WO2004020627A2 - Improved dna and proteins - Google Patents

Abstract

The invention relates to methods utilizing retro-elements for in vivo mutagenesis and directed evolution to produce DNA sequences, RNA sequences, and proteins with new and improved properties. A retro-element is a genetic unit that replicates through a reverse transcription step. Suitable retro-elements with mutagenizing properties are obtained through engineering and directed evolution such that they replicate independently of integration into the host genome. In a method of producing DNA coding for a protein with new or improved properties, a subject DNA is inserted into a suitable retro-element vector, introduced into host cells, these host cells are multiplied, host cells expressing proteins with a desired property are selected, and the DNA isolated. The use of the mutagenizing retro-element vector permits the combination of DNA mutagenesis, recombination, expression of protein and functional selection in one cell, allowing to create higher diversity and to perform continuous directed evolution.

Description

Improved DNA and Proteins

Field of the Invention

The present invention relates to a method of in vivo mutagenesis and directed evolution for the production of DNA with a desired function, in particular DNA coding for proteins with improved properties, and proteins prepared by this method.

Background of the Invention

Industrially produced proteins play an important role in many types of applications. Protein pharmaceuticals, such as cytokines, growth factors, soluble receptors, enzymes, antibodies, and the like, have established themselves as safe and potent drugs for diseases. Other applications of proteins are as food and feed additives, in cosmetics, or as enzymes used for degradation of particular substrates, e.g. in paper industry or as household detergents. The existing proteins of this type are either found as such in the natural environment, or are adapted to the special needs by starting with a protein found in nature and adapting it by alterations of the amino acids making up the protein.

Based upon structure-function information and general experience with proteins, specific modifications of proteins have also been accomplished by targeted modification of DNA coding for such proteins. This approach is limited to proteins where the necessary detailed information is available.

Since the relationship of protein structure and its function is complex and only partially understood, directed modifications based on rational protein design have only limited effects. Therefore methods have been developed for random modifications of DNA coding for proteins and screening of modified protein products in order to identify particular modifications which provide desired additional properties or enhancement of existing properties of proteins of interest.

Random modifications of existing proteins have been accomplished by classical mutagenesis of microorganisms producing such proteins, e.g. by X-ray or UN radiation, or chemical mutagenesis, resulting in mutagenesis of the whole organism. This approach may be especially suitable for the evolution of complex traits involving several genes and proteins. The use of high mutator strains, that have an error-prone DNA polymerase combined with reduced DNA repair activity, is another possibility to mutagenize whole organisms including the genes to be evolved. The disadvantage of these methods is that they are not specific for the sequences of interest and damage the genome of the host organism. The advantage is that the mutations are generated in vivo, and the resulting protein variants are directly expressed.

Through error-prone polymerase chain reaction (PCR), specific DNA sequences can be mutagenized in vitro. For expression of the protein variants, the mutagenized DNA is assembled in a vector and introduced into a cell, or proteins are expressed in vitro through ribosome display. The functional diversity of the proteins that can be screened for an enhanced property is limited by transformation efficiency in the former case, and proper protein folding in the latter. Also, iterations of the procedure are slow and require laborious experimentation.

Retroviruses are known to adapt very efficiently to external selection pressure. For example HIV is known to continuously escape from the human immune response, and to become resistant to drugs designed to restrict its function. This property is based on the error-prone character of retrovirus replication, proceeding through the reverse transcription of the retrovirus RNA to a complementary DNA. The inherently mutagenic replication of retro- viruses is specific for its own RNA. The ability of retroviruses to quickly evolve was used to improve the attributes of the viruses themselves. US 6,168,916 describes a selection process for retroviruses to improve their infectivity, resistance to shear forces, and resistance to human serum. How to use HIV to select a functional sequence element required for virus replication is shown by Berkhout (1993). Even though these examples hint at the power of retrovirus evolution, they remain restricted to elements related to the retrovirus life cycle and required for retrovirus survival itself. The present invention shows entirely new ways of evolving retro-elements towards vectors that have a broad range of applications for the directed evolution of genes and proteins.

Attempts have been made to mutate specific DNA sequences in vivo. WO 97/25410 describes the use of an error-prone DNA polymerase acting on a certain region of a plasmid vector, but only a low mutagenesis rate is achieved. The plasmid vector presented is constructed through rational engineering. Also mentioned is the possibility to use retroviruses or retrotransposons for the same purpose, but no indications are given on how to do that. The need for a new method to adapt retro-elements for use as mutagenizing vectors is evident.

Another system, described in WO 01/70946, achieves a high in vivo mutagenesis rate, but is restricted to one cycle and cannot be achieved iteratively over several generations without intervention by the experimenter. In addition, the heavy bias towards mutation of certain nucleotides severely limits the utility of this system.

The methods to improve proteins by evolutionary approaches are usually grouped under the title of "directed evolution". This is based on the Darwinian principle of generation of diversity and selection of improved variants. Whereas in nature these selection processes take a very long time, mimicking these processes on a laboratory scale and using proper selection criteria, these general evolutionary processes may be accomplished in a time frame of days or hours, opening a way for efficient selection of proteins with new or enhanced functions.

The existing methods of improving protein properties by random modifications and selection for proteins with desired properties have been demonstrated to work in principle. However there is a need for further methods simplifying the combination of random modification and selection in order to have broader applicability and higher efficiency, and to provide a method for in vivo mutagenesis wherein the DNA is mutated in the cell and directly translated into protein. It is further important to have methods at hand which allow a broader diversity in successful mutations.

Summary of the Invention

The present invention relates to methods utilizing retro-elements for in vivo mutagenesis for the directed evolution of DNA sequences, RNA sequences, proteins, and functional entities composed of DNA and/or RNA and/or protein, e.g. vectors, viruses, or enzyme systems to performs series of reactions for a biochemical pathway, in order to produce compounds and functional entities with new and improved properties.

A "retro-element" is a genetic unit that at one point in its life cycle replicates through a reverse transcription step. "Reverse transcription" means the synthesis of DNA on an RNA template, effectively producing a complementary DNA of the RNA template. Examples of such retro-elements according to this definition are listed hereinbelow.

A "free retro-element" is a retro-element that in the cell is found physically separated from other genetic elements or the genome, and depends on its own replication through reverse transcription for reproduction.

A "mutagenizing vector" is a vector that during replication mutagenizes sequences inserted into it.

A "retro-element derived vector" is a vector derived from whole or parts of a retro-element for use as a mutagenizing vector. A retroplasmid is an example for a retro-element derived mutagenizing vector.

A "plasmid" is a linear or circular genetic element, which autonomously replicates within a host cell, and is physically separated from the host genome.

A "retroplasmid" is a plasmid that replicates through a reverse transcription step.

The "stability of replication" of a plasmid is defined as the fraction of the host cells that harbor the plasmid with a selection marker, the plasmid not being integrated in the host genome, after a defined number of host cell divisions under growth conditions selective for presence of the selection marker. "Stable replication" of a retroplasmid means that after several (e.g. between 3 and 40) divisions of the host cells harboring the retroplasmid with a selection marker, the retroplasmid can be found in the majority of cells (e.g. in between 70% and 100% of cells) not integrated into the host genome.

A "DNA sequence function" means a function that originates from a DNA sequence, for example protein products and their functions encoded by a sequence, RNA transcripts and their functions encoded by a DNA sequence, functions of the DNA itself like the ability to bind transcription factors or take on a secondary structure, and functions resulting from the interplay of different factors and functions encoded by a DNA sequence, like entire vectors or biochemical pathways, or, in principle, whole organisms. A "subject sequence" means the DNA sequence that is the subject of mutagenesis on replication, and from which new or improved variants are formed in the process of replication. The subject sequence may encode a function with defined properties, but may also represent a plurality of sequences in the form of a library of DNA sequences. "Subject protein" is the protein product encoded by a subject sequence.

"Sequence similarity" means that two sequences are sufficiently similar that homologous recombination can occur between the two sequences in the cell type used for the experiment. Sequences with a divergence of more than 40% at the nucleotide level are not sufficiently similar for efficient homologous recombination to occur.

"Rate of mutagenesis of at least 10^"6 per nucleotide per round of replication" includes rates of mutagenesis of 10^" per nucleotide per round of replication and higher, e.g. about 1 x 10^" , 1 x 10^"5, 1 x 10^"4, or 1 x 10^"3, or higher. Rates of mutagenesis are calculated according to Preston (1996).

The method of the invention is e.g. applied to the preparation of a modified DNA having a desired function, in particular a modified DNA coding for a retro-element with a desired property.

In one aspect of the invention, the directed evolution approach is itself applied to a retro- element. The invention comprises a method for the preparation of a retro-element with a desired property, wherein a retro-element is inserted into host cells, expressed, and free retro- elements are isolated, in particular a method for the preparation of a retro-element with changed replication properties, wherein a retro-element is inserted into host cells, expressed, and free retro-elements are selected.

A retro-element, which already may be suitably adapted by genetic engineering, is modified through directed evolution such that it e.g. can replicate independently of integration into the host genome. For this purpose, a suitable retro-element is inserted into host cells, the host cells grown and the retro-element expressed, and free retro-elements selected, the selected retro-elements optionally reinserted into fresh host cells for a new round of growth, expression and selection, and the cycle of reinsertion, growth, expression and selection optionally repeated for one or more times, each time adjusting the criteria used for selection until modified retro-elements that can efficiently replicate independently of integration into the host genome are found. Preferred retro-elements replicating independently of integration into the host genome are covalently closed circular retro-elements.

In a particular aspect of the invention, retro-elements are selected that stably replicate as retroplasmids. Preferred retroplasmids are covalently closed circular.

Another aspect of the invention relates to a method of producing DNA having a desired function. In particular the invention relates to a method of producing DNA coding for a protein with a new or improved property, characterized in that a subject sequence to be mutagenized is inserted into a vector derived from a retro-element, the resulting modified vector is introduced into host cells, these host cells are multiplied, host cells expressing proteins with a desired property are selected, selected host cells are further multiplied, the cycle of multiplication and selection is repeated until the expressed proteins have reached a threshold of improved property, vectors are isolated and the DNA coding for the protein recovered.

The invention further relates to a simplified directed evolution process, wherein use of a mutagenizing vector derived from a retro-element permits the combination of DNA mutagenesis, recombination, expression of protein and functional selection in one cell. This greatly simplifies the directed evolution process, allowing to create higher diversity and to perform continuous directed evolution.

In one particular aspect the invention relates to a method of producing DNA coding for a protein with new or improved properties, characterized in that a subject sequence to be mutagenized is inserted into a vector derived from a retro-element, the resulting modified vector is introduced into host cells, these host cells are grown and multiplied under conditions of a continuous constant or increasing selection pressure for desired properties of the protein, surviving cells are separated, vectors are isolated and the DNA coding for the protein recovered.

In a particular embodiment, the invention relates to a method of producing DNA coding for a protein with an improved affinity to a target system characterized in that a DNA coding for a protein with a basic affinity to said target system is inserted into a retro-element vector, the resulting modified vector is introduced into host cells, these host cells are multiplied, host cells expressing proteins with a desired affinity to said target system are selected, the selected host cells further multiplied, and the cycle of multiplication and selection is repeated until the expressed proteins have reached a threshold of improved affinity to said target system, vectors are isolated and the DNA coding for the protein is recovered.

Another aspect of the invention relates to diversified DNA sequences and proteins, the diversity being generated by retro-elements, retro-element derived vectors and retroplasmids through mutagenesis and recombination in vivo.

The invention also relates to methods to select retro-elements with certain properties by separation and purification of free retro-element DNA from the rest of the host DNA.

In yet another aspect the invention relates to the preparation of retro-elements with altered properties through selection by separation and purification of free retro-element DNA from the rest of the host DNA. Preferred altered properties relate to the replication of retro- elements, in particular increased replication rate of retro-elements, the ability of retro- elements to replicate within a heterologous host, and the ability of retro-elements to replicate independently of integration into the host genome, for example the ability of a retro-element to replicate as a retroplasmid.

One aspect of the invention relates to the particular combination of retro-element and host cells used for retro-element evolution. Preferred are host cells that lack endogenous retro- elements with a sequence similarity to the retro-element to be evolved in order that no homologous recombination of endogenous and inserted exogenous retro-elements with each other can occur. Homologous recombination does not efficiently occur between sequences with more than 40% divergence at the nucleotide level (Drouin 2002).

The invention further relates to the recovery of purified retro-elements through a method of in vitro DNA amplification. A preferred in vitro DNA amplification method is the polymerase chain reaction (PCR).

The invention further relates to the method wherein the resulting DNA coding for a protein with improved properties is used to express the protein, and proteins so prepared. The invention further relates to the resulting retro-element derived vectors, and the host cells comprising such vectors.

In another aspect, the invention relates to retro-element derived vectors comprising DNA coding for a protein with a particular property, and to host cells containing such vectors.

In a further aspect, the invention relates to proteins with desired properties obtained by the described methods.

Detailed Description of the Invention

The invention relates to a process for obtaining DNA, RNA, peptides, or proteins through mutagenesis in vivo by use of a reverse transcriptase dependent vector.

The resulting peptides, DNA or RNA can have a range of new or improved properties, for example, structural, enzymatic, catalytic, antigenic or pharmacological properties, or properties of binding, and more generally, new or improved chemical, biochemical, or biological properties. Also, several peptides can be concomitantly evolved towards one of the above mentioned new or improved properties or a new or improved combined function, e.g. to perform a series of reactions for a biochemical pathway, or the replication of a vector. Particular proteins with enhanced properties can be, but are not limited to, antibodies with higher affinity, receptor ligands eliciting higher responses, like growth factors, and biosynthetic enzymes catalyzing specific reactions.

Retro-elements are genetic parasites that inhabit the genomes of all eukaryotes and many prokaryotes. Retro-elements include retroviruses, endogenous retroviruses, LTR retro- transposons, poly(a) retrotransposons, pararetroviruses, retroplasmids, retro-introns, retrotranscripts, and retrons (Coffin 1997). They have in common that at one point in their life cycle their genome is copied from RNA into DNA by reverse transcription. This process is carried out by a reverse transcriptase (RT), usually encoded by the retro-element.

Preferred retro-elements are LTR retrotransposons, ρoly(a) retrotransposons, retroviruses, retroplasmids, retro-introns, retrotranscripts, and retrons, in particular LTR retrotransposons, poly(a) retrotransposons, retroplasmids, retro-introns, retrotranscripts, and retrons. Most preferred are LTR-retrotransposons and retroplasmids.

The LTR-containing retrotransposons comprise a large family of elements that have been identified in all well-studied eukaryotic nuclear genomes. Structurally, their genome includes many retrovirus-like features. LTR-retrotransposons and retroviruses have in common that they are flanked by long terminal repeats (LTRs) and therefore can circularize through homologous recombination of their LTRs. The very large number of LTR retrotransposons isolated thus far segregate phylogenetically into two groups, called the Tyl -copia family and the Ty3-gypsy family. The families can be differentiated both in terais of gross structure and by phylogenetic comparisons of conserved (mostly reverse transcriptase) protein sequences. Both Tyl and Ty3 from the yeast Saccharomyces cerevisiae have been extensively studied (Boeke 1991). Preferred LTR retrotransposons are selected from the Tyl-copia group.

Retroplasmids derive from a completely different branch of the extensive retro-element family than the Ty retrotransposons. Most retro-elements inhabit the cytoplasm and nuclei of host cells; however, a few retro-elements live in organelles. These are members of the "prokaryotic" class of retroplasmids. Mauriceville and Varkud plasmids were isolated from the mitochondria of certain strains of N. crassa, in which they replicate independently of integration into the mitochondrial DΝA through a reverse transcriptase encoded on the plasmid (Kuiper 1988).

It is an important aspect of the invention that the mentioned retrotransposons or retroplasmids use reverse transcriptases as opposed to a DΝA polymerase using DΝA as a template. Reverse transcriptase is less likely to interfere with the DΝA replication of the host cell than an error-prone DΝA polymerase, especially if reverse transcription occurs in specialized virus-like particles in the cytoplasm. Replication is not influenced by the DΝA repair mechanisms of the host cell, which would eliminate or substantially diminish diversity created in the replication process. RΝA-DΝA duplexes produced by reverse transcriptase are likely not to be recognized by the usual DΝA repair machinery. In addition, reverse transcriptases generally have a higher error rate than normal DΝA polymerases.

It is also important to have DΝA as an end product of mutagenesis, and not RΝA. First, RΝA is impractical to work with due to its inherent instability. For most traditional cloning and analysis work, the RNA would have to be reverse transcribed into DNA at some point. Second, recombination frequency between individuals of homologous RNA sequences is very low or even not existent, but there are many cellular systems for DNA homologous recombination. Homologous recombination is an important aspect of any evolutionary process.

The invention relates to the use of retro-elements in the general field of directed evolution. Directed evolution relates to a collection of methods to specifically improve a molecule or a trait of an organism through the Darwinian principle of generation of diversity and selection of improved variants. For example, for the directed evolution of a protein, the first step is the preparation of a diverse set of DNA coding for protein variants, either starting from a DNA coding for a protein with a known function, or also creating a totally random library. The second step is the expression or "display" of these proteins, meaning the translation of the DNA information into actual protein molecules. The third step is the identification of the proteins with a threshold level of a desired property and the isolation of the DNA coding for these proteins. Finally, the recovered DNA is amplified and again diversified for the next round of selection. The procedure is iterated until the desired level of a property is reached. The integration of these steps through mutagenesis in vivo is an essential feature of the invention. Because mutagenized DNA does not need to be introduced into cells by transformation, much higher diversity can be achieved. Furthermore, several cycles of selection and mutagenesis can be performed without intervening isolation of the DNA and retransformation into new host cells, eventually allowing for continuous evolution.

In a first phase, the method of the invention is applied to generate vectors derived from retro- elements suitable for use as a mutagenizing vector. In a second phase, subject sequences are inserted into these vectors for mutagenesis and identification of variants with desired properties. Although it may be possible in certain situations to combine the two phases in one, the above mentioned sequence is a preferred embodiment of the invention. Both phases follow essentially the same general process. Therefore, the two phases are not explicitly differentiated in the following description of this general process.

This general process is now described in relation to the invention. Assembly of retro-elements, insertion of a subject sequence

There exist several alternatives to harness retro-elements for use in directed evolution experiments. Generally, a retro-element is modified or assembled using methods of genetic engineering well known in the art, based on knowledge of the functioning of a retro-element and its host. The retro-element sequences are changed to enhance its function as a mutagenizing vector, for example changing certain amino acids, and/or omitting parts of the sequence, and/or adding new features. Alternatively, a mutagenizing vector is obtained by a directed evolution approach, by generating a variety of mutated vectors, and selecting from that variety the vectors with enhanced properties as a mutagenizing vector. This process is iteratively repeated with increasing selection pressure until the adapted vector fulfils the requirements set for a mutagenizing vector. Although both approaches are possible individually, it is a preferred embodiment of this invention to use a combination of rational engineering of an initial mutagenizing vector, e.g. exchange of the natural promoter for a more active or regulatable one, followed by improvement through directed evolution.

Retro-elements coexist in equilibrium with their natural host, with a replication rate ensuring their proliferation, but at the same time checked by the need of survival of the host. Therefore, retrotransposons have a very low rate of replication and transposition, because every transposition event into a new site changes and possibly damages the host genome. If a retro- element is obtained with a higher activity and is reintroduced into its natural host, it recombines through homologous recombination with the present endogenous wild-type retro- elements and the higher activity is lost. To overcome the problem of recombination with endogenous retro-elements, the retro-element of the invention is expressed and replicated in a heterologous host. The sequences of any retro-elements of the heterologous host and the retro- element introduced should have enough difference to avoid homologous recombination. Different organisms have very different efficiencies for recombination between divergent, but similar sequences (Modrich 1996), but in S. cerevisiae sequences with more than 40% divergence on the nucleotide level have a low occurrence of recombination (Drouin 2002). Once they have entered a host cell, retro-elements can replicate even in distant heterologous hosts (Nakayashiki 1999, Lucas 1995, Hirochika 1996). For example, the Tyl retrotransposon of S. cerevisiae replicates in the yeast P. pastoris. In favorable combinations of heterologous retro-elements and hosts, a wild-type retro-element is sufficiently active to serve as a starting point of improvement by directed evolution. But retro-elements often have a low replication rate in their natural host as well as in a heterologous host. To increase the replication rate, a retro-element or a host is engineered by rational means based on knowledge of the functioning of a retro-element and the host, for example by introduction of a promoter known to be active in the chosen host. Numerous promoters of many hosts are known in the art. Also, an artificial generic promoter can be designed by a person skilled in the art. For certain directed evolution schemes it is advantageous to have a regulatable promoter, such that expression of the retro-element can be turned on or off at a chosen time point. For example, when the expression of a retro-element is putting a large strain on the host, it is advantageous to grow the host before expression of the retro-element. Regulatable promoters are well known in the art and can depend on the type of carbon source, e.g. repression by glucose, activation by galactose or by methanol, or by lactose or an analogue like IPTG. Other regulatable promoters that can be used rely on the tet repressor system, glucocorticoid hormone receptor, the metallothionein promoter, or are regulatable by temperature. If no initial activity of a retro-element is detected in a heterologous host, the retro-element is mutagenized already initially and active variants isolated through one of the methods detailed herein below. Active variants are then further improved through their replication and mutagenesis through reverse transcription, and eventually used to mutagenize a subject sequence introduced into the thus derived retro-element. For example, the Mauriceville retroplasmid from the mitochondria of N. crassa is mutagenized and variants active in E. coli are isolated as detailed further below.

Another way to overcome the problem of recombination of the altered retro-element with endogenous retro-elements in the natural host is to eliminate endogenous retro-elements by eliminating them whole or in part through homologous recombination. Inactivation of specific sequences by homologous recombination ("gene knockout") is a procedure well known in the art and established for a variety of organisms.

Parts of a retrovirus may be used to assemble a retro-element able to replicate within a host. Since uninfected host cells do not contain the retroviral sequences, there is no need to use a heterologous host, and the natural host can be used. For example, sequences required for maturation and export of the virion are deleted, such that the remaining parts of the retrovirus replicate like a retrotransposon. In a further step, the integrase function can be impaired to avoid integration into the host genome and favor replication as an extrachromosomal vector. Retroviruses can directly be used to mutagenize a subject sequence, by inserting a subject sequence into the virus genome at a suitable location. Viruses are isolated carrying variants of the subject sequence with desired properties. This is for example a protein expressed on the surface of the virion, with a threshold affinity to a target system. A maximum of viral trans- acting factors are transferred to the host organism for expression, the remaining retrovirus containing only minimal cώ-acting sequences required for replication, to allow for elevated mutagenesis rates.

Methods for inserting DNA into vectors are well known in the art. For any vector, heterologous DNA has to be inserted such as not to interfere with the basic vector functions. Suitable locations are in between known functional sequence elements, like open reading frames or cts-acting DNA sequences like promoters or origins of replication. In particular, the Tyl retrotransposon contains a Bglll restriction site, after the end of the sequences coding for retrotransposon proteins, and before the 3 -LTR, where heterologous sequences have successfully been inserted (Boeke 1985). Proteins inserted at that site can have their own promoter driving transcription of a mRNA independent of the retrotransposon RNA. It is also possible, though, to express subject proteins to be mutagenized as fusions with retrotransposon proteins, with or without a protease cleavage site at the point of fusion. Alternatively, a frameshift signal or internal ribosome entry site can be used to direct the expression of a subject protein from the retrotransposon RNA. It is also possible to insert subject sequences in suitable locations within the LTR, specifically just before the beginning of the promoter in U3. Additional locations can be identified by inserting a heterologous sequence at a chosen or random location, transforming the vector into its host and selecting for functional vectors. This is especially easy if the heterologous sequence used encodes for an antibiotic resistance gene.

The Mauriceville retroplasmid contains a large open reading frame coding for a reverse transcriptase. Since the functions of the intervening sequences are less well characterized, subject sequences are preferably inserted just before or after this open reading frame. These can have their own promoter or just an additional ribosome attachment site (internal Shine Dalgarno sequence). It is also possible to coordinate expression of the reverse transcriptase with a subject protein through translational coupling. Again, additional locations can be identified by inserting a heterologous sequence at a chosen or random location and testing the viability of the such modified vector. It is possible to delete certain elements present on a wild-type retro-element and to independently express these elements in a host cell, to provide them in trans for the replication of the retro-element (Xu 1990). This is independent of the fact whether a retro- element replicates through an integration step into the host genome or not. Minimal sequences required in cis can be determined by deletion analysis, progressively removing sequences of a retro-element while providing in trans all the elements normally expressed by the retro- element. A retro-element derived vector replicating independently of integration into the host genome can be further optimized as a mutagenizing vector this way. For some types of directed evolution protocols it is of advantage to have a minimum of cts-acting sequences on a mutagenizing vector, because all the sequences present on the vector itself are mutagenized. If there are less functional sequences on the vector, there is less chance of them being impaired by the mutagenesis and higher rates of mutagenesis can be applied. Or, in other words, fewer variants will be lost just due to mutations in the sequences necessary for vector functioning.

Introduction into host

Modified retro-elements are introduced into host cells by methods well known in the art, e.g. by electroporation, packaging into phage, chemical methods using DMSO, PEG, lithium acetate, lithium chloride, calcium chloride, or combinations thereof, also calcium phosphate transfection, DEAE-dextran mediated transfection, cationic lipid-mediated transfection, transduction, infection, or other methods. Such methods are described in many standard laboratory manuals, such as in Ausubel et al., Current Protocols in Molecular Biology (1987).

For easy selection of transformed cells, a marker gene is inserted into the retro-element, for example a gene conferring antibiotic resistance, or a gene for an essential step in a biosynthetic pathway.

Suitable host cells are from any taxonomic origin, including archaea, eubacteria, and eukaryota. Preferred host cells from eubacterial origin are E. coli. Preferred host cells from eukaryotic origin, are ascomycote fungi. Preferred ascomycote fungi are the hemi- ascomycetous yeast. Preferred hemi-ascomycetous yeast are from the family of the Saccharomycetaceae. Preferred Saccharomycetaceae are from the genus Saccharomyces and from the genus Pichia. Preferred species are Pichia pastoris, and Saccharomyces cerevisiae. Likewise preferred are host cells from eukaryotic origin, such as host cells of the phylum chordata. Preferred chordates are mammals. Preferred mammalian cell lines are the mouse NIH 3T3 cells, ratl cells, the monkey COS7 cells, and the human Jurkat and 293T cells.

Preferred host cells are the hemi-ascomycetous yeasts. Preferred hemi-ascomycetous yeasts belong to the genus Pichia.

Generation of diversity: Mutagenesis and recombination

Replication by reverse transcription is mutagenic, with mutation rates ranging from 10^"4 to 10^" mutations per nucleotide per replication cycle for retroviruses and the Tyl retrotransposon (Preston 1996). Growing cells containing a retro-element and expressing that retro-element yields a variety of mutant retro-elements. Variants with specific new or enhanced properties can then be isolated. Within an individual cell, recombination between different variant retro- elements occurs, on the one hand through regular recombination between the double-stranded retro-element DNAs, through action of the recombination mechanisms of the host cell, on the other hand through template switching during reverse transcription of eukaryotic retroviruses and LTR-retrotransposon (Hu 1997, Zhang 1993).

It is also possible to combine mutagenesis by reverse transcriptase with other methods of diversification. These include error-prone PCR or any of the methods for in vivo or in vitro DNA recombination known in the art, including DNA shuffling and step recombination. For example, it may be advantageous to start a directed evolution experiment with an initial diversity created by one of the above mentioned methods, insert this diversity into the in vivo mutagenizing vector and introduce the vector into a host, for further diversification and selection. Or, after a few rounds of diversification and selection through use of the mutagenizing vector, the obtained sequences could be recombined through use of one of the above mentioned DNA recombination methods, in vivo or in vitro.

Screening/selection

"Screening" is, generally speaking, a two-step process in which one first determines which cells, organisms or molecules, do and do not express a detectable marker, or phenotype (or a selected level of marker or phenotype), and then physically separates the cells, organisms or molecules having the desired property. Proteins can be screened via optical methods, like absorbance, fluorescence, luminescence, change of wavelength, color reactions, etc. Positives must then be physically isolated somehow, for example through picking, or by cell sorting (FACS). These processes can be automated in high-throughput screenings.

"Selection" refers to the entire process of identification and isolation. Single-step selection is applied when the method of identification is also the method of isolation. If the desired property of the subject protein is combined with the ability of the cell to grow, host cells are grown under special conditions exerting a selection pressure, and only those cells will multiply which express proteins with the desired threshold property. This method is particularly preferred if the proteins are intracellularly expressed or incorporated into the cell membrane. This way, for example, proteins with the ability to perform a biochemical reaction vital for cell growth, or proteins conferring resistance to a toxic substance, e.g. an antibiotic, can be identified. Through engineering, growth can be conditionally linked to a normally non- vital function, like the interaction of two proteins in the yeast two hybrid system or the protein-fragment complementation assay (Remy 1999). If a host cell and procedure is chosen wherein the protein expressed is displayed at the cell surface or transported into the extracellular space, the usual selection methods can be successfully applied where the affinity to a target system is used as a selection criteria and cells are identified that carry DNA expressing proteins with a threshold affinity to the target system.

For example, proteins can be selected based on their binding properties to another protein, an antigen or chemical structure. For this, it is convenient to express the protein on a cell surface, keeping it linked to its genetic information contained inside the cell. A variety of methods are known for cell surface display of proteins on yeast, bacteria, and phage (reviewed in Wittrup 2001, Sidhu 2000). For example, a sequence coding for a single chain antibody (scFv) is fused with the sequence required for cell surface display. This sequence is inserted into the retro-element derived vector and transformed into the host cell. On the one hand, it is possible to use the sequence of a specific antibody that has some affinity to an antigen that needs to be improved (affinity maturation). By growing the host cells and through replication of the retro- element derived vector, variations of the original antibody are created that can be selected for improved affinities. On the other hand, a library of different antibodies can be inserted into the retro-element derived vector and transformed into a multitude of host cells, each cell containing one or a few different antibodies. Growth and replication of the retro-element derived vector will increase the variety of the library. In both cases, antibodies specifically binding to a chemical structure of interest can be separated from the antibodies with low affinity simply by providing the chemical structure on a solid support, mixing with the antibody library, and washing to eliminate the non-binding antibodies. The solid support can be in the form of beads, plates, columns, or dishes. After washing, the binding antibodies (still connected to the cells they originate from) are eluted, the cells further grown with replication of the retroplasmid, which introduces mutations in the antibody genes that then can be selected for higher affinity in further rounds of binding and washing. The procedure can be iterated until the desired affinity is reached, every time adjusting the binding and washing conditions to select for higher affinities. This procedure provides an easy means of producing high affinity antibodies.

Proteins for use as pharmaceuticals can be improved through the method of the invention. Proteins used as pharmaceuticals are for example antibodies, enzymes, growth factors, cytokines, peptide hormones, soluble receptors, vaccines, parts of proteins from human origin or derived from an infective agent, like a pathogenic microorganism or virus. Improved pharmaceutical proteins can be isolated according to their ability to attach to a pharmacological target, elicit or inhibit a cellular response, escaping recognition by the immune system, or, in the case of a vaccine, eliciting a response by the immune system. Many other pharmacological mechanisms are known in the art, and many of them are accessible to improvement by the method of the invention. As an example, the identification of peptides eliciting a receptor response is outlined: First, the activation of a cellular receptor is linked to a read-out. Many signaling pathways relaying receptor activation are well characterized, and receptor activation can be linked to transcription of a reporter gene, like the green fluorescent protein. If receptor activation involves dimerization of two receptor molecules, dimerization can be detected by the protein-fragment complementation assay (Remy 1999). Then, additional cells are provided, containing a retro-element derived vector with sequences for expression of peptides linked to the surface of the cells. Cells containing the receptor linked to the read-out are mixed with the cells expressing the peptides. Positives are isolated according to their read-out, for example through fluorescence-activated cell sorting (FACS) of cells expressing the green fluorescent protein. Cells expressing active peptides are physically linked to the cells expressing the receptor, due to the interaction of the receptor with the active peptide. Therefore, cells expressing active peptides are copurified, are then further grown to replicate the retro-element derived vector with the peptide sequences, and used for additional rounds of selection and growth, until peptides with the desired threshold activity towards the chosen receptor are obtained.

Whole organisms or functional entities can be screened or selected for a desired property. For example, viruses have been selected for their ability to infect a heterologous host

(US 6,168,916), or viruses have been selected for improved stability and production yields (Powell 2000). An important aspect of this invention is the selection of retro-elements with an elevated replication rate. Endogenous retro-elements, like retrotransposons, can be selected for an elevated replication rate by isolating free non-integrated retrotransposon DNA. Freshly replicated retro-element DNA in many cases resides in the cell for some time as free extra- chromosomal DNA before transposition or elimination through the host. Retro-elements with a high replication rate are selected by isolating this free DNA. In the case of the Tyl retrotransposon, part of the freshly reverse transcribed retrotransposon cDNA is known to circularize in the cell through homologous recombination of the LTRs, and this free circular DNA can be isolated (Eichinger 1988). The mutants with the highest retrotranscription activity are the ones producing the highest amount of free retrotransposon DNA of which a certain proportion will circularize. LTR retrotransposons and retroviruses are preferred retro- elements of the invention because they have the ability to form functional circular retro- elements through homologous recombination of the LTRs.

Since endogenous retro-elements have to coexist with their host, high replication rates must be accompanied by reduced integration to avoid damaging the host genome. Indirectly, non- integrating retro-elements can be selected on this basis. Selection is performed by separating free linear or circular retrotransposon DNA from integrated forms and the rest of the genome, e.g. by agarose gel electrophoresis, chromatography (e.g. size exclusion), or density gradient centrifugation. Covalently closed circular DNA is usually supercoiled in the cell. Covalently closed circular DNA can be isolated e.g. by CsCl density gradient centrifugation, or, in conjunction with a bacterial origin of replication for a plasmid through introduction into E. coli, where circular DNA is required for efficient transformation.

One preferred method is separation based on size of the DNA. A preferred size separation method is agarose gel electrophoresis. Another preferred method is separation due to differences in DNA topologies. A preferred topology difference is the separation of covalently closed circular from open circular and linear DNA. A preferred method to separate covalently closed circular from other DNA topologies by CsCl density gradient centrifugation.

Cells need a variety of chemical building blocks, like nucleotides, amino acids, or enzyme cofactors to sustain their survival and growth, as well as the ability to metabolize molecules as an energy source. These molecules are made through a series of chemical reactions catalyzed by enzymes. Lack of an enzyme in a vital pathway, like nucleotide synthesis, inhibits growth of the cell. For example, a step in the synthesis of the amino acid histidine can be eliminated by deleting parts of the yeast HIS4 gene. Deletion can be accomplished by replacing the functional copy of the gene with one that misses sequences through homologous recombination, a procedure well known in the art. These yeast cells are unable to grow in media lacking the amino acid histidine. DNA sequences encoding sequences similar to the ones coding for the missing enzyme activity are inserted into the retro-element derived mutagenizing vector and introduced into the host lacking the enzyme activity. If the initial population is large enough, it is even possible to provide a random synthetic library of DNA sequences inserted into the mutagenizing vector to start with. Cells are first grown under non- selective conditions to allow replication of the mutagenizing vector and mutagenesis of the inserts. The concentration of histidine in this example is then gradually decreased to favor growth of cells containing vectors with sequences able to produce a minimally active enzyme. By further decreasing the amount of histidine and finally omitting it, enzymes able to catalyze the missing biosynthetic step are evolved. This procedure serves as an example of continuous evolution, requiring minimal intervention by the experimenter, because by providing a limited amount of histidine at the outset, the stringency of the selection is automatically increased as the histidine concentration decreases during the course of the experiment, and the cells that harbor enzymes that are able to complement the decreased histidine concentration are able to grow, at the same time replicating the retro-element derived vector and producing improved variants of the enzyme. The same kind of experiment can be done by providing an energy source to cells that are not capable of metabolizing that source, and evolving enzymes from the mutagenizing vector to catalyze the degradation of the energy source.

Preferred selection pressure for a desired property in the method of the invention is e.g. the scarcity or lack of vital biosynthetic precursor molecule, antibiotic pressure, or affinity of the expressed protein to a target under constant washing conditions. Different selection methods may be combined in an alternating mode, e.g. selection for affinity of the expressed protein to a target alternating with selection based on temperature or pH.^'

DNA isolation/amplification

After selection, vectors containing DNA coding for the desired functions are isolated. Isolation of total DNA or plasmid DNA can be performed through one of the methods known in the art (Ausubel 1987). Depending on the amounts of the collected DNA and requirements of the procedure, the DNA can be amplified by one of the methods known in the art, e.g. the polymerase chain reaction (PCR) or cloning into a standard bacterial plasmid. Also, isolation and amplification can be performed concomitantly, for example by performing PCR directly on total DNA or even a crude cell lysate. The DNA can be analyzed by methods well known in the art and used for the next iteration of the directed evolution procedure.

Iteration of procedure

The cycles of selection, amplification and reinsertion continues until the function the DNA encodes, for example an expressed protein or a replicating vector, has reached the desired threshold of properties. The vectors are then isolated and the DNA coding for the desired protein recovered with methods known in the art, e.g. by PCR or ligation into a standard bacterial plasmid like pUC18.

It is an important aspect of the invention that the cycle of host cell growth and selection (growth comprising mutagenesis and amplification) can be repeated without needing to isolate and reintroduce the DNA. Nevertheless, depending on the specific goal of the directed evolution procedure and the properties selected, it may be necessary to isolate and reintroduce the DNA after one or several iterations of growth and selection.

For example, the procedure has to be iterated to develop a retrotransposon replicating independently of integration into the host genome. First, retrotransposons evolve that just have a higher rate of replication. After further rounds of amplification and selection of extra- chromosomal free retrotransposon DNA, the retro-element DNA has lost the ability to integrate into the genome by retrotransposition, and replication is sufficiently efficient to ensure the presence of the free extra-chromosomal retro-element over several generations. To further increase the stability of the evolved retro-element, it is grown over several generations and then purified and retransformed into fresh host cells. Possible combinations of retro- elements and host cells were discussed herein above. Host cells and retro-elements can be from almost any origin. Besides yeasts and bacteria, mammalian cells in general, and specifically the mouse NIH 3T3 cell line, as well as the human cell lines Jurkat and 293T cells, are preferred. For optimal stability, it may also be necessary to adapt the host strain. This is done by randomly mutagenizing the host strain genome, for example with a chemical mutagen and isolating the strains that harbor the highest amount of free retro-element, visualized by expression of the green fluorescent protein encoded by the retro-element.

Continuous directed evolution

If cells are grown under conditions of a continuous constant or increasing selection pressure for desired properties of the protein, the approach described here provides a simple method with vast numbers of mutations providing the basis for efficient property improvement of the protein without experimental intervention as required by a cycle of mutagenesis, selection, amplification and reinsertion of the vector. Examples of conditions exerting constant pressure are e.g. the lack of a vital biosynthetic precursor molecule, antibiotic pressure, or affinity under constant washing conditions. Increasing selection pressure can be applied e.g. by decreasing the concentration of said biosynthetic precursor molecule in the culture medium, increasing antibiotic concentration, or using more stringent washing conditions.

Said DNA may be used as starting material for further rounds of improvement of protein properties, e.g. applying other selection criteria. For instance a protein may be obtained by the method of the invention which has a desired threshold affinity to a target receptor and a desired thermostability obtained in a second round of property improvement.

The method of the invention can also be applied to create a useful candidate retrotransposon to be used in further inventive production of DNA coding for protein with desired properties. Through iterative selection and amplification of non-integrating abundant variants of a retrotransposon a vector that can be successfully used for mutagenization is obtained. Free retrotransposon variants are isolated from a host cell and amplified by the polymerase chain reaction (PCR), then reinserted into a new host cell, these cells are grown, and then again free retrotransposon elements isolated. In another application of the method of the invention the retroplasmid found in the mitochondria of the fungus Neurospora crassa can be modified to yield a new retroplasmid being able to replicate in E. coli and other suitable bacteria.

Furthermore, the method of the invention can be used to derive retro-element vectors from pathogenic retroviruses, like HIN. These derived vectors can be used in drug development, screening for inhibitors of retro-element replication. In addition, functional mutants of said vectors derived from pathogenic retroviruses can easily be generated and assessed with respect to resistance to anti-retroviral drugs. This way, drugs can be identified active against a wider range of mutant retroviruses.

The following examples are useful to further explain the invention but in no way limit the scope of the invention. A particular retro-element-derived mutagenizing vector is obtained through directed evolution of the Tyl retrotransposon from S. cerevisiae in the heterologous host P. pastoris. Another retro-element-derived particular mutagenizing vector is obtained through host adaptation of the Mauriceville retroplasmid from N. crassa in E. coli.

Brief Description of the Figures

FIG. 1: Schematic representations of plasmids used for the directed evolution of the Tyl retrotransposon in Pichia pastoris. Arrows indicate the transcriptional direction of the corresponding elements. Hatched boxes denote additional intervening Tyl sequences. FIG. 1 A shows the construct with the initial engineered Tyl retrotransposon under the control of the AOX promoter with the pGAP-LTR. FIG. IB shows the plasmid used as a template for the left arm of the overlap extension PCR. FIG. 1C shows the plasmid used as a template for the right arm of the overlap extension PCR.

FIG. 2 shows the sequence of the pGAP LTR (SEQ ID ΝO:8). "TyH3" denotes sequences originating from plasmid pGTyH3, "pGAP" denotes sequences originating from the Pichia pastoris GAP-promoter, "S" denotes sequences originating from synthetic oligonucleotides. Restriction sites Clal, Spel, and Xhol are indicated. U3, R, and U5 denote the respective LTR element. FIG. 3: Overlap extension PCR for the assembly of a linear Tyl construct for the transformation of Pichia pastoris. The lines below the schematic representation of the Tyl linear construct signify PCR products, the numbers on the ends of each line identify the primers used for the PCR reaction. Hatched boxes denote additional intervening Tyl sequences.

Primer 1 (SEQ ID NO:l) 5'-TGACGAACATTGTCGACAATTGGT-3' Primer 2 (SEQ ID NO:2) 5 '-GTTGGGATTCCATTGTTGATAAAGGCTA-3 ' Primer 3 (SEQ ID NO:3) 5'-GAAGTCCACACAAATCAAGATCCGT-3' Primer 4 (SEQ ID NO:4) 5'-GACAATGTTCGTCAAAATGGTGAC-3' Primer 5 (SEQ ID NO:5) 5'-TACGCGATCGCTGTTAAAAGGACAA-3'

Primer 6 (SEQ ID NO:6) 5'-CGCCGGTTGCATTCGATTCCTGTTT-3 ' Primer 7 (SEQ ID NO:7) 5'-CGTTTTCTGGATAGGACGACGAAG-3 '

FIG. 4 is a schematic representation of a plasmid containing an adapted Mauriceville retroplasmid with duplicated promoter regions at the 5 '- and 3 '-ends. "P" means promoter region, "RT" means reverse transcriptase open reading frame, "β-lam" means β-lactamase open reading frame.

Examples

Example 1: Evolution of a retrotransposon replicating independently of integration

Tyl is an LTR retrotransposon from the copia group, whose original host is the yeast S. cerevisiae. The Tyl retrotransposon containing a neomycin resistance marker, pGTyH3-neo (Boeke 1988), is fused to the P. pastoris AOXl promoter, such that the transcription initiation of the AOXl promoter and of Tyl coincide (FIG. 1A). For this, the P. pastoris AOXl promoter is amplified from plasmid pIB4 (Sears 1998) with oligos (SEQ ID NO:9) 5'-GGAATTCCAATTCCTTCTAT-3' and (SEQ ID NO: 10) 5'-TCTCCTCGAGGATAAAAAAAAAGGTTTAAG-3'. These oligos add an EcoRI restriction site at the 5'-end of the promoter, for cloning into the EcoRI restriction site in pUC18, and an Xhol restriction site, for joining to the Xhol restriction site positioned at the transcription start of pGTyH3-neo. The wild-type Tyl promoter from the 3 '-long terminal repeat (LTR) of pGTyH3-neo, which during reverse transcription is translocated to the 5' of the retrotransposon, is replaced by a Pichia pastoris GAP-promoter. For this, a series of PCR amplifications (using as templates plasmid pIB2 (Sears 1998), TyH3 from pGTyH3 (Boeke 1985), and synthetic primers) and cloning steps are performed resulting in the sequence termed "pGAP LTR" (FIG. 2). The sequences downstream of the Xhol site originate from the 5'-LTR of pGTyH3. The AOXl promoter is repressed during conditions when glucose is available as a carbon source, but activated when only methanol is available. The GAP promoter is constitutively active, but since it is initially located in the 3 '-LTR will not transcribe the Tyl retrotransposon. This construct permits to activate the retrotransposon at a specific time and keep it activated independently of methanol and the originally inserted retrotransposon. This assembly is inserted into the general cloning vector pUC18 containing a β-lactamase gene (β-lam) as an antibiotic restistance marker. A fragment containing a functional copy of the Pichia pastoris HIS4 gene is recovered from pIB4 using Aatll and Dral restriction enzymes, and inserted into the Aatll restriction site of pUC18 (FIG. 1A).

The resulting plasmid is linearized with Stul, a single restriction site in the HIS4 gene, and transformed into the P. pastoris strain GS115 (Invitrogen), which is his4 auxotroph, according to the manufacturer's instructions. The transformants are selected on plates with media lacking histidine, containing glucose as a carbon source. Transformants are replated on the same media in addition containing 120μg/ml of the antibiotic G418 to enrich for clones having integrated the entire construct. Media are made according to Burke (1998). The resultant clones are further grown to stationary phase in about 200 ml media lacking histidine, with glucose as a carbon source, which represses the AOXl promoter. Retrotransposon expression is then started by transferring the yeast to about 800 ml media lacking histidine and without glucose, but containing 0.5% methanol, which induces the.4O.A7 promoter and can be used as a carbon source by the yeast Pichia pastoris. The cells are grown in this methanol-containing media during approx. 48h at 20°C, with refeeding to 0.5% methanol after approx. 24h. Three quarters of the cells are then harvested and total DNA extracted according to standard procedure (Burke 1998). One quarter of the cells is transferred to about 800 ml YPD complex media and grown further during 8h at 20°C. From this culture, again three quarters of the cells are used for DNA extraction, and one quarter diluted in about 800 ml YPD complex media and grown further during 16h at 20°C. This entire culture is now used for DNA extraction. According to the stage of evolution of the retrotransposon, different cultures are used for purification of free circular retrotransposons. At the very beginning, free circular retrotransposons can only be isolated as long as expression is driven from the strong AOXl promoter of the genome based retrotransposon, and sometimes from the second culture where the AOXl promoter has been repressed for about two divisions of the cells. After 1 to 4 cycles of growth, expression, selection and reinsertion of retrotransposons, free circular retrotransposons can be isolated from the longest culture where the cells have grown during a total of 20h in YPD media. This shows that through this procedure the replication rate and stability of circular retrotransposons can be improved. This also exemplifies how the criteria for selection are adjusted for each cycle of growth, expression, selection and reinsertion of retrotransposons.

Total DNA of the appropriate culture is migrated on a 0.8% preparative agarose gel in a Tris/acetate/EDTA electrophoresis buffer (Ausubel 1988), containing 0.5 μg/ml ethidium bromide, together with a standard that consists of a non-related supercoiled plasmid of a size (7.2 kb) similar to the circular retrotransposon DNA (6.8 kb), to serve as size marker and carrier DNA for the minute amounts of supercoiled covalently closed circular retrotransposon DNA. Before loading, the DNA is heated to 70°C during 2-3 minutes, and mixed with Ficoll 400 loading buffer (Ausubel 1988). A slot, which is 7 mm large, 1.5 mm wide, and about 10 mm high is loaded with between 40-50 μg of total yeast DNA in a volume of 90 μl containing 1 μg of the marker plasmid. The gel is migrated at 0.5 N/cm during 14-18h. The band indicated by the marker DΝA is generously excised, including the range of about 500 bp (linear size marker) around the indicator plasmid. DΝA is eluted using the Qiagen Qiaquick Gel Extraction kit, pooling the excised agarose of 5 slots for one Qiaquick column. Eluted DΝA is treated with plasmidsafe exonuclease (Epicentre technologies) according to the manufacturer's instructions. The DΝA is then amplified by PCR, using primers (SEQ ID ΝO:l 1) 5'-ACAGCGATCGCGTATTTCGTCTC-3' and (SEQ ID NO:5) 5'-TACGCGATCGCTGTTAAAAGGACAA-3' complementary to the neoR region. The PCR mix contains Mg -free DyNazyme buffer at the concentration suggested by the manufacturer (FINNZYMES), 2.5 mM MgCl₂, 360 μM each dNTP, 0.2 μM each primer, 0.75 Units of DyNazyme EXT DNA polymerase, and between 0.5 μl and 5 μl template. The PCR program includes a hot start with 10 cycles of 94°C for 20 sec, and 68°C for 6 min, and 18 cycles of 94°C for 20 sec, and 68°C for 6 min 20 sec, with an increment time of 20 sec every cycle, on a Techne Genius thermal cycler (Techne, Duxford, England). PCR products are resolved by agarose gel electrophoresis. Since there is too little product to be visible, agar is excised at the expected size and purified using Qiagen Qiaquick gel extraction columns, adding a small amount of pUC18 DNA to act as carrier. A second and third round of the same PCR reaction is performed to generate sufficient amounts of product. For the purpose of clarity, this product is called the primary Tyl PCR product. As a general rule, for all PCR reactions, DyNazyme EXT (FINNZYMES) is used according to the instructions of the manufacturer for long PCR products. Optimal MgCl₂ and template concentrations, as well as annealing temperatures are determined for every PCR reaction. All PCR reactions are accompanied by appropriate controls to check for contamination. Also, setting up of PCR reactions and processing PCR products is strictly separated, except where products serve as template for a subsequent PCR reaction. Using overlap extension PCR, the primary Tyl PCR product is re-integrated into the linear construct shown in FIG. 3. At each step of the overlap extension PCR procedure, PCR products are purified by agarose gel electrophoresis and gel extraction. As a template for reaction using primers 1 and 2, a plasmid is assembled putting the HIS4 gene directly adjacent to the pAOX promoter from Pichia pastoris (FIG. IB), essentially removing pUC18 sequences from in between the two elements compared to the original construct in FIG. 1 A. The plasmid shown in FIG. 1C is made by removing most Tyl sequences from the plasmid shown in FIG. 1 A, through cleavage with Xhol restriction enzyme, and religation of the remaining fragment. This plasmid is serving as a template for reaction using primers 3 and 4 (FIG. 3). These template plasmids are made because use of the entire plasmid of FIG. 1 A as a template will result in carry-over of minute amounts of this plasmid in the subsequent reactions of the overlap extension PCR and contamination of the final PCR product by fragments originating from that plasmid only. The reaction using primers 1 and 5 uses as templates the product from the reaction with primers 1 and 2, and the primary Tyl PCR product. The reaction using primers 6 and 4 uses as templates the product from the reaction with primers 3 and 4, and the primary Tyl PCR product, and a fragment encompassing the open reading frame of the neomycin resistance marker (neoR). Finally, the reaction using primers 7 and 4 uses as templates the product from reaction with primers 1 and 5 and the product from reaction with primers 6 and 4. This final product is transformed into fresh cells of the P. pastoris strain GS115 (Invitrogen), through the spheroblasting method, according to the manufacturer's instructions. Transformants are selected and grown as described above. After repeating the procedure between 1 and 4 times, free circular Tyl retrotransposons can be purified from the longest culture where the cells have grown during a total of 20h in YPD media after induction of retrotransposon expression in media containing methanol. Individual clones of these circular Tyl molecules have an increased replication rate at least 3 times as high as the original inserted Tyl retrotransposon, as determined by quantitative PCR after retransformation into host cells. In this pool of circular Tyl retrotransposons, which contain a variety of different mutations, the open reading frame of the neomycin resistance marker (neoR) is exactly replaced by the open reading frame of the P. pastoris HIS3 gene using overlap extension PCR. The result is a DNA fragment from a presumed circular Tyl molecule with a HIS3 marker, linearized within the HIS3 open reading frame. A second fragment is made complementary to several hundred base pairs on each side of the linearization site. The linearization site is chosen towards the end of the HIS3 open reading frame, and the complementary fragment starts just after the initiator methionine (ATG) of the HIS3 open reading frame, to avoid the possibility that the complementary fragment can confer HIS3 autotrophy on its own. In other words, the two fragments overlap with each other such that after homologous recombination a circular Tyl plasmid with a HIS3 marker is formed. The two fragments are co-transformed into P. pastoris strain lacking the HIS3 gene (Cosano 1998). This P. pastoris strain is made essentially according to the procedure employed for the deletion of the HIS4 gene by Crane (1994). Transformants are selected on media lacking histidine at 20°C, grown further in liquid media lacking histidine, and DNA extracted. The DNA of between hundreds and thousands of individual transformants is analyzed by southern blotting, using the HIS3 sequence as a probe (Ausubel 1988). Transformants are identified that have a proportionally strong signal at the expected sizes for supercoiled and relaxed covalently closed circular DNA, compared to a proportionally weak signal for the bulk of the genomic DNA. The identity of the circular Tyl is confirmed by purification of the covalently closed circular DNA, amplification by PCR, and restriction enzyme digestion. Certain restriction sites are lost due to mutagenesis. Use of several enzymes will nevertheless indicate whether the gross structure still corresponds to the Tyl retrotransposon. The nature of the circular Tyl molecules is further confirmed by DNA sequencing. Because some of these circular Tyl molecules stably replicate as plasmids for several divisions of the host cell, they are now called retroplasmids. The properties of theTyl retroplasmids are further improved by additional rounds of introduction into fresh host cells, growth, and purification of Tyl retroplasmids, combined with backcrossing of the resulting Tyl retroplasmids with Tyl molecules from earlier stages of the evolution procedure (WO 97/07205). The directed evolution of Tyl retroplasmids illustrates a combination of step-wise and continuous evolution. At the beginning, free circular retrotransposons are selected after one cycle of retrotransposon replication, but the number of replication cycles increases after each additional selection. At the last step, evolution is continuous, where retroplasmids are evolved over many generations. This is possible because indirect selection pressure is applied through the necessity of presence of the HIS3 marker gene.

Characterization of evolved Tyl retroplasmids by southern blotting: DNA of a P. pastoris strain lacking his3 sequences, but harboring the evolved retroplasmid with a his3 marker, and DNA of a wild-type P. pastoris strain with one copy per genome of the his 3 sequence are compared with each other through southern blotting. Equal amounts of total DNA of each strain are resolved on an agarose gel and, after transfer to a membrane, hybridized with a probe specific to his3. The majority of the genomic DNA runs in a discrete band towards the top of the gel. The lane with the wild-type strain gives a clear signal reproducing the discrete band of genomic DNA, but no significant and only diffuse signals in the rest of the lane. In the lane of the strain with the retroplasmid, at the location where on the agarose gel a discrete genomic band is visible, a diffuse signal is present that is significantly weaker compared to the signal in the same location in the lane of the wild-type strain. In addition, in the lane of the strain with the retroplasmid, clear and discrete bands are present at about 3.5 Kb and 12 Kb, which corresponds to the sizes expected for covalently closed circular and open circular forms of the retroplasmid DNA under the conditions used. This result shows that if there are any retroplasmids integrated into the host genome, their number is significantly less than one copy per genome, or per cell. Since the cells rely on the presence of the his3 marker for their growth, the his3 marker must be provided by the retroplasmid also visible on the southern blot. The retroplasmid cannot rely on a genome based copy for its propagation, because one such copy would have to be present in every cell, and the retroplasmid must therefore be replicating independently of integration into the genome. Furthermore, the ability to purify retroplasmids and retransform them into fresh host cells, where they are again able to replicate, shows that the purified retroplasmid DNA contains all the elements necessary for replication and that replication is not bound to a genome based copy of the retroplasmid. Finally, the size and the structure of the retroplasmids, as determined by restriction mapping and southern blotting, are consistent with circular DNA structure containing one LTR, arisen through homologous recombination of the two LTRs of an intermediate linear retroplasmid cDNA.

Sequence diversity of evolved Tyl retroplasmids: DNA of Tyl retroplasmids is isolated by gel electrophoresis, gel extraction and PCR, subcloned into a bacterial plasmid vector and the sequences determined by DNA sequencing. Control PCR reactions on a template where the exact sequence is known, are performed to determine the frequency of mutations introduced by the PCR reaction. This control determines that the PCR amplification under the conditions used accounts for less than 1 mutation per 1000 nucleotides. In contrast, a large diversity of sequence variation is detected among the retroplasmid sequences. The sequences can be grouped into phylogenetic families according to their sequence distances, with distances varying from between 1 to 15 mutations per 1000 nucleotides between the individual retroplasmid sequences, on average between 3-4 mutations per 1000 nucleotides. The retroplasmids analyzed are purified from a culture whose cells have gone through 30 divisions starting from a single fransformed colony of P. pastoris. This mutation rate is in general agreement with the known mutation rates for LTR retrotransposons of 10^"4 to 10^" mutations per nucleotide per replication cycle (Preston 1996). In agreement with this result is the fact that digestion of the pool of PCR-amplified retroplasmids with a variety of restriction enzymes yields not only the expected pattern of restriction fragments, but also a general smear and additional bands, indicating that the population is not homogeneous and restriction sites are mutated. The sequence analysis also revealed sporadic occurrences of homologous recombination events that must have taken place between individual retroplasmids. These events can either have happened between double-stranded DNA molecules through regular homologous recombination carried out by cellular elements, or homologous recombination can have occurred by template switching during the reverse transcription process (Hu 1997).

Example 2: Increased resistance ofHIS3 to competitive inhibition by 3-amino-l,2,4-triazole

Imidazoleglycerolphosphate dehydratase (HIS3) is sensitive to competitive inhibition by 3- amino- 1 ,2,4-triazole (AT) (Horecka 2000). A P. pastoris strain lacking the HIS3 gene, containing a Tyl retrotransposon with a HIS3 marker gene replicating independently of integration is used (see Example 1). The strain must have gone through at least 30 divisions starting from a single transformed cell to provide adequate diversity of functional HIS3 variants. Cells are plated at a density of 3x10⁴ cells per petri dish of 15 cm diameter, on minimal media lacking histidine (Burke 1998), and supplemented with AT to a final concentration of 100 mM. About one hundred 15 cm petri dishes are plated and grown during 10 days at 20°C. Outgrowing colonies are harvested, and replated at the same density as above on the same media. Cells are grown for 7 days at 20°C, colonies harvested and pooled, and DNA extracted. HIS3 open reading frames are amplified by PCR, fused with the original promoter sequence for the HIS3 marker from Tyl, and cloned into pIBl (Sears 1998). Individual clones are linearized with the Stul restriction enzyme and transformed into a P. pastoris strain lacking the HIS3 gene and transformants selected on media lacking histidine. Transformants are then identified that have the capacity to grow on media containing 100 mM AT. The corresponding HIS3 open reading frames are analyzed by DNA sequencing. P. pastoris cells expressing these imidazoleglycerolphosphate dehydratases are capable to grow in the presence of 100 mM AT in media lacking histidine.

This example also illustrates the principle of continuous directed evolution. Cells containing HIS3 genes that confer a little resistance to AT can grow, even if slowly, and replicate the contained retrotransposon. These HIS3 genes are thus further mutated, and new beneficial mutations provide an even bigger growth advantage to the cells containing the corresponding HIS3 genes, that are again further mutated.

Example 3: Directed evolution of a bacterial retroplasmid

The Mauriceville retroplasmid replicates through reverse transcription and is found in mitochondria of the fungus Neurospora crassa (Kuiper 1988). The following procedure describes the adaptation of the Mauriceville retroplasmid to replicate in Escherichia coli. Fungal mitochondria utilize a different genetic code than E. coli. The open reading frame (ORF) contained in the Mauriceville retroplasmid, encoding a reverse transcriptase and probably other, undefined functions, contains eight TGA codons coding for a tryptophane in fungal mitochondria. TGA encodes a stop codon in E. coli. Through site-directed mutagenesis, all the eight positions are changed to TGG, coding for a tryptophane in E. coli. For easy selection of transformants, a beta-lactamase (β-lam) encoding sequence is inserted just after the reverse transcriptase open reading frame. Furthermore, to ensure proper expression of the retroplasmid genomic RNA in Escherichia coli, the original Mauriceville promoter (Kennell 1994) is replaced by the early promoter pL of phage HK022 (Cam 1991), including the putL antiterminator sequence (Weisberg 1999) to ensure full length transcription of the Mauriceville genomic RNA (FIG. 4). This modified Mauriceville retroplasmid is assembled in the pUC18-tet plasmid backbone, which has the beta-lactamase sequence conferring resistance to the antibiotic ampicillin replaced by a sequence conferring resistance to the antibiotic tetracycline. Since the 5'- and the 3 '-ends of the genomic RNA are adjacent on the circular DNA molecule, the 0.9 kb Pstl-Sacl region containing the promoter and the 3 '- end is duplicated, to ensure full length expression of the linear genomic RNA (FIG. 4). This construct is transformed into the E. coli high mutator strain XL-1 red (Stratagene) and the cells grown over seven passages, each passage inoculating 0.5 ml of the grown culture into 500 ml fresh culture. Of the last culture, plasmid DNA is isolated according to standard procedures. This plasmid mixture contains predominantly the entire construct. In addition, there is a small amount of plasmids corresponding to the introduced Mauriceville retroplasmid alone, generated trough homologous recombination between the duplicated promoter regions. This plasmid has a size of about 4.5 kb. The same homologous recombination event also creates low amounts of pUC18-tet with one copy of the 0.9 kb Pstl- Sad region as an insert. This plasmid has a size of about 3.5 kb. A small amount of an unrelated plasmid of 4.5 kb size is added to the mixture to serve as marker and carrier for the following purification steps. The entire plasmid mixture is resolved through agarose gel electrophoresis, the marker plasmid together with the mixture of recombined and replicated Mauriceville retroplasmids excised and eluted from the agarose. The obtained sample is purified a second time through gel electrophoresis and elution to remove all traces of the original entire construct. Then, the sample is digested with PvuII restriction enzyme, which has cleavage sites in the pUCl 8-tet backbone but not in the retroplasmid sequences and treated with plasmidsafe exonuclease (Epicentre Technologies) according to the manufacturer's instructions. Retroplasmids are amplified by PCR, using the synthetic oligonucleotides (SEQ ID NO: 12) 5*-GCGAGCCCTATGGCCAAAATTAG-3' and

(SEQ ID NO: 13) 5'-CCATAGGGCTCGCCAAGCAGTGA-3' that are specific to a region in the reverse transcriptase open reading frame of the Mauriceville retroplasmid. These PCR products are used to recreate constructs of the same overall topology as the initial construct of FIG. 4, except that all the original Mauriceville retroplasmid sequences including the flanking promoter regions are replaced by the analogous sequences contained in the PCR products. To preserve the diversity of the retroplasmid mixture contained in the PCR, this is done using overlap extension PCR in a similar fashion as detailed in Example 1, and FIG. 3. The fragments resulting from the overlap extension PCR are ligated and transformed into the E. coli high mutator strain XL-1 red (Stratagene) again for the next round of mutagenesis and selection. The entire process is repeated between 5 to 20 times, until a retroplasmid is obtained that stably replicates over several divisions of the E. coli host.

Example 4: Directed evolution of a β-lactamase

Bacterial resistance to β-lactam antibiotics such as penicillins and cephalosporins is mediated by β-lactamases that cleave the amide bond in the β-lactam ring to generate inactive products. TEM-1 β-lactamase hydrolyzes both penicillins and cephalosporins. However, it cannot efficiently hydrolyze extended-spectrum cephalosporins, such as cefotaxime and ceftazidime. For determination of cefotaxime resistance, 10⁵ cells are spread per LB-Agar plate (10 cm diameter), different plates containing different concentrations of the antibiotic cefotaxime. Since the antibiotic resistance depends on the cell density plated, a similar number of cells are plated throughout the assays and selections. Determined this way, cefotaxime (Sigma, St. Louis MO) has a minimum inhibitory concentration (MIC) of 0.04 μg/ml for E. coli DH10B expressing the TEM-1 β-lactamase inserted into the Mauriceville-derived retroplasmid. E. coli DH10B expressing the TEM-1 β-lactamase inserted in the retroplasmid are grown in liquid LB at 30°C. A total of about 6-10⁶ cells is plated on LB-Agar containing 0.5 μg/ml cefotaxime. Cells are grown at 30°C during 16-20h. Colonies are harvested and pooled, and again a total number of about 6-10⁶ cells is plated at a density of 10⁵ cells per 10 cm-plate on LB-Agar containing 1 μg/ml cefotaxime. This procedure is iterated, each time doubling the cefotaxime concentration. After 10 iterations, reaching about 250 μg/ml cefotaxime, retroplasmids are isolated, retransformed into fresh E. coli DH10B and plated on 250 μg/ml cefotaxime to differentiate between evolved β-lactamases and cells that have acquired elevated cefotaxime resistance due to other reasons. The thus selected plasmids are again isolated, DNA encoding evolved β-lactamases subcloned and analyzed by DNA sequencing. The β-lactamases resulting from these sequences have the ability to efficiently hydrolyse cefotaxime. E. coli expressing these β-lactamases are able to grow in the presence of up to 250 μg/ml cefotaxime.

Example 5: Evolution of the Ty3 retrotransposon in S. cerevisiae.

Ty3 is an LTR retrotransposon from the Gypsy group. It originates from the yeast Saccharomyces cerevisiae. Only between 2 and 3 copies of Ty3 are found in the genome of S. cerevisiae (Hansen 1988). The wild-type Ty3 has a very low replication and transposition frequency. To select Ty3 retrotransposons with an elevated replication rate, concomitant with the inability to integrate into the host genome, the S. cerevisiae strain TMY43 is used where the naturally occurring Ty3 elements have been removed, leaving only solo Ty3 LTRs (Sadeghi 2001). In addition, remaining HIS3 sequences are removed from this strain by homologous recombination. A Ty3 is assembled whose expression is under the control of a regulatable galactose promoter, and which in the 3 'LTR has a HIS3 marker gene inserted, interrupted by an artificial intron (Sadhegi 2001). Ty3 wild-type promoter sequences between the HIS3 marker gene and the TATA box (140 base pairs after the beginning of the sigma element (Hansen 1988)) are replaced by synthetic upstream activating sequences conferring strong constitutive promoter activity (Uemura 1997). This construct is inserted into a yeast integrating vector and transformed into the above S. cerevisiae strain derived from TMY43. With this setup, an analogous procedure as detailed in Example 1 is followed to obtain retrotransposons that stably replicate independently of integration into the host genome for several divisions of the host cells.

Example 6: Directed evolution ofMoloney murine leukemia virus (Mo-MuLV)

Mo-MuLN is a mammalian type C retrovirus that has been extensively studied and used as a retroviral vector for a variety of applications (Coffin 1997). When lacking env sequences,

Mo-MuLN can behave like a retrotransposon within mammalian cells (Heidmann 1988). The following procedure is followed to evolve Mo-MuLN into a vector that can replicate as a plasmid within a mammalian host cell independently of integration into the host genome. A construct is assembled that resembles the wild-type Mo-MuLN but lacks most of the env sequences and contains a puromycin selection marker. Specifically, the pBabe puro plasmid from Morgenstern (1990) is adapted such that the ATG- gag is replaced by the wild-type gag and pol sequences, followed without intervening polyadenylation signal by the SN40 promoter, the puromycin resistance marker and the 3' -LTR of the pBabe puro plasmid. To reduce the frequency of recombination within the 61 bp overlap of the 3' end of the pol sequence and the 5 'end of the env sequence, nucleotides are changed such that the resulting amino acid sequence remains the same, but the homology between the two sequences is reduced (Morgenstern 1990). These changes will reduce the frequency of wild-type virus generation through homologous recombination. Similar changes are introduced in the last 31 codons of the env gene. The resulting env open reading frame is transferred into a suitable expression plasmid with a strong constitutive promoter and a polyadenylation signal, avoiding sequences homologous to the ones found in the above pBabe puro construct with the gag and pol sequences. The env expression plasmid is used to generate a stable packaging cell line essentially following the procedure of Pear (1993). The pBabe puro construct with the gag and pol sequences is introduced into this packaging cell line to make high titre retroviral supernatants (Pear 1993). These supernatants are used to infect about 1x10 NIH 3T3 cells, select infected cells in media containing puromycin, and expand puromycin resistant cells for a few generations. Cells are washed, harvested and total DNA extracted according to standard methods. Circular retroviral molecules are purified through agarose gel electrophoresis similar to the procedure described in Example 1. A PCR strategy analogous to the one detailed in Example 1 is used to amplify circular retroviruses and to recreate the pBabe puro construct with the gag and pol sequences, to preserve the diversity of the isolated circular retroviruses. This construct is then again introduced into packaging cells to make high titre retroviral supernatants for infection of fresh NIH 3T3 cells for the next round of growth and purification of circular retroviruses. By packaging the pBabe puro construct with the gag and pol sequences into retroviruses and using these retroviruses to infect host cells, the use of an inducible promoter can be avoided, since infection is very efficient and a high percentage of the host cells receive the retrovirus construct. The cycles of making retroviral supernatants, infection, growth, purification, and PCR amplification of circular retroviruses is repeated between 5 to 15 times, until circular retroviruses are obtained that for several divisions of the host cell replicate independently of integration into the host cell genome, as well as independently of horizontal transfer between cells. There is no extracellular viral phase, the adapted retroviruses replicate as retroplasmids within the same host cell, and are passed on from one generation to the next through vertical transfer.

References

WO 01/70946. Methods for diversification of single genes in vivo.

WO 00/26397. An unusual retrotransposon from the yeast Candida albicans. WO 97/07205. Method for preparing polypeptide variants

WO 97/25410. A Method for in vivo production of a mutant library in cells.

US 6,168,916. Host adaptation of retroviral vectors. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D., Seidman, J. G., Smith, J. A., and

Struhl, K. (1987). Current Protocols in Molecular Biology (New York: John Wiley & Sons). Berkhout, B., and Klaver, B. (1993). Nucleic Acids Res 21, 5020-5024.

Boeke, J. D., Garfinkel, D. J., Styles, C. A., and Fink, G. R. (1985). Cell 40, 491-500.

Boeke, J. D., and Sandmeyer, S. B. (1991). The molecular and cellular biology of the yeast

Saccharomyces: Genome dynamics, protein synthesis, and energetics, J. R. Broach et al., eds.

(Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press), pp. 193-261. Boeke, J. D., Xu, H., and Fink, G. R. (1988). Science 239, 280-2.

Burke, D., Dawson, D., and Stearns, T., eds. (1998). Methods in Yeast Genetics: A Cold

Spring Harbor Laboratory Course-Manual (Cold Spring Harbor, NY: Cold Spring Harbor

Laboratory Press).

Cam, K., Oberto, J., and Weisberg, R. A. (1991). J Bacteriol 173, 734-740. Crane, D. I., and Gould, S. J. (1994). Curr Genet 26, 443-450.

Coffin, J. M., Hughes, S. H., and Narmus, H. E. (1997). Retroviruses (Cold Spring Harbor,

ΝY: Cold Spring Harbor Laboratory Press).

Cosano, I., Alvarez, P., Molina, M., and Νombela, C. (1998). Yeast 14, 861-867.

Drouin, G. (2002). J Mol Evol 55, 14-23. Eichinger, D. J., and Boeke, J. D. (1988). Cell 54, 955-66.

Hansen, L. J., Chalker, D. L., and Sandmeyer, S. B. (1988). Mol Cell Biol 8, 5245-5256.

Heidmann, T., Heidmann, O., and Nicolas, J. F. (1988). Proc Natl Acad Sci USA 85, 2219-

2223.

Hirochika, H., Otsuki, H., Yoshikawa, M., Otsuki, Y., Sugimoto, K., and Takeda, S. (1996). Plant Cell 8, 725-34.

Hu, W. S., Bowman, E. H., Delviks, K. A., and Pathak, V. K. (1997). J Virol 71, 6028-36.

Kennell, J. C, Wang, H., and Lambowitz, A. M. (1994). Mol Cell Biol 14, 3094-107.

Kuiper, M. T., and Lambowitz, A. M. (1988). Cell 55, 693-704. Lucas, H., Feuerbach, F., Kunert, K., Grandbastien, M. A., and Caboche, M. (1995). EMBO J

14, 2364-73.

Modrich, P., and Lahue, R. (1996). Annu Rev Biochem 65, 101-133.

Morgenstern, J. P., and Land, H. (1990). Nucleic Acids Res 18, 3587-3596. Nakayashiki, H., Kiyotomi, K., Tosa, Y., and Mayama, S. (1999). Genetics 153, 693-703.

Pear, W. S., Nolan, G. P., Scott, M. L., and Baltimore, D. (1993). ProcNatl Acad Sci USA

90, 8392-8396.

Powell, S. K., Kaloss, M. A., Pinkstaff, A., McKee, R., Burimski, I., Pensiero, M., Otto, E.,

Stemmer, W. P., and Soong, N. W. (2000). Nat Biotechnol 18, 1279-82. Preston, B. D. (1996). Proc Natl Acad Sci USA 93, 7427-7431.

Remy, L, and Michnick, S. W. (1999). Proc Natl Acad Sci USA 96, 5394-9.

Sadeghi, N., Rύtz, M. L., and Menees, T. M. (2001). Arch Virol 146, 1919-1934.

Sears, I. B., O'Connor, J., Rossanese, O. W., and Glick, B. S. (1998). Yeast 14, 783-90.

Sidhu, S. S. (2000). Curr Opin Biotechnol 11, 610-6. Uemura, H, Koshio, M., Inoue, Y., Lopez, M. C, and Baker, H. V. (1997). Genetics 147,

521-532.

Weisberg, R. A., and Gottesman, M. E. (1999). J Bacteriol 181, 359-367.

Wittrup, K. D. (2001). Curr Opin Biotechnol 12, 395-9.

Xu, H., and Boeke, J. D. (1990). Mol Cell Biol 10, 2695-702. Yanisch-Perron, C, Vieira, J., and Messing, J. (1985). Gene 33, 103-19.

Zhang, J., and Temin, H. M. (1993). Science 259, 234-8.

Claims

Claims

1. A method for the preparation of a retro-element with changed replication properties, wherein a retro-element is inserted into host cells, expressed, and free retro-elements are selected.

2. A method according to claim 1, wherein a retro-element is inserted into host cells, the host cells grown and the retro-elements expressed, free retro-elements are selected, the selected retro-elements reinserted into fresh host cells for a new round of growth, expression and selection, and the cycle of reinsertion, growth, expression and selection optionally repeated for one or more times, each time adjusting the criteria used for selection until modified retro- elements that can efficiently replicate independently of integration into the host genome are found.

3. A method according to claim 1 or 2, wherein the selected retro-elements replicate independently of integration into the host genome.

4. A method according to claim 1 or 2, wherein the selected retro-elements are stably replicating retroplasmids.

5. A method according to claim 4, wherein stably replicating retroplasmids are selected from a retrovirus replicating independently of horizontal transfer within the same cell.

6. A method according to claim 1 or 2, wherein the selected retro-elements have an increased replication rate compared to the initially inserted retro-element.

7. A method according claim 1 or 2, wherein the retro-element inserted is heterologous to the host cell, and the selected retro-elements have the ability to replicate within the heterologous host.

8. A method according to claim 1 or 2, wherein the host cells lack endogenous retro-elements with a sequence similarity to the retro-element inserted.

9. A method according to claim 1 or 2, wherein the retro-element inserted into host cells comprises a heterologous promoter replacing the wild-type promoter of the retro-element.

10. A method according to claim 1 or 2, wherein the retro-element inserted into host cells consists of a plurality of mutated retro-elements.

11. A method according to claim 1 or 2, wherein free retro-elements are selected by separation of covalently closed circular plasmid DNA from the rest of the host DNA.

12. A stably replicating retroplasmid prepared by the method of claim 4 or 5.

13. A retroplasmid according to claim 12, characterized in that it undergoes mutagenesis at a rate of mutation of at least 10^" per nucleotide per round of replication.

14. A collection of retroplasmids according to claim 12 consisting of a plurality of mutated retroplasmids with on average at least 3 random mutations per 1000 nucleotides.

15. A retroplasmid according to claim 12, characterized in that, when inserted into a host cell, it is integrated into the genome of the host cell in an amount significantly less than one copy per cell.

16. A retroplasmid according to claim 12 comprising a heterologous DNA sequence.

17. Use of a retroplasmid according to claim 16 to produce a diversity of heterologous DNA sequences.

18. Use of a retroplasmid according to claim 17 for diversifying the DNA sequence through combined mutation and recombination.

19. Use of a retroplasmid according to claim 17 for combined mutation, recombination and expression of proteins.

20. Use of a retroplasmid according to claim 17 for combined mutation, recombination, expression and selection of proteins.

21. Use of a retroplasmid according to claim 17 for continuous evolution of DNA or proteins.

22. Use of a retroplasmid according to claim 17 for in vivo mutagenesis of a DNA sequence.

23. A method for the preparation of a DNA coding for a protein with a new or improved property, characterized in that a DNA sequence is inserted into a retroplasmid, the resulting modified retroplasmid is introduced into host cells, these host cells are multiplied, host cells expressing proteins with a desired property are selected, selected host cells are further multiplied, the cycle of multiplication and selection is repeated until the expressed proteins have reached a threshold of improved property, retroplasmids are isolated and the DNA coding for the protein recovered.

24. A method according to claim 23, wherein the host cells comprising the modified retroplasmids are multiplied under conditions of a continuous constant or increasing selection pressure for the desired property of the protein.

25. DNA coding for a protein with improved property prepared by the method of claim 23 or 24.

26. DNA according to claim 25 coding for an improved HIS3 protein, which, when inserted into host cells, confers increased resistance to 3-amino-l,2,4-triazole.

27. DNA according to claim 25 coding for an improved β-lactamase protein, which, when inserted into host cells, confers increased resistance to cefotaxime.

28. Protein with improved property expressed by DNA prepared by the method of claim 23 or 24.

29. An improved HIS3 protein according to claim 28, which, when inserted into host cells, confers increased resistance to 3-amino-l,2,4-triazole.

30. An improved β-lactamase protein according to claim 28, which, when inserted into host cells, confers increased resistance to cefotaxime.