Afma nano tool in membrane biology

Absence of ABCA4 does not significantly alter the structure of the disk rim scaffold Figure 4 , and we therefore suggest that given the compact and ordered structure of the scaffold, ABCA4 is likely not a component of it Figure 5C.

Hence, we propose ABCA4 resides next to the scaffold at the disk rim. However, whether ABCA4 interacts with the rim scaffold or locates at a distance from it remains unclear. Recently, AlphaFold2 has proven its ability to predict the 3D structure of not only monomeric proteins with high accuracy Jumper et al. At this point, both observations are difficult to reconcile. Hence, we can only hypothesize that the repeats resolved in our average are smaller PRPH2-ROM1 complexes which oligomerize to form the disk rim scaffold and enforce the high membrane curvature at the disk rims.

However, a density map with near-atomic resolution would be required to clarify whether the repeats are composed of PRPH2-ROM1 dimers or tetramers. Non-covalently bound PRPH2-ROM1 complexes are known to form higher order oligomers stabilized by disulfide bonds Loewen and Molday, , which are essential for normal disk morphogenesis Milstein et al.

We hypothesize that these disulfide bonds Figure 5B , are responsible for the contacts across rows Figure 3 or between repeats of the peripheral rows Figure 3—figure supplement 3C.

Docking the predicted model of the PRPH2 dimer into the repeats of the central row, however, reveals that the two PRPH-C cysteine residues are not located where we observe these contacts but closer toward the inner membrane leaflet Figure 3—figure supplement 5B , Video This may be explained by errors in the prediction or the dimers assuming a different conformation upon oligomerization and embedding into the highly curved membrane environment of the disk rim.

We could not resolve this heterogeneity as additional structures to sufficient quality by subvolume averaging, but in combination with the inherent flexibility of the disk rim, it might be the reason for the restricted resolution of our averages.

Instead, our analysis in situ resolves three rows of repeats which are also linked by the luminal domain but are rather organized side-by-side Figure 5A. Furthermore, our results raise the question whether the basic building blocks of the disk rim scaffold are PRPH2-ROM1 dimers or tetramers. We propose a mechanism for disk rim curvature formation, where the two diverging transmembrane densities of V-shaped PRPH2-ROM1 complexes displace lipids in the inner membrane leaflet.

While unlinked complexes are able to induce some membrane curvature, their oligomerization into three continuous rows is required to force the membrane into this elongated, highly curved geometry Milstein et al. Our data indicate that the luminal domains of complexes hold the disk rim scaffold together Figure 3C , which is consistent with the fact that most pathological mutations of PRPH2 affect its luminal domain Boon et al. In good agreement with previous work, it is possible that these mutations impair the formation of complexes and their disulfide bond-stabilized oligomerization Chang et al.

Hence, these alterations could impede or completely prevent disk morphogenesis which, in turn, would disrupt the structural integrity of ROS, compromise the viability of the retina and ultimately lead to blindness. To minimize the interval between dissection and plunge-freezing, only one mouse was used for each preparation.

The mouse was euthanized by exposing it to CO 2 for 3—5 min followed by cervical dislocation. The first eyeball was excised with curved scissors and glued Scotch Single-use super glue gel with its sclera side down to a plastic Petri dish. The eye was dissected as follows. First, a slit was made with a scalpel blade, and one blade of fine scissors inserted into the slit. The cornea was cut away and then the lens removed with fine forceps. The retina was transferred into a 1.

To prevent damaging the retina during transfer, the opening of the pipet tip was widened by cutting off its tip. The same procedure was applied to the second eye. The retinas were vortexed at rpm for 1 min to detach ROS. The centrifugation step enriched ROS in the supernatant which was transferred into a fresh tube Figure 1—figure supplement 1A , B.

The combined supernatant was gently mixed by repetitive pipetting four times. The resulting sample was used for plunge-freezing. The total extraction time was 10—20 min. The grids were blotted with a filter paper and a Teflon sheet from the reverse and front side, respectively. Grids were stored in liquid nitrogen until use. Plunge-frozen grids were fixed into custom-made autogrids, mounted into a shuttle Rigort et al. Lamellae were prepared using a Gallium ion beam at 30 kV.

FIB-milling was performed in a stepwise manner using rectangle patterns following similar procedures as in Schaffer et al. For the final cleaning step, a low current of 50 pA was used to obtain lamellae thinner than nm Figure 1—figure supplement 1D , F. For improved conductivity of the final lamella, the grid was again sputter-coated after cryo-FIB preparation with platinum in the Quorum prep-chamber 10 mA, 1 s as previously reported in Mahamid et al.

Cryo-transmission electron microscopy observations were performed using a Titan Krios operated at kV Thermo Fisher Scientific. The individual projection images were recorded as movies dose fractionation mode on a K2 Summit Gatan direct electron detector camera operated in counting mode with an image pixel size of 2.

A fraction of the tomographic tilt-series in this work were acquired with the VPP Danev et al. Alignment and operation of the Volta phase plate were carried out as described previously Fukuda et al.

During automated tilt-series acquisition an autofocusing routine was performed using zero defocus offset with 5 mrad and 10 mrad beam tilt for conventional tilt series and data acquisition with VPP, respectively. Tomographic tilt-series were collected using standard automated acquisition procedures. All datasets are listed in Table 1. The montage tiles were aligned using the IMOD version 4.

Each lamella contained several ROS. In some cases, the ROS ultrastructure was partially distorted. As the distortions were locally confined, tilt-series were exclusively recorded in areas with ROS unperturbed by the sample preparation Figure 1—figure supplement 2A. For each mouse strain and acquisition scheme, data was collected on samples derived from at least three different mice. Prior to tilt-series alignment, the projection images were corrected for beam-induced motion with MotionCor2 Zheng et al.

Tilt-series alignment and tomographic reconstructions were performed using the IMOD Mastronarde and Held, software package version 4. Platinum particles originating from the protective platinum layer which were deposited over the lamella surface during FIB-milling served as fiducials Figure 1—figure supplement 2B.

Final alignment of the tilt-series images was performed using the linear interpolation option in IMOD. In these cases, the density appears dark, that is with a low gray value. To measure the repetitive distances of ROS disk membranes, contours of varying length perpendicular to the disk stack were defined in the disk interior e.

The contours were generated in 3dmod by opening the tomographic volume in the Zap window and creating a new model. Each contour included two points spanning across at least two ROS disks.

Cuboids were cropped along these contours Figure 1—figure supplement 3B. The cuboid voxels were averaged along the base area to obtain a 1D intensity profile of length h Figure 1—figure supplement 3C.

For the thickness calculation of the plasma membrane PM d PM , a total of subvolumes were extracted from five tomograms along the PM, aligned, and subvolume averages were calculated for each tomogram.

A similar approach was used to compute the maximum diameter of the disk rim d DR parallel to the ROS cylinder axis. A total of subvolumes from six tomograms were aligned and averages calculated for each tomogram. To calculate the width of the cytosolic gap at the disk incisure d IN and the distance between the PM and the disk rim d PR , the refined coordinates of the disk rim subvolumes were utilized.

The value of d Shift was determined as the distance from the center of the subvolume average to the outer periphery of the disk rim along L Figure 1—figure supplement 4B. The subvolumes were separated into three groups: group 1 and group 2 comprised subvolumes on opposite sides of the disk incisure; group three contained subvolumes close to the PM.

Only the distance between neighboring disk rims Distance G in Figure 1—figure supplement 5 was not directly measured in the tomograms but calculated as the difference between the unit cell distance and the maximum disk rim diameter Distance B and H, respectively, in Figure 1—figure supplement 5.

Besides subvolume averaging, the distance calculations and the required image processing steps were performed in MatLab aided by the TOM software toolbox Nickell et al.

More dense structures, like proteins, appear darker in tomographic slices which translates into a lower gray value. The results of the automated segmentation and the original tomograms were loaded in Amira v. By comparing the two volumes, segmented patches which did not correspond to membranes were identified and manually removed. Afterwards, neighboring disk membranes of adjacent disks were grouped into pairs Figure 2—figure supplement 1C. The results of the initial automated membrane segmentation correspond to the central membrane plane.

By adding a layer of three voxels on either side of the central plane, the segmentation was grown to a thickness of 7 nm. Additionally, these masks defined the borders of the cytosolic gap between disks which a connector must bridge.

The cytosolic voxels between the membrane masks were normalized separately for each membrane pair to a mean value of zero and a standard deviation of one. This extinguished gradients in the gray value distribution throughout a tomogram caused by heterogeneous lamella thickness and compensated for contrast differences between tomograms. The original Pyto workflow segments connectors between the membranes of adjacent disks by evaluating all cytosolic voxels between the membrane masks as described below.

At each iteration i, the algorithm performs a connectivity segmentation by selecting j groups of voxels v i j based on four conditions:. This defines a relationship among all connectors picked at the individual gray value steps. The connector segmentation as output contains only independent groups of voxels which do not have ancestors.

The original Pyto workflow is sketched in Figure 2—figure supplement 1A. Visual inspection of the segmented connectors and their comparison to the densities observed in the raw tomograms, however, revealed that fewer connectors with a higher volume than expected were segmented Figure 2—figure supplement 2A , D.

This difference is caused by several interconnected elements which were segmented as one connector. Therefore, we customized the original Pyto workflow by applying an additional mask to the tomographic volume prior to the connector segmentation Figure 2—figure supplement 1B. First, a binary mask was created that is one for all voxels with gray value below g max , and elsewhere zero.

Third, a volume with the watershed lines set to zero and elsewhere one was multiplied with the binary mask. The resulting mask was applied to the original tomographic volume. Then Pyto was used to segment connectors in the masked tomogram. A sketch of the customized Pyto workflow and its processing steps applied to the data for one membrane pair are depicted in Figure 2—figure supplement 1B and C, respectively.

The threshold ramp for the original and the customized Pyto workflow was always started at the minimum gray value g min of —2 and ended at maximum gray value g max of —0. The manual segmentation of connectors was performed as follows: initially, the membranes in the tomograms were masked as done for the automated segmentation. Tomographic volumes with the membrane mask applied were loaded into Amira v. To assess the quality of the customized Pyto segmentation approach, the results were compared to the manual segmentation Figure 2—figure supplement 2C , D.

Two major differences are apparent: First, the connectors selected automatically were bulkier than manually picked connectors. This is caused by the Pyto algorithm that picks voxels based on their gray value and their connectivity and evaluates all voxels at once, not in a slice-by-slice manner Figure 2—figure supplement 2E.

Second, fewer connectors were picked manually. This is likely due to inclined structures, which were not observed as continuous connectors in one single tomographic slice, but several successive slices. Consequently, they could be missed manually Figure 2—figure supplement 2E.

Therefore, picking of connectors with the automated segmentation approach is more reliable than the manual segmentation. Ninety percent of the connectors were picked by both methods and the error of the determined connector coordinates was below 2 nm.

This error is small compared to the pixels size of 1 nm and the size of membrane patches with diameters of — nm. Therefore, the shape of the automatically segmented connectors may not be reliable, but their abundance and arrangement in 3D can be quantitatively analyzed.

The tomograms were selected based on a good IMOD tilt-series alignment scores and visual confirmation of well-resolved densities between ROS disks. The connectors and the membrane surface area were divided into two fractions. The disk rim fraction was within 40 nm from the outer periphery of disks rims. The remainder was considered the disk interior fraction.

Based on this definition, connectors were assigned as the disk rim connectors and as disk interior connectors. The local connector concentrations in the membrane fractions were calculated as the number of connectors n fraction per surface area A fraction :.

To compare the determined local concentrations with literature values for ROS proteins, the connector concentrations per full disk membrane were calculated. The total disk membrane area A tot was estimated based on the morphological considerations specified in Figure 2—figure supplement 3A according to:.

The total area of the fractions per disk A fraction tot were evaluated based on the distance threshold of 40 nm from the rim and the assumptions in Figure 2—figure supplement 3A :. The division by two was introduced because a connector links two membranes. Therefore, the segmentation approach detects each connector effectively twice, in contrast to a density attached to only one membrane. The connector density was calculated for each tomogram separately. To do the spatial analysis, each connector was assigned with a central coordinate C con located in the center between the two neighboring membranes Figure 2—figure supplement 3B.

A coordinate based on the center of mass of all connector voxels would result in off-center positions Figure 2—figure supplement 3B which would induce errors in the spatial analysis. Nearest-neighbor distances between connectors were calculated based on C con.

To estimate the connector length L con , the two membrane contact points P mb1 and P mb2 of a connector with both disk membranes were determined Figure 2—figure supplement 3B. L con was calculated as the sum of the distances between the central coordinate and the two contact points according to:. The mean grey value was defined as the average gray value of all connector voxels. The statistical significance of differences between disk rim and disk interior connectors was established with the two-sample Kolmogorow-Smirnow test in MatLab.

Initially, binned subvolumes were extracted from dose-weighted and, if possible, CTF-corrected tomograms. Phi is the angle of the first in-plane rotation around the z-axis.

Theta describes the second rotation around the new x-axis, and Psi the third rotation around the new z-axis. For the gray value representation of subvolume averages, the scale was inverted compared to the raw tomograms. Therefore, density in slices through subvolume averages appears bright, translating into a high grey value.

The initial subvolume extraction points of connectors were defined at their two membrane contact points P mb1 and P mb2 Figure 2—figure supplement 3B as elucidated by the segmentation. Initial Euler angles for Psi and Theta were determined so that the subvolume z-axis was parallel to the local normal vector of the disk membrane. The Phi angles were randomized. The initial alignment brought the membranes into register and refined the initial orientations.

The averages as a result of this alignment indicate a clear density protruding from the membrane into the cytosol, but it appears fuzzy compared to the membrane signal Figure 2—figure supplement 4A and D, for disk rim and disk interior connectors, respectively.

Hence, a different classification approach was chosen. Then, for each subvolume rotational averages around the z-axis were calculated and the resulting 2D images classified. A cylindrical mask focused the classification on the cytosol between disks. Subvolumes of classes indicating a density between the membranes were considered most promising Figure 2—figure supplement 3C. The resulting averages, however, remained featureless and revealed no further structural insights Figure 2—figure supplement 4C , F.

The high rate of putative false-positives indicated by the classification of the connector subvolumes suggests that our segmentation approach is error prone. Most likely, because the segmentation algorithm cannot distinguish two densities in close proximity protruding from opposite disks into the cytosol from an actual connector.

On the other hand, we obtain classes with elongated densities that appear to link neighboring disks. Therefore, we assume that the two types of connectors indeed exist, yet at lower concentration than the initial segmentation determined Figure 2H. Particularly, the disk rim connectors which are frequently seen in our tomograms Figure 2A and B and have been observed before Corless et al.

Splines were manually picked along disk rims. For that, the tomographic volume was visualized in the 3dmod ZAP window, and a new model created. A new contour was defined for each disk rim by adding points along its outer periphery. Initial subvolume extraction points were set along the splines with 1 nm distance. Initial Euler angles for Psi and Theta were assigned so that the local spline direction dictated the orientation of the subvolume z-axis.

The Phi angles of the in-plane rotation were randomized Figure 3—figure supplement 1A. The initial average was composed of a strong density along the z-axis Figure 3—figure supplement 1A. Later, this search was refined. For the initial reference, a subset of subvolumes was aligned against the unstructured, first average. After several iterations, the symmetry was broken until the average converged into the hairpin-like structure of the disk rim.

This initial reference was then used to align the whole dataset. During this step, the subvolume positions converged to the disk rims and a first estimate for all three Euler angles was obtained Figure 3—figure supplement 1B.

This information was used to perform so-called distance-cleaning. At each of the lattice points, the particle with the highest similarity to the subvolume average, as estimated by the cross-correlation score, was kept and all others were discarded, which resulted in a minimal distance of 4 nm between subvolume coordinates. To potentially take the symmetry of the repeats into account, the subvolumes were reoriented by rotating the subvolume z-axis to point toward the disk center into the disk lumen, and the y-axis parallel to the ROS cylinder axis Figure 3—figure supplement 1C.

For this step, the tomograms of the WT conv dataset were preprocessed in Warp version 1. Instead of using the entire preprocessing capabilities implemented in Warp, the motion corrected, non-CTF-corrected, and non-dose-filtered projections of the tilt-series were imported into Warp with the corresponding tilt-series alignment files.

In the first step, the CTF parameters were calculated for each projection in Warp. In a second step, the CTF was estimated for the whole tilt-series taking the tilted geometry of the individual projections into account. The alignment was focused on the central row of density with a wedge-shaped mask that covered four repeats Figure 3—figure supplement 1E.

Classes which indicated a highly ordered and symmetric disk rim scaffold were selected Figure 3—figure supplement 2A , distance-cleaned, and separately aligned in RELION. More by Jingying Li. More by Juan Li. More by Huanghao Yang. Cite this: Anal. Article Views Altmetric -. Supporting Information. Cited By. This article has not yet been cited by other publications. Pair your accounts. Your Mendeley pairing has expired.

Here, a comprehensive overview is provided on an emerging platform: cell-membrane-coating nanotechnology. As a fundamental unit of biology, cells carry out a wide range of functions, including the remarkable ability to interface and interact with their surrounding environment. Instead of attempting to replicate such functions via synthetic techniques, researchers are now directly leveraging naturally derived cell membranes as a means of bestowing nanoparticles with enhanced biointerfacing capabilities.

This top-down technique is facile, highly generalizable, and has the potential to greatly augment existing nanocarriers.