Isolation of the Lateral Border Recycling Compartment using a diaminobenzidine-induced density shift

Abstract

The migration of leukocytes across the endothelium and into tissue is critical to mounting an inflammatory response. The Lateral Border Recycling Compartment (LBRC), a complex vesicular-tubule invagination of the plasma membrane found at endothelial cell borders, plays an important role in the this process. Although a few proteins have been shown to be present in the LBRC, no unique marker is known. Here we detail methods that can be used to characterize a subcellular compartment that lacks an identifying marker. Initial characterization of the LBRC was performed using standard subcellular fractionation with sucrose gradients and took advantage of the observation that the compartment migrated at a lower density than other membrane compartments. To isolate larger quantities of the compartment, we modified a classic technique known as a diaminobenzidine (DAB)-induced density shift. The DAB-induced density shift allowed for specific isolation of membranes labeled with HRP conjugated antibody. Because the LBRC could be differentially labeled at 4°C and 37°C, we were able to identify proteins that are enriched in the compartment, despite lacking a unique marker. These methods serve as a model to others studying poorly characterized compartments and organelles and are applicable to a wide variety of biological systems.

Introduction

Most pathology is arguably due to poorly controlled or misdirected inflammation that results in damage to normal tissues. During inflammation leukocytes are recruited out of the blood stream by a series of molecular interactions between the leukocyte and endothelium. The last step in the process is the migration of leukocytes across the endothelial cell layer that lines blood vessels (reviewed in (1–3)), and is commonly referred to as transendothelial migration (TEM). During TEM, the leukocyte squeezes through the endothelium (typically at endothelial cell borders) to gain access to the tissue beneath. Endothelial cells play an active role in this process, not only by providing critical molecular interactions with the leukocyte, but also by remodeling the junction to allow for the leukocyte to pass. Both of these functions are facilitated by a recently identified perijunctional membrane compartment referred to as the Lateral Border Recycling Compartment (LBRC) (4–6).

The LBRC was first identified by examination of the subcellular localization of Platelet/Endothelial Cell Adhesion Molecule-1 (PECAM) using immuno electron microscopy (6). This analysis showed that PECAM, in addition to the expected junctional localization, was also found in a series of interconnected vesicles and tubules contiguous with and emanating from the junctional plasma membrane. Although these membrane invaginations bear slight resemblance to caveolae, PECAM and the caveolae marker caveolin-1 do not co-localize by immunofluorescence or subcellular fractionation (6). Vascular Endothelial-cadherin (VE-cadherin), an abundant junctional protein found in adherens junctions, is notably absent from the LBRC (6, 7), suggesting that the compartment has a distinct protein make up and is not simply an invagination of the plasma membrane.

Our studies also showed that although the LBRC remains continuously open to the extracellular milieu, it is inaccessible to antibodies and macromolecules larger than 50 kDa when incubated with endothelial cells at 4°C (6). Although the mechanism behind this ‘gating’ at 4°C is currently unknown, this observation allows for an indirect method of tracking the LBRC. After labeling the LBRC and surface pools of PECAM at 37°C with Fab fragments of a non-blocking monoclonal antibody, the cells can be chilled and the surface pool blocked with an unconjugated secondary antibody. The remaining unblocked LBRC pool can then be visualized with fluorescent secondary as it becomes exposed to the surface when the monolayer is returned to 37°C. By following the increase in fluorescence over time, we found that PECAM in the LBRC exchanges with PECAM in the plasma membrane at the junction with a half-time of ~10 min (6). During TEM however, significantly more fluorescence is observed around the migrating leukocyte indicating that the LBRC is selectively delivered to that region of the cell border. Treating the endothelial cells with agents that disrupt microtubules or inhibit kinesin function significantly inhibits TEM, indicating that the targeted delivery of the LBRC is required for efficient transmigration (4). Together these findings indicate that the LBRC functions to deliver both relevant proteins and additional membrane to facilitate TEM.

As with any organelle or subcellular compartment, our understanding of its function would be significantly advanced by a more comprehensive knowledge of its protein constituents. Since its initial identification, only a few other integral membrane proteins have been found in the LBRC, namely poliovirus receptor, JAM-A, and CD99, all of which have been shown to be involved in TEM (5, 7–9). The only method available to identify other components is to screen candidates using immuno electron microscopy, which is inefficient, costly, and limited to antigens accessible from the exterior of the cell to which antibodies are available.

As an alternate to this candidate-based approach, we sought to isolate the LBRC for proteomic analysis to identify proteins that are enriched or unique to it. However, because all of the known LBRC proteins are also found on the plasma membrane, any method based on isolating PECAM (or CD99, JAM-A, or poliovirus receptor) would co-purify plasma membrane. Likewise, because there is no known unique LBRC marker, it would be difficult to judge the purity of the isolated LBRC fractions. The process would be significantly aided if a specific LBRC marker were known. Unfortunately though, we were stuck in a logic loop; needing a specific marker to purify the compartment but needing to isolate the compartment to identify a specific marker.

In an attempt to break out of the loop, we first sought to isolate the LBRC using standard subcellular fractionation. We reasoned that we could exploit two known characteristics of the compartment to follow it during the fractionation process. Previous work using biochemical methods and electron microscopy has shown that the LBRC 1) does not contain VE-cadherin and 2) is inaccessible to antibodies at 4°C (5–7). Although informative, this approach did not yield sufficient material for proteomic analysis. To recover more, we modified a classic technique, the diaminobenzidine (DAB)-induced density shift (10–12). Using this method we recovered samples enriched in the LBRC and were able to identify IQGAP1 and Vimentin as LBRC-associated proteins.

Results

Subcellular fractionation – flotation gradient

Antibodies (e.g. anti-PECAM) incubated with confluent endothelial cell monolayers at 4°C are excluded from the LBRC and only label the surface and junctional pools of the protein. We reasoned that, after fractionation, membranes from the LBRC could then be identified by the absence of both VE-cadherin and the labeling antibody. Toward this end, we labeled several confluent human umbilical vein endothelial cell (HUVEC) monolayers with anti-PECAM antibodies at 4°C. The cells were then washed and collected by scraping. Cells were then homogenized in buffer without detergent to preserve membrane integrity and composition. After clearing large cellular debris, the homogenate was loaded at the bottom of a 10–40% continuous sucrose gradient and centrifuged at ~130,000 × g overnight. During ultracentrifugation, membranes migrate in the gradient to a density that corresponds to their inherent density, determined by their unique lipid and protein makeup. Fractions were collected and their protein constituents recovered using precipitation with trichloroacetic acid (TCA) and analyzed by western blotting.

In this simple fractionation plasma membrane migrated towards the medium density fractions, as determined using the plasma membrane/early endosome marker Transferrin Receptor (TfR) (Figure 1A). VE-cadherin, a marker of endothelial cell junctions, had an identical distribution as TfR as expected since it is also a plasma membrane protein. LAMP2, a marker of late endosomes and vacuoles had a broader distribution, migrating across the medium and high density fractions. Interestingly, PECAM, which is known to localize to both the plasma membrane and LBRC, had a broader distribution and was found co-migrating not only with the other plasma membrane markers but also migrated to several low density fractions. These low density fractions did not contain any of the other markers examined, namely endosomes (early endosome antigen-1, EEA1), lysosomes (lysosomal-associated membrane protein 2, LAMP2), or endoplasmic reticulum (calreticulin). The fractions also did not contain the antibody used to label the surface and junctional PECAM during the initial monolayer incubation at 4°C. Based on the absence of both VE-cadherin and the labeling antibody, we concluded that these low density fractions are enriched in LBRC relative to total membranes and the medium density fractions.

Figure 1. Isolation of the LBRC using subcellular fractionation.

A. Ten dishes (10 cm) of confluent HUVEC were labeled with anti-PECAM antibody (hec7) for 1 hr at 4°C to label surface and junctional PECAM. After extensive washing, cells were collected and homogenized. The resulting homogenate was adjusted to 42% sucrose and placed at the bottom of a 10–40% continuous sucrose gradient. After 18 hr centrifugation at 133,000 × g, fractions were collected and the proteins in each precipitated using TCA, resolved using SDS-PAGE, and analyzed using western blotting. Transferrin Receptor (TfR), Lysosomal Associated Membrane Protein-2 (LAMP2), and Vascular Endothelial Cadherin (VE-Cadherin) were used as markers of endosomes/plasma membrane, lysosomes, and junctions, respectively. Asterisk denotes fractions enriched in LBRC membranes, as determined by the lack of other intracellular markers and the presence of PECAM that was not labeled with antibody (IgG HC) at 4°C. Images shown are representative of more than 5 separate experiments. B. Total membranes (TM) were collected from cell homogenate by centrifugation at 200k × g for 45 min. To analyze the enrichment of other proteins in the LBRC, TM and LBRC fractions were analyzed side-by side. To make comparisons relative to PECAM enrichment, samples amounts were adjusted so that PECAM signal was equal.

The low density fractions containing PECAM but not VE-cadherin or IgG heavy chain (typically 1–3 fractions on the less dense side of those containing plasma membrane) were pooled and analyzed for the presence of several other proteins (Figure 1B). To examine the enrichment of proteins in the LBRC-containing fractions, we compared their relative abundance to a crude membrane preparation that represents total cellular membranes (Total Membranes, TM). In order to analyze the presence of other proteins, we adjusted the amounts of the total membrane and LBRC fraction so that the PECAM signal was the same intensity. This approach allows us to determine if a protein is enriched in the LBRC at least to the extent that PECAM is enriched in the compartment. As seen in Figure 1B, VE-cadherin and the IgG used to label the surface pool of PECAM are essentially absent from LBRC enriched fractions, further confirming the purity of the fractionation method.

With this comparison, we expected that proteins that are excluded from the compartment would be absent in the LBRC fractions, similar to VE-cadherin. On the other hand, a protein that is unique to the LBRC would have a significant signal in the LBRC fraction, possibly more than the signal in the total membrane fraction. Using this method we were able to determine that several other junctional proteins, namely plakoglobin and β-catenin, were not present in the LBRC. We were also able to confirm that kinesin, which is involved in the directed movement of the LBRC during TEM, is at least partially recovered in the LBRC fractions. This relatively simple flotation method is therefore suitable for testing candidate proteins for which antibodies are available without resorting to immuno electron microscopy. Unfortunately though, in testing several different candidate LBRC proteins, we did not identify any that were significantly abundant in the LBRC like PECAM.

Furthermore our attempts to purify a significant amount of this LBRC-enriched fraction proved unsuccessful. Previous findings showed that roughly 30% of total PECAM is localized to the LBRC (6). However, a quantitative comparison of the amount of PECAM in the LBRC fractions to an equivalent amount of total membrane showed that the recovery of LBRC PECAM was lower than expected, around 10% (data not shown). We suspect that the LBRC membranes distributed in a manner that partially overlapped with the distribution of plasma membrane in the flotation gradient. Attempts to modify the methodology (e.g. refining the sucrose density in the relevant range) to obtain better yield or resolution of the two pools of PECAM were unsuccessful.

Subcellular fractionation – DAB-induced density shift

Confident that we could differentially label the two pools of PECAM (i.e. labeling at 4°C only labels the surface pool), we next sought to determine a way to distinguish one pool from the other. In investigating various methods for fractionating cellular components, we came across rarely used method referred to as a diaminobenzidine (DAB)-induced density shift. This method involves loading vesicles with horseradish peroxidase (HRP). After initial purification, the density of the labeled vesicles can be changed biochemically by incubating the sample with the membrane permeant reagents DAB and H₂O₂ for 30 min. HRP catalyzes the oxidation of DAB which then polymerizes into a dense brown precipitate. This reaction is readily used in immunohistochemistry and electron microscopy. Its utility in subcellular fractionation, though first reported in (10), has apparently been forgotten beyond a few now historic reports (11–14). The DAB-density shift relies on the observation that HRP in the vesicle lumen will catalyze the DAB polymerization there, leading to an accumulation of the dense DAB polymers and a corresponding dramatic shift in net vesicle density.

Although we could not directly label the LBRC exclusively, we reasoned that we could produce two samples, one pre-labeled at 4°C where only the surface pool of PECAM is labeled, and another sample pre-labeled at 37°C that would contain both the LBRC and surface pools of PECAM. By isolating the membranes from the pre-labeled cells and subjecting them to the DAB density shift, we could then recover two samples, one containing only plasma membrane and the other containing both the plasma membrane and the LBRC (Figure 2). Simple comparison of the banding pattern of two samples would then reveal the proteins uniquely enriched in the LBRC.

Figure 2. Schematic of the DAB-induced density shift method.

A. Schematic detailing the differential labeling of the plasma membrane (PM) and LBRC with HRP-conjugated antibodies against PECAM. At 4°C, macromolecular reagents, like antibodies, can access the PM and intercellular junction (J) but are inaccessible to the LBRC. However, at 37°C these reagents efficiently label molecules that reside in the LBRC. B. Schematic of the gradient purification method. Cleared homogenate was loaded at the bottom of a 10/32% step gradient. After centrifugation, crude membranes (including the LBRC and plasma membrane) are recovered from the 10/32% interface. This initial purification step removes all heavier membranes and large cellular debris that migrates at the higher density fractions where the DAB density shifted fractions will migrate. Recovered membranes are incubated with DAB and H₂O₂ which causes all vesicles containing the HRP-conjugated antibody to accumulate dense DAB polymers (dark filled vesicles). During centrifugation, these previously low-density vesicles now migrate to the high density fractions.

Because any vesicle that contains HRP will undergo the density shift, we wanted to make sure that the labeling method did not alter PECAM distribution. We were particularly concerned about co-purifying endosomes, since protein ligation by antibody has been shown to induce the internalization of the protein in other systems. This process has been reported to be further enhanced when primary antibodies are cross-linked with secondary antibodies. Therefore, we conjugated anti-PECAM directly to HRP to avoid the need for secondary antibodies and further cross-linking (Figure 3A). Immunofluorescence microscopy was used to determine if this modified antibody faithfully recapitulated PECAM staining. To examine this, confluent endothelial cell monolayers were incubated with unconjugated anti-PECAM and the conjugated HRP-antibody at 37°C for 30 min. Monolayers were then fixed and permeablized and the labeling of the indicated primary was visualized using immunofluorescence (Figure 3B). Although the intensity of the signal was slightly higher in the sample that was stained after fixation and permeablization, no difference in staining pattern was observed between the monolayers pre-labeled with anti-PECAM or anti-PECAM-HRP and the monolayer labeled after fixation. In all three samples the majority of PECAM was detected at the junction, with a minor amount observed on the apical plasma membrane. None of the samples showed significant staining in the form of intracellular punctae that are characteristic of endocytic vesicles. This suggests that this labeling protocol does not induce internalization of PECAM and thus co-purification of endocytic vesicles should not be an issue with the DAB-density shift.

Figure 3. Production and testing of anti-PECAM-HRP antibody.

A. Anti-PECAM antibody (hec7, left lane) was conjugated to HRP as described in the materials and methods. Analysis of the conjugation using non-reducing SDS-PAGE and Coomassie staining showed that most of the antibody had been conjugated to one or two HRP molecules (right lane, annotated on the side as +1 and +2). B. The uptake and internalization of anti-PECAM (left panel) and anti-PECAM-HRP (middle panel) antibodies were examined after incubation for 30 min at 37°C. After the incubation, monolayers were fixed, permeablized and stained with fluorescent secondary antibodies to check for internalization. The staining pattern in monolayers that were pre-incubated the antibodies was indistinguishable from that observed in monolayers stained after fixation and permeablization (right panel) suggesting that pre-incubation with the anti-PECAM-HRP did not induce PECAM internalization. Scale bar = 25 μm.

Having confirmed that the anti-PECAM-HRP labeling did not alter PECAM distribution, we next sought to isolate the membranes labeled by the antibody using the DAB density shift. Toward this end, we labeled HUVEC monolayers with either anti-PECAM or anti-PECAM-HRP at 4°C or 37°C as indicated. Monolayers were collected and homogenized. The homogenization breaks apart cellular membranes that then, with their neighboring proteins and lipids, reform into vesicles, the thermodynamically favorable form of lipids in aqueous solution. Although we could not control the topology of the proteins in the vesicle (i.e. they could form with their extracellular leaflet facing the lumen of the vesicle, trapping the HRP-antibody in the lumen, or vice versa), we reasoned that most of the LBRC membranes would form with the HRP inside, since the organelle is naturally in a semi-vesicular state with this orientation. Membranes from the PM could presumably form either way, although we reasoned that enough would form with the HRP on the inside to achieve sufficient recovery of the membrane.

To remove the high density membranes that would run with the density-shifted vesicles, the homogenized samples were first purified on a 10%/32% step gradient. Crude membranes were recovered from the 10/32% interface and incubated with diaminobenzidine and H₂O₂ for 30 min. During this incubation, vesicles containing HRP accumulate dense DAB polymers. After incubation, the samples were loaded on a 20–50% continuous sucrose gradient and centrifuged at an RCF_ave of 130,000 × g overnight. Fractions were collected and their associated proteins recovered by TCA precipitation. Western blot analysis of samples labeled with unconjugated anti-PECAM showed that the PECAM-containing membranes co-migrated in the medium density fractions with LAMP2, the vacuole/late endosome marker (Figure 4A). However, analysis of samples from monolayers labeled with HRP-conjugated anti-PECAM antibody showed that PECAM-containing membranes had migrated to much higher density fractions, away from medium density fractions containing the LAMP2 marker.

Figure 4. Isolation of LBRC membranes using DAB-induced density shift.

A. Confluent HUVEC (10 × 10 cm dishes) were labeled with anti-PECAM antibodies (Hec7) or anti-PECAM antibodies conjugated to horseradish peroxidase (HRP) at 4°C or 37°C as indicated. Monolayers were then washed and the cells collected, homogenized and clear by low speed centrifugation. This cleared homogenate was first fractionated on a 10/32% sucrose step gradient. After centrifugation, crude membranes were recovered at the 10/32% interface and incubated with DAB and H₂O₂ for 30 min at room temperature. The sample was then loaded on top of a continuous 20–50% sucrose gradient and centrifuged overnight. Fractions were collected and their protein content recovered by TCA precipitation, resolved using SDS-PAGE, and visualized for the indicated protein with western blotting. IgG HC - heavy chain from the initial labeling. Fractions with density shifted membranes have been overlined. The western blots shown are representative of more than 15 separate experiments. B. Fractions containing density-shifted membranes (denoted by the line above them in A) were pooled and their proteins recovered by TCA precipitation and resolved using SDS-PAGE. The indicated proteins were visualized using western blotting. The relative amount between the two density-shifted samples was not adjusted. However, the relative amounts of the “cleared homogenate” and “crude membrane” samples were adjusted so that VE-cadherin and transferrin receptor (TfR) would all be similar and approximately equal to the density-shifted samples, thus allowing for determination of relative enrichments between all the samples.

To confirm the utility of this method we pooled the density-shifted fractions from the two samples and analyzed them for a variety of different markers (Figure 4B). In addition to examining the enrichment between these two samples, we wanted to examine the enrichment relative to the starting material (cleared homogenate) and an intermediate step in the method (crude membranes). To determine the relative enrichment, we adjusted the amounts of the “cleared homogenate” and “crude membrane” samples so that the amount of VE-cadherin and TfR would be similar between these samples and the density-shifted samples. The relative amount between the two density-shifted samples was not adjusted. As seen in Figure 4B, PECAM was slightly enriched in the sample collected from cells labeled at 37°C, possibly because of the additional labeling of PECAM from the LBRC at this temperature. The amounts of VE-cadherin and (TfR) detected in the two density-shifted samples were nearly identical. This makes sense, since neither TfR (6) nor VE-cadherin (5) were previously found to co-localize with LBRC, so these amounts represent the contribution of the plasma membrane to this fraction.

Analysis of other markers shows that the density-shifted fractions are largely devoid of endocytic (Rab5 and EEA1) and lysosomal (LAMP2) markers, compared to earlier steps in the process. This further confirms that the HRP conjugated anti-PECAM antibody did not induce endocytosis of PECAM on a meaningful scale. Caveolin-1 (Cav-1) was also not detected in the density-shifted material, further confirming that the LBRC is distinct from caveolae. We also examined the samples for the presence of kinesin, which is known to be involved in LBRC function during TEM. Interestingly, we could only detect kinesin in the cleared homogenate and not in the crude membranes or either of the density-shifted samples. One possible explanation for this is that a relatively small fraction of the total cellular kinesin interacts with membrane. Alternately, kinesin could be dissociating from its target membranes during the purification. Furthermore, kinesin was not detected in either of the density-shifted fractions, even with larger amounts of sample and longer exposure times.

In determining the amount of material needed to obtain the similar VE-cadherin and TfR signals, we were also able to roughly determine the amount of recovery during the purification method, as least for PECAM. To summarize, our analysis of the initial homogenization step showed roughly 70% of PECAM was recovered after clearing with low speed centrifugation (data not shown). Likewise, we found that this “cleared homogenate” had roughly twice the relative amount of PECAM as the “crude membranes,” which in turn had roughly five times the amount as the 4°C sample. Based on the western blots of the DAB density shift gradients, we only pooled the fractions in which PECAM had substantially shifted away from LAMP2 (recovering 30–50% of the total PECAM, Figure 4A), suggesting that the recovery of PECAM from the gradient using TCA is roughly 40–60% efficient. Thus, all together, we estimate a total recovery of 5–10% PECAM in the 4°C sample, with slightly more in the 37°C sample (70% recovery after homogenization; 50% recovery after isolating crude membranes; 30% shifted to higher density fractions; 60% recovery after precipitation).

Based on the temperature-sensitive differential labeling method described above and elsewhere, we reasoned that the samples from monolayers labeled at 37°C would be significantly enriched in LBRC membranes compared to samples from monolayers labeled at 4°C. To collect enough of each sample to perform proteomic analysis, we confirmed these findings in immortalized HUVEC (iHUVEC) and scaled up the method accordingly. We repeated density shift 10 times with ~10× 10 cm dishes for each sample. Density shifted fractions from each trial were identified, collected, and pooled. The corresponding proteins recovered by TCA precipitation, were resolved using SDS-PAGE and visualized by silver staining (Figure 5A). We adjusted the amount of each sample so that the majority of the bands common to both samples would appear at equivalent intensity. Samples that had been labeled at 4°C and 37°C both presumably contain roughly equivalent amounts of the diverse array of plasma membrane proteins. This was confirmed by western blotting for VE-cadherin and TfR, which were detected at nearly identical amounts between the samples (data not shown). After doing so, several bands were apparent that were significantly enriched in the samples labeled at 37°C. We excised the putative LBRC-associated proteins that ran at ~55, 110, and 200 KDa. Using mass spectrometry, these proteins were identified as the intermediate filament protein vimentin, Major Vault Protein (MVP), and the cytosolic scaffold protein IQ-motif GTPase activating protein 1 (IQGAP1), respectively (Figure 5B). The recovery of vimentin and IQGAP1 as LBRC-associated proteins fits well with recent reports that suggest their involvement in TEM (15–17). The biological role of MVP is currently unknown. Because of this, we decided to leave the characterization of MVP in TEM to other studies.

Figure 5. Identification of enriched proteins from the density shifted fractions.

A. Fractions containing density shifted membranes (as in Figure 4A) were pooled from 12 experiments using iHUVEC. Proteins contained in these fractions were precipitated, resolved using SDS-PAGE, and visualized using silver staining. The relative amounts of the density-shifted material were adjusted slightly so that enriched proteins could be detected more easily. At this relative amount, the majority of bands (presumably from plasma membrane proteins) were roughly equal (denoted by the lines between the lanes). Also, at these relative proportions, the amount of VE-cadherin, detected using western blot, was equivalent (data not shown). Red arrows indicate the putative LBRC-enriched proteins that were excised and sequenced using mass spectrometry. B. Fragments of IQGAP1, major vault protein, vimentin identified by mass spectrometry. Trypsin cleavage site is indicated by a period. Putative flanking residues, as determined by the peptide identification programs, are shown. C. Western blot of IQGAP1 and vimentin in the density shifted fractions as compared to the cleared homogenate and crude membranes. Samples were in the same relative amounts as in Figure 4B.

To confirm that the IQGAP1 and vimentin were enriched in the LBRC-containing, we probed the pooled fractions from the 4°C or 37°C labelings using western blotting. Analysis of these two samples from the representative density shift gradient shown in Figure 4B using the same relative amounts of the samples showed that both IQGAP1 and vimentin are indeed enriched in the sample containing LBRC (Figure 5C). Interestingly, only a relatively small amount of vimentin was detected in the 37°C density-shifted fractions. This could be because only a small fraction of total cellular vimentin is associated with the LBRC. Nevertheless, vimentin was enriched in the 37°C sample compared to the 4°C sample, suggesting a specific association with the LBRC compared to the rest of the plasma membrane.

We wanted to examine the localization of IQGAP1 and vimentin using immunofluorescence. We reasoned that proteins that were recovered based on their association with LBRC should co-localize with it, at least partially. Again this was complicated by the lack of a suitable marker for the LBRC. However, immuno electron microscopy findings suggest that the LBRC is observed primarily at the endothelial cell borders. Also, using immunofluorescence, every confirmed LBRC protein is observed at cell-cell contacts, at least partially. We examined the subcellular localization of IQGAP1 and vimentin using immunofluorescence microscopy (Figure 6). Consistent with previous reports (18, 19), IQGAP1 exhibited both diffuse cytoplasmic staining and significant localization the junction (identified using PECAM). As expected, the staining pattern for vimentin was predominately filamentous. However, there were regions in most cells where vimentin filaments were observed in close apposition to the cell border where an interaction with the LBRC could take place (Figure 6, inset).

Figure 6. Immunofluorescence of IQGAP1 and Vimentin in endothelial cells.

Confluent HUVEC fixed, permeablized, and stained for PECAM (to identify LBRC and cell borders) and either IQGAP1 or Vimentin. Inset shows region of interest (denoted by dashed box) where the localization of IQGAP1 or vimentin to the junction is particularly pronounced. The contrast has been adjusted in the vimentin inset to make individual filaments more visible. Dashed yellow line in inset denotes putative cell borders as determined by PECAM staining. The scale is the same for all four panels (bar = 10 μm).

Discussion

Here we detail an approach for isolating and characterizing the LBRC that is broadly applicable to a variety of intracellular organelles and experimental designs. Until these methods were developed, the only way to determine if a protein was localized to the LBRC was to perform electron microscopy. Although electron microscopy does clearly delineate localization to the LBRC, its use is costly, inefficient, and limited to only those proteins that can be labeled from the outside of the cell and for which an antibody has been developed against their extracellular domain. As such, this method is only suitable for a candidate-based approach and not large scale proteomic studies. Because of this, we sought to develop a method that could be used to purify significant amounts of the compartment for further analysis and the identification of additional LBRC components.

Our initial approach was based on the observation that membranes of the LBRC have a density that is relatively lower than that of most other intracellular membranes including the plasma membrane (Figure 1). The identification of LBRC membranes took advantage of the previous observation that proteins in the LBRC (e.g., PECAM) are inaccessible to antibodies at 4°C (6). Thus, when examined after fractionation on a sucrose gradient, the LBRC-containing fractions were identified by the presence of PECAM and the absence of the labeling antibody. The LBRC-containing fractions were relatively devoid of other intracellular membranes, namely, plasma membrane, endosomes, lysosomes, and endoplasmic reticulum. Analysis of the pooled fractions showed that they contained LBRC PECAM but not surface PECAM (determined by the absence of the labeling antibody).

To determine whether other proteins were present in the LBRC, we normalized the sample amounts for western blotting so that the PECAM signal would be equivalent. We could not normalize to total protein or lipid because we did not recover enough of the LBRC sample to get accurate measurements on these values. Also, without knowing more about the compartment, normalizing to protein or lipid amounts could be misleading (i.e. the protein or lipid content of the LBRC could be dramatically different than that of other membrane compartments). By comparing the abundance of a candidate protein to that of PECAM, we could at least determine if the proteins were present or enriched in the LBRC to the same extent as PECAM (30% of the total PECAM is in the LBRC (6)). Thus, this sucrose fractionation method is a big improvement over immuno electron microscopy, since allows us to perform a candidate based approach for more than one protein at a time and allows us to probe for candidate molecules that are not accessible from the exterior of the cell. Although this method allowed us to determine that several of the proteins involved in the formation of adherens junctions were not detected in the LBRC (5), we unfortunately did not identify any new LBRC proteins. However, we were able to detect kinesin in the LBRC fractions, albeit at a reduced level compared to PECAM. This finding is not unexpected as previous work has shown that kinesin is required for the directed movement of the LBRC during TEM (4). That kinesin is not as enriched as PECAM could either be because it is not required to be as abundant in order to perform its function during TEM or because only a small fraction of the membrane-associated kinesin is interacting with the LBRC.

It is interesting that the LBRC is observed to be less dense than the plasma membrane considering that analysis of the compartment using electron microscopy shows that the LBRC is contiguous with the plasma membrane (6, 7). Because the density of a membrane is determined by its unique protein and lipid content (20), this finding suggests that protein and/or lipid flux between the LBRC and plasma membrane could be gated or regulated. It is currently unclear how this could be accomplished, although we speculate that the high degree of membrane curvature observed in the LBRC could either be the cause or the result of selective protein and lipid enrichment as has been reported in other systems (21–23).

Cholesterol-rich membranes are often reported to have relatively low density and flotation on density gradients is often used to purify such membranes, like endosomes and lipid microdomains or rafts (24–27). The low density of the LBRC membranes could indicate that they are enriched in cholesterol and comprised of lipid microdomains or rafts. Unfortunately, we did not recover enough material using this method to accurately determine the cholesterol content in the LBRC fractions. However, previous work has shown that the LBRC does not contain caveolin-1, a common marker and structural component of endosomes and plasma membrane rafts (26). Thus, even if the LBRC is rich in cholesterol, it is not likely to merely be a subset of these other well characterized domains.

Because the amount of material recovered from this simple flotation gradient was insufficient for proteomic analysis, we modified a classic method, the DAB-induced density shift, to increase our yield. This method was pioneered almost 30 years ago and was initially used to purify endocytic vesicles that contained HRP-conjugated ligands taken up from the bloodstream (10–12). The DAB density shift was subsequently used to examine the endocytic sorting of various receptors, but, despite its apparent utility, largely disappeared from the repertoire of common molecular biology techniques 15 years ago (14, 28, 29). Here we detail an updated variation of this method that uses an antibody conjugated to HRP, purifies a non-endocytic compartment, and is suitable for proteomic analysis. The methods here could be used to isolate any compartment that can be labeled with HRP. By extension, it could presumably be used to purify membranes that have been labeled genetically, i.e. expressing a HRP fused to a marker. One theoretical constraint being that the HRP must be in the lumen of the resulting vesicles in order to trap the polymerized DAB inside. It is important to point out that one of the main benefits of the DAB density shift is that the researcher can control when in the purification method the density shift is induced. For example, in the method here, the lighter crude membranes (which include vesicles from the LBRC) were first collected from the 10/32% interface of a sucrose step gradient. Only after effectively removing all higher density membranes was the DAB/H₂O₂ reaction performed. As a result, only the HRP-labeled density shifted membranes now migrated at the higher density, allowing us to recover a relatively simplified protein profile and identify those proteins that were enriched in the sample containing the LBRC. Simply stated, we first floated LBRC membranes away from contaminants of higher density, and then used the DAB density shift to separate LBRC vesicles from contaminants of equal or lower density.

Because there is no specific marker of the LBRC, we developed this method to purify it from total cellular membranes. This method exploited our ability to differentially label the LBRC at 4°C and 37°C. This allowed us to prepare two samples that were identical except that the one from cells labeled at 37°C also contained the LBRC. Analysis of the density-shifted samples showed that this method effectively purifies the labeled membranes from other cellular membranes including endosomes and lysosomes (Figure 4B). With these samples in hand, we could then examine the enrichment of putative LBRC proteins. We found PECAM slightly enriched in the sample containing the LBRC. Interestingly, kinesin, which is required for LBRC function during TEM, was not detected in the LBRC-enriched fraction, or any membrane fraction for that matter. It is possible that amount of kinesin interacting with the LBRC at steady-state is vanishingly small or that kinesin interaction with the LBRC is dependent on signals that occur during TEM. Alternately, kinesin interaction with the LBRC and/or membranes in general might not be stable enough to survive the long purification method.

To identify novel LBRC components, we compared the banding patterns of these two samples and identified several proteins that were largely enriched in the sample containing the LBRC labeled at 37°C. Using mass spectrometry, these putative LBRC proteins were identified as IQGAP1, vimentin, and MVP. All of these proteins are localized to the cytoplasm, and because of this could not have been identified with the immuno electron microscopy labeling approach discussed earlier. Although neither has been examined for association with the LBRC, the knockdown of either IQGAP1 or vimentin in endothelial cells causes a significant reduction in TEM (15–17), supporting the use of the DAB density shift to purify relevant proteins in TEM. Analysis of the purified samples using western blotting confirmed that both IQGAP1 and vimentin were enriched in the 37°C sample (Figure 5). Furthermore, both IQGAP1 and vimentin were detected, at least partially, at the cell periphery where they would be appropriately positioned to interact with the LBRC (Figure 6). Future experiments will examine the interaction of these molecules with the LBRC using post-embedding immunogold electron microscopy.

Materials and Methods

All procedures involving human subjects and human materials were approved by the Institutional Review Board of Northwestern University Feinberg School of Medicine and performed in accordance with the Helsinki Declaration with the appropriate informed consent.

Antibodies

This study utilized mouse IgG_2a anti-human PECAM clone hec7 (30), and mouse IgG_2a anti-human VE-cadherin clone hec1 (31), all produced in the laboratory via hybridoma methodologies. Mouse anti-human β-catenin was obtained from BD Biosciences (San Jose, California). Rabbit anti-human calreticulin and rabbit anti-EEA1 were obtained from EMD Millipore (Billerica, Massachusetts). Chicken anti-human plakoglobin was a generous gift from Dr. Kathleen Green (32). Chicken anti-human vimentin for use in immunofluorescence microscopy was obtained from Covance (Princeton, New Jersey) while mouse anti-human vimentin clone v9 for use with western blotting and rabbit anti-human caveolin-1 were obtained from Abcam (Cambridge, Massachusetts). Rabbit anti-human Kinesin 5B was purchased from Sigma-Aldrich (St. Louis, MO). Rabbit anti-human Rab5A (S-19), rabbit anti-human IQGAP1, mouse anti-human LAMP2 and mouse anti-human transferrin receptor were obtained from Santa Cruz Biotechnology (Santa Cruz, California). Goat-anti-mouse and goat-anti-rabbit conjugated to HRP were purchased from BioRad Laboratories. Secondary antibodies (Alexa Fluor 488 or Rhodamine Red-X conjugated to either goat anti-mouse, goat anti-rabbit, or goat anti-chicken) were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA).

Isolation and culture of Endothelial Cells

Human Umbilical Vein Endothelial Cells (HUVEC) were isolated from human umbilical cords as previously described (30, 33). At their second passage, isolated HUVEC were either grown to confluence (3–5 days) on coverslip dishes (Mattek, Ashland, MA) for immunofluorescence or 10 cm tissue culture plates for gradient preparations. To collect enough material for large scale fractionation, primary HUVEC were transduced according the method of Moses et al. (34) to generate immortalized HUVEC (iHUVEC) which were then expanded and used at a maximum passage of 25. The results of all experiments performed iHUVEC were repeated and validated with HUVEC where possible.

Horseradish peroxidase (HRP) conjugation

Anti-PECAM antibody (clone hec7) was conjugated with HRP using the EZ-Link Plus Activated Peroxidase Kit (Thermo Scientific Pierce) according to the manufacture’s specifications. Conjugation efficiency was determined by analyzing the resulting sample with SDS-PAGE and Coomassie blue staining. Using this method, the majority of antibody was conjugated to one or two HRP molecules. Unconjugated HRP and residual conjugation buffer was removed and replaced with PBS using an Amicon Ultra Centrifugal Filter (Millipore) with a 100 KDa molecular weight cutoff. The conjugated antibody was stored at −20°C until use.

Immunofluoresence microscopy

Confluent HUVEC were fixed in 2% paraformaldehyde for 10 min, washed three times with PBS, quenched with 75 mM/L glycine for 15 min, and washed three more times with PBS. Blocking buffer (PBS + 5% bovine serum albumin, Sigma-Aldrich) was added for 30 min at room temperature (RT). Cells were then incubated with primary antibody at 10 μg/mL in blocking buffer for 1 hr at RT, washed extensively with PBS, and then incubated with appropriate secondary at 4 μg/mL in blocking buffer for 1 hr at RT protected from light. For the visualization of internalized anti-PECAM-HRP (clone hec7), cells were incubated with 10 μg/mL of the antibody in fresh media for 30 min at 4°C or 37°C before fixation. Monolayers were then washed and fixed as described above and permeabilized for 4 min with 0.1% Triton X-100 (Sigma-Aldrich) before blocking. Samples were imaged using a restoration workstation (Delta Vision 3D; Applied Precision) equipped with an inverted microscope (model IX70; Olympus) using a 60× oil objective.

Subcellular fraction – flotation gradients

Two to three 10 cm dishes of confluent HUVEC or iHUVEC were chilled on ice and washed three times with ice-cold PBS. All subsequent steps were performed on ice with chilled buffers and equipment. Monolayers were then treated with 20 μg/ml anti-PECAM antibody (clone hec7) for 1 hour on ice to label the surface and junctional pools of PECAM. Monolayers were then rinsed three times with PBS and scraped into a combined volume of 1 ml of homogenization buffer: 6 mmol/L imidazole pH 6, 250 mmol/L sucrose, 50 mmol/L NaCl and 1× protease inhibitor cocktail (Complete Mini, Roche Diagnostics). The sample was then homogenized using a cell homogenizer (Isobiotec, Heidelberg, Germany) by 20 passes with a 10 μm-clearance bearing. The crude homogenate was centrifuged at 18,000 × g for 15 min at 4°C. 400 μl of the cleared lysate was mixed with 1.2 ml of 60% sucrose (weight/volume, w/v) and loaded into a thin-wall Ultra-Clear 13×51mm ultracentrifuge tube (Beckman Coulter). The sample was overlaid with a continuous 10–40% (w/v) sucrose gradient buffered with 6 mmol/L imidazole pH 6 and 50 mmol/L NaCl. The flotation gradient was centrifuged in a SW55 Ti rotor at 37,500 rpm (RCF_ave 133,000 × g) for 16 hours at 4°C. 250 μl fractions were collected from the bottom. Proteins were recovered by trichloroacetic acid (TCA) precipitation.

Subcellular fraction – DAB-induced density shift

Cell homogenates for the DAB-induced density shift were homogenized and cleared as described above except that 8–12 monolayers per condition were used instead of 2–3 and samples were homogenized in three batches to minimize clogging of the homogenizer. Additionally, monolayers were incubated with 10 μg/ml of the indicated antibody at either 4°C for 1 hr or 37°C for 30 min. The cleared lysate was loaded on top of a 10%/32% sucrose step gradient (1 ml of each step) in a thin-wall Ultra-Clear 13×51mm ultracentrifuge tube and centrifuged for 2 hr at 37,500 rpm in an SW55 Ti rotor. Crude membranes (~300 μl) were recovered from the 10%/32% interface. 250 μl of crude membrane was mixed with 1.25 ml of 1.2 mg/ml 3,3′-diamiobenzidine tetrahydrochloride (DAB, Sigma) in 6 mmol/L imidazole pH 6, 50 mmol/L NaCl. The DAB solution was passed through a 0.2 μm syringe filter before mixing with the crude membranes. 5 μl of 5% hydrogen peroxide (from a 30% w/v stock, Sigma) was added to the DAB/membrane mixture and rotated end-over-end for 30 min at room temperature in the dark. 1.25 ml of the sample was then loaded onto a 3.75 ml 20–50% continuous sucrose gradient in a thin-wall Ultra-Clear 13×51mm ultracentrifuge tube and centrifuged for 18 hr at 37,500 rpm in an SW55 Ti rotor. 250 μl fractions were collected from the bottom. Proteins were recovered by TCA precipitation.

Trichloroacetic acid (TCA) precipitation

Proteins from the samples recovered from the sucrose gradients were recovered by precipitation using TCA. For simple analysis of the distribution of proteins in the gradients 25–50 μl of sample was precipitated in a final volume of 1 ml of 10% (w/v) TCA (Sigma) in a microfuge tube. 10 μg of cytochrome c (Sigma) was added as a carrier. Samples were incubated on ice for 30 min and centrifuged for 15 min at 18,000 × g at 4°C. The supernatant was replaced with cold acetone and the spin repeated. Protein pellets were then air-dried and resupended in 50 μl of 1% SDS and 10 mmol/L Tris pH 8. To recover proteins for large scale mass spectrometric analysis, the procedure was scaled up accordingly but the cytochrome C carrier was omitted and the final pellet was resuspended in 250 μl.

Western blotting

Aliquots of the samples were mixed with Laemmli loading buffer with β-mercaptoethanol and incubated at 60°C for 30 min. Equivalent amounts loaded onto a 10% polyacrylamide gel and resolved using SDS-PAGE. Proteins were transferred to PVDF and detected using standard western blotting techniques. Blots were probed for the indicated antigen. To examine multiple proteins in 1 set of samples, the membrane was either stripped and reprobed using standard methods or an identical blot was prepared.

Analysis of samples by silver stain

Fractions from the DAB-density shift gradients that contained density shifted vesicles as identify by western blot analysis were pooled and their protein component recovered using TCA precipitation as described above. A small aliquot of the pooled samples was removed and mixed with Laemmli loading buffer with β-mercaptoethanol, incubated at 60°C for 30 min, and resolved on a 10% polyacrylamide gel using SDS-PAGE. Bands were visualized using a Silver Stain Kit for Mass Spectrometry (Thermo Fisher Scientific). This procedure was then repeated with various amounts of each sample to identify the appropriate volumes in which the majority of common bands would be the same intensity. The gel was then re-run using these volumes to directly compare the two samples. This adjustment allowed for identification of the bands that were enriched in the LBRC-containing sample.

Sample preparation and mass spectrometry analysis

Bands that were enriched in the LBRC sample were excised and destained according the instructions accompanying the silver stain kit. Subsequent sample preparation and mass spectrometric analysis was performed by the Chicago Biomedical Consortium/University of Illinois at Chicago Research Resources Center Proteomics and Informatics Services Facility. The in-gel tryptic digestion was carried out by following the protocol described by Kinter and Sherman (35). Briefly, the gel bands were cut into 1 mm³ pieces, rinsed, and dehydrated, and the protein was reduced with DTT and alkylated with iodoacetamide in the dark. Samples were then digested with trypsin at 37 °C in 50 mM ammonium bicarbonate. Each sample was injected onto a reversed phase column (75 micron id 150 mm Zorbax SB300 C-18, Agilent Technologies, CA) connected to a Dionex Ultimate 3000 HPLC system and a Thermo Finnigan LTQ-FT Ultra mass spectrometer equipped with an nanospray interface. The samples were chromatographed using a binary solvent system consisting of A: 0.1% formic acid and 5% acetonitrile and B: 0.1% formic acid and 95% acetonitrile at a flow rate of 200 nl/min. A gradient was run from 15% B to 55% B over 60 minutes. The mass spectrometer was operated in positive ion mode with the trap set to data dependent MS/MS acquisition mode. The instrument is set to complete a mass scan from 400–1800 daltons in one second. Peaks eluting from the LC column that have ions above 1,000 arbitrary intensity units trigger the ion trap to isolate the ion and perform an MS/MS experiment scan after the MS full scan. The instrument’s dynamic ion exclusion filter was set to allow the instrument to record 2 MS/MS spectra for each detected ion to optimize the acquisition of quantitative data in MS mode as well as qualitative data in MS/MS mode and in addition after every MS scans in the ion trap, ions were then transferred to the isolation cell of the 7T magnet in the FT where a high resolution (up to 500,000 at m/z 400) and high mass accuracy (1 ppm) mass spectrum of the precursor ions was obtained.

Data reduction of Tandem MS Spectra

Data analysis was carried out using the Mascot software platform (Matrix Science, UK) Bioworks 3.3 (Thermo Instruments, CA), Sorcerer 2 (Sage-N Research Systems, CA) and the X!Tandem software platform. Raw data was extracted from the MS data files using the data extractor module in the Bioworks software package and then subjected to protein sequence database search by Mascot, Sequest (using the Sorcerer server) and by the X!Tandem software package. The library searching identified the detected proteins from the individual peptides by comparing lists of ion masses obtained in the analysis to lists of ions computer generated from the hypothetical tryptic peptides that could be produced from all the proteins in a given organism’s proteome. The results for all proteins detected on all the servers were collected and collated using the Scaffold 2 package (Proteome Systems, OR).

Synopsis.

The Lateral Border Recycling Compartment (LBRC) is a novel endothelial cell compartment that appears as a multi-vesicular invagination of the plasma membrane. The LBRC is critical for the movement of leukocytes across the endothelium. Here we detail the use of a diaminobenzidine-induced density shift for the purification of the LBRC for proteomic analysis. Vimentin and IQGAP1 were identified from LBRC-enriched fractions.