Effects of Estrogen on Apolipoprotein Secretion by the Human Hepatocarcinoma Cell Line, HepG2*

We have examined the effect of estrogen on the rate of accumulation of apolipoproteins secreted by the human hepatocarcinoma cell line, HepG2. Prior to expo- sure to hormone, we detected less than 300 high-affin-ity, nuclear, estrogen-binding sites/cell. Within 48 h of growth in the presence of 20 nM 17;IB-estradiol this number rose to 3000-3500 sites/cell. Rates of accumulation of two of the major apolipoproteins, apo-C-I1 and apo-A-I increased 2.5- and 2.0-fold, respectively, in response to estrogen treatment. Other major apolipoproteins were not affected at this concentration of hormone. Induction of both proteins was completely antagonized by 20 nM testosterone. The density distri- bution of apolipoproteins secreted by the hepatocytes was similar to that reported using perfused liver sys-tems. The consequences of estrogen treatment were to increase the apo-C-II/apo-(2-111 ratio in very low density lipoproteins as well as to decrease the overall very low density 1ipoprotein:high density lipoprotein ratio.

Numerous studies have demonstrated that estrogenic steroids influence apolipoprotein levels in a wide variety of species (1)(2)(3)(4). However, in mammals the precise mechanisms involved remain poorly delineated. Epidemiological studies indicate that premenopausal women have lower VLDL' and higher HDL than age-matched males (5). This decreased VLDL:HDL ratio has been correlated with a markedly reduced risk of developing atherosclerosis (6). Elevated levels of estrogen that remain within the physiological range, such as those that occur during pregnancy and oral contraceptive therapy, have been associated with hypertriglyceridemia (7,8). However, in contrast, estrogen therapy has been shown to ameliorate the condition of individuals suffering from a familial form of hyperlipoproteinemia known as dysbetalipoproteinemia (Type I11 familial hyperlipoproteinemia) (9). In at least two mammalian species, pharmacological doses of estrogen induce a superficially similar hypolipidemic response (10-12). In these cases, it has been demonstrated that the hormone causes an approximate 10-fold increase in the levels of hepatic lipoprotein receptors that are believed to be responsible primarily for the clearance of chylomicron and VLDL remnants.
The effects of physiological concentrations of the hormone * This work was supported by research grants to R. G. D. from the Ontario Heart Foundation and the Medical Research Council of Canada and an Ontario Heart Foundation postdoctoral fellowship to S-P. Tam. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The abbreviations used are: VLDL, very low density lipoprotein(s); HDL, high density lipoprotein(s); LDL, low density lipoprotein(s); apo, apolipoprotein. on apolipoprotein synthesis and clearance have been less well defined and have relied in large part on in vivo studies of apolipoprotein turnover rates. These data have in turn been used to deduce approximate rates of synthesis. As yet, no data are available to indicate whether or not estrogens alter apolipoprotein production at a pretranslational level with the exception of studies carried out on oviparous species (4,13,14).
In this and the accompanying article (15) we have examined the effect of 178-estradiol on the production of several apolipoproteins in the human hepatoma cell line, HepG2. This cell line has been shown previously to synthesize many serum proteins, including the major apolipoproteins that are characteristic of parenchymal cells (16-18). Thus it is of considerable potential interest as a model system for examining regulation of lipoprotein synthesis and metabolism. At the moment, the effects of sex steroids on protein production in these cells have not been characterized. We have demonstrated in this manuscript: 1) that HepG2 cells do contain appreciable numbers of high-affinity, nuclear, estradiol-binding sites; 2) that physiological concentrations of the hormone increase the secreted levels of two major apolipoproteins, apo-C-I1 and apo-A-I; 3) that this induction can be antagonized by testosterone; and 4) that the density distribution of the major secreted apolipoproteins agrees well with data derived from perfused liver systems.

RESULTS
Growth and maintenance of HepG2 cells in medium supplemented with 178-estradiol (20 nM) or medium containing charcoal-treated serum had no detectable effects on their doubling time or viability when compared with control cultures. The levels of high-affinity estrogen-binding proteins in nuclei isolated from cells cultured in the presence and absence of estrogen and in medium containing complete or charcoaltreated serum were estimated by saturation exchange assay at 37 "C ( Fig. 1) as described under "Methods." When cells were cultured for 2 days in medium containing complete fetal calf serum and 17P-estradiol (2 X lo-' M), the levels of nuclear estrogen-specific binding increased 6-to 7fold relative to control cultures. Exposure of the cells to 2 x M and 2 X 10"j 178-estradiol resulted in an additional 15 and 25% in nuclear binding levels, respectively. Maintenance of the cells in charcoal-treated serum for a period of 2 days resulted in approximately a %fold reduction in specific nuclear binding capacity, relative to control cultures. Addition of 17P-estradiol to the medium under these conditions induced a 13-to 14-fold increase in nuclear binding sites during the subsequent 24-48 h, so that saturation levels were similar in estradiol-treated cells cultured in either complete or charcoaltreated serum (Fig. L4).
Affinities of the nuclear binding sites were examined in both control cells and cells treated with 20 nM 17P-estradiol. A typical titration curve for hormonally treated cells is shown in Fig. 1B. The shape of the curve, especially in the case of nuclei from estrogen-treated cultures, suggested the presence of at least two binding components of different affinities. A Scatchard plot of the binding data obtained with nuclei from treated cells is shown in Fig. 1C. The plot was resolved into two components as described by Rosenthal(28). One of these had a KO of 9.5 x 10"' M, the other a KO of 7.25 X M. In estrogen-treated cells, the high-affinity component was estimated to comprise approximately 25% of the total binding, which, based upon the specific activity of the ligand and the number of cells used per assay, suggests the presence of approximately 3000-3500 sites/nucleus.
The low levels of high-affinity binding in nuclei from control cultures precluded obtaining data of sufficient accuracy to warrant estimation, by Scatchard analysis, of the proportion of binding sites which were of high affinity. In order to obtain such an estimate, we took advantage of the observation that exchange with high-affinity nuclear receptors does not occur a t 0 "C (29). Saturation assays of both control and estrogen-treated nuclei were carried out a t 0 and 37 "C, and the fraction of sites that were of high affinity was calculated from the difference between the two curves. This method also indicated that in both cases the high-affinity sites accounted for approximately 25% of the total estrogen-specific nuclear binding.

Effect of Estrogen on Levels of Secreted Apolipoproteim-
The relative levels of apolipoproteins secreted into the culture medium in the presence and absence of 20 nM 17P-estradiol were estimated by immunoprecipitation with lZ5I-labeled monospecific antibodies that have been characterized previously ( Fig. 2.4). The protocol of hormone additions and medium changes is described under "Methods." Estradiol treatment for a period of 24 or 48 h resulted in a 2.0-fold increase in the level of apo-A-I and a 2.5-fold increase in the level of apo-C-I1 present in the medium, relative to those found in control cultures of similar cell density. No significant alterations in apo-C-111, apo-E, apo-B, or serum albumin levels were detected. The increases in apo-C-I1 and apo-A-I levels were obtained when estradiol was added either to dividing cells or cultures that had been confluent for up to 24 h prior to addition of estrogen. However, a decreased response (data not shown) was observed in cells that had been confluent for longer periods prior to hormone treatment. A comparable series of experiments was carried out to assess the effect of testosterone on apolipoprotein levels in control and estrogen-treated cultures. Testosterone (20 nM) alone had no significant influence on the levels of secreted apolipoproteins but completely antagonized the estrogen-induced increase in apo-C-I1 and apo-A-I (Fig. 2 A ) when both hormones were added to the medium concurrently.
Apo-C-I1 has been reported previously to constitute only a minor component of the protein spectrum synthesized by HepG2 cells. As a further check on the results obtained from immunoprecipitations we analyzed apolipoproteins in the culture medium by Western blotting. An example of an autoradiograph of a protein blot that was incubated with anti-apo-C-I1 and anti-apo-E antibodies is shown in Fig. 2B (Miniprint Section). Only two proteins, of the expected size for apo-E and apo-C-11, react with the antibodies. Densitometry of appropriately exposed autoradiographs using apo-E as an internal control yielded results consistent with those obtained by immunoprecipitation.
Density Class Distribution of Newly Secreted Apolipoproteins-Since the distribution of newly synthesized apolipoproteins secreted by the HepG2 cells has not been analyzed and since estrogen treatment could potentially influence not only the levels of the secreted apolipoproteins but also their distribution, immunoprecipitations were carried out on 4 density fractions isolated from culture medium by sodium bromide density ultracentrifugation. In order to avoid complications arising from equilibration between ne,wly synthesized lipoproteins and those present in fetal calf serum, lipoproteins were removed from the serum prior to its addition to the culture medium, either by flotation or by adsorption to Cab-0-Sil.
Removal of the serum lipoproteins had no effect on the hormonal induction of apo-C-I1 and apo-A-I or the lack of induction of the other apolipoproteins. However, in both the presence and absence of estrogen the levels of several of the apolipoproteins (apo-B, -C-111, and -E) were 15-20% higher in cultures supplemented with lipoprotein depleted as opposed to complete serum (data not shown).
Density fractions corresponding to VLDL (d < 1.006 g/ml), LDL (d = 1.006-1.063 g/ml), HDL (d = 1.063-1.21 g/ml), and free apolipoproteins were isolated from 30-ml aliquots of culture media, and the distribution of individual apolipoproteins in each fraction was determined (Fig. 3). The results obtained with lipoprotein-depleted serum prepared by both methods were in good agreement, exhibiting less than a 10% variation in all cases. In control and estrogen-treated cultures, less than 8% of any of the apolipoproteins were found in a density fraction (d > 1.21) in which the free proteins would be expected to reside. In contrast, 97% of serum albumin was recovered in this fraction. The majority of apo-B (68.5 f 1.3%, n = 4) was present in a VLDL density range, the remainder being in the LDL density fraction. The VLDL,, LDL, and the HDL fractions contained 29.5 f 1.3, 10.5 f 2.1, and 55.3 f 1.5% of the apo-C-I11 and 39.8 f 2.2, 13.3 f 1.7, and 40 & 1.7% of the apo-E, respectively. Estrogen had no effect on the distribution of these apolipoproteins but did significantly alter the distribution of apo-C-11.
In control cultures, 42 f 2, 6.5 f 0.7, and 44 f 2.8% (n = 4 in all cases) of apo-C-I1 were recovered in VLDL, LDL, and HDL fractions, respectively, while the distribution following estrogen treatment was 29.5 f 2.1, 8 f 0.5, and 56.5 f 5.0%. This alteration in distribution reflects a 55 f 11% increase in the amount of apo-C-I1 recovered in the VLDL fraction as opposed to a 182 f 11.3% increase in the HDL fraction. Thus, approximately 65% of the additional apo-C-I1 found after estrogen treatment remains associated with the HDL fraction.
Although estrogen induced a 2-fold increase in the rate of accumulation of secreted apo-A-I, it did not alter the density distribution of the apolipoprotein. In both control and hormonally treated cultures 91.0 f 1.4% of the apo-A-I was present in the HDL density fraction, the remainder being divided more or less equally between the LDL and free protein fractions.

DISCUSSION
Estrogen administration is known to alter several hepatic functions in humans, ranging from increased secretion of specific plasma proteins to alterations in drug metabolism (30). High-affinity, low-capacity, estrogen-specific binding proteins have been identified in livers of several mammals including man, suggesting that the hormone may exert a direct effect on the tissue (30,31). The few studies that have been carried out on human liver have examined only cytosolic fractions. No data have been published on the levels of nuclear high-affinity binding proteins in human liver. Thus, prior to examining the effects of estrogen on apolipoprotein production in the HepG2 cell line we carried out a series of experiments to determine the number of high-affinity, nuclear, estrogen-binding sites in these cells.
The level of nuclear binding sites found in control cultures was extremely low and decreased still further when the cells were cultured in the presence of charcoal-treated rather than complete serum. However, when the cells were maintained in estradiol and complete serum for 36-48 h prior to assay we obtained a 6-to 7"fold increase in total estrogen-specific binding compared with control cultures. From analyses of these data we have attributed the binding to two components.
The high-affinity component has a KD = 9.5 X 10"' M and appears not to undergo steroid exchange at 4 "C. In these

FIG. 3. Effect of estrogen on the distribution of apolipoproteins between various density fractions.
Various density classes of lipoproteins were isolated from cells grown in medium supplemented with lipoproteindeficient calf serum in the absence (-) or presence (+) of 17P-estradiol (20 nM), respectively. VLDL was isolated at d < 1.006 g/ml; LDL was isolated from d = 1.006 to 1.063 g/ml; HDL was isolated from d = 1.063 to 1.21 g/ml, and free proteins were isolated in the d > 1.21 g/ml infranatant. An aliquot of each fraction was analyzed by immunoprecipitation using various '251-labeled monospecific antibodies as described under "Methods." Results presented are the mean of two experiments from cells grown in lipoprotein-deficient serum prepared by ultracentrifuge (-) and Cab-0-Si1 treatment (---).
respects it is similar to high-affinity mammalian receptors characterized in other estrogen target tissues such as uterus (29). The number of high-affinity binding sites detected after estradiol pretreatment of the cells is 3000-3500/cell. While this is low when compared to levels found in uterus, it is in the same range as that found in avian liver following induction by estrogen (32). The lower affinity component we detected has an apparent KO = 7.25 X lo-' M. Rat uterus has been shown to contain a nuclear, lower affinity, estrogen-binding protein, designated Type I1 receptor by Clark (33), that has a Kr, of 3.3 X IO-' M. However, this protein exhibits cooperative binding of estradiol, while the lower affinity component that we identify apparently does not. Its dissociation constant is also comparable to cytosolic "high capacity-lower affinity" binding proteins that have been identified in livers from several species, raising the possibility that some of the lower affinity binding we detect may be attributable to contamination with cytosolic binding proteins (34).
Maintenance of the cells in a wide range of estradiol concentrations indicated that adjustment of the medium to 20 nM in hormone every 12 h was adequate to maintain nuclear binding at 90% of the maximal levels obtainable. This regimen of addition was followed for all experiments. The initial level of hormone used is approximately 20-fold higher than peak circulating concentrations of estradiol found in premenopausal nonpregnant women, but since hepatocytes metabolize estradiol extremely rapidly the effective steady-state concentrations of the hormone in the culture system are likely to be considerably lower. Consequently, it is not possible to equate the nominal concentrations in the cultures with circulating in vivo levels. Under the conditions examined, estradiol was found to increase the rate of accumulation of secreted apo-C-I1 and apo-A-I, 2.5-and 2-fold, respectively, while having no effect on apo-C-111, apo-E, and apo-B.
Apo-C-I1 and apo-A-I are important regulators of triglyceride and cholesterol metabolism. The former is an activator of lipoprotein lipase (35) and the latter an activator of lecithin-cholesterol acyltransferase (36). At the moment, the compositions of nascent VLDL and HDL secreted by human liver are not known. Based upon studies with animal models, both apoproteins are believed to be secreted from the liver associated in a nascent HDL particle, from which apo-C-I1 subsequently transfers to VLDL and chylomicrons (37,38). If this assumption is correct the presence of apo-C-I1 and apo-C-I11 in the VLDL fraction we have examined suggests that redistribution of apoproteins may occur in the culture medium in a manner similar to that observed in uiuo.
The effects of estrogen on apolipoprotein density distribution were to elevate the apo-C-1I:apo-C-I11 ratio in both VLDL and HDL by increasing the total amount of apo-C-I1 in those fractions and to double the apo-A-I content of HDL. Since apo-A-I is the major protein constituent of HDL this results in a significant decrease in the overall VLDL:HDL ratio in the medium. In this respect the response we observe is consistent with epidemiological data indicating that premenopausal women maintain lower VLDL:HDL ratios than age-matched males and with studies indicating that transient elevations in HDL levels can be correlated with ovulation when estradiol levels are at a maximum (7). One of the most recent studies on the effects of estrogen administration on the synthesis and turnover of apolipoproteins in premenopausal women also concluded that the hormone significantly increased the rate of synthesis of apo-A-I, but apo-C-I1 was not examined (39). In addition, these authors reported that they observed a 1.8-fold increase in the rate of VLDL apo-B synthesis, calculated on the basis of steady-state levels and turnover rates, in subjects given 0.1 mg of ethinyl estradiol/ day. This observation is consistent with the results of several studies documenting increases in VLDL levels during estrogen therapy (7, 40, 41). Recent experiments3 using significantly higher levels of estradiol indicate that estradiol levels at least 25-fold higher than those required to increase apo-C-I1 and apo-A-I production are necessary to increase the levels of apo-B secreted by the HepG2 cells. Further experiments with slowly metabolized estrogen analogues will be required before we can place more accurate estimates on the threshold concentrations necessary to induce these two different effects.
In contrast to the results obtained with estrogen, testosterone had no discernible effect on the secreted levels of any of the apolipoproteins examined. However, at an initial concentration of 20 nM it completely prevented the estrogen-induced increase in apo-C-I1 and apo-A-I. At the moment, the mechanism involved is not known. In several species, androgens have been shown to increase hepatic levels of high-capacity, cytoplasmic, estrogen-binding proteins that may act as an intracellular "sink" for the hormone (34). However, such proteins remain to be identified in human liver and in HepGZ cells.
The analyses of apolipoprotein levels reported here have estimated accumulated levels of apolipoproteins in the culture medium and so reflect the balance between rates of synthesis and turnover (42). Data from primary hepatocytes indicate that little if any turnover of newly synthesized apolipoproteins occurs suggesting that rates of accumulation in the medium do provide an acceptable indication of synthetic rates. This has yet to be confirmed for HepG2 cells.
Despite the possibility that some metabolism may be taking place, it does appear, at least in the case of apo-C-I1(15), that the major effect of estrogen is to increase the rate of synthesis of the apolipoprotein and that it does so by elevating the levels of apo-C-I1 mRNA. Similar studies are in progress to determining if this also applies to the induction of apo-A-I, apo-E, and apo-B.