Cytoplasmic lattices store developmentally poised degradative and cytoskeletal complexes in mammalian eggs

Abstract

The cytoplasmic lattice (CPL) in mammalian eggs is essential for early embryonic development but its molecular components, structural organization and functional capacity have remained elusive. Here, using cryo-electron microscopy, we show that the CPL filament in mouse metaphase II eggs contains repeating units with a periodicity of ~37 nm and determine its high-resolution, native structure and complete subunit composition. The CPL architecture organizes maternal-effect proteins, ubiquitination machinery and tubulin into a highly structured reservoir. Maternal-effect proteins form the scaffold of the CPL to sequester a UHRF1–UBE2D3 E3–E2 ubiquitination module and three distinct FBXW–SKP1 E3 ubiquitin ligase components, notably all in activity-excluded states. The CPL further contains αβ-tubulin heterodimers in a GTP-bound state, indicating microtubule-assembly-competent tubulin held in reserve. CPL filaments are capped by a terminal unit that lacks a PADI6 dimer, a scaffold component, suggesting a structural mechanism that prevents further oligomerization. Interactions between neighboring CPL filaments promote the assembly of a three-dimensional network in the egg cytoplasm. Taken together, our work defines how CPL assembly and architecture prime mammalian eggs for ubiquitin-mediated protein degradation and cytoskeletal remodeling during the egg-to-embryo transition.

Li et al. obtain high-resolution, native structures revealing that cytoplasmic lattices in mammalian eggs organize and store degradative and cytoskeletal complexes for early embryonic development.

Main

How cells efficiently organize their interior to support function is a central problem in biology. This challenge is especially critical in large, long-lived cells. Their size, diffusion constraints and long lifespan impose organizational demands far beyond those of typical somatic cells^1–3. Mammalian oocytes are an extreme example. They are among the largest cells in the body and can remain arrested for long periods while maintaining the ability to rapidly undergo meiotic divisions and to support early embryonic development^4,5. How such cells maintain cytoplasmic order, preserve essential molecular activities in a poised state and prevent premature activation of developmental programs remains poorly understood.

Classical models of cellular organization, primarily based on small, rapidly dividing cells, focus on membrane-bound organelles and dynamic cytoskeletal networks. These models do not readily explain how oocytes manage the storage, protection and timely use of vital components over extended timescales. This gap in understanding limits mechanistic insights into key processes that affect reproductive health, such as the causes of age-related decline in oocyte quality and the high rate of aneuploidy in human eggs.

A prominent but poorly understood feature of mammalian oocytes is the cytoplasmic lattice (CPL), a dense fibrillar network that permeates the cytoplasm^6–11. The CPL has been proposed to function as a storage site for essential maternal-effect proteins, which are maternally supplied factors required for early embryogenesis before zygotic genome activation¹². Loss of CPL components, including the maternal-effect proteins PADI6, NLRP5, TLE6 and OOEP^7,9,13,14, results in early embryonic arrest, suggesting its functional importance. However, the molecular architecture of the CPL and the principles by which it contributes to cellular organization and function are only starting to emerge^12,15–18.

Although the organization of CPLs in germinal vesicle (GV)-stage oocytes has been defined to various degrees^15–18 and CPLs persist in metaphase II (MII) eggs¹², their architecture immediately preceding fertilization has not been established. Resolving this state is important for understanding how the CPL may package and organize essential factors in preparation for embryogenesis. However, the structural complexity of the CPL also presents a major technical obstacle, as conventional biochemical isolation approaches are likely to perturb its native architecture, disrupt interfilamental contacts and dissociate more weakly associated factors.

To define the final structural state of the CPL before embryogenesis, we determined its native structure in fertilizable MII eggs. To preserve native CPL architecture, zona pellucida (ZP)-free eggs were deposited onto cryo-electron microscopy (cryo-EM) grids and rapidly ruptured mechanically to generate a thin cytoplasmic layer suitable for imaging. Using single-particle cryo-EM, we determined the structure of the complete CPL building block at 3.5-Å resolution, defined its assembly principles and gained insight into its potential function. The CPL is formed by linear stacking of repeating units with a core composed of maternal-effect proteins, ubiquitination machinery and αβ-tubulin heterodimers. Notably, we found that a UHRF1–UBE2D3 ubiquitination module and three distinct FBXW–SKP1 E3 ubiquitin ligase components are sequestered in their inhibitory states within the CPL. Additionally, the CPL also stores microtubule-assembly-competent, GTP-bound tubulin heterodimers. Preservation of interfilament interactions in our preparations allowed us to define how these higher-order contacts drive CPL fiber assembly. Together, these findings establish the CPL as a structured maternal reservoir of proteostatic and cytoskeletal resources for early embryonic development.

Results

Native structure of the CPL in MII mouse eggs

To understand the ultrastructural organization of the CPL in fully mature eggs, we first characterized mouse MII eggs by negative-stain transmission EM (TEM). In our TEM micrographs of ultrathin sections, we consistently observed highly abundant CPL structures broadly distributed throughout the cytoplasm (Extended Data Fig. 1a). Morphologically, these assemblies constitute a highly ordered fibrous network composed of stacked, lattice-like filament arrays. Quantification of individual CPL units within these arrays demonstrates a wide distribution of lengths, with a median of approximately 340 nm and some long filaments extending beyond 1.2 μm (Extended Data Fig. 1b). This macroscopic architecture and the filament length distribution in MII eggs are highly analogous to the lattice structures previously observed in immature oocytes¹². These observations confirm that the extensive CPL framework is robustly maintained during meiotic maturation, persisting as a dominant ultrastructural feature in the mature egg.

Extended Data Fig. 1. Macroscopic organization and quantification of the CPL network in mouse eggs.

a, Representative TEM micrographs of ultrathin sections from intact mouse eggs. Similar CPL structures were observed in sections from three independently imaged eggs. The right panel is a magnified view of the boxed region in the left panel. Scale bars, 600 nm (left) and 300 nm (right). b, Quantification of the lengths of individual CPL filaments measured from the TEM micrographs. The violin plot shows the distribution of filament lengths in eggs. The solid line indicates the median, and the width of the violin represents the relative density of measurements at each length. Length measurements were obtained from 465 filaments across 5 TEM micrographs from three eggs.

Source data

To further resolve the molecular structure of CPL in mature eggs at high resolution, we developed a purification-free cryo-EM pipeline to visualize the endogenous CPL in the native context of MII eggs. As shown in Fig. 1a, ZP-free MII eggs were directly deposited onto cryo-EM grids. Following a rapid, blotting-based rupture step directly on the EM grid, the egg cytoplasm formed a thin, electron-transparent film across the grid holes, followed by immediate vitrification within seconds (Fig. 1a and Extended Data Fig. 2a). Raw cryo-EM micrographs of these preparations revealed a dense network of distinct filament-like structures, confirming the preservation of the CPL (Fig. 1b). Untilted, high-resolution micrographs were then collected for structure determination by the single-particle cryo-EM approach.

Fig. 1. Native cryo-EM structure of the CPL from mouse MII eggs.

a, Schematic of the purification-free cryo-EM workflow. ZP-free MII eggs were applied directly to EM grids and mechanically ruptured on-grid to generate a thin layer of cytoplasmic content for single-particle cryo-EM analysis. b, Top, representative cryo-EM micrograph showing native CPL filaments. A total of 44,044 movies were collected from five independent datasets, with similar CPL filament features observed across datasets. Arrowheads indicate individual filaments. Bottom, representative 2D class averages. Scale bar, 100 nm. c, Reconstruction of the CPL filament reveals an extended filamentous architecture with a periodicity of ~37 nm. One repeating unit is outlined. Right, schematic representation of the periodic organization of the filament. d, Density map of the CPL repeating unit shown in two views. One ASU is highlighted in color. e, Atomic model of the CPL ASU shown in two orthogonal views, with components colored and labeled. Image in panel a created in BioRender; Li, Y. https://biorender.com/lxqujcz (2026).

Extended Data Fig. 2. Cryo-EM sample preparation and single-particle data processing workflow for the native CPL.

a, Representative intermediate-magnification cryo-EM montage showing egg cytoplasm spread into a thin film across the grid square. Similar features were observed across five independent samples. b, Single-particle cryo-EM data processing workflow.

Initial three-dimensional (3D) reconstruction demonstrated that the CPL adopts an extended filamentous organization composed of regularly repeating structural segments that exhibit an apparent C₂ symmetry and a periodicity of ~37 nm (Fig. 1c). The repeating unit of the CPL filament can in principle be defined in two ways on the basis of two distinct local two-fold symmetry axes along the filament (denoted symcenter 1 and 2 in Extended Data Fig. 3a). To identify the biologically relevant assembly unit, we performed 3D classification analysis centered on each symcenter separately. Classification on the assembly unit defined by symcenter 1 yielded a heterogeneous population, including a major class where half of the assembly was completely absent, indicating an unstable unit. By contrast, the symcenter 2 assembly unit consistently yielded structurally complete classes, which differed only by modest conformational flexibility. This is consistent with contact-surface analysis, which confirmed that the symcenter 2 unit’s interface is double that of the symcenter 1 unit (Extended Data Fig. 3b). Consequently, we define this segment as the biological repeating unit of the CPL (Fig. 1c). Because this repeating unit contains two related halves arranged by local C₂ symmetry, we further extracted the minimal nonredundant building block, defined as the asymmetric unit (ASU). We obtained a consensus reconstruction of the ASU at an overall resolution of 3.5 Å (Fig. 1d). Subsequent focused refinement of the functional units within the ASU further improved the map quality, yielding local resolutions up to 3.0 Å (Tables 1 and 2 and Extended Data Fig. 2b).

Extended Data Fig. 3. Focused classification analyses during CPL structure determination.

a, Classification results for defining the biologically relevant CPL repeating unit. Two local twofold symmetry axes (symcenter 1 and symcenter 2) can be defined along the filament, and focused classification was performed on both candidate C2 assembly units. b, Contact-surface analysis of the two alternative repeating-unit definitions. The symcenter-2 assembly exhibits a doubled buried interface size. c-e, Focused classifications performed on different regions (c, NLRP14’ + UHRF1’ + UBE2D3’; d, FBXW19–SKP1 + FBXW21–SKP1; e, PADI6 dimer 5) of the CPL map. Percentages indicate the fraction of particles assigned to each class.

Table 1.

Data collection and refinement statistics

	CPL ASU (EMD-76334) (PDB 12DL)	CPL ASU consensus (EMD-76335)	CPL filament (EMD-76333)	Local refinement (EMD-76315)	Local refinement (EMD-76321)	Local refinement (EMD-76322)
Data collection and processing
Magnification	81,000	81,000	81,000	81,000	81,000	81,000
Voltage (kV)	300	300	300	300	300	300
Electron exposure (e− per Å2)	50	50	50	50	50	50
Defocus range (μm)	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5
Pixel size (Å)	1.068	1.068	1.068	1.068	1.068	1.068
Final Micrographs (no.)	44,044	44,044	44,044	44,044	44,044	44,044
Initial particle images (no.)	491,871	491,871	491,871	491,871	491,871	491,871
Final particle images (no.)	249,541	249,541	290,887	249,541	249,541	249,541
Symmetry imposed	C1	C1	C1	C1	C1	C1
Map resolution (Å)	3.52	3.52	8.63	3.29	3.10	3.03
FSC threshold	0.143	0.143	Nyquist	0.143	0.143	0.143
Refinement
Model resolution (Å)	3.3
FSC threshold	0.5
Model composition
Nonhydrogen atoms	133,305
Protein residues	16,740
Ligands	16
B factors (Å2)
Protein	118
Ligand	84
R.m.s.d.
Bond lengths (Å)	0.04
Bond angles (°)	0.6
Validation
MolProbity score	1.5
Clashscore	4.1
Poor rotamers (%)	0
Ramachandran plot
Favored (%)	95.03
Allowed (%)	4.88
Disallowed (%)	0.09

Table 2.

Data collection and refinement statistics

	Local refinement (EMD-76323)	Local refinement (EMD-76324)	Local refinement (EMD-76325)	Local refinement (EMD-76326)	Local refinement (EMD-76327)	Local refinement (EMD-76330)	Local refinement (EMD-76331)
Data collection and processing
Magnification	81,000	81,000	81,000	81,000	81,000	81,000	81,000
Voltage (kV)	300	300	300	300	300	300	300
Electron exposure (e− per Å2)	50	50	50	50	50	50	50
Defocus range (μm)	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5	−1.2 to −2.5
Pixel size (Å)	1.068	1.068	1.068	1.068	1.068	1.068	1.068
Final Micrographs (no.)	44,044	44,044	44,044	44,044	44,044	44,044	44,044
Initial particle images (no.)	491,871	491,871	491,871	491,871	491,871	491,871	491,871
Final particle images (no.)	249,541	249,541	249,541	144,983	249,541	197,886	154,218
Symmetry imposed	C1	C1	C1	C1	C1	C1	C1
Map resolution (Å)	3.08	3.08	3.56	3.29	3.49	3.51	4.16
FSC threshold	0.143	0.143	0.143	0.143	0.143	0.143	0.143

The high-quality, native CPL cryo-EM maps enabled unambiguous assignment of the complete CPL protein inventory and accurate atomic model building (Fig. 1e). By integrating AlphaFold3 predictions¹⁹ and structure-similarity searches²⁰ with the experimental density, we defined the complete molecular composition of the CPL (Fig. 2a and Table 3). The resulting model revealed a large assembly of 16 distinct protein components in specific copy numbers, with bound ligands and coordinated metals of functional implications (Fig. 2b). The structure establishes that the intricate macromolecular assembly of the MII CPL is organized into three functional modules: a scaffold of maternal-effect proteins (PADI6, NLRP5, TLE6, OOEP, KHDC3, ZBED3, NLRP4F and NLRP14), ubiquitination machinery (a complete E3 ligase UHRF1 and E2 conjugating enzyme UBE2D3 module and three FBXW-family E3 ligase components, including FBXW18, FBXW19 and FBXW21, each paired with SKP1) and a tubulin assembly unit (αβ-tubulin heterodimer). To support our structural model, we performed quantitative proteomics analysis (Supplementary Table 1) and confirmed that all modeled CPL components were present in the proteomics data, most ranking among the top 10% highly expressed proteins in the egg proteome (Extended Data Fig. 4). These fundamental components form a minimal ASU, which undergoes dimerization and linear polymerization to generate a CPL filament. Multiple filaments crosslink into a higher-order fiber network (Fig. 2a).

Fig. 2. Architecture and functional modules of the CPL.

a, Top left, schematic summary of the CPL ASU and its higher-order organization. Top right and bottom, atomic models of individual CPL components fitted into their corresponding cryo-EM map densities. The cryo-EM maps are displayed as transparent surfaces. The components are grouped by functional modules. Top right, ubiquitination machinery. Middle right, tubulin. Bottom, maternal-effect proteins. b, Representative high-resolution features of the native CPL cryo-EM map, including amino acid side chains, bound ligands and coordinated metal ions.

Table 3.

Complete molecular inventory of the native CPL ASU determined by cryo-EM

Protein name	UniProt ID	Mouse Genome Informatics ID	Number of residues	Number of copies per ASUa	Proteomics rankingb	Protein category
PADI6	Q8K3V4	2655198	682	10	1	Maternal-effect protein
NLRP5	Q9R1M5	1345193	1163	2	4	Maternal-effect protein
TLE6	Q9WVB3	2149593	581	2	27	Maternal-effect protein
OOEP	Q9CWE6	1915218	164	2	118	Maternal-effect protein
KHDC3	Q9CWU5	1914241	440	1	12	Maternal-effect protein
ZBED3	Q9D0L1	1919364	228	1	46	Maternal-effect protein
NLRP4F	L7N1W9	2145528	937	1	28	Maternal-effect protein
NLRP14	Q6B966	1924108	993	2	11	Maternal-effect protein
UHRF1	Q8VDF2	1338889	782	2	14	Ubiquitination machinery
UBE2D3c	P61079	1913355	147	2	434	Ubiquitination machinery
FBXW18	Q3TSA9	3505704	469	1	77	Ubiquitination machinery
FBXW19	Q8C2W8	3505706	466	1	51	Ubiquitination machinery
FBXW21	Q8BI38	2443323	468	0.6	84	Ubiquitination machinery
SKP1	Q9WTX5	103575	163	2.6	41	Ubiquitination machinery
TBA1Cd	P68373	1095409	449	1	22	Tubulin
TBB4Bd	Q9D6F9	107848	444	1	21	Tubulin

^aValues represent the number of copies per ASU.

^bRanking is based on the protein identification scores from our proteomics dataset.

^cModeled here as UBE2D3, although other highly similar UBE2D family members cannot be definitively excluded.

^dSpecific tubulin isotypes could not be distinguished from the cryo-EM data. TBA1C and TBB4B are used as representatives for model building.

Extended Data Fig. 4. Quantitative proteomic profiling of CPL components in the mouse egg.

Each dot represents an identified protein, plotted according to its Log10 abundance area (x-axis) and Log10 identification score (y-axis). A vertical dashed line demarcates the threshold for the top 10% most highly abundant proteins within the dataset. All structurally modeled components of the CPL are explicitly labeled and color-coded based on their functional categories: maternaleffect proteins (green), ubiquitination machinery (red), and tubulin (yellow).

Maternal-effect proteins form the scaffold of CPL

Maternal-effect proteins constitute the main structural framework of the CPL. This framework is organized into two linked modules: a central core composed of two copies of NLRP5–TLE6–OOEP heterotrimer with additional associated components and a peripheral assembly of five PADI6 dimers. Together, these modules define the overall architecture of the CPL ASU and provide the scaffold for recruitment of the remaining functional CPL components.

At the center of the CPL, two sets of NLRP5, TLE6 and OOEP assemble into a heterodimeric module. This module is built around two NLRP5 molecules, which dimerize back to back through their leucine-rich repeat (LRR) domains (Extended Data Fig. 5a), consistent with a previously determined structure²¹. From a common NLRP5–TLE6–OOEP scaffold, the two halves diverge in their associated factors. In one half, the scaffold-associated NLRP4F engages KHDC3 through its N-terminal domain, whereas in the other, the corresponding NLRP4F binds ZBED3 through its C-terminal domain (Extended Data Fig. 5a).

Extended Data Fig. 5. Structural organization of the CPL maternal-effect protein scaffold.

a, The maternal-effect protein core is composed of two copies of NLRP5–TLE6–OOEP heterotrimer with additional associated components. Insets show side views of the two halves and the interface between the two NLRP5 LRR domains. The NLRP4F N-terminal domain (NTD) and C-terminal domain (CTD) are engaged in distinct binding modes in the two halves. b, Five PADI6 homodimers form a pentameric scaffold within each ASU. Insets highlight the inter-dimer contacts mediated by the N-terminal Ig-like subdomains and the C-terminal α/β-propeller domains.

The peripheral scaffold of the CPL is formed by a pentameric array of PADI6 homodimers (PADI6 dimers 1–5; Fig. 2a and Extended Data Fig. 5b), each adopting a head-to-tail configuration. This assembly is stabilized by two classes of interdimer interactions: contacts between the N-terminal Ig-like domains and packing between the C-terminal α/β-propeller domains (Extended Data Fig. 5b). Through its α/β-propeller domain, PADI6 dimer 1 contacts NLRP5 in the core scaffold, thereby linking the peripheral PADI6 assembly to the central NLRP5–TLE6–OOEP scaffold.

The CPL sequesters specialized ubiquitination modules

One of the most prominent features of the CPL structure is its abundance of ubiquitin enzymes. Notably, the structure reveals two distinct classes of sequestered ubiquitination machinery. The first is a UHRF1–UBE2D3 module, comprising the E3 ligase UHRF1 and its cognate E2 enzyme UBE2D3 (Fig. 3a). The E3 and E2 enzymes form a discrete module anchored between the scaffold NLRP14 and PADI6 dimer 2 in a CPL ASU. The 3D classification further showed that, in ~61% of particles, another copy of UHRF1–UBE2D3–NLRP14 assembly is positioned beneath the CPL core through interaction with PADI6 dimer 4, forming an underfoot module (Extended Data Fig. 3c).

Fig. 3. The CPL sequesters specialized ubiquitination machinery.

a, Overall structure of the CPL-bound UHRF1–UBE2D3 E3–E2 module. b, Canonical chromatin-recognition surfaces of UHRF1 are occluded in the CPL context. Left, superposition of previously reported H3 N-terminal tail-bound (PDB 3SOU) and H3K9me3-bound (PDB 4GY5) ligands onto the CPL-bound UHRF1_PHD domain shows a steric clash (highlighted by dashed contour) with surrounding CPL components. Right, superposition of hemimethylated DNA onto the CPL-bound UHRF1_SRA domain (PDB 3CLZ) shows that the canonical DNA-binding mode is incompatible with the assembled CPL environment. c, Canonical ubiquitin-binding modes of UBE2D3 modeled in the CPL context. Left, Ub^B (PDB 4V3L) and Ub^DO (PDB 5IFR and 5D0K) are sterically occluded by surrounding CPL components (highlighted by dashed contours). Right, Ub^DC can be accommodated (PDB 4V3L). d, Three distinct FBXW–SKP1 modules (marked by ovals) are positioned at defined sites within the CPL ASU. FBXW18, FBXW19 and FBXW21 are shown in different shades of purple and SKP1 is shown in gray. e, Structural mapping of sequence variability among CPL-associated FBXW members. Surface representation of the FBXW models colored by sequence conservation among FBXW18, FBXW19 and FBXW21. Strictly conserved residues are colored in deep purple, whereas highly variable residues are colored in white. Neighboring CPL components are shown in gray. f, Schematic model for sequestration and release of ubiquitination activities by the CPL.

UHRF1 is an established epigenetic regulator in early development for histone ubiquitination and DNA methylation maintenance^22,23. Structurally, UHRF1 is a multidomain protein composed of five major domains connected by flexible linkers. In the CPL-bound state, however, all five domains are well resolved and stably docked within the assembly (Fig. 3a). Its function normally depends on accessible chromatin-recognition and catalytic surfaces; the PHD domain binds the histone H3 tail and the SRA domain recognizes hemimethylated DNA. Inspection of the CPL-bound structure, however, shows that these surfaces are occluded by neighboring CPL components (Fig. 3b). This observation supports that the architecture of the CPL traps the UHRF1 module in a structurally constrained, inactive state.

UHRF1 makes extensive direct contact with UBE2D3, effectively cradling the E2 enzyme within the CPL assembly (Fig. 3a), which is consistent with the previously reported PADI6–UHRF1–UBE2D complex²⁴. To assess whether the canonical ubiquitin-binding modes of UBE2D3 can be accommodated in this context, we performed structural modeling. This analysis showed that both the backside-bound ubiquitin (Ub^B)²⁵ and the donor ubiquitin in the open conformation (Ub^DO)^26,27 would sterically clash with surrounding CPL components, whereas donor ubiquitin in the closed conformation (Ub^DC) can be accommodated (Fig. 3c). Thus, the architecture of the CPL appears to restrict UBE2D3 to a catalytically poised E2~Ub state, while disfavoring more accessible ubiquitin-binding configurations.

A second major class of ubiquitin ligases sequestered within the CPL comprises three distinct FBXW-family proteins, each paired with one SKP1 (Fig. 3d). The FBXW proteins act as substrate receptors and SKP1 links them to the core ligase scaffold in canonical SCF (SKP1–CUL1–F-box) E3 ligase complexes, which govern the turnover of essential proteins during embryonic development^28,29. Two stably incorporated FBXW–SKP1 assemblies contact NLRP5 molecules and β-tubulin. Focused classification further showed that a third FBXW is present in ~60% of CPL units (Extended Data Fig. 3d). All three contain the characteristic eight-bladed WD40 domain and are highly similar to one another (Extended Data Fig. 6a). Although homologous FBXW-family members could be fitted into these densities with comparable overall agreement and multiple leading candidates are all highly expressed in oocytes and early embryos (Extended Data Fig. 6b,c), the well-resolved side-chain features in our cryo-EM map allowed confident assignment of three members, FBXW18, FBXW19 and FBXW21 (Extended Data Fig. 6d).

Extended Data Fig. 6. Assignment of CPL-associated FBXW proteins.

a, Overall architecture of a representative CPL-associated FBXW protein, showing the N-terminal F-box domain and the C-terminal WD40 domain. Bottom, top view of the eight-bladed WD40 β-propeller. b, Overlay of AlphaFold3-predicted models for closely related FBXW-family members, illustrating their overall structural similarity. c, Expression heat map of candidate FBXW-family genes across developmental stages and tissues. Expression scores were retrieved from the Bgee database. d, Representative side-chain densities used to distinguish candidate FBXW-family members. Residues shown correspond to positions that differ among highly similar family members and enabled confident assignment of the three CPL-associated FBXW proteins. e, Sequence alignment of FBXW18, FBXW19, and FBXW21. Strictly identical residues are highlighted with a red background. Partially conserved residues are shown in red text, while non-conserved residues are shown in black text.

To understand why these three FBXW-family members are incorporated into the CPL, we aligned their sequences to identify nonconserved positions (Extended Data Fig. 6e) and mapped this sequence variability directly onto our 3D structural model. The variable residues cluster at interfaces with neighboring CPL components (Fig. 3e), indicating that member-specific sequence features help determine CPL binding. This supports selective recruitment of a specialized FBXW subset rather than nonspecific capture of closely related homologs. Structural comparisons with well-characterized FBXW homologs indicate that the conserved substrate-binding site lies on the top surface of the WD40 β-propeller (Extended Data Fig. 7a). In the CPL-bound state, this canonical substrate-binding pocket is buried within the assembly and inaccessible (Extended Data Fig. 7b).

Extended Data Fig. 7. Canonical substrate-binding surfaces of FBXW proteins are buried in the CPL.

a, Representative structures of substrate-bound FBXW-family proteins. In these complexes, substrates bind to the top surface of the WD40 β-propeller, centered around the pore. b, Structures of the three CPL-associated FBXW proteins (ribbons) positioned within the CPL assembly (gray surfaces). In all three cases, the top surfaces of the WD40 domain are largely buried.

Source data

These observations indicate that CPL selectively sequesters a highly specialized set of ubiquitination factors, thereby preventing premature ubiquitination in the egg. The captured machinery points to two distinct functional outputs: UHRF1-mediated chromatin regulation and FBXW-containing, SCF-dependent substrate degradation. Together with the specificity of FBXW molecules bound to the CPL, this suggests that the CPL is not a generic storage depot but a dormant and highly specific ubiquitination platform poised to orchestrate protein homeostasis and chromatin remodeling during early embryonic development (Fig. 3f).

The CPL functions as a reservoir for assembly-competent tubulin

In addition to serving as a reservoir for ubiquitin enzymes, our CPL structure suggests another potentially important storage function: sequestration of tubulin, the fundamental building block of microtubules³⁰. We observed two globular densities that could be unambiguously assigned as an αβ-tubulin heterodimer (Fig. 4a), positioned between the maternal-effect proteins NLRP14 and NLRP5. Our proteomics analysis revealed three highly expressed α-tubulin isotypes and five β-tubulin isotypes in eggs (Supplementary Tables 1 and 2). These isotypes share a high overall sequence identity (>90%), with sequence variations mostly restricted to their intrinsically flexible C-terminal tails (Supplementary Fig. 1a,b), making unambiguous assignment of specific isotypes challenging. Therefore, we conservatively modeled TBA1C and TBB4B as representative placeholders in this study. Notably, the variable sites among isotypes are positioned away from interaction interfaces with other CPL components (Supplementary Fig. 1c), suggesting the CPL recruits tubulin heterodimers through a promiscuous binding mode independent of specific isotypes. Inspection of the nucleotide-binding pockets showed that both the α-subunit and β-subunit are bound to GTP together with a coordinating magnesium ion (Fig. 4a). Whereas α-tubulin constitutively binds GTP, the GTP bound to β-tubulin is typically hydrolyzed to GDP upon incorporation into the microtubule lattice^31,32. The CPL-bound heterodimer is, therefore, captured in a prehydrolysis state before microtubule assembly.

Fig. 4. CPL stores assembly-competent αβ-tubulin.

a, Structure of the CPL-associated αβ-tubulin heterodimer. Insets, map densities for GTP and Mg²⁺ bound to both the α-subunit and the β-subunit. b, Structural comparison of CPL-bound tubulin with representative tubulin conformations in distinct nucleotide and assembly states. GMPCPP is a nonhydrolyzable GTP analog. After superposition of α-tubulin, the Cα r.m.s.d. of β-tubulin was calculated. c, A coordinated metal ion, tentatively assigned as Ca²⁺, is bound by the α-tubulin K40 loop. Insets, coordination environment and coordination geometry. d, Schematic model for CPL-mediated sequestration of tubulin to support developmental competence.

To further define the structural state of CPL-bound tubulin, we compared the sequestered αβ-tubulin heterodimer to tubulin in soluble and microtubule-assembled states (Fig. 4b). Soluble αβ-tubulin is characteristically curved, with an approximately 12° bend, whereas incorporation into the microtubule lattice straightens the heterodimer^33,34. Using α-tubulin for alignment, we measured the Cα root-mean-square deviation (r.m.s.d.) of β-tubulin as a readout of heterodimer conformation. The CPL-bound heterodimer matched most closely to the curved soluble state, especially GTP analog-bound tubulin, and differed markedly from the straight conformation found in the microtubule lattice (Fig. 4b). Together with its nucleotide state, this shows that the CPL captures tubulin in a prehydrolysis conformation characteristic of soluble, assembly-ready αβ-tubulin. Quantitative proteomics further showed that a substantial fraction of total cellular tubulin, in the range of 12–14%, is docked on the CPL (Supplementary Note 1). These findings support a model in which the CPL functions as a tubulin reservoir, sequestering the heterodimers in their assembly-competent state. Such sequestration is expected to prevent spontaneous microtubule polymerization driven by high cytosolic tubulin concentrations, while preserving a reserve that can be rapidly deployed for cytoskeletal remodeling after fertilization (Fig. 4d).

Notably, our structure also uncovered a compact extra density, consistent with a coordinated metal ion within the well-resolved α-tubulin K40 loop (Fig. 4c). The density is positioned in an electronegative pocket formed by residues D39, T41, G44 and E55 of α-tubulin, with coordination geometry and distances most consistent with calcium³⁵. Notably, highly similar Ca²⁺ coordination by the αK40 loop has been consistently observed in numerous previously reported tubulin crystal structures^36–42, although it has been suggested to potentially arise from crystallization conditions⁴³. Our native structure provides evidence in agreement with such a metal-bound state (Extended Data Fig. 8), although we cannot exclude other possibilities given the potential for post-translational modifications within the αK40 loop. In microtubules, this αK40 loop is a conserved regulatory element containing the canonical K40 acetylation site⁴⁴ and it is generally highly flexible and poorly resolved⁴³. The putative metal-coordinating conformation observed here may be favored in the CPL-sequestered state. Although the physiological importance remains to be established, these observations raise the possibility that CPL-associated tubulin may contribute to a calcium-related role for egg activation at fertilization^45,46.

Extended Data Fig. 8. Comparison of the αK40-loop Ca²⁺ site with previously reported crystal structures of α-tubulin.

Structural superpositions of the current cryo-EM model (yellow) with previously reported tubulin crystal structures containing Ca²⁺ at the αK40 loop. Comparisons with PDB IDs 5fnv (cyan), 7dad (olive), and 6n47 (red) reveal close agreement in both the αK40 loop backbone conformation (top panels) and the local metal ion coordination geometry (bottom panels).

Higher-order assembly of CPL fibers

Given the CPL’s likely role as a major reservoir for ubiquitin enzymes and tubulin, it is important to understand how CPL filaments assemble into higher-order fibers capable of accommodating the many copies of these components. To define the assembly mechanism, we docked the repeating C₂ unit into a cryo-EM map of the extended filament using rigid-body fitting. Adjacent repeating units stack in a strictly coaxial manner (Fig. 5a,b and Extended Data Fig. 9a), generating a linear filament without intrinsic curvature or helical twist. This longitudinal linkage is mediated by relatively rigid contacts between PADI6 dimer 5 of one repeating unit and an NLRP5 module of the next (Fig. 5a). In contrast to the rigidity of the interunit interface, flexibility resides within each repeating unit. The two ASUs within the repeating unit can move relative to one another. Alignment of one ASU showed that the other undergoes continuous swinging motions in three dimensions, with the two NLRP4F molecules acting as hinges (Fig. 5b and Extended Data Fig. 9b). These local motions could propagate to generate curved CPL filaments observed in the TEM micrographs (Extended Data Fig. 1a).

Fig. 5. Structural basis of CPL filament assembly, termination and higher-order network formation.

a, Cryo-EM map showing the longitudinal polymerization of adjacent repeating units (i and i + 1). The zoomed-in atomic model reveals that this coaxial stacking is primarily driven by specific interactions between the PADI6 dimer 5 of one unit and the NLRP5 of the adjacent unit. b, Schematic representation of the organization of the CPL filament and its structural flexibility. c, Structural analysis reveals how filaments end and are capped by a terminal unit. Top left, representative micrograph showing the location of internal (blue) and terminal (red) units. Bottom left, particle ratio of internal and terminal units. Right, a 3D reconstruction of the terminal unit (red; right), revealing the absence of PADI6 dimer 5 (marked by ovals). d, Cross-sectional view of the higher-order CPL 3D lattice reconstruction, demonstrating that a central main filament can potentially coordinate with up to four adjacent filaments positioned at its bottom and lateral sides. e, Top and middle, cryo-EM maps and atomic models of a filament pair, revealing that the interfilament interaction is mediated by contacts between PADI6 molecules (dimers 1 and 3) of the main filament and the NLRP14–UHRF1 module of the adjacent filament. Bottom, a side view of the reconstructed filament pair, demonstrating a relative crossing angle of ~4.8° between the two filaments. f, A schematic model summarizing the hierarchical assembly of the 3D CPL network. Repeating units composed of two ASUs polymerize linearly to form a CPL filament. A putative terminal unit may cap the filament end, while defined interfilament interactions crosslink neighboring filaments into a higher-order three-dimensional CPL network.

Extended Data Fig. 9. Structural analysis of intra-filament organization and flexibility.

a, Model docking reveals that consecutive units align in a strictly coaxial manner with an average of ~0° longitudinal rotation. b, The swinging of one ASU relative to the other across three dimensions reveals intra-unit conformational flexibility.

We next investigated how filament growth terminates. A 3D classification analysis identified a distinct class corresponding to half of the repeating unit (a single ASU) and without the next repeat (Fig. 5c and Extended Data Fig. 3a). Mapping these half-units back to the micrographs showed that they localize predominantly to the filament ends, suggesting that they represent putative terminal units of the CPL filament. Structural modeling of these terminal units revealed the absence of PADI6 dimer 5, the critical bridging element that mediates longitudinal stacking between repeats (Fig. 5c and Supplementary Fig. 2). This observation was corroborated by an independent 3D classification focused on the PADI6 dimer 5 region within the ASU, which also identified the subpopulation lacking this dimer (Extended Data Fig. 3e and Supplementary Fig. 2). This class lacking PADI6 dimer 5 was nearly identical in structure to the putative terminal units and accounted for a comparable particle fraction (Supplementary Fig. 2), supporting that the identified terminal units represent a bona fide structural state. The ratio of the half-terminal units to the intact internal units is ~2:7, suggesting that an average filament contains about eight repeating units. Given the ~37-nm axial periodicity, this corresponds to an average filament length of ~300 nm, consistent with the measured median length of individual CPL filaments from our TEM analysis (Extended Data Fig. 1b) and previous observations¹². Together, these observations raise the intriguing possibility that the putative terminal units provide a structural capping mechanism to terminate polymerization.

Extended filament reconstructions additionally revealed connecting densities from neighboring filaments around the central filament, suggesting that one filament can engage up to four adjacent filaments at lateral and basal positions (Fig. 5d). Reconstruction of an interacting filament pair showed that this higher-order connection is mediated mainly by the basal NLRP14–UHRF1 module of one filament and lateral PADI6 dimers 1 and 3 of the neighboring filament (Fig. 5e, top and middle). The two filaments intersect at a shallow angle of ~4.8° (Fig. 5e, bottom), an arrangement that may favor 3D network formation.

Together, these data support a hierarchical model for CPL assembly (Fig. 5f); coaxial stacking of repeating and terminal units generates an apolar, end-capped filament and directional lateral contacts crosslink these filaments into higher-order fibers, ultimately building the large, 3D CPL network essential for early embryonic development.

Discussion

The high-resolution, native structures of the mouse CPL we determined in unfertilized eggs reveal the complete molecular composition, assembly and architectural principles of the lattice. Structural comparison of CPL between the GV^15–17 and the MII stages described here reveals a highly conserved overall architecture, molecular composition and assembly principle (Extended Data Fig. 10), suggesting that the CPL is largely maintained through meiotic maturation. Our structure defines the final prefertilization state of this conserved lattice and shows that its sequestered ubiquitination machinery and tubulin reservoir remain intact in fertilizable eggs.

Extended Data Fig. 10. Structural comparison of the CPLs between GV and MII stages.

a, Schematic of mouse oocyte meiotic progression and developmental stages. b, Structural comparison of the CPLs between GV and MII stages. The MII-stage CPL ASU determined in this study is shown in red. The corresponding GV-stage structures (PDB ID: 9W2M, 9XRL and 9SFP) are shown in beige, blue, and gray.

The CPL is an apparently dormant ubiquitination hub in eggs, incorporating multiple ubiquitination factors in structurally restrained states. UHRF1 is responsible for the ubiquitination of histone H3 to maintain DNA methylation during early embryogenesis. In the UHRF1–UBE2D3 E3–E2 module, the substrate-binding and ubiquitin-binding pockets are mostly sterically occluded by the surrounding subunits of the CPL. Anchoring this module to the CPL can restrict UHRF1 in the cytoplasm, providing spatial and temporal regulation of UHRF1 for its nuclear translocation and epigenetic reprogramming⁴⁷. The oocyte-specific FBXW-family E3 ubiquitin ligase substrate receptors, FBXW18, FBXW19 and FBXW21, are bound to the CPL at three distinct positions, each paired with one adaptor protein SKP1. Their putative substrate-binding pockets are buried within the CPL, rendering their activities inaccessible in eggs. Although the substrates for these E3 ligase receptors remain elusive, we speculate that structural remodeling of the CPL or the release of these E3 ligase components after fertilization may facilitate targeted, ubiquitin-mediated degradation of paternal and/or maternal proteins during egg-to-embryo transition. Thus, our structure supports a model in which the CPL may serve as a storage platform for ubiquitination machinery and future work will be required to test the functional deployment of these factors during egg-to-embryo transition.

The CPL may serve as a reservoir of tubulin for cytoskeletal remodeling. The αβ-tubulin heterodimers anchored in the CPL are bound to GTP-Mg²⁺, adopting the curved, prehydrolysis conformation characteristic of soluble, assembly-ready tubulin and distinct from that in polymerized microtubules. Remarkably, the CPL sequesters ~12–14% of the total tubulin in eggs in a nonpolymeric but readily deployable state to prevent spontaneous microtubule assembly. We propose that the CPL serves as a reserve for microtubule polymerization, supporting the rapid, de novo assembly of an extensive cytoplasmic microtubule network that drives early pronuclear migration in mammalian zygotes^48,49.

Across the many cryo-EM structures of tubulin and microtubule polymers reported to date, the α-tubulin K40 loop has generally remained disordered and becomes structured only under specific conditions, such as enzyme binding^43,44,50. Therefore, it is notable that, in the CPL, we observe a stable αK40 loop conformation coordinated with a putative metal ion, tentatively assigned as calcium on the basis of its coordination geometry and near-identical conformation to numerous Ca²⁺-bound crystal structures. If this site is indeed calcium bound and substantially occupied, the abundance of CPL-bound α-tubulin suggests that the associated calcium pool could contribute appreciably to fertilization-induced calcium spikes⁴⁶. However, the identity, occupancy and physiological relevance of this putative metal-binding state remain to be established. Whether this feature contributes to calcium-related regulation during fertilization or early embryonic development will require future investigation.

The linearly stacked architecture of the CPL appears suitable for cargo sequestration and autoinhibition, resembling the assembly principles of several metabolic enzyme filaments^51,52. This organization may be particularly suited to the extraordinary lifespan of mammalian oocytes. In many species, oocytes are formed before birth and maintained in prolonged arrest⁵³, during which CPLs must preserve their cargo for eventual fertilization. Consistent with this view, CPLs are already assembled with a similar architecture in GV oocytes^15–18. Canonical cytoskeletal polymers such as actin filaments, microtubules and ESCRT-III are typically linked to regulated remodeling⁵⁴. In this context, it remains unclear how CPL components are activated and deployed during fertilization and early embryogenesis, and when and how CPLs disassemble. Although localized CPL remodeling was recently observed at the six-cell or eight-cell stage⁵⁵, whether disassembly occurs rapidly or progressively and how cargo deployment is coordinated with early embryonic stages remain to be determined. The new experimental methodology described here, together with established quantitative cell biology approaches for studying preimplantation embryonic development, will help to address these open questions.

Our approach of directly depositing individual eggs onto cryo-EM grids and rupturing them to expose cytoplasmic materials while preserving native organization opens new avenues to study the ultrastructural basis of cellular architecture in large, long-lived cells. The core CPL components identified here have clear human orthologs or close homologs (Supplementary Table 3) and are detected in human oocyte proteomes⁵⁶, suggesting that the major architectural principles of the CPL are likely conserved in humans. Given that nonsynonymous mutations in CPL genes are linked to female infertility^21,57, this strategy provides a foundation for single-cell structural analysis of human eggs and embryos at high resolution, with the potential to illuminate currently inaccessible aspects of human reproduction and inform assisted reproductive technologies.

Methods

Mouse husbandry, oocyte collection and in vitro maturation into MII eggs

Fully grown, prophase I-arrested oocytes were collected from the ovaries of 6–8-week-old CD1 female mice in M2 medium supplemented with dibutyryl cyclic AMP (dbcAMP) to maintain meiotic arrest. Oocytes were released from arrest by washing through multiple droplets of dbcAMP-free M2 medium and matured in vitro into MII eggs as described previously⁵⁸. Mature eggs were prepared for subsequent experiments by removing the ZP through sequential washing in droplets of Tyrode’s solution. Mice were maintained under specific-pathogen-free conditions on a 12-h light–dark cycle. All animal procedures were approved by the Yale University Institutional Animal Care and Use Committee (protocols 2024-20408 and 2024-20562).

Negative-stain EM imaging

A total of 10–12 ZP-free eggs were transferred onto poly(L-lysine)-coated culture dishes and allowed to settle and attach in PBS. Samples were then fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer for 1 h at room temperature. Samples were rinsed three times with 0.1 M sodium cacodylate and postfixed in 1% osmium tetroxide with 0.8% potassium ferrocyanide in 0.1 M sodium cacodylate for 1 h. Samples were then rinsed once in buffer and three times in water and then transferred to 2% uranyl acetate for 1 h, followed by three rinses in distilled water. Dehydration was performed through a graded ethanol series (50%, 70%, 90% and 100%) and samples were embedded in Embed‑812 epoxy resin (EM Sciences). Ultrathin sections (60 nm) were cut using a Leica UC7 ultramicrotome. Grids were poststained with 2% uranyl acetate and lead citrate before imaging on an 80-kV FEI Tecnai Biotwin TEM instrument equipped with an AMT NanoSprint 15 camera.

Tandem mass tag (TMT)-based quantitative proteomics of mouse eggs

For MII egg proteomics, 500 fully grown, prophase I-arrested oocytes were isolated from 8 C57BL/6 female mice aged 6–12 weeks. Oocytes were maintained in prophase arrest in M2 medium supplemented with dbcAMP, then released and matured in vitro to MII-arrested eggs as described previously⁵⁸. Eggs were then washed three times with several droplets of PBS to remove residual medium and extracellular contaminants and lysed in the RIPA buffer. Protein extracts were processed for TMT-based quantitative proteomics. Peptides were generated by tryptic digestion, labeled with isobaric TMT reagents and fractionated before liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis. MS data were acquired using high-resolution Orbitrap instrumentation and raw data were processed using established pipelines as described previously⁵⁸. Peptide and protein identifications were filtered to a false discovery rate of 1% using a target–decoy approach. Relative protein quantification was based on TMT reporter ion intensities, with normalization across channels to correct for labeling efficiency and sample loading. All proteomic sample preparation, TMT labeling, LC–MS/MS acquisition and initial data processing were performed by the Proteomics Core Facility at the University of Bristol.

Native cryo-EM sample preparation

R2/1 holey carbon gold grids (200-mesh, Quantifoil) coated with a 2-nm continuous carbon film were glow-discharged for 25 s at 20 mA. A total of 5–7 ZP-free MII eggs were manually placed onto each grid. The grids were then processed using a Vitrobot Mark IV (Thermo Fisher Scientific) at 10 °C and 100% humidity. To achieve optimal ice thickness and egg rupture on the grid, a consecutive triple-blotting was applied with a blot force of 20 and sequential blot times of 1 s, 1 s and 3 s. Following the final blot, the grid was immediately plunged and vitrified in liquid ethane.

Cryo-EM data acquisition

Cryo-EM data were acquired on a Titan Krios G2 TEM instrument (Thermo Fisher Scientific) operated at 300 kV. The microscope was equipped with a Gatan BioQuantum energy filter (20-eV slit width) and a Gatan K3 direct electron detector. Automated data collection was performed using SerialEM software using the single-particle cryo-EM approach. Micrographs were recorded in counting mode at a magnification of ×81,000, corresponding to a calibrated physical pixel size of 1.068 Å. Each video was collected with a total dose of 50 e⁻ per Ų. The defocus range was set between −1.2 and −2.5 μm.

Cryo-EM data processing

The acquired videos were processed for motion correction⁵⁹ and contrast transfer function (CTF) correction⁶⁰ using cryoSPARC⁶¹. A subset of 12,231 micrographs was first selected for particle picking. An initial particle set was generated by manual picking followed by two-dimensional (2D) classification and the resulting class averages were used for template-based automatic particle picking in cryoSPARC. After 2D classification, a clean dataset of 62,986 particles was retained. These particles were subjected to ab initio reconstruction and heterogeneous refinement. The most well-defined class was subjected to homogeneous refinement with C₂ symmetry applied, resulting in a 4.8-Å-resolution map that served as the reference for subsequent comprehensive particle picking.

High-resolution 2D template matching was performed on the entire dataset of 44,044 micrographs using GisSPA⁶², using reference templates generated with a 3° angular step, bin 2 scaling and C₂ symmetry. This process yielded an initial set of 491,871 particles. After the removal of duplicate particles, several rounds of local refinement and 3D classification and subset curation, a clean dataset of 290,887 particles was obtained. These particles were recentered to symcenter 2. Subsequent local refinement and CTF refinement yielded a repeating C₂ unit map at an overall resolution of 4.1 Å. To obtain the ASU map, particles were then symmetry-expanded and recentered to the ASU center; duplicate particles were removed. The resulting particles were refined in C₁ symmetry using an ASU mask together with CTF refinement, producing an ASU consensus map at 3.5-Å resolution.

To improve local map quality and resolve heterogeneous regions, focused local refinement and focused 3D classification were performed with region-specific masks. Local refinement yielded maps for the NLRP14 + UHRF1 (the UBL, TTD, PHD and SRA domains) region at 3.1-Å resolution, the tubulin dimer at 3.1-Å resolution, PADI6 dimers 1–4 at 3.3-Å resolution, the NLRP5’ + TLE6’ + OOEP’ + ZBED3 region at 3.0-Å resolution, the NLRP5 + TLE6 + OOEP + KHDC3 region at 3.1-Å resolution, the NLRP4F region at 3.6-Å resolution and the UBE2D3 + UHRF1_RING region at 3.6-Å resolution. Focused classification further resolved heterogeneity in several regions: PADI6 dimer 5 was present in 79% of particles and was refined to 3.5-Å resolution; the FBXW19–SKP1 + FBXW21–SKP1 region separated into classes containing FBXW19–SKP1 alone (42%) or both FBXW19–SKP1 and FBXW21–SKP1 (58%), with the latter refined to 3.3-Å resolution; the NLRP14′ + UHRF1′ + UBE2D3′ module was present in 61% of particles and refined to 4.2-Å resolution. The ASU composite map was generated by combining ten locally refined maps. Application of C₂ symmetry to the ASU composite map produced the final repeating unit composite map. The overall resolutions of the above reconstructed maps were assessed using the gold-standard criterion of Fourier shell correlation (FSC)⁶³ at 0.143 cutoff⁶⁴. For structural visualization and figure preparation, the final reconstructed maps were sharpened using DeepEMhancer⁶⁵.

To analyze the extended filament structure, particles were reextracted with a larger box size (2,048 pixels, bin 4). Local refinement using a main-filament mask yielded a filament map with an 8.6-Å Nyquist-limited resolution. Subsequent 3D classification with an adjacent filament mask was performed to resolve the map of an interacting filament pair.

Model building and refinement

Protein identification and model building were performed using a combination of template fitting, structure prediction and structure-similarity searches. For the NLRP5–TLE6–OOEP core, the published structure (PDB 8H93) was used as the initial model and docked into the map in ChimeraX⁶⁶. For unassigned densities, candidate identities were evaluated in two ways. First, AlphaFold3-predicted models¹⁹ of candidate proteins were docked into the cryo-EM density and assessed for overall agreement. Second, when no clear candidate could be assigned a priori, the density was initially traced as a polyalanine model in Coot⁶⁷. This polyalanine model was then used for a structure-similarity search against the AlphaFold database¹⁹ and the DALI server²⁰ to identify structurally related candidates. Candidate subunits were subsequently fitted into the cryo-EM map, assembled into a composite model and manually adjusted. Final model building was carried out iteratively in Coot, followed by real-space refinement in PHENIX⁶⁸ with secondary-structure, Ramachandran and rotamer restraints to maintain model geometry and improve agreement with the map. The refinement statistics are presented in Tables 1 and 2.

Estimation of CPL-associated tubulin fraction

Protein abundances were obtained from TMT-based quantitative proteomics as summed reporter ion intensities (‘area’). CPL-associated tubulin was estimated on the basis of structural stoichiometry, assuming a 1:1 relationship between tubulin heterodimers and the CPL components NLRP5, TLE6 and OOEP. Total tubulin abundance was calculated by summing all detected α/β-tubulin isoforms. The fraction of CPL-associated tubulin was then calculated by normalizing CPL-derived tubulin estimates to total tubulin abundance. Detailed calculations are provided in Supplementary Note 1.

Reporting summary

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

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-026-01843-2.

Author contributions

Conceptualization, B.M. and Y.X. Methodology, Y.L., J.L., W.Z., C.W., S.T., B.M. and Y.X. Investigation, Y.L., J.L., W.Z. and C.W. Visualization, Y.L., B.M. and Y.X. Project administration, B.M. and Y.X. Supervision, B.M. and Y.X. Funding acquisition, S.T., B.M. and Y.X. Writing—original draft, Y.L., S.T., B.M. and Y.X. Writing—review and editing, Y.L., S.T., B.M. and Y.X.

Peer review

Peer review information

Nature Structural & Molecular Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available. Primary Handling Editor: Dimitris Typas, in collaboration with the Nature Structural & Molecular Biology team.

Funding

This work was supported in part by a Vallee Scholars Award (VS-2024-56), Pew Scholars Award (00037689) and a National Institutes of Health (NIH) grant (R35GM146725) to B.M., a Lalor Foundation Postdoctoral Fellowship to J.L., an NIH National Institute of Child Health and Human Development grant R00HD104924 and a David Sokal Innovation Award of Male Contraceptive Initiative 2024-303 to S.T. and a Yale discretionary fund to Y.X.

Data availability

The cryo-EM density maps generated in this study were deposited to the EM Data Bank under accession numbers EMD-76333 (CPL filament map), EMD-76335 (ASU consensus map), EMD-76334 (ASU composite map), EMD-76315 (focused on PADI6 dimers 1–4), EMD-76321 (focused on NLRP5 + TLE6 + OOEP + KHDC3), EMD-76322 (focused on NLRP5’ + TLE6’ + OOEP’ + ZBED3), EMD-76323 (focused on NLRP14 + UHRF1), EMD-76324 (focused on tubulin), EMD-76325 (focused on NLRP4F), EMD-76326 (focused on FBXW19–SKP1 + FBXW21–SKP1), EMD-76327 (focused on UBE2D3 + UHRF1_RING), EMD-76330 (focused on PADI6 dimer 5) and EMD-76331 (focused on NLRP14′ + UHRF1′ + UBE2D3′). The corresponding atomic model of CPL ASU was deposited to the Protein Data Bank under accession number PDB 12DL. The proteomics data generated in this study were deposited to the PRIDE repository under accession number PXD076290. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at https://doi.org/10.1038/s41594-026-01843-2.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41594-026-01843-2.