NO870630L - STABILIZATION OF UNTABLY DEDICATED REPLICATIONS. - Google Patents

Description

Technical introduction

Genetic stability, at the level of cell populations, as well as at the level of the individual cell, is a universal feature of cell biology.

A larger set of predominantly unknown control mechanisms governs the equal distribution of the cellular chromosomes to daughter cells during cell division, as unequal distribution of the chromosomes will lead to cell death, or at least as far as eukaryotic cells are concerned, to an increased possibility of serious cell disorders.

However, it is not only the distribution of cellular chromosomes that appears to be strictly regulated, as extrachromosomal genetic elements (which can mainly be considered as minichromosomes) of prokaryotic, as well as eukaryotic, cells can be stably maintained in cell populations. The development of plasmids through the exchange of genetic information with host cell genomes may indicate a chromosomal origin of the genetic information that controls the equal distribution of freely replicating molecules, especially those molecules that exist in very few copies per cell. cell or per section (organelle). Such putative chromosomal ancestors need not serve a similar purpose with respect to the problem of chromosome distribution, although this is a tantalizing possibility.

It should be noted that the naturally occurring bacterial plasmids (e.g. R1) which exist in 1-2 copies per bacterial genome, faces a distribution problem of the same order of magnitude as the bacterial chromosome. This means that strict control mechanisms must be in place to ensure the stable maintenance of such plasmids in the population, even in the absence of an external selection pressure that favors the survival of the cells carrying the plasmid.

Control mechanisms for plasmid replication control the plasmid concentration in growing cells, but without an orderly distribution of the plasmid molecules among the daughter cells during cell division, the possibility of cell division leading to a plasmid-free cell is very high, especially for plasmids that exist in only 1-2 copies per cell. bacterial genome. Maintenance functions that control the orderly distribution of replicons have been termed distribution functions.

An alternative replicon-stabilizing function encoded by the ccd locus of plasmid F (Ogura and Hiraga, Proe. Nati. Acad. Sei., 80, 1983, 4784-4788) was shown to involve inhibition of cell division in a situation where only one copy of the F plasmid was present at the time of cell division.

In eukaryotic cells, similar stability functions affecting freely replicating molecules can be encountered in three different situations. First, the genomes of several naturally occurring viruses capable of infecting eukaryotic cells have the potential to exist in a plasmid state, although the natural course of viral infection may prevent this from taking place. Plasmid genomes can be found in cells infected with e.g. Epstein-Barr virus, SV40 and polyomaviruses, papillomaviruses and certain retroviruses, particularly HTLVTII, which is associated with acquired immunodeficiency syndrome and Visnavirus, which originates from sheep. Several of these viruses can be used as gene vectors in eukaryotic cells.

Second, specific regions from eukaryotic chromosomes, e.g. of human origin, are found as naturally occurring extrachromosomally replicating microchromosomes ("double minutes") which are maintained stably during growth, with the number of copies ranging from a few to several hundreds. Using a yeast plasmid that could not replicate in Saccharomyces cerevisiae, autonomously replicating sequences (ars) were identified by inserting into the plasmid fragments of chromosomal DNA from different sources, e.g. yeast, Drosophila and Zea mays (Stinchcomb, D.T. et al, Proe. Nati. Sei., 77. 1930, pp. 4559-4563), and such autonomously replicating sequences can be used to create new replicons for a variety of different cells, but the critical point is the degree to which such replicons can be stably maintained in the cell population without external selection pressure. Third, two types of organelles, mitochondria and plant chloroplasts, are largely self-renewing organelles that contain their own strictly segregated genome, and by organelle division, each newly formed organelle must be supplied with at least one copy of the organelle genome to function. Most mitochondrial species contain from 5-20 genomes, which consequently represents a genome partitioning problem similar to that of low or intermediate copy number bacterial plasmids, and the appearance of genome-free organelles can be expected to occur at a high frequency, leading to the appearance of non- functional organelles if the distribution is not strictly regulated. This expected loss of mitochondria due to different genome distribution can be compensated for by the fact that most cell types contain a large number of mitochondria. A mechanism that ensures the stable maintenance in the population of mitochondria of genetic information introduced into the mitochondria by a gene vector may prove useful in attempts to genetically manipulate certain plant cells (Levings III, C.S. and Fring, D.R., in Genetic Engineering, 1, ed. , J. K. Setlow and A. Hollaender, Plenum Press, 1979, pp. 205-222). In plant chloroplasts, the number of chloroplast genomes varies from 20-80, that is, a situation similar to that of the high copy number plasmids in bacteria, and further, in most plant cells the number of chloroplasts is high. In Chlamydomonas, however, there are only two chloroplasts per cell, and despite the presence of 80 copies of chloroplast genomes in each, a high degree of stability is necessary to ensure the stable maintenance of the chloroplast genome. Genetic manipulations in higher plants may be possible using plant chloroplasts (Bogorad, L., in Genetic Engineering, 1, ed. J.K. Setlow and A. Hollaender, Plenum Press, 1979, pp. 181-204).

The growing interest in vectors engineered to allow the expression of foreign genes in eukaryotic cells makes the control mechanisms involved in genome maintenance in eukaryotic cell populations an important research target.

The strategy developed by the inventors has involved extracting the extensive knowledge of the concepts involved in plasmid stabilization from a well-defined system in E. coli, the R1 plasmid, and attempting extrapolation to 1 ) other replicons,

2) other unrelated bacteria and 3) eukaryotic systems.

The current interest in replicon distribution functions has arisen from instability problems that are often encountered when recombinant DNA techniques are used on an industrial scale.

In prokaryotic systems, the construction of cloning vectors from naturally occurring plasmids has primarily focused on achieving a high copy number which, to some extent, also serves to maintain the plasmid stably in a population, as the possibility of a plasmid-free cell emerging on due to random distribution of plasmid molecules is reduced. It is well documented in prokaryotic, as well as eukaryotic systems, that the insertion of foreign DNA into many otherwise stable cloning vectors can convert these into highly unstable recombinant plasmids. This may be due to a reduction in the copy number, or growth-inhibiting effects of the product(s) coded for by the inserted DNA. Since the possibility of applying external selection pressure, that is to say positively selecting plasmid-carrying cells, is rather limited during cultivation on an industrial scale, it seems to be required to be able to manipulate the stability phenotype of the vector.

A further step towards an improved understanding of the stable maintenance of plasmids in host cells was the discovery that the parB region of plasmid R1 can convert unstable recombinant plasmids unrelated to RI into stably inherited plasmids that are maintained in the E. coli population for hundreds of generations without any external selection pressure (see international patent application PCT/DK83/00086, publication number WO84/01172). The stabilization of plasmids by means of the parB region suggested an applicability of the resulting vectors in large-scale production using genetically engineered bacteria, as no external selection pressure is required to ensure the maintenance of the plasmid during large-scale cultivation.

Description of the invention

The present invention relates, among other things, to replicons that carry genetic information that is responsible for the stable maintenance of the replicon in a cell population, which information is either native to a given replicon or is carried on an inserted fragment or fragments of nucleic acid, as the replicons can also carry a or multiple inserted genes that are not naturally related to the replicons.

The parB stabilization is a very effective replicon stabilization mechanism which has the advantage that the entire mechanism is expressed from a sequence present on one and the same replicon. The inventors have therefore searched for parB-like stabilizing functions among plasmids with the same stability problems as R1, and for putative ancestors of the R1 plasmid stability systems in the chromosomal DNA of S. coli and of other prokaryotic as well as eukaryotic organisms, the ultimate goal being to select or construct from such genetic information, stability-regulating systems that are operative in a given organism. As shown by the original results of this research work, which are described below, stabilization by parB has been found to be an indication of a general and very effective replicon stabilization mechanism that it may be possible to establish in a number of different living cells.

The invention is based on these discoveries and relates in one respect to a replicon which, when contained in a cell, expresses a product capable of killing the cell or its progeny (or a precursor of the product) and which further expresses an antagonist to the killing product ( or a precursor of the antagonist), the antagonist being one that suppresses the killing product (or a precursor of the killing product) in cells containing the replicon, while the aritagonist activity decreases when the replicon is lost from the cells, so that the antagonist (or its precursor) no longer is expressed continuously, which means that the killing product (or its precursor) present in the now replicon-free cell is no longer suppressed by the antagonist, which leads to cell death, with the proviso that the killing product as well as the antagonist for this are not coded for by R1 parB -area when the replicon is a plasmid that can replicate in Gram-negative bacteria.

The parB area itself has previously been described in e.g. international patent application number PCT/DK83/00086, publication number W084/-01172, in connection with the demonstration of the stable maintenance of a number of different plasmids replicating in Gram-negative bacteria. As shown in this patent application, it was assumed that the stabilizing effect could be attributed to a distribution function expressed from the parB region. This distribution function apparently worked by a mechanism that ensured the orderly distribution of the plasmid molecules between the daughter cells during cell division. Recent research has shown that the par3 region is very effective in stabilizing many different types of unstable inherited plasmids which are then very rarely lost from the population in the absence of selection pressure. The stabilizing effect was further shown to be independent of plasmid replication, of the reasons for the instability and, to a large extent, of the bacterial host.

However, further research into the structure and molecular mechanism of parB has surprisingly revealed a new stabilization principle which, as indicated above, includes a cell-killing function and an antagonist therefor. On the basis of hybridization experiments as described below, it is believed that the new stabilization principle is active in most, if not all, types of cells, and has the advantage of being very versatile, meaning that it can be used to stabilize many different types replicon that can be unstable maintained for many different reasons.

In the present context, the term "replicon" denotes a segment of nucleic acid (DNA or RNA) that can replicate autonomously in a given host cell or region of host cells. Thus, the replicon can e.g. be a bacterial plasmid, a bacterial virus, a bacterial chromosome, a eukaryotic plasmid (RNA or DNA), a eukaryotic virus, a eukaryotic autonomously replicating sequence carrying a chromosomal origin of replication, a eukaryotic mitochondrial

DNA molecule or a eukaryotic chloroplast DNA molecule.

For most practical purposes in the technical utilization of the invention, the replicon will not be a native replicon, but will be a replicon containing one or more inserted nucleotide sequences that are not naturally related to the replicon, and which have been inserted into the replicon by which any recombinant DNA technique necessary for the particular replicon and the type of nucleotide sequence to be inserted (the term "inserted" or a gene or fragment of nucleic acid indicates that the gene or fragment of nucleic acid has been introduced into the replicon at some stage during its construction of the final replicon). The inserted nucleotide sequence or sequences can e.g. be a DNA sequence containing a gene which expresses a product which it is desirable to produce in a cell culture, e.g. bacteria containing the replicon, e.g. a plasmid. Such products include a number of different products for medical or technical purposes, in particular polypeptides and proteins or fragments thereof, enzymes and non-proteinaceous products from reactions of enzymes with a compound in the nutrient medium, hormones and other low molecular weight products. Products of eukaryotic genes, especially mammalian genes, are of particular interest. However, the replicon may also be a not previously isolated replicon in its natural state, provided that the replicon is useful either as a vector (into which a non-native gene may be inserted for the purpose of harvesting the relevant gene product from a cell culture containing the replicon containing the inserted gene), or as a production replicon if the replicon itself expressed a valuable product.

A search for regions analogous to par3 in other systems has surprisingly revealed that sequences homologous to parB exist in many different plasmids, bacterial genomes and even in eukaryotic cells, and the molecular mechanism behind the stabilizing effect as described below forms a solid basis for the assumption that a central cellular target, possibly conserved in all cells, reacts with a gene product from these homologous sequences (possibly conserved in all cells) according to the same or a similar principle, so that the parB homologous sequences can be used directly or following relatively simple genetic manipulations leading to the construction of parB-like stabilizing regions, to stabilize any given type of replicon in many different organisms. In the given working examples, the proof for this reasoning is given.

In the present context, the term "precursor" denotes a product which, under the conditions prevailing in the cell at the time in question, either directly or indirectly gives rise to the product whose presence leads to the function in question (either the killing function or the antagonist function).

The toxic substance can be a protein or a precursor thereof, such as a pre-protein that must be modified either by proteolytic cleavage or by the addition of a functional group (including a fatty acid residue) to become toxic, or an mRNA whose translation leads to a toxic protein.

The antagonist can be a gene product, either RNA or protein,

which suppresses the activity of the toxic substance by direct combination with it or by interfering with its targets in the cell, or it can be one that suppresses the turnover of precursors for the toxic product, i.e. pre-protein,

or the translation of mRNA into the actual toxic product.

In most or all living cells, there are genes whose products, when significantly overexpressed, will be lethal or growth-inhibiting for the cell, and which can therefore be used for replicon stabilization in the manner described above.

The mechanism that has now been found to be exerted at the R1 parB region is a killing/antagonist mechanism of the type indicated above. Thus, the mechanism is based on products expressed from two genes present within the R1 parB region, one gene (the R1 parB hok gene, as described below) expresses a product capable of killing the cell, and a second gene ( the R1 parB sok gene, as described below) expresses a product which antagonizes the hok gene product in cells containing a replicon from which the R1 par3 genes are expressed, while on the other hand in the absence of continued expression of the antagonist, e.g. caused by loss of the replicon, the antagonist activity decreases so that the killing product (which has a longer half-life in the cell than the antagonist) is no longer suppressed by the antagonist (ie the translation of the mRNA encoding the killing product is no longer inhibited). It is obvious that this type of mechanism leads to the death of cells that have lost the replicon from which the antagonist function is expressed to ensure that the population of cells does not contain proliferating cells that do not have the replicon, which in turn means that the living and proliferating cells of the population will be cells containing the replicon; in this way, the stable maintenance of the replicon in the population is ensured.

As a result of further research into parB-mediated stabilization, it was surprisingly found that the mechanism governing the stable maintenance of a replicon in a cell population involves the directed expression of a host cell killing function. Stabilization mechanisms involving the controlled expression of a host cell killing function have been described previously, but in all known cases they have involved a separation of the gene encoding the antagonist (e.g. on a plasmid) from the gene encoding the toxic product (e.g. on the chromosome), and the killing function has been limited to one or a very few organisms (see e.g. European Patent Application No. 82306207.0, Publication No. 0.080.848).

According to the present invention, the above stated principle is the basis for e.g. The R1 parB plasmid stabilization function is applicable to a wide range of replicons and consequently in a wide range of host cells, prokaryotic as well as eukaryotic, as a method to achieve stable maintenance of the relevant replicon in a given cell population.

Two aspects of the parB system are very important for 1) the general applicability of this type of stabilization system in living cells and 2) the possibility of constructing stabilization systems corresponding to the parB system that can function in a wide spectrum

organisms:

1. The unexpected discovery of homologous sequences in most/- alie bacteria, as well as in various eukaryotic organisms, which is consistent with the assumption that the target of the lethal function of the hok product is fundamental to most, if not all, living cells. 2. The unique nature of the control mechanism that ensures the survival of all plasmid-bearing cells, and the rapid killing of any plasmid-free cells that occur. Based on genetic evidence, it is believed that the actual sok-encoded regulatory substance is an RNA molecule that prevents translation of hok-encoded mRNA. If one assumes that the sok-encoded RNA is more unstable than the hok-encoded mRNA, the inhibition of the translation of the hok-encoded mRNA will gradually decrease in a cell that has lost the parB<+-> plasmid, and the hok-encoded protein will eventually kill the cell.

On the basis of these assumptions, one thinks that

i) the parB region, or any highly similar region obtained from replicon other than R1, which includes both hok and sok genes, or analogs thereof, should stabilize the maintenance in all types of bacteria of all types of plasmids that is inserted in. This has been confirmed, as many different plasmids have been stabilized more than a hundred times in E. coli after the insertion of parB, and in a bacterium that is phylogenetically very far from E. coli, such as Pseudomonas putida, parB- the region found to be as effective as in E. coli.

ii) Plasmids other than R1 from bacteria other than E. coli may carry sequences largely homologous to parB. It is likely that such sequences can be used in the same way for plasmid stabilization purposes. The thinking behind this is that the hok gene is homologous to a chromosomal RelB-orf3 gene that expresses a product whose activity is identical to that of the hok gene, which supports the theory that the gene has

moved over (transposed) from the chromosome to the plasmid, where it eventually evolved into the parB system. This makes it reasonable to assume that the same process may also have taken place in other replicons. The hybridization experiments strongly suggest that this is in fact the case, as various plasmids unrelated to R1, as well as chromosomal DNA from bacteria other than E. coli carry parB-homologous sequences.

It should be noted that a chromosomal sequence from E. coli that is 55 percent homologous to parB within the hok coding region, yet undetectable by hybridization with the parB probe used, gives a Hok phenotype confirming the above claims .

iii) It will be possible to construct a parB-like system by first isolating the hok gene analogue from either a plasmid or a chromosome, followed by superimposing a control loop similar to that represented by sok in parB; the construct can be performed by inserting a short synthetic DNA sequence (including a promoter) that, when transcribed, expresses a small antisense RNA (antisense RNA) complementary to the portion of hok mRNA or hok-analog mRNA that contains the ribosome binding site. Pairing of antisense RNA to mRNA prevents translation of the mRNA and consequently cell death. Several different such natural regulatory systems (where an antisense RNA prevents the translation of an mRNA) have previously been described, e.g. by Light and Molin, The EMBO Journal 2, No. 1, 1983, pp. 93-98, and artificial systems similar to those proposed above have also been used successfully. Thus, the proposed strategy is logical and realistic and can be applied to e.g. all known bacteria.

According to the explanation given above, the replicons according to the invention can be replicons relating to many different types of cells, such as a bacterial plasmid, a bacterial virus, a bacterial chromosome, a eukaryotic plasmid (RNA or DNA), a eukaryotic virus, a eukaryotic autonomously replicating sequence carrying a chromosomal origin of replication, a eukaryotic mitochondrial DNA molecule or a eukaryotic chloroplast DNA molecule. Similarly, the sequence in the replicon specifying the killing product may conveniently be a sequence from a corresponding system, such as from a bacterial plasmid, a bacterial chromosome, a eukaryotic plasmid, a eukaryotic virus, a eukaryotic autonomously replicating sequence carrying a chromosomal origin of replication, a eukaryotic mitochondrial DNA molecule or a eukaryotic chloroplast DNA molecule.

Similarly, the sequence encoding the antagonist of the killing product may be obtained from a bacterial plasmid, a bacterial chromosome, a eukaryotic plasmid, a eukaryotic virus, a eukaryotic autonomously replicating sequence carrying a chromosomal origin of replication, a eukaryotic mitochondrial DNA molecule or a eukaryotic chloroplast molecule, or it may be a synthetic sequence.

Interesting replicons according to the invention are replicons that carry sequences that operate analogously to the R1 parB system, so that when they are contained in a cell they express a messenger RNA that codes for a product that can kill the cell or its progeny and further express a antagonist that prevents the translation of messenger RNA in cells containing replicons. When the replicons are lost from the cell, the antagonist (or its precursor) is no longer continuously expressed, and its activity decreases, so that the translation of the messenger RNA present in the now replicon-free cell is no longer hindered by the antagonist. Among these interesting replicons are replicons that express an antisense RNA molecule that acts as the antagonist, which through base pairing to at least part of the messenger RNA that codes for the cell-killing product prevents translation of this messenger RNA.

According to what is indicated above, particularly interesting embodiments of the invention are replicons which are DNA molecules comprising a sequence encoding a product (or a precursor of the product) capable of killing the cell or its progeny in a manner which is identical or analogous to that of the product (or precursor) expressed by the R1 parB hok gene. Such replicon further comprises a sequence which codes for an antagonist which functions in a manner which is identical to or analogous to the antagonist expressed by the R1 parB sok gene. Thus, in any replicon of the invention, particularly bacterial plasmids, the R1 parB region may be used as such to ensure the maintenance of the replicon, or it may be suitably modified to ensure proper expression of its hok and sok genes in the relevant host bacteria, e.g. by appropriate changes in the promoters from which these genes are transcribed.

Alternatively, a fragment of nucleic acid expressing a mechanism that ensures replicon maintenance in a population in a manner identical to or analogous to the above-mentioned mechanism can be partially or completely obtained from DNA or RNA more closely related to the host cell in question. A recommended strategy is to select from the genome of a given host cell or from the DNA or RNA of replicons contained in the host cell, those nucleic acid sequences homologous to the R1 hok gene, or to any gene homologous to R1 hok, which when expressed can kill the host cell from which the gene has been isolated. Such sequences can be used in experiments designed to test for the presence of hock- and sock-like activities in the relevant host cell. If both activities are found, this suggests the presence of a useful replicon-stabilizing region within the isolated sequences, while if only a hawk gene-like activity is found, it is possible to construct the necessary regulatory loop (that is, the antagonist) according to the strategy that is suggested above, whereby a potential cell-killing function is converted into a potentially useful replicon-stabilizing function.

The sequence expressing the product capable of killing the cell or its progeny can be inserted separately from the sequence expressing the antagonist, or the two sequences can be present on one nucleic acid fragment, which is the case with e.g. a DNA fragment comprising the R1 parB region. The nucleic acid fragment can be inserted into the replicon by any recombinant DNA technique suitable for the type of replicon in question and for the type of nucleic acid comprising the stabilization function.

One aspect of the invention thus relates to a DNA fragment comprising a sequence that codes for a product that can kill a cell or the progeny of a cell that contains a replicon carrying said fragment (or codes for a precursor of the product), and a sequence which encodes an antagonist of the killing product (or encodes a precursor of the antagonist), the antagonist being one that represses the killing product (or a precursor of the killing product) in cells containing the replicon, while the antagonist activity decreases when the replicon is lost from the cell, so that the antagonist (or its precursor) is no longer continuously expressed, which means that the killing product (or its precursor) present in the now replicon-free cell is no longer suppressed by the antagonist, leading to cell death, with the proviso that DNA the fragment does not comprise the R1 parB region when the replicon carrying the fragment is a plasmid that can replicate in gram-negative bacteria. An interesting embodiment of this page is constituted by a DNA fragment comprising a sequence that codes for a messenger RNA for a product that can kill the cell or the progeny of the cell containing the replicon carrying said fragment (or codes for a precursor of the product ), and a sequence encoding an antagonist that prevents the translation of messenger RNA in cells containing the replicon. When the -replicon is lost from the cell, the antagonist (or its precursor) is no longer continuously expressed, and its activity decreases, so that the translation of the messenger RNA present in the now replicon-free cell is no longer suppressed by the antagonist.

The invention also relates to a living cell containing a replicon as stated above, and a cell culture containing such a living cell. The cell can be a bacterium, a single-celled eukaryotic organism, an animal cell or a plant cell.

Although according to the invention it is most advantageous to provide the killing function and the antagonist for this on the same replicon, as this ensures an easier application of the stabilization principle according to the invention, it is also conceivable that the sequences which code for Hok and Sok respectively the gene products or sequences homologous to these can be inserted on separate replicons in such a way that the stabilizing function according to the invention is expressed when the replicons are introduced into a host cell. In order to obtain correct expression, it may be necessary to suitably modify the promoter sequences of one or both genes. Thus, another aspect of the invention relates to a living cell containing a replicon comprising the R1 parB hok gene or a sequence homologous thereto, and another replicon comprising the R1 parB sok gene or a sequence homologous thereto. It should be noted that in the present context the expression "sequence homologous to this" is to be understood as a sequence which is either homologous to one of the parB genes itself or homologous to any sequence which is homologous to one of the parB genes.

In principle, the hok and sok genes or the homologous sequences can be present on any two different replicons, provided that these can co-exist in compatibility in the same cell. However, the hok gene or the hok-homologous gene is preferably present on a replicon which itself is stably maintained in the cell, such as a chromosome, while the sok gene or the sok-homologous gene is most advantageously inserted on the replicon, whose stable maintenance is desired, that is, typically the replicon that expresses the desired gene product. The replicon carrying the sok gene or the sok homologous gene is therefore usually an extrachromosomal autonomously replicating nucleotide sequence, e.g. a plasmid. Loss of the replicon carrying the sok gene or the sok homologous gene from the cell, so that the sok gene product is no longer continuously expressed, causes cell death according to the principles described above, meaning that only cells containing this replicon are viable .

The invention also relates to a method for producing a stabilized replicon as stated above, which method comprises inserting into a replicon a sequence that codes for a product that can kill a cell containing the replicon or the progeny of the cell (or codes for a precursor of the killing product), and a sequence that codes for an antagonist of the killing product (or that codes for a precursor of the antagonist), the antagonist being one that suppresses the killing product (or a precursor of the product) in cells containing the replicon, while the antagonist activity decreases when the replicon is lost from the cell, so that the antagonist (or its precursor) is no longer continuously expressed, which means that the killing product (or its precursor) present in the now replicon-free cell is no longer suppressed by the antagonist, leading to cell death, with the proviso that the cell-killing product, as well as the antagonist for this, is not coded for wood The R1 parB region when the replicon is a plasmid that can replicate in Gram-negative bacteria.

According to the explanation given above, the sequence or sequences inserted may be a sequence or sequences of DNA (or, where applicable, RNA) originating from a cell related to the replicon, in particular a sequence or sequences which are homologous to the R1 parB region, or the sequence or sequences may originate from the R1 parB region, possibly with modified promoter functions to adapt the sequence to the particular cell type in which the replicon is to be contained.

The invention also relates to a method for producing a gene product encoded by a nucleic acid sequence, which method comprises growing cells containing a replicon as stated above, and harvesting a gene product expressed from the replicon. Cultivation should be carried out in a nutrient medium that is suitable for the relevant cell type, that is, a medium containing the necessary nutrients for this cell type. Breeding will normally be carried out for a large number of generations, e.g. more than 100, to obtain a sufficient production of the valuable product encoded by the nucleic acid sequence, and during this cultivation the loss of the replicon will decrease dramatically due to the replicon stabilizing function indicated here compared to the loss that takes place if no replicon stabilizing mechanism is used. Thus, it is realistic to have a loss of the stabilized replicon of less than 10~<4>cells/generation, normally less than 10~<5>cells/generation, and often less than 10~<6>cells/generation.

In all of the above cases, the invention is based on the ability of nucleotide sequences, such as parB or parB-like sequences, to mediate a phenotype with stable maintenance in a cell population of a replicon carrying the sequences. The stabilization mediating sequences should be inserted into the replicon in such a way that the killing function and the antagonist function which are coded for by e.g. hok or hok-like and sock or sock-like genes are expressed. For them to be effective, it may be advantageous to choose the sequences according to the type of host cell involved, e.g. bacteria, yeast, mold, animal cells or plant cells, and it may be necessary to modify or artificially recreate the correct regulatory loop, e.g. the sock regulation loop or the sock-like regulation loop. The resulting stabilization mechanism will operate according to the principles described above, particularly in a manner analogous to the mechanism explained for R1 parB. Since the central point in the replicon stabilization mechanism is that cells that have lost the replicon are killed by the killing product, i.e. the hog gene product, the present invention has the important industrial utility due to the independence of the described replicon stabilizing mechanism from anything external selection pressure, i.e. the addition of special nutrients, antibiotics or other changes in the cell environment during cultivation on a large scale of cells that include several generations of cell reproduction.

Another important advantage of the present invention is the independence of the described replicon stabilizing mechanism from any particular combinations of host cell and replicon vector that are necessary in the system described in European patent application no. 82306207.0, publication no. 0.080.848. In the technical utilization of the present invention, a wide spectrum of replicon/host cell systems can therefore be considered. The general strategy for forming a replicon-maintaining system, e.g. a system identical to or analogous to the R1 parB system has been described above and will be exemplified as follows.

1. Gram-negative and gram-positive bacteria

Suitable replicons in bacterial host cells can e.g. be plasmids that can replicate in Enterobacteriaceae, e.g. pBR322 or R1 runaway replication plasmids (European patent application no. 83305438.0, publication no. 0.109.150), or capable of replicating in gram-negative bacteria in general, that is, plasmids derived from RSF1010 (Bagdasarian et al, Gene 16, 1981, pp. 237-242), or plasmids that can replicate in Gram-positive bacteria, such as B. subtilis, e.g. pC194 and pUB110 (Lovett and Keggins, Meth. in Enzymol. 68, 1979, pp. 342-357), all of which can be unstable inherited when used as cloning vectors. To stabilize such bacterial plasmids by a parB-like mechanism, a DNA fragment or DNA fragments comprising the R1 parB region can be inserted into the replicon in such a way that R1 hok and -sok expression is transcribed from the native promoters of R1 coupleB. If the R1 hok and -sok genes are not transcribed from these native promoters in the relevant host cell, e.g. due to differences in promoter sequence requirements between E. coli and other bacteria, the native promoter sequences of R1 parB can be replaced by promoters known to be recognized in the relevant host cell, such promoters being either natural promoters or synthetic promoters, such as the 3P02 phage promoter which is active in B. subtilis (Williams et al, Gene 16, 1981, pp. 199-206). If the R1 hok gene product is not lethal to the host cell in question - an absolute requirement to establish the replicon stabilization function for R1 parB - an R1 hok analog sequence can be isolated from either the genome of the host cell in question (or a closely related bacterial species), or from a plasmid that occurs naturally in the host cell in question (or a closely related bacterial species) and then tested for cytotoxic activity in a manner similar to that described in the examples for an E. coli chromosomal homologue of R1 parB, followed by the construction of a control loop that regulates its expression of the newly isolated parB homologue to form a replicon stabilizing system that is specifically adapted to the host cell involved when the constructed piece (construct) is inserted into the relevant replicon.

The use of the hok/sok system or a system analogous to the hok/sok system in bacteria thus includes: a suitable selection of replicon and host cell, insertion into the replicon of the correct sequence comprising a hok/sok system or a system analogous to hok /sok system expressed in the selected host cell, insertion into the stabilized replicon of a gene or genes encoding one or more useful products to be produced on a large scale, introduction of the stabilized recombinant replicon into the bacterial host cell by standard techniques for bacterial transformation, cultivation of the replicon-containing host cells in a culture medium supplemented with the necessary nutrients for several generations, e.g. more than 100 as necessary to achieve the desired cell density, and finally harvest the cells and medium from which any resulting product can be isolated. It should be emphasized that the mechanism of replicon stabilization to which the present invention relates does not require any manipulation of the culture during cultivation to ensure the maintenance of the stabilized replicon in the population, as cells from which the replicon is lost are killed by the hok or hok-like product expressed in the replication-free cell.

2. Yeast cells

The technical utilization of recombinant DNA techniques in eukaryotic systems may be desirable for achieving such post-translational modifications (specific proteolytic cleavages, glycosylation etc.) of primary (eukaryotic) gene products, which modifications are either not made in bacteria at all taken or at best only done in a suboptimal way. A widely used eukaryotic organism is the yeast Saccharomyces cerevisiae, where a naturally occurring plasmid, the 2 \ ± replicon, has been adapted as a vector for the expression of genes that are not naturally related to the 2n replicon in Saccharomyces cerevisiae. Recombinant DNA molecules comprising the 2u replicon with one or more insertions of unrelated DNA fragments are often found to be unstablely inherited in yeast cells. According to the principles outlined above, it is possible to isolate or construct a sequence to be inserted into a yeast replicon, e.g. The 2y. replicon, which sequence uses the principle of replicon stabilization by R1 parB to stabilize the yeast plasmid.

Although the native promoters of the R1 hok and sok genes are unlikely to be functional in Saccharomyces cerevisiae cells, the conservation of hok-like sequences over large evolutionary distances, as suggested by the discovery of hok-homologous sequences in a variety of organisms, that it is realistic to test the product of the R1 hok gene and genes related to R1 hok (e.g. rel3-orf3 or pari, or other genes that originate from bacterial genomes and show sequence homology and the functional level of R1 hok or similar genes isolated from bacterial plasmids) for its ability to kill yeast cells, such as Saccharomyces cerevisiae. In practice, this will involve isolating the coding region of the hok gene or hok-like gene and linking the coding region to a suitable yeast cell promoter, such as the TRF1 promoter (Dobson et al, Nucleic Acids Res. 11, 1983, pp. 2287-2302), with the resulting replicon finally introduced into yeast cells according to standard methods, and the effect of the expression of the hok or hok-like gene examined. If cell death results, a useful hok or hok-like gene has been identified.

Alternatively, sequences identified in yeast DNA from yeast and homologous to parB or reiB-orf3 can be isolated, ligated to an appropriate yeast promoter, inserted into the 2n replicon and, after the recombinant replicon has been introduced into Saccharomyces cerevisiae, tested for their ability to to kill the cell. From a hok gene or a hok-like gene that has been shown to be toxic to yeast, such as e.g. Saccharomyces cerevisiae when expressed, a replicon stabilizing system identical or analogous to the R1 parB system can be generated, by imposition of a regulatory loop (eg a sok gene or sok-like gene driven by an appropriate yeast promoter) as previously described in the discussion of the general strategy. The resulting yeast hok/sok or hok/sok-like sequence can be inserted into any yeast replicon, e.g. The 2u replicon or derivatives thereof, into which genes not naturally related to 2n have been inserted for the purpose of achieving expression of the inserted genes to ensure the stable maintenance of the replicon or derivatives thereof during cultivation. The thus stabilized replicon can be introduced into yeast cells, e.g. Saccharomyces cerevisiae cells, by transformation or protoplast fusion, and after selection, the cells carrying the stabilized replicon can be grown further into a culture on a large scale in a suitable culture medium supplemented with the necessary nutrients. No external selection pressure is necessary to ensure replicon stabilization, as cells from which the replicon is lost are killed by the hok product or the hok-like product expressed in the replicon-free cell. The culture consisting only of cells containing the replicon is then harvested, and any useful product expressed from the replicon can be isolated from either the yeast cells or the culture medium, depending on the nature of the gene and the gene product in question.

3. Mammalian cells

The need for special post-translational modifications may necessitate the expression of certain eukaryotic genes in mammalian cells, that is, of human or animal origin, rather than in bacteria or unicellular eukaryotic organisms. Replicons that can be used as cloning vectors in eukaryotic cells are obtained from chromosomes (ars replicons), from DNA viruses, e.g. SV40 and bovine papillomavirus, or from RNA viruses, e.g. retroviruses. The two DNA viruses mentioned can be maintained in a plasmid state in infected cells, while most retroviruses (RNA-containing viruses) need to be genetically modified to exist as freely replicating DNA molecules rather than chromosomally integrated copies of the viral RNA genome. It can be expected that the same problems with stable maintenance of the replicon involved, as described above, will occur during large-scale cultivation of cells containing the above-mentioned replicons into which genes not naturally related to the replicon have been inserted to achieve expression of a useful product.

A sequence containing a hok/sok system or a system analogous to hok/sok can be constructed in a similar way to that described for the yeast cell system, as soon as a gene that exerts a hok or hok-like effect in the chosen host cell has been identified. A first step would thus be to insert the coding sequence of known hok or hok-like genes regardless of their origin from bacterial genomes or yeast cell genomes into a replicon that can replicate in the host cell in such a way that the expression of the hok gene is obtained by induction of expression. This means that the sequences coding for the hok product or the hok-like product should be supplemented with all necessary regulatory sequences that are necessary for the expression of a gene in the host cell. A promoter sequence suitable for insertion upstream of the hok gene or the hok-like gene can be the tumor virus LTR (long terminal repeat sequence) from the mammary gland of mice, which is inducible with steroid hormones. If cell death takes place when transcription is induced, a hok gene or a hok-like gene has been identified for the relevant host cell, and from this hok gene or hok-like gene a replicon stabilizing system can be constructed by imposing a regulatory loop, e.g. e.g. the sock loop or the sock-like loop as described above.

If none of the available hok genes or hok-like genes of bacterial or yeast origin exert a toxic effect in the relevant mammalian cells, new hok-like sequences can be isolated from a mammalian genome (e.g. the sequences discovered in Tetrahymena mitochondrial DNA and in human cellular DNA) and then tested for hok-like activity when properly expressed. The recommended strategy for the discovery of new hook-like sequences has been suggested above.

The use of the hok/sok-like stabilization mechanism in mammalian cells thus includes: selection of a suitable replicon, e.g. a retroviral vector that can exist as a freely replicating molecule, as has been described for certain retroviruses, and host cell selection depending on the actual sequences that direct the expression of the inserted hok/sok-like genes; insertion into the stabilized replicon of such foreign genes which code for one or more desired products to be produced into the replicon; introducing the stabilized recombinant replicon into the selected mammalian cell line by standard techniques of DNA transfection or microinjection; selection of cells containing the replicon, cultivation of the cells in a culture medium adapted to the cell line involved by addition of necessary nutrients and growth factors with a view to obtaining a large-scale cell culture expressing the gene encoding the useful product; and finally harvesting the culture and isolating the desired product.

It should be noted that for any given fragment shown to express a stabilizing function when inserted into a replicon, it is possible to test whether the resulting stable maintenance of the replicon is attributable to a stabilizing mechanism according to the present invention by the following way: Insertion of the fragment expressing the stabilizing function to be tested into a conditionally replicating vector, such as a rep(ts) vector that can replicate in a given type of host cell, culturing the cells containing the vector under conditions where the vector no longer replicating, but where cell cultivation still takes place, so that the vector is eventually diluted out of the cells, and determination of the number of viable cells compared to a control sample containing the same vector, but not carrying the fragment. If the viability number is significantly lower than for the control sample, as a result of the gradual removal of the antagonist in the vector-free cells, a stabilization mechanism according to the invention has been identified. An example of this test procedure is described in Gerdes et al, "Unique type of plasmid maintenance function: Postsegregational killing of plasmid-free cells", Proe. Nati. Acad. Pollock. USA 83, 1986, pp. 3116-3120.

In connection with the performance of this test, it may also be possible to perform a hybridization test for a possible homology of the fragment to R1 parB or a parB homologue.

Although particular types of replicon adapted to particular cell types have been described in detail above, the general principle of application of the stabilizing mechanism according to the invention is the same, regardless of the type of replicon and cell containing the replicon: the creation on the same replicon of a host-killing function and a antagonist function balanced so that it suppresses the host killing function in cells containing the replicon, thus ensuring the stabilization of the replicon in the cell population.

Description of the drawing

The invention is further described with reference to the drawing, where

fig. 1 shows a deletion map for the parB region. The location of the parA region and parB region within the EcoRI-A fragment of plasmid R1 is shown as black bars. Restriction enzyme sites in the EcoRI-A fragment are described in International Patent Application No. PCT/DK83/00086, Publication No. WO 84/01172. The parB region is located within the 1.9 kb Pstl fragment bounded by coordinates 15.0 to 16.9. The parB region was further mapped to the right of 580 bp of an 880 bp Rsal fragment. The shaded area indicates the smallest parB area. The position of the hok and sok genes within the 580 bp parB region is also shown. A Bglll-Sall fragment containing the / pR promoter and the d857 allele of the A. - suppressor gene was inserted into the pBR322 derivatives carrying different parts of the parB fragment. The position of the inserted fragments and the direction of transcription from ApR are shown below the map of the parB region (arrows). The pR promoters in pKG633, pKG634 and pKG341 read from left to right into the parB region, while A. The pR promoter in pKG171 reads from right to left. Restriction enzyme sites are shown as E (EcoRI), B (Ball), B2 (Bglll), S (Sali), R (Rsal) and P (Fstl).

Fig. 2 shows a map of plasmid pPR95 (13 kb). copA, copB represent replication control genes of plasmid R1, repA represents a gene required for R1 replication, ori is the origin of replication, bla indicates a gene that confers ampicillin resistance on plasmid-carrying cells, parB represents the R1-derived region comprising hok- and sok- the genes that express a maintenance function, deo-lacZ' indicates a translational fusion between the deoC gene and the lacZ gene. lacZ,Y,A represents the lac operon, c1857 represents a gene encoding a temperature-sensitive A repressor that controls Apr promoter activity. Files indicate the direction of transcription. The black bars indicate the extent of the different genes. Restriction enzyme sites are shown as S (Sali), B2 (Bglll), B (BamHI) and E (EcoRI).

Fig. 3a and 3b show the nucleotide sequence of the parB region. The 5<1-> end of the upper DNA strand is arranged on the right. The numbering of the bases is according to the coordinates of the parB region in fig. 1. Ter denotes the stop codon of the open reading frame present in the nucleotide sequence consisting of more than 50 codons. fMet corresponds to the start codon of the open reading frame. The amino acid sequence of the hok gene product starting at position +304 is shown below the DNA sequence - the amino acid abbreviations are standard nomenclature. The sequences designated "-10" and "-35" are the promoter structures of the hok and sok genes. phok and psok denote the hok and sok promoter, respectively. Roman numerals indicate stem-loop structures, with the horizontal arrows on each side indicating the sequence of the stem-loop structures. Vertical arrows indicate the termination of hok mRNA. Fig. 4 shows host cell killing after X. pR-induced activation of the hok gene. Strain JC411 containing either pKG634 (filled symbols) or pKG171 (open symbols) was grown exponentially in A+3 minimal medium supplemented with casamino acids at 30°C. At time 0, the temperature was changed to 42°C and growth of the cultures was monitored as OD450 and viable counts on selective medium (LB plates containing 50 µg/ml ampicillin). Fig. 5 is a photograph of cells sampled 1 hour after the shift of strain JC411 (pKG634) to 42°C. The files point to cells with a clearly changed morphology. Cells with a normal morphology were also seen. Magnification 2000 x. Fig. 6 shows the suppression of host cell killing. Strain JC411 containing either pF634 alone (filled symbols) or pF634 plus pPR633 (open symbols) was grown exponentially in A+B minimal medium supplemented with casamino acids at 30°C. At time 0, the temperature was changed to 42°C, and growth of the cultures was followed by measuring the optical density (OD450) and viable counts on selective medium (LB plates containing 100 µg/kanamycin). Fig. 7a is a comparison of the amino acid sequences of the hok gene product and the relB orf3 gene product. Conserved amino acids are shown in bold, amino acids representing conservative changes are underlined. Fig. 7b shows the alignment of the nucleotide sequences of parB and orf3 of the E. coli relB operon (Bech et al, The EMBO Journal 4, 1985, pp. 1059-1066). the parB sequence is the upper strand, relB-orf3 the lower, coordinates as in fig. 3. Vertical lines indicate conserved nucleotides. The numbers in parentheses are the coordinates of the relB nucleotide sequence as given by Bech et al. The two sequences are aligned so that the start codons of the two reading frames are in the same position - this is indicated by Met at position +304. The termination codons of the two reading frames are indicated by Ter at position +460. Fig. 8a shows 0.75 µg of EcoRI-restricted pooled DNA from strains of E. coli analyzed by filter hybridization using the R1 parB probe. Field 1: R1drd-19, Field 2: R100, Field 3: R386. These fields were exposed for 30 minutes. Lane 4: RP1, lane 5: R6-K, lane 6: plasmid-free E. coli. These fields were exposed for 5 hours. The size of relevant fragments is given in kilobases. Fig. 8b shows 0.75 µg of EcoRI-restricted pooled DNA from strains of E. coli analyzed by filter hybridization using the relB-orf3 probe. Lane 1: R100, Lane 2: R386, Lane 3: plasmid-free E. coli. Time of exposure: 3.5 hours. The size of relevant fragments is given in kilobases. Fig. 9 shows 0.5-0.75 µg Eco-RI restricted pooled DNA from various bacteria analyzed by filter hybridization using the R1 parB probe. The autoradiogram was exposed for 17 hours. Two different photographic exposures of the same autoradiogram are shown. Field 1: Salmonella typhimurium (not mentioned in the text), field 2: Serratia marcescens, field 3: Pseudomonas fluorescens, field 4: Pseudomonas putida, field 5: Proteus vulgaris (not mentioned in the text), field 6: Escherichia coli, field 7 : Bacillus subtilis, lane 8: Bacillus circulans PL236. Sizes of radiolabeled marker (/-restricted by HindIII) are indicated in kilobases. Fig. 10 shows 0.5-0.75 µg EcoRI-restricted pooled DNA from various bacteria analyzed by filter hybridization using the relB-orf3 probe. The autoradiogram was exposed for 17 hours (field 19) and 72 hours (fields 2-7). Lane 1: Serratia marcescens, Lane 2: Pseudomonas fluorescens, Lane 3: Pseudomonas putida, Lane 4: Bacillus subtilis, Lane 5: Bacillus circulans PL236, Lanes 6, 7: Lactobacillus. Sizes of radiolabeled marker (X-limited by HindIII) are given in kilobases. Fig. 11 shows filter hybridization analyzes of DNA from eukaryotic cells using the relB-orf3 probe (lanes 1-4), as well as the R1 parB probe (lanes 5-6). DNA was cleaved with EcoRI (lanes 1-3 and 5-6) or with Pstl (lane 4). Lane 1: 5 µg macronuclear DNA from Tetrahymena thermophila, Lane 2: 5 µg total DNA from Tetrahymena thermophila, Lane 3: 0.25 µg chloroplast DNA from Pisurn sativum, Lane 5: 5 µg total cellular DNA from neuroblastoma, Lane 6: 10 µg pooled cellular DNA from embryonic liver. Sizes of fragments are given in kilobases. Fig. 12 shows a map of the wide host range of plasmid pFN13 parB<+>. ori indicates the origin of replication, mob indicates the sequence that allows the plasmid to be transferred by conjugation to most other Gram-negative species when a conjugative (that is, mobilizing) plasmid is present in the same cell, aphA represents a gene that confers kanamycin resistance, lacZ is the gene that codes for G-galactosidase, p symbolizes the repA promoter from plasmid R1, parB represents the R1-derived plasmid maintenance function that codes for the hok and sok genes. Restriction sites are shown as P (Pstl), E (EcoRI), H (Hindlll), B-| (BarnHI) and B2 (Bglll). Fig. 13 is a model of the regulation of the hok gene. During cell division, one of the daughter cells receives all the parB<+->plasmids present. A. Division cycle of a cell carrying a parB<+->plasmid (hok<+>, sok<+>). Both the hok and sok gene products are produced continuously from the plasmid. The sok product counteracts the hok product, and no effect on cell viability is observed (1a+2a). B. Division cycle of a plasmid-free cell. In the plasmid-free cell, the hok and sok products are present. The sok product breaks down faster than the hok product, which leads to activation of the hok product (3b). This activation leads to damage of a vital cell function and cell death (4b). Fig. 14 shows restriction maps of the translation and transcription fusion vectors based on pJEL124 (A) and pJEL126 (B), where c1857 denotes the gene for a temperature-sensitive repressor controlling XpR promoter activity, oriR1 denotes the origin of replication, bla denotes a gene conferring ampicillin resistance on plasmid-bearing cells, lacZ,Y,A indicate the laz operon, the arrows show the direction of transcription, the black bars indicate the extent of the insertion from the parB region, the shaded areas indicate the hok gene or part of it, the open unshaded areas indicate sok- the gene or part thereof, sokT indicates the target of the Sok gene product and restriction enzyme sites are shown as E-| (EcoRI), B-| (BarnHI) and S (Sali).

Materials and methods

Bacterial strains and plasmids

The following E. coli K-12 strains were used: CSH50 ((lac-pro) rpsL), JC411 (metB leu his argG lacY malA xyl mtl gal rpsL), LE392 (Murray et al, Mol. Gen. Genet. 150 , 1S77, p. 53) and HB101 (F. Bolivar and K. Bachman, Methods Enzymol. 68, 1979, p. 245). JC411 was mainly used in the physiological experiments. CSH50 was mainly used for the genetic experiments. Furthermore, the B. subtilis strain BD 224 (trpC2, thr-5, rec-4)

(Dubnan & Cirigliano, J. Bact. 117, 1974, p. 488) used for

experiments with gram-positive bacteria. The plasmids used are listed in Table 1. Bacteriophage X. 47 (W.A.M. Loenen & W.J. Braunner, Gene 8, 1979, pp. 69-80) was used for plaque hybridization.

The experimental techniques used were standard techniques used in the fields of microbial genetics (J. Miller: Experiments in Molecular Genetics, Cold Spring Karbor, New York, 1972) and genetic manipulation (Davis, Botstein and Roth: A Manual for Genetic Engineering; Advanced Bacterial Genetics, Cold Spring Karbor, Nev/ York, 1980, and Maniatis, Fritsch and Sambrook: Molecular Cloning, Cold Spring Harbour, New York, 1982).

All cells were grown in LB medium (Bertani, J. Bact 62, 1951, p. 293) with 0.2 percent glucose and 1 ug/ml thiamine, or A+B minimal medium (Clark and Maaløe, J. Mol. Biol. 23, 1967, p. 99) supplemented with 0.2 percent glucose and 1 percent casamino acids. The plates used were LA plates containing LB medium and 1.5 percent agar.

McConkey lactose indicator plates were prepared as recommended by the manufacturer (Difco) and X-gal plates were prepared by adding 20-40 µg/ml 5-bromo-4-chloro-indolyl-S-D-galactoside to A+B minimal medium supplemented with 0.2 percent glucose and 1 ug/ml thiamine.

Physical-chemical methods

Clear lysates were prepared according to the method described by Clewell and Helinski, Proe. Nati. Acad. Pollock. USA 62, 1969, pp. 1159-66.

Small-scale production of plasmid DNA was carried out by the method of Birnboim et al, Nucl. Acids Res. 7, 1979, pp. 1513-23.

Large-scale production and analysis of plasmid DNA was performed using dye buoyant density gradient centrifugation according to Stougaard and Molin, Anal. Biochem. 118, 1981, p.181. The restriction endonucleases were used according to the instructions provided by the manufacturer (Boehringer, Mannheim or Biolabs, New England) at 37°C. Double and triple digests were performed by starting with the enzyme that required the lowest salt concentration and then adjusting with additional buffer before adding the next enzyme.

Treatment with exonuclease Bal31 was performed as follows: 0.1 unit of Bal31 was added to 50 µg linear DNA and samples were taken at 1, 2, 4, 8, 16, 32 and 60 min. to 60 mM EDTA, extracted with phenol, ethanol precipitated and resuspended in 20 µl TE buffer. Half of the 20 µl was digested with the appropriate restriction enzyme, which was subjected to agarose gel electrophoresis to determine the average size of the deleted DNA fragments. To the other half, the appropriate linker was added and the mixture ligated for 48 hours with an excess of T4 DNA ligase.

Ligation of restricted plasmid DNA was performed as recommended by the manufacturer, with the exception of blunt-end ligation, where an excess of T4 DNA ligase and ATP was added.

Measurements of S-galactosidase expressed from fusion plasmids were made from exponentially growing cultures, largely as described by Miller, op.eit. Activity units were expressed as OD420 per minute OD450 per ml culture. Rifampicin (Ciba-Geigy) was added at high concentrations (300 µg/ml) to the non-permeable strain CSH50. Induction of the lac operon by addition of IPTG (1 mM) to an exponentially growing culture of CSH50 (F'lac) and simultaneous addition of rifampicin (300 µg/ml) showed that LacZ synthesis stopped within the first minute after the antibiotic was been added.

For the preparation of RNA, culture samples (10 ml) were removed and rapidly cooled to 0°C, spun in a Sorvall SS34 rotor and resuspended in TE buffer containing 0.2 percent SDS, phenol-extracted twice, chloroform-extracted three times, ethanol precipitated and resuspended in 200 µl TE buffer.

Mapping of RNA endpoints using S1 nuclease was performed largely as described by Maniatis et al, op.eit. DNA-RNA hybrids were treated with an excess of S1 enzyme (100 units) at 20°C overnight. End labeling of DNA was performed by standard methods (Maniatis et al. op.eit.)

The calculation of mRNA half-lives from the rifampicin induction curves was performed according to M.L. Pato, P. Bennett, K. von Meyenburg, J. Bact. 116, 1973, pp. 710-718. The remaining capacity of protein synthesis (expressed in B-galactosidase units) which is directly proportional to the concentration of mRNA, (that is, hok-lacZ hybrid mNRA) in the cell, was determined as the distance from the induction curve to the terminal plateau (which was reached after an infinite time ). These protein synthesis capacity values were plotted on a semi-logarithmic scale against time. This type of plotting always gave straight lines, from which the half-life of the mRNA could be read.

Construction of inducible parB derivatives (see also number 1) pKG633: The SalI-BglII fragment of pGU82 containing the cl 857 temperature-tolerant allele of the A. repressor gene and the X pR promoter was inserted into pPR633 in front of the parB region, so that ApR- the promoter reads into the region from left to right (Fig. 1). In an analogous manner, the SalI-BglII fragment of pOU82 was inserted into pPR634 and pPR341, which are Bal31 deletion derivatives of pPR633, leading to pKG634 and pKG341. pKGI71: In pPR171, the SalI-Bglll fragment of pOU82 was inserted in the opposite orientation, leading to pKG171. The positions and orientations of the inserted ApR promoters in relation to the hok and sok genes are shown in fig. 1. pF634: The EcoRI-SalI fragment of pKG634 containing the right 390 bp of the par3 region and the Xd857-pR inducible promoter system was inserted into the unique SalI site in the kanamycin resistance fragment (aphA<+>) of pML31 (D.R. Helinski , J. Bact. 127, 1976, pp. 982-937) by blunt-end ligation (S1 nuclease was used to blunt the ends of the restricted DNA fragments).

The R1-derived translational fusion vector pJEL124 (shown in Fig. 14) containing a polylinker in front of the lacZ gene lacking the first 7 1/3 codon (lacZ') was used for all translational fusions between the hok gene and the lacZ gene. The copy number for the R1-based transcripts and translational fusion vectors used here is approximately 1 per E. coli genome.

Plasmids pKG733 and pKG732: pKG733 was constructed by inserting the BamHI-Sau3A restriction fragment of pPR633 (+1 to +342) into the BamHI site of pJEL124. The Sau3A site at +342 is arranged within the hok gene and positioned so that the fusion of lacZ to this site forms a hybrid reading frame consisting of the first 14 amino acids of the hok gene, followed by the lacZ' coding region. Thus, the expression of β-galactosidase activity from the translational fusion plasmid pKG733 is controlled by signals that normally control the expression of the hok gene. Plasmid pKG732 was constructed by inserting a 190 bp EcoRI-BamHI fragment of pSKS106 containing the lacUV5 promoter in front of the parB-derived insert of pKG733.

Plasmids pKH734 and pKG735: pKG734 was constructed by inserting the 3amHI-Sau3A fragment of pPR634 (+194 to +342) into the BamHI site of pJEL124. Plasmid pKG735 was constructed by inserting the 190 bp EcoRI-BamHI fragment of pSK3106 containing the lacUVo promoter in front of the parB-derived insert of pKG734.

Plasmids pKG741 and pKG742: pKG741 was constructed by inserting the BamHI-Sau3A fragment of pPR341 (+268 to +342) into the BamHI site of pJEL124. Plasmid pKG742 was constructed by inserting the 190 bp EcoRI-BamHI fragment of pSKS106 in front of the parB-derived insert of pKG741.

Plasmids pKG780 and pKG782: A BamHI-BclI restriction fragment of pKG733 encoding the sok gene and the hok' gene joined in frame to the lacZ' gene was inserted in the two possible orientations into the BamHI site of pPR345, a pBR322- plasmid derivative containing the parB region extending from +350 to +580 on the BamHI-EcoRI restriction fragment (Fig. 7).

Promoter-fusion plasmids: In all the promoter-fusion engineered pieces between the hok and sok promoters and the lacZ gene, the R1-based transcripts were inserted into the fusion vector pJEL126 (see Fig. 14), which contains EcoRI and BamHI restriction sites in front of the intact lacZ gene used.

Plasmids pKG730 and pKG736: The EcoRI restriction fragments of pKG732 and pKG735 were inserted into the EcoRI site of pJEL126, leading to pKG730 and pKG736, respectively.

Plasmid pUC341: the BamHI-EcoRI restriction fragment +268 to +580) of pPR341 was replaced with the small EcoRI-BamHI restriction fragment in front of the lacZ gene of pJEL126, leading to pUC341.

Microbiological methods

Distribution Experiment I: The construction of Lac<+->vectors made it possible to determine the Par<+->phenotype of a plasmid simply by plating on non-selective McConkey lactose plates or X-Gal plates. Bacteria (Alac) containing these plasmids give a Lac<+->phenotype which can be easily counted as colored colonies on the indicator plates, while plasmid-free cells have a Lac~ phenotype and appear as colorless colonies.

Distribution Experiment II: (Used for Lac~ plasmids): A colony from a selectable plate (a plate containing an antibiotic) was streaked onto another selective plate. From this plate, a colony was streaked onto an LA plate to form single colonies. From the LA plate, approx. 10 colonies suspended in 1 ml of 0.9 per cent NaCl to a dilution of 10~<4> and 10~<5> respectively. o.1 ml of the diluted mixtures of 10~<4> and 10"" 5 bee spread on LA plates. From these plates, the resistance pattern of 50 colonies (200 colonies if slight instability is expected) was tested on the correct selective plates. The loss frequency (LF value) is then calculated on the basis of the formula

LF=1-( )<<1>/<27>>

where v is the frequency of plasmid-bearing cells assuming that one colony is grown for 27 generations. There is a large statistical fluctuation in this procedure.

Distribution experiment III: Quantitative measurements of the stability of Lac<+->and Lac~ plasmids. One complete colony was taken from a selective plate and resuspended in 1 ml of 0.9 percent NaCl to a concentration of 10 3 cells/ml. 2 x 0.1 ml of the 10~<3> dilution was used to inoculate 2 x 10 ml of LB medium, and the inoculation was carried out at 30°C with shaking. At a cell density of approx. 5x10^ cells/ml, the cultures were diluted 10^ and 10^ times. 0.1 ml of the 10<4> dilution (5 x 10<3> cells) was used to inoculate 10 ml of fresh LB medium and 0.4 ml of the 10^ dilution was spread on McConkey lactose plates, and the plates were inoculated at 30°C overnight. The dilution from 5 x 10^/ml to 5 x 10<2>/ml corresponds to 20 generations of cultivation (2^0), so that the change in the frequency of plasmid-bearing cells from one dilution to the next corresponds to that which takes place during of 20 generations of cultivation. More generally, the LF value can be calculated as follows:

1 = (1 - LF)91 and2= C"LF)<g>2

where1 and 2 are the frequency of plasmid-bearing cells after g-|, respectively and g2 generations, and LF is the frequency of the loss per cell per generation. Consequently, becomes

-]/20 (1 - LF)S1-92

and

LF = 1 - (v-i/v2) (1/(g1~92) )

Using this formula avoids errors due to fluctuations in the number of seeding cells at time zero. A more suitable simplification of the formula is

LF<=><in>( yl 2)/(<g>2-<g>i)

Incompatibility attempts

Plasmids presumed to carry an inserted pair region were subjected to sorting using the observation that two otherwise compatible replicons carrying the same pair region are incompatible with each other, leading to the loss of one or the other of the two in absence of selection pressure. The experiment is carried out by transforming the plasmid to be examined into a bacterial strain carrying another plasmid, selecting for both plasmids on double-selective plates. After plating on a double-selective plate (a plate containing two different antibiotics), the incompatibility was measured either qualitatively or quantitatively.

For the qualitative incompatibility assay, a colony from the double-selective plate was streaked onto an LA plate to form single colonies. About. 10 colonies from this plate were resuspended in 1 ml of 0.9 per cent NaCl and diluted to 10-<4> and 10-<5> respectively. 0.1 ml of the 10-<4-> and 10_<5-> dilutions were spread on LA plates. From these plates, 50 colonies (or 200 colonies if a slight incompatibility was expected) were tested on the correct selective plates. If a Lac<+> plasmid was incorporated into the experiment, McConkey lactose indicator plates were used instead of replicate plating on selective plates. In the case of Lac<+->plasmids, the sorting was thus performed by transformation of suspected Par<+->plasmids into a strain that already contained a Par<+->hybrid plasmid that gave a Lac<+->phenotype. Plasmid compatibility and therefore evidence for a particular Par<+> phenotype of the incoming plasmid is readily detectable by sorting for Lac~ colonies on McConkey plates, showing that the resident Par+ plasmid had been destabilized. An example of this sorting method is described in example 4.

Quantitative incompatibility assays were performed by measuring the loss frequencies of Lac<+> plasmids after establishment of hetero-plasmid populations as described above. The LF values were measured as described under "Distribution test III".

Purification of chromosomal DNA

Total DNA was extracted from bacteria as follows. Cells were harvested by centrifugation, washed twice in 1x TEN buffer (TEN = 10 mM Tris (pH 7.5), 1 mM EDTA, 0.1 M NaCl) and resuspended in 1/10 volume of TEN containing 1 mg /ml lysozyme. After incubation at 37°C for 30 minutes, the protoplasts were lysed by addition of sodium dodecyl sulfate to a final concentration of 1 percent, and proteinase K was added to 0.25 mg/ml. The lysate was incubated at 37°C for 2 hours and then extracted twice with buffered phenol and three times with chloroform. Sodium acetate was added to 0.3 M, and DNA was precipitated by adding 1 volume of isopropanol. The precipitate was washed several times in 96 percent and 80 percent ethanol. Finally, DNA was dissolved in 1 mM Tris, 1 mM EDTA.

Total DNa from Tetrahymena thermophila BVII was prepared according to Nielsen, H. and Engberg, J.: Biochem. Biophys. Acta 825, 1985, pp. 30-33. Macronuclei from Tetrahymena thermophila BVII were isolated (Cech, T.R. et al: Cell 27, 1981, pp. 478-496) and DNA extracted (Maniatis et al, 1932, op.eit., pp. 280-231). rDNA from Tetrahymena thermophila BVII was prepared as described by Engberg, J. et al: J. Mol. Biol. 104, 1976, pp. 455-470.

Chloroplast DNA from Pisum sativum was isolated according to Bookjans, G. et al: Anaiyt. Biochem. 141, 1984, pp. 244-247.

Embryonic liver tissue from a 7-week legal abortion was ground up in physiological saline, and DNA was prepared according to Maniatis et al, 1982, op.eit., pp. 280-231. Similarly, DNA was isolated from a tumor biopsy from a case of neuroblastoma, the isolated DNA was found to contain a several hundredfold enlarged chromosomal region, and similarly the tumor cells were found to contain a number of extrachromosomal minichromosomes by microscopy of meiotic cells.

Isolation of DNA fragments for radioactive labeling

100 micrograms of pPR35 and pBD2724 DNA were digested with EcoRI and EcoRI and HindIII, respectively. The fragments were separated by electrophoresis through a 1 percent agarose gel in Tris-borate buffer at 5 volts per cm for 3 hours. The desired fragments were isolated by

electroelution on top of an NA45 membrane (Schleicher & Schull) according to the manufacturer's guidelines. After recovery of the fragments by elution of the filter in 1.5 M NaCl at 65°C, the fragments were again subjected to purification by agarose gel electrophoresis and NA45 membrane recovery from the gel.

Agarose gel electrophoresis

DNA was cleaved with the appropriate restriction endonucleases according to guidelines provided by the manufacturers. For cellular DNA, 10 units per micrograms of DNA used. The incubation time was 3 hours at 37°C. The generated DNA fragments were separated by electrophoresis through 0.7 percent or 1 percent agarose gels in a Tris-acetate buffer at 0.8 volts per cm for 18 hours and made visible by staining with ethidium bromide.

A molecular weight marker was prepared as follows: wt ΔDNA was restricted with HindIII and end-labeled using the Klenow polymerase from Boehringer, Mannheim, as recommended by the manufacturer in the presence of α-32P-dCTP plus non-radioactive dATP and dGTP. When used as a molecular weight marker, the amount of Tetrahymena macronuclear DNA was added, corresponding to the DNA load of the experimental fields.

Transfer of the DNA fragments from gel to nitrocellulose filter After partial depurination in 0.25 N HCl for 15 minutes at room temperature, denaturation of DNA in the gel, neutralization and subsequent transfer of DNA from gel to a BA35 (Schleicher & Schull) nitrocellulose filter was carried out as described in Maniatis et al, 1982, op.eit., pp. 280-281. The completeness of the transfer was ensured by staining the gel with ethidium bromide after transfer.

Production of radioactively labeled probe

0.3 microgram of the 900 bp parB fragment and 0.3 microgram of the 300 bp relB-orf3 fragment were radiolabeled by nick translation (Maniatis et al, 1982, op.eit.) using 0.25 micromolar a -32P-deoxycytidine triphosphate (3000 Ci per mmol). The unincorporated radioactive precursor was removed using repeated ethanol precipitations.

100 micrograms of E. coli tRNA was added to each preparation as a carrier.

The special activities of the probes were respectively 2-3 x 10^ and 4-5 x 10^ dprn per micrograms of parB and relB-orf3 fragment.

Hybridization

Filters containing DNA transferred from agarose gels were pre-incubated in plastic bags with the hybridization solution (10 ml per 120 cm<2>) for 18 hours at 37°C with constant shaking. The hybridization solution was modified from Wahl et al, Proe. Nati. Acad. Pollock. 76, 1979, pp. 3683-3687 and contained 38 percent deionized formamide, 0.75 M NaCl, 50 mM sodium phosphate, and 10x Denhardt<1>'s solution (50x Denhardt's solution is 0.2 percent bovine serum albumin, 0 .2 percent polyvinylpyrrolidone and 0.2 percent Ficoll).

After preincubation, the denatured radiolabeled probes were applied to appropriate filters. In experiments using the par3 probe, the fragment concentration during hybridization was 3 ng/ml, while the relB-orf3 probe was used at a concentration of 1.3 ng/ml to obtain equimolar concentrations of complementary sequences in the two situations.

Hybridization was performed at 37°C with gentle shaking for 19 hours.

The hybridized filters were washed once for 20 minutes at room temperature in 0.4 x washing buffer, and finally twice for 30 minutes at 60°C in 4 x washing buffer. The wash buffer contained 0.6 M NaCl, 0.1 percent SDS, 0.1 percent sodium pyrophosphate, 50 mM sodium phosphate, pH 6.5. Autoradiography was performed using X-ray films and intensifier screens. The exposure times are indicated in the description of the figures.

The term "filter hybridization assay" is used in the examples to indicate the following sequence of operations: agarose gel electrophoresis of DNA fragments, transfer of the fragments to nitrocellulose filters, hybridization with the appropriate radiolabeled probe, filter washing, and autoradiography of the filter after washing. The data shown in the examples represent autoradiograms obtained by filter hybridization analysis.

The term "homology" is used herein to denote the presence of any degree of complementarity between a given probe and the nucleic acid materials being analyzed.

The degree of homology is expressed as the fraction of complementary bases in a duplex nucleic acid molecule formed between a given probe and the nucleic acid material being analyzed.

The minimum degree of homology detectable is a function of the experimental conditions used during the hybridization and the characteristics of the probe and nucleic acid material being analyzed.

The degree of homology between probe DNA and a filter-bound DNA material was judged from the intensity of the actual hybridization signal compared to the signal intensity observed for a 100 percent homologous filter-bound sequence under the same conditions.

The intensity of the hybridization signal depends mainly on the hybridization speed and the number of filter-bound DNA molecules present in the particular hybridization band. The hybridization rate is largely determined by the concentration of the complementary sequences during the hybridization, the ionic conditions, the temperature and the degree of homology between the probe DNA and the filter-bound molecules. Hybridization rate increases in the presence of non-complementary sequences (Bonner, T.I. et al, J.Hol. Biol. 31, 1973, p. 123), which increases the thermal stability of duplex DNA, 1 percent mismatch between probe and filter-bound DNA leads to a reduction in thermal stability of 1 degree (Maniatis et al, 1982, op.eit., p. 388). The hybridization conditions therefore determine the degree of mismatching that will still give a detectable signal. It should be noted that the conditions used in the present work did not lead to saturation of the filter-bound DNA with probe.

The present set of hybridization and filter conditions subject DNA duplexes to a temperature that is 40°C below the mean melting temperature of perfectly matched duplex DNA in the same ionic environment, that is, the conditions allow the detection of signals from duplexes containing a high degree unpaired bases. The formula used in these calculations is discussed in Beltz, G.A. et al., Meth. Enzymol. 100, 1983, pp. 266-285.

It is estimated that the conditions used detect 100 percent of the maximum hybridization signal obtained from duplicates with from 100 percent down to 80 percent homology, while the signal from a 60 percent homologous duplex is 50 percent of the above stated maximum, see above. Dupeeks with lower

homology than 60 percent gives even weaker signals.

For duplicates with a large mismatch, a signal can be detectable if the exposure time of the autoradiogram can be extended, or if the number of copies of filter-bound complementary molecules can be increased.

Example 1

Deletion mapping of the parB region (see Fig. 1) Construction of pPR95 (Fig. 2)

The construction of pPR95 was carried out as follows: plasmid pOU93 (Gerdes et al, J. Bacteriol. 161, 1985, pp. 292-98) is a pBR322 derivative containing the parB Pstl fragment derived from the EcoRI-A fragment of plasmid RI (Fig. 1). The Pstl fragment is appropriately split into smaller fragments by the restriction enzyme Rsal, as shown in fig. 1. By conventional cloning methods, the largest RsaI fragment (830 bp) was inserted into the SmaI site of the pBR322-derived cloning vector pKP34 (Prentki et al, Gene 17, 1982, pp. 189-96), yielding pPR13. The Smal site of pHP34 is flanked by two EcoRI sites, and therefore the inserted 880 bp RsaI fragment was converted to a 900 bp EcoRI fragment. The thus generated 900 bp EcoRI fragment of pPR13 was cloned into the unique EcoRI site of the miniRI derivative pOU82 (International Patent Application No. PCT/DK83/00086, Publication No. WO 84/01172), yielding pPR95. A drawing of pPR95 is shown in fig. 2. Plasmid pOU82 is unstable inherited due to the lack of any distribution function (International Patent Application No. PCT/DK83/00086, Publication No. WO 84/01172), and has a loss frequency of the order of 10~<2>per. cell per generation.

On the other hand, pPR95 is very rarely lost and is characterized by having a loss frequency of less than 10~4 per cell per generation (measured as described in International Patent Application No. PCT/DK83/00036, Publication No. WO 84/01172), which is the characteristic loss frequency of parB<+>miniRI derivatives. Thus, one can draw the conclusion that the complete parB region is arranged on the 880 bp RsaI fragment judging by the ability of the fragment to stabilize miniRI replicons.

The fine mapping of the parB region was performed as follows: pPR95 was restricted with BarnHI, treated with exonuclease Bal31 and ligated. Before ligation, BamHI oligonucleotide linkers were added. This treatment led to a series of deletion derivatives covering the left part of the parB region. The extension of the deletions was determined by size fractionation of the DNA fragments on agarose gels after the DNA had been treated with the restriction enzymes EcoRI and BarnHI. Next, the exact insertion of the BamHI oligonucleotide linkers was determined by nucleotide sequencing, as described by Maxam and Gilbert (Meth. Enzymol. 65, 1980, pp. 499-566). In this way, a very detailed mapping of the area was achieved. Furthermore, the parB phenotype (determined as described in Materials and Methods) was analyzed for each plasmid derivative. Deletion from pPR95 of the sequence extending from -320 to 0 (see Fig. 1), leading to pPR311 did not alter the parB<+> phenotype. Thus, the remaining 580 bp BamHI-EcoRI fragment in pPR311 must contain the complete parB region. Deleting from the left further into the range completely cancels the stabilizing activity.

Deletions into the right part of the 580 bp parB fragment of pPR311 (see Fig. 1) led to loss of the parB<+> phenotype, so that the parB region extends to a position near this end of the fragment.

Example 2

Nucleotide sequence of the parB region (see Fig. 3)

The nucleotide sequence of the minimal parB region shown in FIG. 3, was obtained using the chemical degradation method described by Maxam and Gilbert (Meth. Enzymol. 65, 1980, pp. 499-566). In the following, a detailed description is given of the essential biological information in the nucleotide sequence of the parB region.

The sequence of the minimal parB region of 580 bp, as defined in Example 1 (see Fig. 1), is shown in Fig. 3. The central and left parts of the area are very rich in dyad symmetries. 580 bp contains three open reading frames consisting of more than 50 codons. The start and stop codons of these reading frames are indicated in fig. 3. The reading frame starting at position +304 and ending at +460 follows a DNA sequence (5'-AGGA-3') similar to the E. coli ribosome binding site (Shine and Dalgarno, Nature (London) 254, 1975, pp . 34-38), which is known to act as the recognition site for ribosomes that initiate translation of mRNA. The polypeptide product of this reading frame is shown below the DNA sequence in fig. 3.

Example 3

Functions expressed from the parB region (see Figs. 4 and 5)

A series of plasmids were constructed from which conditional expression of the putative genes (as suggested from the sequence) in the parB region was achieved by insertion of a fragment carrying the X pR promoter and the Ac1857 gene. The positions of these deposits are indicated on fig. 1. The ApR temperature-inducible promoter system was chosen because the regulatory gene for the X. pR promoter (d 857 gene), as well as Å.pR is arranged on a single Bglll-Sall restriction fragment, furthermore the c1357 allele of the A repressor gene the inserted promoter is inducible at high temperature (42°C), but quiescent (silent) or near quiescent at low temperature (30°C). The BglII-SalI fragment of pOU82 was inserted into plasmids pPR634 and pPR341 by standard cloning procedures, yielding plasmids pKG634 and pKG341, respectively (cf. Materia-

ler and Methods).

At 30°C, cells containing pKG634 and pKG341 grow normally, but induction of ApR (at 42°C) leads to rapid killing of the host cells.

Fig. 4 shows the killing kinetics (viable counts) and growth measured as OD450 after a shift to 42°C of strain JC411 (pKG634). Viable counts drop rapidly (half-life of 2.5 minutes), and the increase in OD450 stops. The presence of an ApR promoter transcribing the parB region in the opposite direction (pKG171) has no effect on cell growth and viability (Fig. 4, comparison).

Microscopic examination (phase contrast) of the cells (JC411/- pKG634) after heat induction of the X pR promoter showed that the cells changed morphology: Most of the material apparently condensed in zones and left the rest of the cell transparent. An illustration of this is shown in fig. 5, where both normal and altered cells are present. The cells that have the characteristic parB-induced appearance are hereafter referred to as "ghost" cells.

As the XpR promoter fragment was inserted just above the start of the 52 amino acid open reading frame (see Example 2), this strongly suggests that the 52 amino acid polypeptide encoded by the open reading frame starting at position +304 (Fig. 3) is responsible for the cell killing, and consequently this gene is called hok (host killing) in what follows.

Example 4

Suppression of the host killing effect expressed by hook (see Fig. 6)

A gene from which a highly toxic product is expressed must of course be regulated. It was therefore assumed that the regulator for hock was also coded for at the parB area. In a first attempt to characterize this regulatory loop, the fragment of pKG634 containing /Cd 857 upstream of the hok gene was inserted into a mini-F plasmid, yielding pF634. Fig. 6 shows the induction of killing by JC411 (pF634) which shows that the killing takes place somewhat slower and less efficiently than is the case for pKG634 in accordance with the low copy number of F compared to pBR322.

A second parB<+->plasmid (pPR633) was then transformed into strain JC411 (pF634) and the induction experiment repeated with this double plasmid strain. As can be seen in fig. 6, the parB region present in trans completely suppresses the transcriptional activation of the hok gene. The parB region thus codes for a suppressor of host killing (the sok gene, suppressor of host killing).

Using this experimental construction as a test method, the sok gene was mapped as follows: Double plasmid strains containing pF634 and one of the deletion derivatives, respectively pPR634, pPR341, pPR154 or pPR171 were constructed and by following the growth pattern of these strains at 42° C, the 3ok phenotype of the deletion derivatives was determined by measuring the growth after the temperature change. The analysis of these deletion derivatives showed that plasmids pPR634 and pPR154 express Sok activity, while plasmids pPR341 and pPR171 express undetectable levels of Sok activity.

The plasmids were also tested for the incompatibility phenotype characteristic of par3<+> (see International Patent Application No. PCT/DK83/00086, Publication No. WO 84/01172), and it was found that plasmids expressing Sok activity also express parB specific incompatibility, while plasmids which are Sok~, as described above, do not express incompatibility. Thus, the parB incompatibility reaction represents an assay method for Sok activity.

In a manner similar to that described for the hok gene, the region required for sok gene activity has been further narrowed. One of the soc~ derivatives used in the mapping procedure, pPR171, contains the parB region extending from coordinates -300 to +288 (Figs. 1 and 3). A restriction fragment containing X d 857 and ApR was inserted into pPR1 71 in such a way that the ApR promoter reads into the sok region of the plasmid, leading to pKG171 (see Materials and Methods).

Plasmid pKG171 was transformed into strain CSH50 containing pOU94. Plasmid pOU94 is a lac<+>parB<+>p15 derivative that is fully stably inherited due to the presence of the parB region on the plasmid. Introduction of other parB<+->plasmids into this strain leads to destabilization of pOU94 due to the incompatibility expressed from parB. At 30°C, the presence of pKG171 did not destabilize pOU94 to any significant extent, while a clear destabilization was detected at 42°C. Therefore, the right-to-left transcription into the parB region of pKG171 leads to activation of the incB region (ie the sok gene).

The results described here further narrow down the sok gene, which must therefore be located between +194 (pPR634) and +288 (pPR171). It has also been suggested that the sok gene promoter reads from right to left (opposite to the hok gene transcription) and is located, at least in part, in the region between +288 (pPR171) and +336 (pPR154). A putative -10 sequence (TATCCT) is located at position +262 and a -35 sequence (TTGCGG) is located at position +285 (Fig. 3) (Kawley and McClure, Nucleic Acids Res. 11, 1983, pp. 2237-2255). It is believed that these sequences constitute the promoter for the sok gene.

Example 5

Further analysis of the hok and sok gene structure and the interaction between the hok and sok transcripts

Regulation of hok gene expression at the translational level by the Sok gene product

To gain further insight into the regulatory loop that controls the expression of the hok gene, a series of translational and transcriptional fusions between the hok gene and the lacZ gene were constructed (see Fig. 14). The β-galactosidase activities expressed from the translational fusion plasmids pKG733, pKG734 and pKG741 were very low, below one unit (as defined in Materials and Methods). Activities that low cannot be accurately measured, and therefore a DNA fragment encoding the regulatable iac promoter was inserted into pKG733, pKG734 and pKG741 upstream of the DNA fragments derived from the parB region, leading to pKG732, pKG735 and pXG742 respectively. These plasmids express low but measurable amounts of B-galactosidase.

Plasmids p3R322 and pPR1 3 (pBR322parB<+>) were individually transformed into strain CSH50 containing the translated fusion plasmids pKG732 (parB derived insert from +1 to +342) or pKG735 (parB insert from +194 to +342). The parB region present in trans was found to reduce the β-galactosidase activity expressed from the fusion plasmids from 6.0 to 3.0 units in the case of pKG732, and from 14.0 to 2.5 units in the case of pKG735. The same experiment was performed with CSK50, which carried pKG730 and pKG736 which are transcriptional fusion plasmids carrying the same inserts as pKG732 and pKG735. No suppression of the β-galactosidase synthesized from these plasmids was observed. One can therefore draw the conclusion that the parB-encoded repressor of hok gene expression (the sok gene product) acts at a post-transcriptional level, probably at the level of translation (see below).

Genetic mapping of the sok gene

Derivatives of p3R322 carrying different segments of the parB region were transformed into CSH50 carrying pKG735, and the effect on hok'-lacZ fusion protein synthesis was measured. Plasmids pPR633 (+1 to +580) and pBR634 (+194 to +580) repressed hok<1->lacZ synthesis from pKG735, whereas pPR341 (+268 to +580) did not. Similarly, pPR154 (-300 to +366) repressed hok'-lacZ synthesis from pKG735, whereas pPR171 (-300 to +283) had no such effect. This suggests that the active repressor gene is located from +194 (pPR634) to +336 (pPR154). The mapping of the sok gene is consistent with the mapping of the repressor gene using a functional analysis of the activity of the sok gene (described in Example 4).

Further characterization of the sok promoter To measure the strength and direction of the sok promoter, various Bal31-generated fragments derived from the parB region (Table 1) were inserted in front of the intact lacZ gene on the transcription fusion vector pJEL126. Plasmids pPR438 (fusion point +194) expressed high levels of S-galactosidase, while pPR341 (fusion point +268) did not express significant levels of activity, suggesting that the sok gene promoter is not intact in PR3 41. These results show that soc- the promoter is at least partially localized to the region between +194 and +268. Plasmid pPR437 (fusion point of +1) expresses an approximately 10-fold lower activity than pPR438, and therefore one must draw the conclusion that 90 percent of the sok gene transcripts terminate between +1 and +194.

A DNA fragment extending from the Fokl site at +226 (labeled 5' end) to the 3au96l site at +482 was hybridized to RNA-prepared strain CSH50 containing pPR633 (pBR322-parB). After S1 nuclease treatment of the DNA-RNA hybrid, the sample was run on a denaturing sequencing gel together with Maxam-Gilbert reactions of the same DNA probe. A weak band appeared corresponding to a transcript starting at +258. This mapping of the 5' end of the sok transcript was consistent with the genetic mapping of the sok promoter described above.

The nucleotide sequence in the region above +260 contains a putative -10 sequence (5'- TATCCT-3') arranged from +267 to +262, and a - 35 sequence (5'- TTGCGC-3') arranged from +290 to +285 (bases corresponding to E. coli promoter consensus sequences according to Hawley and McClure (op.cit.) are shown in underlined letters). Thus, all the information that is currently available, both genetic and physical, confirms the assumption made in example 4 that these sequences constitute the sok gene promoter. The transcription of B-galactosidase from the lac promoter (pKG736) gives a yield of approx. 300 units. The S-galactosidase transcribed from the sok promoter (pPR433) gives a yield of 1500 units. Thus, the sok promoter is a strong promoter with an activity that is approximately 5 times greater than that of the lac promoter.

The target for the sock product is arranged inside the sock

The translational fusion plasmids shown in FIG. 14 was used to map the target of the sok gene product (sok-T). Addition of the sok gene in trans via pPR13 to pKG732 (inserts from +1 to +342) and pKG735 (inserts from +134 to +342) suppressed the synthesis of the hok 1-lacZ fusion proteins expressed by these plasmids, whereas no such effect was evident in the case of pKG742 (insert from +268 to +342, Fig. 14). Thus, the conclusion is drawn that the genetic target for sok-RNA is coded for by the DNA that stretches from +194 to +268 and exactly overlaps the area that codes for sok-RNA itself. Plasmid pKG742 does not express sok RNA, and this plasmid was expected to yield higher amounts of hok<1->lacZ fusion protein than observed (see below).

S1 nuclease mapping of the 5'-end and 3'-end of hook mRNA

An obvious promoter-like structure in the DNA sequence upstream of the hok gene coding frame is indicated in fig. 3. This putative promoter, phok, has the -10 sequence (5'- TATGAT-3') and the -35 sequence (5'-TGGTCA-3') separated by 18 bp, no other obvious promoter-like sequences are found upstream of the coding region of the hok gene.

The 5<1-> end of the hok gene transcript was identified with the 31-nuclease-based technique described in Materials and Methods. The probe used extended from the Fokl site at +229 (labeled 5' end) to the SfaNI site at +13. The probe was hybridized to RNA prepared from strain CSH50 containing pPR633 (pBR322-par3<+>), treated with nuclease-S1 and run on a polyacrylamide gel along with the Maxam and Gilbert reactions of the same DNA probe. Furthermore, the probe was also hybridized to RNA prepared from culture samples that had been taken at different times after the addition of rifampicin to the culture. Before rifampicin addition, a very weak band appears corresponding to a transcript starting at +130, which corresponds to a transcript starting at +130 exactly in accordance with the promoter structure suggested by the DNA sequence. The addition of rifampicin led to a time-dependent enhancement of the hybridization between the DNA probe and the RNA starting at +130: At 25 minutes after the addition of rifampicin,

the hybridization is several orders of magnitude more efficient than before the addition of rifampicin. It is therefore assumed that the most important hok gene transcript starts at position +130. The rifampicin-induced time-dependent enhancement of hybridization between the transcript starting at +130 and the DNA probe used is believed to reflect the attenuation by an inhibitor of the hybridization reaction between the hog mRNA and the DNA probe. This inhibitor is most likely Sok RNA which is complementary to the same part of hok mRNA as the DNA probe used. It should be noted that there was no inhibition of hybridization between the hok mRNA and the DNA probe complementary to the 3<1> end of the hok mRNA.

The 3' end of hok mRNA was also mapped by the 31 nuclease technique. A DNA fragment extending from +342 (Sau3A site) to +580 (EcoRI site) was labeled at the 3' end of the Sau3A site and hybridized to RNA prepared from CSH50 (pPR633), treated with 31-nuclease and run on a polyacrylamide gel along with the Maxam-Gilbert reactions of the DNA probe. The main part (80-90 percent) of the hok gene transcripts terminated at positions +561, 562 and 563, and smaller transcription termination points upstream of these positions were also observed. Position +561 follows a potential stem-loop structure with a GC-rich stem, followed by a series of U's resembling typical rho-independent transcription termination structures (Rosenberg and Court, Ann. Rev. Genet 13, 1979, pp. 319-353).

Thus, the major hok gene transcript begins at +131/132 and ends around +562.

The strength of the hok promoter was determined by transcribing fusion plasmid pKG730 (Fig. 14), which contains the lac promoter and the hok promoters in tandem in front of the intact lacZ gene. The β-galactosidase activity expressed from pKG730 in CSH50 is approx. 300 units. To avoid transcription from the lac promoter, a pBR322 derivative carrying the lac1 gene (pHB4) was transformed into CSH50 (pKG730). This double plasmid strain expresses a B-galactosidase level of 100 units (the lac1 gene present in trans on a multicopy plasmid reduces the activity of the lac promoter to less than 1 percent of the activity of the unrepressed state). Thus, the hok promoter is relatively weak with an activity of approx. 50 percent or less relative to the lac promoter. the sok promoter is approx. 5 times stronger than the lac promoter, and the sok promoter must therefore be at least 10 times more active than the hok promoter.

Induction of the translation of hok mRNA by the addition of rifampicin

Plasmid pKG780 is a pBR322 derivative carrying the sok gene, the hok<1->lacZ hybrid gene obtained from pKG733, part of the lacY gene and the part of the parB region that codes for the hok mRNA termination. Plasmid pKG732 is identical to pKG730, except that the downstream parB insert is replaced with the terminal part of the tetracycline gene obtained from pBR322. Despite the high copy number of these plasmids, they express very low levels of S-galactosidase (approx. 1 unit in both cases), which corresponds to very few 6-galactosidase molecules per cell.

Rifampicin was added (300 µg/ml) to exponentially growing cultures of CSK50 containing either pKG780 or pKG782, and samples were taken to determine β-galactosidase activity. There was no increase in activity in the first 10 minutes after the addition of rifampicin, but thereafter an induction of 6-galactosidase production (and consequently translation of hok'-lacZ mRNA) was observed. The synthesis of the hok'-lacZ fusion protein lasts for approx. 80 minutes in both cases. The half-lives for hok'-lacZ-hok<1-> and hok'-lacZ-tet' mRNA were calculated to be approx. 22 and 20 minutes (see Materials and Methods).

Thus, the insertion of the hok mRNA termination downstream of the LacZ gene did not cause any significant increase in stability of the hok'-lacZ hybrid mRNA.

Conclusions

Detection of RNA transcripts synthesized from the parB region shows that hok mRNA begins at +131/132 and ends around +562, and that sok RNA begins at +260 and ends between +1 and +194. Accordingly, sok RNA is complementary to the leader sequence of hok mRNA (Fig.

3). The analysis of promoter and gene fusions between the hok gene and the lacZ gene shows that the sok gene encodes a repressor of hok gene expression, which acts at the post-transcriptional level. The region from +194 to +336 codes for an active repressor gene that is transcribed in the opposite direction to the hok gene transcription. No obvious protein-coding frame is present in the sok gene region, and one therefore draws the conclusion that the repressor coded for by the sok gene is a small antisense RNA (antisense RNA) which, when hybridized to the complementary hok mRNA, prevents translation of the latter. This observation is supported by the mapping of the target of the sok gene product (sok-T), which was localized to the region coding for the leader sequence of the hok mRNA that is complementary to sok RNA (+194 to +268).

the sok-T motif is arranged upstream of the SD sequence of hok-mRNA, and therefore the derepression of translation by sok-RNA upon hybridization to hok-mRNA can most likely be explained by its long-term effects (obvious to the secondary structure of the mRNA). The region coding for the leader sequence of hok mRNA (and consequently sok-T) is rich in dyad symmetries. hok mRNA has the potential to form three starch loop structures in the range from +194 to +303, shown as I (AG = -16 kcal/mol), II ( G = -20 kcal/mol) and III ( AG = -26 kcal/mol) in fig. 3. sok mRNA is complementary to the region coding for loop I, parts of loop II, but not loop III. The Shine-Dalgarno sequence of the hok gene will, if loop III is formed, be sequestered in the stem of this high-energy loop. It is therefore likely that the formation of loop III is a prerequisite for the suppression of hook mRNA translation.

The 5-loop structures II and III overlap and therefore cannot be formed simultaneously. If loop II is formed, the SD sequence will probably not be sequestered in the secondary structure of the mRNA, and the SD sequence will acquire an "open state" freely available to be translated by the ribosomes. It is therefore believed that the hybridization of sok RNA to hok mRNA changes the pattern of mRNA looping from an "open state" (formation of loop II) to a "closed state" (formation of loop III), yielding an aiRNA that does not can be translated by the ribosomes.

The S-galactosidase activity expressed from pKG472, where the hok'-lacZ gene is transcribed from the lacUV5 promoter was unexpectedly low (20 units), as this plasmid does not contain the sok gene. The region upstream of the SD sequence of pKG742 contains an intact loop III, but not loops I or II. Consequently, loop III can constantly form, without competition from loop I or loop II, and this may be the reason for the low β-galactosidase expression from this mRNA.

The addition of rifampicin to cells containing plasmids pKG780 or pKG782 induced synthesis of hok'-lacZ hybrid mRNA which was found to be very stable with a half-life of approx. 20 minutes. Surprisingly, the hybrid mNRA containing the tetracycline gene mRNA termination (pKG732) is as stable as the hybrid mRNA containing the native hok mRNA termination (pKG780). This result cannot be easily extrapolated to the wild-type hok mRNA, but one can draw the conclusion that the 5' end of the hok mRNA leader sequence (from +1 to +342) encodes a determinant important for the stability of the hok- the mRNA molecule.

The initial time delay in S-galactosidase synthesis is most easily explained as a reflection of the degradation time of the inhibitor of hok<1->lacZ protein synthesis, that is, sok RNA.

The rifampicin-induced time-dependent enhancement of the hybridization reaction between hok mRNA and the DNA probe used for mapping the 5' end of hok mRNA can be most easily explained as follows: It is known that RNA hybridizes several orders of magnitude faster with RNA than with DNA ( Favaloro et al, in Methods in Enzymology 65, 1983, pp. 718-749). Therefore, the presence of sok-RNA (before rifampicin addition) prevents the hybridization between the 5'-end of hok-mRNA and the complementary DNA probe, because hok-mRNA exhibits a faster hybridization reaction with sok-RNA than with the DNA probe (sok-RNA and The DNA probe used in this experiment is complementary to the same part of hok mRNA). After the addition of rifampicin, which prevents the formation of new transcripts, the hybridization between hok mRNA and the DNA probe complementary to the 5' end of hok mRNA increases strongly with time. This enhancement of hybridization reflects the attenuation of the hybridization inhibitor, sok RNA. It should be noted that there is no hindrance of the hybridization between the hok mRNA and a DNA probe complementary to the 3<1> end of the hok mRNA. Significant amounts of hok mRNA are present in the cells even 25 minutes after the addition of rifampicin, suggesting that hok mRNA is unusually stable. These interpretations are in full agreement with the kinetics of hok 1-lacZ fusion protein synthesis observed in the rifampicin induction experiments.

?to fig. 13 is a model explaining the killing of cells that have lost a parB-carrying plasmid during the previous cell division shown: The presence of a parB-carrying plasmid in the cell gives rise to cytoplasmic products of the hok and sok genes, hok mRNA and Sok -RNZ*. hok-mRNA is stable with a half-life of 20 minutes or more, while Sok-RNA, which prevents translation of hok-mRNA, is unstable, but is present in excess. Consequently, in a plasmid-carrying cell, the Sok RNA prevents synthesis of the Hok protein, and the cell remains viable. When, on the other hand, a plasmid-free cell appears, e.g. upon uneven distribution of the plasmid molecules during cell division, the unstable Sok RNA in this cell decreases, leaving the Hok mRNA freely available to be translated by the ribosomes, the Hok protein is synthesized, and killing of the plasmid-free cell follows.

An important testable prediction emerges from the post-segregational depression model of hok gene expression proposed above: All unstable inherited plasmids can be stabilized at the parB region, as the killing mechanism is independent of the actual reason for loss of the plasmid. Accordingly, the parB region has been shown to stabilize many types of replicon ranging from low copy number plasmids (e.g. R1 (see International Patent Application No. PCT/DK83/00086, Publication No. WO 84/01172)) to high copy number plasmids copy number (p15A, pBR322 (see PCT/DK33/00086, Publication No. WO 84/01172) and RSF1010) - all of which have become unstable after genetic manipulation. oriC minichromosomes can also be stabilized at the parB region. Surprisingly, the parB region stabilizes a highly unstable RSF1010-lacZ<+-> derivative in P. putida (see Example 12).

Similar regulatory systems involving regulation of the translation of an mRNA material by an antisense RNA molecule have been described previously (Light and Molin, The EMBO Journal 2, 19S3, pp.93-98, R.W. Simons and N. Kleckner, Cell 34 , 1983,..pp. 683-691, Mizuno et al, Proe. Nati. Acad. Sei., 1984, pp. 1966-1970), and even artificial systems where synthetic sequences have been inserted as codes for antisense -RNA complementary to the leader region of an mRNA has been described (Izant and Weintraub, Cell 36, 1984, pp. 1007-1015).

Example 6

Discovery of an E. coli chromosomal homolog of R1 parB (see Fig.

8a)

Since plasmid evolution has involved an extensive exchange of genetic information between bacterial chromosomes and freely replicating DNA molecules, the chromosomal DNA of E. coli was analyzed for possible ancestral sequences for the R1 parB sequences.

In field 6 on fig. 8a, the total EcoRI-restricted DNA from plasmid-free E. coli JC411 was analyzed by filter hybridization to a parB probe, see Materials and Methods. A fragment of 20 kb can be seen to give a rather weak but definite signal which can also be detected in other fields containing E. coli DNA if they are exposed for the same time (fields 4, 5). The chromosomal sequence is estimated to be approx. 55 percent homologous with parB. The chromosomal sequence is called pari in what follows.

A big question is, of course, to what extent the discovery of homology at the level of the nucleotide sequence also reflects similarity with regard to function of the products coded for by the homologous non-arrays, an aspect that will be dealt with in more detail in example 7.

Example 7

Genetic organization of an E. coli chromosome! homologue of R1 parB and its functional relationship with RI parB (see Fig. 7) the hok gene of plasmid R1 defined in example 3, codes for a polypeptide of 52 amino acids. The amino acid sequence of the hok gene product was compared with a large number of known protein sequences. It was surprisingly found that a 51 amino acid polypeptide encoded by the relB-orf3 gene of the E. coli relB operon (Bech et al, The EMBO Journal 4, 1985, pp. 1059-1066) showed significant homology to the hok product . The amino acid sequences of the two homologous proteins are shown in fig. 7, which shows that 42 percent (22) of the amino acids are identical in the two proteins. For 17 percent (9) of the amino acids, the changes are conservative, which means that an amino acid has been replaced with an amino acid with similar chemical properties (that is, hydrophobicity, charge, etc.), which leads to a total homology of 61 percent. The charged amino acids in particular are well conserved, as are the cysteine residues in positions 16 and 31 (Fig. 7).

The DNA sequences of the hok gene and of relB-orf3 were also compared as shown in fig. 7. The coordinates used in the following are parB<+->sequence coordinates as shown in fig. 3. From coordinates +290 to +460 there is 55 percent homology between the two sequences. It appears from fig. 7 that the conserved region comprises nucleotides on the top and bottom of the protein coding sequence which is arranged from +304 to +460. The conservation of bases outside the coding region suggests that regulatory features of the two genes have also been at least partially conserved.

To show that sequence homology reflects similarity in function, a plasmid carrying the A pR promoter fragment upstream of the rel3-orf3 gene was constructed (see the description of an analogous type of construct used in mapping the hok gene in Example 3).

When A pR-mediated transcription into relB-orf3 is induced, a rapid killing of the cells is observed with kinetics similar to that observed for bacteria containing plasmid pKG341, as described in Example 3. Similarly, all the cells in the culture were transformed into in the hook-characteristic "ghost cells" (see fig. 5).

There is thus a striking homology between the hok gene of plasmid R1 and the relB-orf3 of the E. coli relB operon both at the structural and functional level.

Example 8

parB homologous sequences on different plasmids (see Fig. 8a)

Filter hybridization analysis of pooled EcoRI-restricted DNA from a number of strains of E. coli containing different plasmids was performed using the parB probe (lanes 1-5 of Fig. 8a).

Plasmid R1drd-19 is a member of the R1 plasmid family, from which the parB probe was originally cloned. R1drd-19 is available in two copies per bacterial genome. EcoRI-restricted pooled DNA from E. coli 1005/R1drd-19 is analyzed in lane 1. A strongly hybridizing fragment of 19.5 kb can be seen, the size of which is consistent with the genetic mapping of parB function to the 19.5 kb R1 plasmid (International Patent Application No. PCT/- DK83/00086, Publication No. WO 84/01172).

Plasmid R100 is closely related to RI carrying a transposable element Tn10 within the region equivalent to the 19.5 kb EcoRI-A fragment of R1. The transposon contains the recognition sequence for EcoRI, and consequently a further EcoRI site is introduced into the R1-like EcoRI-A fragment and cleaves this into the two EcoRI-A and EcoRI-D fragments of R100 (Miki et al, J. Bacteriol. 144, 1930, pp. 87-99). These two EcoRI fragments of R100 both contain sequences which, by heteroduplex mapping, have been found to be homologous to sequences present in the F factor (Sharp et al, J. Mol. Biol. 75, 1973, p. 235). A strongly hybridizing fragment of 12.8 kb is seen in lane 2, fig. 8a, thereby mapping the parB region of R100 to the EcoRI-D fragment of R100, within the center of the region of homology between R1 and R100 and F.

This localization of parB within the F-homology region of R100 led to the search for parB-like sequences on plasmids belonging to the incompatibility group, IncFI.

EcoRI-restricted pooled DNA from B210/R386, an E. coli strain containing the IncFI plasmid R386, was analyzed by filter hybridization using the parB probe (lane 3, Fig. 8a).

Plasmid R386, which belongs to the same incompatibility group as F, was found to give a parB hybridization signal corresponding to an EcoRI fragment of 19.5 kb. As this plasmid is present in 0.5-1 copies per genome, suggests the discovery of a signal of ca. one third of the RI 00 signal (lane 2, Fig. 8a) that the degree of homology between R1 parB and the R386 par3-like sequences is 55-60 percent.

The search for parB-related sequences was extended to other incompatibility groups. The plasmid RP1, which belongs to the incompatibility group IncP, was analyzed.

With the parB probe, the total EcoRI-restricted DNA from 1005 (RP1) gives a hybridization signal corresponding to the EcoRI-linearized plasmid (lane 4, Fig. 8a). Furthermore, a hybridizing band of 20 kb is seen corresponding to pari, which was discussed in example 5.

As RP1 is adapted for stable maintenance in a wide range of gram-negative bacterial hosts, the discovery of parB-related sequences on RP1 opens up the possibility of maintenance systems analogous to the R1 parB system, which are operative in E. coli as well as in Pseudomonas putida (Example 11), can function in a variety of bacterial hosts.

Another plasmid, R6-K (IncX incompatibility group), was found to carry sequences with nearly the same hybridization properties as RP1, as evidenced by the presence of a 25 kb EcoRI fragment of R6-K, which hybridizes the parB probe (lane 5, Fig. 8a).

The low copy plasmid F has been analyzed in detail to determine whether the presence of R1 parB hybridization sequences reflects the presence of a stabilization mechanism akin to that of R1 parB.

Two plasmid stabilization functions have been identified within the genome of F, and the corresponding genes (sop (Ogura and Hiraga, Cell 32, 1983, pp. 351-360) and ccd (Ogura and Hiraga, Proe. Nati. Acad. Sei. USA 30, 1933, pp. 4784-4783)) has been localized to the EcoRI fragment spanning map positions 40.3 to 49.5.

Filter hybridization analysis of pooled DNA from E. coli 1005 containing F showed that R1 parB-related sequences were present on a 10.7 kb EcoRI fragment of F (map position 49.5 to 60.2), and additional hybridization analyzes of 1005(F)- DNA digested with EcoRI and/or BarnHI mapped these sequences to a 4.5 kb BamHI-EcoRI fragment extending from map position 55.7 to 60.2. This suggests the presence of a third plasmid stabilizing function within F.

The region of F hybridizing the R1 parB probe was then cloned into a bacteriophage /. vector. EcoRI-digested DNA from 1005(F) was size fractionated by preparative agarose gel electrophoresis, and fragments of 9.5-12 kb were recovered by electroelution from the gel. The fragments were ligated to the EcoRI sites on the left and right arms of ÅL147 and packaged in vitro to give infectious phages (see Maniatis et al, Molecular Cloning, Cold Spring Harbor, New York, 1982, p. 256), which were then used to infect E. coli LE392. Recombinant phages carrying R1 parB-related sequences were identified by plaque hybridization.

From a recombinant phage carrying the 10.7 kb EcoRI fragment containing the RI parB-related sequences, the fragment was isolated and inserted into pUC8 at the EcoRI site. In a resulting plasmid, pNL1, the insert is oriented in such a way that cleavage of the pNLI DNA with BarnHI causes a 4.5 kb fragment carrying the R1 parB-hybridizing sequences to be cleaved.

The 4.5 kb BarnHI fragment from pNL1 was isolated, and the region hybridizing the R1 parB probe was mapped to an 870 bp Rsal fragment by filter hybridization analysis. The 370 bp RsaI fragment was isolated and inserted into the SmaI site of Ml3mp9. A number

recombinant phages carrying R1 parB-related sequences on a 370 bp insert were identified by plaque hybridization. The nucleotide sequence of the inserted DNA was analyzed according to Sanger et al, Froc. Nati. Acad. Pollock. USA 74, pp. 5463-5467.

The nucleotide sequence from part of one of the recombinant phages, mpNL12, comprises 402 bases extending from the RsaI site, and this sequence is 90 percent homologous with the region from +178 to +580 of the R1 parB sequence (Fig. 3). All essential features of the R1 parB region are also found in the F-derived sequence: (1) an open reading frame coding for a protein of 50 amino acids is present corresponding to the R1 hok gene, (2) the ribosome binding site of R1 hok is conserved, (3) the region corresponding to the 3'-untranslated part of R1 parB mRNA, which is thought to be required for hok mRNA stability, is well conserved (90 percent homology), and (4) the putative - 10 and -35 areas of R1 sok are also preserved.

The open reading frame within the F-derived sequence encodes a protein of 50 amino acids that differs only slightly from the R1 specified hook protein. Firstly, two codons in R1-hok have been deleted, namely val-15 and ser-29. Second, two conservative substitutions have taken place, namely leu-16 to val and his-39 to tyr.

It is obvious that the R1 hok gene and the related sequences on F derive from a common ancestral sequence, and furthermore, the conservation of a coding region corresponding to R1 hok strongly suggests that the encoded protein is involved in the stabilization of F.

To test for plasmid-stabilizing properties of the F-derived sequence, the 4.5 kb BamKI fragment from pNL1, which carries the F hook-like sequences, was inserted into pJEL82, a low-copy plasmid with a loss frequency of 10~<2 > per generation (see PCT/DK83/00084, Publication No. WO 84/01171). The resulting plasmid, pJEL82/F, as well as pJEL82, was transformed into E. coli H3101. Cultures of the two strains were grown for 16 hours without selection pressure, and the fraction of plasmid-containing cells (Ap<R>) was determined. The result was as follows:

It was therefore concluded that the 4.5 kb BarnHI fragment carrying R1 parB-related sequences exerts a plasmid stabilizing effect. If the stabilization is due to the presence of the hook-like gene inside the F fragment, the appearance of ghost cells would be expected in cultures of cells containing pJEL82/F grown without selection pressure, see example 3. An overnight culture of cells containing pJEL82/F was found to contain approx. 5 percent ghost cells that were indistinguishable from R1 hock-induced ghost cells.

In the case of F, the demonstration of sequences related to R1 parB by filter hybridization thus reflects the existence of a functionally similar plasmid stabilization mechanism.

Example 9

Stepwise hybridization as a strategy for detection of replicon-stabilizing sequences homologous to parB-related sequences (see Fig. 8b)

Ratio of hybridization determines the degree of homology between a probe and a filter-bound DNA material that is necessary to give a detectable signal, see the discussion in Materials and Methods. Consequently, filter-bound sequences may exist that remain undetectable with the given probe under the given set of hybridization conditions, but may nevertheless encode a hook-like activity, see the discussion of homology versus function in Materials and Methods. This is shown in the following experiment.

As described in Example 6, relB-orf3 represents a chromosomal homologue of RI parB based on sequence comparison data and the functional similarity between hok and relB-orf3. relB-orf3 and flanking sequences, as they are present in plasmid pBD2724, were used as a probe in a filter hybridization analysis of E. coli chromosomal DNA.

Plasmid pBD2724 is a pER322 derivative containing a KincII-MluI fragment from the relB operon of E. coli comprising the relB-orf3 coding sequence (coordinates 1070-1350 according to Bech et al, op.cit.).

In field 3, fig. 3b, the total EcoRI-restricted DNA from plasmid-free E. coli is analyzed by filter hybridization using the relB-orf3 probe. Besides the 20 kb hybridizing fragment which probably represents the above identified pari sequence (Example 6), another hybridizing fragment of 16 kb is detected. Since the intensity of the latter is greater than the intensity of the 20 kb signal, the 16 kb EcoRI fragment must span the E. coli relB-orf3 gene used as a hybridization probe, i.e. the intensity of the 16 kb signal provides a reference, from which degrees of homology can be estimated. The intensity of the pair hybridization signal, which is approx. 3/4 of the relB-orf3 signal, suggesting that pari is ca. 65-70 percent homologous with relB-orf3. Since the 16 kb relB-orf3-carrying fragment is not detected with the parB probe (lane 6, Fig. 8a), RI parB is 50 percent or less homologous to relB-orf3.

In Example 5, the parB probe was found to detect the 20 kbp chromosomal homolog, but not the 16 kbp homolog that represents, according to the above data, relB-orf3. As the latter, as described in example 6, exerts hok-like activity when expressed, it can be assumed that pari will also express hok-like activity or sok-like activity and/or both activities when properly expressed.

The relB-orf3 fragment was used as a probe in filter hybridization analysis of E. coli containing plasmid R100 and R386, both of which contain R1 parB-like sequences (Fig. 3a, lanes 1 and 2). Under the suggested hybridization conditions, the relB-orf3 probe did not detect these sequences (Fig. 8b, lanes 1 and 2), as only the 20 kb pari fragment and the 16 kb relB-orf3-carrying fragment can be seen to hybridize the probe , which suggests that

the absence of hybridization between a probe from a region that expresses hok or hok-like activity and a given DNA material does not exclude that the relevant DNA material can exert hok-like activity if it is properly expressed. The discovery of homology between the DNA material in question and the region that expresses hok or hok-like activity therefore strongly suggests that the DNA material in question will exert hok or hok-like activity if it is expressed correctly.

The above data therefore reveal a useful strategy in the search for regions that exert hok/sok-like activities: A probe representing a region of nucleic acid comprising hok or hok-like genes (e.g. R1 parB) is used to detect homologues sequences (e.g. pari) within the relevant genome (e.g. chromosomal or plasmid DNA) which are then tested for hok or hok-like activity (as was done for the relB-orf3 region) in the appropriate experimental setting, and if it turns out to code for such an activity or activities, they are then themselves used as probes in a second round of hybridisations to define new homologous sequences which are possibly related to the probes used in the first round of hybridisations (e.g. eg R1 parB). This stepwise approach combining nucleic acid hybridization and functional analyzes of the isolated nucleic acid sequences can be adapted as a general strategy to search for genes expressing hok or hok-like activities in genomes that are increasingly separated from E. coli on the evolutionary scale.

Example 10

Detection of parB-related sequences in bacteria (see Figs. 9 and 10)

In the previous examples, it was shown 1) that sequences related to R1 parB are widely distributed among bacterial plasmids isolated from Gram-negative bacteria, and 2) that sequences related to R1 parB are present in the chromosomal DNA of E. coli. These discoveries prompted a search for sequences related to either R1 parB or one of the av. the chromosomal counterparts (E. coli relB-orf3) in DNA from a number of different bacteria as part of either their chromosomal DNA or of plasmids naturally present in these organisms.

Filter hybridization analysis of EcoRI-restricted DNA from Serratia marcescens with the R1 parB probe shows intense hybridization to 3 fragments of 4.1, 2.9 and 2.5 kb (lane 2, Fig. 9). Only the 4.1 kb fragment also hybridizes with the relB-orf3 probe (lane 1, Fig. 10). The parB probe hybridizes an additional 6 fragments. Two of these signals are stronger than the parB signal obtained from the reiB-orf3-carrying 16 kb fragment in E. coli DNA (lane 6, Fig. 9). Hybridization of Serratia marcescens DNA with the E. coli relB-orf3 probe gives a number of weak hybridization signals. It is possible that the strong hybridization bands of 2.5, 2.9 and 4.1 kb are derived from plasmid(s) , although agarose gel electrophoresis did not reveal any high copy number plasmids.

Pseudomonas fluorescens was analyzed as a plasmid-free member of this species. Hybridization of DNA from Pseudomonas fluorescens with R1 parB (lane 3, Fig. 9) shows 8-10 hybridizing fragments, of which 4 show signals with intensities of approx. 33 percent of the chromosomal counterpart of R1 parB (field 6, fig. 9). A single one of these fragments, of approx. 13 kb, also probably hybridizes the E. coli relB-orf3 region (lane 2, Fig. 10). Moreover, the relB-orf3 probe identifies 5 fragments in particular, albeit at low signal intensity. Two of these, of 5.5 and 5.6 kb, are also seen in DNA from Pseudomonas putida when this DNA is analyzed using the relB-orf3 probe (lane 3, Fig. 10). The relB-orf3 probe hybridizes to an additional 5 fragments in Pseudomonas putida DNA, but none of these fragments are recognized by the R1 parB probe (lane 4, Fig. 9). In Pseudomonas putida DNA, the parB probe detects approx. 10 fragments with low signal intensity and a single fairly strong hybridizing fragment of approx. 7.3 kb.

Among Gram-positive bacteria, B. subtilis, B. circulans PL236 and two strains of Lactobacillus were analyzed for the presence of sequences related to either R1 parB or E. coli relB-orf3.

In the case of the parB probe, a single rather strongly hybridizing fragment of 5.2 kb was found in DNA from B. circulans (lane 8, Fig. 9). Very weak signals were obtained from a few

additional fragments of B. circulans DNA. With the relB-orf3 probe, a limited number of hybridizing fragments were seen in DNA from B. subtilis (lane 4, Fig. 10), B. circulans (lane 5, Fig. 10) and Lactobacillus (lanes 6 and 7, Fig. 10 ). The number of relB-orf3 hybridizing fragments varied from 6-11, and all have approximately the same signal intensity. In Lactobacilli, agarose gel electrophoresis has shown the presence of plasmids, suggesting the possibility that at least some of the hybridizing sequences are of plasmid origin. A search for plasmids in B. circulans PL 236 has been negative, suggesting that the sequence of B. circulans DNA hybridizing the R1 parB probe (lane 8, Fig. 9) may be of chromosomal origin.

The above-mentioned experiments show that sequences related to R1 parB and/or E. coli relB-orf3 are widely distributed among bacterial genera and not only Enterobacteriaceae, from which the probes were derived, but also the Gram-positive bacteria.

Example 11

Detection of parB-related sequences in eukaryotic cells (see Fig. 11)

A unicellular eukaryotic organism was examined, namely the ciliate protozoan Tetrahymena thermophila, fig. 11. This organism is characterized by 1) a high number of mitochondrial DNA molecules per cell, and 2) approx. 12,000 copies of ribosomal RNA genes arranged on cell replicating rDNA molecules. Neither the R1 parB probe nor the E. coli relB-orf3 probe detect any fragments in DNA prepared from isolated macronuclei (lane 1, Fig. 11). Keller did not hybridize the probes to the two EcoRI fragments of isolated rDNA (lane 3, Fig. 11). The pooled EcoRI-restricted DNA from Tetrahymena thermophila, which includes mitochondrial DNA, showed two hybridizing fragments of 6.6 kb and 3.3 kb (lane 2, Fig. 11), with the relB-orf3 probe, while the parB probe did not gave some signals. The hybridizing fragments co-migrated with two EcoRI fragments of mitochondrial DNA which were readily detectable by ethidium bromide staining of the gel before DNA transfer.

Pea (Pisum sativum) chloroplast DNA was cleaved with the restriction endonuclease Pstl, and 0.125 µg was analyzed by filter hybridization using the parB and parB-orf3 probes (lane 4, Fig. 11). The latter probe hybridizes to a fragment of approx. 16 kb.

Finally, two samples of human cell DNA were analyzed by filter hybridization following EcoRI restriction. The R1-parB probe gave a (weak) hybridization signal to neuroblastoma DNA (lane 5, Fig. 11), as well as to embryonic liver DNA (lane 6, Fig. 11). The high mitochondrial content of liver tissue may indicate that the observed signal in field 6, fig. 11 is obtained from human mitochondria.

Neuroblastoma DNA was analyzed as other hybridization analyzes had suggested selective enlargement of a small chromosomal region leading to the presence of "double minute" extrachromosomal minichromosomes ("double minutes"), the origin of the hybridization signal in field 5, Fig. 11 is unknown.

Example 12

Complete stabilization of highly unstable recombinant plasmids in P. putida and S. marcescens at the parB region (see fig.

12)

When cloning foreign DNA into Pseudomonas spp., it has been found that the introduced plasmids often become very unstable. This plasmid instability was thought to be caused by the introduction of an unknown host cell function that somehow recognizes and eliminates the foreign DNA (Kim and Meyer, J. Bacteriol. 159, 1984, pp. 678-682). As Pseudomonas spp. are highly relevant for industrial purposes, it was considered important to test the effect of parB on such highly unstable recombinant plasmid derivatives.

Plasmid pKT230 is a cointegrate plasmid consisting of the wide host range vector RSF1010 (which can be stably maintained in a wide range of Gram-negative bacteria) and pACYC177, a derivative of the natural p15 plasmid (Bagdasarian et al, Gene 16, 1981, pp. 237-242 ). A BglII restriction fragment of plasmid pJL217 (Light and Molin, Hol. Gen. Genet. 187, 1982, pp. 486-493), containing the lac genes, was inserted into the unique BamHI restriction site of pXT230, yielding pFN13. Plasmid pFNl3, which has a high copy number, is stably maintained in E. coli with a loss frequency of less than 10-<4> per cell per generation. In P. putida, however, the same plasmid derivative has a loss frequency of 30 per cent per cell per generation. In S. marcescens, the loss rate is 5 per cent per cell per generation.

To examine the effect of parB, the parB EcoRI fragment of pPR95 was inserted into the unique EcoRI site of pFN13. The resulting plasmid was conditioned pFN13 parB<+>, and a map of this is shown in fig. 12. Plasmid pFN13 was mobilized by a conjugative plasmid into P. putida and 3. marcescens, selecting for kanamycin resistance in both cases. The loss frequency for pFNl 3 parB<+>in both P. putida and S. marcescens was less than 10~° per cell per generation, corresponding to an increase in plasmid stability in P. putida of at least 10^ times, and in S. marcescens of at least 10<4> times.

Thus, it can be concluded that the parB<+->region is very effective in ensuring stability of highly unstable recombinant plasmids even in bacterial species distantly related to E. coli.

Example 13

To analyze whether the R1 hok gene exerts a cell-killing effect in Gram-positive bacteria, e.g. Bacillus subtilis, if properly expressed from a B. subtilis promoter, the parB region is isolated from pPR95 (Fig. 2) as an EcoRI fragment. The EcoRI sites are filled in using Klenow polymerase to form blunt ends, and the fragment is then digested with FspI. The 300 bp Fspl-(EcoRI) fragment (spanning map position +286 to +580 in Fig. 3) lacking the hok promoter is isolated and inserted into pSI-1 (Yansura & Henner, Proe. Nat. Acad . Sei. USA 81, 1984, p. 439), which has been digested with Sali and then treated with mung bean nukiease (rnung bean) (Rosenberg et al, Meth. Enzymol. 101, 1983, p. 123) to remove the 5' single-stranded overhangs. The ligation mixture is transformed into E. coli MC1000 (Casabadan and Cohen, J. Holec. Biol. 133. 1980.

pp. 179-207).

This construct places the hok gene containing the hok ribosomal binding site near a synthetic B. subtilis ribosomal binding site present on pSI-1, a B. subtilis/E. coli shuttle vector. The expression of the inserted gene is controlled by the SP01 promoter on pSI-1, whose activity is controlled via the lac operator which exists between the promoter and the synthetic ribosomal binding site (spac-I promoter, Yansura & Henner, op.cit.). pSI-1 also expresses the lac1 gene which codes for the lac repressor.

E. coli transformants are analyzed for IPTG (isopropylthiogalactoside) induced growth inhibition, as the SP01 promoter is active in E. coli. Sorting is performed via replicate plating from plates containing 8 µg/ml chioramphenicol to plates containing chloramphenicol and IPTG (1 mM). Colonies that grow poorly in the presence of IPTG are further analyzed in liquid culture to confirm the phenotype. A number of transformants exhibiting ghost cell formation (see Example 3) when IPTG is added to 1 mM have been identified.

Transformants are further grown in liquid culture to OD500 = 0.2, at which point IPTG is adjusted to 1 mM. Addition of IPTG leads to growth arrest and formation of ghost cells (see example 3) in the culture.

Plasmid DNA is isolated from such IFTG-sensitive E. coii transformants and introduced into B. subtilis BD224 by protoplast transformation (Chang & Cohen, Molec. Gen. Genet. 168, 1979, p. 111) with subsequent selection for chloramphenicol-resistant transformants.

To assay for a possible cytotoxic effect of R1 hok in B. subtilis, the Cm<R->resistant B. subtilis transformants are plated in the absence or presence of 1 mM IPTG. If no growth

or severe growth inhibition is observed in the presence of IPTG, RI hock is correctly expressed. It is then possible to impose a sok-like antisense RNA regulation on the translation of hok mRNA by appropriate constructs, as suggested above.

Claims (35)

1. Replicon which, when contained in a cell, expresses a product capable of killing the cell or its progeny (or a precursor of the product) and which further expresses an antagonist of the cell-killing product (or a precursor of the antagonist), the antagonist being one that suppresses the killing product (or a precursor of the killing product) in cells containing the replicon, while the antagonist activity decreases when the replicon is lost from the cell so that the antagonist (or its precursor) is no longer continuously expressed, meaning that the killing product (or its precursor) present in the now replicon-free cell is no longer suppressed by the antagonist, leading to cell death, with the proviso that the cell-killing product as well as its antagonist is not encoded by the R1 parB region when the replicon is a plasmid that can replicate in gram-negative bacteria. 2. Replicon as stated in claim 1, characterized in that the sequence on the replicon which codes for the killing product (or its precursor) is obtained from a bacterial plasmid, a bacterial chromosome, a eukaryotic plasmid, a eukaryotic virus, a eukaryotic autonomously replicating sequence carrying a chromosomal origin of replication, a eukaryotic mitochondrial DNA molecule or a eukaryotic chloroplast molecule. 3. Replicon as stated in claim 1, characterized in that the sequence and the replicon which encodes the antagonist (or its precursor) are obtained from a bacterial plasmid, a bacterial chromosome, a eukaryotic plasmid, a eukaryotic virus, a eukaryotic autonomously replicating sequence which carries a chromosomal origin of replication, a eukaryotic mitochondrial DNA molecule, or a eukaryotic chloroplast molecule, or is a synthetic sequence. 4. Replicon as stated in one of claims 1-3, characterized in that it contains a nucleotide sequence that is not naturally related to the replicon. 5. Replicon as stated in one of the preceding claims, characterized in that when it is contained in a cell, it expresses a messenger RNA which codes for a product which can kill the cell or its progeny and which further expresses an antagonist which prevents the translation of messenger RNA in cells containing the replicon. 6. Replicon as stated in claim 5, characterized in that the antagonist is an antisense RNA molecule (antisense RNA molecule) which, by base pairing to at least part of the messenger RNA that codes for the cell-killing product, prevents translation of this messenger RNA. 7. Replicon as stated in one of the preceding claims, characterized in that it is a DNA molecule comprising a sequence that codes for a product (or a precursor of the product) that can kill the cell or its progeny in a manner identical to or analogous to the product which (or whose precursor) is expressed by the R1 parB hok gene. 8. Replica as stated in claim 7, characterized in that the sequence which codes for the product which can kill the cell or its progeny (or a precursor of the product) comprises the R1 parB hog gene or a sequence homologous to this. 9. Replicon as stated in one of the preceding claims, characterized in that it is a DNA molecule comprising a sequence that codes for an antagonist that functions in a way that is identical to or analogous to the antagonist expressed by R1 parB sok- the gene. 10. Replicon as stated in claim 9, characterized in that the sequence which codes for the antagonist comprises the R1 parB sok gene or a sequence which is homologous to this. 11. Living cell, characterized in that it contains one replicon comprising the R1 parB hok gene or a sequence homologous to this and another replicon comprising the R1 parB sok gene or a sequence homologous to this. 12. Living cell as stated in claim 11, characterized in that the replicon comprises the R1 parB hok gene or the sequence homologous to this is a chromosome. 13. Living cell as stated in claim 12, characterized in that the replicon comprising the R1 parB sok gene or the sequence homologous to this is an extrachromosomal, autonomously replicating nucleotide sequence. 14. Living cell as stated in claim 13, characterized in that the replicon is a plasmid. 15. Living cell, characterized in that it contains a replicon as stated in one of claims 1-10. 16. Living cell as stated in claim 15, characterized in that it is a bacterium, a unicellular eukaryotic organism, an animal cell or a plant cell. 17. Cell culture, characterized in that it contains a living cell as stated in claim 16. 18. DNA fragment, characterized in that it comprises a sequence which, when the DNA fragment is inserted into a replicon, codes for a product which can kill the cell containing the replicon or the progeny of the cell (or codes for a precursor of the product) and a sequence which codes for a antagonist of the killing product (or encodes a precursor of the antagonist), the antagonist being one that suppresses the killing product (or a precursor of the killing product) in cells containing the replicon, while the antagonist activity decreases when the replicon is lost from the cell so that the antagonist (or its precursor) is no longer continuously expressed, meaning that the killing product (or its precursor) present in the now replicon-free cell is no longer suppressed by the antagonist, leading to cell death, provided that the DNA fragment does not include R1 parB -region when the replicon carrying the fragment is a plasmid that can replicate in gram-negative bacteria. 19. DNA fragment as stated in claim 13, characterized in that it comprises a sequence which, when the DNA fragment is inserted into a replicon, codes for a messenger RNA which codes for a product which can kill cells containing the replicon or the progeny of the cell (or encodes a precursor of the product) and a sequence encoding an antagonist that prevents the translation of the messenger RNA in cells containing the replicon. 20. DNA fragment as stated in claim 19, characterized in that the antagonist is an antisense RNA molecule which, by base pairing to at least part of the messenger RNA that codes for the cell-killing product, prevents translation of this messenger RNA. 21. DNA fragment as set forth in one of claims 18-20, characterized in that it comprises a sequence that codes for a product (or a precursor of the product) that can kill the cell or its progeny in a manner that is identical to or analogous to with the product which (or whose precursor) is expressed by the R1 parB hok gene. 22. DNA fragment as stated in claim 21, characterized in that the sequence which codes for the product which can kill the cell or its progeny (or a precursor of the product) comprises the R1 parB hok gene or a sequence which is homologous thereto. 23. DNA fragment as set forth in one of claims 18-20, characterized in that it comprises a sequence that codes for an antagonist that functions in a way that is identical to or analogous to the antagonist expressed by the R1 parB sok gene. 24. DNA fragment as stated in claim 23, characterized in that the sequence which codes for the antagonist comprises the R1 parB sok gene or a sequence which is homologous to this. 25. Method for producing a replicon as stated in claim 1, characterized in that it comprises inserting, in a replicon, a sequence coding for a product capable of killing a cell containing the replicon or the progeny of the cell (or coding for a precursor of the product) and a sequence coding for an antagonist of the killing product (or encodes a precursor of the antagonist), the antagonist being one that represses the killing product (or a precursor of the product) in cells containing the replicon, while the antagonist activity decreases when the replicon is lost from the cell so that the antagonist (or its precursor) is no longer continuously expressed, which means that the killing product (or its precursor) present in the now replicon-free cell is no longer suppressed by the antagonist, leading to cell death, with the proviso that the cell killing product as well as the antagonist for this are not encoded for at the R1 parB region when the replicon is a plasmid that can replicate in gram-negative bacteria. 26. Method as stated in claim 25, characterized in that a DNA fragment as stated in claim 15 is inserted. 27. Method as stated in claim 25 or 26, characterized in that the replicon is a bacterial plasmid, a bacterial chromosome, a eukaryotic plasmid, a eukaryotic virus, a eukaryotic autonomously replicating sequence carrying a chromosomal origin of replication, a eukaryotic mitochondrial DNA molecule or a eukaryotic chloroplast molecule. 28. Method as stated in one of claims 25-27, characterized in that the sequence or sequences that are inserted is a sequence or sequences of DNA (or possibly RNA) originating in a cell that is related to the replicon. 29. Method as stated in claim 28, characterized in that the sequence or sequences is a sequence or sequences that are homologous to the R1 parB region. 30. Method as stated in claim 28, characterized in that the sequence or sequences is a sequence or sequences originating in the R1 parB area. 31. Method for producing a gene product coded for by a nucleic acid sequence, characterized in that it comprises growing cells containing a replicon as stated in one of claims 1 - 10, and harvesting a gene product expressed from the replicon. 32. Method as stated in claim 31, characterized in that the cultivation is carried out for at least 100 cell generations. 33. Method as stated in claim 32, characterized in that the loss of the replicon is less than 10~<4> /cell/generation. 34. Method as stated in claim 33, characterized in that the loss of the replicon is less than 10~<5> /cell/generation. 35. Method as stated in claim 34, characterized in that the loss of the replicon is less than 10- <f> Vcell/generation.