EasySCP unveils extensive liver zonation at single-cell proteomics resolution

Abstract

The broader application of single-cell proteomics (SCP) in biology has been limited by complex workflows and reliance on specialized instrumentation. Here we present EasySCP, a high-throughput method that integrates FACS-based single-cell sorting, an all-in-one, single-step digestion process in 384-well plates, and sensitive mass spectrometry. EasySCP identifies nearly 5000 proteins from individual HEK293 cell. Applied to female murine liver, EasySCP achieves spatially informed proteomics profiling of hepatocytes zonation, detecting an average of 3500 proteins per hepatocyte and uncovering zonation patterns for 3277 out of 5267 proteins. Building on 215 conserved zonation markers, we further develop hepatocyte spatial status score (HSS) that enables reconstruct liver zonation across single-cell and multi-omics datasets. Together, our study introduces EasySCP, a broadly accessible tool for dissecting cellular heterogeneity at single-cell proteomics resolution in both healthy and disease states, effectively bridging the gap between transcriptomics and functional proteomics.

Single-cell proteomics’ (SCP) application in biology is limited by complex workflows and reliance on specialized instrumentation. Here the authors present EasySCP, a high-throughput SCP method integrating FACS sorting, one-step digestion, and sensitive MS to achieve spatially informed proteomics.

Introduction

The advent of single-cell analysis technologies has transformed the landscape of biological research by uncovering cellular heterogeneity and dynamic biological processes. Among these, single-cell RNA sequencing (scRNA-seq) has provided profound insights into transcriptomics diversity, facilitating the identification of novel cell types and states, and elucidating the molecular mechanisms underlying development and disease progression^1–3. However, scRNA-seq is limited in its ability to directly quantify protein expression and functional states, which are often regulated by post-translational modifications (PTMs) and other regulatory processes⁴. Furthermore, the correlation between RNA and protein expression is often imperfect, highlighting the need for proteome-level analyses at single-cell resolution⁵.

Single-cell proteomics (SCP) has emerged to address this gap by enabling direct quantification of proteins and their PTMs in individual cells. Mass spectrometry (MS)-based SCP has gained momentum due to its capacity to identify and quantify thousands of proteins in an unbiased manner, without reliance on antibodies or targeted probes⁶. Recent advances in MS instrumentation, along with innovations in sample preparation workflows, have propelled SCP to a stage where it can rival the throughput and depth of scRNA-seq^7,8. Workflows such as SCoPE-MS and nanoPOTS have demonstrated proteome coverage exceeding 2000 proteins per cell under optimized conditions^5,9. Recently, identification of 5000 proteins per cell has been realized by applying state-of-the-art mass spectrometer Orbitrap Astral¹⁰.

Despite recent advances, single-cell proteomics remains limited by technical challenges, including sample preparation losses, reliance on specialized equipment, and constrained throughput. These obstacles impede accessibility and hinder the widespread adoption of SCP in biomedical research^6,11.

To address these challenges, we develop EasySCP, a streamlined workflow (Fig. 1) designed to empower single-cell proteomics by providing a simple, accessible, and scalable solution. EasySCP integrates a pre-configured 384-well plate system that enables rapid sample preparation within 2 h, without any pipetting and sample transferring, which is inspired by MARS-Seq and SPID-seq developed in our lab^12,13. EasySCP detects up to 5000 proteins in a single HEK293 cell, ~3500 proteins in primary CD45^- non-immune cells, and around 3200 proteins in CD45⁺ immune cells isolated from MC38 tumor models. Furthermore, by incorporating marker-guided spatial sorting, EasySCP allows for the reconstruction of hepatic zonation proteomes at single-cell resolution. Utilizing both single-cell transcriptomics and proteomics data, we develop HSS to infer the spatial architecture of hepatocytes in diverse conditions, providing insights into liver function and architecture. EasySCP is applicable to a broad range of biological and clinical applications.

Fig. 1. EasySCP, overview of ultra-easy single-cell proteomics workflow.

Step 1: Single cell suspension preparation from tissues or in vitro cultures. Step 2: SuperMix buffer is loaded into 384-well plates using an automatic liquid handling system. Step 3: Single cells are sorted into 384-well plates via FACS. Step 4: Protein digestion is performed in a PCR thermocycler with a heated lid for 2 h. Step 5: Samples are acidified and directly subjected to LC–MS/MS acquisition in the 384-well plate, followed by downstream data processing and bioinformatic analysis.

Results

EasySCP enables an ultra-easy workflow for single-cell proteomics

Recent advances in single-cell proteomics have improved both sample preparation and mass spectrometry (MS) acquisition strategies. However, most existing approaches rely on labor-intensive workflows and specialized instrumentation, limiting scalability and accessibility. To overcome these challenges, we developed EasySCP, a high-throughput, user-friendly method for single-cell proteomics (Fig. 1). EasySCP integrates fluorescence-activated cell sorting (FACS) with an all-in-one lysis and digestion buffer, enabling efficient, single-step protein extraction and digestion in ultra-low volumes. We employed an ultra-low reaction system by injecting a SuperMix (SM) lysis buffer containing mass spectrometry and protease-compatible detergents, endoproteinase LysC and trypsin, and a volatile salt buffer into 384-well plate prior to single cell deposition. This configuration streamlines the workflow and increases accessibility and throughput relative to microfluidic-based systems such as proteoCHIP EVO 96. For single-cell isolation, EasySCP leverages FACS^14–18, a widely adopted, high-throughput, multi-parameter sorting platform that is capable of selecting cells based on cell size, granularity, and fluorescence markers (Supplementary Fig. 1A). This provides a versatile and unbiased cell selection capability superior to dispensing/picking-based systems like cellenONE or TECAN Uno, particularly for targeting phenotypically distinct populations.

Protein digestion is carried out using a heated-lid PCR instrument to minimize evaporation, eliminating the need for external evaporation barriers such as oil overlayers, hexadecane, or controlled humidity chambers. Following digestion, samples are acidified and directly subjected to LC-MS analysis within the same 384-well plate. The entire process is conducted within a single well, thereby preventing sample loss due to pipetting or transfer steps and ensuring maximum yield and reproducibility.

EasySCP enables single-cell proteome analysis in vitro and in vivo

To evaluate the performance of the EasySCP workflow, we first tested its application in mini-bulk proteomics. Fifty HEK293 cells were sorted into a single well by FACS (Supplementary Fig. 1B). Over 7200 proteins and more than 50,000 peptides were identified in a 24-min LC-MS gradient (Fig. 2A). Technical replicates showed high reproducibility, with Pearson correlation coefficients (R) exceeding 0.85 and a low coefficient variation (CV) of 10.09% (Supplementary Fig. 1C, D), demonstrating the method’s robustness and quantitative accuracy.

Fig. 2. EasySCP shows robust reproducibility at single-cell proteomics level both in vitro and in vivo.

A Number of proteins and peptides identified in 50 HEK293 cells (n = 6). Number of proteins (B) and peptides (C) identified in single HEK293 cell in reaction systems of 200 nL (n = 6), 300 nL (n = 6) and 500 nL (n = 6) with MBR for each group. Number of proteins (D) and peptides (E) identified in single HEK293 cell under reaction conditions of 37 °C (n = 6) and 50 °C (n = 8) with MBR for each group. F Localization of proteins identified in single HEK293 cell under the reaction condition of 50 °C (n = 8). G Biological functions of proteins identified in single HEK293 cell under the reaction condition of 50 °C (n = 8). Number of proteins (H) and peptides (I) identified in immune cell (n = 5) and non-immune cell (n = 5). J Heatmap showing differentially expressed proteins between immune cells (n = 5) and non-immune cells (n = 5). Source data is provided as Source Data File 1.

We next evaluated EasySCP for true single-cell proteomics. Due to the ultra-low protein content of single cells (~50–300 pg), large reaction volumes may reduce enzymatic digestion efficiency. To optimize conditions, we tested the digestion performance using 200, 300 and 500 nL reaction volumes with a 2 h incubation at 37 °C. The median number of protein identifications with match between runs (MBR) was 4472 (22,784 peptides) for 200 nL, 4526 (23,385 peptides) for 300 nL, and 4379 (21,379 peptides) for 500 nL (Fig. 2B, C), with the highest number of proteins identified at 300 nL volume. Without MBR, the median number of protein identifications were 2926 (11,755 peptides), 3255 (13,242 peptides), and 2803 (10,828 peptides) for 200, 300, and 500 nL, respectively, showing a similar trend across reaction volumes (Supplementary Fig. 1E, F). Based on these results, 300 nL reaction volume was selected for subsequent experiments.

To further optimize enzymatic digestion, we compared two incubation temperatures 37 °C and 50 °C. With MBR enabled, incubation at 50 °C resulted in increased protein identifications, reaching a median of 4773 proteins with 27,292 peptides, compared with 4517 proteins with 22,757 peptides at 37 °C (Fig. 2D, E). A similar trend was deserved without MBR, where the median number of protein identifications was 3856 (16,651 peptides) at 50 °C and 3251 (13,993 peptides) at 37 °C (Supplementary Fig. 1G, H). All single-cell replicates demonstrated high consistency, with Pearson correlation coefficients (R) above 0.92 (Supplementary Fig. 1I), and a median CV of 22.68% (Supplementary Fig. 1J), confirming high reproducibility and quantitative precision of EasySCP workflow.

Subcellular localization analysis revealed that, under the 50 °C incubation condition, a median of 396 proteins were membrane-associated, 731 mitochondrial, 2196 nuclear, and 2326 cytoplasmic (Fig. 2F). We also identified 61 kinases, 158 transcription factors, and 553 transport proteins (Fig. 2G), underscoring the depth and functional coverage achieved by EasySCP.

Following established single-cell proteomics benchmarking guidelines¹⁹, we included blank controls, single-cell (n = 6), and 20-cell samples (n = 3) to assess background signals and analytical sensitivity. Blank controls yield 45–182 protein identifications (23–180 without MBR), compared with 4298–5144 proteins (3982–4410 without MBR) in single cells and 6508–6571 proteins (6337–6502 without MBR) in 20-cell samples, indicating minimal background interference (Supplementary Fig. 1K).

Additionally, we further compared different reaction systems using single HeLa cells (n = 6 per group). EasySCP achieved a median of 4588 protein identifications (3929 without MBR), compared with 2754 (1969 without MBR) for the method reported by Mann et al.¹⁴, 3884 (3472 without MBR) for Slavov et al.¹⁵, 4922 (4098 without MBR) for Olsen et al.¹⁰, and 3130 (2517 without MBR) for Yun Yang et al.²⁰ (Supplementary Fig. 1L). Together, these results indicated that EasySCP performs competitively relative to current state-of-the-art SCP methods.

While significant progress has been made in single-cell proteomics, most applications were still limited to cell cultures in vitro. To demonstrate EasySCP’s utility in complex tissues in vivo, we applied it to a syngeneic murine colorectal carcinoma model using MC38 cells. Following tumor dissociation, CD45 staining was used to distinguish immune from non-immune cells (Supplementary Fig. 1M). A median of 3205 proteins with 12,339 peptides were detected in immune cells and 3517 proteins with 14,738 peptides in non-immune cells with MBR (Fig. 2H, I). Clustering analysis clearly separated immune and non-immune populations (Supplementary Fig. 1N), highlighting the precision of EasySCP in resolving tumor microenvironment at single-cell resolution. Immune cell replicates showed high correlation (Pearson R > 0.85), except for one outlier likely due to cellular heterogeneity (Supplementary Fig. 1O), while non-immune cell replicates all exceeded R = 0.7 (Supplementary Fig. 1P). Immune cells yield a median of 358 membrane, 437 mitochondrial, 1212 nuclear, and 1727 cytoplasmic proteins (Supplementary Fig. 1Q), including 21 E3 ubiquitin ligases, 72 transcription factors, and 35 transport proteins (Supplementary Fig. 1R). Non-immune cells showed 335 membrane, 451 mitochondrial, 1429 nuclear, and 1890 cytoplasmic proteins (Supplementary Fig. 1S), along with 28 E3 ubiquitin ligases, 86 transcription factors, and 31 transport proteins (Supplementary Fig. 1T). The protein expression profiles clearly distinguished immune from non-immune cells. Immune cells displayed a pronounced signature of macrophage-related functions, characterized by canonical markers (Itgam, Cd68, Mrc1, Arg1), mediators of antigen processing and presentation (Ctss, Fcer1g), regulators of immune signaling (Sirpa, Ptpn6, Vav1, Hck, Lyn), and factors involved in cytoskeleton organization and functional maintenance (Coro1a, Lcp1, Lamp1, Hk3), while non-immune cells predominantly expressed proteins facilitating cell proliferation and DNA replication (Pcna, Top2a, Mcm5, Mcm7, Msh6), alongside proteins contributing to migration and invasion (Cttn, Fscn1, Mxra7), metabolic reprogramming (Phgdh, Fasn, Cad), and epigenetic and transcriptional regulation (Hat1, Brd3, Eif4a3, Abce1) (Fig. 2J).

EasySCP resolves hepatocyte heterogeneity at single-cell resolution

Extending proteomics profiling to complex tissues, where cellular diversity, spatial organization, and microenvironmental factors govern functional specialization, is critical for understanding context-dependent proteome dynamics in development, disease, and homeostasis²¹. The mammalian liver, composed of hepatocytes organized along the porto-cental lobular axis, exhibits pronounced spatial heterogeneity, known as liver zonation, driven by gradients in oxygen, nutrients, morphogens, and hormones delivered via blood flow^22–24. Previous studies have used single-cell RNA sequencing, spatial transcriptomics and bulk proteomics to delineate hepatic zonation^{16,23,25–28}. More recently, Rosenberger et al. and follow-up studies^29–31 demonstrated that deep visual proteomics could achieve deep proteome coverage in both murine and human liver under different disease stages. Here, we integrated EasySCP with zonated surface marker-guided FACS sorting to spatially isolate individual hepatocytes, enabling the comprehensive reconstruction of hepatic zonation at single-cell proteomics resolution (Fig. 3A).

Fig. 3. Single-cell proteomics unveils hepatocyte heterogeneity at the protein level.

A Workflow of liver cell single-cell proteomics analysis. B Eight gates with spatial distribution from CV to PV were defined by flow cytometry analysis based on the fluorescence of CD73 and E-cadherin. C Unsupervised clustering analysis was conducted on the cell distribution of the eight gates. D UMAP visualization of the SCP data. Colors denote relative cellular expression of CD73, Ces1, E-cadherin, and Ass1. E Immunofluorescence staining of portal vein (E-cadherin, Ass1) and central vein (CD73, Ces1) associated protein in mouse liver. Representative micrographs were obtained from at least three independent biological replicates. CV central vein, PV portal vein. Scale bars, 50 μm. Source data is provided as Source Data File 2.

Building on established spatial sorting strategies²³, we performed single hepatocyte isolation using antibodies against the pericentral marker CD73 and periportal marker E-cadherin (Supplementary Fig. 2A). Hepatocytes were split into eight fluorescence-defined populations (Gate 1 to Gate 8) reflecting zonal positions along the central-porto axis, and deposited into 384-well plates pre-loaded with SM buffer for downstream EasySCP processing (Fig. 3B, Supplementary Data 1). We identified a total of 5267 proteins across 608 single hepatocytes, with a median of 3295–3503 proteins per cell across all gates (n = 65–84 per gate) (Supplementary Fig. 2B, C). Comparative analysis with bulk liver proteomics data²³ revealed 2634 overlapping proteins, and 2597 uniquely detected proteins in our single-cell dataset (Supplementary Fig. 2D). Subcellular localization and functional categorization of identified proteins revealed consistent coverage across zones (Supplementary Fig. 2E, F) underscoring the enhanced sensitivity of the EasySCP workflow.

After data filtering and batch-effect normalization, hepatocytes displayed highly heterogeneous, and different sorting gates exhibited clear gradients alongside the porto-central axis (Fig. 3C, Supplementary Fig. 3A, B). Unsupervised clustering grouped hepatocytes into 10 distinct clusters. The number of clusters was guided by quantitative anatomical measurements of liver lobule architecture. We measured the physical distance between CV and PV in 200 independent liver lobules and obtained an average CV-PV distance of ~284.3 μm (Supplementary Data 2). Based on the reported average diameter of hepatocytes²⁹, this distance corresponds to approximately ten sequential hepatocyte layers along the CV-PV axis. Accordingly, partitioning hepatocytes into 10 clusters reflects the intrinsic spatial resolution of hepatocyte layering within the liver lobule. These clusters exhibited a gradient-like distribution across gates 1–8 (Supplementary Fig. 3C). Notably, canonical zonated proteins including CD73, E-cadherin, Cyp1a2 and Cyp2f2 exhibited expected gradient pattern (Fig. 3D, Supplementary Fig. 3D–G), consistent with previous bulk proteomics profiles²³. Beyond known markers, we identified additional proteins with strong zonated signatures across functional categories, including drug and xenobiotic metabolism (Cyp1a2, Cyp2a5, Cyp2f2, Fmo2, Gba2), amino acid metabolism (Gls2, Gldc, Aldh1b1), aldehyde detoxification (Hsd17b6, Aldh3a2) and transport-related protein (Slc1a2, Tomm40l) (Supplementary Fig. 3E–G). Whole-slide immunofluorescence staining confirmed the zonated expression of CD73, E-cadherin, argininosuccinate Synthase 1 (Ass1), and carboxylesterase 1 (Ces1) (Fig. 3E). Ass1 were enriched in periportal zones, whereas Ces1 localized to pericentral regions, validating the spatial proteomics gradients identified by EasySCP (Fig. 3D, Supplementary Fig. 3D). Together, these data demonstrate the capability of EasySCP to resolve liver zonation and uncover spatially organized protein biomarkers at single-cell resolution.

Spatial reconstruction of mouse liver zonation at single-cell proteome level

To characterize liver zonation at single-cell proteome resolution, we analyzed spatial trends in protein expression across hepatocytes sorted from the central vein (CV) to the portal vein (PV) axis. Among the 5267 quantified proteins, 3277 (62%) displayed significant spatial segregation (FDR < 0.05, Kruskal–Wallis test) (Fig. 4A, Supplementary Data 3). Of these zonated proteins, the proteins including key regulators of fatty acid metabolism (Cyp4a10, Scd1, Hacd3, Elovl5, Hsd3b5, Hsd17b13), oxidative stress (Gstp1, Gclc, Dhrs9, Romo1), drug metabolism (Slc22a1, Slc13a2, Gba2) and signaling pathways (Plpp3, Alpk1, Eef2k, Clic5, Mtres1) were identified (Fig. 4B). Cross-validation with spatial proteomics datasets from laser microdissection (Rosenberger et al.²⁹) and bulk proteomics (Ben-Moshe et al.²³) confirmed substantial overlap: 70% zonated proteins (758 out of 1074) from laser microdissection, and 69% (1158 out of 1672) from bulk proteomics (Fig. 4C), supporting the spatial fidelity and depth of our single cell proteomics.

Fig. 4. Spatial zonation of the hepatocyte proteome.

A Zonation of hepatocyte proteins. Proteins are sorted by the COM values in the SCP data. Protein levels were normalized to the maximal level across all FACS gates. B Expression trend of zonated proteins. Average values of the cells in each gate were plotted. The shaded regions are ± s.e.m. C The intersection of zonated protein division. Sort the proteins based on the COM values in the SCP data. Blue indicates proteins with zonated expression detected in both the SCP and Ben-Moshe et al.²³ data, green indicates proteins with zonated expression detected in both the SCP and Rosenberger et al.²⁹ data, while yellow indicates proteins with zonated expression only in the SCP. D Scatterplot showing the average expression levels (log2 scale) of each protein at PV (X-axis) and CV (Y-axis). Yellow denotes differentially expressed proteins enriched in CV, whereas purple highlights those enriched in PV (P.adjust <0.05, |ave_log2FC| > 0.5). Differential expression was assessed using a two-sided Wilcoxon rank-sum test, with multiple testing correction performed using the Benjamini–Hochberg method. E Differentially expressed proteins of KEGG pathway enriched. F, G Levels of peroxisome and gluconeogenesis pathway from CV (left) to PV (right) gates as normalized to the maximal level across all FACS gates. Source data is provided as Source Data File 3.

To further explore region-specific liver functions, we annotated Gate 1-2 cells as CV-associated and Gate 7-8 cells as PV-associated clusters, based on the similarity of the expression profiles (Supplementary Fig. 4A, B). Differential analysis identified 1126 region-enriched proteins (619 CV-enriched, 507 PV-enriched) (Fig. 4D, Supplementary Data 4). Functional enrichment revealed that CV-enriched proteins were associated with biosynthesis of unsaturated fatty acids, fatty acid elongation, peroxisome, and drug metabolism-cytochrome P450 (Fig. 4E, Supplementary Data 5). In contrast, PV-enriched proteins were associated with biosynthesis of amino acids, gluconeogenesis, aminoacyl-tRNA biosynthesis, and oxidative phosphorylation (Fig. 4E). Region-specific pathway scores confirmed that peroxisomal pathways localized to pericentral zones (Fig. 4F), whereas gluconeogenesis dominated portal zones (Fig. 4G), in agreement with established hepatic metabolic gradients.

We further compared our SCP data with two independent spatial transcriptomic datasets from 10x Visium (Guilliams et al.²⁷) and Stereo-seq (Xu et al.²⁵) (Supplementary Data 6). Notably, spatial mRNA-protein correlations were modest (52.01% and 54.76% of mRNA-protein pairs were consistent, and 47.99% and 45.2% were opposite in 10x Visium and 10x Stereo-seq data) but significant (Spearman’s ρ = 0.201–0.227 for all matched genes; ρ = 0.387–0.412 for zonated subsets; P < 10⁻⁴⁵ to 10⁻⁹⁵; Supplementary Fig. 4C, D).

From these compared results, several genes showed consistent zonation at both transcript and protein level, including Glul, Oat, Cyp2e1, Tkfc, and Slc13a3 (Supplementary Fig. 4E), which participate in nitrogen, carbohydrate, and drug metabolism. Conversely, genes involved in mitochondrial function (Slc25a19), signaling (Plcb1), and proteostasis (Psmf1) displayed discordant spatial patterns between transcript and protein level, highlighting post-transcriptional regulatory effects (Supplementary Fig. 4E).

To systematically characterize these discordant patterns along the CV-PV axis, we compared protein zonation profiles with spatial transcriptomic gradients from both datasets. In the Visium dataset (Guilliams et al.²⁷), 699 genes exhibited increasing protein abundance but decreasing RNA levels, whereas 72 genes showing the opposite pattern (Supplementary Fig. 5A). Similarly, the Stereo-seq dataset (Xu et al.²⁵) identified 662 and 194 genes in these respective categories (Supplementary Fig. 5B). Intersecting the two datasets yielded 463 robust genes with protein upregulation but RNA downregulation (Group A) (Supplementary Fig. 5C), and 42 genes with protein downregulation but RNA upregulation (Group B) (Supplementary Fig. 5D).

Functional enrichment analysis revealed that Group A intersection genes were significantly enriched in pathways related to protein translation, protein degradation, mitochondrial metabolism, redox homeostasis, and nucleotide metabolism, with strong enrichment in oxidative phosphorylation (Supplementary Fig. 5E). In contrast, Group B intersection genes were enriched in processes associated with liver development and structural maintenance, extracellular matrix organization, sterol biosynthesis, type I interferon signaling, endoplasmic reticulum membrane organization, lipid droplet regulation, and protein-lipid complex functions (Supplementary Fig. 5F).

To experimentally validate representative cases, we performed immunofluorescence (IF) staining for 20 zonated markers. Among them, Ndufs5 (Supplementary Fig. 5I) and Uqcrq (Supplementary Fig. 5J) are shown as representative cases exhibiting opposite spatial trends at the RNA and protein levels. Consistent with the SCP data, both proteins showed higher expression in the PV region and co-localized with E-cadherin at the protein level (Supplementary Fig. 5G, H). In contrast, spatial transcriptomic maps derived from the Stereo-seq dataset showed that their transcripts co-localize with Nt5e (the gene encoding CD73), rather than Cdh1 (the gene encoding E-cadherin), confirming spatial discordance between transcript and protein levels (Supplementary Fig. 5K, L).

Of these 20 zonated markers, two zonation-specific proteins (Dhrs11 and Fmo3) were validated by IF (Supplementary Fig. 6A). Previously reported zonated proteins were also included as positive controls, confirming the reliability of our staining and spatial interpretation (Supplementary Fig. 6B, C). Specifically, a total of 8 well-established CV-enriched proteins, including Acox1, Aldh2, Cs, Cyp2e1, Ehhadh, Ephx2, Pdha1, and Psmd14, exhibited higher expression in the CV region, whereas 8 PV-enriched proteins, such as Acaca, Etcd, Gldc, Hal, Hsd11b1, Sdhb, Suclg2, and Vdac1, showed preferential localization in the PV region. The spatial distribution patterns of these canonical zonation markers were consistent with previous reports, further supporting the reliability of our immunohistochemical validation and spatial assignment.

Hepatocytes spatial status score accurately predict hepatocyte functional zonation

To further benchmark spatial features of hepatocyte zonation, we performed integrative analysis across multi-omics datasets including our SCP data, proteomics datasets from laser microdissection (Rosenberger et al.²⁹) and spatial bulk proteomics (Ben-Moshe et al.²³), spatial transcriptomic datasets from 10x Visium (Guilliams et al.²⁷), and Stereo-seq (Xu et al.²⁵). A total of 458 proteins were identified to exhibit zonation patterns consistently across these five datasets (total of 1,009,736 cells) (Supplementary Fig. 7A, Supplementary Data 7), which were defined as a set of conserved zonation proteome markers. Utilizing these 458 proteins, we further applied unsupervised clustering to regroup the 608 single hepatocytes into 10 clusters, revealing the most representative hepatocytes enriched in the CV (46 cells) and PV (26 cells) regions (Fig. 5A), and identified 215 significant differential protein expression (147 CV-enriched and 68 PV-enriched) (Supplementary Fig. 7B). Notably, the top five proteins showed gradient changes along the CV and PV axis (Fig. 5B). Leveraging these features, we developed a Hepatocyte Spatial Score (HSS) to quantify spatial distribution patterns of cells based on gene or protein expression. Predicted results by our HSS approach aligned with the experimental sorting thresholds in our SCP data and showed strong correlation with physical spatial distance (Spearman’s ρ = 0.824, P = 1.45 × 10⁻⁸⁵) in laser microdissection-based spatial proteomics (Rosenberger et al.²⁹), capturing the spatial trend along the CV-PV axis (Fig. 5C, D). For external snRNA-seq data (Bravo et al.¹⁶), HSS accurately reconstructs liver zonation by recapitulating the marker expression patterns (Fig. 5E, Supplementary Fig. 7C–E). Across spatial transcriptomics datasets, our HSS revealed gradient transitions along the CV-PV in space and showed strong correlation with unsupervised spatial clustering using whole transcriptomics, consistently in both 10× Visium (Guilliams et al.²⁷) (Supplementary Fig. 7F–I) and Stereo-seq (Xu et al.²⁵) datasets (Fig. 5F, Supplementary Fig. 8A–D). In spatial transcriptomics samples which were collected 8 h after partial (2/3) hepatectomy (PHx), HSS detected preserved overall CV-PV axis structure but pronounced shifts in expression and spatial partitioning, including an expansion of transitional state cells (Supplementary Fig. 8E), which was confirmed by expression changes in the CV marker Cyp2e1 and the PV marker Cyp2f2 at the tissue level (Supplementary Fig. 8F–I). Collectively, these results demonstrated that HSS robustly captures hepatocyte spatial organization, and provides a quantitative framework directly linking hepatocyte molecular features with liver spatial architecture, thereby facilitating the investigation of hepatic physiology and pathology.

Fig. 5. HSS-based reconstruction of hepatocyte spatial zonation.

A Distribution of cell numbers in each gate corresponding to unsupervised clusters, based on core zonation proteins. B Expression of the top 5 significant proteins in differentially expressed proteins, relative to total expression from CV to PV. C Spatial virtual coordinate projection with color mapping of HSS. Distances between hepatocyte layers were estimated from CD73, E-cadherin immunofluorescence sections and CV-PV measurements. Scale bar, 30 µm. D Correlation between HSS and physical spatial distance (Rosenberger et al.²⁹). Relative distance of 0 is at the PV, and 1 is at the CV. The solid red line indicates the fitted linear regression trend; the pink shaded area represents the 95% confidence interval (CI); the yellow shaded area represents the 95% prediction interval (PI). E External snRNA-seq data with spatial distribution predicted by HSS and heatmap of liver partitioning marker expression. F Spatial zoning was performed based on the HSS gradient, and representative CV (Cyp2e1, Glul, Gluo) and PV (Cyp2f2, Ftcd, Rida) region markers in the illustrated areas were spatially visualized. Source data is provided as Source Data File 4.

Discussion

We present EasySCP, a single-cell proteomics method that enables high-throughput, and reproducible proteomics profiling across diverse biological systems. By integrating fluorescence-activated cell sorting (FACS) with a robust sample lysis and digestion system, EasySCP streamlines the proteomics workflow, reducing the time, expertise, and costs traditionally associated with single-cell proteomics.

Compared with previously reported SCP workflows, EasySCP adopts a simplified preparation strategy. In the workflow descripted by Brunner et al.¹⁴, 1 µL of buffer (ACN, Tris) was pre-loaded prior to FACS sorting, followed by the addition of 1 ng Trypsin/LysC after sorting. Similarly, workflows from the Mann group¹⁴ and the Slavov group¹⁵ require enzyme addition after single-cell sorting. In contrast, in EasySCP, proteases are pre-dispensed into the wells prior to sorting, enabling immediate digestion after cell deposition and eliminating the need for a post-sorting enzyme addition step. Methods reported by Olsen et al.¹⁰ and Yun Yang et al.²⁰ more closely resemble a one-pot strategy, where digestion proceeds directly after cell isolation. However, these approaches rely on high-end single-cell isolation systems such as the CellenONE platform.

Additionally, based on functional classification, these platforms can be grouped into microfluidic separation systems, liquid dispensing systems, and cell sorting systems. EvoChip serves as a microfluidic carrier platform typically integrated with the EvoSep LC system for high-throughput LC-MS analysis, but does not perform cell or liquid sorting by itself. In SCP workflows, it is often combined with CellenONE for precise single-cell or low-volume liquid dispensing. Tecan liquid handling systems and the Mantis system are automated liquid dispensing platforms. Tecan systems are designed for general µL-mL liquid handling, while Mantis provides higher precision for nL-µL scale dispensing, but neither offers phenotypic cell selection. The recently introduced Tecan Uno integrates both liquid and single-cell dispensing capabilities; however, it does not support fluorescence-based cell selection and is limited to cells ~9–25 µm in size. CellenONE BSC supports both precise liquid handling and single-cell isolation, equipped with four fluorescence channels that enable selection based on cell size and phenotype. The key advantages of FACS include true multiparametric analysis, allowing simultaneous detection of multiple fluorescence markers along with forward and side scatter signals, which enables refined phenotypic discrimination and precise selection of defined cell types. FACS offers high throughput (often thousands of cells per second), purity and recovery rates, and robust reproducibility.

An additional practical advantage of EasySCP is workflow flexibility. EasySCP reaction plates can be fully prepared in advance, sealed, and stored at −80 °C prior to acquisition. These plates can be transported on dry ice to facilities equipped with cell-sorting instruments for FACS-based isolation, after which the downstream proteomic workflow can proceed according to the standard EasySCP procedure. Importantly, apart from the mass spectrometry acquisition platform, EasySCP does not require additional dedicated or proprietary instrumentation beyond commonly accessible laboratory equipment. In particular, FACS is widely available in institutional core facilities and represents a standard instrument in biological research centers. Unlike specialized single-cell isolation systems that require separate acquisition and maintenance (such as Tecan Uno, CellenONE BSC et al.). This approach provides a platform that is not only scalable and reproducible but also accessible to researchers across a wide range of disciplines, including cancer research, developmental biology, and tissue homeostasis studies. We compared EasySCP with several single-cell proteomics methods in HeLa cells and found that it achieves comparable protein identification depth under strictly matched experimental conditions (Supplementary Fig. 1L, Supplementary Data 8 and Supplementary Data 9). Overall, EasySCP simplifies the preparation process, reduces hands-on time and total processing time, and lowers instrumentation requirements while maintaining competitive analytical performance.

The versatility of EasySCP opens numerous possibilities for future applications in both basic research and clinical settings. In particular, the integration of EasySCP, the pre-loaded 384 well plate acts as capture wells, with spatial proteomics offers an exciting opportunity to dissect in vivo differences in protein expression across various tissue types. We applied EasySCP to murine liver and generated a comprehensive single hepatocyte proteome dataset. After the integrative analysis, we further presented the HSS which could accurately predict the spatial status of the hepatocytes both from transcripts and proteins expression level. The spatial assignment of hepatocytes in this study followed the staining and sorting strategy described by Ben-Moshe et al.²³, using established zonation-associated markers to generate CV- and PV-biased cell populations, rather than being derived from direct in situ spatial proteomics measurements. Rosenberger et al.²⁹ and Xu et al.²⁵ datasets, previously reported zonated proteins were used to establish the HSS score, while the spatial coordinate information provided in the dataset was used solely to validate the spatial relevance of HSS. Spatially resolved proteomics is poised to advance our understanding of liver heterogeneity, liver-related diseases progression, and therapeutic responses. For example, in liver cancer research, EasySCP could be used to map tumor microenvironments and identify key proteins involved in immune evasion or drug resistance, which could ultimately guide the development of personalized therapies. Similarly, EasySCP could be employed in developmental biology to explore the dynamic changes in protein expression during organogenesis or regeneration. We note that recent studies, including those by Rosenberger et al.²⁹ and Weiss et al.³¹, have also achieved spatial single-cell proteomics in liver tissue, reporting 2500–4000 proteins per hepatocyte. Our results show comparable proteome depth and extend FACS-based zonation strategies to a streamlined single-cell workflow integrated with high-sensitivity mass spectrometry.

Additionally, the performance of EasySCP could be further improved by advancing the instrumentation used in the proteomics analysis. For example, integrating more sensitive mass spectrometers with higher resolution and faster acquisition speeds would further improve the quality and depth of data obtained from each sample. These improvements would reduce the technical limitations imposed by current instrumentation and allow for even more accurate profiling of cellular heterogeneity.

Moreover, EasySCP can be seamlessly integrated into emerging multi-omics workflows, enabling simultaneous analysis of proteomics, transcriptomics, metabolomics, or epigenomics. This integration facilitates a more comprehensive understanding of how cellular functions are regulated across multiple molecular layers, providing a systems-level perspective on biological processes³². By bridging the gap between different omics approaches, EasySCP has the potential to significantly advance our ability to map the molecular mechanisms underlying complex diseases, development, aging, and tissue regeneration.

Overall, EasySCP provides a high-throughput, reproducible, and user-friendly workflow of single-cell proteomics, offering a high-throughput, reproducible, and user-friendly method. While improvements in throughput, sensitivity, and instrumentation are still required, the method shows great promise in a wide range of applications, from basic research in tissue biology to clinical applications in personalized medicine. By integrating EasySCP with spatial proteomics technologies, we are poised to uncover insights into tissue microenvironments, disease mechanisms, and therapeutic strategies, ushering in an era of high-resolution molecular mapping.

Methods

Ethical statement

All animal care and experimental procedures are approved by the Ethics Committee of Zhengzhou University (ZZU-LAC20240625[02]) and all institutional and national guidelines for the care and use of laboratory animals are followed.

Cell lines

The human HEK293 cells, and mice MC38 cells were purchased from Shanghai Jinyuan Biotechnology Co., Ltd. Cells were cultured in Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum (VivaCell) and 1% penicillin-streptomycin, and incubated with 5% CO₂ at 37 °C.

Animals and experimental design

In all experiments, wild-type 6-week-old female C57BL/6JOlaHsd mice were purchased from GemPharmatech Co., Ltd (Beijing, China). Mice were provided with water and food ad libitum. Animals were maintained under standard laboratory conditions with 12 h light/dark cycle, 23 ± 2 °C, and 45–60% humidity.

Liver perfusions and hepatocytes staining

Mice were anesthetized with 1.25% tribromoethanol, prepared by dissolving 1.25 g tribromoethanol (T48402; Sigma) and 2.5 mL tert-amyl alcohol (152463; Sigma) of 97.5 mL double steaming water. After the mice were anesthetized, a syringe was inserted into the inferior vena cava, and the liver was perfused with a perfusion buffer consisting of 1.2 mM EDTA in 500 mL HBSS. Shortly after the infusion begins, the portal vein is cut to allow the blood to flow out³³. After ~30 mL of perfusion buffer was infused, 25 mL of digestion buffer was administered, which consisted of perfusion buffer supplemented with 0.5 mg/mL collagenase IV (C4-BIOC; Sigma) and 1.6 mM CaCl₂. After digestion is completed, the liver was transferred into a petri dish containing 25 mL of digestion buffer, lightly chopped and filtered through a 70 μm cell filter to collect cells. Cells were spun down at 50 g for 3 min at 4 °C to obtain the hepatocyte-enriched pellet. The cells were suspended in MACS buffer (2 mM EDTA, and 0.5% BSA in 1× PBS) at a concentration of 10⁶ cells in 100 μL for flow cytometry staining.

Cells were stained with the following BioLegend antibodies at a dilution of 1:200: PE-anti-mouse/human CD324 E-cadherin (147303; BioLegend); APC anti-mouse CD73 (127209; BioLegend); PE/Cy7 anti-mouse CD31 (102410; BioLegend); and APC/Cyanine7 anti-mouse CD45 (103116; BioLegend). To determine the ploidy of hepatocytes, DNA was stained with Hoechst 33342 (5 μg/mL). After incubation on ice for 1 h, add 1 mL MACS buffer solution. Cells were centrifuged for 5 min at 50 × g. at 4 °C and the supernatant was discarded. Next, cells were stained with 7-AAD (5 μg/mL) to later enable the detection of viable cells by FACS.

Immunofluorescence staining

The fresh liver tissues of mice were directly embedded with OCT embedding agent and frozen on dry ice. The frozen blocks were made10 μm thick sections (Leica CM1950). The slices were fixed with 4% PFA (Thermo Fisher Scientific) for 15 min. Wash three times with PBST (1× PBS containing 0.05% Tween20), each time for 5 min. Permeate with 0.3% Triton X-100 in PBS for 5 min. Incubate at room temperature with the blocking solution (PBS containing 5% FBS, 2% BSA, and 0.3% Triton X-100) for 1 h. The sections were incubated overnight with the antibody solution (a blocking solution with a total reaction volume of 200 μL) at 4 °C. The antibody used was PE-anti-mouse/human CD324 E-cadherin (1:400; 147303; BioLegend), APC anti-mouse CD73 (1:400; 127209; BioLegend), Ass1 (1:100; 16210-1-AP; Proteintech), Ces1 (1:100; 16912-1-AP; Proteintech). The slices were washed three times with PBST on the second day. Non-directly labeled type antibodies were incubated at room temperature in the dark for 2 h using secondary antibodies, including CoraLite594-conjugated goat anti-rabbit IgG (1:400; SA00013-4; Proteintech), CoraLite647-conjugated F(ab’)2 Fragment Goat Anti-Mouse IgG (1:400; SA00014-10; Proteintech) and Cy5 conjugated Goat Anti-rabbit IgG (1:200; GB27303; Servicebio). Nucleus was stained with DAPI (1 μg/mL in PBS; HY-D0814; MedChemExpress) for 10 min. Subsequently, the slices were mounted on glass with antifade mountant (G1401, Servicebio). The whole tissue section was scanned using 20× objective lens and stitched by PannoramicMIDI II (3DHISTECH). Multichannel fluorescence scanning was performed to sequentially capture images. Imaging of the liver porto-central axis was viewed and processed using CaseViewer software (3DHISTECH) on digital slides.

Tumor transplantation and cell staining

The MC38 subcutaneous tumor model was established by inoculating tumor cells (1 × 10⁶/100 μL) on both flanks of C57BL/6 mice. Tumors were harvested on 21 days post inoculation. The tissue was chopped up in a petri dish and transferred to a 15 mL centrifuge tube using a wide muzzle. Tumor tissues were digested at 37 °C for 30 min with digestion buffer (1 mg/mL collagenase IV, and 0.1 mg/mL DNase in RPMI 1640 medium containing 20% FBS) to isolate the cells. Digestion was halted using a complete medium and cells were collected using a 40 μm cell filter. Cells were collected at 350 × g for 10 min at 4 °C and red blood cells were lysed. The cells were suspended in MACS buffer at a concentration of 10⁶ cells in 100 μL. Cells were stained with APC/Cyanine7 anti-mouse CD45 (103116; BioLegend) at a dilution of 1:200. Next, cells were stained with 7-AAD (5 μg/mL) to later enable the detection of viable cells by FACS.

HEK293 cell collection

HEK293 cells with 70–80% confluent culture was collected. Count and resuspension cells in MACS buffer at a concentration of 10⁶ cells in 100 μL. Cells were stained with 7-AAD (5 μg/mL) to later enable the detection of viable cells by FACS.

Flow cytometry and cell sorting

Cells were sorted by FACSAria Fusion (BD Biosciences) using a 130 μm nozzle. To minimize potential protein contamination, BD FACSFlow Sheath Fluid (342003) was used as the sheath fluid, which does not contain FBS or other added carrier proteins and is an aqueous buffered solution including sodium fluoride (0.064%). For liver cells, the gating strategy was adapted from Ben-Moshe et al.²³ to enrich spatially stratified hepatocyte populations while maintaining comparable viability and ploidy levels. Briefly, events were sequentially gated as follows: (1) hepatocytes were first selected from total events based on forward scatter area (FSC-A) versus side scatter area (SSC-A) to exclude debris and large aggregates; (2) singlets were selected by excluding margins in FSC-A versus FSC-width plots; additional singlet discrimination was performed using SSC-A versus SSC-width; (3) live cells were selected by excluding 7-AAD-positive dead cells; (4) non-parenchymal cells were excluded by gating out CD31- and CD45-positive populations; (5) tetraploid hepatocytes were identified based on Hoechst staining intensity to minimize ploidy-associated size and protein content differences; (6) To resolve zonated hepatocyte subpopulations along the porto-central axis, we used CD73 and E-cadherin as spatial markers. Positivity thresholds were defined using unstained controls, and the highest intensity observed in unstained cells was set as the cutoff. CD73-positive cells and E-cadherin-positive cells were each subdivided into four equal subpopulations according to fluorescence intensity, resulting in eight populations in total (Gate 1–8). Within the CD73-positive group, Gate 1–4 represented decreasing CD73 intensity. Within the E-cadherin-positive group, Gate 5–8 represented increasing E-cadherin intensity. Single cells from each population were sorted into a pre-treated 384-well plate and subsequent mass spectrometry was performed.

For tumor samples and HEK293 cells, 100 μm nozzle is used for cell sorting. After removing the adherent cells, HEK293 cells sorted individual living cells directly by 7-AAD. The immune and non-immune cells in the tumor samples were distinguished according to CD45 staining, and their individual cells were sorted into the pre-treated 384-well plate, and then analyzed by mass spectrometry.

Sample preparation for the 384-well plate workflow

SuperMix (SM) lysis buffer (0.2% DDM, 10 ng/µL LysC and 20 ng/µL trypsin and 100 mM Tris buffer, pH 8.0) was injected into 384-well plate by automatic instrument (UK Robotics D2). Single cell was directly sorted into prepared 384-well plate containing SM lysis buffer, and then lysed and digested into tryptic peptides at 37 °C or 50 °C for 2 h in a PCR instrument with heat lid. To avoid evaporation of samples, the temperature of the lid of the PCR instrument is at least 10 °C higher than the incubation temperature. Finally, each well was acidified by 4 µL 1% formic acid (FA) and then ready for LC-MS analysis.

LC-MS/MS analysis

Peptides in the 384-well plate were directly loaded onto Aurora elite C18 column (15 cm × 75 µm lD, 1.7 µm C18, IonOpticks, Melbourne, Australia) and eluted at 200 nL/min and a total gradient of 24 min by Vanquish Neo liquid chromatograph. Eluted peptides were analyzed by Orbitrap Astral mass spectrometry. FAIMS Pro Duo with compensation voltage −48 V was applied to reduce background interference. For Orbitrap full scan, precursor ions with the range of 400–800 m/z were analyzed at the resolution of 240,000, and normalized AGC target and maximum injection time were set to 500% and 100 ms, respectively. For Astral DIA MS2 scan, precursor mass range was from 400 to 800 m/z, and isolation window and window overlap were respectively 20 m/z and 0. The isolated ions was fragmented by using higher-energy collisional dissociation with normalized collision energy 25%. The fragment ions with the range of 150−2000 m/z were acquired. Normalized AGC target and maximum injection time were set to 800% and 40 ms, and the loop control time was 0.6 s. For protein identification in Supplementary Fig. 1K, Orbitrap Astral Zoom mass spectrometry was used for data acquisition in a 14 min gradient. FAIMS Pro Duo with compensation voltage −48 V was applied to reduce background interference. For Orbitrap full scan, precursor ions with the range of 400–800 m/z were analyzed at the resolution of 240,000, and normalized AGC target and maximum injection time were set to 500% and 100 ms, respectively. For Astral DIA MS2 scan, precursor mass range was from 400 to 800 m/z, and isolation window and window overlap were respectively 20 m/z and 0. The isolated ions was fragmented by using higher-energy collisional dissociation with normalized collision energy 25%. The fragment ions with the range of 150−2000 m/z were acquired. Normalized AGC target and maximum injection time were set to 800% and 40 ms, and the loop control time was 0.6 s.

Data analysis

Raw data were searched by Spectronant (version 19.4) with a library-free directDIA+ approach against uniport human database or mouse database. Maximum-intensity m/z extraction and precision iRT calibration with local (non-linear) regression were applied, with dynamic MS1 and MS2 tolerances. Identification thresholds were controlled at 1% FDR at the PSM, peptide, and protein group levels, with precursor and protein Q-value cutoffs of 0.01 (protein 0.05 at run level). Decoys were generated using a mutated approach with NN-predicted fragments, and protein inference was performed using IDPicker. Database searching was conducted with Trypsin/P allowing up to two missed cleavages and peptide lengths of 7–52 amino acids; variable modifications included protein N-terminal acetylation and methionine oxidation. Quantification was performed at the MS2 level using peak area with interference correction enabled, based on Q-value-filtered precursors without imputation, and cross-run normalization was applied automatically. Protein quantities were calculated using a Top N approach (1–3 peptides per protein group). In Spectronaut, peptide identification transfer across runs is achieved through its cross-run matching framework based on aligned retention time and mass accuracy, thereby improving identification consistency across samples. The analysis was performed with “Condition” set to Not Defined, meaning that all single cells were treated as part of the same processing batch and eligible for cross-run identification matching. This configuration effectively corresponds to enabling match-between-runs (With MBR). Conversely, when “Condition” is defined to separate runs into distinct groups, cross-run identification transfer is restricted, corresponding to without MBR (Without MBR). For data analysis in Supplementary Fig. 1K, data were searched by DIA-NN (2.2.0) software with a library-free directDIA+ approach against uniport human database. Database searching was conducted with Trypsin/P allowing up to two missed cleavages and peptide lengths of 7–52 amino acids, and variable modifications included protein N-terminal acetylation and methionine oxidation with or without MBR.

Proteomic depth was determined directly from the original Spectronaut output, before any normalization or imputation of missing values. Protein group data were analyzed using R (version 4.4.1) within RStudio (version 2024.09.16). For HEK293 cells, immune and non-immune cells, the Quantity data were preprocessed separately for each cell type by retaining proteins that were detected in at least one cell within the respective type. Pairwise Pearson correlation coefficients between single cells were calculated using the cor function with the method set to “Pearson” and the parameter use = “pairwise.complete.obs” to accommodate missing values. The resulting correlation matrix was visualized as a heatmap using the ComplexHeatmap package. To assess the stability of protein expression, we calculated the coefficient of variation (CV = SD/Mean × 100%) for each protein across cells. The resulting CV values were used for kernel density estimation, and the distribution was visualized using the ggplot2 package.

To explore subcellular localization, we accessed the UniProt database³⁴ to retrieve annotation data for each protein. Proteins with multiple annotated locations were parsed into individual localization entries, and the frequency of each localization category was summarized and visualized. In addition, several functional annotation databases were integrated for protein classification, including the transporter classification database³⁵ (TCDB), E3 ubiquitin ligase database³⁶ (E3 ATLAS), epigenetic regulator database³⁷ (EpiFactors), kinase database³⁸ (KinBase), and transcription factor database³⁹ (AnimalTFDB). For G-protein-coupled receptor (GPCR)-related proteins, filtering and classification were based on molecular function annotations provided in the UniProt database³⁴.

For liver SCP proteomics data, all 650 single hepatocyte samples were initially evaluated, and cells with fewer than 1000 quantified proteins (n = 42) were excluded from downstream analyses. After filtering, 608 cells remained for data-researching and subsequent analysis. Quantity data was preprocessed and filled the missing values with 0, divided each value by the sum of the columns, multiplied by the scale factor 10,000 and applied the logarithmic transformation (log1p) to perform the normalization process. Based on the grouping information of FACS, we retained proteins expressed in at least 6 cells in each group, and a total of 5231 proteins were identified. For zonated gene screening, Benjamini–Hochberg procedure was performed to calculate the FDR, and 3277 proteins with FDR < 0.05 were significantly zonated. Next, all protein profiles were normalized to their maximum across all FACS gates and sorted them by their center-of-mass calculation as described above²³. Heatmaps were generated using the ComplexHeatmap package (version 2.20.0). We integrated the previously published datasets, including those based on laser microdissection (Rosenberger et al.²⁹) and spatial bulk proteomics (Ben-Moshe et al.²³), compared the detection situations with protein partitions. To quantitatively summarize spatial expression gradients, we calculated a center-of-mass (COM) value for each protein, following the method described by Ben-Moshe et al.²³. Briefly, for an expression profile distributed across the eight fluorescence-defined FACS gates (z = 1–8), the COM was calculated as:

$$COM\left(x\right)=\frac{{\sum }_{z=1}^{8}z\cdot x_z}{{\sum }_{z=1}^8{x}_z}$$ 1

where $x_z$ represents the expression level of the protein in gate $z$. The COM provides a quantitative metric for ranking and comparing zonation patterns across datasets.

To visualize the spatial distribution of hepatocytes, the dispersion degree of proteins is calculated and retain those with high variability (variance-to-mean ratio >0.1) to eliminate those with low variability or background noise. Subsequently, dimension reduction analysis was conducted in the Seurat⁴⁰ (v5.3.0). To eliminate the batch effect formed by the sampling date, we applied the FastMNN algorithm for cross-sample integration based on the principal component analysis, and set the integration parameters (k = 85, ndist = 1.14) to improve the matching accuracy and optimize the integration effect. Nonlinear dimensionality reduction was performed using UMAP, and the results were visualized accordingly.

Gene set enrichment analysis for proteomics data is based on spatial sorting to generate hepatocyte subsets enriched in specific regions. Gate 1 and gate 2 population were classified as central veno-associated clusters (CV), and gate 7 and gate 8 population as portal veno-associated clusters (PV). Differentially expressed proteins between CV and PV clusters were identified using the Wilcoxon rank-sum test implemented in FindMarkers. P-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method, and proteins with FDR-adjusted p-values (p.adjust <0.05) and |avg_log2FC|> 0.5 were considered significant. Gene set enrichment analysis was performed using enrichKEGG with KEGG or WikiPathways as functional libraries, and pathways with FDR-adjusted p-values (p.adjust <0.05) were considered significantly enriched.

In two spatial transcriptome datasets, 10× Visium (Guilliams et al.²⁷) and Stereo-seq (Xu et al.²⁵) platforms, the spatial transcriptome was divided into eight equidistant spatial regions as described above. We identified proteins corresponding to 4830 genes from Guilliams et al.²⁷ (2022) and 4905 genes from Xu et al.²⁵ (2024) within the SCP dataset. According to the same analytical method, 1953 and 2299 were zonated genes, respectively, as defined in the respective studies. The average expression values of central vein-related clusters and portal vein-related clusters in the three datasets were statistically calculated respectively, and the biases in the expression of each gene or protein (log2 times change of CV relative to PV) were calculated. Then, calculate the Spearman correlation coefficient to evaluate the consistency between protein and RNA biases. Meanwhile, we also calculated the RNA expression abundance corresponding to the partitions. The expression level of each partition was standardized as the average of all gates, and the expression of genes or proteins was represented by a line graph.

Hepatocyte spatial status score

We defined the Hepatocyte Spatial Status Score (HSS) to quantify the spatial positioning of hepatocytes along CV-PV axis. Based on unsupervised clustering of 458 conserved zonation proteome markers in SCP and FACS sorting gate, we identified representative hepatocyte clusters enriched in CV and PV regions (46 cells in CV and 26 cells in PV). Differential expression analysis was performed to further define 147 CV and 68 PV marker gene sets (P < 0.05 and |avg_log2FC|> 0.5). For each cell, the HSS is defined as the standardized difference of the weighted difference between the GSVA⁴¹ scores of CV and PV gene sets defined as follows:

$$HS{S}_i=\frac{{\sum }_{gϵG_{CV}}{w}_{ig}.K\left(z_{ig}\right)-{\sum }_{gϵG_{PV}}{w}_{ig}.K\left(z_{ig}\right)}{\sqrt{{\sigma }_{CV}^2+{\sigma }_{PV}^2}+ϵ},ϵ=0.01$$ 2

where $z_{ig}$ represents the standardized expression value of gene g in cell i, obtained by centering and scaling expression values across cells. $K\left(z_{ig}\right)$ denotes the Gaussian kernel transformation, which smooths the standardized expression values and emphasizes relative enrichment patterns in a non-parametric manner, and $w_{ig}$ is the rank weight of each gene, reflecting the relative expression rank within each cell, with higher weights assigned to genes with higher expression ranks to reduce the influence of low-abundance or noisy features, as defined in GSVA⁴¹. HSS provides a standardized and continuous single-cell-resolution indicator of spatial positioning along the CV-PV axis. Normalization was performed after calculating the HSS scores within each sample set to ensure comparability of scoring results across different samples. For each cell, the raw HSS score was adjusted by adding the absolute value of the minimum score of the entire set, followed by division by the maximum score of the set. This normalization procedure allowed for consistent evaluation of spatial physiological state changes along the CV-PV axis across different samples. All computational analyses were performed in R (v4.4.1) using the GSVA package (v1.52.3), with additional data handling and visualization performed using Seurat (v5.3.0), tidyverse (v2.0.0) and ggplot2 (v3.5.1).

Statistics & reproducibility

For each experimental condition, at least three independent biological replicate experiments were conducted. Data analyses were primarily performed using R (version 4.4.1) within RStudio (version 2024.09.16), and detailed methods and parameters are described in the Data analysis section. Data from representative experiments, including microscopic images, were obtained from at least three independent biological replicates and showed consistent and reproducible results.

Reporting summary

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

Author contributions

B.G.L., I.A., and B.B.H. conceptualized the EasySCP workflow. G.C.Q. and B.B.H. acquired MS data. J.H.W. prepared all the samples. J.H.W., J.E.X, and K. X. analyzed the data. Y.F. contributed to the data analysis. Q.Y.Z., X.W., S.Y.W., Y.X.S. and Y.L.C. assisted in conducting animal experiments. B.B.H. and J.H.W. curated data. Z.G.D., X.H., T.J.Z. and P.L. supervised the project. B.B.H., J.H.W. and B.G.L. wrote the original manuscript draft. All authors reviewed the manuscript and provided input.

Peer review

Peer review information

Nature Communications thanks Florian Rosenberger who co-reviewed with Kathakali De, and Hai-long Piao for their contribution to the peer review of this work. A peer review file is available.

Funding

B.B.H. discloses support for the research of this work from the National Natural Science Foundation of China [No. 32401230]. Z.G.D. discloses support for the research of this work from Major Science and Technology Projects in Henan Province [No.221100310100]. G.C.Q. discloses support for the research of this work from Shandong Provincial Natural Science Foundation [No. ZR2022QC244]. J.H.W. discloses support for the research of this work from Zhengzhou University Young Student Basic Research Projects (PhD students) [No. ZDBJ202527]. B.G.L. is supported by the Startup Financing from the Tianjian Laboratory of Advanced Biomedical Sciences [No. 2024-KYQD-0117-001] and Zhengzhou University [No. 35220827-25] [No. 35220184-24]. G.C.Q. is supported by Young Taishan Scholars Program of Shandong Province.

Data availability

The proteomics raw data have been deposited in the ProteomeXchange Consortium. The dataset can be accessed under Project accession: PXD070185 and PXD074900. The raw data are also accessible through the iProX database under accession number IPX0013985000 [https://www.iprox.cn//page/subproject.html?id=IPX0013985001] and IPX0015940000. Source data are provided with this paper.

Code availability

The source code is publicly available at: https://github.com/WeiJinghui954/EasySCP and archived at 10.5281/zenodo.19627167⁴².

Competing interests

B.G.L. and B.B.H. are inventors on the patent relating to EasySCP, which has been applied. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74525-8.