Constitution of an inter- and intra- heterogeneous panel of primary HCC cells from the Alb-R26 Met liver tumours

The overall workflow of the present study and the methodologies employed are illustrated in Fig. 1. Briefly, a cohort of Alb-R26Met mice was imaged by photon-counting micro-computed tomography (PC-CT) [40] to identify those carrying tumours. In seven Alb-R26Met mice, we detected between 1 and 13 tumours, varying in their size and liver location (Fig. 2A; Table S3). Eight tumours were dissected and processed as indicated in Fig. 2. Histopathological analyses of Alb-R26Met tumours confirmed that they were exclusively HCC (Fig. 2B-C, Table S4), as previously reported [34, 36]. In particular, we observed loss of physiological architecture as compared to normal mouse liver, with complete loss of the portal tract structure, signs of cytological atypia, fatty change, inflammation, and necrosis in some area, diagnosed by the pathologist as a well differentiated HCC (Fig. 2B-D, Table S4). To note, T20 displayed a cystic histology, as shown by arrows in Fig. 2B. Dissociated cells from dissected tumours were seeded in 2D culture condition to generate primary HCC cells, then characterized morphologically and molecularly. We were able to establish primary HCC cells from all eight Alb-R26Met tumours, which we named HCC16 to HCC23 (Fig. 3A). HCC16 and HCC19 primary cells were originated from two different Alb-R26Met tumours of the same mouse (Table S3). In the Alb-R26Met mice, the Mettg gene is expressed following Cre-mediated excision of a STOP cassette containing the β-Gal gene [37]. By performing X-Gal staining, we found a variability in the proportion of recombined and not-recombined cells across the eight established Alb-R26Met primary HCC cells (Fig. S1). In particular, HCC19 and HCC21 primary cells are exclusively composed of recombinant cells, whereas the others have a variable percentage of recombinant cells, ranging from 99.69% for HCC16 to 7.81% for HCC18 (Fig. S1).

Fig. 1

Schematic representation of the overall study. (A) Spontaneously generated tumours from the Alb-R26Met mouse model were monitored by PC-CT using the PIXSCAN-FLI imaging system. Dissected tumours: 1) were characterized by Haematoxylin and Eosin staining (HES) histological analyses; 2) underwent both mechanical and enzymatical cell dissociation. (B) Dissociated cells were seeded in 2D culture conditions. Primary HCC cells were characterized morphologically and molecularly by RT-qPCR and Western Blots (WB), then used for drug testing. (C) Primary HCC cells were secondarily cultured in 3D condition into Matrigel after few passages in order to generate tumoroids, which were further characterized (created with BioRender.com)

Fig. 2

Alb-R26Met tumours with distinct liver locations and histological characteristics. (A) In vivo PC-CT imaging of spontaneous tumours in the Alb-R26Met model. The eight tumours used for the generation of primary HCC cells and tumoroids are highlighted with a yellow dotted line. (B) HES sections from tumours highlighted in (A) used for the study. Dotted lines show the boundary between nodule in nodule areas. Note: T19 shows areas composed by either small or large cells, exemplifying heterogeneity in cell morphology; T20 displays cystic architecture delineated by arrows; (*) shows nuclear atypia, including high N/C ratio, irregular nuclear membrane, multinucleation; (♦) shows cytoplasmic hyaline globule; (♣) shows macro and micro-vesicular hepatosteatosis. (C) Table summarizing major histological characteristics displayed by the tumours (well differentiated: well diff; Highpower field: HPF). (D) Haematoxylin and Eosin staining of a normal mouse liver section. The doted lines indicate the sinusoids draining into central vein

Fig. 3

Alb-R26Met primary HCC cells display morphological and molecular heterogeneity. (A) Brightfield pictures documenting the different morphologies of the eight Alb-R26Met primary HCC cells. (B-C) E-cadherin (B) and Vimentin (C) immunostaining performed at 48h after seeding the cells on poly-lysine coated coverslips, documenting inter- and intra-tumour heterogeneity regarding these two epithelial and mesenchymal markers, respectively. Arrowheads point to some examples of positive cells. (D) Heatmap reporting the mRNA expression levels of the indicated markers of HCC, proliferation, EMT cancer-stem cell (CSC) like, epithelial, progenitor, HCC prognosis and Wnt/β-Cat, in Alb-R26Met primary HCC cells, compared to control livers (n=2-6) and to a previously characterised Alb-R26Met HCC cell (HCC13) (34). Statistical analyses were performed via two-way ANOVA, followed by Dunnett’s multiple comparison. RQ stands for relative quantification compared to control livers. (E) Dot plot showing the positive correlation between the expression levels of Mki-67 proliferation marker and the growth ratio of Alb-R26Met primary HCC cells. Statistical analyses were performed using one-way ANOVA, followed by Dunnett’s multiple comparison as reported in Fig. S4. Levels of significance: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001. RQ stands for relative quantification compared to control livers

Morphological characterisation of these primary HCC cells showed inter- and intra-heterogeneity (Fig. 3A). Specifically, HCC18, HCC20, and HCC21 cells display predominant fusiform and mesenchymal-like shape, whereas the other cells have a more epithelial-like morphology polygonal in shape (Fig. 3A). Such inter-/intra-heterogeneity is further evidenced by analysing Vimentin (Vim; a mesenchymal marker) and E-cadherin (E-Cad; an epithelial marker) expression, with variable levels among the different primary HCC cells (Fig. 3B-C, S2, S3). Notably, HCC18, HCC19, HCC20, and HCC22 are composed by cells positive for both E-Cad and Vim (Fig. 3B-C, S2, S3), possibly indicating a hybrid state characterising cancer cell aggressiveness [47].

We further analysed their molecular features at the transcriptomic level by performing RT-qPCR analyses. Results showed that HCC16, HCC17, HCC19, HCC20, and HCC21 primary cells express Alpha-fetoprotein (Afp), an HCC marker (Fig. 3D, S4A). HCC18 cells are negative for Afp, although positive for Hsp1a1 coding for HSP70, another HCC marker (Fig. 3D, S4A). We also observed a heterogeneity in the expression of Adamtsl5, another HCC marker [35], essentially expressed in HCC17 and HCC18 primary cells (Fig. 3D, S4A). Instead, all cells are negative for Cltc (coding for Clathrin Heavy Chain 1) and Gpc3 (Glypican-3) liver cancer markers (Fig. 3D, S4A). All primary HCC cells are highly expressing Mki-67, a proliferation marker (Fig. 3D, S4A), indicating high proliferation capacities. Transcriptional evaluation of Cdh1 and Vim revealed that primary HCC cells express various levels of both markers (Fig. 3D, S4A). We additionally found that primary HCC cells exhibit high levels of the progenitor marker Krt19, and of the HCC prognostic markers Cd24a, Cdc20, and Igf2bp3. Interestingly, we also found high levels of Vcam1, a marker of early hybrid Epithelial-Mesenchymal Transition (EMT) (not expressed only in HCC19; Fig. 3D, S4A). On the other hand, we observed low levels of Krt18 (epithelial marker), Twist1 (mesenchymal marker), Zeb1, Fat1, Cd51 (markers of early and late hybrid EMT, respectively). Finally, a slight increase of Nanog, but not Epcam (cancer stem cells markers), was found in most of the cells (Fig. 3D, S4A). Instead, all primary HCC cells are negative for other markers characterizing HCC subtypes [34], such as Glul. Overall, primary HCC cells exhibit high expression variability across genes associated with poor prognosis, progenitor (Igf2bp3, Krt19), proliferation (Mki67), and stemness (Cd24a, Nanog) (Fig. S4B). Collectively, these findings illustrate the heterogeneity of Alb-R26Met primary HCC cells, likely reflecting a broad transcriptomic range, underscoring the added value of integrating gene expression data to morphology-based classification.

Alb-R26 Met primary HCC cells are heterogeneous in their proliferation capability and signalling competences, whereas all resistant to RTKi

The heterogeneous Mki-67 expression levels in Alb-R26Met primary HCC cells prompted us to analyse possible correlation with their growth kinetics. We found that higher Mki-67 levels in primary HCC cells positively correlate with their faster proliferation rate (Fig. 3E, S4A, S5A). In particular, HCC18 and HCC19 cells are characterised by high Mki-67 and growth kinetics, compared with HCC16, HCC20, and HCC22 cells. Through biochemical studies, we next analysed the signalling properties of Alb-R26Met primary HCC cells and found heterogeneous expression levels of the RTK MET and FGFR4, whereas VEGFR2 is predominantly expressed in HCC20 cells (Fig. 4A, S5B). Intrigued by the variable Mettg protein levels across primary HCC cells, we analysed heterogeneity of Mettg transcript levels by reanalysing bulk RNA-seq data from 12 Liver tumours and 6 non-tumour livers from Alb-R26Met mice (Table S5), previously described [36]. The Mettg in Alb-R26Met mice is constituted by an extracellular portion corresponding to the mouse sequence and an intracellular domain corresponding to the human sequence [37, 38]. We found a consistent coverage across the 3′ region of the transcript (encoding the intracellular human portion of MET) in both livers and tumours of Alb-R26Met mice (Fig. S6). Furthermore, these findings document elevated levels of the Mettg transcripts in tumours compared with livers, with variable levels among them (Fig. S6), indicating distinct mechanisms of Mettg regulation across tumours/cells. Heterogeneity among Alb-R26Met primary HCC cells was also observed in the activation levels of downstream pathways, as shown by the phosphorylation status of AKT, MEK, and ERK, and in the expression levels of a cancer-stem cell like marker SOX9 (Fig. 4A, S5B). No evident correlations were observed among marker and signalling levels, similar to what is reported in HCC patients. Next, we explored the effects of Alb-R26Met primary HCC cells to RTKi used in the clinic for treatment. Results showed an overall resistance pattern of all Alb-R26Met primary HCC cells to cabozantinib, lenvatinib, sorafenib, and regorafenib, with slight differences only when high doses were used (Fig. 4B). Together, these results illustrate the degree of heterogeneity on proliferation and signalling of this Alb-R26Met primary HCC cell panel, modelling as well HCC resistance to RTKi.

Fig. 4

Alb-R26Met primary HCC cells are characterised by heterogeneous expression and phosphorylation levels of signals and by resistance to RTKi. (A) Western blot of the indicated signalling proteins. Actin or Tubulin were used for normalization. HCC13 cells were used as a positive control (34). Quantifications are reported in Fig. S5B. Full western blot and ponceau staining are shown in Fig. S10-12. (B) Dot plot reporting the percentage of viable cells after treatment (1, 3, 10 µM) with the indicated RTKi (cabozantinib, lenvatinib, sorafenib, and regorafenib). Statistical analyses were performed using one-way ANOVA followed by Tukey’s multiple comparison. Levels of significance: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001

Alb-R26 Met primary HCC cells are competent to form tumoroids and recapitulate anatomopathological features of primary tumours

The generation of liver tumoroids from biopsies or surgical specimens of patients and/or animal models has been the focus of several attempts and studies over the last years. Nevertheless, the low success rate (26–30%) in generating HCC tumoroids remains a major limitation, challenging their use as disease models. We therefore assessed the possibility of exploiting the clinically pertinent Alb-R26Met HCC model to constitute a panel of tumoroids, while optimizing the methodology for their generation, based on previous reports [14, 15, 32]. First, we hypothesized that a short expansion phase in 2D in vitro culture of freshly dissociated cells from liver tumours may permit an enrichment of cells competent to form tumoroids, without significantly impacting the heterogeneous composition of cell type diversity as observed after several passages. For tumoroid generation, we used Alb-R26Met primary HCC cells cultured in 2D conditions with passages ranging from P4 to P12. Cells were trypsinized and plated into 3D undiluted Matrigel domes (Fig. 5, top raw strategy using suspension plate conditions). We additionally explored the capability of cells to form tumoroids in ultralow adherent plates (Fig. 5, bottom raw strategy), a condition particularly pertinent to assess drug effectiveness on tumoroids [48]. Second, inspired by few protocols reported in the literature for the generation of tumoroids from HCC (summarized in Fig. S7), we considered pertinent to test several slight modifications of culture conditions. The overall outcomes of different experimental settings allowed us to successfully generate Alb-R26Met tumoroids from all of the eight primary HCC cells (Fig. 6A; we named tumoroids as tHCC16 to tHCC23). We observed a morphological heterogeneity across different tumoroids. Whereas a subset of them is characterised by spherical and compact structures (tHCC16, tHCC17, tHCC22, tHCC23), others have spherical structures with some protrusions similar to buds (tHCC18, tHCC19), and few displayed cystic shapes with a cleared centre surrounded by an envelope (tHCC20, tHCC21; Fig. 6A). Anatomopathological analysis of paraffin-embedded tumoroids confirmed that Alb-R26Met tumoroids exhibited HCC morphology, with features resembling primary tumours (Figs. 2B and 6B-C, Table S4). The differentiation of the tumoroids were additionally noted to be the same as the original tumours, with a similar degree of cytologic atypia, thus providing further verification (Table S4). Notably, tHCC20 recapitulates the cystic structure of the corresponding T20 tumour, with trabecular pattern seen next to cyst (Figs. 2B-C and 6B).

Fig. 5

Schematic representation of the generation of tumoroids from primary HCC cells using domes in suspension and ultra-low attachment plates. Top raw reports the procedure used to culture tumoroids in domes within suspension plates. This strategy was used to generate tumoroids from 2D primary HCC cells and to monitor their growth. Bottom raw reports the procedure used to assess tumoroid viability to drug treatments. This was performed on ultra-low adherent plates to avoid that a proportion of seeded cells attach to the plate and grow in 2D, interfering with drug effectiveness interpretation

Fig. 6

Morphological, histological, biological, and molecular characterisation of Alb-R26Met tumoroids. (A) Brightfield pictures documenting the capability of Alb-R26Met primary HCC cells to form tumoroids with distinct morphologies in suspension plates. Scale bar corresponds to 200 µm. (B) HES of Alb-R26Met tumoroids recapitulates the histology of the corresponding primary tumours from which they were generated. Scale bar represents 200 µm. (C) Table summarizing the histological characteristics of Alb-R26Met tumoroids, consistent with the originating tumour histology. (D) Graph reporting the growth capability of the Alb-R26Met tumoroids in suspension plates. The “mean final area” of all the tumoroids was measured before splitting from passage P0 to P5. (E) Heatmap reporting the mRNA expression levels of the indicated markers of tumoroids grown in suspension plates. RQ stands for relative quantification compared to the control liver. Statistical analysis was performed using one-way ANOVA, followed by Dunnett’s multiple comparison. Levels of significance: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001

At biological levels, we found that the Alb-R26Met tumoroids maintain their growth capacity along passages (we tested up to passage 7 for all, to passage 11 for some of them). By monitoring their growth features over time and across passages in suspension plates, we observed that tHCC16 to tHCC20 and tHCC23 are able to generate significantly bigger tumoroids than tHCC21 and tHCC22, displaying smaller size. Moreover, all of them maintained a constant or increased ability to form tumoroids across passages (Fig. 6D, S8A). Collectively, results emphasize the behavioural diversity of this tumoroid panel (Fig. 6D). Tumoroid formation was confirmed using ultra-low adherent plates (Fig. S8B). No major morphological and growth differences were observed on Alb-R26Mettumoroids cultured in Matrigel-based domes in suspension plates versus ultra-low attachment plates.

At transcriptomic level, RT-qPCR analysis revealed molecular heterogeneity between samples (Fig. 6E), as observed in primary HCC cells. Specifically, we found heterogeneous expression levels of Afp, Adamtsl5, Mki-67, Cdh1, Vimentin, Vcam1, Fat1, Itgav, Zeb1, Twist1, Nanog, Epcam, Cd24a, and Cdc20 (Fig. 6E, S9A). Expression levels in Alb-R26Met tumoroids are comparable to those in primary HCC cells, despite slight differences (Figs. 3D and 6E; S4A, S9A). Interestingly, Alb-R26Met tumoroids express higher levels of Adamtsl5 compared to primary HCC cultures (Figs. 3D and 6E; S4A, S9A). Tumoroids showed a certain level of expression variability across genes (Fig. S9B), although to a less extent compared to primary HCC cells (Fig. S4B). Overall, these results illustrate the robustness of the methodology we established to generate tumoroids after a short expansion of primary HCC cells in 2D conditions, with a 100% success rate, ensuring maintenance of their heterogenic features.

Alb-R26 Met tumoroids highlight HCC vulnerability to combinatorial romidepsin plus cabozantinib treatment

By recapitulating tumoral cell-cell and cell-matrix interactions, tumoroids offer a more reliable opportunity to evaluate the effectiveness of anticancer treatments. Moreover, the use of a heterogeneous tumoroid panel enables to assess drug effects on HCC settings with distinct molecular properties. We reasoned to exploit the Alb-R26Met tumoroid panel by assessing the effectiveness of a new treatment we recently discovered [49]. Specifically, by analysing HCC patient data from eight independent cohorts, we found that overexpression of HDAC1 and HDAC2 in HCC patients correlates with decreased overall survival [49]. We documented how romidepsin, a class-I Histone deacetylase (HDAC) inhibitor (HDACi), through a combinatorial effect on distinct molecular regulators, leads HCC cells to a vulnerable state, conferring dependency to RTK signalling support. This is exemplified by a combinatorial action of romidepsin with cabozantinib (RomiCabo) on HCC cells [49]. We therefore used the Alb-R26Met tumoroid panel to assess the robustness of RomiCabo treatment across heterogeneous HCC settings. For drug testing, ultra-low attachment plates are superior to suspension plates for the enhanced size and speed in tumoroid growth. Additionally, ultra-low attachment plates avoid the formation of a cell layer on the bottom at the plate, which can interfere with accurate evaluation of drug effectiveness (Fig. 5). Coherently with results shown in Fig. 4B for primary HCC cells in 2D conditions, Alb-R26Met tumoroids are resistant to cabozantinib alone (1.3 µM; Fig. 7A). When exposed to romidepsin alone, we found that the viability of Alb-R26Met tumoroids is severely compromised compared with vehicle controls, with an IC50 ranging between 0.0029 and 0.0131 µM (Fig. 7B). tHCC18 is exceptional for its moderate response to romidepsin alone, with a 40% of viability even at high doses (Fig. 7B). In RomiCabo conditions, we found two distinct behaviours. Concerning tumoroids already sensitive to romidepsin alone, the addition of cabozantinib either slightly decreases their viability (tHCC19, tHCC23) or does not provide additional effects (tHCC16, tHCC17, tHCC20, tHCC21, tHCC22; Fig. 7B-D, Table S7). Instead, tHCC18 is more resistant to romidepsin alone, and the addition of cabozantinib significantly reduces their viability (Fig. 7B-D). This is illustrated by: (a) the IC50 of romidepsin and RomiCabo at 0.0131 and 0.0072, respectively; (b) the tHCC18 tumoroid viability that drastically drops in RomiCabo condition, as illustrated by a maximal response to 40% with romidepsin alone and a complete response with RomiCabo (Fig. 7B-D). Additionally, by comparing the response of all tumoroids among them, we found that tHCC18 and tHCC21 are more resistant compared to the others, with statistical differences at distinct drug doses (Fig. 7D, Table S7). Together, these results exemplify how the Alb-R26Met tumoroid panel allows modelling heterogeneous behaviour of HCC in drug sensitivity assessment. Furthermore, our data document the potency of romidepsin, alone or in combination with cabozantinib, strengthening the effectiveness of RomiCabo as a new putative therapeutic approach for HCC treatment.

Fig. 7

Responsiveness of Alb-R26Met tumoroids to romidepsin alone or in combination with cabozantinib. (A) Graph reporting the viability of tumoroids to 1.3 µM of cabozantinib, a dose chosen for its clinical relevance in relation to the average amount of drug found in patient serum, and allowing comparative analyses with outcomes using patient-derived tumoroids (49). Values are compared to non-treated control conditions. Statistical analyses were performed using one-way ANOVA, followed by Dunnett’s multiple comparison. Levels of significance: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001. (B) Graphs reporting the viability of tumoroids at 1.3 µM of cabozantinib plus increasing concentration of romidepsin (0.001, 0.003, 0.01, 0.03 µM). Values are compared to non-treated control conditions. The IC50 values are indicated on each graph. Statistical analyses were performed using two-way ANOVA, followed by Sidak’s multiple comparison. Data were further analysed using Bonferroni correction, showing same significance outcomes. Levels of significance: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001. (C) Brightfield pictures of tumoroids at the indicated conditions. Regarding tHCC18 morphology, we observed some cystic structures developing under cabozantinib (1.3 µM) without and with romidepsin (0.01 µM), possibly related to a dying process in relation to the thinner tumoroid borders, different from cystic tumoroid structures (see as example tHCC20 and tHCC21 in Fig. 6A and S7B). (D) Graph reporting the viability of all tumoroids treated with romidepsin plus cabozantinib to compare the treatment efficacy among them. The IC50 values for each tumoroids are reported. Experiments were performed in ultra-low adherent plates