Potential conflict of interest: Nothing to report.
This work was supported by grants from the National Institutes of Health. Sonya V. Iverson and Edward E. Schmidt were supported in part by the Montana Agricultural Experiment Station. Kristin M. Comstock was supported by the National Science Foundation through an undergraduate summer research fellowship. Infrastructure support was provided by the National Institutes of Health through a Centers of Biomedical Research Excellence grant.
The contributions that de novo differentiation of new hepatocyte lineages makes to normal liver physiology are unknown. In this study, a system that uniquely marks cells during a finite period following primary activation of a serum albumin gene promoter/enhancer-driven Cre recombinase (albCre) transgene was used to investigate birthrates of new hepatocyte lineages from albumin (Alb)-naive precursors in mice. Elapsed time was measured with a two-color fluorescent marker gene that converts from expressing tandem dimer Tomato (tdT; a red fluorescent protein) to expressing green fluorescent protein (GFP) following primary exposure to Cre. The accumulation of GFP and the decay of tdT each contributed to a regular fluorescence transition, which was calibrated in vivo. In normal adults, this system revealed that a steady-state level of 0.076% of all hepatocytes had differentiated within the previous 4 days from albCre-naive cell lineages. In comparison with resting adult livers, the relative abundance of these newborn hepatocytes was elevated 3.7-fold in the growing livers of juveniles and 8.6-fold during liver regeneration after partial hepatectomy in adults. Conclusion: Newborn hepatocyte lineages arising from Alb-naive cells contribute to liver maintenance under normal conditions. Hepatocyte lineage birthrates can vary in response to the liver's physiological status. (HEPATOLOGY 2011;)
Hepatocytes are one of the few differentiated cell types that can replicate DNA and proliferate.1, 2 The mass of a mouse liver increases more than 500-fold between embryonic day 14.5 and postnatal day 56 (P56).3 This dramatic growth is associated with (1) changes in the relative proportions of populations of resident cell types (hepatocyte populations expand to replace hematopoietic cells that migrate to the bone marrow),4, 5 (2) increases in the size of hepatocytes, and (3) increases in the total numbers of hepatocytes in the organ.3, 6 The hepatocyte proliferative component of this growth has been estimated to account for at least a 30- to 100-fold increase in the number of hepatocyte genomes during this developmental transition.3
Hepatocytes in adult animals can also transiently proliferate in response to hepatocyte losses resulting from partial hepatectomy, toxic exposures, or other insults that reduce the number of hepatocytes.7, 8 For example, surgical removal of two-thirds of a rodent's liver induces rapid synchronous entry of nearly all remaining hepatocytes into the cell cycle, which results in full regeneration of liver mass in approximately 10 days.9, 10
Likely hidden within the dramatic backdrop of hepatocyte replication during these periods of proliferative growth are the subtle contributions of proliferation of low-abundance prehepatocyte cells and their differentiation into new proliferative hepatocyte lineages. Currently, it is unclear whether prehepatocyte cells, which are variously called hepatic stem cells, oval cells, bipotential cells, and progenitor/stem cells, are all equivalent or represent different types or degrees of lineage commitment; however, all have in common the ability to differentiate into hepatocytes.2, 11, 12 Previous studies have been unable to detect substantial contributions of prehepatocyte cells to postnatal developmental or acute regenerative growth,2, 13-15 and we are unaware of any previous studies showing a role for differentiation of new hepatocyte lineages in normal liver maintenance. However, levels of prehepatocyte cells can be increased by conditions that chronically compromise hepatocytes while impeding hepatocyte proliferation.2, 12, 16-18 Because most previous studies on prehepatocyte cells relied on expansion of their populations by such means, it is unclear which of their ascribed properties truly reflect characteristics of prehepatocyte cell types in the normal liver and which reflect atypical or intermediately differentiated states that arise in response to the induction conditions.19-21 Clearly, there is a need for improved means of studying the activities of prehepatocyte cell types under conditions that may be relevant to normal human health and exposures.
In this study, we combined existing transgenic mouse lines to develop a sensitive system to quantify the levels of newly differentiated hepatocytes that arise from albumin (Alb)-naive cells. Using this system, we show that newborn hepatocyte lineages contribute to developmental liver growth in normal juvenile mice, to liver maintenance in normal adult mice, and to liver regeneration after partial hepatectomy. Our results suggest that newborn hepatocyte lineages may play important roles in the growth, maintenance, and repair of normal liver.
α-Alb, anti-mouse albumin antibody; AdCre, replication-defective adenoviral vector expressing Cre recombinase; Alb, albumin; albCre, mouse serum albumin gene promoter/enhancer-driven Cre transgene; β-ActEnh, chick β-actin enhancer; CDE, choline-deficient and ethionine-supplemented; Cre, cyclization recombinase; DAPI, 4′,6-diamidino-2-phenylindole; GFP, green fluorescent protein; HNF4, hepatocyte nuclear factor 4; loxP, phage P1 locus of cross-over, the Cre-dependent recombination site; P, postnatal day; ROSACreER, B6;129-Gt(ROSA)26Sortm1(cre/Esr1)Nat/J; ROSAmT-mG, Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J; SEM, standard error of the mean; tdT, tandem dimer Tomato.
Materials and Methods
Mouse Lines and Care Conditions.
All animal care and use protocols were approved by the institutional animal care and use committee of Montana State University. In this article, genetic loci are designated by lowercase italics, and genetic quality follows as a superscript, with each allele separated by a slash. A plus sign designates a wild-type allele. For arbitrarily inserted transgenes, a superscripted 2, 1, or 0 is used to indicate whether animals contained the transgene on two chromosomes, on one chromosome, or not at all. A semicolon separates designations for different genetic loci. For example, a ROSAmT-mG/+; albCre0 mouse was heterozygous for the mT-mG allele of the ROSA locus and did not have an albCre transgene. C57Bl/6J, Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J (ROSAmT-mG),22 and B6;129-Gt(ROSA)26Sortm1(cre/Esr1)Nat/J (ROSACreER) mice,23 and mice bearing the serum albumin gene promoter/enhancer-driven Cre recombinase (albCre) transgene [B6.Cg-Tg(Alb-cre)21Mgn/J] were purchased from Jackson Labs (stock numbers 000664, 007576, 004847, and 003574, respectively). Genomic DNA samples were collected before weaning from all pups, and the genotypes were determined by polymerase chain reaction with previously reported primers.24 All mice were maintained under the following conditions: sterilized feed (PicoLab 5058), water, bedding, and enrichment materials; forced-air caging systems with high-efficiency particulate air filters (Tecniplast); and a 14-hour/10-hour light/dark cycle. Administration of a replication-defective adenoviral vector expressing Cre recombinase (AdCre) was performed as described previously.24 When a choline-deficient and ethionine-supplemented (CDE) diet was used, it consisted of choline-deficient rodent chow pellets (MP Biomedicals) and water supplemented with 0.075% (wt/vol) DL-ethionine (Alfa Aesar).17, 25
Tissue Harvesting, Surgeries, Fluorescence Microscopy, and Data Analyses.
Animals of various ages (which are indicated elsewhere in the text and figures) were sacrificed, and liver samples were fresh-frozen in an Optimal Cutting Temperature medium (Tissue-Tec). For Alb immunostaining, the animals were perfused by cardiac puncture/portal draining with 5 to 10 mL of sterile saline to remove cross-reactive serum Alb from vessels and capillaries. For immunostaining, the following were used: polyclonal goat anti-mouse Alb (A90-134A, Bethyl) and an Alexa Fluor 350-labeled donkey anti-goat secondary antibody (A21081, Molecular Probes), as we described previously26; polyclonal rabbit anti–hepatocyte nuclear factor 4 (anti-HNF4; sc-8987, Santa Cruz)27 and an Alexa Fluor 350–labeled donkey anti-rabbit secondary antibody (A-10039, Invitrogen); or a monoclonal rat Monts-4 antibody28 (generously provided by Dr. M. A. Jutila, Montana State University) and an Alexa Fluor 350–labeled goat anti-rat secondary antibody (A-10093, Invitrogen). Two-thirds hepatectomy was performed (when indicated) as described previously.3, 10 Cryosections (5 μm) were fixed in 75% acetone/25% ethanol and were mounted with Fluoromount-G alone or with Fluoromount-G containing 4′,6-diamidino-2-phenylindole (DAPI; Southern Biotech) as indicated (because of the blue emission of Alexa Fluor 350, DAPI was excluded from all immunofluorescence slides). Monochromatic images were captured digitally on a Nikon Eclipse 80i, a Nikon Eclipse E800, or an Olympus BX60 microscope; for each, standard DAPI (blue), fluorescein isothiocyanate (green), or tetramethyl rhodamine isothiocyanate (red) filter sets were used. Within each experiment, all images were captured with the same microscope and camera system. Some micrographs were electronically enlarged or reduced with Photoshop CS3. We set the scale bars in the figures first by photographing a micrometer scale under each magnification and then by subjecting these images to the same electronic enlargement or reduction used for the biological images. For quantitative pixel analyses, monochromatic images were captured with uniform microscope and camera settings for each color, and no further adjustments were made. The exposure settings for quantitative pixel analyses generally underexposed images to prevent signal saturation by strongly fluorescent cells. Green and red pixels were counted with the Histogram function in Photoshop CS3. For other monochromic images, the whole area shown was adjusted with the Autocontrast function in Photoshop. All monochrome image merging was performed with Photoshop CS3 (see Supporting Fig. 1). For merging, images were captured separately for each color channel and were layered with Photoshop. The blue channel formed the background with 100% opacity, the green channel was next with approximately 50% opacity, and the red channel was the top layer with approximately 40% opacity. The opacity values were chosen empirically to produce the best aesthetic balance of blue, red, and green fluorescence (see Supporting Fig. 1). All layers were merged and were uniformly adjusted with the Autocontrast function in Photoshop. When images were overexposed, a black point was set inside the darkest region of a major capillary with the Photoshop CS3 Levels function. With the exception of the images of newly differentiated hepatocytes, all micrographs were representative of the whole organ. Newly differentiated hepatocytes were distributed in clusters. Images were taken to show these, as indicated in the figure legends. All image adjustments were performed uniformly for the entire image. None of these adjustments qualitatively or quantitatively affected the interpretations or conclusions arising from the data; they instead served useful aesthetic functions and helped with the visualization and enumeration of important biological characteristics of each sample. For the quantification of newborn hepatocytes, arbitrary fields of view were counted for a series of sections from each mouse until at least 30 fields were evaluated and 30 reddish hepatocytes were counted. Statistical analyses used at least three biological replicates (i.e., different animals) for each condition. Graphical data are presented as means and standard errors of the mean (SEMs). Significance was tested with the Student t test.
ROSAmT-mG Fluorescent Switch Kinetics in Hepatocytes.
The ROSAmT-mG allele was developed as a strong and ubiquitously expressed two-color fluorescent marker to distinguish Cre-naive cell lineages from Cre-exposed cell lineages in mice.22 In brief, it is a targeted insertion into the nonessential and ubiquitously expressed ROSA26 locus on chromosome 6.29 Expression levels were augmented by the incorporation of enhancer elements from the chick β-actin gene and from cytomegalovirus.22 Downstream of the promoter is a loxP-flanked modified tandem dimer Tomato (tdT) cistron, which is followed by a modified enhanced green fluorescent protein (GFP) cistron.22 The cassette was engineered to direct both fluorescent proteins to the outer membrane.22ROSAmT-mG has no overt affects on mouse physiology.22 A schematic of ROSAmT-mG is shown in Fig. 1A.
Expression of Cre in cells containing ROSAmT-mG results in the excision of the tdT cistron. This uncovers the GFP cistron and causes an irreversible switch in outer membrane fluorescence from red to green (Fig. 1A).22 Because this is a genetic modification, the fluorescence state of the allele is passed to daughter cells during replication. As such, cell lineages transiently expressing Cre at any point in their history are marked as green by ROSAmT-mG; only truly Cre-naive lineages fluoresce red.
Intravascular inoculation of ROSAmT-mG/+ mice with a hepatocyte-tropic, replication-defective AdCre resulted in roughly synchronous conversion of a mosaic subset of hepatocytes from red fluorescence to green fluorescence in vivo (Fig. 1B).3, 24 In mice harboring both ROSAmT-mG and the albCre transgene (ROSAmT-mG/+;albCre1 mice), which expresses Cre under the control of the alb gene promoter and enhancer,30 all differentiated hepatocytes expressed Cre and fluoresced green (Fig. 2B,D), whereas endothelial cells surrounding blood vessels and capillaries as well as other nonhepatocyte liver cells, remained Cre-naive and fluoresced red (Fig. 2C,D).3, 24, 26 Because GFP labeled the outer membranes of only differentiated hepatocytes in ROSAmT-mG/+;albCre1 mouse livers, green fluorescent images revealed subtle zonal differences in hepatocyte cell size and membrane fluorescence (Fig. 2B,D), which corresponded to differences in the density of the endothelial cell–lined capillary networks observed in the red channel (Fig. 2C). Around portal circulation, hepatocytes were smaller, and the green fluorescence was more intense in comparison with that seen around venous circulation. ROSAmT-mG/+;albCre0 mice, which did not express Cre, had red membranes in all cells, yet the zonal pattern of global fluorescence was still observed with increased intensity around portal circulation (Fig. 2E). This zonal variation in hepatocyte membrane fluorescence is reminiscent of previously reported zonal differences in metabolic activity and gene expression31, 32 and may have a related underlying cause.32 Importantly, with ROSAmT-mG/+ mice, this variation appears to be equivalent for both the red and green markers (as discussed later).
ROSAmT-mG as an In Vivo Chronometer.
We previously observed that during regeneration in ROSAmT-mG/+;albCre1 mice, a rare subset of periportal hepatocytes could be found with a substantial level of red fluorescent protein expression on their membranes (Supporting Fig. 1).3 These cells, despite being heterozygous for the ROSAmT-mG allele and therefore able to express only tdT or GFP and not both (Fig. 1A), exhibited both red and green outer membrane fluorescence. Theoretically, a somatic mutation in either the albCre transgene or the ROSAmT-mG allele could cause Cre failure and result in the rare appearance of a purely red hepatocyte in a ROSAmT-mG/+;albCre1 mouse; however, few physiological situations could allow an individual cell to express both tdT and GFP. Most likely, these cells had been actively expressing tdT until very recently and then had converted to GFP expression. During an interim period, one might expect preformed tdT protein to persist coincidently with the newly formed GFP protein accumulating in the membranes, and this would result in a transient period during which both proteins would coexist in single cells. This situation would be expected to occur only in newly differentiated hepatocytes that had very recently activated the albCre transgene and had not yet fully converted from tdT accumulation to GFP accumulation.3 To test this possibility, we immunostained cryosections from perfused regenerating ROSAmT-mG/+;albCre1 mouse livers with an anti-mouse albumin antibody (α-Alb) or no primary antibody (control), and this was followed by a blue fluorescent secondary antibody (Fig. 3A; yellow arrows denote the same point in each vertical series of exposures). The results verified that these reddish cells (circumscribed by fine, white lines in blue immunofluorescence panels) had cytoplasmic Alb protein, and this indicated that they were actively expressing their endogenous alb genes and synthesizing serum Alb. We also immunostained sections of regenerating liver for HNF4, a hepatocyte-specific transcription factor,27 and for Monts-4, a cell surface marker of tissue-resident macrophages, including Kupffer cells28 (Fig. 3B,C). The results showed that these reddish cells contained nuclear HNF4 protein and were not a part of the Monts-4–expressing hepatic cell population. As evidenced by their cell morphology, their cytosolic Alb protein staining, their nuclear HNF4 protein staining, and their absence of surface Monts-4 staining, these cells were differentiated hepatocytes and not Kupffer cells.27, 28, 33-37 Based on to the presence of both tdT and GFP in their membranes, they were recently Cre-naive cells that had activated the albCre transgene but had not yet lost all preexisting tdT. Thus, we conclude that the ROSAmT-mG marker could be used not only to trace Cre-exposed cell lineages but also, in combination with albCre, as a short-term in vivo chronometer for detecting hepatocyte lineages recently undergoing primary differentiation from an Alb- and Cre-naive prehepatocyte cell type.3
To calibrate this chronometer in hepatocytes, we induced synchronous expression of Cre in the hepatocytes of ROSAmT-mG/+ mice by intravascular administration of AdCre.24 Mice were harvested in triplicate over a 9-day time-course thereafter and liver sections were photographed for red and green fluorescence (Fig. 4A,B). Green and red pixels were quantified in regions containing only hepatocyte membranes and green/red pixel ratios were calculated (Fig. 4B,C). The data showed that within a liver, the green/red pixel ratio varied by ≤10%. This verified that the zonal variations in hepatocyte fluorescence intensities were similar for GFP and tdT fluorescence (Fig. 2A,D). The green/red pixel ratios were plotted versus time, and this produced a calibration curve for the rate of fluorescent protein conversion after Cre expression in ROSAmT-mG/+ hepatocytes in vivo (Fig. 4C). We obtained a similar calibration curve by triggering Cre activity in ROSAmT-mG/CreER mice3 with a pulse of 4-hydroxytamoxifen (Supporting Fig. 2).
Although the transition from purely red fluorescent outer membranes to purely green fluorescent outer membranes took roughly 10 days to complete in hepatocytes in vivo, at later times in this transition, the hepatocytes were only subtly red (Fig. 4B,C). In ROSAmT-mG/+;albCre1 livers, for which the fields were dramatically green from the abundance of differentiated hepatocytes yet also were intertwined with red fluorescent endothelial cells (Fig. 2D and Supporting Fig. 1), hepatocytes with low levels of residual red fluorescence could be difficult to discern from older purely green hepatocytes. Pragmatically, newly differentiated hepatocyte lineages were noticeably red for approximately 4 days after Cre expression (Fig. 4C and Supporting Fig. 2), after which they became difficult to identify in screens. Therefore, the values for newborn hepatocyte lineages in this study are for lineages up to approximately 4 days after the primary induction of albCre expression.
Newborn Hepatocyte Lineages in ROSAmT-mG; albCre1 Mice.
In ROSAmT-mG cells, the conversion from red to green is irreversible.22 In ROSAmT-mG/+;albCre1 mice, which have only a single allele of ROSAmT-mG, any purely red fluorescent cell must be on a lineage that has never expressed albCre; any cell that exhibits both red and green fluorescence must have expressed mT recently enough to retain preformed tdT protein in the membrane and must also have expressed albCre to excise the mT cistron and uncover the expression of mG (Fig. 1A). This situation should occur only in newborn hepatocyte lineages during a brief period following the primary expression of the albCre transgene.
Some conditions of hepatotoxicity result in the expansion of prehepatocyte cell populations. One established protocol for enriching prehepatocyte cells in rodent livers is by maintenance of animals on a CDE diet.17, 25, 38, 39 Livers harvested from resting adult ROSAmT-mG/+;albCre1 mice on a normal diet showed that 0.076% of all hepatocytes were ≤4 days old (Fig. 5A,D,E). The maintenance of mice on a CDE diet resulted in increases in the levels of these newborn hepatocytes to 0.4% at 5 days and 0.6% at 7 days (5- and 8-fold higher, respectively; Fig. 5A), and this is consistent with these cells arising anew from the expanded populations of prehepatocyte cells induced by these conditions.
Next, we used ROSAmT-mG/+;albCre1 mice to quantify the percentages of hepatocytes in newborn lineages in juvenile livers undergoing developmental growth and in regenerating adult livers 3 days after partial hepatectomy (Fig. 5B-G). Although 0.076% of all hepatocytes in resting adult livers were newborn, as evidenced by tdT fluorescence in their outer membranes (Fig. 5D,E), during developmental growth (Fig. 5B,C) or regeneration (Fig. 5F,G), 0.28% or 0.65% of all hepatocytes, respectively, were newborn. This indicates that the birth of new hepatocyte lineages contributes to liver maintenance, development, and regeneration in postnatal animals, and this contribution is different under each of these conditions.
Although the liver is known to contain a small population of stem cells, measurement of prehepatocyte cell contributions to normal postnatal liver physiology has proven difficult. In this study, we used the chronometer provided by the intrinsic fluorescent protein accumulation and turnover kinetics following conversion of the ROSAmT-mG marker allele22 in combination with a trigger provided by differentiation-induced expression of albCre30, to quantify levels of hepatocytes arising anew from Alb-naïve cells in mouse livers. Using this system, we were able to quantify for the first time the contributions of birth of new hepatocyte lineages to the growth, maintenance, and regeneration of normal mouse liver.
For resting adult livers, the steady-state level of hepatocytes born within the previous 4 days was 0.076%. Because hepatocytes are proliferative, each newborn hepatocyte detected in this study was expected to give rise to a lineage that might proliferate and contain many cells before its expiration or senescence. We do not yet know how many replicative cycles a typical hepatocyte lineage will undergo before it expires. However, if each newborn hepatocyte lineage were able to undergo six consecutive proliferative cycles before its extinction, for example, then each newborn hepatocyte lineage would eventually yield 64 hepatocytes. In this scenario, a sufficient number of new hepatocyte lineages would be born every 4 days in resting adults to replace 4.9% (64 × 0.076%) of the liver. This would be a substantial rate of hepatocyte lineage turnover. Thus, the apparently subtle rate of birth of new hepatocyte lineages reported here likely translates into a substantial contribution to the liver's cellular dynamics and represents an unexpected level of genome renewal from Alb-naive precursors in this organ.
The system employed here measures only the birth of new hepatocyte lineages from Alb-naive lineages. In contrast, some oval cells express low levels of Alb20, 40-42 and thus may represent a prehepatocyte cell type whose contributions would not have been detected here. Most studies that have characterized oval cells either have used disease states (e.g., cirrhotic and/or cancerous human livers) or have augmented oval cell populations in rodent livers with hepatotoxic treatments.20, 40-43 Importantly, resident prehepatocyte cells in normal livers have gene expression characteristics that are distinct from those in oval cells under conditions that induce their increased abundance. For example, Gleiberman et al.21 showed that in normal resting or regenerating mouse livers, prehepatocytes express a nestin-GFP transgene that is not expressed in the more abundant oval cells induced by hepatotoxic treatments in these mice.21 Also, prehepatocyte cells in normally developing livers20 and resident prehepatocyte cells in unmanipulated adult livers19 do not express Alb. It has been suggested that there may be different levels of maturation within the prehepatocyte cell population, with cells that express Alb representing a more mature stage.19 Other approaches will be needed to measure contributions from cell subpopulations that are not Alb-naive. However, the birthrates that we report here, even if they are possibly underestimated for these reasons, are substantial and will necessarily modify models of hepatocyte lineage life histories. For example, we are unaware of any previous reports showing that birth of new hepatocyte lineages contributes to the maintenance of normal adult liver homeostasis. The observation that new hepatocyte lineages are continuously being born under steady-state conditions (i.e., unmanipulated adults with no liver growth) implies that existing lineages must also be dying and are not, as current models suggest, indefinitely self-renewing.2 A picture is emerging of a largely overlooked hepatocyte lineage life history that involves birth from prehepatocyte cells, multiple rounds of replication, and eventual death of the lineage.
The authors thank J. Prigge, M. Rollins, C. Weisend, and E. Suvorova for their contributions to this study and M. Jutila for the Monts-4 monoclonal antibody.