Target Validation in Drug Discovery

Connecticut

I

Pharmaceutical biotechnology for target validation

Outline

Chapter 1: GENERATION OF TRANSGENIC ANIMALS

Chapter 2: TARGET VALIDATION IN CHEMOGENOMICS

1

GENERATION OF TRANSGENIC ANIMALS

BRIAN P. ZAMBROWICZ, PH.D., KATHLEEN H. HOLT, PH.D., D. WADE WALKE, PH.D., LAURA L KIRKPATRICK, Ph.D. and DEREK E. EBERHART, Ph.D.,     Lexicon Genetics Incorporated, The Woodlands, Texas

The development of molecular biological tools for mutant-mouse production signaled the new era of contemporary mammalian genetics. Recent technology advances coupled with completed sequence of the mouse and human genomes are converging synergistically such that today’s scientists are exploring their fields of interest in arguably the most productive period for research and discovery of modern biological science. Scientists can now take advantage of an unprecedented ability to manipulate the mouse genome using versatile techniques that afford both custom genetic alteration and random mutagenesis approaches. Together, these strategies can be employed to quickly and comprehensively saturate the genome with a high percentage of gene coverage. Productivity estimates of the leading commercial biotechnology groups that have knockout mouse production and analysis capabilities, together with published knockout reports, suggest that roughly one-fourth of the approximate 25,000 genes in the mouse genome have been knocked out and studied to varying degrees. The international mouse-genomics community has recently begun an initiative to produce knockout alleles for all mouse genes and to develop a system for open access to data and reagents for the broader scientific community. Clearly we can look forward to an escalation of discovery productivity in the upcoming years. It will be even more exciting to track how mouse gene-function discoveries will drive the drug discovery industry’s efforts to develop the next generation of therapies for human disease.

I INTRODUCTION

The demonstration that Mendel’s laws of inheritance governing plant traits translated to the inheritance of coat color characteristics in mice launched the classical period of mouse genetics in 1902 (Cuenot, 1902). Over the next several decades, the mouse proved to be a robust experimental model system that afforded unprecedented breakthroughs in understanding mammalian genetics and physiology. In the early 1900s, the first strain of inbred mice—DBA—was derived by C. C. Little, who later became the founder of The Jackson Laboratory (Paigen, 2003). Many of the more than 300 inbred mouse strains in use today were developed during the same time period (Staats, 1980). These mouse strains were the cornerstone of research, which led to the elucidation of fundamental physiological mechanisms, including antibody-mediated tumor transplantation rejection (Gorer, 1937) and retroviral-mediated protooncogene activation and neoplastic transformation (Mulbock and Bentvelzen, 1968). They also fueled the beginnings of the contemporary field of biochemical genetics. Mouse geneticists at the time could not have known the degree of conservation between human and mouse genes. Nevertheless, the idea that human and mouse physiology is highly conserved appears to have been a widely held view considering the common theme of comparative physiological and disease studies in mice and humans reported over the years.

Ultimately, classical mouse genetics laid the groundwork for the launch of the new era of modern mouse genetics, which began around 1980 with the abrupt arrival of molecular biological techniques and the coincident development of germline-competent murine embryonic stem (ES) cell lines. A brilliant scheme, known as gene targeting, was devised to engineer precise genetic manipulations by homologous recombination using positive and negative selection in mammalian tissue culture cells (Mansour et al., 1988). In parallel, pluripotent murine ES cell lines were derived from the inner cell mass of blastocysts and found to be capable of contributing to the formation of an embryo (Bradley et al., 1984). Eventually, genome-wide mutational strategies, such as retroviral gene disruption and chemical mutagenesis, were applied to mutant mouse production. These random mutagenesis strategies have proven to be extremely efficient when used in conjunction with viral-insertion site-mapping techniques (Zambrowicz et al., 1998) and robust point mutation mapping techniques (Hrabe de Angelis et al., 2000). Using the technologies of gene trapping and gene targeting by homologous recombination, Lexicon Genetics has generated and completed the phenotypic analysis of more than 2600 knockout mouse lines to date, which represent approximately 2500 distinct genes. Over the last four years, gene function in mice has been explored by focusing efforts on reducing the amount of functional transcript of a given gene of interest. This method, known as RNA interference, takes advantage of the endogenous cellular machinery dedicated to destroying double stranded RNA resulting from viral infection. For the first time in the history of modern science, scientists now have access to a molecular toolbox filled with unique instruments designed to alter the functional expression of genes in a mammal. Based on the physiological consequences of a given genetic alteration in the mouse, scientists can predict the function of the orthologous gene in man. Coupled with drug development capabilities, these inferences can translate to therapeutic treatments for human disease (Zambrowicz and Sands, 2003; Zambrowicz, Turner, et al., 2003).

II GENERATION OF TRANSGENIC ANIMALS FOR TARGET VALIDATION

A Gene Targeting by Homologous Recombination

Gene targeting by homologous recombination affords deliberate and precise manipulation of genomic sequence in the mouse. Gene targeting in mammalian cells is an exceedingly rare event and remains a bottleneck in the entire process of knockout mouse generation. Approximately 99% of the targeting vectors introduced into ES cells will insert randomly in the genome, as opposed to integrating homologously into the target sequence (Joyner, 1999). Since the early years of gene targeting, efforts have been focused on identifying key variables that might be optimized to achieve higher gene-targeting rates. For example, the use of positive as well as negative selection protocols to reduce the survival of cells containing random integrants (Mansour et al., 1988), using isogenic DNA within targeting sequences (te Riele et al., 1992), increasing the homology arm length (Deng and Capecchi, 1992; Hasty et al., 1991), and limiting the deletion size (Joyner, 1999) have each contributed to overall improved rates of success.

Previously, Lexicon described a λ phage knockout shuttle (λ KOS) system for streamlined construction of targeting vectors (Wattler et al., 1999). This system employs a murine 129/SvEvBrd λ phage genomic library containing negative selection cassettes flanking the genomic inserts (Figure 1.1A). The genomic inserts are 9 kb average length, which allows for deletion strategies upward of 6 kb or greater, while maintaining conveniently sized arms of homology that allow for efficient targeting as well as ease of screening for correct recombination events by Southern blot or long-range PCR. The library has a nine-fold genomic coverage and is arrayed in a 96-well format such that any gene of interest can be recovered efficiently by PCR. To date, we have yet to identify a single gene that was not represented in the library, and more than 2500 genes have been successfully retrieved from this resource. The λ phage clone is readily converted to a high copy plasmid via Cre-mediated excision (Figure 1.1A). In our experience, the genomic plasmids are stable and propagation robust in common E. coli laboratory strains. Rare base-cutter restriction enzyme sites are placed at the 5′ and 3′ boundaries of the deletion region by introduction of a yeast selection cassette via an efficient homologous recombination event in yeast. Finally, a single directional cloning step replaces the yeast selection cassette with an ES cell selection cassette (Figure 1.1B). Our ES cell selection cassettes usually contain a reporter gene, typically LacZ, which can be used to determine the temporal and spatial expression patterns of the targeted gene. The λ KOS system has proven to be robust in terms of both efficiency and scalability, and we have employed this system to generate by far the world’s largest collection of custom mutant mouse lines.

FIGURE 1.1 Construction of gene targeting vectors using the λ KOS system. (A) A schematic of a genomic KOS clone before and after CRE-mediated excision. The positions of the bacterial and yeast origins of replication and selectable markers are shown. The orientation of the HSV-tk genes (negative selection cassettes) and the loxP sites are indicated by arrows. In this illustration, the genomic clone contains exons 1 and 2 of a gene to be targeted. (B) A selection cassette containing both yeast and bacterial selectable markers and flanked by several rare-cutter restriction sites (noted by asterisk*) is PCR amplified using primers containing at their 5′ end 25 to 40 bp of sequence homology flanking the genomic region to be deleted (hatched boxes). In this schematic, a deletion of exons 1 and 2 is planned. Yeast-mediated recombination replaces the genomic region to be deleted with the amplified double-selection cassette. To complete the final targeting vector, the yeast selection cassette is replaced with any desired ES cell-selection cassette using the rare-cutter restriction sites in a single cloning step. The ES selection also contains, in most cases, a reporter gene (LacZ), which can be used to define the expression patterns of the gene targeted ( Wattler et al., 1999).

Of note, in the years pre-dating the completion of sequencing the human and mouse genomes, a partial cDNA tag of a given gene of interest was sufficient for engineering virtually any targeting vector utilizing our system. Several overlapping genomic clones would be isolated from the λ KOS library using exonic probes derived from the cDNA of interest. The genomic plasmid clones would be shotgun sequenced to create a large genomic contig, typically greater than 15 kb in length, within which exons could be defined and the mutation strategy designed. Importantly, external probe sequences, restriction maps, and Southern blot strategies could be derived from the large genomic contigs (Figure 1.2A and 1.2B). Meanwhile, the recent completion of the sequencing of the human and mouse genomes (Lander et al., 2001; Venter et al., 2001) has afforded the advantage that we can now upfront design any mutation strategy with a robust screen design in silico, and then retrieve an appropriate genomic clone from the λ KOS library to begin vector construction. This has resulted in saving much time that was previously spent troubleshooting Southern blot screening strategies.

FIGURE 1.2 Construction of a large genomic contig using overlapping KOS clones. (A) A schematic illustrating multiple KOS clones isolated with an exonic probe derived from the gene of interest. Prior to the sequencing of the mouse genome, these clones would be transposon sequenced to generate a complete working contig in regions where no public sequence existed. External and internal probes could then be amplified and tested on genomic DNA to determine their suitability as probes for the final confirmation of targeted ES cells (B). After electroporation of the target vector, ES cell clones are screened in 96-well format to identify targeted clones. Those clones are then expanded for final confirmation before microinjection into blastocysts.

Recently, we have optimized a PCR-based targeting vector engineering strategy, taking advantage of the sequenced mouse genome as well as newly available, high-processivity DNA polymerases (Figure 1.3A). This PCR-based approach allows us much greater flexibility in mutation design, as we are no longer limited by the size or the placement of the KOS clones. The strategy utilizes a single directional cloning step whereby the two homology arms, the ES cell selection cassette, and the pKO Scrambler NTKV-1901 vector backbone (Stratagene, La Jolla, Calif.) are ligated together. This approach can significantly decrease the amount of hands-on manipulations and streamlines the overall timeline to final targeting vector construction, from an average of 10 weeks for the λ KOS system to as few as 2 weeks for the PCR-based system. In addition, we have adapted the use of long-range PCR as a method to rapidly screen for targeted ES cell clones in a 96-well format (Figure 1.3B). We screen for targeting events by PCR using one primer outside of the arm of homology and a second primer within the selection cassette. Amplified bands are sequence-confirmed, and the corresponding ES cell clones are expanded for final Southern blot confirmation prior to microinjection into blastocysts. The PCR-based approach for identifying correctly targeted ES cell clones compares favorably to Southern blotting as it can be completed quickly and does not require the use of radioactive probes. One disadvantage of the long-range PCR approach for both target vector construction and targeted ES cell clone screening is that time-consuming troubleshooting steps may occasionally be necessary. Based on our experience, the need for PCR troubleshooting often seems to correlate with repetitive sequence stretches or overall high GC content of the target sequences. Nevertheless, having access to the λ KOS and the PCR-based vector construction as well as the targeted clone screening approaches greatly increases our overall throughput.

FIGURE 1.3 Construction of gene targeting vectors and ES cell screening using long-range PCR. (A) Utilizing multiple bioinformatic tools and the public mouse genome sequence, a target gene is scrutinized to determine the best mutation to create. Once a target region is identified, PCR primers are designed to amplify arms of homology 5′ and 3′ of the deleted region (long hatched boxes). The primers immediately adjacent to the deleted region contain sites for rare-cutter restriction enzymes (noted by asterisk*) and the distal primers contain restriction sites for cloning into the pKO Scrambler NTKV-1901 vector (RE). The arms are typically 3- to 5-kb long and are amplified from genomic DNA prepared from the same ES cells used for electroporation to ensure isogenicity. After amplification, the homology arms are ligated in a four-piece reaction to the ES cell-selection cassette and the Scrambler backbone. (B) After electroporation of the target vector, ES cell clones are screened in 96-well format using long-range PCR to identify targeted clones. A primer is placed outside one of the homology arms and used with a primer inside the selection cassette. Positive bands are sequence-confirmed and the clones expanded for final confirmation as shown in Figure 1.2 prior to microinjection.

As of August 2005, almost 2800 gene targeting projects representing over 2400 distinct genes have been initiated at Lexicon, and over 2000 projects have achieved germline transmission. As shown in Figure 1.4, we have focused our gene targeting efforts on druggable gene classes including channels, enzymes, G protein-coupled receptors (GPcRs), kinases, membrane-associated and integral membrane proteins, proteases, secreted proteins, and transporters. We launched a concerted effort to knock out all GPcRs several years ago, and we have virtually completed our phenotypic screen of the GPcR gene class, excluding the odorant and pheromone receptors. Our gene targeting technologies have proven to be equally efficient for mutating all gene classes, with an average of 6.43% targeting efficiency (Figure 1.5). This compares favorably to published results, including a recent report describing an average 3.8% targeting efficiency for 200 genes (Valenzuela et al., 2003). Surprisingly, utilizing lengthy homology arms derived from bacterial artificial chromosomes (BACs) did not afford any advantage in terms of targeting efficiency compared to our system. While the use of large homology arms does allow for very large deletions, it also mandates that a great deal of additional scrutiny is applied to the mutant locus to ensure deletions and rearrangements have not occurred across the entire homology arm length.

FIGURE 1.4 Genes targeted by homologous recombination at Lexicon. As of August 2005, 2761 total gene targeting projects have been initiated representing 2457 distinct genes. The number of total projects initiated includes null mutations, conditionals, humanizations, and point mutations. More than 2000 gene targeting projects have achieved germline transmission at this time. The total number of unique genes is broken down by gene class here. CHA = channels; ENZ = various types of enzymes, including phophodiesterases and phosphatases; GPR = G protein-coupled receptors; KNS = kinases; MAS = membrane and secreted proteins; MEM = membrane proteins; PRT = proteases; SEC = secreted proteins; TRA = transporters; OTHER = other gene classes not considered druggable.

FIGURE 1.5 Targeting efficiency by homologous recombination broken down by gene classes. A dataset comprising 1860 target vectors was complied and used to identify variables that may or may not correlate with targeting efficiency. The dataset is 86% KOS vectors and 14% Scrambler vectors for standard null and conditional projects only. The overall average efficiency in this dataset is 6.43%. The efficiency across gene class is broken down and depicted here. CHA = channels; ENZ = various types of enzymes, including phophodiesterases and phosphatases; GPR = G protein-coupled receptors; KNS = kinases; MAS = membrane and secreted proteins; MEM = membrane proteins; PRT = proteases; SEC = secreted proteins; TRA = transporters; OTHER = other gene classes not considered druggable.

We have built a sizable gene targeting dataset compiled from 1860 targeting vectors, consisting of 5′ and 3′ homology arm sequence, the sequence of the deleted region, and the percent of targeting events based on an average of 400 ES cell clones screened by 96-well Southern blot or PCR per targeting vector. As previously reported by others, whose data was based on a significantly smaller number of targeting vectors, we also observed a positive correlation between increasing overall homology sequence length and gene targeting efficiency (Figure 1.6). Not surprisingly, given that total homology length and deletion size are inversely dependent variables in our λ KOS system, we observed that deletion size negatively correlates with gene targeting efficiency (Figure 1.7). We have scrutinized several variables within our gene targeting dataset with the idea that certain criteria, such as sequence composition or relative position of the intended deletion within a genomic context, might correlate with gene targeting efficiency and therefore elucidate additional core principals to improve the efficiency of gene targeting. Disappointingly, we were unable to find any correlation with targeting efficiency between multiple variables including repeat content, distance from ATG or transcription start site, whether or not the gene was expressed in ES cells, or distance from the nearest gene. Nevertheless, targeting efficiency by homologous recombination is undoubtedly influenced by locus-specific variables we do not yet understand. For example, small changes in homology arms can dramatically increase targeting efficiency (Lexicon unpublished data).

FIGURE 1.6 Total homology positively correlates with targeting efficiency. Total homology refers to the lengths of the 5′ and 3′ arm combined. The average length of homology in the dataset vectors is 7.6 kb, with a high of 12.3 kb and a low of 2.4 kb.

FIGURE 1.7 Deletion size negatively correlates with targeting efficiency. The average deletion size in the dataset is 2.0 kb, with a high of 20.6 kb and a low of 41 bp.

While there have been no major recent advances resulting in significant increases in gene targeting rates since the introduction of isogenic DNA and positive negative selection, many groups have reported new techniques aimed at streamlining the steps of target vector construction and screening for targeting events (Lee et al., 2001; Valenzuela et al., 2003; Wattler et al., 1999). For the most part, these new vector construction approaches have employed either yeast or bacterial recombination systems to avoid tedious cloning steps otherwise dependent on native restriction sites. In addition, new high-fidelity DNA polymerases and real-time PCR have been deployed to speed the screening steps. Despite the many challenges involved with engineering gene knockouts, targeted knockouts for approximately 10% of the known 25,000 murine genes have already been reported in the literature (Austin et al., 2004). To date, Lexicon has achieved germline transmission for more than 2000 lines created by gene targeting, and a sizeable portion of the over 2600 knockouts having completed phenotypic analysis to date were generated by gene targeting. We have demonstrated that it is possible to use gene targeting to create knockout lines on a large scale and in a high-throughput manner.

In addition to serving as the workhorse technology to create standard null alleles for our large-scale screen for valid drug targets (see Section IIE. Genome5000™), the λ KOS system has also proven to be extremely adaptable to designing diverse genetic alterations, such as conditional floxed alleles, knock-ins, and humanizations (Figure 1.8). We have created custom intermediate cloning vectors that allow for directional cloning of the floxed or knock-in cassettes in conjunction with rare-cutter restriction enzyme sites introduced into the genomic clone by a recombination step in yeast (Wattler et al., 1999). A minor variation of this approach also allows for the introduction of point mutations into exons.

FIGURE 1.8 Other types of mutations can be made by gene targeting. Although the majority of gene targeting vectors are designed to create null alleles, our technology allows for the generation of other mutations as well. Conditional alleles: A custom universal cloning cassette has been created that allows for the directional cloning of a genomic region to be flanked by loxP sites. To make this type of mutation, the KOS clone is first subjected to yeast-mediated recombination to introduce the rare-cutter restriction sites (noted by asterisk*) in the appropriate location, ( Wattler et al., 1999), or arms of homology are PCR-amplified carrying the rare-cutter sites. The region to be floxed is cloned into the intermediate vector using other restriction sites (RE). Then that intermediate cassette is cloned together with the arms of homology utilizing the rare-cutter sites. The positions of the loxP sites are indicated with arrows. The neo selection cassette is flanked with FRT sites (diamonds) to allow for FLIP recombinase-mediated excision later if desired. Knock-in or humanization alleles: The open reading frame of a cDNA of interest can be knocked-in to either a ubiquitously expressing locus (e.g., HPRT or ROSA-26) or to another locus. When a human cDNA is knocked-in to and disrupts the corresponding mouse locus, the mutation is termed a humanization. This allele again begins by PCR-amplifying arms of homology with rare-cutter restriction sites in the desired genomic positions. The cDNA of interest is cloned into another custom universal vector using other directional RE sites. This intermediate vector includes two exons of a minigene (black boxes) to provide a splicing event (enhancing expression) and also a floxed neo selection cassette (loxP sites as arrows). The intermediate cassette is then cloned together with the arms of homology utilizing the rare-cutter sites. Point mutations: Individual amino acid residues (white star) can be mutated by incorporation of the desired base changes into the primer used to PCR the 5′ arm of homology. This primer also carries the rare-cutter site (noted by asterisk*) that, when coupled with the rare-cutter site incorporated into the 3′ arm of homology, allows for the cloning of an ES cell-selection cassette into the adjoining intron. The selection cassette can be subsequently excised using