Abstract

   Liver cancer arises from the evolutionary selection of the dynamic
   tumor microenvironment (TME), in which the tumor cell generally becomes
   more heterogeneous; however, the mechanisms of TME-mediated
   transcriptional diversity of liver cancer remain unclear. Here, we
   assess transcriptional diversity in 15 liver cancer patients by
   single-cell transcriptome analysis and observe transcriptional
   diversity of tumor cells is associated with stemness in liver cancer
   patients. Tumor-associated fibroblast (TAF), as a potential driving
   force behind the heterogeneity in tumor cells within and between
   tumors, was predicted to interact with high heterogeneous tumor cells
   via COL1A1-ITGA2. Moreover, COL1A1-mediated YAP-signaling activation
   might be the mechanistic link between TAF and tumor cells with
   increased transcriptional diversity. Strikingly, the levels of COL1A1,
   ITGA2, and YAP are associated with morphological heterogeneity and poor
   overall survival of liver cancer patients. Beyond providing a potential
   mechanistic link between the TME and heterogeneous tumor cells, this
   study establishes that collagen-stimulated YAP activation is associates
   with transcriptional diversity in tumor cells by upregulating stemness,
   providing a theoretical basis for individualized treatment targets.

   Subject terms: Tumour heterogeneity, Cancer stem cells

Introduction

   Most liver cancer cases develop mainly due to evolutionary selection of
   an adverse tumor microenvironment (TME), in which deterministic tumor
   features are preferentially selected for their survival fitness [[46]1,
   [47]2]. Accordingly, complex genomics and TMEs create a molecular
   conundrum for the diagnosis and treatment of liver cancer and
   contribute to therapeutic failure and ultimately lethal outcomes
   [[48]2]. Tumor cells with strong stemness can generate heterogeneous
   subtypes through multidirectional differentiation [[49]3, [50]4].
   However, the underlying mechanisms of TME-mediated heterogeneity in
   liver cancer remain unclear. Molecular characterization of cell
   communities at the single-cell level may help shed light on the complex
   interplay among tumor cells and stromal cells in the TME.

   Liver cancer is etiologically and biologically heterogeneous,
   comprising many molecular subtypes [[51]5, [52]6], which are clinically
   treated as separate entities. Interestingly, there is evidence that
   some molecular subtypes of hepatocellular carcinoma (HCC) and
   intrahepatic cholangiocarcinoma (iCCA) share similar tumor biology and
   key drivers [[53]7]. Accordingly, the hepatic microenvironment could
   direct lineage commitment to either HCC or iCCA in the presence of the
   same oncogenic drivers [[54]8, [55]9]. The TME likely improves the
   capability of tumor cells to grow in poor microenvironment [[56]2,
   [57]8, [58]9]. The cellular components of the TME are highly complex
   resulting in high microenvironmental diversity associated with poor
   prognosis [[59]2, [60]9]. As an important constituent of the TME,
   tumor-associated fibroblast (TAF) is regarded as a promising
   therapeutic target for limiting cancer progression [[61]10]. Studies
   have established that TAF could facilitate cancer progression via
   aberrant extracellular matrix (ECM) remodeling with collagen I (COL-I)
   enrichment [[62]10, [63]11]. However, whether and how TAF regulates the
   heterogeneity of tumor cells is unclear, and requires further
   investigation.

   The Hippo-Yes-associated protein (YAP) signaling is known to regulate
   stem cell homeostasis, tissue regeneration and tumor progression
   [[64]12]. YAP is a major downstream effector of Hippo pathway [[65]12,
   [66]13]. Phosphorylation of YAP results in its cytoplasmic retention
   and degradation [[67]14, [68]15]. When dephosphorylated, YAP can enter
   the nucleus and bind with transcriptional factors to regulate the
   expression of many target genes [[69]16], usually increasing cell
   proliferation and decreasing apoptosis [[70]17]. YAP hyperactivation
   has been observed in many tumors, including liver cancer [[71]13,
   [72]18]. However, whether and how Hippo-YAP signaling responds to TME
   stimuli to affect the tumor heterogeneity of liver cancer remains
   unknown.

   Here, upon assessing the single-cell sequencing results of liver cancer
   patients, we identified transcriptional diversity of tumor cells
   potentially coexisted with the increased stemness. TAF interacted with
   high heterogeneous tumor cells via COL1A1-ITGA2, which activated
   YAP-signaling to regulate transcriptional diversity in tumor cells by
   improving their stemness. Thus, our study uncovered intercellular
   crosstalk between TAF and tumor cells involved in transcriptional
   diversity in liver cancer, suggesting potential targets for liver
   cancer therapy.

Experimental procedures

Cell culture and establishment of primary tumor-associated fibroblasts (TAFs)

   HCC cell lines (Huh7, LM-3, and HepG2) were purchased from ATCC and
   cultured in DMEM media (Gibco) with Fetal Bovine Serum (FBS, 10%;
   Gibco) and Penicillin-Streptomycin (100 U; Gibco) at 37 °C, 5% CO2.
   Fresh human tumor tissues were used to harvest primary TAFs. Tissue
   samples were cut into small pieces (approximately 3 mm^3) and put on to
   six-well cell culture plates with Dulbecco’s Modified Eagle’s Medium
   (DMEM) containing 10% FBS (Gibco), 10 ng/mL basic fibroblast growth
   factor (bFGF), 100 U/mL penicillin and 100 mg/mL streptomycin (Gibco).
   TAFs were starting to migrate out from the small piece of tissues from
   3 to 7 days later. After 2 weeks, the remnants of the tissue were
   carefully removed and subcultured TAFs every 3 days, and passages 3–10
   were used for this study [[73]19, [74]20].

Reagents and antibodies

   DEN (N0756) and antibodies specific for Flag (M3165) were obtained from
   Sigma-Aldrich (MO). Antibodies specific for YAP (sc-101199,
   immunofluorescence [IF]) were purchased from Santa Cruz Biotechnology
   (TX), and those specific for YAP (ab39361, IP/IHC), ITGA2 (ab181548),
   and COL1A1 (ab34710) were purchased from Abcam (MA).

Animal experiments

   Male SD rats (10–12 weeks, 220–250 g) were obtained from the Shanghai
   Experimental Center, Chinese Science Academy, Shanghai. The rats were
   randomly grouped and maintained at an animal facility under
   pathogen-free conditions. All animal experiments were performed
   according to the animal protocols approved by the Shanghai Eastern
   Hepatobiliary Surgery Hospital Animal Care Committee. To induce the
   model of liver cancer, 100 p.p.m. DEN (95 μg/ml) was added to the
   drinking water of rats for 16 weeks. Liver tumors were measured with
   electronic calipers and counted (for tumors with diameters ≥1 mm).
   Liver sections were preserved in 10% neutral-buffered formalin for
   histopathological and immunohistochemistry (IHC) analyses, blood was
   collected, and serum was isolated for biochemical analysis. All
   analysis were conducted in investigator blinded fashion.

Sphere formation assay

   Sphere formation assays were performed in 6-cm culture dishes coated
   with 1% agarose. HPCs suspended in serum-free medium were seeded at a
   density of 5 000 cells/dish and incubated for 3–7 days in low or high
   concentrations of collagen. The numbers and sizes of spheres were
   observed manually under a microscope.

Tissue microarray and IHC staining

   Tissue microarray (TMA) sections of tumor and adjacent nontumor
   specimens were prepared by Shanghai Outdo Biotech Co., Ltd. (Shanghai,
   China). This TMA contains tissues from 77 paired fresh liver carcinoma
   and adjacent tumor tissue samples, and was used to examine the
   expression profiles of YAP, ITGA2 and COL1A1 by IHC. For IHC, TMA
   sections were incubated with anti-ITGA2 antibody (1:200 dilution),
   anti-YAP1 antibody (1:100 dilution), or anti-COL1A1 antibody (1:200
   dilution). IHC staining was scored by two independent pathologists who
   were blinded to the clinical characteristics of the patients. The
   scoring system was based on the intensity and extent of staining:
   staining intensity was classified as 0 (negative), 1 (weak), 2
   (moderate), or 3 (strong); and the staining extent was dependent on the
   percentage of positive cells (out of 200 examined cells) and was
   classified as 0 (<5%), 1 (5–25%), 2 (26–50%), 3 (51–75%), or 4 (>75%).
   According to the staining intensity and staining extent scores, the IHC
   results were classified as 0–1, negative (-); 2–4, weakly positive (+);
   5–8, moderately positive (++), and 9–12, strongly positive (+++).

Immunofluorescent staining

   For antigen colocalization studies, fluorescence immunostaining of
   multiple proteins in tissues was performed with a sequential
   fluorescent method. Primary antibodies of against ITGA2 (1:200
   dilution), YAP (1:100 dilution) and COL1A1 (1:200 dilution) were used.
   Alexa 488-conjugated goat antimouse IgG (Invitrogen, Carlsbad, CA)
   Alexa 561-conjugated goat antirabbit IgG (Invitrogen) and Alexa
   647-conjugated goat antirabbit IgG were used as secondary antibodies.

Real-time quantitative polymerase chain reaction (RT-PCR)

   Total RNA was extracted from cells by using TRIzol Reagent (Invitrogen,
   Carlsbad, CA, USA), and further treated with RNase free DNase (Promega,
   Madison, WI, USA) to eliminate any residual DNA. Complementary DNA was
   prepared by using oligo dT18-primers and MMLV reverse transcriptase
   (Promega). RT-PCR was performed on a Light Cycler 480 system (Roche
   Diagnostics, Mannheim, Germany). The primers used in this experiment
   were as follows: SMAD2, forward, 5′-CCCACTCCATTCCAGAAAAC-3′, and
   reverse, 5′-GAGCCTGTGTCCATACTTTG-3′; SOX9, forward,
   5′-AGGAAGCTGGCAGACCAGTA-3′, and reverse, 5′-ACGAAGGGTCTCTTCTCGCT-3′;
   MYC, forward, 5′-AACAGGAACTATGACCTCG-3′, and reverse, 5′-AGCAGCTCGAA
   TTTCTTC-3′; and FGF8, forward, 5′-CAGTTGGAATTGTGGCAATCAAAG-3′, and
   reverse, 5′-CTTTTGATTTAAGGCAACGAACATTTC-3′.

Patients and follow-up analysis

   The cohort in this study contained 77 patients from January 1997 to
   December 2007. All patients were randomly selected from those with
   liver cancer who underwent hepatectomy in the Shanghai Eastern
   Hepatobiliary Surgery Hospital. Informed consent was obtained from each
   patient under a protocol approved by the Hospital Research Ethics
   Committees. None of the patients were administered preoperative
   treatment, and recurrence was confirmed by contrast-enhanced imaging
   studies or cholangiography according to standard guidelines for liver
   cancer. Overall survival (OS) was defined as the interval between
   surgery and either death or the last follow-up. The data were censored
   at the last follow-up for surviving patients.

Transfection and viral Infection

   Approximately 0.5–2 µg/ml plasmid was transfected using Lipofectamine
   2000 (Invitrogen) according to the manufacturer’s instructions. Where
   indicated, cells were infected with virus expressing GFP (multiplicity
   of infection (MOI), 10) or YAP (MOI, 10) in serum-free medium for the
   indicated times. Eight hours later, the cells were rinsed and cultured
   in fresh medium. After 48 h, the cells were cultured in DMEM
   supplemented with 10% FBS.

ChIP-qPCR

   ChIP was performed as previously described [[75]21]. Briefly, cells
   were cross-linked with freshly prepared formaldehyde (1.42%) for
   15 min, and treated with glycine (125 mM) for 5 min at room
   temperature. After two rounds of washing with ice-cold PBS, the cells
   were scraped and collected by centrifugation. Pelleted cells were
   resuspended in 400 μL of ChIP lysis buffer (50 mM HEPES/KOH, pH 7.5;
   140 mM NaCl; 1 mM EDTA; 1% Triton X-100; 0.1% Na-deoxycholate and
   protease inhibitors) and subjected to sonication with Bioruptor to
   shear the chromatin (30 s on high-power, 30 s off; 20 cycles). After
   sonication, samples were further diluted twice with lysis buffer and
   centrifuged to clear the supernatant. Eighty microliters of supernatant
   (1/10 of total) were directly processed to extract total DNA as
   whole-cell input. The remaining supernatants were transferred to new
   Eppendorf tubes and incubated with either IgG or YAP antibodies (14074,
   Cell Signaling Technology) at 4 °C overnight. Samples were then treated
   with prewashed protein A/G beads (L2118; Santa Cruz) for another 3 h,
   washed five times with the indicated buffers and resuspended in 100 μL
   of 10% Chelex (1421253; Bio-Rad). The samples were boiled for 10 min
   and centrifuged at 4 °C for 1 min. Supernatants were transferred to new
   tubes. After that, another 120 μL of Milli-Q water was added to each
   bead pellet, which was vortexed for 10 s, and centrifuged again to spin
   down the beads. The supernatants were combined together as templates
   for follow-up qPCR analysis.

ChIP-seq

   ChIP-seq was performed based on a previous protocol with minor
   modifications [[76]22]. Cells stably expressing Flag-tagged YAP were
   subjected to the same treatments as described above to obtain cell
   pellets. After that, cells were resuspended in 400 μL ChIP digestion
   buffer (20 mM Tris HCl, pH 7.5; 15 mM NaCl; 60 mM KCl; 1 mM CaCl2 and
   protease inhibitors). To shear the chromatin, cells were digested with
   an appropriate amount of micrococcal nuclease (MNase, M0247S, NEB) at
   37 °C for 20 min to ensure that the majority of chromatin was mono- and
   di-nucleosomes. The reaction was stopped with 2X stop buffer (100 mM
   Tris HCl, pH 8.1; 20 mM EDTA; 200 mM NaCl; 2% Triton X-100; 0.2%
   Na-deoxycholate and protease inhibitors). Samples were further
   sonicated with a Bioruptor at high power for 15 cycles (30 s on, 30 s
   off) and then centrifuged to remove debris. Next, soluble chromatin was
   immunoprecipitated with FLAG antibodies (F3165, Sigma) at 4 °C together
   with prewashed protein A/G beads. After extensive washing with the
   indicated buffers, samples were eluted and reverse cross-linked in
   elution buffer (10 mM Tris HCl, pH 8.0; 10 mM EDTA; 150 mM NaCl; 5 mM
   DTT and 1% SDS) at 65 °C overnight. Then, sequential digestion with
   DNase and Proteinase K was performed, and the DNA was purified with a
   PCR purification kit (B518141, Sangon Biotech). DNA that was
   successfully collected from three ChIP assays was pooled to generate
   libraries with the Ovation Ultra-Low Library Prep kit (NuGEN) according
   to the manufacturer’s instructions. Sequencing was performed on an
   Illumina HiSeq 2500 platform.

Cell dissociation

   HCC tumor tissues and adjacent tissues were collected after surgical
   resection in MACS Tissue Storage Solution (Miltenyi Biotec) in a 50 mL
   conical tube and transported on ice to the laboratory. Briefly, samples
   were first washed with phosphate-buffered saline (PBS), minced into
   small pieces (approximately 1mm^3) on ice, and enzymatically digested
   with collagenase I (Worthington) for 15 min at 37 °C, with agitation.
   After digestion, samples were sieved through a 70 µm cell strainer, and
   centrifuged at 300 g for 5 min. The cell pellet was resuspended in 1 mL
   freezing media (Gibco) for long-term cryopreservation in liquid
   nitrogen. Throughout the dissociation procedure, cells were maintained
   on ice whenever possible, and the entire procedure was completed in
   less than 1 hr.

scRNA-seq data processing

   ScRNA-seq data of tumor samples (n = 15; consistent with the original
   article, the sample would be taken into account only if it contains
   more than 20 tumor cells) (GEO: [77]GSE125449), healthy tissues (n = 4)
   and cirrhotic tissues (n = 3) (GEO: [78]GSE136103) were filtered (both
   gene and cell) and normalized by using the Seurat package (version 3.0)
   [[79]23] in R (version 3.5.3). Genes expressed in fewer than three
   cells per sample were excluded, as were cells that expressed fewer than
   500 genes or had a mitochondrial gene content >20% of the total UMI
   count. The total number of transcripts in each single-cell was
   normalized to 10,000. Highly variable genes were detected according to
   the average expression (between 0.05 and 3) and dispersion (above 0.5)
   of the genes, followed by data scaling (subtracting the average
   expression) and centering (divided by standard deviation). These
   variable genes were considered to account for cell-to-cell differences,
   and were further used for PCA. The first 20 PCs were applied for t-SNE
   analysis according to the eigenvalues.

Identification of nonmalignant cell types

   We extracted the transcriptome data of cells from the expression
   profiles of all the single cells evaluated. Similar to the total cell
   analysis, we first selected variable genes across cells, based on
   criteria of average expression (between 0.05 and 3) and dispersion
   (above 0.5) of the genes. Then, we performed data scaling followed by
   dimension reduction with PCA. The first 20 PCs were selected for t-SNE
   analysis. Different subclusters of cells were revealed on the t-SNE
   plot. We annotated the cells based on known cell lineage-specific
   marker genes as T cells (CD4, CD3E, CD3D, CD3G, CD8A, CD8B), B cells
   (CD79A, SLAMF7, BLNK, FCRL5), TECs (PECAM1, VWF, ENG, CDH5), CAFs
   (COL1A2, FAP, ACTA1, COL3A1, COL6A1), TAMs (CD14, CD163, CD68, CSF1R),
   and HPC-like (EPCAM, KRT19, PROM1, ALDH1A1, CD24).

CNV estimation

   Cells defined as endothelia, fibroblast and macrophage were used as
   references to identify somatic copy number variations with the R