Method of screening single cells for the production of biologically active agents

[0001] This application claims the benefit of U.S. Provisional Application Nos. 61/072,985, filed Apr. 5, 2008 and 61/141,579, filed Dec. 30, 2008, which applications are incorporated herein by reference. This application is related to two other applications filed concurrently herewith on Apr. 3, 2009 which two applications have attorney docket numbers SCTI-0001 and SCTI-0002, which two related applications are incorporated herein by reference in their entirety.

[0002] This invention generally relates to a methods, devices and kits for screening single cells for the production of one or more biologically active agents of interest, such as a protein, nucleic acid, or a protein and the nucleic acid encoding same.

[0003] It is often desirable to screen single cells or small populations of cells for the production of one or more biologically active agents. Such information may be used to determine the status or activity of cells following exposure to a biologically active agent, for example drug candidates such as small molecules, proteins, or siRNAs, to determine the effects of those biologically active agents upon the cells. Such information may also be useful to identify cells that are expressing a particular variant of a biologically active agent, for example, a SNP, or an antibody. Such information may also be useful to identify nucleic acid sequences that encode biologically active agents of interest, for example antibodies, to utilize the frequency of nucleotide sequences in a plurality of cells to predict or model the behavior of the proteins they create or the resulting phenotypes. The methods, devices and kits herein are directed to these and similar utilities.

[0004] Methods, devices and kits for screening single cells for the production of one or more biologically active agents of interest, such as a protein, nucleic acid, or a protein and the nucleic acid encoding same.

[0005] In practicing the subject method, cells that are producing one or more biologically active agents of interest are separated into individual wells. The cells are allowed to produce the one or more biologically active agent in the wells, and the one or more biologically active agents in wells is then contacted with one or more binding agents bound to a solid surface. The binding agent(s) specifically binds the one or more biologically active agents of interest in the wells, and information relating to the cells in the wells is then determined based upon the binding of the one or more biologically active agents of interest to the binding agent or to a target binding protein, for example, information on whether a signaling pathway is active/inactive in the cell of the well. This information is then associated to a particular addressed region of the solid surface so as to identify the well of particular interest. In some aspects of the invention, a biologically active agent is provided to the well, thereby affecting the status and/or activity of the cell in the well.

[0006] In some aspects of the invention, the method further comprises obtaining the nucleotide sequence encoding the at least one biologically active agent of interest, for example the one or more polypeptides comprising a protein of interest, and correlating the information about the binding properties of the biologically active agent of interest with the nucleotide sequence(s) encoding the biologically active agent of interest.

[0007] In some aspects of the invention, only the nucleotide sequence encoding the at least one biologically active agent of interest is obtained. In these aspects, cells that are producing the at least one biologically active agents of interest are separated into individual wells, and the nucleotide sequence encoding the one or more biologically active agents is obtained. This data can then be used to obtain metadata about the biologically active agent of interest, for example, the binding properties of the biologically active agent(s), or the frequency with which the biologically active agent of interest is represented in the cell population that was screened.

[0008] Thus, the present invention provides an efficient method for analyzing a large number of cells individually rather than analyzing a population of cells and reporting the average measurement for the population. Cells are separated and processed in small microwells where diffusion of molecules among microwells is retarded or blocked. Additionally, each microwell, due to its small dimensions, facilitates rapid reaction rates, such as mRNA hybridization or capture of proteins. Furthermore, the proteins captured on the solid substrate form an array on the substrate where their kinetic properties can be analyzed en masse either for their contact on the solid substrate or when contacted with labeled affinity ligands common to all captured proteins or complexes.

[0009] These and other objects, advantages, and features of the invention will become apparent to those persons skilled in the art upon reading the details of the methods and embodiments of the invention as more fully described below.

[0021] As used herein, a biologically active agent is any agent that possesses an activity in a biological system. Examples of biologically active agents include small molecule compounds; polypeptides, e.g. proteins; siRNAs; and oligonucleotides. A plurality of such biologically active agents would include, for example, 2 or more of such agents; in some cases, 3 or more agents; 5 or more agents; 10 or more agents; 20 or more agents; 50 or more agents; 100 or more agents; 500 or more agents; 1000 or more agents; 5000 or more agents; 10,000 or more agents; 30,000 or more agents; 100,000 or more agents; or 1,000,000 or more of such agents.

[0022] As used herein, the terms “tag(s)” refers to a moiety that identifies the physical location of its origin. In some embodiments, a tag may be an oligonucleotide tag(s), wherein the sequence of the oligonucleotide serves as the tag to identify the physical location of its origin.

[0023] The term “oligonucleotide” as used herein includes linear oligomers of natural or modified monomers or linkages, including deoxyribonucleosides, ribonucleosides, anomeric forms thereof, peptide nucleic acids (PNAs), and the like, capable of specifically binding to a target polynucleotide by way of a regular pattern of monomer-to-monomer interactions, such as Watson-Crick type of base pairing, base stacking, Hoogsteen or reverse Hoogsteen types of base pairing, or the like. Usually monomers are linked by phosphodiester bonds or analogs thereof to form oligonucleotides ranging in size from a few monomeric units, e.g. 3-4, to several tens of monomeric units, e.g. 40-60. Whenever an oligonucleotide is represented by a sequence of letters, such as “ATGCCTG,” it will be understood that the nucleotides are in 5′.-0.3° order from left to right and that “A” denotes deoxyadenosine, “C” denotes deoxycytidine, “G” denotes deoxyguanosine, and “T” denotes thymidine, unless otherwise noted. Usually oligonucleotides of the invention comprise the four natural nucleotides; however, they may also comprise non-natural nucleotide analogs. It is clear to those skilled in the art when oligonucleotides having natural or non-natural nucleotides may be employed, e.g. where processing by enzymes is called for, usually oligonucleotides consisting of natural nucleotides are required. “Perfectly matched” in reference to a duplex means that the poly- or oligonucleotide strands making up the duplex form a double stranded structure with one other such that every nucleotide in each strand undergoes Watson-Crick basepairing with a nucleotide in the other strand. The term also comprehends the pairing of nucleoside analogs, such as deoxyinosine, nucleosides with 2-aminopurine bases, and the like, that may be employed. In reference to a triplex, the term means that the triplex consists of a perfectly matched duplex and a third strand in which every nucleotide undergoes Hoogsteen or reverse Hoogsteen association with a basepair of the perfectly matched duplex. Conversely, a “mismatch” in a duplex between a tag and an oligonucleotide means that a pair or triplet of nucleotides in the duplex or triplex fails to undergo Watson-Crick and/or Hoogsteen and/or reverse Hoogsteen bonding.

[0024] As used herein, “nucleoside” includes the natural nucleosides, including 2′-deoxy and 2′-hydroxyl forms, e.g. as described in Komberg and Baker, DNA Replication, 2nd Ed. (Freeman, San Francisco, 1992). “Analogs” in reference to nucleosides includes synthetic nucleosides having modified base moieties and/or modified sugar moieties, e.g. described by Scheit, Nucleotide Analogs (John Wiley, New York, 1980); Uhlman and Peyman, Chemical Reviews 90:543-584 (1990), or the like, with the only proviso that they are capable of specific hybridization. Such analogs include synthetic nucleosides designed to enhance binding properties, reduce complexity, increase specificity, and the like.

[0025] The terms “nucleotide sequence” and “nucleic acid sequence” are used interchangeably herein to describe the sequence of a polynucleotide, for example a gene encoding a polypeptide. Likewise, the terms “sequence determination”, “determining a nucleotide sequence”, and “determining a nucleic acid sequence” are used interchangeably to describe determining the sequence of nucleic acids or nucleotides that make up a polynucleotide, including determination of partial as well as full sequence information of the polynucleotide. That is, the term includes sequence comparisons, fingerprinting, and like levels of information about a target polynucleotide, as well as the express identification and ordering of nucleosides, usually each nucleoside, in a target polynucleotide. The term also includes the determination of the identification, ordering, and locations of one, two, or three of the four types of nucleotides within a target polynucleotide.

[0026] As used herein, “epitope” is the part of a macromolecule that is recognized by the immune system, specifically by antibodies, B cells, or T cells. Although epitopes are usually thought to be derived from nonself proteins, sequences derived from the host that can be recognized are also classified as epitopes. Most epitopes recognized by antibodies or B cells can be thought of as three-dimensional surface features of an antigen molecule; these features fit precisely and thus bind to antibodies.

[0027] As used herein, “hybridization” refers to the process wherein cellular RNA or single stranded DNA interacts with oligonucleotides having substantial sequence complementariy, wherein duplexes are formed in said regions of sequence complementarity.

[0028] As used herein, “microwell” refers to sub millimeter structures with a volume between 1 picoliter and 500 nanoliters. The microwell is typically constructed in a shape that allows dense packing on a planar substrate, i.e.: the shape is triangular, rectangular, or hexagonal. Microwells can be either opened by removing one surface, usually at the top, or closed by placing said top in contact with other surfaces. The microwell can be homogeneous, or constructed out of dissimilar materials, including but not limited to glass, photoresist, or polydimethylsiloxane (PDMS).

[0029] As used herein, “paratope” refers to the part of an antibody that recognizes the epitope of an antigen.

[0030] As used herein, “phOX” refers to 4-{[(Z)-(5-OXO-2-PHENYL-1,3-OXAZOL-4(5H)-YLIDENE)METHYL]AMINO}BUTANOIC ACID, represented by the chemical formula C₁₄H₁₂N₂O₄

[0031] As used herein, “rotamers” refer to low energy side-chain conformations. The use of a build-library of rotamers allows anyone determining or modeling a structure to try the most likely side-chain conformations, saving time and producing a structure that is more likely to be correct. This is, of course, only the case if the rotamers really are the correct low energy conformations. Libraries address this quality issue in a number of ways: they use only very high resolution structures (1.8 Å or better), remove side chains whose position may be in doubt using a number of filters, use the mode rather than the mean of observed conformations (which has a number of advantages), and make efforts to remove systematically misfit conformations.

[0032] As used herein, “mutations” are changes to the nucleotide sequence of the genetic material of an organism. Mutations can be caused by copying errors in the genetic material during cell division, by exposure to ultraviolet or ionizing radiation, chemical mutagens, or viruses, or can occur deliberately under cellular control during processes such as hypermutation. In multicellular organisms, mutations can be subdivided into germ line mutations, which can be passed on to descendants, and, as used herein, “somatic mutations”, which cannot be transmitted to descendants in animals. B cells undergo somatic mutations during the process of affinity maturation.

[0033] To be able to perform their biological function, proteins fold into one or more specific spatial conformations, driven by a number of noncovalent interactions such as hydrogen bonding, ionic interactions, Van der Waals' forces and hydrophobic packing. As used herein, “structure” refers to four distinct aspects of a protein's structure: a) the primary structure is the amino acid sequence of the peptide chains, b) the secondary structure is the highly regular sub-structures (alpha helix and strands of beta sheet) which are locally defined, meaning that there can be many different secondary motifs present in one single protein molecule, c) tertiary structure is the three-dimensional structure of a single protein molecule; a spatial arrangement of the secondary structures, and d) the quaternary structure is the complex of several protein molecules or polypeptide chains, usually called protein subunits in this context, which function as part of the larger assembly or protein complex.

[0034] As used herein, a protein is an organic compound made of amino acids arranged in a linear chain and joined together by peptide bonds between the carboxyl and amino groups of adjacent amino acid residues. Proteins may comprise a single polypeptide chain of amino acids or multiple polypeptide chains of amino acids, for example, multiple polypeptides bound together by disulfide bonds, e.g. an antibody. As used herein, a “protein complex” is a group of two or more associated proteins formed by protein-protein interactions. Protein complexes are a form of quaternary structure. Protein complexes that are stable over time are herein called “cis protein complexes”. Protein complexes that are transient are herein called “trans protein complexes”. The present invention relates to both cis and trans protein complexes.

[0035] As used herein, “kinetic properties” refer to the rates of reaction k_off, k_on, and their ratio K_D between cis and/or trans protein complexes. For a binary protein complex, the dissociation constant K_D is monotonically related to the Gibbs free energy which describes the work obtainable from an isothermal, isobaric process, conditions closely approximated in living systems.

[0036] As used herein, a “target binding protein” or “target protein” is a protein to which a biologically active agent of interest binds. The target binding protein could be a protein that typically binds to the biologically active agent of interest; for example, a cytokine, wherein the biologically active agent of interest is a receptor; or an antigen, wherein the biologically active agent of interest is an antibody. Alternatively, the target binding protein could be a protein that does not typically bind to the biologically active agent of interest; for example, a control cytokine.

[0037] As used herein, “binding agent” or “capture agent” is a molecule used to immobilize a biologically active agent. Capture agents can be oligonucleotides, DNA, RNA, protein, small molecules, peptides, aptamers, etc which have an affinity for their respective natural or artificial ligands.

[0038] As used herein, a solid surface or solid surface is any sort of surface or support. It may be made of glass, plastics, nitrocellulose, polyvinylidene fluoride, or other highly non-reactive materials. A binding/capture agent may be attached, in which case, the binding/capture agent may coat the solid surface, or may be contained to discrete locations on the solid surface, for example spotted in spots, or localized into pads, or configured into a line.

[0039] The term “monoclonal antibody” relates to an antibody chosen from a mixture of different antibodies. All monoclonal antibodies of the same specificity are identical except for natural mutants thereof. The term “antibody” as used herein is understood to mean intact molecules of immunoglobulins as well as fragments thereof (Fab, F(ab′), Fv, scFv).

[0040] As used herein, a “ligand” is a substance that is able to bind to and form a complex with a biomolecule to serve a biological purpose.

[0041] The term “B cell” is used herein to mean an immune cell that develops in the bone marrow and is highly specialized for making immunoglobins and antibodies. A B cell is a lymphocyte which is derived from bone marrow and provides humoral immunity. A B cell recognizes antigen molecules in solution and matures into a plasma cell. Thus, when the term “B cell” is used herein it is intended to encompass cells developed from B cells such as plasma cells.

[0042] The term “plasma cell” is intended to mean a cell that develops from a B lymphocyte in reaction to a specific antigen. Plasma cells are found in bone marrow and blood. A plasma cell may also be called a plasma B cell or plasmacyte and are cells in the immune system which secrete large amounts of antibodies. Plasma cells differentiate from B cells upon stimulation by CD4+ lymphocytes. A plasma cell is a type of white blood cell that produces antibodies and is derived from an antigen-specific B cell. Throughout this application the term “B cell” is intended to encompass “plasma cells” and vice versa. In general both are intended to encompass terms referring to cells which produce antibodies of interest.

[0043] Biologically active agents are characterized by their “binding affinity” to a given target biologically active agent, for example a protein. For example, an antibody is characterized by its affinity to a binding site or epitope. Every antibody is comprised of a particular 3-dimensional structure of amino acids, which binds to another structure referred to as an epitope or antigen.

[0044] The binding of an antibody to its antigen is a simple bimolecular, reversible reaction. If the antibody is represented by Ab and the antigen by Ag, the reaction can be analyzed by standard kinetic theory. Assuming a single binding site the reaction is represented by the equation I as follows:

[0045] $\begin{array}{cc}\mathrm{Ag}+\mathrm{Ab}\begin{array}{c}\underset{\rightharpoonup }{k_1}\\ \stackrel{\leftharpoondown }{k_2}\end{array}\mathrm{Ag}-\mathrm{Ab}.& I\end{array}$

[0046] where Ag-Ab is the bound complex. The forward and reverse binding reactions are represented by rate constants k1 and k2 respectively. The “binding affinity” of the antibody to the antigen is measured by the ratio of complexed to free reactants at equilibrium. The lower the concentration of the reactants at equilibrium, the higher the binding affinity of the antibody for the antigen. In the field of immunology, the binding affinity is represented by an “affinity constant” which is represented by the symbol “K” or sometimes referred to as “Ka”. The “K” is defined by the equation II as follows:

[0047] $\begin{array}{cc}K=\frac{\left[\mathrm{Ag}-\mathrm{Ab}\right]}{\left[\mathrm{Ag}\right]\left[\mathrm{Ab}\right]}=\frac{k_1}{k_2}.& \mathrm{II}\end{array}$
where the brackets denote concentration in moles per liter or liters per mole.

[0048] A typical value for the binding affinity Ka which is also referred to as “K” and is the “affinity constant” which for a typical antibody is in a range of from about 10⁵ to about 10¹¹ liters per mole. The Ka is the concentration of free antigen needed to fill half the binding sites of the antibody present in solution with the antigen. If measured in liters per mole a higher Ka (e.g. 10¹¹) or higher affinity constant indicates a large volume of solvent, a very dilute concentration of free antigen, and as such indicates the antibody has a high binding affinity for the epitope.

[0049] If the Ka is measured in moles per liter a low Ka (e.g. 10⁻¹¹) indicates a less concentrated solution of the free antigen needed to occupy half of the antibody binding sites, and as such a high binding affinity.

[0050] Equilibrium is achieved in order to measure the Ka. More specifically, the Ka is measured when the concentration of antibody bound to antigen [Ag-Ab] is equal to the concentration of the antibody [Ab]. Thus, [Ag-Ab] divided by [Ab] is equal to one. Knowing this, the equation II above can be resolved to the equation III as follows:

[0051] $\begin{array}{cc}K=\frac{1}{\left[\mathrm{Ag}\right]}.& \mathrm{III}\end{array}$

[0052] In equation III the units for K are liters per mole. Typical values in liters per mole are in a range of from about 10⁵ to about 10¹¹ liters per mole.

[0053] The inverse of the above equation is K=[Ag] where the units for K are in moles per liter, and the typical values are in a range of 10-11 to 10⁻⁵ moles per liter.

[0054] The above shows that typical binding affinities can vary over six orders of magnitude. Thus, what might be considered a useful antibody might have 100,000 times greater binding affinity as compared to the binding affinity of what might be considered a different antibody, which is also considered useful.

[0055] Based on the above it will be understood that binding characteristics of an antibody to an antigen can be defined using terminology and methods well defined in the field of immunology. The binding affinity or “K” of an antibody can be precisely determined.

[0056] Those skilled in the art will understand that a high degree of binding affinity does not necessarily translate to a highly effective drug. Thus, when obtaining drug candidates the candidates showing a wide range of binding affinities may be tested to determine if they obtain the desired biochemical/physiological response. Although binding affinity is important, some drug candidates with high binding affinity are not effective drugs and some drug candidates with low binding affinity are effective drugs.

[0057] The term “metadata” is used to describe “data about other data.” An item of metadata may describe an individual datum, or content item, or a collection of data including multiple content items and hierarchical levels, for example a database schema. As used herein, metadata is definitional data that provides information about or documentation of other data acquired in the course of using the present invention. For example, metadata documents data about data elements or attributes, (name, size, data type, etc) and data about records or data structures (length, fields, columns, etc) and data about data (where it is located, how it is associated, ownership, etc.). Metadata may include descriptive information about the context, quality and condition, or characteristics of the data. For example, metadata would be the data about the frequency of representation of a particular antibody in a population of cells, as determined by acquiring data on the sequences of antibodies expressed by a plurality of individual cells in a population.

[0058] Methods, devices and kits are provided for screening single cells for the production of one or more biologically active agents of interest, such as a protein, nucleic acid, or a protein and the nucleic acid encoding same.

[0059] Before the present invention is further described, it is to be understood that this invention is not limited to particular methods and devices described, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present invention will be limited only by the appended claims.

[0060] Where a range of values is provided, it is understood that each intervening value, to the tenth of the unit of the lower limit unless the context clearly dictates otherwise, between the upper and lower limits of that range is also specifically disclosed. Each smaller range between any stated value or intervening value in a stated range and any other stated or intervening value in that stated range is encompassed within the invention. The upper and lower limits of these smaller ranges may independently be included or excluded in the range, and each range where either, neither or both limits are included in the smaller ranges is also encompassed within the invention, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the invention.

[0061] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, some potential and preferred methods and materials are now described. All publications mentioned herein are incorporated herein by reference to disclose and describe the methods and/or materials in connection with which the publications are cited. It is understood that the present disclosure supercedes any disclosure of an incorporated publication to the extent there is a contradiction.

[0062] It must be noted that as used herein and in the appended claims, the singular forms “a”, “an”, and “the” include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to “a cell or sequence” may include a plurality of such cells or sequences and reference to “the well or addresses” may include reference to one or more wells or addresses and equivalents thereof known to those skilled in the art, and so forth.

[0063] The publications discussed herein are provided solely for their disclosure prior to the filing date of the present application. Nothing herein is to be construed as an admission that the present invention is not entitled to antedate such publication by virtue of prior invention. Further, the dates of publication provided may be different from the actual publication dates which may need to be independently confirmed.

[0064] As summarized above, methods, devices and kits are provided for screening single cells for the production of biologically active agents of interest, and optionally for obtaining the nucleotide sequences that encode those biologically active agents. In further describing the invention, the subject methods are described first, followed by a review of the devices and kits for use in practicing the subject methods.

[0065] One embodiment of the method of the invention comprises a method of obtaining information from a plurality of isolated cells. The information may be simultaneously obtained from a large number of cells by placing the large number of cells in individual wells. An attempt is made to include a single cell within a single well. However, when carrying out the method involving hundreds, thousands or even tens of thousands of cells and wells it may be that some wells do not include a cell and other wells include more than one cell.

[0066] A method of the present invention may begin with immunizing an animal and extracting antibody producing cells from that animal. However, the process may begin with the cells already having been extracted and placing the cells into wells of a well tray. Those skilled in the art will recognize that the methodology of the invention can be carried out in a variety of different ways. In one embodiment a well tray comprised of microwells is used. The microwells have a volume sufficient to accommodate a single cell and liquid nutrient to support the cell for a limited period of time during which the cell produces antibodies. Those skilled in the art will recognize that it is desirable to place a single cell within each microwell. However, when the process is actually carried out some wells will not contain a cell and some wells may contain two or more cells. Although this can limit the effectiveness of the invention, the invention can be carried out when only a small percentage of the microwells contain a single cell, e.g. 1%, 5%, 10%, 50% or more of the cells only contain a single cell and the remainder of the wells contain no cell or a plurality of cells, i.e. 2 or more cells. It is desirable if a very high percentage, 70% or more, 80% or more, 90% or more, or 95% or more of the wells contain a single cell and only a single cell. This makes it possible to utilize all of the wells and specifically relate antibodies produced in the well to a single cell. Those skilled in the art will also recognize that even though it is desirable that all of the cells produce antibodies some of the cells may not produce antibodies or may produce antibodies in insufficient amounts to be detectable. The invention that can be carried out when a relatively small number of the cells are actually producing antibodies. For example, it may be that only 1%, 5%, 10% or 50% of the cells in the wells are actually producing antibodies (or other biologically active molecule) in detectable amounts even though it is desirable to obtain a high percentage of antibody producing cells, e.g. 70% or more, 80% or more, 90% or more, or 95% or more of the cells placed in wells are producing antibodies (or other biologically active molecule). The more wells containing a single cell which is producing antibodies the greater the efficiency of the methodology of the invention.

[0067] It is desirable that each of the regions or areas of the surface have the binding agent such as an antibody binding agent such as protein A bound to its surface at that region. However, the invention could be carried out when a relatively small percentage of those areas have a protein binding agent bound to its surface. For example, the invention could operate with only 1%, 5%, 10%, 50% or more of the areas on the surface having a binding agent bound thereto. It is desirable that a high percentage, 70% or more, 80% or more, 90% or more, 95% or more or preferably 100% of the areas have a binding agent bound at the region which corresponds to the well or mircrowell. In a similar manner the microwells may have bound to their surface a polynucleotide sequence such as a sequence which specifically binds to a polynucleotide sequence encoding a portion of an antibody or other protein of interest. It is desirable that all of microwells have the polynucleotide bound to a surface. However, the invention can be operative when only a small number of the wells, e.g. 1%, 5%, 10% or more of the microwells have the polynucleotide bound to its surface. It is preferable that when using this embodiment of the invention that a high percentage of the microwells, e.g. 70% or more, 80% or more, 90% or more, 95% or more or more preferably 100% of the microwells have the polynucleotide sequence bound to its surface.

[0068] The cells are cells such as B cells or plasma cells which produce antibodies and the antibodies which are produced in the wells are brought into contact with a binding agent such as protein A which is bound to a surface which may be a membrane. This surface may have a plurality of addressable regions or will include regions which can be specifically relatable to individual wells. The wells may be in a well tray which has a well density of 100 or more wells per cm² or 1,000 or more wells per cm² and the wells may include a detectable marker which makes it possible to determine the position of a particular well relative to other wells and the marker may be a marker cell or a group of markers which could include a dye, a nucleotide sequence, a radioactive label or a quantum dot.

[0069] After contacting antibodies in the wells with the binding agent a process is carried out to determine binding information relating to the binding of the antibodies which are bound to the surface to an antigen which is preferably a specific known antigen in order to determine information such as the binding affinity of the antibodies to that antigen. The binding information relating to specific areas on the surface is then associated with the particular well in which the antibodies were obtained. Thus, it is possible to simultaneously obtain information from a large number of different wells relating to a large number of different antibodies and associate the antibodies with the wells from which the antibodies were obtained.

[0070] After determining particular antibodies on the areas of the surface which are of interest and associating those with the wells of interest it is possible to obtain specific polynucleotide information from the particular wells which information is generally sequence information regarding messenger RNA obtained from particular wells of interest. The messenger RNA may be obtained by binding messenger RNA in the wells to sequences which selectively bind to sequences which encode antibodies such as sequences which encode light and heavy chains of antibodies.

[0071] The messenger RNA obtained can be converted to cDNA. The cDNA may contain tags which are specific to the well of particular interest. The tag may be used to associate binding information relating to antibodies of particular interest with messenger RNA from wells of particular interest. The invention is described in further detail below with reference to the figures.

[0072] In practicing the subject method, cells that are producing one or more biologically active agents of interest are separated into individual wells. In some embodiments, the cells are substantially the same in every well, for example, cells from cultured cell lines, cells from primary cell preparations, or engineered cells. In other embodiments, the cells are different in each well, that is, they may be of different origins or have genetic differences that make them unique from one another, e.g. plasma cells, or different kinds of cells from the same human subject, or the same kind of cells from different human subjects. In some embodiments, only one cell is placed per well. In other embodiments, multiple cells are placed per well. Typically, the ratio of the number of microwells containing 2 or more cells divided by the number of microwells containing a single cell is less than 20%.

[0073] The cells are allowed to produce the one or more biologically active agent in the wells, and the one or more biologically active agents in the wells is then contacted with one or more binding agents attached to a solid surface. In some embodiments, the binding agent that is attached to the solid surface is the same for all wells. In some embodiments, the binding agent that is attached to the solid is different for different wells. Examples of binding agents include Protein A, Protein G, Protein L, Protein A/G, anti-IgG Fcγ subclass-specific antibodies, and a target protein for said at least one biologically active agent produced by said cell. The solid surface to which these binding agents are attached may be made of glass, plastics, nitrocellulose, polyvinylidene fluoride, or other highly non-reactive materials, and is comprised of addressable regions, such that each address is specifically related to one of the individual wells containing the cell producing the biologically active agent of interest.

[0074] The binding agent(s) on the solid surface specifically binds the one or more biologically active agent(s) of interest produced by the cells in the wells, thereby “capturing” the biologically active agent(s) of interest. Information relating to the cells may then be determined, based upon the binding of the one or more biologically active agents of interest to the binding agent itself. For example, the presence or absence of a biologically active agent in the well can be determined by the presence of absence of binding to the binding agent on the solid surface, wherein the presence/absence of that biologically active agent indicates if a signaling pathway has been activated/deactivated in the cell of the well. Alternatively, information relating to the cells may be determined based upon the binding of the captured one or more biologically active agents of interest to a target binding protein provided subsequently. For example, the binding affinity of the captured biologically active agent for a target protein can be determined by affinity studies using the biologically active agent in its captured state on the solid surface, wherein a particular affinity of the captured biologically active agent for the target protein indicates that the cell that produced the biologically active agent is a cell of interest. This information is then associated to a particular addressed region of the solid surface so as to identify the well of particular interest.

[0075] To address the regions on the solid surface such that they are in registration with the wells comprising the cells of interest, and to provide a biologically active agent whose binding to a binding agent is known and to which the experimental samples may be compared, it may be desirable to include marker cells with a known marked biologically active complex into the cell population under investigation. A marker cell would be a cell that is optionally capable of expressing an antibody, for example, a hybridoma. Such lines are deposited and described in the literature and easily acquired. These cells are stained with an antigen with a distinct fluorophore so they can be identified in the microwells, or genetically engineered to express a fluorescent tag that can be tracked. Likewise, the antibody can be captured on the solid surface and stained with the appropriate antigen with the distinct dye to identify its location on the solid surface. As such, the position of the marker cells on the microarray slide can be used as landmarks to address regions on the solid surface. The biologically active agents of other cells captured on the solid surface can then be matched based on their position relative to the marker cells and the antibodies they produce.

[0076] In some embodiments, a biologically active agent is provided to the cells in the wells, for example, a small molecule compound, an siRNA, a protein, or an oligonucleotide. In some embodiments, this provided biologically active agent is the same in every well. Thus, one provided biologically active agent may be assayed for its impact on a plurality of cell types or on a plurality of signaling pathways. In other embodiments, this provided biologically active agent is different in each well. Thus, a plurality of provided biologically active agents may be assayed for their impact on a single cell type or on a single signaling pathway.

[0077] In some aspects of the invention, the method further comprises obtaining the nucleotide sequence encoding the at least one biologically active agent of interest, for example a polypeptide, or multiple polypeptides if the biologically active agent of interest is a complex of polypeptides, and correlating the information about the binding properties of the biologically active agent of interest with the nucleotide sequence encoding the biologically active agent of interest. For example, the nucleotide sequence may be obtained by obtaining mRNA from the cell, synthesizing cDNA from the mRNA and sequencing the cDNA.

[0078] In some embodiments, the nucleotide sequence encoding the biologically active agent of interest in a select well is obtained. In other embodiments, the nucleotide sequence of the biologically active agent of interest produced by the all of the cells in all of the wells is determined simultaneously. In these embodiments, a well-specific tag, e.g. an oligonucleotide tag, is provided to the wells, and obtaining the nucleotide sequence encoding the at least one biologically active agent comprises incorporating the well-specific tag into the cDNA as it is being synthesized, pooling all of the cDNAs, and sequencing the pooled cDNAs by ultra-high throughput DNA sequencing technology to determine the nucleotide sequence of the at least one biologically active agent of interest in each originating cell. The oligonucleotides may be provided to the wells prior to placing the cells in the wells. Alternatively, the oligonucleotides may be provided to the wells at the same time as of after placing the cells in the well. Typically, the oligonucleotides are attached to surfaces inside the microwells prior to placing the cells in the wells. The oligonucleotides may be at least about 10 nucleotides long, at least about 15 nucleotides long; at least about 20 nucleotides long; at least about 25 nucleotides long; at least about 30 nucleotides long; at least about 40 nucleotides long; at least about 50 nucleotides long; at least about 60 nucleotides long; at least about 70 nucleotides long; at least about 80 nucleotides long; at least about 90 nucleotides long; typically, with less than 100 nucleotides complementary to mRNA coding for the biologically active agent. The oligonucleotides may contain one or more unique tags common to all identical oligonucleotides within the same microwell.

[0079] During the sequencing process, the microwell identifying tags are converted into digital tags in the form of a DNA sequence, which can then be used to associate nucleotide sequences with particular wells (and with one another, if more than one biologically active agent of interest is being sequenced from a well, e.g. a complex of proteins, e.g. the heavy and light chain of an antibody) and hence with particular addressed regions of the membrane. Thus, the information about the binding properties of the biologically active agent of interest can be correlated with the nucleotide sequence encoding the biologically active agent of interest.

[0080] As above, to address the regions on the solid surface such that they are in registration with the wells comprising the cells of interest, and to provide a biologically active agent whose binding to a binding agent is known and to which the experimental samples may be compared, it may be desirable to include marker cells with a known marked biologically active agent into the cell population under investigation, for example, a hybridoma cell line that is optionally capable of expressing an antibody with known sequence information for both heavy and light chain. These cells can be stained with an antigen labeled with a distinct fluorophore or genetically engineered to express a fluorescent marker so they can be identified in the microwells. Likewise, the antibody captured on the array can also be stained with the appropriate antigen with the distinct dye to identify its position on the solid surface. Finally, the marker cells' mRNA is processed and sequenced so the DNA sequence plus the digital tags will positively identify the position of the marker cells on the capture microarray slide. As such, the position of the marker cells on the microarray slide can be matched with those on the solid surface. The mRNA and protein of other cells in the array can be matched based on their position relative to the marker cells, their mRNA and antibodies. These marker cells can also be used to provide a definitive binding affinity to which the binding affinities of the plurality of biologically active agents can be compared.

[0081] In some aspects of the invention, only the nucleotide sequence encoding the at least one biologically active agent of interest is obtained. In these aspects, this data is employed to obtain metadata about the biologically active agent of interest. For example, based upon the nucleotide sequence, metadata on the binding properties of the biologically active agent(s) can then be obtained and the ideal biologically active agent predicted. Similarly, metadata on the frequency that the biologically active agent of interest is expressed by a cells in a population can be obtained, which can be used to predict the preference of an organism for that biologically active agent.

[0082] In some aspects of the invention, other cellular characteristics that may be affected by the production of a biologically active agent of interest by the cell or the presence of a provided biologically active agent may be measured. Such cellular characteristics may include, for example, cellular morphology, membrane potential, surface marker expression, secretion, or intracellular marker expression. Such characteristics may be measured by capturing computer images of cells deposited in the wells and analyzing these images for various parameters, for example, shape, color, opacity and size. These parameters are then associated with the binding affinities of the biologically active agents produced by the cells and/or the nucleotide sequences encoding these biologically active agents utilizing the location of each well.

[0083] The present invention provides an efficient method for analyzing a large number of cells individually rather than analyzing a population of cells and reporting the average measurement for the population. The method can be used to screen hundreds, thousands, tens of thousands, hundreds of thousands, millions or more cells at the same time. The method can be used to a) rank biologically active agents of interest with respect to affinity b) filter DNA sequence data thereby removing contaminating or irrelevant cells from consideration, c) classify the type of cell based on the quantity of biologically active agent produced or d) measure the frequency of cells producing similar biologically active agents of interest.

[0084] The present invention finds utility in the field of drug screening. In such a case, a plurality of biologically active agents to be screened for modulating a cellular signaling pathway, e.g. small molecule compounds, antibodies, or siRNA, is provided to a plurality of microwells in a multi-well device, wherein each microwell comprises a different biologically active agent. Substantially similar cells are provided to each well and allowed to produce a biologically active agent of interest, e.g. a protein, the expression of which will be predictive of efficacy of the biologically active agent being screened. The biologically active agent of interest that is produced by cells in the wells is contacted with a binding agent, e.g. a protein to which the biologically active agent is known to bind, which is, in turn, attached to a solid surface, e.g. the lid of the multi-well device. If the cells are producing the biologically active agent of interest, it is captured by the binding agent at an addressed location. The solid surface is then assayed for the presence of the biologically active agent by ELISA or other such binding assay. The information on the presence/absence of the biologically active agent is then associated to a particular addressed region of the solid surface, which is in turn associated with a well in the multi-well plate. Thus, a plurality of biologically active agents may be screened to determine their efficacy at modulating a cellular signaling pathway.

[0085] The present invention also finds utility in the field of network pharmacology. In such a case, one or more biologically active agents, e.g. a small molecule compound, antibody, or siRNA, is screened to determine its effect on a plurality of cellular signaling pathways. For example, a single biologically active agent is provided to a plurality of microwells in a multi-well device. Substantially similar cells are provided to each well and allowed to produce biologically active agents of interest, e.g. proteins, the production of which will be predictive of the impact of the biologically active agent on a plurality of signaling pathways. The biologically active agent of interest produced by the cells in the wells is contacted with a plurality of binding agent, e.g. proteins, to which the biologically active agents are known to bind, which are, in turn, spotted on a solid surface, e.g. the lid of the multi-well device, wherein each binding agent is spotted at a region on the solid surface that is associated with a well. If the cells are producing the biologically active agents of interest, the biologically active agents of interest are captured by the binding agent at an addressed location. The solid surface is then assayed for the presence/absence of the biologically active agents produced by the cell, for example, by using multi spectrum fluorescent detection on a microarray scanner to identify the different binding agents (e.g. with each labeled with a different colored fluorophore) and quantitating the amount of biologically active agent captured at each binding agent. The information on the presence/absence of the biologically active agents can then be used to determine the state of the cell and the level of activity of cellular signaling pathways. Thus, a single biologically active agent can be screened to determine its efficacy at modulating a plurality of cellular signaling pathways. Note that the effects of combinations of biologically active agents on networks of pathways can also be screened in this manner. In such cases, more than one biologically active agent may be provided in each well, or the agents may be provided individually to each well and in combination to a single well.

[0086] The present invention also finds utility in identifying protein variants of interest. For example, in the field of biomarker discovery, to identify novel single nucleotide polymorphisms (SNPs) that may be predictive of increased/decreased susceptibility to disease, or increased/decreased sensitivity to treatment. In such a case, a biologically active agent, e.g. a protein, and mutants therein have typically been linked to increased/decreased susceptibility to disease or increased/decreased sensitivity to treatment. In such a case, cells from different human subjects are placed in a plurality of wells, wherein the cells from different subjects are in different wells, and the cells are allowed to produce the protein of interest. The protein of interest in the wells is contacted with a binding agent, e.g. a protein to which the protein of interest is known to bind, which is, in turn, bound to a solid surface, e.g. the lid of the multi-well device. The protein of interest is captured by the binding agent at an addressed location, unless the protein of interest comprises a mutation, e.g. a SNP, which modulates the efficacy of binding of the protein of interest to the binding agent. The solid surface is then assayed for the protein of interest by ELISA or other such binding assay to determine the efficacy of binding of the protein of interest to the binding agent on the solid surface. The information on the binding efficacy of the protein of interest is then associated to a particular addressed region of the solid surface, which is in turn associated with a well in the multi-well plate.

[0087] Meanwhile, the nucleotide sequence of the protein of interest in each well is determined by preparing mRNA from the cells in the wells, synthesizing cDNA and sequencing the cDNA. The cDNAs are sequenced en masse, by providing unique oligonucleotides to each well which are incorporated into the cDNA as it is being synthesized, after which the cDNAs are pooled and sequenced by ultra-high throughput DNA sequencing technology to determine the nucleotide sequence of the at least one protein of interest in each originating cell. During the sequencing process, the microwell-identifying tags are converted into digital tags in the form of a DNA sequence, which can then be used to associate nucleotide sequences with particular wells (and with one another, if more than one protein of interest is being sequenced from a well, or the protein is comprised of multiple polypeptides, see for example below) and hence with particular addressed regions of the solid surface. Thus, a plurality of variants of a protein of interest can be screened simultaneously, and the information about the binding properties of these variants can be correlated with the nucleotide sequence encoding these naturally occurring variants. Correlation of this information with data on the subjects' susceptibility to disease and/or sensitivity to treatment can be used to identify novel predictive and prognostic biomarkers.

[0088] It should be noted that in some cases, the protein variants of interest may be comprised of a single polypeptide. In other cases the protein variants of interest are complexes of polypeptides produced from the same exact cell, i.e. cis complexes, wherein it is the complexing of the polypeptides that confers binding properties upon the protein, e.g. an antibody. In still other cases, the protein variants of interest may be complexes of proteins, i.e. trans complexes, wherein it is the complexing of proteins that confers binding properties upon one protein of the complex or even the complex overall. For example, many cell surface receptors responsible for signal transduction possess the ability to form complexes with other receptors via non-covalent bonds. The present invention can also be used to examine these protein complexes for their binding properties on a cell-by-cell basis, so as to identify qualitative changes such as splice variants that affect the formation of these complexes. For example, a cell-by-cell comparison between normal and cancer cells can be made. If the relevant proteins are captured, similar analysis with a labeled ligand of the receptors can be performed to understand the binding characteristics of these complexes on a per cell basis as well as to quantify the receptors and to identify different combinations on a per cell basis. All these pieces of information can be correlated with the observed characteristics of the cell mentioned earlier, including the binding characteristics of the biologically active agent produced by the cell and the nucleotide sequence(s) encoding that biologically active agent. A high content analysis of the cell and its signal transduction machinery can be performed.

[0089] One example of a field wherein the biologically active agent of interest is a complex of polypeptides is the field of antibody discovery. For example, the present invention may be used to screen cells to identify those producing antibodies that have a higher affinity or avidity for an antigen and to determine the nucleotide sequence encoding the binding domains of these antibodies. In such a case, B cells, e.g. plasma cells, are placed in a plurality of wells, wherein each plasma cell is placed in a different well, and the cells are allowed to produce antibodies. The antibodies produced by the cells in the wells are contacted with a binding agent, e.g. Protein G, Protein A, Protein L, Protein A/G, which is bound to a solid surface, e.g. the lid of a multi-well device, such that the antibodies are universally and specifically bound. The captured antibodies are then assayed by any of a number of known binding assays, e.g. ELISA, to determine their affinity and/or avidity for their target protein, i.e. antigen. The information on the binding efficacy of the antibody is then associated to a particular addressed region of the solid surface, which is in turn associated with a well in the multi-well plate.

[0090] Meanwhile, the nucleotide sequences of the heavy and light chains of the antibodies are determined by preparing RNA from the cells in the wells, synthesizing cDNA and sequencing the cDNA. For sequencing en mass, oligonucleotides that are unique to each well are provided to the wells which are incorporated into the cDNA as it is being synthesized, after which the cDNAs are pooled and sequenced by ultra-high throughput DNA sequencing technology to determine the nucleotide sequence of the antibody produced by each plasma cell. During the sequencing process, the microwell identifying tags are converted into digital tags in the form of a DNA sequence, which is used to associate nucleotide sequences of the heavy and light chains with particular wells and with one another and hence with particular addressed regions of the membrane. Thus, the information about the binding properties of antibodies produced by particular cells in particular wells can be correlated with the nucleotide sequence encoding those antibodies. Such information can be used to engineer antibodies with better affinities/avidities for antigens.

[0091] Alternatively, the present invention may be used to obtain the nucleotide sequences of the heavy and light chains independent of using the solid surface element to assay antibody binding affinites, and the frequency of occurrence for a given antibody sequence used to estimate the K_D of that antibody due to the phenomenon of clonal selection during an immune response. This estimated K_D can be correlated with the measured value obtained from captured antibodies. This line of analysis can aid in understanding the process of affinity maturation in vivo and ultimately guide a focused and more efficient effort to increase the affinity of an antibody in vitro.

[0092] With a large number of sequences available for a combination of heavy and light chains of a given antibody, it becomes possible to use the positions of the mutation and the measured equilibrium constant to infer the critical amino acid residues of the paratope on the antibody. The present invention is also useful for these studies, as it provides for acquiring nucleotide sequence for proteins encoded by single cells in a large population.

[0093] With the critical amino acid residues identified, it also becomes possible to infer the epitopes and the structure around the putative epitopes on the antigen. The present invention estimates the free binding energy of the antibody and the putative protein folding. With the potential epitope identified for each antibody, this provides another criterion to classify the antibody obtained for a target antigen. The ability to quickly identify sets of antibodies each against different epitopes on an antigen is an advantage of the invention. This reduces the number of confirmatory tests required given the large number of different antibodies recovered. Deciphering the structures local to the identified epitopes on the antigen can lead to a good understanding of the overall structure of the antigen when the density of the epitope identified becomes high enough. This outcome is also unprecedented. These are all due to the fact that a large number of antibodies can be rapidly recovered using the invention.

[0094] Though the aim to obtain antibodies with the highest K_D and the lowest k_off is in line with pharmaceutical development, the present invention can be used to follow an immune response by taking antibody producing cells such as plasma cells and/or antigen specific B cells from any day after immunization. Furthermore, the present invention can be used to monitor the state of the immune system by profiling the B cell population without selection from mouse spleen or even those B cells in circulation for a human subject. This could be very useful in studying autoimmune diseases and to discover biomarkers or potential drug targets. Certainly, the T cell receptors having a structure similar to antibodies can be analyzed in a similar fashion for studying immune functions. Consequently, the present invention has many potential uses but a major application is likely to be the analysis of gene combinations that are polymorphic within a population of cells, such as the rearranged genes for Ig or T cell receptors or secreted antibodies.

[0095] Similar to phage display technology, the present invention can be used to screen constructs in many display technologies, such as yeast display, bacterial display, and mammalian display. The candidate display clones can be enriched from the library analogous to the manner for selecting antigen specific B cells by flow cytometry. Once enriched, these display clones can be separated into microwells just like antigen specific B cells and processed in a very similar way. A change in the capture sequence has to be made appropriate for the display technology used. However, in contrast to many of these display technologies, candidates can be processed much more rapidly and can characterize their binding properties en masse if the proteins expressed by the display clones are captured as described under Embodiment 1 or Embodiment 2, below.

[0096] It should also be noted that the present invention may also be used to examine other cellular characteristics that may be affected by the production of protein variants by a cell, for example, cellular morphology, membrane potential, surface marker expression, secretion, or intracellular marker expression. Thus, one may record these cellular characteristics on a per cell basis, and then correlate these characteristics with the binding characteristics of the biologically active agent produced by the cell and/or the nucleotide sequence(s) encoding that biologically active agent. As one example, the morphology of individual cells may be used as a measure of the cell's tumorigenic capacity, and those results correlated with nucleic acid sequencing to assess protein variants or gene copy number so as to confirm the diagnosis and reduce false positives. When fluorescent measurements are taken, of cells or surface bound antibodies, a microarray reader produces an image wherein the intensity of various pixels must be associated with the concentration of cell surface markers, or surface bound antibodies, in specific wells. To insure proper registration between the confined cell, the captured mRNA, and optionally the captured protein, it again may be desirable to include marker cells with a known cis protein complex into the cell population under investigation.

[0097] In further describing aspects of the invention, the following description focuses on screening single cells for the expression of antibodies that bind to a particular antigen with a desired affinity, and/or obtaining the nucleotide sequences that encode the heavy and light chains of those antibodies. However, the subject methods and devices thereof also find use in the screening of single cells for other biologically active agents that can be used to acquire information on the status and activity of those cells.

[0098] For example, a plasmocyte 102 is suspended inside microwell 101 in a buffered solution that fills the entire microwell. Microwell 101 may have dimensions 50 microns by 50 microns by 50 microns. Dimensions 105 measure 0 to 50 microns from the bottom of microwell 101 to the top. From one side to the opposite side the dimensions range from −25 microns to 25 microns. DNA oligonucleotides are attached to the bottom surface of microwell 101 in contiguous areas called pads. Pad 103 contains an oligonucleotide complementary to an invariant sequence in the mRNA of plasmocyte 102 that codes for the heavy chain of an antibody. Pad 104 contains an oligonucleotide complementary to an invariant sequence in the mRNA of plasmocyte 102 that codes for the light chain of an antibody. Said invariant sequences are selected so they will bind strongly and physically capture complementary mRNA when said mRNA is released from plasmocyte 102 and diffuses into the solution which may be a cell culture solution in microwell 101. Whereas microwell 101 will function properly in any orientation, for the purposes of illustration in FIG. 1 pad 103 and pad 104 are shown attached to the bottom of microwell 101.

[0099] In the following particular embodiments, the biologically active agents of interest that are being produced by cells are antibodies produced by cells in a host as an immune response to an antigen. In Embodiment 1 and Embodiment 2, the nucleotide sequence encoding the heavy and light chain for each antibody produced by each plasma cell is determined, and the antibodies produced by each plasma cell are captured to form an addressed array in registration with the microwells where the original plasma cells are each confined. A preparation of fluorescently labeled antigen is used at varying concentrations to interrogate antibodies captured at each spot such that both the dissociation constant (k_off) and equilibrium constant (K_D) are measured. The association constants are derived from both k_off and K_D. Such kinetic binding properties are extremely valuable in ranking the antibodies obtained using the present invention since a large number of antibodies are recovered and a means of classifying them is desirable. Furthermore, in the event of an inadequate enrichment for antigen-specific B cells using flow cytometry, these measurements serve to filter out data deriving from contaminating non B cells or non antigen-specific B cells.

[0100] In Embodiment 3 and Embodiment 4, this biological system is used to illustrate how the present invention can be used to determine the nucleotide sequence encoding two constituent macromolecules of a cis protein complex where the two copies of each constituent, heavy and light chains, are held together by multiple covalent bonds. In this case, there is a polymorphic difference in the heavy and light chains of each antibody in each plasma cell generated during the immune response. By practicing the present invention, the combination of the antibody's heavy and light chain primary structure can be determined for each plasma cell analyzed. Since such a combination ensures exquisite specificity of the antibody, knowing the primary structure, that is, amino acid sequence, allows one to genetically engineer the antibody by synthesizing an appropriate construct based on the knowledge of primary structure of the heavy and light chain combination of an antibody and by importing such a construct into an appropriate expression system for unlimited supply of the antibody, which finds diverse applications in a variety of fields including research, diagnostics and therapeutics.

[0101] Though the discussion focuses on mouse antibodies, it is clear that the present invention can be applied to antibodies produced by animals with the ability to generate a humoral immune response. In an aspect of the invention a prerequisite for practicing the present invention is the knowledge of the sequence of the constant region for the heavy and light chain mRNA of an antibody for a given isotype. Since the present invention obtains mRNA and proteins directly from antibody producing cells such as B cells, it clearly obviates the need for cell fusion to generate hybridoma cells.

[0102] Although only the mouse IgG1, IgG2a, and IgG2b isotype heavy chain conserved sequence and the κ light chain sequence are used for capturing the mouse antibody mRNA in the example illustrated, this represents an embodiment rather than a limitation of the invention. It is entirely within the scope of the present invention to use spotted microarrays wherein a mixture of oligonucleotide probes comprising appropriate sequences can be attached and used to capture all isotypes.

[0103] In one of the embodiments, the protein complex under study can be the heavy chain and light chain of an antibody. In this case, integrating structural information with kinetic properties provides insights into the paratopes on the antibody as well as the epitopes on the antigen. To illustrate this utility of the present invention, simulation with antibodies against the hapten 2-phenyl-5-oxazoIone (phOX) was performed.

Embodiment 1

[0104] Microwells may be produced by attaching 50 micron (±25%) high walls onto a flat support and binding oligonucleotides pads inside the wells. Cells are distributed over these microwells in such a manner that it allows a single cell to settle into a microwell. A cover that is coated with a material suitable for attaching proteins of interest is placed over said microwells. For example, to immobilize antibodies on the top cover the cover is coated with an antibody specific capture agent such as Protein A or Protein L (Sigma-Aldrich, St. Louis, Mo.).

Embodiment 2

[0105] In Embodiment 2, a material suitable for attaching proteins of interest is coated on a commercial microwell array, e.g.: a blank PicoTiterPlate device (454 Life Sciences, Branford, Conn.) containing 50 micron hexagonal wells. Cells are distributed over said hexagonal wells in such a manner that allows a single cell to settle into a well. For example, to immobilize antibodies on the surfaces of the wells one coats the surfaces with an antibody specific compound such as Protein A. A custom oligonucleotide array (NimbleGen Systems, Inc., Madison, Wis.) is placed over said PicoTiterPlate to capture and tag mRNAs.

Embodiment 3

[0106] In Embodiment 3, we fabricate microwells by attaching approximately 50 micron walls onto a flat support. Cells are distributed over said microwells in such a manner that allows a single cell to settle into a microwell. A custom oligonucleotide array (NimbleGen Systems, Inc., Madison, Wis.) is placed over said fabricated microwells in order to capture and tag mRNAs. No proteins are captured.

Embodiment 4

[0107] In Embodiment 4, a commercial microwell array, e.g.: a blank PicoTiterPlate device (454 Life Sciences, Branford, Conn.) containing 50 micron hexagonal wells, is used to contain the distributed cells. A custom oligonucleotide array (NimbleGen Systems, Inc., Madison, Wis.) is placed over said commercial microwell array to capture and tag mRNAs. No proteins are captured.

Embodiment 5

[0108] In Embodiment 5, two planar surfaces are separated a small distance (e.g.: 50 microns) by a porous structure. One or both planar surfaces are microarrays. The porous structure contains holes large enough to contain a biological cell (e.g.: 50 microns). The porous structure is temporarily affixed to one planar surface, thereby constructing an array of open microwells. Biological cells are dispersed over the top of said microwells in one of two methods: a) stochastic separation, wherein a concentration of said cells is chosen so that only a single cell usually occupies a single microwell when the cells are randomly dispersed above said microwells; and b) deterministic separation, wherein individual cells are caused to move to predetermined microwells. After a period of time (e.g.: 3 minutes) cells above microwells are allowed to enter said microwells due to their increased density. Reagents common to all microwells are introduced to the region above said microwells and reactions occur within each microwell. Small molecules are allowed to diffuse out of said microwells in a short amount of time (e.g.: 10 seconds). Then the second planar surface is placed over said microwells sealing them and allowing reactions inside the microwells to occur over a longer period of time. Said reactions result in a modification of the microarray(s) by chemicals confined to said microwells. After said microarray(s) are deemed modified, the planar arrays are separated and the porous structure removed. Codes fabricated into said modified microarray(s) are used during analysis to associate microwells with measured results. Said codes may be the physical position on a microarray, or a tag caused to be embedded in the analyzed data. For example, in the case of an oligonucleotide array, said embedded tag may be a unique DNA sequence.

The Multi-Well Device

[0109] Central to the subject methods of the present invention is the confinement of single cells into individual microwells. Accordingly, one aspect of the present invention is a multi-well device comprised of microwells. The microwells serve two purposes: first, to inhibit the diffusion of reactants and products among individual cells; second, to accelerate rates for processes such as nucleic acid hybridization or protein interactions. Accordingly, the microwells of the multi-well device of the present invention are interconnected at a density greater than 100 microwells per cm², and each microwell has a volume between 1 picoliter and 500 nanoliters. The tray may include any number of wells and preferably includes 96 or more wells, 100, 1,000, 10,000, 100,000, 1,000,000 wells or more. Typically, the microwells occupy a density greater than 100 wells per cm². The wells are designed so that they are capable of holding a single cell and the tray surface is designed such that when the cells are placed on and moved about the surface of the tray a single cell is moved into each of the wells. Additionally, the tray may be designed such that the microwells are intermittently accessible to external reagents.

[0110] Also central to some embodiments of the present method is the screening of cells for biologically active agents of interest. Accordingly, the multi-well device of the present invention may further comprise a solid surface to which is attached agents that bind to the biologically active agent of interest, e.g. binding agents. This solid surface may be a membrane, filament, bead, or polystyrene or polypropylene sheet, e.g. a tray top comprised of a plurality of areas wherein each of the area corresponds uniquely with each well on the tray. The areas may be planar or protrude outward to provide more surface area. The areas (and optional protuberances) on the tray top are coated with an antibody binding agent, e.g. protein A, protein G, protein L, protein A/G, goat anti-mouse IgG antibodies, or a target binding protein for the at least one biologically active agent of interest that is being produced by the cells. The coating can include any agent which efficiently and selectively binds to the biologically active agents produced by the cells in the wells.

[0111] In some embodiments, biologically active agents may be provided to the microwells. For example, small molecule compounds, proteins, siRNA or oligonucleotides may be spotted in the microwells. Each of the wells may also or alternatively have coated thereon a polynucleotide, e.g. an oligonucleotide. The polynucleotide may act as a primer to replicate nucleotides in the well after the cells are lysed, for example, in a sequencing reaction.

[0112] The device may be in a kit that includes instructions to use the device in accordance with any method described here. Further, the device may be sold with reagents including PCR reagents.

[0113] The following examples are put forth so as to provide those of ordinary skill in the art with a complete disclosure and description of how to make and use the present invention, and are not intended to limit the scope of what the inventors regard as their invention nor are they intended to represent that the experiments below are all or the only experiments performed. Efforts have been made to ensure accuracy with respect to numbers used (e.g. amounts, temperature, etc.) but some experimental errors and deviations should be accounted for. Unless indicated otherwise, parts are parts by weight, molecular weight is weight average molecular weight, temperature is in degrees Centigrade, and pressure is at or near atmospheric.

[0114] Immunization of Mice

[0115] Referring to FIG. 2, a group of BALB/c mice 201 are immunized intraperitoneally with 50 μg of hen egg lysozyme (HEL, Sigma, St. Louis, Mo.) in complete Freund's adjuvant (Sigma). Six weeks after the first immunization, the mice are immunized intravenously with the same dose of HEL. Two days after the second inoculation, blood is taken from the tail vein of the mice and the presence of polyclonal antibodies against HEL in the sera is determined by an ELISA test. If the titer determined by said ELISA test is satisfactory, the mice are sacrificed the next day and their spleens harvested.

[0116] Isolation of Antigen Specific B Cells

[0117] Cell suspensions of splenocytes are obtained by mechanical disruption of the spleens, and erythrocytes are lysed in 0.16 M chloride ammonium solution at pH 7.4. The resulting single-cell suspensions are stained with a combination of HEL labeled with FITC (Pierce Biotechnology, Inc., Rockford, Ill.) and HEL labeled with PE-Cy5 (Prozyme, San Leandro, Calif.) at optimal dilution. Staining is conducted at 4° C. in 2 steps. First, cells are stained with the FITC-labeled HEL for one hour, using the empirically determined optimal sub-saturating concentration. Cells are then washed three times with FACS buffer. Next, cells are stained with the PE-Cy5 labeled HEL. After staining, cells are washed twice in PBS with 5% FCS.

[0118] Cells are analyzed using a modified dual laser FAC-Star^PLUS™ (Becton Dickinson). Files are acquired using CELLQuest software. In step 202, single B cells with appropriate surface phenotype are sorted for repertoire analysis using the automatic cell-dispensing unit attached to the FAC-Star^PLUS™ and Clone-Cyt software (Becton Dickinson). Cells are sorted into an appropriate volume of PBS supplemented with an RNase inhibitor, SUPERase-In, at a concentration of 1 U/μL. Alternatively, antigen-specific B cells can be isolated using magnetic-assisted cell sorting (MACS) using reagents identifying appropriate cell surface markers. In addition, antigen-specific B cells can be isolated by agarose gel microdrops using a CellSys 101 microdrop maker (One Cell Systems, Inc., Cambridge, Mass., USA) or microdroplets resulting from water-in-oil emulsion. Alternatively, a combination of the methods mentioned thereof can be used to isolate antigen-specific B cells.

[0119] Cell Separation Using PicoTiterPlate (Steps 203)

[0120] A PicoTiterPlate (PTP) device (454 Life Sciences, Branford, Conn.) is prepared and the surface coated with Protein L (Pierce Biotechnology, Inc., Rockford, Ill.) in 0.1 M sodium acetate overnight. The surface is blocked with 3% BSA in PBS for 1 hour. After removing 3% BSA in PBS, the surface is washed 4 times with 0.1% Tween-20 in PBS. The PTP device is then filled with PBS supplemented with SUPERase-In. Air bubbles that might remain in the wells are dislodged by centrifugation, once the PTP device is in the Bead Deposition Device (454 Life Sciences, Branford, Conn.). Using a 2-mL pipette, cells are gently placed on top of the PTP device and let sit for 10 minutes to allow individual cell to fall into a well. Cells that do not go into a well are moved off the PTP device with a coverslip. Cells are lysed with a customary lysing solution including a chaotropic agent such as guanidinium thiocyanate or a detergent such as NP-40 (step 204). If capturing of the protein of interest is desired, the detergent NP-40 is to be used. A solid surface coated with a capture agent such as protein A or protein L is used to cover the microwells to confine diffusion of macromolecules released from the lysed single cell and to simultaneously capture said proteins of interest such as antibodies on the solid surface forming a protein array (step 211).

[0121] Fabrication of Patterned Photoresistant Structures and Membranes.

[0122] Arrays of cylindrical posts of photoresist are fabricated on silicon wafers using standard photolithographic techniques and rigid chrome masks. The arrays of square features measuring 50 μm×50 μm×50 μm with the inter-feature spacing of 50 μm are fabricated using transparencies as photomasks.

[0123] Fabrication of Elastomeric PDMS Membranes.

[0124] The PDMS prepolymer (mix in a 10:1 ratio with a crosslinking catalyst) is spin-coated on the bas-relief of patterned photoresist features mentioned above at 3,000 rpm for 60 second to generate a film that is ca. 45 μm thick. As such, the resulting PDMS membrane possesses square holes defined by the dimension of the photoresist features. The PDMS membranes are cured for 2 h at 60° C. A thicker layer of PDMS prepolymer is added to the edges of the membranes in dropwise fashion. The membranes are kept at 60° C. overnight. The membranes are removed from their supports using tweezers and are cut to the desired sizes along the edges of the support.

[0125] Coating the Holed Membranes with BSA to Minimize Non-Specific Binding by Proteins

[0126] The PDMS membranes are placed on the surface of a glass slide with a few drops of ethanol. Drops of a buffered solution of BSA (1% w/v, in PBS) are placed on the membrane to cover the holes. Since the fluid does not readily fill the hydrophobic pores, a vacuum (ca. 500 mTorr) is applied for 30 seconds and released twice to extract the air trapped in the holes and the BSA is allowed to adsorb to the surfaces for 15 min. The PDMS membranes on the glass slides are then rinsed three times with PBS. The membranes are peeled from the glass slide in the presence of PBS and transferred to a protein array slide (Full Moon BioSystems, Sunnyvale, Calif.) covered with PBS to help seal the membrane onto the slide.

[0127] A Glass Slide Coated with Protein L to Support a Holed PDMS Membrane.

[0128] SuperEpoxy 2 glass slides (TeleChem International, Inc., Sunnyvale, Calif.) are coated with Protein L (Pierce Biotechnology, Inc., Rockford, Ill.). A prepared holed PDMS membrane is positioned on top of the glass slide and is pressed to make uniform contact with the surface. The entire setup is blocked again in a buffered solution of BSA. Again, brief vacuum is applied to remove air trapped in the holes. Blocking is performed for 1 hour, followed by washing three times with PBS. The glass slide coated with protein L supporting a holed PDMS membrane is now ready to capture antibodies from B cells and holes on the PDMS membrane are representative of microwells. In addition to Protein L, it is possible to use Protein G, Protein A/G, goat anti-mouse IgG antibodies, etc.

[0129] General Features of the Probes on NimbleGen Hd-2 Microarray Chips

[0130] The NimbleGen HD-2 microarray chips have 2.1 million probes on a single slide with the array size of 62 mm×14 mm. As such, there is a potential to have 2.1 million unique tags. To facilitate capture, tagging, and subsequent DNA manipulation, general features of a probe for both heavy chain and light chain are utilized wherein an example of said general features for a heavy chain capture probe are: Xho1 cleavage, Primer B by 454 Life Sciences, a 2-base Spacer (always constant) to identify the start of the well identifier tag, a 14 mer well identifier, a mouse gamma heavy chain CH1 IgG1, IgG2a, and an IgG2b isotype heavy chain conserved sequence capture probe. Furthermore, an example of said general features for a light chain capture probe are an Xho1 cleavage site, Primer A by 454 Life Sciences 407, a 2-base spacer (always constant) to identify the start of the well identifier tag, a 14 mer well identifier tag, and a mouse kappa light chain constant region capture probe. Note that the length of the microwell identifier tag is a 14-mer, which has a possible sequence combination of over 268 million, far in excess of what is need for 2.1 million probes. This cushion allows elimination of undesirable sequences such as consecutive triples of the same base and perfect self-complementarity or complementarity between the microwell identifier and heavy chain probe sequence, light chain probe sequence, 454 primer A, or 454 primer. In addition, it is desirable to keep the G+C content between 40-60%. If necessary, a parity code, such as one used in Hamming codes, can be embedded in the well identifier for error-correction purposes. Examples of probes for immunoglobulin heavy chain (SEQ ID NO:1) and light chain (SEQ ID NO:2) polypeptides may be found in FIG. 4.

[0131] The process of capturing released mRNA from lysed cells and conversion into a tagged ds cDNA is illustrated in FIG. 3. A target molecule of mRNA 301 hybridizes to an oligonucleotide probe 203, which is attached to solid surface 306 as in a typical microarray slide comprising non-DNA linker 305, PCR/sequencing primer B of 454 Life Sciences 304, a well identifier tag 303, a capture oligonucleotide 302 complementary to a region in the target mRNA from either the heavy or light chain gene. Reverse transcription 307 uses a reverse transcriptase (Stratagene, La Jolla, Calif.) to synthesize the first strand cDNA 308 by extending the capture oligonucleotide 203 as a primer and copying an appropriate region of the target mRNA template thereby incorporating 5′-methyl dCTP 309. Second strand cDNA reaction 310 replaces the target mRNA by a combined action of RNaseH and E. coli DNA polymerase I, leaving behind a small segment of the original mRNA template at the 5′ end. Reaction 313 using T4 DNA polymerase blunts the 5′ end of the ds cDNA 314. In a ligation reaction 315, an adaptor molecule 316 comprising PCR/sequencing primer A of 454 Life Sciences with a 3′ protruding end to insure orientation specific ligation is attached. In process 318, excess adaptor molecules are removed and in process 319, the attached ds cDNA is cleaved off the solid surface by endonuclease XhoI (New England Biolabs, Ipswich, Mass.), releasing the ds cDNA molecule 320 with the well identifier tag flanked by 454 Life Sciences's PCR/sequencing primers A and B ready to be plugged into the 454 sequencing workflow.

[0132] mRNA and Antibody Capture (Step 205)

[0133] An equal volume of 1% NP-40 is pipetted from one side of the PTP device to make the final NP-40 concentration at 0.5%. The PTP device is incubated for 10 seconds. The top of the PTP device is then covered with the NimbleGen HD-2 microarray chip, which is securely clamped down using an appropriate device. Incubate at 37° C. for 30 minutes. The entire setup is then soaked in RNase free PBS held in a tray to separate the microarray from the PTP device. The PTP device is washed extensively with PBS for 3 times. PTP device now having the antibodies captured from single B cells can be stored at 4° C. The microarray was soaked in PBS held in a tray at 60° C. for 10 minutes to remove non-specifically hybridized RNA species, then extensively washed with PBS for 3 times. The microarray is now ready for downstream manipulation.

[0134] cDNA Synthesis and Adaptor Attachment (Step 206 in FIG. 2)

[0135] The microarray is briefly blotted dry on a piece of kimwipe and 30 μL of reverse transcriptase (RTase) reaction solution containing SuperScriptIII RTase (Invitrogen, Carlsbad, Calif.), appropriate buffer components, and dNTPs except 5′-methy dCTP is immediately added. A cover slip is added and the microarray is placed at 50° C. for 1 hour in a humid environment. The cover slip is removed by soaking in DNA polymerase buffer and wash 3 times. The chip is again blotted dry on a piece of kimwipe. Thirty μL of appropriate buffer containing E. coli DNA polymerase and RNaseH is added, and the samples incubated at 16° C. for 2 hours. The chip is washed and the synthesized double-stranded (ds) cDNA is polished by T4 DNA polymerase for 30 minutes at 37° C. Pre-annealed adaptors with a biotinylated end are added and ligated to blunt-ended cDNA on the chip by T4 DNA ligase at room temperature for 1 hour in an appropriate buffer. Excess adaptors are then removed by extensive washing in 10 mM Tris-HCl (pH 7.5), 0.1 mM EDTA. The slide with the cDNA attached can be stored at 4° C.

[0136] Marker Cells for Registration Between the DNA Pads on the Microarray Slide and the Protein Spots on the Protein Array

[0137] The hybridoma line TIB-228 (American Type Culture Collection, Manassas, Va.) produces an antibody against human CD14 with the isotype IgG2b. The DNA sequence has been determined for its heavy and light chains (GenBank accession number for heavy chain variable region and light chain variable region is AY669065 and AY669066, respectively). Recombinant human CD14 can be obtained commercially (Abnova, Taipei, Taiwan). When practicing the present invention, an appropriate number of cells, in the range of 10 to 50, can be mixed with the enriched antigen-specific B prior to separating them into microwells. CD14 can be labeled with PE-Cy5 to be distinguished from anti-HEL antibodies, which will be stained with FITC-labeled HEL.

[0138] DNA Sequencing by 454 Life Sciences's Technology

[0139] The double stranded (ds) cDNA attached to the NimbleGen chip is cleaved off (step 207) by incubating at 37° C. for 15 minutes with XhoI endonucleases (New England Biolabs, Ipswich, Mass.). The released ds cDNA can be optionally captured by streptavidin beads (MyOne beads, Invitrogen, Carlsbad, Calif.) and stored. Such a strategy minimizes DNA loss during storage. Afterwards, the complementary strand can be melted off the beads and quantified using real time PCR by a 454 primer A and light chain primer with an appropriate probe located within the light chain region Similar reagents for real time PCR can be designed for the heavy chain ds cDNA.

[0140] Once the ds cDNA is quantified, an appropriate amount is retrieved and mixed with appropriate amount of beads for subsequent emulsion PCR in step 208. The beads are processed and sequenced according to the vendor's recommendation (step 209). Including the spacer, the tag, the constant region, and the entire variable region, both the heavy and light chains require about 415 bases of sequence, which is within the limit of the 454 sequencing technology. Sequencing reads in one orientation are sufficient to obtain unambiguous sequence information to reconstruct the heavy and light chains for an antibody. This is because the genomic sequence of the prototype variable sequence is known. Furthermore, it is likely that FR1 are mostly unmutated in the antibody sequence during immune response based on other reported sequences. Finally, the redundancy of the sequence reads per antibody is high. Thus, based on the vendor's specification of 1 million reads per run, it is possible that one can analyze up to 200,000 B cells minus the few marker cells with 5 fold coverage for each B cell. Alternatively, cDNA originating from specific wells identified by binding kinetics measurement can be directly amplified by incorporating a tag specific sequence at the 3-end of the PCR primer.

[0141] After a sequencing run, the following steps are performed to analyze the sequence (steps 210 and 213):

• [0142] a) A list of sequences from a 454 run is obtained;
• [0143] b) The sequences are computationally evaluated and the tag, spacer, and constant region for each sequence identified, thereby determining whether each sequence is an H or L chain sequence. In addition, bases toward the 5′ end of the heavy chain constant region are sequenced and the minute variation therein used to determined the isotype of the captured antibody;
• [0144] c) The constant region sequence and spacer are masked;
• [0145] d) The marker cell H and L chain sequence are identified;
• [0146] e) The tags corresponding to the marker cell H and L chain sequences are identified and examined to determine if they are in close proximity (i.e.: less than a microwell width) on the microarray. If affirmative, these tags identify a microwell. Locations of multiple microwells containing marker cells will establish the physical layout of the other microwells that each may contain a B cell;
• [0147] f) Based on the layout, a tag group for each expected microwell is generated;
• [0148] g) Sequences with identical H chain or L chain are identified;
• [0149] h) The corresponding tag for the identical H chain or L chain sequences are computationally evaluated. If the tags are identical, then the origin for that H chain or L chain is confirmed;
• [0150] i) Using the tags from confirmed H chains and confirmed L chains, the pairs of H tag and L tag are reviewed to see if they fit into expected tag groups. If affirmative, the combination is confirmed;
• [0151] j) Multiple H chain sequences or L chain sequences are aligned in a confirmed combination to obtain a consensus sequence. The consensus will be matched with the prototype genome sequences of the appropriate immunoglobuliln v genes;
• [0152] k) All CDRs and FRs on the paired H chain and L chain sequences are identified. In addition, identify mutated residues compared with the prototype;
• [0153] l) Confirmed combinations are clustered and the frequency for identical combinations tallied to determine if they represent the outcome of a clonal selection event; and
• [0154] m) The DNA sequence is translated into amino acid sequence.

[0155] The construction of approximately 50 micron chemically inert structures is common knowledge to those skilled in the art. They are used in both Embodiment 1 and Embodiment 3, above, of this invention. Briefly, a 34 mm coverslip (Fisher Scientific, Pittsburgh, Pa.) is washed in acetone, isopropanol, methanol, deionized (18 Mohm) water and dried under a stream of nitrogen. It is further cleaned by exposure to at 120 watt oxygen plasma with a base pressure of 80 mTorr and oxygen pressure of 120 mTorr for 10 minutes (Technics 500 II Asher).

[0156] The clean coverslip is carefully centered on the vacuum chuck of a spin coater (Laurell Technologies, North Wales, Pa.) and SU-8 50 (Microchem, Newton, Mass.) negative photoresist is statically dispensed on the center. The coverslip rotates at 500 rpm for 10 seconds and ramps up to 2,000 rpm over the next 30 seconds. The coverslip is soft baked for 20 minutes at 95° C. and cools at room temperature for 10 minutes.

[0157] An emulsion mask is designed using Adobe Photoshop and printed on a Linotronic-Hercules 3300 dpi printer (7.4 micron spot size). The mask is placed in contact with the SU-8 film, covered with a quartz slide to weight it down, and exposed in a Kasper 2001 Contact Mask Aligner for 1 minute at 365 nm and at an intensity of approximately 200 mJ/cm². The exposed photoresist is baked for 2 hours at 95° C. on a level hotplate, soaked in PM Acetate for 30 minutes, and then cured at 200° C. for 1 hour.

[0158] The coated coverslip is soaked in sterile, distilled water overnight and sterilized in 70% ethanol in DI under a UV germicidal lamp for 1 hour. The cover slip is rinsed three times using sterile water and stored for at least 4 hours wet and ready for use. The 4 hour storage time ensures no gas bubbles remain in the wells.

[0159] The dimension of 50 microns is chosen so the microwells comfortably contain a single 10-15 micron plasmocyte. Since the Reynold's number is 56 times the fluid velocity in meters/sec, there is no turbulent flow inside the wells. Therefore, once the wells are filled with fluid no gas bubbles enter them.

[0160] This example refers to Embodiment 3 and Embodiment 4, above. We calculated results from a molecular simulation computing the diffusion of guanidinium thiocyanate (GT) into microwells and back out again. Simulated GT molecules diffused toward and eventually dissolved the cell membrane of a B cells. We computed 20 images of a typical microwell, each computed at a different time. The difference between the first 10 images was 0.1 second; the difference between the last 11 images was 1.0 second.

[0161] At time t=0 seconds, 2000 GT molecules were quickly introduced into an equal volume immediately directly above each microwell. This represents a very low concentration of GT (26 pM) and is much lower than the GT concentration needed to lyse cells (1 M). Since only 2000 molecules were visualized and used to measure diffusion, we imagined only a small fraction of the 1 M GT as labeled with a reporter and only those reporter molecules were used in our calculations. As time progressed, 400 GT molecules diffused into microwells until t=1.0 second. Then the number of GT molecules above the microwell was quickly reduced and maintained at zero. Said molecules inside the microwells diffused out of the microwells, and after 10 more seconds the concentration fell to 13 molecules.

[0162] In our simulation, between 0.5 and 0.8 seconds the cell membrane of plasmocytes substantially dissolved, as did the membrane surrounding the Golgi bodies. Since said mRNA was substantially larger than GT it diffused much more slowly.

[0163] The number of GT molecules used in this simulation can scale to any convenient concentration. If we normalize to the concentration loaded above physical microwells, 21% will diffuse into microwells in 1.0 second, and 10 seconds after flushing the concentration above the microwells, the concentration inside the wells will fall to 0.65%. At time equal 11 seconds all mRNA is still completely contained within the wells. GT at this low concentration does not interfere with the hybridization of RNA to DNA, or DNA to DNA. The microwells are sealed after 11 seconds so that all mRNA continues to remain inside the microwells.

[0164] The diffusivity was calculated from the Stokes-Einstein relation:

[0165] $D=\frac{k_{B}T}{6\pi \eta r}$

[0166] where D is the diffusivity, kB is Boltzman's constant (1.38×10-23 J/K), T is room temperature, η is the viscosity of the buffer (8.94×10-4 Pa/sec), and r is the radius of gyration of GT (5.0 Angstrom) or mRNA (60 Angstrom), respectively. Trajectories were recomputed every Δt=3 milliseconds adding a vector of length √{square root over (6DΔt)} in a random direction chosen from a 482 tessellation of a sphere.

[0167] When the cis- and/or trans-protein complexes are antibodies, our invention provides information about a much larger number of them than has previously been available. By associating the sequenced tags with the original microwells, we identify a population of mRNAs produced by B cells as well as their respective heavy and light chain sequence. Said sequences can infer the somatic mutations inside each B cells.

[0168] We simulated a population of B cells from the published sequences of 21 light and heavy chain antibodies. The sequences were aligned according to recommendations at Web Antibody Modeling. Each sequence was stored with its dissociation constant as measured by Milstein. We also calculated a copy number which represents the number of cells containing a specific mRNA with sequence.

[0169] We examined the variation in amino acids along the length of the light chain the heavy chain independently. At every amino acid location, we computed the probability of an amino acid being different from the most frequent amino acid. If at position i we count f most frequent amino acids out of n, then the probability is computed as:

[0170] $p_i=1-\frac{f}{n}.$
The most variable light and heavy chain amino acid are the most likely locations of the paratope. The paratope can be inferred because we can align a population of sequences.

[0171] Our invention can also measure the K_D of the antigen bound to the paratope, and we can plot the relationship between K_D and probability of amino acid mutations. In this invention, we interpret clusters in the K_D-mutation graphs as evidence of significant immunological activity. When the immune system has found a stable heavy or light chain antibody structure small single-base mutations indicate that it can significantly improve specificity by mutating a single amino acid. Conversely, a single cluster with low K_D indicates discovery of a successful somatic mutation. We strongly associate evidence of successful somatic mutations with paratopes.

[0172] Although great strides have been made by computational chemists, current methods to computationally infer protein tertiary structure using only knowledge about the amino acid sequence and solvent is still challenging, time consuming and error prone. However, significant improvements in computation time and accuracy are available if prior information or constraints are incorporated during putative folding. The Web Antibody Modeling (http://antibody.bath.ac.uk/) has demonstrated antibody models with a high degree of success. Our invention uses the computed structure of the antibodies we discover, and the paratopes identified in example 4 above, to assign realistic probabilities to putative protein folds.

[0173] The original Dead-End Elimination (DEE) algorithm identifies several thousand putative protein foldings. Recent improvements have sped up this algorithm and provided structural and statistical filters that significantly reduce this number to approximately 1,000. The run time is approximately 5.5 hours.

[0174] The present invention can computationally estimate the free binding energy of the antibody and the putative protein folding. To compute this energy estimate, we computationally dock the antibody with the putative protein. There are many methods of doing this, utilizing shape, electron density, or statistical frequencies. Large computing clusters significantly improve computation time. Typical run times on large computers run approximately 8 seconds per docking.

[0175] Our invention compares the computational estimate of free binding energy to the measured K_D and chooses the protein and epitope that most closely matches. The result of our invention is: a) a population of digitally immortalized antibodies against a specific antigen by obtaining the DNA sequences of their H and L chains; b) kinetic metrics of the interaction between each antibody and the antigen; c) computational mapping of the paratopes of each antibodies; d) computational mapping epitopes on the antigen; and e) modeling of the structure of the antigen aided by combining information from a), b), c) and d).

[0176] This example refers to Embodiment 3 and Embodiment 4. The surface of each B cell contains thousands of identical membrane bound B-Cell Receptors (BCRs), and to a lesser extent plasmocytes do as well. We can specifically label them by subjecting the cells to a known concentration of fluorescently labeled antigen which will bind to a fraction of the BCRs, said fractions being closely approximated by the expression:

[0177] $f=\frac{\left[a\right]}{K_D+\left[a\right]}$

[0178] where K_D is the dissociation constant and [a] is the concentration of antigen.

[0179] The fraction f can be estimated by the fluorescence signal. However, large errors are introduced because cells vary in size. Estimates off can be greatly improved by normalizing said fluorescent measurements by the size the of the cell, a dimension well correlated with forward scattered F1 signals from an appropriate cell sorting instrument, such as one from BD Biosciences, San Jose, Calif. Reversing the above formula, since we know the concentration of antigen [a] used to label the cells, the K_D can be estimated from the normalized fluorescent signals f as:

[0180] $K_D=\frac{\left[a\right]\left(1-f\right)}{f}$

[0181] Our invention distributes cells into microwells and captures DNA complementary to mRNA on uniquely tagged oligonucleotides. It is ordinarily difficult to associate the K_D inferred by FACS fluorescence with a tagged oligonucleotide since the cells are pooled prior to said distribution. One way our invention solves this association is by using the relationship between the K_D and the abundance (or frequency) of B cells expressing a particular antibody: the number of B cells expressing identical antibody molecule n is a function of the K_D:

[0182] $n=\frac{c}{K_D^3}$
where c is an arbitrary constant.

[0183] This relationship applies not only to the number of sequences we read from said distributed microwells, it also applies to the antigen positive cells we measure in the FACS. Therefore, the constant c is determined by:

• [0184] a) determining the number of times each sequence has been sampled;
• [0185] b) using an arbitrary value of c, computing the copy number associated with each sequence;
• [0186] c) sorting the sequences in order of said copy number;
• [0187] d) normalizing the side-scattered FACS signal F2 by the forward scattered FACS F1 signal to compute and estimate of the bound BCR surface density;
• [0188] e) sorting the FACS signals by said bound BCR density estimate;
• [0189] f) adjusting the parameter c to insure the total number of sequences equals the total number of FACS signals;
• [0190] g) aligning said sequences with said bound BCR density estimates;
• [0191] h) computing the K_D for each said bound BCR density estimate;
• [0192] i) associating with each sequence the average K_D of said bound BCR density estimate aligned with said sequences.

[0193] We simulated data from the forward scattered F1 signal of a FACS sort of plasmocytes. We displayed the fluorescent amplitudes and the number of cells counted with a specific fluorescent amplitude.

[0194] Starting with our simulated F1 and F2 signals, we followed the procedure labeled as steps a) through i) above. A comparison of the K_D s inferred by our method and K_DS reported in the literature were quite good. The reasonably close match of the measured and predicted K_DS demonstrates the value of this approach.

[0195] Our invention distributes a certain number of cells over a surface populated by microwells. After a short time, e.g.: 3 minutes, said cells settle into the microwells. There is a competition between having as many microwells, and therefore as many cells, as possible, and leaving the microwells large enough that a) sufficient mRNA is captured by tagged oligonucleotides, and b) material from the cell's interior does not substantially interfere with mRNA-DNA hybridization. We anticipate that for cells with a diameter of 10-15 microns such as B cells, a 50 micron well size is a reasonable compromise. Clearly for other cells or different chemistries, the well size may vary.

[0196] Once the size of the microwells is fixed, the number of microwells must be determined. A convenient dimension for the entire array is the size of a microscope cover slip, 34×34 mm, of which only 24×24 mm is used in order to allow room for mechanical support. This leaves room for 230 thousand microwells placed on 50 micron centers.

[0197] As we distribute the cells over the microwells we would like one cell to settle into each microwell. However, if we spread too many cells over the microwells there will be a high probability that two or more cells may settle in the same well. If we spread too few cells over the microwells, we will leave many microwells without a single cell. The statistical optimization of this problem is well known to those skilled in the art. Briefly, consider a single cell entering the volume immediately above the microwells. The probability p a microwell contains said single cell is

[0198] $p=\frac{1}{230\text{,}000}.$
The probability prob that two microwells contain one cells each is, according to the binomial theorem, prob=2p (1−p), since we wish one of the two cells to be in the well while the other cell is anywhere else. For n cells, prob=np(1−p)ⁿ⁻¹. The number of single cell microwells (singlets) equals 230K times prob. Using the same formalism, the number of unoccupied microwells equals 230K(1−p)

[0199] To optimize the number of distributed cells, we computed the probability of multiplets m, m=1−(1−p)ⁿ−np(1−P)ⁿ⁻¹; i.e.: the probability that a microwell will receive 2 or more cells as a function of number of cells. A customary balance, and the one we use here, between multiplet probability and number of cells is 5% which occurs when n=82,000 cells.

[0200] We computed the relationship between number of single cells occupying single microwells and number of cells distributed over the microwells. For example, when the number of distributed cells equals 82,000, the number of single cell wells 57,400.

[0201] Every microwell contains two tagged oligonucleotides to capture Ab mRNAs: one for the heavy chain and one for the light chain. In order to digitally associate these two sequences, at least one molecule from each well containing a cell must be sequenced. Since the cDNA sequences are released and pooled, we must over-sample the sequences to insure at least one sequence from the heavy and light chain of each cell are sequenced.

[0202] The population of sequences is twice the number of cells. The probability p of selecting a particular heavy chain sequence equals the probability q of selecting the light chain sequence:

[0203] $p=q=\frac{1}{164\text{,}000}.$
The probability of not selecting a particular sequence is r=1−p−q. The number of wells with both chains sequenced at least once is computed using the multinomial distribution: prob=164,000(1−rⁿ−nprⁿ⁻¹−npqⁿ⁻¹). The number of wells with both chains sequenced was computed as a function of total number of samples sequenced. A customary over-sequencing rate to insure good coverage is 5×. Using 5×82,000, this number of sequences provides double sequences from 160,000 microwells, for a double sampling percentage of 97.6%.

[0204] The hybridoma line (NQ2-12.4) has a measured K_D of 2.8×10⁻⁷ molar⁻¹. Measured antibody k_on values tend to cluster around 105 sec⁻¹ molar⁻¹, giving a k_off of 2.8×10⁻² sec⁻¹. K_Ds larger than 10⁻⁷ molar⁻¹ are of little interest for research, diagnostics or therapeutics because they bind their respective antigen very weakly, dissociate from their antigen quickly, and substantially lower the probability of obtaining desired outcomes. Tight binding antibodies are much more useful, typically with K_DS less than 10⁻⁸. Our invention takes advantage of the fact that tight binding K_D s typically have k_off values<2.8×10⁻² sec⁻¹, and therefore easily allow the use of fluorescence to measure K_D and k_off in a 24×24 mm antibody array. k_on is computed from

[0205] $k_{\mathrm{on}}=\frac{k_{\mathrm{off}}}{K_D}.$
In this example, we show that our invention works for the extremely weak case of 2.8×10⁻² sec⁻¹ and therefore will work for other commercially interesting antibodies.

[0206] In this embodiment, a microscope coverslip is coated with Protein A. Cells are distributed over the microwells, lysed, and said coverslip placed over the microwells. In this embodiment oligonucleotides are affixed to a surface inside the microwells. If a cells lysed in a microwell contains a large number of antibodies, these antibodies will diffuse throughout the well, into the vicinity of the Protein A and tightly bind, as is well known to those skilled in the art. After an incubation time commensurate with the anticipated range of k_on, and antibody concentration, the coverslip is removed, washed, and subjected to a solution of fluorescently labeled antigen.

[0207] We performed a detailed molecular simulation of the binding of NQ2-12.4 to phOX, using the equilibrium expression: Ab.Ant Ab+Ant. We concentrated on the first 3 minutes after unbound antibody Ab was exposed to two initial concentrations of fluorescently labeled antigen: [Ant₀]=100 nM, and [Ant₀]=200 nM. The evolution of unbound antibody Ab, bound antibody Ab.Ant and free antigen Ant is summarized by the following well known relationships:

[0208] $\begin{array}{c}\frac{d\left[\mathrm{Ab}·\mathrm{Ant}\right]}{dt}=-k_{\mathrm{off}}\left[\mathrm{Ab}·\mathrm{Ant}\right]+k_{\mathrm{on}}\left[\mathrm{Ab}\right]\left[\mathrm{Ant}\right]\\ =-\frac{d\left[\mathrm{Ab}\right]}{dt}\\ =-\frac{d\left[\mathrm{Ant}\right]}{dt}\end{array}$

[0209] Fluorescently labeled antigens are easily measured and closely approximate the binding of unlabeled antigens. If we start with no bound antigen, after soaking the antibodies in antigen for over 3 minutes the fraction f of antibody sites with bound antigen divided by the total number of antibody sites asymptotically approaches equilibrium as:

[0210] $f=\frac{\left[\mathrm{Ab}·\mathrm{Ant}\right]}{\left[\mathrm{Ab}·\mathrm{Ant}\right]+\left[\mathrm{Ab}\right]}=\frac{\left[\mathrm{Ab}·\mathrm{Ant}\right]}{\left[{\mathrm{Ab}}_0\right]}=\frac{\left[\mathrm{Ant}\right]}{K_D+\left[\mathrm{Ant}\right]}$

[0211] If we know the total number of antibodies we could measure the fractional occupancy f directly: we would carefully excite the fluorophores with a well controlled flux of light and count the number of photons scattered into our detector. Based on the fluorescent cross-section and known probability of photobleaching, we could quite closely estimate the number of fluorophores. The difficulty with this approach is we do not know the number of total antibodies, and therefore we cannot form a ratio of measured fluorophores to total fluorophores in order to calculate the ratio f.

[0212] Our invention solves this problem by measuring two or more fluorescent amplitudes. The signal from a fluorescent microarray reader is dependent on the concentration of fluorophores. Usually the relationship is approximately linear. If we measure the same total signal from a fixed deposition of antibodies, and properly correct for known photobleaching effects, then the fractional occupancy f is also linearly related to measured fluorescence: meas₁=α.f₁ for some fractional occupancy f₁ related to some specific antigen concentration [Ant₁]. For the same deposition of antibodies, this is equally true for meas₂, f₂ and [Ant₂] with the same α. Using the relationship for f stated above, we solve for K_D as a function of meas₁, [Ant₁], meas₂, and [Ant₂]:

[0213] $K_D=\frac{{\mathrm{meas}}_2-{\mathrm{meas}}_1}{\frac{{\mathrm{meas}}_1}{\left[{\mathrm{Ant}}_1\right]}-\frac{{\mathrm{meas}}_2}{\left[{\mathrm{Ant}}_2\right]}}$

[0214] Although this equation is written as a function of two fluorescent measurements associated with two antigen concentrations, it is to be understood that more fluorescent measurements associated with more antigen concentrations could easily be used by those skilled in the art of non-linear parameter estimation.

[0215] Once the K_D is determined we solve for α using

[0216] $\alpha =\frac{{\mathrm{meas}}_1}{f_1}=\frac{{\mathrm{meas}}_1\left(K_D+\left[{\mathrm{Ant}}_1\right]\right)}{\left[{\mathrm{Ant}}_1\right]}.$
Knowing α is extremely useful in our invention since we can then convert our measured signals to fractional occupancy. Since we know the number of measured fluorophores from the calibration of the fluorescent reader, the fractional occupancy gives us the total number of antibodies, both bound and unbound. This number correlates highly with the identity of cells, i.e., whether they are plasmocytes or B cells. Plasmocytes often contain additional somatic mutations from their B cell progenitors. Since the B cell receptors on the surface of plasmocytes were created before the somatic mutations, the antibodies may be different. Identifying B cells, as well as identifying high frequency B cells, allows us to better estimate tight binding antibodies.

[0217] The above description requires the accurate measurement of both antigen and fluorescent signal. We use the initial concentration of antigen [Ant₀] as an estimate of [Ant]. This is generally adequate when [Ant₀] is substantially greater than the concentration of total Ab. As estimates of [Ab], we use an antibody density of 22 fmol cm⁻² over an area of 50 micron² in a volume of 125 picoliters for a concentration of 4.4 nM. Our initial concentrations [Ant₀] of 100 and 200 nM are substantially greater than this.

[0218] The measurement of fluorescence must be done in a solution with [Ant₀]=0 to minimize background noise. We measure fluorescent amplitude as a parameter while measuring k_off. Fluorescence decreases as antibody bound antigens are subjected to an antigen free solution. In general, it is difficult to measure [Ab.Ant] at time=0 using a coverslip since it takes a finite amount of time to load the coverslip covered by free solution into an appropriate reader while maintaining a flow of antigen free solution over the coverslip. In our invention, we sample the bound antibody concentration by measuring the fluorescence of fluorescently labeled antigen bound to the antibodies at approximately one minute intervals. Modern microarray readers (e.g.: ArrayIt InnoScan 700, TeleChem International, Inc., Sunnyvale, Calif.) scan 25×24 mm areas in one minute at 10 micron resolution, quite sufficient for 50 micron antibody deposits.

[0219] For example, phOX antigen bound to NQ2-12.4 decreases quite quickly and represents one of the worst antibodies our invention anticipates measuring since its fluorescent decay time is the fastest we anticipate sampling. Tighter antibodies will have decay times orders of magnitude greater. The microarray reader will image approximately 230,000 antibody spots in approximately one minute and therefore infeasible to image all spots continuously or immediately. Our invention samples the signal decrease three times. Typical fluorescent measurements in digital numbers are: 736 in 0.39 seconds, 137 in 1.39 seconds, and 26 in 2.39 seconds.

[0220] In our simulation, we fit the measured data meas to the well known formula for exponential decay: meas=Ae^−kofft. Using the above data and the non-linear parameter estimation features of Mathematica (Wolfram Research Inc., Champaign, Ill.) we compute A to be 1427 and k_off to be 0.028 sec¹.

[0221] This example applies to Embodiment 2 and Embodiment 4, above. NimbleGen oligonucleotide arrays contain staggered oligonucleotide pads 13 microns×13 microns in size. By staggered, it is meant alternating, for example, a chess board, wherein the black squares are staggered. When mating a NimbleGen array with PicoTiterPlate (PTP) hexagonal wells, there will be some oligonucleotide pads that are completely exposed to the microwell, and others that are buried under microwell walls or exposed to multiple microwells. To make our invention work with a NimbleGen oligonucleotide array and PTP hexagonal wells, we must insure that a majority of wells contain at least two pads with a large portion of their area exposed.

[0222] We conducted a simulation wherein 100×100 hexagonal wells were constructed on 50 micron centers with a diameter, that is, a distance between opposing walls, of 44 microns. A grid of points on one micron centers was placed over each hexagonal well and only points falling inside each hexagon were selected. For each selected hexagonal well, the number of points in the unique pad containing it (computed by dividing its location by the size of each pad) was incremented. After processing each point we had a data base that listed for each well the pads it contained and the number of points in each pad. Since we knew the maximum number of points in a pad, we computed the fraction of each pad inside each well.

[0223] Each well contained at least six pads. For each well we selected the six pads with the largest fractions, creating a 10,000×6 array. Considering this to be 10,000 samples of six numbers, the mean of the 6 fractions and their standard deviations, are: the first mean is 1 with a standard deviation of 0; the second mean in 0.99 with a standard deviation of 0.02; the third mean is 0.93 with a standard deviation of 0.07; the fourth mean is 0.81 with a standard deviation of 0.11; the fifth mean is 0.51 with a standard deviation of 0.15; and the sixth mean is 0.4 with a standard deviation of 0.15.

[0224] Every well contained at least one complete pad. 99% contained at least 2 staggered pads. Therefore, our embodiment using NimbleGen oligonucleotide arrays and PTP hexagonal wells allows the heavy and light chains from 99% of the distributed cells to be digitally matched using DNA tags.

[0225] This method is useful when using a robotic microarray spotter (e.g.: SpotBot 2, TeleChem, International, Sunnyvale, Calif.). A 1% low melting point agarose (Invitrogen Corp., Carlsbad, Calif.) solution is dissolved in purified water and poured over the surface of a glass slides at 70° C. (2.0 ml per slide). After gelling of the agarose, slides are dried in air. The dried slide is placed into the microarray spotting equipment and warm water (e.g.: 70° C.) is dispensed onto the agarose. The agarose melts and is washed away with another application of warm water. A hole pattern is fabricated wherein said holes are as reproducible as the microarray spotter's tolerance (approximately 10% for SpotBot 2). Once said holes are formed and cleaned, other molecules can be spotted into the holes using the same microarray spotter. If there is an alignment error, the agarose offer a low resistance to the mircroarray needle. The final result is a hole whose size is controlled by the warm water injection process, and a uniform coating of detection molecules attached to the glass. The molecules coated onto the glass can be, for example, oligonucleotides, proteins, antibodies, or capture molecules, or a mixture of 2 or more oligonucleotides, proteins, antibodies, or capture molecules.

[0226] The generation of unique oligonucleotide tags and their incorporation into oligonucleotides including a region complementary to mRNA and a PCR primer.

[0227] One method of fabricating an oligonucleotide array complementary to mRNA utilizes synthesized oligonucleotides. This is difficult when considering arrays with, say, 40,000 or more oligonucleotides each with a unique tag. An efficient method of producing 40,000 oligonucleotides with unique tags comprises:

• [0228] a) Synthesizing 200 unique oligonucleotides without terminating phosphate groups;
• [0229] b) Synthesizing another 200 unique oligonucleotides with terminating phosphate groups;
• [0230] c) Ligating oligonucleotides from step (a) with oligonucleotides from step (b). Note that oligonucleotides from step (a) cannot ligate with one another because of the missing phosphates. Similarly, oligonucleotides from step (b) cannot ligate with one another because of the extra phosphates.
• [0231] d) Ligating the oligonucleotides from step (c) with the PCR primer.
• [0232] e) Ligating the oligonucleotides from step (d) with the region complementary to the mRNA.

[0233] The value of 40,000 pads is only exemplary and not intended as an additional limitation on the size of the microarray. This example, used in conjunction with a glass slide coated with low melting point agarose, provides a low cost example of Embodiment 5.

[0234] SU-8 photoresist (MicroChem Corporation, Newton, Mass.) was spin-coated onto a 10 cm silicon wafer in accordance with manufacturer's recommendations. A pattern of 50 micron cubic micro-wells was exposed using ultraviolet light from a mask aligner (SUSS MicroTec AG, Schleiβheimer Str. 90, D 85748 Garching b. München) in accordance with manufacturer's recommendations and developed using SU-8 developer (MicroChem). Polydimethylsiloxane (PDMS) (SylGard Elastomer, Ellsworth Adhesives, Germantown, Wis.) was mixed, cured and removed in accordance with manufactuer's recommendations. The bottom tray (FIG. 5) was formed by using a razor blade to cut out an array of 205 by 154 micro-wells (31, 570 micro-wells) and carefully placing it onto a clean and dry microscope slide where it formed a hydrophobic bond with the glass slide. The dimensions and location of the micro-wells exactly matched the location and size of pads of oligonucleotides on a DNA chip (385K CGH array, Roche/Nimblegen, Madison, Wis.). The DNA chip formed the top of the micro-wells (FIG. 6) and was placed above the micro-wells so that most wells were directly beneath DNA and most well walls were in contact with bare glass. When micro-wells this small are used, the positioning, alignment, and subsequent sealing of a DNA slide above PDMS wells is sufficiently delicate that mechanical assistance is generally needed, such as a mask aligner or a custom built apparatus.

[0235] One example of an available mask aligner is the MJB4 four inch manual mask aligner from SUSS MicroTec AG (Schleiβheimer Str. 90, D 85748 Garching b. München). A glass slide is affixed with mylar tape to a 4″ square quartz plate and inserted into the aligner's mask vacuum chuck. PDMS is affixed to a similar plate and inserted into the wafer chuck. The two surfaces are aligned using optics and positioning controls available on the aligner. Once aligned, the PDMS is removed and processing continues until the DNA slide and wells must be aligned. At this time, the PDMS is reinserted into the aligner, the two surfaces (DNA and PDMS) are placed into contact, and the entire assembly is left alone for several minutes while cDNA is constructed from antibody mRNA.

[0236] Although this procedure can be carried out it has several drawbacks: 1) The components to be aligned must be pushed into a UV protected area precluding easy access by people and equipment; 2) there are very few places for excess fluid from the necessarily wet cell suspension deposited on top of the wells; and 3) there are expensive features on the aligner that are not needed.

[0237] The system can be more efficiently used with a custom aligner designed to contain a small amount of fluid, provide easy access for the experimenter and equipment, while providing a reduced set of functions at a reduced cost. An example of a custom aligner 500 is shown in FIG. 9. Four vertical posts 502 screw into aluminum base plate 501 and support aluminum table 503. On aluminum table 503 sits the bottom pressure plate 504. A first plastic insert 505 and a second plastic insert 506 are sandwiched between bottom pressure plate 504 and top pressure plate 507.

[0238] Vacuum port 508 provides close contact between bottom pressure plate 504 and plastic insert 505, as well as top pressure plate 507 and plastic insert 506. The bottom tray, includes micro-wells 511 placed on glass microscope slide 510, and the tray is held firmly to the top side of plastic insert 505 by additional vacuum ports 509. Similarly, a DNA chip 512, having the same dimensions as the microscope slide 510, is held firmly to the bottom side of plastic insert 506 by additional vacuum ports 513.

[0239] When bottom vacuum is released, the bottom micro-wells 511 and glass slide 510 can be carefully moved in two axes by sliding it against plastic insert 509. Once the wells 511 are optimally aligned, vacuum is applied to vacuum ports 509 to keep the slide 510 from moving. A small computer interfaced microscope (not shown) is positioned beneath the aluminum table 503 to image cells in wells using flat field light entering the top of apparatus 500.

[0240] Once alignment is achieved, marks are placed on the plastic insert 509 to record the position of slide 510. Alternatively, objects are placed next to the glass slide 510 to record the position of the slide 510. Then microwells 511 and the glass slide 510 can be removed, processed and returned to alignment apparatus 500 within a positional accuracy of 10 microns. It was found that 10 micron accuracy was adequate to correctly align the microwells 511 with the DNA chips 512 so that specific addresses on the chip 512 could be correctly matched to specific addresses of the microwells 511.

[0241] During alignment and processing, it is often useful to examine the microwells, or cells in the microwells, under a moderate power microscope (e.g.: 100×). The field of view of the microscope seldom encompasses all microwells. Therefore, it is difficult to know exactly which wells are being viewed by examining a single microscope image either in real time or afterwards from a captured image. It may be desirable to know exactly which wells are being viewed without having to count wells from an edge or corner.

[0242] For this purpose, a coded shape such as a particular shape (FIG. 10) may be embedded into the wells. The shape is a tradeoff between three constraints: 1) the shapes should be different enough that epoxy coated wafers can easily reproduce them and, in turn, they can be seen in the PDMS impressions taken from the epoxy coated wafers; 2) the well volume should not be substantially different from well to well so that cells will experience approximately the same microenvironment; and 3) the wall thickness is not reduced in a manner which would weaken the PDMS structure. Fragments of an octagon provided a useful balance between these constraints. Octagon fragments were separated into symmetric or nonsymmetric fragments and used in shapes of FIG. 10.

[0243] In this example a 4×4 block of wells was chosen because under our laboratory microscopes we could easily place 16 wells in the field of view. Referring to FIG. 10, a 16 well block 600 is an example of the use of octagon fragments to embed three numbers in the shape of the wells: 1) the row number; 2) the column number; and 3) the pattern number. In this example, there are a maximum of 64 rows and columns, each numbered 0 through 63. Since there are 4 wells in each row and column per block, this example addresses a maximum of 256 wells in all rows or columns of a pattern. Binary representations are used to encode the row and column numbers. As is well known in computer literature, six binary digits are required to encode values 0 through 63. Symmetrical wells are used to represent row and column 1s and 0s. For example, in FIG. 10, well 605 is a 1 and well 607 is a zero.

[0244] Nonsymmetrical wells are used at the corners of the block. Well 601 provides a starting point and starting direction. In addition, the other corners (602, 603 and 604) provide chirality, i.e. the shape is not identical to its mirror image in left and right handed microscopes images viewed from either the top or the bottom. Therefore, a method is needed to determine which wells go with which blocks without presupposing a particular viewing direction. The three corners of the block that are not the start well have a unique orientation that points into the center of the block. Since there are two such orientations (e.g.: well 602 and 603), the two orientations can be used to represent a binary 1 (well 602) and a binary 0 (well 603) in order to encode additional information. In this example, there are different well patterns each with a slightly different dimension so as to accommodate variations in PDMS shrinkage caused by different curing temperatures. The three corners encode for 8 different values, 0 through 7.

[0245] FIG. 10 shows a block encoding the three values: column 3, row 4 and pattern 5. The block is first located in a microscope's field of view by finding nonsymetrical well 601 which is designed to have the shape of a home-plate in baseball. Following four wells in the direction of well 601, well 602 defines the chirality of the block, in this case to the left. Wells 603 and 604 confirm this chirality. In addition, starting from the high bit, wells 604, 603 and 602 encode for the binary number 101 which, expressed in decimal, is the well pattern 5.

[0246] The column code starts with high bit 606 and ends with low bit 605: 000011, or column 3. The row code starts with high bit 608 and ends with low bit 607: 000100, or row 4.

[0247] This is an illustrative example only. The shapes can easily vary to accommodate different designs and materials, as well as the number of wells per block, the sizes and relative arrangements.

[0248] Insofar as the description above and the accompanying drawings disclose any additional subject matter that is not within the scope of the claims below, the inventions are not dedicated to the public and the right to file one or more applications to claim additional inventions is reserved.

[0249] The preceding merely illustrates the principles of the invention. It will be appreciated that those skilled in the art will be able to devise various arrangements which, although not explicitly described or shown herein, embody the principles of the invention and are included within its spirit and scope. Furthermore, all examples and conditional language recited herein are principally intended to aid the reader in understanding the principles of the invention and the concepts contributed by the inventors to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents and equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure. The scope of the present invention, therefore, is not intended to be limited to the exemplary embodiments shown and described herein. Rather, the scope and spirit of present invention is embodied by the appended claims.