PKD, one of the most prevalent life-threatening human genetic diseases4
, is characterized by the progressive dilation of renal tubules that eventually form cysts, often leading to end-stage renal failure. The homozygous inactivation of most cystic genes leads to a rapid and severe PKD phenotype in mice5,6
. Notably, in the last few years several studies have shown that there is a key postnatal developmental switch that defines the onset and the severity of the polycystic phenotype7–11
. However, the molecular mechanism at the basis of this differential propensity to develop cysts or not remains poorly defined.
We have previously shown that HNF-1β, a transcription factor expressed in the kidney12
, controls the expression of cystic genes, including Pkhd1
(encoding polycystic kidney and hepatic disease-1), Pkd2
(encoding polycystic kidney disease-2) and Umod
inactivation in the prenatal mouse kidney results in severe PKD. Mutations in HNF1B
are very frequent in individuals suffering from renal cysts or dysplasia13,14
, often associated with a particular form of diabetes mellitus, known as maturity onset diabetes of the young type 5 (MODY5).
The drastic phenotype elicited by prenatal inactivation of Hnf1b
has until now prevented the analysis of the function of this protein in already formed, normally shaped renal tubules. To address this question, we adopted a strategy of a postnatal inducible inactivation. With the use of an interferon (Poly(I)–Poly(C))-inducible Cre system (MxCre; ref. 15
) and a floxed allele of Hnf1b16
, we analyzed the effect of Hnf1b
inactivation at key time points during postnatal renal maturation in mouse pups. To monitor the extent of recombination efficiency, we took advantage of a tester allele of the Cre recombinase activity (ROSA26R; ref. 17
). Our results showed that inactivation of Hnf1b
in young pups elicited cystic dilations, even with a limited extent of recombination (). In fact, all mice injected with the inducer of the Cre recombinase between 0 d and 2 d after birth developed polycystic disease 30 d later (). These lesions progressed to generate markedly enlarged kidneys 60 d later (Supplementary Fig. 1
). In contrast, mice induced on day 10 after birth did not develop any cystic lesions within 1 to 6 months later ( and data not shown). This result was not due to a lower extent of gene inactivation, as the degree of recombination in noncystic mice induced in older mice was paradoxically higher than that detected in cystic mice injected at birth (). Moreover, these noncystic mice had lost the expression of HNF-1β protein in the medulla and at the cortico-medullary boundary, as revealed by immunostaining (). All of these data indicate that HNF-1β function is necessary for maintaining normal tubular morphology only during a narrow time window that corresponds to the first few days after birth. Recent studies have monitored the effects of the inactivation of cystic disease-associated genes (Ift88
(encoding polaris), Kif3a
(encoding kinesin-3A) and Pkd1
) at various time points in early postnatal life7–11
. Similar to what we have observed in this study, mice developed a prompt and dramatic PKD phenotype only when these genes were inactivated shortly after birth, whereas, if the inactivation was carried out later, a much slower onset of the cystic phenotype could be observed (after a minimum of 5–6 months)7–11
. Collectively, these data suggest that cyst formation requires more than just the loss of cystic disease–associated gene expression and suggest the existence of unknown cystogenetic triggering events.
Figure 1 Time-restricted cystogenic competence. (a) Morphology of kidneys from mice where Hnf1b was inactivated at either P1 or P10 and killed 30 d later. X-gal staining reveals the extent of Cre recombination (ROSA26R). Scale bar, 1 mm. (b) Percentage of Hnf1b (more ...)
It has been suggested that cell proliferation participates in cyst formation (reviewed in ref. 18
); nevertheless, the dramatic resistance to cyst development from postnatal day 10 (P10) on occurs in a phase when the kidney is still actively growing. To clarify the mechanisms of this restricted cystogenetic competence, we decided to characterize the pattern of renal cell proliferation throughout this transition. Our results demonstrated that cystogenic competence correlates with a particular intense and localized pattern of cell proliferation that characterizes postnatal nephron morphogenesis and elongation (Supplementary Data
and Supplementary Fig. 2
During postnatal kidney maturation, proliferation progressively declines and almost disappears when the organ has reached its final adult size19
. Quiescent renal tubular cells in adults can be forced to reenter the cell cycle after ischemia-reperfusion (I/R) injury20
. We found that the pattern of cell proliferation after I/R is very similar to that of elongating tubules at P0–P2, with locally intense and cell cycle-synchronized proliferation in the cortico-medullary region (Supplementary Fig. 2
). We have previously shown that tubular elongation during early postnatal nephron maturation is linked to oriented cell division21
. Notably, when we analyzed the distribution of the mitotic angles of regenerating tubular cells, we observed a strikingly biased orientation comparable to that observed during nephron morphogenesis (). Ninety-five percent of mitotic angles were smaller than 30°, with an average of 15° (). These data indicate that the morphogenetic pattern of intense, synchronized and oriented cell proliferation is programmed to be reactivated during renal regeneration.
Figure 2 Tubular regeneration elicits distorted cell division and dilations in Hnf1b deficiency. (a) Three-dimensional reconstruction of typical examples of longitudinally (left) or perpendicularly (right) oriented mitoses. Tubules are in green; mitoses are in (more ...)
The correlation between morphogenetic proliferation and cystogenic competence made us wonder whether the locally intense and synchronized proliferation could be the triggering cystogenic event in the context of homozygous deletion of a cystic disease–associated gene. To assess this point, we forced morphologically normal Hnf1b
-deficient quiescent tubules () to reenter the cell cycle by an I/R-induced injury. Mutant mice, in which the gene was deleted around P60, developed marked dilations and cysts 2 weeks after I/R injury (). These lesions occurred at the cortico-medullary boundary, the region where the most intense regenerative proliferation takes place in kidneys after ischemia and where Hnf1b
is efficiently deleted (). In contrast, we never observed cysts in injured control littermates (). Hnf1b
inactivation in quiescent cells did not elicit a prompt increase in proliferative response or tubular dilation (Supplementary Fig. 3
and data not shown). However, once proliferation was triggered, Hnf1b
-deficient kidneys tended to proliferate more than the controls (Supplementary Fig. 3
). In addition, we observed that the mitotic orientation of proliferating mutant cells was significantly distorted compared to the controls (). Taken together, our results indicate that the lack of HNF-1β is a permissive (necessary but not sufficient) condition to elicit tubular dilation. In this context, we showed that proliferative bursts, occurring during tubular elongation or regeneration, are the crucial triggering events that can provoke cystic tubular dilations.
We next tried to elucidate the mechanisms underlying the synergistic effect of proliferation on Hnf1b
deficiency. HNF-1β controls the expression of several genes, including Pkd2
). As shown above, Hnf1b
gene deletion at P0–P2 resulted in a marked PKD phenotype (). Quantitative RT-PCR (qRT-PCR) showed that this phenotype was accompanied by a dramatic decrease in Umod
expression () similar to that observed with the prenatal inactivation of Hnf1b
). The identification of genomic HNF-1 binding sites that are highly conserved throughout evolution allowed us to identify additional potential target genes involved in cell polarity, cystogenesis or both, including Crb3
(encoding crumbs homolog-3), Tcfap2b
(transcription factor AP-2β), Tmem27
(encoding transmembrane protein-27) and Bicc1
(encoding bicaudal C homolog 1) (data not shown). All of these genes contained sites that were bound in vivo
by HNF-1β (Supplementary Fig. 4
). The expression of these genes was markedly downregulated when Hnf1b
was inactivated at P0 (). However, in kidneys where Hnf1b
was deleted after P10, we were surprised to see that there was no drop in the expression of these target genes (). The only exception was Umod
, whose expression dropped to a level similar to that observed after perinatal inactivation ().
Figure 3 Effect of proliferation on HNF-1β target gene expression and chromatin modifications. (a) Renal expression of Umod, Pkhd1, Pkd2, Kif12, Crb3, Tcfap2b, Tmem27 and Bicc1 mRNA transcripts in wild-type or in mutant adult kidneys 3–4 weeks (more ...)
The paradoxical resilient expression of most target genes could have been explained by a particularly long half-life of messenger RNAs that may have compensated and masked a possible drop in the transcription initiation rate after Hnf1b
deletion. To assess this point, we quantified the short-lived heterogeneous-nuclear RNA (hnRNA) that can be detected only in actively transcribed genes, just before splicing has taken place. We found that the transcription initiation rate of these targets was not significantly altered by Hnf1b
deletion in quiescent kidneys in vivo
(Supplementary Fig. 5
). Notably, the decrease in the Pkd2
transcription rate at P0 was smaller than that observed in mature mRNA ( and Supplementary Fig. 5
), indicating that the expression of this gene was probably also controlled at the post-transcriptional level. Immunohistochemistry confirmed that in noncystic, Hnf1b
deficient tubules (deleted at P10–P15 or in adult mice), Pkd2 and Kif12 expression was not affected (Supplementary Fig. 6
and data not shown). However, when proliferation waves occurred in cells that had lost HNF-1β, either via I/R-induced injury or during tubular elongation, the expression of target genes was markedly reduced (Supplementary Fig. 6
). It is noteworthy that Pkd2 or Kif12 expression did not change in the tubules of wild-type mice after ischemia (Supplementary Fig. 6
and data not shown) or in mutant cells that did not proliferate (Supplementary Fig. 7
To further validate these observations, we subjected Hnf1b-mutant adults to I/R-induced injury followed by a continuous administration of BrdU to label all cells that had undergone proliferation after the inactivation. We inspected the kidneys of mice killed long after the proliferative wave had occurred (14 d) for the presence of BrdU incorporation and Pkd2 or Kif12 expression. A careful examination of the double-stained tissue sections showed that Pkd2 and Kif12 staining tended to be lost in BrdU-positive cells that had been forced to proliferate in the absence of HNF-1β (). Collectively, these observations indicate that HNF-1β target genes may be classified into two distinct groups. In the first group (class 1 genes), exemplified by Umod, transcription markedly drops after Hnf1b deletion, regardless of the proliferative status of cells. In the second group (class 2 genes, including Pkhd1, Pkd2, Kif12, Crb3, Tcfap2b, Tmem27 and Bicc1), once activated, transcription no longer requires the persistent and redundant action of HNF-1β if cells remain quiescent. However, this resilient transcription does not persist when proliferation occurs.
It is known that when cells enter mitosis, chromatin is condensed and forms the typical heavily packed mitotic chromosomes. Throughout the progression of these events, gene transcription is shut down2,3
. At the exit from mitosis, gene expression must be reactivated to restore the typical tissue- and differentiation-specific transcriptional program. Our observations could be ascribed to an essential involvement of HNF-1β in inducing the post-mitotic chromatin reopening and gene transcriptional reactivation of class 2 genes. Notably, Hnf1a
, a gene closely related to Hnf1b
, also has a key role in the establishment, maintenance or both of an open chromatin structure at the Pah
gene locus (encoding phenylalanine hydroxylase) during liver development23
. To verify the impact of Hnf1b
deletion on chromatin status in proliferative cells, we carried out chromatin immunoprecipitation with antibodies specific for active chromatin modifications including histone H3 Lys4 trimethylation (H3K4me3) and H3 Lys9 acetylation (H3K9ac) and inactive H3 Lys9 (H3K9) or Lys27 (H3K27) trimethylation (H3K9me3 or H3K27me3). To this end, we used a mouse inner medullary collecting duct cell line (mIMCD3) expressing an inducible dominant-negative form of HNF-1β (ref. 24
). Class 2 genes underwent silencing and heterochromatinization upon HNF-1β inhibition: the acetylation of H3K9 was significantly reduced, whereas the trimethylation of H3K9 and H3K27 was significantly increased (). In addition, when we inspected the same sites in vivo
, we observed a comparable loss of activating chromatin modifications on target genes in perinatal mutant kidneys ( and Supplementary Fig. 8
). In summary, our results suggest that Hnf1b
deletion in a proliferative context has a drastic effect on the chromatin status of its target genes.
During cell cycle progression in higher eukaryotes, after nuclear envelope breakdown, transcription factors are released from chromatin. This leads to the transcriptional silencing that accompanies mitosis. To further characterize the function of HNF-1β, we monitored its localization during the cell cycle. With the use of a HNF-1β–GFP fusion protein, we discovered that, contrary to the classical behavior of transcription factors (shown here by an HNF4α–GFP fusion protein), HNF-1β is not dissociated from chromatin during mitosis (). Rather, upon nuclear envelope breakdown, a substantial fraction of HNF-1β remains associated with the mitotic chromatin traveling with the condensed chromosomal barrels throughout the entire mitotic process ( and Supplementary Video 1). Our results support a previously undescribed concept and indicate that transcription factors such as HNF-1β may be involved in gene expression reprogramming after the silencing accompanying mitotic chromatin condensation is relieved. Thus, HNF-1β could act as a bookmarking factor that marks target genes for a rapid transcriptional reactivation after mitosis.
Figure 4 HNF-1β behaves as a bookmarking factor. (a) Time-lapse microscopy of mIMCD3 cells expressing either HNF1β–GFP or HNF4α–GFP fusion proteins. For each fusion protein, phase contrast (top) and green fluorescence (bottom) (more ...)
mutant mice, a severe polycystic kidney phenotype is observed only when this transcription factor is inactivated in highly proliferating tubular cells. These cells are cycling very intensely and probably need to promptly reestablish the expression of cystic disease–associated genes as soon as the cells reenter interphase. These cystic disease–associated genes are probably necessary to maintain the proper alignment of cell division during tubular elongation. The fact that HNF-1β remains associated with mitotic chromatin during mitosis in these proliferating cells might provide a useful and rapid mechanism to reactivate the expression of crucial target genes necessary for preventing tubular dilation. In Hnf1b
deficiency, proliferation triggers a positive feedback loop leading to the persistence of a spatially distorted hyperproliferation. It is tempting to speculate that an analogous mechanism might explain the unexpected refractory response to cystic gene inactivation in adult mice7–11,25
. The transcription networks downstream of ciliary and cystic genes26,27
, similarly to HNF-1β, could maintain the expression of antiproliferative targets in mutant cells in a quiescent metastable state. Notably, it has been recently shown that PKD model mice treated with roscovitine, a cyclin-dependent kinase inhibitor, show long-lasting decreased cystogenesis28
, further suggesting that inhibition of proliferative bursts may have a beneficial effect on this pathology.
In conclusion, our study indicates that transcription factors such as HNF-1β may control gene expression via two distinct mechanisms. The first involves the maintenance of active transcription from its target promoters. The second involves a new aspect in the control of gene expression and concerns the reprogramming of transcription by counterbalancing the silencing effect induced by mitotic chromatin condensation (). The implications of these observations shed light on the crucial connection between genetically determined predispositions and environmentally induced proliferative stress conditions.