Human Liver Regeneration Is Characterized by the Coordinated Expression of Distinct MicroRNA Governing Cell Cycle Fate



In the absence of adequate compensatory regeneration, overwhelming liver damage can cause acute liver failure (ALF) and death without emergent liver transplantation (LT). Auxiliary LT produces satisfactory outcomes in this setting, with the prospect of native liver regeneration sustaining long-term survival. Since animal models only partially recapitulate human liver regeneration, we investigated the molecular mechanisms controlling it in this unique LT setting, as an exemplar of human liver regeneration. We demonstrate coordinated changes in expression of microRNA (miRNA) during regeneration that drive proliferation, innate immunity and angiogenesis. In contrast, failed regeneration in a similar cohort is associated with distinct miRNA enforcing cell cycle inhibition and DNA methylation. The miRNA expression associated with successful or failed regeneration when recapitulated in vitro, triggered expression of cardinal regeneration-linked genes promoting cell cycle entry or inhibition, respectively. Furthermore, inhibition of miRNA 150, 663 and 503, whose downregulation is associated with successful regeneration, induced cell proliferation which a key determinant of successful regeneration. Our data indicate that human liver regeneration may be orchestrated by distinct miRNA controlling key regeneration-linked processes including hepatocyte proliferation. To our knowledge this is the first characterization of molecular processes associated with human liver regeneration.


acute liver failure


auxiliary liver transplantation




hierarchical cluster analysis




liver transplantation; miRNA, microRNA


non-regeneration group


principal component analysis


partial hepatectomy


regeneration group


Among human solid organs, the liver retains an exceptional ability to regenerate tissue mass and organization due to the proliferative capacity of fully differentiated hepatocytes, thereby maintaining vital life-supporting function after injury [1]. In the clinical setting, regeneration can occur after an overwhelming toxic insult (such as drug poisoning or viral infection) that causes the critical illness syndrome of acute liver failure (ALF) [2]. Failure of regeneration in this setting leads to death without emergent liver transplantation (LT). Unregulated chronic regeneration can lead to cirrhosis and cancer, both of which have a significant impact on global health [3]. Hitherto, it has not been possible to systematically characterize human liver regeneration at a molecular level or investigate the pathways regulating it.

Animal studies, using the rodent partial hepatectomy (PH) model, have elucidated some of the underlying mechanisms mediating liver regeneration [4]. It appears to be initiated by a variety of inflammatory and innate immune activation pathways. TNF, complement (C3a and C5a), TLR and platelet-derived serotonin have all been implicated [5]. These act on specialized resident antigen presenting cells, which elaborate the pivotal cytokine IL-6. This binds to its cognate receptor on hepatocytes to drive cell cycle entry. This early phase is characterized by expression of key cell cycle genes and leads to hepatocyte proliferation [6]. This is accompanied by downregulation in expression of important cell cycle inhibitors like p21 and Tob1 [7, 8]; their inappropriate expression being linked with failure of liver regeneration [8, 9]. A degree of overlap also exists between proliferative pathways implicated in rodent fetal liver organogenesis and adult liver regeneration in the PH model [10].

MicroRNA (miRNA) can regulate the function of hundreds of target messenger RNAs and therefore gene products [11]. As such, they represent ideal candidates for regulating regenerative processes that require rapid and large-scale changes in gene expression patterns. Investigation of their role in liver regeneration using animal models has suggested a role for selected miRNA during the early proliferative phase and the termination phase of regeneration [12, 13].

However, rodent PH models do not entirely recapitulate human liver regeneration in the clinical context. Even in human PH, regeneration often occurs in the context of underlying liver disease, which is lacking in rodent PH. Robust animal models for the regeneration that occurs in the context of human ALF and chronic liver diseases do not currently exist.

We investigated human liver regeneration using the unique model of auxiliary liver transplantation (ALT) [14]. Emergent LT for ALF usually entails the removal of the entire native liver and its replacement with the transplanted organ. Selected patients in this setting can, however, undergo ALT, where a native damaged liver remnant is left in situ adjacent to the transplanted liver. Patients are selected for ALT using King's College criteria for transplantation in ALF. In addition they are typically under 40 years of age, present with an acute rather than sub-ALF phenotype, the etiology is usually drug/toxin induced or seronegative and they are hemodynamically stable. We have previously shown that the native liver can be induced to regenerate in approximately 70% of patients by weaning immunosuppression (IS) and that successful human liver regeneration appears to be driven by hepatocyte proliferation in this setting [14, 15].

We investigated the molecular pathways regulating human liver regeneration in patients with ALF who underwent ALT, using sequential archived tissue samples obtained from their regenerating livers.

We show for the first time in humans, that directed and specific expression of miRNA occurs during successful and failed human liver regeneration, indicating that specific miRNA regulating cell cycle, innate immunity and angiogenesis may regulate it. Furthermore, our findings indicate that miRNA profiling may lead to the discovery of biomarkers that predict the outcome of liver regeneration in a clinical setting of severe liver failure.

Materials and Methods

Patient samples

We analyzed archived liver biopsy samples from 11 patients who had undergone ALT for ALF (Figures 1A and S1). The biopsies in this series were performed between January 1995 and February 2004. They were not protocolized by time post-ALT and were, with the exception of the T = 1 biopsy obtained after arterial reperfusion, carried out to assess for regeneration. These patients were selected on the basis of availability of biopsy samples amenable to RNA extraction and the availability of imaging studies characterizing regeneration outcome. All samples analyzed were obtained prior to institution of IS withdrawal.

Figure 1.

Successful and failed regeneration in ALT. (A) Demographics for the seven patients (R1–7) in the regenerator group (RG) and the four patients (NR1–4) in the nonregenerator group (NRG). Representative CT (B) and HIDA (C) scans for a patient in the RG at early (i), mid (ii) and late (iii) phases during regeneration compared to a patient in the NRG (late phase, B and C, iv) showing no regeneration. Histology showing successful and failed regeneration in the RG (D), i–iii and NRG (E), i–iii, respectively. A = auxiliary (graft) liver; H = hepatocyte foci; N = native liver.

RNA isolation

For isolation of RNA from FFPE blocks, 800 μL of Xylene was added to 10 μm section of tissue. Alternatively, FFPE slides were incubated in Xylene for 10 minutes and transferred into ethanol. Sections were removed and dried for 10 minutes at 55°C. RNA was isolated using the High pure FFPE RNA microkit (Roche Diagnostics, Hertfordshire, UK).

GeneChip® miRNA 1.0 Arrays

Five hundred nanograms total RNA from each sample was enriched using YM-100 columns (Millipore, Watford, UK) and RNA was labeled using the FlashTag™ Biotin RNA Labeling Kit (Genisphere, Hatfield, PA). GeneChip miRNA 1.0 Arrays (Affymetrix, Wooburn, UK) were processed using the GeneChip Hybridization, Wash, and Stain Kit, with fluidics script FS450_0003. Scanning was performed using Affymetrix Command Console Software. Microarray data are MIAME compliant and has been deposited in the GEO repository (accession number: GSE36146).

Transduction of human primary hepatocytes or HUH-7 cells using HIV lentiviral vectors

Human primary hepatocytes (1 × 106 cells/well, Invitrogen, Paisley, UK) were seeded in six-well plates. Cells were infected with miRNA expressing lentiviral vectors at an multiplicity of infection (MOI) of 10 and cultured at 37°C, 5% CO2. HUH-7 cells (4 × 105 cells/well) were seeded in six-well plates, infected with miRNA expressing lentiviral vectors at an MOI of 10 and cultured at 37°C, 5% CO2 in media containing puromycin, in order to select lentivirus-infected cells. mCherry or GFP expression was observed using a Leica fluorescence and images were captured using a Leica D-Lux3 LMS camera.

Ethical approval

All FFPE samples used in this study were dated before September 2006 and had local ethical committee approval for use in the study.

Details of the miRNA array analysis, lentiviral vectors, fluorescence-activated cell sorting, qPCR/RT-PCR gene expression analysis and EdU cell proliferation assay are provided in the Supporting Information.


Successful and failed regeneration in ALT

We investigated the molecular pathways regulating human liver regeneration in 11 patients who had undergone emergent ALT for ALF (Figure S1). All patients had undergone ALT based on well-established clinical prognostic criteria reflecting a low probability of spontaneous regeneration, recovery and survival without LT [14]. As described, patient and graft selection were in accordance with established protocols for ALT. Seven of these patients subsequently demonstrated successful regeneration of their native livers (RG; denoted R1–7) in association with complete withdrawal of IS. The second group consisted of four patients who also underwent ALT but failed to demonstrate native liver regeneration and evidenced no anatomical, vascular or other causes for this failure (NRG; denoted NR1–4). The patient characteristics and time course of successful or failed regeneration are presented in Figure 1A. All patients were treated with standard calcineurin inhibitor-based IS and there were no differences in early graft dysfunction or rejection rates between groups (data not shown). Regeneration is typically quantified using three different modalities: volume expansion of the native liver (and involution of the transplant) is assessed using conventional radiographic imaging (Figure 1B), liver functional recovery is assessed by HIDA scan, a form of nuclear isotope scan measuring bile excretion (Figure 1C) and hepatocellular regeneration using histology (Figures 1D and E). All three modalities were consistent in defining the endpoint of successful liver regeneration in the RG (Figures 1Bi–iii, Ci–iii and Di–iii) and failed regeneration in the NRG (Figures 1Biv, Civ and Ei–iii).

miRNA expression distinguishes phases of liver regeneration

We compared miRNA expression in samples obtained during reperfusion at transplantation (T1) and subsequent sequential biopsies (the second; T2 and third; T3) in patients in the RG, using miRNA microarray analysis. Five of the T2 and one of the T3 biopsies were performed opportunistically at the same time the grafts were biopsied to assess function. Six of the T3 biopsies were performed specifically to assess morphology in the context of potential weaning from IS after demonstration of an increase in volume of the native liver on CT scans. Five of the seven patients progressed to the initiation of IS withdrawal in time frames that reflected the pace of regeneration assessed by volume of the native liver and between 6 and 21 months after transplantation. IS withdrawal was deferred in two patients on the basis of clinical choice rather than differentiating characteristics. Withdrawal of IS was completed in all patients. There was no correlation between the pace of regeneration and age of patient or the etiology of the ALF (Figure 2A). Despite temporal heterogeneity in sample acquisition, supervised hierarchical cluster analysis (HCA) for miRNA expression revealed that a greater similarity of expression existed within each sequential biopsy sample [1-3] than between them for all seven patients (Figure 2B). The cladogram shown above the heat map confirms this interpretation. The analysis clearly distinguished the three groups by miRNA expression, indicating that these groups were distinct and that a consistency of miRNA expression existed within each group. The list of miRNA identified in this analysis and their respective p values are shown in Table S1.

Figure 2.

miRNA expression distinguishes phases of regeneration in the RG. (A) Time of acquisition of sequential biopsies 2 and 3 for miRNA analysis for seven patients (R1–7) in the RG. T1 is always day 0: T2 is day of second biopsy and T3 is day of third biopsy. (B) Hierarchical cluster analysis for all patients with sequential biopsies 1–3 color-coded in the legend. The cladogram shown above demonstrates degree of relatedness between samples by miRNA expression; line-length inversely correlates with similarity. (C) Principal component analysis for the same data set using the same color code (p < 0.0265). miRNA list and associated p values shown in Table S1.

Principal component analysis (PCA) resolves a multidimensional data set by identifying key variables that explain the observed differences. This analysis resolved the data set into three principal components represented in the three-dimensional scatter plot (Figure 2C) and confirmed the clear segregation of sequential biopsies 1–3 by miRNA expression profiles. The separation of biopsies 1–3 by both HCA and PCA indicates that these are functionally distinct in relation to miRNA expression. They also confirm that a commonality of miRNA expression exists within each biopsy group for all patients. The PCA also places T1 on the dominant axis of variance and indicates a closer relationship between T2 and T3 than either of these with T1, an interpretation supported by the cladogram in the HCA. These data are consistent with regeneration commencing at or after T1 and then proceeding through T2 and T3.

Coordinated miRNA expression during early liver regeneration

Since early proliferative events are fundamental to successful regeneration, we focused our analysis on regulation of miRNA expression between T1 and T2 for each patient in the RG. This comparison also minimizes the impact of uncontrolled variables associated with case mix and clinical status. The cluster analysis identified consistent changes in the expression of selected miRNA for all seven patients between T1 and T2 (Figure 3A). Representative examples of the histology samples used in this analysis to generate the miRNA signatures at T1 and T2 are shown in Figure 3(B). This shows severe hepatocyte damage and depletion at T1 (Figure 3B-i) and early hepatocyte regeneration at T2 (Figure 3B-ii). The data set of up- and downregulated miRNA was analyzed using MetaCore (GeneGo Inc.) to delineate the genetic pathways regulated by the miRNA we identified. This algorithm recognizes coregulated components of pathways or biological processes that are impacted on by increased or decreased expression of specific miRNA. Z- and G-scores for any given network represent the number of genes of interest represented in the network and the higher the score, the greater the relevance.

Figure 3.

Coordinated miRNA expression and liver regeneration. (A) Cluster analysis comparing up- and downregulated miRNA between T1 and T2 for the RG (p < 0.01). (B) Representative histology at T1 (i) and T2 (ii) from one patient showing regenerative foci; H. (C) Metacore analysis identified a statistically significant network of downregulated miRNAs (i), and known target genes for individual miRNA (ii). (D) Statistically significant network of upregulated miRNA (i) and known target genes for individual miRNA (ii). Regeneration-linked target genes subdivided into cell proliferation (●), apoptosis (◊), angiogenesis / VEGF signaling (○) and innate immune activation (□) pathways.

Examining downregulated miRNA between T1 and T2, MetaCore identified a network of coregulated species incorporating miRNA-503, miRNA-23a, miRNA-150, miRNA-663 and miRNA-654 of high statistical significance (p = 4.95 × 10−19; Figure 3C-i). Reduced expression of these miRNA, by reversing specific translational inhibition, increases gene expression of key components of pathways implicated in regeneration in animal models (Figure 3C-ii). Reduced expression of miRNA-150 is known to increase expression of TNFα, a key initiator of liver regeneration [16], the antiapoptotic gene Survivin [17], also implicated in liver regeneration and c-Myb, a driver of cell proliferation. Reduced expression of miRNA-503 leads to increased expression of pivotal cell cycle genes including cyclin D1, E1, E2, F, Wee1, CDC25A and CHK1 [18]. Similarly reduced miRNA-663 expression drives increased expression of TGFβ1, JunB/JunD and increased AP-1 activity [19], important mediators of early liver regeneration in animal models. Furthermore, reduced expression of miRNA-23a favors the regenerative, rather than the proapoptotic activity of TNFα [20]. In summary, the pathways activated by coordinated downregulation of these miRNA include those known to be critical for the early phase of liver regeneration by promoting hepatocyte proliferation.

Analysis of upregulated miRNA between T1 and T2 revealed a highly significant network consisting of miRNA-126, miRNA-130a, miRNA-20a, miRNA-520e and miRNA-330 (p = 5.57 × 10−19; Figure 3D-i and ii). Significantly, miRNA-126 is a highly conserved inducer of angiogenesis through enhancement of vascular endothelial growth factor signaling [21, 22]. In addition, increased miRNA-520e expression is known to inhibit expression of the membrane-bound complement regulator CD46 and thereby increase expression of the complement components C4b and C3b [23], key mediators of early liver regeneration in animal models. We validated miRNA expression changes identified in all array experiments and quantified these formally by carrying out targeted quantitative PCR for miRNA as shown in Supplementary Figure S2.

Distinct miRNA expression during failed regeneration

In order to test the specificity of the miRNA profiles we identified in early successful liver regeneration, we carried out HCA on T1 and T2 in the NRG. The cluster analysis of miRNA expression profiles for T1 and T2 in the NRG is shown in Figure 4A. Consistent changes in expression among specific miRNA were observed between sequential biopsies T1 and T2 for all patients. The time of acquisition of biopsy 2 is shown in Figure 4B. Representative examples of the histology samples used to generate the miRNA signatures from T1 and T2 in the NRG are shown in Figure 4(C). At T1 this shows a comparable hepatocyte injury and depletion to that in the RG (Figure 4C-i). However, by T2 there is no comparable hepatocyte recovery and regeneration (Figure 4C-ii). Analysis of downregulated miRNA in this group revealed a network consisting of miRNA-152, DNMT1, AML1/ETO fusion protein, c-FLIP (L), ITGAL (p = 1.59 × 10−4; Figure 4D-i and ii). Reduced expression of miRNA-152 leads to increased expression of DNA methyltransferase 1 (DNMT1). Expression of this gene leads to global DNA hypermethylation and aberrant gene expression [24], resulting in inhibition of proliferation and regeneration.

Figure 4.

Coordinated expression of distinct miRNA during failed regeneration. (A) Cluster analysis comparing up- and downregulated miRNA between sequential biopsies T1 and T2 for the NRG (p < 0.01). (B) Time of acquisition of sequential biopsies obtained for miRNA analysis. T1 is always day 0, T2 is day of second biopsy. (C) Representative histology at T1 (i) and T2 (ii) from a patient showing no regeneration. (D) Statistically significant networks of downregulated miRNA (i) and (E) upregulated miRNA (i) with their respective target genes shown in (D-ii) and (E-ii). Genes are subdivided into those promoting DNA hypermethylation (○), inhibiting cell proliferation (●) or innate immune inactivation (□).

Analysis of upregulated miRNA between T1 and T2 for this group revealed a network consisting of miRNA-150, miRNA-let-7i, HBII-52 snoRNA, c-Src, C/EBPbeta (p = 6.48 × 10−9; Figure 4E-i). As discussed previously, increased miRNA-150 expression leads to reduced expression of TNFα, Survivin and c-Myb, all key mediators of liver regeneration. Increased let-7i leads to decreased TLR4 expression [25], thereby negatively impacting on liver regeneration and leading to cell cycle inhibition.

Initial miRNA expression distinguishes successful and failed liver regeneration

We also analyzed miRNA expression at the time of ALT (T1), directly comparing the RG and NRG. As previously mentioned, all patients were selected for ALT using standard clinical prognostic criteria indicating comparable levels of liver damage and likelihood of spontaneous regeneration. This was corroborated by histologically comparable levels of hepatocyte damage and loss at T1 between the RG and NRG (Figures 1D-i and E-i). Notwithstanding this, distinct subsets of miRNA were expressed at T1 in the RG and NRG, as shown by the segregation of these two groups by PCA analysis (Figure 5A). The list of miRNA down- and upregulated between the RG and NRG at T = 1 are shown in Figure 5Bi and 5Bii, respectively.

Figure 5.

miRNA expression at the time of ALT distinguishes successful and failed liver regeneration. (A) Principal component analysis comparing miRNA expression at T1 in the RG (R1–7) versus the NRG (NR1–4). p < 0.01. (B) Gene list of top 30 downregulated (i) and upregulated (ii) miRNA and associated p values. (C) Metacore analysis identified a statistically significant network of downregulated miRNA (i) and known target genes/associated pathways (ii). (D) Statistically significant network of upregulated miRNA (i) and known target genes/associated pathways (ii).

Metacore analysis for miRNA that were downregulated in the RG compared to the NRG at T1 revealed a statistically significant network consisting of miRNA-200b, miRNA-183, ZEB1 and SP1 (p = 9.21 × 10−8, Figure 5C-i and ii). miRNA-200b is a well-established tumor suppressor that promotes apoptosis, as well as inhibiting cell proliferation, cancer progression and angiogenesis. Hypoxia and HIF-1α stabilization has been shown to inhibit miRNA-200b, resulting in increased expression of several proangiogenic genes [26]. Reduced expression of miRNA-200b also leads to downregulation of fibrogenesis [27] and apoptosis, as well as upregulation of cell proliferation via increased expression of PKC and downregulation of cell cycle inhibitors [28].

Analysis of upregulated miRNA in this group identified a highly statistically significant network composed of miRNA-27a, miRNA-494, miRNA-1224 and miRNA-149 (p = 4.15 × 10−23, Figure 5D-i and ii). Upregulation of miRNA-27a inhibits apoptosis [29] and drives cell proliferation [30, 31] and angiogenesis [32]. Upregulation of miRNA-494 leads to reduced expression of Ca2+/calmodulin-dependent protein kinase (CaMKII), resulting in increased cell proliferation and apoptotic resistance through upregulation of PI3-Kinase, Akt and MDM2 [33]. miRNA-494 also targets PTEN, leading to decreased caspase-3/7 activity and increased cell proliferation [34]. Finally, increased expression of miRNA-149 is associated with reduced apoptotic activity [35], as well as increased cell proliferation through enhanced levels of Jun, c-Myc [36, 37] and inhibition of E-cadherin [38].

The distinct miRNA expression patterns we observed between the RG and NRG at the time of ALT indicate that even at this early stage, both groups were functionally discrete and that regeneration had already commenced at T1 in the RG. Furthermore, the dominance of coregulated miRNA that determine cell cycle fate, apoptosis and angiogenesis suggests that these may directly regulate human liver regenerative outcome.

Regeneration-linked miRNA regulate cell cycle associated gene expression

We established a lentiviral transduction system, in order to investigate whether the miRNA associated with successful or failed regeneration are capable of regulating specific target gene expression in hepatocytes. Primary human hepatocytes were used for gene expression analysis as more tractable transformed cell lines often have aberrant cell cycle machinery. Vectors that either specifically bind and functionally interfere with a given miRNA or increase its expression were used, in order to recapitulate the changes we observed during early human liver regeneration. We selected miRNA that were identified through MetaCore as belonging to a network regulating target genes of relevance to regeneration. Cells expressing fluorescent-tagged miRNA were evident by fluorescence microscopy (Figure 6A, right panel) and were sorted by FACS (Figure 6A, left panel). Lentiviral vector expression was confirmed by PCR (Figure 6B). Target gene expression was normalized to expression in cells transduced with a scrambled vector to minimize the effects of transduction, lentiviral expression and cell sorting on gene expression. Although primary hepatocytes are highly resistant to cell cycle entry in vitro [39], inhibition of miRNA-150, miRNA-663 or miRNA-503 led to statistically significant increased expression of minichromosome maintenance gene 2 (MCM2), a robust marker of cell cycle progression [40] and liver regeneration [41] (Figure 6C). While inhibition of miRNA-150 and miRNA-503 expression did not significantly alter expression of the cardinal cell cycle inhibitor p21, miRNA-663 inhibition led to a statistically significant reduction in expression of p21 (Figure 6D). Downregulation of p21 expression is directly linked with regenerative capacity in animal models [7]. Therefore functional recapitulation of the miRNA changes we observed, which belonged to a network associated with successful liver regeneration, led to changes in hepatocyte gene expression that indicated they were cell cycle competent. This conclusion is supported by the observation that inhibition of miRNA-503 expression also led to statistically significant increased expression of its known target genes for posttranscriptional regulation; cyclin D1, cyclin F, CDC25A and CHK1 (Figure 6E). These cell cycle genes promote proliferative regeneration in animal studies [42]. Inhibition of expression of miRNA-150 led to statistically significant increased expression of Survivin and TNFα while inhibition of miRNA-663 led to increased expression of TGFβ (Figure 6F). These gene products are known targets for regulation by their respective miRNA and as previously described, are implicated in liver regeneration in animal models. On the other hand upregulation of miRNA-150 expression, associated with failed regeneration in our study, led to a statistically significant increase in expression of p21, a principal inhibitor of liver regeneration [9] (Figure 6D).

Figure 6.

In vitro recapitulation of miRNA expression associated with successful and failed regeneration specifically induced cell cycle promoting and inhibiting genes, respectively. (A) FACS profiles (left panel) or fluorescent images (right panel) of primary hepatocytes transduced with: (i) scrambled control vector (□); (ii) miRNA-150 inhibitor (□); (iii) miRNA-663 inhibitor (□), (iv) miRNA-503 inhibitor (□) and (v) miRNA-150 expressor (□); compared to untransduced primary hepatocytes (▪). All constructs coexpressed mCherry except (v), which expressed GFP. Bar (math formula) represents the sorted population. (B) RT-PCR for puromycin expression in sorted samples A–E. pBABEpuro vector (+) and uninfected primary hepatocytes (-) were used as controls. Arrow (→) represents 600 bp marker. Expression of (C) MCM-2 and (D) p21 by qPCR for each miRNA. Expression of known targets for (E) miRNA-503; (F) miRNA-150 and miRNA-663. All expression levels were normalized to scrambled control vector. qPCR data are representative of three different experiments. Error bars indicate SD. *p < 0.05, **p < 0.01, ***p < 0.001.

Inhibition of miRNA-150, miRNA-503 and miRNA-663 promotes proliferation in vitro

We have demonstrated that mimicking miRNA expression changes associated with liver regeneration in primary hepatocytes, induced the expression of cell cycle genes known to be associated with its initiation. In order to test whether these miRNA can directly influence cell cycle, we expressed lentiviral vectors interfering with the function of miRNA-150, miRNA-503 and miRNA-663 in the model cancer cell line HUH-7. After 2 weeks in selection medium in vitro, flow cytometry and immunofluorescence (Figure 7i-ii) demonstrate a homogeneous and high expressing transduced population. Proliferation was assessed using EdU (5-ethynyl-2′-deoxyuridine), incorporated during DNA synthesis and compared to proliferation in cells transduced with a scrambled control vector. EDU incorporation over 5 hours in cells expressing the control vector was <10%, representing the background proliferation in this cell line. Inhibition of all three miRNA, associated with both successful liver regeneration and induction of cell cycle genes, led to between 300% and 450% increase in EDU incorporation over 5 hours (Figure 7-iii). It was not possible to test the effect of increased miRNA-150 expression, as the protocol for EDU incorporation interfered with the detection of GFP and hence of the transduced cell population. Furthermore, our in vitro study did not permit the analysis of cell cycle phenotype resulting from coordinated changes in expression of multiple miRNA that we observed during regeneration in vivo.

Figure 7.

Inhibition of three miRNA downregulated during liver regeneration induced cell proliferation in Huh7 cells. (i) Flow cytometry profiles showing transduced HUH-7 cells (□) compared to untransduced (▪) cells. HUH-7 cells were infected with: (A) scrambled control vector; (B) miRNA-150 inhibitor; (C) miRNA-663 inhibitor; and (D) miRNA-503 inhibitor. Stable expression was obtained by maintaining cells under puromycin selection. All constructs coexpressed mCherry; bar (math formula) represents the proportion of positively transduced cells. (ii) Fluorescence images of samples A–D. (iii) Flow cytometry dot plots showing EdU incorporation in transduced cells over 5 hours. EdU positive cells represent proliferating cells. Results are representative of three independent experiments.


The robust and specific miRNA signatures we demonstrate in association with successful or failed human liver regeneration represent, to our knowledge, the first characterization of molecular changes associated with human liver regeneration. The consistency of miRNA expression we observe within sequential biopsies 1–3, despite disparate intervals from time of ALT, most likely reflect differences in regenerative rate between individuals. These differences may arise from variations in cause, extent and timing of liver injury, as well as the impact of additional clinical cofactors (e.g. inflammation, drugs, underlying liver function and age). In this regard, regeneration in a clinical context differs markedly from experimental models of regeneration where the genetic background, timing and extent of liver injury are tightly controlled. These findings may also have been impacted on by the clinical modalities used to define regenerative outcome, which quantify three different aspects of liver regeneration: hepatocyte mass, liver volume and bile excretion. While these modalities were consistent in defining the phenotype of successful or failed regeneration their kinetics differed during liver regeneration. Despite these caveats and the scarcity of patients, a more systematic analysis of sequential native liver biopsies using standardized time points may shed further light on the possible role played by specific miRNA in phases of liver regeneration.

All patients received standard calcineurin inhibitor-based IS, arterial perfusion was maintained throughout and minimal or no portal venous occlusion was employed. Furthermore, all auxiliary grafts met standard liver transplant criteria for graft to recipient volume and weight ratios. Therefore, it is unlikely that reperfusion injury, graft volume or IS impacted significantly on the differences in early miRNA expression we observed between the RG and NRG. The differences in miRNA expression between the RG and NRG at the time of LT, prior to the institution of IS or the development of reperfusion changes, indicate that they reflect regeneration status and that regeneration is initiated prior to institution of IS withdrawal. The impact of other nonnative liver associated events (including graft-related events like ACR) is difficult to control for in our analysis. Therefore while patient characteristics were carefully matched between the RG and NRG, such events may influence the changes in native liver miRNA expression we observed.

Specific changes in miRNA expression occur during successful regeneration that regulate cell proliferation, innate immune activation and angiogenesis. All of these processes are implicated in successful liver regeneration in animal studies [4, 43]. Our data also indicate that failure to regenerate in humans may result from active cellular processes, including cell cycle arrest and DNA methylation, which are also regulated by specific miRNA. The differences in miRNA expression between regenerating and nonregenerating livers at the time of ALT also indicate that early regeneration may be driven by changes in expression of miRNA regulating proliferation, cell survival, innate immune activation and angiogenesis. This conclusion is supported by the description of changes in miRNA expression accompanying partial rat hepatectomy [13] and the finding that their overall absence impairs liver regeneration in this model system [12]. miRNA represent ideal candidates for regulating regeneration, which requires rapid, large-scale and coordinated changes in gene expression. Differences between the miRNA we identify and those associated with rodent PH models of liver regeneration may reflect species differences as well as differences in regenerative phenotype in relation to rate, presence of inflammation and ALF.

Our experiments cannot delineate whether the miRNA signals we obtain arise predominantly from hepatocyte or other lineages. However, successful liver regeneration in animal models and in this human transplant model is predominantly driven by hepatocyte proliferation [4, 15]. Therefore, the dominant proliferative signature we identify indicates that the miRNA profiles we obtain derive predominantly from changes in hepatocyte expression. This conclusion is supported by the evidence that specific miRNA associated with regeneration can promote both the expression of relevant cell cycle genes and cell proliferation in human hepatocytes.

Understanding the processes that regulate liver regeneration may lead to treatments that optimize its potential in liver failure, thereby minimizing the need for LT. Furthermore, should regulation of regeneration by miRNA be shown to be more generalized, this may allow regeneration to be enforced in tissues that currently lack this capability, such as the brain or spinal cord after injury.

Since tractable miRNA signatures were obtained from limited archived clinical material, our findings indicate that all comparable archives may be amenable to interrogation using such “state of the art” technologies. Our findings also indicate that clinically tractable miRNA-based biomarkers for liver and general regeneration may be feasible. Given the current reliance on prognostic models that are more than 20 years old [44], our findings may lead to the development of novel biomarkers that better predict both the potential for spontaneous liver regeneration/recovery and transplant need, thereby improving patient survival.


We thank E. Aldecoa-Otalora Astarloa for assistance with the miRNA array experiments; the Dhawan Lab for provision of HUH-7 cells; Winston Vetharoy for help with the FACS experiments; and Alberto Sánchez-Fueyo for helpful comments on the manuscript.

Funding source: This study was funded by the NIHR comprehensive Biomedical Research Centre, Guy's and St. Thomas's NHS Trust and the Higher Education Funding Council for England.


The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.