Introduction

Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide, with a high mortality rate, increasing incidence, and limited treatment options. Development of HCC is complex and typically based on a multi-step process involving chronic hepatitis and hepatic fibrogenesis, eventually resulting in malignant transformation of hepatocytes [1]. A key event in liver fibrogenesis is the activation of hepatic stellate cells (HSCs). HSCs are defined as vitamin A-storing pericytes that constitute 5–8% of all liver cells [2]. HSC activation requires induction of pro-fibrotic genes such as Acta2 (encoding alpha-smooth muscle actin), Col1a1 (encoding collagen type I alpha 1), and Pdgfrb (encoding Platelet-derived growth factor receptor-beta) [3]. This eventually results in the transdifferentiation of resting HSC into collagen-producing, proliferating myofibroblasts [4]. This process also involves cell cycle re-entry and proliferation of the normally quiescent HSCs. Basically, the cell cycle is regulated by cyclins and associated cyclin-dependent kinases (Cdks). The transition from the quiescent state of the cell cycle into the phase of DNA synthesis is thought to be regulated by E-type cyclins Cyclin E1 (CCNE1) and Cyclin E2 (CCNE2) in complex with cyclin-dependent kinase 2 (CDK2) [5]. In this context, we previously demonstrated that Ccne1 gene expression is induced in human and murine liver fibrosis [6]. Accordingly, constitutive, genetic knockout of Ccne1, but also systemic knockdown of Ccne1 using small interfering RNA (siRNA) prevented the onset of liver fibrosis in mice in the established carbon tetrachloride (CCl4) model [6,7,8]. However, the major target cell population mediating the pro-fibrotic effect of Ccne1 has not been clearly identified yet.

An abundance of studies has indicated that HSCs are also an important component of the HCC tumor microenvironment. Very recently, a comprehensive study using state-of-the-art single-cell sequencing technology in combination with cell-type specific genetic modifications revealed dual functions of HSCs in hepatocarcinogenesis, demonstrating that quiescent HSCs rather protect from HCC development, while activated HSCs promote HCC [9].

We have recently shown that Ccne1 and Cdk2 are essential for the initiation of HCC in mice [10]. While we were able to demonstrate that hepatocytes especially require Cdk2 for hepatocarcinogenesis, the effector cells for the oncogenic effects of Ccne1 in the liver have not yet been identified. Here, we tried to answer the question of whether the pro-fibrotic effect of Ccne1 is triggered by a function in HSCs and whether this has an effect on downstream hepatocarcinogenesis. We provide evidence that the inactivation of Ccne1 only in HSCs is sufficient to substantially reduce liver fibrosis but also HCC development in CCl4-mediated murine injury models, which is at least partially dependent on its canonical kinase subunit CDK2.

Results

Genetic ablation of Ccne1, specifically in HSCs, significantly reduces CCl4-mediated liver fibrosis

We generated mice with HSC-specific deletion of Ccne1 (Ccne1ΔHSC) and challenged these animals and cre-negative littermates with repeated CCl4 injections for a period of 6 weeks (Fig. 1A). CCl4 treatment per se resulted in a significant increase in aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activities and relative liver weights. However, we did not detect significant differences between Ccne1f/f and Ccne1ΔHSC mice regarding serum transaminase activities or relative liver mass under basal conditions or after CCl4 treatment (Fig. 1B). The livers of CCl4-treated Ccne1f/f control littermates revealed extensive fibrosis as expected, which was significantly reduced in Ccne1ΔHSC mice as evidenced by histological analysis (Fig. 1C), determination of collagen fiber formation (Fig. 1D, E) as well as lower induction of Acta2 (encoding the HSC-activation marker alpha-smooth-muscle actin, i.e., αSMA) and down-regulation of Col1a1 (encoding collagen type 1 alpha 1) (Fig. 1F). In summary, these data already indicate that Ccne1 performs important pro-fibrotic functions in HSCs.

Fig. 1: Genetic ablation of Ccne1, specifically in HSCs, reduces CCl4-mediated liver fibrosis.
figure 1

A Experimental setup: Mice with HSC-specific deletion of Cyclin E1 (Ccne1ΔHSC; n = 4) and control littermates (Ccne1f/f; n = 6) were injected with CCl4 three times a week, beginning with the age of 8 weeks. After 6 weeks (18 injections in total), mice were euthanized and examined for markers of liver damage and fibrosis progression. Control mice of both genotypes and matching ages were left untreated (ut). B Left: Determination of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activities. Values are given as units/liter (U/l). Right: Liver mass index was calculated as the liver weight (LW): body weight (BW) ratio in percent. C Representative Hematoxylin and Eosin (H&E) stained liver paraffin sections from Ccne1f/f and Ccne1ΔHSC mice after 6 weeks of CCl4-treatment. D Representative Sirius Red stainings from formalin-fixed paraffin-embedded liver sections. E Quantification of liver fibrosis from images in (D). Sirius Red-positive image areas were quantified using ImageJ software and calculated as a percentage of total tissue areas. F Determination of Acta2 (encoding alpha-smooth-muscle actin, left) and Col1a1 (encoding collagen type 1 alpha 1, right) mRNA expression by quantitative real-time PCR (qPCR). Expression values were normalized to the expression of the housekeeping gene Gapdh and calculated as fold induction in comparison to untreated Ccne1f/f mice of the same genetic background and age. Data are expressed as means ± SD. *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.

Full size image

Ccne1 in HSCs drives hepatic fibrosis during HCC development

Liver fibrosis is often considered a precursor to hepatocarcinogenesis [11]. In order to test whether the pro-fibrotic function of Ccne1 in HSCs, as identified above, may also trigger the formation of liver cancer, we applied the N-nitrosodiethylamine (DEN)/CCl4 model [12]. This model combines the pro-inflammatory and pro-fibrotic effects of chronic CCl4 challenge with the hepatocarcinogenic effects of single DEN treatment and well reflects human pathogenesis from inflammatory liver injury via fibrosis to HCC. To this end, 2-week-old Ccne1ΔHSC mice and Ccne1f/f control littermates were injected once with DEN, followed by weekly injections of CCl4 from the 6th to the 24th week of life (Fig. 2A). During this period, WT animals usually develop severe liver fibrosis, dysplastic lesions, and macroscopic HCCs.

Fig. 2: Ccne1 in HSCs drives hepatic fibrosis during HCC development.
figure 2

A Experimental setup: Ccne1ΔHSC mice (n = 15) and Ccne1f/f littermates (n = 12) were subjected to the DEN/CCl4 HCC model. Mice were injected once with DEN on day 14 after birth to induce HCC, followed by weekly administration of CCl4 starting at the age of 6 weeks. Animals were euthanized at the age of 24 weeks for subsequent analysis. B Representative Sirius Red stainings from formalin-fixed paraffin-embedded liver sections. Scale bar: 100 µm. C Morphometric quantification of liver fibrosis. Sirius Red-stained images were analyzed for the percentage of stained image areas. D–F Gene expression analysis of (D). Col1a1, E Acta2 and F Pdfgfb (encoding platelet-derived growth factor receptor beta) by qPCR. Gene expression values were normalized to Gapdh expression and calculated as fold induction in comparison to untreated (ut) Ccne1f/f mice of the same genetic background and age. G Immunohistochemical staining for alpha-smooth-muscle actin (αSMA) of liver sections from DEN/CCl4 treated Ccne1f/f and Ccne1ΔHSC mice. Left: Overview; scale bar: 100 µm. Right: enlarged view of the marked area from left; scale bar: 100 µm. H Morphometric quantification of αSMA positive area from the IHC staining shown in (E) using ImageJ software with IHC toolbox. Data are expressed as mean ± SD. *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.

Full size image

As expected from previous publications, WT (i.e., Ccne1f/f) mice developed pronounced liver fibrosis with excessive hepatic collagen deposition upon combined DEN/CCl4 treatment (Fig. 2B, C), which was accompanied by substantial induction of Col1a1 gene expression and upregulation of Acta2 (Fig. 2D, E). Importantly, the inactivation of Ccne1, specifically in HSCs, was sufficient to significantly reduce hepatic fiber formation, collagen expression, and HSC activation under the same conditions (Fig. 2B–E). In addition, deletion of Ccne1 in HSCs led to significantly reduced expression of the Pdgfrb gene encoding platelet-derived growth factor receptor-beta (PDGFR-β, Fig. 2F), which is also typically expressed in activated HSCs [13]. Altogether, these data demonstrate that Ccne1 is a pro-fibrotic key factor in HSCs which is essential for their activation and the onset of liver fibrogenesis in the setting of inflammatory hepatocarcinogenesis.

We analyzed the role of Ccne1 for HSC activation in more detail by performing in situ stainings of αSMA in livers from DEN/CCl4-treated Ccne1f/f and Ccne1ΔHSC mice, thereby confirming significantly reduced αSMA protein expression and thus HSC activation in Ccne1ΔHSC mice (Fig. 2G, H). Diminished αSMA protein expression through Ccne1 depletion was especially evident in areas of tissue damage, where activated HSCs usually proliferate and produce an extracellular matrix (Fig. 2G, boxed areas). Of note, the inactivation of Ccne1 in HSCs did not influence αSMA expression or collagen deposition in healthy mice under basal conditions (see Fig. 1E–G). In addition, the animals did not show a basal phenotype that could be attributed to the deletion of Ccne1 in HSCs. Only the liver-to-body ratio showed a significant reduction in the untreated control animals (Supplementary Fig. 1A–G).

HSC-specific deletion of Ccne1 in mice leads to a reduced tumor burden during inflammatory hepatocarcinogenesis

We next investigated the impact of HSC-specific Ccne1 deletion for hepatocarcinogenesis in the DEN/CCl4 model. Ccne1ΔHSC mice did not reveal significant changes in AST or ALT activities (Fig. 3A) or in the relative liver-to-body weight ratio (Fig. 3B) after DEN/CCl4 treatment. However, the deletion of Ccne1 in HSCs resulted in a significant reduction of absolute tumor numbers and the cumulative tumor size (Fig. 3C, D). In good agreement, the detailed histological analysis of tumorous changes revealed fewer dysplastic areas in Ccne1ΔHSC mice compared to control animals (Fig. 3E, F). Interestingly, the fibrosis severity significantly correlated with the tumor number if all mice from both experimental groups were considered (Supplementary Fig. 2A). However, a more detailed analysis differentiated according to genotype showed that this correlation is only true for Ccne1ΔHSC mice, but not for Ccne1f/f controls (Supplementary Fig. 2B). This further indicates that specific, Ccne1-driven pro-fibrotic signals also trigger hepatocarcinogenesis.

Fig. 3: HSC-specific deletion of Ccne1 in mice leads to a reduced number and size of dysplastic lesions during inflammatory hepatocarcinogenesis.
figure 3

Ccne1ΔHSC mice (n = 15, blue bars) and Ccne1f/f littermates (n = 12, gray bars) were subjected to the DEN/CCl4 HCC model as illustrated in Fig. 2A. Mice were sacrificed at the age of 24 weeks and analyzed for liver injury, liver histology, and HCC progression. A Determination of AST and ALT activities in Units/liter (U/l). B Determination of liver mass index calculated as the liver weight (LW): body weight (BW) ratio in percent. C Macroscopic appearance of explanted livers from Ccne1f/f and Ccne1ΔHSC mice. Representative liver tumors are highlighted by arrows. Scale bar: 1 cm. D Quantification of tumor burden in Ccne1f/f and Ccne1ΔHSC mice by macroscopic evaluation of HCC nodules. Left: Number of HCC nodules (tumor number); right: cumulative tumor diameter representing a measure of tumor size. E Representative H&E stained liver paraffin sections. Dysplastic liver areas are encircled. F Microscopic evaluation of histologic dysplastic areas displayed in (E). Left: The total number of dysplastic lesions per slide for each mouse is shown. Right: The proportion of dysplastic lesion areas was calculated as the lesion area/total liver area ratio in percent. G Determination of Albumin (Alb), alpha-fetoprotein (Afp), and Prominin1 (Prom1, encoding CD133) mRNA expression by qPCR. Expression values were normalized to the expression of Gapdh and calculated as fold induction in comparison to untreated (ut) Ccne1f/f mice. Data are expressed as mean ± SD. *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.

Full size image

We next examined if the absence of Ccne1 in HSCs may affect the expression of genes encoding albumin (Alb), α-fetoprotein (Afp), and CD133 (Prom1), which are established markers for hepatocytes (Alb), de-differentiated hepatoma cells (Afp) or liver cancer stem cells (Prom1) [14,15,16]. As expected, DEN/CCl4 treatment resulted in reduced Alb gene expression in Ccne1f/f and Ccne1ΔHSC mice in comparison to untreated controls indicating HCC formation and thus de-differentiation of hepatocytes (Fig. 3G). However, this reduction of Alb was significantly lower in Ccne1ΔHSC mice, which is in good agreement with the observed low HCC incidence in these mice. In contrast, Afp was up-regulated to the same extent in both Ccne1f/f and Ccne1ΔHSC mice, while deletion of Ccne1 in HSCs resulted in significantly reduced induction of Prom1 upon DEN/CCl4 treatment (Fig. 3G).

Altogether, our data suggest that the pro-fibrotic function of Ccne1 in HSCs strongly contributes to the process of hepatocarcinogenesis.

Deletion of Ccne1, specifically in HSCs, reduces proliferation and inflammation in the liver and modulates the myeloid tumor microenvironment

Ccne1 is mostly known as a cell cycle regulator controlling the G1/S-phase transition [5]. As liver injury generated by the DEN/CCl4 model primarily induces compensatory proliferation and secondary liver cancer only, we next aimed to investigate whether the anti-fibrotic and anti-carcinogenic effects seen in DEN/CCl4-treated Ccne1ΔHSC mice were also linked to an aberrant proliferation profile in the liver. Following DEN/CCl4 treatment, Ccne1ΔHSC mice revealed a significantly reduced hepatic proliferation compared to controls as determined by Ki-67 protein- and corresponding Mki67 gene expression (Fig. 4A–C). We further characterized the proliferating cell compartments in DEN/CCl4 treated Ccne1f/f mice by co-staining of liver sections with Ki-67 and cell type-specific antibodies for hepatocytes (HNF4α), activated HSCs (αSMA) and immune cells (CD45), respectively. As expected, the majority of proliferating cells were identified as hepatocytes, while we detected rather few proliferating activated HSCs and almost no leukocytes with cell cycle activity (Supplementary Fig. 3A, B).

Fig. 4: Deletion of Cyclin E1, specifically in HSCs, reduces proliferation and inflammation in the liver and modulates the myeloid tumor microenvironment.
figure 4

Ccne1ΔHSC mice (n = 15, blue bars) and Ccne1f/f littermates (n = 12, gray bars) were subjected to the DEN/CCl4 HCC model as illustrated in Fig. 2A. Mice were sacrificed at the age of 24 weeks. Explanted livers were analyzed for markers of proliferation and inflammation. A Representative immunofluorescence stainings for Ki-67 (green) on liver cryosections. Total nuclei are stained with DAPI (blue). The dashed line indicates the border of one representative dysplastic nodule. Scale bar: 100 µm. B Quantification of proliferating liver cells shown in (A) in percent. ut: untreated mice. C–E Gene expression analyses by qPCR. Expression values were normalized to the expression of Gapdh and calculated as fold induction in comparison to untreated (ut) Ccne1f/f mice. C Mki67 gene expression. D Gene expression of interleukin 6 (Il6) and tumor necrosis factor-α (Tnf). E Gene expression of chemokine (C–C motif) ligand 5 (Ccl5) and Ccl2. F Immunofluorescence staining for CD11b (green); total nuclei are stained with DAPI (blue). Scale bar: 100 µm. G Quantification of CD11b positive cells from the images shown in (F). Values are given as a percentage of total cells. H Immune cells were isolated from explanted livers, stained with fluorochrome-labeled antibodies, and then analyzed via flow cytometry (FACS). Percentage of TAM1 (defined as CD11b+, GR1−, MHCII+) and TAM2 (defined as CD11b+, GR1+, MHCII−) cells of CD45+ cells are depicted. Data are expressed as mean ± SD. *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.

Full size image

Regarding the pro-inflammatory response upon DEN/CCl4 challenge, we found a significant up-regulation of interleukin 6 (Il6) mRNA expression in fibrosis- and HCC-bearing Ccne1f/f mice, which was reduced to almost background levels of untreated controls after HSC-specific deletion of Ccne1 in Ccne1ΔHSC mice. In contrast, the expression of the tumor necrosis factor-α gene (Tnf) was not significantly different between Ccne1f/f and Ccne1ΔHSC mice under both conditions (Fig. 4D). Thus, Ccne1 seems to be involved in the control of Il6 expression in the fibrotic and tumorous liver independent of Tnf.

HSCs can secrete the CC chemokine ligands 2 (CCL2) and 5 (CCL5) in response to pro-fibrotic signals, thereby inducing an overall chemotactic activity for monocytes (CCL2) and lymphocytes (CCL5), respectively [17]. In good agreement, Ccl2 and Ccl5 gene expression was induced in Ccne1f/f mice after DEN/CCl4 treatment (Fig. 4E). Of note, inactivation of Ccne1 in HSCs was associated with a tendency towards reduced mRNA expression of Ccl2 (p = 0.065) compared to controls indicating that Ccne1 could be directly or indirectly involved in immune cell recruitment.

In order to study this in more detail, we examined the myeloid cell fractions in the DEN/CCl4 treated liver of Ccne1f/f and Ccne1ΔHSC mice by immunofluorescence stainings and FACS analysis. Overall, we found a significantly decreased number of CD11b+ cells in mice with HSC-specific Ccne1 deletion compared to control littermates (Fig. 4F, G). Moreover, FACS analysis revealed an altered polarization pattern of the mature myeloid (i.e., CD45+/CD11b+/Gr1-) compartment, referred to as tumor-associated macrophages (TAM). In detail, Ccne1ΔHSC livers contained elevated numbers of MHCII-expressing TAMs (TAM1, Fig. 4H), which are considered pro-inflammatory or immune-competent [18]. Concomitantly, we detected a reduced number of MHCII-negative TAMs (TAM2, Fig. 4H) in Ccne1ΔHSC HCC livers, which have been described as anti-inflammatory or immune-suppressive before [18]. This signature of reduced CCL2 signaling, myeloid cell infiltration, and TAM2-polarization has previously been associated with a predominantly anti-tumor microenvironment [17], resulting in an attenuated HCC burden [19]. Moreover, analysis of the lymphoid immune cell compartment by FACS did not show any differences between both groups (Supplementary Fig. 4). Altogether, our data indicate that Ccne1 expression in HSCs affects the myeloid composition of the tumor environment during hepatocarcinogenesis.

Deletion of Cdk2, specifically in HSCs, reduces CCl4-induced hepatic fibrogenesis

E-type cyclins are considered regulatory subunits of CDK2. However, recent findings also point to several CDK2-independent functions of CCNE1, for instance, in the context of hepatocarcinogenesis [10, 20]. To clarify whether the identified pro-fibrotic role of CCNE1 depends on its canonical kinase subunit CDK2, we also subjected mice with HSC-specific deletion of Cdk2 (Cdk2ΔHSC mice) to the CCl4 fibrosis model for a total of 6 weeks (Fig. 5A). Liver transaminase activities (i.e., AST and ALT) and body mass indices were not different between WT control (i.e., Cdk2f/f) and Cdk2ΔHSC mice under basal or treatment conditions, respectively (Fig. 5B). Cdk2ΔHSC mice revealed improved liver architecture compared to control animals after CCl4 treatment (Fig. 5C). Importantly, Cdk2ΔHSC mice showed a significant reduction of hepatic fibrosis compared to WT controls as determined by Sirius Red staining of fibers (Fig. 5D, E) and measurement of HSC activation markers Acta2 and Col1a1 (Fig. 5F, G). Thus, Ccne1 and Cdk2 both mediate pro-fibrotic properties in HSCs. Therefore, these data strongly indicate that the pro-fibrotic function of CCNE1 depends on functional CDK2.

Fig. 5: Deletion of Cdk2 specifically in HSCs reduces CCl4-induced hepatic fibrogenesis.
figure 5

A Experimental setup: Cdk2ΔHSC (n = 8, red bars) mice and Cdk2f/f littermates (n = 7, gray bars) at the age of 8 weeks were challenged with CCl4 three times a week for 6 weeks to induce liver fibrosis. Mice were euthanized 48 h after the last injection and analyzed for liver histology, injury, and fibrosis progression. B Determination of AST and ALT activities in Units/liter (U/l). The liver mass index was calculated as the liver weight (LW): body weight (BW) ratio in percent. C Representative H&E-stained liver paraffin sections from Cdk2f/f and Cdk2ΔHSC mice after 6 weeks of CCl4-treatment. D Representative Sirius Red stainings of liver sections. E Morphometric quantification of liver fibrosis. Sirius Red stained images were analyzed for Sirius Red-positive image areas (percentage of total tissue areas) using ImageJ software. F, G Gene expression analysis of F Acta2 and G Col1a1 by qPCR. Expression values were normalized to the expression of Gapdh and calculated as fold induction in comparison to untreated Cdk2f/f mice. ut: untreated. Data are expressed as mean ± SD. *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.

Full size image

Ccne1 drives pro-fibrotic and pro-carcinogenic properties of HSCs through Cdk2-dependent and Cdk2-independent mechanisms

We next aimed to directly evaluate the impact of Ccne1 and Cdk2 on the pro-fibrotic properties of HSCs. To this end, we isolated primary HSCs from Ccne1ΔHSC and Cdk2ΔHSC mice and studied their in vitro viability and pro-fibrotic gene regulation in comparison to cre-negative control cells for a period of 10 days (Fig. 6A). Of note, the cell number of both primary Ccne1f/f and Cdk2f/f control HSCs decreased over time. However, cell number (and thus survival) of both Ccne1- and Cdk2-depleted HSCs was substantially lower and reached only approximately 50 % of density from matching WT control cells after 10 days of cultivation (Fig. 6B). This suggests that loss of Ccne1 and Cdk2 both results in significantly impaired survival of HSCs. As expected, the cre-mRNA expression could be detected in Ccne1ΔHSC and Cdk2ΔHSC HSCs, but not in floxed control cells (Supplementary Fig. 5A). Accordingly, Ccne1 was undetectable in Ccne1ΔHSC cells, but clearly expressed in Ccne1f/f, Cdk2f/f, and Cdk2ΔHSC HSCs. In addition, we could also verify the lack of Cdk2 expression in Cdk2ΔHSC HSCs (Supplementary Fig. 5B). Immunofluorescence microscopy indicated strongly reduced activation (i.e., αSMA and Ki-67 expression) of Ccne1ΔHSC and Cdk2ΔHSC HSCs compared to WT (i.e., Ccne1f/f, Cdk2f/f) cells (Supplementary Fig. 5C), which was further confirmed by qPCR analysis of Acta2 and Col1a1 gene expression (Fig. 6C). Importantly, we found induction of Il6 gene expression in WT HSCs after 10 days of cultivation, which was reduced almost to background levels by deletion of Ccne1, but not through inactivation of Cdk2 (Fig. 6D). This suggests that Ccne1 is involved in controlling IL6 secretion of HSCs in the pro-carcinogenic liver environment. Moreover, Ccne1 and Cdk2 were both required for the induction of Pdgfrb gene expression in HSCs during spontaneous in vitro activation (Fig. 6E) and for upregulation of the Ccl2 gene (encoding the chemoattractant chemokine CCL2) in HSCs (Fig. 6F). In summary, these experiments clearly demonstrate that Ccne1 is essential for growth, survival, activation and secretory functions of HSCs, which are mostly Cdk2-dependent (activation, PDGFR-β, CCL2), while expression of Il6 is controlled by a novel Cdk2-independent function of Ccne1.

Fig. 6: Ccne1 drives pro-fibrotic and pro-carcinogenic properties of HSCs through Cdk2-dependent and Cdk2-independent mechanisms.
figure 6

A Experimental setup. Primary HSCs were isolated from Ccne1ΔHSC and Cdk2ΔHSC mice as well as from respective cre-negative (i.e., Ccne1f/f, Cdk2f/f) littermate controls. Cells were cultivated for up to 10 days (D0–D10) and analyzed at daily intervals. B Growth analysis of HSCs derived from Ccne1ΔHSC/Ccne1f/f (left panel) and Cdk2ΔHSC/Cdk2f/f mice (right panel). The same section of each culture plate was imaged daily and subjected to cell counting. Relative cell numbers were calculated as percent in relation to the cell number at day 1 of each biological replicate. C–F. Gene expression analysis of C Acta2 (left), Col1a1 (right), D Il6, E Pdgfrb, and F Ccl2. G-M. Concomitant inactivation of Ccne1 and Cdk2 in primary HSCs. G. Experimental setup. Primary HSCs were isolated from Cdk2f/f and Cdk2ΔHSC mice and cultivated as before. Four days after isolation, HSCs were transfected with lipid nanoparticles (LNPs) loaded with either scrambled (scr.) siRNA or Ccne1 siRNA. Cells were harvested after 10 days and characterized by qPCR. H Gene expression analysis of Ccne1 in Cdk2f/f and Cdk2ΔHSC mice after siRNA transfection showed approximately 90% knockdown efficiency of Ccne1 siRNA. I Growth analysis of Cdk2f/f (left panel) and Cdk2ΔHSC HSCs transfected with either scrambled (scr.) or Ccne1 siRNA. Analysis was performed as described for panel (B). J–M Gene expression analysis of J Acta2 (left), Col1a1 (right), K Il6, L Pdgfrb, and M Ccl2. All mRNA expression values were normalized to the expression of Gapdh and calculated as fold induction in comparison to either primary Ccne1f/f HSCs (C–F) or Cdk2f/f HSCs treated with scr. siRNA (H, J–M). Data are expressed as mean ± SD. *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.

Full size image

Next, we were interested to test whether concomitant inactivation of Ccne1 and Cdk2 in HSCs might result in synergistic effects. For this purpose, we encapsulated small interfering RNA (siRNA) specific for Ccne1 in lipid nanoparticles (LNP) and transfected primary Cdk2f/f or Cdk2ΔHSC HSCs as illustrated in Fig. 6G. Using this approach, Ccne1 gene expression was reduced to approximately 90% in both HSC groups (Fig. 6H). As a control, cells were transfected with a non-coding scrambled siRNA. In agreement with the data from genetic Ccne1 inactivation (compare Fig. 6B), knockdown of Ccne1 by siRNA in WT (i.e., Cdk2f/f) HSCs also resulted in a significant reduction of cell number over time. However, concomitant inhibition of Ccne1 and Cdk2 in Ccne1 siRNA treated Cdk2ΔHSC HSCs had no significant effect on this reduced viability (Fig. 6I). Concomitant inactivation of Cdk2 and Ccne1 reduced gene expression levels of Acta2 and Col1a1 in HSCs to the same extend as deletion of Cdk2 alone (Fig. 6J). Importantly, these experiments confirmed that inhibition of Ccne1 triggers down-regulation of Il6 gene expression in a Cdk2-independent manner, which could be achieved by Ccne1 siRNA delivery in both Cdk2f/f or Cdk2ΔHSC HSCs (Fig. 6K). Finally, combined inactivation of Cdk2 and Ccne1 down-regulated Pdgfrb and Ccl2 gene expression to the same extent as deletion of Cdk2 alone or Ccne1 siRNA delivery alone (Fig. 6L, M). Taken together, these experiments demonstrate that Ccne1 and Cdk2 jointly control the expression of several pro-fibrotic genes in HSCs, whereas Il6 expression is regulated by Ccne1 in a Cdk2-independent manner.

We finally aimed to evaluate the expression of Ccne1 and related cell cycle genes in resting and activated HSCs during fibrogenesis on a single cell level. To this end, we took advantage of single-cell RNA sequencing (scRNAseq) data that we had generated before [21]. In that study, activated HSCs (i.e., myofibroblasts; MFB) were isolated from WT mice after chronic treatment with CCl4 for 3 weeks and subjected to scRNAseq. As a control, resting HSCs were isolated from untreated mice and processed accordingly. Due to different expression profiles, the cells were separated into five sub-clusters referred to as MFB I to MFB IV as well as resting HSC (Fig. 7A). We analyzed these data with respect to G1/S phase cyclins and Cdks. As our present experiments demonstrated that Il6 is regulated by Ccne1 in HSCs, we also investigated Il6 expression in the scRNAseq data set.

Fig. 7: Identification of activated HSCs with an expression of Ccne1 and Cdk2 after chronic CCl4 treatment by single-cell RNA sequencing (scRNAseq).
figure 7

Expression analysis of cell cycle-related genes from single-cell sequencing data (scRNAseq) of resting and activated HSCs (i.e., myofibroblasts; MFB). A Left: Subset clustering of resting HSCs (HSC) and four different MFB sub-clusters referred to as MFB I–MFB IV from scRNAseq datasets serving as an overview. Data were adapted from [21]. Right: Heatmap showing average expression of G1/S-phase genes, Mki67 and Il6 for each cluster. B t-SNE plots show the relative gene expression strength of selected cell cycle genes and Il6. C Analysis of Ccne1 co-expression with Mki67 and Il6 expression on single cell level. Upper panel: Expression of Ccne1 is associated with Mki67 expression in MFB I and MFB III clusters. Detection (>0 reads per cell) for Ccne1 and Mki67 were assessed for each cluster. Numbers in each tile denote numbers of cells; d: detected; nd: not detected; color shows Chi-Square-residuals of observed versus expected counts. P-values represent a Chi-square test of independence between the detection of the two genes. Lower panel: Expression of Ccne1 is slightly associated with Il6 expression solely in cluster MFB I. Detection (>0 reads per cell) for Ccne1 and Il6 was assessed for each cluster. Color code and calculation were performed as above.

Full size image

Cell-cycle-dependent expression of Ccne1 is typically restricted to the G1/S-phase transition [22, 23]. However, in resting HSCs or MFBs, Ccne1 expression was only detected in 2.2% of all cells. Ccne1 was mostly expressed in cells belonging to the MFB I cluster (Fig. 7B, C). Similarly, Cdk2 (detected in 4.4% of all cells), Ccne2 (detected in 1.3% of all cells), and Ccna2 (detected in 1.6% of all cells) were also, at best, slightly expressed in only a small proportion of HSCs and MFBs (Fig. 7B, C). In good agreement, the number of Mki67-expressing cells (3.1%) was in a similar range in MFBs suggesting that the percentage of proliferating MFBs during fibrogenesis at a distinct time point is rather moderate. The expression profile of each individual cell with respect to cell cycle genes is depicted in Supplementary Fig. 6A, B. Surprisingly, we detected a strong gene expression of D-type cyclins and their interacting kinases Cdk4/Cdk6 throughout all HSC/MFB clusters, while Cdk1 was rather low expressed (Fig. 7A and Supplementary Fig. 6A–C).

We next assessed the expression of Ccne1 and Mki67 for each cluster and found a significant association of both genes in MFB I and MFB III (Fig. 7C, upper panel), suggesting that Ccne1 mediates also in MFBs cell cycle-dependent functions. Il6 expression was detected in only 8.4 % of the cells throughout cluster MFB I–MFB IV (Fig. 7B), lacking significant association with Ccne1 expression except for MFB II where, however, Ccne1 expression was mostly absent (Fig. 7C, lower panel).

In summary, the scRNAseq data suggest that the strong effects of the HSC-specific Ccne1 deletion regarding liver fibrogenesis and HCC development are caused by a small number of actively cycling HSCs and MFBs. The data further implicate that the effects of Ccne1 on Il6 expression in HSCs might be indirect, as a significant co-expression in the same cells could not be verified.

Discussion

HCC is primarily characterized by inappropriate, excessive proliferation of transformed hepatocytes/hepatoma cells. However, besides intrinsic aberrations within hepatocytes, such as de-regulation of cell cycle activators (e.g., Cyclins) or mutations of tumor suppressors (e.g., p53), also extrinsic signals from the tumor environment can contribute to HCC progression. The HCC tumor environment typically consists of cellular components, including tumor-associated macrophages, endothelial cells, immune cells, and activated HSCs, as well as non-cellular components, such as growth factors and extracellular matrix proteins [24]. In previous work, we have demonstrated that the cell cycle mediator Ccne1 is essential for both initiation of liver fibrosis and HCC [6, 7, 10]. However, so far, we either constitutively deleted Ccne1 in all cells or systemically inhibited Ccne1 by siRNA, thereby targeting at least hepatocytes, HSCs, CD45+ cells, and endothelial cells to prevent fibrogenesis [7]. With regard to HCC, we have so far demonstrated that ubiquitous genetic inactivation of Ccne1 or hepatocyte-specific deletion of its interacting kinase subunit Cdk2 inhibited HCC initiation [10], while interventional depletion of Ccne1, but not of Cdk2, in hematopoietic cells, hepatocytes and lymphocytes ameliorated the progression of HCC [20]. Thus, our previous data does not exclude any of the main hepatic cell types from being a Ccne1-dependent driver of liver fibrosis or HCC.

Here, we tested the hypothesis that HSCs are important effector cells for Ccne1-driven liver fibrosis and hepatocarcinogenesis. To this end, we generated HSC-specific Ccne1 knockout mice and challenged these animals with a pure liver fibrosis model (CCl4) as well as with a model reflecting both liver fibrogenesis and hepatocarcinogenesis (DEN/CCl4). Our study revealed that deletion of Ccne1 only in HSCs substantially reduced liver fibrosis in both experimental settings, which was associated with reduced HSC activation. Importantly, the lack of Ccne1 in HSCs also ameliorated HCC initiation and progression. Mechanistically, we demonstrated that the pro-fibrotic effect of Ccne1 in HSCs is dependent on its associated kinase Cdk2. Finally, deletion of Ccne1 in HSCs was associated with reduced expression of genes encoding important signaling molecules such as IL-6, PDGFR-β, and CCL2, which could be an additional key to explain the pro-fibrotic properties of Ccne1.

Thus, our study demonstrates that Ccne1 is an important modulator of HSCs in the course of liver fibrosis and cancer development. This is especially remarkable as HSCs represent only 5–8% of total liver cells [2], and only a few of them (approximately 2%) showed Ccne1 expression at a distinct point in time in our scRNAseq analyses when chronically challenged with CCl4. It must be pointed out that Ccne1 is only expressed at the G1/S-phase transition and is degraded immediately afterward [25]. The duration of the complete S-phase in HSCs is not precisely known; however, in rat hepatocytes, it has been estimated that S-phase lasts approximately 7 h [26]. Thus, our scRNAseq measurements regarding Ccne1 may reflect only a snapshot of (non-synchronized) HSCs that underwent S-phase initiation exactly at the moment of isolation. It is, therefore, very likely that in a longer observation period, many more HSCs will express Ccne1 once. The number of Ccne1-expressing cells detected by scRNAseq was very similar to the number of Cdk2-expressing HSCs and correlated with Mki67-expressing HSCs, implicating that one function of Ccne1 is indeed to drive the HSC cell cycle, in a Cdk2-dependent manner. This is further supported by our observations in primary HSCs confirming that proliferation and survival of HSCs require both Ccne1 and Cdk2. Our current hypothesis on the role of Ccne1 in HSCs is schematically shown in Fig. 8. Accordingly, poor survival and reduced proliferation capacity of HSCs due to Ccne1 or Cdk2-deficiency is one possible mechanism to explain the substantial reduction of liver fibrosis in our study (Fig. 8A). Weakened fibrosis per se in combination with a reduced absolute number of activated HSCs could subsequently result in a modulation of the liver tissue environment that will act less tumorigenic upon a secondary oncogenic stimulation such as DEN treatment. Our conclusion is in perfect agreement with the very recent study of Filliol et al. [9], which shows in a variety of extensive experiments that activated HSCs overall have a tumor-promoting role in mouse models of HCC, including the DEN/CCl4 model. In this context, the authors of that study convincingly demonstrated that the depletion of HSCs in mice resulted in suppression of the DEN/CCl4-mediated HCC development [9]. Thus, it is logical that the reduction of HSC survival and activation (as observed in Ccne1ΔHSC mice) should also result in reduced HCC load after DEN/CCl4-treatment. In addition, our data suggest that gene expression/transcription of several pro-inflammatory/pro-carcinogenic signaling molecules is directly controlled by Ccne1/CCNE1 (Fig. 8B), and this effect is either Cdk2-dependent (e.g., in case of Pdgfrb and Ccl2, Fig. 8C) or Cdk2-independent in case of Il6 gene expression (Fig. 8D). Of note, this finding has not only been proven in a genetic approach (using gene deletion) but also in knockdown experiments using LNP-encapsulated Ccne1 siRNA, pointing to therapeutic applicability of our results.

Fig. 8: Proposed role of Ccne1 and Cdk2 for liver fibrogenesis and fibrosis-induced HCC.
figure 8

A Our data demonstrate that Ccne1 and Cdk2 are both required for the differentiation, proliferation, and survival of HSCs. Accordingly, the inactivation of Ccne1 or Cdk2 in HSCs reduces the pro-fibrotic properties of HSCs, thereby attenuating fibrosis formation upon chronic injury such as the CCl4 challenge. In turn, attenuated fibrosis may impact the tissue environment resulting in less liver tumor formation. B Ccne1 in HSCs triggers liver fibrogenesis or HCC formation directly through transcriptional control of pro-fibrotic or hepatocarcinogenic signaling molecules. C The latter mechanism can be Cdk2-dependent by driving the expression of genes such as Pdgfrb (encoding PDGF-Rβ) and Ccl2 (encoding CCL2). D We demonstrated that Ccne1 is involved in transcriptional control of Il6 in a Cdk2-independent manner, which is likely to contribute to HCC formation.

Full size image

The mechanism by which CCNE1 alone or in combination with CDK2 controls the transcription of pro-fibrotic genes is unclear to date. However, there is overwhelming evidence that CCNE1/CDK2 can indeed control the transcription of target genes via several different mechanisms, as outlined elsewhere in detail [27]. Already as part of its canonical function, CCNE1/CDK2 enables the transcription of E2F-target genes by phosphorylation of the retinoblastoma protein RB. More recent studies suggest that CCNE1/CDK2 is also involved in the modulation of several other transcriptional regulators, such as Id2 [28], Myc [29], or NF-Y [30] through phosphorylation. We, therefore, conclude that CCNE1 alone or in complex with CDK2 can, in general, act as a transcriptional co-factor, and we postulate that this may also be true for the control of some pro-fibrotic genes, although the exact targets of this mechanism have to be evaluated in extensive future studies.

PDGF is the most potent pro-fibrotic factor associated with HSC proliferation and activation. Overexpression of its receptor PDGFR-β in the liver can be linked with enhanced fibrogenesis, steatosis, and HCC in mice [31]. In our study, HSC-specific deletion of Ccne1 was associated with the downregulation of Pdgfrb gene expression in the DEN/CCl4 model in vivo but also in primary HSCs in vitro. This suggests that the downregulation of Pdgfrb in the DEN/CCl4-treated liver of Ccne1ΔHSC mice is best explained by a direct function of Ccne1 on Pdgfrb transcription and not by a simple reduction of the absolute HSC number in the injured liver. Of note, Lrat-Cre-mediated deletion of Pdgfrb inhibited HSC activation, liver fibrosis, and tumorigenesis in Mdr2−/− mice in the study of Filliol et al. [9], which further suggests that down-regulation of PDGFR-β transcription by Ccne1 inhibition is mechanistically related to reduced fibrogenesis and hepatocarcinogenesis in our own experimental setting.

Furthermore, our data demonstrated that also the chemokine CCL2 is transcriptionally regulated by Ccne1 in a Cdk2-dependent fashion both in DEN/CCl4-treated liver and in primary HSCs. CCL2 regulates the migration and infiltration of monocytes/macrophages through combination with its specific receptor CCR2 and has a vital role during liver fibrogenesis [32]. Downregulation of Ccl2 in Ccne1ΔHSC mice could explain the reduced numbers of CD11b+ cells observed in these mice. These results indicate that Ccne1 in HSCs can modulate the tumor microenvironment through CCL2 and also suggests that the interaction of HSCs and TAMs may have an important impact on tumor development and progression, as proposed recently [32]. Our study also revealed that Ccne1 regulates Il6 gene expression in HSCs independent of Cdk2. It has been previously shown that IL6 is essential for proper DEN-induced hepatocarcinogenesis in mice, and Kupffer cells within the liver have been identified as a major source of HCC-inducing IL6 [33]. However, a recent paper detected considerable IL6 expression also in HSCs. In turn, HSC-derived IL6 was able to promote HCC progression in mice through the induction of myeloid-derived suppressor cells (MDSCs), thereby modulating the tumor environment. Of note, the ablation of Il6 gene expression in HSCs substantially attenuated HCC growth in that study [34]. In our work, we detected reduced Il6 expression in the fibrotic liver of DEN/CCl4-treated Ccne1ΔHSC mice, but also in Ccne1-deficient HSCs in vitro, while ablation of Cdk2 in HSCs had no significant effect on Il6 regulation. We, therefore, conclude that the down-regulation of Il6 through Ccne1-depletion in HSCs is significantly involved in the reduced tumor burden of DEN/CCl4-treated Ccne1ΔHSC mice.

Our scRNAseq analyses detected Il6 gene expression only in a few HSCs following CCl4 treatment. Of note, the frequency of Il6-expressing HSCs was very similar to the number of HSCs with detectable Ccne1 expression. However, significant co-expression of Ccne1 and Il6 could at a distinct time point was almost undetectable on a single cell level. We, therefore, assume that Ccne1 induces Il6 in HSCs either with a time delay or indirectly involving cell-cell communications.

Our final conclusion is summarized in Fig. 8: Ccne1 in HSCs is an important modulator of fibrosis-driven HCC through at least 3 mechanisms by (i) enabling proliferation and survival of HSCs, (ii) controlling Cdk2-dependent expression of oncogenic signaling molecules, and (iii) driving Cdk2-independent Il6 expression. Of note, ubiquitous deletion of Ccne1 was similarly effective in preventing liver fibrosis [6] but much more potent in inhibiting HCC initiation [10] when compared to HSC-specific Ccne1 inhibition. Thus, HSCs are the main effector cells for the pro-fibrotic function of Ccne1, but presumably only one out of several effector cells for the oncogenic function of Ccne1. Nevertheless, our data strongly indicate that therapeutic Ccne1 inhibition will be beneficial for the treatment of both liver fibrosis and fibrosis-induced HCC.

Material and methods

Maintenance of mice and general animal experimentation

Animals were housed in a temperature-controlled room with 12-h light/dark cycles and free access to food and water. For this study, we crossed conditional Ccne1 (Ccne1f/f) or Cdk2 (Cdk2f/f) knockout mice [35, 36] with transgenic mice expressing Cre-recombinase under the control of the L-rat promoter [37], resulting in mice lacking Ccne1 or Cdk2 specifically in HSCs (Ccne1∆HSC, Cdk2∆HSC). For experiments, we used male mice in a C57BL/6 background; floxed (f/f), cre-negative littermates served as controls.

For induction of liver fibrosis, 8-week-old mice were injected with CCl4 (0.5 ml/kg i.p., diluted in corn oil) three times a week for 6 consecutive weeks. For induction of liver fibrosis and HCC, we applied the DEN/CCl4 model in which fibrosis is induced by a single injection of N-nitrosodiethylamine (DEN) followed by repeated administration of carbon tetrachloride (CCl4) following a protocol previously described [12] with minor modifications. Briefly, 14-day-old pups were injected with DEN (25 mg/kg, i.p.), followed by weekly injections of CCl4 (0.5 ml/kg, diluted in corn oil) for 18 weeks, starting at the age of 6 weeks. All animals were euthanized 48 h after the last CCl4 injection by cervical dislocation.

Tumor quantification

Livers were removed and analyzed for the presence of visible tumors. Tumor-bearing livers were documented using a Leica MZ16 microscope together with Diskus software (Hilgers, Königswinter, Germany). Separation into individual lobes was followed by measuring the diameter of visible surface nodules in millimeters. Intrahepatic tumors were identified and counted on Hematoxylin & Eosin (H&E) stained liver sections. Each liver section was completely scanned using an Axio Observer.Z1 microscope (Carl Zeiss, Jena, Germany), followed by quantification of all tumor nodules and of the tumor-free area using Image J software (http://rsbweb.nih.gov/).

Isolation and cultivation of primary HSCs

Primary mouse HSCs were isolated from 28 to 35-week-old mice by collagenase perfusion as described recently [6] with the modification that animals were sacrificed before cell isolation. After isolation, living cells were plated on 12-well cell+ plates (Greiner, Kremsmünster, Austria) at a density of 30,000 cells per 12-well in DMEM medium (PAA Laboratories GmbH, Pasching, Austria) supplemented with 10% FCS and 100 U/ml penicillin/streptomycin. After 4 h, cells were attached, and dead cells were removed by washing with PBS and medium exchange. Each day, the same area of the plate was imaged using an Axio Imager.Z1 microscope (Carl Zeiss) and analyzed for cell density.

Generation of lipid nanoparticles and transfection of primary HSCs with siRNA

The used siRNAs were ordered from IDT (Leuven, Belgium) in a lyophilized form and resolved in 0.1 M acetate buffer. Sequences of used siRNAs are listed in Supplementary Table 2.

For generation of LNPs, the lipids Cholesterol, 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC), and 1,2-Dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG 2000) were acquired from Sigma Aldrich (St. Louis, MO). Heptatriconta-6,9,28,31-tetraen-19-yl-(dimethylamino)butanoate (DLin MC3-DMA, MC3) was obtained from Hycultec (Beutelsbach, Germany). Lipids were resolved in absolute ethanol at the molar ratios of 50:10:38.5:1.5 (MC3:DSPC:Cholesterol:DMG-PEG2000). LNP were generated using a NanoAssemblr Spark (Precision Nanosystems, Vancouver, Canada) according to the manufacturer’s instructions. For larger-scale production, we performed LNP generation using two Ph.D. Ultra syringe pumps (Harvard Apparatus, Holliston, MA) connected to a Herringbone microchip (Microfluidics ChipShop, Jena, Germany) and mixing was done at a flow rate ratio of 3:1 (siRNA:lipids). LNP were characterized for size using dynamic light scattering (DLS) and surface charge using Zetapotential (Malvern Zetasizer, Malvern, UK). A Ribogreen assay was done to quantify the RNA content of LNP, as published before [38].

For transfection, primary HSCs were seeded on Petri dishes immediately after isolation and cultivated for 4 days. Immediately before transfection at day 4, HSCs were washed once with PBS and supplied with fresh medium containing siRNA at a concentration of 1 µg/ml encapsulated in LNPs as described above. Transfected HSCs were then cultivated for another 6 days without further treatment.

Single-cell RNA sequencing (scRNAseq) and bioinformatics data analysis

For analysis of HSC gene expression on a single cell level, we used recently published data [21]. In the underlying experiments, wild-type (WT) mice were treated 3 times a week with CCl4 for 3 weeks and sacrificed 36 h after the last treatment. Activated HSCs (i.e., myofibroblasts; MFB) were isolated from these animals and subjected to scRNAseq. As a control, resting HSCs were isolated from untreated mice and processed accordingly. Cells were analyzed by using the Chromium Single Cell 50 kit (10× Genomics, Pleasanton, CA, USA), according to the manufacturer’s protocol. After library generation, sequencing was performed by Illumina sequencing on a NextSeq 550. The primary analysis was done by using an in-house pipeline based on “cellranger” (10× Genomics). Additional analysis was then performed by using the “Seurat” (v2.3.2) package for R (v3.5) (https://www.r-project.org/).

Statistical analysis

For statistical analysis of the results, Graph Pad Prism 8 (Graphpad Software, Inc., USA) was used. All results were first tested for normal distribution using the Shapiro-Wilk test. Comparison of two groups was performed by an unpaired, two-tailed t-test in case of a normal distribution or by Mann Whitney test in case of non-normal distributed results. Comparison of three or more groups differing in one variable (e.g., genotype) was performed using ordinary one-way ANOVA with Tukey’s multiple comparisons test. In the case of groups differing in two variables (e.g., genotype and treatment), two-way ANOVA with Tukey’s multiple comparisons test was performed. All data were presented as mean ± standard deviation (SD) using combined dot plots with bars to indicate values of individual mice or biological replicates. Significances were defined as *p ≤ 0.05; **p ≤ 0.005; ***p ≤ 0.001.