Control of liver size by RNAi-mediated multiplex knockdown and its application for discovery of regulatory mechanisms

Abstract

Background and aims

The Hippo pathway controls organ size through a negative regulation of the transcription co-activator Yap1. The overexpression of hyperactive mutant Yap1 or deletion of key components in the Hippo pathway leads to increased organ size in different species. Analysis of interactions of this pathway with other cellular signals corroborating organ size control is limited in part due to the difficulties associated with development of rodent models.

Methods

Here, we develop a new model of reversible induction of the liver size in mice using siRNA-nanoparticles targeting two kinases of Hippo pathway, namely, mammalian Ste20 family kinases 1 and 2 (Mst1 and Mst2), and an upstream regulator, neurofibromatosis type II (NF2).

Results

The triple siRNAs nanoparticle-induced hepatomegaly in mice phenocopies one observed with Mst1-/- Mst2-/- liver-specific depletion, as shown by extensive proliferation of hepatocytes and activation of Yap1. The simultaneous co-treatment with a fourth siRNA nanoparticle against Yap1 fully blocked the liver growth.

Hippo pathway-induced liver enlargement is associated with p53 activation, evidenced by its accumulation in the nuclei and upregulation of its target genes. Moreover, injections of the triple siRNAs nanoparticle in p53^LSL/LSL mice shows that livers lacking p53 expression grow faster and exceed the size of livers in p53 wild type animals, indicating a role of p53 in controlling Yap1-induced liver growth.

Conclusion

Our data show that siRNA-nanoparticulate manipulation of gene expression can provide the reversible control of organ size in adult animals, which presents a new avenue for the investigation of complex regulatory networks in liver.

Introduction

Maintenance of proper organ size in an organism is a fundamentally important process. The proper function of organs depends on coordinated control of their size during developmental and adult stages. In contrast, loss of organ size control contributes to a number of diseases, including hypertrophy and degenerative diseases(1). Organ size control integrates various factors, including local mechanical, autocrine/paracrine stimuli as well as soluble circulating cues and environmental factors(1-3).

One of the cues critical in controlling organ size and cellular proliferation is the Hippo signaling pathway (Reviewed in(4, 5)). This pathway is composed of an evolutionarily conserved core kinase cassette and upstream modulators (reviewed in(6)). In mammals, the core kinases consist of mammalian Ste20 family kinases 1 and 2 (Mst1 and Mst2), Salvador homolog 1 (Sav1), Large tumor suppressor 1 and 2 (Lats1 and Lats2), and Mps one binder 1 (Mob1)(4, 5). The core cascade is tightly regulated by multiple upstream modulators(6). Neurofibromatosis type II (NF2, also known as Merlin) can activate the core kinase cassette(7), likely through direct binding and recruiting Lats1/2 to the plasma membrane(8). The Hippo pathway negatively regulates its major downstream effector Yes-associated protein 1 (Yap1) through phosphorylation, provoking its degradation and cytoplasmic retention(9, 10). As a potent transcription co-activator, Yap1 can induce genes involved in cellular growth and apoptosis inhibition by association with the TEAD family and other transcription factors(9, 11, 12). Overexpression of S127 mutant Yap1 in the adult liver causes massive hepatomegaly(10). It has been reported that while one allele of either Mst1 or Mst2 is sufficient to maintain embryonic development, Mst1-/-Mst2-/- mice are embryonically lethal (13-15). One copy of either Mst1 or Mst2 can maintain quiescence of hepatocytes, but liver-specific genetic removal of both Mst1 and Mst2 leads to significant liver enlargement(14-16).

Hippo signaling has been shown to interact with several other pathways, including PI(3)K–mTOR(17), Wnt/beta-catenin(18), Insulin/IGF(19, 20), and Sonic hedgehog (Shh) signaling pathway(21) in normal tissue and tumorigenesis. However, deciphering its complex interactions in vivo requires sophisticated approaches allowing to gradually and simultaneous manipulate multiple genes.

Recent advances in synthetic siRNA delivery nanoparticles make it possible to specifically suppress one or more genes simultaneously in a range of species, from rodents to primates(22-25), including humans(26). Through manipulating the compositions of nanoparticles, siRNA delivery systems have shown potent and reversible silencing effects in vivo with high specificity in multiple tissues and cell types, including hepatocytes(22, 23), macrophages(27) and endothelial cells(28). Here we develop a siRNA nanoparticle-based approach to manipulate organ size through inhibition of the Hippo pathway. The deep reduction of the expression levels of Mst1/Mst2/Nf2 is critical to elucidate the role of p53 pathway in the control of liver growth.

Materials and Methods

siRNA synthesis, screening and lipid nanoparticles (LNP) formulation

siRNAs targeting mouse Mst1, Mst2, NF2, Yap1 were designed and screened as previously described(24). The sequences and IC₅₀ values of each siRNA are provided in Supplementary Table 1. The RNA strands were synthesized, characterized and duplexed by Alnylam Pharmaceuticals as previously described(24). Individual siRNAs were formulated into lipid nanoparticles and mixed prior to injections(23).

Cell culture

Hepa1-6, NIH3T3, AML-12 cells were obtained from ATCC and were propagated in DMEM supplemented with 10% FBS. Cells were transfected with siRNA using either Lipofectamine RNAiMAX (Invitrogen) or LNP with siRNA as described elsewhere(24).

AML-12 cells obtained from ATCC were grown on BD BioCoat collagen I coated plasticware. Cells were treated with taurocholic acid (Sigma-Aldrich, time and dose as indicated). Cells were collected in RIPA buffers supplemented with proteases and phosphatases inhibitors (Pierce Bio) for western blot, or fixed with 2% buffered paraformaldehyde solution, followed by permeabilization with 1% Triton X-100 in PBS for immunocytochemistry.

Animals

C57BL/6 mice were purchased from Charles River laboratories. p53^LSL/LSL mice were published elsewhere(29). All animals received humane care, and animal protocols were approved by the Committee on Animal Care at MIT and the Institutional Animal Care and Use Committee of Alnylam Pharmaceuticals, certified by the American Association for Accreditation of Laboratory Animal Care. After 3 days of acclimatization in the animal facility, 7-9 week-old mice were injected via tail vein (i.v.) with either PBS or siRNA in LNP formulations at various concentrations. To restore p53 expression in p53^LSL/LSL mice, animals were treated with tamoxifen (two intraperitoneal doses) before the siRNA treatment. Animals were sacrificed by CO₂ overdose; tissues were harvested at different time points as indicated. Hepatocytes, stellate cells and Kupffer cells were isolated from C57BL/6 mice via collagenase perfusion, density centrifugation and antibody selections as described previously(30-32).

Histological, immunohistological and immunocytochemical analysis

Mouse tissues were fixed in 4% paraformaldehyde. Tissue sections were stained according to standard immunohistochemistry protocols as previously described(33) or with secondary antibodies labelled with Alexa 488, Alexa 555 and Alexa 647 (Invitrogen) to visualize antigen localization. We have used the following primary antibodies: anti-Cytokeratin 18 (Progen), anti-Cytokeratin 19 (Abcam), anti-E-cadherin (BD Biosciences), anti-F4/80 (Biolegends), anti-glutamine synthase (BD Biosciences), anti-Ki67 (Neomarkers), anti-p21 (Santa Cruz), anti-p53 (Leica Biosystems), anti-Yap1 (Cell Signaling), phospho-pH2A.X (Cell Signaling). Phalloidin coupled with Alexa555 was used to visualize F-actin. Analysis of H&E, immunohistochemistry and immunocytochemical images was performed using ImageJ package (NIH).

Western blots and quantification

The liver tissues were homogenized in RIPA buffer (Thermo Fisher) to harvest proteins. Total protein was resolved on TGX gradient gels (BioRad). The following primary antibodies were used: anti-β-actin (Sigma), anti-a-catenin (Cell Signaling), anti-cyclin D1 (Millipore), anti-Mst1 (Cell Signaling), anti-Mst2 (Cell Signaling), anti-NF2 (Sigma), anti-p21 (Santa Cruz), anti-pYap1 (Cell Signaling), anti-Yap1, phospho-pH2A.X. (Abcam) Licor Odyssey Imaging system was used to visualize protein bands. Gray scale images were quantified with ImageJ as described previously(24).

Gene expression analysis

The levels of mRNAs were measured by branched DNA assay (Affymetrix) or quantitative PCR using TaqMan probes and Roche LightCycler 480. Levels of mRNA of the genes-of-interest were normalized to the levels of GAPDH mRNA and then to the average arbitrary value of the control group.

Serum chemistry

Albumin, total protein, glucose, HDL, LDL, direct and total bilirubin, AST, ALT, total bile acids were measured in serum using Beckman Coulter reagents and Olympus Au400 autoanalyser.

Statistical analysis

P values were calculated using the Student's t-tests and the One-Way ANOVA in Prism 5 (GraphPad).

Results

siRNA mediated knockdown of Mst1, Mst2 and NF2 leads to hepatomegaly

We designed and screened sets of chemically modified siRNAs(34) targeting Mst1 and Mst2 (Supplementary Fig. 1). Specifically, 27 siRNAs targeting Mst1 and 28 siRNAs targeting Mst2 were synthesized. The gene silencing efficiency for each siRNA was examined in NIH3T3 cells at 5nM (Supplementary Fig. 1a, c). We found that nine siRNAs targeting Mst1 and nine siRNAs targeting Mst2 can achieve deep knockdown at this dose (Supplementary Fig. 1a, c). A dose response study was performed using the 3 most potent siRNAs for each gene. The IC₅₀ for the most potent siRNA targeting Mst1 and Mst2 were ~0.05nM and 0.2nM, respectively (Supplementary Fig. 1b, d). The most efficient siRNAs were chosen and formulated into lipid nanoparticles (LNP), which have shown hepatocyte specific targeting with high potency(23). Approximately 90% knockdowns of both Mst1 and Mst2 mRNAs were shown in hepatocytes isolated from mice with treatment of LNP-formulated siRNAs against Mst1 (termed as si-Mst1) or Mst2 (termed as si-Mst2), respectively (Supplementary Fig. 2). Animals treated with si-Mst1 alone, si-Mst2 alone, or their combination showed no liver growth (Supplementary Fig. 3). PBS and LNP-formulated luciferase siRNA (termed as si-Control) treated mice served as controls. This result suggested that the remaining expression of Mst1 and Mst2 (~10%) was sufficient to maintain Yap1 inactivation. To further inhibit the Hippo pathway, specific siRNAs against NF2, which is a key protein for maintaining activity of this pathway(7, 8), were developed and screened (Supplementary Fig. 4). We synthesized 27 siRNAs targeting NF2. The gene silencing efficiency for individual siRNA was examined, and five efficient siRNAs were identified (Supplementary Fig. 4a). The IC₅₀ for the most potent siRNA targeting NF2 was ~0.02nM (Supplementary Fig. 4b). LNP-formulated siRNA against NF2 (termed as si-NF2) showed efficient knockdown in vivo (Supplementary Fig. 5). Animals treated with a combination of si-Mst1, si-Mst2 and si-NF2 twice per week for two weeks at the dose of 0.67 mg/kg for each siRNA (hereafter, designated as triple siRNAs nanoparticle) showed 2.5 folds induction of liver size at Day 15 after the first injections of triple siRNAs nanoparticle (Fig 1). Similar results were obtained with alternative siRNAs targeting Mst1/Mst2/NF2 (Supplementary table 1, and data not shown).

Figure 1. Treatment of lipid-based nanoparticles (LNP)-formulated siRNAs targeting Mst1, Mst2 and NF2 (Triple siRNAs) induced hepatomegaly in mice.

C57BL/6 female mice were treated with triple siRNAs, dose matched LNP-siRNA against the luciferase gene (si-Control), or PBS as controls. Livers were taken 15 days after the first injection. a, Macroscopic views of livers. Scale bar is equal to 1 cm. b, Maximal induction of liver size in triple siRNA treated mice. Error bars represent standard error (SEM), p<0.01.

To characterize this model, different combinations and doses of these siRNAs were examined. Combinations of si-Mst1 and si-NF2, or si-NF2 alone did not induce growth, while the combination of si-Mst2 and si-NF2 induced only minor liver growth (Supplementary Fig. 6a). Once we confirmed that all three siRNAs were necessary for induction of the hepatomegaly, we performed dose titration for each siRNA in vivo (Supplementary Fig. 6b-d). Based on these results, we optimized the dose for triple siRNA treatment as 0.3 mg/kg for si-Mst1, 0.6 mg/kg for si-Mst2 and 0.2 mg/kg for si-NF2 (Supplementary Fig. 6e). Our experiments described below used this combined dose.

We further characterized the time course of liver growth induced by siRNA. We found that the growth started at Day 9 and peaked at Day 15 after the first injection of triple siRNAs nanoparticle (4 injections in total) (Fig. 2a). After withdrawal of triple siRNAs nanoparticle treatment, the liver shrank back to its original size at Day 40 (Fig. 2a). In order to examine whether apoptosis is a key mechanism of liver shrinkage, we performed TUNEL assay and western blots of caspases. We found that in contrast with normal livers and enlarged livers, the shrinking livers had a significantly higher number of TUNEL-positive cells (Supplementary Fig. 7a). Furthermore, cleaved caspase 3 and 8 were significantly upregulated in the shrinking livers (Supplementary Fig. 7b). Our data is consistent with a previous study showing that conditional inactivation of Yap1 in enlarged mouse liver led to apoptosis of hepatocytes(10).

Figure 2. In vivo treatment with triple siRNAs induced proliferation of hepatocytes and activation of Yap1.

C57BL/6 female mice were treated with triple siRNAs or dose matched si-Control. Livers were taken at Day 7, 9, 11, 13, 15, 24, 32 or 40 days after the first injection. a, Liver/body weight ratio at different time points. Error bars represent SEM, (N≥3). b, Immunohistochemical analysis of hepatocyte proliferation (Day 11). Ki-67 (grey), Phalloidin (yellow), Hoechst (blue). c, Quantification of (b). Error bars represent SEM, **, p<0.01 (N=3). d, Western blot analysis of Hippo pathway signalling. e, Triple siRNA treatment changed Yap1 localization in liver (Day 11). Yap1 (yellow), nuclei (blue). f, Expression of Yap1-dependent genes were induced by triple siRNAs (Day 11). Error bars represent SEM, **, p<0.01 (N=3).

Remarkably, liver growth induced by the triple siRNAs nanoparticle treatment has an upper limit: prolonged treatment with siRNAs for an additional one or two weeks (up to 9 injections in total, 2 injections per week, 1.1 mg/kg per injection) did not induce further growth of the liver (Fig. 2a), indicating the possibility of mechanism(s) to restrict Yap1-induced hepatomegaly.

To clarify the targeted cell types in liver, mice were treated with siRNA nanoparticles against triple siRNAs. Hepatocytes, Kupffer cells, stellate cells and hepatic mononuclear cells were isolated from the livers. We found that with the formulation (C12-200) and siRNA dose (1.1mg/kg in total) we used, the knockdown effect is efficient (~90%) in hepatocytes (Supplementary Fig. 8a). In contrast, no significant knockdown was observed in Kupffer cells, stellate cells and hepatic mononuclear cells. We previously showed that our siRNA lipid formulation causes deep silencing of integrin β1 in hepatocytes, while no significant knockdown was found in CD31+, aSMA+, F4/80+ cells(35). Because it is difficult to get a specific commercially available antibody for the separation of biliary epithelial cells, we performed an alternative approach to explore whether our lipid nanoparticle could target biliary epithelial cells. We designed and screened a highly potent siRNA against Cytokeratin 19 (CK19), a specific marker for biliary epithelial cells. We found that this potent siRNA knockdown of CK19 has an IC₅₀ of 0.07nM in vitro (Supplementary Fig. 8b). We formulated it into lipid nanoparticles and treated mice. We found it could not induce any silencing of CK19 in mouse liver (Supplementary Fig. 8c). These data indicate that our lipid nanoparticles cannot induce gene knockdown in biliary epithelial cells.

Knockdown of Mst1, Mst2 and NF2 reproduces major features of Yap1 hyperactivation

To explore whether triple siRNAs nanoparticle-induced liver growth was due to cell division, we analyzed the proliferation of hepatocytes using ki-67 staining. We found that proliferation started at Day 7 and peaked at Day 11, at which point about 25% of total hepatocytes were ki-67-positive (Fig. 2b and c). We consistently found that expression of related genes to cell cycle was significantly induced (Supplementary Fig. 9a). Furthermore, we determined whether the proliferation of hepatocytes is due to loss of expression of components in the Hippo pathway and activation of Yap1. The protein levels of Mst1, Mst2 and NF2 in livers of triple siRNAs nanoparticle treated animals were examined by immunoblotting. We observed deep reduction of each protein at Day 11 and 15 (Fig. 2d). In consequence, triple siRNAs nanoparticle treatment diminished phospho-Yap1 (p-Yap1) by about 95% in total liver lysates (Fig. 2d and Supplementary Figure 10), while the combination of si-Mst1 and si-Mst2 treatment only decreased p-Yap1 level by about 50% (Supplementary Fig. 10). Consistently, we observed Yap1 accumulation in the nuclei of hepatocytes in liver sections after treatment with triple siRNAs nanoparticle (Fig. 2e). Interestingly, 12 days after withdrawal of triple siRNAs nanoparticle treatment (as Day 24), Mst1, Mst2, and NF2, as well as p-Yap1 recovered, indicating the feasibility of reversible manipulation of Hippo pathway (Fig. 2d). Furthermore, we found the up-regulation of Yap signature genes including CTGF, Areg, Aurka, Birc5 and AFP (Fig. 2f and Supplementary Fig. 9b,c). In addition, we observed that the growth of the liver was zone dependent. The periportal area, which contains the highest levels of nutrients and oxygen(36), exhibited the most significant expansion, whereas the pericentral area had minimal expansion (Supplementary Fig. 11a). Finally, to rule out the possibility of the involvement of macrophages in liver growth(31), we depleted hepatic macrophages using Clodronate Liposomes(30). We found that triple siRNAs nanoparticle-induced similar growth of liver in control and macrophage-depleted mice (Supplementary Fig. 12).

To analyze functions of enlarged livers, we performed analysis of serum biochemistry (Table 1). We found no significant difference in levels of ALT and AST between PBS, control siRNAs and triple siRNA nanoparticle treatment, indicating integrity of hepatocytes. Meanwhile, other indicators showed pronounced hallmarks of cholestasis, including increased levels of alkaline phosphatase, cholesterol, low density lipoproteins, bilirubin, and bile acids. To further characterize the changes in the liver related to bile flow, we performed immunohistochemical staining of CK19 to visualize bile ducts and gene expression analysis of several genes related to bile acid metabolism and transport. The number of CK19 positive cells and the density of bile ducts were significantly increased after triple siRNAs treatment (Supplementary Fig. 14). This expansion, together with altered gene expression levels, e.g. NTCP, Abcg5, MDR1, MRP4, and Cyp7 (Supplementary Fig. 9d), indicates that triple siRNA treatment induced remarkable cholestasis.

Table 1. Parameters of serum chemistry in mice treated with different combinations of siRNA targeting Hippo pathway.

	PBS	si-Control	Triple siRNA	si-Mst2+si-NF2	si-Mst1+si-NF2
Albumin, g/dl	2.90±0.12	2.89±0.22	3.31±0.15*	3.14±0.17#	2.96±0.11
Total protein, g/dl	4.7±0.1	4.9±0.2	5.5±0.2**	5.2±0.2##	5.0±0.2
Globulin, g/dl	1.82±0.08	2.06±0.05	2.24±0.11*	2.1±0.07#	2.06±0.11
Blood urine nitroqen, mg/dl	19.2±2.3	20.5±2.5	18.5±1.9	21.4±2.8	19.4±1.1
Alkaline phosphatase (ALP), IU/I	101.5±14.4	116.3±17.9	209.2±87.4*	125.8±31.5#	108.7±10.9#
ALT (SGPT), IU/I	19±2.4	24.1 ±8.9	50.9±15.7	52.9±28.7#	19.8±1.6
AST (SGOT), IU/I	47.8±16.8	67.4±29.1	65.9±3.2	83.5±12.1	59.7±11.4
Direct bilirubin, mg/dl	0.04±0.024	0.023±0.015	0.228±0.189*	0.042±0.017#	0.023±0.011#
Total bilirubin, mg/dl	0.178±0.046	0.140±0.036	0.448±0.244*	0.186±0.032##	0.132±0.013#
Total bile acids, mg/dl	6.1±1	8.9±2.6	91.4±33***	25.5±5.6###	15.1±1.8###
Cholesterol, mg/dl	68.5±10.4	76.3±10.2	125.6±20.4***	87.2±6.6###	88.7±5###
Triglycerides, mg/dl	90.9±24.3	80.5±18.2	90.2±10.4	45.4±4.7*,#	70.3±7.5
HDL, mg/dl	39.4±6.9	39.5±6.7	39.1 ±2.4	32.3±2	42.6±1.8
LDL, mg/dl	6.1±0.6	9.9±0.8	27.6±4.5***	23.3±1.9***	14.8±2.6###
Glucose, mg/dl	231.9±46.8	233.4±22.4	152.5±15.7*	208.3±34.8	214.3±48.1

^{*,**, ***}- p<0.05, p<0.01, p<0.001 comparison vs si-Control

^{#, ##, ###}- p<0.05, p<0.01, p<0.001 comparison vs Triple siRNA

The triple siRNA induced hepatomegaly is Yap1 dependent

To validate Yap1-dependent mechanism of the observed phenotype, we depleted Yap1 in triple siRNAs nanoparticle treated animals. siRNAs against Yap1 (si-Yap1) were designed and screened (Supplementary Fig. 11). We synthesized 24 siRNAs targeting Yap1. The gene silencing efficiency for individual siRNA was tested and five efficient siRNAs were identified (Supplementary Fig. 13a). The IC₅₀ for the most potent siRNA targeting Yap1 was ~0.03nM (Supplementary Fig. 13b). We co-treated animals with LNP-formulated si-Yap1 with triple siRNAs nanoparticle. We found that si-Yap1 completely blocked the liver growth induced by triple siRNAs nanoparticle (Fig. 3a), demonstrating that the hepatomegaly induced by triple siRNAs nanoparticles are Yap1-mediated. Co-treatment of si-Yap1 with triple siRNAs nanoparticles diminished both pYap1 and Yap1 expression (Fig. 3b), thereby blocking the proliferation of hepatocytes, as indicated by diminished expression of cell cycle related genes (Fig. 3b and c). Consistently, co-treatment with si-Yap1 also abolished the manifestation of cholestasis associated markers (Supplementary Table 2).

Figure 3. Additional siRNA treatment proved genetic interactions.

C57BL/6 female mice were treated with dose matched si-control, triple siRNAs plus si-Control, or triple siRNAs plus si-Yap1. Livers were taken 11 or 15 days after the first injection. a, weight at Day 15 was recorded. Error bars represent SEM, **, p<0.01 (N≥5). b, Liver lysate was made and western blot performed. c, Analysis of cell cycle related genes in liver tissues (Day 11) by qPCR. Error bars represent SEM, **, p<0.01 (N=3). d, YAP activation induces Notch pathway signaling mRNA expression of Notch pathway genes was determined by RT-qPCR in siLUC, Triple siRNA, or Triple siRNA+siYAP treatment groups,*p<0.05. N=4

A previous study showed that activation of Yap induces Notch pathway activation with expansion of hepatic progenitor cells(37). To explore whether activation of Yap induces Notch pathway activation in the triple siRNA-nanoparticles treated mouse liver, we performed a panel of qPCR analyses for Notch pathway. We found that Notch pathway signature genes including NOTCH1, NOTCH2, Jag1, SOX9, and HES were all significantly upregulated in triple siRNA treated livers (Fig 3d). Further, such induction was abolished by in vivo Yap1 siRNA treatment (Fig 3d). Notch pathway has been suggested as an inducer of biliary specification(38). This is consistent with the expansion of CK19 positive cells in triple siRNA-nanoparticles treated mouse livers (Supplementary Fig. 14).

p53 controls Yap1-induced liver growth

To further characterize the effects of hepatomegaly, we analyzed gene expression patterns of livers following long-term triple siRNAs nanoparticle treatment (Fig. 2a). Intriguingly, we observed the up-regulation of cell cycle checkpoint genes (Supplementary Fig. 7c) and apoptosis related genes (Supplementary Fig. 7e) with treatment of triple siRNAs nanoparticles. Importantly, we noticed significant induction of the cyclin-dependent kinase inhibitor 1a (p21) (Fig 4a and Supplementary Figure 15a). Withdrawal of treatment of triple siRNAs nanoparticle led to diminished p21 expression (Supplementary Fig. 15a). Moreover, co-treatment of si-Yap1 with triple siRNAs nanoparticle abolished p21 induction (Supplementary Fig. 15b), suggesting that such induction is mediated by Yap1, directly or indirectly. Based on these results, we hypothesized that the tumor suppressor protein p53, an essential regulator of p21 transcription(39), could be activated as a result of triple siRNAs nanoparticles treatment. Indeed, the pronounced nuclear staining of p53 indicated its activation with treatment of triple siRNAs nanoparticles (Fig. 4b). To investigate potential interaction between the Hippo pathway and p53, we compared the phenotype of triple siRNAs nanoparticles treatment between wildtype mice and p53^LSL/LSL mice, which harbor a Lox-STOP-Lox cassette in the first intron of p53 and are phenotypically equivalent to p53 null strains(29). In contrast to that liver of wildtype mice, the livers of p53^LSL/LSL mice grew much faster than wildtype mice and exceeded the upper limit of growth (Fig. 5a, b). Remarkably, more than 60% of hepatocytes were ki-67 positive in p53^LSL/LSL mice at Day 14, which was near 3 folds higher than in wildtype mice (Fig. 5c, d). The H&E staining indicates more severe histological abnormalities in p53^LSL/LSL mice than in wildtype animals with triple siRNAs treatment, such as a higher degree of pleomorphism, extreme polyploidy, dysplasia of hepatocytes (likely due to swelling with water/fluids), aneuploidy in a portion of hepatocytes, and intranuclear cytoplasmic inclusions (Fig 5e). To determine wehther a major part of ki-67 positive cells were indeed from hepatocytes, we performed double-staining for keratin 18 (CK18) and ki-67. We found that a major part of the ki-67 cells in both wildtype p53 and p53 KO were also CK18 positive (Fig 5f), indicating that hepatocytes were under active proliferation. To rule out the possibility of different levels of knockdown, we measured the levels of Mst1, Mst2, NF2, and pYap1 and found no differences between wildtype mice and p53^LSL/LSL mice Supplementary Fig. 15c). To further prove that p53 plays a role in controlling liver growth, we restored p53 by Tamoxifen treatment in p53^LSL/LSL; Cre-ER mice, whose stop cassette in p53 gene can be removed by Cre recombinase. We found that restoration of p53 significantly suppressed liver growth induced by triple siRNAs nanoparticle treatment (Fig. 5e and Supplementary Fig. 15d). Collectively, these results indicate that p53 is activated upon triple siRNAs nanoparticle induced hepatomegaly, and loss of p53 permits further expansion of liver size.

Figure 4. p53 was activated in livers of triple siRNAs treated animal.

C57BL/6 mice and p53^LSL/LSL were treated with triple siRNAs or dose matched si-Control. a, The levels of CDKN1A (P21) mRNA in livers of C57BL/6 mice treated with si-Control or triple siRNAs(Day 14). Error bars represent SEM, **, p<0.01 (N=5). b, p53 staining in liver sections. Red arrowheads indicate nuclear accumulation of p53.

Figure 5. p53 plays an important role in restraining liver growth.

a, Macroscopic views of livers (Day 22). b, Liver/body weight ratio at different time points. Error bars represent SEM, *, p<0.05 (N=4). c, Hepatocyte proliferation evidenced by ki-67 staining (Day 14). d, Quantification of ki-67+ cells. Error bars represent SEM, **, p<0.01, *, p<0.05 (N=3). e, H&E staining of the liver sections (Day 22). f, immunofluorescent staining of ki-67 (red) and keratin 18 (green) of the liver sections (Day 22)., scale bars = 100µm. g, p53^LSL/LSL; CreER mice treated with tamoxifen (p53 on) or vehicle controls (p53 off). Liver weight was measured 22 days after the first dose of triple siRNAs. Error bars represent SEM, **, p<0.01 (N≥3).

We have hypothesized that p53 may be activated due to DNA damage caused by the elevated levels of bile acids, a result of cholestasis associated with Yap1-induced hepatomegaly. Using the phosphorylated histone protein (H2A.X) as an indicator of DNA damage(40) we observed that triple siRNA-induced hepatomegaly led to DNA damage (Fig. 6). The intensity of the histone phosphorylation correlated to both the degree of hepatomegaly and p21 levels, i.e. maximal levels of phosphor-H2A.X were observed in the groups treated with triple siRNA; in contrast, co-treatment with triple siRNAs and si-Yap1 did not upregulate phospho-H2A.X. It suggests that p53 activation observed in this model is by activation of checkpoint cascades to DNA damage. To explore this mechanism further, we tested the direct effect of bile acids on p21 activation in murine hepatocyte-derived cell line AML-12. We treated cells with taurocholic acid, a conjugated primary bile salt which is abundant in mice with cholestasis(41). The taurocholic acid treatment led to, strong upregulation of p21 mRNA levels and nuclear accumulation of p21 as well as p53 (Fig 7a and b). To determine whether TCA treatment induced DNA damage in AML12 cells, we treated AML12 cells with TCA, and performed TUNEL assays. We found TUNEL-positive cells appeared in TCA treated cells but not control cells (Fig 7c), indicating severe DNA damage in TCA treated cells.

Figure 6. DNA damage in enlarged liver.

Livers were collected at Day 14 after the first injection of triple siRNAs. DNA damage was evidenced by phospho-H2A.X positive cells (indicated by red arrows).

Figure 7. Bile acids induce activity of p53-pathway in vitro.

AML12 cells were treated with 10mM taurocholic acid (TCA) for 4 hours. Untreated cells serve as a control. a, mRNA levels of CDKN1A(p21) in AML12 cells. Error bars represent SEM. b, immunofluorescent staining, arrowheads indicate nuclear accumulation of p53 (white) and p21 (red). Bar indicates 100µm. C Cells were fixed and stained for TUNEL and counterstained with DAPI, scalebars = 400µm.

Discussion

Our data demonstrates that it is feasible to manipulate organ size reversibly with siRNA nanoparticles in vivo. Application of chemically modified synthetic siRNA does not lead to saturation of the RISC complex and induces minimal immunostimulatory response compared to viral delivery of shRNA(24, 42, 43). Inhibition at the mRNA level of multiple components of the Hippo pathway is sufficient to activate Yap1 in vivo, thereby inducing proliferation of hepatocytes and increasing liver size. Contrary to hepatomegaly induced by complete loss of Mst1/Mst2, one induced by siRNA treatment has an upper limit of growth (about 2.5 fold). Our data indicates that activation of p53 is a key regulator for limiting liver growth. While p53 signaling is essential in tumor suppression by coordinating multiple cellular processes, including activation of DNA repair proteins, arrest of cell cycle, and initiation of cellular apoptosis and senescence (reviewed in(44)), to our knowledge, our results provide the first demonstration that endogenous p53 plays a role in size regulation of the liver. Although, we cannot exclude completely direct activation of p53 by loss of Hippo pathway, e.g. feedback loop through Lats2(45, 46), our results suggest that siRNA mediated activation of Yap1 can lead to DNA damage response in the liver. Bile acids have been shown to induce DNA damage in a certain types of cells and cell death in hepatocytes(47-49). Concordant with previous observations, our data show that Yap1 activation leads to disruption of bile duct homeostasis(7) and provokes a cholestasis-like phenotype.

Crosstalk between Hippo and p53 pathways has been suggested in fly models as well as in mammalian cell line models(46, 50). Our approach allows investigation of pathway interactions in normal livers. It reveals indirect interactions between signaling pathways mediated by tissue remodeling and changes in composition of metabolites. Remarkably, not only is the proliferation rate of hepatocytes in p53^LSL/LSL mice much higher than in the wildtype control, but also the size of hepatocytes in p53^LSL/LSL mice is significantly enlarged (Fig. 5d; Supplementary Fig. 12c, d). It is noteworthy to investigate coordination of p53 and Hippo pathway during the process of regeneration and under the context of pathological conditions for future studies. It is known that abnormal bile acids homeostasis can activate p53 in the mouse liver (51). There is evidence indicating that Yap overactivation can lead to biliary epithelium over-proliferation and irregular shape(7). To our best knowledge, there is no report showing Yap overactivation provokes a cholestasis-like phenotype. Dysfunction of bile ducts may be mediated by Yap-dependent dysregulation of gene expression of bile acid transporters in hepatocytes(52) (Supplementary Fig. 9d) or indirect effects on the bile ducts function through modulation of paracrine/contact signaling between hepatocytes and cholangioytes (Fig. 3d). Extensive expansion of hepatocytes in the periportal zone may contribute to mechanical distortion of bile duct function. Our data suggests that Yap activation leads to abnormal bile acid homeostasis, which in turn triggers p53 activation (Fig 4, 6, 7). Furthermore, our data indicates that loss of p53 led to increased hepatocyte proliferation when Hippo pathway was suppressed (Fig 5). Previous studies reported that no obvious abnormal liver histology was found with short-term deletion of p53; in contrast, deletion of p53 induces liver cancer after 10-12 months in mouse model(53). Several studies showed that loss of p53 did not significantly accelerate phenobarbitone or diethylnitrosamine-induced hepatocellular carcinoma(54-56). Thus, our finding suggests an important connection between p53 and Hippo pathway, two key pathways for maintaining quiescence of hepatocytes. Although p21 is an important downstream effector of p53 and its induction is a recognized marker of p53 activation, the tumour suppressor p53 has a number of targets to exert its functions(57). The p21-deficient mice have a less obvious phenotype than p53-deficient animals(58).

We have shown that siRNA nanoparticle-mediated control of organ size is reversible, as indicated by the shrinking of liver size and restoration of pYap1 after withdrawal of triple siRNAs nanoparticle-treatment (Fig. 2a and d). Livers shrank back to their original size, indicating an intrinsic mechanism for regulating normal organ size. The technique presented in our manuscript provided a flexible way to induce transient gene inactivation and subsequent ability to study the mechanism of liver size control in a new manner. A number of questions remain unclear regarding the process of how the liver returns to the exact normal size, including which signals trigger and stop the apoptosis of hepatocytes, and why only part of cells are undergoing apoptosis. It is possible that multiple factors are involved in the process in the liver, including mechanical forces, oxygen levels and surface and nuclear receptors. The method described here allows dissection of these factors in vivo in a flexible and convenient way. A recent study indicated that a compensatory network can be activated to buffer against deleterious mutations rather than transcriptional knockdown(59). Thus, an in vivo genetic knockdown approach may reveal phenotypes which can not be observed by a knockout approach. As loss of the Hippo pathway is temporary, and the dose of siRNAs can be titrated to induce temporary proliferation in a portion of hepatocytes without substantial increase of organ size (data not shown), we believe that modulating liver size in vivo with siRNA-nanoparticles may provide an efficient tool to explore autonomous and non-autonomous mechanisms for organ size regulation.