INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Studies over the last two decades have shown that alveolar surfactant exists in a number of discrete structural subtypes, namely a lamellar body-like (LB) form, a tubular myelin (TM) form, and a small vesicular (SV) form (1). Surfactant lavaged from the alveolar spaces has been separated into these subtypes by a variety of centrifugation methods on the basis of their sedimentation velocity (2) or their buoyant density (2, 6). These studies have yielded important information about the phospholipid and protein compositions, metabolic relationships, ultrastructures, and the surface tension properties of the various subtypes. More recently, surfactant subtype composition has been examined in a variety of special situations, for example lung development (10) and lung injury (11).

The methods of subtype separation have not been standardized and have varied among studies, leading to uncertainties of interpretation. As analysis of surfactant subtypes promises to yield additional insights into surfactant physiology, it is necessary to know whether the two principal methods of subtype analysisdifferential centrifugation (DC) and equilibrium buoyant density centrifugation (EBDC)yield comparable results, and if they do not, what limitations investigators should bear in mind.

The purpose of the present study was to compare the two widely used methods for the separation of alveolar surfactant into subtypesDC and EBDCand to examine if they yielded results that were similar in terms of the relative proportions of each subtype. Our findings suggest that these two methods yield results that are systematically and significantly different, and that the differences could not be reconciled by minor adjustments in methodology.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

All experiments were performed on CF₁ female mice (Carworth Farms, Portage, MI) age 8 to 10 wk. Mice were specific-pathogen-free and maintained under barrier conditions and fed standard lab chow and water ad libitum until assay.

To label surfactant phospholipids, mice received intraperitoneal injections of 5 to 10 µCi [³H]choline or 1 to 3 µCi [¹⁴C]choline (New England Nuclear, Boston, MA) 6 to 18 h before assay. They were killed by an overdose of intraperitoneal pentobarbital and exsanguinated by transection of the abdominal aorta after which the trachea was exposed and cannulated to obtain surfactant by lavage (lavage buffer 0.15 M NaCl, 2 mM CaCl₂, 5 mM Hepes buffer, pH 7.40 at 37° C) as described previously (6). Lavages from 4 to 10 mice were pooled and centrifuged at 500 × g × 10 min to sediment cells (in previous experiments (6) this initial centrifugation sedimented up to 5% of the total phospholipids in the bronchoalveolar lavage [BAL]). The BAL supernatant was used immediately for subtype separation studies. The yield of surfactant phospholipid was approximately 1.0 mg phospholipid per mouse in 5 ml of BAL (0.2 mg · ml¹) and its specific activity was approximately 2,000 disintegrations per minute (dpm) · mg phospholipid¹ after either isotope.

For DC, separation of surfactant subtypes was based on the methods described by Magoon and coworkers (2) or Veldhuizen and coworkers (3) with modifications as stated in the text. In general, replicate 6-ml aliquots of BAL (sometimes diluted with lavage buffer) were centrifuged in a Ti50.3 fixed-angle rotor at g-forces and times stated in RESULTS at 4° C in a Beckman L8-M ultracentrifuge. The supernatant and pellet were purified for phospholipids by the method of Folch and coworkers (12) and assayed for radioactivity.

For EBDC separation of surfactant subtypes, the method previously described (6) was used. Replicate 2- to 4-ml aliquots of the same pool of BAL as used for DC separation were underlaid with continuous sucrose gradients (8 to 10 ml of 0.8 to 0.1 M sucrose dissolved in lavage buffer) and centrifuged in a Ti70 rotor at 50,000 rpm (196,000 × g) at 8° C for 16 h. In some experiments, as stated in RESULTS, parallel 4-ml samples were centrifuged through 8 ml of similar sucrose gradient in a SW41 rotor at 35,000 rpm (175,000 × g) at 8° C for 16 h. After centrifugation, the gradients were separated from below into 15 to 20 fractions, the densities of alternate fractions were determined by refractive index, and the radioactivity of each fraction was assayed after purification of phospholipids (12).

For analysis of results, radioactivity was taken as an index of phospholipid amount. For DC, the radioactivities of the pellet (large aggregates, LA) and supernatant (small aggregates, SA) were used. For EBDC, the area under each buoyant density peak of radioactivity was used. This was calculated by a software program (Peakfit; Jandel Scientific, San Rafael, CA) which was allowed to find the best fit of the gradient radioactivity profile to three peaks corresponding to "ultraheavy" (UH), "heavy" (H), and "light" (L) subtypes (6). The fit was accepted only when r² exceeded 0.95 (p < 0.0001) and when the peaks were centered at buoyant densities that were consistent with previous reports (6), otherwise the data for that gradient were rejected. Approximately 10% of the gradients were thus rejected. For purposes of presentation, the UH peak, which, when present, was found to be a small component in most gradients, was combined with the H peak into a single datum called either UH + H or simply H subtype. Statistical comparisons of data were performed by Student's t test or analysis of variance (ANOVA).

Surfactant protein B (SP-B) was assayed by dot-blot analysis. Samples of surfactant were treated with trichloracetic (TCA, final concentration 5%) at 4° C for 1 h and sedimented in a microfuge. The pellet was resuspended in 0.1 M NaOH with 1% sodium dodecyl sulfate and serial 2-fold dilutions were spotted onto nitrocellulose membrane. After drying, the membrane was equilibrated with TBST buffer (10 mM Tris pH 8.0, 150 mM NaCl, 0.05% Tween-20) for 10 min, and then blocked with the same buffer containing 5% bovine serum albumin. The membrane was incubated for 3 h at 4° C with the primary antibody, rabbit anti-SP-B (gift of Dr. J. Whitsett, University of Cincinnati), washed three times with TBST, and then incubated for 1 h at 4° C with the secondary antibody, horseradish peroxidase- linked donkey anti-rabbit Ig (Amersham, Arlington Heights, IL). After three further washes the membrane was soaked for 5 min in LumiGlo chemiluminescence substrates (KPL, Gaithersburg, MD) and exposed to film. In a preliminary assay, equal volumes of each fraction of a sucrose gradient of mouse lavages were assayed. Abundant SP-B was detected in the densest 1 to 3 fractions and in fractions that contained UH and H subtypes, but not in fractions that contained L subtype nor in material at the top of the gradient (data not shown), confirming that SP-B could be used as a marker for UH and H subtypes, as previously shown (3, 13).

Mouse serum was obtained by transection of the abdominal vessels of mice that had received an overdose of pentobarbital. The aspirated blood was allowed to clot and was centrifuged to yield serum which was stored at 70° C until use.

The hypothesis tested was that if DC and EBDC methods provide equivalent estimates of subtype amounts, the activity of the DC pellet (LA) would equal that under the UH plus H peaks of EBDC fractions, and that the activity of the DC supernatant (SA) would equal that under the L peak of EBDC fractions.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Four sets of experiments were performed. In each, a pool of BAL surfactant was separated into subtypes by DC and EBDC in replicate parallel assays.

In the first experiment we examined the reproducibility of each method and compared the proportional yields of each subtype by DC and by EBDC. Mice were labeled 18 h prior to assaya time at which all surfactant phospholipid compartments in the lung are of approximately equal specific activity (14). Aliquots from a single pool of BAL of mice were separated by DC (40,000 × g × 15 min, or 10 K · g · h, Ti50.3 fixed-angle rotor) and by EBDC, each in 6-fold replicates. We found (Table 1, upper panel ) that both methods yielded data that were highly internally consistent but that the two methods yielded different results. Almost all the starting counts were recovered in the DC method, but it yielded a substantially and significantly lower proportion of total phospholipid radioactivity in the pellet (LA) than the EBDC method (UH + H subtype), p < 0.0001. Correspondingly, the proportion of activity in the supernatants of DC (SA) was significantly greater than that found at the density corresponding to L subtype on EBDC, p < 0.0001. 

To determine whether the discrepancy between the two methods of subtype separation could be reconciled by increasing the time and g-force of DC, we increased these by a factor of 2 to 48,000 × g × 25 min (or 20 K · g · h) using another pool of BAL in 4-fold replicates. Table 1 (lower panel ) shows that this adjustment in g-force reduced the discrepancy slightly, but a large and significant difference between the two methods remained. In this experiment, EBDC was also performed in a swinging bucket rotor (SW41) with results that were essentially identical to those obtained from the Ti70 fixed-angle rotor, showing consistency between the two EBDC methods.

In several other experiments (data not shown) DC was performed by careful aspiration of the supernatant from above rather than by decanting it (in order to eliminate the possibility that some of the pellet of DC was poured off by decanting), or in SW41 rotor rather than Ti50.3. These measures did not significantly reduce the apparent underrepresentation of LA by DC as compared with EBDC.

Also noticed (Table 1, both panels) is that the sum of radioactivity under the phospholipid peaks in EBDC gradients was consistently less (by approximately 10%) than the sum of radioactivity in the pellet and supernatant of DC separations, p < 0.001. Much of the activity missing from EBDC data could be explained by that recovered from the first 1 to 3 (most dense) fractions and from the last (least dense) fraction of the EBDC gradients (for example, see next experiment, Figure 1, left panel ). Neither of these very dense or very light fractions corresponds to known surfactant subtypes and both were excluded by the curve-fitting program. However, each would be included in the pellet and supernatant fractions of DCsa potential additional source of discrepancy.

View larger version (14K): [in this window] [in a new window]   Figure 1.   EBDC gradients of first experiment. Duplicate gradients of starting BAL (left panel ) and the supernatants of differential centrifugations (DC, 48 K × g × 25 min, right panel ). The peak with buoyant density of 1.05 g · ml1 represents heavy subtype, the peak at 1.025 g · ml1 represents light subtype, showing that the supernatant of DC contains substantial heavy subtype.

As these results suggested that DC did not sediment all of the heavier subtypes, we separated a pool of BAL by DC (20 K · g · h) and then subjected each of the replicate supernatants to subsequent EBDC to determine whether they contained any unsedimented UH + H subtype. The result (Figure 1) shows duplicate EBDC gradients of the starting material (left panel ) and the supernatants of the 20 K · g · h DC (right panel ). The DC supernatants contained a considerable amount of phospholipid that banded at a mean buoyant density of approximately 1.052 g · ml¹ consistent with being H subtype that had not been sedimented by prior DC. This confirms that DC failed to sediment all of the material that corresponds to UH and H subtypes in EBDC.

In the second group of experiments we performed serial DCs and EBDCs of a single pool of BAL. The aim was to determine if additional DCs could reconcile the apparent discrepancy between the DC and EBDC methods of subtype separation, particularly with respect to the apparent underrepresentation of heavier subtypes in the DC pellets. In order to preferentially label the heavier subtypes, mice were labeled 6 h before assay in this experiment. Multiple aliquots of a single pool of BAL were submitted to a series of four identical serial DCs (20 K · g · h, Ti50.3 fixed-angle rotor) and duplicate EBDCs. The pellets and duplicate 1.0-ml aliquots of the supernatants of the first DC were purified for phospholipids and assayed for radioactivity. The remainders of the supernatants were pooled, a portion was removed for EBDC analysis, and the rest was recentrifuged in equal aliquots in the same manner (20 K · g · h, same rotor). The pellets and duplicate 1.0-ml aliquots of the supernatants of the second centrifugation were again assayed as previously described. The remainders of the supernatants were again pooled and recentrifuged in the same manner a total of four times, yielding four replicate sets of pellets and supernatants, each of which was assayed for phospholipid radioactivity, and EBDCs of each of the four sets of DC supernatants.

The results of the four serial DCs are shown in Table 2 on the left; pellets are called P1-P4 and supernatants are called S1-S4; amounts throughout are expressed as percentages of the total starting activity. The first DC sedimented about half of the total initial activity. Recentrifugation of the supernatant (at the same g-force for the same time) yielded only a small additional amount of activity in the second pellet (8.1% of the total starting activity), and further centrifugations sedimented progressively smaller amounts. The sum of activities in the pellets of the four sequential centrifugations was 68.1%, the final supernatant containing the remainder.

EBDC analysis of the starting material and each DC supernatant is shown in Table 2, right panel. This shows that the proportion of starting material that was of UH + H buoyant density (67.5%) was significantly higher than that sedimented by the first DC (50.2%), similar to the result of the first experiment. However, the supernatant of the first DC, called S'1, contained a substantial amount of H subtypeapproximately the same proportion of H to L subtypes as in the starting material and not substantially less, as would be expected if DC preferentially sedimented H subtype. Moreover, the H/L subtype ratio in the subsequent supernatants, called S'2-S'4, seemed to increase with successive DCs. This result implies not only that DC failed to sediment all of the H subtype, but also that some of the L subtype must have been sedimented by each of the serial DCs.

In a parallel experiment using the same DC protocol as shown in Table 2, we assayed the SP-B content of 100-µl aliquots of the starting BAL pool and each DC pellet and supernatant, using SP-B as a marker for heavier subtypes (3, 13). (Each DC pellet was resuspended in a volume of lavage buffer equivalent to the volume of the supernatant, from which 100 µl was removed for SP-B assay to ensure equivalence of sample volumes.) The result (Figure 2) shows that the starting pool contained abundant SP-B. The first DC pellet, P1, contained approximately one-half the amount of SP-B contained in the starting pool, but the supernatant, S1, also contained a substantial amount of SP-B. Moreover, subsequent DC pellets, P2-P4, contained very little SP-B, although SP-B was detectable in subsequent supernatants, S2-S4. This result accords closely with the amounts of heavy subtypes detected by EBDC of the supernatants (Table 2) and tends to confirm that DC fails to sediment all of the heavier subtypes. This experiment was performed three times with essentially similar results.

View larger version (48K): [in this window] [in a new window]   Figure 2.   SP-B dot blots of samples from experiment 2, starting BAL (upper panel ) and serial DC pellets (P1, ...P4, lower left panel ) and supernatants (S1, ...S4, lower right panel ), for comparison with data in Figure 2. Dot blots were made from equivalent volumes of each sample and analyzed in 2-fold serial dilutions, as indicated, showing that the supernatants of DCs contain SP-B consistent with their containing heavy subtype.

In the third set of experiments, we analyzed further the sedimentation behavior of H and L subtypes in DC by making pure preparations of [¹⁴C]-labeled H subtype and [³H]-labeled L subtype as follows. For pure H subtype, BAL was obtained from mice that had received [¹⁴C]choline 6 h before they were killed. After a preliminary low-speed centrifugation, the supernatant was centrifuged at 20,000 × g × 30 min (10 K · g · h, SW41 rotor). The pellet was gently resuspended and dispersed in a glass/Teflon homogenizer (one pass) and used as H subtype. For pure L subtype, BAL was obtained from mice that had received [³H]choline 18 h prior to sacrifice. After a preliminary low-speed centrifugation, the supernatant was centrifuged at 175,000 × g × 90 min (262 K · g · h, SW41 rotor) to sediment heavy subtype. The supernatant was used as L subtype.

Preliminary analysis of [¹⁴C]- and [³H]-choline-labeled preparations, which were subjected to EBDC separately and in combination, showed that the [¹⁴C]-labeled subtype was predominantly of H buoyant density and contained < 5% L subtype as well as abundant SPB as a marker for H subtype; [³H]- labeled subtype contained < 5% H subtype and was devoid of SP-B, consistent with the former [¹⁴C]-labeled material being authentic H subtype, and the latter [³H]-labeled material being authentic L subtype (data not shown).

The [¹⁴C]-H and [³H]-L subtypes were combined and aliquots of the pool were subjected to four separate DC protocols of increasing g-force and centrifugation time in a Ti50.3 fixed-angle rotor, namely 22,400 rpm × 15 min (10 K · g · h), 24,500 rpm × 25 min (20 K · g · h), 27,000 rpm × 40 min (40 K · g · h), and 33,700 rpm × 60 min (90 K · g · h). The pellets and duplicate aliquots of the supernatants were purified for phospholipids and counted.

Figure 3 shows the proportions of each isotope that were found in the pellets (left panel ) and in the supernatants (right panel ) of each DC. The pellets (Figure 3, left panel ) contained predominantly [¹⁴C]-H subtype, and the supernatants (Figure 3, right panel ) contained predominantly [³H]-L subtype. However, there was considerable cross-contamination of both the pellets and the supernatants with heterologous subtype at each g-force · time DC protocol.

View larger version (17K): [in this window] [in a new window]   Figure 3.   DC separation of differentially labeled surfactant subtypes subjected to four parallel DC protocols of increasing g-force · time, experiment 3. Purified [14C]-labeled H subtype was combined with purified [3H]-labeled L subtype and centrifuged at the g-force · time shown on the abscissa. Activities of each isotope found in the pellets as percents of the total amount loaded are shown in the left panel; activities found in the supernatants are shown in the right panel. (Overall recoveries of each isotope were 95.1 ± 5.1% for [14C], and 94.7 ± 2.9% for [3H].)

As before, an aliquot of each DC supernatant was subjected to subsequent EBDC analysis. The supernatant of the 10 K · g · h DC, when analyzed on EBDC, was found to contain 27.5% of the [¹⁴C]-H subtype activity of starting material and 88.6% of the [³H]-L subtype. The supernatant of the 90 K · g · h DC contained 8.0% of the [¹⁴C]-H subtype activity of starting material, but only 52.0% of the [³H]-L subtype (EBDC gradients not shown). These amounts of cross-contamination correspond closely to those seen in Figure 3, right panel, which tends to confirm its estimates of cross-contamination. In addition, aliquots of each of the four DC supernatants were subjected to dot-blot analysis for SP-B and revealed the presence of SP-B consistent with their containing some H subtype in the DC supernatants (data not shown).

Consequently, there was no g-force · time protocol for DC that yielded complete separation of subtypes. Optimal separation was found at 20 K · g · h, but even this protocol yielded no more than 85% of authentic H and L subtypes in the pellet and supernatant respectively, but also simultaneous contamination by up to 15% of the heterologous subtype. This experiment was performed twice with essentially similar results.

In the fourth experiment we explored the effect that the presence of extraneous proteins might have on subtype separation. To aliquots of a pool of BAL we added mouse serum to final concentrations of 0, 1, 3.3, and 10 mg protein · ml¹. Replicate aliquots were separated by DC (48,000 × g × 25 min, 20 K · g · h) or by EBDC. The results of a representative experiment (Table 3) show that DC yielded a smaller proportion of the total phospholipid in the pellet than was found in the UH + H fraction of EBDC (no added serum), as in previous experiments. More importantly, the proportion of total phospholipid recovered in the pellets and supernatants after DC varied with the protein content of the BAL, significantly more being pelleted in samples that contained more protein, p < 0.01. By contrast, the subtype proportions yielded by EBDC did not vary with protein content, nor did their buoyant densities. (Much of the extraneous protein was found in the lowermost 2 to 3 fractions of the gradients and largely separate from the phospholipid peaks.) This experiment was performed a total of three times with minor modifications and with essentially similar results.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The two alternative methods of analysis of surfactant subtype composition, DC and EBDC, have been widely used. However, neither has been standardized, nor have the two methods previously been directly compared with each other. Each method has advantages and drawbacks. The DC method separates surfactant into two fractionsa pellet consisting of LA, and a supernatant consisting of SA of surfactant. It is easy to perform, and affords the rapid assay of many parallel samples, thus providing good statistics. The EBDC method requires more sample preparation, a longer centrifugation time, is somewhat more labor-intensive, and requires a more painstaking analysis and thus does not easily lend itself to replicate analyses. However, it yields a profile of phospholipid peaks in the gradient from which the mass and mean buoyant density of each subtype, UH, H, and L, can be calculated and compared with previously published values.

We compared the two methods in four sets of experiments, analyzing the subtype composition of mouse alveolar surfactant from BAL pools by each method in parallel assays. Both DC and EBDC methods were highly internally consistent among replicate analyses (Table 1). The two EBDC protocolsthat using a fixed-angle rotor and that using a swinging bucket rotoryielded results that were in close agreement with each other (Table 1, lower panel ). However, DC yielded results that varied with the centrifugation conditions (e.g., Figure 3) and that differed markedly and systematically from EBDC results of the same BAL pool.

First, the overall recovery of phospholipids from DC was consistently greater by approximately 10% than that from EBDC. We suspect this is due to the inclusion in the DC pellets and supernatants of small amounts of phospholipids that did not band in EBDC gradientsphospholipids that are found in the most dense and least dense fractions of most EBDC gradients (e.g., Figure 1). These do not correspond to known surfactant subtypes and were intentionally disregarded by the curve-fitting program used to quantitate subtypes in EBDC gradients. But as the amounts of phospholipids at very high and very low densities were relatively small and were of approximately equal size, their inclusion in the assay of large to small aggregate ratios as determined by DC would not constitute a substantial discrepancy between the two methods.

Second, the pellets of DC (LA) consistently contained less phospholipid than was contained in the heavier subtypes identified by EBDC of the same samples. Correspondingly, the DC supernatants (SA) contained relatively more than was identified as L subtype in EBDC separations (Tables 1 and 2, Figure 1). That this was due to a failure of DC to sediment all the heavier subtypes is suggested by the finding that the supernatants of initial DC, when subsequently analyzed by EBDC, were found to contain substantial amounts of the heavier subtypes, (Figure 1 and Table 2), and substantial amounts of SP-B (Figure 3), a constituent associated only with heavier subtypes (13). Indeed, experiment 2 (Table 2) suggests that a substantial proportion of H subtype may not be sedimentable by the DC conditions used in this experiment.

Furthermore, analysis of the data in Table 2 suggested that DC may also sediment some L subtype, a possibility that was supported by the presence of phospholipid not associated with SP-B in the pellets, P2-P4 (Figure 2). Similarly, the third experiment, differential labeling of H and L subtypes, showed that there was cross-contamination of authentic H and L subtypes at each DC protocol employed (Figure 3). Not only did DC fail to sediment a substantial amount of H subtype at relatively low g-forces (10 to 20 K · g · h), but 5% of H subtype remained unsedimented even at relatively high g-forces (90 K · g · h), suggesting that this proportion may exist in small aggregates of relatively high buoyant density. Moreover, this experiment probably underestimated the amount of cross-contamination that would occur with unprocessed BAL, as the method we employed to purify subtypes unavoidably eliminated those fractions that contained surfactant of intermediate buoyant density and sedimentation properties. Figure 3 therefore probably represents the least extent of cross-contamination of subtypes that would occur with unprocessed BAL.

The addition of extraneous protein to the starting BAL, the fourth set of experiments, resulted in substantial variations in apparent subtype proportions as determined by the DC method (Table 3), whereas EBDC of the same samples yielded consistent estimates of subtype proportions at each level of protein content. Although we did not compare subtype separation methods in any experimental condition in which surfactant was contaminated in vivo by increased protein, such as lung damage, there is a probability that the latter would add an additional complication to the accurate estimation of surfactant subtype composition when DC is used as the sole method for estimating subtype compositions in pathologic situations.

We conclude that the DC and EBDC methods of surfactant subtype analysis yield results that are not concordant with each other, disproving the hypothesis that was tested. The discrepancy can be attributed for the most part to failure of DC to sediment a substantial proportion of H subtype while sedimenting some L subtype. Cross-contamination occurred in DC at all g-forces and conditions we explored even when BAL was obtained from normal animals. In situations where BAL may contain excess protein, e.g., when lung damage is being compared with normal controls, the use of DC analysis without EBDC verification is likely to be particularly problematic and could lead to erroneous conclusions concerning surfactant subtype composition.

Because the methods used to separate surfactant subtypes are rarely described in much detail, subtle methodologic variations among investigators may result in discrepancies between findings from different groups. Because of this possibility, and in the interest of reducing such discrepancies, investigators may wish to bear in mind the limitations of whichever method of surfactant subtype analysis they employ.