Abstract
This study examines inhibiting galectin 1 (Gal1) as a treatment option for hepatocellular carcinoma (HCC). Gal1 has immunosuppressive and cancer-promoting roles. Our data showed that Gal1 was highly expressed in human and mouse HCC. The levels of Gal1 positively correlated with the stages of human HCC and negatively with survival. The roles of Gal1 in HCC were studied using overexpression (OE) or silencing using Igals1 siRNA delivered by AAV9. Prior to HCC initiation induced by RAS and AKT mutations, lgals1-OE and silencing had opposite impacts on tumor load. The treatment effect of lgals1 siRNA was further demonstrated by intersecting HCC at different time points when the tumor load had already reached 9% or even 42% of the body weight. Comparing spatial transcriptomic profiles of Gal1 silenced and OE HCC, inhibiting matrix formation and recognition of foreign antigen in CD45+ cell-enriched areas located at tumor-margin likely contributed to the anti-HCC effects of Gal1 silencing. Within the tumors, silencing Gal1 inhibited translational initiation, elongation, and termination. Furthermore, Gal1 silencing increased immune cells as well as expanded cytotoxic T cells within the tumor, and the anti-HCC effect of lgals1 siRNA was CD8-dependent. Overall, Gal1 silencing has a promising potential for HCC treatment
1. Introduction
Hepatocellular carcinoma (HCC) is a leading cause of cancer-related death globally, and early detection and effective treatment options are limited1. Current therapies for advanced-stage HCC only offer modest clinical benefits and are associated with various side effects, resulting in a median survival time of about a year2,3. To overcome these limitations, novel approaches are urgently needed to target the molecular drivers of HCC and shut down pathological signaling.
Galectin1 (Gal1) is a carbohydrate-binding lectin that interacts with glycoconjugate ligands of the extracellular matrix, endothelial cells, and T lymphocytes4, 5, 6, 7, 8. Gal1 is overexpressed in many types of cancers (e.g., liver, colon, breast, and lung) and is involved in multiple aspects of tumorigenesis, including cell proliferation, invasiveness, metastasis, and angiogenesis9, 10, 11, 12, 13. One known mechanism by which Gal1 promotes tumor growth is to induce epithelial to mesenchymal transition (EMT), a critical step for tumor initiation and being invasive14,15. Moreover, Gal1 can lead to the development of sorafenib-resistant cells through the activation of the PI3K/AKT signaling pathway16,17. Thus, Gal1 can be a target for cancer therapeutics. Surprisingly, using multidrug-resistance 2 and Gal1 double knockout mice, loss of Gal1 increases hepatic injury, inflammation, and fibrosis18. It is likely that silencing Gal1 in the germline can lose its immunosuppressive protective role and make the double knockout mice more susceptible to liver injury. Thus, the roles of Gal1 in HCC remain to be characterized.
Using xenograft models, a combination of a Gal1 inhibitor, OTX008, and sorafenib significantly reduced tumor growth13. However, in 2012, a phase 1 clinical trial was conducted to evaluate the effect of OTX008 to treat advanced solid tumors (ClinicalTrials.gov: NCT01724320). The data showed that OTX008 reduced serum Gal1 but with side effects. So far, the outcome of the study has not been released. Thus, there is a need to develop an effective, stable, and safe approach to inhibiting Gal1. Moreover, emerging evidence revealed the significance of gut microbiome and immunity in influencing liver health19,20. Therefore, it is also important to study liver carcinogenesis and treatment using orthotopic mouse models.
The current study examines the efficacy of gene therapy to silence Gal1 in HCC treatment using orthotopic preclinical models. Adeno-associated virus (AAV) serotype 9, which has been demonstrated to be safe in 263 clinical trials, was used to deliver Gal1 for overexpression or silencing21. Our data revealed that Gal1 overexpression prior to tumorigenesis facilitated liver carcinogenesis, while silencing it could prevent and treat HCC as well as prolong survival. Mechanistically, Gal1 silencing acts on several pathways based on locations. Those pathways included regulating translation machinery within the tumor, reducing the interactions between CD45+ cells and matrix formation at the tumor margin, shifting T cell populations, etc. In addition, the HCC treatment effect of Gal1 silencing was cytotoxic T cell dependent. Together, the generated data demonstrated promising outcomes for both preventative and therapeutic applications of AAV9-mediated Gal1 silencing, which selectively targets stromal and tumor cells and exhibited no observable toxicity in preclinical models.
2. Material and methods
2.1. Clinical specimens
Frozen liver cancers (n = 11) and normal livers (n = 9), confirmed by histological evaluation, were obtained from the Translational Pathology Core Laboratory Shared Resource at the University of California, Los Angeles (UCLA, CA, USA). Among them, six tumors and adjacent normal tissues were paired and derived from six HCC patients. The tissue procurement process was approved by the UCLA Institutional Review Board with protocol number 11-2504 approved on 1 February 2011. A microarray slide consisting of 22 HCC specimens was obtained from the UC Davis Pathology Biorepository at UC Davis Health under the UC Davis Comprehensive Cancer Center.
The human tissue procurement process was approved by the UCLA Institutional Review Board with protocol number 11-2504 approved on 1 February 2011.
2.2. Generating HCC mouse models and treatment strategies
Wild-type male and female FVB/N mice were obtained from the Jackson Laboratories (Sacramento, CA, USA) and were housed in regular filter-top cages at 22 °C with a 12 h:12 h light cycle. For hydrodynamic injection, myr-Akt1 and N-RasV12 (1 μg/g body weight) and sleeping beauty transposase plasmid (0.08 μg/g body weight) were diluted in 2 mL PBS filtered (0.22 μm) and injected into the lateral tail vein of 6 weeks-old mice in 7 s. Constructs used in these animal experiments showed long-term expression of genes via a hydrodynamic injection22. Animal experiments were conducted following the National Institutes of Health Guide for the Care and Use of Laboratory Animals under protocols approved by the Institutional Animal Care and Use Committee of the University of California, Davis, CA, USA.
To silence the Gal1, AAV9 (Applied biological material, Richmond, BC, Canada) was used. Target sequences were: target a—81:CGCCAAGAGCTTTGTGCTGAA, target b—178:GTGTGTAACACCAAGGAAGAT, target c—306:AGACGGACATGAATTCAAGTT, and target d—367:GCGGATGGAGACTTCAAGATTAAGTGCGT. The vector size was 6593 bp (Supporting Information Fig. S1A). The vector was tagged with GFP. Gal1 siRNA (1012 genome copy/kg BW) was administered intravenously once. Scramble-AAV9 was used as a control. The same vector was used to overexpress the Gal1, 1012 genome copy/kg BW (Fig. S1B). Blank AAV9 was administered as a control (Fig. S1C).
To understand how immune cells are involved in the therapeutic effects of lgals1 siRNA, CD8 was depleted in mice by i.p. injection of 200 μg/kg body weight of either anti-mouse CD8α (BE0004-1; Bioxcell) or isotype control (BE0089; Bioxcell) two times a week for 3 consecutive weeks.
2.3. Histology and immunohistochemistry
Tissues were fixed in 10% formalin for 16 h and kept in 70% ethanol for 72 h, followed by embedding in paraffin and cutting into 5-μm sections. Hematoxylin and eosin (H&E) staining was performed23. Mouse tumor score was quantified by pathologists based on H&E stained slides using five criteria (i.e., the level of centrilobular vacuolar degeneration, the number of proliferation foci, scirrhous type foci of proliferation mitotic index, and inflammatory cells)24,25. Details are described in Supporting Information Table S1.
Immunostaining was performed with specific antibodies against Gal1 (Abcam, Cambridge, MA) and CD8 (eBioscience, Waltham, MA). The number of positive-staining cells was counted in at least five random microscopic fields (100 × magnification) for each section.
2.4. In situ apoptosis analysis
Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay kit (Abcam, Cambridge, MA, USA) was used. Liver sections were incubated with 3% H2O2 to inactivate endogenous peroxidases after being treated with proteinase K. Biotin labeling of DNA-exposed 3′-OH ends of apoptotic cells was produced by adding terminal deoxynucleotidyl transferase (TdT) enzyme, and TdT labeling reaction mix. Biotinylated DNA was detected by incubation (30 min) with a streptavidin–HRP conjugate. After washing in TBS, DAB substrate was added, and reactions were stopped. The sections were counterstained with Methyl Green, dehydrated, and then coverslipped. Images were taken under a Nikon Eclipse E600 microscope. The number of apoptotic cells was determined by counting positive-staining cells in at least five random microscopic fields (100 × magnification) for each specimen.
2.5. Measurement of biochemical parameters and cytokines
Serum samples were used to detect alanine aminotransferase, aspartate transferase, cholesterol, lipase, and creatine kinase by colorimetric method slide based on published methods (Fujifilm DRI-CHEM SLIDE)26. Mouse interferon-gamma (IFN-γ) and granzyme B were quantified using the ELISA kits (Abcam, MA, USA). To evaluate the collagen content, hepatic hydroxyproline concentration was determined using the colorimetric assay kit (Sigma–Aldrich).
2.6. RNA isolation, quantitative real-time PCR
RNA was extracted using a TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA). cDNA was synthesized using a high-capacity RNA-to-cDNA kit (Applied Biosystems, Carlsbad, CA, USA). Real-time quantitative PCR with reverse transcription (RT-qPCR) was performed on a QuantoStudio5, applied biosystems by Thermo Fisher scientific system using Power SYBR Green PCR master mix (Life Technologies). Primers were designed using Primer3 Input software version 0.4.0. The glyceraldehyde-3-phosphate dehydrogenase (Gapdh) mRNA level served as an internal control. Data generated from healthy livers were used to establish a baseline to calculate the relative expression levels between groups. The relative mRNA level was calculated by the 2−ΔΔCT method.
2.7. RNA sequencing and functional analysis
RNA quality control was performed using Qubit and Bioanalyzer instruments. Libraries were prepared using the NEBNext Ultra II non-directional RNA Library Prep kit. Library quality and concentration were assessed with Labchip and qPCR and subjected to sequencing using Novaseq6000. Differential expression analysis of two conditions/groups (two biological replicates per condition) was performed using the DESeq2 R package (1.20.0). DESeq2 provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting P-values were adjusted using Benjamini and Hochberg’s approach for controlling the false discovery rate. Genes with an adjusted P-value. The DEGs were used for KEGG analyses using the cluster profile package25, and an adjusted P-value of <0.05 was considered a significant event. Moreover, Gene set enrichment analysis (GSEA) was utilized to discover the molecular mechanism of the whole genome at the transcription level rather than the DEGs.
2.8. GeoMx DSP whole transcriptome workflow
Liver sections (4 μm) were used for digital spatial profiler (DSP) whole transcriptome sequencing (NanoString). The panel of morphology markers used included CD45 in addition to SYTO13 nuclear stain and Pan-cytokeratin. Slides were stained with RNAscope probes and GeoMx DSP oligo-conjugated RNA detection probes according to manufacture protocol.
Slides were loaded onto the GeoMX DSP slide holder with 6 mL of Buffer S (GeoMx RNA slide prep kit, Nanostring). Files were configured to associate RNA targets and GeoMx readout barcodes. The appropriate scan type and focus channel were selected to populate fluorescence exposure settings. Scan areas were defined for high magnification. A total of 8 regions of interest (ROIs) with a diameter of 300–600 μm per group were selected (4 tumors and 4 margins). The Geomx software was used to define the area of illumination within each ROI, was further segmented based on morphological markers (CD45, SYTO13, and PanCK), followed by UV-cleaved DSP barcodes from each ROI, were collected, dispensed into a 96-well plate, and counted. These barcodes were tagged with their specific RNA target identification sequence and ROI location during library preparation; therefore, can be matched to their in situ hybridization probes and the unique molecular identifier for read deduplication. The sequenced oligonucleotides were then imported into the GeoMx DSP platform to be integrated with slide images and ROI selections for spatially resolved RNA expression. The FASTQ sequencing files were converted into digital count files using Nanostring’s GeoMx NGS Pipeline software. Quality control checks and data analysis were performed using the GeoMx DSP Data Analysis suite. The data were filtered by the limit of quantitation and then normalized by the third quartile of all counts. Cell deconvolution analyses were performed using the spatialdecon geoscript (v1.3 updated October 2022) available at Nanostring’s Geoscript Hub.
2.9. Cell deconvolution analyses
The spatialdecon geoscript (v1.3 updated October 2022) available at Nanostring’s Geoscript Hub was used to generate cell deconvolution analyses in the GeoMx DSP control center [https://nanostring.com/products/geomx-digital-spatial-profiler/geoscript-hub/].
The abundance of immune cells (e.g., CD8 T cell) was calculated based on the single sample GSEA enrichment score of the expression deviation profile per cell type27,28. Subsequently, the calculated enrichment score was normalized, which is used as the abundance of immune cells (http://bioinfo.life.hust.edu.cn/web/ImmuCellAI/).
2.10. Bioinformatics analysis of human HCC
Normalized mRNA data for liver hepatocellular carcinoma (LIHC) were obtained from the University of California, Santa Cruz (UCSC) Xena browser (https://xenabrowser.net/). Univariate and bivariate expression analyses among cases were performed with the R software. To further explore the prognostic value of the gene-expression signature driven by lgals1 (high or low), cancer patients were divided into two subgroups for survival analyses. Discretization of the lgals1 gene expression data into low- or high-expression levels was performed according to the StepMiner one-step algorithm (http://genedesk.ucsd.edu/home/public/StepMiner/). These two groups were then compared for overall survival and disease-specific survival (Kaplan–Meier curves and log-rank test) using the survival and survminer R packages.
2.11. Statistical analysis
All quantitative data are expressed as mean ± standard error of mean (SEM). Groups were compared using one-way ANOVA followed by a Tukey test. Mann–Whitney test was used to compare two groups. Survival curves were plotted by Kaplan–Meier analysis and compared using the log-rank test. All analyses were carried out using the GraphPad Prism 9.0 statistical software package (San Diego, CA, USA), and P values < 0.05 were considered significant.
3. Results
3.1. Gal1 is overexpressed in human and mouse HCC
Histological normal human livers (n = 9) had low levels of LGALS1 mRNA in comparison with human HCCs (n = 11) revealed by real-time PCR. Among those specimens, six paired specimens were obtained from the same HCC patients (Fig. 1A). The findings were confirmed by immunohistochemistry (IHC) using another cohort of 22 specimens from HCC patients. The number of Gal1-positive cells was quantified using ImageJ. In human HCC, about 80% of cells demonstrated high immunoreactivity of Gal1 (Fig. 1B)…
