WO1996026733A1

WO1996026733A1 - Codominance-mediated toxins

Info

Publication number: WO1996026733A1
Application number: PCT/US1996/002271
Authority: WO
Inventors: Alexander Varshavsky
Original assignee: Alexander Varshavsky
Priority date: 1995-03-01
Filing date: 1996-02-28
Publication date: 1996-09-06
Also published as: EP0812204A1; JPH11502407A; EP0812204A4; AU5132996A

Abstract

Disclosed is a new class of therapeutic reagents referred to as codominance-mediated toxins. These reagents include (1) an effector domain; (2) a first codominant signaling moiety located proximately to a binding domain for a first protein; and (3) a second codminant signaling moiety located proximately to a binding domain for a second protein. The ability to mask and unmask the signaling moiety by the binding of intracellurar reagents, and the codominance of signaling moieties are exploited to design a novel class of reagents having previously unattainable selectivity.

Description

CODOM1NANCE-MEDIATED TOXINS

Background of the Invention

Abnormal cells differ from their normal progenitors and other cells of the same organism in a variety of ways, including protein composition. For example, virus-infected cells contain virus-specific proteins; the levels of certain cellular proteins are also altered as a result of viral infection. Cancer cells, which can grow at sites of their initial emergence and at distant sites that they are capable of colonizing, differ from their normal progenitors in the patterns of gene expression. Some of the tumor-specific proteins are altered versions of normal proteins, in that they are encoded by genes whose mutations were among the causes of a malignant phenotype.

A viral genome encodes relatively few proteins. Since most of these proteins have functional counterparts in cells that the virus infects, effective antiviral drugs remain, by and large, a goal to be reached. With malignant tumors that cannot be eliminated by surgery alone, the problem of finding a drug up to the task is even more complicated, because compositional differences between a tumor cell and its normal progenitor can be subtle and quantitative rather that qualitative. In addition, cells of a tumor are often heterogeneous genetically and in protein composition. These difficulties are a major reason for the failure of present-day therapies to cure most cancers. The problem is not necessarily the insufficient specificity of a drug — some of the cytotoxic compounds used in cancer therapy, for example, methotrexate and vinblastine (which bind, respectively, to dihydrofolate reductase and tubulin), are highly specific for their ligands. The problem is that a single drug target may not define the cell type to be eliminated unambiguously enough.

Another approach to cancer therapy involves the linking of a toxin to an antibody or another ligand (for example, a growth factor) that binds to a target on the surface of tumor cells. The tumor selectivity of these drugs, called immunotoxins or chimeric toxins, is often higher than that of small cytotoxic drugs such as methotrexate. Unfortunately, a surface marker that an immunotoxin recognizes may be present not only on target cells. Moreover, since this marker is often not essential for tumorigenicity, there may be cells in a tumor that lack the marker but are still malignant. These are some of the limitations of present-day immunotoxins. Yet another approach is to increase or redirect the power of the immune system to identify and selectively destroy tumor cells. Immunotherapy of cancer has a long and checkered history. The recent revival of this strategy, brought about by advances in the understanding of antigen presentation and lymphocyte- mediated cell killing, holds the promise of a rational and curative therapy. Since this goal remains to be reached, it is still far from certain that intelligent manipulation of the immune system will prove sufficient for a complete and assured cure of most cancers.

Summary of the Invention The subject invention is based on the discovery of a new and generally applicable set of methods and reagents for eliminating (or modifying) dangerous cells. The invention relates, in one aspect, to a new class of therapeutic reagents referred to as codominance-mediated toxins. These reagents include (1) an effector domain; (2) a first codominant signaling moiety located proximately to a binding domain for a first protein; and (3) a second codominant signaling moiety located proximately to a binding domain for a second protein. The ability to mask and unmask the signaling moiety by the binding of intracellular reagents, and the codominance of signaling moieties are exploited to design a novel class of reagents having previously unattainable selectivity. In another aspect, the invention relates to a method for selectively killing or precluding division of a cell known to contain both a first protein and a second protein. In this method, a DNA construct is provided which encodes 1) an effector domain; 2) a first codominant degron located proximately to a binding domain for a first protein; and 3) a second codominant degron located proximately to a binding domain for a second protein. The DNA construct is introduced into the cell under conditions appropriate for expression of the intracellular degron- dependent, ligand-regulated toxin. Brief Description of the Drawings

Fig. 1 is a diagram showing an tracellular degron-dependent, /igand- regulated toxzn (termed indelin). An indelin designed to kill [Pl ⁺ P2⁺] cells that express proteins PI and P2, but to spare the other cell types — [Pl⁺ P2 ], [PI^" P2⁺] and [PI^" P2^'] — contains a cytotoxic effector domain and two degradation signals (degrons) dl and d2, placed within or near two domains PI* and P2* that bind, respectively, to PI and P2. In cells other than [Pl⁺ P2⁺], at least one of the indelin's degrons, dl and/or d2, is active (unobstructed), resulting in a short-lived and therefore relatively nontoxic indelin. By contrast, in [Pl ⁺ P2⁺] cells both dl and d2 would be masked by PI and P2 (bound, respectively, to the indelin's PI* and P2* domains), resulting in a long-lived and therefore toxic indelin. This design requires that PI, P2, and the indelin reside in the same compartment — the one in which the indelin's toxic domain exerts its effect (it can be the cytosol, the nucleus, and/or another compartment). In addition, the degrons dl and d2 must be active in the relevant compartment; they may or may not be identical — the design is compatible with either choice.

Fig. 2 is a diagram showing an intracellular translocation signal-dependent, igand regulated tox (termed intralin). (A) An intralin designed to kill [Pl ⁺ P2⁺] cells but to spare the other cell types — [Pl⁺ P2 ], [PI^" P2⁺] and [PI^" P2 ] — is a fusion containing an effector domain that is toxic in the cytosol but not in the nucleus and two nuclear localization signals (NLSs), placed within or near two domains PI* and P2* that bind, respectively, to PI and P2. In cells other than [Pl ⁺ P2⁺], at least one of the indelin's NLSs is active (unobstructed), resulting in a nuclear and therefore nontoxic intralin. By contrast, in [Pl⁺ P2⁺] cells both NLSs would be masked by PI and P2 (bound, respectively, to the intralin's PI* and P2* domains), resulting in a cytosolic and therefore toxic intralin. In a [PI ⁺ P2⁺] cell, PI and P2 must be located at least in the cytosol. (B) Same as A but the intralin's effector domain is toxic in the nucleus but not in the cytosol. This intralin would kill [Pl⁺ P2 ], [PI^" P2⁺], and [PI^" P2 ] cells that lack at least one of the two cytosolic proteins, PI and P2, but would spare [Pl⁺ P2⁺] cells that contain both PI and P2. Fig. 3 is a diagram showing a hybrid cødominance-mediated toxin (termed comtoxi ). A hybrid comtoxin contains: an effector domain that is active in the nucleus but not in the cytosol; a degron placed within or near the domain PI*; and a nuclear localization signal (NLS) placed within or near the domain P2*. PI* and P2* bind, respectively, to intracellular proteins PI and P2. As described in the diagram, this comtoxin would kill exclusively [Pl⁺ P2 ] cells, which contain PI but lack P2. In these cells, PI must be located at least in the nucleus, whereas P2 must be a cytosolic protein. Another constraint is that the degron dl must be active at least in the nucleus. The actually indicated state of comtoxin in [Pl ⁺ Pl⁺] cells requires that PI is present in both the cytosol and the nucleus, and that the degron dl is active in both of these compartments. If PI is present largely in the nucleus, or if dl is active only in the nucleus, the metabolic properties of this comtoxin would differ from those indicated, but its selectivity (killing exclusively [Pl⁺ P2 ] cells) would remain the same. Fig. 4 is a diagram showing a split toxin. In this design, two subdomains of a toxic effector domain are separated by an insert whose sequence contains a binding site for an intracellular protein PI. Unlike the PI* domains of other comtoxins (Figs. 1-3), the Pl-binding site of a split comtoxin should be a relatively short (peptide-size) region that remains conformationally flexible unless it is bound by PI. For this design to work, the affinity between subdomains of a toxic domain should be low enough to make their interaction substantially reversible. A flexible insert between subdomains of a toxin can also contain two binding sites for intracellular proteins PI and P2, arranged so that the binding of either PI or P2 would be sufficient to impair the reconstitution of active toxin. The resulting comtoxin would kill exclusively cells that lack both PI and P2. Other split-toxin designs are mentioned elsewhere.

Detailed Description of the Invention

The present invention is based on the identification of a new and generally applicable strategy for eliminating, or modifying, dangerous cells in a multicellular organism. The main idea of this strategy is grounded in the property of codominance which is exhibited by a variety of signals that reside in biopolymers such as, for example, proteins and nucleic acids. To simplify discussion, protein signals and their use in connection with the invention will be used as examples. One of skill in the art will recognize that the principles described herein are transferrable to applications in connection with nucleic acids, for example, and such extensions fall within the scope of this invention. Protein signals include, but are not limited to, degradation signals (degrons) and various translocation signals, including nuclear localization signals (NLSs), which confer on a protein the ability to be transported from the cytosol into the nucleus of a cell.

Two signals present in a protein are termed codominant when each of these signals can exert its effect on the protein independently from, and without a significant interference by, the other signal in the protein. As described in greater detail below, the main idea of this invention is that the rational and novel combination of codominant signaling moieties, protein-interacting domains and an effector domain can result in the creation of a new class of drugs having previously unattainable selectivity. These new pharmaceutical reagents are referred to herein as comtoxins (codominance- ediated toxins), and specific examples are shown in Figs. 1-4.

Figure 1 shows a comtoxin bearing degradation signals. Comtoxins of this class will be referred to herein as indelins (intracellular rfegron-dependent, /igand- regulated toxins). The left-hand panel of Fig. 1, labeled "Cell Type", indicates either the absence (PI^" or P2^"), or the presence (Pl ⁺ or P2⁺) of the first protein or polypeptide (PI) or the second protein or polypeptide (P2) in the cell in question. The right-hand panel shows each of the three required elements of a comtoxin in diagrammatic form. A fusion protein of the type shown in the right-hand panel is delivered into cells in particular by introducing an appropriate expression vector encoding the fusion protein into the cells. Introduction of the expression vector into the cell is followed by transcription and translation of the encoded fusion protein. In this example, the toxic domain is a protein or polypeptide that is toxic at least in the cytosol of the cell. The N- terminal domain of the comtoxin of Fig. 1 comprises a first degron

(dl) which is located proximately to a binding domain (PI^*) for a first intracellular protein. The domain which is flanked by the N-terminal domain and the toxic domain comprises a second degron (d2), located proximately to a binding domain (P2^*) for a second intracellular protein. The first and the second protein are indicated in the right-hand panel as PI and P2, respectively.

For the comtoxin of Fig. 1, two alternative metabolic fates for the fusion protein are possible — the fusion protein is either short-lived or long-lived. If the fusion protein is long-lived, the toxic domain-containing comtoxin will persist in the cytoplasm for a period of time sufficient to kill the cell. If the fusion protein is short-lived, the toxic domain will be degraded by the highly processive ubiquitin-dependent proteolytic pathways that recognize and target either one or both of the comtoxin 's degrons, dl and d2. Whether the fusion is long-lived or short-lived is dependent upon the presence or absence of PI or P2 in the cytosol of the cell. If neither PI nor P2 is present in the cell, the fusion protein is short¬ lived due to the presence of the unobstructed (unmasked) degrons dl and d2. If either PI or P2 is present, but not both, the fusion protein is still short-lived due to the fact that one of the two codominant degrons remains unobstructed

(unmasked) and will retain its ability to signal the degradation of the comtoxin. However, if both PI and P2 are present, both degrons (dl and d2) will be sterically masked, and the fusion protein will be long-lived. Thus, the comtoxin of Fig. 1 is designed to kill cells that express both PI and P2, but spare cell that lack PI and/or P2.

Another example of comtoxin design is shown diagramatically in Fig. 2. The left-hand panel of Fig. 2, labeled "Cell Type" indicates either the absence (PI or P2 ) or the presence (Pl⁺ or P2⁺) of the first protein (PI) or the second protein (P2) in the cell in question. Panel A of Fig. 2 relates to a comtoxin molecule bearing a cytosol-specific toxic domain, whereas Panel B relates to a comtoxin molecule bearing a nucleus-specific toxic domain. In all comtoxin molecules depicted in Fig. 2, the intracellular signaling domain is a nuclear localization signal (NLS). Molecules of comtoxin bearing at least one NLS which is not sterically masked will be transported from the cytosol to the nucleus. Thus, the comtoxin shown in Fig. 2, Panel A, will be transported to the nucleus if neither PI nor P2, or if either PI or P2 (but not both) are present in the cytosol. Since the toxic domain of this comtoxin is cytosol-specific (see below), the toxic effect will result if, and only if, the comtoxin is retained in the cytosol, and this will occur only if both codominant NLSs are sterically masked (i.e. , only when both PI and P2 are present in the cell's cytosol). Comtoxins of the type shown in Fig. 2 are referred to herein as intralins (mtracellular trønslocation signal-dependent, ήgand-regulated toxins).

Panel B of Fig. 2 shows the inverse situation where the effector domain of an intralin is toxic in the nucleus but not in the cytosol (see below). This type of intralin would kill cells that lack PI or P2, or both of them — the selectivity opposite to that of the intralin considered above (Fig. 2 A versus Fig. 2B). Indeed, a comtoxin bearing NLSs would stay in the cell's cytosol if, and only if, both PI and P2 are present in the cytosol, masking both NLSs of the intralin (Fig. 2B). Thus, while an indelin (a comtoxin bearing degrons) (Fig. 1) kills exclusively cells that contain both PI and P2, an intralin whose toxic domain is nucleus-specific (Fig. 2B) spares exclusively these cells. The ability to target cells that lack a set of predetermined proteins is important because cancer cells lack (or contain functionally inactive versions of) certain regulatory proteins ("tumor suppressors") that are present in the normal progenitors of these cells.

It should be noted that counterparts of constructs in Figs. 1 and 2 that bear a single P*-type domain can function as indelins or intralins; however, these constructs are not yet comtoxins (the mechanics of codominance requires more than one P*-type domain). At the same time, the ligand-controlled toxicity of these novel constructs would be analogous to that of comtoxins. Thus, in addition to being a natural intermediate in the designing of a comtoxin, an indelin or an intralin bearing a single P*-type domain can also be used as a drug specific for a single intracellular target.

It should be emphasized that conditional toxins of the present invention which bear a single P*-type domain represent a new drug design as well. While their selectivity is limited to one protein target, this target is intracellular, in contrast to the cell surface-expressed targets of present-day immunotoxins (also called chimeric toxins). The single-P*-type-domain conditional toxins of the present invention are "natural" intermediates in the construction of multiple- P*-type-domain comtoxins, and should also be useful as drugs under conditions in which the targeting of a single, predetermined intracellular protein is sufficient for the goals of a given therapy or cytotoxic cell selection in a mammalian cell culture in vitro.

Another type of comtoxin, referred to herein as a hybrid comtoxin, is shown diagrammatically in Fig. 3. This type of comtoxin bears both degron and NLS signaling moieties. The simplest comtoxin of this class, shown in Fig. 3, contains an effector domain that is toxic in the nucleus but not in the cytosol; a degron placed within or near domain PI^*; and an NLS placed within or near domain P2^*. As before, the PI' and P2^* domains should be able to bind, respectively, to intracellular proteins PI and P2. However, in contrast to a "pure" indelin or intralin (Figs. 1 and 2), this "hybrid" comtoxin would kill exclusively cells that contain nuclear protein PI but lack cytosolic protein P2, for only in such cells would the comtoxin be both nuclear (because its NLS is not masked, owing to the absence of P2) and long-lived (because its degron is masked by the PI -PI* complex).

A comtoxin bearing a degron and an NLS (Fig. 3) would fail to distinguish [Pl ⁺ P2 ] cells from other cell types if both PI and P2 are nuclear proteins. Fig. 4 illustrates a class of conditional toxins whose mode of selectivity addresses, in particular, this problem. A single-domain protein whose subdomains are separated (at a surface loop) by a conformationally flexible insert can adopt a (nearly) normal conformation, in which the insert is extruded to the outside of the folded domain. For example, the in vivo folding of ubiquitin, a 76-residue protein, was shown to be virtually unperturbed by the insertion of an unrelated 80-residue sequence at a site between the two subdomains of ubiquitin (Johnsson and Varshavsky, Proc. Natl. Acad. Sci. USA 91: 10340 (1994)). The idea of a "split" toxin (Fig. 4) stems from these and analogous data, and also from the concept that the extent of conformational flexibility of an insert between two subdomains of a protein can influence both kinetic and equilibrium aspects of the protein's folding. In particular, a conformationally rigid insert would be expected to perturb or preclude the interaction between the protein's subdomains. In the diagram of Fig. 4, two subdomains of a toxic domain are separated by a sequence that contains a binding site (PI^*) for an intracellular protein PI. Unlike the PI^* domains of other comtoxins (Figs. 1-3), the PI^* site of a split comtoxin (Fig. 4) should be a relatively short "peptide-size" (~ 10 to -40 residues) region that remains conformationally flexible unless it is bound by PI. The construct of Fig. 4 would be toxic in cells that lack PI but relatively nontoxic in Pl-containing cells. A flexible insert between subdomains of a toxin can also contain two binding sites for intracellular proteins PI and P2, arranged so that the binding of either PI or P2 would be sufficient to impair the reconstitution of active toxin. The resulting comtoxin would kill exclusively cell that lack both PI and P2. (Note that while a split toxin containing a single PI -binding site represents a novel design and useful therapeutic agent, it is not yet comtoxin in the strict definition, because no codominance and multiple protein binding sites are involved here as yet. The simplest "true" split comtoxin would contain at least two protein- binding sites for proteins PI and P2. To simplify terminology, the construct shown in Fig. 4 will be referred to as a "comtoxin" as well.)

For a split comtoxin to work, the affinity between subdomains of a toxic domain should be low enough to make their interaction substantially reversible. This would preclude irreversible activation of the toxic domain before its encounter with PI (Fig. 4). The affinity between subdomains of the toxic domain can be adjusted, if necessary, through mutational alterations analogous to those that have been used to adjust the affinity between the subdomains of ubiquitin (Johnsson and Varshavsky, Proc. Natl. Acad. Sci. USA 91: 10340 (1994)).

One specific example of a split-toxin design utilizes barnase, a ribonuclease secreted by the bacterium Bacillus amyloliguefaciens. Bamase is a 110-residue protein lacking disulfide bonds. It has been studied extensively as a model for protein structure and folding. In particular, it has been shown that two fragments of barnase (residues 1-36 and residues 37-110) can reassociate upon mixing to form the active enzyme with a nearly normal structure and thermal stability (Sanco and Fersht, J. Mol. Biol. 224: 741 (1992)). These findings, as well as the knowledge of the three-dimensional structure of barnase, and the analogous studies of split ubiquitin referred to above (Johnsson and Varshavsky, Proc. Natl. Acad. Sci. USA 91: 10340 (1994)) indicate that a flexible peptide linker inserted between residues 36 and 37 of barnase would be extruded to the outside of the folded barnase globule upon the association and coalescence of the two subdomains of barnase. A peptide linker that would function as a split-toxin's PI -binding site (Fig.

4) that modulates the reconstitution of active barnase can be chosen from a variety of candidate sequences. For example, it has been shown that synthetic peptides 10-15 residues long that contain a phosphorylated tyrosine residue in a predetermined sequence context can form tight complexes with the so-called SH2 domains of specific proteins (Felder et al. , Mol. Cell. Biol. 13: 1449 (1993)).

SH2 domains are present in a variety of regulatory proteins, including the cytosol- exposed regions of transmembrane receptors. A common feature of SH2 domains is their ability to form tight complexes with sequences containing a phosphotyrosine residue, with each SH2 domain recognizing a specific phosphotyrosine-containing sequence motif. Thus, a split toxin that comprises a split barnase with a relatively short (10-20 residues) linker sequence that connects the two barnase subdomains and binds to a specific SH2-containing intracellular protein would function as a conditional split toxin of the present invention: the activity of this toxin is modulated by the binding of an SH2-containing intracellular protein to the peptide linker, whose conformation (flexible in the absence of a bound SH2 domain, relatively rigid in its presence) would determine the efficiency of reconstitution of active barnase and hence the overall activity of this barnase- based split toxin.

A more complex split comtoxin can be designed, for example, to distinguish [Pl⁺ P2 ] cells from other cells types even if both PI and P2 are nuclear proteins. In addition to a Pl-binding, split-toxin effector domain (Fig. 4), this comtoxin would also bear a nonconditional NLS, a P2^* domain — the ligand of nuclear protein P2, and an adjacent degron that can be masked by a complex between P2 and P2^*. This split comtoxin would kill exclusively cells that contain nuclear protein P2 but lack nuclear protein PI, for only in such cells would the comtoxin be long-lived (because a degron is masked by the P2-P2^* complex) and bear an active toxic domain (because the PI '-containing insert between subdomains of the toxic domain is not bound by PI and therefore remains flexible). Cytosol- specific versions of these designs are possible as well. Since other polymers, for example, RNA, can also fold into ligand-binding domains and bear signals such as degrons, nucleic acid-based comtoxins should also be feasible. Having discussed specific examples of comtoxins, it is important to more fully discuss their elements. In a preferred embodiment, comtoxins are amino acid copolymers (e.g., fusion proteins) which include at least three elements: (1) an effector domain; (2) a first codominant intracellular signaling moiety located proximately to a binding domain for a first intracellular protein; and (3) a second 5 codominant intracellular signaling moiety located proximately to a binding domain for a second intracellular protein. There is no strict requirement that the first and second intracellular protein be different. In addition, it should be recognized that when used in this context, the term "intracellular protein" should be understood to encompass intracellular peptides and polypeptides as well. As discussed previously, it is the binding of the first or second protein to their binding site located proximately to the first or second codominant intracellular signaling moiety that determines the selectivity of the comtoxin by influencing the in vivo half-life and/or intracellular localization of the comtoxin. Comtoxins can be produced by any of the known methods for producing an amino acid copolymer having a predetermined sequence identity. However, the preferred method for constructing and producing a comtoxin employs recombinant DNA techniques. Through the use of such conventional techniques, DNA encoding the required comtoxin elements is isolated from a biological source or sources and modified as necessary using techniques such as site-directed mutagenesis.

Preferably the minimum number of nucleotides required to encode an amino acid segment (e.g., a peptide, polypeptide or protein) which confers the required function is employed. Since each of the required elements of the comtoxin molecule is easily assayable, it is a matter of routine experimentation to determine the minimum length of a DNA fragment which will encode a functional comtoxin element. The resulting DNA fragments are linked together, using standard recombinant DNA techniques, yielding an open reading frame which translates into a fusion protein comprising all of the required comtoxin elements. Over the last decade, many amino acid sequences have been shown to be usable as interdomain linkers — sequences that form a hydrophilic and flexible segment of the polypeptide chain relatively resistant to endoproteases present in the bloodstream, intercellular spaces and inside the cells. A particularly extensive collection of suitable linkers emerged from designs of single-chain antibodies, in which a linker sequence connects a light-chain antigen-binding domain to an analogous heavy-chain domain. The open reading frame, prepared as described above, is inserted into a

DNA expression vector which includes a transcriptional promoter and other sequences required for expression. The choice of expression vectors from among the many available options is largely dependent upon the cell type in which expression is desired. As discussed more fully below, eukaryotic expression vectors are preferred for many applications.

Alternatively, a comtoxin can also be produced by expressing it, through the intermediacy of prokaryotic expression vectors, in bacteria (e.g., E. coif), purifying the resulting overexpressed protein, and contacting the purified protein with target cells directly. When produced for use in this manner, a comtoxin should also bear an additional domain that enables its translocation into the cell's cytosol. (For purposes of clarity it is noted that the term "cytoplasm" denotes the interior of a cell outside of its nucleus, while "cytosol" is the cytoplasmic milieu outside of other membrane-enclosed compartments that reside in the cytoplasm.) The use of such "translocation" domains, present in a variety of natural toxins such as, for example, whole ricin, whole Pseudomonas exotoxin A, and whole diphtheria toxin, is widely reported in the prior art. Typically, such reports relate to conventional, present-day chimeric toxins (Vitetta et al. , lmm. Today 14: 252 (1993); Pastan et al , Annu. Rev. Biochem. 61: 331 (1992)). The fundamental, qualitative differences between the current chimeric toxins and comtoxins of the present invention are: (1) the current chimeric toxins are able to recognize surface markers but not intracellular ones; and (2) the current chimeric toxins are in principle incapable of multiple-target, combinatorial selectivity that is characteristic of comtoxins.

An effector domain is a protein or polypeptide which is able to exert a specific effect (e.g., to cause the death of a cell, or its terminal differentiation) when the effector is delivered to a predetermined intracellular location. Thus, the effector domain of a comtoxin can be derived from a protein or polypeptide which acts as a toxin when delivered to a cell. Many such toxins are known in the art, including, for example, the A-chain of ricin (and analogous plant toxins), the toxic domain of the Pseudomonas exotoxin and the toxic domain of diphtheria toxin. As discussed previously, the intracellular compartment specificity of certain toxic proteins or polypeptides can be exploited in connection with comtoxin design in order to alter the selectivity of a comtoxin whose substrates are located in the cytosol but not in the nucleus. Such proteins or polypeptides include, for example, the diphtheria toxin and the Pseudomonas exotoxin A, both of which inhibit protein synthesis by ADP-ribosylating (and thereby inactivating) elongation factor 2 (EF2). Since the bulk of EF2 is cytosolic, the translocation of comtoxin containing a Pseudomonas-type toxic domain from the cytosol to the nucleus would physically separate a toxin from its substrate. An effector domain can be derived from any protein or polypeptide the introduction of which into a predetermined cellular location would cause cell death. For example, a deoxyribonuclease would act as a toxic effector domain if introduced into the nucleus (but not into the cytosol) of a target cell. Indeed, JE-coRI has been shown to cleave nuclear DNA in vivo (in the yeast Saccaromyces cerevisiae), killing the cells (Barnes and Rine, Proc. Natl. Acad. Sci. USA 82:

1353 (1985)). Examples of toxic domains whose substrates are present in both the cytosol and the nucleus, include, in addition to the A-chain of ricin, ribonucleases such as RNAase A and barnase, which have been used to produce conventional chimeric toxins (Rybak et al. , J. Biol. Chem. 266: 21202 (1991); Prior et al , Cell 64: 1017 (1991)). Since the goal of therapy is to render dangerous cell harmless, the set of useful effectors is not confined to cytotoxic proteins. This set includes, for instance, transcription factors whose presence in a cell results in growth arrest and terminal differentiation (Weintraub, Cell 75: 1241 (1993)). In particular, the expression of a muscle-specific transcription factor such as MyoD or myogenin in nonmuscle cells causes them to differentiate into muscle-like cells. Under normal conditions, these proteins are expressed only in the muscle and only at appropriate times. A comtoxin whose effector domain is based on MyoD, myogenin or an analogous transcription factor should be able to cause a terminal differentiation of the targeted cells but not of other cells that receive the drug.

With regard to codominant intracellular signals, a variety of such signals have been described in the literature. For example, a short in vivo half-life can be conferred on a protein by one or more of distinct degradation signals, or degrons (Varshavsky, Cell 64: 13 (1992); Hershko & Ciechanover, Ann. Rev. Biochem. 61: 761 (1992)). The best understood intracellular degradation signal is called the N-degron (Varshavsky, Cell 64: 13 (1991)). This signal comprises a destabilizing N-terminal residue and an internal lysine (or lysines) of a protein substrate. A set of N-degrons bearing different destabilizing residues is referred to as the N-end rule — a relation between the in vivo half-life of a protein and the identity of its N-terminal residue. The lysine residue of an N-end rule substrate is the site of formation of a multiubiquitin chain, which is required for the substrate's degradation.

Ubiquitin is a protein whose covalent conjugation to other proteins plays a role in a number of processes, primarily through routes that involve protein degradation. The recognition of an N-end rule substrate is mediated by a targeting complex whose components include a ubiquitin-conjugating enzyme (one of several such enzymes in a cell) and a protein called N-recognin or E3. A targeted, multiubiquitinated substrate is processively degraded by the 26S proteasome — a multicatalytic, multisubunit protease. Aspects of the N-end rule, ubiquitin fusions and related technologies are the subject of a number of U.S. Patents issued to Varshavsky et al. , including U.S. Patent Nos. 5,132,213; 5,212,058; 5,122,463; 5,093,242 and 5,196,321, the disclosures of which are incorporated herein by reference. The in vivo half-life of a protein bearing a strongly destabilizing N-terminal residue such as arginine can be as short as 1 minute, whereas an identically expressed and otherwise identical protein bearing a stabilizing N-terminal residue such as valine has a half-life of more than 20 hours, resulting in a greater that 1 ,000-fold difference between the steady-state concentrations of these proteins in a cell. Ubiquitin-dependent proteolytic systems (including the N-end rule pathway) share many components of the 26S proteasome. Differences among these systems encompass their distinct targeting complexes, whose recognins bind to degradation signals other than N-degrons. In addition to the N-degron, the amino acid sequences that function as degradation signals and are transplantable to other proteins include, for example, the "destruction boxes" of short-lived proteins called cyclins (Glotzer et al , Nature 349: 132 (1991); Ciechanover, Cell 79: 13, 1994)) and two specific regions of Matα2, the short-lived transcriptional regulator S. cerevisiae (Hochstrasser and Varshavsky, Cell 61: 691 (1990)).

The property of codominance is not confined to degradation signals. For reasons analogous to those considered above for degrons, the signals that confer on a protein the ability to enter membrane-enclosed compartments (Schatz, Protein Sci. 2: 141 (1993); Osborne and Silver, Annu. Rev. Biochem. 62: 219 (1993); and Dingwall and Lasky, Trends Biochem. Sci. 16: 478 (1991)) will function independently of each other if they are present in spatially distinct regions of the same protein. The discussion herein is confined to nuclear localization signals (NLSs) but is also relevant to comtoxins bearing other autonomous signals, in particular other translocation signals.

Proteins smaller than — 60 kD can enter the nucleus by diffusing through the nuclear pores, but the pore-mediated transport of a larger protein requires the presence of at least one NLS accessible to components of the nuclear translocation system. NLSs are short sequences (10-20 residues) rich in lysine and arginine; their steric accessibility in a target protein appears to be sufficient for their activity as nuclear translocation signals. Many NLS-bearing proteins enter the nucleus shortly after their synthesis in the cytosol, but the transport of some proteins is not constitutive: their NLSs have to be "activated", often by unmasking sterically shielded NLSs. One example relevant to the concept of comtoxins is the mammalian protein pi 10, an NLS -containing precursor of the p50 subunit of the transcription factor NF-KB. pi 10 has been shown to remain in the cytosol because a region in the C-terminal half of pi 10 sterically masks an NLS located in the N- terminal half (p50 region) of the same protein. By contrast, the protein p50 — the product of proteolytic processing of pi 10 — bears an active (unobstructed) NLS and is transported into the nucleus. Other instances of subtle and precise regulation of nuclear uptake through the masking of NLSs in the NF-κB family of transcription factors have also been described (Liou and Baltimore, Curr. Op. Cell Biol 5: All (1993); Henkel et al , Cell 68: 1121 (1992); and Beg and Baldwin, Genes Dev. 7: 2064 (1993)). The concept of the present invention is based on utilizing the property of codominance of signals in biopolymers to produce a new class of reagents, termed comtoxins. Comtoxin function is based on the fact that either the metabolic stability (in vivo half-life) of a comtoxin or its intracellular location (for example, the cytosol versus the nucleus) depend on the occupancy of the comtoxin 's binding domains (referred to as P*-type domains in this discussion) (see Figs. 1-3) by their protein ligands in a target cell. This dependence — one key feature of comtoxins — is made possible by the fact that a signaling moiety in a protein is a distinct area of the protein's surface that is recognized by a large (protein-size) targeting complex specific for a signaling moiety. Given the physical bulkiness of targeting complexes, and relatively large areas occupied by signaling moieties on the surfaces of proteins, even a partial obstruction (steric masking) of a signaling moiety in a protein by another protein globule bound nearby would be sufficient to prevent a productive binding of the signaling moiety by a respective targeting complex. Thus, the placing of a P*-type domain of a comtoxin adjacent to a chosen signaling moiety such as, for example, a degron or an NLS, makes possible a steric masking of the signaling moiety upon binding of an intracellular protein ligand to the P*-type domain of comtoxin (Figs. 1-3). While it is not possible to describe definitively and in advance the range of distances encompassed by the terms "adjacent to" or "proximately to", the above consideration indicates that suitable mutual arrangements of a maskable signaling moiety and a P*-type domain will be relatively numerous for each particular comtoxin 's design. It will therefore be a matter of routine experimentation to synthesize fusion proteins in which the protein-binding domain is located at varying distances from the signaling domain. Given the ease with which the functioning of an intracellular signaling domain can be assayed, a number of such fusion proteins can be assayed in parallel to determine the degree to which binding of a protein interferes with the function of the signaling domain. This impairment or inactivation of an intracellular signaling domain due to steric masking results in the alteration of the intracellular fate of the comtoxin molecule.

The P^*-type domains of a comtoxin are selected to bind to specific, predetermined, intracellular ligands (e.g., specific intracellular proteins PI, P2, etc.). Ideally, the set of intracellular ligands of a comtoxin, the set that defines the cell type to be eliminated, should be chosen by carrying out a detailed survey of proteins produced by a targeted cell population. The protein composition of specific cell types, in particular of normal and tumor-derived human cells, is being determined (for example, Hall et al. , Proc. Natl Acad. Sci. USA 90: 1927

(1993)), but the information gathered thus far is incomplete for most cell types. Therefore a useful strategy is to choose the ligands from a set of intracellular proteins that are already known to be either overproduced in the target cells or absent from them; the latter proteins should be present in most (not necessarily all) nontarget cells.

Consider a tumor cell lacking the wild-type version of a tumor suppressor protein (expressed in the normal progenitors of this cell) and overproducing another protein, perhaps as a result of a gene amplification that contributed to the cell's malignant phenotype. For example, many (but not all) human breast carcinomas lack the tumor suppressor protein p53 or contain its functionally inactive variants. Further, many of these tumors overproduce the protein c-Myc (in addition to several other proteins). A comtoxin that kills exclusively cells which contain a PI protein but lack a P2 protein has the requisite selectivity for such a setting. The PI^* domain of this comtoxin would bind to c-Myc, while the comtoxin 's P2^* domain would bind to wild-type p53 (but not to its mutant variant in a given carcinoma). Even nonselective intracellular delivery of this comtoxin would kill Myc-overexpressing, p53-lacking carcinoma cells but spare most, if not all, other cells of the organism, because normal cells contain p53 — the few, if any, normal cell types that lack p53 are unlikely to overproduce c-Myc. Moreover, the selectivity of this comtoxin can be increased further, if necessary, by adding to it a degron- or NLS-containing P3^* domain that binds to a third intracellular protein, P3, chosen to sharpen the description of target cells.

This example illustrates strategies that can be used with other antitumor comtoxins as well. In the cases that have been analyzed in detail, cancer cells were found to differ from their normal progenitors by a combination of at least the following traits: the absence (or decreased production) of certain proteins expressed in the progenitor cells; the presence of mutant variants of normally expressed proteins; and an overproduction of either wild-type or mutant proteins that were not expressed (or expressed weakly) in the progenitor cells. Thus, it should be possible to select an optimal set of intracellular ligands for the designs of comtoxins against specific cancers, against cells infected by a specific virus, and against other undesirable cells as well. Examples of the latter are cells that form atherosclerotic plaques, and cells of neovascular endothelium, whose selective ablation would cut off the blood supply to a metastic tumor.

In selecting protein binding sites, referred to as P^*-type domains, it should be recognized that a homodimerization of the P^*-type domains in a single comtoxin molecule is undesirable, because it would interfere with the function of the comtoxin. Thus, a P^*-type domain should be able to form a heterodimer with its intracellular partner (a PI '-type protein) while not forming a high affinity homodimer. Examples of alternative P^*-type domains include: (1) natural protein ligands of a Pl-type protein; (2) peptide-size fragments of the natural ligand that retain affinity for a Pl-type protein; (3) single-chain antibodies, or fragments of same, which are specific for a Pl-type protein; and (4) nonpepude, low M_r ligands of a Pl-type protein. This latter possibility is confined to directly delivered comtoxins (as distinguished from comtoxins delivered through the intermediacy of expression vectors).

Single-chain antibodies, listed in item (3) above, are described in the prior art (see, e.g., Whitlow and Filpula, Methods 2: 97 (1991); Vitetta et al , lmm. Today 14: 252 (1993) and Pastan et al , Annu. Review Biochem. 61: 331 (1992)). This class of P^*-type domains includes single-chain antibodies to the junctional regions of tumor-specific fusion proteins produced by tumorigenic chromosome translocations. These mutant proteins form early in the evolution of a tumor cell lineage, are likely to be required for the malignant phenotype, and therefore are an especially pertinent class of Pl-type proteins.

The construction of a single-chain antibody against a specific protein or a specific region of a protein includes, first, the production of a monoclonal antibody with a requisite specificity, and, second, the assembly of a DNA fragment containing an open reading frame that encodes a single-chain counterpart of this monoclonal antibody. The latter step is carried out using Polymerase Chain Reaction (PCR) to amplify and fuse in-frame the relevant non-contiguous regions of the two genes that encode the protein-binding domains of the antibody's heavy chain and the light chain in the hybridoma cells that produce the antibody. Both steps of this process (the production of a hybridoma secreting an antibody of requisite specificity and the PCR-mediated construction of a gene encoding a single-chain version of this antibody) are, by now, standard and routine procedures for those skilled in the art. Moreover, since many potential targets of comtoxins (i.e., the proteins PI, P2, etc.) are oncoproteins, tumor suppressors and other mammalian regulatory proteins, the already existing and growing collection of monoclonal antibodies against many of these proteins (and the corresponding hybridoma cell clones that produce these antibodies) can be tapped for the construction of the corresponding single-chain antibodies, making it possible, in many cases, to bypass the step of producing an initial hybridoma. (This latter step has been made routine by the many years of development and perfection of the technology for producing monoclonal antibodies (see e.g., Harlow and Lane (Eds.), Antibodies, A Laboratory Manual (Cold Spring Harbor Laboratory, Cold Spring Harbor, New York (1988)).) Both in vivo and in vitro applications for the comtoxin molecules of the present invention are disclosed. The typical in vivo application is in a therapeutic regimen designed to eliminate dangerous cells or render them harmless (nondividing). Consider, for example, a situation wherein a cancer has been diagnosed in a patient, and a comtoxin whose selectivity is appropriate for targeting the tumor cells of this patient has been designed as described above. Two alternative delivery routes for this comtoxin are (1) intravascular; or (2) expression of the comtoxin within the target cell.

For delivery via the intravascular route (Gilman et al, Eds., The Pharmacological Basis of Therapeutics (Pergamon Press, New York, 1990), a comtoxin should possess not only a toxic domain but also a domain that mediates the translocation of a protein fusion from the cell surface to the cytosol. This aspect of comtoxins is confined to direct-delivery (as distinguished from expression-based) strategies, and is similar to the analogous aspects of current immunotoxins (Vitetta et al , 1mm. Today 14: 252 (1993); Pastan et al. , Annu. Review Biochem. 61: (1992); and Olsnes et al , Som. Cell Biol. 2: 1 (1991)). If necessary, the initial step of a comtoxin 's delivery as a protein can be made partially cell type-selective by fusing the comtoxin to a domain such as, for example, an antibody that binds to a surface marker on target cells. Crucially, and in contrast to the situation with current chimeric toxins, the said surface marker can be present on more that just target cells without significantly increasing the comtoxin's nonspecific toxicity. Indeed, the conditional toxicity of a comtoxin is decided by the environment it encounters after entering the cell; therefore, a comtoxin would not affect cells whose intracellular protein "signatures" differ from the targeted one. One advantage of the direct-delivery strategy is that it bypasses potential problems associated with the vector-mediated delivery of an indelin. One potential drawback of the direct delivery is a larger size of a multidomain comtoxin (due to the presence of additional domains), in comparison to an otherwise identical indelin that is delivered through the intermediacy of an expression vector. Since the testing of comtoxins using either of these delivery strategies is technically straightforward, involves exclusively the existing technologies, and can be assessed directly and objectively, a prudent experimental approach would be to use both strategies in evaluating a given comtoxin, and to compare the results.

Comtoxins can also be delivered into cells through the intermediacy of expression vectors. Both of these approaches are a part of ongoing efforts to improve bioavailability of protein drugs used in medical interventions, from cytotoxic treatments to gene therapies. The problem of insufficient selectivity is common to all of the current cytotoxic strategies: once the effector reaches its intended intracellular compartment, the cell is likely to be killed irrespective of whether it was a target or an innocent bystander. For example, one difficulty with the current immunotoxins is their nonspecific toxicity — largely, but not only, to the liver. This toxicity, which imposes a limit on both duration and intensity of treatments, stems in part from the clearance of an intravenously administered immunotoxin by cells of the reticuloendothelial system (Vitetta et al , 1mm. Today 14: 252 (1993)).

By contrast, even nonselective delivery of the comtoxins of the present invention would not affect most nontarget cells. This feature of comtoxins will yield a much higher therapeutic index (i.e., a much higher tolerated intensity and duration of treatments). If necessary, the initial step of a comtoxin 's delivery can be made selective by fusing the comtoxin to a domain that binds to a surface marker on target cells. Crucially, this marker can be present on more than just target cells without significantly increasing the comtoxin 's nonspecific toxicity. Comtoxins designed for delivery by an expression vector would lack the "compartment-crossing" domain required for their directly delivered counterparts. The vector can be either a retroviral vector, adenoviral vector, another viral vector, or simply naked DNA within a gene delivery system, for example, a liposome-based delivery system. Both viral vectors and liposome-based gene delivery systems have been successfully used in approaches to gene expression in whole animals. Several types of vectors for gene therapy and other applications are already available (reviewed by Yee et al, Proc. Natl. Acad. Sci. USA 91: 9564 (1994); see also Mulligan, Science 260: 926 (1993) and Anderson, Science 256: 808 (1992)). These vectors can be used for the delivery of comtoxins in cell cultures, in whole animals, and, with appropriate preliminary testing, in human patients as well. Recent advances in the design of viral and plasmid-based vectors (see e.g. , Nabel et al , Proc. Natl. Acad. Sci. USA 90: 11307 (1993) and Mulligan, Science 260 926 (1993)) resulted in tailor-made, nonreplicating vectors that can transfect both growing and quiescent cells, and are either specific for cells that bear a predetermined surface marker or almost nonselective. Such vectors, in use for gene therapy and other applications (for example, Mulligan, Science 260: 926 (1993) and Anderson, Science 256: 808 (1992)), are powerful vehicles for the delivery of comtoxins.

Comtoxins can also be used in a variety of in vitro applications. For example, in a bone marrow culture explant of a leukemia patient, comtoxins can be used to selectively eliminate leukemic cells from a sample of the patient's bone marrow (for subsequent reinfusion) without perturbing the marrow's normal cells. A high-dose chemotherapy with bone marrow stem cell rescue is, at present, a standard, and sometimes curative, treatment for a variety of cancers, including leukemias, lymphomas, and, more recently, glioblastomas and other metastatic solid tumors. A common step in all of the current stem cell-rescue protocols involves withdrawing a portion of the patient's bone marrow prior to the marrow- ablating chemotherapy. The resulting short-term in vitro culture of the marrow cells is then treated to eliminate, if possible, any malignant cells present in it before the reinfusion of the purged bone marrow back to the patient whose resident bone marrow has been ablated by a high-dose chemotherapy.

The current methods for selectively killing malignant cells in the marrow cell culture involve treatments with conventional chimeric toxins (Vitetta et al , 1mm. Today 14: 252 (1993)), which comprise a cytotoxic effector domain and a domain (typically antibody-based) that binds to a structure on the surface of target cells. This design of chimeric toxins is often, but not always, sufficient for eliminating most leukemic cells (the bulk of which bears lineage-specific surface markers) from an explant culture of a bone marrow. Unfortunately, many types of tumor cells, including nonhematologic cancers, are much more heterogeneous, surface marker-wise, than certain types of leukemias. Therefore a complete, assured and selective elimination of metastic tumor cells from an in vitro culture of bone marrow prior to its reinfusion remains, in general, a problem to be solved. By contrast, the comtoxins of the present invention, by virtue of their higher, multitarget selectivity and sensitivity to intracellular markers (as distinguished from cell surface ones) should provide an alternative and more selective means for in vitro bone marrow purging of cancer cells.

EXAMPLES

Example 1 : Construction and testing of an indelin

The initial testing of efficacy and selectivity of an indelin will take place not in the whole animals but in the technically less demanding setting of a mammalian (human) cell culture. Using human cell culture allows the subsequent testing of indelins' efficacy and selectivity in experimental animals such as mice, using, for example, thymus-deficient "nude" mice that do not reject human tumor transplants. With cell cultures, the vector-mediated delivery of an indelin is technically straightforward, because several high-efficiency vectors and protocols for short-term or long-term transfection of mammalian cells (including human cells) are available. Briefly, the testing of a [PI ⁺ P2⁺]-specifιc indelin involves determining whether this reagent, once it is introduced (directly or through the intermediacy of an expression vector) in the targeted [Pl⁺ P2⁺] versus nontargeted ([Pl⁺ P2 ], [PI^" P2⁺], or [PI^' P2 ]) cells, will kill exclusively (or nearly exclusively) [Pl⁺ P2⁺] cells.

The indelin to be designed in this Example is a ultidomain fusion whose C-terminal domain is the A-chain of ricin, linked to the nearest "upstream" domain (P2*) by a short (5 to 20 residues) linker sequence. The domain P2* is linked, in turn, to the domain PI*. A simple-sequence, relatively hydrophilic linker GGGSGGGSGGGSGGGS (in single-letter amino acid abbreviations) will be used initially to join these domains. A first and a second degron will be positioned within or near domains PI* and P2* such that the binding of a protein to PI* inhibits the activity of the first degron and the binding of a protein to P2* inhibits the activity of the second degron. The nucleic acids which encode the various elements of the indelin molecule will be isolated from naturally occurring sources and are assembled by conventional techniques within a suitable expression vector such as, for example, the pSG5 vector sold by Stratagene Inc. This vector is a 4.1 kb E. coli- mammalian cells "shuttle" plasmid containing the SV40 early promoter and other relevant features for expression of inserted genes of interest in mammalian cells. For short sequences like the linker sequence, DNA will be made to order by synthetic techniques. The toxic domain employed will be the ricin A-chain (other cytotoxic effectors, for example, the toxic domain of the Pseudomonas exotoxin or the toxic domain of the diphtheria toxin, can also be used to construct an indelin). The deduced amino acid sequence of the A-chain of ricin is provided, for example, in Funatsu et al (Biochimie 73: 1157 (1991)). In this experiment, no effort will be made to reduce the size of the toxic domain by deletion or other methods. Thus, a cDNA-containing fragment encoding the complete ricin A-chain will be isolated from, for example, the plasmid pRAP229 (Ready et al, Proteins 10: 270 1991)), using conventional techniques.

The degradation signals (degrons), to be positioned within or immediately adjacent to the P*-type domains, can be chosen from among several presently known degrons. The degradation signals employed in initial studies will be the N-degron and the degron defined by the cyclin "destruction box". The N-degron has been dissected biochemically and genetically, and is understood in considerable detail (Varshavsky, Cell 69: 725 (1992)). Several portable variants of the N-degron have been described and analyzed (reviewed by Varshavsky, Cell 69: 725 (1992)). The portable N-degron employed in the initial studies will be the one described by Bachmair and Varshavsky (Cell 56: 1019 (1989)). This N-degron comprises a destabilizing N-terminal residue, such as arginine (Arg), followed by the 45-residue sequence derived from E. coli Lac repressor. The corresponding procedures, sequences, and the plasmids, together with their restriction maps, are described in detail by Bachmair and Varshavsky (Cell 56: 1019 (1989)). For the degron defined by the cyclin "destruction box", the sequences involved are encoded by a specific fragment of the plasmid pAG132 (Glotzer et al , Nature 349: 132 (1991)). In the initial experiments, P*-type domains that bind to at least the following intracellular proteins will be used: 1) the E6 protein of an oncogenic human papillomavirus (HPV) such as HPV16; 2) the E7 protein of the same oncogenic HPV. The E6 and E7 proteins of the oncogenic human papillomaviruses are among the preferred initial intracellular targets for the design and testing of comtoxins of the present invention. A strong association between anogenital cancer, especially cervical cancer in women, and specific human papillomaviruses (HPVs) has been established by a number of detailed studies. Cervical cancer alone is responsible for more than 500,000 deaths worldwide annually. The high-risk HPVs include HPV- 16 and HPV- 18, and these are associated with squamous intraepithelial neoplasias which are potentially precancerous. Altogether, more than 90% of cervical cancers can be shown to contain DNA of one of the high-risk HPV types (Scheffner et al. , Curr. Topics Microbiol Immunol 186: 83 (1995)).

Although an HPV genome encodes approximately six proteins, only two of them, the HPV proteins E6 and E7, are consistently expressed in cervical carcinomas, strongly suggesting that the tumorigenic effect of HPV is mediated specifically by these proteins. Recent research has shown that the E7 protein acts by forming a specific complex with a nuclear protein called retinoblastoma protein (Rb) — a regulatory component of the networks that prevent an uncontrolled cell proliferation. Rb protein thus acts as a tumor suppressor. The formation of a complex between the E7 protein of HPV and Rb protein in an HPV-infected cell interferes with the tumor suppressor function of Rb thereby contributing to the malignant transformation of the corresponding cell lineage.

It has also been shown that the second carcinoma-associated HPV protein, E6, specifically binds to another cellular tumor suppressor protein, called p53. The formation of E6-p53 complex not only perturbs the growth-suppressing function of p53 but also results in the accelerated degradation of the E6-bound p53, decreasing p53 concentration and thereby, in effect, inhibiting the function of p53 in more ways than one (Huibregtse et al, EMBO J. 10: 4129 (1991)). No curative treatment of HPV-derived metastatic carcinomas is available at the present time. The reason, common to this an other incurable metastatic cancers (which are a great majority of cancers), is that neither low molecular mass cytotoxic drugs nor the current chimeric toxins, which recognize a marker on the cell surface, are able to distinguish between the carcinoma and normal cells. By contrast, an E6/E7-specific comtoxin (more specifically, indelin) of the present invention would be sensitive to the presence of E6 and E7 inside the target cells, and therefore would spare normal cells which, by definition, lack these HPV-derived intracellular (intranuclear) proteins.

Before describing the construction of an E6/E7-specific indelin, general requirements for PI* and P2* domains of an indelin (these domains are denoted collectively as P*-type domains) are briefly considered. These requirements are as follows: (i) a P*-type domain should bind to an intracellular (cytosolic or nuclear) protein PI or P2 (these proteins, chosen by the designer, comprise the target cell's protein "signature"); (ii) a P*-type domain should not form a high-affinity homodimer; (iii) a P*-type domain should contain, either as a part of the domain or in its immediate vicinity, a sequence (degradation signal) that confers a short half-life on the indelin in the cytosol and/or in the nucleus.

Given these constraints, a suitable P*-type domain can be either a natural protein ligand of the proteins PI or P2, or, for example, a single-chain antibody that binds to a PI or P2. The use of single-chain antibodies as P*-type domains eliminates the problem of homodimerizaton that a natural ligand might be prone to; it also makes possible a standardization of the design of a P*-type domain, since single-chain antibodies against different protein domains have a common "core" antibody structure. Finally, the use of a single-chain antibody as a P*-type domain makes the vector-mediated delivery of an indelin technically straightforward, because no assembly of a conventional (multi-chain) antibody molecule is required with single-chain antibodies. The actual construction of an E6/E7-indelin will proceed via two routes that produce, separately at first, an E6-indelin and an E7-indelin. These single- P*-type-domain indelins can be tested and, if necessary, fine-tuned in human cell cultures that either lack or express E6 and/or E7 proteins of HPV 16 (see below). After that, a more complex, codominance-based comtoxin (E6/E7-indelin) will be constructed from its single-P*-type-domain precursors (E6-indelin and E7-indelin). It should be emphasized that even these "intermediate", single-P*-type-domain constructs in the design of E6/E7-indelin are expected to be useful drugs on their own. Although a ligand-regulated conditional toxin of this type (E6-indelin or E7-indelin) would be sensitive to just one ligand per drug, the ligands themselves are intracellular proteins (E6 or E7). Thus, these single-P*-type-domain indelins (E6-indelin and E7-indelin) will already possess selectivity distinct from, and unattainable by, the current chimeric toxins, which recognize cell surface markers but are incapable, by virtue of their design, of being sensitive to specific intracellular markers. Since HPV-induced carcinomas produce both E6 and E7 proteins, an E7-indelin (or, alternatively, an E6-indelin) may actually prove sufficient for selective elimination of these carcinoma cells, in a cell culture setting and, later, in a human patient. Why then would one construct and use the more complex, codominance mediated, comtoxin-type E6/E7-indelin? The answer, discussed in the following paragraphs, is two-fold.

First, the ability of an E6/E7-indelin to kill exclusively cells that contain both E6 and E7, and to spare cells that lack even one of these proteins makes the comtoxin-type E6/E7-indelin much less perturbable (than a single-P*-type-domain indelin such as E6-indelin or E7-indelin) by the in vivo "noise" of relatively weak and nonspecific protein-protein interactions. For example, if a single-chain antibody that forms a P*-type domain of E6-indelin and binds, with high affinity, to E6 protein (see below), accidentally binds (crossreacts), even with a relatively low affinity, to a normal cellular protein in a normal cell that lacks E6 and E7 proteins of HPV, this crossreaction may adversely affect the desired selectivity of E6-indelin for the E6/E7-containing carcinoma cells. The same would be true of E7-indelin. By contrast, the comtoxin-type E6/E7-indelin would be remarkably more resistant to this selectivity-reducing perturbation because, for it to occur, both the E6-binding and the E7-binding P*-type domains of this E6/E7-indelin must also bind to normal cellular proteins — a much less probable event in any specific cell lineage. Thus, a comtoxin-type E6/E7-indelin is not only more selective than its single-P*-type-domain counterparts, but its selectivity is in addition more robust against perturbations of the kind described above.

Second, it is noted that by focusing on HPV-induced cancer as the first testing ground for comtoxins of the indelin class, one actually chooses a favorable testing field relative to other comparably prevalent and incurable cancers that are not caused by viruses, but are instead the result of (multiple) mutations in specific cellular genes. As discussed above, these cancer cells are often extremely close in protein composition to their normal progenitors, differing from them largely by the concentrations of specific proteins. For example, a cancer cell may overexpress c-Myc protein and several other proteins, but some of these proteins would also be present, at lower levels, in some of the normal cells of the same organism. In other words, the E6/E7 specificity of HPV-induced carcinomas makes less severe demands on the selectivity of a comtoxin than a typical cancer does. Thus, while the choice of HPV-induced cancers as the first testing ground for comtoxins is justified by the current incurability and high frequency of these cancers, it is also suggested by the more crisply defined compositional difference between HPV-induced carcinoma cells and their normal progenitors. For these reasons, the greater selectivity of comtoxin-type indelins (in comparison to their single-P*-type- domain counterparts) will be crucial for attaining the curative level of target selectivity with most non viral cancers. a. Construction of E7-indelin. At the present time, a variety of high-affinity polyclonal antibodies against

E6 or E7 proteins of HPV 16 are available, but a monoclonal antibody has been prepared, thus far, only against E7 protein (Scheffner et al , EMBO J. 11: 2425 (1992)). The genes encoding E6 and E7 proteins of HPV 16 (and several other strains of human papillomavirus) have been cloned by several groups; the corresponding proteins have been overexpressed in E. coli, purified to homogeneity and used as antigens to raise antibodies (Scheffner et al. , Curr. Topics Microbiol. Immunol 186: 83 (1994), and references therein). The production of a single-chain antibody to E7 protein of HPV 16 will utilize the mouse hybridoma cell line 100201 described by Scheffner et al. (EMBO J. 11: 2425 (1992)) that secretes a monoclonal antibody to E7 protein. The methods for producing single-chain counterparts of specific monoclonal antibodies are extensively described in the prior art (see, e.g., Whitlow and Filpula, Methods 2: 97 (1991); Vitteta et al , lmm. Today 14: 252 (1993), and Pastan et al , Annu. Rev. Biochem. 61: 331 (1992)).

A DNA fragment that encodes a single-chain counterpart of the anti-E7 monoclonal antibody 100201 will be produced using Polymerase Chain Reaction (PCR) and purified genomic DNA of the 100201 hybridoma cells to amplify and fuse in-frame the relevant noncontiguous regions of the two genes that encode the E7-binding domains of the 100201 antibody's heavy chain and the light chain. The PCR-mediated construction of a DNA-based open reading frame (ORF) encoding an antibody specific for a hybridoma of interest is a routine procedure for those skilled in the art. The resulting single-chain 100201 antibody-encoding ORF will be linked in-frame, using standard techniques for manipulating recombinant DNA, to a reading frame encoding the A chain of ricin (see above), using the flexible, relatively protease-resistant and hydrophilic linker GGGSGGGSGGGSGGGS — one of the many suitable linkers described in the prior art, in particular in the art of producing single-chain antibodies (Whitlow and Filpula, Methods 2: 97 (1991)). The resulting larger ORF will encode a fusion of the anti-E7 single-chain antibody (a P*-type domain of the indelin being designed) to the toxic domain (A-chain) of ricin, with the latter domain being the C-terminal domain of the fusion. All of these, as well as subsequent recombinant DNA manipulations prior to the placement of a final construct into a mammalian expression vector (see below) will be carried out in the background of the versatile E. cø/t^'-based vector pUC19 (Maniatis et al, Molecular Cloning, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY (1989)).

To position a degradation signal (degron) in proximity to the anti-E7 single-chain antibody domain, one can choose from a variety of different portable degradation signals (see above). For this specific construction, we shall utilize a version of the N-degron described by Bachmair and Varshavsky (Cell 56: 1019 (1989)). However, it must be emphasized that this choice is by no means unique: both other versions of the N-degron and other degradation signals, unrelated to the N-end rule pathway, should be comparably suitable for the designs of indelins. As described above, the N-degron of Bachmair and Varshavsky (Cell 56: 1019 (1989)) comprises a destabilizing N-terminal residue, such as arginine (Arg), followed by the 45-residue sequence derived from E. coli Lac repressor. Both the nucleic acid sequence and the corresponding amino acid sequence of this N-degron have been described by Bachmair and Varshavsky (Cell 56: 1019 (1989)).

Since an N-degron contains a destabilizing N-terminal residue, it should be placed, by definition, at the N-terminus of an indelin fusion construct. To produce a destabilizing residue at the N-terminus of a protein, we shall utilize the ubiquitin fusion technique developed earlier by the author of the present invention (Varshavsky, Cell 69: 725 (1992); US patents issued to Varshavsky et al , including U.S. Patent Numbers 5, 132,213; 5,212,058; 5,122,463; 5,093,242 and 5, 196,321). Ubiquitin is a 76-residue protein essentially unchanged from yeast to humans. Many previously sequenced recombinant DNA plasmids encoding ubiquitin are available (for example, the plasmid pUB23 by Bachmair et al , Science 234: 179 (1986)).

The fully assembled DNA-based ORF of the Ε7-specific indelin will encode the following protein regions or domains, beginning from the N-terminus:

1) ubiquitin;

2) the N-degron described by Bachmair and Varshavsky (Cell 56: 1019 (1989));

3) the 100201 single-chain antibody to the E7 protein of HPV 16 human papillomavirus; and

4) the toxic domain (A-chain) of ricin. b. Construction ofE-6 indelin.

Analogous procedures will be used to construct a DNA ORF encoding an E6-specific indelin, the only significant differences being as follows. First, because no monoclonal antibodies to the E6 protein of HPV are available at the present time (a variety of polyclonal antibodies have been used in the E6/HPV studies thus far), one step in the construction of the E6-indelin will be to produce a panel of monoclonal anti-E6 antibodies, using E6 purified from E. coli overexpressing this protein (Scheffner et al , Curr. Topics Microbiol Immunol 186: 83 (1994)) and routine techniques for producing monoclonal antibodies (D. Lane, Antibodies, Cold Spring Harbor Laboratory, NY (1989)) that are well known to those skilled in the art. The highest-affinity anti-E6 monoclonal antibodies thus obtained will be converted into single-chain (sFv) antibodies (or rather DNA ORFs encoding such antibodies), using standard techniques described above in the protocol of making the E7-indelin.

Second, instead of the N-degron that will be used to construct the E7- indelin (see above), an internal degradation signal such as, for example, the cyclin destruction box (Glotzer et al , Nature 349: 132 (1991)), will be positioned adjacent to the E6-binding single-chain antibody domain. The rest of the construction protocol will be essentially identical to the one described above for the construction of the E7-indelin. The fully assembled DNA-based ORF encoding the E6-specific indelin will encode the following protein regions or domains, beginning from the N-terminus:

1) ubiquitin;

2) the internal degron specified by the cyclin destruction box (described by Glotzer et al , Nature 349: 132 (1991)); 3) a single-chain antibody to the E6 protein of HPV 16 human papillomavirus; and 4) the toxic domain (A-chain) of ricin.

The resulting E6-indelin and E7-indelin will be tested independently, using essentially identical protocols that utilize methods well known to those skilled in the art. Therefore, for brevity, the description below will deal exclusively with the E7-indelin. First, the DNA ORF encoding the E7-indelin will be positioned downstream of a moderately active promoter such as the metalothionein or Moloney leukemia virus promoter within a mammalian expression vector such as, for example, the pE7Mo plasmid (Barbosa et al, EMBO J. 9: 153 (1990)). The resulting E7-indelin-expressing plasmid will be introduced, under identical conditions (see below), into two otherwise identical test cell cultures that differ by the presence of E7 protein. Such pairs of E7-lacking and E7-expressing human cultures have been described, in particular, by Barbosa et al. (EMBO J. 9: 153 (1990)); they can also be produced de novo from any HPV-lacking human cell line by stably transfecting it with the cloned E7 HPV gene, as described previously for E7 and E6 HPV proteins (for example, Kessis et al. , Proc. Natl Acad. Sci. USA 90: 3988 (1993)). High efficiency transient transfection of mammalian cells with plasmid-based vectors is a well-established art. Specifically, we shall employ the ipø ecrin-mediated transfection protocol provided by Gibco/BRL Inc. , which produces this lipophilic, transfection-enhancing reagent. This protocols results in the introduction of a transfecting plasmid into nearly every cell in a cell culture without, at the same time, a significant toxicity or cell death.

This test will immediately indicate whether the initial, "unadjusted" design of the E7-indelin works as expected, because in this case the bulk of cell expressing E7 protein would die upon transfection with the E7-indelin-expressing plasmid, whereas no such effect will be observed upon transfection (with the same plasmid) of the otherwise identical cells that lack E7 protein. Indeed, in the former cells, E7 would interact with the E7-binding single-chain antibody domain of the incoming E7-indelin, masking the nearby degron. The result would be a long-lived and therefore toxic E7-indelin. By contrast, in cells lacking E7, the degradation signal adjacent to the E7-binding single-chain antibody domain would remain active (unobstructed), resulting in a short-lived and therefore nontoxic E7- indelin.

If the above (desired) effect is not observed, adjustments in the design of E7-intralin will be carried out. First, to simplify the adjustment procedure, the region of the DNA ORF encoding the cytotoxic (ricin) domain of the indelin will be (temporarily) replaced, using standard recombinant DNA techniques, with a short DNA region encoding one of several epitopes for commercially available monoclonal antibodies, for example the 9-residue epitope, termed "ha", derived from hemagglutinin of the influenza virus and the corresponding anti-ha monoclonal antibody available from Boehringer Inc. This modification will make it possible to vary the distance between the N-degron and the E7-binding single- chain antibody domain of the E7-indelin without having to deal with the intrinsic toxicity of the indelin's ricin domain. The adjustment protocols, for this or any other indelin, will vary the distance between a degron and an E7-binding domain of the indelin. The resulting constructs, in which the indelin's toxic C-terminal domain such as ricin had been replaced by an epitope tag such as ha (see above), will be expressed in either E7- containing or E7-lacking human cell cultures, using cell lines and expression vectors described above. Pulse-chase analysis of the metabolic stabilities of these constructs (which differ from each other exclusively by the mutual arrangement and proximity of a degron and an E7-binding domain) will then be carried out to determine directly their in vivo half-lives. In a pulse-chase assay, a set of protein molecules labeled with a radioactive tracer (for example, a radioactive amino acid) for a relatively short period of time (a "pulse") is followed by a "chase", which is accomplished through the immunoprecipitation and electrophoretic analysis (and quantitation) of a protein of interest at different times after the termination of pulse. Pulse-chase analysis is a routine assay; its protocol and logic are well known to those skilled in the art (for example, Bachmair et al , Science 234: 179 (1986)).

The aim of the adjustment of an E7-indelin is to produce a construct that is short-lived (as much as possible) in vivo in the absence of E7 protein, and is long- lived (as much as possible) in the presence of E7 protein. The final, satisfactory construct is fused to the cytotoxic domain (see above), restoring the indelin's organization. The resulting E7-indelin is then retested for conditional toxicity in E7-containing and E7-lacking cell cultures as described above. It should be emphasized that even the very first, initial E7-indelin is likely to possess the required qualities. The adjustment procedure would be utilized only if the metabolic properties of an initial E7-indelin are less than satisfactory according to the criteria stated above.

Exactly the same approach will be used to design, test and, if necessary, readjust the E6-indelin. Thereafter, in a nearly final step in the construction of the E6/E7-indelin, a DNA fragment encoding the region of E6-indelin that contains its E6-binding domain and a nearby degradation signal will be excised from the ORF of the E6-indelin and inserted into the ORF of E7-indelin, between the regions encoding the E7-binding domain and the C-terminal cytotoxic domain. The resulting DNA ORF of the E6/E7-indelin will encode the following protein regions or domains, beginning from the N-terminus:

1) ubiquitin;

2) the N-degron described by Bachmair and Varshavsky (Cell 56: 5019 (1989));

3) the 100201 single-chain antibody to the E7 protein of the HPV16 human papillomavirus (see above);

4) an internal degron specified by the "destruction box" of cyclins (see above); 5) a single-chain antibody to the E6 protein of the HPV16 human papillomavirus; and 6) the toxic domain (A-chain) of ricin.

At the present time, a monoclonal antibody is available for E7 but not E6 protein (see above). Therefore one preliminary step in the construction of E6/E7-indelin will be to produce a panel of monoclonal antibodies to E6 protein (which had been purified from E6-overproducing E. coli), to choose an appropriate (highest-affinity) antibody among those produced, and to use the corresponding hybridoma cells as described above to construct an ORF encoding the corresponding single-chain antibody to E6 protein. Both steps of this protocol (the production of hybridomas secreting a monoclonal antibody of requisite specificity and the PCT-mediated construction of an ORF encoding a single-chain counterpart of this antibody) are standard procedures for those skilled in the art. Moreover, since many potential targets of indelins are oncoproteins, tumor suppressors and other mammalian regulatory proteins, there already exists a large (and rapidly growing) collection of monoclonal antibodies against many of these proteins. The hybridoma cell clones that produce these antibodies can be used for construction of the corresponding single-chain antibodies, making it possible, in many cases, to bypass the initial step of producing a monoclonal antibody to an indelin's target. Example 2: Construction and testing of an intralin

Intralins are comtoxins that bear translocation signals, in particular nuclear localization signals (NLSs) (Figure 2). In the experiments of this Example, an intralin specific for cells infected with the Sindbis virus (see below) will be constructed. The P*-type domains of this intralin, termed "nsP2/4-intralin", will be designed to bind the following "early" proteins of the Sindbis virus: a. the nsP2 protein (a site-specific protease). b. The nsP4 protein (RNA polymerase).

These cytoplasmic (cytosolic) proteins of the Sindbis virus — a plus- stranded RNA virus that infects humans and several other metazoan hosts — are among the preferred (initial) intracellular targets for the design and testing of comtoxins of the present invention. The Sindbis virus is a member of the family of alphaviruses. This family has 26 presently recognized members (reviewed by Strauss and Strauss, Microbiol Rev. 58: 491 (1994)). Many strains of the alphavirus family are a serious threat to human health. For example, eastern and western equine encephalitis viruses cause fatal encephalitis in both North America and South America. The Ross River virus and related alphaviruses cause epidemic polyarthritis in humans; the crippling symptoms of this disease can persist for years. The Sindbis virus is nearly (though not entirely) avirulent for humans, and at the same time is closely related to virulent members of the alphavirus family such as the viruses mentioned above. This makes the Sindbis virus a particularly attractive experimental target for the first-generation intralins designed to kill Sindbis virus-infected cells but to spare uninfected cells, because these intralins can be tested and perfected without major precautions against a viral infection of people; at the same time, a Sindbis virus-specific intralin, once designed and optimized, can readily be modified to convert it into intralins against virulent members of the alphavirus family.

The Sindbis virus and other alphaviruses replicate exclusively in the cytoplasm (more specifically, in the cytosol) of infected cells. Given the logic of intralins in Figure 2A, the cytosolic localization of the Sindbis virus' life cycle is one reason for choosing this virus as the first targets for intralins. The Sindbis genomic RNA is 11,703 nucleotides (nt) in length. It comprises two major regions: a nonstructural domain encoding the nonstructural proteins, including RNA polymerase, and a structural domain encoding the three structural proteins of the Sindbis virion. The nonstructural ("early") proteins nsPl-nsP4 are translated as one or two polyproteins from the genomic RNA itself. These proteins are cleaved to produce nsPl, nsP2, nsP3 and nsP4. Two of these proteins will be the targets of the intralin to be constructed. These are nsP2, which is a site-specific protease that processes (cleaves) the Sindbis polyproteins (including the polyprotein of which nsP2 is initially a part) to produce individual proteins of the Sindbis virus, and nsP4, which is the viral RNA polymerase that replicates the Sindbis genomic RNA (reviewed by Strauss and Strauss, Microbiol Rev. 58: 491 (1994)). Both nsP2 and nsP4 are located in the cytosol, are distinct from normal cellular proteins, are produced early in the infection cycle, and therefore are good targets of an intralin whose function would be to selectively kill virus-infected cells before the formation of significant amounts of mature Sindbis virions in these cells. (It should be noted that, although the Sindbis nsP2 protein is a protease, its highly restricted substrate specificity render it nontoxic to mammalian cells and incapable of cleaving most antibodies, which will be used below as domains that bind nsP2 and nsP4).

The methods and protocols for constructing an intralin are highly similar to those used to construct an indelin. Therefore in the description below we shall consider in some detail only those aspects of the construction of nsP2/4-intralin that substantially differ from the steps in the construction of E6/E7-indelin (see above), and will refer to the indelin protocols in describing the other relevant steps. The assembled DNA ORF of the nsP2/4-intralin will encode the following protein regions or domains, beginning from the N-terminus: a) a single-chain antibody to the nsP2 protein of the Sindbis virus; b) the nuclear localization signal (NLS) Pro-Lys-Lys-Lys-Arg-Lys-Val, positioned in proximity to the domain in item a; c) a single-chain antibody to the nsP4 protein of the Sindbis virus; d) the NLS sequence Pro-Lys-Lys-Lys-Arg-Lys-Val (same as in item b above), positioned in proximity to the domain in item c; and e) the toxic domain (A-chain) of diphtheria toxin. Comments on the construction protocol for the nsP2/4-intralin:

1) the NLS to be used in items b and d above is the one present in the large T antigen of the SV40 virus. This strong and at the same time short, portable NLS has been extensively characterized (Dingwall and Laskey, Trends Biochem. Sci. 16: 478 (1991));

2) single-chain antibodies against nsP2 and nsP4 will be produced exactly as described above for the E6/E7-indelin, with the preparation of anti-nsP2 and anti-nsP4 monoclonal antibodies preceding the construction of ORFs encoding single-chain antibodies to nsP2 and nsP4, respectively;

3) the adjustment strategies for the nsP2/4-intralin (if they prove necessary) will be implemented similarly to the adjustment strategies described above for the E6/E7-indelin. Specifically, the position of an NLS will be varied relative to the position of the nearby nsP2- or nsP4-binding domain of the intralin;

4) in contrast to the E6/E7-indelin, which utilizes the ricin-based cytotoxic domain (see above), the nsP2/4-intralin will utilize the A- chain of diphtheria toxin as its cytotoxic domain. The reason for this alteration stems from the logic of intralin design, illustrated in

Figure 2A: a cytotoxic domain of an intralin of this type should be toxic in the nucleus but not in the cytosol. Indeed, unlike the A- chain of ricin, which is toxic in both the nucleus and the cytosol, the A-chain of diphtheria toxin acts by ADP-rybosylating (and thereby inactivating) the elongation factor 2 (EF2). Since the bulk of EF2 is cytosolic, the translocation of a intralin containing the diphtheria-type toxic domain from the cytosol to the nucleus would physically separate a toxin from its substrate; 5) the testing of efficacy of either intermediate designs (nsP2-intralin and nsP4-intralin) or the final construct (nsP2/4-intralin) will be carried out similarly to the testing of the E6/E7-indelin and its precursors, except that Sindbis virus-infected and uninfected human cells will be used as test targets. The cells and viral strain to be used will be the SW-13 line of human cells and the Toto-1000 strain of the Sindbis virus, as described by Li and Rice (J. Virology 63: 1326 (1989)).

Claims

1. A codominance-mediated toxin, comprising: a) an effector domain; b) a first codominant intracellular signaling moiety located proximately to a binding domain for a first protein; and c) a second codominant intracellular signaling moiety located proximately to a binding domain for a second protein.

2. A codominance-mediated toxin of Claim 1 wherein the first and second intracellular signaling moieties are degradation signals.

3. A codominance-mediated toxin of Claim 1 wherein the first and second intracellular signaling moieties are translocation signals.

4. A codominance-mediated toxin of Claim 3 wherein the translocation signals are nuclear localization signals.

5. A codominance-mediated toxin of Claim 1 wherein the first intracellular signaling moiety is a degradation signal and the second intracellular signaling moiety is a nuclear localization signal.

6. An intracellular degradation signal-dependent, ligand-regulated toxin, comprising: a) an effector domain; b) a first codominant degradation signal located proximately to a binding domain for a first protein; and c) a second codominant degradation signal located proximately to a binding domain for a second protein.

7. An intracellular translocation signal-dependent, ligand-regulated toxin, comprising: a) an effector domain; b) a first codominant translocation signal located proximately to a binding domain for a first protein; and c) a second codominant translocation signal located proximately to a binding domain for a second protein.

8. An intracellular translocation signal-dependent, ligand-regulated toxin of Claim 7 wherein the first and second translocation signals are nuclear localization signals.

9. An intracellular translocation signal-dependent, ligand-regulated toxin of Claim 8 wherein the effector domain is cytosol-specific.

10. An intracellular translocation signal-dependent, ligand-regulated toxin of Claim 8 wherein the effector domain is nucleus-specific.

11. A codominance-mediated toxin, comprising: a) an effector domain; b) a cell surface-binding domain; c) a cytosol-penetrating domain; d) a first codominant intracellular signaling moiety located proximately to a binding domain for a first protein; and e) a second codominant intracellular signaling moiety located proximately to a binding domain for a second protein.

12. A codominance-mediated toxin of Claim 11 wherein the first and second intracellular signaling moieties are degradation signals.

13. A codominance-mediated toxin of Claim 11 wherein the first and second intracellular signaling moieties are translocation signals.

14. A codominance-mediated toxin of Claim 13 wherein the translocation signals are nuclear localization signals.

15. A codominance-mediated toxin of Claim 14 wherein the effector domain is cytosol-specific.

16. A codominance-mediated toxin of Claim 14 wherein the effector domain is nucleus-specific.

17. A codominance-mediated toxin of Claim 11 wherein the first intracellular signaling moiety is a degradation signal and the second intracellular signaling moiety is a nuclear localization signal.

18. A method for selectively killing or precluding division of a cell known to contain both a first protein and a second protein, the method comprising: a) providing a DNA expression construct encoding an intracellular degradation signal-dependent, ligand-regulated toxin, comprising: i) an effector domain; ii) a first codominant degradation signal located proximately to a binding domain for a first protein; and iii) a second codominant degradation signal located proximately to a binding domain for a second protein; and b) introducing the DNA construct from step a) into the cell to be killed under conditions appropriate for expression of the intracellular degradation signal-dependent, ligand-regulated toxin.

19. A method of Claim 18 wherein the cell known to contain a first protein and a second protein is a prokaryotic cell.

20. A method of Claim 18 wherein the cell known to contain a first protein and a second protein is a eukaryotic cell.

21. A method for selectively killing or precluding division of a eukaryotic cell known to contain both a first protein and a second protein in the cytosol, the method comprising: a) providing a DNA expression construct encoding an intracellular translocation signal-dependent, ligand-regulated toxin, comprising: i) a cytosol-specific effector domain; ii) a first codominant nuclear localization signal located proximately to a binding domain for a first protein; and iii) a second codominant nuclear localization signal located proximately to a binding domain for a second protein; and b) introducing the DNA construct from step a) into the cell to be killed under conditions appropriate for expression of the intracellular degron-dependent, ligand-regulated toxin.

22. A method for selectively killing or precluding division of a eukaryotic cell known to lack both a first protein and a second protein in the cytosol, the method comprising: a) providing a DNA expression construct encoding an intracellular translocation signal-dependent ligand-regulated toxin, comprising: i) a nucleus-specific effector domain; ii) a first codominant nuclear localization signal located proximately to a binding domain for a first protein; and iii) a second codominant nuclear localization signal located proximately to a binding domain for a second protein; and b) introducing the DNA construct from step a) into a cell under conditions appropriate for expression of the intracellular translocation signal-dependent, ligand-regulated toxin.

23. A method for selectively killing or precluding division of a eukaryotic cell known to contain both a first protein and a second protein, the method comprising: a) providing a multidomain protein fusion comprising: i) an effector domain; ii) a cell surface-binding domain; iii) a cytosol-penetrating domain; iv) a first codominant intracellular signaling moiety located proximately to a binding domain for a first protein; and iv) a second codominant intracellular signaling moiety located proximately to a binding domain for a second protein; and b) bringing the multidomain fusion of step a) in contact with target cells.

24. A method of Claim 23 wherein the first and second intracellular signaling moieties are degradation signals.

25. A method of Claim 23 wherein the first and second intracellular signaling moieties are translocation signals.

26. A codominance-mediated toxin of Claim 23 wherein the translocation signals are nuclear localization signals.

27. A method of Claim 23 wherein the first intracellular signaling moiety is a degradation signal and the second intracellular signaling moiety is a nuclear localization signal.

28. A fusion protein which exhibits conditional, intracellular ligand-dependent toxicity specific for one predetermined intracellular ligand, the fusion protein comprising: a) an effector domain; and b) an intracellular signaling moiety located proximately to a binding domain for a protein.

29. A fusion protein which exhibits conditional, intracellular ligand-dependent toxicity specific for one predetermined intracellular ligand, the fusion protein comprising: a) an effector domain; b) a cell surface-binding domain; c) a cytosol-penetrating domain; and d) an intracellular signaling moiety located proximately to a binding domain for a protein.