Structural and functional studies of erythrocyte membrane-skeleton by single-cell and single-molecule techniques

As the indispensable oxygen-transporting cells, erythrocytes exhibit extreme deformability and amazing stability as they are subject to huge reversible shear stress and extrusion force during massive circulation in the body. The unique architecture of spectrin-actin-based membraneskeleton is considered to be responsible for such excellent mechanical properties of erythrocytes. Although erythrocytes have been recognized for more than 300 years, myriad questions about membrane-skeleton constantly attract people's attention. Here, we summarize the kinds of distinctive single-cell and single-molecule techniques that were used to investigate the structure and function of erythrocyte membrane-skeleton at macro and micro levels.


Introduction
Erythrocytes are the largest number of blood cells that deliver oxygen from the lungs to the body tissues. The erythrocytes possess a unique biconcavedisk pro¯le that provides excellent deformability and stability to survive the huge shear force and extrusion force during massive circulation in their life. It has been suggested that the characteristic architecture of erythrocyte membrane-skeleton is responsible for such extreme mechanical deformability. 1,2 The organization of membrane-skeleton is rather complex, and current models depict that the membrane-skeleton is principally composed of a lipid bilayer with an attached skeleton formed by a two-dimensional triangular network of the spectrin tetramers linked by junctional complexes consisting of short actin¯laments, protein 4.1, adducin, tropomodulin, tropomyosin and other associated proteins ( Fig. 1). [3][4][5] This membrane-skeleton plays an important role in maintaining the physiological functions of erythrocytes. It has been demonstrated that the deformability decrease resulted from membrane-skeleton defects underlie the generation of various diseases such as diabetes, 6 anemia, 7 spherocytosis, 8 malaria infection, 9 and so on.
The considerable progresses in the knowledge of the structure and function of erythrocyte membrane-skeleton deeply rely on the application of advanced single-cell and single-molecule techniques. For instance, the optical tweezers, 10 micro°uidic chips 11 and micropipette aspiration 12 have emerged as potent methods for investigating membraneskeleton-dependent deformability of erythrocyte at single-cell level. Meanwhile, the sub-cellular nanoscale organization of the membrane-skeleton has been well studied by electron microscopy (EM) 13,14 and atomic force microscopy (AFM). 15 Furthermore, multiple techniques including immuno-electron microscopy (iEM), 16 antibody recognition imaging by AFM 17 as well as super-resolution°uorescence microscopy (SRFM) 18 provide new opportunities to uncover the ultrastructure of the erythrocyte membrane-skeleton at single-molecule level.
The present review focuses on the current understanding of the structure and function of erythrocyte membrane-skeleton based on single-cell and single-molecule level methods.

Membrane-Skeleton-Dependent Deformability of Erythrocyte Studied by Single-Cell Techniques
Devoid of nucleus and intracellular organelles, the extreme deformability of erythrocyte predominantly relies on its unique architecture of membraneskeleton system. In other words, the erythrocyte deformability could directly and e®ectively reveal the membrane-skeleton properties. The better deformability corresponds to healthier organization of membrane-skeleton.

Optical tweezers
Optical tweezers,¯rst achieved in 1986 by Ashkin who was rewarded Noble Prize in 2018, trapped and manipulated dielectric particles by laser beams based on radiation forces of pico-newton scale. 19 As a powerful tool to manipulate atoms, 20 particles 21 and cells, 22 optical tweezers are¯nding increasingly widespread applications in the investigation of various physical and biological processes. Speci¯cally, optical tweezers have shown tremendous potential in measurement of erythrocyte deformability at single-cell level. Early optical tweezers consisted of a single-beam optical trap formed by a laser to manipulate erythrocytes. Individual erythrocytes were stably suspended inside a°ow chamber by application of a single-beam optical tweezers, where the bending rigidity of the erythrocytes was observed to infer the deformability. 23 On the other hand, the erythrocyte deformability could be studied when single-beam optical tweezer dragged erythrocyte at a constant velocity ( Fig. 2(a)). 24 In addition, the ATP-dependent nanoscale membrane°uctuations, an indicator of local erythrocyte membrane-skeleton mechanical properties, have been investigated by various techniques including°ickering spectroscopy, positive phase contrast microscopy and point dark¯eld microscopy. [25][26][27] Furthermore, application of single optical tweezer enabled a highly accurate measurement of the erythrocyte membrane°uctuation amplitudes with microsecond temporal and sub nanometer spatial resolution through laser scattering detection. 28 The optical stretcher, a variant of optical tweezers, uses two nonfocused laser beams from opposite directions to stably trap the erythrocyte ( Fig. 2(b)). The erythrocyte deformability was assessed by shape change when it was stretched out along the laser beam axis. 29 The radiation damage to the cells can be ignored for optical stretcher owing to the defocused beams, which minimizes the light°ux through the cells in comparison to other optical traps using focused lasers.
Distinguishing from the optical stretcher with two counter-propagating-beams, the dual-trap optical tweezer uses two parallel trapping beams that focused in the interior of the erythrocyte. The erythrocyte is stretched when the two beams are separated by a distance (Fig. 2(c)). This developed optical tweezers provided precise control to the erythrocytes, and the stretched length of erythrocytes was measured to assess the deformability. 30,31 Furthermore, a more complex optical tweezer with three optical traps was designed to manipulate erythrocytes from three dimensions simultaneously. 32 In addition, optical tweezers can also manipulate erythrocytes through microbeads attached to them as shown in Fig. 2(d). Based on this technique, the erythrocyte deformability was indicated by the change of projected diameter of the erythrocyte. 33 The elasticity of erythrocyte membrane-skeleton was investigated through applying force to beads attached to erythrocyte ghosts. 34 The relaxation response of the erythrocytes stretched by this optical tweezer was also recorded to assess the deformability by the relaxation time after large deformation. 35 This microbeads-based optical tweezer could induce larger deformation of erythrocytes than that of pure optical tweezers.
Diversi¯ed optical tweezers have been developed to study erythrocyte deformability in di®erent conditions including malaria infection, 36 sickle cell disease, 37 osmolality change 38 and trapping in living animals. 39 It found that erythrocyte deformability underwent a decrease in pathological conditions. 36,37 These works may contribute to an in-depth understanding of the mechanical properties of erythrocyte membrane-skeleton.

Micropipette aspiration
Since the¯rst use of micropipette aspiration for measuring the elastic properties of sea urchin eggs in 1954 by Mitchison and Swann, 40 this method has been widely used in many types of cells, such as neutrophil cells, 41 endothelial cells 42 and chondrocyte cells. 43 In particular, the micropipette aspiration permits quantitative analysis of the erythrocyte deformability. 44 Depending on the diameter of the micropipette and the aspiration pressure, there are three types of erythrocyte aspiration, including aspiration of a small hemispherical portion of membrane-skeleton ( Fig. 3(a)), an aspired tongue within the micropipette ( Fig. 3(b)), and an aspiration of the whole cell into the micropipette (Fig. 3(c)). When a small hemispherical portion of the erythrocyte was aspired into the micropipette, the erythrocyte deformability was assessed by the surface elastic moduli, which could be calculated through the size of the aspired portion and the applied pressure. 45,46 When erythrocyte was aspirated into a micropipette with a long tongue, the shear modulus of membrane-skeleton could be determined by the tongue length. 47 The actin¯laments of the membraneskeleton were found to be obviously modi¯ed when erythrocyte was aspirated with a long tongue inside the micropipette. 48 When the suction pressure increased large enough, the aspirated erythrocyte folded into the micropipette, in which case the bending elastic modulus could be derived from the applied pressure. 49 In comparison with optical tweezers, the micropipette aspiration could supply a local deformation on the erythrocyte membrane-skeleton. Meanwhile, the mechanical properties of the membrane-skeleton can be quantitatively revealed when the micropipette aspiration method combines with theoretical models.

Micro°uidic channels
The micro°uidic channels have also been demonstrated as a potent and useful method for investigation of erythrocyte deformability. Various parameters on the morphology of erythrocytes in micro°uidic channels can be quanti¯ed to assess the deformability. For instance, deformation index (DI ¼ L=D, L is the height, and D is the width of the parachute, Fig. 4(a)) was de¯ned to measure erythrocyte deformation when they folded into a parachute-like shape in the micro°uidic channels. 50,51 The cortical tension, another indicator of the deformability, could be calculated by the deformed erythrocytes when they extrude through a microchannel with repeated funnel pores. 52,53 The transit velocity of the erythrocyte in the micro-°u idic channels was often used to assess the deformability. In detail, a microchannel with repeated taper constrictions was designed to detect deformability di®erence between healthy and P. falciparum-infected erythrocytes based on transit velocity measurement ( Fig. 4(b)). 54 A microchannel with periodically triangle-shaped pillars was fabricated to study the deformability of iron dextrantreated erythrocyte through the transit velocity calculation. 55 A narrow straight micro°uidic channel was applied to investigate the e®ects of drugsinduced actin¯laments assembly/disassembly on erythrocyte deformability determined by alterations in cell transit time. 56 Furthermore, micro°uidic channels possess a unique ability to simulate capillary microcirculation system in vitro, 57,58 especially the blockage of capillaries. For example, the behavior and morphology of poorly deformable erythrocytes in capillary blockages were simulated and described using microchannels with di®erent width. 59 The degradation of the erythrocyte deformability was measured through the applied pressure and°ow rate inside the multi-channel micro°uidic network (Fig. 4(c)). 60 In addition, the deformability of erythrocytes was characterized through the capture status in a biomimetic micro°uidic chip with mechanical constraints that was designed to simulate the human spleen slits (Fig. 4(d)). 61 Based on application of di®erent micro°uidic channels, they all suggested that erythrocyte deformability decreased signi¯cantly under various abnormal conditions, such as malaria-infection, 62,63 blood storage lesion, 64,65 diabetes 66,67 and sickle cell disease. 68,69 Compared with the optical tweezers and the micropipette aspiration, the micro°uidic channels provided a unique opportunity to investigate erythrocyte deformability in a rapid, highthroughput and relatively easy-to-use method at single-cell level.

Ultrastructure of the Erythrocyte Membrane-Skeleton Studied by Single-Molecule Techniques 3.1. Immuno-electron microscopy (iEM)
Taking advantage of the ultrahigh imaging resolution of EM, the sub-cellular ultrastructure of erythrocyte membrane-skeleton has been primarily described by di®erent kinds of EM. 13,14,16,[70][71][72] Early results from negative staining EM (Fig. 5(a)), 70 as well as quick-freezing, deep-etching and rotaryreplication (QFDERR) (Fig. 5(b)) 16 clearly showed that the erythrocyte membrane-skeleton formed a dense triangular network of intersecting straight laments linked by rod-like junctional complexes at the converging nodes at sub-cellular level. Interestingly, the junction-to-junction distances of the triangular network were 200 nm in spread erythrocyte membrane-skeleton 13 and 30-40 nm in nonspread membrane-skeleton. 72 Furthermore, a work used cryo-electron tomography to evaluate the three-dimensional topology in intact, unexpanded erythrocyte membrane-skeleton, which revealed a complex¯lament meshwork with an average edge length of 46 nm (Fig. 5(c)) and a gradual decrease in both the density and the thickness of the network from the center to the edge of the cell at sub-cellular level. 14 Recently, cryo-electron tomography was used to reveal the remodeling of actin¯laments in Plasmodium falciparum-infected erythrocytes. 73,74 The development of iEM, which combines immunogold labeling with above EM techniques, enables the single-molecule measurements of erythrocyte membrane-skeleton architecture. 16,72,75 Conventional electrophoresis and biochemical analysis only showed that erythrocyte membraneskeleton contains 20 major proteins such as spectrin, actin, tropomodulin, band 3, protein 4.1, ankyrin, etc. without actual spatial distribution information. 76,77 Labeling the membrane-skeleton with gold particles that coated with site-speci¯c spectrin antibodies veri¯ed that the¯laments in the meshwork were spectrin molecules based on negative staining EM (Fig. 5(d)) 16  Immunogold labeling of band 3 molecules further demonstrated that integral membrane proteins were organized into macromolecular complexes centered on band 3 protein, which were anchored to the membrane-skeleton. 72 Moreover, immunolocalization of tropomodulin (Fig. 5(e)), actin (Fig. 5(f)) and tropomyosin (Fig. 5(g)) in spread erythrocyte membrane-skeleton provided direct evidence that short¯laments at the central junctions were indeed actin¯laments, and both tropomodulin and tropomyosin were associated with the actin¯laments. 78 iEM could distinguish normal biconcave erythrocytes from abnormally-shaped erythrocytes obtained from patients by distinct patterns of spectrin¯laments. 79

Atomic force microscopy (AFM)
Compared with EM, normal AFM has the advantage of in situ imaging and high sensitivity to small height variations in surfaces at sub-cellular level, thereby it has been widely used in measuring the sizes, shapes and mechanical properties (such as sti®ness, viscoelasticity, hardness and adhesion) of erythrocyte. [80][81][82] AFM nano-mapping revealed that pathological erythrocytes usually lost the unique biconcave-disk shape and exhibited decreased membrane roughness as well as increased membrane sti®ness compared with healthy erythrocytes. [82][83][84] Furthermore, AFM could examine the nanoscale ultrastructure of the erythrocyte membrane-skeleton. For instance, the topography of the membraneskeleton on the cytoplasmic surface of the erythrocyte plasma membrane was directly observed in freeze-dried ghosts at the external surface of the cell without removing the membrane or extending the membrane-skeleton. 17 Additionally, another work demonstrated the applicability of AFM for imaging the two-dimensional membrane-skeleton of untreated as well as¯xed human erythrocytes under near physiological conditions (Fig. 6(a)). 15 The correlation of AFM with single-molecule recognition imaging (also known as topography and recognition imaging, TREC), in which the AFM tip is tethered with speci¯c molecules to recognize the counterpart molecules on the surface (Fig. 6(b)), expands the single-molecule level applications of AFM. 85,86 These recognition pairs can be antigens and antibodies, hemagglutinin and oligosaccharides, ligands and receptors, enzymes and substrates, etc. Applying this combined method, Takeuchi et al. immunogold labeled spectrin with speci¯c antibodies and demonstrated that the threedimensionally folded meshwork was indeed the spectrin network (Fig. 6(c)). 17 In addition, single molecule force spectroscopy (SMFS), a powerful approach derived from TREC imaging with AFM, can study protein interactions at the single-molecule level by detecting tens of piconewton forces. Based on SMFS, Wang et al. observed the membrane proteins of erythrocyte and proposed a novel semi-mosaic model for the erythrocyte membrane organization. [87][88][89] Although application of SMFS sheds new light on the studies of erythrocyte membrane proteins, the anatomy of erythrocyte membrane-skeleton at single-molecule level by AFM still lack a satisfactory level of understanding up to date.

Super-resolution°uorescence microscopy (SRFM)
According to the results of EM and AFM, it is remarkable to note that major discrepancies exist for the length of spectrin tetramers of the membraneskeleton. [13][14][15][16][17] Above studies had di±culty in obtaining the native ultrastructure of the erythrocyte membrane-skeleton due to the extensive sample processing, in which samples were often dried and/or membrane removed. Recently, the rise of SRFM provides new opportunities to visualize actual cellular ultrastructure with nanoscale resolution at single-molecule level with minimal sample damage. 90,91 Three-dimensional stochastic optical reconstruction microscopy (3D-STORM) has been demonstrated as a powerful SRFM to investigate cytoskeletal biology. [92][93][94] Our recent work revealed the native ultrastructure of the erythrocyte membrane-skeleton in membrane-preserved cells based on 3D-STORM (Fig. 7). 18 We achieved molecular speci¯city for six targets, namely the N-and C-termini of -spectrin, protein 4.1, F-actin, tropomodulin (TMOD) and adducin. We¯rst developed an alternative method, in which live erythrocytes were¯rst allowed to adhere to a polylysine-coated coverslip before subse-quent¯xation and labeling ( Fig. 7(a)). It showed that the bottom membrane-skeleton was°at down to the axial resolution limit, which allowed us to obtain optimal 3D-STORM image quality (Figs. 7(b) and 7(c)). Based on STORM data of N-termini of -spectrin, we revealed a 80 nm junction-tojunction distance depending on the analysis of the distance between nearest neighbors (Figs. 7(d) and 7(e)) and two-dimensional autocorrelations (Figs. 7(d) and 7(e)), a length in agreement with relaxed spectrin tetramers under equilibrium, 95 as well as theoretical predictions based on the abundance of spectrin in erythrocytes, 96 but contrasting sharply with previous results from dried or membrane-removed erythrocytes by staining EM cryo-electron tomography (46 nm), 13,14 as well as from neurons under similar super-resolution settings (190 nm). 93 Interestingly, though presented at lower densities, statistics of distances between nearest neighbors still yielded peaks at 80 nm for actin¯laments and actin-capping proteins, TMOD and adducin, indicating that the actin-related targets occupy a subset of the same junctional complexes. Furthermore, crosscorrelation calculation between N-and C-termini of -spectrin indicated a 80 nm junction-to-junction distance (Figs. 7(f)-7(h)). Meanwhile, we found that the dense erythrocyte membrane-skeleton often contained nanoscale voids (Figs. 7(f) and 7(g)), which may behave as structural weak points to facilitate fast changes of the erythrocyte shape during huge circulation. In addition, crosscorrelation calculation between the protein 4.1 and TMOD showed a maximum value at zero intermolecular distance (Figs. 7(i)-7(k)), thus con¯rming co-localization at the nanoscale suggested from the in vitro interactions of puri¯ed proteins. In brief, our STORM data thus call for both experimental and theoretical reassessments of the structure and function of the native erythrocyte membrane-skeleton. Although our results could reveal the native organization of the erythrocyte membrane-skeleton depending on much more careful data analysis, it must admit that the STORM images fail to show direct visual pattern of membrane-skeleton as seen by the EM and AFM.
Recently, Hu® from Zeiss Company developed a new form of super-resolution microscopy named AiryScan confocal°uorescence microscopy. 97 Smith et al. used this super-resolution microscopy to image nonmuscle myosin II (NMII) motors in intact erythrocytes. 98 Results indicated that NMII could form bipolar¯laments, which associate with membraneskeleton F-actin. Furthermore, NMII motor activity modulates interactions with the spectrin-F-actin network to control erythrocyte biconcave disk prole and deformability. This work provided a previously undescribed mechanism for actomyosin regulation of membrane tension, curvature, cell shape and biomechanics.

Conclusion
In conclusion, this paper summarizes the current understanding of the structure and function of erythrocyte membrane-skeleton based on kinds of distinctive single-cell and single-molecule level techniques. Each technique we mentioned has its advantages and also disadvantages, thereby making the integrated application of various techniques, such as correlation of EM and SRFM, to be helpful for researchers in further investigation of the unanswered questions underlying the structure and function of erythrocyte membrane-skeleton.

Con°ict of Interest
The authors declare no con°ict of interest.