As the basic unit of a lumenal membrane staple we designed a chimeric protein, termed CD8_lumenal, that contained a signal sequence, for its proper insertion and translocation across the ER membrane, fused to 1) a fluorescent protein (GFP or DsRed) to track the protein within the cells by confocal microscopy, 2) four repeats of a domain (FM4) designed to self-aggregate in trans in the absence of the disaggregating drug (AP21998) (Rollins et al., 2000), and 3) the CD8 protein, known to be a trans-membrane protein ultimately targeted to the plasma membrane (Hennecke and Cosson, 1993; Lavieu et al., 2010). This CD8_lumenal protein should initially reside in the ER oriented with its FM4-containing aggregation domain facing the lumen, and so long as it remains non-aggregated (in the presence of AP21998) should be effectively transported to the cell surface. On the other hand, if AP21998 is removed while in the ER—or at any later stage of transport through the Golgi—then intra-lumenal adhesions are expected to form as CD8_lumenal aggregates homo-typically in trans, like a cadherin. These staples are expected to remain in the ER if they are planar and too large to be packaged into a small COPII vesicle. Various structural models suggest a planar adhesion plaque will form that closely links two sides of ER lamellae or tubules, with an intra-lumenal separation between the stapled membranes that from first principles could vary from 10 to 23 nm (Figure 1A), depending on the preferred arrangement of trans-aggregating FM4 domains. In particular, however, if repeat FM domains interact like cadherins (Miyaguchi, 2000; Pokutta and Weis, 2007) primarily via their most-membrane distant module (i.e., at their tips), then we would expect a membrane separation in the stapled state close to 25 nm (model 1 in Figure 1A).

Figure 1 with 1 supplement see all

Download asset Open asset

When CD8_lumenal is expressed in HeLa cells in the absence of the disaggregating drug AP21998, numerous, discrete electron-dense plaques are in fact readily observed by electron microscopy (Figure 1B, red square). The staples trigger lumenal constrictions within ER tubules (Figure 1C) whose separation averages 25 ± 5 nm, suggesting that model 1 in which only the distal FM domains are typically bonded prevails in the ER (Figure 1A) most closely approximates the structure of the adhesion plaque. Remarkably, when we used cytosolic staples that harbor the FM domains at the cytosolic face of the membranes (CD8_cytosolic), ER membranes were artificially stacked, confirming the trans–nature of the staples interaction (Figure 1—figure supplement 1). Although we exclusively visualized staples resulting from trans-interactions (two opposed membranes), we cannot rule out that cis-interactions (on the same membrane) may occurs when the physical constraints are favorable.

At the confocal level, the staples showed a reticular pattern typical of ER. Only when AP21998 is added to disaggregate the staples in the ER is CD8_lumenal transported to the cell surface (Figure 2A). Importantly, when the staples are stuck in the ER they did not prevent the production or transport of co-expressed Golgi resident enzymes (ST-RFP), showing that ER staples are not toxic (Figure 2B).

Figure 2 with 1 supplement see all

Download asset Open asset

We also employed biochemical tests of cell extracts to characterize the staples. When tested by SDS-PAGE, the disaggregating drug increased Triton-X100 solubility (without sonication procedure) of CD8_lumenal in cell extracts by a factor 3.9 ± 0.9 (Figure 2C). Note that this relatively modest response of CD8_lumenal to the disaggregating-drug (44% Triton X100-solubility after 4 hr treatment (Figure 2C) can be attributed to the transient transfection procedure. A large portion of the transfected cells showed a very high level of CD8_lumenal over-expression, and did not respond well to the disaggregating drug. As a consequence CD8_lumenal, remained aggregated in the ER even during drug treatment. This also explains the apparent moderate efficiency of transport of disaggregated-CD8 when measured by bulk assay (Figure 3B). Disaggregated CD8_lumenal, expected to be O-glycosylated (Nilsson et al., 1989; Lavieu et al., 2010), appeared as an isoform with reduced SDS-PAGE motility (* Figure 2C). As expected, Jacalin, a lectin that binds D-Galactose residues, exclusively precipitates this isoform (Figure 2D). In addition, metabolic radiolabeling experiments showed that staples (∼5.5 hr half-life) were slightly more stable that disaggregated CD8_lumenal (∼2.5 hr half-life) (Figure 2—figure supplement 1). Such stability is enough to guarantee the presence of the staples during numerous complete rounds of anterograde transport either from ER to Golgi or across the entire Golgi (around 20 rounds if we assume that one round of transport takes 15 min). Note that staples persist in the ER and in the Golgi (Figure 2—figure supplement 1) for the same time period (5–6 hr) before degradation by a cellular machinery that is not yet known.

Figure 3 with 1 supplement see all

Download asset Open asset

Together, these results showed that ER-aggregated CD8_lumenal forms staples that remain within the ER during their lifetime. However, this is a reversible state because adding the disaggregating drug dissolved a portion of the staples and allowed disaggregated CD8_lumenal to behave as a proper anterograde cargo that is glycosylated in the Golgi prior to being targeted to the cell surface membrane.

Temperature blocks have allowed the selective interruption of the secretory pathway in mammalian cells. A classic example is the 20°C temperature block, which slows exit from the trans-most Golgi cisterna (TGN) more than it slows ER → Golgi transport or transport from cis → trans Golgi (Matlin and Simons, 1983). As a result secretory cargoes accumulate in the TGN, from which they can be synchronously released by raising the temperature to the physiological value of 37°C. Temperature blocks in the 15–18°C range slow export of secretory cargo from the ER, but slow anterograde transport in the Golgi stack even more, resulting in limited penetration of the Golgi stack with accumulation in the ER–Golgi intermediate compartment (ERGIC) and the cis-most Golgi cisterna or two (Hobman et al., 1992; Balch et al., 1994; Volchuk et al., 2000).

Following an established procedure (Volchuk et al., 2000) we employed a 16°C temperature block to load and accumulate some of the disaggregated CD8_lumenal into the cis-Golgi of HeLa cells. Then, continuing at 16°C, we washout the disaggregating drug to allow for membrane staples reformation in the cis-Golgi. Finally, to explore the fate of these staples we raised the temperature to 37°C to allow anterograde transport to resume (Figure 3A). Unless it is explicitly stated to the contrary, all of these experiments were performed in presence of cycloheximide to prevent protein synthesis.

An independent, biochemical test confirmed that the unassembled CD8_lumenal had reached the cis-Golgi but not the trans-Golgi at 16°C. Specifically, a portion of disaggregated CD8_lumenal is precipitated by Helix Pomatia (Figure 3—figure supplement 1), a lectin that binds a N-Acetyl-Galactosamine residue that is only added within the cis-Golgi (Roth et al., 1994). In contrast, the lectin Jacalin, which requires trans-Golgi localized modifications for binding (Spicer and Schulte, 1992), did not precipitate CD8_lumenal at 16°C (Figure 3—figure supplement 1), although it did at 37°C (Figure 2D).

To explore the fate of the staples when transport is permitted, we first used a surface biotinylation method and immunofluorescence in intact cells to assess the surface exposure of CD8_lumenal after release of the temperature block to allow anterograde transport to resume. Surprisingly, the CD8_lumenal re-aggregated in the Golgi at 16°C did not reach the surface and was retained in the Golgi as judged by confocal microscopy (Figure 3B,C, lane 2). Golgi retention of the staples is due to controlled aggregation, because this block to transport was removed when the drug was added back, and CD8_lumenal (disaggregated) now moves from the Golgi to the plasma membrane (Figure 3B,C, lane 3). Note that except when mentioned otherwise, we used as a loading control 10% of the pellet fraction (insoluble fraction after sonication), which is almost free of soluble disaggregated-CD8_lumenal (including the glycosylated-form). Again we attributed the apparent low efficiency of CD8_lumenal transport to transient transfection. Cells that respond to the drug show >50% transport efficiency to the PM after 2 hr chase (Figure 3C, lane 3), consistently with the efficiency of transport of regular CD8-GFP (FM4-free) reported previously (Lavieu et al., 2010).

Next, we used confocal microscopy to further independently delineate where the staples accumulated within the Golgi at 16°C and their fate during a chase at 20°C which allows anterograde transport to proceed to the trans-Golgi but not beyond. We confirmed that co-expressed cis- and trans-Golgi fluorescent markers (Grasp65 and Golgin97, respectively) could be well discriminated by confocal microscopy and the degree of separation quantified with Pearson’s coefficient (Figure 4A and graph). As a negative control we showed that ER-staples did not co-localize with either the cis- or trans-Golgi markers (Figure 4B, 4 and graph). Using the same method we demonstrated that the initial localization of membrane staples Golgi staples formed at 16°C overlaps with the cis- but not the trans-Golgi marker. This co-localization with the cis- but not the trans-Golgi marker persists following warm-up to 20°C (Figure 4B, 2, 3 and graph), implying that the staples do not migrate in the cis → trans direction within the stack, even under conditions where this transport process is expected to resume. Again, the block to transport of the staples is entirely due to aggregation, because adding the drug to disaggregate the staples now allows transport and increased co-localization with the trans-Golgi marker following the warm-up to 20°C (Figure 4B, 2 and graph).

Figure 4

Download asset Open asset

Electron microscopy was used to further confirm the results from confocal microscopy. The staples were recognized in thin sections as electron dense plaques within Golgi cisternae (arrows, Figure 5A–C) spanning a gap of 27 ± 3 nm (Figure 5D) between opposing membrane surfaces, similar to their morphology within the ER. Staples formed at 16°C and then chased at 20°C for either 2 or 6 hr (which according to confocal microscopy remained in the cis-Golgi) were as would now be expected exclusively localized at one face of the Golgi, within the very first or more rarely the second and the third cisternae (Figure 5A–C and graph Figure 7C for quantification). Immuno-electron microscopy established that this face is the cis-Golgi since the gold particles recognizing the GFP harbored by the staples co-localized with the gold particles that bind GM130, a cis-Golgi marker (Figure 5E). Confirming this, the staples were at the opposed end of the stack from clathrin, which is a trans-Golgi marker (Figure 5F).

Figure 5 with 1 supplement see all

Download asset Open asset

Consistently, lectin-precipitation experiment showed that staples formed at 16°C and chased at 20°C harbored a cis-glycan signature (precipitated by Helix pomatis) but not the trans-glycan signature (not precipitated by Jacalin) (Figure 3—figure supplement 1, lane 2)

Altogether, the above experiments reveal that membrane staples formed in the first cisterna of the Golgi stack do not move across the stack, even after several hours.

Interestingly, the staples accumulate at the central portion of the cis-most cisternae (Figure 5A–C and Figure 5—figure supplement 1) being excluded from their rims. This could be the consequence of the fixed dimension of the staples, which are too small to fit across a dilated rim (Figure 5D). So, the staples would be expected to preferentially form in the flattened central regions of a cisterna.

Because large soluble aggregates such as collagens can move forward rapidly across the stack (Bonfanti et al., 1998) it was unexpected to observe that membrane staples remain fixed in cis-most cisternae. We therefore wanted to see if staples deposited in later Golgi cisternae might also be static. To deposit staples throughout the Golgi stack we allowed CD8_lumenal to enter the Golgi at 20°C in the presence of drug to maintain it in the disaggregated state. Then, the drug was removed at 20°C. Surface biotinylation established that staples formed at 20°C remained at the Golgi even after the cells were warmed up to 37°C for 2 hr (Figure 5G), similar to the cis-Golgi staples. Electron microscopy revealed that the staples formed at 20°C were present and retained throughout the entire Golgi stack, including the trans-most Golgi cisterna (Figure 5H). This is not surprising because significant back-up of cargo in the stack at 20°C has been previously documented (van Deurs et al., 1988).

The fact that the staples are retained in place at every level of the Golgi stack further emphasizes the static nature of the cisterna on the time scale being studied (5–6 hr) and also rules out the caveat that the apparently stable cis localization of staples at 16°C results from anterograde transport followed by efficient retrieval to the cis cisternae.

To rigorously conclude that the staples are immobile during ongoing anterograde transport, we need to demonstrate positively that anterograde transport continues in the same stacks that retain the staples. To test this, we introduced a well-studied anterograde cargo into cells harboring staples, the VSV-encoded G protein. Specifically, we used a GFP-tagged version of a temperature-sensitive mutant G protein that is retained in the ER at 39°C, but which can exit the ER and be transported to the cell surface via the Golgi when the temperature is subsequently lowered to 32°C (Presley et al., 1997). In these experiments, the staples were labeled with DsRed (DsRed-CD8_luminal) and co-expressed with GFP-tagged VSV-G. We estimated by confocal microscopy that >95% of the cells were co-transfected. We used the 39°C temperature block to retain VSV-G-GFP in the ER, then a fraction of both VSV-G-GFP and disaggregated DsRed-CD8_luminal (+drug) were partially released into the Golgi at 16°C, prior to re-aggregating CD8_lumenal (−drug) at 16°C into cis-Golgi staples. Only then was the temperature raised to 32°C for 2 hr to test the transport of VSV-G-GFP (the fraction localized into the stapled cis-Golgi and the fraction that was still into the ER). Surface biotinylation showed that the amount of VSV-G detected at the surface of cells that harbored staples in their Golgi was 91 ± 12% of the amount of VSV-G detected at the surface of cells harboring a normal Golgi (Figure 6A). This showed that the stapling procedure did not interfere dramatically with the transport of classical anterograde cargo.

Figure 6 with 3 supplements see all

Download asset Open asset

To be certain that the very Golgi harboring static staples continued to engage in productive anterograde transport, we measured the rate of efflux of VSV-G from Golgi areas harboring staples as compared to those that did not. We used photo-activatable GFP-tagged VSV-G (PAGFP-VSV-G [Patterson and Lippincott-Schwartz, 2002]) to activate a cohort of protein within a Golgi area to enable us to measure the release of VSV-G from that same Golgi complex. The rate of the release (half-time ∼ 5 min) of PA-GFP-VSV-G from even long-term (4 hr) stapled Golgi was indistinguishable from that measured from a non-stapled Golgi identified with a red-tagged Golgi marker (ST-RFP) (Figure 6B). This observation and the fact that stapled-Golgi remain properly stacked established that stapled-Golgi is fully functional for transport.

Then we asked if Golgi-staples could interfere with the transport of endogenous large cargo. We used Saos-2 cells, which secrete endogenously produced collagen-I. After transfection with GFP-CD8_lumenal, Saos-2 cells were incubated at 20°C in the presence of the disaggregating drug, followed by drug removal to allow the positioning of the staples within the Golgi (Figure 6C upper panel, lane 1,2 show the glycosylated form of CD8_lumenal). 2 hr after stapling the Golgi membranes, the culture media was removed and cells were incubated at 37°C in the presence of fresh media containing ascorbate to release collagen-I from the ER. After a 2 hr chase the amount of newly secreted collagen-I was determined by TCA precipitation/immunobloting procedure (Figure 6C). Cells harboring stapled-Golgi for 2 hr showed 92.5 ± 13% of freshly secreted collagen-I when compared to cells harboring normal Golgi (set to 100%). This suggests that, as reported above with for VSV-G, anterograde transport of Collagen-I is not altered when staples are positioned within the Golgi. To rigorously assess if Golgi staples had only minor effect on secretion, we further tested the rate of transport of collagen-I and MMP2, two proteins secreted by Saos-2 cells. The kinetics of secretion for both proteins did not show any robust differences whether the Golgi apparatus was stapled or not (Figure 6—figure supplement 1). Again, this suggests that the amount of staples that is loaded in the Golgi membranes during the re-aggregation procedure does not poison the anterograde transport function of the Golgi.

To further test if Golgi staples interfere with the behavior of Golgi membranes, we assessed the redistribution of Golgi resident enzymes into the ER under BFA treatment (Lippincott-Schwartz et al., 1989). Redistribution of Golgi enzymes into the ER was observed whether Golgi staples were present or not (Figure 6—figure supplement 2). Note that consistently with the moderate disaggregation reported above (44% solubilization in Triton X100), some aggregates remained in the ER (arrows Figure 6—figure supplement 2). These ER-remaining aggregates did not perturb the behavior of the Golgi resident enzymes (Golgi targeting and BFA-redistribution).

Importantly, FRAP experiments (Figure 6—figure supplement 3) showed that staples (localized either within the Golgi or within the ER) readily diffuse within the cisternal membranes and do not perturb the diffusion of Golgi resident enzymes; in fact, Golgi-staples diffused only slightly more slowly than disaggregated CD8_lumenal at 20°C (Figure 6—figure supplement 3). This would explain in a satisfying manner why the essential functions of the Golgi (and ER) such as transport and glycosylation are not perturbed by the presence of the staples, since these behave like any cargo or resident protein within the membrane, as further evidenced by the re-distribution of staples (like any other Golgi cargo or resident enzyme) to the ER as a result of BFA treatment.

Retrograde transport is also a major aspect of Golgi function, and it would potentially be of interest to investigate if staples interfere with such function. While we cannot presently rule out this caveat, even if retrograde transport were blocked it would not diminish our conclusion that anterograde transport of both small and large cargo (next section) can occur at normal speeds without movement of cisternae in the stack.

Previously, we found that soluble aggregates that are compositionally similar to the staples (the main difference being that they were not membrane-anchored) rapidly move forward through the Golgi, analogous to collagens (Volchuk et al., 2000). We noted that these soluble aggregates were concentrated at the rims of the Golgi cisternae, analogous to collagens. Given the prima facia contradiction between our current observation of immobile staples and our prior observation of mobile soluble aggregates composed of virtually the same protein, we repeated the earlier studies and compared the fates of soluble aggregates and staples in the same cells.

We first asked whether lumenally-expressed soluble FM4 (FM4-hGH) and membrane-bound FM4 proteins (CD8_lumenal) co-aggregate in the cell. This would be expected because they share the same homotypic adhesion FM modules. As expected, when the two FM-containing proteins were present in the ER they were efficiently co-immunoprecipitated from cell extracts (Figure 7—figure supplement 1). However, when these same proteins were disaggregated and allowed to leave the ER and then re-aggregated for 1 hr in the cis-Golgi at 16°C, or even further in the Golgi stack at 20°C, the Golgi-modified forms of these same proteins were no longer co-immunoprecipitated (Figure 7—figure supplement 1). On the contrary homo-aggregates constituted of soluble cargo (hGH) were efficiently formed even after re-aggregation at 16°C or 20°C (Figure 7—figure supplement 1). This suggests that hetero-aggregates are not formed efficiently within the Golgi cisternae.

Confocal microscopy independently established that the soluble and membrane-bound FM proteins segregated into separate aggregates within the same Golgi area (Figure 7—figure supplement 2) and, electron microscopy confirmed that this separation occurs even within the same cisternal compartment. Even when both reside in the same cisterna, the electron-dense plaques comprising the staples are localized in the central regions of the cisterna while the soluble aggregates are concentrated at the dilated rims (Figure 7E).

Figure 7 with 2 supplements see all

Download asset Open asset

What are the fates of soluble aggregates as compared to staples in the Golgi when anterograde transport is permitted to resume at 37°C? Whereas the Golgi retained the staple protein it released the soluble aggregates to the secretory pathway (Figure 7A). This biochemical result confirms yet again that hetero-aggregates (composed of staples and soluble FM proteins) were not formed efficiently within the Golgi, confirming by yet another approach that the two FM proteins aggregate separately within the Golgi cisternae (even when using a 2:1 membrane:soluble protein ratio).

Electron microscopy was also used to follow the fate of the two types of aggregates. When aggregates accumulated in the cis-Golgi are allowed to move anterograde through but not exit the Golgi following warm-up to 20°C, ≈75% of the staples, as expected, remain at the cis-face of the Golgi, whereas ≈75% of the soluble aggregates are transported to the opposite, trans-face of the same Golgi stacks (Figure 7B and graph).

We conclude that the anterograde transport process permits large soluble aggregates to be carried forward between cisternae that are immobile.

HeLa cells were maintained at 37°C in 5% CO₂ in DMEM (Gibco, Grand Island, NY, United States) supplemented with 10% FBS (Gibco). Saos-2 cells were maintained with the same media supplemented with ascorbate (25 μg/ml). HeLa cells were transfected using lipofectamine 2000 (Invitrogen, Grand Island, NY, United States) as recommended by the manufacturer. Saos-2 cells were transfected using electroporation (Nepa21type II model from Nepa gene, Chiba, Japan).

CD8_lumenal (ss-GFP-FM4-CD8 and ss-DsRed-FM4-CD8) were generated by sequential insertion of GFP (or DsRed), and CD8 encoding sequences into a pC4 ss-FM4 backbone vector (ARIAD), using XbaI/SpeI compatibility and BamH1 restriction site. CD8 was amplified from pC4-CD8-GFP plasmid (Lavieu et al., 2010). A similar strategy was used to create ss-DsRed-FM4hGH and ss-GFP-FM4hGH. PAGFP-VSVG and _tsVSVG-GFP were ordered from Addgene. J Rohrer provided ST-RFP. Grasp65 and Golgin97 encoding plasmids were a gift from Y Wang and S Munro, respectively.

These experiments were performed as described previously (Lavieu et al., 2010) with slight modifications. HeLa cells were grown in six-well plates. All the disaggregation/re-aggregation experiments were performed in HBSS (Gibco) supplemented with 10% FBS (Gibco) and 100 μg/ml cycloheximide (Sigma, St Louis, MO, United States). Disaggregating drug (AP21998; ARIAD, Cambridge, MA, United States) was used at 1 μM. 16°C, 20°C, 32°C and 37°C incubations were performed using temperature-controlled incubators.

Briefly, subconfluent HeLa cells were biotinylated in 1 mg/ml EZ-Link Sulfo-NHS-SS Biotin at 4°C for 30 min and washed two times with PBS. After neutralization with 100 mM glycine pH7 at 4°C for 30 min, cells were washed, detached and collected by centrifugation. Pellets were resuspended in lysis buffer (1% Triton X-100, 5 mM EDTA, 150 mM NaCl pH7.4 and 1% protease inhibitor) for 1 hr at 4°C and sonicated (80W, 10 s pulse). The cell lysate was centrifuged, and the supernatant was collected. Samples were incubated with neutravidin agarose resin at 4°C overnight for Biotin-labeled protein precipitation. The neutravidin interacting proteins were eluted with SDS sample buffer from neutravidin pellets. Extracted proteins were first separated in SDS-polyacrylamide gels and then transferred onto nitrocellulose membranes for immunoblotting. After blocking with fat-free milk, the membranes were incubated with appropriate primary antibodies. Primary antibodies were detected by chemiluminescence using horseradish peroxidase-conjugated secondary antibodies. Fluorographs were quantitatively scanned using the NIH image software.

When TCA precipitation was required, 0.1% FBS was used. Chase media was collected and precipitated overnight at 4°C with 10% TCA. After centrifugation TCA pellets were washed with acetone prior to being resuspended in loading buffer and analyzed by immunoblotting blot as described above.

For lectin precipitation, HeLa cells extracts containing CD8_lumenal were incubated at 4°C with Jacalin-immobilized lectin (Sigma-Aldrich) or Helix pomatia Gel-HPA-immobilized lectin (EY Laboratories, San Mateo, CA, United States) prior to being processed for immunoblotting as described above. For immunoblotting, we used monoclonal anti-GFP (Roche, Brandford, CT, United States), polyclonal anti-mCherry (Biovision, Milpitas, CA, United States), anti-Collagen-I (SP1.D8 from Developmental Studies Hybridoma Bank, Iowa City, IA, United States), and polyclonal anti-MMP2 (Cell Signaling, Boston, MA, United States).

HeLa cells were grown on glass coverslip in 24-well plates. For surface labeling, cells were incubated on ice for an hour with a monoclonal anti-GFP antibody (Roche) prior to being fixed for 10 min with 4% PFA and incubated with an Alexa Fluor 546-coupled antibody. Cells were then processed for confocal microscopy imaging, which was performed in multi-tracking mode on either a Zeiss LSM510 or a Zeiss LSM510 META. Images were analyzed using Zeiss LSM510 software or using ImageJ (co-localization finder plugin). For photo-activation experiments we used a multicolor spinning-disk confocal based on an inverted Olympus microscope (IX-71) and Perkin-Elmer Ultraview system with 5-laser and FRAP/photoactivation. VSVG-PAGFP was photo-activated within the Golgi region (identified by the co-expressed DsRed-tagged protein) at 405 nm (80% laser power) for 5 s. VSVG-PA-GFP was imaged using 488 nm light (10% laser power).

No external funding was received.

We thank ARIAD Pharmaceuticals for providing us with its secretion/aggregation kit reagents. Thanks to Derek Toomre for providing access to the spinning disk microscope and the electroporation apparel. Thanks to Daniel Kuemmel for modeling the trans-complexes. Thanks to Morven Graham and Xinran Liu for providing EM assistance. Thanks to Pietro De Camilli for the gift of the Clathrin antibody. Thanks to Myun Hwa Dunlop for editing the manuscript.

© 2013, Lavieu et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.