The context-dependent role of transforming growth factor-β/miR-378a-3p/connective tissue growth factor in vascular calcification: a translational study

Strategy 1: exogenous TGF-β1 effect

After 7 days of osteogenic media (OM) exposure (the mature phase), treated rat aortic smooth muscle cells (ASMCs) developed prominent Alizarin red (AR) positively stained nodules under gross (Figure 1A, upper row) and microscopic examinations (Figure 1A, lower row), while none was observed in untreated ones. We next incubated ASMCs with increasing concentrations of TGF-β1 and compared calcification propensity to those treated with OM. Results showed that 2.5 and 5 ng/mL TGF-β1 induced 4-fold and 8-fold increased calcification compared to the control group, although the degree of elevation remained lower than that in the OM-treated group (Figure 1B). We subsequently chose 5 ng/mL TGF-β1 in the following experiments. Phenotypically, exogenous TGF-β1 without OM induced mild albeit significant dose-dependent increase in AR positively stained nodules (Figure 1C, 1D, upper row), while TGF-β1 cotreated with OM led to more prominent increase in AR positive nodules compared to that in OM-only and the control groups (Figure 1C, 1D, lower row). OM induced 10-fold greater calcium deposition than the control, while 2.5 and 5 ng/mL TGF-β1 cotreated with OM led to 80% greater calcification severity relative to the OM only group (Figure 1E). We also examined the influences of TGF-β1 on calcification in human ASMCs. TGF-β1 treatment only without OM similarly led to a dose-dependent increase in AR positively stained nodules compared to the control after incubation of 14 days (Figure 1F); this effect took place after TGF-β1 higher than 0.5 ng/mL (Figure 1G). TGF-β1 co-treated with OM also enhanced calcification compared to the OM only group after 7 days of incubation (Figure 1H). We derived that 2 and 5 ng/mL TGF-β1 induced significantly higher calcification severity compared to the OM only group (Figure 1I).

Figure 1. The influence of exogenously administered TGF-β1 on vascular calcification during the mature phase. TGF-β1 exhibited a dose-dependent increase in calcification regardless of OM exposure across ASMC of different species. (A) Gross (upper row) and microscopic (lower row) view of ASMCs treated with OM, AR-stained (B) Calcium quantitation results of ASMCs without or with different TGF-β1 concentrations or OM treatment. (C) Gross and (D) microscopic view of ASMCs treated without or with TGF-β1 alone, OM alone, or in combination, AR-stained. (E) Calcium quantitation results of ASMCs without or with TGF-β1 alone, OM alone, or in combination. (F) Human ASMCs without OM and with different concentrations of TGF-β1, stained with AR. Quantitative results are shown in panel (G). (H) Human ASMCs with OM co-treated with different concentrations of TGF-β1, stained with AR. Quantitative results are shown in panel (I). AR, Alizarin red; ASMC, aortic smooth muscle cell; ctrl, control; OM, osteogenic media; TGF-β1, transforming growth factor-β1.

We next evaluated early (day 3) and mid-term (day 5) influences on calcification posed by TGF-β1. During the early phase of VC, exogenous TGF-β1 without OM did not increase AR positive nodules relative to the control group, while TGF-β1 cotreated with OM led to significantly more AR positive nodules, followed by the OM only group (Figure 2A, 2B). Calcification quantification results also lent support to findings shown on qualitative examinations (Figure 2C). During mid-term calcification, AR positive nodules increased prominently in the OM only and OM cotreated with TGF-β1 groups compared to the control (Figure 2D, 2E), while the TGF-β1 only group exhibited only a trend of increasing calcification compared to the control without significant differences during mid-term calcification (Figure 2F).

Figure 2. The context-dependent influences of TGF-β1 on VC, based on different test strategies, during the early and mid- term phases of calcification. OM and TGF-β1 synergistically increased calcification upon co-treatment. During early (day 3) calcification, ASMCs exposed to TGF-β1 alone, OM alone, or in combination, were stained using AR, with gross (A) and microscopic (B) images shown. Calcium quantitation results of ASMCs without or with TGF-β1 alone, OM alone, or in combination during early calcification are shown in panel (C). During mid-term (day 5) calcification, ASMCs exposed to the same condition were stained with AR (gross, D; microscopic, E) images shown. Calcium quantitation results during mid-term calcification are shown in panel (F). AR, Alizarin red; ASMC, aortic smooth muscle cell; ctrl, control; OM, osteogenic media; TGF-β1, transforming growth factor-β1; VC, vascular calcification.

Strategy 2: Endogenous TGF-β1 expressions upon OM exposure

We next examined TGF-β1 expressions in ASMCs treated with OM. TGF-β1 mRNAs were significantly suppressed throughout the early, mid-term, and mature phase of ASMC calcification, with the degree of suppression shrinking toward the mature phase (early, 58%; mid-term, 18%; and mature, 19%) (Figure 3A). TGF-β1 protein levels did not differ between the OM only group and the control during the early phase (OM to control, 1.023) but declined in the OM only group during the mid-term phase (OM to control, 0.785) (Figure 3B). In addition, we examined the canonical receptors of TGF-β1 signaling, Smads in this model (Supplementary Figure 1). Smad1, Smad2, and Smad5 expressions decreased significantly during the mature phase after OM exposure compared to the control, while Smad2 and Smad5 expressions also decreased significantly during the mid-term and early phases, respectively. However, no significant differences were observed regarding Smad3 expression through the entire calcification period.

Figure 3. Changes in the expressions of TGF-β1 and its potential effectors, miR-378a-3p and CTGF upon OM exposure. TGF-β1 induced an early up-regulation of CTGF, followed by mid-term miR-378a-3p stimulation, while OM suppressed all effectors tested since the early phase of calcification. From early, mid-term to mature VC, OM-induced TGF-β1 (A, mRNAs; B, western blots), miR-378a-3p (early, C; mid-term, D; mature, E), and CTGF (early, F; mid-term, G; mature, H) expressional changes are shown. ctrl, control; CTGF, connective tissue growth factor; OM, osteogenic media; TGF-β, transforming growth factor-β.

Investigating TGF-β1/microRNA/connective tissue growth factor (CTGF) signaling in VC

To verify the biologic consequences of TGF-β1 suppression related to OM exposure, we evaluated the expressions of calcification-regulating miRNA (miR-378a-3p) [10] and connective tissue growth factor (CTGF) levels at different periods of calcification. During the early (Figure 3C), mid-term (Figure 3D), and mature (Figure 3E) phases of calcification, OM consistently and significantly suppressed miR-378a-3p expressions in ASMCs (early, 32%; mid-term 27%; and mature, 17%). Similarly, OM induced significantly lower CTGF expressions in ASMCs over time (early, 42%; mid-term, 33%; and mature, 22%) (Figure 3F–3H, respectively). These findings supported the possibility that miR-378a-3p and CTGF might be downstream effectors of TGF-β1 regarding VC. On the other hand, ASMCs treated with TGF-β1 without OM induced significantly higher miR-378a-3p expression starting from the mid-term (Figure 3D) to the mature phase (Figure 3E) of calcification. The up-regulation of miR-378a-3p induced by TGF-β1 exhibited dose-dependency during the mature phase (Figure 3E). TGF-β1, in the absence of OM exposure, also up-regulated CTGF expressions during the early (Figure 3F), mid-term (Figure 3G), and mature phases (Figure 3H) (early, 71%; mid-term 68%; and mature, 41%). TGF-β1 cotreated with OM led to a milder up-regulation of miR-378a-3p during the early (Figure 3C) and mature phases (Figure 3E) of calcification, and also a milder up-regulation of CTGF during the early (Figure 3F), mid-term (Figure 3G), and mature phases (Figure 3H) of calcification.

To address whether miR-378a-3p served as an upstream regulator of CTGF, we over-expressed miR-378a-3p in ASMCs and evaluated calcification severity as well as CTGF levels. ASMCs with miR-378a-3p overexpression showed significant down-regulation of RUNX2 (Figure 4A, 4B) during the early and mid-term phases, and also alkaline phosphatase (ALP) (Figure 4C) during the mid-term phase of calcification. These findings indicated that miR-378a-3p overexpression suppressed vascular calcification. miR- 389a-3p overexpression did not alter TGF-β1 mRNA levels during the early (Figure 4D) and mid-term phases (Figure 4E) of calcification. On the contrary, miR-378a-3p overexpression did not alter CTGF expressions during the early phase (Figure 4F) but suppressed CTGF expressions during the mid-term phase (Figure 4G). Importantly, miR-378a-3p overexpression acted synergistically with OM exposure to lower CTGF expressions in ASMCs (Figure 4G). The extent of miR- 378a-3p overexpression paralleled that of CTGF suppression (Figure 4H).

Figure 4. Functional evaluation of the role of miR-378a-3p during VC and ex vivo validation. Over-expression of miR-378a-3p led to CTGF suppression while remained neutral regarding TGF-β1 expressions. From early to mid-term VC, the expressions of RUNX2 (early, A; mid-term, B), ALP (C), TGF-β1 (early, D; mid-term, E), and CTGF (early, F; mid-term, G) in ASMCs without or with miR-378a-3p over-expression or OM treatment are shown. (H) The effect on CTGF expression exerted by increasing miR-378a-3p transfection dose. (I) AR staining of ex vivo aortas without or with OM, with aortic expressions of TGF-β1 (J), miR-378a-3p (K), and CTGF (L) shown. ALP, alkaline phosphatase; AR, Alizarin red; ASMC, aortic smooth muscle cell; ctrl, control; CTGF, connective tissue growth factor; OM, osteogenic media; TGF-β, transforming growth factor-β.

Result validation in the ex vivo model

We further evaluated the expressions of TGF-β1 and miR-378a-3p in an organ culture model. Upon OM incubation, animal aortas exhibited prominent circumferential calcification compared to those without OM (Figure 4I). Calcified aorta exhibited a significant decrease in TGF-β1 (Figure 4J), miR-378a-3p (Figure 4K), and CTGF (Figure 4L) expressions compared to the control.

Bioinformatic analysis of the TGF-β1/miR-378a-3p/CTGF relationship

The paths between TGF-β1 and miR-378a-3p, and between miR-378a-3p and CTGF were able to form a regulatory network, resulting in a hierarchical search tree space. The node in brown square, red diamond, grey triangle, and green circle denoted the miRNA, mRNA, source/target and transcription factor, respectively. Bioinformatic analyses showed that TGF-β1 modulated the expressions of miR-378a-3p, which potentially influenced the longevity regulating pathway, cellular senescence, and osteoclast differentiation through targeting dedicated gene clusters (Figure 5A). The down regulation of miR-378a-3p further influenced CTGF expressions in stem cells [11], and this connection might go through one to several layers of mediators illustrated in Figure 5B. In addition, CTGF was also a downstream direct target of the TGF-β1 signaling pathway and participated in osteoblastic differentiation involving ASMCs. Although miR-378a-3p expressions were suppressed during VC, leading to the release of CTGF inhibition, CTGF expressions were also decreased following OM-induced TGF-β1 suppression (Figure 5C). Such a negative feedback signal triggered by miR-378a-3p following TGF-β1 signaling points to a potential balance between pro-calcific and anti-calcific influences following OM exposure [12].

Figure 5. Bioinformatic analyses of potential links within the TGF-β1/miR-378a-3p/CTGF axis, with multiple mediators uncovered. The expression levels of TGF-β1/miR-378a-3p/CTGF were found to decrease in sera of patients with severe aortic arch calcification compared to those without. Hierarchical clustering of potential linking nodes/effectors between TGF-β1 and miR-378a-3p (A), miR-378a-3p and CTGF (B), and between TGF-β1, miR-378a-3p, and CTGF (C). Serum TGF-β1 (D), CTGF (E), and miR-378a-3p (F) levels of older adults according to the absence of AAC or presence of severe AAC are shown. AAC, aortic arch calcification; CTGF, connective tissue growth factor; TGF-β, transforming growth factor-β.

Clinical validation results

We validated our findings in sera from community-dwelling relatively healthy older adults without aortic arch calcification (AAC) and with severe AAC. For circulating miR-378a-3p, we previously showed that miR-378a-3p levels were lower among patients with AAC than those in patients without AAC [13]. In this study, totally 35 participants with a mean age of 75.8 ± 6.0 years and 13 (37.1%) males were identified. Among them, 24 and 11 did not have and had severe AAC, respectively. No significant differences were noted between the two groups regarding their age (p = 0.98), body mass index (p = 0.26), or the prevalence of morbidities including hypertension (p = 0.47) and diabetes mellitus (p = 0.94). The AAC group had a significantly lower circulating TGF-β1 (non-AAC vs. AAC, 9.62 ± 8.25 vs. 1.85 ± 3.36 ng/mL, p = 0.005) (Figure 5D) and CTGF levels (non-AAC vs. AAC, 24.93 ± 6.39 vs. 19.98 ± 7.11 ng/mL, p = 0.048) (Figure 5E) than the non-AAC one. Circulating miR-378a-3p levels were borderline lower in those with severe AAC compared to those without AAC (the former vs. the latter, 0.134 ± 0.207 vs. 0.55 ± 0.763, p = 0.086) (Figure 5F).

Experimental utilities

Rat ASMCs were purchased from American Type Culture Collection (ATCC) (A7r5), and maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 15% fetal calf sera (FCS). Human ASMCs were kind gifts from Dr. Liao and Dr. Shih in Armed Force Taoyuan General Hospital. AR stain and recombinant TGF-β1 were bought from Sigma-Aldrich (St. Louis, MO, USA). Primers for TGF-β, CTGF, RUNX2, ALP, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) were custom-designed by Mission Biotech. (Taipei, Taiwan), with forward/reverse sequences provided in Table 1. The primer for miR-378a-3p was obtained from QIAGEN (Valencia, CA, USA). Primary rabbit anti-rat TGF-β1 (1:1000) and β-actin (1:10000) were purchased from Abcam (Cambridge, UK) and Sigma, respectively, while secondary goat anti-rabbit antibody (1:3000) were purchased from Abcam (Cambridge, UK).

The in vitro model of VC and severity quantitation

The detailed procedures of establishing in vitro VC have been outlined in our prior work [10, 38]. Briefly, ASMCs were seeded onto culture plates until 80% to 90% confluence, when OM containing high phosphate (mixture of NaH₂PO₄ and Na₂HPO₄ leading to 2.5 to 3 mM inorganic phosphate concentration) were administered for durations ranging from 3 to 7 days. Such media were exchanged every 2 days. After treatment, ASMCs were washed with phosphate buffered saline and ethanol-fixed. Two approaches were used to measure calcification extent. First, we used AR to qualitatively examine positively stained calcification nodules in the control and OM groups. Second, we used the acid elution method to extract deposited calcium from ASMCs according to our prior work [38]. For rat ASMCs, we used the calcium quantitation kit (Calcium Detection Kit; Abcam, Cambridge, UK) based on cetylpyridinium chloride extraction, using spectrophotometry of OD 560-575 nm absorbance. For human ASMCs, we used an acid extraction method for recovering Alizarin red dye, followed by ammonium hydroxide neutralization and spectrophotometrical detection of calcium amount at OD 450 nm. Calcium quantitative results were normalized to cell number by counting or protein amount measured by a protein amount assay (Biorad, Hercules, CA, USA).

Reverse transcription-polymerase chain reaction and Western blot

We extracted total RNA from ASMCs and reassured RNA quality using 260/280nm and 260/230nm ratios. This was followed by reverse transcription (RT) and real-time quantitative polymerase chain reaction (qPCR) using the RT kit (Abcam, Cambridge, UK) or miRCURY LNA RT kit (QIAGEN) and SYBR green-based kit (Abcam, Cambridge, UK) or miRCURY LNA SYBR Green PCR kit (QIAGEN), respectively. We calculated fold changes of gene expressions relative to GAPDH or β-actin, using the delta-delta Ct method. The amount of protein of interest was determined using Western blot, with antibodies outlined above.

Testing the role of TGF-β in VC: time- and dose-dependency

We adopted two strategies for evaluating the role of TGF-β in the process of VC, exogenous administration and stimulated expression upon pro-calcific exposure. During the former experiments, we evaluated the effects of different concentrations (1 – 5 ng/mL) of TGF-β1 on the severity of ASMC without and with OM-induced calcification. During the latter experiments, we evaluated TGF-β1 expression levels in OM-induced ASMC calcification at different temporal periods (early, day 3; mid-term, day 5; and mature, day 7). This arbitrary designation of calcification phases was made based on the pathophysiological observations in other in vitro calcification models and calcification-prone animals, which disclosed calciprotein formation in as early as 3 days [39], rapid progression of VC compared to other calcification sites [40], and our prior experimental findings [10, 38]. This was followed by the evaluation of expressions of TGF-β downstream signaling (CTGF) and calcification-regulating miRNA (miR-378a-3p). Existing literature reported that the TGF-β1/CTGF pathway was heavily involved in vascular fibrosis [41], and CTGF up-regulation might contribute to ASMC osteogenic differentiation [19]. We previously showed that miR-378a-3p over-expression suppressed VC progression in ASMCs and animal models [10]. Moreover, TGF-β1 was shown to up-regulate miR-378a-3p expressions in podocytes [42] but whether this action is demonstrable in ASMCs and participates in VC pathogenesis remains unclear.

Ex vivo model of VC

We subsequently tested our in vitro findings in an ex vivo VC model according to our prior work [10, 38]. The animal experiment protocol, which was part of an integrative VC pathophysiology and therapeutic project, has been approved by the Institutional Animal Care and Use Committee (IACUC) of National Taiwan University College of Medicine (NTU-CM) (NO. 20190048). Briefly, wild type C57BL/6 mice were purchased from LASCO (Taipei, Taiwan), housed and acclimatized in the Laboratory Animal Center of NTU-CM with food and water provided ad libitum. Aortas were harvested and subjected to DMEM or osteogenic media, cultured for up to 9 - 14 days. We assessed the expression of TGF-β1 and miR-378a-3p without or with OM and/or exogenously administrated TGF-β1.

microRNA transfection experiments

To dissect the regulatory relationship between TGF-β, miR-378a-3p, and CTGF, we transfected ASMCs with miR-378a-3p and analyzed the mRNA levels of TGF-β1 and CTGF. Rat ASMCs were seeded onto 24-well plates, and we used Lipofectamine RNAiMAX (Invitrogen) to deliver mature miR-378a-3p (miRCURY LNA miRNA Mimics; QIAGEN). After reassuring transfection efficiency, we performed qPCR for TGF-β1 and CTGF. We also examined whether the degree of miR-378a-3p overexpression influenced CTGF expressions.

Bioinformatic analysis of the TGF-β/miR-378a-3p/CTGF relationship

We integrated miRNA-gene-transcription factor (TF) networks which considered multiple transcriptional and post-transcriptional interactions from several databases. We also gathered protein-protein interactions from the Biogrid database (https://thebiogrid.org/), containing 4,964 unique proteins and 8,683 interactions involving rattus norvegicus [43], and the TF-miRNA from TransmiR v2.0 database (http://www.cuilab.cn/transmir) which was manually curated from publications [44]. We then extracted targets of the miRNA using the miRNAtap database in R version 3.4 (https://bioconductor.org/packages/release/bioc/html/miRNAtap.html). We predicted targets of miR-378a-3p using all 5 possible target-predicting databases (DIANA, Miranda, PicTar, TargetScan, and miRDB). These tools predicted miRNA targets based on different binding strategies, and we considered candidate targets of miRNAs predicted by at least three tools, and the union of targets was taken as the target output results for a given miRNA. Finally, we adopted the depth-First search (DFS) strategies to identify reachable nodes and paths from source nodes to target nodes during the construction of miRNA-gene-TF networks.

Clinical validation studies

In this study, we used cryopreserved sera from community-dwelling older adults participating in a cytokine and miRNA study for measuring circulating TGF-β1 and CTGF, based on enzyme-linked immunosorbent assays (ELISAs) (ab100647 and ab261851; Abcam). Their VC severity was determined based on a validated classification scheme for AAC [45], which also demonstrated good applicability to our patients according to our prior work [13, 46]. We identified those without AAC and those with severe AAC, and compared their circulating TGF-β1 and CTGF levels using the Student’s independent t-test, in order to validate the clinical implications of TGF-β1 and CTGF in VC.

Availability of data and material

The raw data for conducting this analysis are will be available upon reasonable request to the corresponding author and approved by relevant authorities.

Code availability

Custom codes were used in this study.