Isolation and Lipid Characterization of Cholesterol-Enriched Fractions in Cortical and Nuclear Human Lens Fibers

Abstract

purpose. Human lens membranes contain unusually high levels of cholesterol and sphingolipids, lipids known to segregate into liquid-ordered domains. The current study was conducted to pursue the determination and characterization of these domains in membranes of clear and cataractous human lenses.

methods. Cortical and nuclear regions of aged clear and cataractous lenses were obtained. After lysis with Triton X-100 at 4°C and sucrose linear-density centrifugation, sedimenting and nonsedimenting fractions (when present) were collected. Phospholipids were analyzed by ³¹P-nuclear magnetic resonance (NMR) and mass spectrometry. Caveolae and raft markers were tested by Western blot analysis.

results. Only samples from clear lenses exhibited a nonsedimenting band. Phospholipid contents were comparable for sedimenting fractions of clear and cataractous membranes. Cholesterol to phospholipid molar ratios in light-density bands were nearly 7, three times greater than in sedimenting fractions. The portion of total cholesterol present in nonsedimenting fractions increased from 5.5% in the cortex to 14% in the nucleus. Two lysophospholipids comprising approximately 10% of all phospholipids in total membranes were undetectable in nonsedimenting fractions. Caveolin-1 was enriched in these fractions.

conclusions. Phospholipid compositional differences between lighter and heavier fractions from clear lenses were relatively minor and could not, alone, account for the substantial enrichment of cholesterol in the lighter fractions. Specific proteins, such as caveolin-1, must recruit cholesterol and induce clustering. Undetectable amounts of light-density domains in cataractous membranes suggest either disruption of these aggregates and thus the function of proteins within them, possibly relevant to lens transparency, and/or greater density of these clusters due to stronger binding of insoluble crystallins to membranes.

Plasma membranes of human lenses exhibit not only an unusually high content of cholesterol (CHol) (∼50%–60% of all lipids), ^{1 2 3 4 5 6 7} but also exceptional features in their phospholipid (PL) composition. Unlike in any other membrane investigated to date, dihydrosphingomyelins (DHSMs) comprise nearly half of all PLs in human lens fibers. ^{8 9 10 11} Compared to sphingomyelins (SMs), the most abundant sphingolipid in other mammalian membranes, DHSMs do not have a trans double bond between the fourth and fifth carbons of the sphingoid base. SMs, although present in human lenses, make up a smaller portion of all PLs, and their content is age-dependent, increasing from 7% in younger lenses (0–15 years old) to 14% in older ones (≥76 years old). ¹²  

The preferential segregation of sphingolipids and cholesterol (CHol) into liquid-ordered domains or “rafts” in biomembranes ¹³ has been the theme of exciting research in the past decade. Several reviews and reports have addressed not only the relevant biological functions of these clusters, but also their physical organization and the nature of the forces that lead to their formation. ^{14 15 16 17 18 19 20 21}  

Rafts have been isolated in a variety of biological membranes by using protocols that rely on the insolubility of these clusters in nonionic detergents (e.g., Triton X-100) at low temperature. ^{22 23} Their presence in natural membranes has been demonstrated conclusively by in situ studies. ^{24 25 26 27 28} In nonhuman mammalian lenses, the detection of caveolin, a protein marker for caveolae, has been reported (Lo WK, Wen XJ, Mills A, ARVO Abstract 3260, 1997). It is believed that rafts and caveolae participate in and/or control multiple functions, including cell signaling and potocytosis and may serve as docking sites for certain pathogens (bacteria, parasites, viruses) and toxins. ^{29 30 31 32 33}  

Given the unusually high contents of CHol and sphingolipids in human lens membranes, the clustering of these lipids into domains of different lipid and/or protein content is highly plausible, as proposed by others. ^{5 34} In vitro x-ray diffraction studies using membrane fractions extracted from human lenses suggested the presence of CHol monohydrate bilayers (3.4 nm thick) within PL plasma membranes. ³⁵  

Because proteins may play a major role in the sequestering of CHol and sphingolipids, ³⁶ our first objective was to determine the possible presence of CHol-enriched domains in cortical and nuclear membranes (with their constitutive proteins) of aged clear and cataractous lenses. Second, we assessed the lipid composition of the various fractions to establish whether differences in lipid contents alone could explain domain formation. We have applied the practical approach—differences in solubility in nonionic detergents at low temperatures—for the isolation of the different fractions. The characterization of lipids was pursued by using nuclear magnetic resonance (NMR) and mass spectrometry (MS), because both approaches offer complementary information and do not require separation and/or derivatization steps. With phosphorous-NMR (³¹P-NMR), PLs are identified according to their headgroup composition. Yet, virtually no details are obtained about their acyl-chain distribution. MS offers an excellent alternative to gather this relevant information, especially through the use of soft-ionization techniques such as matrix-assisted laser desorption ionization (MALDI). ³⁷ Today, MALDI time-of-flight mass spectrometry (MALDI-TOF MS) stands as one of the fastest and most sensitive methods for lipid analysis. It enables not only the assessment of acyl-chain length and the number of sites of unsaturation, but also the quantification of CHol, a key component in the clustering process. In addition, we applied MALDI-TOF MS to obtain the mass spectral profiles of the proteins in the different fractions. 

Methods

Materials

Chemicals used for the preparation of the main buffer (25 mM Tris-HCl [pH 7.6], 150 mM NaCl, and 5 mM EDTA), denoted TNE, lysis buffer (1% Triton X-100 in TNE), phosphate-buffered saline (PBS; pH 7.6), sucrose solutions (5%, 30%, and 40% in TNE), and octyl β-d-glucopyranoside (n-octyl glucoside), as well as the matrix compounds, 2,5-dihydroxybenzoic acid (2,5-DHB) and 3,5-dimethoxy-4-hydroxycinnamic acid (sinapinic acid), and the solvents (trifluoroacetic acid [TFA], methanol, chloroform, acetonitrile) were all purchased from Sigma (St. Louis, MO). N-hexanoyl-sphingosylphosphorylcholine and 1,2-dimyristoyl-sn-glycero-3-phosphatidic acid (DMPA) were used as internal standards and were obtained from Matreya, Inc. (Pleasant Gap, PA). Deuterated cholesterol-2,2,3,4,4,6-d₆ (d₆-CHol) was received from Cambridge Isotope Laboratories, Inc. (Andover, MA). Mouse monoclonal primary antibodies against human caveolin-1, caveolin-2, and glycosylphosphatidylinositol (GPI)-phospholipase D, as well as goat anti-mouse horseradish peroxidase secondary antibody and rat cerebrum lysate were all from BD Transduction Laboratories (Lexington, KY). All reagents were used without further purification. Human lenses were obtained from the University of Louisville Lions Eye Bank (Louisville, KY) and the Lions Eye Bank of Lexington (Lexington, KY). The provisions of the Declaration of Helsinki for research involving human tissues were observed. 

Total Lipid Extraction

Two clear human lenses (70 years old) and three nuclear sclerotic cataractous lenses (66–72 years old) with extended nuclear and cortical opacification were dissected into cortical and nuclear regions. The separation of cortex and nucleus was based on the differences in consistency of the tissue. The harder inner region, denoted nucleus herein, encompasses embryonic and fetal fibers. Lipids were extracted from the obtained tissue materials using a volume of methanol-chloroform (2:1 vol/vol) 30 times greater than the tissue. After sonication for 3 minutes, samples were centrifuged for half an hour. The lipid-containing supernatant was named total membrane fraction. To minimize lipid loss, no further purification steps were performed. 

Extraction of Detergent-Insoluble Fractions

Cortical and nuclear regions were excised from three clear human lenses (70–72 years old) and three nuclear sclerotic cataractous lenses with advanced opacification (72–87 years old). The protocol used for the isolation of detergent-insoluble fractions (DIFs) and further separation by flotation on a sucrose linear density gradient was adapted from a previously described approach ¹⁶ and is illustrated in Figure 1 . Briefly, cortical and nuclear lens regions were homogenized in PBS (pH 7.6) at room temperature. After a 1-hour centrifugation at 5000 rpm and 4°C (50.3 Ti rotor; Beckman Instruments, Inc., Fullerton, CA), a water-insoluble fraction (WIF) was recovered. The lysis buffer was added to the WIF and mixed thoroughly, and after 15 minutes at 4°C, the sample was centrifuged for 30 minutes at 5000 rpm and 4°C (50.3 Ti rotor). The obtained detergent-insoluble fraction (DIF) was brought to 40% sucrose, and a linear sucrose gradient (30%–5%) was laid over it. Samples were then centrifuged for 24 hours at 25,000 rpm at 4°C (SW27 rotor; Beckman Instruments, Inc.). Only for clear lens samples (cortical and nuclear), a single light-scattering band (0.5–1.0 mm thick) was observed in the top one fifth of each tube and denoted nonsedimenting DIF (NSDIF). The NSDIF band was harvested. The pelleted material was also collected and labeled sedimenting DIF (SDIF). 

The procedure for total lipid extraction, described earlier, was followed to extract lipids from each of the DIFs. 

Lipid Assignment and Quantification

Lipids extracted from the total membranes of the cortical and nuclear lens regions, as well as the cortical and nuclear DIFs (NSDIFs and SDIFs), were assayed by both ³¹P-NMR and MALDI-TOF MS. 

³¹P-NMR Lipid Analysis.

The quantification of PLs in lens tissue by ³¹P-NMR spectrometry was demonstrated in the 1980s. ^{38 39} In the current study, we used a similar approach to analyze PLs extracted in a methanol-chloroform mixture. ³¹P-NMR data were acquired on a NMR spectrometer (Inova-500; Varian, Sunnyvale, CA). The following parameters were used: spectral width of 2024.7 Hz (sweep width δ = 10 parts per million [ppm]), 60° pulse, 4 K data points, 1.000-second delay time, and 0.711-second acquisition time at 25°C. Proton-decoupling (500.16 MHz) was used. Spectra were processed with a line broadening of 3.0 Hz and phase correction. Further spectral evaluation was accomplished using GRAMS software (version 386; Galactic Industries Corp., Salem, NH). A 200-μL aliquot of cesium EDTA (Cs⁺-EDTA) reagent ³⁹ was added to 400 μL of each sample before data acquisition. This detergent (Cs⁺-EDTA) yielded narrower ³¹P-NM resonances. After heating at 40°C to 50°C for 15 minutes, samples were allowed to cool down to room temperature before NMR spectral acquisition. To minimize lipid loss, the chloroform-methanol extracts were not subjected to the conventional washing step with aqueous KCl. It is thus likely that remaining proteinaceous components may be responsible for the greater line widths observed in the ³¹P-NMR bands of these extracts compared with those seen in more pure lipid samples. Nevertheless, the compositional data are in agreement with those published previously for human lens membranes. ^{11 12} For the quantification of the PLs, a known amount of DMPA was added as an internal standard. Chemical shifts were referenced to internal phosphatidylcholines (PCs, δ = −0.84 ppm). 

MALDI-TOF MS Lipid Analysis.

Mass spectra were acquired with a Voyager Biospectrometry DE instrument (PerSeptive Biosystems, Framingham, MA), that uses a pulsed nitrogen laser at 337 nm to induce ionization. The extraction voltage was 20 kV. The laser power level was adjusted to obtain high signal-to-noise ratios, while ensuring minimal fragmentation of the parent ions. Samples were directly applied onto the stainless-steel spectrometer plate as 1-μL droplets, followed by the addition of 1 μL of DHB-matrix solution (0.5 M of 2,5-DHB in methanol containing 0.1% TFA). After crystallization at ambient conditions, positive ion spectra were acquired in the reflector mode. At least 10 independent droplets were analyzed for each sample. One mass spectrum represents the average of more than 100 traces. 

For the quantification of PLs and CHol, known amounts of SM with a hexanoyl acyl chain (six carbons long and no unsaturation site), SM(6:0), and d₆-CHol were added to serve as internal standards. To improve mass accuracy, a two-point calibration was performed. The peak related to d₆-CHol ion with a mass-to-charge ratio (m/z) of 375.39 served as the lower mass limit. Because of its high m/z of 815.68, the peak corresponding to the DHSM with nervonoyl chain (24 carbons long and one unsaturation site), DHSM(24:1), was used as the upper mass value. After mass calibration, software developed in our laboratory was used to identify and assign the lipid peaks. The program was designed to screen for the presence of PL-related peaks by comparison of the experimental and theoretical values of m/z corresponding to the monoisotopic peaks of the protonated, sodiated, and potassiated lipid ions. To ensure correct identification of the targeted lipids, a tolerance of Δ(m/z) = 0.06 was set up. This maximum difference between experimental and theoretical values allowed the effective discrimination of lipid-related peaks from those due to the background. The areas of the monoisotopic peaks corresponding to a given lipid were added and referenced to the total area of the internal standard peaks. 

Protein Analysis

To detect possible differences in the protein contents of low- and high-density fractions, we obtained mass spectral profiles with the use of MALDI-TOF MS and tested the presence of several marker proteins reported to be enriched in rafts and caveolae. 

Mass Spectral Protein Analysis.

Ten milligrams of SDIF or 1 mg of NSDIF was dissolved in 300 μL (SDIF) or 30 μL (NSDIF) 1% n-octyl-glucoside solution prepared in 5 mM Tris-HCl (pH 7.6) and 0.1 M mercaptoethanol. A portion of the dissolved material was mixed with an identical volume of the matrix (30 mg of sinapinic acid in a mixture of 250 μL of 6% TFA in distilled water and 250 μL of acetonitrile). A 1-μL droplet of the sample-matrix mixture was spotted onto the stainless-steel MALDI plate, and mass spectra were collected with the same spectrometer used for lipid analysis. 

SDS-PAGE and Western Blot Analysis.

SDIFs and NSDIFs isolated from clear, aged human lens membranes were tested for the presence of human-specific caveolins-1 and -2, known markers for caveolae. In addition, the presence of GPI-phospholipase D was also checked in SDIFs and NSDIFs. Proteins with GPI anchors have been reported to be enriched in rafts. ^{16 26 28} A rat cerebrum lysate was used as positive control for caveolin-2 as well as negative control for caveolin-1 and GPI-phospholipase D. 

The materials obtained from SDIFs and NSDIFs were subjected to SDS-PAGE using 5% to 15% gradient, 1-mm-thick gels. Duplicate runs were obtained for each sample. The samples were boiled 5 minutes before loading, and the final concentration of dithiothreitol in the electrophoresis buffer was 200 mM. Protein (100 μg) was loaded in each well. The gels were run for approximately 1 hour at constant current and then transferred to nylon membranes (Immobilon-P; Millipore, Bedford, MA) for 2 hours at 200 mA. After transfer, the membranes were incubated for 20 minutes with 5% milk to block nonspecific binding sites. The membranes were clamped with a Western blot analysis manifold that isolates 40 channels across the membrane. Antibodies against caveolin-1, caveolin-2, and GPI-phospholipase D were each added in different channels and allowed to hybridize for 45 minutes. The blots were removed from the manifold, washed, and hybridized for 30 minutes with goat anti-mouse horseradish peroxidase as the secondary antibody. The membranes were washed, and antibody binding was detected with the extended-duration chemiluminescence substrate (Pierce, Rockford, IL). Digital detection was performed (PDQuest; Bio-Rad, Hercules, CA) and ratios of the normalized data were used to quantify the possible enrichment of the chosen markers in the NSDIFs. 

Results

Unlike membranes isolated from cataractous tissues, those from cortical and nuclear regions of clear lenses yielded a band of nonsedimenting material after sucrose linear density gradient centrifugation. Because the isolation of DIFs was performed simultaneously for cataractous and clear lenses, the observed lack of low-density fractions in cataractous membranes cannot be attributed to changes in experimental variables. For clear lens membranes, the NSDIF band was not observed when the content of CHol was reduced upon addition of β-methyl cyclodextrin. Consistent with other reports, ^{40 41} treatment of membranes with 30 mM β-methyl cyclodextrin reduced the contents of CHol by 75%. Moreover, no light-density band was obtained when Triton X-100 was not included in the lysis buffer. The following paragraphs describe lipid compositional differences among the fractions isolated from nuclear and cortical membranes of clear adult human lenses, as well as the PL compositions of clear and cataractous membranes. In addition, a brief protein analysis of the DIFs is presented. 

PL Composition by ³¹P-NMR Spectroscopy

The ³¹P-NMR spectral traces obtained for the total membrane extracts of clear and cataractous tissues were comparable within the uncertainty of the measurements. Furthermore, with DHSM and SM comprising 54% ± 5% and 10% ± 3%, respectively, of the total PLs, the composition of total clear membranes is in agreement with results reported in older lenses. ¹² The PL contents of total membrane extracts from clear or cataractous lenses were not significantly different from those acquired for the corresponding (either cortical or nuclear) SDIFs. The NSDIFs, only observed for clear lenses, exhibited compositional differences compared with the corresponding SDIFs. Figure 2 shows the ³¹P-NMR spectra of the cortical NSDIF and SDIF of clear lens membranes. Although the major components were the same in both fractions, two of the minor PLs present in the SDIF spectrum, (unknown U1 at 1.3 ppm and lysophosphatidylglycerol [LPG] at 1.2 ppm) were undetectable in the corresponding cortical NSDIF. The same trends were seen among the different nuclear fractions (data not shown). 

Although the PL components present in the total membranes, SDIF, and NSDIF isolated from the cortex of clear lenses were identical to those in the corresponding nuclear fractions, there were slight variations in their relative content. The most significant change was the decrease in the contents of U1, LPG, and PC from 4.8% ± 0.5%, 3.7% ± 0.5%, and 3.0% ± 0.3%, respectively, in total cortical membranes, to 2.2% ± 0.4%, 2.5% ± 0.3%, and 1.9% ± 0.2% in total nuclear membranes. 

The total amount of PLs was evaluated from the spectral traces obtained for each fraction after addition of DMPA as an internal standard. This approach allowed the quantification of the portion of total PLs present in each NSDIF. As shown in Figure 3 , for normal lens fractions the relative amount of total PLs present in the nuclear NSDIFs was 2.9% ± 0.3%, more than twice that in the cortical fractions (1.3% ± 0.1%). 

PL Acyl-Chain Characterization by MALDI-TOF MS

Although ³¹P-NMR bands allow the quantification of PLs with different head groups, they provide no information on the length and degree of unsaturation of the hydrocarbon chains. MALDI-TOF MS, in contrast, allows the detection of ionizable lipids and, from the value of m/z, it is possible to establish their acyl-chain distribution. There are, however, two drawbacks. Among PLs, zwitterionic lipids bearing a positive charge in their head groups have significantly higher ionization efficiencies than anionic PLs. Therefore, the mass spectra of the human lens lipids provide excellent signals for PCs, SMs, and DHSMs but not for other PLs, such as phosphatidylglycerols (PGs). The second limitation of MALDI-TOF MS lies in the relatively poor reproducibility of the peak intensities. Clearly, this affects the precision of the lipid quantification and makes the addition of internal standards imperative. Because we pursued the quantification of not only the detectable PLs, but also CHol, two internal standards were used, SM(6:0) and d₆-CHol. Neither of these compounds is a constituent of the analyzed fractions. 

Figure 4 shows relevant MALDI-TOF mass spectral regions corresponding to the cortical SDIF from clear lenses, with internal standards added. Besides the protonated ions M+H⁺, the corresponding sodiated M+Na⁺ species were detected. In addition, the potassiated M+K⁺ adducts were also observed for the most abundant PL components. For glycerophospholipids, such as PCs, the total number of carbons and the number of double bonds in the two acyl chains linked to the glycerol-based backbone are listed. In the case of sphingophospholipids, there is only one acyl chain attached to the 18-carbon long sphingoid base, either sphinganine (18:0) for DHSM or sphingosine (18:1) for SM. Therefore, the data shown correspond to the number of carbons and total number of unsaturation sites of that acyl chain. 

CHol was detected at m/z = 369.36 and corresponds to the ion formed by CHol protonation and subsequent water elimination: (CHol+H⁺)-H₂O. ^{42 43} As shown in Figure 4 , the peak for d₆-CHol can be easily discerned from that of natural CHol in the sample, because the differences in their masses is 6 D. 

Although the peaks observed in the nuclear fractions of clear lenses had the same m/z as in the cortical fractions, their relative intensities were not identical. Figure 5 shows the acyl-chain distribution of DHSMs, SMs and PCs in total membranes (white bars), SDIFs (gray bars), and NSDIFs (black bars) corresponding to cortical (Fig. 5A) and nuclear (Fig. 5B) membranes of clear human lenses. The comparison of trends in Figure 5 shows one significant difference in the acyl-chain distribution of the total membrane extracts with respect to the DIFs: Both cortical and nuclear total membranes exhibited a greater relative amount of DHSM(16:0) and a smaller content of DHSM(24:1) than did the SDIFs and NSDIFs. Although there are no other significant differences among the various fractions, the amount of PCs in all nuclear fractions was lower than in the corresponding cortical fractions, in agreement with ³¹P-NMR results presented earlier. 

Within the uncertainty of the quantification (10%–20% relative standard deviation [RSD]), there were no statistically significant differences in the acyl-chain distribution between total membranes from clear and cataractous tissues. 

Determination of CHol-to-PL Molar Ratios

MALDI-TOF MS and ³¹P-NMR spectroscopy served as complementary tools to quantify both CHol and PLs in the various fractions. MALDI-TOF MS was useful in the quantification of zwitterionic PLs, such as PCs, SMs, and DHSMs, and it also allowed the evaluation of CHol. ³¹P-NMR spectra, however, provided quantitative results for all PLs, but not for CHol, because the steroid does not possess phosphorus nuclei. DHSMs, the most abundant PLs in all fractions, were detected with high sensitivity and precision by both methods, and their relative content served as a link between the compositional results obtained through the two methodologies. In other words, MALDI-TOF MS allowed the evaluation of the ratio of CHol to total DHSMs and, through ³¹P-NMR, the amount of DHSMs to all other PLs was determined. 

The total CHol present in NSDIFs was evaluated with the use of d₆-CHol, as previously mentioned. Figure 3 shows that 14.1% ± 2.7% of the CHol found in the total nuclear membranes of clear lenses resides in the NSDIF. This percentage is nearly three times greater than that evaluated for cortical membranes, in which only 5.5% ± 1.1% of the total CHol was present in the NSDIF. 

With the amounts of CHol and PLs quantified, the molar ratio of CHol to PLs was calculated for each fraction. Figure 6 shows these values for total membrane extracts, SDIFs, and NSDIFs obtained for the cortex and nucleus of adult clear human lenses. The unusually high enrichment of CHol in the NSDIFs is obvious in both cortical and nuclear regions that exhibited CHol-to-PL ratios of 6.6 ± 2.0 and 7.0 ± 2.1, respectively. Less dramatic are the differences between the ratios in the total membranes and SDIFs. However, in both cortical and nuclear SDIFs, the ratio of CHol to PL exceeded 2, whereas in the total membrane extracts it was below 2. These results are in agreement with reports that have proposed that at CHol-to-PL ratios exceeding 2, lipids become insoluble in nonionic detergents at low temperatures. ^{44 45}  

Protein Analysis

MALDI-TOF mass spectra of the SDIFs and NSDIFs from clear lens membranes are shown in Figure 7 and indicate that there are proteins present solely in NSDIFs, as well as proteins with higher relative content in NSDIFs than in SDIFs. The quantification of these differences by SDS-PAGE followed by conventional staining with Coomassie blue was pursued. However, the presence of overlapping bands between 18 and 28 kDa did not allow for unequivocal determination of differences between SDIFs and NSDIFs. Because the masses of those proteins with higher intensity in the mass spectral profile of NSDIF (Fig. 7 , asterisks) fell within the range of molecular weights of caveolins, known markers for caveolae, we then pursued the detection of caveolin-1 and -2 by Western blot analysis. Figure 8 shows the positive detection of caveolin-1 (lane 1) in the cortical NSDIF. Two isoforms, one at 22 kDa and a more intense one near 24 to 25 kDa, were observed in both SDIFs and NSDIFs, corresponding to cortical and nuclear regions of clear lenses. However, the relative intensities of the normalized spots indicated an enrichment of 1.42 ± 0.04 in the 24- to 25-kDa isoform in NSDIFs compared with SDIFs. Neither caveolin-2 (Fig. 8 , lane 2) nor the 110-kDa raft marker GPI-phospholipase D (not shown) was detectable under the experimental conditions of this study in any of the fractions. 

Discussion

There are two relevant findings in this study: first, domains highly enriched in CHol are present in membranes of clear human lens fibers and their enrichment is greater than that reported for any other membrane; and second, these domains of low density are undetectable in cataractous tissues. 

For both cortical and nuclear regions of clear lenses, the CHol-to-PL molar ratios were above 2 for the SDIFs, whereas for the total membrane extracts they were slightly below 2. These ratios are within the range reported in total membranes. ^{3 46 47 48} Furthermore, our results on the lack of statistically significant compositional differences in membrane PLs from clear and cataractous tissues are in agreement with other reports. ^{5 49} Moreover, no significant changes in the CHol-to-PL ratios in cataractous versus clear total membranes have been reported. ^{3 5} Yet, in spite of these similarities, only clear lenses exhibited fractions of low density for which the CHol-to-PL ratio was approximately 7. This enrichment of CHol in NSDIFs is greater than that of any reported raft or caveolae. It is evident that the large amounts of the steroid are responsible for the low density of the NSDIFs. Indeed, after depletion of approximately 75% CHol with methyl-β-cyclodextrin, the NSDIFs were no longer observed. 

These findings prompt the following questions: What factors or components might be responsible for the differences in solubility and density in the isolated fractions? What function(s) could be associated with these CHol-enriched domains that are detectable only in membranes from clear lens tissue? 

Our ³¹P-NMR results indicate that, although the PL composition of total membrane extracts from clear lenses and the corresponding SDIFs were comparable, they contained LPG and an unknown PL that were not observed in NSDIFs. U1 has been reported exclusively in the human lens, whereas LPG was detected in human and several nonhuman mammalian lenses. ^{12 50} The relatively high chemical shift of the LPG resonance was related to formation of an H-bond between the hydroxy moiety and the phosphate group. ³⁹ With similar reasoning, it is likely that the resonance at 1.3 ppm may be due to a lysophospholipid. Exclusion of these PLs from NSDIFs may be a consequence of the high order in the CHol-rich domains. With the loss of one acyl chain, the shape of lysolipids is altered and may not be able to fit well in a highly ordered environment. However, given that these two PLs make up less than 10% of all PLs in total membrane extracts or SDIFs, their absence in the NSDIFs may not be a predominant cause but rather a consequence of the remarkably high recruitment of CHol into these domains. 

Regarding acyl-chain distribution of DHSMs, SMs, and PCs, the only variation was observed between the total membrane extracts and the DIFs (both SDIFs and NSDIFs) as seen in Figure 5 . DHSM(24:1) was present in relatively higher levels in both DIFs than in the total membrane extracts (cortical or nuclear fibers of normal lenses). The nervonoyl chain in these lipids is unsaturated, with the cis double bond in the 15th position. We postulate that this unsaturation site creates a bend at a location ideally suited for the accommodation of one (or two) molecule of CHol, as shown in the hypothetical arrangement illustrated in Figure 9 . This packing could maximize hydrophobic interactions, thus enhancing the resistance of the DIFs to detergent solubilization. It is possible that the increase in the relative content of DHSM(24:1) at the expense of DHSM(16:0) in the DIFs (compared with the total membrane extracts) facilitates the partitioning of extra CHol in these fractions and accounts for the increase in the CHol-to-PL ratio from 1.7 (total membranes) to 2.2 (SDIFs). However, because there were no differences in acyl-chain distribution between sedimenting and nonsedimenting fractions, the higher levels of CHol in NSDIFs than in SDIFs could not be explained by changes in acyl-chain length or degree of unsaturation. 

The lipid contents of the NSDIFs in cortical and nuclear fractions of normal lenses were similar, except for a decrease in the relative content of PCs in the nuclear fractions compared with the cortical ones. The portion of lipids present in these domains was, however, more than twice higher in the nucleus than in the cortex. It is possible that this trend is the result of an age-dependent concentration effect. The fibers present in the inner nucleus of an adult lens are those created during embryonic and fetal stages. Because the growth of the human lens is at least one order of magnitude higher in prenatal than in postnatal stages, ^{51 52} it is possible that during gestation the rapidly expanding fiber membranes are more disordered, with higher levels of unsaturated PLs. As these fibers age, the relative amounts of unsaturated glycerolipids may diminish progressively due to oxidation and/or hydrolysis by phospholipases. As a result, sphingolipids with no unsaturation sites or a single site that does not preclude strong interactions with CHol, become relatively more concentrated and can segregate into clusters. 

The lack of substantial variations in the PL content of fractions of different density led us to consider that the unusual enrichment in CHol in NSDIFs could be related to the presence of CHol-recruiting proteins. This possibility has been demonstrated in neuronal rafts isolated from rat brains in which NAP-22, a myristoylated protein with a high content of β-sheet conformation, was reported to be the driving force for clustering. ³⁶ In alignment with this hypothesis, our preliminary mass spectral studies indicated differences in protein profiles obtained for the fractions of low and high density. Furthermore, Western blot analysis confirmed the enrichment of caveolin-1 in the buoyant fractions. This membrane protein is known to have a long hydrophobic region likely to permit or induce its insertion into the lipid bilayer. Furthermore, caveolin-1 has been implicated in intracellular traffic of CHol. ⁵³ We predict that other specific proteins are enriched in these domains and that they may function in concerted ways yet to be unraveled. 

This analysis leads us to hypothesize that the clustering of CHol around specific proteins may be relevant to the control of their structure and thus their function. Moreover, the absence of low-density fractions in cataractous membranes suggests that the disruption of these domains may contribute to loss of transparency. A likely challenge to the organization of these domains could be the massive association of crystallins to membranes in cataractous tissues. This would enhance the density of the clusters and prevent their buoyancy in a sucrose density gradient. The greater association of total membranes with modified crystallins in cataractous tissues than in aged clear lenses has been reported. ⁵⁴ In addition, the abnormal morphology of the membranous envelope surrounding focal opacities has been demonstrated by in situ spectral studies conducted by Duindam et al. ⁵⁵ As a result of excessive protein binding, conformational changes may be induced in either exposed portions of membrane proteins and/or headgroups and interfacial regions of PLs present in the liquid-ordered domains. Either type of modification could impair the trafficking and/or signaling properties often associated with caveolar proteins. 

With the finding of low-density, detergent-resistant domains, cortical and nuclear fibers of clear human lenses can now be added to the expansive list of cells in which these domains are present. The absence of these domains in cataractous tissues with extended opacification suggests that the integrity of the clusters is essential for the function of the proteins sequestered therein. Thorough protein mapping and investigation of the specific roles played by proteins enriched in these domains are of utmost importance, because they may be critical to the maintenance of human lens transparency. 

 Supported by National Eye Institute Grant EY11657.

 Submitted for publication August 2, 2002; revised October 9, 2002; accepted October 24, 2002.

 Disclosure: M. Rujoi, None; J. Jin, None; D. Borchman, None; D. Tang, None; M.C. Yappert, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Extraction of DIFs from human lens membranes. The cortical (or nuclear) region of adult human lenses was homogenized in PBS solution. The water-insoluble fraction was treated with 1% Triton X-100 in TNE buffer. The DIF was subjected to sucrose linear density gradient centrifugation. The light-scattering band observed in the low-density region of the gradient was harvested and labeled NSDIF. The pellet was marked SDIF.

Figure 2.

 

³¹P-NMR spectra of cortical DIFs from adult clear lenses. The ³¹P-NMR spectra obtained for the PLs extracted from the cortical NSDIF (▪) and SDIF ( ) show that LPG (δ = 1.2 ppm) and an unknown U1 (δ = 1.3 ppm) were present only in the SDIF (and total membrane extract, not shown). The high chemical shift of U1 suggests that this PL may be a lysoproduct.

Figure 3.

 

Amount of total (cortical or nuclear) PLs and CHol present in NSDIFs from adult clear lenses. The portions of total membrane PLs (□) and CHol ( ) that are present in the cortical and nuclear NSDIFs were evaluated from ³¹P-NMR and mass spectral data. The average and SD correspond to 10 independent determinations. The relative amount of total lipids present in these fractions was nearly three times higher in the nuclear than in the cortical membranes.

Figure 4.

 

MALDI-TOF MS spectrum of the lipids extracted from the cortical SDIF from adult clear lenses. The shown spectral regions highlight some of the PL-related peaks that correspond to the protonated M+H⁺, sodiated M+Na⁺, and potassiated M+K⁺ ions. These ions were identified by comparison of the experimental and theoretical m/z ratio of the monoisotopic peaks. A tolerance of Δ(m/z) = 0.06 was set up to discriminate against background. CHol was detected as the ion formed on protonation and subsequent water elimination. ^{42 43} For the quantification of PLs and CHol, two internal standards (in italics) were added to the sample: SM(6:0) and d ₆ -CHol, respectively. The length and number of unsaturation sites of the acyl chain connected to the (18:1) sphingosine (for SMs) or (18:0) sphinganine base (for DHSMs) are noted. For PCs, the total length and number of double bonds of the two acyl chains are indicated.

Figure 5.

 

Acyl-chain distributions of PCs, SMs, and DHSMs in total membranes and DIFs for (A) cortical and (B) nuclear membranes from adult clear lenses. From the m/z ratio in the mass spectrum, the acyl-chain composition (length and number of sites of unsaturation) of PCs, SMs and DHSMs was evaluated in total membranes (□), SDIFs ( ) and NSDIFs (▪). The averages and standard deviations were estimated from the mass spectra acquired from 10 independent sample droplets. Compared with the lipids extracted from total membranes, both DIFs exhibited a relative decrease in DHSM(16:0) and an increase in DHSM(24:1).

Figure 6.

 

Molar ratio of CHol to PLs in cortical and nuclear total membranes and DIFs from adult clear lenses. The ratio in total membranes (□) was less than two in both cortical and nuclear membranes, but exceeded two in SDIFs ( ). The relative content of CHol in NSDIFs (▪) was unusually high, exceeding 6 for cortical and nuclear fractions. The average and SD correspond to 10 data sets.

Figure 7.

 

MALDI-TOF MS protein mass analysis of cortical DIFs from aged clear lenses. The mass spectral traces correspond to the NSDIF (top trace) and SDIF (bottom trace) obtained from cortical fiber membranes. Proteins that are either present in the NSDIF exclusively, or with higher relative amount than in the SDIF are indicated by asterisks.

Figure 8.

 

Identification of caveolin-1 in NSDIFs by SDS-PAGE and Western blot analysis. Western blot analysis with antibodies against caveolin-1 (lane 1) and -2 (lane 2) demonstrated the presence of caveolin-1 isoforms at ∼22 and 24 to 25 kDa and the absence of caveolin-2 in NSDIFs. Rap2 (21 kDa) and GRB-2 (24 kDa) were among the molecular weight standards.

Figure 9.

 

Hypothetical representation of the packing of CHol in membranes containing DHSM(24:1). This arrangement illustrates the proposed partitioning of CHol within DHSMs, facilitated by the cis double bond in position 15 of the nervonoyl chain.