presley supp info

SUPPLEMENTARY MATERIAL Manuscript L06538 TH/rst Supplementary Information for “In vivo dissection of COPI and Arf1 dynamics and role in Golgi membrane transport.” Presley, J. F, Ward, T. H., Pfeifer, A.C., Siggia, E. D., Phair, R. D. & Lippincott-Schwartz, J. Figure S1. Arf1-GFP can function as the sole copy of Arf1 in yeast. To show that ARF1 C-terminally tagged with Green Fluorescent Protein (GFP) is fully functional we used a method described recently (Baudin et al., 1993) to modify endogenous S. cerevisiae ARF1 by homologous recombination. GFP was PCR-amplified from pFA6-GFP(S65T)-HIS3Mx6 (Longtine et al., 1998) using the primers 5’-CCGGTGAAGGTTTGTATGAAGGTTTGGAATGGTTAAGTAACAGTTTGAAAcggatccccgggttaattaa-3’ and 5’-TATGTTTCATTTAGTTTATACAA GCGTATTTGATCCATATTCTAGAATTTgaattcgagctcgtttaaac-3’. The underlined sequences correspond to the position where the PCR fragment was introduced in the yeast genome. The selective marker used in this experiment was Schizosaccharomyces pombe his5 which is able to complement the S. cerevisisae HIS3 mutant. The PCR fragment was directly integrated into yeast strain CJY045-4-4 (S288C ura3-52 his3-200 ARF1 arf2::URA3) (from Cathy Jackson) and transformants were analysed by yeast colony PCR as well as fluorescence microscopy. The above FRAP sequence shows that membrane-bound Arf1-GFP rapidly recovers into a photobleached box. This demonstrates rapid exchange between membrane-bound and freely mobile cytoplasmic Arf1-GFP. Figure S2. Arf1[Q71L]-CFP protects COPI from membrane dissociation during BFA treatment. ldlF cells stably expressing COP-YFP were transfected with Arf1-CFP or with GTP-locked Arf1[Q71L]-CFP as indicated and incubated for 24 h at 37° C to allow expression. They were then incubated for 2 min in 5 µg/ml BFA and fixed in 2% paraformaldehyde for 20 min. Images of CFP and YFP (left and right panel pairs) were obtained on a Zeiss 410 confocal microscope. Note that COP-YFP remains Golgi-associated after BFA treatment in cells expressing Arf1[Q71L]-CFP (bottom panels), in contrast to untreated cells (top panels). The same results are observed for COPI in cells expressing Arf1[Q71L] without the CFP tag. These results suggest, therefore, that the CFP tag on Arf1 does not interfere with Arf1’s ability to interact with COPI on membranes or its ability to recruit COPI to membranes. Figure S3. Arf1-HA is lost from Golgi membranes faster than COPI after BFA treatment. ldlF cells on coverslips stably expressing COP-YFP were transfected with Arf1-HA, incubated for 24 h at 37° C to allow expression, and fixed in 2% paraformaldehyde for 20 min. In the lower two panels they were incubated in 5 µg/ml BFA for 30 sec prior to fixation. Fixed cells were then stained with a commercial mouse antibody against HA (Covance Research Products, Cumberland, VA) and a Cy3-labeled secondary anti-mouse antibody. Images were obtained on a Zeiss 510 confocal microscope. These results demonstrate that Arf1 modified with the small HA tag behaves similarly to Arf1-GFP in response to BFA treatment. Both Arf1 constructs dissociate within 30 sec of BFA treatment in a process that is not linked to COPI dissociation.  Figure S4. Aluminium fluoride protects COPI but not Arf1-HA from membrane dissociation during BFA treatment. ldlF cells stably expressing COP-GFP were transfected with Arf1-HA and incubated for 24 h at 37°C to allow Arf1-HA expression and then treated with BFA and/or AlF as indicated. After these treatments, cells were fixed and stained with mouse anti-HA and a TRITC-labeled anti-mouse antibody and examined on a Zeiss 410 confocal microscope using a 63x N.A. 1.4 oil immersion objective. Cells were incubated with AlF for 10 min and then BFA (5 µM) was added for 5 min. Alternatively, cells were treated with only BFA for 5 min. AlF (low) indicates treatment with 50 µM AlCl3 and 50 mM NaF. AlF (high) indicates treatment with 10 mM AlCl3 and 50 mM NaF. The data show that AlF at either high or low doses protects COPI but not Arf1-HA from dissociating from the Golgi during BFA treatment. These data support our conclusion that Arf1 and COPI are differentially sensitive to AlF treatment, as previously suggested by Donaldson et al. (1991). Co-immunoprecipitation of COP-GFP with COP ldlF cells stably expressing COP-GFP were lysed and the COPI complex was immunoprecipitated with a rabbit antibody against COP using protein A beads and standard procedures (Bonifacino, J. S. & Dell’Angelica, E. Immunoprecipitation in Current Protocols in Cell Biology, John Wiley & Sons, Inc., New York, NY, pp. 7.2.1-7.2.21; 1998). The lysate was mixed with beads a second time and immunoprecipitated using a rabbit antibody against GFP (Molecular Probes, Eugene, OR) to detect uncomplexed COP-GFP. The immunoprecipitates were analysed by SDS-PAGE and Western blot using the same anti-GFP antibody. Approximately 2/3 of the GFP signal (62% in one experiment) was immunoprecipitated with the antibody against COP indicating that most COP-GFP was stably associated with the COPI complex. These results are similar to those of Shima et al. (Shima, D. T., Scales, S. J., Kreis, T. E. & Pepperkok, R. Segregation of COPI-rich and anterograde-cargo-rich domains in endoplasmic-reticulum-to-Golgi transport complexes. Curr. Biol. 12, 821-824; 1999). Fig. 1e.movie Time-lapse sequence showing inward movement of peripheral eCOP-GFP-containing structures (from Fig. 1e). Fig. 2b.movie Time-lapse sequence showing photobleach and recovery of eCOP-GFP labeled Golgi (from Fig. 2b). Fig. 2d.movie Time-lapse sequence showing photobleach and recovery of eCOP-GFP labeled pre-Golgi intermediates (from Fig. 2d). Fig. 5c.movie Time-lapse sequence showing photobleach and recovery of eCOP-GFP labeled Golgi in cell held at 4º C (from Fig. 5c). Absence of free COPI-coated vesicles in cytoplasm If a 50-80 nm diameter coated vesicle labeled with eCOP-GFP is free in the cytoplasm, we should be able to detect it by fluorescence microscopy given a range of 50-200 molecules of coatomer per vesicle and the fact that each coatomer contains one eCOP-GFP. Microtubules are easily visible by light microscopy even though their diameter (~20 nm) is below the resolving capacity of the light microscope. Moreover, GFP-labeled nuclear pores (~60 nm diameter) (each containing 16-32 GFP molecules) have been detected using light microscopy on the nuclear envelope (Daigle et al., 2001) as well as clathrin-coated pits (<100 nm diameter) on the plasma membrane (Keen and colleagues). The fact that we do not observe small coated vesicles labeled with eCOP-GFP in the cytoplasm, therefore, suggests that putative COPI-coated vesicles would have to remain predominantly nearby (e.g., tethered to) Golgi and Golgi-derived membranes where they would not be resolved by fluorescence microscopy. Differences in FRAP recovery and BFA release rates for Arf1-GFP The faster Golgi dissociation rate of Arf1-GFP measured in our study compared to that found by Vasudevan et al. (1998) can be explained in part because their experiments were carried out at 22°C, whereas our experiments were carried out at 37°C. We know that Arf1 exchange on and off membranes slows down as a function of temperature (see Fig. 5d), so this could partly explain why their rate of recovery was slower than ours. A second explanation relates to the FRAP protocol used by Vasudevan et al. (1998) We monitored fluorescence recovery after bleaching at 5 sec intervals whereas Vaseduvan monitored recovery at 20 sec intervals. From our experience of analyzing FRAP experiments, we know it is essential to obtain recovery kinetics at early time points to accurately assess a recovery rate. For Arf1-GFP, we found that over 80% of its fluorescence had recovered by 20 sec after bleaching. Because Vasudevan et al., (1998) did not have recovery kinetics prior to that first 20 sec time-point, we believe their estimated exchange rate of Arf1-GFP on and off membranes was too slow. Vasudevan et al., (1998) also reported release kinetics by Arf1-GFP upon BFA treatment that were slower than what we observed. In these experiments, Vasudevan et al., (1998) added a concentrated aliquot of BFA to their cells, whereas we added a pre-mixed BFA solution. When we added a concentrated BFA aliquot rather than a pre-mixed BFA solution to our cells, we observed slowed Arf1-GFP release kinetics comparable to that reported by Vasudevan et al. Thus, we believe that the time taken for BFA to diffuse throughout the solution after being delivered as a concentrated aliquot explains why cells in Vasudevan et al.,’s experiment responded more slowly to BFA. Quantitative Golgi FRAP experiments Quantitative FRAP of eCOP-GFP on Golgi (Fig. 4d) was performed as follows: 1) the entire pool of non-Golgi-associated COPI was photobleached in order to eliminate any free GFP or uncomplexed eCOP-GFP, 2) the cell was given 30 min to allow complete re-equilibration of fluorescent Golgi and cytoplasmic pools, 3) an isolated Golgi fragment containing less than 10% of total Golgi fluorescence was selectively photobleached with high intensity laser light in order to minimize the effect of diffusion on the recovery kinetics, and 4) fluorescence in a region corresponding to the isolated fragment was quantitated with time as was total Golgi and total cytosolic fluorescence. Golgi/cytosol ratio was independent of expression level in all except the brightest 10% of eCOP-GFP expressing ldlF cells indicating that eCOP-GFP in excess of available COPI complex was not a factor except in the bright cells, which were never used in experiments. Recovery of Golgi-associated Arf1 fluorescence was not diffusion limited since: 1) D as measured by FRAP in the cytoplasm was greater than 20 µ2/sec and thus mixing occurred over an entire cell diameter over a timescale only 10% of the recovery time, and 2) identical recovery curves were measured whether the entire Golgi apparatus or small fragments were photobleached. Because Arf1[Q71L]-GFP expressed at high levels was able to protect COPI from dissociating from membranes in the presence of BFA. (see supplementary methods), addition of GFP to Arf1 did not abrogate its ability to interact with COPI. Cycling of Arf1-GFP on and off membranes was inhibited in cells microinjected with GTPgS. Coatomer release kinetics A prediction of the conventional model of sequential coatomer binding and release (Fig. 5a) relates to its requirement for a prebudding state in which COPI polymerizes to form a coat and is not released until a vesicle has budded, whereupon it is released en bloc. This could occur in two ways: 1) by rapid polymerization followed by en bloc release, or 2) by a slow multi-step polymerization process with en bloc release at the last step. We tested both of these mechanisms and found that neither could simultaneously account for both our data (Fig. 4D) and the known morphological distribution of COPI on nascent buds and vesicles. If the bud formation process is characterized by rapid polymerization of COPI followed by even faster en bloc release, then the observed slow exponential release kinetics (t1/2 = 30 sec) clearly do not reflect coat assembly time. Thus, in order to explain the observed slow exponential release of COPI, the release rate must be completely determined by the rate at which COPI is delivered to the prebudding state. If this were the case, however, the steady-state distribution of COPI would be predominantly in surrounding membrane, not in vesicle buds, contrary to the known ultrastructural distribution of COPI. If, instead, the coat formation process is viewed as a slow multistep assembly, then it becomes possible to have the steady-state distribution of COPI in nascent buds and vesicles as reported. However, the predicted kinetics of COPI release would now show a shoulder characteristic of release from the sequential coat assembly processes, which is inconsistent with our data showing exponential release of COPI. Thus, while qualitatively plausible, neither formulation of the en bloc release model is quantitatively consistent with the available data. Kinetic modeling of COPI/Arf1 Cycle The model described in Fig. 4C of the main text was formulated as a system of ordinary differential equations representing conservation of mass in the dynamic cellular chemical system illustrated. Using standard chemical kinetics, processes involving a single substrate (or input) molecule were treated as first order and were characterized by a single first-order rate constant. Second order processes, involving two input molecules, were assumed to proceed at a rate proportional to the product of the two molecular abundances. The constant of proportionality was the corresponding second order rate constant. Because cellular processes proceed at rates that depend on the probability of the corresponding molecular interaction, it is general practice to formulate rate laws in terms of the concentrations of the reactants. In cellular systems, however, the volumes of distribution are nearly always ill defined and we have consequently expressed all molecular abundances in units of molecules per cell. This, of course, requires that the second order rate constants have non-traditional units, but we and others (Weng, G., Bhalla, U. S. & Iyengar, R. Complexity in biological signaling systems. Science 284, 92-96; 1999) have found this to be a useful approach that obviates the need to estimate additional parameters. The model equations are as follows: d/dt(fARFg) = kgef*fARFcyto - kae*fARFg*effector - kac*fARFg*COPcyto - kbleachARF*fARFg d/dt(fARFeffector) = kae*fARFg*effector - kgtpase1*fARFeffector - kbleachARF*fARFeffector d/dt(fARFCOP) = kac*fARFg*COPcyto - kgtpase2*fARFCOP - kbleachARF*fARFCOP d/dt(fARFcyto) = kgtpase1*fARFeffector + kgtpase2*fARFCOP - kgef*fARFcyto d/dt(fARFexch) = kfxarf*fARFg - krxarf*fARFexch - kbleachARF*fARFexch d/dt(fCOPcyto) = kuncoat*fCOPXg - kac*ARFg*fCOPcyto d/dt(fCOPXg) = kgtpase2*ARFfCOP - kuncoat*fCOPXg -kbleachCOP*fCOPXg - kfexch*fCOPXg + krexch*fCOPexch d/dt(ARFfCOP) = kac*ARFg*fCOPcyto - kgtpase2*ARFfCOP - kbleachCOP*ARFfCOP d/dt(fCOPexch) = kfexch*fCOPXg - krexch*fCOPexch - kbleachCOP*fCOPexch where variables beginning with the letter f are the fluorescent, GFP-tagged molecules, ARFg represents Arf1-GTP on Golgi membranes, fARFeffector represents Arf1 complexes with non-COPI effector molecules in Golgi membranes, fARFCOP represents Arf1 in Arf1:COPI complexes in Golgi membranes, fARFcyto represents Arf1-GDP in cytoplasm, fARFexch represents Golgi Arf1-GTP in a small, slowly turning over exchange pool (see below), fCOPcyto represents free COPI in the cytoplasm, fCOPXg represents Golgi COPI bound to cargo molecules or other Golgi proteins, but unassociated with Arf1, ARFfCOP represents COPI in Arf1:COPI complexes, and fCOPexch represents COPI in a small slowly turning over Golgi exchange compartment perhaps bound to other Golgi proteins (see below). The remaining symbols in this system of equations are the rate constants governing the various chemical processes represented by the model. Specifically, kgef governs the combined processes of Arf1-GDP binding to Golgi membranes and subsequent activation by the action of an Arf1 GEF, kae governs the binding of Golgi Arf1-GTP to effectors other than COPI, kac governs binding of (sometimes referred to as recruitment) cytoplasmic COPI to Golgi Arf1-GTP, kbleachARF and kbleachCOP govern the bleaching of fluorescent Arf1 and COPI respectively by intense laser light, kgtpase1 governs the rate of GAP-mediated Arf1-GTP hydrolysis in Arf1 complexed to effectors other than COPI (this, we generally took to be equal to kgtpase2 because changes in kgtpase1 affect only the non-COPI subsystem and more data on other effector systems would be required to resolve it), kgtpase2 governs GTP hydrolysis in Arf1:COPI complexes, kfxarf and krxarf govern the fluxes into and out of the slow Arf1 exchange pool, kuncoat governs the release of COPI into the cytoplasm from Golgi protein or cargo complexes, kfexch and krexch govern fluxes into and out of the slow COPI exchange compartment. All the first order rate constants have units of sec-1. All the second order rate constants have units of (molecules/cell)-1 sec-1. Traditional biochemical values of second order rate constants have units of M-1 sec-1. This difference results from our decision to formulate the differential equations in terms of molecular abundance (molecules per cell), as described above. A corresponding system of equations was written for the non-fluorescent, native molecules. This was required because in the non-steady state induced by BFA, the abundances that constitute the chemical potentials will change and the differential equations will become nonlinear, even for the fluorescent tracer molecules. Thus, by including equations for the non-fluorescent molecules, it is possible to automatically account for the BFA-induced changes in effective rate constants, and to analyze the BFA data and the FRAP data simultaneously. This overall system of equations along with ancillary algebraic relationships defining species conservation and measured quantities, was simulated using the various numerical techniques implemented in Berkeley Madonna (www.berkeleymadonna.com). Both integration and optimization routines in this software package were taken from (Press, W. H., Teukolsky, S. A., Vetterling, W. T. & Flannery, B. P. Numerical recipes in C. The art of scientific computing. 2nd edition. Cambridge University Press, Cambridge, 1992.). In particular, we made extensive use of the Auto-stepsize integrator for model solution and the downhill simplex optimizer for parameter estimation and hypothesis testing. The reader will notice that two of the differential equations (those for fARFexch and fCOPexch) in the above system have no counterparts in the simplified model shown in the main text. These allow the model to account for small, very slow components of the BFA and FRAP kinetics for Arf1 and COPI that are consistently observed. The COPI exchange pool accounts for approximately 28% of the Golgi COPI, and the Arf1 exchange pool accounts for 12% of total Golgi Arf1. Though small, they represent new and potentially interesting features of Arf1 and COPI processing that are revealed by the data, but are, as yet, not understood in terms of their biological counterparts. The model was tested against our experimental data by direct comparison of model solutions to the Golgi Arf1-CFP and eCOP-YFP time-lapse measurements. Bleaching experiments were simulated by setting kbleachARF and kbleachCOP to non-zero values for the duration of the experimental bleach. BFA experiments were simulated by setting kgef to zero after the time of BFA addition. In addition to the BFA and FRAP kinetics, data on the observed abundances and steady-state distribution of Arf1 and COPI were fitted simultaneously. Western blot analysis of our cells revealed approximately 8 x105 total COPI molecules and 8 x 106 total Arf1 molecules per cell. The distribution data showed that at steady-state, 34% of the total cellular Arf1 is associated with the Golgi and 40% of the total cellular COPI is associated with the Golgi. These constraints are profoundly important for estimation of the relative fluxes of Arf1 binding to COPI and Arf1 binding to other effectors. Since these values may have a significant impact on the conclusions reached, and since precise measurements are only beginning to become available, we examined the sensitivity of our conclusions to these total abundances. This analysis is critical because kinetic detection of the COPI-related pool of Golgi Arf1 is facilitated if this Arf1 represents a significant portion of the total Golgi Arf1. Clearly, if 99% of the Golgi Arf1 participates in pathways that do not involve COPI, then the observed fluorescence signals are effectively decoupled; the kinetics of the COPI-associated Arf1 will have little bearing on the overall Golgi Arf1 kinetics and will thus be difficult to resolve. Two principal issues were explored: first, what changes in conclusions would result if total Arf1 were actually 15 times greater than total COPI? The factor 15 was chosen because it represents the largest ratio estimated in the literature. Significantly, little change in the key rate constants devolved from this much larger ratio. For example, kgef changed from 0.0124 to 0.0122 s-1, and kuncoat changed from 0.0258 to 0.0277 s-1. Fits of the FRAP and BFA data were comparable to those in Fig. 4B, and we therefore conclude that the information derived from these data sets is essentially independent of the measured Arf1:COPI ratio. Next we examined the sensitivity of our results to the assumption that the two GTPase rate constants are equal. When optimized independently, the two GTPase rate constants were 0.041 for the non-COPI effector pathway, and 0.032 s-1 for the COPI pathway. Adding this additional adjustable parameter, did not appreciably improve the FRAP and BFA fits as judged by the sums of squares. Some investigators may wish to treat these as independent parameters; others will agree that the additional parameter is unjustified by the small improvement in fit. We provide the single constrained value in the table below. To assist investigators who wish to run this model themselves, the following Table contains all the parameter values used to obtain the fits shown in the main text. The initial steady-state abundances of all the Arf1 and COPI compartments were estimated by setting all the derivatives to zero in the above system of equations, eliminating three of the equations by enforcing mass conservation, and then solving the resulting system of algebraic equations using the ROOTFINDER routines in Berkeley Madonna. Table 1. Rate constants and additional parameters used in the kinetic model. Symbol Value kgef 0.0127 s-1 kae 3.96 x 10-7 (molecules/cell)-1 s-1 kac 1.05 x 10-8 (molecules/cell)-1 s-1 kbleachARF 0.73 s-1 kbleachCOP 1.03 s-1 kgtpase 0.0384 s-1 kfxarf 0.0026 s-1 krxarf 0.0051 s-1 kuncoat 0.0231 s-1 kfexch 7.51 x 10-4 s-1 krexch 0.0012 s-1