Keywords
HCC - genomics - single cell - heterogeneity - liver cancer
Hepatocellular carcinoma (HCC) is a highly lethal malignancy and is a leading cause
of cancer mortality around the world, causing around 700,000 deaths annually.[1] In the U.S., overall cancer-related mortality dropped by an impressive 29% between
1991 and 2017.[2] While the 5-year survival rate for all cancers, when combined, improves to 67%,
patients with HCC continue to have a dismal 5-year survival of around 18%.[2] Moreover, despite advances in locoregional therapies, targeted therapies, and immunotherapy,
the median survival for advanced HCC remains less than 2 years. One of the factors
contributing to the persistent poor survival rates for HCC is the lack of adequate
response to conventional therapies.
After the initial promising approval of tyrosine kinase inhibitor sorafenib for advanced
HCC in 2007, several clinical trials for HCC evaluating other kinase inhibitors like
sunitinib and brivanib failed over the past decade.[3]
[4] We have recently seen the additional approval of lenvatinib as first-line therapy,
and regorafenib, cabozantinib, and ramucirumab as second-line therapies.[5] Promising results from a recent phase III clinical trial of combination of immunotherapy
(atezolizumab) with bevacizumab indicate this combination will now finally displace
sorafenib as first-line therapy for advanced HCC.[3] Despite these results being considered a major breakthrough for HCC, we have to
acknowledge that the combination therapy only marginally improves progression-free
survival by around 2 months when compared with sorafenib. Clearly, there is an urgent
need to develop better therapies for HCC and in order to do that we first have to
understand the reasons for therapy resistance. One of the crucial factors which poses
a major challenge not just for developing new therapies but also for identifying novel
biomarkers or for predicting response to therapy, is tumor heterogeneity. Hence, deciphering
tumor heterogeneity is pivotal to improve clinical outcomes for patients with HCC.
Genomic Landscape of HCC
Next-generation sequencing (NGS) technologies have steadily evolved over the past
15 years allowing us to sequence an exponential number of biological samples, at progressively
lower costs. For instance, the cost of sequencing the human genome has fallen from
$10,000 per megabase of deoxyribonucleic acid (DNA) in 2001, to less than 1 cent per
megabase in 2019.[6] In the pre-NGS era, classical experimental schemes were generally hypothesis-based,
where investigations with appropriate controls were designed to either confirm or
refute a hypothesis. This approach, while incrementally fruitful, led to a myopic
view of biological complexity, since hypotheses are generally limited by the need
for a priori knowledge. The application of nonhypothesis-driven approaches using NGS-based
technologies to sequence whole cancer genomes has raised the veil on the vast underlying
complexity of genetic and epigenetic alterations in cancer cells, thus ushering in
an era of rapid new discoveries.
A classic example of NGS-based discovery is the identification of the TERT promoter mutation. Telomerases are enzymes which maintain telomere length of chromosomes
during rapid cell division and were known to be upregulated in cancers, endowing cancer
cells with immortality.[7] But the mechanisms by which cancer cells increased telomerase expression were not
fully understood. NGS-based whole genome analysis of familial malignant melanoma in
2013 revealed, for the first time, recurrent mutations in the TERT gene promoter, which was associated with higher TERT messenger ribonucleic acid (RNA) levels.[8] Subsequently, this TERT promoter mutation has been found to be one of the most common
genetic events in multiple cancers including HCC.[9] Several large-scale studies have revealed the genomic landscape of HCC and have
identified major somatic mutations, copy number variations (CNVs), and transcriptomic
signatures of HCC.[9]
[10]
[11]
[12]
[13] Few of the most common somatic events in HCC include mutations in the TERT promoter, TP53, CTNNB1, AXIN1, or ARID1A genes and most common CNV involve MYC, MET, CCND1, VEGFA, and FGF19 genes. Mutational signature patterns have also helped in identifying etiologic agents
which drive cancer initiation, like signatures associated with tobacco, aflatoxin,
or aristolochic acid exposure.[9]
[14] Thus, the application of NGS-based sequencing has rapidly led to expansion of our
knowledge of cancer biology thus opening the door to further questions and research.
But it has to be noted that the aforementioned NGS studies have identified genetic
events mostly by sequencing bulk tumors and hence do not capture the heterogeneity
in HCC.
Cancer Heterogeneity in Hepatocellular Carcinoma
Cancer Heterogeneity in Hepatocellular Carcinoma
HCC is known to exhibit exceptional heterogeneity compared with other tumors. Tumor
heterogeneity can mean different things in different contexts—(1) etiologic – HCC
can be caused by varied risk factors ranging from viral hepatitis to nonalcoholic
fatty liver disease to toxins, (2) geographic heterogeneity – HCC has different incidence
and outcomes in different regions of the world, or (3) molecular heterogeneity. In
this review, we will primarily be discussing molecular heterogeneity, which refers
to the variations in genetic events, gene expression patterns, activated pathways,
immune infiltrates, or stromal changes across tumors. Molecular heterogeneity can
be further classified into—(1) interpatient – molecular differences between tumors
of different patients, (2) intertumor – refers to variations between different tumors
in the same patient, or (3) intratumor heterogeneity refers to variations across different
regions of an individual tumor in the same patient ([Fig. 1]). One of the main reasons for the significant tumor heterogeneity seen in HCCs is
that they generally arise in the background of a highly inflamed and fibrotic procarcinogenic
microenvironment created by cirrhosis, which exerts a field effect throughout the
liver. Hence, synchronous and metachronous tumor initiating events occur concurrently
all over the liver, leading to significant inter- and intratumor heterogeneity.
Fig. 1 Types and sources of tumor heterogeneity in hepatocellular carcinoma (HCC). The top
panel present the different types of heterogeneity in HCC. (A) Interpatient – molecular differences between tumors of different patients, (B) intertumor – variations between different tumors in the same patient, and (C) intratumor heterogeneity refers to variations across different regions of an individual
tumor in the same patient. The bottom panel lists the different sources of heterogeneity
in HCC which can result either from – (D) cancer-cell intrinsic variations, (E) differences in tumor immune cell infiltrates, and (F) changes in tumor stroma.
Early studies elegantly uncovered HCC heterogeneity by sampling multiple nodules from
individual patients and studying them using platforms such as immunohistochemistry
(IHC), Sanger sequencing, Southern blot hybridization, or DNA fingerprinting.[15]
[16]
[17] However, recent advances in genomics and single-cell sequencing have opened the
floodgates of cancer research and have provided a rare window into the molecular heterogeneity
in HCC. In this review, we will discuss how multi-omic and single-cell sequencing
technologies have enabled us to navigate and understand tumor heterogeneity in HCC.
Multi-Omic Approaches to Tumor Heterogeneity in HCC
Multi-Omic Approaches to Tumor Heterogeneity in HCC
NGS-based multi-omic analyses have been particularly important in studying tumor heterogeneity.
Understanding and quantifying tumor heterogeneity is crucial since it has several
translational implications for predicting recurrence and response to therapy. Tumors
with high levels of heterogeneity generally do not respond well to therapies since
the selective pressure introduced by the therapy leads to expansion of resistant subclones
or to the emergence of new drug-tolerant clones. Most of the genomic studies have
analyzed sequencing data obtained from bulk tumor tissue leading to limited understanding
of the tumor heterogeneity. Novel platforms such as single-cell sequencing, mass cytometry,
and spatial transcriptomics are now enabling us to decode the complexity of the tumor
microenvironment at a very high resolution. Using these platforms along with serial
sampling and multiregion tumor sampling is allowing us to develop a deeper understanding
of the temporal and spatial heterogeneity of HCC ([Fig. 2]).
Fig. 2 Multi-omic approaches to study tumor heterogeneity in hepatocellular carcinoma (HCC).
Multiregion tumor sampling or serial liquid biopsy sampling can be used to study tumor
heterogeneity. Next-generation sequencing (NGS)-based sequencing technology, single-cell
sequencing, mass cytometry, mass spectrometry, or other high-resolution methods can
be used to study the extracts from multiregion sampling and dissociated single cells
decipher tumor heterogeneity. This knowledge can then be used to stratify patients
and tumors or to predict clinical outcome.
A malignant tumor is not just a mass of rapidly proliferating tumor cells but is a
highly complex quasi-organ which is infiltrated by a plethora of host immune cells,
and stromal cells, is supported by an extracellular matrix and also entwined by blood
vessels which provide oxygen and nutrients.[18] The heterogeneity present in tumors does not necessarily arise solely from the cancer
cells themselves but also from variations in the immune cells, stromal cells, and
other components of the microenvironment (TME) ([Fig. 1]). By extracting and studying single cells from tumors, we are now able to independently
understand the heterogeneity in these discrete compartments of the TME. We will discuss
below the studies exploring heterogeneity either in the cancer cells or in the immune
cell compartment.
Multi-Omic Analysis of Cancer Cell Heterogeneity in HCC
Multi-Omic Analysis of Cancer Cell Heterogeneity in HCC
Cancer cells exhibit high degree of genomic instability and are hence the main source
of heterogeneity in a tumor, we will address them first. To study heterogeneity using
bulk tumor sample sequencing, investigators generally sample multiple discrete regions
from the same tumor or multiple regions from different tumors in the same patient
and compare them using this multiregion targeting strategy.
For instance, Ling et al used laser dissection microscopy to perform honeycomb-like
precise microdissections of a single 35-mm HCC tumor and extracted DNA from 286 separate
regions for genotyping, and performed whole exome sequencing of 23 of these dissected
regions. Each sample contained approximately 20,000 cells.[19] Surprisingly, they found extreme genetic diversity in this tumor with more than
100 million estimated coding region mutations, and 6 to 7 major clones. The 23 exome
sequenced samples within this tumor displayed 20 unique cell subclones. Authors analyzed
these data by modern population genetic theory and concluded that the extreme genetic
diversity implies evolution under the non-Darwinian mode. This theory possibly explains
the high degree of heterogeneity noted in HCC. Moreover, we can easily see how the
prevalence of genetic driver events can be underestimated when we sample and sequence
only one site per tumor for each patient.
In another interesting study, Shi et al performed multiregional whole exome sequencing
of two synchronous HCCs and one cholangiocarcinoma nodule from the same patient and
found that the three primary lesions showed almost no overlapping mutations or CNVs
underscoring intertumor heterogeneity in this patient.[20] Moreover, this patient developed postoperative recurrence and sampling the two recurrent
nodules showed that their mutational profile was concordant with just one of the primary
HCC nodules, not the other two, indicating their origin. This patient had underlying
hepatitis B virus (HBV)-related cirrhosis possibly explaining the emergence of these
synchronous but disparate primary liver cancer lesions. HCC induced by HBV can show
high degrees of intertumor heterogeneity since HBV viral integration into the host
genome is an added driver of carcinogenesis in this subset of HCCs. Xue et al studied
43 separate lesions from 10 patients with multifocal HBV-HCC using exome sequencing.[21] They found that the degree of intratumoral heterogeneity varied widely between patients
with the proportion of ubiquitous mutations ranging from 8 to 97% of all mutations.
While satellite nodules near the main nodule appeared to mostly have shared genetic
events with the primary, metastatic nodules or even tumor thrombus appeared to have
distinct genetic events providing evidence for clonal evolution during tumor progression.
On construction of phylogenetic trees, they found that HBV integrations in driver
genes like MLL4 or TERT occurred both in the trunks and branches of the tree. These results suggest that
viral integrations may occur either early or late during tumor evolution and play
an important role in driving tumor heterogeneity in HCC.
While most studies have explored the exome sequencing profile or the transcriptome
to study tumor heterogeneity, Buczak et al evaluated heterogeneity using large-scale
proteomics.[22] This approach is valuable since analyzing the proteomic data enables us to gain
functional insights into the heterogeneity introduced by the genetic events. Buczak
et al used a mass spectrometry-based proteomic approach to explore inter- and intratumor
spatial heterogeneity in five patients with HCC.[22] Interestingly, authors also explored spatial heterogeneity in HCC by carefully comparing
the proteomic profile of tumor cells microdissected from the center of the tumor versus
those from the periphery, rather than performing random sampling. The small number
of samples did not allow them to make robust conclusions, but the study did demonstrate
prominent interpatient variability in the differential proteomic expression between
the cells at the center of the tumor and those at the periphery. For example, mitochondrial
metabolism appeared to be downregulated in the periphery in multiple samples, compared
with the center. This study illustrates the added complexity of tumor cell spatial
location on functional heterogeneity.
Multi-Omic Analysis of Immune Cell Heterogeneity in HCC
Multi-Omic Analysis of Immune Cell Heterogeneity in HCC
Tumor heterogeneity can arise from the infiltrating host immune and stromal cells
and not just from the cancer cells themselves. With recent promising results for immunotherapy
in HCC, there is increasing interest in understanding the specific immune subsets
present in the tumor microenvironment and how they interact with the cancer cells.
Classic approaches like multiplex IHC and flow cytometry have previously revealed
immune cell heterogeneity and helped identify specific subtypes based on immune infiltrates.
For instance, Kurebayashi et al used multiplex immune cell IHC to evaluate 919 regions
from 158 resected HCC specimens and were able to classify the HCCs into—immune-high,
immune-mid, or immune-low subtypes, with the immune-high subtype having a generally
better prognosis.[23] More recently technologies such as mass cytometry, codetection by indexing, and
T cell receptor (TCR) sequencing are allowing us to directly interrogate the tumor
immune microenvironment more comprehensively.[24]
[25]
[26]
Combining multi-omic sequencing with immune profiling is a powerful strategy to gain
a more global overview into tumor complexity. Losic et al pursued this strategy by
integrating DNA-, RNA-, and TCR-sequencing from 71 spatially discrete regions from
the same tumor-derived 14 human liver cancer specimens to investigate intratumoral
heterogeneity.[27] They found significant heterogeneity not just in tumor-specific antigen expression
in different regions of a tumor, but also in immune infiltrate burden and immune cell
clonality. They also found that the regional clonal immune response contributes significantly
to intratumoral heterogeneity in HCC. To understand this data at a higher resolution,
they followed this up by performing single-cell sequencing of around 20,000 cells
from two geographically distant regions and confirmed significant heterogeneity in
the transcriptional pathways activated in distant regions within the same nodule.[27]
In another study which used similar multi-omic approaches, the investigators performed
DNA-sequencing, RNA-sequencing, mass spectrometry-based proteomics, and also mass
cytometry for immune profiling of 42 discrete samples from eight HCC patients.[28] They found significant intralesional heterogeneity in tumor cells at all levels
ranging from the genome to the proteome. Compared with the cancer cells, the immune
infiltrates in HCC appeared to be less heterogeneous and they were able to cluster
all tumors into three groups based on the immune profile—subtype 1 with competent
immune response and good prognosis, subtype 2 were “cold” tumors with minimal immune
infiltrates, and subtype 3 with higher immunosuppressive infiltrates and exhausted
T cells.[28] These data suggest that studying the tumor immune microenvironment might be a more
tractable approach to predict response to immunotherapy than sequencing just the tumor
cells.
Another interesting feature of multifocal HCC is the coexistence of intrahepatic metastasis,
which are derived from vascular spread of larger primary tumor, and multicentric primary
HCC where the lesions are not related to each other. Dong et al evaluated distinctions
between multicentric HCCs and intrahepatic metastasis by combining data from DNA-sequencing,
RNA-sequencing, multiplex immunostaining, immunopeptidomics, and TCR sequencing of
47 tumors from 15 patients with multifocal HCC.[29] Authors described significant spatiotemporal differences in the transcriptome and
immune profiles between intrahepatic metastasis and multicentric HCCs. They found
that metastatic lesions shared neoantigens and TCR repertoires with the primary lesion
and had lower T cell infiltration along with higher M2-like macrophage infiltration.
However, multicentric HCCs were characterized by higher expression of immune checkpoints
and also higher rates of immune editing. These spatiotemporal changes imply that response
to immunotherapy will differ between those with intrahepatic metastasis versus multicentric
HCC. Moreover, they were able to integrate the data from the sequencing and immune
profiling platforms to develop a classifier which could predict recurrence. Thus,
large-scale multi-omic data from these studies clearly reveal the vast complexity
and heterogeneity in HCC and raise the concern that single-site biopsy from individual
tumors might not be enough to risk stratify patients or to make therapeutic decisions.
Tumor Heterogeneity from Other Stromal Cells
Tumor Heterogeneity from Other Stromal Cells
Apart from immune cells, other stromal cells like cancer-associated fibroblasts (CAFs)
and endothelial cells also contribute to tumor heterogeneity in HCC. CAFs represent
a mixed population of cells which can be derived from local fibroblasts, mesenchymal
stem cells, vascular smooth muscle cells, or even from cancer cells.[30] Multi-omic and single-cell technologies have been successfully used to profile CAFs
from breast and lung cancer and they have identified specific functional subtypes
of CAF.[31]
[32] In general, there appear to be two major phenotypes, a contractile phenotype which
contributes to the tumor matrix and an inflammatory phenotype which modulates the
local immune response via its secretome.[33] Vascular endothelial cells are another source of intratumoral heterogeneity.[34] Ma et al recently used single-cell RNAseq from 19 liver cancer samples and found
that the transcriptional states of nonmalignant cells such as tumor endothelial cells
and CAFs were reprogrammed in tumors with higher heterogeneity and the gene signature
derived from these stromal cells, could be used to discriminate heterogeneity in HCCs.[35]
Single-Cell Sequencing Approaches to Understand Tumor Heterogeneity in HCC
Single-Cell Sequencing Approaches to Understand Tumor Heterogeneity in HCC
Most of the NGS-based multi-omic techniques described above like whole exome sequencing
or RNA-sequencing typically extract DNA or RNA by lysing bulk tissue. Hence, the data
represents the average mutation rate or average gene expression across millions of
cells of heterogeneous origin. While this approach has provided valuable insights
into heterogeneity, it does obscure the differences in expression between the different
kinds of cells thus leading to an underestimation of the biological complexity of
cancer. Single-cell sequencing is an approach where bulk tumors are dissociated to
isolate single cells, followed by DNA or RNA extraction from these cells which are
then sequenced individually.[36] Advances in microfluidic technology and molecular barcoding have made the profiling
of tens of thousands of individual cells not only technically feasible but also cost-effective.
Investigators are now increasingly using single-cell sequencing to tractably perform
large-scale analysis at a single-cell level in multiple cancers, at an unprecedented
resolution. These single-cell sequencing technologies are revolutionizing our ability
to understand cancer and opening a window into cancer heterogeneity. Single-cell DNA
sequencing enables us to define the mutation profile of individual subclones within
tumors and to understand their evolutionary trajectory. Meanwhile, single-cell RNA
sequencing allows us to evaluate the transcriptional profile of individual cancer
cells in comparison to the infiltrating stromal cells or host immune cells. Other
technologies like single-cell DNA methylome sequencing and single-cell transposase-accessible
chromatin with sequencing (ATACseq) promise to reveal novel insights into the epigenetic
landscape of tumors at a single-cell resolution. We will now highlight the major studies
employing these single-cell technologies to study heterogeneity in the cancer cells
and host immune cells in HCC.
Single-Cell Genomics of Cancer Cells in HCC
Single-Cell Genomics of Cancer Cells in HCC
One of the exciting aspects of singlulomics is the opportunity to study how genetic
or epigenetic driver events impact gene expression in an individual cell, thus establishing
the functional relevance of the genetic event to cancer progression. Hou et al pursued
this by exploring the genome, transcriptome, and methylome of each individual cell
using scTrio-seq of 25 single cancer cells derived from a patient with HCC.[37] Upon hierarchical clustering of these 25 cells they found two subtypes one with
copy gains in chromosome 8, 11, and 20; while the second subpopulation showed loss
of chromosomes 4 and 16. They were able to demonstrate that these large-scale CNVs
were associated with corresponding gene expression changes, while these CNVs did not
correspond with the methylomic changes. Interestingly, they found that clustering
based on CNV or methylome sequencing classified the cells into the exact same two
clusters. By overlapping these clusters with the gene expression changes, they showed
that cells in the first subpopulation with higher CNV were less responsive to immune
recognition as they had downregulation of multiple genes in the acute inflammatory
response and complement activation pathways. Thus, the authors used the information
from the triple-omics analysis to infer specific biological properties of individual
subclones within a tumor at a single-cell level, and demonstrate how these findings
may have translational implications in predicting response to immunotherapy.
The study of cancer stem cells (CSCs) has been an area of controversial research and
several investigators have identified divergent populations of liver CSCs based on
expression of surface markers such as CD44, CD90, CD133, or EpCAM.[38] Single-cell sequencing is an ideal platform to explore this further, since we are
now able to evaluate the expression of a multitude of stem cell markers on each individual
cell.[39]
[40] Zheng et al used flow cytometry to sort and isolate CSC subpopulations (CD133 +/CD24 +/EpCAM + )
from two HCC cell lines and 118 cells derived from a single patient with HCC who had
undergone resection.[40] Single-cell whole transcriptome sequencing of these cells revealed significant phenotypic
and functional heterogeneity in self-renewal capacity and differentiation depending
on individual surface marker expression. The transcriptomic heterogeneity and functional
diversity observed in the diverse CSC population highlights the potential origin of
cancer cell heterogeneity.
One of the direct mechanisms by which HBV causes HCC is via integration of the HBV
genome into the host genome eventually leading to molecular transformation of benign
hepatocytes into cancer cells.[41] But several questions remain unanswered, specifically, the frequency of integration
essential for hepatocarcinogenesis, the role of spatial heterogeneity in viral integration
sites within the liver, and the timing of viral integration during tumor evolution.
Duan et al performed single-cell whole genome sequencing to profile 96 cancer cells
and 15 normal liver cells collected from 3 patients with hepatitis B-driven HCC.[42] They showed that all the single cells in a specific HCC clone exhibited the same
HBV integration indicating these integrations are early genetic drivers of cancer
initiation and remain stable during tumor evolution. Also, tumors with multinodular
morphology were typically polyclonal and displayed a high degree of heterogeneity.
Thus, this study highlights how different sites of HBV integrations in the host liver
cell genome can account for the clonal diversity of multifocal HBV-HCC.
Single-Cell Genomics to Study Immune Cells in HCC
Single-Cell Genomics to Study Immune Cells in HCC
Another major advantage of single-cell technology is the ability to analyze the expression
patterns in host immune cells infiltrating the tumor. Zhang et al combined two single-cell
technologies to study > 75,000 single CD45+ immune cells infiltrating HCC derived
from 16 patients with HCC, and compared it to immune cells in the adjacent normal
liver, lymph node, blood, or ascitic fluid.[43] Using single-cell sequencing Zhang et al identified two distinct states of tumor-associated
macrophages (TAMs) in human HCC—one with a myeloid-derived suppressor cell-like profile
and another with a M2-like TAM profile. Further, they observed that the macrophages
with M2-like profile showed expression of genes such as SLC40A1 encoding ferroportin, suggesting that iron metabolism is potentially involved in
macrophage polarization in tumors. Next, they used computational lineage tracing technology
to reveal potential migration patterns of immune cells from the tumor to distant sites
like lymph nodes or ascitic fluid. For instance, they identified a subset of dendritic
cells which were LAMP3+ and appeared to be home to hepatic lymph nodes from the tumor
and crosstalk with lymphocytes in the TME leading to dysfunctional T cells. This analysis
of immune cells from multiple sites provides valuable insight into the dynamic modification
of immune cell profiles during tumor evolution and highlights that cancer is not just
a local phenomenon but is a systemic disease.
Some investigators have taken the approach to focus primarily on systematically interrogating
tumor infiltrating immune cells using single-cell sequencing. Zheng et al performed
single-cell RNA sequencing of around 5,000 T cells isolated from the tumor, adjacent
normal liver, and the peripheral blood of six patients with HCC.[44] The high-resolution data enabled the authors to identify 11 distinct T cell subtypes
in HCC. Exhausted CD8+ T cells were the largest cluster of tumor infiltrating CD + 8
T cells and TCR analysis revealed these exhausted T cells had likely evolved from
other CD8+ T cell subtypes during tumor evolution. Another major T cell subset of
interest was the HCC-specific CCD8 +/LAYN+ regulatory T cells (Tregs) that were clonally
expanded and enriched in the tumor when compared with the nontumorous liver. These
Tregs are known to have an immunosuppressive function and their enrichment could thus
represent a mechanism by which cancer cells evade tumor-specific immune responses.
Thus, the authors were able to use single-cell sequencing to reveal novel subtypes
of immune cells in HCC and infer their developmental trajectory.
Moreover, these data serve as a valuable resource and their analysis can have clinical
implications in understanding resistance to immunotherapy. Wang et al reanalyzed the
single-cell sequencing data generated by Zheng et al along with independent T cell
flow cytometry data from patients with chronic hepatitis B or HCC, to identify mechanisms
for T cell exhaustion. They found that the mechanisms of T cell exhaustion in chronic
hepatitis B without HCC to be distinct from mechanisms responsible for T cell exhaustion
in HCC, since the T cells from these conditions clustered into distinct T cell exhaustion
modules.[44]
[45]
Studies using single-cell analysis have not only provided mechanistic insights into
drug resistance in HCC but also help predict clinical outcomes. By analyzing single-cell
RNA-seq profiles of 19 primary HCC and intrahepatic cholangiocarcinoma patients, it
was found that vascular endothelial growth factor (VEGF) signaling in tumors promotes
an immunosuppressive microenvironment, thus providing a rationale for combination
therapy of immune checkpoint inhibitors with VEGF inhibitors in HCC.[35] Moreover, they were able to link the degree of intratumoral heterogeneity to patient
prognosis, as tumors with higher transcriptomic diversity were associated with patient's
worse overall survival. Thus, ongoing studies in single-cell sequencing show significant
promise to reveal translationally relevant insights into tumor heterogeneity and therapeutic
resistance in HCC.
Using Liquid Biopsy to Study Tumor Heterogeneity
Using Liquid Biopsy to Study Tumor Heterogeneity
The analysis of tumors using biomarkers such as circulating tumor cells (CTCs) or
cell-free DNA (cfDNA) in peripheral blood has been coined “liquid biopsy” since we
are now able to sample tumors noninvasively.[46] The rapid developments in the field of liquid biopsy research have ushered in a
promising phase in biomarker research and liquid biopsy has already shown potential
to help us decipher tumor heterogeneity in HCC.[47]
[48] During tumor clonal evolution, cancer cells sequentially acquire genetic events
throughout the different steps in the metastatic cascade namely invasion, dissemination,
and colonization. Spatial heterogeneity between cancer cells inside the tumor versus
those in the circulation and those at metastatic sites has been an area of great interest.
To study this, Sun et al collected blood either from peripheral vein, peripheral artery,
hepatic veins, infrahepatic inferior vena cava, and portal vein before tumor resection
in 73 patients with HCC.[49] Authors found that CTCs were predominantly epithelial at vascular sites proximal
to the tumor, but transformed to a mesenchymal-like phenotype during dissemination
in the circulation. Tumors that exhibited a high degree of epithelial-mesenchymal
transition activity were noted to have a higher number of total CTCs in the hepatic
veins. Moreover, CTC and circulating tumor burden were observed to predict postoperative
lung metastasis and intrahepatic recurrence, respectively. This study highlights the
value of serial and spatial sampling of the peripheral blood in understanding tumor
heterogeneity and predicting clinical outcomes in HCC.
cfDNAs are very encouraging biomarkers which have the potential to circumvent existing
challenges in tumor sampling. Since cfDNA potentially will contain DNA shed from all
the major tumor nodules, they can help identify, and serially follow, the major genomic
events across multiple tumor regions without the need for repeated biopsies. Huang
et al evaluated the value of cfDNA in tracking tumor heterogeneity.[50] They performed whole exome sequencing and targeted deep sequencing of 32 multiregional
tumor samples along with matched preoperative cfDNA sequencing from five patients
with HCC. Authors found that complementing targeted deep sequencing of a single tumor
specimen with cfDNA sequencing enabled them to identify genetic variation that could
be targeted by Food and Drug Administration (FDA)-approved drugs in 37.1% patients.
This data, though preliminary, provides evidence for the utility of cfDNA analysis
in evaluating tumor heterogeneity.
Experimental Systems to Model Tumor Heterogeneity
Experimental Systems to Model Tumor Heterogeneity
Our current understanding of tumor heterogeneity is limited by the traditional experimental
model systems such as immortalized cell lines or transgenic mouse models. We need
to develop in vitro and in vivo novel model systems which can capture the complexity
of human HCCs. Patient-derived primary cell lines, organoids derived from human HCC
samples, and patient-derived xenografts (PDXs) are promising model systems which can
be employed to study heterogeneity ([Fig. 3]). Tumor samples either from resection or biopsy can be used to establish patient-derived
primary cancer cells which can be maintained in culture for extended periods of time
and can be used for therapeutic drug screening.[51] This strategy can be successfully combined with multiregion sampling to model tumor
heterogeneity in the dish. Gao et al sampled 55 regions from 10 patients with HCC
undergoing curative resection, and isolated primary cancer cells.[52] They maintained these cells in low-passage culture and performed whole-exome sequencing,
copy number analysis, and high-throughput drug screening. They found that cells from
four discrete tumor regions containing genetic alterations like FGF19, DDR2, PDGFRA, and TOP1 were sensitive to corresponding targeted therapeutic agents, thus establishing proof-of-principle
that such in vitro approaches can be used to study tumor heterogeneity in HCC.
Fig. 3 Experimental model systems to study tumor heterogeneity in hepatocellular carcinoma
(HCC). Multiregion tumor targeting can be coupled with tumor dissociation followed
by primary cell culture (A) or organoid culture (B) in vitro to develop patient-derived primary cells or patient-derived organoids.
These dissociated single cells can also be grown as xenografts in immunocompromised
mice to establish patient-derived xenografts (C). All these experimental model systems can be used to interrogate tumor heterogeneity
and to perform therapeutic drug screens.
Organoids are preclinical models where cancer cells are cultured in vitro as three-dimensional
structures.[53] Developing organoids from multiregion sampling is a good experimental approach to
study tumor heterogeneity. Li et al used this approach and generated 27 liver cancer
organoid cell lines by sampling distinct regions of the tumor.[54] They tested a library of 129 cancer drugs using these organoid cell lines and were
able to identify a subset of drugs that appeared pan-effective, thus demonstrating
the utility of cancer organoid drug testing in drug discovery pipelines. PDXs are
established by growing human HCC xenografts in immunocompromised mice. PDX models
have already been successfully used to study tumor progression, identify novel biomarkers,
and to perform preclinical personalized drug screens.[55]
[56]
[57] Using multiregion sampling strategies to develop PDXs can enable investigators to
model intratumor heterogeneity in vivo. While such approaches have been successful
in other cancers in combination with other technologies like exome sequencing,[58]
[59] establishing PDXs in HCC is challenging since the rate of engraftment is traditionally
considered to be low.[55]
Challenges and Future Directions
Challenges and Future Directions
The advances in NGS-based multi-omic sequencing and single-cell sequencing have brought
tremendous excitement and are finally enabling us to comprehensively study tumor heterogeneity
in HCC. But several challenges remain in harnessing these technologies to the fullest.
A relatively unique predicament in HCC is that currently there is no clinical indication
to biopsy the tumor to confirm the diagnosis. The resulting lack of access to tissue
samples has already served as a major roadblock in studying tumor progression and
heterogeneity in HCC. The approval of novel biomarker-stratified therapies will hopefully
provide clinical justification for tissue sampling in patients with HCC in the future.
Until then, investigators should continue to pursue tissue samples acquisition under
research protocols with patient consent. Serial sampling of cfDNA is another promising
tool which can help us bypass the need to obtain multiregion tumor sampling to understand
heterogeneity. Although early results from cfDNA studies are promising, these still
need to be validated in larger cohorts before they can be rolled out to routine clinical
care. Moreover, it is unlikely that CTC or cfDNA will fully capture the heterogeneity
in the primary tumors both due to sensitivity of the assays and also the likelihood
that only a small proportion of the primary tumor cells or tumor DNA are in circulation,
thus limiting our capacity to detect subclonal mutations which are present in only
a subset of HCC cells.
Lastly, tumor heterogeneity is felt to be a major reason for the failure of multiple
drugs in clinical trials for cancer.[60]
[61] There is limited direct evidence that heterogeneity leads to therapeutic resistance
in HCC, since tumors will have to be biopsied before and after treatment to determine
this.[61] But given the degree of heterogeneity observed in HCC, and the experimental data
linking heterogeneity to drug resistance in other cancers, there is significant concern
that heterogeneity could lead to clinical trial failures in HCC.[60]
[62] We need to develop a strategy to account for tumor heterogeneity while enrolling
patients with HCC in future clinical trials. Protocol biopsies followed by single-cell
sequencing before and after treatment can help us understand the evolutionary trajectory
of therapy resistance. Moreover, biomarker-stratified approaches and subtype-enriched
enrollment in clinical trials can hopefully allow us to overcome the challenges introduced
by tumor heterogeneity.
To conclude, rapid developments in multi-omic and singulomic technologies are enabling
us to build quantitative, high-resolution maps of tumor landscapes, allowing us to
delineate cellular lineage and functions at a single-cell level. These advances allow
us to finally realize the full potential of personalized medicine in the care of patients
with HCC.
Main Concepts and Learning Points
Main Concepts and Learning Points
-
HCCs exhibit high degree of interpatient, intertumor, and intratumor heterogeneity.
-
Apart from the cancer cells, immune cells and stromal cells are also drivers of intratumor
heterogeneity.
-
Multitargeted tumor sampling combined with multi-omic sequencing strategies have enabled
us to gain deeper insights into tumor heterogeneity in HCC.
-
Single-cell sequencing has allowed us to create high-resolution maps of the diverse
cell populations in a heterogeneous tumor.
-
Insights gained from the multi-omic and singulomic study of tumor heterogeneity will
help develop novel biomarkers and improve therapeutic targeting of HCC.
