|Home | About | Journals | Submit | Contact Us | Français|
Bmi1 is a member of the Polycomb protein family and represses transcription by modifying chromatin organization at specific promoters. Bmi1 is implicated in the control of stem cell self-renewal and has been shown to regulate cell proliferation, tissue homeostasis, and differentiation. Bmi1 is present in a subpopulation of self-renewing pancreatic acinar cells and is expressed in response to pancreatic damage. We investigated the role of Bmi1 in exocrine pancreas regeneration.
Acute pancreatitis was induced in Bmi1−/− mice with caerulein; pancreatic cell regeneration, differentiation, and apoptosis were assessed. Cultured Bmi1−/− and wild-type primary acini were analyzed in vitro, to determine acinar-specific consequences of Bmi1 deletion. To investigate cell-autonomous vs non–cell-autonomous roles for Bmi1 in vivo, pancreatitis was induced in Bmi1−/− mice reconstituted with a wild-type hematopoietic system.
Bmi1 expression was upregulated in the exocrine pancreas during regeneration after caerulein-induced pancreatitis. Exocrine regeneration was impaired following caerulein administration to Bmi1−/− mice. Pancreata of Bmi1−/− mice were hypoplastic, and the exocrine pancreas was replaced with ductal metaplasia that had increased apoptosis and decreased cell proliferation, compared to that of wild-type mice. Expression of Cdkn2a and p53-dependent apoptotic genes were markedly upregulated in Bmi1−/− pancreas, compared to wild-type mice, after injury. Furthermore, after transplantation of bone marrow from wild-type to Bmi1−/− mice, the chimeric mice had intermediate levels of pancreatic hypoplasia and significant, but incomplete, rescue of impaired exocrine regeneration after caerulein injury.
Bmi1 contributes to regeneration of the exocrine pancreas after caerulein-induced injury through cell-autonomous mechanisms—in part by regulating Cdkn2a expression—and non-cell-autonomous mechanisms.
Bmi1, first identified as a collaborating oncogene through its ability to cooperate with cMYC in the induction of lymphoma1, 2, is a member of the Polycomb group proteins (PCG)3. PCG proteins combine in Polycomb repressive complexes (PRC) to epigenetically regulate gene expression by modifying chromatin organization at specific promoters4. Two major Polycomb repressive complexes (PRC1 and PRC2) have been described. PRC2, which contains EZH2, catalyses the trimethylation of histone H3 at lysine K27 (H3-K27)4. The lysine H3-K27 tag serves as a docking site for recruitment of PRC1, resulting in transcriptional repression via PRC1 mediated ubiquitylation of histone H2A on lysine 119 or protein sumoyation4. Bmi1 is indispensible for the function of the PRC1 complex, making it a key component of PCG mediated gene regulation.
Bmi1 is essential for the self-renewal of several types of adult stem cells including those within the hematopoietic system5, nervous system6, mammary gland7 and prostate8. A similar role has been noted within cancer stem cell populations found in leukemia9, breast7 and lung cancers10. These functions are in part attributed to the ability of the PRC1 complex to bind to and repress the Cdkn2a locus, which encodes both p16 (Ink4a) and p19 (Arf) 4, 11. p16 functions upstream of the tumor suppressor Rb and inhibits cell cycle progression by specifically blocking the cyclin-D/CDK4/6 complex, whereas p19 functions upstream of p53 by stabilizing p53 through inhibition of Mdm2 function11. Indeed, Bmi1 dependent regulation of these pathways has been shown to be important for tissue homeostasis, as evidenced by proliferative arrest and p53-dependent cell death of hematopoietic stem cells in the context of de-regulated p16Ink4a and p19Arf in Bmi1 knock out (KO) mice5.
Mouse models of acute pancreatitis have revealed that the exocrine pancreas possesses remarkable regenerative capacity after inflammatory tissue damage. In response to acute pancreatitis induced by the cholecystokinin analogue caerulein, acini can transiently de-differentiate into proliferative, duct-like structures that express markers characteristic of pancreatic embryonic progenitors, such as Pdx112. Developmental signaling pathways, including the Hedgehog, Notch, and Wnt pathways are also reactivated during exocrine regeneration, and have been shown to be critical for exocrine recovery following caerulein treatment in mice13–16. In the absence of further damage, re-differentiation into functional acinar cells occurs rapidly within one week after cessation of caerulein treatment12–15, 17.
Recently, cell lineage tracing experiments performed in Bmi1-CreER mice have revealed a subpopulation of differentiated pancreatic acinar cells with proliferative capacity18. Long-term Bmi1 lineage tracing experiments in the pancreas demonstrate that Bmi1 lineage-labeled cells remain consistent over a year, demonstrating that Bmi1 is present in a subpopulation of self-renewing pancreatic acinar cells18. In addition, Bmi1 expression is induced in pancreatic acinar cells after acute injury during regeneration18, 19 and is critical for beta cell regeneration through regulation of the Ink4a/Arf locus20. However, the functional role of Bmi1 in exocrine pancreas regeneration remains unknown.
In the present study, we show impaired exocrine pancreas regeneration resulting in organ hypoplasia in Bmi1 KO mice after caerulein-induced injury. The decrease in regenerative capacity was associated with markedly increased apoptosis and decreased cell proliferation. Furthermore, Bmi1 target genes, p16, p19 and downstream p53 pro-apoptotic genes were markedly upregulated in Bmi1 KO pancreas after injury. Finally, using in vivo bone marrow transplantation and in vitro acinar culture approaches, we show that pancreatic hypoplasia and impaired exocrine regeneration is mediated through a combination of cell autonomous and non-cell autonomous mechanisms in Bmi1 KO mice after caerulein injury.
Detailed methods are described in the Supplementary Materials and Methods.
Bmi1GFP/+ knock-in mice have been described previously21. Transgenic mice were on a mixed, background. Littermate Bmi1GFP/+ or wild-type mice were used as controls. All mouse experiments were performed under the approval of the UCSF Institutional Care and Use of Animals Committee (IACUC).
To explore a possible role for Bmi1 during exocrine regeneration, we first ascertained the Bmi1 expression pattern following acute caerulein pancreatitis. As shown previously18, Bmi1 is present in a small subset (~1.5%) of adult acini in wild-type (WT ) mice at the age of 5–8 weeks (Figure 1A). In contrast, we observed a significant expansion of Bmi1 expression in acinar cells (~97%) in WT mice two days after acute caerulein-induced injury (Figure 1A). Seven days after caerulein treatment, following considerable regeneration, exocrine Bmi1 expression was again restricted to rare acini (Figure 1A). This pattern was also confirmed using GFP as a reporter for Bmi1 expression in Bmi1GFP/+ mice, in which the endogenous Bmi1 gene has been replaced through homologous recombination with GFP21. A small subset (~0.9%) of acinar cells were positive for GFP in Bmi1GFP/+ mice before caerulein injury (Figure 1B) compared to nearly ubiquitous GFP expression in acinar cells (~98%) at day 2 post-caerulein treatment (Figure 1B). We also noted GFP expression in damaged acini at 4, 8, and 24 hours after the last injection of caerulein (Supplementary Figure 1). GFP expression did not colocalize with the mature duct marker DBA lectin, or the pancreatic stellate cell markers vimentin and desmin in PBS-treated mice, at day 2 and day 7 post-caerulein treatment (Figure 1C, 1D). These data indicate dynamic Bmi1 expression following caerulein pancreatitis that is mainly limited to acinar cells during exocrine pancreas regeneration.
In order to determine if Bmi1 plays a functional role in exocrine pancreas regeneration, we subjected homozygous Bmi1GFP/GFP mice (Bmi1 KO ) to acute caerulein pancreatitis at 6 weeks of age21. In the absence of damage, pancreatic architecture and exocrine histology was indistinguishable between Bmi1 KO and control mice (data not shown), indicating that Bmi1 is dispensable for exocrine pancreas development. Seven days after caerulein treatment in WT controls, exocrine regeneration was nearly complete and acinar morphology appeared nearly indistinguishable from untreated mice (Figure 2B). Strikingly, Bmi1 KO mice displayed gross pancreatic hypoplasia (Figure 2A). H&E staining and immunostaining for amylase (acini) and cytokeratin19 (CK19, ducts) revealed a dramatic reduction in acinar cells accompanied by a profound increase in duct-like epithelial cells (Figure 2B & 2C). Notably, pancreatic hypoplasia, ductal metaplasia, and loss of acinar cells persisted 3 weeks after caerulein treatment, suggesting that acinar regeneration is indeed blocked, and not merely delayed, in the absence of Bmi1 (Supplementary Figure 2A–2C).
During regeneration, acinar cells temporarily express embryonic progenitor markers, including Pdx1, Hes1, and Sox912, 13. Immunostaining and quantitative RT-PCR (q-PCR) analyses revealed persistent expression of Sox9 and Hes1 in Bmi1KO epithelium at day 7 after injury (Figure 2D, 2E). In contrast, epithelial cells remained negative for Pdx1 (Figure 2D). This suggests that damaged acinar cells were either eliminated and replaced by ductal cells, or remained duct-like and failed to reactivate the complement of differentiation pathways characteristic of acinar regeneration. Indeed, the expression of transcription factors Ptf1a and Mist1, which are expressed in adult acini and crucial for exocrine pancreas differentiation22, 23,24, were dramatically down-regulated in Bmi1 KO pancreas compared to control pancreas 3 days after injury (Figure 2F). Furthermore, H&E and desmin staining showed that the remaining duct structures were embedded in an expanded stromal compartment, absent in regenerating control mice day 7 after injury (Figure 2B, 3A, 3B). The persistent stromal reaction occurred in parallel with prolonged inflammation in Bmi1 KO pancreas at day 7 after injury, as evidenced by sustained upregulation of IL6 expression and active, phospho-Stat3 (Figure 3C, 3D, 3E).
To determine if the requirement for Bmi1 is exclusive to exocrine regeneration following careulein pancreatitis, we also challenged female WT and Bmi1 KO mice with choline deficient ethionine-supplemented (CDE) diet induced pancreatitis25. H&E staining revealed a similar degree of exocrine disorganization and mild edema at day 4 after the initiation of the CDE diet in both control and Bmi1KO female mice (Supplementary Figure 3A). Immunohistochemistory showed that Bmi1 and Sox9 were expressed in acinar cells in Bmi1GFP/+ mice at that time point (Supplementary Figure 3B). Furthermore, we observed coexpression of GFP and clusterin, a marker of immature, regenerating acini13, 14, in acinar cells in Bmi1GFP/+ mice, suggesting that Bmi1 is expressed in damaged acinar cells following CDE diet induced pancreatitis in a similar fashion as what we have observed after caerulein pancreatitis (Fig. 1; Supplementary Figure 3B). At day 8 after initiation of the CDE diet, acinar regeneration was completed in control mice, whereas acinar regeneration remained impaired in Bmi1 KO mice (Supplementary Figure 3C). Thus, Bmi1 is critical for acinar regeneration following both caerulein and CDE induced pancreatitis.
To determine if compromised acinar regeneration correlated with a more severe initial response to caerulein pancreatitis, we analyzed Bmi1 KO mice at day 1 and day 2 post-caerulein injection. H&E staining showed duct-like epithelial cells in both control and Bmi1 KO pancreata at the first two days post-caerulein injection (Supplementary Figure 4A). At day 1, duct-like epithelial cells frequently co-expressed amylase and CK19. At day 2 we detected widespread co-expression of clusterin and Sox9 in both control and Bmi1 KO mice (Supplementary Figure 4C), Furthermore, serum amylase, another parameter for the severity of acute pancreatitis, was indistinguishable between control and Bmi1 KO mice before caerulein injection and at day 1 post-caerulein injection (Supplementary Figure 5A). Thus, Bmi1 null acini appear to respond to acute pancreatitis in a manner comparable to WT acini. However, while double amylase, CK19 positive cells persisted at day 2 in WT mice, the level of amylase expression appeared relatively reduced at the expense of CK19 in duct-like epithelial cells in Bmi1 KO mice (Supplementary Figure 4B). These data indicate that acinar regeneration starts to be compromised at day 2 following caerulein in Bmi1 KO mice.
Prior work from our lab and others has demonstrated that the early stages of caerulein pancreatitis are associated with pancreatic inflammation26–28. To determine whether Bmi1 loss affects inflammatory cell infiltration at the onset of caerulein pancreatitis, we performed staining for markers of hematopoietic cells and macrophages, respectively. Staining for CD45 and F4/80 showed a similar degree of inflammatory infiltrates and macrophages at day 1 and day 2 post-caerulein injection, suggesting that the initial inflammation in the pancreas is comparable between Bmi1 KO and control mice after caerulein injury (Supplementary Figure 6). In addition, to assess another established parameter for the systemic severity of pancreatitis, we analyzed lung histology after caerulein injury. H&E staining showed that the degree of alveolar membrane thickening and infiltration of inflammatory cells were comparable between Bmi1 KO and control mice both at day 1 and day 2 post-caerulein injection (Supplementary Figure 5B). In addition, immunostaining for CD45 and myeloperoxidase (MPO) in the lung revealed that the degree of inflammatory and neutrophil infiltration was comparable between Bmi1 KO and control mice at day 2 post-caerulein injection (Supplementary Figure 5C). Therefore, Bmi1 loss does not appear to significantly alter acute inflammation during caerulein induced pancreatitis.
Next we tested whether the pancreatic hypoplasia in Bmi1 KO mice after injury was associated with changes in apoptosis or cell proliferation. Immunostaining and quantification of cleaved-Caspase3/E-cadherin double-positive cells revealed a dramatic increase in cleaved-Caspase3 positive pancreatic epithelial cells in Bmi1 KO mice compared to control mice at both day 3 and 7 after caerulein-induced injury (Figure 4A and 4C and data not shown). Increased apoptosis was also confirmed by a similar increase in the number of TUNEL/E-cadherin double-positive cells in the Bmi1 KO pancreatic epithelium when compared to controls (Figure 4B). In addition, pancreatitis associated apoptosis was also increased in Bmi1 KO mice compared to WT mice 8 days after the initial CDE treatment (Supplementary Figure 3C).
We next determined whether cell proliferation is affected in Bmi1 KO pancreas after caerulein injury. Immunostaining and quantification demonstrated that the number of cells double positive for the mitosis marker phospho-histone H3 (pHH3) and E-cadherin was significantly decreased in the Bmi1 KO pancreatic epithelium compared to control mice at both day three and day seven after caerulein treatment (Figure 4D and 4E and data not shown). Therefore, both markedly increased apoptosis and decreased cell proliferation contribute to the hypoplastic pancreas observed in Bmi1 KO mice folowing acute caerulein pancreatitis.
Bmi1 has been shown to control cell proliferation during self-renewal in part through controlling transcription of the Cdkn2a locus, encoding the tumor suppressors p16 and p19 4, 11. We found that Bmi1 constrained Cdkn2a, p16, and p19 expression in the pancreas, as PBS-treated Bmi1 KO mice displayed increased expression compared to PBS-treated wild-type controls (Figure 5A). Furthermore, we observed a dramatic increase of both p16 (Ink4a ) and p19 (Arf ) mRNA transcripts in Bmi1 KO pancreas three days after caerulein-injury compared to control pancreas (Figure 5A). Co-immunostaining of p16 and Sox9 confirmed increased p16 expression in pancreatic epithelium undergoing metaplasia in Bmi1 KO mice compared to control mice three days after injury (Figure 5B). p16 impairs cell cycle progression by inhibiting the phosphorylation and subsequent inactivation of Rb. Correlating with increased p16 expression levels, we observed a significant reduction in phospho-Rb levels in Bmi1 KO pancreas compared to control pancreas after injury (Figure 5C). In parallel, p19 expression leads to stabilization of the tumor suppressor p53 via inhibition of its repressor Mdm211. Immunofluorescence revealed increased p53 expression in Bmi1 KO mice compared to control mice 3 days after injury (Figure 5D). In addition, q-PCR analyses revealed that p53 target genes involved in apoptosis, including Noxa, Apaf1, and Bax were significantly upregulated in Bmi1 KO pancreas (Figure 5E). Furthermore, the expression of the cell cycle inhibitor p21, regulated in part through p53, was also significantly elevated in Bmi1 KO pancreas (Figure 5E). Taken together, loss of Bmi1 function during exocrine regeneration results in dramatic upregulation of cell cycle inhibitory mechanisms downstream of p16 and p19, as well as p53 dependent apoptotic genes.
Previous reports have shown that Bmi1 is essential for maintenance of adult hematopoietic stem cells and that Bmi1 KO mice have defective hematopoietic systems with markedly reduced hematopoietic stem cells5, 29. Hematopoietic cell numbers are reduced in Bmi1 KO mice beginning in early development and the absolute numbers of both B lymphoid and myeloid cell populations are significantly reduced in adult Bmi1 KO mice compared to WT mice29. Therefore, it is possible that the profound systemic hematopoietic defects in Bmi1 KO mice and the prolonged inflammation observed in Bmi1 KO mice might contribute to non-cell-autonomous mechanisms that affect exocrine regeneration after caerulein injury.
To address this question, we transplanted WT bone marrow into irradiated WTh and Bmi1 KO mice at 6–9 weeks of age (Figure 6A). Eight weeks after bone marrow transplantation, we confirmed hematopoietic reconstitution in Bmi1 KO chimeric mice by assessing the presence and relative abundance of GFP-negative, CD45-positive cells by flow cytometry (Supplementary Figure 7A) and the presence of the Bmi1 WT allele by PCR from splenic DNA (Supplementary Figure 7B). Transplanted chimeric animals were subjected to caerulein induced pancreatitis and analyzed for pancreatic defects seven days after caerulein treatment. Interestingly, we observed an intermediate level of pancreatic hypoplasia in Bmi1KO chimeric mice that was more severe than what we found in chimeric WT control mice, but less pronounced than the defects observed in non-chimeric Bmi1 KO mice (Figure 6B, 6C). Histologic, morphometric, and immunofluorescence analysis of regenerated acinar area confirmed this observation (Figure 6B, 6D). Immunostaining for E-cadherin/cleaved Caspase 3 revealed increased apoptosis in the pancreas of the reconstituted Bmi1 KO chimeras (Figure 6E) compared to WT reconstituted controls, findings that were supported by increased expression of Cdkn2a, p16 and p19 as well as the p53 dependent pro-apoptotic genes Noxa and Apaf1 in the reconstituted Bmi1 KO chimeric mice (Figure 6F). Although there was a trend towards partially reduced expression of p16, p19, Noxa, and Apaf1 in WT reconstituted Bmi1KO mice compared to Bmi1 KO mice 7 days after caerulein, only Apaf1 expression was significantly decreased (Figure 6F). Therefore, while epithelial Bmi1 may play a major role in suppressing genes that could drive exocrine cell death, Bmi1 appears to function both in pancreatic epithelial cells as well as in hematopoietic cells to regulate acinar cell regeneration after caerulein-induced injury.
To further define the cell autonomous consequences of Bmi1 depletion, we examined changes in potential Bmi1 targets in cultured primary acini treated with 10 nM caerulein for 2 days. Immunostaining confirmed the lack of Bmi1 protein in mutant, amylase positive acini in vitro (Figure 7A–7D). We found that Bmi1 deletion results in significant upregulation of the Cdkn2a locus and its splice variants p16 and p19 even in the absence of caerulein treatment. p16 expression remained high and Cdkna2, p19, and Noxa expression were further increased two days after caerulein treatment (Figure 7E). These data support a cell autonomous role for Bmi1 in preventing expression of anti-proliferative or pro-apoptotic genes. Taken together with the in vivo result showing considerable recovery of exocrine parenchyma in WT reconstituted Bmi1 KO mice, Bmi1 appears to regulate pancreatic exocrine regeneration through a combination of cell-autonomous and non-cell-autonomous mechanisms.
Recent reports have revealed that when damaged, adult acinar cells undergo de-differentiation, regeneration, and in some cases trans-differentiation, depending on the severity of damage and the genetic context in which it takes place13, 30, 31. Here, we show that Bmi1 functions to inhibit apoptosis and permit proliferation during injury-induced acinar regeneration. Bmi1 has been shown to be critical for the self-renewal of several types of adult stem cells. Although a number of recent studies have aimed to define adult pancreatic cells with a “stem-like” capacity to recapitulate multiple cell compartments32–35, acinar cells appear to predominantly regenerate from pre-existing acinar cells that undergo de-differentiation to assume a facultative progenitor state following caerulein pancreatitis12, 15. Interestingly, we observe no gross defects in Bmi1 null pancreata unless challenged with acute chemical or diet induced pancreatitis. Therefore, our data suggests that Bmi1 predominately functions to acutely inhibit apoptosis and support proliferation of regenerating exocrine cells that have transiently changed their differentiation state, which may reflect a general role for Bmi1 in cells undergoing self renewal.
In Bmi1 KO mice, de-regulated expression of p16Ink4a and p19Arf in hematopoietic stem cells results in proliferative arrest and p53-dependent cell death, respectively5. Also, Bmi1 inhibits apoptosis in neurons through repression of p53 dependent apoptosis36. In this study, we observed dramatic upregulation of p16, p19, and p53 dependent apoptotic genes in Bmi1KO pancreata after caerulein injury. Therefore, regulation of Cdkn2a expression appears to be a general mechanism by which Bmi1 contributes to tissue regeneration. However, specific removal of Cdkn2a in acinar cells in the context of Bmi1 mutant mice would need to be performed to determine if other growth inhibitory pathways might depend on pancreatic Bmi1 expression for appropriate regulation during regeneration.
Recently, Mallen-St Clair and colleagues reported that EZH2, an indispensible component of the Polycomb repressive complex 2 (PRC2), plays a critical role in exocrine pancreas regeneration37. They observed similarly compromised acinar regeneration in mice with pancreas specific EZH2 deletion after caerulein-induced pancreatitis as observed in Bmi1 KO mice in this study37. Furthermore, they showed that deleting Ink4a in the absence of EZH2 rescued acinar regeneration, demonstrating a critical involvement of the EZH2/Ink4a axis in this process. Given the fact that the lysine H3-K2 tag catalyzed by PRC2 facilitates PRC1 mediated transcriptional repression of the Ink4a locus4, their work supports Bmi1 dependent control of Ink4a expression as an important component of exocrine regeneration. Notably, deregulated Ink4a expression is also a critical downstream target of Bmi1 and EZH2 in endocrine cells as Bmi1 KO mice and mice in which EZH2 has specifically been deleted in beta cells display similar diabetic phenotypes and deficient beta cell regeneration38.
Our data showing that exocrine regeneration is significantly restored in Bmi1 KO mice reconstituted with WT bone marrow demonstrate that non-cell autonomous effects of Bmi1, possibly due to defects in hematopoietic cells, contribute to the impaired exocrine regeneration in Bmi1 KO mice. While some inflammatory cell types, such as CD4+ T-cells and dendritic cells26 39, have been shown to play roles in both the development of acute pancreatitis and subsequent tissue regeneration, hematopoietic cell types specifically dependent on Bmi1 (such as myeloid cells and B-Cells) have not been investigated in this process. Further investigation is required to determine what roles these cell types might play in maintaining viability, proliferation, and re-establishing tissue architecture during pancreatic regeneration. Interestingly, we show that Bmi1 KO mice display persistent inflammation following caerulein pancreatitis, as evidenced by chronic cytokine expression and Stat3 signaling. Therefore, Bmi1 dependent inflammatory cells might be involved in constraining the inflammatory response during pancreatic regeneration.
Taken together, our findings that Bmi1 is required for normal acinar regeneration have possible implications for pancreatic disease in humans. Bmi1 appears to play a critical role in permitting a proliferative state that is not only required for pancreatic homeostasis, but one that shares similarities with pancreatic cells undergoing both persistent damage, such as in chronic pancreatitis, as well as transformation and malignant progression.
We thank Irving L. Weissman for providing Bmi1GFP/+ mice. We indebted to Cecilia Austin for tissue processing and preparation, Renee Vanderlaan and Ruth Taniguchi for technical help, Limor Landsman for flow cytometry analyses, and all Hebrok lab members for helpful discussion. Work in M.H.’s laboratory was supported by a grant from the NIH (CA112537). A. Fukuda was supported by a post-doctoral Research Fellowship of the Japan Society for the Promotion of Science, a Fellowship from the U.S. National Pancreas Foundation, and the Klein Family Foundation Fellowship. Image acquisition was supported by the imaging core of the UCSF Diabetes and Endocrinology Research Center (DERC) NIH grant P30DK63720. Flow cytometry analysis was supported by the DERC FACS Core.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
A.F.: study concept and design, acquisition and analysis of data, drafting of the manuscript J.P.M.IV: technical support, interpretation of data, critical revision of the manuscript M.H.: study supervision, including study design, analysis and interpretation of data, critical revision of the manuscript.
The authors declare no conflicts of interest.