Samples were acquired from the Helsinki Biobank at Helsinki University Hospital. Informed written consent was collected at the time of sample deposit. The study was approved by an ethics committee at Helsinki University Hospital (HUS/3319/2018) and was performed in accordance with Finnish bylaws. Tumor samples (n=21) were obtained from HB patients treated at Children’s Hospital, Helsinki University Hospital between January 1, 1990 and December 31, 2017. Sampling was performed during surgical resection or liver transplantation (after pre-operative chemotherapy). Normal liver (NL) control samples were collected from organ donors (n=3).

Five µm sections of formalin-fixed paraffin embedded tumor specimens were deparaffinized and immunostained with a monoclonal anti-human CAIX antibody (M75) (25). Immunoperoxidase staining was performed using an automated Lab Vision Autostainer 480 (LabVision Corporation, Fremont, CA, USA) and Power Vision+ Poly-HRP Immunohistochemistry kit reagents (ImmunoVision Technologies Co). The staining protocol included the following steps: (a) rinsing in wash buffer; (b) treatment in 3% H₂O₂ in ddH₂O for five minutes and rinsing with wash buffer; (c) blocking with cow colostrum diluted 1:2 in Tris-buffered saline (TBS) containing 0.05% Tween-20 for 30 minutes and rinsing in wash buffer; (d) incubation with 1:100 diluted M75 for 30 minutes; (e) rinsing in wash buffer three times for five minutes each; (f) incubation with poly-HRP-conjugated anti-mouse IgG for 30 minutes and rinsing in wash buffer three times for five minutes each; (g) incubation in DAB (3,3`-diaminobenzidine tetrahydrochloride) solution (one drop of DAB solution A and one drop of DAB solution B in 1 ml of ddH₂O) for six minutes and rinsing in ddH₂O; (h) CuSO₄ treatment for five minutes to enhance the signal and rinsing in ddH₂O; (i) treatment with hematoxylin for one minute; (j) rinsing with ddH₂O. All steps were performed at room temperature (RT). Imaging was performed using 3DHISTECH Panoramic 250 FLASH II digital slide scanner at Genome Biology Unit (Research Programs Unit, Faculty of Medicine, University of Helsinki Biocenter, Helsinki, Finland).

Raw microarray data of gene expression and clinical data from 53 HB tissue samples and 14 noncancerous liver tissue samples were acquired from the Gene Expression Omnibus (GEO) database of the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/geo/), accession number GSE131329. Microarray data were analyzed with Chipster software (https://chipster.rahtiapp.fi/) (26) using the normalization tool for Affymetrix gene arrays (27, 28). Statistical tests were conducted using the “Two group tests” tool (empirical Bayes as test and Benjamini-Hochberg as p-value adjustment method) (29). Differences in the distribution of continuous variables were assessed using the Mann-Whitney U test. Statistical significance was set to p-value < 0.05. Analyses were conducted with IBM SPSS Statistics version 28.0 (IBM Corp., Armonk, NY, USA).

The human HB cell line HUH6 was purchased from the Japanese Collection of Research Bioresources Cell Bank (Osaka, Japan). HB cell lines established from patient-derived xenografts (PDX; HB-303 and HB-295) were obtained from XenTech (Evry, France). HUH6 cells were maintained in Dulbecco’s modified Eagle medium (DMEM)-GlutaMAX (glucose: 1 g/l) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 µg/ml streptomycin (all from Gibco, Stockholm, Sweden). HB-303 and HB-295 cells were cultured in Advanced DMEM/F12 medium (Gibco) supplemented with 8% FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin, and 20 µM of Y-27632 (SelleckChem, Houston, TX, USA). Cells were routinely maintained at +37°C in a humidified incubator with 5% CO₂. In this study, these conditions represent normoxia (21% O₂) (30). Hypoxic conditions were generated utilizing the XVivo incubation system (partial pressures: 94% N₂, 1% O₂, 5% CO₂) (BioSpherix, Parish, NY, USA). All cell lines were authenticated through short tandem repeat profiling.

Carbonic anhydrase IX/XII inhibitor SLC-0111 (alias: U-104) was purchased from SelleckChem (catalog no. S2866) and dissolved in sterile DMSO as a 10 mM stock solution. Further dilutions were prepared in adequate cell culture medium. Normal cell culture medium supplemented with DMSO was utilized as a control treatment. Incubations were performed for 48 h if not otherwise stated, and the medium was not changed during the incubations.

RNA and protein extraction was performed with Nucleospin RNA/Protein Mini extraction kit (Macherey-Nagel, Düren, Germany) following the manufacturer’s instructions.

HUH6 cells were cultured under normoxic or hypoxic conditions and treated with 100 µM of SLC-0111 or vehicle. RNA was extracted after 48 h incubation. Prior to sequencing, RNA concentration, quality, and integrity were assessed by the Biomedicum Functional Genomics Unit (Helsinki Institute of Life Science and Biocenter Finland, University of Helsinki, Finland) using the TapeStation system (Agilent, Glostrup, Denmark). RNA libraries were prepared applying polyA selection, and Illumina compatible cDNA libraries were constructed by GENEWIZ (Leipzig, Germany). Subsequently, samples were sequenced on Illumina NovaSeq 6000 yielding 2x150bp paired end reads (GENEWIZ). An RNA sequencing dataset containing 11 HB patient samples and 11 NL samples was obtained from the GEO database of NCBI (accession number: GSE151347) (31, 32). FastQC tool was utilized to control quality of the reads (33). Sequenced reads were mapped to human reference genome hg38 using HISAT2 aligner (34). Reads per gene were counted with HTseq (35). To analyze differentially expressed genes (DEGs), edgeR2 tool was employed (36). Cut-off values were set to fold change lg₂ +/-0.8 and Benjamini-Hochberg adjusted p-value <0.05. Preprocessing of data and DEG analysis was carried out with Chipster (26). Gene Ontology (GO) analysis was performed with Enrichr (37, 38). R packages tidyverse and ggplot2 were used for data visualization in R (v. 4.0.3).

RNA was reverse transcribed using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA) following the manufacturer’s instructions. RT-qPCR was carried out with a CFX384 thermocycler instrument (Bio-Rad), and PowerUP SYBR Green Master Mix (Thermo Fisher Scientific) was used for gene amplification. Relative gene expression was assessed using the 2^-ΔΔCT method (39). Geometric mean of ACTB and PPIG expression served as a reference. Primer sequences were: ACTB; GCGTGACATCAAAGAGAAGC (forward), AGGATTCCATACCCAAGAAGG (reverse); CAIX GCCTTTGCCAGAGTTGACGA (forward), TCTGAGCCTTCCTCAGCGAT (reverse); PPIG CAATGGCCAACAGAGGGAAG (forward), CCAAAAACAACATGATGCCCA (reverse).

Proteins (10 µg) were separated by electrophoresis using Mini-Protean TGX Stain-Free Gels (Bio-Rad, Hercules, CA, USA) Next, proteins were transferred onto polyvinyl fluoride membrane. Blocking was performed with 5% non-fat milk in TBS. Primary antibody incubations were performed at +4°C for overnight (anti-human CAIX rabbit IgG at a of dilution 1:1500; NB100-417, Novus Biologicals, Littleton, CO, USA). Secondary antibody incubation was carried out at RT for 1 h (1:10,000; #111-035-144, Jackson ImmunoResearch, West Grove, PA, USA). Protein bands were illuminated using the Enhanced Chemiluminescence detection kit (Amersham ECL reagent; GE Healthcare, Barrington, IL, USA). Quantification was performed with Image Lab Software 6.0 (Bio-Rad). CAIX band intensities were normalized to the amount of total protein in the corresponding lane using stain-free technology (37).

Ibidi-treated cell culture inserts (3-well in µ-dish; Ibidi, Munich, Germany) were used to generate wounds. Cells were seeded into inserts in high density 24 h prior to the experiment. Before treatment initiation with vehicle or 100 µM of SLC-0111, inserts were removed, and cells were washed with phosphate buffered saline (PBS) to eliminate debris. Wounds were imaged at treatment outset (0 h) and after 20 h with an Eclipse TS100 microscope supplemented with the DS-Fi1 digital imaging system (Nikon, Tokyo, Japan). Wounds (16 images/insert) were analyzed with ImageJ software to calculate the percentage of wound closure. The following formula was used: Wound closure (%) = ((W₀ – W_t)/W₀) x 100 (W₀ = Wound area at 0 h and W_t = Wound area at 20 h).

Cells were seeded at a density of 2000 cells/well into 96-well ultra-low attachment plates (PerkinElmer, Waltham, MA, USA) and cultured without disturbance for 48 h at +37°C in a humified incubator with 5% CO₂. After the establishment period, cells were incubated with vehicle or increasing concentrations of SLC-0111 for 48 h.

Cell viability (ATP concentration) was assessed with the ATPLite™ 2D or 3D monitoring system (PerkinElmer) following manufacturer’s instructions. Luminescence was measured with a GloMax microplate reader (Promega, Madison, WI, USA).

HUH6 and HB-303 cells (100 000 cells/well) were grown in 4-well chamber slides coated with collagen I for 24 h. Cells were fixed with 4% paraformaldehyde. Non-specific binding was blocked with UltraVision Protein Block solution (Thermo Scientific, Fremont, CA, USA). Next, cells were incubated with primary antibody at room temperature for 1 h (NB100-417 human anti-rabbit CAIX at 1:1000 dilution, Novus Biologicals, Littleton, CO, USA). Secondary antibody incubation was performed with goat anti-rabbit IgG (H+L) AlexaFluor 647 (1 h, room temperature) at 1:800 dilution (A32733, Invitrogen, Carlsbad, CA, USA). Images were captured with a Zeiss Axio Imager M2 (objective: EC Plan Neofluar 40 X/0.75 Ph2 M27) (Carl-Zeiss, Oberkochen, Germany).

Spheroids were fixed with chilled 100% methanol for 20 minutes at RT. Following washes with PBS, 0.1% Triton-X was utilized to permeabilize the cells. Nonspecific binding was blocked with UltraVision Protein Block Solution (Thermo Fisher). Primary antibody incubation (anti-human CAIX rabbit IgG, at a dilution of 1:100; NB100-417, Novus Biologicals) was performed at RT for 1.5 h. Subsequently, spheroids were incubated with secondary antibody (anti-rabbit IgG (H+L) AlexaFluor 647, at dilution 1:200; A32733, Thermo Fisher) at RT for 1 h. Hoechst (at dilution 1:2000; #62249, Thermo Fisher) was used for nuclear staining. Opera Phenix High Content Screening System was employed to capture images (Perkin Elmer). Imaging was performed in the High Content Imaging and Analysis unit (FIMM, University of Helsinki).

Potential bioactive targets of SLC-0111 were assessed with SwissTargetPrediction (http://www.swisstargetprediction.ch/) (40) and SUPERpred (https://prediction.charite.de) (41) online tools.

Cell experiments were conducted in triplicate. Statistical analyses were carried out with GraphPad Prism (v. 8.4.2; San Diego, CA, USA). Student’s t-test or one-way ANOVA followed with Tukey’s test were utilized to assess statistical significance depending on the experimental setting. p-value < 0.05 was considered significant.

We analyzed 21 specimens of HB (11 male, 10 female) in the Helsinki Biobank. The median patient age at surgery was 3.18 years (0.23-10.83 years). Patient characteristics, treatments, and CAIX expression status are summarized in Table 1 . Three pediatric donor liver samples (age 2.0-8.2 years) were used as normal controls. Consistent with previous studies, in healthy liver CAIX immunostaining was restricted to bile duct cells ( Figures 1A, B ) (16, 17). Over 70% of the HB specimens demonstrated CAIX immunoreactivity; 9/21 had intermediate ( Figures 1C, D ) expression, and 6/21 had high CAIX expression ( Figures 1E, F ). CAIX staining was predominantly membranous in both the HB and healthy liver samples ( Figures 1A–F ). Within HB specimens, CAIX-positive cells were grouped in small clusters in the middle of viable tissue ( Figures 1C, D ) or adjacent to necrotic areas ( Figures 1E, F ), regions presumed to be hypoxic due to limited blood supply.

In the Helsinki cohort, all 5 cases of metastatic HB demonstrated CAIX immunoreactivity ( Table 1 ), suggesting that CAIX expression correlates with advanced disease. We used a larger patient cohort [GSE131329, a dataset containing 53 HB and 14 normal liver samples] to compare CAIX mRNA expression with three clinical variables – occurrence of an unfavorable event, metastasis, and overall survival. Total CAIX expression was higher in normal liver (median 7.485, [IQR 7.308–7.613]) than in HB samples (median 7.260, [IQR 7.115–7.470]) ( Figure 2A ), likely a reflection of CAIX expression in the biliary epithelium of normal tissue. The occurrence of any event was associated with higher CAIX expression (median 7.360 [IQR 7.245-7.640]) than an event-free disease course (median 7.145 [IQR 7.075-7.376]) ( Figure 2B ). Patients with distant metastases had higher CAIX expression (median 7.470 [IQR 7.223-7.638]) than those without metastases (median 7.170 [IQR 7.110-7.380]) ( Figure 2C ). Poor overall survival was associated with elevated CAIX expression (HB median 7.230 [IQR 7.162-7.360] vs. normal liver median 7.105 [IQR 6.999-7.170]) ( Figure 2D ).

We used cell culture models to investigate whether low oxygen tension induces CAIX expression in HB. The immortalized human HB cell line HUH6 or the PDX-derived cell lines HB-295 and HB-303 were cultured in a hypoxic chamber or normoxic incubator for 48 h. All three HB cell lines demonstrated little or no baseline CAIX mRNA expression when cultured under normoxia ( Figures 3A –C). CAIX mRNA expression was markedly upregulated in hypoxic cells compared to normoxic controls, with increases of 740-fold in HUH6 ( Figure 3A ), 165-fold in HB-295 ( Figure 3B ), and 6.7-fold in HB-303 cells ( Figure 3C ). Similarly, hypoxia induced 6- to 50-fold increases in CAIX protein levels ( Figures 3D–I ).

Next, we assessed the impact of 100 µM SLC-0111 on CAIX mRNA and protein expression under normoxic and hypoxic conditions. In normoxia, CAIX mRNA expression remained invariant after SLC-0111 treatment in all three cell models ( Figures 3A–C ). HB cells cultured under hypoxia and treated with SLC-0111 demonstrated a 40-60% reduction in CAIX mRNA expression compared to vehicle treated control cells ( Figures 3A–C ). Following SLC-0111 treatment, levels of CAIX protein decreased significantly in hypoxic HUH6 cells but not in HB-295 or HB-303 cells ( Figures 3D–I ).

To explore the effects of CAIX inhibition on HB cell survival in monolayer and spheroid cultures, we measured ATP concentrations, a surrogate for cell viability, following exposure of cells to increasing amounts of SLC-0111. In monolayer cultures, HUH6 cell viability decreased in a dose-dependent manner both in normoxia and hypoxia ( Figure 4A ). As with HUH6 cells, the impact of SLC-0111 on the viability of HB-295 monolayer cultures was more pronounced in normoxic than hypoxic conditions ( Figure 4C ). HB-303 had a dissimilar response to SLC-0111 than the other two models; cell viability increased with doses of 50-100 µM and with doses of 125-175 µM a modest decrease in viability was observed ( Figure 4E ).

The spatiotemporal distribution of oxygen in solid cancers cannot fully be mimicked in monolayer cell cultures. Instead, spheroids more closely resemble the 3-dimensional architecture of solid tumors, as oxygen levels differ for cells exposed directly to growth medium vs. those located in the inner parts of spheroids (42). To further assess the effects of SLC-0111 on HB cells, we used spheroids cultured in normoxia. SLC-0111 elicited a decrease in viability in all three HB spheroid models ( Figures 4B, D, F ). We also noticed spontaneous expression of CAIX in HB spheroids under normoxic conditions, whereas the cells grown in 2D showed negligible CAIX expression ( Supplementary Figure 1 ).

Several studies have reported decreased cell motility after pharmacological inhibition of CAIX or silencing of the CAIX gene (43–45). In our HB cell models, migration rates decreased significantly after 20 h treatment with 100 µM SLC-0111 compared to control cells both under normoxic and hypoxic conditions ( Figures 5A–C ). SLC-0111 had the most drastic effect on migration in HUH6 cells, wherein motility decreased approximately 70% in normoxia and 40% in hypoxia after 20 h of SLC-0111 treatment ( Figure 5A ). Hypoxia increased the migratory capacity of HB-295 cells compared to normoxic control cells, and SLC-0111 reduced motility in both normoxia and hypoxia ( Figure 5B ). A modest reduction in migration was observed in HB-303 cells treated with SLC-0111. In these cells the decrease was approximately 30% in normoxia and 35% in hypoxia compared to corresponding vehicle treated controls ( Figure 5C ).

As noted above, SLC-0111 decreased viability and motility in HB cells even under normoxic conditions when CAIX expression was undetectable or extremely low, suggesting that the drug may have CAIX-independent effects. To explore transcriptomic changes induced by SLC-0111 treatment, we performed RNA sequencing analysis for HUH6 cells treated with 100 µM SLC-0111 for 48 h under either normoxia or hypoxia. First, we assessed global gene expression alterations triggered by hypoxia compared to baseline expression in normoxia. A total of 2876 DEGs were observed of which 2155 genes were upregulated and 721 were downregulated ( Supplementary Figure 2 ; Supplementary Table 1 ). The three most upregulated protein coding genes were gamma-aminobutyric acid receptor subunit alpha-2 (GABRA2), CAIX, and aquaporin 10 (AQP10) ( Figure 6 ; Supplementary Table 2 ). Regulatory factor X6 (RFX6), acyl-CoA thioesterase 12 (ACOT12), and adrenoceptor alpha 2A (ADRA2A) were the most downregulated protein coding genes under hypoxia in comparison to normoxia ( Figure 6 ; Supplementary Table 2 ).

Next, we assessed the effects of SLC-0111 on the HUH6 cell transcriptome. In normoxia, we observed 304 upregulated genes and 96 downregulated genes after SLC-0111 treatment ( Supplementary Table 1 ). Under hypoxic conditions, SLC-0111 induced upregulation of 175 genes and downregulation of 312 genes ( Supplementary Table 1 ). Altogether, 76 genes were differentially expressed in both normoxic and hypoxic HUH6 cells treated with SLC-0111 ( Figure 7D ). Of these 76 genes, 15 genes were downregulated both in normoxia and hypoxia, 60 genes were upregulated in both conditions, and one gene was differentially regulated in hypoxia and normoxia ( Figure 7A ; Supplementary Table 3 ). Molecular functions associated with these overlapping genes included semaphorin binding, protein-arginine deaminase activity, and protease binding ( Figure 7B ). Metal ion related biological processes were highly overrepresented in SLC-0111 treated cells ( Figure 7C ). We also characterized overlaps in genes dysregulated in HB patient samples and expression alterations caused by SLC-0111 in HUH6 cells. A total of 7411 DEGs ( Supplementary Table 1 ) were noted in HB vs NL, and 35 of these genes were also differentially expressed in HUH6 cells following SLC-0111 administration ( Supplementary Table 1 ). SLC-0111 treatment of HUH6 cells caused upregulation of 20 genes that were downregulated in HB tumor tissue ( Table 2 ), including the tumor suppressor genes MT1G, MT1X, MT2A, OTC, PCK2, PGLYRP2, SERPINC1, and NR1I3. Three genes (FOXJ1, PRRT1, and TSSK5P) were downregulated in SLC-0111 treated HUH6 cells and upregulated in HB tumor tissue ( Table 2 ).

To identify other potential targets for SLC-0111, we performed in silico target prediction analysis with two online tools (SwissTargetPrediction and SUPERpred). We combined the common predicted targets from both tools ( Table 3 ). In addition to CAIX and CAXII, SLC-0111 had high expected probability of binding CAII ( Table 3 ). Other identified targets included histone deacetylase (HDAC) 3, thymidylate synthase (TYSY), nuclear factor NF kappa-B inhibitor kinase alpha (CHUK), mammalian target of rapamycin (mTOR), cyclin dependent kinases (CDKs) 1/2/4/5, and phosphatidylinositol 3-kinases PK3CA, PK3CB, and PK3CG ( Table 3 ).