Mono- and binucleated polyploid hepatocytes (4n, 8n, 16n and higher) are found in all mammalian species, but the functional significance of this conserved phenomenon remains unknown 1-4. Polyploidization occurs through failed cytokinesis, begins at weaning in rodents and increases with age 2,5-7. Previously, we demonstrated that the opposite event, i.e. ploidy-reversal, also occurs in polyploid hepatocytes generated by artificial cell fusion 8-10. This suggested the intriguing possibility that somatic “reductive mitoses” can also happen in normal hepatocytes. Here we show that multipolar mitotic spindles form frequently in mouse polyploid hepatocytes and can result in one-step ploidy-reversal to generate offspring with halved chromosome content. Proliferating hepatocytes produce a highly diverse population of daughter cells with multiple numerical chromosome imbalances as well as uniparental origins. Our findings support a dynamic model of hepatocyte polyploidization, ploidy-reversal and aneuploidy, a phenomenon which we term the “ploidy-conveyor.” We propose that this mechanism evolved to generate genetic diversity and permits adaptation of hepatocytes to xenobiotic or nutritional injury.
We first tested whether normal polyploid hepatocytes can undergo ploidy-reversal in vivo. Highly pure (> 99%; Fig. 1a, Supplementary Fig. 1) octaploid hepatocytes isolated by fluorescence activated cell sorting (FACS) from male mice hemizygous for Rosa26 (lacZ) were transplanted into female Fah-/- mice, a model for selective liver replacement 11, and the frequency of donor-derived hepatocytes was assessed following extensive (>70%) liver repopulation (Supplementary Fig. 2). Hepatocytes from repopulated recipient mice were loaded with Hoechst and fluorescein di-ß-D-galactopyranoside (FDG), a fluorescent substrate of ß-galactosidase (ß-gal), and analyzed by FACS. Donor-derived ß-gal was expressed by octaploid (90 ± 2%), tetraploid (83 ± 5%) and diploid hepatocytes (59 ± 5%) (Fig. 1b). Donor-derived FAH expression was also detected in most octaploid (89 ± 3%), tetraploid (86 ± 4%) and diploid (67 ± 11%) hepatocytes (data not shown). The overall ploidy distribution of donor hepatocytes was the same as found in normal liver of an aged mouse. Diploid and octaploid hepatocytes proliferated at equivalent rates (Supplementary Fig. 3), thus eliminating the possibility the high percentage of near-diploid donor-derived hepatocytes resulted from overgrowth by rare contaminating diploids. Cytogenetics confirmed the presence of reduced-ploidy donor-derived cells (Fig. 1c). Surprisingly, most donor cells were aneuploid, i.e. had numerical chromosome gains and/or losses (Figs. 1d, 1e). In addition to analyzing single hepatocytes, liver sections from repopulated livers were stained for donor markers. While most Fah+ nodules contained both Y-chromosome (mChrY) and ß-gal activity, loss of either marker was found in ~5% of FAH+ repopulation nodules (Supplementary Fig. 4). Together, these findings indicate that most normal polyploid hepatocytes undergo ploidy-reversal and marker segregation when forced to divide extensively.
The transplantation experiments demonstrating ploidy-reversal were done using the Fah knockout mouse. Therefore, the high degree of aneuploidy observed (Fig. 1d) and marker loss (Supplementary Fig. 4) could be attributed to toxic metabolites made by Fah deficient hepatocytes 12. To address this potential artifact, the karyotypes of hepatocytes from wild-type mice were determined. Although chromosome counts clustered around 40, 80 and 160 chromosomes, frequent chromosome gains and losses were detected in adult hepatocytes (Supplementary Fig. 5). At weaning, most hepatocytes were normal diploids, and by adulthood >60% of hepatocytes had numerical abnormalities (Fig. 1e). Thus, normal hepatocytes become aneuploid in adult mice, and this phenomenon is unrelated to Fah deficiency.
Due to the time required for liver repopulation (6-8 weeks) and the dynamic ability of hepatocytes to change ploidy, it was unclear whether the ploidy-reversal in vivo occurred in a single step or gradually over the course of multiple cell divisions. Therefore, we tracked ploidy changes using a short-term (1-2 mitoses) in vitro system. As expected, diploid hepatocytes became binucleated and polyploidized (Supplementary Figs. 6a, 6b). To investigate ploidy-reversal, we followed the fate of polyploid hepatocytes. Concurrent with DNA replication and mitosis (data not shown), the percentage of binucleated cells dropped dramatically, generating populations with >80% mononucleated cells (Supplementary Fig. 6a). After 5 days in culture pure tetraploid hepatocytes had produced daughters with 8c and 2c DNA content (18 ± 1% and 0.9 ± 0.1%, respectively) (Figs. 2a, 2b). Cultured octaploid hepatocytes showed similar ploidy redistribution (Supplementary Fig. 6c, Fig. 2b). The emergence of reduced-ploidy daughter cells over 1-2 cell cycles indicates a single-step mechanism for ploidy-reversal.
To test whether polyploid hepatocyte mitosis could also produce aneuploidy, we analyzed the frequency and nature of chromosome missegregation in vitro. Mice hemizygous for a yeast artificial chromosome containing the human CD46 gene (hCD46) 13 on mChr9 were utilized (Supplementary Figs. 7a-7c). Using fluorescence in situ hybridization (FISH), chromosome signals in hepatocyte nuclei were quantified before and after proliferation (Supplementary Fig. 8a). After 5d expansion by 4c hepatocytes, the ploidy distribution shifted to include cells with 2c and 8c DNA content, as illustrated previously (Fig. 2a). Daughter cells with reduced ploidy (2c cells) and equal ploidy (4c cells) were hybridized with probes for hCD46 and mChr9. Approximately 99% of hepatocytes analyzed directly without prior culture (2c and 4c) contained the appropriate number of FISH signals (Fig. 2c). In contrast, FISH signals were skewed in 2c and 4c daughter cells (12.5% and 21.4%, respectively), indicating a diverse and aneuploid population of daughter cells (Figs. 2c, 2d and Supplementary Fig. 8b). Importantly, ~1/3 of the 2c cells displayed uniparental disomy for mChr9. These data show that proliferating hepatocytes routinely generate a genetically diverse population of daughter cells.
The mechanism by which hepatocytes generated aneuploid or reduced-ploidy daughter cells was unknown. We rationalized that increased numbers of centrosomes in polyploid hepatocytes could lead to multipolar divisions and/or chromosome missegregation. To test this idea, cell divisions by polyploid hepatocytes were analyzed in vitro. In ~half of tetraploid mitoses, bipolar spindles were established and maintained by centrosome clustering (Fig. 3a). The remaining tetraploid hepatocytes contained spindles oriented in a multipolar configuration, with centrosomes oriented on 3-4 distinct poles (reflecting either true multipolar spindles or alignment of prometaphase chromosomes from binucleated hepatocytes) (Figs. 3b, 3c). Octaploid hepatocytes established multipolar spindles with as many as 8 poles (Fig. 3d). However, only ~1% of cells in anaphase or telophase were oriented with tripolar spindles (Fig. 3e). Additionally, hepatocytes with 2 discrete mitotic spindles synchronized in metaphase (Fig. 3f) and anaphase (Fig. 3g) were identified, an event we called “double mitosis.” Mitotic figures nearly identical to those seen in cultured hepatocytes were also observed during hepatocyte proliferation in vivo (Figs. 3i-3l).
The high percentage of multipolar metaphases and much lower frequency of multipolar anaphases/telophases was surprising. We hypothesized multipolar spindles could represent a temporary step in mitosis, and are then reorganized to be bipolar. Similar spindle dynamics were recently documented in cancer cells containing supernumerary centrosomes 14. A consequence of multipolar spindle reorientation is chromosome missegregation. Microtubules from different poles can attach to a single kinetochore, and failure to repair such merotelic attachments can lead to incomplete chromosome segregation 14. Consistent with spindle reorganization, we identified lagging chromosomes in 25-50% of tetraploid hepatocytes undergoing bipolar anaphase (Fig. 3h), suggesting that merotely contributes to marker loss (i.e., aneuploidy and loss-of-heterozygosity) observed in proliferating polyploid hepatocytes.
To determine whether multipolar mitoses produced viable offspring, we monitored hepatocyte divisions by time-lapse microscopy. As expected, diploids completed bipolar cell division with successful (89%) or failed cytokinesis (7%) (Fig. 4a). For analysis of polyploid hepatocytes, we focused mostly on tetraploids, but similar findings were seen with octaploids and non-fractionated hepatocytes (i.e., a mixture of all ploidy classes that were never exposed to Hoechst or subjected to FACS). Nearly 90% of tetraploid hepatocytes divided in a bipolar manner (Fig. 4a, Supplementary Fig. 9, Supplementary Movie 1), and in many time-lapse sequences their daughters (14%) divided again. Approximately 7% of tetraploid mitoses failed to complete cytokinesis (Fig. 4a, Supplementary Fig. 10, Supplementary Movie 2). Half of mononucleated and binucleated tetraploids transitioned from an early multipolar intermediate (as seen by chromosome alignment along multiple axes) to a standard bipolar configuration (Fig. 4c). Mitotic arrest and apoptosis were never seen.
In addition to standard divisions producing 2 daughters, mitosis along multiple axes was seen. For example, 3.2% of tetraploid hepatocytes (mononucleated and binucleated) were captured undergoing tripolar mitosis (Figs. 4a, 4b, Supplementary Movie 3). All of the daughter cells were viable for the duration of the imaging session (up to 16h), and in some cases (15% of daughters) we were able to film subsequent mitoses (Supplementary Fig. 11, Supplementary Movie 4). While ~10% of tripolar divisions completed 3-way cytokinesis, most divisions (~90%) ended in partial failed cytokinesis (Fig. 4d). Nuclear content frequently segregated in a 4:2:2 ratio, which is consistent with 1 tetraploid daughter nucleus and 2 reduced-ploidy diploids (Supplementary Fig. 12). Furthermore, 1.2% of tetraploid hepatocytes completed double mitotic events (Fig. 4a). Double mitoses by either binucleated (Supplementary Fig. 13, Supplementary Movie 5) or mononucleated tetraploids (Supplementary Fig. 14, Supplementary Movie 6) generated 4 distinct nuclei via 2 synchronized mitoses (Fig. 4e). By definition, birth of 4 mononucleated cells from a parental cell represents a ploidy-reversal event. Daughters were viable and appeared healthy throughout imaging (as long as 10h).
Our data demonstrate that hepatocytes can increase (failed cytokinesis) and reduce (multipolar mitosis) their ploidy, thus resulting in the concept of a “ploidy-conveyor.” This dynamic mechanism not only generates numerical chromosome “abnormalities,” but also uniparental chromosome sets. Given that 5-10% of all genes are thought to be monoallelically expressed 15, this segregation pattern produces tremendous genetic heterogeneity.
The pervasive presence of aneuploid genotypes in the liver raises the question of whether this phenomenon serves a physiological purpose. Elegant studies in yeast showed aneuploidy can provide a strong selective advantage in response to multiple environmental stressors 16. Our findings suggest the provocative possibility that hepatocyte polyploidization evolved precisely to result in subsequent ploidy-reversal, aneuploidy and genetic diversity. Therefore, hepatic injury, which produces liver regeneration, could result in selection of hepatocytes that are genetically most resistant to the injury from a pre-existing pool of diverse genotypes. Genetic analysis of hepatocytes following liver injury may reveal favorable genotypes that differ from the germ line. Indeed, our own group has already observed an example of “favorable loss-of-heterozygosity” followed by selection in the Fah deficient mouse 17.