Resin embedded multicycle imaging (REMI): a tool to evaluate protein domains

Abstract

Protein complexes associated with cellular processes comprise a significant fraction of all biology, but our understanding of their heterogeneous organization remains inadequate, particularly for physiological densities of multiple protein species. Towards resolving this limitation, we here present a new technique based on resin-embedded multicycle imaging (REMI) of proteins in-situ. By stabilizing protein structure and antigenicity in acrylic resins, affinity labels were repeatedly applied, imaged, removed, and replaced. In principle, an arbitrarily large number of proteins of interest may be imaged on the same specimen with subsequent digital overlay. A series of novel preparative methods were developed to address the problem of imaging multiple protein species in areas of the plasma membrane or volumes of cytoplasm of individual cells. For multiplexed examination of antibody staining we used straightforward computational techniques to align sequential images, and super-resolution microscopy was used to further define membrane protein colocalization. We give one example of a fibroblast membrane with eight multiplexed proteins. A simple statistical analysis of this limited membrane proteomic dataset is sufficient to demonstrate the analytical power contributed by additional imaged proteins when studying membrane protein domains.

Figure 1. Illustrative depiction of method.

A typical array tomography section through cells suspended in LR White (A) includes a sizable amount of cell volume, but only fragments of membrane (B). We can embed membranes directly to maximize the imageable membrane (C). Our method for accomplishing this is cartooned in (D). To polymerize in the absence of oxygen, we use a specialized preparation chamber (E). A glass-bottomed dish with membrane to be embedded is covered with a sheet of adhesive film. Two holes are punched through the film, used as access points to pump in argon gas and to apply solutions.

Figure 2. LR-White embedding provides a durable platform for multicycle labeling.

(A) Immunolabeling of an HA-expressing mouse fibroblast membrane demonstrates epitope preservation and proximity to the surface of the resin. Antibody elution with a solution of 0.2 M NaOH/0.02% SDS/ddH20 (pH 13) for 20 minutes removes fluorescence between rounds. The epitopes survive repeated immunolabeling, their fluorescence increasing slightly after the first elution through an as-yet uncharacterized process of antigen retrieval. This has been observed across different labeled proteins and secondary antibody choices. (B) When the average brightnesses are equalized, round 1 and round 3 display a highly correlated labeling pattern (Pearson’s r = 0.89), as illustrated by minmax analysis. (C) A color histogram of the previous image, displaying the ratio of green to red, shows that while the two rounds are largely similar, small differences do exist in the relative labeling of puncta. Scale bar: 5 μm.

Figure 3. 8-channel immunolabeling of membrane and cytoskeletal proteins.

(A) Heterogeneous immunolabels applied to an isolated HA-expressing mouse fibroblast membrane provide a diverse selection of spatial properties for analysis. (B) Three membrane proteins can be compared directly, with an ROI (square inset, (C) extracted for numerical analysis. (D) When the threshold for analysis is varied across the dynamic range, an exceedance relationship arises, defined as the ratio of pixels occupied by multiple channels to that expected if the channels were independent. Although correlations arise in all comparisons, the triple-occupancy case displays more sensitivity, from lower thresholds, quickly outpacing the doublets. All scale bars are 5 μm.

Figure 4. Whole-mount resin embedding of cellular monolayers.

REMI also preserves the structure of whole cells, depending on cell type (A). “Typical” cells such as mouse fibroblasts (B) largely retain their structural integrity, and the LR White resin forms a thin layer above their apical membrane. Human adipocytes (C) lose their structural integrity during the embedding process, leaving behind a basal membrane ready for labeling. Scale bar: 10 μm.

Figure 5. Durability of REMI media.

In an extended test of REMI sample durability, mouse fibroblast cells were plated onto a dish with a thin coat of carbon to enhance adhesion, embedded in LRW, and imaged at low resolution with DAPI-containing media (SlowFade Gold). Between each image, the sample underwent three washes of high-pH solution, at 20 minutes per wash, simulating three elution cycles. This was repeated to destruction for each sample; when catastrophic “lifting” of the entire cell layer would occur. Average lifetime of 14 + /−1.6 rounds, n = 3.

Figure 6. Multiround super-resolution (STED) imaging of embedded cells.

The physical properties involved in REMI do not interfere with the functionality of stimulated emission depletion (STED) microscopy of embedded mouse fibroblast cells (A). Left: confocal image, right: STED image. Multiple rounds of super-resolved fine structure labeling is also a possibility in STED (B), and result in images with similar patterns of epitope labeling, even at such high resolution (inset). Minmaxed analysis of the rounds (C) demonstrates a larger degree of epitope fluctuation between rounds, presumably due to stochastic labeling. This becomes evident in the related histogram (D). If the region of analysis is restricted to well-defined puncta, as in the bottom half of the image in C (below dashes), the stochastic labeling is unmistakable (D, inset). Scale bars: 5 μm.

Figure 7. Penetration of immunolabeling in REMI media.

A STED volume image of a mouse fibroblast cell, partially rounded while undergoing mitosis, demonstrates antibody penetration when alpha-Tubulin is labeled for precisely two hours. (A) averaged projections in the lateral and axial dimensions have visually homogeneous labeling. (B) labeling can be mapped onto the distance from the exterior of the cell (lateral and axial profiles, top), optionally removing the bright mitotic spindle labeling (bottom). (C) A plot of average intensity as a function of depth within the cell demonstrates uniform labeling inward to the edge of the distance map when the nucleus is excluded. Volume size: 20μm × 20μm × 5.5μm. Scale bar: 5 μm.