A genetic system for tissue-specific inhibition of cell proliferation

During development, tissue homeostasis, disease progression, or tissue repair and regeneration after injury, cells typically proliferate to expand their numbers or compensate for their loss (Simons and Clevers, 2011; Hogan et al., 2014; Augustin and Koh, 2017; Mohamed et al., 2018). For example, in development, cellular proliferation occurs to ensure proper organ growth and function (Buckingham et al., 2005; Heallen et al., 2011; Tzahor and Poss, 2017). The preliminary coronary vascular plexus expands and sprouts to form new blood vessels in the ventricular wall during heart development (Red-Horse et al., 2010; Tian et al., 2013), and pre-existing coronary endothelial cells re-expand and contribute to neovascularization after myocardial infarction (He et al., 2017a). Likewise, during adult homeostasis stem cells residing in the crypt of the intestine differentiate into transit-amplifying cells that rapidly expand to make more enterocytes as the intestinal wall repeatedly sheds (Barker et al., 2007).

Further, in mouse liver homeostasis, Axin2⁺ hepatocytes residing initially in the peri-central region of the liver lobule expand gradually to populate the liver, reaching the peri-portal region after 1 year of proliferation and migration. During liver injury, Sox9⁺ hybrid hepatocytes in the peri-portal region expand further to re-populate the entire liver lobule (Font-Burgada et al., 2015). Also, during lung injury, stem cells residing in the bronchioalveolar duct junction differentiate and expand to provide bronchiolar or alveolar epithelial cells for lung repair and regeneration (Kim et al., 2005). Cardiomyocytes in the neonatal heart proliferate and heart apex resection stimulates their proliferation even further to regenerate the tissue after such injury (Porrello et al., 2011). Although in the adult heart cardiomyocyte turnover is low, these cells still exhibit detectable proliferation to slowly compensate for the exchange of ∼50% of these cells during a normal life span (Bergmann et al., 2009). Throughout the body, cell proliferation is involved in virtually every biological process that ranges from development to disease to regeneration.

Although cell proliferation is an important and fundamental process, it remains technically challenging to dissect the exact contribution of distinct expanding cell populations in the context of the entire organism in the setting of disease or tissue repair from other complex processes that occur in these settings. Many cell types are typically involved in disease pathogenesis and progression, and their individual expansion or cell type conversion can be either beneficial or detrimental to tissue recovery (Zeisberg et al., 2003; Ubil et al., 2014). For example, after myocardial infarction, coronary endothelial cells proliferate markedly in the infarct border due to tissue hypoxia (Zhou et al., 2011). The injury also stimulates cardiomyocyte proliferation in the border region, which could be beneficial to heart repair (Lin et al., 2014). Meanwhile, during heart repair fibroblasts expand and become myofibroblasts that assist in the repair of the infarct region, but if they do so excessively tissue fibrosis and scar formation occur (Moore-Morris et al., 2014).

Owing to technical challenges, the contribution of cell proliferation of individual cell types to disease progression remains largely unexplored, even though such insight could be valuable. For example, in addition to the cell-autonomous role of cell proliferation, manipulation of the proliferation of one cell type could also provide a unique opportunity to understand the behavior of other cell types in the disease or repair process. Previous studies in liver regeneration showed that biliary epithelial cells or ductal cells do not differentiate into new hepatocytes after injury, and hepatocytes proliferate by themselves to generate new hepatocytes after liver injury (Chu et al., 2011; Malato et al., 2011; Schaub et al., 2014; Tarlow et al., 2014; Yanger et al., 2014). However, a recent study reported that impaired hepatocyte proliferation could induce the conversion of biliary epithelial cells to new hepatocytes (Raven et al., 2017; Deng et al., 2018), indicating that manipulation of cell proliferation could advance our understanding of cell plasticity and tissue regeneration. Therefore, manipulation of cell proliferation may provide a unique approach to understanding the role of lineage expansion of a single cell type in the context of a whole organism's response to injury, while also unraveling the underlying cellular plasticity.

To generate a genetic system that allows inducible cell cycle arrest in a tissue-specific manner, we used controlled expression of the cell cycle inhibitor gene p21 (Cdkn1a) for cell proliferation inhibition (Carlson et al., 2008; Raven et al., 2017). To achieve tissue specificity, we employed a Cre-loxP recombination system to target p21 overexpression, while also expressing GFP to mark such cells (ip21-GFP system). We found that p21-overexpressing cells did not proliferate during homeostasis, disease, or after injury. Further, to facilitate the use of ip21-GFP in combination with another cell lineage tracing system, we also generated a second ip21-GFP system that was responsive to Dre-rox recombination and orthogonal to Cre-loxP system (Anastassiadis et al., 2009). Here, we present examples of how these tools could improve our understanding of the role of cell proliferation and non-cell-autonomous effects on cell fate plasticity during disease and tissue regeneration.

To allow Cre-mediated gene expression, we generated a R26-p21-GFP allele by homologous recombination using a CRISPR/Cas9 strategy (Fig. 1A). GFP was linked with p21 by a self-cleaving 2A peptide so that GFP activity indicated p21 expression in the targeted cells (Fig. 1A). To characterize the R26-p21-GFP mouse line, we first crossed it with a Cre line for the cardiac fibroblast study. Namely, we used a fibroblast-specific CreER line, Col1a2-CreER (Fig. 1B), which is specifically expressed in fibroblasts in the developing heart (Fig. S1). We then crossed Col1a2-CreER mice with R26-p21-GFP mice to generate double-positive mice for subsequent experiments. Tamoxifen treatment of these mice at postnatal day (P) 3 induces Cre-loxP-mediated Stop cassette removal and subsequent p21-GFP expression in Col1a2⁺ fibroblasts (Fig. 1C). As a control reporter line, we used R26-tdTomato (Madisen et al., 2010) for the study of Col1a2-CreER-labeled cells. Whole-mount fluorescence of P8 hearts showed efficient GFP or tdTomato expression in Col1a2-CreER;R26-p21-GFP and Col1a2-CreER;R26-tdTomato tissues, respectively (Fig. 1D). Immunostaining for GFP (surrogate for p21 expression) and PDGFRa (a fibroblast marker) in Col1a2-CreER;R26-p21-GFP heart sections showed that the majority of PDGFRa⁺ cells co-expressed GFP (Fig. 1E).

We also detected highly efficient labeling by Col1a2-CreER;R26-tdTomato (Fig. 1E). Quantification revealed that 89.84±1.47% and 93.32±1.84% of PDGFRa⁺ fibroblasts were labeled in Col1a2-CreER;R26-p21-GFP and Col1a2-CreER;R26-tdTomato hearts, respectively (Fig. 1F). There was negligible reporter activity in both groups in the vehicle-treated controls (Fig. 1F). Immunostaining for GFP and p21 on Col1a2-CreER;R26-p21-GFP showed robust p21 expression in GFP⁺ cells (Fig. 1G), whereas p21 was not detected in tdTomato⁺ fibroblasts in Col1a2-CreER;R26-tdTomato hearts (Fig. 1H). We also confirmed that a subset of PDGFRa⁺ fibroblasts expressed p21 in heart sections from Col1a2-CreER;R26-p21-GFP mice (Fig. 1I). p21 expression was also detected in subsets of cardiomyocytes in both groups (Fig. S2). These data indicate that the R26-p21-GFP allele responded to Cre specifically and efficiently in fibroblasts, which was comparable to a conventional reporter.

To test functionally whether overexpression of p21 inhibits fibroblast proliferation, we stained tissue sections for Ki67 (Mki67) and found that GFP⁺ fibroblasts in heart sections from Col1a2-CreER;R26-p21-GFP mice did not express Ki67, while a subset of tdTomato⁺ fibroblasts in samples from Col1a2-CreER;R26-tdTomato mice did express Ki67 (Fig. 1J). Additionally, immunostaining for 5-ethynyl-2′-deoxyuridine (EdU) and GFP in heart sections from Col1a2-CreER;R26-p21-GFP mice showed no EdU incorporation in GFP⁺ fibroblasts, whereas EdU was readily detected in non-fibroblast lineages, such as cardiomyocytes and endothelial cells (Fig. S3). As a control, EdU was readily detected in tdTomato⁺ fibroblasts of hearts from Col1a2-CreER;R26-tdTomato mice (Fig. 1K). Quantification showed that 11.73±1.27% of fibroblasts incorporated EdU in hearts from Col1a2-CreER;R26-tdTomato mice, but almost no incorporation was observed in hearts from Col1a2-CreER;R26-p21-GFP mice (Fig. 1L). These data demonstrate that R26-p21-GFP could be used to inhibit proliferation of a specific cell lineage, such as fibroblasts (Fig. 1M).

We next functionally tested whether inhibition of fibroblast proliferation could have any significant impact on tissue fibrosis after myocardial infarction (MI). We ligated the left anterior descending coronary artery in mice to induce MI at 2 weeks after tamoxifen treatment, and collected the hearts for analysis at 3 and 7 days following MI surgery (Fig. 2A). Hearts from Col1a2-CreER;R26-p21-GFP mice showed less collagen production compared with those from Col1a2-CreER;R26-tdTomato mice after MI injury (Fig. 2B). Sirius Red staining to measure the degree of fibrosis showed a significantly lower amount of tissue fibrosis in the R26-p21-GFP group compared with the R26-tdTomato group (Fig. 2C).

We then performed immunostaining for EdU and GFP or tdTomato on heart sections and found almost no detectable EdU⁺GFP⁺ fibroblasts in hearts from Col1a2-CreER;R26-p21-GFP mice, but a substantial number of EdU⁺tdTomato⁺ fibroblasts in hearts from Col1a2-CreER;R26-tdTomato mice at both 3 and 7 days following MI (Fig. 2D). Quantitatively, we found that whereas 13.39±1.76% and 2.54±0.52% of fibroblasts incorporated EdU at day 3 and 7 after MI, respectively, in the infarct regions of the hearts from Col1a2-CreER;R26-tdTomato mice, only 0.075±0.020% and 0.051±0.011% of fibroblasts incorporated EdU at day 3 and 7 after MI, respectively, in the infarct regions of hearts from Col1a2-CreER;R26-p21-GFP mice (Fig. 2E). Further, we found that most of the EdU⁺ cells in the hearts of Col1a2-CreER;R26-p21-GFP mice were VE-cadherin (VE-cad; Cdh5)⁺ endothelial cells or CD45 (PTPRC)⁺ hematopoietic cells in the infarcted regions (Fig. 2F,G). We also detected Ki67⁺tdTomato⁺ fibroblasts in the R26-tdTomato group, but Ki67⁺GFP⁺ fibroblasts were not detected in the R26-p21-GFP group (Fig. 2H,I).

To validate this finding further, we collected hearts at 28 days after MI, the stage at which most stable scars form. We stained PDGFRa and GFP or tdTomato on Col1a2-CreER;R26-p21-GFP or Col1a2-CreER;R26-tdTomato heart sections, respectively, and showed that both GFP and tdTomato were colocalized with the fibroblast cell marker (Fig. 3A). There was a noticeable reduction of GFP⁺ fibroblasts compared with tdTomato⁺ fibroblasts in the infarcted region of MI hearts (Fig. 3A). We also observed a significant increase of coronary blood vessels in the infarcted regions of Col1a2-CreER;R26-p21-GFP hearts compared with Col1a2-CreER;R26-tdTomato hearts (Fig. 3B), indicating that inhibition of fibroblast proliferation promoted neovascularization of injured myocardium. With more blood vessel supply, the area of surviving myocardium in the infarcted region of Col1a2-CreER;R26-p21-GFP was significantly greater than that of Col1a2-CreER;R26-tdTomato (Fig. 3C). Sirus Red staining on MI heart sections confirmed decreased scar formation in Col1a2-CreER;R26-p21-GFP hearts compared with littermate controls (Fig. 3D,E). Importantly, reduced scar formation was associated with improved heart function, as evaluated by echocardiography (Fig. 3F). These data suggest that inhibition of fibroblast proliferation significantly reduced scar formation and promoted neovascularization and cardiomyocyte survival.

Independently, we induced inhibition of fibroblast proliferation in a myocardial ischemia-reperfusion model using ip21-GFP mice (Fig. S4). We found that overexpression of p21 in fibroblasts inhibited their proliferation and ameliorated tissue fibrosis after myocardial ischemia-reperfusion injury (Fig. S4). Taken together, these data show that inhibition of fibroblast proliferation through p21 overexpression reduces tissue fibrosis and scar formation.

To show application of the ip21-GFP system in another context, we studied its effect on inhibition of hepatocyte proliferation in liver homeostasis or injury. We crossed Alb-CreER mice, which selectively express CreER in hepatocytes and epithelia cells of kidney via the albumin promoter (Fig. S5), with those harboring the R26-p21-GFP reporter. We then induced p21-GFP expression in Alb⁺ hepatocytes by tamoxifen (Fig. 4A). To detect hepatocyte proliferation, we collected livers from 4-week-old mice and analyzed them for GFP expression. We detected GFP signal in the livers of the tamoxifen-treated, but not vehicle-treated, Alb-CreER;R26-p21-GFP mice (Fig. 4B). These GFP⁺ cells expressed a high level of p21 in mice treated with tamoxifen whereas no p21⁺ cells were detected in mice after vehicle treatment (Fig. 4C, Fig. S6A). Quantitative RT-PCR showed significantly higher p21 expression in hepatocytes isolated from mice treated with tamoxifen compared with vehicle-treated mice (Fig. 4D).

To verify that the expression of p21-GFP was specific to hepatocytes in liver, we co-stained for GFP and the hepatocyte marker HNF4a or the biliary epithelial cell marker CK19 (Krt19) on liver sections, and found that hepatocytes, but not biliary epithelial cells, express GFP (Fig. 4E,F). To test functionally whether the overexpression of p21-GFP inhibits hepatocyte proliferation, we stained liver sections for GFP and Ki67 (Fig. 4G). We found both Ki67⁺ hepatocytes (larger, round nuclei; Fig. 4G, arrowheads) and Ki67⁺ non-hepatocytes (smaller, irregularly shaped nuclei; Fig. 4G, arrows) in the vehicle-treated group. In contrast, we detected Ki67 expression in GFP⁻ non-hepatocytes but not in GFP⁺ hepatocytes in liver sections of tamoxifen-treated mice (Fig. 4G). Quantitative data showed that although there was no significant difference in the number of proliferating non-hepatocytes between vehicle- and tamoxifen-treated samples, the number of proliferating hepatocytes per field was markedly different between the two groups (Fig. 4H).

At the adult stage (8 weeks of age and later), most hepatocytes do not proliferate during tissue homeostasis, but they do proliferate more prominently after injury (Pu et al., 2016). Thus, we next explored whether the ip21-GFP system in the adult liver would have any effect on hepatocyte proliferation after liver injury. We first induced carbon tetrachloride (CCl₄)-mediated liver injury at 2 weeks after tamoxifen or oil treatment and then collected samples 48 h later to check the response of hepatocytes to acute injury (Fig. 4I). Immunostaining for GFP and Ki67 on liver sections from the oil-treated group showed that hepatocytes responded robustly to injury by boosting the cell cycle at an extremely high level, as 25.14±1.91% of hepatocytes expressed Ki67 (Fig. 4J,K). However, we could barely detect any proliferating hepatocytes after injury in the tamoxifen-treated group (Fig. 4J,K). In the same sections, we also detected Ki67⁺ cells that were GFP⁻, indicating that non-hepatocytes proliferate in response to injury in Alb-CReER;R26-p21-GFP mice treated with tamoxifen (Fig. 4J). In the sham controls, we detected very few Ki67⁺ hepatocytes in the oil-treated group and none in the tamoxifen-treated group (Fig. S6B). These data demonstrate that the ip21-GFP system could be used to specifically inhibit hepatocyte proliferation during both tissue homeostasis and in response to injury (Fig. 4L).

We next exposed Alb-CReER;R26-p21-GFP mice to a chronic liver injury model by treating them for 3 weeks with a 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC)-containing diet (Fig. S7A). In both normal and DDC diets, tamoxifen treatment induced robust p21-GFP expression in the liver (Fig. S7B). Whereas all examined HNF4a⁺ hepatocytes in the normal-diet group were GFP⁺, we noticed a subset of HNF4a⁺ hepatocytes that were GFP⁻ in the DDC-fed group (Fig. S7C). Noticeably, most of these HNF4a⁺GFP⁻ cells were restricted to the peri-portal region, where they were closely associated with CK19⁺ biliary epithelial cells (Fig. S7C). Immunostaining for GFP, E-cadherin (E-cad, a peri-portal hepatocyte marker; also known as Cdh1) and glutamine synthetase (GS, a peri-central hepatocyte marker) on liver sections from DDC-treated mice showed that a subset of GFP⁻ hepatocytes were present in the peri-portal regions (Fig. S7D).

We never observed the appearance of GFP⁻ hepatocyte clusters in the normal-diet group (Fig. S7D). Functionally, GFP⁻ hepatocytes, but not GFP⁺ hepatocytes, express Ki67 in the peri-portal regions in the livers of DDC-fed mice (Fig. S7E). Consistent with this observation, we found that GFP⁺ hepatocytes, but not GFP⁻ hepatocytes, expressed p21 (Fig. S7F). These GFP⁻ hepatocytes may emerge from a limited number of unlabeled pre-existing hepatocytes. However, it is also possible that inhibition of hepatocyte proliferation in chronic injury may induce a new hepatocyte generation from other cell lineages, such as has been previously reported for biliary epithelial cells (BECs) (Raven et al., 2017).

Cre-loxP has been widely used and the ip21-GFP system would be compatible with many existing Cre-based tools in current use. However, to inhibit the proliferation of one cell lineage at the same time as genetically tracing or manipulating another cell lineage, an alternative genetic system is needed in addition to the present Cre-loxP system. A previous report indicated that Dre-rox and Cre-loxP were two orthogonal recombination systems, and they could be used simultaneously for more precise evaluation of cell fate plasticity and gene function in vivo (Anastassiadis et al., 2009; Hermann et al., 2014; He et al., 2017b; Pu et al., 2018). We therefore generated a Dre-rox-induced p21-GFP expression tool based on a R26-rox-p21-GFP allele (Fig. 5A), which responds to Dre recombinase and expresses p21-GFP upon AAV8-TBG (human thyroid hormone-binding globulin; also known as SERPINA7)-Dre virus treatment (Fig. 5B).

We did not detect any GFP signal in R26-rox-p21-GFP mice with AAV8-control virus treatment (Fig. 5B), indicating no leakiness of the R26-rox-p21-GFP allele in the absence of Dre expression. Also, given the serotype of the AAV vector, we found that AAV8-TBG-Dre specifically targeted rox sites in hepatocytes (Fig. S8). Immunostaining showed that almost all HNF4a⁺ hepatocytes were GFP⁺ after Dre virus treatment, whereas hepatocytes were GFP⁻ in livers of mice treated with a control virus (Fig. 5C,D). We also found that p21 was expressed in almost all GFP⁺ hepatocytes in liver samples from Dre, but not control, virus-treated mice (Fig. 5E). Although the proliferation of adult hepatocytes was extremely low, we did detect Ki67⁺ hepatocytes in AAV8-control treated mice (Fig. 5F). However, we never detected Ki67⁺ hepatocytes in Dre virus-treated mice out of 65 liver sections from five mice examined (Fig. 5F).

We next exposed the mice to CCl₄ treatment and collected samples for analysis of cell proliferation (Fig. 5G). CCl₄ mainly induced acute injury in the peri-central region (Fig. 5H), which would lead to peri-central hepatocyte death and induction of proliferation of peri-portal hepatocytes as a compensatory mechanism (Pu et al., 2016). Immunostaining for GFP, HNF4a and CK19 showed hepatocyte death in the peri-central region of AAV8-control- or AAV8-Dre-treated mouse livers (Fig. 5I). We detected robust hepatocyte proliferation in the peri-portal region of AAV8-control- but not Dre virus-treated mice (Fig. 5J,K). These data demonstrate that the R26-rox-p21-GFP system could be used to inhibit cell proliferation in specific cell lineages targeted by Dre (Fig. 5L).

To test whether inhibition of the proliferation of one cell lineage can induce another cell lineage to trans-differentiate and compensate for the loss of the targeted cells, we induced chronic liver injury by DDC diet in R26-rox-p21-GFP mice (Fig. S9A). For both normal and DDC diet groups, AAV8-TBG-Dre treatment of R26-rox-p21-GFP mice led to efficient recombination in the liver (Fig. S9B). Immunostaining for GFP, HNF4a and CK19 on liver sections showed that hepatocytes in both peri-central and peri-portal regions were efficiently labeled by GFP (Fig. S9C). Immunostaining with the zonation markers E-cad and GS showed the efficient labeling (>99%) of hepatocytes in peri-portal and peri-central regions of the liver lobule (Fig. S9C). In R26-rox-p21-GFP mice fed the DDC diet, we detected clusters of prominent GFP⁻ hepatocytes appearing in the midst of GFP⁺ hepatocytes (Fig. S9D). Many of these GFP⁻ clones were closely associated with BECs. These GFP⁻ hepatocytes expressed Ki67 but not p21, whereas their surrounding GFP⁺ hepatocytes expressed p21 but not Ki67 (Fig. S9E,F).

These data suggest that p21-GFP expression in hepatocytes inhibited their proliferation, whereas p21⁻GFP⁻ hepatocytes emerge to proliferate as a clonal expansion to generate new hepatocytes during chronic liver injury. These proliferating GFP⁻ hepatocytes are likely derived from a very few pre-existing hepatocytes that have not been targeted by the Dre virus. It is also likely that hepatocyte inhibition spurs the potential of BECs to convert into hepatocytes during liver repair and regeneration. Both mechanisms may happen for generating new hepatocytes.

To address BEC-to-hepatocyte conversion directly, we combined the Dre-rox and Cre-loxP orthogonal systems to inhibit one cell lineage proliferation while simultaneously genetically tracing another cell lineage. Treating R26-rox-p21-GFP;CK19-CreER;R26-tdTomato mice with AAV8-TBG-Dre virus (Fig. 6A), we inhibited hepatocytes proliferation by the Dre-rox system and simultaneously induced CK19-CreER-mediated lineage tracing of BECs in the liver. Virtually all hepatocytes (i.e. HNF4a⁺) expressed p21-GFP and 40.92±1.84% of BECs (i.e. CK19⁺) were labeled with tdTomato in the liver (Fig. 6B-D). We fed mice for 4 weeks with the DDC diet, then switched them to a normal diet, and collected livers for analysis after 5 weeks' recovery (Fig. 6E).

A prominent cluster of a tdTomato⁺ signal was detected in the AAV8-TBG-Dre-treated liver, compared with the AAV8-control-treated group (Fig. 6F). By immunostaining, we found clusters of tdTomato⁺HNF4a⁺ hepatocytes in the peri-portal regions (Fig. 6G), demonstrating that these hepatocytes were derived from pre-labeled BECs. No tdTomato⁺HNF4a⁺ hepatocytes were detected in the AAV8-control-treated mice (Fig. 6G). The BEC-derived hepatocytes expressed the peri-portal zonation marker E-cad but not the peri-central zonation marker GS (Fig. 6H,I). Although p21 was expressed in GFP⁺ hepatocytes, we did not detect any p21 expression in tdTomato⁺ hepatocytes (Fig. 6J). Further, we found that Ki67 was expressed in these tdTomato⁺ hepatocytes, but not in GFP⁺ hepatocytes (Fig. 6K). These data demonstrate that genetic inhibition of hepatocyte proliferation during chronic liver injury induces a BEC-to-hepatocyte conversion (Fig. 6L).

In this study, we generated an inducible genetic system to inhibit cell proliferation in a tissue-specific manner. We presented two examples of how this genetic approach could be used for studying the role of cell proliferation in tissue homeostasis, repair and regeneration. During MI many cell lineages, including endothelial cells, fibroblasts and cardiomyocytes, respond to injury by increasing their proliferation. Unravelling the specific role of the expansion of one cell type in the context of the whole-organ response is crucial for understanding the function of this expansion and also for potential therapeutic targets. It has been reported that ablation of myofibroblasts in the injured myocardium reduces collagen production and scar formation after MI (Kanisicak et al., 2016; Kaur et al., 2016). After MI, the resident pre-existing fibroblasts respond to hypoxic conditions and expand to form essentially all myofibroblasts (He et al., 2017a). As excessive myofibroblasts were a result of cell proliferation from pre-existing fibroblasts, we therefore used the ip21-GFP system to inhibit the fibroblast proliferation and excessive expansion, which resulted in a significant reduction of tissue fibrosis and scar formation. Of note, inhibition of fibroblast proliferation significantly promoted neovascularization, resulting in more cardiomyocyte survival in the infarcted region.

It has been proposed that facultative resident hepatic stem or progenitor cells contribute to new hepatocytes after injury in both fish and mice (Español-Suñer et al., 2012; Choi et al., 2014; He et al., 2014; Rodrigo-Torres et al., 2014). However, genetic lineage-tracing studies in mice indicate that pre-existing hepatocytes are the cellular source of new hepatocytes generated after injury (Schaub et al., 2014; Tarlow et al., 2014; Yanger et al., 2014). Rather than putative stem cell differentiation, adult hepatocytes are generated by self-duplication or proliferation (Chu et al., 2011; Malato et al., 2011). In chronic severe human liver diseases, hepatocyte proliferation was significantly impaired in the context of widespread hepatocyte senescence, indicative of p21 upregulation (Marshall et al., 2005; Richardson et al., 2007). However, the general injury models (e.g. hepatectomy, CCl₄, DDC or bile duct ligation) in mice stimulate rather than inhibit hepatocyte proliferation in the liver, which does not actually recapitulate human chronic liver injury conditions. To mimic human liver injury conditions in a mouse model, a recent study employed gene manipulation to inhibit hepatocyte proliferation in a combination of conventional liver injuries, and then found that cholangiocytes contribute to hepatocytes in this particular condition (Raven et al., 2017). Xie and colleagues also reported that the conversion of cholangiocytes to hepatocytes could occur in mouse after long-term severe liver injuries (Deng et al., 2018). The inhibition of hepatocytes therefore has relevant biological significance for the study of cell lineage plasticity in models of human liver diseases. These studies and ours also suggested that hepatocyte regeneration from biliary cells might occur in human chronic liver disease. Identifying the cellular sources of hepatocyte regeneration could provide insights for developing new therapeutic strategies.

We therefore generated an inducible Cre-loxP-based genetic tool for p21 overexpression to selectively inhibit cell proliferation in any cell type of interest (as long as a relevant Cre-driver is available). To prove the utility of this new system, we showed that Alb-CreER-mediated recombination results in highly efficient and specific overexpression of p21-GFP in hepatocytes, suppressing their proliferation both in homeostasis and after injury. Interestingly, we did observe a cluster of p21⁻ hepatocytes appearing in the peri-portal regions after chronic injury and these cells were tightly associated with ductal cells. It is possible that very rare pre-existing p21⁻ hepatocytes expand during injury. However, due to the restricted regions that associate with BECs in the peri-portal region, it is more likely these p21⁻ hepatocytes were derived from p21⁻ BECs after injury when hepatocyte proliferation is impaired (Raven et al., 2017).

In addition to the Cre-loxP recombination system, we also generated a Dre-rox-mediated p21-GFP expression genetic system. By comparing hepatocyte proliferation, we found that the R26-rox-p21-GFP system works as efficiently and specifically as R26-p21-GFP for suppressing cell proliferation. Because Cre-loxP and Dre-rox are two orthogonal recombination systems (He et al., 2017b), we then employed Dre-rox to inhibit hepatocyte proliferation by p21-GFP and simultaneously used CK19-CreER;R26-tdTomato alleles to trace BECs in the same liver. In such mice fed a DDC diet to induce chronic liver injury, we detected clusters of tdTomato⁺ BECs-derived hepatocytes in the peri-portal regions. Of note, we also observed clusters of tdTomato⁻GFP⁻ in the same liver after injury. We observed that tdTomato⁺ hepatocytes contributed to 0.22±0.021% of the liver by quantification of liver sections stained with tdTomato, GFP and HNF4a. This was partly due to the inefficiency of BEC labeling by CK19-CreER (∼40%). We also found that tdTomato⁻GFP⁻ hepatocytes accounted for 30.5±4.7% of all hepatocytes after DDC injury. The presence of a high proportion of tdTomato⁻GFP⁻ hepatocytes was attributed to incomplete genetic targeting by AAV8-TBG-Dre. Although the majority of pre-existing hepatocytes overexpress p21-GFP after Dre-rox recombination, the very small number of untargeted hepatocytes (GFP⁻) expands significantly during liver injury, maintaining GFP⁻ and exhibiting a lack of tdTomato expression. Owing to the presence of a subset of p21⁻GFP⁻ hepatocytes, the magnitude of lineage conversion from BECs to hepatocytes could be significantly reduced after liver injury. Improving the efficiency of AAV8-TBG-Dre virus infection or using the Alb-ires-DreER line would likely enhance the efficiency of hepatocyte proliferation inhibition, thereby promoting the levels of regeneration of hepatocytes from biliary cells.

A previous study revealed the hepatobiliary hybrid progenitor (HHyP) population in human fetal liver, which suggested the possibility that BECs express other lineage markers after stress (Segal et al., 2019). However, our genetic labeling was induced at 4 weeks before liver injury, so our labeled cells were BECs under normal conditions. It is less likely that we labeled the HHyP population during the time window when we administered the tamoxifen. Although the possible existence of a HHyP population during stress may represent a new interesting direction, our study does not focus on identification of a HHyP population in the adult liver under stress. Even so, our work provides an example of inhibition of one cell type by ip21-GFP leading to the cell-fate conversion of another cell lineage, indicating the non-cell-autonomous role of cell proliferation inhibition during chronic injury.

Taken together, our work provides two inducible genetic tools for inhibition of cell proliferation of specific cell lineages. Combination of R26-p21-GFP or R26-rox-p21-GFP with other genetic tools could further help elucidate both the cell-autonomous and non-cell-autonomous role of cell proliferation in development, disease and regeneration. We expect that dissecting the exact role of proliferation of one cell lineage will provide new insights into disease pathogenesis, while also uncovering potential new therapeutic targets for the treatment of disease and/or improving regenerative medicine.

All mice studies were carried out in accordance with the guidelines in the Institutional Animal Care and Use Committee (IACUC) of the Institute for Nutritional Sciences and the Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Science. The Alb-CreER, CK19-CreER, R26-tdTomato and R26-rox-tdTomato mouse lines have been published previously (Barker et al., 2007; Madisen et al., 2010; Zhang et al., 2016; He et al., 2017b). Col1a2-CreER, R26-p21-GFP and R26-rox-p21-GFP mouse lines were generated by homologous recombination using CRISPR/Cas9 technology. For Col1a2-CreER, the cDNA encoding CreER was inserted into the translational start codon ATG of the Col1a2 gene, followed by a polyadenylation sequence. For R26-p21-GFP, the cDNA encoding CAG-loxp-stop-loxp-p21-P2A-GFP was inserted into the Rosa26 gene locus (between exon 1 and exon 2). A P2A peptide sequence was used to link the p21 coding region and GFP cDNA, allowing translation of both p21 and GFP proteins. For the R26-rox-p21-GFP mouse line, the CAG-rox-stop-rox-p21-P2A-GFP cDNA was generated and inserted into the Rosa26 locus as R26-p21-GFP. These mouse lines were generated by Shanghai Biomodel Organism Co., Ltd., Shanghai, China. All mice involved in studies were maintained on 129/C57BL6 and ICR mixed background. Animals received tamoxifen (Sigma-Aldrich, T5648) dissolved in corn oil at the indicated time. Adult mice received 0.1 mg tamoxifen per gram mouse body weight by oral gavage. Both male and female mice were included in the study.

Collected mouse tissues were fixed in 4% paraformaldehyde (PFA; Sigma-Aldrich) and placed on agar for whole-mount imaging and z-stack imaging using a Zeiss stereomicroscope (Axio Zoom V16).

Immunostaining was performed as previously described. In detail, dissected tissues were fixed in 4% PFA at 4°C for 1 h. Then, tissues were washed in PBS, dehydrated in 30% sucrose overnight at 4°C and embedded in OCT (Sakura). For staining, cryosections (8-10 μm) were washed in PBS, incubated in blocking buffer [5% normal donkey serum (Jackson ImmunoResearch), 0.1% Triton X-100 in PBS] for 30 min at room temperature then stained with the primary antibodies overnight at 4°C. Signals were developed with Alexa fluorophore-conjugated or horseradish peroxidase-conjugated secondary antibodies: Alexa 488 donkey anti-rabbit (Invitrogen, A21206), Alexa 555 donkey anti-rabbit (Invitrogen, A31572), Alexa 647 donkey anti-rabbit (Invitrogen, A31573), Alexa 488 donkey anti-goat (Invitrogen, A11055), Alexa 555 donkey anti-goat (Invitrogen, A21432), Dylight 549 donkey anti-rat IgG(H + L) (Jackson ImmunoResearch, 712-505-153), Alexa 488 donkey anti-rat (Invitrogen, A21808), immPRESS-rat immunoglobulin (goat) (Vector Laboratories, MP-7444-15), immPRESS-goat immunoglobulin (horse) (Vector Laboratories, MP-7405-50), immPRESS-rabbit immunoglobulin (horse) (Vector Laboratories, MP-7401-50). The tyramide signal amplification kit (PerkinElmer) was used to amplify weak signals. Nuclei were counterstained with 4′6-diamidino-2-phenylindole (DAPI, Vector Laboratories). The following antibodies were used: tdTomato (Rockland, 600-401-379, 1:500), tdTomato (ChromoTek, 5F8, 1:100), GFP (Invitrogen, A11122, 1:500), GFP (Abcam, ab6662, 1:500), GFP (Nacalai Tesque, 04404-84, 1:500), PDGFRa (R&D Systems, AF1062, 1:500), VE-cadherin (R&D Systems, AF1002, 1:100), CD45 (eBioscience, 25-0451, 1:400), Troponin I (Abcam, ab56357, 1:200), E-cadherin (Cell Signaling, 3195, 1:100), cytokeratin 19 (CK19, Developmental Studies Hybridoma Bank, TROMA-III, 1:100), HNF4a (Cell Signaling, 3113s, 1:1000), Ki67 (Thermo Scientific, RM-9106-S0, 1:200), GS (Abcam, ab49873, 1:1000), p21 (Abcam, ab188224, 1:500). Immunostaining images were acquired using a Zeiss confocal microscope (LSM710) or a Nikon A1 confocal microscope.

Mice received an intraperitoneal injection of 10 mg kg⁻¹ EdU (Invitrogen, A10044) at the indicated time. To detect the proliferation of fibroblasts in the heart, mice were sacrificed 24 h after the EdU injection. Cryosections were obtained and incubated with primary and secondary antibodies as described above, and EdU signal was developed using the Click-iT EdU Alexa Fluor 647 Imaging Kit (Life Technologies, C10340) following the manufacturer's instructions. Images were acquired using a Zeiss confocal microscope (LSM710) or a Nikon A1 confocal microscope.

Sirius Red staining was designed to assess fibrotic tissue formation after injury and was performed as described previously (Liu et al., 2019). Cryosections were fixed in 4% PFA for 15 min and washed in PBS for 15 min, then fixed overnight in Bouins' solution (5% acetic acid, 9% formaldehyde, 0.9% picric acid). Afterwards, sections were stained with 0.1% Fast Green (Fisher Scientific) for 3 min and incubated in 1% acetic acid for 1 min followed by incubation with 0.1% Sirius Red (Sigma-Aldrich) for 2 min. Sections were rinsed with tap water before incubation into the staining solution. Finally, slides were dehydrated in 100% ethanol twice, cleared in xylene, and mounted with resinous medium. Images were obtained on an Olympus microscope (BX53).

MI was performed in mice as described previously (Li et al., 2018). In detail, mice were anaesthetized with isoflurane and their chest hair was shaved. Mice were ventilated and anaesthetized by tracheal intubation, which was connected to an isoflurane gas machine and respiratory machine (Harvard Apparatus). An incision of ∼1 cm was made between the third and fourth ribs to expose heart. Then the left anterior descending coronary artery was ligated with a suture in the upper third region. The incision was then closed after surgery. The mice were supplied with oxygen for a few minutes until the mice woke up, and they were kept in a warm environment until they displayed normal behavior.

MI-reperfusion injury was performed as described previously (He et al., 2017). In brief, mice were anaesthetized as described for MI injury. Afterwards, the chest skin was disinfected and an incision was made to expose heart. Then, the left anterior descending coronary artery was ligated with a suture and a hose was inserted in between the suture and the ligated site for 30 min. After ligation, the left anterior descending coronary artery was re-perfused by removing the suture and hose. The incision was then closed after surgery. The mice were supplied with oxygen until the mice woke up, and they were kept warm until they recovered normal behavior.

For the CCl₄-induced acute injury, mice were intraperitoneally injected with a single dose of 1 μl/g body weight CCl₄, with the CCl₄ being dissolved at 1:3 in corn oil.

For the DDC diet, mice were fed with standard mouse diet (Harlan Teklad, 5015) containing 0.1% DDC (Sigma-Aldrich).

For p21 overexpression detection, we isolated hepatocytes from the liver of Alb-CreER;R26-p21-GFP mice treated with tamoxifen or oil. Hepatocytes were isolated by standard two-step collagenase perfusion as described previously (Pu et al., 2016). Briefly, mice were anaesthetized and an incision was made in the lower abdomen to expose the liver with inferior vena cava and portal vein when the mouse was no longer responsive. A cannula was inserted into the inferior vena cava and secured with a suture around the cannula. The portal vein was cut and the mouse liver was perfused with perfusion medium for 10-15 min. The liver was next perfused with medium containing collagenase type I (Invitrogen) for 10 min. After adequate digestion, the liver was dissected in a Petri dish and passed through a 70 mm filter followed by centrifugation for 2 min at 50 g. Non-hepatocytes cells remained in the supernatant, which was discarded, and the hepatocytes pellet was re-suspended in DMEM medium. The above washing step was repeated three times to remove non-hepatocytes. Washed cells were further purified by Percoll. Viable cells remaining at the bottom of the Percoll were washed and collected.

Total RNA was extracted from hepatocytes isolated from Alb-CreER;R26-p21-GFP mice treated with tamoxifen or oil. Hepatocytes were lysed with Trizol, and total RNA was extracted according to the manufacturer's instructions (Invitrogen). Then, RNA was reverse transcribed into cDNA using a Prime Script RT kit (Takara). The SYBR Green qPCR master mix (Applied Biosystems) was used and quantitative RT-PCR was performed on a StepOnePlus Real-Time PCR System (Applied Biosystems). Gapdh was used as internal control. The primers for p21 and Gapdh RT-PCR are listed below: Gapdh 5′-TTGTCTCCTGCGACTTCAAC-3′ (forward), 5′-GTCATACCAGGAAATGAGCTTG-3′ (reverse); p21 5′-GGAACATCTCAGGGCCGAAA-3′ (forward), 5′-CAATCTGCGCTTGGAGTGAT-3′ (reverse).

AAV8-TBG-Dre virus was generated by Taitool Bioscience, Shanghai. Mice at the age of 8 weeks were induced with AAV8-TBG-Dre virus or AAV8-TBG-null virus (AAV8-control). The virus was diluted in PBS. Virus was administered by tail vein injection at a dose of 3×10¹¹ once a week for 2 weeks. The tail vein injection volume was limited to 200 μl. Mice were given a 2-week washout period before being placed on a DDC diet.

Genomic DNA was extracted from the mouse tail. Tissues were lysed with Proteinase K overnight at 55°C, followed by centrifugation at maximum speed (21,000 g) for 8 min to acquire supernatant with genomic DNA. DNA was deposited by adding isopropanol, and washing in 70% ethanol. All mice were genotyped with specific PCR primers. See Table S1 for a list of genomic PCR primers.

To determine ejection fraction (EF) as well as fraction shortening (FS), echocardiography was performed on mice at 1 week before MI and 1 week, 4 weeks and 8 weeks after MI, using a digital ultrasound system (Vevo2100 Imaging System, Visual Sonics). The mice were measured under isoflurane anesthesia. All parameters were analyzed using original parameters and accompanying software.

Tissue samples were randomized and blinded for analyzers. Data are presented as mean±s.e.m. Statistical analysis was performed by two-tailed unpaired Student's t-test for comparison of differences between two groups. P<0.05 was considered to be statistically significant.

We thank Dr Guoqiang Gu for sharing CK19-CreER, and Shanghai Biomodel Organism Co., Ltd. for mouse generation.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.183830.reviewer-comments.pdf