| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 278, Issue 32, 30037-30043, August 8, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
||
From the
Department of Woman and Child Health,
Karolinska Institutet, S-171 76 Stockholm, Sweden and the
Laboratory of Physiological Genetics, Institute
of Cytology and Genetics, and the ¶Group of
Functional Genomics, Novosibirsk Institute of Bioorganic Chemistry, Siberian
Branch of Russian Academy of Sciences, 630090 Novosibirsk, Russia
Received for publication, March 3, 2003 , and in revised form, May 12, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Nickel is widely used in modern industry (reviewed in Ref. 6). Inhalation is the primary route of occupational exposure to nickel and other heavy metals, and inhalation of nickel compounds is a common cause of pneumoconiosis (6, 7). AQP3, AQP4, and AQP5 are expressed in the airway epithelia (8–10). AQP3 is located at the apical membrane of human lung epithelium (10). Here, we have examined the effects of Ni2+ on the water permeability of human AQP3, AQP4, and AQP5 expressed in a human lung cell line. Since AQP3 has, when expressed in oocytes, been reported to be pH-sensitive (11), we also examined the effect of extracellular acidification. We show that Ni2+ and pH regulate the water permeability of human AQP3, but not of human AQP4 and AQP5. We also address the question of whether Ni2+ and pH may interact in the regulation of human AQP3.
Identification of the molecular sites responsible for the
Ni2+ and pH sensitivity of AQP3 is important for future
development of therapeutic agents. Histidine, with a
pKa of 
6.5, is the most likely molecular
target for regulation by pH and is also a preferential site for
Ni2+ binding. By performing a series of mutations of
extracellular histidines and amino acids considered to interact with
histidine, we identified several molecular determinants of pH and
Ni2+ sensitivity and at least one common determinant of
Ni2+ and pH sensitivity.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The point mutations in the extracellular loops of human AQP3 were generated by PCR-based mutagenesis using wild-type cDNA as a template. The presence of each point mutation and absence of other modifications were confirmed by sequence analysis of the whole insert.
The transmembrane structure of human AQP3 was predicted using TMHMM Version 2.0 (12). The protein sequences were aligned with ClustalW Version 1.81 (13).
Cell Culture—The human bronchial epithelial cell line BEAS-2b (subpassages 10–36; European Collection of Cell Cultures, Center for Applied Microbiology and Research, Salisbury, Wiltshire, UK) was cultured on coverslips (Bioptechs, Butler, PA) coated with collagen type I and fibronectin (Sigma, Stockholm, Sweden) in Dulbecco's modified Eagle's medium/nutrient mixture F-12 (1:1; Invitrogen, Paisley, Scotland, UK) containing 0.5 units/ml penicillin and 50 µg/ml streptomycin and supplemented with 10% heat-inactivated fetal bovine serum and 2 mM L-glutamine. On the second day of culture, the cells were transiently transfected with cDNA constructs (see above) using CLONfectin (Clontech) according to the manufacturer's protocol. Experiments were performed on the fourth day of culture.
Measurement of Water Permeability—The water permeability (Pf) was measured using a method that we recently described in detail (14). The method allows one to determine the Pf in individual cells within cell monolayers and to compare the Pf in cells that do and do not express GFP-labeled proteins. Briefly, the coverslips with the transfected cells were mounted in a closed perfusion chamber (Focht Live Cell Chamber System; Butler, PA) on the stage of a Zeiss 410 inverted laser scanning microscope and scanned every 2 s with excitation at 488 nm and emission at 515–525 nm. At the beginning of every Pf measurement, an image showing the distribution of GFP-tagged proteins was recorded. Then, with the specimen remaining on the stage of the microscope, the cells were loaded with calcein (Molecular Probes Europe, Leiden, The Netherlands). To measure the Pf, the cells were perfused with 300 mosM PBS (137 mM NaCl, 0.9 mM CaCl2, 0.49 mM MgCl2, 2.7 mM KCl, 1.5 mM KH2PO4, and 8.1 mM Na2HPO4, pH 7.4). The perfusate was then switched to 200 mosM PBS (osmolarity was decreased by reducing the concentration of NaCl to 87 mM). To determine the Pf at a different pH, the pH of both the 300 and 200 mosM perfusion solutions was adjusted to the required value. In solutions at pH 8.0 and higher, CaCl2 and MgCl2 were omitted. In experiments with NiCl2, ZnCl2, and CdCl2, phosphates in all solutions were replaced with 10 mM HEPES, and 1.5 mM KCl was added to preserve potassium concentration.
The Pf was calculated using the following
equation (14):
Pf = 
(1 –
b/V0)(
(A/V)0Vw
0)–1.
The time constant was calculated for every cell from the curve showing
changes in fluorescence intensity inside the cell during osmotic swelling. 1
– b/V0 represents the osmotically active portion of
the cell volume. This portion was calculated using direct volume measurements
performed using the Imaris Image Processing Toolkit (Bitplane, Zurich,
Switzerland) on stacks of images recorded through the calcein-loaded cells at
300 mosM and through the same cells after complete swelling in 200
mosM PBS. For BEAS-2b cells, the osmotically active portion of the
cell volume was found to be 0.15 ± 0.02 (n = 7). The constant
was calculated as the slope of the relative fluorescence
versus the relative osmolarity calibration curve. For BEAS-2b cells,
it was found to be 0.63. (A/V)0 is the initial
cell surface-to-volume ratio. It was calculated using Imaris on stacks of
images recorded through the calcein-loaded cells incubated in 300
mosM PBS. For BEAS-2b cells, the ratio was found to be 5111
± 286 cm–1 (n = 9).
Vw is the partial molar volume of water (18
cm3/mol), and 0 is the initial osmotic
gradient (outside-inside), which was 10–4
mol/cm3 in our experiments.
The basal Pf in BEAS-2b cells and the Pf in transfected cells could vary depending on the lot of CLONfectin used for the transfection and the cell culture subpassage. Therefore, we statistically compared only the Pf in cells that were at a similar subpassage and transfected with the same CLONfectin preparation.
Analysis of the Subcellular Distribution of GFP-tagged AQP3—To determine whether the effects of acidic pH or NiCl2 were due to a changed amount of AQP3-GFP in the plasma membrane, the transfected cells were mounted in the same chamber as used for Pf measurements in the control solution. A stack of images with a vertical displacement of 0.4 µm was recorded. The solution was then changed to a pH 5.5 solution or a solution containing 1 mM NiCl2, and a new stack of images was recorded. The ratio of GFP signal in the plasma membrane to that in the adjacent cytosol was measured using ImageJ software (Research Services Branch, National Institute of Mental Health, Bethesda, MD).2 The subcellular localization of AQP3 was compared in the same cells in two different solutions. Images at the same cell height were taken for the comparison. No swelling or shrinkage of the cells was observed in response to low pH or NiCl2.
Measurement of Intracellular pH—Cells were loaded with the pH-sensitive fluorescent dye 2',7'-bis(2-carboxyethyl)-5(6)-carboxyfluorescein (BCECF) by incubation with 10 µM BCECF/acetoxymethyl ester (Molecular Probes Europe) for 30 min at 37 °C. The cells were mounted in the same chamber used for Pf measurements (see above) on a Zeiss Axiovert 135 microscope using a x40/1.4 epifluorescence oil immersion objective. Fluorescence was alternately excited at wavelengths of 440/10 nm and 470/10 nm; emission fluorescence was collected with a 510/30-nm band pass filter. Data were recorded with a GenIISys image intensifier system connected to a CCD camera (MTI CCD72, Dage-MTI, Inc., Michigan City, IN) and analyzed using RatioTool software from ISee Imaging Systems (Raleigh, NC). At the end of each experiment, the intracellular pH was calibrated using buffer solutions containing 150 mM KCl, 5 mM MES or HEPES, and 10 µM nigericin adjusted to pH 6.0, 6.5 (MES), 7.0, 7.5, and 8.0 (HEPES).
Data Presentation and Analysis—Data are presented as means ± S.E. Statistical analysis was performed using Student's t test. A difference of p < 0.05 was considered statistically significant.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Sensitivity of Human AQP3 to pH and Ni2+—Acidification of the extracellular solution dose-dependently decreased the Pf in cells expressing AQP3-GFP without affecting untransfected cells (Fig. 1e). At pH 5.5, the Pf in transfected cells was not different from that in untransfected cells. The pH effect was similar in cells expressing AQP3-GFP and in cells expressing AQP3+GFP. Inhibition of AQP3-mediated water permeability by acidic pH was rapid and reversible (Fig. 1, f and g). There was no detectable change in the distribution of human AQP3-GFP in cells exposed to low extracellular pH. The ratio of GFP signal in membranes to that in the cytosol was 2.52 ± 0.27 before and 3.00 ± 0.27 (n = 8) after 1 min in pH 5.5 solution, when the AQP3-mediated water permeability was completely inhibited. Notably, the intracellular pH was only little affected when the extracellular pH was reduced from 7.4 to 5.5 (7.34 ± 0.01 and 7.27 ± 0.01, respectively; n = 58).
Nickel also decreased the Pf in cells expressing human AQP3-GFP in a dose-dependent manner (Fig. 2a). The effect was rapid and reversible (Fig. 2b). The Pf in untransfected cells was not affected by Ni2+. There was no detectable change in the distribution of AQP3-GFP in cells exposed to Ni2+. The ratio of GFP signal in membranes to that in the cytosol was 3.89 ± 0.34 before and 4.05 ± 0.31 (n = 8) after 1 min in solution with 1 mM NiCl2, when the AQP3-mediated water permeability was dramatically down-regulated.
|
The effect of Ni2+ was pH-dependent (Fig. 2c). At neutral and acidic pH, the AQP3-mediated water permeability was completely inhibited by 1 mM NiCl2. At pH 7.4 and 8.0, the Pf in transfected cells was decreased by Ni2+, but remained significantly higher than that in untransfected cells.
Specificity of the pH and Ni2+ Effect on Human AQP3— AQP4 and AQP5 are, like AQP3, expressed in the epithelial cells of the lower airways of the human lung (10). BEAS-2b cells were transfected with either human AQP4 or AQP5 using cDNA constructs encoding GFP and the water channels as separate or fusion proteins. The water permeability was increased 6–8-fold in cells expressing AQP4 or AQP5 compared with untransfected cells. Lowering the extracellular pH to 5.5 or exposing these cells to Ni2+ had no effect on the Pf (Fig. 3, a and b).
|
Like nickel, zinc and cadmium are well documented causes of pneumoconiosis (6, 7). Cells transfected with AQP3-GFP were incubated with 1 mM ZnCl2 or CdCl2. The Pf in cells expressing AQP3-GFP or in untransfected cells was not affected by either of these compounds (Fig. 3c).
Extracellular Determinants of the Ni2+ and pH Sensitivity of Human AQP3—To examine the molecular determinants of Ni2+ and pH sensitivity, we performed a series of point mutations in the three extracellular loops of AQP3 (Fig. 4). The extracellular loops were chosen for the following reasons. First, the regulation of AQP3 by Ni2+ and pH was rapid and reversible. Second, the changes in extracellular pH that inhibited the water permeability of AQP3 did not lead to any substantial changes in intracellular pH. Third, inhalation is the main route for Ni2+ exposure. AQP3 is expressed apically in the human airways, and its extracellular loops will be the first target for nickel binding.
|
The following amino acids were chosen for mutation: histidine, which can interact with Ni2+ (15) and which is the only amino acid with a pKa in the range for AQP3 pH sensitivity; serine, aspartate, tryptophan, and tyrosine, since they may interact with histidine (16–18). The results from the mutation studies are summarized in Table I, and the most relevant results are presented in Figs. 5 and 6. All mutants except D219A and W231A (Fig. 5, a and b) were targeted to the plasma membranes of the cells. In two cases (S49A and H154A), mutant AQP3 was water-impermeable. In five cases (H53A, H53F, H53Y, Y124A, and H154Y), the maximal water permeability in cells expressing mutant AQP3 was significantly decreased compared with that in cells expressing wild-type AQP3.
|
|
|
Three residues, Trp128 and Ser152 in the second extracellular loop and His241 in the third extracellular loop, were identified as determinants of AQP3 Ni2+ sensitivity (Fig. 5, c and d). The Pf in cells expressing AQP3(W128A), AQP3(S152A), or AQP3(H241A) was similar to that in cells expressing wild-type AQP3. Ni2+ had no effect on the water permeability of any of these mutants.
None of AQP3 mutants completely abolished the pH dependence of the water channel. In cells expressing AQP3(H53A), AQP3(Y124A), or AQP3(H154F), the range of pH sensitivity was shifted to more alkaline values (Fig. 6, a–c). In cells expressing AQP3(S152A), the range of pH sensitivity was shifted to more acidic pH (Fig. 6d). The water permeability of wild-type AQP3 was completely inhibited at pH 5.5. At this pH, the Pf in cells expressing AQP3(S152A) was 3.9-fold higher compared with that in the surrounding untransfected cells.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
It is well documented that Ni2+ can up- or down-regulate the activity of ion channels (19–24), but this is, to the best of our knowledge, the first demonstration of Ni2+ regulation of the activity of a water channel. The effect of Ni2+ was immediate, reversible, and dependent on at least three extracellular residues. In studies of the mechanisms by which Ni2+ modulates the activity of cyclic nucleotide-gated channels and the epithelial sodium channel, extracellular histidines were found to be the molecular determinants of Ni2+ sensitivity (19, 24). Nitrogen in the imidazole side chain of histidine is the preferred target for Ni2+ binding (15). Human AQP3 has one histidine in the first extracellular loop, two in the second, and one in the third. In this study, we found that one of these four extracellular histidine residues, His241, is crucial for the Ni2+ sensitivity of human AQP3. The side chain of tryptophan has a structure that is similar to that of the imidazole ring of histidine. Therefore, we reasoned that tryptophan may also be a candidate for Ni2+ binding. Indeed, we found that substitution of Trp128 with alanine completely abolished the Ni2+ sensitivity of AQP3. Negatively charged amino acids such as aspartate and glutamate can also bind Ni2+ (25). None of the examined extracellular aspartates proved to be important for the regulation of AQP3 by Ni2+. Glutamate residues are not present in the extracellular loops of human AQP3. Ser152 was the third residue in the extracellular loops of AQP3 that was also found to be crucial for Ni2+ sensitivity. This residue, as well as Tyr124, was studied because the effect of Ni2+ was pH-dependent. Mutation of His241, Trp128, or Ser152 led to complete elimination of Ni2+ sensitivity. This indicates that these three amino acids together are essential for Ni2+ binding to human AQP3.
Mutation of His53 to alanine attenuated and of His154 to alanine abolished the water permeability of AQP3. We hypothesized that the imidazole rings of His53 and/or His154 may participate in stacking interactions, important for the proper conformation of the water path. To address this question, we replaced histidine with the aromatic amino acid phenylalanine. We found that AQP3(H154F), but not AQP3(H53F), had the same maximal water permeability as wild-type AQP3.
Human AQP3 was pH-sensitive in the range from pH 5.5 to 7.0. The imidazole group of histidine is the only amino acid side chain affected within this pH range. At pH 5.5, the group is protonated and positively charged, whereas at pH 7.0, it is electrically neutral. His53 from the first extracellular loop and His154 from the second extracellular loop were found to participate in the regulation of the water permeability by pH. Mutation of Ser152 was also associated with a modification of AQP3 pH sensitivity. Serine is a common partner of histidine in a number of enzymes that have an active-site motif known as the catalytic triad (reviewed in Ref. 16). We suggest that, in AQP3, Ser152 may play a role similar to that in the catalytic triad of enzymes. Tyrosine has been shown to closely interact with histidine in the photosystem of cyanobacteria (17). Cells expressing AQP3(Y124A) had significantly lower maximal water permeability than cells expressing wild-type AQP3. It is possible that the aromatic side chain of Tyr124 may participate in stacking interactions with histidine or some other aromatic residue of AQP3.
The finding that AQP3 water permeability is pH-dependent is in line with previous observations by Zeuthen and Klaerke (11). Two other mammalian AQPs have been found to be pH-sensitive: AQP0, which is expressed in the lens (26), and AQP6, which is expressed in intercalated cells of kidney collecting ducts (27). In contrast to AQP3, the maximal water permeability for AQP0 and AQP6 is low at pH 7.5 and significantly increases at acidic pH. A histidine residue in the first extracellular loop was identified as a determinant of the pH sensitivity of AQP0. Amino acid residues controlling the pH sensitivity of AQP6 were not identified. No attempt has been previously made to identify the molecular determinants of AQP3 pH sensitivity.
Alignment of the protein structures showed that all amino acid residues involved in the regulation of AQP3 by Ni2+ or pH are absent in AQP4 and AQP5. Neither AQP4 nor AQP5 was Ni2+- or pH-sensitive. Nickel and acidic pH can therefore be used as tools to discriminate between the activities of AQP3 on one hand and AQP4 and AQP5 on the other.
This study was performed on a bronchial epithelial cell line derived from the human lung. Nickel is widely used in modern industry, and inhalation is the primary route of occupational nickel exposure. High doses of nickel have been suggested to predispose to asthma, lung fibrosis, and lung cancer (reviewed in Ref. 6). Could interaction of nickel with AQP3 contribute to one or more of these conditions? AQP3 is, like AQP4 and AQP5, abundantly expressed in the airway epithelium. AQP3-null mice have few signs of lung dysfunction except for a small but significant reduction of airway humidification (28, 29). Mice lacking AQP4 and AQP5 also have few signs of lung dysfunction. This has raised the question of whether lung AQPs are important for airway fluid transport. However, there appears to be important species differences with regard to the lung expression of AQPs and particularly with regard to AQP3 expression. AQP3 is present in the apical membranes of bronchial epithelial cells in the human lung, but not in the rodent lung (8–10). Hence, the extracellular loops of AQP3 will, in the human lung, face the airway lumen. Here, we found that Ni2+ binds to the extracellular loops of human AQP3 and that AQP3 is sensitive to extracellular pH. Other investigators have shown that the epithelial sodium channel, which is essential for apical sodium entry into the airway epithelium, is also Ni2+-sensitive (24). Taken together, these observations are compatible with the concept that the initial stages in human lung disease caused by nickel are related to the effect of Ni2+ on apical water and ion transporters. The interaction between nickel and pH becomes interesting in this context. The effect of Ni2+ was more pronounced at acidic and neutral pH than at alkaline pH. Thus, it may be important to ensure that the airway liquid is alkaline in the situation of nickel exposure.
The pH dependence of human lung AQP3 may be of great clinical importance.
The pH of airway surface liquid is 6.8, and it responds rapidly to
changes in systemic acid-base status
(30,
31). In a recent study in
which human airway surface epithelial cells were used, it was found that the
addition of mast cells causes substantial increases in proton secretion
(32). In line with this, it
was found that the pH of exhaled airway vapor condensate is significantly
lower in patients with acute asthma than in healthy controls
(33). In our study, AQP3 water
permeability decreased sharply when the pH fell below 7.0. Interestingly,
there is evidence that airway surface liquid pH may be dynamically regulated
and that activation of the adenylate cyclase-cAMP pathway results in
alkalinization, whereas treatment with histamine and ATP results in
acidification of airway surface liquid
(32,
34,
35).
AQP3 is expressed in several other organs, including the kidney, salivary and lacrimal glands, and skin (9, 36–38). The most prominent phenotype in AQP3 knockout mice is a decrease in urinary concentrating capacity, resulting in severe nephrogenic diabetes insipidus (39). Hence, it will be important to explore the relationship between renal medullopapillary pH and urinary concentrating capacity. Nickel is one of the most common causes of contact allergic dermatitis (6, 40). Nickel inhibition of AQP3 may be a contributor to this condition. AQP3 and another member of the aquaporin family, AQP7, have recently been found in dendritic cells, which play an important role in the innate immune system (41, 42). Mercury, which is well known blocker of AQPs (2–4) and a cause of contact dermatitis (6, 43), inhibits the regulatory cell volume decrease after macropinocytosis in dendritic cells (42). Given the finding that nickel blocks the water permeability of AQP3, it is tempting to speculate that perturbation of volume regulation of dendritic cells is an important contributor to nickel dermatitis and to the inflammatory component of nickel pneumoconiosis.
|
FOOTNOTES |
|---|
|| To whom correspondence should be addressed: Q2:09 Astrid Lindgren Children's Hospital, S-171 76 Stockholm, Sweden. Tel.: 46-8-517-77-327; Fax: 46-8-517-77-328; E-mail: anita.aperia{at}ks.se.
1 The abbreviations used are: AQPs, aquaporins; GFP, green fluorescent
protein; Pf, water permeability; PBS,
phosphate-buffered saline; MES, 4-morpholineethanesulfonic acid.
2 Available at
rsb.info.nih.gov/ij/index.html.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  M. D. Sorani, Z. Zador, E. Hurowitz, D. Yan, K. M. Giacomini, and G. T. Manley Novel variants in human Aquaporin-4 reduce cellular water permeability Hum. Mol. Genet., August 1, 2008; 17(15): 2379 - 2389. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. L. Nemeth-Cahalan, K. Kalman, A. Froger, and J. E. Hall Zinc Modulation of Water Permeability Reveals that Aquaporin 0 Functions as a Cooperative Tetramer J. Gen. Physiol., October 29, 2007; 130(5): 457 - 464. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. N. Nejsum and W. J. Nelson A molecular mechanism directly linking E-cadherin adhesion to initiation of epithelial cell surface polarity J. Cell Biol., July 10, 2007; 178(2): 323 - 335. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H.-F. Huang, R.-H. He, C.-C. Sun, Y. Zhang, Q.-X. Meng, and Y.-Y. Ma Function of aquaporins in female and male reproductive systems Hum. Reprod. Update, November 1, 2006; 12(6): 785 - 795. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Cucu, J. Simaels, J. Eggermont, W. Van Driessche, and W. Zeiske Opposite effects of Ni2+ on Xenopus and rat ENaCs expressed in Xenopus oocytes Am J Physiol Cell Physiol, October 1, 2005; 289(4): C946 - C958. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Zelenina, S. Tritto, A. A. Bondar, S. Zelenin, and A. Aperia Copper Inhibits the Water and Glycerol Permeability of Aquaporin-3 J. Biol. Chem., December 10, 2004; 279(50): 51939 - 51943. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
) was regulated by extracellular pH; the
Pf in untransfected cells (
) and back to 300 mosM (
). The
slope of the curves after the change in osmolarity is proportional to the rate
of cell swelling and is a measure of the water permeability in the cells. The
obvious difference in the rate of swelling of transfected (trace A)
and untransfected (trace B) cell at pH 7.4 disappeared at pH 5.5 and
reemerged when the extracellular pH was returned to 7.4. g, shown is
the mean Pf in transfected (black bars;
n = 15) and untransfected (white bars; n = 28)
cells from the experiment described for f.
and
, the direction of the shift of the pH dependence curve.
