|Home | About | Journals | Submit | Contact Us | Français|
During cytokinesis, regulatory signals are presumed to emanate from the mitotic spindle. However, what these signals are and how they lead to the spatiotemporal changes in the cortex structure, mechanics, and regional contractility are not well understood in any system.
To investigate pathways that link the microtubule network to the cortical changes that promote cytokinesis, we used chemical genetics in Dictyostelium to identify genetic suppressors of nocodazole, a microtubule depolymerizer. We identified 14-3-3 and found that it is enriched in the cortex, helps maintain steady state microtubule length, contributes to normal cortical tension, modulates actin wave formation, and controls the symmetry and kinetics of cleavage furrow contractility during cytokinesis. Furthermore, 14-3-3 acts downstream of a Rac small GTPase (RacE), associates with myosin II heavy chain and is needed to promote myosin II bipolar thick filament remodeling.
14-3-3 connects microtubules, Rac and myosin II to control several aspects of cortical dynamics, mechanics, and cytokinesis cell shape change. Further, 14-3-3 interacts directly with myosin II heavy chain to promote bipolar thick filament remodeling and distribution. Overall, 14-3-3 appears to integrate several critical cytoskeletal elements that drive two important processes cytokinesis shape change and cell mechanics.
Cytokinesis is driven by regional mechanical activities – myosin II-based contractility, actin polymer dynamics, and actin crosslinking. Over the years, we have discovered and have been studying a two-component system of equatorial and global/polar actin-associated proteins that govern these dynamical and mechanical features of the dividing cell cortex. The global pathway is controlled by a Rac-family small GTPase (encoded by the RacE gene), which regulates the distribution of cortical actin crosslinkers and provides resistive stresses to modulate furrow ingression kinetics [1–3]. However, while active cell shape change is ultimately driven and controlled by these cortical mechanical features, the process is regulated spatially and temporally by the mitotic spindle.
The mitotic spindle has two major structures - the central spindle and the astral microtubule network – that regulate the cortex (e.g. [4–6]); however, the molecular bases of these signaling cascades are not well defined. The central spindle signals are mediated partly by kinesin-6-family motors complexed with two general sets of proteins, including MgcRacGap (in metazoans – the centralspindlin complex) or the chromosomal passenger proteins INCENP and Aurora kinase (in Dictyostelium protozoans and higher metazoans) [7–10]. The astral microtubules play an important role in many cell-types, including Dictyostelium, echinoderms and C. elegans embryos where they help direct symmetry breaking [4, 11, 12]. The major signaling pathway activated by these astral microtubules has remained largely elusive though a heterotrimeric G-protein pathway has been implicated in some systems (e.g. ). Yet, Dictyostelium Gβ null mutants (Dictyostelium Gβ is encoded by a single gene) are primarily defective in chemotaxis-assisted scission . Overall, significant gaps remain in our understanding of how the mitotic spindle regulates the regional mechanics that drive cytokinesis cell shape change.
Here, we used nocodazole to disturb microtubules and searched for genes involved in microtubule-related regulatory pathways using cDNA library suppression analysis. Using this chemical-genetic approach, we discovered 14-3-3 over-expression as a genetic suppressor of nocodazole. The 14-3-3 proteins are a family of abundant proteins, which are widely expressed in all eukaryotic cells and are highly conserved from Dictyostelium to mammals. 14-3-3 has been implicated in cytokinesis completion, and the mammalian isoform 14-3-3σ is involved in mitotic translational control [15, 16]. Two isoforms are found in yeast, Drosophila and C. elegans. Dictyostelium cells have only one isoform, which makes Dictyostelium a unique system for 14-3-3 studies. Here, we present evidence that 14-3-3 functions in a pathway linking microtubules, RacE, and myosin II to cortex mechanics, cortical actin wave dynamics, and cytokinesis shape control.
To begin dissecting how the mitotic spindle modulates cytokinesis, we challenged pools of cDNA library transformed cells with nocodazole at the IC50 concentration, the concentration that reduced the growth rate by 50% (measured to be 10 μM). From 100 pools of 1000 cDNA library transformants (100,000 total transformants), we recovered 14-3-3 (sequence analysis in Fig. S1) and a dominant-negative version of enlazin (enl-tr), the Dictyostelium ezrin-radixin-moesin (ERM)-family protein . Both 14-3-3 and enl-tr recapitulated the nocodazole-suppression (Fig. 1A), verifying the genetic interactions with nocodazole. Because enl-tr was originally identified as a genetic suppressor of cortexillin-I mutants , the recovery of 14-3-3 and enl-tr in this genetic selection led us to ask whether 14-3-3 links the microtubule network to the actin cortex to modulate cytokinesis contractility.
To initiate characterization of 14-3-3, we purified a recombinant His-tagged 14-3-3 and generated polyclonal antibodies. In the 14-3-3 overexpression cell-lines, 14-3-3 levels were increased approximately 2-fold relative to control cells (Fig. 1B, C). We also quantified the wild type cellular concentration of 14-3-3, which was 1.4 μM monomer, corresponding to a 0.7 μM dimer (the typical functional unit of 14-3-3) concentration (Fig 1D). This concentration is similar to that found for other actin-associated proteins, including dynacortin (1 μM) , fimbrin (0.6 μM) , and myosin II (3.4 μM) .
We attempted to delete 14-3-3 by homologous recombination. However, this effort yielded highly enlarged, fragile cells (typical of multinucleated cells with a severe cortical defect), which died within a few generations (Fig. S2A–C). Therefore, we down-regulated 14-3-3 expression levels using a hairpin construct (14-3-3hp). Our data confirmed that 14-3-3 is an essential gene as complete silencing by the hairpin similarly killed the cells under standard drug conditions. However, we established a protocol for achieving partial RNA interference, allowing the recovery of wt:14-3-3hp cells with a 60-70% knockdown of both RNA and protein levels (Fig 1B, C). These cells had a slower growth rate in suspension culture (Fig. S2D) and a larger overall cell size (Fig. S2E) as compared with control. DAPI staining revealed that these enlarged 14-3-3hp cells are multinucleated, indicating a cytokinesis defect (Fig. S2E, F). The 14-3-3 levels were inversely correlated with the degree of multinucleation (severity of the cytokinesis defect) (Fig. S2G).
To assess how 14-3-3 contributes to cytokinesis contractility, we quantified the furrow ingression dynamics using our rescaling scheme [2, 3, 17]. In this scheme, wild type cleavage furrows ingress following a slow, nonlinear trajectory. In contrast, mutants in either the global or equatorial pathways divide following a multiphasic trajectory in which the early stage is slow and the late phase accelerates relative to the initial phase. As compared to wild type cells, wt: 14-3-3hp cells have an altered furrow morphology (Fig. 1E) and a multiphasic furrow ingression dynamic (Fig. 1F, G). The DIC time-lapse and the rescaled furrow-thinning dynamics show that down-regulation of 14-3-3 slowed down the initial bridge ingression. Finally, these 14-3-3hp cells also exhibited an elongated bridge length and have a higher probability of asymmetrical division as reflected by the area ratio of the two daughter cells (Fig 1E, H). To summarize, 14-3-3 ensures symmetrical cell division and wild type furrow ingression dynamics. It is worth noting that the morphology and dynamics of the 14-3-3hp cytokinesis is highly reminiscent of myoII null cytokinesis [2, 3].
Next, we analyzed the sub-cellular localization of 14-3-3 by immunocytochemistry and by live-cell imaging of cells expressing 14-3-3-GFP. In fixed cells, 14-3-3 concentrated in the cortex of wild type and 14-3-3 over-expressed cells, but was depleted from the cortex in 14-3-3hp cells (Fig 2A). Similarly, 14-3-3-GFP showed cortical enrichment that is distinct from soluble GFP in live cells (Fig. S2H). Mutation (K49E; ) of the key conserved lysine in 14-3-3’s ligand-binding cleft partially reduced the cortical enrichment of 14-3-3, indicating that the cortical enrichment is dependent on ligand binding (Fig. S2H). During cytokinesis, 14-3-3 was enriched in the global/polar cortex in fixed (Fig. 2B) and live cells (Fig. S2I). We did not see clear co-localization of 14-3-3 with microtubules by either method. To ensure that the 14-3-3-GFP was functional, we confirmed its ability to rescue nocodazole treatment, which was similar to the untagged version (Fig. 1A).
Because 14-3-3 was recovered as a genetic suppressor of nocodazole, we examined the microtubule structures in control cells and in cells with altered 14-3-3 expression levels with or without nocodazole (Fig. 2C). To simplify the analysis, we examined the total microtubule length by epi-fluorescence microscopy when cells were flattened with a sheet of agarose (Fig. 2C). We found that the average microtubule was ~30% shorter in 14-3-3hp cells whereas 14-3-3OE did not significantly alter the average microtubule length. Upon nocodazole treatment, the average length decreased from ~10 μm to 2 μm in control cells, whereas 14-3-3OE cells had longer microtubules than the control cells had.
As 14-3-3 was enriched in the cortex, we sought to determine if alteration of 14-3-3 levels specifically changed the extent and dynamics of microtubule-cortex interactions. Here, we define the microtubule-cortex interaction site as the region of the microtubule detected by total internal reflection fluorescence (TIRF) microscopy, which only illuminates ~200 nm from the cell-surface interface (Fig. S3A). We developed a semi-automated microtubule-tracking algorithm to quantify the microtubule-cortex interaction length and lifetime. We found that the lifetimes of the microtubule-cortex interactions were similar (τ≈5s) for all genotypes (wt control, wt: 14-3-3OE, and wt: 14-3-3hp) with or without nocodazole (data not shown). We also analyzed the microtubule-cortex contact length distributions (Fig. S3B). We found that without nocodazole, perturbation of 14-3-3 levels significantly altered these distributions, whereas with nocodazole, the data interpretation is complicated by the fact that nocodazole-treated cells were more photosensitive.
Because 14-3-3hp cells had a cytokinesis defect (Fig. 1) and an interphase microtubule structural defect (Fig. 2C), we needed to determine if there was a defect in mitotic spindle-associated microtubules. However, it is difficult to detect all astral microtubules making contact with the cortex. Instead, we analyzed the mitotic spindle dynamics, specifically the spindle length change and angular velocity. We could not find any quantitative impact on the mitotic spindle dynamics upon 14-3-3 depletion (Fig. S3C).
The cortical localization of 14-3-3 led us to test whether it is involved in cortical mechanics. We measured an effective cortical tension, using micropipette aspiration (MPA). When wild type cells were treated with nocodazole (10 μM), the cortical tension was reduced by 60% (Fig. 2D). Over-expression of 14-3-3 partially rescued the cortical tension defect of nocodazole-treated cells (raising it from 0.4 to 0.8 nN/μm; Fig. 2D). In contrast, 14-3-3hp cells had a 40% reduction in interphase cortical tension (Fig. 2E), and expression of the ligand-binding mutant 14-3-3(K49E) had a dominant-negative effect on cortical tension, reducing it by 30% compared to wt control (Fig. 2F). These observations suggest that 14-3-3 works with microtubules to control cortical mechanics.
RacE is important for cytokinesis, cortical tension, and the cortical accumulation of the global actin crosslinkers dynacortin and coronin [1, 22]. Though dynacortin and coronin are downstream of RacE, the pathways governed by RacE are largely unknown. Because 14-3-3 also controls cortical tension and is globally distributed during cytokinesis, we tested whether 14-3-3 participates in the RacE signaling pathway. 14-3-3-GFP was expressed in RacE null cells and the localization of 14-3-3-GFP was examined. Without RacE, 14-3-3-GFP failed to accumulate in the cortex (Fig. 3A). However, when the cells were complemented with mCherry-RacE, the cortical localization of 14-3-3 was restored (Fig 3A). To begin to characterize this RacE dependency of 14-3-3’s cortical localization, we tested 14-3-3’s solubility. We lysed cells with 0.1% NP-40 and found that a greater fraction (~30% vs. <~10%) of 14-3-3 sedimented in the pellet fraction when RacE was expressed in the cell and GTP was provided in the lysis buffer (Fig. 3B). Stronger detergents such as 0.1% Triton-X-100 disrupted this GTP- and RacE-dependent fraction, suggesting that the insoluble fraction has a membrane component rather than being purely cytoskeleton-associated. Furthermore, over-expression of 14-3-3 partially rescued the cortical tension defect in RacE null cells (Fig. 2E). RacE null cells have a cortical tension that is 30% of wild type, but when 14-3-3 is over-expressed, the cortical tension was nearly doubled to 55% of wild type. Together, these findings suggest that 14-3-3 participates in the signaling pathway from RacE to the cortical cytoskeleton: 14-3-3 depends on RacE for its cortical localization and its separation into the NP40-insoluble fraction, and over-expression of 14-3-3 partially rescues the mechanical defect in RacE null cells.
Because RacE and 14-3-3 mutants have severe mechanical defects, we examined the actin network of these cells. Here, we expressed GFP-tagged-LimEΔcoil to identify polymeric actin in wild type or wt:14-3-3hp cells and discovered that 14-3-3 modulates cortical actin wave formation (Fig. S3D–H; Table S2). Actin waves are dynamic actin assemblies that form at the cell-substrate interface and propagate along the surface in a manner dependent on Arp2/3 and myosin I, but not myosin II . 14-3-3-GFP itself did not show wave behavior (data not shown). RacE mutants also showed a suppressed cortical actin activity in the same assay though the qualitative features are slightly different in this strain background (data not shown). In combination, these observations indicate that 14-3-3 modulates cortex mechanics downstream of RacE, which may then have an indirect effect on cortical actin wave dynamics.
We next tested whether 14-3-3 can rescue the RacE null cytokinesis and suspension growth defects. As compared to RacE null cells, the cytokinesis morphology was qualitatively rescued (Fig. 3C), and the furrow ingression dynamics were quantitatively slowed down in RacE:14-3-3OE cells, making them more wild type-like (Fig. 3D). Furthermore, 14-3-3 over-expression partially rescued the suspension growth defect of RacE null cells (Fig. 3E). This genetic suppression is specific to RacE null cells since it failed to suppress cortI null (Fig 3F), myoII null and kif12 null cells (kif12 encodes the Dictyostelium kinesin-6-family protein ), all of which are cell-lines with severe cytokinesis defects in suspension culture (data not shown). Thus, 14-3-3 specifically acts in a pathway, which links the microtubule network and RacE and controls cell mechanics and cytokinesis shape progression.
To determine how 14-3-3 controls cytokinesis and cortical mechanics, we used high resolution mass spectrometry to identify novel 14-3-3-binding partners. We expressed a 14-3-3-FLAG protein in cells and isolated associated proteins by anti-FLAG co-immunoprecipitation (Fig. S4A). 14-3-3-FLAG-associated proteins were identified using database searching and prioritized using a spectral counting-based protein quantification and statistical analysis strategy (see Supplemental Experimental Procedures). From this, a list of potential 14-3-3 interactors was assembled (Fig. S4B). As expected, the most abundant protein identified from the anti-FLAG immunoprecipitation was 14-3-3. The third most abundant protein was myosin II heavy chain protein, and other proteins of interest included fimbrin and myosin II heavy chain kinase. Given the roles of 14-3-3 and myosin II in cortical mechanics and cytokinesis, we chose to focus on their interactions first. Using antibodies directed against myosin II heavy chain and 14-3-3, we tested whether endogenous myosin II and 14-3-3 do, in fact, associate (Fig. 4A, B). These endogenous proteins could be reciprocally co-immunoprecipitated (in case the proteins associated indirectly through actin, Mg2+ and ATP were included in the buffers to prevent myosin II from locking onto actin). Using sequence analysis, we found several motif 2 and 3, but no motif 1, consensus 14-3-3 binding sites in the myosin II heavy chain peptide (Fig. S4C). Three of these binding sites overlapped or were adjacent to three critical threonines in the myosin II heavy chain tail that regulate myosin II bipolar thick filament (BTF) assembly . The BTF assembly incompetent phosphomimic myosin II 3xAsp (where the three threonines have been substituted with aspartic acids) failed to co-immunoprecipitate 14-3-3. Therefore, 14-3-3 likely only associates with assembly-competent myosin II.
To determine if this association is functionally relevant, we examined the distribution of GFP-myosin II (expressed in a myoII null background) and found that GFP-myosin II distribution was severely altered in 14-3-3hp cells (Fig. 4C). In particular, myosin II was significantly less uniform, even punctated in 14-3-3hp cells. During cytokinesis, GFP-myosin II could still accumulate at the furrow but did so in a highly aggregated fashion (Fig. 4C). In contrast, 14-3-3 overexpression did not appear to alter the wild type GFP-myosin II distribution (Fig. 4C). Finally, 14-3-3 required myosin II for it to accumulate in the cortex (data not shown).
To determine how 14-3-3 impacts BTF assembly dynamics, we performed fluorescence recovery after photobleaching (FRAP) analysis of GFP-myosin II in the cleavage furrow and interphase cortices (Fig. 4D; Fig. S4D). For all genotypes (wt control, wt:14-3-3OE, and wt:14-3-3hp), GFP-myosin II had a recovery time (τrec) of 3-5 s in interphase and cleavage furrow cortices. However, the immobile fraction was significantly increased in the 14-3-3hp cells in both the cleavage furrow and the interphase cortex (Fig. 4D; Fig. S4D). Overall, we found a general trend in which the immobile fractions increased such that 14-3-3OE < control < 14-3-3hp. These observations indicate that 14-3-3 associates with assembly-competent myosin II, increasing its mobility so that it can remodel along the interphase and cleavage furrow cortices.
To test if 14-3-3 directly impacts myosin II assembly, we performed BTF assembly assays using purified myosin II from Dictyostelium and recombinant Dictyostelium 14-3-3-His purified from E. coli. BTF assembly is highly salt dependent with the optimal concentrations for myosin II BTF assembly occurring in the range of 25 to 100 mM NaCl . Using equimolar (250 nM) 14-3-3 dimer and myosin II hexameric monomer (two heavy chains, two regulatory light chains, and two essential light chains), we measured the percentage of soluble myosin II as a function of salt concentration (Fig. 4E). Under these conditions, 20% of the myosin II remains soluble due to the critical concentration of myosin II in the range of 25–100 mM salt. We found that 14-3-3 increased the fraction of soluble myosin slightly from ~20% to ~30%. This level of soluble myosin II is similar to that measured for the 1xAsp and 2xAsp mutant myosins, rather than the full 3xAsp assembly-incompetent myosin II . We could not detect appreciable increases in the 14-3-3 levels in the insoluble assembled fraction most likely because 14-3-3 interacts in a highly sub-stoichiometric manner with the myosin II monomers in the BTFs. To compare the in vivo FRAP and in vitro BTF assembly results, 14-3-3 shifts myosin II out of the assembled state in vitro and increases the mobile fraction in vivo in order to allow the myosin II to remodel and redistribute around the cell cortex. Because 14-3-3 promotes myosin II mobility and both proteins contribute to cortical mechanics, we asked whether 14-3-3 required myosin II to have an impact on cortical tension. However, 14-3-3OE did not have an effect on the cortical tension of myoII null cells (Fig. 2E). Thus, 14-3-3 works through myosin II to control cortical tension.
Cytokinesis shape change occurs over a fast five-minute time-span, requiring considerable cortical remodeling during this time-frame. Much insight has come from studying the myosin heavy chain kinases, which phosphorylate the critical threonines in the tail of myosin II heavy chain to control bipolar thick filament (BTF) assembly [25, 27]. Most likely what these enzymes do is set the level of the free pool of myosin II monomers (~80%) in the cell. This free pool then is maintained through flux of the assembled and disassembled BTF states and is required to ensure that the contractile network is turned over during processes like motility and cytokinesis. 14-3-3 also impacts this dynamic, preventing the myosin II from forming larger order assemblies and shifting the assembly equilibrium. Importantly, the cellular concentrations of myosin II in the BTF state and the 14-3-3 dimer are identical (0.7 μM), indicating that hypotheses of direct biochemical modulation of BTF assembly by 14-3-3 are reasonable. Four scenarios may be envisioned for how 14-3-3 might modulate BTF dynamics. First, by binding to unphosphorylated monomers, 14-3-3 may suppress myosin II thick filament nucleation and/or elongation so that fewer well structured BTFs are assembled. Second, 14-3-3 could bind transiently to the mature BTF, lowering individual monomer affinity and helping it to release from the BTF. Third, the in vivo system may be more complex where 14-3-3 binds to the BTF, putting the myosin monomers in a configuration that makes them more ideal substrates for myosin heavy chain kinases (MHCK). As MHCK was identified as a potential 14-3-3 protein interactor in our mass spectrometric analysis, such a tertiary complex of myosin BTF-MHCK-14-3-3 might be formed. Fourth, 14-3-3 might form part of the myosin II BTF cortical receptor/anchoring complex. In no system is it well established how myosin II BTFs are anchored to the cortex. In dividing 14-3-3hp cells, the myosin II aggregates appear to dislodge from the cortex, suggesting 14-3-3 might contribute to cortical anchoring. Overall, this system is highly dynamic where all of the measured time-scales occur on the low seconds. The data also demonstrate that if myosin II BTFs are not distributed uniformly, they cannot contribute to cortex mechanics and therefore the 14-3-3hp cortical tension and cytokinesis morphology closely resembles that of the myoII null cells. These observations may prove to be generalized to other systems as mammalian myosin II appeared on the list of potential 14-3-3σ interactors in a proteomics study . Furthermore, 14-3-3 may contribute to myosin II activation in many ways as 14-3-3 has been implicated in regulating mammalian myosin light chain phosphatase .
14-3-3 also functions downstream of RacE: its cortical localization depends on RacE, its solubility depends on RacE and GTP, and 14-3-3 overexpression in RacE nulls partially rescues the growth, cortical tension and cytokinesis defects. Though we were unable at this point to demonstrate direct binding between 14-3-3 and RacE, these data demonstrate that 14-3-3 links at least indirectly to RacE. These data also suggest that the cortical localization of 14-3-3 reflects where active RacE is found. RacE is uniformly distributed around the plasma membrane during cell division  and may regulate one or more of the global/polar-module proteins . Significantly, 14-3-3 enriches all along the cell cortex, but is reduced in the cleavage furrow region. Since myosin II was still able to accumulate in the cleavage furrow cortex in the 14-3-3hp cells, active RacE may promote 14-3-3 accumulation in the polar cortex where it helps remodel the myosin II BTFs.
14-3-3 may also modify the function of several other cortical actin-associated proteins such as dynacortin, fimbrin, coronin, enlazin and LimE . Several of these proteins contain predicted consensus 14-3-3 binding motifs and are found in comparable concentrations as 14-3-3. Fimbrin appeared on the list of potential interactors (Fig. S4B); however, we have been unable so far to recapitulate the fimbrin-14-3-3 interaction using standard co-immunoprecipitation assays. This is possibly due to fimbrin’s short cortical association time (260-ms) . Interestingly, 14-3-3 proteins often bind to phosphorylated ligands, and a number of phosphorylated proteins reside in the cortex (e.g. dynacortin is a phosphoprotein ). Similarly, sea urchin embryos accumulate multiple cortical phosphoproteins during cell division, and this accumulation is inhibited by nocodazole-treatment .
Finally, microtubules contribute to cortical mechanics, probably through a signaling pathway that includes 14-3-3. In return, 14-3-3 contributes to the steady-state microtubule structures and modulates cortical mechanics through RacE and myosin II (Fig. 4F). Therefore, this is not a linear pathway, but what appears to be a circular (feedback) system between the microtubules and the cortex. The combination of these observations then begs the question whether the manner in which the spindle regulates the cortex is really through a feedback system where the spindle directs the cortex and the cortex directs the spindle . In this case, cytokinesis symmetry breaking may occur through bidirectional communication between these two structures. Similar ideas have been suggested elsewhere [33, 34]. Overall, 14-3-3 coordinates three major cytoskeletal elements – microtubules, actin, and myosin II – to control two critical processes, cytokinesis fidelity and cortical mechanics.
Experimental procedures for Dictyostelium strains, cell culture and plasmids, cDNA library screening, cell growth in suspension and growth rate determination, RNAi analysis, DAPI staining and nuclei/cell ratio determination, antibody production and Western analysis, immunocytochemistry, cytokinesis analysis, microtubule analysis, actin wave analysis, effective cortical tension measurements, anti-FLAG co-immunoprecipitation, mass spectrometry analysis, 14-3-3-myosin II co-immunoprecipitation, FRAP analysis, and in vitro bipolar thick filament assembly may be found in the Supplemental Data. For statistical analysis, significance was determined using a two-tailed Student’s t-test (ST), Mann-Whitney nonparametric test (MW), or by comparing proportions (CP) where the standard error is defined as SE= √(f(1−f)/n) where f is the fraction of cells showing a behavior and n is the sample size. The p-values are provided for each test, and for CP, the z-values are also provided.
We thank the members of the Robinson lab for helpful comments on the manuscript and Cathy Kabacoff, in particular, for help with generating cell-lines. This work was supported by an American Heart Fellowship (to Q.Z.), NIH grant GM066817 (to D.N.R.), ACS grant RSG CCG-114122 (to D.N.R.), NSF grant CCF 0621740 (to P.A.I. and D.N.R.), NIH grant GM86704 (to P.A.I. and D.N.R.), and NIH grant GM082834 (to U.S.E).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.