- Email: [email protected]
Solubility of retinoids in water
ARCHIVES
OF BIOCHEMISTRY
Vol. 257, No. 2, June,
AND
RIOPHYSICS
pp. 297-304.
1991
Solubility of Retinoids in Water Ete Z. Szutsl and Ferenc I. Harosi Laboratory of Sensory Physiology, Marine and Department
Received
of Physiology,
December
Boston
17, 1990, and in revised
Biological Laboratory, Woods Hole, Massachusetts [Jniuersity School of Medicine, Boston, Massachusetts
form
February
6, 1991
Spectrophotometric and radioactive techniques were used to measure the water solubility of retinaldehyde, retinol (vitamin A), and retinoic acid under physiological conditions. Hydration decreases the molar extinction coefficient of these substances and shifts their absorption peak bathochromically (10 nm for retinal and = 1 nm for the rest). We find their solubility to be about 0.1 pM at room temperature, pH 7.3 (with experimental values being 0.11 fiM for retinaldehyde, 0.06 yM for retinol, and acid). To prevent oxidative degra0.21 pM for retinoic dation of retinol, which is the most labile retinoid, our argon-saturated buffer solutions contained physiological levels of ascorbate or a-tocopherol. To the best of our knowledge, water solubility of these compounds has not yet been previously reported. Although the measured solubilities are relatively low, they are significant and may account for the movement of retinoids through the aqueous phase as observed by others during exchange with binding proteins and during intervesicular transfer in the absence of binding proteins. Diffusion of uncomplexed retinoids through the aqueous phase can be a major pathway for transport over subcellular distances. L 1991
Academic
Press,
Inc.
Retinoids, which include vitamin A (retinol) and its chemical derivatives, regulate and control diverse physiological functions. Because these compounds are lipophilic, binding proteins act as carriers for their inter- and intracellular transport. Several types of retinoid-binding proteins are known to exist [for a recent review, see (l)]. Irrespective of their cellular location, all retinoid-binding proteins share the ability to transfer (bind and release) their ligand. The actual molecular mechanism of this transfer reaction has been in dispute. One commonly cited mechanism involves membrane receptors, which act as docking sit,es for the binding protein and which may also ’ 7’0 whom
correspondence
should
be addressed.
02543; 02118
catalyze the loading and unloading of the ligand [e.g., (a)]. An early observation (3) suggested that receptors may be unnecessary for transfer, as retinoids could partition into the aqueous phase. This suggestion was later corroborated by several investigators (4-6), who directly measured retinoid transport among vesicles and also between vesicles and binding proteins. Retinol was found to pass through the aqueous phase, whether retinol-binding protein from the serum (4, 5) or interphotoreceptor retinoid-binding protein (IRBP)’ from the eye (6) mediated the transfer process. Although an aqueous transfer mechanism requires retinoids to be sufficiently water-soluble, solubilities were neither cited nor investigated in these studies. In fact, it appears that water solubility has not yet been measured for any retinoid. The only quantitative statement available in the literature is a citation to unpublished data that solubility of retinol is less than 1 nM (3). In the absence of this information, we set out to measure the solubility” of the physiologically most relevant retinoids. For these studies we specifically chose spectrophotometric techniques as our primary assay, because (i) electronic absorption spectra are indicative of chemical properties and purity, (ii) absorption spectra of retinoids are well characterized, and (iii) spectroscopic measurements allow direct monitoring of a solute within an aqueous environment, without the need for additional handling or processing that may perturb the results. Complicating factors included light sensitivity of the solute and its susceptibility to oxidative damage, as well as ’ Abbreviations used: IRBP. interphotoreceptor retinoid-binding protein; A,,,,, wavelength of peak absorbance; t,,,, molar extinction coefficient at A,,,. ” We use the term “soluhility” in its general meaning. Soluhility of a solid is generally defined as the amount of dissolved solute in equilibrium with its solid state. As the nature of dissolution is not a part of the definition, soluhility of an amphipathic molecule includes all its dissolved forms (monomers, multimers, and micelles). For example, the Merck Index gives the soluhility of sodium dodecyl sulfate as “1 gram dissolves in 10 ml water” which is equivalent to 0.35 M, about two orders of magnitude ahove critical micelle concentration for this detergent.
295 Copyright All
rights
c 1991 by Acxlcmic Press, 01 reproduction in any form
Inc. resen~d.
298
SZUTS
AND
its strong tendency to adsorb to glassware. This report, which is limited to establishing the solubility values, forms the basis of future studies on other biologically relevant parameters. Some of our results were previously presented in abstract form (7). MATERIALS
AND
METHODS
Materials. Retinal (Eastman Kodak Co., Rochester, NY), retinol, and retinoic acid (both from Fluka Chemical Corp., Ronkonkoma, NY) were purchased as all-trans isomers and stored at -20°C under argon. New England Nuclear (Boston, MA) supplied [3H]retinol (47 Ci/mmol, 21 ELM) in ethanol with 2.32 mM d-ol-tocopherol. [I-i”C]Ethanol was prepared by American Radiolabeled Chemicals, Inc. (St. Louis, MO). Spectroscopy grade ethanol and hexane were purchased from Quantum Chemical Corp. (Tuscola, IL) and Fluka Chemical Corp., respectively. Solvents for HPLC were obtained from Fisher Scientific Co. Glassware and cuvettes were thoroughly cleaned between experiments with successive rinses of acetone, chromic acid, and water. Solubilities were tested in a physiological saline solution (also called buffer) that contained 150 mM NaCl, 1 mM Na-phosphate, pH 7.2-7.3, dissolved in HPLC-grade water. To reduce oxidation, all polar solvents were thoroughly purged with argon for 0.5-l h. prior to use and in the case of retinol the buffer also contained ascorbate or o-tocopherol. Purging with argon reduced oxygen tension in the buffer by 95%, as measured with an oxygen electrode (Orbisphere Lab., Geneva, Switzerland). Spectroscopy. Spectral analysis was performed with a photodiodearray uv-vis spectrophotometer (HP 8452A that was operated with HP 89500A Chemstation; Hewlett-Packard Co., Palo Alto, CA). Most samples were measured in lo-cm long cylindrical quartz cuvettes (30 ml capacity). Illustrated spectra represent “raw data,” obtained in single scans without any additional processing. Zero absorbance was assigned to flat portions of the spectra on the longwave side of the main absorption band. Stock solutions were prepared by dissolving the crystalline retinoids in either argon-purged ethanol or hexane and were handled in dim red light. Retinoid solutions were stored for short periods at 4”C, and monitored regularly before use by recording their absorption spectra. They were discarded for any discernable distortion. As a result, retinol stocks were prepared on the day of the experiment,, while ethanolic solutions of retinal and retinoic acid were usually good for several days. In absolute ethanol, all-trans ret,inal absorbed maximally at 382 nm with a molar extinction coefficient of 42.4 (k3.3) X 10” (M-cm)-‘, in excellent agreement with a previous report (8). For all-trans.retinoic ’ at 341 nm in ethanol acid we measured emarto be 49.7 X 10” (M-cm) that agrees well with 45.2 X lo” (9). We determined retinol’s extinction in ethanol, coeflicient at its h,,, of 325 nm to be 38.3 X lo” (M-cm)-’ which is less than its previously published value of 45.1 X 10” (10). In spite of its lower extinction, our retinol eluted as a single peak by HPLC analysis with a purity of >99%, in agreement with the manufacturer’s specifications for the chemical. HPLC. Retinoids were extracted from the physiological solution with hexane (3.0 ml hexane to 28 ml saline) with an efficiency of 92%, as determined with [3H]retinol. About 100 ~1 of the extract was injected for HPLC analysis, performed on a Waters system (Waters Associates Inc., Milford, MA) with an Alltech silica column (Econosphere Silica 5 pm; 250 mm X 4.6 mm; Alltech Associates Inc., Deerfield, IL). The mobile phase was run at 1 ml/min as follows: pure hexane (solvent A) for 5 min after injection, linear gradient of solvent B (ethylacetate/ methanol, 9/l) at I%/min for 20 min, linear gradient of 3% B/min for 10 min, and finally constant 50% B for 10 min. The amount of water in the “saline-saturated” hexane samples [5 mM, (ll)] did not significantly affect retention times (21.4 F 0.2 min for retinal and 26.6 min for retinal). Dilution of ethanolic retinoid solutions with buffer. For these sets of experiments, a series of ethanolic stock solutions were mixed with buffer
HAROSI so that final concentration of total retinoid remained constant for the mixtures (at about 0.3 PM), while the ratio of ethanol to water was varied. The mixtures were analyzed spectroscopically. Each dilution was performed directly in a clean cylindrical quartz cuvette. As long as the final ethanol concentration of the mixture was 50% or greater, the dilution factor completely accounted for final absorbance indicating that 6 for each retinoid remained unchanged from its value in pure ethanol. However, adsorption to the cuvette and changes in e became progressively pronounced when ethanol was <250/o. To control for these factors, we used the following protocol: (i) the absorbance spectrum of the solution to be analyzed was recorded (initial spectrum); (ii) the content of the cuvette was decanted into a Aask and mixed with an equal volume of pure ethanol, so that final ethanol in the flask was 350%; (iii) 28 ml of pure ethanol was added into the cuvette to dissolve all adsorbed retinoid and the absorbance spectrum was measured (spectrum of residue-incuvette); (ivj the mixture in the flask was transferred to a new clean cuvette and its spectrum recorded (spectrum of 250% ethanol mixture). Throughout this procedure, cuvettes and flasks were weighed to accurately measure both transferred and retained volumes. The baseline for all the spectral measurements came from a control cuvette that was processed in parallel with and in the same manner as the test cuvette, but lacked retinoid. From the three recorded spectra, the following variables were unambiguously calculated: mol of retinoid dissolved in the initial solution to be analyzed (calculated from absorbance and volume of 350% ethanol mixture), solubility and tmaxof retinoid in the initial solution (calculated from previously derived “mol of retinoid dissolved” and from volume and absorbance of initial solution), and mol of quartzadsorbed retinoid (from absorbance and volume of residue-in-cuvette). Solubility experiments. In these experiments, a controlled amount of solid retinoid was deposited in excess into the cylindrical cuvette, by pipetting l-10 ~1 of ethanolic stock on the bottom of the cuvette and by evaporating the ethanol with a stream of argon for 5 min. The drying process removed 99.99% of the ethanol, as measured with [‘4C]ethanol. After the argon purge, 28 ml argon-saturated buffer was pipetted into the cuvette. Additional argon was blown through its two ports before the cuvette was stoppered with teflon plugs and placed on a gentle mixer. The absorbance of the aqueous solution (buffer spectrum) was measured 30-60 min after mixing. The same identical analysis was performed with it as described above for the diluted ethanolic stocks, generating an analogous set of three spectra (buffer spectrum, spectrum of >50% ethanol mixture, and spectrum of residue-in-cuvette). Within the spectrophotometer cuvette, retinoid can exist in three “phases:” freely dissolved in bulk phase, adsorbed to quartz surfaces, and aggregated at the air-water interface. The solution decanted from the cuvette is not representative of the cuvette’s bulk solution if the decanting process transfers significant amounts of retinoids from the two kinds of interfaces. The extent of these effects was measured in experiments with [“Hlretinol, by comparing spectral measurements with the radioactivity contained in 1 ml buffer, sampled entirely from the bulk phase. Such analysis indicated that the contribution of the surface efIects was negligible. All experiments were performed at room temperature (21-23°C).
RESULTS
Effect of hydration on spectral absorbance. The effect of water on the spectral properties of retinoids was determined in a series of experiments in which ethanolic retinoid solutions were diluted with buffer. Fig. 1 illustrates the general aspects of hydration as determined for retinaldehyde. As the mole ratio of ethanol to water decreases (representing increased hydration) the first observed effect is a bathochromic shift. This shift (from 382 nm in pure ethanol to 392 nm for ~25% ethanol) is ac-
SOLUBILITY
OF RETINOIDS
,./ 20 t
FIG. 1. Effect of hydration on X,,, and molar extinction coefficient for all-trans-retinaldehyde. A, normalized absorbance spectrum in buffer (solid curve) and in absolute ethanol (dashed curve). Note the bathochromic shift and broadening of the peak upon hydration. B and C, X,,, and fmsxvariation as function of ethanol concentration, plotted as mole ratio, in the mixed solvent system of ethanol/water. For mole ratios of alO-“, measurements were made with mixtures of ethanol/buffer into which ethanolic stock was directly added. The data point at 10 -s mole ratio refers to a buffer measurement for which the ethanolic stock was dried before the buffer was added. For all points, added stock contained a constant amount of retinal (9 nmol) so that the final retinal concentration was 0.325 pM, if completely dissolved. Solutions were mixed for about 30 min before spectral analysis. Data points are the averages of at least three independent determinations. Error bars represent standard deviation, and if absent, are less than the symbol size. Conversion between ethanol/water mole ratio, ethanol molarity, and volume percent are 10’ = 17.1 M, 99.7% v/v; 10 = 16.6 M, 97%; 1 = 13.1 M, 76.4%; 10-i = 4.2 M, 24.5%; lo-’ = 0.54 M, 3.1%; 10m3= 55 mM, 0.32%; 10 ’ = 5.5 mM, 0.03%; 10m6= 55 PM, 3 X 10m4%; lo-* = 0.55 PM, 3 X 10 ‘%.
companied by an increase in the half-width of the main absorption band (Figs. 1A and 1B). Upon additional hydration, the amplitude of the band (molar extinction) diminishes, beginning at about 25% ethanol (Fig. 1C). The observed decrease in E,,,~~,which in buffer drops ultimately
IN WATER
299
to about 43% of the extinction in pure ethanol, is not accompanied by additional shifts in X,,,. We interpret the observed effects to represent titration of two sites or regions on retinaldehyde: one near the polar end, the other along the polyene chain. The wavelength shift occurs at a mole ratio of 1, indicating that water and ethanol molecules can equally displace each other near the polar end. However, invasion by water molecules at the second site along the conjugated bonds is resisted by the hydrophobic interactions between ethanol and the polyene chain. As a result, the solution must contain 100 or more water molecules per ethanol before a significant drop in peak extinction occurs. The solubility of retinaldehyde in a mixed ethanol/water solvent system increases with increasing ethanol fractions, as expected. When the ethanol mole ratio was 20.1, all the retinaldehyde was dissolved, yielding a concentration of 0.33 PM in these experiments. As we reduced the ethanol mole ratio, the solubility of retinaldehyde in the bulk phase diminished to 0.14,0.11, and 0.11 PM at ethanol concentrations of 3%, 0.3%, and 0.03%, respectively. In addition, the dissolved retinaldehyde had a lower extinction coefficient, as shown in Fig. 1C. Remaining retinaldehyde was adsorbed to the walls of the cuvette. Although we did not study the time-course of this precipitation, our spectrophotometric observations were made about 30 min after the introduction of the solute into the ethanol/buffer mixture. The surprising outcome of these experiments is that the solubilizing action of ~3% ethanol solutions is very low, even though the ratio of organic solvent to solute is still high (e.g., mole ratio of ethanol-to-retinal is 1.5 X 10” at 3% ethanol). The data point at a mole ratio of lo--’ in Fig. 1B and Fig. 1C represents buffer solutions equilibrated with argon-dried retinoid sample. The amount of ethanol remaining after drying was measured with [‘4C]ethanol and was found to be 0.01% of the initial ethanol volume, which was mainly bound nonspecifically to the cuvette. Using this value, we calculate the mole ratio of ethanol to water in our buffers to be lOPa, equivalent to the presence of about 0.6 /*M ethanol. This amount of retained organic solvent does not appear to have altered the solubility of retinaldehyde in buffer, because (i) we observed the same solubility even when ethanol mole ratio was decreased another lo-fold (by the use of smaller volumes of more concentrated stocks), and (ii) the ethanol concentration had to exceed 3% before increased solubilizing action of the organic solvent could be demonstrated, as discussed above. Solubility of retinaldehyde. As previously described at greater length, the central feature of our spectral analysis for water solubility is that the solute’s concentration in buffer can be determined from the absorbance of the solution before and after it is diluted with ethanol. We chose
300
SZUTS
AND
Absorbance spectra for solubility determination of all-transFIG. 2. retinaldehyde. Retinal’s spectrum in buffer was recorded first in a locm cuvette (bottom trace). The buffer was decanted from the cuvette and mixed with an equal volume of 100% ethanol and recorded with a separate cuvette. The spectrum of this buffer-ethanol mixture was doubled for display here (middle trace) to clearly demonstrate the relative extinction coefficient in the two media without the dilution effect. The top trace is the spectrum of retinaldehyde remaining in the initial cuvette, with the residue solubilized in 100% ethanol.
equal volume dilutions for all retinoid experiments, because such a dilution yields a 50% ethanol solution for which retinoid adsorption to the cuvette is absent and for which emaxis known to be the same as for pure ethanol. Representative experimental spectra for calculating retinaldehyde’s relative extinction coefficient and solubility are presented in Fig. 2. The average solubility we measured for all-trans-retinaldehyde in our buffer was 0.11 PM (Table I). The purity of retinal dissolved in buffer was verified by HPLC techniques. Buffer solutions were extracted with hexane and the absorption spectrum of the extract was measured before it was analyzed by HPLC (Fig. 3). Based on chromatograms obtained at three wavelengths (370,
TABLE
HAROSI
320, 280 nm), the sample’s content was at least 95% alltruns-retinal. Assuming that the efficiency of extraction is the same for retinoids and oxidation products, the result with HPLC verifies that all-trarzs-retinal was the predominant compound present in buffer. Thus, the existence of retinal in an aqueous environment is confirmed by two independent analytical techniques. Solubility of retinol. As reported by others (12), retinol is unstable in aqueous solutions in the absence of antioxidants. To prevent its degradation, we supplemented our regular argon-purging of solutions with the addition of either ascorbate to the buffer (1 mM) or a-tocopherol to the ethanolic stock solutions (mole ratio of retinol to tocopherol being 7-9). Not only have these antioxidants been used previously with retinoids (13,14), they are also physiological. For example, the ascorbate concentration is 0.5-l mM in the retina and pigment epithelium (15) and the concentration of vitamin E is -0.3 mM in rod outer segments (16). Being a highly polar molecule, vitamin C is not expected to affect the solubility of retinol. On the other hand, its main absorption band [h,,, = 266 is very near to that of nm, hx = 12 X 10” (M-cm)m1] retinol so it interferes with the spectroscopic determination. In order to avoid this problem, we extracted retinol from buffer with hexane (as performed for HPLC analysis), thereby effectively separating vitamin C from it. Knowing the extraction efficiency, retinol’s concentration in buffer was calculated from its spectrum in the hexane phase (Fig. 4). This way (method #l in Table I) we obtained a solubility of 0.05 PM for retinol in our buffer solutions. Due to the indirect nature of this determination, method #l probably yields the least accurate result. Spectral interference was greatly reduced with the second antioxidant. Although cr-tocopherol also absorbs in the same spectral region as retinol, its contribution is
I
Solubility and Extinction Coefficient of All-trans.retinoids Relative extinction coefficient (buffer relative to that in ethanol) (%)
Retinoid Retinaldehyde Retinol (vitamin Method 1b Method 2 Method 3 Retinoic acid
in Water
Extinction coefficient (t,,,) in buffer” (M-cm) ’ X lo”
8 (n = 7)
17.9 * 3.5 (n = 7)
46 k 19 (n = 5)
17.8 f 7.3 (n = 5)
75 k
37.6 f 1.0 (n = 3)
42 +
Concentration
in buffer
(FM) 0.108 + 0.016
(n = 7)
0.051 0.079 0.056 0.212
(n (n (n (n
A)
2 (n = 3)
f f f *
0.017 0.026 0.010 0.022
= = = =
8) 5) 5) 3)
’ The absolute value of the extinction coefficient in buffer was calculated from the experimentally determined value for relative extinction 42.4 X 10” for retinal, 38.3 X 10” for retinol, and 49.7 X lOa for retinoic acid (all coefficient (column 1) and from the following ernar in ethanol: being our determinations). * Method 1: buffer contained 1 mM ascorbate and the spectrophotometric analysis is based on hexane extract; method 2: buffer contained etocopherol (mole ratio of retinol/tocopherol = 7-9) and analysis performed spectrophotometrically; method 3: same as method 2 except data from radioactive assays.
SOLUBILITY
OF
RETINOIDS
IN
WATER
301
FIG. 4.
Absorbance spectra of all-trans.retinol in aqueous buffer (solid (dashed curves; X,,, = 325 nm). curve; X,,, = 326 nm) and in n-hexane Top trace, absorbance of the hexane phase of a control experiment in which the dried retinol sample was dissolved in hexane before a similar volume of buffer was added. Middle trace, absorbance of the hexane phase after extraction from buffer. Bottom trace, retinol in aqueous buffer. In this experiment, the retinol stock contained n-tocopherol (the mole ratio of retinol to tocopherol was 9).
FIG. 3. Retinaldehyde extraction from buffer into hexane. A, retinal’s absorbance spectrum in buffer prior to extraction, as measured in 10. cm long cuvette (solid curve). Retinal’s absorbance in hexane following extraction, as measured in l-cm long cuvette (dashed curve). B, HPLC chromatogram of same hexane extract. The major peak is all-transretinaldehyde, as determined by standards. Minor peaks marked with * are solvent impurities. Top trace indicates gradient profile and is a plot of percentage B. Time delay between gradient controller and uv detector is 8.7 min. Column was Econosphere silica, 5 +m; 250 mm X 4.6 mm; sample was 80 ~1 hexane extract; flow rate, 1 ml/min; mobile phase A, hexane; mobile phase B, ethylacetate/methanol (9/l); pressure, 400 psi: detector X, 370 nm.
small because (i) it has a lower t,,, [2 X 10” (M-cm).. ’ at x max = 286 nm] and (ii) lower concentrations of it are needed for effective protection of retinol. Since vitamin E is already present in commercially supplied solutions of radioactive retinol, we could use the same samples for bothspectrophotometricandradioactiveanalyses.Retinol solubility by spectrophotometric analysis was 0.08 PM in the presence of vitamin E (Fig. 4 and method #2 in Table I). The corresponding result with radioactivity was found to be 0.06 yM (method #3 in Table I). When the experimental results of the three methods are averaged, the solubility of retinol is 0.06 PM. Solubility of retinoic acid. Compared with retinal, retinoic acid’s behavior to hydration is anomalous. We find the peak absorbance of the acid in absolute ethanol and pure buffer to be essentially the same at 341 nm (Fig. 5). This value, however, does not remain stationary in a mixed solvent system of ethanol and water: in 50% and 98% ethanol, we find a hypsochromic shift of 3 and 4 nm, respectively. The molar extinction coefficient in the
aqueous phase decreases by only -25% from that in ethanol. Solubility of retinoic acid was determined spectrophotometrically to be 0.21 PM (Table I). DISCUSSION
Solubility values. Our results indicate that the three unesterified retinoids have an aqueous solubility concentration of about 0.1 PM, plus or minus a factor of two. As we kept the time interval for dissolution relatively short (30-60 min), the experimental values could be considered an underestimate of the actual solubilities. However, additional considerations suggest that dissolved solutes were essentially in equilibrium with the solid phase in our solutions. In the binary solvent systems that we prepared by mixing buffer with ethanol solutions, retinoid solubility was limited at low ethanol concentrations, even though
FIG. 5.
Absorbance spectra for solubility determination of all-transretinoic acid. Middle trace, retinoic acid’s spectrum in buffer. Top trace, twice the absorbance of the 50/50 ethanol/buffer mixture. Bottom trace, absorbance of the ethanol-dissolved cuvette residue.
302
SZUTS
AND
retinoid was fully solvated by ethanol at the time of mixing. This was illustrated for retinaldehyde, for which we found final concentrations to be 0.11-0.14 PM in 0.03-3% ethanol (Fig. 1). In other experiments, in which retinaldehyde was dried before the introduction of the aqueous buffer, final concentration reached 0.11 pM (Fig. 2 and Table I). Thus, solubility approached the same value whether the procedure started with undissolved solute (solid) or with solute dissolved in ethanol. Therefore, our experimental measurements on retinoid solubility were probably near equilibrium. The elution profiles observed in reverse-phase chromatography are indicative of relative solubilities [for review, see (17)]. This is because partitioning between aqueous mobile and organic stationary phases is the primary mechanism of separation by reverse-phase HPLC, leading to chromatograms where polar substances elute first from the column. For retinoids, the observed order of elution is retinoic acid, followed by retinol and then by retinaldehyde (14,18). Since greater polarity increases the water solubility of a substance, the order of elution suggests that retinoic acid is the most, and retinaldehyde the least, soluble. The fact that we found solubility of retinol to be the lowest of the three retinoids may be due to the use of antioxidants. In the absence of antioxidants, retinoids and especially vitamin A rapidly decompose in aqueous solutions. We confirm the observation of others (13, 14) that the presence of ascorbate and a-tocopherol retards retinoid degradation. However, the use of ascorbate interferes with the spectroscopy of aqueous solutions of retinol and made solubility determinations indirect and possibly less accurate. The use of a-tocopherol may have also interfered in these measurements, due to its hydrophobic nature. Its water solubility, which remains unknown, is probably less than that of retinol on the basis of its longer retention time by reverse-phase HPLC (19). When ethanolic stock solutions containing both retinol and tocopherol were dried in our experiments, the two vitamins could have become associated with each other. Such an association may have retarded the availability of retinol for dissolution and may have been the cause for an underestimated solubility value. Additional studies are needed to fully determine the effects of tocopherol on retinoid solubility. Retinoid aggregation. Being amphipathic molecules, retinoids may self-aggregate to form complexes that in principle can range from dimers to micelles. Because our solubility determinations were made under conditions where dissolved solute was in equilibrium with its solid state, our experimentally derived solubilities are correct even if micelles formed in our buffers. However, micelle formation is unlikely for the very reason that retinoid solubility is limited. Because the number and size of micelles can grow with increased amounts of added solute,
HAROSI
the solubility of micelle-forming compounds should exceed 0.1 yM. The notion that retinoids form micelles was suggested by two reports (20,21) that studied the polarization of retinol’s fluorescence in buffers that contained x 1 PM retinol and N 1% ethanol. In these studies, the endogenous fluorescence of retinol was found to be more polarized in the aqueous mixture than in pure ethanol, indicating the loss of rotational freedom in the mixture. This result was interpreted by both groups as evidence for micelle formation. However, there are two other possible explanations: adsorption of retinol to cuvette walls (found by us to occur whenever ethanol < 25%) and precipitation of retinol (as it goes out of solution) due to limited solubility. Similar precipitation is known to occur for cholesterol, which forms a suspension of “microcrystals” above its solubility limit of ~10 nM in water (22, 23). With a fluorimetric probe technique used to measure critical micelle concentration (24), we found no evidence for micelle formation by either retinol or retinaldehyde (Harosi and Szuts, unpublished). Although micelle formation is unlikely, the presence of smaller aggregates (such as dimers or trimers) is certainly possible. Of the three retinoids, dimerization is expected for retinoic acid, because hydrogen-bonding between -COO- and -COOH groups is known to be especially strong for long chain fatty acids. Smith and Tanford (25) have shown that tenacious hydrogen bonds are responsible for the formation of dimers and higher aggregates by carboxylic acids at neutral pH, even below critical micelle concentration. However, the extent of dimerizationlaggregation, which they could not determine, remains unknown. The mechanism of multimer formation by retinoids, similarly remains to be investigated.4 Physiological relevance. Is a water solubility of 0.1 PM for retinoids physiologically relevant? The cellular requirement of any retinoid is best characterized for the photoreceptors of the vertebrate retina. Photoreceptors contain about 3.5 mM rhodopsin (27), the visual pigment, whose chromophore is II-cis-retinaldehyde. Photoexcitation leads to hydrolysis of rhodopsin, releasing free alltrans-retinal which is rapidly converted to all-trans-retinol within the photoreceptors. Resynthesis of 11-cis-retinal from all-trans-retinol involves the cells of the pigment epithelium, which abut the receptor cells [ref. (2) for review]. Thus, there is a constant traffic of retinoids between ’ We determined the absorbance spectra of the retinoids in various solvents (not illustrated). In dimethyl sulfoxide, which is a strong hydrogen bond acceptor (26), we found a large shift in X,,, for retinoic acid, but smaller shifts for retinol and retinaldehyde. The corresponding X,,, for retinoic acid, retinol, and retinaldehyde were respectively 341, 325, and 382 nm in ethanol and 361,332, and 386 nm in dimethyl sulfoxide. These results are consistent with the notion that hydrogen-bonding among solute molecules may occur in aqueous solutions of retinoic acid, whereas the same is less likely for retinol and even less for retinaldehyde.
SOLUBILITY
OF
RETINOIDS
these two cell layers of the retina across the interphotoreceptor space. At most levels of illumination a steady state is achieved where the rate of photoactivation equals the rate of rhodopsin regeneration, supported by retinoid transport. For example, when ambient light intensity bleaches 10% of the cone pigments in the human eye, the regeneration rate is about 4% per minute (28). Similarly, regeneration is about 1% per minute in a live frog after 90% of its rhodopsin has been bleached (3). Surprisingly, it is possible to satisfy such rates of delivery by free diffusion alone from our aqueous solubility values. This conclusion is based on the application of diffusion theory (29) to a model system in which a prolate target is bombarded by solute molecules, whose concentration is constant in the surrounding medium. In this analysis, we take the cylindrically shaped outer segment of rod cells to be the target (with I= 60 pm, r = 3 pm) and we assume that (i) 11-&-retinal diffuses as a monomer with D = 0.5 X lo-” cm”/s and (ii) all collisions with the target lead to transfer to the plasma membrane. With these conditions, a solubility of 0.1 PM yields a diffusion rate of 2 X 10R molecules/min to a single rod outer segment. This rate represents the arrival of enough retinoid to regenerate -6% of the rhodopsin in the cell every minute. Hence, the solubility values we determined are sufficiently high to meet the metabolic requirements of photoreceptors without invoking any binding proteins. These results are based on the assumption that the aqueous medium surrounding the cells is saturated with retinoids. The actual free concentration of these compounds remains unknown. Photoreceptors were used for this mathematical model, because they are known to metabolize retinoids at a high rate. However, the analysis and conclusion would be valid for other cells, especially those with reduced metabolic requirements. Our results may explain several unresolved observations in retinal physiology: (a) the absence of a binding protein within rod photoreceptors, where >90% of the chromophore must cross the aqueous cytoplasm separating plasma and disk membranes, (b) the absence of confirmation for the existence of predicted receptor sites for IRBP on the extracellular surfaces of both photoreceptors and epithelial cells, and (c) the nonspecificity of IRBP for the aldehyde and alcohol forms of vitamin A and t,heir various isomers (2). Although binding proteins are not an absolute requirement for retinoid transport over short distances (comparable to the size of cells), IRBP may yet serve as an adjunct of transport in the retina. For example, it can support transport rates in excess of the value permitted by water solubility alone. The relatively high concentration of IRBP in the extracellular space (30), which is about two orders of magnitude greater than the aqueous solubility of retinoids, increases the saturability of the medium surrounding the receptors, thereby increasing transfer
IN
303
WATER
rates between source and target. Additionally, IRBP appears to prevent the loss of retinoids by diffusion through the aqueous phase of the retina. Therefore, IRBP and other cellular binding proteins can serve to localize and concentrate retinoids to regions of special need. Thus, we agree with others who have expressed similar ideas for the visual cycle in the vertebrate eye (6) and for the role of retinoic acid in the developing chick limb bud (31). In summary, we have investigated the aqueous solubility of three of the most common retinoids. Their solubility could be demonstrated provided the necessary precautions are taken to exclude air from the solutions and to use antioxidants, such as vitamins C and E. When going either from ethanol-dissolved samples or dried retinoid stocks, we found the aqueous solubilities to be about 0.1 PM. Although relatively low, such a solubility is physiologically significant. We suggest that a quantitative knowledge of aqueous solubility is necessary for a detailed description of retinoid transport, including the role and the molecular mechanism of action of retinoid-binding proteins. ACKNOWLEDGMENTS We thank Dr. Gregor Jones for helpful discussions and Drs. Rosalie Crouch and Venkat Mani for assistance with HPLC. This research was supported in part by grants from the National Institutes of Health (Biomedical Research Support S07-RR05547, EY04876, EY02399) and National Science Foundation (BNS-8912108).
REFERENCES B., Green, 1. Blomhoff, Science 250, 399-404.
M. H., Berg,
T., and Norum,
K. R. (1990)
C. D. B. (1984) in The Retinoids (Sporn, M. B., Roberts, 2. Bridges, A. B., and Goodman, D. S., Eds.), Vol. 2, pp. 125176, Academic Press, New York. 3. Yoshikami,
S., and Nell,
G. N. (1978)
Scknce
200,
139%1395.
G. (1987) Hiochim. Biophys. Actn 901, 4. Fex, G., and Johannesson, 255-264. 29, 3878-3883. 5. Noy, N., and Xu, Z.-J. (1990) Biochrmistry 6. Ho, M-T. P., Massey, J. B., Pownall, H. J., Anderson, R. E., and Hollyfield,
J. G. (1989)
J. Viol.
Chem.
264,
928-935.
Visual Sci. 7. Harosi, F. I., and Szuts, E. Z. (1990) Inu. Opthalmol. 31(4), 112. 8. Hubbard, R., Brown, P. K., and Bownds, D. (1971) in Methods in Enzymology pp. 615-65.3,
(McCormick, D. B., and Wright, Academic Press, San Diego.
I,. D., Eds.),
Vol.
18,
9. Kofler, M., and Rubin, S. H. (1960) Vitam. Harm. 18, 315-339. J., and Heller, J. (1974) J. Rid. Chem. 249, 47124719. 10. Horwitz, 11. Snyder, I,. R., and Kirkland, J. J. (1979) Introduction Liquid Chromatography, 2nd ed., p. 248, Wiley, New 12. Crouch, R. K., Mani, V., Beischel, Ophthalmol. Vis. Sci. 31(4), 112.
C., and Goletz,
13. Kaegi, H. H. (1984) in The Retinoids (Sporn, and Goodman, D. S., Eds.), Vol. 1, p. 1477179, York.
14. Frolik,
to Modern York. P. (1990)
Inv.
M. B., Roberts, A. B., Academic Press, New
C. A., and Olson, J. A. (1984) in The Retinoids vol. 1 (Sporn, M. B., Roberts, A. B., and Goodman, D. S., Eds.), Vol. I, p. 182 233, Academic Press, New York.
304
SZUTS
AND
15. Lai, Y.-L., Fong, D., Lam, K.-W., Wang, H.-M., and Tsin, A. T. C. (1986)
Curr.
Eye Res. 5, 933-938.
16. Dilley, R. A., and McConnell, D. G. (1970) J. Membr. Biol. 2, 317~ 323. 17. Lyman, W. J., Reehl, W. F., and Rosenblatt, D. H. (1982) Handbook of Chemical Property Estimation Methods: Environmental Behavior of Organic Compounds, pp. l-2, McGraw-Hill Co., New York. 18. McCormick, A. M., Napoli, J. L., and DeLuca, H. F. (1980) in Methods in Enzymology (McCormick, D. B., and Wright, L. M., Eds.), Vol. 67, pp. 220-233, Academic Press, San Diego. 19. Catignani, G. L. (1986) in Methods in Enzymology (Chytil, F., and McCormick, D. B., Eds.), Vol. 123, pp. 215-219, Academic Press, San Diego. 20. Radda, G. K., and Smith, D. S. (1970) FEBS Lett. 9, 287-289. 21. Cogan, U., Kopelman, M., Mokady, S., and Shinitzky, M. (1976) Eur. J. Biochem. 65, 71-78. 22. Renshaw, P. F., Janoff, A. S., and Miller, K. W. (1983) J. Lipid Res. 24, 47-51.
HAROSI 23. Dawidowicz, 202.
E. A. (1987) Cum. Topics
24. De Vendittis,
E., Palumbo, G., Parlato, G., and Bocchini,
Anal.
Riochem.
Membr.
Tramp.
29,
175-
V. (1981)
115, 278-286.
25. Smith, R., and Tanford,
C. (1973) Proc. N&l. Acad. Sci. USA 70,
289-293.
26. Katz, L., and Penman, S. (1966) J. Mol. Biol. 27. Harosi, F. I. (1975) J. Gen. Physiol.
66,
15, 220-231.
357-382.
28. Rushton, W. A. H. (1972) in Handbook of Sensory Physiology (Dartnall, H. J. A., Ed.), Vol. VII/l, pp. 364-394, Springer-Verlag, New York. 29. Berg, H. C. (1983) Random Walks in Biology, p. 29, Princeton
Univ.
Press. 30. Adler, A. J.. and Evans, C. D. (1985) in The Interphotoreceptor Matrix in Health and Disease, (Adler, A. J., and Bridges, C. D. B., Eds.), pp. 65-88, A. R. Liss, New York. 31. Eichele, G. (1989) Trelzds Gen. 5, 246-251.
OF BIOCHEMISTRY
Vol. 257, No. 2, June,
AND
RIOPHYSICS
pp. 297-304.
1991
Solubility of Retinoids in Water Ete Z. Szutsl and Ferenc I. Harosi Laboratory of Sensory Physiology, Marine and Department
Received
of Physiology,
December
Boston
17, 1990, and in revised
Biological Laboratory, Woods Hole, Massachusetts [Jniuersity School of Medicine, Boston, Massachusetts
form
February
6, 1991
Spectrophotometric and radioactive techniques were used to measure the water solubility of retinaldehyde, retinol (vitamin A), and retinoic acid under physiological conditions. Hydration decreases the molar extinction coefficient of these substances and shifts their absorption peak bathochromically (10 nm for retinal and = 1 nm for the rest). We find their solubility to be about 0.1 pM at room temperature, pH 7.3 (with experimental values being 0.11 fiM for retinaldehyde, 0.06 yM for retinol, and acid). To prevent oxidative degra0.21 pM for retinoic dation of retinol, which is the most labile retinoid, our argon-saturated buffer solutions contained physiological levels of ascorbate or a-tocopherol. To the best of our knowledge, water solubility of these compounds has not yet been previously reported. Although the measured solubilities are relatively low, they are significant and may account for the movement of retinoids through the aqueous phase as observed by others during exchange with binding proteins and during intervesicular transfer in the absence of binding proteins. Diffusion of uncomplexed retinoids through the aqueous phase can be a major pathway for transport over subcellular distances. L 1991
Academic
Press,
Inc.
Retinoids, which include vitamin A (retinol) and its chemical derivatives, regulate and control diverse physiological functions. Because these compounds are lipophilic, binding proteins act as carriers for their inter- and intracellular transport. Several types of retinoid-binding proteins are known to exist [for a recent review, see (l)]. Irrespective of their cellular location, all retinoid-binding proteins share the ability to transfer (bind and release) their ligand. The actual molecular mechanism of this transfer reaction has been in dispute. One commonly cited mechanism involves membrane receptors, which act as docking sit,es for the binding protein and which may also ’ 7’0 whom
correspondence
should
be addressed.
02543; 02118
catalyze the loading and unloading of the ligand [e.g., (a)]. An early observation (3) suggested that receptors may be unnecessary for transfer, as retinoids could partition into the aqueous phase. This suggestion was later corroborated by several investigators (4-6), who directly measured retinoid transport among vesicles and also between vesicles and binding proteins. Retinol was found to pass through the aqueous phase, whether retinol-binding protein from the serum (4, 5) or interphotoreceptor retinoid-binding protein (IRBP)’ from the eye (6) mediated the transfer process. Although an aqueous transfer mechanism requires retinoids to be sufficiently water-soluble, solubilities were neither cited nor investigated in these studies. In fact, it appears that water solubility has not yet been measured for any retinoid. The only quantitative statement available in the literature is a citation to unpublished data that solubility of retinol is less than 1 nM (3). In the absence of this information, we set out to measure the solubility” of the physiologically most relevant retinoids. For these studies we specifically chose spectrophotometric techniques as our primary assay, because (i) electronic absorption spectra are indicative of chemical properties and purity, (ii) absorption spectra of retinoids are well characterized, and (iii) spectroscopic measurements allow direct monitoring of a solute within an aqueous environment, without the need for additional handling or processing that may perturb the results. Complicating factors included light sensitivity of the solute and its susceptibility to oxidative damage, as well as ’ Abbreviations used: IRBP. interphotoreceptor retinoid-binding protein; A,,,,, wavelength of peak absorbance; t,,,, molar extinction coefficient at A,,,. ” We use the term “soluhility” in its general meaning. Soluhility of a solid is generally defined as the amount of dissolved solute in equilibrium with its solid state. As the nature of dissolution is not a part of the definition, soluhility of an amphipathic molecule includes all its dissolved forms (monomers, multimers, and micelles). For example, the Merck Index gives the soluhility of sodium dodecyl sulfate as “1 gram dissolves in 10 ml water” which is equivalent to 0.35 M, about two orders of magnitude ahove critical micelle concentration for this detergent.
295 Copyright All
rights
c 1991 by Acxlcmic Press, 01 reproduction in any form
Inc. resen~d.
298
SZUTS
AND
its strong tendency to adsorb to glassware. This report, which is limited to establishing the solubility values, forms the basis of future studies on other biologically relevant parameters. Some of our results were previously presented in abstract form (7). MATERIALS
AND
METHODS
Materials. Retinal (Eastman Kodak Co., Rochester, NY), retinol, and retinoic acid (both from Fluka Chemical Corp., Ronkonkoma, NY) were purchased as all-trans isomers and stored at -20°C under argon. New England Nuclear (Boston, MA) supplied [3H]retinol (47 Ci/mmol, 21 ELM) in ethanol with 2.32 mM d-ol-tocopherol. [I-i”C]Ethanol was prepared by American Radiolabeled Chemicals, Inc. (St. Louis, MO). Spectroscopy grade ethanol and hexane were purchased from Quantum Chemical Corp. (Tuscola, IL) and Fluka Chemical Corp., respectively. Solvents for HPLC were obtained from Fisher Scientific Co. Glassware and cuvettes were thoroughly cleaned between experiments with successive rinses of acetone, chromic acid, and water. Solubilities were tested in a physiological saline solution (also called buffer) that contained 150 mM NaCl, 1 mM Na-phosphate, pH 7.2-7.3, dissolved in HPLC-grade water. To reduce oxidation, all polar solvents were thoroughly purged with argon for 0.5-l h. prior to use and in the case of retinol the buffer also contained ascorbate or o-tocopherol. Purging with argon reduced oxygen tension in the buffer by 95%, as measured with an oxygen electrode (Orbisphere Lab., Geneva, Switzerland). Spectroscopy. Spectral analysis was performed with a photodiodearray uv-vis spectrophotometer (HP 8452A that was operated with HP 89500A Chemstation; Hewlett-Packard Co., Palo Alto, CA). Most samples were measured in lo-cm long cylindrical quartz cuvettes (30 ml capacity). Illustrated spectra represent “raw data,” obtained in single scans without any additional processing. Zero absorbance was assigned to flat portions of the spectra on the longwave side of the main absorption band. Stock solutions were prepared by dissolving the crystalline retinoids in either argon-purged ethanol or hexane and were handled in dim red light. Retinoid solutions were stored for short periods at 4”C, and monitored regularly before use by recording their absorption spectra. They were discarded for any discernable distortion. As a result, retinol stocks were prepared on the day of the experiment,, while ethanolic solutions of retinal and retinoic acid were usually good for several days. In absolute ethanol, all-trans ret,inal absorbed maximally at 382 nm with a molar extinction coefficient of 42.4 (k3.3) X 10” (M-cm)-‘, in excellent agreement with a previous report (8). For all-trans.retinoic ’ at 341 nm in ethanol acid we measured emarto be 49.7 X 10” (M-cm) that agrees well with 45.2 X lo” (9). We determined retinol’s extinction in ethanol, coeflicient at its h,,, of 325 nm to be 38.3 X lo” (M-cm)-’ which is less than its previously published value of 45.1 X 10” (10). In spite of its lower extinction, our retinol eluted as a single peak by HPLC analysis with a purity of >99%, in agreement with the manufacturer’s specifications for the chemical. HPLC. Retinoids were extracted from the physiological solution with hexane (3.0 ml hexane to 28 ml saline) with an efficiency of 92%, as determined with [3H]retinol. About 100 ~1 of the extract was injected for HPLC analysis, performed on a Waters system (Waters Associates Inc., Milford, MA) with an Alltech silica column (Econosphere Silica 5 pm; 250 mm X 4.6 mm; Alltech Associates Inc., Deerfield, IL). The mobile phase was run at 1 ml/min as follows: pure hexane (solvent A) for 5 min after injection, linear gradient of solvent B (ethylacetate/ methanol, 9/l) at I%/min for 20 min, linear gradient of 3% B/min for 10 min, and finally constant 50% B for 10 min. The amount of water in the “saline-saturated” hexane samples [5 mM, (ll)] did not significantly affect retention times (21.4 F 0.2 min for retinal and 26.6 min for retinal). Dilution of ethanolic retinoid solutions with buffer. For these sets of experiments, a series of ethanolic stock solutions were mixed with buffer
HAROSI so that final concentration of total retinoid remained constant for the mixtures (at about 0.3 PM), while the ratio of ethanol to water was varied. The mixtures were analyzed spectroscopically. Each dilution was performed directly in a clean cylindrical quartz cuvette. As long as the final ethanol concentration of the mixture was 50% or greater, the dilution factor completely accounted for final absorbance indicating that 6 for each retinoid remained unchanged from its value in pure ethanol. However, adsorption to the cuvette and changes in e became progressively pronounced when ethanol was <250/o. To control for these factors, we used the following protocol: (i) the absorbance spectrum of the solution to be analyzed was recorded (initial spectrum); (ii) the content of the cuvette was decanted into a Aask and mixed with an equal volume of pure ethanol, so that final ethanol in the flask was 350%; (iii) 28 ml of pure ethanol was added into the cuvette to dissolve all adsorbed retinoid and the absorbance spectrum was measured (spectrum of residue-incuvette); (ivj the mixture in the flask was transferred to a new clean cuvette and its spectrum recorded (spectrum of 250% ethanol mixture). Throughout this procedure, cuvettes and flasks were weighed to accurately measure both transferred and retained volumes. The baseline for all the spectral measurements came from a control cuvette that was processed in parallel with and in the same manner as the test cuvette, but lacked retinoid. From the three recorded spectra, the following variables were unambiguously calculated: mol of retinoid dissolved in the initial solution to be analyzed (calculated from absorbance and volume of 350% ethanol mixture), solubility and tmaxof retinoid in the initial solution (calculated from previously derived “mol of retinoid dissolved” and from volume and absorbance of initial solution), and mol of quartzadsorbed retinoid (from absorbance and volume of residue-in-cuvette). Solubility experiments. In these experiments, a controlled amount of solid retinoid was deposited in excess into the cylindrical cuvette, by pipetting l-10 ~1 of ethanolic stock on the bottom of the cuvette and by evaporating the ethanol with a stream of argon for 5 min. The drying process removed 99.99% of the ethanol, as measured with [‘4C]ethanol. After the argon purge, 28 ml argon-saturated buffer was pipetted into the cuvette. Additional argon was blown through its two ports before the cuvette was stoppered with teflon plugs and placed on a gentle mixer. The absorbance of the aqueous solution (buffer spectrum) was measured 30-60 min after mixing. The same identical analysis was performed with it as described above for the diluted ethanolic stocks, generating an analogous set of three spectra (buffer spectrum, spectrum of >50% ethanol mixture, and spectrum of residue-in-cuvette). Within the spectrophotometer cuvette, retinoid can exist in three “phases:” freely dissolved in bulk phase, adsorbed to quartz surfaces, and aggregated at the air-water interface. The solution decanted from the cuvette is not representative of the cuvette’s bulk solution if the decanting process transfers significant amounts of retinoids from the two kinds of interfaces. The extent of these effects was measured in experiments with [“Hlretinol, by comparing spectral measurements with the radioactivity contained in 1 ml buffer, sampled entirely from the bulk phase. Such analysis indicated that the contribution of the surface efIects was negligible. All experiments were performed at room temperature (21-23°C).
RESULTS
Effect of hydration on spectral absorbance. The effect of water on the spectral properties of retinoids was determined in a series of experiments in which ethanolic retinoid solutions were diluted with buffer. Fig. 1 illustrates the general aspects of hydration as determined for retinaldehyde. As the mole ratio of ethanol to water decreases (representing increased hydration) the first observed effect is a bathochromic shift. This shift (from 382 nm in pure ethanol to 392 nm for ~25% ethanol) is ac-
SOLUBILITY
OF RETINOIDS
,./ 20 t
FIG. 1. Effect of hydration on X,,, and molar extinction coefficient for all-trans-retinaldehyde. A, normalized absorbance spectrum in buffer (solid curve) and in absolute ethanol (dashed curve). Note the bathochromic shift and broadening of the peak upon hydration. B and C, X,,, and fmsxvariation as function of ethanol concentration, plotted as mole ratio, in the mixed solvent system of ethanol/water. For mole ratios of alO-“, measurements were made with mixtures of ethanol/buffer into which ethanolic stock was directly added. The data point at 10 -s mole ratio refers to a buffer measurement for which the ethanolic stock was dried before the buffer was added. For all points, added stock contained a constant amount of retinal (9 nmol) so that the final retinal concentration was 0.325 pM, if completely dissolved. Solutions were mixed for about 30 min before spectral analysis. Data points are the averages of at least three independent determinations. Error bars represent standard deviation, and if absent, are less than the symbol size. Conversion between ethanol/water mole ratio, ethanol molarity, and volume percent are 10’ = 17.1 M, 99.7% v/v; 10 = 16.6 M, 97%; 1 = 13.1 M, 76.4%; 10-i = 4.2 M, 24.5%; lo-’ = 0.54 M, 3.1%; 10m3= 55 mM, 0.32%; 10 ’ = 5.5 mM, 0.03%; 10m6= 55 PM, 3 X 10m4%; lo-* = 0.55 PM, 3 X 10 ‘%.
companied by an increase in the half-width of the main absorption band (Figs. 1A and 1B). Upon additional hydration, the amplitude of the band (molar extinction) diminishes, beginning at about 25% ethanol (Fig. 1C). The observed decrease in E,,,~~,which in buffer drops ultimately
IN WATER
299
to about 43% of the extinction in pure ethanol, is not accompanied by additional shifts in X,,,. We interpret the observed effects to represent titration of two sites or regions on retinaldehyde: one near the polar end, the other along the polyene chain. The wavelength shift occurs at a mole ratio of 1, indicating that water and ethanol molecules can equally displace each other near the polar end. However, invasion by water molecules at the second site along the conjugated bonds is resisted by the hydrophobic interactions between ethanol and the polyene chain. As a result, the solution must contain 100 or more water molecules per ethanol before a significant drop in peak extinction occurs. The solubility of retinaldehyde in a mixed ethanol/water solvent system increases with increasing ethanol fractions, as expected. When the ethanol mole ratio was 20.1, all the retinaldehyde was dissolved, yielding a concentration of 0.33 PM in these experiments. As we reduced the ethanol mole ratio, the solubility of retinaldehyde in the bulk phase diminished to 0.14,0.11, and 0.11 PM at ethanol concentrations of 3%, 0.3%, and 0.03%, respectively. In addition, the dissolved retinaldehyde had a lower extinction coefficient, as shown in Fig. 1C. Remaining retinaldehyde was adsorbed to the walls of the cuvette. Although we did not study the time-course of this precipitation, our spectrophotometric observations were made about 30 min after the introduction of the solute into the ethanol/buffer mixture. The surprising outcome of these experiments is that the solubilizing action of ~3% ethanol solutions is very low, even though the ratio of organic solvent to solute is still high (e.g., mole ratio of ethanol-to-retinal is 1.5 X 10” at 3% ethanol). The data point at a mole ratio of lo--’ in Fig. 1B and Fig. 1C represents buffer solutions equilibrated with argon-dried retinoid sample. The amount of ethanol remaining after drying was measured with [‘4C]ethanol and was found to be 0.01% of the initial ethanol volume, which was mainly bound nonspecifically to the cuvette. Using this value, we calculate the mole ratio of ethanol to water in our buffers to be lOPa, equivalent to the presence of about 0.6 /*M ethanol. This amount of retained organic solvent does not appear to have altered the solubility of retinaldehyde in buffer, because (i) we observed the same solubility even when ethanol mole ratio was decreased another lo-fold (by the use of smaller volumes of more concentrated stocks), and (ii) the ethanol concentration had to exceed 3% before increased solubilizing action of the organic solvent could be demonstrated, as discussed above. Solubility of retinaldehyde. As previously described at greater length, the central feature of our spectral analysis for water solubility is that the solute’s concentration in buffer can be determined from the absorbance of the solution before and after it is diluted with ethanol. We chose
300
SZUTS
AND
Absorbance spectra for solubility determination of all-transFIG. 2. retinaldehyde. Retinal’s spectrum in buffer was recorded first in a locm cuvette (bottom trace). The buffer was decanted from the cuvette and mixed with an equal volume of 100% ethanol and recorded with a separate cuvette. The spectrum of this buffer-ethanol mixture was doubled for display here (middle trace) to clearly demonstrate the relative extinction coefficient in the two media without the dilution effect. The top trace is the spectrum of retinaldehyde remaining in the initial cuvette, with the residue solubilized in 100% ethanol.
equal volume dilutions for all retinoid experiments, because such a dilution yields a 50% ethanol solution for which retinoid adsorption to the cuvette is absent and for which emaxis known to be the same as for pure ethanol. Representative experimental spectra for calculating retinaldehyde’s relative extinction coefficient and solubility are presented in Fig. 2. The average solubility we measured for all-trans-retinaldehyde in our buffer was 0.11 PM (Table I). The purity of retinal dissolved in buffer was verified by HPLC techniques. Buffer solutions were extracted with hexane and the absorption spectrum of the extract was measured before it was analyzed by HPLC (Fig. 3). Based on chromatograms obtained at three wavelengths (370,
TABLE
HAROSI
320, 280 nm), the sample’s content was at least 95% alltruns-retinal. Assuming that the efficiency of extraction is the same for retinoids and oxidation products, the result with HPLC verifies that all-trarzs-retinal was the predominant compound present in buffer. Thus, the existence of retinal in an aqueous environment is confirmed by two independent analytical techniques. Solubility of retinol. As reported by others (12), retinol is unstable in aqueous solutions in the absence of antioxidants. To prevent its degradation, we supplemented our regular argon-purging of solutions with the addition of either ascorbate to the buffer (1 mM) or a-tocopherol to the ethanolic stock solutions (mole ratio of retinol to tocopherol being 7-9). Not only have these antioxidants been used previously with retinoids (13,14), they are also physiological. For example, the ascorbate concentration is 0.5-l mM in the retina and pigment epithelium (15) and the concentration of vitamin E is -0.3 mM in rod outer segments (16). Being a highly polar molecule, vitamin C is not expected to affect the solubility of retinol. On the other hand, its main absorption band [h,,, = 266 is very near to that of nm, hx = 12 X 10” (M-cm)m1] retinol so it interferes with the spectroscopic determination. In order to avoid this problem, we extracted retinol from buffer with hexane (as performed for HPLC analysis), thereby effectively separating vitamin C from it. Knowing the extraction efficiency, retinol’s concentration in buffer was calculated from its spectrum in the hexane phase (Fig. 4). This way (method #l in Table I) we obtained a solubility of 0.05 PM for retinol in our buffer solutions. Due to the indirect nature of this determination, method #l probably yields the least accurate result. Spectral interference was greatly reduced with the second antioxidant. Although cr-tocopherol also absorbs in the same spectral region as retinol, its contribution is
I
Solubility and Extinction Coefficient of All-trans.retinoids Relative extinction coefficient (buffer relative to that in ethanol) (%)
Retinoid Retinaldehyde Retinol (vitamin Method 1b Method 2 Method 3 Retinoic acid
in Water
Extinction coefficient (t,,,) in buffer” (M-cm) ’ X lo”
8 (n = 7)
17.9 * 3.5 (n = 7)
46 k 19 (n = 5)
17.8 f 7.3 (n = 5)
75 k
37.6 f 1.0 (n = 3)
42 +
Concentration
in buffer
(FM) 0.108 + 0.016
(n = 7)
0.051 0.079 0.056 0.212
(n (n (n (n
A)
2 (n = 3)
f f f *
0.017 0.026 0.010 0.022
= = = =
8) 5) 5) 3)
’ The absolute value of the extinction coefficient in buffer was calculated from the experimentally determined value for relative extinction 42.4 X 10” for retinal, 38.3 X 10” for retinol, and 49.7 X lOa for retinoic acid (all coefficient (column 1) and from the following ernar in ethanol: being our determinations). * Method 1: buffer contained 1 mM ascorbate and the spectrophotometric analysis is based on hexane extract; method 2: buffer contained etocopherol (mole ratio of retinol/tocopherol = 7-9) and analysis performed spectrophotometrically; method 3: same as method 2 except data from radioactive assays.
SOLUBILITY
OF
RETINOIDS
IN
WATER
301
FIG. 4.
Absorbance spectra of all-trans.retinol in aqueous buffer (solid (dashed curves; X,,, = 325 nm). curve; X,,, = 326 nm) and in n-hexane Top trace, absorbance of the hexane phase of a control experiment in which the dried retinol sample was dissolved in hexane before a similar volume of buffer was added. Middle trace, absorbance of the hexane phase after extraction from buffer. Bottom trace, retinol in aqueous buffer. In this experiment, the retinol stock contained n-tocopherol (the mole ratio of retinol to tocopherol was 9).
FIG. 3. Retinaldehyde extraction from buffer into hexane. A, retinal’s absorbance spectrum in buffer prior to extraction, as measured in 10. cm long cuvette (solid curve). Retinal’s absorbance in hexane following extraction, as measured in l-cm long cuvette (dashed curve). B, HPLC chromatogram of same hexane extract. The major peak is all-transretinaldehyde, as determined by standards. Minor peaks marked with * are solvent impurities. Top trace indicates gradient profile and is a plot of percentage B. Time delay between gradient controller and uv detector is 8.7 min. Column was Econosphere silica, 5 +m; 250 mm X 4.6 mm; sample was 80 ~1 hexane extract; flow rate, 1 ml/min; mobile phase A, hexane; mobile phase B, ethylacetate/methanol (9/l); pressure, 400 psi: detector X, 370 nm.
small because (i) it has a lower t,,, [2 X 10” (M-cm).. ’ at x max = 286 nm] and (ii) lower concentrations of it are needed for effective protection of retinol. Since vitamin E is already present in commercially supplied solutions of radioactive retinol, we could use the same samples for bothspectrophotometricandradioactiveanalyses.Retinol solubility by spectrophotometric analysis was 0.08 PM in the presence of vitamin E (Fig. 4 and method #2 in Table I). The corresponding result with radioactivity was found to be 0.06 yM (method #3 in Table I). When the experimental results of the three methods are averaged, the solubility of retinol is 0.06 PM. Solubility of retinoic acid. Compared with retinal, retinoic acid’s behavior to hydration is anomalous. We find the peak absorbance of the acid in absolute ethanol and pure buffer to be essentially the same at 341 nm (Fig. 5). This value, however, does not remain stationary in a mixed solvent system of ethanol and water: in 50% and 98% ethanol, we find a hypsochromic shift of 3 and 4 nm, respectively. The molar extinction coefficient in the
aqueous phase decreases by only -25% from that in ethanol. Solubility of retinoic acid was determined spectrophotometrically to be 0.21 PM (Table I). DISCUSSION
Solubility values. Our results indicate that the three unesterified retinoids have an aqueous solubility concentration of about 0.1 PM, plus or minus a factor of two. As we kept the time interval for dissolution relatively short (30-60 min), the experimental values could be considered an underestimate of the actual solubilities. However, additional considerations suggest that dissolved solutes were essentially in equilibrium with the solid phase in our solutions. In the binary solvent systems that we prepared by mixing buffer with ethanol solutions, retinoid solubility was limited at low ethanol concentrations, even though
FIG. 5.
Absorbance spectra for solubility determination of all-transretinoic acid. Middle trace, retinoic acid’s spectrum in buffer. Top trace, twice the absorbance of the 50/50 ethanol/buffer mixture. Bottom trace, absorbance of the ethanol-dissolved cuvette residue.
302
SZUTS
AND
retinoid was fully solvated by ethanol at the time of mixing. This was illustrated for retinaldehyde, for which we found final concentrations to be 0.11-0.14 PM in 0.03-3% ethanol (Fig. 1). In other experiments, in which retinaldehyde was dried before the introduction of the aqueous buffer, final concentration reached 0.11 pM (Fig. 2 and Table I). Thus, solubility approached the same value whether the procedure started with undissolved solute (solid) or with solute dissolved in ethanol. Therefore, our experimental measurements on retinoid solubility were probably near equilibrium. The elution profiles observed in reverse-phase chromatography are indicative of relative solubilities [for review, see (17)]. This is because partitioning between aqueous mobile and organic stationary phases is the primary mechanism of separation by reverse-phase HPLC, leading to chromatograms where polar substances elute first from the column. For retinoids, the observed order of elution is retinoic acid, followed by retinol and then by retinaldehyde (14,18). Since greater polarity increases the water solubility of a substance, the order of elution suggests that retinoic acid is the most, and retinaldehyde the least, soluble. The fact that we found solubility of retinol to be the lowest of the three retinoids may be due to the use of antioxidants. In the absence of antioxidants, retinoids and especially vitamin A rapidly decompose in aqueous solutions. We confirm the observation of others (13, 14) that the presence of ascorbate and a-tocopherol retards retinoid degradation. However, the use of ascorbate interferes with the spectroscopy of aqueous solutions of retinol and made solubility determinations indirect and possibly less accurate. The use of a-tocopherol may have also interfered in these measurements, due to its hydrophobic nature. Its water solubility, which remains unknown, is probably less than that of retinol on the basis of its longer retention time by reverse-phase HPLC (19). When ethanolic stock solutions containing both retinol and tocopherol were dried in our experiments, the two vitamins could have become associated with each other. Such an association may have retarded the availability of retinol for dissolution and may have been the cause for an underestimated solubility value. Additional studies are needed to fully determine the effects of tocopherol on retinoid solubility. Retinoid aggregation. Being amphipathic molecules, retinoids may self-aggregate to form complexes that in principle can range from dimers to micelles. Because our solubility determinations were made under conditions where dissolved solute was in equilibrium with its solid state, our experimentally derived solubilities are correct even if micelles formed in our buffers. However, micelle formation is unlikely for the very reason that retinoid solubility is limited. Because the number and size of micelles can grow with increased amounts of added solute,
HAROSI
the solubility of micelle-forming compounds should exceed 0.1 yM. The notion that retinoids form micelles was suggested by two reports (20,21) that studied the polarization of retinol’s fluorescence in buffers that contained x 1 PM retinol and N 1% ethanol. In these studies, the endogenous fluorescence of retinol was found to be more polarized in the aqueous mixture than in pure ethanol, indicating the loss of rotational freedom in the mixture. This result was interpreted by both groups as evidence for micelle formation. However, there are two other possible explanations: adsorption of retinol to cuvette walls (found by us to occur whenever ethanol < 25%) and precipitation of retinol (as it goes out of solution) due to limited solubility. Similar precipitation is known to occur for cholesterol, which forms a suspension of “microcrystals” above its solubility limit of ~10 nM in water (22, 23). With a fluorimetric probe technique used to measure critical micelle concentration (24), we found no evidence for micelle formation by either retinol or retinaldehyde (Harosi and Szuts, unpublished). Although micelle formation is unlikely, the presence of smaller aggregates (such as dimers or trimers) is certainly possible. Of the three retinoids, dimerization is expected for retinoic acid, because hydrogen-bonding between -COO- and -COOH groups is known to be especially strong for long chain fatty acids. Smith and Tanford (25) have shown that tenacious hydrogen bonds are responsible for the formation of dimers and higher aggregates by carboxylic acids at neutral pH, even below critical micelle concentration. However, the extent of dimerizationlaggregation, which they could not determine, remains unknown. The mechanism of multimer formation by retinoids, similarly remains to be investigated.4 Physiological relevance. Is a water solubility of 0.1 PM for retinoids physiologically relevant? The cellular requirement of any retinoid is best characterized for the photoreceptors of the vertebrate retina. Photoreceptors contain about 3.5 mM rhodopsin (27), the visual pigment, whose chromophore is II-cis-retinaldehyde. Photoexcitation leads to hydrolysis of rhodopsin, releasing free alltrans-retinal which is rapidly converted to all-trans-retinol within the photoreceptors. Resynthesis of 11-cis-retinal from all-trans-retinol involves the cells of the pigment epithelium, which abut the receptor cells [ref. (2) for review]. Thus, there is a constant traffic of retinoids between ’ We determined the absorbance spectra of the retinoids in various solvents (not illustrated). In dimethyl sulfoxide, which is a strong hydrogen bond acceptor (26), we found a large shift in X,,, for retinoic acid, but smaller shifts for retinol and retinaldehyde. The corresponding X,,, for retinoic acid, retinol, and retinaldehyde were respectively 341, 325, and 382 nm in ethanol and 361,332, and 386 nm in dimethyl sulfoxide. These results are consistent with the notion that hydrogen-bonding among solute molecules may occur in aqueous solutions of retinoic acid, whereas the same is less likely for retinol and even less for retinaldehyde.
SOLUBILITY
OF
RETINOIDS
these two cell layers of the retina across the interphotoreceptor space. At most levels of illumination a steady state is achieved where the rate of photoactivation equals the rate of rhodopsin regeneration, supported by retinoid transport. For example, when ambient light intensity bleaches 10% of the cone pigments in the human eye, the regeneration rate is about 4% per minute (28). Similarly, regeneration is about 1% per minute in a live frog after 90% of its rhodopsin has been bleached (3). Surprisingly, it is possible to satisfy such rates of delivery by free diffusion alone from our aqueous solubility values. This conclusion is based on the application of diffusion theory (29) to a model system in which a prolate target is bombarded by solute molecules, whose concentration is constant in the surrounding medium. In this analysis, we take the cylindrically shaped outer segment of rod cells to be the target (with I= 60 pm, r = 3 pm) and we assume that (i) 11-&-retinal diffuses as a monomer with D = 0.5 X lo-” cm”/s and (ii) all collisions with the target lead to transfer to the plasma membrane. With these conditions, a solubility of 0.1 PM yields a diffusion rate of 2 X 10R molecules/min to a single rod outer segment. This rate represents the arrival of enough retinoid to regenerate -6% of the rhodopsin in the cell every minute. Hence, the solubility values we determined are sufficiently high to meet the metabolic requirements of photoreceptors without invoking any binding proteins. These results are based on the assumption that the aqueous medium surrounding the cells is saturated with retinoids. The actual free concentration of these compounds remains unknown. Photoreceptors were used for this mathematical model, because they are known to metabolize retinoids at a high rate. However, the analysis and conclusion would be valid for other cells, especially those with reduced metabolic requirements. Our results may explain several unresolved observations in retinal physiology: (a) the absence of a binding protein within rod photoreceptors, where >90% of the chromophore must cross the aqueous cytoplasm separating plasma and disk membranes, (b) the absence of confirmation for the existence of predicted receptor sites for IRBP on the extracellular surfaces of both photoreceptors and epithelial cells, and (c) the nonspecificity of IRBP for the aldehyde and alcohol forms of vitamin A and t,heir various isomers (2). Although binding proteins are not an absolute requirement for retinoid transport over short distances (comparable to the size of cells), IRBP may yet serve as an adjunct of transport in the retina. For example, it can support transport rates in excess of the value permitted by water solubility alone. The relatively high concentration of IRBP in the extracellular space (30), which is about two orders of magnitude greater than the aqueous solubility of retinoids, increases the saturability of the medium surrounding the receptors, thereby increasing transfer
IN
303
WATER
rates between source and target. Additionally, IRBP appears to prevent the loss of retinoids by diffusion through the aqueous phase of the retina. Therefore, IRBP and other cellular binding proteins can serve to localize and concentrate retinoids to regions of special need. Thus, we agree with others who have expressed similar ideas for the visual cycle in the vertebrate eye (6) and for the role of retinoic acid in the developing chick limb bud (31). In summary, we have investigated the aqueous solubility of three of the most common retinoids. Their solubility could be demonstrated provided the necessary precautions are taken to exclude air from the solutions and to use antioxidants, such as vitamins C and E. When going either from ethanol-dissolved samples or dried retinoid stocks, we found the aqueous solubilities to be about 0.1 PM. Although relatively low, such a solubility is physiologically significant. We suggest that a quantitative knowledge of aqueous solubility is necessary for a detailed description of retinoid transport, including the role and the molecular mechanism of action of retinoid-binding proteins. ACKNOWLEDGMENTS We thank Dr. Gregor Jones for helpful discussions and Drs. Rosalie Crouch and Venkat Mani for assistance with HPLC. This research was supported in part by grants from the National Institutes of Health (Biomedical Research Support S07-RR05547, EY04876, EY02399) and National Science Foundation (BNS-8912108).
REFERENCES B., Green, 1. Blomhoff, Science 250, 399-404.
M. H., Berg,
T., and Norum,
K. R. (1990)
C. D. B. (1984) in The Retinoids (Sporn, M. B., Roberts, 2. Bridges, A. B., and Goodman, D. S., Eds.), Vol. 2, pp. 125176, Academic Press, New York. 3. Yoshikami,
S., and Nell,
G. N. (1978)
Scknce
200,
139%1395.
G. (1987) Hiochim. Biophys. Actn 901, 4. Fex, G., and Johannesson, 255-264. 29, 3878-3883. 5. Noy, N., and Xu, Z.-J. (1990) Biochrmistry 6. Ho, M-T. P., Massey, J. B., Pownall, H. J., Anderson, R. E., and Hollyfield,
J. G. (1989)
J. Viol.
Chem.
264,
928-935.
Visual Sci. 7. Harosi, F. I., and Szuts, E. Z. (1990) Inu. Opthalmol. 31(4), 112. 8. Hubbard, R., Brown, P. K., and Bownds, D. (1971) in Methods in Enzymology pp. 615-65.3,
(McCormick, D. B., and Wright, Academic Press, San Diego.
I,. D., Eds.),
Vol.
18,
9. Kofler, M., and Rubin, S. H. (1960) Vitam. Harm. 18, 315-339. J., and Heller, J. (1974) J. Rid. Chem. 249, 47124719. 10. Horwitz, 11. Snyder, I,. R., and Kirkland, J. J. (1979) Introduction Liquid Chromatography, 2nd ed., p. 248, Wiley, New 12. Crouch, R. K., Mani, V., Beischel, Ophthalmol. Vis. Sci. 31(4), 112.
C., and Goletz,
13. Kaegi, H. H. (1984) in The Retinoids (Sporn, and Goodman, D. S., Eds.), Vol. 1, p. 1477179, York.
14. Frolik,
to Modern York. P. (1990)
Inv.
M. B., Roberts, A. B., Academic Press, New
C. A., and Olson, J. A. (1984) in The Retinoids vol. 1 (Sporn, M. B., Roberts, A. B., and Goodman, D. S., Eds.), Vol. I, p. 182 233, Academic Press, New York.
304
SZUTS
AND
15. Lai, Y.-L., Fong, D., Lam, K.-W., Wang, H.-M., and Tsin, A. T. C. (1986)
Curr.
Eye Res. 5, 933-938.
16. Dilley, R. A., and McConnell, D. G. (1970) J. Membr. Biol. 2, 317~ 323. 17. Lyman, W. J., Reehl, W. F., and Rosenblatt, D. H. (1982) Handbook of Chemical Property Estimation Methods: Environmental Behavior of Organic Compounds, pp. l-2, McGraw-Hill Co., New York. 18. McCormick, A. M., Napoli, J. L., and DeLuca, H. F. (1980) in Methods in Enzymology (McCormick, D. B., and Wright, L. M., Eds.), Vol. 67, pp. 220-233, Academic Press, San Diego. 19. Catignani, G. L. (1986) in Methods in Enzymology (Chytil, F., and McCormick, D. B., Eds.), Vol. 123, pp. 215-219, Academic Press, San Diego. 20. Radda, G. K., and Smith, D. S. (1970) FEBS Lett. 9, 287-289. 21. Cogan, U., Kopelman, M., Mokady, S., and Shinitzky, M. (1976) Eur. J. Biochem. 65, 71-78. 22. Renshaw, P. F., Janoff, A. S., and Miller, K. W. (1983) J. Lipid Res. 24, 47-51.
HAROSI 23. Dawidowicz, 202.
E. A. (1987) Cum. Topics
24. De Vendittis,
E., Palumbo, G., Parlato, G., and Bocchini,
Anal.
Riochem.
Membr.
Tramp.
29,
175-
V. (1981)
115, 278-286.
25. Smith, R., and Tanford,
C. (1973) Proc. N&l. Acad. Sci. USA 70,
289-293.
26. Katz, L., and Penman, S. (1966) J. Mol. Biol. 27. Harosi, F. I. (1975) J. Gen. Physiol.
66,
15, 220-231.
357-382.
28. Rushton, W. A. H. (1972) in Handbook of Sensory Physiology (Dartnall, H. J. A., Ed.), Vol. VII/l, pp. 364-394, Springer-Verlag, New York. 29. Berg, H. C. (1983) Random Walks in Biology, p. 29, Princeton
Univ.
Press. 30. Adler, A. J.. and Evans, C. D. (1985) in The Interphotoreceptor Matrix in Health and Disease, (Adler, A. J., and Bridges, C. D. B., Eds.), pp. 65-88, A. R. Liss, New York. 31. Eichele, G. (1989) Trelzds Gen. 5, 246-251.