Research Article

Multiplexed protein maps link subcellular organization to cellular states

See allHide authors and affiliations

Science  03 Aug 2018:
Vol. 361, Issue 6401, eaar7042
DOI: 10.1126/science.aar7042

You are currently viewing the abstract.

View Full Text

Log in to view the full text

Log in through your institution

Log in through your institution

Making multiplexed subcellular protein maps

Being able to visualize protein localizations within cells and tissues by means of immuno-fluorescence microscopy has been key to developments in cell biology and beyond. Gut et al. present a high-throughput method that achieves the detection of more than 40 different proteins in biological samples across multiple spatial scales. This allows the simultaneous quantification of their expression levels in thousands of single cells; captures their detailed subcellular distribution to various compartments, organelles, and cellular structures within each of these single cells; and places all this information within a multicellular context. Such a scale-crossing dataset empowers artificial intelligence–based computer vision algorithms to achieve a comprehensive profiling of intracellular protein maps to measure their responses to different multicellular, cellular, and pharmacological contexts, and to reveal new cellular states.

Science, this issue p. eaar7042

Structured Abstract


Obtaining multiplexed molecular readouts from large numbers of single cells in situ is an important technological goal to facilitate scientific discoveries in basic and translational research. Various methods have been developed in recent years to achieve spatially resolved multiplexed measurements of the abundance of large sets of mRNAs or proteins in biological samples. These technologies have brought the promise that, through large-scale efforts, all functionally relevant cell types of an organism will emerge from such multiplexed data in an unbiased manner. Furthermore, these cell types will be able to be mapped within their physical context within a tissue.


The involvement of an mRNA or protein (state) in cellular function depends on its specific intracellular location and interaction with other molecules and cellular structures. Moreover, the phenotype of an individual cell is determined by the functional state, abundance, morphology, and turnover of its organelles and cytoskeletal structures. To functionally interpret molecular multiplexed information, such measurements will thus need to resolve the intracellular length scale.


We report a simple, robust, and nondegrading protocol that achieves 40-plex protein staining in the same biological sample using off-the-shelf antibodies called iterative indirect immunofluorescence imaging (4i). In conjunction with high-throughput automated microscopy and computer vision, 4i allows highly reproducible multiplexed measurements from surface areas of several mm2 subsampled by pixels of 165 nm by 165 nm. This approach simultaneously captures functionally relevant properties that emerge at the cell population, cellular, and intracellular level. 4i can thus quantify the influence of local cell crowding on protein abundance, the effect of cell cycle position on protein phosphorylation, patterns of protein subcompartmentalization, and organelle morphology all in the same single cell and across thousands of cells.

We developed a data-driven computer vision approach that generates multiplexed protein maps (MPMs). MPMs comprehensively quantify intracellular protein composition with high spatial detail in large numbers of single cells. They are not confounded by the specific relative geometry and orientation of an individual cell. Thus, MPMs allow systematic comparisons of subcellular spatial protein distribution between single cells that experience different cell cycle states, microenvironments, growth conditions, or exposure to drugs. Using the example of subcellular relocalization of epidermal growth factor (EGF) receptor upon exposure to EGF, we demonstrate that MPMs allow the systematic definition of cellular states undetectable by multiplexed whole-cell measurements. The findings are functionally relevant and can be connected to multiple molecular and phenotypic properties apparent at the cellular and cell population scale.


By preventing photocrosslinking during imaging, 4i enables multiplexed immunofluorescence with off-the-shelf antibodies, both in small-scale and high-throughput experiments. 4i datasets cover multiple length scales, eliminating the need for extrapolation or inference in the interpretation of results. Integration of these length scales in one dataset reveals a richness of scale-crossing connections that current models of biological processes do not yet consider. These connections determine how gene expression is adapted to the cellular state, how a cell type is determined, how a pathological cellular phenotype emerges, or how a tumor cell responds to a drug.

Iterative indirect immunofluorescence imaging (4i).

4i obtains 40-plexed protein readouts at high spatial detail in thousands of cells with the use of off-the-shelf antibodies. Multiplexed protein maps derived from such images provide a comprehensive quantitative description of compartmentalized intracellular protein composition. These maps can identify new cellular states and allow quantitative comparisons of intracellular organization between single cells in different cell cycle states, microenvironments, and drug treatments.


Obtaining highly multiplexed protein measurements across multiple length scales has enormous potential for biomedicine. Here, we measured, by iterative indirect immunofluorescence imaging (4i), 40-plex protein readouts from biological samples at high-throughput from the millimeter to the nanometer scale. This approach simultaneously captures properties apparent at the population, cellular, and subcellular levels, including microenvironment, cell shape, and cell cycle state. It also captures the detailed morphology of organelles, cytoskeletal structures, nuclear subcompartments, and the fate of signaling receptors in thousands of single cells in situ. We used computer vision and systems biology approaches to achieve unsupervised comprehensive quantification of protein subcompartmentalization within various multicellular, cellular, and pharmacological contexts. Thus, highly multiplexed subcellular protein maps can be used to identify functionally relevant single-cell states.

View Full Text