MiR7 family (MiR7a‐1, MiR7a‐2 and MiR7b) are located on three separate chromosomes (Figure S4B), and their concurrent upregulation in the absence of ZSWIM8 likely resulted from post‐transcriptional regulation. We examined the spatial and temporal expression of MiR7a, which was among the top 30 most abundant microRNAs in P2 mouse brain (Figure 4C). Using the miRNAscope technique, we established that MiR7 was widely present in the neonatal brain and highly expressed in the cerebral cortex (Figure 4D), consistent with previous results (Zhao et al. 2020; Kleaveland et al. 2018). Conditional deletion of ZSWIM8 in the embryonic brain by Nestin‐driven Cre caused upregulation of overall MiR7 intensity, especially in the S1 cortex and the corpus callosum (Figure 4D,E). This phenotype is consistent with the small RNA‐Seq results and has also been confirmed by qRT‐PCR (Figure S4C,D). Furthermore, the level of MiR7a was highest at P0, rapidly dropped to less than 50% at P7, and remained largely steady afterwards (Figure 4F). It is noteworthy that ZSWIM8 protein level in the brain showed an opposite trend from P0 to P7 (Figure 4G). The miRNAscope data further revealed brain region‐specific differences of MiR7a dynamics in the S1 cortex and the corpus callosum at P2, P7, and P14 (Figure 4H). Such spatiotemporal dynamics supports a regulatory function of MiR7 in brain development, which could be modulated by ZSWIM8.

FIGURE 4.

Deletion of ZSWIM8 causes upregulation of a subset of microRNAs in the neonatal mouse brain. (A) Whole brain tissues from three pairs of littermates (Z ^f/f vs. Z ^f/f ;N‐Cre) at P2 were collected for small RNA‐seq analysis. The x‐axis represents mean miRNA expression levels in Z ^f/f brains. The y‐axis represents log₂‐transformed fold‐change (FC) of the corresponding miRNAs in Z ^f/f ;N‐Cre brains (CKO). Highlighted in orange are miRNAs with significantly increased expression in Z ^f/f ;N‐Cre animals (p‐adjust < 0.05) and a mean expression level > 1000 FPKM in Z ^f/f . (B) Hierarchic clustering of all miRNAs significantly upregulated in Z ^f/f ;N‐Cre animals (p‐adjust < 0.05). Data from three pairs of littermates are demonstrated. Blue and orange colors represent low and high expression levels, respectively. (C) Mean expression levels of the top 30 most abundant miRNAs in P2 Z ^f/f whole brains (N = 3). (D) Detection of MiR7a in coronal brain sections by the miRNAscope in situ hybridization assay from P1 littermates (Z ^f/f and Z ^f/f ;N‐Cre) by MiRNAscope. The intensity of signals is illustrated by pseudo color. Scale bar, 100 μm. (E) Representative miRNAscope images of MiR7a in the S1 cortex (top, scale bar = 20 μm) and the corpus callosum (bottom, scale bar = 10 μm) of Z ^f/f and Z ^f/f ;N‐Cre animals. Nuclei are shown by DAPI staining (blue). The miRNAscope in situ hybridization assay allows visualization of MiR7a in cytosolic granules surrounding DAPI‐labeled nuclei. (F) qRT‐PCR quantification of MiR7a levels in whole brains from wild‐type (WT) C57BL/6J mice along the indicated time course. Brain tissues from three mice were collected for each time point, and MiR7 levels were normalized against the mean at P0. *p < 0.05 relative to P0 (Student's unpaired t‐test). (G) The protein level of ZSWIM8 in WT forebrain samples at the indicated stages was determined by immunoblotting. E15, embryonic day 15. (H) Dynamics of MiR7a expression detected by miRNAscope in the S1 cortex (top, scale bar = 20 μm) and the corpus callosum (bottom, scale bar = 10 μm) of WT mice at the indicated neonatal stages. Nuclei are shown by DAPI staining (blue).

2.5. ZSWIM8 Ablation in the Oligodendrocyte Lineage Leads to Hypomyelination

Since Z ^f/f ;N‐Cre mice lack ZSWIM8 in both neurons and glial cells in the CNS, we wondered whether the observed AGO2/MiR7 upregulation and hypomyelination occurred cell‐autonomously within oligodendrocytes. In the corpus callosum and S1 cortex of P14 Zswim8 ^f/f mice, Zswim8 mRNA was detected in both Pdgfra ⁺ and Cnp ⁺ cells by RNAscope (Figure 5A and Figure S5A), but was significantly more abundant in the latter (Figure 5B). Next, we generated a Zswim8 ^f/f ;Cnp‐Cre mouse strain (Z ^f/f ;C‐Cre), in which ZSWIM8 was specifically eliminated in premyelinating oligodendrocytes (pre‐OLs) when the Cnp promoter started to express. Reduced ZSWIM8 expression in Cnp ⁺ but not in Pdgfra ⁺ cells confirmed specific knockout of ZSWIM8 in differentiating oligodendrocytes (Figure 5A and Figure S5A). Notably, Cnp mRNA drastically decreased in Z ^f/f ;C‐Cre mutants compared to Z ^f/f , consistent with downregulation of CNP proteins in the forebrain of P14 Z ^f/f ;N‐Cre mutant animals (Figure 5C). Additionally, in agreement with results from Z ^f/f ;N‐Cre mice, anti‐MBP IHC showed clear hypomyelination in the corpus callosum and S1 cortex of Z ^f/f ;C‐Cre animals at P14 (Figure 5D and Figure S5B). Both the myelinated area in the corpus callosum (Figure 5E) and the length of myelin sheath in S1 deep layers (Figure S5C,D) significantly decreased in Z ^f/f ;C‐Cre brains. These data corroborate that ZSWIM8 is essential for maturation of oligodendrocytes and formation of myelin sheath in a cell‐autonomous fashion.

FIGURE 5.

Conditional deletion of ZSWIM8 in premyelinating OLs impairs OL maturation and causes hypomyelination in the corpus callosum. (A) Representative RNAscope images of Zswim8, Pdgfra and Cnp in the corpus callosum (coronal sections) of P14 Z ^f/f and Z ^f/f ;C‐Cre mice. Scale bar, 50 μm. Zoom‐in images of the boxed regions are shown on the right (scale bar = 10 μm). White arrowheads and arrows point to Pdgfra ⁺ or Cnp ⁺ cells that co‐expressed Zswim8, respectively. (B) Integrated cellular levels of Zswim8 mRNA (AU, arbitrary unit) detected by RNAscope in Cnp ⁺ pre‐OLs and Pdgfra ⁺ OPCs within the corpus callosum of P14 Z ^f/f mice (N = 40 cells). ***p < 0.001 (Student's unpaired t‐test). (C) Integrated cellular levels of Cnp and Pdgfra mRNA detected by RNAscope in the corpus callosum from P14 Z ^f/f and Z ^f/f ;C‐Cre mice in (A) were quantified (N = 40 cells). ***p < 0.001 (Student's unpaired t‐test). (D, E) Immunohistochemistry images of MBP at the midline of the corpus callosum (coronal sections) from P14 Z ^f/f and Z ^f/f ;C‐Cre animals. Scale bar, 20 μm. Quantification of the myelinated areas is shown in (E). N = 3 pairs of littermates. **p < 0.01 (Student's unpaired t‐test).

2.6. MiR7 Acts Downstream of ZSWIM8 to Inhibit Oligodendrocyte Maturation and Function

A previous in vitro study suggested that MiR7 was involved in oligodendrocyte differentiation, during which its level must be tightly controlled (Zhao et al. 2012). We wondered whether MiR7 dysregulation in ZSWIM8‐deleted mouse brain could underly the impaired OL development and hypomyelination in vivo. By combining microRNAscope and IHC staining, we first confirmed the presence of MiR7 signals in the cytosol of OLIG2⁺ oligodendrocyte lineage cells (Figure 6A), which compose the largest cell population in the corpus callosum at neonatal stages (Figure S6A). We then performed gain‐of‐function experiments to demonstrate MiR7 activity in vivo. To this end, we used MiR7 agomir, a chemically modified double‐stranded RNA that can mimic the function of endogenous MiR7. Compared to other common miRNA mimics, agomirs show higher affinity for cell membrane and better stability, particularly suitable for in vivo studies. After validating its expression and efficiency in vitro (Figure S6B–F), we injected control or MiR7 agomir into bilateral ventricles in WT brains at P3, when endogenous MiR7 drastically drops, and then examined OPCs and OLs by IHC after a week (Figure 6B). In the corpus callosum, although total OLIG2⁺ cells and OPCs (OLIG2⁺ PDGFRA⁺) showed no difference between control and MiR7 agomir‐treated animals, the density and percentage of mature OLs (OLIG2⁺ CC1⁺) were significantly reduced by MiR7 agomirs (Figures 6C–E). This result points to an inhibitory function of MiR7 in oligodendrocyte maturation and is consistent with phenotypes of ZSWIM8‐deleted animals. To determine whether MiR7 acts downstream of ZSWIM8 to regulate OL development, we carried out loss‐of‐function studies using MiR7 antagomirs, which are chemically modified single‐strand miRNAs that complement with and suppress endogenous MiR7 activity. Control or MiR7 antagomirs were injected into bilateral ventricles of Z ^f/f and Z ^f/f ;C‐Cre mice at P3 following the same scheme shown in Figure 6B. Maturation of OLs was then examined at P14 by OLIG2 and CC1 staining. In control antagomir‐treated groups, Z ^f/f ;C‐Cre mutant brains that lacked ZSWIM8 again showed a decrease in the thickness of corpus callosum and in the percentage of OLIG2⁺ CC1⁺ mature OLs as compared to Z ^f/f (Figure 6F–H), recapitulating our earlier observations (Figure 1 and Figure S1). The density of mature OLs in Z ^f/f ;C‐Cre animals also showed a subtle decrease compared to Z ^f/f (Figure 6I, right). However, in Z ^f/f ;C‐Cre mice receiving MiR7 antagomirs, the thickness of the corpus callosum was restored to the level of control animals (Figure 6F,G), and the density and percentage of mature OLs were also effectively rescued (Figure 6H,I). In addition, MiR7 antagomirs elevated Cnp expression that was drastically weakened in Z ^f/f ;C‐Cre mice (Figure S6B). These results support our hypothesis that MiR7 is a downstream mediator of ZSWIM8 function. We noticed that, in Z ^f/f controls, MiR7 antagomirs mildly reduced the corpus callosum thickness and increased the total OLIG2⁺ cells (Figure 6G,I), suggesting that the activity of MiR7 must be tightly restricted to an optimal level to ensure proper maturation of oligodendrocytes. Together, these data indicate that dysregulation of MiR7 (among other microRNAs) is a major mechanism mediating the biological effects of ZSWIM8 loss on oligodendrocyte maturation in vivo.

FIGURE 6.

ZSWIM8 supports oligodendrocyte maturation by controlling MiR7. (A) Detection of MiR7 in OLIG2⁺ cells by combined miRNAscope and immunohistochemistry assays in the corpus callosum from WT mice at neonatal stages. OLIG2⁺ cells are the major cell population in the corpus callosum during this time window. MiR7 signals were observed in cytosolic granules surrounding OLIG2 antibody‐marked nuclei. Scale bar, 10 μm. (B) Illustration of the expression profiles of oligodendrocyte lineage markers during development. The experimental design of In Vivo injections of MiR7 agomir is shown at the bottom. (C) WT brains that received control or MiR7a agomir were analyzed on P10 for OLIG2, PDGFRA, and CC1 expression in the corpus callosum by immunohistochemistry. Scale bar, 20 μm. (D) The density of OLIG2⁺ cells, OLIG2⁺ PDGFRA⁺ OPCs, and OLIG2⁺ CC1⁺ mature OLs in the corpus callosum from mice treated with control or MiR7a agomir were quantified. *p < 0.05 (Student's unpaired t‐test). (E) The percentage of OPCs or mature OLs among all OLIG2⁺ cells in the corpus callosum from mice treated with control or MiR7a agomir was quantified. *p < 0.05 (Student's unpaired t‐test). (F) Immunohistochemistry of OLIG2 and CC1 in the corpus callosum (coronal sections) from P14 Z ^f/f and Z ^f/f ;C‐Cre brains that received control or MiR7 antagomir on postnatal day 3 (P3). DAPI stains the nuclei. The corpus callosum is demarcated by white dotted lines. Scale bar, 20 μm. (G) Quantification of the thickness of corpus callosum in (A). N = 3 for each group. *p < 0.05, **p < 0.01, and ^# p < 0.05 (one‐way ANOVA) relative to indicated samples. The body of the corpus callosum in 8–12 coronal brain sections of each brain sample was used for quantification. (H) The percentage of OLIG2⁺ CC1⁺ mature OLs in all OLIG2⁺ cells was quantified (one‐way ANOVA). **p < 0.01 and ^# p < 0.05 (one‐way ANOVA) relative to indicated samples. ns, not significant. (I) Densities of OLIG2⁺ and OLIG2⁺ CC1⁺ cells in the P14 corpus callosum of the indicated groups were quantified. *p < 0.05 and ^# p < 0.05 (one‐way ANOVA) relative to indicated samples. ns, not significant.

2.7. Transcriptomic Analysis Revealed Aberrant Expression of Myelination‐ and Disease‐Related Genes in ZSWIM8‐Deficient Oligodendrocytes

Finally, since dysregulation of microRNAs/miRBPs may broadly perturb the transcriptome, we performed RNA‐Seq in order to fully understand how loss of ZSWIM8 impairs oligodendrocyte development in postnatal brains. O4‐enriched cells (OL lineage) from whole brains of P14 animals were analyzed, and numerous differentially expressed genes (DEGs) were identified in the ZSWIM8‐null samples (Figure 7A; Table S4). Focusing on DEGs with > 2000 read counts, we found 110 of them downregulated and 40 of them upregulated by ZSWIM8 deletion. A considerable portion of these genes (47 downregulated and 15 upregulated) are known to be enriched in OPCs, myelinating oligodendrocytes and newly formed oligodendrocytes isolated from mouse cortex (Figure 7B,C and Figure S7A; Table S5) (Zhang et al. 2014), while the remaining DEGs likely originated from OL lineage cells from other parts of the whole brain. More than 100 of those downregulated genes (Figure 7A) were predicted to be targets of miRNAs that were elevated in ZSWIM8 KO brains, including MiR7 family members (Figure 7D). Moreover, our database and literature search showed that many of these genes have been implicated in various developmental and neurological disorders, and several genes are known to participate in the myelination process (Figure 7D and Figure S7B) (Kaushansky et al. 2010; Vetro et al. 2021; Gothie and Kennedy 2024; Zhou et al. 2012; O'Malley and Isom 2015; Nicolas et al. 2018). Together, the coordinated action of ZSWIM8, AGO2, and MiR7 in OL maturation has illustrated the biological importance of IDR‐directed PQC and microRNA homeostasis in the developing brain (Figure 7E). A better understanding of the operating principles of such systems will bring new insights into the pathogenesis and treatment of related diseases.

FIGURE 7.

Transcriptomic analysis identified dysregulated expression of myelination and disease genes in ZSWIM8‐deficient oligodendrocytes. (A) RNA‐seq analysis of O4⁺ cells isolated from whole brain tissues of three pairs of P14 littermates (Z ^f/+ ;C‐Cre vs. Z ^f/f ;C‐Cre). Relative expression levels of each gene in control samples are shown on the x‐axis, and their log₂‐transformed fold‐change (FC) in Z ^f/f ;C‐Cre brains are shown on the y‐axis. Significantly increased or decreased genes in Z ^f/f ;C‐Cre animals (|log₂FC| > 1, p‐value < 0.05 and p‐adjust < 0.05) with read counts > 2000 in control brains are highlighted in red and cyan, respectively. (B, C) Overlaps between down‐regulated mRNAs (with read counts > 2000 in Z ^f/+ ;C‐Cre samples, B) or up‐regulated mRNAs (with read counts > 2000 in Z ^f/f ;C‐Cre samples, C) and mRNAs highly expressed in myelinating oligodendrocytes, newly formed oligodendrocytes, and OPCs. (D) Chord diagram showing prediction results from TargetScanMouse 7.2, where downregulated mRNAs (as in A) were analyzed for their potential regulation by miRNAs upregulated in ZSWIM8‐null brains using. Genes known to express in the OL lineage (as in B, C) are highlighted in bold. (E) A schematic model (created with BioRender, https://www.biorender.com) demonstrating that ZSWIM8 participates in development of neurons and oligodendrocytes through different mechanisms. In our previous study ZSWIM8 clears misfolded DAB1 and regulates migration of neuronal progenitor cells (Wang et al. 2023). In this study, ZSWIM8 regulates oligodendrocyte development and microRNA stability through its substrate AGO2.

3. Discussion

Critical to the function of CNS, oligodendrocyte differentiation and myelination involve drastic changes in cell morphology, gene expression and proteome reshaping. How transcription regulation and PQC systems are coordinated in this process is an intriguing question. Here, we have contributed several pieces of evidence toward addressing this question. First, the conserved E3 ligase ZSWIM8 is required for OL maturation and myelination, which extends its biological relevance beyond the neuronal compartment in the CNS. Second, our work in mice and human cells demonstrate that the “disorder‐targets‐misorder” mechanism is conserved in mammals, and that CRL^ZSWIM8 is one of the first mammalian E3s known to utilize this mechanism for broad regulation of IDPs. We provided a part list of potential ZSWIM8 substrates that can be further studied in the context of brain development. Third, ZSWIM8 ubiquitinates AGO2 in a manner that depends on AGO2‐miRNA binding, representing an unprecedented mechanism for E3‐substrate recognition. And lastly, our work presented the first in vivo evidence that ZSWIM8‐dependent downregulation of MiR7 in OPCs is essential for OL maturation.

In our previous study, we first reported that Z ^f/f ;Nestin‐Cre mice exhibited severely impaired NPC migration and reduced survival (Wang et al. 2023). Now we show that oligodendrocyte development is also defective in these mice, which may exacerbate brain failure that culminates in perinatal lethality (Wang et al. 2023). The OL differentiation phenotype of Z ^f/f ;Nestin‐Cre mice was fully recapitulated in the Z ^f/f ;Cnp‐Cre mice. The spatiotemporal pattern of ZSWIM8 deletion differs between these two Cre lines, while they produce the same myelination defects, strongly arguing that the phenotype is a direct consequence of Zswim8 ablation rather than artifacts from different Cre activities. We also note that no discernible neuronal cell death was seen in Z ^f/f ;Nestin‐Cre mice (Wang et al. 2023), thereby excluding neuron‐derived damage signals that might indirectly interfere with OL differentiation in this model. Together, these results unambiguously demonstrate that ZSWIM8 functions cell‐autonomously within the OL lineage to secure the myelination process.

Use of the Z ^f/f ;Nestin‐Cre mice, in which ZSWIM8 is broadly deleted in the CNS since early development, facilitated our proteomic studies and identification of the large number of proteins regulated by ZSWIM8. Particularly enlightening was the finding that those proteins upregulated in ZSWIM8‐null brains have higher contents of IDR. These IDPs generally pose a formidable challenge to the quality control system in cells since their disordered nature defies clearance by effectors that recognize finite structures. Rather, a fuzzy mode of substrate recognition has been proposed for quality control enzymes such as the San1 E3 ubiquitin ligase (Caygill and Brand 2017). ZSWIM8 is a new addition to this category of E3 ligases, but unlike previously known ones that only target select substrates, our results suggest that ZSWIM8 is a versatile regulator of a wide array of IDPs. In this sense, ZSWIM8 resembles Hsp90, a molecular chaperone that can cope with a wide range of targets/clients with unstructured sequences (Kolhe et al. 2023; Rosenbaum et al. 2011). It is also interesting to note that the ZSWIM8 homolog in C. elegans, EBAX‐1, can directly bind Hsp90 and form a PQC triage controlling the refolding or degradation of misfolded proteins (Wang et al. 2013).

ZSWIM8 is a large protein of over 1800 amino acids in length. Except for the N‐terminal domains, the remainder of this protein is poorly annotated in terms of structure and function. And yet the number, sizes and spacing of large IDRs in this region are conserved from worm to human, strongly suggesting functional importance (Figure 3B) (Wang et al. 2013). Deletion of these IDRs abolished the activity of ZSWIM8 toward its substrates, in accordance with the “disorder‐targets‐misorder” notion. It will be important to determine the specific contributions of each individual IDR in substrate recognition in the future. On the other hand, we noticed that WT ZSWIM8 itself is relatively unstable, and deletion of either IDRs or the BC Box led to appreciable stabilization of the protein (Figure 3F,K). This result suggests that ZSWIM8 could auto‐ubiquitinate and regulate its own level, considering that it is an IDR‐rich protein. Alternatively, ZSWIM8 might be targeted by other IDR‐directed E3(s), the identity of which would be interesting to find out. Moreover, how ZSWIM8 expression is dynamically controlled during brain development (Figure 4G) and its physiological function in other tissues/organs are all outstanding questions worth pursuing.

Although both ELAV1 and AGO2 can be ubiquitinated by ZSWIM8, they differ in the molecular details. As a typical IDR‐protein, ELAV1 recognition and ubiquitination by ZSWIM8 share a similar PQC mechanism as seen previously with the other ZSWIM8/EBAX‐1 substrates (Wang et al. 2013, 2023). ELAV1 antagonizes processing of miRNAs including MiR7 (Kumar et al. 2017; Lu et al. 2014). ZSWIM8 selectively targets misfolded ELAV1 for degradation to protect the activity of properly folded ELAV1 toward miRNA. In ZSWIM8‐null tissues, normal ELAV1 proteins are either functionally perturbed or diluted by misfolded ELAV1, thus insufficient to keep the level of MiR7 under control.

Distinct from ELAV1, AGO2 is a rather structured protein with weakly disordered properties in the PAZ domain (Figure 3E). Its ubiquitination by ZSWIM8 requires the participation of miRNA as we demonstrated by MiR7 overexpression and the miRNA binding‐deficient AGO2 mutant (Figure 3). This finding adds a new piece of evidence that RNAs, including miRNA shown here and previously reported circRNAs and lncRNAs (Kingston et al. 2022; Huang et al. 2019; Shi et al. 2021; Yoon et al. 2013), can act as important mediators of RBP degradation by the UPS. ZSWIM8 itself has not been categorized as an RBP, and currently there is no evidence indicating direct interaction between ZSWIM8 and miRNAs. Instead, considering the established model of TDMD (Han et al. 2020) and our biochemical data, MiR7‐stimulated AGO2 degradation probably involves conformational change of AGO2 that alters surface charge, exposes cryptic binding sites and/or ubiquitination sites favored by the ZSWIM8 ligase. Nonetheless, we cannot rule out the possibility that other entities of the AGO‐miRNA complex (e.g., other RBPs) in the cell may also contribute to ZSWIM8 binding. Future structural studies are required to fully resolve this issue.

During the preparation of this manuscript, two new studies systematically analyzed changes in miRNA profiles caused by whole‐body ZSWIM8 deletion (Shi et al. 2023; Jones et al. 2023). From a different route focusing on proteomic changes during OL maturation, we arrived at the same conclusion that ZSWIM8 degrades AGO2 in the RISC complex and consequentially deprotects the bound miRNAs, leading to TDMD (Shi et al. 2023, 2020; Jones et al. 2023; Han et al. 2020). Loss of ZSWIM8 disrupts this self‐limiting system, thus causes aberrant expression of miRNAs such as MiR7. All of the upregulated miRNAs we found in the Zswim8 ^f/f ;Nestin‐Cre brains were also shown to be increased in whole‐body ZSWIM8 knockouts (Shi et al. 2023; Jones et al. 2023). Aside from MiR7 that is abundantly expressed in neonatal brains, the other miRNAs upregulated in ZSWIM8‐null tissues (e.g., the Chr12F1 cluster) have not been functionally associated with OL development. The mechanism behind their upregulation (transcriptional or post‐transcriptional) and the biological effects thereof await further investigation. In addition, two well‐studied miRNAs for oligodendrocyte development, MiR‐219 and MiR‐338, were not altered in ZSWIM8 mutant mice in this study and others, suggesting their stability is not regulated by ZSWIM8 (Shi et al. 2023; Jones et al. 2023). These results, together with recent findings from worm, fly and human cell culture systems (Kingston et al. 2022; Sheng et al. 2023; Li et al. 2021; Kingston and Bartel 2021; Donnelly et al. 2022; Stubna et al. 2024) further corroborate the conserved physiological relevance of ZSWIM8 in regulating a specific set of miRNAs.

Multiple studies have reported involvement of MiR7 in brain development and neurological diseases (Zhao et al. 2020). As one of the most abundant miRNAs in the brain, MiR7 functions in various brain regions regulating the expression of several protein‐coding and non‐coding RNAs (Zhao et al. 2020; Kleaveland et al. 2018). Whether ZSWIM8 regulates other developmental processes in the brain via controlling AGO2/MiR7 turnover requires further study. Although no ZSWIM8 mutations have been implicated in human neurological diseases yet, Lessel et al. (2020) reported that patients with heterozygous mutations in AGO2 exhibited brain abnormalities upon MRI examination, with more than half of the patients showing malformations of the corpus callosum. This phenotype strikingly resembles the hypomyelination and callosal thinning we observed in Zswim8 mutant mice. A very recent study by Fan et al. (2023) showed that cKO of AGO2 in Schwann cells resulted in severe demyelination in the peripheral nervous system, again echoing phenotypes of ZSWIM8 deficiency. Therefore, therapeutic strategies aimed at modulating this pathway—for instance, using small molecules to enhance ZSWIM8 activity or antisense oligonucleotides to suppress MiR‐7—may hold promise for treating demyelinating disorders. Unveiling more ZSWIM8 substrates by combining genetic, biochemical and multiomic approaches will greatly advance our understanding of quality control of IDPs in health and disease.

4. Materials and Methods

Data are shown as mean ± SEM unless otherwise noted. Comparison of two groups was performed by two‐tailed unpaired Student's t‐test and multiple comparisons were performed by one‐way ANOVA using GraphPad Prism 6.0. Data are shown as mean ± SEM unless otherwise noted.

All other detailed information of materials and methods can be found in Appendix S1, including mice, plasmids, cell culture, immunocytochemistry, biochemistry, western blot, endogenous co‐immunoprecipitation, immunohistochemistry, RNAscope and microRNAscope, TEM, iBAQ quantitative mass spectrometry and GO analyses, sRNA‐Seq and analysis, O4⁺ oligodendrocytes isolation by magnetic‐activated cell sorting, mRNA sequencing and analyses, computational analysis of IDRs and stereoscopic injection.

Author Contributions

Jing Lei: experimental execution, data analysis, visualization, writing – original draft, writing – review and editing. Siming Zhong: visualization, experimental execution, data analysis, writing – review and editing. Rong Fan: experimental execution, data analysis. Xin Shu: experimental execution, data analysis, writing – original draft. Guan Wang: experimental execution, data analysis. Jiansheng Guo: experimental execution, data analysis. Shuting Xue: experimental execution, data analysis. Luqian Zheng: experimental execution, data analysis. Aiming Ren: resources. Junfang Ji: resources. Bing Yang: resources. Shumin Duan: resources, visualization. Zhiping Wang: conceptualization, supervision, resources, writing – original draft, writing – review and editing. Xing Guo: conceptualization, supervision, resources, writing – original draft, writing – review and editing.

Funding

This work was supported by the National Key Research and Development Program of China grants 2023YFF1204400 (to X.G.) and 2016YFA0501000 (to Z.W.), and the National Natural Science Foundation of China grants 31671039 (to Z.W.) and 32071257 (to X.G.).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Mass spectrometry analyses of Z ^f/f ;N‐Cre and Z ^f/f P14 forebrains. Related to Figure 1. Mass spectrometry data were analyzed by the iBAQ (intensity‐based absolute quantification) method. Both of absolutely intensity and iBAO intensity were shown in the table. Adjusted iBAQ intensity ratios were defined as $\frac{{\text{iBAQ intensity}}_{\mathrm{CKO}}-{\text{iBAQ intensity}}_{\text{Ctrl}}}{{\text{iBAQ intensity}}_{\mathrm{CKO}}+{\text{iBAQ intensity}}_{\text{Ctrl}}}$. A protein with an adjusted iBAQ intensity ratio ≥ 0.15 or ≤ −0.15 was considered up‐ or down‐regulated.

Table S2: Small RNAseq analyses of Z ^f/f and Z ^f/f ;N‐Cre P2 brains. Related to Figure 3. miRNA expression was quantified with Unique Molecular Identifiers (UMI), and DEGseq was used for analysis of differentially expressed genes defined as | log₂(fold‐change) | > 0 and p‐adjust < 0.05 (Benjamini–Hochberg).

Table S3: The detailed information of upregulated miRNAs. Related to Figure 3. The chromosomal localization and aliases of mouse (mmu) miRNAs were obtained from the National Center for Biotechnology Information (NCBI). Normalized abundance data were acquired from small RNA sequencing analyses in Figure 3A.

Table S4: Differential expression genes of mRNA sequencing analyses of O4⁺ cells from Z ^f/f and Z ^f/f ;N‐Cre. Related to Figure 7. O4⁺ cells were isolated from whole brain tissues of three pairs of P14 littermates (Z ^f/+ ;C‐Cre vs. Z ^f/f ;C‐Cre). Differential expression analysis was performed using the DESeq2 R package (1.20.0), and only RNAs with FPKM > 5 in control samples were used for subsequent analyses. The resulting p‐values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate. |log₂FoldChange| > 1, p‐value < 0.05 and p‐adjust < 0.05 were set as the threshold for significantly differential expression.

Table S5: The expression of differential genes in oligodendrocyte lineages. Related to Figure 7. Genes labeled with “−” are downregulated in Z ^f/f ;C‐Cre/+ and those labeled with “+” are upregulated in Z ^f/+ ;C‐Cre/+. Only genes with read counts > 2000 are summarized in the table. Differentially expressed genes labeled “√” are highly expressed in myelinating oligodendrocytes, newly formed oligodendrocytes and oligodendrocyte precursor cells. The reference oligodendrocyte‐lineage cell RNA‐seq database is available at https://brainrnaseq.org. In this database, only genes with FPKM > 20 is considered highly expressed.

Table S6: Key resources table.

Data S1: Supporting information.

Data Availability Statement

microRNA sequencing dataset is available on https://www.ncbi.nlm.nih.gov/sra/PRJNA1008351. Mass spectrometry dataset is available on https://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD043323 or https://www.iprox.cn/page/project.html?id=IPX0006627000. RNA sequencing datasets are available at https://www.ncbi.nlm.nih.gov/sra/PRJNA1138904.