WO2018215624A1

WO2018215624A1 - Method for image-based flow cytometry and cell sorting using subcellular co-localization of proteins inside cells as a sorting parameter

Info

Publication number: WO2018215624A1
Application number: PCT/EP2018/063729
Authority: WO
Inventors: Michael Knop; Patrick THEER; Iordania CONSTANTINOU; André SCHULZE; Michael JENDRUSCH; Frederik GOERLITZ
Original assignee: Universitaet Heidelberg; Deutsches Krebsforschungszentrum
Priority date: 2017-05-24
Filing date: 2018-05-24
Publication date: 2018-11-29

Abstract

The present invention relates to methods for the physical sorting of cells, such as flow cytometry methods. In particular, the invention relates to methods in which the co-localization of at least two detectable markers is utilized to physically sort (and ultimately separate) cells. The present invention further relates to devices capable of performing said methods, to uses of the co-localization of the at least two detectable markers for the physical sorting of cells, to kits comprising at least two detectable markers and instructions for performing the methods, to uses of said kits, as well as to cells or cell pools that have been sorted, i.e. have been isolated or obtained, with such methods.

Description

METHOD FOR IMAGE-BASED FLOW CYTOMETRY AND CELL SORTING USING SUBCELLULAR CO-LOCALIZATION OF PROTEINS INSIDE CELLS

AS A SORTING PARAMETER

FIELD OF THE INVENTION

The present invention relates to methods for the physical sorting of cells, such as flow cytometry methods. In particular, the invention relates to methods in which the co-localization of at least two detectable markers is utilized to physically sort (and ultimately separate) cells.

The present invention further relates to devices capable of performing said methods, to uses of the co-localization of the at least two detectable markers for the physical sorting of cells, to kits comprising at least two detectable markers and instructions for performing the methods, to uses of said kits, as well as to cells or cell pools that have been sorted, i.e. have been isolated or obtained, with such methods. BACKGROUND OF THE INVENTION

In biotechnology and biomedicine, flow cytometry denotes a laser- or other light- source-dependent method used for the characterization of cells and populations of cells via counting of said cells and the detection and quantification of biomarkers. It is used for cell characterization and cell sorting. In flow cytometry, a stream of liquid-suspended cells is passing an electronic detection system that detects fluorescence and other light-dependent signals, such as light scattering, suitable for multiparametric analyses of complex cell samples such as blood or suspended tissue samples. Routinely, commercial instruments are able to detect the abundance of biomarkers by measuring the intensity of fluorescence signals emitted from labels or dyes that are used to detect the signals. Typically, signal width, height and total intensity are measured and further characterization of cell size and granularity is possible using light scattering and pulse width.

Microscopes are central to cellular studies as only they provide insights into the dynamic organization of cells. The development of new microscopes and microscopic techniques is a main driver of progress in cellular research. High throughput microscopy opened up the use of subcellular information for systemic studies or screening purposes, including drug development. Improvements seek to increase the throughput, to improve data analysis and to implement different microscopy techniques to maximize the information acquired from the samples ('high content microscopy'). A caveat of high throughput microscopy is that it requires ordered sample arrays in order to correlate the retrieved information with a particular sample. This contrasts with fluorescence-activated cell sorting (FACS), a flow cytometric technique that allows the isolation of cells, where cells are moving in front of the detector and total amount of emitted fluorescence is used as a signal to identify individual cells and cell populations and specimen that can be physically sorted based on a predefined total signal into distinct pools for further identification or analysis.

The combination of fluorescence microscopy that provides images of cells with sub-structural resolution and the cell sorting capacity of flow cytometers would require the acquisition and rapid analysis of images of cells. Sorting occurs based on patterns or structures detected by the image analysis algorithms. However, reliable pattern or structure recognition and the identification and quantification of subcellular features that are biologically meaningful is difficult and hard to implement into high speed cell sorters. The reasons are that (i) there is only a limited amount of time available for the computational analysis of the cells, and (ii) biological variability of subcellular structures, such as the different organelles, their size, and abundance is very high and exhibits strong cell-to-cell differences. Therefore, for many applications and analyses it is very difficult to derive descriptors that are informative and reliable enough to serve as a specific predefined information to distinguish wanted from unwanted cells and to isolate only the wanted cell(s). Accordingly, alternative methods need to be used for cell sorting when e.g. biologically meaningful subcellular features should be used as a sorting criterion.

A combination of fluorescence microscopy with FACS would enable the physical sorting of cells into populations with distinct biological features derived from image analysis of microscopic images of each individual cell. Such biological features can be very rich in information and hence possess enormous potential for biotechnology and biomedical research. However, to achieve efficient image- based sorting of cells in high speed, it is required that the images are analyzed while the cells are being transported between the imaging site and the sorting site of the instrument for physical cell sorting. For the comprehensive analysis of complex samples that can consist of thousands to many millions of cells, it is furthermore required to analyze cells with high speeds, i.e. many cells per seconds. To overcome the speed and precision limitations imposed by complex image analysis routines and cell-to-cell differences in cell populations, it is required to use image analysis procedures that are fast and informative and that provide robust and functionally relevant quantifications of subcellular localization patterns.

Some recent developments seek to implement FACS-like cell sorting capabilities into fluorescence microscopy. Methods based on a continuous flow of cells ('continuous flow methods') image the cells while they are passing by the objective lens followed by rapid image analysis and decision making to trigger sorting and gating of the cells into two or more vessels where they are collected. Faithful sorting of desired cells means that the decision whether or not a cell should be sorted has to be made as accurately as possible in a very short time span while the cells are migrating through the system from the imaging location to the physical sorting location, which makes image analysis also time-critical. In conventional FACS, the available window for decision-making is usually in the range of <1 ms.

Accordingly, the technical problem underlying the present invention is to provide methods for the physical sorting of cells of interest using biologically relevant subcellular features derived from image analysis of a microscopic image of each individual cell as sorting criterion. Preferably, such methods should allow for the rapid sorting of reasonable cell numbers in the thousands or more. Further, the present invention should provide devices capable of performing said methods.

SUMMARY OF THE INVENTION

According to a first aspect, the invention provides a method for the physical sorting of cells of interest, comprising the steps of:

i. providing cells that are labeled with at least two detectable markers in a fluid;

ii. applying a flow to the fluid;

iii. acquiring at least one electronic image of an individual cell;

iv. detecting the signals of the at least two detectable markers within the at least one electronic image;

v. determining the degree of overlap of the spatial patterns of the signals of said at least two markers as a measure for the degree of spatial co- localization of said at least two detectable markers; and

vi. in case the degree of overlap is within a predefined range, physically separating said cell from cells with a degree of spatial co-localization not in the predefined range.

In a second aspect, the present invention relates to a device for acquiring electronic images of individual cells and physically sorting cells of interest, comprising:

a) a microfluidic system;

b) light sources for the differential excitation of at least two detectable markers;

c) a multichannel detection system that is capable of generating spatially resolved two- or three-dimensional localization maps representing the subcellular distribution of the at least two detectable markers in the cell; d) hardware that is configured to execute analysis software computing the degree of co-localization of the at least two detectable markers in the localization maps generated in step (c); and

e) a cell-sorting unit.

The present invention further provides, in a third aspect, the use of the degree of spatial overlap of the signals of at least two detectable markers in an electronic image of a cell of interest that is labeled with said at least two detectable markers as a criterion for the physical sorting of said cell of interest. Thus, according a fourth aspect the present invention provides a cell or cell pool that has been isolated with the method according to the first aspect of the present invention, or a cell clone derived from said cell or cell pool.

Moreover, in a fifth aspect, the present invention also provides a kit comprising at least two detectable markers and instructions for performing the method according to the first aspect of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

As stated above, in a first aspect, the present invention relates to a method for the physical sorting of cells of interest, comprising the steps of:

ii. applying a flow to the fluid;

iii. acquiring at least one electronic image of an individual cell;

vi. in case the degree of spatial co-localization is within a predefined range, physically separating said cell from cells with a degree of spatial co-localization not in the predefined range.

The method according to the first aspect allows a cell sorting based on the relative positions of the specific molecules within the cell. Subcellular localization of molecules, such as proteins, nucleic acids or lipids is determined by the biological and biophysical properties of a molecule. For small molecules it is mostly the place of their synthesis or whether they are hydrophobic enough to enter and transverse biological membranes or their active transport by specific transporters that determines their localization inside cells. Individual localizations can be inside a compartment or organelle, inside or attached to a specific membrane, or localization in the cytoplasm.

For proteins, the localization is usually determined by amino acid sequence features, so called targeting sequences as well as sequences that function for covalent addition of molecules, e.g. lipid residues, and membrane segments. Based on these sequence properties proteins acquire a specific localization inside a cell, to one or multiple structures or dispersed throughout the cytoplasm. The information that is inherent to the distribution and localization of proteins and molecules inside cells is not dependent of the total abundance of a protein.

In classical flow cytometry, where total signals are analyzed, the term "sorting" is used in different ways. First, it is used for physical sorting of cells and objects into one or more distinct cell pools. Second, it is also frequently used for the computational sorting of data-points into different categories. However, as used herein, the terms "sorting" and "physical sorting" refer to the first definition, i.e. the actual physical sorting of cells into one or more distinct individual cell pools which can be separately used for further identification, analysis or experimentation. Given the small sizes of cells in the micrometer range, this poses many challenges, one of which is addressed in this invention.

Cells of interest to be sorted using the methods of the present invention are not limited in any manner and can be any type of cells or cell lines that are of interest for a particular question or application. Suitable cells and cell lines include any kind of prokaryotic, e.g. archaeal or bacterial, cells, as well as any kind of eukaryotic cells, including e.g. yeast cells, plant cells, insect cells, and mammalian cells. In particular embodiments, the cells can be part of larger cell assemblies such as organoids or embryos.

The cells used in the method according to the first aspect are labeled with at least two, i.e. 2, 3, 4, 5, or more, detectable markers, wherein two detectable markers are preferred. The terms "marker" and "detectable marker" as used herein relate to any moieties that can be detected in the framework of the methods of the present invention. The markers are able to emit a signal, in particular a light signal, such as a fluorescent light signal. The light signal is preferably emitted upon excitation of the marker. These markers are preferably distinguishable from each other, e.g. by the wavelength of the light each fluorophore is emitting. Alternative modes of fluorophore discrimination are also thinkable, e.g. based on the lifetime of the emitted electrons. According to one embodiment, at least one marker is a fluorophore. According to a further embodiment, all markers are fluorophores and the fluorophores emit light at different wavelengths. Typically, markers are either chemical compounds that are fluorescent in particular localizations in the cell, such as fluorescent molecules that are fluorescent as a function of a physico-chemical parameter (e.g. pH, Calcium- levels, etc.), or they are chemical compounds or proteins such as antibodies that are coupled to fluorescent moieties. In this case, the chemical compound (such as a lipid molecule) or the protein (such as the antibody) are targeting the markers to a specific substructures or the localization of a target inside the cell.

Due to the marker, the substructure or target, i.e. labeled moiety, can be detected based on the coupled fluorescent moiety. Fluorescent moieties can be fluorescent dyes, quantum dots, fluorescent proteins, e.g. green fluorescent proteins (e.g. GFP) or red fluorescent proteins (e.g. mCherry) or any other type of molecule or nanostructure that is able to emit light. Suitable fluorescent moieties, fluorescent dyes, quantum dots and fluorescent proteins are not particularly limited and are known in the art.

Cells of interest can be labeled with detectable markers, e.g. fluorescent markers, by methods known in the art. In this context, the labeling of cells according to the present invention includes the labeling of cells with affinity ligands, antibodies, antibody fragments or antibody mimetics which in turn are conjugated with the marker, e.g. a fluorescent dye, the labeling of cells with dyes that enrich in particular compartments of the cells, such as e.g. lysotracker for lysosomes and mitotracker for mitochondria, as well as the expression of fluorescent proteins, fluorescent protein conjugates, dye-linking tags, such as e.g. SNAP-Tag or Halo- Tag, or the like in the cells. Detectable markers can also be expressed from transfected RNA or DNA or genomically-incorporated DNA inside the cell. To obtain cells labeled with multiple detectable markers, any of these techniques for cell labeling can also be combined with each other, i.e. different types of detectable markers can be used together to obtain cells with two or more labelings.

As the labeled moiety occurs in a multitude of copies within the cell, a marker produces a multitude of signals. The multitude of signals relating to the same marker, and i.e. to the same cellular component that is labeled, forms a signal pattern. The signal pattern may be described as an area containing the individual signals corresponding to the marker. As the electronic image may be a three- dimensional (3D) image based on image acquisition in multiple focal planes, e.g. multifocal plane microscopy, the signal pattern may also be a volume. According to one embodiment, the spatial patterns define an area or volume and the overlap of the area or volume of the spatial patterns is calculated based on overlap- determining parameters. The overlap-determining parameters include parameters that relate to (i) the total area and the shape of the spatial pattern, (ii) the relative area the spatial pattern occupies relative to the cell or (iii) parameters that describe the intensity distributions within the occupied area. There are many different overlap- determining parameters that can be used to define spatial distribution patterns within cells. The overlap-determining parameters can be combined to complex algorithms for more specific definition of the overlap of two signal patterns. Such algorithms can be handled by deep learning and neuronal network based computing methods, to better specify the similarity of the localization of two markers.

The term „co-localization" as used herein in particular refers to spatial co- localization of marker signals within a cell. Accordingly, the terms "co-localization" and "spatial co-localization" are used interchangeably.

To describe co-localizations, basic concepts and notations from set theory can be used to describe the relationship of the relative amount of spatial occupancy of one marker relative to the other marker. In addition, numerical measures can be used to quantify absolute and relative areas and associated signal intensities.

Co-localization can be quantified as a fraction or percentage of the area or volume of the spatial pattern of the signal of a marker relative to this parameter of the other marker. Thus according to one embodiment, the degree of overlap is determined by calculating the overlapping area or volume of the spatial patterns of the two signals, calculating the area or volume of the full spatial pattern of the signal of the first marker and calculating the quotient of the overlapping area or volume to the full area or volume. Accordingly, the co-localization is a measure that describes to what extend two different entities occupy the same space inside a cell.

The marker signals can either fully co-localize, i.e. the signal patterns obtained for both markers overlap completely, or they can co-localize partially, i.e. when the signal patterns do not fully overlap. The latter scenario is also referred to as partial spatial co-localization.

The sorting criterion is whether the degree of spatial co-localization is within a predefined range. The range is predefined based on the specific biological question. For example, the question may be whether a first moiety labeled with a first marker shows a high degree of co-localization to a second moiety or subcellular structure with a second marker. In this case, the upper limit of the predefined range is 100 % and the lower limit is 40 %, preferably 60 %, more preferably 70 %, most preferably 80 %. Alternatively, the question may be whether a first moiety labeled with a first marker shows a low degree of co-localization to a second moiety or subcellular structure with a second marker. In this case, lower limit of the predefined range is 0 and the upper limit is 60 %, preferably 40 %, more preferably 30 %, most preferably 20 %.

Of course, the question can also relate to a specific degree of co-localization, such as whether a first entity, i.e. the first marker has a specific degree of co- localization. In this regard, the predefined range would be X ± Y, wherein X may be 10 %, 20 %, 30 %, 40 %, 50 %, 60 %, 70 %, 80 % and wherein the Y is e.g. 10 % or 5 %.

Partial spatial co-localization can be defined, for example, by the degree of overlap of the detected marker signals of about 90%, about 80%, about 70%, about 60%, about 50%, about 40%, about 30%, about 20%, or about 10%. The degree of overlap of the detected marker signals can also be in a range of 90% to 10%, 80% to 20%, 70% to 30%, or 60% to 40%.

Possible scenarios of partial spatial co-localization are illustrated in Figure 6. Examples for a low degree of partial spatial co-localization show an overlap of the marker signals of about 40%, preferably 30%, more preferably 20%, most preferably 10%. Examples of a high degree of partial spatial co-localization show an overlap of the marker signals of about 60%, preferably 70%, more preferably 80%, most preferably 90%.

The degree of co-localization may not only be defined in terms of a specific percentage. For some applications, it will only be relevant if there is a "high degree of co-localization", a "low degree of co-localization" or "no co-localization". As shown in the Figures, dependent on the cellular component or subcellular structure that is labeled, a variety of shapes are found for the signal patterns. The determination of the degree of co-localization may also include a factor for the shape of the signal pattern.

Importantly, determination of the co-localization based on the overlap of the signal patterns is independent of the total amount of each marker present in the cell and thus independent on signal intensity. Thus, according to one embodiment the determination of the degree of spatial co-localization is independent of the intensities of the signals of the markers. Since conventional flow cytometry permits measuring the total intensity of a signal per cell only, conventional fluorescence associated cell sorting (FACS), which is based on flow cytometry, is not capable of sorting cells based on subcellular localization and fluorescence distributions and co-localization based criteria. Preferably, at least one of the at least two detectable markers labels a structure such as a protein or a modification or specific state of a protein of interest (e.g. a phospho-epitope or a particular conformation or folding state of a protein). According to one embodiment, the first of the at least two detectable markers labels a protein of interest. In this context, the protein of interest is not limited in any manner and includes any protein that is expressed or is presumably expressed by the cells of interest and modification of a protein.

Further, it is preferred that at least one of the at least two detectable markers labels a cellular substructure of interest or a second protein of interest. According to one embodiment, the second of the at least two detectable markers labels a cellular substructure of interest. In this context, the cellular substructure of interest is not limited in any manner and is preferably a cell organelle, including e.g. the cell nucleus, mitochondria, the Golgi apparatus, any other cell organelle, or a particular area of the cell surface.

In case one detectable marker labels a protein of interest or a first structure and the other detectable marker labels a subcellular structure of interest or a protein at a subcellular localization, the methods of the present invention are suitable to sort cells based on a quantification of the fraction of both detectable markers that exhibit the same localization inside the cell, as determined by analysis of cells in the steps described in (iii) to (vi).

According to the present invention, steps (iii) to (vi) are performed for each individual cell to be analyzed.

In step (iii) of the method of the present invention, an electronic image of the individual cell is acquired. The electronic image of the individual cell is a microscopic electronic image of the cell. Methods for the acquisition of an electronic image of a cell, i.e. a microscopic electronic image of the cell, are not particularly limited and are known in the art. They include any methods for the generation of spatial maps of signals of the markers in cells.

According to one embodiment, the acquisition of the electronic image is performed using a fluorescence microscope, which is preferably equipped with one or more high-speed camera(s).

Said microscope is preferably equipped with an objective lens, e.g. a 60* or 100* lens, having a high numerical aperture (NA), as well as high resolution and light transmission, to achieve excellent detection of fluorescence. Respective microscopes and cameras are known in the art. According to one embodiment of the first aspect, the image acquisition of step (c) is performed using an objective with a numerical aperture of at least 1.0, preferably at least 1.2, more preferably at least 1.3, most preferably at least 1.4.

According to one embodiment, more than one electronic image is acquired per cell by using different wavelengths of the excitation beam and separation of the emission images for simultaneous detection. These acquired images are preferably multi-channel images, wherein the number of channels preferably corresponds to the number of fluorescent markers used. In the case of using two fluorescent markers, said images are preferably dual-channel images. The electronic images may be recorded with one or more cameras. For example, the multi-channel images can be acquired by multiple cameras. For two markers, two cameras may be used to acquire the image. The term "camera" as used herein refers to any device that is capable of acquiring spatially resolved images of a cell.

Due to the application of a flow to the fluid in which the cells are delivered in a fluid stream, in particular in such a way that cells can be imaged and subsequently analyzed individually, i.e. one after the other. Methods for generating such flows of cells are not particularly limited and are known in the art, often referred to as 'microfluidics'. According to one embodiment of the method, the flow rate of the fluid applied in step (ii) is at least 1 nL/sec, preferably at least 100 nL/sec, more preferably at least 1 μΙ_/εβο, most preferably at least 100 μί/εβα This flow rate allows the sorting of a high number of cells within short time.

In step (iv) the signals of the at least two detectable markers within the at least one electronic image are detected. The detection is followed by the determining of the degree of overlap of the spatial pattern of the signals of said at least two markers as a measure for the degree of spatial co-localization of said at least two detectable markers.

In this context, the present invention includes methods in which more than two, e.g. 3, 4, 5, or more detectable markers are used, and wherein more than two or all of the used markers co-localize with each other. No matter how many markers are used, co-localization always refers to a detection of two (or more) markers at the same subcellular localization. For multicellular assemblies (e.g. organoids), co-localization can either refer to subcellular localizations within individual cells of the organoids, or to a substructure of the multicellular assembly, e.g. specific cells within the assembly. The degree of co-localization can be measured from the dual- or multi-channel images. Preferably, co-localization of the at least two detectable markers is quantified based on the overlap of the localization of one marker with the localization of the other marker(s). An overlap of the localizations of two markers can be detected when both (or multiple markers) exhibit a predefined degree of localization to the same spatial area in the cell. In this context, co-localization is preferably to areas that constitute only a part or a fraction of the image of the entire cell.

In step (vi) of the method of the present invention, a cell with a degree of co- localization within the predefined range is separated from cells with a degree of co-localization outside of the predefined range, i.e. is directed into a specific compartment or vessel. This separation is also referred to as physical sorting the cells.

The sorting or separation of the cells/cell pools into different vessels is achieved by gating the stream of cells at the right time point so that the cell for which the degree of co-localization has been determined can be individually directed into one of the different vessels based on the information if the degree of co- localization is in the predefined range or not. Methods for the physical sorting of cells are not particularly limited and are known in the art.

By assessing co-localization as a parameter for the localization of a marker of interest it is advantageously not necessary to have any a priori knowledge about the shape and location of the structure of the marker of interest and to use such information to detect specific cells, where this localization is either present or absent.

In addition to the spatial signal pattern derived degree of co-localization, image analysis can measure other parameters of the cells as well (e.g. intensity of the signals in different channel, size of the cells, shape of the cells, etc.). Therefore, in specific embodiments, the co-localization sorting criterion can be combined with one or more of these other parameters, so that the sorting criterion is only fulfilled if also another parameter is in a predefined value, for example to sort only cells above a certain size.

In preferred embodiments of the methods of the present invention, signal detection according to step (iv) and determination of the overlap according to step (v) are performed directly after image acquisition in step (iii). The signal detection and determination of the overlap of the spatial patterns of the signals of the at least two markers is also referred to as image analysis.

In further embodiments, the cell sorting, i.e. separation of the cells, in step (vi) is performed directly after image analysis according to steps (iv) and (v). In yet further preferred embodiments, steps (iii), (iv), (v) and (vi) are performed directly after each other for each individual cell to be analyzed.

In particular and preferred embodiments of the methods of the present invention, the image analysis of step (iv), preferably steps (iii) and (iv), more preferably steps (iii), (iv), (v) and (vi) are performed in an automated manner, i.e. image acquisition, image analysis, and physical sorting of each cell is performed in an automated manner. This includes an automated, i.e. computerized, evaluation of whether the sorting criterion, i.e. the degree of co-localization of two or more detectable markers, is met for each individual cell. Respective automated methods according to the present invention are able to perform steps (iii) to (vi) for each individual cell to be analyzed within the time that it needs for the individual cells to flow from the site of image acquisition to the site of physical cell sorting into different vessels, e.g. within at most 5 ms, preferably at most 4 ms, at most 3 ms, at most 2 ms, or at most 1 ms.

Preferred implementations of steps (iii) to (vi) enable the analysis and sorting of many cells per second, the more the better from 10 cells per second to more than 100,000 cells per second, e.g. of at least 50 cells/second, preferably at least 100 cells/second, at least 200 cells/second, at least 300 cells/second, at least 500 cells/second at least 750 cell/second, or at least 1000 cells/second. According to one embodiment of the method of the first aspect, the steps (iii) to (vi) are carried out at a speed of 1 cell/second, preferably, 50 cells/second more preferably 100 cells/second.

According to one embodiment, one electronic image per cell is acquired. Other methods use the splitting of the excitation beam and an identical acquisition light path to acquire more than one electronic image of the along its path in the fluid stream. The acquisition of only one image has the advantage that it reduces the time for image analysis. The small size of the images allows a fast image analysis, which is important to minimize the time between the image acquisition and the sorting event.

According to one embodiment of the first aspect, the excitation beam and detection light path for image acquisition move with the cell at the same speed of the cell to allow a constant image acquisition of the moving cell. This makes it possible to obtain a single electronic image of the cell with a high resolution of and a high signal intensity of the marker.

For a reliable determination of the spatial patterns of the marker signals, which are the basis for the co-localization, the resolution of the structures of the cell is important. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 3 μηη. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 2 μηη. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 1.0 μηη. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 0.5 μηη. Alternatively, the resolution of the structures of the cell may be defined as a percentage of the theoretical diffraction limit of the used optical set up of the microscope. Thus, according to one embodiment, the structures of the cell have a resolution in the electronic image of less than 20 % above the diffraction limit. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 40 % above the diffraction limit. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 60 % above the diffraction limit. According to one embodiment, the structures of the cell have a resolution in the electronic image of less than 80 % above the diffraction limit.

According to one embodiment, the distance between objective and cells during image acquisition is constant with a variability in the range of below 8 μηη, preferably below 4 μηη, more preferably below 2 μηη. The low variability has the advantage that the images of different cell are better comparable leading to a high accuracy of the sorting.

In this context, the methods of the present invention are preferably performed within the framework of a flow cytometric procedure.

(a) a microfluidic system;

(b) light sources for the differential excitation of at least two detectable markers;

(c) a multichannel detection system that is capable of generating spatially resolved two- or three-dimensional localization maps representing the subcellular distribution of the at least two detectable markers in the cell; (d) hardware that is configured to execute analysis software computing the degree of co-localization of the at least two detectable markers in the localization maps generated in step (c); and

(e) a cell-sorting unit.

The function of the microfluidic system (a) of the devices of the present invention is two-fold. First, the main function of the microfluidics system is to produce a steady stream of cells that are moved in a fluid, first to the area of cell imaging and then from there to the area where the cells are sorted. Thus, in a preferred embodiment, said microfluidic system is capable of producing and transporting a steady stream of cells within the device. In particular, embodiments, the microfluidics system (a) can also be used to concentrate the cells before they are streamed through the device and imaged and then sorted.

According to one embodiment, the microfluidic system contains an inlet, a cell- concentration unit, a cell-focusing unit, and an imaging unit. Second, a function of the microfluidics systenn is to physically sort individual cells into channel(s) so that single cells or pools of cells with the same properties, i.e. the same degree of co-localization of markers, can be analyzed in isolation from the rest of the cells. In this context, several methods that allow to physically sort cells in a microfluidics device based on various different principles are known in the art, from electrostatics, to optical traps, laser pulses, and piezo-electric channel opening/closing. Respective microfluidic systems are not particularly limited and are known in the art.

The light sources (b) of the devices of the present invention are preferably laser light sources. In this context, due to the high speed of the flowing cells, exposure times are preferably very short, in order to prevent motion blur in the acquired images. This generally requires laser light sources whose output can be modulated on a short time-scale. Respective laser light sources are known in the art and include e.g. diode lasers, and acousto-optically modulated (AOM) solid- state lasers.

According to one embodiment, the beams emitted from said light sources are moveable within the imaging unit in direction of the flow of the fluid. This allows an excitation of an individual cell while it moves with the fluid stream. According to one embodiment, the detection light path of the detection system is moveable within the imaging unit in direction of the flow of the fluid. The excitation beam and the detection light path are preferably steered by the same mirrors. Thus, detection light path is moveable in combination with the beams emitted from the light sources. This combination allows the excitation of the markers in an individual cell and an image acquisition while the cell is moving. This in turn makes it possible to obtain a single electronic image of the cell with a high resolution of and a high signal intensity of the marker.

The multichannel detection system (c) of the device of the present invention is preferably a fluorescence microscope equipped with one or more high-speed camera(s). The fluorescence microscope is preferably equipped with an objective lens or oil immersion objective lens, e.g. a 60* or 100* lens, having a high numerical aperture (NA), e.g. a NA of 1.3 or better. Respective microscopes and cameras are known in the art. The term "camera" as used herein refers to any device that is capable of acquiring spatially resolved images of a cell.

The multichannel detection system (c) is preferably suitable for the generation of the localization maps on the required time scale (preventing motion blur) and repetition rate (cell image- and sorting-rate). This in general requires means for the collection and spectral discrimination of the (fluorescence) light generated using the light sources (b). For example, in a camera based system, the fluorescence light ennitted from the cell is typically collected by a high-NA microscope objective, spectrally separated by multichroic mirrors and filters and imaged onto either multiple cameras (one camera per spectral channel) or onto separate areas (one area per spectral band) of a single camera. While high quantum efficiency and small read noise are a well-sought feature in most cameras, here, due to the required short exposure times, it is of paramount importance. In addition, cameras for image based sorting are required to feature high frame rates and short global-shutter as well as response times (preferentially both in the low με range or better). Respective detection systems, microscopes, lenses, and cameras are known in the art.

The analysis hardware (d) of the devices of the present invention preferably supports multicore operation allowing for the parallelized computation of co- localization parameters. Respective hardware is not particularly limited and is known in the art. However, preferably, the hardware supports multicore operation, which allows for the parallelized computation of many different co-localization parameters at essentially no additional analysis time and thus a fine tunable multivariate approach to cell discrimination. While multicore operation is supported by most standard CPUs, their number of cores is relatively small. Often a better choice are GPUs (graphic cards) which provide many more cores and in addition an architecture streamlined for the processing if image data which allows a much higher degree of parallelization to rapidly process the image data and analyze co-localization. Alternatively, specialized hardware implementations can be used to further accelerate the analysis.

The sorting unit (e) of the devices of the present invention functions to physically sort individual cells into channel(s) so that single cells or pools of cells with the same properties, i.e. the same degree of co-localization of markers, can be analyzed in isolation from the rest of the cells. In this context, several methods that allow to physically sort cells in a microfluidics device based on various different principles are known in the art, from electrostatics, to optical traps, laser pulses, and piezo-electric channel opening/closing. Respective sorting units are not particularly limited and are known in the art.

In this aspect, all relevant definitions and limitations as defined above for the methods of the present invention apply in an equal manner. In particular, the cells of interest, detectable markers, means for the analysis of co-localization, and means for physically sorting the cells are as defined above.

In a third aspect, the present invention relates to the use of the degree of spatial overlap of the signals of at least two detectable markers in an electronic image of a cell of interest that is labeled with said at least two detectable markers as a criterion for the physical sorting of said cell of interest.

In a fourth aspect, the present invention relates to a cell or cell pool that has been isolated with the method according to the first aspect of the present invention, or a cell clone derived from said cell or cell pool. In this aspect, the cells are as defined above. Methods for generating cell clones from single cells or cell pools are not particularly limited and are known in the art.

In a fifth aspect, the present invention relates to a kit comprising at least two detectable markers and instructions for performing the method according to the first aspect of the present invention. In this aspect, the detectable markers are as defined above.

As such, the present invention also relates to the use of the kits of the fifth aspect for performing the methods of the present invention.

The present invention advantageously identified the co-localization of at least two detectable markers in a cell of interest as a feature that is easily detectable and in a manner that is robust, informative and fast enough to allow the use thereof as a sorting criterion in flow cytometric cell sorting based on dual-/multi-channel images (image-activated cell sorting; IACS). Further, due to the almost unlimited number of possible marker combinations, said feature advantageously can be very rich in information and hence possesses enormous potential for biotechnology and biomedical research.

As used herein, the term "about" is intended to be a modifier of ± 10% of the specified value. As an example, the term "about 5%" is intended to encompass the range of 4.5 to 5.5%.

The terms "comprising/comprises", "consisting of/consists of, and "consisting essentially of/consists essentially of are used herein in an interchangeable manner, i.e. each of said terms can expressly be exchanged against one of the other two terms.

The present invention addresses the challenges associated with image processing during a flow cytometry experiment and elucidates on methods to rapidly process relevant information. In particular, co-localization of fluorescence to subcellular structures is used as a way to rapidly assess relevant information about a protein or marker inside a cell and to use this information for physical cell sorting. It is concluded that using co-localization according to the present invention as a measure for protein localization makes image-based flow cytometry-associated physical cell sorting a reliable method to perform many different types of assays or experiments so that it can be used for biomedical analytics, for drug development pipelines or to screen for cellular effects of drugs, as well as applications in the analysis of samples from patients and many other applications.

Further, it is demonstrated that co-localization of a protein with a marker, a dye or another protein serves as a robust and rapidly computable measure to quantify the functional localization of a protein, compatible with image-based flow cytometry and cell sorting. By using the co-localization according to the present invention as a measure to assess the information about a protein's assignment to a particular cellular structure, it is possible to avoid the use of complex pattern recognition algorithms to identify particular structures. However, the present invention can also be used in combination with such pattern or structure recognition methods where specific areas of the cells are identified using criteria other than co-localization.

Moreover, co-localization is independent on the exact shape, size, or subcellular localization of a structure. This makes it a robust and easy to compute measure for the functional localization of a marker.

The present invention describes the use of subcellular co-localization as a parameter for cell sorting using flow cytometry instrumentation that incorporates optical imaging of cells with subcellular resolution for the purpose of defining criteria for the isolation and sorting of specific cells.

Here, a simple, robust, and reliable method for the detection of specific and biologically meaningful protein localizations compatible with fast image analysis and sorting of the cells directly after imaging, as well as respective devices and kits, are presented. It is proposed to use co-localization metrics as a criterion for cell sorting. Instead of pattern recognition to identify cellular substructures and to decide on whether a protein would localize to a specific structure, it is proposed to use metrics that quantify to what extent the protein of interest does co-localize with a reference protein or marker dye for a specific organelle as criterion for image-based flow cytometry associated cell sorting. Such metrics are insensitive to cell-to-cell differences in organelle and substructure shapes and it is shown herein that it allows robust cell sorting. FIGURES

Figure 1 shows a schematic drawing of the major components of an image- activated cell sorting system based on fluorescence microscopy and the detection of two or more proteins/markers for spatial co-localization to distinguish cells and for rapid decision-making and subsequent physical cell sorting.

The minimal system capable of performing physical cell sorting using the present method of spatial co-localization consists minimally of the following components:

- Microfluidic system 1 for streaming and physical sorting of cells,

- a dual-/multi-channel imaging device (i.e. microscope) suitable to acquire images of streaming cells,

- a processing unit for image processing where the said method of spatial co-localization detection is executed and a decision is made whether a cell should be physically sorted or not based on a predefined degree of spatial co-localization,

- a controller that controls and synchronizes cell flow, image acquisition, image processing, and physical cell sorting.

In detail, an optical system such as a microscope with camera is used to image cells with high speed and resolution and dual-/multi-channel detection of markers (see Figure 1 ). The acquired image is transmitted to the computer/processing unit where the image is analyzed (cell detection, spatial co-localization analysis of markers) using e.g. a CPU, GPU, FPGA or similar computing device. The computer compares the result of the analysis with the predefined value (from the user) and then generates signals via a signal generator for cell sorting. In Figure 1 , the signal generator provides signals for signal receivers that control the different parts of the microfluidic system, e.g. the sorting unit (i), the pumps to control the flow of the cells (ii), and the valves to control the fluids and cells (iii). Moreover, the processor synchronizes all components of the system to track cells and to trigger imaging and sorting at the right time.

Figure 2 shows schematics of a microfluidic system implementation for the continuous delivery of cells to the imaging site as well as for cell sorting based on the spatial co-localization method described in this invention.

As shown in Figure 2, the microfluidic system contains seven major parts, some of which are optional, as indicated. Cells are delivered via a cell-delivery unit 7 comprising inlets 2 through which cells 3, which can be provided as a cell mixture and, if necessary, the sheath fluids are delivered. Next, the cells pass the cell concentration unit 8, where the cell stream becomes more ordered and focused to the channel's center, in this case with the help of sheath fluids being injected from the two outer inlets 2. To control cell spacing, the cells can pass through a cell-focusing unit 9. In the case when cells are too far apart, the outlets/inlets 4 can be used for liquid removal. In the case when the cells are too close to each other the outlet/inlets 4 can be used to add liquid in order to space the cells apart to a desired spacing. Next, cells passing the cell-detection unit 10 are detected and their speed is calculated using two or more images of the cell taken at known time intervals. Using the calculated speed, the time the cell will be arriving at the cell-imaging unit 1 1 can be estimated for the precise capturing of the fluorescence image. Following image capturing and processing the cell arrives at the cell- sorting unit 12. A number of different techniques can be used for the physical sorting of the cells. In this example, sorting using electric fields using an electrode 5 is demonstrated. Finally, the sorted cells are led to their respective outlets in the collection unit 13. For the purpose of this example, two outlets were used, but any reasonable number of outlets should be possible.

Figure 3 shows the beam path in a schematic microscope according to one embodiment of the invention. A multi-channel fluorescence microscope suitable for the acquisition of multi-channel images from cells with different markers according to one embodiment of the invention comprises a microfluidic system 1 , an objective lens 14, multichroic beam splitter 15, a multichannel light source 16, mirrors 17, tube lenses 18, and an image/signal detector 19.

Figure 4 is schematic drawing of the method of spatial co-localization-based physical cell sorting.

Figure 4A illustrates the cell flow in the imaging and sorting unit of the microfluidic system that is used to stream the cells in front of the fluorescence microscope objective and to the site of physical sorting. The image shows a single yeast cell in a microfluidics unit. The cell is marked with a circle generated by an algorithm for rapid detection of streaming cells. Figure 4B schematically illustrates the beam path of the microscope unit for simultaneous multicolor fluorescence imaging for detection of streaming cells using rapid exposure times and methods to prevent motion blur. In Figure 4C a schematic drawing of data processing paths are shown. Images acquired by the microscopy unit are transferred to the data analysis unit and analyzed for protein or marker co-localization. Depending on the degree of spatial co-localization for each cell, an individual sorting decision is taken and the cell is sorted into one of the channels as soon as it is streamed to the sorting point, as illustrated in Figure 4A.

Figure 5 shows the discrimination of cells for physical cell sorting based on fast computing of protein/marker co-localization.

Figures 5A and 5B show images of yeast cells used as test sample. Cells from two different strains are shown. Each strain expresses two different proteins tagged with green fluorescent protein (GFP) or a red fluorescent protein (mCherry). In Figure 5A, signal overlap of the GFP and mCherry marker protein at the site of the nucleolus is detected. In contrast, no overlap can be shown in Figure 5B in images of cells expressing a GFP marker protein located at the nucleolus and a mCherry marker localizing to the cell periphery. Bright field images, fluorescence images and a color-merged image are shown. Bar - 5 μηη. Figure 5C shows an example calculations of parameter specifying co-localization of the two labeled proteins in the strains of Figure 5A (dots) and 5B (triangles). Each dot/triangle represents one cell. In Figure 5D the time of computation that is required to identify the cells (Hough transform) and compute the parameter for co-localization is shown.

Figure 6 is a schematic illustration of possible scenarios of spatial subcellular co- localization of markers of dyes within a cell 3. In general, localizations of a marker or dye at distinct subcellular structures 22 or the cell periphery 21 are possible (see Figure 6A). Moreover, it is also possible that the signal of a marker or a dye is detected in a part of the cell 23 or the entire cell 24 as depicted in Figures 6B and 6C.

Different scenarios are thinkable when defining spatial co-localization patterns. Such scenarios are illustrated in Figures 6D to 6I. Figures 6D to 6F show spatial co-localization patterns of signals detected for two different markers 25, 26 in two subcellular regions within a cell 3. No co-localization is detected when the signal of marker 25 and the signal of marker 26 do not overlap (see Figure 6D). Full spatial co-localization can be defined when both markers (25 and 26) localize to the same subcellular structure and their signals overlap 27 to the full extent, as illustrated in Figure 6E. Figure 6F shows one scenario of possible partial spatial co-localization, which can be defined as the area of both markers 25 and 26 exhibits some but not full overlap 27. In addition, spatial co-localization can be defined when more than two distinct subcellular regions to which markers 25 and 26 localize are detected (see Figures 6G to 6I). Similar to the scenario presented in Figure 6D, no co-localization is shown in Figure 6G as no overlap of the signals of marker 25 and marker 26 is detected. In Figure 6H, only one subcellular region shows total overlap 27 of marker signals 25 and 26 whereas two additional subcellular regions showing signal for marker 25 or marker 26 do not overlap. Such a scenario is regarded as another partial spatial co-localization. Figure 6I illustrates another possible scenario for partial spatial co-localization. In this scenario, the signal for marker 25 overlaps nearly to its full extent 27 with the signal of marker 26. Such an overlap is regarded as a high degree of partial spatial so-localization. EXAMPLES

Material and Methods:

Instruments and computing.

The instrument used for flow cytometry image-based cell sorting is based on a custom-made inverted epi-fluorescence wide field microscope that incorporates a high NA (numerical aperture) lens (1.4 NA, 60*), a high speed camera, a white light source and laser light sources for dual fluorescence color imaging, and a multi-view image splitter to enable the simultaneous detection of two fluorescence channels and a bright field channel on the same camera chip. The system is equipped with autofocus and electronics for high-speed image acquisition and reduction of motion blur and encompasses high-speed data transmission capable for in-line image analysis using a CPU- or GPU-based image-processing unit. In the example shown, a Hough transform was used for cell detection, and for co- localization analysis cross correlation and the correlation coefficient were computed. Microfluidics of cells was done using a PDMS (polydimethylsiloxane)- based microfluidics unit and cell gating based on results from the computational analysis of the images was achieved either using electrical fields or a laser-based motion direction unit.

Example 1 - Imaging and cell sorting speeds.

The instrument and microfluidics unit was designed to be able to image and physically sort cells into two populations at speeds up to 100 cells/second. Higher speeds are possible via adaptation of the system, i.e. by using a more powerful laser to obtain sufficiently short exposure times of the cells. Sorting of cells into > 2 different populations is possible and requires adaptations of the microfluidics devices. Depending of the number of cells imaged per second, the time available for image analysis and decision-making (Figure 4a, t) can become very short because of the fast movement of cells from the image acquisition site to the cell- sorting site. Hence, the decision whether a cell should be sorted has to be made in a short time, before the cell reaches the site of physical sorting, i.e. the branch point of the channel within the sorting unit where the cell is directed physically into one or the other channel. These two places, the image acquisition and the channel branch point, are ideally located within the field of view of the microscope to provide sufficient control over the sorting process. The exact layout of the microfluidics device and the field of view of the optical system (in the case of the present system: 200 μηη) and the cell movement speeds determine time between imaging and sorting (t). Typically, this value is in the range of < 1 ms up to a few milliseconds. This time window is the period during which computational analysis of the image has to occur (Figure 4c). This requires fast transfer of the image from the camera to the image-processing unit.

Example 2 - Test samples

To test the system, yeast cells were used, which carried two different proteins labeled either with a green or a red fluorescent protein expressed inside the same cells using genomic insertions of the fluorescent protein tags. Examples of strains where the two proteins show co-localization or no co-localization are shown in Figure 5a and 5b. Such test cells constitute a challenging test sample, since yeast cells are small (in the range of 3 to 5 μηη) and the expression levels of the fluorescent proteins is usually not high when expressed form genomic loci. In order to visualize the cellular distribution of proteins in these cells with sufficient sensitivity and resolution, an imaging system with a 60* or 100* and 1.4 NA objective lens is required.

Example 3 - Evaluation of co-localization as a robust and fast computable measure of protein localization

A critical step in image-based cell sorting is the analysis of subcellular localization patterns. Depending on the biological or experimental problem, different types of protein or marker localizations have to be expected and distinguished from each other in order to identify the cells of interest. A typical scenario encountered in fluorescence microscopy would be to distinguish cells where a protein does localize to a particular structure, i.e. a specific organelle such as the Golgi or the mitochondria, from cells where this protein is localized elsewhere. To evaluate how fast and how reliable different cells can be distinguished based on subcellular protein co-localization, various parameters that specify co-localization based on many different criteria were computed for the strains shown in Figure 5a and 5b and many more such yeast strains with many different localizations. The plot in Figure 5c exemplifies this by showing that large populations of cells from the strains shown in Figure 5a and 5b segregate into two populations. The time needed to segment the cell in the bright field image and to analyze the co- localization is a fraction of a ms (Figure 5d) which is fully compatible with in line co-localization analysis of cells and physical sorting in an image-based flow cytometry system.

Example 4 - Image acquisition

The quality of the images used for sorting is comparable to that obtained by a typical high-end wide-field fluorescence microscope equipped with a 60* or 100* 1.4 oil-immersion high NA objective, i.e. it exhibits a close-to diffraction-limited resolution of around 250 nm and a similar signal-to-noise ratio (SNR). Thus, image blur caused by cell motion ('motion blur') is kept to an absolute minimum (ideally well below the diffraction limit, i.e. < 100-200 nm). This limits the maximal possible exposure time. To achieve short exposure times, intense illumination pulses using collimated laser beams with beam-widths that match or only slightly exceed the cell diameter are required. In order to obtain sufficient SNR, cameras with extremely low read-noise (the limiting noise factor apart from shot noise) such as a sCMOS or EMCCD are used. For co-localization analysis of proteins with a marker protein for cell sorting, dual-view detection of two fluorescence channels simultaneously on the same camera chip is used. Moreover, precise image acquisition triggering and high data transfer rates for subsequent image analysis, e.g. using GPU based image processing, are employed.

Example 5 - Image analysis

Image analysis is a multistep procedure including initial image segmentation to isolate the cell from the background. To investigate how fast such basic image operations can be implemented, standard LabView segmentation functions were used to detect yeast cells in 256x256 pixel images. It was found that cell identification and segmentation required < 1 ms. A field of view (FOV) fitting the diameter of cells (4-8 μηη) can further reduce image size (approx. 100x200 pixel) and thus reduces segmentation time.

In order to optimize the system, co-localization analysis is implemented according to the present invention, since this is a simple and rapid way to score important and highly unique information about a protein without the need of sophisticated pattern recognition and cell classification. To test this, two different co-localization parameters of the GFP and mCherry signal were calculated, i.e. Persons (PCC) and Maeders (MCC) coefficient for a series of 256x256 pixel test images and processing speeds of ~1 and -2.5 ms, respectively, were found.

Subsequent image analyses and segmentations retrieves many different features (related to intensity, shape and texture) that can be used for cell classification. Moreover, unsupervised deep learning algorithms can be used for visual grouping using a metric learning procedure for appearance similarity. As this approach is founded on a highly parallel convolutional neural network, parallelization on a GPU is used. It has been established that data transfer from the camera to a GPU is fast enough to be compatible with GPU-based image analysis. Conclusion

Fluorescence microscopy retrieves information about the distribution and abundance of fluorescent dyes or other fluorescent labels with high contrast and sensitivity. Using functionalized dyes, i.e. dyes coupled to antibodies or genetically expressed dye-reporters (e.g. fluorescent proteins or dye-linking tags), it is possible to detect the entire pool of a given protein inside the cell. Depending on the localization of the protein to different organelles or other cellular structures, such as the nucleus or the cytoskeleton, different characteristic patterns can be observed. A trained scientist is able to derive from such images a hypothesis about the specific association of the labeled protein with organelles or structures, e.g. whether the protein localizes to the nucleus, the mitochondria or the Golgi, just to name a few examples. Alternatively, for assay or drug development, co- localization can be used as a criterion for the read out as a consequence of assay condition or treatment to isolate cells for analysis. For precise characterization of the readout, i.e. whether the localization of a reporter can be quantified, dual color fluorescence microscopy images and co-localization with another reporter with well-known localization are typically made. The present invention shows that co- localization is a reliable and fast computable parameter suitable for cell sorting. There are many ways how co-localization can be computed, most of which are based on simple algorithms and routines that all deliver more or less similar results and can be computed with similar speeds. The optimal algorithm for a particular experiment or assay can be rapidly tested, e.g. by human inspection of the images or by running a test experiment with control samples.

The use of co-localization as a measure of protein localization for image-activated cell sorting provides several advantages. In particular, the computation of co- localization as a means to deduce the localization of a protein does not require any information about structures or patterns of different organelles. Therefore, no deep-learning procedures or classifier training is needed to decide whether a protein localizes to a specific location, or not. Moreover, co-localization is robust versus biological variability of the shapes and dimension of organelles and subcellular structures in different cells. Such variability is very difficult to quantify in individual cells. Moreover, protein localization classifiers are typically not useful for the assessment of individual cells but instead require images from many cells of the same type in order to arrive at a decision. This is obviously not possible in image-activated cell sorting, since here individual cells are investigated and a decision has to be made based on single cell images.

Protein co-localization can be used in any physical cell-sorting instrument where the subcellular distribution of proteins, markers, or dyes constitutes the relevant information for the assay or experiment. Given optimized computation, much higher speeds are reachable and application in human health, where even extreme scenarios, such as the purification or isolation of specific cells from human blood, are possible.

In summary, the present invention provides a simple, robust, and reliable method for the detection of specific and biologically meaningful protein localizations compatible with ultra-fast image analysis and sorting of the cells directly after imaging, as well as respective devices and kits. The use of co-localization metrics as a criterion for cell sorting is proposed. Instead of pattern recognition to identify cellular substructures and to decide on whether a protein would localize to a specific structure, it is proposed to use metrics that quantify to what extent the protein of interest does co-localize with a reference protein or marker dye for a specific organelle as criterion for image-based flow cytometry associated cell sorting. Such metrics are insensitive to cell-to-cell differences in organelle and substructure shapes and it has been shown herein that it allows robust physical cell sorting.

List of Reference signs Microfluidic system

Inlets for cell and sheath fluid delivery from reservoirs Cell

Outlets/inlets for liquid removal/addition

Electrode

Outlets for collection of sorted cells

Cell-delivery unit

Cell-concentration unit

Cell-focusing unit

Cell-detection unit

Cell-imaging unit

Cell-sorting unit

Collection unit

Objective lens

Multichroic beam splitter

Multichannel light source

Mirror

Tube lens

Image/ signal detector

Cell boundary

Localization to cell periphery

Localization to one or more subcellular structures Localization to a part of the cell

Localization to the entire cell

Signal marker 1

Signal marker 2

Spatial co-localization signal of marker 1 and 2

Claims

A method for the physical sorting of cells of interest, comprising the steps of:

ii. applying a flow to the fluid;

iii. acquiring at least one electronic image of an individual cell;

The method of claim 1 , wherein spatial patterns define an area or volume and the area or volume of the spatial patterns is calculated on the basis of pattern determining parameters.

The method of claim 1 or 2, wherein the degree of overlap is determined by calculating the overlapping area or volume of the spatial patterns of the two signals, calculating the area or volume of the full spatial pattern of the signal of the first marker and calculating the quotient of the overlapping area or volume to the full area or volume.

4. The method of any one of claims 1 to 3, wherein the lower limit of the predefined range is 0 and the upper limit is 60 %, preferably 40 %, more preferably 30 %, most preferably 20 %.

5. The method of any one of claims 1 to 3, wherein the upper limit of the predefined range is 100 % and the lower limit is 40 % preferably 60 %, more preferably 70 %, most preferably 80 %.

6. The method of any one of claims 1 to 5, wherein the first of the at least two detectable markers labels a protein of interest.

7. The method of any one of claims 1 to 6, wherein the second of the at least two detectable markers labels a cellular substructure of interest.

8. The method of any one of claims 1 to 7, wherein steps (iii), (iv), (v) and (vi) are performed directly after each other for each individual cell to be analyzed.

9. The method of any one of claims 1 to 8, wherein determination of the degree of spatial co-localization is independent of the signal intensities.

10. The method of any one of claims 1 to 9, wherein steps (iv) and (v), preferably steps (iii), (iv) and (v), more preferably steps (iii), (iv), (v) and (vi) are performed in an automated manner.

1 1. The method of claim 10, wherein steps (iii), (iv) and (v) and (vi) are performed for each individual cell within at most 5 ms.

12. The method of any one of claims 1 to 1 1 , wherein the flow rate of the fluid applied in step (ii) is at least 1 nL/sec, preferably at least 100 nL/sec, more preferably at least 1 μΙ_/εβο, most preferably at least 100 μΐ,/εβα

13. The method of any one of claims 1 to 12, wherein the steps (iii) to (vi) are at a speed of 1 cell/second, preferably, 50 cells/second more preferably 100 cells/second.

14. The method of any one of claims 1 to 13, wherein one electronic image per cell is acquired.

15. The method of claim 14, wherein the excitation beam and detection light path for image acquisition move with the cell, in particular at the same speed of the cell to allow a constant image acquisition of the moving cell.

16. The method of any one of claims 1 to 15, wherein more than one electronic image is acquired per cell by using different wavelengths of the excitation beam and separation of the emission images for simultaneous detection, using one or more cameras, or using two, or using more cameras.

17. The method of any one of claims 1 to 16, wherein step (iii) is performed using a fluorescence microscope equipped with one or more high-speed camera(s).

18. The method of any one of claims 1 to 17, wherein step (iii) is performed using an objective with a numerical aperture of at least 1.0, preferably at least 1.2, more preferably at least 1.3, most preferably at least 1.4.

19. The method of any one of claims 1 to 18, wherein the distance between objective and cell during image acquisition is constant with a variability in the range of -below 8 μηη, preferably below 4 μηη, more preferably below 2 μηη.

20. The method of any one of claims 1 to 19, wherein the structures of the cell have a resolution in the electronic image of less than 3 μηη, preferably less than 2 μηη, more preferably less than 1.0 μηη, most preferably at least 0.5 μηη.

21. A device for acquiring electronic images of individual cells and physically sorting cells of interest, comprising:

a) a microfluidic system;

c) a multichannel detection system that is capable of generating spatially resolved two- or three-dimensional localization maps representing the subcellular distribution of at least two detectable markers in the cell; d) hardware that is configured to execute analysis software computing the degree of co-localization of the at least two detectable markers in the localization maps generated in step (c); and

e) a cell-sorting unit.

22. The device of claim 21 , wherein the microfluidic system contains an inlet, a cell-concentration unit, a cell-focusing unit, and an imaging unit.

23. The device of claim 21 or claim 22, wherein the beams emitted from the light sources are moveable within the imaging unit in direction of the flow of the fluid.

24. The device of any one of claims 21 to 23, wherein the detection system comprises a detection light path, which is moveable within an imaging unit in direction of the flow of the flu id, in particular in combination with the beams emitted from the light sources.

25. The device of any one of claims 21 to 24, wherein the light sources are laser light sources.

26. The device of any one of claims 21 to 25, wherein the multichannel detection system is located at the imaging unit and is a fluorescence microscope equipped with a high-speed camera.

27. Use of the degree of spatial overlap of the signals of at least two detectable markers in an electronic image of a cell of interest that is labeled with said at least two detectable markers as a criterion for the physical sorting of said cell of interest.

28. A cell or cell pool that has been isolated with the method of any one of claims 1 to 20, or a cell clone derived from said cell or cell pool.

29. A kit comprising at least two detectable markers and instructions for performing the method of any one of claims 1 to 20.