Aged mice develop brain calcification primarily in the thalamus region. Brain calcification is commonly detected in elderly individuals. We evaluated whether aged mice also develop brain calcification by susceptibility-weighted imaging (SWI) (35), which is an advanced in vivo imaging technique for the detection and quantification of mouse brain calcification, given the associated high sensitivity and accuracy of this imaging modality (36, 37). Susceptibility-weighted images of male mice showed hypointensities only in the thalamic region of 22-month-old male mice (Figure 1A, yellow arrows). No hypointensity areas were found in 3-month-old male mice. Phase map analysis revealed similar findings (Supplemental Figure 1A, yellow arrows; supplemental material available online with this article; https://doi.org/10.1172/JCI168447DS1). Five of 6 (83.3%) 22-month-old male mice showed hypointensities in the thalamus (Figure 1B). Surprisingly, we detected no hypointensities in any brain regions of female mice at 3 months or 22 months of age (Figure 1, A and B). Quantification of the total calcification load in the aged male mice revealed an average of 0.3531 ± 0.161 mm3 volume of brain calcification (Figure 1C). The calcification loads of the 5 samples were located at the same thalamic region (i.e., slices 8–13 from the caudal side of the thalamus) (Figure 1D). To cross-validate the finding of hypointensities by SWI MRI, we performed micro-CT (μCT) scanning of the samples. Hyperdense lesions were indeed detected only in the thalamus of the aged male mice (Supplemental Figure 1B). There were no hyperdense lesions found in 3-month-old male mice or in female mice at either age. Furthermore, immunofluorescence staining of tissue sections from the samples with hypointensities identified by SWI revealed multiple nodules in thalamus that highly expressed the bone matrix protein osteopontin (OPN) (Supplemental Figure 1C), a well-established marker of brain calcification (35, 38). Thus, the hypointensity lesions detected by SWI scanning were indeed brain calcification.

Aged male mice develop brain calcification at the thalamic regions. (A) Axial and coronal SWI sequence images of brains from 3-month-old male mice (3M-M), 22-month-old male mice (22M-M), 3-month-old female mice (3M-F), and 22-month-old female mice (22M-F). Calcifications (yellow arrows) were observed as black structures in the thalamic region on susceptibility-weighted images. n = 6–9. (B) Calcification incidence calculation based on the SWI analysis in A. (C) The volume of calcification in different brain regions, including the cortex, hippocampus, thalamus, and hypothalamus, in 22-month-old male mice was determined on the basis of SWI analysis. n = 5. (D) Quantification of the volume of calcification in individual serial sections (750 μm thick) by SWI throughout the entire thalamus, from caudal to rostral aspects. n = 5. (E) Alizarin red staining of brain tissue sections from 3- and 22-month-old male mice. Boxed areas are shown at higher magnification (×10) in the corresponding panels on the right. The calcification nodule is shown in red. n = 9. (F) Calculation of the calcification incidence in 3- and 22-month-old male mice based on the histology stainings. n = 9. (G) Double-immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against CD13 and OPN. n = 5. (H) Quantification of the volume of the OPN+ calcified nodules in G. n = 5. The Imaris 3D reconstruction method was used to quantify the number and volume of calcification. (I and J) Double-immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against OPN and OCN (I) or OPN and CatK (J). n = 3. Scale bars: 100 μm (G, I, and J). (K) Immunofluorescence staining was performed on brain sections from male mice at 3 and 22 months of age using an antibody against ALPL. n = 3. Scale bar: 100 μm. Data are shown as the mean ± SD. **P < 0.01, by unpaired, 2-tailed Student’s t test (H).

We conducted a more comprehensive characterization of the brain calcification using male mice at the ages of 3 months and 22 months. Consistent with the SWI results, Alizarin red staining of brain tissue sections detected clusters of calcified nodules in the thalamus of 22-month-old mice but not in that of 3-month-old mice (Figure 1E). Calcified nodules were detected in the thalamus region in 7 of 9 mice at the age of 22 months, but none of the 3-month-old young mice developed thalamic calcification (Figure 1F). Moreover, OPN was highly expressed in these lesions in the thalamus (Figure 1G). Costaining with the blood vessel pericyte (PC) marker CD13 confirmed that the OPN-stained aggregates were associated with vessels (Figure 1G). Quantification of the calcification volume revealed dramatic differences between young and old mice (Figure 1H). In addition, co-immunofluorescence staining showed that OPN was well localized with the osteoblast-secreted protein osteocalcin (OCN) (Figure 1I) and the osteoclast-secreted protein cathepsin K (CatK) (Figure 1J) in the calcification lesions. Vessel-associated calcifications have previously been found in the thalamus in old mice with increased expression of ALPL (39), which is an enzyme that regulates mineralization in bone and other tissues. We also detected an increase in ALPL expression in the thalamus of aged mice relative to expression levels in young mice (Figure 1K). Taken together, these results demonstrate that brain vessel–associated calcifications develop in male mice during natural aging and that the lesions were located primarily in a specific thalamic region.

Overexpression of PDGF-BB in preosteoclasts is sufficient to induce brain calcification. We previously found that elevated circulating PDGF-BB is a key mediator for age-associated vascular pathologies (4, 32). Here, we also detected dramatically increased serum PDGF-BB concentrations in 22-month-old male mice relative to concentrations in 3-month-old male mice (Figure 2A). Relative to young (3-month-old) male mice, serum PDGF-BB levels in aged (22-month-old) male mice were nearly 3-fold higher. Surprisingly, contrary to the elevated serum PDGF-BB concentration observed in aged male mice, 22-month-old female mice exhibited lower serum PDGF-BB concentrations than did young female mice (Figure 2B). Therefore, there was a sex difference (linear regression model, sex-by-age interaction effect: P = 7 × 10–7) in the alteration of circulating PDGF-BB levels with age. To further examine whether increased serum PDGF-BB levels are involved in age-associated brain calcification, we took advantage of an established conditional Pdgfb-transgenic mouse line (PdgfbcTG), in which Pdgfb is overexpressed in skeletal TRAP+ preosteoclasts (4, 40). PDGF-BB+ cells were detected in bone and bone marrow of WT littermates (Supplemental Figure 2), consistent with our previous finding that preosteoclasts are a major cell type that express PDGF-BB in bone and bone marrow (4, 33). The number of PDGF-BB+ cells in PdgfbcTG mice was substantially higher than in WT mice (Supplemental Figure 2), confirming the successful overexpression of PDGF-BB in the bone tissue of this conditional transgenic mouse line. Notably, serum PDGF-BB levels in PdgfbcTG mice were more than 2-fold higher relative to levels in WT mice at a young age (6 months old) (Figure 2C), mirroring the abnormality in aged male mice. In our previous study, we discovered that the elevated levels of circulating PDGF-BB in aged mice or PdgfbcTG mice were primarily generated by bone/bone marrow preosteoclasts, rather than being derived from blood myeloid cells (4). Here, we examined whether cells in brain produce excessive PDGF-BB in aged mice and PdgfbcTG mice. We first conducted an analysis of a publicly available single-cell RNA-Seq (scRNA-Seq) data set (Gene Expression Omnibus [GEO] GSE129788), in which scRNA-Seq was performed to profile the cellular composition and transcriptomes of young and old mouse brains (41). Although many brain cell types, such as endothelial cells (ECs), VSMCs, PCs, microglia (MGs), and hemoglobin-expressing vascular cells (Hb-VCs), express Pdgfb in young mice, none of these cell types expressed more Pdgfb in aged mice (Supplemental Figure 3A). Moreover, immunofluorescence staining of brain tissue sections from PdgfbcTG mice showed no increase in the numbers of PDGF-BB–expressing MGs (TMEM119+ and Iba1+ cells) (Supplemental Figure 3, B and C) or PCs (CD13+ cells) (Supplemental Figure 3D). Therefore, local brain cells did not overexpress PDGF-BB in aged mice or in PdgfbcTG mice.

Preosteoclast-secreted PDGF-BB is sufficient to induce brain calcification. (A–C) ELISA measurements of serum PDGF-BB concentrations in 3- and 22-month-old male mice (n = 5) (A), 3- and 22-month-old female mice (n = 6) (B), and 6-month-old PdgfbcTG mice and WT littermates (n = 4) (C). (D) Calculation of calcification incidence in 6-month-old PdgfbcTG mice and WT littermates. n = 9. (E) Representative images of Alizarin red staining of brain tissue sections from 6-month-old PdgfbcTG mice and WT littermates. The calcification nodule is shown in red. Boxed areas are shown at a higher magnification (×10) in the corresponding panels on the right. n = 9. (F) Double-immunofluorescence staining of frozen brain tissue sections from 6-month-old PdgfbcTG using antibodies against CD13 and OPN. n = 3. (G) OPN expression in hippocampus, thalamus, and cortex from PdgfbcTG mice and WT littermates was measured by Western blot analysis. n = 5. (H) The relative intensity from G was calculated using ImageJ. n = 5. (I) Immunofluorescence staining of frozen brain tissue sections from PdgfbcTG mice and WT littermates using antibodies against ALPL. n = 3. Boxed areas are shown at higher magnification (×5) in the corresponding panels on the right. n = 3. Scale bars: 100 μm (E and I); 200 μm (F). Data are shown as the mean ± SD. *P < 0.05, **P < 0.01, and ***P < 0.001, by unpaired, 2-tailed Student’s t test (A–C and H).

We then assessed whether brain calcification would occur in young PdgfbcTG mice, similar to the change observed in aged male mice. As expected, calcification lesions, as detected by Alizarin red staining, were found in the thalamus of 6-month-old PdgfbcTG mice but not in the age-matched WT mice (Figure 2E). We detected calcified nodules in 6 of 9 PdgfbcTG mice, and only 1 of 9 WT littermates had thalamic calcification (Figure 2D). OPN+ aggregates associated with vessels were also found in PdgfbcTG mice but not in WT mice (Figure 2F). Western blot analysis consistently showed upregulation of OPN expression in the thalamus of PdgfbcTG mice relative to WT mice (Figure 2, G and H). We also detected an increase in ALPL expression in the thalamus of young (6-month-old) PdgfbcTG mice relative to age-matched WT mice (Figure 2I).

Normalization of circulating PDGF-BB abolishes age-associated brain calcification. We next assessed whether normalizing the level of circulating PDGF-BB using a genetic approach would attenuate cerebral vascular calcification in aged mice. We took advantage of an established conditional Pdgfb-KO mouse line (PdgfbcKO) (4, 33, 40), in which circulating PDGF-BB levels can be reduced by approximately one-half. Consistent with the previous observation, we found that aged WT mice had increased serum PDGF-BB concentrations, whereas PdgfbcKO mice had dramatically reduced serum PDGF-BB concentrations. The reduced PDGF-BB levels were almost equivalent to those in healthy young mice (Figure 3A). While calcified nodules were present in the thalamus of aged Pdgfbfl/fl (WT) mice, as detected by Alizarin red and von Kossa staining, the lesions were not detected in the same region in any of the age-matched PdgfbcKO mice tested (Figure 3, B and C). Immunofluorescence staining of brain tissue sections showed vessel-associated OPN+ nodules in the thalamus of aged WT mice. However, OPN+ nodules were not found in thalamus of PdgfbcKO mice (Figure 3, D and E). Therefore, the aged conditional Pdgfb-KO mice with normalized circulating PDGF-BB levels did not develop age-associated cerebral vascular calcification like the aged WT mice did.

Deletion of Pdgfb from bone preosteoclasts alleviates brain calcification. (A) ELISA measurement of serum PDGF-BB concentrations in 17-month-old PdgfbcKO male mice and WT littermates. n = 4–5. (B) Calculation of the calcification incidence in PdgfbcKO mice and WT littermates based on histology. n = 6. (C) Representative images of Alizarin red– and von Kossa–stained images for 17-month-old PdgfbcKO mice and WT littermates. Calcification nodules are shown in red and black, respectively. Boxed areas are shown at higher magnification (×40) in the corresponding panels on the right. n = 6. (D) Double-immunofluorescence staining of frozen brain tissue sections from 17-month-old PdgfbcKO mice and WT littermates using antibodies against CD13 and OPN. n = 3–4. (E) Quantification of the volume of OPN+ calcified nodules in D. n = 3–4. The Imaris 3D reconstruction method was used to quantify the number and volume of calcification. Scale bars: 100 μm (C and D). Data are shown as the mean ± SD. **P < 0.01 and ***P < 0.001, by unpaired, 2-tailed Student’s t test (A and E).

Recombinant PDGF-BB induces brain vessel calcification and activates the osteogenic program in an ex vivo cerebrovascular culture system. To further characterize the molecular changes underlying cerebral vascular calcification, we developed an ex vivo cerebral microvessel culture system, in which the intact microvessel fragments isolated from the entire brain were cultured (Figure 4A). Importantly, using this method, brain vascular cell compositions, including CD31+ ECs, CD13+ PCs, and GFAP+ astrocyte endfeet, were consistently yielded (Figure 4B) and were able to respond to various treatments. We then incubated the cultured cerebral microvessels with recombinant mouse PDGF-BB or vehicle in the presence or absence of the mineralization inducer phosphate plus calcium (Pi+Ca). Although there were no evident mineral deposits in the microvessel incubated with Pi+Ca alone, the addition of 20 ng/mL PDGF-BB markedly increased vessel calcification, as visualized by Alizarin red staining (Figure 4C). Moreover, a higher concentration of PDGF-BB (50 ng/mL) increased more vessel calcification, as detected by Alizarin red, von Kossa, and ALPL staining (Figure 4D).

PDGF-BB treatment promotes calcification in an ex vivo cerebrovascular model. (A) Schematic diagram shows the ex vivo brain vessel isolation and culturing procedure (details are described in Methods). (B) Characterization of microvessel preparations. Representative images of immunofluorescence analysis of isolated microvessels stained with antibodies for each cell component: endothelial marker CD31 (green), PC marker CD13 (red), and astrocyte marker GFAP (green). (C and D) The isolated brain microvessels were incubated for 16 hours under the indicated culture conditions. The concentrations of Pi and Ca were 2.6 mM and 2.7 mM, respectively. Calcification was detected by Alizarin red, von Kossa, and ALPL staining. Scale bars: 100 μm.

Previous studies of arterial calcification demonstrated that pathological medial calcification is a process analog to bone mineralization with VSMCs entering an osteoblast-like differentiation program with increased expression of bone-specific transcription factors and osteogenic differentiation–associated factors (42, 43). To assess whether PDGF-BB induces similar changes in cerebral microvessels, we took advantage of real-time profiler PCR array technology to chart the changes in gene expression of 84 genes related to osteogenic differentiation, including growth factors and transcription factors mediating osteogenesis. Intriguingly, the majority of osteogenesis-associated genes (80 of a total of 84 genes) were upregulated, and only 4 genes were downregulated in PDGF-BB–treated microvessels compared with vehicle-treated vessels (Figure 5A), indicating a broad activation of the osteogenic differentiation program by PDGF-BB. Following the analysis, genes that exhibited a P value of less than 0.05 were selected. A heatmap was then generated to visualize the expression patterns of these genes. Among the top upregulated genes, there were many bone extracellular matrix protein–encoding genes (e.g., collagen family member Mmps, Bglap, Spp1, and integrins), growth factor genes (Bmps, Ihh), and transcription factor genes promoting osteoblast differentiation (e.g., Runx2, Sp7, Smad1, Gli1, Dlx5) (Figure 5B). We further validated the gene array analysis using quantitative real-time PCR (qRT-PCR) analysis, in which PDGF-BB induced upregulation of Runx2 (Figure 5C), Alpl (Figure 5D), and Spp1 (Figure 5E) levels in cerebral microvessels. These results from the ex vivo experiments suggest that PDGF-BB promoted brain microvessel calcification by directly activating the osteogenic differentiation program. We also conducted an analysis of a publicly available scRNA-Seq data set, in which scRNA-Seq was performed to profile the cellular composition and transcriptomes of young and old mouse brains (41). We analyzed and compared transcriptional changes between young and old cell types specifically in 6 cerebral vascular cell types: arachnoid barrier cells (ABCs), ECs, PCs, VSMCs, Hb-VCs, and vascular and leptomeningeal cells (VLMCs). The genes included in our analysis are osteogenic differentiation–associated genes that were identified from our PCR array analysis as being upregulated in response to PDGF-BB. Analysis of the scRNA-Seq data set (GEO GSE129788) showed that many genes related to osteogenic differentiation were upregulated across the vascular network (Figure 5F). In particular, the violin plot showed upregulation of Alpl gene expression in aged versus young murine cerebral vascular cells (log fold change = 0.462, Bonferroni-adjusted P = 1.44 × 10–12; Figure 5G).

PDGFBB activates osteogenic differentiation–associated genes in the cerebral vasculature. (A and B) Osteogenic gene array analysis of cultured brain microvessels treated with or without 20 ng/mL PDGF-BB. (A) Expression levels of 84 osteogenesis-related genes are compared. Expression levels (2–ΔCt) of these genes are plotted on a logarithmic scale. The 2 boundary lines above and below the center partition line indicate the threshold of 1.5-fold upregulation and downregulation between the groups. Genes with at least 1.5-fold higher expression in PDGF-BB–treated vessels compared with vehicle-treated vessels are shown as red dots (the gene names and fold changes are listed in B). (B) Heatmap shows statistically significantly regulated genes (>1.5-fold and P < 0.05) in PDGF-BB–treated versus vehicle-treated microvessels. n = 2. (C–E) qRT-PCR analysis of Runx2, Alpl, and Spp1 mRNA expression levels in isolated mouse brain vessels with different concentrations of PDGF-BB treatment for 16 hours. n = 3. (F) Dot plots of osteogenic gene expression in different subtypes of vascular cells from young and aged mouse brains by analysis of a scRNA-Seq data set (GEO GSE129788) (see detailed information in Methods). (G) Violin plot shows substantially increased expression of Alpl in cerebral vascular cells in the aging mouse brain. *P < 0.05, **P < 0.01, and ***P < 0.001, by ordinary 1-way ANOVA for multiple-group comparisons (C–E).

PDGF-BB activates p-PDGFRβ/p-ERK/RUNX2 signaling in PCs to upregulate osteoblast differentiation genes. Since the blood vessel cells isolated from mouse brain in our ex vivo experiments contained a mixture of vascular cells, we assessed which cell type or types were key players in the process of calcification during aging. We first examined the changes in PDGF-BB/PDGFRβ signaling in brain PCs, which have abundant PDGFRβ expression (44–46). PCs isolated from mouse brain were treated with PDGF-BB at 30 ng/mL (equivalent serum PDGF-BB level in aged mice) for different durations. Western blot analysis showed a persistent downregulation of PDGFRβ in PDGF-BB–treated cells relative to vehicle-treated control cells (Supplemental Figure 4A). Immunocytofluorescence staining demonstrated that PDGFRβ was mostly distributed on the plasma membrane of PCs, with some protein expressed in the nucleus in the vehicle-treated control cells, whereas PDGFRβ expression on the cell membrane was markedly reduced with PDGF-BB treatment (Supplemental Figure 4B), suggesting that the PDGF-BB ligand induced downregulation of cell-surface PDGFRβ. This result is in line with our recent finding that persistent exposure of brain PCs to abnormally high concentrations of PDGF-BB leads to reduced PDGFRβ expression by inducing ectodomain shedding of PDGFRβ from PCs (32). However, despite the lowered cell-surface expression of PDGFRβ, PDGF-BB rapidly induced the activation of PDGFRβ signaling, as evidenced by increased PDGFRβ phosphorylation at Tyr751, along with the activation of downstream phosphorylated ERK (p-ERK) (Figure 6, A and B). p-PDGFRβ and p-ERK were maintained at higher levels in longer-term PDGF-BB–treated cells compared with those without PDGF-BB treatment (Supplemental Figure 4B and Figure 6, C and D). Moreover, PDGF-BB treatment markedly upregulated the further downstream effectors RUNX2 and OPN (Figure 6E). The results suggest that even when cell-surface PDGFRβ was reduced in PCs due to persistent high levels of PDGF-BB during aging, the remaining low level of PDGFRβ was still sufficient to trigger robust activation of the downstream p-ERK/RUNX2 signaling pathway. Increased p-PDGFRβ and OPN expression in PCs was also confirmed by immunofluorescence staining of brain tissue sections from aged mice (Figure 6F) and cytofluorescence staining of isolated mouse brain PCs with PDGF-BB treatment (Figure 6G). Moreover, qRT-PCR results showed greatly increased expression of Runx2 in PDGF-BB–treated PCs relative to vehicle-treated control cells (Figure 6H). Sp7 and Spp1, direct Runx2 downstream targets and key osteoblast differentiation genes, were also upregulated in cells with PDGF-BB treatment. As PDGFRβ is also expressed in astrocytes, we assessed whether osteoblast differentiation genes were stimulated in astrocytes by PDGF-BB treatment at the same concentration used in the PC culture. We observed no differences in mRNA expression of Osx or Bglap in PDGF-BB–treated compared with vehicle-treated astrocytes at various time points (Figure 6I), indicating transdifferentiation of PCs but not astrocytes toward osteoblast-like lineage cells.

PDGF-BB stimulates p-PDGFRβ/p-ERK/RUNX2 signaling in PCs to activate multiple key osteoblast differentiation genes. (A–D) Primary mouse brain PCs were exposed to recombinant mouse PDGF-BB at a concentration of 30 ng/mL for shorter time periods (A and B) and longer time periods (C and D). Expression of the indicated proteins was detected by Western blot analysis (A and C). The relative intensities of the proteins were quantified using ImageJ (B and D). (E) Primary mouse brain PCs were exposed to recombinant mouse PDGF-BB at a concentration of 30 ng/mL for 4 and 8 hours. Western blot analysis of RUNX2 and OPN expression. (F) Double-immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against CD31 and p-PDGFRβ. n = 3. (G) Primary mouse brain PCs were subjected to an 8-hour treatment with recombinant mouse PDGF-BB. Double immunocytochemical staining was performed using antibodies against PDGFRβ and OPN. Boxed areas are shown at a higher magnification in the corresponding panels to the right. n = 3. (H) Primary mouse brain PCs were treated with 30 ng/mL PDGF-BB for 4 and 8 hours. qRT-PCR analysis was conducted to assess the expression levels of osteogenic marker genes, including Spp1, Runx2, and Sp7. n = 6. (I) Primary brain astrocytes were treated with 30 ng/mL PDGF-BB for 4, 8, and 16 hours. qRT-PCR was conducted to assess the expression levels of osteogenic markers, including Sp7 and Bglap. n = 3. Scale bars: 100 μm (F and G). Relative fold-change results are shown as the mean ± SD. *P < 0.05, **P < 0.01, and ****P < 0.0001, by ordinary 1-way ANOVA for multiple-group comparisons (H and I).

PDGF-BB activates the phosphate transporter Slc20a1 in astrocytes. It has been reported that PDGF-BB increases the expression of Slc20a1, a type III sodium-dependent Pi transporter, in aortic SMCs (47, 48). Slc20a1 is distributed in many CNS cell types, including astrocytes, neurons, and ependymocytes (41). We then tested whether PDGF-BB also activates Slc20a1 expression in cerebral microvessels, as PiT1 (Slc20a1-encoded protein) is a key regulator of skeletal mineralization. As we expected, recombinant mouse PDGF-BB dose-dependently increased the mRNA expression of Slc20a1 in the ex vivo–cultured cerebral vessels (Figure 7A). Consistently, Western blot analysis also showed that PiT1 expression was elevated by PDGF-BB treatment in a dose-dependent manner (Figure 7, B and C). The PDGF-BB failed to increase Slc20a2 and Xpr1 expression in our ex vivo microvessel assays (Supplemental Figure 5, A and B). We assessed PiT1 expression in the thalamus of aged male mice. Double-immunofluorescence staining of brain tissue sections demonstrated that PiT1 expression was undetectable in 3-month-old young mice, whereas a PiT1+ signal was detected within the calcified lesions that were associated with CD13+ and CD31+ brain capillaries in 22-month-old mice (Figure 7, D and E). We also tested whether PiT1 is required for PDGF-BB–induced brain vessel calcification by adding phosphonoformic acid (PFA), a specific inhibitor of phosphate transporter, in the PDGF-BB–treated ex vivo cerebral microvessels (49, 50). While PDGF-BB stimulated calcification of the microvessels, adding low (1 mM) and high (3 mM) dosages of PFA efficiently blocked the effect (Figure 7J). The result suggests that PiT1 mediates PDGF-BB–induced cerebral vascular calcification. We further characterized the specific cell type(s) within the cerebral vasculature that are responsible for PDGF-BB–induced Slc20a1 expression. We postulated that PCs and/or astrocytes might be the targets, as both cell types have abundant PDGFRβ expression. Our results showed that PDGF-BB stimulated the upregulation of Slc20a1 gene expression in astrocytes in a dose- and time-dependent manner (Figure 7, F and G). Consistently, PDGF-BB at high concentration induced upregulation of PiT1 protein expression in astrocytes as detected by Western blot analysis (Figure 7H), whereas it failed to exert the same effect on PCs (Figure 7I). Furthermore, GFAP+ astrocytes accumulated surrounding the OPN+ calcified nodules (Figure 7K). Importantly, while a PiT1+ signal was not detected in the thalamic astrocytes of WT control mice, many GFAP+ astrocytes exhibited PiT1 expression in PdgfbcTG mice (Figure 7L), providing in vivo evidence for the involvement of astrocytes in the mineralization process of brain calcification.

PDGF-BB upregulates phosphate transporters PiT1 (Slc20a1) in astrocytes. (A) qRT-PCR analysis of Slc20a1 mRNA expression in isolated mouse brain microvessels with different dosages of PDGFBB treatment for 16 hours. n = 4. (B) Western blot analysis of PiT1 protein expression in isolated mouse brain microvessels with different dosages of PDGF-BB treatment for 16 hours. n = 3. (C) The relative density of PiT1 was calculated using ImageJ. n = 3. (D) Double-immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against CD13 and PiT1. n = 3. (E) Double-immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against CD31 and PiT1. n = 3. (F and G) qRT-PCR analysis of Slc20a1 mRNA expression in primary brain astrocytes treated with PDGF-BB at different dosages (F) and for different durations (G). n = 3. (H) Western blot analysis of PiT1 protein expression in brain astrocytes treated with PDGF-BB at different dosages. (I) Western blot analysis of PiT1 protein expression in brain PCs treated with 50 ng/mL PDGF-BB at different time points. (J) Isolated brain microvessels were incubated under different culture conditions indicated for 16 hours with 2.6 mM Pi, 2.7 mM Ca, and 20 ng/mL PDGF-BB. The concentrations of the Pi transporter inhibitor PFA were 1 mM and 3 mM. Calcification was detected by Alizarin red staining. (K) Double-immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against GFAP and OPN. n = 3. (L) Double-immunofluorescence staining of frozen brain tissue sections from WT and PdgfbcTG mice using antibodies against GFAP and PiT1. n = 3. Scale bars: 100 μm (J–L); 200 μm (D and E). Relative fold-change results are shown as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by ordinary 1-way ANOVA for multiple-group comparisons (A, C, F and G). Magnification value is ×10 for D, E, and K.

We next attempted to gain mechanistic insights into PDGF-BB–induced Slc20a1 expression in cerebral vessels. Runx2, the master organizer in controlling the osteoblast differentiation program (51), was among one of the top genes that differentially changed in PDGF-BB–treated versus vehicle-treated cerebral microvessels in the osteogenesis PCR array (Figure 5, B and C). It is noteworthy that a consensus RUNX2 binding site (TGTGGT) (52) was found in Slc20a1 promoter regions, nearby the transcription start site (Figure 8A). We then used the ChIP assay to test whether RUNX2 could bind to the Slc20a1 promoter in cerebral microvessels in response to PDGF-BB treatment. Chromatin from the PDGF-BB– and vehicle-treated cerebral microvessels was immunoprecipitated using a specific antibody against RUNX2. Genomic DNA fragments bound to RUNX2 were analyzed by PCR using random primers designed to include the putative RUNX2 binding sites in the Slc20a1 promoter region (Figure 8A). No genomic DNA was pulled out in the immunoprecipitates from normal mouse IgG (Figure 8B, lane 1). As a positive control, RNA PolII antibody efficiently pulled out DNA of GAPDH (Figure 8B, lane 3). Importantly, there was only a weak binding of RUNX2 at the Slc20a1 promoter in vehicle-treated cerebral vessels (Figure 8B, lane 2 in upper panel), and the binding was dramatically enhanced in the PDGF-BB treatment group (Figure 8B, lane 2 in lower panel). Therefore, RUNX2 indeed bound the Slc20a1 promoter region for its gene transcription. Accumulating evidence demonstrated that RUNX2 is a target of the kinases MAPK, ERK, and AKT, which induce RUNX2 protein phosphorylation to enhance its activity (53–58). Indeed, we observed that PDGF-BB induced upregulation of p-RUNX2 (Figure 8C, upper panel, and Figure 8D). Consistently, immunofluorescence staining of brain tissue sections showed strong p-RUNX2+ signal at the calcified lesions in the thalamus of aged mice but not in that of young mice (Figure 8E). It has been recognized that in VSMCs, PDGF-BB can activate ERK MAPKs (59, 60) and PI3K/Akt signaling (61, 62), which may induce RUNX2 phosphorylation. We then tested this assumption in the ex vivo cerebral vessel culture system. Indeed, we found that recombinant PDGF-BB treatment greatly enhanced the phosphorylation of ERK in a dose-dependent manner compared with the vehicle-treated group, whereas the phosphorylation of Akt was not stimulated by PDGFF-BB (Figure 8, F and G). Taken together, these findings indicate that elevated PDGF-BB in a pathological context bound to its receptor in vascular cells, leading to a robust activation of the p-PDGFRβ/p-ERK/RUNX2 signaling cascade. As a transcription factor, RUNX2 played a dual role. First, it activated the transcription of downstream osteogenic genes in PCs, promoting their transdifferentiation into osteoblast-like cells. Second, it directly stimulated gene transcription of the phosphate transporter Slc20a1 in astrocytes, thereby leading to an impaired phosphate balance to promote mineralization of the brain vasculature (Figure 8H).

PDGF-BB activates Slc20a1 gene transcription through ERK/RUNX2 signaling. (A) Schematic representation of the Slc20a1 promoter region. The location of the consensus RUNX2 binding site (TGTGGT) and the regions chosen for PCR amplification by the primers in the ChIP-qPCR assays are indicated. (B) Cerebral microvessels isolated from brain were treated with 20 ng/mL PDGF-BB or vehicle. Chromating DNA was immunoprecipitated using a specific antibody against RUNX2 or mouse IgG (negative control). DNA fragments were amplified with primers specific for the Slc20a1 promoter. As a positive control, an antibody against RNA polymerase II (RNA PolII) was used for immunoprecipitation, and primers specific for GAPDH were used for PCR. (C) Cerebral microvessels isolated from brain were treated with 20 ng/mL PDGF-BB or vehicle. Western blot analysis of p-RUNX2 and total RUNX2 (t-RUNX2). n = 3. (D) The relative density of p-RUNX2 to t-RUNX2 was calculated using ImageJ. n = 3. (E) Immunofluorescence staining of frozen brain tissue sections from 3- and 22-month-old male mice using antibodies against RUNX2. n = 3. Scale bar: 100 μm. Magnification ×10. (F) Cerebral microvessels isolated from brain were treated with increasing concentrations of PDGF-BB. Western blot analysis of p-ERK and p-AKT. n = 3. (G) Relative densities of p-ERK and t-ERK were calculated using ImageJ (n = 3). (H) Schematic model showing the molecular mechanisms underlying PDGF-BB–induced brain vascular calcification. All data are shown as the mean ± SD. *P < 0.05 and **P < 0.01, by ordinary 1-way ANOVA for multiple-group comparisons (G). *P < 0.05, by unpaired, 2-tailed Student’s t test for 2-group comparisons (D).