Secretion of von Willebrand factor by endothelial ... - Semantic Scholar

6 downloads 0 Views 3MB Size Report
Apr 29, 2014 - Contributed by Maurice B. Burg, March 16, 2014 (sent for review ...... Shroff R, Hignett R, Pierce C, Marks S, van't Hoff W (2006) ... Adam SS, Key NS, Greenberg CS (2009) D-dimer antigen: Current concepts and future.
Secretion of von Willebrand factor by endothelial cells links sodium to hypercoagulability and thrombosis Natalia I. Dmitrieva1 and Maurice B. Burg1 Renal Cellular and Molecular Biology Section, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD 20892 Contributed by Maurice B. Burg, March 16, 2014 (sent for review September 20, 2013)

blood clotting

| HUVEC | osmotic stress

C

ardiovascular thrombotic events are a leading cause of mortality and morbidity worldwide (1–3). A critical step in thrombosis is inappropriate initiation of the coagulation cascade. The trigger is not always apparent. In some cases the triggering mechanism is vascular injury, similar to the initiation of hemostasis. For example, arterial thrombosis can be triggered by rupture of atherosclerotic plaques that leads to endothelial damage. However, venous thrombosis can also occur in the presence of an intact endothelium, making the triggering mechanism less apparent (1, 4, 5). von Willebrand factor (vWF) and platelets are the key components of blood that initiate clots. vWF is a large multimeric glycoprotein. It is produced by endothelial cells, stored in the form of ultralarge (UL) vWF multimers in Weibel–Palade bodies, then secreted by exocytosis of the Weibel–Palade bodies (6, 7). Upon secretion, UL vWF multimers are cleaved by ADAMTS13 protease into smaller and less hemostatically active multimers that circulate in blood. Clot formation initiates when platelets bind to specific binding sites on vWF through the platelet membrane glycoprotein Ib. While vWF is circulating in blood in the globular latent form, the binding sites are not exposed. A special elongated confirmation is necessary for binding. This conformation occurs, for example, after secretion of vWF from endothelial cells before cleavage by ADAMTS13 protease, or under high shear stress in stenosed arteries. Numerous coagulation factors are then activated, resulting in growth of blood clots (4, 5, 8–10). A number of population-based studies have

www.pnas.org/cgi/doi/10.1073/pnas.1404809111

suggested that elevated levels of vWF are a risk factor for thrombogenesis, especially in patients with previous cardiovascular diseases and older people (9). Hypernatremia has been defined as a rise in serum sodium to above 145 mmol/L (11). The most common cause of hypernatremia is reduced body water. Dehydration and hypernatremia are often accompanied by thrombosis. Thus, complications owing to thrombosis, such as disseminated intravascular coagulation, intracranial thrombosis, and peripheral thrombosis are commonly encountered during the hypernatremia associated with dehydration (12–14), the hyperosmolality associated with diabetes (15), and heatstroke (16). Increased sodium concentration is a common feature of the thrombosis in these studies. This led to recognition that hypernatremia might contribute to the thrombosis observed in such cases, but the mechanism had not been identified (17). In the present study we demonstrate that high NaCl increases production and secretion of vWF from endothelial cells in cell culture and in tissues of mice in vivo. In mice, the modest increase of blood sodium caused by mild water restriction raises endothelial vWF secretion sufficiently to increase coagulability of blood and induce thrombosis. Further, analysis of data from the Atherosclerosis Risk in Communities (ARIC) Study demonstrates that serum sodium is positively associated with the level of vWF in blood and the risk of stroke in humans. Our results identify a mechanism for hypernatremia-induced thrombosis and suggest that hydration and salt intake are modifiable factors that affect coagulability and thrombosis through high salt-dependent secretion of vWF from endothelial cells. Fig. S1 summarizes the findings and the proposed implications of our study. Significance Cardiovascular thrombotic events are a leading cause of mortality and morbidity worldwide. Hypercoagulability increases the risk of thrombosis. This study shows that elevated sodium concentration stimulates endothelial production of a key initiator of the clotting cascade, von Willebrand factor, leading to increased coagulation and thrombogenesis. In everyday life, increases in blood sodium often occur as the result of insufficient drinking, excessive water loss, or high salt intake. Therefore, our results indicate that water and salt intake are modifiable factors affecting coagulability and risk of thrombosis. In clinical practice, sustained elevation of plasma sodium is a part of widely used hypertonic saline therapy. Our results suggest that monitoring of coagulation during the therapy might be beneficial to prevent thrombotic complications. Author contributions: N.I.D. and M.B.B. designed research; N.I.D. performed research; N.I.D. analyzed data; and N.I.D. and M.B.B. wrote the paper. The authors declare no conflict of interest. 1

To whom correspondence may be addressed. E-mail: [email protected] or [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1404809111/-/DCSupplemental.

PNAS | April 29, 2014 | vol. 111 | no. 17 | 6485–6490

PHYSIOLOGY

Hypercoagulability increases risk of thrombi that cause cardiovascular events. Here we identify plasma sodium concentration as a factor that modulates blood coagulability by affecting the production of von Willebrand factor (vWF), a key initiator of the clotting cascade. We find that elevation of salt over a range from the lower end of what is normal in blood to the level of severe hypernatremia reversibly increases vWF mRNA in endothelial cells in culture and the rate of vWF secretion from them. The high NaCl increases expression of tonicity-regulated transcription factor NFAT5 and its binding to promoter of vWF gene, suggesting involvement of hypertonic signaling in vWF up-regulation. To elevate NaCl in vivo, we modeled mild dehydration, subjecting mice to water restriction (WR) by feeding them with gel food containing 30% water. Such WR elevates blood sodium from 145.1 ± 0.5 to 150.2 ± 1.3 mmol/L and activates hypertonic signaling, evidenced from increased expression of NFAT5 in tissues. WR increases vWF mRNA in liver and lung and raises vWF protein in blood. Immunostaining of liver revealed increased production of vWF protein by endothelium and increased number of microthrombi inside capillaries. WR also increases blood level of D-dimer, indicative of ongoing coagulation and thrombolysis. Multivariate regression analysis of clinical data from the Atherosclerosis Risk in Communities Study demonstrated that serum sodium significantly contributes to prediction of plasma vWF and risk of stroke. The results indicate that elevation of extracellular sodium within the physiological range raises vWF sufficiently to increase coagulability and risk of thrombosis.

Results High Salt-Induced vWF Secretion: Human Umbilical Vein Endothelial Cells. vWF is critical for the development of venous thrombosis

when blood flow is restricted by stenosis in mice (18). The thrombosis depends on the level of vWF. Thus, vWF knock-out mice do not develop the thrombi, and mice that have low vWF have reduced thrombosis. We hypothesized that the thrombosis that occurs during hypernatremia might also be caused by increased vWF secretion, albeit from intact endothelial cells, in response to the elevated NaCl in blood. To test whether high NaCl affects production of vWF, we exposed human umbilical vein endothelial cells (HUVECs) in culture to the range of NaCl that occurs in humans during hypernatremia. We added up to 55 mmol/L of NaCl, elevating the medium osmolality from 270 mosmol/kg (lower end of the normal physiological range) to up to 380 mosmol/kg (severe hypernatremia). HUVECs adapt well up to at least 380 mosmol/kg, maintaining a normal appearance (Fig. S2A), and logarithmic growth (Fig. 1A), for weeks. High NaCl reversibly increases secretion of vWF (Fig. 1B), and the rate of secretion remains elevated for up to 5 wk, provided that NaCl remains elevated (Fig. S2B). Immunofluorescent staining demonstrates that HUVECs secrete long strings of vWF multimers while exposed to high NaCl, but, when not exposed to high NaCl, retain most of their vWF compactly stored in Weibel–Palade bodies (Fig. 1C). When NaCl is elevated, the level of intracellular vWF does not change much (Fig. 1D), indicating that the newly synthesized vWF is secreted nearly as fast as it is made. The rate of secretion returns to the basal level soon after NaCl is lowered, demonstrating that the salt-dependent increase is reversible (Fig. 1B). High NaCl also reversibly elevates vWF mRNA, consistent with increased transcription (Fig. 1F). Having found a salt-dependent increase of vWF secretion from HUVECs, we questioned whether high salt produces the same effect on endothelial cells in vivo. The kidney was a convenient organ to start with because of the wide range of NaCl normally present in the different parts of the kidney. The level of NaCl in blood perfusing the cortex of the kidney is the same as that in systemic blood, whereas blood perfusing the renal medulla normally contains very high NaCl. In fact, vWF is higher in inner and outer renal medullas of mice than it is in the renal cortex (Fig. 1E and Fig. S2C). This finding supports the conclusion that a high salt-dependent increase of vWF production by endothelial cells occurs in vivo, as well as in cell culture. Role of the Osmotically Regulated Transcription Factor NFAT5 in High Salt-Induced vWF Secretion. NaCl is a functionally impermeant

solute, so elevation of extracellular NaCl is hypertonic. High NaCl causes osmotic efflux of water from cells, leading to decreased cell volume and increased intracellular ionic strength. Adaptive cellular responses are activated to compensate for the dehydration and its consequences. NFAT5 is the master transcription factor that is activated by hypertonicity. Hypertonicity increases expression of NFAT5 mRNA and protein, increases NFAT5 transcriptional and transactivating activities, and mediates transcription of many NFAT5 target genes that are directly or indirectly involved in adaptation to high NaCl (19). Therefore, we tested whether vWF is an NFAT5 target gene in HUVECs. High NaCl elevates vWF mRNA, which is consistent with increased transcription (Fig. 1F). The vWF promoter contains an osmotic response element (20) (ORE, NFAT5 binding site) close to the transcription start site (Fig. 2D). In HUVECs, high NaCl increases NFAT5 mRNA (Fig. 2C) and NFAT5 protein (Fig. 2A) in correlation with the salt-dependent increases of vWF mRNA and protein secretion (Fig. 1 B and F). In addition, the NFAT5 target, aldose reductase, also increases (Fig. 2B), consistent with up-regulation of NFAT5 transcriptional activity. Finally, ChIP analysis confirms binding of NFAT5 to the ORE in 6486 | www.pnas.org/cgi/doi/10.1073/pnas.1404809111

Fig. 1. High NaCl increases production and secretion of vWF from vascular endothelial cells. (A–D and F) HUVECs were exposed to media in which NaCl was elevated for 4 d to the total osmolality indicated in the figure panels. (A) HUVECs adapt to the range of elevated NaCl that occurs in hypernatremia, maintaining logarithmic growth (see also Fig. S2A for images of the cells). (B) High NaCl increases secretion of vWF, but when the elevated NaCl is lowered to the control level (270 mosmol/kg) for 2 d vWF secretion returns to its control level (mean ± SEM, *P < 0.05, t test, n = 5, linearly dependent on NaCl concentration, P = 0.0009). (C) At 270 mosmol/kg vWF multimers are compactly stored in Weibel–Palade bodies (arrow), but when NaCl is increased for 4 d the cells release long fibers of uncondensed vWF multimers (arrow). Green, vWF; blue, nuclei stained with DAPI. The lower panel is a higher magnification of the upper panel. (D) High NaCl has little effect on the level of intracellular vWF protein. (Upper) Representative Western blot image. (Lower) Quantification, relative to 270 mosmol/kg, normalized to tubulin (mean ± SEM, *P < 0.05, t test, n = 3). (F) High NaCl increases vWF mRNA in HUVECs. vWF mRNA level returns to the basal when NaCl is lowered for 2 d (mean ± SEM, *P < 0.05, t test, n = 5, linearly dependent on NaCl concentration, P < 0.0001). (E) vWF protein is higher in the kidney medulla, where interstitial NaCl is very high, than in kidney cortex, where the interstitial concentration of NaCl is similar to that in systemic blood. Immunohistochemical staining for vWF (brown). See also Fig. S2C for colocalization with CD31staining (endothelial cell marker).

the vWF gene promoter in proportion to the level of NaCl (Fig. 2D). These results indicate that high NaCl-induced increase of NFAT5 activity contributes to the increased vWF production in endothelial cells. Dehydration and Hypertonic Signaling in Mice. We next tested whether hypernatremia results in increased production of vWF in endothelial cells in vivo. We chose dehydration as a model to increase NaCl in vivo. Dehydration is defined as net loss of body water resulting from decreased water intake or increased water loss. Dehydration leads to elevated osmolality of plasma and other extracellular fluids (21). We controlled the amount of water that the mice consumed by feeding them with gel food containing 30% water as their only Dmitrieva and Burg

source of water intake or, as a control, feeding the same food, but with free access to drinking water (Fig. 3A). To assess the degree of dehydration, we measured urine osmolality, body weight, serum sodium, serum osmolality, and plasma protein concentration. Water restriction increased urine osmolality, indicating activation of the renal concentrating mechanism to conserve water (Fig. 3B). The water-restricted mice did not lose weight, but their growth was retarded, so they weighed about 2% less than control mice by the end of 9 d of water restriction (Fig. 3E). Plasma protein concentration was unchanged (Fig. 3D). The dehydration of our water-restricted mice is of a magnitude recognized to be quite mild (22). Still, serum sodium and serum osmolality increased (Fig. 3C). Thus, our model produced a small increase of plasma sodium within the physiological range. Despite the fact that the increase in serum sodium and osmolality (Fig. 3C and Table S1) were very modest, we detected activation of a hypertonic response in several tissues, as evidenced by increased expression of NFAT5 and its transcriptional target, aldose reductase (Fig. 3F). Dehydration, vWF Secretion, and Thrombosis in Mice. In HUVECs, high NaCl increases vWF production (Figs. 1 and 2). To see Dmitrieva and Burg

Positive Association of Serum Sodium Concentration with Blood Level of vWF and Stroke Risk in Humans. To access the relevance of

our findings to humans, we analyzed whether serum sodium and vWF have a positive association in humans, using data from the ARIC Study. ARIC is a study of cardiovascular disease in a cohort of 15,792 45- to 64-y-old persons sampled from four US communities in 1987–1989 (24). We used the results of the baseline clinical examination of the participants during their first visit. To evaluate effect of Na on vWF, we conducted multiple regression analysis with the following predictor variables: serum sodium, glucose, and estimated glomerular filtration rate (eGFR) (Fig. 6A). See SI Materials and Methods, Fig. S5, and Tables S2–S7 for the reasons those variables were selected, information about the basic descriptive statistics for the variables, zero-order correlations between the variables, and regression coefficients. The overall model is statistically significant [F(3, 14,675) = 210, P < 0.001] with all variables significantly contributing to predicting plasma level of vWF (P < 0.001) (Fig. 6A and Tables S4 and S5). The positive regression coefficient for serum Na+ indicates that the increase in Na+ is accompanied by increased vWF, consistent with overall findings of our study. Additionally, a 3D plot of blood vWF level vs. serum sodium concentration and age demonstrates that higher levels of vWF occur in participants with higher concentrations of serum sodium (Fig. 6B). The same is true for 10-y stroke risk (Fig. 6C). The stroke risk at first visit was retrospectively calculated for every participant of the ARIC Study based on the study outcomes and included in ARIC datasets (25). Multivariable regression analysis of 10-y stroke risk with sodium, glucose, and eGFR as predictor variables demonstrates that plasma sodium significantly contributes to prediction of stroke (Table S6). Diabetes and chronic kidney disease (CKD) lead to increased PNAS | April 29, 2014 | vol. 111 | no. 17 | 6487

PHYSIOLOGY

Fig. 2. (A–C) High NaCl-induced secretion of vWF from HUVECs is accompanied by increased expression of the osmoregulated transcription factor NFAT5. HUVECs were exposed to high NaCl, as in Fig 1. (A) High NaCl increases NFAT5 protein. (Upper) Representative Western blot. (Lower) Quantification, relative to 270 mosmol/kg, normalized to tubulin (mean ± SEM, *P < 0.05, t test, n = 4, linearly dependent on NaCl concentration, P < 0.0001). (B) High NaCl increases expression of aldose reductase (AR), which is a transcriptional target of NFAT5. Western blot, analyzed as in A (mean ± SEM, *P < 0.05, t test, n = 3, linearly dependent on NaCl concentration, P = 0.006). (C) High NaCl increases NFAT5 mRNA. Quantification is relative to 270 mosmol/kg (mean ± SEM, *P < 0.05, t test, n = 5, linearly dependent on NaCl concentration, P < 0.0001). The NFAT5 mRNA returns to the basal level when NaCl is lowered for 2 d. (D) High NaCl increases binding of NFAT5 to the DNA element that is an NFAT5 binding site in the promoter of the vWF gene. (Upper) Diagram showing location of the NFAT5 binding site upstream of transcription start site (TSS) of the vWF gene, and the positions of primers that were used to analyze NFAT5 binding by ChIP. (Lower) ChIP results, relative to 270 mosmol/kg (mean ± SEM, *P < 0.05, t test, n = 3).

whether mild dehydration, caused by water restriction, affects production of vWF by endothelial cells in mouse tissues, we analyzed vWF mRNA and protein in tissues of water-restricted mice. Mild water restriction increases vWF mRNA significantly in the liver and lungs and also increases it in some other tissues, but not to a statistically significant degree (Fig. 4A). We measured vWF protein in endothelial cells by immunohistochemistry of the liver tissue sections. The mild water restriction significantly increases vWF in endothelial cells of the liver (Fig. 4 B and C and Fig. S3). vWF apparently rises in endothelial cells, as evidenced by correlation of staining for vWF with staining for the endothelial cell marker CD31 (Fig. S4). We hypothesized that dehydration-induced increase of secretion of vWF from endothelial cells in vivo might elevate vWF in blood and promote thrombosis. To assess the effect of water restriction on coagulation, we analyzed some of the factors depicted on Fig. 5A. Interaction of vWF with platelets activates thrombin and converts soluble plasma fibrinogen into insoluble, cross-linked fibrin polymers that stabilize blood clots (thrombi) (5, 10). Concurrently, clots are degraded by activation of other factors, principally the protease plasmin. Degradation of the fibrin polymers by plasmin leads to the appearance of fibrin degradation products (FDPs) in blood. D-dimer is the FDP whose level is used as a clinical indicator of ongoing coagulation. D-dimer increases in thrombotic conditions such as disseminated intravascular coagulation, deep venous thrombosis, and pulmonary embolism (5, 10, 23). We analyzed the levels of vWF and D-dimer in mouse blood and the number of microthrombi in the liver. Water restriction increases vWF and D-dimer in plasma (Fig. 5B), as well as the number of microthrombi in capillaries of the liver (Fig. 5 D and E). In addition, plasminogen activator inhibitor 1 (PAI-1) also increases (Fig. 5C), which limits the activation of plasmin and favors thrombosis. Thus, mild water restriction increases endothelial cell secretion of vWF enough to elevate the level of vWF in blood and to promote thrombosis in mice.

Fig. 3. Water restriction elevates serum sodium and activates hypertonic signaling in mouse tissues. To elevate NaCl in vivo, mice were subjected to water restriction for 9 d. (A–E) The water restriction produced mild subclinical dehydration (mean ± SEM, n = 5, *P < 0.05, t test). (A) Experiment design. To limit the amount of water, mice were fed with gel food containing 30% water and were not given any additional water. Control group were fed the same gel food but had free access to water. (B) Water restriction increased urine osmolality. (C) Water restriction increased serum sodium (by 5 mmol/L) and serum osmolality. See also Table S1 for other serum parameters. (D) Water restriction did not change plasma protein concentration. (E) Water restriction retarded growth but did not acutely reduce weight. (F) Water restriction increased mRNA of NFAT5 and of its transcriptional target, aldose reductase (AR) in several mouse tissues.

levels of VWF (26, 27). Consistently, there is significant zero-order correlation of plasma glucose and eGFR with level of vWF (Table S3). To assess whether a positive association of plasma sodium with vWF and risk of stroke is still present in participants without diabetes and CKD, we performed the analysis

Fig. 4. Water restriction increases vWF in endothelial cells of mice (mean ± SEM, n = 5, *P < 0.05, t test). Water was restricted for 9 d. (A) vWF mRNA increased in liver and lung. (B and C) Water restriction increased vWF protein in endothelial cells in the liver. (B) Representative images from immunohistochemical staining for vWF protein (brown) in the liver tissue sections. See Fig. S4 for pattern of blood capillaries in the liver (same sections stained for endothelial cells marker CD31). (C) Quantification of vWF in tissue sections. See Materials and Methods and Fig. S3 for details about image analysis.

6488 | www.pnas.org/cgi/doi/10.1073/pnas.1404809111

on a cohort of participants without diabetes and without CKD (eGFR >60 mL·min−1·1.73 m−2) who had normal weight [body mass index (BMI) = 18.5–25 kg/m2]. In this “healthy” cohort (n = 3,345), higher levels of vWF and risk of stroke occur also in participants who have higher concentrations of plasma sodium (Fig. 6D and Table S7). In summary, the analyses indicate that serum sodium is positively associated with the level of vWF and the risk of stroke in humans. Discussion A model that describes the link between sodium, vWF, and thrombosis is shown on Fig. S1. In this study we have shown that elevation of extracellular NaCl stimulates production of vWF by vascular endothelial cells both in cell culture (HUVECs) and in mice in vivo. In HUVECs, secretion of vWF and activation of the osmoregulated transcription factor NFAT5 increase in parallel when NaCl is added to elevate osmolality of the medium over the range from 270 mosmol/kg, which is the lower end of the normal physiological range, to 380 mosmol/kg, which is the high end of severe hypernatremia. In mice, elevation of serum sodium by 5 mmol/L as the result of water restriction increases expression of NFAT5 and stimulates endothelial vWF production in several tissues. The mild dehydration increases vWF secretion enough to elevate vWF significantly in systemic blood and to cause thrombosis, as evidenced by the appearance of microthrombi and activation of fibrinolysis. Further, blood sodium and vWF are positively associated in humans. Our findings identify the mechanism probably involved in the thrombosis seen in cases of hypernatremic dehydration (12–14), hyperosmolality associated with diabetes (15), and heatstroke (16). In clinical practice, infusion of hypertonic saline to elevate plasma sodium up to 155 mmol/L, and even higher, is routinely used in head injury, ischemic stroke, and intracerebral hemorrhage and is the subject of heated debate (28, 29). The results of our study indicate that elevation of plasma sodium during such therapy could increase blood coagulability, leading to thrombotic complications. There are also broader implications of our study, as follows. Inadequate water intake increases plasma sodium concentration (21, 30), as does the high consumption of salt that is prevalent Dmitrieva and Burg

worldwide (31). Thus, 400 mL of soup containing 6 g of salt elevates plasma sodium for many hours by about 2 mmol/L (32), and increasing salt intake from 600 mg to 10 g per day for several days elevates plasma sodium by about 3 mmol/L (33). How much salt and water should be consumed, and how the consumption affects health, are issues that are the focus of intensive research and discussion (30, 34, 35). Perhaps monitoring the level of vWF or other coagulation markers could enhance investigation of these issues. Subclinical dehydration may impair health because of increased vWF. Old age predisposes one to dehydration because of decreased thirst, impaired urinary concentrating ability, and increased insensible water loss (36, 37). Diabetes mellitus predisposes one to dehydration because of the osmotic diuresis caused by high urinary glucose. Further investigation is required to determine whether chronic subclinical dehydration in the elderly and in diabetics is the reason for elevated vWF and increased risk of thrombosis. Do dehydration and hypernatremia cause hypercoagulability in humans? There are previous indications that they do. Namely, increased circulating vWF and risk of thrombosis were found to be associated with diabetes (26, 38), which predisposes one to dehydration; small clinical trials demonstrated association of platelet aggregation (39), plasma fibrinogen, D-dimer, and vWF (40) with high salt intake, and the 200-fold increase in deep vein thrombosis in shelters in Japan after the 2011 earthquake was attributed to dehydration (41). Is health in humans likely to be adversely affected by hypercoagulability and increased vWF induced by subclinical Dmitrieva and Burg

Fig. 6. Plasma sodium is positively associated with blood level of vWF and 10-y risk of stroke in the ARIC Study. (A) Multiple linear regression analysis was used to assess effect of serum sodium on level of vWF. The results demonstrate that serum Na+ as well as glucose and eGFR significantly contribute to predicting the level of vWF. See SI Materials and Methods, Fig. S5, and Tables S2–S5 for details of the analysis. (B–D) Three-dimensional mesh plots, visualizing level of vWF and risk of stroke, as functions of serum sodium concentration and age. (B and D) Level of vWF is higher in participants with higher levels of serum sodium, both in B [all ARIC participants (n = 14,679)] and in D [cohort of 3,345 participants without diabetes who had eGFR >60 mL·min−1·1.73 m−2 and normal weight (BMI = 18.5–25 kg/m2)]. (C) Ten-year risk of stroke is increased in participants with higher level of serum sodium. All ARIC participants were included in the analysis. See also Tables S6 and S7.

PNAS | April 29, 2014 | vol. 111 | no. 17 | 6489

PHYSIOLOGY

Fig. 5. Water restriction increases vWF in the blood of mice and activates thrombogenesis (mean ± SEM, n = 5, *P < 0.05, t test). Water was restricted for 9 d. (A) Overview of blood clotting. Increase of vWF causes platelet activation and aggregation, leading to coagulation and formation of fibrin clots (thrombi). Concurrent activation of fibrinolysis begins degrading clots, which increases D-dimer in blood. (Note that measurement of D-dimer is also a clinical test for thrombogenesis.) (B) Water restriction increases vWF and D-dimer in plasma of mice. (Upper) Western blots. (Lower) Quantification, normalized to control. (C) Water restriction increases PAI-1 in plasma of mice. Note that PAI-1 inhibits fibrinolysis, which delays degradation of fibrin clots. (D) Representative images of immunohistochemical staining (brown) of liver tissue sections for fibrin and for the endothelial cell marker CD31. The staining for fibrin identifies microthrombi (arrows) inside liver capillaries. (E) Quantification of microthrombi in tissue sections from liver. Water restriction increases the number of microthrombi per square millimeter.

dehydration? That seems likely, considering that hypercoagulability is linked to cardiovascular disease and atherosclerosis (42), and increased vWF is a risk factor for thromboembolism, myocardial infarction, and stroke (7, 26, 43). Do other factors besides elevated NaCl contribute to the increased blood vWF and coagulability induced by water restriction? In addition to elevating blood sodium, water restriction increases antidiuretic hormone, and antidiuretic hormone increases vWF in blood (44). Infusion of the antidiuretic hormone analog desmopressin (DDAVP) is used clinically to treat bleeding disorders by elevating vWF (45). However, the increased production of vWF caused by high NaCl in HUVECs occurs independent of any increase of antidiuretic hormone (Fig. 1). In addition, the therapeutic use of DDAVP involves much greater antidiuretic hormone activity than the maximal level reached during dehydration (17, 45, 46). Furthermore, DDAVP produces only a transient increase of vWF in blood because it causes release of preexisting intracellular stores of vWF, whereas we observe a prolonged transcriptionally regulated increase in mice. Therefore, although the mild water restriction that we used undoubtedly elevated antidiuretic hormone, that increase does not explain the prolonged increase of vWF that we observe. Increased blood viscosity and hemoconcentration as a result of dehydration might also contribute to increased coagulability and thrombogenesis. However, our water restriction protocol does not cause acute weight loss (Fig. 3E) or increased plasma protein concentration (Fig. 3D), which speaks against these possibilities. In summary, we find that high salt-induced secretion of vWF from endothelial cells causes hypercoagulability, which could

induce thrombosis in conditions associated with hypernatremia, including dehydration, old age, diabetes, high dietary salt intake, and hyperosmotic therapy. We propose that hydration and salt intake are modifiable factors that affect coagulability and thrombogenesis. Materials and Methods

Blood was collected from the tails. Total plasma protein was measured using BCA Protein Assay (Pierce). vWF and D-dimer in plasma were measured by Western blot. PAI-1 in plasma was measured using Luminex immunoassay technique (48). PAI-1 agarose beads and assay reagents were obtained from EMD Millipore. Paraffin embedding of mouse tissues and immunohistochemical staining were performed as described (49). To quantify vWF protein expression on the diaminobenzidine-stained sections, we used the yellow-CMYK channel method (50). The Python script that performs yellowCMYK channel extraction and quantification is available upon request. Further details of the study methods are provided in SI Materials and Methods.

HUVEC cells and growth medium were obtained from ATCC. Osmolality of this medium (control medium) was 270 mosmol/kg. High NaCl medium was prepared by adding NaCl to the total osmolality of 290–380 mosmol/kg. vWF was measured in HUVEC supernatants using the vWF Human ELISA Kit (Abcam). HUVECs protein extraction, Western blot, immunofluorescent staining, and ChIP were performed as described (47). Target sequences in ChIPed DNA were quantified by real-time PCR with SYBR-Green PCR Kit (Qiagen). RNA was extracted from HUVECs with RNeasy Mini Kit (Qiagen) and from mouse tissues using AllPrep DNA/RNA Mini Kit (QIAGEN). mRNA was quantified by quantitative RT-PCR using TaqMan gene expression assays (Life Technologies). All mouse studies were done under approval of the National Heart, Lung, and Blood Institute animal study protocols. Mice were water-restricted by feeding them with gel food containing 30% water.

ACKNOWLEDGMENTS. We thank Daniil A. Kitchaev for writing Python script for CMYK yellow channel extraction from immunohistochemistry images, Daniela Malide and Christian Combs at the National Heart, Lung, and Blood Institute (NHLBI) Light Microscopy Core facility for help with microscopy, Leigh Samsel at NHLBI Flow Cytometry Core Facility for help with Luminex analysis, and Eric S. Leifer at the NHLBI Office of Biostatistics Research for expert advice on the clinical data analysis. This manuscript was prepared using Atherosclerosis Risk in Communities research materials obtained from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center. The study was supported by the Intramural Program of NHLBI.

1. Goldhaber SZ, Bounameaux H (2012) Pulmonary embolism and deep vein thrombosis. Lancet 379(9828):1835–1846. 2. Roger VL, et al.; American Heart Association Statistics Committee and Stroke Statistics Subcommittee (2011) Heart disease and stroke statistics—2011 update: A report from the American Heart Association. Circulation 123(4):e18–e209. 3. Mackman N (2008) Triggers, targets and treatments for thrombosis. Nature 451(7181): 914–918. 4. Engelmann B, Massberg S (2013) Thrombosis as an intravascular effector of innate immunity. Nat Rev Immunol 13(1):34–45. 5. Reininger AJ (2008) Function of von Willebrand factor in haemostasis and thrombosis. Haemophilia 14(Suppl 5):11–26. 6. Michaux G, Cutler DF (2004) How to roll an endothelial cigar: The biogenesis of Weibel-Palade bodies. Traffic 5(2):69–78. 7. Sadler JE (2005) New concepts in von Willebrand disease. Annu Rev Med 56:173–191. 8. Chen JM, López JA (2005) Interactions of platelets with subendothelium and endothelium. Microcirculation 12(3):235–246. 9. Denis CV, Lenting PJ (2012) von Willebrand factor: At the crossroads of bleeding and thrombosis. Int J Hematol 95(4):353–361. 10. Versteeg HH, Heemskerk JWM, Levi M, Reitsma PH (2013) New fundamentals in hemostasis. Physiol Rev 93(1):327–358. 11. Adrogué HJ, Madias NE (2000) Hypernatremia. N Engl J Med 342(20):1493–1499. 12. van Amerongen RH, Moretta AC, Gaeta TJ (2001) Severe hypernatremic dehydration and death in a breast-fed infant. Pediatr Emerg Care 17(3):175–180. 13. Hbibi M, et al. (2012) Severe hypernatremic dehydration associated with cerebral venous and aortic thrombosis in the neonatal period. BMJ Case Rep, 10.1136/ bcr.07.2011.4426. 14. Shroff R, Hignett R, Pierce C, Marks S, van’t Hoff W (2006) Life-threatening hypernatraemic dehydration in breastfed babies. Arch Dis Child 91(12):1025–1026. 15. Keenan CR, Murin S, White RH (2007) High risk for venous thromboembolism in diabetics with hyperosmolar state: Comparison with other acute medical illnesses. J Thromb Haemost 5(6):1185–1190. 16. Bouchama A, et al. (2012) Tissue factor/factor VIIa pathway mediates coagulation activation in induced-heat stroke in the baboon. Crit Care Med 40(4):1229–1236. 17. Grant PJ, Tate GM, Hughes JR, Davies JA, Prentice CR (1985) Does hypernatraemia promote thrombosis? Thromb Res 40(3):393–399. 18. Brill A, et al. (2011) von Willebrand factor-mediated platelet adhesion is critical for deep vein thrombosis in mouse models. Blood 117(4):1400–1407. 19. Burg MB, Ferraris JD, Dmitrieva NI (2007) Cellular response to hyperosmotic stresses. Physiol Rev 87(4):1441–1474. 20. Ferraris JD, Williams CK, Ohtaka A, García-Pérez A (1999) Functional consensus for mammalian osmotic response elements. Am J Physiol 276(3 Pt 1):C667–C673. 21. Senay LC, Christensen ML (1965) Changes in blood plasma during progressive dehydration. J Appl Physiol 20(6):1136. 22. Cheuvront SN, Kenefick RW, Charkoudian N, Sawka MN (2013) Physiologic basis for understanding quantitative dehydration assessment. Am J Clin Nutr 97(3):455–462. 23. Adam SS, Key NS, Greenberg CS (2009) D-dimer antigen: Current concepts and future prospects. Blood 113(13):2878–2887. 24. Williams OD (1989) The Atherosclerosis Risk in Communities (ARIC) Study: Design and objectives. The ARIC investigators. Am J Epidemiol 129(4):687–702. 25. Chambless LE, Heiss G, Shahar E, Earp MJ, Toole J (2004) Prediction of ischemic stroke risk in the Atherosclerosis Risk in Communities Study. Am J Epidemiol 160(3):259–269. 26. Blann AD (2006) Plasma von Willebrand factor, thrombosis, and the endothelium: The first 30 years. Thromb Haemost 95(1):49–55. 27. Shen L, et al. (2012) Von Willebrand factor, ADAMTS13 activity, TNF-α and their relationships in patients with chronic kidney disease. Exp Ther Med 3(3):530–534.

28. Diringer MN (2013) New trends in hyperosmolar therapy? Curr Opin Crit Care 19(2): 77–82. 29. Grände PO, Romner B (2012) Osmotherapy in brain edema: A questionable therapy. J Neurosurg Anesthesiol 24(4):407–412. 30. Armstrong LE (2012) Challenges of linking chronic dehydration and fluid consumption to health outcomes. Nutr Rev 70(Suppl 2):S121–S127. 31. Brown IJ, Tzoulaki I, Candeias V, Elliott P (2009) Salt intakes around the world: implications for public health. Int J Epidemiol 38(3):791–813. 32. Suckling RJ, He FJ, Markandu ND, MacGregor GA (2012) Dietary salt influences postprandial plasma sodium concentration and systolic blood pressure. Kidney Int 81(4):407–411. 33. He FJ, Markandu ND, Sagnella GA, de Wardener HE, MacGregor GA (2005) Plasma sodium: Ignored and underestimated. Hypertension 45(1):98–102. 34. He FJ, Macgregor GA (2012) Salt intake, plasma sodium, and worldwide salt reduction. Ann Med 44(Suppl 1):S127–S137. 35. Valtin H (2002) “Drink at least eight glasses of water a day.” Really? Is there scientific evidence for “8 x 8”? Am J Physiol Regul Integr Comp Physiol 283(5):R993–R1004. 36. Cowen LE, Hodak SP, Verbalis JG (2013) Age-associated abnormalities of water homeostasis. Endocrinol Metab Clin North Am 42(2):349–370. 37. Dmitrieva NI, Burg MB (2011) Increased insensible water loss contributes to aging related dehydration. PLoS ONE 6(5):e20691. 38. Carr ME (2001) Diabetes mellitus: a hypercoagulable state. J Diabetes Complications 15(1):44–54. 39. Nara Y, et al. (1984) Dietary effect on platelet aggregation in men with and without a family history of essential hypertension. Hypertension 6(3):339–343. 40. Liu F, et al. (2011) Potassium supplement ameliorates salt-induced haemostatic abnormalities in normotensive subjects. Acta Cardiol 66(5):635–639. 41. Ueda S, Hanzawa K, Shibata M, Suzuki S (2012) High prevalence of deep vein thrombosis in tsunami-flooded shelters established after the great East-Japan earthquake. Tohoku J Exp Med 227(3):199–202. 42. Loeffen R, Spronk HMH, ten Cate H (2012) The impact of blood coagulability on atherosclerosis and cardiovascular disease. J Thromb Haemost 10(7):1207–1216. 43. Wiman B, et al. (2000) Plasma levels of tissue plasminogen activator/plasminogen activator inhibitor-1 complex and von Willebrand factor are significant risk markers for recurrent myocardial infarction in the Stockholm Heart Epidemiology Program (SHEEP) study. Arterioscler Thromb Vasc Biol 20(8):2019–2023. 44. Nussey SS, Bevan DH, Ang VTY, Jenkins JS (1986) Effects of arginine vasopressin (AVP) infusions on circulating concentrations of platelet AVP, factor VIII: C and von Willebrand factor. Thromb Haemost 55(1):34–36. 45. Mannucci PM (1997) Desmopressin (DDAVP) in the treatment of bleeding disorders: the first 20 years. Blood 90(7):2515–2521. 46. Grant PJ, Davies JA, Tate GM, Boothby M, Prentice CR (1985) Effects of physiological concentrations of vasopressin on haemostatic function in man. Clin Sci (Lond) 69(4): 471–476. 47. Dmitrieva NI, Cui K, Kitchaev DA, Zhao K, Burg MB (2011) DNA double-strand breaks induced by high NaCl occur predominantly in gene deserts. Proc Natl Acad Sci USA 108(51):20796–20801. 48. Khan SS, Smith MS, Reda D, Suffredini AF, McCoy JP, Jr. (2004) Multiplex bead array assays for detection of soluble cytokines: Comparisons of sensitivity and quantitative values among kits from multiple manufacturers. Cytometry B Clin Cytom 61(1):35–39. 49. Dmitrieva NI, Burg MB (2007) High NaCl promotes cellular senescence. Cell Cycle 6(24):3108–3113. 50. Pham N-A, et al. (2007) Quantitative image analysis of immunohistochemical stains using a CMYK color model. Diagn Pathol 2:8.

6490 | www.pnas.org/cgi/doi/10.1073/pnas.1404809111

Dmitrieva and Burg

Supporting Information Dmitrieva and Burg 10.1073/pnas.1404809111 SI Materials and Methods Cell Culture. Primary human umbilical vein endothelial cells (HUVECs) (PCS-100–010; ATCC) were grown in Vascular Cell Basal Medium (PCS-100–030; ATCC) supplemented with Endothelial Cell Growth Kit-BBE (PCS-100–040; ATCC) and antibiotics (PCS-999–002; ATCC). Osmolality of this medium (control medium) was 270 mosmol/kg. High-NaCl medium was prepared by adding NaCl to the total osmolality of 290–380 mosmol/kg. All experiments were performed on logarithmically growing cells at about 80% confluence. To elevate NaCl, control medium was replaced by the high-NaCl medium. For proliferation curves shown on Fig. 1A, cells were counted at every passage. For all experiments, except those measuring long-term proliferation, cells were used at passages P1–P4. Quantification of vWF secretion by HUVECs. vWF was measured in HUVEC supernatants using the von Willebrand factor (vWF) Human ELISA Kit (Ab108918; Abcam). Cells grown on 10-cm dishes were exposed to high NaCl. To maintain logarithmic growth, cells were reseeded to six-well plates 2 d after NaCl was increased and maintained in high NaCl for 2 d more. Some cells were returned to control medium for 2 d after 4 d in high NaCl. vWF secretion was measured during the last 20 h of exposure to high Na by replacing the media with fresh ones during that time. Then, the medium was collected, centrifuged at 3,000 × g for 10 min, and vWF concentration was measured in the supernatants. Cells were lysed and total protein measured as described below for Western blots. The amount of vWF secreted in 20 h was calculated from the collected volume of culture medium and concentration of vWF, normalized by total cell protein. The rate of secretion of vWF, when NaCl was elevated, was normalized by the rate in control medium. Extraction of Protein from HUVECs and Western Blot. Cells were rinsed with PBS, adjusted with NaCl to the same osmolality as the medium, then lysed with RIPA buffer [50 mM Tris·HCl, 1% Nonidet P-40, 150mM NaCl, 1mM EDTA, 1 mM NaF, 1 mM Na3VO4, and protease inhibitors (Roche Diagnostics)]; 3× Laemmli sample buffer (Cell Signaling) was added to the lysates and samples were boiled for 5 min. Sample loading onto gels was equalized according to the total protein concentration measured before addition of Laemmli buffer. Primary antibodies were against vWF (A0082; Dako), NFAT5 (sc-13035; Santa Cruz Biotechnology), α-tubulin (691251; MP Biomedicals), and aldose reductase (kindly provided by Peter Kador, University of Nebraska Medical Center, Omaha, NE). Secondary antibodies were labeled with Alexa Fluor 680 nm dye (Invitrogen). Immunoblots were scanned and integral fluorescence (IF) from each band was measured using Odyssey Infrared Imaging System (Li-COR Biosciences). Immunofluorescent Detection of vWF in HUVEC Cells. Cells grown on eight-chamber slides were fixed for 10 min in 2% (vol/vol) formaldehyde (18814; Polysciences, Inc.) at room temperature, washed with PBS, permeabilized with 0.1% Triton X-100 in PBS, and blocked with 3% (wt/vol) BSA for 1 h at room temperature. Slides were incubated with primary antibodies for vWF (A0082; Dako) at 4 °C overnight, followed by secondary antibodies labeled with Alexa Fluor 488 nm (green emission) (4412; Cell Signaling) at room temperature for 1 h. After two washes with PBS, cells were stained with 2.5 μg/mL DAPI (DNA stain) (D1306; Invitrogen) and mounted with ProLong Gold antifade Dmitrieva and Burg www.pnas.org/cgi/content/short/1404809111

reagent (P36930;Invitrogen). Pictures were taken by confocal microscopy using a Zeiss LSM 510 microscope (Carl Zeiss MicroImaging) with a 63× N.A. 1.4 oil-immersion objective. RNA Extraction from HUVEC Cells and Quantification by Real-Time PCR. Cells were exposed to high NaCl for 4 d before measure-

ment. Some cells were then returned to control medium for 2 d. Total RNA was extracted with RNeasy Mini Kit (74104; Qiagen). The RNA (1 μg) was converted to cDNA by reverse transcription using TaqMan high-capacity cDNA RT kit (4374966; Applied Biosystems). vWF and NFAT5 mRNA levels were measured by real-time PCR using human TaqMan gene expression assays (4331182; Life Technologies) on a 7900HT Fast Real-Time PCR System (Applied Biosystems). Quantification was performed by the Comparative CT method; 18S ribosomal RNA was used as an endogenous control (4319413E; Applied Biosystems) and was amplified in the same tube with the target. The relative copy number [2 – (Ct(vWF) – Ct (18S rRNA)] of a target was calculated for each sample and normalized to the copy number in the corresponding control sample (270 mosmol/kg). The experiment was repeated five times. ChIP. ChIP was performed using the Enzymatic Chromatin IP Kit (9003; Cell Signaling Technology) as described (1). HUVECs grown on 15-cm dishes were cross-linked with 1% formaldehyde for 10 min. Chromatin was digested with MNase to generate mainly mononucleosomes with a minor fraction of dinucleosomes. For each sample, chromatin containing 7 μg of DNA was immunoprecipitated with 10 μg of anti-NFAT5 antibody (13035; Santa Cruz Biotechnology) or with anti-normal rabbit IgG (2729; Cell Signaling Technology). Protein and RNA in samples were enzymatically digested, and the DNA was further purified. The number of DNA fragments containing target sequences in input chromatin and in chromatin immunoprecipitated (IP) with antiNFAT5 and IgG were quantified by real-time PCR with SYBRGreen PCR Kit (204054; Qiagen). Three target sequences were quantified, one containing the NFAT5 binding site and two nearby sites outside of the NFAT5 binding site (the following section gives the target sequences). Quantification was performed by comparative CT method. The relative copy number of each target sequence for each IP sample was calculated as 2 – (Ct(IP DNA) – Ct (Input DNA)). The number of copies of each target sequence in an NFAT5 ChIP was normalized by the copy number in an IgG ChIP. The experiment was repeated three times. Data are presented relative to control (270 mosmol/kg). Primer Design for NFAT5 ChIP. Three primer pairs specific for the three regions close to the transcription start site of the vWF gene were designed using the Primer-BLAST tool (www.ncbi.nlm.nih. gov/tools/primer-blast/). One of these regions contains the NFAT5 binding sequence (TGGAAATGTCC); the other two regions do not, and served as negative controls. The regions of the vWF gene that were tested by ChIP are shown in Fig. 2D. Sequences of the primers are as follows: (i) (−) control 1: forward primer (GATTCTGCCTCTGGTGCCAT), reverse primer (CACATACCTCGCACTCTGCT); (ii) NFAT5 binding site: forward primer (CAGGGTACCAGAAGTGGGTG), reverse primer (GGTGAAGGTGGGGAGTGATG); and (iii) (−) control 2: forward primer (CTGTGAAGGTTCTGCGGAGT), reverse primer (CCCCAACAGGAGATGGCATT). Specificity of the primers was verified by gel electrophoresis of the PCR products. Each produced a single band. 1 of 7

Mice. Mice were purchased at age 3 mo from Taconic (129S6,

RNA Extraction from Mouse Tissues and Analysis by Real-Time PCR.

129SVE; Taconic Farms, Inc.) and housed in the National Heart, Lung, and Blood Institute (NHLBI) animal facility. All mouse studies were done under approved NHLBI animal study protocols and mice were housed in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

Tissues were collected in Allprotect Tissue Reagent (76405; Qiagen) at the end of the experiments and stored at −20 °C. Total RNA was extracted using AllPrep DNA/RNA Mini Kit (80204; Qiagen). RNA from lung was extracted using RNeasy Fibrous Tissue Mini Kit (74704; Qiagen). cDNA conversion and quantification of vWF, NFAT5, and aldose reductase (AR) mRNA by real-time PCR was performed as described above for HUVEC cells. Mouse TaqMan Gene Expression Assays (4331182; Life Technologies) for NFAT5, vWF, and AR were used.

Water Restriction (9 d). The experiment design is shown in Fig. 3A. All mice received gelled food containing 30% water [1.7 mL of deionized water + 4 g of balanced purified rodent diet (AIN76A; Research Diets) + 57 mg of agar per 5.7 g of the food]. Food was provided in excess in individual cups so the mice ate what they wanted. The control group had free access to water. Water-restricted mice did not get any additional water. Five control and five water-restricted mice were tested. For urine collection, mice were housed in metabolic cages (Hatteras Instruments). Urine was collected in vials under mineral oil to prevent evaporation. Collection vials were replaced every 24 h and urine osmolality was measured with a Fiske Model 210 Freezing-Point Micro-Osmometer (Fiske Associates). Mouse Blood Collection. Blood was collected from the tails of the mice. The mice were placed under a heat lamp for 3–5 min to warm them and dilate blood vessels. After warming, the mice were placed in a restraining device from which their tails protruded. A small nick was made in the tail with a sterile scalpel blade. Four drops of the blood were collected into a tube containing 20 μL of heparin sodium (1,000 U/mL). Pressure was applied to the nick for 15–30 s to stop the flow of blood. Tubes were centrifuged at 1,000 × g for 10 min. Plasma was transferred to new tubes then centrifuged for 10 more minutes at 10,000 × g to remove platelets. The platelet-free plasma was stored at −80 °C. Collection tubes were weighed before and after the blood collection and after the plasma was transferred to calculate how much the plasma was diluted by the heparin. This dilution factor was used to calculate the concentrations of plasma proteins, vWF, and D-dimer. To prepare serum, blood was collected in the same way, but without heparin. The blood was allowed to clot by leaving it undisturbed at room temperature for 30 min, then the clot was removed by centrifuging at 4 °C for 10 min at 1,800 × g. Measurement of Total Protein, vWF, and D-Dimer in Mouse Plasma.

Total protein concentration in mouse plasma was measured using BCA Protein Assay Kit (23227; Pierce). vWF and D-dimer in the plasma were measured by Western blot. Equal volumes of plasma were loaded on the gel. Before loading, 3× Laemmli sample buffer supplemented with DTT (7722; Cell Signaling) was added to the samples and the mixtures were boiled for 5 min. Immunoblots used primary antibodies against vWF (A0082; Dako) and D-dimer (bs-3514R; Bioss) and secondary antibodies labeled with Alexa Fluor 680 nm dye (Invitrogen). The immunoblots were scanned and integral fluorescence from each band was measured using Odyssey Infrared Imaging System (LI-COR Biosciences). Measurement of Plasminogen Activator Inhibitor 1 in Plasma. Plasminogen activator inhibitor 1 (PAI-1) was measured using a beadbased sandwich Luminex immunoassay technique (2). PAI-1 agarose beads and assay reagents were purchased from EMD Millipore (MCVD1-77AK). The assay was performed according to the manufacturer’s instructions and PAI-1 was measured on a Luminex 100 System. Measurement of Serum Sodium, Potassium, Blood Urea Nitrogen, Glucose, and Osmolality. The chemical parameters were measured

with a Vitros 250 Chemistry System (Ortho-Clinical Diagnostics). Serum osmolality was measured with a vapor pressure osmometer (Vapro 5520; Wescor, Inc.). Dmitrieva and Burg www.pnas.org/cgi/content/short/1404809111

Immunohistochemical Staining of Paraffin-Embedded Mouse Tissue Sections. Mouse tissues were fixed for 48 h in 4% paraformalde-

hyde at 4 °C then embedded in paraffin. Sections were cut and mounted on silanized slides by American Histolabs. Sections were stained with antibodies against vWF (A0082; Dako), CD31 (DIA 310; Dianova), and fibrinogen (4440-8004; AbD Serotec), as previously described (3). Visualization was with diaminobenzidine (DAB) (D22187; Molecular Probes), which generates a browncolored oxidation product upon reaction with the HRP-labeled secondary antibody. Development of the brown DAB color was monitored to prevent oversaturation that would prevent quantification. A Nikon E800 Widefield Microscope was used for photography. Quantification of vWF Protein Expression on Mouse Tissue Sections.

To quantify vWF protein expression on the DAB-stained sections (discussed above), we used the yellow-CMYK channel method described by Pham et al. (4). They demonstrated that images in the CMYK yellow channel correlate with the immunohistochemistry DAB stain intensity. We extracted the CMYK yellow channel from the images of liver tissue sections stained with DAB for vWF (brown color) and with hematoxylin for nuclei (blue) (Fig. 4B) and quantified average yellow intensity per pixel as a measure of vWF protein expression. We processed the images according to Pham et al. (4), implemented with a Python script. The source image ðR; G; BÞ is transformed to the final image ðR′; G′; B′Þ as follows:  R′; G′; B′ = ðY ; Y ; 0Þ ð1 − BÞ where Y = 255p 255 − K and K = 255 − maxðR; G; BÞ: This transformation computes the yellow channel of the image by the RGB-to-CMYK format conversion and then loads that channel value back into the standard RGB image format. The effect of this transformation is to extract the yellow channel of the image. Examples of original images and images in the yellow channel after processing are given on Fig. S3. The Python script that performs this transformation is available upon request. Images of liver sections from five control and five water-restricted mice were analyzed; 10–15 images were taken from each section and their average intensity was calculated.

Statistical Analysis. Comparisons were performed by the nonpaired Student t test. P < 0.05 was considered significant. All data with n = 5 were tested by the Kolmogorov–Smirnov method for normality. All except AR mRNA passed the test. For AR, statistical significance was ascertained by the nonparametric Mann– Whitney test. For n < 5 (measurements of NFAT5 and AR proteins in HUVEC cells) normal distribution was assumed and Student t test was used. vWF protein secretion, vWF mRNA, NFAT5 protein and mRNA, and AR protein levels in HUVEC cells were analyzed for linear dependence on the level of NaCl. All of the statistical analyses were done using GraphPad InStat Software. 2 of 7

SI Results Analysis of Association Between Plasma Sodium and Blood Level of vWF in Humans. To test for possible relation of plasma sodium

about age is still preserved in the model through those correlations. Second, BMI was eliminated to reduce redundancy based on its relatively high correlation with glucose, which is a more interesting variable for our study, because glucose contributes to plasma osmolality and directly affects endothelia. Third, eGFR was included because it significantly correlates with vWF level but not with Na+ and has a weaker correlation with glucose than other variables. In summary, our model includes the least correlated variables satisfying the requirement of independence in multivariable analysis, but at the same time preserves information about other variables, such as age and BMI, through correlations with them. For the final multivariable analysis, we transformed the vWF variable to make its distribution normal (Fig. S5). Table S4 summarizes the results of the regression: The overall regression is statistically significant for all of the variables, namely, plasma Na+, glucose, and eGFR, contributing to the plasma level of vWF [F(3, 14,675) = 210, P < 0.001]. We next removed eGFR from the model to see whether Na+ and glucose can predict vWF without eGFR. The two-variable regression still is statistically significant, with both Na+ and glucose being significant predictors of vWF level [F(2, 14,676) = 289, P < 0.001] (Table S5). Because regression coefficients are positive, we conclude that an increase in plasma Na+ and/or glucose is accompanied by increased vWF.

concentration with that of vWF in humans, we analyzed the correlation of plasma sodium and vWF using data from the Atherosclerosis Risk in Communities (ARIC) Study. The data were obtained from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center. Therefore, the results of our analysis do not necessarily reflect the opinions or views of the ARIC Investigators. ARIC is a study of cardiovascular disease in a cohort of 15,792 persons sampled from four US communities in 1987–1989. At baseline, 45- to 64-y-old members of sampled households in Minneapolis, Minnesota (selected suburbs); Forsyth County, North Carolina; Washington County, Maryland; and Jackson, Mississippi were enrolled (5). We used the results of the baseline clinical examination of the participants during their first visit. The data were analyzed using SAS (SAS Institute Inc.) and SigmaPlot (Systat Software) software. We conducted multivariable regression analysis to assay the relation between the concentrations of serum sodium and vWF (6). Complete data were available for 14,679 participants. Participants with missing data were removed from the analysis. The following variables from the ARIC Study were used: sodium (mmol/L), vWF (IU/dL), glucose (mmol/L), blood urea nitrogen (BUN, mg/dL), creatinine (mg/dL), albumin (g/dL), body mass index (BMI, kg/m2), and age (years). Estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease equation (7): eGFR = 170 × Serum Creatinine−0.999 × Age−0.176 × (0.762 if Female) × (1.118 if Black) × BUN−0.170 × Albumin+0.318. We chose the variables to include in the analysis based on their known ability to affect the level of vWF. Diabetes, age, and chronic kidney disease (CKD) are all associated with increased levels of vWF (8, 9). Therefore, we included age, eGFR, glucose, and BMI in the list of candidates for inclusion in multivariable model as covariates to plasma sodium. Table S2 and Fig. S5 show the basic statistics and the distribution histograms for the chosen variables. Some of them are skewed. Therefore, for initial assessment of zero-order correlations, we used Spearman’s rank correlation coefficient, a nonparametric measure of statistical dependence between two variables (Table S3). The zero-order correlation analysis did not reveal any association between sodium and vWF. However, both vWF and Na+ levels are significantly correlated with other variables. Therefore, we proceeded with multiple regression analysis. We included Na+, glucose, and eGFR in the final multivariable model. The reasons for this selection are (6) as follows. First, age was eliminated because its distribution is very far from normal (Fig. S5), and age is significantly correlated with all variables included in model (Table S3). Information

tively calculated for participants in the ARIC Study based on the study outcomes and included in ARIC datasets (10). We used this information to test whether there is an association between plasma sodium concentration and the predicted stroke risk. The data were available for 13,630 participants. Participants with missing data were removed from the analysis. We conducted a multivariable regression analysis similar to the one that we performed to assess association between sodium and vWF. The analysis demonstrates that each of the predictor variables (plasma sodium, glucose, and eGFR) has a significant independent zero-order correlation with 10-y stroke risk (Table S6). In the multivariable model, all of the variables significantly (P < 0.001) contribute to the predicted risk of stroke [F(3, 13,626) = 866, P < 0.001] (Table S6). We next performed the same analysis on the cohort of participants without diabetes and without CKD (eGFR >60 mL·min−1·1.73 m−2) who had normal weight (BMI = 18.5–25 kg/m2). In this “healthy” cohort (n = 3,345), all of the variables significantly (P < 0.001) contribute to predicted risk of stroke (Table S7). Thus, plasma sodium concentration is positively associated with 10-y risk of stroke predicted based on outcomes of the ARIC Study.

1. Dmitrieva NI, Cui K, Kitchaev DA, Zhao K, Burg MB (2011) DNA double-strand breaks induced by high NaCl occur predominantly in gene deserts. Proc Natl Acad Sci USA 108(51):20796–20801. 2. Khan SS, Smith MS, Reda D, Suffredini AF, McCoy JP, Jr. (2004) Multiplex bead array assays for detection of soluble cytokines: Comparisons of sensitivity and quantitative values among kits from multiple manufacturers. Cytometry B Clin Cytom 61(1):35–39. 3. Dmitrieva NI, Burg MB (2007) High NaCl promotes cellular senescence. Cell Cycle 6(24): 3108–3113. 4. Pham N-A, et al. (2007) Quantitative image analysis of immunohistochemical stains using a CMYK color model. Diagn Pathol 2:8. 5. Williams OD (1989) The Atherosclerosis Risk in Communities (ARIC) Study: Design and objectives. The ARIC investigators. Am J Epidemiol 129(4):687–702.

6. Royston P, Sauerbrei W (2008) Selection of variables. Multivariable Model-Building: A Pragmatic Approach to Regression Analysis Based on Fractional Polynomials for Continuous Variables, eds Royston P, Sauerbrei W (Wiley, New York), pp 23–52. 7. Levey AS, et al.; Modification of Diet in Renal Disease Study Group (1999) A more accurate method to estimate glomerular filtration rate from serum creatinine: A new prediction equation. Ann Intern Med 130(6):461–470. 8. Shen L, et al. (2012) Von Willebrand factor, ADAMTS13 activity, TNF-α and their relationships in patients with chronic kidney disease. Exp Ther Med 3(3):530–534. 9. Blann AD (2006) Plasma von Willebrand factor, thrombosis, and the endothelium: the first 30 years. Thromb Haemost 95(1):49–55. 10. Chambless LE, Heiss G, Shahar E, Earp MJ, Toole J (2004) Prediction of ischemic stroke risk in the Atherosclerosis Risk in Communities Study. Am J Epidemiol 160(3):259–269.

Dmitrieva and Burg www.pnas.org/cgi/content/short/1404809111

Analysis of Association Between Plasma Sodium and Stroke Risk in the ARIC Study. The 10-y stroke risk at first visit was retrospec-

3 of 7

Fig. S1. Model summarizing the proposed relation between sodium, vWF, and thrombosis. Elevation of NaCl in extracellular fluid leads to graded increase of endothelial cell vWF mRNA, vWF protein, and of the rate of secretion of vWF from them. Increased vWF provides an explanation for the occurrence of thrombosis during hypernatremia. Even a small increase in plasma sodium (5 mmol/L), as the result of water restriction, increases vWF secretion from endothelial cells enough to increase blood vWF and stimulate thrombogenesis. Our findings suggest that hypernatremia-induced vWF secretion caused by insufficient water or excessive salt intake adversely affects cardiovascular health by increasing coagulability of blood and thus facilitating thrombotic events.

Fig. S2. High NaCl increases production and secretion of vWF from vascular endothelial cells. (A) Images of HUVEC cells grown in high-NaCl medium for 15 d. HUVEC cells adapt well to medium in which NaCl concentration is increased to total osmolality up to 380 mosmol/kg. (B) High NaCl increases vWF secretion rate from HUVEC cells and secretion stays elevated for many days as long as NaCl stays elevated. Primary HUVEC cells were exposed to media in which level of NaCl was elevated to total osmolality 380 mosmol/kg. The cells were exposed to such high-NaCl medium for 34 d. Cells were maintained in logarithmic growth by passaging them when they reached about 80% confluency. (C) Expression of vWF protein is higher in mouse renal medulla, where interstitial NaCl normally is high, than in cortex, where interstitial NaCl concentration is similar to that in systemic blood. Kidney tissue sections from normal mice show immunohistochemical localization of vWF in endothelial cells. Brown staining indicates vWF or CD31, a marker of endothelial cells; blue staining (hematoxylin) indicates nuclei. vWF is high in renal medullary endothelial cells but is almost entirely absent in the renal cortex.

Dmitrieva and Burg www.pnas.org/cgi/content/short/1404809111

4 of 7

Fig. S3. Representative examples of CMYK yellow channel extraction from images of liver tissue sections stained for vWF using DAB chromogen and counterstained with hematoxylin. (Left) Original images. (Right) Images of extracted CMYK yellow channel. The pattern and intensity of the signal in the CMYK yellow channel (Right) corresponds in pattern and intensity to the brown DAB staining of vWF (Left). Also see SI Materials and Methods and Fig. 4 for full description of the vWF quantification.

Fig. S4. Pattern of vWF staining correlates with the pattern of staining for the endothelial cell marker, CD31. Images of liver tissue sections stained for vWF and CD31, using DAB chromogen (brown) and counterstaining with hematoxylin (blue, nuclei).

Dmitrieva and Burg www.pnas.org/cgi/content/short/1404809111

5 of 7

Fig. S5.

Frequency distribution histograms for variables used in analysis of association of sodium with vWF and risk of stroke in the ARIC Study.

Table S1. of mice

Effect of water restriction on serum parameters

Serum variable

Control

Sodium, mmol/L Osmolality, mosmol/L BUN, mg/dL Glucose, mg/dL Potassium, mmol/L

145.1 294.4 28.3 132.4 5.9

± ± ± ± ±

Water restriction

0.5 2.5 1.6 3.3 0.1

150.2 302.4 37.3 136.2 5.4

± ± ± ± ±

1.3** 3.6* 4.0* 9.8 0.1*

Mice were subjected to water restriction for 9 d. Blood was collected from tails and serum was prepared as described in Materials and Methods. *P < 0.05; **P < 0.01.

Table S2. Basic descriptive statistics for the variables used in analysis of the association between sodium and vWF in the ARIC Study Variable

Mean

SD

5–95% percentiles

141.0 118.7 27.7 6.1 69.1

2.4 48.6 5.4 2.3 12.4

137–145 58–206 20.6–37.6 4.7–9.6 51–90

+

Na , mmol/L vWF, IU/dL BMI, kg/m2 Glucose, mmol/L eGFR, mL·min−1·1.73 m−2

Table S3. Spearman’s rank correlation coefficients for the variables used in analysis of the association between sodium and vWF in the ARIC Study Variable +

Na vWF Age BMI eGFR

vWF

Age

BMI

eGFR

Glucose

0.00

0.06*** 0.19**

0.03* 0.12** 0.00

0.00 −0.05*** −0.23** −0.02

−0.02* 0.14** 0.14** 0.32** 0.01

*P < 0.001; **P < 1.00E-06; ***P < 1.00E-09.

Dmitrieva and Burg www.pnas.org/cgi/content/short/1404809111

6 of 7

Table S4. Multiple regression analysis of plasma level of vWF (transformed) with plasma Na+, glucose, and eGFR as predictor variables (ARIC Study, n = 14,679) Independent variable Intercept Na+, mmol/L Glucose, mmol/L eGFR, mL·min−1·1.73 m−2

Regression coefficient bj

SE of bj

t

P

3.918 0.0050 0.0344 −0.0018

0.189 0.0013 0.0014 0.0003

20.8 3.8 24.1 −7.2