Abstract

   Beyond dyslipidemia, inflammation contributes to the development of
   atherosclerosis. However, intrinsic factors that counteract vascular
   inflammation and atherosclerosis remain scarce. Here we identify
   insulin-like growth factor binding protein 6 (IGFBP6) as a
   homeostasis-associated molecule that restrains endothelial inflammation
   and atherosclerosis. IGFBP6 levels are significantly reduced in human
   atherosclerotic arteries and patient serum. Reduction of IGFBP6 in
   human endothelial cells by siRNA increases inflammatory molecule
   expression and monocyte adhesion. Conversely, pro-inflammatory effects
   mediated by disturbed flow (DF) and tumor necrosis factor (TNF) are
   reversed by IGFBP6 overexpression. Mechanistic investigations further
   reveal that IGFBP6 executes anti-inflammatory effects directly through
   the major vault protein (MVP)–c-Jun N-terminal kinase (JNK)/nuclear
   factor kappa B (NF-κB) signaling axis. Finally, IGFBP6-deficient mice
   show aggravated diet- and DF-induced atherosclerosis, whereas
   endothelial-cell-specific IGFBP6-overexpressing mice protect against
   atherosclerosis. Based on these findings, we propose that reduction of
   endothelial IGFBP6 is a predisposing factor in vascular inflammation
   and atherosclerosis, which can be therapeutically targeted.
     __________________________________________________________________

   Atherosclerosis is the hallmark pathology of coronary artery disease
   (CAD) and has long been a significant cause of death globally^[73]1.
   Atherosclerotic plaques typically develop in areas of disturbed flow
   (DF), such as arterial branches or vascular bends. DF alters the
   morphology and cytoskeleton of endothelial cells (ECs) through altered
   biochemical signaling and gene expression, ultimately leading to
   endothelial dysfunction. Conversely, unidirectional laminar flow (UF)
   usually occurs in the straight part of the blood vessel, and vessels
   exposed to this area are protected against atherosclerosis^[74]2–[75]9.

   DF instigates endothelial dysfunction partially through inducing
   inflammation in ECs^[76]10. Inflammation leads to endothelial injury,
   which, in turn, promotes the expression of a variety of adhesion
   molecules, fueling the adhesion and migration of circulating leukocytes
   to the activated endothelium and ultimately leading to plaque
   formation^[77]11,[78]12. Therefore, targeting inflammation has been
   recognized as a necessary means to combat atherosclerosis^[79]13. More
   recently, in a retrospective analysis of 31,245 patients receiving
   statin therapy, residual inflammation risk as assessed by
   high-sensitivity C-reactive protein (hs-CRP) was a stronger predictor
   of risk for future cardiovascular events and death than residual
   cholesterol risk as assessed by low-density lipoprotein cholesterol
   (LDL-C)^[80]14,[81]15. In addition, low-dose colchicine reduces the
   incidence of major adverse cardiovascular events (MACE) in patients
   with cardiovascular disease^[82]16. Likewise, monoclonal antibodies
   targeting the inflammatory cytokines interleukin (IL)-6 and IL-1β (by
   ziltivekimab and canakinumab, respectively) have shown promising
   anti-atherosclerotic effects in clinical trials^[83]17,[84]18.
   Therefore, the identification of novel anti-inflammatory therapies
   holds promise for reducing residual inflammation and cardiovascular
   risk.

   To identify novel anti-inflammatory targets in ECs, we selected RNA
   sequencing (RNA-seq) datasets from human umbilical vein endothelial
   cells (HUVECs) exposed to atherosclerotic protective therapy (statins
   and UF) as well as expression datasets from patients with CAD with
   unstable plaques. We discovered that insulin-like growth factor binding
   protein 6 (IGFBP6) may serve as a potential vascular
   homeostasis-associated molecule. Previous studies demonstrated that
   IGFBP6 plays a crucial role in regulating the cell cycle, cell
   proliferation and migration^[85]19–[86]21. Also, compared to stable
   plaques from patients with CAD, the expression level of IGFBP6 in
   unstable plaques was decreased^[87]22. However, the precise role and
   mechanism of IGFBP6 in endothelial homeostasis and atherosclerosis
   remain unclear.

   Here we demonstrate that global or EC-specific deficiency of Igfbp6
   accelerates the progression of atherosclerosis in mice, whereas
   EC-specific overexpression of Igfbp6 ameliorates atherosclerosis.
   Mechanistically, IGFBP6 inhibits endothelial cell inflammation and
   monocyte adhesion through the major vault protein (MVP)–c-Jun
   N-terminal kinase (JNK)/nuclear factor kappa B (NF-κB) pathway. Of
   translational impact, the expression of IGFBP6 is decreased in serum
   and atherosclerotic lesions of patients with CAD. Overall, these data
   suggest that IGFBP6 is mechanoresponsive, anti-inflammatory and
   atheroprotective.

Results

IGFBP6 is decreased in human atherosclerotic plaques and regulated by blood
flow

   To identify novel regulators of endothelial homeostasis, we treated
   HUVECs with atorvastatin for transcriptomic sequencing ([88]GSE176531)
   and selected differentially expressed genes (DEGs) with P < 0.05 and
   log[2]fold change (FC) > 1.5. In addition, DEGs in sequencing datasets
   from UF-treated HUVECs ([89]GSE20739 and [90]GSE87534) and CAD patient
   samples ([91]GSE163154 and [92]GSE41571) were compared. The combined
   analysis of five datasets yielded a total of three genes: ADAM
   metallopeptidase with thrombospondin type 1 motif 1 (ADAMTS1), cellular
   communication network factor 3 (CCN3, also known as NOV) and IGFBP6.
   ADAMTS1 and NOV were reported to be involved in endothelial
   inflammation and atherosclerosis^[93]23,[94]24. However, the role and
   mechanism of IGFBP6 in endothelial homeostasis and atherosclerosis
   remain obscure. Thus, in the present study, we focused on the role of
   IGFBP6 as a potential regulator of endothelial homeostasis ([95]Fig.
   1a).

Fig. 1 |. IGFBP6 is decreased in human atherosclerotic plaques and regulated
by blood flow.

   Fig. 1 |
   [96]Open in a new tab

   a, Identification of IGFBP6 as a novel endothelial
   homeostasis-associated molecule. Statin-upregulated genes
   ([97]GSE176531) and UF-upregulated genes ([98]GSE20739 and
   [99]GSE87534) were overlapped with downregulated genes in unstable
   atherosclerotic plaques ([100]GSE163154 and [101]GSE41571). b, The
   expression of IGFBP6 in macroscopically intact tissues versus atheroma
   plaques, stable plaques versus unstable plaques in public datasets from
   patients ([102]GSE43292 (n = 32) or [103]GSE163154 (stable plaques: n =
   16; unstable plaques: n = 27)). c, Expression of IGFBP6 in tissues and
   organs of patients with CAD (n = 600) and healthy individuals (n = 250)
   in the STARNET database. AOR, atherosclerotic aortic wall; LIV, liver;
   SF, subcutaneous fat; SKLM, skeletal muscle; VAF, visceral fat. d,
   Representative images of immunofluorescence staining of IGFBP6 (red),
   VE-cadherin (green) and DAPI (blue). The expression of IGFBP6 was
   downregulated in the endothelium of human aortic atherosclerotic
   lesions compared to normal controls (n = 3). Scale bar, 25 μm. e, Serum
   samples were collected from normal controls and patients with CAD to
   verify the IGFBP6 content by ELISA (control: n = 12; CAD: n = 19). f,
   IGFBP6 mRNA expression was downregulated comparing DF to UF in the
   public microarray in HUVECs (n = 3, [104]GSE20739). g, HUVECs were
   subjected to DF or UF for 24 h. KLF2, KLF4 and IGFBP6 mRNA expression
   were downregulated in DF-treated HUVECs (n = 3). h, Schematic
   representation of a mouse model with PCL. i, Igfbp6 was decreased in
   the intimal RNA of mouse carotid arteries 4 d after PCL surgery (n =
   5). j, Immunofluorescence staining for IGFBP6 (red) and DAPI (blue)
   showed that IGFBP6 was reduced in the carotid arteries of mice at 4
   weeks after PCL surgery (n = 5). Green: auto-fluorescence of the
   elastic lamina (EL). Scale bar, 25 μm. k, Immunoblotting showing that
   the IGFBP6 was lower in the inner curvature of the AA than in the TA in
   C57BL/6J mouse aortas (n = 6). l, En face immunofluorescence staining
   for IGFBP6 (red), VE-cadherin (green) and DAPI (blue) in C57BL/6J mouse
   aortas, showing decreased IGFBP6 expression in the AA compared to the
   TA (n = 10). Scale bar, 50 μm. Data are shown as mean ± s.d.
   Statistical analysis was performed by two-tailed Student’s t-test for
   b, e, k and l, by Welch’s t-test for c and f and by paired t-test for g
   and i.

   By mining two additional and independent published datasets
   ([105]GSE43292 and [106]GSE163154), we observed that IGFBP6 mRNA
   expression was reduced in carotid artery plaques compared to the
   adjacent non-plaque tissues. Additionally, unstable plaques had lower
   levels of IGFBP6 mRNA than stable plaques ([107]Fig. 1b). We
   subsequently validated IGFBP6 expression in tissues and organs of
   patients with CAD (n = 600) and healthy individuals (n = 250) from the
   Stockholm-Tartu Atherosclerosis Reverse Network Engineering Task
   (STARNET) database^[108]25–[109]27. The results showed that IGFBP6
   expression was significantly lower in atherosclerotic aortic tissues of
   patients with CAD compared to healthy controls ([110]Fig. 1c). To
   extend the translational significance of our study in human patients,
   we analyzed the arterial expression of IGFBP6 in control subjects and
   patients with CAD. In comparison to healthy coronary arteries, the
   expression level of IGFBP6 in the intima of human coronary arteries
   with atherosclerotic plaques was decreased ([111]Fig. 1d). Furthermore,
   the level of IGFBP6 was reduced in the serum of patients with
   atherosclerotic plaques compared to that in healthy individuals
   ([112]Fig. 1e).

   Because the expression of IGFBP6 is upregulated by UF, we next explored
   the potential role of IGFBP6 in responding to different types of shear
   stress. We first examined its expression in HUVECs by mining microarray
   data of ECs treated with DF and UF ([113]GSE20739). The mRNA expression
   of IGFBP6 was lower in HUVECs treated with DF compared to those treated
   with UF ([114]Fig. 1f). We also validated the downregulation of IGFBP6
   mRNA expression in HUVECs exposed to DF compared to UF using
   quantitative real-time polymerase chain reaction (qRT–PCR) ([115]Fig.
   1g). Because the IGFBP family has seven members, we next examined the
   expression of IGFs and different IGFBPs in UF-treated and DF-treated
   HUVECs. Our results showed significant increases in IGFBP5 and IGFBP6
   ([116]GSE20739) ([117]Extended Data Fig. 1a,[118]b). We also measured
   the mRNA levels of IGFBPs in HUVECs and found that IGFBP6 had a higher
   expression level than other IGFBPs ([119]Extended Data Fig. 1c).
   Because our previous study reported that IGFBP5 is mechanoresponsive
   and anti-inflammatory^[120]28, in the present study our goal was to
   explore the potential role of IGFBP6 in regulating endothelial function
   and atherosclerosis.

   To verify blood flow-dependent regulation of IGFBP6 expression in vivo,
   we obtained intimal RNA from carotid arteries of mice after partial
   carotid ligation (PCL) surgery for 4 d ([121]Fig. 1h). qRT–PCR analysis
   showed that the expression of Igfbp6 mRNA in the endothelium of the
   left carotid artery (LCA; exposed to DF) was decreased compared to LCA
   from sham group after 4 days of PCL ([122]Fig. 1i). Consistently,
   immunofluorescence staining demonstrated a reduction in IGFBP6 protein
   expression after PCL ([123]Fig. 1j). IGFBP6 protein expression was also
   quantified in different regions of the aorta, specifically the inner
   curvature of aortic arch (AA; region of DF) and thoracic aorta (TA;
   region of UF). We found that IGFBP6 protein expression was lower in the
   AA than in the TA ([124]Fig. 1k). En face staining of the mouse aortic
   endothelium also showed lower expression of IGFBP6 in the AA than in
   the TA ([125]Fig. 1l).

   Collectively, these results suggest a potential role of IGFBP6 as a
   flow-responsive gene that may execute important biological functions in
   vascular homeostasis and atheroprotection.

IGFBP6 deficiency increases atherosclerotic plaque formation in mice
receiving PCL surgery

   To validate whether IGFBP6 suppresses atherosclerosis in vivo, we first
   constructed Igfbp6 global knockout mice (Igfbp6^−/−) by deleting the
   exon 2 of Igfbp6. We confirmed Igfbp6 knockout by genotyping
   ([126]Extended Data Fig. 2a) and qRT–PCR. Decreased Igfbp6 mRNA
   expression was observed in tested mouse tissues and organs (adipose
   tissue, liver, heart, lung, colon and aorta) of Igfbp6^−/− mice
   compared to wild-type (WT) mice ([127]Fig. 2a). Male 8-week-old
   Igfbp6^−/− mice and their WT littermate controls were subsequently
   infected with AAV8-Pcsk9^D377Y (to reduce LDLR levels in mouse livers
   and induce atherosclerosis; [128]Extended Data Fig. 2b) via tail vein
   injection concurrent with feeding with a high-cholesterol diet (HCD)
   for 2 weeks. Afterwards, we performed PCL surgery on the mice and
   continued HCD feeding for 3 weeks ([129]Fig. 2b).

Fig. 2 |. IGFBP6 deficiency increases atherosclerosis in mice receiving PCL
surgery.

   Fig. 2 |
   [130]Open in a new tab

   a, Igfbp6 transcript level was successfully deleted in various tissues
   and organs in Igfbp6^−/− mice (n = 3). b, Schematic diagram
   illustrating the experimental scheme. c, Mouse serum was used to verify
   IGFBP6 deletion by ELISA in Igfbp6^−/− mice (n = 4). d,
   Immunofluorescence staining for IGFBP6 (red) and DAPI (blue) showed
   that IGFBP6 was reduced in the RCAs of Igfbp6^−/− mice (n = 3). EL
   denotes elastic lamella that display auto-fluorescence (green). Scale
   bar, 25 μm. e, Increased plaque formation in male Igfbp6^−/− mice
   receiving PCL (WT: n = 15; Igfbp6^−/−: n = 13). Scale bar, 2 mm. f,
   Representative Oil Red O–stained carotid arteries of male WT and
   Igfbp6^−/− mice. The panel on the right shows the quantification of the
   Oil Red O–positive plaque area (WT: n = 15; Igfbp6^−/−: n = 10). Scale
   bars, 50 μm. g, Immunofluorescence staining for VCAM-1 (green) and DAPI
   (blue) showed that VCAM-1 was increased in carotid cryosections from
   Igfbp6^−/− mice (n = 6). Scale bars, 100 μm and 25 μm. Data are shown
   as mean ± s.d. Statistical analysis was performed by two-tailed Welch’s
   t-test for a, by Mann–Whitney U-test for c and by Student’s t-test for
   e and f. W, weeks.

   No significant differences were observed between Igfbp6^−/− mice and
   their WT littermates regarding the levels of plasma total cholesterol
   (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C),
   LDL-C, aspartate aminotransferase (AST) or alanine aminotransferase
   (ALT) ([131]Extended Data Fig. 2c). As IGFBP6 is a secreted protein, we
   used ELISA to measure IGFBP6 levels in the serum of Igfbp6^−/− mice.
   IGFBP6 was depleted in the knockout mice ([132]Fig. 2c).
   Immunofluorescence staining also showed that IGFBP6 was drastically
   reduced in the RCA of Igfbp6^−/− mice ([133]Fig. 2d). Most notably,
   compared to WT mice, the area of atherosclerotic plaques in the LCA of
   Igfbp6^−/− mice was increased ([134]Fig. 2e). Oil Red O and
   immunofluorescence staining showed an increase in the area of lipid
   accumulation and vascular cell adhesion molecule-1 (VCAM-1) positivity
   in the endothelium layer of the LCA in Igfbp6^−/− mice ([135]Fig.
   2f,[136]g). These findings indicate that the absence of Igfbp6
   increases atherosclerotic lesions induced by DF in mice.

EC-specific knockout of IGFBP6 aggravates atherosclerosis in mice

   To further explore the role of EC-derived IGFBP6 in atherosclerosis
   development in hypercholesterolemic mice, we infected male and female
   Igfbp6^ECKO (Cdh5-Cre; Igfbp6^flox/flox) mice and Igfbp6^WT
   (Igfbp6^flox/flox) littermate controls with AAV8-Pcsk9^D377Y
   ([137]Extended Data Fig. 3a,[138]b). En face staining of the aortic
   intima of mice showed a specific ablation of IGFBP6 in the ECs of
   Igfbp6^ECKO mice compared to Igfbp6^WT control mice ([139]Fig. 3b).
   Compared to control mice, Igfbp6 mRNA levels in the intimal layer but
   not in the medial layer of Igfbp6^ECKO mice aorta were decreased
   ([140]Fig. 3c). ELISA also indicated that serum levels of IGFBP6 in
   Igfbp6^ECKO mice were significantly reduced by approximately 70%
   ([141]Fig. 3d), whereas EC-specific deletion of Igfbp6 did not induce
   changes in mRNA expression of other members of the Igfbps in the aortic
   intima ([142]Extended Data Fig. 3c), consolidating the specificity of
   IGFBP6 knockout in mouse aortic endothelium.

Fig. 3 |. EC-specific knockout of IGFBP6 aggravates atherosclerosis in mice.

   Fig. 3 |
   [143]Open in a new tab

   a, Schematic figure showing the experimental strategy. b, En face
   immunofluorescence staining for IGFBP6 (red), VE-cadherin (green) and
   DAPI (blue) in the aortas of mice. Decreased IGFBP6 expression in the
   intima of Igfbp6^ECKO mice was observed (n = 5). Scale bar, 50 μm. c,
   The expression of Igfbp6 mRNA in the intima and non-intima lysate of
   the aorta was detected by qRT–PCR (n = 5). d, IGFBP6 level in the serum
   of Igfbp6^ECKO mice was detected by ELISA (n = 4). e,f, Representative
   en face images of Oil Red O staining of atherosclerotic lesions of the
   aorta in male (e) (Igfbp6^WT: n = 8; Igfbp6^ECKO: n = 7) and female
   Igfbp6^ECKO mice and respective controls (f) (Igfbp6^WT: n = 8;
   Igfbp6^ECKO: n = 10). Scale bar, 2 mm. g,h, Representative images of
   atherosclerotic lesions in the aortic root stained with Oil Red O in
   male (g) (Igfbp6^WT: n = 8; Igfbp6^ECKO: n = 7) and female Igfbp6^ECKO
   mice and respective controls (h) (Igfbp6^WT: n = 8; Igfbp6^ECKO: n =
   11). Scale bar, 200 μm. i, Representative Oil Red O–stained
   cryosections of brachiocephalic arteries from male Igfbp6^ECKO mice and
   control mice. The quantification of the Oil Red O–positive plaque area
   is shown in the right panel (Igfbp6^WT: n = 7; Igfbp6^ECKO: n = 9).
   Scale bar, 50 μm. j, Massonʼs trichrome staining showed collagen
   expression in plaques, and decreased collagen expression in the aortic
   roots of male Igfbp6^ECKO mice was observed as compared to control mice
   (Igfbp6^WT: n = 8; Igfbp6^ECKO: n = 11). Scale bar, 200 μm. k, H&E
   staining of aortic root lesions was performed in male Igfbp6^ECKO mice
   and control mice, and the area of the necrotic plaque core was
   quantified (Igfbp6^WT: n = 8; Igfbp6^ECKO: n = 11). Scale bar, 200 μm.
   l,m, Immunofluorescence staining of α-SMA (red) and CD68 (green) in the
   diseased area of the aortic root in male (l, Igfbp6^WT: n = 8;
   Igfbp6^ECKO: n = 7) and female Igfbp6^ECKO mice as compared to control
   mice (m, Igfbp6^WT: n = 6; Igfbp6^ECKO: n = 8). Scale bar, 50 μm. Data
   are shown as mean ± s.d. Statistical analysis was performed by
   two-tailed Student’s t-test for c (left panel), e–k, l (right panel)
   and m (left panel), by Mann–Whitney U-test for c (right panel) and by
   Welch’s t-test for d and l (left panel) and m (right panel). W, weeks.

   At the end of 12 weeks of HCD feeding, no phenotypic differences were
   observed between Igfbp6^ECKO mice and Igfbp6^WT littermates regarding
   body weight, blood glucose or the levels of plasma TC, TG, HDL-C,
   LDL-C, AST or ALT ([144]Extended Data Fig. 3d). However, Oil Red O
   staining of mouse aorta and aortic sinus revealed a significant
   increase in atherosclerotic plaque area in both male and female
   Igfbp6^ECKO mice compared to the respective controls ([145]Fig.
   3e–[146]h). In addition, EC-specific deletion of Igfbp6 also increased
   the accumulation of plaques in the brachiocephalic artery ([147]Fig.
   3i). Massonʼs trichrome staining revealed a decrease in collagen
   content in the plaques of Igfbp6^ECKO mice ([148]Fig. 3j). Hematoxylin
   and eosin (H&E) staining of the aortic root showed that Igfbp6 ablation
   in ECs increased the size of lesional necrotic core ([149]Fig. 3k). In
   addition, α-SMA expression was decreased in the aortic plaques of
   Igfbp6^ECKO mice compared to the control group, whereas CD68^+
   macrophage content was increased ([150]Fig. 3l,[151]m). These data
   suggest that EC-specific knockout of Igfbp6 increases atherosclerosis
   in both male and female mice and increases features of plaque
   instability.

EC-specific overexpression of IGFBP6 ameliorates atherosclerosis in ApoE^−/−
mice

   To correlate IGFBP6 level and plaque progression, we established a
   mouse model of progressive atherosclerosis from early (HCD feeding for
   6 weeks) to advanced stage (HCD feeding for 18 weeks) of
   atherosclerosis. The plaque area of the en face aorta and aortic sinus
   of ApoE^−/− mice increased significantly with plaque progression
   compared to C57BL/6J control mice ([152]Fig. 4a). We observed that the
   expression of IGFBP6 in plaques of ApoE^−/− mice was reduced compared
   to healthy aortas ([153]Fig. 4b). Similar data were obtained in the
   serum of ApoE^−/− mice fed an HCD for 6–18 weeks ([154]Fig. 4c). These
   data implicate decreased expression level of IGFBP6 during
   atheroprogression.

Fig. 4 |. EC-specific overexpression of IGFBP6 ameliorates atherosclerosis in
ApoE^−/− mice.

   Fig. 4 |
   [155]Open in a new tab

   a, Representative images of atherosclerosis plaque development in the
   gross whole aorta and aortic sinus (stained with Oil Red O) of male
   C57BL/6J mice fed with a normal chow diet for 6 weeks and male ApoE^−/−
   mice fed with an HCD for 6 weeks or 18 weeks (n = 5). Scale bars, 1 mm
   and 200 μm. b, Immunofluorescence staining of IGFBP6 (red), CD31
   (green) and DAPI (blue) in the diseased area of the aortic root in male
   C57BL/6J and ApoE^−/− mice (n = 5). Scale bar, 25 μm. c, Serum samples
   were collected from male C57BL/6J and ApoE^−/− mice to verify IGFBP6
   levels by ELISA (n = 5). d, Schematic figure showing the experimental
   strategy. e, Immunofluorescence staining of IGFBP6 (red) in the
   diseased area of the aortic root in male AAV9-EC-Igfbp6 mice (n = 5).
   Scale bar, 50 μm. f, The expression of Igfbp6 mRNA in the intima and
   non-intima lysate of the aorta was detected by qRT–PCR (n = 5). g–i,
   Representative images of Oil Red O–stained aortas (g, n = 14), sections
   of aortic sinus (h, n = 9) and sections of brachiocephalic arteries (i,
   n = 10). Scatter plots show the statistical evaluation of the Oil Red
   O–positive areas. Scale bars, 2 mm (g and i) and 200 μm (h). j,
   Massonʼs trichrome staining showed increased collagen expression in the
   aortic roots of male AAV9-EC-Igfbp6 mice (n = 9). Scale bar, 200 μm. k,
   H&E staining of aortic root lesions was performed in male
   AAV9-EC-Igfbp6 mice, and the area of the necrotic core of the plaque
   was quantified (AAV9-EC-Con: n = 9; AAV9-EC-Igfbp6: n = 11). Scale bar,
   200 μm. l,m, Immunofluorescence staining of α-SMA (red) and CD68
   (green) in the diseased area of the aortic root and brachiocephalic
   arteries in male AAV9-EC-Igfbp6 mice (l, AAV9-EC-Con: n = 9 or n = 13;
   AAV9-EC-Igfbp6: n = 12; m, AAV9-EC-Con: n = 7 or n = 10;
   AAV9-EC-Igfbp6: n = 11 or n = 9). Scale bar, 50 μm. Data are shown as
   mean ± s.d. Statistical analysis was performed by two-tailed one-way
   ANOVA followed by Bonferroni’s test for c, by Student’s t-test for f–j
   and l (right panel), by Welch’s t-test for k and by Mann–Whitney U-test
   for l (left panel) and m. W, weeks.

   To determine whether IGFBP6 overexpression in ECs ameliorates
   atherosclerosis, we employed an EC-directed AAV9 vector driven by the
   ICAM-2 promoter to achieve EC-specific overexpression of Igfbp6
   (AAV9-EC-Igfbp6) in 8-week-old male ApoE^−/− mice. Littermate controls
   were injected with a control virus (AAV9-EC-Con). Concurrently, all
   mice were fed an HCD ([156]Fig. 4d) for 12 weeks. Increased IGFBP6
   levels were observed in aortic root and aortic intima in AAV9-EC-Igfbp6
   mice compared to control mice ([157]Fig. 4e,[158]f and [159]Extended
   Data Fig. 4a). At the same time, EC-specific overexpression of Igfbp6
   did not induce changes in mRNA expression of other members of the
   IGFBPs in the aortic intima ([160]Extended Data Fig. 4b). Terminally,
   no notable phenotypic differences were observed between the
   AAV9-EC-Igfbp6 mice and the control mice in terms of body weight, blood
   glucose and the levels of plasma TC, TG, HDL-C, LDL-C and AST, whereas
   the level of ALT tended to decrease ([161]Extended Data Fig. 4c). Oil
   Red O staining of the mouse aorta, aortic root and brachiocephalic
   arteries showed that EC-specific overexpression of Igfbp6 reduced the
   atherosclerotic plaque area ([162]Fig. 4g–[163]i). Massonʼs trichrome
   staining revealed that EC-specific overexpression of Igfbp6 increased
   collagen expression in atherosclerotic plaques ([164]Fig. 4j). In
   addition, H&E staining showed that EC-specific overexpression of Igfbp6
   reduced the size of the necrotic core area in plaques ([165]Fig. 4k).
   AAV9-EC-Igfbp6 mice also exhibited increased α-SMA expression in the
   brachiocephalic arteries and aortic root as well as decreased CD68^+
   macrophage infiltration ([166]Fig. 4l,[167]m). Collectively, these
   results suggest that EC-specific overexpression of Igfbp6 ameliorates
   atherosclerotic lesions in mice and enhances several features of plaque
   stability.

IGFBP6 inhibits endothelial inflammation

   Next, we further explored the role of IGFBP6 in endothelial
   inflammation. After treatment with various pro-inflammatory factors
   (IL-1α, IL-1β, tumor necrosis factor (TNF), interferon-gamma (IFN-γ)
   and lipopolysaccharide (LPS)), we observed significant downregulation
   of IGFBP6 mRNA levels under various inflammatory factor treatments
   ([168]Fig. 5a). To assess whether IGFBP6 exerts anti-inflammatory
   effects in ECs, we first conducted experiments on HUVECs and human
   aortic endothelial cells (HAECs). We used siRNA (si-IGFBP6) or IGFBP6
   overexpression adenovirus (Ad-IGFBP6) to knock down or overexpress
   IGFBP6, respectively, and assessed their impacts on TNF-induced
   inflammatory response. First, we confirmed successful IGFBP6 knockdown
   by ELISA ([169]Fig. 5b). In IGFBP6-silenced HUVECs, both the mRNA and
   protein levels of VCAM-1, intercellular adhesion molecule-1 (ICAM-1)
   and E-selectin (SELE) were significantly elevated ([170]Fig.
   5c,[171]d). In contrast, in IGFBP6-overexpressing HUVECs, both the mRNA
   and protein levels of VCAM-1, ICAM-1 and SELE were decreased compared
   to control ([172]Fig. 5e,[173]f). Of functional relevance, knockdown of
   IGFBP6 increased the adhesion of THP-1 monocytes to HUVECs ([174]Fig.
   5g), whereas overexpression of IGFBP6 inhibited THP-1 monocyte adhesion
   to HUVECs ([175]Fig. 5h). The anti-adhesive effect of IGFBP6 was
   recapitulated in HAECs ([176]Extended Data Fig. 5a–[177]f).

Fig. 5 |. IGFBP6 regulates EC inflammation.

   Fig. 5 |
   [178]Open in a new tab

   a, HUVECs were treated with various inflammatory stimuli (IL-1α, IL-1β,
   TNF, IFN-γ and LPS) for 24 h, and the mRNA expression of VCAM-1 and
   IGFBP6 was detected (n = 3–4). b, HUVECs were transfected with si-NC or
   si-IGFBP6 for 48 h and then treated with TNF for 6 h. The expression of
   IGFBP6 was determined by ELISA (b, n = 3). c,d, The expression of
   VCAM-1, ICAM-1 and SELE was determined by qRT–PCR (c, n = 5) and
   western blot (d, VCAM-1 and ICAM-1: n = 9; SELE: n = 3). e,f, HUVECs
   were transfected with Ad-NC or Ad-IGFBP6 for 24 h and then treated with
   TNF for 6 h. The expression of IGFBP6/Flag, VCAM-1, ICAM-1 and SELE was
   determined by qRT–PCR (e, n = 3) and western blot (f, VCAM-1 and
   ICAM-1: n = 5; SELE: n = 3). g, Number of adherent THP-1 cells to
   HUVECs treated with si-NC or si-IGFBP6 in the presence of TNF (n = 6).
   Scale bar, 50 μm. h, Number of adherent THP-1 cells to HUVECs treated
   with Ad-NC or Ad-IGFBP6 in the presence of TNF (n = 6). Scale bar, 50
   μm. i,j, HUVECs were transfected with Ad-NC or Ad-IGFBP6 for 24 h and
   then treated with UF or DF for 24 h, respectively. The expression of
   VCAM-1 and ICAM-1 was detected by qRT–PCR (i, n = 3) and western blot
   (j, n = 4). k, Number of adherent THP-1 cells to HUVECs treated with
   Ad-NC or Ad-IGFBP6 in the presence of DF or static (n = 4). Scale bar,
   50 μm. l, After intraperitoneal injection of murine TNF (500 ng per
   mouse) for 4 h, the leukocytes of mice were stained with Rhodamine 6G,
   and the number and migration rate of leukocytes in mesenteric vessels
   were observed and counted (n = 3). Scale bar, 100 μm. Data are shown as
   mean ± s.d. Statistical analysis was performed by two-tailed Student’s
   t-test for l (left and middle panel), by paired t-test for a, b, d and
   f, by two-way ANOVA (repeated measures) followed by Bonferroni’s test
   for c, e, g, i and j, by two-way ANOVA followed by Tukey’s test for h
   and k and by Welch’s t-test for l (right panel). Veh, vehicle.

   Furthermore, RNA-seq data indicate that IGFBP6 overexpression led to
   changes of DEGs related to TNF signaling pathway, fluid shear stress
   and atherosclerosis ([179]Extended Data Fig. 5g,[180]h). Specifically,
   overexpression of IGFBP6 in ECs reduced the mRNA expression of adhesion
   molecules (VCAM-1, ICAM-1 and SELE) and chemokines (C-C motif chemokine
   ligand 2 (CCL2), C-X-C motif chemokine ligand 1 (CXCL1), CXCL3, CXCL6
   and CXCL8) ([181]Extended Data Fig. 5i). Subsequently, we investigated
   whether IGFBP6 exerts anti-inflammatory effects in ECs exposed to DF.
   Compared to UF, exposure to DF upregulated the expression of ICAM-1 and
   VCAM-1 genes in HUVECs, which was reversed by IGFBP6 overexpression
   ([182]Fig. 5i). In parallel, DF-induced VCAM-1 protein upregulation was
   reversed by IGFBP6 overexpression ([183]Fig. 5j). As a consequence,
   overexpression of IGFBP6 inhibited the adhesion of THP-1 monocytes to
   HUVECs exposed to DF ([184]Fig. 5k).

   Finally, we evaluated whether Igfbp6 impacts leukocyte adhesion in vivo
   by intravital microscopy. We injected TNF into the peritoneal cavity of
   male Igfbp6^ECKO mice and Igfbp6^WT control mice at 4–6 weeks of age to
   induce leukocyte adhesion. We observed that the EC-specific knockout of
   Igfbp6 increased the number of rolling and adhering leukocytes in the
   mesenteric vessels of mice and slowed the movement of leukocytes
   ([185]Fig. 5l). These data indicate that IGFBP6 alleviates endothelial
   inflammation induced by inflammatory factors and prevents leukocyte
   adhesion to activated endothelium, further consolidating the protective
   role of IGFBP6 against endothelial dysfunction.

IGFBP6 is a transcriptional target of KLF2

   KLF2 is a master transcription factor responsive to shear stress and
   regulates the expression of a battery of mechanosensitive genes
   associated with vascular homeostasis^[186]29–[187]31. By analyzing the
   human IGFBP6 promoter region, we identified multiple putative binding
   sites of KLF2. Therefore, we examined the possibility of whether KLF2
   could transcriptionally regulate IGFBP6 expression in ECs. We thus
   treated HUVECs with three different statins (atorvastatin, simvastatin
   and rosuvastatin) as well as resveratrol at different concentrations
   for 24 h. We observed a concentration-dependent upregulation of KLF2
   and IGFBP6 mRNA expression in statin-treated and resveratrol-treated
   ECs ([188]Fig. 6a–[189]d).

Fig. 6 |. IGFBP6 is a transcriptional target of KLF2.

   Fig. 6 |
   [190]Open in a new tab

   a–d, HUVECs were treated with escalating concentrations of indicated
   chemicals (atorvastatin (a), simvastatin (b), rosuvastatin (c) and
   resveratrol (d)) for 24 h. Then, the mRNA expression of KLF2 and IGFBP6
   was detected (n = 5). e, HUVECs were treated with si-NC or si-KLF2 for
   48 h, and the mRNA expression of KLF2 and IGFBP6 was detected (n = 3).
   f, HUVECs were treated with siRNA or si-KLF2 for 24 h and then treated
   with DMSO or atorvastatin for 24 h, and the mRNA expression of KLF2 and
   IGFBP6 was detected (n = 5). g,h, HUVECs were treated with Ad-NC or
   Ad-KLF2 for 24 h, and the expression of KLF2 and IGFBP6 was determined
   by qRT–PCR (g, n = 4) and ELISA (h, n = 3). i, Dual-luciferase reporter
   assay showed (293T) that KLF2 regulated IGFBP6 promoter activity in
   HEK293T cells transfected with different mutants of IGFBP6 promoters (n
   = 5). j, In Ad-KLF2-transduced HUVECs, chromatin
   immunoprecipitation–PCR showed that KLF2 bound to the IGFBP6 promoter
   (n = 3). Data are shown as mean ± s.d. Statistical analysis was
   performed by two-tailed one-way ANOVA (repeated measures) followed by
   Bonferroni’s test for a and b (left panel) and d and f, by
   Kruskal–Wallis test for b (right panel), by Mann–Whitney U-test for g
   (right panel), by Welch’s t-test for i (Ad-NC WT versus Ad-KLF2 WT), by
   one-way ANOVA followed by Tukey’s test for i (Ad-KLF2 WT versus Ad-KLF2
   MT1 versus Ad-KLF2 MT2), by paired t-test for c, e and g (left panel)
   and j and by Student’s t-test for h.

   We then employed siRNA (si-KLF2) or Flag-tagged adenovirus (Ad-KLF2) to
   knock down or overexpress KLF2 in HUVECs, respectively. We found that
   KLF2 positively regulates the mRNA expression of IGFBP6 ([191]Fig.
   6e–[192]h). Also, we found that KLF2 depletion partially reversed the
   upregulation of IGFBP6 mRNA expression induced by atorvastatin
   ([193]Fig. 6f). Next, we constructed various plasmid vectors (including
   empty vector (basic), vector containing the IGFBP6 promoter (WT) and
   vectors containing mutations that deleted all CACCC (MT1) or GGGTG
   (MT2) sequences in the IGFBP6 promoter) for the dual-luciferase
   reporter assay. The results showed that overexpression of KLF2
   increased the activity of the WT IGFBP6 promoter and the MT1 IGFBP6
   promoter but not the MT2 IGFBP6 promoter, indicating that KLF2 could
   enhance the activity of the IGFBP6 promoter and bind to the GGGTG
   sequence in the IGFBP6 promoter ([194]Fig. 6i). Chromatin
   immunoprecipitation–PCR also confirmed that KLF2 binds to the IGFBP6
   promoter region ([195]Fig. 6j). Taken together, these findings
   collectively indicate that IGFBP6 is a bona fide transcriptional target
   of KLF2. Because KLF4 regulates a plethora of overlapping genes as
   KLF2, we also confirmed that KLF4 also regulates the expression of
   IGFBP6. The results suggest that IGFBP6 may also be a transcriptional
   target of KLF4 ([196]Extended Data Fig. 6a,[197]b).

IGFBP6 interacts with MVP and suppresses endothelial inflammation via
MVP-dependent inhibition of p65 and JNK phosphorylation

   To unravel the direct mechanism of the anti-inflammatory effects of
   IGFBP6 in ECs, we precipitated IGFBP6 from overexpressed ECs followed
   by mass spectrometry. Based on protein-specific peptide segments and
   scores, MVP had a high ranking score ([198]Fig. 7a).
   Co-immunoprecipitation assays revealed an interaction between exogenous
   IGFBP6 and endogenous MVP in ECs ([199]Fig. 7b). Co-transfection of
   Flag-tagged IGFBP6 and HA-tagged MVP in HEK293T cells revealed that
   IGFBP6 interacted with MVP ([200]Fig. 7c). Immunofluorescence staining
   also showed that IGFBP6 co-localized with MVP in the cytoplasm of ECs
   ([201]Fig. 7d). Through surface plasmon resonance (SPR) experiments, we
   found that MVP can directly bind to the IGFBP6 protein in a
   concentration-dependent manner ([202]Fig. 7e).

Fig. 7 |. IGFBP6 interacts with MVP and suppresses endothelial inflammation
via MVP-dependent inhibition of p65 and JNK phosphorylation.

   Fig. 7 |
   [203]Open in a new tab

   a, HUVECs were treated with Ad-NC or Ad-IGFBP6 for 24 h before
   treatment with TNF for 6 h. Lysates were immunoprecipitated with
   anti-Flag-M2 beads. List of IGFBP6-binding proteins identified by
   LC–MS/MS. b,c, Immunoprecipitation (IP) was performed with the
   indicated antibodies in HUVECs (b) and HEK293T cells (c). d,
   Immunofluorescence staining for IGFBP6 (red), MVP (green) and DAPI
   (blue) in HUVECs (n = 4). Scale bar, 10 μm. e, SPR analysis of the
   direct interaction of MVP and IGFBP6. f, HUVECs were transfected with
   si-NC or si-MVP for 48 h and treated with TNF for 6 h. Whole cell
   lysate of treated cells was immunoblotted to detect MVP, ICAM-1, P-p65,
   p65, P-JNK and JNK (n = 3). g,i, ECs were transfected with si-NC or
   si-IGFBP6 for 48 h and then treated with TNF for different times. The
   protein expression of P-p65, p65, P-JNK and JNK was determined in HAECs
   (g) and HUVECs (i) (n = 3). h,j, ECs were transfected with Ad-NC or
   Ad-IGFBP6 for 24 h and then treated with TNF for different times. The
   protein expression of P-p65, p65, P-JNK and JNK was determined in HAECs
   (h) and HUVECs (j) (n = 3). k, HUVECs were transfected with adenovirus
   (Ad-NC or Ad-IGFBP6) and siRNA (si-NC or si-MVP) for 48 h and then
   treated with TNF for 6 h. THP-1 monocyte adhesion assay was performed
   (n = 3). Scale bar, 50 μm. l, The protein expression of Flag-tagged
   IGFBP6, MVP, ICAM-1, P-p65, p65, P-JNK and JNK was determined in HUVECs
   (n = 3). m,n, Immunofluorescence staining of P-JNK (red) and JNK
   (green) in the diseased area of the aortic root in male Igfbp6^ECKO
   mice (m, Igfbp6^WT: n = 8; Igfbp6^ECKO: n = 13) and male AAV9-EC-Igfbp6
   mice (n, AAV9-EC-Con: n = 8; AAV9-EC-Igfbp6: n = 6). Scale bar, 50 μm.
   Data are shown as mean ± s.d. Statistical analysis was performed by
   two-tailed Student’s t-test for k, by Mann–Whitney U-test for m and by
   Welch’s t-test for n.

   Previous studies demonstrated that MVP suppresses inflammation by
   inhibiting the TNF receptor-associated factor 6 (TRAF6)–NF-κB and
   apoptosis signal-regulating kinase 1 (ASK1)–JNK signaling pathways in
   macrophages, leading to protection against atherosclerosis in
   mice^[204]32,[205]33. We demonstrated that MVP silencing in HUVECs
   increased ICAM-1 and SELE mRNA expression levels, increased monocyte
   adhesion to HUVECs and increased the phosphorylation of JNK and p65
   ([206]Fig. 7f and [207]Extended Data Fig. 6c,[208]d). Moreover, we
   induced endothelial cell inflammation by treating HUVECs and HAECs with
   TNF at different timepoints. In IGFBP6-silenced HAECs, the expression
   of phosphorylated JNK at serine 183 and threonine 185 was increased,
   whereas the overexpression of IGFBP6 significantly decreased the
   expression of phosphorylated JNK and p65 ([209]Fig. 7g,[210]h). In
   HUVECs, similar results were obtained ([211]Fig. 7i,[212]j).
   Interestingly, MVP silencing reversed the preventive effect of IGFBP6
   on monocyte adhesion ([213]Fig. 7k) and JNK and p65 phosphorylation
   ([214]Fig. 7l). Immunofluorescence staining of the mouse aortic sinus
   revealed that EC-specific knockout of Igfbp6 increased JNK
   phosphorylation in plaques, whereas EC-specific overexpression of
   Igfbp6 exhibited the opposite effects ([215]Fig. 7m,[216]n).

   To identify the specific mechanism whereby IGFBP6 inhibits inflammation
   through MVP, we explored the mechanistic details downstream of
   IGFBP6–MVP interaction. In previous reports, MVP in macrophages
   inhibited downstream JNK and p65 activation by inhibiting the
   dimerization of ASK1 and TRAF6 (refs. [217]32,[218]33). We first
   verified in HUVECs that MVP can also bind to ASK1 ([219]Extended Data
   Fig. 7a). The interaction between exogenous IGFBP6 and endogenous ASK1
   in ECs was further validated by co-immunoprecipitation ([220]Extended
   Data Fig. 7b). We next evaluated the effect of IGFBP6 on the
   dimerization of ASK1. Our findings verified the exogenous interaction
   between ASK1 and MVP and the inhibition of ASK1 dimerization by MVP.
   IGFBP6 also binds with ASK1 and increases the interaction between MVP
   and ASK1. Most importantly, IGFBP6 promoted the binding of MVP to ASK1,
   further inhibiting the ASK1 dimerization ([221]Extended Data Fig. 7c).
   As a result, IGFBP6 downregulates the phosphorylation of JNK downstream
   of ASK1 dimerization.

   We then employed AlphaFold 3–based molecular docking to predict the
   binding sites of IGFBP6 and MVP and pre-selected three possible binding
   sites according to screening conditions, such as hydrogen bond distance
   ([222]Extended Data Fig. 7d). We constructed four mutants of IGFBP6,
   including three mutants with a single point mutation to alanine
   (E48A/R65A/R103A) and one mutant carrying triple point mutations (E48A
   + R65A + R103A). All these plasmid vectors carried flag tags and were
   transfected into HEK293T cells together with HA-MVP vector. Our results
   showed that a single mutation at E48 or R65 or R103 could slightly
   reduce the interaction between IGFBP6 and MVP with varying potencies,
   whereas mutation of all three potential binding sites could
   significantly dampen the binding between IGFBP6 and MVP ([223]Extended
   Data Fig. 7e). We then explored whether the anti-inflammatory effect of
   IGFBP6 depended on these key binding sites. We constructed IGFBP6
   adenovirus with mutations at three binding sites and transfected into
   HUVECs for downstream functional assays. The results showed that
   overexpression of WT IGFBP6 could significantly inhibit the
   inflammation of ECs, whereas this anti-inflammatory effect was weakened
   in ECs infected with the triple mutants of IGFBP6, indicating that
   these three binding sites were at least partially involved in relaying
   the anti-inflammatory IGFBP6–MVP signaling axis ([224]Extended Data
   Fig. 7f). These results suggested that IGFBP6 exerted anti-inflammatory
   effects by binding to MVP.

   It was reported that MVP can be detected in human serum as a biomarker
   of disease^[225]34. We speculate whether IGFBP6 is secreted
   extracellularly with MVP to play an anti-inflammatory role. First, we
   detected MVP protein extracellularly in conditioned media from
   MVP-overexpressed ECs ([226]Extended Data Fig. 8a). We next explored
   the possibility of the extracellular interaction between IGFBP6 and
   MVP. Our data demonstrated that MVP and IGFBP6 secreted in concentrated
   endothelial conditioned media do not interact with each other
   ([227]Extended Data Fig. 8a,[228]b). To further validate whether
   secreted IGFBP6 exerts anti-inflammatory effects, we treated HUVECs
   with different amounts of recombinant human IGFBP6 protein before
   stimulation with TNF. Our data showed that extracellular IGFBP6 had no
   observable anti-inflammatory effects. The phosphorylation of JNK and
   p65 was not affected either ([229]Extended Data Fig. 8c,[230]d).
   Similarly, the conditioned medium from HUVECs overexpressing IGFBP6 did
   not elicit appreciable anti-inflammatory effects ([231]Extended Data
   Fig. 8e). Subsequently, we further evaluated the anti-inflammatory
   effect of extracellular MVP and found that different concentrations of
   recombinant MVP had no anti-inflammatory effect on ECs ([232]Extended
   Data Fig. 8f). In summary, these findings indicate that IGFBP6 exerts
   its anti-inflammatory effects mainly by interacting with intracellular
   MVP rather than extracellular MVP.

   Taken together, IGFBP6 inhibits JNK and p65 through its interaction
   with intracellular MVP, thereby reducing the expression of adhesion
   molecules and inhibiting monocyte adhesion to ECs, ultimately combating
   endothelial inflammation.

Discussion

   In this study, we demonstrated that IGFBP6 is a negative regulator of
   endothelial inflammation and leukocyte adhesion, thereby inhibiting the
   development of atherosclerosis (Graphic Abstract and [233]Extended Data
   Fig. 9). This conclusion is backed by multiple findings from cultured
   human ECs, various mouse models and human patients. The salient
   findings of the present study can be summarized as follows: (1) the
   expression of IGFBP6 is decreased in serum and atherosclerotic tissues
   of patients with CAD; (2) gain and loss of function of endothelial
   Igfbp6 in atheroprone mouse models suggest that Igfbp6 ameliorates the
   progression of atherosclerosis in mice; and (3) IGFBP6 is a
   transcriptional target of KLF2 that inhibits EC inflammation and
   monocyte adhesion through the MVP–JNK/NF-κB pathway.

   KLF2 is a key regulator of endothelial homeostasis and
   atheroprotection. Some chemicals have been shown to increase KLF2
   expression or activity, such as statins and
   resveratrol^[234]30,[235]35,[236]36. In the present study, we found
   that both statins and resveratrol upregulate IGFBP6 expression in ECs
   by activating KLF2, suggesting that IGFBP6 is druggable and that its
   upregulation may contribute to the pleiotropic and atheroprotective
   effects of statins and resveratrol.

   To elucidate the molecular mechanism by which IGFBP6 suppresses
   inflammation, we identified MVP as an IGFBP6-interacting protein that
   mediates the anti-inflammatory effects of IGFBP6. MVP was shown in
   macrophages to limit atherosclerosis through the ASK1–JNK and NF-κB
   signaling pathway^[237]32,[238]33. Similar to the role of MVP in
   macrophages, we reported that IGFBP6 also decreased the phosphorylation
   of JNK and NF-κB (p65) dependent on MVP in endothelial cells. We
   further found that IGFBP6 also interacts with ASK1, and IGFBP6
   mitigates the dimerization of ASK1 by binding to MVP. We predicted the
   potential binding sites of IGFBP6 and MVP by molecular docking and
   confirmed the key binding domains in this interaction. All these lines
   of evidences collectively pinpoint the anti-inflammatory effects of
   IGFBP6 in ECs through binding with MVP.

   The IGFBP family consists of six high-affinity proteins (IGFBP1–6) and
   one low-affinity protein (IGFBP7) for IGF, which regulate IGF
   utilization, cell proliferation and differentiation^[239]37. The role
   of IGFBPs in atherosclerosis is being recognized. For example, IGFBP1
   attenuates atherosclerosis in mice through the PI3K–Akt–eNOS
   pathway^[240]38. IGFBP2 is a positive predictor of MACE in patients
   with acute coronary syndrome^[241]39. Cell senescence promotes fibrous
   cap degradation and ultimately leads to atherosclerotic plaque rupture
   through IGFBP3 (ref. [242]40). When compared to non-lesional areas,
   IGFBP4 is upregulated in the arterial lesion area of ApoE^−/−
   mice^[243]41. Our previous study reported that IGFBP5 is
   mechanoresponsive and anti-inflammatory^[244]28. IGFBP7 expression is
   increased in patients with atherosclerotic aortic aneurysm^[245]42. It
   is known that IGFBPs play different roles in cardiovascular diseases,
   including atherosclerosis, because different IGFBPs have different
   tissue expression levels and may exert different pathophysiological
   functions under different disease contexts. The expression levels of
   IGFBPs in ECs varied greatly, with IGFBP1–5 being lowly expressed in
   HUVECs, whereas IGFBP6 and IGFBP7 had relatively high expression levels
   ([246]Extended Data Fig. 1c). According to the previous database mining
   and the mechanoresponsive nature of IGFBP6, we observed that, compared
   to IGFBP7, IGFBP6 is an EC-enriched IGFBP member that responds to blood
   flow shear stress. Nevertheless, the exact role of IGFBP6 in vascular
   biology and diseases remains largely unknown.

   Previous studies showed that IGFBP6 is essential for regulating cell
   cycle, cell proliferation and cell migration^[247]19–[248]21. In ECs,
   IGFBP6 inhibits angiogenesis independently of IGF2, leading to a
   significant reduction in the number of blood vessels in zebrafish
   embryos as well as a marked decrease in the blood vessel density of
   BALB/c nude mouse xenografts^[249]20. The anti-angiogenic role of
   IGFBP6 aligns with the atheroprotective effects of IGFBP6 because
   neovascularization may promote the progression of atherosclerotic
   disease and renders the plaque unstable^[250]43. Also, it was reported
   that, compared to stable plaques in patients with CAD, the expression
   level of IGFBP6 in unstable plaques was decreased^[251]22. However, the
   direct role and mechanism of IGFBP6 in endothelial homeostasis and
   atherogenesis remain unclear. In our study, global or EC-specific
   deletion of Igfbp6 exacerbated HCD-induced or DF-induced
   atherosclerotic lesion development. In contrast, overexpression of
   Igfbp6 in ECs ameliorated HCD-induced atherosclerotic lesion
   development. These whole battery of gain-of-function and
   loss-of-function studies in mice suggests that endothelial IGFBP6
   negatively regulates the progression of atherosclerotic lesions and
   highlights the potential of IGFBP6-elevating strategies for treating
   atherosclerotic diseases.

   We recognize that this study has several limitations. First, IGFBP6 was
   also expressed in non-ECs of blood vessels, such as vascular smooth
   muscle cells (VSMCs). However, it remains unexplored in this study
   whether VSMC-derived IGFBP6 has similar atheroprotective effects.
   Second, the paracrine functions of IGFBP6 secreted from ECs on
   neighboring VSMCs or macrophages remain to be explored. Finally, from
   the translational perspective, liposome nanoparticles carrying IGFBP6
   mRNA to deliver IGFBP6 specifically to plaque ECs are needed to achieve
   EC-targeted IGFBP6 delivery for cardiovascular therapeutics.

   In conclusion, our study suggests that IGFBP6 is mechanoresponsive,
   anti-inflammatory and atheroprotective. The elevation of endothelial
   IGFBP6 genetically or pharmacologically may represent a strategy to
   reduce residual inflammatory risk for CAD.

Methods

Experimental mice and treatment

   Male ApoE^−/− mice at 8 weeks of age were purchased from GemPharmatech
   (no. T001458). Igfbp6^flox/flox and Igfbp6^−/− mice were sourced from
   the Shanghai Model Organisms Center (no. NM-KO-205448 and no.
   NM-CKO-2107216). Igfbp6^ECKO mice were generated by crossing
   Igfbp6^flox/flox mice with Cdh5-Cre mice (Shanghai Model Organisms
   Center, no. 017968), resulting in EC-specific deletion of Igfbp6.
   Eight-week-old male ApoE^−/− mice were subjected to EC-specific
   overexpression of Igfbp6 using an EC-specific AAV9 vector driven by the
   ICAM-2 promoter (AAV9-EC-Igfbp6, 1 × 10^12 vg per mouse)^[252]44. In
   addition to ApoE^−/− mice, hepatic LDLR protein was downregulated by a
   single dose of AAV8-Pcsk9^D377Y (3 × 10^11 vg per mouse) injected into
   the tail vein^[253]45. The AAVs were constructed by Vigene Biosciences.
   All mice were fed an HCD (Research Diets, no. D12108C; 20% fat and
   1.25% cholesterol) to induce atherosclerosis. Mice were housed in
   specific pathogen-free (SPF) standard cages with a 12-h light/dark
   cycle and allowed to freely consume standard diets with a temperature
   of 20–24 °C and a humidity of 30–70%. The animal study protocols were
   approved by the Animal Care and Use Committee of University of Science
   and Technology of China (USTCACUC212401038) and carried out in
   accordance with the guidelines for the care and use of laboratory
   animals from the National Institutes of Health.

Human samples

   Human samples were obtained from The First Affiliated Hospital of the
   University of Science and Technology of China, with the research
   involving these samples approved by the hospital’s ethics committee
   (2023KY-383 and 2024KY-397). Patient tissue samples were obtained from
   the coronary arteries of heart transplant recipients with heart failure
   after myocardial infarction (n = 3), and healthy human tissue samples
   were taken from coronary arteries of abandoned heart transplant donors
   (n = 3). Serum samples from healthy individuals were obtained from a
   health examination center (n = 12), whereas those from patients with
   CAD were obtained from the department of cardiac surgery (n = 19).
   Serum was used for ELISA experiments, and coronary artery tissues were
   embedded in formalin and sectioned continuously after embedding in
   optimal cutting temperature compound (OCT; Sakura, no. 4583) for
   immunofluorescence staining and other experiments. This study was
   conducted in accordance with the Declaration of Helsinki, and all
   participants provided informed consent. Patient clinical
   characteristics are provided in [254]Supplementary Tables 1 and [255]2.

Chemicals and reagents

   Recombinant human TNF (no. 300–01A), IL-1α (no. 200–01A), IL-1β (no.
   200–01B) and IFN-γ (no. 300–02) were purchased from PeproTech,
   dissolved in sterile distilled water and then diluted with 0.1% BSA
   (Sigma-Aldrich, no. 9048–46-8) in PBS to a final concentration of 10 ng
   ml^−1 for in vitro experiments. The amount of recombinant murine TNF
   (BBI, no. C600052–0005) was 500 ng per mouse for intravital microscopy.
   LPS (no. [256]T11855), atorvastatin (no. 134523–00-5), simvastatin (no.
   79902–63-9), rosuvastatin (no. 287714–41-4) and resveratrol (no.
   501–36-0) were purchased from TargetMol. LPS was added to HUVECs at a
   final concentration of 1 μmol L^−1. Calcein AM (Beyotime Biotechnology,
   no. 148504–34-1) was used to stain THP-1 monocytes. Rhodamine 6G
   (Sigma-Aldrich, BCCF1630) was used to stain leukocytes in mice. The
   antibodies used in this study included anti-IGFBP6 (Abcam, ab219560,
   lot no. GR3219721–4; 1:1,000), anti-Flag (Sigma-Aldrich, no. F1804;
   1:2,000) and anti-VCAM-1 (Zenbio, no. 381014, lot no. L28JU0P; 1:1,000)
   for immunoblotting; anti-VCAM-1 (Cell Signaling Technology (CST), no.
   39036; 1:100) for immunofluorescence; and anti-ICAM-1 (Zenbio, no.
   200350-D6; 1:1,000), anti-SELE (BioVision, no. 3631; 1:1000), anti-MVP
   (Huabio, no. ET1705–69, lot no. HK0904; 1:1,000), anti-HA (Proteintech,
   no. 81290–1-RR, lot no. 00116648; 1:1,000), anti-P-p65 (CST, no. 3033S,
   lot no. 19; 1:1,000), anti-p65 (CST, no. 6956S, lot no. 10; 1:1,000),
   anti-P-JNK (CST, no. 81E11, lot no. 17; 1:1,000), anti-JNK
   (Proteintech, no. 66210–1-Ig, lot no. 10023911; 1:1,000), anti-ASK1
   (CST, no. 3762; 1:1000), anti-CD144 (BD Biosciences, no. 555289 and no.
   555661, lot no. 1076252; 1:100), anti-CD31 (BD Biosciences, no. 553370,
   lot no. 1005853; 1:100), anti-CD68 (BD Biosciences, no. 566386; 1:100),
   anti-α-SMA (Proteintech, no. 14395–1-AP, lot no. 00102366; 1:100),
   anti-GAPDH (Proteintech, no. 60004–1-Ig, lot no. 10029187; 1:10,000),
   anti-β-actin (Proteintech, no. 66009–1-Ig, lot no. 10029234; 1:1,000)
   and anti-IgG (Proteintech, no. 30000–0-AP; 1:100; B900620 and B900630).

Cell culture and shear stress experiments

   Different donors of HUVECs (cat. no. FC-0003, lot no. 08119/10128/04
   608/07815/06290/04827/08478/06279) and HAECs (cat. no. FC-0014, lot no.
   02357/01921/04675) were purchased from Lifeline Cell Technology. In
   this study, 3rd–7th generation HUVECs and HAECs were used^[257]46. The
   cells were cultured in ECM medium, which contained 1× endothelial cell
   growth supplement (ScienCell), 5% FBS and 1% penicillin–streptomycin
   antibiotics. HEK293T cells were purchased from the American Type
   Culture Collection (ATCC) using high-glucose DMEM containing 10% FBS
   (Sigma-Aldrich, no. 1943609–65-1) and 1% penicillin–streptomycin
   antibiotics (Sigma-Aldrich, RNBL3872). The human monocyte line THP-1
   was purchased from the ATCC in RPMI 1640 medium supplemented with 10%
   FBS (Biosharp, no. BL303A). All the above cells were cultured at 37 °C
   and 5% CO[2]. In the in vitro laminar flow shear stress experiment,
   confluent HUVEC monolayers were seeded on 100-mm plates, and a
   cone-and-plate shear stress device (Jiangsu Stoli Instrument Co., Ltd.)
   was used to generate DF or UF for 24 h^[258]47. The flow system was
   enclosed in a chamber maintained at 37 °C and ventilated with 95%
   humidified air and 5% CO[2].

Monocyte adhesion assay

   HUVECs or HAECs were treated with adenovirus or siRNA. THP-1 cells
   cultured in RPMI 1640 medium were counted to ensure that the cell
   density was approximately 10^5 cells per milliliter. THP-1 cells were
   then stained with the 5 μM fluorescent dye Calcein-AM and incubated at
   37 °C in the dark for 30 min. After centrifugation, the cell pellet was
   washed three times with RPMI 1640 medium, and the THP-1 cells were
   resuspended and evenly added to each well. The cells were incubated at
   37 °C in the dark for 30 min. Non-adherent cells were removed by
   washing three times with pre-warmed ECM medium and then images were
   captured under a microscope (Carl Zeiss Vision, Axio Vert.A1) to
   quantify the number of adherent cells.

Intravital microscopy of leukocyte–endothelial interaction in mesenteric
veins in mice

   Four- to six-week-old male Igfbp6^ECKO mice and littermate control
   Igfbp6^WT mice with body weights ranging from 16 g to 20 g were
   intraperitoneally injected with recombinant murine TNF (500 ng per
   mouse) to induce leukocyte adhesion in mice. Four hours later,
   Rhodamine 6G was injected into the tail vein to stain the leukocytes in
   the mice. Then, the mice were anesthetized, and the abdomen of the mice
   was sterilized. A median laparotomy was performed on the mice, and
   several drops of pre-warmed saline were dropped into the abdominal
   cavity to keep the tissue moist. Mesenteric veins were spread out in a
   100-mm dish, and 200–300-μm mesenteric veins were placed in the center
   of the field of view of the microscope. Images were then acquired using
   an inverted microscope (Nikon Ti2-E) equipped with a Yogokawa CSU-W1
   Spinning Disk Confocal Scanner Unit^[259]48. Fluorescence was
   visualized with the 488-nm laser, combined with emission filter
   (B525/50), recorded by an sCMOS camera (Photometrics, Prime95B).
   Imaging data were analyzed using NIS-Elements AR software (version
   5.20.00)^[260]48. Leukocyte–endothelial interaction was recorded for 1
   min in 3–5 different veins from each mouse. The amount of rolling and
   adherent leukocytes was finally quantified, and the rolling velocity
   was determined by measuring the time a single leukocyte would roll
   steadily over the endothelium simultaneously across a 100-μm distance.

Serum lipid profile

   After anesthesia with 2.5% isoflurane gas, blood was collected from the
   mice through the orbital venous plexus, and serum was collected by
   centrifugation at 3,000 r.p.m. for 10 min at room temperature. Lipid
   levels were measured by Servicebio Technology using a detection kit
   (Ranto) to measure serum TC (no. S03042), TG (no. S03027), HDL-C (no.
   S03025), LDL-C (no. S03029), ALT (no. S03030) and AST (no. S03040).

Partial ligation of the LCA

   Eight-week-old male Igfbp6^−/− and WT mice were injected with 3 × 10^11
   vg per mouse AAV8-Pcsk9^D377Y (Vigene Biosciences, Addgene plasmid no.
   58376; [261]http://n2t.net/addgene:58376; RRID: Addgene_58376; gifted
   from Jacob Bentzon) via the tail vein and concurrent with feeding with
   an HCD^[262]45,[263]49,[264]50. After 2 weeks, all mice were
   anesthetized with 2.5% isoflurane gas, and the ventral midline of the
   neck was opened approximately 3–4 mm. The left external carotid artery
   (ECA) was ligated first using a 6–0-gauge suture; then, the internal
   carotid artery (ICA) was ligated together with the occipital artery
   (OA); and finally, only the superior thyroid artery (STA) was retained,
   and the mouse wound was sutured. Mice were observed daily for 3–4 d
   after surgery, and mice with dehiscence of the wound were disinfected.
   After 3 weeks, the mice were euthanized; the bilateral carotid arteries
   were fixed with 4% paraformaldehyde (PFA); and the dehydrated carotid
   arteries were stored under OCT using 30% sucrose and immediately
   frozen. Frozen sections were stained using Oil Red O or
   immunofluorescence.

Plasmid construction and viral or siRNA transfection

   Recombinant adenoviruses encoding Flag-tagged human IGFBP6 (Ad-IGFBP6;
   [265]NM_002178.3), human KLF2 (Ad-KLF2; [266]NM_016270.4), Flag-tagged
   human IGFBP6 mutation (Ad-IGFBP6[E48A+R65A+R103A]) and control
   adenovirus (Ad-NC) were constructed by Vigene Biosciences. siRNA
   (si-IGFBP6, si-KLF2 and si-MVP) or control (si-NC) for human IGFBP6,
   KLF2 and MVP ([267]NM_017458.3) or control (si-NC) were purchased from
   RiboBio Co., Ltd. The sequences of all siRNAs are provided in
   [268]Supplementary Table 3. Lots of plasmids (HA-MVP, Myc-MVP,
   ASK1-Flag, ASK1-HA and IGFBP6-Flag (E48A, R65A and E103A)) were
   constructed by Vigene Biosciences. All plasmid sequences are listed in
   [269]Supplementary Table 4. Adenovirus was used at a titer of
   multiplicity of infection (MOI) = 1, and the cells showed no obvious
   toxicity. Lipofectamine 2000 (Invitrogen, no. 11668–019) was used to
   transfect ECs and HEK293T cells. Cells cultured in 12-well plates were
   transfected with 20 nmol L^–1 siRNA or 1 μg of plasmid per well, and
   HEK293T cells were transfected using 4 μg in 100-mm dishes for 24 h.

Total RNA extraction, RNA-seq and qRT–PCR

   RNA was isolated using RNeasy kits (Yishan Biotech). After quality
   control of RNA, a library for RNA-seq was established at the Beijing
   Genomics Institute. In brief, mRNAs were separated from total RNA using
   magnetic beads attached to oligonucleotides (dT). First-strand cDNA was
   generated by reverse transcription with random hexamer primers, and
   then the second-strand cDNA was synthesized. The synthesized cDNA was
   subjected to end repair, and 3′ adenylation was then performed.
   Adapters were attached to the ends of these 3′ adenylated cDNA
   fragments. The cDNA fragments were amplified by PCR, purified with
   AMPure XP Beads (Thermo Fisher Scientific, Agencourt) and dissolved in
   EB solution. The library was validated on a Bioanalyzer system (Agilent
   Technologies, no. G2939BA). The double-stranded PCR products were
   thermatively denatured and circularized using the splint oligo
   sequence. The single-strand circular DNA (ssCir DNA) was formatted into
   the final library. The library was amplified using phi29 to generate
   DNA nanospheres (DNA-NSs), each containing over 300 copies of a single
   molecule. The DNA-NSs were loaded into the patterned nanoarray, and
   single-end 50 (paired-end 100/150) base reads were produced by
   combinatorial probe-anchor synthesis (cPAS).

   Total RNA was reverse transcribed into cDNA using reverse transcription
   kits (TaKaRa, no. RR037A). Upon completion of the reverse transcription
   process, real-time PCR was carried out using a Roche LC96 real-time PCR
   detection system using AceQ qPCR SYBR Green Master Mix (Vazyme, no.
   Q111–02). The relative expression of RNA was calculated by the 2^−ΔΔCt
   method, and GAPDH was used as a normalization control. Primer sequences
   are shown in [270]Supplementary Table 5.

DEG analysis in Gene Expression Omnibus datasets

   Transcriptome sequencing datasets in this study were collected from the
   Gene Expression Omnibus (GEO) ([271]https://www.ncbi.nlm.nih.gov/gds/)
   database and analyzed using the online software GEO2R
   ([272]https://www.ncbi.nlm.nih.gov/geo/geo2r/). The GEO datasets used
   in this study included [273]GSE41571, [274]GSE163154, [275]GSE20739,
   [276]GSE87534 and [277]GSE1176531.

Co-immunoprecipitation

   Whole cell lysates were prepared with lysis buffer (50 mM Tris-HCl (pH
   7.4), 140 mM NaCl, 5% glycerol and 1% Triton X-100) containing protease
   inhibitors (MCI, no. HY-K0010) and phosphatase inhibitors (Yeasen, no.
   20109ES05) for 1 h on ice, and proteins were incubated with 2 μg of
   specific antibodies in continuous rotation at 4 °C overnight. Then, 30
   μl of Protein A/G Magnetic (MCE, no. HY-K0202) was added and incubated
   for 2 h with constant rotation speed at 4 °C. The beads were adsorbed
   on a magnetic stand and washed three times with wash buffer (50 mM
   Tris-HCl (pH 7.4), 140 mM NaCl and 0.1% Triton X-100). After the last
   wash, the supernatant was aspirated and discarded, and precipitated
   proteins were subsequently resuspended in sample buffer and boiled at
   95 °C for 10 min. The material obtained by immunoprecipitation of cell
   lysates was subjected to western blot analysis.

SPR assay

   The binding affinity was measured by SPR using a BIAcore T200 (GE
   Healthcare) instrument. Recombinant human IGFBP6 (Sino Bio, no.
   13026-H08H) was diluted in sodium acetate solution (pH 5.0) with a
   final concentration of 20 μg ml^−1. IGFBP6 was immobilized on a CM5
   sensor chip (GE Healthcare) by amine coupling to reach appropriate
   1,036 resonance units (RU) density. The running buffer contained a 1×
   HBS buffer. Instrument settings remained at 25 °C.

   Then, 10 concentrations of the recombinant human MVP (Abnova, no.
   H00009961-P01) (0 μg ml^−1, 0.01953125 μg ml^−1, 0.0390625 μg ml^−1,
   0.078125 μg ml^−1, 0.15625 μg ml^−1, 0.3125 μg ml^−1, 0.625 μg ml^−1,
   1.25 μg ml^−1, 2.5 μg ml^−1 and 5 μg ml^−1) were injected at a flow
   rate of 30 μl min^−1. The contact time was 2 min for IGFBP6 and MVP;
   the dissociation time was 2 min for IGFBP6 and MVP. A blank
   immobilization was performed for one of the sensor chip surfaces to use
   for correction of the binding response. Sensorgrams were analyzed using
   Igor Pro version 6.1 (WaveMatrix).

Western blotting

   Whole lysates isolated from cell and tissue samples were lyzed with
   RIPA buffer supplemented with protease inhibitors (MCE, no. HY-K0010)
   and phosphatase inhibitors (Yeasen, no. 20109ES05) and boiled at 95 °C
   for 10 min. Samples were separated by SDS–PAGE, transferred to
   nitrocellulose membranes (Pall) and subsequently incubated at room
   temperature with blocking buffer (LI-COR) for 1 h. After blocking, the
   membranes were incubated with the primary antibody at 4 °C overnight.
   The membranes were then washed with Tris-buffered saline (TBS)
   containing 0.1% Tween 20 three times for 10 min. The membrane was
   subsequently incubated with IRDye 680RD goat anti-mouse IgG (no.
   925–68070) or IRDye 800CW goat anti-rabbit IgG (no. 926–32211) (LI-COR;
   1:15,000 dilution) for 45 min at room temperature. Finally, the blots
   were visualized using an Odyssey CLx Infrared Imaging System (LI-COR).

Immunofluorescence

   HUVECs cultured in 35-mm glass-bottom dishes were fixed with 4% PFA for
   20 min and then washed three times with PBS. The cells were
   permeabilized with PBS containing 0.1% Triton X-100 for 20 min to
   facilitate membrane penetration. Afterwards, the cells were blocked
   with PBS containing 10% goat serum for 1.5 h, followed by incubation
   with rabbit anti-IGFBP6 and mouse anti-MVP primary antibodies (1:100)
   overnight at 4 °C. The cells were then washed again and incubated at
   room temperature for 1 h with secondary antibodies (Alexa Fluor goat
   anti-rabbit 546 (no. [278]A48254) and Alexa Fluor goat anti-mouse/rat
   488 (no. A48255), 1:1,000). After washing with PBS, DAPI staining
   solution (Beyotime Biotechnology) was added for 0.5 h. A laser scanning
   confocal microscope (Leica, TCS SP8 X) was used to capture the images.
   Immunofluorescence quantification was performed using ImageJ software.
   For immunofluorescence staining of mouse vascular tissue, after OCT
   removal with PBS for 10 min, the tissue was blocked with PBS containing
   10% goat serum for 1.5 h, and subsequent steps and cell staining were
   consistent. Immunostaining of murine aortic roots and human coronary
   arteries with isotype control IgG was performed as negative control
   ([279]Extended Data Fig. 10a,[280]b).

En face staining and RNA flushing from aortic endothelium

   After euthanasia, 12-week-old mice were perfused slowly for 5 min with
   saline containing 40 USPU ml^−1 heparin injected through the left
   ventricle, followed by 10 min with pre-cooled PBS containing 4% PFA (pH
   7.4). Subsequently, the aorta was dissected from surrounding adipose
   tissues, opened longitudinally before incubation with PBS containing
   0.1% Triton X-100 for 10 min and blocked with goat serum in TBS
   containing 2.5% Tween 20 for 1 h. Next, aortas were incubated with a
   1:100 dilution of VE-cadherin and IGFBP6 antibodies in dilution buffer
   (TBS containing 3% BSA and 2.5% Tween 20) at 4 °C overnight. After
   three rinses for 5 min with washing solution (TBS containing 2.5% Tween
   20), the aortas were incubated with Alexa Fluor 488 or 546 secondary
   antibodies (1:1,000 dilution) at room temperature for 1 h. The nuclei
   were stained with DAPI. Finally, after three more rinses in washing
   solution, the aorta was spread on a slide with the aortic endothelium
   facing the lens. A laser scanning confocal microscope (Leica, TCS SP8
   X) was used for imaging.

   In mice, RNA was extracted from ECs collected from the AA or the TA.
   Specifically, mice were anesthetized and euthanized; the left ventricle
   was perfused with PBS; and the aorta was separated after removal of
   adipose tissue. After washing with PBS, the aorta (tunica intima) was
   rinsed with 300 μl of TRIzol in a 1-ml syringe. The remaining part of
   the adventitia of the blood vessel was then placed into TRIzol to
   collect RNA (tunica media). RNA was isolated using the TRIzol method.

ELISA

   ELISA kits (Thermo Fisher Scientific, no. EMIGFBP6 and no. EHIGFBP6)
   were used to quantify IGFBP6 in mouse serum and cell supernatant as
   well as in human serum. IL-6 and IL-1β in human serum were quantified
   using kits (no. VAL102 and no. VAL101). ELISAs were carried out
   following the manufacturer’s protocol. In brief, samples were added to
   wells coated with antibodies and incubated at room temperature for 2 h.
   The solution was discarded, and the plate was washed four times.
   Detection antibody bound to the fixed target protein was added to each
   well and incubated at room temperature for 2 h. Excess detection
   antibody was washed off, followed by the addition of streptavidin-HRP
   and incubation at room temperature in the dark for 30 min. The plates
   were washed four times again and incubated for another 30 min at room
   temperature in the dark after adding substrate. Stop solution was then
   added, and the absorbance was read at 450 nm after thoroughly mixing
   the contents.

IGFBP6 promoter cloning and dual-luciferase reporter assay

   To investigate the precise relationship between IGFBP6 and KLF2,
   several plasmids were used, including the empty plasmid (basic-luc),
   and the plasmid was generated containing the IGFBP6 promoter (2,000-bp
   length before the transcription start site) cloned upstream from
   firefly luciferase (WT-luc). A version of the IGFBP6 promoter reporter
   was produced with deletion mutations of all CACCC KLF2 consensus sites
   (MT1-luc), and a version was produced with deletion mutations of all
   GGGTG sites (MT2-luc). All reporter plasmids were produced by
   Genomeditech. Primers for reporter and mutant genes are provided in
   [281]Supplementary Table 4. HEK293T cells were transfected with a
   mixture of luciferase reporter plasmid and control plasmid containing
   basally regulated renilla luciferase, and the cells were transfected
   with Lipofectamine 2000 (Thermo Fisher Scientific, no. 11668019) and
   cultured for 1 d. Then, the cells were collected and lysed, and the
   Dual-Luciferase Reporter Assay Kit (Vazyme, no. DL101–01) was tested by
   a multifunctional microplate reader (Molecular Devices, SpectraMaxiD3).
   Luciferase activity was measured using SoftMax Pro 7.1 software. The
   relative luciferase activity was determined by calculating the ratio of
   firefly luciferase to renilla luciferase values.

Chromatin immunoprecipitation

   The EZ-Magna ChIP A/G Chromatin Immunoprecipitation Kit (Sigma-Aldrich,
   17–10086) was used to prepare chromatin from Ad-KLF2-treated or
   Ad-NC-treated HUVECs. In brief, cells were fixed with 1% formaldehyde
   for 10 min to crosslink DNA to proteins, and the reaction was quenched
   with glycine. Next, the nuclei were lysed with nuclear lysis buffer and
   passed through an ultrasonic cell grinder (Ningbo Scientz
   Biotechnology, SCIENTZ-IID) to cleave chromatin into small 200–1,000-bp
   fragments. Immunoprecipitation with Flag antibody and Protein A/G beads
   was then performed by rotating incubation at 4 °C overnight. The next
   day, the magnetic beads were washed using a magnetic rack, followed by
   elution of chromatin and reverse crosslinking (incubation at 62 °C for
   2 h with shaking), followed by boiling at 95 °C for 10 min. Finally,
   the DNA was purified for PCR analysis. The IGFBP6 promoter was
   amplified by PCR using multiple primers, followed by nucleic acid gel
   electrophoresis. After gel extraction, the DNA was quantified by qPCR.
   Details of the primers used are provided in [282]Supplementary Table 5.

Analysis of plaque composition

   To determine the extent of atherosclerotic plaques in mice, mouse
   hearts that had been perfused from the left ventricle with PBS were
   collected and fixed in 4% PFA for 1 d before dehydration in 30% sucrose
   for 1 d. Subsequently, OCT-embedded hearts were sliced horizontally
   along the aortic axis in the direction of the AA. H&E and Massonʼs
   trichrome staining were performed to determine the necrotic core area
   of the aortic root and plaque collagen expression. ImageJ software was
   applied to determine the necrotic core area and collagen expression of
   the plaque, and this area divided by the plaque area was used to
   calculate the plaque lesions.

En face Oil Red O staining

   Mice were euthanized and aorta were isolated. Aorta were cut open in
   PBS under the dissection microscope to expose atherosclerotic plaques.
   After fixation with 4% PFA for 1 d, the mixture was removed and washed
   with PBS for 5 min. The mixture was mixed with 60% isopropanol (Sangon
   Biotech, no. 67–63-0), soaked for a short time and stained with freshly
   prepared filtered Oil Red O (Poly Scientific, no. s1849–32OZ) at 37 °C
   for 15 min. The solution was discarded and then washed with PBS after
   washing with 60% isopropanol. The en face aorta was mounted onto tissue
   slides with the intimal layer facing up. An MShot Image Analysis System
   (Mingmei Photoelectric Technology) recorded images. The plaque area was
   determined using ImageJ software, and the percentage of plaque area was
   calculated by the ratio of plaque area to total vessel area.

Conditioned medium preparation

   Conditioned media from ECs infected with Ad-IGFBP6 were collected.
   Media were concentrated using Macrosep Advance (Pall, no. MAP010C37).
   Co-immunoprecipitation was performed using the concentrated media.

Liquid chromatography with tandem mass spectrometry analysis

   Liquid chromatography with tandem mass spectrometry (LC–MS/MS) analysis
   was conducted using a timsTOF Pro mass spectrometer (Bruker) coupled
   with a nanoElute (Bruker). The peptides were loaded onto a C18
   reversed-phase analytical column (Thermo Fisher Scientific, Easy
   Column, 25 cm long, 75 μm inner diameter, 1.9 μm resin) in 95% buffer A
   (0.1% formic acid in water) and separated using a linear gradient of
   buffer B (99.9% acetonitrile and 0.1% formic acid) at a flow rate of
   300 nl min^−1. The mass spectrometer was operated in positive ion mode
   with an electrospray voltage of 1.5 kV. Precursors and fragments were
   analyzed using the TOF detector within a mass range of m/z 100–1,700.
   The timsTOF Pro was operated in parallel accumulation serial
   fragmentation (PASEF) mode, with PASEF mode data collection under the
   following conditions: ion mobility coefficient (1/KO) value was set
   from 0.6 Vs cm^−2 to 1.6 Vs cm^−2; one MS and 10 MS/MS PASEF scans were
   performed. Active exclusion was enabled with a release time of 24 s.

Molecular docking prediction

   In this study, the three-dimensionl structures of both IGFBP6 and MVP
   proteins were derived from predictive models generated by the AlphaFold
   3 online tool ([283]https://alphafoldserver.com/). Protein–protein
   molecular docking was performed using the online tool HDOCK SERVER,
   with detailed parameters referring to the provided example
   tutorials^[284]51. The Protein Interfaces, Surfaces and Assemblies
   (PISA) service, available at the European Bioinformatics Institute, can
   be accessed via the following link:
   [285]http://www.ebi.ac.uk/pdbe/prot_int/pistart.html. The final results
   were analyzed and visualized using PyMOL version 2.6 software.

Statistical analysis

   Data are presented as the mean ± s.d. The Shapiro–Wilk normality test
   was used to assess the distribution of the data. For comparisons
   between two groups with normally distributed data, Student’s t-test was
   applied when variances were equal, whereas Welch’s correction was used
   if the F-test indicated unequal standard deviations. For the
   experiments using HUVECs/HAECs from different donors, the two-sided
   paired t-test was adopted because of the large differences among
   different donors. Comparisons among three or more groups were conducted
   using one-way or two-way ANOVA followed by Tukey’s or Bonferroni’s
   multiple comparison test to evaluate differences among multiple groups.
   For data that were not normally distributed, non-parametric tests were
   used. The Mann–Whitney test was used for comparisons between two
   independent groups, and the Kruskal–Wallis test and Dunn’s multiple
   comparison test were used for comparisons among multiple groups. For
   experiments involving HUVECs/HAECs with different donors, one-way or
   two-way ANOVA (repeated measures) followed by Bonferroni’s test was
   used. Representative images were chosen to represent the group mean
   most accurately of all available data. P values less than 0.05 were
   considered statistically significant. Statistical analyses were
   performed using GraphPad Prism 10.0 (GraphPad Software) and SPSS
   version 27.0.

Extended Data

Extended Data Fig. 1 |. The gene expression values of IGFs and IGFBPs.

   Extended Data Fig. 1 |
   [286]Open in a new tab

   a, The mRNA expression counts of IGFs and IGFBPs in UF- and DF-treated
   HUVECs. (GEO accession No. [287]GSE20739, n = 3). b, A heatmap showing
   the differentially expressed genes (DEGs) between the two groups.
   Upregulated genes are labelled in red and downregulated genes are shown
   in blue. c, The mRNA expression of IGFBP1-7 in HUVECs under static
   conditions (n = 3). Statistical analysis was performed by two-tailed
   Welch’s t test for a, by one-way ANOVA followed by the Tukey’s test for
   c.

Extended Data Fig. 2 |. Construction and genotype identification of IGFBP6
global knockout mice.

   Extended Data Fig. 2 |
   [288]Open in a new tab

   a, Identification of genotypes of Igfbp6^−/− mice. b, The LDLR protein
   level in liver tissues of mice after PCSK9 injection as detected by
   Western blot. Liver tissues from male C57BL/6 J and Ldlr^−/− mice were
   used as controls. c, Analysis of ALT, AST, TG, TC, HDL-C, and LDL-C
   levels in male WT and Igfbp6^−/− mice (n = 13–15). Statistical analysis
   was performed by two-tailed Student’s t test, Mann–Whitney U test and
   Welch’s t test for c.

Extended Data Fig. 3 |. Construction of EC-specific IGFBP6 knockout mice,
genotype identification and serum biochemistry.

   Extended Data Fig. 3 |
   [289]Open in a new tab

   a, Schematic view of the generation of Igfbp6^ECKO mice. b,
   Identification of genotypes of Igfbp6^ECKO mice. c, The mRNA expression
   of Igfbps in the aortic intima of male Igfbp6^ECKO mice and control
   mice (n = 5). d, Analysis of body weight, blood glucose, ALT, AST, TG,
   TC, HDL-C, and LDL-C levels (n = 7–11). Statistical analysis was
   performed by two-tailed Student’s t test and Mann–Whitney U test for c,
   d, and by Welch’s t test for d (Male: LDL-C).

Extended Data Fig. 4 |. Validation and serum biochemistry of EC-specific
IGFBP6 overexpression mice.

   Extended Data Fig. 4 |
   [290]Open in a new tab

   a, En face immunofluorescence staining IGFBP6 (red), VE-Cadherin
   (green), and DAPI (blue) in mouse aorta, showing increased IGFBP6
   expression in the aortic endothelium of male AAV9-EC-Igfbp6 mice (n =
   3). Scale bar: 50 μm. b, The mRNA expression of Igfbps in the aortic
   intima of male AAV9-EC-Igfbp6 mice and control mice (n = 5). c,
   Analysis of body weight, blood glucose, ALT, AST, TG, TC, HDL-C, and
   LDL-C levels in mouse serum (n = 10). Statistical analysis was
   performed by two-tailed Mann–Whitney U test and by Student’s t test for
   b, c, and by Welch’s t test for c (ALT).

Extended Data Fig. 5 |. IGFBP6 suppresses inflammation in HAECs and
RNA-sequencing of IGFBP6 overexpressed ECs indicates the anti-inflammatory
effect of IGFBP6.

   Extended Data Fig. 5 |
   [291]Open in a new tab

   a and b, HAECs were transfected with control si-NC or si-IGFBP6 for 48
   h and then treated with TNF-α for 6 h. The expression of IGFBP6,
   VCAM-1, ICAM-1 was determined by qRT-PCR (a, n = 3) and Western blot
   (b, n = 4). c and d, HAECs were transfected with Ad-NC or Ad-IGFBP6 for
   24 h and then treated with TNF-α for 6 h. The expression of
   Flag-IGFBP6, VCAM-1, ICAM-1, SELE and CCL2 was determined by qRT-PCR
   (c) and Western blot (d) (n = 3). e and f, HAECs were treated as
   indicated and THP-1 monocyte adhesion assay was performed and the
   number of adherent monocytes were quantified (n = 3). Scale bar: 50 μm.
   g, HUVECs were transfected with Ad-NC or Ad-IGFBP6 for 24 h and then
   treated with TNF-α for 6 h. The treated cells were analyzed by
   transcriptome sequencing (n = 3). h, Bubble diagram of the KEGG pathway
   enrichment analysis of DEGs. i, Heatmap diagram showing the DEGs
   between the two groups. Upregulated genes are labelled in red and
   downregulated genes are shown in blue. Scale bar: 50 μm. Statistical
   analysis was performed by two-tailed paired t test for a (left panel),
   b, d, by two-way ANOVA (repeated measures) followed by the Bonferroni’s
   test for a (right panel), c, by Student’s t test for e, f.

Extended Data Fig. 6 |. KLF4 up-regulates the expression of IGFBP6 and MVP
depletion promotes inflammation in HUVECs.

   Extended Data Fig. 6 |
   [292]Open in a new tab

   a and b, HUVECs were treated with Ad-NC or Ad-KLF4 for 24 h, and the
   expression of KLF4, NOS3 and IGFBP6 was determined by qRT-PCR (a, n =
   5), and ELISA (b, n = 3). c, HUVECs were transfected with si-NC or
   si-MVP for 48 h and then treated with TNF-α for 6 h. The expression of
   MVP, ICAM-1and SELE was determined by qRT-PCR (n = 3). d, HUVECs were
   treated as indicated and THP-1 monocyte adhesion assay was performed
   and the number of adherent monocytes were quantified (n = 3). Scale
   bar: 50 μm. Statistical analysis was performed by two-tailed paired t
   test for a, by Student’s t test for b, by two-way ANOVA (repeated
   measures) followed by the Bonferroni’s test for c, and by two-way ANOVA
   followed by the Tukey’s test for d.

Extended Data Fig. 7 |. The anti-inflammatory effect of IGFBP6 depends on the
binding site of IGFBP6 to MVP.

   Extended Data Fig. 7 |
   [293]Open in a new tab

   a, In HUVECs, Co-IP experiments were performed on MVP and ASK1 (n = 3).
   b, HUVECs were treated with adenovirus to overexpress IGFBP6. After 42
   h of treatment, TNF-α was added for 6 h. After that, flag-tagged IGFBP6
   was pulled down and ASK1 (n = 3) was detected. c, Representative Co-IP
   and Western blot assays revealed dimerization of ASK1 in HEK293T cells
   transfected with indicated plasmid vectors (n = 3). d, Molecular
   docking was performed using Alphafold3 to predict the binding sites of
   IGFBP6 and MVP. e, Representative Co-IP and Western blot assays showing
   the binding domains of IGFBP6 to MVP in HEK293T cells (n = 3). f,
   HUVECs were treated with IGFBP6 and mutated IGFBP6 adenovirus for 42 h,
   followed by TNF-α for 6 h. The expression levels of Flag, VCAM-1,
   ICAM-1 were detected by Western blot (n = 3).

Extended Data Fig. 8 |. Secreted IGFBP6 and MVP do not affect inflammation in
HUVECs.

   Extended Data Fig. 8 |
   [294]Open in a new tab

   a and b, HUVECs were treated with adenoviruses (Ad-MVP and Ad-IGFBP6)
   for 7 days, during which the supernatant was continuously collected.
   Ultrafiltration tubes were used to concentrate the protein in the
   supernatant, and flag (a, n = 3) or HA (b, n = 3) antibodies were used
   for Co-IP, respectively, and finally verified by Western blot. c,
   HUVECs were treated with different concentrations of IGFBP6 recombinant
   protein for 42 h, and then treated with TNF-α for 6 h, and indicated
   proteins were detected by Western blot (n = 3). d, HUVECs were treated
   with different concentrations of IGFBP6 recombinant protein for 3 h,
   and then treated with TNF-α for 10 mins, and indicated proteins were
   detected by Western blot (n = 3). e, 80% of confluent HUVECs were
   treated with Ad-NC or Ad-IGFBP6 in the presence of TNF. After that,
   whole cell lysate was collected for Western blot (left 4 lanes,
   control). In parallel experiments, the treated cells were washed with
   PBS and new complete medium was supplemented. Then, conditioned media
   were collected to treat HUVECs before lysate was collected for Western
   blot (right 4 lanes) (n = 3). f, HUVECs were treated with different
   concentrations of MVP recombinant protein for 3 h, and then treated
   with TNF-α for 10 mins, and indicated proteins were detected by Western
   blot (n = 3).

Extended Data Fig. 9 |. Graphical Abstract.

   Extended Data Fig. 9 |
   [295]Open in a new tab

   IGFBP6 serves as a novel endothelial homeostasis-associated molecule
   which is mechanoresponsive and confers anti-inflammation and
   atheroprotection via binding with intracellular MVP and suppresses JNK
   and p65 phosphorylation.

Extended Data Fig. 10 |. Immunostaining of murine aortic roots and human
coronary arteries with isotype control IgG.

   Extended Data Fig. 10 |
   [296]Open in a new tab

   a and b, Representative images of immunofluorescence staining with rat
   IgG, rabbit IgG or mouse IgG for mouse and human plaque tissues (n =
   3). Scale bar = 50 μm.

Supplementary Material

   Supplemental Data
   [297]NIHMS2048285-supplement-Supplemental_Data.pdf^ (298.5KB, pdf)

Acknowledgements