Tumor phylogeography reveals block-shaped spatial heterogeneity and the mode of evolution in Hepatocellular Carcinoma

Liu, Xiaodong; Zhang, Ke; Kaya, Neslihan A.; Jia, Zhe; Wu, Dafei; Chen, Tingting; Liu, Zhiyuan; Zhu, Sinan; Hillmer, Axel M.; Wuestefeld, Torsten; Liu, Jin; Chan, Yun Shen; Hu, Zheng; Ma, Liang; Jiang, Li; Zhai, Weiwei

doi:10.1038/s41467-024-47541-9

Download PDF

Article
Open access
Published: 12 April 2024

Tumor phylogeography reveals block-shaped spatial heterogeneity and the mode of evolution in Hepatocellular Carcinoma

Nature Communications volume 15, Article number: 3169 (2024) Cite this article

1522 Accesses
28 Altmetric
Metrics details

Subjects

Abstract

Solid tumors are complex ecosystems with heterogeneous 3D structures, but the spatial intra-tumor heterogeneity (sITH) at the macroscopic (i.e., whole tumor) level is under-explored. Using a phylogeographic approach, we sequence genomes and transcriptomes from 235 spatially informed sectors across 13 hepatocellular carcinomas (HCC), generating one of the largest datasets for studying sITH. We find that tumor heterogeneity in HCC segregates into spatially variegated blocks with large genotypic and phenotypic differences. By dissecting the transcriptomic heterogeneity, we discover that 30% of patients had a “spatially competing distribution” (SCD), where different spatial blocks have distinct transcriptomic subtypes co-existing within a tumor, capturing the critical transition period in disease progression. Interestingly, the tumor regions with more advanced transcriptomic subtypes (e.g., higher cell cycle) often take clonal dominance with a wider geographic range, rejecting neutral evolution for SCD patients. Extending the statistical tests for detecting natural selection to many non-SCD patients reveal varying levels of selective signal across different tumors, implying that many evolutionary forces including natural selection and geographic isolation can influence the overall pattern of sITH. Taken together, tumor phylogeography unravels a dynamic landscape of sITH, pinpointing important evolutionary and clinical consequences of spatial heterogeneity in cancer.

PERCEPTION predicts patient response and resistance to treatment using single-cell transcriptomics of their tumors

Article 18 April 2024

A distinct Fusobacterium nucleatum clade dominates the colorectal cancer niche

Article Open access 20 March 2024

Inferring gene regulatory networks from single-cell multiome data using atlas-scale external data

Article Open access 12 April 2024

Introduction

Tumors are complex ecosystems organized into heterogeneous three-dimensional structures, how tumor and non-tumor cells distribute and evolve spatially is a central question in tumor evolution and treatment response^1,2. Even though intratumor heterogeneity (ITH) has been extensively characterized across many cancer types^3,4, previous studies often surveyed generic ITH from the limited number of sectors (e.g., n = 3–5) without in-depth exploration of spatial intra-tumor heterogeneity (sITH)^5,6. Although spatial sampling has been employed to address specific biological questions such as phenotypic differences between central and peripheral regions⁷, microenvironmental spatial heterogeneity^8,9,10 as well as the evolutionary origin of metastasis^7,11, the overall landscape of sITH across cancer types are largely unknown. In addition, previous work often surveyed the ITH at the genetic level, understanding the joint evolution of genotypic and phenotypic ITH at the macroscopic (i.e., whole tumor) level remains understudied.

Parallel to genomic surveys of ITH, many evolutionary models including linear¹², branched^1,4,13, neutral^14,15, and punctuated evolution^16,17,18,19 have been proposed for tumor evolution. However, none of these models are explicit regarding the spatial organization of ITH. Even though computational modeling of tumors as spatially expanding populations has been extensively explored^{6,7,14,20,21,22,23,24,25}, the gap between theoretical modeling and empirical data is still quite large²⁶. As shown in recent computational simulations^20,22, the mode of tumor evolution can be heavily influenced by the spatial organization of tumor cells (i.e., tissue architecture)^20,21,22,27 varying across different tissue types²⁴. Characterizing sITH and studying the mode of spatial tumor evolution is pivotal for understanding the history of tumor evolution.

In this work, we employ a phylogeographic approach and sample exomes and transcriptomes extensively from 235 sectors across 13 HCC patients, generating one of the largest datasets for studying sITH. By integrating physical coordinates of tumor sectors with genomic and transcriptomic information, we dissect spatial distributions of distinct subclones and depict their evolution at the genotypic and phenotypic levels. We find that tumor heterogeneity in HCC segregates into spatially variegated blocks with large genotypic and phenotypic differences. By dissecting the transcriptomic heterogeneity, we discover that 30% of patients had a “spatially competing distribution” (SCD), where different spatial blocks have distinct transcriptomic subtypes co-existing within a tumor, capturing the critical transition period in disease progression. Population Genetic analysis unravel a strong signal of natural selection in SCD patients suggesting the importance of adaptive evolution in driving disease progression. Taken together, we unravel a dynamic landscape of sITH and pinpoint important evolutionary and clinical consequences of spatial heterogeneity.

Results

A phylogeographic approach for surveying sITH

Thirteen patients who underwent surgical treatment for HCC were recruited for this study. As a surgical cohort, most of these patients have early-stage disease and are all hepatitis B positive (Supplementary Data 1). Even though several multi-regional approaches have been conducted for HCC (e.g., recently reviewed in Craig et al.²⁸), the number of sectors taken was often 3–5 and the sITH has not been sufficiently explored. In order to systematically survey the sITH across the whole tumor, we sampled multiple sectors from a central slice of the tumor in a honeycomb manner, resulting in 5–30 sectors (mean = 17) depending on the size of the tumor (Fig. 1a and Supplementary Fig. 1). Adjacent normal liver tissues were also harvested as the genomic control. In total, we have taken 222 tumor sectors and 13 normal samples, resulting in one of the largest datasets for studying sITH in HCC and across cancer types^7,21.

**Fig. 1: The cellular phylogeographic sampling and the IBD pattern.**

In order to survey the genotypic and phenotypic heterogeneity, whole exome sequencing (WES, mean coverage of 98x) and RNA-seq were carried out for all samples (“Methods”, Supplementary Data 2). With somatic variants called across all the samples, we found that CTNNB1, TP53, and chromosomal 1q amplifications are common truncal events shared across sectors (Supplementary Fig. 2a, c). Among all the patients, a wide range of tumor mutation burden (TMB) was found (1.5–11.0 perMb) and tumors with high TMB tend to enrich for mutational signatures such as aristolochic acid (AA) signature (SBS22, Supplementary Fig. 2b). In general, these patients had a similar overall genomic profile as several earlier studies including TCGA²⁹ and the PLANET cohort³⁰ (e.g., Supplementary Fig. 2d and Supplementary Data 2).

Isolation-by-distance relationship across the genotypic and phenotypic space

Multiple hypotheses have been raised regarding the spatial organization of tumor heterogeneity^6,23. For example, in colorectal cancer (CRC), previous studies found that subclones tend to be well organized in geographically distinct regions in pre-cancerous lesions. However, this clear segregation will rapidly progress to a spatially variegated distribution in carcinoma²³, suggesting that spatial mixing might be a hallmark of cancer. In order to survey the sITH, we measured genetic differentiation between multiple tumor populations (sectors) based on the genetic distance which characterizes the proportion of private mutations between two sectors (“Methods”). When we correlated the genetic distances with the physical distances of tumor sectors, a wide range of correlations were found across the patients. For example, SH06 has very strong genetic differentiation across sectors and shows a strong positive relationship between genetic differentiation and the physical distance (Fig. 1b). In evolutionary genetics, positive correlation between physical and genetic distance is known as the isolation-by-distance (IBD) relationship³¹. In addition to overall genetic differentiation, when we performed clonal deconvolution of multiple tumor sectors and measured distances in clonal composition between tumor sectors (“Methods”), we also observed the IBD relationship consistently (Supplementary Figs. 3 and 4). Thus, the IBD relationship is robust across multiple distance metrics (Supplementary Note 1 and Supplementary Figs. 3 and 4). Throughout the patient cohort, most patients show strong IBD pattern among the sectors (Fig. 1c).

Parallel to the genetic differentiation, when we surveyed the phenotypic (transcriptomic) differentiation based on the dissimilarity across tumor sectors (“Methods”), strong IBD relationships were also found across patients (Fig. 1b, d) and were robust to many confounding factors such as tumor purity and copy number alterations (Supplementary Note 2 and Supplementary Figs. 5–7). Correlating the genetic differentiation with the transcriptomic difference, we found that large proportions of phenotypic variation can be explained by genetic differences (Fig. 1b, e). This suggests that genetic changes, even though vary greatly across tumors, can have a very strong impact on the phenotypic evolution (Fig. 1e). Taken together, we observed an IBD pattern where physically closer sectors are genotypically and phenotypically more similar across all the tumors^32,33.

Spatially variegated blocks in HCC

The IBD distribution provides an intuitive overview of sITH along one dimension. However, the spatial distribution of sITH within solid tumors is significantly under-explored. The phylogeographic sampling provides a comprehensive means surveying the spatial heterogeneity across the tumors. Focusing on the 10 patients with at least 10 sectors (mean = 21 sectors per tumor), we found that even though IBD is consistent across patients, many patients have a distorted linear relationship. For example, two parallel linear relationships exist in patient SH05 and many nearby samples have surprisingly high genetic differentiation (top left corner of Fig. 2a). In evolutionary genetics, distorted IBD relationship is often driven by uneven spatial distribution³⁴. By calculating local Geary’s C, a spatial statistic that measures the local spatial autocorrelation³⁵, we found multiple regions within the tumor have a much lower correlation between genetic information and spatial location (Supplementary Fig. 8). By applying the phylogenetic analysis, principal component analysis (PCA) as well as clonal deconvolution analysis to patient SH05 (“Methods”, Supplementary Note 3 and Supplementary Fig. 9), we indeed discovered two major clades of tumor sectors clearly separating the samples into two groups (Fig. 2b, c). Thus, the strong differentiation between subgroups within SH05 generates strong discontinuity in the IBD relationship.

In order to systematically explore the overall landscape of spatial heterogeneity across the 10 patients, we clustered the tumor sectors based on the phylogenetic distance between the samples (“Methods”). To choose the optimal number of clusters and compare the spatial segregation of samples between patients, we calculated the Calinski-Harabasz (CH) Index which measures the ratio of genetic differences (i.e., phylogenetic distance) between clusters vs within clusters (“Methods”). Using the CH index, we found the optimal number of clusters are often between two to three (e.g., SH05, Fig. 2e), even though higher numbers of clusters are found for a few patients (e.g., SH03, Fig. 2f). Interestingly, when we inspected the geographic distribution of these clusters, samples from the same cluster often locate in spatially continuous regions within the tumor, partitioning the tumors into variegated blocks (Fig. 2e–g and Supplementary Fig. 3). Thus, we referred “spatially distinct clusters” that partitioning the tumor as “spatial blocks” of the cancer. Strikingly, the genetic differentiation between blocks is often much higher than genetic differentiation within blocks (Fig. 2d), and regions near the block boundary tend to have low spatial correlation (Supplementary Fig. 8). Thus, the heterogeneity within HCC is spatially variegated with regional homogeneity within blocks, but large differentiation between blocks. Parallel to the natural world, this spatial distribution is very similar to our planet where tectonic plates partition the surface of the earth into discrete regions³⁶. For example, two blocks exist in patient SH05 and a central ridge of high regional genetic diversity (θ_R) was found at the interface of two blocks in SH05 (“Methods”, Fig. 2e and Supplementary Fig. 10). Across the patients, the spatial blocks had a wide variety of shapes (Fig. 2e–g and Supplementary Fig. 3). Taken together, we found that HCC has block-shaped spatial heterogeneity across the tumors (Fig. 2g and Supplementary Fig. 3).

Block-shaped phenotypic heterogeneity mirrors the genotypic heterogeneity

As tumors often have block-shaped heterogeneity at the genetic level, we wondered how the transcriptomic heterogeneity might co-evolve spatially within a tumor. Using SH05 as an example, when we analyzed the transcriptomic divergence (Fig. 2h), the phylogenetic relationship (Fig. 2i) as well as the PCA map (Fig. 2j), we found similar block-shaped transcriptomic heterogeneity largely matching the corresponding pattern at the genetic level. When we clustered the transcriptomic profiles into optimal number of clusters, we observed similar spatial blocks between the phenotypic and genotypic level, even though differences do exist in several patients such as SH11 and SH13 (Supplementary Fig. 3). Taken together, phenotypic heterogeneity largely mirrors the genotypic heterogeneity with similar spatial pattern across the tumors.

Virtual micro-dissection of bulk transcriptome identified four HCC subtypes

The block-shaped spatial heterogeneity indicates non-gradual sITH in HCC and the spatial heterogeneity is driven by the geographic distribution of multiple cell types including tumor, stroma, and immune cells. In order to understand biological processes and cellular components driving the transcriptomic heterogeneity in HCC, we employed a reference-free deconvolution method³⁷ to deconvolute the bulk transcriptome into subcomponents (known as compartments)³⁸. In order to capture as much transcriptomic heterogeneity as possible for HCC, we used the TCGA-LIHC cohort (n = 369) as a reference set and identified three compartments positively correlated with tumor purity (i.e., phenotypes associated with tumor cells) and one compartment negatively correlated with tumor purity reflecting the microenvironmental changes (Supplementary Fig. 11). Using functional enrichment analysis of factor genes associated with each compartment³⁸, we found that compartments positively correlated with tumor purity are enriched for metabolism, cell cycle, Wnt pathway respectively, while the compartment negatively correlated with tumor purity is enriched for extra-cellular matrix (ECM) function (Fig. 3a and Supplementary Fig. 11). Across the cohort, downregulated metabolism and increased cell cycle are strongly associated with disease progression³⁹ and patient survival in HCC (Fig. 3b–d and Supplementary Fig. 11).

**Fig. 3: Virtual deconvolution identifies four HCC subtypes.**

Combining all four compartments, we discovered four distinct transcriptomic subtypes for HCC (Fig. 3e, “Methods”). The first group is enriched for CTNNB1 mutations and has strong activation of the Wnt pathway. Since this subgroup significantly overlaps with the CTNNB1 subtype from Chiang et al.⁴⁰, we thus name this subgroup as the Wnt subtype. Among the other three subgroups, one subtype had molecular phenotypes closest to the normal with low levels of cell cycle, high expression of metabolic pathways, and activated ECM. As patients from this group are enriched for early-stage tumors with good survival (Fig. 3f, g), we name it as the classic subtype. We also identified one subtype with the most advanced phenotypes including high cycle cell and decreased metabolic functions. Since patients in this subtype significantly match the proliferation subtype from Chiang et al.⁴⁰ and are enriched for late-stage patients (Fig. 3e–g), we name it as the proliferation subtype. The other subtype has phenotypes in between the classic and proliferation subtype and was named as the hybrid (i.e., intermediate) subtype³⁸. Across classic, hybrid, and proliferation subtypes, we observed an increasingly higher proportion of late-stage patients with worse survival (Fig. 3f, g).

In order to explore genomic changes differentiating these subtypes, we systematically compare the tumor and non-tumor components across subtypes. As the Wnt subtype is strongly associated with CTNNB1 mutations and represents a unique subclass of tumors, we focused on differences between the other three subtypes. When we compare genetic changes in the tumor cells, we noticed that a strong increase in genome instability between the classic and the other two subtypes, accompanying the upregulation of cell cycle related pathways (Fig. 3h, p-value < 10⁻⁶). When we compared the non-tumor components, in particular the immune cell composition (“Methods”), the hybrid subtype has an immune-exclusive status (Fig. 3e, i and Supplementary Fig. 12), while the proliferation subtype is immunologically exhausted with higher M0 macrophages and Tregs (Fig. 3j). Taken together, a collection of genotypic and phenotypic changes is associated with subtype differences in HCC (Supplementary Fig. 13).

Multiple transcriptomic subtypes spatially coexist within a tumor

With the virtual microdissection approach, we have systematically explored the transcriptomic heterogeneity between patients. Within patient tumors, we discovered block-shaped spatial heterogeneity with large phenotypic and genotypic differences between spatial blocks. We wondered how this sITH within patients might be linked to inter patient differences. By clustering all our samples with the TCGA patients and assigning subtypes to the samples (Supplementary Fig. 14), four patients (SH05, SH06, SH10, and SH12) were found to have mixed transcriptomic subtypes within the same tumor. SH05 and SH10 have mixtures of hybrid and proliferation subtypes (Fig. 4a–c), while SH06 has a mixture of classic and proliferation subtypes (Fig. 4d). The spatial partition of subtypes matches almost perfectly with spatial blocks at the transcriptomic level (Fig. 4a, c, d and Supplementary Fig. 3). When we performed differential expression analysis between different spatial blocks within a tumor, we found that transcriptomic differences within patients are highly concordant with differences between transcriptomic subtypes at the inter-patient level (Fig. 4b, e, i). Inspecting genetic differences between different spatial blocks in the patients with mixed transcriptomic subtypes, we found that putative oncogenic drivers are often associated with the spatial blocks with more aggressive phenotypes. One particularly striking observation is a significant increase in chromosomal copy number alterations (CNA) in SH06 as the sectors evolved from the classic subtype to the proliferation subtype (Fig. 4f, g), matching the differences in CNAs observed at the cohort level (Fig. 3h). Even though large chromosomal changes have been repeatedly found driving punctuated evolution in tumor initiation^16,17,18,19, the observation here suggests that large chromosomal changes might also be driving tumor progression see Discussion. The fourth patient (SH12) had only five sectors and was not included in the 10 patients surveying sITH. A mixture of the Wnt and classic subtypes exist in SH12, correlating with a subclonal CTNNB1 mutation (Fig. 4h, i).

**Fig. 4: Multiple transcriptomic subtypes within tumors.**

The role of Darwinian selection in shaping the spatial heterogeneity

The spatially co-existing transcriptomic subtypes in different geographic locations of the tumor is very similar to an era in ancient Chinese history known as the Warring States (Fig. 5a)⁴¹. During the Warring States period (475–221 BCE), seven different states coexisted and competed with one another until the rise of the Qin dynasty which unified China into the first empire. Thus, we name the spatial distribution as the spatially competing distribution (SCD). As tumors can only start with one transcriptomic subtype, the coexistence of multiple subtypes primed a fundamental question: what are the evolutionary forces driving the origin of the SCD? From the subtype comparison, we observed that tumor cells with more aggressive subtypes often have additional genetic changes (e.g., much higher genome instability, Fig. 3h) and advanced phenotypes (e.g., higher cell cycle, Fig. 3e), representing derived populations further evolved from the benign subtype during the late stage of tumor progression. Thus, the spatially competing distribution might represent a critical stage of disease progression where phenotypically aggressive subtypes have emerged and are in the process of replacing relatively benign groups (Fig. 5a).

**Fig. 5: Darwinian selection in patients with the spatially competing distribution.**

One prediction from the neutral theory is that the size of a clone depends on the time of origin and we would expect later-arising clones to be in minor frequencies under neutrality⁴². However, we found the size of the derived (i.e., aggressive) subtypes are often much larger than the ancestral (i.e., benign) subtypes, contrasting the prediction from the neutral evolution. For example, in patient SH05, the tumor region with less aggressive phenotypes (i.e the hybrid subtype) is much smaller (n = 5 sectors) than the area with the aggressive phenotype (i.e., the proliferation subtype, n = 25 sectors, Fig. 5a). Using the classical imbalance index in phylogenetics known as the Sackin index S, we also found significant evidence for clonal asymmetry for SH05 (p-value < 10⁻⁴, Fig. 5b). In all three patients in the SCD, more aggressive transcriptomic subtypes always occupy larger proportions of the tumors (i.e., clonal asymmetry, Fig. 5b), rejecting neutrality for these patients. In order to understand the prevalence and significance of this observation, when we extended the analysis to the PLANET cohort³⁰, we observed clonal asymmetry for all three additional patients in the SC distribution where phenotypically derived clones are consistently larger than ancestral clones (Fig. 5b).

Parallel to the clonal asymmetry at the phenotypic level, the other evidence for ongoing positive selection at the genetic level was the existence of subclones at intermediate frequencies. Using MOBSTER⁴³, we detected subclonal selection in all tumors with the SC distribution (Fig. 5c, d and Supplementary Fig. 15), which suggests that Darwinian selection can be the major evolutionary force driving tumor evolution in the SC distribution. Interestingly, when we extended the same statistical tests to the full cohort, we found varying levels of non-neutral evolution across the cohort, suggesting that Darwinian selection together with several other evolutionary forces (e.g., geographic isolation) can contribute to the block-shaped spatial heterogeneity (see Discussions).

sITH can significantly confound patient diagnosis and treatment

The extensive sITH raised an important question: how this spatial heterogeneity might affect clinical practice. From the literature, we curated a list of diagnostic and treatment-related biomarkers, including a transcriptomic predictor for the micro-vascular invasion (MVI) and immunotherapy biomarkers such as the GEP score (pan cancer) and the inflammatory score (HCC) (Supplementary Data 3). By plotting the variation of these biomarkers, we discovered extensive heterogeneity between and within patients (Fig. 6a and Supplementary Fig. 16). Comparing the variance of these biomarkers across patients (Fig. 6b), we found that a few patients such as SH06, SH08, and SH09 tend to have higher variance in these biomarkers. In general, the variance of these biomarkers is rather heterogeneous across patients.

**Fig. 6: Spatial heterogeneity of biomarkers.**

In addition to intra-tumor heterogeneity, when we look at the distribution of these biomarkers across transcriptomic subtypes (Fig. 6c), we found that some biomarkers correlate strongly with disease progression. For example, the MVI score has increasingly higher values from the classic to the proliferation subtype. On the other hand, the TIDE score had relatively similar values across subtypes, suggesting that different biomarkers can have different trajectories with disease progression. When we dissected the spatial distribution of these biomarkers across patients, we found that some biomarkers had highly concordant spatial distribution with spatial blocks (e.g., the GEP score, Fig. 6d), while others have a heterogeneous spatial distribution irrespective of the spatial blocks (e.g., Cytolytic scores, Fig. 6e and Supplementary Figs. 17 and 18). This suggests that spatial segregation of biomarkers across the tumor might not always follow the overall sITH. Taken together, sITH can significantly confound patient clinical diagnosis and treatment and a single biopsy is not enough for clinical decisions (see Discussion).

Discussion

Using a phylogeographic approach and one of the largest datasets for studying sITH^7,21, we have systematically dissected the genotypic and phenotypic spatial heterogeneity across a large number of tumor sectors from 13 HCC patients. Contrary to the spatial mixing observed in CRC, spatial heterogeneity in HCC mirrors the geography with an isolation-by-distance pattern. The spatial heterogeneity in HCC is organized into spatial blocks with large genotypic and phenotypic divergence. Using a virtual dissection approach, we discovered a spatial distribution defined as the spatially competing distribution where spatial blocks within a tumor have different transcriptomic subtypes in 30% of patients. The spatially competing distribution represents a critical transitional stage where multiple phenotypically distinct subpopulations coexist and compete during disease progression. By integrating the spatial distribution and clonal composition, we found that Darwinian selection is the driving force for disease progression in patients with SC distribution and can also contribute to the block-shaped sITH. Taken together, we have generated one of the largest datasets for studying macroscopic sITH and unraveled a dynamic landscape of sITH.

The study of sITH provides several insights into tumor evolution. First, limited sampling of sITH commonly practiced in the field might lead to the “blind man and an elephant” bias. For example, we can imagine different mode of tumor evolution can be inferred if we take samples within (linear and neutral evolution) or between different spatial blocks (branched and non-neural evolution). The extensive spatial sampling provides a holistic overview of the whole tumor and different modes of tumor evolution observed earlier might be confounded by sITH. Secondly, even though we found that natural selection is an important factor driving the spatial heterogeneity in the patients with SC distribution, they are exceptional cases with both large phenotypic and genotypic differences. It remains unknown what are the evolutionary forces driving the collective pattern of spatial heterogeneity including block-shaped heterogeneity and clonal asymmetry. Even though varying levels of selective signal across the patient cohort were detected, previous studies in evolutionary genetics suggest that geographic isolation together with a suite of non-selective forces can also contribute to the spatial pattern⁴⁴. Taken together, this study pointed directions how we can dissect the evolutionary forces driving the spatial heterogeneity and how we may formulate more rigorous methods in detecting natural selection at both genotypic and phenotypic levels (Supplementary Note 4).

The study of spatial heterogeneity also highlighted challenges facing patient diagnosis and treatment. Even though truncal driver mutations can be well represented by a single biopsy⁴⁵, many of the diagnostic and treatment biomarkers are phenotypic measurements and often have extensive spatial variation. For example, the immune microenvironment tends to have high diversity and does not always follow the genetic diversity observed in tumor cells (Supplementary Fig. 18). Moreover, when we compare the biomarker distribution between the central and peripheral regions of the tumors, no significant differences were found except the vascular invasion (Supplementary Fig. 19). As we found earlier, ancestral clones can reside in different parts of the tumors and tumor growth can take a wide variety of shapes which might yield complex spatial distribution without systematic differences between inside and outside of the tumor. Thus, how to design spatially guided approaches for patient diagnosis and treatment can be a promising direction for the community.

Methods

Patient recruitment and spatial sampling

From the Ditan Hospital affiliated with Capital Medical University in Beijing, we recruited 13 treatment naïve HCC patients with approval from the Institutional Review Board (IRB) of the Ditan Hospital. Signed informed consent was obtained from each patient before surgery. After surgical operation, a single slice of tumor sample was harvested from the center of the tumor and multiple tumor sectors (3 mm or 5 mm in diameter) were harvested across the 2-D space depending on the size of the tumor (Fig. 1a). In total, 222 tumor sectors ranging from 5–30 regions per tumor (T) and 13 matched adjacent non-tumor liver tissues (N) were harvested (Supplementary Fig. 1). The physical coordinates of the tumor sectors were subsequently extracted based on the sampling image. Sex and/or gender were not considered in the study design.

Genomic sequencing and variant calling

DNA and RNA were extracted from tissue samples using AllPrep DNA/RNA Mini Kit(Qiagen). Exome libraries were constructed using the Agilent SureSelect Human All Exon V6. RNA sequencing libraries were generated using NEBNext® UltraTM RNA Library Prep Kit from Illumina® (NEB, USA) following the manufacturer’s recommendations. Both exome and RNA sequencing were conducted at Novogene Co., Ltd. with 150 bp paired-end reads.

After quality control which includes trimming adapters and removing low-quality reads, we mapped short reads to the hg19 reference genome using BWA MEM (v0.7.17)⁴⁶. Duplicated reads were subsequently marked using MarkDuplicates and base quality scores were recalibrated using GATK4 (v4.1.3.0)⁴⁷. Somatic variants were called using Mutect2⁴⁸. Variants were further filtered using GATK FilterMutectCalls with default parameters and were annotated with Oncotator (v1.9.9.0)⁴⁹. Somatic copy number variations as well as tumor purity were inferred using Sequenza (v3.0.0)⁵⁰. The genome instability index (GII) is defined as the proportion of the genome with abs(log2(CN/2)) > 0.2, where CN is the copy number of the focal genomic region⁵¹.

Signature analysis

We used SigProfiler⁵² to perform signature analysis. We first used SigProfilerMatrixGenerator⁵³ (v1.2.1) to create a mutation matrix for all types of mutations. We then applied SigProfilerExtractor (v1.1.4) to extract de-novo mutational signatures from the mutational matrix and further deconvolute de-novo signatures to the COSMIC signatures (v3.2)⁵².

Curating putative driver events in HCC

A list of 62 driver genes for HCC was extracted from a recent integrative analysis of HCC genomes⁵⁴. Significantly mutated copy number alterations were extracted by reanalyzing the TCGA-LIHC dataset using GISTIC2 (v2.0.23) algorithm⁵⁵ (Supplementary Fig. 2a, d).

Transcriptomic analysis

The raw RNA sequencing reads were aligned to the reference genome using STAR (v2.7.2b)⁵⁶ and further quantified using RSEM (v1.3.1)⁵⁷. We used DEseq2 (v1.30.1)⁵⁸ to normalize expression levels of different samples and performed differential expression analysis. Transcriptomic distance between two samples was calculated as the (1-Spearman correlation) between two samples using top most variable 3000 genes (based on median absolute deviation, MAD). In order to explore the influence of tumor purity on tumor spatial phenotypic heterogeneity, we identified genes that were positively correlated with tumor purity in our data as those with Spearman correlation coefficient greater than 0.5 and used this subset of genes to calculated the transcriptomic distance (Supplementary Note 2). In order to explore the influence of copy number variation on tumor phenotypic spatial heterogeneity, we constrained the gene list to genes with no copy number variation in all sectors of a given patient and calculated transcriptomic distances based on this subset of genes (Supplementary Note 2).

Phylogenetic, clonal deconvolutional, and population genetic analysis of tumor samples

The phylogenetic and clonality analysis followed similar steps as our previous work³⁰. The genetic distance between two sectors was calculated as the Hamming distance between the two mutation lists called from two sectors. We used the Analysis of Phylogenetics and Evolution (APE) from R to perform the phylogenetic analysis. We first tabulated the presence and absence of mutations across sectors for each patient, maximum parsimony⁵⁹ was employed to construct the phylogenetic relationship. In addition to the maximum parsimony, Neighbor-joining algorithm⁶⁰ was also employed to construct the phylogenetic relationship based on the Hamming distance calculated using the presence and absence of mutations. We also used the 1-Spearman correlation as the distance at the transcriptomic level to construct the neighbor-joining tree at the transcriptomic (RNA) level.

Based on somatic variants and local copy number variations, PyClone (v1.13.1)⁶¹ was employed to dissect the clonal composition of multiple tumor samples, which infers the predicted number of clones, clonal structure, mutation assignment, and mutation VAF distribution in each clone. Before PyClone analysis, we have performed these steps: 1) take a union of all the SNVs across all sectors for a given patient; 2) extract the number of reference and alternative allele count at the SNV positions for all sectors (based on Samtools mpileup). 3) We required the SNVs to have a frequency > 0.05 in at least one sector (to filter away possible false positives). Clonality inference was performed based on SNVs only as variant frequency estimation for indels are much less accurate.

Similar to our previous study³⁰, CITUP (v0.1.0)⁶² was used to construct the optimal clonal phylogeny from the clonal composition inferred from PyClone. We first filtered away clones with only one mutation from the PyClone output and subsequently took the top 10 clones with the highest number of mutations³⁰ (CITUP analysis with too many clones can be computationally too intensive⁶²). Clonal phylogenies were visualized using the ggplot2 package in R.

Alternative approaches correlating the genetic distance and physical distance

In addition to correlating the genetic distance and physical distance, we also explored alternative approaches. In particular, when we correlated the phylogenetic distance as well as the clonal distance with physical distance, we observed a consistent linear relationship (i.e., IBD relationship, Supplementary Note 1). The phylogenetic distance was extracted from the phylogenetic relationship. The clonal distance is measured as the differences in the clonal composition and is calculated as the Euclidian distance between samples based on their clonal composition. To be more precise, if there are n clones presented in two samples with proportions (p₁, p₂…p_n) and (q₁, q₂, q₃…q_n), we calculated the distance in the clonal composition between two samples as $\sqrt{{\sum }_{i}^{n}{\left({p}_{i}-{q}_{i}\right)}^{2}}$.

Spatial correlation and the Geary’s C statistic

For each sector i, the local Geary’s C index ${C}_{i}$ is calculated as the following formula:

$${C}_{i}=\frac{{2n}^{2}}{{\Sigma }_{i}{\Sigma }_{j}{D}_{{ij}}^{2}}\mathop{\sum }\limits_{j}{w}_{{ij}}{D}_{{ij}}^{2},$$

(1)

where $j$ represents another sector in the tumor, and $n$ represents the total number of sectors in the patient. ${w}_{{ij}}$ denotes the spatial weight between sector $i$ and $j$. We employed the canonical form of ${w}_{{ij}}$ by taking the inverse of the physical distance between sector i and j. ${D}_{{ij}}$ represents the genetic distance between sector $i$ and $j$ which we used the Hamming distance between mutations found in sector i and j.

To test the significance of the observed value, we calculated the empirical p-value of the observed local Geary’s C. We randomly permuted the sectors spatially and recalculated the Geary’s C and computed the proportion of replicates where the permuted ${C}_{i}{{{{{\rm{s}}}}}}$ are less than the observed ${C}_{i}$. In case of neighborhood with strong spatial autocorrelation, the local Geary’s C will be small because of smaller values of ${D}_{{ij}}$. On the contrary, regions with weak spatial autocorrelation will have very large p-values because of large values of ${D}_{{ij}}$ in the local neighborhood.

Spatial blocks and the spatial heterogeneity of the tumor

In order to explore the spatial distribution of ITH, we clustered the genetic profiles of tumor sectors following these steps. 1) Based on the parsimony tree, we cut the phylogenetic tree into all possible k clusters for a given number of clusters k (k = 2−5). 2) In order to compare the clusters within and between clusters, we computed a statistic Calinski-Harabasz Index (CH-index)^63,64. CH-index is commonly applied for clustering evaluation^63,64. It is calculated as the ratio of the inter-cluster dispersion and the intra-cluster dispersion, where dispersion is measured by the mean sum of distances. For a collection of $N$ samples that are partitioned into $K$ clusters, the sum of distances within and between clusters can be calculated as:

$$S{S}_{{within}}=\mathop{\sum }\limits_{k=1}^{K}\mathop{\sum }\limits_{i=1}^{{n}_{k}}\mathop{\sum }\limits_{j=i+1}^{{n}_{k}}\frac{d\left(i,\, j\right)}{{n}_{k}},$$

(2)

$$S{S}_{{between}}=\mathop{\sum }\limits_{i=1}^{N}\mathop{\sum }\limits_{j=i+1}^{N}\frac{d\left(i,\, j\right)}{N}-S{S}_{{within}},$$

(3)

where, ${n}_{k}$ is the number of samples in cluster $k$ ($k={{{{\mathrm{1,2}}}}},\ldots,K$), and $d(i,\, j)$ denotes the distance between pairwise samples $i$ and $j$. The CH-index is then calculated as:

$${CH}=\frac{{{SS}}_{{between}}\times ({N} - {K})} {{{{\mbox{SS}}}}_{{within}}\times (K-1)}.$$

(4)

We picked the optimal number of clusters according to the CH-index (maximum value). In the clustering of the genetic pattern, we used the phylogenetic distance as the distance metric. In the clustering of the transcriptomic profile, we use the phylogenetic distance in the transcriptomic tree as the distance metric.

In order to display the spatial heterogeneity across the 2-D space, we calculated ITH values at grid points distributed along the 2-D space (with a grid size of 2 mm in distance). We defined the local region as a circular area centered around the grid point and we have chosen the diameter as the median distance between all sectors of that tumor. The regional diversity (i.e., θ_R) is defined as the average difference between all pairs of samples in the circular area. For pairs of samples, the genetic (DNA) distance is calculated as the proportion of branch mutations (i.e., non-shared mutations) among all mutations found in two samples. In order to explore how variation in sequencing depth might affect θ_R, we down-sampled the sequencing depth of all sectors to the minimum sequencing depth across all the sectors for a given patient. We found that the re-estimated θ_R values are highly similar to the original values (Supplementary Fig. 10). In order to depict the spatial variation in diversity, we generated a spatial heatmap with a coloring hue proportional to the level of θ_R using the “stat_density_2d” method from the R package ggplot2.

Virtual deconvolution and the transcriptomic subtypes of HCC

Raw expression counts for the TCGA LIHC cohort (n = 369) were downloaded from the TCGA website and subsequently normalized using Deseq2⁵⁸. We employed DECODER³⁷ to identify subcomponents (known as compartments) and infer compartment weights for each sample using the non-negative matrix factorization (NMF) method. We set 10,000 resampling to train the seeding matrix and DECODER identified five different compartments.

To annotate the biological functions of each compartment, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of the factor genes from each compartment using the R package “ClusterProfiler”. We have chosen four compartments with significant correlation to tumor purity as the base for identifying HCC subtypes. The fifth compartment has no correlation to tumor purity and factor genes associated with the fifth compartment are not enriched for any biological function. For each compartment, we chose the top 100 factor genes from each compartment and then used the combined set of factor genes across all four compartments to perform the consensus clustering using the ConsensusClusterPlus from the R package. In the consensus clustering, we resampled 80% samples and features (expression) and hierarchical clustering was performed for each resampled dataset. We set the number of iterations to be 500 and used 1- Spearman correlation distance as the distance metric. The cumulative distribution function (CDF) curves of the consensus matrix were used to determine the optimal number of clusters. In order to assign transcriptomic subtypes to our patient samples, we combined LIHC dataset with our cohort and reperformed the clustering.

Annotating samples for known transcriptomic subtypes

To examine whether our clustering results are consistent with known transcriptomic subtypes from the literature^39,40,65,66, we downloaded pathways associated with the transcriptomic subtypes curated in MSigDB⁶⁷ and applied GSVA analysis to each sample to compute a pathway score for each sample. Samples were then assigned to the subtype with the highest pathway score.

Immune score and immune cell proportion

Immune scores estimated by ESTIMATE⁶⁸ were retrieved from ESTIMATE’s website at https://bioinformatics.mdanderson.org/estimate/. Immune cell proportions were estimated using CIBERSORTx⁶⁹. ConsensusTME (v0.0.1.9) was also used to infer the composition of the tumor microenvironment⁷⁰. We calculated the immunological regional diversity in a similar fashion as the genetic and transcriptomic level. In particular, we used the dissimilarity distance (i.e., 1-Spearman correlation) between samples based on the proportion of immune cells estimated by CIBERSORTx.

Neutrality tests

In order to test for neutrality, we performed several statistical tests: 1) To test for clonal asymmetry, we applied the Sackin test to test if the tree shapes fit the null hypothesis of equal fitness (i.e., Yule process). We used the “sackin”.test from “apTreeshape” (v1.5-0.1) to perform the test⁷¹. 2) In order to test the existence of the selected subclone, we used MOBSTER (v0.1.1)⁴³ to identify the selected subclone at intermediate frequencies. In order to obtain a corrected variant allele frequency (VAF) for MOBSTER, we first used Samtools mpileup to compute the raw VAF for all the somatic variants across all tumor sectors of a given patient. To adjust for tumor purity and local copy number, PyClone was used to infer the Cancer Cell Fraction (CCF). Sites with corrected VAF < 0.01 were removed before the subclonal deconvolution using MOBSTER.

Biomarkers for HCC

To explore spatial heterogeneity in diagnostic and treatment-related biomarkers, we collected 8 biomarkers including TIDE⁷², GEP⁷³ and IMPRES⁷⁴, Cytolytic⁷⁵, Immune class⁷⁶, inflammatory⁷⁷, vascular invasion⁷⁸ and metastasis⁷⁹ score. TIDE, GEP, and IMPRES score were calculated using the original method. The other biomarkers were calculated using GSVA based on their marker genes. In order to compare the biomarker distribution between the central and peripheral regions of the tumor, samples that are within r/2 from the central point of the tumor were defined as the center of the tumor, where r is the radius of the tumor.

Statistical analysis

Linear regression was performed based on the least-square method. Multiple test correction is based on the Benjamini-Hochberg (BH) approach. When testing differences in two categories (e.g., Figs. 3h–j and 4g), the Wilcox test (two-sided test) is employed.

Reporting summary

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

Data availability

The raw data of exome and RNA sequencing have been deposited in the Genome Sequence Archive (GSA) hosted at the National Genomics Data Center (HRA002112). The TCGA-LIHC dataset was obtained from the Genomic Data Commons (GDC) data portal [https://portal.gdc.cancer.gov/projects/TCGA-LIHC]. Source data are provided in this paper.

Code availability

The scripts for analyzing the data and plotting the results are available at GitHub hosted at https://github.com/LiuXiaodong1/HCC_SPH.

References

Greaves, M. & Maley, C. C. Clonal evolution in cancer. Nature 481, 306–313 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
McGranahan, N. & Swanton, C. Biological and therapeutic impact of intratumor heterogeneity in cancer evolution. Cancer Cell 27, 15–26 (2015).
Article CAS PubMed Google Scholar
Junttila, M. R. & de Sauvage, F. J. Influence of tumour micro-environment heterogeneity on therapeutic response. Nature 501, 346–354 (2013).
Article ADS CAS PubMed Google Scholar
McGranahan, N. & Swanton, C. Clonal heterogeneity and tumor evolution: past, present, and the future. Cell 168, 613–628 (2017).
Article CAS PubMed Google Scholar
Tanaka, N. et al. Whole-tissue biopsy phenotyping of three-dimensional tumours reveals patterns of cancer heterogeneity. Nat. Biomed. Eng. 1, 796–806 (2017).
Article CAS PubMed Google Scholar
Waclaw, B. et al. A spatial model predicts that dispersal and cell turnover limit intratumour heterogeneity. Nature 525, 261–264 (2015).
Article ADS CAS PubMed PubMed Central Google Scholar
Zhao, Y. et al. Selection of metastasis competent subclones in the tumour interior. Nat. Ecol. Evol. 5, 1033–1045 (2021).
Article CAS PubMed PubMed Central Google Scholar
Carmona-Fontaine, C. et al. Metabolic origins of spatial organization in the tumor microenvironment. Proc. Natl Acad. Sci. USA 114, 2934–2939 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Hoefflin, R. et al. Spatial niche formation but not malignant progression is a driving force for intratumoural heterogeneity. Nat. Commun. 7, ncomms11845 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Lloyd, M. C. et al. Darwinian dynamics of intratumoral heterogeneity: not solely random mutations but also variable environmental selection forces. Cancer Res. 76, 3136–3144 (2016).
Article CAS PubMed PubMed Central Google Scholar
Ryser, M. D. et al. Minimal barriers to invasion during human colorectal tumor growth. Nat. Commun. 11, 1280 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Fearon, E. R. & Vogelstein, B. A genetic model for colorectal tumorigenesis. Cell 61, 759–767 (1990).
Article CAS PubMed Google Scholar
Nowell, P. C. The clonal evolution of tumor cell populations. Science 194, 23–28 (1976).
Article ADS CAS PubMed Google Scholar
Ling, S. et al. Extremely high genetic diversity in a single tumor points to prevalence of non-Darwinian cell evolution. Proc. Natl Acad. Sci. USA 112, E6496–E6505 (2015).
Article CAS PubMed PubMed Central Google Scholar
Williams, M. J., Werner, B., Barnes, C. P., Graham, T. A. & Sottoriva, A. Identification of neutral tumor evolution across cancer types. Nat. Genet. 48, 238–244 (2016).
Article CAS PubMed PubMed Central Google Scholar
Baca, S. C. et al. Punctuated evolution of prostate cancer genomes. Cell 153, 666–677 (2013).
Article CAS PubMed PubMed Central Google Scholar
Gao, R. et al. Punctuated copy number evolution and clonal stasis in triple-negative breast cancer. Nat. Genet. 48, 1119–1130 (2016).
Article CAS PubMed PubMed Central Google Scholar
Stephens, P. J. et al. Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell 144, 27–40 (2011).
Article MathSciNet CAS PubMed PubMed Central Google Scholar
Wang, Y. et al. Clonal evolution in breast cancer revealed by single nucleus genome sequencing. Nature 512, 155–160 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Chkhaidze, K. et al. Spatially constrained tumor growth affects the patterns of clonal selection and neutral drift in cancer genomic data. PLoS Comput. Biol. 15, e1007243 (2019).
Article CAS PubMed PubMed Central Google Scholar
Fu, X. et al. Spatial patterns of tumor growth impact clonal diversification in a computational model and the TRACERx Renal study. Nat. Ecol. Evol. 6, 88–102 (2022).
Article PubMed Google Scholar
Noble, R. et al. Spatial structure governs the mode of tumor evolution. Nat. Ecol. Evol. 6, 207–217 (2022).
Sottoriva, A. et al. A big bang model of human colorectal tumor growth. Nat. Genet. 47, 209–216 (2015).
Article CAS PubMed PubMed Central Google Scholar
Sun, R. et al. Between-region genetic divergence reflects the mode and tempo of tumor evolution. Nat. Genet. 49, 1015–1024 (2017).
Article CAS PubMed PubMed Central Google Scholar
Zhai, W. et al. The spatial organization of intra-tumor heterogeneity and evolutionary trajectories of metastases in hepatocellular carcinoma. Nat. Commun. 8, 4565 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Bozic, I. & Wu, C. J. Delineating the evolutionary dynamics of cancer from theory to reality. Nat. Cancer 1, 580–588 (2020).
Article PubMed Google Scholar
West, J., Schenck, R. O., Gatenbee, C., Robertson-Tessi, M. & Anderson, A. R. A. Normal tissue architecture determines the evolutionary course of cancer. Nat. Commun. 12, 2060 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Craig, A. J., von Felden, J., Garcia-Lezana, T., Sarcognato, S. & Villanueva, A. Tumour evolution in hepatocellular carcinoma. Nat. Rev. Gastroenterol. Hepatol. 17, 139–152 (2020).
Article PubMed Google Scholar
Cancer Genome Atlas Research Network. Electronic address, w.b.e. & Cancer Genome Atlas Research, N. Comprehensive and Integrative Genomic Characterization of Hepatocellular Carcinoma. Cell 169, 1327–1341.e23 (2017).
Article Google Scholar
Zhai, W. et al. Dynamic phenotypic heterogeneity and the evolution of multiple RNA subtypes in hepatocellular carcinoma: the PLANET study. Natl. Sci. Rev. 9, nwab192 (2022).
Article CAS PubMed Google Scholar
Slatkin, M. Isolation by distance in equilibrium and non-equilibrium populations. Evolution 47, 264–279 (1993).
Article PubMed Google Scholar
Nielsen, R. et al. Tracing the peopling of the world through genomics. Nature 541, 302–310 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Novembre, J. et al. Genes mirror geography within Europe. Nature 456, 98–101 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Petkova, D., Novembre, J. & Stephens, M. Visualizing spatial population structure with estimated effective migration surfaces. Nat. Genet. 48, 94–100 (2016).
Article CAS PubMed Google Scholar
Geary, R. C. Incorporated Statistician. Vol. 5, p. 115–146 (Wiley, 1954).
Greene, M. T. Alfred Wegener Science, Exploration, And The Theory Of Continental Drift (JHU Press, 2015).
Peng, X. L., Moffitt, R. A., Torphy, R. J., Volmar, K. E. & Yeh, J. J. De novo compartment deconvolution and weight estimation of tumor samples using DECODER. Nat. Commun. 10, 4729 (2019).
Article ADS PubMed PubMed Central Google Scholar
Moffitt, R. A. et al. Virtual microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal adenocarcinoma. Nat. Genet. 47, 1168–1178 (2015).
Article CAS PubMed PubMed Central Google Scholar
Lee, J. S. et al. Classification and prediction of survival in hepatocellular carcinoma by gene expression profiling. Hepatology 40, 667–676 (2004).
Article CAS PubMed Google Scholar
Chiang, D. Y. et al. Focal gains of VEGFA and molecular classification of hepatocellular carcinoma. Cancer Res. 68, 6779–6788 (2008).
Article CAS PubMed PubMed Central Google Scholar
Donald Harper, M. K. Books of Fate and Popular Culture in Early China, 560 (BRILL, 2017).
Kimura, M. & Ohta, T. The age of a neutral mutant persisting in a finite population. Genetics 75, 199–212 (1973).
Article CAS PubMed PubMed Central Google Scholar
Caravagna, G. et al. Subclonal reconstruction of tumors by using machine learning and population genetics. Nat. Genet. 52, 898–907 (2020).
Article CAS PubMed PubMed Central Google Scholar
Marquitti, F. M. D., Fernandes, L. D. & de Aguiar, M. A. M. Allopatry increases the balance of phylogenetic trees during radiation under neutral speciation. Ecography 43, 1487–1498 (2020).
Article ADS Google Scholar
Reiter, J. G. et al. Minimal functional driver gene heterogeneity among untreated metastases. Science 361, 1033–1037 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
Article CAS PubMed PubMed Central Google Scholar
McKenna, A. et al. The Genome Analysis toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010).
Article CAS PubMed PubMed Central Google Scholar
Cibulskis, K. et al. Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat. Biotechnol. 31, 213–219 (2013).
Article CAS PubMed PubMed Central Google Scholar
Ramos, A. H. et al. Oncotator: Cancer Variant Annotation tool. Hum. Mutat. 36, E2423–E2429 (2015).
Article PubMed PubMed Central Google Scholar
Favero, F. et al. Sequenza: allele-specific copy number and mutation profiles from tumor sequencing data. Ann. Oncol. 26, 64–70 (2015).
Article CAS PubMed Google Scholar
Chin, S. F. et al. High-resolution aCGH and expression profiling identifies a novel genomic subtype of ER negative breast cancer. Genome Biol. 8, R215 (2007).
Article PubMed PubMed Central Google Scholar
Alexandrov, L. B. et al. The repertoire of mutational signatures in human cancer. Nature 578, 94–101 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Bergstrom, E. N. et al. SigproFilerMatrixGenerator: a tool for visualizing and exploring patterns of small mutational events. BMC Genomics 20, 685 (2019).
Article PubMed PubMed Central Google Scholar
Kaya, N. A. et al. Genome instability is associated with ethnic differences between Asians and Europeans in hepatocellular carcinoma. Theranostics 12, 4703 (2022).
Article CAS PubMed PubMed Central Google Scholar
Mermel, C. H. et al. GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers. Genome Biol. 12, R41 (2011).
Article PubMed PubMed Central Google Scholar
Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
Article CAS PubMed Google Scholar
Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinform. 12, 323 (2011).
Article CAS Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Article PubMed PubMed Central Google Scholar
Nixon, K. C. The Parsimony ratchet, a new method for Rapid Parsimony Analysis. Cladistics 15, 407–414 (1999).
Article PubMed Google Scholar
Saitou, N. & Nei, M. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4, 406–425 (1987).
CAS PubMed Google Scholar
Roth, A. et al. PyClone: statistical inference of clonal population structure in cancer. Nat. Methods 11, 396 (2014).
Article CAS PubMed PubMed Central Google Scholar
Malikic, S., McPherson, A. W., Donmez, N. & Sahinalp, C. S. Clonality inference in multiple tumor samples using phylogeny. Bioinformatics 31, 1349–1356 (2015).
Article CAS PubMed Google Scholar
Caliński, T. & Harabasz, J. A dendrite method for cluster analysis. Commun. Stat. 3, 1–27 (1974).
MathSciNet Google Scholar
Studer, M., Ritschard, G., Gabadinho, A. & Müller, N. S. Discrepancy analysis of state sequences. Sociol. Methods Res. 40, 471–510 (2011).
Article MathSciNet Google Scholar
Boyault, S. et al. Transcriptome classification of HCC is related to gene alterations and to new therapeutic targets. Hepatology 45, 42–52 (2007).
Article CAS PubMed Google Scholar
Hoshida, Y. et al. Integrative transcriptome analysis reveals common molecular subclasses of human hepatocellular carcinoma. Cancer Res. 69, 7385–7392 (2009).
Article CAS PubMed PubMed Central Google Scholar
Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 (2015).
Article CAS PubMed PubMed Central Google Scholar
Yoshihara, K. et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat. Commun. 4, 2612 (2013).
Newman, A. M. et al. Robust enumeration of cell subsets from tissue expression profiles. Nat. Methods 12, 453–45 (2015).
Article CAS PubMed PubMed Central Google Scholar
Jimenez-Sanchez, A., Cast, O. & Miller, M. L. Comprehensive benchmarking and integration of tumor microenvironment cell estimation methods. Cancer Res. 79, 6238–6246 (2019).
Article CAS PubMed Google Scholar
Bortolussi, N., Durand, E., Blum, M. & Francois, O. apTreeshape: statistical analysis of phylogenetic tree shape. Bioinformatics 22, 363–364 (2006).
Article CAS PubMed Google Scholar
Jiang, P. et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat. Med. 24, 1550 (2018).
Article CAS PubMed PubMed Central Google Scholar
Ayers, M. et al. IFN-gamma-related mRNA profile predicts clinical response to PD-1 blockade. J. Clin. Invest. 127, 2930–2940 (2017).
Article PubMed PubMed Central Google Scholar
Auslander, N. et al. Robust prediction of response to immune checkpoint blockade therapy in metastatic melanoma. Nat. Med. 24, 1545–1549 (2018).
Article CAS PubMed PubMed Central Google Scholar
Rooney, M. S., Shukla, S. A., Wu, C. J., Getz, G. & Hacohen, N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell 160, 48–61 (2015).
Article CAS PubMed PubMed Central Google Scholar
Sia, D. et al. Identification of an immune-specific class of hepatocellular carcinoma, based on molecular features. Gastroenterology 153, 812–826 (2017).
Article CAS PubMed Google Scholar
Sangro, B. et al. Association of inflammatory biomarkers with clinical outcomes in nivolumab-treated patients with advanced hepatocellular carcinoma. J. Hepatol. 73, 1460–1469 (2020).
Article CAS PubMed PubMed Central Google Scholar
Villanueva, A. et al. Combining clinical, pathology, and gene expression data to predict recurrence of hepatocellular carcinoma. Gastroenterology 140, 1501–1512.e2 (2011).
Article CAS PubMed Google Scholar
Roessler, S. et al. A unique metastasis gene signature enables prediction of tumor relapse in early-stage hepatocellular carcinoma patients. Cancer Res. 70, 10202–10212 (2010).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We would like to thank Chung-I Wu, Rasmus Nielsen, Xionglei He, Miao Xu, Yanhua Qu and Zhengting Zou, Jianbin Chen for their constructive comments. This work is funded by the National Science Foundation of China (grant no. 32293192/32293190, 31970566, 92259303 and 32000407). L.M. is supported by the National Key R&D Program of China (2019YFA0709501) and the National Natural Science Foundation of China (11971459). W.Z. is supported in part by the Strategic Priority Research Program of the Chinese Academy of Sciences XDPB17 (XDPB17) and the National Key R&D Program of China (grant 2018YFC1406902 and 2018YFC0910400).

Author information

These authors contributed equally: Xiaodong Liu, Ke Zhang.
These authors jointly supervised this work: Liang Ma, Li Jiang, Weiwei Zhai.

Authors and Affiliations

Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, Chinese Academy of Sciences, Beijing, China
Xiaodong Liu, Dafei Wu, Tingting Chen, Zhiyuan Liu, Sinan Zhu, Liang Ma & Weiwei Zhai
School of Life Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, China
Xiaodong Liu
Department of General Surgery, Beijing Ditan Hospital, Capital Medical University, No. 8, Jingshun East Street, Chaoyang District, Beijing, P.R. China
Ke Zhang, Zhe Jia & Li Jiang
Genome Institute of Singapore, Agency for Science, Technology and Research, Singapore, Singapore
Neslihan A. Kaya & Torsten Wuestefeld
University of the Chinese Academy of Sciences, Beijing, China
Tingting Chen & Zhiyuan Liu
Centre for Quantitative Medicine, Program in Health Services and Systems Research, Duke-NUS Medical School, Singapore, Singapore
Sinan Zhu & Jin Liu
Institute of Pathology, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany
Axel M. Hillmer
School of Biological Sciences, Nanyang Technological University, Singapore, Singapore
Torsten Wuestefeld
Bioland Laboratory (Guangzhou Regenerative Medicine and Health Guangdong Laboratory), Guangzhou, China
Yun Shen Chan
CAS Key Laboratory of Quantitative Engineering Biology, Shenzhen Institute of Synthetic Biology, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China
Zheng Hu
Center for Excellence in Animal Evolution and Genetics, Chinese Academy of Sciences, Kunming, China
Weiwei Zhai

Authors

Xiaodong Liu
View author publications
You can also search for this author in PubMed Google Scholar
Ke Zhang
View author publications
You can also search for this author in PubMed Google Scholar
Neslihan A. Kaya
View author publications
You can also search for this author in PubMed Google Scholar
Zhe Jia
View author publications
You can also search for this author in PubMed Google Scholar
Dafei Wu
View author publications
You can also search for this author in PubMed Google Scholar
Tingting Chen
View author publications
You can also search for this author in PubMed Google Scholar
Zhiyuan Liu
View author publications
You can also search for this author in PubMed Google Scholar
Sinan Zhu
View author publications
You can also search for this author in PubMed Google Scholar
Axel M. Hillmer
View author publications
You can also search for this author in PubMed Google Scholar
Torsten Wuestefeld
View author publications
You can also search for this author in PubMed Google Scholar
Jin Liu
View author publications
You can also search for this author in PubMed Google Scholar
Yun Shen Chan
View author publications
You can also search for this author in PubMed Google Scholar
Zheng Hu
View author publications
You can also search for this author in PubMed Google Scholar
Liang Ma
View author publications
You can also search for this author in PubMed Google Scholar
Li Jiang
View author publications
You can also search for this author in PubMed Google Scholar
Weiwei Zhai
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

W.Z., L.J. and L.M. co-lead the project. X.L., K.Z., Z.J., D.W., T.C. and L.J. collected the patient samples. X.L., N.A.K., S.Z. and L.M. performed the analysis. X.L. and W.Z. wrote the original manuscript with inputs from K.Z., N.A.K., Z.J., D.W., T.C., Z.L., S.Z., A.M.H., T.W., J.L., Y.S.C., Z.H., L.M. and L.J.

Corresponding authors

Correspondence to Liang Ma, Li Jiang or Weiwei Zhai.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Jacob Househam and the other anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Reporting Summary

Source data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Liu, X., Zhang, K., Kaya, N.A. et al. Tumor phylogeography reveals block-shaped spatial heterogeneity and the mode of evolution in Hepatocellular Carcinoma. Nat Commun 15, 3169 (2024). https://doi.org/10.1038/s41467-024-47541-9

Download citation

Received: 04 August 2022
Accepted: 26 March 2024
Published: 12 April 2024
DOI: https://doi.org/10.1038/s41467-024-47541-9

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.