Fetal-derived adrenomedullin mediates the innate immune milieu of the placenta (original) (raw)
Adm-null placentas exhibit parietal trophoblast giant cell death and reduced labyrinth vessel branching. We previously showed that Adm is expressed in trophectoderm cells of the preimplantation blastocyst and that its expression is increased when trophoblast stem cells are differentiated into the giant cell lineage (31). Using in situ hybridization, we found that this robust expression was maintained in vivo within fetal parietal trophoblast giant cells (TGCs; polyploidy derivatives of the primary mural trophectoderm cells that line the implantation site; refs. 32–34) at E9.5 and E13.5 (Figure 1, A–C), consistent with previous studies (35). There was also strong expression of Adm surrounding the invading ectoplacental cone at E9.5 and at the innermost border of the maternal-fetal interface (Figure 1, A and B). Adm expression persisted in TGCs at E13.5, and stromal cells of the maternal decidua expressed moderate levels of Adm (Figure 1C). The chorionic plate, embryo, fetal membranes, labyrinth, and spongiotrophoblast-containing junctional zone expressed little to no Adm.
Fetal loss of Adm causes trophoblast apoptosis at the maternal-fetal interface. (A) In situ hybridization of Adm gene expression in WT E9.5 mouse placentas, revealing robust expression in parietal TGCs. (B) Digital zoom of boxed region in A (enlarged ×2-fold), showing punctuate Adm staining in TGCs lining the ectoplacental cone (epc) at the innermost border of the maternal-fetal interface and little to no expression in the chorionic plate (cp), embryo (emb), or fetal membranes (fm). (C) Adm expression persisted in TGCs at E13.5 (arrows) and was diffusely expressed in stromal cells throughout the maternal decidua (de). The labyrinth (lb) and spongiotrophoblast-containing junctional zone (jz) expressed little to no Adm. (D) H&E staining of placentas from Adm+/+ and Adm–/– littermates revealed largely normal structures, with no difference in the thickness of the labyrinth layer (Supplemental Figure 1). Digital zoom of central part of placentas is shown at the right of each image (enlarged ×1.5-fold). (E) TUNEL staining of placentas from Adm+/+ and Adm–/– littermates showed a prominent band of apoptosing cells in Adm–/– placentas that colocalized to TGC location and correlated with the high level of Adm expression in these cells. For better clarity and data presentation, the original colors from the captured images in E were inverted using Adobe Photoshop. Original magnification, ×4 (A–C). Scale bars: 1 mm (D); 500 μM (E).
Although Adm–/– mice die at midgestation with lymphatic vascular defects (36), we wondered whether genetic dosage of fetal Adm could also contribute to pregnancy outcomes. Histological and morphometric analyses showed that Adm–/– placentas appeared overtly normal, with well-formed layers and labyrinth thickness that were indistinguishable from those of Adm+/+ littermates (Figure 1D and Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI67039DS1). Because a hallmark function of AM peptide is vasodilation, some have speculated that its expression in placenta may be associated with maintaining low-resistance blood flow through the placental vasculature and umbilical cord (37). Contrary to this hypothesis, we found that loss of Adm had no effect on placental impedance, as calculated by umbilical cord blood flow using in vivo Doppler ultrasound (Supplemental Figure 2).
Despite their overtly normal appearance and function, E13.5 Adm–/– placentas had a prominent band of apoptosing cells that spatially localized to the innermost border of the maternal-fetal interface (Figure 1E), where _Adm_-expressing parietal TGCs are located, which suggests that AM is an essential survival factor for these cells. Consistent with this finding, qRT-PCR analysis of trophoblast marker genes revealed that Adm–/– placentas had significantly reduced expression of placental lactogen 1 — a specific marker of parietal TGCs — while the expression of other TGC lineage markers and spongiotrophoblast markers did not differ significantly from that of WT placentas (Supplemental Figure 3). Even though Adm is not abundantly expressed in the labyrinth, we also found a modest, yet significant, reduction in the labyrinth-expressing genes Gcm1, JunB, and Gab1 (Supplemental Figure 3).
Further characterization of the labyrinth layer showed marked defects in the vascular patterning of fetal vessels in Adm–/– placentas. Isolectin B4 staining of the ECM surrounding fetal labyrinth vessels revealed a highly branched network in Adm+/+ placentas, but abnormally large and underbranched fetal vessels in Adm–/– placentas (Figure 2A). Using nucleated and non-nucleated blood cells as markers for fetal and maternal blood sinuses, respectively, we performed quantitative morphometric analysis of the labyrinth vasculature (Figure 2B). Consistent with our finding of normal labyrinth thickness and normal placental impedance, the total area of vascular space in Adm–/– labyrinth did not differ from that in Adm+/+ placentas (Figure 2C). However, we found that individual fetal sinuses of Adm–/– placentas were significantly larger than those of Adm+/+ placentas (P < 0.01; Figure 2D). This increase in the size of fetal sinuses was accompanied by a concomitant reduction in the size of maternal sinuses in Adm–/– versus Adm+/+ placentas (P < 0.02; Figure 2D).
Reduced branching of fetal labyrinth vessels in Adm–/– placentas. (A) Isolectin B4 staining revealed a highly branched network in Adm+/+ placentas, in contrast to abnormally large and underbranched fetal vessels in Adm–/– placentas. (B) Fetal (blue) and maternal (red) blood sinuses demarcated within the labyrinth layer. (C) Quantitative morphometric analysis of total sinus space per field. (D) Quantitation of individual sinus area revealed that fetal sinuses of Adm–/– placentas were significantly larger than those of Adm+/+ placentas, with a concomitant reduction in size of maternal sinuses. n = 12 placentas per genotype. *P < 0.02; #P < 0.01. (E and F) Scanning electron microscopy of vascular corrosion casts of maternal placental vasculature, at low power (E) and higher magnification (F). The large holes (white arrows) were indicative of large fetal vessels. Images are representative of n = 4 placentas per genotype. (G) Alkaline phosphatase staining of labyrinth revealed no structural or quantitative differences in chorionic villus cells. Data are mean ± SEM. Scale bars: 50 μM (A, B, and G); 1 mm (E); 100 μM (F).
To better visualize the architecture and structure of the labyrinth vasculature, we made vascular corrosion casts of the placentas by perfusing pregnant female mice with a liquid polymer. Fetal placental tissue was enzymatically digested away from the hardened maternal casts in order to reveal the interdigitation of maternal and fetal vasculature within the labyrinth placentas of Adm+/+ and Adm–/– littermates. Therefore, in the scanning electron micrographs of Figure 2, E and F, the open spaces represent the footprint of the fetal vasculature as it was interdigitated between the maternal blood sinuses within the central labyrinth layer. Adm–/– placentas had large holes (Figure 2E, arrowheads) and higher-magnification views demonstrated vast spacing between the maternal sinuses (Figure 2F), reflective of the large and under-branched phenotype of Adm–/– fetal vessels. Taken together, these data demonstrated a failure of fetal vessels to branch appropriately in Adm–/– placentas.
Because shallow invasion of chorionic villi can lead to blunted fetal vascular branching, we used alkaline phosphatase staining to evaluate the extent of chorionic villus formation and found no significant difference between WT and Adm–/– placentas (integrated density, Adm+/+, 11.8 ± 2.0 AU; Adm–/–, 11.3 ± 2.0 AU; Figure 2G). This finding demonstrated that the fetal vascular branching defects of Adm–/– placentas were inherent to the fetal vasculature, rather than secondary to defects in chorionic villus invasion or formation. Collectively, the data presented in Figures 1 and 2 showed that compared with those of their WT littermates, Adm–/– placentas exhibited increased apoptosis of parietal trophoblast cells accompanied by reduced fetal vessel branching.
Failed SA remodeling and reduced uNK cells in Adm-null placentas. Because several clinical studies have shown that maternal levels of AM are reduced in preeclampsia (23), we evaluated another characteristic pathology of preeclampsia in Adm–/– placentas: remodeling of maternal SAs. As expected, SAs of WT placentas progressively lost staining for SMA as they approached the maternal-fetal interface (Figure 3A). In contrast, SAs from Adm–/– placentas retained a thick smooth muscle layer throughout the decidua that was nearly twice the thickness measured in WT placentas (P < 0.01; Figure 3, A and B). The shedding of VSMCs surrounding SAs is associated with localized degradation of the ECM, a process in which the cathepsin family of proteases has been shown to play a predominant role (10, 38). As expected, SAs from Adm+/+ placentas were tightly surrounded by a prominent band of cathepsin staining, whereas cathepsin staining was weak and diffuse throughout the deciduas of Adm–/– placentas and did not surround the maternal SAs (Figure 3C), consistent with a failure of these SAs to shed their smooth muscle coverage. The enlargement of maternal SAs at midgestation is also associated with proliferation of vascular endothelial cells, a process termed reendothelialization. Using BrdU incorporation assays, we readily observed this robust endothelial cell proliferation in all SAs of Adm+/+ placentas, but this was rarely evident in the SAs of Adm–/– placentas (Figure 3D).
Fetal Adm dosage influences maternal SA remodeling and placental uNK cell content. (A) Anti–α-SMA staining showed that decidual SAs of E13.5 Adm–/– placentas retained thick coverage of VSMCs (arrowheads) compared with those of Adm+/+ littermates. (B) Morphometric analysis showed a statistically significant increase in the thickness of SMA staining surrounding the SAs of Adm–/– versus littermate Adm+/+ placentas. *P < 0.01. (C) Cathepsin staining of decidual SAs. (D) BrdU incorporation assays showed marked proliferation of vascular endothelial cells (arrowheads) in all SAs of Adm+/+ placentas, but this was rarely evident in SAs of Adm–/– placentas. (E) Perforin staining of uNK cells within deciduas at E13.5. (F) Quantitation of perforin+ uNK cells showed a significant reduction in Adm–/– versus Adm+/+ placentas. *P < 0.05. For all analyses, n = 10 placentas per genotype. Scale bars: 50 μM.
Numerous studies have pointed to maternal uNK cells as essential mediators of SA remodeling, and the precise factors contributing to their recruitment, activation, and function remains an area of intense study. Therefore, we stained and counted the number of uNK cells using 2 different uNK markers that identify mature (perforin-containing) and/or activated (Ly49G2+) uNK cells. Consistent with the phenotype of failed SA remodeling, we found that deciduas of Adm–/– placentas had half the number of perforin-containing uNK cells observed in Adm+/+ littermates (P < 0.05; Figure 3, E and F). Similarly, levels of Ly49G2+ uNK cells were also reduced by approximately half in Adm–/– placentas, although the difference did not reach statistical significance.
Taken together, our findings demonstrated that lack of fetal AM caused numerous placental pathologies, including reduced fetal vessel branching, failed SA remodeling, and reduced number of uNK cells. We also noted that the incidence and degree of the Adm–/– placental phenotype was strictly and exclusively correlated with the genotype of the individual Adm–/– conceptus, not influenced by the genotype of neighboring conceptuses.
Adm–/– placental phenotype is independent of maternal genotype. A confounding factor to the present findings is that the maternal genotype used in the above studies was haploinsufficient for AM; we have previously shown that during early gestation, Adm+/– females exhibit reduced fertility due to abnormal implantation (30, 31). To eliminate this confounding factor, we performed ovary transplantations in which Adm+/+ ovaries or Adm+/– ovaries (containing gene-targeted _Adm_-null gametes) were transplanted into WT recipient females (referred to herein as Adm+/+ ovary→WT or Adm+/– ovary→WT recipients, respectively), which were then bred to Adm+/– males (Figure 4A). This approach allowed us to evaluate the consequences of loss of fetal AM within the context of a WT maternal uterus and immune system and to circumvent the inherently low offspring resulting from embryo transfer approaches that we previously experienced with the 129S6/SvEv strain (30). The average success rate of ovarian transplants (the percentage of pregnant mice relative to the total undergoing transplantation) was 57.5% and did not differ based on recipient or donor genotype. The transplanted ovaries exhibited overtly normal histology, with numerous ovarian follicles and corpora lutea (Supplemental Figure 4), suggestive of normal estrogenic function. In addition, compared with Adm+/+ donor ovaries, Adm+/– donor ovaries did not adversely affect the recipient females’ breeding fecundity, since both exhibited similar recovery times to first parturition (Adm+/+ ovary→WT recipients, 56.3 days; Adm+/– ovary→WT recipients, 50.6 days; Figure 4C) and similar litter sizes (Figure 4D).
Ovary transplantation reveals independence of maternal genotype on Adm–/– placental phenotype. (A) Ovary transplantation protocol. Donor ovaries from Adm+/– or Adm+/+ female mice were surgically sutured in the ovarian bursa of WT 129S6/SvEv recipient mice after removal of their own ovaries. The recipient females were then bred to Adm+/– male mice to generate Adm+/+, Adm+/–, or Adm–/– offspring. (B) Mendelian ratios of offspring were as expected, and Adm–/– mice exhibited characteristic embryonic edema (arrowhead) at E13.5. (C and D) Recovery time to first parturition (C) and average litter size (D). Numbers within bars denote total number of litters analyzed per breeding. (E–I) Adm–/– placentas from Adm+/– ovary→WT recipient females displayed the same pathological phenotypes of Adm–/– placentas from Adm+/– females — including (E) reduced fetal vessel branching in the labyrinth, (F) retention of SMCs (arrowheads) around maternal SAs, (G) reduction in endothelial cell proliferation (arrowheads) in maternal SAs, and (H and I) significantly reduced DBA+ uNK cell numbers — compared with Adm+/+ littermates. *P < 0.05. For ovary transplant studies, n = 6–10 placentas per genotype. Scale bars: 50 μM.
The cross of Adm+/– males to Adm+/– ovary→WT recipient females resulted in the expected Mendelian ratio of fetal genotypes and the previously characterized phenotype of extreme embryonic edema and lethality for Adm–/– fetuses (Figure 4B and refs. 36, 39). Consistently, the placental phenotypes of underbranched fetal labyrinth vessels, poorly remodeled SAs, and significantly reduced uNK cell numbers were completely recapitulated in Adm–/– placentas born to Adm+/– ovary→WT recipient females compared with Adm+/+ littermate placentas and Adm+/+ placentas derived from Adm+/+ ovary→WT recipient females (Figure 4, E–I). Thus, loss of Adm from fetal tissues directly contributed to the observed placental pathologies, even in the context of genotypically normal maternal Adm.
Recapitulation of phenotypes in placentas genetically lacking the AM receptor. AM peptide can bind to and signal through the G protein–coupled receptor calcitonin receptor–like receptor (Calcrl; encoding CLR). We have previously shown that mice with genetic deletion of Calcrl phenotypically recapitulate the embryonic lethality and lymphatic vascular defects observed in Adm–/– mice (36, 40), thereby providing genetic and in vivo evidence to substantiate CLR as a bona fide AM receptor.
Using in situ hybridization, we consistently found high levels of Calcrl expression in WT placentas that spatially and temporally colocalized with that of the peptide ligand Adm. Specifically, Calcrl expression was robustly present in parietal TGCs at both E9.5 and E13.5 (Figure 5, A–C). In contrast to the diffuse expression of Adm surrounding the ectoplacental cone and stromal decidual cells, expression of Calcrl was concentrated to endothelial-like cords of the E9.5 decidua (41) and to maternal endothelial cells lining the maternal SAs of E13.5 placentas (Figure 5C, inset).
Fetal loss of Calcrl recapitulates Adm–/– placental phenotypes. (A) In situ hybridization of Calcrl gene expression in WT E9.5 mouse placentas revealed expression in parietal TGCs, similar to the pattern of Adm expression. Unlike Adm, Calcrl was expressed at moderate levels in the embryo and fetal membranes. There was also robust expression of Calcrl in cord-like structures throughout the early decidua. (B) Digital zoom of boxed region in A (enlarged ×2-fold), showing punctuate Calcrl staining in TGCs lining the ectoplacental cone at the innermost border of the maternal-fetal interface and little to no expression in the chorionic plate. (C) Calcrl expression was diffuse throughout the labyrinth at E13.5 and robustly expressed in the maternal endothelial cells lining the decidual SAs (arrows, inset; enlarged ×2-fold). (D) H&E staining of placentas from Calcrl+/+ and Calcrl–/– littermates revealed largely normal structures, with no appreciable differences in layers. Digital zoom of central part of placentas is shown at the right of each image (enlarged ×1.5-fold). (E–H) Calcrl–/– placentas at E13.5 displayed the same pathological phenotypes of Adm–/– placentas — including (E) reduced fetal vessel branching in the labyrinth, as visualized by isolectin staining, (F) retention of SMCs around maternal SAs, and (G and H) significantly reduced number of DBA+ uNK cells — compared with Calcrl+/+ littermate placentas. *P < 0.001. n ≥ 6 placentas analyzed per genotype. Data are mean ± SEM. Original magnification, ×4 (A–C). Scale bars: 1 mm (D); 50 μM (E and G); 10 μM (F).
Like Adm–/– placentas, the Calcrl–/– placentas appeared overtly normal and contained well-formed placental layers and normal histomorphometric features (Figure 5D). Nevertheless, the Calcrl–/– placentas exhibited the same abnormal phenotypes of Adm–/– placentas, including reduced fetal vessel branching, retention of SMCs surrounding maternal SAs, and significantly reduced uNK cell numbers compared with Calcrl+/+ littermate placentas (Figure 5, E–H). Thus, the recapitulation of the Adm–/– placental phenotype in Calcrl–/– placentas further supports the role of CLR as an AM receptor within this tissue and indicates that lack of AM signaling within the fetal compartment is causally associated with these phenotypes.
Overexpression of fetal Adm reverses the placental phenotype and drives maternal uNK cell recruitment to the decidua. The absence of apoptosis within decidual regions that contain uNK cells (Figure 1E) suggests that the reduced number of uNK cells in Adm–/– placentas may be caused by abnormalities in their recruitment to the tissue, rather than their loss through cell death. To determine whether fetal AM can directly influence the recruitment of uNK cells to the decidua, we generated and characterized a novel gene-targeted mouse model in which AM expression was increased approximately 3-fold (Figure 6, A–C). The gene targeting was designed to increase AM expression approximately 3-fold via stabilization of mRNA levels through genetic modification of the 3′ untranslated region (3′UTR) of the endogenous Adm gene.
Genetic overexpression of fetal Adm reverses the placental preeclampsia phenotypes and drives uNK recruitment to the decidua. (A) Targeting vector for generation of Admhi/hi mice consisted of (a) a 6-kb genomic fragment of the Adm gene isolated from a 129S6/SvEv genomic phage library and containing all 4 exons and 5′UTR and 3′UTR of the Adm gene, (b) the bovine growth hormone polyA sequence (bGH 3′UTR), (c) 2 tandem copies of the 1.2-kb 5′ insulator sequence from chicken β-globin gene (2XIns), (d) 1.3 kb of pMC1 promoter–driven neomycin (Neo), (e) 80 bp of AU/U-rich element of the mouse c-fos gene (ARE), and (f) 2 loxP recombination sites. The latter 5 elements were cloned as a cassette, 23 bp downstream of the endogenous Adm stop codon. (B) Southern blot analysis on genomic DNA confirmed correct targeting of the Admhi allele. (C) Adm gene expression in placentas from Adm+/+ and Admhi/hi mice, analyzed by quantitative RT-PCR, showed a significant 3-fold increase in gene expression level. *P < 0.05. (D–G) Admhi/hi placentas (D) appeared histologically comparable to Adm+/+ placentas and showed (E) highly branched fetal labyrinth vessels, (F) appropriate SA remodeling, and (G) reendothelialization. (H and I) DBA staining revealed that Admhi/hi placentas had a significant 30% increase in uNK cell numbers compared with Adm+/+ littermate placentas. *P < 0.05. Data are mean ± SEM. Scale bars: 1 mm (D); 50 μM (E–H).
Adm+/hi mice were viable and fertile and were intercrossed in order to compare placental phenotypes between Adm+/+ and Admhi/hi littermates. The average litter size of Admhi/hi females was 6.04 live births, which was not significantly different than the 6.60 pups per litter observed for WT C57BL/6J mice of the comparable genetic background. Admhi/hi placentas born to Adm+/hi intercrossed mice appeared overtly normal and had appropriately branched labyrinth fetal vessels that were indistinguishable from Adm+/+ littermates (Figure 6, D and E). Moreover, the SAs of Admhi/hi placentas were appropriately remodeled and did not differ from those of Adm+/+ littermates (Figure 6, F and G). Interestingly, Dolichos biflorus agglutinin (DBA) staining and counting of uNK cells revealed that Admhi/hi placentas had a significant 30% increase in the number of uNK cells within the decidua compared with that of Adm+/+ littermate placentas (Figure 6, H and I). These data indicate that 3-fold overexpression of Adm is compatible with normal placental development and, moreover, that fetal-derived AM can actively promote the recruitment of uNK cells to the placenta.
AM dose-dependently alters the chemokine, cytokine, and MMP profiles of uNK cells in vivo and in vitro. uNK cells constitute the largest proportion of immune cells in the decidua, and so we expected and found that the dynamic fluctuations in uNK cell recruitment between Adm–/– and Admhi/hi placentas were reflected by concomitant changes in the expression of numerous chemokines and cytokines. Specifically, the expression of Ccl7, Ccl17, Cxcl9, Cxcl10, Xcl1, and TNF were downregulated in Adm–/– placentas and concomitantly upregulated in Admhi/hi placentas compared with their respective WT controls (Figure 7A). To determine whether AM peptide directly causes secretion of chemokines and cytokines from uNK cells, we isolated uNK cells and first determined that these cells expressed high levels of Calcrl and survived up to 48 hours in culture without evidence of apoptosis, as evaluated by TUNEL and trypan blue exclusion (Supplemental Figure 5 and data not shown). Then, using a Luminex-based detection system, we found that treatment of uNK cells with 10 nM AM stimulated marked secretion of a variety of chemokines and cytokines (Figure 7B), several of which have established functions in reproductive immunology (i.e., CXCL10, GM-CSF, and IL-23) (42). Furthermore, we found that AM treatment dose-dependently increased the gene expression and activity of MMP9, but not MMP2, in uNK cells (Figure 7, C–E), consistent with the previously described functions of uNK-derived MMP9 in SA remodeling (43). Collectively, these data suggest that fetal-derived AM potently influences the immune milieu of the placenta in 2 ways: first by recruiting, then by activating, uNK cells to secrete chemokines, cytokines, and MMPs, which are important contributors to SA remodeling.
AM is a direct activator of uNK cells in vivo and in vitro. (A) Using a mouse Chemokines and Receptors array platform (see Methods), the expression of 85 chemokines, cytokines, and related proteins was evaluated in placental RNA extracts from Adm–/–, Admhi/hi, and respective WT littermates. Admhi/hi placentas showed significant elevations in numerous cytokines and chemokines that were concomitantly downregulated in Adm–/– placentas. *P < 0.05. (B) uNK cells isolated from E10.5–E12.5 WT placentas were cultured in vitro and treated with 10 nM AM. Screening for more than 25 chemokines and cytokines from media samples revealed significant 2- to 4-fold increases in secretion of numerous chemokines and cytokines and a prominent reduction in the secretion of CCL5, IL-16, and IL-23 in response to AM treatment. All changes were statistically significant compared with untreated control (P < 0.05). (C) Isolated WT uNK cells exhibited dose-dependent increases in Mmp9 gene expression, which (D and E) correlated with increased MMP9 zymography activity. Mmp2 gene expression levels and zymography activity were not affected by AM treatment. *P < 0.05.
AM-treated uNK cells stimulate VSMC apoptosis. Finally, we sought to determine whether the effects of AM on isolated uNK cells directly relate to the phenotype of reduced SA remodeling in Adm–/– placentas. To test this, we developed an in vitro cell culture system in which primary mouse VSMCs were treated with control media or uNK-conditioned media that had been supplemented or not with 10 nM AM peptide. As predicted, treatment of VSMCs with control media and AM-supplemented control media had no obvious effects on cell growth, morphology, or apoptosis (Figure 8, A and B). In contrast, treatment of VSMCs with uNK-conditioned media resulted in dramatic changes in cellular morphology and apoptosis rate. Importantly, these effects were markedly and significantly exacerbated when the uNK-conditioned media was supplemented with AM, with a nearly 4-fold increase in the ratio of proapoptotic Bax to antiapoptotic Bcl2 gene expression (Figure 8C). Thus, we conclude that AM can promote the secretion of a cocktail of chemokines and cytokines from uNK cells that can in turn induce apoptosis of VSMCs.
Conditioned media from AM-treated uNK cells causes VSMC apoptosis. (A and B) Confluent monolayers of cultured mouse VSMCs were placed for 12 hours in control media or uNK-conditioned media treated or not with 10 nM AM peptide, then imaged (A) and stained with TUNEL (B). (C) After treatment with uNK-conditioned media, expression levels of the proapoptotic Bax and antiapoptotic Bcl2 genes were analyzed from VSMC RNA lysates by quantitative RT-PCR; data are expressed as Bax/Bcl2 ratio. *P < 0.05, 1-way ANOVA. Data are mean ± SEM. Original magnification, ×20 (A).