Archives of Toxicology

, Volume 88, Issue 9, pp 1749–1763

MicroRNA-34a is dispensable for p53 function as teratogenesis inducer

Authors

  • Eyal Mor
    • Sackler Faculty of MedicineTel-Aviv University
  • Lin He
    • Division of Cellular and Developmental Biology, Molecular and Cell Biology DepartmentUniversity of California at Berkeley
    • Sackler Faculty of MedicineTel-Aviv University
    • Sackler Faculty of MedicineTel-Aviv University
Reproductive Toxicology

DOI: 10.1007/s00204-014-1223-9

Cite this article as:
Mor, E., He, L., Torchinsky, A. et al. Arch Toxicol (2014) 88: 1749. doi:10.1007/s00204-014-1223-9

Abstract

The tumor suppressor protein p53 is a powerful regulator of the embryo’s susceptibility to diverse teratogenic stimuli, functioning both as a teratogenesis inducer and suppressor. However, the targets that p53 engages to fulfill its functions remain largely undefined. We asked whether the microRNA (miRNA) miR-34 family, identified as one of the main targets of p53, mediates its function as a teratogenesis inducer. For this, pregnant ICR-, p53- and miR-34a-deficient mice, as well as rats, were exposed to 5-aza-2′-deoxycytidine (5-aza), a teratogen inducing limb reduction anomalies (LRA) of the hindlimbs in mice and either the hindlimbs or forelimbs in rats. Using hind- and forelimb buds of 5-aza-exposed embryos, we identified that the miR-34 family members are the most upregulated miRNAs in mouse and rat limb buds, with their increase level being significantly higher in limb buds destined for LRA. We showed that p53 mediates the 5-aza-induced miR-34 transcription followed by met proto-oncogene and growth-arrest-specific 1 target suppression in embryonic limb buds. We demonstrated that p53 regulates the teratogenic response to 5-aza acting as a teratogenesis inducer albeit miR-34a deletion does not affect the susceptibility of mice to 5-aza. Overall, our study thoroughly characterizes the expression and regulation of miR-34 family in teratogen-resistant and teratogen-sensitive embryonic structures and discusses the involvement of epigenetic miRNA-mediated pathway(s) in induced teratogenesis.

Keywords

miRNAmicroRNAmiR-34LimbsDevelopmentp53Teratogens

Introduction

A variety of roles have been shown for the p53 protein to date (Lane and Levine 2010; Vousden and Prives 2009). Among these is its role as a powerful regulator of the embryo’s susceptibility to harmful maternal stimuli and environmental embryopathic stresses (teratogens) (Torchinsky et al. 2005; Mirkes 2008). Yet, relatively little is known about the molecular mechanisms through which p53 regulates the response to teratogenic stimuli and additional in vivo studies in embryos are needed in order to reveal these mechanisms. Indeed, it has been well demonstrated that the set of genes engaged by p53 depends on the cell type and the nature and intensity (dose) of the stress (Lane and Levine 2010; Vousden and Prives 2009). This observation strongly suggests that a p53 signaling pathway formed in embryos exposed to stress may differ significantly from that formed in adults. Furthermore, evidence that p53 is capable of functioning both as a teratogenesis suppressor and inducer (Torchinsky and Toder 2010) indicates considerable diversity in the pathways through which p53 regulates the teratogenic response. Taking this into account, it is worthwhile taking note of studies that propose microRNAs (miRNAs) to be important components of diverse signaling pathways formed by p53 to fulfill its functions (He et al. 2007; Raver-Shapira et al. 2007; Chang et al. 2007; Tarasov et al. 2007; Bommer et al. 2007; Corney et al. 2007).

miRNAs are a group of small, ~22-nt-long non-coding RNAs derived from longer RNA precursors (pre-miRNA) of ~70 to 100 nt, which are processed from primary transcripts (pri-miRNA). miRNAs guide posttranscriptional repression of protein-coding genes by basepairing with the 3′ untranslated region (3′ UTR) of target mRNAs (Bartel 2009). One possible manner by which miRNAs regulate their gene targets is fine-tuning of their expression, while miRNAs control the expression of a large proportion of mammalian genes (Friedman et al. 2009; Hornstein and Shomron 2006; Krol et al. 2010). During recent years, compelling evidence suggests that miR-34 family members (miR-34a, b and c) are among the main downstream targets that p53 engages in order to regulate apoptosis (Raver-Shapira and Oren 2007; Feng et al. 2011; Hermeking 2010). Meanwhile, apoptosis in embryos is tightly regulated and the failure of this regulation inevitably entails maldevelopment (Toder et al. 2002). Together, the above data motivated us to ask whether the miR-34 family is involved in the p53-mediated response to teratogenic insults that activates apoptosis. Additionally, the paradigm proclaiming miR-34 family as an important component of the p53 network is mainly based on studies performed in in vitro models. An in vivo study performed recently in mice carrying a targeted deletion of all three members of the miR-34 family (Concepcion et al. 2012) gravely questions this paradigm. This study shows that miR-34 family function is not required for p53-mediated apoptosis induced by ionizing radiation in thymocytes of the miR-34 knockout mice. Further research using in vivo models is undoubtedly topical and come to support or dispute the findings of this study (Concepcion et al. 2012). Moreover, in vivo studies exposing embryos to teratogens, followed by tracking p53 and miR-34 levels, are of unique perspective given the different function of many genes in embryos versus adult cells and the central role of p53 in teratogenic responses.

Pregnant mice and rats exposed to such a teratogen as 5-aza-2′-deoxycytidine (5-aza) were used as models in this work. The choice of the models was based on in vivo studies demonstrating that 5-aza induces limb reduction anomalies (LRA) of the hindlimbs in mice, whereas in rats, it is able to induce LRA of either the hindlimbs or forelimbs, depending on the developmental stage at the time of exposure (Rogers et al. 1994; Branch et al. 1996, 1999; Rosen and Chernoff 2002). Thus, using 5-aza, we could monitor the expression of miRNAs and their target genes in embryonic structures destined to be malformed, simultaneously with those that resist teratogen effect and develop normally within the same embryo. In addition, the above studies have also provided evidence, suggesting that 5-aza-induced excessive apoptosis and suppression of cell proliferation are key intermediate steps in the pathogenesis of 5-aza-induced limb anomalies. In parallel, the possibility that p53 may be involved in mediating those cell responses has been suggested by experiments in different cell cultures, demonstrating that 5-aza treatment activates p53 and induces p53-dependent apoptosis as well as inhibits cell proliferation via p53-dependent activation of cyclin-dependent kinase inhibitor 1A (P21) (Karpf et al. 2001; Zhu et al. 2004).

In this study, four basic experimental models were used (see Fig. 1): (1) mice treated on day 10 of gestation (GD10, Carnegie stages ~11 to 12) exhibiting hindlimb LRA and intact forelimbs (Fig. 1a); (2) rats treated on GD 12 (Carnegie stage ~12) also exhibiting hindlimb LRA and intact forelimbs (Fig. 1b); (3) mice treated on GD9 (Carnegie stages ~9 to 10) exhibiting intact limbs (Fig. 1c); and (4) rats treated on GD 11 (Carnegie stage ~10) exhibiting forelimb LRA and intact hindlimbs (Fig. 1c).
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig1_HTML.gif
Fig. 1

Treatment schedule used to generate limb reduction anomalies (LRA) in 5-aza-treated mice and rats. Top time points at which 5-aza was injected (GD—Gestation day; CS—Carnegie stage). Bottom images of the limb phenotype demonstrated by fetuses exposed to a teratogenic dose of 5-aza at the indicated time points

Using these models, as well as p53 and miR-34a knockout mice, we asked: (1) how exposure to 5-aza affects the global expression pattern of miRNAs in embryonic limb buds. After narrowing down to a specific miRNA family, we further asked whether: (2) 5-aza-induced alterations in miR-34 family expression are organ and species specific; (3) by which mechanisms does 5-aza alter miR-34 family expression and are they p53 dependent; (4) which are the miR-34 family targets in limb buds of 5-aza-exposed embryos; and lastly (5) how does the limb phenotype correlate with the expression of p53, miR-34 family and its targets.

Results

5-Aza modifies miRNA expression in the fore- and hindlimb buds

The expression profile of a comprehensive set of miRNAs was first tested in the hindlimb and forelimb buds of mouse and rat embryos exposed to 5-aza treatment at teratogenic doses 0.5 and 1.0 mg/kg on GD 10 and GD 12, respectively. This treatment regime induces LRA of the hindlimbs but not of the forelimbs (Rosen and Chernoff 2002; Branch et al. 1999). RNA was collected 24 h following 5-aza treatment. We observed that out of hundreds of miRNAs, miR-34a and miR-34c stood out as the two most upregulated miRNAs in hindlimb buds of both mice and rats (Supp. Table 1; Fig. 2a), i.e., in the limbs destined to be malformed. At the same time, in the teratologically resistant forelimb buds, upregulation of miR-34a was far less prominent than that in the hindlimb buds and no upregulation of miR-34c was detected (Fig. 2a). When exposing mice to a sub-threshold dose of 5-aza (0.15 mg/kg), we observed that the extent to which 5-aza affects the expression of miR-34a and c is dose dependent (Fig. 2b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig2_HTML.gif
Fig. 2

miRNA profilings indicate miR-34 family elevation in limbs destined to be malformed. miRNA expression profile in limb buds tested 24 h following 5-aza injection. a miRNA expression profile in mouse (left) and rat (right) hindlimb buds destined to exhibit LRA and forelimb buds that develop normally. b miRNA expression profile in hindlimb buds of mice embryos exposed to a teratogenic dose (0.5 mg/Kg) of 5-aza versus a sub-threshold dose (0.15 mg/Kg). c Expression of miRNAs that showed a dose-dependent upregulation in mice and rat hindlimbs but no or very limited upregulation in the forelimbs. ‘M’/’N’ indicates malformed/normal limb development

In the above experiments, along with miR-34a and c, three additional miRNAs (miR-194, miR-370 and miR-192) showed dose-dependent upregulation in mice and rat hindlimbs (Fig. 2c). Yet, 5-aza-induced alterations in the expression level of these miRNAs were far less prominent than those registered for miR-34a and c. Additionally, unlike miR-34a and c, the expression level of these miRNAs did not differ significantly in the hind- and forelimb buds although tended to be higher in the former.

We thus further focused on the miR-34 family members miR-34a and c as well as on miR-34b. We note that miR-34b was absent from the miRNA profiling platform we used, yet given its potential functional role, we tested it using other methods in subsequent experiments.

The expression of miR-34 family is higher in an embryonic structure destined to be malformed

Whereas in mice 5-aza induces only LRA of the hindlimbs, in rats, it can also induce LRA of the forelimbs. These observations allowed us to examine whether the extent of 5-aza-induced miR-34 family activation in limb buds is associated with the occurrence of LRA or is it organ- or species-type dependent. We therefore examined the expression of miR-34 family in the two additional models: rats exposed to 5-aza on GD 11 which exhibit LRA of the forelimbs but not of the hindlimbs and mice exposed to 5-aza on approximately the same developmental stage (on GD 9) exhibiting no LRA. As expected, the level of miR-34a expression was higher in limb buds destined for LRA than in limb buds destined to develop normally, in both mice and rats (Fig. 3a). Remarkably, however, the extent of miR-34a elevation in limb buds destined to be malformed was comparable in all models. Similarly, it was also comparable in limb buds that develop normally in all models.
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig3_HTML.gif
Fig. 3

miR-34 family elevation level is correlated with the limb developmental fate. a Distinction between miR-34a, b and c expression in the four models of limb development toxicity. Expression was determined 24 h following 5-aza treatment. ‘M’/’N’ indicates malformed/normal limb development. b Expression kinetics of miR-34a, b and c and their pri-miRNA forms in mice hindlimb buds destined to be malformed and resistant forelimb buds in mice exposed to 5-aza on GD 10. (a + b) Fold change values were calculated relative to control animals and U6-snRNA as endogenous control. Results were analyzed statistically using the GT2 test for multiple comparisons and presented as 95 % comparison intervals for the means. Means with intervals that do not overlap are significantly different (indicated by asterisk). Means with intervals, which do not reach 1 (the level of expression in controls), differ significantly from controls

The expression level of miR-34b and c (Fig. 3b) was found to be far lower than that of miR-34a. Yet, the expression pattern of these miRNAs in fore- and hindlimb buds of rats and mice did not generally differ from that demonstrated by miR-34a.

Thus, the above results indicate that miR-34a is the member of the miR-34 family most affected by 5-aza exposure and therefore mainly addressed in further experiments. These results also suggest that the extent to which 5-aza activates the miR-34 family in the embryo is dependent more on the susceptibility of the tested organ to 5-aza-induced teratogenic stimuli at the moment of exposure than on organ or species specificity.

miR-34 genes are transcriptionally induced by 5-aza

Given that each miRNA is a derivative of its transcribed primary transcript (pri-miRNA), we measured the level of the RNAs processed to yield the miR-34 family mature miRNAs following exposure to 5-aza. We observed that the expression of both the pri-miR-34a transcript and the transcript common for miR-34b and miR-34c was elevated in the mice limb buds following 5-aza treatment on GD 10. This elevation was of higher magnitude in the hindlimb buds destined to be malformed. (Fig. 3b). Plausibly, the expression of these pri-miRNA forms preceded that of the mature forms of miR-34 family in the hindlimb buds: These pri-miRNA forms reached there peak at 6 h after treatment and then decreased toward 24 and 48 h, whereas the elevation of the mature forms of the miR-34 family in the hindlimb buds was observed as early as 6 h after treatment, peaked at 24 h and was abolished at 48 h (Fig. 3b).

Thus, these data suggests that miR-34 genes are transcriptionally induced by 5-aza and are then processed to yield their mature active forms.

Met and Gas1 are targeted by miR-34a in embryonic limb buds

Each miRNA can potentially bind a large set of target genes. In order to deepen our understanding of the potential role of the miR-34 family in distorted or normal limb development, we explored its putative gene targets. We retrieved these genes by screening for those that fulfilled the following criteria: (1) miR-34 family bioinformatic putative prediction of target genes; (2) genes that function in one of the relevant pathways associated with 5-aza-induced limb maldevelopment; and (3) genes that showed a downregulation at the same time when the miRNA levels were elevated. The last two gene lists were based on our broad gene expression profiling in the mice hindlimbs buds that are destined to be malformed, and in the forelimbs that develop normally, 24 h following 5-aza treatment on GD 10 (data not shown). The overlap of the above datasets retrieved 51 genes (Fig. 4a), of those, 25 possess a conserved miR-34 binding site on their 3′-UTRs. We subsequently used all four models described above to screen for the genes that showed a dysregulation in a limb phenotype-dependent manner. We observed that growth arrest specific 1 (Gas1), platelet-derived growth factor receptor, alpha polypeptide (Pdgfra), met proto-oncogene (Met) and 3′-phosphoadenosine 5′-phosphosulfate synthase 2 (Papss2) were downregulated in limb buds destined to be malformed. Furthermore, these genes were only modified to a lower or null extent in limb buds that develop normally (Fig. 4b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig4_HTML.gif
Fig. 4

miR-34 family and its putative targets show an inversely correlated expression. a A Venn diagram-assisted target search by crossing data from ‘Relevant genes’ (genes with a relevant role for limb development), ‘Putative targets’ for mice miR-34 according to TargetScan and ‘Dysregulated genes’ which are downregulated in mice hindlimbs destined to be malformed or upregulated in mice resistant forelimb buds (see “Materials and methods” section). b The expression of selected genes 24 h following 5-aza treatment in the four models (left). Expression kinetics in mice hindlimb buds destined to be malformed together with resistant forelimb buds in mice exposed to 5-aza on GD 10. ‘M’/’N’ indicates malformed/normal limb development. The statistics methods are as described in Fig. 3

While only Met and Pdgfra were previously shown to be targeted by miR-34a (Li et al. 2009; Silber et al. 2012), we tested which of these genes are genuine targets of miR-34a in the limb development context. Using the murine C3H10T1/2 cell line that is used as a model of limb buds at this developmental stage (Shea et al. 2003), we observed that the treatment of these cells with 5-aza results in miR-34a induction after 24 h (data not shown). We have shown above that the extent of miR-34a elevation, induced by 5-aza, is associated with the organ susceptibility to 5-aza-induced teratogenic stimuli. Thus, we wanted to mimic this effect in the cell line and test the inverse correlation between the miRNA levels and its target genes. For this purpose, we transfected miR-34a into C3H10T1/2 cells in two doses that resulted in an average overexpression of ~5- and ~20-fold (Fig. 5a). Out of the putative target genes from our screen (Gas1, Pdgfra, Met and Papss2), two, Met and Gas1, demonstrated a miR-34a dose-dependent downregulation in our cell system (Fig. 5a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig5_HTML.gif
Fig. 5

miR-34a regulates the expression of Met and Gas1 in a dose-dependent manner. a miR-34a, Met and Gas1 expression in C3H10T1/2 cells 24 h following miR-34a low (L; 4.9-fold) and high (H; 18.3-fold) transfections. Fold change was calculated relative to a scrambled RNA transfection, and U6-snRNA or Gapdh were used as endogenous controls. Values are presented as mean ± SEM (N > 3). b A region from the mouse Gas1 3′-UTRs spanning the miR-34 binding site was cloned into the psiCHECK-2 plasmid downstream of a Renilla luciferase reporter gene. This plasmid or its mutated version was transfected 24 h after miR-34a transfection into HEK293T cells. Renilla luciferase activity was measured 24 h later and normalized to firefly luciferase activity. Similarly, transfections with a scrambled miRNA and a mutated version (mut) of miR-34a binding site we used as controls. Values are presented as mean ± SEM (N > 3). c An illustration of miR-34a and its binding site on the mouse Gas1 3′-UTR (or its mutated version). P value < 0.05, 0.002 is indicated by */**, respectively (two-sided t test)

While Met is an established target of miR-34a (Li et al. 2009), we asked whether miR-34a regulation of Gas1 is also based on direct interaction between the miRNA and the gene transcript. For this purpose, we employed the Luciferase reporter assay. A region of mouse Gas1 3′-UTR spanning miR-34 family binding site was cloned downstream to a Renilla Luciferase reporter gene. This reporter plasmid was transfected following miR-34a transfection into HEK293T cells. Relative expression of the Renilla Luciferase reporter was then measured compared to that of a Firefly Luciferase reporter, a control transfection and the 3′-UTRs mutated in the miRNA binding site. We observed that miR-34a reduces the Renilla Luciferase-Gas1 activity to 0.76-fold (Fig. 5b, c; P value < 0.002), indicating a direct regulation of miR-34a upon Gas1 expression.

p53 acts as an inducer of 5-aza teratogenesis

Earlier in vitro studies implied that p53 may be involved in the mechanisms underlying 5-aza teratogenesis (Karpf et al. 2001; Zhu et al. 2004). In our current work, we used p53 heterozygous mice exposed to 5-aza on GD 10 (a model of 5-aza-induced LRA of the hindlimbs; reproductive performance in Supp. Table 2) to test this assumption. First, using Western blot analysis, we showed that the exposure to 5-aza increases the level of p53 protein in fore- and hindlimb buds of p53+/+ embryos and that this increase is more prominent in the hindlimb buds (Fig. 6a). Then, we evaluated the teratogenic response of p53+/− mice to 5-aza by examining the long bones of the hindlimbs (the femur, tibia and fibula) of GD 18 embryos and recorded abnormalities such as incomplete ossification, non-ossified bone, misshaped bone and absence of bone (Wise et al. 1997) (which indicate increasing levels of limb abnormalities). We observed that in females injected with a high dose of 5-aza (1 mg/kg), the level of postimplantation death of embryos exceeded 60 %, being significantly higher compared with that registered in controls (7.9 %) as well as in females exposed to a low dose (0.5 mg/kg) of 5-aza (10.5 %) (Supp. Table 2). Yet, a departure from the Mendelian 25:50:25 % genotype ratio was not observed in surviving fetuses. The pattern of 5-aza-induced anomalies and their incidence were found to be practically identical in p53+/+ and p53+/− fetuses (Fig. 6b). Yet, the response of p53−/− fetuses to 5-aza-induced teratogenic stimuli differed dramatically from that of their p53-positive counterparts. Indeed, when females were treated with 1.0 mg/kg 5-aza, more than 40 % of p53-positive fetuses exhibited such anomalies as a non-ossified bone, a misshaped bone and an absence of bone. At the same time, all p53 knockout fetuses had less severe anomalies such as incomplete ossification, while other anomalies were absent. Furthermore, a lower dose of 5-aza, 0.5 mg/kg, turned out to be a threshold teratogenic dose for p53−/− embryos, inducing incomplete ossification in single fetuses only. Yet, the proportion of their p53-positive counterparts exhibiting non-ossified and/or misshaped bone remained high enough (Fig. 6b). These results strongly suggest that p53 acts as an inducer of 5-aza teratogenesis.
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig6_HTML.gif
Fig. 6

p53-null embryos are less sensitive to 5-aza-induced LRA. a Representative Western blot of phosphorylated p53 in 5-aza-treated (‘T’) mice hindlimbs destined to be malformed and resistant forelimbs, with respective controls (‘C’). Quantification was performed in the hind and forelimbs relative to their controls and the actin levels. b Percentage of the different p53 genotypes’ fetuses suffering from the indicated anomalies following 5-aza treatment on GD 10. be Alizarin red and Alcian blue staining of cartilage and bone. Limbs were examined on GD 18. c Normal long bones: femur, tibia and fibula. df Representative long bones anomalies: d incomplete ossification of Tibia and Fibula; e all long bones are misshaped, incomplete ossification of the femur and no ossification of the fibula; f Incomplete ossification of the tibia while the femur and fibula are absent

p53 mediates the 5-aza-induced miR-34 family activation and miR-34a targets suppression

We tested the expression of all members of the miR-34 family in limb buds of p53-positive and p53-negative embryos, 24 h following exposure to 5-aza. We observed that the level of miR-34 family expression increase in limb buds of p53+/+ and p53+/− embryos was practically identical, being higher, as expected, in the hindlimb than in forelimb buds (Fig. 7a). The level of miR-34 family expression increase in limb buds of p53−/− embryos was significantly lower than that recorded in their p53+/+ and +/− counterparts. Remarkably, however, it was equal in the forelimb and hindlimb buds and it was higher than that in controls. These data suggest that a p53-dependent mechanism is the sole mechanism underlying the different levels of the miR-34 family activation in limb buds of different fates following 5-aza treatment. Nevertheless, a p53-independent mechanism through which 5-aza activates the miR-34 family in embryonic limb buds also exits. We noticed that the promoters of the mouse and rat miR-34 genes, in addition to p53 binding sites, contain CpG islands, similarly to the previously addressed human promoters (Hiroki et al. 2012; Mazar et al. 2011; Kalimutho et al. 2011; Nalls et al. 2011; Chen et al. 2012) (Supp. Figure 1). 5-Aza is a demethylating agent and has been shown to directly demethylate these CpG islands in humans, implying its potential to use this mechanism to activate miR-34 family in mouse limb buds. Yet, our experiments have demonstrated that the methylation of the miR-34 family promoters in mouse limb buds is very poor (Supp. Figure 1) seemingly eliminating the possibility of promoter methylation as the mode of p53-independent regulation of miR-34 (a, b and c).
https://static-content.springer.com/image/art%3A10.1007%2Fs00204-014-1223-9/MediaObjects/204_2014_1223_Fig7_HTML.gif
Fig. 7

miR-34 family elevation is p53-dependant, while miR-34a targets suppression is p53 and miR-34a dependent. Expression was tested 24 h following 5-aza treatment on GD 10. a miR-34a, b, c expression in hindlimbs (HL) and forelimbs (FL) of different p53 genotypes’ mice embryos. bMet and Gas1 expression in the hindlimbs of different p53 genotypes’ mice embryos. cMet and Gas1 expression in the hindlimbs of different miR-34a genotypes’ mice embryos. The statistics methods are as described in Fig. 3

Then, we tested the expression of miR-34a established targets, Met and Gas1 in the hindlimbs of p53-positive and p53-negative embryos. We observed that the 5-aza-induced inhibition of these genes was apparent in p53+/+ embryos but was partly (for Met) and almost completely (for Gas1) attenuated in p53−/− embryos (Fig. 7b).

miR-34a regulates Met and Gas1 expression but not 5-aza teratogenesis

The above data suggested that p53 mediates the 5-aza-induced suppression of Met and Gas1 in limb buds destined to be malformed via miR-34 family activation. In order to test whether miR-34 activation is a pathogenic event for 5-aza teratogenesis, we utilized miR-34a heterozygous mice (Choi et al. 2011) exposed to 5-aza on GD 10 (a model of 5-aza-induced LRA of the hindlimbs; reproductive performance in Supp. Table 3). No hindlimbs external anomalies, nor anomalies of the long bones, were observed in GD 18 miR-34a−/− fetuses of control mice. When mice were exposed to 0.5 mg/kg 5-aza, downregulation of miR-34a established targets, Met and Gas1, was detected in the hindlimb buds of miR-34a+/+ but not in −/− fetuses (Fig. 7c), suggesting that miR-34a is a powerful regulator of these genes in the hindlimb buds of in vivo developing embryos. At the same time, we observed that the pattern and intensity of the long bone anomalies were practically identical in miR-34a+/+ and miR-34a−/− fetuses (Supp. Table 4), not differing from those registered in p53+/+ fetuses of 5-aza-treated p53 heterozygous females. This result suggests that miR-34a plays a redundant function in the pathway(s), which p53 engages to function as an inducer of 5-aza teratogenesis.

Discussion

5-Aza is a well-known teratogen inducing limb reduction anomalies (LRA) such as phocomelia and meromelia in mice and rats in an organ- and species-specific fashion. Our study shows that p53 regulates the teratogenic response to 5-aza acting as a teratogenesis inducer. While addressing how 5-aza affects the expression profile of miRNAs in the limb buds, we observed that miR-34 family members, and most profoundly miR-34a, are the most upregulated miRNAs in mouse and rat limb buds. Further experiments revealed that 5-aza activates the miR-34 family by affecting their primary transcripts with a higher magnitude in the limb buds destined to be malformed. It has also been observed that 5-aza activates these miRNAs in both a p53-dependent and independent manner and, importantly, the activation was found to be much more prominent when mediated by p53. Altogether, the above data support earlier observations demonstrating the miR-34 family as a regulatory target of p53 (Raver-Shapira and Oren 2007; Feng et al. 2011; Hermeking 2010).

The clear correlation between miR-34a and its targets’ expression, and the fate of the developing limb (malformed versus intact), implies a regulatory function for miR-34a. Yet, in experiments in miR-34a heterozygous mice, we observed that the teratogenic response of miR-34a knockout embryos did not differ from that of their miR-34a-positive counterparts. This result indicates that miR-34a is dispensable for the function of p53 as a teratogenesis inducer.

The function of p53 as a teratogenesis inducer is firstly associated with its ability to activate the pro-apoptotic signal transduction pathway in response to teratogen-induced apoptotic stimuli (Torchinsky and Toder 2010). The miR-34 family members are suggested to be key components of p53 networks controlling apoptosis (Raver-Shapira and Oren 2007; Feng et al. 2011; Hermeking 2010). Do the miR-34b and miR-34c genes, which are co-activated by 5-aza in p53-dependent manner, compensate for miR-34a loss, exhibiting redundancy in miRNA activity, or do they act as independent apoptosis inducers?

Obviously, we cannot rule out such scenarios. It is worth noting, though, that while positive feedback loops exist between p53 and miR-34a but not between p53 and miR-34b/c [miR-34a activates p53 via suppression of sirtuin 1 (Sirt1) or via myelocytomatosis oncogene (Myc)-mediated pathway] (Chen and Hu 2012), it is conceivable that miR-34a is a far more powerful apoptosis inducer than miR-34b/c. Additionally, the level of expression was suggested to be a factor influential enough in determining the ability of miR-34 to induce apoptosis, as shown for miR-34a (Liu et al. 2011). Meanwhile, in our study, the expression level of miR-34b/c in the limb buds sensitive to 5-aza-induced teratogenic stimuli was significantly lower than that of miR-34a. In light of this evidence, it seems unlikely that miR-34b/c may fully compensate for miR-34a loss in embryos responding to 5-aza.

Thus, our results arguing against an essential role for the miR-34 family in the function of p53 as an inducer of 5-aza teratogenesis seemingly conflicts with studies indicating the miR-34 family members as important components of the p53-controlled pro-apoptotic network (Liu et al. 2011). This discrepancy is eliminated, though, if we take in mind that these studies were performed mainly in in vitro cell-based models, which are hardly suitable for studies addressing developmental toxicity due to their failure to adequately represent the full scope of developmental complexity (Lee et al. 2012). Thus, our in vivo results support Concepcion et al. study (Concepcion et al. 2012) that questions the paradigm of miR-34 role in p53-mediated apoptosis. While other tissue- and/or context-specific studies might prove otherwise (Jain and Barton 2012), we expend Concepcion et al. model to developmental toxicity.

Furthermore, the set of genes targeted by any miRNA depends on the cellular context (Chen and Hu 2012). Therefore, it is hardly surprising that, whereas there is a plethora of genes regulating apoptosis detected as miR-34a targets in various cultured cells, in the limb buds of embryos exposed to 5-aza only two genes, Met and Gas1, were shown to be genuine miR-34a targets. Can these genes be involved in determining the response of embryos to 5-aza-induced apoptotic stimuli?

Such a possibility is suggested by studies in in vitro models demonstrating that both Met and Gas1 are able to counteract and promote apoptosis in a cell-context-dependent manner (Tulasne and Foveau 2008; Nakamura et al. 2011; Martinelli and Fan 2007). At the same time, 5-aza-induced teratogenic insult causes not only suppression of Met and Gas1: we observed that it also activates many pro-apoptotic genes, including Myc, Bak1, Bax, Phlda3 and Itgb3 bp (data not presented) as well as caspase 3 (Casp3) (Torchinsky et al. 2012). And noteworthy, both heterozygous Gas1 and Met mice demonstrate normal phenotype of the long bones (Lee et al. 2001; Bladt et al. 1995). Together, the above data suggest that miR-34a-mediated suppression of Met and Gas1 may not significantly affect the intensity of excessive apoptosis initiated by 5-aza in the hind limb buds.

In conclusion, although our study suggests no functional role of miR-34a in teratogen-initiated p53-mediated apoptosis, the possibility that the miR-34 family may regulate the susceptibility of embryos to teratogens remains realistic enough. First, as mentioned above, p53 is able to act not only as a teratogenesis inducer but also as a protector and its protective role may firstly be associated with processes such as antioxidant defense and DNA repair (Torchinsky and Toder 2010). It is conceivable that miR-34 family, activated following a teratogenic exposure, can act as a component of a protective mechanism regulating these p53 activities. Second, our current study as well our previous work with cyclophosphamide (CP) (Gueta et al. 2010) shows that these teratogens are able to activate the miR-34 family acting at a sub-threshold teratogenic dose, i.e., a dose inducing no structural anomalies. At the same time, it has been shown that distorted expression of the miR-34 family in embryos results in the appearance of offspring having no external anomalies at birth but exhibiting trabecular bone architecture indicative of bone loss in adulthood (Bae et al. 2012; Wei et al. 2012). And our recent study has shown that a single exposure of pregnant mice to a sub-threshold dose of 5-aza also results in bone loss in adult offspring (Torchinsky et al. 2012). It is undoubtedly important to reveal whether other teratogens can affect offspring in this way and to what extent the miR-34 family is involved in mediating this phenomenon. Lastly, we observed that 5-aza and CP are able to activate the miR-34 family in a p53-independent fashion. This observation implies a scenario in which the miR-34 family plays a regulatory role in response to embryophatic stresses not activating the p53 pathway, while this scenario warrants further investigation. It is obvious that in vivo models should be the first choice for such studies.

Materials and methods

Ethics statement

Animal experiments were approved by the Ethics Committee for Animal Use of Tel Aviv University.

Animals

Six- to-eight-week-old ICR mice and Sprague–Dawley rats were obtained from the Tel Aviv University animal facility. Breeding pairs of p53 knockout mice (Jacks et al. 1994) and miR-34a knockout mice (Choi et al. 2011) were received from Prof. Moshe Oren (Weizmann Institute of Science, Israel) and Dr. Lin He (University of California, Berkeley), respectively. A colony of p53 and miR-34a heterozygous mice is being maintained at the Tel Aviv University Animal Facility.

The animals were maintained on a 14-h light/10-h dark cycle with food and tap water ad libitum. To obtain pregnancy, females were caged with males for 3 h, from 7 to 10 am (darkness), and the presence of a vaginal plug (for mice) or spermatozoids in a vaginal smear (for rats) at 11 a.m. was designated as day 0 of pregnancy (GD 0).

Genotyping of the p53 knockout mice was performed as described elsewhere (Pekar et al. 2007). Briefly, DNA was extracted from tails, and PCR was performed using 3 primers (Sigma): 5′-ACAGCGTGGTGGTACCTTAT-3′, 5′-TATACTCAGAGCCGGCCT-3′, and 5′-CTATCAGGACATAGCGTTGG-3′ under the following conditions: 94 °C for 3 min, 40 cycles of (94 °C for 30 s, 55 °C for 30 s, 72 °C for 1 min), 72 °C for 5 min. Genotyping of the miR-34a knockout mice was performed as for p53 knockout mice, with these exceptions: The presence of a miR-34a allele was identified using the primers (Sigma): Forward: 5′-CCAGCTGTGAGTAATTCTTTG-3′, Reverse: 5′-ACAATGTGCAGCACTTCTAG-3′, producing a 102-bp band. The presence of miR-34a knockout allele was identified by the presence of a lacZ sequence using the primers: Forward: 5′-CGTCACACTACTTCTGAACGTCG-3′, Reverse: 5′-CAGACGATTCATTGGCACCATGC-3′.

5-Aza-2′-deoxycytidine (5-aza) treatment and testing of limb phenotype

The treatment schedules were chosen based on results of both our and previous studies addressing 5-aza teratogenesis (Rosen and Chernoff 2002; Branch et al. 1999; Torchinsky et al. 2012). These studies revealed that in mice and rats, 5-aza induces limb reduction anomalies (LRA) of hindlimbs if it is injected on GD 10 and GD 12, at approximately Carnegie stage 12, respectively. Yet, in rats 5-aza also induces LRA of the forelimbs when injected on GD 11, at approximately Carnegie stage 10, whereas no LRA was recorded in mice exposed to 5-aza on GD9 paralleling the same Carnegie stage in rats. These results were reproduced in our preliminary experiments with mice and rats maintained in our animal facility (Fig. 1). Correspondingly, in this study, ICR mice were intraperitoneally injected with 5-aza-2′-deoxycytidine (Sigma) (5-aza) at a dose of 0.5 mg/kg (a teratogenic dose) or 0.15 (a sub-threshold teratogenic dose) (in 0.2 ml saline/20 g body weight) on GD9 or GD10. Rats were injected with 5-aza at a dose of 1.0 mg/kg (in 1.0 ml saline/100 g body weight) on GD 11 or GD 12. The p53 and miR-34a heterozygous females were injected with 5-aza on GD 10 at doses 0.5 or 1 mg/kg (in 0.2 ml saline/20 g body weight). Pregnant females injected with saline were used as a control throughout the study.

To compare the susceptibility of the different p53 or miR-34a genotypes’ embryos to 5-aza-induced teratogenic insult, female mice were killed by cervical dislocation on GD 18, the uteri were removed, and implantation sites, resorptions and live fetuses were recorded. Then, fetuses were stained using Alizarin red and Alcian blue protocol (Inouye 1976), and long bones of the hindlimbs (femur, tibia and fibula) were examined for absence, misshaping, incomplete or non-ossification (Wise et al. 1997).

RNA extraction

Hind and forelimbs were dissected at the indicated times following 5-aza treatment on GD 11,12 for rats and GD 9,10 for mice. Only the forelimbs of mice treated on GD 9 were dissected since the hindlimb buds cannot be precisely separated from the torso at this stage of development. Total RNA was purified using the RNeasy kit (Qiagen) for dissected limbs or Trizol (Invitrogen) for C3H10T1/2 cells. RNA quality was assessed using NanoDrop ND-1000 spectrophotometer (Thermo Scientific).

MiRNA profiling and screening

First-strand cDNA was synthesized from total RNA using Megaplex reverse transcriptase reaction with the High Capacity cDNA kit (Applied Biosystems). cDNA and TaqMan Universal PCR Master Mix (No AmpErase UNG; Applied Biosystems) was then transferred into a loading port on a Rodent TLDA card A according to the manufacturer’s instructions. PCR amplification was carried out using ABI Prism 7900HT Sequence Detection System under the following conditions: 2 min at 50 °C, 10 min at 95 °C, 40 cycles of (30 s at 95 °C and 1 min at 60 °C). miRNA relative levels were calculated based on the comparative threshold cycle (Ct) method. The RQs were calculated using the equation: RQ = 2−ΔΔCt.

Gene expression microarray analysis

Gene expression analysis was performed using Affymetrix GeneChip Mouse Gene 1.0 ST arrays according to the instruction manual, as described in the Affymetrix Web site (28,853 genes across 770,317 distinct probes; http://www.affymetrix.com). A total of eight arrays were carried out in biological duplicates. Gene profiling array was carried out on CEL files using Partek Genomics Suite™, version 6.5 (http://www.partek.com). Data were normalized and summarized with the robust multi-average method (Irizarry et al. 2003), followed by analysis of variance (ANOVA).

MiR-34 target screening

Putative miR-34 family targets were retrieved by screening for targets that fulfilled the following criteria: (1) bioinformatics target prediction: putative targets of the mouse miR-34 family according to TargetScan (Lim et al. 2003); (2) relation of the gene function to one of the gene ontology terms enriched in gene expression assays in mice hindlimbs that are destined to exhibit LRA (24 h following 5-aza treatment): skeletal system development, regulation of developmental process, cell adhesion and regulation of cell differentiation (according to ‘DAVID Bioinformatics Resources’ (Huang et al. 2009a, b); gene annotation enrichment analysis; data not shown); (3) miRNA target shows an inversely correlated expression: genes that are downregulated in mice hindlimb buds by more then 1.3-fold 24 h following 5-aza treatment.

MiRNA expression analysis by real-time polymerase chain reaction (PCR)

First-strand cDNA was synthesized from total RNA using a MultiScribe reverse transcriptase reaction with the High Capacity cDNA kit (Applied Biosystems) and TaqMan MicroRNA Assay RT primer (Applied Biosystems) for specific miRNA or U6-snRNA. Mixtures containing cDNA, TaqMan Universal PCR Master Mix (No AmpErase UNG; Applied Biosystems) and TaqMan MicroRNA Assay Real Time probe (Applied Biosystems) for each miRNA were loaded onto 96-well plates, while PCR amplification and results analysis were carried out as described above in ‘miRNA profiling and screening’ in “Materials and Methods” [thermal cycler conditions: 2 min at 50 °C, 10 min at 95 °C, 40 cycles of (15 s at 95 °C and 1 min at 60 °C)].

mRNA and pri-miRNA expression analysis by real-time PCR

cDNA synthesis of RNA isolated from limb buds or cells was carried out by the MultiScribe reverse transcriptase reaction using the High Capacity cDNA kit (Applied Biosystems). Mixtures containing cDNA, specific primers (Sigma; see below) and Power SYBR green PCR master mix (Applied Biosystems), were loaded on 96-well plates, and PCR amplification was performed as described under ‘miRNA profiling and screening’ (thermal cycler conditions: 2 min at 50 °C, 10 min at 95 °C, 40 cycles of (15 s at 95 °C and 1 min at 60 °C) and dissociation curve cycle of 15 s at 95 °C, 15 s at 60 °C and 15 s at 95 °C). Result analysis was executed as described under ‘miRNA profiling and screening,’ using glyceraldehyde-3-phosphate dehydrogenase (Gapdh) as an endogenous control.

The primers used were the following: Gapdh(mouse & rat)-Fwd: 5′ AAA GTG GAT GTC GTC GCC ATC AAT GAT 3′, Gapdh-Rev: 5′ CTG GAA GAT GGT GAT GGG ATT TCC ATT 3′; Met(mouse)-Fwd: 5′ CCA AGC AGT TCA GCA CGT AG 3′, Met-Rev: 5′ ATG ATA GAC ACA GCC AAA ATG C 3′; Met(rat)-Fwd: 5′ TCT GAA GCT TTG TTG TGT ACG G 3′, Met-Rev: 5′ TTG AAG AGA CTG CTT GCT TCC 3′; Gas1(mouse & rat)-Fwd: 5′ GCG AAT CGG TCA AAG AGA AC 3′, Gas1-Rev: 5′ CTTC GTC GTA GTA GTC GTC CAG 3′; Papss2-(mouse & rat)-Fwd: 5′ GAG GTG GCC AGG CTC TTT 3′, Papss2-Rev: 5′ GGG CAT TCT CAC GAT CCT T 3′; Pdgfra(mouse)-Fwd: 5′ TCT GGT CCT CAG CTG TCT CC 3′, Pdgfr -Rev: 5′ TCA TTC TCG TTT GGG AGG AT 3′; Pdgfra(rat)-Fwd: 5′ TCG AAG GCA GGC ACA TTT AT 3′, Pdgfra-Rev: 5′ AAT GAC TAA AGA ATC CGT CAT GC 3′; pri-miR-34 primers from (He et al. 2007): mmu-pri-mir-34a-Fwd: 5′ CTG TGC CCT CTT GCA AAA GG 3′, mmu-pri-mir-34a-Rev: 5′ GGA CAT TCA GGT GAG GGT CTT G 3′; mmu-pri-mir-34b/c-Fwd: 5′ GGC AGG AAG GCT CCA GAT G 3′, mmu-pri-mir-34b/c-Rev: CCT CAC TGT TCA TAT GCC CAT TC 3′.

Bisulfite sequencing analysis

Bisulfite sequencing analysis was performed as previously described (Lodygin et al. 2008). Genomic DNA was extracted from hindlimb buds dissected from GD 10 mice, and 2 ug of DNA was treated with bisulphite using the EpiTect Bisulphite Kit (QIAGEN). This kit converts unmethylated cytosine to uracil, leaving methylated cytosine are unaffected. Enzymatically methylated control DNA (CpGenome Universal Methylated DNA, Chemicon) was used as positive control. For mmu-miR-34a promoter analysis, bisulphite-treated genomic DNA was used as a template to amplify fragments of 539 bp covering the 469-bp-long CpG island (see Supp. Figure 1) using the primers (cytosine converted to uracil is indicated in italics): Fwd: TTG TTT GTT ATA GGTTTA TTT AGG TTT, Rev: ACA AAC CCA AAC ACA ACC CCT AC. For mmu-miR-34b/c promoter, amplification was performed for two fragments of 337 bp and 395 bp, together covering the 648-bp-long CpG island. For all reactions, FastStart Taq DNA Polymerase (Roche) was used as follows: 95 °C for 5 min, 30 cycles of (95 °C for 30 s, 55 °C for 40 s and 64 °C for 1 min) and 64 °C for 7 min. Gel-purified PCR products were sub-cloned into a TOPO-TA vector (Invitrogen). For each PCR fragment, 10–12 individual clones were sequenced on both strands and the methylation status of each CpG pair was examined.

5-Aza treatment and miRNA transfection of C3H10T1/2

For 5-aza treatment, murine mesenchymal cell line C3H10T1/2 (Shea et al. 2003) was seeded in 12-well plates (0.7 × 105 cells) in DMEM supplemented with 10 % FBS and 1 % Pen/Strep (penicillin/streptomycin). 5-Aza-2′-deoxycytidine (100 µM; Sigma) was added at 90–100 % confluence. For miR-34a transfection, cells were seeded in 12-well plates (0.7–0.9 × 105 cells) in DMEM supplemented with 10 % FBS. Transfection of pre-miR-34a (Ambion) was performed after 24 h using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer’s instructions, with 20 pmol (~28 nM; low transfection)/80 pmol (~115 nM; High transfection) or a scrambled RNA. In all C3H10T1/2 experiments, cells were harvested for RNA purification after 24 h.

Gas1 3′ UTR plasmid

Fragments of ~400 bp of Gas1 3′ UTR spanning the miR-34 binding site was cloned into the XhoI-NotI restriction site downstream of the Renilla Luciferase Reporter gene of the psiCHECK-2 plasmid (Promega) that also contains a Firefly Luciferase Reporter (used as control) under a different promoter. For this purpose, the 3′ UTR fragments were PCR-amplified from mice genomic DNA and XhoI–NotI restriction sites were added (italics), using the primers: Fwd: 5′ ACA CTC GAG TCC ATC GGT AAT GCT CAG TG 3′, Rev: 5′ AAG GTC AAG CGG CCG CTA CAA AGT ACA GCA ACT GGT A 3′. The miRNAs binding sites were site-directed mutated (4 bases in the seed region) by a PCR of the plasmid using the enzyme PfuUltra II Fusion HS DNA Polymerase (Genex), and the PCR : 95 °C for 2 min, 16 cycles of (95 °C for 20 s, 58 °C for 20 s and 72 °C for 2 min) and 72 °C for 3 min. The primers used for mutagenesis are those indicated here (target nucleotides in italics), along with the complementary reverse primer: 5′ CCT AAA GCT CGG TAC CAA TAT CTA GGA AAA CCT C 3′. Products were incubated with DpnI (New England BioLabs), and the mutated plasmid was sequenced to confirm mutagenesis prior to use.

Dual-Luciferase assay

HEK293T cells were seeded in 24-well plates in DMEM supplemented with 10 % FBS and 1 % Pen/Strep (penicillin/streptomycin). After 24 h, cells were transfected with pre-miR-34a or scrambled RNA (Ambion) using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer’s instructions. After additional 24 h, the cells were transfected with the plasmid psiCHECK-2 containing the desired 3′ UTR with or without site-directed mutations using the TransIT-LT1 Transfection Reagent (Mirus), according to the manufacturer’s instructions. After additional 24 h, firefly and renilla luciferase activities were measured using the Dual-Luciferase Reporter Assay System kit (Promega) and a Veritas microplate luminometer, according to the manufacturer’s instructions.

Western blotting

Mice limbs were collected and lysed in IGEPAL lysis buffer (Sigma-Aldrich), and quantified using the Bio-Rad protein assay (Bio-Rad Laboratories). 60 μg of the protein samples were resolved on 10 % SDS PAGE gels, transferred to a nitrocellulose membrane (Whatman), blocked with 5 % milk in TBS-tween or 1 % BSA- and 10 % sodium azide in TBS-tween, and incubated in a primary antibody diluted in blocking solution at 4 °C overnight. Actin (Clone C4 (MAB1501, Millipore) and phospho-p53 (Ser15) (9284, Cell Signaling Technology) with the corresponding HRP-conjugated secondary antibodies were used. Detection was carried out using the ECL method (Thermo scientific) and developed using X-ray film (Kodak). Band intensities were quantified using Image-J, while relative expression levels were normalized to the expression of Actin.

Statistical analysis

The expression of tested miRNAs/mRNAs was evaluated using between three and four samples of fore- and hindlimbs prepared from embryos taken from different litters. The GT2-method for multiple comparisons (Sokal and Rohlf 1995) was used to analyze the data. For the statistical analysis of data characterizing 5-aza-induced hindlimb anomalies in p53+/− and miR-34a+/− mice, an embryo was used as an independent variable, and the chi-square test was run to look for an association across treatment groups and embryonic genotypes as described (Fleiss 1981). The two-tailed level of significance of differences was equal to 0.05 for all tested parameters.

Acknowledgments

The authors would like to acknowledge Avital Gilam and Daphne Doron for their appreciated contribution. The Shomron laboratory is supported by the Wolfson Family Charitable Fund; Claire and Amedee Maratier Institute for the Study of Blindness and Visual Disorders; Israel Cancer Research Fund; Earlier.org—friends for an earlier breast cancer test; I-CORE Program of the Planning and Budgeting Committee of the Israel Science Foundation (Grant No. 41/11).

Supplementary material

204_2014_1223_MOESM1_ESM.doc (30 kb)
Supplementary material 1 (DOC 30 kb)
204_2014_1223_MOESM2_ESM.jpg (46 kb)
Supplementary material 2 (JPEG 45 kb)
204_2014_1223_MOESM3_ESM.xls (93 kb)
Supplementary material 3 (XLS 93 kb)
204_2014_1223_MOESM4_ESM.doc (30 kb)
Supplementary material 4 (DOC 30 kb)
204_2014_1223_MOESM5_ESM.doc (30 kb)
Supplementary material 5 (DOC 30 kb)
204_2014_1223_MOESM6_ESM.doc (28 kb)
Supplementary material 6 (DOC 28 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2014