Anal. Chem. 2007, 79, 5139-5142
Metabolic Cytometry. Glycosphingolipid Metabolism in Single Cells Colin D. Whitmore,† Ole Hindsgaul,‡ Monica M. Palcic,‡ Ronald L Schnaar,§ and Norman J. Dovichi*,†
Department of Chemistry, University of Washington, Box 351700, Seattle, Washington 98195, Carlsberg Laboratory, Gamle Carlsberg Vej 10, DK-2500, Valby Copenhagen, Denmark, and Departments of Pharmacology and Neuroscience, The Johns Hopkins School of Medicine, 725 North Wolfe Street, Baltimore, Maryland 21205
Glycosphingolipids are found on all vertebrate cells and constitute major cell surface determinants on all nerve cells, where they contribute to cellular diversity and function. We report a method for the analysis of glycosphingolipid metabolism in single cells. The ganglioside GM1 was tagged with the fluorescent dye tetramethylrhodamine. This labeled compound was taken up and metabolized by a culture of pituitary tumor (AtT-20) cells. After 50 h, the cells were formalin fixed. Cells were aspirated into a fused-silica capillary and lysed, and components were separated by capillary electrophoresis with a laser-induced fluorescence detector. All metabolic products that retained the fluorescent dye could be detected at the low-zeptomole level. A total of 54 AtT-20 cells were individually analyzed using this procedure. The electrophoretic profiles were remarkably reproducible, which facilitated identification of components based on the migration time of fluorescently labeled standards. Eleven components were detected, and the average peak height of these components spanned more than 2 orders of magnitude, so that trace metabolites can be detected in the presence of abundant components. The most highly abundant components generated 10% relative standard deviation in normalized abundance. The average cell took up roughly 2 amol (106 copies) of the labeled substrate. This method allows determination of cell-to-cell diversity and regulation of glycosphingolipid metabolism.
Glycosphingolipids consist of a “glyco” head that is composed of a series of saccharide groups and a “sphingo” tail that can vary in length but is typically an 18-carbon nonpolar amino alcohol. Gangliosides are sphingolipids that have one to four sialic acids attached to the head and have a long-chain fatty acid attached to the sphingosine to create a ceramide tail. (The number of sugar and sialic acid groups gives rise to the common naming of gangliosides. The subscript letter indicates the quantity of sialic acid groups: A, M, D, and T stand for asialo, monosialo, disialo, and trisialo. The number originated from the order of peak * To whom correspondence should be addressed. E-mail: chem.washington.edu. † University of Washington. ‡ Carlsberg Laboratory. § The Johns Hopkins School of Medicine. 10.1021/ac070716d CCC: $37.00 Published on Web 06/14/2007
dovichi@
© 2007 American Chemical Society
appearance in thin-layer chromatography and therefore indicates the length of the sugar chain. GM1 has the longest chain, GM2 lacks the terminal galactose, and GM3 has one saccharide less.) Gangliosides are predominately located on the outer membranes of neurons, where they account for 20-25% of lipids.1 The nonpolar tail is integrated into the membrane, but the head groups are exposed; this exposure allows ganglioside metabolism to be crucially involved in a range of extracellular interactions. The role of gangliosides and their metabolites in cellular proliferation, apoptosis, and differentiation has been the subject of extensive research.2 There are several severe genetic disorders associated with disruption of metabolism of these compounds. Tay-Sachs disease, in which large amounts of GM2 accumulate, is a prominent example. There is significant cell-to-cell heterogeneity in the amount of gangliosides and their metabolites. Mice were prepared that could not synthesize lipids larger than GM3 and GD3.3 They had axon degeneration at 43 times the rate of control animals, but this degeneration only took place in a minority of cells. The hypothalamus of mice with Tay-Sachs was stained to reveal GM2 storage, which revealed that accumulation occurred unevenly among the neurons.4 The diversity of neuronal phenotypes includes differential glycosphingolipid metabolism. Antibody staining and fluorescence microscopy are commonly used to characterize heterogeneity in ganglioside expression.4 However, there are several limitations to the use of antibodies. First, antibodies are not available for all structures. Second, antibodies exhibit cross-reactivity. Third, while different fluorescent stains can be used to visualize several targets simultaneously, correcting for spectral cross talk is difficult, which limits the dynamic range of the measurement. We have reported an approach for monitoring complete metabolic pathways in single cells.5 Compounds labeled with highly fluorescent dyes can be detected with exquisite sensitivity by capillary electrophoresis and laser-induced fluorescence.5-9 If (1) Tettamanti, G. Glycoconjugate J. 2004, 20, 301-317. (2) Colombaioni, L.; Garcia-Gil, M. Brain Res. Rev. 2004, 46, 328-355. (3) Sheikh, K. A.; Sun, J.; Liu, Y. J.; Kawai, H.; Crawford, T. O.; Proia, R. L.; Griffin, J. W.; Schnaar, R. L. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 75327537. (4) Platt, F. M.; Neises, G. R.; Reinkensmeier, G.; Townsend, M. J.; Perry, V. H.; Proia, R. L.; Winchester, B.; Dwek, R. A.; Butters, T. D. Science 1997, 276, 428-431. (5) Krylov, S. N.; Zhang, Z. R.; Chan, N. W. C.; Arriaga, E.; Palcic, M. M.; Dovichi, N. J. Cytometry 1999, 37, 14-20. (6) Cheng, Y. F.; Dovichi, N. J. Science 1988, 242, 562-564. (7) Wu, S.; Dovichi, N. J. J. Chromatogr. 1989, 480, 141-155.
Analytical Chemistry, Vol. 79, No. 14, July 15, 2007 5139
Figure 1. Schematic diagram of the metabolic pathway examined. Ceramide:TMR is the lipid tail that has had its fatty acid chain replaced with the highly fluorescent dye tetramethylrhodamine.
a substrate is tagged with a fluorescent label, any metabolic transformations that preserve that label can be detected at very low levels. In our ganglioside studies, GM1 is fluorescently labeled with tetramethylrhodamine, a highly fluorescent dye that does not prevent uptake or metabolism of the substrate. We have created a capillary electrophoresis method for the separation and detection of labeled GM1 and its metabolites with subzeptomole detection limit.9 Pagano and co-workers have used similar fluorescently labeled compounds to examine glycosphingolipid localization and transport.10,11 That group has developed a method of prompting cellular uptake of these fluorescent probes by complexing them to defatted bovine serum albumin.11 In this paper, we report analysis of sphingolipid metabolism in single cells, and we present its application to the characterization of the metabolism of GM1 in pituitary tumor (AtT-20) cells. The metabolic pathways under consideration are shown in Figure 1. EXPERIMENTAL METHODS Reagents. The synthesis of tetramethylrhodamine-labeled GM1, GA1, GM2, GA2, GM3, LacCer, GlcCer, and Cer has been previously described.12 Briefly, the preparation involved acylation of the homogeneous C18 lyso forms of GM1, LacCer, GlcCer, and Cer with the N-hydroxysuccinimide ester of a β-alanine tethered 6-tetramethylrhodamine (TMR) derivative, followed by conversion using galactosidase, sialidases, and sialytranserase enzymes. Figure 2 presents the structure of the labeled GM1. Cells and Cell Culture. AtT-20 cells were grown in Dulbecco’s modified Eagle medium (DMEMH10F5) with 10% horse serum in the presence of antibiotics. The cells were incubated at 37 °C with 5% CO2. Uptake of Fluorescent GM1 and Homogenate Preparation. Fluorescent GM1 was complexed to defatted bovine serum albumin in a 1:1 ratio in a procedure adapted from Pagano.11 The 1.3 mg of TMR-labeled GM1 was dissolved in 400 µL of water, 760 µL of ethanol, and 40 µL of chloroform. The 75-µL aliquots of this solution were evaporated in a centrifuge under vacuum for 1 h. One aliquot was dissolved in 50% ethanol. This aliquot was mixed by vortexing with 10 mL of growth medium enhanced only with 3.4 mg of defatted bovine serum albumin. The resulting medium (8) Zhao, J. Y.; Dovichi, N. J.; Hindsgaul, O.; Gosselin, S.; Palcic, M. M. Glycobiology 1994, 4, 239-242. (9) Whitmore, C. D.; Olsson, U.; Larsson, A.; Hindsgaul, O.; Palcic, M. M.; Dovichi, N. J. Electrophoresis. In press. (10) Pagano, R. E.; Martin, O. C. Biochemistry 1988, 27, 4439-4445. (11) Marks, D. L.; Singh, R. D.; Choudhury, A.; Wheatley, C. L.; Pagano, R. E. Methods 2005, 36, 186-195. (12) Larsson, E. A.; Olsson, U.; Whitmore, C. D.; Martins, R.; Tettamanti, G.; Schnaar, R. L.; Dovichi, N. J.; Palcic, M. M.; Hindsgaul, O. Carbohydr. Res. 2006.
5140
Analytical Chemistry, Vol. 79, No. 14, July 15, 2007
contained a 5 mM 1:1 complex of fluorescent GM1 and BSA. The cells were incubated in this medium for 50 h. After incubation, the growth medium was removed by pipet. The cells were detached with 1X trypsin at 37 °C for 5 min. Cells were suspended in 10 mL of phosphate-buffered saline (PBS), transferred to a 15-mL centrifuge tube, and spun for 5 min; supernatant was removed and replaced with fresh PBS. This washing procedure was performed five times. A cellular homogenate was prepared by spinning the cells, removing the PBS, and mixing the cells with 50 µL of a 1% SDS solution.12 Cells were then sonicated at 60% duty cycle for ∼30 min. Solutions were stored on ice before analysis. For single-cell analysis, cells were fixed with freshly prepared 4% formaldehyde in PBS at room temperature for 12 min. After treatment, cells were spun and the formalin solution was discarded. The reaction was quenched by resuspending the cells in 10 mM glycine in PBS. After bubbling subsided (∼15 min), the cells were spun and the supernatant was removed. Cells were resuspended in a fresh aliquot of glycine in PBS. Cells were then washed in PBS five times before storage at 4 °C. Cells were then rinsed three times before use. Capillary Electrophoresis. The capillary electrophoresis system is similar to others used in this group.5-9,13,14 Briefly, a 37-cm-long, 20-µm-i.d., and 150-µm-o.d. uncoated fused-silica capillary was used for the separation. The instrument was equipped with a postcolumn sheath-flow cuvette for fluorescence detection. Fluorescence excitation was provided by a 10-mW frequency-doubled Nd:YAG laser operating at 532 nm. Emission was collected by a 0.45 NA microscope objective, passed through a 580 DF40 band-pass filter, imaged onto a GRIN-lens coupled fiber optic, and detected by a single-photon counting avalanche photodiode. The signal was recorded on a PC, and data were processed using Matlab running on a Macintosh computer. The cell was lysed on column with 0.4% Triton X-100 in water15 by loading the cell between plugs of this buffer. Injection was accomplished by exposing the sheath flow at the terminus of the capillary to a 1-m drop for 1-s increments, creating enough of a siphon to inject a small volume.16 A cell was visually located on a PHEMA-coated slide17 with an inverted microscope, and the capillary was lowered to it using a set of micromanipulators. The separation was performed using a running buffer composed of 10 mM sodium tetraborate, 35 mM sodium deoxycholate (SDC), and 5 mM methyl-β-cyclodextrin. Sodium deoxycholate was chosen as the anionic surfactant because it forms “weak” micelles that allow for the partition of molecules that have very hydrophobic components, such as the tails of the glycosphingolipids. Methyl-β-cyclodextrin adds another dimension to the separation; its neutral charge increases the mobility of lipids that interact with it relative to their interactions with SDC. We have discussed the partition of lipids into this pseudostationary phase elsewhere.18 The running voltage was 18 kV. (13) Dovichi, N. J.; Hu, S. Curr. Opin. Chem. Biol. 2003, 7, 603-608. (14) Kraly, J. R.; Jones, M. R.; Gomez, D. G.; Dickerson, J. A.; Harwood, M. M.; Eggertson, M.; Paulson, T. G.; Sanchez, C. A.; Odze, R.; Feng, Z.; Reid, B. J.; Dovichi, N. J. Anal. Chem. 2006, 78, 5977-5986. (15) Pang, Z.; Al-Mahrouki, A.; Berezovski, M.; Krylov, S. N. Electrophoresis 2006, 27, 1489-1494. (16) Krylov, S. N.; Starke, D. A.; Arriaga, E. A.; Zhang, Z.; Chan, N. W.; Palcic, M. M.; Dovichi, N. J. Anal. Chem. 2000, 72, 872-877. (17) Krylov, S. N.; Dovichi, N. J. Electrophoresis 2000, 21, 767-773. (18) Zhang, l.; Hu, S.; Cook, L.; Dovichi, N. J. Electrophoresis 2002, 23, 30713077.
Figure 2. Structure of GM1 and GM1-TMR. Symbols are shown in the inset for Figure 1.
Figure 3. Electropherograms of 54 AtT-20 single cells incubated with the fluorescently labeled GM1. Electropherograms are ordered based on the amplitude of the ceramide peak (far right).
Data were treated with a five-point median filter to remove spikes due to bubbles and cellular debris and then convoluted with an 84-point-wide Gaussian function with 10-point standard deviation. Peak alignment was performed using a two-point procedure based on the GM1 and Cer peak positions.19 RESULTS Separation and Identification of Metabolites in Single Cells. In preliminary studies, AtT-20 cells were incubated for periods between 2 h and several days with the fluorescently labeled GM1. The quantity of metabolites increased with incubation time up to 50 h. Therefore, a 50-h incubation was chosen for single cell experiments to produce the most complex metabolic profile. Uptake of the lipid was confirmed via fluorescence microscopy. Fifty-four individual cells were analyzed (Figures 3-5). Peak identification was accomplished by comparing the data with an electropherogram generated from fluorescently labeled standards.9 Most of the anticipated metabolic products of the labeled GM1 were observed, as were five components that are yet to be identified. It is speculated that these correspond to the synthesis of the more complex gangliosides that lie to the bottom and right of the metabolic pathway in Figure 1; as we show in the Supporting Information, these cells produce significant amounts of GD3 and GD1a. A blank injection of supernatant produced a featureless electropherogram. (19) Li, X. F.; Ren, H.; Le, X.; Qi, M.; Ireland, I. D.; Dovichi, N. J. J. Chromatogr., A 2000, 869, 375-384.
Figure 4. Comparison of single-cell, homogenate, and standard electropherograms. The bottom trace (solid) is generated by electrophoresis of a set of eight fluorescently labeled standards. The middle trace (dashed) is generated by electrophoresis of a homogenate prepared by incubating AtT-20 cells with labeled GM1. The top trace (dotted) shows the average of the 54 single-cell electropherograms presented in Figure 3.
Figure 5. Expanded view of the data of Figure 4. Electropherogram of standards (bottom, solid), cellular homogenate (middle, dashed), and the average of 54 single cells (top, dotted) are expanded to highlight low-amplitude components. One unidentified component (Un1) is off scale.
Heterogeneity Displayed in Single Cells from an AtT-20 Culture. A two-point peak alignment algorithm was employed, based on the GM1 and Cer peaks, to allow comparison of the singlecell data.19 After alignment, Figure 3, all components were in excellent registration. A nonlinear regression analysis was used Analytical Chemistry, Vol. 79, No. 14, July 15, 2007
5141
bands and a faint GM2 band in the thin-layer chromatogram. These data correlate with the single-cell results, where we observe strong GM1 and weaker GM2 signals. The thin-layer experiment stains sialic acid and is not able to observe neutral breakdown products. The thin-layer results suggest that it will be valuable to prepare the fluorescent GD3 and GD1a sphingolipid standards to test for comigration with the unknown peaks in the single-cell electropherogram. In particular, the addition of a sialic acid increases migration time by ∼20 s; the unknown peak 1 has a migration time that is consistent with GD1a.
Figure 6. Histogram of uptake of fluorescent substrate. The peak amplitudes were summed for the 54 single-cell electropherograms of Figure 3. The smooth curve is the least-squared fit to an exponential decay.
to fit a Gaussian function to the GM2 peak; the relative standard deviation in migration time was 0.04% after alignment. The excellent reproducibility in peak position allowed us to calculate the average electropherogram from the 54 cells, Figures 4 and 5. The average electropherogram showed remarkable agreement with the signal generated from a cellular homogenate for major components. Several minor components, particularly unknown component 3, did show differences between the homogenate and single-cell data, which may reflect differences in sample preparation for single cells and the homogenate.20 Next, nonlinear regression analysis was used to fit a set of Gaussians to the 11 peaks; these Gaussians had centers fixed at the position of the peaks in the average electropherogram, Figure 4, and width equal to the observed peak standard deviation of 0.5 s. Peak amplitude was forced to be non-negative. Substrate uptake was estimated from the sum of the peak amplitudes in each electropherogram. The total fluorescent signal corresponded to 1.6 amol (106 copies) of tetramethylrhodamine, and had an exponential distribution, Figure 6, with a large cellto-cell variation (RSD ) 80%). This variation may be due to differences in substrate uptake, cell lysis efficiency, the phase of the cell in the cell cycle, and photobleaching of the fluorescent tag before analysis. Peak heights were normalized to the total signal for each cell, and these normalized peak heights were averaged across the set of cells. The average normalized peak height varied by over 2 orders of magnitude. The substrate, GM1, generated about half of the signal, followed by GM2 and Cer, each of which accounted for roughly a quarter of the remaining signal. The relative standard deviation in the normalized peak heights was inversely proportional to peak height (r ) 0.91). Highly abundant components had small relative standard deviations in normalized peak height (∼10%). Components near the detection limits had much higher standard deviations, which approach 100%. We employed thin-layer chromatography to analyze sphingolipid content of a homogenate prepared from this cell line (Supporting Information).21 We observe strong GD1a, GD3, and GM1 (20) Krylov, S. N.; Arriaga, E.; Zhang, Z.; Chan, N. W.; Palcic, M. M.; Dovichi, N. J. J. Chromatogr., B 2000, 741, 31-35. (21) Schnaar, R. L. Methods Enzymol. 1994, 230, 348-70.
5142
Analytical Chemistry, Vol. 79, No. 14, July 15, 2007
CONCLUSIONS We have demonstrated the measurement of substrate and 10 putative sphingolipid metabolites in single pituitary tumor (AtT20) cells using capillary electrophoresis. A simple two-point migration time normalization procedure generated extremely reproducible peak positions. The experiment provides at least 2 orders of magnitude dynamic range, which facilitates detection of trace metabolites in the presence of abundant components. Relative standard deviation is inversely related to peak amplitude; abundant components can be detected with 10% cell-to-cell precision. There is relatively little heterogeneity in sphingolipid metabolism in this cultured cell line. As noted above, much higher heterogeneity is observed for sphingolipid expression in primary neurons. Five components did not comigrate with our standards. They are likely generated by catabolism of GM3 to the GD, GT, and larger structures. Standards for these compounds are being synthesized and will be available for spiking experiments. Glycosphingolipid expression varies among and within cell populations, during cellular differentiation and between normal cells and their transformed counterparts. Metabolic cytometry will be used to determine the variation of glycosphingolipid metabolic flux within a population of cells and to identify distinct glycosphingolipid metabolic patterns associated with identifiable cellular subpopulations. We anticipate applying these techniques to populations of stem cells, differentiating cells, mature (differentiated) cells, and transformed cells and correlating differences in glycosphingolipid flux with independent microscopic markers to discover new relationships between cellular phenotype and glycosphingolipid metabolic state. Few laboratories are equipped to monitor sphingolipid metabolism at these levels. However, our protocol employs formalinfixed cells, which are stable under routine storage and shipping conditions. As a result, it will be possible for investigators to employ these substrates in their laboratories to treat primary cells, which can then be fixed and shipped to a suitably equipped analysis laboratory. ACKNOWLEDGMENT This work was supported by Grant R33DK70317 from the National Institutes of Health. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review April 11, 2007. Accepted June 5, 2007. AC070716D
