Dynamic actin filaments control the mechanical behavior of the human red blood cell membrane.

Short, uniform-length actin filaments function as structural nodes in the spectrin-actin membrane skeleton to optimize the biomechanical properties of red blood cells (RBCs). Despite the widespread assumption that RBC actin filaments are not dynamic (i.e., do not exchange subunits with G-actin in the cytosol), this assumption has never been rigorously tested. Here we show that a subpopulation of human RBC actin filaments is indeed dynamic, based on rhodamine-actin incorporation into filaments in resealed ghosts and fluorescence recovery after photobleaching (FRAP) analysis of actin filament mobility in intact RBCs (~25-30% of total filaments). Cytochalasin-D inhibition of barbed-end exchange reduces rhodamine-actin incorporation and partially attenuates FRAP recovery, indicating functional interaction between actin subunit turnover at the single-filament level and mobility at the membrane-skeleton level. Moreover, perturbation of RBC actin filament assembly/disassembly with latrunculin-A or jasplakinolide induces an approximately twofold increase or ~60% decrease, respectively, in soluble actin, resulting in altered membrane deformability, as determined by alterations in RBC transit time in a microfluidic channel assay, as well as by abnormalities in spontaneous membrane oscillations (flickering). These experiments identify a heretofore-unrecognized but functionally important subpopulation of RBC actin filaments, whose properties and architecture directly control the biomechanical properties of the RBC membrane.

INTRODUCTION

The mammalian red blood cell (RBC) is structurally unique in that it contains no nucleus, intracellular organelles, or transcellular cytoskeleton. Underlying the RBC plasma membrane is the membrane skeleton, a two-dimensional isotropic cytoskeletal network with quasi-hexagonal symmetry. The membrane skeleton consists of nodes of short (∼37-nm-long) actin filaments interconnected by long strands of (α1β1)2-spectrin tetramers (Fowler, 2013). The actin filaments of the RBC membrane skeleton are each attached to five to seven (α1β1)2-spectrin tetramers stabilized along their lengths by two tropomyosin isoforms, TM5b (Tpm1.9) and TM5NM1 (Tpm3.1), capped at their pointed and barbed ends by tropomodulin1 (Tmod1) and αβ-adducin, respectively, and possibly bundled by dematin (functions reviewed in Fowler, 2013). Current models of the spectrin-actin membrane skeleton presume that the actin filaments function as stable nodes or linkages, whose subunits do not dynamically assemble or disassemble (Fowler, 2013). The RBC membrane skeleton is the prototypical membrane-associated cytoskeleton, serving as a platform for understanding the assembly and maintenance of specialized membrane protein domains in diverse cell types, such as the transverse tubules and intercalated disks of cardiomyocytes, costameres of skeletal muscle fibers, axon initial segments and nodes of Ranvier in neurons, and lateral membrane domains of polarized epithelial cells (Bennett and Gilligan, 1993; Bennett and Baines, 2001; Bennett and Lorenzo, 2013). Therefore, understanding the basis for membrane skeleton assembly, stability, and connectivity is of broad biological importance.

Several studies have indicated that the architecture of RBC actin filament nodes mediates the connectivity of the membrane skeleton and imparts the RBC with the biconcave shape and biomechanical properties optimized for efficient passage through the microvasculature. RBC actin filament uncapping via targeted genetic deletion of Tmod1 or the α or β subunit of adducin in mice leads to abnormal RBC shapes (spheroelliptocytosis), with increased osmotic fragility and reduced deformability, resulting in hemolytic anemias of varying severities (Gilligan ; Muro ; Porro ; Chen ; Robledo ; Moyer ). Furthermore, simultaneous deletion of β-adducin and the dematin headpiece domain required for actin bundling results in more-pronounced spherocytosis and a correspondingly more-severe anemia (Chen ; Liu ). In the case of Tmod1 deletion, the short actin filaments in the membrane skeleton become more variable in length (Moyer ), implying altered actin assembly/disassembly, although this has not been directly demonstrated. Moreover, actin filament assembly in the RBC membrane skeleton has been implicated in the origin of RBC membrane “flickering” (i.e., spontaneous and ATP-dependent vibrational oscillations of the RBC membrane; Burton ; Levin and Korenstein, 1991; Tuvia , 1998; Costa ; Betz ; Yoon ). This has been supported by experimental observations of reduced flickering after chemical inhibition of actin filament barbed-end assembly (Tuvia ).

The uniform lengths of the short RBC actin filaments and their locations at nodes of the spectrin-actin membrane skeleton have led to the current assumption that RBC actin filaments are not dynamic (i.e., do not exchange subunits with G-actin in the cytosol during normal RBC homeostasis). However, whether RBC actin filaments might be dynamic and, if so, whether actin dynamics regulates membrane skeleton connectivity and/or RBC mechanical properties are questions that have never been rigorously addressed. Several lines of evidence have hinted, but not directly proven, that a dynamic process of actin subunit turnover may occur in RBCs. First, the RBC cytosol contains ∼10 μg/ml (0.24 μM) G-actin, which is slightly greater than the barbed-end critical concentration (Pinder and Gratzer, 1983; Pollard ), suggesting that soluble actin subunits may exchange with subunits at the barbed ends. Consistent with this, a cytosolic actin population has been visualized via immunogold labeling of sections of intact RBCs (Cyrklaff ). Second, it has been shown that infection of human RBCs with the malaria-causing parasite Plasmodium falciparum results in dramatic remodeling of RBC actin filaments into an aberrantly branched network in the cytosol to facilitate export of virulence factors (Cyrklaff ; Rug ). Presumably, such reorganization requires dynamic disassembly/reassembly of actin filaments.

Here, we directly probed the mechanisms of actin dynamics in RBCs via treatment with the actin-disrupting drugs cytochalasin-D (CytoD), latrunculin-A (LatA), and jasplakinolide (Jasp). CytoD inhibits actin subunit association and dissociation at barbed ends; LatA destabilizes and depolymerizes dynamic actin filaments by binding to and sequestering actin monomers, driving the F:G-actin balance toward the G-actin state; and Jasp stabilizes dynamic actin filaments, driving the F:G-actin balance toward the F-actin state (MacLean-Fletcher and Pollard, 1980; Cooper, 1987; Coue ; Spector ; Sampath and Pollard, 1991; Bubb , 2000; Morton ; Allingham ; Holzinger, 2009). Our results show that rhodamine-actin (rho-actin) subunits incorporate into discrete sites within the membrane skeleton in resealed RBC ghosts, suggesting that a subpopulation of RBC actin filaments is indeed available for subunit exchange. Rho-actin subunit incorporation can be blocked via treatment with CytoD, demonstrating that subunit exchange occurs at barbed ends. Moreover, fluorescence recovery after photobleaching (FRAP) analysis revealed a mobile subpopulation of RBC actin filaments (∼25–30% of total actin filaments), whose mobility can be partially attenuated via treatment with CytoD. Consistent with subunit assembly/disassembly, treatment with LatA or Jasp induces an approximately twofold increase or ∼60% decrease, respectively, in soluble actin. Neither LatA nor Jasp treatment affected RBC osmotic fragility, but both LatA and Jasp treatment resulted in altered membrane deformability and flickering. Collectively, these results provide the first direct evidence for the existence of a dynamic and mobile RBC actin filament subpopulation and that actin filament architecture directly controls the mechanics of the human RBC membrane.

RESULTS

Characterization of RBC actin subunit exchange and filament mobility

A previous study measured a G-actin concentration of ∼0.24 μM in human RBC cytosol, based on the ability of G-actin to inhibit DNase I activity (Pinder and Gratzer, 1983). We reinvestigated the concentration of cytosolic G-actin in RBCs by using Western blots to compare the relative proportions of actin in Triton X-100–extracted membrane skeletons versus RBC cytosolic fractions. This experiment revealed that ∼96.3% of human RBC actin is membrane skeleton associated, whereas ∼3.7% is cytosolic (Figure 1, A and B), similar to our previous analyses of mouse RBC actin (Moyer ). This distribution of insoluble versus cytosolic actin is comparable to the distribution observed in mouse skeletal muscle, in which ∼5–10% of total actin is cytosolic (Gokhin and Fowler, 2011), but differs markedly from cell types with rapidly remodeling actin cytoskeletons, such as migrating vascular endothelial cells, in which ∼50–60% of total actin is cytosolic (Hinshaw ; Fischer ). Assuming ∼500,000 actin subunits in the human RBC membrane skeleton (Pinder and Gratzer, 1983; Fowler, 1996) and a typical human RBC volume of 90 fl (Kaushansky ), we calculate a total (membrane skeleton–associated plus cytosolic) actin concentration of ∼9.6 μM in RBCs (Figure 1B). Moreover, assuming that RBCs do not contain G-actin–sequestering proteins, our data reveal a cytosolic G-actin concentration of ∼0.36 μM (Figure 1B), which is ∼50% greater than the value of ∼0.24 μM reported previously (Pinder and Gratzer, 1983) and halfway between the actin barbed-end and pointed-end critical concentrations of 0.1 and 0.6 μM, respectively (Pollard ).

FIGURE 1:

Inhibition of barbed-end exchange with CytoD blocks dynamic incorporation of rho-actin subunits into the RBC membrane skeleton. (A) Western blots for actin and Tmod1 in total human RBC extracts, RBC cytosolic fractions (supernatants), and Triton X-100–extracted membrane skeletons (pellets). (B) Tabulation of actin and Tmod1 stoichiometry in RBCs. Note that ∼96.3% of actin and, for comparison, virtually all Tmod1 is associated with the membrane skeleton. Cytosolic and total actin molecules per RBC were based on a value of 500,000 actin subunits/membrane skeleton (Pinder ; Pinder and Gratzer, 1983), and actin concentrations were based on an average human RBC volume of 90 fl (Kaushansky ). (C) Phase-contrast (left) and fluorescence (right) micrographs of human RBC ghosts incubated in 2.4 μM rho-actin and either DMSO (top) or 0.5 μM CytoD (bottom) for 10 min on ice and then resealed and incubated for 30 min at 37ºC. In DMSO-treated ghosts, rho-actin localizes to bright puncta along the ghost membrane (white arrowhead), indicating incorporation into the membrane skeleton. In CytoD-treated ghosts, incorporation of rho-actin into the membrane skeleton is markedly reduced, as indicated by fewer and fainter rho-actin puncta (black arrowhead). (D) Western blots of human RBC ghosts incubated in 2.4 μM rho-actin and either DMSO or 0.5 μM CytoD for 10 min on ice and then resealed and incubated for 30 min at 37ºC. Ghosts were then extracted by Triton X-100 and centrifuged to obtain cytosolic fractions (supernatants) and Triton X-100–extracted membrane skeletons (pellets). Western blotting was performed using an anti-rhodamine antibody to assess distribution of exogenous rho-actin (top), an anti–β-actin antibody to assess distribution of endogenous RBC actin (exclusively β-actin; Pinder and Gratzer, 1983); middle), and an anti–pan-actin antibody to assess distribution of total actin (exogenous rho-actin plus endogenous RBC actin; bottom).

Next, to directly address whether dynamic actin subunit assembly or exchange occurs between the RBC membrane skeleton and cytosol, we resealed human RBC ghosts in the presence of rho-actin for 30 min and then visualized rho-actin incorporation. In dimethyl sulfoxide (DMSO)–treated ghosts, rho-actin localized to discrete sites around the RBC circumference, consistent with dynamic incorporation of rho-actin subunits into the membrane skeleton (Figure 1C). This effect was dramatically attenuated in ghosts treated with 0.5 μM CytoD to inhibit barbed-end exchange (Figure 1C), indicating that barbed-end exchange mediates rho-actin incorporation into the RBC membrane skeleton, consistent with a cytosolic G-actin concentration of ∼0.36 μM (Figure 1B). However, the discrete foci of rho-actin incorporation suggest that only a subpopulation of actin filaments is available for barbed-end incorporation within the experiment's 30-min time frame. We confirmed CytoD inhibition of rho-actin incorporation into the membrane skeleton by Western blotting of Triton X-100–extracted membrane skeletons versus cytosolic fractions using an anti-rhodamine antibody (Figure 1D). Additional evidence in favor of dynamic subunit exchange comes from our observation that a small amount of endogenous RBC actin is displaced from the membrane skeleton fraction to the cytosolic fraction in the presence of rho-actin (compare Figure 1, A and D), as determined by Western blotting using an antibody against β-actin, the singular endogenous actin isoform in RBCs (Pinder and Gratzer, 1983).

To determine whether inhibition of barbed-end exchange might also influence overall RBC actin filament stability, we incubated intact RBCs with increasing concentrations of CytoD and performed Western blots to detect changes in soluble actin. We observed no changes in the amount of soluble actin in RBCs treated with CytoD concentrations up to 1 μM (Figure 2, A and D), indicating that inhibition of barbed-end exchange does not lead to net assembly or disassembly of RBC actin filaments. This finding was supported by Alexa 488–phalloidin staining and total internal reflection fluorescence (TIRF) microscopy of CytoD-treated RBCs, which showed diffuse and uniform F-actin staining across the RBC membrane, indistinguishable from DMSO-treated RBCs (Figure 2G). We conclude that the incorporation of rho-actin into discrete foci in the membrane skeleton is likely due to rho-actin subunit exchange with actin subunits at barbed ends of a subpopulation of the endogenous actin filaments in the membrane skeleton.

FIGURE 2:

Treatment with LatA and Jasp (but not CytoD) alters the soluble actin pool in human RBCs without affecting the diffuse localization pattern of F-actin determined by TIRF microscopy. (A–C) Western blots of Triton X-100–extracted supernatants prepared from intact RBCs treated with the indicated concentrations of (A) CytoD, (B) LatA, or (C) Jasp and then lysed with Triton-lysis buffer. (D–F) Quantitation of Western blots in A–C. Note that CytoD has no effect on the soluble actin pool, 10 μM LatA results in approximately twofold increase in soluble actin that does not increase further with increasing concentrations of LatA, and 5 μM Jasp results in ∼60% decrease in soluble actin that does not decrease further with increasing concentrations of Jasp. In all experiments, Ponceau S served as a loading control for normalization of actin levels. Error bars indicate mean ± SEM of n = 3 lanes. *p < 0.01. (G) Representative TIRF micrographs of RBCs stained with Alexa 488–phalloidin to label F-actin associated with the RBC membrane. Note the diffuse and uniform localization of F-actin in all drug treatment groups. Variability in RBC size and apparent “folds” in the RBC membrane are artifacts arising from cytocentrifugation onto glass slides.

Although our rho-actin incorporation experiments indicate turnover of at least some individual RBC actin filaments, an alternative approach is required to determine the fraction of actin filaments that are mobile in intact RBCs. To study this question, we labeled the actin filaments in the membrane skeleton of intact living RBCs with a cell-permeable infrared-fluorescent Jasp conjugate, silicon-rhodamine (SiR)-Jasp, which was developed to study actin mobility in living cells and previously referred to as SiR-actin (Lukinavicius ). SiR-Jasp is particularly advantageous for investigations of RBCs due to the nonoverlap of the emission spectrum of SiR with the absorption spectrum of hemoglobin (Lukinavicius ). As expected, SiR-Jasp efficiently labeled RBC actin filaments, based on bright rim staining visualized by wide-field fluorescence microscopy (Figure 3A). SiR-Jasp was also confirmed to selectively label the membrane skeleton by using TIRF microscopy (unpublished data), similar to Alexa 488–phalloidin staining of fixed and permeabilized RBCs imaged by TIRF microscopy (Figure 2G).

FIGURE 3:

FRAP analysis of SiR-Jasp–stained human RBCs reveals F-actin mobility that can be reduced by CytoD treatment. (A) Example fluorescence micrographs of SiR-Jasp–stained RBCs before photobleaching, at the time of photobleaching, and during postphotobleaching recovery. Yellow arrow indicates photobleached area. (B) Example FRAP data. The black trace represents an original, uncorrected recovery curve measured in a photobleached area. The red trace represents nonspecific, off-target bleaching measured in same RBC but away from the photobleached area. The green trace represents the recovery curve corrected for nonspecific, off-target bleaching during repetitive image acquisition. The blue trace represents exponential curve fitting used to calculate F-actin mobile fraction and half-time of fluorescence recovery. Black arrow indicates the time of photobleaching. (C) Dotplot of F-actin mobile fraction in RBCs in buffer with or without fixation in 0.5% acrolein before SiR-Jasp staining. (D, E) Dotplots of (D) F-actin mobile fraction and (E) half-time of F-actin fluorescence recovery for RBCs treated with either DMSO or 1 μM CytoD. Each dot represents data from one RBC. *p < 0.05; **p < 0.001.

We performed FRAP analysis of intact, living SiR-Jasp–labeled RBCs by photobleaching a 1-μm-diameter spot, followed by collection of fluorescence images for at least 15 min (Figure 3B). After applying correction for nonspecific photobleaching arising from repetitive image acquisition, as described previously (Quadri ; Melhorn ), we found that recovery of SiR-Jasp fluorescence occurred within minutes. By applying exponential curve fitting to the corrected fluorescence versus time traces (Reits and Neefjes, 2001), we calculated that an average of ∼25–30% of the SiR-Jasp-labeled RBC actin filaments comprises a mobile subpopulation (Figure 3C). The observed fluorescence recovery was due to bona fide F-actin mobility and was not an artifact resulting from association/dissociation of the SiR-Jasp probe, because fixation in 0.5% acrolein before SiR-Jasp staining abolished fluorescence recovery and reduced the apparent mobile fraction of F-actin to almost 0% without affecting the SiR-Jasp labeling of the RBCs (Figure 3C). Intriguingly, inhibition of barbed-end exchange via CytoD treatment slightly but significantly reduced the mobile fraction of RBC actin filaments (Figure 3D), although the half-time of fluorescence recovery was unaffected (Figure 3E), suggesting that barbed-end assembly partially accounts for actin filament mobility in RBCs. We conclude that approximately one-third of the filaments in the RBC actin filament network possess a heretofore-unrecognized capacity for forming new filaments, elongating existing filaments, undergoing micrometer-scale rearrangements, or a combination thereof. In addition, barbed-end actin subunit exchange at the single-filament level influences two-dimensional F-actin mobility in the plane of the membrane skeleton.

Treatment with LatA and Jasp perturbs RBC actin filament assembly

Treatment with LatA or Jasp provides a pharmacological means to alter F:G-actin balance for studying dynamic actin filament network function in living cells (for examples, see Ayscough ; Wang , 2014; Pappas ). As expected from LatA's monomer-sequestering function (Coue ; Spector ; Morton ), LatA treatment of intact RBCs for 4 h led to an approximately twofold increase in soluble actin at a LatA concentration of 10 μM, and no additional increase in soluble actin was observed when the LatA concentration was raised to 20 μM (Figure 2, B and E). Similarly, as expected from Jasp's ability to stabilize actin filaments by inhibiting actin subunit dissociation from the filaments (Bubb , 2000; Holzinger, 2009), Jasp treatment of intact RBCs for 4 h led to ∼60% decrease in soluble actin at a Jasp concentration of 5 μM, and no additional decrease in soluble actin was observed when the Jasp concentration was raised to 10 μM (Figure 2, C and F). Because the overwhelming majority of RBC actin is associated with the insoluble membrane skeleton fraction (Figure 1, A and B), these changes in soluble actin do not result in concomitant decreases in insoluble actin of a sufficient magnitude to be detectable by Western blotting (unpublished data). The idea that the membrane skeleton remains grossly intact after drug treatment is supported by the observation that LatA treatment had no effect on the levels of soluble Tmod1, α-adducin, or TM5NM1 (Supplemental Figure S1) and that TIRF microscopy of LatA- and Jasp-treated RBCs found no changes in F-actin localization as compared with DMSO-treated RBCs (Figure 2G). Collectively, these data indicate that LatA or Jasp treatment provides a means to specifically perturb assembly/disassembly of a small subpopulation of RBC actin filaments.

Disruption of actin filament assembly/disassembly alters RBC membrane mechanics

To determine the role that actin filament assembly/disassembly might play in regulating RBC membrane mechanics, we subjected LatA- and Jasp-treated RBCs to a battery of functional assays. First, we observed that treatment with a LatA or Jasp concentration sufficiently high to disrupt RBC actin filaments (10 or 5 μM, respectively) had no effect on the osmotic fragility of RBCs (Figure 4). This indicates preserved membrane extensibility and ratio of surface area to volume (Mohandas and Chasis, 1993), which is consistent with the lack of discernible effects of LatA or Jasp on RBC sizes and shapes visualized by TIRF microscopy of Alexa 488-phalloidin–stained RBCs (Figure 3G). However, when we measured RBC transit times through 5-μm-wide microfluidic channels, mimicking RBC passage through capillaries (Shevkoplyas , 2006; Ghiran ), we observed decreased transit times for both LatA- and Jasp-treated RBCs (Figure 5), indicating enhanced membrane deformability. Enhanced membrane deformability was observed at 0.2 μM LatA and 0.5 μM Jasp, which are drug concentrations whose effects on RBC actin filament assembly/disassembly are not within the detection sensitivity of Western blotting for actin in RBC supernatants (compare Figure 2, A–F, and Supplemental Figure S2). Enhanced membrane deformability at such low drug concentrations highlights the remarkable sensitivity of RBC deformability to RBC actin filament assembly.

FIGURE 4:

LatA or Jasp treatment does not affect the osmotic fragility of human RBCs. Curves depict the percentage of lysed RBCs as a function of buffer osmolarity for (A) RBCs treated with either DMSO or 10 μM LatA and (B) RBCs treated with either DMSO or 5 μM Jasp. Error bars indicate mean ± SEM of n = 12 replicates. Note that RBCs in A and B were obtained from different human donors and hence have somewhat different osmotic fragilities in the presence of DMSO alone.

FIGURE 5:

LatA or Jasp treatment increases the membrane deformability of human RBCs. (A) Representative video stills of RBC entry, passage, and egress through a microfluidic channel in response to fluid flow. (B, C) Boxplots depict RBC transit times through a microfluidic channel, from entry to egress, for RBCs treated with the indicated concentrations of (B) LatA or (C) Jasp. Decreased transit times of LatA- and Jasp-treated RBCs signify increased membrane deformability. Each boxplot reflects n ≥ 30 RBCs. *p < 0.01. Note that RBCs in B and C were obtained from different human donors and hence have slightly different transit times in the presence of DMSO alone.

RBC transit time in a microfluidic channel can be expressed as the entry time (i.e., time required for initial RBC shape change upon channel entry) plus the persistence time in the channel before egress from the channel. When we compared these distinct modes of RBC kinematics, we observed that LatA treatment decreases both entry time and persistence time in the channel, whereas Jasp treatment only decreases entry time, with little effect on persistence time in the channel (Figure 6). We conclude that reduced transit time of LatA-treated RBCs is attributable to both reduced entry time and persistence time in the channel, whereas reduced transit time of Jasp-treated RBCs is solely attributable to reduced entry time.

FIGURE 6:

LatA and Jasp treatment have differential effects on entry time vs. persistence time in a microfluidic channel. Boxplots depict (A, C) RBC entry times and (B, D) persistence times in a microfluidic channel for RBCs treated with the indicated concentrations of (A, B) LatA or (C, D) Jasp. Note that LatA treatment decreases both entry time and persistence time in the channel, whereas Jasp treatment only decreases entry time, with no effect on persistence time in the channel. Each boxplot reflects n ≥ 30 RBCs. *p < 0.01. Note that RBCs in A and B vs. C and D were obtained from different human donors and hence have slightly different entry time and persistence time in the presence of DMSO alone.

Next we used LatA and Jasp treatment to examine the role of actin filament assembly and disassembly in RBC flickering—spontaneous oscillations of the RBC membrane (Levin and Korenstein, 1991; Tuvia ; Costa ; Betz )—which are reduced via CytoD inhibition of F-actin's ATPase activity (Tuvia ). We analyzed RBC flickering by phase-contrast video microscopy of biconcave RBCs, with computational analysis of the coefficient of variance over time for the amplitudes of RBC membrane oscillations (Figure 7A; Ghiran ). We found that F-actin disassembly induced via treatment with 1 μM LatA significantly increased the variance of flickering, whereas F-actin stabilization via treatment with 1 μM Jasp had the opposite effect (Figure 7, B and C). Therefore actin filament stability is inversely related to the variance of flickering. This result differs from our microfluidic channel assays, for which F-actin destabilization via LatA treatment and F-actin stabilization via Jasp treatment induced nearly identical increases in RBC deformability (Figure 5). Collectively our microfluidic channel and flickering assays support a model in which different actin filament subpopulations are differentially assembled and disassembled in different modes of RBC kinematics.

FIGURE 7:

LatA and Jasp treatment have differential effects on the membrane flickering of human RBCs. (A) Example heat maps depicting variances of flickering amplitudes across the RBC surface. (B, C) Dotplots depict coefficients of variance of flickering amplitudes for RBCs treated with the indicated concentrations of (B) LatA or (C) Jasp. Note that LatA treatment increases the variance of flickering amplitudes, whereas Jasp treatment decreases the variance of flickering amplitudes. Each dot represents data from one RBC. *p < 0.05. Note that RBCs in B and C were obtained from different human donors and hence have slightly different variances of flickering amplitudes in the presence of DMSO alone.

DISCUSSION

Filament-level actin subunit dynamics in RBCs

Research on RBCs has long assumed that membrane skeleton-associated actin filaments are static and do not undergo dynamic subunit exchange (Fowler, 2013). However, our data challenge this dogma and argue in favor of dynamic actin subunit exchange between the cytosol and an actin filament subpopulation in the RBC membrane skeleton, mediated by barbed-end dynamics. First, the soluble actin concentration in RBCs (0.36 μM) is between the barbed-end and pointed-end critical concentrations of 0.1 and 0.6 μM, respectively (Pollard ). Second, we observed that CytoD treatment inhibits rho-actin incorporation at discrete foci in the membrane skeleton of RBC ghosts. These experiments were performed with RBCs from adult peripheral blood, with minimal (<1%) contamination by reticulocytes, which lack a mature spectrin-actin membrane skeleton with its full cohort of associated actin-binding proteins (Chasis ; Kaushansky ; Liu ). Thus barbed-end exchange appears to be the predominant mechanism facilitating dynamic actin subunit incorporation into the RBC membrane skeleton. Moreover, an RBC cytosolic actin concentration intermediate with respect to the barbed-end and pointed-end critical concentrations suggests that at least some RBC actin filaments might undergo treadmilling (i.e., subunit assembly at barbed ends with disassembly at pointed ends; Bugyi and Carlier, 2010).

Although our data do not allow us to determine whether many filaments partially assemble or disassemble or whether small numbers of entire filaments polymerize or depolymerize, it is instructive to compare the changes in cytosolic actin with the amount of polymerized actin filaments in the membrane skeleton. For example, our drug treatment experiments indicate that the fraction of actin subunits capable of assembling and disassembling from RBC actin filaments via barbed-end exchange is rather low. The approximately twofold increase in soluble actin after LatA treatment reflects an increase from 3.7% to 7.4% soluble actin over the 4-h incubation at 37°C used here, corresponding to depolymerization of ∼1000–1600 whole filaments out of a population of ∼30,000–40,000 filaments in the membrane skeleton, or, alternatively, slightly <1 subunit from each of the 30,000–40,000 filaments (Fowler, 1996). Similarly, the ∼60% decrease in soluble actin after Jasp treatment reflects a decrease from 3.7% to 1.5% soluble actin, corresponding to polymerization of ∼600–900 new filaments in the membrane skeleton, or <1 subunit onto each of the 30,000–40,000 filaments (Fowler, 1996). Such subtle changes in F-actin content and organization in response to actin-disrupting drugs differ markedly from the robust changes in F-actin organization observed in LatA- and Jasp-treated nonerythroid cells (Ayscough ; Bubb ; Pappas ) and may reflect specialization of the spectrin-based membrane skeleton. Our calculations, in conjunction with the observed foci of rho-actin incorporation into RBC ghosts, favor the existence of dynamic subpopulations or “hotspots” within the membrane skeleton where actin filaments assemble and disassemble, rather than uniform turnover across the entire RBC actin network.

Network-level actin filament mobility in RBCs

Our FRAP experiments also revealed a previously unrecognized long-range mobility of the RBC actin filament network. The simplest explanation for F-actin mobility in the spectrin-actin membrane skeleton of a quiescent biconcave RBC might be the occurrence of stochastic spreading and unspreading events within the network, facilitated by conformational changes in the (α1β1)2-spectrin strands (Bloch and Pumplin, 1992). During such events, rearrangements of (α1β1)2-spectrin strands would presumably be driven by lateral movements of ankyrin/band3 complexes in the lipid bilayer (Golan and Veatch, 1980; Kodippili , 2012). This, in turn, would result in lateral displacement of (α1β1)2-spectrin-bound actin filaments. Occurrence of such events would also imply that the degree of spreading of the RBC membrane skeleton is heterogeneous across the RBC membrane, which could lead to spatial heterogeneities in the force distribution across the membrane, possibly driving RBC membrane curvature and biconcave shape (Mohandas ; Mohandas and Chasis, 1993).

An alternative explanation for the observed F-actin mobility in RBCs would be spontaneous dissociation/reassociation of (α1β1)2-spectrin/F-actin linkages, leading to changes in membrane skeleton architecture and actin filament positioning. However, the presence of five to seven high-affinity (α1β1)2-spectrin attachments on each RBC actin filament appears to make such events relatively unlikely (Bennett and Baines, 2001; Fowler, 2013). It also remains unclear why inhibition of barbed-end actin exchange at the single-filament level via CytoD treatment results in a small but statistically significant decrease in the mobile fraction of F-actin at the membrane-skeleton level. A possibility is that barbed-end assembly (i.e., more ATP-actin subunits) might influence the affinity of (α1β1)2-spectrin for F-actin or that CytoD preferentially binds to and directly alters the mobility of a biochemically distinct subset of mobile RBC actin filaments. Additional biochemical experiments are required to distinguish among these or other possibilities.

It is instructive to compare actin mobility in RBCs with actin mobility in nonerythroid cells. RBC actin filaments are structurally similar to the thin filaments of striated muscle sarcomeres in that they are both long-lived cytoskeletal structures, have precisely regulated and highly uniform lengths, are capped at both ends, and are coated along their length with tropomyosin (Fowler, 1996). However, sarcomeric thin filaments are many times longer than RBC actin filaments (∼1 μm vs. 37 nm long, respectively) and exhibit both barbed-end and pointed-end assembly when rho-actin is microinjected into living myocytes (Littlefield ), unlike RBC actin filaments, which only exhibit barbed-end assembly when RBC ghosts are resealed in the presence of rho-actin. Nevertheless, FRAP analysis of sarcomeric thin filaments in cultured cardiomyocytes and skeletal myvotubes expressing GFP-actin has identified a mobile actin fraction of ∼25% and fluorescence recovery occurring on the minutes time scale (Littlefield ; Skwarek-Maruszewska ; Pappas ), similar to our findings in RBCs. By contrast, FRAP analysis of the highly dynamic, dendritic actin filament network in the lamellipodium of migrating cells identifies an actin mobile fraction of nearly 100% and fluorescence recovery occurring within seconds (Lai ). Collectively these studies argue against simple categorization of actin filament structures as either “static” or “dynamic” and instead favor a continuum of dynamic states on which both short- and long-lived actin filament structures might reside.

Actin filament function in RBC biomechanics

We confirmed that it is possible to perturb F-actin assembly/disassembly in RBCs via treatment with LatA or Jasp, consistent with these drugs' destabilizing or stabilizing effects on dynamic actin filaments, respectively (Coue ; Spector ; Bubb , 2000; Morton ; Holzinger, 2009). We exploited this in our biomechanical assays, which revealed that RBC actin filaments play important roles in specific modes of RBC deformation. Neither LatA nor Jasp treatment affected RBC osmotic fragility, indicating unchanged membrane extensibility and ratio of surface area to volume (Kaushansky ), but both LatA and Jasp treatment increased membrane deformability, as determined by a microfluidic channel assay. Moreover, our data are the first to reveal different modes of RBC actin filament function during passage through a microfluidic channel—namely, microchannel entry time is both LatA and Jasp sensitive, whereas persistence time in the microchannel is LatA but not Jasp sensitive, suggesting that different subpopulations of actin filaments with different assembly/disassembly and mechanical properties are recruited during each of these functional modes.

The importance of RBC actin filaments in regulating RBC deformability was first suggested by experiments showing abnormal membrane deformability in resealed RBC ghosts treated with phalloidin (Nakashima and Beutler, 1979), which has a similar actin filament–stabilizing effect as Jasp, and later by a series of gene-targeted mouse models lacking various actin-capping proteins in RBCs (Gilligan ; Muro ; Porro ; Chen ; Robledo ; Moyer ). (Treatment with a CytoD analogue, CytoB, has also been shown to increase RBC osmotic fragility and membrane deformability [ Beck ], but this could be due to off-target binding to the glucose transporter and metabolic effects [ Bloch, 1973; Taverna and Langdon, 1973; Lin and Spudich, 1974; Jung and Rampal, 1977].) Increased RBC deformability due to LatA or Jasp treatment could be due to changes in the number and/or steric accessibility of (α1β1)2-spectrin-binding sites on some of the short actin filaments, which may lead to rearrangement of the spectrin-actin lattice and suboptimal force distribution across the plasma membrane. Such a mechanism is consistent with mathematical modeling of RBC actin filament nodes, whose mechanical behavior is sensitive to the spatial arrangement of (α1β1)2-spectrin attachments (Sung and Vera, 2003; Vera ).

In addition to increasing deformability, we also discovered that LatA or Jasp treatment increases or decreases the variance of intact RBC membrane flickering amplitudes, respectively (i.e., actin filament stability is inversely related to the variance of flickering). This extends work by Tuvia showing that phalloidin treatment, which is expected to have a similar effect as Jasp treatment, decreases the variance of flickering in RBC ghosts. However, results by Betz differed from ours in that they showed that no effect of LatA on flickering; we attribute this difference to the fact that they examined solely flickering amplitudes, whereas we focused on flickering variances. We propose two possible models: 1) Actin filament assembly/disassembly in the proximity of the RBC membrane may result in transient contacts between the lipid bilayer and dynamic actin subunits at filament ends, which, when integrated over space and time, manifest as 0.2- to 30-Hz flickering events (Brochard and Lennon, 1975). 2) Alternatively, actin filament assembly “pushing” on the lipid bilayer may stabilize the bilayer by enhancing membrane tension. Such a pushing mechanism might contribute to the spatial heterogeneity of the biomechanical properties of the membrane (Picas ). Future work will require careful biophysical approaches and analyses of RBCs with targeted deletions of actin-binding proteins to understand the structural basis for actin filament recruitment and force transmission that underlie RBC biomechanical properties and shapes.

MATERIALS AND METHODS

RBC handling

Human blood was drawn into BD Vacutainer tubes spray coated with K2EDTA at the Normal Blood Donor Service at the Scripps Research Institute (La Jolla, CA), according to an Institutional Review Board–approved protocol for blood collection from human subjects (11-5773). Freshly drawn blood was centrifuged for 10 min at 1000 × g, serum and buffy coat were aspirated off, and RBCs were resuspended in 20 volumes of 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)–buffered saline (HBS) optimized for human RBCs (145 mM NaCl, 5 mM KCl, 1 mM MgCl2, 10 mM glucose, 10 mM HEPES, 2 mM adenosine, pH 7.4). RBCs were washed three times by centrifugation at 1000 × g, aspiration of the supernatant, and resuspension in 20 volumes of HBS. Washed RBCs were then stored on ice for up to 2 d. In some experiments, membrane skeletons were prepared from washed RBCs or ghosts by addition of four volumes of Triton-lysis buffer (145 mM NaCl, 5 nM KCl, 1 mM MgCl2, 10 mM HEPES, 5 mM dithiothreitol [DTT], 2.5% Triton X-100, pH 7.4) and protease inhibitor cocktail (1:1000; Sigma-Aldrich, St. Louis, MO) to the packed cells, followed by centrifugation of the membrane skeletons at 15,000 × g through an underlaid shelf of Triton-lysis buffer plus 20% sucrose and subsequent separation of supernatants and pellets.

Actin-disrupting drugs

CytoD, LatA, and Jasp were purchased from Sigma-Aldrich, and stock solutions were prepared in DMSO and stored at −20°C until use.

Rho-actin incorporation

Rho-actin was prepared from purified rabbit skeletal muscle actin by labeling on cysteine 374 as described (Tait and Frieden, 1982; Littlefield ). Rho-actin was labeled at an efficiency of ∼0.3 mol rhodamine/mol actin and stored in liquid N2 at 5 mg/ml in a buffer containing 4 mM Tris, 0.1 mM CaCl2, 0.2 mM Mg2ATP, 2 mM DTT, and 0.02% NaN3 (pH 8.0). RBCs were washed three times in phosphate-buffered saline (PBS) at 4°C, resuspended to 10% hematocrit in PBS, and pretreated with 0.5 μM CytoD in 0.001% DMSO, DMSO alone, or no additions for 30 min at room temperature in PBS. RBCs were collected by sedimentation in an Eppendorf microfuge, the supernatant was removed, and RBCs were lysed by adding 10 μl of packed RBCs to 40 μl of 0.5 mM Mg2ATP and 10 mM sodium phosphate (pH 7.4) containing 2.4 μM rho-actin with or without 0.5 μM CytoD (final concentration) and incubated for 10 min on ice. Lysed RBCs were resealed by the addition of 5 μl of 1 M KCl and 10 mM MgCl2 for 30 min at 37°C, followed by the addition of 1 ml of cold PBS and centrifugation for 5 min at 12,000 rpm in an Eppendorf microfuge to remove excess extracellular rho-actin. Note that this and all subsequent wash buffers contained 500 nM CytoD or DMSO, as appropriate. Resealed ghosts were washed twice in PBS with 0.01% bovine serum albumin (PBS/BSA) at 4°C by resuspension and sedimentation and then resuspended in 800 μl PBS/BSA.

In some experiments, 200 μl of the ghost suspension was allowed to settle onto a carbonate-coated coverslip for 30 min at room temperature. Ultraclean “squeaky clean” coverslips were prepared as described (Waterman-Storer, 2001) and carbonate coated as described (Perlmann ). Excess fluid was aspirated off, and coverslips were washed in 200 μl of PBS/BSA, fixed in 1.6% paraformaldehyde (PFA) in PBS for 15 min at room temperature, and then washed twice in PBS. Ghosts were observed as wet mounts with a Zeiss Axioskop fluorescence microscope using a 63× Plan-Apochromat oil-immersion objective lens (numerical aperture [NA] 1.4) with an Optovar setting of 1.6. Images were collected with a 20-s exposure time using a charge-coupled device (CCD) camera and a Y1300 Interline chip (Princeton Instruments, Trenton, NJ). In other experiments, 10 μl of the ghost suspension was used to prepare membrane skeletons, as described, and subjected to Western blotting, as follows.

Western blotting

Washed RBCs in HBS were incubated for 4 h in a 37°C water bath in the presence of DMSO or the indicated drugs. The supernatants (soluble fractions) of Triton-lysis buffer–extracted RBCs were separated from the pellets (membrane skeleton fractions) as described. Soluble fractions were solubilized in 1/5 volume of 5× SDS sample buffer, boiled for 5 min, electrophoretically separated on 4–20% Tris-glycine gradient minigels for 1 h at 200 V, and transferred to nitrocellulose (pore size, 0.2 μm) in transfer buffer containing 20% methanol. Blots were stained with 0.2% Ponceau S in 3% trichloroacetic acid to control for loading, blocked for 2 h in 4% BSA plus 1% goat serum in PBS at room temperature, and then incubated in primary antibody diluted in Blitz buffer (4% BSA, 10 mM NaHPO4, 150 mM NaCl, 1 mM EDTA, 0.2% Triton X-100, pH 7.4) overnight at 4°C. Primary antibodies were as follows: mouse monoclonal anti-actin (C4, 1:10,000; EMD Millipore, Billerica, MA), mouse monoclonal anti-rhodamine (11H10, 1:1000; Santa Cruz Biotechnology, Dallas, TX), mouse monoclonal anti–β-actin (AC-74, 1:1000; Sigma-Aldrich), rabbit polyclonal anti-human Tmod1 (R1749bl3c, 1:1000; Moyer ), rabbit polyclonal anti–α-adducin (1:1000; a gift from Vann Bennett, Duke University, Durham, NC), and sheep polyclonal anti-TM5NM1 (AB5447, 1:1000; EMD Millipore). After washing in PBS plus 0.1% Triton X-100, blots were incubated in either horseradish peroxidase– or 680LT-conjugated secondary antibody diluted in Blitz for 1 h at room temperature. After washing again in PBS plus 0.1% Triton X-100, bands were visualized either using enhanced chemiluminescence followed by exposure to film or a LI-COR Odyssey infrared imaging system. Background-corrected band intensities were densitometrically quantified using ImageJ. Western band intensities were normalized to the sum of the intensities of the nonhemoglobin bands on the corresponding Ponceau S–stained blot.

TIRF microscopy

RBCs washed in HBS with glucose and adenosine (see earlier description) were incubated for 4 h in a 37°C water bath in the presence of DMSO or the indicated drugs. RBCs were then fixed overnight in 4% PFA at 4°C, washed three times in HBS, permeabilized for 10 min in 0.3% Triton X-100, blocked in 4% BSA plus 1% goat serum, and stained with Alexa 488–phalloidin (Life Technologies, Carlsbad, CA) for F-actin. After additional washing, RBCs were deposited onto glass slides using a Thermo-Fisher Cytospin 4 Cytocentrifuge at 1000 rpm for 3 min, and coverslips were mounted onto the slides using Fluoro-Gel aqueous mounting medium (Electron Microscopy Sciences, Hatfield, PA). Images were collected at room temperature on a Nikon Eclipse Ti inverted microscope with a 100× Apochromat oil objective lens (NA 1.49) and TIRF illumination in conjunction with a Photometrics CoolSNAP HQ2 CCD camera (Roper Scientific, Tucson, AZ). Images were collected using NIS-Elements 3.2 software (Nikon, Melville, NY) and processed using Volocity 5.3.2 software (Improvision, Waltham, MA).

FRAP

RBC actin filaments were labeled by incubating washed RBCs (1 × 109 RBCs at a hematocrit of 10%) in Ca2+- and Mg2+-containing Hank's balanced salt solution (HBSS) supplemented with 1 μM SiR-Jasp (Lukinavicius ) for 30 min at room temperature. In some experiments, RBCs were fixed in 0.5% acrolein for 5 min at room temperature before SiR-Jasp staining (Melhorn ). Cells were then washed twice, resuspended in HBSS containing 0.05% immunoglobulin G–free BSA, and seeded on a slide for 5 min before FRAP analysis. Images were acquired at a rate of 1 image/min using a 60× UPlanApo objective (NA 1.42) mounted on an Olympus BX62 microscope and a cooled QImaging Emc2 electron-multiplying CCD camera. FRAP was performed using a 488-nm Laser Vector Photomanipulation unit controlled by SlideBook 5.5 software (Intelligent Imaging Innovations, Denver, CO) and analyzed using Slidebook's FRAP analysis module, as described previously (Melhorn ). At least 20 RBCs were analyzed for each condition. RBCs that did not have a standard biconcave shape were excluded from analysis.

Osmotic fragility

Lysis of intact RBCs as a function of osmolarity was measured as previously described (Gilligan ; Moyer ). Briefly, washed RBCs were treated with DMSO or the indicated drugs and then diluted to a final hematocrit of 5% in HBS. Small volumes of the RBC suspension (10 μl) were diluted in buffers of known osmolarities (range 0–280 mOsm/l, pH 7.4), incubated at room temperature for 20 min, and centrifuged for 5 min at 1000 × g. Supernatants were removed and transferred to 96-well plates, and percentage lysis was calculated from the absorbance at 540 nm.

Microfluidic measurement of RBC deformability

Two-dimensional microfluidic arrays were used as previously described (Ghiran ). Briefly, washed RBCs (5 μl) were loaded into the inlet reservoir of a microchannel and driven into a capillary-like area by lowering the reservoir tube on the opposite side of the apparatus. Once RBCs reached the microchannel, the reservoir was raised to an appropriate height that allowed control RBCs from healthy donors to pass through the length (25 μm) of the capillary in ∼3 s. The cells were video recorded using a 40× Ph2 Plan Fluorite objective (NA 0.75) mounted on a TE300 Nikon inverted microscope, using a QImaging Retiga EXi CCD camera controlled by iVision 4.7 software (BioVision Technologies, Exton, PA) at a rate of 10 frames/s. Movies were analyzed offline, frame by frame, and any RBCs that were unusually shaped, overlapped, or clustered were excluded from measurement. At least 30 RBCs were counted for each experimental condition.

Membrane flickering analysis

Spontaneous RBC membrane oscillations (flickering) were measured as previously described (Ghiran ). Positive-low, phase-contrast, time-lapse images of RBCs seeded on microscope slides were recorded for 10 s at a rate of 33 frames/s using a 100× UPlanApo phase contrast objective (NA 1.35) on an Olympus BX62 microscope. Flickering was measured using iVision 4.0.9 software (BioVision Technologies). At the end of each recording, an intensity projection step of the image stack was performed to identify and exclude RBCs that drifted during recording. The intensity of scattered light was used to calculate, pixel by pixel, the coefficient of variance at each point within the RBC and display the results as a pseudocolor amplitude map.

Statistics

Significant differences between two groups were detected using Student's t test. Significant differences among three or more groups were detected using one-way analysis of variance with post hoc Fisher's protected least significant difference tests. Statistical analysis was performed in Excel (Microsoft, Redmond, WA). Significance was defined as p < 0.05.