Background & Summary

The liver is a critical multifunctional organ, serving as a central coordinator for metabolic homeostasis and contributing to the eradication of various xenobiotic compounds and toxins1. The ability to manipulate multi-tasks for the liver mainly depends on the spatial zonation, which is based on the highly structured repeating anatomical units termed liver lobules2. Owing to the blood flow and morphogens, hepatocytes along the lobular axis, which is from portal veins to central veins, are exposed to different physicochemical environments, resulting in differential expression profiles with a further tendency to distinct functions3,4,5. On the basis of hepatocytic differences, liver lobules are divided into three zones, zone 1–3 from portal veins to central veins, with various essential liver functions6.

In the 20th century, researchers used several technologies to explore liver zonation characteristics, including in-situ hybridization, immunohistochemistry, and microdissection combined with transcriptome sequencing7,8,9. However, the precision and depth of these studies are limited.

Nowadays, single-cell RNA sequencing makes it possible to measure the genome-scale information with a more increased resolution and has revealed that 50% of hepatocyte genes are expressed in a zonation manner10,11,12. Nevertheless, dissociating tissue into single-cell suspensions loses inherent spatial information, which can not be reconstructed in silico completely13,14. Moreover, on reviewing the literature, the proportion of hepatocytes was less than what should be expected15,16,17,18,19. Such bias may be due to the size of hepatocytes and their intolerance to tissue dissociation methods.

To address these issues, we conducted spatial transcriptomics on the normal human liver with 10X Genomics Visium technology and obtained 6581 high-quality human liver spots from two organ donors (liver 1 and 2). Consistent with the proportion in normal liver, cells in these data were largely hepatocytes20.They were divided into four subgroups along the lobular axis with the combination of unbiased classification and location of spots. The data contain detailed spatial information of normal liver, providing a reliable reference for liver disease research.

Methods

We introduce a summary of the liver spatial transcriptome method. The whole procedure included the acquisition of human liver tissue, preparing frozen sections, and Visium sample processing (Fig. 1A).

Fig. 1
figure 1

Spatial transcriptome reveals the cell populations of the normal human liver. (A) Overview of the Visium process using human liver tissue samples. (B) Uniform manifold approximation and projection (UMAP) plot showing the unbiased classification of liver cells. (C) Split UMAP plot showing the batch effect between the two different liver samples. (D) Heatmap showing the marker genes of each cluster, highlighting the top marker genes for each cluster. (E) Violin plot illustrating the selected marker genes of each cluster.

Full size image

Ethical approval

We received approval from the Ethics Committee of the Affiliated Hospital of Yunnan University, and signed informed consent was obtained from all patients.

Human liver tissue procurement

Fresh human liver samples (Supplementary Table S1) were collected at the Affiliated Hospital of Yunnan University. Samples were obtained from patients undergoing partial liver resection for hepatic hemangioma. Normal liver tissues were obtained at least 1 cm away from hemangioma, and the HE images support the normal liver diagnosis(Supplementary Figure 1A,B).

First, fresh samples were obtained from the operating theatre and immediately rinsed three times using pre-cooled sterile saline to wash away any residual blood and red blood cells on the surface. The surface was blotted dry using sterile gauze, the whole process taking about 1 min.

A small amount of pre-cooled optimal cutting temperature (OCT) compound (Sakura Finetek, Torrance, CA) was first added to cover the bottom in a 7 mm*7 mm embedding mold, followed by gentle clamping of the liver tissue block with forceps and placing it in the mold. Sufficient OCT was then added to completely cover the specimen, taking care to prevent the appearance of air bubbles. The whole process was completed on ice.

The tissue blocks were then immediately placed in dry ice boxes for quick-frozen storage and transferred to a deep cryogenic refrigerator for storage. Subsequent sample quality control and tissue optimization processes were carried out.

Frozen sections preparation and quality control

The liver tissue block was mounted onto the specimen disc of a cryostat (Leica Microsystems) using OCT. The cryomicrotome was precooled to −20 °C, and the liver tissue block was mounted onto the specimen disc using OCT compound. Frozen sections were cut into 10 μm thick. 3–5 pieces of sections were collected into pre-cooled 1.5 ml centrifuge tubes for total-RNA extraction.

The RNAeasy™ Animal RNA Isolation Kit with Spin Column (R0024; Beyotime, Shanghai, China) was used to obtain total RNA from frozen sections according to the manufacturer’s guidelines. RNA concentration was measured by nanodrop (ThermoFisher Scientific, MA, USA) and RNA quality was assessed by Qsep1 bioanalyzer (BiOptic Inc., Taiwan, China). Samples with RNA integrity number (RIN) above 8.0 were used for sequencing.

Tissue Optimization

Visium spatial tissue optimization (TO) slides contain eight mRNA capture areas with oligonucleotides, and each capture area is defined by an etched frame. Prior to library preparation, the Visium Spatial Tissue Optimization workflow allows the user to optimize permeabilization conditions for a tissue of interest.

10 μm tissue sections from the same sample were placed onto the Capture Areas of the TO slides. These sections fixed and stained, as described in Tissue Fixation & Staining Demonstrated Protocols (CG000238 Rev D), and then permeabilized for 4 min, 8 min, 12 min, 16 min, 20 min, 24 min, 28 min, respectively. The mRNA released during permeabilization bound to capture probes on the slide, and fluorescently labeled nucleotides were used to visualize synthesized cDNA. Tissue was enzymatically removed, leaving fluorescent cDNA covalently linked to oligonucleotides on the TO slide. Fluorescent cDNA and bright-field images were scanned by a 3DHISTECH Pannoramic Midi scanner (3D HISTECH Ltd., Budapest, Hungary). The permeabilization time that results in maximum fluorescence signal with the lowest signal diffusion is optimal. Thus we chose 8 min for our final permeabilization time.

Visium gateway gene expression process

The Visium Gateway Gene Expression Slide includes 2 Capture Areas (6.5 × 6.5 mm). The Capture Area has ~5,000 gene expression spots. 10 μm tissue sections from the 2 samples were placed onto the Capture Areas of the Visium slides. Tissue sections placed on these Capture Areas were fixed, stained, imaged and permeabilized as described in the previous step, and cellular mRNA was captured by the primers on the gene expression spots. Second Strand Mix was added to the tissue sections on the slide to initiate second strand synthesis. The 10x Genomics Visium Spatial Gene Expression Reagent Kits user guide (CG000239 Rev D) was followed to prepare the spatial transcriptome library. All the cDNA generated from mRNA captured by primers on a specific spot shared a common Spatial Barcode. The cDNA concentration was detected by a Qubit4.0 fluorometer (Invitrogen). Libraries were generated from the cDNA and sequenced and the Spatial Barcodes were used to associate the reads back to the tissue section images for spatial gene expression mapping.

The sequencing process

Libraries were loaded at 300 pM and sequenced on a NovaSeq 6000 System (Illumina) using a NovaSeq S4 Reagent Kit (200 cycles, 20027466, Illumina), at a sequencing depth of approximately 250–400 M read-pairs per sample. Sequencing was performed using the following read protocol: read 1, 28 cycles; i7 index read, 10 cycles; i5 index read, 10 cycles; read 2, 91 cycles.

Visium raw data processing

Raw FASTQ files (Supplementary Table S2) and histology images were processed with the Space Ranger software version 1.3.1, which uses STAR v.2.5.1b for genome alignment21, against the Cell Ranger hg38 reference genome “refdata-cellranger-GRCh38-3.0.0”, available at: (http://cf.10xgenomics.com/supp/cell-exp/refdata-cellranger-GRCh38-3.0.0.tar.gz).

Use of STutility for quality control (QC) and second data analysis after correction of batch effect

We used the R (version 4.1.2, https://www.r-project.org/) and STutility R package (https://github.com/jbergenstrahle/STUtility). We used the “MergeSTData” function to merge the two liver datasets. According to the median number of genes in the liver samples, spots with <200 genes were filtered. As the spatial transcriptome did not capture doublet cells and “soup-RNA,” we did not set an upper limit on the threshold. After QC, 6581 high-quality liver spots were obtained. The relationship between the mRNA reads and the number of mRNAs was detected and visualized (Fig. 2E).

Fig. 2
figure 2

Quality control (QC) of human liver spatial transcriptome data. (A,C) HE staining of the 2 samples. (B, D) Visium Plots showing the spatial location of each cluster. (E) Bar and Feature plots illustrating the number of genes, unique molecular identifiers (UMIs), and total counts of 2 liver samples in each spot.

Full size image

Genes were normalized using the Sctransform package, and high-variable genes (n = 3000) were conducted to the following principal-component analysis (PCA)22. The harmony package was used to correct the batch effect between the two samples23. 30 corrected PCs were selected as the input for uniform manifold approximation and projection (UMAP). We detected the batch effect between the two different liver samples (Fig. 1C). With a resolution of 0.5, spots were clustered by the “FindClusters” function and classified into 15 different spot types. Next, we used the “FindAllMarkers” function to find differentially expressed genes between each type of spots (Fig. 1D,E, Supplementary Table S3). Markers used to define cell types included HAMP, GLUL, CYP3A4, VIM, IGKC, SPP1, CYP1A2, CD163, IGHM, HBA1, ACTA2, and COL3A1. Since the Visium technology obtained a mixed transcriptome of multiple cells in a spot, we defined three specific regions (C5, C8, C9) in combination with spatial locations (Supplementary Figure 1D).

Immunohistochemical validation

All immunohistochemical images were adapted from the Human Protein Atlas24.

Data Records

Raw sequencing data have been deposited into the NCBI sequence read archive (SRA) database under accession number PRJNA82140325. All processed data have been uploaded to the figshare database26. These data include filtered_feature_bc_matrix.h5, cloupe file and spatial files.

Technical Validation

The liver specimens were collected freshly dissected from organ donors, one 57 years-old male and one 39 years-old female. (Supplementary Table S1, Methods). The median genes per spot were 1980, higher than that from previous scRNA-seq dataset10,18 (Fig. 2E).

After QC, 6581 high-quality spots were further analyzed. According to the marker genes, cells were classified into 15 clusters that contained spots in the range of 48–1379 spots per cluster, corresponding to hepatocytes, biliary epithelial cells, liver endothelial cells, B cells and et al. (Fig. 1B,D). The clusters were visualized using UMAP, precisely matching with the HE staining results (Fig. 2A,C).

The results showed that most cells were hepatocytes, consistent with the natural condition. The hepatocytes were extracted and classified into four  sub-clusters. The embeddings of these four sub-clusters on the UMAP were precisely along the lobular axis, thus naming them as Zone 1 (near the portal veins), Zone 2-1, Zone 2-3, Zone 3 (near the central veins), respectively (Fig. 3A). To visualize cells in different zonation with Visium, we determined the region of portal veins and central veins with the help of HE staining and delimited the border of liver lobules (Fig. 2B,D). The differential genes related to the liver zonation, including the peri-portal area genes HAL, HAMP, and CRP and the peri-central area genes GLUL, CYP3A4, and CYP1A1, were located (Fig. 3 C). The IHC staining figures from the human protein atlas further confirmed these results (Fig. 4A–D). In sum, we contributed a reliable spatial transcriptomics atlas, benefiting the study of liver heterogeneity and providing a reference for spatial research of the liver disease.

Fig. 3
figure 3

Spatial transcriptome revealing the zonation distribution of hepatocytes. (A) UMAP plot showing the fine clusters of hepatocytes. (B) HE and Visium Plots showing the zonation of the hepatocytes. (C) Visium and UMAP plots showing the expression levels of each gene on the tissue.

Full size image
Fig. 4
figure 4

Representative IHC plots verifying the zonal distribution of differential genes at the protein level.

Full size image