Thoracic Aortic Aneurysm (TAAD)-causing Mutation in Actin Affects Formin Regulation of Polymerization

More than 30 mutations in ACTA2, which encodes α-smooth muscle actin, have been identified to cause autosomal dominant thoracic aortic aneurysm and dissection. The mutation R256H is of particular interest because it also causes patent ductus arteriosus and moyamoya disease. R256H is one of the more prevalent mutations and, based on its molecular location near the strand-strand interface in the actin filament, may affect F-actin stability. To understand the molecular ramifications of the R256H mutation, we generated Saccharomyces cerevisiae yeast cells expressing only R256H yeast actin as a model system. These cells displayed abnormal cytoskeletal morphology and increased sensitivity to latrunculin A. After cable disassembly induced by transient exposure to latrunculin A, mutant cells were delayed in reestablishing the actin cytoskeleton. In vitro, mutant actin exhibited a higher than normal critical concentration and a delayed nucleation. Consequently, we investigated regulation of mutant actin by formin, a potent facilitator of nucleation and a protein needed for normal vascular smooth muscle cell development. Mutant actin polymerization was inhibited by the FH1-FH2 fragment of the yeast formin, Bni1. This fragment strongly capped the filament rather than facilitating polymerization. Interestingly, phalloidin or the presence of wild type actin reversed the strong capping behavior of Bni1. Together, the data suggest that the R256H actin mutation alters filament conformation resulting in filament instability and misregulation by formin. These biochemical effects may contribute to abnormal histology identified in diseased arterial samples from affected patients.     

An intermediate form of ADP-F-actin

With yeast actin, contrary to other actins, filament formation, ATP hydrolysis, and Pi release are concurrent at low actin concentrations, the condition usually employed to assess actin polymerization. This observation leads to a question concerning the conformation of the filament barbed end that might be recognized by specific actin-binding proteins. To try to detect possible new actin polymer conformations that might be intermediate in the pathway leading to mature F-actin, we monitored the change in intrinsic tryptophan fluorescence of yeast and muscle actins polymerized at pH 6 to accelerate the rate of filament formation. This allowed temporal resolution of the Pi release process from the slower process of polymerization. With both actins, we detected a biphasic instead of the usual monophasic fluorescence change, a rapid decrease that tracks with filament formation followed by a slower rebound (the second phase). This second phase postpolymerization conformational change requires Pi release and occurs nearly coincident with its release. The addition of Pi causes this second phase response to disappear, and the inclusion of Pi during polymerization prevents its appearance. At pH 7.5, with higher yeast actin concentrations to accelerate polymerization, a two-phase fluorescence change is also observed. In this case, the second phase change lags substantially behind Pi release. Pi release could also be resolved from polymer formation. V159N yeast actin, hypothesized previously as remaining in a postpolymerization ATP-like state, exhibits the same two-phase intrinsic tryptophan fluorescence behavior as wild-type yeast actin. Together, these observations demonstrate the presence of an intermediate filament state between ADP-Pi and mature ADP-F-actin.     

Talin binds to actin and promotes filament nucleation

Platelet talin binds to actin in vitro and hence is an actin binding protein. By four different non-interfering assay conditions (fluorescence, fluorescence recovery after photobleaching, (FRAP), dynamic light scattering and DNase-I inhibition) we show that talin promotes filament nucleation, raises the filament number concentration and increases the net rate of actin polymerization but has no inhibitory effect on filament elongation. Binding of talin to actin occurs at a maximal molar ratio of 1:3 as determined by fluorescencetitration under G-buffer conditions. The overall binding constant was approximately 0.25 microM.     

Actin hydrophobic loop 262-274 and filament nucleation and elongation

The importance of actin hydrophobic loop 262-274 dynamics to actin polymerization and filament stability has been shown recently with the use of the yeast mutant actin L180C/L269C/C374A, in which the hydrophobic loop could be locked in a “parked” conformation by a disulfide bond between C180 and C269. Such a cross-linked globular actin monomer does not form filaments, suggesting nucleation and/or elongation inhibition. To determine the role of loop dynamics in filament nucleation and/or elongation, we studied the polymerization of the cross-linked actin in the presence of cofilin, to assist with actin nucleation, and with phalloidin, to stabilize the elongating filament segments. We demonstrate here that together, but not individually, phalloidin and cofilin co-rescue the polymerization of cross-linked actin. The polymerization was also rescued by filament seeds added together with phalloidin but not with cofilin. Thus, loop immobilization via cross-linking inhibits both filament nucleation and elongation. Nevertheless, the conformational changes needed to catalyze ATP hydrolysis by actin occur in the cross-linked actin. When actin filaments are fully decorated by cofilin, the helical twist of filamentous actin (F-actin) changes by ∼ 5° per subunit. Electron microscopic analysis of filaments rescued by cofilin and phalloidin revealed a dense contact between opposite strands in F-actin and a change of twist by ∼ 1° per subunit, indicating either partial or disordered attachment of cofilin to F-actin and/or competition between cofilin and phalloidin to alter F-actin symmetry. Our findings show an importance of the hydrophobic loop conformational dynamics in both actin nucleation and elongation and reveal that the inhibition of these two steps in the cross-linked actin can be relieved by appropriate factors.     

The C terminus of formin FMNL3 accelerates actin polymerization and contains a WH2 domain-like sequence that binds both monomers and filament barbed ends

Formin proteins are actin assembly factors that accelerate filament nucleation then remain on the elongating barbed end and modulate filament elongation. The formin homology 2 (FH2) domain is central to these activities, but recent work has suggested that additional sequences enhance FH2 domain function. Here we show that the C-terminal 76 amino acids of the formin FMNL3 have a dramatic effect on the ability of the FH2 domain to accelerate actin assembly. This C-terminal region contains a WASp homology 2 (WH2)-like sequence that binds actin monomers in a manner that is competitive with other WH2 domains and with profilin. In addition, the C terminus binds filament barbed ends. As a monomer, the FMNL3 C terminus inhibits actin polymerization and slows barbed end elongation with moderate affinity. As a dimer, the C terminus accelerates actin polymerization from monomers and displays high affinity inhibition of barbed end elongation. These properties are not common to all formin C termini, as those of mDia1 and INF2 do not behave similarly. Interestingly, mutation of two aliphatic residues, which blocks high affinity actin binding by the WH2-like sequence, has no effect on the ability of the C terminus to enhance FH2-mediated polymerization. However, mutation of three successive basic residues at the C terminus of the WH2-like sequence compromises polymerization enhancement. These results illustrate that the C termini of formins are highly diverse in their interactions with actin.     

Individual actin filaments in a microfluidic flow reveal the mechanism of ATP hydrolysis and give insight into the properties of profilin

The hydrolysis of ATP associated with actin and profilin-actin polymerization is pivotal in cell motility. It is at the origin of treadmilling of actin filaments and controls their dynamics and mechanical properties, as well as their interactions with regulatory proteins. The slow release of inorganic phosphate (Pi) that follows rapid cleavage of ATP gamma phosphate is linked to an increase in the rate of filament disassembly. The mechanism of Pi release in actin filaments has remained elusive for over 20 years. Here, we developed a microfluidic setup to accurately monitor the depolymerization of individual filaments and determine their local ADP-Pi content. We demonstrate that Pi release in the filament is not a vectorial but a random process with a half-time of 102 seconds, irrespective of whether the filament is assembled from actin or profilin-actin. Pi release from the depolymerizing barbed end is faster (half-time of 0.39 seconds) and further accelerated by profilin. Profilin accelerates the depolymerization of both ADP- and ADP-Pi-F-actin. Altogether, our data show that during elongation from profilin-actin, the dissociation of profilin from the growing barbed end is not coupled to Pi release or to ATP cleavage on the terminal subunit. These results emphasize the potential of microfluidics in elucidating actin regulation at the scale of individual filaments.     

Subdomain location of mutations in cardiac actin correlate with type of functional change

Determining the molecular mechanisms that lead to the development of heart failure will help us gain better insight into the most costly health problem in the Western world. To understand the roles that the actin protein plays in the development of heart failure, we have taken a systematic approach toward characterizing human cardiac actin mutants that have been associated with either hypertrophic or dilated cardiomyopathy. Seven known cardiac actin mutants were expressed in a baculovirus system, and their intrinsic properties were studied. In general, the changes to the properties of the actin proteins themselves were subtle. The R312H variant exhibited reduced stability, with a T(m) of 53.6 °C compared to 56.8 °C for WT actin, accompanied with increased polymerization critical concentration and Pi release rate, and a marked increase in nucleotide release rates. Substitution of methionine for leucine at amino acid 305 showed no impact on the stability, nucleotide release rates, or DNase-I inhibition ability of the actin monomer; however, during polymerization, a 2-fold increase in Pi release was observed. Increases to both the T(m) and DNase-I inhibition activity suggested interactions between E99K actin molecules under monomer-promoting conditions. Y166C actin had a higher critical concentration resulting in a lower Pi release rate due to reduced filament-forming potential. The locations of mutations on the ACTC protein correlated with the molecular effects; in general, mutations in subdomain 3 affected the stability of the ACTC protein or affect the polymerization of actin filaments, while mutations in subdomains 1 and 4 more likely affect protein-protein interactions.     

Nucleation limits the lengths of actin filaments assembled by formin

Formins stimulate actin polymerization by promoting both filament nucleation and elongation. Because nucleation and elongation draw upon a common pool of actin monomers, the rate at which each reaction proceeds influences the other. This interdependent mechanism determines the number of filaments assembled over the course of a polymerization reaction, as well as their equilibrium lengths. In this study, we used kinetic modeling and in vitro polymerization reactions to dissect the contributions of filament nucleation and elongation to the process of formin-mediated actin assembly. We found that the rates of nucleation and elongation evolve over the course of a polymerization reaction. The period over which each process occurs is a key determinant of the total number of filaments that are assembled, as well as their average lengths at equilibrium. Inclusion of formin in polymerization reactions speeds filament nucleation, thus increasing the number and shortening the lengths of filaments that are assembled over the course of the reaction. Modulation of the elongation rate produces modest changes in the equilibrium lengths of formin-bound filaments. However, the dependence of filament length on the elongation rate is limited by the number of filament ends generated via formin’s nucleation activity. Sustained elongation of small numbers of formin-bound filaments, therefore, requires inhibition of nucleation via monomer sequestration and a low concentration of activated formin. Our results underscore the mechanistic advantage for keeping formin’s nucleation efficiency relatively low in cells, where unregulated actin assembly would produce deleterious effects on cytoskeletal dynamics. Under these conditions, differences in the elongation rates mediated by formin isoforms are most likely to impact the kinetics of actin assembly.     

Inorganic phosphate regulates the binding of cofilin to actin filaments

Inorganic phosphate (Pi) and cofilin/actin depolymerizing factor proteins have opposite effects on actin filament structure and dynamics. Pi stabilizes the subdomain 2 in F-actin and decreases the critical concentration for actin polymerization. Conversely, cofilin enhances disorder in subdomain 2, increases the critical concentration, and accelerates actin treadmilling. Here, we report that Pi inhibits the rate, but not the extent of cofilin binding to actin filaments. This inhibition is also significant at physiological concentrations of Pi, and more pronounced at low pH. Cofilin prevents conformational changes in F-actin induced by Pi, even at high Pi concentrations, probably because allosteric changes in the nucleotide cleft decrease the affinity of Pi to F-actin. Cofilin induced allosteric changes in the nucleotide cleft of F-actin are also indicated by an increase in fluorescence emission and a decrease in the accessibility of etheno-ADP to collisional quenchers. These changes transform the nucleotide cleft of F-actin to G-actin-like. Pi regulation of cofilin binding and the cofilin regulation of Pi binding to F-actin can be important aspects of actin based cell motility.     

A mammalian actin substitution in yeast actin (H372R) causes a suppressible mitochondria/vacuole phenotype

To determine the reason for the inviability of Saccharomyces cerevisiae with skeletal muscle actin, we introduced into yeast actin the first variant muscle residue from the C-terminal end, H372R. Arg is also found at this position in non-yeast nonmuscle actins. The substitution caused retarded growth on glucose and an inability to use glycerol as a sole carbon source. The mitochondria were clumped and had lost their DNA, the vacuole appeared hypervesiculated, and the actin cytoskeleton became somewhat depolarized. Introduction of the second muscle actin-specific substitution, S365A, rescued these defects. Suppression was also achieved by introducing the four acidic N-terminal residues of muscle actin in place of the two found in yeast actin. The H372R substitution results in an increase in polymerization-dependent fluorescence of Cys-374 pyrene-labeled actin. H372R actin polymerizes slightly faster than wild-type (WT) actin. Yeast actin-related proteins 2 and 3 (Arp2/3) accelerates the polymerization of H372R actin to a much greater extent than WT actin. The two suppressors did not affect the rate of H372R actin polymerization in the absence of an Arp2/3 complex. In contrast, the S365A substitution dampened the rate of Arp2/3 complex-stimulated H372R actin polymerization, and the addition of the four acidic N-terminal residues caused this rate to decrease below that observed with WT actin in the presence of Arp2/3. Structural analysis of the mutations suggests the presence of stringent steric and ionic requirements for the bottom of actin subdomain 1 and also suggests that there is allosteric communication through subdomain 1 within the actin monomer between the N and C termini.     

Actin filament uncapping localizes to ruffling lamellae and rocketing vesicles

Regulated actin filament assembly is critical for eukaryotic cell physiology. Actin filaments are polar structures, and those with free high affinity or barbed ends are crucial for actin dynamics and cell motility. Actin filament barbed-end-capping proteins inhibit filament elongation after binding, and their regulated disassociation is proposed to provide a source of free filament ends to drive processes dependent on actin polymerization. To examine whether dissociation of actin filament capping proteins occurs with the correct spatio-temporal constraints to contribute to regulated actin assembly in live cells, I measured the dissociation of an actin capping protein, gelsolin, from actin in cells using a variation of fluorescence resonance energy transfer (FRET). Uncapping was found to occur in cells at sites of active actin assembly, including protruding lamellae and rocketing vesicles, with the correct spatio-temporal properties to provide sites of actin filament polymerization during protrusion. These observations are consistent with models where uncapping of existing filaments provides sites of actin filament elongation.     

The regulation of actin polymerization and the inhibition of monomeric actin ATPase activity by Acanthamoeba profilin

Profilin inhibits the rate of nucleation of actin polymerization and the rate of filament elongation and also reduces the concentration of F-actin at steady state. Addition of profilin to solutions of F-actin causes depolymerization. The same steady state concentrations of polymerized and nonpolymerized actin are reached whether profilin is added before initiation of polymerization or after polymerization is complete. The KD for formation of the 1:1 complex between Acanthamoeba profilin and Acanthamoeba actin is in the range of 4 to 11 microM; the KD for the reaction between Acanthamoeba profilin and rabbit skeletal muscle actin is about 60 to 80 microM, irrespective of the concentrations of KCl or MgCl2. The critical concentration of actin for polymerization and the KD for the actin-profilin interaction are independent of each other; therefore, a change in the critical concentration of actin alters the amount of actin bound to profilin at steady state. As a consequence, the presence of profilin greatly amplifies the effects of small changes in the actin critical concentration on the concentration of F-actin. Profilin also inhibits the ATPase activity of monomeric actin, the profilin-actin complex being entirely inactive.     

Formin Cdc12’s specific actin assembly properties are tailored for cytokinesis in fission yeast

Formins generate unbranched actin filaments by a conserved, processive actin assembly mechanism. Most organisms express multiple formin isoforms that mediate distinct cellular processes and facilitate actin filament polymerization by significantly different rates, but how these actin assembly differences correlate to cellular activity is unclear. We used a computational model of fission yeast cytokinetic ring assembly to test the hypothesis that particular actin assembly properties help tailor formins for specific cellular roles. Simulations run in different actin filament nucleation and elongation conditions revealed that variations in formin’s nucleation efficiency critically impact both the probability and timing of contractile ring formation. To probe the physiological importance of nucleation efficiency, we engineered fission yeast formin chimera strains in which the FH1-FH2 actin assembly domains of full-length cytokinesis formin Cdc12 were replaced with the FH1-FH2 domains from functionally and evolutionarily diverse formins with significantly different actin assembly properties. Although Cdc12 chimeras generally support life in fission yeast, quantitative live-cell imaging revealed a range of cytokinesis defects from mild to severe. In agreement with the computational model, chimeras whose nucleation efficiencies are least similar to Cdc12 exhibit more severe cytokinesis defects, specifically in the rate of contractile ring assembly. Together, our computational and experimental results suggest that fission yeast cytokinesis is ideally mediated by a formin with properly tailored actin assembly parameters.     

Fesselin, a synaptopodin-like protein, stimulates actin nucleation and polymerization.

Fesselin is a proline-rich actin binding protein that has recently been isolated from smooth muscle [Leinweber, B. D., Fredricksen, R. S., Hoffman, D. R., and Chalovich, J. M. (1999) J. Muscle Res. Cell Motil. 20, 539-545]. Fesselin is similar to synaptopodin [Mundel, P., Heid, H. W., Mundel, T. M., Krüger, M., Reiser, J., and Kriz, W. (1997) J. Cell Biol. 139, 193-204] in terms of its size, isoelectric point, and sequence although synaptopodin is not present in smooth muscle. The function of fesselin is unknown. Evidence is presented here that fesselin accelerates the polymerization of actin. Fesselin was effective on actin isolated from either smooth or skeletal muscle at low ionic strength and in the presence of 100 mM KCl. At low ionic strength, fesselin decreased the time for 50% polymerization to about 1% of that in the absence of fesselin. The lag phase characteristic of the slow nucleation process of polymerization was eliminated as the fesselin concentration was increased from very low levels. Fesselin did not alter the critical concentration for actin but did increase the rate of elongation by approximately 3-fold. The increase in elongation rate constant is insufficient to account for the total increase in polymerization rate. It is likely that fesselin stabilizes the formation of actin nuclei. Time courses of actin polymerization at varied fesselin concentrations and varied actin concentrations were simulated by increasing the rate of nucleation and both the forward and reverse rate constants for elongation.     

Inhibition of actin filament depolymerization by the Dictyostelium 30,000-D actin-bundling protein

We have studied the effect of the Dictyostelium discoideum 30,000-D actin-bundling protein on the assembly and disassembly of pyrenyl-labeled actin in vitro. The results indicate that the protein is a potent inhibitor of the rate of actin depolymerization. The inhibition is rapid, dose dependent, and is observed at both ends of the filament. There is little effect of 30-kD protein on the initial rate of elongation from F-actin seeds or on the spontaneous nucleation of actin polymerization. We could detect little or no effect on the critical concentration. The novel feature of these results is that the filament ends are free for assembly but are significantly impaired in disassembly with little change in the critical concentration at steady state. The effects appear to be largely independent of the cross-linking of actin filaments by the 30-kD protein. Actin cross-linking proteins may not only cross-link actin filaments, but may also differentially protect filaments in cells from disassembly and promote the formation of localized filament arrays with enhanced stability.     

Formin Cdc12’s specific actin assembly properties are tailored for cytokinesis in fission yeast

Formins generate unbranched actin filaments by a conserved, processive actin assembly mechanism. Most organisms express multiple formin isoforms that mediate distinct cellular processes and facilitate actin filament polymerization by significantly different rates, but how these actin assembly differences correlate to cellular activity is unclear. We used a computational model of fission yeast cytokinetic ring assembly to test the hypothesis that particular actin assembly properties help tailor formins for specific cellular roles. Simulations run in different actin filament nucleation and elongation conditions revealed that variations in formin’s nucleation efficiency critically impact both the probability and timing of contractile ring formation. To probe the physiological importance of nucleation efficiency, we engineered fission yeast formin chimera strains in which the FH1-FH2 actin assembly domains of full-length cytokinesis formin Cdc12 were replaced with the FH1-FH2 domains from functionally and evolutionarily diverse formins with significantly different actin assembly properties. Although Cdc12 chimeras generally support life in fission yeast, quantitative live-cell imaging revealed a range of cytokinesis defects from mild to severe. In agreement with the computational model, chimeras whose nucleation efficiencies are least similar to Cdc12 exhibit more severe cytokinesis defects, specifically in the rate of contractile ring assembly. Together, our computational and experimental results suggest that fission yeast cytokinesis is ideally mediated by a formin with properly tailored actin assembly parameters.     

Modulation of actin polymerization by the spectrin-band 4.1 complex

The effect of human erythrocyte spectrin dimer and band 4.1 on the polymerization of actin was studied by two independent methods: by following the increase in fluorescence of actin covalently conjugated to N-pyrenyl-iodoacetamide (pyrenylactin) and by following the increase in light scattered by actin polymers. Both techniques indicated that the complex of spectrin dimer and band 4.1, but neither spectrin nor band 4.1 alone, stimulates the rate of nucleation (decreases the lag phase of polymerization) and stabilizes oligomers of F-actin. While the band 4.1-spectrin complex, but not spectrin alone, had no immediate effect on the rate of polymerization after the lag phase, it eventually decreases the rate of actin filament growth when the molecular ratio of actin-spectrin-band 4.1 approaches the physiological range. The complex has no detectable effect on the critical actin concentration and does not significantly alter the apparent order of the nucleation reaction.     

Mutational analysis of arginine 177 in the nucleotide binding site of beta-actin.

Actin ADP-ribosylated at arginine 177 is unable to hydrolyze ATP, and the R177 side chain is in a position similar to that of the catalytically essential lysine 71 in heat shock cognate protein Hsc70, another member of the actin-fold family of proteins. Therefore, actin residue R177 has been implicated in the mechanism of ATP hydrolysis. This paper compares wild-type beta-actin with a mutant in which R177 has been replaced by aspartic acid. The mutant beta-actin was expressed in Saccharomyces cerevisiae and purified by DNase I-affinity chromatography. The mutant protein exhibited a reduced thermal stability and an increased nucleotide exchange rate, suggesting a weakened interdomain connection. The ATPase activity of G-actin and the ATPase activity expressed during polymerization were unaffected by the R177D replacement, showing that this residue is not involved in catalysis. In the presence of polymerizing salts, ATP hydrolysis by both wild-type Mg-beta-actin and the mutant protein preceded filament formation. With the mutant actin, the initial rate of ATP hydrolysis was as high as with wild-type actin, but polymer formation was slower, reached lower steady-state levels, and the polymers formed exhibited much lower viscosity. The critical concentration of polymerization (Acc) of the mutant actin was increased 10-fold as compared to wild-type actin. Filaments formed from the R177D mutant beta-actin bound phalloidin.     

Yeast actin with a subdomain 4 mutation (A204C) exhibits increased pointed-end critical concentration.

Characterizing mutants of actin that do not polymerize will advance our understanding of the mechanism of actin polymerization and will be invaluable for the production of short F-actin structures for structural studies. To circumvent the problem of expressing dominant lethal nonpolymerizing actin in yeast, we adopted a cysteine engineering strategy. Here we report the characterization of a mutant of yeast actin, AC-actin, possessing a single pointed-end mutation, A204C. Expression of this mutant in yeast results in actin-polymerization-deficient phenotypes. When copolymerized with wild-type actin, ATP-AC-actin is incorporated into filaments. ADP-AC-actin participates in the nucleation and elongation of wild-type filaments only at very high concentrations. At low concentrations, ADP-AC-actin appears to participate only in the nucleation of wild-type filaments, suggesting that Ala-204 is involved in modulating the critical concentration of the pointed end of actin.     

Binding of phosphate to F-ADP-actin and role of F-ADP-Pi-actin in ATP-actin polymerization

Our previous work (Carlier, M.-F., and Pantaloni, D. (1986) Biochemistry 25, 7789-7792) had shown that F-ADP-Pi-actin is a major intermediate in ATP-actin polymerization, due to the slow rate of Pi release following ATP cleavage on filaments. To understand the mechanism of ATP-actin polymerization, we have prepared F-ADP-Pi-actin and characterized its kinetic parameters. 32Pi binds to F-ADP-actin with a stoichiometry of 1 mol/mol of F-actin subunit and an equilibrium dissociation constant Kpi of 1.5 mM at pH 7.0 Kpi increases with pH, indicating that the H2PO-4 species binds to F-actin. ADP-Pi-actin subunits dissociate much more slowly from filament ends than ADP-actin subunits; therefore, the stability of filaments in ATP is due to terminal ADP-Pi subunits. The slow rate of dissociation of ADP-Pi-actin also explains the decrease in critical concentration of ADP-actin in the presence of Pi reported by Rickard and Sheterline (Richard, J. E., and Sheterline, P. (1986) J. Mol. Biol. 191, 273-280). The effect of Pi on the rate of actin dissociation from filaments is much more pronounced at the barbed end than at the pointed end. Using gelsolin to block the barbed end, we have shown that the two ends are energetically different in the presence of ATP and saturating Pi, but less different than in the absence of Pi. The results are interpreted within a new model for actin polymerization. It is possible that phosphate binding to F-actin can regulate motile events in muscle and nonmuscle cells.