KLF4 is a key determinant in the development and progression of cerebral cavernous malformations

Abstract

Cerebral cavernous malformations (CCMs) are vascular malformations located within the central nervous system often resulting in cerebral hemorrhage. Pharmacological treatment is needed, since current therapy is limited to neurosurgery. Familial CCM is caused by loss-of-function mutations in any of Ccm1, Ccm2, and Ccm3 genes. CCM cavernomas are lined by endothelial cells (ECs) undergoing endothelial-to-mesenchymal transition (EndMT). This switch in phenotype is due to the activation of the transforming growth factor beta/bone morphogenetic protein (TGFβ/BMP) signaling. However, the mechanism linking Ccm gene inactivation and TGFβ/BMP-dependent EndMT remains undefined. Here, we report that Ccm1 ablation leads to the activation of a MEKK3-MEK5-ERK5-MEF2 signaling axis that induces a strong increase in Kruppel-like factor 4 (KLF4) in ECs in vivo. KLF4 transcriptional activity is responsible for the EndMT occurring in CCM1-null ECs. KLF4 promotes TGFβ/BMP signaling through the production of BMP6. Importantly, in endothelial-specific Ccm1 and Klf4 double knockout mice, we observe a strong reduction in the development of CCM and mouse mortality. Our data unveil KLF4 as a therapeutic target for CCM.

Keywords: CCM; EndMT; KLF4; TGFβ‐BMP; endothelial cells.

© 2015 The Authors. Published under the terms of the CC BY 4.0 license.

Figures

Figure 1. KLF4 is determinant for CCM development in vivo

1. Representative images of WT, iCCM1, and iCCM1/KLF4 mouse brains at P12 (n = 5 for each genotype). Scale bar: 500 μm.

2. Representative confocal analysis of VE‐CADHERIN (green) and KLF4 (red) in vascular lesions (dotted area) and pseudo‐normal cerebellar vessels (white arrowheads) of WT, iCCM1, and iCCM1/KLF4 mice (n = 4 in each group). VE‐CADHERIN identifies ECs; DAPI visualizes nuclei. Scale bar: 50 μm.

3. qRT–PCR of both Ccm1 and Klf4 expression performed on freshly isolated brain ECs derived by WT, iCCM1, and iCCM1/KLF4 mice at P12. Fold difference in gene expression is relative to WT mice. Data are mean ± SD (n = 4–5/group). A two‐tailed unpaired t‐test was performed. Ccm1: ***P = 0.0004; Klf4: ***P = 0.0003, ****P = 1.43E‐05.

4. Quantification of number and size of vascular lesions in the cerebellum of iCCM1 and iCCM1/KLF4 mice at P12. Columns represent means ± SD (n = 3 for each genotype from 2 litters). A two‐tailed unpaired t‐test was performed comparing iCCM1 versus iCCM1/KLF4. < 5,000 μm2: **P = 0.0018; 5–1,000 μm2: *P = 0.0117; > 10,000 μm2: **P = 0.0019.

5. Kaplan–Meier survival curve of WT, iCCM1, and iCCM1/KLF4 mice (n = 8 for each group). Mantel–Cox statistical test was performed: ***P < 0.0001.

Figure EV1. Klf4 is critical for cavernoma development and progression in the retina

1. A

   Isolectin B4 staining (IB4, used to identify vasculature) on WT, iCCM1 and iCCM1/KLF4 retinae at P12. Dotted area highlights macroscopic differences in the extension of CCM lesion area between iCCM1 and iCCM1/KLF4 mice. Images are representative of five mice for each genotype. Scale bar: 500 μm.

2. B

   Representative immunostaining (one out of three performed, n = 3 in each group) for KLF4 (light blue) in the retinae of WT, iCCM1, and iCCM1/KLF4 mice. Vasculature at the periphery of the retina is shown after isolectin B4 staining (red). Scale bar: 60 μm.

3. C–E

   Quantification of percentage of retinal area covered by vascular lesions (C), vascular front density at the leading edge of the plexus (D), and average distance covered by the growing vessels measured as vascular progression (E) in retinae from WT, iCCM1, and iCCM1/KLF4 mice at P12. Data are mean ± SD (n = 5 for each genotype from three different litters). A two‐tailed unpaired t‐test was performed. ***P = 0.0003, **P = 0.004.

Figure EV2. KLF4 amount is increased upon loss of Ccm1, Ccm2, and Ccm3

1. qRT–PCR analysis of Klf4 during disease progression at different times (P3, P5, and P9) after tamoxifen‐induced Ccm1 recombination (P1) in freshly isolated brain ECs derived from WT and iCCM1 mice. Data are mean ± SD (n = 3 in each group). Fold changes are relative to WT animals. A two‐tailed unpaired t‐test was performed. *P = 0.0143, **P = 0.002, ##P = 0.001.

2. Representative immunostaining of KLF4 (red) in combination with PECAM1 (green, to identify ECs) in brain sections from iCCM2, iCCM3 mice, and their relative WT controls (one out of three performed). Cell nuclei are visualized with DAPI; dotted area highlights lesion area. Scale bars: 50 μm.

Figure EV3. Increased ERK5 phosphorylation is responsible for KLF4 upregulation and KLF4‐dependent EndMT in the absence of both CCM2 and CCM3

1. qRT–PCR analysis of Klf4 in WT ECs transfected with either siRNA directed to anyone of the three Ccm genes or control siRNA. Data are presented as mean ± SD (n = 3). Results are shown as fold changes relative to control siRNA‐treated ECs. A two‐tailed unpaired t‐test was performed. ***P = 0.0001, ###P = 0.0009, **P = 0.0017.

2. WB analysis of pERK5, ERK5, and KLF4 in lung‐derived WT and CCM2 KO ECs treated with XMD8‐92 or vehicle for 72 h. Both pERK5/ERK5 ratio and KLF4 amount normalized over vinculin, the loading control, were quantified by densitometry scan. WB results are representative of three independent observations.

3. qRT–PCR of Ccm2, Klf4, Bmp6, and some EndMT markers in WT and CCM2 KO ECs treated with XMD8‐92 or vehicle for 72 h. qRT–PCR results are shown as mean ± SD (n = 3), and fold changes are relative to vehicle‐treated WT ECs. A two‐tailed unpaired t‐test was performed. Ccm2: ***P = 0.0003; Klf4: ***P = 0.0002, **P = 0.006; Bmp6: **P = 0.0018, ****P = 8.7E‐05; Fsp1: **P = 0.0013, ****P = 6.5E‐05, ***P = 0.0008; Sca1: **P = 0.001, ##P = 0.0025, *P = 0.029; Id1: **P = 0.0013, ##P = 0.0029.

4. WB analysis of pERK5, ERK5, and KLF4 in lung‐derived WT and CCM3 KO ECs treated with XMD8‐92 or vehicle for 72 h. Both pERK5/ERK5 ratio and KLF4 were normalized and quantified as in (A).

5. qRT–PCR of Ccm3, Klf4, Bmp6, and some EndMT markers in WT and CCM3 KO ECs treated with XMD8‐92 or vehicle for 72 h quantified as in (C). Ccm3: ****P < 0.00001; Klf4: ****P < 0.00001; Bmp6: ****P < 0.00001, ***P = 0.0009, *P = 0.01; Fsp1: ***P = 0.0002, ****P < 0.00001, ####P = 3.8E‐05; Sca1: ****P < 0.00001, ####P = 4.79E‐05, **P = 0.0013; Id1: **P = 0.0039, ##P = 0.0017.

Source data are available online for this figure.

Figure 2. KLF4 is increased in a human brain cell line upon loss of CCM1 and in cerebral lesions of CCM1 patients

1. Left panel: Ccm1 and Klf4 relative mRNA levels in CCM1 (siCCM1) and control (siCTRL) siRNA‐treated hCMEC/D3. The result is shown as fold changes in gene expression in siCCM1‐treated versus control. Data are presented as mean ± SD (n = 3). A two‐tailed unpaired t‐test was performed. ****P < 0.00001, ####P = 1.2E‐05. Right panel: WB analysis of KLF4 amount in siCCM1‐ and siCTRL‐treated hCMEC/D3. Tubulin is used as a loading control. These data are representative of three independent observations.

2. Immunohistochemical analysis performed on serial sections of brain tissue derived from a familial CCM1 patient. Hematoxylin and eosin (left panels), CLAUDIN5 (central panels), and KLF4 (right panels) stainings were performed in brain lesion and normal peri‐lesion vessels as control brain tissue. Higher magnification images of KLF4 staining of the boxed regions are shown. CLAUDIN5 identifies ECs and arrowheads mark endothelial KLF4‐positive nuclei. Scale bar: 100 μm. These data are representative of three independent observations.

Source data are available online for this figure.

Figure 3. KLF4 regulates the EndMT switch in CCM1 KO ECs

1. A–D

   Cultured lung‐derived WT and CCM1 KO ECs were lentivirally transduced with shRNA directed to either Klf4 (shKLF4) or control sequence (shCTRL). (A) qRT–PCR of mesenchymal (Fsp1, Id1), stem cell‐like (Sca1), and endothelial markers (VE‐cadherin and Claudin5) in WT shCTRL, WT shKLF4, CCM1 KO shCTRL, and CCM1 KO shKLF4 ECs. Data are mean ± SD (n = 3). Fold difference in gene expression is relative to WT shCTRL ECs. A two‐tailed unpaired t‐test was performed. Klf4: ****P < 0.00001, ***P = 0.0008, **P = 0.007; Fsp1: ***P = 0.0001, ****P = 3.9E‐05, ###P = 0.0002; Sca1: ***P = 0.0001, ****P < 0.00001; Id1: ***P = 0.0001; Ve‐cadherin: ***P = 0.0001; Claudin5: ****P = 1.37E‐05. (B) WB of EndMT markers in WT shCTRL, WT shKLF4, CCM1 KO shCTRL, and CCM1 KO shKLF4 ECs. Vinculin is the loading control. These data are representative of three independent observations. (C) Proliferation rate of WT shCTRL, WTshKLF4, CCM1 KO shCTRL, and CCM1 KO shKLF4 ECs cultured for 5 days. Columns represent mean ± SD (n = 8). A two‐tailed unpaired t‐test was performed. 3 days of culture: ***P = 0.0001, **P = 0.0045, ##P = 0.0017; 5 days of culture: ***P = 0.0001, **P = 0.0029, ##P = 0.0040. (D) Migration rate measured in a wound assay of WT shCTRL, CCM1 KO shCTRL, and CCM1 KO shKLF4 ECs. Mean ± SD is shown (n = 6). A two‐tailed unpaired t‐test was performed. ***P = 0.0004, ###P = 0.0009.

2. E

   Cultured lung‐derived WT ECs were lentivirally transduced with a full‐length murine Klf4 (LentiKLF4) or empty vector (Mock). qRT–PCR (left panel) and WB (right panel) of EndMT markers in Mock and LentiKLF4 ECs. qRT–PCR data are mean ± SD (n = 3) and the fold changes are relative to Mock ECs. A two‐tailed unpaired t‐test was performed. **P = 0.002, ***P = 0.0007, ###P = 0.0001. WB results are representative of three independent observations. Tubulin is the loading control.

Source data are available online for this figure.

Figure EV4. ERK5‐KLF4 axis regulates EndMT in primary brain CCM1 KO ECs

1. Upper panel: qRT–PCR of both Ccm1 and Klf4 gene recombination efficiency after in vitro TAT‐Cre‐recombinase treatment of primary brain ECs derived from WT, Ccm1fl/fl, and Ccm1fl/fl/Klf4fl/fl mice to originate cultured WT, CCM1 KO, and CCM1‐KLF4 KO brain ECs, respectively (10 mice for each genotype per experimental replicate). Data are presented as mean ± SD (n = 3). Fold changes are relative to WT brain ECs. A two‐tailed unpaired t‐test was performed. Ccm1: **P = 0.001; Klf4: ****P < 0.00001, ####P = 5.2E‐05. Lower panel: WB of KLF4 in WT, CCM1 KO, and CCM1‐KLF4 KO brain ECs described above. Vinculin is the loading control. These data are representative of three independent observations.

2. Representative confocal analysis (out of three performed) of KLF4 (red) and VE‐CADHERIN (green) in WT, CCM1 KO, and CCM1‐KLF4 KO primary brain ECs obtained as in (A). Scale bar: 30 μm.

3. qRT–PCR of EndMT markers in WT, CCM1 KO, and CCM1‐KLF4 KO brain ECs obtained as in (A). Data are mean ± SD (n = 3). Fold changes are relative to WT ECs. A two‐tailed unpaired t‐test was performed. Fsp1: ***P = 0.0006, **P = 0.0011; Sca1: ****P < 0.00001, ####P = 7.7E‐05; Id1: **P = 0.0017; Slug: *P = 0.02, #P = 0.01; Bmp6: **P = 0.003.

4. WB of FSP1, SCA1, ID1, and SLUG in WT, CCM1 KO, and CCM1‐KLF4 KO brain ECs obtained as described in (A). Tubulin is the loading control. These data are representative of three independent observations.

5. Analysis of pERK5 and ERK5 protein levels in WT, CCM1 KO, and CCM1‐KLF4 KO brain ECs obtained as in (A). pERK5/ERK5 ratio normalized over tubulin is indicated.

6. WB analysis of pERK5, ERK5, KLF4, FSP1, SCA1, ID1 (left panel), and qRT–PCR of Bmp6 (right panel) in WT and CCM1 KO brain ECs, obtained as in (A), and then treated with XMD8‐92 or vehicle for 72 h. qRT–PCR results are shown as mean ± SD (n = 3) and fold changes are relative to vehicle‐treated WT ECs. A two‐tailed unpaired t‐test was performed. **P = 0.001, ***P = 0.0006, ****P = 6.3E‐05. WB results are representative of three independent observations, and tubulin is the loading control.

Source data are available online for this figure.

Figure EV5. EndMT marker expression is reduced in the absence of KLF4 ex vivo and in vivo

1. A

   qRT–PCR of some EndMT markers in freshly isolated brain ECs from WT, iCCM1, and iCCM1/KLF4 mice analyzed at P12. Fold changes are relative to WT animals. Data are mean ± SD from a representative experiment out of three.

2. B–D

   Representative confocal analysis of (B) PECAM1 (green) and FSP1 (red), (C) GLUT1 (green) and SCA1 (red), or (D) PECAM1 (green) and ID1 (red) in normal cerebellar vessels of WT mice and vascular lesions (dotted area) of both iCCM1 and iCCM1/KLF4 mice (n = 4 in each group). PECAM1 identifies ECs; DAPI visualizes nuclei. Scale bars: 50 μm.

Figure 4. KLF4 increases Bmp6 expression and SMAD1 phosphorylation in ECs

1. qRT–PCR of Bmp6 expression in WT shCTRL, WT shKLF4, CCM1 KO shCTRL, and CCM1 KO shKLF4‐cultured ECs. Data represent the mean ± SD (n = 3). Fold changes are relative to CCM1 KO shCTRL ECs. A two‐tailed unpaired t‐test was performed. ****P < 0.00001.

2. WB of pSMAD1 and SMAD1 in WT shCTRL, WT shKLF4, CCM1 KO shCTRL, and CCM1 KO shKLF4 ECs. pSMAD1/SMAD1 ratio normalized over tubulin was quantified by densitometry scan. These data are representative of three independent observations.

3. Immunohistochemical analysis of PECAM1 (left panels) and BMP6 (central panels) performed on serial sections of cerebellum derived from WT, iCCM1, and iCCM1/KLF4 mice at P12. Higher magnification images of the boxed regions are shown in the right panels. Black arrowheads mark ECs and dotted area indicates the Purkinje cell layer used as a positive control of the staining. Scale bar: 100 μm. These data are representative of three independent observations (n = 3 for each genotype).

4. ChIP analysis of KLF4 binding to Bmp6 promoter. Putative KLF4 binding sites identified by MatInspector are indicated. The levels of DNA are normalized to input. Columns are mean ± SD of triplicates from a representative experiment out of three.

5. Transcriptional reporter assay performed in HEK‐293 cells transfected with Bmp6 promoter reporter plasmid together with an empty vector, a full‐length KLF4 or a mutant KLF4 lacking the DNA‐binding zinc finger domains (KLF4 ∆ZnF). Red boxes in the picture indicate KLF4 binding sites validated by ChIP. Fold change in the Bmp6 promoter activity is relative to empty vector‐transfected cells. Data are mean ± SD (n = 3). A two‐tailed unpaired t‐test was performed. ****P < 0.00001, ####P = 5.8E‐05.

Source data are available online for this figure.

Figure 5. KLF4 promotes EndMT through BMP6 upregulation

1. A

   qRT–PCR of Bmp6 in Mock and LentiKLF4‐cultured ECs. Data are mean ± SD (n = 3). Fold changes are relative to Mock ECs. A two‐tailed unpaired t‐test was performed. **P = 0.0034.

2. B

   WB of pSMAD1 and SMAD1 in Mock and LentiKLF4 ECs. Vinculin is the loading control. pSMAD1/SMAD1 ratio normalized over vinculin was quantified by densitometry scan. These data are representative of three independent observations.

3. C

   ChIP of KLF4 interaction with Bmp6 promoter in Mock and LentiKLF4 ECs. Putative KLF4 binding sites identified by MatInspector are indicated. The levels of DNA are normalized to input. Columns are mean ± SD of triplicates from a representative experiment out of three.

4. D–F

   Mock and LentiKLF4‐cultured ECs were lentivi rally transduced with shRNA directed to either Bmp6 (shBMP6) or control sequence (shCTRL). (D) qRT–PCR analysis of Bmp6 in Mock shCTRL, Mock shBMP6, LentiKLF4 shCTRL, and LentiKLF4 shBMP6 ECs. qRT–PCR data represent the mean ± SD (n = 3) and the fold changes are relative to Mock shCTRL ECs. A two‐tailed unpaired t‐test was performed. ***P = 0.0008, **P = 0.004, ##P = 0.003. WB of pSMAD1 and SMAD1 (E) or EndMT markers (F) in Mock shCTRL, Mock shBMP6, LentiKLF4 shCTRL, and LentiKLF4 shBMP6‐cultured ECs. pSMAD1/SMAD1 ratio was quantified as in (B). These WB data are representative of three independent observations. Tubulin is the loading control. The fold changes of FSP1, SCA1, and ID1 normalized over tubulin were quantified by densitometry scan. These WB data are representative of three independent observations.

Source data are available online for this figure.

Figure 6. KLF4 directly regulates some EndMT marker expression

1. A

   qRT–PCR (left panel) and WB (right panel) analyses of KLF4 in cultured lung WT and KLF4 KO ECs. qRT–PCR data represent the mean ± SD (n = 3) and the fold changes are relative to WT ECs. A two‐tailed unpaired t‐test was performed. ***P = 0.0006. GAPDH is the loading control in WB. These WB data are representative of three independent observations.

2. B

   WB of pSMAD1 and SMAD1 in WT and KLF4 KO ECs left untreated or treated with recombinant BMP6 for 4 h. pSMAD1/SMAD1 ratio normalized over GAPDH was quantified by densitometry scan. These WB data are representative of three independent observations.

3. C

   qRT–PCR of Fsp1, Sca1, and Id1 in WT and KLF4 KO ECs stimulated with BMP6 for 96 h. qRT–PCR data are mean ± SD (n = 3). Fold changes in gene expression in BMP6‐treated versus untreated ECs. A two‐tailed unpaired t‐test was performed. **P = 0.0032, ***P = 0.0007.

4. D–F

   ChIP analysis of KLF4 binding to the promoters of Fsp1 (D), Sca1 (E), and Id1 (F) in WT and CCM1 KO cultured ECs. The positions of the putative KLF4 binding sites identified with MatInspector in the promoters of the genes analyzed are indicated. The levels of DNA are normalized to input. Columns are mean ± SD of triplicates from a representative experiment out of three performed.

5. G, H

   Transcriptional reporter assays performed in HEK‐293 cells transfected with either Fsp1 (G) or Sca1 (H) promoter reporter plasmid together with an empty vector, KLF4 or KLF4 ∆ZnF. In the schematics, red boxes indicate KLF4 binding sites validated by ChIP, while grey boxes indicate KLF4 binding sites not validated by ChIP. Fold change in the promoter activity is relative to empty vector‐transfected cells. Data are mean ± SD (n = 3). A two‐tailed unpaired t‐test was performed. ****P < 0.00001.

Source data are available online for this figure.

Figure 7. In the absence of CCM1 increased ERK5 phosphorylation is responsible for KLF4 upregulation and KLF4‐dependent EndMT

1. A

   WB of phoshorylated ERK5 (pERK5) and total ERK5 in freshly isolated brain ECs from WT (n = 2) and iCCM1 (n = 4) mice from two different litters at P12. VE‐PTP measures the endothelial content and vinculin is the loading control. pERK5/ERK5 ratio normalized over vinculin and VE‐PTP was quantified by densitometry scan.

2. B

   WB of pERK5, ERK5, and KLF4 in cultured lung‐derived WT ECs either Ccm1 (siCCM1) or control (siCTRL) siRNA‐treated alone or in combination with two siRNA targeting Erk5 (siERK5#1,2). pERK5/ERK5 ratio normalized over vinculin is indicated. These data are representative of three independent observations.

3. C–E

   Treatment of cultured lung‐derived WT and CCM1 KO ECs with either 5 μM of XMD8‐92 or the vehicle (72 h). (C) WB analysis of pERK5, ERK5, and KLF4. pERK5/ERK5 ratio was quantified as in (B). (D) Left panel: evaluation of pSMAD1 and SMAD1 protein levels. pSMAD1/SMAD1 ratio normalized over vinculin is shown. These data are representative of three independent observations. Right panel: qRT–PCR of Bmp6. The data represent the mean ± SD (n = 3). Fold changes are relative to vehicle‐treated CCM1 KO ECs. A two‐tailed unpaired t‐test was performed. **P = 0.001, *P = 0.01. (E) Representative WB of EndMT markers (one out of three performed). Tubulin is the loading control.

4. F

   qRT–PCR analysis of Klf4, Mef2A, Mef2C, and Mef2D (left panels) and WB of KLF4 (right panels) in cultured WT and CCM1 KO ECs treated with either control siRNA (siCTRL) or siRNA directed to any of Mef2A (upper panels), Mef2C (central panels), or Mef2D (lower panels). qRT–PCR data represent the mean ± SD (n = 3) and the fold changes are relative to WT siCTRL ECs. A two‐tailed unpaired t‐test was performed. Upper panels: ***P = 0.0002, ###P = 0.0005, ****P = 4.09E‐05, **P = 0.001; central panels: ***P = 0.0005 (WT siCRTL vs WT siMEF2C), ***P = 0.0006 (KO siCRTL vs KO siMEF2C), ###P = 0.0001, ****P < 0.00001, **P = 0.007; lower panels: ****P = 5.6E‐05 (WT siCRTL vs WT siMEF2D), ****P = 1.53E‐05 (KO siCRTL vs KO siMEF2D), ####P = 3.1E‐05. For WB, vinculin is used as a loading control and data are representative of three independent observations.

Source data are available online for this figure.

Figure 8. MEKK3‐MEK5‐dependent ERK5 phosphorylation is responsible for KLF4 upregulation

1. Left panel: WB analysis of pERK5, ERK5, KLF4, and MEKK3 in WT and CCM1 KO ECs treated with either two different siRNAs against Mekk3 (siMEKK3 #1 and siMEKK3 #2) or a control sequence (siCTRL). Both pERK5/ERK5 ratio and KLF4 amount normalized over vinculin were quantified by densitometry scan. WB results are representative of three independent observations. Right panel: qRT–PCR of Klf4 in WT and CCM1 KO ECs treated with two different siRNAs directed to Mekk3 (siMEKK3 #1 and siMEKK3 #2) or a control sequence (siCTRL). qRT–PCR results are shown as mean ± SD (n = 3), and fold changes are relative to siCTRL‐treated WT ECs. A two‐tailed unpaired t‐test was performed. ***P = 0.0002, **P = 0.002, ****P = 2.5E‐05, ####P = 3.1E‐05, ###P = 0.0007.

2. WB analysis of pERK5, ERK5, KLF4, and MEK5 (left panel) and qRT–PCR of Klf4 (right panel) in WT and CCM1 KO ECs treated with a siRNA against MEK5 or a control sequence (siCTRL) quantified as in (A). **P = 0.0035, ****P < 0.00001, ####P = 3.5E‐05.

3. WB analysis of pERK5, ERK5, and KLF4 (upper panel) and qRT–PCR of Klf4 (lower panel) in WT and CCM1 KO ECs treated with BIX‐02189 or vehicle for 48 h quantified as in (A). Fold changes are relative to vehicle‐treated WT ECs. A two‐tailed unpaired t‐test was performed. ***P = 0.0002, ****P = 2.1E‐05.

4. Schematic model of KLF4 activity and regulation during CCM pathogenesis.

Source data are available online for this figure.